PHYSICAL REVIEW
B
VOLUME 46, NUMBER 2 1JULY 1992-IIEPR
linewidth
of
local
magnetic
moments
diluted
in
Ce
intermediate-valence
compounds
Pablo
A.
VenegasDepartamento de Ftsica, Faculdade de Ciencias,
RESP,
Campus Bauru, R.Engenheiro Luis Coube s/n, Bauru, Sgo Paulo, BraziGaston
E.
BarberisInstituto de Fisiea, Universidade Estadual de Campinas (UNICAMP), 13081Campinas, Sito Paulo, Brazil (Received 23 December 1991)
Anomalous thermal behavior on the EPR linewidths has been observed for Gd impurities diluted in
Ce„Al
„B„(A
=La,
Y,B
=Ir,
Os,Rh,Pd) intermediate-valence compounds. In this work we show thatthe exchange interaction between the local magnetic moments and the intermediate-valence host ions has an important contribution to the relaxation rates ofthe local moments. We calculated the relaxa-tion, using the Redfield formalism and the ideas contained in the interconfigurational fluctuation model
ofHirst. We show that the exchange interaction contribution has an exponential dependence on the ex-citation energy ofthe intermediate-valence ions.
I.
INTRODUCTIONCe intermetallic compounds have been extensively studied in the literature and many
of
them show anomalies typicalof
intermediate-valence (IV) alloys.'By doping Ce compounds with Gd + ions, their
ESR
spectra allow us to study locally the influence
of
the intermediate-valence Ce ions. In metals, the local mo-mentsESR
linewidth,hH,
isexpectedto
increase linearly with the temperature (Korringa ). In contrast, inCe„A,
„Pd3
(A=Ag,
Y)and Ce„La& „Oszthe GdESR
spectra show a nonlinear increase
of
hH
in the tempera-ture range4.2&
T&300
K.
At low temperatures the slope d(b,H)ldT
is smaller than in the isostructuralnon-IV compounds MPd3
(M=Sc,
Y,
La), whereas at high temperatures the resonance line is strongly broadened and the slope asymptotically approaches the value mea-sured in LaPd3. The temperature dependenceof
hH
be-comes nonlinear and cannot be explained only by invok-ing the usual Korringa mechanism.EPR
studies in metallic hosts, which show interconfigurational fluctuations, suggest that thein-direct exchange interaction between the local magnetic impurities and the Ce
4f
electrons have an appreciable contribution to the linewidthof
the magnetic impurities.The Ce interconfigurational fluctuations are transferred via the Ruderman-Kittel-Kasuya-Yoshida
(RKKY)
in-teraction to the Gd site. These fluctuations generate an effective alternating magnetic field, which relaxes the Gd
spins. A similar mechanism was also used for the inter-pretation
of
the NMR relaxation ratesof
noble metals with Kondo impurities, for perturbed angularcorrela-tion measurements
of
' Cein (La&„Y„)A12
alloys, ' andfor Gd
ESR
on concentrated Van Vleck paramagnets."
We use this mechanism to explain the anomalous
behav-ior
of
the linewidthof
Gd diluted in the Ceintermediate valence compounds.The Gd linewidth can be calculated with use
of
the ideas contained in the "interconfigurational fluctuation"model
of
Hirst.''
This model has been used toexplainvarious physical measurements and particularly magnetic susceptibility and Mossbauer studies in intermediate-valence compounds. In their approach the expectation
value
of
the measured physical quantity (isomer shift, magnetic moment,etc.
)iscalculated by "averaging" overall possible
4f"
configurations. The occupation probabil-ityof
the n configuration is usually expressed asP„(T)
~
exp(E„/T„)
—
withT„=T+b„.
"
'
HereE„
and b,
„are
the energy and widthof
the4f"
configuration. The quantity measured thus depends ex-ponentially on the excitation energy,E,„,
defined as the energy required to delocalize a single electron from the4f"
configuration (i.e.
,E,
„=E„E„&).
Al—though thistheoretical approach is phenomenological, it successfully explains most
of
the Europium Mossbauer spectra and susceptibility studies in mixed-valence compounds. ''
In this work we use this approach and the Redfield
for-malism' to calculate the contribution
of
an excitedconfiguration
of
the host Ce ionsto
the linewidthof
the impurity spectra. We fit the experimental data andcom-pare with other theoretical approaches.
'
II.
