Fabrication of Carbon-

Based Nanomaterials

Prof. Dr. Ibrahim M. El-Sherbiny

Professor of Nano & Materials Sciences
University of Science and Technology, Zewail City
Carbon

• Melting point: ~ 3500oC

• Atomic radius: 0.077 nm
• Basis in all organic componds
• About 10 M. carbon componds
Carbon is an amazing atom!

It can make chains (Catenation) …………….

It can make flat planes ……………

It can make very hard 3D materials …

Bonding

Graphite – sp2
Diamond – sp3
These wonderful characteristics
and exciting flexibility of the
carbon atom lies behind all of
organic and bio-chemical
compounds. It is also the solid
base for the wide range of c-
based nanomaterials that some
of them will be covered in this
section.
Carbon-Based Nanomaterials

• Fullerene
• Tubes
• Cones
• Carbon black
• Horns
• Rods
• Foams
• Nanodiamonds
 1. Fullerene
”The most symmetrical large molecule”
• Discovered in 1985
- Nobel prize Chemistry 1996, Curl, Kroto, and Smalley

• C60, also 70, 76 and 84.

- 32 facets (12 pentagons and 20 hexagons) Epcot center, Paris
- prototype

~1 nm

Architect: R. Buckminster Fuller

 1. Fullerene
• Symmetric shape
→ lubricant
• Large surface area
→ catalyst
• High temperature (~500oC)
• High pressure
• Hollow
→ caging particles
• Ferromagnet?
- polymerized C60
- up to 220oC
 1. Fullerene
• Chemically stable as graphite
- most reactive at pentagons
• Crystal by weak van der Waals force
• Superconductivity
- K3C60: 19.2 K
- RbCs2C60: 33 K

Kittel, Introduction to Solid

State Physics, 7the ed. 1996.
2. Carbon Nanotubes (CNT)
History/Timeline
2. Carbon Nanotubes (CNT)

History/Timeline
 2. Carbon Nanotubes (CNT)

Structure

 Formed from the hexagonal graphene carbon structure

 Can form single-wall (SWNT) or multi-wall (MWNT)
structures
 Ends (tips) can be open or closed (capped)
 High aspect ratio (length = ~1000 x diameter)
• Discovered 1991, Iijima

"Helical microtubules of graphitic carbon", S. Iijima, Nature 354,

56 (1991)

MWNT “discovered” in 1991 by Sumio

Iijima
– Diameter of ~5-10 nm
– Concentric cylinders of carbon with
a length of several mm
MWNT were produced by slightly changing the arc
discharge conditions in the Krätschmer-Huffman
apparatus (e.g. dc instead of ac).

A thick carbon deposit grew on the cathode. Inside the

hard amorphous carbon shell there was a soft material
that contained the nanotubes.

In order to produce single-wall nanotubes it was

necessary to introduce some transition metals into the
graphite burned during the arc discharge.
So, why is IIjima universally regarded as the
discoverer of nanotubes?

- His pictures were better (better microscopes)

- He was at the right place at the right time. The

research community was ready for the carbon
nanotubes after all the excitement around the fullerenes.

- Some (3) theoretical papers had been published on the

structure and electronic properties of single-walled
nanotubes
2. Carbon Nanotubes (CNT)

Roll-up vector:
 
Ch  n a1  m a2
Wrapping (10,0) SWNT
(Animation)
Wrapping (10,10) SWNT
(armchair)
Wrapping (10,5) SWNT (chiral)
Single walled nanotubes

Different cap geometries

Some Pictures of Carbon
Nanotubes
Properties of Carbon Nanotubes
• High aspect ratio
• Electrical
• Thermal
• Defects
• One-Dimensional Transport
• Mechanical
 High aspect ratio

High aspect ratio:

Length: length
typical few μm
 1000
diameter
→ quasi 1D solid

Diameter:
as low as 1 nm
 Electrical Properties
• If the nanotube structure is
armchair then the electrical
properties are metallic
• If the nanotube structure is
chiral then the electrical
properties can be either
semiconducting with a very
small band gap, otherwise the
nanotube is a moderate
semiconductor
• In theory, metallic nanotubes
can carry an electrical current
density of 4×109 A/cm2 which
is more than 1,000 times
greater than metals such as
copper
 Thermal Properties
• All nanotubes are expected to be very good thermal
conductors along the tube, but good insulators laterally to
the tube axis.

• It is predicted that carbon nanotubes will be able to

transmit up to 6000 watts per meter per Kelvin at room
temperature; compare this to copper, a metal well-known
for its good thermal conductivity, which transmits 385
watts per meter per K.

• The temperature stability of carbon nanotubes is

estimated to be up to 2800oC in vacuum and about 750oC
in air.
 Defects
• Defects can occur in the form of atomic
vacancies. High levels of such defects can lower
the tensile strength by up to 85%.

• Because of the very small structure of CNTs, the

tensile strength of the tube is dependent on its
weakest segment in a similar manner to a chain,
where the strength of the weakest link becomes
the maximum strength of the chain.
 Mechanical
 Carbon nanotubes are very flexible

http://www.ipt.arc.nasa.gov/gallery.html
Nanoscience Research Group
University of North Carolina (USA)
http://www.physics.unc.edu/~rsuper/research/
 Mechanical, Cont.
Carbon nanotubes are the strongest ever known material.
• Young Modulus (stiffness):
Carbon nanotubes 1250 GPa
Carbon fibers 425 GPa (max.)
High strength steel 200 GPa

• Tensile strength (breaking

strength)
Carbon nanotubes 11- 63 GPa
Carbon fibers 3.5 - 6 GPa
High strength steel ~ 2 GPa

• Elongation to failure : ~ 20-30 %

• Density:
Carbon nanotube (SW) 1.33 – 1.40 gram / cm3
Aluminium 2.7 gram / cm3

• Carbon nanotubes have the strongest tensile strength of any material known.
• It also has the highest modulus of elasticity.
 One-Dimensional Transport
(Electron confinement)
• Due to their nanoscale
dimensions, electron
transport in carbon
nanotubes will take place
through quantum effects
and will only propagate
along the axis of the tube.
Because of this special
transport property, carbon
nanotubes are frequently
referred to as “one-
dimensional.”
 3. Cones

Li et al. Nature 407 (2000) 409.

• Discovered 1994 (closed form) Ge & Sattler
1997 (open form) Ebbesen et al.

• Closed: same shape as HIV capsid

• Possible scale-up production (open form)
• Storage?
→ Hydrogen

19.2 o 38.9 o 60.0 o 84.6 o 112.9 o

Scale bar: 200 nm Krishnan, Ebbesen et al. Nature 388 (2001) 241.
 4. Carbon black
Large industry
- mill. tons each year

• Tires, black pigments, plastics,

dry-cell batteries, UV-protection
etc.
• Size: 10 – 400 nm
 Production Methods

1. Arc discharge (AD)

2. Laser ablation (LA)

3. Catalyzed chemical vapor deposition

(CCVD)
1. Arc Discharge Method
 2. Laser ablation

Smalley
The laser ablation method provided a breakthrough at
the time.

The material was almost ”pure”. Very little amorphous

carbon or other particles.

High quality, low defect single-walled nanotubes

Most of the early, pioneering electronic studies were

carried out with these nanotubes

Very nice studies of the growth mechanism dynamics

BUT: expensive and still difficult to have real control

over growth.
3. CVD Nanotube Production
 Arc discharge Chemical vapor Laser ablation
method deposition (vaporization)
Connect two graphite rods Place substrate in oven, Blast graphite with intense
to a power supply, place heat to 600 C, and slowly laser pulses; use the laser
them millimeters apart, and add a carbon-bearing gas pulses rather than
throw switch. At 100 amps, such as methane. As gas electricity to generate
carbon vaporizes in a hot decomposes it frees up carbon gas from which the
plasma. carbon atoms, which NTs form; try various
recombine in the form of conditions until hit on one
NTs that produces prodigious
amounts of SWNTs
Can produce SWNT and Easiest to scale to Primarily SWNTs, with a
MWNTs with few structural industrial production; long large diameter range that
defects length can be controlled by
varying the reaction
temperature
Tubes tend to be short with NTs are usually MWNTs By far the most costly,
random sizes and and often riddled with because requires
directions defects expensive lasers
Writing & Printing

Carbon – graphite

C60: 1000x better resolution

than ink (Xerox)
 CNT / polymer composites
• Current technology
- carbon black
- 10 – 15 wt% loading
- loss of mechanical
properties

• CNT composites
- 0.1 – 1 wt% loading
- low perculation treshold
CNT / polymer composite

• Transparent electrical conductor

- Thickness: 50 – 150 nm
- High flexibility

Wu et al. Science
305 (2004) 1273.
Electric devices
 Application

Flat screen displays

• Field emission

Plasma TV

Saito et al., Jpn. J. Appl. Phys. 37 (1998) L346.

 Transistor
• Vacuum tubes • Semiconductor, Si-based
- Nobel prize 1906, Thomson. - Nobel prize 1956, Shockley, Bardeen,
and Brattain.
- 2000, Kilby.

IBM, 1952.
 Transistor

• SWCNT
- 2.6 GHz, T = 4 K
- Logical gates

Emitter Collector

Base

Bachtold, Dekker et al. Li et al. Nano Lett. 4 (2004) 753.

Science 294 (2001) 1317.
Antenna
 Antenna
• Dipole Radio wave:
~ 3/4 m
m
3 10 8
c s ~ 100 MHz
f  
 3m

• Nanotube
Optical wave:
L
 ~ L / 2 ~ 500 nm

Dekker, Phys. Today May

(1999) 22
Eldrid Svåsand, IFE, Kjeller
Atomic Force Microscopy
 Atomic force microscopy

• Tube or cone
• Chemical probe

Wong, Lieber et al.

Nature 394 (1998) 52.
 Space Elevator

• The Idea
– To create a tether from earth to some object
in a geosynchronous orbit. Objects can then
crawl up the tether into space.
– Saves time and money
• The Problem
– 62,000-miles (100,000-kilometers)
– 20+ tons
 Space Elevator

Pictures from
http://www.space.com/businesstechnology/technology/space_elevator_020327-1.html
 Space Elevator

• The Solution: Carbon Nanotubes

– 10x the tensile strength (30GPa)
• 1 atm = 101.325kPA
• 10-30% fracture strain
• Further Obstacles
– Production of Nanofibers
• Record length 4cm
– Investment Capital: $10 billion
 Some applications of Carbon
Nanotubes include the following
• Micro-electronics /
semiconductors • Nanotube actuator
Conducting Composites Molecular Quantum wires
Controlled Drug Delivery/release Hydrogen Storage
Artificial muscles Noble radioactive gas storage
Supercapacitors Solar storage
Batteries Waste recycling
Field emission flat panel displays Electromagnetic shielding
Field Effect transistors and Dialysis Filters
Single electron transistors Thermal protection
Nano lithography Nanotube reinforced
Nano electronics composites
Doping Reinforcement of armour and
Nano balance other materials
Nano tweezers Reinforcement of polymer
Data storage Avionics
Magnetic nanotube Collision-protection materials
Nanogear Fly wheels"
 Hydrogen storage

2 H2(g) + O2(g) → 2 H2O (l) + energy

Mg2NiH LaNi5H6 H2 (liquid) H2 (200 bar)

3.16 wt% 1.37 wt%
Schlapbach & Züttel, Nature 414 (2001) 353
 Hydrogen storage
• Aim: 5 - 7 wt% H2

• SWCNT
- Dillon et al. (1997) : 8 wt%
(questionable)
- Tarasov et al. (2003): 2.4
wt% reversible, 25 bar H2, -
150oC.

• Cones
- Mealand & Skjeltorp,
(2001) US Patent 6,290,753

Eldrid Svåsand, IFE Kjeller

 Commercial
• Companies: ~ 20 worldwide
- Carbon Nanotechnologies Inc. (CNI)
- SES Research
- n-Tec

• Prices:
- Tubes: pure SWCNT $500 / gram (CNI)
MWCNT € 20-50 / gram (n-Tec)
- C60 : pure $100-200 / gram (SES Research)
- Cones: Multi € 1 / gram (n-Tec)

- Gold : $10 / gram

Warnings & Health Hazards

• Environment and health

• No scale-up production of fullerenes and tubes
• No scale-up design, yet.
 Warnings & Health Hazards
• According to scientists at the National
Institute of Standards and Technology, carbon
nanotubes shorter than about 200 nanometers
readily enter into human lung cells similar to
the way asbestos does, and may pose an
increased risk to health.

• Carbon nanotubes along with the majority of

nanotechnology, are an unexplored matter,
and many of the possible health hazards are
still unknown.
 References
• http://www.news-medical.net/news/22799.aspx
• Chae, H.G.; Kumar, S. (2006). "Rigid Rod Polymeric Fibers". Journal
of Applied Polymer Science 100:791-802: 791.
doi:10.1002/app.22680.
• Hong, Seunghun; Sung Myung (2007). "Nanotube Electronics: A
flexible approach to mobility". Nature Nanotechnology 2: 207–208.
doi:10.1038/nnano.2007.89
• Meo, S.B.; Andrews R. (2001). "Carbon Nanotubes: Synthesis,
Properties, and Applications". Crit. Rev. Solid State Mater. Sci.
26(3):145-249: 145. doi:10.1080/20014091104189.
• Kolosnjaj J, Szwarc H, Moussa F (2007). "Toxicity studies of carbon
nanotubes". Adv Exp Med Biol. 620: 181–204. PMID 18217344
• Ebbesen, T. W.; Ajayan, P. M. (1992). "Large-scale synthesis of
carbon nanotubes". Nature 358: 220–222. doi:10.1038/358220a0
Materials Science-Poland, 33(3), 2015, pp. 566-578
http://www.materialsscience.pwr.wroc.pl/
DOI: 10.1515/msp-2015-0079

Graphene synthesis: a Review

S. S AQIB S HAMS∗ , RUOYU Z HANG , J IN Z HU

Ningbo Key Laboratory of Polymer Materials, Ningbo Institute of Industrial Technology, Chinese Academy of Sciences,
Ningbo, Zhejiang 315201, P. R. China

Graphene has achieved a great amount of popularity and interest from the science world because of its extraordinary
physical, mechanical and thermal properties. Graphene is an allotrope of carbon, having one-atom-thick planar sheets of sp2
bonded carbon atoms densely packed in a honeycomb crystal lattice. Many methods to synthesize graphene have been developed
over a short period and we believe it is necessary to create a list of the most notable approaches. This article focuses on the
methods to synthesize graphene in an attempt to summarize and document advancements in the synthesis of graphene research
and future prospects.
Keywords: graphene; synthesis; nanomaterials; graphite; 2D; biomass
© Wroclaw University of Technology.

1. Introduction explained by Geim et al. [8], “graphene is a single

atomic plane of graphite, which – and this is essen-
Graphene is perhaps the simplest form of car- tial – is sufficiently isolated from its environment
bon and definitely the thinnest material ever pro- to be considered free-standing”. Graphene strikes
duced. It has one-atom-thick planar sheets of sp2 as a remarkable candidate for making functional
bonded carbon atoms densely packed in a hon- nanocomposites having excellent properties. The
eycomb crystal lattice [1]. It is useful to under- fact that its precursor (graphite) is readily avail-
stand graphene as a few layers of graphite. In able makes it all the more tempting [9]. Defect-
this context, the extraordinary properties of hon- free graphene presents outstanding physical prop-
eycomb carbon are not really surprising. In mid- erties, such as high intrinsic mobility and ballistic
dle ages, the layered morphology and weak disper- transport, high thermal conductivity and Young’s
sion forces between adjacent sheets were utilized modulus, an optical transmittance of almost 98 %
to make marking instruments, similar to our use and a large specific surface area [1, 10, 11]. Much
of pencils today. Recently, graphite is being used attention has been given to the characterization of
as dry lubricant, instead of lavishly priced hexago- graphene and its derivatives so we will not be dis-
nal boron nitride and molybdenum disulphide [2]. cussing the properties of graphene. This review ar-
Graphite is also used in electrodes because of high ticle focuses on the brief history and recent ad-
in-plane electrical conductivity (∼104 /Ω·cm) and vancement in synthesis of graphene.
in industrial blast furnaces because of its excel-
lent thermal conductivity (∼3000 W/m·K) [3, 4].
The term ‘graphene’ was first used by Boehm et
2. Synthesis of graphene
al. [5], while describing single layer carbon foils. Graphene has been in the limelight for some
It was derived from graphite and the suffix ‘-ene’. time now and many researchers have been work-
Graphite flakes are basically a stack of graphene ing on graphene synthesis, thus, several methods
layers with an interplanar spacing of 3.5 Å [6] and have been reported for the exfoliation of graphite
having C–C bond length of about 0.142 Å [7]. Like into graphene. They can be divided into two
main categories: the top-down approach and the
∗ E-mail: s.saqib.shams@gmail.com bottom-up approach. Graphite is simply a stack

Unauthenticated
Download Date | 12/26/17 9:59 AM
 Graphene synthesis: a Review 567

of graphene layers and to be able to separate

these layers into individual sheets of graphene,
van der Waals forces need to be overcome [13].
This approach is referred to as the top-down ap-
proach. Several challenges are associated with
this approach, like surface defects that occur dur-
ing sheet separation and the separated sheets re-
agglomerating afterwards. In general, the top-down
approaches offer low yields and are tedious pro-
cedures [13]. While the top-down approach fo-
cuses on breaking graphene precursor (graphite)
into atomic layers from a stack (Fig. 1), bottom-
up approach implements carbon molecules as
building blocks; typically these carbon molecules
are obtained from alternative sources (Fig. 2).
Even though the bottom-up approach is not suit-
able for the production of graphene sheets with
large surface area, this approach offers the pos-
sibility of manufacturing graphene nano-ribbons
and graphene dots (so-called nanoflakes) in large
quantities [14].
Initially after the discovery of graphene, the
most common method exercised was the mi-
cromechanical cleavage of graphite to synthesize
high quality, defect-free graphene: this method
is also called “scotch tape”, drawing or peel off
method [15]. Shortly after sometime, chemical
vapour deposition (CVD) was used as a viable
alternative [16, 17]. However, it was soon re-
alised that these procedures offer a low production
yield and particularly in the case of micromechan-
ical cleavage they were time consuming that hin-
dered effective and full-exploitation of these mate-
Fig. 1. (a) AFM image of graphene. The folded region rials [18]. Even though researches have fine-tuned
exhibiting a relative height of ∼4 Å clearly indi- the technique to provide high quality crystallites
cates that it is a single layer (Copyright National up to 100 µm in size, making it better separated
Academy of Sciences, USA). (b) A graphene (Fig. 1) in intercalated graphite compounds (so-
sheet freely suspended on a micrometer-size
called GIC) [1].
metallic scaffold. (c) Scanning electron micro-
graph of a relatively large graphene crystal, Another route alternative to the micromechani-
which shows that most of the crystal’s faces cal cleavage is exfoliation of graphite or its deriva-
are zigzag and armchair edges as indicated by tives, mainly graphite oxide (GO), which makes
blue and red lines and illustrated in the in-
high production yield achievable, making the pro-
set. 1D transport along zigzag edges and edge-
related magnetism are expected to attract signifi- cess cost-effective and scalable [19]. Exfoliation
cant nano-tubes attention [1]. (d) Raman spectra procedures have been scaled up and now com-
of graphite, GO and graphene sheets [12]. (re- panies are able to sell graphene in large quan-
printed with permission). tities [20]. These are examples of the top-down

Unauthenticated
Download Date | 12/26/17 9:59 AM
568 S. Saqib Shams, Ruoyu Zhang, Jin Zhu

approaches of graphene synthesis, where precur- witnessed by TEM and AFM analysis (Fig. 1).
sors of graphene are stripped down to few lay- However, the uneven thickness of films obtained by
ers of graphite by chemical, mechanical or thermal this method and high production cost due to low
treatments. yield render this method unusable for mass pro-
duction. These methods are a prime example of
the top-down approach, where a graphene precur-
sor (graphite, GO, etc.) is taken apart layer by layer,
forming graphene sheets. The group from Man-
chester University including the two Nobel Prize
winners, Geim and Novoselov, obtained graphene
by micro-mechanical alleviation of graphite. They
used ‘the scotch tape’ or peel off method to repeat-
edly split graphite crystals into increasingly thin-
ner pieces. The tape containing optically transpar-
ent flakes was then dissolved in acetone; a few steps
later, the flakes containing both multi-layer and
monolayer graphene were sedimented on a silicon
wafer to be studied under a microscope. This tech-
nique was later modified by researchers to avoid the
stage, where graphene floated in a liquid. Such de-
velopments eventually led to graphene flakes with
sizes larger than 1 mm which were visible to the
naked eye [21]. Ability to isolate graphene caused
the current research boom. Prior to it, free-standing
atomic planes were thought not to exist [15]. It was
widely believed by scientists that such structures
would be unstable, if scaled down to nanometers
and were expected to scroll and buckle [22, 23].
Some scientists also believe that the intrinsic mi-
croscopic roughening on the scale of 1 nm might
play a role in the stability of 2D crystals [24].

Novoselov et al. [26] applied similar exper-

imental approaches to other materials obtaining
free-standing atomic planes of boron nitride, mica,
Fig. 2. Schematic showing bottom-up and top-down dichalcogenides and complex oxides, but none of
graphene synthesis approaches. them have managed to achieve as much attention
so far. Jayasena et al. [27] reported another ap-
proach towards mechanical exfoliation which used
2.1. Top-down approach a diamond wedge to literally scrape off graphene
2.1.1. Mechanical exfoliation from a graphite source (highly ordered pyrolitic
Mechanical or micromechanical exfoliation graphite, so-called HOPG), aided by ultrasonic
method has been a turning point in the history of oscillations. This, in principle, shares the same
graphene. It is still the key synthesis technique ideology as the approach observed by Geim et
for obtaining high quality graphene for research al., where layers of graphene are mechanically
purposes since it delivers high quality graphene removed from the carbon source. However, due
films sized from 5 to 10 µm as it has been to the ability to control frequency of oscillations

Unauthenticated
Download Date | 12/26/17 9:59 AM
 Graphene synthesis: a Review 569

by this method, in terms of properties, is identical

to what is achievable by sonication of graphite in
solvents and surfactants. Because of the simplicity
of this method, it seems viable for industrial scale
manufacturing of graphene. Interestingly, this
method can also be applied to exfoliate other lay-
ered materials, like boron nitride and molybdinum
disulfide [28]. An additional step to help improve
the quality of graphene obtained by GO, which,
as a precursor, can be obtained by many methods,
has been proposed by Zhang et al. who has chosen
the method reported by McAllister et al. [29] and
treated it under a hot press at 1500 °C for 5 min, un-
der uniaxial pressures of 40 MPa. GO sheets start to
get reduced to graphene sheets between 800 °C to
1500 °C. This method can help obtain gram-scale
highly crystalline graphene. High temperature and
pressure treatment of GO sheets offer a step further
Fig. 3. Schematic of some popular graphene synthesis from thermal treatment of GO to reduce them into
techniques along with their respective features, few layer graphene sheets. The addition of high
and their potential applications [25]. (re-printed pressure into the system further improves the pro-
with permission). portion of GO sheets being reduced to graphene
sheets. This technique also stands as a viable con-
tender for being scaled up [30].
and contact pressure, this allows more control
resulting in consistent properties of thin layer 2.1.2. Graphite intercalation
graphene. As highlighted by Jayasena et al., this Graphite can also be reduced to graphene via
approach requires better understanding of the rela- intercalation. Different chemical species can be in-
tion between the edge formation and ultrasonic vi- serted between graphite interlayer space to produce
brations to be able to explain nanohorns-like struc- graphite intercalation compounds (GIC). Due to
ture formation around the edges. It should also be the presence of these intercalants between graphite
interesting to study localized temperature increase layers, the distance between them increases. This
caused by friction. This method has the potential also changes the properties of graphene because
to provide high quality graphene for experimen- the increased interlayer distance affects electronic
tal purposes, but does not seem suitable for be- coupling between graphene layers. Different inter-
ing scaled up. Keith et al. have recently reported calants could lead to GIC having different set of
graphene synthesis by shear exfoliation in sta- properties usable for applications focusing on elec-
bilizing liquids (N-methyl-2-pyrrolidone, sodium trical, thermal and magnetic performance [31]. Lee
cholate, etc.) in which graphite is immersed and et al. [32] have studied the effects of intercalation
mixed at local shear rates higher than 104 s−1 . This on capacitance in an electrode and have reported
approach attempts to replace the usage of sonica- improved electro-conductivity of MnO2 due to the
tion as means to separate layers, resulting in a less presence of GIC.
energy consuming and scalable approach. It also
allows production of defect-free and un-oxidized 2.1.3. Nanotube slicing
graphene. As noted by Keith et al., shearing rate Carbon nanotubes have also been used as a pre-
must be higher than ∼104 s−1 , which is achievable cursor for graphene. Graphene nanoribbons which
in a simple kitchen blender, too. Graphene obtained are a few microns in size can be obtained by this

Unauthenticated
Download Date | 12/26/17 9:59 AM
570 S. Saqib Shams, Ruoyu Zhang, Jin Zhu

approach [25]. Such methods offer high yield and supercapacitors, sensors, etc. Reduction of GO has
good quality graphene. Theoretically, carbon nano- generally been preferred for graphene reduction be-
tubes could be cut open to form 2D graphene cause of the lower degree of exfoliation achieved
sheets [33]. Several methods have been developed, by graphite and expandable graphite [39–41]. GO
like plasma etching of nanotubes partly embedded can be exfoliated by sonication in water but the
in polymer film [34] or placing carbon nanotubes resulting material is electrically insulating, so it
in a solution containing potassium permanganate should be reduced to restore its conductivity. Re-
and sulfuric acid [35]. High quality graphene can duction can be performed by chemical or ther-
be obtained from nanotube slicing. Chen et al. [36] mal treatment [18]. Rapid heating has also been
developed a nanotube slicing technique which is reported as a possible solution resulting in high
suitable for industrial use. Carbon nanotubes were yields of dispersed carbon powder with a few per-
abraded between ground-glass surface under a con- cent of carbon flakes [42]. Also, by dispersing ox-
stant load of 0.06 N/cm2 causing friction to slice idized and chemically processed graphite in wa-
them into graphene sheets. Graphene obtained by ter, and using paper-making techniques, the mono-
this approach can be used in applications, such layer flakes in a form of a single sheet bonded
as FET, interconnects, NEM and composites [25]. very powerfully can be obtained. These sheets,
However, given the novelty of the precursor (CNT), called graphene oxide paper, have a measured ten-
this method might not be financially viable for sile modulus of 32 GPa [43]. Graphene based mem-
commercial production. branes are impermeable to all gases and liquids
except water, which evaporates through the mem-
2.1.4. Pyrolysis method brane, like it did not exist [44]. GO reduction un-
Pyrolysis offers ease of process and can be der ambient atmosphere with the aid of HCl has
scaled up for production. However, the yield is also been reported [45]. Gurunathan et al. have de-
low and there are impurities in the final product. veloped a novel method of GO reduction with the
In this method, a 1:1 molar ratio of sodium (2 g) help of biomass obtained from lysogeny broth bac-
and ethanol (5 mL) are heated in a sealed reac- teria. In their reported method, bacteria grown from
tor vessel at 220 °C for 72 hours to obtain the lysogeny broth was incubated for 21 hours in an Er-
graphene precursor: solid solvothermal product. It lenmeyer flask, at 120 rpm and 37 °C. The resultant
is then rapidly pyrolysed and the remaining prod- culture was centrifuged at 10000 rpm to obtain the
uct washed with deionized water (100 mL). The biomass used for rGO synthesis. 200 mg of Bacil-
suspended solid is then vacuum filtered and dried lus marisflavi biomass was added to the GO dis-
in a vacuum oven at 100 °C for 24 hours. This persion (0.5 mg/mL) and stirred for 72 h at 37 °C,
solvothermal reaction method can deliver graphene followed by centrifugation to remove excess bacte-
yield of approx. 0.1 g per 1 mL of ethanol-typically ria [46]. While chemical methods make it possible
yielding 0.5 g per reaction [37]. to synthesize graphene at lower temperatures and
Alternative methods: Pyrolysis has also been cost, graphene obtained by such methods contains
used to obtain carbon forms bearing high energy high density of surface defects. Graphene obtained
density for super capacitor application; interest- from this approach is also of a low purity.
ingly these were derived from biomass and their
properties bear resemblance with graphene [38]. 2.1.6. Electrochemical exfoliation
Parvez et al. reported an interesting method
2.1.5. Reduction of graphite oxide (GO) of exfoliating graphene from graphite using
Graphene nanoflakes/powder, from a nm to a an electrochemical method. They used 0.1 M
few µm in size, is obtainable by chemical reduc- H2 SO4 solution as electrolyte, while graphite
tion of graphite oxide (GO). Graphene obtained flakes as anode and platinum wires as cathode.
by this method is suitable for use in conductive A positive charge of +10 V was used to the
inks and paints, polymer fillers, battery electrodes, system and graphite flakes began to dissolve in the

Unauthenticated
Download Date | 12/26/17 9:59 AM
 Graphene synthesis: a Review 571

solution. After 2 minutes, the voltage bias was can be used in such process [50, 51]. A single-
removed and the exfoliated graphitic material cycle yield of up to 5 % is obtainable from this
was collected by vacuum filtration. The sample process and about 70 % of the graphene obtained
was then washed with water repeatedly to ensure by this process is 1 nm thick having electrical
removal of residual acidic content. The obtained conductivity of up to 5000 S·m−1 [52]. When
powder was dispersed in DMF, resulting in EG using the solvent aided systems, graphene tends
sheets [47]. Lu et al. have proposed a one-pot to re-stack together after sonication because of
method to synthesize different nanocarbon forms, van der Waals forces. To avoid such problems,
including graphene, by exfoliation of graphite surfactants or dispersing agents can be added in
in ionic liquids. They experimented with in- the solution before sonication, so the graphene
corporation of 1-butyl-3-methylimidazolium sheets would not re-stack. This method can be used
tetraflouroborate [BMIm][BF4 ] in water used as to obtain graphene sheets from graphite without
an electrolyte for the electrochemical exfoliation chemical modification. It also offers graphene
of graphite. Static potentials of 1.5 to 15 V were yields as high as 5.33 mg/mL. Because of the use
applied. The exfoliated particles were washed with of an ionic liquid (1-hexyl-3-methyl-imidazolium
water and ethanol to achieve a neutral pH [48]. hexafluorophosphate or HMIH), this process is
stable and green [51]. Sodium dodecylbenzene
2.1.7. Sonication sulfonate (SDBS) also tends to stabilize graphene
High quality, un-oxidized graphite and layers and keeps them from aggregation [53]. Liu
graphene flakes can be obtained by sonica- et al. have reported a method using GO suspension
tion [49]. This method is a prime example of the (via sonication) to obtain 3D hydrogel as a support
top-down approach towards graphene synthesis for microbe immobilization and BOD biosensor.
since, in this method, ultrasonic energy is used Neutral Red (NR) solution was added into 20 mL
to separate graphene layers stacked together in a of 2 mg/mL GO suspension, which was sonicated
precursor. But this method requires a large amount further for 10 min and then the solution was kept at
of energy since sonication is the only energy 180 °C for 5 h in a teflon-lined autoclave, followed
source and it can be a concern, while scaling up by temperature reduction to room temperature.
this method [27]. Also, removal of impurities is The obtained hydrogel was washed thoroughly
a problem that occurs in this process. Graphene with deionized water before characterization [54].
obtained by this technique can be used in the
field of polymer fillers, transparent electrodes and π-π interaction has a great potential to be used
sensors [25]. Graphene can be dispersed in certain as an assistance in this regard. Aromatic ring con-
solvents, like N-methyl-pyrrolidone, at concen- taining organic molecules could be used to prevent
trations of up to 0.01 mg/mL. Such solutions can graphene from restacking [55–59]. It is also pos-
be used in a variety of processes to incorporate sible to use the high energy interface between two
graphene in polymers and composites, such as immiscible liquids to hold the graphene sheets af-
spray coating, vacuum filtration, solvent casting ter sonication, thus, prevent them from re-stacking.
or drop casting [49]. While sonication presents Heptane and water are a viable couple to be used
an easy approach towards making graphene, it as immiscible liquids. Graphene sheets obtained in
poses challenges for industrialization in terms this process are up to 95 % transparent and con-
of power consumption. Solvent aided sonication ductive. This process allows obtaining macroscopic
is a simple modification of sonication method graphene coatings on glass surfaces with the help
and an obvious step ahead. Graphene is obtained of chlorosilanes, containing no more than 3 lay-
from graphite through sonication with the help ered graphene stacks [60]. This method has a great
of a solvent (NMP, TEA, etc.), where graphene potential for use in optoelectronics and solar pan-
can be separated from graphite with the help of els. GO can also be reduced to graphene sheets via
centrifugation [49]. NMP or other ionic liquid sonochemical treatment. The pH of GO solution is

Unauthenticated
Download Date | 12/26/17 9:59 AM
572 S. Saqib Shams, Ruoyu Zhang, Jin Zhu

modified to be 10 by adding NaOH and a small conditions (140 kGy doses). The reduced GO
amount of hydrazine. The solution should then be was then separated by centrifugation, washed with
subjected to sonication for 2 h. Because the tem- alcohol a couple of times and dried at 60 °C
perature is unregulated in this method, the solution under vacuum [70]. Laser scribing is a process
might reach temperatures of up to 65 °C by the end similar in mechanism to pyrolysis; however, the
of sonication [12, 61–64]. Hydrazine reacting with energy source used here is a laser. GO is
GO in the presence of ultrasonic irradiation causes coated on a DVD which is then treated inside a
acoustic cavitation resulting in more reactive sites LightScribe DVD drive to directly reduce GO films
for epoxide and hydroxyl removal. Also, the tem- to graphene [71]. This technology has the poten-
perature induced by the irradiation itself causes re- tial to lead us to self-powering smart garments,
duction of GO into graphene since GO is not ther- outfitted with piezoelectric patches to harvest en-
mally stable above 200 °C [12]. ergy from body movements. Power source to cam-
ouflage uniforms is one possible application [72].
2.1.8. Ball milling Other methods have been reported with the same
In principle, ball milling is a sub-category of phenomenon but with different light sources, such
mechanical exfoliation but many variations of this as UV or xenon flash [73, 74], or different kinds
method have been developed, so this method de- of lasers, such as femtosecond laser, pulsed laser,
serves to be discussed separately. Ball milling has etc. [75–79]. In addition, Zhou et al. have re-
been in use for mixing and particle size reduction ported a method to selectively reduce GO by laser
for a long time. Scientists have used this parti- writing [80].
cle size reduction approach to exfoliate graphene
sheets from graphite, sometimes using magnetic 2.2. Bottom-up approach
assistance [65] and sometimes using chemical [66,
67] or solvent assistance [68]. Magnetic assistance 2.2.1. Growth from metal-carbon melts
provides filtration of metallic particles present in In this simple method, a carbon source (e.g.
the precursors, while chemical assistance prevents graphite powder) is placed in contact with a transi-
the sheets from re-stacking using chemical inter- tion metal and heated at high temperatures, enough
actions, such as π-π [66]. Such techniques pro- to melt the metal. Once carbon starts to dissolve
vide FLGs (65 layers) and offer high production in the molten metal, temperature can be lowered
yield. Compared to graphite, a larger surface area to decrease solubility of carbon resulting in exces-
of graphene (153.9 m2 /g) can be obtained using this sive carbon precipitation. The precipitate can be
process [69] and impressively low ID /IG ratio of 0.6 skimmed off to give different forms of carbons,
can be obtained [66]. e.g. thick graphite, few layers of graphene (FLG)
and single layers of graphene (SLG). Nickel seems
2.1.9. Radiation based methods most suited for this process since it is not Raman
Radiation based methods offer graphene pro- active [81]. Ruthenium and iridium are known to
duction in a relatively short period of time. High be usable too but the quality of graphene is not as
quality graphene is obtainable by this method but good as what is achievable from nickel and cop-
the yield is low. Also, we believe radiation con- per [82, 83]. There have been several different vari-
cerns will not allow such methods to be used in- ations of this technique reported [81, 84]. However,
dustrially regardless of the fact that this process the most appreciated method seems to be graphene
is financially not viable for industrialization. GO growth on nickel films using CVD. Nickel film is
can also be reduced to graphene sheets using elec- heated up to 900 to 1000 °C under inert condi-
tron beam irradiation. According to the reported tions (Ar gas used). Methane gas is introduced into
method, an aqueous dispersion of GO and iso- the system along with Ar. This enables the carbon
propyl alcohol was put in a sealed plastic bag from the methane gas to be absorbed on the nickel
and irradiated for 10 min under 2 MeV/10 mA film. Methane supply is later discontinued and the

Unauthenticated
Download Date | 12/26/17 9:59 AM
 Graphene synthesis: a Review 573

system is allowed to cool down (under Ar flow). the low yield is obtained. So this method is not suit-
This enables the carbon to diffuse out of nickel able for industrial manufacturing.
surface and form graphene sheets [85–88]. Simi-
2.2.3. Dry ice method
lar to nickel based method, this method uses Cu
substrate and methane gas depositing carbon on the
According to dry ice method, graphene can be
surface, under high temperature and low pressure.
obtained by burning 3g of Mg ribbon inside a dry
ice bowl, covered up by another dry ice (solid CO2 )
However, atmospheric pressure leads to multilayer
graphene on both copper and nickel [89]. It is pos-
slab. Mg needs to be completely burnt in CO2 and
sible to grow large films of graphene using this
the residue is stirred overnight in 100 mL of 1 M
method [87, 90, 91]. In contrast to methane gas,
HCl. Both Mg and MgO are soluble in water. So,
ethane and propane tend to produce bilayer coat-
the mixture is filtered and the residue is washed
ings; this could be due to their molecular structure
with deionized water till pH turns neutral. To re-
containing a higher concentration of carbon [92].
move water content, the residue needs to be dried
overnight under vacuum at 100 °C resulting in a
2.2.2. Epitaxial growth on silicon carbide graphene yield of 680 mg (92 %) [104].
(SiC)
Graphene can be epitaxially grown on SiC 2.2.4. Deposition
substrate ideal to be used in transistors and cir- There have been a lot of methods reported
cuits because of the thin graphene films obtain- for deposition-based graphene synthesis, primar-
able by this method (>50 µm). In this method, ily because of its practicality and ease of scal-
graphene is obtained by heat treatment of sili- ability [47, 105–110]. Chemical Vapor Deposi-
con carbide (SiC) at 1100 °C [93]. This method tion (CVD) techniques lead to thin graphene films
leads to obtaining epitaxial graphene, while the (675 cm) [25]. Because of the large size of
size of graphene flakes remains dependent upon these films, they are suitable for use in applica-
the size of SiC wafers. The surface of SiC influ- tions, such as touch screens, smart windows, so-
ences the thickness, mobility and carrier density lar cells, flexible LCD and OLED. CVD meth-
of graphene obtained in this system as observed ods enable growth of good quality, large area
in research [94]. Graphene obtained from this pro- graphene films [17, 85, 90]. Most of the CVD
cedure tends to have weak anti-localization, un- techniques are scalable but are financially not
like exfoliated graphene [95]. On the other hand, viable. Spin coating enables nanotube-reinforced
similar to graphene obtained by drawing or peel graphene to be made. Dodecyl-functionalized sin-
off method, SiC-epitaxial graphene displays ex- gle wall carbon nanotubes (DF-SWCNT) are used
tremely large, temperature independent mobility in this process via spin coating and annealing,
but not as high as exfoliated graphene [96]. Even providing stronger, flexible and more conductive
without transfer, graphene obtained through this graphene. HIP co-produced SWCNT were used
method displays massless Dirac fermions [85, 96– in the reported method [111]. DF-SWCNT and
101]. In multi-layered epitaxial graphene, weak trichloromethane solutions were made as reported
van der Waals forces responsible for multilayer co- by Liang et al. [112]. Solution containing DF-
hesion do not necessarily impact electronic prop- SWCNT and trichloromethane was dispersed on a
erties of individual sheets within a stack. This ef- pre-treated Cu foil using spin-coating which was
fect is related to the symmetry of interlayer inter- then annealed for 15 min at 1080 °C under H2 and
actions [102]. This phenomenon does not occur in Ar flow, while keeping the chamber pressure con-
other cases and electronic properties are affected, stant at 933.25 Pa. The samples were reported to be
e.g. in bulk graphite. A 2-inch SiC wafer can offer removed and allowed to cool down to room temper-
cut off frequencies of 100 GHz [103]. This tech- ature afterwards [113]. Supersonic spray offers a
nique offers high quality graphene but at high pro- simple and inexpensive option to deposit graphene
duction cost because SiC substrate is expensive and films on to a variety of substrates, in contrast

Unauthenticated
Download Date | 12/26/17 9:59 AM
574 S. Saqib Shams, Ruoyu Zhang, Jin Zhu

to CVD. This method does not actually synthe- layer-by-layer deposition. Successful transfer and
size graphene but improves it. With this treatment, creation of multilayer films were obtained by
graphene with little or no defects can be obtained, dipping indium tin oxide (ITO) coated glass slides
having ID /IG ratios as low as 0.22 [114]. This sys- in poly(diallyldimethylammonium) (PDDA) aque-
tem uses kinetic spray deposition that benefits from ous solution (1 wt.%, pH = 9.37), to have the
the supersonic acceleration of droplets through a substrate positively charged. The sheets were kept
converging and diverging de Laval nozzle [47]. in the solution for 15 min. To remove the ex-
R-Go gets atomized upon exposure to high gas (air) cess solution, the sheets were rinsed with ultra-
stream [114]. In dip coating, GO is deposited on a pure water and dried under continuous nitrogen
substrate (typically, silica or quartz) and then re- flow. The obtained substrates, coated and positively
duced into graphene via thermal treatment under charged, were then immersed in a poly(sodium 4-
continuous Ar or H2 flow. This method allows bet- styrenesulfonate) modified graphene sheets (PSS-
ter control of film thickness [115]. This method can GS) solution (1 mg·mL−1 , pH = 9.6) for 20 min,
also be beneficial for producing solar panels and followed by washing and drying. This created a
other optoelectronic applications. It offers excel- uniform layer of PSS-GS on the substrate. In or-
lent conductivity and optical transparency in visi- der to obtain multilayers, the PSS-GS coated sub-
ble and near-infrared regions, chemical and thermal strate was positively charged again in PDDA aque-
stability along with the benefit of transferability ous solution, creating another layer of PDDA just
between substrates. Vacuum filtration can lead to as the initial layer. To create a MnO2 layer, the
graphene films by filtering GO suspension through substrates were dipped in a colloidal suspension
a cellulose ester membrane with an average pore of MnO2 nanosheets (0.2 g·L−1 , pH = 9.15) fol-
size of 25 nm. As the solvent drains out, the GO lowed by washing and drying. The final multilayer
forms a layer on the cellulose substrate. Since GO sheets of ITO/ (PDDA/PSS-GS/PDDA/MnO2 )n is
tends to block pores, it does not allow further pas- suitable for use in supercapacitors [119].
sage of the solvent once a pore is filled, what leads
Table 1. Comparison of intensity ratios (D and G bands)
the remaining GO suspension to other (vacant) between graphene obtained by different tech-
pores, making this a self-regulating system. The niques.
thickness of these layers can be controlled by the
amount of solution or its concentration. Once fil- Method ID /IG Reference
tration is completed, the substrate (cellulose) layer Wedged based exfoliation >0.6 [27]
can be dissolved by placing the film sandwich side- Shearing 0.18 [28]
ways in acetone (or other suitable solvents) be- Hot Press Method >1.2 [30]
cause the GO film obtained by this process can be Nanotube Slicing 0.704 [36]
used in lithographic processes. Once an indepen-
Nickel Based 0.24 [84]
dent GO film is obtained, it requires to be reduced
Dry Ice Method 0.84 [104]
to graphene films [106]. Several methods have been
Electrochemical Exfoliation 0.4 [47]
reported to perform this operation, including the
Solvent aided Sonication 0.85 [51]
method in which ordered nano and macroscale
Rebar Graphene 0.1 [113]
FLG is exfoliated from a pre-patterned and pris-
tine HOPG surface and printed on a semiconduc- Supersonic Spray 0.22 [114]
tor substrate by an electrostatic force, reported by Dip Coating 0.81 [115]
Liang et al. [112]. However, a combination of hy- Vacuum Filtration 0.85 [117]
drazine vapor exposure and low-temperature an- Electron Beam Irradiation 1.51 [70]
nealing treatment renders efficient results [117,
118]. Li et al. have reported a method to synthe- Table 1 shows a comparison of D and G band
size multilayer films based on graphene and man- intensity ratios. The intensity ratio or ID /IG is a
ganese dioxide (MnO2 ) sheets using electrostatic measure of disorder degree and average size of the

Unauthenticated
Download Date | 12/26/17 9:59 AM
 Graphene synthesis: a Review 575

sp2 domains [70]. It can also be used to calculate References

the crystallite size of graphene (or graphite) sam- [1] G EIM A.K., N OVOSELOV K.S., Nat Mater., 6 (2007),
ples using Tuinstra-Koenig relation [120]. Lower 183.
degree of structural defects in graphene leads to [2] A LLEN M.J., T UNG V.C., K ANER R.B., Chem. Rev.,
110 (2009), 132.
smaller ID /IG numbers. [3] E NOKI T., S UZUKI M., E NDO M., Graphite Intercala-
tion Compounds and Applications, Oxford University
Press, New York, 2003.
3. Conclusions [4] D ELHAES P., Graphite and precursors, CRC Press,
Amsterdam, 2001.
[5] B OEHM H.P., S ETTON R., S TUMPP E., Pure. Appl.
Graphene has gained an immense amount of Chem., 66 (1994), 1893.
fame and scientific attention from the time of its in- [6] C ASTILLO -M ARTINEZ E., C ARRETERO -G ONZALEZ
ception. This is primarily because of its extraordi- J., S OVICH J., L IMA M.D., J. Mater. Chem. A, 2
(2014), 221.
nary properties and the potential it carries for high
[7] PAULING L., The nature of the chemical bond and the
performance composites. Up-scaling graphene pro- structure of molecules and crystals: an introduction to
duction is a challenge no less than synthesis of modern structural chemistry, Cornell University Press,
high quality graphene itself since many methods Ithaca (NY), 1960.
are not suitable for scaling up. Scientists have cho- [8] G EIM A.K., Science, 324 (2009), 1530.
[9] KOTOV N.A., Nature, 442 (2006), 254.
sen different approaches to synthesize graphene us- [10] R AO C., B ISWAS K., S UBRAHMANYAM K., G OVIN -
ing top-down and bottom-up approaches. Bottom- DARAJ A., J. Mater. Chem., 19 (2009), 2457.
up approaches, like CVD, provide excellent con- [11] S OLDANO C., M AHMOOD A., D UJARDIN E., Car-
trol of thickness and quality, while being unsuit- bon, 48 (2010), 2127.
[12] K RISHNAMOORTHY K., K IM G.-S., K IM S.J., Ultra-
able for industrialization. However, it can be used son. Sonochem., 20 (2013), 644.
for producing graphene for high-end applications, [13] E DWARDS R.S., C OLEMAN K.S., Nanoscale, 5
such as touch screen LCD, OLED, solar panels, etc. (2013) 38.
In contrast to bottom-up approach, top-down ap- [14] WARNER J.H., S CH ÄFFEL F., BACHMATIUK A.,
R ÜMMELI M.H., Graphene: Fundamentals and emer-
proach comprises of breaking down graphene pre-
gent applications, Elsevier, Oxford, 2012.
cursors, such as graphite, GO, etc. into few layer [15] N OVOSELOV K.S., G EIM A.K., M OROZOV S.,
graphene. Many different techniques have surfaced J IANG D., Z HANG Y., D UBONOS S., Science, 306
so far to achieve this goal. Methods, such as shear- (2004), 666.
[16] DATO A., R ADMILOVIC V., L EE Z., P HILLIPS J.,
ing, provide easy and scalable approach. Graphene F RENKLACH M., Nano Lett., 8 (2008), 2012.
is being synthesized on an industrial scale [121], [17] R EINA A., J IA X., H O J., N EZICH D., S ON H.,
however, currently available methods rely on batch B ULOVIC V., Nano Lett., 9 (2008), 30.
production and offer problems associated with it, [18] V ERDEJO R., B ERNAL M.M., ROMASANTA L.J.,
L OPEZ -M ANCHADO M.A., J. Mater. Chem., 21
for example difficulty in controlling consistency of (2011), 3301.
properties which is vital for industrial scale produc- [19] PARK S., RUOFF R.S., Nat. Nanotechnol., 4 (2009),
tion. We believe graphene might not be a ‘one size 217.
fits all’ case, where one particular kind of graphene [20] S EGAL M., Nat. Nanotechnol., 4 (2009), 612.
[21] G EIM A.K., M ACDONALD A.H., Phys. Today, 60 (8)
would be suitable for all applications which is one (2007), 35.
of the reasons why graphene synthesized by differ- [22] S HENDEROVA O., Z HIRNOV V., B RENNER D., Crit.
ent methods is suitable for use in different appli- Rev. Solid State, 27 (2002), 227.
[23] S AKAMOTO J., H EIJST VAN J., L UKIN O.,
cations. Since new methods to synthesize graphene S CHL ÜTER A.D., Angew. Chem. Int. Edit., 48
keep getting more and more interesting, we hope to (2009), 1030.
see a solution that provides not just a fair balance [24] M EYER J.C., G EIM A.K., Nature, 446 (2007), 60.
between scalability and quality, but the best of both. [25] M ITTAL G., D HAND V., R HEE K.Y., PARK S.-J.,
L EE W.R., J. Ind. Eng. Chem., 21 (2015), 11.
The aim of this article is to provide an insight into [26] N OVOSELOV K., J IANG D., S CHEDIN F., B OOTH T.,
graphene synthesis technology by providing an in- K HOTKEVICH V., M OROZOV S., P. Natl. Acad. Sci.
troduction and comparative study. USA, 102 (2005), 10451.

Unauthenticated
Download Date | 12/26/17 9:59 AM
576 S. Saqib Shams, Ruoyu Zhang, Jin Zhu

[27] JAYASENA B., S UBBIAH S., Nanoscale Res Lett., 6 [52] Z HOU M., T IAN T., L I X.F., S UN X.D., Z HANG J.,
(2011), 95. C UI P., Int. J. Electrochem. Sc., 9 (2014), 810.
[28] PATON K.R., VARRLA E., BACKES C., S MITH R.J., [53] L OTYA M., H ERNANDEZ Y., K ING P.J., S MITH R.J.,
K HAN U., Nat. Mater., 13 (2014), 624. N ICOLOSI V., K ARLSSON L.S., J. Am. Chem. Soc.,
[29] M C A LLISTER M.J., L I J.-L., A DAMSON D.H., 131 (2009), 3611.
S CHNIEPP H.C., A BDALA A.A., L IU J., Chem. [54] L IU L., Z HAI J., Z HU C., G AO Y., WANG Y., H AN
Mater., 19 (2007), 4396. Y., Biosens. Bioelectron., 63 (2015), 483.
[30] Z HANG Y., L I D., TAN X., Z HANG B., RUAN X., L IU [55] X U Y., BAI H., L U G., L I C., S HI G., J. Am. Chem.
H., Carbon, 54 (2013), 143. Soc., 130 (2008), 5856.
[31] Z HAN D., S UN L., N I Z.H., L IU L., FAN X.F., [56] H AO R., Q IAN W., Z HANG L., H OU Y., Chem. Com-
WANG Y., Adv. Funct. Mater., 20 (2010), 3504. mun., 48 (2008), 6576.
[32] L EE H., K ANG J., C HO M.S., C HOI J.-B., L EE Y., J. [57] PATIL A.J., V ICKERY J.L., S COTT T.B., M ANN S.,
Mater. Chem., 21 (2011), 18215. Adv. Mater., 21 (2009), 3159.
[33] B RUMFIEL G., Nature, 10 (2009), 1038. [58] E NGLERT J.M., R ÖHRL J., S CHMIDT C.D., G RAUP -
[34] J IAO L., Z HANG L., WANG X., D IANKOV G., DAI NER R., H UNDHAUSEN M., H AUKE F., Adv. Mater.,
H., Nature, 458 (2009), 877. 21 (2009), 4265.
[35] KOSYNKIN D.V., H IGGINBOTHAM A.L., S INITSKII [59] S U Q., PANG S., A LIJANI V., L I C., F ENG X.,
A., L OMEDA J.R., D IMIEV A., P RICE B.K., Nature, M ÜLLEN K., Adv. Mater., 21 (2009), 3191.
458 (2009), 872. [60] W OLTORNIST S.J., OYER A.J., C ARRILLO J.-M.Y.,
D OBRYNIN A.V., A DAMSON D.H., ACS Nano, 7
[36] C HEN J., C HEN L., Z HANG Z., L I J., WANG L.,
(2013), 7062.
J IANG W., Carbon, 50 (2012), 1934.
[61] D ENG C., H U H., G E X., H AN C., Z HAO D., S HAO
[37] C HOUCAIR M., T HORDARSON P., S TRIDE J.A., Nat.
G., Ultrasonics., 18 (2011), 932.
Nanotechnol., 4 (2008), 30.
[62] P INJARI D.V., PANDIT A.B., Ultrasonics., 18 (2011),
[38] B ISWAL M., BANERJEE A., D EO M., O GALE S., En- 1118.
erg. Environ Sci., 6 (2013), 1249. [63] S AFARIFARD V., M ORSALI A., Ultrasonics., 19
[39] C HEN G., W U D., W ENG W., W U C., Carbon, 41 (2012), 823.
(2003), 619. [64] R AMADOSS A., K IM S.J., J. Alloy. Compd., 544
[40] R AMANATHAN T., S TANKOVICH S., D IKIN D., L IU (2012), 115.
H., S HEN H., N GUYEN S., J. Polym. Sci. Pol. Phys., [65] L EE J.K., L EE K., L EE K.I., G AP L.J., I L L.G.,
45 (2007), 2097. Ball-milled graphene nano-powder or ribbon purify-
[41] D REYER D.R., PARK S., B IELAWSKI C.W., RUOFF ing method, involves separating magnetic impurities
R.S., Chem. Soc. Rev., 39 (2010), 228. during stirring suspension using magnet, where impu-
[42] E SWARAIAH V., A RAVIND S.S.J., R AMAPRABHU S., rities are incorporated into graphene powder during
J. Mater. Chem., 21 (2011), 6800. ball-milling, Korea Institute of Science and Technol-
[43] D IKIN D.A., S TANKOVICH S., Z IMNEY E.J., P INER ogy, Seoul, p. 7.
R.D., Nature, 448 (2007), 457. [66] L EON V., Q UINTANA M., H ERRERO M.A., F IERRO
[44] NAIR R., W U H., JAYARAM P., G RIGORIEVA I., J.L.G., H OZ DE LA A., P RATO M., Chem. Commun.,
G EIM A., Science, 335 (2012), 442. 47 (2011), 10936.
[45] S HEN B., L U D.D., Z HAI W.T., Z HENG W.G., J. [67] L IN T., TANG Y., WANG Y., B I H., L IU Z., H UANG
M ATER . C HEM . C, 1 (2013), 50. F., Energ. Environ Sci., 6 (2013), 1283.
[68] B ORAH M., DAHIYA M., S HARMA S., M ATHUR
[46] G URUNATHAN S., H AN J.W., E PPAKAYALA V., K IM
R.B., D HAKATE S.R., Mater. Focus, 3 (2014), 300.
J.-H., Int. J. Nanomed., 8 (2013), 1015.
[69] L IU L., X IONG Z., H U D., W U G., C HEN P., Chem.
[47] PARVEZ K., L I R., P UNIREDD S.R., H ERNANDEZ Y.,
Commun., 49 (2013), 7890.
H INKEL F., WANG S., ACS Nano, 7 (2013), 3598.
[70] PAN D., WANG S., Z HAO B., W U M., Z HANG H.,
[48] L U J., YANG J.-X., WANG J., L IM A., WANG S., WANG Y., Chem. Mater., 21 (2009), 3136.
L OH K.P., ACS Nano, 3 (2009), 2367. [71] E L -K ADY M.F., S TRONG V., D UBIN S., K ANER
[49] H ERNANDEZ Y., N ICOLOSI V., L OTYA M., B LIGHE R.B., Science, 335 (2012), 1326.
F.M., S UN Z., D E S., Nat. Nanotechnol., 3 (2008), [72] M ILLER J.R., Science, 335 (2012), 1312.
563. [73] C OTE L.J., C RUZ -S ILVA R., H UANG J., J. Am. Chem.
[50] A LZARI V., N UVOLI D., S COGNAMILLO S., P IC - Soc., 131 (2009), 11027.
CININI M., G IOFFREDI E., M ALUCELLI G., J. Mater. [74] G AO E., WANG W., S HANG M., X U J., Phys. Chem.
Chem., 21 (2011), 8727. Chem. Phys., 13 (2011), 2887.
[51] N UVOLI D., VALENTINI L., A LZARI V., S COG - [75] A BDELSAYED V., M OUSSA S., H ASSAN H.M.,
NAMILLO S., B ON S.B., P ICCININI M., J. Mater. A LURI H.S., C OLLINSON M.M., E L -S HALL M.S.,
Chem., 21 (2011), 3428. J. Phys. Chem. Lett., 1 (2010), 2804.

Unauthenticated
Download Date | 12/26/17 9:59 AM
 Graphene synthesis: a Review 577

[76] H UANG L., L IU Y., J I L.-C., X IE Y.-Q., WANG T., [100] T ZALENCHUK A., L ARA -AVILA S., K ALABOUKHOV
S HI W.-Z., Carbon, 49 (2011), 2431. A., PAOLILLO S., S YV ÄJ ÄRVI M., YAKIMOVA R.,
[77] C HICHKOV B., M OMMA C., N OLTE S., A L - Nat. Nanotechnol., 5 (2010), 186.
VENSLEBEN VON F., T ÜNNERMANN A., Appl. Phys. [101] L ARA -AVILA S., K ALABOUKHOV A., PAOLILLO
A, 63 (1996), 109. S., S YV ÄJ ÄRVI M., YAKIMOVA R., FAL’ KO V.,
[78] S OKOLOV D.A., S HEPPERD K.R., O RLANDO T.M., arXiv:09091193, 2009.
J. Phys. Chem. Lett., 1 (2010), 2633. [102] H ASS J., VARCHON F., M ILLAN -OTOYA J.-E.,
[79] T RUSOVAS R., R ATAUTAS K., R A ČIUKAITIS G., S PRINKLE M., S HARMA N., H EER DE W.A., Phys.
BARKAUSKAS J., S TANKEVI ČIEN Ė I., N IAURA G., Rev. Lett., 100 (2008), 125504.
Carbon, 52 (2013), 574. [103] L IN Y.-M., D IMITRAKOPOULOS C., J ENKINS K.A.,
[80] Z HOU Y., BAO Q., VARGHESE B., TANG L.A.L., FARMER D.B., C HIU H.-Y., G RILL A., Science, 327
TAN C.K., S OW C.H., Adv. Mater., 22 (2010), 67. (2010), 662.
[81] A MINI S., G ARAY J., L IU G., BALANDIN A.A., A B - [104] C HAKRABARTI A., L U J., S KRABUTENAS J.C., X U
BASCHIAN R., J. Appl. Phys., 108 (2010), 094321. T., X IAO Z., M AGUIRE J.A., J. Mater. Chem., 21
[82] S UTTER P.W., F LEGE J.-I., S UTTER E.A., Nat. (2011), 9491.
Mater., 7 (2008), 406. [105] B LAKE P., B RIMICOMBE P.D., NAIR R.R., B OOTH
[83] P LETIKOSI Ć I., K RALJ M., P ERVAN P., B RAKO R., T.J., J IANG D., S CHEDIN F., Nano Lett., 8 (2008),
C ORAUX J., N’ DIAYE A., Phys. Rev. Lett., 102 (2009), 1704.
056808. [106] E DA G., FANCHINI G., C HHOWALLA M., Nat. Nan-
[84] W EATHERUP R.S., BAYER B.C., B LUME R., otechnol., 3 (2008), 270.
D UCATI C., BAEHTZ C., S CHL ÖGL R., Nano Lett., [107] L I D., M ÜLLER M.B., G ILJE S., K ANER R.B., WAL -
11 (2011), 4154. LACE G.G., Nat. Nanotechnol., 3 (2008), 101.
[85] K IM K.S., Z HAO Y., JANG H., L EE S.Y., K IM J.M., [108] R AHAMAN M., I SMAIL A.F., M USTAFA A., Polym.
K IM K.S., Nature, 457 (2009), 706. Degrad. Stabil., 92 (2007), 1421.
[86] Z HANG Y., Z HANG L., Z HOU C., Accounts Chem. [109] KO Y.U., C HO S.-R., C HOI K.S., PARK Y., K IM
Res., 46 (2013), 2329. S.T., K IM N.H., J. Mater. Chem., 22 (2012), 3606.
[87] BAE S., K IM H., L EE Y., X U X., PARK J.-S., Z HENG [110] YAMAGUCHI H., E DA G., M ATTEVI C., K IM H.,
Y., Nat. Nanotechnol., 574 (2010), 574. C HHOWALLA M., ACS Nano., 4 (2010), 524.
[88] R AFIEE J., M I X., G ULLAPALLI H., T HOMAS A.V., [111] N IKOLAEV P., B RONIKOWSKI M.J., B RADLEY
YAVARI F., S HI Y., Nat. Mater., 11 (2012), 217. R.K., ROHMUND F., C OLBERT D.T., S MITH K.,
[89] L ENSKI D.R., F UHRER M.S., J. Appl. Phys., 110 Chem. Phys. Lett., 313 (1999), 91.
(2011), 013720. [112] L IANG F., S ADANA A.K., P EERA A., C HATTOPAD -
[90] L I X., C AI W., A N J., K IM S., NAH J., YANG D., HYAY J., G U Z., H AUGE R.H., Nano Lett., 4 (2004),
Science, 324 (2009), 1312. 1257.
[91] L EVENDORF M.P., RUIZ -VARGAS C.S., G ARG S., [113] YAN Z., P ENG Z., C ASILLAS G., L IN J., X IANG C.,
PARK J., Nano Lett., 9 (2009), 4479. Z HOU H., ACS Nano, 8 (2014), 5061.
[92] WASSEI J.K., M ECKLENBURG M., T ORRES J.A., [114] I RISSOU E., L EGOUX J.-G., RYABININ A., J ODOIN
F OWLER J.D., R EGAN B., K ANER R.B., Small, 8 B., M OREAU C., J. Therm. Spray Techn., 17 (2008),
(2012), 1415. 495.
[93] S UTTER P., Nat. Mater. 8 (2009), 171. [115] WANG X., Z HI L., M ÜLLEN K., Nano Lett. 8 (2008),
323.
[94] O HTA T., B OSTWICK A., M C C HESNEY J., S EYLLER
[116] L IANG X., C HANG A.S.P., Z HANG Y., H ARTENECK
T., H ORN K., ROTENBERG E., Phys. Rev. Lett., 98
B.D., C HOO H., O LYNICK D.L., C ABRINI S., Nano
(2007), 206802.
Lett., 9 (1) (2009), 467.
[95] M OROZOV S., N OVOSELOV K., K ATSNELSON M., [117] S TANKOVICH S., D IKIN D.A., P INER R.D.,
S CHEDIN F., P ONOMARENKO L., J IANG D., Phys. KOHLHAAS K.A., K LEINHAMMES A., J IA Y., W U
Rev. Lett., 97 (2006), 016801. Y., N GUYEN S.T., RUOFF R.S., Carbon, 45 (7)
[96] J OBST J., WALDMANN D., S PECK F., H IRNER R., (2007), 1558.
M AUDE D.K., S EYLLER T., http://arxiv.org/ [118] WATCHAROTONE S., D IKIN D.A., S TANKOVICH S.,
abs/0908.1900,2009. P INER R., J UNG I., D OMMETT G.H.B., E VME -
[97] S HEN T., G U J., X U M., W U Y., B OLEN M., C A - NENKO G., W U S.-E., C HEN S.-F., L IU C H .,-P.,
PANO M., Appl. Phys. Lett., 95 (2009), 172105. N GUEN S.T., RUOFF R.S., Nano Lett., 7 (7) (2007),
[98] W U X., H U Y., RUAN M., M ADIOMANANA N.K., 1888.
H ANKINSON J., S PRINKLE M., Appl. Phys. Lett., 95 [119] L I Z., WANG J., L IU X., L IU S., O U J., YANG S.,, J.
(2009), 223108. Mater. Chem., 21 (2011), 3397.
[99] A LEXANDER -W EBBER J., BAKER A., JANSSEN T., [120] G ÓMEZ -NAVARRO C., W EITZ R.T., B ITTNER A.M.,
T ZALENCHUK A., L ARA -AVILA S., K UBATKIN S., S COLARI M., M EWS A., B URGHARD M., K ERN N.,
Phys. Rev. Lett., 111 (2013), 096601. Nano Lett., 7 (11) (2007), 3499.

Unauthenticated
Download Date | 12/26/17 9:59 AM
news & views

be explained by a reversible protonation quantum dots and metallic nanoparticles, for the creation of bigger assemblies.
and deprotonation of the MPA co-ligand which would allow plasmonic interactions Overcoming these challenges could lead to
in neutral and alkaline pH. These results between the two to be studied in detail3. numerous applications in nanophotonics
suggest that the programmed quantum Challenges for the future include and nanoelectronics. ❐
dot complexes could be used to monitor developing a better understanding of how
enzymatic reactions that involve a pH the experimental conditions and other Yan Liu is at the Biodesign Institute and in the
change in cellular compartments. An factors (such as the length of the various Department of Chemistry and Biochemistry, Arizona
obvious future direction is to explore domains in the DNA ligands) influence the State University, Tempe, Arizona 85287, USA.
quantum dot complexes that can respond size and valency of the quantum dots, and e-mail: yan_liu@asu.edu
to other environmental triggers, such extending the approach to other materials
as changes in ion concentrations or the systems to make quantum dots that are References
presence of biomarkers. The strategy less toxic and emit over a wider range of 1. Choi, C. L. & Alivisatos, A. P. Annu. Rev. Phys. Chem.
61, 369–389 (2010).
developed by Kelley and co-workers wavelengths. Combining this approach 2. Tikhomirov, G. et al. Nature Nanotech. 6, 485–490 (2011).
could also be readily used to create hybrid with DNA scaffolds4 would increase control 3. Jin, Y. & Gao, X. Nature Nanotech. 4, 571–576 (2009).
structures containing both semiconductor over long-range order, which would allow 4. Seeman, N. C. Annu. Rev. Biochem. 79, 65–87 (2010).

GRAPHENE ELECTRONICS

Thinking outside the silicon box

The best applications of graphene will be those that exploit its basic characteristics, rather than try to change them.

