- Email: [email protected]
Lattice imaging of silicide-silicon interfaces
Thin Solid Films, 93 (1982) 91-97
91
LATTICE I M A G I N G O F S I L I C I D E - S I L I C O N I N T E R F A C E S * L. J. CHEN t AND J. W. MAYER Department o f Materials Science and Engineering, Cornell University, Ithaca, N Y 14853 (U.S.A.) K.N. TU I B M T. J. Watson Research Center, Yorktown Heights, NY10598 (U.S.A.) T. T. SHENG Bell Laboratories, Murray Hill, NJ07974 (U.S.A.)
The interfaces of both epitaxial and non-epitaxial silicides and silicon were investigated by the direct lattice imaging method using cross-sectional samples. Non-epitaxial CoSi 2 on silicon was observed to have a curved interface. Epitaxial CoSi 2, however, was found to be smooth within a facet. No evidence of an amorphous layer at the interface was obtained. Epitaxial NiSi2 on Si(001) was found to be heavily faceted. The facets are on {111} and {100) planes with the former more frequently observed. The interface between Si(lll) and NiSi 2 is also faceted but less so than that for Si(001). The interface is very rough on a large scale. Straight boundary lines corresponding to faceted planes were observed which indicated that the interfaces on an atomic scale were quite smooth. Defect clusters and planar defects were also observed at the interfaces.
1. INTRODUCTION Thin films of silicides have found widespread use in integrated circuits as Schottky barriers, ohmic contacts, low resistivity gates and interconnects. Most silicide studies have been concentrated on electrical measurements, reaction kinetics and the identification of phases formed during reaction 1. A great deal of information has been obtained on electrical properties, activation energies of silicide growth, the sequence of phase formation and the predominant diffusing species during the reaction, yet much less is known about the structural and morphological aspects of silicide-silicon interfaces. The interfaces control many technically important properties of silicide contacts and interconnects such as Schottky barrier height, contact resistance and corrosion stability; consequently it is important to advance our knowledge of these interfaces. Moreover, there is a considerable basic interest in the atomic structure of the silicide-silicon interface, e.g. the existence of a thin * Paper presented at the Symposium on Thin Films and Interfaces, Boston, MA, U.S.A., November 16-19, 1981. "~Permanent address: Department of Materials Science and Engineering, National Tsing Hua University, Hsinchu, Taiwan. 0040-6090/82/0000-0000/$02~75
© Elsevier Sequoia/Printed in The Netherlands
92
L.J. C ~
et al.
amorphous layer at the interface has been proposed in order to predict the first silicide phase in transition metal-silicon systems2. F611 et al. have studied the epitaxial interfaces of Pd2Si/Si, NiSi2/Si as well as NiSi/Si by transmission electron microscopy (TEM) using the direct lattice imaging method on cross-sectional samples 3'4. The smoothness of the interfaces was revealed. No evidence of an amorphous interfacial layer was obtained. The resistivity of the silicides is the single most important criterion in considering them for the metallization in integrated circuits. Of all the silicides, CoSi2 has one of the lowest resistivities (18-20 ~ cm) s. Similar to NiSi2, CoSi2 has a cubic CaF2 structure with lattice constant ao = 5.37 ~ and the lattice mismatch with diamond cubic silicon is only about 1.1%. Both CoSi2 and NiSi2 are known to form epitaxially on low index planes of silicon6'7. A sketch for comparison of the CaF2 and silicon structures is shown in Fig. 1. It is straightforward to extend lattice imaging studies of the interfaces of other silicides to include epitaxial CoSi2. In addition, since a great majority of the silicides do not possess an epitaxial relationship with respect to the silicon matrix, it is also desirable to study the interfaces between non-epitaxial silicides and silicon. In this paper, we repoit the results of the study of both epitaxial and non-epitaxial interfaces of CoSia and silicon. For comparison, we have also studied many cross-sectional samples of epitaxial NiSi2 on silicon.
