Optimization of low temperature GaN buffer

layers for halide vapor phase epitaxy growth of

bulk GaN

Carl Hemmingsson and Galia Pozina

Linköping University Post Print

N.B.: When citing this work, cite the original article.

Original Publication:

Carl Hemmingsson and Galia Pozina, Optimization of low temperature GaN buffer layers for

halide vapor phase epitaxy growth of bulk GaN, 2013, Journal of Crystal Growth, (366), 61-

66.

http://dx.doi.org/10.1016/j.jcrysgro.2012.12.016

Copyright: Elsevier

http://www.elsevier.com/

Postprint available at: Linköping University Electronic Press

http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-90144

1

Optimization of low temperature GaN buffer layers for halide vapour phase

epitaxy growth of bulk GaN

Carl Hemmingsson and Galia Pozina

Department of Physics, Chemistry and Biology (IFM), Linköping University, S-581 83

Linköping, Sweden

Corresponding author: Carl Hemmingsson

Phone: +46 13 282627

Mobile: +46 73 0651264

Fax: +46 13 142337

Email: cah@ifm.liu.se

Abstract

We have studied growth and self-separation of bulk GaN on c-oriented Al2O3 using low

temperature (LT) GaN buffer layers. By studying the X-ray diffraction (XRD) signature for the

asymmetric and symmetric reflections versus the LT-GaN thickness and V/III precursor ratio,

we observe that the peak width of the reflections is minimized using a LT buffer thickness of

~100 -300 nm. It was observed that the V/III precursor ratio has a strong influence on the

morphology. In order to obtain a smooth morphology, the V/III precursor ratio has to be more

than 17 during the growth of the buffer layer. By using an optimized LT buffer layer for

growth of a 20 µm thick GaN layer, we obtain a XRD peak with a full width at half maximum

of ~400 and ~250 arcsec for (002) and (105) reflection planes, respectively, and with a dark pit

density of ~2.2 108 cm

-2. For layers thicker than 1 mm, the GaN was spontaneously separated

and by utilizing this process, thick free freestanding 2’’ GaN substrates were manufactured.

Keywords: A1: Crystal morphology, Interfaces, Roughening; A2. Growth from vapour; A3

Hydride vapour phase epitaxy; B1. Nitrides; B2. Semiconducting III-V materials

PACS classification codes: 81.05.Ea; 81.10.Bk; 81.15.Cd; 68.35.Ct; 78.55.-m; 78.60.Hk

2

1. INTRODUCTION

The development of devices based on III-V nitrides has been hampered due to the

difficulties in fabrication of large native single crystal substrates and today, still most of the III-

V nitride based devices are grown on substrates such as Al2O3 or 6H-SiC. Heteroepitaxial

growth results in stressed layers with high density of defects and problems with cracking due to

the large lattice mismatch and difference in thermal expansion coefficient between the

substrate and the layer. Due to the large defect density the operating lifetime and the internal

quantum efficiency drops with higher current density, thus for more demanding devices such

as visible or UV laser diodes (LDs), high brightness power light emitting diodes (LED) for

white light illumination, and possibly for high performance high power transistors, epitaxial

lateral overgrowth techniques (ELOG) or native GaN substrates with a dislocation density

below 106 cm

-2 are necessary. There are few companies today that can offer GaN substrates

commercially grown by halide vapour phase epitaxy (HVPE), ammonothermal growth (an

analogue of hydrothermal growth used for production of quartz) or high nitrogen pressure

solution (HNPS) growth (for example, see Ref. [1-5]), however, the production cost is still too

high for a broader commercialization. Thus, there is still a strong need to develop growth

techniques for manufacturing of low cost GaN substrates.

