Intrinsic paramagnetism and aggregation of manganese dopants in SrTiO$_3$

0.75 MB·

by  Andrej Zorko

Intrinsic paramagnetism and aggregation of manganese dopants in SrTiO
3
A. Zorko,1,2, M. Pregelj,1 H. Luetkens,3 A.-K. Axelsson,4 and M. Valant5
∗
1Jožef Stefan Institute, Jamova c. 39, SI-1000 Ljubljana, Slovenia
2EN–FIST Centre of Excellence, Dunajska c. 156, SI-1000 Ljubljana, Slovenia
3Laboratory for Muon Spin Spectroscopy, Paul Scherrer Institute, CH-5232 Villigen, Switzerland
4London South Bank University, 103 Borough Road, London SE1 0AA, London, UK
5University of Nova Gorica, Vipavska 13, SI-5000 Nova Gorica, Slovenia
(Dated: March 25, 2014)
Usinglocal-probemagnetic-characterizationtechniquesofmuonspinrelaxationandelectronspin
4
1 resonancewehaveinvestigatedtheMn-inducedmagnetismofthewide-band-gapperovskiteSrTiO3.
Ourresultsclearlydemonstratethatthisdilutedmagneticoxideremainsparamagneticdowntolow
0
2 temperatures for both doping cases, i.e., when Mn substitutes for Sr or Ti. In addition, both
experimental techniques have revealed that the distribution of individual Mn2+ and Mn4+ ions is
r
nonrandom,as theseions partially aggregate intonanosized clusters.
a
M
PACSnumbers: 75.50.Pp,75.30.Hx,76.30.-v,76.75.+i

2
2
I. INTRODUCTION regation of magnetic species into Mn O -like clusters
3 4
and thus to be extrinsic.25,30 Moreover,such magnetism
]
l wasshowntocriticallydependonsynthesisconditions.30
e Oxide semiconductors with a fraction of the host
- cations substituted by magnetic ions, i.e., diluted mag- Thus, the intrinsic magnetic ground state of the Mn-
r doped SrTiO remains elusive and calls for an in-depth
t netic oxides (DMOs), have been presently intensively in- 3
.s vestigated for potential spintronic applications.1–8 The magnetic study. Furthermore, the role and size of su-
t perexchangeinteractionsbetweendopants33inSrTiO is
a quest for new multiferroic/magnetoelectric DMO mate- 3
m rials that would be of great technological importance unclear. These magnetic interactions could further be
influenced by a seemingly inevitable presence of oxygen-
- is, however, hampered by the general concern whether
d their magnetism, coupled or uncoupled to a polar state vacancies(VO) inthese materials,34 which mightleadto
n of the material, is undoubtedly intrinsic.9–12 Lately, bound magnetic polarons.35 This mechanism could then
o explain the ferromagnetic ordering of some Mn-doped
there has been huge interest in the electronically and
c SrTiO materials,34 but needs to be critically assessed.
[ chemically doped wide-band-gap perovskite SrTiO3. In 3
this incipient ferroelectric that is highly polarizable, In order to address all the above-mentioned pending
2 issues about the Mn-doped SrTiO , we have performed
yet it misses the electric freezing even at T 0 3
v due to quantum fluctuations,13 a small concentr→ation an in-depth local-probe investigation. Here, we combine
7
of dopants can drastically change its electric and mag- results of muon spin relaxation (µSR) and electron spin
0
resonance (ESR) investigations. A further insight into
6 netic properties. To name only the most prominent ex-
0 amples, superconductivity,14,15 a magnetoelectric multi- the dopant-induced electronic states in SrTiO3 has been
. glass state,16,17a two-dimensional electron gas,18,19 and providedby diffuse reflectancespectroscopy(DRS) mea-
1
0 ferromagnetism20–22 have been reported in the bulk or surement. We show that all the investigated samples,
including the A-site doped ones, remain paramagnetic
4 at the surface of SrTiO3 and at interfaces with other in-
1 sulating oxides. down to low temperatures and that the Mn ions tend to
: Manganese impurities in SrTiO , in particular, have form dopant-reach islands. Similar nonrandom distribu-
v 3
tions of dopants on a nanoscale – spinodal decomposi-
i beenthoroughlyinvestigated,becausetheycanbemulti-
X tioninto regionswith high/lowconcentrationof dopants
valentandbecauseoftheirpolarandmagneticmoments.
r Theyareheldresponsibleforarelaxor-typepolarstate23 and the same crystal structure as the semiconducting
a when Mn2+ substitutes for Sr2+ (at the dodecahedral host–wereobservedbefore indifferentdiluted magnetic
semiconductors.5,36–39 However, in doped SrTiO such
A site), while on the other hand, the material remains 3
quantum paraelectric when Mn4+ substitutes for Ti4+ aggregationhas not been observedyet and is fundamen-
(at the octahedral B site).24,25 The polar instability ob- tally different from the recently reported40 tendency of
the Mn2+ ions to segregate in grain boundaries.
servedinthecaseoftheA-sitedopingisduetooff-center
location of the much smaller Mn2+ ion with respect to
the Sr2+ ionandwasconfirmedbyexperiments26–28 and
theory.29 The polar freezing has been further argued to II. EXPERIMENTAL DETAILS
induce a spin-glass state for certain compositions, where
a magnetoelectric coupling should lead to a novel mag- Powdersamplesweresynthesizedaccordingtothesyn-
netoelectricmultiglass.16,17 Theexistenceofsuchastate thesis protocols thoroughly explained in Ref. 30. Our
has been recently hotly debated.30–32 As an alternative, study benefits from a systematic synthesis approachem-
the frozen magnetism was suggested to be due to seg- ploying high-energy homogenization of starting reagents
2

andvariousdifferentsinteringconditions(oxygenpartial
pressures were varied by applying either O2 or N2 at- etry 0.2 (a) 1.6 K 30 K 70 K
mosphere in post treatment) thus allowing a production m
m 0.1
of high-quality powder samples of the cubic perovskite y
s
SrTiO , nominally doped on, both, A and B sites in the a 0.0
3 R
doping range from 1 to 3%. For each doping concentra- S
-0.1
tionandachosendopingsite,theO andN post-treated
2 2
-0.2
samples were obtained from the same parentcompound.
