RAPID COMMUNICATIONS PHYSICAL REVIEW B VOLUME 57, NUMBER 2 1 JANUARY 1998-II Voltage jumps in current-voltage characteristics of Bi2Sr2CaCu2O81d superconducting films: Evidence for flux-flow instability under the influence of self-heating Z. L. Xiao, P. Voss-de Haan, G. Jakob, and H. Adrian Institut für Physik, Universität Mainz, Staudinger Weg 7, D-55099 Mainz, Germany ~Received 8 August 1997! We present the observation of voltage jumps at high flux-flow velocities in current-voltage characteristics of c-axis-oriented Bi2Sr2CaCu2O81d superconducting thin films. The experimental results agree quantitatively with the prediction for viscous flux-flow instability including consideration of a finite heat-removal rate from the samples. We obtained a characteristic magnetic field B T , which separates the regions where the instability is dominated by nonthermal and heating mechanisms. The inelastic scattering rate 1/t in and the diffusion length l e of the quasiparticles have also been extracted. @S0163-1829~98!50302-5# Some 20 years ago Larkin and Ovchinnikov ~LO! predicted a vortex instability resulting in a voltage jump in the current-voltage (I-V) characteristics at high flux-flow velocities.1 The underlying idea is that in the flux-flow region the viscous friction, which a vortex experiences during its movement, is a nonmonotonic function of the velocity * . Above this critical with a maximum at a critical value y LO velocity, the viscous force decreases leading to an even higher velocity accompanied by a still lower viscosity and so on and so forth, giving rise to the above instability. Experimental investigations on this phenomenon have been done both in low-temperature2–5 and high-temperature6–8 superconducting thin films. The LO theory allows to extract the temperature dependence of the inelastic scattering times of quasiparticles. The published experimental results could be summarized as follows: ~1! The voltage jumps are intrinsic to the vortex system and are not artifacts due to thermal runaway; a reasonable agreement between experimental results and LO theory has been found.2–7 ~2! Both the critical current at which the voltage jumps occur2–5,7 and the critical vortex velocity corresponding to the instability3,5,7,8 exhibit significant magnetic field dependence. This behavior is not predicted by LO theory, and the magnetic-field dependence of the critical velocity makes the extracted values for the scattering rate uncertain. ~3! The voltage jumps could appear in the vortex glass or collective creep state which correspond to a negative curvature of the I-V characteristics in doublelogarithmic plots.5,7 Such voltage jumps could be caused by other mechanisms given below, for example, as predicted by the theory of self-organized criticality ~SOC!,9 voltage instabilities could appear near the pinning-depinning transition by thermally activated jumps of vortices which trigger a chain reaction of vortex movements leading to avalanches of diverging size. Recently the magnetic-field dependence of the critical current and the critical voltage has been explained theoretically by Bezuglyj and Shklovskij ~BS! considering the effect of self-heating on the flux-flow instability predicted by Larkin and Ovchinnikov. They have successfully interpreted the experimental data of classic superconducting indium films.10 In this paper we report experimental evidence for the BS theory in high-T c superconductors. Because the instability 0163-1829/98/57~2!/736~4!/$15.00 57 caused by SOC and other mechanisms appears only before or near the flux line depinning, the possible corresponding interpretation could be ruled out if flux-flow behavior V}I is observed before the voltage jump. Due to its low magnetic irreversibility field,11 the Bi2Sr2CaCu2O81d superconductor should be an excellent candidate for the observation of the flux-flow behavior in a large range of temperatures and magnetic fields. In our experiment we have investigated the I-V characteristics of Bi2Sr2CaCu2O81d superconducting films. Both the typical behavior of flux flow and the voltage jumps were observed at lower and higher vortex velocities, respectively. Our experimental data agree quantitatively with the prediction of BS theory for the viscous flux-flow instability influenced by self-heating. Bi2Sr2CaCu2O81d superconducting films of different thicknesses ~typically 400 nm! were deposited by sputtering on @100# oriented SrTiO3 substrates.12 As revealed by x-ray analysis, all films were purely c-axis oriented. Silver contact pads were evaporated and bridges ~length 