Ga
2
O
3
Cite as: Appl. Phys. Lett. 117, 032101 (2020); https://doi.org/10.1063/5.0012579 Submitted: 02 May 2020 . Accepted: 07 July 2020 . Published Online: 20 July 2020
Nguyen Tien Son , Quoc Duy Ho , Ken Goto, Hiroshi Abe , Takeshi Ohshima , Bo Monemar, Yoshinao Kumagai, Thomas Frauenheim, and Peter Deák
COLLECTIONS
This paper was selected as an Editor’s Pick
ARTICLES YOU MAY BE INTERESTED IN
A 1.86-kV double-layered NiO/β-Ga2O3 vertical p–n heterojunction diode
Applied Physics Letters
117, 022104 (2020);
https://doi.org/10.1063/5.0010052
Point defects in Ga2O3
Journal of Applied Physics
127, 101101 (2020);
https://doi.org/10.1063/1.5142195
A review of Ga2O3 materials, processing, and devices
Electron paramagnetic resonance and theoretical
study of gallium vacancy in b-Ga
2
O
3
Cite as: Appl. Phys. Lett. 117, 032101 (2020);doi: 10.1063/5.0012579 Submitted: 2 May 2020
.
Accepted: 7 July 2020.
Published Online: 20 July 2020
Nguyen TienSon,1,a) Quoc DuyHo,2,3 KenGoto,4HiroshiAbe,5 TakeshiOhshima,5 BoMonemar,1 YoshinaoKumagai,4,6ThomasFrauenheim,2,7and PeterDeak2
AFFILIATIONS
1Department of Physics, Chemistry and Biology, Link€oping University, SE-58183 Link€oping, Sweden
2Bremen Center for Computational Materials Science, University of Bremen, P.O. Box 330440, D-28334 Bremen, Germany 3Faculty of General Sciences, Can Tho University of Technology, 256 Nguyen Van Cu, Can Tho 94108, Vietnam
4Department of Applied Chemistry, Tokyo University of Agriculture and Technology, Koganei, Tokyo 184-8588, Japan
5National Institutes for Quantum and Radiological Science and Technology, 1233 Watanuki, Takasaki, Gunma 370-1292, Japan 6Institute of Global Innovation Research, Tokyo University of Agriculture and Technology, Koganei, Tokyo 184-8588, Japan 7Beijing Computational Science Research Center, 10 East Xibeiwang Road, Haidian District, Beijing 100193, China
and Computational Science and Applied Research Institute, Shenzhen, China a)Author to whom correspondence should be addressed:tien.son.nguyen@liu.se
ABSTRACT
Unintentionally doped n-type b-Ga2O3becomes highly resistive after annealing at high temperatures in oxygen ambient. The annealing
pro-cess also induces an electron paramagnetic resonance (EPR) center, labeled IR1, with an electron spin of S ¼ 1/2 and principal g-values of gxx¼ 2.0160, gyy¼ 2.0386, and gzz¼ 2.0029 with the principal axis of gzzbeing 60from the [001]direction and gyyalong the b-axis. A
hyperfine (hf) structure due to the hf interaction between the electron spin and nuclear spins of two equivalent Ga atoms with a hf splitting of 29 G (for69Ga) has been observed. The center can also be created by electron irradiation. Comparing the Ga hf constants determined by EPR with corresponding values calculated for different Ga vacancy-related defects, the IR1 defect is assigned to the double negative charge state of either the isolated Ga vacancy at the tetrahedral site (VGa(I)2 ) or the VGa(I)–Gaib–VGa(I)complex.
VC 2020 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license (http://
creativecommons.org/licenses/by/4.0/).https://doi.org/10.1063/5.0012579
Gallium oxide (Ga2O3) belongs to ultra-wide bandgap
semicon-ductors and exists in five different polytypes, of which monoclinic b-Ga2O3is most thermally stable. The wide bandgap (4.7 eV)1and
the availability of substrates with well-controlled n-type doping, which can be fabricated by different methods with low costs, make b-Ga2O3
an attractive material for transparent contact layers in optical devices operating in the deep ultraviolet spectral region2and for high-power electronics.3In recent years, remarkable progress in developing Ga2O3
power electronics has been made.3However, knowledge on impurities and defects, especially intrinsic defects, in b-Ga2O3is still rather poor.
