1: \documentclass[aps,twocolumn,showpacs,amsmath,amssymb]{revtex4}
2: \usepackage{graphicx}
3: \usepackage{bm}
4: %\usepackage{times}
5: %\usepackage{dcolumn}
6: %\usepackage{natbib}
7:
8: \def\br{ \bm{r} }
9: \def\bR{ \bm{R} }
10: \def\bk{ \bm{k} }
11: \def\bq{ \bm{q} }
12: \def\brho{ \bm{\rho} }
13: \def\im{ \,\mathrm{Im}\, }
14: \def\re{ \,\mathrm{Re}\, }
15:
16: \begin{document}
17:
18: \title{CePt$_3$Si: an unconventional superconductor without inversion center}
19:
20: \author{K. V. Samokhin, E. S. Zijlstra, S. K. Bose}
21:
22: \address{Department of Physics, Brock University, St.Catharines, Ontario, Canada L2S 3A1}
23: \date{\today}
24:
25:
26: \begin{abstract}
27: Most superconducting materials have an inversion center in their
28: crystal lattices. One of few exceptions is the recently discovered
29: heavy-fermion superconductor CePt$_3$Si [E. Bauer {\em et al},
30: Phys. Rev. Lett. \textbf{92}, 027003 (2004)]. In this paper, we
31: analyze the implications of the lack of inversion symmetry for the
32: superconducting pairing. We show that the order parameter is an
33: odd function of momentum, and that there always are lines of zeros
34: in the excitation energy gap for one-component order parameters,
35: which seems to agree with the experimental data. The
36: superconducting phase can be non-uniform, even without external
37: magnetic field, due to the presence of unusual gradient terms in
38: the Ginzburg-Landau free energy. Also, we performed {\em ab
39: initio} electronic structure calculations for CePt$_3$Si, which
40: showed that the spin-orbit coupling in this material is strong,
41: and the degeneracy of the bands is lifted everywhere except along
42: some high symmetry lines in the Brillouin zone.
43: \end{abstract}
44:
45: \pacs{74.20.Rp, 74.70.Tx, 71.20.Be}
46:
47: \maketitle
48:
49: \section{Introduction}
50: \label{sec: Intro}
51:
52: In the past two decades, a number of novel superconducting
53: materials have been discovered where order parameter symmetries
54: are different from an $s$-wave spin singlet, predicted by the
55: Bardeen-Cooper-Schrieffer (BCS) theory of electron-phonon mediated
56: pairing. From the initial discoveries of unconventional
57: superconductivity in heavy-fermion compounds, the list of examples
58: has now grown to include the high-$T_c$ cuprate superconductors,
59: ruthenates, ferromagnetic superconductors, and possibly organic
60: materials. In most of these materials, there are strong
61: indications that the pairing is caused by the electron
62: correlations, in contrast to conventional superconductors such as
63: Pb, Nb, {\em etc}. Non-phononic mechanisms of pairing are believed
64: to favor a non-trivial spin structure and orbital symmetry of the
65: Cooper pairs. For example, the order parameter in the high-$T_c$
66: superconductors, where the pairing is thought to be caused by the
67: anti-ferromagnetic correlations, has the $d$-wave symmetry with
68: lines of zeroes at the Fermi surface.
69:
70: A powerful tool of studying unconventional superconducting states
71: is symmetry analysis, which works even if the pairing mechanism is
72: not known. Mathematically, superconductivity is a consequence of
73: the breaking of the gauge symmetry $U(1)$ in the full symmetry
74: group of the normal state: ${\cal G}=S_M\times U(1)$, where $S_M$
75: is a magnetic space group, which includes usual space group
76: operations (i.e. translations, rotations, inversion, {\em etc})
77: and time reversal operation $K$. In a non-magnetic crystal,
78: $S_M=S\times{\cal K}$, where $S$ is the space group, and ${\cal
79: K}=\{E,K\}$. In a magnetic crystal, $K$ enters $S_M$ only in
80: combination with other symmetry elements. The superconducting
81: state is said to be ``unconventional'' if, in addition to the
82: gauge symmetry, some other symmetries from ${\cal G}$ are broken.
83: The group-theoretical analysis of unconventional superconducting
84: states in non-magnetic crystals was developed in Refs.
85: \cite{VG85,UR85}. Recently, it was extended to include the
86: ferromagnetic case \cite{FMSC,FominFMSC,MineevFMSC}.
87:
88: In almost all previous studies, it was assumed that the crystal
89: has an inversion center, which leads to degenerate electron bands
90: and makes it possible to classify the Cooper pair states according
91: to their spin (or pseudospin if the spin-orbit coupling is taken
92: into account). The even (singlet) and odd (triplet) components of
93: the superconducting order parameter can then be studied separately
94: \cite{Book}. Although this is the case in most superconductors,
95: there are some exceptions. Early discussion of the possible loss
96: of inversion symmetry associated with a structural phase
97: transition in V$_3$Si, which is an A15 type superconductor, can be
98: found in Ref. \cite{AB65}. Later, a C15 superconductor HfV$_2$ was
99: found to undergo a transition from a cubic to a
100: non-centrosymmetric body-centered orthorhombic structure
101: \cite{LZ72}. The possible existence of superconductivity was
102: reported in ferroelectric perovskite compounds SrTiO$_3$
103: \cite{SHC64} and BaPbO$_3$-BaBiO$_3$ \cite{BV83}. More recently,
104: there has been a renewed theoretical interest in this problem,
105: mainly in the context of superconductivity in two-dimensional (2D)
106: systems, such as Cu-O layers in YBCO \cite{Edelstein}, or surface
107: superconductivity on Tamm levels \cite{GR01,BG02}. According to
108: Ref. \cite{Edelstein,GR01}, in the absence of inversion symmetry
109: the order parameter becomes a mixture of spin-singlet and
110: spin-triplet components, which leads, for instance, to the Knight
111: shift attaining a non-zero value at $T=0$. Other peculiar
112: properties of non-centrosymmetric superconductors include a jump
113: of magnetic induction at the surface \cite{LNE85}, and the
114: possibility of non-uniform helical superconducting phases due to
115: the presence of first-order gradient terms in the Ginzburg-Landau
116: (GL) functional \cite{MS94}.
117:
118: This work is motivated by a recent experimental discovery of
119: superconductivity with $T_c\simeq 0.75K$ in CePt$_3$Si
120: \cite{exp-CePtSi}, which is the first known heavy-fermion material
121: without inversion center. (It should be mentioned that
122: incommensurate density modulations might break inversion symmetry
123: in a heavy-fermion compound UPt$_3$, as pointed out in Ref.
124: \cite{Min93}.) Our goal is two-fold. First, we study the symmetry
125: of electron bands, calculate the electronic structure, and
126: estimate the magnitude of the spin-orbit coupling in the normal
127: state of CePt$_3$Si. Second, we use general symmetry arguments to
128: analyze the gap symmetry and spatial dependence of the order
129: parameter in a three-dimensional non-centrosymmetric tetragonal
130: superconductor, assuming a strong spin-orbit coupling and the
131: clean limit. The paper is organized as follows. In Sec. \ref{sec:
132: Symmetry of bands}, we study symmetry of the single electron
133: states. In Sec. \ref{sec: Bands}, the results of the electronic
134: band structure calculations are reported. Sec. \ref{sec: OP}
135: focuses on the symmetry of the superconducting order parameter,
136: possible locations of the gap nodes, and also spatial structure of
137: the superconducting phase, using the phenomenological
138: Ginzburg-Landau approach. Sec. \ref{sec: Concl} concludes with a
139: discussion of our results.
