We report on the van der Waals epitaxy of high-quality single-crystalline AlN and the demonstration of AlGaN tunnel junction deep-ultraviolet light-emitting diodes directly on graphene, which were achieved by using plasma-assisted molecular beam epitaxy. It is observed that the substrate/template beneath graphene plays a critical role in governing the initial AlN nucleation. In situ reflection high energy electron diffraction and detailed scanning transmission electron microscopy studies confirm the epitaxial registry of the AlN epilayer with the underlying template. Detailed studies further suggest that the large-scale parallel epitaxial relationship for the AlN epilayer grown on graphene with the underlying template is driven by the strong surface electrostatic potential of AlN. The realization of high-quality AlN by van der Waals epitaxy is further confirmed through the demonstration of AlGaN deep-ultraviolet light-emitting diodes operating at ∼260 nm, which exhibit a maximum external quantum efficiency of 4% for an unpackaged device. This work provides a viable path for the van der Waals epitaxy of ultra-wide bandgap semiconductors, providing a path to achieve high performance deep-ultraviolet photonic and optoelectronic devices that were previously difficult.

III-nitride semiconductors have emerged as the materials of choice for a broad range of device applications, including optoelectronics in the ultraviolet (UV) and visible regions and high power and high frequency electronics.1–5 They have also shown tremendous promise for applications in quantum information and renewable energy.6–8 To date, however, due to the lack of native substrates at the economic scale, conventional III-nitride devices are often heteroepitaxially grown on foreign substrates, such as Al2O3, Si, and SiC.9 The resulting large lattice mismatch and large difference in thermal expansion coefficients produce stressed materials with high density threading dislocations, stacking faults, and point defects,10,11 which severely limit the device performance. To date, the best reported external quantum efficiency (EQE) for AlGaN deep-UV light-emitting diodes (LEDs) is limited to ∼11% at ∼265 nm12 and is well below 1% for LEDs operating at shorter wavelengths.4 

Issues associated with lattice and thermal mismatches can be potentially addressed by using two-dimensional (2D) material assisted van der Waals epitaxy (vdWE).13–18 For example, using graphene (Gr) as the buffer layer, conventional heteroepitaxy can be substituted by vdWE, wherein the effect of the lattice mismatch between the III-nitride epilayer and the Gr/host substrate is no longer significant, due to the weak atomic interaction at the 2D/3D interface.17,19–21 Significantly, the device heterostructures grown on Gr can be released and transferred onto any other foreign substrates or templates, including metals and polymers, providing distinct opportunities for heterogeneous integration, efficient thermal management, and flexible/conformal device and system fabrication.14–16,22 The dangling-bond-free inert surface of pristine Gr, however, suppresses the nucleation of III-nitrides,16 severely limiting the growth of high-quality continuous epilayers. In this regard, defects have been intentionally created on Gr to promote the nucleation and formation of III-nitride epilayers by metal-organic chemical vapor deposition (MOCVD).17,23–27 Compared to MOCVD, molecular beam epitaxy (MBE) is a non-equilibrium growth process, which provides unique opportunities to achieve high-quality GaN and AlN epilayers directly on pristine Gr. To date, however, there were only a few reports on Gr-assisted growth of III-nitride nanowires by MBE.28,29 The realization of high-quality III-nitride epilayers, particularly AlN, on pristine Gr has remained elusive.

In this Letter, we report on the vdWE and characterization of AlN grown on Gr by using plasma-assisted MBE. High-quality single crystalline AlN grown on Gr was achieved through controlled nucleation by migration enhanced epitaxy (MEE) on Gr. The effect of the crystallographic structure underneath Gr on the vdWE of AlN is explored using Gr/sapphire and Gr/AlN templates as substrates. The epitaxial registry from the underlying template is confirmed by reflection high energy electron diffraction (RHEED) and scanning transmission electron microscopy (STEM) analysis. We have further demonstrated AlGaN deep-UV LEDs on Gr, which operate at ∼260 nm and exhibit a maximum EQE of 4% for an unpackaged device.