(a) (b) Fig. 1. Comparison of(a) the cubic GaF 2 structure (©, fluorine; O, calcium) and (b) the diamond cubic silicon (©) structure. 2. EXPERIMENTAL PROCEDURES
Nickel and cobalt thin films about 300-400 ,~ in thickness were deposited at room temperature onto (001)- and (111)-oriented silicon wafers by electron gun evaporation with a base vacuum better than 10- 7 Torr. The disilicides were formed by annealing treatments in a helium atmosphere or in a vacuum better than 10- ~ Torr. Conventional TEM specimens were prepared by chemical thinning from the silicon side; the deposited side was covered with an electronic wax for protection during chemical polishinga. Cross-sectional samples were prepared using the procedures outlined by Sheng and Chang a. The electron microscopy was performed in a JEOL 200 CX microscope operated at 200 kV or in a Siemens 102 microscope operated at 120 kV. 3. RESULTS AND DISCUSSIONS 3.1. Cobalt silicides
Polycrystalline CoSi 2 was formed by reacting cobalt thin films on silicon in a
LATTICE IMAGING OF INTERFACES
93
helium atmosphere at 850 °C for up to 4 h. The grains were found to be 0.1-0.5 ~tm in size. Figure 2 shows a cross-sectional view of a sample with polycrystalline CoSiz on Si(001). The interfaces between CoSiz grains and the silicon substrate were generally curved. For a CoSiz layer about 1500/~ thick, the estimated roughness was about 300 4. For samples annealed at 950 °C for 1 h, only epitaxial CoSi z was observed.
!6
Fig. 2. Bright field micrograph of a cross-sectional sample with polycrystalline CoSi 2 on silicon. Fig. 3. Lattice imaging of a cross-sectional sample with a CoSi 2 grain on silicon, rill [112] for silicon.
Figure 3 is an example of high resolution lattice imaging of a cross-sectional sample oriented along the [112] direction of the silicon substrate. Only (111) lattice fringes of silicon with an interplanar spacing of 3.14 ~ were observed; (220) fringes with a spacing of 1.92 ~ were too narrow to be resolved with the electron microscope utilized. No fringes were observed in the CoSi 2 grain and there were no extra diffraction spots corresponding to CoSi2 visible in the diffraction pattern. However, whether the absence of fringes in the CoSi2 grain was due to the effects of thickness or to the orientation of the CoSi2 grain was not determined. The interface region appeared to be brighter than other regions in the micrograph. The bright contrast could arise from the stress induced during deposition or during thermally induced reaction near the interface. One of the difficulties encountered in the high resolution imaging of cross-sectional samples, particularly when the silicide layer does not exhibit an epitaxial relationship with respect to the substrate silicon, is that lattice fringes may not be observed at the same time. A difference in ion milling rates between CoSi 2 and silicon during sample preparation may produce a thin foil with a substantial difference in thickness between the CoSi 2 and the silicon. The sample preparation may also produce artifacts, e.g. stress near the interface was relieved during the ion milling, which may obscure the true identity of the interface structure. For non-epitaxial silicides, the silicide layer usually did not have the "correct" orientation with respect to the substrate silicon so that when the sample was oriented with respect to silicon to reveal the lattice fringes of silicon, silicide fringes
94
L.J. CHENe t al.
were usually not visible. Only occasionally does a CoSi 2 grain possess just the "correct" orientation, so that lattice fringes in both the CoSi2 grain and the silicon matrix can be observed simultaneously. Such an example is shown in Fig. 4. The spacing of the CoSi2(111) fringes is about 1~o less than that of the Si(111) fringes. The fringes were always observed to extend right to the interface of the silicide and the matrix silicon so that no amorphous layer was evident at the interface.