140:
141:
142:
143: \section{Symmetry of electron bands}
144: \label{sec: Symmetry of bands}
145:
146:
147: The symmetry group of the normal paramagnetic state: ${\cal
148: G}=S\times{\cal K}\times U(1)$, where $S$ is the space group of
149: the crystal, and $U(1)$ is the gauge group. In the case of
150: CePt$_3$Si, the space group is P4mm (No. 99), which is generated
151: by (i) lattice translations by the primitive vectors
152: $\bm{a}=a(1,0,0)$, $\bm{b}=a(0,1,0)$, $\bm{c}=c(0,0,1)$ of a
153: tetragonal lattice, and (ii) the generators of the point group
154: $G=\mathbf{C}_{4v}$: the rotations $C_{4z}$ about the $z$ axis by
155: an angle $\pi/2$ and the reflections $\sigma_x$ in the vertical
156: plane $(100)$. Spatial inversion $I$ is not an element of the
157: symmetry group.
158:
159: In the absence of inversion symmetry, spin-orbit (SO) coupling
160: plays a crucial role due to its band-splitting effect. At non-zero
161: SO coupling, the single-particle wavefunctions are linear
162: combinations of the eigenstates of the spin operator $s_z$:
163: $\langle\br|\psi\rangle=u(\br)\chi_\uparrow+
164: v(\br)\chi_\downarrow$. Since the normal state Hamiltonian $H_0$
165: is invariant with respect to the crystal lattice translations, the
166: eigenfunctions are the Bloch spinors $\psi_{\bk,n}(\br)$ belonging
167: to the wave vectors $\bk$ in the Brillouin zone, which can be
168: written in the form
169: \begin{equation}
170: \label{psi_pm}
171: \langle\br|\bk,n\rangle=u_{\bk,n}(\br)\chi_\uparrow+
172: v_{\bk,n}(\br)\chi_\downarrow.
173: \end{equation}
174: The corresponding eigenvalues $\epsilon_{n}(\bk)$ describe the
175: band dispersion of free electrons.
176:
177: In the presence of both the time-reversal and inversion symmetries
178: the bands are two-fold degenerate at each $\bk$. Indeed, the
179: states $|\bk,n\rangle$ and $KI|\bk,n\rangle$ correspond to the
180: same $\bk$, have the same energy, and are orthogonal (in addition,
181: these two states are degenerate with another pair of orthogonal
182: states, $K|\bk,n\rangle$ and $I|\bk,n\rangle$, which correspond to
183: $-\bk$). In this case, $n=(\nu,\pm)$, where $\pm$ labels the
184: linearly independent Bloch states at a given band index $\nu$.
185: There is a freedom in choosing the basis functions
186: $|\bk,\nu,+\rangle$ and $|\bk,\nu,-\rangle$. The most frequently
187: used convention is that they should transform under rotations $R$
188: similar to the spin eigenstates $|\bk,\nu,\uparrow\rangle$ and
189: $|\bk,\nu,\downarrow\rangle$, i.e.
190: \begin{equation}
191: \label{pseudospin def}
192: R|\bk,\nu,\alpha\rangle=
193: U_{\alpha\beta}(R)|R\bk,\nu,\beta\rangle,
194: \end{equation}
195: where $\alpha,\beta=\pm$ and $U(R)$ is the spin-rotation matrix:
196: for a rotation by an angle $\theta$ around some axis $\bm{n}$:
197: $U(R)=\exp[-i(\theta/2)(\bm{\sigma}\cdot\bm{n})]$ ($\bm{\sigma}$
198: are Pauli matrices). The practical recipe for constructing the
199: basis of the Bloch states is as follows: first choose a state
200: $|\bk,\nu,+\rangle$ at each $\bk$ in the irreducible part of the
201: first Brillouin zone, then act on it by $KI$ to obtain an
202: orthogonal state $|\bk,\nu,-\rangle$ at the same $\bk$, and
203: finally act on $|\bk,\nu,\pm\rangle$ by the elements of the point
204: group and use the prescription (\ref{pseudospin def}) to obtain
205: the pairs of the basis functions belonging to the star of $\bk$.
206: The single-electron states constructed in this way are referred to
207: as the pseudospin states \cite{UR85}. It is the presence of the
208: inversion center that makes the bands double degenerate in a
209: non-magnetic crystal:
210: $\epsilon_{\nu,+}(\bk)=\epsilon_{\nu,+}(-\bk)=\epsilon_{\nu,-}(\bk)$
211: at all $\bk$.
212:
213: In contrast, if the crystal lacks the inversion symmetry, then the
214: degeneracy of the single-electron bands $\epsilon_n(\bk)$ is
215: lifted everywhere, except from some points or lines of high
216: symmetry. In particular, the bands always touch at
217: $\bk=\mathbf{0}$ (the $\Gamma$ point), because of the Kramers
218: theorem: time-reversal symmetry means that $|\bk,n\rangle$ and
219: $K|\bk,n\rangle\sim|-\bk,n\rangle$ have the same energy. In the
220: case of CePt$_3$Si, one can show that the spin-split bands touch
221: along the $\Gamma Z$ line (i.e. along the [001] direction), and
222: there are no other symmetry-imposed degeneracies in the bands
223: crossing the Fermi level (see Sec. \ref{sec: Bands} below). In the
224: limit of zero SO coupling (which is not applicable to CePt$_3$Si),
225: but still without an inversion center, the symmetry of the system
226: contains $SU(2)$ spin rotations, in addition to the space group,
227: time reversal, and the gauge group. Then, the bands at each $\bk$
228: are double degenerate.
229:
230: Let us now show that the transformation properties of the
231: single-electron wave functions in the absence of inversion center
232: are different from those of the spin or pseudospin eigenstates,
233: see Eq. (\ref{pseudospin def}). Mathematically, the wave functions
234: in each band transform according to irreducible co-representations
235: of the Type II magnetic space group $S_M=S\times{\cal K}$
236: \cite{co-reps} (one should use co-representations because the
237: time-reversal operation $K$ is anti-unitary). In addition, the
238: co-representations must be double-valued because the states
239: $|\bk\rangle$ are spin-$1/2$ spinors, so that any rotation by
240: $2\pi$ changes their sign: $C_{4z}^4|\bk\rangle=-|\bk\rangle$,
241: $\sigma_x^2|\bk\rangle=-|\bk\rangle$, and also
242: $K^2|\bk\rangle=-|\bk\rangle$ (the band index $n$ is omitted). The
243: double-valuedness can be dealt with in the standard fashion by
244: using the ``double-group'' trick \cite{LL-3}: one introduces a
245: fictitious new symmetry element $\bar E$, which commutes with all
246: other elements and satisfies the conditions
247: $C_{4z}^4=\sigma_x^2=\bar E$, $\bar E^2=E$, and also $K^2=\bar E$.
248:
249: The co-representations of $S_M$ can be derived from the usual
250: representations of the unitary component, which in our case
251: coincides with the space group $S$. For each $\bk$ in the
252: Brillouin zone, the basis of an irreducible representation of $S$
253: is formed by the Bloch states corresponding to the star of $\bk$.
254: The state $K|\bk\rangle$ belongs to the wave vector $-\bk$, but
255: the irreducible representations of $S$ corresponding to $\bk$ and
256: $-\bk$ are inequivalent (because of the absence of inversion
257: symmetry) and must therefore be regarded as belonging to a single
258: ``physically irreducible'' representation of twice the dimension
259: \cite{LL-3}. In terms of co-representations this means that the
260: Bloch states $|\bk\rangle$ and $K|\bk\rangle$ belong to the same
261: two-dimensional irreducible co-representation of $S\times{\cal K}$
262: \cite{co-reps}. Thus, the appropriate basis of the Bloch states is
263: formed by the wave functions $|G\bk\rangle$ corresponding to the
264: star of $\bk$, and also by their time-reversed counterparts
265: $K|G\bk\rangle$, which can be combined in a set of bispinors
266: $|\Psi_{\bk}\rangle=(|\bk\rangle,K|\bk\rangle)^T$. All the states
267: from this set have the same energy:
268: $\epsilon(\bk)=\epsilon(G\bk)=\epsilon(-\bk)$.
