JOURNAL OF ELECTRON MICROSCOPY TECHNIQUE 18:117-120 (1991)
Large-Area Plan-View Sample Preparation for GaAs-Based Systems Grown by Molecular Beam Epitaxy DAVID J. HOWARD, DAVID C. PAINE, AND ROBERT N. SACKS Division of Engineering, Brown University, Providence, Rhode Island 02912 (D.J.H., D.C.P); United Technologies Research Center, East Hartford, Connecticut 06108 (R.N.S.)
KEY WORDS
TEM, Chemical liftoff, AlAs, InGaAs, Plan-view TEM
ABSTRACT
We describe a method for plan-view transmission electron microscopy (TEM) sample preparation that takes advantage of extreme etch-rate selectivity in GaAs and AlAs in HF/H,O solutions. GaAs/In,Ga,~,As/GaAs strained-layer films (x = 0.05, 0.10, 0.19, 0.22) were chemically lifted off using this technique and were mounted on Cu TEM grids such that TEM transparent areas of up to 1 x 2 mm of constant thickness (196.4 nm) could be viewed. This simple, large-area plan-view technique uses only chemical methods and significantly extends the usefulness of TEM for the evaluation of crystal quality in GaAs-based epitaxial systems. The method requires the growth of a release layer of AlAs (10 nm thick) prior to the layered structure of interest.
INTRODUCTION Recently, controversy has arisen concerning the question of critical thickness and the onset of strain relaxation in lattice-mismatched films of In,Ga,.,As on GaAs substrates. The controversy is largely due to the wide variation in sensitivity of the various techniques used to detect strain relaxation. Transmission electron microscopy (TEM),being a direct imaging technique, is among the most sensitive of the available methods, but it suffers from the relatively small areas that can be viewed in a single, conventional plan-view sample. Standard preparation of semiconducting materials for TEM involves a combination of mechanical, chemical, and ion-beam thinning techniques that produce a thin wedge-shaped disk that contains relatively small areas (typically 100-150 pm in diameter) of electron transparent material. With such small areas, TEM studies of interfacial misfit dislocations in strained-layer systems are impractical unless the defects are spaced less than 50-100 pm apart. Such heavily defected material is not of interest for device applications. We present a new technique for sample preparation designed to increase the utility of TEM for the evaluation of misfit defect densities in strained-layer GaAs systems. The large-area plan-view technique described in this paper is a chemical liftoff procedure that produces square centimeters of continuous electron-transparent material of constant thickness. The useful area is limited only by the size of the TEM specimen holder and the dimensions of the Cu ring used t o support the layer. With this technique, the film thickness is uniform across the entire sample and, in general, is well known. Consequently, contrast due to thickness fringes and concerns about strain relaxation during sample preparation for in situ experiments (e.g., Hull and Bean, 1989) are minimized. It also provides an attractive alternative to x-ray topography and photoluminescence for the evaluation of as-grown and postprocessed molecular beam epitaxy (MBE) crystal quality. Such TEM analysis will also reveal crystallographic data impor-
0 1991 WILEY-LISS, INC
tant to the improved understanding of dislocation nucleation, multiplication, and interaction mechanisms. This information is critical to the development of growth conditions that minimize defects in strainedlayer systems.
MATERIALS AND METHODS For the fabrication of plan-view samples, we made use of a recent report (Yablonovitch et al., 1987) that describes a technique for detaching thin epitaxial films from GaAs substrates. We have extended this technique and demonstrated its utility for fabricating samples for plan-view TEM evaluation of GaAs-based systems. The two essential results from the earlier work are 1) the extreme selectivity in etch rates between AlAs and GaAs in HF/H,O solutions and 2) the tendency of liquid thermoplastic wax to contract upon hardening, generating an internal tensile stress. The process can be summarized as follows. An AiAs release layer 20-100 nm thick is grown (usually by MBE) on a GaAs substrate. Because AlAs and GaAs are almost exactly lattice-matched, subsequent layers grown on top of the AlAs are unaffected by the presence of this liftoff layer. The epitaxial film grown on top of the release layer is then capped with thermoplastic wax. When submerged in HF/H,O solutions, the AlAs is dissolved, thus undercutting-and ultimately detaching-the thin film grown on top of the AlAs release layer. In this process, the role of the wax is twofold. First, it protects the surface of the thinlayered structure from chemical attack during processing and provides buoyancy for flotation once the film has fully detached from the substrate. Second, and more importantly, the forces imparted by the tension in
Received February 20, 1990; accepted in revised form July 10, 1990. Address reprint requests to David J . Howard, Division of Engineering, Box D, Brown University, Providence, RI 02912.
D.J. HOWARD m AL.
118
essary (and sufficient) to allow the escape of gaseous etchant products from the reaction area that would otherwise be trapped and stop the reaction. In our experience, the wax-induced curvature is elastically accommodated in the thin film (as expected a t room temperature in a covalent solid) and does not act to increase the density of dislocations. The details of our application of the liftoff process for the fabrication of TEM specimens are a s follows.
Thermoplastic Wax
GaAs InGa As x I-x GaAs 1
AlAs GaAs Buffer Layer
i
I
GaAs Substrate
Material growth The expitaxial layers for this work were grown in a Varian GEN I1 molecular-beam epitaxy system. During growth, the substrate temperature was maintained at 520°C. Dimeric As generated by a solid As cracker was used for all growths, and V/III flux ratios were approximately 12. The growth began with a 100-nmthick buffer layer of GaAs, followed by 10-nm-thick layer of AlAs, 80 nm of GaAs, 36.4 nm of In,Gal.,As, and finally a 80-nm-thick GaAs cap. A total of four of these structures were grown, all with identical film thicknesses, but with increasing Indium concentrations (x = 0.05, 0.10, 0.19, 0.22). The alloy compositions were established using double-crystal rocking curves, which were obtained from each of the wafers after growth using a Bede QCl table-top difractometer (Tanner, 1989).
I Fig. 1. Schematic view of the layered structure (not to scale). Note that the top three layers and the surface coating of wax are being undercut as the AlAs liftoff layer is etched away in HF:H,O.
the wax act to slightly curl the thin film that is being chemically undercut. The radius of curvature due to the wax, although undetected by the naked eye, are nec-
Fig. 2. Very low-magnification TEM montage showing the large area of electron-transparent material available with the large-area plan-view technique.
LARGE-AREA PLAN-VIEW TECHNIQUE
ing them to slide free and float to the surface. The gentle nudging was necessary in order to sever wax that overlapped the edges of the trilayer and adhered to the substrade during processing. 3 . Next, the floating layers were scooped out of the HF/H,O solution with filter paper and mounted on Cu TEM support rings (Pella # 1GC12H) with Torr Seal vacuum epoxy. The vacuum epoxy, chosen for its low vapor pressure and low curing temperature, was allowed to cure for 24 hours at room temperature. 4. When the epoxy dried, the black wax was dissolved by carefully passing the samples through a series of trichloroethylene baths until clean. The mounted films were now on the order of 1,OOOA thick and were unprotected and very fragile. A low-magnification montage of a typical TEM sample produced by this technique is shown in Figure 2.
as-grown
X x
-
calculated
x
00
01
02
03
04
119
05
X
Fig. 3. A graph of average dislocation spacing and calculated spacing (from theory of Matthews and Blakeslee, 1974) vs. Indium concentration. Note that the variations in dislocation spacing represent a five orders of magnitude range, from millimeters to nanometers.
Liftoff The primary goal during the development of the liftoff procedure for our TEM studies was to obtain a fast, easy method for detaching a series of trilayer structures, consisting of GaAs/In,Ga,.,As/GaAs (shown schematically in Fig. l ) , from their substrates without introducing additional defects. The procedure can be viewed as a four-step process. 1. First, a n acid-resistant thermoplastic wax (black wax) was dissolved (1:1 by volume) in trichloroethylene. This wax solution was applied in drops with a stirring rod to the GaAs surface to a thickness greater than 1 mm and was allowed to cure for 20 minutes at room temperature. Black wax was chosen because it is chemically resistant to HF/H,O solutions, and it contracts upon curing generating a tensile stress within the wax. The wax also protects the surfaces of the films to be studied. The wafer is then baked a t 100°C for 30 minutes to allow the wax to melt and spread uniformly over the surface. 2. After cooling, samples were cleaved by scribing the backside of the wafer, using a diamond point scribe, into approximately 3 x 3 mm sizes. These samples were then submerged, wax side up, in a 25% HF/H,O solution. After 80 minutes, most of the samples had fully detached from the substrate and floated to the surface of the HF/H,O solution. The few samples that did not fully detach from the substrate were gently nudged with a small, rounded-end stirring stick, caus-
Note that two steps were taken to prevent the curling that typically accompanies large areas of freestanding films thin enough to be electron transparent; the thin films were rigidly fixed to Cu support rings prior to wax removal, and the trilayers were symmetrically strained in design through the sandwiching of the (strained) In,Ga,.,As layer between two layers of GaAs of equal thickness. In addition, it was essential that the effects of the tensile forces in the wax be felt all the way t o the edges of the sample (since the undercutting reaction would be stopped if their was inadequate transport of reaction by-products). In our experience, the simple painting of wax solution onto precut samples was not effective. This was because i t was difficult to apply the wax close enough to the edges of the sample without extending over onto the AlAs or substrate. If the wax was not close enough to a n edge, the uncovered region would not be subject to the tensile forces necessary to open a reactive pathway in the AlAs layer beneath it. If, on the other hand, the wax overlapped onto the AlAs layer and substrate, this would cause additional problems such as a decrease in the exposed etching and outdiffusion area (which would increase process time) and trilayer adhesion to the substrate even after complete dissolution of the AlAs layer. For these reasons, the cleaving process was important because i t fractured the wax and specimen together, leaving no gap or overhang between the trilayer and the wax, thus maximizing the tension effect for fast, successful liftoff.
RESULTS AND DISCUSSION Using the liftoff technique, TEM samples similar to those shown in Figure 2 were fabricated as part of a n investigation (Paine et al., 1989) of strain relaxation in GaAs/In,Ga,-,As/GaAs systems. These samples included a range (x = 0.05, 0.10, 0.19, 0.22) of Indium concentrations and dislocation densities (0 to < 1 dislocation1100 nm2). In one case, a n average dislocation spacing of less than 10 nm was observed using weak beam TEM techniques. In another case, individual dislocations were spaced greater than 1 mm apart and were identified and characterized. The success of this study shows that the liftoff process is unaffected by the density of defects or the In concentrations in the films.
120
D.J. HOWARD ET AL.
The wide range of defect densities that can be determined by using TEM in conjunction with the liftoff technique is evident from Figure 3. In this figure; dislocation spacings that range five orders of magnitude, (from nanometers to millimeters) are plotted. Some advantages of the liftoff technique are that it is purely chemical; no grinding or polishing is necessary. It is comparatively easy and rapidly produces extremely large, useful areas of thin film. The samples produced are, in our experience, extremely flat (minimizing bend contours) and are by design uniform in thickness (eliminating thickness contours). Direct TEM observation, unlike other commonly used film evaluation methods such as x-ray topography or photoluminescence, allows crystallographic information to be obtained from individual defects. Although this technique requires the growth of a n AlAs layer prior t o the growth of the desired GaAs-based structure, AlAs is almost exactly lattice matched with GaAs and thus does not affect the as-grown defect density. The technique we have described provides a powerful tool for
the investigation of dislocation mechanics and strain relaxation a s well a s a quick and accurate method of evaluating the crystalline quality in single and multilayer GaAs films.
ACKNOWLEDGMENTS Discussions with Mil Branestofort of United Technologies are gratefully acknowledged. REFERENCES Hull, R., and Bean, J.C. (1989) Thermal stability of Si/Ge,Si,~,/Si heterostructures. ADDLPhvs. Lett.. 5518. Matthews, J.W., and BiakesGe, A.E. i1974) Defects in epitaxial multilayers. J. Crystal Growth, 27:118. Paine, D.C., Howard, D.J., Luo, D.W., Sacks, R.N., and Estrich, T.C. (19891 The relaxation of In,Ga, .,As/GaAs strained multilayers. Presented a t the 1989 MRS Fall Meeting, Boston, MRS Svmuosia "~ Proceedings Volume 160. Tanner, K. (1989) X-ray topography and precision diffractometry of semiconducting materials. J. Electrochem. SOC., 136:3438. Yablonovitch, E., Gmitter, T., Harbison, J.P., and Bhat, R. (1987) Extreme selectivity in the lift-off of epitaxial GaAs films. Appl. Phys. Lett., 51:2222.
Large-Area Plan-View Sample Preparation for GaAs-Based Systems Grown by Molecular Beam Epitaxy DAVID J. HOWARD, DAVID C. PAINE, AND ROBERT N. SACKS Division of Engineering, Brown University, Providence, Rhode Island 02912 (D.J.H., D.C.P); United Technologies Research Center, East Hartford, Connecticut 06108 (R.N.S.)
KEY WORDS
TEM, Chemical liftoff, AlAs, InGaAs, Plan-view TEM
ABSTRACT
We describe a method for plan-view transmission electron microscopy (TEM) sample preparation that takes advantage of extreme etch-rate selectivity in GaAs and AlAs in HF/H,O solutions. GaAs/In,Ga,~,As/GaAs strained-layer films (x = 0.05, 0.10, 0.19, 0.22) were chemically lifted off using this technique and were mounted on Cu TEM grids such that TEM transparent areas of up to 1 x 2 mm of constant thickness (196.4 nm) could be viewed. This simple, large-area plan-view technique uses only chemical methods and significantly extends the usefulness of TEM for the evaluation of crystal quality in GaAs-based epitaxial systems. The method requires the growth of a release layer of AlAs (10 nm thick) prior to the layered structure of interest.
INTRODUCTION Recently, controversy has arisen concerning the question of critical thickness and the onset of strain relaxation in lattice-mismatched films of In,Ga,.,As on GaAs substrates. The controversy is largely due to the wide variation in sensitivity of the various techniques used to detect strain relaxation. Transmission electron microscopy (TEM),being a direct imaging technique, is among the most sensitive of the available methods, but it suffers from the relatively small areas that can be viewed in a single, conventional plan-view sample. Standard preparation of semiconducting materials for TEM involves a combination of mechanical, chemical, and ion-beam thinning techniques that produce a thin wedge-shaped disk that contains relatively small areas (typically 100-150 pm in diameter) of electron transparent material. With such small areas, TEM studies of interfacial misfit dislocations in strained-layer systems are impractical unless the defects are spaced less than 50-100 pm apart. Such heavily defected material is not of interest for device applications. We present a new technique for sample preparation designed to increase the utility of TEM for the evaluation of misfit defect densities in strained-layer GaAs systems. The large-area plan-view technique described in this paper is a chemical liftoff procedure that produces square centimeters of continuous electron-transparent material of constant thickness. The useful area is limited only by the size of the TEM specimen holder and the dimensions of the Cu ring used t o support the layer. With this technique, the film thickness is uniform across the entire sample and, in general, is well known. Consequently, contrast due to thickness fringes and concerns about strain relaxation during sample preparation for in situ experiments (e.g., Hull and Bean, 1989) are minimized. It also provides an attractive alternative to x-ray topography and photoluminescence for the evaluation of as-grown and postprocessed molecular beam epitaxy (MBE) crystal quality. Such TEM analysis will also reveal crystallographic data impor-
0 1991 WILEY-LISS, INC
tant to the improved understanding of dislocation nucleation, multiplication, and interaction mechanisms. This information is critical to the development of growth conditions that minimize defects in strainedlayer systems.
MATERIALS AND METHODS For the fabrication of plan-view samples, we made use of a recent report (Yablonovitch et al., 1987) that describes a technique for detaching thin epitaxial films from GaAs substrates. We have extended this technique and demonstrated its utility for fabricating samples for plan-view TEM evaluation of GaAs-based systems. The two essential results from the earlier work are 1) the extreme selectivity in etch rates between AlAs and GaAs in HF/H,O solutions and 2) the tendency of liquid thermoplastic wax to contract upon hardening, generating an internal tensile stress. The process can be summarized as follows. An AiAs release layer 20-100 nm thick is grown (usually by MBE) on a GaAs substrate. Because AlAs and GaAs are almost exactly lattice-matched, subsequent layers grown on top of the AlAs are unaffected by the presence of this liftoff layer. The epitaxial film grown on top of the release layer is then capped with thermoplastic wax. When submerged in HF/H,O solutions, the AlAs is dissolved, thus undercutting-and ultimately detaching-the thin film grown on top of the AlAs release layer. In this process, the role of the wax is twofold. First, it protects the surface of the thinlayered structure from chemical attack during processing and provides buoyancy for flotation once the film has fully detached from the substrate. Second, and more importantly, the forces imparted by the tension in
Received February 20, 1990; accepted in revised form July 10, 1990. Address reprint requests to David J . Howard, Division of Engineering, Box D, Brown University, Providence, RI 02912.
D.J. HOWARD m AL.
118
essary (and sufficient) to allow the escape of gaseous etchant products from the reaction area that would otherwise be trapped and stop the reaction. In our experience, the wax-induced curvature is elastically accommodated in the thin film (as expected a t room temperature in a covalent solid) and does not act to increase the density of dislocations. The details of our application of the liftoff process for the fabrication of TEM specimens are a s follows.
Thermoplastic Wax
GaAs InGa As x I-x GaAs 1
AlAs GaAs Buffer Layer
i
I
GaAs Substrate
Material growth The expitaxial layers for this work were grown in a Varian GEN I1 molecular-beam epitaxy system. During growth, the substrate temperature was maintained at 520°C. Dimeric As generated by a solid As cracker was used for all growths, and V/III flux ratios were approximately 12. The growth began with a 100-nmthick buffer layer of GaAs, followed by 10-nm-thick layer of AlAs, 80 nm of GaAs, 36.4 nm of In,Gal.,As, and finally a 80-nm-thick GaAs cap. A total of four of these structures were grown, all with identical film thicknesses, but with increasing Indium concentrations (x = 0.05, 0.10, 0.19, 0.22). The alloy compositions were established using double-crystal rocking curves, which were obtained from each of the wafers after growth using a Bede QCl table-top difractometer (Tanner, 1989).
I Fig. 1. Schematic view of the layered structure (not to scale). Note that the top three layers and the surface coating of wax are being undercut as the AlAs liftoff layer is etched away in HF:H,O.
the wax act to slightly curl the thin film that is being chemically undercut. The radius of curvature due to the wax, although undetected by the naked eye, are nec-
Fig. 2. Very low-magnification TEM montage showing the large area of electron-transparent material available with the large-area plan-view technique.
LARGE-AREA PLAN-VIEW TECHNIQUE
ing them to slide free and float to the surface. The gentle nudging was necessary in order to sever wax that overlapped the edges of the trilayer and adhered to the substrade during processing. 3 . Next, the floating layers were scooped out of the HF/H,O solution with filter paper and mounted on Cu TEM support rings (Pella # 1GC12H) with Torr Seal vacuum epoxy. The vacuum epoxy, chosen for its low vapor pressure and low curing temperature, was allowed to cure for 24 hours at room temperature. 4. When the epoxy dried, the black wax was dissolved by carefully passing the samples through a series of trichloroethylene baths until clean. The mounted films were now on the order of 1,OOOA thick and were unprotected and very fragile. A low-magnification montage of a typical TEM sample produced by this technique is shown in Figure 2.
as-grown
X x
-
calculated
x
00
01
02
03
04
119
05
X
Fig. 3. A graph of average dislocation spacing and calculated spacing (from theory of Matthews and Blakeslee, 1974) vs. Indium concentration. Note that the variations in dislocation spacing represent a five orders of magnitude range, from millimeters to nanometers.
Liftoff The primary goal during the development of the liftoff procedure for our TEM studies was to obtain a fast, easy method for detaching a series of trilayer structures, consisting of GaAs/In,Ga,.,As/GaAs (shown schematically in Fig. l ) , from their substrates without introducing additional defects. The procedure can be viewed as a four-step process. 1. First, a n acid-resistant thermoplastic wax (black wax) was dissolved (1:1 by volume) in trichloroethylene. This wax solution was applied in drops with a stirring rod to the GaAs surface to a thickness greater than 1 mm and was allowed to cure for 20 minutes at room temperature. Black wax was chosen because it is chemically resistant to HF/H,O solutions, and it contracts upon curing generating a tensile stress within the wax. The wax also protects the surfaces of the films to be studied. The wafer is then baked a t 100°C for 30 minutes to allow the wax to melt and spread uniformly over the surface. 2. After cooling, samples were cleaved by scribing the backside of the wafer, using a diamond point scribe, into approximately 3 x 3 mm sizes. These samples were then submerged, wax side up, in a 25% HF/H,O solution. After 80 minutes, most of the samples had fully detached from the substrate and floated to the surface of the HF/H,O solution. The few samples that did not fully detach from the substrate were gently nudged with a small, rounded-end stirring stick, caus-
Note that two steps were taken to prevent the curling that typically accompanies large areas of freestanding films thin enough to be electron transparent; the thin films were rigidly fixed to Cu support rings prior to wax removal, and the trilayers were symmetrically strained in design through the sandwiching of the (strained) In,Ga,.,As layer between two layers of GaAs of equal thickness. In addition, it was essential that the effects of the tensile forces in the wax be felt all the way t o the edges of the sample (since the undercutting reaction would be stopped if their was inadequate transport of reaction by-products). In our experience, the simple painting of wax solution onto precut samples was not effective. This was because i t was difficult to apply the wax close enough to the edges of the sample without extending over onto the AlAs or substrate. If the wax was not close enough to a n edge, the uncovered region would not be subject to the tensile forces necessary to open a reactive pathway in the AlAs layer beneath it. If, on the other hand, the wax overlapped onto the AlAs layer and substrate, this would cause additional problems such as a decrease in the exposed etching and outdiffusion area (which would increase process time) and trilayer adhesion to the substrate even after complete dissolution of the AlAs layer. For these reasons, the cleaving process was important because i t fractured the wax and specimen together, leaving no gap or overhang between the trilayer and the wax, thus maximizing the tension effect for fast, successful liftoff.
RESULTS AND DISCUSSION Using the liftoff technique, TEM samples similar to those shown in Figure 2 were fabricated as part of a n investigation (Paine et al., 1989) of strain relaxation in GaAs/In,Ga,-,As/GaAs systems. These samples included a range (x = 0.05, 0.10, 0.19, 0.22) of Indium concentrations and dislocation densities (0 to < 1 dislocation1100 nm2). In one case, a n average dislocation spacing of less than 10 nm was observed using weak beam TEM techniques. In another case, individual dislocations were spaced greater than 1 mm apart and were identified and characterized. The success of this study shows that the liftoff process is unaffected by the density of defects or the In concentrations in the films.
120
D.J. HOWARD ET AL.
The wide range of defect densities that can be determined by using TEM in conjunction with the liftoff technique is evident from Figure 3. In this figure; dislocation spacings that range five orders of magnitude, (from nanometers to millimeters) are plotted. Some advantages of the liftoff technique are that it is purely chemical; no grinding or polishing is necessary. It is comparatively easy and rapidly produces extremely large, useful areas of thin film. The samples produced are, in our experience, extremely flat (minimizing bend contours) and are by design uniform in thickness (eliminating thickness contours). Direct TEM observation, unlike other commonly used film evaluation methods such as x-ray topography or photoluminescence, allows crystallographic information to be obtained from individual defects. Although this technique requires the growth of a n AlAs layer prior t o the growth of the desired GaAs-based structure, AlAs is almost exactly lattice matched with GaAs and thus does not affect the as-grown defect density. The technique we have described provides a powerful tool for
the investigation of dislocation mechanics and strain relaxation a s well a s a quick and accurate method of evaluating the crystalline quality in single and multilayer GaAs films.
ACKNOWLEDGMENTS Discussions with Mil Branestofort of United Technologies are gratefully acknowledged. REFERENCES Hull, R., and Bean, J.C. (1989) Thermal stability of Si/Ge,Si,~,/Si heterostructures. ADDLPhvs. Lett.. 5518. Matthews, J.W., and BiakesGe, A.E. i1974) Defects in epitaxial multilayers. J. Crystal Growth, 27:118. Paine, D.C., Howard, D.J., Luo, D.W., Sacks, R.N., and Estrich, T.C. (19891 The relaxation of In,Ga, .,As/GaAs strained multilayers. Presented a t the 1989 MRS Fall Meeting, Boston, MRS Svmuosia "~ Proceedings Volume 160. Tanner, K. (1989) X-ray topography and precision diffractometry of semiconducting materials. J. Electrochem. SOC., 136:3438. Yablonovitch, E., Gmitter, T., Harbison, J.P., and Bhat, R. (1987) Extreme selectivity in the lift-off of epitaxial GaAs films. Appl. Phys. Lett., 51:2222.