Fig. 4. Exampleofa CoSi2 grain havingthe "correct"orientationso that latticefringesin both the C o S i 2 and the siliconmatrix can be observed. Fig. 5. High resolutionlatticeimage showingthe precursor ofepitaxialgrowth of CoSi2on silicon. For silicides having an epitaxial relationship to the silicon matrix, the stress around the interface may make the diffraction conditions for the epitaxial silicide and the silicon matrix different. As a consequence, although the so-called axial diffraction condition 1° (high symmetry condition) can be easily achieved in the thicker parts of the silicon by using Kikuchi lines, in interface regions thin enough for lattice imaging, the lattice fringes, if obtained at all, are no longer structural images. The interpretation of the lattice images is therefore more complicated. Nevertheless, the lattice images still provide valuable information not obtainable otherwise. Some of the CoSi2 grains exhibited an epitaxial relationship with respect to the silicon matrix as indicated by the presence of a regular dislocation network. Figure 5 shows an example of a sample oriented along the [001] direction of an Si(lll) matrix. The angle between CoSi2(111) fringes and Si(111) planes was measured to be about 70 °. No amorphous layer was observed at the interface. It is of particular interest that dark and bright bands, about 30/~ in thickness, were observed at the interface. Slight tilting of the sample away from the axial diffraction condition did not eliminate "band" contrast. The contrast may arise from the stress induced at the interface during the silieide formation. They were not thickness fringes, since the thickness fringes are of the order of hundreds of ~ngstr6ms under the diffraction condition for imaging. Close examination of the interface revealed that the first dark band exhibited a twinning relationship with respect to the silicon matrix. The occurrence of twinning at the interface between NiSi2 and silicon was reported by F611 et al. 3 and was also found by Chen e t al. Ix for CoSi2-Si interfaces. It is reasonable to conjecture that what we observed heretofore was the precursor of the epitaxial growth ofCoSi 2 on silicon, since the top layers ofCoSi 2 were non-epitaxial.
LATTICE IMAGINGOF INTERFACES
95
It is to be noted that the interface is relatively smooth as revealed by the straight boundary line. The straight boundary line was found to extend to more than 0.5 larn in length.
3.2. Nickel silicides NiSi2 was formed by annealing nickel thin films, 300-400/~ in thickness, at 800 °C for 0.5-1 h in a helium atmosphere. The silicides were found to be completely epitaxial with respect to the silicon matrix as was indicated by the existence of interface dislocation networks and by the analysis of diffraction patterns. Figure 6 shows a bright field fiat-on view of the sample illustrating the configuration of the dislocation network. The dislocations were identified to be of edge type with ~(112) Burgers' vectors. The average spacing between dislocations was about 1000/Yt which is close to the theoretically expected value of 970 ~ taking the lattice constant of NiSi2 to be a 0 = 5.406 A. Dislocation networks with 1(110) edge-type Burgers' vectors were also observed. Figure 7 shows a low magnification cross-sectional view of NiSi 2 on Si(001).
Fig. 6. Bright field micrograph of a flat-on sample showing the configuration of the dislocation network at the interfaceof NiSi2 and silicon. Fig. 7. Low magnificationcross-sectionalviewof NiSi2 on Si(001)showingthat the interfacesare heavily faceted.
Fig. 8. Lattice imaging of NiSi2 on Si(001).Defect clusters and planar defectsare indicated by arrows.
96
L.J. CHENe t al.
The interfaces are heavily faceted. The facets are on {111} and {100} planes with the former more frequently observed. The interface is very rough on a large scale. The interface between Si(111) and NiSiz is also faceted, but less so than for NiSi z on Si(001). Figure 8 shows a direct lattice imaging micrograph of NiSi 2 on Si(001). Flat boundary lines corresponding to facets were observed. The (111) fringes show an offset crossing the (100) interfaces similar to that observed by F611 e t al. a Detailed examinations showed the existence of irregularities at the interfaces. Defect clusters and extra planes parallel to the interface were observed at the interface as marked by circles A and B respectively. Similar defects were also found in many other examples. Although the possibility exists that the contrast was due to an artifact of the specimen preparation, it is also possible that they were genuine defective structures nucleated at the impurity sites of the interfaces. 4. CONCLUSIONS TEM studies, particularly high resolution lattice imaging of cross-sectional samples, provide valuable information on the structural and morphological aspects of silicide-silicon interfaces which is not obtainable otherwise. Non-epitaxial CoSi2 on silicon was shown to have curved interfaces with a roughness of 300 ~ for silicides about 1500/~ thick. Lattice fringes of CoSi2 were found to extend right to the interface so that there was no evidence of an amorphous layer at the interface. The precursor of epitaxial CoSi2 on silicon was also observed. The interfaces, up to 0.5 lain in length, appear to be very smooth. A twinning relationship between CoSi2 and silicon was also obtained. For epitaxial NiSi 2 on silicon, the interface dislocations were analyzed to be of edge type with ½(110) and-~(112) Burgers' vectors. The spacing of about 1000/~ between dislocations corresponds very well to the theoretically expected value 970 ~. The interfaces were found to be heavily faceted. The facets are on {111} and {100} planes with the former more frequently observed. The interface between Si(111) and NiSi 2 is also faceted, but less so than that for Si(001). The interface is very rough on a large scale. Straight boundary lines corresponding to facets were generally observed. Defect clusters and planar defects, possibly induced by the presence of impurities at the interface, were also detected. ACKNOWLEDGMENTS
The authors wish to thank L. S. Hung for technical assistance and H. C. Cheng and C. Y. Hou for help in the preparation of the manuscript. This work was supported in part by the National Science Foundation through the Materials Science Center at Cornell University. REFERENCES 1 K . N . Tu and J. W. Mayer, in J. M. Poate, K. N. Tu and J. W. Mayer (eds.), Thin Films-Interdiffusion and Reactions, Wiley, New York, 1978, p. 359. 2 R . M . Walser and R. W. Bene, Appl. Phys. Lett., 28 (1976) 624.