269:
270: Since the function $G|\bk\rangle$ belongs to the wave vector
271: $G\bk$, one can write
272: $G|\bk\rangle=e^{i\varphi_{\bk}(G)}|G\bk\rangle$. The undetermined
273: phase factors come from the freedom in choosing the phases of the
274: Bloch states and realize a representation of the point group in
275: this basis (it is assumed that the Bloch states are single-valued
276: and continuous functions throughout the Brillouin zone). Using the
277: commutation of $G$ and $K$ we have $GK|\bk\rangle=KG|\bk\rangle=
278: K\exp[i\varphi_{\bk}(G)]|G\bk\rangle=
279: \exp[-i\varphi_{\bk}(G)]K|G\bk\rangle$. The co-representation
280: matrices in the basis of bispinor wave functions can be obtained
281: from the relations
282: \begin{eqnarray}
283: \label{trans G}
284: && G|\Psi_{\bk}\rangle=\left(\begin{array}{cc}
285: e^{i\varphi_{\bk}(G)} & 0 \\
286: 0 & e^{-i\varphi_{\bk}(G)} \\
287: \end{array}\right)|\Psi_{G\bk}\rangle\\
288: \label{trans bE}
289: && \bar E|\Psi_{\bk}\rangle=\left(\begin{array}{cc}
290: -1 & 0 \\
291: 0 & -1 \\
292: \end{array}\right)|\Psi_{\bk}\rangle\\
293: \label{trans K}
294: && K|\Psi_{\bk}\rangle=
295: \left(\begin{array}{cc}
296: 0 & 1 \\
297: -1 & 0 \\
298: \end{array}\right)|\Psi_{\bk}\rangle
299: \end{eqnarray}
300: [we have taken into account that $K(K|\bk\rangle)=-|\bk\rangle$].
301: Note that the multiplication rules for the co-representation
302: matrices are different from usual unitary group representations.
303: For example, $D(G_1G_2)=D(G_1)D(G_2)$ and $D(GK)=D(G)D(K)$, but
304: $D(KG)=D(K)D^*(G)$ and $D(K^2)=D(K)D^*(K)$ \cite{co-reps}.
305:
306: The general symmetry arguments given above can be illustrated
307: using an exactly-solvable three-dimensional generalization of the
308: Rashba model, which was originally proposed to describe the
309: effects of symmetry lowering near the surface of a semiconductor
310: \cite{Rashba60} and recently applied in Refs.
311: \cite{Edelstein,GR01} to quasi-2D non-centrosymmetric
312: superconductors. Consider a single band $\epsilon_0(\bk)$ in a
313: crystal described by the point group $\mathbf{C}_{4v}$. At zero SO
314: coupling, the band is two-fold degenerate due to spin. The absence
315: of reflection symmetry in the $xy$ plane implies the existence of
316: an internal electric field in the crystal, whose average over a
317: unit cell is non-zero. In the Rashba approximation, this
318: non-uniform field is replaced by its average, introducing a
319: constant vector $\bm{n}\parallel\hat z$. When a non-zero SO
320: coupling is switched on, it can described by an additional term in
321: the Hamiltonian:
322: \begin{equation}
323: \label{Rashba}
324: H_{so}=\alpha\sum\limits_{\bk}\bm{n}\cdot(\bm{\sigma}_{\sigma\sigma'}
325: \times\bk)\;a^\dagger_{\bk,\sigma}a_{\bk,\sigma'},
326: \end{equation}
327: where $\sigma,\sigma'=\uparrow,\downarrow$ is the $z$-axis spin
328: projection, and the states $|\bk,\sigma\rangle$ are the Bloch
329: spinors at zero SO coupling. Diagonalization of the full
330: single-electron Hamiltonian, $H=H_0+H_{so}$, gives two bands
331: \begin{equation}
332: \label{Rashba bands}
333: \epsilon_1(\bk)=\epsilon_0(\bk)+\alpha|\bk_\perp|,\quad
334: \epsilon_2(\bk)=\epsilon_0(\bk)-\alpha|\bk_\perp|
335: \end{equation}
336: ($|\bk_\perp|=\sqrt{k_x^2+k_y^2}$), which satisfy the condition
337: $\epsilon_{1,2}(\bk)=\epsilon_{1,2}(-\bk)$ and additional
338: symmetries from the point group. Also, the bands touch along the
339: line $\bk\parallel\hat z$. It is easy to see that one cannot use
340: pseudospin to label these bands, because the time reversal $K$
341: transforms the bands into themselves:
342: $|\bk,n\rangle\to|-\bk,n\rangle$ ($n=1,2$), while the pseudospin
343: states would be transformed into one another.
344:
345: To conclude this section, it should be mentioned that the
346: superconductivity in CePt$_3$Si seems to occur in the presence of
347: anti-ferromagnetic (AFM) order \cite{exp-CePtSi}, although no data
348: have been reported on the structure of the magnetic phase or the
349: magnitude of the staggered moment. Although the symmetry analysis
350: above was done assuming a paramagnetic normal state, our results
351: can be easily generalized for an AFM case. For a staggered
352: magnetization directed along the $z$ axis, the only change one has
353: to make in the symmetry group is to replace $K$ with
354: $KT_{\bm{a}}$, which combines the time reversal operation with a
355: lattice translation. Then, Eq. (\ref{trans K}) is replaced by
356: \begin{equation}
357: \label{trans KT}
358: KT_{\bm{a}}|\Psi_{\bk}\rangle=
359: \left(\begin{array}{cc}
360: 0 & e^{i\bk\bm{a}} \\
361: -e^{-i\bk\bm{a}} & 0 \\
362: \end{array}\right)|\Psi_{\bk}\rangle,
363: \end{equation}
364: so that $(KT_{\bm{a}})^2|\bk\rangle=-e^{-2i\bk\bm{a}}|\bk\rangle$.
365:
366: From the expressions (\ref{trans G})--(\ref{trans KT}), one
367: obtains the transformation rules for the creation operators of
368: electrons in the Bloch states $|\bk\rangle$. Although there is
369: some freedom in choosing the phase factors, we will see in Sec.
370: \ref{sec: OP} that the physically relevant properties are
371: insensitive to the choice of $\varphi_{\bk}(G)$.
372:
373:
374:
375: \section{Electronic structure}
376: \label{sec: Bands}
377:
378: CePt$_3$Si crystallizes in the same tetragonal structure as
379: CePt$_3$B, with space group P4mm (No. 99). The lattice parameters
380: are $a = 4.072$ {\AA} and $c = 5.442$ {\AA}. Ce is at the 1(b)
381: site ($1/2$, $1/2$, $0.1468$), Pt at the 2(c) site ($1/2$, $0$,
382: $0.6504$) and at the 1(a) site ($0$, $0$, $0$). The $z$ coordinate
383: of the latter site was chosen to be zero. Si is at the 1(a) site
384: ($0$, $0$, $0.4118$). These structural parameters all derive from
385: single-crystal x-ray data \cite{exp-CePtSi}, and can be assumed to
386: be sufficiently accurate for our purposes.
