Determination of Thermodynamic and Transport Properties of Non-Stoichiometric Oxides

The purpose of this chapter is to describe methods for determining the oxygen stoichiometry (bulk and surface oxygen activities) and transport properties in non-stoichiometric oxides using solid electrolyte cells. These oxides are generally referred to as “mixed ionic-electronic conductors” (MIEC). It is customary to consider that in a MIEC, the ionic or electronic transport number is higher than 0.01 and that the total electrical conductivity is sufficiently high, i.e., higher than 10-5 S.cm-1. Emphasis will be given on sources of error. Devices for overcoming these sources of error will be described. As the aim of this chapter is limited and cannot be an exhaustive review, the readers are referred to excellent general handbooks dealing with solid state electrochemistry, with chapters devoted to electrochemistry of non-stoichiometric oxides [Gellings & Bouwmeester, 1997; Kharton, 2011; Rickert, 1982; Sorensen, 1981] or review papers [Heyne, 1982; Weppner & Huggins, 1978].


Introduction
The purpose of this chapter is to describe methods for determining the oxygen stoichiometry (bulk and surface oxygen activities) and transport properties in non-stoichiometric oxides using solid electrolyte cells.These oxides are generally referred to as "mixed ionic-electronic conductors" (MIEC).It is customary to consider that in a MIEC, the ionic or electronic transport number is higher than 0.01 and that the total electrical conductivity is sufficiently high, i.e., higher than 10 -5 S.cm -1 .Emphasis will be given on sources of error.Devices for overcoming these sources of error will be described.As the aim of this chapter is limited and cannot be an exhaustive review, the readers are referred to excellent general handbooks dealing with solid state electrochemistry, with chapters devoted to electrochemistry of non-stoichiometric oxides [Gellings & Bouwmeester, 1997;Kharton, 2011;Rickert, 1982;Sorensen, 1981] or review papers [Heyne, 1982;Weppner & Huggins, 1978].

Control and measurement of oxygen activity
Precise control and monitoring of the oxygen pressure in the experimental chamber is required for the determination of thermodynamic and transport properties in MIECs.Electrochemical devices have been developed since more than thirty years, allowing the control of the oxygen pressure in the 1 -10 -27 bar range in various gas mixtures or under partial vacuum.Solid electrolyte microprobes have also been proposed for the local determination of the oxygen activity on the surface of a non-stoichiometric oxide.

Measurement of oxygen pressure in a gas phase and under partial vacuum
The conventional oxygen sensor, shown Figure 1 a, is based on a closed yttria-stabilized zirconia tube [Kleitz et al., 1992].The electrodes are made of platinum paste.The outer electrode, in contact with air, forms the reference electrode.The cell emf obeys the Nernst law: A miniaturized sensor (Figure 1 b) with enclosed metal-metal oxide reference system (Pd-PdO, Co-CoO) has been developed and commercialized1 [Fouletier & Vitter, 1980].The main advantage of this device concerns the possibility of in situ measurements without circulation of the analyzed gas.
It has been shown that under partial vacuum or for high temperature operation, the sensor could become inaccurate due to the disturbing effect of the oxygen semipermeability flux through the zirconia tube, reaching the measuring electrode.To overcome the problem, it has been proposed to use the "zirconia point electrode" shown in Figure 1 c.The platinum measuring electrode is replaced by a conical piece of zirconia (see also Figure 4 a) pressed in contact with the inner surface of the zirconia tube, which is not platinized.The actual measuring electrode is a platinum coating deposited on the top of the zirconia probe.In this design, due to the shape of the probe, the oxygen semipermeability flux does not reach the measuring electrode since the current follows the path of lowest resistance.Consequently, the measuring electrode remains in equilibrium with the gas phase.As developed in Section 2.2.1, this zirconia point electrode can be used for the measurement of oxygen activity on the surface of a non-stoichiometric oxide.

Pump-sensor device
This device is widely used for the control and monitoring of oxygen content in gas mixtures flowing in the experimental chamber.The gas circuit is schematized in Figure 2 a.When nominally pure gases (Ar, N 2 , He, etc.) are used, according to the Faraday law, the oxygen mole fraction X in the flowing gas obeys the following equation: where X° is the oxygen mole fraction in the gas supplied to the pump, D is the gas flow rate in L.h -1 NTP and I is the current intensity in A. As shown in Figure 2 b, the theoretical equation ( 2) is verified in the 10 -7 -1 mole fraction range: the oxygen mole fraction is determined by the oxygen sensor (using Nernst equation ( 1)) for various current intensities I passing through the pump [Fouletier et al., 1975].3) and (4) [Caneiro et al., 1981;Fouletier et al., 1984].
The same device can be used for monitoring the composition of CO 2 -CO and Ar-H 2 O-H 2 mixtures.In that case, pure (H 2 , CO 2 ) or premixed gases, such as Ar -5 % H 2 are used.
According to the Faraday's law, the equilibrium oxygen pressure vs. the current intensity passing through the pump obeys the following equations: where q is the hydrogen mole fraction in the feed gas, and T is the absolute temperature in the experimental chamber.
The association of an electrochemical oxygen pump and an oxygen sensor allows the monitoring of oxygen partial pressure in a flowing gas in the range 1 bar -10 -27 bar with an accuracy of 2 %.