Commercial single monolayer (ML) Gr grown on Cu(111) by CVD was first transferred onto the target sapphire and AlN template substrates (supplementary material Fig. S1). Details for the wet transfer and subsequent cleaning processes are described in the supplementary material. As illustrated in Fig. 1(a), the scanning electron microscopy (SEM) image shows a flat and clean surface after Gr transfer. The dark lines and areas in Fig. 1(a) originate from the boundaries and the nucleation of Gr on copper, respectively [supplementary material Fig. S1(a)]. The atomic force microscopy (AFM) image of the Gr/AlN template, illustrated in Fig. 1(b), shows a root mean square (RMS) roughness of 0.72 nm for a 5 × 5 μm2 area. The surface largely maintains the features of the underlying AlN template, except for the Gr boundaries. Figure 1(c-iii) presents the typical Raman spectrum of the transferred Gr with two characteristic peaks at the G band (1586 cm−1) and 2D band (2682 cm−1), while the defect-related D band peak (1347 cm−1) is negligible. The intensity ratio of 2D peak to G peak (I2D/IG) is around 2, confirming that it is single layer Gr.30 AlN/Gr samples were subsequently grown using a Veeco GENxplor MBE system equipped with a radio frequency (RF) plasma-assisted nitrogen source. The Gr/substrates were baked and outgassed at 200 °C and 600 °C in the MBE load-lock chamber and preparation chamber for 2 h, respectively, to obtain a clean surface. The growth conditions included a nitrogen flow rate of 0.3 standard cubic centimeters per minute (sccm), a RF plasma forward power of 350 W, an Al beam equivalent pressure (BEP) in the range of 6.0 × 10−8 Torr to 1.0 × 10−7 Torr, and a substrate temperature of 900 °C (thermocouple reading). The initial AlN nucleation layer was grown using MEE, with a thickness of ∼30 nm, followed by the growth of the 500 nm AlN epilayer. The entire growth procedure was monitored in situ using RHEED. All the AlN samples were grown on pristine transferred Gr. A thin (several MLs) Al layer was first deposited on Gr before striking the N2-plasma.

FIG. 1.

(a) SEM and (b) AFM images of Gr transferred from copper foil onto the AlN template. (c) Comparison of Raman spectra at different stages: (i) sapphire, (ii) AlN template, (iii) 1 ML Gr/AlN template, (iv) AlN/Gr/AlN template, and (v) residual Gr/AlN template after AlN peeling-off.

FIG. 1.

(a) SEM and (b) AFM images of Gr transferred from copper foil onto the AlN template. (c) Comparison of Raman spectra at different stages: (i) sapphire, (ii) AlN template, (iii) 1 ML Gr/AlN template, (iv) AlN/Gr/AlN template, and (v) residual Gr/AlN template after AlN peeling-off.

Close modal

To elucidate the effect of the underlying substrate on Gr assisted vdWE, studies of the initial nucleation of AlN on Gr/sapphire and Gr/AlN templates were performed. The RHEED patterns recorded at different growth stages are shown in Figs. 2(a)–2(f). Before the deposition of AlN, blurry broad RHEED patterns of Gr can be observed along any of the azimuths [Figs. 2(a)–2(c)], due to the mosaic morphology of the transferred Gr layer. The weak streaky RHEED patterns of the underlying substrate are observable as well [marked as red arrows in Figs. 2(a)–2(c)]. The MEE technique (alternately supplying Al and N atoms) was used to improve the nucleation morphology [supplementary material Fig. S2(a)].31 For growth on Gr/sapphire, during the nucleation process, the RHEED patterns gradually changed from blurry streaky to dashed spotty. Unexpectedly, two sets of AlN RHEED patterns appeared at the same azimuth [Fig. 2(d)], suggesting that the nucleation layer has two kinds of in-plane domains, even though the surface has a small RMS roughness of ∼0.75 nm [Fig. 2(g)]. In the case of the Gr/AlN template, the RHEED patterns transformed from blurry streaky into dashed streaky, and only one set of RHEED patterns appeared along the [112¯0] and [11¯00] azimuths [marked as blue arrows in Figs. 2(e) and 2(f)]. The alignment of RHEED patterns is consistent with that of the AlN underneath Gr, indicating that the top AlN nucleation layer has the same crystallographic orientation with the bottom AlN template. Figure 2(h) shows that a high density of large-size AlN nuclei formed on Gr/AlN template, with a corresponding RMS roughness of ∼2.98 nm. The height fluctuations of the nuclei grown by MEE were reduced significantly compared to the one grown using the conventional method (RMS ∼ 5.75 nm) [supplementary material Fig. S2(b)]. Shown in Fig. 2(i), the nucleation layer grown directly on the AlN template has a smoother surface (RMS ∼ 0.98 nm), while the density of larger-size nuclei is much lower than that on the Gr/AlN template. This result suggests that the AlN nucleation is suppressed by the inert surface of Gr.17 The Raman spectrum acquired after the AlN nucleation exhibits clear peaks of the D band and G band with a dramatic reduction of the 2D band peak [Fig. 1(c-iv)]. It confirms the presence of Gr after the epitaxy of AlN although the defects increased on the Gr surface after AlN epitaxy.

