- Email: [email protected]
Origins of defects in self assembled GaP islands grown on Si(001) and Si(111)
Thin Solid Films 357 (1999) 53±56 www.elsevier.com/locate/tsf
Origins of defects in self assembled GaP islands grown on Si(001) and Si(111) V. Narayanan a,*, N. Sukidi b, K.J. Bachmann b, S. Mahajan a a
Department of Chemical, Bio and Materials Engineering, Arizona State University, Tempe AZ 85287-6006, USA Department of Materials Science and Engineering, North Carolina State University Box 7919 Raleigh, NC 27695-7919, USA
b
Abstract Microstructures of GaP epitaxial islands grown on Si(001) and Si(111) by chemical beam epitaxy have been investigated by transmission electron microscopy (TEM). Results indicate that planar-defect free GaP islands of sizes ,20 nm can be produced at 5608C on Si(001). Some of the islands are faceted on {111} and {113} planes. Subsequent planar defect formation occurs due to stacking errors on the smaller {111} facets of GaP islands that may be P-terminated. These stacking errors are attributed to the low surface mobility on P-terminated facets. A high density of planar defects is observed in smaller islands grown on Si(001) at 4208C, a consequence of reduced atomic mobility at low temperatures that leads to {111} stacking errors. Wurtzite GaP has been observed to coexist with the zinc-blende polytype in some of the islands grown on Si(111) at 5608C. q 1999 Elsevier Science S.A. All rights reserved. Keywords: GaP islands; Si substrates; Defects; High resolution transmission electron microscopy
GaP and Si have a lattice mismatch of 0.37% at room temperature. Therefore, the two materials constitute an nearly ideal combination for the integration of Si and III± V technologies. A number of previous studies have indicated that GaP nucleates on Si in the form of islands [1,2]. These islands exhibit a high density of planar defects despite the low lattice mismatch. Further, the island shape and planar defect morphology is remarkably similar to that observed in GaAs/Si(001), a system with a lattice mismatch of 4% [3,4]. This suggests that planar defect formation in GaP/Si (001) may not arise from mis®t stresses, but rather from growth mistakes on faceted GaP islands. To understand the formation of planar defects, we have investigated the initial stages of heteroepitaxy on Si(001) and Si(111).
water rinse. GaP was deposited on Si using pulsed chemical beam epitaxy (CBE) [5] wherein, the heated Si substrate was exposed to pulses of tertiarybutylphosphine (TBP) and triethylgallium (TEG) with a steady hydrogen background pressure. The duration of the source vapor cycle tsvc repeated throughout the growth run was made up of the widths of the TBP and TEG pulses and pauses in between. The temperature range investigated was 420±5608C and the overall pressure during deposition was between 10 24 and 10 25 Torr. Using a cycle time tsvc of 5 s, TEG ¯ow of 0.05 sccm was pulsed into the reactor for 300 ms per cycle under continuous TBP and hydrogen ¯ow of 0.6 sccm and 5.0 sccm, respectively [7]. Structural analysis of as-grown layers was performed in cross-section by high resolution transmission electron microscopy (HRTEM) on a JEM-4000EX microscope operating at 400 kV and with a point to point resolution of 0.17 nm.
2. Experimental details
3. Results and discussion
The Si wafers were cleaned using an RCA clean. This consisted of a 10-min dip in a 1:1:5 solution of NH4OH, H2O2 and de-ionized (DI) water maintained at 758C, a 5min rinse in DI water, a 10-min dip in a 1:1:5 solution of HCl, H2O2 and DI water also kept at 758C and a rinse in DI water. This was followed by a buffered HF etch and a ®nal DI
Fig. 1 reveals the morphology of GaP islands after 80 s of growth at 4208C. This island has an average height of 8 nm and width of 18 nm and is faceted on {111} and {113} planes. Twins are seen parallel to the {111} planes. Fig. 2 shows an HRTEM image of `deposits' on the {111} facets of GaP islands grown on Si(001) at 5608C. The accommodation of lattice mis®t beyond a critical layer thickness between GaP and Si would require the
1. Introduction
* Corresponding author.
