PHYSICAL REVIEW B 66, 115415 共2002兲

Density-functional-theory calculations of charged single-walled carbon nanotubes

Ji Luo, L.-M. Peng, Z. Q. Xue, and J. L. Wu*
Department of Electronics, Peking University, Beijing 100871, China
共Received 3 April 2002; revised manuscript received 26 June 2002; published 26 September 2002兲
The electronic structures of charged carbon nanotube cluster models were calculated with the density-
functional-theory 共DFT兲 method. With the increasing number of extra electrons, 共a兲 a potential barrier is
formed near the tube wall; its height increases and the width at the highest occupied molecular-orbital
共HOMO兲 level decreases; 共b兲 the HOMO energy increases linearly, giving chances to electrons to tunnel
through the barrier and escape; 共c兲 the total energy first decreases and then increases, presenting a perfect
parabolic curve and indicating an optimum number of extra electrons; and 共d兲 electrical field near the tube was
worked out. A comparison between the tube and a perfect conductor demonstrates some of their resemblances,
but the exact properties such as the tube capacitance depend on the quantum theory calculations. The total-
energy variation obeys Koopmans’ theorem and although calculated with DFT method, it can be expressed in
terms of the tube’s classical parameters such as capacitance and work function. The electronic charge rear-
rangement after the addition of extra electrons is presented and this might be the determinant of the electronic
structures of charged carbon nanotubes.

DOI: 10.1103/PhysRevB.66.115415 PACS number共s兲: 61.48.⫹c, 73.22.Dj, 36.40.Wa, 31.15.Ew

Carbon nanotubes have attracted extensive research inter- a 共10,0兲 zigzag open-ended SWNT containing 200 carbon
est both experimentally and theoretically since their atoms. Zero, two, four, six, and eight extra electrons were
discovery.1,2 Their unique physical properties originating added to the models consecutively in the calculations. Mod-
from their unique structures have displayed bright applica- els were supposed to have perfect atomic structures. Extra
tion perspectives. Carbon nanotubes are expected to be one electrons might cause atomic structure relaxation. Yet calcu-
of the hopeful materials of the next-generation electronic lations of Changwook Kim et al. showed that capped tubes
devices,3 and now they have already manifested their advan- withstand field strengths up to 2 V/Å 共Ref. 7兲 and Lou,
tage as an efficient electron source through their excellent Nordlander, and Smalley have used perfect SWNT cluster
field-emission properties.4 Theoretical studies and experi- models to study their electronic properties in strong electrical
mental researches stimulated calculations of their electronic fields with maximum local field strength up to 2 V/Å 共Ref.
structures such as energy bands, density of states, etc., which 6兲. In our calculations, the number of extra electrons is small
have long since been accomplished, and now an extensive 共at most eight兲 compared with the total electron number of
literature exists.5 Their electronic structures in the external the neutral models 共more than one thousand兲 and so is the
electrical fields have also been calculated theoretically.6,7 variation of the total electronic charge density after adding
Other researchers studied carbon nanotubes with defects,8 the extra electrons. At the same time, the maximum electrical
deformation,9 and finite length.10 But charged carbon nano- field strength generated by the extra electrons is smaller than
tubes seem to have been less examined, especially the effects
of extra charges on their electronic structures. Whereas
charged carbon nanotubes may play a key role in the electron
tunneling of the future electronic devices. Experimentally,
short carbon nanotubes could be used as Coulomb islands or
quantum dots in single-electron devices, where they may be
reasonably supposed to carry extra charges.11 Charged car-
bon nanotubes may also provide a possible theoretical model
when their field-emission properties are investigated. In the
situation of field emission, strong electrical field exists
around the nanotube tip, and this field is accompanied by the
accumulation of negative charges, or, the field itself should
be the result of this charge distribution. Hence the potential
distribution near the charged carbon nanotubes and their
electronic structures might provide us some understanding
about their emission behaviors in some aspects.
In this paper, we report our density-functional-theory
共DFT兲 calculations of the electronic structures of charged
single-walled carbon nanotubes 共SWNT’s兲. The cluster mod- FIG. 1. Single-walled carbon nanotube 共SWNT兲 models. 共a兲
els for SWNT’s are shown in Fig. 1: Model one is a 共5,5兲 Model 1: SWNT 共5,5兲 capped at two ends with semispheres of a C60
armchair SWNT capped at two ends with semispheres of a molecule; it has 190 carbon atoms. 共b兲 Model 2: Open-ended
C60 molecule and containing 190 carbon atoms; model two is SWNT 共10,0兲 containing 200 carbon atoms.

