Maximizing cubic phase gallium nitride surface coverage on nano-patternedsilicon (100)

R. Liu and C. Bayrama)

Department of Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, Urbana,Illinois 61801, USA and Micro and Nanotechnology Laboratory, University of Illinois at Urbana-Champaign,Urbana, Illinois 61801, USA

(Received 10 April 2016; accepted 17 July 2016; published online 28 July 2016)

Here we investigate the hexagonal-to-cubic phase transition in metalorganic-chemical-vapor-

deposition-grown gallium nitride enabled via silicon (100) nano-patterning. Electron backscatter

diffraction and depth-resolved cathodoluminescence experiments show complete cubic phase GaN

surface coverage when GaN deposition thickness (hc), etch depth (td), and opening width (p) obey

hc � 1:06p� 0:75td; in line with a geometrical model based on crystallography. Cubic GaN unifor-

mity is studied via electron backscatter diffraction and cathodoluminescence measurements. Atomic

force microscopy reveals a smooth cubic GaN surface. Phase-transition cubic GaN shows promising

optical and structural quality for integrated photonic devices. Published by AIP Publishing.[http://dx.doi.org/10.1063/1.4960005]

Thanks to their direct bandgap across the entire visible

spectrum and ultra violet, Gallium Nitride (GaN) semicon-

ductors and its compounds (AlGaInN) have transformed the

visible light emitting diode (LED) industry and are now

being explored for RF/power transistors. Almost without

exception, LEDs and other kinds of GaN devices including

RF/power transistors are grown on three- or six-fold symme-

try surfaces (e.g., Al2O3, SiC, and Si(111) substrates) due to

phase stability. The resulting GaN is therefore the six-fold

symmetric hexagonal phase (wurtzite) GaN (h-GaN). The

non-centrosymmetric nature (i.e., Ga and N atoms are not

interchangeable in the lattice) of the hexagonal crystal

arrangement leads to residual spontaneous and piezoelectric

polarization fields.1 Both these polarization fields are along

the h0001i growth direction, which is also the carrier injec-

tion direction in vertical transport devices, such as LEDs,

lasers, and detectors, and thus are detrimental to recombina-

tion dynamics and device efficiency.1–7 On the other hand,

cubic phase (zincblende) GaN (c-GaN) does not possess

these polarization fields. Other advantages of c-GaN in pho-

tonic devices include cleavage planes and a higher optical

gain.8 As such, there exists a need for a reliable approach

for fabricating c-GaN for applications ranging from

polarization-free photonics,9 normally off transistors,10

room-temperature ferromagnetism,11 high-temperature spin-

tronics,12 and single photon emitters.13 Yet, c-GaN is one of

the least studied materials due to its phase instability and ten-

dency to revert to the more stable h-GaN.

Conventional c-GaN is grown via planar epitaxy onto

substrates such as GaAs,14 3 C-SiC,15 Si (100),16 and

MgO,17 or via impurity incorporation.18,19 However, the

c-GaN formed through these approaches suffers from high

defectivity and phase mixing. Recently, c-GaN formation via

phase transition on patterned substrates with V-shaped

[Si{111}-Si{�1�1 1}]20–22 or U-shaped [Si{111}-Si{100}-

Si{�1�1 1}]23,24 groove structures was introduced by metalor-

ganic chemical vapor deposition (MOCVD). Based on the

crystallographic geometry that h-crystal h0001i direction and

c-crystal h111i direction are equivalent, it has been shown if

two h-phase h0001i growth fronts merge at an angle of

�109.5� (i.e., the angle between the two Ga-N bonds in the

tetrahedral bonding), c-phase would form after the seam.

Anisotropic nano-patterning of Si(100) substrates are typi-

cally used to create U-shaped (or V-shaped if etched all the

way) grooves with a crystallographic angle of 54.74�

between the (100) and (111)Si surfaces. Thus, GaN selective

MOCVD-growth on Si(111) facets of a U-groove leads to

two h-GaN growth fronts meeting at an angle of 54.74� � 2

� 109.5�, which is exactly the angle required to facilitate the

transition of the h-GaN into c-GaN after coalescence. Some

practical applications of this technology involve localized

GaN devices. As the transistor technology is within tens of

nanometers, it is indeed feasible to use such localized epitax-

ial high quality materials for GaN sub-micron transistors.

