ARTICLE IN PRESS

Journal of Crystal Growth 303 (2007) 314317 www.elsevier.com/locate/jcrysgro

Critical thickness calculations for InGaN/GaN

Abstract A model based on the overall energy balance is used to calculate the critical thickness for an InGaN epitaxial layer on a GaN substrate. The critical thickness values as a function of the indium content are found to be lower than values predicted by models proposed by Fischer or People and Bean. We also used the energy balance model to estimate the effect of the hexagonal symmetry of wurtzite materials on the critical thickness; this results in reduction of the critical thickness by as much as 20% of its corresponding isotropic value. From the small amount of experimental data available we conclude that the energy balance model is more appropriate for describing the critical thickness of the InGaN/GaN material system than the models developed by Fischer or People and Bean. r 2007 Elsevier B.V. All rights reserved.
PACS: 61.72.y; 68.37.Lp; 81.05.Ea; 81.15.Aa Keywords: A1. Line defects; B1. Nitrides; B2. Semiconducting gallium compounds

1. Introduction Gallium nitride (GaN) and its alloys have become important semiconductor materials as they allow the design and production of devices with unique optoelectronic properties. Light emitting diodes (LEDs) based on wurtzite InGaN/GaN can produce bright light with colours ranging from green to near ultra-violet. Despite the rapid research developments in the III-nitride semiconductor area, there are still crucial issues to be addressed in order to increase the efciency of LEDs for the future success of solid-state lighting. An example is the difculties associated with the growth of layers with high indium content for the longer wavelengths, the increased lattice mismatch resulting in the formation of dislocations in the active region that may decrease the efciency of LEDs. A recent transmission electron microscopy (TEM) study of InGaN/GaN quantum well (QW) structures revealed the presence of dislocation half-loops which contained mist dislocation (MD) [1]. While experiments are being carried out to study the effect of the InGaN QW growth temperature on the dislocation formation, the present
Corresponding author. Tel.: +44 1223 334404; fax: +44 1223 34437.

E-mail address: [email protected] (D. Holec). 0022-0248/$ - see front matter r 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.jcrysgro.2006.12.054

study will investigate theoretically MD formation in the InGaN/GaN material system. The generation of MDs is a well-known phenomenon in hetero-epitaxy, in which a thin lm is grown on a substrate with signicantly different lattice parameters. Below a certain lm thickness, called the critical thickness hc , a hetero-epitaxial lm may be grown pseudomorphically on a substrate, while a relaxation of mist strain via plastic ow occurs for thicker lms h4hc . In this sense, the most common mechanism of plastic relaxation is through the formation of MDs. Several theoretical models for calculating the critical thickness in isotropic materials have been published over the years [24]. Although these account reasonably well for the strain relaxation processes occurring in cubic systems such as SiGe/Si and InGaAs/GaAs, there has not been, to the best of our knowledge, any report attempting to estimate the effect of hexagonal symmetry on the hc values. In this study we will present calculations of the critical thickness for an InGaN/GaN single-layer system using a model based on balancing energy changes in the system [5] and the effect of the hexagonal symmetry of the III-nitride materials will be evaluated. Calculated hc values will be compared with experimental values reported for various InGaN/GaN systems.

ARTICLE IN PRESS
D. Holec et al. / Journal of Crystal Growth 303 (2007) 314317 315

2. Critical thickness model Freund and Suresh [5] proposed a simple model to calculate hc values based on a balance of energy (hereafter called the energy balance model). Compared to the elastically strained pseudomorphic epilayer, one straight innitely long dislocation, which has glided from the free surface down into the hetero-interface (see Fig. 1), changes the total energy (per unit length) of this system by DE dE d dW , dl dl (1)

15

12 A, B, C [GPa]

A B C

where dE d =dl is the dislocation energy and dW =dl is the work done by the system during the dislocation glide to partially relax the mist strain. For DE40, the dislocation introduces some excess energy compared to the pseudomorphic layer and thus it is not energetically favourable. Conversely, for DEo0, the dislocation relieves some energy and thus reduces the overall energy stored in the system. Therefore, a critical thickness criterion reads: DE 0. (2)

0 0 0.1 0.2 indium content x 0.3

Fig. 2. Values of coefcients from the expression of the dislocation energy (see Eq. (3)).

where h is the layer thickness, cij are the elastic constants and the mist strain m is given by m as al . al (5)

Following the procedure described by Steeds [6], we have calculated numerically dE d =dl assuming an innitely long straight dislocation lying on the hexagonal c-plane. The result can be written in a general form dE d R Ab2 Bb2 Cb2 ln , e;k e;? s r0 dl (3)

as and al are in-plane lattice constants of a substrate and a thin layer, respectively. 3. Results For the isotropic case, where an analytical formula for dE d =dl is available, the presented model gives the same results as the critical thickness model developed by Matthews and Blakeslee [2]. The calculated hc values versus the indium content x of the InGaN/GaN system for both the isotropic and the hexagonal symmetry approximations are shown in Fig. 3. The differences between the two curves in Fig. 3 become quite signicant for the alloys with a higher indium content as is demonstrated in Fig. 4 where the predicted hc values can be reduced by as much as 20% in the examined composition range due to the hexagonal symmetry. For example, for indium content 20% (corresponding usually to green light emitting InGaN/GaN materials), the isotropic hiso is 4.3 nm, c whereas with the hexagonal symmetry the model predicts hhex 3:6 nm. It is worth noting that these values are very c close to the thicknesses of the QWs in InGaN-based LEDs. The comparison with some other frequently used models, in particular Fischers [3] and People and Beans [4], are shown in Fig. 5 together with published experimental data, obtained using photoluminescence (PL) [9,10] and mobility and conductivity measurements [9]. Generally, the experimental results seem to follow Fischers model rather than the energy balance models curve. However, PL and electrical measurements are indirect observations based on the assumption that the onset of