0040-6090/99/$ - see front matter q 1999 Elsevier Science S.A. All rights reserved. PII: S 0040-609 0(99)00474-5
54
V. Narayanan et al. / Thin Solid Films 357 (1999) 53±56
Fig. 1. A GaP Island on Si(001) after 80 s of growth at 4208C.
Fig. 2. A deposit on the {111} facet of a GaP island grown on Si(001) at 5608C.
nucleation of structurally necessary mis®t dislocations at the interface. The separation (S) between these dislocations is given by [6] S aGaP aSi =2 jaGaP 2 aSi j where aGaP, aSi are the unstrained lattice parameters of GaP and Si respectively. The factor of two arises from the fact that 608 mixed dislocations of the type a/2k011l (which can glide on {111} planes inclined to the surface and dissociate into 308 and 908 partials to form stacking faults) only accommodate half the lattice mis®t of Lomer (908) edge dislocations, because the Burgers vector is inclined to the interface. The separation distance between planar defects in Fig. 1 is much smaller than the calculated distance of 72 nm. In addition, the presence of thick twins cannot be due to the dissociation of a single 608 mixed dislocation. Further, nucleation of dislocation half-loops from the free surface beyond the critical thickness can also lead to the formation of stacking faults, however, the activation energy required to nucleate such loops is considerable and the probability of nucleation for mis®ts , 2% is quite low [8]. The implication of these observations is that lattice mismatch may not play a role in the formation of planar defects within the epitaxial islands. The presence of deposits on the {111} facet (Fig. 2) has also been observed in plan-view micrographs [7] on the shorter {111} facets of the islands which may be P-terminated and this is depicted in Fig. 3a. The observation that contiguous to the deposit the stacking sequence of the {111} facet changes from a zinc-blende to the wurtzite {0001} implies that there are kinetic constraints to growth on this facet which allow the transformation. The faint outline of the wurtzite structure in the deposit combined with the observation that these deposits are noticed only on shorter {111} facets of these islands imply different growth condi-
tions on this facet. The presence of the deposit on the Pterminated {111} facet may be attributed to a reduced surface mobility for adatoms due to the higher density of dangling bonds compared to the Ga-terminated facet. Thus, incoming adatoms are immobilized as they land and may not reach the equilibrium positions. Formation of planar defects is related to stacking errors on the {111} facets during growth. Under equilibrium conditions, atoms will tend to arrange themselves into low energy con®gurations, and the correct stacking sequence of the {111} planes will be maintained. However, at low temperatures as in Fig. 1, atoms are not suf®ciently mobile to arrange themselves in low energy positions. Thus, atoms
Fig. 3. (a) A deposit is formed on a {111}P facet and (b) formation of an intrinsic stacking fault on a {111} facet.
V. Narayanan et al. / Thin Solid Films 357 (1999) 53±56
arriving at a {111} facet may occupy incorrect sites. For the sake of discussion, assume that the {111} facet is terminated by atoms in the A position. For correct stacking the atoms in the next layer should occupy B sites. When a stacking error occurs, atoms will occupy the C sites. The position of the next layer of atoms decides the nature of the fault. If the atoms deposit on the A site and continue depositing in the correct sequence, the result will be ¼ABCA|CABC¼ and an intrinsic fault is formed as depicted in Fig. 3b. If the atoms deposit on the B site, the result will be ¼ABCA|C|BCABC¼ and an extrinsic fault results. Finally if atoms deposit on the B site and continue the deposition in an inverse sequence, the result will be ¼ABCA|CBAC¼ and a twin is formed. In addition, a higher probability of stacking errors on the shorter {111}P facet is expected because of reduced surface mobility which prevents atoms from attaining low energy con®gurations. Fig. 4 shows a GaP island grown for 80 s at 5608C that is faceted on {112} and {113} planes. In contrast to Fig. 1, these islands are free of planar defects and do not exhibit {111} facets. The absence of planar defects could be attributed to two factors. First, enhanced atomic mobility at higher temperatures allows the atoms to arrange themselves in low energy con®gurations for larger sizes of nuclei. Therefore, planar defect-free islands can be produced for larger sizes of nuclei at higher temperatures. Second, the presence of {112} and {113} facets and the absence of {111} facets on which growth mistakes could lead to stacking faults, may also prevent planar defect formation. Islands that grow on Si(111) at 5608C exhibit two different morphologies. Fig. 5 depicts a `hut' shaped island, in which wurtzite GaP with ¼ABAB¼ stacking of its basal plane coexists with zinc-blende GaP such that the c-axis of the wurtzite polytype is parallel to the (111) plane normal of the zinc-blende polytype. Fig. 6 shows an island with only the zinc-blende polytype. Since, islands take a longer time
Fig. 4. Planar defect free island grown on Si(001) after 80 s of growth at 5608C.
