Dependence of the electronic structure of SrRuO
3
and its degree of correlation
on cation off-stoichiometry
Wolter Siemons,
1,2
Gertjan Koster,
1,
*
Arturas Vailionis,
1
Hideki Yamamoto,
1,3
Dave H. A. Blank,
2
and Malcolm R. Beasley
1
1
Geballe Laboratory for Advanced Materials, Stanford University, Stanford, California 94305, USA
2
Faculty of Science and Technology and MESA
⫹
Institute for Nanotechnology, University of Twente, P.O. Box 217,
7500 AE Enschede, The Netherlands
3
NTT Basic Research Laboratories, 3-1 Wakamiya Morinosato, Atsugi-shi, Kanagawa 243-0198, Japan
共Received 2 February 2007; revised manuscript received 21 May 2007; published 24 August 2007
兲
We have grown and studied high quality SrRuO
3
films grown by molecular beam epitaxy as well as pulsed
laser deposition. By changing the oxygen activity during deposition, we were able to make SrRuO
3
samples
that were stoichiometric 共low oxygen activity兲 or with ruthenium vacancies 共high oxygen activity兲. This ability
to control the ruthenium stoichiometry has permitted us to make a systematic study of the dependence of the
degree of electron correlation in SrRuO
3
on stoichiometry using transport and photoemission experiments. We
have compared the measured ultraviolet photoemission spectroscopy spectra with calculated density of states
spectra and offer explanations for the large observed differences between the two.
DOI: 10.1103/PhysRevB.76.075126 PACS number共s兲: 73.50.⫺h
INTRODUCTION
SrRuO
3
is a material of uncommon scientific interest. It is
unusual among the correlated complex oxides in being a me-
tallic itinerant ferromagnet. It exhibits so-called bad metal
behavior at high temperatures, but is clearly a Fermi liquid at
low temperatures. It is also of widespread utility as a con-
ducting electrode in many devices utilizing complex oxides
of various kinds. One of the important open questions in the
physics of SrRuO
3
is its degree of electron correlation and
how that correlation affects the physical properties of the
material. A related question is what factors control the degree
of correlation. Recently Kim et al.
1
have shown using pho-
toemission data how the degree of correlation varies system-
atically in the ruthenate family of complex oxides as a
whole. Toyota et al.,
2,3
using transport and photoemission
measurements, have studied the degree of correlation as a
function of SrRuO
3
film thickness 共at the few monolayer
level兲 and shown that a correlation-driven insulator-to-metal
transition occurs as film thickness increases. In this paper, we
demonstrate that the degree of correlation depends on sto-
ichiometry and show how this dependence correlates with
various physical properties of the material.
At the same time, our results solve the heretofore mystery
of why the transport properties of SrRuO
3
are so sensitive to
how it is synthesized, particularly in thin film form.
4
More
specifically, we show that this sensitivity is not a simple
matter of defect scattering but, rather, once again is related to
stoichiometry and the degree of correlation.
The properties observed in the ruthenate family of com-
plex oxides range from very good metals 共 RuO
2
兲 to insula-
tors 共Y
2
Ru
2
O
7
兲. Recently, Kim et al.
1
suggested that this
change in electronic behavior can be attributed to a change in
electron-electron correlation. They come to this conclusion
based on fits of experimental core-level photoemission spec-
tra 关x-ray photoemission spectroscopy 共XPS兲, Cox et al.
5
兴
using dynamic mean field theory.
6
The 3d peaks of ruthe-
nium are compared for ruthenates put in order of metallicity,
and a systematic relative shift in spectral weight from the
so-called screened peak to the unscreened peak is observed.
To explain these shifts, these authors
1
use a model in which
the fitting parameters are the Hubbard U and the bandwidth
W of the ruthenium 4d band. The outcome of their analysis is
that the ratio U /W correlates with the ratio of the screened
and unscreened peaks, from which one can deduce that the
stronger the screened peak, the less correlation 共i.e., the more
metallic兲.
In ultraviolet photoemission spectroscopy 共UPS兲 spectra,
a similar shift in spectral weight is observed. In this case, the
t
2g
peak, which is very close to the Fermi level and referred
to as the coherent peak 共i.e., the quasiparticle band near E
F
兲,
is reduced as correlation increases, and a broad peak around
1.5 eV starts to rise. This latter peak is called the incoherent
peak 共i.e., the remnant of the Hubbard bands 1–2 eV above
and below E
F
兲. Both SrRuO
3
and CaRuO
3
can be classified
by their XPS and UPS spectra in this way, despite, for ex-
ample, similar transport properties at room temperature.
Based on the classification by XPS spectra, CaRuO
3
would
be a more correlated system than SrRuO
3
. However, the
samples studied in Ref. 1 could have suffered from the same
off-stoichiometry as will be discussed in this paper 共for in-
stance, due to the surface or sample preparation processes兲.
Viewed from the perspective of the work by Kim et al.,
1
in this paper, we show that similar systematic variations in
the degree of correlation can occur within a single ruthenate
as a function of stoichiometry. In particular, it turns out that
one source of disorder or off-stoichiometry can be varied in
SrRuO
3
thin films by changing the deposition conditions or,
what turns out to be more or less equivalent, the deposition
technique. Specifically, we demonstrate that variation of va-
cancies on the ruthenium site gives rise to a systematic
change in the degree of correlation. Moreover, the transport
properties of our samples are clearly linked to their photo-
emission spectra 共XPS and UPS兲 and to the crystal unit cell
parameters. This correlation with the unit cell volume V
c
permits us to use V
c
as a surrogate variable for the ruthenium
deficiency, a dependency which is well known from
literature.
4
This change in volume also alters the average
Ru-O-Ru bond angle, which is known to be closely con-
nected to the d-band width
7
and, therefore, electron correla-
PHYSICAL REVIEW B 76, 075126 共2007兲
1098-0121/2007/76共7兲/075126共6兲 ©2007 The American Physical Society075126-1
tion. SrRuO
3
appears to be a system where these effects of
correlation, which are normally not easily accessible, can be
studied in a systematic fashion, but we suspect that the un-
derlying physics is generic.
EXPERIMENT
The thin film samples reported in this paper are grown by
two different methods: molecular beam epitaxy 共MBE兲 and
pulsed laser deposition 共PLD兲. The samples are grown in the
same vacuum chamber with a background pressure of
10
−9
Torr. All films are grown on TiO
2
terminated SrTiO
3
substrates, prepared according to Koster et al.
8
The MBE
samples are deposited in an oxygen pressure of 10
−5
Torr
and at a substrate temperature of 700 °C at a rate of about
1 Å/s from separate ruthenium and strontium sources, which
enables us to vary the ratio of ruthenium to strontium. Elec-
tron impact emission spectroscopy is used to control the
deposition rate of the electron beam heated sources. Typical
thickness of the films grown ranges from 200 to 300 Å. Dur-
ing growth, atomic oxygen is provided by an Astex SXRHA
microwave plasma source. By adjusting oxygen flow and
generator wattage 共200–600 W兲, the amount of atomic oxy-
gen can be controlled.
9
During growth, reflection high-
energy electron diffraction 共RHEED兲 is used to monitor the
morphology of the samples.
For PLD a KrF excimer laser produces a 248 nm wave-
length beam with typical pulse lengths of 20–30 ns. A rect-
angular mask shapes the beam and a variable attenuator per-
mits variation of the pulse energy. The energy density on the
target is kept at approximately 2.1 J/cm
2
. Before each run
the rotating stoichiometric SrRuO
3
target is preablated for
2 min at 4 Hz. Films are deposited with a laser repetition
rate of 4 Hz, with the substrate temperature at 690 °C. The
oxygen background pressure was 10
−5
Torr as in the MBE
case; however, no atomic oxygen was provided for the PLD
films. The thickness of the films is in the same range as for
the MBE samples.
X-ray diffraction 共XRD兲 experiments were performed on
a Panalytical X’pert thin film diffractometer to determine
both the thickness and the crystalline quality of the samples.
UPS measurements were performed in situ in a vacuum
chamber attached to the main growth chamber, which has a
base pressure of ⬍5⫻ 10
−10
Torr, with a VG Scientific
ESCAlab Mark II system 共nonmonochromatic, helium dis-
charge兲. XPS measurements were performed in situ with the
same system 共nonmonochromatic Al K
␣
兲. This experimental
setup is crucial for obtaining reliable data since the surface of
SrRuO
3
is known to change when exposed to air
10
in a way
that gives rise to surface states in the spectra. Electrical
transport properties were measured with a Quantum Design
Physical Property Measurement System.
By varying the deposition conditions, we found that the
properties of the films changed. The films so obtained were
divided into three groups: ruthenium rich, near stoichio-
metric, and ruthenium poor. The ruthenium rich samples are
typically made by using a relatively low oxygen partial pres-
sure and/or low atomic oxygen flux and depositing a little
excess ruthenium. The nearly stoichiometric samples are
single phase and are obtained by tuning the oxygen activity
together with the use of a Ru/Sr ratio close to 1. The ruthe-
nium poor samples are created by using relatively high par-
tial oxygen pressures and a Ru/Sr ratio greater than or equal
to 1 共the ruthenium flux does not seem to be the determining
factor here兲.
It turns out to be impossible to grow SrRuO
3
with stron-
tium vacancies. It is possible though to grow samples that are
stoichiometric or with ruthenium vacancies 共for reasons that
will be discussed later兲. These samples are single phase and
have the SrRuO
3
crystal structure with no precipitations. The
precise stoichiometry is difficult to determine, but, based on
the combined analysis 共XRD and XPS兲, we can make the
above distinction without further specifying the exact com-
position of individual samples in each group. All three
sample types can be obtained by MBE, but samples made by
PLD using a stoichiometric target are typically ruthenium
poor. In the remainder of this paper, we will focus on
samples that are either 共near兲 stoichiometric or ruthenium
poor.
RESULTS
We first present our resistivity measurements. In Fig. 1,
the normalized resistance is plotted as a function of tempera-
ture for three samples: a stoichiometric MBE, a ruthenium
poor MBE, and a ruthenium poor PLD film. The normaliza-
tion factor is the room temperature resistivity. For the sto-
ichiometric samples the room temperature resistivity 共in the
bad metal regime兲 is around 190
⍀ cm, which compares
well to values found in literature for polycrystalline and
single crystal samples 共150–200
⍀ cm兲. When ruthenium
vacancies are introduced, the value increases markedly to
roughly 300
⍀ cm. The room temperature resistivity varies
FIG. 1. 共Color online兲 Resistance behavior as a function of tem-
perature for a nearly stoichiometric MBE grown 共black solid兲,a
ruthenium poor MBE grown 共red dashed兲, and a PLD grown 共or-
ange dotted兲 SrRuO
3
film. The resistance has been normalized with
respect to the values at 300 K. The inset shows the derivative of
each line; T
c
has been set as the point where the derivative is maxi-
mal. The resistivity 共
兲 of the stoichiometric MBE grown SrRuO
3
at room temperature is about 190
⍀ cm and its carrier density at
4 K is 2⫻ 10
22
cm
−3
.
SIEMONS et al. PHYSICAL REVIEW B 76, 075126 共2007兲
075126-2
systematically with the amount of ruthenium vacancies, as
shown in Fig. 2共b兲 共discussed below兲. Also of interest is the
residual resistivity ratio 共RRR兲, which is defined as the re-
sistivity at 300 K divided by the value at 4 K. For the best
single crystals, values between 50 and 100 are reported,
11
and similar results can be obtained for optimized thin films.
12
Like the room temperature resistivity, the RRR is very sen-
sitive to changes in stoichiometry. For the stoichiometric
sample in Fig. 1 共black solid line兲, the RRR is 26, which is
excellent for the thickness of these samples, whereas for both
ruthenium poor films 共orange dotted and red dashed lines兲,
the RRR drops to around 7. Note that for the majority of
PLD films in the literature, which fall in the ruthenium poor
class 共see below兲, values of 5 or less are reported.
13,14
Next we turn to the Curie temperature of our films, which
are shown in the inset of Fig. 1, where the temperature de-
rivatives of the resistance near the transition temperature are
compared. The transition temperature T
C
in bulk samples is
160 K, whereas in all of our thin film samples, T
C
is reduced
by about 10 K 共Ref. 15兲 due to strain. From literature, we
know that when a sufficient amount of ruthenium vacancies
are introduced, T
C
is lowered further.
4
This gives us some
indication of the order of magnitude of the vacancy density.
On this basis, we estimate that for the range of samples we
studied, the vacancy concentration is much smaller than a
few percent.
In order to get a quantitative indicator of the ruthenium
deficiency, we examined the change in lattice parameters in
going from the stoichiometric to ruthenium poor samples. In
Fig. 2共a兲, we show
/2
scans for two thick films 共330 nm兲,
one grown with MBE in low oxygen pressure and one with
PLD in high oxygen pressure, to illustrate the difference in
the d spacing for the out-of-plane 共110兲 direction of SrRuO
3
when different deposition techniques are used. These films
are much thicker than the samples compared in other parts of
the paper to show that the films remain perfectly crystalline
and strained up to very large thicknesses. In addition, these
scans reveal the very high degree of crystallinity of the
samples by the existence of finite size fringes in the scans for
both samples. As can be seen, the out-of-plane lattice con-
stant is larger for the film grown in high oxygen pressures,
explained by a lower ruthenium content.
4
Since the c axis 共in
plane兲 of the SrRuO
3
is fixed by epitaxy 共the films are fully
strained for all sample thicknesses in this paper兲, the increase
in the 共110兲 direction 共linear combination of the a and b
axes兲 is directly proportional to an increase in unit cell vol-
ume, which we take as our measure of the ruthenium defi-
ciency in Fig. 2共b兲.
We obtained the lattice parameters by refining on six re-
flections. For samples which were classified as having low
ruthenium content, we find a unit cell with larger volume.
This increase in the case of ruthenium vacancies is consistent
with earlier measurements done on ruthenium deficient
single crystal samples.
4
The bulk values taken from literature
are a=5.572 Å, b=5.533 Å, c=7.849 Å, and V
bulk
=242.0 Å
3
共Ref. 4兲 for single crystals and correspond to an
unstrained lattice, unlike the films grown here on SrTiO
3
.
In Fig. 2共b兲, we show how changes in the transport data
correlate with the volumes of the orthorhombic unit cell for
various films studied. The room temperature and low tem-
perature resistivities as well as the RRRs show a clear trend,
with smaller volumes showing the lowest resistivities and
highest RRRs. These films are typically found in our second
group made by MBE 共oxygen activity tuned to optimize oxi-
dation and ruthenium sticking兲. On the other end, PLD films
appear to have larger volumes and poorer transport proper-
ties. Note that the Curie temperature should also show a
similar trend, although the change in T
C
is small as shown in
Fig. 1.
To get a clearer understanding of how or whether the
electronic properties of SrRuO
3
vary as a function of stoichi-
ometry, we measured the UPS and XPS spectra of some of
our samples. In Fig. 3共a兲, the UPS spectra are plotted for the
same samples that were used for the transport data in Fig. 1.
First, for the stoichiometric SrRuO
3
sample grown by MBE,
the spectrum shows a peak at the Fermi energy, correspond-
ing to the ruthenium t
2g
band and a valley at binding energy
⬃1.5 eV. This peak has been observed before by Kim et al.
16
and is also expected based on models.
17
In ruthenium poor
(a)
(b)
FIG. 2. 共Color online兲共a兲 Comparison of
/2
XRD scans of
two thick 共 330 nm兲 films grown by PLD 共orange dashed兲 and MBE
共black solid兲. The fringes indicate good crystalline quality of the
films; the shift in peak position indicates that the PLD film has a
larger out-of-plane lattice parameter 关共110 direction兲兴. 共b兲 A com-
parison between the volume of the SrRuO
3
unit cell in thin films on
SrTiO
3
substrates grown by either MBE or PLD and their transport
properties: room temperature 共red circles兲 and low temperature re-
sistivity 共blue triangles兲 on the right y axis and residual resistance
ratio 共black squares兲 on the left y axis. The dashed lines are guides
for the eyes.
DEPENDENCE OF THE ELECTRONIC STRUCTURE OF… PHYSICAL REVIEW B 76, 075126 共2007兲
075126-3
samples, as can be seen in Fig. 3共a兲, for both the PLD film
and the ruthenium poor MBE film, a second, broad peak
appears at ⬃1.5 eV, the so-called incoherent peak. Note that
these spectra are normalized on the t
2g
peak to make their
comparison easier.
Examples of XPS core-level spectra for one stoichio-
metric sample and one ruthenium poor sample are plotted in
Fig. 3共b兲. For clarity, we reduced these spectra of the ruthe-
nium 3d doublet by subtracting the strontium 3p
1/2
peak fit-
ted with a Gaussian, which overlaps at lower binding energy.
Note that our in situ measurements are free of carbon con-
tamination and its complications near the surface region.
10
This is confirmed by the absence of the C 1s peak, which
normally also overlaps with the Ru 3d peaks. The stoichio-
metric sample shows more spectral weight for the Ru 3d
peaks at lower binding energy compared to the ruthenium
poor sample. Peak fitting using three sets of Gaussians 共the
area of the peaks in each spin orbit doublet pair have a 2:3
ratio兲 shows that at a binding energy of ⬃278 eV an extra
curve 共yellow filled兲 is necessary to fit the data for stoichio-
metric SrRuO
3
, which is arguably the so-called screened
peak typical for SrRuO
3
. On the other hand, the data for the
ruthenium poor sample are dominated by the unscreened
peak.
DISCUSSION
Before discussing the data explicitly, we would first like
to point out that the stoichiometry of the samples is ex-
tremely dependent on the oxygen activity during deposition.
For MBE, this can be independently varied by controlling
the flux of molecular or atomic oxygen. Apparently, the
sticking of ruthenium 共possibly through the formation of
RuO
4
, which is very volatile
18,19
兲 is a function of oxygen
activity: at relatively low oxygen activity, the stoichiometry
is mostly determined by the supplied Sr/Ru ratio. When ex-
cess ruthenium is supplied, RHEED, scanning electron mi-
croscopy, Auger spectroscopy, and transmission electron mi-
croscopy reveal precipitation of RuO
2
, and the Ru 3d core-
level spectra show very strong screened peaks typical for
RuO
2
. At intermediate oxygen activity it is most favorable to
achieve perfect stoichiometry: RHEED shows two dimen-
sional growth and XPS and/or UPS show almost perfect
spectra typical for SrRuO
3
. The “best” values for resistivity
and RRR are obtained here. Finally, at high oxygen activity
ruthenium vacancies are unavoidable, independent of the ra-
tio Sr/Ru in the supplied vapor: RHEED shows two dimen-
sional growth, layer by layer or even step flow,
20
and XRD
clearly shows a modification in the unit cell, indicative of
ruthenium vacancies.
4
We believe the last scenario to be the case also for PLD
films, where a high 共atomic兲 oxygen pressure exists within
the plume 共in addition to the background activity兲 and very
little can be done to avoid this. PLD offers the advantage of
making the same quality film every time, albeit ruthenium
poor. On the other hand, we have shown that, even when
ruthenium vacancies are present, the crystallinity of the ma-
terial remains the same compared to the stoichiometric films
and therefore cannot be used to explain any of the observed
variations in transport and photoelectron spectroscopy 共PES兲.
(a)
(b)
FIG. 3. 共Color online兲共a兲 UPS spectra at room temperature of
three SrRuO
3
films: a MBE film 共black solid兲, a ruthenium poor
MBE film 共red dashed兲, and a PLD film 共orange dotted兲共same
samples as in Fig. 1兲. All samples show states at the Fermi edge and
aRut
2g
peak close to the Fermi level. The spectra have been nor-
malized on the t
2g
peak. 共b兲 XPS detail spectra at room temperature
of the Ru 3d doublet of a stoichiometric film 共top兲 and a ruthenium
poor film 共bottom兲 including fitting of the spectra using three spin
orbit coupled doublets represented by Gaussians 共two pairs orange
dashed or dotted, the third pair yellow filled兲; the overlapping
Sr 2p
1/2
peak 共Gaussian兲 was subtracted for both spectra. In the
spectrum of stoichiometric film, more pronounced screened peaks
at lower binding energy 共yellow filled兲 are needed to give a good fit,
indicative of less correlated behavior. For the ruthenium poor spec-
trum, the screened peaks are almost nonexistent.
SIEMONS et al. PHYSICAL REVIEW B 76, 075126 共2007兲
075126-4
There seems no indication that oxygen vacancies play a sig-
nificant role, but we cannot distinguish this from the forma-
tion of ruthenium vacancies by looking at the response of
lattice constants only. In a study of similar ruthenium defi-
cient samples, oxygen vacancies were not observed.
4
Resistivity is affected both at low temperatures with lower
RRR values and at higher temperatures with a variation
in lower room temperature resistivity for stoichiometric
samples. At low temperatures, defect scattering dominates
transport properties, resulting in the systematic behavior of
the resistivity at low temperature and the RRR values, in-
creasing and decreasing, respectively, almost exponentially
with unit cell volume 共more defects兲. On the other hand, at
room temperature in the bad metal regime, the resistivity
increases more linearly with the unit cell volume. This indi-
cates that more than just scattering is contributing at room
temperature. It would be interesting to study the high tem-
perature behavior 共beyond the Ioffe-Regel limit
21
兲 as a func-
tion of stoichiometry, which is part of a planned future en-
deavor.
Now we turn to the observed differences in PES and their
correlation with transport properties as a function of ruthe-
nium stoichiometry. First, it is logical to exclude the ruthe-
nium rich samples from this comparison because both trans-
port and PES would be a mixture of SrRuO
3
and precipitated
RuO
2
. Both UPS and XPS spectra of films with ruthenium
vacancies are distinctly different from their stoichiometric
counterparts. Our data of the Ru 3d peaks in SrRuO
3
with
vacancies suggest that these ruthenium poor samples show
almost no screened peaks, whereas the stoichiometric
samples show pronounced screened peaks. This effect is
most often associated with the material becoming more cor-
related, which has been quantified by Kim et al.;
1
see also
our earlier discussion. In the same spirit, ruthenium poor
samples show spectral weight at ⬃1.5 eV in He
I UPS,
which has previously been attributed to an incoherent peak in
a picture of a strongly correlated system. When the ⬃1.5 eV
peak is absent, the t
2g
peak at the Fermi level becomes more
pronounced. It is noteworthy here that the t
2g
peak at the
Fermi level has not yet been observed with scraped bulk
samples, indicating that thin film specimens may provide a
better opportunity in investigating the intrinsic electronic
structure of SrRuO
3
with photoemission spectroscopy as has
been pointed out by Kim et al.
16
This shift in spectral weight
is to be expected when the spectra are compared with density
of states 共DOS兲 calculations.
17,22,23
The calculations show a
t
2g
peak with almost the same height as the O 2p peak. We
can consider this discrepancy in terms of a self-energy cor-
rection to the one-electron band structure.
7,16,17,23,24
Due to
the k dependence of the self-energy, the spectral weight at
the Fermi level is reduced by a factor m
k
/m
b
, where m
k
is
called the “k mass” and m
b
is the bare band mass. By com-
paring with band calculations, we estimate m
k
/m
b
to be 0.3
for the sample with ruthenium vacancies and 0.6 for the
stoichiometric sample. Combining these values with the
mass enhancement factor m
*
/m
b
as derived from the
electronic specific heat
␥
, one obtains the quasiparticle
weight Z =共m
k
/m
b
兲/共m
*
/m
b
兲. For m
*
/m
b
, we have used the
average value of 4.1 as determined by two different
measurements.
17,22
This leads to values of Z
−1
of 13 and 7 for
the ruthenium poor and stoichiometric samples, respectively.
The value of 13 for the ruthenium poor samples is in line
with what has been reported by most authors.
16,17,23
The
value of 7 for the stoichiometric samples is slightly lower
than what was reported by Kim et al.
16
on their samples. The
high values for Z
−1
have been used by others
17,23
to explain
the reduction of the t
2g
peak at the Fermi level due to elec-
tron correlation. Our work shows that the intensity of the t
2g
peak is very sensitive to the ruthenium stoichiometry and
that UPS spectra of stoichiometric films look much more like
the calculated DOS than ruthenium poor films. We conclude
that ruthenium stoichiometry 共and possibly surface states
contribute as well兲 has a much larger impact on PES spectra
than previously assumed and could provide for a large part
an explanation for the observed photoemission results in this
paper, but also by others. Of course, the change in the PES
spectra could be caused by a change in correlation due to the
change in unit cell and the Ru-O-Ru bond angle, but it is also
possible that the vacancies themselves caused it. Therefore
we propose that besides a change in correlation, the transfer
of spectral weight could also be caused by inelastic processes
at the ruthenium vacancies. The t
2g
orbitals that form the
conduction band form states spreading in two dimensions. To
define the energy of a state to within the observed 0.5 eV
peak, it would be necessary to construct a packet of band
states, which, for a band of width 4 eV, would extend
4/0.5=8 cube edges. The corresponding
4
2
⬇50 ruthenium
sites would generally contain at least one vacancy, even at
only 2% vacancies. It would be reasonable to expect that the
removal of a band electron might cause shake-off excita-
tions, vibrational or electronic, of the vacancy. This would
increase the energy needed to eject the electron, as can be
seen in both the UPS and XPS spectra.
CONCLUSIONS
We have grown SrRuO
3
thin films on SrTiO
3
under vari-
ous conditions and conclude that this material exhibits a
range of properties due to a subtle change in stoichiometry
on the ruthenium site related to the oxidation conditions dur-
ing deposition. Resistivity, x-ray diffraction, UPS, and XPS
all seem to indicate that this change in behavior is due to a
changing electron-electron correlation, although we also
point out that contributions from inelastic processes at va-
cancy sites as well as from surface states cannot be ignored.
These results shed light on the well-known sensitivity of the
properties of SrRuO
3
to its synthesis conditions. Equally im-
portant, they suggest a clear path to more quantitative com-
parisons with theory.
ACKNOWLEDGMENTS
One of us 共G.K.兲 thanks the Netherlands Organization for
Scientific Research 共NWO, VENI兲. W.S. thanks the Nano-
technology network in the Netherlands, NanoNed. We would
also like to thank Walt Harrison, M. Naito, Kookrin Char,
Ann Marshall, Jim Reiner, Guus Rijnders, Ted Geballe, and
Robert Hammond for helpful discussions. This work was
supported by the DOE BES with additional support from
EPRI.
DEPENDENCE OF THE ELECTRONIC STRUCTURE OF… PHYSICAL REVIEW B 76, 075126 共2007兲
075126-5