0163-1829/2002/66共11兲/115415共5兲/$20.00 66 115415-1 ©2002 The American Physical Society

JI LUO, L.-M. PENG, Z. Q. XUE, AND J. L. WU PHYSICAL REVIEW B 66, 115415 共2002兲

2 V/Å. Thus as an approximation the effects of atomic relax-

ation have been neglected and the atomic structures of these
models were considered as not varying with extra electrons
in our calculations.
Calculations were carried out with the Amsterdam density
functional 共ADF兲 program which is based on the Kohn-Sham
approach to the density-functional theory.12–14 In the calcu-
lations, double-␨ Slater-type 2s and 2p basis functions were
used for the valance electrons of carbon atoms and their 1s
electrons were kept frozen, which led to fairly reasonable
results and at the same time limited the computational costs.
The local-density approximation 共LDA兲 for the exchange
and correlation energy given by Vosko, Wilk, and Nusair15
was applied to the calculation of the potential during self-
consistent field iterations and after that, general gradient ap-
proximation method suggested by Becke16 and Perdew17 was FIG. 2. Total electronic charge density of a four-electron-
used to calculate the energy. To evaluate the reliability of our charged 共6,0兲 SWNT containing 96 atoms. The section passes
calculations with ADF package, we used this program to through the tube axis and a series of carbon atoms. Contour plots
calculate the electronic structures of a single C60 molecule. are from ADF program 共a兲 and from CASTEP program 共b兲. Contours
The calculation gives its highest occupied molecular-orbital of values from 0.02 to 3.98 a.u. with a step of 0.04 a.u. are pre-
共HOMO兲 energy as ⫺7.83 eV, leading to a 7.83 eV ioniza- sented in each plot. 共c兲 is a gray-scale graph of the difference be-
tion potential, in good agreement with that reported in the tween the charge density of the charged tube and that of the neutral
literature.6,18 one, from ADF program. Difference ranging from ⫺0.01 共the
The charged nanotube cluster model carrying n extra elec- brightest area兲 to ⫹0.01 a.u. 共the darkest area兲 is presented and the
trons was treated as a molecular system with M fixed carbon gray background indicates zero difference.
nuclei, 4M ⫹n valence electrons, and 2M frozen core elec-
trons. Total electronic charge density 共we will refer to it sim- ⫻22 Å 3 unit cell. CASTEP calculation with a cutoff energy of
ply as charge density hereafter兲 and electrostatic potential of 310 eV was carried out on this supercell structure. The
this molecular system were calculated on the grid points in a charge density of this tube carrying four extra electrons cal-
chosen volume. The electrostatic potential is defined in culated with the ADF program and with CASTEP program is
illustrated in Figs. 2共a兲 and 2共b兲 respectively. Here the four
atomic units as
extra electrons roughly correspond to eight electrons for the

兺a 兩 r⫺Ra a兩 ⫹ 冕 兩 r⫺r⬘兩 dr⬘ ,

larger model 1 and model 2 in Fig. 1, that is, one extra
Z n 共 r⬘ 兲 electron for every 25 carbon atoms. The two methods gave
V 共 r兲 ⫽⫺ 共1兲
very similar charge density contour plots except in the re-
gions near the nuclei where the plane-wave method could not
where n(r) is the charge density, while Ra and Z a denote the give correct values because it employs pseudopotential and
position and the charge number of the ath nucleus, respec- where the values from the ADF program are not presented in
tively. V(r) can be explained as the electrostatic potential of the plot. Both results show that the charge density has large
an electron at the place r in the electrical field of all the values near the tube wall where carbon atoms are located.
nuclei and the charge density of the molecule, assuming that Figure 2共c兲 shows the charge-density variation after four ex-
the molecule is not polarized by the test electron. Thus tra electrons are added to the neutral tube, calculated with the
“V(r) gives the electrical field. V(r) is also found not to be ADF program. Instead of simply increasing its density, the
strongly affected by the inclusion of electron correlation.19 electronic charge rearranges itself such that in some regions
As an approximation, this quantity can be viewed as the the charge density increases and in some regions it decreases.
one-particle potential that an electron of the molecule system But the variation is small and mainly takes place near the
feels. Finally, energies of the molecular orbitals were calcu- tube wall. Results in Figs. 2共a兲 and 2共b兲 qualify the local-
lated as eigenvalues of the self-consistent Fock matrix, and orbital method in the calculations of the SWNT models with
the total energy of the charged model was also worked out small extra charges. For the larger SWNT’s 共model 1 and
relative to the neutral model. model 2 in Fig. 1兲, calculations with the more efficient ADF
Since we are investigating charged systems, there arises program gave the charge density of the similar features.
the question of the adequacy of the localized basis set used in Figure 3 illustrates the electrostatic potential contour plots
the ADF program. Because in this situation, the electrons in a vertical section of the charged SWNT models. Far away
could relax into states of more extended character and the from the models, the potential is very much like that gener-
extra charges could spatially occupy regions far from the ated by a negative point charge. In the interior of the models,
tube wall. To clarify this problem, a comparative calculation electrostatic potential is approximately uniform, showing an
of a small 共6,0兲 SWNT containing 96 carbon atoms was car- isopotential region with zero field. In the regions near the
ried out with the ADF program and with a plane-wave-based tube wall, structures of the potential distribution could be
DFT program CASTEP.20 After structure relaxation with a found. For instance, closed curves indicate the existence of
semiempirical method, the tube was put into a 12⫻12 extreme values. For more information, potentials on the axis