X-ray diffraction patterns showed no indication of sec-
o3n%d-adroypepdhassaemspinlesanuyseodfitnheouinrvsetsutdigyatwederesatmheplesas.meThaes metry0.2 (b) 13.06 KK 7152 0K K
m
in Ref. 30. In the following, our samples are labeled by y
s
a
XcY,whereX =A,Bdenotesthetwosites,cthedoping R "1/3"
S0.1
concentrationinpercent,andY =O ,N theposttreat-
2 2
mentatmosphere;e.g. A3O standsforthe3%nominally
2
A-site Mn-dopedsamplethatwaspost-treatedinthe O
2 0.0
atmosphere. LKT (10 mT)
y
ESMRa.gTnheetisµmSRondaatloacwalesrceacleolwleacsteindspoencttehdewGiethneµrSalRPaunrd- metr0.2
m
poseSurface-Muon(GPS)instrumentattheSwissMuon sy
Source(SµS),PaulScherrerInstitute(PSI),Switzerland. R a0.1 (c)
The measurements were performed in the weak trans- S 0 mT
2.5 mT 40 mT
verse field (wTF) of 3 mT, in zero field (ZF) and in
10 mT 160 mT LKT (0 mT)
various longitudinal fields (LF) applied along the initial 0.0
0 2 4 6 8
muonpolarizationinthe temperaturerange1.6 120K. t ( s)
−
The ESR measurements were performed between 5 and
FIG. 1: The temperature-dependent µSR asymmetry of the
300Kin the X band(at9.5 GHz)ona home-builtspec-
trometer. The DRSwasperformedatroomtemperature A3O2 samplein (a)wTF and(b)ZF.(c)Field decouplingof
the longitudinal µSR relaxation in various LFs at 1.6 K for
on the Lambda 650 UV-vis spectrophotometer (Perkin-
the same sample. The solid lines in panels (a), (b) and (c)
Elmer) equipped with a diffusion sphere.
correspondtofitstothemodels(4),(1)and(3),respectively.
The dashed line in (b) shows the expected position of the
"1/3-tail", and in (c) the prediction of the static Lorentzian
III. RESULTS Kubo-Toyabemodel.
A. Muon spin relaxation
Toassessthereportedlyintrinsicpresenceofthefrozen tryofmuondecaydetectedbyemittedpositrons–thatis
spin-glass state in the A-site doped SrTiO samples16,17 proportionaltoP(t)isplottedfortheA3O sample. The
3 2
on a local scale, we first resort to µSR, a technique wTF measurements [Fig. 1(a)] are generally accepted as
well known for its extreme sensitivity to local magnetic aneffectivetooltotracethevolumefractionofthefrozen
fields.41 The muons, that are initially almost100% spin- spins, as the internal fields are typically well above the
polarizedstopinthe sampleandexperiencealocalmag- applied field and thus cause a loss of the precession am-
netic field B , which leads to a time-dependent muon plitude. In diluted spin glasses though, the average in-
µ
polarizationP(t). Itsformreflectsamagnitude,distribu- ternal field may be of the order of only 0.1–1 mT, i.e.,
tion andfluctuations ofB anddepends ofthe magnetic much smaller than the applied wTF. However, the spin-
µ
coupling of the muon with its surroundings. If the field glasstransitionshouldstillmanifestinasharpincreaseof
is rapidly fluctuating, ν γ B , where ν denotes the the muon depolarization rate below the transition tem-
µ µ
≥
field fluctuation frequency and γ = 2π 135.5 MHz/T perature, because the on-set of random internal fields
µ
×
is the muon gyromagnetic ratio, P(t) will decay mono- broadens the distribution of local fields.44 Experimen-
tonically with time in ZF. A quasi-static local field, on tally, we observe a smooth temperature dependence of
the other hand, causes a coherent oscillation of the ZF the transverse relaxationrate λ in the wTF experiment
t
muonpolarizationthatisdampedbyslowmuondynam- (vide infra), with no notableirregularities(Fig.2)atthe
ics (ν γ B ) and by a field distribution. The result- reportedspin-glasstransitionofaround40K.16,17,31 The
µ µ
≪
ing depolarization remains nonmonotonic even for ran- observed damping of the wTF signal can in principle be
domly oriented frozen dense (e.g, pertinent to nuclear either due to dephasing in an inhomogeneous static field
magnetism42) or diluted magnetic moments (e.g, arising or due to spin dynamics. These two effects can usually
from a dilute-alloy spin glass43). beseparatedintheZFexperiment. InZF,thedynamical
In Fig. 1, the µSR asymmetry G(t) – spatialasymme- fieldsinadilutedmagneticsystemleadtothemonotonic
3

TABLE I: XANES-derived occupancy of the Mn2+ (Mn4+) A3N2
ions on the A (B) sites in SrTiO3,30 the width of the field l A3O2
distribution at the muon site a/γµ and the field fluctuation 0.3 B3O2
frequency ν for thethreeinvestigated samples.
sample A-site(Mn2+) B-site (Mn4+) a/γµ ν -1 (s) 0.2 t A3O2
(%) (%) (mT) (MHz)
A3N2 82 18 0.67(5) 4.4(2)
0.1
A3O2 36 64 0.75(10) 9.1(5)
B3N2 13 87
B3O2 9 91 0.85(10) 18.3(4) 0.0
0 20 40 60 80 100 120
T (K)
FIG. 2: The temperature dependence of the longitudinal λ
l
"root-exponential" relaxation43 andtransverseλtmuonrelaxationratesinMn-dopedSrTiO3.
Solid lines are a guide to theeye.