Oxygen pressure domains of ideal response
It should be pointed out that the accuracy of the equilibrium oxygen pressure control depends on the buffer capacity δ of the gas.The buffer capacity can be defined as the number of moles of oxygen required for changing the chemical potential of 1 kJ/mole of gas.The variation of δ with oxygen pressure, at various temperatures, for Ar-O 2 and CO 2 /CO mixtures is given in Fig. 3.It has been shown previously that the buffer capacity of the gas has to be higher than 10 -6 mole, at 800°C.Consequently, at this temperature, the pressure domains in which the oxygen pressure is accurately controlled are: 1 -10 -6 bar in Ar-O 2 mixtures and 10 -10 -10 -27 bar in CO-CO 2 mixtures.Obviously, these pressure domains depend on temperature.The same type of curve is obtained with H 2 /H 2 O mixtures.
Fig. 3. Variation of the buffer capacity of inert gas-O 2 and CO-CO 2 mixtures, with indication of the oxygen pressure domains of correct utilization of the oxygen sensor at 800°C [Fouletier, 1982/83].

Control and measurement on an oxide surface
Basically, the oxide ion activity in a MIEC can be determined using the electrochemical cell: Pt -Reference system / oxide electrolyte / MIEC -Pt.The measurement of the emf of the electrochemical chain allows the determination of the oxygen activity in the MIEC.The following sections are devoted to the description of devices allowing the determination of the oxide activity on the surface of a MIEC, and of the deviation from equilibrium of the oxide surface.Gold is used as metallic point electrode due to its poor catalytic activity for oxygen electrode reaction; consequently, gold can be considered as an electronic probe on the MIEC and: ).The MIEC surface is considered as an equipotential.The emf E of the chain can be easily obtained from the following assumptions: -Electronic equilibrium at interfaces α and β: Obviously, if the surface of the sample is in equilibrium with the gas phase, the emf is nil.
Provided the oxygen pressure in the gas phase near the surface is known (as an example using a microsensor described in Section 2.1), the oxygen activity on the surface can be determined (see Section 4.1).Similar designs using calcia-stabilized zirconia cells have been tested for continuous monitoring of oxygen activity on the surface of growing scale during high temperature oxidation of metals [Akida et al., 2008].

Oxygen minisensor
Mini-probes (Figure 6) have been developed for the determination of thermodynamic properties of MIEC as a function of composition.Zirconia-or thoria-based tubes, a few mm in diameter, have been used.A metal-metal oxide system serves as a reference.Such cells have been used for the measurement of oxygen potential in urania-based solid solutions or for continuous control of oxygen redistribution in UO 2+x under a thermal gradient [Ducroux et al., 1980;Une & Oguma, 1982].

Determination of nonstoichiometry in oxides
Various methods based on the measurement of different physico-chemical characteristics have been developed.Most of the methods measure changes of the non-stoichiometry ratio x; however, with the help of a defect model, the absolute value of x can often be obtained [Riess & Tannhauser, 1982].Thermogravimetry under controlled oxygen pressure is the most widely used method [Caneiro et al., 1982[Caneiro et al., , 2011;;Kharton et al.].Other methods are based on the coulometric titrations.The oxygen activity in MIECs can also be determined using potentiometric cells involving a solid electrolyte.In the following sections, only the techniques involving solid electrolytes designs will be described.

