PHYSICAL REVIEW B VOLUME 59, NUMBER 2 1 JANUARY 1999-II Onset of flux penetration into a thin superconducting film strip A. V. Kuznetsov and D. V. Eremenko Department of Quantum Electronics, Moscow State Engineering Physics Institute, 115409 Moscow, Russia V. N. Trofimov Laboratory of Nuclear Problems, Joint Institute for Nuclear Research, 141980, Dubna, Moscow region, Russia ~Received 26 June 1998! Penetration of vortices into a thin superconducting film and the edge barrier arising in a transverse magnetic field are considered. The vortex Gibbs energy is calculated on the base of the Kogan’s solution for the Pearl’s vortex situated near the film edge. The penetration field and the magnetization in the critical state are calculated. The influence of the edge barrier on magnetization of high-T c thin films is estimated. @S0163-1829~99!13901-8# I. INTRODUCTION Application of thin films of high-temperature superconductors ~HTSC’s! in superconducting devices requires precise knowledge of the films’ response to external fields. Recently the critical-state problem has been solved for a film inserted in a magnetic field applied normally to the film.1,2 Penetration of the magnetic flux into a film in the presence of an edge barrier has been described by the geometrical barrier model.3–5 Both models give the magnetic response when spatial distribution of the critical current or its dependence on the flux density are known. Since they deal with currents and flux averaged over several intervortex distances, neither entry of the first vortex into a film nor features of flux penetration following from peculiarities of thin-film vortices are treated in the frame of these models. As was shown by Pearl6 thin-film vortices interact with external and self-induced fields primarily through the free space adjacent to the film, where no screening currents flow. As a sequence, the vortical current nonexponentially decreases with a distance from the vortex core and the vortex self-energy depends on both film dimensions7,8 and a vortex position in the film.7,9 In addition, vortices interact with shielding current flowing throughout the film in an external field. All this leads to strong dependence of the vortex Gibbs energy on a position in the film. It is well known that the Bean-Livingston barrier for the entry of vortices exists at the surface of bulk superconductors10–12 and at the film edges.7,13 The barrier was analyzed in a thorough discussion of the Meissner response of pinning-free thin films given by Likharev.7 A set of integral equations for computing the Gibbs free energy was obtained. However, generality of the analysis makes it difficult to use the obtained results. Some simplification was introduced to account for the edge barrier in the critical state. It was shown that the vortex self-energy rises towards the film center so vortices are attracted by the edges.7 The interaction of a vortex with its image outside the edge was used to describe the attraction,13 but the exact solution by Kogan9 for a vortex situated near the edge contradicts such an approximation. To calculate interaction of vortices with the shielding current I the distribution I}1/AW2x was used near the edge where vortices are nucleated.13 This distribution diverges at the edge position W while the edge current is 0163-1829/99/59~2!/1507~7!/$15.00 PRB 59 limited.7,14 To obtain the edge current one should solve the London equation playing a major role in the physics of thin superconducting films.4,7,9,14 In the present paper we compute the London equation and using Kogan’s solution calculate the Bean-Livingston barrier in a film strip. We treat the case of strong pinning which is inherent in real films14 because of interaction of vortices with defects, dislocations and tensions arising at a film-substrate interface. The paper is organized as follows. In Sec. II the Meissner response is analyzed. Using Kogan’s solution we calculate the vortex