115415-2
DENSITY-FUNCTIONAL-THEORY CALCULATIONS OF . . . PHYSICAL REVIEW B 66, 115415 共2002兲

FIG. 3. Electrostatic potential contour plots of 共a兲 eight-

electron-charged model 1 and 共b兲 six-electron-charged model 2. In
共a兲, the section passes through the model axis and one of the carbon
atoms in the end pentagon; in 共b兲, it passes through the model axis
and a series of carbon atoms in the tube wall. See Fig. 1 for the
SWNT models.

of charged model 1 are plotted in Fig. 4共a兲. In these plots,

potential wells correspond to the places of the semisphere
caps of the tube, and charge density has large values in these
wells. In the interior of the model, potentials on the axis
show no variation. While in Fig. 4共b兲, potentials on a line in
the wall of model 2 are illustrated. They exhibit the charac-
FIG. 4. Electrostatic potentials on the axis of charged model 1
teristics of a typical periodic lattice potential, with the val-
共a兲 and on a line in the wall of charged model 2 共b兲. Long-dashed
leys corresponding to the places of the carbon nuclei, the
lines illustrate the potentials of perfect conductors, and line seg-
higher peaks corresponding to the centers of the hexagons,
ments represent HOMO energy. The numbers of extra electrons are
and the lower peaks corresponding to the middle points of marked in the plots.
the carbon-carbon bonds. In both cases 关Figs. 4共a兲 and 4共b兲兴,
when extra electrons are added, the potential between the for electrons in such orbitals to tunnel through the barrier
tube ends is lifted and its values at infinity remain zero, and escape. We use simple WKB method to estimate
hence potential barriers are formed at the tube ends. The the transmission coefficient 共in atomic units兲: T
height of the barriers increases linearly while their width at ⫽exp关⫺2兰ba冑2(V(z)⫺E)dz 兴 where E corresponds to the
HOMO energy decreases with the number of extra electrons. HOMO energy, a and b are the abscissas of the interesting
Potential remains high in the interior except for the neutral points of the straight line E and the potential curve V(z). 22
model 共in this case it is slightly lower in the interior than at For model 1 with eight extra electrons, transmission coeffi-
infinity兲. Dividing the extra charge by the interior potential, cient at the HOMO energy is about 0.004, and the corre-
we estimate the capacitance of the SWNT cluster models as sponding barrier width is about 4 Å.
about 9⫻10⫺20 F. It is interesting that the height of the po- In the potential wells, strong electrical fields exist to con-
tential barrier, the interior potential, and the HOMO energy fine electrons. After extra electrons are added, there appears
all increase 共approximately兲 linearly with the number of ex- another one near but outside the tube wall, which tends to
tra electrons; slopes of these lines give roughly the same pull electrons out of the model. Strength of this field also
value: the reciprocal of the capacitance of the cluster.21 This increases with the number of extra electrons. This is similar
to the case of field emission. Figure 5共b兲 shows the maxi-
is shown in Fig. 5共b兲. In formula, the HOMO energy E n of
mum strength of this electrical field for model 1. In fact,
the n-electron-charged tube could be expressed as 共in SI兲
several 共at most eight兲 extra electrons will lead to the field
strength near the tube tip as strong as 1.4 V/Å, comparable
e2 with that for the field emission, which is of the order of
E n⫽ n⫹E 0 , 共2兲
C 107 V/cm. When electrons of the system have relatively
large probabilities to escape, and the electrical field near the
where C denotes the capacitance, E 0 is the HOMO energy of tube is strong enough, calculations might become inappropri-
the neutral tube, and e is the fundamental charge. ate for not including the extended basis functions and for the
Since the HOMO energy increases linearly with the num- possible unstability of the tube’s atomic structures. Thus
ber of extra electrons and four extra electrons are enough not more than eight extra electrons were considered in our
to lift it from about ⫺7 eV to above zero, there are chances calculations.