Tomás Palacios

G
raphene, a single layer of Infrared photons
carbon atoms arranged in a Metal
honeycomb lattice, is on its way
to revolutionizing electronics. The
combination of perfect charge-carrier
confinement with very high carrier
mobility in a two-dimensional structure
makes it an ideal material for highly
scaled electronics. These properties are
attracting more and more electrical
engineers to the ‘flatland’ of graphene,
and the effort is paying off 1. Recently,
transistors with frequency performance in
excess of 300 GHz have been reported2 and
companies and research groups around the
Doped silicon Silicon Dielectric Si MOSFET
world are engaged in a race to demonstrate
circuits of increasing complexity 3–5.
However, in spite of this progress, the Figure 1 | Schematic diagram of an infrared detector in which graphene nanoribbons (top) are integrated
most promising applications of graphene onto a prefabricated silicon chip. Graphene-based ambipolar electronic devices and various types of
in electronics for the immediate future are sensor could also be readily integrated onto silicon chips. Figure used and modified with permission from
unlikely to involve competing directly with Kyeong-Jae Lee, MIT.
conventional transistors.
Although graphene transistors are
very fast, at present they lack traditional As I argued at a recent conference in Ambipolar electronics is based on a new
current saturation, have poor on–off Boston organized by Nature Publishing family of nonlinear devices that rely on the
current ratios, and are much slower Group, ‘Graphene: The Road to Applications’, fact that the conduction and valence bands
(typically 100 times) in real circuit amplifiers and digital logic may not be the in graphene meet at a point called the Dirac
applications. These serious limitations best applications for this amazing material, point. This band structure makes it possible
are the result of graphene not having a at least in the short term. In the past few to continuously sweep the Fermi level from
bandgap, and they are often overlooked months, there have been several proposals the conduction band to the valence band by
in the fast-moving world of research. for graphene devices and circuits that attempt just changing the gate voltage of a field-
They mean that graphene transistors, to exploit the fundamental properties of the effect transistor. This in turn changes the
although excellent tools to probe the material, including its zero bandgap, rather channel conduction from electron-based to
transport properties of this material, still than trying to overcome them. Examples hole-based and gives graphene transistors
have a long way to go to compete with include ambipolar electronics5–10, infrared their characteristic V-shape transfer
state-of-the-art silicon and compound detectors11,12 and heterogeneous graphene– curve. This ambipolar conduction, which
semiconductor electronics. silicon systems13. is responsible for the poor performance

464 NATURE NANOTECHNOLOGY | VOL 6 | AUGUST 2011 | www.nature.com/naturenanotechnology

© 2011 Macmillan Publishers Limited. All rights reserved

 news & views

of amplifiers and logic devices based on variety of graphene-based devices onto applications of any sufficiently new and
graphene, actually enables a new class a prefabricated silicon chip and connect innovative technology have always been —
of nonlinear radiofrequency electronic the devices to the chip through the top- and will continue to be — applications
devices. Frequency multipliers5,6, , high- level metallization (Fig. 1). In an early created by that technology.”14 There are
frequency mixers7 and digital modulators8,9 demonstration of this concept 13, a silicon many applications that benefit from today’s
have all been recently demonstrated using chip was covered with a graphene layer that graphene, but the best are still to come. ❐
this principle, with operating frequencies was patterned to form interconnections
as high as 16 GHz and radiofrequency and delay lines that were controlled by Tomás Palacios is in the Department of
efficiencies in excess of 90% (ref. 10). the chip. A huge amount of research Electrical Engineering and Computer Science,
Infrared photodetectors are another has been devoted to the development of Massachusetts Institute of Technology, Cambridge,
class of graphene device that shows higher methods for integrating new materials Massachusetts 02139, USA.
performance than its silicon counterpart (such as III–V semiconductors) with e-mail: tpalacios@mit.edu
today, despite its relative immaturity 11,12. silicon: with graphene, the integration is
Again, these devices benefit from the zero remarkably simple. References
bandgap of graphene, which allows them Graphene is an amazing material 1. Schwierz, F. Nature Nanotech. 5, 487–496 (2010).
to absorb light from the infrared to the looking for an application. Meanwhile, 2. Liao, L. et al. Nature 467, 305–308 (2010).
3. Palacios, T., Hsu, A. & Wang, H. IEEE Commun. Mag.
ultraviolet with almost equal strength. the electronics industry is a trillion-dollar
48, 122–128 (June 2010).
Graphene photodetectors have already giant looking for the next switch. Although 4. Lin, Y-M. et al. Science 332, 1294–1297 (2011).
been fabricated at very high frequencies graphene may very well provide that switch, 5. Moon, J. S. et al. IEEE Electron Dev. Lett. 32, 270–272 (2011).
and at wavelengths that cannot be accessed it is important to keep our eyes open to 6. Wang, H., Nezich, D., Kong, J. & Palacios, T.
IEEE Electron Dev. Lett. 30, 547–549 (2009).
with silicon-based devices. These devices other opportunities, particularly those 7. Wang, H., Hsu, A., Wu, J., Kong, J. & Palacios, T.
will have an important role in the next that result naturally from the properties of IEEE Electron Dev. Lett. 31, 906–908 (2010).
generation of infrared night-vision graphene. Ambipolar nonlinear electronics, 8. Yang, X., Liu, G., Baladin, A. A. & Mohanram, K. ACS Nano
4, 5532–5538 (2010).
systems, optical communications and broadband photodetectors and advanced 9. Harada, N., Yagi, K., Sato, S. & Yokoyama, N. Appl. Phys. Lett.
chemical analysis. sensors are some of them, and they can be 96, 012102 (2010).
Finally, a key property of graphene, easily integrated onto silicon platforms. It is 10. Wang, H., Hsu, A., Kim, K. K., Kong, K. & Palacios, T. IEEE
International Microwave Symposium (Baltimore, 5–10 June 2011).
which is often forgotten, is its ability to tempting to try to fit any new material into 11. Xia, F., Mueller, T., Lin, Y-M., Valdes-Garcia, A. & Avouris, P.
integrate with almost any substrate. This existing technologies and needs. However, Nature Nanotech. 4, 839–843 (2009).
means that it does not need to replace we should never forget the words of the 12. Bonaccorso, F., Sun, Z., Hasan, T. & Ferrari, A. C. Nature Photon.
4, 611–622 (2010).
silicon to be useful: rather, graphene semiconductor pioneer and Nobel Prize 13. Lee, K-J., Qazi, M., Kong, J. & Chandrakasan, A. P.
can complement silicon. In the near winner, Herb Kroemer, which he calls his IEEE Trans. Electron Dev. 57, 3418–3425 (2010).
future, it will be possible to transfer a lemma on new technology: “The principal 14. Kroemer, H. Rev. Mod. Phys. 73, 783–793 (2001).

BIOCOMPUTING

DNA computes a square root

Complex molecular circuits with reliable digital behaviour can be created using DNA strands.

Yaakov Benenson

T
he field of molecular computing In molecular logic, analogy is often the inputs tends to disappear, particularly
will shortly enter its third decade, drawn between the concentration in multistep reactions. Counteracting
but basic terms used by its of a molecule and logic values: low these natural trends with the experimental
practitioners are still the subject of lively concentrations represent ‘false’ or ‘0’ and construction of large ‘digital’ networks is of
debate. The terms ‘logic’ and ‘digital’, high concentrations ‘true’ or ‘1’. To launch a value for two reasons. First, it might help to
for example, are frequently objected to computation, concentrations of molecules unravel the ‘mystery of life’, suggesting how
on the grounds that molecules cannot designated as inputs are adjusted to set information processing in cells (including
be reduced to zeros and ones. However, their values to true or false as desired; seemingly Boolean transformations2)
it is only habit that allows electronic this triggers downstream processes that could be implemented in practice.
switches to be viewed as a ‘natural’ fit for eventually generate products designated Second, it can clarify whether the digital
abstract models of computation, but not as outputs. In a well-designed digital paradigm is in fact the best approach to
molecules. In fact, such models can be molecular circuit, the outputs depend on molecular computing.
projected onto any physical system; the inputs in a Boolean fashion, for example, Qian and Winfree build on rich
ultimate test is whether the projection an AND logic function is implemented by a previous work that has explored DNA–
is useful. Writing in Science, Lulu Qian process in which the output is high only if DNA interactions and the resulting
and Erik Winfree of the California input 1 is high and input 2 is high. interconversion of single-stranded (ss), as
Institute of Technology now report how The more complex these Boolean well as partially and fully double-stranded
digital abstraction can guide the design relations become, the more difficult it DNA species, without any chemical change
of DNA-based molecular circuits of is to implement corresponding circuits to the DNA oligomers themselves3–5. The
unprecedented complexity 1. because the high/low distinction between key to their design lies in a systematic

NATURE NANOTECHNOLOGY | VOL 6 | AUGUST 2011 | www.nature.com/naturenanotechnology 465

© 2011 Macmillan Publishers Limited. All rights reserved

 ARTICLES
PUBLISHED ONLINE: 26 JUNE 2011 | DOI: 10.1038/NNANO.2011.94

Bi- and trilayer graphene solutions

Chih-Jen Shih1, Aravind Vijayaraghavan1,2, Rajasekar Krishnan1, Richa Sharma1, Jae-Hee Han1,3,
Moon-Ho Ham1, Zhong Jin1, Shangchao Lin1,4, Geraldine L.C. Paulus1, Nigel Forest Reuel1,
Qing Hua Wang1, Daniel Blankschtein1 and Michael S. Strano1 *

Bilayer and trilayer graphene with controlled stacking is emerging as one of the most promising candidates for post-
silicon nanoelectronics. However, it is not yet possible to produce large quantities of bilayer or trilayer graphene with
controlled stacking, as is required for many applications. Here, we demonstrate a solution-phase technique for the
production of large-area, bilayer or trilayer graphene from graphite, with controlled stacking. The ionic compounds iodine
chloride (ICl) or iodine bromide (IBr) intercalate the graphite starting material at every second or third layer, creating
second- or third-stage controlled graphite intercolation compounds, respectively. The resulting solution dispersions are
specifically enriched with bilayer or trilayer graphene, respectively. Because the process requires only mild sonication, it
produces graphene flakes with areas as large as 50 mm2. Moreover, the electronic properties of the flakes are superior to
those achieved with other solution-based methods; for example, unannealed samples have resistivities as low as ∼1 kV
and hole mobilities as high as ∼400 cm2 V–1 s–1. The solution-based process is expected to allow high-throughput
production, functionalization, and the transfer of samples to arbitrary substrates.

G
raphene, with sheets consisting of fewer than 10 stacked however, they all share the limitation that the produced flakes
layers of sp 2-hybridized carbon lattice1, combines outstand- have dimensions less than those suitable for conventional photo-
ing electronic and mechanical properties2–4. Owing to their lithography, a manufacturing method for economical mass pro-
distinct electronic band structures, AB-stacked bilayer2,3,5 and tri- duction22–27. Both GO and direct-exfoliation methods also lack
layer4 graphene have extraordinary potential for next-generation direct control over the layering and stacking of the exfoliated
optoelectronic and microprocessor applications. However, these flakes28. In contrast, when combined with the well-developed inter-
promising materials require the development of new synthesis calation chemistry of graphite29,30, it becomes possible to obtain
methods to effectively control the number of AB-stacked layers by large-size graphene dispersions with layering control. Graphene
means of graphite exfoliation and processing. There has been sheets from covalent graphite intercalation compounds (GIC)31–33,
substantial progress in the production of graphene through the ionic GIC34–36 and expanded graphite37 have been reported.
micromechanical cleavage of mono- and few-layer graphene6 from Nevertheless, the produced graphene is found to be either signifi-
highly-ordered pyrolytic graphite (HOPG). However, this approach cantly defective following the intercalation–expansion process31–33
lacks a viable mechanism for scalable manufacturing. Epitaxial or remains too small in area for scalable applications34,35. In
growth on an SiC substrate7 provides advantages in this respect, addition, all liquid-phase methods can only produce size- and
but imposes constraints on the choice of substrate. Recent results layer-polydispersed graphene28,38. In the present Article, we
have demonstrated chemical vapour deposition (CVD) for graphene demonstrate an important link between the stage number of a
growth and transfer to various substrates8–10. However, for bi- or tri- non-covalent, precursor GIC and the resulting mode in the layer
layer applications, SiC and CVD graphene films tend to be turbos- number dispersion of the resulting exfoliated solution, allowing
tratic, where slight deviation from the AB stacking destroys the for solutions enriched in bi- and trilayer graphene. The efficiency
unique electronic structures of the bilayer and trilayer graphene, of the approach is such that a milder ultrasonic processing is suffi-
making it electronically similar to monolayer graphene11,12. cient, so larger flake sizes are recovered, enabling devices produced
Alternatively, exfoliation of pristine graphite into a liquid phase is by conventional photolithography.
easily scalable and allows for more precise chemical modification
in solution13,14. If layering can be controlled during dissolution, Stage-controlled intercalation
this would provide a viable route for the mass production of AB- HOPG was chosen as a high-quality graphite source, and we used
stacked bi- and few-layer graphene on arbitrary substrates for elec- the halogen intercalants ICl (to form Stage-2 ionic GIC) and IBr
tronic devices. The liquid-phase production of chemically converted (to form Stage-3 ionic GIC). The staging phenomenon is related
graphene from graphene oxide (GO) is a high-yield method for to the effect of long-range lattice strain29. Once the exposed
monolayer graphene-based solutions15–19. Despite producing high- surface graphite planes have interacted with the intercalated
concentration dispersions with large flake areas, a substantial species, the intercalant molecules are introduced into the host
number of defects are introduced during the reduction material from the exposed end surfaces, and sequentially form
process20,21, and the intrinsic properties of graphene are only par- layered structures. In this scheme, Stage-2 and Stage-3 GICs have
tially restored22. Liquid-phase exfoliation of graphite in organic sol- every second or third layer of the graphite lattice intercalated,
vents23–25 and surfactant aqueous solutions26,27 are promising routes respectively30, as shown in Fig. 1a. The detailed intercalation
to the production of pristine graphene dispersions. To date, method is described in Supplementary Section S1. The stage of

1
Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA, 2 School of Computer Science, The
University of Manchester, Manchester M13 9PL, UK, 3 Department of Energy IT, Kyungwon University, Seongnam, Gyeonggi-do 461-701, South Korea,
4
Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA. *e-mail: strano@mit.edu

NATURE NANOTECHNOLOGY | VOL 6 | JULY 2011 | www.nature.com/naturenanotechnology 439

© 2011 Macmillan Publishers Limited. All rights reserved.

ARTICLES NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2011.94

a b c d
(0 0 2) HOPG
(0 0 4)
(0 0 6)

(0 0 4) (0 0 8)
IBr 3rd Stage GIC
(0 0 12)
(0 0 5) (0 0 6) (0 0 7) (0 0 9) (0 0 10) (0 0 11)

÷(Intensity) (a.u.)
(0 0 3)
(0 0 6) ICl 2nd Stage GIC

(0 0 9)
(0 0 4) (0 0 5) (0 0 7) (0 0 8)

(0 0 2) IBr 3rd Stage EG

(0 0 4)

(0 0 2) ICl 2nd Stage EG

(0 0 4)

20 30 40 50 60 70 80 90
2
e

i ii iii iv v

Figure 1 | Graphene dispersions from ionic graphite intercalation compounds (GIC). a, Three-dimensional computer-generated molecular models (carbon,
grey; iodine, red; bromine, yellow; chlorine, cyan) of HOPG (top), IBr Stage-3 GIC (middle) and ICl Stage-2 GIC (bottom). b, Comparison of XRD patterns
of HOPG, IBr Stage-3 GIC, ICl Stage-2 GIC, IBr Stage-3 expanded graphite (EG) and ICl Stage-2 EG. For IBr Stage-3 GIC, an additional three peaks appear
between two of the three main peaks corresponding to the (0 0 2), (0 0 4) and (0 0 6) planes in HOPG, which is a clear signature of Stage-3 GIC (the
intercalant layer inserts between every three graphite layers). Analogous XRD patterns also apply to the ICl Stage-2 GIC. Both Stage-3 and Stage-2 expanded
graphites show much weaker (0 0 4) peaks than the HOPG, while the (0 0 6) peak becomes unobservable. c, Photographs of HOPG (top), Stage-3 GIC
(middle) and Stage-3 EG (bottom) (the HOPG and GIC materials are held up with tweezers with the edges toward the viewer). d, Photographs of HOPG
(top), Stage-2 GIC (middle), and Stage-2 EG (bottom). e, Photograph of steps involved in forming suspensions of EGs in 2 wt% sodium cholate aqueous
solution. As-prepared Stage-3 (i) and Stage-2 (ii) expanded materials floating on the solution, followed by 30 min homogenization (iii) and 10 min sonication
(iv). (v) Clear and grey graphene solutions after 2,000 r.p.m. centrifugation.

the GIC denotes the crystallographic arrangement of intercalant and The obtained Stage-2 and Stage-3 GICs were then used as pre-
graphite layers, with a superlattice structure confirmed by X-ray dif- cursors of expanded graphite. Briefly, the GICs were annealed at
fraction (XRD) patterns along the c-axis39. The top three profiles in 800 8C for 5 min under an argon atmosphere. The IBr or ICl inter-
Fig. 1b illustrate the XRD patterns of HOPG, IBr Stage-3 GIC and calants between the graphene layers were volatilized rapidly and
ICl Stage-2 GIC, respectively. The crystallographic evidence removed completely during the ‘thermal shock’ process41. The sub-
(Supplementary Section S2) confirms that halogen molecules inter- stantial, anisotropic volume changes for Stage-3 and Stage-2
calate every three and two graphite layers. expanded samples are shown in Fig. 1c,d, respectively. During the
According to the XRD analysis, it should be noted here that thermal expansion, the bilayer and trilayer crystal domains are iso-
the intercalant peaks in the GIC superlattice exhibit reduced lated and crosslinked by grain boundaries with stacking faults
intensities and peak broadenings compared to perfect superlattice formed during the intercalation process. It is also expected that
crystals39. This suggests that the synthesized Stage-2 GIC is domi- small sub-domains may collapse and recombine to form thicker,
nated by a bilayer superlattice structure, but includes minority larger crystallites. The symmetry-breaking of the graphite lattice
crystal phases (including graphite and GIC sub-domains). In for Stage-3 and Stage-2 expanded materials is shown in the XRD
addition, the lateral non-uniformity of grain sizes, which is analysis in the bottom two profiles in Fig. 1b. The peak positions
described by the Daumas–Hérold model40, also plays an impor- corresponding to the (0 0 2) and (0 0 4) planes are almost identical
tant role. These imperfections arise partly because of the fact to those in HOPG. This indicates that the intercalant layers are
that the intercalation process is not completely optimized29, and removed and the graphite lattice parameters restored. However,
partly because of the polycrystalline nature of the HOPG. Better the relative intensity of the (0 0 4) peak is decreased from the
control of the intercalation kinetics29 is currently under investi- HOPG case, and no (0 0 6) peak is observed for either expanded
gation with a view towards achieving 100% intercalation to the sample. Hence, the sublattices in both materials consist of fewer
desired stage. than four graphite layers, with long-range order greater than

440 NATURE NANOTECHNOLOGY | VOL 6 | JULY 2011 | www.nature.com/naturenanotechnology

© 2011 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2011.94 ARTICLES
a i ii b
Ring II
1

SiO2 substrate SiO2 substrate

Ring I
iii iv
Ring 10 m
Capillary flow
2
c
SiO2 substrate SiO2 substrate 10 m

v vi
Contact line
Big flake
1
moves back

SiO2 substrate SiO2 substrate

Figure 2 | On-chip separation method based on graphene size, using the ‘coffee-ring effect’. a, Schematic of the separation process. (i) A droplet of
graphene dispersion on the SiO2 substrate with contact angle u1. (ii) Because the evaporation rate is greatest at the drop edge, some graphene flakes deposit
along the edge of the drop at the beginning. The presence of these flakes facilitates pinning of the contact line at the drop edge. (iii) The outward capillary
flow of liquid is induced from the bulk of the drop, and the contact angle u2 becomes gradually smaller. (iv) The liquid flow carries graphene flakes out
towards the edge of the drop, which results in the formation of a ‘coffee ring’. (v) By controlling the rate of evaporation, all small flakes collect together to
form the coffee ring, and particularly large flakes are found inside the coffee ring. (vi) When the contact angle u2 becomes too small, the contact line moves
back and a smaller droplet forms with contact angle u1 again. b, Low- (left) and high- (right) magnification images of a matrix of small graphene solution
droplets (50 nl each) on a SiO2 wafer, spaced by 4 mm, using a microprinter. Bright dots are associated with surfactant residues after drying. c, Low- (left)
and high- (right) magnification images of the ‘coffee ring’.

a G b c d
= 633 nm 2D = 633 nm 20 nm
D
Monolayer
Monolayer
Bilayer
Bilayer
Intensity (a.u.)

Intensity (a.u.)

3−4 layer 3−4 layer 2 nm 5 m

e f
5 layer 5 layer nm
2 m 2 m 10
5−10 layer 5
5−10 layer
0
> 10 layer
> 10 layer
–5
1,200 1,600 2,000 2,400 2,800 2,400 2,600 2,800
–10
Raman shift (cm−1) Raman shift (cm−1)

Figure 3 | Characterization of graphene flakes. a,b, Raman spectra (excitation wavelength l ¼ 633 nm) for graphene flakes with different numbers of
stacked layers. Note that for multilayer graphene flakes, the D-peaks (1,340 cm21, a) are absent, but for the relatively smaller monolayer graphene flake, the
weak D-peak arises due to edge effects. 2D Raman spectra (b) corresponding to specific numbers of stacked layers of our graphene flakes show the same
features observed using the Scotch-tape method on the same substrate. c, HRTEM images of the edges of bilayer (left) and trilayer (right) graphene
flakes. d, TEM image of a representative multilayer graphene flake (area, 50 mm2). e,f, Representative AFM images of bilayer (e) and 3–4-layer graphene
flakes (f).

six-layer almost completely absent. Note that, compared with between crystallites and expel argon gas trapped inside voids. This
natural graphite, in which the graphene layers are strongly bound pretreatment process is required, because ultrasonication is only
by van der Waals interactions, the systematic crystalline defects effective for dispersions inside the liquid, although we expect that
that crosslink bilayer and trilayer crystallites in expanded materials the agitation will result in some degree of recombination and aggre-
should be far easier to break up in a liquid-phase dispersion by gation of crystallites in expanded graphite. The protocol of ultrasoni-
mild ultrasonication. This also provides a link between the GIC cation implemented here involves significantly lower strength
stage number and layer number dispersion in the resulting solution, compared to direct sonication of graphite24,27. This appears to pre-
and hence a route to bi- and trilayer solutions. serve large flakes areas in the work reported here. Sonication con-
ditions that are too mild result in ineffective exfoliation of
Size separation of graphene flakes crosslinked defects, so most of the material pellets during centrifu-
The dispersion process for expanded graphite is shown in Fig. 1e. gation (the yield is estimated as 3.0%; Supplementary Section S3).
Detailed materials and methods associated with the dispersion To locate and isolate large flakes from graphene solutions for
process are described in Supplementary Section S3. The main further characterization, an on-chip separation method based on
purpose of homogenization is to disassemble weak crosslinks size using the ‘coffee-ring effect’42–44 was developed. A schematic

NATURE NANOTECHNOLOGY | VOL 6 | JULY 2011 | www.nature.com/naturenanotechnology 441

© 2011 Macmillan Publishers Limited. All rights reserved.

ARTICLES NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2011.94

a b c Bilayer Monolayer
sub-domain sub-domain

100 100 k8,1

10 10
Flake area ( m2)

Flake area ( m2)

Defects that cross-link
sub-domains
1 1 1
d
0.8 Monolayer

concentration
Normalized
Ref. 24 Ref. 24 0.6
0.1 0.1
0.4
Bilayer
Ref. 27 Ref. 27
0.2
Trilayer
0.01 0.01 0
1 2 3−4 5 5−10 >10 1 2 3−4 5 5−10 >10 0 1 2 3 4 5 6
Graphene layer Graphene layer Dimensionless time (kt)

e 1 1 f 1 1
Monolayer Monolayer
0.8 0.8 0.8 0.8
concentration

concentration

concentration

concentration
Bilayer
Normalized

Normalized

Normalized

Normalized
Monolayer Bilayer
0.6 Trilayer 0.6 0.6 0.6
Trilayer
0.4 0.4 0.4 0.4 40% Monolayer
Bilayer 35%
Bilayer
0.2 0.2 15% 0.2 0.2
Trilayer 9%
8% 3% Trilayer
0 0 0 0
0 1 2 3 4 5 6 0.01 0.1 1 0 1 2 3 4 5 6 0.01 0.1 1
Dimensionless time (kt) Dimensionless time (kt) Dimensionless time (kt) Dimensionless time (kt)

Figure 4 | Evidence for layer- and size-controlled graphene dispersions from different GICs. Distributions for number of stacked layers and the area of
graphene flakes inside ‘coffee rings’ from IBr Stage-3 (a) and ICl Stage-2 (b) GICs. Regions occupied by graphene solutions produced from direct exfoliation
of natural graphite24,27 are also shown for comparison. c, Schematic describing the exfoliation process of a 9-layer graphite lattice into 8- and 1-layer products
with rate constant k8,1 for a Stage-2 EG system. d, Calculated concentrations of mono- (blue curve), bi- (red curve) and tri- (green curve) layer graphene
dispersions as a function of dimensionless time using graphite with all interlayer bond strengths equal. e, Calculated concentrations for Stage-3 EG with every
third interlayer bonds weaker. Adjacent graph shows the predicted time (vertical line) corresponding to the experimentally observed layer distribution for
Stage-3 solutions. f, Calculated concentrations for Stage-2 EG with every third interlayer bond weaker. Adjacent graph shows the predicted time (vertical line)
corresponding to the experimentally observed layer distribution for Stage-2 solutions.

illustration of the mechanism is shown in Fig. 2a. In the formation AB stacked. For the monolayer graphene, a single Lorentzian at
of size-sorted graphene rings using the coffee-ring effect, the size of 2,660 cm21 is seen, and with increasing number of layers, the
the droplet, the concentration of graphene and the substrate temp- other spectral component at 2,690 cm21 gradually dominates. The
erature all play important roles45,46. The optimized process is flakes were categorized as monolayer, bilayer, 3–4-layer, 5-layer,
described in Supplementary Section S4. The on-chip separation 5–10-layer and .10-layer graphene after comparing the obtained
method can be further scaled up by (i) increasing the concentration Raman 2D peak shapes with those in the literature47,48. It is desirable
of graphene and (ii) using micro/nanoprinting technology. To to quantify the trilayer graphene yield from the Stage-3 GIC.
demonstrate this, a matrix of graphene droplets (50 nl each) with However, the 2D signature in Raman spectroscopy is unable to
a droplet spacing of 4 mm was deposited onto a SiO2–Si substrate unambiguously distinguish 3- and 4-layer graphene. It can only dis-
using a microprinter, as shown in Fig. 2b. During the evaporation tinguish 3- or 4-layer flakes from bilayer and 5-layer flakes, and this
process, 3–4 rings form for each drop and, in the end, residual sur- fact should be recognized during the analysis of flake-thickness dis-
factant segregates in the central region, which can be rinsed away tribution discussed below. The D-peak (1,340 cm21), which is a
using deionized water, as shown in Fig. 2c (left image). The larger measure of defects and chemical functionalization, is absent
flakes in the solution can then be found inside each ring, which is except in monolayer graphene flakes. We note that the monolayers
composed of the smaller flakes. We also noticed that the proposed observed are significantly smaller than the bilayer and thicker flakes,
method can self-align graphene ribbons along the radial direction and that the D-peak in monolayers results from symmetry-breaking
of the ring (right image in Fig. 2c). at the edges47, because the laser spot is the same size as the flake area
and includes the flake edges as well as the interior. The exfoliated
Evidence for layer-controlled graphene dispersions graphene retains its pristine structure and does not contain signifi-
Raman spectroscopy was used to characterize the isolated graphene cant defects or functionalization as a result of the intercalation,
flakes. The full Raman spectra and 2D peaks for representative gra- expansion or solution processing. Additional Raman analysis is pre-
phene flakes are shown in Fig. 3a,b, respectively. The 2D peaks cor- sented in Supplementary Figs S7 and S8.
responding to specific numbers of stacked layers in our exfoliated Large graphene flakes can also easily be found on a transmission
graphene flakes exhibit the same layering dependence as that electron microscopy (TEM) grid. Figure 3c shows high-resolution
observed in micromechanically cleaved graphene on the same sub- TEM (HRTEM) images of edges of bilayer and trilayer flakes.
strate47,48, indicating that the solution-processed graphene flakes are Figure 3d shows a TEM image of a representative few-layer graphene

442 NATURE NANOTECHNOLOGY | VOL 6 | JULY 2011 | www.nature.com/naturenanotechnology

© 2011 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2011.94 ARTICLES
a b 10,000
Exfoliated bi- and trilayer graphene

1,000

Hole mobility (cm2 V–1 s–1)

CVD bilayer
graphene
TG 100
Our current
10 work

1
Reduced
GO
D S 0.1

3 m 0.01
10 100 1,000 10,000 100,000
20 m 2D resistivity ( )

c 130 d 4 1.4
VTG = −4 V
3
120 VTG = −2 V
VTG = 0 V 2

2D resistivity (k )
110
VTG = 2 V
1
VTG = 4 V
IDS( A)

100

VTG(V)
0
90
−1
80 −2
70 −3
VDS = 0.1 V
60 −4 0.8
−60 −40 −20 0 20 40 60 −60 −40 −20 0 20 40 60
VBG(V) VBG(V)
e 75 f 1.5 2.0
VTG = 1.5 V
VTG = 1.0 V
70 1.0
VTG = 0.5 V
VTG = 0 V

2D resistivity (k )
65 VTG = −0.5 V 0.5
VTG = −1.0 V
IDS( A)

VTG(V)

60 0.0

55 −0.5

50 −1.0
VDS = 0.1 V
45 −1.5 1.5
−60 −40 −20 0 20 40 60 −60 −40 −20 0 20 40 60
VBG(V) VBG(V)

Figure 5 | Electronic characteristics of bilayer and ∼3–4-layer graphene devices in the presence of a perpendicular electric field at room temperature.
a, Low-magnification image of our fabricated top-gate (TG) device on a specific graphene flake. b, 2D resistivities and hole mobilities for several unannealed
FET devices using our bilayer- or trilayer-enriched graphene dispersions. Reported values corresponding to micromechanically exfoliated bi- and trilayer
graphene52, CVD bilayer graphene10 and reduced graphene oxide18,51,53,54 are also shown for comparison. c,d, Transfer currents (c) and 2D resistivity
(d) as functions of top-gate voltage (VTG) and bottom-gate voltage (VBG) of a representative 3–4-layer graphene device. e,f, Transfer currents (e) and
2D resistivity (f) as functions of VTG and VBG of a representative bilayer graphene device.

with an area of 50 mm2. Electron diffraction patterns in that we successfully produced layer-controlled, AB-stacked graphene
Supplementary Fig. S4 show typical Bragg reflection intensity dispersions. For comparison, we have indicated the regions occupied
ratios24 for monolayer (Supplementary Fig. S4a) and few-layer by graphene solutions produced from direct processing of natural
(Supplementary Fig. S4b) graphene. Additional TEM images of large graphite24,27. Another important feature of the distributions produced
few-layer graphene (.10 mm2) are shown in Supplementary Fig. S5. in this work is that the flake area increases as the layer number
An HRTEM image of the edge of a monolayer flake is shown in increases, with few-layer graphene flakes being an order of magnitude
Supplementary Fig. S6. Atomic force microscopy (AFM) images of larger than monolayers. It is clear that the preponderance of bilayer or
the isolated bilayer and 3–4-layer graphene flakes are shown in trilayer graphene flakes in solution corresponds to Stage-2 or Stage-3
Fig. 3e,f, respectively. The number of stacked layers was confirmed superlattice sub-domains created during the intercalation process.
by in situ Raman spectroscopy (Supplementary Figs S9 and S10). Relatively thick flakes are generated either from higher-stage sub-
Systematic statistical analysis of size and thickness distributions for domains in GICs or due to the recombination between crystallites
graphene flakes inside concentric rings was carried out by a combi- during post-expansion and homogenization processes. Alternatively,
nation of optical microscopy and Raman spectroscopy. Figure 4a,b small monolayer flakes appear to be exfoliated from larger bilayer
shows the distributions of graphene flakes from IBr Stage-3 and and trilayer flakes because ultrasonication was applied. We believe
ICl Stage-2 GICs, respectively. Distributions with modes centred at that further optimization of the intercalation process towards 100%
bilayer flakes (40%) and 3–4-layer flakes (35%) were obtained for Stage-2 or Stage-3 GIC would further narrow the layer size distribution
Stage-2 and Stage-3 expanded graphite, respectively. This indicates around the desired number of layers. It is also noteworthy that the

NATURE NANOTECHNOLOGY | VOL 6 | JULY 2011 | www.nature.com/naturenanotechnology 443

© 2011 Macmillan Publishers Limited. All rights reserved.

ARTICLES NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2011.94

method reported here may also be extended to monolayer graphene monolayer graphene oxide (0.01–12 cm2 V21 s21)18,51,53,54, as
exfoliation by carrying out Stage-1 intercalation with ICl49. shown in Fig. 5b. A number of factors limit our mobility. The extra-
The kinetics of the exfoliation and dispersion of graphite ordinary mobility of monolayer graphene has only been observed on
materials in the liquid phase using the ultrasonication process is suspended graphene samples at low temperatures1, and supported
analogous to polymer decomposition50. Consider a solution con- graphene has limited mobility due to the scattering effects of the
taining graphite (graphene) materials, Pi , which has an i-layer underlying substrate that are still under investigation56. In addition,
(1 ≤ i ≤ m, where m is the maximum layer number considered) scattering and doping from adsorbed impurities (e.g. oxygen57, sur-
structure. For all possible layer numbers i, the generation rate of factant and photoresist) on graphene surfaces could reduce mobi-
the i-layer graphene in the sonication process can be written as lity58. We also found that the appearance of the Dirac point and
the mobilities of graphene FET devices strongly depend on the
d[Pi ] m 
i−1
photoresist lift-off process. For some of our fabricated devices, we
= ki,j−i [Pj ] − kj,i−j [Pi ] (1) did not observe the Dirac point, and the electronic characteristics
dt j=i+1 j=1
showed unipolar behaviour (Supplementary Fig. S12). Figure 5d
presents a two-dimensional contour plot of 2D resistivities of a
where [Pi] denotes the concentration of Pi. Details of the kinetic trilayer graphene device as a function of top-gate voltage (VTG)
model and numerical solutions can be found in Supplementary and bottom-gate voltage (VBG). It is encouraging that the resistivity
Section S7. Using natural graphite as the raw material, the ‘exfolia- decreases with increasing perpendicular electric field, as observed in
tion rate constant’ can be considered independent of the layer ABA-stacked trilayer graphene4. The unique electronic properties
number (that is, ki,j ¼ k). In this limit, the layer distribution of exfo- suggest that the graphene flake is ABA-trilayer. On the other
liated products remains broad13, with increasing sonication time hand, for our bilayer device (Fig. 5f ), the Dirac point moves along
biasing the distribution towards monolayer27 instead of terminating the diagonal, and the 2D resistivities reach a maximum at the
at any specific layer number. The simulated concentrations as a upper-left and lower-right corners. This behaviour is a clear signa-
function of time (Fig. 4d) show that the bilayer or trilayer graphene ture of bilayer graphene55.
will never dominate. On the other hand, considering the Stage-3
and Stage-2 starting materials, it follows that the exfoliation Conclusions
should be easier for every three and two layers in the materials, We have illustrated an advanced concept for the layer-controlled
respectively. We simulate this by increasing the rate constants for production of pristine large-area graphene dispersions. The use of
every three (Stage-3) or two (Stage-2) layers by a factor of 10, say; ionic graphite intercalation compounds to produce Stage-2 and
the simulated concentrations as a function of dimensionless time Stage-3 GICs using ICl and IBr as the ionic intercalants have been
are shown in Fig. 4e,f, respectively. In this scheme, the k8,1 rate con- shown to be excellent precursors for the production of bilayer and
stant is assigned to a 9-layer graphite exfoliated into an 8- and trilayer graphene dispersions. When combined with an on-chip sep-
1-layer for a Stage-2 expanded graphite shown in Fig. 4c and aration method, a population of large-area graphene flakes is pro-
Supplementary Section S7. Note that the model predicts that trilayer duced, such that conventional photolithography is enabled for
(Fig. 4e) and bilayer (Fig. 4f ) graphene are both enriched with mild top-gate device fabrication. Our present approach comprises the
sonication conditions (short dimensionless time kt , 0.5). At the only viable route (at this time) to producing AB-stacked bi- and
specific dimensionless time, 0.055kt, the calculated trilayer (35%, trilayer graphene on arbitrary substrates on a large scale, and the
Fig. 4e right) and bilayer (40%, Fig. 4f right) graphene concen- measured mobility values for the devices produced in the present
trations are quantitatively consistent with experimentally observed work are the highest for any solution-dispersed material to date.
values using Stage-3 and Stage-2 GICs as the raw materials, respect- The changes in 2D resistivities of our trilayer and bilayer devices
ively. Therefore, by introducing proper graphite materials with a as functions of VTG and VBG exhibit the same characteristics as
selective laminar structure around a specific layer, as well as mild observed for the case of micromechanically exfoliated graphene.
sonications, the produced solution should be able to reflect the
original composition of crystallites directly and form layer- Received 15 April 2011; accepted 19 May 2011;
controlled dispersions. published online 26 June 2011

Electronic properties of bi- and trilayer graphene flakes References

The large flakes isolated in this work enabled the fabrication of field- 1. Geim, A. K. & Novoselov, K. S. The rise of graphene. Nature Mater. 6,
effect transistor (FET) devices using conventional ultraviolet photo- 183–191 (2007).
2. Ohta, T. et al. Controlling the electronic structure of bilayer graphene. Science
lithography. Several top-gate devices were fabricated (details on 313, 951–954 (2006).
device fabrication can be found in Supplementary Section S6). 3. Zhang, Y. B. et al. Direct observation of a widely tunable bandgap in bilayer
Figure 5a shows a top-view image of the graphene device. graphene. Nature 459, 820–823 (2009).
Figure 5b–f presents the electronic characterization data of represen- 4. Craciun, M. F. et al. Trilayer graphene is a semimetal with a gate-tunable band
tative bilayer and trilayer devices in the presence of a perpendicular overlap. Nature Nanotech. 4, 383–388 (2009).
5. Zhou, S. Y. et al. Substrate-induced bandgap opening in epitaxial graphene.
electric field. All measurements were carried out under ambient Nature Mater. 6, 916–916 (2007).
conditions. The 2D resistivities (the resistance of each device nor- 6. Novoselov, K. S. et al. Two-dimensional atomic crystals. Proc. Natl Acad. Sci.
malized by its channel length and width) of the graphene flakes at USA 102, 10451–10453 (2005).
zero gate bias are about 1 kV, which is one order of magnitude 7. Berger, C. et al. Ultrathin epitaxial graphite: 2D electron gas properties and a route
lower than the one corresponding to a film of pristine graphene dis- toward graphene-based nanoelectronics. J. Phys. Chem. B 108, 19912–19916 (2004).
8. Sutter, P. W., Flege, J. I. & Sutter, E. A. Epitaxial graphene on ruthenium. Nature
persions24, and three orders of magnitude lower than the one corre- Mater. 7, 406–411 (2008).
sponding to reduced graphene oxide18,51. The transfer characteristics 9. Reina, A. et al. Large area, few-layer graphene films on arbitrary substrates by
show an ambipolar behaviour, and the calculated hole mobility is chemical vapor deposition. Nano Lett. 9, 30–35 (2009).
400 cm2 V21 s21. This value is 3–7 times lower than that 10. Lee, S., Lee, K. & Zhong, Z. H. Wafer scale homogeneous bilayer graphene films
from micromechanically cleaved bilayer and few-layer graphene by chemical vapor deposition. Nano Lett. 10, 4702–4707 (2010).
11. Castro Neto, A. H., Guinea, F., Peres, N. M. R., Novoselov, K. S. & Geim, A. K.
(1,500–3000 cm2 V21 s21)52 on a SiO2 substrate, a little bit The electronic properties of graphene. Rev. Mod. Phys. 81, 109–162 (2009).
lower than CVD deposited bilayer graphene (580 cm2 V21 s21)10, 12. dos Santos, J. M. B. L., Peres, N. M. R. & Castro, A. H. Graphene bilayer with a
but 1–5 orders of magnitude higher than that from reduced twist: electronic structure. Phys. Rev. Lett. 99, 256802 (2007).

444 NATURE NANOTECHNOLOGY | VOL 6 | JULY 2011 | www.nature.com/naturenanotechnology

© 2011 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2011.94 ARTICLES
13. Coleman, J. N. Liquid-phase exfoliation of nanotubes and graphene. Adv. Funct. nanoparticles using hydrodynamic forces in cylindrical droplets. Nano Lett. 7,
Mater. 19, 3680–3695 (2009). 2693–2700 (2007).
14. Park, S. & Ruoff, R. S. Chemical methods for the production of graphenes. 44. Sharma, R. & Strano, M. S. Centerline placement and alignment of anisotropic
Nature Nanotech. 4, 217–224 (2009). nanotubes in high aspect ratio cylindrical droplets of nanometer diameter. Adv.
15. Niyogi, S. et al. Solution properties of graphite and graphene. J. Am. Chem. Soc. Mater. 21, 60–65 (2009).
128, 7720–7721 (2006). 45. Shen, X. Y., Ho, C. M. & Wong, T. S. Minimal size of coffee ring structure.
16. Stankovich, S. et al. Synthesis of graphene-based nanosheets via chemical J. Phys. Chem. B 114, 5269–5274 (2010).
reduction of exfoliated graphite oxide. Carbon 45, 1558–1565 (2007). 46. Sangani, A. S., Lu, C. H., Su, K. H. & Schwarz, J. A. Capillary force on particles
17. Li, D., Muller, M. B., Gilje, S., Kaner, R. B. & Wallace, G. G. Processable aqueous near a drop edge resting on a substrate and a criterion for contact line pinning.
dispersions of graphene nanosheets. Nature Nanotech. 3, 101–105 (2008). Phys. Rev. E 80, 011603 (2009).
18. Tung, V. C., Allen, M. J., Yang, Y. & Kaner, R. B. High-throughput solution 47. Ferrari, A. C. et al. Raman spectrum of graphene and graphene layers. Phys. Rev.
processing of large-scale graphene. Nature Nanotech. 4, 25–29 (2009). Lett. 97, 187401 (2006).
19. Wang, S. et al. High mobility, printable, and solution-processed graphene 48. De Marco, P. et al. Rapid identification of graphene flakes: alumina does it better.
electronics. Nano Lett. 10, 92–98 (2010). Nanotechology 21, 255703 (2010).
20. Gomez-Navarro, C. et al. Atomic structure of reduced graphene oxide. Nano 49. Ghosh, D., Gangwar, R. & Chung, D. D. L. Superlattice ordering in graphite-IC1
Lett. 10, 1144–1148 (2010). single-crystals and fibers. Carbon 22, 325–333 (1984).
21. Bagri, A. et al. Structural evolution during the reduction of chemically derived 50. Dotson, N. A. Polymerization Process Modeling, xvi, 371 p. (VCH, 1996).
graphene oxide. Nature Chem. 2, 581–587 (2010). 51. Eda, G., Fanchini, G. & Chhowalla, M. Large-area ultrathin films of reduced
22. Becerril, H. A. et al. Evaluation of solution-processed reduced graphene oxide graphene oxide as a transparent and flexible electronic material. Nature
films as transparent conductors. ACS Nano 2, 463–470 (2008). Nanotech. 3, 270–274 (2008).
23. Blake, P. et al. Graphene-based liquid crystal device. Nano Lett. 8, 1704–1708 52. Nagashio, K., Nishimura, T., Kita, K. & Toriumi, A. Mobility
(2008). variations in mono- and multi-layer graphene films. Appl. Phys. Exp. 2,
24. Hernandez, Y. et al. High-yield production of graphene by liquid-phase 025003 (2009).
exfoliation of graphite. Nature Nanotech. 3, 563–568 (2008). 53. Su, C. Y. et al. Electrical and spectroscopic characterizations of ultra-large
25. Shih, C. J., Lin, S., Strano, M. S. & Blankschtein, D. Understanding the reduced graphene oxide monolayers. Chem. Mater. 21, 5674–5680 (2009).
stabilization of liquid-phase-exfoliated graphene in polar solvents: molecular 54. Gomez-Navarro, C. et al. Electronic transport properties of individual
dynamics simulations and kinetic theory of colloid aggregation. J. Am. Chem. chemically reduced graphene oxide sheets. Nano Lett. 7, 3499–3503 (2009).
Soc. 132, 14638–14648 (2010). 55. Wang, F. et al. Direct observation of a widely tunable bandgap in bilayer
26. Lotya, M. et al. Liquid phase production of graphene by exfoliation of graphite graphene. Nature 459, 820–823 (2009).
in surfactant/water solutions. J. Am. Chem. Soc. 131, 3611–3620 (2009). 56. Bruna, M., Vaira, A., Battiato, A., Vittone, E. & Borini, S. Graphene strain
27. Green, A. A. & Hersam, M. C. Solution phase production of graphene with tuning by control of the substrate surface chemistry. Appl. Phys. Lett. 97, 021911
controlled thickness via density differentiation. Nano Lett. 9, 4031–4036 (2009). (2010).
28. Green, A. A. & Hersam, M. C. Emerging methods for producing monodisperse 57. Gunlycke, D., Li, J., Mintmire, J. W. & White, C. T. Altering low-bias transport in
graphene dispersions. J. Phys. Chem. Lett. 1, 544–549 (2010). zigzag-edge graphene nanostrips with edge chemistry. Appl. Phys. Lett. 91,
29. Dresselhaus, M. S. & Dresselhaus, G. Intercalation compounds of graphite. Adv. 112108 (2007).
Phys. 30, 139–326 (1981). 58. Adam, S., Hwang, E. H., Galitski, V. M. & Das Sarma, S. A self-consistent theory
30. Chung, D. D. L. Graphite. J. Mater. Sci. 37, 1475–1489 (2002). for graphene transport. Proc. Natl Acad. Sci. USA 104, 18392–18397 (2007).
31. Li, X. L. et al. Highly conducting graphene sheets and Langmuir–Blodgett films.
Nature Nanotech. 3, 538–542 (2008). Acknowledgements
32. Ang, P. K., Wang, S. A., Bao, Q. L., Thong, J. T. L. & Loh, K. P. High-throughput The authors acknowledge fundings from the 2009 US Office of Naval Research Multi
synthesis of graphene by intercalation — exfoliation of graphite oxide and study University Research Initiative (MURI) on Graphene Advanced Terahertz Engineering
of ionic screening in graphene transistor. ACS Nano 3, 3587–3594 (2009). (GATE) at MIT, Harvard and Boston University. M.S.S. is also grateful for a 2008 Young
33. Lee, J. H. et al. One-step exfoliation synthesis of easily soluble graphite and Investigator Program Award (YIP) from the US Office of Naval Research. D.B. and
transparent conducting graphene sheets. Adv. Mater. 21, 4383–4387 (2009). S.L. are grateful for financial support from the DuPont/MIT Alliance. C.J.S. is grateful for
34. Li, X. L., Wang, X. R., Zhang, L., Lee, S. W. & Dai, H. J. Chemically derived, partial financial support from the David H. Koch Fellowship. The authors acknowledge
ultrasmooth graphene nanoribbon semiconductors. Science 319, support from the Institute for Soldier Nanotechnologies at MIT, funded by a grant from
1229–1232 (2008). the Army Research Office. XRD and TEM analyses were performed at the MIT Center
35. Behabtu, N. et al. Spontaneous high-concentration dispersions and liquid of Materials Science and Engineering (CMSE), supported by S. Speakman and Y. Zhang.
crystals of graphene. Nature Nanotech. 5, 406–411 (2010). Fruitful discussions with D.D.L Chung, J. Kong, A. Hsu and J.H. Kim are also
36. Zheng, J. A. et al. High quality graphene with large flakes exfoliated by oleyl gratefully acknowledged.
amine. Chem. Commun. 46, 5728–5730 (2010).
37. Lee, J. H. et al. The superior dispersion of easily soluble graphite. Small 6,
58–62 (2010). Author contributions
38. Loh, K. P., Bao, Q. L., Ang, P. K. & Yang, J. X. The chemistry of graphene. C.J.S, A.V., R.K. and M.S.S. conceived and designed the dispersion experiments. C.J.S. and
J. Mater. Chem. 20, 2277–2289 (2010). R.K. implemented the dispersion method. C.J.S., A.V., R.S., G.P. and M.H.H. fabricated
39. Sasa, T., Takahash, Y. & Mukaibo, T. Crystal structure of graphite bromine FET devices. C.J.S. and M.S.S. developed the mathematical model. A.V., R.S., Z.J. and J.H.H.
lamellar compounds. Carbon 9, 407–416 (1971). performed the TEM and AFM analysis. C.J.S. performed the XRD and Raman analysis.
40. Daumas, N. & Herold, A. Relations between phase concept and reaction Q.H.W., S.L. and N.F.R. did additional experiments for revising the manuscript. C.J.S., D.B.
mechanics in graphite insertion compounds. Comptes Rendus Hebdomadaires and M.S.S. wrote the manuscript with input from A.V. All authors discussed the results and
Des Seances De L Academie Des Sciences Serie C 268, 373–382 (1969). commented on the manuscript.
41. Nikitin, Y. A. & Pyatkovskii, M. L. Formation and properties of materials based
on thermally expanded graphite. Powder Metall. Met. Ceram. 36, 41–45 (1997). Additional information
42. Deegan, R. D. et al. Capillary flow as the cause of ring stains from dried liquid The authors declare no competing financial interests. Supplementary information
drops. Nature 389, 827–829 (1997). accompanies this paper at www.nature.com/naturenanotechnology. Reprints and
43. Sharma, R., Lee, C. Y., Choi, J. H., Chen, K. & Strano, M. S. Nanometer permission information is available online at http://www.nature.com/reprints/.
positioning, parallel alignment, and placement of single anisotropic Correspondence and requests for materials should be addressed to M.S.S.

NATURE NANOTECHNOLOGY | VOL 6 | JULY 2011 | www.nature.com/naturenanotechnology 445

© 2011 Macmillan Publishers Limited. All rights reserved.

 ARTICLES
PUBLISHED ONLINE: 20 MARCH 2011 | DOI: 10.1038/NNANO.2011.30

Transforming C60 molecules into graphene

quantum dots
Jiong Lu1, Pei Shan Emmeline Yeo1,2, Chee Kwan Gan2, Ping Wu2 and Kian Ping Loh1 *

The fragmentation of fullerenes using ions, surface collisions or thermal effects is a complex process that typically leads
to the formation of small carbon clusters of variable size. Here, we show that geometrically well-defined graphene
quantum dots can be synthesized on a ruthenium surface using C60 molecules as a precursor. Scanning tunnelling
microscopy imaging, supported by density functional theory calculations, suggests that the structures are formed through
the ruthenium-catalysed cage-opening of C60. In this process, the strong C60–Ru interaction induces the formation of
surface vacancies in the Ru single crystal and a subsequent embedding of C60 molecules in the surface. The fragmentation
of the embedded molecules at elevated temperatures then produces carbon clusters that undergo diffusion and
aggregation to form graphene quantum dots. The equilibrium shape of the graphene can be tailored by optimizing the
annealing temperature and the density of the carbon clusters.

G
raphene is of great interest because of its unique electronic, may combine rapidly on the surface to form large graphitic sheets,
thermal and mechanical properties, and it shows promise or may be confined on surface defects to form featureless clusters.
as a material for use in nanoelectronics1–7. Graphene field- A narrow temperature or coverage window may exist where the
effect transistors can be scaled to shorter channel lengths and interparticle diffusion length favours the growth of GQDs with geo-
higher speeds without encountering the adverse short channel metric uniformity. Here, we report a mechanistic approach to the
effects that restrict the performance of existing devices8. Although synthesis of a series of atomically defined GQDs by the metal-catalysed
the vertical scaling of the device can be pushed to the ultimate cage-opening of C60.
limit of a single atomic layer in graphene, the problem of lateral
scaling is not trivial and requires rather involved top-down lithogra- Geometrically well-defined GQDs on Ru(0001)
phical approaches. Currently, top-down lithographical, solution- It is known that a graphitic carbon layer can be derived from the
phase chemical methods and surface-assisted coupling and cyclode- surface-catalysed decomposition of C60 adlayers on reactive tran-
hydrogenation of linear polyphenylenes have been applied to sition metals. Supplementary Fig. S1 illustrates the transformation
produce graphene nanoribbons (GNRs)4,7,9 and graphene of a high coverage of C60 (Q . 0.7 monolayer (ML)) to a single
quantum dot (GQD)-based materials4,6,7,10,11. Bandgap engineering layer of graphene covering a Ru(0001) surface after annealing the
can be implemented in GNRs and GQDs due to quantum confine- sample at 1,200 K for 5 min. Within a coverage range of
ment6,11 and edge effects9, which is useful for realizing the potential 0.2 ML , Q , 0.7 ML C60 , the short diffusion distance between
of graphene as a transistor. To date, the majority of solution-phase the fragmented molecules favours the aggregation of these carbon
methods produce irregularly sized and shaped graphene sheets due clusters, resulting in the growth of larger-sized and irregularly
to the intrinsic randomness of the defect-mediated exfoliation or shaped graphene nano-islands (Supplementary Fig. S2). To grow
cutting process of the precursor graphitic flakes. To produce GQDs, the interparticle diffusion length must be sufficiently long
highly regular graphene nanostructures, a fabrication process that (mean distance between C60 molecules, 15+3 nm) or the interpar-
is driven by thermodynamics, as in crystal growth, should be ticle diffusion velocity must be sufficiently low to limit diffusional
more suitable than defect-mediated fragmentation processes. aggregation of the decomposed C60 fragments. By controlling
In principle, graphene and fullerenes are interconvertible. these factors, GQDs with well-defined geometrical shapes can be
Inspired by the surface-catalysed cyclodehydrogenation of polyaro- assembled from the carbon clusters derived from C60. Figure 1
matic precursors to form fullerenes and the direct transformation of shows the temperature-dependent generation of GQDs following
graphene to fullerenes12,13, we explored the possibility that atomic- the annealing of the Ru surface with a low coverage of 0.08 ML
ally precise GQDs could be generated from the metal-catalysed C60. Triangular GQDs with a 15+1% yield (apparent lateral size,
cage-opening of C60. The idea is simple: every C60 molecule has 2.7 nm) were produced following 1 min annealing at 725 K
the same size and shape, so the fragmentation of fullerene, if (Fig. 1a,b). We also observed parallelogram-shaped (2.7 × 4.2 nm)
carried out under well-controlled conditions, should produce uni- (Fig. 1c), trapezoid-shaped (2.7 × 4.8 nm) (Fig. 1d) and hexagon-
formly sized GQDs. Carbon nanotubes, onions and graphitic shaped (10 nm) (Fig. 1f ) GQDs, but these have lower yields com-
domains have been generated from the optical heating of functiona- pared to the smaller hexagonal mushroom-shaped dots (0.9 nm)
lized fullerenes14 and the decomposition of fullerenes on metal sur- on the surface (Fig. 1a, inset). Further annealing of the sample at
faces15,16. However, to date, the opening of the fullerene cage to form 825 K for 1 min produced 5 nm, perfectly hexagonal GQDs,
GQDs has not been reported. Moreover, the fragmentation mechan- which have a 30+2% yield (Fig. 1e), while the small-sized
ism of C60 on metal surfaces is not well understood. Depending on (,5 nm) carbon clusters and GQDs disappear. Although some
the diffusion rate of the C60 fragments on a surface, these fragments GQDs synthesized here do contain point defects that affect the

1
Department of Chemistry, National University of Singapore, 3 Science Drive 3, Singapore 117543, 2 Institute of High Performance Computing, 1 Fusionopolis
Way, #16-16 Connexis, Singapore 138632. * e-mail: chmlohkp@nus.edu.sg

NATURE NANOTECHNOLOGY | VOL 6 | APRIL 2011 | www.nature.com/naturenanotechnology 247

© 2011 Macmillan Publishers Limited. All rights reserved.

ARTICLES NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2011.30

a b c g

1.2

Z (nm)
ED
2.5 Å

0 1 2 3 V

dI⁄dV (a.u.)
X (nm)
10 nm 1 nm 2 nm IV

d e f
III

II

−1.5 −1.0 −0.5 0.0 0.5 1.0 1.5

2 nm 20 nm 3 nm Vbias (V)

Figure 1 | STM images of GQDs formed by decomposition of 0.08 ML C60 on Ru(0001). a, A 0.08 ML C60/Ru sample after annealing at 725 K for 2 min.
Inset: magnified view of mushroom-shaped dots. b–d, Magnified views of triangular (2.7 nm, b), parallelogram-shaped (2.7 × 4.2 nm, c), trapezoid-shaped
(2.7 × 4.8 nm, d) GQDs. Inset to c: line contour taken along the green line in b. e,f, Hexagon-shaped GQDs (5 nm and 10 nm) obtained after further
annealing the sample at 825 K for 2 min. g, Representative local STS data for differential conductances dI/dV of the GQDs in b (I), c (II), e (III), f (IV) and
for giant monolayer graphene on Ru(0001) (V). Tunnelling parameters: V ¼ 0.5 V, I ¼ 0.1 nA; V ¼ 0.3 V, I ¼ 0.2 nA for the inset images in a and e; V ¼ 0.3 V,
I ¼ 0.2 nA (b,c); V ¼ 0.3 V, I ¼ 0.1 nA (d,f).

periodic moiré pattern of graphene on the periphery (Fig. 1e and indicated in the dI/dV spectra (Fig. 1), the triangular-shaped, paral-
Fig. 4h), the majority of the GQDs show well-defined geometries. lelogram-shaped and hexagon-shaped (with lateral dimension 5 nm
It is expected that size confinement and the associated edge and 10 nm) GQDs exhibit an energy gap of 0.8 eV, 0.6 eV, 0.4 eV and
effects of these two-dimensional nanostructures will give rise to 0.25 eV, respectively (see Supplementary Information for more
electronic properties that are different from those of bulk materials. details). The bandgap of GQDs increases with a reduction in
Atomically resolved topography reveals that the 2.7 nm GQDs with their lateral size, which is analogous to bandgap widening in
zigzag edges have 2.5 Å triangular lattice structures (‘3-for-6’) (inset GNRs with a narrowing of ribbon width22. The experimental gap
of Fig. 1c) in both the moiré valley regions (centre area in Fig. 1b) (Eg) versus size (L) relation for GQDs follows the relationship
and maxima regions (circled area of Fig. 1b), an arrangement that Eg(eV) ¼ 1.77+0.12 eV nm/L 0.9+0.1 using a least-squares fit
is unlike that of the large-sized graphene sheets, where the hexago- (Supplementary Fig. S4), which is a close match to the predicted
nal honeycomb lattice pattern is present for the moiré maxima scaling trend of Eg (eV) ¼ 1.68 eV nm/L from quantum confine-
regions only17–19 (Supplementary Fig. S1). The lattice mismatch ment1,23. The slight deviation can be explained by the interaction
between graphene and substrate gives rise to the characteristic between the GQDs and the Ru(0001) substrate. As the size of the
moiré-type pattern. The vertical buckling of the graphene in the GQDs increases, the discrete energy levels centred at 0.5 eV and
scanning tunnelling microscopy (STM) data18 gives rise to a 0.8 Å 0.9 eV for the 2.7 nm GQDs gradually vanish, and a typical dip at
difference in the apparent height between the bright parts and the –0.4 eV for epitaxial graphene appears when the size of the GQDs
dark region under a sample bias of 0.3 V. The corrugation reaches 5 nm, which corresponds to the charge neutral Dirac point
between bright and dark regions is also dependent on the sample of graphene (ED). At this energy, the linear p bands intersect at a
bias applied, which means the electronic effects in this system single point in k-space so that charge carriers vanish19,24,25. For
result in stronger corrugation than the actual geometric roughness GQDs larger than 10 nm, a pronounced dip appears at –0.35 eV
of the graphene layer. The carbon atoms in the valley region interact and the tunnelling conductance increases sharply with energy for
strongly with the Ru substrate, so the sublattice symmetry of gra- both filled and empty states. A negative shift of the Dirac point by
phene is broken. Only the atoms of one particular sublattice are 0.35 eV is consistent with a substantial electron doping of the gra-
therefore imaged by STM. In the area where the graphene is phene adlayer by the Ru substrate.
buckled and not interacting with the substrate (circled area in
Fig. 1b), triangular GQDs exhibit a ‘3-for-6’ image. Several factors Ruthenium-catalysed cage-opening of C60
could give rise to this, such as the presence of topological frustra- How do the fullerene cages open and transform into GQDs? A com-
tion20, which results in the appearance of spin states near the plete fragmentation of a monolayer of C60 film would produce five
Fermi level that are localized on one particular sublattice of the graphene layers if all carbon atoms remain on the surface. However,
GQDs (Supplementary Fig. S3), or arising from an edge-induced for a Ru surface adsorbed with monolayer to multilayer C60 films,
interference pattern21. only single-layer graphene was observed on the Ru(0001) surface
after annealing at 1,200 K for 5 min (Supplementary Fig. S1).
Measuring the bandgap of the GQDs Therefore, it can be inferred that there is substantial gasification
The correlation between the electronic properties of GQDs with their of the decomposed carbon fragments. To probe the decomposition
atomic structure and lateral size was investigated using spatially process of the fullerene and to examine how the surface-retained
resolved scanning tunnelling microscopy (STS). The effect of fragments can transform into GQDs, we tracked the evolution of
zigzag and armchair edges on the electronic properties can be a lone C60 molecule on Ru(0001) (Fig. 2a). When the substrate
excluded, because the GQDs here have uniform zigzag edges. The covered with 0.03 ML C60 was gently heated to 450 K for 10 min,
size-dependent bandgap of the GQDs is shown in Fig. 1g. As the C60 transformed into structures with threefold rotational

248 NATURE NANOTECHNOLOGY | VOL 6 | APRIL 2011 | www.nature.com/naturenanotechnology

© 2011 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2011.30 ARTICLES
a b 9 725 K c I II
6

Z (Å)
3
0
0 1 2 3 4
X (nm)
III IV

10 nm 3 nm

d 725 K e 825 K f
3 nm 2 nm
70
60

Fraction (%)
50
40
30
82 20
5K
10
0.6 72
5K
Do 0.81.0 0
ts
siz 1.2
e ( 1.4
nm 1.6
)

Figure 2 | STM images of the C60-derived clusters after annealing a 0.03 ML film of C60 on Ru(0001). a, 0.03 ML C60 on Ru (0001). Inset: magnified
view of one C60 molecule after annealing the sample at 450 K. b,d,e, Bright dots observed after flash annealing at 725 K for 2 min: mushroom-shaped dots
(b) and flower-shaped dot (d); and three flower-shaped dots obtained at 825 K (e). Insets (b,d,e): magnified views of these dots. Top inset of b: line contour
of an individual C60 molecule (green curve) and bright dots in b (white curve). c, Surface diffusion and combination of C60-derived fragments (STM image
size, 4.5 × 4.2 nm). f, Histogram of dot size distribution at different annealing temperatures: 725 K for 2 min and 825 K for 1 min. Tunnelling parameters:
V ¼ 1.2 V, I ¼ 0.08 nA, V ¼ 100 mV, I ¼ 0.3 nA for the inset image in a; V ¼ 0.3 V, I ¼ 0.2 nA (b); V ¼ 1.2 V, I ¼ 0.1 nA (c); V ¼ 0.3 V, I ¼ 0.25 nA (d);
V ¼ 0.3 V, I ¼ 0.16 nA (e).

symmetry (Fig. 2a, inset). One interpretation is that the C60 mol- video; Supplementary Movie 1). Annealing the sample to 500–550 K
ecule is lying with one six-membered ring parallel to the initiates thermal hopping of the C60 molecules and dissociation of
Ru(0001) substrate, as has been observed for C60 molecules on C60 clusters on the terrace. In this temperature range, the diffusion
Pt(110) and Cu(111) surfaces26,27. After flash annealing for 2 min barrier of the isolated C60 molecule is overcome and it moves to the
at 725 K, numerous bright spots that are distinctly different from on-top adsorption sites on Ru(0001) (Supplementary Table 1).
the starting C60 were found on the surface (Fig. 2b,d). The apparent Simultaneously, the Ru atom under the C60 hops out to form a
height for these bright spots is 2.7 Å, representing a decrease of 60% vacancy, allowing the C60 to sink lower into the surface (with a
from the initial height of 6.5 Å for C60 under a similar bias voltage. decreased apparent height of 0.5+0.1 Å; Supplementary Fig. S6).
The lateral size is also significantly reduced (Fig. 2b, inset). Such an adatom-vacancy mechanism has been observed by Felici
Following a detailed investigation, three distinct structures were and colleagues for C60 on Pt(111)29 and is predicted by Li and
observed at different annealing temperatures (Fig. 2b,d,e). colleagues to occur for C60 on Ag(111) surface30. At the higher
Annealing to 725 K produced flower-shaped dots with a threefold annealing temperature of 650 K, embedded C60 molecule decomposed
symmetry (0.7 nm) (Fig. 2d), which constitutes 23% of the to form carbon clusters, as shown in Fig. 3b.
decomposition products (Fig. 2f ). Hexagonal mushroom-shaped To obtain an atomistic insight into the decomposition mechan-
dots with a lateral width of 0.9 nm were also observed with a 67% ism of C60 on the Ru surface, the adsorption of a C60 molecule on a
yield (Fig. 2b). After annealing at 825 K, the smaller flower- five-layer Ru(0001) supercell slab was simulated using density func-
shaped dots appear to have merged into larger ones (1.2 nm) tional theory (DFT). The model considers the adsorption of C60 on
(Fig. 2e). The magnified view of a 1.2 nm dot (Fig. 2e, inset) a Ru vacancy site (on-top_vac model) and a non-vacancy site (on-
shows that it originates from the combination of three flower- top model) (Supplementary Table 1). The configuration in which
shaped dots, which are joined at a central protrusion spot. the adatom-vacancy mechanism operates is the most energetically
Magnified STM images reveal a triangular lattice pattern for the favourable. The bond lengths between the C atoms of the bottom
C60-derived dots rather than the hexagonal-shaped honeycomb hexagon ring and the closest Ru atoms (grey–pink bonds in
lattice pattern of giant graphene. In addition, the atoms at the per- Fig. 3c) are, on average, 2.0% shorter in the on-top_vac than in
iphery of the clusters bond strongly to the substrate, resulting in an the on-top configuration. Likewise, the bond lengths between the
upward lifting (0.28+0.02 Å in the apparent height for the mush- second-tier carbon atoms of the C60 and the closest Ru atoms
room-shaped dots) of the centre region relative to the edge, (black–pink bonds in Fig. 3d) are on average 2.5% shorter in the
forming a dome-like shape28. The strong bonding of peripheral on-top_vac configuration. The stronger interaction of the lower
carbon atoms to the substrate suggests that the edges of the clusters hemisphere carbon atoms with the Ru substrate indicates the weak-
are hydrogen-free. ening of certain C–C bonds in the C60 molecule. Figure 3d shows the
How are the carbon clusters generated from the opening of the bond lengths of the bottom hemisphere of the optimized C60 of the
C60 cage? The STM images in Fig. 3 and Supplementary Fig. S6 on-top_vac configuration. The C–C long bonds, which are between
provide an insight into this process (the thermal diffusion and a hexagon and pentagon ring in C60 (labelled in red) in Fig. 3d, have
sinking of C60 molecules are also visualized in a STM-recorded been extended by 2.6% on average compared to the perfect C60

NATURE NANOTECHNOLOGY | VOL 6 | APRIL 2011 | www.nature.com/naturenanotechnology 249

© 2011 Macmillan Publishers Limited. All rights reserved.