387:
388: We calculated the electronic structure of CePt$_3$Si with the
389: full-potential (FP) linear augmented-plane wave (LAPW) method
390: \cite{Bla90,Mad01,Sch02}, which is based on density functional
391: theory \cite{Koh65}. For the exchange and correlation potential we
392: used the local density approximation \cite{pw92,footnote_gga}
393: (LDA). We performed a non-magnetic calculation, neglecting the AFM
394: order observed experimentally below $2.2$ K \cite{exp-CePtSi}. The
395: muffin-tin radii of the atoms were chosen as $2.11$ $a_0$. A
396: typical number of plane waves in our basis set was 580. The
397: electronic ground state was calculated self-consistently on a grid
398: of 4212 $\mathbf{k}$ points in the entire Brillouin zone.
399:
400: For the electronic structure of alloys containing heavy elements,
401: such as Ce and Pt, the SO coupling can in general not be
402: neglected. In particular, as shown in Sec. \ref{sec: OP}, for the
403: analysis of the symmetry properties of the superconducting order
404: parameter it is important to obtain an estimate of the SO
405: splitting of the bands near the Fermi energy. The SO coupling has
406: been included in our calculations using the ``second variational
407: treatment'', as discussed by MacDonald {\it et al} \cite{SO}. In
408: this approach, first the eigenstates are calculated in the absence
409: of the SO interaction. Then, the SO interaction is included in a
410: perturbative way, where the eigenstates up to a certain cut-off
411: energy, calculated without the SO interaction, are used as the
412: basis states. In our calculations this cut-off energy was
413: approximately 22 eV above the Fermi energy.
414:
415: Although the CePt$_3$Si structure doesn't have inversion symmetry,
416: the band energies still satisfy the relation
417: $\epsilon_n(\mathbf{k}) = \epsilon_n(-\mathbf{k})$, due to the
418: time reversal symmetry of the single-electron Hamiltonian (see
419: Sec. \ref{sec: Symmetry of bands}). Figure \ref{fig_bands1} shows
420: the band structure of CePt$_3$Si without SO coupling calculated
421: along some high symmetry lines. The bands were plotted according
422: to increasing energy. A complete analysis of band crossings based
423: on the character of eigenfunctions was not carried out. However,
424: the bands labeled $\beta$ and $\gamma$ do cross between the
425: symmetry points $X$ and $M$, and so do the $\beta$ and $\gamma$
426: bands between $R$ and $A$. The $\gamma$ band and the first dotted
427: band above the Fermi energy have the same symmetry between $M$ and
428: $\Gamma$, and hence are unlikely to cross there. The labels of the
429: bands were chosen according to the band index, not the band
430: character, simply to relate various parts of the Fermi surface to
431: the bands crossing the Fermi level. In the electronic density of
432: states (not shown) there is a peak at 0.4 eV above the Fermi
433: energy, which is due to the unfilled Ce-4f electrons. We found
434: that the electrons at the Fermi energy are predominantly of Ce-4f
435: character.
436:
437: Figure \ref{fig_bands} shows the band structure of CePt$_3$Si with
438: SO coupling \cite{SO}. As in Fig. \ref{fig_bands1} we connected
439: bands with the same band index, ignoring band crossings. In Fig.
440: \ref{fig_bands}, the bands near the Fermi energy $\epsilon_F$ are
441: split by an amount of at most $50-200$ meV. This splitting
442: vanishes along the lines $\Gamma - Z$ and $M - A$.
443:
444: Figure \ref{fig_fs} shows the cross sections of the Fermi surface
445: of CePt$_3$Si in the presence of SO coupling. The splitting of the
446: bands at the Fermi energy due to the SO coupling is prominent in
447: this figure. The sheets of the Fermi surface labeled $\alpha$ and
448: $\beta$ are hole-like. The $\gamma$ sheets are electron-like. The
449: $\alpha$ sheet consists of a feature around the $Z$ point. The
450: $\beta$ band gives rise to a larger feature around the $Z$ point
451: and, in addition, to a feature around the $X$ point [and the
452: symmetry related point $Y = (0,0.5,0)$]. The $\gamma$ bands give
453: rise to a feature around the line $M - A$, which is almost
454: dispersionless in $k_z$, as well as to a small feature around the
455: $\Gamma$ point. We further noted that the $\alpha$ bands
456: contribute 1.9\% and 3.5\% respectively to the density of states
457: at the Fermi energy. Similar contributions coming from the SO
458: split $\beta$ and $\gamma$ bands are 25\% and 45\% and 15\% and
459: 9.0\%, respectively.
460:
461:
462:
463: \section{Superconducting order parameter}
464: \label{sec: OP}
465:
466: \subsection{Symmetry analysis}
467:
468: The single-electron states $|\bk,n\rangle$ can be used as a basis
469: for constructing the Hamiltonian which takes into account the
470: Cooper pairing between electrons. We have $H=H_0+H_{sc}$, where
471: \begin{equation}
472: \label{H_0}
473: H_0=\sum\limits_n\sum\limits_{\bk}
474: \epsilon_n(\bk)c^\dagger_{\bk,n}c_{\bk,n},
475: \end{equation}
476: is the free-electron part, with the chemical potential absorbed
477: into the band energies. As follows from the results of Secs.
478: \ref{sec: Symmetry of bands} and \ref{sec: Bands}, the electronic
479: bands $\epsilon_n(\bk)$ are non-degenerate, except along the line
480: $\bk\parallel\hat z$, and invariant under all operations from the
481: point group $\mathbf{C}_{4v}$ and also under inversion:
482: $\epsilon_n(\bk)=\epsilon_n(-\bk)$. The Fermi surface of
483: CePt$_3$Si consists of several sheets (see Sec. \ref{sec: Bands}),
484: all of which can in principle participate in the formation of the
485: superconducting order.
486:
487: Assuming a BCS-type mechanism of pairing, the interaction between
488: the band electrons in the Cooper channel can be written in the
489: following form:
490: \begin{equation}
491: \label{H BCS}
492: H_{sc}=H^{(1)}_{sc}+H^{(2)}_{sc}+H^{(3)}_{sc},
493: \end{equation}
494: where
495: \begin{eqnarray}
496: \label{H1}
497: H^{(1)}_{sc}=\frac{1}{2}\sum\limits_{n}\sum\limits_{\bk,\bk'}
498: V^{(1)}_n(\bk,\bk')c^\dagger_{\bk,n}c^\dagger_{-\bk,n}
499: c_{-\bk',n}c_{\bk',n}
500: \end{eqnarray}
501: \begin{eqnarray}
502: \label{H2}
503: H^{(2)}_{sc}=\frac{1}{2}\sum\limits_{n\neq m}\sum\limits_{\bk,\bk'}
504: V^{(2)}_{nm}(\bk,\bk')c^\dagger_{\bk,n}c^\dagger_{-\bk,n}
505: c_{-\bk',m}c_{\bk',m}
506: \end{eqnarray}
507: \begin{eqnarray}
508: \label{H3}
509: H^{(3)}_{sc}=\frac{1}{2}\sum\limits_{n\neq m}\sum\limits_{\bk,\bk'}
510: V^{(3)}_{nm}(\bk,\bk')c^\dagger_{\bk,n}c^\dagger_{-\bk,m}
511: c_{-\bk',m}c_{\bk',n}.
512: \end{eqnarray}
513: The potentials $V$ are non-zero only inside the energy shells of
514: width $\omega_c$ (the cutoff energy) near the Fermi surface.
515:
516: Treating the Cooper interaction between the electrons with
517: opposite momenta in the mean-field approximation, we obtain
518: \begin{equation}
519: \label{H_sc}
520: H_{sc}=\frac{1}{2}\sum\limits_{\bk}\sum\limits_{nm}
521: \Bigl[\Delta_{nm}(\bk)c^\dagger_{\bk,n}
522: c^\dagger_{-\bk,m}+\mathrm{h.c.}\Bigr].
