Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Masa-aki Ozaki
Department of Physics, Kyoto University, Kyoto, 606 Japan
Kazushige Machida*
Department of Physics, Kyoto University, Kyoto, 606 Japan
and Laboratory of Atomic and Solid State Physics, Clark Hall, Cornell Uniuersity, Ithaca, 1Vew York, 14853 2501-
(Received 20 June 1988)
Stimulated by recent remarkable findings which lead to a common feature of all four known
heavy-fermion superconductors — UPt3, (U, Th)Be», URu2Siz, and CeCu2Siz, namely, a close prox-
—
imity between the magnetism and unconventional superconducting pairing state we present a gen-
eral framework on how to identify the superconducting classes. Taking into account the existence
of the spin-density wave (SDW) above T, in UPt3 and URu2Si2 and below T, in (U, Th)Be» and
CeCu&Si2, we enumerate group theoretically the pairing states compatible with given space-group
symmetry according to the Landau subordination scheme for the second-order phase transition.
This enables us to eftectively narrow down the possible states. These enumerated states are quite
general, independent of particular models or pairing interaction forms. Then we give some physical
arguments on the relative stability among the allowed states. Maximally utilizing experimental and
theoretical data available at present, we attempt to decrease the number of the possible states and
obtain the most plausible candidates for each material. To confirm the emerging universal feature
that the ground state in heavy-fermion superconductors is the coexisting phase with the magnetism,
further experiments are proposed, based on our results.
HFS's by summarizing our work and examining other' s the space (P), spin rotation (S), gauge (@) and time re-
work in the context of our work. versal (t ) group
The notation used in this paper are the same as those
used in our previous papers. ' ' We use the SchonAies G, =PXSXR, (2. 1)
notation for point groups and their elements, and refer to where R =N+tN.
a book by Bradley and Cracknell for the general group The presence of the SDW lowers this symmetry and
theory. the system acquires the invariance group G C: Gp.
Generally, the invariapce group 6
can be obtained '
II. ENUMERATION AND CLASSIFICATION from the irreducible representation over real number field
OF PAIRING STATES (from now on we denote it as "rep") of Go such that
lr(k+K) =lr(k), with K being a reciprocal vector CeCuzSiz) or by the linear specific-heat behavior. Al-
whose fundamental vectors are g; (i =1,2, 3) is obtained though no one has succeeded in describing correctly this
by projecting the periodic function transforation, once the coherent state is established we
can start out with the renormalized normal Fermi liquid
exp[2~i(n&k, +nzk 2+n3k3)] (2.9) where the quasiparticle masses have become very heavy
on the irreducible component of -the rep y of H where' by the mass renormalization. The residual interactions
n,
is an integer and between the quasiparticles are also expected to be rela-
tively weak, permitting us to treat it in terms of a mean-
k k )g$ + k2g2+ k3g3 ~ field theory (we will discuss some phenomena deviated
from this behavior in Sec. VII), namely, we start with the
following mean-field Hamiltonian with two kinds of in-
2. The superconducting states from which the SDW bifurcates teractions; the intra-atomic Coulomb interaction U, re-
We consider the superconducting states which allow a sponsible for SDW, and the attractive interaction
certain SDW to bifurcate via a second-order transition. g „„(k, ) for SC, which are minimum ingredients to
. k
Below Tz ( & T, ) two orderings coexist. The enumera- discuss the interplay between them:
tion of the pairing states is much the same except that we H =Hp+Hsow+Hsc (2. 12)
now consider the space group P in the normal state. If
we consider the translationally invariant pairing function
only, then the overall group Gp is given by
Ho g E„C„C„ (2. 13)
l&(k )&0, and l&&(k )&0 (class III) is intermediate between odd. The previous group-theoretic classifications' ' in
them. the strong spin-orbit coupling limit based on point group
The complete classification scheme which includes the symmetry do not allow the odd-parity state with the line
odd-parity states is as follows: node, and this conclusion hardly changes even if we ex-
tend it to space-group symmetry, as will be seen. These
class I: (E, O), (0~~, 0), and (Oj, E), experimental facts sharply contradict with the
class II: (E, E ), ( 01,E ), and (0~, 0 ), classification scheme. Furthermore, according to Scala-
pino et al. and Miyake et al. , antiferromagnetic Auc-
class III: (E, E+0), (O~~~, E+0), and (O„E+0), tuations as a pairing mechanism which are observed ex-
perimentally, directly by inelastic neutron scatterings
in which (x, y ) denotes that the parity is x (odd or even) and indirectly by specific-heat measurement, favor the
and the translational symmetry by Q is y (odd, even, or even-parity d-wave-like state over the odd-parity states.
l
mixed). And and mean whether the spin-quantization
~~ This also leads us to a puzzle.
axis is perpendicular or parallel to the magnetization M. Quite recently a totally unexpected discovery has been
Under the presence of SDW, the pairing states in class I reported by Aeppli et al. via the neutron diffraction that
should be chosen because their T, 's are highest among UPt3 exhibits an antiferromagnetic transition at Tz =5 K
these classes, and always coexist with the SDW below T, . with the saturation moment -0.
02pz/U atom. The
By solving the coupled self-consistent Eqs. (2. 16) and superconductivity coexists with the SDW below
(2. 17) for each class, it is found that among the states in
class I the states, whose SC nodes maximally overlap
T & T, -0. 5 K. We notice that the SDW pattern is ex-
actly the same as in the dilutely doped systems (U, Th)Pt3
with the SDW gap in reciprocal space, are stabilized be- (Ref. 29) and U(Pt, Pd)3 (Refs. 30 and 31), except that the
cause the condensation energies of two orderings are -0.
moments are 6pz/Uatom in the latter two materi-
gained most effectively. (Actually, we have done it nu-
merically' '"
for a two-dimensional model system. The
als. This finding inevitably urges us to reexamine the pre-
vious thoughts on the UPt3 problem, particularly on the
quantitative conclusion referred to here does not change
nature of the pairing to resolve the preceding puzzles.
for the three-dimensional case. ) Apparently this con-
Here we point out that we could narrow down the can-
clusion is model independent. Note that the magnetiza-
didates of the possible pairing states because the presence
tion M, as a function of temperature, is strongly
(moderately) suppressed below T, in class II (I), yielding
of the SDW greatly lowers symmetry compared to that in
the paramagnetic state, severely restricting the forms of
a good indicator to distinguish between the two classes.