ARTICLES NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2011.30

a b
e-C60 d-C60

i-C60

c d 1.440 1.441 1.442 1.439

1.399 1.400
1.429
1.453 1.442 1.443 1.453

1.386 1.386
1.455 1.455
1.480 1.412 1.479
1.452
1.452
1.479 1.481
1.441 1.438 1.456
1.455 1.400
1.399
1.442 1.415 1.413 1.441 1.440
1.440
1.442
1.429 1.441
1.442 1.454 1.454 1.429
1.440 1.444

1.441 1.480 1.479 1.443

1.440
1.440
1.400
1.399
1.452 1.453
1.443

Figure 3 | Three-dimensional STM images of a carbon cluster derived from the decomposition of embedded C60 molecules on Ru(0001), and the
simulated ‘on-top_vac’ configuration of a C60 molecule on Ru(0001). a, Constant-current image at 600 K (e-C60 , embedded C60; i-C60 , intact C60; image
size, 5 × 3 nm2). b, Constant current image at 650 K (d-C60 , decomposition of embedded C60; image size, 5 × 3 nm2). c, On-top_vac configuration of the
C60 molecule. The single-sided arrow indicates the top-down point of view, from which d is derived. d, C–C bond lengths of the bottom hemisphere of the
C60 in c. The top hemisphere is not shown for clarity.

molecule, and the C–C short bonds between two hexagon rings in C60 In the case of C60-derived clusters, owing to the limited mobility
(labelled in blue) have been lengthened by 3.0%. Although the of the clusters and purely surface-mediated growth, each aggrega-
calculations were performed at 0 K, we surmise that these lengthened tion event is limited to only a few clusters, leading to the formation
bonds constitute a fault line that can be ruptured by thermal energy. of GQDs. We cannot preclude that growth of the GQDs arises from
At higher temperatures, sufficient energy can be generated from the ‘one-by-one’ addition of carbon adatoms from carbon clusters
resonance of electron–phonon and phonon–molecular vibronic rather than the coalescence of whole clusters, due to the limited
coupling31 to rupture the fullerene cages into two unsymmetrical temporal resolution of STM.
hemispheres along the fault line. The surface-retained fragment A lower diffusion coefficient (in the range of 10215 to
derived from the bottom hemisphere of the ruptured C60 evolves 10 216
cm2 s21) for these C60-derived clusters enables our dynamic
eventually into the observed surface-stabilized flower-shaped and STM to capture the cluster movement at high temperature
mushroom-shaped clusters on the Ru surface, while the top (Fig. 2c and Supplementary Fig. S6). The GQDs derived are equili-
hemisphere of the C60 cage may desorb into the gas phase. brium structures with minimized edge free energy35, exhibiting
highly symmetrical threefold and sixfold structures that are stable
Peculiarity of a C60 precursor in the formation of GQDs to 1,000 K. (Fig. 5 and Supplementary Fig. S9). Some non-equili-
One question is whether the GQDs obtained here are unique to the brium structures such as parallelogram- and trapezoid-shaped
C60 precursor. At a fundamental level, the question is whether the GQDs, which were produced in much lower yield, were also
diffusional dynamics of the carbon clusters derived from C60 is observed. It is instructive to see that these non-equilibrium struc-
unique for the growth of GQDs, as opposed to growth mediated tures transform into equilibrium shapes by the migration and
by adatom carbon diffusion. To answer this question, we compare detachment of edge atoms during annealing (Fig. 5). We captured
the carbon nanostructures produced by C60 and C2H4 the dynamic reconstruction of the boundary of these structures
(a commonly used precursor in chemical vapour deposition) in using STM video (Supplementary Movie 2). The important stages
the initial stage of graphene growth. A pictorial summary compar- in the fragmentation processes, which follow the trapezoid  paral-
ing the growth of GQDs from C60-derivied carbon clusters and lelogram  triangle sequence, are captured and highlighted by the
large-sized graphene islands from C2H4 is shown in Fig. 4. brown circles in Fig. 5. Once the islands reach their equilibrium
First, at a low dosage of C2H4 (,1 Langmuir (L)), nucleation of shapes, the growth will cease and their shapes are retained on the
C adatoms to form carbon clusters was observed to occur primarily surface (Fig. 5 and Supplementary Fig. S9) to elevated temperatures.
at the step edges of the substrate32. It is well known that catalytic In conclusion, we have demonstrated the formation of regularly
dehydrogenation is more facile at these sites and the carbon sized graphene quantum dots on Ru(0001) substrate using C60 ,
dimer is stabilized by the step geometry33. In contrast, the majority which acts as a unique precursor compared to C2H4. STM
of C60 molecules were observed to adsorb on the terraces (Fig. 2a). imaging provides direct evidence of the Ru-catalysed cage-
Second, the much higher mobility of carbon adatoms derived from opening of C60 and the assembly of its fragments into surface-stabil-
hydrocarbon decomposition, as well as the tendency for smaller ized carbon clusters. Owing to the restricted mobility of these
carbon species to dissolve in the bulk and segregate34, makes it dif- clusters, the aggregation event can be more readily size-controlled
ficult to form small GQDs. Instead, large-sized, irregularly shaped than with carbon adatoms derived from a hydrocarbon source.
graphene islands are obtained by the Ostwald ripening process32. On thermally activated diffusion, these clusters coalesced to form

250 NATURE NANOTECHNOLOGY | VOL 6 | APRIL 2011 | www.nature.com/naturenanotechnology

© 2011 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2011.30 ARTICLES
a e
10−15 to 10−16 cm2 s−1

b f

725 K
825 K

c d g h

40 nm 40 nm 10 nm 10 nm

Figure 4 | Comparison of the growth mechanism of graphene nanoislands and quantum dots using C2H4 and C60. a–d, Mechanisms using C2H4. Highly
mobile carbon adatoms from the dehydrogenation of C2H4 (a). Nucleation of the C adatom occurs at the step edges (at ,1 L dose of C2H4) (b). Large-sized,
irregular shaped graphene islands are generated readily (at 1 L , Q , 10 L dose of C2H4). (1 Langmuir (L) ¼ 1 × 1026 torr s.) Corresponding STM images for
the growth of graphene islands from C2H4 (c,d). e–h, Mechanisms using C60. The majority of C60 molecules adsorb on the terrace, and these decompose to
produce carbon clusters with restricted mobility (e). Temperature-dependent growth of GQDs with different equilibrium shape from the aggregation of the
surface diffused carbon clusters (f). Corresponding STM images for the well-dispersed triangular and hexagonal equilibrium shaped GQDs produced from
C60-derived carbon clusters (g,h). Tunnelling parameters (c,d, g,h): V ¼ 0.5 V, I ¼ 1 nA.

a b c

0 70' 90'
d e f

240' 310' 330'

g

1220'

Figure 5 | Series of STM images monitoring the transformation of trapezium-shaped GQDs to triangular-shaped GQDs at 1,000 K. The numbers in the
images indicate the time lapse in seconds. Tunnelling parameters (a–g): V ¼ 1.4 V, I ¼ 0.3 nA; image size, 25 × 12 nm2.

geometrically well-defined GQDs. The adatom-vacancy model rel- and sizes may also exhibit magnetic properties, and can potentially
evant to the molecular embedding and fragmentation of C60 on form building elements for logic gates in ultrafast high-density
the Ru surface may have generic validity for semiconductor or insu- spintronic devices.
lator substrates, which can exhibit strong substrate–carbon bonding;
the templated growth of GQDs on such surfaces may allow nano- Methods
electronic applications to be developed. Arising from size effects The experiments were performed in an ultrahigh-vacuum (UHV) chamber at a base
or topological frustration, graphene clusters of different shapes pressure of 2 × 10210 mbar. The STM used was a SPECS STM 150 Aarhus unit.

NATURE NANOTECHNOLOGY | VOL 6 | APRIL 2011 | www.nature.com/naturenanotechnology 251

© 2011 Macmillan Publishers Limited. All rights reserved.

ARTICLES NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2011.30

In addition to STM topographic images, STS data were collected that were an 19. Sutter, E., Acharya, D. P., Sadowski, J. T. & Sutter, P. Scanning tunneling
arithmetic average of values measured at 50–70 different points (depending on microscopy on epitaxial bilayer graphene on ruthenium (0001). Appl. Phys. Lett.
GQD size) over the surface of the GQDs. The Ru(0001) crystal was depleted of 94, 133101 (2009).
carbon impurities by repeated cycles of argon ion sputtering at room temperature 20. Wang, W. L., Yazyev, O. V., Meng, S. & Kaxiras, E. Topological frustration in
p(Ar) ¼ 5 × 1025 mbar, 1.0 keV, followed by annealing in an O2 pressure of graphene nanoflakes: magnetic order and spin logic devices. Phys. Rev. Lett. 102,
2 × 1027 mbar at 1,000 K, then flashing to 1,600 K. The vacuum system was also 157201 (2009).
equipped with Knudsen cells (MBE-Komponenten) for the evaporation of 21. Ritter, K. A. & Lyding, J. W. The influence of edge structure on the electronic
C60 housed in the preparation chamber. During deposition, the substrate was held properties of graphene quantum dots and nanoribbons. Nature Mater. 8,
at room temperature. The deposition rate of C60 was calibrated by counting the 235–242 (2009).
surface density of the large-scale STM images with coverage below 1 ML (one layer 22. Barone, V., Hod, O. & Scuseria, G. E. Electronic structure and stability of
of fully covered C60). semiconducting graphene nanoribbons. Nano Lett. 6, 2748–2754 (2006).
Spin-polarized simulations based on DFT were conducted using the SIESTA 23. Berger, C. et al. Electronic confinement and coherence in patterned epitaxial
code36. The local density approximation for the exchange-correlation functional by graphene. Science 312, 1191–1196 (2006).
Ceperley and Alder was used37. We applied Troullier–Martins pseudopotentials38. In 24. Xu, K., Cao, P. G. & Heath, J. R. Scanning tunneling microscopy characterization
the case of Ru, relativistic and nonlinear core corrections were added. Double-z plus of the electrical properties of wrinkles in exfoliated graphene monolayers. Nano
polarization localized basis orbitals were used, with the C (2s, 2p) and Ru (4d, 5s) Lett. 9, 4446–4451 (2009).
electrons treated as valence. The C60 molecule was modelled on a five-layer Ru slab 25. Zhang, Y. B. et al. Giant phonon-induced conductance in scanning tunnelling
model (Fig. 3) with a 16.08 × 13.93 × 26.44 Å3 supercell and a 3 × 3 × 1 spectroscopy of gate-tunable graphene. Nature Phys. 4, 627–630 (2008).
Monkhorst–Pack sampling scheme. STM image simulations were conducted under 26. Larsson, J. A. et al. Orientation of individual C60 molecules adsorbed on
the simplifying assumptions of the Tersoff–Hamann theorem39, which states that Cu(111): Low-temperature scanning tunneling microscopy and density
STM images may be modelled by calculation of the local density of states (LDOS) functional calculations. Phys. Rev. B 77, 115434 (2008).
around the Fermi level of the surface of interest. 27. Casarin, M. et al. Strong bonding of single C60 molecules to (1×2)-Pt(110): an
STM/DFT investigation. J. Phys. Chem. C 111, 9365–9373 (2007).
Received 1 December 2010; accepted 15 February 2011; 28. Lacovig, P. et al. Growth of dome-shaped carbon nanoislands on Ir(111): the
published online 20 March 2011 intermediate between carbidic clusters and quasi-free-standing graphene. Phys.
Rev. Lett. 103, 166101 (2009).
References 29. Felici, R. et al. X-ray-diffraction characterization of Pt(111) surface
1. Geim, A. K. & Novoselov, K. S. The rise of graphene. Nature Mater. 6, nanopatterning induced by C60 adsorption. Nature Mater. 4, 688–692 (2005).
183–191 (2007). 30. Li, H. I. et al. Surface geometry of C60 on Ag(111). Phys. Rev. Lett. 103,
2. Novoselov, K. S. et al. Electric field effect in atomically thin carbon films. Science 056101 (2009).
306, 666–669 (2004). 31. Schulze, G. et al. Resonant electron heating and molecular phonon cooling in
3. Choucair, M., Thordarson, P. & Stride, J. A. Gram-scale production of graphene single C60 junctions. Phys. Rev. Lett. 100, 136801 (2008).
based on solvothermal synthesis and sonication. Nature Nanotech. 4, 32. Coraux, J. et al. Growth of graphene on Ir(111). New J. Phys. 11, 023006 (2009).
30–33 (2009). 33. Chen, H., Zhu, W. G. & Zhang, Z. Y. Contrasting behavior of carbon nucleation
4. Li, X. L., Wang, X. R., Zhang, L., Lee, S. W. & Dai, H. J. Chemically derived, in the initial stages of graphene epitaxial growth on stepped metal surfaces. Phys.
ultrasmooth graphene nanoribbon semiconductors. Science 319, Rev. Lett. 104, 186101 (2010).
1229–1232 (2008). 34. Sutter, P. W., Flege, J. I. & Sutter, E. A. Epitaxial graphene on ruthenium.
5. Li, D., Muller, M. B., Gilje, S., Kaner, R. B. & Wallace, G. G. Processable aqueous Nature Mater. 7, 406–411 (2008).
dispersions of graphene nanosheets. Nature Nanotech. 3, 101–105 (2008). 35. Mo, Y. W., Swartzentruber, B. S., Kariotis, R., Webb, M. B. & Lagally, M. G.
6. Ponomarenko, L. A. et al. Chaotic Dirac billiard in graphene quantum dots. Growth and equilibrium structures in the epitaxy of Si on Si(001). Phys. Rev.
Science 320, 356–358 (2008). Lett. 63, 2393–2396 (1989).
7. Kosynkin, D. V. et al. Longitudinal unzipping of carbon nanotubes to form 36. Soler, J. M. et al. The SIESTA method for ab initio order-N materials simulation.
graphene nanoribbons. Nature 458, 872–875 (2009). J. Phys. Condens. Matter 14, 2745–2779 (2002).
8. Schwierz, F. Graphene transistors. Nature Nanotech. 5, 487–496 (2010). 37. Ceperley, D. M. & Alder, B. J. Ground state of the electron gas by a stochastic
9. Cai, J. M. et al. Atomically precise bottom-up fabrication of graphene method. Phys. Rev. Lett. 45, 566–569 (1980).
nanoribbons. Nature 466, 470–473 (2010). 38. Troullier, N. & Martins, J. L. Efficient pseudopotentials for plane-wave
10. Pan, D. Y., Zhang, J. C., Li, Z. & Wu, M. H. Hydrothermal route for cutting calculations. Phys. Rev. B 43, 1993–2006 (1991).
graphene sheets into blue-luminescent graphene quantum dots. Adv. Mater. 22, 39. Tersoff, J. & Hamann, D. R. Theory of the scanning tunneling microscope.
734–738 (2010). Phys. Rev. B 31, 805–813 (1985).
11. Li, L. S. & Yan, X. Colloidal graphene quantum dots. J. Phys. Chem. Lett. 1,
2572–2576 (2010). Acknowledgements
12. Otero, G. et al. Fullerenes from aromatic precursors by surface-catalysed K.P.L acknowledges funding support from an NRF-CRP grant ‘Graphene related materials
cyclodehydrogenation. Nature 454, 865–819 (2008). and devices’ (R-143-000-360-281) as well as MOE Tier 2 grant ‘Structure and dynamics of
13. Chuvilin, A., Kaiser, U., Bichoutskaia, E., Besley, N. A. & Khlobystov, A. N. molecular self-assembled films’ (R-143-000-344-112).
Direct transformation of graphene to fullerene. Nature Chem. 2, 450–453 (2010).
14. Krishna, V., Stevens, N., Koopman, B. & Moudgil, B. Optical heating and rapid
transformation of functionalized fullerenes. Nature Nanotech. 5, Author contributions
J.L. and K.P.L. conceived and designed the experiments. J.L. performed the STM and STS
330–334 (2010).
measurement. P.S.E.Y. carried out theoretical calculations. C.K.G and W.P contributed
15. Cepek, C., Goldoni, A. & Modesti, S. Chemisorption and fragmentation of
analysis tools. K.P.L supervised the project. All authors discussed the results and
C60 on Pt(111) and Ni(110). Phys. Rev. B 53, 7466–7472 (1996).
analysed the data.
16. Swami, N., He, H. & Koel, B. E. Polymerization and decomposition of C60 on
Pt(111) surfaces. Phys. Rev. B 59, 8283–8291 (1999).
17. Marchini, S., Gunther, S. & Wintterlin, J. Scanning tunneling microscopy of Additional information
graphene on Ru(0001). Phys. Rev. B 76, 075429 (2007). The authors declare no competing financial interests. Supplementary information
18. Wang, B., Bocquet, M. L., Marchini, S., Gunther, S. & Wintterlin, J. Chemical accompanies this paper at www.nature.com/naturenanotechnology. Reprints and
origin of a graphene moire overlayer on Ru(0001). Phys. Chem. Chem. Phys. 10, permission information is available online at http://npg.nature.com/reprintsandpermissions/.
3530–3534 (2008). Correspondence and requests for materials should be addressed to K.P.L.

252 NATURE NANOTECHNOLOGY | VOL 6 | APRIL 2011 | www.nature.com/naturenanotechnology

© 2011 Macmillan Publishers Limited. All rights reserved.

 LETTERS
PUBLISHED ONLINE: 6 FEBRUARY 2011 | DOI: 10.1038/NNANO.2010.283

Fast DNA sequencing with a graphene-based

nanochannel device
Seung Kyu Min‡, Woo Youn Kim†‡, Yeonchoo Cho and Kwang S. Kim*

Devices in which a single strand of DNA is threaded through a high surface-to-volume ratio23. Indeed, such sensitivity allows the
nanopore could be used to efficiently sequence DNA1–9. measurement of Hall conductance in graphene even at room temp-
However, various issues will have to be resolved to make this erature24. In contrast to narrow GNRs, carbon nanotubes25 show
approach practical, including controlling the DNA translocation more conductance channels as well as double to multiple stacking,
rate, suppressing stochastic nucleobase motions, and resolving which significantly reduces the characteristic perturbations of the
the signal overlap between different nucleobases4,7. Here, we electric current towards the noise level. This means that GNRs are
demonstrate theoretically the feasibility of DNA sequencing superior to carbon nanotubes (see Supplementary Information).
using a fluidic nanochannel functionalized with a graphene To verify our design hypothesis, we performed comprehensive
nanoribbon. This approach involves deciphering the changes state-of-the-art transport simulations18,19. The trajectory of molecu-
that occur in the conductance of the nanoribbon10,11 as a lar structures for the transport calculations was obtained using
result of its interactions with the nucleobases via p–p stack- molecular dynamics (MD) simulations at room temperature
ing12,13. We show that as a DNA strand passes through the (the NAMD package26) for a ssDNA composed of eight bases
nanochannel14, the distinct conductance characteristics of the (5′ -GCATCGCT-3′ ) located with solvents and counterions in a
nanoribbon15–17 (calculated using a method based on density nanochannel. The nanochannel was composed of silicon nitrate14,
functional theory coupled to non-equilibrium Green function and was equipped with a narrow GNR with a width of 1 nm
theory18–20) allow the different nucleobases to be distinguished (Supplementary Movie)15. The ssDNA was dragged with a force of
using a data-mining technique and a two-dimensional transient 30 kcal mol21 Å21 (0.25 nN per base) to mimic an applied
autocorrelation analysis. This fast and reliable DNA sequencing strong electric field. This strong force was chosen simply for the
device should be experimentally feasible in the near future. sake of computational efficiency, and provided a favourable
In spite of the significant promise of pioneering work on the outcome, but could also be reduced substantially (to an
development of rapid and inexpensive DNA sequencing3–9, a prac-
tical and reliable method to guarantee single-base resolution has yet
a
to be devised. The stochastic motion of a single-stranded DNA
(ssDNA) through a nanopore lacks well-defined interactions between
the nanopore and the nucleobases4. Orientational fluctuations of
rce
nucleobases cause overlaps of the current distributions between ing fo
Pull
different bases7, so it would be advantageous to develop a device
that could hold each base firmly while the base was being interro- NR
AG
gated. We therefore designed a hybrid device composed of a nanos- NA
cale fluidic channel functionalized by a graphene nanoribbon ssD te
stra
(GNR) that interacts with each nucleobase by means of a p–p inter- sub
Si 3N
4
action (Fig. 1). The GNR forms a bridge across a nanochannel14
between opposing tip electrodes (in the same way in which a nano-
lens can be attached to a near-field scanning optical microscope
tip21). The nanochannel provides a pathway along which an
ssDNA can be driven by an external field, and the graphene controls b
the orientational fluctuation of individual bases by stacking them on
its surface by means of the dispersion force between p-systems. The
p–p interaction is sufficiently strong to sustain the stacking struc-
tures for current measurements, while being weak enough to d
allow the ssDNA to be translocated by an external force. θ
Here, we use a narrow GNR15 that can be synthesized using a
recently developed bottom-up approach. As a DNA passes through
a nanochannel, each nucleobase interacts with the GNR via p–p
stacking. The ballistic conductance of the GNR diminishes at Figure 1 | DNA base stacking on a graphene nanodevice during its passage
specific energies corresponding to the characteristic p-molecular through a fluidic nanochannel. a, Schematic of a nanochannel device with
orbitals (MO) via the Fano resonance22, which enables different an armchair GNR (AGNR) through which a ssDNA passes. The water
bases to be distinguished. The conductance of graphene is highly molecules and counterions in the nanochannel are not depicted. b,
sensitive to a small change on its surface due to its unrivalled Instantaneous snapshot from a simulation (d, stacking distance; u, tilt angle).

Center for Superfunctional Materials, Department of Chemistry and Department of Physics, Pohang University of Science and Technology, Hyojadong,
Namgu, Pohang 790-784, Korea; † Present address: Department of Chemistry, KAIST, Daejeon 305-701, Korea; ‡ These authors contributed equally to this
work. * e-mail: kim@postech.ac.kr

162 NATURE NANOTECHNOLOGY | VOL 6 | MARCH 2011 | www.nature.com/naturenanotechnology

© 2011 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2010.283 LETTERS
a 7 b 3
Pristine
6
Ade
5 Gua
2

Transmission

DOS (a.u.)
Cyt
4
Thy
3
1
2

0 0
−2 −1 0 1 2 −2 −1 0 1 2
E − EF (eV) E − EF (eV)

c d

Figure 2 | Electronic and transport properties of a GNR device stacked optimally with DNA bases. a, Transmission curves (in G0). b, Density of states.
c,d, MOs of a GNR with stacked Cyt at energies (E 2 EF) of 20.9 eV (c) and 21.4 eV (d). The sharp dip in transmission of Cyt at 21.4 eV occurs due to the
strong Fano resonance between its MO and graphene (d).

experimental value of 0.25 pN per base14) to provide an increased (Fig. 2c,d), as can be noted from the conspicuous peaks in the
and steadier base–GNR holding time for more reliable measure- density of states for the whole system (Fig. 2b). For example, the
ment. An analysis of structural changes confirmed that each base strong resonance between graphene and the Cyt MO at E –
formed a stable stacking structure over the entire transit. Such p- EF ¼ 21.4 eV (EF is Fermi energy) partially blocks the ballistic con-
stacking is much stronger than the H–p interactions between ductance of the graphene (Fig. 2d).
water molecules and graphene, so the stacking structures are not sig- In experiments, the transmission function can be replaced by con-
nificantly influenced by solvent effects. The p-stacking energies ductance measured at a small bias as a function of gate voltage Vg:
(20 kcal mol21)27 are sufficiently strong to form stable stacked
structures of nucleobases on graphene in water at room temperature.   2e2  
With regard to ensuring experimental feasibility, we investigated G Vg = T m
h
how to control the sliding motion of the ssDNA on graphene.
Sliding friction is uniform because of the non-local character of where h is Planck’s constant and m ¼ EF 2 eVg is the chemical poten-
the dispersion force, as can be noted in the extremely smooth tial adjusted by the gate voltage. Measuring conductance rather than
motion observed in carbon nanotubes28,29. The translational speed current allows a clearer distinction of the change in the signal,
of a DNA strand in a nanochannel can be made steady by using because conductance through the GNR device changes by a unit of
another wide GNR (in addition to the GNR used to measure the quantum conductance G0 (¼2e 2/h) in its interaction with DNA
currents) on which a number of bases are simultaneously stacked bases. For instance, the 50% relative reduction from 2G0 to 1G0
(frictional force is proportional to GNR width30). For practical and the dip width of 30 meV in the conductance of this GNR
experiments, the force used in the computations could be reduced device (see Supplementary Information) is experimentally measur-
by three orders of magnitude to an experimental value14. The fric- able, being well above the noise level at room temperature. At each
tional force of ssDNA on a large graphene sheet, which can be esti- DNA translocation step, the two-dimensional conductance of the
mated from the frictional force between two neighbouring shells in a GNR has to be measured for a whole sweep of chemical potential,
carbon nanotube28, offsets a dragging force such as that exerted by which is experimentally controlled via a backgate capacitively
an externally applied electric field. A DNA strand can then slide on a coupled to the GNR. The rate-determining step is then the voltage
graphene surface smoothly and steadily with controlled speed. The sweep. The transit time of each base should be controlled so as to
negatively charged phosphate groups translocate the whole DNA be longer (10 ms) than the gate voltage sweeping time (0.1 ms,
against the direction of the electric field, while the counterions sol- 10 MHz); this lies at the limit of present experimental capability.
vated in water move along the direction of the field. The transmission function in the energy range of interest is not
To promote the identification of each base, we investigated the affected significantly when the geometries of the bases on the GNR
quantum conductance of the graphene-based device using our are changed, or in response to environmental effects (such as back-
Postrans package19. A narrow armchair GNR with a width of bones, water, hydrated phosphate groups and counterions). This is
1 nm and a bandgap calculated to be 1.16 eV was considered. because the other molecular groups have much larger energy gaps
Figure 2a shows the transmission curves of the nucleobase–GNR between the occupied and unoccupied MOs than nucleobases, and
complexes as a function of energy, T(E). The pristine GNR shows there are cancellation effects due to the similar background environ-
integer transmission values corresponding to the number of con- ments and the random motions of the environmental species.
duction channels. However, the GNR with a stacked base presents The transmission data in Fig. 3a were obtained using density
a transmission curve with sharp drops at the characteristic MO functional theory coupled to non-equilibrium Green’s function
energies of each base, due to the Fano resonance between the loca- (DFT-NEGF) method for the time-evolving series of base–GNR
lized MOs of the nucleobase and the continuum p-band of the GNR geometries, from the MD trajectories for the whole sequence of

NATURE NANOTECHNOLOGY | VOL 6 | MARCH 2011 | www.nature.com/naturenanotechnology 163

© 2011 Macmillan Publishers Limited. All rights reserved.

LETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2010.283

a 2 transmission drops appearing at a specific energy E within a tiny

SC time interval D around time t:
1.5
t
1
H(E, t; D) = D(E, t ′ )dt ′
E − EF (eV)

−0.5 t−D
SG
−1 SA/C where D(E, t) is the number of peaks in the energy interval (E, E þ 1)
−1.5 SG/T at time t (1 is the histogram energy interval), and D characterizes the
−2 sampling size of the time data set.
0.25 0.5 0.75 1.0 1.25 1.5 However, the histogram in Fig. 3a, taken alone, is not sufficient to
Time (ns) achieve single base resolution. For automatic sequencing, we devised
b GT2
a data-mining approach for more simplistic identification (see
GT3 GT4
SG/T GT1 Methods). We then plotted the probability function projected
onto each base, Px(t)(x ¼ G/T/C/A) (Fig. 3b,c). In this way, each
AC1 AC2 AC3 AC4
SA/C base is uniquely separated from the others, allowing single base res-
olution in the test sequencing for 5′ -GCATCGCT-3′ (see
G1 G2
SG Supplementary Information).
Much more reliable sequencing is achieved by clear recognition
C1 C2 of the interaction interval of each individual base with the GNR,
SC C3
removing the possibility of the introduction of ambiguities if a
0.25 0.5 0.75 1.0 1.25 1.5 series of bases of the same type pass beneath the GNR. To this
Time (ns) end, we have developed a normalized two-dimensional transient
c autocorrelation function (TACF) for the given time spant :
G C A T C G (C) T
1 PG t0 +t 1
Amplitude

PC C(t, t0 ; t) = dt ′ dE J(E, t ′ )J(E, t + t ′ )/

PA t0 −1
0 PT
t0 +t 1
dt ′ dE [J(E, t ′ )]2
0.25 0.5 0.75 1.0 1.25 1.5 t0 −1
Time (ns)

d 0.1 1.0
This gives the autocorrelation of a given event function J(E, t)
between two time intervals of duration t separated by time t.
Here, t is smaller than the passage time of any single base. In the
t (ns)

situation where the transient correlation between different bases is

zero, the two-dimensional TACF shows a distinct triangular
0.0 0.9 pattern when one base passes out and another base comes into
0.25 0.5 0.75 1.0 1.25 1.5
t0 (ns) the interaction region (Fig. 3d; Supplementary Fig. S2). This helps
us determine the passage time of each base. Then, from Fig. 3c,d,
e base sequencing is possible (Fig. 3e).
G
C In summary, we have reported an ultrasensitive graphene-
A embedded nanochannel device that effectively controls the motion
T of nucleobases via p–p interaction and deciphers the ultrasensitive
characteristic Fano resonance-driven conductance of individual
0.25 0.5 0.75 1.0 1.25 1.5
Time (ns)
bases, one by one, in real time. In practice, the resolution of sequen-
cing patterns will be much clearer as the DNA translocation slows
Figure 3 | Simulation results of the transport properties of 5′ -GCATCGCT-3′ .
down with a dragging force much weaker than that used in the
a, Time-dependent histogram of the transmission peak positions
calculations. Any further ambiguities can be resolved by either
(red, maximal values; blue, minimal values). The histogram shows features
parallel or repetitive serial measurements, because the ssDNA is
of Cyt (SC), Gua (SG), Ade/Cyt (SA/C) and Gua/Thy (SG/T) corresponding
reusable. The experimental implementation of this device would
to the band centred around E 2 EF ¼ Es ¼ 1.8, 20.65, 21.2 and 21.65 eV,
allow ultrafast and reliable DNA sequencing, which is essential for
respectively. The height of each box to the right of the histogram represents
collecting wide-ranging data related to differences/similarities
the energy range of integration for each Sbase. b, Plots of Sbase(t) versus time
between diverse DNA sequences in the field of bioinformatics.
as an ssDNA passes. The sequence GCATCGCT can be inferred from the
four curves. c, Resolving of the sequence by Pbase(t), as a function of
Methods
The DFT-NEGF transport calculations18–20 based on the local density
Sbase(t). d, Plot of two-dimensional TACF, C(t, t0; t ¼ 0.025 ns). The black approximation and using double-z plus polarization basis sets were performed in a
dotted lines each indicate an instant at which a (subsequent) DNA base fully self-consistent way. These calculations were carried out for all the time-evolving
passes out; white dashed lines each indicate an instant at which a geometries and the entire sequence of the DNA interacting with the GNR, using
(subsequent) DNA base comes in. e, Final base sequence obtained from room-temperature MD simulation trajectories including all surrounding water
molecules and counterions in the nanochannel, which was composed of silicon
analysis of c and d. Colours and heights are given in relative intensity. nitrate and with a narrow GNR (width, 1 nm).
For automatic sequencing analysis using the above electric conductance data, a
5′ -GCATCGCT-3′ interacting with the GNR, including all data-mining approach was required. As specific drops in the transmission curves at
surrounding water molecules and counterions in the nanochannel. 1.8, –0.9, –1.4 and –1.6 eV in the absence of water (Fig. 2a) indicated stacking of the
corresponding nucleobases on the graphene, the histogram H(E, t; D) showed
To identify each base, we manipulated the transmission raw data similar characteristic patterns in a realistic GNR–nanochannel device containing
using our data-mining technique. In the first step, we introduced whole DNA in water with counterions (Fig. 3a). H(E, t; D) was integrated over a
a time-dependent histogram H(E, t; D) to represent the number of specified energy range [Es – Gs/2, Es þ Gs/2] to obtain the number of transmission

164 NATURE NANOTECHNOLOGY | VOL 6 | MARCH 2011 | www.nature.com/naturenanotechnology

© 2011 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2010.283 LETTERS
drops appearing within the given energy range and time interval D around t: 14. Liang, X. & Chou, S. Y. Nanogap detector inside nanofluidic channel for fast
real-time label-free DNA analysis. Nano Lett. 8, 1472–1476 (2008).
Es +Gs /2 15. Cai, J. et al. Atomically precise bottom-up fabrication of graphene nanoribbon.
S(Es , t; D) = H(E, t; t)dE Nature 466, 470–473 (2010).
Es −Gs /2
16. Kosynkin, D. V. et al. Longitudinal unzipping of carbon nanotubes to form
graphene nanoribbons. Nature 458, 872–876 (2009).
The energy values Es (1.8, –0.65, –1.2 and –1.65 V for Cyt, Gua, Ade/Cyt and 17. Jiao, L., Zhang, L., Wang, X., Diankov, G. & Dai, H. Narrow graphene
Gua/Thy, respectively) were chosen from the histogram in Fig. 3a. The small nanoribbons from carbon nanotubes. Nature 458, 877–880 (2009).
coloured bars at the right side of Fig. 3a show the ranges of Gs at a given energy Es. 18. Kim, W. Y. & Kim, K. S. Tuning molecular orbitals in molecular electronics and
To unravel some partial overlaps, we used the probability projecting each base. spintronics. Acc. Chem. Res. 43, 111–120 (2010).
For example, the probability of Gua (Cyt) is obtained by projecting SG,T (SA,C) 19. Kim, W. Y. & Kim, K. S. Prediction of very large values of magnetoresistance in a
into SG (SC): graphene nanoribbon device. Nature Nanotech. 3, 408–412 (2008).
20. Kim, W. Y. & Kim, K. S. Carbon nanotube, graphene, nanowire, and molecule
PG (t) = SG SG,T ; PC (t) = SC SA,C based electron and spin transport phenomena using the nonequilibrium Green’s
function method at the level of first principles theory. J. Comput. Chem. 29,
The probability of Thy (Ade) can then be obtained by projecting out PG (PT) from 1073–1083 (2008).
SG,T (SA,C): 21. Lee, J. Y. et al. Near-field focusing and magnification through self-assembled
nanoscale spherical lenses. Nature 460, 498–501 (2009).
PT (t) = SG,T − PG ; PA (t) = SA,C − PC 22. Fano, U. Effects of configuration interaction on intensities and phase shifts.
Phys. Rev. 124, 1866–1878 (1961).
23. Schedin, F. et al. Detection of individual gas molecules adsorbed on graphene.
Thus, the data-mining technique, including two-dimensional TACF analysis of
Nature Mater. 6, 652–655 (2007).
C(t, t0; t), for the time evolution of the conductance renders the automatic
24. Novoselov, K. S. et al. Room-temperature quantum hall effect in graphene,
sequencing feasible.
Science 315, 1329 (2007).
Received 7 October 2010; accepted 22 December 2010; 25. Meng, S. et al. DNA nucleoside interaction and identification with carbon
nanotubes. Nano Lett. 7, 45–50 (2007).
published online 6 February 2011 26. Phillips, J. C. et al. Scalable molecular dynamics with NAMD. J. Comput. Chem.
26, 1781–1802 (2005).
References 27. Antony, J. & Grimme, S. Structures and interaction energies of stacked
1. Collins, F. S. et al. A vision for the future of genomics research. Nature 422, graphene–nucleobase complexes. Phys. Chem. Chem. Phys. 10,
835–847 (2003). 2722–2729 (2008).
2. Service, R. F. The race for the $1000 genome. Science 311, 1544–1546 (2006). 28. Cummings, J. & Zettl, A. Low-friction nanoscale linear bearing realized from
3. Shendure, J. & Ji, H. Next-generation DNA sequencing. Nature Biotechnol. 26, multiwall carbon nanotubes. Science 289, 602–605 (2000).
1135–1145 (2008). 29. Hong, B. H. et al. Extracting subnanometer single shells from ultralong
4. Branton, D. et al. The potential and challenges of nanopore sequencing. Nature multi-walled carbon nanotubes. Proc. Natl Acad. Sci. USA 102,
Biotechnol. 26, 1146–1153 (2008). 14155–14158 (2005).
5. Garaj, S. et al. Graphene as a subnanometre trans-electrode membrane. Nature 30. Mo, Y., Turner, K. T. & Szlufarska, I. Friction laws at the nanoscale. Nature 457,
467, 190–193 (2010). 1116–1119 (2009).
6. Meller, A., Nivon, L. & Branton, D. Voltage-driven DNA translocations through
a nanopore. Phys. Rev. Lett. 86, 3435–3438 (2001).
7. Zwolak, M. & Di Ventra, M. Colloquium: physical approaches to DNA Acknowledgements
sequencing and detection. Rev. Mod. Phys. 80, 141–165 (2008). This work was supported by the National Research Foundation (National Honor Scientist
8. Lagerqvist, J., Zwolak, M. & Di Ventra, M. Fast DNA sequencing via transverse program: 2010-0020414, WCU:R32-2008-000-10180-0, EPB Center: 2009-0063312, GRL)
electronic transport. Nano. Lett. 6, 779–782 (2006). and KISTI (KSC-2008-K08-0002). The authors thank D. R. Mason and N. Kim
9. Zwolak, M & Di Ventra, M. Electronic signature of DNA nucleotides via for discussions.
transverse transport. Nano. Lett. 5, 421–424 (2005).
10. Novoselov, K. S. et al. Electric field effect in atomically thin carbon film. Science Author contributions
306, 666–669 (2004). S.K.M. and W.Y.K. worked together. Y.C. assisted in the calculations and analysis.
11. Kim, K. S. et al. Large-scale pattern growth of graphene films for stretchable K.S.K. supervised the project. S.K.M., W.Y.K. and K.S.K. wrote the paper together.
transparent electrodes. Nature 457, 706–710 (2009).
12. Kim, K. S., Tarakeshwar, P. & Lee, J. Y. Molecular clusters of p-systems:
theoretical studies of structures, spectra and origin of interaction energies. Additional information
Chem. Rev. 100, 4145–4185 (2000). The authors declare no competing financial interests. Supplementary information
13. Lee, E. C. et al. Understanding of assembly phenomena by aromatic–aromatic accompanies this paper at www.nature.com/naturenanotechnology. Reprints and
interactions: benzene dimer and the substituted systems. J. Phys. Chem. A 111, permission information is available online at http://npg.nature.com/reprintsandpermissions/.
3446–3457 (2007). Correspondence and requests for materials should be addressed to K.S.K.

NATURE NANOTECHNOLOGY | VOL 6 | MARCH 2011 | www.nature.com/naturenanotechnology 165

© 2011 Macmillan Publishers Limited. All rights reserved.

 news & views

NANOPORES

Graphene opens up to DNA

It might be possible to sequence DNA by passing the molecule through a small hole in a sheet of graphene.

Zuzanna S. Siwy and Matthew Davenport

T
he nanopores that are found in the a b c
membranes of cells have an important
role in biology and, along with
artificial nanopores fabricated in materials
such as silicon nitride, they have also been
used to detect single molecules and, in some
cases, extract more detailed information
such as the size and shape of the molecules1.
The capabilities of these systems are
influenced by the diameter of the nanopore
and also by the thickness of the membrane 5 nm 10 nm 10 nm
that contains it. In general, however, these
membranes are too thick to be able to
probe the local structure of a molecule as it Figure 1 | Images of nanopores sculpted in graphene by the electron beam of a transmission electron
passes through the nanopore. To read the microscope and used in the work of a, Golovchenko and colleagues6, b, Dekker and colleagues7 and
sequence of nucleotides in a DNA molecule, c, Drndić and colleagues8. The rings surrounding the pore in (c) are indicative of the number of graphene
for example, a membrane of sub-nanometre layers that form the membrane. Parts b and c reproduced with permission from: b, ref. 7, © 2010 ACS;
thickness is required. c, ref. 8, © 2010 ACS.
Graphene is a two-dimensional sheet
of carbon atoms arranged in a honeycomb
lattice. The material has remarkable The three teams — led by Jene be viable for DNA detection8. Furthermore,
mechanical, electronic and optical properties Golovchenko of Harvard, Cees Dekker of the nanopores could differentiate between
and, as a result, is at present being investigated Delft and Marija Drndić of UPenn — all DNA that passed through the pore in an
for a variety of applications, including flexible address the issue of DNA detection using extended form and that which passed
touch-screen devices2–4. This mechanical graphene nanopores. DNA is a long, through in a folded form (Fig. 2c).
stability also means that it is possible to use spaghetti-like molecule made up of repeating Despite these exciting developments,
an electron beam to sculpt a nanopore in a units called nucleotides. Each nucleotide is the central question remains: is single-
suspended sheet of graphene5. Writing in composed of one of four bases — adenine base resolution with a graphene nanopore
Nature and Nano Letters, three independent (A), cytosine (C), guanine (G) or thymine feasible? All three groups found that the
groups based at Harvard University and (T) — attached to a backbone of sugar and translocation events are too fast to be
Massachusetts Institute of Technology 6, Delft phosphate groups. Only one nucleotide can resolved by the existing detection electronics.
University of Technology 7, and the University reside in the ultrathin graphene pore at a However, the Harvard–MIT team went a
of Pennsylvania8 now report using graphene time, so it should be possible to sequence by step further and quantified the resolution
membranes as nanopore sensors. observing the translocation of consecutive obtainable with a graphene nanopore using
The basis of nanopore detection is the nucleotides through the nanopore. numerical simulations6. Golovchenko and
resistive pulse technique, or Coulter-counter The Harvard–MIT team used graphene colleagues considered the translocation
method, in which a membrane containing a membranes 1 or 2 layers thick6, the Delft of a cylindrical molecule whose diameter
single nanopore is sandwiched between two team membranes 1–8 layers thick7 and changed discontinuously from 2.2 nm to
reservoirs of electrolytic solution, such as the UPenn team membranes 3–15 layers 2.0 nm (diameters that are consistent with
aqueous potassium chloride. Ions are driven thick8, and all report initial DNA detection the dimensions of base pairs in double-
through the pore by applying an electric experiments that are an important step stranded DNA). Using a membrane
potential difference across the membrane, towards DNA sequencing (Fig. 1). The thickness derived from experiment (0.6 nm),
resulting in an ionic current that can be membranes (prepared by chemical vapour as well as a more conservative estimate
measured. When the electrolyte contains deposition6,8 or exfoliation from graphite7) of thickness (1.5 nm), they showed that
larger charged molecules, such as DNA or each contain single pores with diameters graphene nanopores should inherently
proteins, these are also driven through the of 5–25 nm drilled in the graphene sheet possess sub-nanometre resolution. In the
pore, causing a transient dip in the ionic with an electron beam. All groups presented case of the 0.6-nm membrane, the resolution
current. Thus, each current dip represents a membranes that could detect double- was determined to be 0.35 nm — nearly
passage, or translocation, of a biomolecule, stranded DNA molecules with lengths that identical to the length of a single base pair.
with the magnitude and duration of the varied from 400 to 48,000 base pairs (Fig. 2). On a more fundamental level, the
pulse being indicative of the molecule’s Even membranes with significant variability electrolyte/graphene interface presents an
radius and length, respectively. in the baseline current levels were found to intriguing physical system. The Harvard–

nature nanotechnology | VOL 5 | OCTOBER 2010 | www.nature.com/naturenanotechnology 697

© 2010 Macmillan Publishers Limited. All rights reserved

news & views

a b This work on graphene nanopores is

Add DNA
certain to be the precursor to further studies
in a variety of fields. Within the scope of
nanopore sensing, there are numerous

1 ns
0.2 s
avenues to explore using the graphene
itself as an electrode to control the local
electric potential and translocation rates, or
c to monitor the transverse conductance of
individual nucleotides as they pass through
a pore9. Previous theoretical work suggests
that the latter technique, if used in the proper
geometry, could provide DNA sequencing
with a 0% error rate10. Combined with recent

1 ns
2 ms advances in the production of large-area
high-quality graphene (sheets with up to a
+ + +
30-inch diagonal)4, the results open the door
– – –
to a host of interesting explorations including
the development of new membrane systems
Figure 2 | Translocation of DNA through a graphene nanopore. a, Double-stranded DNA threading a for biosensing. ❐
graphene nanopore. Each coloured segment represents a different nucleotide. The software VMD 1.8.7,
PyMOL1.2r1, and TubeGen 3.3 was used in preparation of the image. b, Characteristic conductance versus Zuzanna S. Siwy and Matthew Davenport are in the
time signals of a graphene nanopore before and after the addition of DNA. c, The conductance signature Department of Physics and Astronomy, University of
(top) of various DNA conformations (bottom) as they translocate through a graphene nanopore. Parts b California, Irvine, California 92697, USA.
and c reproduced with permission from ref. 7, © 2010 ACS. e-mail: zsiwy@uci.edu

References
1. Dekker, C. Nature Nanotech. 2, 209–215 (2007).
MIT team observed that although the understood if an ionic-current-based DNA 2. Novoselov, K. S. et al. Science 306, 666–669 (2004).
3. Neto, A. H. C., Guinea, F., Peres, N. M. R., Novoselov, K. S. &
ion permeability of an intact graphene sequencing algorithm is to be employed. Geim, A. K. Rev. Mod. Phys. 81, 109–162 (2009).
membrane was negligible compared with Perhaps the most striking question 4. Bae, S. et al. Nature Nanotech. 5, 574–578 (2010).
that of a membrane that contained a that arises from these reports relates to the 5. Fischbein, M. D. & Drndić, M. Appl. Phys. Lett.
93, 113107 (2008).
nanopore, it was not zero and it was different dimensionality of the graphene pore: the 6. Garaj, S. et al. Nature 467, 190–193 (2010).
for different cations, which indicates that pore was seen to behave as both a cylinder 7. Schneider, G. F. et al. Nano Lett.
electrolyte–graphene interactions need to of finite length (with its conductance being 10, 3163–3167 (2010).
be taken into account6. Evidence of DNA– proportional to the square of the diameter)7 8. Merchant, C. A. et al. Nano Lett.
10, 2915–2921 (2010).
graphene interactions was also observed6–8. and as an infinitesimally thin circular aperture 9. Zwolak, M. & Di Ventra, M. Rev. Mod. Phys. 80, 141–165 (2008).
These interactions will need to be carefully (with conductance proportional to diameter)6. 10. Postma, H. W. Ch. Nano Lett. 10, 420–425 (2010).

NANOElEcTRONicS

Nanoribbons on the edge

Arrays of graphene nanoribbons are fabricated on structured silicon carbide substrates using self-organized growth,
without lithography and with well-controlled widths.

John A. Rogers

G
raphene is attractive for electronics better switching capabilities, but they are large enough to allow efficient transistor
because of its exceptional properties difficult to manufacture owing to the very switching3. However, practical circuits
and the relative ease with which it can small widths that are required. Writing in require very narrow widths (less than
be integrated into transistors and circuits. In Nature Nanotechnology, Michael Sprinkle, ~10 nm), with exceptional uniformity in the
the form of large-area planar sheets, it can be Walter de Heer and colleagues at the Georgia lateral dimension, and well-ordered edges.
processed using straightforward adaptations Institute of Technology and the Institut Recent reports suggest that such ribbons
of methods that are already in widespread Néel in Grenoble report a clever growth can be achieved by ultrasonic tearing and
use by the semiconductor industry. However, technique that yields organized, aligned exfoliation of larger sheets4, chemical
graphene in its native single-layer form has arrays of narrow ribbons of graphene on ‘unzipping’ of carbon nanotubes5 and
zero electronic bandgap, which prevents structured substrates of silicon carbide covalent linking of assemblies of molecular
transistor devices made with it from being (SiC), with the potential to enable switching building blocks6. In their present state of
turned off completely. This makes graphene properties suitable for broad classes of development, however, these strategies do
unusable in power-efficient digital circuits electronic systems2. not offer the levels of engineering control
(though it can still be used in radio frequency Early theoretical work suggested that over dimensions or spatial layouts needed
electronics)1. Narrow graphene ribbons have graphene ribbons would have bandgaps for realistic applications. Lithographic

698 nature nanotechnology | VOL 5 | OCTOBER 2010 | www.nature.com/naturenanotechnology

© 2010 Macmillan Publishers Limited. All rights reserved

 LETTERS
PUBLISHED ONLINE: 3 OCTOBER 2010 | DOI: 10.1038/NNANO.2010.192

Scalable templated growth of graphene

nanoribbons on SiC
M. Sprinkle1, M. Ruan1, Y. Hu1, J. Hankinson1, M. Rubio-Roy1, B. Zhang1, X. Wu1, C. Berger1,2
and W. A. de Heer1 *

In spite of its excellent electronic properties, the use of Hall effect21, and suggest that this fact may be used to produce
graphene in field-effect transistors is not practical at room nanoribbons by novel fabrication methods.
temperature without modification of its intrinsically semimetal- It has long been known22 that SiC{0001} surfaces exhibit step
lic nature to introduce a bandgap1–4. Quantum confinement bunching in various environments. Recent systematic studies have
effects can create a bandgap in graphene nanoribbons, but found a greater propensity for step bunching on the (0001) face23–26,
existing nanoribbon fabrication methods are slow and often with vicinal miscuts towards k1100l displaying bunching of
produce disordered edges that compromise electronic proper- parallel steps into (110n) ‘nanofacets’ up to 4–5 unit cells in
ties2–4. Here, we demonstrate the self-organized growth of gra- height, oriented at an angle of 258 to the basal plane24,25. It has
phene nanoribbons on a templated silicon carbide substrate been suggested that such nanofacet formation may be understood
prepared using scalable photolithography and microelectronics as a minimization of surface free energy24,27. Steps perpendicular
processing. Direct nanoribbon growth avoids the need for to the directions k1100l are strongly favoured on (0001), such that
damaging post-processing. Raman spectroscopy, high-resol- steps formed macroscopically towards k1120l are microscopically
ution transmission electron microscopy and electrostatic zigzagged, with segments perpendicular to k1100l23–26. The (0001)
force microscopy confirm that nanoribbons as narrow as face, in contrast, seems to form steps without restriction on orien-
40 nm can be grown at specified positions on the substrate. tation25. Step-bunched nanofaceting has not been observed there
Our prototype graphene devices exhibit quantum confinement previously25,26, but we find (Supplementary Fig. S1) that a (110n)
at low temperatures (4 K), and an on–off ratio of 10 and facet is induced by pre-processing (see below). These results are
carrier mobilities up to 2,700 cm2 V21 s21 at room temperature. qualitatively true of both 6H- and 4H-SiC polytypes25. It is
We demonstrate the scalability of this approach by fabricating perhaps expected that epitaxial graphene growth should proceed
10,000 top-gated graphene transistors on a 0.24-cm2 SiC chip, first on nanofacets, given the weaker bonding of silicon atoms
which is the largest density of graphene devices reported there, and this has been observed on etching-induced (110n)19
to date. and (112n)28 nanofacets.
Epitaxial graphene on SiC, prepared by thermal decomposition We propose exertion of control over the natural step-bunching
of SiC, with the same as a built-in electronics-grade substrate, is mechanism to prepare a crystal facet for self-organized graphene
an exciting new electronic material that presents the possibility of growth (Fig. 1). Given the previous discussion, the best choice for
room-temperature ballistic devices1,5,6. Extremely high carrier this purpose may be (110n). Controlled facets are achieved by photo-
mobilities, exceeding 250,000 cm2 V21 s21 at room temperature, lithographic definition of nickel lines on a SiC substrate perpendicu-
have been observed7, and it has been shown to be compatible lar to the k1100l direction; these lines are transferred into the SiC by a
with traditional top-down processing techniques8,9. Its linear band fluorine-based reactive ion etch (RIE), which, while relatively simple
structure10 allows ambipolar tuning of conduction and direct technologically, allows nanometre precision in the etch depth. As
application to radiofrequency devices11,12, and, in addition to depicted in Fig. 1, it is the etch depth that ultimately defines the
nanoribbon fabrication, bandgaps have been introduced by selective width of the nanoribbons prepared. Etch depths of 20 nm were
chemical treatment13,14. The nanoribbon approach, which intro- readily achieved in this work (see Fig. 3), resulting in a ribbon
duces a bandgap inversely proportional to ribbon width15, is prom- width (40 nm) sufficiently narrow to give a sizable bandgap at
ising in that it has shown large gating effects2,4. room temperature2–4. Much narrower widths should be achievable
The morphology of epitaxial graphene on SiC is highly influ- with further development. After removal of the nickel mask and
enced by the underlying SiC structure. In well-controlled graphene cleaning, the crystal is heated to elevated temperatures (1,200–
growth conditions5,6,16–18, nominally on-axis SiC retains an ordered 1,300 8C) at intermediate vacuum (1024 torr) for 30 min, inducing
terrace structure that originates in the miscut angle of the SiC wafer. SiC step flow. The abrupt step relaxes to a (110n) facet, and the temp-
Many interpret these steps as being problematic19, but scanning tun- erature is elevated to .1,450 8C within 1.5 min, maintained for
nelling microscopy measurements have consistently observed that 10 min for graphene growth, then allowed to cool naturally, falling
the graphene lattice is continuous over such steps. This is true of below 1,300 8C within 0.25 min (refs 1,5,16).
0.5-nm 4H-SiC half-unit cell steps and few-nanometre steps This careful control of growth temperature, time and atmosphere
where step bunching occurs1,20. We find that this remains true allows selective growth on the facet, as shown by Raman mapping in
at a much larger scale, over prepared SiC steps as large as Fig. 2f. The intensity of the 2D Raman band (2,700 cm21) charac-
150 nm (Supplementary Fig. S1). These observations help explain teristic of graphene is mapped over a 100-nm SiC step and adjacent
the transport measurements in which underlying SiC steps appear (0001) faces. Little to no intensity is observed on the horizontal sur-
to have little effect on mobility5,21, or observation of the quantum faces, but significant intensity is seen at the step edge, indicating the

1
Georgia Institute of Technology, Atlanta, Georgia 30332-0430, USA, 2 CNRS/Institut Néel, BP166, 38042 Grenoble, France.
* e-mail: deheer@electra.physics.gatech.edu

NATURE NANOTECHNOLOGY | VOL 5 | OCTOBER 2010 | www.nature.com/naturenanotechnology 727

© 2010 Macmillan Publishers Limited. All rights reserved.

LETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2010.192

a a _ b
Etch depth (0001)

c d

_
SiC (0001) + RIE
e f

_
b (110n)
facet

(0001)

0 1,000
2D intensity (a.u.)
g

24°

Graphene ·0001Ò
nanoribbon _
·1100Ò _
·1120Ò
Graphene SiC

d Al2O3
Gate Figure 2 | Raman and TEM observations of graphene grown selectively on
Drain SiC nanofacet (110n) with n ≈ 8. a, Optical micrograph of pre-patterned
100-nm step on the SiC(0001) face following graphene growth. b, Raman
map (lateral resolution, 1 mm; grid, 0.25 mm) of the 2D peak intensity at
this location indicates preferential graphene growth on the (110n) facet.
c,d, Optical micrograph (c) and Raman map (d) of step on SiC(0001)
following exposure to directional O2 RIE. e,f, Optical micrograph (e) and
Raman map (f) demonstrating fully selective growth on SiC(0001) without
Source post-treatment. Scale bars in a,c,e, 2 mm. g, HRTEM cross-sectional images
SiC
of a similar step on (0001) confirm preferential growth on the (110n) facet.
Scale bar, 2 nm (and for all insets).
Figure 1 | Process for tailoring of the SiC crystal for selective graphene
growth and device fabrication. a, A nanometre-scale step is etched into SiC corresponding to the high-index SiC facet (110n) as expected, where
crystal by fluorine-based RIE. b, The crystal is heated to 1,200–1,300 8C (at n ≈ 8. The precise facet obtained is dependent on processing tempera-
low vacuum), inducing step flow and relaxation to the (110n) facet. c, Upon ture23. It must be noted that apparent imperfections in the graphene
further heating to 1,450 8C, self-organized graphene nanoribbon forms on sheets may be introduced by the HRTEM specimen (thin slice) polish-
the facet. d, Complete device with source and drain contacts, graphene ing process or the high-energy electron beam during imaging.
nanoribbon channel, Al2O3 gate dielectric and metal top-gate, as Electrostatic force microscopy (EFM, related to Kelvin probe FM)
shown in Fig. 5b. allows observation of narrower epitaxial graphene nanoribbons
(Fig. 3). This probing of the local relative workfunctions differen-
presence of graphene there. Note that the lateral resolution of the tiates epitaxial graphene on the nanofacet from the surrounding sub-
Raman instrument, at 1 mm, is much larger than the facet width strate (this correlates to mapping of the 2D Raman line), highlighting
and the mapping grid spacing. self-organized nanoribbons as shown in Fig. 3c. Here, the SiC step is
Cross-sectional high-resolution transmission electron microscopy 20 nm, and the observed graphene ribbon width is 40 nm. Note the
(HRTEM) images (with slightly thicker graphene for visibility) three-dimensional structure, with epitaxial graphene interconnects
(Fig. 2g) confirm preferential growth and enable observation of the bridged by nanoribbons, all self-organized on the prepared template.
nanofacet. Graphene is observed on the facet, with only partial Ribbon samples formed on (0001) (Figs 4a and 5b) were prepared
layers on the horizontal (0001) plane. The facet angle observed, 248, for electrical measurement by exposure to an extremely short direc-
is in agreement with atomic force microscopy (AFM) measurements tional O2 RIE to remove any graphene fragments from the horizontal
(not shown) of 24–288 across multiple samples and locations, (0001) surface. This was verified by Raman mapping, as shown in

728 NATURE NANOTECHNOLOGY | VOL 5 | OCTOBER 2010 | www.nature.com/naturenanotechnology

© 2010 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2010.192 LETTERS
a b c

Figure 3 | AFM/EFM observation of self-organized epitaxial graphene nanoribbons of width ∼40 nm. a, AFM topography of patterned SiC structure on
(0001) before graphene growth. Horizontal trench of depth 20 nm (to become the active device area) is flanked by trenches of depth 100 nm (to become
interconnects). b, AFM (topography) detail of white outlined area in a following graphene growth. SiC step bunching and step flow cause the patterned
steps to become atomically flat semicircular plateaus atop an atomically flat terrace. c, Corresponding EFM amplitude (relative workfunction) detail
highlights graphene nanoribbons on (110n) relative to surrounding substrate. Given the finite size of the conducting probe tip (radius, 20 nm), graphene
nanoribbon width is estimated at 40 nm. Wider interconnect-like graphene ribbons at left and right join the nanoribbon device to the larger circuit. Scale
bars and colour scales (far right) in a,b and c are 1 mm, 500 nm, 500 nm, 0–140 nm, 0–87 nm and 3.7–4.4 V, respectively. See electrical transport
data in Fig. 4c.

a 1.4 b
77 K 1.8
20 K
1.6
1.2
15 K 1.4

1
R (kΩ ☐−1)

1.2
G (mS ☐)

10 K
1
0.8
0.8

0.6
0.6
4K
0.4

0.4 0.2
−20 −10 0 10 20 −20 −10 0 10 20
Vsd (mV) Vg − Vg Dirac (V)
c 0.2 3.8 d 60
0.8
0.6
R (kΩ ☐−1)

50
0.4
0.1 3.6
0.2
40
0
R (kΩ ☐−1)
G (mS ☐)

−5 0 5
I (mA)

Vg (V)
0 3.4 30

20
−0.1 3.2
10

−0.2 3 0
−1 −0.5 0 0.5 1 −0.2 0 0.2
Vsd (V) Vg (V)

Figure 4 | Electronic transport measurement of graphene nanoribbons. a, Conductance versus source–drain voltage as a function of temperature.
A bandgap is observed at 4 K, as expected for a ribbon of this width (250 nm). Inset, optical micrograph of test structure. Scale bar, 15 mm; channel length,
3 mm. b, Room-temperature resistance versus top-gate voltage, Vg , relative to voltage at which the Dirac point is observed, Vg Dirac ≈ 24 V, for two
representative self-organized nanoribbons, red and blue. Asymmetry and mobility degradation are due to interaction with the NO2 adhesion layer. Inset,
optical micrograph of top-gated devices as studied here. Scale bar, 3 mm; gate length, 450 nm. c, Room-temperature current (red) and corresponding
conductance (blue) versus source–drain voltage for the 40-nm-wide nanoribbon device shown in Fig. 3. Constant (minimum) conductance line, dashed
black, is plotted for reference. Near-linearity is expected2 in this ungated measurement due to large intrinsic electron density. d, Gated measurements of
40-nm-wide nanoribbon devices at room temperature (inset) and 4 K.

NATURE NANOTECHNOLOGY | VOL 5 | OCTOBER 2010 | www.nature.com/naturenanotechnology 729

© 2010 Macmillan Publishers Limited. All rights reserved.

LETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2010.192

a dimension8,11,12. The Dirac point (resistance maximum) is typically

observed at Vg Dirac ≈ 24 V, corresponding to an electron density at
Vg ¼ 0 V of 3 × 1012 cm22, in agreement with numerous measure-
ments of the first graphene layer above the SiC interface1,5,6,9,12. This
indicates that the first graphene layer is modulated by the top-gate
and the graphene channel is not more than a few layers in thickness.
The gating efficiency further confirms the selectivity of the growth.
Recently devised dielectric adhesion methods are expected to result
in improved performance by reducing interaction between the
graphene channel and dielectric stack30.
Room-temperature four-terminal current versus source–drain
voltage measurement of narrow (width 40 nm) ribbons (Fig. 3) is
shown in Fig. 4c. Although a near-linear response is expected2 in
this ungated device due to large intrinsic electron density, as discussed
b 1.1
above, nonlinearity is observed. Conductance does not decrease
appreciably at 77 K. Gated measurements of similar devices at room
temperature and 4 K are shown in Fig. 4d. The on–off ratio is 10
at room temperature, and .25 at 4 K, with excellent symmetry. The
1.05 gate in this case is gold on graphene, with an inherently large
contact resistance; further details will be published at a later date.
Photolithographic processing allows fabrication of a large
number of devices at higher density. An array of top-gated graphene
G (mS ☐)

1 transistors prepared on the (0001) face of a 4 × 6 mm2 SiC chip with

SiC etch depth 100 nm (ribbon width, 250 nm) is shown in
Fig. 5a. A single device (source, drain, channel and gate, as illus-
trated in Fig. 1) occupies a 35 × 65 mm2 area, so the 0.24-cm2
G chip accommodates more than 10,000 transistors. This density
0.95
S D was limited primarily by the size of the probe tips used for electrical
measurement, but it is, to our knowledge, the highest density of gra-
phene devices achieved to date. The room-temperature gating effect
0.9 is plotted in Fig. 5b.
−20 −10 0 10 20 It should be noted that there are probably fundamental differ-
Vg (V) ences in graphene growth among the possible SiC facets, analogous
to the dramatic differences in growth speed and layer orientation
Figure 5 | Graphene transistor array with a density of 40,000 devices observed on the (0001) and (0001) faces6,16, and the (110n) facet
per cm2 prepared on SiC(0001). a, Optical micrograph. Scale bar, 100 mm. chosen here is possibly not the most desirable in terms of selectivity
b, Room-temperature ambipolar gating effect: conductance versus gate and quality of graphene produced. This is particularly true of facets
voltage. Inset, an individual FET consisting of source (S), drain (D), graphene prepared on the (0001) surface, where there is apparently more
channel and gate (G). Scale bar, 20 mm; channel length, 7 mm. freedom in facet choice. This is a topic of ongoing investigation.
These results demonstrate that graphene growth on non-tra-
Fig. 2d, and extensive electrical probing confirmed a lack of measur- ditional crystal faces is viable and useful in device fabrication, par-
able conductivity on the horizontal surfaces. As discussed above, on ticularly for the production of nanoribbons on a large scale, and
(0001) (Figs 2e–g, 3 and 4b), the selective graphene growth required fabrication of graphene transistors at a density greater than
no post-processing, and device yield was better than 90%. Metal 40,000 per cm2 represents a milestone in the development of gra-
contacts were deposited for four-terminal measurements without phene electronics. Refinement of this approach will come with
gate and two-terminal measurements with top gates. In the latter further reduction of ribbon width, development of three-dimen-
case, the graphene surface was functionalized by NO2 (ref. 29), fol- sional device structures, and optimization of dielectric and facet
lowed by atomic layer deposition (ALD) of Al2O3 and lift-off of a selection. Importantly, damage to ribbon edges by violent cutting
metal gate (see Fig. 1). processes such as O2 etching has been eliminated by a coalescence
Four-terminal measurements of prepared ribbons (Fig. 3 and of top-down and bottom-up lithographies. Pre-patterning of the
insets to Fig. 4a,b) yield sheet resistances of 180–1,000 VA21, SiC substrate is, in general, a new and promising direction in the
values typically observed in as-grown planar graphene6. Figure 4a development of epitaxial graphene electronics, as complex struc-
shows a series of conductance versus source–drain voltage curves tures and applications are readily envisaged.
taken between 77 K and 4 K. The behaviour is metallic at high
temperatures, but quantum confinement is clearly manifested in Methods
the nonlinearity observed at 4 K, indicating the presence of a Substrates were nominally on-axis research-grade semi-insulating 4H-SiC (Cree,
small bandgap, as expected of this 250-nm-wide ribbon, and in Inc). Arrays of nickel lines were defined on the (0001) or (0001) SiC crystal face by a
agreement with previous reports5. standard photolithographic lift-off process, and transferred into the SiC by a 43%
SF6/23% O2/33% Ar RIE operating at 30 mtorr. Radiofrequency power was tuned to
Top-gated two-terminal measurement of similar devices at room give a SiC etch rate of 8 Å s21, allowing fine control of the etch depth. Ultrasonic
temperature is presented in Fig. 4b. Although the NO2-functionali- treatment in nitric acid removed nickel from the SiC surface, and further cleaning
zation, one of a handful of recent techniques for enabling adhesion and graphene growth proceeded as described previously15,16. O2 RIE operating at
of high-k dielectrics to graphene, is known to dramatically degrade 100 mtorr was tuned to give a graphene etch rate of 1 Å s21, and etch time on
mobility, on–off ratio and electron–hole symmetry30, we observe (0001) faces was several to a maximum of ten seconds. Samples were mounted with
the (110n) facet parallel to ion flux. Contacts were defined by electron-beam or
large field-effect mobilities between mFE ¼ 900 and photolithographic lift-off of 5 nm palladium/60 nm gold. Atomic layer deposition
2,700 cm2 V21 s21, values that are comparable to or better than (ALD) of 39 nm Al2O3 was performed as described by Williams and colleagues29 in
those previously reported at room temperature for ribbons of this a commercial Cambridge Nanosystems Savannah ALD system before lift-off of an

730 NATURE NANOTECHNOLOGY | VOL 5 | OCTOBER 2010 | www.nature.com/naturenanotechnology

© 2010 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2010.192 LETTERS
aluminium top-gate. Al2O3 does not deposit uniformly without an adhesion layer, 16. Hass, J., de Heer, W. A. & Conrad, E. H. The growth and morphology of epitaxial
and NO2 functionalization was used in this case. The Al2O3 dielectric constant as multilayer graphene. J. Phys. Condens. Matter 20, 323202 (2008).
deposited is k ¼ 6. Raman mapping was performed with an excitation wavelength of 17. Emtsev, K. V. et al. Towards wafer-size graphene layers by atmospheric pressure
532 nm, lateral resolution of 1 mm and grid spacing of 0.2520.5 mm. 2D intensity graphitization of silicon carbide. Nature Mater. 8, 203–207 (2009).
was taken at 2D maxima near 2,725 cm21. HRTEM measurements were performed 18. Virojanadara, C. et al. Substrate orientation: a way towards higher quality
by Evans Analytical Group (Raleigh, NC, USA), with an acceleration voltage of monolayer graphene growth on 6H-SiC(0001). Surf. Sci. 603, L87–L90 (2009).
200 kV. EFM data were obtained on a Park XE-70 in enhanced non-contact EFM 19. Robinson, J. et al. Nucleation of epitaxial graphene on SiC(0001). ACS Nano 4,
mode with 2 VAC at v ¼ 17 kHz on the tip, sample floating, and v response (shown 153–158 (2010).
in Fig. 3) monitored by lock-in amplifier. Electrical transport measurements were 20. Hass, J. R. Structural Characterization of Epitaxial Graphene on Silicon Carbide.
performed at atmospheric pressure. Field-effect mobility was calculated according to PhD thesis, Georgia Institute of Technology. Preprint at http://hdl.handle.net/
mFE ¼ dG/dVg LW/C (refs 8,12), where G, L, W and C are square conductance, 1853/26654 (2008).
channel length, width and dielectric capacitance, respectively, after subtracting 21. Wu, X. et al. Half integer quantum Hall effect in high mobility single layer
contact resistance determined by four-terminal measurement of the device epitaxial graphene. Appl. Phys. Lett. 95, 223108 (2009).
(Supplementary Fig. S2). 22. Kong, H. S., Glass, J. T. & Davis, R. F. Chemical vapor deposition and
characterization of 6H-SiC thin films on off-axis 6H-SiC substrates. J. Appl. Phys.
Received 30 June 2010; accepted 31 August 2010; 64, 2672–2679 (1988).
published online 3 October 2010 23. Syväjärvi, M., Yakimova, R. & Janzn, E. Step-bunching in SiC epitaxy: anisotropy
and influence of growth temperature. J. Cryst. Growth 236, 297–304 (2002).
References 24. Nakajima, A., Yokoya, H., Furukawa, Y. & Yonezu, H. Step control of vicinal
1. Berger, C. et al. Ultrathin epitaxial graphite: 2D electron gas properties and a 6H-SiC(0001) surface by H2 etching. J. Appl. Phys. 97, 104919 (2005).
route toward graphene-based nanoelectronics. J. Phys. Chem. B 108, 25. Nie, S. et al. Step formation on hydrogen-etched 6H-SiC{0001} surfaces. Surf.
19912–19916 (2004). Sci. 602, 2936–2942 (2008).
2. Han, M. Y., Brant, J. C. & Kim, P. Electron transport in disordered graphene 26. Borovikov, V. & Zangwill, A. Step bunching of vicinal 6H-SiC{0001} surfaces.
nanoribbons. Phys. Rev. Lett. 104, 056801 (2010). Phys. Rev. B 79, 245413 (2009).
3. Tapasztó, L., Dobrik, G., Lambin, P. & Biró, L. P. Tailoring the atomic structure 27. Nakagawa, H., Tanaka, S. & Suemune, I. Self-ordering of nanofacets on vicinal
of graphene nanoribbons by scanning tunnelling microscope lithography. SiC surfaces. Phys. Rev. Lett. 91, 226107 (2003).
Nature Nanotech. 3, 397–401 (2008). 28. Tanaka, S., Morita, K. & Hibino, H. Anisotropic layer-by-layer growth of
4. Jiao, L., Zhang, L., Wang, X., Diankov, G. & Dai, H. Narrow graphene graphene on vicinal SiC(0001) surfaces. Phys. Rev. B 81, 041406 (2010).
nanoribbons from carbon nanotubes. Nature 458, 877–880 (2009). 29. Williams, J. R., DiCarlo, L. & Marcus, C. M. Quantum Hall effect in a
5. Berger, C. et al. Electronic confinement and coherence in patterned epitaxial gate-controlled p–n junction of graphene. Science 317, 638–641 (2007).
graphene. Science 312, 1191–1196 (2006). 30. Farmer, D. B. et al. Utilization of a buffered dielectric to achieve high field-effect
6. de Heer, W. A. et al. Epitaxial graphene. Solid State Commun. 143, carrier mobility in graphene transistors. Nano Lett. 9, 4474–4478 (2009).
92–100 (2007).
7. Orlita, M. et al. Approaching the dirac point in high-mobility multilayer Acknowledgements
epitaxial graphene. Phys. Rev. Lett. 101, 267601 (2008). The authors acknowledge helpful conversations with A. Zangwill, V. Borovikov and
8. Kedzierski, J. et al. Epitaxial graphene transistors on SiC substrates. IEEE Trans. F. Ming. This research was supported by the W.M. Keck Foundation, the National Science
Electron Dev. 55, 2078–2085 (2008). Foundation under grant no. DMR-0820382, and the Nanoelectronics Research Initiative
9. Li, X. et al. Top- and side-gated epitaxial graphene field effect transistors. Phys. Institute for Nanoelectronics Discovery and Exploration (INDEX).
Status Solidi A 207, 286–290 (2010).
10. Sprinkle, M. et al. First direct observation of a nearly ideal graphene band
structure. Phys. Rev. Lett. 103, 226803 (2009).
Author contributions
W.A.dH. and C.B. conceived the project, and W.A.dH., C.B., M.S. and M.R. designed the
11. Moon, J. et al. Epitaxial-graphene RF field-effect transistors on Si-face 6H-SiC
experiment. W.A.dH. supervised the project, with assistance from C.B. M.S., M.R. and C.B.
substrates. IEEE Electron Dev. Lett. 30, 650–652 (2009).
performed the experiment. Y.H., J.H. and B.Z. helped with sample preparation. M.R.R.
12. Lin, Y.-M. et al. 100-GHz transistors from wafer-scale epitaxial graphene. Science
helped with Al2O3 deposition. X.W. and J.H. assisted with low-temperature measurement.
327, 662 (2010).
M.S. and C.B. analysed the data, and M.S. wrote the paper.
13. Wu, X. S. et al. Epitaxial-graphene/graphene-oxide junction: an essential step
towards epitaxial graphene electronics. Phys. Rev. Lett. 101, 026801 (2008).
14. Bekyarova, E. et al. Chemical modification of epitaxial graphene: spontaneous Additional information
grafting of aryl groups. J. Am. Chem. Soc. 131, 1336–1337 (2009). The authors declare no competing financial interests. Supplementary information
15. Nakada, K., Fujita, M., Dresselhaus, G. & Dresselhaus, M. S. Edge state in accompanies this paper at www.nature.com/naturenanotechnology. Reprints and
graphene ribbons: nanometer size effect and edge shape dependence. Phys. Rev. permission information is available online at http://npg.nature.com/reprintsandpermissions/.
B 54, 17954–17961 (1996). Correspondence and requests for materials should be addressed to W.A.dH.