Thermogravimetric investigation under controlled oxygen partial pressure
Thermal stability and oxygen content versus temperature and oxygen partial pressure are very important parameters in the case of non-stoichiometric compounds.Oxygen nonstoichiometry (δ) is generally measured using electronic microbalances under controlled atmospheres [Kharton et al., 2008;Tsipis et al., 2008].About 1 g of powder sample is placed in a silica basket suspended by platinum wires from the beam of the microbalance.The oxygen partial pressure is controlled by the introduction of gas mixtures (Ar-O 2 , Ar-H 2 , dry or wet) into the sample chamber [Nakamura et al., 2009a], [Kiselev et al., 2011;Nakamura et al., 2009b].First of all, the equilibrium between the sample and the surrounding gas phase is checked by controlling that both the weight of the sample and the oxygen partial pressure (recorded by a zirconia sensor introduced into the sample chamber) reached constant value.
Figure 7a shows a typical weight change curve of Nd 2-x NiO 4+δ in Ar-H 2 mixed atmosphere to determine the starting oxygen content of the sample [Zhao et al., 2008].The change in oxygen content is determined from variation of the weight of the sample Δw S according to the following relation: where Δδ, M S , M O , w S are the variation of oxygen nonstoichiometry, the molar formula weight of the sample and oxygen atom and the weight of the specimen, respectively.The experimental error due to the buoyancy is negligibly small compared to the weight variation of the sample due to release or incorporation of oxygen.
In the case of figure 7a, two weight changes are observed.The first one occurring at 350°C corresponds to the loss of interstitial oxygen and the reduction of Ni 3+ to Ni 2+ .The second weight loss is assigned to the total reduction of Nd 2 NiO 4+δ to Nd 2 O 3 and Ni metal.The absolute value of the oxygen content is determined from the weight change of the sample during the decomposition in H 2 atmosphere.The decomposition reaction can be expressed by: From the weight change, the δ value is deduced and the Ni 3+ concentration can be estimated.This mixed valency parameter can be linked to the electronic conductivity of the compound.
The change of molar Gibbs energy ΔG° for reaction (8) represents the partial molar Gibbs energy of oxygen atoms per 1 mol and can be expressed as follows [Caneiro et al., 2011;Patrakeev et al 1995]: where R, T and PO 2 are the gas constant, the temperature and the oxygen partial pressure, respectively.Taking into account the Gibbs-Helmholtz equation, and equation ( 9), the changes of standard partial molar enthalpy ΔΗ° and entropy ΔS° of oxygen release process per 1 mol of oxygen atoms can be deduced from equations ( 10) and ( 11) respectively: In order to determine the standard thermodynamic quantities, experimental dependences of the non-stoichiometry (δ) versus temperature and oxygen partial pressure are determined (see Fig. 7b) [Mauvy et al., 2009].If the plots into the coordinates R/2 Ln PO 2 vs. (1/T) δ and RT/2 Ln PO 2 vs. (T) δ give linear relationship within the values of δ studied, it allows to access to standard partial molar enthalpy ΔΗ° and entropy ΔS° (Fig. 8a).This linear behaviour indicates that the values of standard thermodynamic quantities are essentially independent of the temperature in the working temperature range.These values can be estimated from linear regression coefficients at given oxygen content.Figure 8b shows the calculated values of the standard enthalpy ΔΗ° for oxygen release process for Nd 1.95 NiO 4+δ compound.It can be noticed that a linear behaviour of both thermodynamic parameters ΔΗ° and ΔS° versus δ, can be related to a random distribution of non-interacting point defects as components of ideal solution.Generally, when the nonstoichiometry increases, noticeable deviation from linearity can be observed.In the case of Nd 1.95 NiO 4+δ oxide, ideal-solutionlike state means that the interaction among defect species is nearly constant regardless of the defect concentration.

Pump-sensor device
The pump-sensor device described in section 2.1.1 can be used for the determination of the stoichiometry ratio in oxides using the device shown Fig. 9 a [Meas et al., 1978].The experimental reactor R is connected to an oxygen sensor OS (using a zirconia probe as measuring electrode), an oxygen pump OP and a mechanical pumping system PS.A feedback controller FBC compares the sensor emf E given by the sensor to a set value E S and adjust the oxygen pressure in the experimental vessel by passing an appropriate current through the electrochemical pump.As a preliminary treatment, the system was initially outgassed under a pressure lower than 10 -8 bar and then equilibrated with a wellcontrolled low oxygen pressure, typically 10 -7 bar.Starting from this equilibrium state, a typical experiment is simply performed by changing the set voltage E S and recording the resulting pumping current passing through the oxygen pump and its integral which is proportional to the oxygen amount exchanged between the solid and the gas.Following an identical procedure, measurements were first carried out without sample in the reaction vessel.The quantities of oxygen involved in the gas-solid equilibrium are calculated by difference.This device has been used for oxygen adsorption studies on zeolite or on stabilized zirconia [Meas et al. 1978].As an example, in figure 9 b, is plotted the variation of the stoichiometry ratio of CeO 2-x as a function of oxygen partial pressure, at 1273 K [Fouletier et al., 1982].

