Superconducting to spin glass state transformation in ß-pyrochlore KxOs2O6

1 MB·English

by  C. C. Lee

Superconducting to spin glass state transformation in β-pyrochlore K Os O
x 2 6
C. C. Lee1, W. L. Lee2, J. -Y. Lin3, C. C. Tsuei4, J. G. Lin1, and F. C. Chou1,5∗
1Center for Condensed Matter Sciences, National Taiwan University, Taipei 10617, Taiwan
2Institute of Physics, Academia Sinica, Taipei 11529, Taiwan
3Department of Physics, National Chiao Tung University, HsinChu 30076, Taiwan
4IBM, Yorktown Heights, NY 10598, U.S.A. and
5National Synchrotron Radiation Research Center, HsinChu 30076, Taiwan
(Dated: January 20, 2011)
β-pyrochore KOs O , which shows superconductivity below ∼ 9.7K, has been converted into
2 6
KxOs2O6 (x <∼ 32-12) electrochemically to show spin glass-like behavior below ∼ 6.1K. Room tem-
1 perature sample surface potential versus charge transfer scan indicates that there are at least two
1 two-phase regions for x between 1 and 0.5. Rattling model of superconductivity for the title com-
0 pound has been examined using electrochemical potassium de-intercalation. The significant reduc-
2 tionofsuperconductingvolumefractionduetominorpotassiumreductionsuggeststheimportance
n of defect and phase coherence in the rattling model. Magnetic susceptibility, resistivity, and spe-
a cificheatmeasurementresultshavebeencomparedbetweenthesuperconductingandspinglass-like
J samples.

9
1 PACSnumbers: 74.70.-b,75.50.Lk,74.25.Bt,74.25.F-

]
n I. INTRODUCTION
o
c
- Frustrationhasbeenoneofthemostintriguingcooper-
r ative phenomena in magnetism. The geometrically frus-
p
trated system becomes one fascinating class of material
u
s with emergent novel physical properties by its complex
. nature,includingthelayeredtriangularandKagomelat-
t
a ticein2D,andthe3Dpyrochlorestructurewithacorner-
m
sharing tetrahedral sublattice.1,2 The finding of the first
- superconducting 3D pyrochlore transition metal oxides
d
Cd Re O with T ∼ 1 K in 2001 was surprising.3 The
n 2 2 7 c
o following search of superconductivity in the pyrochlore
c familyhasraisedTc upto9.7KinAOs2O6 withA=Cs,
[ Rb, and K since.4–6

1 Themostgeneralformulaof3DpyrochloreA B O O’
2 2 6
v isusuallydescribedincubicofFd¯3m(No.227)symmetry
9 with A at 16d (1,1,1), B at 16c (0,0,0), O at 48f (x,1,1) FIG. 1: (Color online) (a) The osmium ions form 3D frus-
3 2 2 2 8 8 trated pyrochlore structure. (b) The potassium ions can be
andO’at8b(3,3,3).1 The3DpyrochloreA B O O’can
6 8 8 8 2 2 6 viewedasrattlinginacageformedwith18oxygens, andthe
3 be best described as the frustrating conner-shared tetra- cages are connected through tunnels formed with 6 oxygens
. hedralasshowninFig1,whereeitheroneorbothAand as indicated by a hexagonal mirror plane. Ideally K ions sit
1
B sublattices can be magnetic. Different valence combi- at 8b site as shown in the center of O cage, and the center
0 18
1 nations including A+3/B+3, A+2/B+4, and A+/B+5 are of the O6 pore corresponds to the 16d site.
1 possible, although the preferred site of monovalent ion
: occupies O’ site (8b) instead of A site (16d) as a result
v
of lower and stable electrostatic potential.7 the midpoint connecting two Os O cages within the
i 12 18
X The deficient type pyrochlore with monovalent ion oc- large pore center formed by 6 oxygens.
