Supercond. Sci. Technol. 13 (2000) 356–361. Printed in the UK PII: S0953-2048(00)09495-1 Current distribution and ac susceptibility response of a thin superconducting disc in an axial field: a theoretical approach S A Aruna, P Zhang, F Y Lin, S Y Ding and X X Yao Department of Physics and National Laboratory of Solid State Microstructures, Centre for Advanced Studies in Science and Technology of Microstructures, Nanjing University, Nanjing 210093, People’s Republic of China Received 12 November 1999 Abstract. Within the framework of the thermally activated process of the flux line or flux line bundles, and by time integration of the 1D equation of motion of the circulating current density J(ρ, t), which is suitable for thin superconducting films (R d, λ), we present numerical calculations of the current profiles, magnetization hysteresis loops and ac susceptibility χn = χn + iχn for n = 1, 3 and 5 of a thin disc immersed in an axial time-dependent external magnetic field Ba (t) = Bdc + Bac cos(2π νt). Our calculated results are compared with those of the critical state model (CSM) and found to prove the approximate validity of the CSM below the irreversibility field. The differences between our computed results and those of the CSM are also discussed. 1. Introduction Since the discovery of high temperature superconductors (HTSCs), most magnetic data have been taken on thin films in the form of discs, slabs or strips with the applied magnetic field perpendicular to the film surface. To obtain maximum signal, most magnetization experiments are often performed with thin specimens in a perpendicular field [2], because in this orientation, even weak magnetic fields can penetrate the film due to large demagnetization effects. One may account for these geometry effects by introducing a demagnetization factor. Sun et al [3] assumed the Bean critical state model to describe the magnetic response of a thin disc and obtained the hysteristic moment and temperature dependent ac susceptibility. An exact critical state model for thin discs in transverse field was recently presented by Mikheenko and Kuzovlev [4]. They took into account the current distribution flowing in the Meissner state (vortex-free region, Bz = 0) in addition to the current density |J | = Jc , which is assumed constant and flows in the vortex penetrated region and provided analytical solutions for the magnetic response of the disc. This model was extended by Zhu et al [5] who took into account the fact that the current density should not change abruptly from the vortex-penetrated region to the vortex-free region, i.e.|J | Jc . Based on these two models, Clem et al [1] gave a complete description of ac susceptibility calculations for curves of χ ∼ Bac in the critical state model, 0953-2048/00/040356+06$30.00 © 2000 IOP Publishing Ltd where Bac is the amplitude of the ac applied perpendicular field. For ac magnetic response, Gurevich and Brandt [6] pointed out that experimental results are usually analysed within the framework of two opposite theoretical approaches. The first one is based on the CSM below the irreversibility line and the other is the linear flux dynamics above the irreversibility line, where the current–voltage characteristic is ohmic [7]. According to them, the CSM completely neglects the dissipative processes in the critical or subcritical state J Jc , where Jc (T , B) is assumed to be independent of the electric field E. Thus the CSM can cause some inaccuracy in describing the ac magnetic response of thin superconducting films. This situation was changed after a series of articles [8–10] were written by Brandt, in which he developed a general formulation governing the behaviour of thin superconducting films