- Email: [email protected]
Reflection electron microscopy studies of GaP(110) surfaces in UHV-TEM
Keflection electron microscopy studies of CL?P(110) surfaces in UHV-TEM
Received 1 September
f992; accepted for pub~~eation 7 f)ct~ber 1992
Reflection electron microscopy (REM) and reflection high energy electron diffraction (RHEEDI were used in a WWV electran microscope to study the effects of various surface treatments on the topography and crystal structure of cleaved GaP(l10) surhces. Reported ion-milling and annealing cleaning treatments resulted in rough surface topography, and in good 1 x 1 RHEED patterns. Annealing at 800°C resulted in surface smoothing, but also led to dissociation of GaP with viscous flow of a Ga-rich molten phase on the surface. Regions on the surface which were not covered by the molten phase maintained the 1 x 1 reconstruction typical of the clean GaP(110) surface.
trahiglr Cleaved GaPfllOj s?lrfaces are of interest because of their recent use in nitridation studies 111 and for growth of transition-metal Schottky barriers [Xl. Clean surfaces have been produced by in-situ cleavage or by ion-milling and annealing of ex-situ cleaved or polished surfaces [l-5]. Low energy electron diffraction studies have found a 1 X 1 reconstruction on the clean C110) surface [3,4j, However, no imaging studies have been reported ~~ara~ter~~~ng the to~~~ap~~ of the re~o~st~~ted surface. A~~ea~~~g tempe~tures for c~rn~o~~d semiconductors are usuaIIy limited by preferential sublimation of one of the components, In the case of Gap, annealing at 550°C has been used for cleaning in UHV [3,5], but in reported growth experiments of GaP epitaxial layers, or on GaP substrates, temperatures up to 900°C have been used t&71. In this work, reflection electron microscopy CREMI and reflection high energy electron djffra~t~on (RHEEDf, performed in an ul-
I
Also at Department of Physics, Arizona State Un&x&y, Tempe, AZ X5287, USA.
003%602g/93/$06.00
vacuum
transmj~jonelectrori
micro-
scope, were used to compare the to~gra~~y and structure of the GaP0fOI surface resulting from these various treatments.
2. Experimentwl conditions Samples for REM were prepared from GaP(001) wafers by cleavage in air immediatety before insertion in the microscope. The studied (110) surfaces had typical dimensions of 3 mm along the fzia]erystaf d&e&ion and fCfL-200 jm akmg the [Wlf direction. Etectron transparent samples were prepared for additional investigation of some of the high temperature annealing processes observed in REM. These samples were thinned to perforation using a brominated methanol etch following standard mechanical dimpling and polishing of 3 mm disks cut from (001) wafers, The TEM samples were rinsed in TCE, acetone and methanol before loading in the microscope. The Fhifips 43OST microscope, modified for UHV o~r~t~on and in-situ surface treatment by Gatan @], was used at INI keV gor most of this study. After bake-out, the base pressure in the
0 1993 - Elsevier Science Publishers B.V. All rights reserved
M. Gajdardziska-Josifovska et al. / REMof GaP(ll0) in UHV-TEM
UHV-section of the microscope was typically 2-5 x 10v9 Torr. The REM samples were cleaned by in-situ ion-milling using 1 keV Ar ions incident at 20” to the (110) surface with Ar partial pressure of 2 x lop6 Torr. A single-tilt specimen holder with Ta furnace was used to anneal the surfaces within the objective lens pole-piece region. The base vacuum in this region was improved by liquid-nitrogen-cooled cryofingers: the vacuum between the cryoblades cannot be measured directly but was sufficient to reproducibly observe the Si(1107 x 7 surface reconstruction [9]. Ex-situ analytical microscopy was performed on TEM samples in a Philips 4OOST-FEG TEM operated at 100 keV.
