Seonyoung
Yoo
a,
Sihyuk
Choi
a,
Jeeyoung
Shin
b,
Meilin
Liu
c and
Guntae
Kim
*a
aInterdisciplinary School of Green Energy, and KIER-UNIST Advanced Center for Energy, Ulsan National Institute of Science and Technology (UNIST), Ulsan 689-798, Korea. E-mail: gtkim@unist.ac.kr; Fax: +82 52 217 2909.
bDepartment of Mechanical Engineering, Dong-Eui University, 995 Eomgwangno, Busan-jin-gu, Busan 614-714, Korea.
cSchool of Materials Science and Engineering, Georgia Institute of Technology, 771 Ferst Drive, N.W., Atlanta, GA 30332-0245, USA.
First published on 10th April 2012
While La1−xSrxMnO3 (LSM) has been widely used as a cathode material for SOFCs based on YSZ electrolytes at high temperatures,3,4 it is inadequate for use in an intermediate temperature range due to reduced ionic and electronic conductivity and diminished catalytic activity at lower temperatures.5–7
Recently, mixed ionic and electronic conductors (MIECs) have received tremendous attention as potential cathodes for IT-SOFCs. MIECs based on transition metal (e.g. Mn, Fe, Co, and Ni) oxides have been extensively investigated. Among various MIECs, cobalt containing oxides showed superior electrocatalytic activity than oxides with predominant electronic conductivity (and little ionic conductivity) such as lanthanum manganese. In particular, La1−xSrxCo1−yFeyO3−δ (LSCF)-based cathodes have attracted much attention for IT-SOFCs. However, the long-term stability of LSCF-based cathodes is still a concern.6–8
As a mixed conductor derived from the K2NiF4-type materials, La2NiO4 has attracted significant attention for possible application as IT-SOFC cathodes. Its advantages include high oxygen ionic and electronic conductivity, moderate thermal expansion coefficient (TEC), and high electrocatalytic activity toward oxygen reduction under oxidizing conditions.9–11 Other materials such as lanthanum cobaltite perovskite also exhibit good conductivities; however, the large thermal expansion mismatch with other cell components may lead to thermo-mechanical problems.
Ruddlesden–Popper compounds are comprised of alternating perovskite and rock-salt layers, as shown in Fig. 1. The number of perovskite layers increases with n in this structure, leading to the formation of higher order Ruddlesden–Popper phases, La3Ni2O7 and La4Ni3O10, which is argued to allow faster ionic and electronic transport.6,11,12 These effects are primarily attributed to increased concentration of Ni–O–Ni bonds, which are responsible for electronic conduction due to progressive delocalization of the p-type electronic charge carriers, and enhanced vacancy-migration or oxygen ion diffusivity.4,11,12 Currently, many of the studies on Lan+1NinO3n+1+δ (n = 1, 2, or 3) focused on the structure and electrochemical properties of the bulk phases, with little attention to the basic thermodynamic properties. To date, the characteristics of Lan+1NinO3n+1+δ infiltrated into a scaffold of YSZ and the actual configuration of a porous cathode fabricated by infiltration, are still unknown. Because of the unique microstructures and possible interactions between the two phases, the behaviour of this Lan+1NinO3n+1+δ–YSZ could be very different from that of a pure Lan+1NinO3n+1+δ phase. Further, redox properties related to oxygen thermodynamics such as oxidation enthalpies and entropies of Lan+1NinO3n+1+δ (n = 1, 2, and 3), and redox stability have not been reported for the intermediate temperature range.
Fig. 1 The structure of Lan+1NinO3n+1 (n = 1, 2, and 3). |
In this study, we characterized non-stoichiometric variations of oxygen and electrical conductivities of Lan+1NinO3n+1+δ (n = 1, 2, and 3) infiltrated into porous YSZ as a function of oxygen partial pressure in a temperature range of 923–1023 K. Redox behavior was evaluated using coulometric titration and the electrical conductivity was determined using 4-probe conductivity measurement.