in axial field by deriving the equation of motion for the sheet current using the E = Ec (J /Jc )n material law equation with finite creep exponent n. Most of the works by Brandt treated a finite thickness of cylinder samples where the applied field is parallel to the axis of the cylinder. Evidently, this can also be used in the case of superconducting thin films. Hence, based on this general formulation suggested by Brandt, it is possible to deal with the problem of the extent to which the CSM is approximately valid to analyse the experimental data for ac magnetic responses. In this paper, based on the general formulation by Brandt for the sheet current in thin superconducting films, we attempt Thin disc in axial field to study the flux dynamics by numerical time integration of the 1D integral equation of motion of the circulating current density J(ρ, t) using the nonlinear E(J ) electric field material law E = E(J )J/|J |. In HTSCs, the CSM is only assumed to remain approximately valid well below the irreversibility field. Based on the CSM, [1] calculated the field and current distributions, ac susceptibility components for n = 1, 3, 5 and the magnetization hysteresis loops as a function of the ac field amplitude. Similarly, although temperaturedependent ac susceptibility responses are usually calculated, we calculate the ac response in the flux creep regime as a function of the ac field amplitude Bac for the purpose of comparison and verification of the CSM below the irreversibility line. 2. Theoretical model; equation of motion of the sheet current We consider a thin disc immersed in an applied timedependent magnetic field Ba (t) = Bdc + Bac cos(2π νt) normal to the film surface, where Bac is the ac field amplitude and Bdc is a constant dc biased field situated at the origin. Evidently, shielding currents with current density J are developed and flow in the plane of the disc in a circumferential direction J = J eφ . The total vector potential due to current and applied field as given by [2] is A(ρ, t) = R 0 G(ρ, ρ )J (ρ , t)d dρ + ρ Ba (t) 2 (1) where d is the thickness of the film and G(ρ, ρ ) is the Green function containing information about the specimen shape and is defined as ρ cos φ dφ µ0 π G(ρ, ρ ) = . (2) 2 2π 0 [ρ + ρ 2 − 2ρρ cos φ ]1/2 For the purpose of numerical calculations, we divide the disc into N concentric loops of width R/N spaced equidistantly and then approximate the average current density as J (i) flowing in each loop with outer radius iR/N, inner radius (i− 1)R/N and average radius (i − 1/2)R/N [11]. Expressing the induction law × E = −Ḃ = − ×Ȧ in the form E = −Ȧ gives Ȧ = −E(J ), and the discretized form of equation (1) becomes −E(ρi , t) = N ρj G(ρi , ρj )ρ J˙(ρj , t)d + Ḃa (t) 2 j =1 i = 1, 2, . . . , N. G(ρi , ρj )G−1 (ρj , ρk )(ρ)2 = δik where the prefactor Ec = BV0 and V0 is the flux creep velocity. Next we introduce the current density J -dependent activation energy µ Uc Jc U (J ) = −1 . (7) µ J Equation (7) is a popular theoretical power law U (J ) relationship with much experimental support [12–14] that describes the commonly used thermally activated processes, where Uc is a characteristic energy scale and Jc is the critical current density in the absence of thermal fluctuations. The glassy exponent µ is a universal constant in the vortex-glass theory with a value of about unity. In the collective-creep theory, µ = 1/7 for the case of single vortex creep (J ∼ Jc ); µ = 3/2 for small bundle creep (J < Jc ) and µ = 7/9 in the large bundle pinning regime (J Jc ) [15]. Besides, equation (7) becomes U (J ) = Uc ln( JJc ) as µ → 0 and U (J ) = Uc (1 − JJc ) for µ = −1 (Anderson–Kim model). The electric field materials law therefore takes the form µ Uc J Jc E = Ec exp − −1 (8) Jc T J where the extra factor J /Jc is introduced to provide a gradual cross-over to the viscous flux flow regime [16, 17]. We can normalize equation (4) by introducing , J˜ = JJc , dimensionless quantities: ρ̃ = Rρ , ρ̃ = ρ R Jc Rd Ẽ = EEc , B̃ = BB0 and t˜ = tt0 where t0 = µ0 E and c B0 = µ0 Jc d. Here, J˜ is in unit of Jc , B̃ in unit B0 = µ0 Jc d RJc d and t˜ in unit t0 = µ0 E . Equation (4) can be therefore c rewritten as N ρ̃j ∂ B̃(t˜) ∂ J˜(ρ̃i , t˜) −1 ˜ G̃ (ρ̃i , ρ̃j ) − Ẽ(ρ̃j , t ) − = 2 ∂ t˜ ∂ t˜ j =1 where (3) (4) (5) with ρ = R/N . In the flux creep regime, the material law equation E(J ) becomes strongly nonlinear and can be generally described by U (J ) E(J ) = Ec exp − . (6) T i = 1, 2, . . . , N In the general case of nonlinear E(J ) and arbitrary sweep rate Ḃa (t), the time integration of equation (3) has to be performed numerically, hence the time derivative should be moved out from the integral to obtain J˙i (t) as an explicit functional of J and Ḃa (t) [12]. Thus, inverting equation (3) yields the following linear differential equation N ρj ˙ i , t)d = J,(ρ G−1 (ρi , ρj )ρ −E(ρj , t) − Ḃa (t) 2 j =1 i = 1, 2. . . . , N where G−1 (ρi , ρj ) is related to G(ρ, ρ ) through the delta function G̃−1 (ρ̃i , ρ̃j ) = G−1 (ρi , ρj )µ0 Rρ. (9) (10) Equation (9) completely describes the magnetic response of thin superconducting films for some concrete cases of Ba (t). For Ḃa (t) = 0, the last term vanishes and we obtain the relaxation data if a critical state of Jc (ρ) = Jc (Ba ) = constant is assumed for a biased Ba . If Ḃa (t) = constant and the initial conditions are zero, then we can obtain the initial magnetization of the sample, but in this case, the self-field of the current density must be calculated too. For the case of ac magnetic response, we have Ba (t) = Bdc + Bac cos(2πνt), where ν is normalized as ν̃ = νν0 with 357 S A Aruna et al c ν0 = µ 0 E . Substituting equation (8) into equation (9), we RJc d obtain N ∂ J˜(ρ̃i , ν̃ t˜) G̃−1 (ρ̃i , ρ̃j ) = ∂(ν̃ t˜) j =1 µ ˜ −J (ρ̃j , ν̃ t˜) Uc 1 −1 exp − × ν̃ µT J˜ ρ̃j + 2π B˜ac sin(2π ν̃ t˜) (11) 2 Bdc = 0.5 T, the authors obtained the value of the parameter µ = 0.62. Studies [18,19] have shown that at low fields (Bdc = 0.5 T) in the vortex-glass-collective-creep model, the Jc (T ) and Uc (T ) dependences are Hereafter, where G̃−1 (ρi , ρj ) = G−1 (ρi , ρj ) µ0NR . neglecting the tildes over the dimensionless quantities in equation (11), the equation of motion for the sheet current becomes N ∂J (ρi , νt) G−1 (ρi , ρj ) = ∂(νt) j =1 µ −J (ρj , νt) Uc 1 −1 × exp − ν µT J ρj + 2π Bac sin(2πνt) i = 1, 2, . . . , N (12) 2 which can be expressed in matrix form as respectively, where Jc (0) = 9.06×1010 A m−2 , the activation energy Uc (0) = 338 K, the flux creep velocity V0 = 10 m s−1 [18] and the reduced temperature t = T /Tc . In [18], although experimental data for the parameters appearing in the material law equation (7) are readily available for values of Bdc in the range 0.5 to 6.0 T, we shall only make use of the necessary data as given above at Bdc = 0.5 T for our numerical calculations. For the other quantities, we chose typical values for the disc radius R = 2.5×10−3 m, thickness d = 2.0 × 10−7 m and N = 48. Since we intend to verify the approximate validity of the critical state model in calculating the ac susceptibility dependence on ac magnetic field amplitudes Bac , we chose an arbitrary fixed temperature T = 60 K well below the irreversibility line, in which case thermally activated flux creep can considerably affect the macroscopic electrodynamics of these materials [7, 20] where the flux creep is described by equation (7). From equation (15), with Tc = 90 K and T = 60 K, we obtained