Techniques for isolation and electrical characterization of individual grain boundaries in polycrystalline superconductors G. Schindler, C. Sarma, D.G. Haase*, C.C. Koch and A.I. Kingon North Carolina State University, Department of Materials Science and Engineering, and *Department of Physics, Raleigh, NC 27695-7907, USA Received 3 June 1993; revised 30 September 1993
The electrical characterization of single grain boundaries in polycrystalline high Tc superconductors is a useful means for understanding the low critical current densities encountered in these materials. Various techniques, such as thinning, depositing of contact pads, bonding, mounting and patterning, for the preparation of samples with isolated grain boundaries are described and discussed. Preliminary results show that Josephson junctions as well as grain boundaries with flux pinning behaviour can be found. Keywords: high Tc superconductors; grain boundaries; Josephson junctions
Improvement in the critical current density Jc is required if the high Tc superconducting oxides are to become replacements for conventional conductors or for traditional low T~ superconductors in many engineering applications. It has been demonstrated that in epitaxiai thin films, YBa2Cu3OT_8(YBCO) is capable of carrying high current densities, even at 77 K in magnetic fields of a few tesla, and at lower temperatures in very high fields 1. However, less progress has been made in solving the so-called weaklink problem at grain boundaries which limits Jc in bulk polycrystalline YBCO. It has been shown that the transport critical current density of bulk sintered polycrystalline YBCO is extremely low (--~102103Acm 2 at 77K) even in zero applied magnetic fields, and drops rapidly in the presence of a few hundred gauss 2. The microstructural features believed responsible for these low Jc values include: 1, microcracking due to thermal and transformational stresses3'4; 2, chemical variations at the grain boundaries, including second phases and impuritiesS; 3, structural variations at boundaries such as dislocations, or incoherence of CuO2 planes cross the boundary; 4, the intrinsic anisotropy of the superconducting properties6; and 5, the absence of flux pinning sites within grains. The 'grain boundary problem', i.e. the above points, is the focus of this paper. If microcracking and chemical inhomogeneities at the grain boundaries can be eliminated by careful processing techniques the misorientation angle between adjacent grains generally dictates
the current capability across the boundary. Dimos et al. TM illustrated the effect of intrinsic anistropy of the superconducting properties and the presence of structural disorder at grain boundaries on Jc with artificially fabricated YBCO thin film grain boundaries on bicrystals of SrTiO3. Intergranular Jc values were found to decrease rapidly with increasing misorientation angle, even for a symmetric (100) tilt boundary. Babcock et al. 9 however showed that some naturally occurring high angle boundaries in flux-grown bicrystals can form low energy configurations and that weak-link behaviour at such boundaries is significantly suppressed. Larbalestier e t al. 1° identified three kinds of transport behaviour across these flux grown bicrystals. These were: 1, flux pinning (FP); 2, Josephson junction (J J); and 3, resistive (R) behaviour. In general the transition from FP to JJ to R behaviour was observed as the misorientation angle increased. However, FP was observed in some cases at high angles (e.g. 28°) and JJ at relatively low angles (10°). At the present time no unique relation between grain boundary structure and Jc has been established but it is clear that decreasing the misorientation between adjacent grains is desirable. Thus, enhanced transport Jc values have been found, first in magnetically aligned materials 11 and then melt-processed superconductors 12. To understand the effects of grain structure, chemistry and orientation on Jc, direct transport measurements across well characterized boundaries are needed in bulk polycrystalline oxide superconductors by analogy to the work referred to above on thin
Cryogenics 1994 Volume 34, Number 4 287
Techniques for isolation and electrical characterization: G. Schindler et a l. films7"8 and bicrystals9'1°. Schindler et al. 13 were the first to report current measurements across single grain boundaries in polycrystalline oxide superconductors. They thinned polycrystailine DyBaeCu307_~ ceramics with coarse 100-200mm diameter grains to thickness less than the grain diameter, sputtered Au pads through a metal mask for contacts and used laser patterning to separate the grain boundaries. All the boundaries studied in this investigation exhibited Josephson junction behaviour. The present work concerns the isolation of grain boundaries in a wider range of polycrystalline oxide superconductors including melt-textured specimens with relatively high values of J¢. This paper describes in detail the problems associated with the thinning of the bulk superconductors as well as the contact methods and patterning to isolate individual grain boundaries for electrical measurements. Some preliminary results of electrical transport measurements on individual grain boundaries are also reported.
