Characterization of the Crystalline Quality on GaN on Sapphire

and Ternary Alloys

Hiroshi Amano, Tetsuya Takeuchi, Shigeo Yamaguchi, Christian Wetzel, and Isamu Akasaki

Department of Electrical and Electronic Engineering, Meijo University, Nagoya, Japan 468- 8502

SUMMARY

The growth of GaN on sapphire by OMVPE using a

low-temperature deposited buffer layer was characterized

by in situ TEM observation. The crystalline quality, espe-

cially tilting and twisting of the mosaic GaN crystal, was

characterized by conventional X-ray diffraction together

with grazing-incidence X-ray diffraction. Strain and relaxa-

tion of the AlGaN and GaInN on GaN were also charac-

terized by asymmetrical X-ray diffraction. © 1998 Scripta

Technica, Electron Comm Jpn Pt 2, 81(10): 48�54, 1998

Key words: Low temperature deposited buffer

layer; mosaic structure; twist and tilt; strain and relaxation.

1. Introduction

Group III nitride-based blue and green light emitting

diodes (LED) have already been commercialized, and laser

diodes (LD) in the near ultra-violet (UV) and deep purple

region have been fabricated. Moreover, a high-power field-

effect transistor that operates in the GHz region, and also a

solar-blind high-sensitivity UV detector have been con-

structed. All of these novel devices are fabricated on highly

lattice-mismatched substrates, such as sapphire or 6H-SiC.

This is quite extraordinary if we compare it with other

systems, such as GaAs-based devices or InP-based devices,

in which lattice matching is believed to be the most impor-

tant requirement.

In this paper, we will report the growth modes, espe-

cially the initial growth stage, of GaN on sapphire at the

atomic scale, and the crystalline properties of GaN on

sapphire. Ternary alloys such as AlGaN and GaInN are also

important for the fabrication of novel devices. We also

report on the crystalline quality of ternary alloys on GaN.

2. Growth Mode of GaN on Sapphire at

Atomic Scale

When c-axis�oriented GaN is grown on (0001) sap-

phire, the in-plane lattice mismatch can be as large as 16.1%

at room temperature. Therefore, it is quite difficult to grow

high-quality GaN directly on a sapphire substrate. In the

following discussion, we assume fairly controlled condi-

tions for the growth of nitrides by organometallic vapor

phase epitaxy (OMVPE). Figure 1(a) shows schematically

the growth process of GaN directly grown on sapphire

substrate. In the initial stage, three-dimensional GaN is-

lands are partially formed. The origin of GaN island forma-

tion is either an atomic step side of sapphire or a dislocation

site. In order to relieve the large stress caused by the

difference of the in-plane lattice constants between GaN

and sapphire, the islands might be grown with slight tilting

and/or twisting compared with the crystal orientation of the

sapphire. In the next stage, due to a difference in the growth

rate of GaN on sapphire and that on GaN, or, in other words,

due to the difference in the consumption rate of the source

gases (in this case Ga-alkyls such as TMGa or TEGa),

selective growth occurs. The consumption rate of Ga-alkyls

is expected to be much higher on a GaN surface. Therefore,

CCC8756-663X/98/100048-07

© 1998 Scripta Technica

Electronics and Communications in Japan, Part 2, Vol. 81, No. 10, 1998Translated from Denshi Joho Tsushin Gakkai Ronbunshi, Vol. J81-C-II, No. 1, January 1998, pp. 65�71

48

the three-dimensional GaN islands with tilting and twisting

become larger and larger until they coalesce. Large crystal

defects are formed at the interfaces between islands. Such

GaN shows strong n-type conductivity and low electron

mobility, and p-type GaN had never previously been

achieved.

In order to improve the crystalline quality of GaN on

sapphire, first, a single crystalline AlN layer was interme-

diately grown between the GaN and sapphire substrate by

molecular beam epitaxy [1]. However, the residual donor

concentration was still in excess of 1019 cm�3, and the Hall

mobility was less than 50 cm2/Vs. By using a low-tempera-

Fig. 1. (a) Schematic viewgraph of the OMVPE growth of GaN directly on sapphire. (b) Schematic viewgraph of the

crystallization of a low-temperature AlN buffer layer deposited by OMVPE on sapphire, observed by TEM with high-

temperature sample holder. [8] (c) Schematic viewgraph of the initial growth of GaN on crystallized AlN buffer layer. [9]

49

ture deposited AlN [2] or GaN [3] buffer layer before the

growth of GaN at high temperature, the situation was

changed dramatically, and high-quality GaN with a low

residual donor concentration, a high electron mobility, a

narrow X-ray rocking curve, free exciton emission, and

stimulated emission at room temperature, among other

features, have been obtained. This finding led to the first

successful p-type GaN [4, 5], pn-junction LED [4], and

blue/UV-LED [6], and to observation of the quantum size

effect [7].

