Japanese Journal of Applied

Physics

REGULAR PAPER

Inhomogeneity-induced high temperature ferromagnetism in n-type

ferromagnetic semiconductor (In,Fe)As grown on vicinal GaAs
substrates
To cite this article: Pham Nam Hai et al 2020 Jpn. J. Appl. Phys. 59 063002

View the article online for updates and enhancements.

This content was downloaded from IP address 130.237.165.40 on 10/06/2020 at 14:17

Japanese Journal of Applied Physics 59, 063002 (2020) REGULAR PAPER
https://doi.org/10.35848/1347-4065/ab9401

Inhomogeneity-induced high temperature ferromagnetism in n-type ferromagnetic

semiconductor (In,Fe)As grown on vicinal GaAs substrates
Pham Nam Hai1,2* , Munehiko Yoshida1, Akihide Nagamine1, and Masaaki Tanaka2,3
1
Department of Electrical and Electronic Engineering, Tokyo Institute of Technology, 2-12-1 Ookayama, Meguro, Tokyo 152-0033, Japan
2
Center for Spintronics Research Network (CSRN), The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan
3
Department of Electrical Engineering and Information Systems, The University of Tokyo, 7-3-1 Hongo, Bunkyo, Tokyo 113-8656, Japan
*
E-mail: luosz@jlu.edu.cn
Received April 13, 2020; revised May 7, 2020; accepted May 17, 2020; published online June 2, 2020

We systematically investigate the electrical and magnetic properties of n-type ferromagnetic semiconductor (In,Fe)As thin films grown on vicinal
GaAs(001) substrates with different lattice-relaxation layers and Fe-doping techniques. We show that spinodal decomposition and ferromagnetism
can be enhanced in the (In,Fe)As thin films grown on InAs lattice-relaxation layers under the nucleation growth mode, while they are suppressed in
those grown on the GaSb/AlSb/AlAs (from top to bottom) lattice-relaxation layers under the step-flow growth mode. We demonstrate that room-
temperature ferromagnetism and good electrical properties can be obtained at the same time by using the Fe-delta-doping technique in
combination with the step-flow growth mode. © 2020 The Japan Society of Applied Physics

field Zener model for n-type FMSs,18) high temperature

1. Introduction ferromagnetism in (In,Fe)As has not been realized. Because
Ferromagnetic semiconductors (FMSs) are important spin- TC is proportional to xn1/3 in the mean-field Zener model,
tronic materials inheriting properties of both semiconductors where x is the Fe concentration, increasing either x or n will
and ferromagnets. One of the most important topics in FMS help enhance ferromagnetism.19,20) The former method
research is to fabricate high quality FMSs with high- (increasing x) has been employed successfully in p-type
temperature intrinsic ferromagnetism. The most intensively (Ga,Fe)Sb and n-type (In,Fe)Sb to achieve high-temperature
studied FMS is (Ga, Mn)As,1,2) but its maximum Curie ferromagnetism. In both (Ga,Fe)Sb and (In,Fe)Sb, we can
temperature is limited to 200 K.3) Recently, we have grown dope a surprisingly large amount of Fe while maintaining the
and studied the properties of Fe-doped narrow-gap FMSs, zinc-blende crystal structure.9,11,12) We found that TC of
such as n-type (In,Fe)As,4–6) p-type (Ga,Fe)Sb,7–9) n-type (In, (Ga,Fe)Sb monotonously increases with x, and reaches room
Fe)Sb,10–12) and insulating (Al,Fe)Sb.13) Among them, (In, temperature (300 K) when x is higher than 23%.9) Similarly,
Fe)As is unique because it is an electron-induced n-type FMS TC of (In,Fe)Sb can reach 385 K at x = 35%.12) Thus, it is
with relatively high electron mobility and quantum coher- reasonable to expect intrinsic room-temperature ferromag-
ency. In (In,Fe)As, Fe atoms replace partially In atoms in the netism in (In,Fe)As if high enough Fe can be doped in the
cation (group-III) sites and are in the isoelectronic Fe3+ state. InAs matrix without Fe cluster precipitation.
Thus, the Fe atoms play the role of local magnetic moments, In (In,Fe)As, however, it is difficult to dope a large amount
while free electron carriers are supplied independently by co- of Fe (x > 10%) without destroying the zinc-blende crystal
doped donors such as Si or Be double donors. (In,Fe)As shows structure. Furthermore, heavy doping of Be double donors in
a Curie temperature TC of up to several tens of Kelvin at an (In,Fe)As can increase n up to a maximum value of only
electron density n as low as 6 × 1018 cm−3. It is found that 1.8 × 1019 cm−3.4) Therefore, realization of TC higher than
electron carriers reside in the conduction band with small 100 K in (In,Fe)As by homogeneous Fe doping is still a
electron effective mass (0.03 –0.17m0, where m0 is the challenge. On the other hand, it is possible to increase the
free electron mass),6) high electron mobility (up to local Fe concentration by utilizing the spinodal decomposi-
1000 cm V−1 s−1), and consequently long quantum coherence tion phenomenon21–23) or Fe-delta-doping technique while
length of the electron wavefunctions. In fact, we observed maintaining the zinc-blende crystal structure. (The term
quantum size effects in (In,Fe)As quantum wells as thick as “spinodal decomposition” is widely used to indicate phase
40 nm, and showed that the ferromagnetism in those quantum decomposition in FMSs. This is misleading because phase
wells can be explained by the overlapping of the electron decomposition is not always spinodal;22) it can be either
wavefunctions and the Fe ions.14–16) Furthermore, we demon- binodal or spinodal depending on the polarity of ∂F/∂x,
strated “wavefunction engineering of ferromagnetism” in (In, where F is the free energy function of the FMS and x is the
Fe)As quantum wells by changing the overlapping of the concentration of magnetic impurities. However, for historical
wavefunctions and the Fe ions without large change of the reasons, we will use the term “spinodal decomposition”
electron density.16) This method can be applied to low-power throughout this paper to indicate phase decomposition.)
and high-speed electrical control of ferromagnetism. We also The spinodal decomposition results in local ferromagnetic
observed large spontaneous spin splitting in the conduction domains with higher Fe concentrations. Recently, Sakamoto
band of (In,Fe)As by using tunneling spectroscopy in Esaki et al. has used X-ray magnetic circular dichroism (MCD) to
diode structures.17) These results firmly confirmed the intrinsic investigate the magnetization process in (In,Fe)As:Be.24)
ferromagnetism in (In,Fe)As. They found the existence of nanoscale ferromagnetic do-
Although we obtained TC up to several tens of K at mains at temperatures much higher than the macroscopic
n ∼ 1018 cm−3 in (In,Fe)As, which is two orders of magni- TC. The magnetic moment per ferromagnetic domain is
tude higher than the TC ∼ 100 mK predicted by the mean- 200–300 μB (corresponding to 40–60 Fe atoms), and the
063002-1 © 2020 The Japan Society of Applied Physics
Jpn. J. Appl. Phys. 59, 063002 (2020) P. N. Hai et al.