GZF(t,ν)=G0e−√4a2t/ν =G0e−√λlt, (1)
where G denotes the initial asymmetry and a/γ the 2∆2ν t
0 µ Here, Gd(t,∆,ν) = e−ν2+(γµBLF)2 is the dynamical re-
field-distribution width. Static internal fields, on the
laxation function for a field distribution with the width
otherhand,inspinglassesleadtoanonmonotonicsingle-
∆/γ atagivenmuonsite. Inadilutedspinsystem∆is
µ
dip relaxation function
further distributed, which is typically modeled with the
GLKT(t)=G 1 + 2(1 at)e at , (2) Gaussianfunction ρ(∆,a)= π2∆a2e−a2/2∆2.43 A simul-
ZF 0 3 3 − − taneousfitofvariousLFdataqsetsbetween0and320mT
(cid:20) (cid:21)
[Fig. 1(c)] to the model (3) yields the field-distribution
knownastheLorentzianKubo-Toyabe(LKT)function.43 widtha/γ =0.75(10)mTandthefluctuationfrequency
µ
In the A3O sample, both models give a similar quality ν = 9.1(5) MHz in the A3O system at 1.6 K. We note
2 2
of fit at 1.6 K [see ZF data in Fig. 1(c)], with either the though that the modeling of the decoupling experiment
dynamical longitudinal relaxation rate λl = 0.30(1) s−1 does not perfectly match the experimental data at inter-
[model (1)] or the static field-distribution width a/γ = mediate fields [see the 40-mT curve in Fig. 1(c)], which
µ
0.32(2) mT [model (2)]. We note that in the latter case suggeststhat there are regionsin the sample with differ-
the dip in GLKT(t) falls outside the experimental time ent ν. For instance, a region with larger ν would exhibit
ZF
window, therefore, its existence and the leveling of the faster relaxation than modeled for a single fluctuation
dataat G0 atlongertimes thatischaracteristicforpow- rate.
3
der samples41 can not be verified. A similar, dynamical behavior was found also in the
In order to distinguish between the two scenarios, we other two investigated samples: A3N and B3O . The
2 2
turn to the LF experiment. In the static case, already corresponding field-distribution widths and fluctuation
the applied field of the order of a few times a/γ notice- frequencies at 1.6 K obtained from µSR relaxation in
µ
ably narrows the corresponding local-field distribution, various LFs are summarized in Tab. I. Both parameters
which results in a significantly decreased depolarization. are found to increase with the increasing occupancy of
For demonstration, we show in Fig. 1(c) the prediction the B site, despite the fact that the average magnetic-
of the static LKT model [a/γ =0.32(2) mT] in the ap- moment size decreases – Mn4+ (S = 3/2) on B sites vs.
µ
plied field43 of 10 mT, which should almost completely Mn2+ (S =5/2) on A sites.
removetherelaxation. Thisisevidentlyinconsistentwith The temperature dependence of the relaxation rates
theexperimentandthusunambiguouslyprovesthatlocal is shown in Fig. 2 for all three samples. The longitudi-
magnetic fields at the muon sites are dynamical. Since nal relaxation rates λ were obtained by fitting the ZF
l
magnetic fields from nuclei are generally regarded static data to the "root-exponential"decay [Eq. (1)], while the
on the muon time scale, this finding demonstrates that transverse relaxation rates λ correspond to fits of the
t
thegroundmagneticstateoftheA3O sampleisdynam- wTF data to the model
2
ical.
In the case of dynamical internal fields with a single GwTF(t)=G0cos(γµBwTFt+ϕ)e−√λtt, (4)
fluctuation rate ν, the µSR asymmetry is given in the
applied external field B by43 whereϕ denotesa smalltilt ofthe initialmuonpolariza-
LF
tionfromthebeamdirection. First,westressthatλ and
l
G (t,ν)= ∞G (t,∆,ν)ρ(∆,a)d∆. (3) λt are nearly equal at each temperature, revealing that
LF d the contribution of inhomogeneous static fields to λ is
Z0 t
4
1 (a) A3N2 (c) B3N2 (a) (b)
40
0 C3
3 T)30
nits) -1 nits) 300 K B (m20
ESR signal (arb. u --3201 (b) A3O2 (d) B3O2 R signal T (arb. u12 5112200505 0000K KKKK -1 mu/mol))10045 (cC)2 b C1
-1 C1 ES0 5 K 3 0 (e 23 3
C2 1
-2 (
C3 C1 C2 1/ 1
-3 -1 0
270 300 330 360 390 270 300 330 360 390 270 300 330 360 390 0 100 200 300
B (mT) B (mT) B (mT) T (K)
FIG.3: Theroom-temperatureESRspectraof3-%Mn-doped FIG. 4: Temperature-dependent ESR spectra of the B3O2
SrTiO3samples(circles)andtheirfitswithathree-component sample. Theindividualspectra are displaced for clarity. The
model(seetextfordetails). Thethreeindividualcomponents lowest-fieldpeaksofbothsextetsaremarkedbydashedlines.
are displaced vertically for clarity. (b) The temperature dependence of the ESR line widths of
the three components. (c) Comparison of the inverse ESR
susceptibility of the broad component 1/χ3 = 1/(I3χESR)
marginal. This again confirms the dynamical magnetic andbulksusceptibility 1/χb,measured in thefieldof 10mT.
Solid lines are linear fitsof thehigh-temperaturedata.
state in the A3O sample at 1.6 K. Second, a minimum
2
of both relaxation rates is observed around 75 K. In the
A3N sample, the temperature dependence of λ is very
2 l
similar, while in the B3O sample this minimum is shal- tense single-ion-anisotropy split ESR sextets were also
2
lower. The increase of the relaxationrates with decreas- observed and assigned to the Mn2+-VO centers46 and
ing temperature below 75 K is typical of slowing-down Mn3+-VOcenters46,47thatshouldlocallycompensatethe
of spin fluctuations at low temperatures, as λ = 4a2.43 excess charge of vacancies.