100 mm, width 10 mm! were patterned by photolithography and wet chemical etching, allowing the measurement of I-V curves using the standard four point method. External magnetic fields up to 6 T, parallel to the c axis of the film could be applied. The measurements were conducted with rectangular current pulses of 1 s length and an interval time of 3 s between pulses, with the direction of the current always perpendicular to the field. For the present work two films were analyzed. The experimental results for both samples agree, thus in the following we concentrate on the sample with superconducting critical temperature ~midpoint! T c 586.0 K and a normalstate resistivity of 66 mV cm at 100 K. Figure 1~a! shows the I-V characteristics at 77 K for different magnetic fields from 0.25 T up to 6 T. In order to examine the vortex state at lower vortex velocities, the data are plotted with a double-logarithmic scale. Within our experimental resolution the I-V characteristics in low magnetic fields exhibit a clearly discontinuous voltage jump at a well defined critical current I * . As an example for the data in a magnetic field of 0.25 T, the definition of the experimentally observed critical current I * and critical voltage V * are indicated by the dotted lines. The dashed line represents the slope for typical flux-flow behavior V}I. Figure 1~a! shows R736 © 1998 The American Physical Society RAPID COMMUNICATIONS 57 VOLTAGE JUMPS IN CURRENT-VOLTAGE . . . R737 FIG. 2. Magnetic-field dependence of ~a! the critical current density J * and ~b! the critical vortex velocity y * at temperatures as assigned to each curve. FIG. 1. ~a! Current-voltage characteristics of a c-axis oriented Bi2Sr2CaCu2O81d bridge at 77 K. The magnetic fields, from lower right to upper left are B50.25, 0.50, 0.75, 1, 1.25, 1.5, 2, 3, 4, 5, and 6 T. The dashed line represents the expected behavior V}I for flux flow at lower vortex velocities. The definitions of the critical current I * and the critical voltage V * are indicated by dotted lines. ~b! The magnetic-field dependence of the flux-flow resistivity in the low current limit. The solid lines are guides to the eye. that all curves at not too high vortex velocities are parallel to the dashed line. In addition the flux-flow resistivity r f as plotted in Fig. 1~b! shows the typical linear dependence on B. This result positively demonstrates that the vortex system is already in the flux-flow state before the voltage jumps occur. Thus for the cause of the voltage jumps we can directly exclude the mechanisms of SOC, hot spot,13 and depinning of the flux lines,14 all of which should occur before or at the flux line depinning, i.e., below the flux-flow regime. Vortex lattice crystallization15 could lead to a jumplike voltage increase in the I-V curves after the flux line depinning with the critical current shifted to a higher value with increasing temperature. However, as shown later, the critical current I * observed in our experiments decreases with increasing temperature. Thus we conclude that the mechanism based on flux-flow instability is most suitable to explain the voltage jumps observed in our samples. According to LO theory a voltage instability in I-V characteristics is expected to occur when the vortex velocity reaches the critical value * 51.02~ D/ t in! 1/2~ 12T/T c ! 1/4 y LO ~1! where D is the quasiparticle diffusion coefficient; t in is the inelastic scattering time of quasiparticles. The critical veloc- * and the critical current I LO * , at which the LO instaity y LO bility occurs, are predicted to be field independent. As shown in Figs. 1 and 2, however the experimentally obtained critical current I * and hence the current density J * ~5I * /S, where S is the area of the bridge’s cross section! depend strongly on the magnetic field at all temperatures. The critical velocity y * ~5V * /BL, where L is the length between the voltage contacts! exhibits significant field dependence at higher temperatures ~e.g., 79 K!, while at lower temperatures ~e.g., 65 K! this dependence seems to disappear. These results are consistent with observations both in low-T c superconducting Al,2,3 In,3 Sn,2,3 and Ta/Ge ~Ref. 5! films, and high-T c superconducting YBa2Cu3O72d thin films.7,8 Recently it has been shown in a theoretical work by Bezuglyj and Shklovskij ~BS! that such a magnetic-field dependence could be caused by the unavoidable heating of quasiparticles due to the finite rate of removing the power dissipated in the sample.10 According to BS theory, the quasiparticles should have a higher temperature T * than the experimentally measured temperature T ~bath temperature!. In an extension of the LO approach considering the finite heat removal rate, the magneticfield dependence of the critical current density J * and critical electric field E * (E * 5V * /L5 y * B) are given by J* ~ 3t21 ! 