Preparation of samples Bulk ceramic samples resemble a three dimensional network of grain boundaries. In order to isolate one of these grain boundaries for electrical measurement, the connectivity of the sample must be reduced. The first step is to grind the sample until a thickness smaller than the typical grain size is reached. This gives a planar array of grain boundaries. Contact pads are deposited on the surface of the sample and gold wires are attached to them. The last step involves an etching process to separate the sample in two parts that are connected only by a small bridge crossing the chosen grain boundary. In this way all side effects by other grain boundaries are eliminated and the properties of a single grain boundary can be measured. This technique assumes that the critical current is limited by the grain boundaries and that the grains themselves have a much higher Jc. The processing steps involved in the preparation of the samples will be discussed in detail in the following section. For our measurements disc shaped YBazCu307_~ samples were grown by the melt-textured growth technique by M. Daeumling*. The dimensions were approximately 10 mm in diameter and 1 mm in thickness. These samples were embedded in epoxy resin with some amount of CuO powder. One side of the sample was first lapped using 600/xm grit size sand paper and low water ethanol as a lubricant. After lapping the sample was polished with 6, 1 and 0.25/zm diamond polishing compound. The sample was then mounted with the polished side on a glass slide. The other side of the sample plus epoxy was lapped using 240, 400 and 600 grit size sand paper. As the sample was thinned the epoxy-CuO mixture became transparent. This served as an indicator for the thickness. When the desired thickness was obtained, the sample was polished using the same procedure described above. *The samples were prepared by M. Daeumling, IBM T.J. Watson Research Center, Yorktown Heights, NY, USA
Cryogenics 1994 V o l u m e 34, N u m b e r 4
The samples with gold contacts were mounted on polycrystalline MgO substrates using a special temperature resistant adhesive (Ablebond 84-3, Ablestik Laboratories, CA, USA). Polycrystalline MgO has a coefficient of thermal expansion close to that of YBCO and thus reduces the stress on the sample during cooling. The MgO substrate with the sample was then mounted on a 16-lead side braze chip carrier using the same adhesive. The adhesive was cured at 150°C for l h . The particular mounting adhesive was chosen because it is not rendered soft by heating during wire bonding (see below) and did not produce gas bubbles between the YBCO and the substrate. Several other adhesives were tested and found to degrade the YBCO superconducting properties. There was concern that neither Tc nor the temperature width of the superconducting transition should be altered by effects of polishing or mounting. For convenience the superconducting transitions were characterized by measurement of the a.c. susceptibility with a simple a.c. inductance bridge 14. The sample slides were clamped on top of a flat 1.0 cm outer diameter inductance coil. The samples were roughly centred on the coil axis and were in contact with a calibrated platinum thin film thermometer. The glass slide was located between the coil and the sample and the platinum thermometer was held to the sample with a copper spring clamp. At intervals in the thinning process the sample thickness was determined and the superconducting transition measured in the susceptibility cryostat. The technique is simple to implement but sensitive to the geometry of the sample. As pointed out by Claassen et al.15 such a.c. susceptibility measurements indicate the critical value of the superconducting shielding current induced by the a.c. magnetic field from the inductance coil. The technique probes a surface depth much smaller than the thickness of the films (>10~m); therefore there should be no effects due to the decreasing sample thickness. Also the measurement technique was sensitive only to the surface of the sample facing the glass slide, i.e. the unpolished portion of the sample. In such measurements it is usually assumed that the sample width is much larger than the coil diameter TM. Because of the polishing process the lateral dimensions of the samples (~-3×3mm 2) decreased as more material was removed. It was possible to reproducibly measure transition temperatures, but at the smallest sample size there were background signals due to the copper sample clamp and the thermometer. In preliminary tests, a number of different glues were used to bond bulk samples to the glass slides, including '5-minute epoxy' (Devcon 5-Minute Epoxy Gel, Devcon Corp., IL, USA), M-Bond 610 (Measurements Group Inc., NC, USA) and contact cement (Contact Cement No. 00442, Conros Corp., MI, USA). As the samples were thinned it was noted that the superconducting transition temperatures did not change, but the temperature width of the transition increased. For each sample the transition width was defined as the temperature range from the onset of a change in the a.c. susceptibility to its completion in a diamagnetic state. This change in the transition width was used as the primary indicator of the quality of superconducting
Techniques for isolation and electrical characterization: G. Schindler et al.