Moreover, such GaN/Si offers excellent waveguides and

enables GaN-on-Si photonics as a natural cleavage plane

occurring for cubic phase GaN on Si (100). In this respect,

the implications of the localized GaN-on-Si epitaxy results

are diverse and very motivating for GaN-Si community.

Here, we investigate the method of forming c-GaN in

U-shaped grooves, propose a geometrical model based on

crystallography to describe the hexagonal-to-cubic phase

transition in these grooves, and study the effects of nano-

patterning parameters on c-GaN formation on CMOS-

compatible Si(100) (i.e., on-axis 750-lm-thick Si(100)

substrate). The details of substrate preparation and epitaxial

growth can be found in our earlier works.23,24 GaN materials

deposited by MOCVD on various U-grooves are studied via

temperature-dependent and depth-resolved cathodolumines-

cence (CL) and electron backscatter diffraction (EBSD).

These studies are conducted on a JEOL 7000F analytical

SEM fitted with Gatan MonoCL3 (slit width and grating

a)Present address: Innovative Compound semiconductor (ICOR) Laboratory,

Urbana, Illinois 61801, USA. Electronic mail: cbayram@illinois.edu.

Telephone: þ1 (217) 300-0978. Fax: þ1 (217) 244-6375. URL: http://

www.icorlab.ece.illinois.edu.

0003-6951/2016/109(4)/042103/4/$30.00 Published by AIP Publishing.109, 042103-1

APPLIED PHYSICS LETTERS 109, 042103 (2016)

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 139.179.98.143 On: Thu, 28 Jul 2016

14:48:38

combination yielding a wavelength resolution of 5 nm) and a

HLK Technology EBSD system to detect crystal phase and

orientation, respectively.

Based on our proposed crystallographic geometry

modelling (see supplementary material for modeling/calcula-

tion details25), a schematic of the growth pattern of GaN in a

U-groove seen from the side is shown in Figure 1(a). For

complete c-GaN surface coverage in these U-grooves, the

critical GaN deposition thickness (hc) is shown25 obeying

hc ¼ 1:06p� 0:75tdð Þ= 1� tan a0:71

� �, where p is the opening

width of the U-groove at the dielectric/substrate interface, tdis the silicon etch depth, and a is the dielectric sidewall angle

(Fig. 1(a)).25 Additionally, we define a quantity, fill factor

(ff) as ff¼ h/hc, to quantify the ratio between the actual and

the ideal GaN deposition thickness for complete c-GaN sur-

face coverage. Here, h is defined as the GaN deposition

thickness above Si(100) substrate level (Figs. 1(b)–1(d)).

Figures 1(b)–1(d) show cross-sectional scanning elec-

tron microscope (SEM) images of GaN in three different

U-grooves. Figure 1(b) is close to an ideal case where GaN

has a perfect fill (h � hc) for the patterning parameters of

td¼ 80 nm, p¼ 300 nm, and a � 2�. The top c-GaN(100) sur-

face is seen very flat. Figure 1(c) shows an under-filled case

(h < hc, ff � 45%) with td¼ 70 nm, p¼ 640 nm, and a � 4�.On the surface, a small amount of the flat c-GaN(100) can be

seen in the center flanked by two h-GaN ð1�101Þ planes,

which make a 7� angle to the (100) planes of both the Si sub-

strate and c-GaN. For c-GaN to cover the entire surface,

more deposition is needed. Figure 1(d) shows an over-filled

case (h > hc, ff � 170%) in which too much GaN has been

deposited for td¼ 70 nm, p¼ 190 nm, and i a� 1�. There is

no identifiable crystal plane on the surface, implying over-

deposition of GaN (h > hc) causes the over-filled GaN to

phase mix and lose its crystallinity.