3. Experimental results Fig. 1 shows a REM image (a) and a RHEED pattern (b) of the cleaved surface which remained
1063
unchanged following bake-out of the microscope. The surface had long flat terraces with steps running along the [ liO] direction and the RHEED pattern indicated a bulk terminated surface with some contamination. It should be noted that the topography of the cleaved surface depends on the direction in which the force had been applied during cleavage. For example, REM images from GaP(110) surfaces which had been obtained by cleavage of a GaP(1111 wafer [lo] show a wider distribution of step heights compared to the case in this study when (100) wafers were used. After 20 min ion-milling to remove the contamination layer [3], the RHEED pattern was completely diffuse indicating that the surface had become fully disordered. Annealing at 550°C for 4 h was used to re-establish the surface crystallinity, with the onset of crystallization visible in the RHEED patterns at 325°C. The 1 X 1 surface reconstruction observed previously in LEED, was present on the hot surface and it remained when
Fig. 1. REM images and RHEED patterns from as-cleaved GaP(110) surface ((a) and (b)) and surface cleaned by ion-milling and annealing at 550°C ((c) and cd)).
1064
M. Gajdardziska-Josiforska et al. / REM of GaP(110) in UHV-TEM
Fig. 2. (a) REM image of ion-milled GaP(110) surface after annealing at 780°C (arrows indicate regions covered with molten phase) with RHEED pattern inset; (b) BF image of same sample, tilted edge-on, showing protruding amorphous structures.
the sample was cooled to room temperature, as denoted by the rods in the RHEED patterns in fig. Id. The REM image (fig. lc), however, showed that the clean surface had much smaller terraces and was, in general, rougher than the as-cleaved surface. Cycles of ion milling and annealing did not improve the flatness of the surface, nor did prolonged annealing (up to 20 h) or lower annealing temperatures (450°C). It should be noted that no changes could be induced on the cleaved
surfaces by heating to 550°C without prior ionmilling, suggesting that the surface contamination layer inhibits the mobility of the GaP surface atoms. However, heating non-milled samples to 800°C for N 1 min, resulted in a 1 x 1 reconstruction and a surface topography similar to that in fig. lc. For both as-cleaved and ion-milled surfaces, prolonged annealing at temperatures close to 800°C (77o”C-810°C) caused an increase of the
Fig. 3. Flow of viscous molten phase is illustrated by three REM images from a video sequence of annealing of ion-milled GaPt 110) surface at 800°C.
terrace sizes (fig. 2aj while preserving the 1 X I re~onst~ction on these terraces (fig. Za, inset). fn #mpetitian with this smoothing was a process of dissociation of the GaP with fo~ati~n of two distinct phases. An amorphous phase was observed most readiIy in association with GaP chips which normaly stick to the surface during cleavage. Fuzzy strings can be discerned in the REM image in fig. Za, but these are more clearly seen in the bright field TEM image in fig. 2b obtained from the annealed REM sample after orienting the 1210) surface parallel to the incident electron beam. The second phase observed at N 800°C was a viscous liquid that resulted in the covered regions which appear dark in fig. 2a, The formation of the liquid phase was observed to be faster for the
0
2.40
4.80
7.20
EI?ERGY [KV]
9.60
12.00
GaP chips relative to the flat surface areas. Elec-
&on-beam irradiation of these particles further speeded the formation of the liquid phase presumably due to additional IocaI heating provided by the beam. The larger volume of liquid originating from these particIes was observed to flow slowly and randomly on the surface. Examples were observed when the melted phase would pass over a region on the surface without changing its topography, and then return to the same region after tens of seconds. Such a case is illustrated in fig. 3 which shows three frames from a video recording of this flow process. For flatter parts on the surface the coverage was observed to be thinner and to originate from surface steps. After heating sessions of close to 1 h, the samples wouid typically be completely covered with the
100
120
140
160
180
200
ENERGY LOSS [e.Vf
Fig. 4. Images and spectra from t~nsmission ~a~~~~ sample annealed at 800°C: (a) BF image &owing initial stages of dissociation with facetting and amorphous residue; fb) DF image after prolonged heating showing part&s; {c) X-ray spectra from GaP crystal f- - - - - -1 and typical particle f -k fd> EELS spectrmn of amorphous residue.