THE MODELThe spin Hamiltonian describing the Ce ions in the n
configuration can be written
H=H, f+H,
,where the crystal-field Hamiltonian for Ceis
H,
t=
g
[8404(
J'")+8606(J'")
] Iwith
04(J'")
and06(J'")
the cubic combinationsof
Stevens' equivalent operatorsof
the fourth and sixth de-gree, respectively,84
and86
are the crystalline-field pa-rametersof
the fourth and sixth degree, respectively, andJ'
'isthe total angular momentumof
the1th Ce ion. The
Zeeman Hamiltonian for Ceis
PABLO A. VENEGAS AND GASTON
E.
BARBERIS 46where
pz
is the Bohr magneton, gz is the Lande factor, and H isthe external magnetic field.The Gd Hamiltonian can be written as
H'
=H,
'f+H,
'+H„.
(4)III.
THELINEWIDTHIn our metallic host the impurity
ESR
linewidthcalcu-lations must consider the energy transfer between the im-purity spin
S,
the spinof
the host rare-earth ions (Ce), the spinsof
the conduction electrons as well as the lattice, in the presenceof
static external magnetic field, and asmall alterating field. In the present work we shall assume thatthe system is in the unbottlenecked regime,
i.
e.,the con-duction electrons and the host magnetic ions are in equi-librium with the lattice. In addition we shall assume thatthe transverse susceptibility associated with the impurity
at the impurity resonance frequency, coo, is much larger
than those
of
the other spin systems at the samefrequen-cy. In this limit one can consider only the transverse magnetization,
M,
of
the impurities and neglect thein-teraction
of
therf
field with the magnetic host ions and the conduction electrons. In other words, the magnetic host ions and the conduction electrons are "passive dissi-pative systems."
On the other hand, the static magnetic field splits the Gd resonance in2S
=7
fine-structure lines, which are well resolved at low temperatures. In the present case, forsimplicity, we assume that the finestruc-ture is collapsed and suppose the Gd with an effective spin
S
=
—,'.
With these assumptions the impuritylinewidth can be expressed as
EH=a+bT+EIF
(6)where a isthe residual linewidth, b isthe usual Korringa
contribution originating from the impurity-conduction-electron exchange interaction, and EIF is the impurity linewidth due to the exchange interaction with the Ce
host ions. The last contribution arises from the
RKKY
coupling between the Ceand Gd ions, which transfers the
Cefluctuations tothe Gd site.
The
ESR
relaxationof
an impurity due to (stable)ex-cited configurations
of
the host ions can be calculated us-ing the Bloch-Redfield-Wangsness formalism, ' in analo-gy with impurity relaxation in Van-Vleck compounds."
Then the relaxation rate can be written asHere
H,
'f
andH,
' are defined in a similar form asH,
f andH„with
the obvious substitutionof
J'"
by the spinof
the Gd,S.
The Gd impurity interacts with the neighbor-ing host Ceions via exchange interaction:Zp
H,
„=
—
(g—
1)8
g
SJ'",
1=1
where
8
is the exchange intergal between the two ions and zo is the numberof
Ceneighbor ionsof
eachimpuri-ty.
For
simplicity, however, we have considered only the first Ceneighbors. We calculate the relaxation rateof
the impurityif
it does not create a perturbationof
the Cehost Hamiltonian.
bH=
[()((gJ—
1)] [k
(coo)+k,
(coo)+2k„(0)],
(7) 2A'where co& is the impurity resonance frequency.
kqq(co)(q
=x,
y,z) isthe Fourier transformof
the spectral functions defined ask
=zoK
(co) with Kqq(co) defined asKqq(cu)=
f
(5J
(t)5J )e
dt(10)
In the present case, the Ce interconfigurational fluctua-tion frequency, Ace, is much greater than the impurity resonance frequency, coo. Physically for our model it means that for a very short correlation time, the spectral density
of
the fluctuating field is"white"
to the frequen-cies far above the resonance frequency. So that if we view the problem from a rotating frame (that is one where the magnetization is rotating at the Larmor fre-quency), thex,
y, and z directions are equivalent and the effectsof
the external magnetic field in the fluctuationspectra can be neglected. In this casewe have that
K,
„(a)0)
=—K„„(0);
K
(coo)—
=
K (0)
and for cubic symmetryK„„(0)
=
Ky(0)
=K„(0)
.
(12)
With these assumptions the calculation
of
spectral functions (seeRef.
11)leads to the following expressionforthe linewidth:
,
[~(gJ
—
1)]'zoF(o)
y
p.
I(«I
J,
I«&
I',
(14) $2where
Ina)
andE„are
the eigenvectors and eigenvaluesof
H,F(0)
a frequency distribution function due to the finite widthof
the Ce4f"
states, andp„
=
exp(E„
/T)/Z,
withZ
—
=
g„,
exp(E„,
/T).