115415-3
JI LUO, L.-M. PENG, Z. Q. XUE, AND J. L. WU PHYSICAL REVIEW B 66, 115415 共2002兲

charged SWNT cluster models manifested the HOMO en-

ergy increase, and demonstrated the linearity of this increase
with respect to the number of extra electrons. However, the
total energy of the charged SWNT’s exhibits more complex
variations. For both the tube models in our calculations, their
total energy first decreases and then increases with the addi-
tion of extra electrons, indicating an optimum electron num-
ber that corresponds to the minimum energy. This is illus-
trated in Fig. 5共a兲. The variation can be simulated very well
with parabolic curves. The parabola for model 1 can be ex-
pressed as 共in eV兲

E 共 n 兲 ⫽0.99n 2 ⫺5.76n⫺0.03, 共3兲

where E(n) denotes the total energy of the n-electron-

charged model relative to that of the neutral model. This
leads to the energy difference when successive electrons are
added as 共in eV兲

E 共 n 兲 ⫺E 共 n⫺1 兲 ⫽1.98n⫺6.75. 共4兲

On the other hand, the linear fit of the line in Fig. 5共b兲 gives
the HOMO energy of model 1 to be 共also in eV兲

E n ⫽2.02n⫺6.74. 共5兲

Thus roughly E n ⫽E(n)⫺E(n⫺1), the HOMO energy of

the n-electron-charged tube is the energy needed to add an
FIG. 5. Parabolic total-energy variations of the SWNT’s with electron to the (n⫺1)-electron-charged tube. This is Koop-
respect to the number of extra electrons 共a兲 and linearity of some mans’ theorem, which can be proved in the Hartree-Fock
properties of model one 共b兲. The slope of the lines in 共b兲 equals to method.24 Here it also holds in our DFT calculations of
the reciprocal of the cluster’s capacitance. For the SWNT models charged SWNT models. In view of formula 共2兲 we get 共in SI兲

冉 冊
see Fig. 1.
e2 2 e2
In Ref. 6, the authors show that SWNT’s can screen the E共 n 兲⫽ n ⫹ ⫹E 0 n, 共6兲
2C 2C
external field, demonstrating their similarity to perfect con-
ductors. Our calculations indicate the approximate zero elec- where C, E 0 , and e have the same meaning as in Eq. 共2兲. It
trical field in the interior of the charged SWNT’s, manifest- is interesting that Eq. 共6兲 might also be valid for longer
ing the same resemblance. For a detailed comparison SWNT’s and at least for metallic SWNT’s E 0 could be sub-
between the charged SWNT’s and the structureless perfect stituted by their work function: E 0 ⫽⫺ ␾ . Thus formula 共6兲
conductors of the same dimensions, we calculated the elec- expresses the DFT total-energy variation in terms of classical
trostatic potential of the latter by applying laws of classical parameters. The regularities in HOMO level change and
electrostatics and numerically solving Laplacian equation. total-energy variation are the results of the electronic charge
Figure 4共a兲 contrasts the axis potential of the eight-electron- rearrangement after the addition of extra electrons because
charged model 共model 1兲 with that of its perfect conductor according to DFT, all the properties of a system in its ground
counterpart, while in Fig. 4共b兲 a similar comparison is made state are determined by the charge density. Since the charge
on a line in the wall of the six-electron-charged model distribution in the SWNT is different from that in the struc-
共model 2兲. About 5 Å away from the tube ends, potentials tureless conductor, semiclassical theory could not lead to
match perfectly, while near the tube, potentials show consid- these regularities. Hence for SWNT’s quantum mechanics
erable difference. The potential in the perfect conductor is calculations might be necessary in order to investigate their
several eV higher than that in the SWNT, and this leads to a electronic properties exactly.
smaller capacitance for the perfect conductor than for the In conclusion, we conducted DFT calculations of the elec-
SWNT (6.9⫻10⫺20 F vs 8.7⫻10⫺20 F for model 1兲. More- tronic structures of charged SWNT cluster models. Emphasis
over, in the perfect conductors, the potential distribution was placed on their electrostatic potential distribution and
structures near the SWNT walls such as potential valleys and their energy variations with respect to the number of extra
potential peaks all disappear due to their structurelessness. electrons. A comparison between the SWNT models and the
In the semiclassical theory of the single-electron tunnel- structureless perfect conductors demonstrates that quantum
ing, when an extra electron is added, a quantum dot gets an mechanics calculations are indispensable for understanding
energy increase in the form of the charging energy, thus its the exact properties of the charged SWNT’s. Similarly, the
HOMO level will be lifted.23 Our DFT calculations on the semiclassical theory of the single-electron tunneling of quan-

