JOURNAL OF ELECTRON MICROSCOPY TECHNIQUE 17:212-220 (1991)
Extending the Limit of Atomic Level Grain Boundary Structure Imaging Using High-Resolution Electron Microscopy WILLIAM KRAKOW IBM Research Dzuiszon, T J . Watson Research Center, Yorktown Heights, New York 10598
KEY WORDS
Grain boundaries, Interfaces, Bicrystals
It has been possible to image 2 = 2111 111121.8°tilt boundaries in thin Au films and ABSTRACT to deduce their atomic arrangements. These results represent a n electron microscopic resolution level of 1.43 A, attainable with a small amount of image processing, which produces interpretable structure images. This substantial improvement over other recent grain boundary studies, which required about 1.9-2.0 A resolution, clearly demonstrates that many more tilt grain boundary orientations are now accessible instead of a limited subset. level under optimum defocus conditions for 200-kV It has always been important to extract the highest microscopes with a spherical aberration coefficient resolution possible from a n electron microscope to re- around unity (C,= 1.0 mm). Some of the newer 200-kV veal finer structural details about defects in materials. microscopes have improved objective lens aberration To that end, the level of detail observed a t tilt grain coefficients of - 0.4-0.5 mm and therefore have a the(Yanaka et al., boundaries viewed edge-on, has steadily increased with oretical resolution level of -1.9-2.0 the improvement of microscopes as objective lens aber- 1988; Hewitt et al., 1989). However, intermediate voltrations have decreased and the accelerating voltages age microscopes operating a t 300 and 400 kV can eiincreased. The earliest atomic level lattice plane imag- ther equal or surpass the performance levels of these ing was possible in L l l O l orientations of semiconduc- 200-kV instruments with the added convenience of intors such as Si and Ge, where {Ill}lattice planes could creased specimen tilt for specimen orientational alignbe imaged in the bright-field mode with -3.1 A reso- ment purposes. Higher-voltage microscopes have only recently been lution in a 100-kV-type microscope (Bourret and Desapplied to imaging grain boundary structure a t the 2 A seaux, 1982; Bourret et al., 1982; Vaudin et al., 1983). resolution level in perhaps the last year or two. One of This was soon followed by structure imaging of the atom pair columns of various elemental semiconductor the first of these studies involved in the investigation tilt grain boundary misorientations along the same of the Z = 5/[0011,36.9"(310)symmetrical tilt boundary common rotation axis using 200-kV microscopes (D'An- in Au imaged with a 300-kV incident beam energy terroches and Bourret, 1984; Bourret and Bacmann, (Cosandey et al., 1988). This endeavor just resolved the 1986). In all these former cases, the locations of the 2 lattice spacings by second zone imaging; noisy improjected atom pairs, which are separated by -1.3 A, ages resulted, since the contrast transfer function is were deduced, and the appropriate bonding was then reduced to about 20% of the optimum transfer due to added to determine the interfacial structural units, partial coherence (Krakow, 1984). Most of the few which consisted of five- and seven-membered rings, other studies related to grain boundary structure have where each atom was most often tetrahedrally bonded. been performed with a 400-kV-type microscope in True atomic resolution a t tilt grain boundaries was which the microscope objective lens has a first crosssuccessfully achieved in metal systems using 200-kV over in the contrast transfer function under optimum microscopes. The most intensively investigated system defocus (Scherzer focus) of -1.75 A (Bourret and Peof choice was the I1101 orientation of Au, which has the nisson, 1987). In this type of microscope, it has recently most open projected lattice structure, where { l l l } lat- been demonstrated for several [OOl] tilt boundaries in tice planes of - 2.3 A are imaged on each side of the Ge that structure images can be obtained and that sevinterface. For this orientation, several coincidence eral structural variations of a I;= 13 boundary are cases have been studied, which include 2; = 11 (113) found (Bourret and Rouviere, 1988; Rouviere and Bour(Ishida et al., 1983); Z = 3 (111)(Ichinose and Ishida, ret, 1988). In the case of metals, I have recently inves1981; Krakow and Smith, 1987); 2 = 3 (112) (Krakow tigated l O O l 1 tilt boundaries in Au required that {ZOO} and Smith, 1987); and the Z = 19 (331) (Krakow et al., 1986). Here the axial illumination bright-field images were formed from the second zone of the phase-contrast transfer function. Even these larger spacings are someReceived October 19, 1989; accepted in revised form December 19, 1989. what beyond the first crossover in the contrast transfer Address reprint requests to Dr. William Krakow, IBM Research Center, Yorkfunction, which occurs around the 2.3 A resolution town Heights, NY 10598.
INTRODUCTION
a
a
~
c,
1991 WILEY-LISS, INC
ATOMIC LEVEL GRAIN BOUNDARY IMAGING
crossed lattice planes be imaged on both sides of the interface as a necessary condition to provide an atomic level interpretation a t the 2 A resolution level (Krakow, 1989a,b). Direct microscope evidence of structural multiplicity in a low-angle metal grain boundary was demonstrated for a Z = 17/[001128.1°(410) symmetrical tilt boundary, which has separated dislocation cores (Krakow, 1989a). Multiplicity was shown t o occur in a high-angle Z = 5/[001] 53.1" (210) symmetrical boundary (Krakow, 1989b). Aside from these recent studies, there have been no attempts to visualize boundary structures beyond the -2 A resolution level, where only a few specialized studies have been made to form lattice images even from perfect crystals. Among the earliest attempts a t including larger regions of reciprocal space, hence supposedly higher-resolution imaging, involved imaging [OOl] Au films in which the objective aperture included Bragg reflections out to the (620) spots (Hashimoto et al., 1977). It was found that atom column images had an internal fine structure that could be attributed to the higher-order Bragg reflections. This early work was performed on a 100-kV microscope; hence no useful structure information was present beyond the 2 A lattice spacings. Here, the contrast transfer function was oscillating rapidly, which further complicated the image interpretation. In a more recent study at 200 kV, a -30 A thick [ l l l ] Au thin film was imaged which showed crossed sets of (220) bulk lattice spacings of 1.43 A in the bright-field axial illumination mode (Krakow, 1984). The corresponding (220) reflections are the closest Bragg spots to the origin of reciprocal space for this orientation. In this case, the finite beam divergence of the incident illumination in the microscope introduced considerable attenuation in the transfer function in the reciprocal space region, where the Bragg reflections occurred to about 5-10% of the strong part of the transfer function, which is near unity. This in turn reduced the bulk lattice fringe intensity to where it was comparable to a monolayer type of atomic surface image, as confirmed by computer simulations in that report. Furthermore, in that earlier study, no attempt was made to determine atom positions, since image features would have been severely shifted by the objective lens aberrations. More recently, as part of the initial atomic imaging characterization and performance assessment of a 400kV microscope, various orientations of crystalline Ge were imaged by orienting specimens with a goniometer stage (Bourret and Penisson, 1987).Here structure images were observed for perfect crystals along the , , and zone axis orientations. Atom column images were obtained with the 1.4 distance between atom pairs clearly resolved in the orientation case with white atom contrast at an unusual defocus condition. However, no attempt was made at that time to image defect structures at that resolution level. Instead, Bourret and Penisson concentrated their efforts on oriented tilt boundary spacings corresponding to (110) lattice planes in Mo of 2.2 A. Therefore, to that point, no structure images of tilt grain boundaries with projected bicrystal grain spacings well below the 2 level have been reported
213
on. It is the intent in this paper to present structure images from a engineered bicrystalline specimens of Au containing tilt boundaries with a [111] common axis. An effective 1.43 A point resolution is required in each crystal grain and in the interfacial region.
EXPERIMENTAL PKOCEDURES In the present study, thin film Au bicrystals containing many tilt boundaries were prepared from -40 A thick epitaxially grown single crystals on (111)Ag substrates, which were also grown epitaxially on cleaved mica (Krakow, 1982). After the Au (111)films were removed from their substrates, two free-standing films were placed in contact on a microscope grid and heated in a microscope equipped with a hot stage to observe the interaction. Upon heating slowly to avoid hole formation to between 250-300°C, the films first fuse forming twist boundaries. Upon further heating, these boundaries migrate to the different surfaces of the composite film in different regions, producing regions that are a single crystal and separated by tilt boundaries. A number of these structures were fabricated at various misorientation angles, several of which were then at or near major coincidence orientations with symmetrical boundary segments. High-resolution electron microscopy was then performed in a JEOL 4000EX microscope operating at 400 kV and equipped with a high-resolution pole piece with a spherical aberration coefficient of Cs = 1.0 mm. Since 1.43 A lattice planes were of interest in this orientation, images were recorded at magnifications of x 800,000 on electron image sheet film with exposure times of
Extending the Limit of Atomic Level Grain Boundary Structure Imaging Using High-Resolution Electron Microscopy WILLIAM KRAKOW IBM Research Dzuiszon, T J . Watson Research Center, Yorktown Heights, New York 10598
KEY WORDS
Grain boundaries, Interfaces, Bicrystals
It has been possible to image 2 = 2111 111121.8°tilt boundaries in thin Au films and ABSTRACT to deduce their atomic arrangements. These results represent a n electron microscopic resolution level of 1.43 A, attainable with a small amount of image processing, which produces interpretable structure images. This substantial improvement over other recent grain boundary studies, which required about 1.9-2.0 A resolution, clearly demonstrates that many more tilt grain boundary orientations are now accessible instead of a limited subset. level under optimum defocus conditions for 200-kV It has always been important to extract the highest microscopes with a spherical aberration coefficient resolution possible from a n electron microscope to re- around unity (C,= 1.0 mm). Some of the newer 200-kV veal finer structural details about defects in materials. microscopes have improved objective lens aberration To that end, the level of detail observed a t tilt grain coefficients of - 0.4-0.5 mm and therefore have a the(Yanaka et al., boundaries viewed edge-on, has steadily increased with oretical resolution level of -1.9-2.0 the improvement of microscopes as objective lens aber- 1988; Hewitt et al., 1989). However, intermediate voltrations have decreased and the accelerating voltages age microscopes operating a t 300 and 400 kV can eiincreased. The earliest atomic level lattice plane imag- ther equal or surpass the performance levels of these ing was possible in L l l O l orientations of semiconduc- 200-kV instruments with the added convenience of intors such as Si and Ge, where {Ill}lattice planes could creased specimen tilt for specimen orientational alignbe imaged in the bright-field mode with -3.1 A reso- ment purposes. Higher-voltage microscopes have only recently been lution in a 100-kV-type microscope (Bourret and Desapplied to imaging grain boundary structure a t the 2 A seaux, 1982; Bourret et al., 1982; Vaudin et al., 1983). resolution level in perhaps the last year or two. One of This was soon followed by structure imaging of the atom pair columns of various elemental semiconductor the first of these studies involved in the investigation tilt grain boundary misorientations along the same of the Z = 5/[0011,36.9"(310)symmetrical tilt boundary common rotation axis using 200-kV microscopes (D'An- in Au imaged with a 300-kV incident beam energy terroches and Bourret, 1984; Bourret and Bacmann, (Cosandey et al., 1988). This endeavor just resolved the 1986). In all these former cases, the locations of the 2 lattice spacings by second zone imaging; noisy improjected atom pairs, which are separated by -1.3 A, ages resulted, since the contrast transfer function is were deduced, and the appropriate bonding was then reduced to about 20% of the optimum transfer due to added to determine the interfacial structural units, partial coherence (Krakow, 1984). Most of the few which consisted of five- and seven-membered rings, other studies related to grain boundary structure have where each atom was most often tetrahedrally bonded. been performed with a 400-kV-type microscope in True atomic resolution a t tilt grain boundaries was which the microscope objective lens has a first crosssuccessfully achieved in metal systems using 200-kV over in the contrast transfer function under optimum microscopes. The most intensively investigated system defocus (Scherzer focus) of -1.75 A (Bourret and Peof choice was the I1101 orientation of Au, which has the nisson, 1987). In this type of microscope, it has recently most open projected lattice structure, where { l l l } lat- been demonstrated for several [OOl] tilt boundaries in tice planes of - 2.3 A are imaged on each side of the Ge that structure images can be obtained and that sevinterface. For this orientation, several coincidence eral structural variations of a I;= 13 boundary are cases have been studied, which include 2; = 11 (113) found (Bourret and Rouviere, 1988; Rouviere and Bour(Ishida et al., 1983); Z = 3 (111)(Ichinose and Ishida, ret, 1988). In the case of metals, I have recently inves1981; Krakow and Smith, 1987); 2 = 3 (112) (Krakow tigated l O O l 1 tilt boundaries in Au required that {ZOO} and Smith, 1987); and the Z = 19 (331) (Krakow et al., 1986). Here the axial illumination bright-field images were formed from the second zone of the phase-contrast transfer function. Even these larger spacings are someReceived October 19, 1989; accepted in revised form December 19, 1989. what beyond the first crossover in the contrast transfer Address reprint requests to Dr. William Krakow, IBM Research Center, Yorkfunction, which occurs around the 2.3 A resolution town Heights, NY 10598.
INTRODUCTION
a
a
~
c,
1991 WILEY-LISS, INC
ATOMIC LEVEL GRAIN BOUNDARY IMAGING
crossed lattice planes be imaged on both sides of the interface as a necessary condition to provide an atomic level interpretation a t the 2 A resolution level (Krakow, 1989a,b). Direct microscope evidence of structural multiplicity in a low-angle metal grain boundary was demonstrated for a Z = 17/[001128.1°(410) symmetrical tilt boundary, which has separated dislocation cores (Krakow, 1989a). Multiplicity was shown t o occur in a high-angle Z = 5/[001] 53.1" (210) symmetrical boundary (Krakow, 1989b). Aside from these recent studies, there have been no attempts to visualize boundary structures beyond the -2 A resolution level, where only a few specialized studies have been made to form lattice images even from perfect crystals. Among the earliest attempts a t including larger regions of reciprocal space, hence supposedly higher-resolution imaging, involved imaging [OOl] Au films in which the objective aperture included Bragg reflections out to the (620) spots (Hashimoto et al., 1977). It was found that atom column images had an internal fine structure that could be attributed to the higher-order Bragg reflections. This early work was performed on a 100-kV microscope; hence no useful structure information was present beyond the 2 A lattice spacings. Here, the contrast transfer function was oscillating rapidly, which further complicated the image interpretation. In a more recent study at 200 kV, a -30 A thick [ l l l ] Au thin film was imaged which showed crossed sets of (220) bulk lattice spacings of 1.43 A in the bright-field axial illumination mode (Krakow, 1984). The corresponding (220) reflections are the closest Bragg spots to the origin of reciprocal space for this orientation. In this case, the finite beam divergence of the incident illumination in the microscope introduced considerable attenuation in the transfer function in the reciprocal space region, where the Bragg reflections occurred to about 5-10% of the strong part of the transfer function, which is near unity. This in turn reduced the bulk lattice fringe intensity to where it was comparable to a monolayer type of atomic surface image, as confirmed by computer simulations in that report. Furthermore, in that earlier study, no attempt was made to determine atom positions, since image features would have been severely shifted by the objective lens aberrations. More recently, as part of the initial atomic imaging characterization and performance assessment of a 400kV microscope, various orientations of crystalline Ge were imaged by orienting specimens with a goniometer stage (Bourret and Penisson, 1987).Here structure images were observed for perfect crystals along the , , and zone axis orientations. Atom column images were obtained with the 1.4 distance between atom pairs clearly resolved in the orientation case with white atom contrast at an unusual defocus condition. However, no attempt was made at that time to image defect structures at that resolution level. Instead, Bourret and Penisson concentrated their efforts on oriented tilt boundary spacings corresponding to (110) lattice planes in Mo of 2.2 A. Therefore, to that point, no structure images of tilt grain boundaries with projected bicrystal grain spacings well below the 2 level have been reported
213
on. It is the intent in this paper to present structure images from a engineered bicrystalline specimens of Au containing tilt boundaries with a [111] common axis. An effective 1.43 A point resolution is required in each crystal grain and in the interfacial region.
EXPERIMENTAL PKOCEDURES In the present study, thin film Au bicrystals containing many tilt boundaries were prepared from -40 A thick epitaxially grown single crystals on (111)Ag substrates, which were also grown epitaxially on cleaved mica (Krakow, 1982). After the Au (111)films were removed from their substrates, two free-standing films were placed in contact on a microscope grid and heated in a microscope equipped with a hot stage to observe the interaction. Upon heating slowly to avoid hole formation to between 250-300°C, the films first fuse forming twist boundaries. Upon further heating, these boundaries migrate to the different surfaces of the composite film in different regions, producing regions that are a single crystal and separated by tilt boundaries. A number of these structures were fabricated at various misorientation angles, several of which were then at or near major coincidence orientations with symmetrical boundary segments. High-resolution electron microscopy was then performed in a JEOL 4000EX microscope operating at 400 kV and equipped with a high-resolution pole piece with a spherical aberration coefficient of Cs = 1.0 mm. Since 1.43 A lattice planes were of interest in this orientation, images were recorded at magnifications of x 800,000 on electron image sheet film with exposure times of