Gibbs energy and describe the Bean-Livingston barrier arising in thin films. In Sec. III influence of the barrier on flux penetration is analyzed for the case of strong pinning. We account for the vortex self-energy gradient in the balance of the forces acting upon a vortex and calculate the magnetization on the base of the critical-state equations by McDonald and Clem.2 In Sec. IV we discuss suppression of the barrier by defects at the edge and calculate the penetration field in which vortices begin to enter a strip. Section V contains the barrier-related estimations made for YBa2 Cu3 O72 d films. In Sec. VI a brief summary of our results is given. In the Appendix the barrier-related saturated magnetization is calculated. II. VIRGIN STATE Let us start with an infinite superconducting film strip inserted in a uniform magnetic field H applied normally to the strip. We choose the origin of the coordinate frame at the midpoint of the strip with the y axis lengthwise and the x axis in the lateral direction, so that the edges are located at (2W,y,0) and (W,y,0) ~see Fig. 1!. The film thickness d is small in comparison with the penetration depth l. We neglect the change of a current density J across the thickness and consider shielding current I5 * d/2 2d/2Jdz. The strip is wide in comparison with the screening length, W@L 52l 2 /d. Because of the symmetry of the current and flux distribution in the strip we analyze only the x>0 half of the strip. In this section we treat the virgin state in which no vortices are in the film in the zero field. Then the field increases 1507 ©1999 The American Physical Society A. V. KUZNETSOV, D. V. EREMENKO, AND V. N. TROFIMOV 1508 FIG. 1. Sketches of the superconducting strip, the shielding current distribution and the vortex Gibbs energy. Vectors in the energy sketch represent directions of a force f acting upon a vortex. up to a small value and the shielding Meissner current induces in the strip. The current distribution is calculated from the London equation14 1 L ]I 5 2 ]x 2p E W I ~ x 8 ! dx 2W x 8 2x 2H. ~1! The first term at the right-hand side of Eq. ~1! represents the demagnetization field produced by the current itself. Integrating over x and calculating the integration constant from the boundary condition I(0)50 following from the odd current symmetry I(x)52I(2x), one writes I5 W pL E U 1 ln 0 U 2xH u1x/W . I ~ u ! du2 u2x/W L ~2! We calculated the current distributions presented in Fig. 2. In the central part of the strip the current is described by the well-known dependence7,14 PRB 59 FIG. 3. The vortex Gibbs energy. The energy is presented in units of the self-energy of a vortex situated at the strip center. The applied field h5H/H * is taken in units of H * 5 f 0 / m 0 L ALW. I52 2Hx AW 2 2x 2 ~3! . The larger the W/L ratio, the smaller the difference between Equation ~3! and the calculated distribution. Eq. ~3! diverges like 1/AW2x at the edge while the calculated current is limited by the maximal value I m 5H A2 p W/L. The edge current is close to the dependence14 i511( z / p )ln(gz/4) where i52I/I m , z 5(W2x)/L, and g 50.577 21 . . . is the Euler constant. As seen from Fig. 2, independently of the W/L ratio the deviation of the current distribution from Eq. ~3! occurs at a distance &L from the edge. In the belt z <1 the current is fitted by i5 @ 1 1( z / p )ln(gz/4) # exp(0.357z 1.2) with an accuracy of a few percent. Because of demagnetization the local field and current strongly increase at the edge with applied field. Vortices spontaneously nucleate at the edge via dynamic suppression of the order parameter.15 The self-energy of a vortex situated near the edge of a wide superconducting film is9 «5 F S D S DG f 20 rc W2x F0 2F 0 4 m 0L 2L L , ~4! where f 0 is the flux quantum, the vortex core radius r c is of the order of the coherence length j ,F 0 5Y 0 2H0 , and H0 , Y 0 are the zero-order Struve and the second-kind Bessel functions in the notation of