www.intechopen.com
Stoichiometry and Materials Science -When Numbers Matter 184

Coulometric titration
The coulometric titration method can be defined as charge controlled mass transfer of mobile component between electrodes of electrochemical cell.This method is a very sensitive control of the composition of the nonstoichiometric phase, combined with emf measurements that yield very accurate thermodynamic information [Tretyakov et al., 1997].
As reported in Fig. 10, an yttria stabilized zirconia tube is used as electrolyte for a galvanic cell [Mizusaki et al., 1991;Nakamura et al., 2009a].Platinum paste is painted on the outside of the tube to stick the Pt mesh whereas Pt mesh is attached to the sample powder as the inside electrode.After evacuation and fulfilment with Ar gas procedures, the amount of oxygen, which is extracted or incorporated to the sample, is controlled by the electric charge passed through the cell.After specified amount of electric charge is applied, the electromotive force is measured to determine the equilibrium oxygen partial pressure in the tube.By considering that the amount of oxygen inside the tube is negligibly small, the oxygen amount, which migrated between the sample oxide and the gas phase, is small enough to be neglected.Then, Δδ can be calculated according to the relation: where I , t, F and M S are the current, the time, the Faraday constant and the molar weight of the sample, respectively.
Typical examples of coulometric titration curves are reported in Fig. 11 for La 2 NiO 4+δ compound at different temperatures [Nakamura et al, 2009b].
It can be noticed that data points near the plateau of δ versus log PO 2 region contain larger uncertainty than other data points because equilibrium potential varies easily by the small variation of oxygen content near the plateau region.According to Wagner theory, the slope of the δ versus log PO 2 curve for nonstoichiometric compounds shows minimum value at the stoichiometric composition [Wagner, 1971].
Investigations with oxygen concentration cells revealed that the major problem is a nonelectrochemical transport of oxygen through oxide ion electrolytes.Such leakage introduces uncontrolled and excessive changes in the oxide composition and appears to be the principal cause of instability. .

Gas release method
The stoichiometry ratio in a non-stoichiometric oxide can be monitored by passing a current through it [Fouletier & Kleitz, 1978;Fouletier et al, 1982].The experimental set-up is schematized, Fig. 12a.It consists of an inert gas cylinder, an oxygen pump, the experimental vessel and an oxygen sensor.The sample is a cylindrical pellet with platinum electrodes deposited on its bases.The sample is electrochemically reduced by passing a direct current.The amount of oxygen extracted from the sample, and the corresponding variation of the stoichiometry ratio, is determined by integration of the oxygen content deduced from the downstream oxygen sensor emf (Fig. 12b) [Levy et al., 1988].As an example, for the oxide Zr 1-y Y 2y O 2+y-x , (y is the dopant concentration), the stoichiometry ratio x is calculated from the equation: where E(t) represents the emf of the oxygen sensor during the reduction process, E(0) is the sensor emf before electrochemical reduction, D the gas flow rate, V m the volume of a mole of gas, M S the sample molecular weight, and m the sample weight.

EMF method
According to the pioneering works of Kiukkola & Wagner [Kiukkola & Wagner, 1957], the electrochemical chain is the following: Pt -Reference system / oxide electrolyte / MIEC -Pt.As described in Section 2.2.1, the emf of the cell allows the measurement of the oxygen activity in the MIEC, provided that there is no mass transfer through the oxide surface.The reference system is either a gas (pure oxygen, air) or a metal-metal oxide mixture (Fe-FeO, Cu-Cu 2 O, Pd-PdO, etc.).Examples of measurements on non-stoichiometric oxides are numerous [Mari et al., 1977;Nakamura & Fujino, 1987;Otobe et al., 2009;Porat & Riess, 1994].
The two main problems concern the reactivity between the MIEC and the solid electrolyte and the appearance of electronic conductivity in the solid electrolyte inducing polarization phenomena.The appropriate choice of the solid electrolyte can noticeably reduce the chemical reactivity: as an example, ceria-based electrolyte or apatite is less reactive than stabilized zirconia [Mauvy et al., 2009].Double-electrolyte cells have been proposed to extend the oxygen activity range of thermodynamic cell measurements compared to the range of a single-electrolyte cell arrangement [Shores & Rapp, 1971;Tretyakov & Muan, 1969].
We will focus on the use of cone-shaped MIEC.The experimental set-up is schematized, Figure 13a, and a photograph of a nickelate point electrode is shown, Figure 13b.Air is used as a reference electrode, and the cone-shaped MIEC is gently pressed in contact with the solid electrolyte (zirconia-based pellet, apatite, etc.).If available, a solid electrolyte closedend tube can be used.The main advantages concern the possibility of simultaneous measurements with various MIEC materials, the ease to handle, the reactivity between the MIEC and the solid electrolyte is minimized, a reduced response time to reach thermodynamic equilibrium.
As an example, in Figure 13c is plotted the variation of the oxygen activity (a O ) in the nickelate phase Ni 1.95 NiO 4+ as a function of the activity of oxygen in the gas ( 21/2 O P) .T h e equilibrium between the nickelate and the gas phase is observed in the whole investigated temperature range.

