Alexander
Schmid
*,
Ghislain M.
Rupp
and
Jürgen
Fleig
Institute of Chemical Technologies and Analytics, Vienna University of Technology, Getreidemarkt 9, Vienna, A-1060, Austria. E-mail: alexander.e164.schmid@tuwien.ac.at
First published on 5th April 2018
La0.6Sr0.4FeO3−δ (LSF) thin films of different thickness were prepared by pulsed laser deposition on yttria stabilized zirconia (YSZ) and characterized by using three electrode impedance spectroscopy. Electrochemical film capacitance was analyzed in relation to oxygen partial pressure (0.25 mbar to 1 bar), DC polarization (0 m to −600 m) and temperature (500 to 650 °C). For most measurement parameters, the chemical bulk capacitance dominates the overall capacitive properties and the corresponding defect chemical state depends solely on the oxygen chemical potential inside the film, independent of atmospheric oxygen pressure and DC polarization. Thus, defect chemical properties (defect concentrations and defect formation enthalpies) could be deduced from such measurements. Comparison with LSF defect chemical bulk data from the literature showed good agreement for vacancy formation energies but suggested larger electronic defect concentrations in the films. From thickness-dependent measurements at lower oxygen chemical potentials, an additional capacitive contribution could be identified and attributed to the LSF|YSZ interface. Deviations from simple chemical capacitance models at high pressures are most probably due to defect interactions.
![]() | (1) |
The equilibrium defect chemistry of LSCF has already been investigated on bulk samples by different methods, for example, coulometric titration, thermogravimetry, statistical thermodynamics calculations, carrier gas titration and electronic conductivity measurements.5,20–28 Much less data exist on the defect chemical relations upon electrochemical polarization, that is, under voltage bias.5,10,11 Moreover, the exact defect chemistry of LSCF thin films has hardly been investigated so far; thin film defect chemistry is not necessarily the same as for bulk materials, e.g. due to possible strain or interfacial effects. One experimental method particularly suited for investigating the defect chemistry of thin films is the measurement of the chemical capacitance, a capacitive property of mixed ionic and electronic conductors (MIECs) that depends on the charge carrier concentrations. This chemical capacitance reflects the ability of an oxide to change its stoichiometry in response to a change in oxygen chemical potential in the material.29 In the case of dilute defects and only one relevant electronic defect (eon), the chemical capacitance is given by30
![]() | (2) |
This equation already shows the direct relation between charge carrier (i.e. defect) concentration and chemical capacitance. Detailed studies of chemical capacitances were performed on ceria thin films under reducing conditions.31–34 Some data exist on LSF in air and humidified hydrogen,7,35 on LSCF in air,2,35 and on La0.6Sr0.4CoO3−δ in the range between 0.25 and 1 bar oxygen and under polarization.5,12,35 However, the chemical capacitance measurements of LSCF with in-depth quantitative analysis are still missing. More generally, chemical capacitance measurements are not yet an established routine tool for defect chemical investigations of thin films.
In this study, we use electrochemical impedance spectroscopy to investigate the defect chemistry of La0.6Sr0.4FeO3−δ thin films via their chemical capacitance. A three-electrode approach allows variation of the electrode polarization (overpotential) and, together with variation of the oxygen partial pressure, defect chemical data become accessible over an oxygen chemical potential range spanning the equivalent of 15 orders of magnitude of oxygen partial pressure. Thickness dependent capacitance measurements reveal the existence of an interfacial capacitance in addition to the bulk chemical capacitance of the film. Defect formation enthalpies and entropies can be compared with those of dilute bulk defect models from the literature. Differences between measured thin film chemical capacitances and values predicted by literature data are discussed.