NATURE NANOTECHNOLOGY | VOL 5 | OCTOBER 2010 | www.nature.com/naturenanotechnology 731

© 2010 Macmillan Publishers Limited. All rights reserved.

 LETTERS
PUBLISHED ONLINE: 20 JUNE 2010 | DOI: 10.1038/NNANO.2010.132

Roll-to-roll production of 30-inch graphene films

for transparent electrodes
Sukang Bae1†, Hyeongkeun Kim1,3†, Youngbin Lee1, Xiangfan Xu5, Jae-Sung Park7, Yi Zheng5,
Jayakumar Balakrishnan5, Tian Lei1, Hye Ri Kim2, Young Il Song6, Young-Jin Kim1,3, Kwang S. Kim7,
Barbaros Özyilmaz5, Jong-Hyun Ahn1,4 *, Byung Hee Hong1,2 * and Sumio Iijima1,8

The outstanding electrical1, mechanical2,3 and chemical4,5 prop- There are three essential steps in the roll-to-roll transfer14–16
erties of graphene make it attractive for applications in flexible (Fig. 1a): (i) adhesion of polymer supports to the graphene on the
electronics6–8. However, efforts to make transparent conducting copper foil; (ii) etching of the copper layers; and (iii) release of
films from graphene have been hampered by the lack of effi- the graphene layers and transfer onto a target substrate. In the
cient methods for the synthesis, transfer and doping of gra- adhesion step, the graphene film, grown on a copper foil, is attached
phene at the scale and quality required for applications. Here, to a thin polymer film coated with an adhesive layer by passing
we report the roll-to-roll production and wet-chemical doping between two rollers. In the subsequent step, the copper layers are
of predominantly monolayer 30-inch graphene films grown by removed by electrochemical reaction with aqueous 0.1 M
chemical vapour deposition onto flexible copper substrates. ammonium persulphate solution (NH4)2S2O8 (ref. 17). Finally, the
The films have sheet resistances as low as ∼125 V A21 with graphene films are transferred from the polymer support onto a
97.4% optical transmittance, and exhibit the half-integer target substrate by removing the adhesive force holding the gra-
quantum Hall effect, indicating their high quality. We further phene films. When using thermal release tapes11,12, the graphene
use layer-by-layer stacking to fabricate a doped four-layer films are detached from the tapes and released to counter-substrates
film and measure its sheet resistance at values as low as by thermal treatment (Fig. 1). The third step is not necessary when
∼30 V A21 at ∼90% transparency, which is superior to com- the target substrate is directly attached to the copper foil in the first
mercial transparent electrodes such as indium tin oxides. step by permanent adhesion.
Graphene electrodes were incorporated into a fully functional Figure 2a–c presents photographs of the roll-based synthesis and
touch-screen panel device capable of withstanding high strain. transfer process. An 8-inch-wide tubular quartz reactor (Fig. 2a) is
Conventional transparent electrodes make use of indium tin used in the CVD system, allowing a monolayer graphene film to be
oxide (ITO) and are commonly used in solar cells, touch sensors synthesized on a roll of copper foil with dimensions as large as
and flat panel displays. It has a sheet resistance of less than 30 inches in the diagonal direction (Fig. 2c). A temperature gradient
100 V A21, optical transparency of 90%, and unlimited scalability. usually exists that depends on the radial position inside the tubular
In contrast, the best reported sheet resistance of graphene grown on reactor. In our preliminary work, this sometimes resulted in
nickel substrates by chemical vapour deposition (CVD) is inhomogeneous growth of the graphene on the copper foils. To
280 V A21 (refs 6,9,10), on the centimetre scale. Also, the CVD solve this problem, a 7.5-inch quartz tube wrapped with a
methods based on thermally deposited nickel thin films inevitably copper foil was inserted and suspended inside the 8-inch quartz
require a rigid substrate that can withstand temperatures close to tube. In this way, the radial inhomogeneity in the reaction tempera-
1,000 8C and an etching process to remove the metal catalyst ture could be minimized (Supplementary Fig. S1). In the first step of
layers. These are therefore the main obstacles to the direct use of synthesis, the roll of copper foil is inserted into a tubular quartz tube
graphene on as-grown substrates6,9,10. Finding a way in which to and then heated to 1,000 8C with flowing 8 s.c.c.m. H2 at 90 mtorr.
transfer graphene films onto a foreign substrate is essential11,12. After reaching 1,000 8C, the sample is annealed for 30 min without
Recently, Li and colleagues discovered and demonstrated a CVD changing the flow rate or pressure. The copper foils are heat-treated
method that used centimetre-scale copper substrates13, opening a to increase the grain size from a few micrometres to 100 mm, as
new route to large-scale production of high-quality graphene films we have found that the copper foils with larger grain size yield
for practical applications. Graphene growth is therefore no longer higher-quality graphene films, as suggested by Li and colleagues13
limited to the use of rigid substrates; instead, large flexible copper (Supplementary Fig. S3). The gas mixture of CH4 and H2 is
foils may be used in the form of a roll-type substrate fitting inside then flowed at 460 mtorr with rates of 24 and 8 s.c.c.m. for 30 min,
a tubular furnace to maximize the scale and homogeneity of the pro- respectively. Finally, the sample is rapidly cooled to room temperature
duced graphene films. The flexibility of the graphene and copper (10 8C s21) with flowing H2 under a pressure of 90 mtorr.
foils further allows efficient etching and transfer processes that After growth, the graphene film grown on copper foil is attached
use a cost- and time-effective roll-to-roll production method. to a thermal release tape (Jinsung Chemical Co. and Nitto Denko

1
SKKU Advanced Institute of Nanotechnology (SAINT) and Center for Human Interface Nano Technology (HINT), Sungkyunkwan University, Suwon
440-746, Korea, 2 Department of Chemistry, Sungkyunkwan University, Suwon 440-746, Korea, 3 School of Mechanical Engineering, Sungkyunkwan
University, Suwon 440-746, Korea, 4 School of Advanced Materials Science and Engineering, Sungkyunkwan University, Suwon 440-746, Korea, 5 NanoCore &
Department of Physics, National University of Singapore, Singapore 117576 & 117542, 6 Digital & IT Solution Division, Samsung Techwin, Seongnam
462-807, Korea, 7 Center for Superfunctional Materials, Department of Chemistry, Pohang University of Science and Technology, Hyojadong, Namgu,
Pohang 790-784, Korea, 8 Nanotube Research Center, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba 305-8565 &
Faculty of Science and Engineering, Meijo University, Nagoya 468-8502, Japan; † These authors contributed equally to this work.
* e-mail: byunghee@skku.edu; ahnj@skku.edu

574 NATURE NANOTECHNOLOGY | VOL 5 | AUGUST 2010 | www.nature.com/naturenanotechnology

© 2010 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2010.132 LETTERS
Graphene on
Polymer support polymer support Released
polymer support

Target substrate
Graphene on Cu foil Cu etchant Graphene on target

Figure 1 | Schematic of the roll-based production of graphene films grown on a copper foil. The process includes adhesion of polymer supports, copper
etching (rinsing) and dry transfer-printing on a target substrate. A wet-chemical doping can be carried out using a set-up similar to that used for etching.

a d
Stencil mask

8 inch

Screen
39 inch printer

b e

After
heating

Before
heating

c f

1st
30 2nd
inc
h

Figure 2 | Photographs of the roll-based production of graphene films. a, Copper foil wrapping around a 7.5-inch quartz tube to be inserted into an 8-inch
quartz reactor. The lower image shows the stage in which the copper foil reacts with CH4 and H2 gases at high temperatures. b, Roll-to-roll transfer of
graphene films from a thermal release tape to a PET film at 120 8C. c, A transparent ultralarge-area graphene film transferred on a 35-inch PET sheet.
d, Screen printing process of silver paste electrodes on graphene/PET film. The inset shows 3.1-inch graphene/PET panels patterned with silver electrodes
before assembly. e, An assembled graphene/PET touch panel showing outstanding flexibility. f, A graphene-based touch-screen panel connected to a
computer with control software. For a movie of its operation see Supplementary Information.

Co.) by applying soft pressure (0.2 MPa) between two rollers. from the tape to the target substrate (Fig. 2b). By repeating
After etching the copper foil in a plastic bath filled with copper these steps on the same substrate, multilayered graphene films
etchant, the transferred graphene film on the tape is rinsed with can be prepared that exhibit enhanced electrical and optical
deionized water to remove residual etchant, and is then ready to properties, as demonstrated by Li and colleagues using wet-
be transferred to any kind of flat or curved surface on demand. transfer methods at the centimetre scale19. Figure 2c shows the
The graphene film on the thermal release tape is inserted between 30-inch multilayer graphene film transferred to a roll of 188-mm-
the rollers together with a target substrate and exposed to thick polyethylene terephthalate (PET) substrate. Figure 2d shows
mild heat (90–120 8C), achieving a transfer rate of 150– a screen-printing process used to fabricate four-wire touch-screen
200 mm min21 and resulting in the transfer of the graphene films panels18 based on graphene/PET transparent conducting films

NATURE NANOTECHNOLOGY | VOL 5 | AUGUST 2010 | www.nature.com/naturenanotechnology 575

© 2010 Macmillan Publishers Limited. All rights reserved.

LETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2010.132

a 20,000 b 100
No. of layers (Tr at 550 nm) = 1 (97.4%)
3 2 (95.1%)
3
4 2D 95 3 (92.9%)
2
15,000 2 4 (90.1%)
1

Transmittance (%)
1
Intensity (a.u.)
10 μm 90
0 0
100
1
10,000 85

Transmittance (%)
G 2
95
80 Before HNO3
After HNO3 3
5,000
D No. of layers = 4 90
3 75 4
2 200 400 600 800
1 Wavelength (nm)
0 70
1,500 2,000 2,500 3,000 200 400 600 800 1,000
Raman shift (cm−1) Wavelength (nm)

c 6,000 d 1,600 Number of layers

Before HNO3
2,678 1 2 3 4 C1s
After HNO3 0.20
ΔEF = ~130 meV ~0.12 eV
5,000 4,000 2D 0.15
Intensity (a.u)

ΔΦ (eV)
No D peaks 0.10
2,695 1,200
4,000 2,000
0.05

Intensity (a.u.)
Intensity (a.u)

0.00
0
3,000 1,200 1,300 1,400 1,500 0 60 120 180 240 300 360
Raman shift (cm−1) Doping time (s)
1,582 800
~0.83 eV
2,000 G 1,600
After doping

1,000 400 Before doping

C=O C−O C−C
0
1,520 1,560 1,600 1,640 2,560 2,640 2,720 2,800 292 290 288 286 284 282
Raman shift (cm−1) Binding energy (eV)

Figure 3 | Optical characterizations of the graphene films prepared using layer-by-layer transfer on SiO2/silicon and PET substrates. a, Raman spectra of
graphene films with different numbers of stacked layers. The left inset shows a photograph of transferred graphene layers on a 4-inch SiO2(300 nm)/silicon
wafer. The right inset is a typical optical microscope image of the monolayer graphene, showing .95% monolayer coverage. A PMMA-assisted transfer
method is used for this sample. b, UV–vis spectra of roll-to-roll layer-by-layer transferred graphene films on quartz substrates. The inset shows the UV
spectra of graphene films with and without HNO3 doping. The right inset shows optical images for the corresponding number of transferred layers
(1 × 1 cm2). The contrast is enhanced for clarity. c, Raman spectra of HNO3-doped graphene films, showing 18 cm21 blueshift both for G and 2D peaks.
D-band peaks are not observed before or after doping, indicating that HNO3 treatment is not destructive to the chemical bonds of graphene. d, XPS peaks of
monolayer graphene films transferred on SiO2/Si substrates, showing typical redshift and broadening of carbon 1s peaks (C1s) caused by p-doping. The inset
shows work-function changes (DF) with respect to doping time (lower x-axis) and number of stacked layers (upper x-axis), measure by UPS.

(Supplementary Fig. S12). After printing electrodes and dot 2.2–2.3% for an additional transfer, implying that the average
spacers, the upper and lower panels are carefully assembled thickness is approximately a monolayer (Fig. 3b)23.
and connected to a controller installed in a laptop computer The unique electronic band structure of graphene allows modu-
(Fig. 2e–f ), which shows extraordinary flexibility as we will lation of the charge carrier concentrations in dependence on an
discuss later (Fig. 4d). The scalability and processability of CVD electric field induced by gate bias24 or chemical doping25, resulting
graphene and the roll-to-roll methods presented here are expected in enhancement of sheet resistance. We tried various types of
to enable the continuous production of graphene-based electronic chemical doping methods, and found that nitric acid (HNO3) is
devices at large scales. very effective for p-doping of graphene films (Supplementary
The graphene films seem to be predominantly composed of Fig. S10). Figure 3c shows Raman spectra of the graphene films
monolayers when analysed using Raman spectra (Fig. 3a). before and after doping with 63 wt% HNO3 for 5 min. The large
However, atomic force microscope (AFM) and transmission elec- peak shift (Dn ¼ 18 cm21) indicates that the graphene film is
tron microscope (TEM) images often show bilayer and multilayer strongly p-doped. The shifted G peak is often split near the ran-
islands (Supplementary Figs S5,S6). As the graphene layers are domly stacked bilayer islands, as shown in Fig. 3c. We hypothesize
transferred one after another19, the intensities of the G- and 2D- that the lower graphene layer, which is screened by top layers,
band peaks increase together, but their ratios do not change signifi- experiences a reduced doping effect, leading to G-band splitting.
cantly. This is because the hexagonal lattices of the upper and lower In X-ray photoelectron spectra (XPS), the C1s peaks corresponding
layers are randomly oriented, unlike in graphite, so the original to sp 2 and sp 3 hybridized states are shifted to lower energy, similar to
properties of each monolayer remain unchanged, even after stacking the case for p-doped carbon nanotubes25. However, multilayer
into multilayers20,21; this is clearly different from the case of multi- stacking results in blueshifted C1s peaks. We suppose that weak
layer graphene exfoliated from graphite crystals22. The randomly chemical bonding such as the p–p stacking interaction causes
stacked layers behave independently without significant change in descreening of nucleus charges, leading to an overall increase in
the electronic band structures, and the overall conductivity of the core electron binding energies. We also find that the work functions
graphene films appears to be proportional to the number of of graphene films as estimated by UV photoelectron spectroscopy
stacked layers19. The optical transmittance is usually reduced by (UPS) are blueshifted by 130 meV with increasing doping time

576 NATURE NANOTECHNOLOGY | VOL 5 | AUGUST 2010 | www.nature.com/naturenanotechnology

© 2010 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2010.132 LETTERS
a b 104
300 Roll-to-roll transfer This work Ref. 6
Ref. 19 e
Wet transfer with PMMA This work hen
Theory Ref. 25 g rap
ed
250 Roll-to-roll + HNO3 doping err )
Ref. 28 nsf 9
tra (ref. 1
Sheet resistance (Ω☐−1)

Sheet resistance (Ω☐−1)

103 et-
W
200 n Ni
wn o
e gro
Gra phen s e
otube en
n nan ph
150 102 Carbo gra
R
R2 ing
dop
) e+
100 ula tion p hen
(calc gra
ITO R2R
101 tion)
50 e (calcula
Graphen

0
100
1 2 3 4 70 75 80 85 90 95 100
Number of layers Transmittance (%)

c 30
d 200
12 ITO (ref. 30)
T = 295 K
10 T=6K 3 R2R graphene
25 Tensile
8
R (kΩ)

ΔR/Ro
6 2 Compressive
150
Magnetoresistance (kΩ)

20
4
1
15 2
0 0
−20 0 20 40

ΔR/Ro
10 100 0 2 4
Vbg (V)
Strain (%)
5
0
Rxx 50
−5
Rxy
−10
0
−15
−30 −20 −10 0 10 20 30 40 1 2 3 4 5 6
Vbg (V) Strain (%)

Figure 4 | Electrical characterizations of layer-by-layer transferred and HNO3-doped graphene films. a, Sheet resistances of transferred graphene films
using a roll-to-roll (R2R) dry-transfer method combined with thermal release tapes and a PMMA-assisted wet-transfer method. b, Comparison of sheet
resistance from this research and transmittance plots taken from other references. The dashed arrows indicate the expected sheet resistances at lower
transmittance. The scheme is borrowed from ref. 19. c, Electrical properties of a monolayer graphene Hall bar device in vacuum. Four-probe resistivity (left
bottom inset) is measured as a function of gate voltage in the monolayer graphene Hall bar shown in the right inset at room temperature (black curve) and
T ¼ 6 K (red curve). The QHE effect at T ¼ 6 K and B ¼ 9 T is measured in the same device. The longitudinal resistivity rxx and Hall conductivity sxy are
plotted as a function of gate voltage. The sequence of the first three half-integer plateaus corresponding to n ¼ 2, 6 and 10, typical for single-layer graphene,
are clearly seen. The Hall effect mobility of this device is mHall ¼ 7,350 cm22 V21 s21 at 6 K (5,100 cm22 V21s21 at 295 K). Scale bar (inset), 3 mm.
d, Electromechanical properties of graphene-based touch-screen devices compared with ITO/PET electrodes under tensile strain. The inset shows the
resistance change with compressive and tensile strain applied to the upper and lower graphene/PET panels, respectively.

(Fig. 3d, inset). The multiple stacking also changes the work func- increase in the overall sheet resistance. Because additional layers
tions (Fig. 3d, inset; see also Supplementary Fig. S7 and Table S1), are not directly affected by the adhesion with the substrate
which could be very important in controlling the efficiency of surface, the sheet resistance of multilayers prepared by the roll-to-
photovoltaic26 or light-emitting devices based on graphene trans- roll method does not differ much from that for the wet-transfer
parent electrodes27. case. The p-doping with HNO3 clearly enhances the electrical prop-
The electrical properties of graphene films formed using layer- erties of graphene films, and is more effective in roll-to-roll pro-
by-layer stacking methods were also investigated. Usually, the cesses. The sheet resistance of the p-doped four-layer graphene
sheet resistance of graphene film with 97.4% transmittance is as film with 90% optical transmittance is as low as 30 V A21,
low as 125 V A21 (Fig. 4a) when it is transferred by a soluble which is superior to common transparent electrodes such as ITO
polymer support such as polymethyl methacrylate (PMMA)9,10,19. and carbon-nanotube films (Fig. 4b)28.
The transferrable size achievable using wet-transfer methods is Standard electron-beam lithography has been used to fabricate
limited to less than a few inches of wafer because of the weak mech- graphene Hall bars on conventional 300-nm SiO2/Si substrates
anical strength of spin-coated PMMA layers. However, the scale of (Fig. 4c). The left inset of Fig. 4c shows the four-terminal resistance
roll-to-roll dry transfer assisted by a thermal release tape is in prin- of such samples as a function of backgate voltage (Vbg) at both room
ciple unlimited. In the process of roll-to-roll dry transfer, the first temperature (in black) and at a low temperature (T ¼ 6 K) and zero
layer sometimes shows approximately two to three times larger magnetic field. We observe the graphene specific gate bias depen-
sheet resistance than that of the PMMA-assisted wet-transfer dence of the resistance with a sharp Dirac peak and an effective
method. As the number of layers increases, the resistance drops Hall mobility of 7,350 cm2 V21 s21 at low temperatures. This
faster compared to the wet-transfer method (Fig. 4a). We postulate allows the observation of the quantum Hall effect (QHE)29 at 6 K
that the adhesion of the first layer with the substrate is not strong and a magnetic field of B ¼ 9 T (Fig. 4c, right). The fingerprint of
enough for complete separation of the graphene films from the single-layer graphene, the half-integer quantum Hall effect, is
thermal release tapes (Supplementary Fig. S4). As a result, there observed with plateaus at filling factors of n ¼ 2, 6 and 10 at
can be mechanical damage on the graphene films, leading to an Rxy = 12 , 16 and 101 (h/e 2), respectively. Although the sequence of

NATURE NANOTECHNOLOGY | VOL 5 | AUGUST 2010 | www.nature.com/naturenanotechnology 577

© 2010 Macmillan Publishers Limited. All rights reserved.

LETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2010.132

the plateaus remains for both the electron side and the hole side, 17. Jo, G. et al. Etching solution for etching Cu and Cu/Ti metal layer of liquid
there is a slight deviation from the fully quantized values on the crystal display device and method of fabricating the same. US patent,
6,881,679 (2005).
hole side. 18. Hecht, D. S. et al. Carbon nanotube film on plastic as transparent electrode for
Finally, the electromechanical properties of graphene/PET resistive touch screens. J. Soc. Inf. Display 17, 941–946 (2009).
touch-screen panels were tested (Fig. 4d; Supplementary Fig. S2). 19. Li, X. et al. Transfer of large-area graphene films for high-performance
Unlike an ITO-based touch panel, which easily breaks under transparent conductive electrodes. Nano Lett. 9, 4359–4363 (2009).
2–3% strain, the graphene-based panel resists up to 6% strain; 20. Hass, J. et al. Why multilayer graphene on 4H-SiC(000–1) behaves like a single
sheet of graphene. Phys. Rev. Lett. 100, 125504 (2008).
this is limited not by the graphene itself, but by the printed silver 21. Sprinkle, M. et al. First direct observation of a nearly ideal graphene band
electrodes (Fig. 4d)30. structure. Phys. Rev. Lett. 103, 226803 (2009).
In summary, we have developed and demonstrated the roll-to- 22. Ferrari, A. C. et al. Raman spectrum of graphene and graphene layers. Phys. Rev.
roll production of graphene on ultralarge copper substrates. Lett. 97, 187401 (2006).
The multiple transfer and simple chemical doping of graphene 23. Nair, R. R. et al. Fine structure constant defines visual transparency of graphene.
Science 320, 1308 (2008).
films considerably enhances its electrical and optical 24. Das, A. et al. Monitoring dopants by Raman scattering in an electrochemically
properties. Given the scalability and processability of roll-to-roll top-gated graphene transistor. Nature Nanotech. 3, 210–215 (2008).
and CVD methods and the flexibility and conductivity of graphene 25. Geng, H.-Z. et al. Effect of acid treatment on carbon nanotube-based flexible
films, we anticipate that the commercial production of large-scale transparent conducting films. J. Am. Chem. Soc. 129, 7758–7759 (2007).
26. Schrivera, M., Reganb, W., Losterb, M. & Zettl, A. Carbon nanostructure–aSi:H
transparent electrodes, replacing ITO, will be realized in the
photovoltaic cells with high open-circuit voltage fabricated without dopants.
near future. Solid State Commun. 150, 561–563 (2010).
27. Wu, J. et al. Organic light-emitting diodes on solution-processed graphene
Received 11 March 2010; accepted 28 May 2010; transparent electrodes. ACS Nano 4, 43–48 (2010).
published online 20 June 2010; corrected online 25 June 2010 28. Lee, J.-Y., Connor, S. T., Cui, Y. & Peumans, P. Solution-processed metal
nanowire mesh transparent electrodes. Nano Lett. 8, 689–692 (2008).
References 29. Cao, H. L. et al. Electronic transport in chemical vapor deposited graphene
1. Geim, A. K. & Novoselov, K. S. The rise of graphene. Nature Mater. 6, synthesized on Cu: Quantum Hall effect and weak localization. Appl. Phys. Lett.
183–191 (2007). 96, 122106 (2010).
2. Lee, C., Wei, X., Kysar, J. W. & Hone, J. Measurement of the elastic properties 30. Cairns, D. R. et al. Strain-dependent electrical resistance of tin-doped indium
and intrinsic strength of monolayer graphene. Science 321, 385–388 (2008). oxide on polymer substrates. Appl. Phys. Lett. 76, 1425–1427 (2000).
3. Bunch, J. S. et al. Impermeable atomic membranes from graphene sheets. Nano
Lett. 8, 2458–2462 (2008).
4. Elias, D. C. et al. Control of graphene’s properties by reversible hydrogenation: Acknowledgements
evidence for graphene. Science 323, 610–613 (2009). This work was supported by the National Research Foundation of Korea (NRF), funded
5. Wang, X. et al. N-doping of graphene through electrothermal reactions with by the Ministry of Education, Science and Technology (2009-0081966, 2009-0082608,
ammonia. Science 324, 768–771 (2009). 2009-0083540, 2009-0090017, World Class University R33-2008-000-10138-0, National
6. Kim, K. S. et al. Large-scale pattern growth of graphene films for stretchable Honor Scientist Program), the Research Centre of Breakthrough Technology Program
transparent electrodes. Nature 457, 706–710 (2009). through the Korea Institute of Energy Technology Evaluation and Planning (KETEP),
7. Kim, D.-H. et al. Stretchable and foldable silicon integrated circuits. Science 320, funded by the Ministry of Knowledge Economy (2009-3021010030-11-1), Singapore
507–511 (2008). National Research Foundation (NRF-RF2008-07) & NUS NanoCore, and T.J. Park Junior
8. Sekitani, T. et al. A rubberlike stretchable active matrix using elastic conductors. Faculty Fellowship. The authors thank R. Ruoff (University of Texas at Austin) and P. Kim
Science 321, 1468–1472 (2008). (Columbia University) for helpful comments, W.S. Lim, K.D. Kim and Y.D. Kim (SKKU)
9. Reina, A. et al. Large area, few-layer graphene films on arbitrary substrates by for assistance in XPS analysis, and Samkwang Well Tech Co. for assistance with the touch-
chemical vapor deposition. Nano Lett. 9, 30–35 (2009). panel fabrication.
10. Cai, W. W. et al. Large area few-layer graphene/graphite films as transparent
thin conducting electrodes. Appl. Phys. Lett. 95, 123115 (2009). Author contributions
11. Lee, Y. et al. Wafer-scale synthesis and transfer of graphene films. Nano Lett. 10, B.H.H. planned and supervised the project, with assistance in supervision from J.H.A.
490–493 (2010). Y.-J.K., B.O., K.S.K. and S.I. provided advice for the project. B.H.H., S.B. and H.K. conceived
12. Caldwell, J. D. et al. Technique for the dry transfer of epitaxial graphene onto and carried out the experiment. B.H.H., J.H.A. and B.O. analysed the data and wrote the
arbitrary substrates. ACS Nano 4, 1108–1114 (2010). manuscript. X.X., J.B., Y.Z. and B.O. fabricated the QHE devices, and carried out the
13. Li, X. et al. Large-area synthesis of high-quality and uniform graphene films on measurements. Y.L. and Y.I.S. helped with the fabrication of touch-screen panels and
copper foils. Science 324, 1312–1314 (2009). electromechanical analysis. J.S.P., H.R.K. and T.L. helped with the doping experiment.
14. Ahn, S. H. & Guo, L. J. High-speed roll-to-roll nanoimprint lithography on
flexible plastic substrates. Adv. Mater. 20, 2044–2049 (2008).
15. Yerushalmi, R., Jacobson, Z. A., Ho, J. C., Fan, Z. & Javey, A. Large scale, highly Additional information
ordered assembly of nanowire parallel arrays by differential roll printing. Appl. The authors declare no competing financial interests. Supplementary information
Phys. Lett. 91, 203104 (2007). accompanies this paper at www.nature.com/naturenanotechnology. Reprints and
16. Chang, Y. K. & Hong, F. C. The fabrication of ZnO nanowire field-effect permission information is available online at http://npg.nature.com/reprintsandpermissions/.
transistors by roll-transfer printing. Nanotechnology 20, 195302 (2009). Correspondence and requests for materials should be addressed to B.H.H. and J.H.A.

578 NATURE NANOTECHNOLOGY | VOL 5 | AUGUST 2010 | www.nature.com/naturenanotechnology

© 2010 Macmillan Publishers Limited. All rights reserved.

 LETTERS
PUBLISHED ONLINE: 14 FEBRUARY 2010 | DOI: 10.1038/NNANO.2010.8

Graphene nanomesh
Jingwei Bai1, Xing Zhong2, Shan Jiang2, Yu Huang1,3 * and Xiangfeng Duan2,3 *

Graphene has significant potential for application in elec- the molecular weight while keeping the volume ratio constant can
tronics1–5, but cannot be used for effective field-effect transis- result in vertical aligned domains with variable centre-to-centre dis-
tors operating at room temperature because it is a semimetal tance (Supplementary Fig. S1)27. After UV exposure and glacial acid
with a zero bandgap6,7. Processing graphene sheets into nano- development to remove PMMA domains, a porous structure of the
ribbons with widths of less than 10 nm can open up a bandgap PS matrix, with hexagonally arranged nanoholes vertically penetrat-
that is large enough for room-temperature transistor oper- ing through the film, is obtained (Fig. 2b). The subsequent etching
ation8–19, but nanoribbon devices often have low driving cur- process drills into the underlying SiOx layer, leaving a SiOx nano-
rents or transconductances18,19. Moreover, practical devices mesh that can serve as the mask for the following oxygen plasma
and circuits will require the production of dense arrays of etch to form GNMs (Fig. 2c). The oxide mask can be easily
ordered nanoribbons, which remains a significant chal- removed by briefly dipping the sample into HF solution, and the
lenge20,21. Here, we report the production of a new graphene GNM structure can be observed using scanning electron microscopy
nanostructure—which we call a graphene nanomesh—that can (SEM; Fig. 2d).
open up a bandgap in a large sheet of graphene to create a
semiconducting thin film. The nanomeshes are prepared a b
using block copolymer lithography and can have variable Block copolymer film
periodicities and neck widths as low as 5 nm. Graphene nano- Graphene on substrate
mesh field-effect transistors can support currents nearly 100
times greater than individual graphene nanoribbon devices,
and the on–off ratio, which is comparable with the values Anneal, develop
achieved in individual nanoribbon devices, can be tuned by d c
varying the neck width. The block copolymer lithography SiO2 nanomesh mask RIE Porous PS film
approach used to make the nanomesh devices is intrinsically with PS residue film
scalable and could allow for the rational design and fabrication
of graphene-based devices and circuits with standard
semiconductor processing. O2 plasma
We report a graphene nanomesh (GNM) comprising single- or e f
few-layer graphene into which a high-density array of nanoscale HF dip
Graphene nanomesh
holes are punched (Fig. 1). For convenience, we define two critical
structural parameters in a GNM: the ‘periodicity’, defined as the
centre-to-centre distance between two neighbouring nanoholes,
Lift off
and the ‘neck width’, defined as the smallest edge-to-edge distance
between two neighbouring nanoholes in the nanomesh. g
Figure 1 illustrates the present approach to fabricating GNMs.
For initial demonstrations, mechanically peeled graphene flakes
were used as the starting material, although the approach described
here can be readily extended to graphene films obtained through
chemical exfoliation or chemical vapour deposition22–25. A 10-nm-
thick silicon oxide (SiOx) film is first evaporated onto graphene as
the protecting layer and also as the grafting substrate for the sub- Figure 1 | Schematic of fabrication of a graphene nanomesh. a, Pristine
sequent block copolymer nanopatterning26. The poly(styrene- graphene flake on top of a silicon oxide substrate, forming the starting
block-methyl methacrylate) (P(S-b-MMA)) block copolymer thin material. b, The graphene flake is covered by a thin layer of evaporated SiOx
film with cylindrical domains normal to the surface is then fabricated and a thin film of spin-coated block-copolymer P(S-b-MMA). The SiOx is
and used as the etching template26, and a CHF3-based reactive-ion used as a protecting layer and grafting substrate for the subsequent block-
etch (RIE) process followed by oxygen plasma etch is used to copolymer nanopatterning process. c, The P(S-b-MMA) block-copolymer
punch holes into the graphene layer (see Methods for details). film is annealed and developed, leaving the porous PS matrix as the
An atomic force microscopy (AFM) image of the annealed nanomesh template for further patterning. d, Fluoride-based reactive ion
P(S-b-MMA) thin film shows the cylindrical poly(methyl methacry- etching (RIE) to penetrate the oxide layer, partially degrade the PS film, and
late) (PMMA) domains packed hexagonally within the polystyrene form the SiOx nanomesh hard mask. e, Graphene in the exposed area is
(PS) matrix (Fig. 2a). The domain centre-to-centre distance is found etched away by O2 plasma. f, Following an HF dip to remove the oxide mask,
to be 39 nm when using P(S-b-MMA) with a molecular weight of a GNM is obtained. g, Free-standing GNM can be lifted off the substrate by
77,000 g mol21 and a PS:PMMA volume ratio of 70:30. Changing etching away the underlying silicon oxide.

1
Department of Materials Science and Engineering, University of California, Los Angeles, California 90095, USA, 2 Department of Chemistry and
Biochemistry, University of California, Los Angeles, California 90095, USA, 3 California Nanosystems Institute, University of California, Los Angeles,
California 90095, USA. * e-mail: yhuang@seas.ucla.edu; xduan@chem.ucla.edu

190 NATURE NANOTECHNOLOGY | VOL 5 | MARCH 2010 | www.nature.com/naturenanotechnology

© 2010 Macmillan Publishers Limited. All rights reserved.
NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2010.8 LETTERS
To unambiguously determine the GNM structure, we carried out a b
extensive transmission electron microscopy (TEM) studies (see
Methods for details). Figure 3a–h shows a series of TEM images
of GNMs with variable periodicities or neck widths. For better
image contrast, most TEM images are taken for GNMs with a rela-
tively large thickness (5–10 nm), although images of GNMs with
one or few layers of graphene have also been obtained, but with typi-
cally lower contrast (Supplementary Fig. S2). TEM studies clearly
show the GNM structure with nearly hexagonally arranged nano-
holes, consistent with the SEM studies. Figure 3a,e shows TEM
images of a GNM obtained from a copolymer mask, with molecular
weight of 77,000 g mol21, a nanohole periodicity of 39 nm, and c d
average neck width of 14.6 nm. The nanoholes observed here are
slightly larger than the PMMA domains in the self-assembled
block copolymer pattern due to slight over-etching in the fabrica-
tion process. Importantly, the nanomesh structure can be readily
observed over many micrometres and is apparently only limited
by the size of the starting graphene flakes (Supplementary Fig. S3).
Our fabrication process shows great versatility in controlling
both the neck width and the periodicity, independently. For
example, the neck width can be tuned through controlled over-
etching during the fabrication process. Figure 3b,f shows TEM
images of a nanomesh with a periodicity of 39 nm, but smaller
average neck with of 11.2 nm, achieved through controlled lateral Figure 2 | Images illustrating the steps of the nanomesh fabrication
over-etching of the PS mask. Figure 3c,g shows a nanomesh with process. a, AFM phase contrast image of the annealed block-copolymer film
an average neck width of 7.1 nm, obtained through additional on graphene, showing hexagonal-packed PMMA domains in the PS matrix.
over-etching. Furthermore, it is also possible to tune the mesh b, SEM image of a porous PS film obtained by selectively removing the
periodicity and neck width by using block copolymers of different PMMA domains. c, SEM image of the SiOx nanomesh mask after reactive
molecular weights27. Figure 3d,h shows a much denser GNM with ion etching with the PS mask. d, SEM image of a GNM structure after
smaller periodicity of 27 nm and average neck width of 9.3 nm, removing the top SiOx mesh mask. Scale bars, 100 nm.
obtained using smaller-molecular-weight block-copolymer thin
film as the mask template. pads as the source–drain contacts, a highly doped p-type silicon
As the GNM neck width represents the smallest dimension that substrate as the back gate, and a 300-nm thermal oxide as the gate
controls charge transport through the system, we carried out statisti- dielectric (Fig. 4a,b). Figure 4c,d shows the electrical transport
cal analysis of the neck widths of the GNMs obtained with variable characteristics of a typical GNM transistor with estimated neck
etching conditions or different block-copolymer masks (Fig. 3i–l). width of 10 nm. Drain current (Id) versus drain–source voltage
The histograms resulting from the statistical analysis show that a (Vd) characteristics at various gate voltages (Vg) for the GNM tran-
series of GNMs with variable neck widths of 14.6, 11.2, 7.1 and sistor show typical p-channel transistor behaviour (Fig. 4c). The
9.3 nm has been achieved. The standard deviation of the GNM hole doping observed in the GNM is similar to that of GNR
neck width is, in general, less than 2 nm. The smallest neck width devices, and can be attributed to edge oxidation in the O2 plasma
observed so far is 5 nm; this was achieved through even more process or physisorbed oxygen from the ambient and other
aggressive over-etching (Supplementary Fig. S4). In this deep nano- species during the sample preparation steps. Plots of Id versus Vg
metre regime, this aggressive over-etching may lead to some necks (Fig. 4d) at a constant Vd ¼ 210, 2100 and 2500 mV show little
in the nanomesh breaking due to a finite standard deviation, and current when the Vg is more positive than a threshold voltage
the yield can be limited. However, the as yet incomplete optimiz- (Vth) of 20.6 V, and Id increases nearly linearly when Vg increases
ation of the block copolymer self-assembly process28,29 could lead in the negative direction. The plot shows the device has an on-
to a more uniform nanomesh that will afford more aggressive state conductance as high as 550 mS at Vg ¼ 210 V, which is com-
over-etching to achieve even smaller neck widths. These TEM parable to an array of 50–100 parallel GNR devices18,19. The Id–Vg
studies clearly demonstrate that highly uniform GNMs can be plots also show the device has an on–off ratio greater than 10, which
obtained with controllable periodicities and neck widths using is comparable to GNR field-effect transistors (FETs) with widths in
block copolymer nanolithography. the range 10–15 nm (refs 18,19). Importantly, the on–off ratio of the
This ability to control nanomesh periodicity and neck width is GNM devices can be readily tuned by varying the neck width.
very important for controlling electronic properties, because Additional studies show a smaller on–off ratio of 6 for a GNM
charge transport properties are highly dependent on the width of device with a larger neck width (15 nm), and a higher on–off
the critical current pathway. In the case of graphene nanoribbons ratio of more than 100 for a GNM device with a smaller neck
(GNRs), both theoretical and experimental works have shown that width (7 nm; Fig. 4e, Supplementary Fig. S5). This observation
the size of the conduction bandgap is inversely proportional to is consistent with GNR devices, in which the conduction bandgap
the ribbon width8–19. GNMs can be viewed as a highly intercon- and on–off ratio are inversely proportional to the width of the criti-
nected network of GNRs. Therefore, a narrow neck width is cal current pathway. Similar to GNRs, the opening of the conduc-
needed to gain enough bandgap for sufficient gate response and tion bandgap in our GNM structure can be attributed to a
on–off ratio, and a denser mesh structure can enable higher combination of multiple factors, including lateral quantum confine-
current delivery. ment8–10 and a localization effect resulting from edge disorder such
To investigate the electronic properties of the GNMs, we fabri- as variable edge roughness or absorbed species in the oxygen plasma
cated GNM-based transistors and carried out electrical transport etching process11,12. Determining the exact origin of the bandgap
studies. A three-terminal device was made using the GNM as the and investigation of the impact of exact edge termination will be
semiconducting channel, electron-beam evaporated titanium/gold an interesting topic in future studies. We believe the on–off ratio

NATURE NANOTECHNOLOGY | VOL 5 | MARCH 2010 | www.nature.com/naturenanotechnology 191

© 2010 Macmillan Publishers Limited. All rights reserved.
LETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2010.8

a b c d

e f g h

i j 25 k l
25 30
20 30
20
Counts

Counts

Counts
Counts

15 20
15 20
10 10
10 10
5 5
0 0 0 0
5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20
Neck width (nm) Neck width (nm) Neck width (nm) Neck width (nm)

Figure 3 | TEM studies of graphene and thin-layer graphite nanomesh. a,e, TEM images of GNMs with a periodicity of 39 nm and neck width of 14.6 nm
obtained with block copolymer P(S-b-MMA) of molecular weight 77,000 g mol21. b,f, GNM with periodicity of 39 nm and neck width of 11.2 nm obtained
with controlled over-etching. c,g, GNM with a periodicity of 39 nm and neck width of 7.1 nm, obtained with additional over-etching. d,h, GNM with periodicity
of 27 nm and neck width of 9.3 nm, obtained with block copolymer P(S-b-MMA) of a smaller molecular weight of 47,700 g mol21. The dark strips in the
background of the images originate from the lacey carbon on the TEM grids. Scale bars, 200 nm (a–d) and 100 nm (e–h). i–l, Histograms of the GNM neck
width for e–h with average neck widths of 14.6, 11.2, 7.1 and 9.3 nm, and standard deviations of 2.0, 1.9, 1.5 and 1.3 nm, respectively.

a b
Source

GNM

Drain

c 0.0 d e −10−4
−0.5 V
−10−4 −0.1 V
−0.01 V
−10−5
−0.1 4V
Id (mA)

Id (A)

Id (A)

2V
−10−5
0V
−2 V −10−6 ~15 nm
−0.2 −4 V ~10 nm
−6 V ~7 nm
−10−6
−8 V −10−7
−10 V
−0.3
−0.4 −0.2 0.0 −10 0 10 −10 0 10
Vd (V) Vg (V) Vg (V)

Figure 4 | Room-temperature electrical properties of a graphene nanomesh device. a, Schematic of a GNM-FET. The device is fabricated on a heavily doped
silicon substrate with 300-nm SiO2 as the gate dielectric. The electronic measurement was carried out in ambient conditions at room temperature, without
removing the top oxide layer. b, SEM image of a GNM device made from nanomesh with a periodicity 39 nm and neck width of 10 nm. Scale bar,
500 nm. c, Drain current (Id) versus drain–source voltage (Vd), recorded at different gate voltages for a GNM device with a channel width of 2 mm and
channel length of 1 mm. The on-state conductance at Vg ¼ 210 V is comparable to an array of 100 parallel GNR devices. d, Transfer characteristics for the
device in c at Vd ¼ 210 mV, 2100 mV and 2500 mV. The ratio between Ion and Ioff for this device is 14 at Vd ¼ 2100 mV. e, Transfer characteristics at
Vd ¼ 2100 mV for GNMs with different estimated neck widths of 15 nm (device channel width 6.5 mm and length 3.6 mm), 10 nm (channel width 2 mm
and length 1 mm) and 7 nm (channel width 3 mm and length 2.3 mm).

192 NATURE NANOTECHNOLOGY | VOL 5 | MARCH 2010 | www.nature.com/naturenanotechnology

© 2010 Macmillan Publishers Limited. All rights reserved.
NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2010.8 LETTERS
can be further increased on additional shrinking of the GNM neck copolymer. Controlled over-etching here could be used to enlarge the nanoholes in
width. These studies clearly demonstrate that the formation of the PS nanomesh. CHF3 plasma (50 W, 6 mtorr) was then used to punch holes into
the evaporated SiOx to expose the underlying graphene layer. Additional O2 plasma
GNMs can effectively enable a semiconductor thin film with was used to completely etch away an exposed region of graphene or to further
neck-width-tunable electronic properties. undercut the graphene nanomesh.
It is important to compare the performance of GNM devices
with those of bulk graphene or GNRs. The on-state conductivity Microscopic characterization. TEM samples were prepared by spin coating a
for 10-nm and 7-nm GNM devices is 1–2 orders of magnitude 300-nm-thick layer of PMMA polymer resist onto the graphene mesh substrate,
lower than that of bulk graphene5, but comparable to that of which was then baked at 100 8C. The PMMA–graphene mesh film was then lifted off
in HF solution and transferred onto the lacey film coated copper grid. After removal
GNRs of similar width18,19, although the total drive current can be of the PMMA with acetone vapour, the sample was characterized by a JEOL 1200
50–100 times greater than the GNR devices given the larger TEM operating at 80 kV. Tapping-mode AFM was carried out with a Veeco 5000
device width of the GNM device. The on–off ratio for our GNM system. SEM imaging was performed on a JEOL 6700F unit operated at 5 kV.
device is comparable to that of similar-width GNRs18,19 with the
highest on–off ratio observed in our 7-nm graphene nanomesh, Received 21 September 2009; accepted 12 January 2010;
5–6 times larger than the largest on–off ratios reported in bulk published online 14 February 2010
graphene5. It should be noted that it is usually much more challen-
ging to obtain a high on–off ratio (.100) than a medium on–off References
ratio (10) in a GNM device while maintaining a high on-state con- 1. Novoselov, K. S. et al. Electric field effect in atomically thin carbon films. Science
ductivity, because the high on–off ratio devices require a reduction 306, 666–669 (2004).
in the neck width through aggressive over-etching, which can lead to 2. Scott Bunch, J., Yaish, Y., Brink, M., Bolotin, K. & McEuen, P. L. Coulomb
oscillations and Hall effect in quasi-2D graphite quantum dots. Nano Lett. 5,
partial breakage of the GNM current pathway. It is also important to
287–290 (2005).
recognize that the requirement to achieve a similar on–off ratio in a 3. Zhang, Y. B., Tan, Y. W., Stormer, H. L. & Kim, P. Experimental observation
bigger GNM device is much more stringent than in a smaller single of the quantum Hall effect and Berry’s phase in graphene. Nature 438,
GNR device, as each conducting channel needs to be small enough 201–204 (2005).
to enable a large on–off ratio. The ability to achieve a high on–off 4. Geim, A. K. & Novoselov, K. S. The rise of graphene. Nature Mater. 6,
183–191 (2007).
ratio and high current in our large GNM devices clearly highlights 5. Tan, Y.-W., Zhang, Y., Stormer, H. L. & Kim, P. Temperature dependent electron
the relative uniformity of the current GNM film. transport in graphene. Eur. Phys. J. Spec. Top. 148, 15–18 (2007).
In conclusion, we have demonstrated a new graphene nanostruc- 6. Novoselov, K. S. et al. Two-dimensional gas of massless Dirac fermions in
ture—graphene nanomesh—that can effectively open up a conduc- graphene. Nature 438, 197–200 (2005).
tion bandgap in a large piece of graphene to form a continuous 7. Meric, I., Han, M. Y., Young, A. F., Ozyilmaz, B., Kim, P. & Shepard, K. L.
Current saturation in zero-bandgap, topgated graphene field-effect transistors.
semiconducting thin film. We have shown that GNMs with variable Nature Nanotech. 3, 654–659 (2008).
periodicities and neck widths down to 5 nm can be prepared using 8. Nakada, K., Fujita, M., Dresselhaus, G. & Dresselhaus, M. S. Edge state in
block copolymer lithography. Using such a nanomesh as a semicon- graphene ribbons: nanometer size effect and edge shape dependence. Phys. Rev.
ducting channel, we have demonstrated room-temperature transis- B 54, 17954–17961 (1996).
tors with driving currents nearly 100 times larger than that of an 9. Son, Y.-W., Cohen, M. L. & Louie, S. G. Energy gaps in graphene nanoribbons.
Phys. Rev. Lett. 97, 216803 (2006).
individual GNR device, but with a comparable on–off ratio that is 10. Barone, V., Hod, O. & Scuseria, G. E. Electronic structure and stability of
tunable by varying the neck width. The block copolymer lithogra- semiconducting graphene nanoribbons. Nano Lett. 6, 2748–2754 (2006).
phy fabrication of GNMs is an intrinsically scalable approach and 11. Li, T. C. & Lu, S.-P. Quantum conductance of graphene nanoribbons with edge
can therefore enable a continuous semiconducting GNM thin film defects. Phys. Rev. B 77, 085408 (2008).
that can be processed using standard semiconductor processing to 12. Sols, F., Guinea, F. & Castro Neto, A. H. Coulomb blockade in graphene
nanoribbons. Phys. Rev. Lett. 99, 166803 (2007).
fabricate integrated devices/circuits with characteristics designed 13. Han, M. Y., Ozyilmaz, B., Zhang, Y. & Kim, P. Energy band-gap engineering of
to meet specific circuit requirements. The availability of such graphene nanoribbons. Phys. Rev. Lett. 98, 206805 (2007).
GNMs in the deep nanometre regime will also provide an interesting 14. Ozyilmaz, B., Jarillo-Herrero, P., Efetov, D. & Kim, P. Electronic
system for fundamental investigation of transport behaviour in the transport in locally gated graphene nanoconstrictions. Appl. Phys. Lett.
highly interconnected graphene network, and will enable exciting 91, 192107 (2007).
15. Chen, Z. H., Lin, Y.-M., Rookes, M. J. & Avouris, P. Graphene nano-ribbon
opportunities in highly sensitive biosensors and a new generation electronics. Physica E 40, 228–232 (2007).
of spintronics30. 16. Li, X. L., Wang, X. R., Zhang, L., Lee, S. & Dai, H. J. Chemically derived,
ultrasmooth graphene nanoribbon semiconductors. Science 319,
Methods 1229–1232 (2008).
Preparation of the block copolymer template. Graphene thin film was obtained by 17. Kosynkin, D. V. et al. Longitudinal unzipping of carbon nanotubes to form
mechanically peeling natural graphite onto a silicon substrate with 300-nm thermal graphene nanoribbons. Nature 459, 872–876 (2009).
oxide. A 10-nm SiOx thin film was evaporated onto graphene using an electron- 18. Jiao, L. Y., Zhang, L., Wang, X. R., Diankov, G. & Dai, H. J. Narrow graphene
beam evaporator operating at 1  1026 torr. This fresh oxide surface was nanoribbons from carbon nanotubes. Nature 458, 877–880 (2009).
functionalized with P(S-r-MMA) random copolymer (Polymer Source; Mn , 11,000; 19. Bai, J., Duan, X. & Huang, Y. Rational fabrication of graphene nanoribbons
Mw/Mn , 1.15; styrene 55 mol%, functionalized a-hydroxyl and v-tempo moiety) by using a nanowire etch mask. Nano Lett. 9, 2083–2087 (2009).
spin coating a thin film from 1 wt% toluene solution. This film was annealed at 20. Lu, W. & Lieber C. M. Nanoelectronics from the bottom up. Nature Mater. 6,
170 8C for 72 h to anchor the polymer onto the oxide surface via the end hydroxyl 841–850 (2007).
group, and then rinsed with toluene to remove unanchored polymer. P(S-b-MMA) 21. Duan, X. Assembled semiconductor nanowire thin films for high-performance
with a molecular weight of 77,000 g mol21 (PS-PMMA, 55,000–22,000; Mw/Mn , flexible macroelectronics. MRS Bull. 32, 134–141 (2007).
1.09) and 47,700 g mol21 (PS-PMMA, 35,500–12,200; Mw/Mn , 1.04) were 22. Park, S. & Ruoff R. S. Chemical methods for the production of graphenes. Nature
purchased from Polymer Source and dissolved in toluene. The block copolymer thin Nanotech. 4, 217–224 (2009).
film was prepared on the neutralized surface by spin coating 1 wt% filtered polymer 23. Tung, V. C., Allen, M. J., Yang, Y. & Kaner, R. B. High-throughput solution
solution at 2,500 r.p.m. followed by annealing at 180 8C for 12 h. The thickness was processing of large-scale graphene. Nature Nanotech. 4, 25–29 (2009).
determined to be 30–35 nm by AFM profile. The annealed film was exposed to 24. Li, X. et al. Large-area synthesis of high quality and uniform graphene films on
295 nm UV irradiation for 30 min under vacuum. The degraded PMMA domains copper foils. Science 324, 1312–1314 (2009).
were removed by immersing in glacial acid for 20 min, followed by extensive 25. Reina, A. et al. Large area, few-layer graphene films on arbitrary substrates by
water rinsing. chemical vapour deposition. Nano Lett. 9, 30–35 (2009).
26. Hawker, C. J. & Russell, T. P. Block copolymer lithography: merging ‘bottom-up’
Formation of the graphene nanomesh. We used a reactive ion etcher (STS MESC with ‘top-down’ processes. MRS Bull. 30, 952–966 (2005).
Multiplex Advanced Oxide Etcher) to etch down to the graphene layer. An O2 27. Xu, T. et al. The influence of molecular weight on nanoporous polymer films.
plasma process (50 W, 4 mtorr) was first used to remove exposed random Polymer 42, 9091–9095 (2001).

NATURE NANOTECHNOLOGY | VOL 5 | MARCH 2010 | www.nature.com/naturenanotechnology 193

© 2010 Macmillan Publishers Limited. All rights reserved.
LETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2010.8

28. Cheng, J. Y., Mayes, A. M. & Ross, C. A. Nanostructure engineering by template Science Fellowship. X.D. acknowledges partial support by the NIH Director’s New
self-assembly of block copolymers. Nature Mater. 3, 823–828 (2004). Innovator Award Program, part of the NIH Roadmap for Medical Research, through grant
29. Park, S. et al. Macroscopic 10-terabit-per-square-inch arrays from block no. 1DP2OD004342-01.
copolymers with lateral order. Science 323, 1030–1033 (2009).
30. Son, Y. W., Cohen, M. L. & Louie, S. G. Half-metallic graphene nanoribbons. Author contributions
Nature 444, 347–349 (2006). X.D., Y.H. and J.B. conceived and designed the experiments. J.B., X.Z. and S.J. performed
the experiments. J.B. collected and analysed the data. J.B., Y.H. and X.D. wrote the paper.
Acknowledgements All authors discussed the results and commented on the manuscript.
The authors acknowledge technical support regarding TEM from the Electron Imaging
Center for Nanomachines (EICN) at University of California, Los Angeles, and for device Additional information
fabrication from the Nanoelectronics Research Facility at University of California, Los The authors declare no competing financial interests. Supplementary information
Angeles. We thank R. Kaner and D. Neuhauser for discussions, J. Chen and C. Liu for accompanies this paper at www.nature.com/naturenanotechnology. Reprints and
assistance in statistics analysis, and F.X. Xiu for assistance in block copolymer processing. permission information is available online at http://npg.nature.com/reprintsandpermissions/.
Y.H. acknowledges support from the Henry Samueli School of Engineering and an Applied Correspondence and requests for materials should be addressed to Y.H. and X.D.

194 NATURE NANOTECHNOLOGY | VOL 5 | MARCH 2010 | www.nature.com/naturenanotechnology

© 2010 Macmillan Publishers Limited. All rights reserved.
 feature

Selling graphene by the ton

Michael Segal
Small start-up companies are making large volumes of graphene, the world’s thinnest material, for applications
such as composites and electrodes.

G
raphene is, by any measure, a manufactured will not end up in advanced

VorBeck MATeriAlS
story of extremes. Consisting of devices such as transistors. Instead, most
a few sheets of carbon atoms, it is applications fall into two broad and not
the strongest material ever measured1, always exclusive categories: composites
it has a thermal conductivity more than and electrodes. Composites consist of
double that of diamond2 and has a charge graphene dispersed into a host matrix,
mobility that is among the highest of any and take on some of the properties of
semiconductor3. But just as remarkable graphene as a result. These may include
has been the speed at which this material strength, stiffness, and electrical or thermal
has moved from the research laboratory to conductivity. For example, some containers
the marketplace. used to store and transport volatile fuels are
The exceptional electrical properties of made from polymer or rubber composites.
graphene were discovered only five years Incorporating graphene would increase
ago in experiments that involved manually their stiffness by an order of magnitude,
peeling samples weighing picograms from and allow them to conduct electrical
bulk graphite4. Today the total production currents, reducing the probability of
output of various flavours of graphene electrical discharges and increasing safety.
exceeds 15 tons per year, produced in Similarly, packaging made from
more than 40,000 square feet of facilities, a graphene composite could shield
and this is set to increase to more than electronics from radio frequency signals,
200 tons per year within the next year conduct current to external devices or even
or two. “We actually have a company store information. Vorbeck is focusing
asking the question right now, ‘When on printable conductive inks (a distinct
will I get 5 tons?’” says Bor Jang, the chief class of composites), for applications in
executive officer of Angstron Materials, printed electronics such as radio-frequency
which is based in Dayton, Ohio. “We can’t A shipment of Vor‑x — a functionalized graphene identification tags, low-resolution displays
produce enough.” manufactured by Vorbeck Materials — ready to and backlights, sensors, flexible connectors
be sent to a development partner. and packaging. Graphene makes these inks
“The graphene is already both highly conductive and easy to process.
Graphene-based electrodes take
there — all you have to do The use of acids can lead to oxidation of advantage of the material’s electrical
the graphite, and Vorbeck has licensed conductivity, transparency and
is peel it off.” an approach developed at Princeton surface area. One attractive target is
Bor Jang, Angstron Materials. University5,6 to convert graphite oxide into the replacement of indium tin oxide,
graphene. Angstron, meanwhile, has also which is used in LCD displays and
At present, three small US companies developed a process that avoids chemical photovoltaic devices. Graphene offers
account for the bulk of graphene intercalation and oxidation altogether. easier processing and, potentially, much
manufacturing capacity: Angstron These approaches have the common lower costs. Shortages of indium have
Materials; Vorbeck Materials in Jessup, advantage of being inexpensive. The low driven up its cost from 10 cents per gram
Maryland; and XG Sciences in East cost results from a relatively simple and in 2002 to about 60 cents per gram today,
Lansing, Michigan. The companies take the low-energy treatment process, and a cheap after reaching a peak of one dollar per
same basic manufacturing approach, which starting material: graphite is mined and gram in 2007. Another target is energy
is to break apart graphite into the sheets sold at a cost of cents per gram. Jang puts storage, potentially replacing the graphite
of graphene that make it up — usually it simply: “The graphene is already there — electrodes found in batteries, capacitors
by intercalating acids between them. The all you have to do is peel it off.” and fuel cells. “If you have an inexpensive
resulting intermediate is then thermally or graphene nanoplatelet material and can
mechanically treated to extract graphene What is graphene good for? control its concentration, orientation,
platelets (or nanoplatelets). These platelets Although much of the early excitement surface chemistry and so on, it has the
can consist of a single layer or multiple around graphene centred on its potential for making a significant impact
layers of graphene, and measure hundreds semiconducting properties and exotic on performance,” says Lawrence Drzal,
of nanometres to tens of microns across. physics, the tons of material now being chief scientist of XG Sciences and professor

612 nature nanotechnology | VOL 4 | OCTOBER 2009 | www.nature.com/naturenanotechnology

© 2009 Macmillan Publishers Limited. All rights reserved

 feature

of chemical engineering and materials and the conservatism of their larger in its composites, and two or three tons
science at Michigan State University. “We counterparts is partly a matter of scale. of Angstron’s production capacity will be
are spending more effort in developing Even though tens of tons is orders of devoted to single-layer graphene. However,
new applications in energy storage than magnitude beyond laboratory production about a fifth of Angstron’s output will be
anything else,” says Jang. scales, it is also orders of magnitude below 30-layer graphene, and XG’s nanoplatelets
Energy applications are at the core of industrial chemical company scales. “We’d are typically 5 nm thick, corresponding
another young company, Graphene Energy, consider a few tons per year to be a pretty to more than 16 layers of graphene. Hahn
Inc., spun out of Rod Ruoff ’s laboratory at small volume of material compared with says that Dow is “still working through
the University of Texas at Austin. Graphene what we typically manufacture,” says the process of exactly what it means to call
Energy is focusing on ultracapacitors, which Hahn. “Dow would probably need to something graphene.”
fill a niche between the high-energy/low- see a bigger opportunity for it to make Part of this process involves developing
power densities of batteries, and the high- economic sense.” metrics that allow a quantitative
power/low-energy densities of standard At the other end of the market are description of a particular graphene or
dielectric capacitors. Ultracapacitors require start-up companies serving pristine, graphene composite sample. The most
a conductive and high surface area electrode, single-layer or few-layer graphene in small common metrics are specific surface area,
and graphene fits the bill. “There is reason quantities for research into fundamental in units of area per mass, and thermal
to believe that the specific capacitance could science as well as applications that are and electrical conductivity. Single-layer
be improved significantly, by a factor of two further from commercialization, such as graphene has a specific surface area of
or three,” says Ruoff. “This would be, for graphene-based transistors. For example, approximately 2,600 m2 g−1, whereas the
many people, absolutely game-changing.” Graphene Industries, a company spun best samples produced by industry have
Although Graphene Energy is agnostic out of Andre Geim’s group at Manchester values between 1,000 and 1,800 m2 g−1, even
about whether it will manufacture its own University7, sells individually characterized those that are characterized by electron
graphene or purchase it from an external flakes one at a time. This part of the market microscopy as being single layer. This is
supplier, it will need large volumes of the may also soon be served by chemical due in part to the agglomeration of the
material: “To have a long-range impact vapour deposition techniques, which material when it is assembled into a bulk
in ultracapacitors you’re going to need produce large-area and high-quality sample, device or composite. Similarly,
thousands to tens of thousands of metric graphene8, but these samples are likely to the thermal conductivity of a single sheet
tons per year,” says Ruoff. be expensive relative to graphite-derived of graphene has been measured to be
graphene in the near term. 4,000 W m−1 K−1 (ref. 2), which is three
orders of magnitude greater than that of the
“chemists and physicists look What is graphene? typical graphene composite.
at the properties of a single Although the International Union of Pure “I make the distinction that chemists
and Applied Chemistry (IUPAC) defines and physicists look at the properties of a
graphene sheet, whereas graphene as a single layer of carbon atoms, single graphene sheet, whereas engineers
engineers look at an assembly researchers routinely talk of bilayer, trilayer look at an assembly of graphene or other
and multilayer graphene. Industry stretches nanoparticles into a device,” says Drzal. The
of graphene or other the definition further still, to include any method used to manufacture a composite
nanoparticles into a device.” highly exfoliated graphite product. Vorbeck material or device, therefore, is as
lawrence Drzal, XG Sciences.
uses predominantly single-layer graphene important as the properties of the graphene

Although these applications have

attracted the undivided attention of start-
up companies, larger chemical companies
like Dow, 3M, BASF and DuPont
have so far taken a more conservative
approach. BASF has, for example, formed
a partnership with Vorbeck to jointly
develop products. Dow, on the other hand,
has investigated graphene for over five
years, but remains unconvinced about its
prospects. The company initially focused
on cable-shielding applications, but costs
proved to be uncompetitive. “I’ve yet to
see applications that are really demanding
AnGsTROn MATERiALs

large quantities of these products and

that’s a concern both for graphene and for
carbon nanotubes, although we recognize
that much of the development is still
early-stage,” says Steve Hahn, a senior
scientist with Dow’s Ventures and Business
Development division.
The discrepancy between the A fuel cell bipolar plate manufactured by Angstron. The compression‑moulded thermoplastic composite
enthusiasm of the start-up companies contains 60% graphene by weight.

nature nanotechnology | VOL 4 | OCTOBER 2009 | www.nature.com/naturenanotechnology 613

© 2009 Macmillan Publishers Limited. All rights reserved

feature

used. This fact, combined with the variety inks, graphene also competes with one: “Graphene will gradually replace
of metrics appropriate for different devices its close older relative, the carbon nanotubes in many applications,” he
and materials, means adherence to a nanotube. “In almost all of the areas says. Indeed, Angstron’s literature claims
strict single-layer definition is not useful. that we are developing with our initial that graphene trumps nanotubes on a
Instead, industrial researchers focus on customers, they have definitely evaluated number of other fronts: purity, ease of
what is ‘good enough’ for their application. nanotubes,” says John Lettow, president functionalization, thermal conductivity,
At Dow, Hahn has settled on a surface of Vorbeck. “That’s been step one in the impact on viscosity, uniformity of
area figure of 500 m2 g−1, which he calls “a process.” Big companies like Bayer and dispersion, loading, barrier properties and
bit of a benchmark”. And researchers at Showa Denko are already producing capital investment.
Graphene Energy have achieved what they hundreds of tons of nanotubes per year,
call “excellent” ultracapacitor performance with plans to produce thousands of tons in Past and future
using electrodes with surface areas between the near term. Graphene manufacturers believe that
400 and 700 m2 g−1, which is well below the In competing with nanotubes, graphene they have solved many of the technical
theoretical maximum for graphene. makers point to cost, surface area, ease of challenges involved in high-volume
processing and, possibly, safety as its main production, and consider the material
“i think graphene is now ready advantages. The cost advantage derives that they are now producing as ready for a
from the process: nanotube synthesis variety of applications: “I think graphene
for commercialization.” usually requires a higher energy input, is now ready for commercialization,” says
John lettow, Vorbeck. a high-purity gaseous feedstock and Lettow. In their view a more important
catalysts. “I think there’s a perception challenge is to ensure that real applications
Industry has extended not only the that graphene might in the long run be are developed in time to provide the
definition of graphene, but also its history, a pretty inexpensive commodity-type demand for these high volumes. In the past
questioning the claim that it is a new material compared with carbon nanotubes,” many companies making nanomaterials
material. “The original synthetic work for says Ruoff. have often focused on volume manufacture
graphene goes back into the late 1800s,” The remaining advantages largely before applications were mature, leading
says Hahn. “It is hard to look at graphene derive from geometry. The fact that about to a number of them consuming large
and say it is a new material.” This early half the surface area of a nanotube is amounts of cash, and then going bankrupt.
work focused on intercalated or lamellar relatively inaccessible (because it is inside “That lesson is very well learnt from carbon
compounds of graphite, and by 1970 single- the nanotube) is a critical drawback nanotubes,” says Lettow.
layer graphene on nickel substrates, called for ultracapacitors and other energy The graphene start-up companies
monolayer graphite, had been isolated9,10. applications. “Graphene’s planar geometry discussed here have, so far, spent only a
In the 1980s, Union Carbide was producing reduces tangling and clumping relative to small fraction of the money that was spent
graphite particles under the trade name nanotubes,” says Lettow, “and leads to films by high-volume nanotube manufacturers
Grafoil, with surface areas of 20 m2 g−1, with lower resistances.” over the equivalent period. Moreover, they
roughly the surface area of what might are moving quickly towards marketable
today be called 100-layer graphene11. “Graphene might in products by maintaining close relationships
While Angstron, XG and Vorbeck with development partners, which are
have all applied for patent protection of the long run be a pretty absorbing the majority of their production
their own fabrication approaches, this inexpensive commodity- output. “When single development partners
long history complicates the securing are buying tens of kilograms and putting
of such rights. “If manufacturers just type material compared with it into composites at 2% or 10% loading,
use intercalation and exfoliation, the carbon nanotubes.” their production of pilot products is in the
intellectual property is not very defensible,” tons,” says Jang. “Next year you will see
says Jang. “Another way to do it is to patent rod ruoff, Graphene energy. a breakthrough. ❐
the applications.”
Intellectual property protection by This difference in geometry may also Michael Segal is an Associate Editor for
patents is therefore supplemented by the lead to a safety advantage for graphene. Nature Nanotechnology.
development of in-house, equipment- “The interesting thing about graphene
specific ‘know-how’, particularly related nanoplatelets is that they are nano in references
to questions of scaling and efficiency. one dimension rather than three,” says 1. Lee, C., Wei, X., Kysar, J. W. & Hone, J. Science
321, 385–388 (2008).
“Anybody could probably go into a Drzal. This may mean that it less likely to 2. Balandin, A. A. et al. Nano Lett. 8, 902–907 (2008).
laboratory and generate this material at the move through biological barriers. “From 3. Chen, J.-H., Jang, C., Xiao, S., Ishigami, M. & Fuhrer, M. S.
one gram level,” says Drzal. “But scaling everything we’ve seen so far, potentially Nature Nanotech. 3, 206–209 (2008).
up the process to the point where you’re there’s much less of a health risk associated 4. Novoselov, K. S. et al. Science 306, 666–669 (2004).
going to make significant quantities is with it.” Both Drzal and Jang believe that 5. Li, J.-L. et al. Phys. Rev. Lett. 96, 176101 (2006).
6. Schniepp, H. et al. J. Phys. Chem. 110, 8535–8539 (2006).
a challenge.” handling requirements for graphene will
7. Nature Nanotech. 3, 523 (2008).
be similar to that for coal dust and other 8. Obraztsov, A. N. Nature Nanotech. 4, 212–213 (2008).
that other carbon nanomaterial small particles. 9. Eizenberg, M. & Blakely, J. M. Surf. Sci. 82, 228–236 (1970).
In addition to competing with established For Jang, the combination of cost and 10. Park, S. & Ruoff, R. S. Nature Nanotech. 4, 217–224 (2009).
materials such as graphite and metallic a reduced health risk is a compelling 11. Chung, D. D. L. J. Mater. Sci. 22, 4190–4198 (1987).

