1: % Dear Editor,
2: %
3: % We submit this text for publication in Physical Review Letters.
4: %
5: % Two figs are attached as files.eps
6: %
7: % Vladimir Kogan
8: % Ames Lab, Physics
9: % Ames IA 50011-3020, USA
10: % 515/294-8080, fax: 515/294-0689
11: %
12: % Suggestions for possible referees:
13: %
14: % 1. G.Blatter, BLATTERJ@ITP.ETHZ.CH
15: % 2. A.Koshelev, Argonne, koshelev@msd.anl.gov
16: % 3. L. Bulaevskii, Los Alamos, lnb@viking.lanl.gov
17: % 4. J.Thompson jrt@utk.edu, D. Christen, Oak Ridge, DKC@ORNL.GOV
18: % 5. P. Gammel, Lucent, Bell Labs, plg@lucent.com
19: % 6. M. Franz, franz@benz.pha.jhu.edu (check the address)
20: %
21: \documentstyle[aps,prl,epsfig,multicol]{revtex}
22: %\documentstyle[aps,prl,epsfig]{revtex}
23: %\documentstyle[aps,prl]{revtex}
24: %\documentstyle[aps,preprint]{revtex}
25: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
26: %TCIDATA{Created=Fri May 08 03:57:26 1998}
27: %TCIDATA{LastRevised=Tue May 12 07:44:10 1998}
28: %\input {epsf}
29: \begin{document}
30: \draft
31: %\twocolumn[\hsize\textwidth\columnwidth\hsize\csname
32: %@twocolumnfalse\endcsname
33: \title{Effect of fluctuations on vortex lattice structural transitions in superconductors}
34: \author{A. Gurevich}
35: \address{Applied Superconductivity Center
36: University of Wisconsin, Madison, Wisconsin 53706}
37: \author{V.G. Kogan}
38: \address{Ames National Laboratory and Department of Physics and Astronomy, Iowa State University,
39: Ames Iowa 50011}
40: \date{\today}
41: \maketitle
42: \begin{abstract}
43: The rhombic-to-square transition field $H_{\Box}(T)$ for cubic and
44: tetragonal materials in fields along [001] is evaluated using the nonlocal
45: London theory with account of thermal vortex fluctuations. Unlike extended
46: Ginzburg-Landau models, our approach shows that the line $H_{\Box}(T)$
47: and the upper critical field $H_{c2}(T)$ do not cross
48: due to strong fluctuations near $H_{c2}(T)$ which suppress the
49: square anisotropy induced by the nonlocality. In increasing fields, this
50: causes re-entrance of the rhombic structure in agreement with recent
51: neutron scattering data on borocarbides.
52: \end{abstract}
53: \pacs{PACS numbers: \bf 74.60.-w, 74.60.Ge, 74.70.Dd}
54: %\narrowtext
55: \begin{multicols}{2}
56:
57: Complex behavior of vortex lattices (VLs) even in cubic superconductors
58: like Nb has been known for a long time \cite{Nb}. Recent progress in
59: understanding the evolution of VLs with the magnetic field $H$ and
60: temperature $T$ became possible due to availability of large high quality
61: crystals of borocarbides \cite{Paul}, which triggered a score of
62: small-angle neutron scattering (SANS) \cite{n4,n7,n8}, scanning tunneling
63: \cite{n5,sakata}, and decoration experiments \cite{vinn}.
64: Of a particular interest is the ubiquitous structural transition
65: between rhombic and square VLs in increasing fields along the
66: $c$ axis
67: %in cubic Nb \cite{Nb}, V$_3$Si \cite{v3si} and
68: of tetragonal borocarbides \cite{n4,n7,n8}.
69:
70: For this case, the vortex repulsion is {\it isotropic} within the
71: standard London and Ginzburg-Landau (GL) theories, so that vortices
72: should always form the hexagonal Abrikosov lattice, which provides the
73: maximum vortex spacing for a given flux density $B/\phi_0$ ($\phi_0$ is
74: the flux quantum). There is no coupling between the VL and the crystal
75: in these models; as a consequence, the VL orientation is arbitrary and no
76: VL structural transitions are expected. A full microscopic
77: theory of the mixed state contains this coupling, but involves
78: self-consistent calculations of the gap and current distributions, a
79: formidable task even for materials with the GL parameter
80: $\kappa\sim 1$ \cite{micro}. The situation simplifies in high-$\kappa$
81: materials (like borocarbides), for which one can utilize a more
82: transparent nonlocal London model\cite{nl}. Within this approach, the VL
83: coupling to the crystal is provided by the basic nonlocal
84: relation between the current density and the vector potential,
85: $J_{\alpha}({\bf r})=
86: \int Q_{\alpha\beta}({\bf r}-{\bf r}')A_{\beta}({\bf r}')d^3{\bf r}'$,
87: where the kernel $Q$ depends on the Fermi surface\cite{nl,book}, the
88: pairing symmetry\cite{Franz}, and the field orientation. Here, we
89: consider cubic or tetragonal s-wave materials in fields along the $c$ axis
90: so that $Q ({\bf r})$ has the square symmetry.