523: \end{equation}
524: Here the diagonal matrix elements $\Delta_{nn}(\bk)$ represent the
525: Cooper pairs composed of quasiparticles from the same sheet of the
526: Fermi surface, and the off-diagonal matrix elements
527: $\Delta_{nm}(\bk)$ with $n\neq m$ represent the pairs of
528: quasiparticles from different sheets. From the anti-commutation of
529: the fermionic operators, it follows that $\Delta_{nn}(\bk)$ are
530: odd functions of $\bk$, but $\Delta_{nm}(\bk)=-\Delta_{mn}(-\bk)$
531: with $n\neq m$ do not have a definite parity.
532:
533: A considerable simplification occurs if to assume that the
534: superconducting gaps are much smaller than the interband energies.
535: The band structure calculations of Sec. \ref{sec: Bands} show that
536: typically the SO band splitting $E_{so}$ is of the order of
537: $50-200\;\mathrm{meV}$ (between the bands derived from the
538: degenerate spin-up and spin-down bands at zero SO coupling), which
539: exceeds the superconducting gap by orders of magnitude. In this
540: situation, the formation of interband pairs is energetically
541: unfavorable, and the amplitudes $\Delta_{nm}$ with $n\neq m$
542: vanish. The origin of the suppression of these types of pairing is
543: similar to the well-known paramagnetic limit of singlet
544: superconductivity \cite{CC62}: the interband splitting $E_{so}$
545: cuts off the logarithmic singularity in the Cooper channel, thus
546: reducing the critical temperature. Although the condition
547: $E_{so}\gg T_c$ is violated very close to the poles of the Fermi
548: surface where the spin-split bands touch, the off-diagonal Cooper
549: pairing in the vicinity of these points is still suppressed due to
550: the phase space limitations. We also neglect the possibility of
551: the Cooper pairs having a non-zero momentum, i.e. $\langle
552: c^\dagger_{\bk+\bq,n}c^\dagger_{-\bk,m}\rangle\neq 0$
553: [Larkin-Ovchinnikov-Fulde-Ferrell (LOFF) phase] \cite{LOFF}.
554: Although the critical temperature of the resulting non-uniform
555: superconducting state can be higher than that of the uniform
556: state, this would not be sufficient to overcome a large depairing
557: effect of the SO band splitting.
558:
559: Thus, the interband pairing (\ref{H3}) can be neglected, and
560: $\Delta_{nm}(\bk)=\Delta_n(\bk)\delta_{nm}$ \cite{And84}. This is
561: reminiscent of the situation in ferromagnetic superconductors,
562: where only the same-spin components of $\Delta$ survive the large
563: exchange band splitting \cite{FMSC}. It follows from Eqs.
564: (\ref{trans G})--(\ref{trans K}) and anti-linearity of $K$ that
565: \begin{eqnarray*}
566: &&G(c^\dagger_{\bk,n}c^\dagger_{-\bk,n})G^{-1}=
567: c^\dagger_{G\bk,n}c^\dagger_{-G\bk,n}\\
568: &&\bar E(c^\dagger_{\bk,n}c^\dagger_{-\bk,n})\bar
569: E^{-1}=c^\dagger_{\bk,n}c^\dagger_{-\bk,n}\\
570: &&K(\lambda c^\dagger_{\bk,n}c^\dagger_{-\bk,n})K^{-1}=
571: -\lambda^*c^\dagger_{-\bk,n}c^\dagger_{\bk,n}\\
572: &&=\lambda^*c^\dagger_{\bk,n}c^\dagger_{-\bk,n},
573: \end{eqnarray*}
574: where $\lambda$ is an arbitrary constant. We have the following
575: transformation rules for the order parameter under the elements of
576: ${\cal G}$:
577: \begin{equation}
578: \label{Delta transforms} \left.\begin{array}{ll}
579: G:& \Delta_n(\bk) \to \Delta_n(G^{-1}\bk),\\
580: K:& \Delta_n(\bk) \to \Delta^*_n(\bk).
581: \end{array}\right.
582: \end{equation}
583: Thus, the order parameter components transform like scalar
584: functions. There is no need for double groups, since $\bar E$ is
585: equivalent to $E$ when acting on $\Delta$. In the case of an AFM
586: normal state, $K$ should be replaced with $KT_{\bm{a}}$.
587:
588: The superconducting order parameter on the $n$th sheet of the
589: Fermi surface transforms according to one of the irreducible
590: representations $\Gamma$ of the normal state point group
591: $\mathbf{C}_{4v}$. It can be represented in the form
592: \begin{equation}
593: \label{Delta expansion}
594: \Delta_{n,\Gamma}(\bk)=\sum\limits_{i=1}^{d_{\Gamma}}
595: \eta_{n,i}\phi_{\Gamma,n,i}(\bk),
596: \end{equation}
597: where $d_{\Gamma}$ is the dimension of $\Gamma$,
598: $\phi_{\Gamma,n,i}(\bk)$ are the basis functions (which are
599: different on different sheets of the Fermi surface in general),
600: and $\eta_{n,i}$ are the order parameter components that enter,
601: e.g., the GL free energy and can depend on coordinates
602: \cite{Book}. Despite the absence of inversion center in the
603: crystal, the order parameters $\Delta_n$ have a definite parity,
604: namely they are all odd with respect to $\bk\to-\bk$. The odd
605: irreducible representations of $\mathbf{C}_{4v}$ are listed in
606: Table \ref{table1}. Since $d_\Gamma=1$ or $2$, the order parameter
607: in each band can have one or two components.
608:
609: If we neglect the interband pairing described by $H^{(2)}_{sc}$,
610: see Eq. (\ref{H2}), then the order parameters $\Delta_n$ are
611: completely decoupled, in particular, they all have different
612: critical temperatures $T_{c,n}$. However, there is no reason to
613: expect these interband terms to be small. If they are taken into
614: account, then all $\Delta_n(\bk)$ are non-zero, so the
615: superconductivity will be induced simultaneously on all sheets of
616: the Fermi surface. The simplest way to see how this works is to
617: use the GL free energy functional, which contains all possible
618: uniform and gradient terms invariant with respect to ${\cal G}$.
619: For a one-dimensional representation $\Gamma$ (the generalization
620: for two-dimensional representations is straightforward), we obtain
621: the following expression for the uniform terms in the free energy
622: density:
623: \begin{equation}
624: \label{FGLuniform}
625: F_{uniform} = \sum\limits_{n,m}A_{nm}(T)\eta_n^*\eta_m+F_4,
626: \end{equation}
627: where $F_4$ stands for the terms of fourth order (and higher), and
628: $A_{ij}$ is a real symmetric matrix. The off-diagonal elements of
629: $A$ correspond to the interband pairing. The critical temperature
630: $T_c$ is defined as the maximum temperature at which $A$ ceases to
631: be positive definite. Below $T_c$ all $\eta_n$ are non-zero and
632: proportional to a single complex number $\eta$, such that
633: $\eta_n=\varepsilon_n\eta$, where $\varepsilon_n$ are constants
634: that can be found by minimizing $F_{uniform}$. Therefore, the
635: components of a one-dimensional order parameter corresponding to
636: the representation $\Gamma$ are given by
637: \begin{equation}
638: \label{1D}
639: \Delta_n(\bk)=\eta\varepsilon_n\phi_{\Gamma,n}(\bk).
640: \end{equation}
641: Although the basis functions $\phi$ are different in general, they
642: all have the same symmetry.
643:
644: Our phenomenological theory cannot determine which pairing channel
645: corresponds to the highest critical temperature. In a recent work,
646: Frigeri {\em et al} \cite {FAKS03} proposed a microscopic model
647: for superconductivity in a non-centrosymmetric crystal, treating
648: the SO coupling as a perturbation with a Rashba-type Hamiltonian.
