The London theory of the crossing-vortex lattice in highly anisotropic layered superconductors

S.E. Savelev , J. Mirkovi+, and K. Kadowaki c

Institute of Materials Science, The University of Tsukuba, 1-1-1 Tennodai, Tsukuba 305-8573, Japan, and CREST, Japan Science and Technology Corporation (JST), Japan A novel description of Josephson vortices (JVs) crossed by the pancake vortices (PVs) is proposed on the basis of the anisotropic London equations. The eld distribution of a JV and its energy have been calculated for both dense (a < J ) and dilute (a > J ) PV lattices with distance a between PVs, and the nonlinear JV core size J . It is shown that the shifted PV lattice (PVs displaced mainly along JVs in the crossing vortex lattice structure), formed in high out-of-plane magnetic elds Bz > 0 / 2 s2 [A.E. Koshelev, Phys. Rev. Lett. 83, 187 (1999)], transforms into the PV lattice trapped by the JV sublattice at a certain eld, lower than 0 / 2 s2 , where 0 is the ux quantum, is the anisotropy parameter and s is the distance between CuO2 planes. With further decreasing Bz , the free energy of the crossing vortex lattice structure ( PV and JV sublattices coexist separately) can exceed the free energy of the tilted lattice (common PV-JV vortex structure) in the case of s < ab with the in-plane penetration depth ab if the low (Bx < 0 /2 ) or high ab (Bx 0 /s2 ) in-plane magnetic eld is applied. It means that the crossing vortex structure is realized in the intermediate eld orientations, while the tilted vortex lattice can exist if the magnetic eld is aligned near the c-axis and the ab-plane as well. In the intermediate in-plane elds 0 /2 Bx 0 /s2 , the crossing vortex structure with the trapped PV sublattice ab seems to settle in until the lock-in transition occurs since this structure has the lower energy with respect to the tilted vortex structure in the magnetic eld H oriented near the ab-plane. The recent experimental results concerning the vortex lattice melting transition and transitions in the vortex solid phase in Bi2 Sr2 CaCu2 O8+ single crystals are discussed in the context of the presented theoretical model. PACS numbers: 74.60.Ge, 74.60.Ec, 74.72.Hs

arXiv:cond-mat/0105370v1 [cond-mat.supr-con] 18 May 2001

The mixed state of high temperature superconductors is complex and rich with various vortex phases1,2 . Besides the vortex lattice described by 3D anisotropic Ginzburg-Landau model, the new types of vortex structures can occur within the large part of the phase diagram of the mixed state where the coherence length along the c-axis is smaller than the distance between CuO2 planes. In such a case, the magnetic eld, aligned with the caxis, penetrates a superconductor in the form of quasi two-dimensional pancake vortices (PVs)3 while the eld applied parallel to the ab-plane generates Josephson vortices (JVs) in the layers between CuO2 planes4,5 . In magnetic elds tilted with respect to the c-axis, PVs and JVs can form a common tilted lattice5 or exist separately as a crossing (combined) lattice6,7 . The tilted lattice represents the inclined PVs stacks in elds applied close to the c-axis while, at higher angles, the pieces of JVs linking PVs are developed5,6 . The crossing lattice is another structure containing both a PV stack sublattice and a JV sublattice which coexist separately. The vortex-solid phase diagram in the tilted magnetic elds was rst proposed by Bulaevskii, Ledvij, and Kogan6. According to their model, which does not take into account the interaction between PV and JV sublattices in the crossing lattice structure, the tilted lattice is formed for all orientations of the magnetic eld until the lock-in transition8 occurs if the in-plane London penetration depth ab is larger than the Josephson vortex core with size s ( is the anisotropy parame-

ter and s is the distance between CuO2 planes). In the opposite limit, s > ab , the tilted lattice transforms into the crossing lattice (as the magnetic eld is inclined away from the c-axis) at a certain angle before the lock-in transition happens6 . Later, the possibility of the coexistence of two vortex sublattices with dierent orientations was analyzed numerically by comparing the free energy of such system with the free energy of mono-oriented tilted vortex lattice at dierent eld orientations and dierent absolute values of the external magnetic eld for the case of 3D anisotropic (London model) superconductors911 as well as layered (LawrenceDoniach model12 ) superconductors13 . According to that analysis11,13 performed for = 50 160, the crossing lattice can be energetically preferable in the quite low 2 2 magnetic elds (B = Bz + Bx 0 /2 ) in the inab termediate eld orientations 0 < 1 < < 2 < /2 with = arctan(Bx /Bz ) (Bz and Bx are the eld component along the c-axis and parallel to the ab-plane, respectively). However, the interaction of two coexisting vortex sublattices was not considered in those works911,13 . Recently, Koshelev7 has studied the case of extremely anisotropic superconductors s ab and has shown that the crossing lattice can occupy substantially larger region of the vortex lattice phase diagram in the oblique elds due to the renormalization of the JV energy EJ through the interaction of a Josephson vortex and the PV sublattice. In addition, such interaction leads to the attraction of PVs to JVs14,7 at low out-of-plane magnetic elds Bz

(the some sort of pinning eect). This pinning may induce transitions between dierent substructures of the crossing lattice structure. However, there is still no theoretical investigation how PV sublattice can inuence on the JV lattice in the crossing vortex structure in the case of moderate anisotropic superconductors with s < ab . In this regard, the phase diagram6 of the vortex lattice for strongly anisotropic layered superconductors should be reconsidered (at least for the case of s < ab ) by taking into account the renormalization of EJ and the pinning of PVs by JVs. The vortex structures in highly anisotropic layered superconductors are usually studied on the basis of the nonlinear discrete Lawrence-Doniach model12 , but this model is quite complex and the detailed analysis of the vortex system is complicated. On the other hand, the layerness of superconductors can be ignored on scales larger than the size of the nonlinear Josephson vortex core. Therefore, the linear anisotropic London model could be applied for a study of the vortex crossing lattice outside JV cores. In this paper we introduce the extended London theory which allows to describe the crossing lattice as well as to calculate the energy EJ and the eld distribution of JV in the presence of the crossed PV sublattice. It is shown that with decreasing the perpendicular magnetic eld the pancake sublattice transforms from the shifted sublattice characterized by one component PV displacement along JVs to the trapped sublattice where JVs are occupied by PV rows. The comparison of the free energies of the tilted lattice and the crossing vortex structure for the case ab > s indicates that in low (Bx 0 /2 ) and high (Bx 0 /s2 ) in-plane elds, ab the tilted lattice can exist if the vector B = Bx ex + Bz ez is directed close to the c-axis as well as near the ab-plane, while the crossing lattice is realized in the elds oriented far enough from the crystal symmetry axes. Furthermore, in the intermediate in-plane elds the tilted vortex lattice exists only at the magnetic eld orientations near the c-axis whereas crossing vortex structure settles in the wide angular range until the lock-in transition happens. This paper is organized as follows. The general equations for the magnetic eld distribution and the energy of Josephson vortex in the presence of the pancake lattice are derived in section I. The dense pancake lattice is studied in section II. It is shown that, in the limit s ab and 0 /(s)2 Bz 0 2 s2 /4 , our ab model reproduces the results which were earlier obtained by Koshelev7 , while the shear deformation of the PV lattice signicantly renormalizes the JV energy at the higher out-of-plane elds. Section III is devoted to the dilute PV lattice. It is described how the novel vortex substructure with the PV lattice trapped by the JV lattice can be realized at low Bz . The phase diagram of the vortex-solid phase in the tilted magnetic elds is considered for the case of ab > s in section IV while the recent experimental results are discussed in section V.