Above 75 K, on the other hand, the increase of thνe re- Theroom-temperatureX-bandESRspectraofallfour
laxation rates is rather unusual. 3%-doped samples are compared in Fig. 3. They can be
nicely fit to a sum of two Lorentzian-broadened sextets
(components C1 and C2) and a much broader C3 com-
ponent that also exhibits the Lorentzian line shape. We
B. Electron spin resonance
note that the Lorentzian shape can either result from
motional/exchangenarrowing48 ofthe absorptionline or
In order to obtain further insight into the dynamical from the Lorentzian distribution of the local fields in a
magnetic state of the Mn-doped SrTiO , we next turn random dilute system due to 1/r3-type interactions.49
3
to the ESR measurements. ESR has proven highly in- The position and the splitting of the observed sextets
formative in determining the Mn-ion valence and its lo- arecharacteristicoftheMn2+ ionsattheAsite(C1sex-
cal surroundings in SrTiO . The substitution of Mn4+ tet) and the Mn4+ ions at the B site (C2 sextet);45,46
3
for Ti4+ on the B site was observed by this technique see Tab. II. We stress that in neither of our samples did
long ago,45 while the successful Mn2+ for Sr2+ substi- we observe any Mn3+ resonances, which include an an-
tution on the A site has been claimed only recently.27 gular independent (-1 1) transition that is in X band
At high temperatures, both cases are characterized by expected around 0.6 T↔.47 Therefore, we attribute all the
distinctive hypefine-split ESR sextets, with the g fac- observed ESR components to the Mn2+ and Mn4+ ions.
tor g = 1.994 (2.004) and the hyperfine coupling a = Moreover,wenotethatourcalibrationoftheESRinten-
7.5 mT (8.8 mT) for Mn4+ (Mn2+).45,46 Such sextets sity with a standard sample revealed a nice correspon-
are consistent only with a cubic local crystal-field sym- dencebetweenthe ESR susceptibilityχ andthe bulk
ESR
metry and negligible exchange interactions. For the off- susceptibility χ , measured with a SQUID magnetome-
b
center position of the Mn2+ ions at the A sites, the ab- ter(seeTab.II).ThisagreementprovesthatESRdetects
sence of the expected single-ionmagnetic anisotropyis a all the spins in the Mn-doped SrTiO samples and that
3
consequence of fast ionic motion between energetically all three distinct components of the ESR spectrum are
equivalent sites.27 Moreover, in reduced B-site doped intrinsic. Moreover,we find a relatively good correspon-
SrTiO samples, Mn2+ was even argued to reside on dence (exceptforthe A3N sample)betweenthe relative
3 2
the B site,25,34 reflecting in the notably different hyper- intensity I and the occupancyof the Asites with Mn2+
1
fine coupling a = 8.4 mT and a significantly reduced g ions (see Tab. I). Therefore, the broad ESR component
factor.34 Finally,inreducedsinglecrystals,muchless in- is mainly attributed to the Mn4+ spins on the B site.
5
TABLEII:ParametersofthethreeESRcomponentsoftheMn-dopedSrTiO3 samples,theirroom-temperatureESRintensities
χESR and bulk susceptibilities χb measured by SQUIDin theapplied field of 10 mT.
Component 1 Component 2 Component 3
Sample I1 g1 a1 ∆B1 I2 g2 a2 ∆B2 I3 g3 a3 ∆B3 χESR χb
(%) (mT) (mT) (%) (mT) (mT) (%) (mT) (mT) (emu/mol) (emu/mol)
A3N2 26(5) 2.001 8.8 5.2(3) 1 1.978 7.9 3.5(5) 73(5) 1.998 / 32(2) 3.2(9)×10−4 2.6×10−4
A3O2 27(5) 2.000 8.7 6.0(3) 1 1.982 8.0 3.5(5) 72(5) 1.998 / 43(2) 1.7(5)×10−4 2.3×10−4
B3N2 4(2) 2.002 8.9 4.5(3) 1 1.991 7.6 3.5(3) 95(2) 1.998 / 35(2) 1.8(5)×10−4 1.6×10−4
B3O2 3(2) 2.002 8.8 5.5(5) 1 1.990 7.7 3.5(4) 96(2) 1.997 / 28(2) 2.5(7)×10−4 2.1×10−4
The temperature evolution of the three ESR compo- the C3 component, on the other hand, is more or less
nentsisshowninFig.4fortheB3O sample,inwhichthe temperatureindependentdowntoaround50Kandthen
2
broad C3 component is the most intense. The C1 sextet increases notably down to 5 K. This line-width increase
disappearsbelow 120K,similarlyasobservedbefore.27 can be interpreted as a fingerprint of a slowing down
∼
This canbe attributedtofreezingofthe Mn2+ ionicmo- of spin fluctuations due to developing spin correlations.
tion between the six energetically equivalent off-center Once again, in accordance with the µSR results, we ob-
positions within the A-site dedecahedron, which results serve no anomalies in any of the ESR components (and
in a much larger magnetic anisotropy and, consequently, in any of the investigated samples) around 40 K, where
inamorecomplicatedanisotropy-splitESRpowderspec- the spin-glassstatewassuggestedto setinfor the A-site
trum below 120 K.27 A gradual slowing down of the doped samples.16,17,31 Our measurements thus provide a
∼
Mn2+ hopping with decreasing temperature reflects al- local-probe verification that such a spin state is not in-
readyabove120K in the increasingline width ofthe C1 trinsic and that the Mn-doped SrTiO samples rather
3
component. The C2 sextet, on the other hand, remains remain paramagnetic down to the lowest temperatures.
essentiallyunchangeddownto5K,withtheexceptionof In order to determine the origin of the broad ESR
its intensity that exhibits a Curie-type increase, typical component that is dominant in all samples (see I in
3
for free magnetic moments. The intensity of the broad Tab. II), we have performed a doping-dependent study
componentχ =I χ thatdominatesχ followsχ for all four different sample types. In Fig. 5, the ESR
3 3 ESR ESR b
[Fig. 4(c)]. They both yield a small antiferromagnetic spectra in the doping range c = 1 3% are shown for
−
Weiss temperature of θ = 15(5) K. The line-width of the families with the least intense (AcO ) and the most
− 2
intense (BcO ) broad component. In all cases we ob-
2
serve that the C3 linewidth notably and systematically
narrows with the increasing doping concentration, e.g.,
from 40(2) to 28(2) mT for the B1O and B3O sam-
4 (a) (b) 2 2
ples,respectively. Atthesametime,itsrelativeintensity
A1O2 B1O2 increasesonlyslightly(notethattheincreaseofthespec-
s) 3 trumheightwithdopingispredominantlyduetolinenar-
nit rowing, as the height is proportional to 1/∆B2), mostly
u
arb. 2 A2O2 B2O2 at the expense of the C2 sextet, while the relative in-
al ( tensity of the C1 sextet remains unchanged within the
gn experimental error bars.