So far we consider the case Tz T„ the converse is ) the pairing functions allowed. We fully utilize the advan-
tage that we now have the detailed knowledge of the
also true. Namely, in T& & T„ in order that the two or-
Fermi-surface topology obtained by the band calcula-
derings coexist, the states should belong to class I and the tions. These are known to be in excellent agreement
SDW gap and SC nodes on the Fermi surface (FS) should with the de Haas —van Alphen experiments.
maximally overlap to minimize the destructive interfer-
ence. B. Group theoretical classification
Apart from the relationship between two orderings, the
pairing state itself is required to have some properties We enumerate possible pairing states compatible with
from physical reasons; pairing states with the lesser num- the crystal symmetry of UPt3 (hcp structure, space group
ber of the nodal points, or lines of the pairing function P63/mmc) and also with the SDW symmetry observed.
within the first Brillouin zone (1 BZ), should be chosen. The positions of the U atoms in the unit cef1 are given
This is analogous to the situation in spherical symmetry by (a/v 3, 0, — c/2) and (a/v 3, a/2, c/2), and the
like He where the pairing states with the lesser angular equivalent points. The lattice translational vectors are
momentum (s, p, d, etc. ) are considered a priori as a can-
didate. v'3 a
a, ——, 0, tz=(O, a, O), t3=(0, 0, c) . (3. 1)
T=L(t„t2, t3) is the three-dimensional translation group. P(Q)(s;Iw;+t)=exp[ —iQ (w;+t)]D2h(s;) . (3.4)
Since K; w. =0 for all i and j, where K; =S; 'Q — Q, the
ray representation of the little cogroup P(Q) has the fac- The observed SDW state is derived by the rep B3„of
tor system D2h. We have
k] + k ky k2 aky k3 (3.11)
3a
functions as will be seen shortly. Table I indicates the
states thus obtained. Combining these orbital functions
l(k) with the spin functions (l . =io o ), we get a com-
plete list of the pairing functions, b, (k), both with even
and odd parity.
Similarly, we can enumerate the states in the strong
FIG. 1. The upper half of the first Brillouin zone of the hcp spin-orbit coupling limit where the space and spin sym-
crystal. metries are coupled. The relevant space group becomes
4150 MASA-AKI OZAKI AND KAZUSHIGE MACHIDA 39
TABLE I. The odd- and even-parity states in the absence of the spin-orbit coupling within the sector
in~ = I for UPt, . The pairing function b(k) with even parity is given by h, (k) =I, ro and that with odd
parity is l, 7„,lj- 7, and l, „(7y+i 7, ), where = A1, B1, B„and B3. j
Sector
(n1 n2 n3 ) even Qdd
v'3 — V3 —
(+1,0, 0) lg = cos ak cos
a
k lg = sin ak cos
a
k
2 2 3M 2 2
l~ = sin &3
ak
.
sin —
a
k lg = cos V3
ak sin —
a
k
1g 2 2 2Q 2 2
(0, +1,0) l& =cos(ak ) l~ 2Q =sin(ak )
the double group»D6I„and the invariance group G II B, the following states belong to I for odd parity:
of the SDW state is given by
b, '"(k ) =l~ (k )r„
G =[( D+rI D)
=sin(v'3ak /2)cos(ak» /2)r„,
+I ti n(D 2+&nD2)]L(2t„t2, t3)4 (3. 12)
b, '
'(k)=lq (k)r„
where
=cos(&3ak /2)sin(ak» /2)r„,
nDQ [Ey ( C2zu2z ~t3/2), (C~~ u2„~t3/2), (C~»u2» ~0)] (3. 15)
'(k ) =I~ (k )r, =sin(ak )r
(3. 13)
By considering only the translationally invariant pair- bi '(k)=l~ (k)r =sin(ck, )r
ing function, the relevant rep of G is reduced to the
rep of and for even parity:
in Table II. in the absence of the spin-orbit coupling (Table I). This
choice is reasonable because (1) under the SDW the states
C. Possible states in I are more stable than those in II, and (2) the suppres-
Having enumerated the possible states in Tables I and sion of M(T) below T, is rather moderate (see Fig. 3 tn
II, we now classify these into states, less competitive Ref. 5), indicating that the pairing state should belong to
(class I) and competitive (class II) with the SDW. Ac- I rather than II (compare Figs. 6 and 8 in Ref. 14). We
cording to our classification scheme introduced in Sec. notice that the Knight-shift experiment points to the pos-
TABLE II. The odd- and even-parity states in th strong spin-orbit limit within the sector ~n~ = I for
UPt3. The pairing function b (k ) with even parity is A, (k ) = lj 7p and that with odd parity is given by
the linear combination of the form: 5& (k) =C1l& 7, +c2lg 7y+C3lg 7g) Ag (k) =c, lg 7,
Q 1M 2M 3M 1M M
+ c3 lp 2M
Sector
(n1 n2 n3 ) even odd
v'3 — —
(+1,0, 0) l = cos ak X cos
a
k l = cos —
3
ak
.
sin
a
k
2
l =sin &3
ak
.
sin —
a
k l = cos —
a
k
.
sin
&3
ak
2 2 2 2
(0, +1,0) l~ =cos(ak ) l~
2Q
=sin(ak )
Auctuations at the zone-boundary positions along the The corresponding reciprocal vectors are
I
g, =
=2" ( — 1 1 1), g~=
=2~ (1 1, —1) .