It is known that annealing in O2 ambient can transform n-type
b-Ga2O3 into semi-insulating material,4 while annealing in N2 will
return it to n-type conducting material.5However, the defect responsi-ble for compensation of the shallow donors in creating high-resistive b-Ga2O3has not been identified. The role of intrinsic defects, such as
the Ga vacancy (VGa), in such compensation processes has been
sug-gested by recent theoretical calculations.6
Although bulk b-Ga2O3has been available since many years, few
electron paramagnetic resonance (EPR) studies of intrinsic defects have been reported. In neutron-irradiated b-Ga2O3, Kananen et al.7
observed an EPR spectrum with S ¼ 1/2, showing a hyperfine (hf) structure due to the interaction between the electron spin and the nuclear spins of two equivalent Ga atoms. The center was tentatively assigned to the doubly negatively charged Ga vacancy (VGa2).7Another
hf structure with a smaller hf splitting was also observed simulta-neously in darkness at the low- and high-field ends of the first spec-trum. Based on the observation of the reduction of the hf constant by a factor of two, a model of an isolated negatively charged VGahaving
S ¼ 1 was suggested.7In this model, two holes are localized on two neighboring OIIand OIIIof a GaIIat an octahedral site so that the hf
interaction is assumed to be reduced by a factor of two. However, with different numbers of equivalent Ga atoms (two for OIIand three for
OIII), the hf structure at OIIand OIIIshould be different and the
assumption of a half hf splitting for this center is questionable. In a later EPR study of n-type and Fe-doped semi-insulating b-Ga2O3 irradiated with protons, Bardeleben and co-workers8
observed two EPR centers (EPR1 and EPR2) having an electron spin of S ¼ 1/2 and a hf structure from a hf interaction with two equiva-lent Ga atoms. The EPR1 center is the same S ¼ 1/2 EPR center, which was assigned to VGa2in Ref.7, while EPR2 is the same center
that was previously observed in neutron-irradiated b-Ga2O3 after
low-temperature x-ray excitation and assigned to a self-trapped hole (STH).9However, calculations in Ref.8do not support quantitatively the models of V2Gaand a STH for EPR1 and EPR2 centers,
respec-tively. So far, all reported EPR centers in irradiated materials show a nearly isotropic hf structure from the interaction with two equivalent Ga atoms with a hf splitting in the range of 13 G. The EPR1 center was later assigned to the VGa(I)–Gaib–VGa(I)complex in the (2)
charge state although its hf splitting is only about two third of the calculated hf values.10
In this Letter, we use EPR to identify the defect that is responsible for the compensation of the shallow donor in b-Ga2O3annealed at
high temperatures under O2ambient. For confirmation of its intrinsic
origin, electron-irradiated materials are also studied. Combining EPR and theoretical calculations, we show that the defect is the double neg-ative charge state of either the Ga vacancy at the tetrahedral site [VGa(I)2 ] or the VGa(I)–Gaib–VGa(I)complex.
The material used in this study is unintentionally doped (UID) b-Ga2O3(001) and (010) substrates grown by the “edge-defined
film-fed growth (EFG)” technique using Ga2O3powder of 5 N grade purity
(99.999%) with the main contaminant being silicon (Si). The melted material is pulled by a seed crystal through a capillary die made of irid-ium with a growth rate of 15 mm/h. The ambient of the growth reactor is a mixed gas of nitrogen (98%) and oxygen (2%). The details of the EFG growth are described elsewhere.5,11 The substrate size is 20 mm 10 mm 0.65 mm. The concentration of different impurities was measured by secondary ion mass spectrometry (SIMS), and the four impurities with high concentrations are (given in units of 1017cm3) Si 1.1–7.1, Fe 0.1–4.2, Sn 0.05–0.2, and Mg 0.03–0.8. Here, the concentration is the lowest at the center of the plate and increases along the length to higher values near the edges (8 mm from the center).5All samples were annealed at 1450C in N2for full activation of the shallow donor. Annealing for reducing the
conductivity was performed at 1450C in O2ambient for 3 h. Some
O2-annealed samples were irradiated by 2 MeV-electrons at room
tem-perature with fluences of 1 1018 and 4 1018cm2. EPR
measure-ments were performed on an X-band (9.4 GHz) E500 Bruker spectrometer equipped with a continuous He-flow cryostat, allowing the regulation of the sample temperature from 4 K to room temperature.