FIG. 2.

RHEED patterns of (a)–(c) before and (d)–(f) after AlN nucleation on Gr along the same azimuth as labeled. The red and blue arrows correspond to the RHEED patterns of the host substrates and AlN nucleation layer, respectively, while the blurry broad RHEED patterns in (a)–(c) come from the transferred Gr. AFM images of the AlN nucleation layer grown on (g) Gr/sapphire, (h) Gr/AlN template, and (i) AlN template using the MEE technique.

FIG. 2.

RHEED patterns of (a)–(c) before and (d)–(f) after AlN nucleation on Gr along the same azimuth as labeled. The red and blue arrows correspond to the RHEED patterns of the host substrates and AlN nucleation layer, respectively, while the blurry broad RHEED patterns in (a)–(c) come from the transferred Gr. AFM images of the AlN nucleation layer grown on (g) Gr/sapphire, (h) Gr/AlN template, and (i) AlN template using the MEE technique.

Close modal

Based on the atomic interaction model proposed by Kim et al.20 and Kong et al.,21 the significantly different nucleation behaviors on Gr/sapphire and Gr/AlN templates can be qualitatively explained by the crystallographic difference between sapphire (R-3c) and AlN (P63mc). III-nitrides are strongly polarized materials, with large spontaneous polarization along the c-axis, which introduces considerable surface electrostatic potential fluctuation.21 Due to the difference in surface electrical properties, such as the surface dipole moment and piezoelectric response coefficient (|e31|AlN ∼ 0.58 C/m2 vs |e31|sapphire ∼0.1 C/m2),32,33 the potential fluctuation on the sapphire surface would be significantly weaker than that on AlN. The potential fluctuation from the underlying AlN surface can transmit through 1-ML-Gr, and thus, the impinging adatoms on the 1-ML-Gr/AlN surface can be stabilized at the energetically favorable sites, whereas it is less favorable for adatoms to follow the epitaxial registry of the underlying sapphire substrate due to the much weaker surface potential. To demonstrate this mechanism, we also performed the same growth on the Gr/GaN template and observed a similar nucleation behavior as on the Gr/AlN template. It is worthwhile to mention that the heteroepitaxy of AlN on Gr/sapphire further contributes to uncontrollable initial nucleation, compared to that on Gr/AlN. This issue can be alleviated by introducing defects on Gr.23–25 

