Introduction to Dilute Magnetic Semiconductors

Nan Zheng∗
Course: Solid Sate II, Instructor: Ebio Dagotto, Semester: Spring 2008,
Department of Physics and Astronomy, The University of Tennessee Knoxville
(Dated: April 13, 2008)
This paper begins with a brief introduction to the field of spintronics and dilute magnetic semi-
conductor (DMS), the second part discusses in detail on three typical DMS materials–(Ga,Mn)As,
(Ga,Mn)N and Transition metal doped oxide. Next mechanism of DMS ferromagnetism along with
its magneto transport properties are discussed. Following that, an important sample preparation
technique in DMS– Molecular Beam Epitaxial (MBE) – will be presented. Finally, conclusion and
outlook are made towards the future study on DMS.

PACS numbers:

I. BRIEF INTRODUCTION: II. DMS MATERIALS

DMS AND SPINTRONICS
Two major criteria are considered to select the most
promising materials for semiconductor spintronics. First,
the ferromagnetism should be retained to practical tem-
Since late 1980s, people have noticed that in many peratures namely room temperature. Second, it would
semiconductor crystals, substitution of a transition metal be a major advantage if there were already an existing
element for a host element adds local magnetic moments technology base for the material in other applications.
to the systems’s low-energy degrees of freedom[1, 2].
As early as late 1960 to early 1970, Oxide doped with
These doped materials are known as dilute magnetic
Eu2+ [3] and spinel structured composite (for example,
semiconductors (DMSs), where usually a transition el-
ZnCr2 Se4 [4]) are studied as magnetic semiconductors.
ement is substituted on a small fraction, x, of a host
However, structures of those composites are different
semiconductor element sites. The study of DMS became
from Si or GaAs, the crystals are very hard to produce
flourished in recent ten years since high quality samples
in experiment, their low Curie temperature Tc (50K or
are available through experiment. It is widely believed
lower), strong insulation and poor semiconducting trans-
that DMS are ideal material for spintronics. The fol-
port property[5] further hampered their value in applica-
lowing paragraph will explain what is spintronics and its
tion.
applications.
Later on, studies have been spreaded on dilute mag-
netic semiconductors including transition metal (mainly
The term ”spintronics” stands for spin transition elec- Mn) doped II-VI, IV-VI and II-V compound semiconduc-
tronics. As well known today, integrated circuits and tors, typical examples are: II-VI: (Zn,Mn)Se, (Cd,Co)Se,
high-frequency devices, used for information processing (Hg,Fe)Te; IV-VI: (Sn,Mn)Te, (Pb,Mn)Te, (Pb,Eu)Te,
and communications, have had great success through etc. Mn doped II-VI semiconductors are especially fo-
controlling the charge of electrons in semiconductors. cused on, typical materials are (Zn,Mn)Se etc.
Mass storage of information – indispensable for informa- For a long time, due to much lower solubility of mag-
tion technology – is carried out by magnetic recording netic ions in III-V semiconductors compared to II-VI
(hard disks, magnetic disks...) using electronic spins in semiconductors, along with its poor stability, not much
ferromagnetic materials. It is then quite natural to ask if studies are done on III-V DMSs. A breakthrough was
both the charge and spin of electrons can be used at the made by using molecular beam epitaxy (MBE), a thin-
same time to enhance the performance of devices. This film growth technique in vacuum that allows one to work
is the main idea of spintronics, which is widely expected far from equilibrium. Using MBE technique, Munekata
to be the future solution to downsize current microelec- etc. successfully made III-V material InMnAs, and fer-
tronic devices into size of even nanometers. The realiza- romagnetism was observed in p type InMnAs[2]. Based
tion of functional spintronic devices requires materials on the work above, in 1996, Ohno etc. made the first
with ferromagnetic ordering at operational temperatures Mn doped dilute magnetic semiconductor (Ga,Mn)As[6].
compatible with existing semiconductor materials. Be- This material was grown as a Ga1−x Mnx As(x = 0.015 ∼
ing a ferromagnetic semiconductor with favorable exper- 0.017) thin film on a semi-insulated GaAs(001) substrate.
imental properties, dilute magnetic semiconductors will The fraction of Mn ions in the sample was as high as
promisingly suit this need. (3 ∼ 7) × 1020 cm−1 , which was far higher than the sol-
ubility of GaAs in thermal equilibrium. Furthermore,
because that (Ga,Mn)As were grown on GaAs film, its
crystal structure has a good similarity to GaAs. Due to
∗ Electronic address: nzheng@utk.edu its remarkable properties, (Ga,Mn)As attracted a lot of
 2