× × 5 minute epoxy
0--0 Contact cement
A AM Bond 610
Thickness (p.m) Figure 1 Transition width as measured by a.c. susceptibility versus sample thickness for samples glued with different adhesives. The lines are guides for the eye
properties of the thinned samples. In Figure 1 we display the transition width for samples bonded with the three cements. In all three cases no significant degradation of the samples are indicated until the thicknesses are less than 40/zm. It is seen that the transitions of the samples bonded with 5-minute epoxy broadened only slightly as the thickness decreased to less than 20/~m. The samples bonded with the other adhesives systematically degraded when thinned to less than 40/~m. To establish electrical contacts to grains, techniques used in microelectronics are suitable. Since direct bonding to the superconductor has turned out not to be feasible, gold pads in a suitable pattern must be sputtered onto the surface for subsequent wire bonding. Grain sizes of the melt-textured grown samples are typically, on the order of 1 mm. A pattern of 100 x 100/~m~ pads with a 200-300#m centre-to-centre distance gives a suitable coverage on the sample for four point measurement across the grain boundaries. For the fabrication of the pads, the techniques of photolithography, shadow mask deposition and ion etching have been attempted. Photolithography gave very sharply defined pads and the mask alignment for a suitable distribution of pads was relatively easy. The major drawback of this technique was that the regions of the sample where gold pads were going to be located were exposed to water during the developing of the pattern. This exposure led to the formation of an insulating layer under the pads and thus a high contact resistance. Shadow mask deposition involved putting a metal foil (i.e. Ni), which had holes corresponding to a desired pattern, on the top of the samples during sputtering. A strong rare earth magnet was placed under the sample to pull the nickel mask straight during sputtering. The mask was made of a 6#m thick nickel foil using photolithography and subsequent etching in dilute HNO3. The nickel foil was not completely smooth, so a small gap between the superconductor and the foil was not always avoidable. This resulted in slightly larger pads with 'washed out' edges. The typical size for a pad deposited this way was 150 x 150/~m2. To produce pads by ion etching, the whole sample was first covered with a 0.5-1/~m thick layer of gold. Photoresist was spun on the gold layer and patterned.