Transport properties investigations in MIEC oxides
Because of the charge equilibrium within the oxide, the electronic partial conductivity (electrons or holes) σ el and the ionic partial conductivity σ i are linked to the oxygen nonstoichiometry and, consequently, are two key parameters characterizing a MIEC.The methods developed for deconvoluting the ionic and electronic contributions to the total conductivity of a MIEC allow either the measurement of the partial conductivities, the diffusion coefficient of mobile defects or the transport numbers, either t el or t ion .Only the methods involving a solid electrolyte cell are briefly described.Reviews papers can be referred [Heyne, 1982;Rickert, 1982;Riess, 1997;Weppner & Huggins, 1978].

Oxygen permeation
Assuming that there is no oxygen gradient in the gas phase, the overall oxygen permeation rate through a non-stoichiometric oxide may be schematically decomposed into three elementary steps: ionic defect bulk diffusion (step 1), surface exchange between oxygen and oxygen vacancies (step 2 (high pressure side, )).As previously proposed [Bouwmeester et al., 1992] the membrane may be divided into three zones, as schematically shown in Fig. 14a. ) [Heyne, 1977;Wagner, 1957].
According to this theory, the steady oxygen permeation flux density 2 O j (in mol.cm -2 .s - ) controlled by bulk diffusion in a mixed conductor is given by: where all the symbols have their original meaning and L is the membrane thickness.
www.intechopen.comIn the case of mixed conductors in which the electronic conduction predominates, i.e., σ e >> σ i , the integral in the Wagner relation involves only σ i over the oxygen partial pressure gradient: According to Möbius [Möbius, 1986], 2 O j is the oxygen permeation flux density and 2 O J i s the specific oxygen permeability (in mol.cm -1 .s - ). - In case of simple defect model, neglecting the formation of defect association, the oxygen nonstoichiometry δ and the conductivity σ i are proportional to 2 1/n O P. Substitution into Eq.15 with subsequent integration leads to a simplified expression for oxygen permeation: is the value of the ionic conductivity at unit oxygen pressure.

www.intechopen.com
As indicated by Bouwmeester [Bouwmeester et al., 1994], for high values of n, Eq. ( 16) is obtained by expanding Eq. ( 17) as a power series and truncation after the first term.Kleitz et al. [Kleitz et al., 1973;Fouletier et al., 1975] were the first to demonstrate that the existence of a nonvanishing semipermeability flux through a solid electrolyte induces a deviation from equilibrium on both sides of the membrane.They expanded the Wagner theory to account for partial control of surface reactions on the transport kinetics through stabilized zirconia.This approach has been applied to mixed ionic-electronic oxides [Bouwmeester et al., 1994;Chen et al., 1997;Geffroy et al., 2011;Xu & Thomson, 1999].
According to equation 16, in case of negligible effect of the surface exchange rate, the oxygen permeation flux density 2 O j is proportional to [ln(P rich /P lean )], and the specific oxygen permeability 2 O J is independent of the membrane thickness [Kharton et al., 1999].
In case of limiting effect of the surface exchange rate, the specific oxygen permeability increases with the membrane thickness due to a decreasing role of the exchange kinetics (see Figure 15).[Kharton et al., 1999].
Another approach was to define the ratio h = k/D* where k is the surface exchange coefficient and D*, the tracer diffusion coefficient [Carter et al., 1992;Steele, 1992] or a critical thickness L d (which is the reciprocal of the parameter h), at which the oxygen flux is half of that expected for a diffusion controlled process [Bouwmeester et al., 1994)].The higher the L d characteristic thickness (or the smaller the h value), the more the surface exchange contributes to the oxygen permeation flux.Examples of L d values are given in [Gellings & Bouwmeester, 1997, p. 505]: L d varies from 0.2 µm for La 0.5 Sr 0.5 MnO 3-δ to 0.03 cm for La 0.6 Sr 0.4 Co 0.4 Ni 0.6 O 3-δ at 700°C.
The ionic conductivity can be deduced from the measured oxygen semipermeability flux using the following equation, in which the oxygen activities P* are taken into account instead of the oxygen pressures in the gas: www.intechopen.com

Stoichiometry and Materials Science -When Numbers Matter 190
The oxygen activity on both sides of the pellet, i.e., , can be measured using ceramic point electrodes as described in section 2.2.1 [Geffroy et al., 2011;Mauvy et al., 2007].
Another approach has been to correlate the oxygen permeation flux to oxygen partial pressures [Xu & Thomson, 1999].Assuming that the electronic conductivity is high and constant within the MIEC membrane, the oxygen flux can be written as; ( ) In case of mixed control of the oxygen flux (bulk diffusion and surface exchange), the concentration of oxygen vacancies at both faces of the membrane is governed by the reaction: where k f and k r are the forward and reverse reaction rate constants.
Under stationary conditions, the oxygen permeation can be expressed as: leading to: The three parameters D V , k f and k r can be estimated by fitting the experimental results as function the oxygen pressures on both sides of the membrane.