Figure 1(a)shows the EPR spectrum of UID b-Ga2O3after
acti-vating the shallow donor by annealing in N2ambient at 1450C
mea-sured at 15 K for the magnetic field along the [001] direction, i.e., B?(001). In this material, the dominant EPR signal is the Si shallow donor, which can be detected from 4 K to room temperature without problems with microwave coupling in the cavity. In addition, weak sig-nals of the neutral iron center at the octahedral site, Fe3þGa(II),12,13and
the neutral Cr3þcenter14are also observed. In this sample, the Fe3þ
center is compensated by the shallow donor and mostly in the negative charge state Fe2þ. The electrons emitted to the conduction band mini-mum (CBM) from the Fe3þ/Fe2þlevel at room temperature can be captured to other deep levels during the cooling down process in the experiment, activating a small part of Fe in the neutral charge state Fe3þ. This process is possible since the Fe3þ/Fe2þlevel is located a bit below the CBM: 0.6–0.7 eV as measured by absorption15 or 0.78 eV as determined by deep level transient spectroscopy and calculations.16
After annealing at 1450C in O
2ambient for 3 h, the sample
becomes semi-insulating. The signal of the shallow donor vanishes, while the Fe3þsignal recovers and becomes dominating as shown in Fig. 1(b). The line at 2840 G is from Fe3þ at tetrahedral site
Ga(I),12,13which is not detected in samples with weak Fe3þ signals,
such as in N2-annealed or irradiated samples inFigs. 1(a)and1(c),
respectively. This Fe3þGa(I)line is confirmed by comparing the
spec-trum inFig. 1(b)with a spectrum in an Fe-doped sample measured at room temperature where only the Fe3þspectrum is present. In addi-tion, a weak spectrum with a resolved hf structure is observed [see the inset ofFig. 1(b)]. For easy reference, we label this EPR center as IR1. FIG. 1. (a) EPR spectrum measured at 15 K in as-grown b-Ga2O3after activation
of the shallow donor by annealing at 1450C in N2ambient, showing the
dominat-ing signal of the shallow donor, the weak lines from the neutral Fe3þcenter at the octahedral Ga(II)site,12,13and the signal of the neutral Cr3þcenter.14(b) EPR
spec-trum measured at 55 K after annealing the sample at 1450C in O
2ambient for 3 h
showing the signals from Fe3þ, Cr3þ, and a new center IR1. (c) The IR1 spectrum in a sample annealed in O2 and consequently irradiated with a fluence of
4 1018
cm2. The insets show the hf structure of the IR1 center. All the spectra were measured with a field modulation of 2 G and a resolution of 7000 G/4096 data points.
Figure 1(c)shows the IR1 spectrum in an O2-annealed sample and
subsequently irradiated by electrons with a fluence of 4 1018cm2. The intensity ratio between the central line and other hf lines appears to be slightly different compared to the hf structure in the annealed sample. This is an artifact caused by low-resolution scanning (1.7 G/data point) in a single scan, where a peak can be randomly missed in the data collection. It will be shown later in a higher resolu-tion spectrum that the hf structures shown in the inset ofFigs. 1(b) and1(c)have the same intensity ratio between the central line and other hf lines. We have noticed that the intensity of the IR1 spectrum gets stronger significantly after electron irradiation and increases upon increasing the fluence of irradiation from 2 1018to 4 1018cm2. We notice that the Fe3þsignal in irradiated samples is much weaker compared to that in the O2-annealed sample. The reason for this will
be discussed later.