=2K (1, —1 1), g3= (4.2)
The space group P can be written by the SDW characterized by Q=(2'/a )(0, 0, 1); the [001]
instability (D4 case), or the SDW by Q=(7r/a)(1, 1, 1);
P= g(S, ~w, )L(t„t„t,), (4.3) the [111]instability (D3 case). The fifth column denotes
the types of the gap nodes; plane, line, or none. The actu-
where S,. EO&, w; =(0, 0, 0) for S,. HO, and al topology of the gap node is determined by the intersec-
tion of the above node and the Fermi surface. We have
Wj 2 (t]+t2+ t3) also constructed the explicit forms of the order parame-
ters, where only selected l(k) are shown such that the
for 5, EIO. The overall symmetry group Go of the nor-
number of the nodes is the smallest ones within 1 BZ.
mal state is given by
G, =PxSxR, (4.4)
C. Possible pairing states
when the spin-orbit coupling is absent. The pairing states
are characterized by the rep of Go Having enumerated the possible states compatible with
G =PS 'R, i =0, 1 . (4.5) crystal symmetry, we now go on to the next step. As
mentioned before, the SC gap vanishes at lines on FS for
When the pairing function is translationally invariant, UBe, 3 and probably also for Th-doped systems. We can
this reduces to eliminate the states with nodes vanishing at points in
Tables III and IV. This also excludes the odd-parity
G O=Oq(3S'g R 2, (4.6) states in the strong-orbit-coupling case. In order that the
SDW characterized by Q=(2'/a )(0, 0, 1) occurs via a
where 0& consists of 3,
, A2, E, T, , and T2 representa- second-order transition, the pairing state must have D4
tions. We obtain the pairing functions with even parity as a subgroup. These requirements leave the states with
in Table III and with odd parity in Table IV. Note that even parity and with odd parity as shown in Table V. We
in the even-parity case the states are equivalent to ones have characterized these remaining states as regarding
listed by Volovik and Gor'kov, ' and the results in the the classes I, II, or III (third column) and the matching
odd-parity T]„case are those listed by us. The fourth condition between the SDW gap and the line nodes
column in Tables III and IV indicates explicitly whether (fourth column) where F ( U) denotes a good (bad) overlap
each invariance group contains the symmetries D3 or D4. between them. As for the even-parity states, under the
This provides the information on whether each order pa- conditions we imposed, the following three states are
rameter can exhibit, via a second-order phase transition, most plausible:
39 SUPERCONDUCTING CLASSES OF HEAVY-FERMION MATERIALS 4153
ak, ak ak
a'T2'
2g
(k) =sin
2
sin
2
+i sin 7p e
Even-parity pairing functions and their invariance groups for O„jn (IJ,Th)Be/3 (x Qk /2 y ak /2 and
z =ak, /2).
l] ~p D, xR D4 line
lz&p D4(Dz }XR D4 plane
(l]+ilz) jP O(D2 )' D3, D4 line
(lI, lz): (cosx cosz+cosz cosy —2cosx cosy, &3(cosx cosz —cosy cosz))
(cos2x + cos2y —2 cos2z, &3(cos2y — cos2x ) )
(2cos(2z)cosx cosy —cos(2x)cosy cosz —cos(2y)cosz cosx, ~3[cos(2x)cosy cosz —cos(2y)cosx cosz])
TABLE IV. Odd-parity pairing functions and their invariance groups for OI, in (U, Th)Be» (x = ak, /2, y = aky /2, z = ak, /2).
Representation Order parameter Invariance group Subgroup Type of node
A1„ IP (1+Iu )OA(e, ) XR" D3, D4 plane
I(~ +i~y} (1+Iu „)(1+tu 2 ) OA (e, )' D3, D4 plane
(I1 l2& l3 ): (s1nx (cosz cosy ), slny {cosx cosz ), slnz{ cosy cosx ) )
(sinx [cos(2y )cosz — —cos(2y )cosx ] )
&
cos(2z )cosy ], siny [cos(2z )cosx —cos(2x )cosz ], sinz [cos(2x )cosy
'within the same representation the linear combination of l(k )'s is also allowed.
In these triplet superconducting states all states are invariant for Iw%. and I.(t], t2, t3). Thus in this column (1+IwFr}L(t1,t2, t3) is
omitted for brevity.
'A(e, ) = [u(e„8)9~0 ~ 8~2m j.
0 '(D2) = T(D2)+ C2bu2xtT(D2), T(D2) =D2+ C31(25./3)D2+ C31( —2'/3)D2.
IIO{D2 ) IIT{D2 )+ C2b u2 IIT{D2 ) IIT{D2 ) D2+ C31u (e 2~/3)D2 + C31u {e 2~/3)D2'
IIC3 = E+ C3+1u (e„2m./3)+ C31u (e„— 2m/3).
~C2, =E+ C2, .
IIC31 — E+ C3+1u ( C3+1 )(25./3)+ C31u {C» )( —2m/3).
q C4 = E + C4z u 2a {n/2 ) + C4z u 2a ( /2 ) + C2z ~ FT
D 2{C2a ) C2a C2b ~C2a +
II 3 IIC3+ C2bu2y I C3
D4(D 2 ) =D 2 + C4zFTD 2s D 2 = E + C2a + C2b + C2z ~
C2b E+ C2b.
"IIC31 =E+ C3+1u (C3+1 )(2m/3}+ C31u (C31)( — 2m/3).
'IIC 4=E+ C4+, u2, (5./2)+ C4, u2, ( —S./2)+ C2, %..
D 2( C2b ) C2b + C2a ~C2b C2b E + C2b &
TABLE V. Favorable states among the enumerated states in Tables III and IV, which have node planes and D4 as a subgroup
(x =ak„/2, y =aky/2, and z =ak, /2). The "matching" column denotes the matching between the SDW gap and the SC node; favor-
able (F) or unfavorable ( U).
Representation States l(k) Class Matching
A1g cos(2x )cos(2y )cos(2z ) II F
A2g cos( 3x ) [cos( 2y ) cosz —cos(2z) cosy ]+cyclic III U
cosz( cosx — cosy ) I U
cos(2x ) —cos(2y ) II U
T1g cosx [sin(2y )sinz — sin(2z)siny ] I F
cosx [sin(2y)sinz — sin(2z )siny ]
+i cosy [sin(2z )sinx —sin(2x )sinz] III F
T2g siny sinz I
sinz{siny+i sinx ) I F
cos(2x )sin(2y )sin(2z ) II F
sin(2z )[cos(2x )sin(2y )+i sin(2x )cos(2y )] F
A, „ sin(3x ) [sin(2y )sinz — sin(2z )siny ]+cyclic III F
A2„ sin(2z )sinx siny + cyclic III F
sinx siny sinz I&II F
sin( 3x ) [sin(2y )sinz + sin( 2z) siny ] + cyclic III F
T1u sinz(cosx + cosy ) I&II F
sinx (cosz+ cosy )+ siny(cosx + cosz ) III U
sin(2z ) I&II F
sin(2x )+ s1n(2y ) III U
sin(2z )cos(2x )cos(2y ) I&II F
sin(2x )cos{2y )cos(2z ) + sin(2y )cos(2x ) cos{2z ) I&II U
T2. sinz(cosx — cosy ) I&II F
sinz [cos(2x )cosy — cos(2y )cosx ] I&II F
4156 MASA-AKI OZAKI AND KAZUSHIGE MACHIDA 39
It is worthwhile to mention that since, for example, we can provide a possible phase diagram which should be
a"' checked experimentally as shown in Fig. 4: Since there is
T2 (k+Q x
)=aI"
T2 (k), no second transition observed so far in x &x, 017, „=0.