r cupying 8b (O’-site) instead of 16d (the regular A-site) The origin of superconductivity found in
a
is called β-type pyrochlore, in constrast to the regular (K,Rb,Cs)Os O has been proposed by a rattling
2 6
α-type A B O but written as AB O usually, which is model and tested experimentally,8 where effective inter-
2 2 7 2 6
thecommonstructureofthetitlecompoundofKOs O . action between two electrons is proposed to be coming
2 6
KOs O has a frustrating structure for Os sublattice in from the anharmonic potential generated by the rattling
2 6
corner-sharedOsO octahedraasshowninFig.1(a). An- motion of the K+ ions caged in Os O clusters. Such
6 12 18
otheraspecttodescribethesystemforKandOrelation- rattling-induced superconductivity has been partially
shipisshowninFig.1(b),whereKiscagedby18oxygen supported by the photoemission study as well.9 On
ions while sitting in the K-tetrahedral center. As seen in the other hand, the strong dependence of unit cell size
Fig. 1(b), the empty 16d site, which is usually occupied on T , i.e., the higher T from Cs to K family with
c c
byhighervalenceAinregularα-A B O ,islocatednear progressively larger ion size,4 seems to suggest that an
2 2 7
2
even higher Tc is waiting to be found if Na substitution 1.4
becomes possible. Although Na has been introduced f)
e
irnotuotethferomsystKeOmssOucc,1e0ssftuhlelyntohnrsotuogichhiaonmeitornicexpcrhoadnugcet Cl r 1.2
2 6 g
Na Os O ·H O turns out to be non-superconducting A 1.0
dow1.n4 to2 26K.2Structure refinement suggests that Na g/ Chronopotential
ion does not occupy the original K site at 8b within s A 0.8 Quasi OCP
v
the Os12O18 cage, but at the 32e site, which is in the t 0.6
neighborhood of 16d site near the large O pore center, ol
6 V
i.e., Na1.4Os2O6·H2O practically belongs to the regular P( 0.4
α-type pyrochlore instead of the β-type pyrochlore such C
asWKOhisle2ON6a. substitution failed to produce the identical ,Ocp0.2 1JN=1 N.5aAC/glO /PC solution
V 0.0 4
structure of KOs O and to raise T as expected, we de-
2 6 c 0 50 100 150 200 250 300 350
cided to use an alternative route on tuning K content to
Q (Coul/g)
test the validity of the proposed rattling-induced super-
conductivity model further. Using the electrochemical
chronoamperemetrymethod,wehavereducedpotassium
FIG. 2: (Color online) Surface potential and quasi-
contentfromthepristineKOs O toK Os O (x∼0.5)
2 6 x 2 6 equilibriumopencircuitpotential(OCP)versuschargetrans-
successfully. Wefindthatthesystemcanbetransformed fer Q for a working electrode made of compressed KOs O
2 6
fromsuperconductingtospinglassstatebelowTg ∼6.1K powder,wheresurfacepotentialisobtainedfromachronoam-
afterpotassiumisreducedtohalfofitsoriginallevel. The paremetry scan of constant current density of 1.5 A/g (red
resistivityandspecificheatmeasurementresultsarealso line)andthequasi-equilibriumOCP(blackline)isdescribed
reported for a detailed comparative study. in the text. Samples of different potassium content are pre-
pared using applied voltages V =OCP to produce samples
ap
with specific x and verified by EPMA.
II. EXPERIMENT
charge transfer current level had reached time indepen-
Polycrystalline sample KOs O was prepared from
2 6 dentconstantbackground. Potassiumcontentcanbere-
reagent-grade oxide powders of KO and OsO mixed
2 2 duced progressively from the original 1 to about ∼0.5 by
with an appropriate molar ratio (KO :OsO = 1:2) in
2 2 applying constant anodic potentials from -0.05 to 1.25
an argon-filled glove box and pressed into a pellet with
V relative to the Ag/AgCl Reference electrode. Crys-
mass of ∼0.167 g. The compressed pellet was wrapped
tal structure were analyzed by Bruker D8 X-ray diffrac-
with gold foil, sealed in an evacuated glass tube and
tometer. Magnetic properties were measured using a
heated to 450 ◦C for 72 hours. To control the oxygen
SQUID magnetometer (Quantum Design SQUID-VSM).
partial pressure, a 0.06g Ag O pellet was added to the
2 Since powder samples were prepared using a close oxida-
sealed glass tube to create an oxygen partial pressure at
tion environment at a relatively low temperature of 450
about ∼12 atm at 450 ◦C. Potassium content of the as- ◦C, all following resistivity and specific heat measure-
prepared sample is near 1 per formula unit based on the
ment were done on cold compressed pellet under 6 GPa
electron probe microanalysis (EPMA) using a spray-on
for 20 minutes to reach maximum density. Resistivity
layer of polycrystalline powder, which is consistent with
measurements were carried out by a resistance bridge at
that reported from sample prepared and analyzed using
low current excitation using standard four-probe geome-
the identical method as before.4
try. Since the hydration is a reversible process, we made
Additional potassium content tuning was accom- all physical property characterization on samples dried
plished through an electrochemical de-intercalation pro- at 175 ◦C for 8 hours beforehand.
cess. Anelectrochemicalcellwassetupusing1NNaClO
4
in propylene carbonate (PC) as electrolyte, Ag/AgCl
reference electrode, and platinum as counter electrode.
