Professional Documents
Culture Documents
JPSJ 79 114709
JPSJ 79 114709
FULL PAPER
KEYWORDS: large Fermi surface, Dingle temperature, metamagnetism, Cex La1 x Ru2 Si2 , coherent potential
approximation (CPA), continuous-time quantum Monte Carlo (CT-QMC)
DOI: 10.1143/JPSJ.79.114709
A randomly disordered alloy can be dealt with by the as TK 0:1. With these parameters, the antiferromag-
coherent potential approximation (CPA).23,24) On the other netic ordering is suppressed by formation of the Kondo
hand, strong local correlation can be taken into account by singlet.28)
the dynamical mean-field theory (DMFT),25) which can be
combined with the CPA.10–19) We shall give a brief review of 3. Effect of Magnetic Field at x ¼ 1
this framework for application to the Kondo lattice model. We begin with the pure system (Ce end) under magnetic
We assume, for the electronic state, the translational field. Figure 1(a) shows the magnetization process. We
symmetry of the lattice. This assumption can be realized by define H1 and H2 as lower and upper fields of the
taking an average over the random configuration of two ions. magnetization plateau, respectively. The appearance of the
Within a local approximation for the self-energy, the single- magnetization plateau is explained in a quasiparticle band
particle Green function for conduction electrons is given by picture, which also gives an account of the change of the
FS topology.21,22) In the heavy-fermion system, an almost
G ðk; zÞ ¼ ½z k þ ðzÞ1 : ð2Þ
dispersionless quasiparticle band is located close to the
The self-energy ðzÞ incorporates both the coherent Fermi level. Mixing with a conduction band results in a
potential and the many-body effect due to local interactions. hybridization gap, which is of the order of TK in the Kondo
We find an optimal self-energy from the solution of a single- lattice.29) Under the magnetic field, one of spin components
impurity problem in an effective medium. We define a cavity is filled up to the energy gap. The magnetization does not
Green function G0 ðzÞ by change in this region, namely, H2 H1 TK .
Let us observe the change of the FS topology. In the
G0 ðzÞ1 ¼ G ðzÞ1 þ ðzÞ; ð3Þ
P DMFT, k-dependence of the Green function enters only
where G ðzÞ ¼ N 1 k Gðk; zÞ. We omit the argument z and through k as shown in eq. (2). Hence, we regard ¼ k
index hereafter for simplicity. The local Green function G as ‘‘momentum’’ hereafter. The Fermi momentum can
with ¼ A, B is defined for the local moment site A and the be traced by the renormalized chemical potential
non-magnetic site B. It is given by Re ð0Þ, since the FS of each spin component appears
at the momentum where ¼ Re ð0Þ is satisfied.29)
G ¼ G0 þ G0 t G0 ; ð4Þ
Figure 1(b) shows Re ð0Þ as a function of H. There
where t is the single-site t-matrix, and the interaction has are three regions, and correspondingly we introduce three
been assumed to be local. ‘‘momenta’’: L , S and sc , which give a FS involving
In the CPA, the cavity Green function G0 is determined so ðnc þ 1Þ=2, nc =2, and nc electrons, respectively.
that the local Green function, hGi, averaged over two kinds Region I (0 H < H1 ): Without magnetic field,
of sites is identical to that of the whole lattice, G : Re ð0Þ corresponds to L and the FS involves localized
spins. Upon turning on the magnetic field, the FS splits, but
G ¼ hGi ¼ xGA þ ð1 xÞGB : ð5Þ
the average keeps at L .
This CPA condition can be written in a form which does not Region II (H1 < H < H2 ): At H1 ’ 0:006, one band
include G0 . Using eq. (3), eq. (4) is rewritten in terms of G passes the band edge c ¼ 1 indicating that the chemical
as potential is located in the energy gap. Then, only the other
spin forms the FS, which has the momentum sc . Namely,
G ¼ G þ G t~ G ; ð6Þ
the FS of a single spin-component involves nc electrons in
where this region, and the magnetization shows the plateau.
ð Þ Region III (H > H2 ): By further increasing the magnetic
t~ ¼ ; ð7Þ field, both spin components again contribute to the FS.
1 G ð Þ
For H > H2 ’ 0:126, the spin average of Re ð0Þ is
with 1 1
¼ t þ G0 . The CPA condition (5) is then written located at S , indicating the small FS composed of both spin-
as components.
