Growth and structural characteristics of metastable β-In2Se3 thin films on H-terminated Si(111) substrates by molecular beam epitaxy
Shen Yi-Fan1, Yin Xi-Bo2, Xu Chao-Fan2, He Jing2, Li Jun-Ye2, Li Han-Dong2, †, Zhu Xiao-Hong1, ‡, Niu Xiao-Bin2
College of Materials Science and Engineering, Sichuan University, Chengdu 610064, China
School of Materials and Energy, University of Electronic Science and Technology of China, Chengdu 610054, China

 

† Corresponding author. E-mail: hdli@uestc.edu.cn xhzhu@scu.edu.cn

Project supported by the National Key Research and Development Program of China (Grant Nos. 2018YFA0306102 and 2018YFA0306703), the National Natural Science Foundation of China (Grant Nos. 61474014 and U1601208), and the Sichuan Science and Technology Program, China (Grant Nos. 2019YJ0202 and 20GJHZ0229).

Abstract

Epitaxial growth and structural characteristics of metastable β-In2Se3 thin films on H-terminated Si(111) substrates are studied. The In2Se3 thin films grown below the β-to-α phase transition temperature (453 K) are characterized to be strained β-In2Se3 mixed with significant γ-In2Se3 phases. The pure-phased single-crystalline β-In2Se3 can be reproducibly achieved by in situ annealing the as-deposited poly-crystalline In2Se3 within the phase equilibrium temperature window of β-In2Se3. It is suggeted that the observed γ-to-β phase transition triggered by quite a low annealing temperature should be a rather lowered phase transition barrier of the epitaxy-stabilized In2Se3 thin-film system at a state far from thermodynamic equilibrium.

1. Introduction

In2Se3 is an intriguing member of the group IIIA–VIA semiconductors with different phases and crystal structures, such as layered α-phase, layered β-phase, and defect wurtzite γ phase.[1] The most interesting and extensively studied phase of In2Se3 is layered α-phase, not only because of its superior optoelectric[2,3] and novel ferroelectric properties,[46] but also due a lot to its phase stability at normal pressure and temperature, thus it is widely fabricated by various thin film growth technologies.[710] In contrast to the α-In2Se3, the β- and γ-phases are metastable at normal pressure and temperature. The metastable In2Se3 phase shows more potential applications in serving as phase-change memorydevices,[1115] though challenges remain in achieving these metastable phases (especially thin film structures for application purpose) with structural stability at normal pressure and temperature. Furthermore, the β-phase shows more excellent performance than the α-phase, especially in the optoelectric field. It has a larger photoresponse wavelength range and shorter photic response time to a ferroelectric polymer or CuInSe2.[16,17] Meanwhile, the β-phase core field effect transistor also shows a higher field effect electron mobility, a higher carrier sheet density and a lower contact resistance than α-phase core field-effect transistor.[18]

Until now, there have been only a few references on the preparation of β- and γ-In2Se3 thin films. The γ-In2Se3 was grown by molecular beam epitaxy (MBE).[19,20] Compared with metastable γ-In2Se3, the β-In2Se3 exhibits a much lower phase equilibrium temperature; therefore, the phase-change memory devices at ambient temperatures are more appealing. It was reported that β-In2Se3 thin films could be synthesized from elemental In and Se sources by pulsed laser deposition (PLD) technique[21] or from γ-InSe compound sources by physical vapor transport (PVT) technique.[22] The β-In2Se3 could also be obtained through chemical routes such as chemical vapor deposition (CVD) technique.[23] Nevertheless, almost no research work has been reported on achieving controllable high-quality β-In2Se3 thin films.

In this work, we attempt to prepare metastable β-In2Se3 thin film by MBE. However, the direct MBE growth of In2Se3 within the phase equilibrium temperature window of β-In2Se3 (below 573 K) leads irreversible γ-phase In2Se3 to form. Hence, a growth strategy involving the low-temperature MBE deposition plus in situ post-deposition annealing is developed in order to suppress undesired γ-phase nucleation and then to achieve phase-pure β-In2Se3. It is demonstrated that the In2Se3 thin films grown from room temperature (RT) to 453 K are poly-crystalline, consisting of strained β-In2Se3 and a small quantity of γ-In2Se3 phases. It is very interesting to note that the post-annealing of the as-deposited poly-crystalline In2Se3 thin layer at a temperature (∼ 573 K) much lower than the β-to-γ phase transition point (623 K) results in the formation of single-crystalline β-In2Se3 without any impure phases, suggesting full strain relaxation of the pristine β-In2Se3 phase and an uncommon phase transition path from γ to β phase in the In2Se3 film. The phase transition kinetics in the epitaxially quenched In2Se3 layer upon being annealed is discussed for depicting the observed uncommon phase transition phenomenon.