LATTICE IMAGING OF INTERFACES
3 4 5 6 7 8 9 10 11
97
H. F611, P. S. Ho and K. N. Tu, J. Appl. Phys., 52 (1981) 250. H. F611, P.S. H o a n d K . N. Tu, Philos. Mag. A,45(1982)31. S.P. Murarka, J. Vac. Sci. Technol., 17 (1980) 775. H. Ishiwara, in J, E. E. Baglin and J. M. Poate (eds.), Proc. Symp. on Thin Film Interfaces and Interactions, Los Angeles, CA, 1979, Electrochemical Society, Princeton, NJ, 1980, p. 159. J.C. Bean and J. M. Poate, Appl. Phys. Lett., 37 (1980) 643. L.J. Chert, L. S. Hung and J. W. Mayer, Appl. Surf. Sci., to be published. T.T. Sheng and C. C. Chang, IEEE Trans. Electron Devices, 23 (1976) 531. R. Sinclair, in J. J. Hren, J. I. Goldstein and D. C. Joy (eds.), Introduction to Analytical Electron Microscopy, Plenum, New York, 1979, p. 507. L.J. Chen, J. W. Mayer and K. N. Tu, Proc. Syrup. on Thin Films and Interfaces, Boston, MA, November 16-19, 1981, in Thin Solid Films, 93 (1982) 135.
91
LATTICE I M A G I N G O F S I L I C I D E - S I L I C O N I N T E R F A C E S * L. J. CHEN t AND J. W. MAYER Department o f Materials Science and Engineering, Cornell University, Ithaca, N Y 14853 (U.S.A.) K.N. TU I B M T. J. Watson Research Center, Yorktown Heights, NY10598 (U.S.A.) T. T. SHENG Bell Laboratories, Murray Hill, NJ07974 (U.S.A.)
The interfaces of both epitaxial and non-epitaxial silicides and silicon were investigated by the direct lattice imaging method using cross-sectional samples. Non-epitaxial CoSi 2 on silicon was observed to have a curved interface. Epitaxial CoSi 2, however, was found to be smooth within a facet. No evidence of an amorphous layer at the interface was obtained. Epitaxial NiSi2 on Si(001) was found to be heavily faceted. The facets are on {111} and {100) planes with the former more frequently observed. The interface between Si(lll) and NiSi 2 is also faceted but less so than that for Si(001). The interface is very rough on a large scale. Straight boundary lines corresponding to faceted planes were observed which indicated that the interfaces on an atomic scale were quite smooth. Defect clusters and planar defects were also observed at the interfaces.
1. INTRODUCTION Thin films of silicides have found widespread use in integrated circuits as Schottky barriers, ohmic contacts, low resistivity gates and interconnects. Most silicide studies have been concentrated on electrical measurements, reaction kinetics and the identification of phases formed during reaction 1. A great deal of information has been obtained on electrical properties, activation energies of silicide growth, the sequence of phase formation and the predominant diffusing species during the reaction, yet much less is known about the structural and morphological aspects of silicide-silicon interfaces. The interfaces control many technically important properties of silicide contacts and interconnects such as Schottky barrier height, contact resistance and corrosion stability; consequently it is important to advance our knowledge of these interfaces. Moreover, there is a considerable basic interest in the atomic structure of the silicide-silicon interface, e.g. the existence of a thin * Paper presented at the Symposium on Thin Films and Interfaces, Boston, MA, U.S.A., November 16-19, 1981. "~Permanent address: Department of Materials Science and Engineering, National Tsing Hua University, Hsinchu, Taiwan. 0040-6090/82/0000-0000/$02~75
© Elsevier Sequoia/Printed in The Netherlands
92
L.J. C ~
et al.