649: Assuming a strong interband pairing interaction which induces
650: order parameters of the same amplitude on both sheets of the Fermi
651: surface [see Eq. (\ref{Rashba bands})], they predicted a certain
652: gap symmetry for CePt$_3$Si, which seems to correspond to the
653: two-dimensional $E$ representation in our classification.
654:
655:
656:
657: \begin{table}
658: \caption{\label{table1} The character table and the examples of
659: the odd basis functions for the irreducible representations of
660: $\mathbf{C}_{4v}$.}
661: \begin{ruledtabular}
662: \begin{tabular}{|c|c|c|c|r|}
663: $\Gamma$ & $E$ & $C_{4z}$ & $\sigma_x$ &
664: $\phi_{\Gamma}(\bk)$\\ \hline
665: $A_1$ & 1 & 1 & 1 & $k_z$ \\ \hline
666: $A_2$ & 1 & 1 & $-1$ & $(k_x^2-k_y^2)k_xk_yk_z$ \\ \hline
667: $B_1$ & 1 & $-1$ & 1 & $(k_x^2-k_y^2)k_z$ \\ \hline
668: $B_2$ & 1 & $-1$ & $-1$ & $k_xk_yk_z$ \\ \hline
669: $E$ & 2 & 0 & 0 & $k_x$\ ,\ $k_y$
670: \end{tabular}
671: \end{ruledtabular}
672: \end{table}
673:
674:
675: \subsection{Gap zeros}
676:
677: The symmetry considerations can help find the zeros in the energy
678: spectrum of Bogoliubov quasiparticles, which are responsible for
679: peculiarities in the low-temperature behavior of unconventional
680: superconductors \cite{Book}. The gap structure of the order
681: parameter belonging to the two-dimensional representation $E$ of
682: $\mathbf{C}_{4v}$ depends on the superconducting phase, i.e. on
683: the values of the components $\eta_1$ and $\eta_2$, which in turn
684: are determined by minimizing the free energy of the
685: superconductor. In contrast, the gap nodes for the one-dimensional
686: order parameters are required by symmetry. Although the momentum
687: dependence of the order parameter is different on different sheets
688: of the Fermi surface, see Eq. (\ref{1D}), the locations of
689: symmetry-imposed gap zeros are the same.
690:
691: One can easily show that the order parameter corresponding to
692: $A_1$ vanishes on the plane $k_z=0$, so that the energy gap has
693: lines of nodes at the equators of all sheets of the Fermi surface.
694: Indeed,
695: \begin{eqnarray}
696: \label{C2z}
697: C_{2z}\phi_{A_1}(\bk)=\phi_{A_1}(-k_x,-k_y,k_z)\nonumber\\
698: =\phi_{A_1}(\bk)=-\phi_{A_1}(-\bk),
699: \end{eqnarray}
700: therefore, $\phi_{A_1}(k_x,k_y,0)=0$. In a similar fashion, one
701: can prove the existence of lines of zeros at $k_z=0$ for all other
702: one-dimensional representations. Also, the basis functions of
703: $A_2$ and $B_2$ have lines of zeros at $k_x=0$ or $k_y=0$, while
704: the basis functions of $A_2$ and $B_1$ have lines of zeros at
705: $k_x=\pm k_y$. The examples of the basis functions that have only
706: zeros imposed by symmetry are given in Table \ref{table1}.
707:
708: From the results of Sec. \ref{sec: Bands} it follows that some of
709: the sheets of the Fermi surface cross the boundaries of the
710: Brillouin zone. As seen from Eq. (\ref{C2z}), the order parameters
711: corresponding to all one-dimensional representations vanish at
712: $k_z=\pm\pi/c$, i.e. at the top and bottom surfaces of the
713: Brillouin zone, because $(k_x,k_y,\pi/c)$ and $(k_x,k_y,-\pi/c)$
714: are equivalent points. In addition, for the same reason the basis
715: functions of $A_2$ and $B_2$ have lines of zeros at $k_x=\pm\pi/a$
716: or $k_y=\pm\pi/a$.
717:
718: The gap nodes for one-dimensional order parameters are present at
719: the same locations on all sheets of the Fermi surface and can
720: disappear only if the interband pairing interactions
721: $H^{(3)}_{sc}$, see Eq. (\ref{H3}), are taken into account.
722:
723:
724:
725:
726: \subsection{Helical superconducting states}
727:
728: In addition to the uniform terms (\ref{FGLuniform}), the GL
729: functional for the order parameter corresponding to a
730: one-dimensional representation of $\mathbf{C}_{4v}$ also contains
731: gradient terms
732: \begin{eqnarray}
733: \label{FGLgrad1}
734: F_{grad,1} &=&
735: \sum\limits_{n,m}\Bigl[K^\perp_{nm}(\bm{D}_\perp\eta_n)^*
736: (\bm{D}_\perp\eta_m)\nonumber\\
737: && +K^z_{nm}(D_z\eta_n)^*(D_z\eta_m)\Bigr],
738: \end{eqnarray}
739: where $\bm{D}=\bm{\nabla}+i(2\pi/\Phi_0)\bm{A}$, $\Phi_0=\pi\hbar
740: c/e$ is the flux quantum, $\bm{A}$ is the vector potential, and
741: $\bm{D}_\perp=(D_x,D_y)$. The coefficients $K^\perp_{nm}$ and
742: $K^z_{nm}$ are real symmetric matrices. The terms
743: (\ref{FGLgrad1}), which are of the second order in $\bm{D}$, are
744: present in any multi-band tetragonal superconductor with the order
745: parameter corresponding to a one-dimensional representation of the
746: symmetry group. However, in the absence of an inversion center,
747: one can have additional terms in the GL functional, which satisfy
748: all the necessary symmetry requirements but are of the first order
749: in gradients \cite{MS94}. In our case, they can be written in the
750: form
751: \begin{equation}
752: \label{FGLgrad2}
753: F_{grad,2}=\sum\limits_{n,m}L_{nm}(\eta_n^*D_z\eta_m-
754: \eta_m^*D_z\eta_n),
755: \end{equation}
756: where $L_{nm}$ is a real anti-symmetric matrix, which is non-zero
757: only if the interband pairing (\ref{H2}) is present. The terms
758: (\ref{FGLgrad2}) lead to the possibility that the superconducting
759: state which appears immediately below $T_c$ can be non-uniform,
760: even without external magnetic field.
761:
762: Consider for simplicity only two bands participating in
763: superconductivity, i.e. $n=1,2$. In this case, the matrix $L$ in
764: Eq. (\ref{FGLgrad2}) has only one non-zero element:
765: $L_{12}=-L_{21}=\lambda/2$. The critical temperature for the
766: superconducting state
767: \begin{equation}
768: \label{helical}
769: \eta_1(\br)=\eta_{1,0}e^{iqz},\quad \eta_2(\br)=\eta_{2,0}e^{iqz}
770: \end{equation}
771: is determined by the stability condition of the quadratic terms
772: (both uniform and gradient) in the GL functional towards formation
773: of a state with non-zero $\eta_{n,0}$. From Eqs.