614 nature nanotechnology | VOL 4 | OCTOBER 2009 | www.nature.com/naturenanotechnology

© 2009 Macmillan Publishers Limited. All rights reserved

news & views

enzyme cascades. First, a one-to-one complexes reported so far are all static Arizona State University, Tempe, Arizona 85287, USA.
DNA-protein conjugation technique should constructs. It is desirable to incorporate DNA e‑mail: hao.yan@asu.edu
be adopted to ensure better control over nanomechanical devices7,8 into the assembly
the spatial orientation of the proteins and to achieve dynamic complexes with activities References
1. Wilner, O. I. et al. Nature Nanotech. 4, 249–254 (2009).
minimal loss of their biological function after that are regulated by biologically relevant
2. Seeman, N. C. Nature 421, 427–431 (2003).
conjugation and immobilization. The ‘click signal inputs such as pH change, salt gradient 3. Aldaye, F. A., Palmer, A. L. & Sleiman, H. F. Science
chemistry’ conjugation method reported by or the addition of a nucleic acid strand. 321, 1795–1799 (2008).
Distefano’s group6 seems to be a promising Indeed, it is foreseeable that, by merging the 4. Rinker, S., Ke, Y., Liu, Y., Chhabra, R. & Yan, H., Nature Nanotech.
candidate for this purpose. Second, although addressability of DNA structure with the 3, 418–422 (2008).
DNA-directed assembly has been intensively versatile functionality of protein libraries, 5. Niemeyer, C. M., Koehler, J. & Wuerdemann, C. ChemBioChem
3, 242–245 (2002).
investigated in two dimensions, expanding self-assembling, dynamic and functional
6. Duckworth, B. P. et al. Angew. Chem. Int. Ed.
the organizational power of DNA scaffolds biomolecular networks are in reach. ❐ 46, 8819–8822 (2007).
to three-dimensional motifs would truly 7. Ding, B. & Seeman, N. C. Science 314, 1583–1585 (2006).
mimic the way protein–protein complexes Chenxiang Lin and Hao Yan are at the Department of 8. Yin, P., Choi, H. M. T., Calvert, C. R. & Pierce, N. A. Nature
are assembled in nature. Third, the bienzyme Chemistry and Biochemistry & The Biodesign Institute, 451, 318–322 (2008).

cheMIcAl VAPoUR DePoSItIoN

Making graphene on a large scale

Graphene samples with areas of several square centimetres and excellent electrical and optical properties have
been fabricated using chemical vapour deposition.

Alexander N. obraztsov

g
raphene — a single layer of carbon a b c
atoms bonded into a two-dimensional Carbon precipitates
Carbonaceous gas
hexagonal network — has been onto surface
the subject of intense research since it
Ni
was first isolated in free-standing form in
2004 (refs 1, 2). Just as related forms of
carbon such as fullerenes and nanotubes
had previously been found to have many
remarkable properties, so too did free- d e
standing graphene. Among the most
unexpected and promising of these were its Chemical etchant
thermodynamic stability, extremely high
charge-carrier mobilities, and mechanical
stiffness, the last of which distinguishes it Free-standing graphene
from monoatomic metallic films3.
At first, graphene was produced by the
Figure 1 | A chemical vapour deposition route for large-scale graphene production. a, A thin layer of nickel
mechanical exfoliation, or peeling of thin
(Ni) is deposited onto a substrate (not shown): the crystallinity of this layer is controlled by its thickness,
layers, of thick graphite1. Although this was
annealing and the nature of the substrate. b, The Ni layer is heated to about 1,000 °C and exposed to a
an easy way to get micron-size flakes suitable
carbonaceous gas environment. Carbon atoms are generated at the Ni surface and diffuse into the metal.
for laboratory research, it was difficult to see
c, As the nickel is cooled down, the carbon atoms precipitate out of the nickel layer and form graphene on
how it would scale to produce the large areas
its surface. The graphene samples are expected to form on the (111) faces of Ni crystallites. d, The graphene
necessary for some of its most promising
membrane is detached from the Ni layer by gentle chemical etching. e, The resulting free-standing graphene
potential applications. Subsequently,
layer is transferred onto appropriate substrates (not shown). Colours illustrate the difference in temperature
graphene was grown by epitaxy 4, which
(orange, hot; blue, cold) and chemical composition (orange, gas; blue, etchant).
requires ultrahigh vacuum conditions and is
therefore expensive, and by various chemical
methods3,5,6, which produce graphene with
degraded electrical properties. have made a breakthrough by showing films on nickel substrates from vapour-phase
The fabrication of high-quality, that samples of single- and few-layer carbon10. The CVD approach to producing
large-area graphene remains one of graphene with areas of square centimetres graphene (Fig. 1) relies on dissolving carbon
the hottest topics in nanotechnology can be manufactured with a chemical into the nickel substrate, and then forcing
and materials science today. Now three vapour deposition (CVD) technique, and it to precipitate out by cooling the nickel.
independent research groups, headed transferred to other substrates. The thickness and crystalline ordering of
by Byung Lee Hong of Sungkyunkwan The three approaches are similar to each the precipitated carbon is controlled by
University 7, Jing Kong of MIT8 and other, and also similar to previous work on the cooling rate and by the concentration
Yong Chen from Purdue University 9, the deposition of nanometre-thick graphite of carbon dissolved in the nickel. This

212 nature nanotechnology | VOL 4 | APRIL 2009 | www.nature.com/naturenanotechnology

© 2009 Macmillan Publishers Limited. All rights reserved

 news & views

concentration is in turn controlled by the substrates used to grow epitaxial graphene electronic devices. These include quantum
type and concentration of the carbonaceous would reduce the number of defects in the coherent devices where phase coherent
gas, and the thickness of the nickel layer. samples by reducing the mismatch between electron transport is used for new device
After a chemical etching of the nickel, the the nickel and graphene lattices11. Another functionalities or improved performance.
graphene membrane detaches and can be serious obstacle to CVD technology is that The scalability of CVD graphene technology
transferred to another substrate. This direct the large difference in the thermal expansion may hasten the transition from processing
CVD synthesis provides high quality layers coefficients of nickel and graphite leads to single electronic devices — which has
of graphene without intensive mechanical or ripple structures in the samples7,9,10. This will been demonstrated already 1,7–9 — to the
chemical treatments. need to be overcome to achieve the planar production of complicated integrated
The graphene films demonstrate excellent graphene film topology required for micro- circuits containing numerous active and
electronic and optical properties: a sheet patterning of electronic devices. passive graphene-based components. ❐
resistance as low as ~280 Ω cm−2 (ref. 7), a The successful demonstration of
carrier mobility as high as 3,700 cm2 V−1 s−1 large-area graphene from CVD is a Alexander N. Obraztsov is in the Department of
(ref. 7) and a 90% optical transparency 7,8. promising step towards the industrial Physics, Moscow State University, Moscow 119991,
Although these values are inferior to production of graphene-based electronic Russia, and in the Department of Physics and
those obtained for a perfect single layer of and optoelectronic devices. The use of Mathematics, University of Joensuu, PO Box 111,
graphene1, they are nevertheless useful for large-scale CVD graphene as flexible and FI‑80101 Joensuu, Finland.
applications such as flexible and transparent optically transparent electrical contacts e‑mail: obraz@polly.phys.msu.ru;
electrodes for solar cells and liquid crystal has been demonstrated by the teams led alexander.obraztsov@joensuu.fi
displays, and various high-frequency by Hong 7 and Kong 8. These properties are
electronic and optoelectronic devices. The attractive for applications, including liquid References
1. Novoselov, K. S. et al. Science 306, 666–669 (2004).
combination of low cost, scalability and crystal devices, solar cells and organic 2. Novoselov, K. S. et al. Nature 438, 197–200 (2005).
advanced material properties displayed by light-emission devices, where graphene may 3. Booth, T. J. et al. Nano Lett. 8, 2442–2446 (2008).
the CVD approach is exceptional. provide potentially significant advantages 4. Berger, C. et al. Science 312, 1191–1196 (2006).
Even though the further development (in terms of carrier mobility and mechanical 5. Eda, G., Fanchini, G. & Chhowalla, M. Nature Nanotech.
of large-scale CVD deposition of graphene strength and flexibility) over conventional 3, 270–274 (2008).
6. Park, S. & Ruoff, R. S. Nature Nanotech. 4, 217–224 (2009).
for industrial purposes will be mostly by metal oxides.
7. Kim, K. S. et al. Nature 457, 706–710 (2009).
engineering, fundamental research will Chen and colleagues9 also unambiguously 8. Reina, A. et al. Nano Lett. 9, 30–35 (2009).
also have a part to play. Replacing the demonstrate that the extraordinary 9. Cao, H. et al. Preprint at <http://arxiv.org/abs/0901.1136> (2009).
polycrystalline substrates used by the electronic properties of CVD graphene 10. Obraztsov, A. N. et al. Carbon 45, 2017–2021 (2007).
three groups with the single-crystal nickel may be practically exploited in submicron 11. Grüneis, A. & Vyalikh, D. V. Phys. Rev. B 77, 193401 (2008).

PRoBe MIcRoScoPy

Beneath the surface

The Fermi surface of a crystalline solid
is the boundary between states that are

© 2009 aaas
occupied by electrons at absolute zero
temperature and states that are empty at
this temperature. It is usually measured in
momentum space, rather than real space,
by techniques such as angle-resolved
photoemission spectroscopy. Now Martin
Wenderoth, of the University of Göttingen,
and colleagues have shown that it is possible
to image the Fermi surface of a solid in real
space with a scanning tunnelling microscope
(STM; Science 323, 1190–1193; 2009).
In a typical STM measurement an
electron from the tip of the microscope
tunnels into a surface and becomes a bulk
electron wave with amplitude that decays not spherical, electrons can be focused along copper(111) surface; the image on the right
with distance. If a defect atom is present certain directions, resulting in pronounced is 3.5 nm across and shows one cobalt atom
beneath the surface, the electron wave can interference patterns at the surface. These below a copper(100) surface. The insets
be reflected, interfering with the incoming patterns reflect information about the flow are 4 nm across and show the local electron
wave to form a standing-wave pattern. of electrons in the bulk of the material, and density of states derived from the images
When the Fermi surface is spherical, a weak hence the shape of the Fermi surface. (left insets), and from theory (right insets).
interference pattern is observed at the The STM image on the left is 9 nm
surface. However, when the Fermi surface is across and shows four cobalt atoms below a owAIN VAUghAN

nature nanotechnology | VOL 4 | APRIL 2009 | www.nature.com/naturenanotechnology 213

© 2009 Macmillan Publishers Limited. All rights reserved

 review article
Published online: 29 march 2009 | doi: 10.1038/nnano.2009.58

Chemical methods for the production of graphenes

Sungjin Park1 and Rodney S. Ruoff1*

Interest in graphene centres on its excellent mechanical, electrical, thermal and optical properties, its very high specific surface
area, and our ability to influence these properties through chemical functionalization. There are a number of methods for gen-
erating graphene and chemically modified graphene from graphite and derivatives of graphite, each with different advantages
and disadvantages. Here we review the use of colloidal suspensions to produce new materials composed of graphene and
chemically modified graphene. This approach is both versatile and scalable, and is adaptable to a wide variety of applications.

T
he development of various methods for producing graph­ene graphite in the presence of strong acids and oxidants. The level of
— a single layer of carbon atoms bonded together in a hex- the oxidation can be varied on the basis of the method, the reaction
agonal lattice — has stimulated a vast amount of research in conditions and the precursor graphite used. Although extensive
recent years1. The remarkable properties of graphene reported so research has been done to reveal the chemical structure of graphite
far include high values of its Young’s modulus (~1,100 GPa)2, frac- oxide, several models are still being debated in the literature.
ture strength (125 GPa)2, thermal conductivity (~5,000 W m−1K−1)3, Solid-state 13C NMR spectroscopy of graphite oxide and recently
mobility of charge carriers (200,000 cm2 V−1 s−1)4 and specific sur- of 13C-labelled graphite oxide favours the model shown in Fig. 1a; the
face area (calculated value, 2,630 m2 g−1)5, plus fascinating trans- sp2-bonded carbon network of graphite is strongly disrupted and a sig-
port phenomena such as the quantum Hall effect6. Graphene and nificant fraction of this carbon network is bonded to hydroxyl groups
chemically modified graphene (CMG) are promising candidates or participates in epoxide groups29–32. Minor components of carboxylic
as components in applications such as energy-storage materi- or carbonyl groups are thought to populate the edges of the layers in
als5, ‘paper-like’ mater­ials7,8, polymer composites9,10, liquid crystal graphite oxide. This indicates that further work with solid-state NMR
devices11 and mechanical resonators12. on 13C-labelled graphite oxide is necessary, along with (for example)
Graphene has been made by four different methods. The first titration with fluorescent tags of carboxylic and other groups to iden-
was chemical vapour deposition (CVD) and epitaxial growth, such tify their spatial distribution on individual graphene oxide platelets
as the decomposition of ethylene on nickel surfaces13. These early derived from graphite oxide as discussed further below.
efforts (which started in 1970) were followed by a large body of Graphite oxide thus consists of a layered structure of ‘graphene
work by the surface-science community on ‘monolayer graphite’14. oxide’ sheets that are strongly hydrophilic such that intercalation
The second was the micromechanical exfoliation of graphite15. of water molecules between the layers readily occurs33. The inter-
This approach, which is also known as the ‘Scotch tape’ or peel-off layer distance between the graphene oxide sheets increases revers-
method, followed on from earlier work on micromechanical exfo- ibly from 6 to 12 Å with increasing relative humidity33. Notably,
liation from patterned graphite16. The third method was epitaxial graphite oxide can be completely exfoliated to produce aqueous
growth on electrically insulating surfaces such as SiC (ref. 17) and colloidal suspensions of graphene oxide sheets by simple sonication
the fourth was the creation of colloidal suspensions. (Fig. 1b)34 and by stirring the water/graphite oxide mixture for a
Micromechanical exfoliation has yielded small samples of long enough time35. The measurement of the surface charge (zeta
graphene that are useful for fundamental study. Although large- potential) of graphene oxide sheets36 shows that they have negative
area graphene films (up to ~1 cm2) of single- to few-layer graphene charges when dispersed in water. This suggests that electrostatic
have been generated by CVD growth on metal substrates18–20, and repulsion between negatively charged graphene oxide sheets could
graphene-type carbon materials have been produced by substrate- generate a stable aqueous suspension of them. A considerable body
free CVD21, radio-frequency plasma-enhanced CVD22, aerosol of work37,38 on such aqueous colloidal suspensions was carried out
pyrolysis23 and solvothermal synthesis24, the uniform growth of in the 1950s and 1960s. Such graphene oxide sheets probably have
single-layer graphene is still a challenge. In this review, we dis- a similar chemical structure to the layers in graphite oxide and are
cuss the production of graphene and CMG from colloidal suspen- a promising starting material in the generation of colloidal suspen-
sions made from graphite, derivatives of graphite (such as graphite sions of other CMGs through chemical tuning. Filtration of CMG
oxide) and graphite intercalation compounds. This approach is suspensions has produced free standing paper-like materials7,36,39–41
both scalable, affording the possibility of high-volume produc- that have a layered structure (Fig. 1c, d). Significant advances have
tion, and versatile in terms of being well-suited to chemical func- also been made in using homogeneous suspensions of CMG sheets
tionalization. These advantages mean that the colloidal suspension to produce thin films, which can be relevant to transparent and
method for producing graphene and CMG could be used for a electrically conductive thin-film applications, among others36,39–44.
wide range of applications.
Unreduced graphene oxide sheets
Graphenes from graphite oxide Several authors have stated that homogeneous colloidal suspen-
Since it was first prepared in the nineteenth century25,26, graphite sions of graphene oxide in aqueous and various organic solvents
oxide has been mainly produced by the Brodie25, Staudenmaier27 can be achieved by simple sonication of graphite oxide8,34,45–47.
and Hummers28 methods. All three methods involve oxidation of The hydrophilic graphene oxide can be easily dispersed in water

Department of Mechanical Engineering and the Texas Materials Institute, University of Texas at Austin, One University Station C2200, Austin, Texas
1

78712-0292, USA. *e-mail: r.ruoff@mail.utexas.edu

nature nanotechnology | VOL 4 | APRIL 2009 | www.nature.com/naturenanotechnology 217

© 2010 Macmillan Publishers Limited. All rights reserved.
review article Nature nanotechnology doi: 10.1038/nnano.2009.58

a 2+
2+
2 2+ 2 2+
2+ 2+
2+
2+ 2+ 2
2+ 2+
2 2 2
2
2
2+ 2+ 2+ 2 2+
2+ 2+ 2+
2 2+ 2+
2
2 2
2 2+
2 2
2+ 2+
2+ 2+ 2+
2

b 4

Height (nm)
3
2
1
0
0.0 0.5 1.8 1.5 2.0
Distance (μm)

4
Height (nm)

3
2
1
0
0.0 0.5 1.8 1.5 2.0
Distance (μm)
c d
4
Height (nm)

3
2
1
0
0.0 0.5 1.8 1.5 2.0
250 nm Distance (μm)

Figure 1 | Graphite oxide and graphene oxide. a, Chemical structure of graphite oxide30. For clarity, minor functional groups, carboxylic groups and
carbonyl groups have been omitted at the edges. Reproduced with permission from ref. 30. © 1998 Elsevier. b, An AFM image of exfoliated graphene
oxide sheets47; the sheets are ~1 nm thick. The horizontal lines indicate the sections corresponding (in order from top to bottom) to the traces shown on
the right. Reproduced with permission from ref. 47. © 2007 Elsevier. c, Photograph of folded graphene oxide paper7 (© 2007 NPG). d, A scanning electron
microscope image of the cross-section of the graphene oxide paper, showing layered structure7 (© 2007 NPG).

(at concentrations up to 3 mg ml−1)8,34,45,47, affording brown/ SOCl2 activation of the COOH groups) was used in ref. 49 to
dark-brown suspensions. (See Table 1 for a list of solvents used, modify graphene oxides by long alkyl chains with 20 wt% yield.
the concentrations of colloidal suspensions, the lateral dimensions Interestingly, chemical modification of an alternative starting
and heights of graphene oxide sheets, and the type of precur- material, graphite fluoride, with alkyl lithium reagents produced
sor material used, be it graphite oxide or graphite or expandable alkyl-chain-modified graphene sheets that could be dispersed in
graphite.) The exfoliation to achieve graphene oxide sheets has organic solvents after sonication50.
been most typically confirmed by thickness measurements of the
single graphene sheet (~1-nm height on substrates such as mica) Reduced graphene oxides
using atomic force microscopy (AFM). Although the chemical modification of graphene/graphite oxide or
Graphite oxide can be dispersed directly in several polar solvents graphite fluoride can generate homogeneous colloidal suspensions,
such as ethylene glycol, DMF, NMP and THF at about 0.5 mg ml−1 the resulting CMGs are electrically insulating owing to disruption
(ref. 46). It has also been shown that the chemical modification of of the ‘graphitic’ networks. On the other hand, the reduction of the
graphene oxide sheets by organic molecules yields homogeneous graphene oxide by chemical methods (using reductants such as
suspensions in organic solvents45; reaction of graphite oxide with iso- hydrazine47,51,52, dimethylhydrazine9, hydroquinone53 and NaBH4
cyanate groups produced isocyanate-modified graphene oxide sheets (refs 42 and 54), thermal methods55,56 and ultraviolet-assisted
that are well dispersed in polar aprotic solvents. It was proposed that methods57 has produced electrically conducting CMGs. (See
carbamate and amide functional groups are generated by the reac- Table 2 for a list of electrical properties of graphene-based materi-
tion of isocyanate with hydroxyl and carboxyl groups (Fig. 2a)45. als generated using their suspensions.)
The amide-coupling reaction48 between the carboxyl The reduction of aqueous graphene oxide suspension by hydra-
acid groups of graphene oxides and octadecylamine (after zine at the pH of the suspension when used as made results in

218 nature nanotechnology | VOL 4 | APRIL 2009 | www.nature.com/naturenanotechnology

© 2010 Macmillan Publishers Limited. All rights reserved.
Nature nanotechnology doi: 10.1038/nnano.2009.58 review article
Table 1 | Comparison of a set of chemical approaches to produce colloidal suspensions of CMG sheets
Ref. Starting materials Dispersible solvents Concentration (mg ml−1) Lateral size Thickness (nm)
34 GO/MH Water 1 — —
36 GO/MH Water 0.5 Several hundred nm ~1
39 GO/MH Water 0.1 — ~1.7
40 GO/MH Water 7 Several hundred nm ~1
42 GO/H Water/methanol, 3–4 Several hundred nm ~1.2
acetone, acetonitrile
mixed solvents
45 GO/MH DMF, NMP, DMSO, HMPA 1 ~560 nm ~1
46 GO/H Water, acetone, ethanol, 0.5 100–1,000 nm 1.0–1.4
1-propanol, ethylene glycol,
DMSO, DMF, NMP, pyridine,
THF
49 GO/O DMF, THF, CCl4, DCE 0.5 — 0.5–2.5
50 Graphite fluoride DCB, MC, THF 0.002–0.54 1,600 nm ~0.95
51 GO/S DMF, DMAc, NMP 1 Several hundred nm 1.8–2.2
52 GO/MH Hydrazine 1.5 Up to 20 μm × 40 μm ~0.6
54 GO/S THF <0.48 — 1–2

55, 56, 10 GO/S NMP, DMF, DCB, THF, 0.1 100–2,500 nm 1.1–3.5 (ave. 1.75)
nitromethane
57 GO/H Ethanol 1 Several hundred nm ~2
59 Graphite powder NMP, DMAc, GBL, DMEU 0.01 Several μm 1–5
60 GIC NMP 0.15 Several hundred nm ~0.35
61 EG DCE 0.0005 Nanoribbon (width 1–1.8
<10 nm)
62 EG DMF — ~250 nm ~1
63 EG Water, DMF, DMSO 0.015–0.020 Several hundred nm to a 2–3 (2–3 layers of
few μm graphene)
64 Graphite rod DMF, DMSO, NMP 1 500–700 nm ~1.1
GO, graphite oxide; MH, modified Hummers method; H, Hummers method; O, their own method; S, Staudenmaier method; EG, expandable graphite; GIC, graphite intercalation compound; DMF,
dimethylformamide; DMAc, N,N′-dimethylacetamide; DMSO, dimethylsulphoxide; NMP, N-methylpyrrolidone; THF, tetrahydrofuran; MC, dichloromethane; DCE, 1,2-dichloroethane; DCB,
1,2-dichlorobenzene; HMPA, hexamethylphosphoramide; GBL, γ-butyrolactone; DMEU, 1,3-dimethyl-2-imidazolidinone.

agglomerated graphene-based nanosheets, and, when dried, a CMG that was dispersible in DMF, N,N′-dimethylacetamide, and
black powder (Fig. 2b)47 that is electrically conductive (powder NMP at concentrations up to 1 mg ml−1 (ref. 51).
conductivity, ~2 × 102 S m−1). Elemental analysis (atomic C/O Colloidal suspensions of modified graphenes decorated with
ratio, ~10) of the reduced graphene oxides measured by com- small organic molecules or nanoparticles have also been reported.
bustion revealed the existence of a significant amount of oxygen, In ref. 39, the reduced graphene oxide sheets were functionalized
indicating that reduced graphene oxide is not the same as pris- using pyrenebutyric acid (a well-known organic molecule48 with
tine graphene. Theoretical calculations of the reduction of graph- a strong adsorption affinity for the graphitic plane via π stack-
ene oxide (the model used for graphene oxide had the graphene ing). The aqueous graphene oxide suspension was reduced using
decorated with hydroxyl and epoxide groups) suggest that reduc- hydrazine in the presence of pyrenebutyric acid, yielding a black
tion below 6.25% of the area of the graphene oxide (C/O = 16 in aqueous colloidal suspension (0.1 mg ml−1) of CMG adsorbed by
atomic ratio) may be difficult in terms of removing the remaining pyrenebutyric acid. Its paper-like material, prepared by filtration,
hydroxyl groups58. showed moderate electrical conductivity (2 × 102 S m−1).
Homogeneous colloidal suspensions of electrically conducting The suspension (<0.48 mg ml−1) of gold-nanoparticle-modified
CMGs have been produced by chemical reduction with dimethyl­ graphene sheets in THF was generated by the reaction of NaBH4
hydrazine or hydrazine in the presence of either polymer or and octadecylamine-modified graphene oxide49 and then the addi-
−
surfactant9,34. The reduction of an aqueous suspension contain- tion of AuCl4 to the suspension54. The gold nanoparticles (diam-
ing a mixture of graphene oxide sheets and poly(sodium 4-sty- eter, ~5–11 nm) were anchored to the modified graphene sheets.
renesulphonate) afforded an aqueous black suspension of reduced Graphene modified with titanium dioxide nanoparticles has also
graphene oxide sheets coated in the polymer34. The reduction of been studied57; ultraviolet irradiation of this TiO2/graphene oxide
isocyanate-modified graphene oxide in the presence of polystyrene hybrid in ethanol was used to reduce the graphene oxide sheets
generated a suspension of reduced graphene oxide sheets in DMF (Fig. 2d), producing a black suspension of TiO2-attached CMG
that could then be ‘crashed out’ with methanol to yield a compos- sheets in ethanol. The TiO2 was suggested to act as a photocata-
ite with well-dispersed and electrically conductive CMG sheets9. lyst, transferring photoelectrons from the TiO2 to the graphene
Reduction of sodium dodecylbenzenesulfonate (SDBS)-wrapped oxide sheets57.
graphene oxide with hydrazine and then its chemical modifica- A few methods for creating colloidal suspensions of graph-
tion (Fig. 2c) with aryl diazonium salt produced SDBS-wrapped ene sheets without the help of stabilizers or surfactants have

nature nanotechnology | VOL 4 | APRIL 2009 | www.nature.com/naturenanotechnology 219

© 2010 Macmillan Publishers Limited. All rights reserved.
review article Nature nanotechnology doi: 10.1038/nnano.2009.58

2 2+ 2 2+
+2 2
a
2
2
2+ 2+
2 2 2+
2+

51&2
51&2
&2
5
1+
2
2
5 2 2 1 5
2 2 2 +
+1
2
2
+ +1 5
1 2 +
2 + 2 2 1
5 2 1 5
5
2
2

b c R

OH N2H4 H2O pH 10
O OH H
80 °C, 24 h
O
H COOH COOH
H
– +
O OH BF4N2 R H
O

SDBS-wrapped GO 1a–4a
RT, 1 h
1a, b R = CI R R
250 nm 2a, b R = NO2
3a, b R = OCH3 1b–4b
4a, b R = Br

e
e

(1) (2) (3)

h–

h TiO2

Figure 2 | CMG oxide sheets. a, Proposed reaction of graphene oxide sheets with isocyanates forming carbamate (left oval) and amide (right oval)
functionalities45. b, A scanning electron microscope image of aggregated graphene oxide sheets chemically reduced with hydrazine monohydrate47. Parts
a and b reproduced with permission from ref. 47. © 2007 Elsevier. c, Starting with SDBS-wrapped graphene oxide, reduction and functionalization of
intermediate SDBS-wrapped CMG with diazonium salts51. RT, room temperature. Reproduced with permission from ref. 51. © 2008 ACS. d, TiO2–graphene
hybrid and its proposed response under UV excitation57. Reproduced with permission from ref. 57. © 2008 ACS. e, Chemical route to produce aqueous
suspension of reduced graphene oxide36. (1) Oxidation of graphite to synthesize graphite oxide. (2) Exfoliation of graphene oxide in water by sonication
of graphite oxide. (3) Controlled reduction of graphene oxide sheets by hydrazine yielding a colloidal suspension of conductive CMG sheets, which are
stabilized by electrostatic repulsion (© 2008 NPG).

been reported. An aqueous suspension (0.5 mg ml−1) of reduced that shifting to pH 10 converts neutral carboxylic groups to nega-
graphene oxide sheets under basic conditions (pH 10) was tively charged carboxylate groups, so that when the interior of the
described in ref. 36. The graphene oxide was reduced by hydrazine, graphene oxide sheets are reduced by hydrazine, the negatively
and excess hydrazine was removed by dialysis. It was suggested charged particles do not agglomerate36; see Fig. 2e. Paper-like

220 nature nanotechnology | VOL 4 | APRIL 2009 | www.nature.com/naturenanotechnology

© 2010 Macmillan Publishers Limited. All rights reserved.
Nature nanotechnology doi: 10.1038/nnano.2009.58 review article
Table 2 | Electrical properties of conductive CMG-based materials
Ref. Forms Electrical properties
36, 41 Free-standing paper Conductivity (102 S cm−1): 72 (air dried),
120 (dried at 220 °C), 350 (dried at 500 °C)
39 Free-standing paper Conductivity: 2 × 102 S m−1
40 Free-standing paper Conductivity: 6.9 × 102 S m−1
42 Thin evaporated film (~3 μm thick) on a glass slide Conductivity: 12.5 × 102 S m−1
47 Powder Conductivity: 2 × 102 S m−1
55, 56 Powder Conductivity: (10–23) × 102 S m−1
57 Drop-cast film on a substrate Resistance: 30.5 kΩ
59 Thin film (~30 nm thick) on the alumina membrane by filtration Conductivity: 65 × 102 S m−1
61 Single graphene nanoribbon sheet Semiconducting behaviour
62 Multilayer Langmuir–Blodgett film Resistance: 8–150 kΩ

materials made by filtration and then dried in air showed good able graphites allows production of dispersions of high-quality
electrical conductivity (~7,200 S m−1)36. graphene sheets (almost pristine graphene; see below). Colloidal
In other work, an aqueous suspension (7 mg ml−1) of partially suspensions of graphene sheets in organic solvents such as
reduced graphene oxides was produced, under basic conditions, by NMP were obtained by sonication of graphite powder in ref. 59.
hydrazine reduction of KOH-modified graphene oxides40. It was Although the concentration (0.01 mg ml−1) of the suspension and
suggested that the K+ ions and the carboxylate anions at the edges yields (1–12 wt%) of single-layer graphene were not high, high-
of the CMG sheets formed ion pairs. The electrical conductivity of quality graphene was reported. ‘Good’ solvents for a homogeneous
air-dried ‘paper’ made from this material was ~690 S m−1. colloidal suspension were classified using a model that included
In a third approach42, a graphene-based suspension was pro- the surface energy of graphene and the cohesive energy of the
duced in three reaction steps: (1) ‘pre-reduction’ of graphene oxide solvent. Measurements on very thin films made from these sheets
by NaBH4; (2) sulphonation by aryl diazonium salt of sulphanilic yielded, for example, an electrical conductivity of ~6,500 S m−1 at
acid; and (3) ‘post-reduction’ of aqueous sulphonated graphene ~42% of optical transparency. After the films had been dried at
(2 mg ml−1, pH 3–10) with hydrazine. The sulphonated graphene 400 °C, residual NMP had not been completely removed and was
sheets of step 2 could be dispersed in water (2 mg ml−1, pH 3–10), estimated to be 7% by weight59.
and after the post-reduction step the CMG sheets could be dis- In ref. 60, a homogeneous colloidal suspension of graphene
persed in mixed solvents of water with certain organic solvents. It sheets or ribbons was produced by stirring the ternary potassium
was suggested that covalent functionalization in the CMG sheets salt K(THF)xC24 (which is a graphite intercalation compound) in
by sulphonyl groups was occurring. The electrical conductivity of NMP. The sheets or ribbons (length, ~40 mm) were measured to
the thin evaporated film on a glass slide after drying at 120 °C was be 0.35–0.4 nm thick (ambient scanning tunnelling microscopy
reported to be 1,250 S m−1 (ref. 42). on highly oriented pyrolytic graphite substrate(s) and by ambi-
A suspension of reduced graphene oxide sheets (some with lat- ent AFM on mica substrate(s)) close to the interlayer separation
eral dimensions of up to ~20 mm × 40 mm) in anhydrous hydra- in graphite 0.335 nm thick. The suspension was air sensitive, but
zine was produced by stirring graphene oxide paper in anhydrous after drying by being dropped onto a surface, the resulting deposits
hydrazine (10 ml for 15 mg of graphite oxide) for one week in a were stable when exposed to air.
nitrogen-filled dry box52. The resulting CMG on SiO2 was reported Polymer-coated graphene sheets in organic solvents obtained
to be ~0.6 nm in height, measured by AFM. The anhydrous using expandable graphite have been reported in refs 61 and 62.
hydrazine thus acts both as a reduction agent of graphene oxide Commercial expandable graphite was thermally treated at high
sheets and as a dispersing solvent in this approach. The individual temperature (1,000 °C) for a short time (1 min)61. Further sonica-
CMG sheets were tested as a field-effect transistor and showed tion of the expanded material in dichloroethane solution of poly(m-
p-type behaviour52. phenylenevinylene-co-2,5-dioctoxy-p-phenylenevinylene) (PmPV)
Thermal treatment of graphite oxide is another route that produced a suspension (0.1 mg ml−1) of semiconducting single-
has been used to obtain reduced platelets. Rapid heating layer or few-layer CMG nanoribbons (of width less than ~10 nm)
(>2,000 °C min−1) up to 1,050 °C exfoliates as well as reduces with adsorbed polymer61. Another approach taken was exfoliation
graphite oxide, yielding a black powder55,56. The platelets have a by heating at 1,000 °C for 1 min, re-intercalation with oleum and
similar oxygen content (atomic C/O ratio, 10.3; elemental analysis expansion with tetrabutylammonium hydroxide (Fig. 3a)62. The final
by combustion) to that of hydrazine-reduced graphene oxide, and suspension of graphene sheets, coated in 1,2-distearoyl-sn-glycero-
a powder conductivity in the range of (1–2.3) × 103 S m−1 has been 3-phosphoethanolamine-N-[methoxy(polyethyleneglycol)-5000]
reported55,56. The thermally reduced graphene sheets could be dis- (DSPE-mPEG), was produced by sonication in DMF, with ~90%
persed in several organic solvents (0.1 mg ml−1) and their colloidal of the sheets reported as individual CMG sheets on the basis of
suspension was used to fabricate a set of composites of reduced AFM measurements62.
graphenes and polymers10. An aqueous suspension of organic-molecule-coated graph-
ene sheets using expandable graphite was reported in ref. 63. The
Graphenes from other graphite derivatives expandable graphite was rapidly heated to 1,000 °C for 1 min and
Graphite, graphite intercalation compounds or expandable the resulting expanded graphite was mixed with 7,7,8,8-tetra­
graphites (which are a particular type of graphite intercalation cyanoquinodimethane (TCNQ) in the presence of several drops
compound) have been used as starting materials in obtaining of dimethyl sulphoxide (DMSO). Then a black suspension (15–
colloidal dispersions of single-layer graphene sheets. Ideally, the 20 mg ml−1) of single- or few-layer TCNQ-coated graphene sheets
use of graphite, graphite intercalation compounds or expand- in water was produced by sonication of the dried mixture. It was

nature nanotechnology | VOL 4 | APRIL 2009 | www.nature.com/naturenanotechnology 221

© 2010 Macmillan Publishers Limited. All rights reserved.
review article Nature nanotechnology doi: 10.1038/nnano.2009.58

a X X
(1) (2)

X X
TBA
X X X X
DMF
X X

X X

O
(3)
O
O

n
X O
O

n
Sonication -O
P O
O O
O P -
Phospholipid-PEG X O
OO O
O
O O OO X
O
X

Graphene sheet
X

b
Graphite rod anode

A
+ –
Water

Ionic
Graphite rods liquid

Ionic liquid and water Mixture

Figure 3 | Graphene sheets from graphite derivatives. a, Production of colloidal suspension of polymer-coated CMG sheets from graphite intercalation
compounds62. (1) Exfoliated graphite reintercalated with sulphuric acid between the layers. (2) Insertion of tetrabutylammonium hydroxide (TBA) into the
intercalated graphite. PEG, polyethyleneglycol. (3) CMG sheets coated with polymer molecules and a photo of the suspension in dimethylformamide.
(© 2008 NPG). b, Graphene sheets ionic-liquid-modified by electrochemistry using graphite electrodes64. Left, experimental set-up diagram; right,
exfoliation of the CMG sheets from the graphite anode. Reproduced with permission from ref. 64. © 2008 Wiley.

suggested that TCNQ anions adsorb on graphene sheets and after sonication. The average thickness of the CMG sheets from
stabilize them in water63. The TCNQ-stabilized graphene sheets such a DMF suspension was measured by AFM and was ~1.1 nm.
were also dispersed in highly polar solvents (DMF and DMSO). It was proposed that the imidazolium radicals, electrochemically
Electrochemical treatment of graphite was used in ref. 64 to driven, covalently attached to graphene sheets through the break-
generate a colloidal suspension of CMG sheets. The commer- ing of π bonds64.
cial graphite rods used as anode and cathode electrodes were
immersed in a phase-separated mixture of water and imidazolium- Challenges and perspectives
based ionic liquids (Fig. 3b). After potentials of 10–20 V had been Colloidal suspensions are of great importance in the preparation
applied, it was claimed that ionic-liquid-functionalized graphene of many types of materials, and the suspensions of CMGs hold
sheets originating from the graphite anode precipitated. Dried great promise in this regard. Looming issues in terms of wide-
CMG could be re-dispersed in polar aprotic solvents (1 mg ml−1) scale applicability include scalability (yield, quantity, cost, etc.),

222 nature nanotechnology | VOL 4 | APRIL 2009 | www.nature.com/naturenanotechnology

© 2010 Macmillan Publishers Limited. All rights reserved.
Nature nanotechnology doi: 10.1038/nnano.2009.58 review article
the safety of the solvents used and the removal from the product 20. Reina, A. et al. Large area, few-layer graphene films on arbitrary substrates by
material (if necessary) of residual solvents or stabilizers used in chemical vapor deposition. Nano Lett. 9, 30–35 (2009).
21. Dato, A., Radmilovic, V., Lee, Z., Phillips, J. & Frenklach, M. Substrate-free
the colloid. It is also worth emphasizing that although colloidal
gas-phase synthesis of graphene sheets. Nano Lett. 8, 2012–2016 (2008).
suspensions are normally only regarded as stable if they persist 22. Wang, J. J. et al. Free-standing subnanometer graphite sheets. Appl. Phys. Lett.
for very long periods of time, dispersions of CMGs might remain 85, 1265–1267 (2004).
stable long enough to be processed into something else. 23. Campos-Delgado, J. et al. Bulk production of a new form of sp carbon:
Another critical issue is related to our understanding of the crystalline graphene nanoribbons. Nano Lett. 8, 2773–2778 (2008).
chemical structure(s) of CMG sheets and their reaction mecha- 24. Choucair, M., Thordarson, P. & Stride, J. A. Gram-scale production of
graphene based on solvothermal synthesis and sonication. Nature Nanotech.
nisms. The better our knowledge of the chemistry of these
4, 30–33 (2009).
materials, the better the graphene-based composites, thin films, 25. Brodie, B. C. Sur le poids atomique du graphite. Ann. Chim. Phys. 59, 466 (1860).
paper-like materials and so on that we can make. For example, the 26. Schafhaeutl, C. On the combination of carbon with silicon and iron, and other
prospects for sensors based on CMG65 will hinge on our ability to metals, forming the different species of cast iron, steel, and malleable iron.
chemically tune the CMG for each sensing modality. Phil. Mag. 16, 570–590 (1840).
So far the graphenes derived by the reduction of graphene oxide 27. Staudenmaier, L. Verfahren zur Darstellung der Graphitsaure. Ber. Deut. Chem.
Ges. 31, 1481 (1898).
have contained a significant amount of oxygen and, possibly, sig-
28. Hummers, W. S. & Offeman, R. E. Preparation of graphitic oxide. J. Am. Chem.
nificant numbers of defects. Thermal annealing of reduced graphene Soc. 80, 1339 (1958).
oxide sheets has produced enhanced results62,66, and finding routes for 29. He, H., Riedl, T., Lerf, A. & Klinowski, J. Solid-state NMR studies of the
complete restoration of the sp2 carbon network of pristine graphene structure of graphite oxide. J. Phys. Chem. 100, 19954–19958 (1996).
is of interest. (The graphenes produced from graphite intercalation 30. He, H., Klinowski, J., Forster, M. & Lerf, A. A new structural model for graphite
compounds or expandable graphite may have fewer defects, although oxide. Chem. Phys. Lett. 287, 53–56 (1998).
31. Lerf, A., He, H., Forster, M. & Klinowski, J. Structure of graphite oxide
they are also produced in lower yields and are less amenable to func-
revisited. J. Phys. Chem. B 102, 4477–4482 (1998).
tionalization than graphenes derived from graphene oxide.) 32. Cai, W. et al. Synthesis and solid-state NMR structural characterization of
Finally, we mention the development of new reaction routes 13
C-labeled graphite oxide. Science 321, 1815–1817 (2008).
and starting materials as an alternative. The worldwide supply 33. Buchsteiner, A., Lerf, A. & Pieper, J. Water dynamics in graphite oxide
of natural graphite has been estimated at 800,000,000 tonnes67. investigated with neutron scattering. J. Phys. Chem. B 110, 22328–22338 (2006).
If graphene or very thin platelets of multilayer graphene could 34. Stankovich, S. et al. Stable aqueous dispersions of graphitic nanoplatelets via
the reduction of exfoliated graphite oxide in the presence of poly(sodium
be produced on a large scale by CVD from various precursors,
4-styrenesulfonate). J. Mater. Chem. 16, 155–158 (2006).
new routes for creating colloidal suspensions might also be 35. Jung, I. et al. Simple approach for high-contrast optical imaging and
found, and the supply of graphene/few-layer graphene might be characterization of graphene-based sheets. Nano Lett. 7, 3569–3575 (2007).
enormously increased. 36. Li, D., Muller, M. B., Gilje, S., Kaner, R. B. & Wallace, G. G. Processable
aqueous dispersions of graphene nanosheets. Nature Nanotech.
3, 101–105 (2008).
References 37. Boehm, H. P., Clauss, A., Fischer, G. O. & Hofmann, U. Das
1. Geim, A. K. & Novoselov, K. S. The rise of graphene. Nature Mater. Adsorptionsverhalten sehr dunner Kohlenstoff-Folien. Anorg. Allg. Chem.
6, 183–191 (2007). 316, 119–127 (1962).
2. Lee, C., Wei, X., Kysar, J. W. & Hone, J. Measurement of the elastic properties 38. Boehm, H. P., Eckel, M. & Scholz, W. Uber den Bildungsmechanismus des
and intrinsic strength of monolayer graphene. Science 321, 385–388 (2008). Graphitoxids. Anorg. Allg. Chem. 353, 236–242 (1967).
3. Balandin, A. A. et al. Superior thermal conductivity of single-layer graphene. 39. Xu, Y., Bai, H., Lu, G., Li, C. & Shi, G. Flexible graphene films via the filtration
Nano Lett. 8, 902–907 (2008). of water-soluble noncovalent functionalized graphene sheets. J. Am. Chem. Soc.
4. Bolotin, K. I. et al. Ultrahigh electron mobility in suspended graphene. Solid 130, 5856–5857 (2008).
State Commun. 146, 351–355 (2008). 40. Park, S. et al. Aqueous suspension and characterization of chemically modified
5. Stoller, M. D., Park, S., Zhu, Y., An, J. & Ruoff, R. S. Graphene-based graphene sheets. Chem. Mater. 20, 6592–6594 (2008).
ultracapacitors. Nano Lett. 8, 3498–3502 (2008). 41. Chen, H., Muller, M. B., Gilmore, K. J., Wallace, G. G. & Li, D. Mechanically
6. Zhang, Y., Tan, Y.-W., Stormer, H. L. & Kim, P. Experimental observation of the strong, electrically conductive, and biocompatible graphene paper. Adv. Mater.
quantum Hall effect and Berry’s phase in graphene. Nature 438, 201–204 (2005). 20, 3557–3561 (2008).
7. Dikin, D. A. et al. Preparation and characterization of graphene oxide paper. 42. Si, Y. & Samulski, E. T. Synthesis of water soluble graphene. Nano Lett. 8,
Nature 448, 457–460 (2007). 1679–1682 (2008).
8. Park, S. et al. Graphene oxide papers modified by divalent ions—Enhancing 43. Eda, G., Fanchini, G. & Chhowalla, M. Large-area ultrathin films of reduced
mechanical properties via chemical cross-linking. ACS Nano 2, 572–578 (2008). graphene oxide as a transparent and flexible electronic material. Nature
9. Stankovich, S. et al. Graphene-based composite materials. Nature Nanotechnol. 3, 270–274 (2008).
442, 282–286 (2006). 44. Wang, X., Zhi, L. & Müllen, K. Transparent, conductive graphene electrodes for
10. Ramanathan, T. et al. Functionalized graphene sheets for polymer dye-sensitized solar cells. Nano Lett. 8, 323–327 (2008).
nanocomposites. Nature Nanotech. 3, 327–331 (2008). 45. Stankovich, S., Piner, R., Nguyen, S. T. & Ruoff, R. S. Synthesis and
11. Blake, P. et al. Graphene-based liquid crystal device. Nano Lett. exfoliation of isocyanate-treated graphene oxide nanoplatelets. Carbon
8, 1704–1708 (2008). 44, 3342–3347 (2006).
12. Bunch, J. S. et al. Electromechanical resonators from graphene sheets. Science 46. Paredes, J. I., Villar-Rodil, S., Martinez-Alonso, A. & Tascón, J. M. D. Graphene
315, 490–493 (2007). oxide dispersions in organic solvents. Langmuir 24, 10560–10564 (2008).
13. Eizenberg, M. & Blakely, J. M. Carbon monolayer phase condensation on 47. Stankovich, S. et al. Synthesis of graphene-based nanosheets via chemical
Ni(111). Surf. Sci. 82, 228–236 (1970). reduction of exfoliated graphite oxide. Carbon 45, 1558–1565 (2007).
14. Aizawa, T., Souda, R., Otani, S., Ishizawa, Y. & Oshima, C. Anomalous bond of 48. Tasis, D., Tagmatarchis, N., Bianco, A. & Prato, M. Chemistry of carbon
monolayer graphite on transition-metal carbide surfaces. Phys. Rev. Lett. nanotubes. Chem. Rev. 106, 1105–1136 (2006).
64, 768–771 (1990). 49. Niyogi, S. et al. Solution properties of graphite and graphene. J. Am. Chem. Soc.
15. Novoselov, K. S. et al. Electric field effect in atomically thin carbon films. 128, 7720–7721 (2006).
Science 306, 666–669 (2004). 50. Worsley, K. A. et al. Soluble graphene derived from graphite fluoride. Chem.
16. Lu, X., Yu, M., Huang, H. & Ruoff, R. S. Tailoring graphite with the goal of Phys. Lett. 445, 51–56 (2007).
achieving single sheets. Nanotechnology 10, 269–272 (1999). 51. Lomeda, J. R., Doyle, C. D., Kosynkin, D. V., Hwang, W.-F. & Tour, J. M.
17. Berger, C. et al. Electronic confinement and coherence in patterned epitaxial Diazonium functionalization of surfactant-wrapped chemically converted
graphene. Science 312, 1191–1196 (2006). graphene sheets. J. Am. Chem. Soc. 130, 16201–16206 (2008).
18. Kim, K. S. et al. Large-scale pattern growth of graphene films for stretchable 52. Tung, V. C., Allen, M. J., Yang, Y. & Kaner, R. B. High-throughput solution
transparent electrodes. Nature 457, 706–710 (2009). processing of large-scale graphene. Nature Nanotech. 4, 25–29 (2008).
19. Sutter, P. W., Flege, J.-I. & Sutter, E. A. Epitaxial graphene on ruthenium. 53. Wang, G. et al. Facile synthesis and characterization of graphene nanosheets.
Nature Mater. 7, 406–411 (2008). J. Phys. Chem. C 112, 8192–8195 (2008).

nature nanotechnology | VOL 4 | APRIL 2009 | www.nature.com/naturenanotechnology 223

© 2010 Macmillan Publishers Limited. All rights reserved.
review article Nature nanotechnology doi: 10.1038/nnano.2009.58
54. Muszynski, R., Seger, B. & Kamat, P. V. Decorating graphene sheets with gold 63. Hao, R., Qian, W., Zhang, L. & Hou, Y. Aqueous dispersions of TCNQ-anion-
nanoparticles. J. Phys. Chem. C 112, 5263–5266 (2008). stabilized graphene sheets. Chem. Commun. 6576–6578 (2008).
55. Schniepp, H. C. et al. Functionalized single graphene sheets derived form 64. Liu, N. et al. One-step ionic-liquid-assisted electrochemical synthesis of ionic-
splitting graphite oxide. J. Phys. Chem. B 110, 8535–8539 (2006). liquid-functionalized graphene sheets directly from graphite. Adv. Funct. Mater.
56. McAllister, M. J. et al. Single sheet functionalized graphene by oxidation and 18, 1518–1525 (2008).
thermal expansion of graphite. Chem. Mater. 19, 4396–4404 (2007). 65. Liu, Z., Robinson, J. T., Sun, X. & Dai, H. PEGylated nanographene
57. Williams, G., Serger, B. & Kamat, P. V. TiO2-graphene nanocomposites. UV-assisted oxide for delivery of water-insoluble cancer drugs. J. Am. Chem. Soc. 130,
photocatalytic reduction of graphene oxide. ACS Nano 2, 1487–1491 (2008). 10876–10877 (2008).
58. Boukhvalov, D. W. & Katsnelson, M. I. Modeling of graphite oxide. J. Am. 66. Jung, I. et al. Effect of water vapor on electrical properties of individual reduced
Chem. Soc. 130, 10697–10701 (2008). graphene oxide sheets. J. Phys. Chem. C 112, 20264–20268 (2008).
59. Hernandez, Y. et al. High-yield production of graphene by liquid-phase 67. Olson, D. W. Graphite (natural). USGS Mineral Commodity Summary 2008,
exfoliation of graphite. Nature Nanotech. 3, 563–568 (2008). 74–75 (US Geological Survey, 2008); available at <http://minerals.usgs.gov/
60. Valles, C. et al. Solutions of negatively charged graphene sheets and ribbons. minerals/pubs/commodity/graphite/mcs-2008-graph.pdf>.
J. Am. Chem. Soc. 130, 15802–15804 (2008).
61. Li, X., Wang, X., Zhang, L., Lee, S. & Dai, H. Chemically derived, ultrasmooth
graphene nanoribbon semiconductors. Science 319, 1229–1232 (2008). Acknowledgements
62. Li, X. et al. Highly conducting graphene sheets and Langmuir–Blodgett films. R.S.R. acknowledges previous support from NASA and current support from the state of
Nature Nanotech. 3, 538–542 (2008). Texas, and DARPA iMINT and DARPA CERA for support of graphene-related research.

224 nature nanotechnology | VOL 4 | APRIL 2009 | www.nature.com/naturenanotechnology

© 2010 Macmillan Publishers Limited. All rights reserved.
 LETTERS
PUBLISHED ONLINE: 7 DECEMBER 2008 | DOI: 10.1038/NNANO.2008.365

Gram-scale production of graphene based on

solvothermal synthesis and sonication
Mohammad Choucair1, Pall Thordarson1 and John A. Stride1,2 *

Carbon nanostructures have emerged as likely candidates for a

wide range of applications, driving research into novel synthetic
techniques to produce nanotubes, graphene and other carbon-
based materials. Single sheets of pristine graphene have been
isolated from bulk graphite in small amounts by micromechani-
cal cleavage1, and larger amounts of chemically modified
graphene sheets have been produced by a number of
approaches2–7. Both of these techniques make use of highly
oriented pyrolitic graphite as a starting material and involve
labour-intensive preparations. Here, we report the direct
chemical synthesis of carbon nanosheets in gram-scale
quantities in a bottom-up approach based on the common
laboratory reagents ethanol and sodium, which are reacted to
give an intermediate solid that is then pyrolized, yielding a
fused array of graphene sheets that are dispersed by mild soni-
cation. The ability to produce bulk graphene samples from non-
graphitic precursors with a scalable, low-cost approach should
take us a step closer to real-world applications of graphene.
Graphite is the most common allotrope of carbon. The name is
derived from the Greek verb graphein, ‘to write’, which relates very Figure 1 | Example of the bulk quantity of graphene product. The image
literally to the compound we now know as graphene, as single sheets consists of approximately 2 g of sample.
of graphene were first isolated by simply tracing a sample of bulk
graphite across a substrate in a process known as micromechanical and crystallinity, suggesting an amorphous or disordered
cleavage1. Although theoretical work by Landau8 and others graphitic structure18,19.
suggests that true two-dimensional structures cannot be isolated, The graphene sheets produced herein have been characterized by
recent work has resolved this apparent anomaly by showing that a number of physical methods that clearly demonstrate the nanos-
free-standing graphene sheets are actually buckled, with the tructured form of the carbon product. Representative transmission
ripples extending into the third spatial dimension9. electron microscopy (TEM) images of the as-synthesized graphene
Further attempts to synthesize isolated graphene have been based samples post-sonication are shown in Fig. 2, and scanning electron
on intercalation2, sonication in various solvents3,4 and the chemical microscopy (SEM) images of the sample before sonication are
reduction of graphite to yield few-layer graphite oxide5–7,10. shown in Fig. 3. The Raman spectra of the synthesized graphene
Approaches developed for the production of carbon nanotubes were recorded using both green (514 nm) and red (633 nm) laser
have also been explored, but so far they have only been able radiation, along with samples of natural graphite and charcoal for
to produce graphitic film11. However, single- and few-layer comparison (see Supplementary Information, Fig. S2). The
graphene sheets have been grown epitaxially by the chemical spectra for natural graphite and charcoal agree well with the litera-
vapour deposition of hydrocarbons on metal and nonmetal sub- ture20, and the graphene samples gave Raman signals in the green
strates12–14, substrate-free deposition15 and by thermal decompo- laser similar to that of highly defective graphite—a broad D-band
sition of SiC16,17. centred at 1,353 cm21 and a G-band at 1,590 cm21, which have
Given the lack of a reliable top-down approach for the large-scale similar intensities such that IG/ID  1.16. Samples characterized
production of graphene, attention has turned to bottom-up in the red laser showed similar results with an increased intensity
approaches that might be able to deliver the economies of scale of the D-peak across all samples.
that are found in the chemical and pharmaceutical industries. We The definitive evidence that the carbon sheets obtained by this
have demonstrated that single-layer graphene can be synthesized method are only a single atomic layer thick was obtained with
by low-temperature flash pyrolysis of a solvothermal product of atomic force microscopy (AFM) (Fig. 4; see also Supplementary
sodium and ethanol, followed by gentle sonication of the nanopor- Information, Figs S15 and S16). Multiple cross-sections were
ous carbon product (see Methods). The product of this reaction is recorded of the sample laid onto a mica substrate, and a number
then washed in water and dried to obtain pure graphene (Fig. 1). of step heights were measured across the relief of the sample. The
Solvothermal reactions have been reported to lead to the production step heights measured between the surface of the sheets and the sub-
of carbon nanosheets, although such materials do not appear strate were consistently found to be 4+1 Å, proving them to be only
to consist of single sheets having both high degrees of planarity a single atom thick.

1
School of Chemistry, University of New South Wales, Sydney 2052, Australia, 2 Bragg Institute, Australian Nuclear Science and Technology Organisation,
PMB 1, Menai, New South Wales 2234, Australia; *e-mail: j.stride@unsw.edu.au

30 NATURE NANOTECHNOLOGY | VOL 4 | JANUARY 2009 | www.nature.com/naturenanotechnology

© 2009 Macmillan Publishers Limited. All rights reserved.
NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2008.365 LETTERS

Figure 2 | TEM images of the agglomerated graphene sheets. a,b, The same sample region is viewed at different magnifications and clearly shows the
extent of sheet formation and the tendency for sheets to coalesce into overlapped regions. Scale bars, 200 nm (a) and 1,000 nm (b). An inherent sheet-like
structure displaying an intricate long-range array of folds is evident. As the images are taken in transmission mode, the relative opacity of individual sheets
is a result of interfacial regions with overlap between individual sheets. The sheets extend in lateral dimensions over micrometre length scales, ranging from
1  1027 to greater than 1  1025 m.

Figure 3 | SEM image of the as-synthesized graphene structures. a,b, The bulk graphene product obtained from the pyrolysis of the solvothermal product is
highly porous, but consists of individual sheets. Scale bars, 15 mm (a) and 6 mm (b). The entire image consists of individual graphene sheets held into a
porous structure that typically extends over more than 1  1024 m, with the presence of numerous cavities, or holes. The graphene is therefore initially
obtained as fused sheets, weakly held into a foam-like structure that is then easily separated into individual sheets by sonication in ethanol for
several minutes.