The SEM images of 500-nm-thick AlN epilayers grown on Gr/sapphire and Gr/AlN templates are shown in Figs. 3(a) and 3(b), respectively. The AlN epilayer grown on Gr/sapphire has many grain boundaries, whereas the AlN epilayer grown on the Gr/AlN template shows a continuous and smooth surface. The incomplete coalescence for AlN/Gr/sapphire is mainly due to the non-uniform in-plane domains of the nucleation layer, which was confirmed by the RHEED patterns [Fig. 2(d)]. For the AlN grown on the AlN template with and without the Gr interlayer, the main difference lies in the initial nucleation stage [Figs. 2(h) and 2(i)], while the subsequent epilayer growth yields a similar smooth surface for both cases [supplementary material Fig. S3]. X-ray diffraction (XRD) phi-scan of the (102) plane was carried out on the AlN/Gr/AlN template, wherein a single set of diffraction peaks repeat every 60° [inset of Fig. 3(b)], showing the large-scale parallel epitaxial growth of AlN on the Gr/AlN template. The electron backscatter diffraction (EBSD) map further confirmed a (0001) wurtzite single-crystalline orientation in the AlN epilayer, as indicated in red by the inverse pole figure (IPF) color triangle [Fig. 3(c)]. Together with the RHEED pattern observations, it is evident that the AlN epilayer and AlN template beneath Gr exhibit an excellent parallel epitaxial relationship of (0001) [112¯0](epilayer)ǁGrǁ(0001) [112¯0](template). Using Ni as a stressor,16,20 the AlN epilayer on Gr can be subsequently exfoliated. The Raman peaks of Gr can be clearly detected on the residual substrate after AlN peeling-off, shown in Fig. 1(c-v), demonstrating that the Gr layer still exists after growth and exfoliation, offering an opportunity to reuse the Gr/AlN template.16 

FIG. 3.

SEM images of AlN grown on (a) Gr/sapphire and (b) Gr/AlN template. The inset in (b) shows the XRD phi-scan of the (102) plane of the AlN/Gr/AlN template. (c) Corresponding EBSD map of (b).

FIG. 3.

SEM images of AlN grown on (a) Gr/sapphire and (b) Gr/AlN template. The inset in (b) shows the XRD phi-scan of the (102) plane of the AlN/Gr/AlN template. (c) Corresponding EBSD map of (b).

Close modal

Cross-sectional high-angle annular dark field STEM (HAADF-STEM) was collected using a Cs aberration corrected JEOL 3100R05 microscope (300 keV, 22 mrad) and a 120 mm camera length. The HAADF-STEM images highlight the AlN/Gr/AlN structure showing the interface density consistent with the lighter at. wt. of carbon [Figs. 4(a) and 4(b)]. The atomic-resolution STEM reveals the highly parallel epitaxial relationship of the AlN epilayer with the AlN template beneath Gr exhibiting the stacking manner of Al atoms is consistent on both sides of the Gr layer nearly without any shift or rotation [Fig. 4(b)]. Fast Fourier transform (FFT) of the HAADF-STEM image in the interface region illustrates a highly single crystalline wurtzite structure, indicating that AlN/Gr/AlN share the same crystallographic orientation [Fig. 4(c)]. Although a carbon layer is clearly visible in the STEM image (a projection of the crystal), the Gr lattice was not noticeable and the layer thickness corresponds to ∼1 nm—likely due to the defects generated during the growth and the atomic steps on the AlN template surface. The XRD rocking curves of the (002) and (102) planes for the 1-μm-thick AlN template and 500-nm-thick AlN grown on the AlN template with and without the Gr interlayer are plotted in Fig. 4(d). The full-width-at-half-maximum (FWHM) values can be used to estimate dislocation densities in AlN through the mosaic model.34 The edge dislocation density decreases from 2.1 × 1010 cm−2 to 1.3 × 1010 cm−2, i.e., a 38% reduction by introducing Gr, while the screw dislocation density is maintained at the same level of ∼1.4 × 107 cm−2. The reduction of edge dislocations is likely attributed to (i) the lattice continuity of conventional epitaxy that is interrupted by the Gr layer and van der Waals interface, providing a possible path to block the propagation of edge dislocations from the underlying AlN template; (ii) the three-dimensional nucleation, which is often used in Al(Ga)N epilayer growth,35 offering another approach to reduce the edge dislocations. These two filtering effects of the threading dislocations were observed in the bright field STEM image of the AlN/Gr/AlN interface (supplementary material Fig. S4). The photoluminescence (PL) spectrum of the AlN/Gr/AlN template shows strong near-band-edge emission at ∼210 nm without any significant defect emission (supplementary material Fig. S5), demonstrating the high-quality AlN epilayer grown on Gr.