Various U-grooves (with p ranging from 200 to 1110 nm

and td from 65 to 85 nm) are fabricated, and GaN is depos-

ited in these various U-grooves to create a variety of ff’sranging from 30% to 210%. Cathodoluminescence is used to

study c- and h-GaN emission properties; it provides insights

into the composition of the surface by properly adjusting the

electron acceleration voltage and deliberately creating elec-

tron-hole-pairs at specific depths.26

Figure 2 shows the (a) 280 K and (b) 5.7 K CL spectra

of twelve different types of GaN on U-groove structures

using electron beam current of 2 nA and acceleration voltage

of 2 kV. Our control sample is the p ¼ 1 nm, where there is

no opposing wall or h-GaN growth front, and therefore

no phase transition, leading to pure h-GaN. The ff, p, and

td of U-grooves are listed beside their respective curves in

Fig. 2(a).

For the two U-grooves with ff of 90% and 95%, which

correspond to the scenarios closest to the ideal case (h � hc)

for complete cubic phase surface coverage (see SEM image

in Fig. 1(b)), the CL emission spectra show strong c-GaN

presence with band to band photon energies of 3.22 eV at

280 K (Fig. 2(a)) and 3.29 eV at 5.7 K (Fig. 2(b)), marked

with the dotted vertical lines. These two U-grooves exhibit

additional luminescence mechanisms such as the two donor-

acceptor pairs (DAP1 and DAP2) as reported earlier.26

For ff < 90% (the bottom five U-grooves in Fig. 2), CL

emissions are dominated by the h-GaN, as seen in the ff¼ 0(p ¼ 1 nm) U-groove at 3.42 eV and 3.49 eV (dashed lines)

at 280 K and 5.7 K, respectively. These U-grooves corre-

spond to the cross-sectional SEM image in Fig. 1(c), with a

varying portion of c-GaN in the center, flanked by two

ð1�101Þ planes of h-GaN. As the ff approaches 100% from

30%, c-GaN emission becomes observable and intensifies,

resulting in a superposition of both h- and c-GaN, indicating

that the amount of c-GaN in the center relative to the entire

groove opening is increasing.

FIG. 1. (a) Schematic defining critical deposition thickness (hc), etch depth (td), opening width (p), and sidewall angle (a). Cross-sectional SEM views of GaN

deposited in a U-groove when (b) (h � hc) complete c-GaN surface coverage is achieved (ff � 95%), (c) (h < hc) the U-groove is under-filled with GaN (ff� 40%), and (d) (h > hc) the U-groove is over-filled with GaN (ff � 120%). This leads to phase-mixing. The white scale bars in the bottom right of each figure

correspond to 100 nm.

FIG. 2. CL spectra of GaN on various U-grooves at (a) 280 K and (b)

5.7 K. When ff� 100%, mostly h-GaN emission (dashed lines) is observed

suggesting incomplete c-GaN surface coverage at 280 K and 5.7 K. When

ff � 100%, strong cubic band to band transition is observed (dotted lines)

at 280 K and cubic DAPs are observed at 5.7 K. When ff � 100%, both

c- and h-GaN emissions are recorded at 280 K, which become dominantly

c-GaN at 5.7 K.

042103-2 R. Liu and C. Bayram Appl. Phys. Lett. 109, 042103 (2016)

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 139.179.98.143 On: Thu, 28 Jul 2016

14:48:38

For ff > 100% (the top five U-grooves in Fig. 2), the CL

spectra show a mixture of h- and c-GaN emissions. These

U-grooves correspond to the cross-sectional SEM image in

Fig. 1(d), where the U-groove is over-filled, and no clear

crystal plane is distinguishable. At 280 K, h- and c-GaN

emissions of various amplitudes can be seen without any par-

ticular pattern. The CL spectra suggest that once hc is passed,

the subsequently deposited GaN is of mixed-phase.

Interestingly, despite the various intensities of h-GaN emis-

sion at 280 K across these U-grooves, all of them show pre-

dominantly the band to band emission of c-GaN at 5.7 K.