Due to the potential formation of hydrated form,11 the III. RESULTS AND DISCUSSIONS
KOs O sample working electrode was heated to 175 ◦C
2 6
for 8 hours before the electrochemical experiment and A. Electrochemical de-intercalation
non-aqueouselectrolytewasused. Theworkingelectrode
was composed of polycrystalline KOs O compressed in SinceKionsrattlewithinthelargeOs O cageof∼1
2 6 12 18
pellet form. An initial chronopotentiometry scan with ˚Aanharmonicallyassuggestedbythedetectablephonons
current density of 1.5 A/g was used to trace the quasi- from fine spectral structure of ARPES experiment,9 and
equilibrium sample potential first. Final samples were the ion exchange has been shown to be possible between
preparedwiththechronocoulometrytechnique,applying Na and K ions,10 plus the fact that Os ion has 5d elec-
voltagesfrom-0.05to1.25V/Ag-AgCluntiltheinduced trons of rich valence variation from +2 to +6, we believe
3
that it is highly possible that the reduction of K con-
tent can raise the average valence of Os from the current (a) 1) 2) Pristine KOs2O6
mixed valence of Os+5.5 in KOs O , to the Os+6 state ) 1 2 * OsO
2 6 s 3 2 2
in Os2O6 as a result of the completely emptied cations nit 1) ( (
that potentially occupy the 16d and 8b sites. To explore u 11
the existence of possible stable phases as a function of b. ( 0) 2)
sxbitnhoetidecneoiruamKcmsax-elpsaOlrtepseivo2poeOnsalsr6piieb,nrdloweNceKewashOhxsaCestv2nooeOOCce6om2l.On,1pt42rSl−ooipylmioetsidthlaaaeasdrsKnsitoueolrmelbevtecehddtler,eot-soirctnehattderhetumreiccnitsacgialaoamflnrtoidpomoelne-f Intensity (ar (400) (331) (511) (44(531) (533)(62(444)
surface and react with the surface K, and the follow-
* *
ing K ion self diffusion proceeds to reach a new equilib- * * *
rium at a lower average total K level for the bulk, which
10 20 30 40 50 60
should be allowed according to the relatively high elec- 2θ (deg.)
trical and ionic conductivity for the similar pyrochlore
structure compounds at room temperature.13 (b) (622)
Chronopotentiometry technique was applied to the )
s (533)
system on an electrochemical cell constructed as nit
KOs O /1NNaClO inPC/Pt. Non-equilibriumsurface u
poten2tia6l (Vcp) vers4us total charge (Q) passing through b. 1.05V
r
sample working electrode is shown in Fig. 2 under ap- a
(
plied constant current density of 1.5 A/g. The initial y
t 0.85V
open circuit potential V◦ of ∼0.1 V has been raised to si
∼ 0.85 V after nearly one half of the K+ ions had been en 0.20V
reduced from the original, assuming 100% efficiency of nt
I
charge transfer. Continued de-intercalation process re-
duced potassium level further until side reaction was ob- Pristine
served after surface potential had reached beyond ∼ 1.2
59.5 60.0 60.5 61.0 61.5
V. Based on the preliminary chronopotentiometry scan 2θ (deg.)
shown in Fig. 2, three plateau shape regions correspond-
ing to the two-phase character can be roughly identified,
i.e., the first extremely narrow region near V ∼ 0.2-0.4,
themiddleplateaubetween0.5-1.1,andthefinalplateau FIG. 3: (Color online) X-ray diffraction (XRD) patterns for
(a) the as-grown KOs O , and (b) various de-intercalated
above ∼1.3V. A significantly higher charge transfer was 2 6
observed for Vcp >∼ 1.2V due to side reactions, as ob- KglxesO.s2O6 samples which show slightly shifted diffraction an-
served from the black deposit on the counter electrode
Pt surface and volatile vapor bubbling at the cathode.
SincetheblackdepositfoundonthePtcounterelectrode
Thereexistatleasttwodistinguishablestablephasesthat
disappeared after being exposed to the air, it is possible
are separated by the two-phase plateau regions: one is
that the deposit was the highly toxic and volatile OsO .
4 proximate to the pristine KOs O phase and the sec-
2 6
A separate quasi-equilibrium open circuit potential
ond has stoichiometry close to K Os O as later ver-
0.5 2 6
(OCP) versus Q has been obtained from the same piece
ified by the EPMA chemical analysis. Following this
of KOs O working electrode in situ using chronopoten-
2 6 quasi-equilibrium OCP plot, samples of specific potas-
tiometry scan after the current is off for 260 seconds.
sium content reported in this study have been prepared
Since the background current cannot be estimated ac-
using corresponding V =OCP and final potassium con-
ap
curately before the I(t) curve is fully saturated to the
tentsweredeterminedbasedonEPMAchemicalanalysis.
background level, and there may be other undetectable
Samples prepared using different V are summarized in
ap
reactions (either anodic or cathodic) included in the to-
Table I.