density of ferromagnetic domains is nearly the same as the Be

impurities. The existence of nanoscale ferromagnetic do-
mains is the manifestation of spatial fluctuations of Fe
concentration at the nanoscale due to spinodal decomposition
in (In,Fe)As:Be.25) In fact, we have confirmed spinodal
decomposition in (In,Fe)As by investigating the local Fe
concentration in nano-scale areas using the three-dimensional
atom probe, which will be presented elsewhere. Such local
ferromagnetic domains can have local ferromagnetism with
TC much higher than the macroscopic TC of the matrix.
However, because the ferromagnetic domains are distributed
randomly, they are magnetically independent of each other
and become superparamagnetic at high temperatures. If we
can control the positions of such nano-scale domains so that
they are spatially and magnetically connected, high-tempera-
ture macroscopic ferromagnetism can be realized in
(In,Fe)As. Note that the spinodal decomposition has also
been utilized in (Ga,Mn)As to fabricate zinc-blende MnAs
nanoparticles with Curie temperature up to 360 K.26,27)
In this paper, we attempt to control the spinodal decom-
position process in (In,Fe)As by growing (In,Fe)As thin films
on vicinal GaAs(001) substrates to realize high-temperature
ferromagnetism. The vicinal GaAs substrates are miscut from
the (001) surface toward the [110] direction, and have atomic
steps along the [1̄10] direction on their surface. There are two
possible growth modes of (In,Fe)As on vicinal GaAs(001)
substrates. One is the well-known step-flow growth mode, in
which incident atoms (both In and Fe) migrate on the surface
toward the step edges and stick to them, rather than nucleate
on the surface. In this mode, the steps travel laterally on the Fig. 1. (Color online) (a) and (b) Schematic cross-sectional structure of
surface as the growth proceeds. Because the position of the type-A and type-B (In,Fe)As samples, respectively. The thickness of the
steps is continuously changing with time, spinodal decom- AlAs, AlSb, and GaSb buffer layers of type-B samples are 10 nm, 30 nm,
position is suppressed, and the Fe atom distribution becomes and 70 nm, respectively. (c)–(f) RHEED patterns taken along the [1̄10]
azimuth after the MBE growth of (c) the InAs layer in sample A2, (d) the
more homogeneous [see Fig. 1(h)]. The other is the nuclea- AlSb layer in sample B3, (e) the (In0.947,Fe0.053)As layer of sample A2, and
tion growth mode, in which incident atoms nucleate on the (f) the (In0.947, Fe0.053)As layer in sample B3. (g) and (h) Idealized schematic
surface. Because the steps do not move on the surface, their three-dimensional structure of type-A and type-B samples, respectively.
positions are unchanged, and they can provide fixed nuclea- Here, the top InAs cap is omitted for clarity.
tion sites for spinodal decomposition of Fe because of
attractive interaction between Fe atoms. Therefore, we can
expect that laterally unmoving steps enhance spinodal Table I. List of (In,Fe)As samples and the corresponding off-angle of
decomposition at their positions, resulting in more inhomo- vicinal GaAs(001) substrates, lattice-relaxation layers (from top to bottom),
and Curie temperature (TC).
geneous distribution of Fe; higher Fe concentration is
expected at the steps, while lower Fe concentration is Off-angle Lattice-relaxa- TC (K) TC (K)
expected at the terrace areas between steps [see Fig. 1(g)]. Sample (degree) tion layers (terrace) (step)
Thus, we can expect two types of (In,Fe)As grown on vicinal A1 0 [=GaAs InAs 10
GaAs(001) substrates; (i) the more homogeneous (In,Fe)As (001) just-cut]
grown under the step-flow mode, and (ii) the more inhomo- A2 2 20 >300
geneous (In,Fe)As grown under the nucleation mode. By A3 6 30 >300
A4 10 10 >300
investigating their electrical and magnetic properties, we can
A5 19.5 [=GaAs <10 >300
evaluate the roles of spinodal decomposition in (In,Fe)As, (114)A]
and establish methods for obtaining inhomogeneity-induced B1 0 [=GaAs GaSb/AlSb/ <10
high-temperature ferromagnetism in (In,Fe)As. (001) just-cut] AlAs
B2 2 <10
2. Experiments B3 6 <10
B4 10 <10
The studied (In,Fe)As layers were grown by low-temperature B5 19.5 [=GaAs <10
molecular beam epitaxy (LT-MBE) on various vicinal GaAs (114)A]
(001) substrates with off-angles of 2°, 6°, 10°, and 19.5° C 2 150
[GaAs(114)A] toward the [110] direction, as summarized in D 2 305