FIG. 1: Computed values of Curie temperatures Tc for various

FIG. 2: Curie temperature vs. Mn concentration for DMS
p-type semiconductors containing 5% of Mn per cation (2.5%
(Ga,Mn)As[14]
per atom) and 3.5 × 1020 holes per cm3 [12, 13]

interests on DMS studies as soon as it appeared in the again to non-metal[15].

experiment. Study of origin of ferromagnetism in (Ga,Mn)As has
Typically the observed curie temperature of DMS is always attracted lots of attentions. H. Ohno etc be-
mostly below 50K, the highest Tc on (Ga,Mn)As is only lieves that it is hole charges who lead to ferromagnetism
110K. So an important step for DMS to be applied in in (Ga,Mn)As[6]. They discovered that the number of
reality is to improve its Tc . Many new DMS materi- Mn ions in (Ga,Mn)As is of the same order compared
als are discovered in recent research, for example Mn to those of hole charges, which supported their opin-
doped CdGeP2 [7], CrAs[8], (Ti,Co)O2 [9], (Zn,Co)O[10] ion. Nowadays, it is still not fully understood the rea-
and (Zn,Ni)O[11], etc. In 2000, Dietl explained curie son of ferromagnetism in (Ga,Mn)As. Based on the
temperature in Ga1−x Mnx As and Zn1−x Mnx Te theoret- study of (Ga,Mn)As, Other DMS materials with even
ically using Zener Model, and predicted the existence of higher Curie temperature has been discovered, for exam-
room temperature DMSs, as in Fig. 1[12]. ple, (Ga,Mn)N, Co: TiO2 etc. However, they are all not
Recent research are mainly focused on the following as good as (Ga,Mn)As in terms of experimental capabil-
DMS materials: ity and compatibility with current semiconductor indus-
try. Till now, (Ga,Mn)As is the most promising DMS
material in practice.
A. (Ga,Mn)As

GaAs are already in use in a wide variety of electronic B. (Ga,Mn)N

equipment in the form of electronic and optoelectronic
devices, including cellular phones, compact disks, and in In Dietl’s theoretical calculation, Curie temperature of
many other applications. Therefore, the introduction of Mn doped GaN is the highest among various semicon-
magnetic semiconductors based on GaAs opens up the ducting compound, thus (Ga,Mn)N has a wide spread
possibility of using a variety of magnetic phenomena not interests in researchers. In 2001, Zajat etc made highly
present in conventional nonmagnetic GaAs material in doped (Ga,Mn)N using an ammonothermal method,
the optical and electrical devices already established. which has paramagnetism, and they believe that p type
In 1996, Ohno firstly used MBE technique to produce (Ga,Mn)N can gain ferromagnetism[16]. In the same
Ga1−x Mnx As thin film with a largely enhanced solubil- year, Reed etc reported that they grew a 2µm thick single
ity of the order of 1020 cm−1 . Through study of its mag- crystal on (0001) sapphire substrates using metal organic
netic transportation, it was found that the Curie tem- chemical vapor deposition (MOCVD) method, and then
perature varied as a function of doping factor x, and a laser deposition technique was used to deposit Mn on
its behavior satisfied Tc = 2000x ± 10K[6]. Later F. the Nitride samples[17]. Mn doping was achieved by de-
Matsukura etc discovered that the Curie temperature of position and annealing at different temperatures ranging
(Ga,Mn)As reached its highest 110K when x = 0.05, from 250–800◦ C. The experiment implies that Curie tem-
as shown in Fig. 2. When Mn concentration was re- perature of grown samples will depend on growing and
duced, Curie temperature Tc would also decrease, when annealing conditions, which are between 220∼370K[18].
Mn concentration x went below 0.005, ferromagnetism Theodoropoulou etc used a similar method and achieved
would disappear[14]. Moreover, as Mn concentration in Mn doped P type GaN, with a high dose of 3∼5%. The
(Ga,Mn)As increases, the transportation properties ex- observed Curie temperature is lower than theoretical pre-
perienced series of changes, i.e., non-metal to metal and dictions, the author believes it is due to lower concentra-
 3