91:
92: The kernel $Q(r)$ decays over the nonlocality range
93: $\rho = f(T,\ell)\xi_0$, where $\ell$ is the mean-free path for nonmagnetic
94: scatterers and $\xi_0$ is the BCS zero-$T$ coherence length. The function
95: $f$ decreases slowly with $T$ and is suppressed strongly by scattering
96: \cite{nl,book}. The nonlocality adds to the intervortex interaction a
97: short-range potential $V(x,y)$ with the symmetry of the crystal. In the
98: low field limit, $V$ is irrelevant and the VL is triangular; still,
99: $V$ removes the orientational degeneracy and locks the VL onto certain
100: crystal direction. With decreasing intervortex spacing $a(B)$, the
101: potential $V $ drives the triangular VL into a square at a field
102: $H_{\Box}(T)$. The transition curve $H_{\Box}(T)$ is the subject of
103: this work.
104:
105: The nonlocal London model describes correctly the observed structure and
106: orientation of rhombic VLs in small fields \cite{n8} and the structural
107: evolution toward the square \cite{vinn}. Moreover, the fast increase of
108: $H_{\Box}$ with increasing impurity concentration predicted by the model
109: has been verified by SANS \cite{imp}. Still, there is
110: an open question on what happens when the applied field $H > H_{\Box}$
111: keeps increasing. The nonlocal London model per se implies that the
112: square VL should persist all the way up to $H_{c2}$ since the shorter the
113: intervortex distance, the stronger the role of the square-symmetric
114: potential $V$. This conjecture is supported by calculations based on the
115: GL theory extended to include the 4th order gradient terms
116: \cite{n5,gl1,gl2}; the model predicts that the lines $ H_{\Box}(T)$
117: and $H_{c2}(T)$ should cross as sketched in the inset to Fig. 1.
118:
119: However, recent SANS data for ${\rm LuNi_2B_2C}$ \cite{esk} indicate
120: that instead of crossing $H_{c2}(T)$, the line $H_{\Box}(T)$
121: curves up to avoid $H_{c2}$ and becomes two-valued, see Fig. 1.
122: The upper branch of $H_{\Box}(T)$ marks the re-entrant transition of
123: the rhombic VL in increasing fields, which then evolves toward the
124: triangular VL as $H\to H_{c2}$. In this Letter we argue that vortex
125: fluctuations combined with nonlocal effects can account for this
126: unexpected behavior.
127:
128: Below we evaluate the mean-squared amplitude $\overline {u^2}$ of thermal
129: vortex fluctuations using the elastic energy of the deformed VL. Unlike
130: the case of {\it isotropic} superconductors, an important
131: role in this energy is played by VL {\it rotations} relative to
132: the crystal. We show that $\overline{u^2}$ remains finite at the
133: transition line $H_{\Box}(T)$, but diverges as $B$ approaches $H_{c2}(T)$.
134: In the vicinity of $H_{c2}(T)$, the anisotropic nonlocal potential $V(x,y)$
135: is averaged out by fluctuations, and the interaction becomes isotropic.
136: As a result, the rhombic VL becomes preferable, turning into the
137: hexagonal Abrikosov VL as $B\to H_{c2}$. The re-entrance of the rhombic VL
138: in increasing fields occurs if $\overline{u^2}\sim\xi_0^2 $, i.e., the
139: amplitude of fluctuations needed to wash out the nonlocal effects is much
140: smaller than that required for the VL melting. This re-entrance,
141: therefore, can happen in nearly cubic materials, in which vortices are not
142: split into ``pancakes" and fluctuations are weak. We evaluate the shape
143: of the $H_{\Box}(T)$ curve using model parameters of ${\rm
144: LuNi_2B_2C}$ for which in the clean limit $\rho(T)\simeq
145: (1\div 2)\xi_0$\cite{book,***}. We reproduce the shape of this curve
146: seen in the SANS data \cite{esk}. \\
147:
148: Let us consider the equilibrium square VL with
149: diagonals along [100] and [010] directions, which
150: we take as $x$ and $y$ axes. We are interested mainly in the high field
151: region, where the VL can be considered as incompressible \cite{ehb}.