si 1
R
S
E A3O2 B3O2
0
IV. DISCUSSION
-1 C1 C2 C1 C2 The intensity of the unstructured broad ESR compo-
300 330 360 390 300 330 360 390 nent C3 dominates the ESR spectra in all investigated
B (mT) B (mT) samples. Although such component with a Lorentzian
FIG. 5: Doping-dependent normalized ESR spectra of the line shape is regularly observed in Mn-doped SrTiO3,
SrTiO3 samples(solidlines) atroom temperature. Thespec- especially for B-site doping,25,33,34,50 its origin remains
tracorrespondingtodifferentconcentrationsaredisplacedfor unclear. Its alleged absence in some cases, e.g., as in
clarity. The thick dashed lines show the broad-line (C3 com- Ref. 51, should be taken with caution, given the fact
ponent)contributiontoeachspectrum,whilethethinvertical that due to its broadly distributed spectral weight with
lines indicate the lowest-field peaks of both sextets. respecttothemuchnarrowerindividuallineswithinsex-
6
tets, it can easily be overlooked. It was ascribed before stantial experimentally observed decrease of ∆B with
3
tothe presenceofthe segregatedMnTiO phase.33 Since increasing doping concentration. The latter would de-
3
this phase orders magnetically at T = 65 K, causing creasethe distance between magnetic moments andthus
N
severebroadeningoftheESRlineclosetoT anditsdis- increase the average dipolar interactions. In scenario
N
appearance below T ,52 while in our samples the width (iii) the magnetic anisotropy responsible for the width
N
of the broad ESR line changes gradually and smoothly of the C3 ESR component is induced by oxygen vacan-
down to 5 K, this can not be the case here. Next, in re- cies. A vacancy in the B-site oxygen octahedron in-
ducedSrTi Mn O samples,thebroadESRcompo- ducesalargeuniaxialsingle-ionanisotropythatwasesti-
1 x x 3 δ
nent was co−rrelated w−ith a ferromagnetic response, that matedasD =0.80KfortheMn2+-V centerinreduced
O
was either claimed intrinsic34 or extrinsic.51 Again, our SrTiO single crystals.46 Since k D/gµ ∆B 20, iso-
3 B B 3
∼
samples show no ferromagnetic characteristics. latedMn2+-V centers wouldyielda complex andmuch
O
Inorderto clarifyits originandthe correspondingrel- broader ESR spectrum.
evance for the magnetism of the Mn-doped SrTiO3, we The C3 ESR component thus rather speaks in fa-
have critically assessedpossible ESR broadening mecha- vor of the scenario (iv) where strongly interacting mag-
nisms. AstheisotropicexchangeinteractionJ commutes netic moments on the B site are responsible for the ob-
withtheSz operatorandthereforeyieldsadelta-function served behavior. In this case, large isotropic exchange
ESRresponse,onlyafinitemagneticanisotropygivesafi- (k J gµ ∆B) leads to exchange narrowing and to
B B
niteESRlinewidth.48 First,weconsiderthecaseofwell- a featu≫reless Lorentzian line shape with the line width
isolated magnetic species (J 0). The dipolar interac- ∆B M /k Jgµ .48 Considering the dipolar interac-
tionbetweenmagneticmomen∼tsisalwayspresent,there- tion ≃withi2n dBense cBlusters (taking into account also the
fore, we have calculated its contribution to the ESR line nonsecular terms that renormalize48 M by 10/3)we es-
2
width. In the slow modulation limit (kBJ gµB∆B, timate the required exchange interaction as J 3 K,
where kB, µB, and B0 are the Boltzman co≪nstant, the while for the single-ion anisotropy scenario the r∼equired
Bohr magneton, and the applied magnetic field, respec- exchange coupling is of the order J k D2/gµ ∆B =
B B 3
tively) the second moment of the absorption line arising 15 K. Indeed, the former prediction∼, yielding the mean-
from the secular part of the dipolar interaction,48 field Weiss temperature θ = nS(S + 1)J/3 20 K,
| | ∼
where n=6 denotes the number of nearest neighbors in
M = µ0 2 3S(S+1)(gµB)4 3cos2θjk −1 2 , a cubic lattice, is in excellent agreement with the exper-
2 (cid:16)4π(cid:17) 4 *Xk (cid:0) rj6k (cid:1) +j imIenntparlinvcailpulee,θt=he−si1z5a(b5l)eKm.agnetic exchange (J 3 K)
(5)
∼
yields the ESR line width ∆Bd √M /gµ . Here, the that is needed to account for the broad ESR compo-
2 B
≃ nent could arise from two distinct origins: either from
sum runs over all the neighboring sites of a chosen site
j at the distance r that are occupied by spins S, θ magnetic clusters, where sizable superexchange interac-
jk jk
is the angle that the r vector makes with the external tion between nearest-neighbor-occupiedB sites was the-
jk oretically predicted,33 or from a donor-impurity medi-
magnetic field, and ... denotes powder averaging. In
h ij ated interaction35 in the limit of diluted and homoge-
the case of diluted magnetic lattices, one can sum over
neously dispersed dopants. In order to further inspect
allthe lattice sitesandthe dilute-limitline widthis then
given by ∆Bd(c) = c ∆Bd, because M is an additive the latter scenario, which stems from oxygen vacancies
2
quantityforthe1/r3-ty·peinteractions.43Wefindtheline and has been recently claimed responsible for ferromag-
widths ∆Bd =205 mT and ∆Bd =134 mT in the limit netism of reduced Mn-doped SrTiO3 crystals,34 we have
A B
of either fully occupied A-sites with Mn2+ ions or B- performed DRS measurements, which can provide infor-
sites with Mn4+ ions, respectively. In the scenario (i) mation on in-gap states for band-gap insulators such as
of diluted dopants, these then yield ∆Bd(3%) = 6 mT SrTiO3, with the band gap of 3.2 eV.53 In Fig. 6 we
and ∆Bd(3%) = 4 mT, in disagreemenAt with the en- show the DRS spectra [F(R)E]1/2 as a function of the
B
incoming-light energy E, which are obtained from the
hanced width of the C3 component. The scenario (ii) of
enhanced dipolar interaction within Mn-rich clusters of reflectance (R) spectra of the samples by the Kubenka-
SrTiO3 was suggested before to explain the broad ESR Munk transformation F(R) = (1−2RR)2.54 Similarly as in
component.25 We note though that for dense magnetic Ref. 34, we observed a pronounced dopant-induced in-
clusters the ∆Bd = 205 mT and ∆Bd = 134 mT line gapstate ataround2meVthatisabsentinthe undoped
A B
widthsshouldbeobserved. Theseexceedtheexperimen- SrTiO sample.34 Thisstatecanbeinterpretedasadeep
3
tal ones by an order of magnitude. Moreover,the Gaus- vacancy level, which has been predicted recently to be
sian distribution of internal fields would be expected,48 occupied by a single electron due to a strong Coulomb
which is clearly incompatible with the experimental line repulsion.55 As such, in contrast to shallower donor-
shape. Alternatively, reduced line widths could occur electron states with much larger Bohr radius (typically
if the B sites within clusters were only partially oc- between5and10Å)observedinsomeotherband-gapin-
cupied; e.g., 22% magnetic-moment density on B sites sulatingoxides,35 thisin-gapstateiswelllocalized55 and
would yield the ESR line width of 30 mT. However, can not provide extended magnetic coupling in a low-
even partially occupied clusters can not explain the sub- doping limit. Moreover, the effective coupling mediated
7

the host matrix has been detected.5,38
4
(a) The apparent systematic mismatch between the rela-
1/2 E ] 3 tive ESR intensity of the C1 component I1 (see Tab. II)
F(R) 2 A1N2 atinodn tnheear-seitdegeocsctruupcatnucrye (dXetAeNrmEiSn)edspbeyctrxo-srcaoypya3b0so(srepe-
[ A1O2 Tab. I) indicates that also Mn2+ ions at the A sites ag-
1 B1N2 gregate, although only partially. Since the Mn[Mn]O
3
B1O2 cubic perovskite structure is unstable and a perovskite-
0
type structure stabilizes only at high pressures and high
4
(b) temperatures,56 it would be structurally highly unfavor-
1/2 E ] 3 able ifboth Mn2+ andMn4+ ions enteredthe same clus-
R) ters. Therefore, a formation of separate MnTiO3 and
[F( 2 AA22NO22 SforrMmneOr 3conmapnoouclnudsteinrsbiuslkexapdeoctpetds.thWe eilmnoenteitethcartystthael
1 B2N2 structure52 that is significantly different from the cubic
0 B2O2 perovskite structure of SrTiO3 as the Mn2+ ions reside
4 (c) in oxygen octahedra. SrMnO3, on the other hand, can
form a cubic perovskite polymorph.57 This then proba-
1/2 E ] 3 bly explains the different tendency of Mn2+ and Mn4+
R) 2 A3N2 to aggregate. Quite importantly, we point out that the
F( A3O2 percentageoftheaggregatedMn2+ ionsinSrTiO3 iscor-
[
1 B3N2 related to the sample quality. There is a considerably
B3O2 larger mismatch between I1 and the A-site occupancy,30
0 i.e.,higherlevelofMn2+ aggregationintheA3N thanin
2 3 4 2
theA3O sample,while,onthe otherhand,the presence
E (eV) 2
ofstructuraldefectsdeterminedbytransmissionelectron
FIG. 6: The room-temperature diffuse reflectance spectra of
microscopy(TEM) is muchlowerinA3N , asits micros-
2
(a) 1%, (b) 2% and (c) 3% Mn-dopedSrTiO3 samples. tucture is more uniform.30
Last, we return to the µSR results. Again, the homo-
geneouspictureofthedonor-electron-mediatedexchange
by the donor electrons is always ferromagnetic, because
can not explain the rather slow fluctuation rates of local
itisquadraticinthe dopant–to–donor-electronmagnetic
magnetic fields at the muon stopping sites (see Tab. I).
interaction. This contradicts our experimental observa-
The derived fluctuation rate ν = 20 MHz is orders of
tions that reveal antiferromagnetic interaction between
magnitudebelowtheexpectedrateforexchange-coupled
Mn ions contributing to the broad C3 ESR component.
moments, k J/h 60 GHz (J = 3 K). In the inho-
We arethereforeinclinedtothe scenarioofdopantag- B ∼
mogeneous picture of clusters, on the other hand, the
gregation,where the enhancedESR line width of the C3
volume fraction of such clusters will be very small for
componentcanbe explainedby aninterplayofthe dipo-
the low doping concentrations of our samples. There-
larandsuperexchangeinteractionswithin denseMn-rich
fore,the muons will predominantly couple to the diluted
cluster of SrTiO . With increasing the doping concen-
3 isolated moments outside the clusters, which fluctuate
tration, the fraction of the isolated Mn4+ ions on the B
due to much smaller dipolar interactions. A small frac-
sites decreases (see the C2 component in Fig. 5), as one
tion of muons that do stop inside or close to clusters is
would expect. Moreover, the decreasing line width with
likely responsible for the overestimationof the predicted
increasing doping suggests that the exchange-narrowing
LF decoupling at intermediate fields [Fig. 1(c)].
mechanism is becoming more effective. Although ESR
is not best suited for quantifying the cluster sizes, like
some other nano-characterization tools,5,38 this experi-
V. CONCLUSIONS
mental finding is a likely fingerprint of nanosized clus-
ters. Namely, for small clusters, an average exchange
coupling of a particular Mn4+ ion to its neighbors, be- In conclusion, our muon spin relaxation and electron
ing a scalar sum, is more affected by nonmagnetic Ti4+ spin resonance investigations of Mn-doped SrTiO have
3
neighbors than the averagedipolar interaction, resulting demonstrated unambiguously that this system, indepen-
fromavectorsum. Thiseffectwould,however,dieoutin dentofthenominaldopingsiteandpost-treatmentatmo-
bigger clusters where the percentage of the moments on sphere, intrinsically remains paramagnetic down to low
the cluster surface becomes small compared to the mo- temperatures. Both,µSRdepolarizationcurvesaswellas
ments residing inside the cluster. Thus, the scenario of ESR spectra failed to detect any irregularities in the A-
a bulk phase segregation can safely be dismissed. The site-dopedsamples at the presumedspin-glasstransition
situation rather resembles other diluted magnetic semi- temperature around 40 K. In addition to several recent
conductors, where nano-organization of dopants within reportssuggestinganextrinsic nature offerromagnetism
8
in many DMOs, our study, disproving the intrinsic spin servedinthe ESR studies ofthe Mn-dopedSrTiO pow-
3
freezing in the Mn-doped SrTiO materials, serves as a dersandcrystals. ThiscomponentemergesfromtheMn-
3
warning that the phase-segregation processes in not op- rich clusters of SrTiO and is due to exchange coupling
3
timally processed materials can lead to very faint com- and dipolar interactions between dopants.
positional inhomogeneities that are far below the limits
ofconventionalanalyticaltechniques. Moreover,wehave
discovered site-dependent aggregation of the dopants in
Acknowledgments
Mn-rich clusters within the SrTiO matrix. Quite im-
3
portantly, such spinodal decomposition is correlated to
the quality of the sample’s microstructure. Moreover, it This work has been supported by the Slovenian Re-
is less favorable for the Mn2+ ions that partially remain searchAgency Programs P1-0125,P2-0377 and Projects
isolated in diluted regions, as revealed by the remaining N2-0005,Z1-5443. TheµSRpartofthisworkisbasedon
ESRsextet. Finally,ourstudyhasalsounraveledtheori- experimentsperformedatthe SwissMuonSource(SµS),
ginofthedominantbroadESRcomponentcommonlyob- Paul Scherrer Institute, Villigen, Switzerland.
∗ Electronic address: [email protected] J. B 71, 407 (2009).
1 Y.Matsumoto, M. Murakami, T. Shono,T. Hasegawa, T. 18 A.Ohtomo and H. Y.Hwang, Nature 427, 423 (2004).
Fukumura,M.Kawasaki,P.Ahmet,T.Chikyow,S.Koshi- 19 W. Meevasana, P. D. C. King, R. H. He, S-K. Mo, M.
hara, and H.Koinuma, Science 291, 854 (2001). Hashimoto,A.Tamai,P.Songsiriritthigul,F.Baumberger,
2 P.Sharma,A.Gupta,K.V.Rao,F.J.Owens,R.Sharma, and Z-X.Shen,Nat. Mater. 10, 114 (2011).
R. Ahuja, J. M. Osorio Guillen, B. Johansson, and G. A. 20 A. Brinkman, M. Huijben, M. van Zalk, J. Huijben, U.
Gehring, Nat.Mater. 2, 673 (2003). Zeitler, J. C. Maan, W. G. van der Wiel, G. Rijnders,
3 S. B. Ogale, R. J. Choudhary,J. P. Buban,S. E. Lofland, D. H. A. Blank, and H. Hilgenkamp, Nat. Mater. 6, 493
S. R. Shinde, S. N. Kale, V. N. Kulkarni, J. Higgins, C. (2007).
Lanci, J. R. Simpson, N. D. Browning, S. Das Sarma, H. 21 L. Li, C. Richter, J. Mannhart, and R. C. Ashoori, Nat.
D. Drew, R. L. Greene, and T. Venkatesan, Phys. Rev. Phys.7, 762 (2011).
Lett. 91, 077205 (2003). 22 J. A. Bert, B. Kalisky, C. Bell, M. Kim, Y. Hikita, H. Y.
4 T. Dietl, Nat. Mater. 2, 646 (2003). Hwang, and K.A. Moler, Nat. Phys.7, 767 (2011).
5 T. Dietl, Nat. Mater. 9, 965 (2010). 23 A.Tkach,P.M.Vilarinho,andA.L.Kholkin,Appl.Phys.
6 Y. Yamada, K. Ueno, T. Fukumura, H. T. Yuan, H. Shi- Lett. 86, 172902 (2005).
motani, Y. Iwasa, L. Gu, S. Tsukimoto, Y. Ikuhara, and 24 A.Tkach,P.M.Vilarinho,andA.L.Kholkin,ActaMater.
M. Kawasaki, Science332, 1065 (2011). 53, 5061 (2005).
7 G. Chen, C. Song, C. Chen, S. Gao, F. Zeng, and F. Pan, 25 D. Choudhury, S. Mukherjee, P. Mandal, A. Sundaresan,
Adv.Mater. 24, 3515 (2012). U.V.Waghmare,S.Bhattacharjee, R.Mathieu,P.Lazor,
8 X.Yang,Y.Li,C.Shen,B.Si,Y.Sun,Q.Tao,G.Cao,Z. O. Eriksson, B. Sanyal, P. Nordblad, A. Sharma, S. V.
Xu,and F. Zhang, Appl.Phys.Lett. 103, 022410 (2013). Bhat,O.Karis,andD.D.Sarma,Phys.Rev.B84,125124
9 J. M. D. Coey and S. A. Chambers, MRS Bull. 33, 1053 (2011).
(2008). 26 A.Tkach,P.M.Vilarinho,andA.L.Kholkin,ActaMater.
10 N. Izyumskaya, Ya. Alivov, and H. Morkoç, Crit. Rev. 54, 5385 (2006).
Solid State Mater. Sci., 34, 89 (2009). 27 V. V. Laguta, I. V. Kondakova, I. P. Bykov,
11 M.A.Garcia, E.FernandezPinel,J.delaVenta,A.Que- M. D. Glinchuk, A. Tkach, P. M. Vilarinho, and L. Jas-
sada, V. Bouzas, J. F. Fernández, J. J. Romero, M. S. trabik, Phys.Rev.B 76, 054104 (2007).
Martín González, and J. L. Costa-Krämer, J. Appl.Phys, 28 I.Levin,V.Krayzman,J.C.Woicik,A.Tkach,andP.M.
105, 013925 (2009). Vilarinho, Appl.Phys.Lett. 96, 172902 (2010).
12 M. Valant, T. Kolodiazhnyi, A.-K. Axelsson, G. S. Babu, 29 I. V. Kondakova, R. O. Kuzian, L. Raymond, R. Hayn,
N. M. Alford, Chem. Mater. 22, 1952 (2010). and V. V.Laguta, Phys.Rev.B 79, 134117 (2009).
13 K. A. Müller and H. Burkard, Phys. Rev. B 19, 3593 30 M. Valant, T. Kolodiazhnyi, I. Arčon, F. Aguesse, A.-K.
(1979). Axelsson, and N. M. Alford, Adv.Funct. Mater. 22, 2114
14 J.F.Schooley,W.R.Hosler,andM.L.Cohen,Phys.Rev. (2012).
Lett. 12, 474 (1964). 31 A. Tkach, P. M. Vilarinho, W. Kleemann, V. V. Shvarts-
15 N.Reyren,S.Thiel,A.D.Caviglia,L.FittingKourkoutis, man, P. Borisov, and S. Bedanta, Adv. Funct. Mater. 23,
G. Hammer, C. Richter, C. W. Schneider, T. Kopp, A.- 2229 (2013).
S. Rüetschi, D. Jaccard, M. Gabay, D. A. Muller, J.-M. 32 M. Valant, T. Kolodiazhnyi, I. Arčon, F. Aguesse, A.-K.
Triscone, and J. Mannhart, Science317, 1196 (2007). Axelsson, and N. M. Alford, Adv.Funct. Mater. 23, 2231
16 V. V. Shvartsman, S. Bedanta, P. Borisov, W. Kleemann, (2013).
A. Tkach, and P. M. Vilarinho, Phys. Rev. Lett. 101, 33 R.O.Kuzian,V.V.Laguta,A.-M.Daré,I.V. Kondakova,
165704 (2008). M.Marysko,L.Raymond,E.P.Garmash,V.N.Pavlikov,
17 W. Kleemann, S. Bedanta, P. Borisov, V. V. Shvartsman, A. Tkach, P. M. Vilarinho, and R. Hayn, Europhys. Lett.
S.Miga,J.Dec,A.Tkach,andP.M.Vilarinho,Eur.Phys. 92, 17007 (2010).
9
34 S. Middey, C. Meneghini, and S. Ray, Appl. Phys. Lett. R.W. Collins, Phys.Rev. B16, 4761 (1977).
101, 042406 (2012). 47 D. V. Azamat, A. Dejneka, J. Lancok, V. A. Trepakov,
35 J. M. D. Coey, M. Venkatesan, and C. B. S. Fitzgerald, L. Jastrabik, and A. G. Badalyan, J. Appl. Phys. 111,
Nat. Mater. 4, 173 (2005). 104119 (2012).
36 G.Martínez-Criado, A.Somogyi,S.Ramos,J.Campo, R. 48 A. Bencini and D. Gatteschi EPR of Exchange Coupled
Tucoulou,M.Salome,J.Susini,M.Hermann,M.Eickhoff, Systems (Springer-Verlag, Berlin, 1990).
and M. Stutzmann,Appl.Phys.Lett. 86, 131927 (2005). 49 C. Helm and M. W. Klein, Phys. Rev. Lett. 35, 1783
37 S.Kuroda,N.Nishizawa,K.Takita,M.Mitome,Y.Bando, (1975).
K. Osuch,and T. Dietl, Nat. Mater. 6, 440 (2007). 50 C. B. Azzoni, M. C. Mozzati, A. Paleari, V. Massarotti,
38 A. Bonanni, Semicond. Sci. Technol. 22, R41 (2007). M. Bini, and D. Capsoni, Solid State Commun. 114, 617
39 A. Bonanni, A. Navarro-Quezada, Tian Li, M. Wegschei- (2000).
der, Z. Matěj, V. Holý, R. T. Lechner, G. Bauer, M. 51 D. Choudhury, B. Pal, A. Sharma, S. V. Bhat, and D. D.
Rovezzi, F. D’Acapito, M. Kiecana, M. Sawicki, and T. Sarma, Sci. Rep. 3, 1433 (2013).
Dietl, Phys. Rev.Lett. 101, 135502 (2008). 52 J. J. Stickler, S. Kern, A. Wold, and G. S. Heller, Phys.
40 H. Yang, P. G. Kotula, Y. Sato, M. Chi, Y. Ikuhara, and Rev.164, 765 (1967).
N. D.Browning, Mater. Res. Lett. 2, 16 (2014). 53 K. van Benthem, C. Elsässer, and R. H. French, J. Appl.
41 A. Yaouanc and P. Dalmas de Réotier Muon Spin Rota- Phys.90, 6156 (2001).
tion, Relaxation and Resonance (Oxford University Press, 54 P. Kubenka,J. Opt.Soc. Am.38, 448 (1948).
Oxford, 2011). 55 C. Lin and A. A. Demkov, Phys. Rev. Lett. 111, 217601
42 R. S. Hayano, Y. J. Uemura, J. Imazato, N. Nishida, T. (2013).
Yamazaki, and R.Kubo, Phys.Rev.B 20, 850 (1979). 56 S. V. Ovsyannikov, A. M. Abakumov, A. A. Tsirlin, W.
43 Y. J. Uemura, T. Yamazaki, D. R. Harshman, M. Senba, Schnelle, R. Egoavil, J. Verbeeck, G. Van Tendeloo, K.
and E. J. Ansaldo, Phys. Rev.B 31, 546 (1985). V. Glazyrin, M. Hanfland, and L. Dubrovinsky, Angew.
44 D.E.Murnick,A.T.Fiory,andW.J.Kossler,Phys.Rev. Chem. Int.Ed. 52, 1494 (2013).
Lett. 36, 100 (1976). 57 R. Søndenå, S. Stølen, P. Ravindrano, and T. Grande,
45 K. A.Müller, Phys. Rev.Lett.2, 341 (1952). Phys.Rev. B 75, 214307 (2007).
46 R. A. Serway, W. Berlinger, K. A. Müller, and