(4.8) the pairing state should belong to Class II which
suppresses T& to quite a low value. As x increases, the
with Q„=(2m. /a )(1,0, 0), Q =(2m/a )(0, 1, 0), and nesting becomes improved. Then in x the pairing )x„
state changes into Class I and induces the SDW at T, 2
Q, =(2m/a )(0, 0, 1), the pairing state which breaks cubic
below in which the two orderings coexist to effectively
symmetry restricts the SDW pattern to appear below T, 2,
minimize the two kinds of the condensation energies.
that is, in this example the SDW vector should be Q or
Therefore, there must exist a phase boundary between
Q, . We note that b, 'T'2g (k) is most possible because the Class II in x & x,„and the coexisting phase: I + SDW in
node planes within 1 BZ are least among the three states.
As for the odd-parity states, the following six states are
)
x x,„. In fact, as demonstrated by Maple, the pairing
states in both regions behave quite differently under Gd
equally plausible among Table V: doping. As x increases, (E, E)H (Class II) transforms
ak ak ak, into (E, O) H (Class I) in the even-parity case, or (O~~~, O)
b, ~ (k ) =sin Sln sin +z into (Oi, 0 ) in the odd-parity case.
2 2 2
a' 0,'c—
—, —,
a' —
a' 0,
1 1 1 1 1 1
g, =2~ 0 'a' c
go=2m g3=2~ (5.2)
The SDW ordering vector Q is given by line nodes of the SC gap on FS. The Knight shift of Si
NMR below T, does not change a.ppreciably, implying ei-
Q=-,'(gi+g2+g3) (5.3)
ther an odd-parity or even-parity state in a large impurity
which corresponds to I Z vector in 1 BZ. The magneti- scattering limit. The determination of the parity is still
zation M is parallel to the tetragonal axis (c axis). an open question. In the following we seek both possibili-
As for the superconducting state coexisting with the ties in this system.
SDW, the power-law behaviors of the temperature depen- According to Norman et ah. , who have done the band
dences characterize the NMR relaxation rate ( —T ) structure calculation on this system, there indeed exists a
found by Kohara et al. and ultrasonic attenuation nesting feature between the electron Fermi surface at I
(— T ) by Fukase et al. , 5 both being consistent with the and hole Fermi surface at Z along the I Z line, coinciding
39 SUPERCONDUCTING CLASSES OF HEAVY-FERMION MATERIALS 4157
B. Group-theoretic enumeration
k, = (
—ak +ak 3 +ck, ),' k2=
2 (ak, —ak +ck, ), k3=
3
1
(ak +ak —ck ), (5.6)
4 3
4
TABLE VI. Possible states with even and odd parity for URu2Si2. The pairing function is given by (1) 17p, (2) 17„17„,and
l(7 +i 7y ) for one-dimensional representations with even and odd parity, respectively. The two-dimensional representations (l&, 12);
1, 7p, (1] +il2 )7p and (1 +12 )7p for even Parity, and I [7„(l +il2 )7„1&7„,I 7 +127y (1[+1/2 )7z) 1& (7 + l 7y ), and 1]7 +il~7y for odd
& & &
parity.
Sector
(n] n2 n3 ) Representation States l(k) class Matching
ak„ ak ck,
(1,0,0) cos cos cos
2 2 2
ak ak ck,
(0, 1,0) cos cos sin I&II
2 2 2
ak„ aky ak,
(0,0, 1) B,„ sin sin sin I&II
2 2 2
ak ak ck,
B2g sin sin cos
2 2 2
l
sin —(k —k
a
2
) sin,
. «.
2
' sin
. —(k + k
2
a
) sin
ck,
2
U
E„ sin —
a
2
(k —k ) cos,
«,
2
.
' sin
—
a
2
(k„+k ) cos
ck,
2
I&II
(1,1,0) cos(ck, ) II
sin(ck, ) I&II
is distinctly the best choice because (1) the number of the This is a first attempt toward this direction for these im-
node planes is the smallest, (2) it belongs to class I, (3) the portant materials.
SDW gap and the SC nodes are maximally overlapped, The pSR experiment on CeCu2 &Siz points to two possi-
and (4) the intersection with the FS s gives rise to line bilities to explain the appearance of the internal field
nodes, which is consistent with the experiments. below T-0. 8 K, roughly the same temperature of the SC
transition (T, -0.
7 K); SDW or spin-glass transition.
VI. CeCu2Si2 The independent Si NMR experiments ' detect a sud-
den broadening of the distribution of the internal fields at
A. Preliminaries a certain temperature T~ under an applied field. The
zero-field extrapolation of T& in the NMR results
This material is the first heavy-fermion superconductor smoothly comes to T~ -0.
8 K, indicating that both
which opened a new branch of physics. Yet it keeps giv- probes are seeing the same phenomenon. In addition, ac-
ing us surprises. The recent finding of the static magnetic cording to Kitaoka et al. , there exists further evidence
ordering by the zero-field p, SR [Ref. 6(a)] urges us to for static SDW order competing the SC order: (1) The T
reconsider the physics of this system from the outset. dependence of the nuclear relaxation rate (T, ' ) shows a
39 SUPERCONDUCTING CLASSES OF HEAVY-FERMION MATERIALS 4159
board peak at around T-0. 8 K under zero field. (2) Un- B. Group-theoretical classi6cation and possible states
der an applied field the internal field probed by the NMR
ceases to grow below T, . (3) The material CeCu, 9Siz
which exhibits no SC does not show the magnetic transi-
The possible pairing states both for even and odd pari-
tion either. In view of the close proximity between the
SC and SD%' which is a common characteristic in the ty, which occur from the normal state and are compatible
with the given space group (I4/mmm, D~s ), are listed in
HFS family, we regard it as a SDW transition.