DAP1 can also be seen in four of the five U-grooves, and its

intensity seems to correlate to the intensity of c-GaN emis-

sion at 280 K. We believe that such dominance of c-GaN

emission at 5.7 K, even in mixed phase GaN materials, is

due to the carrier confinement in c-GaN as its bandgap

(3.31 eV) is smaller than that of h-GaN (3.51 eV) due to the

enhanced carrier mobility at low temperature.26 The confine-

ment is most apparent in the ff¼ 70% period. At 280 K, only

a small amount of c-GaN emission can be seen along with a

very strong h-GaN emission. At 5.7 K, the emission spec-

trum shows predominantly the typical low temperature

c-GaN emission mixed with some low temperature h-GaN

emission, suggesting an increase in the ratio of radiative

recombination in c-GaN over h-GaN.

To study the GaN phase along the vertical growth direc-

tion h100i, the CL electron acceleration voltage and the

recorded emission spectra are varied. The relationship

between the maximum electron penetration depth and the pri-

mary electron acceleration voltage is described by Kanaya-

Okayama formula D ¼ 0:0267�A�E5=3

0

q�Z8=9 , where D is the maximum

penetration depth, A the atomic weight, E0 the beam energy, qthe density, and Z the weighted average atomic number.27

Figure 3 inset shows the maximum electron penetration depth

as a function of electron acceleration voltage for GaN materi-

als, assuming the similar behavior in both h- and c-GaN.

Figure 3 shows the CL spectra of the optimized c-GaN (ff� 95%) U-groove with various electron acceleration voltages

(from 1 to 8 kV) under a constant current of 2 nA at room tem-

perature. At voltages up to 2 kV (D ¼ 42 nm), no h-GaN emis-

sion is observed. This indicates that the GaN is composed of

almost entirely of c-GaN, in good agreement with our model-

ling. Beginning at 3 kV (d ¼ 84 nmÞ, h-GaN emission in the

CL spectrum is observed. As voltage increases, more electron-

hole-pairs (EHPs) are created in the deeper parts of the

U-grooves; as a result, more EHPs recombine in the h-GaN

region, emitting at the bandgap of h-GaN, as shown in Fig. 3.

Further electron backscatter diffraction (EBSD) meas-

urements are carried out in order to study the distribution

and location of the different GaN phases on the U-groove

surfaces. The EBSD setup can sample the very top layer

(�10 nm) of the sample when a beam current of 2 nA and

acceleration voltage of 15 kV is used. Figures 4 and S4 (Ref.

25) show the plane-view EBSD images overlaid on top of

the SEM images of the (a) ff¼ 40%, (b) ff¼ 65%, and (c)

ff¼ 95% U-grooves. The red, blue, and grey pixels indicate

hexagonal, cubic, and indistinguishable crystal phase,

respectively, and have the spatial dimension of 20 nm by

20 nm. EBSD identifies and confirms the GaN formed in the

middle as the cubic phase, with its (001) plane on the top sur-

face and its (110), ð1�10Þ planes parallel and perpendicular to

the U-grooves, respectively. As the ff of the U-groove

approaches 100%, the surface coverage of c-GaN increases,

until the entire surface is composed of only c-GaN, as shown

in Fig. 4(c). Electron backscatter diffraction shows that the

c-GaN formed through the phase transition in a U-groove is

of single phase and proves U-groove nano-patterning is a

feasible and reliable method of forming single phase c-GaN

FIG. 3. CL spectra of GaN on a U-groove with ff ¼ 95% are shown as a

function of electron acceleration voltage. Inset shows calculated maximum

electron penetration depth for GaN material. At 2 kV (range¼ 44 nm) and

below, electrons are not able to penetrate the c-GaN to reach h-GaN. At

3 kV (range¼ 84 nm) and above, EHPs are being created in the h-GaN that

corresponds to increase in h-GaN emission in the CL spectra.