115415-4
DENSITY-FUNCTIONAL-THEORY CALCULATIONS OF . . . PHYSICAL REVIEW B 66, 115415 共2002兲

tum dots is insufficient for deriving energy variations of the This work was supported partly by the MOST of China,
charged SWNT’s due to its incapability of describing elec- the National Natural Science Foundation of China, and the
tronic charge rearrangement in the SWNT models when ex- Key Project in Scientific and Technological Research of the
tra electrons are added. Ministry of Education of China.

13
*Corresponding author. Email address: jlwu@pku.edu.cn ADF-2.3.0,
Theoretical Chemistry, Vrije Universiteit, Amsterdam.
1
S. Iijima, Nature 共London兲 354, 56 共1991兲. 14
G. te Velde and E. J. Baerends, J. Comput. Phys. 99, 84 共1992兲.
2
S. Iijima and T. Ichihashi, Nature 共London兲 363, 603 共1993兲; D. 15
S. H. Vosko, L. Wilk, and M. Nusair, Can. J. Phys. 58, 1200
S. Bethune, C. H. Kiang, M. S. Devries, G. Gorman, R. Savoy, 共1980兲.
J. Vazquez, and A. Beyers, ibid. 363, 605 共1993兲. 16
A. D. Becke, Phys. Rev. A 38, 3098 共1988兲.
3
C. Dekker, Phys. Today 52共5兲, 22 共1999兲. 17
J. P. Perdew, Phys. Rev. B 33, 8822 共1986兲.
4
J. M. Bonard, J. P. Salvetat, T. Stockli, L. Forro, and A. Chatelain, 18
M. S. Dresselhaus, G. Dresselhaus, and P. C. Eklund, Science of
Appl. Phys. A: Mater. Sci. Process. A69, 245 共1999兲. Fullerenes and Carbon Nanotubes 共Academic, New York,
5
R. Saito, G. Dresselhause, and M. S. Dresselhause, Physical 1996兲.
Properties of Carbon Nanotubes 共Imperial College, London, 19
Michael Springborg, Density-Functional Methods in Chemistry
1998兲.
6 and Materials Science 共Wiley, New York, 1997兲.
L. Lou, P. Nordlander, and R. E. Smalley, Phys. Rev. B 52, 1429 20
Cambridge Serial Total Energy Package, version 4.2.
共1995兲. 21
7 Because the line representing the interior potential in Fig. 5共b兲
Changwook Kim, Bongsoo Kim, Seung Mi Lee, Chulsu Jo, and
does not pass through the origin exactly, the tube capacitance
Young Hee Lee, Appl. Phys. Lett. 79, 1187 共2001兲.
8 derived from its slope is slightly different from the result of
J.-C. Charlier, T. W. Ebbesen, and Ph. Lambin, Phys. Rev. B 53,
11 108 共1996兲. dividing the extra change by the interior potential; this is a prob-
9
A. Rochefort, P. Avouris, F. Lesage, and D. R. Salahub, Phys. lem of how we define the capacitance of the tube.
Rev. B 60, 13 824 共1999兲.
22
David Bohm, Quantum Theory 共Prentice-Hall, New York, 1951兲.
23
10
A. Rochefort, D. R. Salahub, and P. Avouris, J. Phys. Chem. B David K. Ferry and Stephen M. Goodnick, Transport in Nano-
103, 641 共1999兲. structures 共Cambridge University Press, Cambridge, UK, 1997兲;
11
Marc Bockrath, David H. Cobden, Paul L. McEuen, Nasreen G. Marc A. Kastner, Phys. Today 46共1兲, 24 共1993兲.
24
Chopra, A. Zettl, Andreas Thess, and R. E. Smalley, Science Attila Szabo and Neil S. Ostlund, Modern Quantum Chemistry:
275, 1922 共1997兲. Introduction to Advanced Electronic Structure Theory
12
W. Kohn and L. J. Sham, Phys. Rev. 140, A1133 共1965兲. 共McGraw-Hill, New York, 1989兲.

115415-5