The superconducting properties have been investigated Tables VII and VIII. They are obtained from
extensively for a long time. There is, however, little con-
sensus as to the precise nature of the pairing function, al- exp[2iri(n&k, +nzk~+n3k3)]
though it is certainly unconventional. As T decreases,
the system becomes a Fermi-liquid state around T — = 1 —2 within the sector ~n;~ ~ l. (Note that the space group is
K below which T& T= const is obeyed. At T„T, ' does the same as in URuzSiz, thus keeping the same rotations. )
not exhibit enhancement characteristics in an ordinary Since the experiments strongly point to the even-parity
BCS superconductor, indicative of unconventional pair- state, we consider only this possibility hereafter. We
ing. At lower temperatures, T, ' ~ T,which is con- have enumerated twelve even-parity states in Table VII.
sistent with line nodes on FS. The thermal conductivity We can reduce this number if we suppose that the SD%'
( — T) is also consistent with the line node. The system vector Q is commensurate with the underlying lattices:
shows a substantial Knight shift ' below T, for both com- namely, Q =(2ir/a, 0, 0) or Q, =(0, 0, 2ir/c), then we
ponents parallel and perpendicular to an external field, can specify the operation
implying that the parity of the pairing function is
definitely-even, not odd parity pinned to the crysta11ine b, (k+Q) =+A.(k ) .
axis.
There is unfortunately neither the band structure cal- Under this assumption, only a few states remain which
culation available to examine a nesting feature, nor in- belong to class I, all others belong to class II unfavorable
formation of the SDW pattern in this system. to the SDW. Namely, we obtain
TABLE VII. Superconducting states with even parity from the normal state in D4&. The states are enumerated within the sector
~n;~ 1. The classification in the fourth co1umn is done according to l(k+ Q ) =+l(k ) where Q=(2ir/a, 0 0) or (0 0 2ir/c ).
~
cos
ak„aky
cos cos
ck,
2 2 2
cos(ak„)+ cos(aky )
cos(ck, )
cos(ak )cos(aky )
[cos(ak„) + cos(ak» ) ]cos(ck, )
AqgS none
B,gS —cos(aky )
cos(ak
)
—
[cos(ak„) cos(ak» )]cos(ck, )
ak aky ck,
BpgS sin
2
sin cos
2 2
sin(ak„)sin(aky )
TABLE VIII. Superconducting states with odd parity from the norma1 state in D&I, . The states are enumerated within the sector
fn, ]~i.
Representation Order parameter Irreducible basis l(k)
A, „gS' l(k)r, none
l(k )(z„+i~y )
t'
ak~
A, „gS' akim akim
l(k )~, cos cos cos
2 2 2
l(k )(w +i ~y ) sin(ck, )
ak ck,
B,„S '
l(k)~, sin
2
sin(ak~ )sin ' 2
l(k)(~ +i~ )
l(k)(~ +i~ )
ak ak ck,
b, „(k =cos
[g
)
2
cos
2
cos
2
&p
b, s (k)=sin
ak
sin
gk„ck,
cos &p p
2g 2 2 2
= sin —( k + k ck,
g'E" ( k ) ) sin 7p (6. 1)
2 2
ck,
b, '~'(k )
= sin —
a
(k + +k y ) +1 sin —
a
(k X —k ) sin 7p
2 2
We cannot proceed further to narrow down the possible superconductors; UPt3, (U, Th)Be», URuzSi~, and
states without the detailed band. calculation or informa- CeCu2Si2, namely, a close proximity between the magne-
tion on the Fermi surface topology and the structure of tism and unconventional pairing state, we gave a general
the SDW. framework on how to identify the superconducting
classes, compatible with given space-group symmetry.
VII. DISCUSSIONS AND SUMMARY We took into account the existence of the SDW above T,
Stimulated by recent remarkable findings which lead to in UPt3 and URuzSiz and below T, in (U, Th)Be» and
a common feature of all four known heavy-fermion CeCu2Si2 to enumerate the pairing states. The Landau
39 SUPERCONDUCTING CLASSES OF HEAVY-FERMION MATERIALS 4161
subordination scheme for the second-order phase transi- the lower critical field Hc& below T,2, which leads them to
tion effectively narrows down the possible states. The propose the phase transition between different pairing
enumerated states are quite general, independent of par- states where the lower-temperature phase gains an addi-
ticular models or pairing interaction forms. Then we tional SC condensation energy by opening an energy gap
gave some physical arguments on the relative stability of a different state on a different part of the FS. It is pos-
among the allowed states. Within the limited data avail- sible, however, to gain additional condensation energy by
able at present, we attempted to reduce the number of the inducing the anomalous pairing state:
possible pairing states and pointed out the most plausible
candidates for each material. ~Q ~ (Ck+Qtc —kL ~
Here we give supplementary discussions for some of which is allowed under the coexistence phase between SC
the materials and propose possible experiments. and SDW because a coupling term 6Mb & in the free en-
ergy expression is always possible on symmetry reason,
A. Upt, and this coupling inevitably induced 6& which is not
The ultrasonic-attenuation measurements, under an translationally symmetric.
applied field by Muller et al. and gian et al. , detect a
sharp anomaly well below H, 2, indicative of a field- C. URu2Siz
induced phase transition. Volovik interprets it as aris- According to the specific-heat and NMR experiments,
ing from the phase transition associated with vortex- 30 —40% of the total Fermi surface is annihilated by the
lattice structure change. He assumes that a pairing state formation of the SDW below Tz. A.s mentioned in Sec.
belonging to the two-dimensional E representation V, the nesting does not seem to be so good since only a
among D6(E) at zero field is split into the states belong-
small portion of the FS is gapped by the SDW. We point
ing to the one-dimensional representation due to the hex- out, however, that to evaluate the portion of the gapped
agonal symmetry breaking in the basal plane under a region on the FS, we must consider the effective masses in
field. This interesting interpretation, however, does not
that region, which are dynamically enhanced, in addition
match with our identification in Sec. III. In our
to the topology of the Fermi-surface shape.