In case of GaN bulk growth using HVPE, the growth is started in most cases on c-

oriented Al2O3 substrates. Due to the heteroepitaxial substrate mismatch, a high density of

threading dislocations (TD) is formed. However, with increasing thickness of the GaN layer,

the TD density decreases significantly due to dislocation annihilation processes. In order to

minimize the stress, facilitate the removal of the substrate and reduce the dislocation density a

number of different techniques based on buffer layers have been explored such as growth on

AlN [6], ZnO [7], AlInN [8], InGaN/GaN [9], TiN [10], and GaN nanorods [11] as well as

epitaxial lateral overgrowth (ELOG) [12-14]. These techniques require several growth steps

3

using different types of growth equipment, and in case of ELOG, deposition of a dielectric

such as silicon nitride or silicon oxide and photolithography is necessary. One alternative

method, which does not require photolithography, is the two step deposition process involving

growth of a low temperature (LT) GaN buffer layer. This technique has been used in both

metal organic chemical vapor deposition (MOCVD) [15, 16] and HVPE [17]. Using this

technique, it has been shown that for thicker GaN layers (> 600 µm) the bulk GaN layer is

spontaneously separated from the substrate which means that many time-consuming or

complicated process steps such as mechanical lapping of sapphire [18], chemical etching [19],

laser lift-off [20], ELOG [21, 22] and void assisted separation [23] can be avoided. Thus, there

are several advantages using LT-GaN buffers since the number of process steps, and hence, the

manufacturing cost can be reduced. However, the role of LT-GaN layer on sapphire substrates

has not yet been fully understood [24].

In this paper, we have investigated growth of bulk GaN on Al2O3 substrates with LT-

GaN buffer layers utilizing HVPE method. A number of thick (up to 2 mm) GaN layers were

fabricated on LT buffer layers of various thicknesses grown at different V/III precursor ratios.

Using an optimized LT-GaN layer, we have demonstrated the feasibility to produce free-

standing GaN substrates. To study the morphology, structural and optical properties of the GaN

layers, X-ray diffraction (XRD), scanning electron microscopy (SEM), atomic force

microscopy (AFM), cathodoluminiscence (CL), white light reflectance spectroscopy (WLRS)

and optical microscopy have been used. The study has shown that both the V/III ratio used for

the growth and the thickness of the LT-GaN buffer layer strongly influence the quality of the

subsequent thick GaN layer.

4

2. EXPERIMENTAL

The growth was done in a vertical HVPE reactor at atmospheric pressure where the carrier

gases and precursors are delivered from the bottom. GaCl, NH3 and N2 or a mixture of N2 and

H2 is used as gallium source, nitrogen source and carrier gas, respectively. The GaCl is

synthesized in-situ by flowing HCl over a quartz boat containing liquid gallium held at 700-

750oC. A more detailed description of the growth equipment is given in Ref. 25. In order to

estimate partial pressure of process gases at the substrate surface and assure laminar flow

conditions, numerical simulations using the commercial software CFD-ACE were done. For

additional details about the simulations, see Ref. 26. In the experiments, Al2O3 (0 0 0 1)

substrates were used.

Figure 1 illustrates the different growth steps. Before growth of the LT-GaN buffer layer,

the sapphire substrate was exposed to NH3 during 10 min at 1050 oC. Thereafter, the

temperature was lowered to 600 oC and the LT-GaN buffer was grown. The V/III ratio at the

substrate was varied in the range 6-70 with a constant GaCl partial pressure of 2.8 mbar.

Within this range of V/III ratio the thickness of the buffer layers were varied in the range of 30

- 370 nm. N2 was used as a carrier gas in this step. After the deposition of the buffer layer, the

temperature was increased in presence of NH3 to 1050 o

C and the layer was annealed for 10

min in order to improve the crystallinity of the layers. Finally, the temperature was lowered to

1000oC where the thick high temperature GaN (HT-GaN) layer was grown. The growth was

done with a GaCl partial pressure at the substrate of ~20 mbar and a V/III ratio of 4. As carrier

gas, a mixture of H2 and N2 was used. With these process parameters, the typical growth rate

was ~200 µm/hr. The thickness of the HT-GaN was varied in the range between 20 µm and 2

mm.