063002-2 © 2020 The Japan Society of Applied Physics

Jpn. J. Appl. Phys. 59, 063002 (2020) P. N. Hai et al.

Table I. For comparison, we also grew reference (In,Fe)As and can trigger the spinodal decomposition in the top
layers on a GaAs(001) just-cut substrate. To reduce the very (In,Fe)As thin films. In contrast, the AlSb layer (and the
large lattice mismatch between (In,Fe)As (lattice following GaSb layer, not shown here) can preserve the steps
constant ∼6.06 Å) and GaAs (lattice constant ∼ 5.65 Å), structures of the vicinal GaAs substrate even after lattice
lattice-relaxation layers are inserted between the GaAs relaxation between them has occurred, as clearly seen by their
substrates and the (In,Fe)As layers. We have grown four inclined streaky RHEED patterns in Fig. 1(d). Figures 1(e)
types of 20 nm thick (In,Fe)As thin films with various type of and 1(f) show the RHEED patterns of the (In,Fe)As layer of
lattice-relaxation layers and Fe doping techniques. Type A sample A2 and sample B3, respectively. The surface of the
are (In,Fe)As thin films grown on InAs lattice-relaxation (In,Fe)As layer of sample A2, which was grown on the InAs
layers, while type B are those grown on a stack of GaSb/ lattice-relaxation layer, is slightly rough [see Fig. 1(e)],
AlSb/AlAs (from top to bottom layers) lattice-relaxation which is due to the influence of defects or local strains
layers. As shown below, the InAs lattice-relaxation layers inherited from the steps on the GaAs substrate. In contrast,
promote nucleation growth mode, while the GaSb/AlSb/AlAs the surface of the (In,Fe)As layer of sample B3 is smooth
lattice-relaxation layers promote the step-flow mode. The Fe with clear inclined streaky RHEED patterns [see Fig. 1(f)],
concentration x in the (In1−x,Fex)As layers of type A and B indicating the step-flow mode of (In,Fe)As even at low
are 5.3%. Type C and type D are those grown by an Fe-delta- temperature in this sample.
doping technique on the GaSb/AlSb/AlAs lattice-relaxation From the RHEED observations mentioned above, we can
layers (More details of the growth procedure will be expect that samples A2–A5 have stronger spinodal decom-
described later). During the MBE growth, reflection high- position and more inhomogeneous distribution of Fe, while
energy electron diffraction (RHEED) was used to observe the samples B2–B5 have more homogenous distribution of Fe,
crystallinity and surface morphology of the samples. After compared with the reference sample A1 grown on a
the thin-film growth, electrical properties, magneto-transport GaAs(001) just-cut substrate under the usual condition. The
properties, and magneto-optical properties of those (In,Fe)As idealized schematic three-dimensional (3D) sample structure
thin films were investigated using various methods such as and the expected distribution of Fe are depicted in Figs. 1(g)
MCD spectroscopy and anomalous Hall effect (AHE). The and 1(h) for samples A2–A5 and B2–B5, respectively. [Note
crystal structure of one sample (A2) was investigated by that in Figs. 1(g) and 1(h), the height of the steps is one
scanning transmission electron microscopy (STEM). monolayer (ML), and is magnified for better illustration. The
InAs cap layer is omitted.] In samples A2–A5, one can
3. Results
expect areas with relatively high Fe concentration induced by
3.1. Samples A1–A5 and B1–B5 defects or local strains inherited from the steps on the GaAs
3.1.1. Crystal growth. Figures 1(a) and 1(b) show the substrate, and terrace areas in between with relatively low Fe
schematic cross-sectional structure of samples A1–A5 (type A) concentration. In contrast, in samples B2–B5, we expect that
and samples B1–B5 (type B), respectively. The growth the Fe concentration is more homogenous thanks to the step-
procedure of these samples is as follows. First, we grew a flow mode.
50 nm thick GaAs buffer layer at a substrate temperature (TS) Figure 2 shows a representative STEM lattice image of
of 580 °C. After that, we grew a 10 nm thick InAs lattice- sample A2 projected along the [1̄10] azimuth. The STEM
relaxation layer at TS = 500 °C for type A, or a stack of 10 nm image indicates that the crystal structure of the (In,Fe)As
thick AlAs (TS = 500 °C), 30 nm thick AlSb (TS = 470 °C), layer is zinc-blende with some dislocations, but there is no
and 70 nm thick GaSb (TS = 470 °C) lattice-relaxation layers precipitation of body-centered-cubic or face-centered-cubic
for type B, respectively. These lattice-relaxation layers reduce Fe nanoclusters or other types of crystal structure. This
the lattice mismatch between (In,Fe)As and GaAs. After
cooling the samples to 236 °C, we started growing a 20 nm
thick (In0.947,Fe0.053)As layer, doped with double-donor Be
atoms of 1 × 1019 cm−3. Finally, we grew a 3 nm InAs cap
layer.
Figures 1(c)–1(d) show the RHEED patterns taken along
the [1̄10] azimuth of the InAs lattice-relaxation layer of
sample A2 (2°-off substrate) and the AlSb lattice-relaxation
layer of sample B3 (6°-off substrate), respectively. We see
clear difference between these layers. The InAs layer was
grown at a high temperature of 500 °C, allowing fast surface
migration of In atoms and formation of lattice-relaxed InAs
layer with a relatively flat surface, as evidenced by the
streaky patterns in Fig. 1(c). In this case, however, the steps
on the surface of the vicinal GaAs substrate were buried, and
no step-flow growth of (In,Fe)As on the InAs lattice-
relaxation layer can be expected. Nevertheless, we expect
that the steps on the GaAs substrate can leave traces as defect Fig. 2. Cross-sectional scanning transmission electron microscopy
(STEM) lattice image of sample A2, consisting of 3 nm thick InAs cap/20
sites or local strains at the GaAs/InAs interface, which can nm thick (In0.947,Fe0.053)As/10 nm thick InAs lattice-relaxation buffer layer,
remain up to the InAs surface. In spinodal decomposition, grown on a 2°-off GaAs(001) substrate. Note that the 3 nm thick InAs cap
such local defects or strains are known to be nucleation sites and about 2 nm thick (In,Fe)As were oxidized.