classification scheme there are no states belonging to E
representation because the hexagonal symmetry is al-
ready broken by the preexisting SDW as shown in Sec. D. Theoretical issues
III B. In our opinion, these experiments deserve further
theoretical investigation. Here we like to discuss some controversial issues on
the theoretical side of the heavy-fermion superconduc-
tivity: Although we have assigned an odd-parity state as
a most possible state in UPt3, a prevailing idea contra-
g. (U, „Th„)Bel3 dicts this choice in favor of an even-parity state for the
following two reasons: (a) All the odd-parity states in the
The origin of the second transition at T, 2 for x & 0. 017 strong spin-orbit limit have point nodes. (s) The attrac-
is much debated and actively discussed from various as- tive interaction arising from the SDW fluctuations
pects. There are two opinions; one is due to the SDW favors even parity d-wave-like states over odd-parity
which is to coexist with the SC as ours, and the other is states.
due to the phase change between different classes of pair- Concerning (1), we have already given detailed argu-
ing states. The most thorough study on the latter point ments in Sec. IIID in connection with UPt3. We only
of view is done by Sigrist and Rice, who account for the point out here the experimental fact that all four HFS's
appearance of the internal field below T, 2 by identifying show line nodes of the SC gap in common, and refer the
the low-temperature phase as some nonunitary state. A argument by Miyake again that the effective spin-orbit-
crucial difference between ours and their theory lies in coupling field by a Cooper pair is not necessarily the
the different interpretation on the origin of the static strong limit.
internal field; whether it is a static long-range order Concerning (2), as pointed out in Sec. III D already,
(SDW) as ours, or randomly distributed internal fields the RPA type calculation done by Scalapino et al. can
carried by individual Cooper pairs as othe''s. This could lead to an attractive interaction for some of the odd-
be distinguished by the neutron diffraction experiment. parity states. The important factor here is the com-
An interesting aspect of the phase diagram in T versus bined symmetry; parity symmetry and the translation
x of these materials is the possible existence of the quar- symmetry by Q.
ternary critical point at x,„-=0.017, where the four As for the origin of the pairing mechanism, the spin-
second-order transition lines meet as shown in Fig. 4. Auctuation theory proposed by Scalapino et al. should
The missing fourth line, which separates the states in be extended to the case where the SDW is ordered. The
Class I coexisting with SDW and the states in Class II, spectrum of the dynamical spin susceptibility, which
might run nearly vertically, making its detection hard. plays an important role of mediating the attractive in-
We believe that much physics can be explored near this teraction, is certainly modified by the long-range order.
concentration region. This would show up in a more delicate interplay between
In this connection we should mention that the SDW in the position of the gap node. (We only know preliminary
these HFS's itself is very anomalous, reAecting the results on UPt3. ) The doping experiments in (U, Th), Pt3
dynamical nature of the so-called Kondo coherent state and U(Pt, Pd)3 should be done for small doping regions to
in the normal state. The saturation moments (po) are ex- see the relationship between the suppression of T, and
tremely small compared to the high-temperature effective rapid growth of the spontaneous moment.
moment p, s. (CeCuzSi2' pp 0, 1/p ft 2 68' UPt3.. (2) (U, Th )Be,3 is an extremely interesting system.
0.02/2. 6-2. 9; and URu2Si2. 0.02/3. 51). The remark- To establish the phase diagram T versus x, the experi-
able example is that dilute doping of the Th or Pd in ment near x -0. 017 is waited. We believe that there is a
UPt3 increases po by a factor of 30, yet the SDW patterns good chance to hit the missing fourth transition line. We
are exactly the same. The T dependence of the SDW or- also like to see the specific-heat behavior in UBe, 3 and
der parameter in UPt3 deviates from the mean-field be-
havior.
)
also the system with x 0. 017 at very low temperatures
to confirm the line nodes in both regions.
All experiments on the four HFS's point to line nodes (3) The detailed neutron diffraction in URu2Si2 should
of the SC gap in common ' except for the specific-heat be done to see the subtle change of M below T, associated
measurements' on CeCu2Siz ( —T )and'UBe&3 with the coexistence. The existing data '"' available are
(— T ) and the London-penetration-depth measure- too coarse to see this interplay.
ment on UBe» indicative of point nodes. As pointed (4) The elastic neutron diffraction experiment on
out by Varma, these thermodynamic quantities might CeCuzSi2 is highly required to confirm the long-range na-
not reAect the topology of the gap node because the ture of the SDW below T&-0. 7 K under zero field and
asymptotic Fermi-liquid state is not reached even below also under applied field. The change of the magnetic ex-
T„making the identification of the node topology from citation spectrum below T„which can be observed by
equilibrium properties unreliable compared to transport the inelastic neutron experiment, might help to identify
properties. Although we do not have a definite answer the origin of the pairing mechanism.
why the line node is common, we do observe that the
pairing states in which we take into account the lattice
ACKNOWLEDGMENTS
periodicity often have periodically placed node planes
perpendicular to the translation vector whose intersec- We thank M. Kato for collaboration on some of the
tion with multisheeted Fermi surfaces easily gives rise to present work, and M. Norman and T. Oguchi for provid-
the line nodes of the SC gap. ing their results prior to publication. We are grateful to
We propose the following experiments to identify the Y. Kitaoka, K. Asayama, Y. Onuki, T. Komatsubara, G.
pairing states: Aeppli, J. L. Smith, and D. E. MacLaughlin for discus-
(1) The electron tunneling experiment helps to locate sions on their experiments.
*Present address: Department of Physics, Kyoto University, Gutsmiedl, F. Gross, K. Andres, H. D. Yang, and R. N. Shel-
Kyoto, 606 Japan. ton, Bull. Amer. Phys. Soc. 33, 363 (1988).
'T. M. Rice, Jpn. J. Appl. Phys. 26, Suppl. 26-3, 1865 (1987); A. ' K. Machida and M. Kato, Phys. Rev. Lett.
58, 1986 (1987).
I. Goldman, ibid. 26, 1887 (1987); Z. Fisk, H. R. Ott, and G. M. Kato and K. Machida, J. Phys. Soc. Jpn. 56, 2136 (1987).
Aeppli, ibid. 26, 1882 (1987). ~3M. Kato, K. Machida, and M. Ozaki, Jpn. J. Appl. Phys. 26,
~Z. Fisk, D. W. Hess, C. J. Pethick, D. Pines, J. L. Smith, J. D. Suppl. 26-3, 1245 {1987).
Thompson, and J. O. Willis, Science 239, 33 (1988). M. Kato and K. Machida, Phys. Rev. B 37, 1510 (1988).