The Lan+1NinO3n+1 (n = 1, 2, and 3) were characterized using X-ray diffraction (XRD) and scanning electron microscopy (SEM). X-Ray powder diffraction measurements (Rigaku diffractometer, Cu Ka radiation) were performed to confirm the structure with a scan rate of 0.5° min−1 and a range of 20° < 2θ < 60°.
The oxidation/reduction state can be characterized precisely by coulometric titration as a function of p(O2).13,14 This involves placing the oxide sample in a sealed container, separated from the atmosphere by an O2−-conducting membrane such as yttria-stablized zirconia (YSZ). The YSZ tube (McDanel Advanced Ceramic Technologies, Z15410630) was used both to pump oxygen out of the system and to sense the equilibrium p(O2) inside the tube. Lan+1NinO3n+1 (n = 1, 2, and 3) were located in a sealed container at various temperatures of interest and equilibrated by subjecting it to a flow of 5% O2 in Ar. The sample was then isolated in the tube and the equilibrium p(O2) was measured with an oxygen sensor. Electrodes on both sides of the YSZ tube were used to measure the potential across the membrane, and the potential could be related to the p(O2) through the Nernst equation. A specific amount of charge could be passed across the tube using a BioLogic Potentiostat, with 1 coulomb of charge being equivalent to 2.6 μmol of O2.
The electrical conductivity of the slabs, 2 mm × 5 mm × 10 mm in size, was measured in air by the four-probe method with a BioLogic Potentiostat. The measurements were performed starting from 1023 to 373 K with 323 K measurement intervals.
Fig. 2 XRD patterns of Lan+1NinO3n+1 (n = 1, 2, and 3)–YSZ sintered at 1123 K: each symbol indicates Lan+1NinO3n+1 (n = 1, 2, and 3), La2NiO4+δ (•), La3Ni2O7−δ (▼), La4Ni3O10−δ (♦), and YSZ (✓). |
As is well known, the microstructure of an electrode may influence the reaction kinetics, charge and mass transport processes, and hence fuel cell performance. For example, three phase boundaries (TPBs) at the electrode/electrolyte interfaces are the most active sites for electrochemical reactions in SOFCs. Smaller grain size often results in longer TPB length and potentially higher electrochemical performance.13 SEM images of the Lan+1NinO3n+1–YSZ fired at 1123 K are presented in Fig. 3a, b, and c, respectively, for n = 1, 2, and 3. It is clearly seen that the grain size of the Lan+1NinO3n+1 becomes smaller with the increasing value of n.
Fig. 3 SEM images of the (a) La2NiO4+δ–YSZ, the (b) La3Ni2O7−δ–YSZ, and the (c) La4Ni3O10−δ–YSZ annealed at 1123 K. |
Fig. 4 The isotherms of the Lan+1NinO3n+1 (n = 1, 2, and 3)–YSZ at 923–1023 K, (a) La2NiO4+δ, (b) La3Ni2O7−δ and (c) La4Ni3O10−δ. The solid curves are fitting curves calculated by the proposed defect model using data only before decomposition. |
There is a considerable change in oxygen non-stoichiometry throughout the p(O2) range for all compounds. The data show that the isotherms of Lan+1NinO3n+1–YSZ (n = 1, 2, and 3) have similar shapes, suggesting that the reduction mechanisms of the scaffolds are quite similar. As temperature decreases, the decomposition p(O2) becomes lower. With an increase of n, the isotherms are extended to the left, indicating that La4Ni3O10−δ has higher redox stability than La2NiO4+δ down to a lower p(O2) at the same temperature. This is due possibly to the higher number of perovskite layers in La4Ni3O10−δ and, accordingly, stronger interaction between the molecules in the lattice.
The partial molar enthalpy and entropy of oxygen can be calculated from the slopes of the isotherms. The Gibbs free energy, ΔG, is related to the equilibrium constant, K, and p(O2) as follows,
(1) |
At a constant δ, the partial molar enthalpy of oxygen at various temperatures is shown by the Gibbs–Helmholtz equation.
(2) |
And the partial molar entropy can be obtained by using the Maxwell relation as follows.