I. JOSEPHSON VORTEX IN THE PRESENCE OF PANCAKE VORTEX LATTICE: GENERAL EQUATIONS

We consider a Josephson vortex crossed with the pancake lattice in the framework of the modied London model. On scales which are much larger than both the distance between CuO2 planes and the in-plane coherence length ab , the pancake vortex stack could be considered as an ordinary vortex line at temperatures significantly lower than the evaporation temperature3,15 . The same approach can be also used for the description of the Josephson vortex far from the nonlinear core. The JV current acts on PVs through the Lorentz force causing their displacements along JV, which can be interpreted as a local inclination of the PV lines away from the c-axis. In turn, the local tilt of the PV stacks induces an additional current along the c-axis which redistributes the bare JV eld. Such physical picture can be described with one-component PV displacement u = (u, 0, 0) which does not depend on the x-coordinate (Fig. 1). The free energy functional FP J can be written as FP J = 1 8 d3 R h2 + hp hp p

+ hJ 2 + hJ hJ + 2hp hJ + 2hp hJ , (1) where hp and hJ are the magnetic elds of PV lines and JV, respectively, and is the penetration-depth tensor, = (/x, /y, /z). In the considered coordinate system the tensor has only the diagonal components xx = yy = 2 , zz = 2 with anisotropic penetration c ab depths ab and c . The eld hp is determined by the displacement u of PVs through the London equation (see, for instance,2 ) hp + hp ez +

= 0
i

d z (2)

u(Yi , z )ex z

(r Ri () u(Yi , z )ex ), z

where Ri () = (Xi , Yi , z ) is the equilibrium position of z the i-th PV line, r = (x, y, z) while ez and ex are unit vectors along the z and x axes, respectively. (Here we have accepted that the parametric equation ri () (with z parameter z ) describing the i-th vortex line takes the form of x() = Xi + ui (), y() = Yi , z() = z.) The z z z z in-plane coordinates of the unshifted lines Xi and Yi are expressed through the distances a and b (see Fig. 1 b) between PVs and PV rows as Xi = al/2 + aj and Yi = bl with integer l and j. In our approach, the eld of the Josephson vortex also obeys the London equation hJ 2 ab 2 hJ 2 hJ 2 = 0 (y)(z), c 2 z y 2 (3)

in u. Using the integral representation of -function, the equation (2) can be rewritten as hp + hp +
,

D
\

= 0
i

d z

d3 q iqr iqRi () 1 2 z ez (1 iqx ui () qx u2 ()) z e e i z (2)3 2 ui () z (1 iqx ui ())ex . z z (4)

E

-

FIG. 1. The JV crossed by PV stacks which are shifted by the currents induced by JV: a) 3D sketch depicts the deformation of the PV lattice in the dierent CuO2 planes, b) 2D sketches show the deformation of the PV lattice in a CuO2 plane for the dense (left) and dilute (right) PV lattices. The lled circles correspond to the unshifted PVs while the open ones represent the PVs shifted due to the interaction with JV currents. The shaded area images the nonlinear JV core region. The dashed-dotted lines mark the rows of the unshifted PVs while the dashed curve shows the deviation of PVs from these rows.

The eld hp of PV lines changes on dierent space scales. The rst scale is determined by the characteristic gradient of the displacement u(y, z) and usually is much larger than the distance a between PVs. The second scale is dened by the discreteness of the PV lattice and it is about a. To separate the contribution to the free energy from these scales, we introduce the Fourier variables u(ky , kz ):
/b

u(Yi , z) =
/b

dky 2

dkz u(ky , kz )ei(ky Yi +kz z) , 2

(5)

and u(ky , kz ) = b
Yi

dzu(Yi , z)ei(ky Yi +kz z) ,

(6)

where -functions should be smoothed on a scale of the Josephson vortex core. The JV core size along the z-axis is xed by the interlayer distance s, while the core length along the y-direction is limited by the condition that the current along the c-axis can not exceed the maximum interlayer current jc c0 /(8 2 2 s). In the presence c of PVs, the current across the layers consists of both the current of JV itself and the current born by the local inclination of PV lines. Therefore, the core size along the y-axis J can be renormalized in the presence of PVs and should be calculated self-consistently (see next section). Furthermore, the space variable y could be replaced by y y0 in the argument of the -function with 0 y0 b since the PV lattice can be arbitrary shifted from the center of JV. However, we take y0 = 0 which corresponds to the energetically more preferable position7 . Next, in order to nd the distribution of the magnetic eld in the vortex system and the energy of JV, we will minimize the free energy (1) as a functional of the displacement u. The elds hp and hJ can be obtained using equations (2,3) with the displacement u xed by the minimization of (1). Then, the energy of JV, EJ , dened as the dierence of the free energies (1) with and without JV, will be derived. This energy includes the self energy of JV and the change of the free energy of the PV lattice born by the interaction with JV. We will use the elastic approximation, i.e., the free energy (1) and the magnetic eld of PVs (2) will be expanded up to the second order 3

where the domain of variation of kz is restricted by the inequality |kz | 1/s born by the layerness of the system. Substituting (5) into (4) and using the well-known equality z z i d exp i(k q)Ri () = (2)3 (Bz /0 ) Q (k q Q), where Q = (Qx , Qy , 0) are the vectors of the reciprocal lattice (Qx = 2m/a, Qy = (2n + m)/b with integer m and n), one gets the expansion of the eld of PVs in series with respect to the displacement u: hp = h(0) + h(1) [u] + h(2) [u], np = n(0) + n(1) [u] + n(2) [u], p p p p p p h(i) + h(i) = n(i) p p p where n n
(0) p (1) p

(7)

= e z 0
i

2 (r R ), i dky dkz 4 2

= Bz
Q

(ez iQx + ex ikz )u(ky , kz )eiQx x ei(ky Qy )y eikz z , dky dkz dky dkz u(k)u(k ) n (2) = Bz p 16 4
Q
1 Q2 ez Qx kz ex ei(kz +kz )z eiQx x ei(ky +ky Qy )y . x 2 (8)

The term of hp with Q = 0 corresponds to the continuous approximation and varies on the large scale, while terms with Q = 0 are related to the eld components changing on the scale of a. It is easy to see that only terms with Qx = 0 will give contribution to the part of the free energy describing the interaction between PVs and JV, because the eld hJ does not depend on x and all terms with Qx = 0 vanish after integration over x. Therefore, it is con(1) (1) venient to divide hp and np into two components: (Q) (1) (Q) (Q) (1) np = np + ex n , hp = hp + ex h , where hp and p p np include summands with Qx = 0 while h and n do p p not vary with x. Then, the free energy functional Fcross , containing only terms dependent on the displacement u, can be introduced as Fcross = FP J FP FJ , where FP and FJ are the free energies of the unperturbed PV lattice and the bare JV, respectively. Using equations (7), we obtain the expression for Fcross as Fcross = 1 8 1 + 8
(Q) (Q) d3 R np hp + 2h(0) n(2) p p (Q)

U44 = + +

2 Bz 4

Qx =0

Q2 x 2 1 + 2 kz + 2 Q2 + 2 (ky Qy )2 ab ab x ab Q2 x + 2 (Qy ky )2 ab

1+ 1+

2 Q2 ab x
2 2 kz ab

(1 +

2 kz 2 Q2 + 2 (k Q )2 + c x y c y 2 2 2 2 (c ab )Qx kz 2 2 kz + 2 Q2 + 2 (ky Qy )2 ) ab ab x ab

1 . 2 (1 + 2 Q2 + 2 (ky Qy )2 + 2 kz ) c x c ab

(12)

d3 R h n + 2hJ n . p p p

(9)

The expressions for U44 and U66 represent sums over the reciprocal lattice vectors with Qx = 0, while the summation in the last term of (10) is performed only over the reciprocal lattice vectors with Qx = 0. The function f (q) in (10) appears due to smoothing of function in eq. (3) and can be evaluated as f (q) 1 in the rectangular region |qz | 1/s, |qy | 1/J and f 0 outside that area. The summation in the expression (11) for the shear elastic energy was done by Brandt16 in the limit ky /b:
2 U66 = C66 ky ,

The rst contribution comes from the terms with Qx = 0 and depends only on the short-scale variations of hp . It is determined by the shear deformation and the tilt deformation. The second part describes the interaction of PVs with JV and with the current generated by PVs along the y-axis. Using the equations (8), the free energy Fcross can be rewritten in term of Fourier variables u(ky , kz ): Fcross dky dkz 1 (U66 (ky ) + U44 (ky , kz )) u(k)u(k) = 2 4 2 Bz 0 dky dkz ikz u(k) + 4 4 2
Qx =0

(13)

where the shear elastic modulus C66 is expressed as C66 = (Bz 0 )/(8ab )2 for a0 = 0 /Bz < ab , while C66 = ab /(6a0 )2 /(42 )2 exp(a0 /ab ) for 0 ab a0 > ab . The tilt energy was obtained in2,17 : U44 = + for kz U44 = Bz 0 k2 ln 1 + 2 z 2 4 32 2 ab ab + K0
2 kz 4 ab ln 2 c 2 ab 2 2 K0 + (kz / 2 ) + 2 c