tal charge transfer, we cannot convert the obtained total
charge transfer Q (Coul/g) into actual amount of K de-
intercalated accurately, i.e., assuming one electron loss
corresponds to one K+ de-intercalated from the KOs O B. X-ray diffraction
2 6
surface. There are three plateau regions to be noted
in the quasi-equilibrated OCP vs. Q diagram, one nar- Figure3(a)showsX-raydiffraction(XRD)patternfor
row plateau onset near ∼ 0.15V, the second onset near the as-grown KOs O powder at room temperature. All
2 6
0.55V, and the final plateau onset near ∼ 0.90V which diffractionpeakscanbeindexedbyFd3mstructurewith
must correspond to the side reaction as discussed above. lattice constant a = 10.073(3) ˚A. A few extra peaks la-
4
beled with asterisk correspond to the existence of OsO
2 0 (a)
impurity phase and it has been reported that the ex-
FC
istence of KOsO and OsO impurity phases are un-
4 2
avoidable for the polycrystalline sample prepared using g) -2
similarpreparationconditions.4,11 However, wefindthat /
3
m
the KOsO impurity can be washed off with propylene
4 c
carbonate (PC) solution effectively. Fig. 3(b) illustrates -30 -4 0.0
the XRD patterns at high angles for the de-intercalated 1
(
KOs2O6 treated with electrochemical de-intercalation at H -0.2
various applied voltages, which indicates the occurrence M/ -6 -0.4
ofaslightlyenlargedcellsizeasaresultofapplicationus- ZFC
-0.6
inganevenminimaloverpotentialaslowasVap∼0.15V. H=25Oe V =0.20V
It is obvious that high angle diffraction peaks shift to -8 Pristine -0.8 app
0 5 10 15 20
loweraftertheelectrochemicalprocessingandnoobserv-

0 5 10 15 20
able extra diffractions peaks to suggest extra impurity
phases generated otherwise .
0.012 (b) FC 0.10
The lattice parameters calculated from high angle
diffraction peaks have been indexed with Fd3m space 0.08
group and summarized in Table I. We find that the lat- ) 0.06
g
tice size has an insignificant (<∼0.01 ˚A) jump from the 3m/0.008 0.04
pristine superconducting sample of KOs O in general
2 6 c 0.02
and levels off quickly to be nearly independent of V 3
after V ∼0.2V and x. Such change occurs immediatealyp -10 0.00
ap ( ZFC 0 5 10 15 20
after a small overpotential is applied with Vap as low as H 0.004
∼ 0.2V to induce K reduction, which is rather surpris- M/
ing since the corresponding potassium is reduced by less
H=25Oe
than 1 % only as indicated in Fig. 2. Water intercala-
Spin glass
tionisacommonphenomenonforKOs O sample,which
2 6 0.000
causes deterioration of superconductivity and is accom-
0 5 10 15 20
panied by a slight lattice size increase.11 In fact, most
Temperature (K)
reported lattice parameters of the pristine KOs O were
2 6
possibly overestimated as a result of potential water in-
tercalation, whereas our data fall closest to the sample
prepared and processed strictly in a dry box.11 Since the FIG.4: (Coloronline)Temperaturedependenceofdcaverage
hydration is a reversible process, we made all physical susceptibilityforKxOs2O6 measuredunderappliedmagnetic
property characterization on samples dried at 175 ◦C for fieldof25Oe. (a)showssuperconductingsampleofthepris-
tineKOs O andtheinsetshowssamplewithslightlyreduced
8 hours beforehand. Contrary to what is expected from 2 6
KusingV =0.20V,thelatterhassimilarT onsetbutwith
potassiumreductionthroughde-intercalationbyprogres- ap c
much lower superconducting volume fraction. (b) shows spin
sively higher V , the lattice size increases while it is ex-
pectedtodecreaapsewhenmoresmallerOs6+ ionsaregen- glasssamplesofKxOs2O6(x<∼2/3)preparedusingVap=1.16
V (T ∼ 3.3K) and V =1.05 V (inset, T ∼ 6.1K), respec-
g ap g
erated. It is highly possible that average valence of Os
tively.
higher than +5.5 may not be preferred in the pyrochlore
structure and induces oxygen defect formation, which is
partly supported by the fact that most compounds of
Fig.4(a). Suchdrasticreductionisconcomitantwiththe
high average Os valence prefer perovskite type crystal
relatively sharp increase in lattice size within the same
structure, e.g., Ba CaOsO and Ba LiOs O .14,15
2 6 3 2 9 narrow range below <∼ 0.20V also shown in Table I, be-
yondwhichthereisnomorelatticechangeandthesuper-
conducting volume fraction is reduced to the trace level
C. Magnetic susceptibility beforefallingintothespinglassregion. Infact,asimilar
drastic change on the surface potential (OCP) has also
Bulk superconductivity of T ∼ 9.7 K has been veri- been observed at the same time for low V in the range
c ap
fied for the pristine KOs O sample as shown in Fig. 4 of ∼ 0.16-0.26V (first plateau) as shown in Fig. 2. It is
2 6
byitssuperconductingvolumefractionof∼20%fromthe clear that superconductivity exists within an extremely
field-cooled data without demagnetization factor correc- narrowwindowofstoichiometryKOs O andissensitive
2 6
tion due to particle geometry. Interestingly, supercon- tothelatticesize. Fromthepointofchemicalpressurein
ducting volume fraction is significantly reduced after a singlebandmodelofconventionalBCSpicture,theeasily
minor (∼ 1%) potassium reduction using Vap >∼ 0.2 V enlargedcellsizereducesDOSeffectivelytokillsupercon-
whileT remainsaboutthesame,asshownintheinsetof ductivity. If we follow the rattling model interpretation,
c
5
TABLE I: Summary of sample preparation and physical properties.