We note that even in Region III, the quasiparticle involves
ht~i ¼ xt~A þ ð1 xÞt~B ¼ 0; ð8Þ
the localized spins near H2 , since the energy gap cannot be
which leads to formed without coupling with localized spins in the Kondo
hti lattice. Namely, the ‘‘small FS’’ does not mean the localized
¼ ; hti ¼ xtA þ ð1 xÞtB : ð9Þ character of 4f electrons.22) The energy gap gradually
1 þ G0 hti
vanishes with decreasing 4f-contribution to the quasiparticle
This equation gives the self-energy in the CPA. For x ¼ 1, for H H2 TK .
the equation is reduced to that in the pure Kondo lattice Transition of the FS topology is known as Lifshitz
model.26) The effective medium G0 is determined self- transition, which is distinguished from thermodynamic
consistently from eqs. (2), (3), and (9). phase transition.30,31) The transitions between Regions I
We solve the effective impurity problem using the and II, or II and III in the Kondo lattice are identified as the
continuous-time quantum Monte Carlo (CT-QMC) method Lifshitz transition in the sense that there is no discontinuity
27)
adapted to the Kondo model. We use ffi a semi-circle density
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi in physical quantities at finite temperature. However, as
of states ðÞ ¼ ð2=DÞ 1 ð=DÞ2 with the band cutoff important characteristics in the Kondo lattice, the transition
D ¼ 1, and use the following numerical parameters: nc ¼ is caused by quasiparticles, which incorporate strong local
0:9 for the number nc of conduction electrons per site, correlation. This difference appears in the renormalization
and J ¼ 0:34. The resultant Kondo temperature is estimated factor z defined by
114709-2
Person-to-person distribution by the author only. Not permitted for publication for institutional repositories or on personal Web sites.
(a) x
H1
f
H2 κS κL
0.8 0.8
c H=0
total
nc(κ)
0.6 0.9
1
1
M
0.4 0.8
-0.2 0.2 0.6 1
0.6 κ
0.2
0.4 κS
κL
0 0.2 0
0.2
0 0.4
0 0.05 0.1 0.15 0.2 -1 -0.5 0.6 x
0 0.8
H κ 0.5 1 1
(b)
1 κc Fig. 2. (Color online) Momentum distribution for H ¼ 0. The temper-
κL ature is taken as T ¼ 0:001 for 0:9 x 1 and T ¼ 0:0025 for
κ sc 0 x 0:88, which are low enough to neglect the broadening by finite
0.5 majority spin
minority spin temperature in the scale of the figure. The (red) squares on the bottom and
μσ – ReΣ σ (0)
in the inset show the momenta where dnc ðÞ=d takes a local maximum.
0 κS
μ
-0.5 in the FS transition, becomes lighter on the plateau region.
The effective mass strongly depends on the spin component
-1 –κ c on both sides of the transition, while the spin dependence is
negligible for H H2 . This enhancement has been dis-
0 0.05 0.1 0.15 0.2 cussed in the high-field region.32–34)
H
4. Effect of Disorder
(c) 0.6
Now let us examine the effect of disorder on the FS. We
0.5 discuss the momentum distribution of conduction electrons
majority spin X
0.4 minority spin hcyk ck i ¼ T G ð; in Þein 0þ
nc ðÞ: ð11Þ
n
zσ
0.3
Figure 2 shows ncþ ðÞ ¼ nc ðÞ
nc ðÞ for H ¼ 0. Con-
0.2 cerning 0 x 0:9, the temperature is much lower than TK ,
and the results may be regarded as those in the ground state.
0.1 In the case of x ¼ 0, nc ðÞ has a discontinuity of unit
magnitude according to the Fermi–Dirac distribution func-
0 tion. The Fermi momentum is determined only by the
0 0.05 0.1 0.15 0.2
number of conduction electrons. At x ¼ 1, on the other hand,
H
a small discontinuity appears at the momentum where the
Fig. 1. (Color online) Magnetic field dependences of (a) magnetization Fermi volume includes localized spins.29) This large FS
M, (b) Fermi momentum ¼ Re ð0Þ, and (c) spin-dependent disappears at x ¼ 0:9 ¼ nc . The volume of the large FS
renormalization factor z . The Fermi level lies in the energy gap of the increases by removing localized spins toward x ¼ nc . This
majority spin band for H1 < H < H2 . Parameters are chosen as x ¼ 1,
J ¼ 0:34, nc ¼ 0:9, and T ¼ 0:0025.