1/2 52&t 3/4 , J0 ~ 3t11 ! ~2! E * ~ 12t ! ~ 3t11 ! 5 . E 0 2&t 3/4 ~ 3t21 ! 1/2 ~3! with t5 @ 11b1(b 2 18b14) 1/2# / @ 3(112b) # and b 5B/B T . The introduced normalizing current density J 0 and electric field E 0 are RAPID COMMUNICATIONS R738 Z. L. XIAO, P. VOSS-DE HAAN, G. JAKOB, AND H. ADRIAN J 0 52.62s n /e 0 ~ D t in! 21/2k B T c ~ 12T/T c ! 3/4, ~4! E 0 51.02B T ~ D/ t in! 1/2~ 12T/T c ! 1/4, ~5! 57 which are independent of the magnetic field. Here s n is the conductivity in normal state, e 0 is the electron charge, and k B is the Boltzmann constant. The characteristic magnetic field B T depends on the heat transfer coefficient h and the time t in of quasiparticle energy relaxation B T 50.374k 21 B e 0 R h h t in , ~6! where R h 5( s n d) 21 is the sheet resistance and d is the thickness of the film. Furthermore, the ratio of the critical electric fields with and without the influence of the quasipar* 5t 1/4(3t21) 1/2/&, where ticle heating is found as E * /E LO * 5 y LO * B. It follows that, for B!B T , corresponding to t E LO '1, the heating effect is not important and the critical electric field E * is approximately equal to the predicted theoret* . In the other limit of B@B T , t reaches its ical value E LO minimum of 1/3; hence the heating effect dominates and the * . Thereexperimentally observed E * is far smaller than E LO fore the parameter B T separates the regions where the influence of self-heating is of minor and of major importance for the flux-flow instability. As shown above, we obtained experimentally the critical current density J * and the critical electric field E * from the instability point as defined in Fig. 1~a!. Equations ~2! and ~3! are the parametric form of the corresponding E * and J * taken at different magnetic fields at a fixed bath temperature T. For suitably chosen J 0 and E 0 at the different temperatures, the experimental data should collapse on a common curve. The corresponding results are given in Fig. 3~a!. The symbols represent the experimental data shown in Fig. 2 and the solid line is calculated from Eqs. ~2! and ~3!. Rather good agreement between theory and our experimental data could be found. The determined J 0 and E 0 , both of which decrease with increasing temperature, are given in the inset of Fig. 3~a!. Following the approach of Bezuglyj and Shklovskij, the inelastic quasiparticle scattering rate 1/t in could be extracted from Eqs. ~4! and ~5! using the obtained J 0 , E 0 in combination with B T , which can be obtained by fitting the measured J * (B) at fixed temperatures to the theoretical curve Eq. ~2!. From Eqs. ~2! and ~3! the magnetic-field dependence of the dissipated power could be expressed as J * E * /J 0 E 0 5(1 2t). Thus B T can be extracted through less complicated fitting. The related results are shown in Fig. 3~b!. The magnetic field B T exhibits a strong temperature dependence varying from 5 T at 65 K to 0.4 T at 79 K. In our experiments the values of b5B/B T fall in the range between 0.25 and 2. The rate of the critical electric field with and without thermal * varies from 0.82 ~0.1 T! to 0.47 ~0.75 T! at influence E * /E LO 79 K, whereas at 65 K it changes from 0.71 ~2.5 T! to 0.55 ~6 T!. With these results one can also easily explain why the magnetic-field dependence of the critical velocity y * at 79 K is stronger than at 65 K as shown in Fig. 2~b!. As indicated above, the measured E * , especially obtained in low magnetic fields, is not too far away from the expected critical * of LO theory. In fact, the extracted inelastic quavalue E LO siparticle scattering time t in of indium films differs by about FIG. 3. Scaling of the data in Fig. 2 according to Eqs. ~2! and ~3!. The solid lines are the theoretical curves and the symbols ~identical to Fig. 2! are experimental data. The temperature dependence of the related parameters E 0 , J 0 , and B T is given in the insets of ~a! and ~b!, respectively. a factor of 2 with and without consideration of the heating effect;10 in other systems reasonable critical velocities and inelastic quasiparticle scattering rates 1/t in were obtained without considering heating effect.3–7 The corresponding inelastic quasiparticle scattering rate 1/t in for different temperatures is given in Fig. 4. Both the inelastic quasiparticle