Ref. 16. The vortices interact with the shielding current. Though a vortex in a strip carries less than f 0 of a flux, the interaction of such a vortex with the shielding current is still described in terms of the common Lorentz force9 f L 5 f 0 I. Integrating the Lorentz force one obtains the interaction energy and writes the vortex Gibbs energy FIG. 2. Distribution of the Meissner shielding current. The inset represents the current near the strip edge. The curves are calculated for W/L5100 ~solid! and 2500 ~dashed!. The dotted curves correspond to Eq. ~3! ~the right curve in the inset! and the edge current calculated in Ref. 14. g5«1 f 0 E W x I ~ x 8 ! dx 8 . ~5! The energy calculated from Eqs. ~2!, ~4!, and ~5! is shown in Fig. 3 for some values of applied field. As seen, the potential barrier preventing an entry of vortices exists in the PRB 59 ONSET OF FLUX PENETRATION INTO A THIN . . . strip. The vortices arising at the edge are pushed out of the strip by the force f52¹g since their energy rises toward the strip inner part. There should be no appreciable thermal activation over the barrier the scale energy of which is large, f 20 /4m 0 L* 63104 K for L&1 m m. Vortices penetrate the strip central part only when the Gibbs energy maximum is located at the edge. Using the equation ¹g50 relating the position of the Gibbs energy maximum x m to the applied field we obtained that I m is of the order of the depairing current when x m ;W2r c . Note that only qualitative estimations can be made in a belt of a width r c ; j adjacent to the edge since the London description and, consequently, Eqs. ~2! and ~4! fail there. When H exceeds the field at which the Gibbs energy maximum is located at the edge, vortices enter the strip and migrate toward its central part. The migration is affected by pinning. In the next section we consider the arising critical state accounting for both the pinning and the vortex energy gradient. FIG. 4. Magnetization curves calculated for i p 51 in the presence ~top! and absence of the edge barrier. follows from the critical-state model for a strip with nonuniform critical current.2 Here a is the flux front position related to the applied field by the equation III. CRITICAL STATE In presence of pinning vortices migrate under the influence of the following forces. The self-energy gradient and the Lorentz force tend to move a vortex, the pinning force fp tends to prevent its displacement. If fp exceeds 2¹g the vortex is pinned. In the opposite case it moves to the point where fp 5¹g. That point is situated far from the edge when pinning is weak. As a result, a region where vortices entering the strip are accumulated, is separated from the edge by a wide (@L) region which is free of flux. Such a flux penetration has been described in the geometrical barrier model.3,4 We assume that pinning is strong enough, so a small (;L) displacement from the edge leads to stopping of the vortex. In this case a region of the critical state forms at the edge and grows with the field. Therefore flux penetration can be treated in the frame of the critical-state model. Flowing in the critical-state region the critical current I c provides a balance of the forces acting upon a vortex. From the equation fp 5¹g one obtains f 0 I c 5« 8 2 f px . Here « 8 denotes the derivative with respect to x. We take the pinning force independent of the flux density and express fp in terms of the pinning current I p 5const.0. Since fp acts oppositely to the direction of the vortex displacement, f px 5 f 0 I p , when flux enters the strip and f px 52 f 0 I p when flux exit it. Calculating the self-energy gradient from Eq. ~4! accounting that16 Y 08 52Y 1 and H08 52/p 2H1 we write the critical current I c5 f0 S D G F 2 W2x 2F 1 6i p , L 4 m 0L p 2 ~6! where i p 54 m 0 L 2 I p / f 0 , signs plus and minus correspond to the exit and entry of flux, respectively. The current distribution throughout the strip I5 H I c~ x ! , 2 p a<x, Aa 2 2x 2 E W a I c ~ x 8 ! dx 8 ~ x 8 22x 2 ! Ax 8 2a 2 , 2 x,a, ~7! 