152: Then the displacement of vortices $u_i(x,y;z)$ ($i=x,y$)
153: satisfy div${\bf u}=\partial_xu_x+\partial_yu_y =0$ and the symmetric
154: strain tensor
155: $u_{ij}=(\partial_ju_i+\partial_iu_j)/2$ has only two independent
156: components: $u_{xx}$ and $u_{xy}$. Since the VL orientation is locked on
157: the crystal, rotations of VL about
158: [001], $\omega_{xy}=(\partial_xu_y-\partial_yu_x)/2$, cause an energy
159: increase. Then the energy density $U$ of the deformed VL reads
160: \cite{mir}:
161: \begin{eqnarray}
162: 2 U=c_s u_{xx}^2 + c_x u_{xy}^2 + c_{\omega}\omega_{xy}^2
163: +c_{44}(\partial_z{\bf u})^2 .
164: \label{f}
165: \end{eqnarray}
166: Our notation is motivated by the following. A uniform
167: displacement $u_x = \epsilon x$, $u_y = -\epsilon y$ ($ \epsilon\ll 1$ is
168: a constant) is of a special interest: it is this deformation
169: that transforms the square above $H_{\Box}$ into a rhombus for
170: $B < H_{\Box}$. This deformation was named ``squash"\cite{kc}, so
171: that $c_s$ is the squash modulus. Since this is a second order phase
172: transition, $c_s$ must vanish at $B=H_{\Box}$. Further, the ``simple"
173: $x$-directed shear is $u_x = \epsilon y$, $u_y=0$ so that only
174: $u_{xy}\ne 0$; thus the notation $c_x$ for this shear mode. The shear
175: energy depends on the shear direction, but the corresponding moduli can
176: all be expressed in terms of $c_x$ and $c_\omega$ \cite{mir}. We take
177: all moduli in Eq. (\ref{f}) except the tilt $c_{44}$ as practically
178: nondispersive \cite{ehb}.
179:
180:
181:
182: %Let us consider the equilibrium square VL with diagonals along [100]
183: %and [010] directions, which we take as $x$ and $y$ axes. Then
184: %the elastic energy density $U$ has the form
185: % \begin{eqnarray}
186: % 2U=c_{11}(div{\bf u})^2+ c_xu_{xy}^2-c_su_{xx}u_{yy}+ \nonumber \\
187: %c_{\omega}(\partial_xu_y-\partial_yu_x)^2/4+c_{44}[(\partial_zu_x)^2
188: %+(\partial_zu_y)^2] \label{f}
189: % \end{eqnarray}
190: %Here $u_{ik}=(\partial_i u_k+\partial_k u_i)/2$ is the strain tensor
191: %for the 2D vortex displacements ${\bf u}$,
192: %$c_{44}$ and $c_{11}$ are the tilt and bulk moduli, $c_s$ and $c_x$ are
193: %two shear moduli required by the tetragonal symmetry, $c_{\omega}$ is the
194: %rotational modulus, which account for the fact that the VL orientation is
195: %locked on the crystal, so rigid rotations of the VL about
196: %[001] increase in the elastic energy . The VL is stable if $c_s>0$,
197: %$c_h>0$, and $c_{\omega}>0$.
198:
199: %For an incompressible VL $(c_{11}\gg c_x)$, the in-plane strain tensor
200: %has only 2 independent components, $u_{xx}$ and $u_{xy}$, because
201: %$u_{xx}+u_{yy}=0$. In this case the moduli $c_s$ and $c_x$ correspond to
202: %two distinct deformation modes. A uniform "squash" mode
203: %$u_x = \epsilon x$, $u_y = -\epsilon y$ with $\epsilon\ll 1$ is of
204: %special importance because it is the deformation
205: % that transforms the square VL above $H_{\Box}$ into a rhombic VL for
206: %$B < H_{\Box}$. The corresponding``squash" modulus $c_s(B)$\cite{kc}
207: %and the elastic energy $U_s=c_s\epsilon^2/2$ both vanish at $B\approx
208: %B_{\Box}$, since $c_s(B) = c_0(B/B_{\Box}-1)$ \cite{mir}. Another
209: %shear mode $u_x =\epsilon y$, $u_y=0$ corresponds to the modulus $c_x$
210: %which remains finite at $B_{\Box}$.