(14)

K0 = 2/b, while 3.68

2 Bz 0 ln(a2 /ab ) 2 0 + (4ab )4 32 2 2 c ef 2 2 kz = C44 f kz

f (kz , ky Qy ) (Bz /20 )ikz u(k) , 2 1 + 2 kz + 2 (ky Qy )2 c ab with the shear energy U66 =
2 Bz 4

(10)

(15) for kz K0 . Performing summation over Qy in the second term of equation (10) (see Appendix A), we nally obtain the free energy functional: Fcross = dky dkz 2 2 1 (U44 + U66 ) u(k)u(k) 2 , (16)

Qx =0

Q2 x 1 + 2 Q2 + 2 (ky Qy )2 ) ab x ab (11)

Q2 x , 1 + 2 (Q2 + Q2 ) x y ab

and the tilt energy

Bz Bz u(k)u(k) ikz (kz , ky ) u(k) ikz 4 20 0 b 2 1 + 2 kz ab sinh cosh

2 1 + 2 kz b/c ab

where is dened by the equation (kz , ky ) = 2c

2 1 + 2 kz b/c cos ky b ab

(17)

for ky < min(/b, 1/J ) and kz < 1/s while 0 outside that rectangular area. In the case of small values of wave vector k (ky /b and kz /b), the discreteness of PV lattice is irrelevant and the function coincides with the Fourrier image of the bare JV eld, but is modied substantially for larger ky or kz . The minimization of the free energy functional (16) determines the displacement u as u(k) = ikz (k) Bz . 2 2 4 U44 + U66 + (Bz kz /40 )(k) (18)

scales larger than the distance between PV lines even near the JV core. Thus, the continuous approximation is applicable in the whole space. In this case |ky | < 1/J < /b and, therefore, the cosine and hyperbolic functions in (17) can be expanded in the series. Hence, the function can be rewritten as = 0 . 2 2 1 + 2 kz + 2 ky c ab (21)

In order to describe the eld distribution of a JV in the crossing lattice, the averaged magnetic induction BJ along the x-direction generated by both JV and inclined PV lines can be introduced. By substituting the found displacement (18) into equations (7,8), the magnetic induction of JV BJ = hJ + h is rewritten as p BJ = +Bz
Qy

Substituting this expression for into (19,20) and omitting the dierence between k and q, the equations for the dense PV lattice (which determine the eld distribution and the energy of JV) are deduced as
1/J

BJ = and EJ =

1/J

dqz dqy 0 eiqy y+iqz z 2 2 (2)2 1 + 2 qy + 2 qz c ab ikz u(k) dky dkz ei(ky Qy )+ikz z , 2 1 + 2 (k Q )2 + k 2 2 (2) y c y z ab (19)

2 1 + 2 qy + c

1/s 0 eiqy y+iqz z 2 q 2 + q 2 B 2 /(4(U ab z 44 z z

dqy 2

1/s

dqz 2

2 + C66 ky ))

(22)

2 0 8

1/J 1/J

dqy 2

1/s 1/s

dqz 2 (23)

where qy and qz are the wave vectors of a bare JV (|qy | < 1/J , |qz | < 1/s), while the wave vectors k of the PV lattice are restricted also by the rst Brillouin zone of the PV lattice (|ky | < min(1/J , /b), |kz | < 1/s). To get the energy EJ it is necessary to add Fcross to the energy of the JV itself. Obviously, the energy of a JV in the presence of PV lines is always lower than one of a bare JV. Indeed, for the displacement of PVs u determined by equation (18), the energy Fcross takes the minimum value which is smaller than zero, since Fcross = 0 at u = 0. Finally, the energy of JV in the crossing lattice obeys the equation EJ = 2 0 8 1 dqy dqz 2 2 (2)2 1 + 2 qy + 2 qz c ab
2 kz (k)(k) dky dkz . (20) 2 2 (2)2 U44 + U66 + (kz Bz /40 )(k)

1 . 2 q 2 + 2 q 2 + B 2 q 2 /(4(U 2 1 + c y 44 + C66 qy )) z z ab z

The last undened parameter, J , can be obtained from the condition |BJ (y J , z = 0)/y| (4/c)jc : J 2 s c
1/J 1/s

dqy
1/J

2 2 1 + 2 qy + 2 qz + c ab

dqz 1/s 2 qy 2 2 qz Bz /(4(U44

2 + C66 qy ))

(24)

2 Bz 32 2

Equations (18-20) together with the condition that the current density along the c-axis should be smaller than the maximum current density jc , determine completely the behavior of the PV lines and the JV. However, in further analysis it is convenient to investigate the dense (s a) and dilute (s a) pancake lattices separately.
II. JOSEPHSON VORTEX IN THE PRESENCE OF DENSE PV LATTICE

The region of integration is shown in Fig. 2a. The rectangular domain of possible wave vectors replaces the usual elliptical one due to a peculiar core structure of JV. In anisotropic London model, the core of an ordinary vortex is dened by the elliptical stream line of the persistent current having the depairing value18 . However, in our case the maximum value of qz is determined by the layerness of the medium while the largest value of qy is restricted by the Josephson critical current along the c-axis. The rectangular domain of wave vectors (Fig. 2 a) can be divided into screened (|kz | 1/b) and remote(|kz | < 1/b) subdomains. The rst one corresponds to the region where one can roughly neglect the weak logarithmical dependence on kz in eq. (14) to express the tilt energy as:
2 U44 U 44 + C 44 kz , 2

(25)

For the case of the dense PV lattice, many PV rows are placed on the nonlinear JV core (Fig. 1b, left sketch). It means that the magnetic eld of a bare JV varies on 5

with U 44 = (Bz 0 /32 2 4 ) ln(1 + kz /(2 + b2 )), ab ab C 44 =

Bz 0 32 2 2 c

ln

2 ab 2 +(k 2 / 2 )+2 b z c

and kz

1/bs.

and

\

EJ

D


-


2 0 2 ef f 16 ab c

cut C66 ab + ln 2 J U 44 ab J

(27)

6FUHHQHG  UHJLRQ 5HPRWH  UHJLRQ

-

6FUHHQHG  UHJLRQ

8

-

while the renormalized penetration depth ef f is exab pressed as ef f = ab 2 + ab

2 Bz . 4U 44

where cut min(c , min(b, max(bc /ef f , ab

bc ab /ef f ))), ab

]

(28)

for the dense PV lattice (a) and the dilute PV lattice (b). In the dense case (a), the region of the available wave vectors k of the PV lattice coincides with the accessible domain of wave vectors q; the continuous approximation (21) for is always valid. The approximation (25) is correct in the screened domain of wave vectors, while it fails in the remote region. In the dilute case (b), the continuous approximation is correct only in the region |qy | /b; |qz | /b; the non-screened region |qy | > /b is not accessible for k. The continuous approximation for (k) is broken in the pinning region |ky | < /b; /b |kz | 1/s.

The physical reason of the renormalization of the in-plane penetration depth is related to the screening of the JV eld by currents born by the local inclination of PV lines. From eq. (26) it is easy to see that the size of the non3LQQLQJ 3LQQLQJ 6  UHJLRQ linear JV core also decreases due to the interaction of  UHJLRQ JV and PVs. The similar conclusion was given earlier E by Koshelev7 , who considered the additional phase vari1RQVFUHHQHG  UHJLRQ ation of the order parameter born by the displacement - of PVs. However, the shear contribution to the renormalization of EJ and J was neglected in7 , which could 6FUHHQHG be done only for J > ab (see equations (26) and (27)). - 6FUHHQHG - 8  UHJLRQ  UHJLRQ 6 In the opposite case, i.e., when the London penetration 5HPRWH depth exceeds the JV core size, the shear deformation  UHJLRQ becomes relevant and, as a result, J decreases with Bz slower than it was proposed in7 . To understand how the eld of JV is distributed in 1RQVFUHHQHG  UHJLRQ the real space, we rederive the results considering the free energy functional of the displacement u dened as a function of the spatial coordinates. In the limit s a, the JV eld varies on scales larger than the distance between PVs even near the JV core. This means that the eld hp along the x-axis can be averaged out on the FIG. 2. The integration region in equations (19) and (20) scale larger than a:
\