FIG. 4.

(a) HAADF-STEM overview of the cross-sectional AlN/Gr/AlN structure showing a darker contrast interface consistent with a carbon layer. (b) Atomic-resolution HAADF-STEM of the interface exhibiting the parallel epitaxial relationship of the top AlN. Al atoms (purple balls) are embedded in (b) to show the atomic stacking manner. (c) FFT of the atomic-resolution image presented in (b), indicating the crystallographic orientation of the AlN epilayer consistent with that of the underlying AlN template. (d) XRD rocking curves of the (002) and (102) planes for the AlN template, AlN/AlN template, and AlN/Gr/AlN template.

FIG. 4.

(a) HAADF-STEM overview of the cross-sectional AlN/Gr/AlN structure showing a darker contrast interface consistent with a carbon layer. (b) Atomic-resolution HAADF-STEM of the interface exhibiting the parallel epitaxial relationship of the top AlN. Al atoms (purple balls) are embedded in (b) to show the atomic stacking manner. (c) FFT of the atomic-resolution image presented in (b), indicating the crystallographic orientation of the AlN epilayer consistent with that of the underlying AlN template. (d) XRD rocking curves of the (002) and (102) planes for the AlN template, AlN/AlN template, and AlN/Gr/AlN template.

Close modal

To further confirm the high material quality of AlN, AlGaN-based deep-UV LED structures were grown on the AlN/Gr/AlN template. The device schematic is shown in Fig. 5(a). 500-nm-thick n-AlGaN was first grown as the bottom contact layer, which is followed by the AlGaN quantum well active region designed for emission at 260 nm. A graded p-AlGaN layer was then grown with the Al composition varying from 85% to 65%. A p-AlGaN layer with a thickness of 50 nm was further grown, above which a 2.5-nm-thick GaN layer and a 50-nm-thick n-AlGaN top contact layer were grown. Detailed growth conditions and the tunnel junction (TJ) design are described in previous publications.12,36 The electroluminescence (EL) characteristic of the deep-UV LED on the AlN/Gr/AlN template is shown in Fig. 5(b). The device area size is 40 × 40 μm2. It exhibits strong and highly stable emission at a peak wavelength of ∼260 nm, with no obvious shift in the peak position with varying injection current densities (J), indicating a substantial stress relaxation due to the vdWE.25 With an injection current density of 5 A/cm2, the peak intensity of the EL spectrum from a deep-UV LED with the Gr interlayer is nearly twice as strong as that without the Gr interlayer grown under identical conditions [inset of Fig. 5(b)]. The output power of the deep-UV LEDs was measured directly from the backside of the as-fabricated devices (unpackaged) using pulsed voltage (with a repetition rate of 10 kHz and a 1% duty cycle) to minimize the heating effects (the details are described in the supplementary material).12,36 As shown in Fig. 5(c), the maximum measured EQE is 4% for deep-UV LEDs on Gr, which is significantly higher than that of the LEDs without the Gr interlayer (3%), and it is among the best reported values for AlGaN deep-UV LEDs operating at this wavelength.4 The improved device performance may originate from the reduction of dislocations and defects, as well as the relaxation of strain.24,25,35 It is worthwhile to mention that these demonstrations are our initial studies of AlGaN TJ deep-UV LEDs. With further optimizations of the design and growth parameters, AlGaN TJ deep-UV LEDs with significantly improved performance have been demonstrated recently.12 

FIG. 5.

(a) Schematic of the AlGaN deep-UV LED structure grown on the AlN/Gr/AlN template. (b) EL spectra of deep-UV LED on the AlN/Gr/AlN template. The inset shows the EL spectra of the deep-UV LEDs with and without the Gr interlayer. (c) EQE vs current density of deep-UV LEDs with and without the Gr interlayer.

FIG. 5.