211:
212: Next, we calculate the mean-squared vortex displacement
213: $\overline{u^2}(T,B)=\overline{u_x^2}+\overline{u_y^2} $.
214: Writing the total elastic energy in the Fourier space and utilizing the
215: equipartition theorem, we obtain \cite{rem1}:
216: % for $c_{11}\gg c_x$:
217: \begin{equation}
218: \overline{u^2}=\int\frac{qdqdk_zd\phi}{(2\pi)^3}\frac{2T}{q^2
219: c(\phi)/4+c_{44}(k_z,q)k_z^2},
220: \label{u}
221: \end{equation}
222: where $k_x=q\cos\phi$, $k_y=q\sin\phi$, and
223: $ c(\phi)=c_{\omega}+ c_x\sin^22\phi+ c_s\cos^22\phi$. With the large
224: fields tilt modulus $c_{44}(k)=c_{44}(0)/(1+\lambda^2k^2)$, integration
225: over $k_z$ and $q$ gives $\overline{u^2}=\overline{u_0^2} \eta$. Here
226: $\overline{u_0^2} \sim T\lambda/a^2\sqrt{c_xc_{44}(0)}$ is the
227: mean-squared displacement for the triangular VL with $c_{66}=c_x$ and
228: $c_{44}(0)=B^2/8\pi$ \cite{blat,ehb}. Taking $\overline{u_0^2} $ for a
229: hexagonal VL in a uniaxial material for ${\bf B}|| c$\cite{ehb}, we arrive
230: at
231: \begin{eqnarray}
232: \overline{u^2}&=&16\sqrt{2}\pi^2\lambda_a\lambda_c\xi_a T\eta
233: /\phi_0^2s(b)\,,
234: \label{uu} \\
235: s^2&=&b(1-b)^3\ln(2+1/\sqrt{2b}).
236: \label{s}
237: \end{eqnarray}
238: Here, $\lambda_{a,c}$ are penetration depths,
239: $b=B/H_{c2}$, and
240: \begin{equation}
241: \eta={2\sqrt{c_x}\over \pi} \int_0^{\pi}\frac{d\phi}{
242: \sqrt{c(\phi)}}=\frac{4}{\pi}\sqrt{\frac{c_x}{c_x+c_{\omega}}}
243: {\bf K}\Bigl(\frac{ c_x- c_s}{c_{\omega}+ c_x}\Bigr),
244: \label{r}
245: \end{equation}
246: where ${\bf K}(m)$ is a complete elliptic integral.
247:
248: As follows from Eqs. (\ref{uu})-(\ref{r}), $\overline{u^2}(T,B)$ remains
249: {\it finite} at the instability point $c_s=0$. Thus, vortex fluctuations
250: do not affect the mean-field character of the second order transition at
251: $H_{\Box}$. This unusual behavior occurs because the rotational
252: modulus $c_{\omega}$ is {\it finite}, otherwise $\overline{u^2} $ would
253: diverge logarithmically as $B\to H_{\Box}$.
254:
255: The moduli $c_x$ and $c_{\omega}$ in intermediate fields, $H_{c1}\ll B\ll
256: H_{c2}$, are estimated as $c_x\simeq \phi_0 B/(8\pi\lambda)^2$ and
257: $c_{\omega}\simeq\rho^2B^{2}/2\pi\lambda^2\sqrt{b}$ \cite{app}. For
258: $c_{\omega}/c_x\sim \rho^2/a\xi\ll 1$, we obtain
259: $\eta\simeq (2/\pi )\ln (16c_x/c_{\omega})$. Hence, the nonlocality
260: affects $\overline{u^2}$ via the weakly-varying factor
261: $\eta\sim\ln(\xi/\rho b^{1/4})$, i.e., $\overline{u^2}(T,B)$ is insensitive
262: to detailed behavior of $c_{\omega}(T,B)$ and $c_x(T,B)$.
263: Since $\overline{u^2}\propto (H_{c2}-B)^{-3/2}$ diverges only at $H_{c2}$,
264: the fluctuations suppress the weak VL-crystal coupling mainly near
265: $H_{c2}$. \\
266:
267: We now turn to evaluation of the curve $H_{\Box}(T)$. The
268: free energy $F$ of fluctuating VL can be written as:
269: \begin{equation}
270: F=\frac{B^2}{8\pi}\sum_G\frac{e^{-(C\xi^2 + \overline{u^2})G^2/2}}
271: {1+\lambda^2G^2+\rho^2\lambda^2G_x^2G_y^2}\,.