55
Fig. 5. A GaP island on Si(111) after 300 s of growth at 5608C showing both wurtzite (WZ) and zinc-blende (ZB) polytypes.
to nucleate on the (111) surface as compared to the (001) surface [7], a high supersaturation is needed for growth. Thus, in the `hut' shaped island, the growth environment is rich with the active species before nucleation and this ensures a high initial growth rate. Even though the equilibrium polytype for GaP is zinc-blende, a high local supersaturation may provide the necessary kinetic pathway to stabilize the wurtzite polytype. Hence, the wurtzite polytype may be preferred over the zinc-blende polytype under such conditions. As the growth rate stabilizes after the initial spurt, the zinc-blende structure is again the stable polytype. The coexistence of the wurtzite and zinc-blende polytypes for GaN has been reported by Yang et al. [9] though the polytypes were produced by changing the growth temperature. This is the ®rst time that the coexistence of the two polytypes has been observed at the same growth temperature. It is envisaged that the presence of two different types
Fig. 6. A GaP island on Si(111) after 300 s of growth at 5608C showing only the zinc-blende (ZB) polytype.
56
V. Narayanan et al. / Thin Solid Films 357 (1999) 53±56
of island morphologies on Si(111) under the same nominal growth conditions, could be a consequence of variation in the concentration of active species across the wafer. In Fig. 6 a lower supersaturation of species prevents the formation of the wurtzite polytype and low energy facets develop at an early stage.
Acknowledgements
4. Conclusions
[1] F. Ernst, P. Pirouz, J. Appl. Phys. 64 (9) (1988) 4526. [2] F. Ernst, P. Pirouz, J. Mater. Res. 4 (4) (1989) 834. [3] D. Gerthsen, D.K. Bigelsen, F.A. Ponce, J.C. Tramontana, J. Cryst. Growth 106 (1990) 157. [4] A. Vila, A. Cornet, J.R. Morante, Mater. Lett. 31 (1997) 339. [5] J.T. Kelliher, J. Thornton, N. Dietz, G. Lucovsky, K.J. Bachmann, Mater. Sci. Eng. B 22 (1993) 97. [6] S. Mahajan, Mater. Res. Soc. Symp. Proc. 410 (1996) 3. [7] N. Sukidi, K.J. Bachmann, V. Narayanan, S. Mahajan, J. Electrochem. Soc. 146 (3) (1999) 1147. [8] J.W. Mathews, A.E. Blakeslee, S. Mader, Thin Solid Films 33 (1976) 253. [9] J.W. Yang, J.N. Kuznia, Q.C. Chen, M.A. Khan, T. George, M. De Graef, S. Mahajan, Appl. Phys. Lett. 67 (19) (1995) 3759.
We have shown that defect free GaP islands on Si(001) of sizes , 20 nm can be produced by pulsed CBE at 5608C. Planar defect formation is attributed to stacking errors that occur on shorter {111} facets of epitaxial GaP islands. The high density of planar defects in smaller sized nuclei at lower temperatures is related to low atomic mobility. Wurtzite GaP has been shown to coexist with zinc-blende domains within islands grown on Si(111). This is attributed to a high species supersaturation which kinetically stabilizes the wurtzite polytype.