1066
M. Gaj~r~z~ka-Jos~ou~ka
et al. / REM of GaPllfOl in U~-~EM
liquid phase and after cooling to room temperature gave no distinct RHEED patterns, nor REM images. Under an optical microscope the surface of these samples looked metallic as opposed to the orange translucent color of the as-cleaved surfaces. A TEM sample was used to further study the dissociation process. As in the reflection case without ion-milling, the changes of the TEM specimen started at close to 800°C. The initial stages of dissociation are recorded in the bright field micrograph in fig. 4a. The faint rounded edges of amorphous material denote the original edges of the GaP crystal before heating. At the elevated temperature, crystal modification started by facetting of the original curved surface into (110) surfaces (arrowed in fig. 4a) leaving behind the amorphous residue. After prolonged heating, the edges of the specimen have moved by microns, thin areas of the specimen have disappeared, and dark particles along with amorphous films were visible over the surface or at edges of the specimen. Fig. 4b shows a dark field image recorded with a diffraction spot from one of the particles after the sample has been cooled to room temperature. These particles were crystalline but were not in epitaxial relationship with the GaP crystal. The annealed transmission sample was used for qualitative analytical study of the composition of the crystalline particles and the amorphous layers. Fig. 4c shows X-ray spectra recorded from the GaP crystal (broken line) and from one of the larger crystalline particles, demonstrating that these particles are predominantly composed of Ga. The amorphous regions were susceptible to radiation damage at the beam conditions needed for X-ray studies, but electron-energy loss spectra from these regions show that they are phosphorus rich (fig. 4d).
4. Conclusions The 1 X 1 surface reconst~ction on GaP(110) surfaces was confirmed by RHEED studies in a UHV-TEM. The topography of the surfaces cleaned by ion-milling and annealing at 550°C
was studied by REM and was found to be rough compared to as-cleaved surfaces. No modifications in topography or crystal structure could be induced by annealing at 550°C without prior ionmilling, but flash heating to 800°C did produce the 1 X 1 reconstruction accompanied by rough topography similar to that of the ion-milled case. Annealing at close to 800°C resulted in dissociation of GaP which was faster in thin crystals compared to (110) surface of bulk samples. Electron-beam-enhanced melting and flow of viscous liquid were seen at these temperatures by REM. Amorphous residue was observed in TEM. Upon cooling to room temperature the thin amorphous residue was found by EELS to be P-rich, while the non-epitaxial crystalline particles were found to be Ga-rich.
Ackuowledgemen t This work is based upon research conducted at the Center for High Resolution Electron Microscopy, which is supported by the National Science Foundation under grant DMR89- 13384.
References [l] H.I. Starnberg, P. Soukiassian and T. Kendelewicz, Surf. Sci. 269/270 (1992) 915. [2] D.A. Evans, T.P. Chen, Th. ChassC, K. Horn, M. von der Emde and D.R.T. Zahn, Surf. Sci. 269/270 (1992) 979. [3] C.B. Duke, A. Paton, W.K. Ford, A. Kahn and J. Carelli. Phys. Rev. B 24 (1981) 562. [4] C.B. Duke, A. Paton and A. Kahn, J. Vat. Sci. Technol. A 2 (1984) 515. [S] J. Kanasaki, H. Yamashita, A. Okano, K. Hattori, Y. Nakai, N. Itoh and R.F. Haglund. Jr., Surf. Sci. Lett. 257 (1991) L642. [6] M. Umeno, T. Jimbo and T. Soga, J. Cryst. Growth 98 (1989) 188. [7] M.R. Leys, M.-E. Pistol, H. Titze and L. Samuelson, J. Electron. Mater. 18 (1989) 25. [S] D.J. Smith, J. Podbrdsky, P.R. Swann and J.S. Jones, Mater. Res. Sot. Symp. Proc. 139 (1989) 289. [9] D.J. Smith, M. Gajdard~iska-Josifovska, P. Lu, M.R. McCartney, J. Podbrdsky, P.R. Swann and J.J. Jones, Ultramicroscopy 49 (1993126. [lo] N. Yamamoto and J.C.H. Spence, Thin Solid Films 104 (1983) 43.