C. M. Varma, Physica 148B, 17 (1987). K. Machida and M. Kato, Jpn. J. Appl. Phys. 1, 277 (1988).
4K. Miyake, J. Magn. Magn. Mater. 63dk64, 411 (1987). 6K. Machida and M. Ozaki, Physica C 157, 53 (1989).
5G. Aeppli, E. Bucher, C. Broholm, J. K. Kjems, J. Baumann, '"G. E. Volovik and L. P. Gor'kov, Zh. Eksp. Teor. Fiz. 88,
and J. Hufnagl, Phys. Rev. Lett. 60, 615 (1988). 1412 (1985) [Sov. Phys.— JETP 61, 843 (1985)].
(a) Y. J. Uemura, W. J. Kossler, X. H. Yu, H. E. Schone, J. R. K. Ueda and T. M. Rice, Phys. Rev. B 31, 7114 (1985).
Kempton, C. E. Stronach, S. Barth, F. N. Gygax, B. Hitti, A. i E. I.
Blount, Phys. Rev. B 32, 2935 (1985).
Schenck, C. Baines, W. F. Lank ford, Y. Onuki, and T. M. Ozaki, K. Machida, and T. Ohmi, Prog. Theor. Phys. 74,
Komatsubara (unpublished); (b) C. Broholm, J. K. Kjems, W. 221 (1985); 75, 442 (1986).
J. L. Buyers, P. Matthews, T. T. Palstra, A. A. Menovsky, M. Ozaki, J. Math. Phys. 26, 1514 (1985); 26, 1521 (1985); and
and J. A. Mydosh. , Phys. Rev. Lett. 58, 1467 (1988). also see Prog. Theor. Phys. 62, 1183 (1979); 67, 83 (1982); 67,
R. H. HeA'ner, D. W. Cooke, and D. E. MacLaughlin, in 415 (1982).
Theoretical and Experimental Aspects of Valence Flucutations C. J. Bradley and A. P. Cracknell, The Mathematical Theory
and Heavy Fermions, edited by L. C. Gupta and S. K. Malik of Symmetry in Solids (Clarendon, Oxford, 1972).
(Plenum, New York, 1987). Y. Kohori, T. Kohara, H. Shibai, Y. Oda, T. Kaneko, Y. Ki-
8S. Barth, H. R. Ott, F. N. Gygax, B. Hitti, E. Lippelt, A. taoka, and K. Asayama, J. Phys. Soc. Jpn. 56, 2263 (1987); 57,
Schenck, C. Baines, B. van den Brandt, T. Konter, and S. 395 {1988).
Mango, Phys. Rev. Lett. 59, 2991 (1987) ~
"D. J. Scalapino, E. Loh, Jr. , and J. E. Hirsch, Phys. Rev. B 34,
H. Nakamura, Y. Kitaoka, K. Asayama, and J. Flouquet, The 8190 (1986).
Annual Meeting of Japan Physical Society at Koriyama, K. Miyake, S. Schmitt-Rink„and C. M. Varma, Phys. Rev. B
April, 1988. 34, 6554 (1986), and also see M. T. Beal-Monod, C. Bourbon-
ioA. K. Gangopadhyay, J. S. Schilbling, E. Schubert, P. nais, and V. J. Emery, Phys. Rev. B 34, 7716 (1986).
39 SUPERCONDUCTING CLASSES OF HEAVY-FERMION MATERIALS 4163
G. Aeppli, A. Goldman, G. Shirane, E. Bucher, and M. -Ch. 56H. A. Mook, B. D. Gaulin, G. Aeppli, Z. Fisk, and J. L.
Lux-Steiner, Phys. Rev. Lett. 58, 808 (1987)
~
Smith, Bull. Amer. Phys. Soc. 32, 594 {1987).
G. E. Brodale, R. A. Fisher, N. E. Phillips, G. R. Stewart, and 57K. U. Neumann, H. Cappelman, Z. Fisk, J. L. Smith, and K.
A. L. Giorgi, Phys. Rev. Lett. 57, 234 (1986). R. A. Ziebeck, Solid State Commun. 60, 641 (1986). They
The previous muon experiment already suggests a magnetic claim antiferromangetic Auctuations at different positions
phenomenon in UPt3. D. W. Cooke, R. H. Heffner, R. L. along [111] direction in reciprocal space, but their result
Hutson, M. E. Schillaci, J. L. Smith, J. O. Willis, D. E. Mac- seems less accurate. [See Goldman's article (Ref. 1) on this
Laughlin, C. Boekema, R. L. Lichti, A. B. Denison, and J. point. ]
Oostens, Hyperfine Interact. 31, 425 (1986). M. R. Norman, W. E. Pickett, H. Krakauer, and C. S. Wang,
A. I. Goldman, G. Shirane, G. Aeppli, B. Batlogg, and E. Phys. Rev. B 36, 4058 (1987).
Bucher, Phys. Rev. B 34, 6564 (1986); A. P. Ramirez, B. M. B. Maple (unpublished).
Batlogg, E. Bucher, and A. S. Cooper, Phys. Rev. Lett. 57, T. Kohara, Y. Kohori, K. Asayama, Y. Kitaoka, M. B. Ma-
1072 (1986). ple, and M. S. Torikachvili, Solid State Commun. 59, 603
3 P. Frings, B. Renker, and C. Vettier, J. Magn. Magn. Mater.
(1986).
63464, 202 (1987). T. T. M. Palstra, A. A. Menovsky, J. van den Berg, A. J. Dirk-
J. J. M. Franse, K. Kadowaki, A. A. Menovsky, M. van maat, P. H. Kes, G. J. Nieuwenhuys, and J. A. Mydosh,
Sprang, and A. de Visser, J. Appl. Phys. 61, 3380 (1987). Phys. Rev. Lett. 55, 2727 (1985).
T. Oguchi, A. J. Freeman, and G. W. Crabtree, Phys. Lett. A 6~M. B. Maple, J. W. Chen, Y. Dalichaouch, T. Kohara, C.
117, 428 (1986). Rossel, M. S. Torikachvili, M. W. McElfresh, and J. D.
C. S. Wang, M. R. Norman, R. C. Albers, A. M. Boring, W. Thompson, Phys, Rev. Lett. 56, 185 (1986).
E. Pickett, H. Krakauer, and N. E. Christensen, Phys. Rev. B D. E. MacLaughlin, D. W. Cooke, R. H. Heffner, R. L. Hut-
35, 7260 (1987). son, M. W. McElfresh, M. E. Schillaci, C. Boekema, R. L.