The hf structure of IR1 measured for Bjj[001]is shown inFig. 2 together with the simulated spectrum for a hf interaction between an electron spin S ¼ 1/2 with nuclear spins of two equivalent Ga atoms having a hf splitting of 28.78 G for69Ga (nuclear spin: I ¼ 3/2, nuclear magnetic moment: l/lN¼ 2.01659, and natural abundance: 60.11%)
and 36.55 G for71Ga (I ¼ 3/2, l/lN¼ 2.56227, 39.89%). The ratio
between the hf splitting of71Ga and69Ga, A(71Ga)/A(69Ga) ¼ 36.55/ 28.78 ¼ 1.26998, is in excellent agreement with the ratio between their magnetic moments, l(71Ga)/l(69Ga) ¼ 2.56227/2.01659 ¼ 1.27060. The hf splitting for69Ga is rather isotropic, varying between 28.6 G and 29.1 G. The simulated EPR spectrum inFig. 2is produced using the WINEPR program from Bruker.
The IR1 spectrum is from an electron spin S ¼ 1/2 center. Figure 3shows the angular dependence of the positions of the central line of the IR1 spectrum with the magnetic field rotating in the (010) plane (perpendicular to the b-axis) and in the plane containing the b-axis and the [001] axis. The angular dependence of IR1 can be described by the spin Hamiltonian,
H ¼ lBB g S þX
iIi Ai S: (1)
Here, g is the gyromagnetic tensor, Aiis the hf tensor representing the
hf interaction between the electron spin (S ¼ 1/2) and the nuclear spin Ii(I ¼ 3/2 for both69Ga and71Ga), and lBis the Bohr magneton.
From the best fit to the experimental data, the principal values of the g-tensor are determined to be gxx¼ 2.0160, gyy¼ 2.0386, and
gzz¼ 2.0029. The principal gzzaxis of the g-tensor is 60 from the
[001]direction, and the gyyaxis is parallel to the b-axis of the crystal.
Within the experimental error in hf splitting of 0.5 G, the A-tensors are isotropic: 29.0 G for69Ga and 36.8 G for71Ga. The simulation of the angular dependence of the central line using Eq.(1)and obtained spin Hamiltonian parameters is also shown inFig. 3as a solid curve. The analysis is performed using the Visual-EPR software.17
Figure 4(a)shows the EPR spectrum measured at room tempera-ture in the irradiated sample after annealing at 850C in N
2ambient
for 20 min. At this annealing temperature, the EPR signal of the shal-low donor does not recover and the IR1 signal is still detected. After annealing at 1100C, the IR1 signal almost disappears and a strong
signal of the shallow donor is observed [Fig. 4(b)].
Increasing the annealing temperature to 950C, the concentra-tion of VGais reduced, leading to the activation of a part of the total
concentration of the shallow donor, and its weak signal is detected [Fig. 5(a)]. Annealing at 1100C activates a larger part of the shallow
donor, leading to a further decrease in the Fe3þsignal (due to the
com-pensation by the shallow donor), as shown inFig. 5(b).
A likely reason for the semi-insulating behavior after oxygen treatment is the increase in the concentration of the main compensat-ing acceptor. High-temperature oxygen treatment is expected to create oxygen interstitials and gallium vacancies. The former is a hole trap, while the latter was proposed to be the main compensating accep-tor.6,18 The fact that the EPR signal could be produced by electron FIG. 2. EPR spectrum of the IR1 center (cyan) and the simulation (red) with the hf
interaction with two equivalent Ga atoms for the hf splitting: 28.78 G for69Ga and