063002-3 © 2020 The Japan Society of Applied Physics

Jpn. J. Appl. Phys. 59, 063002 (2020) P. N. Hai et al.

(a) have higher resistivity. At low temperatures, all type-A

samples show insulating behavior, although they are degener-
ated semiconductors with n > 3 × 1018 cm−3 at room tempera-
ture. This indicates that these type-A samples have considerable
scattering centers that cause metal–insulator transition (MIT) at
low temperatures. Interestingly, we observed a trend that the
higher the off-angle is, the more rapidly the resistivity increases
with decreasing temperature, indicating that stronger MIT
occurs. This shows a clear correlation between the off-angle
and the density of scattering centers, supporting our suggestion
(b) that steps on the surface of GaAs create defects or local strains
in the InAs lattice-relaxation layers, which in turn affect the
growth of the top (In,Fe)As layers. The off-angle-dependence of
the electron mobility μ at 300 and 10 K [inset in Fig. 3(a)] also
supports our argument; μ decreases with increasing the off-
angle and lowering temperature. In contrast, the resistivity of
type-B samples shows metallic behavior in a wide range of
temperature, indicating that these samples have good crystal
quality (except for sample B5 which is weakly insulating).
Furthermore, we found that samples B2–B5 have improved μ at
Fig. 3. (Color online) (a) and (b) Temperature dependence of the resistivity of low temperature compared with that of sample B1, as shown in
type-A and type-B (In,Fe)As samples, respectively. The insets of (a) and (b) the inset of Fig. 3(b). This indicates that the step-flow growth
show the electron mobility at 10 K and 300 K versus off-angle of the substrates. mode of (In,Fe)As layers on top of GaSb/AlSb/AlAs lattice-
relaxation layers helped improve the crystal quality.
indicates that any high-temperature ferromagnetism observed Next, we investigate the relationships between the growth
in this sample should due to the locally high Fe-concentration mode of (In,Fe)As and their magnetic properties. First, in
areas of zinc-blende crystal structure, and not due to order to prove the intrinsic ferromagnetism of the (In,Fe)As
precipitated Fe nanoclusters. thin films, we investigate the magneto-optical properties of
3.1.2. Electrical, magneto-optical, and magneto- the (In,Fe)As thin films using MCD spectroscopy in a
transport properties. Figures 3(a) and 3(b) show the reflection setup. Here, the reflection MCD intensity is
temperature dependence of the resistivity of samples 90 (Rs+ - Rs-) 1 dR
expressed as p 2R
∝ R dE
DE, where Rs+ and Rs- are
A1–A5 and B1–B5, respectively. At room temperature,
sample A1, which was grown on a GaAs(001) just-cut the reflectance for the incident right (s+) and left (s-) circular
substrate, has the lowest resistivity, while samples A2–A5 polarized light, respectively, R is the total reflectance, E is the

(a) (c) (e)

(b) (d) (f)

Fig. 4. (Color online) (a) and (b) MCD spectra at 5 K of type-A and type-B (In,Fe)As samples, respectively, under an external magnetic field of 1 T applied
perpendicular to the film plane. Black arrows indicate the optical critical point energies of InAs. (c) and (d) Normalized MCD spectra at 0.2, 0.5, and 1 T of
samples A2 and B3, respectively. (e) and (f) MCD intensity-magnetic field hysteresis at various temperatures of sample A3 and B3, respectively, measured at
the photon energy of E1. The insets of (e) and (f) show the MCD hysteresis at 300 K.