amorphous layer at the interface has been proposed in order to predict the first silicide phase in transition metal-silicon systems2. F611 et al. have studied the epitaxial interfaces of Pd2Si/Si, NiSi2/Si as well as NiSi/Si by transmission electron microscopy (TEM) using the direct lattice imaging method on cross-sectional samples 3'4. The smoothness of the interfaces was revealed. No evidence of an amorphous interfacial layer was obtained. The resistivity of the silicides is the single most important criterion in considering them for the metallization in integrated circuits. Of all the silicides, CoSi2 has one of the lowest resistivities (18-20 ~ cm) s. Similar to NiSi2, CoSi2 has a cubic CaF2 structure with lattice constant ao = 5.37 ~ and the lattice mismatch with diamond cubic silicon is only about 1.1%. Both CoSi2 and NiSi2 are known to form epitaxially on low index planes of silicon6'7. A sketch for comparison of the CaF2 and silicon structures is shown in Fig. 1. It is straightforward to extend lattice imaging studies of the interfaces of other silicides to include epitaxial CoSi2. In addition, since a great majority of the silicides do not possess an epitaxial relationship with respect to the silicon matrix, it is also desirable to study the interfaces between non-epitaxial silicides and silicon. In this paper, we repoit the results of the study of both epitaxial and non-epitaxial interfaces of CoSia and silicon. For comparison, we have also studied many cross-sectional samples of epitaxial NiSi2 on silicon.
(a) (b) Fig. 1. Comparison of(a) the cubic GaF 2 structure (©, fluorine; O, calcium) and (b) the diamond cubic silicon (©) structure. 2. EXPERIMENTAL PROCEDURES
Nickel and cobalt thin films about 300-400 ,~ in thickness were deposited at room temperature onto (001)- and (111)-oriented silicon wafers by electron gun evaporation with a base vacuum better than 10- 7 Torr. The disilicides were formed by annealing treatments in a helium atmosphere or in a vacuum better than 10- ~ Torr. Conventional TEM specimens were prepared by chemical thinning from the silicon side; the deposited side was covered with an electronic wax for protection during chemical polishinga. Cross-sectional samples were prepared using the procedures outlined by Sheng and Chang a. The electron microscopy was performed in a JEOL 200 CX microscope operated at 200 kV or in a Siemens 102 microscope operated at 120 kV. 3. RESULTS AND DISCUSSIONS 3.1. Cobalt silicides
Polycrystalline CoSi 2 was formed by reacting cobalt thin films on silicon in a
LATTICE IMAGING OF INTERFACES
93
helium atmosphere at 850 °C for up to 4 h. The grains were found to be 0.1-0.5 ~tm in size. Figure 2 shows a cross-sectional view of a sample with polycrystalline CoSiz on Si(001). The interfaces between CoSiz grains and the silicon substrate were generally curved. For a CoSiz layer about 1500/~ thick, the estimated roughness was about 300 4. For samples annealed at 950 °C for 1 h, only epitaxial CoSi z was observed.
!6
Fig. 2. Bright field micrograph of a cross-sectional sample with polycrystalline CoSi 2 on silicon. Fig. 3. Lattice imaging of a cross-sectional sample with a CoSi 2 grain on silicon, rill [112] for silicon.
Figure 3 is an example of high resolution lattice imaging of a cross-sectional sample oriented along the [112] direction of the silicon substrate. Only (111) lattice fringes of silicon with an interplanar spacing of 3.14 ~ were observed; (220) fringes with a spacing of 1.92 ~ were too narrow to be resolved with the electron microscope utilized. No fringes were observed in the CoSi 2 grain and there were no extra diffraction spots corresponding to CoSi2 visible in the diffraction pattern. However, whether the absence of fringes in the CoSi2 grain was due to the effects of thickness or to the orientation of the CoSi2 grain was not determined. The interface region appeared to be brighter than other regions in the micrograph. The bright contrast could arise from the stress induced during deposition or during thermally induced reaction near the interface. One of the difficulties encountered in the high resolution imaging of cross-sectional samples, particularly when the silicide layer does not exhibit an epitaxial relationship with respect to the substrate silicon, is that lattice fringes may not be observed at the same time. A difference in ion milling rates between CoSi 2 and silicon during sample preparation may produce a thin foil with a substantial difference in thickness between the CoSi 2 and the silicon. The sample preparation may also produce artifacts, e.g. stress near the interface was relieved during the ion milling, which may obscure the true identity of the interface structure. For non-epitaxial silicides, the silicide layer usually did not have the "correct" orientation with respect to the substrate silicon so that when the sample was oriented with respect to silicon to reveal the lattice fringes of silicon, silicide fringes
94
L.J. CHENe t al.