EMF -Faradic efficiency methods
The following cell has been used for the measurement of the ionic transport in a MIEC: Pt, O 2 (P') / MIEC / Pt, O 2 (P") According to Wagner theory, the emf of the cell obeys the equation: where ion t is the oxide ions transport number.In case of a pure ionic conductor, equation ( 23) becomes the Nernst law E th (see equation ( 1)).Equation ( 23) can often be simplified as: where ion t is the average ionic transference number.
This method has been applied to oxide electrolytes.However, it should be pointed out that the ionic transport number has to be not too small.In case of a non-stoichiometric oxide the resulting oxygen permeation flux will polarize one or both electrodes, and this method is not recommended.
The oxygen ionic transport number t O can also be measured using the Faradic Efficiency method, i.e., t O is the ratio between the oxygen current and the total current driven through the sample by an applied electrical field.However, in case of noticeable electrode polarization, the measured transport number can differ from the actual value.
The set-up is schematized in figure 16a.The cell includes an oxygen pump, an oxygen sensor and the studied membrane, all parts being separated by insulating layers.Figure 16b compares the results obtained with these techniques on La 0.9 Sr 0.1 Ga 0.8 Mg 0.2 O 3-δ : due to polarization phenomena, the classical methods give an overestimated electronic transport number.

Patterson diagrams
This method was initially proposed by Patterson [Patterson, 1971].It is based on the 2 O P dependence of the electrical conductivity.In Figure 17a is plotted the variation of log σ as functions of log ( 2 O P ) and of the reciprocal temperature.As shown in Figure 17b, at a given www.intechopen.comStoichiometry and Materials Science -When Numbers Matter 192 temperature, three oxygen pressure domains can often be defined.In the medium oxygen pressure range, the conductivity is constant and σ = σ i ; at low 2 O P , the conductivity increase is ascribed to an additional n-type electronic conductivity (σ = σ i + σ n ) and at high 2 O P t h e conductivity increase is due to an additional p-type conductivity (σ = σ i + σ p ). Assuming that the ionic conductivity is constant over the whole oxygen pressure range, the partial electronic conductivity can be easily calculated.This method can be considered as a fast and simple screening tool.It should be pointed out that the technique is convenient for electronic transport numbers higher than 1 % and when there is an oxygen partial pressure domain in which the conductivity is purely ionic.

Hebb-Wagner method
The Hebb-Wagner polarization technique has been developed either for the determination of electron and hole conductivity in ionic conductors [Hebb, 1952;Joshi & Wagner, 1975;Wagner, 1957] or for the measurement of ionic conductivity in MIECs [Riess, 1996;Wiemhöfer et al., 2002].Basically, the method consists in using a reversible electrode and blocking electrodes to suppress the predominant charge carrier and thus enable measurement of the minority species.The main limitations of the method have been reviewed [Riess, 1996] and new experimental set-ups have been proposed.
The initially proposed cell is schematized in Figure 18a.Assuming that the MIEC is cationic conductor (mobile defects: i M • ), the interface M/MIEC (1) is a reversible electrode, and the interface C/MIEC (4) is an ion blocking electrode.The polarization of the voltage is chosen so that the mobile species tend to be depleted from the blocking electrode.Under steadystate condition, the ionic current is eliminated (I = I el ) and the theoretical I-V relation, according to the Wagner's theory is the following:  [Riess, 1992b].
An analogous method has been developed for determining the ionic conductivity in a MIEC.The blocking electrode for electronic defects is a solid electrolyte (SE), that conducts the same ions as the MIEC ( i M • defects, is the experiment described, see Figure 18b): (-) M/Solid Electrolyte ( i M • )/MIEC ( i M • , e', h • )/M (+) Multi-electrode set-ups have been proposed by Riess, eliminating experimental problems such as errors due to overpotential at both electrodes or allowing the simultaneous measurement of σ el and σ i in a MIEC [Riess, 1992a[Riess, , 1992b[Riess, , 1996]].