where be;k , be;? and bs are components of the Burgers vector (an edge component lying on 0 0 0 1 plane, an edge component perpendicular to 0 0 0 1 plane and a screw component, respectively). The symbols R and r0 represent the outer and inner cut-off radii of the elastic media adjoining the dislocation [7]. Calculated values of constants A, B and C as functions of In concentration are shown in Fig. 2. We used in our calculations the slip system proposed by Srinivasan et al. [8] who concluded that the possible active slip system on which MDs can glide down from an InGaN free surface, was 1h1 1 2 3i=f1 1 2 2g. The work dW =dl is 3 given by dW c11 c12 c33 2c2 13 m be;k h, dl c33
free surface slip plane
e

(4)

id

h 2r 0

layer

gl

substrate
Fig. 1. A cross-sectional view of a lmsubstrate hetero-interface illustrating the glide of an innitely long straight dislocation (perpendicular to the plane of this sketch) from a free surface.

ARTICLE IN PRESS
316 D. Holec et al. / Journal of Crystal Growth 303 (2007) 314317

1000
isotropic approximation hexagonal symmetry

100

10

1 0 0.1 0.2 indium content x 0.3

strain relief has an immediate visible effect on these parameters. On the other hand, TEM affords a reliable way to directly image dislocations through dark-eld imaging analysis of different reciprocal lattice vectors g. Due to its local and non-ambiguous analysis characteristics, TEM imaging does not incur in the possible overlook of dislocations due to averaging processes that may take place in other techniques. Recent TEM analysis of InGaN/ GaN single (SQW) and multiple quantum well structures (MQWs) revealed the presence of MDs in the InGaN QW of 3 nm thickness and with an indium content of 20% (see Fig. 6) but not in MQWs of similar thickness but with 16% indium [1]. 4. Conclusions

Fig. 3. The calculated critical thickness against composition curves for the single-layer Inx Ga1x N=GaN system. The two curves correspond to the isotropic approximation [2] and to the energy balance model with the proper hexagonal symmetry of wurtzite materials.

critical thickness hc [nm]

We have calculated critical thickness values for the Inx Ga1x N=GaN system using the energy balance model.

0.92

hc hexagonal / hc isotropic

0.88

0.84

0.80

0.76 0 0.1 0.2 indium content x 0.3

Fig. 4. The effect of the hexagonal symmetry on the critical thickness values for the Inx Ga1x N=GaN single-layer system. The hexagonal symmetry lowers the CT values by about 1020% when compared to the isotropic case.

Fig. 6. Weak-beam dark-eld TEM micrograph of an In0:2 Ga0:8 N=GaN single quantum well structure grown on sapphire substrate and GaN underlying layer g5g 1 1 2 0.

1000 critical thickness hc [nm]

100

10

1 0 0.1 0.2 indium content x 0.3

Fig. 5. Comparison of the critical thickness values calculated for Inx Ga1x N=GaN single-layer system using different models with those obtained experimentally by photoluminescence [9,10] and by TEM [1]. As regards the data taken from Ref. [1], only the 16% MQW (empty square symbols) did not show the presence of MDs.

ARTICLE IN PRESS
D. Holec et al. / Journal of Crystal Growth 303 (2007) 314317 317

Consideration of the proper hexagonal symmetry of wurtzite (In)GaN material results in lowering the hc values by about 1020% for the range of indium content 0oxo0:3 when compared to the corresponding isotropic values. From the small amount of experimental data available we conclude that the energy balance model is more appropriate for describing the critical thickness of the InGaN/GaN material system than the models developed by Fischer et al. [3] and People and Bean [4].

References
[1] P.M.F.J. Costa, R. Datta, M.J. Kappers, M.E. Vickers, C.J. Humphreys, D.M. Graham, P. Dawson, M.J. Godfrey, E.J. Thrush, J.T. Mullins, Phys. Stat. Solidi (A) 203 (2006) 1729. [2] J.W. Matthews, A.E. Blakeslee, J. Crystal Growth 27 (1974) 118. [3] A. Fischer, H. Kuhne, H. Richter, Phys. Rev. Lett. 73 (1994) 2712. [4] R. People, J.C. Bean, Appl. Phys. Lett. 47 (1985) 322. [5] L.B. Freund, S. Suresh, Thin Film Materials: Stress, Defect Formation, and Surface Evolution, Cambridge University Press, Cambridge, 2003. [6] J.W. Steeds, Introduction to Anisotropic Elasticity Theory of Dislocations, Clarendon Press, Oxford, 1973. [7] J.P. Hirth, J. Lothe, Theory of Dislocations, Krieger Publishing Company, Malabar, FL, 1982. [8] S. Srinivasan, L. Geng, R. Liu, F.A. Ponce, Y. Narukawa, S. Tanaka, Appl. Phys. Lett. 83 (2003) 5187. [9] M.J. Reed, N.A. El-Masry, C.A. Parker, J.C. Roberts, S.M. Bedair, Appl. Phys. Lett. 77 (2000) 4121. [10] C.A. Parker, J.C. Roberts, S.M. Bedair, M.J. Reed, S.X. Liu, N.A. El-Masry, Appl. Phys. Lett. 75 (1999) 2776.

Acknowledgements The EPSRC and the EU project PARSEM (Contract Number MRTN-CT-2004-005583) are acknowledged for nancial support. D.H. would like to thank Professor Mick Brown for many fruitful discussions.