2. Experimental procedures

The growth of In2Se3 thin films was carried out in a customized MBE chamber with a base pressure of ∼ 3 × 10−10 mbar (1 bar = 105 Pa). The Si(111) substrates were ultrasonically cleaned in acetone, alcohol, and deionized water in sequence for three rounds. Then they were deoiled in H2SO4 (98%) plus H2O2 (30%) solution. To remove the surface oxidized layer and obtain H-terminated surface, the cleaned Si substrates were etched in 40% HF solution for some minutes. Prior to growth, the H-terminated Si(111) substrates were degassed at a temperature below 453 K for 12 h. Ultrapure In ingot (99.9999+%) and Se pellets (99.999%) were evaporated from a standard Knudsen cell and a cracker cell, respectively. The beam equivalent pressure of In and Se were measured by a beam flux monitor. A large Se:In beam equivalent pressure ratio of ∼ 15 : 1 was employed to ensure the Se-rich environment, which proved vital in our experiments for suppressing the formation of undesired InSe compound during growth. Since the epitaxial growth of layered selenides was fully cation dominated in any case,[24,25] the growth rate of In2Se3 thin films could be determined by the In flux. Accordingly, a growth rate of ∼ 1 nm per min of In2Se3 film was achieved at an In beam equivalent pressure of 10 × 10−8 mbar, which was further confirmed by ex situ thickness measurements. The substrate temperature was measured by a thermocouple mounted near the substrate. The post-annealing procedure of as-deposited In2Se3 was held for 5 min then followed by a programmed quenching process. The surface structure evolution of In2Se3 during deposition was in situ monitored by reflection high energy electron diffraction (RHEED). The surface morphologies, crystallinities, and lattice vibration properties of obtained In2Se3 films were inspected by scanning tunnel microscopy (STM), high resolution x-ray diffraction (HRXRD), and Raman spectroscopy, respectively.

3. Results and discussion

Figure 1(a) shows a set of streaky 1 × 1 RHEED patterns from H–Si(111), indicating smooth morphology of H-passivated Si substrate after being etched in HF acid. Upon deposition, the In–Se film became amorphous, and then gradually crystallized as the growth procedure proceeded. Such an amorphous-to-crystalline transition at the initial stage of epitaxy was also observed during the growth of α-In2Se3 on H–Si(111).[10] The spotty RHEED patterns (Figs. 1(b) and 1(c)) suggest a typical island growth mode of the In–Se compound. Interestingly, the transmitted RHEED patterns manifest a zincblende structure of the islanded surface layer of In–Se. Such an islanded In–Se surface layer was observed to decorate on the growth front in the whole growth process. As is well known, the covalent selenides such as ZnSe usually crystallized into the wurzite or zincblende structure.[26] However, wurzite or zincblende In–Se did not exist in the In–Se phase diagram. Consequently, the epitaxial zincblende In–Se surface layer was inferred to be in a state that is far from thermodynamic equilibrium. The zincblende-type In–Se layer persisted on the surface of the In2Se3 film until the growth temperature was raised to higher than 453 K. Figure 1(d) shows a streaky 1 × 1 pattern of In2Se3 with the growth temperature raised to ∼ 460 K. By carefully comparing the difference in streak spacing between substrate and epifilm, an in-plane lattice constant of ∼ 4 Å was determined, which is nearly equal to the a-axis lattice constant of β-In2Se3. When the growth temperature was further increased to ∼ 523 K, a set of surface patterns gradually appeared with the in-plane lattice constant slightly increasing up to 4.1 Å, as depicted in Figs. 1(e) and 1(f), revealing the nucleation of γ-In2Se3 on the growth front. Although the surface morphology evolution was precisely monitored by RHEED, the structural characteristics of the epilayer beneath the growth front are still lacking. Hence, the HRXRD measurements were carried out in order to inspect thoroughly the structure of the as-deposited film. Figure 1(g) shows two typical XRD θ–2θ spectra for In2Se3 samples grown at relatively low temperatures (RT to 453 K) and high temperatures (453 K–573 K). Only (006) and (0012) diffraction peaks of γ-In2Se3 could be observed in the high temperature grown In2Se3 sample (red curve). While for the low temperature grown one, a series of XRD peaks from other phases were observed to coexist with those of the γ-In2Se3 phase (black curve). The positions of these XRD peaks, shifting to high angles as compared with the positions of the (00n) diffraction peaks of either α-In2Se3 or β-In2Se3, are closer to those of (00n) diffraction peaks of β-In2Se3. We thus speculate that the c-axis compressed β phase and γ phase coexisted in the low temperature grown sample. To support this hypothesis, Raman spectroscopy was used to further examine the lattice bonding nature and strain state of the epi-film. As shown in Fig. 1(h), there are no other peaks except the 150 cm−1 one that was observed in the high temperature grown sample (red curve) and belongs to the zone center mode of unstrained γ phase.[27] While in the low temperature grown sample, Raman peaks from both β- and γ-In2Se3 were observed (black curve). The 148-cm−1 Raman peak of γ-In2Se3 was slightly blue-shifted, suggesting a strained state of γ-In2Se3 phase. However, the peaks at 109 cm−1 and 205 cm−1 matched well to those of unstrained β-In2Se3,[17] suggesting that the c-axis compressive strain, as revealed by XRD measurements, might be effectively accommodated within the van der Waals gaps, thus leaving the strong chemical bonds in the covalent cell of layered β-In2Se3 intact.