the temperature dependent critical current density Jc (T = 60 K) = 1.73 × 1010 A m−2 and the corresponding activation energy barrier Uc (T = 60 K) = 271.2 K. Note that ν0 = µ0 RJEcc(T )d has different values for different temperatures since Jc (T ) is temperature dependent, thus in our case, we have ν0 (T = 60 K) = 4.59 × 105 Hz. Equation (12) can then be readily integrated by the Runge– Kutta method starting with Bac (t) = 0 and the known vector Ji (0) at time t = 0, and then increasing Bac (t) gradually to obtain the solution J (ρ, t). Note that this numerical integration is stable if the time step dt is sufficiently small. The magnetic moment m of the disc in dimensionless quantities in transverse magnetic field has only a z component, which can be calculated from 2 ∂J (ρi , νt) −1 G (ρi , ρj )F (j ) = ∂(νt) j where F (j ) is given by F (j ) = −J (ρj , νt) exp ν − Uc µT (13) µ 1 −1 J ρj (14) 2π Bac sin(2πνt). 2 We shall numerically time integrate this 1D integral equation (equation (12)) for the sheet current to obtain the ac response and then check the extent to which the critical state model is valid below the irreversibility line. + 3. Numerical calculations and discussion We have performed computations of flux creep in a uniform magnetic dc field Bdc , with an ac field Bac (t) = Bac cos(2π νt) superimposed by numerical time integration of equation (12). If the value of Bdc is large enough compared with the self-field produced by shielding currents, then the self-field and the ac field can be neglected and we can approximate the field to be just equal to the dc field Bdc . In order to calculate the ac susceptibility from equation (12), the values of R and N must be known, and by combining equations (2), (5) and (10), the quantity G̃(ρ̃i , ρ̃j ) can be explicitly obtained. Meanwhile, the quantities µ, Uc /T , Jc , Bdc , Bac and ν appearing in equation (12) must be given too. Also, to obtain the relationship between the practical and normalized quantities Bac , ν, t, ρ, ρ and J , it is necessary to know the values of the parameters in the expressions defining the dimensionless quantities. As an example, we shall make direct use of the data from [18] for the above mentioned physical parameters in equation (7), which have been obtained by the authors to fit their experimental results for a YBa2 Cu3 O7 thin film whose critical temperature Tc = 90 K. According to [18], for 358 Jc (T ) = (1 + t 2 )−1/2 (1 − t 2 )5/2 Jc (0) (15) Uc (T ) = Uc (0)[1 − t 4 ] (16) and m=π 1 ρ 2 J (ρ) dρ. (17) 0 From the periodic time-dependent magnetization Mz (t), the ac susceptibility components χn and χn can be computed from 2π 1 Mz (t) cos(2πnνt) d(2πνt) (18) χn (Bac ) = π Bac 0 and χn (Bac ) = 1 πBac 2π Mz (t) sin(2π nνt) d(2π νt). (19) 0 Below, we shall present our typical calculation results followed by discussions. Thin disc in axial field In the field sweeping process, the electric field E induced at the sample perimeter is given in [22] as 0.4 χ''1/χ0 0.2 χ'1/χ0 & χ''1/χ0 0.0 χ'1/χ0 -0.2 -0.4 -0.6 Bdc=0.5 T -0.8 f =5000 Hz Peak: χ''1=0.3195 χ0 -1.0 -1.2 0.1 1 10 Bac (a) 4 χ"3/χ0 χ'3/χ0 χ'3/χ0 & χ''3/χ0 2 0 -2 -4 -6 Bdc=0.5 T -8 f= 5 kHz -10 -12 0.01 0.1 1 10 Bac (b) 0.010 χ'5/χ0 0.005 χ'5/χ''0 & χ''5/χ0 Shown in figure 1(a) are the normalized real (χ1 /χ0 ) and imaginary (χ1 /χ0 ) parts of the ac susceptibility versus Bac at frequency f = 5000 Hz under Bdc = 0.5 T, where χ0 is the real part of the ac susceptibility of fundamental frequency response when Bac is small enough. The sharp decrease in χ1 is a consequence of diamagnetic shielding (inductive response), and the values of χ1 correspond to energy dissipation in the sample (resistive response). The peak value in χ1 is the point of maximum