Hostname: page-component-586b7cd67f-tf8b9 Total loading time: 0 Render date: 2024-11-26T17:16:12.187Z Has data issue: false hasContentIssue false

Formation and stability of the prismatic stacking faultin wurtzite (Al,Ga,In) nitrides

Published online by Cambridge University Press:  13 June 2014

P. Ruterana*
Affiliation:
Laboratoire d’Etudes et de Recherches sur les Matériaux –Institut des Sciences de la Matière et du Rayonnement, UPRESA-CNRS 6004, 6 Bd Maréchal Juin, 14050 Caen Cedex, France
A. Béré
Affiliation:
Laboratoire d’Etudes et de Recherches sur les Matériaux –Institut des Sciences de la Matière et du Rayonnement, UPRESA-CNRS 6004, 6 Bd Maréchal Juin, 14050 Caen Cedex, France
G. Nouet
Affiliation:
Laboratoire d’Etudes et de Recherches sur les Matériaux –Institut des Sciences de la Matière et du Rayonnement, UPRESA-CNRS 6004, 6 Bd Maréchal Juin, 14050 Caen Cedex, France
*
1Author for correspondance: email : [email protected], Tel: 33 2 31 45 26 53, Fax: 33 2 31 45 26 60

Abstract

The formation of the {1 10} stacking fault, which has two atomic configurations in wurtzite (Ga,Al,In)N, has been investigated by high resolution electron microscopy and energetic calculations. It originates from steps at the SiC surface and it can form on a flat (0001) sapphire surface. A modified Stillinger-Weber potential was used in order to investigate the relative stability of the two atomic configurations. They have comparable energy in AlN, whereas the 1/2<10 1>{1 10} configuration is more stable in GaN and InN. In GaN layers, only the 1/2<10 1>{1 10} configuration was observed. The 1/6<20 3> configuration was found in small areas inside the AlN buffer layer where it folded rapidly to the basal plane, and when back into the prismatic plane, it took the 1/2<10 1>{1 10} atomic configuration.

Type
Research Article
Copyright
Copyright © 1996 Materials Research Society

Introduction

Extensive research effort is being undertaken on wide band gap of GaN based semiconductors for their very promising device possibilities. Efficient laser diodes have been fabricated for emission in the blue range[Reference Nakamura, Senoh, Nagahama, Iwasa, Yamada, Mitsushita, Kiyoku and Sugimoto1] in layers grown on sapphire by Metal Organic Vapor Epitaxy (MOVPE). The active GaN layers contain large densities of crystallographic defects[Reference Ponce, Bour, Götz, Johnson, Helava, Grzegory, Jun and Porowski2,Reference Vermaut, Ruterana, Nouet, Salvador and Morkoç3], among which, one finds {1

10} planar defects which have recently been called double positioning boundaries (DPBs)[Reference Tanaka, Kern and Davis4,Reference Smith, Chandrasekar, Sverdlov, Botchkarev, Salvador and Morkoç5], stacking mismatch boundaries (SMBs)[Reference Sverdlov, Martin, Morkoç and Smith6] or inversion domain boundaries (IDBs)[Reference Rouvière, Arlery, Niebuhr and Bachem7]. These faults have recently been investigated using High Resolution Electron Microscopy (HREM) and Convergent Beam Electron Diffraction (CBED), and it was shown that they are stacking faults on top of both sapphire and SiC[Reference Vermaut, Ruterana, Nouet, Salvador and Morkoç8-Reference Vermaut, Ruterana and Nouet10]. In fact these planar defects have already been studied in the sixties and two displacement vectors have been measured by conventional microscopy[Reference Drum11,Reference Blank, Delavignette, Gevers and Amelinckx12]. In wurtzite ZnS, Blank et al[Reference Blank, Delavignette, Gevers and Amelinckx12] were the first to study the planar defects which folded from basal to prismatic {1
10} planes, and to interpret them as stacking faults whereas other authors considered them to be thin lamella of the sphalerite phase in CdS[Reference Chadderton, Fitzgerald and Yoffe13]. In this work, we have investigated these faults in (Al,Ga)N epitaxial layers by HREM and modified the Stillinger-Weber[Reference Stillinger and Weber14] (SW) potential in order to analyze their stability in AlN, GaN, InN. The theoretical results were found to correlate with the HREM observations for AlN and GaN.