Received 1 September
f992; accepted for pub~~eation 7 f)ct~ber 1992
Reflection electron microscopy (REM) and reflection high energy electron diffraction (RHEEDI were used in a WWV electran microscope to study the effects of various surface treatments on the topography and crystal structure of cleaved GaP(l10) surhces. Reported ion-milling and annealing cleaning treatments resulted in rough surface topography, and in good 1 x 1 RHEED patterns. Annealing at 800°C resulted in surface smoothing, but also led to dissociation of GaP with viscous flow of a Ga-rich molten phase on the surface. Regions on the surface which were not covered by the molten phase maintained the 1 x 1 reconstruction typical of the clean GaP(110) surface.
trahiglr Cleaved GaPfllOj s?lrfaces are of interest because of their recent use in nitridation studies 111 and for growth of transition-metal Schottky barriers [Xl. Clean surfaces have been produced by in-situ cleavage or by ion-milling and annealing of ex-situ cleaved or polished surfaces [l-5]. Low energy electron diffraction studies have found a 1 X 1 reconstruction on the clean C110) surface [3,4j, However, no imaging studies have been reported ~~ara~ter~~~ng the to~~~ap~~ of the re~o~st~~ted surface. A~~ea~~~g tempe~tures for c~rn~o~~d semiconductors are usuaIIy limited by preferential sublimation of one of the components, In the case of Gap, annealing at 550°C has been used for cleaning in UHV [3,5], but in reported growth experiments of GaP epitaxial layers, or on GaP substrates, temperatures up to 900°C have been used t&71. In this work, reflection electron microscopy CREMI and reflection high energy electron djffra~t~on (RHEEDf, performed in an ul-
I
Also at Department of Physics, Arizona State Un&x&y, Tempe, AZ X5287, USA.
003%602g/93/$06.00
vacuum
transmj~jonelectrori
micro-
scope, were used to compare the to~gra~~y and structure of the GaP0fOI surface resulting from these various treatments.
2. Experimentwl conditions Samples for REM were prepared from GaP(001) wafers by cleavage in air immediatety before insertion in the microscope. The studied (110) surfaces had typical dimensions of 3 mm along the fzia]erystaf d&e&ion and fCfL-200 jm akmg the [Wlf direction. Etectron transparent samples were prepared for additional investigation of some of the high temperature annealing processes observed in REM. These samples were thinned to perforation using a brominated methanol etch following standard mechanical dimpling and polishing of 3 mm disks cut from (001) wafers, The TEM samples were rinsed in TCE, acetone and methanol before loading in the microscope. The Fhifips 43OST microscope, modified for UHV o~r~t~on and in-situ surface treatment by Gatan @], was used at INI keV gor most of this study. After bake-out, the base pressure in the
0 1993 - Elsevier Science Publishers B.V. All rights reserved
M. Gajdardziska-Josifovska et al. / REMof GaP(ll0) in UHV-TEM
UHV-section of the microscope was typically 2-5 x 10v9 Torr. The REM samples were cleaned by in-situ ion-milling using 1 keV Ar ions incident at 20” to the (110) surface with Ar partial pressure of 2 x lop6 Torr. A single-tilt specimen holder with Ta furnace was used to anneal the surfaces within the objective lens pole-piece region. The base vacuum in this region was improved by liquid-nitrogen-cooled cryofingers: the vacuum between the cryoblades cannot be measured directly but was sufficient to reproducibly observe the Si(1107 x 7 surface reconstruction [9]. Ex-situ analytical microscopy was performed on TEM samples in a Philips 4OOST-FEG TEM operated at 100 keV.