were usually not visible. Only occasionally does a CoSi 2 grain possess just the "correct" orientation, so that lattice fringes in both the CoSi2 grain and the silicon matrix can be observed simultaneously. Such an example is shown in Fig. 4. The spacing of the CoSi2(111) fringes is about 1~o less than that of the Si(111) fringes. The fringes were always observed to extend right to the interface of the silicide and the matrix silicon so that no amorphous layer was evident at the interface.
Fig. 4. Exampleofa CoSi2 grain havingthe "correct"orientationso that latticefringesin both the C o S i 2 and the siliconmatrix can be observed. Fig. 5. High resolutionlatticeimage showingthe precursor ofepitaxialgrowth of CoSi2on silicon. For silicides having an epitaxial relationship to the silicon matrix, the stress around the interface may make the diffraction conditions for the epitaxial silicide and the silicon matrix different. As a consequence, although the so-called axial diffraction condition 1° (high symmetry condition) can be easily achieved in the thicker parts of the silicon by using Kikuchi lines, in interface regions thin enough for lattice imaging, the lattice fringes, if obtained at all, are no longer structural images. The interpretation of the lattice images is therefore more complicated. Nevertheless, the lattice images still provide valuable information not obtainable otherwise. Some of the CoSi2 grains exhibited an epitaxial relationship with respect to the silicon matrix as indicated by the presence of a regular dislocation network. Figure 5 shows an example of a sample oriented along the [001] direction of an Si(lll) matrix. The angle between CoSi2(111) fringes and Si(111) planes was measured to be about 70 °. No amorphous layer was observed at the interface. It is of particular interest that dark and bright bands, about 30/~ in thickness, were observed at the interface. Slight tilting of the sample away from the axial diffraction condition did not eliminate "band" contrast. The contrast may arise from the stress induced at the interface during the silieide formation. They were not thickness fringes, since the thickness fringes are of the order of hundreds of ~ngstr6ms under the diffraction condition for imaging. Close examination of the interface revealed that the first dark band exhibited a twinning relationship with respect to the silicon matrix. The occurrence of twinning at the interface between NiSi2 and silicon was reported by F611 et al. 3 and was also found by Chen e t al. Ix for CoSi2-Si interfaces. It is reasonable to conjecture that what we observed heretofore was the precursor of the epitaxial growth ofCoSi 2 on silicon, since the top layers ofCoSi 2 were non-epitaxial.
LATTICE IMAGINGOF INTERFACES
95
It is to be noted that the interface is relatively smooth as revealed by the straight boundary line. The straight boundary line was found to extend to more than 0.5 larn in length.
3.2. Nickel silicides NiSi2 was formed by annealing nickel thin films, 300-400/~ in thickness, at 800 °C for 0.5-1 h in a helium atmosphere. The silicides were found to be completely epitaxial with respect to the silicon matrix as was indicated by the existence of interface dislocation networks and by the analysis of diffraction patterns. Figure 6 shows a bright field fiat-on view of the sample illustrating the configuration of the dislocation network. The dislocations were identified to be of edge type with ~(112) Burgers' vectors. The average spacing between dislocations was about 1000/Yt which is close to the theoretically expected value of 970 ~ taking the lattice constant of NiSi2 to be a 0 = 5.406 A. Dislocation networks with 1(110) edge-type Burgers' vectors were also observed. Figure 7 shows a low magnification cross-sectional view of NiSi 2 on Si(001).
Fig. 6. Bright field micrograph of a flat-on sample showing the configuration of the dislocation network at the interfaceof NiSi2 and silicon. Fig. 7. Low magnificationcross-sectionalviewof NiSi2 on Si(001)showingthat the interfacesare heavily faceted.
Fig. 8. Lattice imaging of NiSi2 on Si(001).Defect clusters and planar defectsare indicated by arrows.
96
L.J. CHENe t al.
The interfaces are heavily faceted. The facets are on {111} and {100} planes with the former more frequently observed. The interface is very rough on a large scale. The interface between Si(111) and NiSiz is also faceted, but less so than for NiSi z on Si(001). Figure 8 shows a direct lattice imaging micrograph of NiSi 2 on Si(001). Flat boundary lines corresponding to facets were observed. The (111) fringes show an offset crossing the (100) interfaces similar to that observed by F611 e t al. a Detailed examinations showed the existence of irregularities at the interfaces. Defect clusters and extra planes parallel to the interface were observed at the interface as marked by circles A and B respectively. Similar defects were also found in many other examples. Although the possibility exists that the contrast was due to an artifact of the specimen preparation, it is also possible that they were genuine defective structures nucleated at the impurity sites of the interfaces. 4. CONCLUSIONS TEM studies, particularly high resolution lattice imaging of cross-sectional samples, provide valuable information on the structural and morphological aspects of silicide-silicon interfaces which is not obtainable otherwise. Non-epitaxial CoSi2 on silicon was shown to have curved interfaces with a roughness of 300 ~ for silicides about 1500/~ thick. Lattice fringes of CoSi2 were found to extend right to the interface so that there was no evidence of an amorphous layer at the interface. The precursor of epitaxial CoSi2 on silicon was also observed. The interfaces, up to 0.5 lain in length, appear to be very smooth. A twinning relationship between CoSi2 and silicon was also obtained. For epitaxial NiSi 2 on silicon, the interface dislocations were analyzed to be of edge type with ½(110) and-~(112) Burgers' vectors. The spacing of about 1000/~ between dislocations corresponds very well to the theoretically expected value 970 ~. The interfaces were found to be heavily faceted. The facets are on {111} and {100} planes with the former more frequently observed. The interface between Si(111) and NiSi 2 is also faceted, but less so than that for Si(001). The interface is very rough on a large scale. Straight boundary lines corresponding to facets were generally observed. Defect clusters and planar defects, possibly induced by the presence of impurities at the interface, were also detected. ACKNOWLEDGMENTS
The authors wish to thank L. S. Hung for technical assistance and H. C. Cheng and C. Y. Hou for help in the preparation of the manuscript. This work was supported in part by the National Science Foundation through the Materials Science Center at Cornell University. REFERENCES 1 K . N . Tu and J. W. Mayer, in J. M. Poate, K. N. Tu and J. W. Mayer (eds.), Thin Films-Interdiffusion and Reactions, Wiley, New York, 1978, p. 359. 2 R . M . Walser and R. W. Bene, Appl. Phys. Lett., 28 (1976) 624.
LATTICE IMAGING OF INTERFACES
3 4 5 6 7 8 9 10 11
97
H. F611, P. S. Ho and K. N. Tu, J. Appl. Phys., 52 (1981) 250. H. F611, P.S. H o a n d K . N. Tu, Philos. Mag. A,45(1982)31. S.P. Murarka, J. Vac. Sci. Technol., 17 (1980) 775. H. Ishiwara, in J, E. E. Baglin and J. M. Poate (eds.), Proc. Symp. on Thin Film Interfaces and Interactions, Los Angeles, CA, 1979, Electrochemical Society, Princeton, NJ, 1980, p. 159. J.C. Bean and J. M. Poate, Appl. Phys. Lett., 37 (1980) 643. L.J. Chert, L. S. Hung and J. W. Mayer, Appl. Surf. Sci., to be published. T.T. Sheng and C. C. Chang, IEEE Trans. Electron Devices, 23 (1976) 531. R. Sinclair, in J. J. Hren, J. I. Goldstein and D. C. Joy (eds.), Introduction to Analytical Electron Microscopy, Plenum, New York, 1979, p. 507. L.J. Chen, J. W. Mayer and K. N. Tu, Proc. Syrup. on Thin Films and Interfaces, Boston, MA, November 16-19, 1981, in Thin Solid Films, 93 (1982) 135.