Short-circuit method ("zero driving force" method)
The ionic conductivity in a MIEC can be determined using the "short-circuiting" method [Riess, 1991].Both electrodes are reversible systems with different compositions.The electronic current is brought to zero by short-circuiting the MIEC on a low impedance amperometer (V = 0).Consequently, it can be demonstrated that the ionic resistance R i obeys the following equation [Riess, 1991]: where E th is the Nernst voltage determined by the compositions of the reversible electrodes, and I is the short-circuit current.
The main advantage of the method is that the contribution of electronic defects to the conductivity is eliminated without requirement of a blocking electrode.The sources of errors have been discussed [Riess, 1991[Riess, , 1992b[Riess, , 1997] ] and the method was applied to mixed conducting pyrochlores [Riess et al., 1992].

Conductivity vs. non-stoichiometry ratio
The electrical conductivity of electrochemically-reduced oxides is measured by impedance spectroscopy.The gas circuit is schematized in Figure 19a.During the first stage (indexes 1) the sample is reduced electrochemically (according to the procedure described in Section 3.2.3)and the stoichiometry ratio is determined using the oxygen sensor.Then the gas flowing in the experimental cell is purified using an oxygen getter (indexes 2).It has been checked that the sample composition remains constant over several days.The electrical conductivity of the reduced sample is measured as a function of temperature.Levy et al., 1988).

Relaxation methods
In chemical relaxation experiments, an abrupt change of chemical potential of one of the constituent elements, usually PO 2 for oxide materials is imposed on a sample under constant temperature.Physical properties such as weight and volume of the sample are recorded versus time until a new thermodynamically equilibrium state is reached.Because electrical conductivity is much more sensitive to change of oxygen chemical potential in the atmosphere than are the other properties as weight, considerable changes in conductivity can be observed even when the oxygen partial pressure change of the corresponding nonstoichiometry is very small.This makes the conductivity relaxation method more easily applicable to a wider variety of materials than other methods such as thermogravimetry  et al., 2000;Ma et al. 1997;ten Elshof et al., 1997].The transient behaviour in the re-equilibration process is recorded and analyzed by fitting the relaxation data to the solution of Fick's second law with appropriate boundary conditions [Crank, 1975].
A classical experimental setup, used for this type of study, is schematically drawn in Fig. 20a.Electrical conductivity relaxation experiments are performed using the four-probe method.Four platinum wires are connected to the sample using platinum paste (see Fig. 20b).Various atmospheres surrounding the sample are obtained by flowing a mixture of oxygen and nitrogen into the sample chamber, using two mass flow controllers or by introducing air.In both cases, the same flow rate is used.During the change of atmosphere, the flow rate is high enough to stabilize PO 2 inside the cell within a short duration compared with the relaxation time.Oxygen partial pressure steps of small amplitude are required to assign the relaxation process to a defined oxygen partial pressure.Moreover, this amplitude has to be small enough to assume constant the mobility of the defects during the relaxation process.In the case of La 2 Cu 0.5 Ni 0.5 O 4+δ compound, oxygen is incorporated into the crystal lattice during the oxidation process or released from the lattice during the reduction step, as it can be observed on Fig. 21 when the sample is subjected to a sudden change of PO 2 [Mauvy et al., 2004].
During the relaxation process, nonstoichiometry spreads through the sample by lattice diffusion, which is driven by concentration gradients of defects.The ionic conductivity of La 2 Cu 0.5 Ni 0.5 O 4+δ can only be contributed by oxygen vacancies ( O V •• ) and interstitial oxygen ions (O i '').Theoretically, the oxygen vacancy concentration should decrease with increase in oxygen partial pressure.So, if the oxygen vacancy is the controlling defect, then one should expect the ionic conductivity to decrease with increase PO 2 .However, experimental observations (see Fig. 21), indicate that ionic conductivity increases.This suggests that O i '' is the predominant defect in this working conditions (high PO 2 ).Because the mobility of cationic defects is much lower than that of oxide defects and holes, chemical diffusion can be considered as the process of diffusion of interstitial oxide ions and counterdiffusion of holes.The transient behaviour in the re-equilibration process can be described by the second Fick law [Sitte, 2001].Taking into account the aspect ratio of the sample, the thickness of the pellet (slab like) controls the kinetic of the oxygen re-equilibration process.Consequently, the diffusion can be treated as a one-dimensional mechanism leading to the following equation: If the chemical diffusion is the rate-determining step, the following equation can be derived from the second Fick law, as suggested by Crank [Crank, 1975] where L is the diffusion length and t the time.M t / M ∞ represent the ratio of the mass of diffused oxide ions at time t, to that obtained for an infinite time.
Assuming constant the number of charges and the mobility of the charge carriers, the apparent conductivity is given by integrating the local conductivity all over the sample.
Finally, the conductivity ratio can be directly related to the mass ratio:   This observation confirms that the relaxation process is not controlled by the surface exchange, but by the bulk diffusion (i.e., this latter process is the limiting step).