3. Experimental results Fig. 1 shows a REM image (a) and a RHEED pattern (b) of the cleaved surface which remained
1063
unchanged following bake-out of the microscope. The surface had long flat terraces with steps running along the [ liO] direction and the RHEED pattern indicated a bulk terminated surface with some contamination. It should be noted that the topography of the cleaved surface depends on the direction in which the force had been applied during cleavage. For example, REM images from GaP(110) surfaces which had been obtained by cleavage of a GaP(1111 wafer [lo] show a wider distribution of step heights compared to the case in this study when (100) wafers were used. After 20 min ion-milling to remove the contamination layer [3], the RHEED pattern was completely diffuse indicating that the surface had become fully disordered. Annealing at 550°C for 4 h was used to re-establish the surface crystallinity, with the onset of crystallization visible in the RHEED patterns at 325°C. The 1 X 1 surface reconstruction observed previously in LEED, was present on the hot surface and it remained when
Fig. 1. REM images and RHEED patterns from as-cleaved GaP(110) surface ((a) and (b)) and surface cleaned by ion-milling and annealing at 550°C ((c) and cd)).
1064
M. Gajdardziska-Josiforska et al. / REM of GaP(110) in UHV-TEM
Fig. 2. (a) REM image of ion-milled GaP(110) surface after annealing at 780°C (arrows indicate regions covered with molten phase) with RHEED pattern inset; (b) BF image of same sample, tilted edge-on, showing protruding amorphous structures.
the sample was cooled to room temperature, as denoted by the rods in the RHEED patterns in fig. Id. The REM image (fig. lc), however, showed that the clean surface had much smaller terraces and was, in general, rougher than the as-cleaved surface. Cycles of ion milling and annealing did not improve the flatness of the surface, nor did prolonged annealing (up to 20 h) or lower annealing temperatures (450°C). It should be noted that no changes could be induced on the cleaved
surfaces by heating to 550°C without prior ionmilling, suggesting that the surface contamination layer inhibits the mobility of the GaP surface atoms. However, heating non-milled samples to 800°C for N 1 min, resulted in a 1 x 1 reconstruction and a surface topography similar to that in fig. lc. For both as-cleaved and ion-milled surfaces, prolonged annealing at temperatures close to 800°C (77o”C-810°C) caused an increase of the
Fig. 3. Flow of viscous molten phase is illustrated by three REM images from a video sequence of annealing of ion-milled GaPt 110) surface at 800°C.
terrace sizes (fig. 2aj while preserving the 1 X I re~onst~ction on these terraces (fig. Za, inset). fn #mpetitian with this smoothing was a process of dissociation of the GaP with fo~ati~n of two distinct phases. An amorphous phase was observed most readiIy in association with GaP chips which normaly stick to the surface during cleavage. Fuzzy strings can be discerned in the REM image in fig. Za, but these are more clearly seen in the bright field TEM image in fig. 2b obtained from the annealed REM sample after orienting the 1210) surface parallel to the incident electron beam. The second phase observed at N 800°C was a viscous liquid that resulted in the covered regions which appear dark in fig. 2a, The formation of the liquid phase was observed to be faster for the
0
2.40
4.80
7.20
EI?ERGY [KV]
9.60
12.00
GaP chips relative to the flat surface areas. Elec-
&on-beam irradiation of these particles further speeded the formation of the liquid phase presumably due to additional IocaI heating provided by the beam. The larger volume of liquid originating from these particIes was observed to flow slowly and randomly on the surface. Examples were observed when the melted phase would pass over a region on the surface without changing its topography, and then return to the same region after tens of seconds. Such a case is illustrated in fig. 3 which shows three frames from a video recording of this flow process. For flatter parts on the surface the coverage was observed to be thinner and to originate from surface steps. After heating sessions of close to 1 h, the samples wouid typically be completely covered with the
100
120
140
160
180
200
ENERGY LOSS [e.Vf
Fig. 4. Images and spectra from t~nsmission ~a~~~~ sample annealed at 800°C: (a) BF image &owing initial stages of dissociation with facetting and amorphous residue; fb) DF image after prolonged heating showing part&s; {c) X-ray spectra from GaP crystal f- - - - - -1 and typical particle f -k fd> EELS spectrmn of amorphous residue.