scattering rate 1/t in itself and its temperature dependence are comparable to those reported for YBa2Cu3O72d thin films.6,7 The strong temperature depen- FIG. 4. Temperature dependence of inelastic quasiparticle scattering rate 1/t in and diffusion length l e 5(D t in) 1/2 extracted from Eqs. ~4! and ~5! using the value of J 0 , E 0 , and B T obtained in Fig. 3. RAPID COMMUNICATIONS 57 VOLTAGE JUMPS IN CURRENT-VOLTAGE . . . dence of the inelastic quasiparticle scattering rate 1/t in is most probably due to the spin-fluctuation phenomenon.16 The diffusion length l e 5(D t in) 1/2, which plays an important role in this theory, must be larger than the core size, in order for the excitations to leave the core,10 and larger than the intervortex distance to get spatial homogeneity.8 The calculated values of the diffusion length from Eq. ~4! using the obtained J 0 are presented also in Fig. 4. Clearly l e is larger than not only the vortex core size ~typically 2 j 0 55 nm! in Bi2Sr2CaCu2O81d superconducting films17 but also larger than the intervortex distance ~about 50 nm in a magnetic field of 1 T!. Thus we can conclude that, in the range of magnetic fields and temperatures we have studied, the nonequilibrium distribution of the quasiparticles extends over the whole volume of the superconductor. The heat transfer coefficient obtained from Eq. ~6! varies from 100 W/cm2 K at 79 K to 130 W/cm2 K at 65 K. This value is lower than the typical heat transfer coefficient 103 W/cm2 K of the film-substrate boundary.18 As shown in Ref. 18, the temperature rise in the film can also be caused from the heat flow in the substrate. The temperature difference between the surface of the substrate and the bath is given as:18 DT s 5 a Pt 0 / $ 2L(Dt 0 ) 1/2@ 4(Dt 0 ) 1/21W # c p % , where a '4/p 1/2, P is the dissipated power in the film, t 0 is the duration of the current pulse, D is the heat diffusion 1 A. I. Larkin and Yu. N. Ovchinnikov, Sov. Phys. JETP 41, 960 ~1976!. 2 L. E. Musienko, I. M. Dmitrenko, and V. G. Volotskaya, JETP Lett. 31, 567 ~1980!. 3 W. Klein, R. P. Huebener, S. Gauss, and J. Parisi, J. Low Temp. Phys. 61, 413 ~1985!. 4 A. V. Samoilov et al., Phys. Rev. Lett. 75, 4118 ~1995!. 5 B. J. Ruck et al., Phys. Rev. Lett. 78, 3378 ~1997!. 6 S. G. Doettinger et al., Phys. Rev. Lett. 73, 1691 ~1994!. 7 Z. L. Xiao and P. Ziemann, Phys. Rev. B 53, 15 265 ~1996!. 8 S. G. Doettinger, R. P. Huebener, and A. Kühle, Physica C 251, 285 ~1995!. 9 O. Pla and F. Nori, Phys. Rev. Lett. 67, 919 ~1991!; L. Legrand et al., Europhys. Lett. 34, 287 ~1996!; Z. L. Xiao and P. Ziemann, Physica C 282-287, 2363 ~1997!. R739 constant of the substrate, and L and W are the length and width of the bridge, respectively. Using c p 51 J/cm3 K, D 50.18 cm2/s ~Ref. 18! and the values of W(510 m m), L (5100 m m) and t 0 (51 s) we derived a heat transfer coefficient h s @ 5 P/(LWDT) # of the SrTiO3 substrate of about 500 W/cm2 K. Thus the effective heat transfer coefficient from film to bath is about 300 W/cm2 K, which is reasonably consistent with our experimental value. The disparity may be due to an underestimated E 0 arising from inhomogeneities in the film.10 In conclusion, we have attributed to the observation of voltage instability in current-voltage characteristics at high flux-flow velocities in high-T c superconducting films of Bi2Sr2CaCu2O81d. A quantitative agreement between our data and the theory for viscous flux flow under the influence of power dissipation during the vortex motion was found. Having integrated the effects of the unavoidable quasiparticle heating into the quantitative analysis of the experimental data, we hope that our results will stimulate further investigation on the viscous flux-flow instability. We would like to thank Th. Kluge and P. Haibach for experimental advice and support. The project was financially supported by the SFB 262. A. I. Bezuglyj and V. A. Shklovskij, Physica C 202, 234 ~1992!. K. Karada et al., Phys. Rev. Lett. 71, 3371 ~1993!. 12 P. Wagner et al., Physica C 215, 123 ~1993!. 13 W. J. Skocpol, M. R. Beasley, and M. Tinkham, J. Low Temp. Phys. 16, 145 ~1974!; A. Vl. Gurevich and R. G. Mints, Rev. Mod. Phys. 59, 941 ~1987!. 14 W. Henderson, E. Y. Andrei, M. J. Higgins, and S. Bhattacharya, Phys. Rev. Lett. 77, 2077 ~1996!. 15 A. E. Koshelev and V. M. Vinokur, Phys. Rev. Lett. 73, 3580 ~1994!. 16 S. M. Quinlan, D. J. Scalapino, and N. Bulut, Phys. Rev. B 49, 1470 ~1994!. 17 F. Gollnik et al., Physica C 235-240, 1933 ~1994!. 18 S. K. Gupta et al., Physica C 206, 335 ~1993! and references therein. 10 11