trend is in contrast to a rigid band picture that the volume
decreases in proportion to the number of localized spins. It
can be understood by considering that x ¼ nc is a condition
" #1 for insulator. Namely, the Fermi momentum varies toward
@ Im ðin Þ c ¼ 1 at x ¼ nc .
z ¼ 1 ; ð10Þ For magnetic field of H1 < H . H2 , nc ðÞ depends on x
@ n n !þ0
in a similar manner as that in Fig. 2, except that the FS in the
inverse of which corresponds to the mass enhancement of Ce side appears only in the minority spin. On the other hand,
the quasiparticle. Since now the energy gap is due to the when H H2 , the FS can be connected continuously from
collective Kondo singlet, Im ð!Þ has a peak centered at x ¼ 0 to x ¼ 1. Figure 3 shows nc ðÞ for H ¼ 0:2, which
the energy gap as in the Mott insulator.25) This peak leads is much larger than H2 and hence much larger than TK .
to a steep slope in the real part Re ð!Þ. As a result, The FS’s of majority and minority spins split as the Ce
the renormalization factor takes a small value when the concentration increases. The size of each component is
quasiparticle level crosses the Fermi level, as shown in opposite to the corresponding number of local spins because
Fig. 1(c). Consequently, the effective mass for the majority of the effective field through the exchange interaction.
spin is enhanced toward the magnetic field where the FS We define an energy scale T that characterizes the
transition occurs, while the other spin, which is not involved scattering rate of the quasiparticles:29)
114709-3
Person-to-person distribution by the author only. Not permitted for publication for institutional repositories or on personal Web sites.
0.6
H = 0 TS
ncσ (κ) TL TL
H = 0.2 0.5 H = 0.15 T S
0.20 T S (T = 0.0025)
1 majority spin (T = 0.001)
0.4
0.8
0.6
Tγ
0.3 TS H=0
0.4 (T = 0.0025)
κL 0.2
0.2 0 H = 0.15
0.2 0.1
0 0.4
-1 -0.5 0.6 x
0 0.8 H = 0.2
κ 0.5 1 1
0
0 0.2 0.4 0.6 0.8 1
x
0.8 H2
κ 0.5 1 1 0.1
(II) Single-component FS
Fig. 3. (Color online) Momentum distribution nc ðÞ for H ¼ 0:2 and
T ¼ 0:0025. Upper and lower panels correspond to majority and minority 0.05
spins, respectively. The (red) squares on the bottom show the momentum
where dnc ðÞ=d takes the maximum. (I) Large FS H1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x
1 dnc ðÞ 1
T ¼ ; ð12Þ
4 d FSð Þ
Fig. 5. (Color online) The FS topology in x–H plane for J ¼ 0:34 and
nc ¼ 0:9. The open square (red line) represents H of metamagnetic
where ¼ S, L, and the subscript FS( ) indicates that the behavior. The closed square (blue line) represents x above which the
derivative is evaluated at the corresponding FS. At x ¼ 0, indication of the quasiparticle is seen in nc ðÞ. The dashed line and circle
the FS gets blurred according to TS ¼ T, since nc ðÞ is symbol represent x where TS takes the maximum. The open (red) and
given by the Fermi–Dirac distribution function. On the other closed (blue) symbols signify anomaly due to majority and minority
spins, respectively.
hand, the FS at x ¼ 1 depends more strongly on temperature:
TL ¼ T=z2 .29) Except for x ¼ 0 and x ¼ 1, T is finite even
at T ¼ 0 because of disorder. In this case, T can be
regarded as the Dingle temperature. whether H is larger or smaller than H2 . When H . H2 , the
Figure 4 shows T as a function of x for several values of heavy-fermion FS of one spin component appears around
H. At H ¼ 0, TS monotonously increases as x increases. x ¼ nc upon increasing x. In addition, if H < H1 , the other
Namely, the La FS is not connected to the Ce side, although spin component shows up for larger x. In the special case
it leaves a trace of the ‘‘small FS’’. Instead, the ‘‘large FS’’ of H ¼ 0, both spin components appear simultaneously at
appears at x nc , appearance of which is indicated by rapid x ¼ nc , at which the system becomes insulating. In this
decrease of TL . For H < H2 , the x-dependence of T is magnetic-field region, H . H2 , the Dingle temperature of the
similar to that of H ¼ 0. Note that TL is defined only for the La FS monotonously increases against x. On the other hand,
minority spin in the range H1 < H < H2 . On the other hand, when H H2 , the La FS is connected continuously to the Ce
when H > H2 , TS becomes zero at both x ¼ 0 and x ¼ 1 side. In this case, the Dingle temperature has a maximum at a
having a maximum at 0:5 . x . 0:8. Namely, the FS in the certain value of x. As shown in Fig. 5, the maximum point
La side remains a true FS up to the Ce end. tends to converge at x ¼ 0:5 in high field, where the system
can be regarded as a nonmagnetic binary alloy.