272: \label{fl}
273: \end{equation}
274: Here the sum runs over
275: $G_x = \pi\sqrt{2}(m-n)/a\tan^{1/2}(\beta/2)$,
276: $G_y=\pi\sqrt{2}(m+n)\tan^{1/2}(\beta/2)/a$, which form the reciprocal
277: lattice;
278: $\beta$ is the apex angle of the rhombic
279: cell, and $m,n$ are integers. The last term in the denominator
280: describes the nonlocal correction to the London theory
281: \cite{nl,gl1}. The factor $\exp(-C\xi^2G^2/2)$ accounts for the
282: finite size of the vortex core \cite{ehb}, which provides
283: a cutoff in the London model. The $B$ dependent quantity $C$ was
284: estimated theoretically as $0.5<C<4$ \cite{core}. By and large, the data
285: on the field dependence of SANS form-factors are consistent with
286: $C\simeq1$ \cite{peter}, the value we adopt below.
287:
288: Fluctuations enter the energy (\ref{fl}) via the Debye-Waller
289: factor $\exp(-\overline{u^2}G^2/2)$, which accounts for the thermal
290: smearing of the vibrating VL\cite{blat}.
291: For $\overline{u^2}>\rho^2$, the fluctuations wash out the nonlocal
292: corrections in $F$, making the averaged vortex interaction isotropic.
293: Hence, the re-entrant transition can be driven by thermal vortex
294: displacements of the order of $\rho\sim\xi_0$.
295: The instability field $H_{\Box}(T)$ thus can occur well
296: below the melting line, which in materials of interest here is rather
297: close to $H_{c2}$.
298:
299: To find $H_{\Box}(T)$ at which $c_s = \partial^2F/
300: \partial\beta^2|_{\beta =\pi/2}=0$, we
301: differentiate $F$ of Eq. (\ref{fl}) and obtain:
302: \begin{equation}
303: \sum_{mn}\frac{e^{-pg}}{d}\Bigl[(2pmn)^2+\Bigl(\frac{8m^2n^2}{d}-g\Bigr)
304: \Bigl(p+\frac{1}{d}\Bigr)\Bigr] =0\,.
305: \label{cs}
306: \end{equation}
307: Here, $p=[\pi C+ \chi(T)\eta(t, b)/s(b)]b$,
308: $g=m^2+n^2$, $d=\mu+g+\zeta (b)(m^2-n^2)^2$,
309: $\mu=1/2\pi b\kappa^2$.
310: The dimensionless control parameters $\chi$ and $\zeta$ quantify the
311: amplitude of thermal displacements and the nonlocal corrections:
312: \begin{equation}
313: \chi=\frac{16\sqrt{2}\pi^3\lambda_a\lambda_cT}{\phi_0^2\xi_a},
314: \qquad
315: \zeta =\frac{\pi}{2}b\left(\frac{\rho}{\xi}\right)^2 .
316: \end{equation}
317: Note that nonlocality enters also the exponent $p$ in Eq. (\ref{cs})
318: via the parameter $\eta$.
319:
320: For the further analysis, we assume that $\lambda (T)
321: =\lambda(0)/(1-t^2)^{1/2}$ and $\xi (T)=\xi(0)/(1-t^2)^{1/2}$,
322: where $t=T/T_c$ (qualitatively, our results do not change if other
323: plausible $T$ dependences are used).
324: Then
325: \begin{eqnarray}
326: \chi&=&\chi_0 t/\sqrt{1-t^2}, \qquad\qquad\quad
327: \zeta=\zeta_0b(1-t^2),
328: \label{aob} \\
329: \chi_0&=&\frac{16\sqrt{2}\pi^3\lambda_a(0)\lambda_b(0)T_c}
330: {\phi_0^2\xi_a(0)}, \quad
331: \zeta_0 =\frac{\pi}{2}\left(\frac{\rho}{\xi_0}\right)^2,
332: \label{ab}
333: \end{eqnarray}
334: In the clean limit $\zeta_0\sim 1$; with increasing scattering,
335: $\zeta_0(\ell)$ drops fast \cite{nl}. For
336: ${\rm LuNi_2B_2C}$ with $T_c=16\,$K, $\xi_0\approx 70$\AA,
337: $\lambda_a(0)\approx 10^3$\AA, and
338: $\lambda_c(0)\approx 1.2\times 10^3$\AA, we estimate $\chi_0\approx
339: 6.4\times 10^{-3}\ll 1$. The smallness of $\chi_0$ indicates that
340: fluctuations contribute little to the thermodynamics of stable VLs
341: \cite{blat,ehb}. As shown below, being crucial on the upper branch of
342: $H_{\Box}(T)$, fluctuations are negligible on the lower branch.