No damage to the sample due to water contact was done during development of the resist, because the gold layer served as an effective protection for the superconductor. The gold which was not protected by photoresist was then etched using 100eV argon ions. This technique delivered a very good pattern of gold pads. However, the adhesion of the pads to the superconductor seemed to be not as high as with the other techniques. A different patterning technique is required for measurement of samples with smaller grain sizes (=100/zm), because fairly large pads for bonding are needed. First, the whole superconductor was covered with chemically vapour deposited SiO> Small windows were opened up through this layer by reactive ion etching (RIE) using S F 6 o r C F 4 gas. Gold pads were then sputtered on the SiO2 layer which were big enough to cover the windows as well as some part of the SiO2 layer. Thus wire bonding could be done on the part of the pads on the SiO2 layer which were electrically connected to the superconductor through the windows. The major drawback of this technique was the formation of an insulating layer on the superconductor during the RIE process as currently implemented. For routine use it has been found most useful to produce the pads by sputtering through a mask or by argon ion etching. The as-sputtered gold pads proved to have low adhesion to the polished surface of the superconductor. Since annealing of the gold contacts was shown to improve their properties 17, a similar approach was followed here. Rapid thermal annealing up to 450°C for 20-60s was not sufficient to improve the adhesion. Longer annealing at 450°C for 5-10 h was necessary for strong adhesion to allow wire bonding. Slow heating (0.5°Cmin ~) was necessary to avoid cracking of the sample when the glue was burning. The oxygen annealing of the sample after thinning it from the bulk ensured that oxygen content was homogeneous throughout the sample. The samples prepared using argon ion etching had to be annealed at 600°C for 10h to ensure sufficient adhesion of the pads to the surface. Gold wires (25#m diameter) were attached to the samples using thermosonic bonding techniques. This involved heating of the sample to 90-100°C. At this step it is crucial that the glue under the sample not be rendered soft by the heating and that there be no air bubbles in it. Two different techniques were available. Ball bonding gives a high strength bond to the gold pads but needs a large pad. Wedge bonding which is more often used for our samples does not require large pad size. However the bond strength is not as high as in the former case. Laser cutting was used to separate the sample into two parts that are connected only by a small bridge across the desired grain boundary. The bridges were 50-100/~m wide and 100-200/~m long. Figure 2 displays an optical micrograph of a sample in a chip carrier, while Figure 3 shows an SEM micrograph of such an isolated grain boundary. The laser used was a pulsed Nd-YAG laser (A = 1.064/~m) operated with a pulse frequency of 5kHz. The average power striking the sample was 0.75 W. The laser beam was focussed to 30#m diameter using a
1994 V o l u m e 34, N u m b e r 4
Techniques for isolation and electrical characterization: G. Schindler et all ~!!i~!! i:i!i! ii
the adhesive under the sample, since it did not appear during the early stages of the process. The damage done to the sample due to the heating is difficult to evaluate. Undisturbed twin boundaries within the grains could be observed as close as lOp~m to the cut, which indicates that there was no massive loss of oxygen from the sample. The thick layer of debris made observation closer to the cut impossible.
Figure 2 Optical micrograph of sample mounted in chip carrier
Figure 3 SEM micrograph of sample showing bond wires and laser cuts
microscope lens. The sample was placed on a computer controlled x-y stage and could be moved continuously under the lens with a speed of 10 mm s-~. Five passes were needed to cut the superconductor completely, so each part of the cut was hit by approximately 75 laser pulses. An SEM picture of a thin polished sample with an isolated structure is shown in Figure 4. The brightness of the interior of the structure is due to charging, which clearly indicates an electrical isolation from the rest of the sample. Debris from the cutting was deposited on the sample in the vicinity of the cut, so it is important to do the wire bonding before cutting the sample. The black colour of the debris probably results from evaporating
Figure 4 SEM picture of a thin polished sample including a structure isolated by laser cutting. The charging of the interior of the structure indicates an electrical isolation from the rest of the sample
Cryogenics 1994 Volume 34, Number 4