]

h 2 p ab

2 h 2 h u p p 2 = n = Bz . c 2 2 z y z

(29)

However, the short range variations of the eld hp give the shear energy U66 and the tilt energy U44 . After ignoring the slow logarithmic dependence on k in the expression for U44 , one can conclude that the density of the tilt energy in the real space is U 44 u2 (y, z), while the density of the shear energy is U66 = C66 (u/y)2. Thus, the free energy functional is expressed as Fcross = 1 8 d3 R 4C66 u y
2

The expression (25) is completely wrong in the remote region in which it is necessary to use equation (15). By using approximation (25) for the tilt energy, the integrals in equations (23) and (24) are easily evaluated (Appendix B) and we get J max c s ef f ab , c s ef f ab ab (26)

+ 4U 44 u2 + h Bz p

u u . + 2hJ z z

(30)

The rst three terms represent the elastic energy (born by shear, electromagnetic tilt and Josephson coupling tilt rigidity, respectively), but the last term is related to the 6

interaction of the PV lines with the current generated by JV. In order to get the complete set of equations for the displacement u and the averaged magnetic induction BJ , we have minimized the functional (30) and have added together equations (3) and (29): 4C66 2u BJ + 4U 44 u 2Bz = 0, y 2 z u 2 BJ 2 BJ 2 = 0 (y)(z) + Bz . BJ 2 c ab 2 z y 2 z

(31)

This set of equations is applicable if the continuous approximation is valid (J a) and, strictly speaking, only when the tilt energy U44 (kz ) can be replaced by the constant U 44 . The last condition fails on the distances far from JV (z 2 + y 2 / 2 > b2 ). In this remote region, the ef constant U 44 has to be substituted by C44 f 2 /z 2 . Besides, if ab > s, the parameter U 44 should be replaced by C 44 2 /z 2 near the JV core (z < ab /). Even though we consider only the situation when the set of equation (31) is valid, i.e., the case ab < s and the region z 2 + y 2 / 2 < b2 , the solution of equations (31) seems to be quite complicated. The relation between the displacement u and the magnetic induction BJ , which is obtained from the rst equation of (31), becomes nonlocal due to the shear rigidity of the PV lattice: u= Bz 8 C66 U 44

d y

BJ (, z) |y|/ y y e z

(32)

where = ab C66 /(U 44 2 ) ab is the characterisab tic length of a nonlocality. However, the nonlocality is irrelevant if the space scale of the variation of BJ is substantially large then , i.e., if J ab . In such a case, the equations (31) for BJ and u can be decoupled Bz BJ , 4U44 z 2 BJ 2 BJ BJ (ef f )2 2 = 0 (y)(z). c ab 2 z y 2 u=

(33)

with the length = ef f . However, the set of equaab ab tions (31) becomes incorrect in the region z 2 +y 2 / 2 > b2 which cuts o that length as b. Thus, the expresab sion (35) coincides with the earlier obtained equation (27) in the studied case J > ab . The results (34,35) can be interpreted in terms of the eective anisotropy parameter ef f = c /ef f which governs the JV lattice. ab Since ef f > ab , the eective anisotropy ef f is reab duced in the presence of PVs with respect to the bare one = c /ab . The similar anisotropy ef f was earlier introduced7 as a ratio ef f = J /s, but these two different denitions of ef f give the same value in the case J > ab when the shear deformation is irrelevant. Here, we discuss how the core size and the JV energy are changed with the magnetic induction Bz if s > ab . For quite high magnetic inductions Bz B1 = (0 /2 ) (s/ab )2 , the size of the nonlinear core J is ab smaller than ab and the shear contribution to the free energy is important. The second logarithmic term in the JV energy (27) can be omitted, and the core size obeys the equation J (Bz ) sa. With decreasing of induc1/4 tion, the core size increases proportionally to Bz and reaches ab at Bz B1 . At low elds, the shear interaction between rows is irrelevant, the JV core size becomes J = c s/ef f sa/ab , and the energy of JV is deterab mined by the logarithmic term in (27). Below the eld 0 /2 , at which the distance a between PVs exceeds ab ab , the currents generated by PVs practically do not inuence on the JV eld and, thus, the renormalization of ab , J and the EJ vanishes. In the case of ab > s the physical picture is dierent from the previous situa tion. The core size obeys the law J sa at elds Bz > 0 /(s)2 . Below this eld, the eective value of the in-plane London penetration depth ef f 2 /a (28) is ab ab still larger than ab , while the JV core size is saturated as J = s. This means that JV eld shows dierent behavior far from JV (z 2 + y 2 / 2 > b2 / 2 ), where the redistribution due to the local inclination of PV lines is still important, and close to the JV core.
III. JOSEPHSON VORTEX IN THE PRESENCE OF DILUTE PV LATTICE.

The equation (33) for induction BJ is the London equation with the renormalized in-plane penetration depth ef f . Therefore, the eld distribution BJ , not far from ab the center of the Josephson vortex (z 2 + y 2 / 2 b2 ), can be approximated as BJ = 0 2ef f c ab K0 z 2 /(ef f )2 + y 2 /2 , c ab (34)

where K0 (x) is a modied Bessel function of zero order. Using the free energy functional (30) and equations (31), it is easy to show that the energy of JV is determined by the eld in its center, i.e., EJ = 0 /(8)BJ (y J , z s): EJ = 2 0 16 2 ef f c ab ln( /s) ab (35)

Far from the JV center, z 2 + y 2 / 2 > b2 / 2 , the JV eld varies slowly which causes the smooth variation of the displacement u even for the case of the dilute PV lattice (a > s). In that spatial region, the continuous approximation is still valid. On the other hand, near the JV core (|y| < b), the JV current increases quite fast inducing a large displacement of the PV stack placed on the center of JV. In this case, the continuous approximation is not applicable. To describe such physical situation, we consider the wave vector area of k divided into two domains (Fig. 2b). In the rst interval |ky | < /b and |kz | < /b, the function can be still roughly approximated by equation 7