(a) Schematic of the AlGaN deep-UV LED structure grown on the AlN/Gr/AlN template. (b) EL spectra of deep-UV LED on the AlN/Gr/AlN template. The inset shows the EL spectra of the deep-UV LEDs with and without the Gr interlayer. (c) EQE vs current density of deep-UV LEDs with and without the Gr interlayer.

Close modal

In summary, we have demonstrated the vdWE of high-quality single-crystalline AlN and AlGaN deep-UV LEDs on Gr using plasma-assisted MBE. It is observed that the substrate/template beneath Gr plays a critical role in governing the initial AlN nucleation to form a high density of large-size nuclei with uniform in-plane domains. The strong surface electrostatic potential of the underlying AlN template enables a large-scale parallel epitaxial relationship for AlN/Gr/AlN. This epitaxial registry is further verified by detailed RHEED and STEM studies. The high-quality AlN grown on Gr is also confirmed by the demonstration of AlGaN TJ deep-UV LEDs operating at ∼260 nm, which exhibit a maximum EQE of 4% for the direct on-wafer measurement. Such devices can also be readily peeled off, thereby providing efficient thermal management and enabling high efficiency flexible deep-UV LEDs, high performance UV laser diodes, and high frequency and high power electronics.

See the supplementary material for the (1) wet transfer process of Gr, (2) schematic of the MEE technique, (3) surface morphology of AlN grown on the AlN template with and without the Gr interlayer, (4) evolution of dislocations at the AlN/Gr/AlN interface, and (5) PL spectrum of the AlN/Gr/AlN template.

This work was supported by the Army Research Office (ARO) under Contract No. W911NF19P0025 and the Blue Sky Program in the College of Engineering at the University of Michigan.