343:
344: Strictly speaking, $H_{\Box}(T)$ should be calculated
345: {\it self-consistently} taking into account the effect of fluctuations
346: on the relevant moduli \cite{scha}.
347: However, since $\overline{u^2}$ is finite at $H_{\Box}$, we may neglect
348: the thermal softening of $c_x$ and $c_{\omega}$. Then,
349: $H_{\Box}(T)$ is just a root of Eqs. (\ref{cs}),
350: which we find numerically.
351: The factor $\eta$ which enters
352: $p$ in Eq. (\ref{cs}) is a much weaker function of $t$ and $b$ than
353: $\chi(t)/s(b)$ ($\eta$
354: varies from 1.6 for $c_{\omega}/c_x=1$ to 3 for $c_{\omega}/c_x=0.1$).
355: For this reason we disregard variation of $\eta$, adopting
356: $\eta=2.7$ for ${\rm LuNi_2B_2C}$ with $c_{\omega}/c_x\simeq
357: 0.2$ at $H_{\Box}$ \cite{mir}.
358:
359: The results of the numerical solution of Eq. (\ref{cs}) are shown in Fig.
360: 2. It is seen that fluctuations do give rise to the
361: re-entrant square-to-rhombus transition in high fields, in a
362: qualitative agreement with SANS data of Fig. 1. In fact,
363: fluctuations change radically the VL phase diagram in high fields,
364: while weakly affecting the low-field branch of $H_{\Box}$.
365: The difference between $H_{\Box}(T)$ and
366: $H_{c2}(T)$ is significant (except the low-$T$ clean limit) which
367: justifies the use of the London model. As the ratio $\rho_0/\xi_0$
368: decreases (e.g., due to nonmagnetic impurities),
369: the region of the square VL on the $H-T$ diagram shrinks.
370: The raise of the lower branch of $H_{\Box}(T)$ has been seen on ${\rm
371: Lu(Co_xNi_{1-x})_2B_2C}$, for which the mean-free path
372: $\ell$ was suppressed by the Co doping \cite{imp}.
373:
374: It is worth noting that although the calculated curves $H_{\Box}(T,\ell)$
375: reproduce correctly qualitative features of the SANS data of Fig. 1,
376: actual position of the upper branch is sensitive not only to the accuracy
377: with which we know the elastic moduli and the parameters for their
378: evaluation, but also to the precise value of the empirical cutoff constant
379: $C$, see Eq. (\ref{f}). The information on the actual position
380: of the upper branch of $H_{\Box}(T,\ell)$ is still scarce, and we hope to
381: refine our approximations when the data are available. \\
382:
383: Now we comment briefly on the possible effect of the VL transition
384: on the flux pinning. Since the instability of the square VL at $H_{\Box}$
385: is not accompanied by divergence of $\overline{u^2} $, the critical current
386: density $J_c$ evaluated within the collective pinning theory,
387: should not be sensitive to the VL transition. Indeed, the correlation
388: function of vortex displacements $\langle u({\bf r})u({\bf r'})\rangle$
389: can be evaluated with the help of Eq. (\ref{u}) in which $T$ is replaced by
390: $\gamma_p\exp i{\bf k}({\bf r-r'})$, where $\gamma_p$ is the pinning
391: parameter\cite{blat}. At $H_{\Box}$, the squash modulus vanishes,
392: but $\langle u({\bf r})u({\bf r'})\rangle$ remains of the same order as
393: for a triangular London VL, to the accuracy
394: of the weak logarithmic factor $\eta\sim \ln(\xi/\rho b^{1/4})\sim 1$.
395: Thus, contrary to the claim of Ref. \cite{gl2}, neither the pinning
396: correlation length nor $J_c$ are significantly affected by the squash
397: softening near $H_{\Box}$.
398:
399: The two-valued $H_{\Box}(T)$ and the re-appearance of the
400: triangular VL at $H\to H_{c2}$ are generic features which are not
401: limited to nonmagnetic borocarbides. The
402: low-$T$ SANS experiments on antiferromagnetic
403: ${\rm TmNi_2B_2C}$ have revealed the triangular VL near $H_{c2}$, which
404: evolves into a square as the field {\it decreases}\cite{es}. At
405: this stage the effect of antiferromagnetic ordering upon VLs is unclear,
406: but fluctuations could certainly contribute to the re-entrant
407: VL transition in ${\rm TmNi_2B_2C}$ as they do in ${\rm LuNi_2B_2C}$.