1509 H5 1 p E I c dx W a Ax 2 2a 2 ~8! . Integrating the current distribution one calculates magnetization of the strip m5 1 Wd E W ~9! Ixdx. 0 Substituting Eq. ~7! for I, dividing the integral into two parts a W W *W 0 . . . dx5 * 0 dx * a . . . dx 8 1 * a . . . dx, inverting the integration order in the double integral, integrating * a0 . . . dx and summing the result with the second part we obtain m5 1 2Wd E W a Ic 2x 2 2a 2 Ax 2 2a 2 dx. ~10! Calculating both H and m from Eqs. ~8! and ~10! for a set of flux front positions one constructs a magnetization curve. When the barrier is absent, i.e., I c 5I p 5const, calculations give the known dependences1 H 21 5(I p / p ) cosh (W/a), m52(WI p /2d) A12(a/W) 2 . For strip with W5100L and i p 51 we computed the magnetization curve presented in Fig. 4 with m5 2(WIp/2d)tanh(pH/Ip) following from the dependences mentioned above. As seen, the edge barrier results in an increase of the magnetization. Using Eq. ~6! in the frame of the critical-state model2 one calculates the magnetization in a decreasing field as well as hysteretic magnetization curve. This problem cannot be solved in the present work because of its complexity. However, we analyzed the saturated remanent magnetization m rs remaining in the strip when H decreases from a high value down to zero. We found that the measurement of m rs and the saturated virgin magnetization m s allows one to separate the barrier-related and pinning-related contributions to the magnetization. In the saturated state the critical current flows throughout the strip. Its distribution is given by Eq. ~6! where the sign of i p is positive for the remanent state and negative for the 1510 PRB 59 A. V. KUZNETSOV, D. V. EREMENKO, AND V. N. TROFIMOV FIG. 5. Ratio of the saturated magnetization calculated for i p 51 and L/r c 5250. The top curve represents the ideal barrier, the other curves correspond to the barrier suppressed by edge imperfection. For a detailed explanation see Sec. IV. virgin state. From Eqs. ~6! and ~9! one obtains that the integrals containing pinning current annihilate in the sum m s 1m rs ; * « 8 xdx. On the contrary, the integrals containing the self-energy gradient annihilate in the difference m s 2m rs ; * i p xdx. In the Appendix we calculated the ratio of the barrier-related and the pinning-related magnetization S D m s 1m rs 4L L . ln 10.81 . m s 2m rs p i p W r c Since the barrier results in an increase of the critical current only near the edge, its influence on the magnetization is independent of a strip width while the pinning-related magnetization is proportional to W. Therefore, as seen from Fig. 5 where dependence of the ratio on W is depicted, the barrier slightly affects the magnetization of wide strips. The vortex self-energy and the critical current primarily change in a narrow belt of a width ;L adjacent to the edge ~see Fig. 3! so a damage of the strip in this region strongly affects the penetration of vortices through the barrier. In the next section we discuss the onset of flux penetration accounting for the edge imperfection. IV. PENETRATION FIELD This is a general point that the surface roughness must suppress the Bean-Livingston barrier. Indeed, in the recent study by Bass et al.17 it was shown that the surface roughness leads to both a decrease of the barrier energy and a shift of the barrier maximum from the surface, however no dramatic suppression of the barrier has been obtained. For example, for the root-mean-square surface roughness ;0.4l the decrease of the energy barrier from its value for ideal surface is only about 10%.17 Note that such a roughness value is quite large since in the treated case of a type-II superconductor the barrier is located at a distance ;l from the surface. For a thin film one also expects no strong suppression of the barrier by the edge unevenness. As follows from the