Conclusion
Determination of nonstoichiometry in oxides is a key point in the search for new materials for electrochemical applications.In recent decades, owing to their current and potential applications (electrodes in fuel cells, insertion electrodes, membranes of oxygen separation, gas sensors, catalytic materials, etc.), various methods of precise characterization of MIECs have been proposed, either the measurement of the defect concentrations and the stoichiometric ratio as functions of the oxide composition, of the surrounding oxygen pressure and of temperature, or the transport properties.There are different methods to determine the electrical properties of MIECs and, more specifically, the ionic and electronic contributions.The most appropriate method depends on different parameters, i.e., the total electrical conductivity of the studied oxides, the ionic and electronic transport numbers, the temperature domain, etc.As examples, the characterization of solid electrolytes requires the determination of electronic transport numbers as low as 10 -4 ; oxygen leakage currents through interconnect materials in solid oxide fuel cells require the measurement of ionic conductivity in a practically pure electronic conductor; in ceramic membranes devoted to oxygen permeation, one of the most important criteria concerns the high ionic and electronic conductivities of the oxide.In the case of MIEC oxides used as oxygen electrode, the electrocatalytic activity is related to the properties of mixed conduction and, therefore, of nonstoichiometry.The determination of the level of ionic conductivity is fundamental for the material selection but this characterization is generally difficult to carry out because of the high electronic contribution to the total transport properties.
In this chapter all the proposed methods have not been described.The objective was to focus on the sources of errors, which may render the results useless.Deviation from equilibrium of the MIEC surface due to oxygen semipermeability flux can lead to erroneous measurements.Experimental set-ups, which allow overcoming these experimental difficulties, were described.
Fig. 4. (a): Ceramic point electrode; (b): Schematic drawing of the cell allowing the measurement of the oxygen activity on the surface of a MIEC.The potential change in the electrochemical chain can be summarized as (E = φ Pt-I -φ Pt-II ): also assumed that there is no oxygen flux at the interface between the ceramic point electrode and the pellet (
Fig. 9. (a): Experimental device for the study of gas-solid reaction under constant oxygen pressure using a pump-sensor device, G: vacuum gauge, LN: liquid nitrogen trap; (b): Variation of the stoichiometry ratio of CeO 2-x at 1273 K.
Fig. 13.(a): Experimental set-up for emf measurement; (b): Photograph of a nickelate point electrode (Ni 1.95 NiO 4+δ ); (c): Variation of the oxygen activity coefficient in the nickelate as a function of temperature for various oxygen activity values in the gas phase [Mauvy et al., 2009].
corresponding oxygen activities on both surfaces of the membrane.The experimental cell for oxygen permeation measurements, including metallic and ceramic point electrodes is shown Figure14 b.The chemically driven oxygen flux through a mixed-conducting oxide was first modelled using Wagner's theory, assuming that both oxide surfaces are in equilibrium with the imposed gas atmospheres ( Fig. 14.(a) Schematic drawing of the oxygen pressure gradients within a permeation membrane, (b) Permeation set-up, with magnified drawing of the ceramic point electrode.
Fig. 15.(a): Variation of the oxygen permeation flux density (a) and of the specific oxygen permeability (b) with the oxygen pressure gradient, from[Kharton et al., 1999].

Figure
Figure 19b gives the Arrhenius plot of the electrical conductivity of stabilized zirconia (ZrO 2 -Y 2 O 3 , 12 m/o) as a function of the stoichiometry ratio.
www.intechopen.comDetermination of Thermodynamic and Transport Properties of Non-Stoichiometric Oxides 195 analysis [Lane

Fig. 20 .
Fig. 20.Schematic drawing of the experimental setup (a) and picture (b) of the 4-electrode sample holder.
σ(0), σ (t) and σ (∞) denote the apparent conductivity at t = 0 (initial), at time t (in the course of relaxation) and for t  ∞ (after reaching a new equilibrium state), respectively.The experimental data and the fitting curve in the form of fractional conductivity change as function of time are shown in Fig.22.The chemical diffusion coefficient D  is then obtained by least-square fitting of the relaxation data.

Fig. 23
Fig.23shows the Arrhenius plots of D  measured on two different samples of the same composition La 2 Cu 0.5 Ni 0.5 O 4+δ , but with two different thicknesses.Oxygen chemical diffusion coefficients, obtained for both samples, are equal over a wide temperature range.This observation confirms that the relaxation process is not controlled by the surface exchange, but by the bulk diffusion (i.e., this latter process is the limiting step).