Results

The stacking fault atomic structure

Two atomic models exist in the literature for the { 11

0 } prismatic fault in wurtzite materials as originally characterized by conventional electron microscopy in the 60s by Drum[Reference Drum11] and Blank[Reference Blank, Delavignette, Gevers and Amelinckx12], respectively. Their projections are shown in figure 1 along [0001].

Figure 1. Atomic configurations of the { 11

0 }stacking fault seen along [0001], a) The ½ [ 10
1 ] displacement vector, b) The 1/6 [ 20
3 ] stacking fault

Formation mechanisms

A. Formation on (0001) 6H-SiC surface

The [0001] SiC surface exhibits steps of various heights. If one ignores the mismatch along the c axis which is small, the GaN and SiC can be geometrically consid

A 6H-portion under a step can be decomposed into a faulted 2H stacking, and depending on the type of step, the decomposition may result in a displacement vector between the two crystallites which grow on the adjacent terraces (fig.2). In fact between the two polytypes, the only displacement vectors d which can result from the decomposition are those of the hcp lattice ( the intrinsic I1, RI1 = 1/6 < 20

3 >, I2 , RI2 = 1/3 <10
0 > and the extrinsic E, RE = 1/2 <0001> ). Among them, I1 and E have 1/2c component along the c axis and may lead to the formation of an extended defect in the epitaxial layer (fig.2). The displacement vector identical to I2 is confined in the interface and may contribute to release the misfit strain. When all the step heights are taken account, it was found that the largest fraction of the faults that will be formed are I1 as shown in table 1

Figure 2. Generation of a prismatic stacking fault on a I1 type of step on the (0001) surface of 6H-SiC

Table 1. Type of defects forming at an interface step between 2H-GaN and 6H-SiC

B. Growth on (0001) sapphire

The continuation of the two hcp anionic lattices may be described by three types of polyhedra at the interface. They are built with oxygen (O) and nitrogen (N) at their vertices, the metal (M) at the centre: one tetrahedron (3O, M, N) that leads to upward polarity, one tetrahedron (O, M, 3N) and one octahedron (3O, M, 3N) that lead to downward polarity. All these factors are taken into account in order to analyze how the resulting translation can be accommodated in the GaN layers by the formation of planar defects in the epitaxial layer: stacking faults (SF) or inversion domains boundaries (IDB). For the wurtzite structure, three models of IDBs have been proposed[Reference Austerman and Gehman15-Reference Potin, Ruterana and Nouet18].

B.1. Growth on a single terrace

On an oxygen terminated surface, there are four ways to build a hcp stacking of nitrogen atoms, they are related to each other by one of the three stacking fault vectors, RSF, of the wurtzite structure (RI1, RI2 and RE). Since two families of tetrahedral sites are available for the Ga3+ cation, this leads to eight GaN stackings linked by the operator WCE = RSF or WCE = RSF + m with opposite polarities(fig.3). In the case of Al-terminated surface, the number of stackings is reduced to three polyhedra due to the aluminium position in the Al2O3 structure. The systematic analysis of the polyhedra combinations allows to determine the relative proportions of the planar defects induced by the operator WCE. For instance, if the interface polyhedra are exclusively downward tetrahedra, the RSF will be only RI1, and combinations of downward tetrahedra and octahedra lead to the formation of RI1 and RI2 faults in equal proportions which may explain the formation of stacking faults inside the epitaxial layers[Reference Barbaray, Potin, Ruterana and Nouet19-Reference Barbaray, Potin, Ruterana and Nouet21].