FIG. 4: Room temperature hysteresis for bulk ZnO(Sn) im-

planted with 3 at.% Mn[1]
FIG. 3: Magnetization vs. temperature for (Ga,Mn)N sample
grown by MBE with ∼9 at.% Mn. The extrapolation of the
curve is based on a mean-field approximation[20, 21] Co doped TiO2 is also observed to have room tem-
perature ferromagnetism. In 2001, Matsumoto reported
that Co:TiO2 (doped Co below 5%) film made by
tion of holes[19]. Fig. 3 shows the temperature depen- PLD displays room temperature ferromagnetism, and its
dence of the magnetization for a sample with 9 at.% Mn, Curie temperature Tc > 400K[9]. Another experiment
yielding an estimated Tc of 940K using a mean-field ap- achieved greater than 300K Curie temperatue in the same
proximation. material but made by oxygen plasma assisted molecular
beam epitaxy (OPA-MBE)[27]. Later on, Co doped TiO2
film with high Curie temperature were also achieved by
C. Transition metal doped oxide several other experiment on various substrates[28–30].
Mn doped SnO2 is predicted as another DMS. H.
Semiconducting oxide has wide energy gap, good light Kimura discovered magnetoresistance in Mn doped SnO2
transmission, can be doped to get n type semiconductor. at low temperature 5K[31].
These preferable properties made them widely used in
practice[22]. Thus, there are also many ongoing research
in oxide DMS, which is now mainly focused on transition III. MOLECULAR BEAM EPITAXIAL
metal doped ZnO and TiO2 materials. TECHNIQUE
In 1999, Fukumura etc produced Mn doped ZnO thin
film using Pulsed Laser Deposition (PLD). ZnO and The major obstacle in making III-V semiconductors
Mn3 O4 are heated at 900◦ C and mixed with a certain magnetic has been the low solubility of magnetic elements
percentage for 12 hours. Concentration of electron car- (such as Mn) in the compounds. Because the magnetic
riers is of the order of 1019 cm−3 . In the grown sam- effects are roughly proportional to the concentration of
ple Zn1−x Mnx O, considerable magnetoresistance is ob- the magnetic ions, one would not expect a major change
served at low temperatures[23]. Two years later, Sato in properties with limited solubility of magnetic impuri-
etc proved theoretically that Mn, Fe, Co and Ni doped ties, of the order of 1018 cm−3 or less.
ZnO DMSs may have stable ferromagnetism[24]. Ueda A breakthrough was made by using molecular beam
etc reported ferromagnetism in Co doped ZnO thin film epitaxy (MBE), a thin-film growth technique in vacuum
made by PLD method, but only with a low repeatability that allows one to work far from equilibrium. When a
(about 10%)[10]. Jin et al produced Co doped ZnO using high concentration of magnetic elements is introduced
laser molecular-beam epitaxy (MBE), but no ferromag- in excess of the solubility limit, formation of the second
netism was observed[25]. Cho et al produced CoFe doped phase occurs if conditions are near equilibrium. How-
ZnO – [Zn1−x (Co0.5 Fe0.5 )x O] thin film – using reactive ever, when crystal is grown at low temperature by MBE,
magnetron co-sputtering and fast annealing. The film there is not enough thermal energy available to form the
sample achieved has x ≥ 0.15, and has the same crys- second phase, and yet there still exists a local potential
tal structure of ZnO. It is indicated that annealing can landscape that allows epitaxial growth of a single-crystal
greatly enhance self magnetization and carrier concentra- alloy. The effort to grow new III-V based DMSs by low-
tion of samples, thus increase Curie temperature to even temperature MBE was rewarded with successful epitaxial
room temperature[26]. Fig. 4 shows the magnetization growth of uniform (In,Mn)As films on GaAs substrates in
behavior at 300K for a Sn-doped ZnO sample implanted 1989, where partial ferromagnetic order was found, and
with 3 at.% Mn, clearly indicating ferromagnetism[1]. ferromagnetic (Ga,Mn)As in 1996.
 4

FIG. 5: Material deposition process in MBE[32]. FIG. 6: Illustration of a MBE construction[32].