energy dissipation. Our numerical calculations give this peak value χ1 = 0.3195χ0 , which is larger than that calculated by Clem et al in [1] based on the CSM. It is a well known fact that the nonlinear response of the superconductor when flux pinning is present is reflected in the presence of higher harmonics in the voltage wave form and the appearance of odd susceptibility coefficients (n = 3, 5, . . .). All the even harmonics are zero due to the symmetry of the magnetization hysteresis loops. Figures 1(b) and 1(c) show plots of the normalized real and imaginary parts of the ac susceptibility components for third (n = 3) and fifth (n = 5) harmonics respectively. Further comparing our calculation results with those of the CSM, we observed that the general features for the ac susceptibility for n = 1, 3 and 5 are similar to those obtained in the CSM. The differences are only quantitative in nature. For instance, our results give the peak in χ1 at χ1 = 0.3195χ0 with a corresponding value of χ1 = −0.3977χ0 , while χ1 = 0.241χ0 and χ1 = −0.382χ0 as calculated in [1] based on the CSM. We also see that our calculation results for χ3 and χ5 are very similar in appearance but with the minimum in χ3 occurring at Bac < 1 while that in χ5 appears at Bac > 1. In contrast to our results, the minima in both χ3 and χ5 in [1] appear at Bac > 1. Note that the normalized value in [1] is Bd = µ0 Jc d/2, while in our case it is B0 = µ0 Jc d, thus leading to a difference by a factor of 2 in comparing the corresponding values of Bac , which has already been checked. Another major difference is the extra wiggle appearing in χ5 near Bac = 1 in the CSM [1], which is not evident in our result. Our calculation results give a continuous curve at the frequency f = 5000 Hz. Now we consider the frequency effect of ac susceptibility. Figure 2 is a plot of (χ1 /χ0 ) and (χ1 /χ0 ) dependences on Bac at three frequencies f = 100 kHz, 5000 Hz, and 250 Hz under Bdc = 0.5 T. There are two distinguished characters in these curves. First, both the peak in χ1 and the step in χ1 shift to lower Bac with decreasing frequency. Secondly, the width of the transition in χ1 broadens as the frequency decreases in accordance with experimental results. Our results are evidently beyond those of the critical state model. However, Van der Beek and Kes [21] pointed out that for nonlinear ac magnetic responses, then one assumes that J is not Jc , but rather determined by the actual form of U (J ), in other words, J is a function of the field sweep rate J = J (Ḃ), which is expressed via the thermally activated energy barrier U (J ) as follows Ec U [J (Ḃ)] = T ln . (20) E χ''5/χ0 0.000 -0.005 -0.010 Bdc=0.5T -0.015 f =5 kHz 0.1 1 10 Bac (c) (χ1 ) and imaginary (χ1 ) parts of n = 1 (fundamental-frequency) complex ac susceptibility components normalized to χ0 , as a function of ac field amplitude Bac = Bdc /B0 where B0 = µ0 Jc d at f = 5000 Hz and Bdc = 0.5 T. A peak appears at χ1 = 0.3195χ0 with a corresponding χ1 = −0.3977χ0 . (b) Real (χ3 ) and imaginary (χ3 ) parts of n = 3 (third harmonic) complex susceptibility components normalized to χ0 , as a function of ac field amplitude Bac as in (a) at f = 5000 Hz and Bdc = 0.5 T. (c) Real (χ5 ) and imaginary (χ5 ) parts of n = 5 (fifth harmonic) complex susceptibility components normalized to χ0 , as a function of the ac field amplitude Bac at f = 5000 Hz and Bdc = 0.5 T. Figure 1. (a) Real 359 S A Aruna et al 0.4 0.0 χ''1/χ0 0.2 -0.1 a 0.0 J/Jc χ'1/χ0 & χ''1/χ0 -0.2 χ'1/χ0 -0.2 c -0.4 b b c -0.3 a d -0.6 Bdc= 0.5 T a f =100 kHz b f =5 kHz c f =250 HZ -0.8 -1.0 -0.5 -1.2 0.1 1 e -0.4 -0.6 0.0 10 0.2 0.4 Bac 0.6 0.8 1.0 ρ /R (a) Figure 2. Normalized ac susceptibility components at fundamental