Applied voltage(V) Lattice constant(˚A) Magnetic Behavior T /T S.C. vol. fraction x
c g EPMA
pristine 10.072 S.C. 9.7K 20.1% 1.00(4)
0.16 10.077 S.C. 9.7K 17.4% *
0.20 10.080 S.C. 9.7K 3.9% 1.01(3)
0.24 10.079 S.C. 9.7K 4.9% *
0.50 10.079 S.C. 9.7K 1.2% 0.97(2)
0.64 10.081 S.C. 9.7K 5.8% *
0.85 10.078 Spin glass 4.6K * 0.65(6)
1.05 10.078 Spin glass 6.1K * *
1.16 10.078 Spin glass 3.3K * 0.48(6)
therequirementofhighlystoichiometricKOs O implies 8
2 6
Curie-Weiss law fitting
theimportanceofphasecoherenceandintoleranceofde-
fects under rattling model, which can be compared with

) 2.0
the similar phenomenon of hydration reduced supercon- e6
ductivity, especially when the water molecule occupies ol 1.5
m
the 32e site which is near potassium (8b site) and re- 3/ 1.0
duces phonon frequency significantly.11 cm4 0.5
After passing the two-phase region using Vap >∼ 0.85V -30
as indicated by the rough plateau region in Fig. 2, single (1 0.0 Pristine
pshhoawseasasmpipnlegloafssK-lixkOesZ2OFC6/(FxC<∼ir32r-e12v)erhsaibslebebeenhafovuionrdbteo- M/H 2 -0.5 0 100 200 300

lowT ∼6.1K,asseeninFig.4(b). Thisisthefirsttime
g
observation of spin glass-like phase in the β-K Os O ZFC @ H=1T Spin glass
x 2 6 0
system. Consider the possible random distribution of
0 100 200 300
potassium ions after the reduction, the impact of potas- Temperature (K)
siumionrandomdistributiononitsnearbyOsspins,and
thenaturalfrustrationrequirementforOsspinsin3Dpy-
rochlore structure, provide the perfect necessary condi-
FIG.5: (Coloronline)Normalstatemagneticsusceptibilityof
tions for a canonical spin glass state with localized spins K Os O asafunctionoftemperatureunderappliedfieldof1
x 2 6
on Os.19 While it is difficult to secure reliable structure Teslaforthe(a)pristinepowdersampleand(b)potassiumde-
refinement on the low temperature synthesized powder intercalated sample using V = 1.05 V. The corresponding
ap
samples, it is reasonable to assume that the randomness low field data are shown in Fig. 4.
couldbecomingfromtheincreasedoccupancyof16dsite
(i.e., the preferred site of high valence A in α-A B O )
2 2 7
when the original 8b site is becoming half empty. We spins is not justified for an itinerant system; in fact, the
find that as the potassium content is reduced progres- relativelylowandweaktemperaturedependenceofχ(T)
sivelyunderhigherV ,thereisnosignificantchangeon does suggest that the contribution seems mostly coming
ap
the superconducting transition temperature T and spin from itinerant Pauli paramagnetic spin contribution,
c
glass-like transition T , although there is clear relative assuming all other temperature independent terms from
g
superconductingvolumefractionchange,whichindicates core diamagnetic and Van Vleck paramagnetic contribu-
that only two single phases co-exist, and is in agreement tions are canceled out. On the other hand, unlike the
withthemiddleV-QplateauasshowninFig.2. Inaddi- usually strongly magnetic 3d and 4d elements, the more
tion, the application of overpotential to the sample elec- spreading 5d orbitals make the electron correlations and
trodecouldalsoinducepossiblepotassiumdiffusionfrom spin-orbit interactions more complex and unexpected.
the original 8b site to the 16d site similar to that of the For example, the semiconducting-metal transition near
regular α-pyrochlore, which can also partly explain the 225KfoundinCd Os O pyrochlorehasbeenexplained
2 2 7
significant reduction of superconducting volume fraction in Slater mechanism of local band antiferromagnetic
when only less than ∼ 1% potassium is reduced electro- ordering.16
chemically as shown in Table I. Enforcing a Curie-Weiss law fitting for χ(T) as
High field magnetic susceptibilities for both the χ +C/(T-θ) to both the superconducting and spin glass
◦
pristine superconducting KOs O and the non- samplesreturnssatisfactoryfittedvaluesofC∼0.061and
2 6
superconducting de-intercalated K Os O are shown 0.035 cm3K/mole plus θ ∼ -106K and -8K, respectively.