This work has been supported by the DOD-MURI Grant F49620-95-1-0447. References
Origins of defects in self assembled GaP islands grown on Si(001) and Si(111) V. Narayanan a,*, N. Sukidi b, K.J. Bachmann b, S. Mahajan a a
Department of Chemical, Bio and Materials Engineering, Arizona State University, Tempe AZ 85287-6006, USA Department of Materials Science and Engineering, North Carolina State University Box 7919 Raleigh, NC 27695-7919, USA
b
Abstract Microstructures of GaP epitaxial islands grown on Si(001) and Si(111) by chemical beam epitaxy have been investigated by transmission electron microscopy (TEM). Results indicate that planar-defect free GaP islands of sizes ,20 nm can be produced at 5608C on Si(001). Some of the islands are faceted on {111} and {113} planes. Subsequent planar defect formation occurs due to stacking errors on the smaller {111} facets of GaP islands that may be P-terminated. These stacking errors are attributed to the low surface mobility on P-terminated facets. A high density of planar defects is observed in smaller islands grown on Si(001) at 4208C, a consequence of reduced atomic mobility at low temperatures that leads to {111} stacking errors. Wurtzite GaP has been observed to coexist with the zinc-blende polytype in some of the islands grown on Si(111) at 5608C. q 1999 Elsevier Science S.A. All rights reserved. Keywords: GaP islands; Si substrates; Defects; High resolution transmission electron microscopy
GaP and Si have a lattice mismatch of 0.37% at room temperature. Therefore, the two materials constitute an nearly ideal combination for the integration of Si and III± V technologies. A number of previous studies have indicated that GaP nucleates on Si in the form of islands [1,2]. These islands exhibit a high density of planar defects despite the low lattice mismatch. Further, the island shape and planar defect morphology is remarkably similar to that observed in GaAs/Si(001), a system with a lattice mismatch of 4% [3,4]. This suggests that planar defect formation in GaP/Si (001) may not arise from mis®t stresses, but rather from growth mistakes on faceted GaP islands. To understand the formation of planar defects, we have investigated the initial stages of heteroepitaxy on Si(001) and Si(111).
water rinse. GaP was deposited on Si using pulsed chemical beam epitaxy (CBE) [5] wherein, the heated Si substrate was exposed to pulses of tertiarybutylphosphine (TBP) and triethylgallium (TEG) with a steady hydrogen background pressure. The duration of the source vapor cycle tsvc repeated throughout the growth run was made up of the widths of the TBP and TEG pulses and pauses in between. The temperature range investigated was 420±5608C and the overall pressure during deposition was between 10 24 and 10 25 Torr. Using a cycle time tsvc of 5 s, TEG ¯ow of 0.05 sccm was pulsed into the reactor for 300 ms per cycle under continuous TBP and hydrogen ¯ow of 0.6 sccm and 5.0 sccm, respectively [7]. Structural analysis of as-grown layers was performed in cross-section by high resolution transmission electron microscopy (HRTEM) on a JEM-4000EX microscope operating at 400 kV and with a point to point resolution of 0.17 nm.
2. Experimental details
3. Results and discussion
The Si wafers were cleaned using an RCA clean. This consisted of a 10-min dip in a 1:1:5 solution of NH4OH, H2O2 and de-ionized (DI) water maintained at 758C, a 5min rinse in DI water, a 10-min dip in a 1:1:5 solution of HCl, H2O2 and DI water also kept at 758C and a rinse in DI water. This was followed by a buffered HF etch and a ®nal DI
Fig. 1 reveals the morphology of GaP islands after 80 s of growth at 4208C. This island has an average height of 8 nm and width of 18 nm and is faceted on {111} and {113} planes. Twins are seen parallel to the {111} planes. Fig. 2 shows an HRTEM image of `deposits' on the {111} facets of GaP islands grown on Si(001) at 5608C. The accommodation of lattice mis®t beyond a critical layer thickness between GaP and Si would require the
1. Introduction
* Corresponding author.