experiments in which the surface of a superconductor was damaged by electron irradiation, there is strong suppression of the surface barrier under the influence of the surface defects.18 To understand this fact one should go beyond the Bean-Livingston approximation which deals with an already nucleated vortex but is unable to analyze the vortex nucleation. Transition to the mixed state in a type-II superconductor occurs as a result of sufficiently large fluctuations of the order parameter.11 The vortex nuclei in the form of a semicircle, the ends of which touch the superconductor surface, are produced near the surface. There exists a minimal critical size ~and energy! of the vortex nucleus for which the nucleus is not ‘‘closed up,’’ continues to develop further, and penetrates into the superconductor.11,12 The critical energy decreases and the fluctuation frequency increases in the regions with suppressed order parameter, therefore the surface defects suppress the surface barrier. In the case of a thin film the numerical simulation of the time-dependent Ginzburg-Landau equation completed by the appropriate Maxwell equations gives the following description for the vortices nucleation.15 The flux penetration into a thin superconducting film occurs via the dynamic suppression of the order parameter on the macroscopic scale and can be viewed as the nucleation of the extended droplets of the normal phase in the superconducting sample. The droplets, emerging at the film edge, stretch for the distance w from the edge. This distance is of the order of some tens of j for narrow film strips ( j !W!L) and is expected to be as large as l for wide ones. The droplets contain multiple topological charge and therefore they are unstable with respect to splitting into singly charged vortices. The time scale of formation and splitting of the droplets is ;1029 s. When the droplets split, the vortices arise inside the film and some of them are situated at the distance ;w from the edge. In the presence of the defects the normal phase tries to settle at the defect sites where the order parameter is already suppressed. The droplets percolate along easy paths connecting the defect sites and split into the vortices which are pinned at the defects. So far we treat a strip with ideal edges. Real samples, however, have the edge defects which are introduced by intergrain boundaries, if the film is polycrystalline, and by lithography because of the decreasing of the film thickness and damage of the superconductor. Via the dynamic suppression of the order parameter the vortices arise in the defect regions in a sufficiently low applied field. One estimates w ;d for a lithography-related defects. In the case of the polycrystalline film w is of the order of the grain diameter, as shown in Fig. 6. Consider first the zero-pinning behavior. If the maximum of the vortex Gibbs energy is situated far from the edge, x m ,W2w, vortices nucleating in the defect regions are pushed out of the strip. When x m >W2w vortices enter the strip and migrate under the influence of the force 2¹g acting towards the strip center. The penetration field H p , in which the first vortex enters the strip, is calculated from the equation ¹g50 solving for x5W2w. Pinning tends to hold a vortex at the site of its nucleation. The first vortex is pinned when ¹g and fp becomes equal at some field. From the equation ¹g5fp one calculates the penetration field when pinning is present m 0H p5 f0 F 1 22/p 2i p L ALW 4 A2 p i . ~11! PRB 59 ONSET OF FLUX PENETRATION INTO A THIN . . . 