Fig. 1. (a) RHEED pattern of H-terminated Si(111) surface, taken along Si[] azimuth; (b) RHEED pattern of ZB-structure after the film growth, taken along Si[1] azimuth; (c) RHEED pattern of ZB-structure after the film growth, taken along 30° rotating from Si[1] azimuth, represented by blue lines; (d) RHEED pattern of growth beginning; (e) RHEED pattern after the film growth; (f) RHEED pattern after film growth taken along 30° rotating from Si[1] azimuth; (g) XRD; (d) Raman spectra of In2Se3 thin films.

The growth of In2Se3 thin films is very complex under the non-equilibrium of thermodynamics in an MBE system, which contains strained β phase and further unexpected γ phase. The unexpected γ phase is stable after growth, indicating that an effective approach to crossing its growth temperature range (453 K–573 K) is required for obtaining the β phase. Therefore, a post-annealing treatment (at a temperature slightly higher than 573 K) followed by a rapid quench is used to drive the change of thermodynamically non-equilibrium state to a thermodynamically equilibrium state. The rapid quench rate is higher than 60 K/s. Figures 2(a)2(d) illustrate the RHEED patterns after growth and post-annealing. The zincblende-structured In–Se surface exists in the whole growth process, while changing to a new pattern after the post-annealing and rapid quench treatment as shown respectively in Figs. 2(a) and 2(b). In Fig. 2(b), there are several bright points in the three stripes, which are an indicator of layered structure. The similar RHEED patterns can be seen along the Si[] azimuth and by rotating 30° from Si[] azimuth after post-annealing as shown in Figs. 2(c) and 2(d). The transformation in RHEED patterns demonstrates that the In2Se3 thin film actually changes into a layered structure by the post-annealing and rapid quench treatment. Furthermore, it is revealed by XRD spectra in Fig. 2(e) that the In2Se3 thin film changes from poly-crystalline to single-crystalline by the post-annealing and rapid quench treatment with all the relevant peaks becoming narrower and showing higher intensity. Likewise, the peak related to γ (006) is not observable after the rapid quench. The lattice constant of rapidly quenched sample is calculated to be 3.99 Å by RHEED patterns, which is very close to that of standard β phase (4.00 Å). Hence, it is supposed that a strain relaxation of β phase occurs after the post-annealing and rapid quench treatment. In addition, the Raman spectra are tested with data shown in Fig. 2(f). The blue shifts of peaks at 109 cm−1 (shifted to 110 cm−1) and 205 cm−1 (shifted to 207 cm−1), both of which are related to the A1(LO + TO) symmetry mode of β phase, demonstrate again the strain relaxation of β phase after the post-annealing and rapid quench process. Meanwhile, the peak at 148 cm−1, which is related to the zone center mode of γ phase, disappears after the heat treatment. As a consequence, both the XRD and Raman spectra evidence the phase transition from a mixture of strained β phase and γ phase to a pure β phase, driven by the “post-annealing and rapid quench”-induced strain relaxation.