0040-6090/99/$ - see front matter q 1999 Elsevier Science S.A. All rights reserved. PII: S 0040-609 0(99)00474-5
54
V. Narayanan et al. / Thin Solid Films 357 (1999) 53±56
Fig. 1. A GaP Island on Si(001) after 80 s of growth at 4208C.
Fig. 2. A deposit on the {111} facet of a GaP island grown on Si(001) at 5608C.
nucleation of structurally necessary mis®t dislocations at the interface. The separation (S) between these dislocations is given by [6] S aGaP aSi =2 jaGaP 2 aSi j where aGaP, aSi are the unstrained lattice parameters of GaP and Si respectively. The factor of two arises from the fact that 608 mixed dislocations of the type a/2k011l (which can glide on {111} planes inclined to the surface and dissociate into 308 and 908 partials to form stacking faults) only accommodate half the lattice mis®t of Lomer (908) edge dislocations, because the Burgers vector is inclined to the interface. The separation distance between planar defects in Fig. 1 is much smaller than the calculated distance of 72 nm. In addition, the presence of thick twins cannot be due to the dissociation of a single 608 mixed dislocation. Further, nucleation of dislocation half-loops from the free surface beyond the critical thickness can also lead to the formation of stacking faults, however, the activation energy required to nucleate such loops is considerable and the probability of nucleation for mis®ts , 2% is quite low [8]. The implication of these observations is that lattice mismatch may not play a role in the formation of planar defects within the epitaxial islands. The presence of deposits on the {111} facet (Fig. 2) has also been observed in plan-view micrographs [7] on the shorter {111} facets of the islands which may be P-terminated and this is depicted in Fig. 3a. The observation that contiguous to the deposit the stacking sequence of the {111} facet changes from a zinc-blende to the wurtzite {0001} implies that there are kinetic constraints to growth on this facet which allow the transformation. The faint outline of the wurtzite structure in the deposit combined with the observation that these deposits are noticed only on shorter {111} facets of these islands imply different growth condi-
tions on this facet. The presence of the deposit on the Pterminated {111} facet may be attributed to a reduced surface mobility for adatoms due to the higher density of dangling bonds compared to the Ga-terminated facet. Thus, incoming adatoms are immobilized as they land and may not reach the equilibrium positions. Formation of planar defects is related to stacking errors on the {111} facets during growth. Under equilibrium conditions, atoms will tend to arrange themselves into low energy con®gurations, and the correct stacking sequence of the {111} planes will be maintained. However, at low temperatures as in Fig. 1, atoms are not suf®ciently mobile to arrange themselves in low energy positions. Thus, atoms
Fig. 3. (a) A deposit is formed on a {111}P facet and (b) formation of an intrinsic stacking fault on a {111} facet.
V. Narayanan et al. / Thin Solid Films 357 (1999) 53±56
arriving at a {111} facet may occupy incorrect sites. For the sake of discussion, assume that the {111} facet is terminated by atoms in the A position. For correct stacking the atoms in the next layer should occupy B sites. When a stacking error occurs, atoms will occupy the C sites. The position of the next layer of atoms decides the nature of the fault. If the atoms deposit on the A site and continue depositing in the correct sequence, the result will be ¼ABCA|CABC¼ and an intrinsic fault is formed as depicted in Fig. 3b. If the atoms deposit on the B site, the result will be ¼ABCA|C|BCABC¼ and an extrinsic fault results. Finally if atoms deposit on the B site and continue the deposition in an inverse sequence, the result will be ¼ABCA|CBAC¼ and a twin is formed. In addition, a higher probability of stacking errors on the shorter {111}P facet is expected because of reduced surface mobility which prevents atoms from attaining low energy con®gurations. Fig. 4 shows a GaP island grown for 80 s at 5608C that is faceted on {112} and {113} planes. In contrast to Fig. 1, these islands are free of planar defects and do not exhibit {111} facets. The absence of planar defects could be attributed to two factors. First, enhanced atomic mobility at higher temperatures allows the atoms to arrange themselves in low energy con®gurations for larger sizes of nuclei. Therefore, planar defect-free islands can be produced for larger sizes of nuclei at higher temperatures. Second, the presence of {112} and {113} facets and the absence of {111} facets on which growth mistakes could lead to stacking faults, may also prevent planar defect formation. Islands that grow on Si(111) at 5608C exhibit two different morphologies. Fig. 5 depicts a `hut' shaped island, in which wurtzite GaP with ¼ABAB¼ stacking of its basal plane coexists with zinc-blende GaP such that the c-axis of the wurtzite polytype is parallel to the (111) plane normal of the zinc-blende polytype. Fig. 6 shows an island with only the zinc-blende polytype. Since, islands take a longer time
Fig. 4. Planar defect free island grown on Si(001) after 80 s of growth at 5608C.
55
Fig. 5. A GaP island on Si(111) after 300 s of growth at 5608C showing both wurtzite (WZ) and zinc-blende (ZB) polytypes.
to nucleate on the (111) surface as compared to the (001) surface [7], a high supersaturation is needed for growth. Thus, in the `hut' shaped island, the growth environment is rich with the active species before nucleation and this ensures a high initial growth rate. Even though the equilibrium polytype for GaP is zinc-blende, a high local supersaturation may provide the necessary kinetic pathway to stabilize the wurtzite polytype. Hence, the wurtzite polytype may be preferred over the zinc-blende polytype under such conditions. As the growth rate stabilizes after the initial spurt, the zinc-blende structure is again the stable polytype. The coexistence of the wurtzite and zinc-blende polytypes for GaN has been reported by Yang et al. [9] though the polytypes were produced by changing the growth temperature. This is the ®rst time that the coexistence of the two polytypes has been observed at the same growth temperature. It is envisaged that the presence of two different types
Fig. 6. A GaP island on Si(111) after 300 s of growth at 5608C showing only the zinc-blende (ZB) polytype.
56
V. Narayanan et al. / Thin Solid Films 357 (1999) 53±56
of island morphologies on Si(111) under the same nominal growth conditions, could be a consequence of variation in the concentration of active species across the wafer. In Fig. 6 a lower supersaturation of species prevents the formation of the wurtzite polytype and low energy facets develop at an early stage.
Acknowledgements
4. Conclusions
[1] F. Ernst, P. Pirouz, J. Appl. Phys. 64 (9) (1988) 4526. [2] F. Ernst, P. Pirouz, J. Mater. Res. 4 (4) (1989) 834. [3] D. Gerthsen, D.K. Bigelsen, F.A. Ponce, J.C. Tramontana, J. Cryst. Growth 106 (1990) 157. [4] A. Vila, A. Cornet, J.R. Morante, Mater. Lett. 31 (1997) 339. [5] J.T. Kelliher, J. Thornton, N. Dietz, G. Lucovsky, K.J. Bachmann, Mater. Sci. Eng. B 22 (1993) 97. [6] S. Mahajan, Mater. Res. Soc. Symp. Proc. 410 (1996) 3. [7] N. Sukidi, K.J. Bachmann, V. Narayanan, S. Mahajan, J. Electrochem. Soc. 146 (3) (1999) 1147. [8] J.W. Mathews, A.E. Blakeslee, S. Mader, Thin Solid Films 33 (1976) 253. [9] J.W. Yang, J.N. Kuznia, Q.C. Chen, M.A. Khan, T. George, M. De Graef, S. Mahajan, Appl. Phys. Lett. 67 (19) (1995) 3759.
We have shown that defect free GaP islands on Si(001) of sizes , 20 nm can be produced by pulsed CBE at 5608C. Planar defect formation is attributed to stacking errors that occur on shorter {111} facets of epitaxial GaP islands. The high density of planar defects in smaller sized nuclei at lower temperatures is related to low atomic mobility. Wurtzite GaP has been shown to coexist with zinc-blende domains within islands grown on Si(111). This is attributed to a high species supersaturation which kinetically stabilizes the wurtzite polytype.
This work has been supported by the DOD-MURI Grant F49620-95-1-0447. References