frequency (n = 1) under Bdc = 0.5 T at frequencies (a) f = 100 kHz, (b) 5000 Hz, (c) 250 Hz. 0.6 0.4 0.6 J/Jc 0.2 0.4 f e d c b a 0.0 -0.2 0.2 g Mz(t)/Msat -0.4 0.0 -0.6 0.0 0.2 0.4 0.6 0.8 1.0 ρ /R -0.2 a b (b) c d e -0.4 -0.6 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 Figure 4. (a) Normalized azimuthal current density distributions (J /Jc ) against radius ρ/R during one cycle in the disc in increasing magnetic field, Bac (t): (a) 0.0499; (b) 0.2455; (c) 0.4669; (d) 0.6426; and (e) 0.7943. (b) Normalized azimuthal current density distributions (J /Jc ) against radius ρ/R during one cycle in the disc in decreasing magnetic field, Bac (t): (a) 0.7943; (b) 0.6425; (c) 0.2455; (d) 0.1329; (e) −0.2454; (f) −0.4669; (g) −0.7943. B'ac Figure 3. Magnetization Mz (t) hysteresis loops normalized to the saturation magnetization Msat = π/3 against the field Bac = Bac (t) − Bdc at different ac field amplitudes: (a) 0.1019; (b) 0.2033; (c) 0.4019; (d) 0.5074; (e) 0.6342. E(R) = (R/2)(dBac /dt). If we approximate E in the flux penetration region as E(R), then we can actually determine J (Ḃ) from equation (20). Our calculation results therefore verify that the prediction of Van der Beek and Kes is a much better approximation than the original Bean critical state model, because in our case, the energy dissipation must be taken into account since Uc (T )/T = 4.5 for T = 60 K. This is a main conclusion of this paper, because according to Brandt, if Uc /T (or n) is large enough (e.g. n = 51), then the CSM can be approximated. A detailed discussion of this problem is beyond the scope of this paper. Here, we only emphasize that in order to compare our calculated results of the frequency effect of ac susceptibility with those of the CSM, it is necessary to use J (Ḃ) instead of Jc as suggested by Van der Beek and Kes [21]. Detailed calculation results 360 of this problem and its comparison with the above model will be considered elsewhere. When an ac magnetic field is applied to the sample surface, the sample is taken through a complete hysteresis loop. Shown in figure 3 is a plot of the magnetization hysteresis loops normalized to the saturation magnetization Msat = π/3 versus Bac = Ba (t) − Bdc at different field amplitudes when the applied time-dependent field completes one cycle. Here, the magnetization depends on Bac , thus for small Bac values, the corresponding current distributions are small. We clearly see that different ac field amplitudes Bac imply different hysteresis loops. In other words, different magnetizations give different J , hence from our calculations, the magnetization Mz (t) depends on the field sweep rate. This calculation verifies that the current J is actually field sweep rate dependent, i.e. J = J (Ḃ) where Ḃ = 2π νBac . Each current distribution corresponds to a different hysteresis loop as shown in figure 3. Figure 4(a) is a plot of the normalized current distribution J /Jc versus ρ/R for increasing field Bac (t) values. The weak applied magnetic field Bac (t) = 0.0499 simply induces Thin disc in axial field azimuthal screening supercurrents which flow in circular patterns around the disc with virtually no flux penetration as shown by curve a in figure 4(a). However, as the magnitude of Bac (t) increases with time (Bac (t) = 0.7943), the current penetrates deeper into the sample as shown by curve e in figure 4(a). The corresponding current distribution for one cycle in decreasing field is shown in figure 4(b), where vortices penetrate in the opposite direction at Bac (t) = −0.7943. Note that curve e in figure 4(a) is a mirror reflection of the current distribution (f) in figure 4(b) for one cycle with the same magnitude but opposite in sign. It is clearly evident from this figure that J is smaller than Jc in the vortex penetration region (see figure 4(b), curve f). Since our calculated Uc /T is only about 4.5, this value is not sufficiently large enough to satisfy the condition for using the Bean critical state model. Therefore, our calculation results simply show that completely neglecting the energy dissipation in the Bean CSM cannot provide accurate results for ac magnetic responses, especially at different ac field frequencies. 