5. Summary and Discussions In realistic situations, the chemical potential is not in the
We summarize the above results in Fig. 5. The Ce side hybridization gap, because of an energy dispersion of 4f
is classified into three regions according to the FS topology: level and/or existence of plural conduction bands on the
(I) large FS, (II) single-component FS, and (III) ‘‘small’’ FS. Fermi level. In this case, a situation may occur where the
This transition is caused by the quasiparticle band, whose magnetization-plateau region disappears in Fig. 5. Then, we
energy gap H2 H1 is of the order of TK . The effect of expect a metamagnetic behavior without the plateau around
disorder on the FS is qualitatively different depending on H1 ’ H2 TK .
114709-4
Person-to-person distribution by the author only. Not permitted for publication for institutional repositories or on personal Web sites.
We comment on another situation where nc is away from 5) R. Settai, T. Kubo, T. Shiromoto, D. Honda, H. Shishido, K.
half filling. For nc 1, the coherence temperature T Sugiyama, Y. Haga, T. D. Matsuda, K. Betsuyaku, H. Harima, T. C.
Kobayashi, and Y. Ōnuki: J. Phys. Soc. Jpn. 74 (2005) 3016.
becomes much smaller than the single-impurity Kondo 6) Y. Matsumoto, M. Sugi, N. Kimura, T. Komatsubara, H. Aoki, I.
temperature TK , while T ’ TK at nc ’ 1.35) The feature Satoh, T. Terashima, and S. Uji: J. Phys. Soc. Jpn. 77 (2008) 053703.
caused by the large FS such as regions (I) and (II) appears at 7) R. G. Goodrich, N. Harrison, A. Teklu, D. Young, and Z. Fisk: Phys.
T T . The indication of the large FS, indicated by the Rev. Lett. 82 (1999) 3669.
8) M. Endo, S. Nakamura, T. Isshiki, N. Kimura, T. Nojima, H. Aoki, H.
solid lines in Fig. 5, would be obscure at T & T , although
Harima, and S. Kunii: J. Phys. Soc. Jpn. 75 (2006) 114704.
the properties at T ¼ 0 are the same as in Fig. 5. 9) N. Harrison, U. Alver, R. G. Goodrich, I. Vekhter, J. L. Sarrao, P. G.
We finally discuss realistic materials such as Pagliuso, N. O. Moreno, L. Balicas, Z. Fisk, D. Hall, R. T. Macaluso,
Cex La1x Ru2 Si2 . We consider that the FS transition in pure and J. Y. Chan: Phys. Rev. Lett. 93 (2004) 186405.
CeRu2 Si2 is the Lifshitz transition caused by the quasipar- 10) A. Yoshimori and H. Kasai: Solid State Commun. 58 (1986) 259.
ticle band.22) Actually, an analysis of transport coefficients 11) Z. Li and Y. Qiu: Phys. Rev. B 43 (1991) 12906.
12) P. Schlottmann: Phys. Rev. B 46 (1992) 998.
indicates disappearance of a FS with one of spin components 13) R. Shiina: J. Phys. Soc. Jpn. 64 (1995) 702.
under magnetic field.36) Our numerical calculation demon- 14) S. Wermbter, K. Sabel, and G. Czycholl: Phys. Rev. B 53 (1996) 2528.
strates that this FS transition gives rise to a metamagnetic 15) E. Miranda, V. Dobrosavljevic, and G. Kotliar: Phys. Rev. Lett. 78
transition as shown in Fig. 1(a). In considering substitution (1997) 290.
effect in the multi-band system, the Dingle temperature is 16) T. Mutou: Phys. Rev. B 64 (2001) 165103; T. Mutou: Phys. Rev. B 64
(2001) 245102.
more informative rather than the dHvA frequencies, since 17) S. Burdin and P. Fulde: Phys. Rev. B 76 (2007) 104425.
each frequency may vary with keeping the total volume of 18) U. Yu, K. Byczuk, and D. Vollhardt: Phys. Rev. Lett. 100 (2008)
the FS. Recent experiments have revealed a contrastive 246401.
behavior depending on direction of the magnetic field: dHvA 19) C. Grenzebach, F. B. Anders, G. Czycholl, and T. Pruschke: Phys.