(3) |
The partial enthalpies of oxidation for the Lan+1NinO3n+1–YSZ (n = 1, 2, and 3), calculated from eqn (2), are presented in Fig. 5. The oxidation enthalpies (−ΔH) are a strong function of oxygen non-stoichiometry in Lan+1NinO3n+1 (n = 1, 2, and 3).
Fig. 5 Partial molar enthalpy of oxidation (−ΔH) at 973 K of the Lan+1NinO3n+1 (n = 1, 2, and 3)–YSZ, (a) La2NiO4+δ, (b) La3Ni2O7−δ and (c) La4Ni3O10−δ. |
The partial molar enthalpies of oxidation at 10−5 atm are plotted in Fig. 6. The higher partial molar enthalpy for La4Ni3O10−δ–YSZ relative to that of Lan+1NinO3n+1 (n = 1 and 2) means that it would be more stable at roughly the same p(O2).
Fig. 6 Partial molar enthalpy of oxidation (−ΔH) at 973 K of the Lan+1NinO3n+1 (n = 1, 2, and 3)–YSZ at p(O2) = 10−5 atm. |
The partial molar entropies of oxidation, calculated from the differences in the Gibbs free energies and the enthalpies, are presented in Fig. 7. The values of −ΔS become smaller for higher δ, implying that the probability for interstitial oxygen formation decreases with excess oxygen. In other words, there are fewer sites for the interstitial oxygen formation reaction in the scaffolds as the amount of excess oxygen increases (Fig. 7a).9,15–17
Fig. 7 Partial molar entropy of oxidation (−ΔS) at 973 K of the Lan+1NinO3n+1 (n = 1, 2, and 3)–YSZ, (a) La2NiO4+δ, (b) La3Ni2O7−δ and (c) La4Ni3O10−δ. |
It can also be argued that fewer sites will be available for oxygen vacancy formation in the scaffolds as the concentration of oxygen vacancies increases. This is consistent with the finding that the values of −ΔS become smaller for higher δ (Fig. 7b and c).9,18
(4) |
Thus, electroneutrality approximation requires that the number of electron holes should be about twice the number of oxygen interstitials.
[h•] = 2[Oi′′] = 2δ | (5) |
In terms of δ and p(O2), the equilibrium constant for eqn (4) can be expressed as
(6) |
where γ1 and γ2, denote the activity coefficients of Oi′′and h•, respectively.
Thus, the Gibbs free energy change for eqn (4) can be expressed as follows.
(7) |
The second term in eqn (7) represents the deviation from the Gibbs free energy change for an ideal system (activity coefficient = 1). As a first order approximation, the deviation is assumed to be a linear function of the oxygen non-stoichiometry, δ.16,19
ΔGex = −RTlnγ1γ22 aδ | (8) |
Thus, the constant a reflects the degree of interaction between the defects and the lattice ions. A positive a suggests that the formation of interstitial oxygen is easier than in an ideal system while a negative a implies that it becomes more difficult than in an ideal system.16 As a approaches zero, the system reduces to the ideal system (i.e. there is no interactions between the defects and the lattice ions). In light of eqn (8), eqn (7) can be rewritten as
(9) |
The relationship between δ and p(O2) as described by eqn (9) can then be used to estimate the parameters K and a from experimental data.
The formation of oxygen interstitials in La2NiO4+δ–YSZ is electrically compensated by the formation of electron holes to satisfy electroneutrality. The oxygen non-stoichiometry data for all La2NiO4+δ–YSZ were fitted to the models described above. The theoretical curves calculated for La2NiO4+δ–YSZ using eqn (9) are presented as solid lines in Fig. 4(a), with the fitted K and a values listed in Table 1. Although there was some small deviation, the observed non-stoichiometric behaviour of La2NiO4+δ–YSZ can be well explained using the proposed defect model over a wide range of p(O2) except the area where decomposition is expected. The equilibrium constant K increases exponentially with temperature, implying that the formation of oxygen interstitials is highly affected by temperature, as also shown in the isotherms. This is further confirmed by the observation that the slope of δ versus p(O2) at higher temperatures is steeper than that at lower temperatures, reflecting the relative ease of interstitial oxygen formation at high temperatures.