063002-4 © 2020 The Japan Society of Applied Physics

Jpn. J. Appl. Phys. 59, 063002 (2020) P. N. Hai et al.

photon energy, and DE is the spin-splitting energy (Zeeman sample A1 [grown on a GaAs(001) just-cut substrate], there is no
dR
splitting energy) of a FMS. The MCD spectrum reflects dE , hysteresis. However, for samples A2–A5 [grown on vicinal
which is unique for the FMS in question. The MCD spectrum of GaAs(001) substrates], clear hysteresis was observed.
an intrinsic FMS would show strongly enhanced extrema at the Furthermore, the hysteresis becomes stronger with increasing
optical critical point energies of the host semiconductors. On the the off-angle and the step density. To further confirm the high-
other hand, if a semiconductor sample contains second-phase temperature ferromagnetic phase in type-A samples, we mea-
metallic magnetic precipitates, the MCD spectrum would show sured the AHE of samples A1–A5 at 300 K. Figures 5(f)–5(j)
a broad background and no particular features related to the host show the anomalous Hall resistance (AHR) of samples A1–A5,
semiconductor, because the host semiconductor would remain normalized by their value at 0.5 T. We observed the same
non-magnetic in this case. Therefore, MCD is a powerful tool to hysteresis curves of AHR as those of MCD as a function of the
distinguish whether the magnetization in the samples comes off-angle. For comparison, we show in Figs. 5(k)−5(o) the
from the intrinsic FMS phase or metallic second-phase pre- MCD − H characteristics of samples B1–B5 at 300 K, which
cipitations. At the same time, the MCD intensity is also show no hysteresis. These suggest that the different growth
proportional to DE, allowing us to obtain the magnetization mode between samples A2–A5 and B2–B5 results in the high-
hysteresis of the measured material as a function of the external temperature ferromagnetic phase in samples A2–A5, while such
magnetic field H. These are the great advantages of MCD over phase is absent in samples B2–B5.
more conventional techniques, such as superconducting The co-existence of the high-TC and low-TC phase in
quantum interference devices, which cannot distinguish the samples A2–A5 and the very low-TC phase in samples B2–
origin of the observed magnetization.28) Figures 4(a) and 4(b) B5 are consistent with spinodal decomposition tailored by the
show the MCD spectra of samples A1–A5 and B1–B5, lattice-relaxation layers and the growth mode of (In,Fe)As, as
respectively, at 5 K and under a magnetic field of 1 T applied illustrated in Figs. 1(g) and 1(h). In samples A2–A5, the
perpendicular to the film plane. The MCD spectra of both type- spinodal decomposition is likely enhanced by the existence
A and type-B samples show strong extrema at the optical critical of defects or local strains inherited from the atomic steps on
point energies E1 (2.61 eV), E1 + D1 (2.88 eV), and E0 the vicinal GaAs substrates, resulting in local ferromagnetic
(4.39 eV) of InAs.29) These results indicate that (In,Fe)As thin domains with high Fe concentrations and high TC. Since
films grown on vicinal substrates preserve the zinc-blende band these step-induced high-TC ferromagnetic domains are spa-
structures, consistent with the STEM observation in Fig. 2. tially connected along the steps, they manifest themselves as
There is, however, difference in the magnetic field response of the macroscopic high-TC phase. On the other hand, the
the MCD spectra between type-A and type-B samples. terrace areas between those domains have lower Fe concen-
Figures 4(c) and 4(d) show the normalized MCD spectra by tration, and appear as the low-TC phase. In samples B2–B5
their intensity at E1 of sample A2 and B3, measured with where the growth proceeds by the step-flow growth mode,
applied magnetic fields of 0.2, 0.5, and 1 T. The normalized the spinodal decomposition is suppressed because the step
MCD spectra of sample B3 shows nearly perfect overlapping on position is changing with time. Thus, the Fe atoms are
a single spectrum in the whole photon energy range, indicating distributed more randomly, and the TC becomes very low.
that the MCD spectra come from a single-phase ferromagnetism These magnetic responses are consistent with the normalized
of the entire (In,Fe)As film. In contrast, the normalized MCD MCD spectra shown in Figs. 4(c) and 4(d), and conclusions
spectra of sample A3 show clear difference in the photon energy from the previous XMCD study of (In, Fe)As.24)
range between 3.2 and 4.2 eV, indicating that there are different 3.2. Samples C and D
phases in this sample. These magnetic responses of the MCD 3.2.1. Crystal growth. In the previous section, we have
spectra are consistent with our expectation for the inhomoge- shown that room-temperature ferromagnetism can be obtained
neous (homogeneous) distribution of Fe in type-A (type-B) by utilizing step-induced spinodal decomposition for (In,Fe)As
samples, respectively, as shown in Figs. 1(g) and 1(h). grown on the InAs lattice-relaxation layers in type-A samples.
Next, we analyze the magnetic field dependence of the MCD However, those type-A samples have many crystal defects and
intensity (MCD − H characteristics). Figures 4(e) and 4(f) show relatively poor electrical properties. In this section, we
the representative MCD − H characteristics at different tem- demonstrate an Fe-delta-doping technique in combination
peratures of sample A3 and B3, respectively, measured at the with the step-flow growth mode on the GaSb/AlSb/AlAs
optical critical point energy E1. One can see that both samples buffers to achieve high-TC ferromagnetism as well as good
show weak hysteresis at low temperatures, indicating the electrical properties. The delta-doping technique enables
existence of a low TC ferromagnetic phase. Other samples doping locally high concentration of magnetic atoms in a
show similar MCD − H characteristics. TC of all the samples semiconductor while maintaining its crystal structure, and it
was estimated by the Arrott plots (MCD2 − H/MCD) and was was used to obtain high TC up to 250 K in
listed in the 4th column of Table I. One can see that samples (Ga,Mn)As.30) Figures 6(a) and 6(b) show the structure of
B1–B5 have very low TC (<10 K), while samples A1–A5 have two new samples C and D, respectively. First, we grew GaSb/
slightly higher TC that do not exceed 30 K. However, there is a AlSb/AlAs lattice-relaxation layers on 2°-off vicinal GaAs
high-TC phase up to room-temperature in sample A3, as shown substrates, in the same way as sample B2. After cooling the
by the magnified hysteresis loop at 300 K [see inset of samples to 236 °C, we first grew 4 MLs of Be-doped InAs
Fig. 4(e)]. In contrast, there is no such high-TC phase in sample (referred to as InAs:Be). For sample C, we used the Fe-delta-
B3 [see inset of Fig. 4(f)]. doping technique to grow 1 ML of Be-doped FeAs (referred to
In order to clarify the origin of the high-TC phase, we as FeAs:Be), i.e. 1 ML of tetrahedral Fe–As bonding, and
systematically investigate the MCD − H characteristics of sam- 2 MLs of InAs:Be, and repeated this growth process for 22
ples A1–A5 at 300 K, which are shown in Figs. 5(a)–5(e). For cycles. For sample D, we grew 0.5 ML of FeAs:Be followed
063002-5 © 2020 The Japan Society of Applied Physics
Jpn. J. Appl. Phys. 59, 063002 (2020) P. N. Hai et al.