0.5 2 6
in Fig. 5. Since the pristine sample is metallic above T , TheCurieconstantsaresignificantlylowerthanthatcal-
c
the application of Curie-Weiss law to describe localized culated from the theoretical value for a completely lo-
6
calized spin with average valence of +5.5 in high-spin
state.10 The effective moment for the spin glass sam- 12 Pristine
Spin glass
ple calculated from fitted C value is ∼ 0.53 µ . For
B
K0.5Os2O6 in a completely localized ionic picture for Os 10 2.10
)
ion with octahedral oxygen ligands, the average Os va- m
lence is +5.75, i.e., 25% Os5++75% Os6+. Comparing c 8 2.05
-
the fitted µ with the theoretical µ estimated from Ω
various posseifbfle low spin (LS)/high sepfifn (HS) combina- m 6 2.00
tions, the most probable case is that both are in the LS (x 1.95
x
state(i.e., 25%Os5+=1/2and75%Os6+=0)withathe- ρ 4
1.90
oreticalspinonlyµ of0.43µ . TheLSstateassump- 0 50 100 150 200
eff B
tionisindrasticcontrasttomostoftheOs5+ compounds 2
with double perovskite structure in HS state, such as
La NaOsO ,22 and the triple perovskite Ba LiOs O of 0
2 6 3 2 9 0 50 100 150 200
mostlyHSOs5.5+,14 while5dorbitalsaremorespreading
Temperature (K)
compared with the usually magnetic 3d/4d compounds,
which makes the electron correlation and spin-orbit cou-
pling more complex. Current spin only LS implication
FIG. 6: (Color online) Resistivity versus temperature for
could be misleading still and its impact to the spin glass
the superconducting KOs O and spin glass-like K Os O
2 6 0.5 2 6
formation remains unclear. Further characterization us-
samples. The spin-glass sample has T ∼3.3K as shown in
g
ing larger single crystal of improved homogeneity is nec- Fig.4(b). TheinsetshowsadetailedviewfortheK Os O
0.5 2 6
essary. sample.
The Weiss temperature θ for the pristine KOs O is
2 6
comparabletothatfoundinNa Os O ·H O,10 butθ is
1.4 2 6 2
foundtobesignificantlyreducedforthespinglasssample tering dominates at higher temperature, which gives
K Os O ,i.e.,theantiferromagneticcorrelationamong residual resistivity ratio (RRR) = ρ(250K)/ρ(10K)∼2.
0.5 2 6
isolatedspinsaresignificantlyreducedforthelatter. The Ontheotherhand, ρ(200K)∼1.9mΩ-cmforspinglass-
generation of a canonical spin glass state requires satis- like K Os O sample of T ∼3.3K (see Fig. 4(b)) is 6
0.5 2 6 g
faction of both frustration and randomness conditions, folds smaller than that for the pristine superconducting
which has led to the puzzling observation on the find- KOs O . It only increases by 5% as T drops to 4.2 K
2 6
ing of spin glass state in pyrochlore Y Mo O of unclear as shown in the inset of Fig. 6. The significantly weaker
2 2 7
originofrandomness.17Currentobservationofspinglass- T dependence in ρ for K Os O stands in great con-
0.5 2 6
like behavior in K Os O implies that the K+ ions (or trasttothatofthesuperconductingKOs O sample. As
0.5 2 6 2 6
oxygen defects, if any) must distribute randomly, most pointed out by Nagao et al.,8 electron-phonon coupling
probably between 16d and 8b sites, and lead to the ran- is significantly enhanced by the rattling of alkali ions,
domness on top of the tetrahedrally coordinated frus- which makes KOs O a strong-coupling superconductor
2 6
trating Os spin substructure. Because of the limitation with the highest T in pyrochlore family. We therefore
c
on sample quantity and powder inhomogeneity, detailed attribute the strong suppression of T-dependence of ρ
exploration of spin glass properties using ac spin suscep- for the spin glass-like K Os O is associated with the
0.5 2 6
tibility measurement should be performed once sizable removalofrattlers(K+ ion),whichshutsofftheelectron-
single crystal sample is available. phononcouplingthatleadstothesuperconductivity. Al-
beit with the complications from the grain boundary
scattering and also the disordering from K+ ion extrac-
tion,theweakerTdependenceofρduetotheabsenceof
D. Resistivity
rattlingphononsisfurthersuggestedbythespecificheat
measurements to be discussed next. Further investiga-
Resistivity data for both the pristine superconducting
tion using single crystal sample is needed to clarify this
sample KOs O and K Os O with spin glass-like be-
2 6 0.5 2 6 issue.
havior are shown in Fig. 6. The onset of superconduc-
tivity occurs at T∼9.7K for KOs O is confirmed, with
2 6
ρmonotonicallyincreaseswithtemperatureinaconcave
curvature. Thisbehavioragreeswithdatapreviouslyob- E. Specific heat
tained using single crystal sample,18 where anharmonic
potential resulting from the K+ ion rattling plays an im- Specificheatmeasurementresultsforboththepristine
portant role. However, we do notice that ρ(200K)=12.5 and the K+ de-intercalated samples are shown in Fig. 7.
mΩ-cm is 50 folds higher than that in a single-crystal. KOs O has a pronounced anomaly onset ∼ 9.7 K due
2 6
This suggests a significant contribution from the grain to the superconducting phase transition. C/T extrapo-
boundary scattering in our cold-compressed polycrys- lates to zero at T ∼ 0, which indicates a nearly 100%
talline sample. Nevertheless, the electron-phonon scat- superconductingvolumefractioninKOs O . Hiroiet al.