FIG. 4. Top-view SEM images with EBSD overlay of GaN on an U-groove with (a) ff¼ 40%, (b) ff¼ 65%, and (c) ff ¼ 95%. Each pixel has resolution of

20� 20 nm, and red, blue, and grey colors correspond to h-GaN, c-GaN, and indistinguishable phases, respectively. The crystal orientation of c-GaN is identi-

fied and labelled in (c). The white scale bars in the bottom right of each figure correspond to 200 nm. (d) AFM surface image of the c-GaN region with an

RMS roughness of 2.65 A.

042103-3 R. Liu and C. Bayram Appl. Phys. Lett. 109, 042103 (2016)

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 139.179.98.143 On: Thu, 28 Jul 2016

14:48:38

surface. Atomic force microscopy surface image of the c-GaN is shown in Fig. 4(d), having a root-mean-square rough-

ness of 2.65 A.

In conclusion, we show that the hexagonal-to-cubic

phase transition is a reliable method to form pure cubic

phase GaN. The U-groove parameters, etch depth, opening

width, and deposition thickness are shown to be of impor-

tance for complete cubic phase surface coverage. A geomet-

rical model based on the crystallography is proposed and

validated by EBSD and CL studies—proving c-GaN show-

ing excellent phase uniformity and smooth surface. Overall,

phase-transition cubic GaN shows promising optical and

structural quality for integrated photonic devices.

This work was carried out in the Micro and

Nanotechnology Laboratory and Frederick Seitz Materials

Research Laboratory Central Research Facilities, University

of Illinois at Urbana-Champaign, IL, USA. The authors

acknowledge support from Dr. James Mabon.

1M. Shur and R. Gaska, IEEE Trans. Electron Devices 57(1), 12 (2010).2T. Oto, R. G. Banal, K. Kataoka, M. Funato, and Y. Kawakami, Nat.

Photonics 4, 767–771 (2010).3G. Verzellesi, D. Saguatti, M. Meneghini, F. Bertazzi, M. Goano, G.

Meneghesso, and E. Zanoni, J. Appl. Phys. 114, 071101 (2013).4J. Cho, E. F. Schubert, and J. K. Kim, Laser Photonics Rev. 7(3), 408–421

(2013).5D. S. Meyaard, G.-B. Lin, J. Cho, E. F. Schubert, H. Shim, S.-H. Han,

M.-H. Kim, C. Sone, and Y. S. Kim, Appl. Phys. Lett. 102, 251114

(2013).6S. Karpov, Opt. Quantum Electron. 47(6), 1293–1303 (2015).7C. Weisbuch, M. Piccardo, L. Martinelli, J. Iveland, J. Peretti, and J. S.

Speck, Phys. Status Solidi A 212(5), 899–913 (2015).8C.-H. Kim and B.-H. Han, Solid State Commun. 106(3), 127–132 (1998).9D. J. As, Microelectron. J. 40, 204–209 (2009).

10R. Granzner, E. Tschumak, M. Kittler, K. Tonisch, W. Jatal, J. Pezoldt, D.

As, and F. Schwierz, J. Appl. Phys. 110, 114501 (2011).11V. A. Chitta, J. A. H. Coaquira, J. R. L. Fernandez, C. A. Duarte, J. R.

Leite, D. Schikora, D. J. As, K. Lischka, and E. Abramof, Appl. Phys.

Lett. 85, 3777–3779 (2004).12J. H. Bub, A. Schaefer, T. Schupp, D. J. As, D. Hagele, and J. Rudolph,

Appl. Phys. Lett. 105, 182404 (2014).13S. Kako, M. Holmes, S. Sergent, M. Burger, D. J. As, and Y. Arakawa,

Appl. Phys. Lett. 104, 011101 (2014).14D. J. As, A. Richter, J. Bush, M. Lubbers, J. Mimkes, and K. Lischka,

Appl. Phys. Lett. 76, 13–15 (2000).15S. F. Chichibu, T. Onuma, T. Aoyama, K. Nakajima, P. Ahmet, T.

Chikyow, T. Sota, S. P. DenBaars, S. Nakamura, T. Kitamura, Y. Ishida,

and H. Okumura, J. Vac. Sci. Technol., B 21(4), 1856–1862 (2003).16C. H. Wei, Z. Y. Xie, L. Y. Li, Q. M. Yu, and J. H. Edgar, J. Electron.