The SEM and CL images were done with a Leo 1500 Gemini scanning electron

microscope equipped with a MonoCL system from Oxford Instruments using a 1200 lines/mm

5

grating blazed at 500 nm and detected by a Peltier-cooled photomultiplier. Structural

characterization was carried out with a Philips MRD XRD system with K1 (=1.54Å)

radiation source operating at 45 kV and 40 mA. A 11 mm2 point source and a parallel plate

collimator were used to analyze the diffracted radiation. The AFM measurements were

performed with a Veeco Dimension 3100 SPM system in tapping mode. The thickness of the

LT-GaN buffer layers were determined by white light reflectance spectroscopy (WLRS)

measurements. For very thin layers (<80 nm) when this technique cannot be used, the thickness

was estimated from the growth time.

3. RESULTS AND DISCUSSION

Figure 2 shows SEM images of the morphology as a function of LT-buffer layer thickness

grown with a V/III ratio of 37, before annealing [Fig.2 (a-d)], after annealing at 1050 oC for 10

min [Fig.2 (e-h)] and optical microscopy images after the buffer layers have been overgrown

with a 20 µm thick HT-GaN layer [Fig.2 (i-l)] .

For LT-buffer layers before annealing with a thickness of 100 nm or more [Fig.2 (b)-(d)],

we observe a continuous film consisting of GaN nanocrystallites with some inclusions of larger

crystallites. With increasing thickness, the size of the crystallites is decreasing. For the thinnest

film [Fig.2 (a)], large crystallites are dominating the surface and a continuous layer of GaN has

not been formed.

The subsequent annealing step [Fig.2 (e)-(h)] recrystallizes the layer and in case of LT-

buffers with a thickness of 100 nm or thicker, a continuous film consisting of hillocks with

some pits has been formed. We have also observed that with increasing thickness larger pits

and hillocks appear. For the thinnest buffer (Fig. 2(e)), some of the isolated crystallites become

larger at the expenses of the smaller ones due to Ostwald ripening.

6

Figure. 2 (i-l) shows the result when we grow a 20 µm thick HT-GaN film on the annealed

LT-GaN buffer layer. For the thinnest one (Fig. 2(i)), the morphology of the overgrown GaN

layer becomes poor since the substrate is only partially covered with LT-GaN. By using thicker

and fully coalesced buffer layers, we can grow single crystalline material (Fig. 2(j)-(l)).

However, we do observe a degradation of the morphology with increasing LT-buffer thickness

which we attribute to the increase of the pit size with increasing thickness of the LT-buffer

layer after annealing. Thus, the optimal morphology of the overgrown GaN layer is achieved

using a LT- GaN buffer layer with a thickness about 100-300 nm.

The influence of V/III-ratio on the morphology for LT buffer layers with a thickness of

~100 nm is illustrated before and after annealing at 1050oC for 10 min in Fig. 3 (a-d) and (e-h),

respectively. After annealing, these LT-buffers were overgrown with a 20 µm thick HT-GaN

layer and the morphology was studied by optical microscopy as shown in Fig.3 (i-l). All

studied LT-GaN films with an optimal thickness of ~100 nm were uniform for all studied V/III

ratios (Fig.3 (a-d)). However, with reduction of the V/III-ratio, the density of larger GaN

crystallites increases. After annealing at 1050oC (Fig. 3(e-h)), a rather dramatic change of the

morphology was observed. For high V/III ratio, a surface mainly consisted of hillocks was

observed, while for the lower V/III ratio the surface appeared to have crater-like features. The

morphology of the LT-GaN film after annealing has a strong influence on the quality of the

HT-GaN, as can be seen in Fig. 3(i-l). Thus, in order to obtain a smooth surface with a low pit

density, the growth should be performed on buffer layers synthesized in a nitrogen rich

environment, i.e. with a V/III ratio exceeding 17. However, in these films cracking formation

has been observed. This is a common problem of heteroepitaxy of GaN on Al2O3, when tensile

stress is generated by an island coalescence mechanism in the beginning of the growth on

substrates with large lattice mismatch [27, 28]. For HT-GaN layers grown on non-optimized

7

LT buffers, i.e. obtained with too low V/III ratio or which are too thin or too thick, no cracking

due to tensile stress was detected.