The crystallinity of the sheets was determined with selected area value of 1,692 m2 g21. The apparent disparity is due to the dissipation
electron diffraction (SAED; Fig. 5), which showed that over length of free sheets in dilute solution in contrast to the solid state.
scales that are at least the order of the coherence length of the elec- Finally, X-ray photoemission spectroscopy (XPS) indicated that the
tron beam, the graphene sheets are crystalline. Indeed the image initial crude carbon product obtained directly from the pyrolysis of the
shows a number of features that demonstrate the sheet-like
p nature solvothermal ethoxide product contained sodium oxide, Na2O, as a
of the material. Only the in-plane hk(d/n) and h0( 3d/n) graphite 16.57% impurity by weight. However, further washing with distilled
reflections were observed, where d is the C – C bond length of graph- water removed this completely. The elemental analysis of the residual
ite (1.35 Å), with no indication of the (00l ) peaks that result from graphene, as determined by XPS, was 86.4% C and 13.6% O by atomic
the interplanar correlations. The fact that the diffraction spots are composition (or by mass, 82.65% C and 17.35% O), and natural
relatively diffuse and of constant intensity indicates that the graphite was determined to be 97.8% C, 1.8% O, 0.2% Cl and
sample consists of free sheets that may not be exactly perpendicular 0.2% S by atomic composition (by mass, 96.6% C, 2.3% O, 0.5% Cl
to the incident beam9. and 0.6% S). These figures should be compared with those obtained
The bulk conductivity of the material when pressed into a disc of from elemental analysis by combustion (see Supplementary
15 mm diameter was 0.05 S m21, which is higher than the value Information): graphene was 78.5% C, 2.7% H and 18.8% O; graphite
for charcoal (1  1027 S m21; ref. 21) but much lower than that was 94.5% C, 0.1% H and 5.4% O. Most of the oxygen present in all
for graphite (3  104 S m21; see Supplementary Information of these samples is believed to be in the form of adsorbed water: this
and ref. 22). This is consistent with highly conductive sheets conclusion is supported by the release of comparable quantities
having multiple interfaces rather than a highly disordered amor- of oxygen under thermo-gravimetric analysis (TGA), with a
phous product. In addition, surface area measurements indicate 7.8% weight loss of the graphene product being recorded at around
that in the solid state, the fused graphene structure has a 110 8C (see Supplementary Information, Fig. S2).
Langmuir23 (nitrogen, 77 K) surface area of 612 m2 g21, whereas in Although the overall synthetic process is conceptually very
dispersed ethanol solutions, methylene blue adsorption24 led to a simple, a delicate interplay has been identified in the solvothermal

NATURE NANOTECHNOLOGY | VOL 4 | JANUARY 2009 | www.nature.com/naturenanotechnology 31

© 2009 Macmillan Publishers Limited. All rights reserved.
LETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2008.365

250
11
Sample
200 Substrate

150

Intensity
22

100
33
b 50

0
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
2θ (deg)
250
10
Sample
c 200 20 Substrate

30
150

Intensity
10 100 40
8
50
Height (Å)

6 50 60
4 0
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
2
2θ (deg)
0
0 100 200 300 400 500 600
Distance along x-axis (nm) Figure 5 | SAED pattern of graphene at 300 keV. a,b, Integrated intensities
p
of the in-plane hk (d/n) reflections (a) and the h0 ( 3d/n) reflections (b).
10
Peaks are labelled with the Miller indices based upon a hexagonal lattice.
8 c, Digitized image of a photographic image plate of the same data, showing
Height (Å)

6
the two sets of overlapping hk and h0 reflections, off set by 198 as a result
of the exposure of two graphene sheets. The image is a composite of two
4 patterns, rotated relative to one another by 198, with one strong (h0)
2 reflection leading to secondary diffraction peaks about one of the reciprocal
lattice points, which are also rotated by 198 to one of the main series of
0
0 100 200 300 400 500 600 lines. This is consistent with the beam striking a region of the sample in
Distance along x-axis (nm) which there is a partial overlap of two sheets with no interplanar stacking
coherence between them and in which the lower layer (the second sheet in
Figure 4 | AFM image of graphene in tapping mode. a, Topography image the direction of travel of the beam) is slightly rotated (198) with respect to
(600  600 nm; height scale, 0–10 Å; scale bar, 100 nm). b,c, Height the first.
profiles (600 nm along the x-axis) obtained from positions b and c indicated
by white arrows in a. To the bottom of the image, a narrow ridge, of around to the small isolated nature of the trapped ethanol as ignition
1  1027 m width, extends beyond the bulk of the sheet. The profile across points, nucleation of the sheets occurs around those regions rich
this point of the image is fully consistent with those observed elsewhere and in ethanol, in what we have labelled a ‘popcorn effect’. The presence
clearly shows the narrow valley between the ridge and the neighbouring of excess ethanol (30%) in the clathrate-like structure has been
sheet. The presence of irregularly shaped sheets is a result of the sonication found to be thermodynamically unstable, having a tendency to
of the fused sheets, which fragment upon sonication. evaporate out of the structure when exposed to air. Experiments
performed on non-clathrated, crystalline sodium ethoxide were
graphene precursor. When prepared from a 1:1 molar ratio of not found to result in the formation of graphene sheets, indicating
sodium and ethanol, the simple oxidation of sodium and reduction that the solvothermal reaction conditions are essential to produce
of ethanol to crystalline sodium ethoxide was not found to be the the metastable ethanol-rich graphene precursor.
dominant product. Under the conditions of the closed, heated reac- The production of large quantities of high-purity graphene
tion vessel, the alcoholic solution becomes increasingly saturated sheets will allow for the development of large-scale applications of
with the metal alkoxide as it forms, and as a result of the auto- this unique material. The study of the relative surface areas of flat
generated pressure (1  102 bar), the free alcohol is encapsulated and porous graphenes is currently under way. Chemically syn-
into the metal alkoxide in a clathrate-like structure. TGA conducted thesized graphene is now available in sufficient quantities to
on the product of the solvothermal step confirmed that the solid seriously consider its use in a wide range of applications, including
contained 29.0+2.5% by weight of water and ethanol. The solid solvo- composites reinforcement, electronic devices, batteries, sensors,
thermal product is not purely the ethoxide– ethanol clathrate, as hydrogen storage and catalysis.
some metal hydroxide may also form in the presence of water. In
addition, some residual sodium may also initially be dispersed Methods
All solvothermal reactions were performed in a Teflon-lined Parr Instrument
into the solid metal alkoxide, but rapidly hydrolyses to sodium Company 4749 reactor having a maximum volume of 23 ml. A typical synthesis
hydroxide when exposed to water vapour in the air, consistent consists of heating a 1:1 molar ratio of sodium (2 g) and ethanol (5 ml) in a sealed
with the discoloration of the solid from light grey to white. Owing reactor vessel at 220 8C for 72 h to yield the solid solvothermal product—the

32 NATURE NANOTECHNOLOGY | VOL 4 | JANUARY 2009 | www.nature.com/naturenanotechnology

© 2009 Macmillan Publishers Limited. All rights reserved.
NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2008.365 LETTERS
graphene precursor. This material is then rapidly pyrolysed, and the remaining 7. Li, D., Müller, M. B., Gilje, S., Kaner, R. B. & Wallace, G. G. Processable aqueous
product washed with deionized water (100 ml). The suspended solid is then vacuum dispersions of graphene nanosheets. Nature Nanotech. 3, 101 – 105 (2008).
filtered and dried in a vacuum oven at 100 8C for 24 h. The final yield of graphene is 8. Landau, L. D. Zur theorie der phasenumwandlungen II. Phys. Z. Sowjetunion 11,
approximately 0.1 g per 1 ml of ethanol—typically yielding 0.5 g per 26 – 35 (1937).
solvothermal reaction. 9. Meyer, J. C. et al. The structure of suspended graphene sheets. Nature 446,
The AFM measurements were performed with a Pico SPM II system (Molecular 60 – 63 (2007).
Imaging) in acoustic a.c. mode. Ethanol suspensions of graphene were applied 10. Dikin, D. A. et al. Preparation and characterization of graphene oxide paper.
directly on freshly cleaved mica, allowed to dry, and the measurements performed in Nature 448, 457– 460 (2007).
air at ambient temperature and pressure. The cantilevers used were silicon 11. Krishnan, A. et al. Graphitic cones and the nucleation of curved carbon surfaces.
cantilevers NSG01 (NT-MDT) with a spring constant of 3– 10 N m21, and Nature 388, 451– 454 (1997).
resonance frequency of 110– 190 kHz. Images were analysed using PicoScan 5 12. Land, T. A., Michely, T., Behm, R. J., Hemminger, J. C. & Comsa, G. STM
image-processing software supplied by Molecular Imaging. The figures shown in investigation of single layer graphite structures produced on Pt(111) by
this article were obtained in topography mode and have been levelled by first-order hydrocarbon decomposition. Surf. Sci. 264, 261– 270 (1992).
plane subtraction, as indicated. The height profiles were all obtained from 13. Nagashima, A., Nuka, K., Itoh, H., Ichinokawa, T. & Oshima, C. Electronic states
topography images that have been levelled by first-order plane subtraction. No other of monolayer graphite formed on TiC(111) surface. Surf. Sci. 291, 93– 98 (1993).
data manipulation was done on the images. 14. Malesevic, A. et al. Synthesis of few-layer graphene via microwave
X-ray photoemission spectroscopy (XPS) measurements were conducted with a plasma-enhanced chemical vapour deposition. Nanotechnology 19,
ESCALAB220i-XL (VG Scientific) system. Standard conditions required a vacuum 305604 (2008).
better than 2  1029 mbar, with a monochromatic aluminium Ka source and source 15. Dato, A., Radmilovic, V., Lee, Z., Phillips, J. & Frenklach, M. Substrate-free
power of 200 W (10 kV  20 mA). The spot size used was 1 mm in cross scans, gas-phase synthesis of graphene sheets. Nano Lett. 8, 2012 – 2016 (2008).
with a pass energy of 100 eV for wide scans. TEM images were obtained using a 16. Forbeaux, I., Themlin, J. M. & Debever, J. M. Heteroepitaxial graphite on
Philips CM200, and SEM images were obtained using a Hitachi S900. 6H-SiC(0001): Interface formation through conduction-band electronic
The SAED images were recorded at 300 keV using a JEOL 3000F at structure. Phys. Rev. B 58, 16396 –16406 (1998).
sample-to-plate distances of 800 and 1,500 mm for the sample and substrate, with 17. van Bommel, A. J., Crombeen, J. E. & van Tooren, A. LEED and Auger electron
exposure times of 0.1 and 0.3 s, respectively. The sample was deposited from observations of the SiC(0001) surface. Surf. Sci. 48, 463– 472 (1975).
suspension onto amorphous carbon substrates and dried by evaporation. 18. Kuang, Q. et al. Low temperature solvothermal synthesis of crumpled carbon
Photographic image plates were taken and subsequently digitized into nanosheets. Carbon 42, 1737 – 1741 (2004).
16-bit images and treated using the public domain software ImageJ 1.38 19. Shen, J.-M. & Feng, Y.-T. Formation of flower-like carbon nanosheet
(http://rsb.info.nih.gov/ij/). Profiles were taken along the preferred orientations for aggregations and their electrochemical application. J. Phys. Chem. C 112,
all six symmetry equivalent reflection sets for each graphene sheet (the image is a 13114– 13120 (2008).
composite of two sets of reflections off set by 198) and then numerically averaged. 20. Kawakami, M., Karato, T., Takenaka, T. & Yokoyama, S. Structure analysis of
The substrate profile was obtained by performing the same procedure across the coke, wood charcoal and bamboo charcoal by Raman spectroscopy and their
diffuse rings and normalized for exposure time. reaction rate with CO2. ISIJ Int. 45, 1027 – 1034 (2005).
TGA was performed on a HiRes Modulated TGA 2950 Thermo-Gravimetric 21. Coutinho, A. R., Rocha, J. D. & Luengo, C. A. Preparing and characterizing
Analyzer (TA Instruments). Data was interpreted using Thermal Advantage biocarbon electrodes. Fuel Process. Technol. 67, 93 – 102 (2000).
v1.1A software. 22. Deprez, N. & McLachlan, D. S. The analysis of the electrical conductivity
Raman spectroscopy was performed on an inVia Renishaw Raman microscope of graphite powders during compaction. J. Phys. D: Appl. Phys. 21,
using green (514 nm) laser excitation, and on a Renishaw R-2000 Raman 101– 107 (1988).
microscope for red (633 nm) laser excitation. Scans were taken on an extended range 23. Gregg, S. J. & Sing, K. S. W. Adsorption, Surface Area and Porosity (Academic
(100 – 3,500 cm21) and the exposure time was 60 s for the green laser and 10 s for Press, 1982).
the red laser. Samples were sonicated in ethanol and drops applied to a glass slide for 24. Boehm, H. P., Clauss, A., Fisher, G. O. & Hofmann, U. Das adsorptionsverhalten
observation. The sample was viewed using a green laser apparatus under a maximum sehr dünner kohlenstoff-folien. Z. Anorg. Allg. Chem. 316, 119– 127 (1962).
magnification of 50, and a red laser apparatus under a magnification of 100.

Received 24 July 2008; accepted 30 October 2008; Acknowledgements

This research was supported by the University of New South Wales (UNSW), the
published online 7 December 2008 Faculty Research Grant Program (FRGP), Goldstar grants and an Australian Postgraduate
Award (APA) to M.C. AFM measurements were carried out courtesy of P.T. at the
References University of Sydney.
1. Novoselov, K. S. et al. Electric field effect in atomically thin carbon films. Science
306, 666 – 669 (2004).
2. Li, X. et al. Highly conducting graphene sheets and Langmuir – Blodgett films.
Author contributions
J.A.S. and M.C. conceived and designed the experiments and developed the interpretation
Nature Nanotech. 3, 538– 542 (2008).
of results. M.C. synthesized all samples. P.T. assisted in the measurement and interpretation
3. Hernandez, Y. et al. High-yield production of graphene by liquid-phase
of the AFM data. All authors discussed the results and commented on the manuscript.
exfoliation of graphite. Nature Nanotech. 3, 563 –568 (2008).
4. Li, X., Wang, X., Zhang, L., Lee, S. & Dai, H. Chemically derived, ultrasmooth
graphene nanoribbon semiconductors. Science 319, 1229 –1232 (2008). Additional information
5. Stankovich, S. et al. Synthesis of graphene-based nanosheets via chemical Supplementary Information accompanies this paper at www.nature.com/
reduction of exfoliated graphite oxide. Carbon 45, 1558 – 1565 (2007). naturenanotechnology. Reprints and permission information is available online at http://npg.
6. Stankovich, S. et al. Graphene-based composite materials. Nature 442, nature.com/reprintsandpermissions/. Correspondence and requests for materials should be
282– 286 (2006). addressed to J.A.S.

NATURE NANOTECHNOLOGY | VOL 4 | JANUARY 2009 | www.nature.com/naturenanotechnology 33

© 2009 Macmillan Publishers Limited. All rights reserved.
 LETTERS
PUBLISHED ONLINE: 9 NOVEMBER 2008 | DOI: 10.1038/NNANO.2008.329

High-throughput solution processing of

large-scale graphene
Vincent C. Tung1†, Matthew J. Allen2†, Yang Yang1 * and Richard B. Kaner1,2 *

The electronic properties of graphene, such as high charge which is much larger than the theoretical value of 0.34 nm found in
carrier concentrations and mobilities, make it a promising can- graphite. This increased thickness may be attributed to unreduced
didate for next-generation nanoelectronic devices1–3. In particu- surface hydroxyl and epoxide groups. Such functionalities are detri-
lar, electrons and holes can undergo ballistic transport on the mental to the electrical properties of graphene. Third, aqueous disper-
sub-micrometre scale in graphene and do not suffer from the sions are not ideal for deposition as the high surface tension of water
scale limitations of current MOSFET technologies2,3. However, leads to aggregation during the evaporation process. Finally, even if
it is still difficult to produce single-layer samples of graphene1,3 GO is perfectly deposited, reduction methods tend to neglect the
and bulk processing has not yet been achieved, despite strenu- area in direct contact with the substrate. Attempts have been made
ous efforts to develop a scalable production method4,5. Here, to complete the reduction stage in solution, but sheets tend to aggregate
we report a versatile solution-based process for the large-scale due to the attractive forces between layers and an overall decrease
production of single-layer chemically converted graphene over in hydrophilicity12,13.
the entire area of a silicon/SiO2 wafer. By dispersing graphite Here, we present a solution-based approach for producing
oxide paper in pure hydrazine we were able to remove large-scale, single-layer chemically converted graphene. Dispersing
oxygen functionalities and restore the planar geometry of the GO paper in pure hydrazine creates stable dispersions of hydrazi-
single sheets. The chemically converted graphene sheets that nium graphene (HG) through the formation of counter-ions.
were produced have the largest area reported to date (up to These hydrazine colloids are readily deposited onto a variety of sub-
20 3 40 mm), making them far easier to process. Field-effect strates, producing uniform films of single- and/or few-layer chemi-
devices have been fabricated by conventional photolithography, cally converted graphene. Photographs of GO paper and HG are
displaying currents that are three orders of magnitude higher presented in Fig. 1a,b respectively, along with three-dimensional
than previously reported for chemically produced graphene6. molecular models of GO before and after reduction. By controlling
The size of these sheets enables a wide range of characteriz- the concentration and composition of these dispersions, films of a
ation techniques, including optical microscopy, scanning desired morphology and surface coverage can be produced. This
electron microscopy and atomic force microscopy, to be non-destructive method preserves the scalability of GO, ultimately
performed on the same specimen. forming reduced sheets much larger than those previously
Single-layer samples of graphene are currently made either by a reported6,11–14. This increase in size vastly reduces the effort necess-
laborious drawing method1 or through the high-temperature ary for electronic testing and renders characterization possible by a
reduction of silicon carbide7,8. In the drawing method, highly oriented number of microscopic techniques. In fact, the scale of graphene
pyrolytic graphite (HOPG) is repeatedly pealed using cellophane tape produced allows one of the most comprehensive studies yet con-
and deposited onto a silicon substrate. Yields for this method are extre- ducted on the characterization of chemically converted graphene.
mely low and involve searching the substrates for single sheets among The reduction of GO by hydrazine vapours is well known6,12,16–19,
a myriad of multilayered flakes. Even when likely specimens are ident- but here GO films are dispersed directly into hydrazine. Bubbles
ified with an optical microscope, atomic force microscopy (AFM) and rapidly form along the film surface due to the chemical reduction,
Raman techniques must be performed in order to provide conclusive which likely produces NO2 and N2 as byproducts. After several
evidence of their identity1,9,10. Alternatively, the reduction of silicon hours, no GO paper can be visually observed in the dispersion
carbide can be used to produce very small domains of single-layer and bubbling subsides, indicating good dispersion and complete
samples; however, temperatures .1,100 8C are needed and produ- reduction, respectively. As an explanation for the new dispersions
cing large domains remains a challenge7,8,11. High-temperature pro- created, we suggest the formation of HG composed of a negatively
cessing is also unlikely to be compatible with fabrication techniques charged, reduced graphene sheet surrounded by N2Hþ 4 counter-
used for most electronic applications. Hence, current methods of pro- ions. Hydrazinium compounds are known to readily disperse in
ducing graphene samples have very low throughput and do not appear hydrazine20. The resultant dispersions are stable for months with
to be scalable up to wafer size. little aggregation due to charge stabilization of counter-ion pairs.
An alternative method for creating single sheets starting from Stable aqueous dispersions have been reported by others16, but
graphite oxide (GO) has been suggested12,13. Graphite can be oxidized differ slightly in their composition. Under aqueous conditions
to produce GO and then exfoliated to create stable aqueous dispersions and low concentrations of hydrazine and high pH, carboxylic acid
of individual sheets. After deposition, GO may be reduced to graphene groups retain their negative charge, allowing adjacent sheets to
either chemically or by means of thermal annealing6,13. However, this repel one another. In anhydrous hydrazine, a greater number of
method has drawbacks. First, many of the resulting sheets are found to these carboxylic acids are likely reduced and the positive charge of
be wrinkled or folded when examined by AFM. Second, cross-sectional their N2Hþ 4 counter-ions becomes responsible for stabilization.
step heights of more than 1 nm are often observed for a single sheet13, The dispersions are similar in that both use hydrazine as the

1
Department of Materials Science and Engineering and California NanoSystems Institute, University of California, Los Angeles, California 90095, USA,
2
Department of Chemistry and Biochemistry and California NanoSystems Institute, University of California, Los Angeles, California 90095, USA; † These
authors contributed equally to this work; * e-mail: yangy@ucla.edu; kaner@chem.ucla.edu

NATURE NANOTECHNOLOGY | VOL 4 | JANUARY 2009 | www.nature.com/naturenanotechnology 25

© 2009 Macmillan Publishers Limited. All rights reserved.
LETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2008.329

N2H4

N2H4

Figure 1 | Chemically converted graphene suspensions. a,b, Photographs of 15 mg of graphite oxide (GO) paper in a glass vial (a) and the resultant
hydrazinium graphene (HG) dispersion after addition of hydrazine (b). Below each vial is a three-dimensional computer-generated molecular model of GO
(carbon in grey, oxygen in red and hydrogen in white) and chemically converted graphene, respectively, suggesting that removal of –OH and –COOH
functionalities upon reduction restores a planar structure.

reducing agent, but are quite independent due to the different

solvent systems—water and pure hydrazine. Note that HG may be
dried and resuspended in the organic solvents dimethyl sulphoxide
(DMSO) or N,N-dimethyl formamide (DMF) for deposition, as has
been reported for other hydrazinium complexes20. This approach
has the advantage of avoiding the requirement for spin-coating
from a solvent of such toxicity as hydrazine.
Purification of HG solutions can be carried out with various
combinations of ultrasonication, dilution and centrifugation.
Briefly, dispersions of large (20  20 mm) single sheets are created
by precipitating out higher density multilayers by means of centrifu-
gation. Such dispersions may be very useful for electronic appli-
cations due to the ease of making contact with the length scales of
these relatively large single sheets. Alternatively, large sheets and
aggregates may be fractured into uniform dispersions of smaller
single sheets ( for example, 1  1 mm) by repeated ultrasonication.
These small sheets readily make continuous films upon deposition
and may prove to be useful as ‘transparent’ conductors21. Note
that even these smaller sheets are large when compared to those pre-
pared by either the peeling method or by chemical reduction of SiC,
Figure 2 | SEM image of a large, single, chemically converted graphene
which typically have dimensions areas of no more than 0.2 mm2.
sheet. Such sheets are readily identified in SEM on silicon/SiO2 substrates
The initial characterization of HG is carried out by spin-coating onto
after spin-coating from hydrazine dispersions.
a silicon/SiO2 substrate and by examination through an optical micro-
scope. Optical microscopy was originally used to identify single–layer,
chemically converted graphene1. Here, however, we use optical images the SiO2 to a chemically converted graphene sheet of 0.6 nm for
primarily to observe the surface coverage of the depositions. For more a given cross-section (Fig. 3d). Figure 3d shows the bimodal fre-
detailed structural information, scanning electron microscopy (SEM) quency distribution of all points recorded during an entire AFM
images were collected, with a representative image of a large single scan, which also shows a peak separation of 0.6 nm. Exfoliated
sheet (20  40 mm) shown in Fig. 2. Under optimized operating con- samples prepared by the annealing method in ultrahigh vacuum
ditions (see Methods), single, double and multilayer chemically con- (UHV) also show typical step heights for single-layer exfoliated
verted graphene may be differentiated by opacity in SEM. sheets of 0.6 nm (ref. 29). The apparent 0.3 nm discrepancy in thick-
Although SEM images can be used to find single sheets of chemi- ness compared to theoretical values for graphene suggests the pre-
cally converted graphene, the images are representative of electronic sence of some residual oxygen functionalities on the sheets22.
structure, not topography. Hence, AFM is needed to establish the Nevertheless, the agreement between peeled samples and those
thickness and surface roughness of single sheets. For comparison, a measured here provides significant evidence that the planar structure
representative SEM and its corresponding AFM height images are of graphene has been largely re-established upon reduction.
shown in Fig. 3a,b, respectively. Height profiles show steps from Although reduced GO samples have been studied by AFM before,

26 NATURE NANOTECHNOLOGY | VOL 4 | JANUARY 2009 | www.nature.com/naturenanotechnology

© 2009 Macmillan Publishers Limited. All rights reserved.
NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2008.329 LETTERS

5 μm
1 μm

Height profile (nm)

2
1
0
10 nm −1 0.63 nm
−2
−3
0.0 0.5 1.0 1.5
Distance (μm)
2

Height profile (nm)

1

−1 0.6 nm

−2
0 1 2 3 4
Distance (μm)
2
Height profile (nm)

5 μm −1 0.59 nm

−2
0 nm 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Distance (μm)

0.1
Frequency

0.08
0.06
0.04
0.02
1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4
(nm)

Figure 3 | Images of an individual chemically converted graphene sheet. a–c, Markers are used to obtain SEM (a) and AFM (b,c) images of the same
specimen. d, Height profiles taken along the white, red and blue solid lines in c highlight step heights of 0.6 nm. The frequency distribution (bottom) of all
heights collected across the area is bimodal.

the size limitations of single sheets have not previously permitted and requires a sophisticated lithographic setup, making it difficult
such quantitative comparisons. to scale up. Moreover, this lithographic process can degrade
In an investigation of chemically converted graphene, the device performance by introducing contaminants if the photore-
characterization of electrical properties and fabrication of elec- sist is not entirely removed23,24. The size of the chemically con-
tronic devices are of great importance. Owing to the size limit- verted graphene sheets reported here, however, provides far
ations of most graphene samples, electron-beam lithography has fewer constraints for the effective fabrication of devices. The
been the only reliable method for producing the patterns necess- large sheets can instead be processed by means of conventional
ary for electrical testing. In a typical electron-beam process, single photolithography on SiO2 substrates to provide gold source–
sheet candidates are located by SEM, markers provided, and a drain top contacts. Figure 4 provides a schematic of our FET
polymethyl methacrylate photoresist selectively exposed. After design (Fig. 4a) and a photograph, optical image, and SEM of a
removing the photoresist, gold source and drain electrodes are working device (Fig. 4b). Electrode separation channel lengths
then evaporated onto the surface to provide a top contact con- of 7 mm can be used and no alignment is necessary in order to
figuration for the graphene device. This method is laborious reliably produce single-sheet devices. This is due to the nearly

NATURE NANOTECHNOLOGY | VOL 4 | JANUARY 2009 | www.nature.com/naturenanotechnology 27

© 2009 Macmillan Publishers Limited. All rights reserved.
LETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2008.329

Drain
Graphene

Source

Silicon dioxide

Silicon

1 μm

Figure 4 | Arrays of working, chemically converted graphene transistors. Field-effect transistors are produced by spin-coating a well-dispersed chemically
converted graphene solution onto SiO2 substrates, followed by registering gold source–drain electrodes on top of the single sheets. a, Schematic view of a
chemically converted graphene field-effect device. b, Photograph (left), optical image (middle) and SEM image (right) of a working device with a channel
length of 7 mm.

1.2
−5.0 × 10–3 1.0
10–3
ISD (mA)

0.8 10–4
0.6
−4.0 × 10–3 10–5
0.4

0.2 10–6
–60 –40 –20 0 20 40 60
−3.0 × 10–3
ISD (A)

VG (V) 10–7
ISD (A)

10–8
−2.0 × 10–3
10–9
VG = 0 V
VG = −10 V 10–10
−1.0 × 10–3
VG = −30 V Reduced graphene
10–11
VG = −50 V Graphite oxide
0.0 10–12
0.0 −0.2 −0.4 −0.6 −0.8 −1.0 0.0 −0.2 −0.4 −0.6 −0.8 −1.0
VSD (V) VSD (V)

Figure 5 | Electrical characteristics of a chemically converted graphene field-effect transistor. a, Current (ISD)/voltage (VSD) characteristics of chemically
converted graphene devices. As the gate voltage is varied from 0 to –50 V, the conductance of the device increases, indicative of a p-type semiconductor
under ambient conditions. The inset shows a ISD –VG characteristic curve at VSD ¼ 0.1 V. b, Current (ISD)/voltage (VSD) plots taken for GO films before and
after reduction indicate a 1  108-fold decrease in sheet resistance.

uniform and tuneable distribution of single sheets over a large considerably larger channel lengths (7 mm versus 500 nm as
area of the wafer (1.5  1.5 cm). previously reported). The results show current responses nearly
More than 50 devices were tested to confirm the electrical output comparable to those of graphene produced by the peeling method
performance, with all measurements carried out under ambient con- (up to 100 mA at VSD ¼ 10 mV and channel lengths of 0.5 mm)1.
ditions. p-Type behaviour is readily and reproducibly achieved at Figure 5b shows output curves for parent GO and our chemically
gate voltages ranging from 0 to –60 V. Figure 5a shows the output converted graphene, indicating a 108-fold decrease in resistance. We
and transfer characteristics of a typical field-effect device. Output attribute the increase in conductivity to more complete reduction of
VSD/ISD curves show more than 5 mA at source–drain voltages of GO by anhydrous hydrazine. Graphene samples produced using the
only –1 V. This represents a considerable advantage over previously drawing method should be understood to represent the ultimate
reported values for reduced GO (25 mA at VSD ¼ 25 V)6 at reduction limit of our devices. We are, as of yet, unable to confirm

28 NATURE NANOTECHNOLOGY | VOL 4 | JANUARY 2009 | www.nature.com/naturenanotechnology

© 2009 Macmillan Publishers Limited. All rights reserved.
NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2008.329 LETTERS
the ambipolar effect as obtained with peeled graphite under ambient 6. Gilje, S., Han, S., Wang, M. S., Wang, K. L. & Kaner, R. B. A chemical route to
conditions. Residual hydroxyl groups may inhibit n-type behaviour, graphene for device applications. Nano Lett. 7, 3394 – 3398 (2007).
7. Berger, C. et al. Ultrathin epitaxial graphite: 2D electron gas properties and
as it is not observed at positive gate biases. Hydroxyl groups can a route toward graphene-based nanoelectronics. J. Phys. Chem. B 108,
serve as deep electron traps and prohibit their function as free carriers, 19912– 19916 (2004).
as is well documented in silicon field-effect transistors6,11. Several 8. Berger, C. et al. Electronic confinement and coherence in patterned epitaxial
methods for the chemical modification of reduced samples to graphene. Science 312, 1191 – 1196 (2006).
achieve n-type behaviour have been devised and are currently under 9. Novoselov, K. S. et al. Two-dimensional atomic crystals. Proc. Natl Acad. Sci.
USA 102, 10451– 10453 (2005).
investigation25–27. These include functionalization, for example, alky- 10. Castro, A., Guinea, F. & Peres, N. M. Drawing conclusions from grapheme.
lation, of parent GO materials before suspension in hydrazine. Phys. World, 33 – 37 (November 2006).
The large size of the chemically converted graphene sheets 11. Ohta, T., Bostwick, A., Seyller, T., Horn, K. & Rotenberg, E. Controlling the
produced here enable Raman spectroscopy to be carried out on a electronic structure of bilayer graphene. Science 313, 951 – 954 (2006).
working FET. This provides a non-destructive method for charac- 12. Stankovich, S., Piner, R. D., Nguyen, S. T. & Ruoff, R. S. Synthesis and
exfoliation of isocyanate-treated graphene oxide nanoplatelets. Carbon 44,
terizing graphene28–30. D (1,350 cm21), G (1,600 cm21), 2D 3342 – 3347 (2006).
(2,700 cm21) and S3 (2,950 cm21) graphitic peaks are present in 13. Stankovich, S. et al. Synthesis of graphene-based nanosheets via chemical
the spectra of the reduced samples. An increase of the D/G ratio reduction of exfoliated graphite oxide. Carbon 45, 1558 – 1565 (2007).
upon reduction of GO is observed. Although a prominent D peak 14. Gomez-Navarro, C. et al. Electronic transport properties of individual
chemically reduced graphene oxide sheets. Nano Lett. 7, 3499 – 3503 (2007).
is typically an indication of disorder in the Raman of peeled gra-
15. Hummers, W. S. Jr & Offeman, R. E. Preparation of graphitic oxide. J. Am.
phene, our observations are consistent with previous reports of Chem. Soc. 80, 1339 (1958).
reduced GO12. The strong D peak in chemically converted graphene 16. Li, D., Mueller, M. B., Gilje, S., Kaner, R. B. & Wallace, G. G. Processable aqueous
should be further investigated, as presently it is somewhat paradox- dispersion of graphene nanosheets. Nature Nanotech. 3, 101 – 105 (2008).
ical (see Supplementary Information for further discussion). If truly 17. Stankovich, S. et al. Stable aqueous dispersions of graphitic nanoplatelets via the
reduction of exfoliated graphite oxide in the presence of poly(sodium
pristine, even chemically converted graphene should have a D peak 4-styrenesulfonate). J. Mater. Chem. 16, 155– 158 (2006).
only near the edges where sp3 carbons are necessarily present. 18. Liu, P. & Gong, K. Synthesis of polyaniline-intercalated graphite oxide by an
In summary, the chemically modified GO materials presented in situ oxidative polymerization reaction. Carbon 37, 706– 707 (1999).
here are the largest chemically converted graphene samples pro- 19. Bourlinos, A. B. et al. Graphite oxide: chemical reduction to graphite and surface
duced to date and can be readily processed in a reliable, scalable modification with primary aliphatic amines and amino acids. Langmuir 19,
6050 – 6055 (2003).
method. This technique is extremely versatile and can be used to 20. Mitzi, B. D., Copel, M. & Chey, S. J. Low-voltage transistor employing
create a myriad of coatings and geometries necessary for device a high-mobility spin-coated chalcogenide semiconductor. Adv Mater. 17,
applications and a full range of characterization techniques. We 1289 – 1293 (2005).
believe that the large scale of these single sheets represents a break- 21. Wang, X., Zhi, L. & Müllen, K. Transparent, conductive graphene electrodes for
through in fabrication and could pave the way for new and innova- dye-sensitized solar cells. Nano Lett. 8, 323– 327 (2008).
22. Schniepp, C. H. et al. Functionalized single graphene sheets derived from
tive experiments on single-layer, chemically converted graphene. splitting graphite oxide. J. Phys. Chem. B 110, 8535 – 8539 (2006).
23. Ishigami, M., Chen, J. H., Cullen, W. G., Fuhrer, M. S. & Williams, E. D. Atomic
Methods structure of graphene on SiO2. Nano Lett. 7, 1643 – 1648 (2007).
Dispersions of GO were produced using a modified Hummers’ method from 24. Moser, J., Barreiro, A. & Bachtold, A. Current-induced cleaning of graphene.
graphite powder15. Typical dispersions were 2% wt/vol GO in water and could be Appl. Phys. Lett. 91, 163513 (2007).
diluted to various concentrations for use. Uniformity of a given dispersion was 25. Eda, G., Fanchini, G. & Chhowalla, M. Large-area ultrathin films of reduced
ensured by heating to 60 8C with repeated ultrasonication. In order to form GO graphene oxide as a transparent and flexible electronic material. Nature
paper, aqueous samples were subjected to vacuum filtration through a 0.22-mm Nanotech. 3, 270 – 274 (2008).
alumina membrane. This filtration process lasted 24 h, after which the resultant 26. Li, X., Wang, X., Zhang, L., Lee, S. & Dai, H. Chemically derived, ultrasmooth
films were left to dry under ambient conditions. Each dry, matte black GO film was graphene nanoribbon semiconductors. Science 319, 1229 – 1232 (2008).
then carefully peeled from its membrane. 27. Li, X. et al. Highly conducting graphene sheets and Langmuir – Blodgett films.
For reduction and suspension as HG, 15 mg of the GO film was dispersed Nature Nanotech. 3, 538– 542 (2008).
directly into a 98% anhydrous hydrazine solution (10 ml) in a nitrogen-filled dry 28. Ferrari, A. C. et al. Raman spectrum of graphene and graphene layers. Phys. Rev.
box and allowed to stir for one week. Lett. 97, 187401 (2006).
To begin characterization, HG suspensions were spin-coated onto silicon/SiO2 29. Graf, D. et al. Spatially resolved Raman spectroscopy of single- and few-layer
substrates. Following deposition, the films were thermally annealed at 150 8C in graphene. Nano Lett. 7, 238 – 242 (2007).
order to remove all hydrazine and to return the HG samples to chemically 30. Calizo, I., Balandin, A., Bao, W., Miao, F. & Lau, C. N. Temperature dependence
converted graphene. of the Raman spectra of graphene and graphene multilayers. Nano Lett. 7,
SEM images were collected under a variety of accelerating voltages and probe 2645 – 2649 (2007).
currents in an attempt to improve contrast and resolution. Ultimately, a relatively
small accelerating voltage (2 – 5 kV) and probe current (5 – 8 mA) proved most
appropriate as they reduced the penetration depth of incident electrons and more Acknowledgements
directly probed surface species. This work has been partially supported by the National Science Foundation (NSF;
DMR-0507294), the NSF-IGERT programme (M.J.A.) and the Air Force Office of Scientific
Received 7 July 2008; accepted 16 October 2008; Research (FA95500710264; V.C.T.). The authors also thank: L. Gomez, A. Steig, G. Yang
and R. Chen for their expertise with instrumentation; UCLA CNSI Pico Lab for imaging;
published online 9 November 2008 S. Gilje for his pioneering efforts at UCLA with graphite oxide; and W. Hou and P. Li for
many fruitful discussions.
References
1. Castro Neto, A. H., Guinea, F., Peres, N. M. R., Novoselov, K. S. & Geim, A. K.
The electronic properties of graphene. Available at http://arxiv.org/abs/ Author contributions
0709.1163 (2007). V.C.T. and M.J.A. conceived and performed the experiments and measurements. Y.Y. and
2. Bunch, J. S. et al. Electromechanical resonators from graphene sheets. Science R.B.K. conceptualized and directed the research project. All authors discussed the results
315, 490 – 493 (2007). and contributed to the manuscript.
3. Staley, N. et al. Lithography-free fabrication of graphene devices. Appl. Phys.
Lett. 90, 143518 (2007). Additional information
4. Viculis, L. M., Mack, J. J. & Kaner, R. B. A chemical route to carbon nanoscrolls. Supplementary Information accompanies this paper at www.nature.com/
Science 299, 1361 (2003). naturenanotechnology. Reprints and permission information is available online at http://npg.
5. Shioyama, H. & Akita, T. A new route to carbon nanotubes. Carbon 41, nature.com/reprintsandpermissions/. Correspondence and requests for materials should be
179 – 181 (2003). addressed to Y.Y. and R.B.K.

NATURE NANOTECHNOLOGY | VOL 4 | JANUARY 2009 | www.nature.com/naturenanotechnology 29

© 2009 Macmillan Publishers Limited. All rights reserved.
 EDITORIAL

Vol.3 No.9 September 2008 www.nature.com/naturenanotechnology

Graphene 2.0
Once the preserve of physicists, graphene is now attracting the attention of growing numbers
of chemists, who are discovering new ways to produce this remarkable material.

In January 2008 Nature Nanotechnology a modern-day string-and-sealing-wax the observation of various spin-based
published a News & Views article entitled approach that involved the use of ‘scotch phenomena, including the Kondo effect:
“Calling all chemists” by Rod Ruoff1 tape’ to peel the graphene from a sample there are also open questions about
about chemical techniques for producing of crystalline graphite, and a more high- the stability of true two-dimensional
graphene — individual layers of carbon tech approach that involved the epitaxial materials6 (according to Landau7 and
atoms — from graphite. At that time growth of graphene on silicon carbide other famous theoretical physicists, two-
chemists had enjoyed most success with substrates3. A series of breakthroughs in dimensional structures are not stable),
the production of chemically modified fundamental physics quickly followed2,4. and the parallels between the relativistic
graphene, notably graphene oxide, (The theory of graphene has a much charge carriers found in graphene and
as opposed to pure graphene. “The longer history, dating back to 1947, at other areas of physics have still to be
‘chemistry part’ of the graphene story least, when Philip Wallace of the Chalk explored fully. Theoretical predictions
has only just begun,” wrote Ruoff, “and River Laboratory in Canada published for the minimum conductivity in
because of the extraordinary number of a paper entitled “The band theory graphene also differ from experimental
levers that chemistry affords, there will of graphite”5). measurements by a factor of π. However,
be a growing dialogue between chemists this has not stopped research into the
and researchers in other fields (such as Chemists have been busy nitty gritty details that need to be fully
physics, materials and processing) who, understood before attention turns to
otherwise, might only have worked with pulling levers ever since. the long-term goal of making electronic
graphene in its pristine form.” devices for real-world applications8.
Well, chemists have been busy pulling However, both of these approaches had Beyond electronics, researchers are
levers ever since, as can be seen in a their drawbacks. The epitaxial approach, looking to exploit the properties of
number of recent papers, including two which has since been demonstrated on graphene in applications as diverse
in this issue (as described by Jun Zhu in other substrates, is expensive, but it may as NEMS9, advanced capacitors10 and
the News & Views article on page 528). well be suited for electronic applications. drug delivery11.
Hongjie Dai and co-workers report The mechanical exfoliation or cleavage Back in the lab, meanwhile, chemists
on page 538 that a combination of approach, on the other hand, is slow and are looking for ways to go beyond making
reintercalation — which basically involves labour-intensive (an optical microscope is samples with good electrical properties
the insertion of other substances, such as needed to find single- or few-layer sheets and produce graphene that is pristine.
nitric acid, between the graphene layers — of graphene among the material peeled Given the rate of recent progress, this
and sonication produces graphene sheets off) and is unlikely to ever be able to might happen sooner rather than later.
that exhibit high electrical conductance at produce graphene in the volumes needed
room temperature. They also went on to for applications. However, this method has References
make large transparent Langmuir–Blodgett since been developed to the stage where 1. Rouff, R. Nature Nanotech. 3, 10–11 (2008).
films. Jonathan Coleman and co-workers, members of the group at Manchester 2. Geim, A. K & Novoselov, K. S. Nature Mater. 6, 183–191 (2007).
on the other hand, used organic solvents University that invented this approach 3. Berger, C. et al. J. Phys. Chem. B 108, 19912–19916 (2004).
4. van den Brink, J. Nature Nanotech. 2, 199–201 (2007).
to produce the graphene in solution have set up a start-up company, Graphene 5. Wallace, P. R. Phys. Rev. 71, 622–634 (1947).
(page 563), from which they made semi- Industries (see page 523). 6. Meyer, J. et al. Nature 446, 60–63 (2007).
transparent films and polystyrene–graphene Fast forward to 2008 and research 7. Landau, L. D. Phys. Z. Sowjetunion 11, 26–35 (1937)
8. Freitag, M. Nature Nanotech. 3, 455–457 (2008).
composites with good electrical properties. on graphene is blossoming in a number
9. Bunch, J. S. et al. Science 315, 490–493 (2007).
As is well known, the experimental of different directions. Research into 10. Vivekchand, S. R. C. et al. J. Chem. Sci. 120, 9–13 (2008).
rise of graphene2 started in 2004 with the fundamentals is continuing, such as 11. Liu, Z., Robinson, J. T., Sun, X. & Dai, H. arxiv.org/abs/0807.4959

nature nanotechnology | VOL 3 | SEPTEMBER 2008 | www.nature.com/naturenanotechnology 517

© 2008 Macmillan Publishers Limited. All rights reserved.
 ARTICLES

High-yield production of graphene by

liquid-phase exfoliation of graphite
YENNY HERNANDEZ1†, VALERIA NICOLOSI1†, MUSTAFA LOTYA1, FIONA M. BLIGHE1, ZHENYU SUN1,2,
SUKANTA DE1,2, I. T. McGOVERN1, BRENDAN HOLLAND1, MICHELE BYRNE3, YURII K. GUN’KO2,3,
JOHN J. BOLAND2,3, PETER NIRAJ2,3, GEORG DUESBERG2,3, SATHEESH KRISHNAMURTHY2,3,
ROBBIE GOODHUE4, JOHN HUTCHISON5, VITTORIO SCARDACI6, ANDREA C. FERRARI6
AND JONATHAN N. COLEMAN1,2*
1
School of Physics, Trinity College Dublin, Dublin 2, Ireland
2
Centre for Research on Adaptive Nanostructures and Nanodevices (CRANN), Trinity College Dublin, Dublin 2, Ireland
3
School of Chemistry, Trinity College Dublin, Dublin 2, Ireland
4
Department of Geology, School of Natural Sciences, Trinity College Dublin, Dublin 2, Ireland
5
Department of Materials, University of Oxford, Parks Road, Oxford OX1 3PH, UK
6
Engineering Department, University of Cambridge, 9 JJ Thomson Avenue, Cambridge CB3 0FA, UK
†
These authors contributed equally to this work.
*e-mail: colemaj@tcd.ie

Published online: 10 August 2008; doi:10.1038/nnano.2008.215

Fully exploiting the properties of graphene will require a method for the mass production of this remarkable material. Two main
routes are possible: large-scale growth or large-scale exfoliation. Here, we demonstrate graphene dispersions with concentrations
up to 0.01 mg ml21, produced by dispersion and exfoliation of graphite in organic solvents such as N-methyl-pyrrolidone. This
is possible because the energy required to exfoliate graphene is balanced by the solvent –graphene interaction for solvents whose
surface energies match that of graphene. We confirm the presence of individual graphene sheets by Raman spectroscopy,
transmission electron microscopy and electron diffraction. Our method results in a monolayer yield of 1 wt%, which could
potentially be improved to 7 –12 wt% with further processing. The absence of defects or oxides is confirmed by X-ray
photoelectron, infrared and Raman spectroscopies. We are able to produce semi-transparent conducting films and conducting
composites. Solution processing of graphene opens up a range of potential large-area applications, from device and sensor
fabrication to liquid-phase chemistry.

The novel electronic properties of graphene have been well in order to make useful devices, either by mechanical transfer or
documented1; the charge carriers behave as massless Dirac through solution processing.
fermions2, and novel effects such as an ambipolar field effect3, a Recently, a large number of papers have described the dispersion
room-temperature quantum Hall effect4 and the breakdown of and exfoliation of graphene oxide (GO)18–21. This material consists of
the Born – Oppenheimer approximation5 have all been observed. graphene-like sheets, chemically functionalized with compounds
A graphene monolayer has also been demonstrated as a such as hydroxyls and epoxides, which stabilize the sheets in
transparent electrode in a liquid crystal device6. However, as was water22. However, this functionalization disrupts the electronic
the case in the early days of nanotube and nanowire research, structure of graphene. In fact GO is an insulator23 rather than a
graphene suffers from a problem that is common to many novel semi-metal and is conceptually different from graphene. Although
materials — the lack of a method for producing it at high yields. the functional groups can be removed by reduction, so far this
The standard procedure used to make graphene is leaves a significant number of defects, which continue to disrupt
micromechanical cleavage7. This gives the best samples to date, the electronic properties, remain18,22. Thus, a non-covalent,
with carrier mobilities up to 200,000 cm2 V21 s21 (refs 8– 10). solution-phase method to produce significant quantities of defect-
However, the single layers so obtained form a negligible fraction free, unoxidized graphene is urgently required. In this paper we
amongst large quantities of thin graphite flakes. Furthermore, it propose one such method.
is difficult to envisage how to scale up this process to mass Here we show that high-quality monolayer graphene can be
production. Alternatively, growth of graphene is also commonly produced at significant yields by non-chemical, solution-phase
achieved by annealing SiC substrates; however, these samples are exfoliation of graphite in certain organic solvents. This work
in fact composed of a multitude of domains, most of them builds upon over 50 years of study into chemical exfoliation of
submicrometre in scale, and they are not spatially uniform in graphite24. Previously, intercalated graphite could be partially
number or size over larger length scales11–13. A number of works exfoliated by reactions involving the intercalant25, through thermal
have also reported graphene growth on metal substrates14–17, but shock26 or by acid treatment of expandable graphite27. However, to
this would require transfer of the sample to insulating substrates date, such methods have given thin graphite sheets or graphene

nature nanotechnology | VOL 3 | SEPTEMBER 2008 | www.nature.com/naturenanotechnology 563

© 2008 Macmillan Publishers Limited. All rights reserved.
ARTICLES
fragments27 rather than large-scale graphene monolayers. The
response to this problem has so far been the exfoliation of
chemically modified forms of graphene such as GO or A B C D E
functionalized graphene20,22,28. However, such materials are not
graphene, as they are insulators containing numerous structural
defects22,28 that cannot, so far, be fully removed by chemical
treatment22. Our method results in high-quality, unoxidized
monolayer graphene at yields of 1 wt%. We show that the
process could potentially be improved to give yields of up to
4,000
12 wt% of the starting graphite mass with sediment recycling. As a
solution-phase method it is versatile, it may be scaled up, and it 3,000

α (L g–1 m–1)
can be used to deposit graphene in a variety of environments and
2,000
substrates not available using cleavage or growth methods.
Furthermore, it can be used to produce graphene-based 1,000 NMP DMA
composites or films, a key requirement for many applications, GBL DMEU
0
such as thin-film transistors, conductive transparent electrodes for 400 600 800 1,000 1,200 1,400
indium tin oxide replacement or for photovoltaics. Wavelength (nm)
20
DISPERSION OF GRAPHITE λ = 660 nm
15 α ~ 2,460 L g–1 m–1

A /I (m–1)
10 NMP
Recently, carbon nanotubes have been successfully exfoliated in a GBL
small number of solvents such N-methylpyrrolidone (NMP)29–33. 5 DMA
DMEU
Such exfoliation occurs because the strong interaction between 0
solvent and nanotube sidewall means that the energetic penalty for 0 2 4 6 8 10
exfoliation and subsequent solvation becomes small34. We suggest Concentration (µg ml–1)
that similar effects may occur between these solvents and sol
Solvent surface energy, E sur (mJ m–2)
graphene. To test this we prepared a dispersion of sieved graphite
40 50 60 70 80 90 100
powder (Aldrich product 332461, batch number 06106DE) in
centrifugation (µg ml–1 )

NMP (spectrophotometric grade, .99.0%) by bath sonication 10 10

Concentration after

% remaining after
(see Supplementary Information, Section S1.2). After sonication

centrifugation
we obtained a grey liquid consisting of a homogeneous phase and sol
γ = E sur – TSsur
sol
5 5
large numbers of macroscopic aggregates. As with nanotube
dispersions30,32, these aggregates could be removed by mild
centrifugation, giving a homogeneous dark dispersion. Such 0 0
dispersions, prepared at different graphite concentrations are 10 20 30 40 50 60 70 80
shown in Fig. 1a. Although moderate levels of sedimentation and Solvent surface tension, γ (mJ m–2)
aggregation occur within three weeks of centrifugation, the
dispersions remain of high quality at least five months after
preparation (see Supplementary Information, Section S2.4). Figure 1 Optical characterization of graphite dispersions. a, Dispersions of
In order to find the concentration after centrifugation, we graphite flakes in NMP, at a range of concentrations ranging from 6 mg ml21 (A)
passed the graphite dispersion through polyvinylidene fluoride to 4 mg ml21 (E) after centrifugation. b, Absorption spectra for graphite flakes
(PVDF) filters. Careful measurements of the filtered mass, dispersed in NMP, GBL, DMA and DMEU at concentrations from 2 to 8 mg ml21.
accounting for residual solvent, gave the concentration of c, Optical absorbance (l ex ¼ 660 nm) divided by cell length (A/l ) as a function
dispersed phase after centrifugation. This procedure was repeated of concentration for graphene in the four solvents NMP, GBL, DMA and DMEU
for three other solvents known to successfully disperse showing Lambert – Beer behaviour with an average absorption coefficient of
nanotubes34: N,N-Dimethylacetamide (DMA), g-butyrolactone ka 660l ¼ 2,460 L g21 m21. The x-axis error bars come from the uncertainty in
(GBL) and 1,3-dimethyl-2-imidazolidinone (DMEU). These measuring the mass of graphene/graphite in solution. d, Graphite concentration
dispersions were then characterized by UV–vis –IR measured after centrifugation for a range of solvents plotted versus solvent
absorption spectroscopy, with the absorption coefficient plotted surface tension. The data were converted from absorbance (660 nm) using
versus wavelength (Fig. 1b). The spectra are featureless in the A/l ¼ ka 660lC with ka 660l ¼ 2,460 L g21 m21. The original concentration, before
visible–IR region as expected35. Each of these four dispersions centrifugation, was 0.1 mg ml21. The y-axis error bars represent the standard
was diluted a number of times and the absorption spectra deviation calculated from five measurements. Shown on the right axis is the
recorded. The absorbance (660 nm) divided by cell length is percentage of material remaining after centrifugation. On the top axis, the surface
plotted versus concentration (Fig. 1c), showing Lambert –Beer tension has been transformed into surface energy using a universal value for surface
behaviour for all solvents, ka660l ¼ 2,460 L g21 m21. entropy of S sol
sur  0.1 mJ K
21 22
m . The horizontal arrow shows the approximate
Thus, it is clear that graphite can be dispersed in some solvents. range of the reported literature values for the surface energy of graphite39–42.
As we will show, the graphite is almost completely exfoliated to
multilayer structures with ,5 layers in NMP, GBL and DMEU, if
not other solvents. In addition, significant quantities of approximately calculate in this case to be (see Supplementary
individual monolayers are also present. The question is what Information, Section S6.0):
solvent properties lead to this exfoliation of graphite?
Such exfoliation can only occur if the net energetic cost is
DHmix 2
very small. In chemistry, this energy balance is expressed as the  ðd  dsol Þ2 f ð1Þ
enthalpy of mixing (per unit volume), which we can Vmix Tflake G

564 nature nanotechnology | VOL 3 | SEPTEMBER 2008 | www.nature.com/naturenanotechnology

© 2008 Macmillan Publishers Limited. All rights reserved.
 ARTICLES

30

Counts
20

10
N = 100
0
0 2 4 6 8 10 12 14
Number of layers per sheet

Figure 2 Electron microscopy of graphite and graphene. a, SEM image of sieved, pristine graphite (scale bar: 500 mm). b, SEM image of sediment after
centrifugation (scale bar: 25 mm). c – e, Bright-field TEM images of monolayer graphene flakes deposited from GBL (c), DMEU (d) and NMP (e), respectively (scale
bars: 500 nm). f,g, Bright-field TEM images of a folded graphene sheet and multilayer graphene, both deposited from NMP (scale bars: 500 nm). h, Histogram of the
number of visual observations of flakes as a function of the number of monolayers per flake for NMP dispersions.

p i
where di ¼ (Esur ) is the square root of the surface energy of phase i, crystallite size in the starting powder was .150 mm, the
Tflake is the thickness of a graphene flake and f is the graphene preparation procedure must result in tearing of the crystallites.
volume fraction. Reminiscent of the Hildebrand–Scratchard This process may be similar to sonication-induced fragmentation
equation36, this shows the enthalpy of mixing is dependent on the of carbon nanotubes43.
balance of graphene and solvent surface energies. For graphite, the
surface energy is defined as the energy per unit area required to EVIDENCE OF EXFOLIATION TO GRAPHENE
overcome the van der Waals forces when peeling two sheets apart.
From equation (1), we expect a minimal energy cost of It is possible to investigate the state of the material remaining
exfoliation for solvents whose surface energy matches that dispersed using transmission electron microscopy (TEM) by
of graphene. To test this, we dispersed graphite in a wide range of dropping a small quantity of each dispersion onto holey carbon
solvents. By measuring the optical absorbance after mild grids. Crucially, this technique is simpler than that previously
centrifugation and using the absorption coefficient (660 nm) to used to prepare graphene for TEM44, which involved under-
transform absorbance into concentration, we can quantify the etching of graphene placed on a silicon substrate. Immediately
amount of graphite flakes dispersed as a function of solvent apparent in the present technique is the advantage of having
surface energy (calculated from surface tension37,38, see Fig. 1 graphite dispersions. Figure 2c–g shows bright-field TEM images
caption) as shown in Fig. 1d. As predicted, the dispersed of the objects typically observed, which generally fall into three
concentration shows a strong peak for solvents with a surface classes. The first class, as shown in Fig. 2c– e, comprises
energy very close to the literature values of the nanotube/ monolayer graphene. Second, in a number of cases we observe
graphite surface energy39–42 (that is,  70 –80 mJ m22). Coupled folded graphene layers (Fig. 2f ). Third, we find bilayer and
with equation (1), this strongly suggests that not only is the multilayer graphene (Fig. 2g). In all cases, these objects have
enthalpy of mixing for graphite dispersed in good solvents very lateral sizes typically of a few micrometres. In some cases the
close to zero, but the solvent –graphite interaction is van der sheet edges tend to scroll and fold slightly (see Supplementary
Waals rather than covalent. In addition, it predicts that good Information, Fig. S3b). However, we rarely observe large objects
solvents are characterized by surface tensions in the region of with thickness of more than a few layers. Thus we believe that, in
40 –50 mJ m22. Also, we can tell from these data that for the best these samples, graphite has been extensively exfoliated to give
solvent (benzyl benzoate), 8.3% by mass of the original material monolayer and few-layer graphene. By analysing a large number
remained after centrifugation. (For NMP, 7.6% remained.) of TEM images, paying close attention to the uniformity of the
It is crucial to ascertain the exfoliation state of the material that flake edges, we can generate flake thickness statistics as shown in
remains dispersed after centrifugation. First we examined the state Fig. 2h. From these data we can estimate the number fraction of
of the initial graphite powder. Scanning electron microscopy monolayer graphene (number of monolayers/total number of
(SEM) studies (Fig. 2a) show the starting powder to consist of flakes observed) in NMP dispersions as 28%. This corresponds to
flakes of lateral size ,500 mm and thickness ,100 mm. In a solution-phase monolayer mass fraction (mass of all
comparison, the sediment separated after centrifugation contains monolayers/mass of all flakes observed) of 12 wt%, leading to
flakes, which are much smaller, with lateral size measured in tens an overall yield (mass of monolayers/starting graphite mass) of
of micrometres with thicknesses of a few micrometres (Fig. 2b). 1 wt% (see Supplementary Information, Table S2 and Section
Clearly, sonication results in fragmentation of the initial flakes, S2.3). In fact, we also find that the sediment can be recycled to
with the largest removed by centrifugation. We note that, as the produce dispersions with number and mass fractions of

nature nanotechnology | VOL 3 | SEPTEMBER 2008 | www.nature.com/naturenanotechnology 565

© 2008 Macmillan Publishers Limited. All rights reserved.
ARTICLES

–1–120
0–110 –1010

1–210 –2110

Intensit y (a.u.) Intensity (a.u.)

200 200 i 14
12 N = 45
100
Intensit y (a.u.)

150 10 Multilayer

Counts
0 8
100
6
50 h 600 4 Graphene
300 2
0
0
0.0 0.5 1.0 1.5 0
0.0 0.5 1.0 1.5 2.0
0.0 0.5 1.0 1.5 I{1100}/I{2110}
Distance (Å–1)
Distance (Å–1)

Figure 3 Evidence of monolayer graphene from TEM. a,b, High-resolution TEM images of solution-cast monolayer (a) and bilayer (b) graphene (scale bar 500 nm).
c, Electron diffraction pattern of the sheet in a, with the peaks labelled by Miller – Bravais indices. d,e, Electron diffraction patterns taken from the positions of the
black (d) and white spots (e), respectively, of the sheet shown in b, using the same labels as in c. The graphene is clearly one layer thick in d and two layers thick
in e. f – h, Diffracted intensity taken along the 1 – 210 to – 2110 axis for the patterns shown in c – e, respectively. i, Histogram of the ratios of the intensity of the
f1100g and f2110g diffraction peaks for all the diffraction patterns collected. A ratio .1 is a signature of graphene.

monolayer graphene that we have measured to be 18% and plot a line section through the (1– 210)–(0 –110) –( –1010) –
7 wt%, respectively. This suggests the possibility of full sediment (– 2110) axis for the patterns in Fig. 3c–e in Fig. 3f –h. In
recycling and the eventual increase of the yield towards Fig. 3f,g we see that the inner peaks, (0– 110) and (– 1010), are
7 –12 wt% (relative to the starting graphite mass). more intense than the outer ones, (1– 210) and ( – 2110),
confirming that that both the flake in Fig. 3a and the region
IDENTIFICATION OF MONOLAYERS BY ELECTRON DIFFRACTION marked by the black dot in Fig. 3b are monolayers. Conversely,
Fig. 3h shows inner peaks that are less intense than the outer
A more definitive identification of graphene can be made by ones, confirming that the area around the white dot in Fig. 3b
analysis of electron diffraction patterns45. As an example of this, consists of more than one layer. Further confirmation of the
Fig. 3a,b shows what appear to be a graphene monolayer and a presence of monolayer graphene can be found by measuring the
graphene bilayer, respectively. Figure 3b is particularly interesting diffraction peak intensity as a function of tilt angle (see
as the right side of the flake consists of at least two layers, Supplementary Information, Section S2.8).
whereas on the left side, a single monolayer protrudes. Figure 3c We can use the fact that the ratio of the intensity of the {1100}
shows the normal-incidence electron diffraction pattern of the to the {2110} peaks gives an unambiguous local identification of
flake in Fig. 3a. This pattern shows the typical sixfold symmetry monolayer versus multilayer to provide information on the yield
expected for graphite/graphene44,45, allowing us to label the of monolayer graphene. We measured the diffraction pattern of
peaks with the Miller–Bravais (hkil) indices. Figure 3d,e 45 flakes before measuring the intensity ratio If1100g/If2110g.
shows normal-incidence selected-area diffraction patterns for the These ratios are plotted as a histogram in Fig. 3i. We get a
flake in Fig. 3b, taken with beam positions close to the black bimodal distribution, with peaks centred at If1100g/If2110g ¼ 0.35
and white dots, respectively. This means we expect one and If1100g/If2110g ¼ 1.5, representing multilayer and monolayer
pattern (Fig. 3d) to reflect monolayer graphene and the other graphene, respectively. These results agree well with reported
(Fig. 3e) to reflect multilayer graphene. In both cases we see a experimental intensity ratios of If1100g/If2110g  0.4 for bilayer
hexagonal pattern similar to that in Fig. 3c. The main difference graphene and If1100g/If2110g  1.4 for monolayer graphene45.
between Fig. 3d and Fig. 3e is that for the multilayers (Fig. 3e), Alhough these data suggest a yield of 51% monolayer graphene,
the f2110g spots appear to be more intense relative to the this is certainly an overestimate, as selected-area electron
f1100g spots. This is an important observation, as for diffraction can give monolayer-like patterns for multilayers, such
multilayers with Bernal (AB) stacking, computational studies as that in Fig. 3b, when the beam is incident on a protruding
have shown that the intensity ratio is If1100g/If2110g , 1, whereas monolayer. Better statistics can be found by counting the number
for monolayers it is If1100g/If2110g . 1 (ref. 46). Virtually all the of layers per flake, as shown in Fig. 2h. However, we can use
objects identified in all the images as multilayers displayed a ratio electron diffraction to check the accuracy of our image analysis,
of If1100g/If2110g , 1, demonstrating that AB stacking is showing that we can reproducibly use it to identify monolayer
predominant in these samples46. graphene, thus confirming the results presented in Fig. 2h. The
This identification of AB stacking in these thin multilayers presence of monolayers was also confirmed by measuring TEM
allows us to differentiate monolayer from multilayer graphene by identified layers by Raman spectroscopy (see Supplementary
inspection of the intensity ratio If1100g/If2110g. To do this, we Information, Section S2.9).

566 nature nanotechnology | VOL 3 | SEPTEMBER 2008 | www.nature.com/naturenanotechnology

© 2008 Macmillan Publishers Limited. All rights reserved.
 ARTICLES
bilayer, demonstrate that our process does not introduce
G
significant structural defects47, such as epoxides covalently
2D
1: Bulk graphite
bonded to the basal plane22. In addition, we recorded Raman
D
spectra for individual flakes deposited on marked TEM grids,
allowing us to identify monolayers, bilayers and multilayers from
Intensit y (a.u.)

2: NMP cast film, large flake

both the TEM image and the shape of the 2D band, confirming
the quality of our exfoliation (see Supplementary Information,
3: NMP cast film, small flake
Section S2.9). Furthermore, X-ray photoelectron spectroscopy, as
shown in Fig. 4b (see Supplementary Information, Section S3.2)
4: NMP cast individual bilayer and infrared spectroscopy (see Supplementary Information,
Section S3.3) show the absence of oxidization typically associated
with GO (refs 18,19). These experiments again confirm that we
1,250 1,500 1,750 2,500 2,750 3,000 can produce high-quality, unoxidized graphite and graphene
Raman shift (cm–1) flakes in solution.

FURTHER CHARACTERIZATION OF LIQUID-PHASE EXFOLIATION

C-N C
C=O Ring CH3 We can briefly illustrate the potential of this method of graphite
exfoliation by using it to make thin graphene films. Raman and
Intensity (a.u.)

N O
288 287 286 285 284 SEM analyses show that these films consist predominately of thin
graphite flakes with fewer than five layers (see Supplementary
Information, Section S1.4). X-ray photoelectron spectroscopy
measurements show that these films have 11 wt% residual NMP
after drying at room temperature at 1  1023 mbar. This value
remained unchanged after a subsequent vacuum anneal at 400 8C
(see Supplementary Information, Section S3.2). Combustion
290 288 286 284 282 analysis gave an NMP content of 10 wt% after room-temperature
Binding energy (eV) drying (1  1023 mbar), which can be reduced to ,7 wt% after
annealing (see Supplementary Information, Section S3.4). These
films have conductivities of 6,500 S m21, similar to reduced
Figure 4 Evidence for defect-free graphene. a, Raman spectra for
graphene oxide films19, and optical transparencies of 42% (see
bulk graphite (1), a vacuum filtered film with the laser spot focused on a large
Supplementary Information, Section S4.0).
(5 mm) flake (2), a vacuum filtered film with the laser spot focused on a small
We also demonstrate polystyrene–graphene composites at high
(1 mm) flake (3), a large (10 mm) bilayer (4). Note that for spectra 2 and 4,
volume fraction. We measured the conductivity of such composites
the D line is absent, indicating that virtually no defects are present. For the
to be 100 S m21 (see Supplementary Information, Section S5.0)
small flake (spectrum 3), a weak D line is apparent, consistent with edge
for 60–80 vol% films, comparable to the most conductive
effects. b, A carbon 1s core-level XPS spectrum for a thin film (30 nm),
polymer–nanotube composites48 and significantly higher than
vacuum-deposited from a graphene dispersion and dried in a vacuum oven at
those quoted for graphene-oxide-based composites20. Finally, we
room temperature. The Shirley background has been subtracted for clarity. Main
deposited graphene monolayers and multilayers on SiO2 surfaces by
fit line represents graphitic carbon (C – C). The remainder, 286 eV, can be very
means of spray coating, demonstrating that this processing method
well fitted considering only residual NMP without the need for any oxide lines.
can potentially be used to prepare samples for microelectronic
The smaller fit lines represent residual NMP; Cring , carbon in the NMP ring
applications (see Supplementary Information, Section S2.7).
bonded to two hydrogen atoms; C – N, carbon in the NMP molecule bonded to a
nitrogen atom; C ¼ O, carbon in the NMP ring double bonded to an oxygen
atom. Left inset: enlarged view of the NMP fit lines (combined and individual). CONCLUSION
Right inset: structure of NMP.
We have demonstrated a scalable method to produce high-quality,
unoxidized graphite and graphene flakes from powdered graphite.
By using certain solvents, graphene can be dispersed at
EVIDENCE FOR DEFECT-FREE GRAPHENE concentrations of up to 0.01 mg ml21. These dispersions can
then be used to deposit flakes by spray coating, vacuum filtration
Although Fig. 1d suggests a van der Waals type solvent– graphene
or drop casting. By adding polymers they can be turned into
interaction, it is crucial to definitively rule out any inadvertent
polymer–composite dispersions. We believe that this work opens
basal-plane functionalization, which could alter the electronic
up a whole new vista of potential applications from sensor or
structure. Figure 4a shows Raman spectra of three different flakes
devices to transparent electrodes and conductive composites.
with the spectrum of bulk graphite for comparison (see
Supplementary Information, Section S1.4). Spectra 2 and 3 were Received 2 May 2008; accepted 2 July 2008; published 10 August 2008.
measured on thin films prepared by vacuum filtration onto References
alumina filters by focusing the spot on a large diameter (5 mm) 1. Geim, A. K. & Novoselov, K. S. The rise of graphene. Nature Mater. 6, 183 –191 (2007).
and a small diameter (1 mm) flake, respectively. Spectrum 4 2. Novoselov, K. S. et al. Two-dimensional gas of massless Dirac fermions in graphene. Nature 438,
197– 200 (2005).
was measured on a significantly large bilayer (.10 mm) (see 3. Novoselov, K. S. et al. Electric field effect in atomically thin carbon films. Science 306, 666–669 (2004).
Supplementary Information, Sections S1.1 and S2.9). The G line 4. Zhang, Y. B., Tan, Y. W., Stormer, H. L. & Kim, P. Experimental observation of the quantum Hall
effect and Berry’s phase in graphene. Nature 438, 201– 204 (2005).
(1,580 cm21) and 2D line (2,700 cm21) are clearly visible in 5. Pisana, S. et al. Breakdown of the adiabatic Born–Oppenheimer approximation in graphene. Nature
all cases. However, the D peak (1,350 cm21) is only visible Mater. 6, 198 –201 (2007).
6. Blake, P. et al. Graphene-based liquid crystal device. Nano Lett. 8, 1704–1708 (2008).
in the spectrum of the very small flake, as expected due to edge 7. Novoselov, K. S. et al. Two-dimensional atomic crystals. Proc. Natl Acad. Sci. USA 102,
effects47. These data, in particular the spectrum for the individual 10451–10453 (2005).

nature nanotechnology | VOL 3 | SEPTEMBER 2008 | www.nature.com/naturenanotechnology 567

© 2008 Macmillan Publishers Limited. All rights reserved.
ARTICLES
8. Bolotin, K. I. et al. Ultrahigh electron mobility in suspended graphene. Solid State Commun. 146, 33. Bergin, S. D. et al. Exfoliation in ecstasy: liquid crystal formation and concentration-dependent
351 –355 (2008). debundling observed for single-wall nanotubes dispersed in the liquid drug g-butyrolactone.
9. Morozov, S. V. et al. Giant intrinsic carrier mobilities in graphene and its bilayer. Phys. Rev. Lett. 1, Nanotechnology 18, 455705 (2007).
016602 (2008). 34. Bergin, S. D. et al. Towards solutions of SWNT in common solvents. Adv. Mater. 20,
10. Du, X., Skachko, I., Barker, A. & Andrei, E. Y. Approaching ballistic transport in suspended graphene. 1876 –1881 (2007).
Nature Nanotech. 3, 491 –495 (2008). 35. Abergel, D. S. L. & Fal’ko, V. I. Optical and magneto-optical far-infrared properties of bi-layer
11. Berger, C. et al. Electronic confinement and coherence in patterned epitaxial graphene. Science 312, graphene. Phys. Rev. B 75, 155430 (2007).
1191– 1196 (2006). 36. Hildebrand, J. H., Prausnitz, J. M. & Scott, R. L. Regular and related solutions 1st edn (Van Nostrand
12. Ohta, T. et al. Morphology of graphene thin film growth on SiC(0001). New J. Phys. Reinhold Company, New York, 1970).
023034 (2008). 37. Lyklema, J. The surface tension of pure liquids—thermodynamic components and corresponding
13. Zhou, S. Y. et al. Origin of the energy bandgap in epitaxial graphene—Reply. Nature Mater. 7, states. Coll. Surf. A 156, 413 –421 (1999).
259– 260 (2008). 38. Tsierkezos, N. G. & Filippou, A. C. Thermodynamic investigation of N,N-dimethylformamide/toluene
14. Marchini, S., Gunther, S. & Wintterlin, J. Scanning tunnelling microscopy of graphene on Ru(0001). binary mixtures in the temperature range from 278.15 to 293.15 K. J. Chem. Therm. 38, 952–961 (2006).
Phys. Rev. B 76, 075429 (2007). 39. Benedict, L. X. et al. Microscopic determination of the interlayer binding energy in graphite. Chem.
15. de Parga, A. L. V. et al. Periodically rippled graphene: Growth and spatially resolved electronic Phys. Lett. 286, 490 –496 (1998).
structure. Phys. Rev. Lett. 1, 056807 (2008). 40. Girifalco, L. A. & Good, R. J. A theory for the estimation of surface and interfacial
16. Sutter, P. W., Flege, J. –I. & Sutter, E. A. Epitaxial graphene on ruthenium. Nature Mater. 7, energies. 1. Derivation and application to interfacial tension. J. Phys. Chem. 61, 904 –909 (1957).
406– 411 (2008). 41. Hodak, M. & Girifalco, L. A. Fullerenes inside carbon nanotubes and multi-walled carbon
17. Pan, Y. et al. Millimetre-scale, highly ordered single crystalline graphene grown on Ru (0001) surface. nanotubes: optimum and maximum sizes. Chem. Phys. Lett. 350, 405 –411 (2001).
ArXiv:0709.2858 (2008). 42. Zacharia, R., Ulbricht, H. & Hertel, T. Interlayer cohesive energy of graphite from thermal desorption
18. Eda, G., Fanchini, G. & Chhowalla, M. Large-area ultrathin films of reduced graphene oxide as a of polyaromatic hydrocarbons. Phys. Rev. B 69, 155406 (2004).
transparent and flexible electronic material. Nature Nanotech. 3, 270 –274 (2008). 43. Jeong, S. H. et al. Preparation of aligned carbon nanotubes with prescribed dimensions: Template
19. Li, D. et al. Processable aqueous dispersions of graphene nanosheets. Nature Nanotech. 3, synthesis and sonication cutting approach. Chem. Mater. 14, 1859– 1862 (2002).
101– 105 (2008). 44. Meyer, J. C. et al. The structure of suspended graphene sheets. Nature 446, 60– 63 (2007).
20. Stankovich, S. et al. Graphene-based composite materials. Nature 442, 282– 286 (2006). 45. Meyer, J. C. et al. On the roughness of single- and bi-layer graphene membranes. Solid State
21. Wang, X., Zhi, L. J. & Mullen, K. Transparent, conductive graphene electrodes for dye-sensitized solar Commun. 143, 101– 109 (2007).
cells. Nano Lett. 8, 323– 327 (2008). 46. Horiuchi, S. et al. Carbon nanofilm with a new structure and property. Jpn J. Appl. Phys. Lett. 42,
22. Stankovich, S. et al. Synthesis of graphene-based nanosheets via chemical reduction of exfoliated L1073 –L1076 (2003).
graphite oxide. Carbon 45, 1558– 1565 (2007). 47. Ferrari, A. C. et al. Raman spectrum of graphene and graphene layers. Phys. Rev. Lett. 97,
23. Jung, I. et al. Simple approach for high-contrast optical imaging and characterization of graphene- 187401 (2006).
based sheets. Nano Lett. 7, 3569 –3575 (2007). 48. Blighe, F. M., Hernandez, Y., Blau, W. J. & Coleman, J. N. Observation of percolation-like scaling, far
24. Dresselhaus, M. S. & Dresselhaus, G. Intercalation compounds of graphite. Adv. Phys. 30, from the percolation threshold, in high volume fraction, high conductivity polymer– nanotube
139– 326 (1981). composite films. Adv. Mater. 19, 4443 –4447 (2007).
25. Viculis, L. M., Mack, J. J. & Kaner, R. B. A chemical route to carbon nanoscrolls. Science 299, Supplementary Information accompanies this paper at www.nature.com/naturenanotechnology.
1361– 1361 (2003).
26. Chen, G. H. et al. Preparation and characterization of graphite nanosheets from ultrasonic Acknowledgements
powdering technique. Carbon 42, 753 –759 (2004). We acknowledge the Centre for Research on Adaptive Nanostructures and Nanodevices and Science
27. Li, X. L. et al. Chemically derived, ultrasmooth graphene nanoribbon semiconductors. Science 319, Foundation Ireland for financial support and Nacional de Grafite (Brazil) for supplying flake graphite.
1229– 1232 (2008). V.N. wishes to thank the EU project ESTEEM for facilitating access to the microscopy facilities in Oxford.
28. Niyogi, S. et al. Solution properties of graphite and graphene. J. Am. Chem. Soc. 128, A.C.F. acknowledges funding from the Leverhulme Trust and the Royal Society.
7720– 7721 (2006).
29. Furtado, C. A. et al. Debundling and dissolution of single-walled carbon nanotubes in amide Author contributions
solvents. J. Am. Chem. Soc. 126, 6095–6105 (2004). J.N.C. conceived and designed the experiments. Y.H., V.N., M.L., F.M.B., Z.S., S.D., B.H., M.B., P.N., S.K.,
30. Giordani, S. et al. Debundling of single-walled nanotubes by dilution: observation of large R.G. and V.S. performed the experiments. I.T.McG., R.G., A.C.F. and J.N.C. analysed the data. Y.K.G.,
populations of individual nanotubes in amide solvent dispersions. J. Phys. Chem. B 110, J.J.B., G.D., R.G., J.H., A.C.F. and J.N.C. contributed materials/analysis tools. A.C.F. and J.N.C. co-wrote
15708–15718 (2006). the paper. Y.H. and V.N. contributed equally to this work. All authors discussed the results and
31. Landi, B. J., Ruf, H. J., Worman, J. J. & Raffaelle, R. P. Effects of alkyl amide solvents on the dispersion commented on the manuscript.
of single-wall carbon nanotubes. J. Phys. Chem. B 108, 17089–17095 (2004).
32. Hasan, T. et al. Stabilization and ‘Debundling’ of single-wall carbon nanotube dispersions in Author information
N-methyl-2-pyrrolidone (NMP) by polyvinylpyrrolidone (PVP). J. Phys. Chem. C 111, Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/.
12594–12602 (2007). Correspondence and requests for materials should be addressed to J.N.C.