Mater. 29(3), 317–321 (2000).17V. D. C. Garcia, I. E. O. Hinostroza, A. E. Echavarria, E. L. Luna, A. G.

Rodriguez, and M. A. Vidal, J. Cryst. Growth 418, 120–125 (2015).18S. Dhara, A. Datta, C. T. Wu, Z. H. Lan, K. H. Chen, Y. L. Wang, C. W.

Hsu, C. H. Shen, L. C. Chen, and C. C. Chen, Appl. Phys. Lett. 84,

5473–5475 (2004).19Y. Cui, V. K. Lazorov, M. M. Goetz, H. Liu, D. O. Robertson, M.

Gajdardziska-Josifovska, and L. Li, Appl. Phys. Lett. 82, 4666–4668

(2003).20S. C. Lee, X. Y. Sun, S. D. Hersee, and S. R. J. Brueck, J. Cryst. Growth

279, 289–292 (2005).21B. Reuters, J. Strate, A. Wille, M. Marx, G. L€ukens, L. Heuken, M.

Heuken, H. Kalisch, and A. Vescan, J. Phys. D: Appl. Phys. 48, 485103

(2015).22M. T. Durniak, A. S. Bross, D. Elsaesser, A. Chaudhuri, M. L. Smith, A.

A. Allerman, S. C. Lee, S. R. J. Brueck, and C. Wetzel, Adv. Electron.

Mater. 2, 1500327 (2016).23C. Bayram, J. Ott, K.-T. Shiu, C.-W. Cheng, Y. Zhu, J. Kim, M. Razeghi,

and D. K. Sadana, Adv. Func. Mater. 24(28), 4492 (2014).24C. Bayram, C.-W. Cheng, D. K. Sadana, and K.-T. Shiu, U.S. patent

9,048,173 (2 June 2015).25See supplementary material at http://dx.doi.org/10.1063/1.4960005 for

supplementary geometrical modelling.26R. Liu and C. Bayram, J. Appl. Phys. 120, 025106 (2016).27K. Kanaya. and S. Okayama, J. Phys. D.: Appl. Phys. 5, 43 (1972).

042103-4 R. Liu and C. Bayram Appl. Phys. Lett. 109, 042103 (2016)

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 139.179.98.143 On: Thu, 28 Jul 2016

14:48:38

Supplementary Online Material 1

Supplementary Online Material for:

“Maximizing Cubic Phase Gallium Nitride Surface Coverage

on Nano-patterned Silicon (100)”

R. Liu and C. Bayram1

Department of Electrical and Computer Engineering, University of Illinois at Urbana-

Champaign, Illinois, 61801, USA

Micro and Nanotechnology Laboratory, University of Illinois at Urbana-Champaign,

Illinois, 61801, USA

This pdf file includes:

- Supplementary Geometrical Model Based on Crystallography

- Figure S1

- Figure S2

- Uniformity Study of the Cathodoluminescence and Electron Backscatter Diffraction

Measurements

- Figure S3

- Figure S4

1 Innovative COmpound semiconductoR (ICOR) Laboratory; Email: cbayram@illinois.edu;

Webpage: icorlab.ece.illinois.edu; Phone: +1 (217) 300-0978; Fax: +1 (217) 244-6375

Supplementary Online Material 2

Supplementary Geometrical Model Based on Crystallography

Figure S1 shows the model sketch of GaN in a U-groove. The V- shaped red dashed lines refer to

the phase boundaries where hexagonal-to-cubic phase transitions occurs1, and the top (red) dashed

line indicates the c-GaN top surface. Under the selective growth conditions1, 2, GaN nucleates on

the Si(111) facets only and these h-GaN growth fronts meet at a 109.48° angle in the middle of the

groove. After the middle of the growth fronts meet (point A in both figures), GaN grown on top

will phase transition to cubic phase. Under these experimental / crystallographic observations, the

geometrical modelling is carried out as follows:

In the following derivation 𝑥𝑖 and 𝑦𝑖 are various dimensions as shown in the Fig. S1 and Fig. S2,

ℎ𝑐 is the critical GaN deposition thickness (defined as the GaN deposition height above Si(100)

that maximizes cubic phase GaN coverage on the U-groove surface), 𝑡𝑑 is the etch depth, 𝑝 is the

opening width, and 𝛼 is the oxide sidewall angle.