408: Similar behavior was seen in ${\rm YNi_2B_2C}$ \cite{Mona}. Another
409: candidate for studying effects of vortex fluctuation is
410: $\rm V_3Si$, in which the rhombic VL was observed at $T<5^{\circ}$K and
411: $H=10\,$kOe ($H_{\Box}\simeq 15\,$kOe); as $T$ increases at the fixed
412: field, the rhombic VL evolves toward the hexagonal one as $T\to T_{c2}(H)$
413: \cite{n7}.\\
414:
415: In conclusion, we present a model of the structural VL transition at
416: $H_{\Box}(T)$ affected by thermal fluctuations of vortices.
417: We show that the curves $H_{\Box}(T)$ and $H_{c2}(T)$ do not cross,
418: instead $H_{\Box}(T)$ becomes two-valued. The upper branch of
419: $H_{\Box}(T)$ corresponds to a re-entrant transition of the
420: rhombic VL, in accordance with recent SANS observations
421: on ${\rm LuNi_2B_2C}$.\\
422:
423:
424: We are grateful to M. Dodgson and M. Yethiraj for useful discussions. M.
425: Eskildsen kindly provided the data for Fig.1. The work of AG was
426: supported by the NSF MRSEC (DMR 9214707) and AFOSR. Ames Laboratory is
427: operated for US DOE by the Iowa State University under contract No.
428: W-5405-Eng-82.
429:
430:
431:
432: \begin{references}
433:
434: \bibitem{Nb} {\it Anisotropy Effects in
435: Superconductors}, edited by H. Weber, Plenum, New York, 1977;
436: D.K. Christen {\it et al.}, \prb {\bf 21}, 102 (1980).
437:
438: \bibitem{Paul}P.C. Canfield, {\it et al.}, Physics Today, {\bf 51}, 40 (1998).
439:
440: \bibitem{n4}U. Yaron {\it et al.}, Nature {\bf 382}, 236 (1996);
441: M.R. Eskildsen {\it et al.}, \prl {\bf 78}, 1968 (1997); {\bf 79}, 487 (1998).
442:
443: \bibitem{n7}M. Yethiraj {\it et al.}, \prl{\bf 78}, 4849 (1997);
444: {\bf 82}, 5112 (1999).
445:
446: \bibitem{n8}D. McK Paul {\it et al.}, \prl{\bf 80}, 1517 (1998).
447:
448: \bibitem{n5}Y. De Wilde {\it et al.}, \prl{\bf 78}, 4273 (1997).
449:
450: \bibitem{sakata}H. Sakata {\it et al.}, \prl{\bf 84}, 1583 (2000).
451:
452: \bibitem{vinn}V.Ya. Vinnikov {\it et al.}, \prb {\bf 64}, 02445XX (2001).
453:
454: %\bibitem{v3si}D.K. Christen {\it et al.}, Physica B{\bf 135}, 369
455: %(1985); V.G. Kogan {\it et al.}, \prl{\bf 79}, 741 (1997).
456:
457: \bibitem{micro}U. Klein, J. Low Temp. Phys. {\bf 69}, 1 (1987);
458: M. Ichioka, A. Hasegawa, and K. Machida, \prb {\bf 59},
459: 8902 (1999).
460:
461: \bibitem{nl}V.G. Kogan {\it et al.}, \prb{\bf 54}, 12386 (1996);
462: {\bf 55}, R8693 (1997).
463:
464: \bibitem{book}V.G. Kogan, P. Miranovi\' c, and D. McK Paul, in
465: {\it The Superconducting State in Magnetic Fields,} Series on Directions in
466: Condensed Matter Physics, v. 13; Edited by C.A.R. Sa de Melo, World Scientific,
467: Singapore (1998).
468:
469: \bibitem{Franz}M. Franz, I. Affleck, and M.H.S. Amin, \prl {\bf 79}, 1555 (1997).
470:
471: \bibitem{imp}K.O. Cheon {\it et al.}, \prb{\bf 58}, 6463 (1998);
472: P.L. Gammel {\it et al.}, \prl{\bf 82}, 4082 (1999).
473: %; V.G. Kogan {\it et al.,} \prb{\bf 62}, 9077 (2000).