In order to find out how the surface roughness of the buffer layer influences the HT-GaN

quality, AFM measurements where done. The roughness of 100 nm thick LT-GaN buffer

layers before annealing is shown as a function of V/III ratio in Fig. 4(a). Fig. 4(b) shows how

the surface roughness develops with the LT-buffer thickness grown with a fixed V/III ratio of

37. As can be seen in Fig. 4(a) the roughness of the buffer layer increases step-like for samples

grown with a V/III ratio less than 37. The surface also became rougher with increasing LT

buffer thickness, although the tendency is less. The step-like increase in roughness for low

V/III ratios correlates with the drastic change from a hillock-like to a crater-like morphology

for LT buffers after annealing as was shown in Fig. 3 (e-h). The rougher surface of LT buffers

grown with low V/III ratios might prevent an efficient diffusion of species, lowering the

surface mobility during annealing, and hence, prevent formation of hillocks.

To determine crystallographic properties of the GaN buffer layers, XRD 2 scans were

done. Figure 5 shows the results for two LT-GaN layers with a thickness of ~250 nm grown

with a V/III-ratio of 17 and 37, respectively. The position of the peaks corresponds to wurtzite

GaN confirming that the LT-GaN buffer is wurtzite. Samples grown with a V/III ratio of 17

show except the main (001) peak additional features related to (100) and (101) reflections

indicating formation of misoriented domains. However, for LT-buffers obtained with a V/III-

ratio of 37 only peaks related to the c-planes (001) are observed.

Fig. 6 shows the results of XRD measurements for 20 µm thick HT-GaN layers grown on

different LT-GaN buffer layers. Fig. 6 (a) shows full width half maxima (FWHM) of XRD

rocking curves of the symmetric (002) and asymmetric (105) reflections versus thickness of the

LT-GaN buffer layers grown with a fixed V/III-ratio of 37. FWHM of the rocking curve for

both the asymmetric and symmetric reflections shows a non-monotonic behavior with a

8

minimum in the range of ~100-300 nm. Thus, according to these measurements, an optimal

buffer thickness is confirmed to be in the range of ~100-300 nm. The FWHM for a 20 µm thick

HT-GaN grown on such LT-buffers was about 400 and 250 arcsec for the symmetric (002) and

for the asymmetric (105) reflections, respectively. The increasing FWHM in case of the

thinnest LT-buffer layer is probably associated with the poor film coverage as shown in Fig.

2(a) while for the thicker buffers the broadening of the XRD peaks is likely related to increased

pit density in LT-buffers as was illustrated in Fig. 2(h).

We have also studied the XRD FWHM for the HT-GaN samples with a thickness of 20

µm grown on LT-GaN buffers obtained at different V/III ratio. With increasing V/III ratio, the

FWHM of the asymmetric (105) reflection dropped while the FWHM of the symmetric (002)

reflection is slightly increased as shown in Fig. 6(b).

The rocking curve FWHM has been associated with the TD density as the dislocations

distort particular reflection planes. The symmetric reflection such as the (002) reflection is

insensitive to pure edge type TD while scans related to the asymmetric planes like (105) are

progressively more sensitive to the edge-type TD [29]. Using this interpretation, our results

indicate that with increasing V/III-ratio, there is a decrease in the density of TD with an edge

component, but no significant change in the density of TD with screw components. It has been

shown in Ref. [30] that edge type threading dislocations are localized at grain boundaries.

Thus, a reduction in the density of edge type TD indicates larger diameter of the grains in the

film and, consequently, the grain diameter is increased in the HT-GaN films grown on LT-

buffer layers fabricated with higher V/III ratios.

To further investigate the nature of the TD, additional characterization by CL was

performed. Fig. 7 (a) shows a panchromatic CL image of a 20 µm thick HT-GaN layers grown

on a 120 nm thick LT buffer layer, which was produced using a V/III ratio of 70. For

comparison we show in Fig. 7(b) a CL image of a 20 µm thick HT-GaN layer grown on a 2 µm

9

thick GaN template grown by MOCVD on sapphire. It is clear that the dark pit density (DPD)

is similar for both layers, and is estimated to ~2 x108 cm

-2. Fig. 7(c) shows the DPD for several

20 µm thick HT-GaN layers fabricated on LT-GaN buffers of different thickness grown with a

V/III ratio of 17, 37 and 70, respectively. From this data we can conclude that the DPD is

practically independent of the LT-GaN buffer layer thickness and V/III-ratio. To demonstrate

that the DPD drops with increasing thickness of the HT-GaN layer we show in Fig. 7 (d) a CL

image of a 500 µm thick GaN layer grown on a 170 nm thick LT-buffer. This image reveals

that the DPD is ~2 x107 cm

-2, i.e. it has dropped one order of magnitude. The dark pits

observed in CL is associated with TD [31], however, there has been a lot of discussion

concerning the interpretation of the DPD. It has been argued that the dark pits are associated

with TD having a pure screw component (burgers vector 0001b ) [32-34] while other

authors claims that this type of dislocation does not give rise to dark pits [35]. Considering the

results from the XRD measurements it is tempting to draw the conclusion that the DPD is

associated with screw type TD since the density of edge type TD dropped with increasing

V/III-ratio. However, it should be noted that CL is mainly probing the near surface region (1-2

µm) of the film while with XRD the whole volume is investigated.

One important aspect of LT-GaN buffers is that the sapphire substrate is spontaneously

separated from the GaN layer. In order to study this process, we have grown a number of

thicker HT-GaN layers with a diameter of 2 inch and we have found that for layers thicker than

~1 mm, the substrate is spontaneously released. Figure 8 shows the backside directly after

growth (a) and front side after lapping and mechanical polishing (b) of a layer having initial

thickness of 2 mm grown on a 200 nm-thick LT-GaN buffer obtained at a V/III ratio of 37.

The sapphire has been almost completely separated from the middle of the GaN layer.

However, at the rim of the boule some sapphire was still left so further improvement of the

process has to be done. The remaining sapphire was lapped off and the wafer was polished

10

from both sides. After polishing, it was revealed that the crystal had some cracks though they

have not penetrated the whole wafer so it still remains a diameter of 2 inch. The origin of the

cracks is unknown but there are several likely reasons such as cracking during cooling after the

growth and/or due to a non-optimized polishing process.

4. SUMMARY

We have studied growth of HT-GaN on LT-GaN buffers using different V/III ratios and

buffer growth times. The HT-GaN layers grown on LT-GaN buffers obtained with a V/III ratio

of 35 or 70 and a thickness of 100-300 nm demonstrate the smoothest surface morphology and

structural quality. The DPD of ~2108 cm

-2 measured on 20 µm thick HT-GaN layers were

found to be similar to DPD in HT-GaN layers grown on 2 µm-thick MOCVD fabricated GaN

templates. For a 500 µm-thick HT-GaN layer, DPD has dropped about one order of magnitude.

By using an optimized LT-GaN buffer layer, thick free freestanding 2’’ GaN substrates were

manufactured by spontaneous partial self-separation.

11

ACKNOWLEDGMENTS

This work was supported by the Swedish Research Science Council and the Swedish

Energy Agency.

12

REFERENCES

[1] Kyma Technologies www.kymatech.com, (2012, Aug. 21).

[2] Lumilog, France www.lumilog.com, (2012, Aug. 21).

[3] TopGan Ltd, www.topgan.fr.pl, (2012, Aug. 21).

[4] Ammono S.A., ammono.com, (2012, Dec. 11).

[5] T. Hashimoto, F. Wu, J.S. Speck, S. Nakamura, Nature Materials, 6 (8), (2007), 568

[6] E. Valcheva, T. Paskova, S. Tungasmita, P.O.A. Persson, J. Birch, E.B. Svedberg, L.