elds Bz > 0 /(s)2 . The origin of this behavior is that the additional current along the c-axis induced by tilted PV stacks is much smaller than jc near JV core in the eld interval 0 /2 Bz 0 /(s)2 , but the inclinaab tion of all PV lines can still cause the renormalization 2 /b dqy /b dqz of the JV eld on scales larger than a. In the eld inEJ (a s) 0 0 / 2 s2 , the main contribution terval 0 /2 < Bz 8 /b 2 /b 2 ab to the Josephson vortex energy (37) comes from the rst 1 term related to the tilt elastic rigidity (born by Josephson 2 2 2 2 1 + 2 qy + 2 qz + Bz qz /(4(U44 + U66 )) c ab coupling of PVs) and shear elasticity of the PV lattice. 2 b Strictly speaking, from our rough estimation of (20), we 0 + ln can not conclude how strongly EJ is suppressed in that 16 2 c ab s eld interval, i.e. in the presence of the dilute PV lat1/s 3 dkz Bz 0 tice. Nevertheless, the pinning energy (last term in (37)) . (36) 128 3 a2 2 /b U44 + Bz 0 kz /(8c ab a) c ab could be the same order of magnitude as the rst and the second terms in (37) in elds Bz 0 /2 and may ab The rst term comes from the spatial region far from the decrease EJ substantially. center of JV while the second and the third terms are Another interesting possibility arising due to the related to the vicinity of the JV center. The screening crossing lattice pinning is the rearrangement of the of the bare JV eld vanishes near JV (non-screened PV lattice in the presence of the JV sublattice. In the region in k-space) which determines the second term in in-plane magnetic elds Bx , JVs form a triangular lat(36). The last term in (36) represents the energy gain tice with distances aJ and bJ between JVs (see inset in due to the strong interaction between the PV line placed Fig. 3a). In general, the PV sublattice and the JV subon the JV core and the JV currents (the energy gain lattice are not commensurate: aJ = pb with integer p. of a PV stack placed on a JV in the limit s ab This means that the considered one-component displace7 and Bz 0 was calculated by Koshelev ). Since the ment of PVs u = (u(y, z), 0, 0) (the shifted PV lattice last term is sensitive to the mutual position of JV and shown in Fig. 3a) does not provide the energy gain comthe nearest PV line, this contribution can be called as ing from the crossing lattice pinning since the PV rows the crossing lattice pinning. Using the results of Apcan not occupy the centers of JVs. However, the PVs can pendix B and taking into account that the evaluation be rearranged in order to occupy all JVs (the trapped 2 Bz 0 kz /(8c ab a) C 44 kz is held in the pinning rePV lattice shown in Fig. 3b) if the PV lines shift also gion (kz > /b), the energy of JV is nally obtained: along the y-direction: u = (ux (x, y, z), uy (x, y, z), 0). The crossing lattice pinning decreases the free energy 2 0 2 C66 ab 21 C 44 2 3 2 ab of the trapped PV lattice, while the additional shear EJ + 2 2 ef f 2 ef f U 44 ab b ab U 44 2 b 16 c ab deformation acts in the opposite way through increasing ab the free energy. For the case Bx < Bz , the energy gain 2 2 0 0 b cut + + ln ln related to the trapped PV lattice is calculated by norb 16 2 c ab s 16 2 c ef f ab malizing the last term of equation (37) per unit volume: 2 0 b s , (37) b s Bx 0 arctan . arctan Etr = 4ac U 44 /C 44 sb + C 44 /U 44 4ac U 44 /C 44 sb + C 44 /U 44 (39) 2 2 where = Bz 0 /(32 c ab C 44 U 44 ) < 1 is the dimensionless function depending quite slowly on Bz and But, in order to trap the PV lattice, the total displacethe numerical parameters 1 and 2 are about unity. ment of PVs along the y-axis between the two nearest JV Next, we will discuss how the renormalization of the rows, i.e., on the scale aJ , should be about b. Following JV energy comes in with increasing of the z-component the simple analysis21 , the extra shear deformation (inset of the magnetic eld. At low elds, Bz 0 /2 in Fig. 3b) is about b/b b/aJ (b is the change of ab (a ab ), the rst term and the last term in (37) can the distance between rows of PVs) and the energy loss be omitted and the expression for the energy of a bare Eshear can be estimated as: JV reported earlier in19,20 is reproduced 2 b Bx Eshear C66 (40) C66 2 c 0 aJ Bz . (38) ln EJ = 16 2 c ab s with numerical constant 1. For the case s ab , the shear elastic energy (40) is strongly suppressed in the For the case ab > s, the renormalization of JV enelds Bz < 0 /(s)2 where the crossing lattice pinning ergy becomes relevant at Bz 0 /2 , i.e., earlier than ab is active, since C66 is exponentially small if a > ab (13). the JV core size starts to decrease which occurs only in 8

2 2 0 /(1 + 2 ky + 2 kz ), while 0 b/(2c ab kz ) c ab in the second region |ky | < /b and /b < |kz | < 1/s (pinning region in Fig. 2b). Following this approach, the energy of JV is evaluated as




-

D
-

-

E

-

Therefore, the trapped PV lattice seems to be realized as soon as a > s. In the opposite case, ab s, the transformation22 from the shifted PV lattice to the trapped PV lattice occurs when the energy gain Etr exceeds the energy loss Eshear . It happens in a certain out-of-plane eld between the eld 0 /(s)2 , at which the crossing lattice pinning is activated, and the eld Bz 0 /2 , where the shear elastic energy rapidly deab creases. Next, we discuss the dierence between the considered trapped state and the chain state proposed for the crossing lattice7 . The trapped state is related to the rearrangement of PVs on the scale aJ between the nearest rows of JVs. On the other hand, the chain state is associated with the creation of an extra PV row (an interstitial in the PV lattice) on a JV, but the inuence of the neighbouring JVs is completely ignored. As a result, the trapped and chain states have the different in-plane eld dependence of the out-of-plane transition elds. The out-of-plane transition eld23 between the shifted and trapped PV lattices does not depend on Hab in contrast to the Hab -dependent out-of-plane eld7 of the destruction of the chain state. Since the analysis7 is correct only in the case of s ab and a ab , the transformation of the PV lattice discussed here seems to be more likely in the case ab > s.
IV. PHASE DIAGRAM OF VORTEX LATTICE IN TILTED MAGNETIC FIELDS

F

FIG. 3. The dierent substructures of the crossing lattice: a) the shifted PV lattice characterized by one-component displacement along JVs, b) the PV lattice trapped by the JV sublattice for the case when the distance aJ between JVs exceeds the distance b between PV rows, c) the trapped PV lattice for aJ < b, i.e., in the case of the eld orientations very close to the ab-plane. The dotted lines depict the rows of unperturbed lattice (crossings of these lines and lled circles mark the positions of unshifted PVs) in all sketches. Dashed-dotted lines indicate the rows of the trapped PV lattice which are deformed in order to match with JV sublattice. The arrows directed from the lled circles to the open ones show the two component displacement of PVs for the cases represented in sketches b) and c). Inset in a): the JV sublattice with lattice parameters aJ and bJ (lines mark the CuO2 planes). Inset in b): the additional deformation of the PV lattice which is required for trapping of PVs by JVs. The upper sketch is the equilateral triangle of the unperturbed PV lattice which is incommensurate with JV lattice (aJ = pb with integer p). The lower sketch is the isosceles triangle of the PV lattice matched with the JV sublattice (aJ = p(b + b)).

In this section we discuss the vortex lattice structures formed at dierent eld orientations. The tilted lattice consists of mono-oriented vortices and transforms continuously from the tilted PV stacks in elds near the c-axis (Fig. 4a) to the long JV strings connected by PV kinks for the eld orientations close to the ab-plane (Fig. 4b). On the other hand, the tilted lattice is topologically different from the crossing vortex structure (Fig. 4c), and they replace each other via phase transition6 . For the analysis of the vortex phase diagram in tilted elds, the free energy of the crossing and tilted vortex structures will be compared. We concentrate on the case s < ab , when, according to Bulaevskii et al.6 and Koshelev20 , the tilted lattice is energetically preferable above the lockin transition8 . We will consider a thin superconducting platelet with the c-axis perpendicular to the plate. In this geometry the lock-in transition occurs at very low elds6 Bz (1 nz )0 /(42 ) ln(s/ab ) with demagab netization factor nz (1 nz 1). For the eld oriented close enough to the c-axis, tan = Bx /Bz , the free energy of the tilted lattice Ft can be tilt 2 2 evaluated as Ft = Ft0 + 1 C44 (k = 0)Bx /Bz in analogy 2 7 0 to the analysis given in ref. . Here, Ft represents the free energy in the absence of the in-plane magnetic eld, tilt while the tilt modulus is expressed as C44 (k = 0) = ef f ef f 2 Bz /4 + C44 with C44 dened in (15) for the case of Bz 0 /(42 ). As a result, we have: ab

tice while the interaction of PVs and JVs is taken into account through the renormalization of the JV energy: Fc
MMM

2 B2 0 Bz Hc2 Bx Bz + x+ EJ . + ln 2 2 8 32 ab Bz 8 0

(42)

D

+ tion (20) in which the lower limits of integrations are

c

The renormalized JV energy, EJ , is dened by equa-

ab

restricted by the conditions qy , ky 1/aJ and qz , kz 1/bJ . The tilted lattice is energetically preferable in the elds 2 oriented near the c-axis because Ft Bx , while Fc Bx , i. e., Ft < Fc for low Bx . The phase boundary between the tilted lattice and the crossing structure can be obtained from the condition Ft = Fc which is rewritten in the form: Bx

E

ab

2 EJ Bz . 0 1.842 /(4ab )4 + 0 Bz /(64 2 2 ) ln(Hc2 /Bz ) c 0 (43)

F

ab

The transition from the tilted lattice to the crossing structure occurs at the eld oriented quite close to the c-axis for high anisotropic superconductors due to: 1) the high energy cost of the inclination of PV stacks in the tilted lattice related to the electromagnetic interaction of PVs, and 2) the decrease of the JV energy in the crossing lattice structure. For the dense PV lattice Bz 0 /(s)2 and ab > s, equation (43) can be simplied: Bx C66 22 ab U 44 2 J ef f ab ab
2 Bz . + ln(Hc2 /Bz ) Bz 2

0 4.3 2 2 ab

(44)
FIG. 4. The 3D sketches of the dierent vortex structures in the tilted magnetic eld with the components Hc and Hab along the c-axis and in the ab-plane, respectively: a) the tilted vortex lattice near the c-axis (TI), when the current between CuO2 planes is much smaller than the critical value jc , i. e., the Josephson strings linking PVs are not developed; b) the tilted vortex lattice far away from the c-axis (TII), when the JV strings are formed; c) the crossing vortex lattice.