La2NiO4+δ | log K | a (J mol−2) |
923 K | 0.336 | −2.42 E+05 |
973 K | −1.130 | −9.88 E+04 |
1023 K | −1.792 | −7.68 E+04 |
Since the value of −a for La2NiO4+δ–YSZ decreases with increasing temperature, the degree of interactions between the defects and the lattice ions diminishes at higher temperatures.
(10a) |
NiNiX ↔ NiNi′ + h• | (10b) |
The combination of eqns (10a) and (10b) are the dominant defect reaction in these materials. Thus, charge neutrality requires that the effective negative charge on the lattice ions be balanced by the positive charges of electron holes and oxygen vacancies, as described below.
[NiNi′] = [h•] + 2[Vo••] or [h•] = [NiNi′] − 2δ | (11) |
Here, the concentration of the charged lattice ions changes with non-stoichiometry as well. Since the electronic conductivity (σh) is much greater than the ionic conductivity (σv) for these materials, however, the total conductivity (σ) is dominated primarily by the transport of electron holes, i.e.
(12) |
where q is the charge of electron and μh is the drift mobility of electron holes.
In light of this approximation, the equilibrium constant for eqn (10a) can be approximated by
(13) |
where γ3 and γ4 denote the activity coefficients of Oox and Vo••, respectively.
The Gibbs free energy change, ΔG, of eqn (10a) can be expressed in terms of the equilibrium constant, K, as in eqn (14a) and (14b) for La3Ni2O7−δ and La4Ni3O10−δ, respectively.
(14a) |
(14b) |
We define the second term in eqn (14a) and (14b) as the deviation from the standard free energy change of the ideal system, similar to the approximation made for the La2NiO4+δ case.20
(15) |
(16a) |
(16b) |
which can be rewritten as,
(17a) |
(17b) |
The oxygen non-stoichiometry data (δ values) and the electrical conductivity data (σ values) at different partial pressures of oxygen collected for the two cathodes (La3Ni2O7−δ–YSZ and La4Ni3O10−δ–YSZ) were curve fitted to the models again. The theoretical curves calculated for using eqn (17a) and (17b) are presented as solid lines in Fig. 4(b) and 4(c), respectively, with the fitted K* and a values listed in Table 2. The theoretical curves again show quite good agreement with the experimental data before decomposition. Oxygen vacancy formation becomes more difficult compared to the ideal system when a is negative.20,21
La3Ni2O7−δ | log K* | a (J mol−2) | La4Ni3O10−δ | log K* | a (J mol−2) |
923 K | 1.308 | −8.53 E+05 | 923 K | 0.049 | −4.13 E+05 |
973 K | 1.662 | −4.27 E+05 | 973 K | 0.412 | −3.15 E+05 |
1023 K | 1.959 | −2.17 E+05 | 1023 K | 1.112 | −1.89 E+05 |
For La2NiO4+δ, the electrical conductivities increased with p(O2) due mainly to increased concentration of mobile interstitial oxygen in the perovskite layers. Since the predominant defects in La2NiO4+δ are interstitial oxygen and electronic holes, an increase in the concentration of oxygen interstitial results in an increase in the concentration of electronic holes, which leads to increased electronic conductivity of the material.14,15
Fig. 8 The electrical conductivity of Lan+1NinO3n+1 (n = 1, 2, and 3)–YSZ at various temperatures from 1023 K to 373 K in air, (●) La2NiO4+δ, (■) La3Ni2O7−δ, and (▲) La4Ni3O10−δ. |
Fig. 9 The electrical conductivities of Lan+1NinO3n+1 (n = 1, 2, and 3)–YSZ, (a) La2NiO4+δ, (b) La3Ni2O7−δ, and (c) La4Ni3O10−δ at (●) 923 K, (■) 973 K, and (▲) 1023 K at various p(O2) (atm). |
For La3Ni2O7−δ and La4Ni3O10−δ, in contrast, an increase in oxygen partial pressure would lead to a decrease in concentration of oxygen vacancies and an increase in the concentration of electronic holes, which in turn increase the p-type electrical conductivity.
This journal is © The Royal Society of Chemistry 2012 |