1511 for a vortex situated in the film center.8 Using m .4 f 0 W(2 ln 211)/p,9 Eq. ~4! and Ref. 16 we write m 0 H c1 f . FIG. 6. Sketch of the strip near the edge. The polycrystalline film ~grey! consists of single-crystal grains of a characteristic diameter D which are separated by the intergrain boundaries ~lines!. The bottom cross sections indicate that the edge rounding varies along the strip. The dark grey color marks the region of the edge damage related to lithography. w is a characteristic width of the edge imperfection due to granularity or lithography. Here the current i and function F 1 are taken at w/L. Note that the same dependence of the penetration field on a sample width H p }1/AW was obtained earlier for a thin flat type-II superconductor.3,5 The penetration field as a function of w and i p is depicted in Fig. 7. H p is strongly suppressed by increase of w. The stronger, the pinning, the lower the penetration field. Moreover, H p is zero for some w. This means that any vortex arising in the infinitesimal field must be pinned. At the same time the penetration field is finite. In a weak field the minimum of the sample free energy is achieved when the film is in the Meissner state. The Meissner state becomes metastable with respect to formation of the mixed state when energies of these states become equal with field increase. The corresponding field, named the lower critical field of film H c1 f , is calculated from the equation H c1 f 5«/ m where the moment m and the energy are taken FIG. 7. The penetration field vs w and i p ~inset!. The field h p 5H p /H * is presented in units of H * 5 f 0 / m 0 L ALW. S D L f0 ln 10.81 . 16~ 2 ln 211 ! LW r c When H,H c1 f no vortices nucleate in a film, therefore H c1 f is the lowest limit for the penetration field. When the first vortex is pinned at the edge the vortical current joins with the shielding one. The edge current rises, therefore, the probability of nucleation of successive vortices increases. The critical state forms at the edge. In the region x<W2w the critical current is described by Eq. ~6!. In the belt x>W2w the current depends on unknown parameters of superconductor in the defect regions. As a contribution of the edge current to the saturated magnetization is as small as 2w/W!1 for a wide strip, it is quite reasonable to neglect the distribution of the edge current and to estimate it by some average value. We restrict the edge current by I c at x5W 2w and calculate in the Appendix the ratio of the barrierrelated to the pinning-related magnetization: F S m s 1m rs 2L w 2 . F 01 F 12 m s 2m rs i p W L p DG . ~12! Here functions F are taken at w/L. As follows from Fig. 5, where the ratio is presented for different w/L, the barrier slightly affects the magnetization if vortices arise at the distance ;L from the edge. When edges are not specially prepared by lithography, the magnetization is well described by the conventional critical-state model, since w@L for such films. V. DISCUSSION Because of the presence of the shielding current flowing throughout the film the barrier discussed above differs from the Bean-Livingston barrier in bulk superconductors.10–12 Vortices in films of the layered superconductor consist of pancakes which are able to surmount the edge barrier one after another. Therefore, the obtained results are hardly applied to films of layered superconductors. Let us estimate some barrier-related parameters. There is the most interest in HTSC thin-film superconducting devices operated at liquid-nitrogen temperature. So we take a strip of a YBa2 Cu3 O72 d film with thickness d50.15 m m and width 2W5100 m m as an example. For HTSC films with the c axis perpendicular to the film the restriction on a film thickness d!2l is rewritten as d !2l/ Ae . Due to strong anisotropy ~the anisotropy parameter e .0.04 for YBa2 Cu3 O72 d ) HTSC films of thickness d;l are quite thin. High-quality epitaxial YBa2 Cu3 O72 d films consist of islandlike grains 0.25 m m in diameter.19 Thus the film thickness and the diameter of grains are of the same order. At T577.4 K for l.0.22 m m ~Ref. 20! and a typical value of the pinning-related critical current density19,21 J c .23106 A/cm2 , one calculates L.0.65 m m and i p .3. Taking w/L50.25 which corresponds to w;d, we estimate the penetration field H p 5H c1 f .0.1 Oe ~Ref. 22! and obtain