Figure 1(b) schematically shows the crystallization

process while ramping up the temperature of the low-tem-

perature deposited AlN buffer layer (LT-AlN) on sapphire,

and Fig. 1(c) shows schematically the crystal growth proc-

ess of the high-temperature grown GaN layer (HT-GaN) on

crystallized LT-AlN as observed by in situ transmission

electron microscopy (TEM) [8] and high resolution scan-

ning electron microscopy (SEM) [9], respectively. In most

cases, the thickness of the LT-AlN is about 20 to 100 nm

and the deposition temperature is about 350 to 600 °C. As-

deposited LT-AlN consists of very fine crystallites. During

ramping up from the deposition temperature to the epitaxial

temperature, about 950 to 1100 °C, the fine crystallites

merge to form large grains. This atomic rearrangement

occurs from the sapphire side and proceeds to the surface

side. When the substrate temperature reaches the epitaxial

temperature, truncated hexagonal mesa islands surrounded

by (0001) and {101_

1} facets are formed, and the whole

LT-AlN layer is then composed of such islands, forming a

brick-like structure. Compared to GaN islands grown di-

rectly on sapphire, the process of lattice mismatch relaxa-

tion should be different. In the case of HT-GaN growth on

LT-AlN, selective growth like that of GaN directly grown

on sapphire may not occur, but orientational selectivity

occurs. Largely misoriented crystallites are surrounded by

dislocations and embedded. Finally, the crystal orientation

of the whole HT-GaN layer becomes aligned.

It is well known that an HT-GaN layer grown on

sapphire using LT-buffer layer has a mosaic structure com-

posed of columns with diameters of 0.1 mm to a few mm.

Next, methods of characterizing the crystalline quality of

these layers using X-ray diffraction, and the relationship

between these properties with other properties such as

optical properties and electrical properties, are discussed.

Each column itself contains few threading dislocations.

Threading dislocations are formed around each column

because the columns are slightly misoriented with respect

to each other. The misorientation is mainly divided into two

parts, tilting and twisting. Tilting is the misorientation of

the c-axis of the columns, while twisting is the in-plane

rotation of the columns. X-ray diffraction itself is a method

of measuring the diffraction profile in the reciprocal lattice

space. By using X-ray diffraction, not only the lattice

spacing, but also the tilting and twisting of the mosaic

structure can be characterized. Figure 2(a) shows schemati-

cally the X-ray diffraction profile from GaN having a

mosaic structure around the (0002) diffraction spot in the

Fig. 2. Schematic viewgraph of the X-ray diffraction

mapping in the three-dimensional reciprocal lattice space

from the GaN having mosaic spread; (a) shows X-ray

diffraction mapping around the (0002) diffraction;

(b) shows that around (101_0) diffraction. a Shows

diffraction profile taken by w-mode scan. b Shows

diffraction profile taken by 2q/w-mode scan

with analyzer crystal.

50

three-dimensional reciprocal lattice map (RLM). In the

ordinal X-ray diffraction measurement, the w-mode and

2q/w-mode are traces of �a� and �b�, respectively. If the

dispersion of the incident X-rays is negligibly small and

thickness of GaN is thick enough, �a� and �b� reflect the

tilt profile and fluctuations in the lattice constant, respec-

tively [10, 11]. In many cases, the tilt profile is wider than

the fluctuation in the lattice constant, which results in a

disk-like profile of X-ray diffraction in RLM. Table 1 shows

one example of the Dwc and Dc of GaN grown on sapphire

by OMPVE using an LT-AlN buffer layer. The lattice con-

stant c can be precisely measured by using such symmetri-

cal diffraction, while in a plane lattice, constant a can be

measured in combination with symmetrical diffraction and

asymmetrical diffraction such as (112_

4), (202_

4), (101_5). In

order to characterize twisting of the layer, we cannot use

such asymmetrical diffraction. We need to rock the sample

perpendicular to the rocking direction of symmetrical dif-

fraction measurement. As shown in Fig. 2(b), (1010) dif-

fraction is promising for measurement of both in plane

lattice constant a and twisting. In Fig. 2(b), �a� shows the

w-mode rocking curve profile, which reflects twisting of

the film, while �b� shows the 2q/w-mode profile which

reflects fluctuations in the lattice constant a. Experimental

results on Dwa and Da are given in Table 1.