IV. MAGNETIC MECHANISM AND

In a ultra-high vacuum, a beam of atoms or, more gen- PHYSICAL PROPERTIES
eral, a beam of molecules is directed towards a crystalline
substrate such that the atoms or molecules stick at the A. Exchange interaction
substrates surface forming a new layer of deposited ma-
terial, as shown in Fig. 5. The difference between MBE
and other material deposition methods as e.g. thermal When magnetic ions are doped into semiconductors,
vacuum evaporation is as follows. MBE does not only they will substitute part of positive ions, their local
deposit material like it is done by conventional evapora- spin momentum has strong spin-spin exchange interac-
tion techniques, but using the very low rates of impinging tion with carriers, which further will lead to different be-
atoms, migration on the surface and subsequent surface havior of charge carriers, thus make DMS different from
reactions ensure the controlled epitaxial growth of a new normal semiconductors. Spin-spin exchange interaction
layer. Simply spoken: Every atom reaching the surface of is the key point through which various magnetic moments
the heated substrate has enough time to migrate around are formed and DMS materials differ from non-magnetic
and find his place to build up a new crystal lattice. With materials. In DMS, exchange interaction include interac-
lots of physics applied, chemistry and engineering work tions between s electrons in conducting band, p electrons
done since beginning MBE in the late 60s and the early in valence band and d electrons from magnetic ions (sp-d
70s to reach state of the art MBE. exchange interaction) and interactions between d elec-
trons from magnetic ions(d-d exchange interaction)[33].
The typical construction of a MBE is shown in Fig. 6.
Soaleck etc indicated that it is the distance between
At the left hand side there are the effusion cells to provide
nearest Mn2+ ions that determines the exchange integral
the molecular beam for either the bulk constituents or
factor after analyzed many experimental results. Experi-
the dopants. These cells can be thermal evaporation cells
ments show that exchange interaction between magnetic
(Knudsen cells), cells for gaseous media or plasma sources
ions is due to superexchange mediated by negative ions
as well. In front of them is a shutter, this means a plate
in distorted crystal lattice[33].
which could be brought into the beam for switching the
beam on and off. Opposite to the cells and the shutters Fig. 7 shows the operative mechanisms for magnetic
is the substrate, mounted on a heatable and rotatable ordering in Mn doped DMS (Ga,Mn)As. There are two
substrate holder. basic approaches to understand the magnetic properties
of dilute magnetic semiconductors. The first one is based
The whole system is in ultra-high vacuum environment on mean field theory which originates in the model of
to guarantee formation of a molecular beam (without vac- Zener[34]. The theories that fall into this general model
uum, the atoms or molecules leaving the effusion cells will implicitly assume that the dilute magnetic semiconduc-
be scattered at residual gas molecules and never form a tor is a more-or-less random alloy, in another word, the
beam directed towards substrate) and purity and there- doping atoms will substitute randomly for the lattice con-
fore quality of the grown layer. Often there are instru- stituents. Within these theories, there are differences in
ments for in-situ analysis like RHEED attached to the how the free carriers are assumed to interact, as shown in
growth chamber. A cryopanel around the sample and Fig. 8. The second class of approaches suggests that the
the cells absorbs residual gases and provides a clean sub- magnetic atoms form small (a few atoms) clusters that
strate environment. produce the observed ferromagnetism[35].
 5

FIG. 9: Magnetotransport data from GaMnN grown at TG =

925◦ C on a GaN buffer[1]