1066
M. Gaj~r~z~ka-Jos~ou~ka
et al. / REM of GaPllfOl in U~-~EM
liquid phase and after cooling to room temperature gave no distinct RHEED patterns, nor REM images. Under an optical microscope the surface of these samples looked metallic as opposed to the orange translucent color of the as-cleaved surfaces. A TEM sample was used to further study the dissociation process. As in the reflection case without ion-milling, the changes of the TEM specimen started at close to 800°C. The initial stages of dissociation are recorded in the bright field micrograph in fig. 4a. The faint rounded edges of amorphous material denote the original edges of the GaP crystal before heating. At the elevated temperature, crystal modification started by facetting of the original curved surface into (110) surfaces (arrowed in fig. 4a) leaving behind the amorphous residue. After prolonged heating, the edges of the specimen have moved by microns, thin areas of the specimen have disappeared, and dark particles along with amorphous films were visible over the surface or at edges of the specimen. Fig. 4b shows a dark field image recorded with a diffraction spot from one of the particles after the sample has been cooled to room temperature. These particles were crystalline but were not in epitaxial relationship with the GaP crystal. The annealed transmission sample was used for qualitative analytical study of the composition of the crystalline particles and the amorphous layers. Fig. 4c shows X-ray spectra recorded from the GaP crystal (broken line) and from one of the larger crystalline particles, demonstrating that these particles are predominantly composed of Ga. The amorphous regions were susceptible to radiation damage at the beam conditions needed for X-ray studies, but electron-energy loss spectra from these regions show that they are phosphorus rich (fig. 4d).
4. Conclusions The 1 X 1 surface reconst~ction on GaP(110) surfaces was confirmed by RHEED studies in a UHV-TEM. The topography of the surfaces cleaned by ion-milling and annealing at 550°C
was studied by REM and was found to be rough compared to as-cleaved surfaces. No modifications in topography or crystal structure could be induced by annealing at 550°C without prior ionmilling, but flash heating to 800°C did produce the 1 X 1 reconstruction accompanied by rough topography similar to that of the ion-milled case. Annealing at close to 800°C resulted in dissociation of GaP which was faster in thin crystals compared to (110) surface of bulk samples. Electron-beam-enhanced melting and flow of viscous liquid were seen at these temperatures by REM. Amorphous residue was observed in TEM. Upon cooling to room temperature the thin amorphous residue was found by EELS to be P-rich, while the non-epitaxial crystalline particles were found to be Ga-rich.
Ackuowledgemen t This work is based upon research conducted at the Center for High Resolution Electron Microscopy, which is supported by the National Science Foundation under grant DMR89- 13384.
References [l] H.I. Starnberg, P. Soukiassian and T. Kendelewicz, Surf. Sci. 269/270 (1992) 915. [2] D.A. Evans, T.P. Chen, Th. ChassC, K. Horn, M. von der Emde and D.R.T. Zahn, Surf. Sci. 269/270 (1992) 979. [3] C.B. Duke, A. Paton, W.K. Ford, A. Kahn and J. Carelli. Phys. Rev. B 24 (1981) 562. [4] C.B. Duke, A. Paton and A. Kahn, J. Vat. Sci. Technol. A 2 (1984) 515. [S] J. Kanasaki, H. Yamashita, A. Okano, K. Hattori, Y. Nakai, N. Itoh and R.F. Haglund. Jr., Surf. Sci. Lett. 257 (1991) L642. [6] M. Umeno, T. Jimbo and T. Soga, J. Cryst. Growth 98 (1989) 188. [7] M.R. Leys, M.-E. Pistol, H. Titze and L. Samuelson, J. Electron. Mater. 18 (1989) 25. [S] D.J. Smith, J. Podbrdsky, P.R. Swann and J.S. Jones, Mater. Res. Sot. Symp. Proc. 139 (1989) 289. [9] D.J. Smith, M. Gajdard~iska-Josifovska, P. Lu, M.R. McCartney, J. Podbrdsky, P.R. Swann and J.J. Jones, Ultramicroscopy 49 (1993126. [lo] N. Yamamoto and J.C.H. Spence, Thin Solid Films 104 (1983) 43.