Figure S2 shows the zoomed in region indicated by the dotted square in Fig. S1. Points A, B, C in

both figures (Fig. S1 and S2) correspond to the same exact locations. The crystallographic angles

(54.74° between the (100) and (111) Si surfaces) are shown accordingly.

From Fig. S2, we see that:

𝒕𝒂𝒏𝟑𝟓. 𝟑° =𝒚𝟐𝒙𝟑

(S1)

𝒕𝒂𝒏𝟓𝟒. 𝟕° =𝒕𝒅𝟐𝒙𝟏

(S2)

𝒙𝟏 + 𝒙𝟑 =𝒑

𝟐 (S3)

𝒚𝟑 = 𝒚𝟐 +𝒕𝒅𝟐

(S4)

Rearranging (S1), (S2), and (S3), we get:

𝒚𝟐 = 𝒙𝟑𝒕𝒂𝒏𝟑𝟓. 𝟑° (S5)

𝒙𝟏 =𝒕𝒅

𝟐𝒕𝒂𝒏𝟓𝟒. 𝟕° (S6)

Supplementary Online Material 3

𝒙𝟑 =𝒑

𝟐− 𝒙𝟏 (S7)

Substituting (S6) into (S7):

𝒙𝟑 =𝒑

𝟐− 𝒙𝟏 =

𝒑

𝟐−

𝒕𝒅𝟐𝒕𝒂𝒏𝟓𝟒. 𝟕°

(S8)

Then using (S8) in (S5):

𝒚𝟐 = 𝒙𝟑𝒕𝒂𝒏𝟑𝟓. 𝟑° =(𝒑

𝟐−

𝒕𝒅

𝟐𝒕𝒂𝒏𝟓𝟒.𝟕°) 𝒕𝒂𝒏𝟑𝟓. 𝟑° (S9)

From Figure S1,

𝒙𝟒 = 𝒑 + 𝟐 𝒉𝑪 𝒕𝒂𝒏 𝜶 (S10)

𝒚𝟏 + 𝒚𝟑 = 𝒉𝑪 + 𝒕𝒅 (S11)

𝒕𝒂𝒏𝟑𝟓. 𝟑° =𝒙𝟒𝟐𝒚𝟏

(S12)

Rearranging (S12) using (S10), and (S11) using (S4):

𝒚𝟏 =𝒙𝟒

𝟐𝒕𝒂𝒏𝟑𝟓. 𝟑°=𝒑 + 𝟐 𝒉𝑪 𝒕𝒂𝒏 𝜶

𝟐𝒕𝒂𝒏𝟑𝟓. 𝟑° (S13)

𝒉 = 𝒚𝟏 + 𝒚𝟑 − 𝒕𝒅 = 𝒚𝟏 + 𝒚𝟐 −𝒕𝒅

𝟐 (S14)

Substituting (S13) and (S9) in (S14), we get:

𝒉𝑪 =𝒑 + 𝟐 𝒉𝑪 𝒕𝒂𝒏 𝜶

𝟐𝒕𝒂𝒏𝟑𝟓. 𝟑°+ (

𝒑

𝟐−

𝒕𝒅𝟐𝒕𝒂𝒏𝟓𝟒. 𝟕°

) 𝒕𝒂𝒏𝟑𝟓. 𝟑° −𝒕𝒅𝟐

(S15)

Solving for ℎ𝑐

𝒉𝑪 = (𝟏 −𝒕𝒂𝒏𝜶

𝒕𝒂𝒏 𝟑𝟓. 𝟑°)