—
Ce ions ffuctuate between the
4f
and the4f
' configurations.If
we assume the4f
configuration as the ground state (withJ,
=0),
the contributionof
the excited4f
configuration to the impurity linewidth can be writ-ten as—F. /T
ex
IF where
2
=
(2~/fi
)[8(
g1—
1)]zoF (0)
I(1a
IJ,
I1a
&I0 0
k
(~)=
f
g
5J'"'(t)
g
5J'"'
e'
'dt.
qq 2 q q
k=1 k=1
Here
( )
denotes a thermal average and 5Jq")the ffuctua-tion in the q componentof
the total angular momentumof
Cedue to the interconfigurational fluctuations, defined as5J(k)
—
J(k)(
J(k))
q q q
In the absence
of
pair correlations,i.
e., terms likeEPRLINE%'IDTH OFLOCAL MAGNETIC MOMENTS
DILUTED.
.
.
913and
E,
„
is the excitation energy required to delocalize asingle electron from the Ce
4f
'configuration.I800
IV. RESULTSAND DISCUSSIQN I400 Using the expression obtained in
Eq.
(15)we cancalcu-late the total linewidth
of
the Gd resonance, given byEq.
(7). Figures 1 and 2 plot the calculated linewidth as a function
of
the temperature for selected casesof
Ce La) „Osz and
Ce„Y)
„Pd3
powered samples. Theresidual linewidth, which is sample dependent, is not
con-sidered in the plots (see figure captions).
To
fit the datawe adjust the parameters b, A, and
E,„,
where b represents the"apparent"
Korringa parameter. This is done so because we are not considering the exchange nar-rowing effectof
the fine structureof
Gd at lowT.
' As stated previously, here we assume an effective spinS=
—,'for the magnetic ion. Rigorously, the linewidth calcula-tions have to consider the contribution
of
the finestruc-ture associated with the spin
S
=
—',of Gd.
However, the exchange narrowing effect collapses the spectra in a sin-gle line, which can be treated by anS
=
—,' effective spin.This collapse, at low T,ismore evident forbig values
of
b and the resonance iswell described by this efFective spin.For
the cases where b is small and the fine-structurecon-tribution, not considered here, is important we use an ap-parent Korringa parameter, associated to the
S
=
—,'effective spin. An example
of
the last case isthe Gd di-luted in CePd3 resonance. Comparing our theoreticalre-sults with the experimental data we can see that in most
of
the cases we got an excellent fit. The same calculations have been made for Ce(Pd&„Ag„)3
and Ce(Pd&„Rh„)3
with similar results. According to our model, at low
T
the main contribution tothe linewidth isoriginated in the usual Korringa mechanism. At high temperature the populationof
the excited Ce4f
configuration isin-500
—IOOO CI IJJ
600
200
50 I00 150 200
TEMPERATURE (K)
250 300 350
FIG.
2. Same asFig.1forCe„Y& „Pd3.creased, and the exponential contribution given by hi+ is the most important. The parameters used to fit the ex-perimental data are shown in Table
I.
Looking at the values
of
the fitting parameters theeffect
of
the Ce concentration seems clear. The increase in the excitation energy value when the Ce concentrationisreduced agree with the interconfigurational fluctuation model.' ' The model predicts an increase
of
theexcita-tion energy when we cross from the intermediate valence
to the magnetic regime. This is the case, for example, when we go from CeOsz to LaOsz. On the other hand, the A value depends on the strength
of
the Cefluctuationspectra and we can expect an increase with the Ce con-centration. The small value
of
b obtained for high Ceconcentrations, at low T, agree with the prediction
of
the "hybridization hole" model for Gd diluted in the CePd3 powered sample and with that obtained by us ' for the monocrystalline spectraof
the sample compound. Thisresult agrees also with that obtained by Hirst, for low T, but not with the result obtained in
Ref.
19,which pre-dicts a higher value for b. However, the thermalbehav-ior
of
the linewidth obtained by the latter authors agree,at least qualitatively, with that obtained here.
300
X
I-C5
& 300
TABLE
I.
Obtained values ofthe parameters b, a, andE,
„
that fitEq.(6)for diluted Gd. The last column shows the exper-imental value ofE,
„published in Ref. 23, tobetaken as a com-parison.I00
b [G/IC] A [G]
E,
„(theor.
)E,
„(expt.
)50 I00 I50 200
TEMPERATURE (K)
250 350
FIG.