B. Magnetotransport Properties
FIG. 7: Illustration of carrier mediated ferromagnetism in di-
lute magnetic semiconductor (Ga,Mn)As. Mn2+ ions sit on 1. Negative Magnetoresistance
trivalent Ga sites (triangles), and therefore act as electron
acceptors (providing holes) as well as producing a magnetic When magnetic ions are doped into semiconductor to
moment (arrows). form DMS, spin from magnetic ions can capture carri-
(A) The holes are thought to mediate ferromagnetic coupling
ers into its ferromagnetic spin cluster, due to spin-spin
between the magnetic Mn2+ ions.
(B) Below certain temperatures, a percolation network is exchange coupling between carriers and magnetic ions.
formed in which clusters of the holes are delocalized and hop The above procedure produces so called bounded mag-
from site to site, which is an effective mechanism for aligning netic polarizer (BMP). However, with increased external
Mn moments within the cluster network. field, BMP will be destroyed more and more, thus lead
to more and more leased carriers to conduct. Therefore,
DMS will display a negative magnetoresistance at low
temperature.
H. Ohno studied Ga1−x Mnx As DMS. The samples dis-
played as both metal and insulator depend on Mn dose.
He discovered that for metallic samples, negative magne-
toresistance will intense with decreased temperature, and
reached its highest at Curie temperature Tc . While for
insulating samples, negative magnetoresistance was still
stronger even with temperature lower than Tc , and ex-
ternal magnetic field effected strongly on sample’s mag-
netoresistance at low temperatures[6]. Fig. 9 shows clear
negative magnetoresistance (11% at 10K) on a GaMnN
sheet.

2. Abnormal Hall Effect

Abnormal Hall effect is observed in DMS materials

with external magnetic field, which is another important
FIG. 8: Schematic of role of carriers in two scenarios for characteristic fact on DMS magnetization. This effect
carrier-induced ferromagnetism in dilute magnetic III-V semi- can be expressed by the following relation:
conductors.
Scenario 1: Delocalized holes form a Fermi liquid. Coulomb R0 RS
effects are approximated. RHall = B+ M, (1)
Scenario 2: Holes form an almost localized tight-binding d d
band. Fluctuation effects due to randomness and hole-hole where R0 is normal Hall Effect factor, RS is abnormal
correlations are approximated. Hall Effect factor. It is observed that RS is proportional
 6

to surface Hall resistance Rsheet , i. e., RS = dcRsheet , vides the possibility to manipulate electron charge and
where c is a constant independent of temperature. spin at the same time, and opens a new field in
semiconducting technology to invent new electronic de-
vices. Although research on DMS application is still
3. Enhanced Magnetophotonic Effect at the stage of exploration, this field has already
present its wide spread future in practice. For ex-
Enhanced magnetophotonic effect is another unique ample, inserting DMS material into magnetic metal
property of DMS, angle change of polarization plane and semiconductor, one can realize ejection of carriers
(Faraday angle) can display magnitude of interactions with spin polarization into non-magnetic semiconduc-
between d, p and s electrons. Theoretical analysis indi- tor, and this technique can be applied to make spin
cates that, Faraday angle θF can be expressed as: polarized light-emitting diode[36]. While for certain
√ ferromagnetic/non-ferromagnetic multi-layer structures,
F0 β − α h2 ω 2 like GaMnAs/AlGaAs/GaMnAs[37] etc, by tuning tem-
θF ≈ M 2 l, (2) perature and voltage, one can control carrier concentra-
2hc gMn µB (Eg − h2 ω 2 )3/2
tion in semiconducting layer and magnetic coupling be-
where F0 is a constant, l is sample thickness, hω is tween magnetic layers, this property can be used to pro-
the energy of incoming photons, Eg is energy gap, M duce new magnetic-controled and photonic-controled de-
is magnetic intensity of the sample. Here, the value of vices.
β − α will vary according to different doping percentage
of composite atoms.
Experiment implies that Faraday Rotation Effect is
much stronger in III-V DMS like (Ga,Mn)As than the
original compound GaAs, with a 6 × 104 deg/cm Faraday
Rotation at temperature 10K and magnetic field 0.1T[2].
The field of DMS is developing in a remarkably fast
paste. nowadays researchers mainly focus on fundamen-
V. CONCLUSION AND OUTLOOK tal aspect, however with deeper and deeper understand-
ing through theoretical and experimental studies, the col-
Dilute magnetic semiconductors (DMS) has both semi- laboration between fundamental and applied research on
conducting and ferromagnetic properties, which pro- DMS will be increasingly intense and widespread.

[1] S. J. Peaton, C. R. Abernathy, D. P. Norton, A. F. B 57, R2037 (1998).