−𝟏

[(𝒑

𝟐−

𝒕𝒅

𝟐𝒕𝒂𝒏 𝟓𝟒. 𝟕°) 𝒕𝒂𝒏 𝟑𝟓. 𝟑° +

𝒑

𝟐𝒕𝒂𝒏𝟑𝟓. 𝟑°−𝒕𝒅𝟐] (S16)

Supplementary Online Material 4

Simplify and plugging in values for the tangents, we have the relationship between critical

thickness and the patterning parameters:

𝒉𝑪 =[𝟏. 𝟎𝟔𝒑 − 𝟎. 𝟕𝟓𝒕𝒅]

(𝟏 −𝒕𝒂𝒏 𝜶𝟎. 𝟕𝟏 )

(S17)

If 𝛼 is negligible ( ≈ 0°), the relationship simplifies to:

𝒉𝑪 = [𝟏. 𝟎𝟔𝒑 − 𝟎. 𝟕𝟓𝒕𝒅] (S18)

Supplementary Online Material 5

Figure S1. GaN deposition in U-groove model (under selective growth conditions1, 2) is shown

with dimensions labelled. The red dashed triangle indicates the central region where GaN has

phase transitioned from hexagonal to cubic.

Supplementary Online Material 6

Figure S2. Zoomed-in schematic of the boxed region indicated with continuous black box in Fig.

S1. Points labeled as A, B, C in Fig. S2 corresponds to the same points A, B, C in Fig. S1.

Supplementary Online Material 7

Uniformity Study of the Cathodoluminescence and Electron

Backscatter Diffraction Measurements

To study the uniformity, large area (<1000 µm2) cathodoluminescence (CL) mapping on GaN with

various fill factors are carried out. Figure S3 shows the mapping of four representative grooves

with ff’s of (from top to bottom): 30%, 70%, 95%, and 120% at 5.7 K. In Fig. S3a, plain-view

SEM images of the grooves are shown, with high magnification insets showing individual groove.

Strips of c-GaN with various widths can be seen sandwiched between h-GaN strips. As the ff

increases from 30%, the c-GaN strips become more dominant, and completely covers the surface

when ff reaches 95%, similar to Fig. 1b in the manuscript. Once the ff becomes greater than 120%,

the h-GaN strips reappear at random places, sandwiching the c-GaN strips. The corresponding CL

mapping of the grooves in Fig. S3a at 355 nm, 390nm, and 425 nm are shown in Fig. S3b, c, d,

respectively. It is observed that as the ff increases, h-GaN exciton emission (355 nm) disappears,

and the DAP2 of c-GaN emission (425 nm) increases. The 390 nm emission corresponds to both

the DAP1 of c-GaN and the LO-phonon of the DAP1 of h-GaN, and therefore cannot be used to

accurately delineate the distribution of the two phases of GaN. The exciton emission of c-GaN

(378 nm) cannot be imaged easily, unfortunately, due to its relatively weak intensity compare to

other emissions. The zero phonon line of DAP1 of h-GaN also coincides with it, further undermines

the wavelength’s ability to distinguish the two phases of GaN.

Supplementary Online Material 8

Figure S3. CL mapping of the (top to bottom) ff = 30%, 70%, 95%, 120%. (a) plain-view SEM

images with a high magnification SEM image. Monochromatic CL mapping at (b) 355 nm, (c) 390

nm, (d) 425 nm of the corresponding periods in (a) are shown.

Supplementary Online Material 9

Figure S4. Multi-period EBSD overlay on SEM images of (a) ff = 40%, (b) ff = 65% (c) ff = 95%

periods using a beam current of 2 nA and acceleration voltage of 15 kV. Data is taken at RT.

Supplementary Online Material 10

REFERENCES

1 C. Bayram, J. Ott, K.-T. Shiu, C.-W. Cheng, Y. Zhu, J. Kim, M. Razeghi, and D.K. Sadana,

Adv. Func. Mater. 24 (28) 4491(2014).

2 C. Bayram, C.-W. Cheng, D.K. Sadana, and K.-T. Shiu, U.S. Patent 9,059,075 (issued June 16,

2015).