4. Summary and conclusion We have considered a thin superconducting disc in a uniform perpendicular dc magnetic field Bdc superimposed with an ac field component Bac cos 2π νt. We successfully invoked the general formulation by Brandt for the equation of motion for the azimuthal sheet current in a thin film disc, and used the vortex-glass-collective-creep model U (J ) = (Uc /µ)[(Jc /J )µ − 1] (where µ = 0.62) to numerically time integrate the 1D equation of motion, from which we obtained the current distribution, magnetization curves and the ac susceptibility components (n = 1, 3 and 5) using experimental data for YBa2 Cu3 O7 thin film [18]. The differences between our calculated results and those of the CSM [1] are only quantitative in nature. We know that ac susceptibility response from the Bean CSM is independent of frequency. If we consider the collective-creep model, then from our calculations, the ac response is actually frequency dependent and the Bean CSM is not accurate in describing this property (i.e. flux-creep model). From the arguments of Van der Beek and Kes, if one uses J (Ḃ) instead of Jc in the flux penetration region, the frequency-dependent ac response of thin films can still be explained by this method even if the energy dissipations due to thermally activated flux diffusion are considered. Acknowledgments This work was jointly supported by the National Centre for Research and Development on Superconductivity of China under grant No AJ-4012 and by the Chinese National Science Foundation for Doctoral Education. References [1] Clem J R and Sanchez A 1994 Phys. Rev. B 50 9355 [2] Brandt E H 1996 Phys. Rev. B 54 4347 [3] Sun J Z, Shcaren M J, Bourne L C and Shrieffer J R 1991 Phys. Rev. B 44 5275 [4] Mikheenko P N and Kuzovlev Yu E 1993 Physica C 204 229 [5] Zhu J, Mester J, Lockhart J and Turneaure J 1993 Physica C 212 216 [6] Gurevich A and Brandt E H 1997 Phys. Rev. B 55 12 706 [7] Blatter G, Feigelamn M V, Geshkenbein V B, Larkin A I and Vinokur V M 1995 Rev. Mod. Phys. 66 1125 [8] Brandt E H 1994 Phys. Rev. B 49 9024 [9] Brandt E H 1994 Phys. Rev. B 50 4034 [10] Brandt E H 1994 Phys. Rev. B 50 13 833 [11] Xu X N, Sun A M, Qin M J, Ding S Y, Jin X, Yao X X and Yan S L 1997 Physica C 291 315 [12] Brandt E H 1998 Phys. Rev. B 58 6506 [13] Kung P J, Marley M P, McHenry M E, Willis J O and Coulter Y 1992 Phys. Rev. B 46, 6427 [14] Ossandon J G, Thompson J R, Christen D K, Sales B C, Sun V R and Ley K W 1992 Phys. Rev. B 46 3050 [15] Sun Y R, Thompson J R, Chen Y J, Christen D K and Goral A 1993 Phys. Rev. B 47 14 481 [16] Feigel’man M V, Geshkenbein V B and Larkin A I 1990 Physica C 167 177 Feigel’man M V, Geshkenbein V B and Vinokur V M 1991 Phys. Rev. B 43 [17] Fisher M P A 1989 Phys. Rev. Lett. 62 1415 Fisher D S, Fisher M P A and Hus D A 1991 Phys. Rev. B 43 130 [18] Wen H H, Schnack H G, Grissen R, Dam B and Rector J 1995 Physica C 241 353 [19] Griessen R, Wen H H, Van Dalen A J J, Dam B, Rector J, Schnack H G, Libbrecht S, Osquiguil E and Bruynseraede Y 1994 Phys. Lett. 72 1910 [20] Brandt E H 1995 Rep. Prog. Phys. 58 1465 [21] van der Beek C J and Kes P H 1991 Phys. Rev. B 43 13 032 [22] Wen H H, Zhao Z X, Wijingaarden R J, Rector J, Dam B and Griessen R 1995 Phys. Rev. B 52 4583 361