1.
I.
Akasaki
,
Rev. Mod. Phys.
87
,
1119
(
2015
).
2.
S.
Nakamura
,
Rev. Mod. Phys.
87
,
1139
(
2015
).
3.
Y.
Sun
,
K.
Zhou
,
Q.
Sun
,
J.
Liu
,
M.
Feng
,
Z.
Li
,
Y.
Zhou
,
L.
Zhang
,
D.
Li
, and
S.
Zhang
,
Nat. Photonics
10
,
595
(
2016
).
4.
D.
Li
,
K.
Jiang
,
X.
Sun
, and
C.
Guo
,
Adv. Opt. Photonics
10
,
43
(
2018
).
5.
E. A.
Jones
,
F. F.
Wang
, and
D.
Costinett
,
IEEE J. Emerging Sel. Top. Power Electron.
4
,
707
(
2016
).
6.
S.
Deshpande
,
J.
Heo
,
A.
Das
, and
P.
Bhattacharya
,
Nat. Commun.
4
,
1675
(
2013
).
7.
X.
Sun
,
P.
Wang
,
B.
Sheng
,
T.
Wang
,
Z.
Chen
,
K.
Gao
,
M.
Li
,
J.
Zhang
,
W.
Ge
,
Y.
Arakawa
,
B.
Shen
,
M.
Holmes
, and
X.
Wang
,
Quantum Eng.
1
,
e20
(
2019
).
8.
B.
Zhou
,
X.
Kong
,
S.
Vanka
,
S.
Chu
,
P.
Ghamari
,
Y.
Wang
,
N.
Pant
,
I.
Shih
,
H.
Guo
, and
Z.
Mi
,
Nat. Commun.
9
,
3856
(
2018
).
9.
D. A.
Neumayer
and
J. G.
Ekerdt
,
Chem. Mater.
8
,
9
(
1996
).
11.
X.
Wu
,
P.
Fini
,
E.
Tarsa
,
B.
Heying
,
S.
Keller
,
U.
Mishra
,
S.
DenBaars
, and
J.
Speck
,
J. Cryst. Growth
189–190
,
231
(
1998
).
12.
A.
Pandey
,
W.
Shin
,
J.
Gim
,
R.
Hovden
, and
Z.
Mi
,
Photonics Res.
8
,
331
(
2020
).
13.
J.
Yu
,
L.
Wang
,
Z.
Hao
,
Y.
Luo
,
C.
Sun
,
J.
Wang
,
Y.
Han
,
B.
Xiong
, and
H.
Li
, “Van der Waals epitaxy of III-nitride semiconductors based on 2D materials for flexible applications,”
Adv. Mater.
(published online 2019).
14.
K.
Chung
,
C.-H.
Lee
, and
G.-C.
Yi
,
Science
330
,
655
(
2010
).
15.
Y.
Kobayashi
,
K.
Kumakura
,
T.
Akasaka
, and
T.
Makimoto
,
Nature
484
,
223
(
2012
).
16.
J.
Kim
,
C.
Bayram
,
H.
Park
,
C.-W.
Cheng
,
C.
Dimitrakopoulos
,
J. A.
Ott
,
K. B.
Reuter
,
S. W.
Bedell
, and
D. K.
Sadana
,
Nat. Commun.
5
,
4836
(
2014
).
17.
Y.
Qi
,
Y.
Wang
,
Z.
Pang
,
Z.
Dou
,
T.
Wei
,
P.
Gao
,
S.
Zhang
,
X.
Xu
,
Z.
Chang
,
B.
Deng
,
S.
Chen
,
Z.
Chen
,
H.
Ci
,
R.
Wang
,
F.
Zhao
,
J.
Yan
,
X.
Yi
,
K.
Liu
,
H.
Peng
,
Z.
Liu
,
L.
Tong
,
J.
Zhang
,
Y.
Wei
,
J.
Li
, and
Z.
Liu
,
J. Am. Chem. Soc.
140
,
11935
(
2018
).
18.
T.
Li
,
C.
Liu
,
Z.
Zhang
,
B.
Yu
,
H.
Dong
,
W.
Jia
,
Z.
Jia
,
C.
Yu
,
L.
Gan
, and
B.
Xu
,
AIP Adv.
8
,
045105
(
2018
).
19.
Y.
Alaskar
,
S.
Arafin
,
D.
Wickramaratne
,
M. A.
Zurbuchen
,
L.
He
,
J.
McKay
,
Q.
Lin
,
M. S.
Goorsky
,
R. K.
Lake
, and
K. L.
Wang
,
Adv. Funct. Mater.
24
,
6629
(
2014
).
20.
Y.
Kim
,
S. S.
Cruz
,
K.
Lee
,
B. O.
Alawode
,
C.
Choi
,
Y.
Song
,
J. M.
Johnson
,
C.
Heidelberger
,
W.
Kong
,
S.
Choi
,
K.
Qiao
,
I.
Almansouri
,
E. A.
Fitzgerald
,
J.
Kong
,
A. M.
Kolpak
,
J.
Hwang
, and
J.
Kim
,
Nature
544
,
340
(
2017
).
21.
W.
Kong
,
H.
Li
,
K.
Qiao