Rev. B 77 (2008) 115125.
signals vanish by the substitution on both La and Ce sides
20) H. Watanabe and M. Ogata: Phys. Rev. B 81 (2010) 113111.
for H ? c, while the dHvA amplitude Aosc shows a 21) S. Watanabe: J. Phys. Soc. Jpn. 69 (2000) 2947.
minimum at 0 < x < 1 for H k c.37) Because the magnet- 22) K. Miyake and H. Ikeda: J. Phys. Soc. Jpn. 75 (2006) 033704.
ization in this system shows a strong Ising anisotropy, we 23) F. Yonezawa and K. Morigaki: Prog. Theor. Phys. Suppl. 53 (1973) 1.
can regard the results for H ? c as low-field region, and the 24) R. J. Elliott, J. A. Krumhansl, and P. L. Leath: Rev. Mod. Phys. 46
results for H k c as high-field region. Then, the tendency is (1974) 465.
25) A. Georges, G. Kotliar, W. Krauth, and M. J. Rozenberg: Rev. Mod.
consistent with our results for the Dingle temperature T , Phys. 68 (1996) 13.
considering Aosc / expðT Þ. The degree of x-dependence 26) J. Otsuki, H. Kusunose, and Y. Kuramoto: J. Phys. Soc. Jpn. 78 (2009)
in the Dingle temperature indicates how much 4f electrons 014702.
contribute to each branch. 27) J. Otsuki, H. Kusunose, P. Werner, and Y. Kuramoto: J. Phys. Soc.
Jpn. 76 (2007) 114707.
Acknowledgments 28) R. Peters and T. Pruschke: Phys. Rev. B 76 (2007) 245101.
29) J. Otsuki, H. Kusunose, and Y. Kuramoto: Phys. Rev. Lett. 102 (2009)
We thank H. Aoki and Y. Matsumoto for useful 017202.
discussions. This work was supported by a Grant-in-Aid 30) I. M. Lifshitz: Sov. Phys. JETP 11 (1960) 1130.
for Scientific Research on Innovative Areas ‘‘Heavy Elec- 31) Y. M. Blanter, M. I. Kaganov, A. V. Pantsulaya, and A. A. Varlamov:
trons’’ (No. 20102008) from the Ministry of Education, Phys. Rep. 245 (1994) 159.
32) A. Wasserman, M. Springford, and A. C. Hewson: J. Phys.: Condens.
Culture, Sports, Science, and Technology, Japan. Matter 1 (1989) 2669.
33) D. M. Edwards and A. C. M. Green: Z. Phys. B 103 (1997) 243.
34) J. Otsuki, H. Kusunose, and Y. Kuramoto: J. Magn. Magn. Mater. 310
1) J. M. Luttinger: Phys. Rev. 119 (1960) 1153. (2007) 425.
2) T. Sakakibara, T. Tayama, K. Matsuhira, H. Mitamura, H. Amitsuka, 35) S. Burdin, A. Georges, and D. R. Grempel: Phys. Rev. Lett. 85 (2000)
K. Maezawa, and Y. Ōnuki: Phys. Rev. B 51 (1995) 12030. 1048.
3) H. Aoki, S. Uji, A. K. Albessard, and Y. Ōnuki: Phys. Rev. Lett. 71 36) R. Daou, C. Bergemann, and S. R. Julian: Phys. Rev. Lett. 96 (2006)
(1993) 2110. 026401.
4) H. Shishido, R. Settai, H. Harima, and Y. Ōnuki: J. Phys. Soc. Jpn. 74 37) Y. Matsumoto, N. Kimura, H. Aoki, M. Kimata, T. Terashima, S. Uji,
(2005) 1103. T. Okane, and H. Yamagami: J. Phys. Soc. Jpn. 79 (2010) 083706.
114709-5