774: (\ref{FGLuniform},\ref{FGLgrad1},\ref{FGLgrad2}), one has the
775: following equation for $T_c(q)$:
776: \begin{equation}
777: \label{Tcq eq}
778: \det\left|\begin{array}{cc}
779: A_{11}+K_{11}q^2 & A_{12}+K_{12}q^2+i\lambda q\\
780: A_{12}+K_{12}q^2-i\lambda q & A_{22}+K_{22}q^2
781: \end{array}\right|=0,
782: \end{equation}
783: where $A_{11}(T)=a_1(T-T_1)$ and $A_{22}(T)=a_2(T-T_2)$, with
784: $T_1$ and $T_2$ having the meaning of the critical temperatures
785: for the bands 1 and 2 respectively in the absence of any interband
786: coupling. The phase transition temperature is obtained by the
787: maximization of $T_c(q)$ with respect to $q$. It is easy to show
788: that the maximum critical temperature corresponds to a state with
789: $q\neq 0$ (a ``helical'' state \cite{MS94}) if the following
790: condition is satisfied:
791: \begin{equation}
792: \label{condition q}
793: \lambda^2+2A_{12}K_{12}-A_{11}(T_{c,0})K_{22}-A_{22}(T_{c,0})K_{11}>0,
794: \end{equation}
795: where $T_{c,0}=T_c(q=0)$. However, even if this condition is
796: violated and the phase transition occurs from the normal state to
797: a uniform superconducting state, there remains a possibility that
798: this uniform state becomes unstable towards the formation of a
799: helical state at a lower temperature. To find this instability,
800: one has to include non-linear terms in the free energy. Because of
801: a large number of the phenomenological parameters in the GL
802: functional with the higher-order terms, the phase diagram of this
803: system is quite rich \cite{MS94}. In particular, there exist
804: various types of helical phases with $q\neq 0$, separated from one
805: another and from the uniform phase by additional phase transitions
806: below $T_c$.
807:
808: It should be emphasized that the origin of the helical
809: superconducting states is different from that of the LOFF
810: non-uniform states \cite{LOFF}. In terms of the GL functional, the
811: LOFF state corresponds to the sign change of the second-order
812: gradient term at some values of the parameters (e.g. of the
813: external magnetic field), while our helical instability occurs
814: because of the presence of the first-order gradient terms.
815:
816:
817:
818:
819: \section{Conclusion}
820: \label{sec: Concl}
821:
822:
823: We have shown that the order parameter in a non-centrosymmetric
824: superconductor with strong spin-orbit coupling has only intra-band
825: components and is always odd with respect to $\bk\to-\bk$, which
826: is a consequence of Pauli principle. This should be contrasted to
827: the case of zero spin-orbit coupling, in which the bands are
828: two-fold degenerate. In that limit, one cannot separate the odd
829: and even components of $\Delta$ because of the lack of inversion
830: symmetry, so the order parameter does not have a definite parity.
831:
832: The Fermi surface of CePt$_3$Si consists of three pairs of sheets,
833: $\alpha$, $\beta$, and $\gamma$, each split by the spin-orbit
834: coupling. Our band structure calculations reveal that the states
835: at the Fermi level are predominantly of Ce-4\textit{f} character.
836: These states are affected strongly by spin-orbit coupling, which
837: leads to the band splitting energy as high as $50-200$ meV. The
838: splitting vanishes along the $\Gamma$ -- \textit{Z} and \textit{A}
839: -- \textit{M} symmetry lines. By far the biggest contribution to
840: the density of states at the Fermi level comes from the $\beta$
841: sheets.
842:
843: Although the large value of the spin-orbit band splitting excludes
844: the superconducting states that correspond to the pairing of
845: electrons from different sheets of the Fermi surface, one can
846: expect that the interband pair scattering will induce gaps of the
847: same symmetry on all sheets of the Fermi surface. Possible gap
848: structure of CePt$_3$Si depends on the dimensionality of the order
849: parameter. If the order parameter corresponds to a one-dimensional
850: representation of the group $\mathbf{C}_{4v}$, then the gap has
851: line nodes where the Fermi surface crosses the high-symmetry
852: planes or the boundaries of the Brillouin zone: at
853: $k_z=0,\pm\pi/c$ for all 1D representations; at $k_x=0,\pm\pi/a$
854: and $k_y=0,\pm\pi/a$ for $A_2$ and $B_2$; at $k_x=\pm k_y$ for
855: $A_2$ and $B_1$.
856:
857: The presence of the gap nodes would manifest itself, e.g. in a
858: power-law behavior of thermodynamic and transport characteristics
859: at $T\to 0$. Although the gap symmetries on all sheets of the
860: Fermi surface should be the same, their magnitudes may be
861: different. The experimental data, e.g. a reduced value of the
862: specific heat jump at $T_c$ \cite{exp-CePtSi}, indicate that only
863: some parts of the Fermi surface have non-zero superconducting
864: gaps, while others remain normal. If this is indeed the case, then
865: the specific heat would drop as $C(T)/T\propto \mathrm{const}\,+a
866: T$ at low temperatures (with the constant contribution coming from
867: the normal sheets of the Fermi surface), which seems to agree with
868: the experimental data of Ref. \cite{exp-CePtSi} in zero field.
869: More detailed information about the pairing symmetry can be
870: obtained only if the precise location of the line nodes is known.
871:
872: The absence of inversion symmetry can also have interesting
873: consequences for the spatial structure of the superconducting
874: phase. We showed that, in contrast to the centrosymmetric case,
875: the Ginzburg-Landau free energy can now contain additional terms
876: which are of the first order in gradients. Such terms can make the
877: superconducting phase unstable towards the formation of a
878: non-uniform (helical) state even at zero magnetic field. For the
879: order parameters corresponding to the one-dimensional
880: representations of $\mathbf{C}_{4v}$, this possibility exists only
881: if the interband pairing is taken into account.
882:
883:
884:
885:
886: \section*{Acknowledgements}
887:
888: The authors thank B. Mitrovic for useful discussions, and also D.
889: Agterberg, E. Bauer, and especially V. Mineev for stimulating
890: correspondence. This work was supported by the Natural Sciences
891: and Engineering Research Council of Canada.
892:
893:
894:
895: \begin{thebibliography}{99}
896:
897:
898: \bibitem{VG85}
899: G. E. Volovik and L. P. Gor'kov, Sov. Phys. JETP \textbf{61}, 843
900: (1985) [Zh. Eksp. Teor. Fiz. \textbf{88}, 1412 (1985)].
901:
902: \bibitem{UR85}
903: K. Ueda and T. M. Rice, Phys. Rev. B \textbf{31}, 7114 (1985).
904:
905: \bibitem{FMSC}
906: K. V. Samokhin and M. B. Walker, Phys. Rev. B \textbf{66}, 024512
907: (2002); K. V. Samokhin and M.~B.~Walker, Phys. Rev. B \textbf{66},
908: 174501 (2002); K. V. Samokhin, Phys. Rev. B \textbf{66}, 212509
909: (2002).
910:
911: \bibitem{FominFMSC}
912: I. A. Fomin, JETP Lett. \textbf{74}, 111 (2001) [Pis'ma Zh. Eksp.
913: Teor. Fiz. \textbf{74}, 116 (2001)]; I. A. Fomin, JETP
914: \textbf{95}, 940 (2002) [Zh. Eksp. Teor. Fiz. \textbf{122}, 1089
915: (2002)].
916:
917: \bibitem{MineevFMSC}
918: V. P. Mineev, Phys. Rev. B \textbf{66}, 134504 (2002); V. P.
919: Mineev, T. Champel, preprint cond-mat/0306471, to be published in
920: Phys. Rev. B.
921:
922: \bibitem{Book}
923: V. P. Mineev and K. V. Samokhin, {\em Introduction to
924: Unconventional Superconductivity} (Gordon and Breach, London,
925: 1999).
926:
927: \bibitem{AB65}
928: P. W. Anderson and E. I. Blount, Phys. Rev. Lett. \textbf{14}, 217
929: (1965).
930:
931: \bibitem{LZ72}
932: A. C. Lawson and W. H. Zachariasen, Phys. Lett. \textbf{38A}, 1
933: (1972).
934:
935: \bibitem{SHC64}
936: J. F. Schooley, W. R. Hosler, and M. L. Cohen, Phys. Rev. Lett.
937: \textbf{12}, 474 (1964).