The low frequency semicircle shows a clear dependence on oxygen pressure and DC voltage. It can be fitted to a parallel R-CPE element and a series resistance, where the constant phase element (CPE) is used to model an imperfect capacitor with impedance ZCPE = Q−1(iω)−P, see the circuit in Fig. 2.37 Experimental P-values close to one (typically between 0.91 and 1) indicate nearly ideal capacitive behavior. For certain measurement parameters, especially at low oxygen partial pressure, only part of the low frequency semicircle was accessible by the frequency range employed here. However, the capacitive contribution could always be determined from the measured data points with small fit errors (typically <1%). From the fit results (Rpol, Q, P), the area-specific capacitance C was calculated by
![]() | (3) |
Hence, the dominating semicircle feature of the spectrum is interpreted in accordance with similar spectra found in the literature for LSCF electrodes2,3,7,35 or similar mixed conducting electrodes.5,14,31,39 This means that the resistance of the large semicircle is caused by the oxygen surface exchange reaction and the capacitance reflects the chemical capacitance of the film. The ionic and electronic across-plane transport resistances, as well as the in-plane electronic transport resistance, are negligible compared to the surface exchange resistance. Contributions from the LSF/YSZ interface can also be neglected, as the measured impedance spectra do not show any distinct interfacial (intermediate frequency) features. Moreover, from the literature and our own previous experiments with different current collector geometries, we can conclude that the electrode volume above the current collector grid is inactive towards oxygen exchange and does not contribute to the chemical capacitance due to the large in-plane ionic sheet resistance.7 Therefore, the chemical capacitance was normalized to the free thin film volume (73.5% of the total film volume for three-electrode samples, and 25% for microelectrode samples), i.e. it was calculated according to
![]() | (4) |
![]() | ||
Fig. 3 Equilibrium chemical capacitance at 600 °C (circles) as a function of oxygen partial pressure, measured on a 40 nm thin LSF film, and fit to a power law (solid line). |
Fig. 4 displays the chemical capacitances under DC polarization, where the electrode overpotential η was calculated from the voltage between the working and reference electrode UDC by correcting for ohmic losses in the electrolyte (Roffset·IDC), i.e.
η = UDC − Roffset·IDC. | (5) |
The Cchemvs. η curves measured in different atmospheres are very similar except for a shift on the overpotential axis. This voltage shift Δη between curves for different partial pressures (p1 and p2) equals the Nernst voltage calculated from these partial pressures according to
![]() | (6) |
To further quantify and compare the impacts of oxygen partial pressure and electrode polarization, we have to consider the distribution of the oxygen chemical potential in a three-electrode system. We define an oxygen chemical potential of the surrounding atmosphere, which is assumed to be constant in the entire apparatus, i.e. at all electrodes. Pure oxygen at 1 bar is used as a reference for the oxygen chemical potential, i.e. μ1barO = 0. Without applied DC bias, all electrodes (working electrode WE, counter electrode CE and reference electrode CE) are in equilibrium with the atmosphere, thus their oxygen chemical potentials (μWEO, μCEO and μREO) are equal to μatO. The oxygen chemical potential in any electrode relative to 1 bar oxygen is thus given by
![]() | (7) |
When applying a DC voltage between the working and counter electrode, the resulting chemical potential distribution strongly depends on the spatial position of the rate limiting step. In our case (dense thin film electrodes with current collector grids), we assume the following conditions, in agreement with the literature:2,3,7,35
• The electron transport in the electrode film is fast. Thus, electronic transport resistances can be neglected and the electrochemical potential of electrons is constant in the entire electrode film. This is reasonable as LSF has a high electronic conductivity, the across-plane transport length is short (40 nm) and the lateral electron transport is augmented by the current collector.
• The across-plane oxygen transport is fast compared to the surface exchange reaction. Therefore, the entire electrode reaction is surface limited, and the electrochemical potential of oxide ions is constant in the electrode film.
• The YSZ|LSF interface does not contribute any significant transport resistance, therefore no electrochemical potential discontinuity occurs at the interface.
Based on these assumptions, the oxygen chemical potential and the electrochemical potentials of oxide ions and electrons are distributed as depicted in Fig. 5. A current flow IDC is imposed between the working and counter electrode, causing a voltage UDC, i.e. an electron electrochemical potential difference between the working and reference electrode. No current flows through the reference electrode, thus it remains in equilibrium with the gas phase, i.e. at μREO = μatO. The DC voltage drop
![]() ![]() | (8) |
![]() | (9) |
![]() | (10) |
μWEO − μREO = 2e·(UDC − IDC·Rion) = 2e·ηWE, | (11) |
![]() | (12) |
Another point of view on the oxygen chemical potential can be introduced by defining an equivalent oxygen pressure of the working electrode .5 This is related to the oxygen chemical potential in the working electrode by
![]() | (13) |
![]() | (14) |
We may now plot the chemical capacitance versus the electrode oxygen chemical potential (relative to 1 bar oxygen), i.e. versus the equivalent oxygen pressure. This leads to a perfect match of all curves measured in different atmospheres, as shown in Fig. 6. We can thus conclude that the measured capacitance of the LSF films is solely defined by the chemical potential of oxygen within the thin film, regardless of the actual atmospheric oxygen partial pressure or electrode overpotential. This also strongly supports the assumptions made above, particularly the absence of a chemical potential gradient across the LSF film.
The interpretation of the CPE in the impedance analysis in terms of a chemical capacitance is thus reasonable and reflects the defect chemical state of LSF. In the regimes between 2.5 × 10−4 bar and 2.0 × 10−1 bar oxygen (Fig. 3) and from 10−7 to 10−12 bar equivalent oxygen pressure (Fig. 6), the Cchem curves can be fitted to a power law, with exponents of −0.21 (Fig. 3) and 0.20 (Fig. 6), respectively.
Fig. 8 displays the measured chemical capacitance, normalized to the active electrode area, as a function of film thickness at different oxygen chemical potentials. The measured capacitances show a linear dependence on the film thickness, as expected for a bulk chemical capacitance. However, fitting these data to a linear function and extrapolating to zero film thickness yields an intercept on the capacitance axis of 0.3 to 0.45 mF cm−2. Thus, the measured capacitance can be separated into a volume-specific chemical capacitance and a volume independent contribution, most likely due to an interface, either LSF|YSZ, LSF|Pt or LSF|air. Grain boundaries as the origin can be excluded because their contribution should again scale with thickness. Similar volume independent contributions to the measured chemical capacitance have also been found on ceria.31 At high oxygen chemical potentials, the volume related chemical capacitance is large and masks the volume independent interfacial capacitance. Close to the capacitance minimum, however, the interface-related capacitance contributes 60% (for 28 nm LSF) to 25% (for 116 nm LSF) to the entire electrode capacitance.
The interfacial capacitance depends only weakly on the oxygen chemical potential, see the inset in Fig. 8, and the measured dependence might easily be an artifact due to the extrapolation for only a few thickness values. To investigate the origin of this capacitance, samples with platinum covered working electrodes were measured. Covering the LSF surface with platinum increases the LSF|Pt interfacial area and eliminates the LSF|air interface. Fig. 9 shows the capacitance measured on the samples with and without platinum on top of LSF as a function of oxygen chemical potential for different film thicknesses. There is almost no difference between the samples with and without the Pt cover, which strongly suggests that the interfacial capacitance is located at the LSF|YSZ interface rather than at the LSF|Pt interface. This conclusion is further supported by a comparison of capacitance measurements on samples with different fractions of the YSZ surface covered by a current collector grid (25% for the three-electrode samples and 75% for the microelectrodes), see Fig. 10. When normalized to the active LSF volume on YSZ, see eqn (4), very good agreement is found. However, which atomistic mechanism at the LSF|YSZ interface causes the corresponding almost pO2 independent interfacial capacitance of about 400 μF cm−2 remains an open question.
![]() | ||
Fig. 11 Brouwer diagram for bulk LSF at 600 °C, based on the literature data,22 with chemical capacitance calculated according to eqn (22). |
For mixed conducting oxides, the chemical capacitance is defined in ref. 30 as
![]() | (15) |
![]() | (16) |
μO = μO2− + 2μh = −μV + 2μh, | (17) |
![]() | (18) |
![]() | (19) |
μi = μ0i + kT![]() ![]() | (20) |
![]() | (21) |
![]() | (22) |
Based on these defect chemical considerations and the thermodynamic data of ref. 22, the volume-specific chemical capacitance of bulk LSF can be calculated and the corresponding curve is given in Fig. 11. At high oxygen partial pressures, the calculated chemical capacitance is determined by the oxygen vacancy concentrations and thus shows a slope of −0.5. Towards lower oxygen partial pressures, it passes a maximum with equally important oxygen vacancies and electron holes. Then, it decreases with decreasing hole concentration (slope 0.25) and reaches a minimum for the electronically intrinsic point (ch = ce). Finally, Cchem again increases towards even more reducing conditions, in accordance with the increasing electron concentration.
This dilute defect model qualitatively explains the shape of the measured chemical capacitance curve, and in the low μO regime, the exponent of the chemical capacitance versus equivalent oxygen partial pressure curve (0.2) is indeed close to the expected value of 0.25. However, two serious deviations become obvious when quantitatively comparing the calculated and measured chemical capacitances, see Fig. 10. At high oxygen partial pressures (above 10 mbar), the measured chemical capacitance decreases in a much shallower manner (−0.21) than predicted by the dilute model (−0.5), and the capacitance value at the minimum is much larger than expected.
The deviation from the ideal behavior for high oxygen partial pressures may be due to hole/hole interactions, since the expected electron hole concentration in this regime is very high (0.4 per unit cell). An excess energy of hole formation due to this interaction might make further oxygen incorporation less favorable, leading to a smaller decrease in the oxygen vacancy concentration, and thus a smaller decrease in chemical capacitance when increasing the oxygen partial pressure. Similar hole/hole interactions were also found for other perovskite materials, such as Ba1−xLaxFeO3−δ and La1−xSrxCoO3−δ.42,43 One reason for the increased capacitance around the expected minimum was already identified above and attributed to the LSF|YSZ interface. This interfacial contribution can be subtracted and in the following, the remaining volume-specific chemical capacitance is used to derive thermodynamic data for oxygen incorporation and electron/hole pair formation.
![]() | ||
Fig. 12 Corrected chemical capacitance of LSF (116 nm thin film) at different temperatures as a function of equivalent oxygen partial pressure. The interface capacitance obtained from film thickness variation at 600 °C (see the inset in Fig. 8) has been subtracted. (Outside the μO range of Fig. 8, the averaged value of Cint was subtracted.) Symbols are measured data, and lines are fits to the bulk defect model according to eqn (22). |
These data were fitted based on the dilute bulk defect model and eqn (22) to extract equilibrium constants for the oxygen incorporation reaction
![]() | (23) |
![]() | (24) |
![]() | (25) |
![]() | ||
Fig. 13 Equilibrium constants for the oxygen incorporation reaction (Kox) and for the electron/hole pair formation (Ki) obtained from Fig. 12versus inverse temperature and fits according to eqn (25). |
This study | Literature (bulk) | |
---|---|---|
ΔHox (kJ mol−1) | −94 | −95.62 ± 4.18 |
ΔSox (J mol−1 K−1) | −50 | −54.27 ± 4.43 |
ΔHi (kJ mol−1) | 26 | 95.75 ± 2.05 |
ΔSi (J mol−1 K−1) | −81 | −21.63 ± 2.13 |
The resulting oxygen incorporation entropy and enthalpy agree very well with the thermodynamic data of bulk LSF.22 However, the deduced electron/hole pair formation enthalpy and entropy are both lower than those for bulk LSF.22,24,25 Possible reasons include the effects of film strain and its changes with changing oxygen chemical potential, and some differences in the exact stoichiometry (cation vacancies) or different pressure ranges used to deduce data, see below. Additionally, grain boundaries in our nanocrystalline thin films may cause deviations from the literature bulk data. Since the oxidation state in the grain boundaries is probably different than in the grains, a shifted Cchemvs. pO2 curve results, which adds to the grain chemical capacitance. A more detailed analysis, however, would require further data points in the very low pO2 region beyond the capacitance minimum.
These experiments demonstrate that chemical capacitance measurements are a powerful alternative method for analyzing the defect concentrations and defect thermodynamics of oxide materials, particular for thin films. Such an approach to the defect chemistry by Cchem measurements even has some advantages compared to common gravimetric studies. Since the chemical capacitance probes primarily minority charge carriers, it might be more sensitive to defect interactions causing vacancy changes in the hole conducting regime (see above). Moreover, compared to gravimetric studies, a less broad pO2 regime is already sufficient to obtain thermodynamic data. This is detailed in Fig. 14 for different equilibrium constants. In gravimetric studies, both steps indicating weight loss for increasing vacancies have to be observed. In Cchem studies, however, only the regime between the capacitance maximum and minimum is required.
![]() | ||
Fig. 14 Calculated chemical capacitance for LSF (solid lines) for different equilibrium constants and the corresponding oxygen stoichiometry (dashed lines), both based on the dilute defect model, see eqn (22). A much narrower pO2 span (only from capacitance maximum to capacitance minimum) is sufficient to obtain both equilibrium constants by chemical capacitance measurements, compared to thermogravimetric measurements (both steps in the oxide concentration are required). Arrows indicate typical ranges required for a fit analysis of Cchem or weight analysis. |
• The chemical capacitance solely depends on the oxygen chemical potential in the electrode film, independent of the source affecting this potential (atmospheric oxygen pressure or external DC voltage).
• The shape of the chemical capacitance versus oxygen chemical potential curve is in qualitative agreement with expectations from bulk defect chemical models and already indicates the relevant minority defects determining the chemical capacitance.
• For very small chemical capacitances, i.e. for moderately reducing conditions, an additional interfacial contribution to the capacitance in the range of 400 μF cm−2 comes into play and can be attributed to the LSF|YSZ interface.
• The corrected volume-specific chemical capacitance of the thin films can be reasonably well approximated by a defect chemical model with dilute defects, thus mass action constants for the oxygen exchange reaction and the electron/hole formation can be determined.
• The temperature dependence of the determined defect chemical data revealed the enthalpy and entropy of the oxygen incorporation reaction and of the electron–hole formation.
• Oxygen incorporation enthalpies and entropies of the LSF films agree very well with the literature data obtained on macroscopic samples. Electronic defect formation enthalpies and entropies, however, differ, leading to higher electronic defect concentrations in the thin films. Moreover, defect interactions between electron holes may play a role at high oxygen partial pressures, leading to deviations from the Cchem values predicted by the dilute defect model.
• The measurements showed that bias and oxygen partial pressure dependent chemical capacitance measurements can be a powerful tool for the analysis of the defect chemistry of thin films, and particularly to reveal details of the minority defects.
This journal is © the Owner Societies 2018 |