2 6
7
were described with dominant β T5 term and β ∼ 0.8
Pristine 5 3
This anomalous phonon contribution has been argued
500 Spin glass
to be due to the anharmonicity which led to the alkali
) 50
-2K ion rattling. The investigation of the phonon term for
-1l400 40 KOs2O6 has been claimed to be hindered by its super-
o conductivity with higher T . Nevertheless, based on our
m c
300 30 results shown in Fig. 7, both the magnitude and tem-

J perature dependence of C(T) for KOs O above T are
m 2 6 c
20
T (200 0 10T2 (K22)0 30 vtleirnygpdhiffoenroenntcofnrotrmibuthtioosneisofbeKli0e.v5eOds2tOob6,ewsimhiillearthtoetrhaatt-
/
C of CsOs O and RbOs O . The Debye temperature de-
100 2 6 2 6
rived from β is ∼ 300 K for K Os O , in contrast to
3 0.5 2 6
C = γT + βT 3 + βT 5 fitting the nearly zero contribution for the reported supercon-
0 3 5 ducting samples of (Cs,Rb)Os O .8 The revival of the
0 50 100 150 200 250 2 6
lattice contribution of the Debye phonons clearly sug-
T2 (K2)
geststheabsenceofrattlinginK Os O . Consequently,
0.5 2 6
the rattling-induced superconductivity is not realized in
K Os O owing to this drastic change of the electron-
FIG. 7: (Color online) Specific heat measurement results of 0.5 2 6
phonon interaction, which is consistent with the ρ(T)
pristine superconducting KOs O and K Os O . This par-
2 6 0.5 2 6
data as shown in Fig. 6. In K Os O , γ=27.8 mJ/mol
ticularbatchofreducedpotassiumhasaspinglass-liketran- 0.5 2 6
sition of ∼ 3.3K as revealed by the ZFC/FC hysteresis onset K2 is much smaller than γ=70 mJ/mol K2 of KOs2O6.
in magnetic susceptibility as shown in Fig. 4(b). It is not clear whether this smaller γ is due to the lower
K+ level or the change in band structure owing to the
absence of the rattling electron-phonon interaction.
havereportedtheobservationofanadditionalfirstorder
phasetransition near7.5K before, which canonly be ob-
servedinsinglecrystalbutneverinpowdersamples.20,21
The origin of this additional phase transition below T
c
hasbeensuggestedtobeduetothemultipolartransition
driven by the octupolar oscillation component.23 While
IV. CONCLUSIONS
current specific heat measurement is based on powder
sample of greater inhomogeneity comparing with single
In summary, spin glass-like behavior below ∼ 6.1 K is
crystal, it is not surprising for failing to observe such
observed when the potassium level is reduced below ∼
multipolar isomorphic structural transition of first order
2-1 from its original level of 1 through an electrochem-
in nature. 3 2
ical de-intercalation process. The existence of spin glass
K Os O carriesnosignofsuperconductivityandbe-
0.5 2 6 statehasbeensupportedbythethermalhysteresisbelow
haves like a normal metal. Furthermore, while the mag-
T in dc spin susceptibility and the deviation from the
netization measurements reveal a spin-glass like signa- g
fitting of specific heat in C(T)=γT+β T3+β T5 beyond
ture near ∼ 3.3K (see Fig. 4(b)), there is no apparent 3 5
electronic and phonon contributions below T . The ran-
associated specific heat peak in this case. The lack of g
domness and frustration requirement for the existence
C(T) peak is exactly one of the features for spin glass
of spin glass state could be justified by the randomly
phase transition,19 owing to the much released entropy
distributed K+ vacancies and the frustrating tetrahedral
above T . To further illustrate the spin glass signature
g coordinated Os sub-structure. Homogeneous and sizable
ofC(T)inK Os O ,C(T)isfitbyC=γT+β T3+β T5
0.5 2 6 3 5 singlecrystalsampleisdesirableforfurtherinvestigation
aboveT2 >10K2,whereγTandβ T3+β T5 correspond
3 5 ontheoriginofthisnewlyfoundspinglassphaseandits
to the electronic and phonon contributions. The fit-
deeper connection to the superconducting state.
ted results are γ=27.8 mJ/mol K2, β =0.607 mJ/mol
3
K4 and β5=0.00049 mJ/mol K6. As seen in the inset Basedontherattlingmodelinterpretationofsupercon-
of Fig. 7, there exists a small but clear extra contri- ductivity,thecrossoverfromthesuperconductingtospin
bution beyond the electronic and phonon contributions, glass state must be related to the strong suppression of
which could well be attributed to the spin glass phase. electron-phonon coupling responsible for the occurrence
ThisspinglasscontributionmakesC(T)deviatefromthe of superconductivity, which has been supported by both
C=γT+β3T3+β5T5 at T2 <∼ 10 K2, in accord with the thetransportandthermodynamicpropertymeasurement
observedmagneticsusceptibilitycuspwithonsetnear3.3 results. While no clear phase diagram has been mapped
K. using the existing powder samples, the proximity of su-
The phonon contribution to C(T) in K Os O is an- perconducting and spin glass states has been once more
0.5 2 6
other intriguing subject. In both superconducting sam- demonstrated to be similar to that found in the high T
c
ples of CsOs O and RbOs O , phonon contributions and iron pnictide superconducting systems.
2 6 2 6
8
Acknowledgment M-002-021.
FCC acknowledges the support from National Science
Council of Taiwan under project number NSC-98-2119-
∗ Electronic address: [email protected] 12 G.J.Shu,A.Prodi,S.Y.Chu,Y.S.Lee,H.S.Sheu,and
1 J. S. Gardner, M. J. P. Gingras and J. E. Greedan, Rev. F. C. Chou, Phys. Rev. B 76, 184115 (2007).
Mod. Phys. 82,53 (2010). 13 A. W. Sleight, J. E. Gulley, and T. Berzins, Solid Stat.
2 A. P. Ramirez, Annu. Rev. Mater. Sci. 24, 435 (1994). Chem. of Energy Conversion and Storage, Chapter 11, p.
3 M. Hanawa, Y. Muraoka, T. Tayama, T. Sakakibara, J. 195.
Yamaura,andZ.Hiroi,Phys.Rev.Lett.87,187001(2001). 14 K.E.Stitzer,A.E.Abed,M.D.Smith,M.J.Davis,S.J.
4 S.Yonezawa,Y.Muraoka,andZ.Hiroi,J.Phys.Soc.Jpn. Kim, J. Darrie,t and H. C. Z. Loye, Inorg. Chem. 42, 947
73, 1655 (2004). (2003).
5 S.Yonezawa,Y.Muraoka,Y.Matsushita,andZ.Hiroi,J. 15 K. Yamamura, M. Wakeshima, Y. Hinatsu, J. Solid Stat.
Phys. Soc. Jpn. 73, 819 (2004). Chem. 179, 605 (2006).
6 S.Yonezawa,Y.Muraoka,Y.Matsushita,andZ.Hiroi,J. 16 D. Mandrus et al., Phys. Rev. B, 63, 195104 (2001).
Phys.: Condens. Matter 16, L9 (2004). 17 J.E.Greedan,M.Sato,andXuYan,SolidStateCommun.
7 J. Pannetier, J. Phys. Chem. Solids 34, 583 (1973). 59, 895 (1986).
8 Y. Nagao, J.I. Yamaura, H. Ogusu, Y. Okamoto, and Z. 18 T.DahmandK.Ueda,Phys.Rev.Lett.99,187003(2007)
Hiroi, J. Phys. Soc. Jpn. 78, 064702 (2009). 19 J. A. Mydosh, Spin glass: an experimental introduction,
9 T. Shimojima, Y. Shibata, K. Ishizaka, T. Kiss, A. Taylor & Francis (1993).
Chainani,T.Yokoya,T.Togashi,X.-Y.Wang,C.T.Chen, 20 J. I. Yamaura, M. Takigawa, O. Yamamuro, and Z. Hiroi,
S. Watanabe, J. Yamaura, S. Yonezawa, Y. Muraoka, Z. J. Phys. Soc. Jpn. 79, 043601 (2010).
Hiroi, T. Saitoh, and S. Shin, Phys. Rev. Lett. 99, 117003 21 Z. Hiroi, S. Yonezawa, and J. I. Yamaura, J. Phys. Con-
(2007) dens. Matter 10, 145283 (2007).
10 Y.G. Shi, A.A. Belik, M. Tachibana, M. Tanaka, Y. Kat- 22 William R. Gemmill, Mark D. Smith, Ruslan Prozorov,
suya, K. Kobayashi, K. Yamaura, E. T. Muromachi, J. andHans-ConradzurLoye,Inorg.Chem.,44,2639(2005).
Solid Stat. Chem. 182, 881 (2009). 23 K. Hattori and H. Tsunetsugu, arXiv:1101.0647v1.
11 R.Galati,C.Simon,C.S.Knee,P.F.Henry,B.D.Rain-
ford, and M. T. Weller: Chem. Mater. 20, 1652 (2008).