Fig. 2. (a) RHEED pattern of ZB-structure; RHEED patterns after the post-annealing, taken along (b) Si[1 1] azimuth, (c) Si[] azimuth, and (d) rotating 30° from Si[] azimuth; (e) XRD spectra of low temperature grown sample and rapidly quenched sample; (f) Raman spectra of low temperature grown sample and rapidly quenched sample; (g) XRD spectra of slowly and rapidly quenched samples; (h) precise XRD spectra of β (006) diffraction peak for slowly and rapidly quenched samples; (i) Raman spectra of slowly and rapidly quenched samples.

To better understand the influence of quench rate, a further experiment is performed and then a representative sample is obtained. This sample experiences a slow quench (The quench rate is lower than 10 K/s) by keeping the other conditions the same as those for the rapidly quenched sample. It is witnessed that similar RHEED patterns are achieved after the post-annealing and quench. But some differences cannot be neglected in the subsequent XRD and Raman characterizations. Figure 2(g) shows the XRD θ–2θ scan patterns of rapidly and slowly quenched In2Se3 samples. The diffraction peaks of slowly quenched sample lie between standard α-In2Se3 (JCPDS 34-0455) and β-In2Se3. All the high index peaks, such as (0015), exhibit obviously a left shift (red line), compared with those correponding peaks of the rapidly quenched sample, which indicates that this sample contains the α-In2Se3. Moreover, all the peaks are broadened and none of the peaks are split in the XRD spectrum, which further indicates that this sample is a mixture of α-In2Se3 and β-In2Se3. The peak broadening of (006) is also observed in a precise scan of omega-2 theta mode as shown in Fig. 2(h). The full width at half maximum (FWHM) of XRD profile for slowly quenched sample is much larger than that of rapidly quenched counterpart and these two peaks are both well distributed. The Raman spectra of the rapidly and slowly quenched samples demonstrate the same trend as shown in Fig. 2(i). A peak broadening at 108 cm−1 is noticed for the slowly quenched sample, which is related to the combination of peaks from α-In2Se3 (104 cm−1) and β-In2Se3 (110 cm−1). Likewise, the peak at 203 cm−1 is indexed to the combination of A1(LO+TO) symmetry mode from α-In2Se3 and β-In2Se3.[18] This result is in good agreement with the XRD spectrum.

After the post-annealing and rapid quench, the sample is further tested with STM. Typical morphology images of rapidly quenched In2Se3 sample are obtained as shown in Figs. 3(a) and 3(b). In the image on a large scale (see Fig. 3(a)), a featured morphology with crystalline grains arranged by steps is clearly seen. The orientation of thin film is consistent in the plane, which is represented by green triangles in Fig. 3(a). There is a flat area in the top part as shown in Fig. 3(b). Accordingly, a distance measurement across the step is obtained on a small scale with data shown in Figs. 3(c) and 3(d). In these images, the step height is measured to be 0.958 nm, which is very close to the thickness of a quintuple layer (1 QL = 0.95 nm–0.97 nm[28]) in the β-In2Se3. Thus, it is demonstrated that the sample after rapid post-annealing process is still composed of layered β-In2Se3 although the overall morphology is rough.

Fig. 3. STM images of 30-nm-thick In2Se3 thin film with post-annealing and rapid quench, showing (a) surface morphology in a large scale of 300 nm × 300 nm; (b) surface morphology in a medium scale of 70 nm × 70 nm, represented by black square in panel (a); (c) surface morphology with a step in small scaled area 42 nm × 30 nm, represented by the black rectangle in panel (b); (d) step height is characterized to be 0.958 nm after correction.