(a) (b) (c) (d) (e)

(f) (g) (h) (i) (j)

(k) (l) (m) (n) (o)

Fig. 5. (Color online) (a)–(e) Magnetic field dependence of the MCD intensity at room temperature for type-A samples, measured at the photon energy of E1.
(f)–(j) Magnetic field dependence of AHR at room temperature for type-A samples. (k)–(o) Magnetic field dependence of the MCD intensity at room
temperature for type-B samples. The sign of MCD data are reversed (−MCD) for better comparison with AHR data.

3.2.2. High-temperature ferromagnetism. Figures 7(a)

and 7(b) show the MCD spectra of samples C and D,
respectively, at 5 K and under a magnetic field of 1 T applied
perpendicular to the film plane. We confirmed the spectral
features of the zinc-blende crystal structure, which are similar
to those of (In,Fe)As grown under usual conditions in our
previous studies. The MCD intensity is as strong as about
100 mdeg at E1. This huge MCD intensity corresponds to an
internal magnetic field of 100 T acting on the InAs host,32)
which cannot be explained by any stray fields from Fe
metallic precipitations acting on non-magnetic InAs, but by
the s,p–d exchange interactions between the Fe localized
d-electrons and InAs. We conclude that the Fe-delta-doped
(In,Fe)As layers in samples C and D also preserve the zinc-
blende band structure.
Figures 7(c) and 7(d) show the MCD − H hysteresis at
various temperatures of samples C and D, respectively,
measured at the photon energy E1. Figures 7(e) and 7(f)
Fig. 6. (Color online) (a) and (b) Schematic cross-sectional structure of show the Arrott plot of samples C and D, respectively. TC of
samples C and D, respectively. The thickness of the AlAs, AlSb, GaSb sample C is 150 K, while that of sample D is 305 K. The
lattice-relaxation layers are the same as those of type-B samples. (c) and electron concentration (n) and mobility (μ) are estimated to
(d) Idealized schematic three-dimensional structure of samples C and D, be n = 1.35 × 1018 cm−3 and μ = 60 cm2 V−1 s−1 for sample
respectively.
C, and n = 1.2 × 1019 cm−3 and μ = 122 cm2 V−1 s−1 for
sample D at room temperature. The sample D has better
by 0.5 ML of InAs:Be, and repeated this for 66 cycles. The crystal quality than sample C, and its resistivity is one order
total Fe-delta-doped (In,Fe)As layer thickness is 66 MLs of magnitude smaller than that of type-A and type-B samples.
(=20 nm) for both samples C and D. Finally, we grew a The ratio of TC between samples D and C (2.03) agrees well
3 nm thick InAs cap layer. Figures 6(c) and 6(d) show the with the ratio of n1/3 between them (2.07), which is consistent
idealized schematic structure of sample C and D, respectively. with the “electron-induced” ferromagnetism observed in our
In sample C, 1 ML of FeAs:Be is separated by 2 ML of InAs. previous studies for (In,Fe)As. Thus, by using the Fe-delta-
In sample D, each 0.5 ML FeAs:Be is connected vertically, doping technique in combination with the step-flow mode,
forming a planar FeAs/InAs superlattice, which is an analogy we achieved both room-temperature ferromagnetism and
to the AlAs/GaAs semiconductor planar superlattice grown on good electrical properties with high electron concentration
vicinal GaAs substrates.31) and high electron mobility in sample D.
063002-6 © 2020 The Japan Society of Applied Physics
Jpn. J. Appl. Phys. 59, 063002 (2020) P. N. Hai et al.