B.2. Growth on adjacent terraces

A step, between two terraces on which two GaN islands grow, introduces a translation TS between them. Its component along c axis can be reduced in RE=1/2 <0001> and a residual translation TR. The number of RE is calculated to minimize the absolute value of this residual translation. For instance, the step between two A terraces with a height of c/3 in Al2O3 (0.433 nm) is equivalent in GaN to 0.835 cGaN, or : TS = 2RE + TR = TR with TR = - 0.165 cGaN ≈ cGaN/6. By considering all the steps between A or B terraces (A or B layers of the hcp stacking), the TS translation can be written as : TS = aRE + TR with a = 0 or 1 and TR may take four values with the approximation

(table.2) : So, the operator relating adjacent terraces must involve this additional parameter, TR :
(1)

As this residual translation is never equal to c/2, it cannot be simply accommodated by the formation of prismatic stacking faults. The operator describing IDBs can be written in order to point out the c/8 translation found inside some of the models in addition to the mirror operation (1). When compared to the residual translation TR, the introduction of the IDB containing the c/8 translation may help in minimizing the shift along c at most of the steps; it is reduced to 1/24 cGaN[Reference Ruterana, Potin, Béré, Barbaray, Nouet, Hairie, Paumier, Botchkarev, Salvador and Morkoç20]. Therefore, in the case of steps, the residual translation cannot be minimized by introducing a stacking fault. During epitaxial growth of GaN layers on (0001) sapphire, the occurrence of basal and prismatic stacking faults is not a straightforward result of the steps at the substrate surface. They form on coalescence of islands whose growth has been initiated with different polyhedra[ Reference Ruterana, Potin, Béré, Barbaray, Nouet, Hairie, Paumier, Botchkarev, Salvador and Morkoç20 ].

Table 2. Residual translation in GaN as a function of the step height in Al2O3

Figure3. Four of the 8 possible stacking of wurtzite (Ga,Al)N on the (0001) sapphire surface, the polyhedra configurations at the interface are shown

Relative stability

The original SW potential is short and it cannot be used in order to predict the values of the non ideal c/a ratio and the vector u which characterizes the displacement along the c axis between the two hcp sublattices in the wurtzite structure. It is unable to reach the third nearest neighbour atom which is responsible of the deviation of c/a ratio from the ideal value. We have added a Gaussian term which was fitted in order to allow the interaction potential to reach the third nearest neighbour. It is given by:

(2)

where dij is the bond length between atoms i and j, r1 the center of the Gaussian is situated between the second and third neighbours distances and Δr1 is the width of the Gaussian function. According to the magnitude of r1 and the sign of C, it is possible to adjust the c/a ratio with respect to the ideal value. The potential obtained in the previous section is applied to investigate the atomic structure of the { 11

0 } stacking fault. The above atomic models were taken into account with the starting configurations corresponding to the 1/6 < 2
03 > and 1/2 < 1
01 > stacking fault displacement vectors as discussed in the previous paragraphs. We applied the method of lattice relaxation to a set of 1984 atoms containing the defect for the two atomic models. As seen in table 3, the formation energy of each configuration depends on the compound for AlN, GaN and InN. In AlN, the two atomic configurations of the { 11
0 } stacking fault have a comparable value of formation energy. This is in complete agreement with our HREM observations in GaN layers and in the AlN buffers22

Table 3. Formation energy in me V/

2 for the relaxed atomic configurations of the {11
0 } prismatic stacking fault, in parenthesis are the values of reference 23

Conclusion

The two configurations of the stacking fault have already been investigated by conventional electron microscopy, one in AlN and the other in ZnS respectively. They both were found to fold easily from the basal to the prismatic plane of these wurtzite materials. In the basal plane both have the 1/6< 20

3 > displacement vector of the I1 stacking fault. In the {11
0}plane, they differ by 1/6< 10
0 >, which is worth less than 4% in energetic balance between the two vectors (using the b2 criterion) in the three nitrides compounds.