568 nature nanotechnology | VOL 3 | SEPTEMBER 2008 | www.nature.com/naturenanotechnology

© 2008 Macmillan Publishers Limited. All rights reserved.
LETTERS

Highly conducting graphene sheets and

Langmuir–Blodgett films
XIAOLIN LI1, GUANGYU ZHANG1, XUEDONG BAI2, XIAOMING SUN1, XINRAN WANG1, ENGE WANG2
AND HONGJIE DAI1*
1
Department of Chemistry and Laboratory for Advanced Materials, Stanford University, Stanford, California 94305, USA
2
Institute of Physics, Chinese Academy of Sciences, Beijing 100080, China
*e-mail: hdai@stanford.edu

Published online: 1 August 2008; doi:10.1038/nnano.2008.210

Graphene is an intriguing material with properties that are free SO3), and inserted tetrabutylammonium hydroxide (TBA,
distinct from those of other graphitic systems1–5. The first 40% solution in water) into the oleum-intercalated graphite
samples of pristine graphene were obtained by ‘peeling off ’2,6 (Fig. 1a) in N,N-dimethylformamide (DMF, see Methods). We
and epitaxial growth5,7. Recently, the chemical reduction of then sonicated the TBA-inserted oleum-intercalated graphite
graphite oxide was used to produce covalently functionalized (Fig. 1b) in a DMF solution of 1,2-distearoyl-sn-glycero-
single-layer graphene oxide8–15. However, chemical approaches 3-phosphoethanolamine-N-[methoxy(polyethyleneglycol)-5000]
for the large-scale production of highly conducting graphene (DSPE-mPEG) for 60 min to form a homogeneous suspension.
sheets remain elusive. Here, we report that the exfoliation – Centrifugation was used to remove large pieces of material from
reintercalation –expansion of graphite can produce high- the supernatant (Fig. 1c; see also Methods). This method resulted
quality single-layer graphene sheets stably suspended in in large amounts of GS suspended in DMF and can be transferrable
organic solvents. The graphene sheets exhibit high electrical to other solvents including water and organic solvents.
conductance at room and cryogenic temperatures. Large We used atomic force microscopy (AFM) to characterize the
amounts of graphene sheets in organic solvents are made into materials deposited on the substrates from the supernatant and
large transparent conducting films by Langmuir– Blodgett observed that 90% of the graphene sheets were single-layer GS
assembly in a layer-by-layer manner. The chemically derived, of various shapes and sizes (Fig. 1d). For the hundreds of GS
high-quality graphene sheets could lead to future scalable measured, we found that the average size of single-layer GS was
graphene devices. 250 nm (see Supplementary Information, Fig. S2a) and the
Several methods have been explored to date to obtain graphene average topographic height was 1 nm (see Supplementary
in solution phase by means of chemical routes. Graphite oxide Information Figs. S2b and S3). Transmission electron microscopy
(GO) has been prepared by harsh oxidation using Hummer’s (TEM, Fig. 1e) and electron diffraction (ED, Fig. 1f ) were used
method16. The as-made GO obtained in this way is electrically to characterize the single-layer GS. The ED pattern of our GS was
insulating, but chemical reduction9,10,15 partially recovers the similar to that of peeled-off graphene23, suggesting a well-
conductivity, albeit at values orders of magnitude below that of crystallized, single-layer graphene structure.
pristine graphene. Irreversible defects and disorder exist in the GO Our starting expandable graphite was prepared by chemical
sheets9,10. The reduced GO exhibits non-metallic behaviour, with intercalation of sulphuric acid and nitric acid24. Upon heating,
the conductance decreasing by about three orders of magnitude they exfoliated violently due to volatile gaseous species released
upon cooling to low temperatures13, whereas pristine graphene is from the intercalant. Most of the exfoliated graphite was still in a
nearly metallic2,17. Recently, we obtained high quality graphene multilayer graphene form25. In order to obtain single-layer GS,
nanoribbons (GNR) by sonicating thermally exfoliated graphite in we used a process of reintercalation with oleum, a chemical
a 1,2-dichloroethane (DCE) solution of poly(m-phenylenevinylene- known to strongly debundle carbon nanotubes through
co-2,5-dioctoxy-p-phenylenevinylene) (PmPV)18. However, the intercalation26. TBA is a molecule capable of inserting into and
yield was low and most of the ribbons had two or more layers. expanding the distance between heavily oxidized graphite layers27.
Despite these and other efforts8–15,19–22, solution phase derivation We proposed that TBA could insert into oleum-intercalated
of single-layer graphene with high electrical conductivity from graphite to increase the distance between adjacent graphitic layers
widely available parent graphite materials has not been achieved on (Fig. 1a), facilitating the separation of GS upon sonication in a
a large scale. The production of stable suspensions of graphene in surfactant solution27. The success of this step was made evident
organic solvents is also an important goal in chemical processing by showing that without this TBA treatment step, with the rest of
and other areas. the method being otherwise identical, the yield of single-layer GS
In the current work, to make high-quality graphene sheets (GS), was extremely low (see Supplementary Information, Fig. S1, for
we started by first exfoliating commercial expandable graphite control experiments). We also found that DMF was a better
(160 –50 N, Grafguard) by brief (60 s) heating to 1,000 8C in solvent than water for our method. DSPE-mPEG is a surfactant
forming gas. We then ground the exfoliated graphite, capable of suspending nanotubes28, another important factor in
reintercalated it with oleum (fuming sulphuric acid with 20% obtaining a homogeneous suspension of GS.

538 nature nanotechnology | VOL 3 | SEPTEMBER 2008 | www.nature.com/naturenanotechnology

© 2008 Macmillan Publishers Limited. All rights reserved.
 LETTERS

X X

X X

TBA
X X X X
DMF
X X

X X

O
O

n
X O
O

n
-O
Sonication P
O O
O
O P -
X O
Phospholipid-PEG OO O
O
O OO X
O
O
X

Graphene
sheet (GS)
X

Figure 1 Chemically derived single-layer GS from the solution phase. a, Schematic of the exfoliated graphite reintercalated with sulphuric acid molecules (teal
spheres) between the layers. b, Schematic of TBA (blue spheres) insertion into the intercalated graphite. c, Schematic of GS coated with DSPE-mPEG molecules and
a photograph of a DSPE-mPEG/DMF solution of GS. d, An AFM image of a typical GS several hundred nanometres in size and with a topographic height of 1 nm
(see Supplementary Information for height details). The scale bar is 300 nm. e, Low- magnification TEM images of a typical GS several hundred nanometres in size.
The scale bar is 100 nm. f, Electron diffraction (ED) pattern of an as-made GS as in e, showing excellent crystallization of the GS.

Our weak oleum treatment condition (soaking in oleum at binding energy corresponding to small amounts of C–O species9,29.
room temperature for one day) is important to obtain high- These species were removed by 800 8C annealing in H2, indicating
quality GS without excessive chemical functionalization and thus the formation of high-quality graphene (Fig. 2b). The annealed GS
property degradation. The conjugate graphene plane is largely exhibited the same XPS spectrum as a pristine highly oriented
free of irreversible modifications throughout the treatment steps. pyrolytic graphite (HOPG) crystal (Fig. 2b), confirming the lack of
Room-temperature oleum treatment is much less oxidative than significant defects or covalent modifications of sp2 carbon in the
Hummer’s method, as is clear from the infrared (IR) spectra, final GS product.
which show that as-made GS has significantly fewer functional We propose the schematic structures of the intermediate and
groups (Fig. 2a,b) than as-made Hummer’s GO (Fig. 2d,e). The final products of our GS and Hummer’s GO as shown in Fig. 2c,f.
IR spectrum of as-made GS (Fig. 2a) showed weaker signals for Oxidization of our intermediate, as-made GS was relatively mild,
the carboxylic groups than the Hummer’s GO (shaded areas in and the few covalently attached functional groups such as the
Fig. 2a,d)29. X-ray photoelectron spectroscopy (XPS) (Fig. 2b) of carboxylic (seen in the IR spectrum, Fig. 2a) and hydroxyl
our as-made GS showed small but noticeable signals at higher groups were most likely at the edges of the as-made GS (Fig. 2c).

nature nanotechnology | VOL 3 | SEPTEMBER 2008 | www.nature.com/naturenanotechnology 539

© 2008 Macmillan Publishers Limited. All rights reserved.
LETTERS
HOOC

OH

HO OH
OH
100
GS as-made 1,800 GS as-made OH
GS annealed OH
95 1,500 HOPG
HO
Transparency (%)

COOH
90 1,200

Counts (a.u.)
85 900 800 ºC

C–H
in H2

O–H
80 600
C–O

300
75
High quality
0
70
1,000 2,000 3,000 4,000 275 280 285 290
Wavenumber (cm–1) Binding energy (eV)

COOH
OH
HO OH
OH OH
HO
HOOC O
100 COOH
GO as-made 1,800 GO as-made COOH

95 GO annealed
1,500 HOPG HO COOH
Transparency (%)

90 1,200
Counts (a.u.)

800 ºC
85 900
C=C
C=O

in H2
O–H

80 600
C–O

300
75 Disordered
0
70
1,000 2,000 3,000 4,000 275 280 285 290
Wavenumber (cm–1) Binding energy (eV)

Figure 2 Comparison of GS and GO sheets. a, IR spectrum (400 – 4,000 cm21) of as-made GS. The shaded region from 1,400 to 1,900 cm21 includes the signal
of the carboxylic groups. b, XPS spectra of as-made, annealed GS and a HOPG crystal. Note the similarity between the spectra of the annealed GS and HOPG.
c, Schematic of the atomic structure of as-made (top) and annealed GS. d, IR spectrum (400 – 4,000 cm21) of as-made GO. The shaded region from 1,400 to
1,900 cm21 includes the signal of the carboxylic groups. e, XPS spectra of as-made, annealed GO and a HOPG crystal. f, Schematic of the atomic structure of
as-made (top) and annealed GO.

This was supported by the fact that the electrical conductivity of gate dielectrics (see Methods and Supplementary Information).
our as-made GS was similar to that of 800 8C vacuum-annealed A typical resistance for 100-nm-wide GS (Fig. 3a) at room
GS (Fig. 3b,c), an unlikely result if the graphene plane was temperature is 10 –30 kV (Fig. 3b,c). The average resistance
heavily modified covalently. The Hummer’s GO was heavily histogram (error bar is the standard deviation) for large numbers
oxidized, with disrupted conjugation in the plane, missing of devices showed that the room-temperature resistance of as-
carbon atoms in the plane30, and abundant functional groups made GS was similar to that of annealed GS devices (for both
such as epoxide, hydroxyl, carbonyl and carboxyl at both the palladium and titanium/gold contacted devices), and about 100
edges and in the plane (Fig. 2f )9,10. Importantly, these abundant times lower than annealed GO (Fig. 3b). As-made GO devices
functional groups weaken the van der Waals interactions between without annealing were all electrically insulating. This result
the layers of GO and make them hydrophilic, which is the reason strongly supports the proposed atomic structures of GS and GO
for the occurrence of single-layer GO exfoliation in aqueous (Fig. 2c,f ) and the assertion that our GS comprise nearly pristine
media to form stable suspensions without the need for an graphene. Our thermally annealed GS retained high electrical
insertion agent such as TBA or the assistance of a surfactant for conductivity, with only a slight increase in resistance at low
suspension. Thermal annealing removed some of the functional temperatures (for both palladium and titanium/gold contacted
groups but was unable to completely repair the holes and other devices), in strong contrast to annealed GO, which was insulating
irreversible defects formed within the plane of the Hummer’s GO at low temperatures (Fig. 3c). Devices comprised of as-made GS
sheets (Fig. 2f )9,10. showed reduced metallic characteristics over annealed GS devices
We fabricated single-GS electrical devices with as-made and (but were still .1,000 times more conducting than GO devices)
annealed GS and Hummer’s GO. We used palladium or with a larger increase in resistance at low temperatures (Fig. 3c).
titanium/gold as source/drain (S/D) metal contacts (channel This suggested the as-made GS contained a small amount of
length L  100 nm), a Pþþ-Si backgate and 500-nm SiO2 for the disorder in the structures.

540 nature nanotechnology | VOL 3 | SEPTEMBER 2008 | www.nature.com/naturenanotechnology

© 2008 Macmillan Publishers Limited. All rights reserved.
 LETTERS

104
107
S 103

Resistance ( x 106 Ω)
Resistance (Ω)
102
106
GS
101

105 100
D

10–1
104
GS 800 °C GS GO 800 °C 0 50 100 150 200 250 300
annealed as-made annealed T (K)

Figure 3 Electrical characterization of a single GS. a, AFM image of a typical device with a single GS (thickness 1 nm, single layer) bridging the channel
(channel length L  100 nm GS with titanium/gold contacts and silicon backgate) between the source (S) and drain (D) electrodes. The scale bar is 200 nm.
b, Mean resistance histograms for 10 devices each of as-made GS, annealed GS and annealed GO. The resistances of as-made GS and annealed GS were similar
(within the error bars representing statistical variations between the GS devices) indicating the high quality of our as-made GS. c, Resistances of as-made GS (green
curve), GS annealed at 800 ºC with a titanium/gold contact (black); GS annealed at 800 ºC with a palladium contact (blue), and GO annealed at 800 ºC (red). The
resistance of GS, particularly the annealed GS, showed only a very small drop in conductance (similar to some of the peel-off pristine graphene samples reported in
the literature) at low temperature.

100 100
Resistance
150 Transparency
90 95

100
80 90

50
70 85

60 0 80
400 800 1,200 1,600 0 1 2 3 4

Figure 4 Large-scale Langmuir-Blodgett (LB) films of GS. a, A photograph of a two-layer GS LB film on quartz with part of it left clear. The scale bar is 10 mm.
b, Transparency spectra of one- (black curve), two- (red curve) and three-layer (green curve) GS LB films. The transparency was defined as the transmittance at a
wavelength of 1,000 nm. c, Resistances (red) and transparencies (blue) of one-, two- and three-layer LB films. The small percentage of bilayer and few-layer GS in
our sample and GS overlapping in the LB film over the substrate contributed to the transparency loss.

To explore the utility of our high-quality GS, we transferred three-layer LB films on quartz (Fig. 4a) afforded sheet resistances
large quantities of GS from DMF to organic solvent DCE, with of 150, 20 and 8 kV at room temperature (Fig. 4c) and
excellent stability against agglomeration. The fact that our as- transparencies (defined as transmittance at a wavelength of
made GS was stably suspended in DCE without the use of 1,000 nm) of 93, 88 and 83%, respectively (Fig. 4b,c). With a
additional surfactant indicates high hydrophobicity of the three-layer LB film, a sheet resistance of 8 kV can be achieved
graphene, consistent with a low degree of graphene oxidation and with a transparency greater than 80%, which compares
covalent functionalization. In contrast, Hummer’s GO was highly favourably with reduced GO films11,12. The conductance and
hydrophilic and completely insoluble in organic solvents. The transparency of our films are comparable to those made of GS
organic stability of our GS enabled Langmuir– Blodgett (LB) formed by sonication of natural graphite in DMF31. This is the
films to be made on various transparent substrates, including first time that high-quality GS have been assembled by the LB
glass and quartz (see Methods and Supplementary Information) technique in a layer-by-layer manner on large substrates. Note
to produce transparent and conducting films. This was done by that with the same method we also succeeded in making GS
adding GS suspensions onto the water subphase, vaporizing the using pristine graphite flakes as the starting material, and the
DCE solvent from the water surface, compressing the floating GS, structural, electrical and spectroscopic properties of the GS made
and transferring the GS LB film onto a substrate by dip-coating. from pristine flakes were similar to those made from expandable
The GS floated on water due to the hydrophobicity within the graphite. Thus, our large-scale synthesis of GS and the ability to
sheet. The edges of the GS contain functional groups, giving rise process it in various solvents for assembly enables its potential
to planar amphiphilic species. We were able to transfer the use in high-performance, scalable applications such as solar cells
GS repeatedly to achieve multilayer films. The one-, two- and using transparent conducting films.

nature nanotechnology | VOL 3 | SEPTEMBER 2008 | www.nature.com/naturenanotechnology 541

© 2008 Macmillan Publishers Limited. All rights reserved.
LETTERS
METHODS one-layer LB film we then transferred another layer GS film onto it by repeating
the LB procedure. We were able to obtain multilayer LB films by this layer-by-
PREPARATION OF GRAPHENE SHEET SUSPENSION layer transfer method. The transparency of the GS films was measured with a
Our single-layer GS preparation started by exfoliating expandable graphite Cary 6000i spectrophotometer using pure quartz as the background. The
(160 –50 N of Grafguard) at 1,000 8C in forming gas for 60 s. Exfoliated graphite transparency was defined as the transmittance at a wavelength of 1,000 nm.
(10 mg) was then ground with NaCl crystallites for 3 min, forming a uniform
greyish mixture. Small pieces of exfoliated graphite were separated and collected Received 31 March 2008; accepted 27 June 2008; published 1 August 2008.
by dissolving the NaCl with water and then filtration. The resulting sample was
then treated with oleum at room temperature for a day. After complete References
1. Geim, A. K. & Novoselov, K. S. The rise of graphene. Nature Mater. 6, 183 –191 (2007).
removal of acid by filtration and repeated washing, the resulting sample was 2. Novoselov, K. S. et al. Electric field effect in atomically thin carbon films. Science 306,
ultrasonicated using a cup-horn sonicator in a DMF (10 ml) solution of TBA 666– 669 (2004).
(130 ml) for 5 min. The suspension was held at room temperature for 3 days to 3. Novoselov, K. S. et al. Two-dimensional gas of massless Dirac fermions in graphene. Nature 438,
allow the TBA to fully insert into the graphene layers. A 5 ml suspension was then 197– 200 (2005).
4. Zhang, Y. B., Tan, Y. W., Stormer, H. L. & Kim, P. Experimental observation of the quantum Hall
taken out and bath-sonicated with DSPE-mPEG (Laysan Bio.) (15 mg) for 1 h, effect and Berry’s phase in graphene. Nature 438, 201– 204 (2005).
forming a homogeneous suspension. After centrifuging the suspension at 5. Berger, C. et al. Electronic confinement and coherence in patterned epitaxial graphene. Science 312,
24,000g for 3 min, we obtained a black suspension with predominantly single- 1191–1196 (2006).
6. Novoselov, K. S. et al. Two-dimensional atomic crystals. Proc. Natl Acad. Sci. USA 102,
layer GS retained in the supernatant. 10451–10453 (2005).
7. Berger, C. et al. Ultrathin epitaxial graphite: 2D electron gas properties and a route toward graphene-
CHARACTERIZATION OF GRAPHENE SHEETS based nanoelectronics. J. Phys. Chem. B 108, 19912–19916 (2004).
8. Dikin, D. A. et al. Preparation and characterization of graphene oxide paper. Nature 448,
AFM images were taken of the GS with a Nanoscope IIIa multimode instrument.
457– 460 (2007).
The samples were prepared by soaking a SiO2 substrate (pretreated with a 4 mM 9. Stankovich, S. et al. Stable aqueous dispersions of graphitic nanoplatelets via the reduction of
3-aminopropyltriethoxysilane (APTES) water solution for 20 min) in the exfoliated graphite oxide in the presence of poly(sodium 4-styrenesulfonate). J. Mater. Chem. 16,
graphene suspension for 20 min, rinsing with water and blow-drying with argon. 155– 158 (2006).
10. Stankovich, S. et al. Synthesis of graphene-based nanosheets via chemical reduction of exfoliated
The substrate was calcined to 350 8C and annealed at 800 8C in H2 before AFM. graphite oxide. Carbon 45, 1558–1565 (2007).
The IR spectrum (400 –4,000 cm21) was measured using a Nicolet IR100 FT-IR 11. Gilje, S., Han, S., Wang, M. S., Wang, K. L. & Kaner, R. B. A chemical route to graphene for device
spectrometer with pure KBr as the background. After removal of the surfactant applications. Nano Lett. 7, 3394 –3398 (2007).
12. Li, D., Muller, M. B., Gilje, S., Kaner, R. B. & Wallace, G. G. Processable aqueous dispersions of
by filtration and repeated washing, the graphene sample was collected and graphene nanosheets. Nature Nanotech. 3, 101– 105 (2008).
ground with KBr. The mixture was dried and compressed into a transparent 13. Gomez-Navarro, C. et al. Electronic transport properties of individual chemically reduced graphene
tablet for measurement. We characterized our GS using a JEOL 2010F FEG oxide sheets. Nano Lett. 7, 3499 –3503 (2007).
14. Wang, X., Zhi, L. J. & Mullen, K. Transparent, conductive graphene electrodes for dye-sensitized solar
transmission electron microscope (TEM) at an accelerating voltage of 120 kV. cells. Nano Lett. 8, 323 –327 (2008).
The TEM samples were prepared by drying a droplet of the graphene suspension 15. Bourlinos, A. B. et al. Graphite oxide: chemical reduction to graphite and surface modification with
on a lacey carbon grid. High-resolution XPS measurement was carried out using primary aliphatic amines and amino acides. Langmuir 19, 6050– 6055 (2003).
16. Hummers, W. S. & Offeman, R. E. Preparation of graphite oxide. J. Am. Chem. Soc. 80, 1339 (1958).
an SSI S-Probe monochromatized XPS spectrometer, which used Al (Ka) 17. Tan, Y. W., Zhang, Y. B., Stormer, H. L. & Kim, P. Temperature dependent electron transport in
radiation as a probe. The analysis spot size was 150  800 mm. Sample graphene. Eur. Phys. J. 148, 15–18 (2007).
preparation involved removal of the surfactant by filtration and repeated 18. Li, X. L., Wang, X. R., Zhang, L., Lee, S. W. & Dai, H. J. Chemically derived, ultrasmooth graphene
nanoribbon semiconductors. Science 319, 1229– 1232 (2008).
washing, and depositing the materials onto a silicon substrate by repeated drop-
19. Stankovich, S. et al. Graphene-based composite materials. Nature 442, 282 –286 (2006).
drying. The GO sample was prepared by depositing materials onto a silicon 20. Schniepp, H. C. et al. Functionalized single graphene sheets derived from splitting graphite oxide.
substrate by repeated drop-drying. The HOPG sample was used for XPS J. Phys. Chem. B 110, 8535–8539 (2006).
measurement without any treatment. 21. Yu, A. P., Ramesh, P., Itkis, M. E., Bekyarova, E. & Haddon, R. C. Graphite nanoplatelet– epoxy
composite thermal interface materials. J. Phys. Chem. C 111, 7565–7569 (2007).
22. Niyogi, S. et al. Solution properties of graphite and graphene. J. Am. Chem. Soc. 128,
GRAPHENE SHEET AND GRAPHITE OXIDE DEVICE FABRICATION 7720 –7721 (2006).
GS and GO were deposited onto a 500-nm SiO2/Pþþ Si substrate (pretreated with 23. Meyer, J. C. et al. The structure of suspended graphene sheets. Nature 446, 60– 63 (2007).
24. Greinke, R. A. et al. Expandable graphite and method. US patent 6416815 B2 (2002).
4 mM APTES solution). After removal of the surfactant by 350 8C calcination
25. Han, J. H., Cho, K. W., Lee, K.-H. & Kim, H. Porous graphite matrix for chemical heat pumps.
and 800 8C H2 annealing, we used electron-beam lithographic patterning Carbon 36, 1801 –1810 (1998).
followed by electron-beam evaporation of palladium (20 nm) or titanium 26. Ericson, L. M. et al. Macroscopic, neat, single-walled carbon nanotube fibres. Science 305,
(1.5 nm)/gold (20 nm) to randomly form the source and drain electrodes 1447 –1450 (2004).
27. Liu, Z. H., Wang, Z. M., Yang, X. J. & Ooi, K. Intercalation of organic ammonium ions into layered
(channel length 100 nm, width 2 mm) on the substrate. The sample was graphite oxide. Langmuir 18, 4926– 4932 (2002).
then annealed in argon at 300 8C for 15 min to improve the contacts between the 28. Kam, N. W. S., O’Connell, M., Wisdom, J. A. & Dai, H. J. Carbon nanotubes as multifunctional
source and drain metal and the GS/GO in the channel region. biological transporters and near-infrared agents for selective cancer cell destruction. Proc. Natl Acad.
Sci. USA 102, 11600– 11605 (2005).
29. Hontoria-Lucas, C., Lopez-Peinado, A. J., Lopez-Gonzalez, J. de D., Rojas-Cervantes, M. L. &
LANGMUIR–BLODGETT FILM FABRICATION Martin-Aranda, R. M. Study of oxygen-containing groups in a series of graphite oxides: physical and
The DMF suspension of GS was centrifuged at 24,000g for 1 h to remove the chemical characterization. Carbon 33, 1585 –1592 (1995).
surfactants. The aggregates were then resuspended in fresh DMF by a brief 30. Kuznetsova, A. et al. Enhancement of adsorption inside of single-walled nanotubes: opening the
entry ports. Chem. Phys. Lett. 321, 292– 296 (2000).
sonication step. This centrifugation and resuspending process was repeated three 31. Blake, P. et al. Graphene-based liquid crystal device. Nano Lett. 8, 1704–1708 (2008).
times. The GS samples were then resuspended in fresh DCE and the
Supplementary Information accompanies this paper at www.nature.com/naturenanotechnology.
centrifugation and resuspending process repeated a further three times to ensure
complete removal of the DSPE-mPEG. The resulting GS were suspended in DCE Acknowledgements
by a 5-min sonication step. GS LB films were made using a commercial KSV- We thank Graftech for providing expandable graphite samples. This work was supported in part by Intel,
Minimicro 2000 LB trough. Approximately 1.2 ml of GS/DCE suspension was the Microelectronics Advanced Research Corporation Materials, Structures and Devices (MARCO MSD)
Focus Centre and the Office of Naval Research.
added to a water subphase in the LB trough. A platinum plate was used to
monitor the surface tension during compression of the GS on the water subphase Author contributions
by moving the two opposing barriers towards each other. At a target surface H.D. and X.L. conceived and designed the experiments. X.L. and G.Z. performed the experiments and
pressure of 27 mN m21, GS were compressed to form a dense LB film analysed the data. H.D. and X.L. co-wrote the manuscript. All authors discussed the results and
transferrable onto a solid substrate (up to 1  1 inch2) by slowly pulling the commented on the manuscript.

substrate out of the aqueous subphase. The transferred GS LB film was typically Author information
calcined at 350 8C to remove DSPE-mPEG and TBA residues before transparency Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/.
and resistance measurement. After calcination of the quartz substrate with a Correspondence and requests for materials should be addressed to H.D.

542 nature nanotechnology | VOL 3 | SEPTEMBER 2008 | www.nature.com/naturenanotechnology

© 2008 Macmillan Publishers Limited. All rights reserved.
 LETTERS

Tailoring the atomic structure of

graphene nanoribbons by scanning
tunnelling microscope lithography
LEVENTE TAPASZTÓ1*, GERGELY DOBRIK1, PHILIPPE LAMBIN2 AND LÁSZLÓ P. BIRÓ1
1
Research Institute for Technical Physics and Materials Science, H-1525 Budapest, Hungary
2
Facultes Universitaire Notre Dame de la Paix, 61 Rue de Bruxelles, B-5000 Namur, Belgium
*e-mail: tapaszto@mfa.kfki.hu

Published online: 8 June 2008; doi:10.1038/nnano.2008.149

The practical realization of nanoscale electronics faces two major

challenges: the precise engineering of the building blocks and
their assembly into functional circuits1. In spite of the
exceptional electronic properties of carbon nanotubes2, only
basic demonstration devices have been realized that require
time-consuming processes3–5. This is mainly due to a lack
nm
of selective growth and reliable assembly processes for 0.5 nm
1
nanotubes. However, graphene offers an attractive alternative.
Here we report the patterning of graphene nanoribbons and 40 20
80 40
bent junctions with nanometre-precision, well-defined widths 120 nm 60 80 100 nm

and predetermined crystallographic orientations, allowing us

to fully engineer their electronic structure using scanning
tunnelling microscope lithography. The atomic structure and Figure 1 Graphene nanostructures patterned by STM lithography.
electronic properties of the ribbons have been investigated by a, 3D STM image of a 10-nm-wide and 120-nm-long graphene nanoribbon.
scanning tunnelling microscopy and tunnelling spectroscopy b, An 8-nm-wide 3088 GNR bent junction connecting an armchair and
measurements. Opening of confinement gaps up to 0.5 eV, a zigzag ribbon.
enabling room-temperature operation of graphene
nanoribbon-based devices, is reported. This method avoids the
difficulties of assembling nanoscale components and may control of their structure with nanometre precision is of
prove useful in the realization of complete integrated circuits, paramount importance.
operating as room-temperature ballistic electronic devices6,7. With standard e-beam lithographic methods, GNRs down to a
High electron mobility and long coherence length make few tens of nanometres in width have been realized recently12,13; this
graphene a subject of intense focus for nanoscale electronic seems to be the limit of this technique1. However, GNRs of just a
applications, even for the realization of room-temperature few nanometres width are required to obtain energy gaps
ballistic (dissipation-free) electronics8. However, a major setback adequate for room-temperature operation, because the gap scales
in the development of graphene-based field-effect transistors is inversely with ribbon width10,12. It is obvious that this scale
the inability to electrostatically confine electrons in graphene, cannot be achieved by e-beam lithography. Moreover, electron
because a single layer of graphite remains metallic even at the lithographic methods have difficulties in controlling the
charge neutrality point9. In order to overcome this problem, a crystallographic orientation of the ribbons.
way to open a gap in the electronic structure of graphene has Scanning probe microscopy techniques and in particular
to be found. A straightforward solution is to pattern the scanning tunnelling microscopy (STM) might offer the solution.
graphene sheet into narrow ribbons, which can be viewed as STM combines the capability of atomic-resolution imaging with
unrolled single-walled carbon nanotubes (CNTs). Owing to the the ability to locally modify the surface of the samples14. The
quantum mechanical constriction of electronic wavefunctions in modification of the graphite surface by STM has an almost
the direction perpendicular to the axis of the ribbon, a 20-year-old history15. However, previous works have focused on
confinement-induced gap can open. Theoretical works have studying the mechanism of surface modification by analysing pit
predicted that the width and crystallographic orientation of the formation in graphite. Although the potential for etching lines
graphene nanoribbons (GNRs) have a strong dependence on has also been demonstrated16, no further significant advance has
this gap10,11, offering us the opportunity to tailor their been reported. In our work we combine the STM feature of
electronic structure. However, it also imposes strict surface modification with atomic-resolution imaging in order to
requirements on the fabrication of GNR-based devices, because engineer nanostructures with almost atomically precise structures
in order to provide sufficient reproducibility, the accurate and predetermined electronic properties.

nature nanotechnology | VOL 3 | JULY 2008 | www.nature.com/naturenanotechnology 397

© 2008 Macmillan Publishers Limited. All rights reserved.
LETTERS
1 nm

Figure 2 Atomic structure of GNRs. a, Atomic-resolution STM image (20  20 nm2, 1 nA, 200 mV) of a 15-nm-wide GNR displaying an atomically flat and
defect-free structure. The colour scale bars encode the height of the imaged features. b,c, Magnified images of the defect-free lattice taken at the centre of the
ribbon (b) and position-dependent superstructures near the edges (c). d, Identification of crystallographic orientation from the triangular lattice observed in
atomic-resolution STM images of HOPG-supported GNR. e, Theoretical STM image of the superstructures at the edges of the ribbon.

1 nm
λ = 0.41 nm interference

λ = 0.25 nm atomic period

0 5.0 10.0
(nm)

Eg = 0.18 eV

dI/dV (a.u.)

–0.4 –0.2 0 0.2

Vbias (V)

Figure 3 Electronic structure of GNRs. a, Constant-current STM image (12  12 nm2, 1 nA, 100 mV) of a 10-nm-wide armchair GNR displaying confinement-induced
standing electron wave patterns (stripes parallel to the axis of the ribbon). The colour scale bars encodes the height of the imaged features. b, Average line-cuts
revealing the period of the observed oscillation, which clearly differs from the periodicity of the underlying atomic structure. c, 2D Fourier transformation of the STM
image. d, Representative tunnelling spectra (STS) taken on the ribbon, revealing an energy gap of 0.18 eV (zero density of states (DOS) marked by horizontal lines).

Cutting of GNRs was carried out by applying a constant bias lithographic parameters (2.4 V bias potential and 2.0 nm s21 tip
potential (significantly higher than the one used for imaging) velocity) we were able to cut GNRs with suitably regular edges,
and simultaneously moving the STM tip with constant velocity which constitutes a great advance towards the reproducibility of
in order to etch the desired geometry fitted to the GNR-based devices. Furthermore, more complex graphene nano-
crystallographic structure, which was known from previous architectures can be tailored by STM lithography. As a basic
atomic-resolution STM imaging (see Methods for details). demonstration, Fig. 1b shows an 8-nm-wide, 30 8 GNR bent
Figure 1a shows a 10-nm-wide and 120-nm-long graphene junction connecting an armchair and a zigzag ribbon, giving rise
nanoribbon etched by STM lithography. By setting the optimal to a metal –semiconductor molecular junction7. After the

398 nature nanotechnology | VOL 3 | JULY 2008 | www.nature.com/naturenanotechnology

© 2008 Macmillan Publishers Limited. All rights reserved.
 LETTERS

1 nm

0.4

(nm)
0.3

0.2

0 0.4 0.8 1.2

(nm)

Figure 4 Tight-binding computation of the STM image of GNRs. 0.5

Topographical STM image at constant current calculated for a 1.7-nm-wide
W = 2.5 nm
armchair GNR at a tip potential V t ¼ 0.5 V and tip – GNR distance of 0.5 nm.

Depth (nm)
0

patterning process, in situ atomic-resolution STM images could be

achieved, revealing the atomic structure of the GNRs. The ribbon
shown in Fig. 2a is 15 nm wide and its axis has a 0 5.0 10.0 15.0
Length (nm)
crystallographic orientation close to the zigzag direction. The
atomic-resolution image reveals an atomicallypffiffiffi flat
pffiffiand
ffi defect-free
structure far from the edges, where a 3  3 R 30 8 type
superstructure pattern can be observed. The origin of this pattern Eg = 0.5 eV
dI /dV (a.u.)

is the interference of the electrons scattered at the irregularities of

the edges17,18 (Fig. 2e). For electronic device applications these
ribbons have to be deposited on an insulating substrate.
However, for atomic-resolution STM and scanning tunnelling
spectroscopy (STS) investigation of GNRs, the surface of highly
oriented pyrolytic graphite (HOPG) provides an ideal substrate, –0.6 –0.4 –0.2 0 0.2 0.4
as it allows the ribbon to remain flat without inducing additional Vbias (V)
defects or changing its topology19–21. The intrinsic characteristics
of GNRs can be studied this way.
When imaging a 10-nm-wide armchair GNR at low bias voltage Figure 5 GNRs for room-temperature electronics. a, STM image
(100 mV), oscillations in the electron density distribution parallel (15  15 nm2, 1 nA, 100 mV) of a 2.5-nm-wide armchair GNR. The colour scale
to the axis of the ribbon, reminiscent of a Fabry– Perot electron bars encodes the height of the imaged features. b, Average line-cut of the STM
resonator, were observed, as shown in Fig. 3a. The periodicity of image revealing the real width of the ribbon. c, Representative STS spectra
the observed oscillations was 0.4 nm, which clearly differs from taken on the narrow ribbon showing an energy gap of about 0.5 eV (zero DOS
the period of the underlying atomic structure (0.246 nm), and marked by horizontal lines).
corresponds to the Fermi wavelength of electrons in graphene
(Fig. 3b). These findings are also supported by Fourier
transformed images, which clearly indicate the presence of a
lower frequency oscillation (red circles) than the atomic across the ribbon. The presence of continuous interference
periodicity (yellow circles) in Fig. 3c. To correctly interpret patterns along the whole length of the ribbon in the experimental
the measurements we have performed theoretical modelling of images is spectacular evidence of phase-coherent quantum
STM images of GNRs based on a tight-binding p-electron billiard (a standing electron wave) in GNRs at room temperature,
Hamiltonian method22. Previously, the method has been used which demonstrates their behaviour as electronic waveguides,
successfully in the interpretation of atomic-resolution STM even under ambient conditions. Both real and reciprocal space
images of CNTs. Figure 4 shows the calculated constant-current images (Fig. 3b,c, respectively) reveal a single characteristic
STM image of a 1.7-nm-wide armchair GNR. Stripes of high oscillation period, indicating the one-dimensional (1D) nature of
electron density distribution running parallel to the axis of the the electronic structure of the narrow GNRs. (See Supplementary
ribbon are clearly visible. The period of these oscillations is about Information, Figs S3 and S4, for bias-dependent and larger-area
0.37 nm, which is close to the experimentally observed STM images, respectively.)
periodicity of 0.41 nm measured in Fig. 3. The small difference To experimentally investigate the electronic band structure of
between the calculated and measured values might occur due to GNRs, STS measurements have been performed. Representative
the doping of the GNR edges in air. At low bias voltages the STM STS spectra taken on the 10-nm-wide armchair GNR are shown
measurements map the (square modulus of the) electronic in Fig. 3d. The dI/dV quantity, which is proportional to the local
wavefunction near the Fermi level23, so we attribute these electronic density of states (LDOS), revealed the presence of van
oscillations to the quantum mechanical confinement of electrons Hove singularities, a signature of a 1D electronic structure, as is

nature nanotechnology | VOL 3 | JULY 2008 | www.nature.com/naturenanotechnology 399

© 2008 Macmillan Publishers Limited. All rights reserved.
LETTERS
well known for CNTs (ref. 24). (See Supplementary Information, sample was then rotated in order to set the desired crystallographic orientation of
Fig. S2, for a comparison with the LDOS of HOPG.) the ribbon axis (edges), and the graphene layer was cut by applying a constant
Furthermore, the distance between the first pair of van Hove bias potential (significantly higher than the one used for imaging) and
simultaneously moving the STM tip with constant velocity in order to etch the
singularities gives the value of the energy gap24. For the 10-nm-
desired geometry. Good results were obtained for positive sample biases. The
wide armchair GNR, a gap value of 0.18 eV was found, which microscopic mechanism of etching is not yet fully understood. Most likely the
seems to suit well the theoretical rule for separation of the energy breaking of carbon –carbon bonds by field-emitted electrons combined with the
levels due to the geometrical constriction of wavefunctions25 electron-transfer-enhanced oxidation of the graphene is responsible for the
Eg(W) ¼ p.n0/W  (2 eV . nm)/W, where W is the width of the etching29. The whole lithographic process was controlled by a custom-written
ribbon and n0 is the Fermi velocity of electrons in graphene. Our computer code enabling us to manage several parameters. The applied bias
results concerning the electronic waveguide behaviour of GNRs potential and the velocity of the tip during patterning were found to be the
are also in accordance with those derived from transport critical parameters. The optimal parameters were slightly dependent on the
measurements25,26 at low temperatures and for much microstructure of the mechanically etched tips, but typically varied in the range
wider ribbons. of 2.2–2.6 V for the bias voltages and 1.0–5.0 nm s21 for the tip velocities. Here
we note that during the patterning we used tip velocities that were three orders of
Another advantage of STM lithography, beyond the precise
magnitude lower than those reported in ref. 16, and moved the tip only once
engineering of the structure of GNRs, is the potential for along the lines to be etched, instead of scanning it several hundred times with
downscaling. We were able to etch GNRs down to a width of high velocity. After optimizing the parameters of the lithographic process we
2.5 nm, corresponding to 10 carbon ring units along the width were able to cut single-layer trenches (0.335 nm deep), down to a width of 1 nm,
of the ribbon (Fig. 5a). The parameters used for lithography and several hundreds of nanometres long, with well-defined edges. (See
were 2.28 V bias potential and 1.0 nm s21 tip velocity. The width Supplementary Information, Fig. S1, for experimental evidence of patterning
and crystallographic orientation of the GNR shown in Fig. 5a only a single graphene layer.) The STM lithographic patterning showed good
correspond to an unrolled (10,0) zigzag single-walled CNT. The stability and reproducibility, even under ambient conditions. After several
STS measurements performed on the 2.5-nm-wide ribbon shown etching processes the quality of the STM tip was still good enough to achieve
atomic-resolution images of the GNRs, enabling us to investigate in situ their
in Fig. 5c reveal an energy gap of about 0.5 eV, in excellent
atomic structure.
agreement with first-principles theoretical calculations11. Tight
binding calculations predict a gap of 1.2 eV . nm/W, which
is also in good agreement with our results. Furthermore, the Received 16 January 2008; accepted 1 May 2008; published 8 June 2008.
gap value of 0.5 eV is comparable with that of Ge (0.67 eV),
allowing the room-temperature operation of GNR-based
References
electronic devices. 1. Avouris, P., Chen, Z. & Perebeinos, V. Carbon-based electronics. Nature Nanotech.
However, unlike CNTs, which are seamless graphitic structures 2, 605– 615 (2007).
2. Avouris, P. Molecular electronics with carbon nanotubes. Acc. Chem. Res. 35, 1026 –1034 (2002).
rolled into a perfect cylinder, GNRs have edges. Because very 3. Tans, S. J., Verschueren, A. R. M. & Dekker, C. Room-temperature transistor based on a single carbon
narrow GNRs are needed to achieve the desired gap, the effect of nanotube. Nature 393, 49–52 (1998).
4. Yao, Z., Postma, H. W. C., Balents, L. & Dekker, C. Carbon nanotube intramolecular junctions.
edges can be critical. It has been shown theoretically that edge Nature 402, 273 –276 (1999).
irregularities might induce electronic states within the gap region27. 5. Keren, K., Berman, R. S., Buchstab, E., Sivan, U. & Braun, E. DNA-templated carbon nanotube field-
effect transistor. Science 302, 1380 –1382 (2003).
Our STM measurements also reveal oscillations in the electronic 6. Areshkin, D. A. & White, C. T. Building blocks for integrated graphene circuits. Nano Lett. 7,
density of states (DOS) of the 2.5-nm-wide ribbon. These 3253–3259 (2007).
oscillations are less regular than the ones observed for wider 7. Yan, Q. et al. Intrinsic current –voltage characteristics of graphene nanoribbon transistors and effect
of edge doping. Nano Lett. 7, 1469 –1473 (2007).
ribbons; their orientation encloses a 308 angle with the ribbon axis, 8. Geim, A. K. & Novoselov, K. S. The rise of graphene. Nature Mater. 6, 183 –191 (2007).
and the corresponding energy lies within the expected gap, which 9. Novoselov, K. S. et al. Two-dimensional gas of massless Dirac fermions in graphene. Nature 438,
197– 200 (2005).
suggests that the imaged states are probably related to edge disorder. 10. Barone, V., Hod, O. & Scuseria, G. E. Electronic structure and stability of semiconducting graphene
Although SPM methods are usually characterized by low nanoribbons. Nano Lett. 6, 2748–2754 (2006).
11. Son, Y. W., Cohen, M. L. & Louie, S. G. Energy gaps in graphene nanoribbons. Phys. Rev. Lett. 97,
throughput, the manufacturing of large-scale integrated circuits 216803 (2006).
of graphene by the SPM lithographic process presented above is 12. Han, M. Y., Ozylmaz, B., Zhang, Y. & Kim, P. Energy band-gap engineering of graphene nanoribbons.
Phys. Rev. Lett. 98, 206805 (2007).
not unrealistic, as ‘Millipede’-type28 scanning probe microscopes 13. Chen, Z., Lin, Y. M., Rooks, M. J. & Avouris, P. Graphene nano-ribbon electronics. Physica E 40,
(SPM) are able to work in parallel, with more than 1,000 tips, 228 –232 (2007).
14. Tseng, A. A., Notargiacomo, A. & Chen, T. P. Nanofabrication by scanning probe microscope
which can significantly increase the efficiency. lithography: A review. J. Vac. Sci. Technol. B 23, 877– 894 (2005).
In summary, we have developed an STM lithography-based 15. Albrecht, T. R. et al. Nanometer-scale hole formation on graphite using a scanning tunneling
technology that allows the engineering of graphene with true microscope. Appl. Phys. Lett. 55, 1727– 1729 (1989).
16. McCarley, R. L., Hendricks, S. A. & Bard, A. J. Controlled nanofabrication of highly oriented
nanometre precision. We etched semiconducting ribbons from pyrolytic graphite with the scanning tunneling microscope. J. Phys. Chem. 96, 10089–10092 (1992).
graphene with predetermined energy gap values up to 0.5 eV, 17. Mizes, H. A. & Foster, J. S. Long-range electronic perturbations caused by defects using scanning
tunneling microscopy. Science 244, 559– 562 (1989).
allowing their operation at room temperature. These GNR 18. Tapasztó, L. et al. Electron scattering in a multiwall carbon nanotube bend junction studied by
devices also show a phase-coherent behaviour, even under scanning tunneling microscopy. Phys. Rev. B 74, 235422 (2006).
19. Ishigami, M., Chen, J. H., Cullen, W. G., Fuhrer, M. S. & Williams, E. D. Atomic structure of
ambient conditions. Furthermore, STM lithography offers us the graphene on SiO2. Nano Lett. 7, 1643 –1648 (2007).
opportunity for patterning more complicated architectures, even 20. Rutter, G. M. et al. Scattering and interference in epitaxial graphene. Science 317, 219– 222 (2007).
21. Gomez-Navarro, C. et al. Electronic transport properties of individual chemically reduced graphene
complete integrated circuits, from networks of GNRs. One can oxide sheets, Nano Lett. 7, 3499 –3503 (2007).
imagine the feasibility of a variety of electronic devices based on 22. Meunier, V. & Lamnin, P. Tight-binding computation of the STM image of carbon nanotubes. Phys.
our technique, which could open up new directions in the Rev. Lett. 81, 5588–5591 (1998).
23. Tersoff, J. & Hamann, D. R. Theory of the scanning tunneling microscope. Phys. Rev. B 31,
experimental realization of graphene-based electronics. 805 –813 (1985).
24. Wildöer, J. W. G., Venema, L. C., Rinzler, A. G., Smalley, R. E. & Dekker, C. Electronic structure of
atomically resolved carbon nanotubes. Nature 391, 59–62 (1998).
METHODS 25. Berger, C. et al. Electronic confinement and coherence in patterned epitaxial graphene. Science 312,
1191 –1196 (2006).
26. Miao, F. et al. Phase-coherent transport in graphene quantum billiards. Science 317,
In order to pattern a single graphene sheet on the surface of a HOPG sample, we 1530 –1533 (2007).
used a commercial DI Nanoscope III STM operating under ambient conditions. 27. Yoon, Y. & Guo, J. Effect of edge roughness in graphene nanoribbon transistors. Appl. Phys. Lett. 91,
73103 (2007).
Pt– Ir tips proved to be the most suitable for both imaging and lithography. First, 28. Knoll, A. et al. Integrating nanotechnology into a working storage device. Microelectron. Eng. 83,
atomic-resolution images were taken on the atomically flat graphene sheet. The 1692 –1697 (2006).

400 nature nanotechnology | VOL 3 | JULY 2008 | www.nature.com/naturenanotechnology

© 2008 Macmillan Publishers Limited. All rights reserved.
LETTERS

Large-area ultrathin films of reduced

graphene oxide as a transparent and
flexible electronic material
GOKI EDA, GIOVANNI FANCHINI AND MANISH CHHOWALLA*
Materials Science and Engineering, Rutgers University, 607 Taylor Road, Piscataway, New Jersey 08854, USA
*e-mail: manish1@rci.rutgers.edu

Published online: 6 April 2008; doi:10.1038/nnano.2008.83

The integration of novel materials such as single-walled carbon concept electrodes for organic photovoltaics (see Supplementary
nanotubes and nanowires into devices has been challenging, but Information). A GO aqueous suspension can be readily obtained
developments in transfer printing and solution-based methods from exfoliation of graphite through oxidation26,27 (see
now allow these materials to be incorporated into large-area Methods). Methods such as drop casting20,24, rapid freezing by
electronics1–6. Similar efforts are now being devoted to making spraying21 and dip coating22 from GO suspension have been used
the integration of graphene into devices technologically to obtain isolated individual and multilayered sheets or thin
feasible7–10. Here, we report a solution-based method that films. In order to reproducibly achieve uniform thin films with a
allows uniform and controllable deposition of reduced controllable number of GO layers over large areas, we have used
graphene oxide thin films with thicknesses ranging from a the vacuum filtration method, which has been used widely to
single monolayer to several layers over large areas. The opto- deposit highly uniform single-walled carbon nanotube (SWNT)
electronic properties can thus be tuned over several orders of thin films2,28,29.
magnitude, making them potentially useful for flexible and Vacuum filtration involves the filtration of a GO suspension
transparent semiconductors or semi-metals. The thinnest films through a commercial mixed cellulose ester membrane with an
exhibit graphene-like ambipolar transistor characteristics, average pore size of 25 nm. As the suspension is filtered through
whereas thicker films behave as graphite-like semi-metals. the ester membrane, the liquid is able to pass through the pores,
Collectively, our deposition method could represent a route for but the GO sheets become lodged. The permeation rate of the
translating the interesting fundamental properties of graphene solvent is controlled by the accumulation of the GO sheets on
into technologically viable devices. the pores so as the number of GO layers increases at a given
Electronic devices constructed from a single layer of graphite, location on the porous membrane, the rate of filtration decreases,
referred to as graphene11, have received significant attention. but does so to a lesser degree at thinner or uncovered regions.
Graphene is a 0 eV bandgap semiconductor in which the filled The process is therefore self-regulating, which allows reasonably
valence band touches the empty conduction band, thus giving good nanoscale control over the film thickness by simply varying
rise to peculiar properties12 that could have particularly either the concentration of the GO in the suspension or the
interesting applications in electronic devices13–16. The discovery of filtration volume. The GO flakes on the filter membrane can then
isolated graphene obtained from the simple mechanical cleaving be transferred by placing the membrane with the film side down
‘Scotch tape method’11 has made fabrication of devices on onto a substrate and dissolving the membrane with acetone,
individual graphene sheets straightforward. Effort is also under leaving behind a uniform GO thin film (see Methods for details).
way to grow large-area epitaxial graphene17,18. In addition, The yield of the transfer process is nearly 100%, independent of
promising approaches based on transfer printing of exfoliated the substrate, indicating that van der Waals interactions give rise
graphene onto electrodes on different substrates for large-scale to sufficiently strong cohesive forces within the film and also
integration have been reported recently7–9. between the GO sheets and the substrate to obtain a well adhered
In addition to individual sheet devices, efforts to obtain uniform film. Indeed, the as-deposited thin films of GO are able
graphene-based composites through the reduction of graphene to withstand typical lithographic processes (rinsing, blowing with
oxide (GO) in solution and incorporation into hosts have also dry nitrogen and deposition of electrodes) without any evidence
yielded promising results19. Recently, there have been reports of of delamination. A GO thin film covering an area of 10 cm2 on
non-composite reduction of GO into graphene using chemical an ester membrane is shown in Fig. 1a, and transferred films on
routes and high-temperature annealing20–25. The chemical glass and plastic substrates are shown in Fig. 1b,c, respectively.
approach is appealing because it opens a route for the deposition The thicknesses obtained by atomic force microscope (AFM)
of graphene from solution, allowing devices to be fabricated on profilometry, ellipsometry and Raman spectroscopy were found
virtually any surface. to be 1–2 nm for the films deposited at a filtration volume of
We describe a simple and reproducible method to uniformly 20 ml and 3–5 nm at 80 ml, suggesting that the thinnest films
deposit between one and five layers of graphene from reduced consist of single layers of GO (ref. 23) (concentration ¼
GO in the form of thin films to create transistors and proof-of- 0.33 mg l21; see Supplementary Information for experimental

270 nature nanotechnology | VOL 3 | MAY 2008 | www.nature.com/naturenanotechnology

© 2008 Nature Publishing Group
 LETTERS

Figure 1 Thin films of solution-processed GO. a – c, Photographs of GO thin films on filtration membrane (a), glass (b) and plastic (c) substrates.

0 5 10 15 0 5 10 15

5 5 1
2
3–5
>5
10 10

N=1
N>5
Intensity (a.u.)

2,500 2,600 2,700 2,800

ω (cm–1)

Figure 2 Characterization of reduced GO thin films using Raman spectroscopy. a,b, Raman maps (15 mm  12 mm) for 20 ml (a) and 80 ml (b) reduced GO thin
films, in which the different colours indicate different numbers of graphene layers, as shown in the key. c, Raman spectra near the 2D peak for a single-layer region
(N ¼ 1) and for a multiple-layer region (N . 5). d,e, Optical micrographs of 20 ml (d) and 80 ml (e) films showing the different densities of the overlapped regions
(darker colours) between the graphene sheets (scale bar ¼ 20 mm).

details and AFM images). It should be noted that pure single-layer graphene31. Careful analysis of the spectra allowed us to monitor
graphene flake has a thickness of 0.34 nm, corresponding to the the 2D peak shifts and thus map the number of reduced GO
interlayer spacing of graphite, but a GO sheet is 1 nm thick layers in the thin films. The Raman maps over 15 mm  12 mm
due to the presence of functional groups, structural defects and spatial regions for the 20 ml and 80 ml reduced GO thin films
adsorbed water molecules25,30. are shown in Fig. 2a and b. The 2D peak shift between one and
We also investigated the number of reduced GO layers using more than five layers of graphene was found to be approximately
Raman spectroscopy by monitoring the second-order zone 40 cm21 wavenumbers, as indicated by the actual measured
boundary phonons peak at 2,700 cm21, referred to as the G0 or peaks shown in Fig. 2c.
2D peak31 (see Methods). Raman spectra of reduced GO thin The Raman maps are consistent with the AFM data in that the
films typically show the presence of the usual D, G and 2D percolating regions consist of 1–2 and 3–5 layers in both the 20
peaks31 (see Supplementary Information). The prominent D peak and 80 ml thin films. However, it can be seen from Fig. 2a that
(absent in mechanically cleaved graphene) clearly indicates the 1–2 layers are more predominant in the 20 ml film compared
presence of structural imperfections induced by the attachment to the 80 ml film, where 3–5 layers are more readily visible.
of hyrodxyl and epoxide groups on the carbon basal plane. The In addition, thicker regions (.5 layers) are also visible in the
intensity of the 2D peak with respect to the D and G peaks is Raman maps, which likely arise from incomplete exfoliation of
small due to disorder, and thus requires care during acquisition GO in suspension24. The optical images of the 20 ml and 80 ml
and analysis. Nevertheless, the shift in this peak can be used as a films are shown in Fig. 2d and e to indicate the degree of
simple non-destructive tool for analysing the number of layers in overlapping (darker regions) among the GO layers. The darker

nature nanotechnology | VOL 3 | MAY 2008 | www.nature.com/naturenanotechnology 271

© 2008 Nature Publishing Group
LETTERS
regions are clearly less visible in the 20 ml film. Our thickness results
suggest that the thin films are uniform in that they contain 1–5 layers Reduced
of reduced GO. However, the slight variation in thicknesses in 10 Reduced and annealed
10
Raman maps of both films point to the fact that a better control

)
–1
of the size and shape of the suspended GO sheets will be essential

Sheet resistance (Ω
if films of exactly a single monolayer are to be deposited.
After deposition, the insulating GO must be reduced to 108
graphene through exposure to hydrazine vapour and/or
annealing in inert conditions19–21,23–25,30,32 to render the material
electrically conductive. We found that the hydrazine vapour alone
is not sufficient to achieve maximum reduction, and annealing 106
alone requires relatively high temperatures (.550 8C)22. Efficient
reduction of the GO thin films was therefore achieved through a
combination of hydrazine vapour exposure and low-temperature
0 100 200 300 400 500
annealing treatment (see Supplementary Information for X-ray
Effective filtration volume (ml)
photoelectron spectroscopy (XPS) results)30,32. The reduction
of GO yields thin films with properties resembling those of
graphene. More interestingly, by controlling the amount of
100
reduced GO on the surface, it is possible to tune the
optoelectronic properties of the thin films as summarized in
Fig. 3a,b. It can be seen from Fig. 3a that the sheet resistance of
the hydrazine-treated GO thin films is independent of the 90

Transmittance (%)
filtration volume except at very high values (.300 ml). However,
annealing at 200 8C in nitrogen (or vacuum) leads to a dramatic
reduction in the sheet resistance (1  105 V A21). The lowest 80
sheet resistance value we obtained was 43 kV A21. The
saturation of sheet resistance in Fig. 3a above a critical filtration
volume is probably due to the fact that reduction is only effective 70 As-deposited
for the uppermost layers. The corresponding transmittances as a Reduced
function of the filtration volume at l ¼ 550 nm for the as- Reduced and annealed
deposited GO, chemically reduced GO and chemically reduced
0 100 200 300 400 500
and annealed GO are shown in Fig. 3b (see Supplementary
Effective filtration volume (ml)
Information for transmission versus wavelength plots). It can be
seen that the chemically reduced and annealed GO leads to a
decrease in the transparency of thin films that is lower than that Figure 3 Electrical and optical properties of reduced GO thin films. a,b, Sheet
for reduced and non-annealed GO, also consistent with the resistance (a) and transmittance (b) at l ¼ 550 nm as a function of filtration
increase in the Drude background obtained from spectroscopic volume for reduced GO thin films. Plots are shown for thin films with different
ellipsometry (see Supplementary Information). reduction steps. The sheet resistance for as-deposited GO thin films is out of scale.
In order to translate the opto-electronic properties into devices,
we fabricated thin-film transistors (TFTs) with reduced GO thin
films. Of the numerous (.100) TFT devices we tested, all
showed uniform transfer characteristics regardless of the channel characteristics (see Methods and Supplementary Information)
length (21 mm or 210 mm, SiO2 thickness ¼ 300 nm, channel was found to be 1 cm2 V21 s21 for holes and was lower at low
width ¼ 400 mm). The relatively long channel lengths ensured temperatures. The electron mobilities ( 0.2 cm2 V21 s21) were
that the transport was bulk limited and the role of contacts was generally lower than the hole mobilities at ambient conditions,
not substantial in our devices. The transfer characteristics as a and the reverse was true in vacuum. The lower overall mobilities
function of temperature for the 20 and 80 ml reduced GO thin in our devices compared to those achieved in individual reduced
films are shown in Fig. 4a,b, respectively, together with a GO flakes (2– 200 cm2 V21 s21) could be attributed to scattering
photograph of the devices (Fig. 4c). The low-temperature at the junctions formed by overlapping flakes. The electron and
measurements exhibit ambipolar characteristics, comparable to hole currents from TFTs and conductivity versus temperature for
graphene. This is remarkable, because transport between the the 20 ml thin films are shown in Fig. 4d,e, respectively. The
source and drain electrodes in our large-scale devices occurs over temperature dependence is unusual but consistent with the
several graphene sheets. The p-type oxygen doping effect11, which anomalous behaviour found in exfoliated graphite33, in which
increases the current and shifts the threshold voltage to positive conduction is metallic-like below 50 K and then crosses over to
voltages, is dramatically reduced for both the 20 ml and 80 ml activated transport above this temperature.
devices when measurements are performed in vacuum. The A method is reported for uniform and controllable deposition
primary differences between the two devices are that the current of 1–5 nm graphene thin films from solution at room temperature
in the 80 ml TFTs is higher and the 20 ml device exhibits a on a variety of substrates, from the reduction of GO. The vacuum
sharper turn on behaviour, which is consistent with conduction filtration method allows the deposition of very thin films of 1 –2
occurring primarily through one or two layers of graphene. In layers of reduced GO that are semiconducting, and thicker films
addition, the ‘V’ shape of the ambipolar graphene transfer that are semi-metallic. We have demonstrated that the sheet
characteristics is more pronounced for the low-temperature resistance of the thin films can be tuned over six orders of
measurements of the 20 ml TFTs, suggesting the semiconducting magnitude and the transparencies from 60 to 95%. The deposition
nature of the material. The mobility of the devices in ambient of uniform thin films allows the simple fabrication of TFTs on
conditions calculated from the linear regime of the transfer various substrates without the use of extensive lithography. Our

272 nature nanotechnology | VOL 3 | MAY 2008 | www.nature.com/naturenanotechnology

© 2008 Nature Publishing Group
 LETTERS
8
4.2 K
100 30 K
6 100 K
150 K
200 K
250 K
292 K

I sd (μA)
4

I sd (nA)
299 K
292 K-air

10

Vsd = 10 V 2 Vsd = 1 V

–100 –50 0 50 100 –100 –50 0 50 100

Vg (V) Vg (V)

120 Electron (Vg = 100 V)

10–1
100 Hole (Vg = –100 V)
80

Conductivity (S m–1)
60
I sd (nA)

40

S D

Vsd = 10 V Vg = 0 V
10–2
20

0 50 100 150 200 250 300 0 50 100 150 200 250 300
Temperature (K) Temperature (K)

Figure 4 TFT devices based on reduced GO thin films. a,b, Source – drain current as a function of gate voltage for 20 ml (a) and 80 ml (b) films measured at
different temperatures. Measurements were conducted in vacuum unless otherwise noted. c, An optical micrograph of the actual device. The channel consists of
several reduced GO sheets percolating across the source (S) and drain (D) electrodes. The channel length is 21 mm. d, Electron and hole currents at maximum and
minimum gate voltages as a function of temperature for the 20 ml film. e, Film conductivity as a function of temperature (film thickness of 2 nm is assumed)
exhibiting anomalous behaviour.

results could provide a pathway for the translation of graphene-based the peak position of the second-order Raman D-peak, f(2D), considerably
research from the fundamental to the technological realm. blueshifts with increasing number of layers in graphene. Therefore, we were able
to reconstruct the number of layers in our samples by mapping f(2D) in our
films. The 2D peak positions we assumed for one, two and a few layers of
METHODS graphene in order to reconstruct the maps were taken from other work31 (see
Supplementary Information, Table S1). Such values of f(2D) allowed us to obtain
Slurry of GO was obtained from SP-1 graphite (Bay Carbon) using the modified the maps reported in Fig. 2a and b, which were reconstructed from the raw data
Hummers method27. The concentration of the slurry was determined by drying GO using the Matlab 6.5 Contour routine. Comparison of spectra recorded from
over phosphorus pentoxide in a vacuum desiccator for a week and was found to be thick (.5 layers) and thin (1 layer) regions of the 80 ml films are reported in
49 mg ml21. The slurry was diluted in water and sonicated to achieve exfoliated Fig. 2c. In addition to the 20 and 80 ml samples, we also measured the number of
sheets of GO. The effective filtration volume mentioned in the manuscript indicates layers for a 50 ml sample (see Supplementary Information).
volumes of 0.33 mg l21 suspension required to achieve the mass deposited by Bottom-gated TFTs were fabricated by depositing and reducing GO on p-Si
suspension of other concentrations. For example, films obtained by filtrating 60 ml substrates with 300 nm thermal oxide. p-Si was used as the gate electrode. We
of 0.66 mg l21 suspension are indicated as a 120 ml film. A dilute GO suspension arbitrarily deposited gold source and drain electrodes with channel lengths of
(0.33–2.64 mg l21) was vacuum-filtrated using a mixed cellulose ester membrane 21 mm and 210 mm to investigate the uniformity of the thin films. The mobility was
with 25 nm pores (Millipore). The membranes with captured GO were then cut calculated using m ¼ (L/WCoxVsd)(DIsd/DVg), where L and W are the channel
into sizes of choice, wetted with deionized water, and pressed against the substrate length and width, Cox the gate oxide capacitance, Vsd the source–drain voltage,
surface with the GO side in contact with the substrate. The GO film was allowed to Isd the source–drain current and Vg the gate voltage. The linear regime of the
dry and adhere to the substrate at room temperature under a 1 kg weight for at least transfer characteristics was used to obtain DIsd/DVg (see Supplementary
10 h. The weight was removed and the membrane was dissolved using acetone Information for details).
(successively pure baths) to leave a GO thin film on the substrate. The films were
then rinsed with a methanol wash and dried by blowing with dry nitrogen. Similar
to the SWNT thin films, the adhesion of the GO thin films was found to be Received 4 February 2008; accepted 5 March 2008; published 6 April 2008.
sufficiently strong to prevent delamination during processing.
Unpolarized Raman spectra were recorded at room temperature on reduced
GO thin films deposited on 300-nm-thick SiO2. A Renishaw InVia Raman References
1. Ahn, J. H. et al. Heterogeneous three-dimensional electronics by use of printed semiconductor
microscope equipped with a 100 objective was used. The excitation source was nanomaterials. Science 314, 1754–1757 (2006).
the 633 nm line from a He– Ne laser and the spot size was estimated to be 2. Wu, Z. C. et al. Transparent, conductive carbon nanotube films. Science 305, 1273 –1276 (2004).
less than 2 mm. Low power and special attention was used to avoid laser heating 3. Snow, E. S., Perkins, F. K., Houser, E. J., Badescu, S. C. & Reinecke, T. L. Chemical detection with a
single-walled carbon nanotube capacitor. Science 307, 1942– 1945 (2005).
of the specimen. Sample areas up to 15  12 mm were mapped using a precision 4. Artukovic, E., Kaempgen, M., Hecht, D. S., Roth, S. & Gruner, G. Transparent and flexible carbon
xyz stage and the Renishaw 2.0 mapping software. According to recent studies31, nanotube transistors. Nano Lett. 5, 757– 760 (2005).

nature nanotechnology | VOL 3 | MAY 2008 | www.nature.com/naturenanotechnology 273

© 2008 Nature Publishing Group
LETTERS
5. Kang, S. J. et al. High-performance electronics using dense, perfectly aligned arrays of single-walled 26. Stankovich, S. et al. Stable aqueous dispersions of graphitic nanoplatelets via the reduction of
carbon nanotubes. Nature Nanotech. 2, 230 –236 (2007). exfoliated graphite oxide in the presence of poly(sodium 4-styrenesulfonate). J. Mater. Chem.
6. Fan, Z. et al. Wafer-scale assembly of highly ordered semiconductor nanowire arrays by contact 16, 155 –158 (2006).
printing. Nano Lett. 8, 20 –25 (2008). 27. Hirata, M., Gotou, T., Horiuchi, S., Fujiwara, M. & Ohba, M. Thin-film particles of graphite oxide 1:
7. Liang, X., Fu, Z. & Chou, S. Y. Graphene transistors fabricated via transfer-printing in device-active High-yield synthesis and flexibility of the particles. Carbon 42, 2929– 2937 (2004).
areas on large wafer. Nano Lett. 7, 3840– 3844 (2007). 28. Hu, L., Hecht, D. S. & Gruner, G. Percolation in transparent and conducting carbon nanotube
8. Chen, J. H. et al. Printed graphene circuits. Adv. Mater. 19, 3623– 3627 (2007). networks. Nano Lett. 4, 2513–2517 (2004).
9. Meitl, M. A. et al. Transfer printing by kinetic control of adhesion to an elastomeric stamp. 29. Unalan, H. E., Fanchini, G., Kanwal, A., Du Pasquier, A. & Chhowalla, M. Design criteria for
Nature Mater. 5, 33 – 38 (2006). transparent single-wall carbon nanotube thin-film transistors. Nano Lett. 6, 677– 682 (2006).
10. Li, D., Muller, M. B., Gilje, S., Kaner, R. B. & Wallace, G. G. Processable aqueous dispersions of 30. Stankovich, S. et al. Synthesis of graphene-based nanosheets via chemical reduction of exfoliated
graphene nanosheets. Nature Nanotech. 3, 101 –105 (2008). graphite oxide. Carbon 45, 1558–1565 (2007).
11. Novoselov, K. S. et al. Electric field effect in atomically thin carbon films. Science 306, 31. Ferrari, A. C. et al. Raman spectrum of graphene and graphene layers. Phys. Rev. Lett. 97,
666– 669 (2004). 187401 (2006).
12. Geim, A. K. & Novoselov, K. S. The rise of graphene. Nature Mater. 6, 183 –191 (2007). 32. Watcharotone, S. et al. Graphene– silica composite thin films as transparent conductors. Nano Lett.
13. Schedin, F. et al. Detection of individual gas molecules adsorbed on graphene. Nature Mater. 7, 1888 –1892 (2007).
6, 652– 655 (2007). 33. Uher, C. & Sander, L. M. Unusual temperature dependence of the resistivity of exfoliated graphites.
14. Han, M. Y., Ozyilmaz, B., Zhang, Y. B. & Kim, P. Energy band-gap engineering of graphene Phys. Rev. B 27, 1326 –1332 (1983).
nanoribbons. Phys. Rev. Lett. 98, 206805 (2007).
15. Barbolina, I.I. et al. Submicron sensors of local electric field with single-electron resolution at room
temperature. Appl. Phys. Lett. 88, 013901 (2006). Supplementary information accompanies this paper at www.nature.com/naturenanotechnology.
16. Tombros, N., Jozsa, C., Popinciuc, M., Jonkman, H. T. & van Wees, B. J. Electronic spin transport and
spin precession in single graphene layers at room temperature. Nature 448, 571 –574 (2007). Acknowledgements
17. Berger, C. et al. Ultrathin epitaxial graphite: 2D electron gas properties and a route toward graphene-
The authors would like to acknowledge Yun-Yue Lin and S. Miller for fabricating organic photovoltaic
based nanoelectronics. J. Phys. Chem. B 108, 19912– 19916 (2004).
devices. A detailed study on the photovoltaic devices will be reported elsewhere. We also acknowledge
18. Coraux, J., N’Diaye, A. T., Busse, C. & Michely, T. Structural coherency of graphene on Ir(111).
A. Kanwal for help with the low-temperature measurements, and thank O. Celik for help with XPS
Nano Lett. 8, 565 –570 (2008).
analysis and A. Mann for allowing us the use of the Raman instrument. This work was funded by the
19. Stankovich, S. et al. Graphene-based composite materials. Nature 442, 282– 286 (2006).
National Science Foundation CAREER Award (ECS 0543867).
20. Gomez-Navarro, C. et al. Electronic transport properties of individual chemically reduced graphene
oxide sheets. Nano Lett. 7, 3499–3503 (2007).
21. Gijie, S., Han, S., Wang, M., Wang, K. L. & Kaner, R. B. A chemical route to graphene for device Author contributions
applications. Nano Lett. 7, 3394– 3398 (2007). G.E. and M.C. conceived and designed the experiments. G.E. carried out the exfoliation, dispersion and
22. Wang, X., Zhi, L. & Mullen, K. Transparent, conductive graphene electrodes for dye-sensitized solar deposition processes in addition to conducting the opto-electronic and device measurements. G.F.
cells. Nano Lett. 8, 323– 327 (2007). performed the Raman analysis. M.C. wrote the paper with input from G.E. All the authors discussed the
23. Jung, I. et al. Simple approach for high-contrast optical imaging and characterization of graphene- results and commented on the manuscript throughout the review process.
based sheets. Nano Lett. 7, 3569 –3575 (2007).
24. Schniepp, H. C. et al. Functionalized single graphene sheets derived from splitting graphite oxide.
J. Phys. Chem. B 110, 8535– 8539 (2006). Author information
25. McAllister, M. J. et al. Single sheet functionalized graphene by oxidation and thermal expansion of Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/.
graphite. Chem. Mater. 19, 4396 –4404 (2007). Correspondence and requests for materials should be addressed to M.C.