The coexistence of strained β phase and γ phase in the low temperature grown sample is believed to stem from the multiple random nucleations on the initial amorphous growth and in the subsequent nucleus growth process. The strain of β phase in that sample might originate from the H-terminated Si(111) substrate. There is a large lattice mismatch of 4.2% between β-In2Se3 and Si(111) in a axis (β-In2Se3: 4.00 Å, Si(111): 3.84 Å). Under such a circumstance, the lattice constant along the a axis is supposed to be smaller than 4.00 Å for the strained β phase, whereas the lattice constant along the c axis is also smaller as revealed already by the right shift in the XRD spectrum of low temperature grown sample. As a result, the unit-cell volume of strained β phase is compressed by ∼ 3.8% compared with that of the standard β phase. A similar volume decrease can be found in a previous research report,[29] in which it is reported that α phase has an interlayer-glide-driven isosymmetric phase transition to β phase with a volume decrease of ∼ 7% in the high pressure environment (0.8 GPa). Therefore, the strained β phase might be a kind of metastable phase arising from the thermodynamically non-equilibrium growth, and the thin epifilm is fully relaxed to single-crystalline β-In2Se3, which is driven by the strain relaxation after post-annealing. Furthermore, it can be observed in the Raman spectra that the peaks present blue shift, implying a collapse of epitaxial equilibrium and an expansion of volume after post-annealing. Meanwhile, the γ phase also transforms into β-In2Se3 phase after post-annealing, no matter the quench is rapid or slow. It is verified in the Raman spectra, as shown in Fig. 2(h), that the peak at 148 cm−1, which belongs to the γ phase, cannot be observed for the post-annealed sample. Thus, an uncommon γ-to-β phase transition occurs in our In2Se3 thin film. The phase transitions in In2Se3 generally stem from some high temperature or pressure effects. A previous article[12] indicated that the α-In2Se3 phase changes into β-In2Se3 after post-annealing at 573 K and the β-In2Se3 phase is stable. A similar experiment was also conducted by another research group,[18] while the temperature used (623 K) was higher. These two articles demonstrate that the β-In2Se3 is stable when it came from a high temperature environment or a post-annealing process. An uncommon γ-to-β phase transition was similarly observed in the high pressure environment (3.2 GPa to 3.7 GPa), as reported previously,[30] which demonstrated that the β phase was more stable than the γ phase in the high pressure environment and such a phase transition was not reversible. The similar phase transition was also observed with other material systems in the high pressure environment.[31] The uncommon phase transition observed in this work might originate from more complicated reasons, which will be studied systematically in our future work.

The high rate of quench proves to be important to retain the single crystal quality of β-In2Se3 thin films. It is verified that a non-strained-β-to-α-phase transition happens in the slow quench process, which is observed in the precise omega-2 theta scan as already shown in Fig. 2(h). The obvious broadening at (006) peak of the slowly quenched sample indicates that the In2Se3 phase of this sample is an intermediate state during phase transition, coming probably from the short-range diffusion, which is however not a nucleation. This phase transition, which is ascribed to the slow quench with temperature gradually decreasing down to room temperature (passing through the β-to-α phase transition point), can be regarded as a spinodal-type process.[32] The rapid quench ends in less than 5 s, thereby suppressing kinetically the transition from non-strained β to α phase, while the slow quench promotes the short-range diffusion. Therefore, if single-crystalline β-In2Se3 thin films are desired, the rate of quench should be as high as possible to avoid the existence of α phase, thereby demonstrating that the non-strained β-In2Se3 after post-annealing is completely retained by the rapid quench, giving an important reference to obtaining single-crystalline β-In2Se3 thin films.

4. Conclusions

In this work, single-crystalline β-In2Se3 thin films are obtained by a low temperature MBE growth (below 453 K) followed by post-annealing and rapid quench. When the growth temperature is below 453 K, the samples are In2Se3 thin films with strained β-In2Se3 and γ-In2Se3 co-existing. When the temperature is above 453 K but lower than 573 K, the samples are completely γ-In2Se3 thin films. Very importantly, beginning with a low temperature growth (below 453 K) and then post-annealing at a temperature slightly higher than 573 K together with rapid quench, a single-crystalline β-In2Se3 thin film sample can be obtained. It is speculated that the transition from a mixture of strained β and γ phases to a pure β phase is driven by the strain relaxation with post-annealing and rapid quench. Eventually, a high rate of quench is the key to avoiding the spinodal-type-β-to-α-phase transition.