that the barrier-related saturated magnetization is only 1% of 1512 A. V. KUZNETSOV, D. V. EREMENKO, AND V. N. TROFIMOV the pinning-related one. Thus the barrier slightly affects the high-field magnetization of YBa2 Cu3 O72 d films. Note, that the barrier-related magnetization can be observed in films with lower J c . For example, we calculated (m s 1m rs )/(m s 2m rs ).0.17 at T515 K for the Nd1.85Ce0.15CuO42 d film strip with W/L5100.23 To estimate the barrier influence on low-field magnetization, ac losses, and noises, one should numerically solve the critical-state equations2 with the critical current given by Eq. ~6!. When a flux front position is located near the edges, it is evident that the barrier holds in flux penetration. Indeed, the authors of Ref. 13 observed the barrier-related decrease of magnetic hysteresis in superconducting quantum interference device magnetometers caused by penetration of vortices near the edges. One expects also that the barrier influence is significant in high-frequency fields when pinning is suppressed. In this case the barrier-related frequency is apparently restricted by a time of the film-vortex nucleation which is of the order of 1029 s. 15 VI. CONCLUSION In this paper we considered the Bean-Livingston barrier in a thin superconducting film. The present analysis is a part of the general problem of flux penetration through an edge barrier in a thin flat superconductor. The cases of type-I and type-II superconductors have been analyzed earlier ~see Refs. 5,24, and references therein!. In conclusion we point out the main results obtained in the present work. ~1! The Gibbs energy of a vortex situated near the edge of a thin superconducting film strip has been calculated and the Bean-Livingston barrier has been described. ~2! The critical-state model has been extended to account for the edge barrier and the magnetization has been calculated. ~3! An entry of the first vortex into a film has been analyzed and the penetration field has been calculated. Integrating by parts the first integral we accounted that «(W)50. Using F 0 ( z ).(2/p ) @ ln(2/z )2 g # , z 5r c /2L!1 for the first term and * u0 F 0 d z .(2/p ) @ ln(2u)1g#, u5W/L @1 for the second term16 one writes m s 1m rs .2 S ~A1! and 2 m s 1m rs 5 f 0 Wd 52 E W 0 2 « 8 xdx52 f 0 Wd F S D E f0 rc L F0 2 2 m 0 Ld 2L W E W 0 W/L 0 ~A2! D ~A3! When the critical current is restricted at the edge region W2w<x by the constant value I c 5« 8 / f 0 6I p , where « 8 is taken at W2w, one writes for the barrier-related magnetization m s 1m rs 5 2 f 0 Wd 52 2 FE W2w 0 F x« 8 ~ x ! dx1« 8 E E W f0 ~ W2w ! F 0 2L 2 m 0 LWd S w ~ 2W2w ! 2 2F 1 2L p DG xdx W2w W/L w/L G F 0~ z ! d z . Here functions F are taken at w/L. Using the integral representation16 of F 0 one writes E b a F 0d z 5 2 p E 5 2 p E ` e 2t dt 0 E Az z d b 2 a 1t 2 b1 Ab 2 1t 2 e 2t ln dt, a1 Aa 2 1t 2 ` 0 and APPENDIX m s 2m rs 52WI p /d, D m s 1m rs 4L L . ln 10.81 . m s 2m rs p i p W r c m s 1m rs 52 In the case of the ideal edge barrier the critical current I c 5« 8 / f 0 6I p flows throughout the strip. From Eqs. ~4! and ~9! one calculates for I p 5const S f0 4L L 2W ln 2 ln . p m 0 Ld e g r c W e g L As L/r c ;2l 2 /d j &2 k , we estimate for k 5l/ j >50 and W/L>50 that the last term at the right-hand side is less than 1.5% of the first one. Neglecting it and dividing Eq. ~A2! by Eq. ~A1! we write for the ideal barrier ACKNOWLEDGMENTS The authors are indebted to Dr. A. A. Ivanov and K. V. Klementev for helpful discussions. We are grateful to Professor A. P. Menushenkov for his support of the present investigation. PRB 59 F5 FE L 2 W p 2 2 S S D G W/L1 A~ W/L ! 2 1t 2 e 2t ln 0 w2 2L ` F f0 w 2 F 01 F 12 2F , 2 m 0 Ld L p F 12 w/L1 A~ w/L ! 