(a) (c) (e) (g)

(b) (d) (f) (h)

Fig. 7. (Color online) (a) and (b) MCD spectrum at 5 K of samples C and D, respectively, under an external magnetic field of 1 T applied perpendicular to
the film plane. Black arrows indicate the optical critical point energies of InAs. (c) and (d) MCD intensity-magnetic field hysteresis at various temperatures of
samples C and D, respectively, measured at the photon energy of E1. (e) and (f) Arrott plots of the MCD intensity of samples C and D, respectively.
(g) Temperature dependence of the resistivity of sample D measured when the current was applied along the [110] and [1̄10 ] direction. (h) Normalized
temperature dependence of the resistivity of sample D by their values at room temperature.

The next question is whether the room-temperature ferro- we conclude that Fe-delta-doping induced inhomogeneity in
magnetism in sample D originates from the Fe-delta-doping sample D survives and is the origin of the observed room-
induced inhomogeneity, i.e. the planar FeAs/InAs super- temperature ferromagnetism. Finally, we note that high-TC
lattice, whose idealized schematic structure is shown in ferromagnetism is only observed in Fe-delta-doped
Fig. 6(d). During the crystal growth, such a superlattice (In,Fe)As thin films grown on vicinal GaAs substrates with
may collapse due to interdiffusion of Fe and In atoms the step-flow mode. In Fe-delta-doped (In,Fe)As reference
between the Fe-delta-doped FeAs domains and the InAs samples grown on GaAs(001) just-cut substrates, TC was
domains. In the ultimate case, the FeAs and InAs domains only several tens of K.
may fuse together to form an (In,Fe)As alloy with the
nominal Fe concentration of 50%. To find whether the planar 4. Discussion
FeAs/InAs superlattice in sample D survives such interdiffu- Under the assumption of the mean-field Zener model, electron-
sion, we measured the resistivity of sample D when the induced ferromagnetism at temperatures higher than 1 K in n-
current is applied along the [110] and [1̄10] direction. If type FMSs would not be possible.18) In previous works,
interdiffusion of Fe and In atoms was completed, there would however, we have already demonstrated electron-induced
be no difference between the resistivity r[110] and r[110
¯ ] for ferromagnetism up to several tens of K in bulk-like n-type
(In,Fe)As with Fe concentrations less than 10%.4) In this work,
the current flows along the [110] and [1̄10] direction. On the
we have further demonstrated ferromagnetism up to room-
hand, if interdiffusion of Fe and In is not completed, we can
temperature in (In,Fe)As thin films with ferromagnetic do-
expect some difference between them. Using the serial
mains having locally high Fe concentrations. Even higher TC
(parallel) resistor model for the FeAs/InAs superlattice
can be obtained by increasing the electron concentration n.
when the current follows along the [110] ([1̄10]) direction,
r FeAs + r InAs 2r FeAs Dr InAs
Note that the maximum n obtained so far in (In,Fe)As is
we obtain r[110] = 2
and r[110
¯ ]= r FeAs + r InAs
, where 1.8 × 1019 cm−3, which is much lower than the maximum hole
r FeAs and r InAs are the resistivity of the FeAs and InAs density (∼1021 cm−3) in (Ga,Mn)As. Thus, there is still more
domains. From these formulas, we expect r[110] > r[110 room for increasing n and TC. High n may be realized by
¯ ] for
doping group-VI elements, such as Te or Se, as donors in
r FeAs ≠ r InAs. Figures 7(g) shows the temperature depen-
(In,Fe)As. From the data of samples C and D, we predict that
dence of the resistivity of sample D measured when the the ultimate TC of such (In,Fe)As films would be given by
current was applied along the [110] and [1̄10] direction, while TC = 305 K × (n/1.2 × 1019 cm−3)1/3. That means, very high
Fig. 7(h) shows those normalized by the values at room TC of 618 K can be obtained at n = 1 × 1020 cm−3.
temperature. One can see that r[110] is not only larger than
r[110
¯ ] but also increases more rapidly than r[110 ¯ ] with
5. Conclusion
decreasing temperature, indicating the existence of FeAs/ In conclusion, we have shown that it is possible to realize
InAs domains with different electrical characteristics. Thus, high temperature ferromagnetism in (In,Fe)As thin films

063002-7 © 2020 The Japan Society of Applied Physics

Jpn. J. Appl. Phys. 59, 063002 (2020) P. N. Hai et al.