Reference
[1] Han G Chen Z G Drennan J Zou J 2014 Small 10 2747
[2] Island J O Blanter S I Buscema M Van der Zant H S J Castellanos-Gomez A 2015 Nano Lett. 15 7853
[3] Jacobs-Gedrim R B Shanmugam M Jain N Durcan C A Murphy M T Murray T M Matyi R J Moore R L Yu B 2014 ACS Nano 8 514
[4] Ding W J Zhu J B Wang Z Gao Y F Xiao D Gu Y Zhang Z Y Zhu W G 2017 Nat. Commun. 8 14956
[5] Cui C J Hu W J Yan X X Addiego C Gao W P Wang Y Wang Z Li L Z Cheng Y C Li P Zhang X X Alshareef H N Wu T Zhu W G Pan X Q Li L J 2018 Nano Lett. 18 1253
[6] Poh S M Tan S J R Wang H Song P Abidi I H Zhao X X Dan J D Chen J S Luo Z T Pennycook S J Castro Neto A H Loh K P 2018 Nano Lett. 18 6340
[7] Zhai T Y Fang X S Liao M Y Xu X J Li L Liu B D Koide Y Ma Y Yao J N Bando Y Golberg D 2010 ACS Nano 4 1596
[8] Ho C H Lin C H Wang Y P Chen Y C Chen S H Huang Y S 2013 ACS Appl. Mater. Interfaces 5 2269
[9] Li Q L Liu C H Nie Y T Chen W H Gao X Sun X H Wang S D 2014 Nanoscale 6 14538
[10] Li H D Ren W Y Wang G Y Gao L Peng R M Li H Zhang P Y Shafa M Tong X Luo S Y Zhou Z H Ji H N Wu J Niu X B Wang Z M 2016 J. Phys. D: Appl. Phys. 49 145108
[11] Chen J Du G Liu X Y 2015 Chin. Phys. 24 057702
[12] Choi M S Cheong B K Ra C H Lee S Bae J H Lee S Lee G D Yang C W Hone J Yoo W J 2017 Adv. Mater. 29 1703568
[13] Pandiana M Matheswarana P Gokul B Sathyamoorthy R Asokan K 2018 Appl. Surf. Sci. 449 55
[14] Lee H Kang D H Tran L 2005 Mater. Sci. Eng. 119 196
[15] Jin B Kang D Kim J Meyyappan M Lee J S 2013 J. Appl. Phys. 113 164303
[16] Wu G J Wang X D Wang P Huang H Chen Y Sun S Shen H Lin T Wang J L Zhang S T Bian L F Sun J L Meng X J Chu J H 2016 Nanotechnology 27 364002
[17] Zheng Z Q Yao J D Yang G W 2017 ACS Appl. Mater. Interfaces 9 7288
[18] Feng W Gao F Hu Y X Dai M J Liu H Wang L F Hu P A 2018 ACS Appl. Mater. Interfaces 10 27584
[19] Chaiken A Nauka K Gibson G A Lee H Yang C C Wu J Ager J W Yu K M Walukiewicz W 2003 J. Appl. Phys. 94 2390
[20] Okamoto T Yamada A Konagai M 1997 J. Cryst. Growth 175--176 1045
[21] Zheng Z Q Yao J D Xiao J Yang G W 2016 J. Mater. Chem. 4 8094
[22] Balakrishnan N Staddon C R Smith E F Stec J Gay D Mudd G W Makarovsky O Kudrynskyi Z R Kovalyuk Z D Eaves L Patanè A Beton P H 2016 2D Mater. 3 025030
[23] Zhou S Tao X Gu Y 2016 J. Phys. Chem. 120 4753
[24] Ren W Y Li H D Gao L Li Y Zhang Z Y Long C J Ji H N Niu X B Lin Y Wang Z W 2017 Nano Res. 10 247
[25] Li H D Yu S P Li Y Channa A I Ji H N Wu J Niu X B Wang Z M 2019 Appl. Phys. Lett. 115 041602
[26] Li H D Wang Z Y Guo X Wong T L Wang N Xie M H 2011 Appl. Phys. Lett. 98 043104
[27] Yuan Y F Cao X R Sun Y Su J Liu C M Cheng L Yuan L H Zhang H Li J 2017 RSC Adv. 7 46431
[28] Zhang F Wang Z Dong J Y Nie A M Xiang J Y Zhu W G Liu Z Y Tao C G 2019 ACS Nano 13 8004
[29] Ke F Liu C L Gao Y Zhang J K Tan D Y Han Y H Ma Y Z Shu J F Yang W G Chen B Mao H K Chen X J Gao C X 2014 Appl. Phys. Lett. 104 212102
[30] Rasmussen A M Mafi E Zhu W G Gu Y McCluskey M D 2016 High Press. Res. 36 549
[31] Huang Y P Huang X L Wang X Zhang W T Zhou D Zhou Q Liu B B Cui T 2019 Chin. Phys. 28 096402
[32] Li H D Zhang X N Zhang Z Mei Z X Du X L Xue Q K 2007 J. Appl. Phys. 101 106102