,
Y.
Kim
,
K.
Lee
,
Y.
Nie
,
D.
Lee
,
T.
Osadchy
,
R. J.
Molnar
,
D. K.
Gaskill
,
R. L.
Myers-Ward
,
K. M.
Daniels
,
Y.
Zhang
,
S.
Sundram
,
Y.
Yu
,
S.-H.
Bae
,
S.
Rajan
,
Y.
Shao-Horn
,
K.
Cho
,
A.
Ougazzaden
,
J. C.
Grossman
, and
J.
Kim
,
Nat. Mater.
17
,
999
(
2018
).
22.
H.
Ci
,
H.
Chang
,
R.
Wang
,
T.
Wei
,
Y.
Wang
,
Z.
Chen
,
Y.
Sun
,
Z.
Dou
,
Z.
Liu
,
J.
Li
,
P.
Gao
, and
Z.
Liu
,
Adv. Mater.
31
,
1901624
(
2019
).
23.
Z.
Chen
,
X.
Zhang
,
Z.
Dou
,
T.
Wei
,
Z.
Liu
,
Y.
Qi
,
H.
Ci
,
Y.
Wang
,
Y.
Li
,
H.
Chang
,
J.
Yan
,
S.
Yang
,
Y.
Zhang
,
J.
Wang
,
P.
Gao
,
J.
Li
, and
Z.
Liu
,
Adv. Mater.
30
,
1801608
(
2018
).
24.
H.
Chang
,
Z.
Chen
,
W.
Li
,
J.
Yan
,
R.
Hou
,
S.
Yang
,
Z.
Liu
,
G.
Yuan
,
J.
Wang
, and
J.
Li
,
Appl. Phys. Lett.
114
,
091107
(
2019
).
25.
Z.
Chen
,
Z.
Liu
,
T.
Wei
,
S.
Yang
,
Z.
Dou
,
Y.
Wang
,
H.
Ci
,
H.
Chang
,
Y.
Qi
,
J.
Yan
,
J.
Wang
,
Y.
Zhang
,
P.
Gao
,
J.
Li
, and
Z.
Liu
,
Adv. Mater.
31
,
1807345
(
2019
).
26.
Y.
Feng
,
X.
Yang
,
Z.
Zhang
,
D.
Kang
,
J.
Zhang
,
K.
Liu
,
X.
Li
,
J.
Shen
,
F.
Liu
,
T.
Wang
,
P.
Ji
,
F.
Xu
,
N.
Tang
,
T.
Yu
,
X.
Wang
,
D.
Yu
,
W.
Ge
, and
B.
Shen
,
Adv. Funct. Mater.
29
,
1905056
(
2019
).
27.
Y.
Jia
,
J.
Ning
,
J.
Zhang
,
C.
Yan
,
B.
Wang
,
Y.
Zhang
,
J.
Zhu
,
X.
Shen
,
J.
Dong
,
D.
Wang
, and
Y.
Hao
,
Adv. Opt. Mater.
8
,
1901632
(
2020
).
28.
V.
Kumaresan
,
L.
Largeau
,
A.
Madouri
,
F.
Glas
,
H.
Zhang
,
F.
Oehler
,
A.
Cavanna
,
A.
Babichev
,
L.
Travers
,
N.
Gogneau
,
M.
Tchernycheva
, and
J.-C.
Harmand
,
Nano Lett.
16
,
4895
(
2016
).
29.
S.
Fernández-Garrido
,
M.
Ramsteiner
,
G.
Gao
,
L. A.
Galves
,
B.
Sharma
,
P.
Corfdir
,
G.
Calabrese
,
Z.
de Souza Schiaber
,
C.
Pfüller
,
A.
Trampert
,
J. M. J.
Lopes
,
O.
Brandt
, and
L.
Geelhaar
,
Nano Lett.
17
,
5213
(
2017
).
30.
A. C.
Ferrari
,
J.
Meyer
,
V.
Scardaci
,
C.
Casiraghi
,
M.
Lazzeri
,
F.
Mauri
,
S.
Piscanec
,
D.
Jiang
,
K.
Novoselov
,
S.
Roth
, and
A. K.
Geim
,
Phys. Rev. Lett.
97
,
187401
(
2006
).
31.
R. G.
Banal
,
M.
Funato
, and
Y.
Kawakami
,
J. Cryst. Growth
311
,
2834
(
2009
).
32.
A.
Costales
,
A. K.
Kandalam
,
A.
Martín Pendás
,
M.
Blanco
,
J.
Recio
, and
R.
Pandey
,
J. Phys. Chem. B
104
,
4368
(
2000
).
33.
A. B.
Georgescu
and
S.
Ismail-Beigi
,
Phys. Rev. Appl.
11
,
064065
(
2019
).
34.
M.
Moram
and
M.
Vickers
,
Rep. Prog. Phys.
72
,
036502
(
2009
).
35.
R.
Gutt
,
L.
Kirste
,
T.
Passow
,
M.
Kunzer
,
K.
Köhler
, and
J.
Wagner
,
Phys. Status Solidi B
247
,
1710
(
2010
).
36.
A.
Pandey
,
W.
Shin
,
X.
Liu
, and
Z.
Mi
,
Opt. Express
27
,
A738
(
2019
).

Supplementary Material