36.55 G for71Ga.
FIG. 3. Angular dependence of the positions of the central line of the IR1 spectrum vs the angle of the magnetic field in the plane containing the b-axis and the [001] axis (u¼ 90, h¼ 0–90 as shown in the inset) and in the (010) plane
(u¼ 90–270, h¼ 90). The angle between the principal g
zzaxis of the g-tensor
is 60from the [001]direction (u¼ 150, h¼ 90). The microwave frequency is
irradiation and increases with the increasing fluence of irradiation also points toward a vacancy as origin. The paramagnetic states of defects in b-Ga2O3have recently been studied by theoretical methods.10EPR
centers with S ¼ 1/2 can be expected from the (2) charge state of VGa. It was shown that the unpaired spin associated with VGa2is
local-ized on one oxygen neighbor and has a hf interaction with two gallium neighbors of that oxygen, as is the case in the present measurement. Based on the charge transition levels calculated in Ref.18, it was con-cluded that, in the Fermi-level window typical of unintentionally doped samples, only the tetrahedrally coordinated VGa(I) vacancy is
likely to be in the (2) charge state,10and so we restrict further discus-sions to VGa(I). The hf parameters have been calculated by using a
GGAþU approach with U ¼ 4 eV, yielding 21–22 G. However, the strength of the hf interaction strongly depends on the localization of the defect state, and it is expected that an optimized hybrid functional, which satisfies the generalized Koopmans’ theorem, will provide a more reliable result for that.19The HSE (0.26,0.00) functional used in Ref.18is Koopmans-compliant and has provided accurate hf data for substitutional Mg in b-Ga2O3,20and so we have used it here to
calcu-late the hf parameters of the tetrahedral vacancy. (The details of the computational parameters are given in Ref. 18.) As it was shown,6 VGa(I)can easily diffuse in the (3) charge state, and the lowest energy
configuration is actually at halfway between two equivalent sites, when a Ga(I)atom is at an interstitial position between them (see Fig. 8 of
Ref.10). In the (2) charge state, we find this VGa(I)–Gaib–VGa(I)
con-figuration (referred to as the b-complex) to be more stable than the initial (and final) state of the movement by 0.6 eV. We have calculated the hf parameters of both and find strong isotropic interaction with two Ga atoms in both cases. Assuming69Ga, we obtain 32.9 G for VGa(I)and 31 G for the b-complex. The results for both configurations
are fairly close to the experimental value of 29 G for the IR1 center. A slightly higher calculated hf value, as compared to the experimental one, is expected to be due to the limited size of the supercell that results in a high spin density. From hf parameters, both defect models, the isolated VGa(I)and the b-complex in the double negative charge
state, can be the candidate for the IR1 center.
In UID samples, the Fe3þ center is compensated by the shallow donor and becomes negatively charged (Fe2þ). After annealing in O2,
VGaand the b-complex with an increased concentration can compensate
the shallow donor and pull the Fermi level down. For observation of the signal of VGa(I)2 (or the b-complex) simultaneously with the strong Fe3þ
signal in darkness as seen inFig. 1(b), the Fermi level should be located between the (1j2) level of the b-complex [at EC1.42 eV (Ref. 10)] and the Fe3þ/Fe2þlevel [at EC0.78 eV (Ref.16)].
In irradiated samples, the Fe3þ signal becomes much weaker
[Fig. 1(c)], indicating that the Fermi level lies above the Fe3þ/Fe2þ
level. The (2j3) levels of VGa(I)and the b-complex have been
calcu-lated to be at EC0.67 eV and EC0.74 eV, respectively.10,18It is,
therefore, possible that only the former is in the paramagnetic (2) state and IR1 is due to VGa(I)only. However, since the difference of
0.07 eV between the (2j3) levels is within the computational accu-racy, we cannot conclusively say whether IR1 can be assigned to VGa(I)2
or the b-complex.
Acceptor impurities with a STH, such as Mg,9can also have a hf interaction with two equivalent Ga atoms, but their ground state should be in the lower half of the bandgap. Transition metals (TMs) can give rise to acceptor levels close to the CBM, e.g., Fe. However, TMs have FIG. 4. EPR spectra of irradiated b-Ga2O3after annealing in N2ambient at (a)
850C and (b) 1100C measured at room temperature for Bjj[001]. The
anneal-ing time is 20 min.