1t 2 p 2 DG 2 dt1 w F L 0 ~A4! . We estimate that F can be neglected with an accuracy of 4% or better for w/L>0.1 and W/L>50. Dividing Eq. ~A4! by Eq. ~A1! we finally write «dx G F 0~ z ! d z . F S m s 1m rs 2L w 2 . F 01 F 12 m s 2m rs i p W L p DG . ~A5! PRB 59 ONSET OF FLUX PENETRATION INTO A THIN . . . E. H. Brandt and M. Indenbom, Phys. Rev. B 48, 12 893 ~1993!; E. Zeldov, J. R. Clem, M. McElfresh, and M. Darwin, ibid. 49, 9802 ~1994!. 2 J. McDonald and J. R. Clem, Phys. Rev. B 53, 8643 ~1996!; P. N. Mikheenko and Y. E. Kuzovlev, Physica C 204, 229 ~1993!. 3 E. Zeldov, A. I. Larkin, V. B. Geshkenbein, M. Konczykowski, D. Majer, B. Khaykovich, V. M. Vinokur, and H. Shtrikman, Phys. Rev. Lett. 73, 1428 ~1994!; T. Schuster, M. V. Indenbom, H. Kuhn, E. H. Brandt, and M. Konczykowski, ibid. 73, 1424 ~1994!; M. Benkraouda and J. R. Clem, Phys. Rev. B 53, 5716 ~1996!. 4 I. L. Maksimov and A. A. Elistratov, Pis’ma Zh. Eksp. Teor. Fiz. 61, 204 ~1995! @JETP Lett. 61, 208 ~1995!#; M. Yu. Kupriyanov and K. K. Likharev, Fiz. Tverd. Tela ~Leningrad! 16, 2829 ~1974! @Sov. Phys. Solid State 16, 1835 ~1974! 5 A. Kuznetsov, D. V. Eremenko, and V. N. Trofimov, Phys. Rev. B 56, 9064 ~1997!. 6 J. Pearl, Appl. Phys. Lett. 5, 65 ~1964!. 7 K. K. Likharev, Izv. Vuzov Radiofiz. 14, 909 ~1971! @Radiophys. Quantum Electron. 14, 714 ~1971!#; 14, 819 ~1971! @14, 722 ~1971!#. 8 A. L. Fetter, Phys. Rev. B 22, 1200 ~1980!. 9 V. G. Kogan, Phys. Rev. B 49, 15 874 ~1994!. 10 C. P. Bean and J. D. Livingston, Phys. Rev. Lett. 12, 14 ~1964!; J. R. Clem, in Proceedings of the 13th Conference on Low Temperature Physics, edited by K. D. Timmerhaus, W. J. O’Sullivan, and E. F. Hammel ~Plenum, New York, 1974!, Vol. 3, p. 102. 11 V. P. Galaiko, Zh. Éksp. Teor. Fiz. 50, 1322 ~1966! @Sov. Phys. JETP 23, 878 ~1966!#; B. V. Petukhov and V. R. Chechetkin, ibid. 65, 1653 ~1973! @ibid. 38, 827 ~1974!#. 12 L. Burlachkov, Phys. Rev. B 47, 8056 ~1993!; L. Burlachkov, V. B. Geshkenbein, A. E. Koshelev, A. I. Larkin, and V. M. Vinokur, ibid. 50, 16 770 ~1994!. 13 J. Z. Sun, W. J. Gallagher, and R. H. Koch, Phys. Rev. B 50, 13 664 ~1994!. 14 A. Larkin and Y. Ovchinnikov, Zh. Éksp. Teor. Fiz. 61, 1221 ~1971! @Sov. Phys. JETP 34, 651 ~1972!#. 1 15 1513 I. Aranson, M. Gitterman, and B. Y. Shapiro, Phys. Rev. B 51, 3092 ~1995!; I. Aranson, B. Y. Shapiro, and V. Vinokur, Phys. Rev. Lett. 76, 142 ~1996!. 16 Handbook of Mathematical Functions with Formulas, Graphs and Mathematical Tables, Natl. Bur. Stand. Appl. Math. Ser. No. 55, edited by M. Abramowitz and A. Stegun ~US GPO, Washington, DC, 1965!. Following from F n ( z )5 @ 2( z /2) n / Ap G(n 11/2) # * `0 e 2 z t (11t 2 ) n21/2dt where G is the gamma function, the integral representations F 0 ( z )5(2/p ) * `0 e 2 z t / A11t 2 dt and F 1 ( z )5(2 z / p ) * `0 e 2 z t A11t 2 dt can be used for computing. F 0 .(2/p ) @ ln(2/z )2 g # for z !1 and F 0 .2/pz for z @1. Here g is the Euler constant. F 1 .2/pz for z !1 and F 1 .(2/p )(1 11/z 2 ) for z @1. 17 F. Bass, V. D. Freilikher, B. Ya. Shapiro, and M. Shvartser, Physica C 260, 231 ~1996!. 18 M. Konczykowski, L. Burlachkov, Y. Yeshurun, and F. Holtzberg, Phys. Rev. B 43, 13 707 ~1991!; N. Chikumoto, M. Konczykowski, N. Motohira, K. Kishio, and K. Kitazawa, Physica C 185-189, 1835 ~1991!. 19 M. Kawasaki, J. Gong, M. Nantoh, T. Hasegawa, K. Kitazawa, M. Kumagai, K. Hirai, K. Horiguchi, M. Yoshimoto, and H. Koinuma, Jpn. J. Appl. Phys., Part 1 32, 1612 ~1993!; R. E. Muenchausen, M. Hawley, S. R. Foltyn, X. D. Wu, R. C. Dye, F. H. Garzon, G. L. Skofronic, and A. H. Carim, Physica C 199, 445 ~1992!. 20 W. N. Hardy, D. A. Bonn; D. C. Morgan, R. Liang, and K. Zhang, Phys. Rev. Lett. 70, 3999 ~1993!; D. A. Bonn, S. Kamal, K. Zhang, R. Liang, D. J. Baar, E. Klein, and W. N. Hardy, Phys. Rev. B 50, 4051 ~1994!. 21 A. Kuznetsov, A. A. Ivanov, D. V. Eremenko, and V. N. Trofimov, Phys. Rev. B 52, 9637 ~1995!. 22 We took r c 5 j leading to L/r c 52 k l/d and k 584 @J. Gohng and D. K. Finnemore, Phys. Rev. B 46, 398 ~1992!#. 23 We used the parameters of the Nd1.85Ce0.15CuO42 d films presented in Ref. 21. 24 A. Kuznetsov, D. V. Eremenko, and V. N. Trofimov, Phys. Rev. B 57, 5412 ~1998!.