Hultman and B. Monemar, Appl. Phys. Lett., 76 (2000), 1860

[7] T. Detchprohm, K. Hiramatsu, H. Amano and I. Akasaki. Appl. Phys. Lett., 61 (1992),

2688.

[8] C. Hemmingsson, M. Boota, R. O. Rahmatalla, M. Junaid, G. Pozina, J. Birch, and B.

Monemar, J. Cryst. Growth, 311 (2009), 292

[9] D. Gogova, D. Siche, A. Kwasniewski, M. Schmidbauer, R. Fornari, C. Hemmingsson,

R. Yakimova and B. Monemar, Phys. Status Solidi C, 7 (2010), 1756

[10] Y. Oshima, T. Eri, M. Shibata, H. Sunakawa and A. Usui, Phys. Status Solidi A,

194 (2002), 554

[11] C. L. Chao, C. H. Chiu, Y. J. Lee, H. C. Kuo, P.-C. Liu, J. D. Tsay, and S. J. Cheng,

Appl. Phys. Lett., 95 (2009), 051905

[12] A. Usui, H. Sunakawa, A. Sakai, and A.A. Yamaguchi, Jpn. J. Appl. Phys., 36 (1997),

L899

[13] G. Nataf, B. Beaumont, A. Bouillé, S. Haffouz, M. Vaille, and P. Gibart J. Cryst.

Growth, 192 (1998), 73

[14] F. Lipski, T. Wunderer, S. Schwaiger and F. Scholz, Phys. Status Solidi A, 207

(2010), 128

13

[15] I. Akasaki, H. Amano, Y. Koide, K. Hiramatsu and N. Sawaki, J. Cryst. Growth, 98

(1989), 209

[16] M. Sumiya, N. Ogusu, Y. Yotsuda, M. Itoh, S. Fuke, T. Nakamura, S. Mochizuki, T.

Sano, S. Kamiyama, H. Amano, and I. Akasaki, J. Appl. Phys., 93 (2003), 1311

[17] P.R. Tavernier, E.V. Etzkorn, Y. Wang and D.R. Clarke, Appl. Phys. Lett., 77 (2000),

1804

[18] H. M. Kim, J. E. Oh and T. W. Kang , H. M. Kim, J. E. Oh and T. W. Kang, Mater.

Lett., 47 (2001), 276

[19] K. Motoki, T. Okahisa, S. Nakahata, N. Matsumoto, H. Kimura, H. Kasai, K.

Takemoto, K. Uematsu, M. Ueno, Y. Kumagai, A. Koukitu and H. Seki, Mater. Sci. Eng.,

B 93 (2002), 123

[20] M.K. Kelly, R.P. Vaudo, V.M. Phanse, L. Gorgens, O. Ambacher and M. Stutzmann,

Jpn. J. Appl. Phys., 38 (1999), L217

[21] D. Gogova, A. Kasic, H. Larsson, C. Hemmingsson, B. Monemar, F. Toumisto, K

Saarinen, L. Dobos, B. Pécs, P. Gibart and B. Beaumont, J. Appl. Phys., 96 (2004), 799

[22] P. Gibart, B. Beaumont, and P. Vennegues, in “Nitride Semiconductors, Handbook on

Materials and Devices, Eds P Ruterana, M Albrecht and J Neugebauer, Wiley-VCH (2003),

45.

[23] Y. Oshima, T. Eri, M. Shibata, H. Sunakawa, K. Kobayashi, T. Ichihashi, and A. Usui,

Jpn. J. Appl. Phys., 42 (2003), L1

[24] M. Sumiya, N. Ogusu, Y. Yotsuda, M. Itoh, S. Fuke, T. Nakamura, S. Mochizuki, T.