References

Nakamura, S., Senoh, M., Nagahama, S., Iwasa, N., Yamada, T., Mitsushita, T., Kiyoku, H., and Sugimoto, Y., Jpn. J. Appl. Phys. 35, L74 (1996).CrossRefGoogle Scholar
Ponce, F.A., Bour, D. P., Götz, W., Johnson, N.M., Helava, H. I., Grzegory, I., Jun, J., and Porowski, S., Appl. Phys. Lett. 68, 917 (1995).CrossRefGoogle Scholar
Vermaut, P., Ruterana, P., Nouet, G., Salvador, A., and Morkoç, H., Inst. Phys. Conf. Ser. 146, 289 (1995).Google Scholar
Tanaka, S., Kern, R. S., and Davis, R. F., Appl. Phys. Lett. 66, 37 (1995).CrossRefGoogle Scholar
Smith, D. J., Chandrasekar, D., Sverdlov, B., Botchkarev, A., Salvador, A., and Morkoç, H., Appl. Phys. Lett. 67, 1803 (1995).Google Scholar
Sverdlov, B.N., Martin, G. A., Morkoç, H., and Smith, D. J., Appl. Phys. Lett. 67, 3064 (1995).CrossRefGoogle Scholar
Rouvière, J. L., Arlery, M., Niebuhr, R., and Bachem, K., Inst. Phys. Conf. Ser. 146, 285 (1995).Google Scholar
Vermaut, P., Ruterana, P., Nouet, G., Salvador, A., and Morkoç, H., Mater. Res. Symp. Proc. 423, 551 (1996).Google Scholar
Vermaut, P., Ruterana, P., Nouet, G., Salvador, A., and Morkoç, H., Phil. Mag. A75, 239, (1997).CrossRefGoogle Scholar
Vermaut, P., Ruterana, P., and Nouet, G., Phil. Mag. A76,1215, (1997).CrossRefGoogle Scholar
Drum, C. M., Phil. Mag., 11, 313 (1964).CrossRefGoogle Scholar
Blank, H., Delavignette, P., Gevers, R., and Amelinckx, S., Phys. Stat. Sol. 7, 747 (1964).CrossRefGoogle Scholar
Chadderton, L.T., Fitzgerald, A.F., and Yoffe, A.D., Phil. Mag. 8, 167 (1963)CrossRefGoogle Scholar
Stillinger, F.H., and Weber, T. A., Phys. Rev. B31, 5262 (1985).CrossRefGoogle Scholar
Austerman, S.B., and Gehman, W.G., J. Mater. Sci. 1, 249 (1966)CrossRefGoogle Scholar
Holt, D.B., J. Chem. Sol. 30, 1297 (1969)CrossRefGoogle Scholar
Romano, L.T., Northrup, J.E., and O’Keefe, M.A., Appl. Phys. Lett. 69, 2394 (1996)CrossRefGoogle Scholar
Potin, V., Ruterana, P., and Nouet, G., J. Appl. Phys. 82, 2176 (1997)CrossRefGoogle Scholar
Barbaray, B., Potin, V., Ruterana, P., and Nouet, G., Diamond and Related Materials, 8, 314 (1999)CrossRefGoogle Scholar
Ruterana, P., Potin, V., Béré, A., Barbaray, B., Nouet, G. , Hairie, A., Paumier, E. , Botchkarev, A. Salvador, , , A., and Morkoç, H., Phys. Rev. B59, 15917 (1999)CrossRefGoogle Scholar
Barbaray, B., Potin, V., Ruterana, P., and Nouet, G. Phil. Mag. A in pressGoogle Scholar
Vermaut, P., Ruterana, P., and Nouet, G., Appl. Phys. Lett. 74, 694 (1999)CrossRefGoogle Scholar
Northrup, J.E., Appl. Phys. Lett. 72 , 2316 (1998)CrossRefGoogle Scholar
Figure 0

Figure 1. Atomic configurations of the { 110 }stacking fault seen along [0001], a) The ½ [ 101 ] displacement vector, b) The 1/6 [ 203 ] stacking fault

Figure 1

Figure 2. Generation of a prismatic stacking fault on a I1 type of step on the (0001) surface of 6H-SiC

Figure 2

Table 1. Type of defects forming at an interface step between 2H-GaN and 6H-SiC

Figure 3

Table 2. Residual translation in GaN as a function of the step height in Al2O3

Figure 4

Figure3. Four of the 8 possible stacking of wurtzite (Ga,Al)N on the (0001) sapphire surface, the polyhedra configurations at the interface are shown

Figure 5

Table 3. Formation energy in me V/2 for the relaxed atomic configurations of the {110 } prismatic stacking fault, in parenthesis are the values of reference 23