1. Temperature dependence of the linewidth for different values of the concentrationx
in Ce La& „Os2. Thefull lines are the plot ofthe theoretical expression [Eq. (6)].The
fitting parameters can be found in Table I(a). The residual
linewidths are not realisitic, for the data was shifted in order to avoid superimposed curves.
1
0.95
0.9
0.85
0.8
1
0.91
0.8
1.27 1.31
3.28
45
5.18
0.54
0.69
0.76
(a) Ce„La, „Os2 9547
9580 7364 3112
94
672 678 849 1132 1681
(b) Ce„Y, „Pd3
1543 576
1170 609
642 517
914 PABLO A. VENEGAS AND GASTON
E.
BARBERIS 46It
is important to observe that in contrast to the hy-bridization hole model, here we have supposed a constant densityof
states asin a normal metal. The nonlinearcon-tribution to the linewidth, according to the present mod-el, is originated in the exchange interaction
of
the mag-netic impurities with the Ce ions. Our calculations, re-garding the hybridization hole model, permit a quantita-tive descriptionof
the linewidth, but using a lesser num-berof
adjustable parameters. In the other hand, is neces-sary topoint out that for small valuesof x,
where the AIF value is small in comparison with the Korringacontribu-tion, the numerical simulation
of
AH leads to unexpected small valuesof
E,
„.
However, as we can clearly seein theCe
La,
„Os2 case, b asymptotically approaches theLaOs2 b value when
x
goes to0.
Note that the values obtained for the excitation energy
are, within the experimental error, close to that founded by Sereni, Olcese, and Rizzuto. This agreement with experimental results supports this interpretation.
ACKNOWLEDGMENTS
The authors want to acknowledge the International Center
of
Theoretical Physics, Trieste, Italy, for the hos-pitality received there, where partof
this work was per-formed. This work was partially supported by CNPQand
FAPESP,
Brazil.'J.
M.Robinson, Phys. Rep. 51,1(1979).J.
Korringa, Physica 16, 601 (1950).H.Schaeffer and
B.
Elschner, Z.Phys. 53,109(1983).4T. Gambke,
B.
Elschner,R.
Kremer, and M.Schanz,J.
Magn. Magn. Mater. 36,124(1983).5C.
Y.
Huang andK.
Sugawara,J.
Low-Temp. Phys. 28, 229 (1977).H.Alloul, Physica 86-88B,449(1977).
7D.
E.
MacLaughlin, O.Penna, and M.Lysak, Phys. Rev.B23, 1039(1981).A. Narath, in Magnetism, edited by V. H. Suhl (Academic,
New York, 1973), p. 150.
B.
Giovaninni, P.Pincus, G.Gladstone, and A.J.
Heeger,J.
Phys. (Paris) Colloq. 321, C1-163 (1971).M.Mekata, T.Uemura, and
T.
Kuriyama, Hyperfine Interact. 10,833(1981).D. Davidov, V. Zevin,
R.
Levin, D. Shaltiel, andK.
Ba-berschke, Phys. Rev.B15,2771(1977).L.L.Hirst, Phys. Kondens. Mater. 11,255(1970).
~3L.L.Hirst, in Magnetism and Magnetic Materials (San Fran
cisco, 1974),Proceedings ofthe 20th Annual Conference on
Magnetism and Magnetic Materials, AIP Conf. Proc. No.24, edited byC. D.Graham, G.H.Lander, and
J. J.
Rhyne (AIP,New York, 1975).
L.L.Hirst, in Valence Instabilities and Narrow Band
Phenom-ena, edited by
R.
D.Parks (Plenum, New York, 1977),p.3. 'sD.K.
Wohlleben, in Proceedings ofthe International Conference on Valence Fluctuations in Solids, edited by L.Falicov, W. Hanke, and M.
B.
Maple (North-Holland, Amsterdam, 1981).'
E.
R.Bauminger, D.Froindlich,I.
Nowik, S.Ofer,
I.
Feiner,and
I.
Mayer, Phys. Rev.Lett. 30,1053(1973).B. C.Sales and D.Wohlleben, Phys. Rev. Lett. 35, 18(1975); 35,1240(1975).
' A.G.Redfield, Adv. Magn. Res. 1,1(1965); IBM
J.
Res.Dev. 1, 19 (1957),see also C. P. Slichter, Principles ofMagnetic Resonance, edited by M. Cardona, P. Fulde, and H.J.
Queisser, Springer Series in Solid State Sciences, 2nd ed.,Vol. 1(Springer, New York, 1978).
' M. M. Ochi and O. L. T.de Menezes, Solid State Commun.
75,355(1990).