474:
475: \bibitem{gl1}I. Affleck, M. Franz, and M.H.S. Amin, \prb {\bf 55},
476: R704 (1997).
477:
478: \bibitem{gl2}B. Rosenstein and A. Knigavko, \prl{\bf 83}, 844 (1999).
479:
480: \bibitem{esk}M.R. Eskildsen {\it et al.} \prl {\bf 86}, 5148 (2001).
481:
482: \bibitem{***} S.B. Dugdale {\it et al.,} \prl {\bf 83}, 4824 (1999).
483:
484: \bibitem{ehb}E.H. Brandt, Rep. Prog. Phys. {\bf 58}, 1465 (1995).
485:
486: \bibitem{mir} A detailed discussion of the moduli is given by P.
487: Miranovi\'c and V.G. Kogan, cond-mat/0105630.
488:
489: \bibitem{kc}V.G. Kogan and L.J. Campbell, \prl {\bf 62}, 1552
490: (1989).
491:
492: \bibitem{rem1} Since div${\bf u}=0$, it is helpful to write $\bf u$ in
493: terms of a function $A(x,y)$ such that
494: $u_x=\partial_yA, u_y=-\partial_xA$.
495:
496: \bibitem{blat}G. Blatter {it et al.,} Rev. Mod. Phys. {\bf 66},
497: 1125 (1994).
498:
499: \bibitem{app}The modulus $c_{\omega}=\partial^2_{\psi\psi}F$ is
500: determined by %23
501: the energy change due to rotation of the VL over the angle
502: $\psi\to 0$. Substituting $G_x\to G_x\cos\psi - G_y\sin\psi$,
503: $G_y\to G_x\sin\psi + G_y\cos\psi$ into Eq. (\ref{fl}), we obtain in
504: the first order in $\rho^2$:
505: $c_{\omega}=(B^2\rho^2/4\pi\lambda^2)\sum_{mn}'(1-8m^2n^2/g^2)\exp(-pg)$,
506: where $g=m^2+n^2$, and $p =\pi C b$. For $p\ll 1$, the double sum
507: is $2\sqrt{\pi /p}$,
508: thus $c_{\omega}=B^2\rho^2/2\lambda^2\sqrt{p\pi}$.
509:
510: \bibitem{core}E.H. Brandt, J. Low Temp. Phys. {\bf 73}, 355 (1988);
511: A. Yaouanc, P.D. de Reotier, and E.H. Brandt, \prb {\bf 55}, 11107 (1997).
512:
513: \bibitem{peter}R.N. Kleiman {\it et al.,} \prl{\bf 69}, 3120 (1992);
514: P.L. Gammel, {\it et al.,} {\it ibid.} {\bf 72}, 278 (1994).
515:
516: \bibitem{scha} P.M. Platzman and H. Fukuyama, \prb{\bf 10}, 3150
517: (1974); D.C. Fisher, {\it ibid.} {\bf 26}, 5009 (1982).
518:
519: \bibitem{es}M.R. Eskildsen {\it et al.}, Nature {\bf 393}, 242 (1998).
520:
521: \bibitem{Mona}M. Yethiraj and D.K.Christen (unpublished).
522:
523: %\bibitem{song} K.J. Song {\it et al.,} \prb {\bf 59}, R6620 (1999).
524:
525: \end{references}
526:
527: \begin{figure}
528: \epsfxsize= 0.7\hsize
529: \centerline{
530: \vbox{
531: \epsffile{fig1.eps}
532: }}
533: \vskip \baselineskip
534: \caption{The transition line $H_{\Box}(T)$ (circles) in ${\rm LuNi_2B_2C}$
535: observed by SANS [17]. The inset shows
536: $H_{\Box}(T)$ (dashed line) predicted by the extended GL theory [6]
537: without vortex fluctuations. The solid lines shows $H_{c2}(T)$.
538: }
539: \label{fig1}
540: \end{figure}
541:
542:
543: \begin{figure}
544: \epsfxsize= 0.7\hsize
545: \centerline{
546: \vbox{
547: \epsffile{fig2.eps}
548: }}
549: \vskip \baselineskip
550: \caption{The transition lines $H_{\Box}(T)$ obtained by numerically
551: solving Eq. (\ref{cs}) for $\chi_0=0.0064$, $C=1$,
552: and a few ratios of $\rho/\xi_0$. The dashed line is $H_{c2}(T)$.
553: }
554: \label{fig2}
555: \end{figure}
556:
557:
558: \end{multicols}
559: \end{document}
560: