supporting online material for - science · supporting online material for a perovskite oxide...
TRANSCRIPT
www.sciencemag.org/cgi/content/full/science.1212858/DC1
Supporting Online Material for
A Perovskite Oxide Optimized for Oxygen Evolution Catalysis from Molecular Orbital Principles
Jin Suntivich, Kevin J. May, Hubert A. Gasteiger, John B. Goodenough, Yang Shao-Horn*
*To whom correspondence should be addressed. E-mail: [email protected]
Published 26 October 2011 on Science Express DOI: 10.1126/science.1212858
This PDF file includes: Materials and Methods
Figs. S1 to S7
Tables S1 to S5
Schemes S1 and S2
References (30–47)
2
Materials and Methods
Material synthesis. The perovskites oxides were synthesized with a co-precipitation method as
described previously (26) with the exception of BSCF, which was synthesized with a nitrate
combustion method. In the BSCF synthesis, alkaline earth and transition-metal nitrates (all
99.998+% Sigma-Aldrich) were prepared at 0.2 M concentration, to which glycine (>99%
Sigma-Aldrich) was added at 0.1 M concentration. The mixture was heated until vigorously
evaporating, at which point, sparks are emitted as a result of combustion; it was then calcined at
1100°C under air atmosphere for 24 hours. All samples were found to be in phase-pure form
from X-ray diffraction with the exception of LaNiO3 (~2% NiO). All structure parameters,
particle size distributions, estimated surface areas, and estimation of eg-electron filling of surface
transition-metal ions are listed in tables S3-5.
Surface area estimation of oxides. Particle sizes and surface areas were quantified to determine
mass specific surface areas, As, using a spherical geometry approximation:
𝐴! ≈∑𝜋𝑑!
∑ 1 6 𝜌𝜋𝑑! =6𝜌∑𝑑!
∑𝑑! =6
𝑑!/!𝜌
where ρ is the oxide bulk density, d is the diameter of individual particles determined from
scanning electron micrographs (FEI-Phillips XL30 and JEOL 6320 FV), and dv/a is the
volume/area averaged diameter. The value of As for non-supported oxide catalysts was also
estimated via nitrogen Brunauer, Emmett, and Teller (BET) area measurements for LaNiO3 and
BSCF, which had the values of ~5.6 m2 g−1 and ~0.4±0.2 m2 g−1 (Micromeritics 2020, error bar
based on three independent BET measurements), respectively. As the BET areas were
reasonably consistent with the SEM-based value of ~3.5 m2 g−1 and ~0.2 m2 g−1, it is believed
that a comparison of the specific activity of the various oxides using the SEM-based As values is
3
reasonable. We however caution that the accuracy of such an estimate in our experience is only
within a conservative factor of ~3.
Electrochemical characterization. The oxide electrode consisted of perovskite oxides, acetylene
black (Chevron), and neutralized Nafion® (Ion power); they were prepared at a loading of 0.25
mgox cm-2disk, 0.05 mgAB cm-2
disk, and 0.05 mgNafion cm-2disk as described elsewhere (18).
Electrochemical measurements were conducted with a rotating-disk electrode (Pine) using a
VoltaLab PST050 potentiostat. The electrolyte was prepared from Milli-Q water (18 MΩ•cm)
and KOH pellets (99.99% weight, Sigma-Aldrich). All measurements were collected under O2
saturation (ultra-high-grade purity, Airgas) to ensure the O2/H2O equilibrium at 1.23V vs. RHE
at a rotation rate of 1600 rpm. The analysis of OER kinetic currents was capacitance-corrected
by taking the average between positive and negative-going scans, and then iR-corrected (fig. S6).
The specific OER activity was obtained from normalizing the kinetic current by the surface area
of each oxide estimated from particle size distribution. Given the limitation of the accuracy of
the surface area measurement (a factor of 3) and the thin-film loading (additional factor of 3), we
estimate the intrinsic error associated with our assessment of the intrinsic activity to be
approximately an order of magnitude. The rotating ring disk electrode (Pine) with Pt-ring was
calibrated using the same rotating ring disk electrode configuration with a thin-film IrO2 as O2
evolution catalyst (N ~ 0.19) and a glassy carbon electrode with a potassium ferricyanide couple
(N ~ 0.21). Generated O2 was collected at a ring potential of 0.4 V vs. RHE.
Synthesis of rutile IrO2 nanoparticles. IrO2 were synthesized in a 2-step reaction. Briefly, a
solution of either IrCl4·xH2O in tetralin and oleylamine was prepared at room temperature and
then heated to 60 °C under Ar atmosphere for 20 minutes, followed by an injection of
tetrabutylammonium borohydride in chloroform. The reaction mixture was heated up to 200 °C
4
for 19 hours. Afterward, the nanoparticles were precipitated by ethanol addition and collected by
centrifugation at room temperature before re-dispersing in hexane, which was followed by a 500
°C heat-treatment for 20 hrs under O2 atmosphere. The sample was found to be in rutile (P 42/m
n m) phase. The surface area was estimated from transmission electron micrographs (JEOL
2010) using the same particle distribution calculation as the perovskites.
Determination of eg filling of LaNi0.5Mn0.5O3 and LaCu0.5Mn0.5O3. The eg fillings of
LaNi0.5Mn0.5O3 and LaCu0.5Mn0.5O3 are complicated by the fact that these compounds undergo
charge-disproportionation between Mn and Ni/Cu atoms. To estimate the eg filling, we track the
oxidation states of each with X-ray Absorption Spectroscopy (XANES.) From a known-
relationship where E0 (maximum inflection point of the absorption edge) scales with oxidation
state (30), we estimate the Mn electronic configuration to be t2g3eg
0.3 for both LaNi0.5Mn0.5O3 and
LaCu0.5Mn0.5O3, and Ni to be t2g6eg
1.7 for LaNi0.5Mn0.5O3 and Cu to be t2g6eg
2.7 for LaNi0.5Mn0.5O3.
When we apply the eg filling from each B-site atom to the OER volcano plot, we have found that
the use of eg filling from Ni is most consistent with the observed activity for LaNi0.5Mn0.5O3 and
Mn for LaCu0.5Mn0.5O3 (fig. 2). Had we used the eg filling of average B-site atoms, or Mn for
LaNi0.5Mn0.5O3 or Cu for LaCu0.5Mn0.5O3, the eg would have resulted in an underestimation of the
OER activity. The consistency eg filling selection with our volcano plot also leads us to propose
that the active site for the OER is the Ni atom for LaNi0.5Mn0.5O3 and the Mn atom for
LaCu0.5Mn0.5O3, where the Mn atom in LaNi0.5Mn0.5O3 has too little eg electron filling and Cu
atoms in LaCu0.5Mn0.5O3 has too many eg electrons, rendering their catalytic properties inactive.
Determination of eg filling of Ba0.5Sr0.5Co0.8Fe0.2O3-δ compound. The eg filling for the mixed
A-site and B-site BSCF is highly complicated by the presence of four distinct cations and a large
concentration of oxygen vacancies inside the host structure. We measure the oxygen vacancy
5
concentration with thermogravimetry (TGA), where we found δTGA to be ~0.4. While this δTGA
is in agreement with the literature value (31), it alone still cannot determine the oxidation state of
Co and Fe atoms. Therefore we resort to the use of XANES to estimate the Co oxidation state.
Using CoO (Sigma-Aldrich) and LaCoO3 as model compounds, we estimate the oxidation state
of the Co to be ~2.8+ (fig. S7). Considering the previous works of Harvey et al. (27-28) and
Arnold et al. (32), whose XANES and X-ray Photoemission Spectroscopy studies have shown
that the Co oxidation state in BSCF is partially reduced below 3+ and that the Fe oxidation state
is partially oxidized above 3+, our estimated oxidation state is therefore not unexpected. Taking
into account that the surface Co state in BSCF has been reported to be in an intermediate spin
state (28), similar to the case of LaCoO3, the electronic configuration of the Co in BSCF is likely
t2g5eg
~1.2.
Ball-milling of Ba0.5Sr0.5Co0.8Fe0.2O3-δ compound. The as-synthesized BSCF powder was ball-
milled with a Fritsch pulverisette 6 planetary mill. A 5:2 mass ratio of 1 mm zirconia milling
beads to BSCF was prepared with 10 mL of N-methyl-2-pyrrolidinone (NMP) solvent. The
materials were milled in a zirconia crucible at 500 rpm for 8 hours, with a 30-minute rest period
every one-hour. The BSCF-NMP mixture was then dried in a furnace at 195 °C in air, and then
at 240 °C in vacuum overnight to evaporate off any remaining solvent. The powder was then
finely ground by a mortar and pestle before electrochemical characterization. We caution,
however, that this ball-milling method does not result in a sufficient miniaturization of the BSCF
particle; a strategy based on nanoscale oxide synthesis will likely produce more uniform,
high-surface-area oxides. We studied the ball-milling sample at both 0.25 mgox cm-2disk, and 0.05
mgox cm-2disk to ensure that there is no transport limitation in our measurement. The mass and
specific activities were found to be the same regardless of the loading.
6
Figure S1. (A-B) Tafel plot of the OER specific activity of all
oxides studied in this work. All the activities are iR-corrected,
and normalized to the catalyst surface area (listed in Table S5).
7
Figure S2. The role of B-O covalency on the OER activity of
perovskite oxides. The potentials at 50 µA/cm2ox as a function
of the B-O covalency, which is estimated by the normalized soft
X-ray absorbance at eg filling = 1. Error bars represent standard
deviations of at least 3 measurements. The B-O covalency data
was from reference (26).
8
Figure S3. X-ray Diffraction Pattern of Ba0.5Sr0.5Co0.8Fe0.2O3-δ.
The experiment was conducted at room temperature using Cr
source. The pattern revealed a space group of P m -3 m (simple
cubic) with a lattice parameter of 3.99 Å.
9
Figure S4. Preliminary stability result for (A) BSCF and (B)
ball-milled BSCF. No OER activity degradation was observed
within the timeframe of the experiment (~1-2 hours). The
incremental activity of ball-milled BSCF over time is attributed to
the removal of remaining ball-milling solvent at longer period.
10
Figure S5. Role of eg electron on the ORR/OER catalysis.
Overpotentials at 50 µA cm-2ox for (A) OER and (B) ORR as a
function of eg-filling. Symbols vary with B ions (Cr, red;, Mn,
orange; Fe, beige; Co, green; Ni, blue; mixed B, purple). Error
bars represent standard deviations of at least three independent
measurements. The grey lines are for guiding purpose only.
11
Figure S6. Ohmic and capacitive corrections of the
as-measured OER activity of example perovskite oxide
catalyst. The as-measured OER activity of the La0.5Ca0.5CoO3-δ
thin-film composite catalyst (“raw”, solid red line) is
capacity-corrected by taking an average of forward and backward
(positive and negative-going) scans. The capacity-corrected
OER current (short-dash blue line) is then ohmically corrected
with the measured ionic resistance (≈45 Ω) to yield the final
electrode OER activity (long-dash green line).
12
Figure S7. Co K-edge XANES of CoO, LaCoO3, and BSCF.
(A) The Co K-edge spectra. (B) The edge position (E0, defined
to be the energy at the highest first derivative of the absorbance)
of BSCF fits right in between CoO and LaCoO3. By using the
known metal oxidation vs. E0 relationship, we estimate the Co
oxidation state to be ~2.8+.
13
Table S1. Benchmark OER activities of various state-of-the-art oxide catalysts.
All the OER activities were ohmic-corrected at 0.4 V overpotential. Specific activity is
normalized to the catalyst specific area following the methodology as described elsewhere (18).
Compounds (electrolyte)
Electrode i (mA cm-2
disk) Loading
(per cm-2disk)
Mass i (A g-1
ox) Specific i
(μA cm-2ox)
IrO2 (Nafion® (33)) ~500 2 mgIrO2 ~250 n.a.
RuO2 (1 M H2SO4 (34)) ~5-1,000 n.a. n.a. ~200
NiCo2O4 (30 w/o KOH (35)) ~300 10 mgNiCo2O4 ~30 ~60
CoxMn3-xO4 (CoMnO-B)
(0.1 M KOH (36))
0.3 mA (area not given)
0.025 mgoxide (area not given)
~12 ~11 (using 112 m2 g-1)
La-doped Co3O4 (1M KOH (37))
~8,000 (estimated*) ~3 mgLa-Co3O4
†
~2,700† (estimated*)
~2,200‡
(estimated*)
Co-phosphate (0.1 M KBi, pH 9 (38)) 0.1-1.5 0.1-2.7 μmolCo
~10 (A g-1
Co) n.a.
IrO2 (0.1M KOH) ~18 0.05 mgIrO2 ~360 ~500
BSCF (0.1M KOH) ~20 0.25 mgBSCF ~80 ~40,000
Ball-milled BSCF (0.1M KOH)
~100 (estimated) 0.25 mgBSCF
~500 (estimated)
~20,000 (estimated)
* Extrapolated activity from last 5 data points (overpotential ~ 0.2 – 0.25V). † This mass activity value reflects only the mass change of the Ni substrate before and after La-
Co3O4 loading. This calculation does not include the possibility that Ni substrate can react with Co3O4 during electrode “drying” at 350 °C (1.5 hours) to form additional NiCo2O4 phase, which will increase the amount of active catalyst.
‡ We use the reported Rf value for specific activity normalization as the BET value (25.4 m2 g-1) does not capture additional active-phase and area formation during the Co3O4 – Ni electrode heat-treatment. Had we used the BET value (25.4 m2 g-1), specific i would be ~ 11 mA cm-2
ox.
14
Table S2. Benchmark OER activities of various state-of-the-art perovskite oxide catalysts.
All the OER activities were ohmic-corrected and normalized to the actual surface area at 0.4 V
overpotential in basic electrolyte (>13 pH). Methodology for specific activity normalization is
given elsewhere (18).
Compounds (reference) Electrode i (mA cm-2
disk) Loading
(mg cm-2disk)
Mass i (A g-1)
Specific i (μA cm-2
ox)
LaNiO3 (Bockris & Otagawa (39))
~20-2,000 (extrapolated) n.a. n.a. 30-360
(extrapolated)
La1-xSrxCoO3 (Bockris & Otagawa (39)) ~2-40 n.a n.a ~13-50
La1-xSrxFeO3 (Bockris & Otagawa (39)) ~2-10 n.a n.a 3-30
La1-xSrxMnO3 (Bockris & Otagawa (39)) ~0.1-1 n.a n.a 0.2-1
La1-xSrxCoO3 (Matsumoto et al. (18)) ~3-10 n.a n.a n.a.
La1-xSrxFeO3 (Wattiaux et al. (20)) ~0.1-2.5 n.a n.a n.a.
La1-xSrxMnO3 (Matsumoto & Sato (19)) ~0.2-0.6 n.a n.a n.a.
LaNiO3 (this work) ~3.5 0.25 ~14 ~370
La1-xCaxCoO3 (this work) ~0.5-3 0.25 ~2-15 ~250-800
La1-xCaxFeO3 (this work) ~0.2-4 0.25 ~0.8-16 ~90-1,500
La1-xCaxMnO3 (this work) ~0.03-0.4 0.25 ~0.12-1.6 ~20-100
15
Table S3. Summary of literature information on the spin state of the perovskite oxides
Valence Spin state Assignment Example of reference
LaCrO3 Cr3+ n/a t2g3 LS and HS identical
LaMnO3 Mn3+ H.S. t2g3 eg
1 Magnetization (40)
LaFeO3 Fe3+ H.S. t2g3 eg
2 Mossbäuer (41)
LaCoO3 Co3+ I.S. t2g5 eg
1 Magnetization (42)
LaNiO3 Ni3+ L.S. t2g6 eg
1 Extrapolation from RNiO3 magnetization (43, 44)
LaNi0.5Mn0.5O3 Mn3.7+ Ni2.3+
H.S. (Mn) L.S. (Ni)
t2g3 eg
0.3 (Mn)
t2g6 eg
1.7 (Ni)
See Supplementary Method
LaCu0.5Mn0.5O3 Mn3.7+ Cu2.3+
H.S. (Mn) L.S. (Cu)
t2g3 eg
0.3 (Mn)
t2g6 eg
2.7 (Cu) See Supplementary
Method
La0.5Ca0.5MnO3- δ (Assume δ = 0) Mn3.5+ H.S. t2g
3 eg0.5 Magnetization (40),
X-ray Emission (45)
La0.5Ca0.5FeO3- δ (Assume δ = 0) Fe3.5+ H.S. t2g
3 eg1.5 Mossbäuer (41, 46)
La0.75Ca0.25FeO3-δ (Assume δ = 0) Fe3.25+ H.S. t2g
3 eg1.75 Mossbäuer (41, 46)
La0.5Ca0.5CoO3-δ (δTGA = 0.21) Co3+ I.S./H.S. t2g
4.5 eg1.5 La1-xCaxCoO3-δ
extrapolation (47)
LaMnO3+δ (Assume δ = 0.1) Mn3.2+ H.S. t2g
3 eg0.8 X-ray Emission (45)
Ba0.5Sr0.5Co0.8Fe0.2O3-δ (δTGA = 0.40)
Co2.8+ Fe>3+
I.S. (Co) H.S. (Fe)
t2g5 eg
1.2 (Co)
t2g3 eg
<2 (Fe) X-ray Photoemission and
Absorption (32)
16
Table S4. The derived lattice parameters of the oxide model compounds in this work
Space group a(Å) b(Å) c(Å)
LaCrO3 P n m a 5.48 7.76 5.51
LaMnO3 P n m a 5.66 7.72 5.53
LaFeO3 P n m a 5.57 7.85 5.56
LaCoO3 R -3 c 5.44 5.44 13.09
LaNiO3 R -3 c 5.46 5.46 13.14
LaNi0.5Mn0.5O3 P n m a 5.46 7.74 5.51
LaCu0.5Mn0.5O3 P n m a 5.48 7.77 5.52
La0.5Ca0.5MnO3- δ P n m a 5.42 7.65 5.43
La0.5Ca0.5FeO3- δ P n m a 5.55 7.84 5.55
La0.75Ca0.25FeO3-δ P n m a 5.52 7.80 5.52
La0.5Ca0.5CoO3-δ P n m a 5.41 7.59 5.36
LaMnO3+δ R -3 c 5.52 5.52 13.35
Ba0.5Sr0.5Co0.8Fe0.2O3-δ P m -3 m 3.99 3.99 3.99
17
Table S5. Characterizations of the oxides studied in this work.
The number averaged diameter, dnumber, the volume-area averaged diameter, dv/a, and the specific
surface area, As, were obtained from particle size distribution measurements as described
elsewhere (18). Note that we also measured the BET area for LaNiO3 and BSCF, which were
found to be within a factor of 2 of the specific area determined from SEM.
dnumber (µm) dv/a (µm) As (m2 g-1)
LaCrO3 0.64 (±0.25) 0.83 1.1
LaMnO3 1.05 (±0.52) 1.51 0.6
LaFeO3 0.71 (±0.34) 1.01 0.9
LaCoO3 0.78 (±0.40) 1.10 0.7
LaNiO3 0.20 (±0.06) 0.24 3.5
LaNi0.5Mn0.5O3 0.34 (±0.11) 0.81 1.1
LaCu0.5Mn0.5O3 0.58 (±0.28) 0.80 1.1
La0.5Ca0.5MnO3- δ 0.92 (±0.44) 0.62 2.1
La0.5Ca0.5FeO3- δ 0.62 (±0.31) 0.89 1.1
La0.75Ca0.25FeO3-δ 0.36 (±0.22) 0.59 1.8
La0.5Ca0.5CoO3-δ 0.43 (±0.23) 0.63 1.6
LaMnO3+δ 1.39 (±0.58) 1.81 0.5
Ba0.5Sr0.5Co0.8Fe0.2O3-δ 0.84 (±1.27) 7.01 0.2
Ba0.5Sr0.5Co0.8Fe0.2O3-δ (ball-milled)
0.23 (±0.09) 0.30 3.9
18
Scheme S1. The geometry of eg orbital in octahedral
metal-oxygen complex (15-16). In this configuration, eg-
symmetry parentage electron serves as the σ* orbital electron.
The dz2 symmetry orbital is shown as an example.
19
Scheme S2. Proposed OER mechanism on perovskite
transition-metal oxide catalysts. Both (A) the OER and (B) the
ORR proceeds via 4 electron transfer steps. In the OER case,
the RDS are the O-O bond formation (2 in OER) and the proton
extraction of the oxy-hydroxide group (3 in OER). For the ORR,
the RDS are either the surface hydroxide displacement (4 in
OER) or the surface hydroxide regeneration (1 in OER).
1
Additional References
30. I. Arcon, B. Mirtic, A. Kodre, Determination of valence states of chromium in calcium chromates by using X-ray absorption near-edge structure (XANES) spectroscopy. J. Am. Ceram. Soc. 81, 222 (1998). doi:10.1111/j.1151-2916.1998.tb02319.x
31. J. Jung, S. T. Misture, D. D. Edwards, Oxygen stoichiometry, electrical conductivity, and thermopower measurements of BSCF (Ba0.5Sr0.5CoxFe1–xO3–δ, 0 ≤ x ≤ 0.8) in air. Solid State Ion. 181, 1287 (2010). doi:10.1016/j.ssi.2010.06.033
32. M. Arnold, Q. Xu, F. D. Tichelaar, A. Feldhoff, Local charge disproportion in a high-performance perovskite. Chem. Mater. 21, 635 (2009). doi:10.1021/cm802779f
33. E. Rasten, G. Hagen, R. Tunold, Electrocatalysis in water electrolysis with solid polymer electrolyte. Electrochim. Acta 48, 3945 (2003). doi:10.1016/j.electacta.2003.04.001
34. L. D. Burke, O. J. Murphy, J. F. Oneill, S. Venkatesan, The oxygen electrode. Part 8. Oxygen evolution at ruthenium dioxide anodes. J. Chem. Soc. Faraday Trans. 73, 1659 (1977).
35. C. R. Davidson, G. Kissel, S. Srinivasan, Electrode kinetics of the oxygen evolution reaction at NiCo2O4 from 30% KOH: Dependence on temperature. J. Electroanal. Chem. 132, 129 (1982). doi:10.1016/0022-0728(82)85012-2
36. F. Y. Cheng et al., Rapid room-temperature synthesis of nanocrystalline spinels as oxygen reduction and evolution electrocatalysts. Nat. Chem. 3, 79 (2011). doi:10.1038/nchem.931 Medline
37. R. N. Singh et al., Novel electrocatalysts for generating oxygen from alkaline water electrolysis. Electrochem. Commun. 9, 1369 (2007). doi:10.1016/j.elecom.2007.01.044
38. A. J. Esswein, Y. Surendranath, S. Y. Reece, D. G. Nocera, Highly active cobalt phosphate and borate based oxygen evolving catalysts operating in neutral and natural waters. Energy Environ. Sci. 4, 499 (2011). doi:10.1039/c0ee00518e
39. J. O. Bockris, T. Otagawa, Mechanism of oxygen evolution on perovskites. J. Phys. Chem. 87, 2960 (1983). doi:10.1021/j100238a048
40. C. Ritter et al., Influence of oxygen content on the structural, magnetotransport, and magnetic properties of LaMnO3+δ. Phys. Rev. B 56, 8902 (1997). doi:10.1103/PhysRevB.56.8902
41. U. Russo, L. Nodari, M. Faticanti, V. Kuncser, G. Filoti, Local interactions and electronic phenomena in substituted LaFeO3 perovskites. Solid State Ion. 176, 97 (2005). doi:10.1016/j.ssi.2004.06.001
42. J. Q. Yan, J. S. Zhou, J. B. Goodenough, Bond-length fluctuations and the spin-state transition in LCoO3 (L = La, Pr, and Nd). Phys. Rev. B 69, 134409 (2004). doi:10.1103/PhysRevB.69.134409
43. M. L. Medarde, Structural, magnetic and electronic properties of RNiO3 perovskites (R = rare earth). J. Phys. Condens. Matter 9, 1679 (1997). doi:10.1088/0953-8984/9/8/003
2
44. J. S. Zhou, J. B. Goodenough, B. Dabrowski, P. W. Klamut, Z. Bukowski, Enhanced susceptibility in LNiO3 perovskites (L = La, Pr, Nd, Nd0.5Sm0.5). Phys. Rev. Lett. 84, 526 (2000). doi:10.1103/PhysRevLett.84.526 Medline
45. T. A. Tyson et al., Valence state of Mn in Ca-doped LaMnO3 studied by high-resolution Mn K-beta emission spectroscopy. Phys. Rev. B 60, 4665 (1999). doi:10.1103/PhysRevB.60.4665
46. J. Li, Investigation of orthorhombic perovskite La1–xCaxFeO3–y (0 ≤ x ≤ 0.5). Phys. Scr. 45, 62 (1992). doi:10.1088/0031-8949/45/1/011
47. J. Mastin, M. A. Einarsrud, T. Grande, Crystal structure and thermal properties of La1–
xCaxCoO3–δ (0 ≤ x ≤ 0.4). Chem. Mater. 18, 1680 (2006). doi:10.1021/cm052547o