of 13
Spin-polarized tunneling spectroscopic studies of the intrinsic heterogeneity and pseudogap
phenomena in colossal magnetoresistive manganite La
0.7
Ca
0.3
MnO
3
C. R. Hughes,
1
J. Shi,
1
A. D. Beyer,
1
and N.-C. Yeh
1,2
1
Department of Physics, California Institute of Technology, Pasadena, California 91125, USA
2
Kavli Nanoscience Institute, California Institute of Technology, Pasadena, California 91125, USA

Received 31 March 2010; revised manuscript received 3 September 2010; published 26 October 2010

Spatially resolved tunneling spectroscopic studies of colossal magnetoresistive

CMR

manganite
La
0.7
Ca
0.3
MnO
3

LCMO

epitaxial films on

LaAlO
3

0.3

Sr
2
AlTaO
6

0.7
substrate are investigated as functions
of temperature, magnetic field and spin polarization by means of scanning tunneling spectroscopy. Systematic
surveys of the tunneling spectra taken with Pt/Ir tips reveal spatial variations on the length scale of a few
hundred nanometers in the ferromagnetic state, which may be attributed to the intrinsic heterogeneity of the
manganites due to their tendency toward phase separation. The electronic heterogeneity is found to decrease
either with increasing field at low temperatures or at temperatures above all magnetic ordering temperatures.
On the other hand, spectra taken with Cr-coated tips are consistent with convoluted electronic properties of
both LCMO and Cr. In particular, for temperatures below the magnetic ordering temperatures of both Cr and
LCMO, the magnetic field-dependent tunneling spectra may be quantitatively explained by the scenario of
spin-polarized tunneling in a spin-valve configuration. Moreover, a low-energy insulating energy gap

0.6 eV
commonly found in the tunneling conductance spectra of bulk metallic LCMO at
T
0 may be attributed to a
surface ferromagnetic insulating phase, as evidenced by its spin-filtering effect at low temperatures and van-
ishing gap value above the Curie temperature. Additionally, temperature-independent pseudogap

PG

phenom-
ena existing primarily along the boundaries of magnetic domains are observed in the zero-field tunneling
spectra. The PG becomes strongly suppressed by applied magnetic fields at low temperatures when the tun-
neling spectra of LCMO become highly homogeneous. These findings suggest that the occurrence PG is
associated with the electronic heterogeneity of the manganites. The observation of lateral and vertical elec-
tronic heterogeneity in the CMR manganites places important size constraints on the development of high-
density nanoscale spintronic devices based on these materials.
DOI:
10.1103/PhysRevB.82.134441
PACS number

s

: 75.47.Lx, 73.40.Gk, 72.25.

b, 71.27.

a
I. INTRODUCTION
“Spintronics” is a new paradigm of electronics based on
the spin-dependent charge transport in magnetic
heterostructures.
1
,
2
It has emerged as one of the most active
research fields in recent years because of the potential advan-
tages of nonvolatility, faster processing speed, small power
dissipation for high-density device integration as opposed to
conventional semiconducting systems,
1
and better coherence
promising for quantum information technology.
2
Among the physics issues associated with spintronics,
knowledge of spin-polarized quantum tunneling and trans-
port across interfaces is particularly important for developing
high-quality and reproducible spintronic devices. From the
perspective of materials, ferromagnets with higher degrees of
spin polarization and higher Curie temperatures are desirable
candidates for use in spintronic devices. In this context, the
manganese oxides
Ln
1−
x
A
x
MnO
3

Ln
: trivalent rare-earth
ions and
A
: divalent alkaline-earth ions

, also widely known
as manganites that exhibit colossal magnetoresistance

CMR

effects,
3
8
appear to be promising spintronic materi-
als because of their half metallicity in the ferromagnetic
state
9
11
so that the degree of spin polarization is nearly
100%. Nonetheless, experimental data and theoretical calcu-
lations have suggested that the ground states of the mangan-
ites tend to be intrinsically inhomogeneous as the result of
their strong tendencies toward phase separation and the
phase separation may involve domains of ferromagnetic met-
als, ferromagnetic insulators, and antiferromagnetic charge-
and orbital-ordered insulators.
9
,
12
16
In fact, it has been dem-
onstrated numerically that the double-exchange interaction
and half metallicity alone cannot account for the large mag-
nitude of negative magnetoresistance in ferromagnetic
manganites,
9
and that the tendencies toward phase separation
in the ground state even for the nominally metallic ferromag-
netic phases play an essential role in the occurrence of the
CMR effects.
9
From the viewpoint of technological applica-
tions, the intrinsic electronic heterogeneity of the manganites
becomes a relevant concern for consistently fabricating min-
iaturized spintronic devices with high areal densities. In this
context, understanding of the spatial varying physical char-
acteristics of the manganites at the microscopic scale will be
important for spintronic applications based on these
materials.
In addition to the tendency toward phase separation in
bulk manganites, various experimental findings have sug-
gested that the surfaces of manganites appear to differ from
the bulk characteristics.
16
For instance, scanning tunneling
spectroscopic

STS

studies of nominally metallic manganite
epitaxial thin films
17
20
and single crystals
13
have always re-
vealed either a small energy gap or a pseudogap

PG

near
the Fermi surface. Although the occurrence of an energy gap
in a nominally metallic manganite could not be explained by
either band-structure calculations
10
,
11
or bulk electrical trans-
port measurements,
21
,
22
the tunneling spectra of the metallic
manganite over a wide energy range except near the Fermi
PHYSICAL REVIEW B
82
, 134441

2010

1098-0121/2010/82

13

/134441

13

©2010 The American Physical Society
134441-1
level were actually consistent with theoretical calculations of
the bulk electronic density of states.
17
,
18
Hence, it is reason-
able to conjecture that the surface of manganites may be a
thin layer with chemical compositions differing from those
of the bulk. Thus, lower-bias ballistic electrons injected from
the scanning tunneling microscopy

STM

tip could be more
sensitive to the surface state, whereas ballistic electrons in-
jected with higher bias could penetrate deeper into the bulk.
In other words, the spectroscopic characteristics at higher
bias voltages may be more representative of the bulk density
of states, whereas those at lower bias voltages may be better
related to the surface state. This conjecture is consistent with
the x-ray photoemission spectroscopic

XPS

studies of
La
0.67
Ca
0.33
MnO
3
that inferred a surface predominantly ter-
minated by an insulating layer of MnO
2
.
23
Another interesting feature associated with the mangan-
ites is the occurrence of PG phenomena.
9
Theoretical studies
using Monte Carlo simulations
24
,
25
suggest that the density
of states

DOS

in the manganites should exhibit PG charac-
teristics with significant spectral depletion at the chemical
potential and broad DOS peaks both above and below the
chemical potential. These theoretical findings have been cor-
roborated by photoemission experiments on bilayer mangan-
ites above their magnetic ordering temperatures.
26
Theoreti-
cally, the occurrence of PG may be regarded as a precursor
of phase separations in forms of magnetic clusters,
24
,
25
which
is analogous to the widely studied PG phenomena in cuprate
superconductors where the appearance of PG is attributed to
the onset of preformed pairs and competing orders.
27
,
28
How-
ever, whether the PG phenomena are common among differ-
ent types of manganites and whether the physical origin of
PG in the manganites is indeed associated with the onset of
mixed phases have not been extensively verified by experi-
ments.
To address the aforementioned issues of phase separa-
tions, insulating surface layers and the PG phenomena in the
manganites, we report in this work spatially resolved tunnel-
ing spectroscopic studies of La
0.7
Ca
0.3
MnO
3

LCMO

epitax-
ial films by means of both regular and spin-polarized STM

Refs.
29
33

STM and SP-STM

. The specific calcium
doping level of
x
=0.3 was chosen because it corresponded to
a nominal metallic phase with nearly the highest Curie tem-
perature

T
C

270 K

and the most spatial homogeneity
among the Ca-doped manganites. The evolution of the spa-
tially resolved tunneling spectra was studied systematically
with temperature, magnetic field, and the degree of spin po-
larization, which provided information about the spatial
scales of stoichiometric inhomogeneity and the average size
of ferromagnetic domains. Additionally, comparison with
band-structure calculations suggested that the spectral char-
acteristics taken with regular STM were consistent with
those of the DOS of the manganite, whereas data taken with
SP-STM may be understood in terms of the product of a
spin-dependent tunneling matrix and the joint density of
states between the SP-STM tip material and the manganite.
On the other hand, the evolution of the surface energy gap
with temperature, magnetic field, and the degree of spin po-
larization was found to be consistent with the spin-filtering
effect of a surface ferromagnetic insulating

FI

phase.
15
Fi-
nally, PG phenomena were found to persist at temperatures
well above all magnetic ordering temperatures in the absence
of external magnetic fields. The PG features were suppressed
by moderate magnetic fields at low temperatures when the
tunneling conductance of LCMO became spatially homoge-
neous.
The findings presented in this work provide quantitative
experimental accounts for the presence of phase separations,
surface states, and PG phenomena in the manganites. In
comparison with other STM studies of the manganites, while
previous investigations have revealed spatially inhomoge-
neous spectra in manganites with different doping levels,
12
14
this work provides the first high-field SP-STM studies on the
manganites and demonstrates the general application of high-
field SP-STM techniques to the investigation of spatially in-
homogeneous magnetic materials.
II. EXPERIMENTAL
The LCMO films used in this study were epitaxially
grown on

LaAlO
3

0.3

Sr
2
AlTaO
6

0.7
substrates by means of
pulsed laser deposition techniques. The substrates were cho-
sen because of their small lattice mismatch

about 0.3%

with the bulk LCMO, which ensures minimized strain in the
resulting films to prevent excess strain-induced effects on the
electronic properties.
21
,
22
,
34
The films were deposited in a
100 mTorr oxygen background pressure to a thickness of

110

10

nm with the substrate temperature kept at
650 °C, and were subsequently annealed at the same tem-
perature for 2 h in 100 Torr O
2
and then cooled slowly to
room temperature. The epitaxy of these films was confirmed
by x-ray diffraction, and the film quality was further verified
by atomic force microscopy and STM measurements, show-
ing terraced growth with step heights corresponding to one
c
-axis lattice constant of the bulk LCMO, as exemplified in
Fig.
1

a

. Further characterizations were conducted using a
superconducting quantum interference device

SQUID

mag-
netometer by Quantum Design for magnetization measure-
ments, showing a Curie temperature of 270

10 K as exem-
plified in Fig.
1

b

for one of the LCMO epitaxial films.
Additionally, the coercive fields of the LCMO films

H
C
LCMO

for magnetic fields perpendicular to the plane of films were
determined by the SQUID magnetometer and were found to
range from 0.02 to 0.04 T as exemplified in Fig.
1

c

.In
general, LCMO thin films fabricated under aforementioned
conditions have been shown to yield high-quality epitaxy via
high-resolution x-ray diffraction studies, magnetization, and
transport measurements.
21
,
22
,
35
Prior to STM measurements, samples were first etched in
a 0.5% bromine in pure ethanol solution and then rinsed in
pure ethanol to remove surface contaminants such as the car-
bonates. Each etched sample was immediately loaded onto
our cryogenic STM system while kept under excess pressure
of helium gas during the loading process. It is worth noting
that the presence of surface contaminants is common among
the perovskite oxides that contain alkaline-earth metals

A
=Ca, Sr, and Ba

because the alkaline-earth elements are
highly reactive to H
2
O and CO
2
.
36
Consequently, the re-
moval of surface nonstoichiometric alkaline-earth com-
pounds such as
A
O,
A
CO
3
, and
A

OH

2
upon chemical etch-
HUGHES
et al.
PHYSICAL REVIEW B
82
, 134441

2010

134441-2
ing and ethonal rinsing would tend to yield slight Ca
deficiency on the sample surface relative to its bulk stoichi-
ometry unless long-time etching was carried out.
36
Further,
sample surfaces without proper removal of the contaminants
often exhibit excess concentrations of alkaline-earth metals
in the XPS studies.
36
,
37
The tunneling studies were conducted with our home-
made cryogenic STM with a base temperature
T
=6 K and a
superconducting magnet capable of magnetic fields up to
H
=7 T. At
T
=6 K the STM system was under ultrahigh
vacuum with a base pressure

10
−10
mbar. Regular tunnel-
ing spectroscopic studies were conducted with atomically
sharp Pt/Ir tips, while spin-polarized studies were made by
means of Pt/Ir tips evaporatively coated with 15–30 mono-
layers of Cr metal prepared in a separate evaporation system,
following the procedures reported previously.
29
,
30
Here we
note that thin-film Cr is ferromagnetic
29
,
30
and the Curie tem-
perature

T
C
Cr

of our Cr-coated tip was found to be much
higher than room temperature. Therefore, it is justifiable to
assume that electrical currents from the Cr-coated tip were
always spin polarized. This assumption was subsequently
verified by field-dependent tunneling conductance measure-
ments, and the degree of spin polarization

15%

was also
estimated, as elaborated in Sec.
III
. The coercive field

H
C
Cr

of the Cr-coated tips was empirically determined by studying
the tunneling conductance vs. magnetic field
H
for electrical
currents tunneling from the Cr-coated tip to a permanent
magnet NdNi. The magnetization of NdNi was opposite to
the applied magnetic field, so that for fields exceeding the
coercive field of the Cr-coated tip, tunneling from the spin-
polarized tip to NdNi became forbidden near the Fermi level
and the tunneling conductance approached zero, as shown in
Fig.
1

d

for small bias voltages at
V
=

15,

35, and

58 mV. The
H
C
Cr
value thus determined was 0.8

0.1 T,
comparable to the values determined by similar tunneling
conductance measurements outlined in Refs.
29
and
30
.
Thus, we have established the condition
H
C
LCMO

H
C
Cr

1.0 T for all Cr-coated tips.
In the following studies, the STM system was operated at
T
=6 K for
H
=0, −0.3, and 3.0 T, and also at
T
=77 and 300
K for
H
=0. Both Pt/Ir- and Cr-coated tips were used in these
studies. Therefore, the field-dependent studies were carried
out at
T

T
C
LCMO

T
C
Cr
and under three conditions:
H
=0,
H
C
LCMO


H


H
C
Cr
with
H

0, and

H


H
C
Cr
with
H

0.
These conditions ensured that the effects of spin-polarized
currents were investigated for three different magnetic con-
figurations in LCMO, which are

1

randomly oriented mag-
netic domains,

2

aligned LCMO magnetic domains that
were antiparallel to the spin-polarized current, and

3

aligned LCMO magnetic domains that were parallel to the
spin-polarized current.
The spectroscopic measurements consisted of tunneling
current

I

versus bias voltage

V

spectra taken from
V
=−3 to 3 V at each pixel on a

128

128

pixel grid located
over a

500

500

nm
2
area for zero-field studies at 77 and
300 K. On the other hand, the scanned sample area for field-
dependent measurements at 6 K was reduced to

250

90

nm
2
because of the much reduced scanning range of
the piezoelectric material at low temperatures and the experi-
mental preference to complete a full spectroscopic scan with-
out interruptions by the need of liquid-helium transfer. For
all tunneling spectra, the typical junction resistance was kept
at

100 M

. It was verified that the tunneling spectroscopy
and topography were both independent of slight variations in
the tip height relative to the sample surface under this range
of junction resistance. For consistent analysis of all experi-
mental data taken under different conditions on multiple
samples, the tunneling spectra to be discussed below were all
processed into normalized differential conductance,

dI
/
dV

/

I
/
V

G
̄
, because this quantity minimizes extrin-
sic effects incurred by slight variations in the sample-tip
separation, and so best represents the material characteristics
of the tunnel junction.
For comparison of the spectral characteristics as functions
of temperature, magnetic field and spin polarization, it would
have been ideal to conduct all studies over identical sample
areas. However, in practice we were only able to investigate
the field-dependent spectra over the same sample area by
FIG. 1.

Color online

a

Typical STM topography of a
La
0.7
Ca
0.3
MnO
3
epitaxial film on

LaAlO
3

0.3

Sr
2
AlTaO
6

0.7
sub-
strate over a

500

500

nm
2
sample area, showing atomically flat
surfaces and steps of one lattice constant height.

b

Magnetization

M

vs temperature

T

data of the same LCMO epitaxial film taken
under an applied magnetic field
H
=0.01 T, showing
T
C
LCMO
=270

10 K with detailed temperature dependence near
T
C
LCMO
given in the inset.

c

Main panel:
M
-vs-
H
hysteresis curve of the
same LCMO film taken at a constant temperature
T
=7 K and for
H
from −0.2 to 0.2 T. Inset: the same
M

H

curve as in the main panel
and for −0.05
H
0.05 T, showing a LCMO coercive field
H
C
LCMO

0.02 T.

d

Normalized tunneling conductance vs
H
spec-
tra obtained from tunneling electrical currents from a Cr-coated tip
to a permanent magnet, a NiNd single crystal, at constant bias volt-
ages
V
=

15,

35, and

58 mV and for
T
=6 K. The magnetic
polarization of NiNd was opposite to the applied magnetic field.
Therefore, when the applied field exceeded the coercive field of the
Cr-coated tip, the spin-polarized tunneling from the Cr-coated tip to
NiNd became forbidden near the Fermi level, yielding nearly zero
tunneling conductance. The coercive field thus determined is

H
C
Cr

=0.8

0.1 T.
SPIN-POLARIZED TUNNELING SPECTROSCOPIC
...
PHYSICAL REVIEW B
82
, 134441

2010

134441-3
keeping the measurements at a constant temperature
T
=6 K. This limitation was because changes in the measure-
ment temperature would result in drifts of the STM tip and
replacing the STM tip

e.g., from Pt/Ir- to Cr-coated tips

would lead to a different sample area upon reapproaching the
tip to the sample. Given the intrinsic heterogeneity of
LCMO, meaningful comparison among data taken either
with different STM tips or at different temperatures could
only be made if statistical consistency in the spectral charac-
teristics could be established. This premise was indeed veri-
fied in our investigation, as elaborated further in Sec.
III
.
III. RESULTS AND ANALYSIS
Based on the methods outlined in Sec.
II
, systematic stud-
ies of the tunneling spectral evolution with temperature,
magnetic field, and spin polarization were carried out to ad-
dress the issues of phase separation, surface state, and PG
phenomena in the nominally metallic manganite epitaxial
thin films of LCMO on

LaAlO
3

0.3

Sr
2
AlTaO
6

0.7
sub-
strates.
A. Spectral characteristics
Our detailed surveys of the LCMO tunneling spectra over
relatively large sample areas and multiple samples with both
Pt/Ir- and Cr-coated tips at
T
=77 K


T
C
LCMO

T
C
Cr

and
H
=0 revealed three types of representative tunneling spectra,
which are labeled as
,
, and
types for convenience.
Examples of the
and
types of spectra taken with a Pt/Ir
tip are shown, respectively, in Figs.
2

a

and
2

b

, and rep-
resentative spectra of the
and
types taken with a Cr-
coated tip are illustrated in Figs.
2

e

and
2

f

.
For the dominant
type of spectra, there are four primary
characteristic features, including two major conductance
peaks at energies of
=
U
+
and −
U
, and two smaller peaks
flanking a low-energy insulating gap at the Fermi level. The
type of spectra is consistent with our previous single-point
spectroscopic studies.
17
In the case of the
-type spectra as
exemplified in Fig.
2

b

, only one pair of peak features at
=
U
+
and
=−
U
may be identified, and this type of spec-
tra typically reveals a wider tunneling gap. The
-type spec-
tra are essentially PG-like, as shown in Fig.
2

f

. Specifically,
the characteristic energies
U


associated with the broad peak
features of the
-type spectra are much smaller than those
found in the other two types of spectra, and the conductance
values at
=
U
+

and
=−
U

are also much smaller. More-
over, in contrast to the completely vanished DOS over a
finite range of energies in the
- and
-type spectra, the DOS
in the
-type spectra only vanishes at the Fermi level and
remains finite for all finite energies

0.
FIG. 2.

Color online

Characteristic features in the regular and spin-polarized tunneling spectra of LCMO taken at
T
=77 K and
H
=0.

a

A normalized
-type tunneling conductance spectrum taken with a Pt/Ir tip, showing maximal conductance at
=−
U
and
=
U
+
that
corresponds to the maximal DOS associated with the majority and minority bands of LCMO, respectively

Ref.
10

. Additional features
associated with an insulating energy gap at
=−

and
=

+
are observed. Here the values of


are determined by identifying the
energies where the second derivatives of the tunneling current

I

relative to the bias voltage

V

reach maxima, as indicated in the inset.

b

A normalized
-type tunneling conductance spectrum taken with a Pt/Ir tip, showing one set of conductance peaks at
=−
U
and
=
U
+
.
The insulating gap values


are determined in the same way as in

a

and are shown in the inset.

c

The histograms of the characteristic
energies
U

and


derived from the tunneling spectra taken with a Pt/Ir tip over a

500

500

nm
2
sample area. Here

U
+

refers to the
maximal count of the positive characteristic energy that is associated with the DOS peak in LCMO and

+

refers to the positive PG energy
associated with the
-type spectra.

d

Energy histograms of
U

and


derived from the tunneling spectra taken with a Pt/Ir tip over a

500

500

nm
2
sample area different from the area studied in

c

, showing statistically similar results.

e

A normalized
-type tunneling
conductance spectrum taken with a Cr-coated tip, showing similar features to the spectra taken with a Pt/Ir tip, although the values of
U

and


differ slightly.

f

A normalized
-type tunneling conductance spectrum taken with a Cr-coated tip, showing PG-like behavior with
one set of relatively low conductance peaks at
=−
U

and
=
U
+

and vanishing DOS at

0. Here we use the notation

to represent
features associated with the PG-like spectra.

g

Energy histograms of
U

and


derived from the tunneling spectra taken with a Cr-coated
tip over a

500

500

nm
2
sample area.

h

Energy histograms of
U

and


derived from the tunneling spectra taken with a Cr-coated tip
over a

500

500

nm
2
sample area different from that in

g

. Here the values of the hard insulating gap and those of the PG appear to
merge, whereas the distribution of
U

appear to be comparable to that shown in

g

.
HUGHES
et al.
PHYSICAL REVIEW B
82
, 134441

2010

134441-4
For consistent comparison of the low-energy features
among three different types of spectra, we associate

+
and


or

+

and −


in the case of the
type

with the
energies where the derivatives of the low-energy tunneling
conductance reached the maximum, as shown in the insets of
Figs.
2

a

,
2

b

,
2

e

, and
2

f

. The histograms for all four
characteristic energies are combined in Figs.
2

c

and
2

d

for spectra taken with Pt/Ir tips over two different

500

500

nm
2
areas, and in Figs.
2

g

and
2

h

for spectra
taken with Cr-coated tips over two different

500

500

nm
2
areas. In general, we find that



0.6 eV and




0.4 eV at 77 K for data taken with Pt/Ir tips. The ap-
parent spatial variations in the characteristic energies are
manifestations of the intrinsically heterogeneous nature of
the manganites, even for the most conducting and homoge-
neous LCMO composition considered in this work.
While the characteristics of each type of spectra taken
with Pt/Ir- and Cr-coated tips were qualitatively similar, as
exemplified in Figs.
2

a

and
2

e

and also manifested by the
histograms shown in Figs.
2

c

and
2

g

, and Figs.
2

d

and
2

h

, careful inspections of the spectral details reveal quan-
titative differences. These differences may be attributed to
the energy-dependent DOS of Cr, which will be discussed
further in Sec.
IV
. Therefore, in the following discussion we
only refer to the data taken with Pt/Ir tips as representative of
the DOS of LCMO.
According to band-structure calculations,
10
,
11
the values
of
U

in the bulk DOS of LCMO are correlated directly with
the Ca-doping level and are well defined for a given Ca-
doping level
x
, as exemplified in Fig.
3

a

for specific doping
levels considered in band-structure calculations. Therefore,
the finite range of
U

values manifested by the histograms in
Figs.
2

c

,
2

d

,
2

g

, and
2

h

imply spatially varying Ca-
doping levels. Nonetheless, the dominating value of

U


varies between 1.0 and 1.2 eV, which correspond to local
Ca-doping levels between
x
=0.33 and
x
=0.25, in good
agreement with the nominal doping level of our sample
x
=0.3 if we estimate the Ca-doping level by assuming mono-
tonic
U

-vs-
x
dependence as shown in Fig.
3

a

. Addition-
ally, the apparent positive correlation between most regions
of the maps of
U
+
and

+
values over the same

500

500

nm
2
sample area, as shown in Figs.
3

b

and
3

c

and
further quantified by the map of cross correlations
38
between
U
+
and

+
in Fig.
3

d

, is suggestive of a common physical
origin for the spatial variation in
U
+
and

+
. In other words,
the intrinsic electronic heterogeneity in the manganites is
responsible for the empirical observation of spectral varia-
tions at low temperatures and in zero field. Further, from Fig.
3

a

we find that positive correlation between
U
+
and

+
implies that larger

+
values are associated with lower Ca-
doping levels

larger
U
+
values

.
As noted before, the sample areas studied with different
STM tips and at different temperatures are generally not
identical. Therefore, for meaningful comparison of the spec-
tral evolution with temperature and spin polarization, it is
necessary to establish statistical consistency of the spectral
characteristics obtained from one

500

500

nm
2
area with
those obtained from another

500

500

nm
2
area of the
same sample. Similarly, for different samples prepared under
the same fabrication conditions, it is also necessary to estab-
lish the same statistical consistency. In this context, we show
in Figs.
2

c

,
2

d

,
2

g

, and
2

h

the histograms of
U

and


obtained from the tunneling spectra on different sample
areas at 77 K with Pt/Ir- and Cr-coated tips. As evidenced by
the similarities of Figs.
2

c

and
2

d

, and Figs.
2

g

and
2

h

, the spectral characteristics over different

500

500

nm
2
areas appear to be statistically consistent for the
same type of STM tips and at the same temperature.
B. Temperature dependence
At
T
=300 K the LCMO epitaxial thin films studied in
this work were in the paramagnetic phase. Hence, the spec-
tral characteristics at
T
=300 K were quite different from
those observed in the bulk ferromagnetic state at
T
=77 K,as
shown in Figs.
4

a

and
4

b

, and Figs.
4

d

and
4

e

for
exemplified spectra taken at 300 K and with Pt/Ir- and Cr-
coated tips, respectively. Comparison of Figs.
4

a

and
4

b

,
and Figs.
4

d

and
4

e

with Figs.
2

a

and
2

b

, and Figs.
2

e

and
2

f

indicates several important contrasts. First, the
large DOS peaks associated with
U

for ferromagnetic
LCMO became much suppressed in the paramagnetic state
and the
U


values become much smaller than
U

. Second,
the surface insulating gap found around the Fermi level at 77
K either completely disappeared

Figs.
4

b

and
4

e

or be-
came a PG

Figs.
4

a

and
4

d

, as summarized by the his-
tograms of the PG



in Figs.
4

c

and
4

f

, where large
counts at both zero and the PG energies


+


are shown. The
vanishing gaps at 300 K for some of the spectra cannot be
accounted for by thermal smearing alone and are therefore
FIG. 3.

Color online

a

Ca-doping

x

dependence of the DOS
peak energies
U

based on known band-structure calculations

Refs.
10
and
11

showing decreasing
U
+
and increasing
U
with
increasing Ca doping. Apparent correlation between the spatial dis-
tribution of the DOS peak energy
U
+
and that of the surface insu-
lating gap

+
is manifested by comparing the

+
map in

b

and the
U
+
map

c

, both over the same

500

500

nm
2
sample area. The
correlation is further quantified by the map of cross correlation
between
U
+
and

+
in

d

Ref.
38

, showing positive correlation
throughout most of the sample area. A smaller fraction of anticor-
related regions occurs along the domain boundaries. These regions
are associated with the occurrence of PG and exhibit the
-type
spectra.
SPIN-POLARIZED TUNNELING SPECTROSCOPIC
...
PHYSICAL REVIEW B
82
, 134441

2010

134441-5
suggestive of a magnetic phase transition occurring at a
mean transition temperature between 77 and 300 K. On the
other hand, the nearly temperature-independent PG energies
are suggestive of a completely different physical origin.
Third, slight differences were found between the spectra
taken with the Pt/Ir tip and those taken with the Cr-coated tip
at 300 K, as manifested in Figs.
4

a

and
4

b

, and Figs.
4

d

and
4

e

. The differences occurred because spectra taken
with the former were representative of the DOS of LCMO in
the paramagnetic phase, whereas those taken with the latter
consisted of convoluted DOS of the paramagnetic LCMO
and the ferromagnetic Cr-coated tip.
In addition to the temperature-dependent spectral charac-
teristics, the spatial variations in the tunneling conductance
also revealed temperature-dependent evolution. Specifically,
the tunneling conductance in the paramagnetic state was gen-
erally more homogeneous than that in the ferromagnetic
state, because the tendency toward phase separations only
occurred in the ferromagnetic state of LCMO, as manifested
by the constant-bias tunneling conductance maps in Figs.
5

a

and
5

c

for room-temperature spectra taken at
=

U
+


with the Pt/Ir- and Cr-coated tips, respectively. In con-
trast, the tunneling conductance in the ferromagnetic state
was significantly more inhomogeneous, as exemplified in
Figs.
5

d

and
5

f

for tunneling conductance taken at 77 K
and for
=

U
+

. Here

U
+

is defined as the most commonly
occurring
U
+
values obtained from the histograms in Fig.
2
.
The statistical distributions of the conductance at
=

U
+

for
77 K and at
=

U
+


for 300 K are summarized by the his-
tograms in Figs.
5

b

and
5

e

for spectra taken with the
Pt/Ir- and Cr-coated tips, respectively. While the histograms
at
T
=300 K were statistically similar between spectra taken
with Pt/Ir- and Cr-coated tips as shown in Fig.
5

b

,at77K
the tunneling conductance distributions for spectra taken
with the Cr-coated tip revealed an overall shift toward higher
conductance than those taken with the Pt/Ir tip. The apparent
differences between the histograms obtained with Pt/Ir- and
Cr-coated tips from LCMO at 77 K are suggestive of differ-
ent effects associated with regular and spin-polarized tunnel-
ing into spatially inhomogeneous LCMO in its ferromagnetic
phase.
Similarly, the tunneling conductance maps for
=


+


taken at 300 K with Pt/Ir- and Cr-coated tips are shown in
Figs.
6

a

and
6

c

, respectively, whereas those for
=


+

taken at 77 K with Pt/Ir- and Cr-coated tips are shown in
Figs.
6

d

and
6

f

. These maps again reveal spatially more
homogeneous tunneling conductance in the paramagnetic
state. For completeness, the statistical distributions of the
tunneling conductance at
=


+


for
T
=300 K and
=


+

for
T
=77 K are summarized by the histograms in
Figs.
6

b

and
6

e

. Here


+

denotes the most commonly
occurring insulating gap value at positive bias from the his-
tograms in Fig.
2
for
T
=77 K and


+


represents the most
commonly found PG values from the histograms in Fig.
4
for
T
=300 K.
C. Magnetic field dependence
Although differences between the tunneling spectra taken
with Pt/Ir tips and those taken with Cr-coated tips are readily
FIG. 4.

Color online

Comparison of the tunneling spectral
characteristics taken with Pt/Ir- and Cr-coated tips at
T
=300 K and
H
=0.

a

A PG-like spectrum taken with a Pt/Ir tip, showing sig-
nificantly suppressed
U


values relative to the peak energies
U

found in the spectrum of Fig.
2

a

. The PG values



are deter-
mined from the peaks of the

d
2
I
/
dV
2

-vs-
V
spectrum.

b

Another
typical type of spectra taken with a Pt/Ir tip, showing vanishing
gaps as detailed in the inset.

c

Histograms of the PG values



and the characteristic energies
U


obtained by using a Pt/Ir tip over
a

500

500

nm
2
area at 300 K, showing suppressed
U


values
relative to the
U

values found at 77 K, as well as large counts of
vanishing gaps

shown by the arrows at
=0

and PG at
=





.

d

A typical PG-like spectrum taken with a Cr-coated tip, showing
suppressed
U


values relative to the peak energies
U

found in the
spectrum of Fig.
2

e

. The PG values



are determined from the
peaks of the

d
2
I
/
dV
2

-vs-
V
spectrum.

e

Another typical type of
spectra taken with a Cr tip, showing vanishing gaps as detailed in
the inset.

f

Histograms of PG values



and the characteristic
energies
U


obtained by using a Cr-coated tip over a

500

500

nm
2
area at 300 K, showing suppressed
U


values relative
to the
U

values found at 77 K, as well as large counts of vanishing
gaps

shown by the arrows at
=0

and PG at
=





.
FIG. 5.

Color online

Comparison of the high-bias tunneling
spectral characteristics taken with Pt/Ir- and Cr-coated tips at
H
=0.

a

A

500

500

nm
2
tunneling conductance map taken with a
Pt/Ir tip at
=

U
+


and
T
=300 K.

b

Histograms of the tunneling
conductance obtained by using a Pt/Ir tip and a Cr-coated tip at
T
=300 K and for
=

U
+


.

c

A

500

500

nm
2
tunneling conduc-
tance map taken with a Cr-coated tip at
=

U
+


and
T
=300 K.

d

A

500

500

nm
2
tunneling conductance map taken with a Pt/Ir
tip at
=

U
+

and
T
=77 K.

e

Histograms of the tunneling con-
ductance obtained by using a Pt/Ir tip and a Cr-coated tip at
T
=77 K and for
=

U
+

.

f

A

500

500

nm
2
tunneling conduc-
tance map taken with a Cr-coated tip at the characteristic energy
=

U
+

and
T
=77 K.
HUGHES
et al.
PHYSICAL REVIEW B
82
, 134441

2010

134441-6
visible at
H
=0 and for
T

T
C
LCMO
, additional magnetic field-
dependent investigations are necessary to provide better
quantitative understanding for spin-polarized tunneling in
LCMO. As described previously, the degree of spin polariza-
tion may be controlled by keeping
T

T
C
LCMO
and by apply-
ing oppositely directed magnetic fields with magnitudes sat-
isfying either the condition
H
C
LCMO


H


H
C
Cr
or

H


H
C
Cr
.
The applied fields
H
=−0.3 T and
H
=3.0 T chosen in this
work are consistent with the required conditions.
In Figs.
7

a

7

c

representative normalized spectra taken
in the same area with a Cr-coated tip and at
T
=6 K are
shown for
H
=0, −0.3, and 3.0 T. It is apparent that these
tunneling spectra evolved significantly with magnetic field
with the statistical field-dependent spectral evolution for a

250

90

nm
2
sample area summarized by the histograms
of the characteristic energies
U
+
and

+
in Figs.
7

d

and
7

e

. In particular, the nonmonotonic field dependence of

U
+

and


+

is noteworthy. Further, the normalized tunnel-
ing conductance map also revealed significant and nonmono-
tonic field dependence, as exemplified in Figs.
8

a

8

c

and
in Figs.
9

a

9

c

for spatially resolved tunneling conduc-
tance maps taken at the characteristic energies
=

U
+

and


+

, respectively, and for
H
=0, −0.3, and 3.0 T over the
same

250

90

nm
2
sample area. We find that the spatially
inhomogeneous conductance map at
H
=0 and for
=

U
+

became more homogeneous in finite fields, reaching overall
highest conductance for
H
=−0.3 T, as manifested statisti-
cally by the histograms of conductance in Figs.
8

a

8

c

for
H
=0, −0.3, and 3.0 T. Additionally, all conductance maps
taken at
=

U
+

appear to correlate with those at
=


+

when we compare Figs.
8

a

8

c

with Figs.
9

a

9

c

.
For comparison, similar magnetic field-dependent spec-
troscopic studies were conducted under the same conditions
with a Pt/Ir tip. The representative tunneling spectra taken in
the same area with a Pt/Ir tip at
T
=6 K and for
H
=0, −0.3,
and 3.0 T are shown in Figs.
10

a

10

c

. Overall these tun-
neling spectra exhibit different field-dependent evolution
when compared with the spectra taken with a Cr-coated tip,
as statistical manifested by the histograms of the character-
istic energies
U
+
and

+
in Figs.
10

d

and
10

e

. In particu-
FIG. 6.

Color online

Comparison of the low-bias tunneling
spectral characteristics taken with Pt/Ir- and Cr-coated tips at
H
=0.

a

A

500

500

nm
2
tunneling conductance map taken with a
Pt/Ir tip at
=


+


and 300 K.

b

Histograms of the tunneling
conductance obtained by using a Pt/Ir tip and a Cr-coated tip at 300
K for
=


+


.

c

A

500

500

nm
2
tunneling conductance map
taken with a Cr-coated tip at
=


+


and for
T
=300 K.

d

A

500

500

nm
2
tunneling conductance map taken with a Pt/Ir tip
at
=


+

and
T
=77 K.

e

Histograms of the tunneling conduc-
tance obtained by using a Pt/Ir tip and a Cr-coated tip at 77 K and
for
=


+

.

f

A

500

500

nm
2
tunneling conductance map
taken with a Cr-coated tip at the characteristic energy
=


+

and
for
T
=77 K.
FIG. 7.

Color online

Comparison of magnetic field-dependent
spectral characteristics taken with a Cr-coated tip at
T
=6 K.

a

Normalized conductance

dI
/
dV

/

I
/
V

G
̄
vs bias voltage

V

spectrum for
H
=0.

b

G
̄
vs
V
spectrum for
H
=−0.3 T.

c

G
̄
vs
V
spectrum for
H
=3.0 T.

d

Histograms of the characteristic energy
U
+
over the same

250

90

nm
2
sample area for
H
=0, −0.3, and
3.0 T.

e

Histograms of

+
over the same

250

90

nm
2
sample
area at
H
=0, −0.3, and 3.0 T.

f

Temperature evolution of the
histograms of

+
over a

250

90

nm
2
sample area at
H
=0,
showing downshifts in the insulating gap values with increasing
temperature. Specifically, two types of gap values a
t6Kmaybe
attributed to an insulating surface gap and a pseudogap. The former
vanishes and the latter persists at 300 K.
FIG. 8.

Color online

Comparison of magnetic field-dependent
high-bias conductance maps

=

U
+

taken over the same

250

90

nm
2
sample area with a Cr-coated tip at
T
=6 K.

a

Normal-
ized conductance map for
H
=0.

b

Normalized conductance map
for
H
=−0.3 T.

c

Normalized conductance map for
H
=3.0 T.

d

Histograms of the normalized conductance at
=

U
+

and for
H
=0, −0.3, and 3.0 T, showing highest mean conductance at
H
=
−0.3 T when the spin polarization of the tunneling currents is an-
tiparallel to the magnetization of LCMO.
SPIN-POLARIZED TUNNELING SPECTROSCOPIC
...
PHYSICAL REVIEW B
82
, 134441

2010

134441-7
lar, we note that the histograms of
U
+
values appear to shift
monotonically up to higher energies with the increasing mag-
nitude of magnetic field. Similarly, the insulating gap values

+
appear to shift up monotonically and the distributions
become sharper with the magnitude of increasing magnetic
field. The spectral characteristics obtained with Pt/Ir tips are
therefore only dependent on the magnitude of applied fields,
which are in stark contrast to the spectra obtained with Cr-
coated tips that are strongly dependent on the direction of the
applied magnetic field. Additionally, the tunneling conduc-
tance
G
̄
taken with Pt/Ir tips at
=

U
+

generally increases
with increasing

H

, as exemplified by the conductance maps
in Figs.
11

a

11

c

for
H
=0, −0.3, and 3.0 T and also sum-
marized by the conductance histograms in Fig.
11

d

.
The general trend of increasing tunneling conductance
and spatial homogeneity at
=

U
+

with increasing mag-
netic field for data taken with the Pt/Ir tip is consistent with
the CMR nature of the manganites because better alignment
of magnetic domains and increasing mobility with increasing
magnetic field results in enhanced electrical conductance
across different magnetic domains. We further note that the
increasing spatial homogeneity in the tunneling conductance
with increasing

H

also agrees with previous STS reports on
LCMO,
12
although previous reports focused on tunneling
conductance studies near the Curie temperature and only at
one constant energy
=3.0 eV. Additionally, we find that
the conductance
G
̄
at
=


+

decreases
with increasing

H

,
as shown in Figs.
12

a

12

c

. The opposite field dependence
of
G
̄


for
=


+

to that for
=

U
+

suggests that
U
+
and

+
are associated with different characteristics of the LCMO
sample. This point will be elaborated in the following analy-
sis and discussion.
The apparent contrasts between aforementioned field-
dependent spectra taken a
t 6 K with Cr-coated tips and those
taken with Pt/Ir tips, as shown in Figs.
7
12
, are all consis-
tent with spin-polarization tunneling in the former. To
achieve more quantitative understanding of the spin-
FIG. 9.

Color online

Comparison of magnetic field-dependent
low-bias conductance maps

=


+

taken over the same

250

90

nm
2
sample area with a Cr-coated tip at
T
=6 K.

a

Normal-
ized conductance map for
H
=0.

b

Normalized conductance map
for
H
=−0.3 T.

c

Normalized conductance map for
H
=3.0 T.

d

Histograms of the normalized conductance at


+

at
H
=0, −0.3,
and 3.0 T, showing lowest mean conductance at
H
=−0.3 T when
the spin polarization of the tunneling currents is antiparallel to the
magnetization of LCMO.
FIG. 10.

Color online

Comparison of magnetic field-
dependent spectral characteristics taken with a Pt/Ir tip at
T
=6 K.

a

Normalized conductance
G
̄

dI
/
dV

/

I
/
V

vs energy


spec-
trum for

a

H
=0 T,

b

H
=−0.3 T, and

c

H
=3.0 T.

d

Histo-
grams of the characteristic energies
U
+
obtained from the same

250

28

nm
2
sample area under
H
=0, −0.3, and 3.0 T, showing
monotonic shift toward the right with increasing

H

.

e

Histograms
of surface gap energies

+
obtained from the same

250

28

nm
2
sample area under
H
=0, −0.3, and 3.0 T, showing
monotonic increase in


+

with increasing

H

.

f

Comparison of
the zero-field temperature evolution of

+
, showing a wide distri-
bution of insulating gap values a
t 6 K and a persistent pseudogap at

+


0.4 eV at 300 K.
FIG. 11.

Color online

Comparison of magnetic field-dependent
high-bias tunneling conductance maps taken at
=

U
+


1336 meV over the same

250

28

sample area with a Pt/Ir tip
at
T
=6 K and for

a

H
=0 T,

b

H
=−0.3 T, and

c

H
=3.0 T,
showing monotonic increase in the homogeneity and the value of
G
̄
with increasing

H

.

d

Histograms of the normalized tunneling
conductance
G
̄
at
=

U
+

at
H
=0, −0.3, and 3.0 T.
HUGHES
et al.
PHYSICAL REVIEW B
82
, 134441

2010

134441-8