grown on vicinal GaAs(001) substrates by utilizing either 7) N. T. Tu, P. N. Hai, L. D. Anh, and M. Tanaka, Appl. Phys. Lett. 105,
spinodal decomposition or Fe-delta-doping techniques. 132402 (2014).
8) N. T. Tu, P. N. Hai, L. D. Anh, and M. Tanaka, Phys. Rev. B 92, 144403
In particular, we demonstrated that spinodal decomposition (2015).
and the corresponding high-temperature ferromagnetism in 9) N. T. Tu, P. N. Hai, L. D. Anh, and M. Tanaka, Appl. Phys. Lett. 108,
(In,Fe)As thin films can be enhanced by growing (In,Fe)As 192401 (2016).
on the InAs lattice-relaxation layers under the nucleation 10) N. T. Tu, P. N. Hai, L. D. Anh, and M. Tanaka, Appl. Phys. Lett. 112,
122409 (2018).
growth mode, or suppressed by growing them on the GaSb/ 11) N. T. Tu, P. N. Hai, L. D. Anh, and M. Tanaka, Appl. Phys. Express 11,
AlSb/AlAs lattice-relaxation layers under the step-flow 063005 (2018).
mode. We further demonstrated that room-temperature ferro- 12) N. T. Tu, P. N. Hai, L. D. Anh, and M. Tanaka, Appl. Phys. Express 12,
magnetism and good electrical properties can be obtained at 103004 (2019).
13) L. D. Anh, D. Kaneko, P. N. Hai, and M. Tanaka, Appl. Phys. Lett. 107,
the same time by using the Fe-delta-doping technique in 232405 (2015).
combination with the step-flow mode. Our techniques can 14) L. D. Anh, P. N. Hai, and M. Tanaka, Appl. Phys. Lett. 104, 042404 (2014).
also be applied to other FMSs to obtain high-temperature 15) D. Sasaki, L. D. Anh, P. N. Hai, and M. Tanaka, Appl. Phys. Lett. 104,
142406 (2014).
ferromagnetism, which is necessary for semiconductor-based
16) L. D. Anh, P. N. Hai, Y. Kasahara, Y. Iwasa, and M. Tanaka, Phys. Rev. B
spintronic devices operating at room temperature. 92, 161201(R) (2015).
17) L. D. Anh, P. N. Hai, and M. Tanaka, Nat. Commun. 7, 13810 (2016).
Acknowledgments 18) T. Dietl, A. Haury, and Y. Merle d’Aubigne, Phys. Rev. B 55, R3347(R)
This work is supported by the CREST program (No. (1997).
19) T. Dietl, H. Ohno, F. Matsukura, J. Cibert, and D. Ferrand, Science 287,
JPMJCR1777) of the Japan Science and Technology 1019 (2000).
Agency (JST), and the Spintronics Research Network of 20) T. Dietl, H. Ohno, and F. Matsukura, Phys. Rev. B 63, 195205
Japan (Spin-RNJ). P. N. H. acknowledges support from the (2001).
Kato Foundation for Promotion of Science, Iketani Science 21) K. Sato, H. Katayama-Yoshida, and P. H. Dederichs, Jpn. J. Appl. Phys. 44,
L948 (2005).
and Technology Foundation, and Toray Science Foundation. 22) P. N. Hai, S. Yada, and M. Tanaka, J. Appl. Phys. 109, 073919 (2011).
ORCID iDs 23) T. Dietl, K. Sato, T. Fukushima, A. Bonanni, M. Jamet, A. Barski,
S. Kuroda, M. Tanaka, P. N. Hai, and H. Katayama-Yoshida, Rev. Mod.
Pham Nam Hai https://orcid.org/0000-0001-6685-4912 Phys. 87, 1311 (2015).
Masaaki Tanaka https://orcid.org/0000-0003-3599-663X 24) S. Sakamoto, L. D. Anh, P. N. Hai, G. Shibata, Y. Takeda, M. Kobayashi,
Y. Takahashi, T. Koide, M. Tanaka, and A. Fujimori, Phys. Rev. B 93,
035203 (2016).
25) N. D. Vu, T. Fukushima, K. Sato, and H. Katayama-Yoshida, Jpn. J. Appl.
1) H. Ohno, A. Shen, F. Matsukura, A. Oiwa, A. Endo, S. Katsumoto, and Phys. 53, 110307 (2014).
Y. Iye, Appl. Phys. Lett. 69, 363 (1996). 26) M. Yokoyama, H. Yamaguchi, T. Ogawa, and M. Tanaka, J. Appl. Phys. 97,
2) T. Hayashi, M. Tanaka, T. Nishinaga, H. Shimada, H. Tsuchiya, and 10D317 (2005).
Y. Otuka, J. Cryst. Growth 175–176, 1063 (1997). 27) P. N. Hai, S. Ohya, M. Tanaka, S. E. Barnes, and S. Maekawa, Nature 458,
3) L. Chen, X. Yang, F. Yang, J. Zhao, J. Misuraca, P. Xiong, and S. von 489 (2009).
Molnar, Nano Lett. 11, 2584 (2011). 28) K. Ando, Science 312, 1883 (2006).
4) P. N. Hai, L. D. Anh, S. Mohan, T. Tamegai, M. Kodzuka, T. Ohkubo, 29) R. R. L. Zucca and Y. R. Shen, Phys. Rev. B 1, 2668 (1970).
K. Hono, and M. Tanaka, Appl. Phys. Lett. 101, 182403 (2012). 30) A. M. Nazmul, T. Amemiya, Y. Shuto, S. Sugahara, and M. Tanaka, Phys.
5) P. N. Hai, L. D. Anh, and M. Tanaka, Appl. Phys. Lett. 101, 252410 (2012). Rev. Lett. 95, 017201 (2005).
6) P. N. Hai, D. Sasaki, L. D. Anh, and M. Tanaka, Appl. Phys. Lett. 100, 31) M. Tanaka and H. Sakaki, Appl. Phys. Lett. 54, 1326 (1989).
262409 (2012). 32) K. Ando and H. Munekata, J. Magn. Magn. Mater. 272, 2004 (2004).

063002-8 © 2020 The Japan Society of Applied Physics