938:
939: \bibitem{BV83}
940: V. V. Bogatko and Yu. N. Venevtsev, Sov. Phys. Solid State
941: \textbf{25}, 859 (1983) [Fiz. Tverd. Tela \textbf{25}, 1495
942: (1983)].
943:
944: \bibitem{Edelstein}
945: V. M. Edelstein, JETP \textbf{68}, 1244 (1989) [Zh. Eksp. Teor.
946: Fiz. \textbf{95}, 2151 (1989)]; V. M. Edelstein, Phys. Rev. Lett.
947: \textbf{75}, 2004 (1995).
948:
949: \bibitem{GR01}
950: L. P. Gor'kov and E. I. Rashba, Phys. Rev. Lett. \textbf{87},
951: 037004 (2001).
952:
953: \bibitem{BG02}
954: V. Barzykin and L. P. Gor'kov, Phys. Rev. Lett. \textbf{89},
955: 227002 (2002).
956:
957: \bibitem{LNE85}
958: L. S. Levitov, Yu. V. Nazarov, and G. M. Eliashberg, JETP Lett.
959: \textbf{41}, 445 (1985) [Pis'ma Zh. Eksp. Teor. Fiz. \textbf{41},
960: 365 (1985)].
961:
962: \bibitem{MS94}
963: V. P. Mineev and K. V. Samokhin, JETP \textbf{78}, 401 (1994) [Zh.
964: Eksp. Teor. Fiz. \textbf{105}, 747 (1994)].
965:
966: \bibitem{exp-CePtSi}
967: E. Bauer, G. Hilscher, H. Michor, Ch. Paul, E. W. Scheidt, A.
968: Gribanov, Yu. Seropegin, H. No\"el, M. Sigrist, and P. Rogl, Phys.
969: Rev. Lett. \textbf{92}, 027003 (2004).
970:
971: \bibitem{Min93}
972: V. P. Mineev, JETP Lett. \textbf{57}, 680 (1993) [Pis'ma Zh. Eksp.
973: Teor. Fiz. \textbf{57}, 659 (1993)].
974:
975: \bibitem{co-reps}
976: E.~Wigner, {\it Group Theory and Its Applications to the Quantum
977: Mechanics of Atomic Spectra} (Academic Press, NY, 1959);
978: C.~J.~Bradley and B.~L.~Davies, Rev. Mod. Phys. \textbf{40}, 359
979: (1968).
980:
981: \bibitem{LL-3}
982: L. D. Landau and E. M. Lifshitz, {\em Quantum Mechanics}, 3rd
983: Edition (Butterworth-Heinemann, Oxford, 2002).
984:
985: \bibitem{Rashba60}
986: E. I. Rashba, Sov. Phys. Solid State \textbf{2}, 1109 (1960).
987:
988: \bibitem{Bla90}
989: P. Blaha, K. Schwarz, P. Sorantin, and S. B. Trickey, Comp. Phys.
990: Commun. \textbf{59}, 399 (1990).
991:
992: \bibitem{Mad01}
993: G. K. H. Madsen, P. Blaha, K. Schwarz, E. Sj\"ostedt, and L.
994: Nordstr\"om, Phys. Rev. B \textbf{64}, 195134 (2001).
995:
996: \bibitem{Sch02}
997: K. Schwarz, P. Blaha, and G. K. H. Madsen, Comp. Phys. Commun.
998: \textbf{147}, 71 (2002).
999:
1000: \bibitem{Koh65}
1001: W. Kohn and L. J. Sham, Phys. Rev. \textbf{140}, A1133 (1965).
1002:
1003: \bibitem{pw92}
1004: J. P. Perdew and Y. Wang, Phys. Rev. B \textbf{45}, 13244 (1992).
1005:
1006: \bibitem{footnote_gga}
1007: Some trial calculations using the linear muffin-tin orbital method
1008: [O.~K.~Andersen, Phys. Rev. B \textbf{12}, 3060 (1975)] produced
1009: similar results for the LDA and the generalized gradient
1010: approximation [J. P. Perdew, J. A. Chevary, S. H. Vosko, K. A.
1011: Jackson, M. R. Pederson, D. J. Singh, and C. Fiolhais, Phys. Rev.
1012: B \textbf{46}, 6671 (1992)].
1013:
1014: \bibitem{SO}
1015: A. H. MacDonald, W. E. Pickett, D. D. Koelling, J. Phys. C: Solid
1016: St. Phys. \textbf{13}, 2675 (1980).
1017:
1018: \bibitem{CC62}
1019: B. S. Chandrasekhar, Appl. Phys. Lett. \textbf{1}, 7 (1962);
1020: A.~M.~Clogston, Phys. Rev. Lett. \textbf{9}, 266 (1962).
1021:
1022: \bibitem{LOFF}
1023: A. I. Larkin and Yu. N. Ovchinnikov, Sov. Phys. JETP \textbf{20},
1024: 762 (1965) [Zh. Eksp. Teor. Fiz. \textbf{47}, 1136 (1964)];
1025: P.~Fulde and R. A. Ferrell, Phys. Rev. \textbf{135}, 550 (1964).
1026:
1027: \bibitem{And84}
1028: That there is only one type of pairing possible in
1029: non-centrosymmetric systems with spin-orbit coupling, was pointed
1030: out in P. W. Anderson, Phys. Rev. B \textbf{30}, 4000 (1984).
1031:
1032: \bibitem{FAKS03}
1033: P. Frigeri, D. F. Agterberg, A. Koga, and M. Sigrist, preprint
1034: cond-mat/0311354 (unpublished).
1035:
1036: \end{thebibliography}
1037:
1038:
1039: \newpage
1040:
1041: \begin{figure}
1042: \includegraphics[angle=270,width=8cm]{bands1}
1043: \caption{\label{fig_bands1}(Color online) Band structure
1044: calculated without SO coupling.
1045: Three bands (labeled $\alpha$, $\beta$, and $\gamma$) cross the Fermi energy.}
1046: \end{figure}
1047:
1048: \begin{figure}
1049: \includegraphics[angle=270,width=8cm]{bands}
1050: \caption{\label{fig_bands}(Color online) Band structure of CePt$_3$Si with
1051: SO coupling.
1052: Bands are split due to the SO interaction (cf. Fig. \ref{fig_bands1}).
1053: The bands that cross the Fermi energy, are plotted as solid lines.}
1054: \end{figure}
1055:
1056: \begin{figure}
1057: \includegraphics[angle=270,width=6cm]{fs1}
1058: \includegraphics[angle=270,width=6cm]{fs2}
1059: \includegraphics[angle=270,width=6cm]{fs3}
1060: \includegraphics[angle=270,width=6cm]{fs4}
1061: \includegraphics[angle=270,width=6cm]{fs5}
1062: \includegraphics[angle=270,width=6cm]{fs6}
1063: \includegraphics[angle=270,width=6cm]{fs7}
1064: \caption{\label{fig_fs}(Color online) Fermi surface of CePt$_3$Si.
1065: The sub-figures (a)--(g) show cross sections for different values of $k_z$
1066: ($k_z$ is measured in units of $2\pi/c$, and $k_{x,y}$ are measured in units of
1067: $2\pi/a$).
1068: Only one quarter of the Brillouin zone is shown.
1069: By applying the reflection symmetries in the $k_y = 0$ and $k_z = 0$ planes,
1070: the Fermi surface in the entire Brillouin zone can be recovered.
1071: The labels of the sheets correspond to the labels in Fig. \ref{fig_bands1}.
1072: Dots indicate the $\mathbf{k}$ points for which we calculated the
1073: band energies.
1074: The Fermi surface was obtained by interpolating between these points.}
1075: \end{figure}
1076:
1077:
1078:
1079: \end{document}
1080: