Electronic Structure of Charge- and Spin-controlled Sr_{1-(x+y)}La_{x+y}Ti_{1-x}Cr_{x}O_{3}

0.28 MB·English

by  H. Iwasawa

Electronic Structure of Charge- and Spin-controlled Sr La Ti Cr O
1−(x+y) x+y 1−x x 3
H. Iwasawa,1,2 K. Yamakawa,1 T. Saitoh,1,∗ J. Inaba,3 T. Katsufuji,3,4
M. Higashiguchi,5 K. Shimada,6 H. Namatame,6 and M. Taniguchi5
1Department of Applied Physics, Tokyo University of Science, Shinjuku-ku, Tokyo 162-8601, Japan
2National Institute of Advanced Industrial Science and Technology, Tsukuba, Ibaraki 305-8568, Japan
3Department of Physics, Waseda University, Tokyo 169-8555, Japan
6 4PRESTO, Japan Science and Technology Corporation, Saitama 332-0012, Japan
0 5Graduate School of Science, Hiroshima University, Higashi-Hiroshima 739-8526, Japan
0 6Hiroshima Synchrotron Radiation Center, Hiroshima University, Higashi-Hiroshima 739-8526, Japan
2 (Dated: February 6, 2008)

n
a WepresenttheelectronicstructureofSr1−(x+y)Lax+yTi1−xCrxO3investigatedbyhigh-resolution
photoemissionspectroscopy. InthevicinityofFermilevel,itwasfoundthattheelectronicstructure
J
were composed of a Cr 3d local state with the t32g configuration and a Ti 3d itinerant state. The
1 energylevelsoftheseCrandTi3dstatesarewellinterpretedbythedifferenceofthecharge-transfer
1
energy of both ions. The spectral weight of the Cr 3d state is completely proportional to the spin

concentration x irrespective of the carrier concentration y, indicating that the spin density can be
]
l controlled byx asdesired. Incontrast, thespectral weight oftheTi3d state isnot proportional to
e y,dependingon theamount of Cr doping.
-
r
t PACSnumbers: 71.20.-b,79.60.-i,71.55.-i
s
.
t
a
m If the charge and spin degrees of freedom of electrons tration (ns and nc), respectively. In fact, magnetic and
canbecontrolledjustasonedesigned,itispossibletoen- transport measurements showed that the paramagnetic
-
d gineera next-generationofdevices mergingconventional Curie constant increases with x, and a semiconducting
n electronics with magnetoelectronics [1]. So far, how- resistivity decreases with increasing y [6]. Judging from
o
ever, there is no report that has succeeded in handling thesignchangeoftheWeisstemperature,thecorrelation
c
[ magnetic and semiconducting properties independently. between Cr ions should vary from antiferromagnetic to
Magnetic semiconductors (MSs) are one of the best can- ferromagnetic interaction with from y=0 to y>0. This
1
didates among the spin-electronic devices utilizing the ferromagetic interaction is probably caused by the ex-
v
both properties. Indeed, following a successful synthe- change interaction between the Cr 3d (spin) and Ti 3d
4
1 sis of the prototypical In1−xMnxAs and Ga1−xMnxAs (carrier) electrons, but the paramagnetic Curie temper-
2 [2, 3, 4], several room temperature (RT) ferromagnetic ature (T ) of this system is still low below ∼10 K unfor-
c
1 MSs such as ZrTe:Cr [5] have recently been discovered, tunately. Nevertheless, it is not evident that the doped
0 leading a continuous effort to understand the electronic Cr3+ really form the local spin of S=3/2 since both Ti
6
and magnetic properties of these materials. However, and Cr t orbitals will be partially filled. In this sense,
0 2g
/ it is still difficult to control their magnetic and semi- despite the low Tc, SLTCO is worthy to be investigated
t
a conducting properties independently, and this difficulty as a prototype of next-generation MSs in which the two
m stems from that the both properties depend upon only degrees of freedom can be controlled.
- theone-transition-metal(1-TM)elementdopingineither
d The primary focus of this Letter is to elucidate the
III-V-based or RT ferromagnetic MSs.
n electronic structure of SLTCO as well as a possibility
o of an independent control of charge and spin degrees of
Alternatively, a Cr-doped perovskite-type titanate
c
freedom by high-resolution photoemission spectroscopy
v: Sr1−(x+y)Lax+yTi1−xCrxO3 (SLTCO) is based on a new (PES). The result demonstrates that the near-E elec-
concept of independent control of the charge and spin F
Xi degrees of freedom by 2-TM elements (Ti and Cr); the tronic structure consist of the Cr 3d3 local state and Ti
3d itinerant state. We will show that the relative en-
r end compound SrTiO of this family is a band insula-
a 3 ergy levels of these 3d states can be described by the
tor with a wide band gap of 3.2 eV. It is well known
difference of the effective charge-transfer energy (∆ )
that carrier doping can be realized by La3+ substitution eff
of both ions, which is one of the basic parameters in
for Sr2+, which introduces electrons (y) into Ti 3d con-
theelectronicstructureofconventional1-TMperovskite-
duction band. A further manipulation of the resulting
type oxides[7]. We discuss thatanappropriatechoiceof
Sr La TiO is to introduce Cr3+ by replacing SrTiO
1−y y 3 3 two TM elements can realize a new magnetic system in
with LaCrO by the amount of x. Because Cr3+ (3d3)
3 which local spins and charge carriers are independently
ions are usually magnetic in oxides, the above substitu-
controlled.
tion of Cr3+ for Ti4+ would realize a “spin doping” of
S=3/2 local moment at the Ti sites. In other words, x PolycrystallinesamplesofSr La Ti Cr O
1−(x+y) x+y 1−x x 3
and y nominally represent “spin” and “carrier” concen- with (x, y)=(0.1, 0.1), (0.2, 0.1), (0.2, 0.2), and (0.2,
2
(a) 50 eV (b) 50 eV To seehow the chargeandspindegreesoffreedomare
45.6 eV 45.6 eV controlled, we have investigated a doping dependence of
D PES spectra. Panel (a) of Fig. 2 presents the valence-
C
b. units) b. units) batan1d50speeVc.trAatofifrSstLTsiCghOt,ftohresfeovuerraslpceocmtraplosfeitaitounrsestaAke−n
y (ar y (ar D identified in Fig. 1 (a) can be observed. Owing to
nsit nsit the largerphotoionizationcross section of the Ti and Cr
nte nte 3d states relative to one of the O 2p state, the near-EF
I I B structures are much clearer and sharper than in Fig. 1
(a). An apparent doping dependence of the features B
A'
B A A and A can be seen easily in the near-E [panel (b)] and
F
-10 -8 -6 -4 -2 0 -3 -2 -1 0 the very near-E [panel (c)] region spectra, respectively.
Binding Energy (eV) Binding Energy (eV) F
Panels (d) and (e) summarize this doping dependence,
where the integration windows are from -3.3 to -2.1 eV
(B) and from -0.4 to 0.1 eV (A).
FIG. 1: (Color online). Resonant PES spectra of (0.2, 0.15)
in (a) the full valence-band and (b) the near-E region. The WecanfindtwoimportantfactsaboutCrdopingfrom
F
spectraare takenat 45.6 eV (dashedline) and 50.0 eV(solid thefigures. First,panels(a)and(b)clearlyshowthatthe
line),whichcorrespondtotheTiandCr3p−3donresonance, feature B is located at a deep level far below E (∼2.7
F
respectively.
eV). This is demonstrating that the doped Cr 3d elec-
trons do not participate in electric conduction. Second,
it is likely that the doped Cr ions are always trivalent
0.15) were prepared by solid-state reaction [6]. We will with the t3 localconfigurationbecause the peak shift of
2g
simply denote them as (0.1, 0.1), etc. All the measure- the Cr 3d and O 2p bands due to the doping of both x
mentswerecarriedoutatBL-1ofHiroshimaSynchrotron and y is not observed in the spectra. This is contrary
Radiation Center (HSRC), Hiroshima University using to the previous PES spectra of La Sr CrO , in which
1−x x 3
a SCIENTA ESCA200 electron analyzer [8, 9]. In or- the Cr3dpeakandO2pbandsshowsystematicchanges
der to obtain clean surface, we fractured the samples in with increasing Cr4+ localized state by replacing La3+
situ in ultrahigh vacuum (better than 1×10−10 Torr) at for Sr2+ [13]. Here it may be still possible that they be-
50 K. The spectra were taken at 50 K using selected come all tetravalent, although such a situation is very
photon-energies (45.6, 50.0, and 150 eV), and the over- unlikely. However, this can be ruled out from the result
all experimental energy resolution was set at 50 meV. of panels (d) and (e); panel (d) shows that the spectral
The backgrounds due to the unscattered electrons were weight of Cr 3d state is completely proportional to x ir-
subtracted from all the spectra. For comparison, a spec- respectiveofy,indicatingthattheCrionsdonotchange
trum of Sr1−(x+y)Lax+yTi1−xVxO3 (SLTVO) with (0.2, their valence. Panel (e) also shows that spectral weight
0.2)wasalsorecordedunder the same experimentalcon- of coherent part has a linear relationship with y [except
ditions as above. for (0.1, 0.1), which will be discussed later]. If the Cr
In order to identify the Ti and Cr 3d states, we have ionsaretetravalent,(x,y)=(0.2,0.1)and(0.2,0.2)mean
first performed a resonant PES measurement for (0.2, nc=0.2+0.1 and 0.2+0.2, respectively. This is, however,
0.15). Figure 1 shows the resonant PES spectra of (0.2, resulting in the spectral weightratio of0.4/0.3≈1.33be-
0.15) taken at 45.6 eV (dashed line) and 50.0 eV (solid tween the two samples, which is far from the observed
line), corresponding to the Ti and Cr 3p−3d on reso- linear relationship. Therefore,we believe that the doped
nance, respectively. In Fig. 1 (a), one can observe the CrionsformtheS=3/2(t3 )local-spinstateandxgives
2g
four spectral features denoted as A to D. By compar- ns of this local spin, althougha more precise determina-
ison with the reported PES spectra of La Sr TiO tion would be done using core-level PES.
x 1−x 3
[10, 11, 12], the leading feature A (-0.7 eV) is attributed On the other hand, panel (e) shows that the spec-
to the Ti 3d state, and the dominant features C (-5.1 tralweightof(0.1,0.1)deviatesfromthe linearrelation-
eV) and D (-6.9 eV) have primarily the O 2p character. ship. This indicates that the charge carrier is not con-
Consequently, the remaining feature B (-2.7 eV) is the trolled solely by y. In other words, the density of states
mostlikelyCr3dstate. Theseassignmentsareconfirmed near E decreases with increasing Cr doping despite the
F
by the resonantenhancementof the features A andB at same carrier concentration. This is in perfectly agree-
each absorption threshold as shown in Fig. 1 (b). Note ment with the electrical resistivity data [6]. Moreover,
that the Ti 3d state further split into the quasiparticle thepeakpositionofthefeatureAapparentlyshiftsaway
band and the remnant of the lower Hubbard band cor- from E by ∼50 meV upon Cr doping without changing
F
responding to the one-electron excitation from d1 to d0 y as shown in panel (c). This is, hence, not a rigid-
of the insulating LaTiO , namely, the coherent (A) and band shift and what is observed is actually the shift of
3
incoherent(A′) parts,consistent with the literature [12]. the midpoint of the leading edge and a strong spectral
3
(a) h νT == 15500 K eV (b) (((000...221,,, 000...211))) (c) arb. units)32..00 (d)
D C (0.2, 0.2) Bof (
a 1.0
nits) eak are (((000...221,,, 000...211)))
b. u (0.2, 0.1) P 00 0.1 0.2 0.3
y (ar Spin concentration (x)
Intensit B ((00..22A,, 00..12)) CrB 3d A'Ti 3dA A Aof (arb. units)32..00 (e)
a 1.0
(0.1, 0.1) (0.1, 0.1) k are ((00..22,, 00..21))
ea (0.1, 0.1)
P 0
-10 -8 -6 -4 -2 0 -4 -3 -2 -1 0 -0.5-0.4-0.3-0.2-0.1 0 0.1 0 0.1 0.2 0.3
Binding Energy (eV) Carrier concentration (y)
FIG. 2: (Color). (a) Valence-band spectra taken at 150 eV. (b) Same as (a) in the near-E region in an expanded scale. (c)
F
Very near-E region spectra. (d) and (e) Intensity plots of the features B and A, respectively. Spectral weight of features B
F
and A is set to unityat (0.1, 0.1) and (0.2, 0.1). The integration windows are given in thetext.
weight suppression, both of which are signals of opening
a pseudogap. However, since the lineshape of the fea- (a) Sr0.6La0.4Ti0.8M0.2O3 LaxSr1-xTiO3 (b)
(M=Cr, V)
Ti 3d
ture A is still quasiparticle-like, the charge carriers are
nvioctinciotymopflettheelymleotcaal-liinzesdulaatnodrttrhaenssiytisotnem. Uinsdveerrsyucinhcthire- units) D C La1-xSrxCrO3 E
cduommsintaanncteasn,tdhieflaonsgu-ffiracniegnetCpooutleonmtibalindtiesorardcteironexbisetcso,mthees y (arb. SLTCO E
well-knownCoulombgap will appear [14, 15, 16]. In our ensit M =
nt Cr B
case, the Cr doping x should give rise to a strong poten- I E
A SLTVO Cr 3d
tialdisorderatTisiteswithasmallnumberofthecharge
carriersy. Hence we believe that this pseudogapis most V B'
likely the Coulomb gap. Further investigation is needed E
-10 -8 -6 -4 -2 0 O 2p V 3d E
on this issue, but it is beyond the scope of this Letter. Binding Energy (eV) F
ctoerfmuFtchisin.eaalPCllrayron,-leewalnie(ndad)tVihose-cfduFeoslipsegce.htd3orowtsnihti∆aocnwesaffsttretuohsfcetSTuvrr0Mae.l6eLoneaflce0etm.-h4bTeeanin20t-d.s8TMspMpla0e.yc2sstyOrsaa3- Energy (eV)-1012 (cT)iO4+ffset by 4.0 MMM334+++ (((RRReeefff... 111887))) 8765eff (e∆ SLTCOLoCcar l3 sdpinCTair r3ider (dE)
(M=Cr, V). Comparing the two spectra, one can easily ng -2 Ti3+ 4 V) SrFeMoO Mo 4d
attributethefeatureB′(-1.8eV)totheV3dstate. Panel ndi V3+ 2 6 Carrier
Bi-3 This work Cr3+ 3
(b) displays a schematic electronic structure of the four
-4 2 E
compounds deduced from our results and the literature 0 1 2 3 Fe 3d
[11, 12, 13]. The distance between the Cr 3d state and Number of 3d electrons Local spin E
F
the O 2p band in SLTCO is virtually identical to that
in La Sr CrO [13]. On the other hand, the Cr 3d
1−x x 3
FIG. 3: (Color). (a) Valence-band spectra of
peak in SLTCO is located at a deeper energy level than
in La Sr CrO because the E location is determined Sr0.6La0.4Ti0.8M0.2O3 (M=Cr, V) taken at 150 eV. (b)
1−x x 3 F and (d) Schematic electronic structure of SLTCO compared
by Ti 3d state. In other words, the combination of (1) with the analogues and the double perovskite Sr FeMoO .
2 6
the difference of the ∆eff’s (D∆eff) of the Ti and Cr 3d (c) Bindingenergy of spectralfeatures and ∆eff’s areplotted
states and (2) the E location determined by the Ti 3d against the number of 3d electrons, where ∆ for M4+
F eff
staterealizesthe Cr3dlocalizedstate (andhence the lo- includes an offset by 4.0 eV. Since thecentroid of A must be
calspins)inthebandgapbetweenO2pandTi3dbands above EF, the upper error bar of the Ti4+ is expected to be
large.
in SLTCO.
4
This is quantitatively demonstrated in panel (c), in of Sr La Ti Cr O by high-resolution pho-
1−(x+y) x+y 1−x x 3
which the binding energy of the observed spectral fea- toemission spectroscopy. Near-E electronic structures
F
tures is compared with ∆ ’s deduced from the analysis show two-component structure of the Cr 3d local state
eff
of PES spectra for M3+ [17] and from the empirical re- andTi3ditinerantstate. OwingtothelargeD∆ ofCr
eff
lation for M3+ and M4+ [18]. The panel revealsthat an andTiions,thedopedCr3delectronsrealizeS=3/2(t3 )
2g
appropriate offset for ∆ for M3+ or M4+ reproduces local-spin state at a deep level far below E (∼2.7 eV).
eff F
the relative energy of these features very well [19]. As a An appropriatechoice of 2-TMelements providesan op-
consequence, the panel predicts that V ions in SLTVO portunitytocreateanewmagneticsysteminwhichlocal
is not V4+ but V3+ because the B’ just locates between spins and charge carriers are independently controlled.
Ti3+ and Cr3+. It should be noted that the electronic We wouldliketo thank S.Kaneyoshifor usefuldiscus-
structure is not solely determined by D∆eff or ∆eff but sions. ThisworkwaspartlysupportedbyaGrant-in-Aid
also by the effective O 2p-TM d hybridization (Teff) and for COE Research(13CE2002)by MEXT of Japan. The
theeffectiveon-sited-dCoulombinteractionUeff [20]. In- synchrotron radiation experiments have been done un-
deed, (∆eff-Ueff) is observed directly in the PES spectra der the approval of HSRC (Proposal Nos. 03-A-58 and
instead of ∆eff. However, the effects of Ueff and Teff do 04-A-39).
not appear in the energy difference among the features
becauseU andT changesystematicallywithnumber
eff eff
ofdelectrons[21]. Thisiswhytherelativebindingenergy
of the features directly reflects D∆ . Therefore, D∆
eff eff
∗
is an essential parameter to determine the backbone of Authorto whom correspondence should be addressed.
Electronic address: [email protected]
the electronic structure of 2-TM compounds.
Owing to the large (∼2.7 eV) D∆ , SLTCO has lo- [1] S. A. Wolf et al.,Science 294, 1488 (2001).
eff [2] H. Munekataet al.,Phys.Rev. Lett.63, 1849 (1989).
cal spins and charge carriers. This indicates that one [3] H. Ohno,Science, 281, 951 (1998).
can introduce a localized spin at a deep energy level [4] Y. Ohnoet al.,Nature 402, 790 (1999).
by an appropriate choice of two elements with a rela- [5] H. Saito, V. Zayets, S. Yamagata, and K. Ando, Phys.
tively large D∆ . In addition, such a situation should Rev. Lett.90, 207202 (2003).
eff
be realized not only in the above randomly-substituted [6] J. Inaba and T. Katsufuji, Phys. Rev. B 72, 052408
2-TM compounds, but also in the ordered 2-TM com- (2005).
[7] J. Zaanen, G. A.Sawatzky,and J. W. Allen,Phys. Rev.
pounds, namely, the ordered double perovskite-type ox-
Lett. 55, 418 (1985).
ides. Panel (d) compares the schematic electronic struc-
[8] K.Shimadaet al.,Nucl.Instrum.MethodsPhys.Res.A
ture of SLTCO and a typical half-metallic double per- 467-468, 504 (2001).
ovskite Sr2FeMoO6 [22]. In the case of Sr2FeMoO6, the [9] K. Shimada et al.,Surf. Rev.Lett. 9, 529 (2002).
locationofE is determined by the Mo 4d(+Fe 3d) t [10] A. Fujimori et al.,Phys. Rev.B 46, R9841 (1992).
F 2g↓
stateandtheFe3d5t3 e2 statearelocatedbetweenthe [11] Y. Aiura et al.,Surf. Sci. 515, 61 (2002)
O 2p and Mo 4d state2sg,↑reg↑flecting a large D∆ between [12] T. Yoshidaet al.,Europhys.Lett. 59, 258 (2002).
eff [13] K.MaitiandD.D.Sarma,Phys.Rev.B54,7816(1996).
Fe3+ and Mo5+. That is to say, the Ti and Cr 3d states
[14] A.L.EfrosandB.I.Shklovskii,J.Phys.C8,L49(1975).
correspondtotheFe3dandMo4dstates,althoughthere [15] J. H. Davies and J. R. Franz, Phys. Rev. Lett. 57, 475,
existsthe essentialdifferencebetweenthem, n -to-n (x-
s c (1986).
to-y)ratioandtherandomness. Thiscomparisonfurther [16] D. D. Sarma et al.,Phys. Rev.Lett. 80, 4004, (1998).
implies that SLTCO could be double-exchange type fer- [17] T.Saitoh,A.E.Bocquet,T.Mizokawa,andA.Fujimori,
romagnet like La Sr MnO [23]. Phys. Rev.B 52, 7934 (1995).
1−x x 3
Although it is simplified, our interpretation employ- [18] A. Fujimori et al., J. Electron Spectrosc. Relat. Phe-
nomen. 62, 141 (1993).
ing D∆ givesa usefulview forunderstanding the elec-
eff [19] Because the electron doping yields the N- and (N+1)-
tronicstructureof2-TMsystemsaswellasnew-material
electron degenerated ground state, we can compare
designing. Forinstance,itisinterestingtoapplyittothe D∆ of the N-electron system with the (N-1) photoe-
eff
electronic structure of interface layers, which has hardly mission final states.
beenstudiedbyPESinspiteofitsimportanceforindus- [20] TheHund’sruleenergystabilizationalsoplaysanimpor-
trial applications [24]. In the heterointerfaces, the elec- tant role in the compounds which have very large local
tronic structure should be considered as quasi-two-TM spins.
[21] Not only ∆ but also U must be taken into account
system, and hence it is essential to take account of the eff eff
forthed5 configurationbecausebothbecomelargestdue
∆ ’s of the two TM ions in each thin film. This would
eff to the large positive multiplet correction.
give essential information about the electronic structure [22] T. Saitoh et al.,Phys. Rev.B 66, 035112 (2002).
of heterointerfaces,particularly on the amount of charge [23] For review, see Colossal Magetoresistive Oxides, edited
transfer from one to the other [24]. by Y. Tokura(Gordon and Breach, New York,2000).
Inconclusion,wehavestudiedthe electronicstructure [24] H.Kumigashiraetal.,Appl.Phys.Lett.84,5353(2004).