3. Hetereoepitaxy of AlGaN and GaInN

Ternary Alloys on GaN

In order to fabricate high-performance LEDs or LDs,

heteroepitaxy of ternary alloys is essential. The binary

compounds AlN, GaN, and InN are lattice-mismatched

with each other. Therefore, understanding the crystalline

quality of AlGaN and GaInN, especially the strain and

relaxation caused by lattice mismatch, is one of the most

critical issues.

The in-plane lattice mismatch between AIN and GaN

is at most 2.4%, while that of InN and GaN can be as large

as 11%. In this paper, we characterize the crystalline quality

of AlGaN and GaInN on GaN, with AlN and InN molar

fraction of less than 0.3 for both ternary alloys.

In order to characterize the macroscopic strain fea-

tures of ternary alloys or the relaxation of the dislocations,

X-ray diffraction is very useful. By measuring the X-ray

diffraction profile in the RLM around the asymmetrical

diffraction spot, for example, (112_4) or (202

_4) in the recip-

rocal lattice map, we can directly characterize the strain and

relaxation situation of the ternary alloys on binary GaN.

Figure 4 schematically shows the RLM around (0004),

(202_4) and (101

_0) diffraction spots of the GaInN and Al-

GaN grown on GaN. The abscissa is proportional to the

inverse of the lattice constant a, while the ordinate is

proportional to the inverse of the lattice constant c. In the

figure, both the fully relaxed case and the fully strained case

are shown. If the abscissas of the peak positions of the

asymmetrical diffraction of the ternary alloys are just

aligned to that of GaN, the in-plane lattice constants of the

ternary alloys are the same as that of GaN. This is the fully

strained case, sometimes called coherent growth. But if the

peak positions of the asymmetrical diffraction of the ternary

alloys are almost on the same line toward (0000) as for GaN,

the ternary alloys should be fully relaxed by dislocations,

and so on. That is sometimes called the fully relaxed, or free

standing case. It is very clear from this figure that to

determine the alloy composition of the strained ternary

alloys, special attention is necessary. When the ternary

alloys are fully relaxed, we can determine the alloy compo-

sition by precisely measuring the 2q/w-mode symmetrical

diffraction profile and assuming Vegard�s law. Otherwise,

if ternary alloys are grown coherently, we measure the

2q/w-mode symmetrical diffraction and have to calculate

Table 1. FWHMs of X-ray diffraction profiles of GaN

grown by OMVPE obtained by 2q/w-mode

and w-mode scan

[arc sec.]

(0002) (101_

0)d [mm]

Dc Dwc Da Dwa

23 406 55 599 1.81

Fig. 3. Configuration for measuring the (101_0) X-ray

diffraction from c-plane GaN having small offset.

51

Fig. 4. Two-dimensional reciprocal space mapping around (202_4) diffraction of AlGaN and GaInN grown on GaN. Both

those of coherently grown AlGaN and GaInN and perfectly relaxed AlGaN and GaInN are represented.

Fig. 5. (a) AlGaN and (b) GaInN grown coherently on GaN. Actual critical layer thickness above which ternary alloys

show lattice relaxation is thought to be much thicker than these results.

52

by assuming Hooke�s law [12]. If the layers are strained but

partially relaxed, we need to measure both a and c.

Figure 5 shows the experimentally determined thick-

ness of AlGaN and GaInN, which showed coherent growth

on GaN. Theoretical results of Matthews and Blakeslee [13]

and of Fisher and others [14], both of whose approaches

assume a balance between the strain energy and energy of

generation of misfit dislocation, are shown for comparison.

As shown in the figure, the range of thickness of ternary

alloys in which the ternary alloys show coherent growth is

much wider than that of the misfit dislocations estimated

from the above-mentioned theory, and the critical thickness

for lattice relaxation is expected to be much thicker, as

shown by the arrow.

4. Conclusions

The growth modes of GaN on sapphire, the crystal-

line quality of GaN on sapphire, and the crystalline quality

of the ternary alloys AlGaN and GaInN on GaN are sur-

veyed. The density of threading dislocations in the GaN

layer grown on sapphire by OMVPE using LT-buffer layer

ranges from 108 to 1011 cm�2. Recently, by the using the

lateral epitaxial overgrowth technique, commonly used in

silicon-on-insulator growth [15], high-quality GaN with a

dislocation density of 106 to 107 cm�2 GaN has been

achieved [16]. Lateral epitaxial growth reduces threading

of dislocations. GaN with a much lower dislocation density

or even dislocation-free GaN will surely improve the per-

formance of LEDs, LDs, or FETS. They are also important

for identification of the intrinsic properties of nitrides.

Acknowledgments. This work was partly sup-

ported by the Ministry of Education, Science, Sports and

Culture of Japan (contract nos. 07505012, 0875083,

09450133, the High-Tech Research Center Project, and the

Japan Society for Promotion of Science�s �Research for the

Future� Program in the Area of Atomic Scale Surfaces and

Interface Dynamics (project Dynamical Process and Con-

trol of Buffer Layer at Interfaces in Highly Mismatched

Systems).

REFERENCES

1. S. Yoshida, S. Misawa, and S. Gonda. Appl. Phys.

Lett., 42, 427 (1983).

2. H. Amano, N. Sawaki, I. Akasaki, and Y. Toyoda.

Appl. Phys. Lett., 48, 353 (1986).

3. S. Nakamura. Jpn. J. Appl. Phys., 30, L1705 (1991).

4. H. Amano et al. Jpn. J. Appl. Phys., 28, L2112 (1989).

5. S. Nakamura, M. Senoh, and T. Mukai. Jpn. J. Appl.

Phys., 30, L1708 (1992).

6. I. Akasaki et al. Inst. Phys. Conf. Ser., 129, 851

(1992).

7. K. Itoh et al. Jpn. J. Appl. Phys., 30, 1924 (1991).

8. H. Amano et al. International Conference on Silicon

Carbide, III-Nitrides and Related Material, Stock-

holm, Aug. 31�Sep. 5 (1997).

9. K. Hiramatsu et al. J. Crystal Growth, 115, 628

(1991).

10. N. Itoh and K. Okamoto. J. Appl. Phys., 63, 1486

(1988).

11. Y. Koide et al. Jpn. J. Appl. Phys., 27, 1156 (1988).

12. T. Takeuchi et al. Jpn. J. Appl. Phys., 36, L177 (1997).

13. J.W Matthews and A.E. Blakeslee. J. Crystal Growth,

32, 265 (1974).

14. A. Fischer, H. Kuhne, and H. Richter. Phys. Rev.

Lett., 73, 2712 (1994).

15. See for example, Y. Kunii, M. Tabe, and K. Kajiyama.

Jpn. J. Appl. Phys., 21, 1431 (1982).

16. A. Usui, H. Sunakawa, A. Sakai, and A. Yamaguchi.

Jpn. J. Appl. Phys., 36, L899 (1997).

53

AUTHORS (from left to right)

Hiroshi Amano (nonmember). Education: Nagoya University, B.S., 1983, D.Eng. Academic appointments: Nagoya

University, research associate, 1988; Meijo University, lecturer, 1992. Research interests: group III nitride semiconductors.

Membership: Applied Physics Society of Japan, IEE Japan, MRS, APS.

Tetsuya Takeuchi (nonmember). Education: Nagoya University, B.S., 1990; M.S., 1992; currently doctoral candidate.

Industry positions: Hewlett-Packard Research Laboratories, 1992. Research interests: group III-V semiconductor crystal

growth. Membership: Applied Physics Society.

Shigeo Yamaguchi (nonmember). Education: Kyoto University, B.S., 1991; D.Eng., 1997. Academic appointments:

Meijo University High-Tech Research Center, 1997. Research interests: group II-VI semiconductor crystal growth, exciton

optics, group III nitride semiconductors. Membership: Applied Physics Society.

Christian Wetzel (nonmember). Education: Technical University of Munich, Diplom-Physiker, 1988; Ph.D., 1993.

Academic appointments: University of California, Berkeley; Materials Science Division, Lawrence Berkeley National Labora-

tory; Meijo University High-Tech Research Center, 1997. Research interests: electronic band and defect structure of low-di-

mensional and wide-bandgap semiconductor systems; high-sensitivity cross-modulation, external perturbation by hydrostatic

pressure; dopants and light emission in group III nitrides.

Isamu Akasaki (member). Education: Kyoto University, B.S., 1952; Nagoya University, D.Eng., 1964. Industry positions:

Kobe Industries, Ltd. (now Fujitsu, Ltd.), researcher, 1952�1959; Matsushita Research Ltd. (head of Tokyo Basic Research

Laboratory No. 4, 1964�1974; head of Semiconductor Division, 1974�1981). Academic appointments: Nagoya University,

research associate, lecturer, associate professor (1959�1964), professor, 1981�1992, professor emeritus; Meijo University,

professor, 1992. Membership: Applied Physics Society of Japan, IEE Japan, APS, MRS, ECS.

54