Next, we will study the eld orientations close to the ab-plane, Bx > Bz . Here, the electromagnetic interaction between PVs in the tilted lattice is not so important and the free energy in the low c-axis elds Bz < 0 /2 ab is reduced6 to: Ft
2 0 HJ Bz 0 Bx Bx ln + + , 8 32 2 ab c s2 Bx 4

(45)

B2 B2 0 Bz Hc2 Ft z + + x ln 8 32 2 2 Bz 8 ab + 3.68
2 2 Bx Hc2 Bx 0 2 0 ln , + 4 B2 2 2 B 2(4ab ) z 64 c z Bz

(41)

2 where Hc2 = 0 /2ab . The rst two terms form the free energy for Bx = 0. The third term is the in-plane magnetic energy, the fourth one comes from the electromagnetic interaction of the inclined PVs, and the last contribution is connected with the Josephson coupling of PVs. The free energy of the crossing lattice Fc consists of two contributions from the PV sublattice and the JV sublat-

where HJ = 0 /(42 ) ln(s/ab ). The rst two terms ab are related to the energy of JV strings while the last one is associated with the energy cost of the formation of PV kinks. In more detail, the PV kink generates the inplane current which decreases with the distance r from the kink center as 1/r up to a critical radius r0 of the region with 2D behavior where the current along the c-axis is about the maximum possible current jc 6 . At larger distances, the in-plane current decays exponentially. Simple evaluation gives r0 = s for the PV kink6 . We note, that the tilted vortex lattice in the considered angular range of the magnetic eld orientations seems to exist as a kink-walls substructure, where kinks (belonging to dierent vortices) are collected in separated walls 10

+.2H

parallel to the yz-plane20 . For the kink-wall substructure of the tilted lattice, the contribution to the free energy (45), attributed to the PV kinks, is slightly reduced in the high in-plane magnetic elds Bx > 0 /s220 , which can be taken into account through renormalization HJ = 0 /(82 ) ln(Hc2 /Bx ). ab In the considered eld interval, Bx > Bz , Bz 0 /2 , the renormalization of the JV energy in the ab crossing lattice structure vanishes. However, the interaction of PV and JV sublattices still manifests itself through the crossing lattice pinning: 0 Hc1 Bz 0 Bx Fc = ln + + 8 32 2 ab c s2 Bx 4
2 Bx

  

7,

&,

    

7, &,

+.2H

Bz

Bx 0 arctan 16 2 2 ab

x H /H0

Bx /H0 x + Bx /H

, (46)

     

with Hc1 = (the critical radius r0 for the in-plane currents of a PV stack is about ab 6 ), x H0 = 0 /s2 and H = 0 /2 . The third term is the ab energy of the unperturbed PV lattice, while the last term corresponds to the crossing lattice pinning contribution which can signicantly decrease the free energy Fc in the in-plane eld interval 0 /2 Bx 0 /s2 . The ab dierence between the crossing lattice pinning contributions to the free energy in the cases of low Bx < Bz and high Bx > Bz in-plane elds (see equations (39), (46)), emerges because the number of PV lines is sucient to occupy all JVs (Fig. 3b) at Bx < Bz while some JV strings do not carry PV rows (Fig. 3c) in the opposite case. By analyzing equations (45) and (46), we can conclude that, at least for Bx 0 /2 and ab 2 Bx 0 /s , the tilted lattice exists near the ab-plane since the condition Ft < Fc is held due to the inequality HJ < Hc1 . The tilted lattice is replaced by the crossing lattice with increasing the out-of-plane magnetic eld above Bz = Bx /. However, it is dicult to determine the contour of the possible phase line between the crossing and tilted vortex structures, since it requires the more precise calculations of the free energies Fc and Ft in the region Bx < Bz . In the intermediate in-plane magnetic elds 0 /2 Bx 0 /s2 , the crossing ab lattice pinning could make the crossing structure to be more energetically preferable with respect to the tilted lattice. In that case, the crossing lattice (CII, Fig. 3b) with a < aJ transforms into the crossing lattice (CIII, Fig. 3c) with the extremely dilute PV sublattice a > aJ at the angle = arctan Bx /Bz arctan . Therefore, we nd a complicated picture of phase transitions between the tilted vortex structure and the crossing vortex structure in the case s < ab . The proposed phase diagram24 is shown in gure 5. As it was suggested earlier7 and according to our calculations by using equation (43), the tilted vortex structure (TI) of inclined PV stacks (see Fig. 4a) can be replaced by the crossing lattice quite close to the c-axis (see phase diagram obtained for = 500, inset in Fig. 5). 11

0 /(42 ) ln(ab /ab ) ab

7,,

&,,

  

,-

2H

&,,,


7,,
 

+,-N2H





FIG. 5. The proposed phase diagram of the vortex solid phase in the oblique magnetic elds calculated using equations (39), (40), (43), (45) and (46) with parameters mentioned in the text, T = 45 K, = 100 and = 1. The dotted line is the line Bx = Bz , while the shaded area marks the region inside which the transition from the crossing lattice C(II) to the tilted lattice T(II) or the crossing lattice C(III) happens. The arrows from enframed TI and TII are directed toward the regions where these vortex structures are realized. Inset: the part of the phase diagram close to the c-axis for strongly anisotropic superconductors with = 500 (s > ab ) and T = 45 K; the solid line marking the transition from the tilted lattice (TI) to the crossing lattice is obtained from eq. (43), while the dashed line, corresponding to the same transition, is calculated by using eq. (6) of Ref.7 . The parameters are chosen to give some insight to the behavior of Bi2 Sr2 CaCu2 O8 in the oblique magnetic elds.

Nevertheless, in the high c-axis magnetic elds, the calculated in-plane magnetic elds of this transition are higher than ones obtained by using the model7 . This dierence comes from the shear contribution to EJ which was omit-

ted in ref.7 . The crossing lattice, which exists in a wide angular range, can have dierent substructures. At high enough out-of-plane elds Bz > 0 / 2 s2 , the shifted PV sublattice is realized in the crossing lattice structure (CI, Fig. 3a). In this substructure, the JV currents shift the PVs mostly along the x-axis. The shifted phase can transform into the trapped PV lattice (CII, Fig. 3b), when the energy gain related to the crossing lattice pinning exceeds the energy needed for the additional shear deformation (the dashed line in Fig. 5 separating CI and CII has been obtained from the condition Etr = Eshear ). Around the line Bx = Bz , the lattice CII can be changed by the tilted lattice (TII) with JV strings linked by PV kinks (Fig. 4b) or by the crossing lattice structure (CIII, Fig. 3c) at which all PV stacks are placed on a few JVs. The domain in the Hc Hab phase diagram with the lattice CIII is determined by the condition Fc < Ft where Ft and Fc are dened by the equations (45) and (46), respectively. With increasing temperature, the region of the lattice CIII becomes narrower and disappears at a certain temperature (see Fig. 6). The proposed phase diagram suggests the possibility of the re-entrant tiltedcrossing-tilted phase transition as the magnetic eld (at least with low B 0 /2 or high B 0 /s2 absoab lute value) is tilted away from the c-axis to the ab-plane. Such possibility for low elds was earlier mentioned in the works11,13 in which the interaction of crossed sublattices was not considered. Moreover, we note that the instability of the tilted lattice was found numerically by Thompson and Moore25 (for 100) only at the intermediate eld orientations 0 < 1 < < 2 < 90 , which could also support the discussed scenario. The parameters taken for the Hc Hab phase diagram at T = 45 K (Fig. 5) and the Hab T phase diagram at the magnetic eld orientation Bz = Bx / (Fig. 6) were chosen to give some insight into the behavior of vortex array in BSCCO in the tilted magnetic elds (see further discussion) as 2 A A ab = 2000/ 1 T 2 /Tc , ab = 30/ 1 T /Tc , , Tc = 90 K, = 100 ( = 500 for inset in s = 15 A Fig.5 and = 150 for inset in Fig. 6).
V. CONCLUSION

pearance of the hexagonal order along the c-axis found by neutron measurements32 . With further tilting of the magnetic eld, the linear dependence of the c-axis meltm ing eld component Hc (Hab ) abruptly transforms into a weak dependence2729 , which, as was shown28 , can not be explained in the frame of the model7 . Such behavior suggests a phase transition in the vortex solid phase in tilted magnetic elds in Bi2 Sr2 CaCu2 O8+ , which was detected in the recent ac magnetization measurements29,31 .



+,-N2H

    

+,-N2H



7,,

 7,,     &,,,       

&,,,

7.

        

7.

This theoretical investigation was partially motivated by the recent intensive experimental studies of the vortex lattice melting transition2630 as well as transitions in the vortex solid phase29,31 in Bi2 Sr2 CaCu2 O8+ single crystals. The observed linear decay of the c-axis m melting eld component Hc with in-plane elds26,27 was interpreted7 as an indication of the crossing vortex lattice. Thus, the tilted lattice could be replaced by the crossing vortex structure quite near the c-axis. According to our calculations, the angle where such transition may occur is about 7 at Bz 100 Oe and 500 while that angle reaches 14.5 in the higher out-of-plane eld Bz = 500 Oe, which correlates well with the disap-

FIG. 6. The Hab T phase diagram of the vortex solid phase at the eld oriented near the ab-plane (Bx /Bz = = 100). The region of the crossing lattice phase (CIII) becomes narrower and nally disappears with increasing temperature. Inset: the same phase diagram for = 150. The dotted lines correspond to the experimenm 2 tally found temperature dependence Hab 1 T 2 /Tc of the in-plane characteristic elds of the vortex lattice melting transition in Bi2 Sr2 CaCu2 O8+ in the tilted elds30 (for instance, symbols represented in the inset exhibit the temperature dependence of the maximum in-plane eld of the vortex lattice melting transition).

12

As was mentioned by Ooi et al.31 , the behavior of the new anomaly of the local magnetization in BSCCO attributed to the phase transition in the vortex solid slightly reminds the peak eect related to the vortex pinning, which, in turn, could be induced by the vortex trapping by planar defects21 . Such analogy, as well as a very weak depen dence of the c-axis magnetic eld component Hc of the novel phase transition on in-plane magnetic elds31 in a wide angular range, could suggest the tansition from the shifted PV sublattice (CI) to the trapped PV sublattice (CII) in the crossing lattice structure. Our estrap timation of the c-axis eld Hc of the transition from trap CI to CII (Hc 450 Oe for = 100 at T = 45 K (see Fig. 5))33 is in a reasonable agreement with experimental ndings31 Hc (T = 45 K) 430 Oe. Near the ab-plane, the properties of the observed anomaly changes abruptly and the eld Hc sharply goes to zero31 , which may indicate a transformation23 in the JV lattice or a trace of the phase transition from TII to CIII. At higher in-plane elds, the found step-wise behavior28 of the vortex lattice melting transition may be related to the existence of one more phase transition in the solid phase. Interestingly enough, all characteristic in-plane elds of the vortex lattice melting transition depend on temperature30 2 proportionally to 1 T 2 /Tc , which is similar to the calculated temperature dependence of the phase transition from CIII to TII (see Fig. 6). In summary, we discussed the crossing lattice structure in a strongly anisotropic layered superconductor in the framework of the extended anisotropic London theory. The renormalization of the JV energy in the crossing lattice structure was calculated in the cases of the dense PV lattice as well as the dilute PV lattice. It was shown, that the crossing lattice pinning can induce the rearrangement of the PV sublattice in the crossing lattice structure as soon as the out-of-plane magnetic eld becomes lower than a certain critical value. The free energy analysis indicates a possibility of the re-entrant tiltedcrossing-tilted lattice phase transition with inclination of the magnetic eld away from the c-axis to the ab-plane in the case of ab > s.
APPENDIX A. APPROXIMATE SUMMATION IN EQUATION (10)

with real numbers r, t, and . Next the sum 1 (ky , kz ) = 0 Qy =2n/b f (kz , ky 2 Qy )/(1 + 2 kz + 2 (ky Qy )2 ) needs to be estimated. c ab By using inequailty kz 1/s, we obtain f (kz , ky Qy ) fJ (ky Qy ) where fJ 1 for |ky Qy | 1/J and zero otherwise. In the case of the dense PV lattice (b J ) we retain only the term with n = 0 in the sum and get the expression 1 (ky > 1/J ) = 0 and 1 (ky 1/J ) = 2 2 0 /(1+2 kz +2 ky ). Taking into account the inequality c ab 2 ky < 1/J 1/b and 1 + 2 kz b/c < b/(s) 1, ab one can rewrite 1 (ky < 1/J , b J ) (ky , kz ). Thus, we come to the equation (16). For the case of the dilute PV lattice (J b), many terms give contributions to the sum 1 N 2 0 n=N 1/(1+2 kz +2 (ky 2n/b)2 ), since inequalc ab ity |ky 2n/b| < 1/J is held until n exceeds N 1. Thus, the function 1 is estimated as: 1 (ky , kz ) (ky , kz ) 0 b
1/J

dx 2 1 + 2 kz + 2 x2 c ab (A2)

Finally, we obtain 1 = (1 (ky , kz )) with (ky , kz )

2 cosh( 1 + 2 kz b/c ) cos(ky b) ab 2 sinh( 1 + 2 kz b/c ) ab

2 arctan

c 2 1 + 2 kz ab

(A3)

At |kz | 1/s, it is easy to show that J /b 1 and the approximation 1 = used in (16) is excellent. Only for kz 1/s, the function 1 can dier from the function by a factor about unity in the case of the dilute PV lattice (the factor is 1 0.5 in the framework of our rough consideration). However, the correct estimation of the value of the factor depends on the type of the smoothing function and requires more precise analisys than one in the framework of the London approach. Therefore, we can always assume 1 = in our semiquantitative consideration.
APPENDIX B. EVALUATION OF INTEGRALS IN EQUATIONS (23,24,36)

In this appendix, the integrals in equations (23,24,36) are evaluated. We start with the dense PV lattice, a f (kz , ky 2n/b) (Bz /20 )ikz u(k) 0 ikz u(k) = J . In the region qz < 1/b, the tilt energy is small (15). 2 1 + 2 kz + 2 (ky 2n/b)2 c ab n= Therefore, the denominators of the integrands in (23,24) 2 ikz 1 (ky , kz )u(k) + (Bz /20 )kz (ky , kz )u(k)u(k). (A1) are substantially larger than ones in the case qz > 1/b and we can roughly neglect the contribution related to The equation (17) for (ky , kz ) can be directly obtained the region qz < 1/b. Using equation (25) for the tilt by using the well-known mathematical equality energy in the domain qz > 1/b, the integrals (23,24) can be rewritten as follows: sinh(r) 1 = r2 + (t + 2n/)2 2r cosh(r) cos(t) n

Our aim is to sum the sequence:

13

1/J

1/s

I =2
1/J

dqy
1/b

dqz

2 f (qy ) 2 2 2 2 2 2 1 + 2 qy + 2 qz + Bz qz /(4(U 44 + C 44 qz + C66 qy )) c ab

, (B1)

2 2 where f (qy ) = 2 /32 3 for (23) and f = J qy for (24). After multiplying numerator and denominator by factor of 0 2 2 U 44 + C 44 qz + C66 qy , the expression (B1) is reduced to 1/J 1/s 2 2 2 (U 44 + C 44 qz + C66 qy )f (qy ) 2 2 2 2 2 (1 + 2 qy + 2 qz )(U 44 + C66 qy ) + (1 + 2 qy + 2 qz )C 44 + c c ab ab
2 Bz 4

I=2
1/J

dqy
1/b

dqz

2 qz

(B2)

2 2 2 The term (1 + 2 qy + 2 qz )C 44 can be neglected with respect to Bz /4 in the denominator. Indeed, the maximum c ab 2 2 2 2 2 2 2 value of c qy C 44 (see eq. (25)) is about c C 44 1/J Bz 0 /J Bz , while the maximum value of 2 qz C 44 is ab 2 2 2 c C 44 1/J (J /s) which is even smaller, since J < s (see eq. (26)). Next, the integration in (B2) over qz can be taken easily: 1/J 2 U 44 + C66 qy (qy ) 2 2 Bz /4 + 2 (U 44 + C66 qy ) ab

I=4
1/c

2 dqy f (qy )

1 C 44 1 + 2 /4 + 2 (U 2) s c qy Bz 44 + C66 qy ab

(B3)

where the function , dened as (qy ) = arctan

1 qy s 2 2 1 + Bz /(42 (U 44 + C66 qy )) ab

arctan

1 qy b

2 2 1 + Bz /(42 (U 44 + C66 qy )) , ab

determines the lower cutting value 1/cut of qy . Namely, by using (26) we can assume that the argument of the rst arc tangent in the last expression is larger than 1, for most of values of qy . Then, the value of (qy ) is about /2 if the argument of the second arc tangent is smaller than 1; otherwise is close to zero. Thus, we obtain the expression for cut : b 2b cut 1/2 2 Bz 2 1 + 42 (U +C 2 ) (ef f )2 b2 2 b2 2 b2 2 ab 44 66 cut 1 2 + + 4 2 2 ab 2 1
ab

min(c , min(b, max(bc /ef f , ab

bc /ef f ))), ab

where we denote = C66 /U 44 ab and take into account that cut should be smaller than c . As a result, the integral (B3) can be evaluated as:
1/J 2 dqy f (qy )

I =4
1/c

C 44 1 + 2 2 /4 + 2 (U 2) s Bz 44 + C66 qy ab

1/J 1/cut

2 dqy f (qy )

1 c qy

2 U 44 + C66 qy 2 2 Bz /4 + 2 (U 44 + C66 qy ) ab

(B4)

In case of the dense PV lattice, the rst integral is small and can be neglected. In addition, we also can omit the term 2 2 2 C66 qy in the expression Bz /4 + 2 C66 qy . Finally, we have ab I(a J ) = 2 1 c ef f ab
1/J 1/cut 2 f (qy )dqy qy

1+

C66 2 qy . U 44

(B5)

14

2 2 After taking f (qy ) = J qy in (B5) and by ignor2 ing C66 qy /U 44 in the case J > ab or by neglecting unity in the opposite case, we obtain the expres-

sion J max

c s ef f ab

c s ef f ab

It coincides with

(26) since ab . The energy of JV (27) is eas2 ily derived if one puts f (qy ) = 2 /(32 3 ) in (B5). 0 Next, we roughly estimate the rst integral in equation 2 2 2 (36). The inequality (1 + 2 qy + 2 qz )C 44 Bz is c ab still correct in the domain qz /b, qy 1/b for Bz 0 /2 . Thus, we can get the estimation (B4) c also for the dilute case (a > s). However, in contrary to the dense PV lattice, the contribution related to the rst term in (B4) remains important: I(a J ) = 4 + C 44 U 44 b(ef f )2 ab 2 c ef f ab
2 /b 1/cut 1 /b 1/c 2 dqy f (qy )

dqy

2 f (qy ) qy

1+

C66 2 qy . (B6) U 44

Here, we have introduced numerical parameters 1 and 2 , since the upper limits of integration are not well dened. The corresponding contribution to the energy is obtained from the last equation for f = 2 /32 3 as pre0 sented in the text.

On leave from All Russian Electrical Engineering Institute, 111250 Moscow, Russia + On leave from Faculty of Sciences, University of Montenegro, PO Box 211, 81000 Podgorica, Montenegro, Yugoslavia 1 G. Blatter, M. V. Feigelman, V. B. Geshkenbein, A. I. Larkin, V. M. Vinokur, Rev. Mod. Phys. 66, 1125 (1994). 2 E. H. Brandt, Rep. Prog. Phys. 58, 1465 (1995). 3 A. I. Buzdin and D. Feinberg, J. Phys. (Paris) 51, 1971 (1990); S. N. Artemenko and A. N. Kruglov, Phys. Lett. A 143, 485 (1990); J. R. Clem, Phys. Rev. B 43, 7837 (1991). 4 L. Bulaevskii and J. Clem, Phys. Rev. B 44, 10234 (1991). 5 D. Feinberg, Physica 194C, 126 (1992). 6 L. N. Bulaevskii, M. Ledvij, and V. G. Kogan, Phys. Rev. B 46, 366 (1992). 7 A. E. Koshelev, Phys. Rev. Lett. 83, 187 (1999). 8 D. Feinberg and C. Villard, Phys. Rev. Lett. 65, 919 (1990). 9 A. Sudbo, E. H. Brandt, and D. A. Huse, Phys. Rev. Lett. 71, 1451 (1993). 10 L. L. Daemen, L. J. Campbell, A. Yu. Simonov, V. G. Kogan, Phys. Rew. Lett. 70, 2948 (1993). 11 G. Preosti and P. Muzikar, Phys. Rev. B 48, 9921 (1993). 12 W. E. Lawrence and S. Doniach, in Proceedings of 12th International Conference on Low Temperature Physics, Kyoto 1970, edited by E. Kanda (Keigaku, Tokyo, 1970), p. 361.

M. Benkraouda and M. Ledvij, Phys. Rev. B 51, 6123 (1995). 14 D. A. Huse, Phys. Rev. B 46, 8621 (1992). 15 L. N. Bulaevskii, S. V. Meshkov, D. Feinberg, Phys. Rev. B 43, 3728 (1991). 16 E. H. Brandt, J. Low Temp. Phys. 26, 735 (1977); the expression of C66 for a0 > ab was presented, for instance, in G. Blatter, V. Geshkenbein, A. Larkin, H. Nordborg, Phys. Rev. B 54, 72 (1996). 17 A. E. Koshelev and P. H. Kes, Phys. Rev. B 48, 6539 (1993). 18 V. G. Kogan and L. J. Campbell, Phys. Rev. Lett. 62, 1552 (1989). 19 J. R. Clem, M. W. Coey, and Z. Hao, Phys. Rev. B 44, 2732 (1991). 20 A. E. Koshelev, Phys. Rev. B 48, 1180 (1993). 21 I. F. Voloshin et al., JETP 84, 1177 (1997). 22 One of the possible physical pictures of the replacement of the shifted PV lattice by the trapped PV lattice is the sequence of the phase transitions. At each of those transitions, the distance L between the nearest JV rows, occupied by PV rows, increases by aJ and takes values L = aJ (lattice in Fig. 3b), L = 2aJ (every second JV row is occupied), etc. 23 The structural transitions between the Josephson vortex lattices with dierent periods4 , which are related to the layerness of the medium, cause the nonmonotonic dependence of aJ on the in-plane eld28 . Such behavior of the JV sublattice can induce the complicated alternation of the shifted and trapped states. 24 For simplicity, we have assumed that the demagnetization factor nz is very close to unity, i.e., the lock-in transition and Meissner state are far below the considered region of Bz . 25 A. M. Thompson and M. A. Moore, Phys. Rev. B 55, 3856 (1997). 26 S. Ooi, T. Shibauchi, N. Okuda, and T. Tamegai, Phys. Rev. Lett. 82, 4308 (1999). 27 J. Mirkovi, E. Sugahara, and K. Kadowaki, Physica c 284B-288B, 733 (2000). 28 J. Mirkovi, S. E. Savelev, E. Sugahara, and K. Kadowaki, c Phys. Rev. Lett. 86, 886 (2001). 29 M. Konczykowski, C. J. van der Beek, M. V. Indenbom, and E. Zeldov, Physica 341C-348C, 1213 (2000). 30 J. Mirkovic, S. Savelev, E. Sugahara, and K. Kadowaki, Meeting Abstracts of the Physical Society of Japan, vol. 55, issue 2, part 3, p. 504, 55th Annual Meeting, September 22-25, 2000; J. Mirkovic, S. Savelev, E. Sugahara, and K. Kadowaki, Physica C, to be published. 31 S. Ooi, T. Shibauchi, K. Iteka, N. Okuda, T. Tamegai, Phys. Rev. B 63, 20501(R) (2001). 32 J. Suzuki, N. Metoki, S. Miyata, M. Watahiki, M. Tachiki, K. Kimura, N. Kataoka, and K. Kadowaki, Advances in Superconductivity XI, proc. of the 11th International Symposium on Superconductivity (ISS98), November 16-19, 1998, Fukuoka, Japan, p. 553 (Springer Verlag). 33 Note that the anisotropy parameter of BSCCO can change in quite wide range depending, in particular, on the oxygen stoichiometry (see for instance G. Balestrino et al., Phys. Rev. B 51, 9100 (1995)).

13

15