FIG. 5. EPR spectrum of irradiated b-Ga2O3after annealing in N2ambient at (a)
950C and (b) 1100C measured at room temperature for Bjja, showing the
increase in the shallow donor signal and the decrease in the Fe3þsignals, while
never shown any observable hf interaction with neighboring atoms in Ga2O3or other semiconductors due to their highly localized unpaired
electrons. Therefore, IR1 should be related to an intrinsic defect. In summary, we have observed the EPR IR1 center in n-type UID b-Ga2O3after annealing in O2ambient at 1450C or electron
irradiation. The center has an electron spin S ¼ 1/2 and shows an almost isotropic hf interaction with the nuclear spins of two equivalent Ga atoms. Comparing the observed hf constant (29 G for69Ga) with the calculated hf parameters for the (2) charge state of the Ga vacancy and its associated defect, VGa(I)–Gaib–VGa(I), and from their
charge transition levels, the IR1 center is assigned to the double nega-tive charge state of either the Ga vacancy at the tetrahedral site, VGa(I)2 ,
or the VGa(I)–Gaib–VGa(I)complex.
Support from DFG Grant No. FR2833/63-1 and HLRN Grant No. hbc00027 for Q.D.H., T.F., and P.D. and from The Institute of Global Innovation Research, Tokyo, University of Agriculture and Technology, Japan, for Y.K. and B.M. is acknowledged.
DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding author upon request.
REFERENCES
1
H. H. Tippins,Phys. Rev.140, A316 (1965).
2K. Shimamura, E. G. Vıllora, K. Domen, K. Yui, K. Aoki, and N. Ichinose,Jpn.
J. Appl. Phys., Part 244, L7 (2005).
3M. Higashiwaki and G. H. Jessen,Appl. Phys. Lett.112, 060401 (2018).
4T. Oshima, K. Kaminaga, A. Mukai, K. Sasaki, T. Masui, A. Kuramata, S.
Yamakoshi, S. Fujita, and A. Ohtomo,Jpn. J. Appl. Phys., Part 152, 051101 (2013).
5
A. Kuramata, K. Koshi, S. Watanabe, Y. Yamaoka, T. Masui, and S. Yamakoshi,Jpn. J. Appl. Phys., Part 155, 1202A2 (2016).
6J. B. Varley, H. Peelaers, A. Janotti, and C. G. Van de Walle,J. Phys.: Condens.
Matter23, 334212 (2011).
7
B. E. Kananen, L. E. Halliburton, K. T. Stevens, G. K. Foundos, and N. C. Giles,
Appl. Phys. Lett.110, 202104 (2017).
8
H. J. von Bardeleben, S. Zhou, U. Gerstmann, D. Skachkov, W. R. L. Lambrecht, Q. D. Ho, and P. Deak,APL Mater.7, 022521 (2019).
9B. E. Kananen, N. C. Giles, L. E. Halliburton, G. K. Foundos, K. B. Chang, and
K. T. Stevens,J. Appl. Phys.122, 215703 (2017).
10
D. Skachkov, W. R. L. Lambrecht, H. J. von Bardeleben, U. Gerstmann, Q. D. Ho, and P. Deak,J. Appl. Phys.125, 185701 (2019).
11N. T. Son, K. Goto, K. Nomura, Q. T. Thieu, R. Togashi, H. Murakami, Y.
Kumagai, A. Kuramata, M. Higashiwaki, A. Koukitu, S. Yamakoshi, B. Monemar, and E. Janzen,J. Appl. Phys.120, 235703 (2016).
12
M. L. Meilman, Sov. Phys. Solid State 11, 1403 (1969) [Fiz. Tverd. Tela 11, 1730 (1969)].
13S. Bhandari and M. E. Zvanut,J. Appl. Phys.127, 065704 (2020). 14
T. H. Yeom, I. G. Kim, S. H. Lee, S. H. Choh, and Y. M. Yu,J. Appl. Phys.93, 3315 (2003).
15S. Bhandari, M. E. Zvanut, and J. B. Varley,J. Appl. Phys.126, 165703 (2019). 16
M. E. Ingebrigtsen, J. B. Varley, A. Y. Kuznetsov, B. G. Svensson, G. Alfieri, A. Mihaila, U. Badst€ubner, and L. Vines,Appl. Phys. Lett. 112, 042104 (2018).
17
Seehttp://www.visual-epr.com/for information on the visual-EPR software.
18P. Deak, Q. D. Ho, F. Seemann, B. Aradi, M. Lorke, and T. Frauenheim,Phys.
Rev. B.95, 075208 (2017).
19
S. Lany and A. Zunger,Phys. Rev. B80, 085202 (2009).