3~M. R. Norman, R. C. Albers, A. M. Boring, and N. E. Lichti, and J. Oostens, Phys. Rev. B 37, 3153 (1988).
Christensen, Solid State Commun. 68, 245 (1988). T. Kohara, Y. Kohori, K. Asayama, Y. Kitaoka, M. B. Ma-
L. Taillefer, R. Newbury, G. G. Lonzarich, Z. Fisk, and J. L. ple, and S. Torikachvili, Jpn. J. Appl. Phys. 26, Suppl. 26-3,
Smith, J. Magn. Magn. Mater. 63&64, 372 (1987). 1247 (1987).
L. Taillefer and G. G. Lonzarich, Phys. Rev. Lett. 60, 1570 65T. Fukase, Y. Koike, T. Nakanomyo, Y. Shiokawa, A. A.
(1988). Menovsky, J. A. Mydosh, and P. H. Kes, Jpn. J. Appl. Phys.
B. Batlogg (private communication). 26, Suppl. 26-3 1249 (1987).
3sK. Miyake, in Theory of Heavy Fermions and Valence Fluctua Y. Kitaoka, K. Ueda, Y. Kohori, T. Kohara and K. Asayama,
tions, edited by T. Kasuya and T. Saso (Springer-Verlag, New Jpn. J. Appl. Phys. (to be published).
York, 1985), p. 256. M. R. Norman (private commuriication).
M. Kato and K. Machida {unpublished). 8Y. Kitaoka and K. Asayama (private communication).
4oF. C. Zhang and T. K. Lee, Phys. Rev. B 35, 3651 (1987). Y. Kitaoka, K. Ueda, K. Kohori, K. Asayama, Y. Onuki, and
W. Putikka and R. Joynt, Phys. Rev. B 37, 2372 (1988). T. Karnatsubara, J. Magn. Magn. Mater. 52, 341 (1985).
4~M. R. Norman, Phys. Rev. Lett. 59, 232 (1987); Phys. Rev. B 7oG. Spam, W. Lieke, U. Gottwick, F. Steglich, and N. Grewe,
37, 4987 (1988). J. Magn. Magn. Mater. 47448, 521 (1985).
43C. J. Pethick, D. Pines, K. F. Quader, K. S. Bedall, and G. E. K. Ueda, Y. Kitaoka, H. Yamada, Y. Kohori, T. Kohara, and
Brown, Phys. Rev. Lett. 57, 1955 (1986). K. Asayama, J. Phys. Soc. Jpn. 56, 867 (1987).
44E. Bucher, J. P. Maita, G. W. Hull, R. C. Fukton, and A. S. There exists a band calculation on CeCu&Si~: J. Sticht, N.
Cooper, Phys. Rev. B 11, 440 (1975). d'Ambrumenil, and J. Kubler, Z. Phys. B 65, 149 (1986).
45R. H. Heffner, D. W. Cooke, Z. Fisk, R. L. Hutson, M. E. However, it is hard to see the FS topology from their calcula-
Schillaci, J. L. Smith, J. O. Willis, D. E. MacLaughlin, C. tion.
Boekema, R. L. Lichti, A. B. Denison, and J. Oostens, Phys. V. Muller, Ch. Roth, D. Mauer, E. W. Scheidt, K. Luders, E.
Rev. Lett. 57, 1255 (1986). Bucher, and H. E. Bornrnel, Phys. Rev. Lett. 58, 1224 (1987).
D. E. MacLaughlin, C. Tien, W. G. Clark, M. D. Lan, Z. Fisk, 74Y. J. Qian, M. -F. Xu, A. Schenstrom, H. -P. Baum, J. B. Ket-
J. L. Smith, and H. R. Ott, Phys. Rev. Lett. 53, 1853 (1984). terson, D. Hinks, M. Levy, and B. K. Sarma, Solid State
47R. Joynt, T. M. Rice, and K. Ueda, Phys. Rev. Lett. 56, 1412 Commun. 63, 599 {1987).
(1986). 75G. E. Volovik, J. Phys. C 21, L221 (1988).
G. E. Volovik and D. E. Khmel'nitskii, Pisma Zh. Eksp. Teor. K. Machida, K. Nokura, and T. Matsubara, Phys. Rev. Lett.
Fiz. 40, 469 (1984) [JETP Lett. 40, 1299 (1984)]. 44, 821 (1980).
L. P. Gor'kov and P. A. Kalugin, Pisma Zh. Eksp. Teor. Fiz. This point is also stressed by several authors in connection
41, 208 (1985) [JETP Lett. 41, 253 (1985)]. with high-T, oxide superconductors: K. Machida and M.
V. V. Moshchalkov and K. Svozil, Phys. Lett. A 120, 356 Kato, Jpn. J. Appl. Phys. 26, Suppl. 26-3, 1150 (1987); M.
(1987). Sigrist and T. M. Rice, Z. Phys. B 68, 9 (1987); J. F. Annett,
5'U. Rauchschwalbe, C. D. Bredel, F. Steglich, K. Maki, and P. M. Randeria, and S. R. Renn, Phys. Rev. B 38, 4660 (1988);
Fulde, Europhys. Lett. 3, 757 (1987). M. Ozaki and K. Machida, J. Phys. Soc. Jpn. 57, 398 {1988);
U. Rauchschwalbe, F. Steglich, G. R. Stewart, A. L. Giorgi, M. Kato, Y. Suzmura, Y. Okabe, and K. Machida, J. Phys.
P. Fulde, and K. Maki, Europhys. Lett. 3, 751 (1987). Soc. Jpn. 57, 726 (1988).
P. Kumar and P. Wolfe, Phys. Rev. Lett. 59, 1954 (1987). D. Einzel, P. J. Hirschfeld, F. Gross, B. S. Chandrasekhar, K.
54I. A. Lukyanchuk and V. P. Mineev, Pisma Zh. Eksp. Teor. Andres, H. R. Ott, J. Beuers, Z. Fisk, and J. L. Smith, Phys.
Fiz. 47, 460 (1988) [JETP Lett. 47, 543 (1988)]. Rev. Lett. 56, 2513 (1986).
5~M. Sigrist and T. M. Rice (unpublished).