The electrical measurements were carried out using the four terminal technique. It is important that the voltage leads are located directly on the grains forming the grain boundary to be investigated. The current leads may also be connected to an adjacent grain. The R(T) measurements were taken using a computer controlled current source (Keithley 224). A scanner allowed the simultaneous measurement of up to four samples with different bias currents. The current voltage characteristics were taken using a current source driven by a ramp generator with a maximum output of _+100mA. The voltage drop across the samples was measured with a Keithley 195-A voltmeter with a 0.1/~V resolution. The samples were measured in a probe inserted in a liquid helium container. The probe contained a coil which allowed the application of magnetic field up to 8 mT. The samples could be mounted either perpendicular or parallel to the field vector. The contact resistance of the samples was measured using the contact configuration shown in Figure 5. This eight terminal technique allows separation of the contributions of the superconductor, the goldsuperconductor interface and the bond wires. Wires labelled A were used for feeding the current to the sample, while wires labelled B were used to pick up the voltage drop within the superconductor. The wires labelled C measured the voltage drop of a larger portion of the superconductor and the contribution of the gold-superconductor interfaces. For T < Tc the contribution of the superconductor is zero and only the interface resistance was measured. To obtain the
Figure 5 Eight-terminal configuration of sample and chip carrier used for determining contact resistance. The pairs of wires are used as follows: A, to provide bias current; B, to measure voltage drop in the sample; C, to measure the contact resistance and the superconductor. The D wires are used in combination with the C wires, to determine the resistance of the bond wires
Techniques for isolation and electrical characterization: G. Schindler et el. 100
100 200 T e m p e r a t u r e (K)
Figure 6 Resistivity of superconductor-gold pad interface temperature. The pad size was 150x 150#m 2 and the bias current 1 mA versus
interface resistance for T > Tc, the signal of the B wires multiplied by a suitable factor to accommodate the effect of the different geometry had to be subtracted. Eventually the resistance of the gold wires could be obtained by subtracting the signals of the C wires from that of the D wires. The observed interface resistance showed nonmetallic behaviour with a steep increase at low temperature (see Figure 6). No deviation from Ohm's law was found at T = 4 K . The interface resistivity was 20/xl) cm 2 at T = 77 K and 50/x11 cm 2 at T = 4 K. For the pad sizes used, this results in a resistance of 0.211 ( T = 7 7 K ) and 0.5-0.611 ( T = 4 K ) , respectively. The bond wires showed metallic behaviour with resistivities of 0.5/zOcm ( T = 7 7 K ) and 0.05/.dlcm ( T = 4 K). At low temperature the wires made only a minor contribution to the total resistance.
Preliminary results The R(T) characteristics showed metallic behaviour and a steep transition with Tc -- 91 K (see Figure 7). No indication of damage to the sample due to polishing or laser cutting could be found here. Most of the grain boundaries showed a negative curvature in their current-voltage (I-V) characteristics at I > I~. They also showed little or no change of their I-V characteristics in applied magnetic fields up to 0.6-
Figure 8 Current-voltage characteristics of sample showing flux pinning behaviour in zero field and with an applied field of 8 mT (T= 85 K). Only a very small change in the I - V characteristics due to the magnetic field could be observed
8roT (see Figure 8). These observations can be explained in terms of flux flow rather than Josephson effect. However, some grain boundaries were found that exhibit the characteristics of a Josephson junction (see Figure 9). The I-V characteristics resembled those of a Josephson junction in the high damping limit (tic ~ 0) 18"19, which accounts for the lack of hysteresis. The lcR N product was found for this junction was 15/zV and was far below the expected value predicted by the Ambegaokar/Baratoff theory 2°. These observations are in accordance with other measurements taken on grain boundaries in polycrystalline superconductors 13. However, a high 'excess current' Iex was observed, so the I-V characteristics approach the line U = RN(I-Ie×) for I>>lc. The critical current was sensitive to very small magnetic fields (see Figure 10). A remarkable drop of Ic could be observed in fields as low as 0.1mT. However, no oscillations of Ic that are typical for homogeneous Josephson junctions could be found. The melt-textured growth samples contain a high concentration of Y2BaCuO5 inclusions. These inclusions make an inhomogeneous current distribution within the grain boundary highly probable and thus account for the lack of oscillations in the It(B) characteristics. The changes from flux pinning to Josephson 100-
~ 0.2 0.1 0
100 200 Temperature (K)
Figure 7 Resistivity v e r s u s temperature of thin polished sample of bulk YBa2CuaO7
0 Voltage (~V)
Figure 9 Current-voltage characteristic of a grain boundary showing Josephson effect (T= 85 K)
Cryogenics 1994 Volume 34, Number 4
Techniques for isolation and electrical characterization: G. Schindler e t 30
thank Professor C. Casey and J. Swartz, Department of Electrical Engineering, Duke University, and G. Simpson of Cree Research, for assistance with bonding, Dr U. Varshney and J. Glass of Arcova for the laser cutting, Dr M. Daeumling of IBM for supply of melttextured specimens and Dr T. Shaw of IBM for useful discussions.
A A A
AA AA AAAAA 0
AA AAA AA A ]
-0.3 -0.2 -0.1 0 0.1 0.2 Magnetic field (naT)
Figure 10 Critical current v e r s u s magnetic field of grain boundary showing Josephson effect (T= 85K). A remarkable decrease of/¢ at very small magnetic fields was found
behaviour are not correlated to different sample preparation or sintering conditions. Both types of grain boundaries could be found in the same sample at opposite sides of the same grain.
Conclusions In summary, we have demonstrated various techniques for the investigation of single grain boundaries in YBa2Cu307_~ bulk ceramics. Preliminary results show that Josephson junctions as well as flux pinning behaviour can be observed, even in the same sample.
Acknowledgements The authors would like to acknowledge that the research was undertaken under the auspices of a grant from the Department of Energy, Basic Energy Sciences. M. Hartsell and M. Andreas made useful contributions to the research. We would like to
292 Cryogenics 1994 Volume 34, Number 4
1 Chaudhari, P., Koch, R.H., Laibowitz, R.B., McGuire, T.R. et al. Phys Rev Lett (1987) 58 2684 2 Peterson, R.L. and Ekin, J.W. Physica C (1989) 157 325 3 Nakahara, S., Fisanick, G.J., Yan, M.F., van Dover, R.B. etal. J Crystal Growth (1987) 85 639 4 Clarke, D.R., Shaw, T.M. and Dimos, D. J Amer Cer Soc (1989) 72 1103 5 Kroeger, D.M., Choudhury, A., Brynestad, J., Williams, R.K. ct al. J Appl Phys (1988) 64 331 6 Worthington, T.K., Gallagher, W.J., Kaiser, D.L., Holtzberg, F.H. et al. Physica C (1988) 153/155 32 7 Dimos, D., Chaudhari, P., Mannhart, J. and LeGoues, F.K. Phys Rev Left (1988) 61 219 8 Dimos, D., Chaudhari, P. and Mannhart, J. Phys Rev B (1990) 41 4038 9 Babcock, S.E., Cai, X.Y., Kaiser, D.L. and Larbalestier, D.C. Nature (1990) 347 167 10 Larbalestier, D.C., Babcock, S.E., Cai, X.Y., Field, M.B. et al. Physica C (1991) 185/189 351 11 Farrell, D.E., Chandrasekhar, B.S., DeGuire, M.R., Fang, M.M. et al. Phys Rev B (1987) 36 4025 12 Jin, S., Tiefel, T.H., Sherwood, R.C., van Dover, R.B. et aL Phys Rev B (1988) 37 7850, 13 Schindler, G., Seebacher, B., Kleiner, R., Miiller, P. et al. Physica C (1992) 196 1 14 Polturak, E., Wilen, L., Cohen, D. and Koren, G. Rev Sci lnstrum (1990) 61 1759 15 Claassen, J.H., Reeves, M.E. and Soulen Jr., R.J. Rev Sci Instrum (1991) 62 996 16 Kittel, C., Fahy S. and Louie, S.G. Phys Rev B (1988) 37 642 17 Ekin, J.W., Larson, T.M., Bergren, N.F., Nelson, A.J. et al. Appl Phys Lett (1988) 52 1819 18 Steward, W.C. Appl Phys Lett (1968) 12 277 19 McCumber, D.E. J Appl Phys (1968) 39 3113 20 Ambegaokar, V. and Baratoff, A. Phys Rev Lett (1963) 10 486 and (1963) 11 104