Hebard, Y. D. Park, L. A. Boatner, and J. D. Budai, [15] A. Oiwa, S. Katsumoto, Endo, and et al., Solid State
Materials Science and Engineering R. 40, 137 (2003). Communications 103, 209 (1997).
[2] H. Munekata, H. Ohno, S. von Molnar, and et al, Phys. [16] M. Zajat, R. Doradzin, J. Gosk, and et al., Appl. Phys.
Rev. Lett. 63, 1849 (1989). Lett. 78, 1276 (2001).
[3] E. L. Nagaev, Physics of Magnetic Semiconductors (MIR [17] M. L. Reed, M. K. Ritums, H. H. Stadelmaier, and et al.,
moscow, 1983). Materials Letters 51, 500 (2001).
[4] D. J. Craik, Magnetic Oxides (Wiley New York, 1975). [18] M. L. Reed, N. A. El-Masry, H. H. Stadelmaier, and
[5] H. Ohno, Science 281, 951 (1998). et al., Appl. Phys. Lett. 79, 3473 (2001).
[6] H. Ohno, A. Shen, F. Matsukura, and et al, Appl. Phys. [19] N. Theodoropoulou, A. F. Hebard, M. E. Overberg, and
Lett. 69, 363 (1996). et al., Appl. Phys. Lett. 78, 3475 (2001).
[7] G. A. Medvedkin, T. Ishibashi, T. Nishi, and et al., Jpn. [20] S. Sonoda, S. Shimizu, T. Sasaki, and et al., J. Cryst.
J. Appl. Phys. 39, L949 (2000). Growth 1358, 237 (2002).
[8] H. Akinaga, T. Manago, and M. Shirai, Jpn. J. Appl. [21] T. Sasaki, S. Sonoda, Y. Yamamoto, and et al., J. Appl.
Phys. 39, L1118 (2000). Phys. 91, 7911 (2002).
[9] Y. Matsumoto, M. Murakami, T. Shono, and et al., Sci- [22] T. Minani, MRS Bull. 25, 38 (2000).
ence 291, 854 (2001). [23] T. Fukumura, Z. Jin, A. Ohtomo, and et al., Appl. Phys.
[10] K. Ueda, H. Tabata, and T. Kawai, Appl. Phys. Lett. 79, Lett. 75, 3366 (1999).
988 (2001). [24] K. Sato and H. Katayama-Yoshida., Jpn. J. Appl. Phys.
[11] T. Wakano, N. Fujimura, Y. Morinaga, and et al., Phys- 40, L334 (2001).
ica E 10, 260 (2001). [25] Z. Jin, T. Fukumura, M. Kawasaki, and et al., Appl.
[12] Dietl, H. Ohno, and F. Matsukura, Science 287, 1019 Phys. Lett. 78, 3824 (2001).
(2000). [26] Y. M. Cho, W. K. Choo, H. Kim, and et al., Appl. Phys.
[13] Dietl, H. Ohno, and F. Matsukura, Phys. Rev. B 63, 195 Lett. 80, 3358 (2002).
(2001). [27] S. A. Chambers, S. Thevuthasan, R. F. C. Farrow, and
[14] F. Matsukura, H. Ohno, A. Shen, and et al., Phys. Rev. et al., Appl. Phys. Lett. 79, 3467 (2001).
 7

[28] W. K. Park, R. J. O. Hertogs, J. S. Moodera, and et al., [33] Y. Liu and L. Zhang, Progress in Physics 14, 83 (1994).
J. Appl. Phys. 91, 8093 (2002). [34] C. Zener, Phys. Rev. B 81, 440 (1951).
[29] N. J. Seong, S. G. Yoon, C. R. Cho, and et al., Appl. [35] M. V. Schilfgaarde and O. N. Myrasov, Phys. Rev. B 63,
Phys. Lett. 81, 4209 (2002). 233205 (2001).
[30] Y. Matsumoto, R. Takahashi, M. Murakami, and et al., [36] R. Fiederling, M. Keim, G. Reusher, and et al., Nature
Jpn. J. Appl. Phys., part 2 40, L1204 (2001). 402, 787 (1999).
[31] H. Kimura, T. Fukumura, H. Koinuma, and et al., Appl. [37] N. Akiba, F. Matsukura, A. Shen, and et al., Appl. Phys.
Phys. Lett. 80, 94 (2001). Lett. 73, 2122 (1998).
[32] A. Gerhard, http://www.wsi.tu-
muenchen.de/E24/resources/facilities.htm.