274 nature nanotechnology | VOL 3 | MAY 2008 | www.nature.com/naturenanotechnology

© 2008 Nature Publishing Group
 ARTICLES

Processable aqueous dispersions of

graphene nanosheets
DAN LI1*, MARC B. MÜLLER1, SCOTT GILJE2, RICHARD B. KANER2 AND GORDON G. WALLACE1*
1
ARC Centre of Excellence for Electromaterials Science, Intelligent Polymer Research Institute, University of Wollongong, NSW 2522, Australia
2
Department of Chemistry and Biochemistry, Department of Materials Science and Engineering, and California NanoSystems Institute, University of
California, Los Angeles, California 90095-1569, USA
*e-mail: danli@uow.edu.au; gwallace@uow.edu.au

Published online: 27 January 2008; doi:10.1038/nnano.2007.451

Graphene sheets offer extraordinary electronic, thermal and mechanical properties and are expected to find a variety of applications.
A prerequisite for exploiting most proposed applications for graphene is the availability of processable graphene sheets in large
quantities. The direct dispersion of hydrophobic graphite or graphene sheets in water without the assistance of dispersing agents
has generally been considered to be an insurmountable challenge. Here we report that chemically converted graphene sheets
obtained from graphite can readily form stable aqueous colloids through electrostatic stabilization. This discovery has enabled us
to develop a facile approach to large-scale production of aqueous graphene dispersions without the need for polymeric or
surfactant stabilizers. Our findings make it possible to process graphene materials using low-cost solution processing techniques,
opening up enormous opportunities to use this unique carbon nanostructure for many technological applications.

Graphene, a new class of two-dimensional carbon nanostructure, graphite appears to be a much more efficient approach to bulk
has attracted tremendous attention from both the experimental production of graphene sheets at low cost2–5.
and theoretical scientific communities in recent years1. This As recently demonstrated by Ruoff and co-workers2,4, the
unique nanostructure holds great promise for potential solution-based route involves chemical oxidation of graphite to
applications in many technological fields such as nanoelectronics, hydrophilic graphite oxide, which can be readily exfoliated as
sensors, nanocomposites, batteries, supercapacitors and hydrogen individual graphene oxide (GO) sheets by ultrasonication in
storage1. However, a lack of an efficient approach to producing water (Fig. 1). Graphene oxide, which is electrically insulating,
processable graphene sheets in large quantities has been a major can be converted back to conducting graphene by chemical
obstacle to exploiting most proposed applications. reduction, for example, using hydrazine. Unfortunately, previous
Like carbon nanotubes and many other nanomaterials, a key work2,4 has shown that, unless stabilized by selected polymers,
challenge in the synthesis and processing of bulk-quantity chemically converted graphene (CCG) sheets obtained through
graphene sheets is aggregation. Graphene sheets, which have a this method precipitate as irreversible agglomerates owing to
high specific surface area, unless well separated from each other, their hydrophobic nature. The resulting graphene agglomerates
tend to form irreversible agglomerates or even restack to form appear to be insoluble in water and organic solvents2, making
graphite through van der Waals interactions. This problem has further processing difficult.
been encountered in all previous efforts aimed at large-scale It is well known that exfoliated graphite oxide (or GO) can
production of graphene through chemical conversion or thermal form well-dispersed aqueous colloids8–12. Our study on the
expansion/reduction2–5. The prevention of aggregation is of surface charge (zeta potential) of as-prepared GO sheets shows
particular importance for graphene sheets because most of their that these sheets are highly negatively charged when dispersed in
unique properties are only associated with individual sheets. water (Fig. 2a), apparently as a result of ionization of the
Aggregation can be reduced by the attachment of other molecules carboxylic acid and phenolic hydroxyl groups that are known to
or polymers onto the sheets4–6. However, the presence of foreign exist on the GO sheets13,14. This result suggests that the
stabilizers is undesirable for most applications. New strategies to formation of stable GO colloids should be attributed to
produce relatively clean graphene sheets in bulk quantity while electrostatic repulsion, rather than just the hydrophilicity of GO
keeping them individually separated are required. as previously presumed2. Given that carboxylic acid groups
Graphite, consisting of a stack of flat graphene sheets, is are unlikely to be reduced by hydrazine under the given
inexpensive and available in large quantities from both natural reaction conditions2, these groups should therefore remain in
and synthetic sources. This ordinary carbon material is likely the the reduced product as confirmed by our FT-IR analysis
most readily available and least expensive source for the (Fig. 2b). The presence of carboxylic acid groups suggests
production of bulk graphene sheets. Mechanical cleavage of that the surface of the graphene sheets in aqueous solution
graphite originally led to the discovery of graphene sheets7 and is should still be charged after reduction. We surmised that
the process currently used in most experimental studies of the electrostatic repulsion mechanism that makes GO colloids
graphene1. However, the low productivity of this method makes stable could also enable the formation of well-dispersed
it unsuitable for large-scale use. Chemical conversion from graphene colloids.

nature nanotechnology | VOL 3 | FEBRUARY 2008 | www.nature.com/naturenanotechnology 101

© 2008 Nature Publishing Group
ARTICLES

(1) (2) (3)

Figure 1 Scheme showing the chemical route to the synthesis of aqueous graphene dispersions. 1, Oxidation of graphite (black blocks) to graphite oxide
(lighter coloured blocks) with greater interlayer distance. 2, Exfoliation of graphite oxide in water by sonication to obtain GO colloids that are stabilized by electrostatic
repulsion. 3, Controlled conversion of GO colloids to conducting graphene colloids through deoxygenation by hydrazine reduction.

10 100 We note that if GO dispersions with concentrations less than

0 GO 0.5 mg ml21 are reduced by hydrazine under appropriate
Zeta potential (mV)

Transmittance (%)

–10 conditions (see Methods), the particle size of the resulting CCG
90
–20 CCG sheets does not increase after the reduction is complete (Fig. 3a).
–30 No sediment is observed even after the dispersion has been
–40 80
centrifuged at 4,000 r.p.m. for several hours. Atomic force
GO
–50 CCG microscopy (AFM) shows that the resulting CCG sheets that are
1 2 3 4 5 6 7 8 9 10 11 12 4,000 3,000 2,000 1,000
cast on a silicon wafer are flat, with a thickness of 1 nm
pH Wavenumber (cm–1) (Fig. 3b). These results indicate that, similar to the original
GO dispersion, the as-prepared CCG sheets remain separated
in the dispersion.
Figure 2 Surface properties of GO and CCG. a, Zeta potential of GO and CCG The colloidal nature of the resulting CCG dispersions is further
as a function of pH, in aqueous dispersions at a concentration of confirmed by two experiments typically conducted in colloid
0.05 mg ml21. b, FT-IR spectra of GO and CCG. The absorption band at around science: investigations of the Tyndall effect and the salt effect. A
1,700 cm21 is attributed to carboxyl groups. The absorption of CCG sheets at this diluted CCG dispersion gives rise to the Tyndall effect, in which
range is observable but not as prominent as that observed for GO, likely due to the a laser beam passing through a colloidal solution leaves a
overlapping of the strong absorption of graphene sheets in this region. discernible track as a result of light scattering (Fig. 3c). Adding
an electrolyte solution such as sodium chloride into a CCG
dispersion leads to immediate coagulation (Fig. 3d). These
As demonstrated in many colloid experiments15, including our observations are characteristic of a lyophobic colloid stabilized
previous work on the synthesis of stabilizer-free conducting through electrostatic repulsion and can be explained using the
polymer aqueous colloids16, the colloidal stability of an classical Derjaguin – Landau– Verwey–Overbeek theory15.
electrostatically stabilized dispersion is strongly dependent on pH, It is worth pointing out that in water, ammonia and hydrazine
the electrolyte concentration, and the content of dispersed dissociate to generate ionic species that act as electrolytes.
particles. By controlling these parameters, we now find that Therefore, as with the addition of sodium chloride (Fig. 3d), the
chemically converted graphene sheets are indeed able to form overuse of the two chemicals can result in destabilization of the
stable colloids through electrostatic stabilization. Graphene oxide resulting dispersions. We note that if the amount of hydrazine
dispersions can be directly converted to stable graphene colloids exceeds the optimal level described in the Methods section
through hydrazine reduction under controlled conditions (Fig. 1). (hydrazine:GO ¼ 7:10 by weight), the stability of the dispersion
The use of polymeric or surfactant stabilizers is not required. We decreases with increasing concentration of hydrazine. For
have found that the complete removal of metal salts and acids, example, if the weight ratio of hydrazine to GO is increased to
which often remain in the starting graphite oxide, is critical to 7:1, agglomeration occurs in about one day. In order to obtain
stability. These residual electrolytes can neutralize the charges on stable dispersions in this case, excess hydrazine must be
the sheets, destabilizing the resulting dispersions. immediately removed from the resulting dispersions.
In order to obtain maximal charge density on the resulting Additionally, like other lyophobic colloids, the colloidal stability
graphene sheets, ammonia is added to the reaction solution to of the resulting dispersions is also dependent on the
increase the pH to around 10. Volatile ammonia can be easily concentration of CCG. The reduction of GO dispersions with a
removed after the graphene sheets are processed into solid films concentration greater than 0.5 mg ml21 leads to gelation over time.
or composites. The use of excess hydrazine also renders the The feasibility of forming stable graphene dispersions through
dispersion basic. However, hydrazine is highly toxic and its use electrostatic stabilization is further supported by our zeta potential
should be minimized. Additionally, we find that during the analysis. As shown in Fig. 2a, the zeta potential of the reduced
reduction process, the graphene sheets at the water/air interface graphene dispersion is pH dependent, which is consistent with
tend to agglomerate upon water evaporation and a layer of black the fact that the ionization of carboxylic acid groups is strongly
solid gradually appears on the liquid surface. This problem can related to pH. Although the magnitude of the zeta potential is
be effectively avoided by adding a layer of water-immiscible lower than that of the original GO sheets at the same pH, the
liquid (such as mineral oil) to the solution to eliminate the zeta potential is below –30 mV when the pH is greater than 6.1
air/water interface. and can reach –43 mV when the pH approaches 10. Zeta

102 nature nanotechnology | VOL 3 | FEBRUARY 2008 | www.nature.com/naturenanotechnology

© 2008 Nature Publishing Group
 ARTICLES
2.0 60 min
15 min
1.2 10 min
8 min
6 min

Absorbance
1.5 4 min
0.8 2 min
Average particle size (mm)

0 min

0.4
1.0

GO
0.0
Reduced with N2H4 but without NH3 200 400 600 800 1,000
1.5 Reduced with N2H4 and NH3 Wavelength (nm)

Figure 4 UV-vis absorption spectra showing the change of GO dispersions

0.0 as a function of reaction time. The absorption peak of the GO dispersion at
0 5 10 15 20 231 nm gradually redshifts to 270 nm and the absorption in the whole
Time (h) spectral region (.231 nm) increases with reaction time, suggesting that
the electronic conjugation within the graphene sheets is restored upon
hydrazine reduction.

potential values more negative than – 30 mV are generally

considered to represent sufficient mutual repulsion to ensure the
stability of a dispersion, as is well known from colloidal science15.
The formation of stable graphene dispersions enables the
reaction process to be monitored using UV-vis spectroscopy. As
shown in Fig. 4, the absorption peak of the GO dispersion at
231 nm gradually redshifts to 270 nm and the absorption in the
whole spectral region (.231 nm) increases with reaction time,
suggesting that the electronic conjugation within the graphene
1 nm
sheets is restored upon hydrazine reduction. Little increase in
absorption is found after 1 h, indicating completion of the
reduction within that period. This experiment also suggests that,
200 nm like conjugated polymers, the electronic conjugation level of
graphene is chemically controllable, offering possibilities to tailor
the optical and electrical properties of graphene sheets. We find
that graphene sheets with different reduction levels can all form
stable dispersions if they are prepared using the procedure
described in the Methods section.
As shown with carbon nanotubes17–20, the dispersion of
nanomaterials in solution is crucial to advancing many
technological applications. Owing to their hydrophobic nature,
NaCl the direct dispersion of graphite or graphene sheets in water has
been generally considered unattainable. This work, for the first
time, suggests that ordinary natural graphite, when treated
appropriately by chemical means, can readily disperse in water to
generate stable graphene colloids without the need for any
polymeric or surfactant stabilizers. Of great significance is that
the successful formation of relatively pure graphene dispersions
enables the use of conventional low-cost solution-phase
processing techniques to create new graphene-based materials
Figure 3 Colloidal and morphological characterization of CCG dispersions. and devices. For example, we have observed that a single layer of
a, Effect of the addition of ammonia on the dispersion state of CCG sheets, CCG sheets can be deposited on a substrate by simply drop-
characterized by measuring average particle sizes over a long period of time. casting from a dilute CCG dispersion (Fig. 3b), which provides a
The photographs shown in the inset were taken two days after the reduction facile approach to obtain single graphene sheets for device
reaction was complete with (left) and without (right) the addition of ammonia. fabrication or studies on the properties of individual sheets.
The concentration of the starting GO solution is 0.5 mg ml21. b, Tapping- Recent work has demonstrated that the graphene sheets obtained
mode AFM image of CCG sheets with a height profile (blue curve; scale bar, by hydrazine reduction of GO sheets appear to act as a p-type
1 nm) taken along the red line. The sample was prepared by drop-casting a semiconductor21 and the electrical conductivity exhibits a field-
dilute CCG dispersion onto a silicon wafer. c,d, The Tyndall effect and the salt effect response21,22, suggesting that our processable CCG sheets
effect confirming the colloidal nature of the CCG dispersions. The salt effect could be an exciting material for use in future nanoelectronics.
experiment highlights the importance of complete removal of residual salts and Uniform graphene films can also be readily formed on a
acids from the starting GO solution. membrane filter by vacuum filtration of as-reduced dispersions.

nature nanotechnology | VOL 3 | FEBRUARY 2008 | www.nature.com/naturenanotechnology 103

© 2008 Nature Publishing Group
ARTICLES
many fields, including membranes, anisotropic conductors and
supercapacitors24. Our preliminary measurements show that the
graphene paper obtained from direct filtration of our stable CCG
dispersions gives a tensile modulus up to 35 GPa, which is close
to that of the GO paper. We expect that our strong, conductive,
flexible and thermally stable graphene paper should be more
attractive than non-conductive, less thermally stable GO paper
for practical applications.
Spraying techniques such as air-brushing can also be used to
produce conductive graphene coatings on various substrates. Like
many other lyophobic colloids, once the graphene colloids are
dried, they are not redispersible in water, rendering as-prepared
graphene coatings water-resistant. Of particular significance is
that, owing to the high aspect ratio of the graphene sheets, a very
100
thin graphene coating, which is almost transparent, can result in
the formation of a continuous conducting network. Figure 5b
90
shows a transmittance spectrum of a sprayed CCG coating on a
Transmittance (%)

glass slide. The coating gives a sheet resistivity of 2.0  107V A21
80
at room temperature, and the transmittance in the visible
wavelength range is higher than 96%. The conductivity of this
70 as-sprayed coating is sufficient for antistatic applications.
Antistatic coatings are vital to the safety of materials, machinery
60 and individuals across many different industries25. Our work may
lead to the development of a new generation of antistatic
50 coatings that can combine electrical conductivity with
400 600 800 1,000
Wavelength (nm)
transparency, excellent thermal and chemical stability, water
resistance and low production cost. Graphene dispersions may
thus find immediate practical uses.
0.15 14 Additionally, the highly charged state of the CCG sheets in
12 water makes it possible to use the well-known layer-by-layer
electrostatic assembly technique26–29 to build up complex and
10
Absorbance

0.10 controllable graphene-based nanosystems with other functional

8 molecules, polymers and nanostructures. We have demonstrated
6 the feasibility of this process by alternately immersing a quartz
0.05 4 slide in a dilute CCG dispersion and a typical cation
2 polyelectrolyte—poly(diallyldimethylammonium chloride). As
confirmed by the absorption spectra (Fig. 5c), CCG sheets can be
0.00 successfully assembled using this simple approach. Note that thin
200 400 600 800 films of graphite oxide sheets have been previously prepared
Wavelength (nm)
using this technique8–12. However, to make the resulting graphite
oxide film conducting, an additional reduction step is needed.
This reduction process is likely to be detrimental to composites
Figure 5 Examples demonstrating that films made of CCG sheets can be
containing more delicate molecular structures such as
easily fabricated from CCG dispersions using various solution-phase
biomolecules or conjugated polymers. It has been extensively
processing techniques. a, A 10-mm-thick CCG film or paper prepared by
demonstrated that self-assembled multilayered electroactive films
vacuum filtration of a CCG dispersion through an alumina membrane. The film
hold great potential in many applications such as sensors and
exhibits a shiny metallic lustre. A CCG strip (top-right inset) cut from the film is
neuroprosthetic devices26–29. It would be reasonable to expect
bent to demonstrate its flexibility. b, A transmission spectrum of a CCG coating
that the successful formation of graphene colloids will open up
deposited on a glass slide by air-brush spraying of a CCG solution. The
possibilities to use this powerful electrostatic assembly technique
transmittance in the visible light range is greater than 96%. c, UV-vis spectra of
to manipulate graphene sheets for creating many new and
polycation/CCG films prepared by a layer-by-layer electrostatic self-assembly
potentially useful nanosystems.
technique. The absorbance increases linearly with an increase in the number of
In summary, we have demonstrated that aqueous graphene
assembly cycles (denoted above each curve), indicative of the successful
dispersions can be readily formed by controlled chemical
assembly of CCG sheets on the substrate.
conversion of GO colloids without the need for either polymeric
or surfactant stabilizers. Chemically converted graphene can now
be viewed as a special water-soluble conducting macromolecule
Free-standing films or graphene paper can be peeled off the that can be simply obtained from graphite. Graphene sheets
membrane. The films are bendable and exhibit a shiny metallic should be superior to normal synthetic conducting polymers in
lustre (Fig. 5a). The conductivity of graphene paper prepared terms of thermal and chemical stability and mechanical strength,
using the procedure described in the Methods section is found to and more competitive than carbon nanotubes in terms of
be 7,200 S m21 at room temperature, which is comparable to production cost. Furthermore, as shown with carbon
that of chemically modified single-walled carbon nanotube nanotubes18–20, the successful dissolution of graphene sheets in
paper23. We note that Ruoff and co-workers have recently solution as well as the residual carboxylic groups on the sheets
demonstrated that strong GO paper can be prepared using a enables the use of solution-phase chemistry to further modify
similar filtration strategy24. The resulting paper could find use in graphene sheets for achieving new functionalities.

104 nature nanotechnology | VOL 3 | FEBRUARY 2008 | www.nature.com/naturenanotechnology

© 2008 Nature Publishing Group
 ARTICLES
The dispersant-free feature offers a great deal of flexibility in the gripped using film tension clamps with a clamp compliance of about
creation of novel graphene-based nanocomposites with many other 0.2 mm N21. The tensile tests were performed in the controlled strain rate mode
molecules and nanostructures. The ease of synthesis and the with a preload of 0.001 N and a strain ramp rate of 0.01% min21.
exceptional solution-phase processability of CCG sheets make
this inexpensive and conductive nanostructure attractive not only Received 27 September 2007; accepted 18 December 2007;
for future nanoelectronics, but also for large-scale applications in published 27 January 2008.
both conventional technological fields, such as transparent
antistatic coatings and electrochemical devices, and emerging
areas such as flexible/transparent electronics, high-performance References
nanocomposites, nanomedicines and bionic materials. We believe 1. Geim, A. K. & Novoselov, K. S. The rise of graphene. Nature Mater. 6, 183 –191 (2007).
2. Stankovich, S. et al. Synthesis of graphene-based nanosheets via chemical reduction of exfoliated
that the work presented here will provide a significant step graphite oxide. Carbon 45, 1558– 1565 (2007).
forward to bringing graphene materials much closer to real- 3. McAllister, M. J. et al. Single sheet functionalized graphene by oxidation and thermal expansion of
graphite. Chem. Mater. 19, 4396– 4404 (2007).
world applications. 4. Stankovich, S. et al. Stable aqueous dispersions of graphitic nanoplatelets via the reduction of
exfoliated graphite oxide in the presence of poly(sodium 4-styrenesulfonate). J. Mater. Chem. 16,
METHODS 155– 158 (2006).
5. Niyogi, S. et al. Solution properties of graphite and graphene. J. Am. Chem. Soc. 128,
7720–7721 (2006).
SYNTHESIS 6. Stankovich, S. et al. Graphene-based composite materials. Nature 442, 282 –286 (2006).
Graphite oxide was synthesized from natural graphite (SP-1, Bay Carbon) by a 7. Novoselov, K. S. et al. Electric field effect in atomically thin carbon films. Science 306,
666– 669 (2004).
modified Hummers method as originally presented by Kovtyukhova and 8. Kotov, N. A., Dekany, I. & Fendler, J. H. Ultrathin graphite oxide – polyelectrolyte composites
colleagues10,30. As-synthesized graphite oxide was suspended in water to give a prepared by self-assembly: transition between conductive and non-conductive states. Adv. Mater. 8,
brown dispersion, which was subjected to dialysis to completely remove residual 637– 641 (1996).
9. Cassagneau, T., Guerin, F. & Fendler, J. H. Preparation and characterization of ultrathin films
salts and acids. Ultrapure Milli-Qw water was used in all experiments. As-
layer-by-layer self-assembled from graphite oxide nanoplatelets and polymers. Langmuir 16,
purified graphite oxide suspensions were then dispersed in water to create a 7318–7324 (2000).
0.05 wt% dispersion. Exfoliation of graphite oxide to GO was achieved by 10. Kovtyukhova, N. I. et al. Layer-by-layer assembly of ultrathin composite films from micron-sized
ultrasonication of the dispersion using a Brandson Digital Sonifier (S450D, graphite oxide sheets and polycations. Chem. Mater. 11, 771– 778 (1999).
11. Hirata, M., Gotou, T. & Ohba, M. Thin-film particles of graphite oxide 2: Preliminary studies for
500 W, 30% amplitude) for 30 min. The obtained brown dispersion was then internal micro fabrication of single particle and carbonaceous electronic circuits. Carbon 43,
subjected to 30 min of centrifugation at 3,000 r.p.m. to remove any unexfoliated 503 –510 (2005).
graphite oxide (usually present in a very small amount) using an Eppendorf 5702 12. Szabo, T., Szeri, A. & Dekany, I. Composite graphitic nanolayers prepared by self-assembly between
finely dispersed graphite oxide and a cationic polymer. Carbon 43, 87– 94 (2005).
centrifuge with a rotor radius of 14 cm. In a typical procedure for chemical 13. Lerf, A., He, H. Y., Forster, M. & Klinowski, J. Structure of graphite oxide revisited.
conversion of graphite oxide to graphene, the resulting homogeneous dispersion J. Phys. Chem. B 102, 4477–4482 (1998).
(5.0 ml) was mixed with 5.0 ml of water, 5.0 ml of hydrazine solution (35 wt% in 14. Szabo, T. et al. Evolution of surface functional groups in a series of progressively oxidized graphite
oxides. Chem. Mater. 18, 2740 –2749 (2006).
water, Aldrich) and 35.0 ml of ammonia solution (28 wt% in water, Crown 15. Everett, D. H. Basic Principles of Colloid Science (The Royal Society of Chemistry, London, 1988).
Scientific) in a 20-ml glass vial. The weight ratio of hydrazine to GO was about 16. Li, D. & Kaner, R. B. Processable stabilizer-free polyaniline nanofiber aqueous colloids.
7:10. After being vigorously shaken or stirred for a few minutes, the vial was put Chem. Commun. 3286– 3288 (2005).
17. Baughman, R. H., Zakhidov, A. A. & de Heer, W. A. Carbon nanotubes—the route toward
in a water bath (95 8C) for 1 h. Unless specifically stated, graphene dispersions applications. Science 297, 787 –792 (2002).
prepared according to the above procedure were used for further characterization 18. Niyogi, S. et al. Chemistry of single-walled carbon nanotubes. Acc. Chem. Res. 35, 1105 –1113 (2002).
and film fabrication in this work. Note that the concentration of hydrazine in the 19. Bahr, J. L. & Tour, J. M. Covalent chemistry of single-wall carbon nanotubes. J. Mater. Chem. 12,
reduction mixture can be varied from 0.0175 wt% (used in the above procedure) 1952 –1958 (2002).
20. Sun, Y. P., Fu, K. F., Lin, Y. & Huang, W. J. Functionalized carbon nanotubes: properties and
to 1.75 wt%. However, in order to obtain stable dispersions when the applications. Acc. Chem. Res. 35, 1096– 1104 (2002).
concentration is greater than 0.0175 wt%, excess hydrazine in the resulting 21. Gilje, S., Han, S., Wang, M., Wang, W. & Kaner, R. B. A chemical route to graphene for device
dispersions must be removed by dialysis against a 0.5% ammonia solution applications. Nano Lett. 7, 3394 –3398 (2007).
22. Gómez-Navarro, C. et al. Electronic transport properties of individual chemically reduced graphene
once the reduction is complete. A RN2H4/GO ratio of 7:10 appears to be an optimal oxide sheets. Nano Lett. 7, 3499 –3503 (2007).
ratio for producing stable dispersions of highly conducting graphene sheets 23. Skakalova, V., Kaiser, A. B., Dettlaff-Weglikowska, U., Hrncarikova, K. & Roth, S. Effect of chemical
(see Supplementary Information). treatment on electrical conductivity, infrared absorption, and Raman spectra of single-walled carbon
nanotubes. J. Phys. Chem. B 109, 7174 –7181 (2005).
The graphene paper, shown in Fig. 5a, was made by filtration of a dispersion 24. Dikin, D. A. et al. Preparation and characterization of graphene oxide paper. Nature 448,
through an Anodisc membrane filter (47 mm in diameter, 0.2 mm pore size; 457 –460 (2007).
Whatman), similar to the method reported for making GO paper24. The 25. Kirchmeyer, S. & Reuter, K. Scientific importance, properties and growing applications of
poly(3,4-ethylenedioxythiophene). J. Mater. Chem. 15, 2077– 2088 (2005).
graphene paper was cut by a razor blade into rectangular strips of approximately 26. Decher, G. Fuzzy nanoassemblies: toward layered polymeric multicomposites. Science 277,
4 mm  15 mm for mechanical testing. 1232 –1237 (1997).
27. Hammond, P. T. Form and function in multilayer assembly: new applications at the nanoscale.
CHARACTERIZATION Adv. Mater. 16, 1271– 1293 (2004).
28. Tang, Z. Y., Wang, Y., Podsiadlo, P. & Kotov, N. A. Biomedical applications of layer-by-layer assembly:
UV-vis absorption and/or transmission spectra were obtained using a Shimadzu from biomimetics to tissue engineering. Adv. Mater. 18, 3203–3224 (2006).
UV 1601 spectrophotometer. The spectra were taken from the reaction mixture 29. Jan, E. & Kotov, N. A. Successful differentiation of mouse neural stem cells on layer-by-layer
(diluted by a factor of 30) at different times. The dispersion/aggregation state of assembled single-walled carbon nanotubes composite. Nano Lett. 7, 1123– 1128 (2007).
30. Hummers, W. S. & Offeman, R. E. Preparation of graphite oxide. J. Am. Chem. Soc. 80, 1339 (1958).
CCG sheets in water was monitored by measuring their average particle size
using a Malvern Zetasizer Nano-ZS particle analyser. Note that the particle size
measurement on this instrument is based on the assumption that the particles Acknowledgements
are spherical, so the instrument is unable to give the absolute sizes of graphene D.L. and G.G.W. acknowledge support from the Australian Research Council. R.B.K. thanks the
Microelectronics Advanced Research Corporation for financial support.
sheets. Nevertheless, the measurements obtained provide a means of monitoring
Correspondence and requests for materials should be addressed to D.L. or G.G.W.
dispersion stability. Attenuated total reflectance FT-IR spectra of free-standing Supplementary information accompanies this paper on www.nature.com/naturenanotechnology.
films prepared by vacuum filtration were recorded on a Nicolet AVATAR 360
FTIR spectrometer with a Smart OMNI Sampler with a germanium crystal. AFM Author contributions
images were taken in tapping mode with the SPM Dimension 3100 from Veeco. D.L. conceived and designed the experiments, R.B.K. and G.G.W. were involved in discussions on the
Conductivity measurements of free-standing CCG films prepared by vacuum design and interpretation of the experiments, and D.L., M.B.M. and S.G. performed the experiments.
filtration were carried out on a Jandel RM3 Test Unit using a four-point-probe D.L., R.B.K. and G.G.W. co-wrote the paper. All the authors discussed the results and commented on
the manuscript.
head with a pin-distance of about 1 mm. Mechanical tensile tests were conducted
with a Q800 Dynamic Mechanical Analyzer (TA Instruments). The samples were Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

nature nanotechnology | VOL 3 | FEBRUARY 2008 | www.nature.com/naturenanotechnology 105

© 2008 Nature Publishing Group
NEWS & VIEWS

Graphene

Calling all chemists

Graphene has potentially useful electronic properties but it is difficult to produce and process
on large scales. Working with chemically modified forms of graphene — such as graphene
oxide — may provide an alternative.

Rod Ruoff graphite into individual layers. Other

is in the Department of Mechanical Engineering, layered materials, such as various clays and
University of Texas at Austin, Austin, micas, have been readily dispersed in water
Texas 78712, USA. to supply individual layers but, perhaps
 e-mail: r.ruoff@mail.utexas.edu surprisingly, this has never been achieved
for graphite in any solvent. Another

T
here has been an explosion of alternative might seem to be ‘expandable
J. O. Sofo, Penn State

interest in the physics of graphene — graphite’, an intercalation compound that

individual layers of graphite in which can expand in volume by a factor of up to
each carbon atom is bonded to three other 1,000 upon rapid heating6. However, the
carbon atoms — since it was first deposited platelets in such expanded graphites are
by mechanical exfoliation in 20041,2. In this roughly 50 nm thick and contain hundreds
approach, which has spawned a cottage of stacked graphene layers.
industry of ‘peeling and rubbing’ in labs Figure 1 Schematic diagram of a graphene sheet So, given the limits of existing ‘physical’
around the world, the large majority of that has been chemically modified by hydrogenating approaches, does the remarkably versatile
platelets are actually too thick to be of the carbon atoms in well-defined regions to define chemistry of carbon offer promising
use in fundamental physics experiments, ‘ribbons’ of graphene. Simulations of the electronic alternatives? For example, could chemical
so well-trained eyes are needed to find properties of chemically modified graphene will be a tuning be used to alter the physical
the relatively few monolayer flakes that useful guide to experimenters. For example, how wide properties of graphene sheets? Indeed,
are produced. do the hydrogenated regions (yellow) need to be to could nanoelectronic devices be made
The results of experiments in which define high-mobility graphene pathways? Does the from chemically functionalized graphene?
these flakes are ‘wired up’ by electron- orientation of the ribbon with respect to the graphene Rather than etching a hypothetical
beam lithography and then studied under lattice (such as along the ‘armchair’ or ‘zigzag’ graphene coating on a 300-mm silicon
a variety of conditions, including high directions) influence the electronic properties? See, for wafer to create a pattern of highly
magnetic fields and ultralow temperatures, example, simulations for a fully hydrogenated graphene conducting graphene nanoribbons, would
suggest that graphene-based devices might sheet (called graphane because it is an alkane) it be better to chemically functionalize
perform well enough at room temperature by Jorge Sofo and co-workers14 at Pennsylvania selected regions so that the carbon atoms
to have a future in commercial-scale State University. in these regions were bonded to four
nanoelectronics3. The large mobility of the other atoms, rather than three? In this
carriers in graphene is also noteworthy approach the tetravalently bonded atoms
because speed is often central to device would jealously hold on to their charge
performance. If it were possible to for sensing pressure, as components in carriers, rather than sharing them openly
uniformly ‘coat’ 300-mm silicon wafers nanoelectromechanical systems, or in like the trivalently bonded atoms, thereby
with a monolayer or perhaps bilayer4 chemical sensing owing to their high defining the desired pattern of graphene
of graphene, standard semiconductor surface area. nanoribbons (see Fig. 1). Input from
processing techniques could, perhaps, Direct growth or deposition on theorists about the likely performance of
pattern graphene into ‘nanoribbons’ for industrial-scale silicon wafers should different devices made using this approach
commercial devices. be targeted for nanoelectronics, but will be valuable7.
The remarkable ‘flatness’ of graphene significant progress has also been made in We have talked so far about
(when it is deposited on really flat the epitaxial growth of graphene layers on possibilities, but what progress has already
substrates) also suggests that it might single-crystal 4-inch silicon carbide wafers been made with respect to developing
eventually be possible to fabricate devices under high temperature and ultrahigh- chemically modified graphene (CMG)?
containing multiple layers of graphene vacuum conditions5. However, the One approach taken by my group has been
and other materials (as long as the combination of high cost and small wafer to use colloidal suspensions of graphite
processing conditions for the subsequent size might limit the widespread application oxide, which has led to the possibility of
layers do not harm the graphene layers of this approach. distributing CMG in polymer matrices8 or
underneath). In addition to the possibility So, in the absence of a reliable and in thin layers of silica glass9, and of layering
of low-power, high-density and high- cost-effective technique for coating CMG to achieve a paper-like material
speed switches and logic elements, 300‑mm silicon wafers, what other with interesting mechanical properties10.
graphene-based devices may also find other alternatives might be considered? One In separate experiments we have used the
applications — as atom-thick membranes possibility might be to exfoliate bulk deposition and dry-down of droplets from

10 nature nanotechnology | VOL 3 | JANUARY 2008 | www.nature.com/naturenanotechnology

© 2008 Nature Publishing Group
 NEWS & VIEWS
aqueous suspensions of graphene oxide substrates and then converted them to interest to the nanoelectronics industry.
(which is one type of CMG) to produce electrically conductive CMG that allowed Nevertheless, whatever happens, graphene
relatively large platelets on substrates, and for voltage gating. Individual platelets and chemically modified graphene
then mapped their optical properties before were identified by light microscopy, and will remain of interest for fundamental
and after heat treatment11. A compelling their electrical properties were studied as a research. It may be that pristine
reason for interest in such colloidal function of temperature or prior exposure graphene is essential for applications in
suspensions is that all the graphene oxide to hydrazine or hydrogen plasma. nanoelectronics. However, it could be that
platelets in the suspension seem to be This work is still at an early stage and some degree of chemical functionalization
individual layers. Alas, this heavily oxidized some questions remain to be answered. In may be tolerated, or perhaps even found to
form of graphene is electrically insulating, converting graphite to graphite oxide, have be beneficial.
but the platelets can be made electrically significant defects been introduced into the
References
conductive by exposing them to chemical individual layers? Are unwanted defects 1. Novoselov, K. S. et al. Science 306, 666–669 (2004).
reductants or heat. introduced when making the aqueous 2. Lu, X. K., Yu, M. F., Huang, H. & Ruoff, R. S. Nanotechnology
The ‘chemistry part’ of the graphene suspension of graphene oxide platelets? 10, 269–272 (1999).
3. Geim, A. K. & MacDonald, A. H. Physics Today 60, 35–41
story has only just begun and because If one is able to keep the basic carbon (August 2007).
of the extraordinary number of levers skeleton intact without too many defects, 4. Min, H. K., Sahu, B., Banerjee, S. K. & MacDonald A. H.
that chemistry affords, there will be a will it be possible to manipulate the surface Phys. Rev. B 75, 155115 (2007).
growing dialogue between chemists and chemistry, and perhaps even ‘return’ 5. Berger, C. et al. Science 312, 1191–1196 (2006).
6. Celzard, A., Mareche, J. F. & Furdin, G. Prog. Mater. Sci.
researchers in other fields (such as physics, to pristine graphene, albeit by a rather 50, 93–179 (2005).
materials and processing) who, otherwise, circuitous route? 7. Chernozatonskii, L. A., Sorokin, P. B. & Brüning J. W.
might only have worked with graphene Although it should be possible to Appl. Phys. Lett. 91, 183103 (2007).
8. Stankovich, S. et al. Nature 442, 282–285 (2006).
in its pristine form. As an example of a position the larger CMG platelets11
9. Watcharotone, S. et al. Nano Lett. 7, 1888–1892 (2007).
first step, teams led by Richard Kaner of where needed on a silicon wafer for 10. Dikin, D. A. et al. Nature 448, 457–460 (2007).
UCLA12, Klaus Kern of the Max Planck further processing, and to perhaps create 11. Jung, I. et al. Nano Lett. 7, 3569–3575 (2007).
Institute in Stuttgart and the EPFL13, and ‘striped’ graphene platelets (see Fig. 1) 12. Gilje, S., Han, S., Wang, M., Wang, K. L & Kaner, R. B.
Nano Lett. 7, 3394–3398 (2007).
myself (I. Jung, D. Dikin, R. Piner & R.S.R, that could also be assembled on silicon 13. Gómez-Navarro, C. et al. Nano Lett., 7, 3499–3503 (2007).
manuscript in preparation) have recently wafers through appropriate patterning, it 14. Sofo, J. O., Chaudhari, A. S. & Barber, G. B. Phys. Rev. B
deposited graphene oxide platelets onto is not yet clear if this approach will be of 75, 153401 (2007).

Nanometrology
Island hopping
Developing new measurement standards
that are based on the fundamental
physical constants, such as Planck’s 20
constant and the mass of the electron, is 5
15
a top priority in metrology laboratories
Current (pA)

around the world. However, the unit of 10 0

electrical current, the Ampere, is still
5
defined in terms of the forces acting on –5
wires carrying currents, which is why 0
400
a number of groups are working on Gat
4 1 200 1
e am Bia
alternative standards. plitu 2 0 s vo 0
ltag –200 0
de )
A central element in some of these set (e e (µ )
(e ) 0 –1 ate off V) –400 –1 set (e
G
Gate off
approaches is a ‘turnstile’ that opens
and closes at a well-defined frequency, f,
allowing a well-defined number of tunnel barriers. The electrons must left, which was recorded at a frequency
electrons, N, to pass through the turnstile tunnel through these barriers to enter or of 12.5 MHz, shows N increasing from
every time it opens, leading to a current leave the island. The device is controlled zero up to ten. The plateaux in the other
I=Nef, where e is the charge of the by a d.c. voltage across the source and image, which was recorded at 20 MHz,
electron. Writing in Nature Physics, drain, and a combination of a.c. and correspond to N=1, 0 and –1 (that is, a
Jukka Pekola and co-workers at the d.c. voltages that are applied to a metal current flowing in the opposite direction).
Helsinki University of Technology and gate electrode. A quantum standard for current
Stony Brook University report a new Pekola and co-workers measure the would bring it into line with the
type of turnstile based on a hybrid current through the turnstile as these standards for resistance and voltage,
single-electron transistor (Nature Phys. voltages are varied, with the plateaux in which are based on the quantum
doi:10.1038/nphys808; 2007). the data (left) demonstrating that they Hall effect and the Josephson effect
The hybrid device consists of an are able to control N. In both figures respectively, thus completing a
island of superconducting aluminium the vertical axis is the current and the metrological triangle for electrical units
that is separated from copper source and horizontal axes are different components based on Ohm’s law.
drain electrodes by aluminium oxide of the voltages. The figure on the far  Peter Rodgers

nature nanotechnology | VOL 3 | JANUARY 2008 | www.nature.com/naturenanotechnology 11

© 2008 Nature Publishing Group
 EDITORIAL

Vol.2 No.4 April 2007 www.nature.com/naturenanotechnology

Carbon does it again

First isolated less than three years ago, graphene is currently the hottest topic in condensed-
matter physics and materials science. Expect further breakthroughs in fundamental physics
and, possibly, the advent of graphene-based electronics.

Once upon a time, the wonder and behave as particles called Dirac fermions cut out from a graphene sheet1. Doing
diversity of the material world was that cannot be slowed down because their this on an industrial scale will require
exemplified by diamond and graphite. velocity is independent of their energy breakthroughs in both epitaxial growth
How could it be, teachers would ask (which explains the high mobility). and patterning technology.
incredulous students, that these two very Examples of other novel quantum There is, of course, nothing new about
different materials — one incredibly phenomena — including several that the promise of novel nanoscale electronics,
beautiful and expensive, the other occur at room temperature — are still and graphene has lots of competition in
completely commonplace — were being discovered3. this area. In the short-term, therefore, it
made of the same stuff? And so it In recent months, attention has also is much more likely to have an impact on
remained until the discovery of fullerene turned to the structural and mechanical fundamental physics because, as Dirac
molecules, also known as the ‘third form properties of graphene. Researchers in fermions, the electrons in graphene do not
of carbon’, in 1985. the US have made nanoelectromechanical behave like those in ordinary materials.
Carbon nanotubes followed relatively systems in which the resonator is made Using graphene to explore the connections
quickly, in 1991, and proved to be even of just one or a few graphene sheets4, and between condensed-matter physics and
more interesting than fullerenes from a European group has shown that single various fundamental phenomena in
both a pure and an applied point of graphene sheets are not perfectly flat5. quantum electrodynamics is likely to prove
view. Nanotubes were often described as highly fruitful6.
rolled-up sheets of carbon atoms, According to Geim and Novoselov
and these sheets were occasionally
Graphene displays a “only the tip of the iceberg has
referred to as graphene, but no one remarkable range of novel been uncovered so far [and] most
seemed that interested in working with electronic behaviour. experimental groups working on
graphene itself. graphene have not published even a
All that changed in 2004 when single paper on the subject”1. They go
Andre Geim, Kostya Novoselov and co- The latter result is actually good news on to say that this has been “a truly
workers at Manchester University and because it has been firmly established, frustrating experience for theorists”,
elsewhere showed that graphene (a.k.a. by theory and experiment, that perfect although this has not stopped the
‘the Mother of all graphitic forms’) could two-dimensional crystals cannot exist in latter from publishing papers on
be isolated from graphite simply using a free state. everything from Andreev reflection
adhesive tape. They soon discovered that So what challenges must be to zitterbewegung (a process that
writing on silicon with a piece of graphite overcome before we see ‘graphenium is normally understood in terms
was an even better method, while a group inside’1 stickers on our laptops? Well of interactions between matter and
at Georgia Tech developed an epitaxial the fact that graphene is usually a metal antimatter). Indeed, like the electrons in
growth process that might be suitable for is not an advantage for conventional graphene, research in the field appears to
mass-production1,2. semiconductor applications, although it be unstoppable.
New physics flowed thick and is possible to engineer a semiconductor
fast: it was soon discovered that the bandgap. However, rather than integrate References
1. Geim, A. K. & Novoselov, K. S. Nature Mater. 6,
electron mobility of graphene was graphene into conventional silicon
183–191 (2007).
an order of magnitude higher than devices, a more radical approach 2. Neto, A. C., Guinea, F., & Peres, N. M. Phys. World 19, 33–37
that of commercial silicon wafers, would be to make devices directly from (November 2006).
which has obvious implications for graphene: in other words, the conducting 3. van den Brink, J. Nature Nanotech. 2, 199–201 (2007).
4. Bunch, J. S. et al. Science 315, 490–493 (2007).
electronics. Moreover, graphene displays channels, quantum dots, interconnects 5. Meyer, J. C. et al. Nature 446, 60–63 (2007).
a remarkable range of novel electronic and other elements needed to make truly 6. Katsnelson, M. I. & Novoselov, K. S. Preprint at
behaviour: in particular, the electrons nanoscale electronic circuits would be <http://arxiv.org/cond-mat/0703374> (2007).

nature nanotechnology | VOL 2 | APRIL 2007 | www.nature.com/naturenanotechnology 191

 NEWS & VIEWS

GRAPHENE

From strength to strength

With exciting new results appearing every week, graphene is one of the hottest topics in
physics, and may also form the basis of a new approach to electronics a decade from now.

Jeroen van den Brink a

1 μm
b
is at the Instituut-Lorentz for Theoretical Physics,
Leiden University, 2300 RA Leiden,
The Netherlands.
e-mail: brink@lorentz.leidenuniv.nl

t is one of those seemingly simple

I questions that kids come up with: “How

does a pencil work?” A scientifically
knowledgeable parent will dutifully explain
that the grey stuff inside the pencil is called
graphite, that graphite is like a stack of
pancakes with each pancake being a layer of Figure 1 Graphene-based superconducting transistor. a, An atomic force microscope image and b, a schematic
carbon atoms, and that when the pencil is representation of a layer of graphene between two superconducting electrodes (yellow). Graphene is not a
moved across a piece of paper, the pancakes superconductor but it can support a supercurrent when placed between two superconductors. A gate electrode
peel off to leave marks on the paper. But (not shown) is used to control the supercurrent, which can be carried by electrons or holes6. The red and blue
parents know that such an answer will only cones in b represent the valleys in the electronic structure of graphene (see Box 1).
provoke more questions such as: “can you see
one of the pancakes?”
Less than three years ago Andre Geim,
Kostya Novoselov and co-workers at excess charge on the graphene sheet. There Hall effect in graphene at room temperature.
Manchester University in the UK and the also seems to be ample room for optimizing This is possible because the energy that
Institute for Microelectronics Technology in the manufacturing process of graphene, so it sets the scale for the quantum Hall effect in
Chernogolovka, Russia, discovered that single is little wonder that researchers at companies ordinary semiconductors is proportional to
layers of graphite — also known as like Intel and IBM have started to work on the inverse effective mass of the electrons,
graphene — really can be seen1. The graphene. But will physicists and engineers but in graphene, owing to its unique
technique they used basically involves be able to convert the undoubted promise of electronic structure, the effective mass of the
drawing with a piece of graphite on a silicon graphene into real-world electronic devices? conduction electron vanishes.
substrate. If the substrate has been prepared Three recent discoveries indicate that Moreover, in work submitted for
with an ultrathin layer of silicon dioxide on things are moving in the right direction. publication, Leonid Ponomarenko,
top, the graphene flakes, which are just half Writing in Science, Novoselov and co-workers Fredrik Schedin, Novoselov and Geim report
a nanometre thick, become visible under an report further evidence of the remarkable the first results on an all-graphene single-
ordinary optical microscope. electronic properties of graphene4. It has been electron transistor (SET) that also
More surprises followed in 2005 when known for more than two decades that the operates at room temperature (ref. 5;
Geim, Novoselov and co-workers and, resistance of a very pure two-dimensional L. A.Ponomarenko, F. Schedin,
separately, Philip Kim, Horst Stormer and conductor changes in a very particular way K. S. Novoselov and A. K. Geim, manuscript
colleagues at Columbia University in the when an external magnetic field is applied. in preparation). As its name suggests, the
USA, drove an electrical current through In the classical Hall effect, which was SET is the ultimate transistor — operating
a graphene sheet2,3. In addition to various discovered in 1879, the Hall resistance is at the smallest possible length scales and
novel electronic phenomena (see Box 1), they proportional to the magnetic field. But in the manipulating the smallest possible currents.
found that the electron mobility quantum Hall effect, which was first observed The SET can be viewed as a quantum dot or
of graphene can be as high as in very pure semiconductors in 1980, the Hall ‘box’ for holding electrons, with the electrons
10,000 cm2 V–1 s–1 at room temperature, resistance is quantized and increases in units being able to enter and leave the quantum dot
which is about ten times higher than the of h/νe2 as the magnetic field is increased, via separate electrodes.
mobility of commercial silicon wafers. where h is Planck’s constant, ν is an integer The operation of a SET relies on electrons
Mobility is often severely limited by and e is the charge on the electron. hopping between the quantum dot and the
structural imperfections, but the structural This quantum Hall effect cannot be leads. The key point is that electrons can only
purity of graphene means that electrons can observed in conventional semiconductors pass through the dot one by one — because of
travel huge distances — 300 nanometres or above about 30 K because thermal the Coulomb repulsion between the electrons
more — without being scattered. Moreover, fluctuations wash out the delicate quantum it is simply not possible for two electrons to
this impressive mobility is little affected by effects that are responsible for it. be on the dot at the same time. If the device
changes in temperature or by the presence of Novoselov et al. have now seen the quantum is ‘open’, current can flow from one lead to

nature nanotechnology | VOL 2 | APRIL 2007 | www.nature.com/naturenanotechnology 199

NEWS & VIEWS
© CHRIS EWELS a b c

Figure 2 The structure of graphene. a, In graphene, each carbon atom is bonded to three other carbon atoms to form a two-dimensional honeycomb structure that displays a
variety of novel electronic properties. b, c, Scanning electron micrographs showing a single graphene sheet (b) and thicker layers of graphite (c).

the other via the quantum dot. However, if a superconducting electronic devices consume Delft team was able to maintain the quantum
single electron is kept hostage on the dot by, less power than ordinary electronics and also coherence needed to make a superconducting
for instance, applying an external electric field, offer ultrafast switching times. Graphene transistor (Fig. 1).
other electrons cannot pass and the SET is is not superconducting by itself but, like Just as the current in a conventional
‘closed’. The electric field that opens and closes most normal conductors, it can support transistor is controlled by a gate voltage,
the SET is supplied by a gate electrode near a supercurrent over short distances when the supercurrent in the Delft transistor is
the quantum dot. The Manchester team used placed between two superconducting manipulated by a gate electrode placed under
electron-beam lithography and isotropic dry electrodes owing to a phenomenon known the graphene sheet. The voltage on the gate
etching to make the quantum dot, the leads as the Josephson effect. For this to happen electrodes controls the density of charge
and the gate electrodes — all from the same the graphene must not destroy the quantum carriers that are injected onto the sheet, and
sheet of graphene. coherence between the electrons in the the sign of this voltage determines if these
Writing in Nature, meanwhile, Cooper pairs that enter the graphene sheet charge carriers are electrons or holes. As the
Hubert Heersche and co-workers from from the superconducting electrodes. By supercurrent can be carried by electrons
the Delft University of Technology in the reducing the dimensions of their device to or holes, the transistor is said to be bipolar.
Netherlands report the first graphene-based the nanoscale, and by taking advantage of Another unusual feature is that the graphene
superconducting transistor6. In principle, the intrinsic cleanliness of graphene, the supports a finite supercurrent, even when the

Box 1 Why is graphene so special?

In a graphene sheet, each carbon atom the interaction between the electrons and by experiment, but there is still a lot for
forms bonds with three other carbon atoms the periodic potential of the honeycomb theorists to think about, including the
to produce a two-dimensional honeycomb lattice results in a band structure in which recent prediction that p–n junctions in
structure (Fig. 2). These carbon–carbon the valence and conduction bands touch at graphene could be used to focus electrons
bonds are very strong, which renders just two points, rather than overlapping, and and, possibly, demonstrate the electronic
graphene stable, even when it has been this leads to the formation of ‘valleys’ in the equivalent of a negative refractive index in
cut into nanoscale structures. Indeed, last electronic structure (Fig. 1b). At these optics12. The electrons in graphene also have
month Jannik Meyer of the Max Planck so-called Dirac points, the electrons a degree of freedom that is not found in
Institute in Stuttgart and co-workers behave as relativistic quasiparticles called other electronic systems because they can be
reported that they had been able to freely Dirac fermions. At first this came as a at one of the two Dirac points, and theorists
suspend a graphene membrane on a surprise but it is now well established. are exploring the possibility of exploiting
microfabricated scaffold in air10. Although Dirac fermions always move with a this in devices13.
the membranes are strongly rippled and velocity that is independent of their energy Other condensed-matter theorists are,
corrugated, they do not rupture. It is this and direction. This is a property that they for example, taking on the basic challenge
mechanical stability that sets the stage for share with photons, which always move at of exploring the consequences of the
electronic stability. the speed of light, c, irrespective of their Coulomb interactions between Dirac
Graphene is a zero-gap semiconductor, energy. Although the Dirac fermions in fermions, and their colleagues in other fields
an electronic state of matter that is precisely graphene are much slower — only moving at are also getting involved. Cosmologists, for
in between a semiconductor and a metal. about c/300 — the important point is that it is instance, are interested in the behaviour
Semiconductors are characterized by an impossible to slow them down, which means of relativistic quantum particles in curved
appreciable energy gap between the valence that the charge mobility is very high11. spaces, so the results of experiments on
and conduction bands, whereas in metals So far, most theoretical predictions curved graphene sheets could provide them
these two bands largely overlap. In graphene about graphene have been confirmed with food for thought14.

200 nature nanotechnology | VOL 2 | APRIL 2007 | www.nature.com/naturenanotechnology

 NEWS & VIEWS
density of charge carriers is zero. Moreover, the large-scale manufacture of such devices of graphene-based nanoelectronics looks
all the superconducting transistors made by with reliable characteristics will be a challenge. possible, if difficult. The main obstacles
the Delft group can support a supercurrent, The use of graphene ‘nanoribbons’ concern large-scale manufacturing and the
making the ability to sustain a Josephson could be another solution because the challenge of patterning graphene at the atomic
effect yet another robust feature of graphene. electronic properties of the ribbon depend level. However, it took 50 years for silicon
The development of commercial on the type of edge it has: simple theoretical technology to reach maturity, so the parents of
graphene-based nanoelectronics will considerations suggest that all ribbons with today would not be irresponsible if they told
encounter a number of technological so-called zigzag edges are metallic, whereas their kids that when they grow up, they might
challenges. First and foremost, is it possible those with ‘armchair’ structures can be well find themselves using devices made from
to produce high-quality graphene wafers semiconductors (with a bandgap) or metals, carbon pancakes just one atom thick.
in sufficient quantities? Options include depending on their width. However, some
dissolving graphite and depositing it as a theorists have recently suggested that all References
1. Novoselov, K. S. et al. Science 306, 666–669 (2004).
monolayer on a substrate in a bottom-up graphene ribbons are semiconducting7. At
2. Novoselov, K. S. et al. Nature 438, 197–200 (2005).
approach, or burning off silicon in the top present the edges are naturally passivated 3. Zhang, Y., Tan, J., Stormer, H. L. & Kim, P. Nature 438,
layer of silicon carbide crystals in a by hydrogen atoms, but they could be 201–204 (2005).
top-down approach. functionalized by attaching different chemical 4. Novoselov, K. S. et al. Science 315, 1379 (2007).
5. Geim, A. K. & Novoselov, K. S. Nature Mater. 6, 183–191 (2007).
Another problem in the transistors made groups, providing potential for sensor 6. Heersche, H., Jarillo-Herrero, P., Oostinga, J. B.,
so far is that there is a noticeable leakage applications. Preliminary reports on graphene Vandersypen, L. M. K. & Morpurgo, A. F. Nature 446, 56–59 (2007).
current, even when the device is meant to ribbons are appearing8,9, but the challenge of 7. Son, Y.-W., Cohen, M. L. & Louie, S. G. Phys. Rev. Lett. 97,
switched off. This is related to the unstoppable routinely cutting and pasting graphene with 216803 (2006).
8. Han, M. Y., Ozyilmaz, B., Zang, Y, & Kim, P. Preprint at
nature of the electrons in graphene (see the required level of precision will require the <http://arxiv.org/cond-mat/0702511> (2007).
Box 1) and can only be remedied by creating very best tools from nanotechnology. 9. Chen, Z., Lin, Y.-M., Rooks, R. J. & Avouris, P. Preprint at
a bandgap. One way to solve this problem is Although graphene research is still in <http://arxiv.org/cond-mat/0701599> (2007).
10. Meyer, J. et al. Nature 446, 60–63 (2007).
to use bilayer graphene, because, according its infancy, the number of experimental
11. Katsnelson, M. I., Novoselov, K. S. & Geim, A. K. Nature Phys. 2,
to theory, a voltage difference between the breakthroughs in the past three years has 620–625 (2006).
two layers will generate a bandgap. However, been stunning: the observation of the 12. Cheianov, V. V., Fal’ko, V. & Altshuler, B. L. Science 315,
making bilayer graphene on a large scale is quantum Hall effect at room temperature, 1252–1255 (2007).
13. Rycerz, A, Tworzydło, J. & Beenakker, C. W. J. Nature Phys. 3,
probably even more difficult than making the first graphene-based SET, and the bipolar 172–175 (2007).
single-layer graphene. SETs could circumvent superconducting transistor are just a few of 14. Cortijo, A. & Vozmediano, M. A. H. Nucl. Phys. B 763,
the problems with leakage currents, but again many highlights. Moreover, the development 293–308 (2007).

NANOPARTICLES

A very versatile nanocapsule

Numerous copies of a pumpkin-shaped molecule can be linked together to form a
nanocapsule shell that can trap compounds inside. The outer surface of this capsule can be
decorated with other species by plugging them into the cavities of the hollowed-out pumpkins.

Darrell W. Kuykendall and international edition of Angewandte Chemie, — under some conditions their presence
Steven C. Zimmerman* Kimoon Kim and colleagues at Pohang can be somewhat fleeting, with dissociated
are in the Department of Chemistry at the University of Science and Technology in components often observed.
University of Illinois at Urbana-Champaign, Korea describe an impressively versatile One way to enhance the stability of these
Urbana, Illinois 61801, USA. synthesis of polymer nanocapsules with structures is to covalently link the constituent
*e-mail: sczimmer@uiuc.edu surfaces that can be easily and reversibly parts. In general, the crosslinkable groups
decorated with other molecules2. are attached — either covalently or non-
n the past half-century, scientists have One of the many different strategies covalently — to a core template. The core is

I made enormous strides towards creating

nanoscale structures and devices. Often
these efforts are inspired by biology, not
for making synthetic nanocapsules takes
advantage of self-assembly — the method by
which well-ordered architectures are formed
removed by chemical degradation, yielding
architectures that will ideally retain the shape
of the central template6. Such ‘core–shell’
least because nature has been making spontaneously from small molecules3 or approaches are now well established1, as
functional nanosystems using the ‘bottom- polymers4. For example, convex aromatic are hybrid construction approaches that
up’ approach for millennia. Scientists compounds can be pieced together along use a combination of self-assembly and
have, with varying degrees of success, their edges through hydrogen bonds to templating7. The new approach taken by
been attempting to mimic the basic capsid form hollow spheres that encapsulate small Kim and co-workers to make a nanocapsule
architecture of viruses in order to create molecules5. However, these supramolecular is fundamentally different from these
nanoscale capsules for applications ranging assemblies are held together by relatively strategies — the building blocks are simply
from targeted drug delivery to functional weak non-covalent bonds and their stability stitched together using traditional chemical
materials development1. Now, writing in the depends strongly on their environment reactions (rather than non-covalent

nature nanotechnology | VOL 2 | APRIL 2007 | www.nature.com/naturenanotechnology 201