Sano, S. Kamiyama, H. Amano, and I. Akasaki, J. Appl. Phys., 93 (2003), 1311

[25] C. Hemmingsson, P.P. Paskov, G. Pozina, M. Heuken, B. Schineller and B. Monemar,

Superlattices Microstruct., 40 (2006), 205

[26] C. Hemmingsson, G. Pozina, M. Heuken, B. Schineller and B. Monemar

14

J. Crystal Growth , 310 (2008), 906

[27] E.V. Etzkorn and D.R. Clarke, J. Appl. Phys., 89 (2001), 1025

[28] W.D. Nix and B.M. Clemens, J. Mater. Res., 14 (1999), 3467

[29] B. Heying, X.H Wu,, S. Keller, Y. Li, D. Kapolnek, B.P. Keller, S.P. Denbaars and

J.S. Speck, Appl. Phys. Lett., 68 (1996), 643

[30] R. Chierchia, T. Böttcher, H. Heinke, S. Einfeldt, S. Figge and D. Hommel, J. Appl.

Phys., 93 (2003), 8918

[31] S.J. Rosner, E.C. Carr, M. J. Ludowise, G. Girolami, and H.I. Erikson, Appl. Phys.

Lett., 70 ( 1997), 420

[32] E.G. Braze, M.A. Chin and V. Narayanamurti, Appl. Phys. Lett., 74 (1999), 2367

[33] G. Koley and M.G. Spencer, Appl. Phys. Lett., 78 (2001), 2873

[34] T. Hino, S. Tomiya, T. Miyajima, K. Yanashima, S. Hashimoto and M. Ikeda, Appl.

Phys. Lett., 76 (2000), 3421

[35] M. Albrecht, J.L. Weyher, B. Lucznik, I. Grzegory and S. Porowski, Appl. Phys. Lett.,

92 (2008), 231909

15

Figure captions

Fig. 1 Schematic diagram of the growth process.

16

Fig. 2 SEM images of the morphology as a function of buffer layer thickness for LT-GaN

layers grown with a V/III ratio of 37, before annealing (a-d), after annealing at 1050 oC for 10

min (e-h) and microscopy images after they have been overgrown with a 20 µm thick HT-GaN

layer (i-l) .

17

Fig. 3 SEM images of the morphology as a function of V/III-ratio for LT buffer layers with a

thickness of ~100nm, before annealing (a-d), after annealing at 1050 oC for 10 min (e-h) and

microscopy images after they have been overgrown with a 20 µm thick HT-GaN layer (i-l).

Fig. 4 (a) Roughness (RMS) for a 100 nm thick LT-GaN before annealing using various V/III

ratios and (b) roughness of LT-GaN buffers before annealing versus thicknesses using a V/III

ratio of 37. The line is guidance for the eye.

18

Fig. 5 2 XRD-scans of 250 nm thick LT-GaN buffer layers grown with a V/III- ratio of 17

and 37.

Fig. 6 FWHM of XRD rocking curves of the symmetric (002) and asymmetric (105) reflection

of 20 µm thick HT-GaN layers grown on LT-GaN buffer layers of (a) different thicknesses

using a V/III-ratio of 37 and (b) different V/III ratios using a LT-GaN buffer thickness of 100

nm.

19

Fig. 7 Panchromatic CL images taken at room temperature for (a) 20 µm thick GaN layer on a

120 nm thick LT-GaN buffer layer grown with a V/III ratio of 70, (b) 20 µm thick GaN layer

grown on a 2 µm thick MOCVD GaN template on sapphire, (c) dark pit density for 20 µm

thick GaN layers using various LT-GaN buffer layer thicknesses and V/III ratios and (d)

panchromatic CL image of a 500 µm thick GaN layer grown on a 170 nm LT-GaN buffer layer

obtained with a V/III ratio of 70.

1x107

100 150 200 250

Dark

pit

den

sit

y [

cm

]-2

Buffer thickness [nm]

V/III:17

V/III:37

V/III:70

1x108

1x109

300

(c)

2 µm

(d)

2 µm

(b)

2 µm

(a)

20

Fig. 8 Images of a 2 mm thick GaN layer grown on a LT-buffer layer (thickness=200nm, V/III-

ratio=37). (a) backside (nitrogen side) with some remaining sapphire and (b) front side

(gallium side) after removal of sapphire and polishing.