Standard Article

You have free access to this content

Methane reforming kinetics, carbon deposition, and redox durability of Ni/8 yttria-stabilized zirconia (YSZ) anodes

Advances in Electrocatalysis, Materials, Diagnostics and Durability

Performance degradation

High-temperature fuel cells

  1. E. Ivers-Tiffée1,
  2. H. Timmermann1,
  3. A. Leonide1,
  4. N. H. Menzler2,
  5. J. Malzbender2

Published Online: 15 DEC 2010

DOI: 10.1002/9780470974001.f500063

Handbook of Fuel Cells

Handbook of Fuel Cells

How to Cite

Ivers-Tiffée, E., Timmermann, H., Leonide, A., Menzler, N. H. and Malzbender, J. 2010. Methane reforming kinetics, carbon deposition, and redox durability of Ni/8 yttria-stabilized zirconia (YSZ) anodes. Handbook of Fuel Cells. .

Author Information

  1. 1

    Universität Karlsruhe (TH), Karlsruhe, Germany

  2. 2

    Institute of Energy Research, IEF, Jülich, Germany

Publication History

  1. Published Online: 15 DEC 2010

1 Introduction

  1. Top of page
  2. Introduction
  3. Experimental
  4. Conversion of Hydrocarbon Fuels
  5. Carbon Formation and Degradation Effects
  6. Redox Durability
  7. Summary and Conclusions
  8. Acknowledgment
  9. References

Solid oxide fuel cell (SOFC) anodes should exhibit high efficiency with respect to charge transport in the solid and transport processes in the gas phase as well as catalytical and electrochemical reaction steps. As the extent to which these processes occur is (mostly) space dependent, state-of-the-art anode structures consist of various functional layers that differ in thickness, microstructure, and chemical composition. The most commonly used and, hence, best-investigated materials system in this regard is an anode cermet (ceramic–metal) composed of metallic, electronic-conducting, and catalytically active nickel and ionic-conducting yttria-stabilized zirconia (YSZ).

Anodes for stationary SOFC systems are most likely to be operated with natural gas at temperatures above 800 °C, whereas mobile SOFC applications such as auxiliary power units (APUs) operated below 800 °C use (liquid) hydrocarbons. The fuel (gasoline or diesel) cannot be converted by 100% in the reformer and so a reformate containing higher hydrocarbons has to be converted at the anode. The catalytic conversion takes place in the near-surface region of the anode substrate, whereas the electrochemical oxidation of H2 to H2O (or of CO to CO2) proceeds close to the electrolyte/anode interface (cf. region II in Figure 1). Outside of region II, the anode is subject to various processes that lead to an aging and, thus, degradation in the electrochemical performance of the anode.

thumbnail image

Figure 1. Operating regions for Ni/8YSZ anodes in stationary and mobile SOFC systems. Regions of carbon deposition (I) and reoxidation (III) surround the area of electrochemistry (II). The dominating process depends on oxygen partial pressure

The metallic phase within the anode cermet, mostly nickel, is oxidized at higher oxygen partial pressures (cf. region III in Figure 1). At high fuel utilization, this reoxidation of Ni to NiO can be caused by a breakdown of the fuel gas supply or by frequent start/stop operations without auxiliary gas (N2). In most cases, this is accompanied by a measurable volume change of the (initially) metallic Ni phase, resulting in a damage to (or degradation of) the whole anode structure.1 Sequential cyclic redox can quickly lead to an aging of the anode or even a complete failure of the cell (fracture of the thin film electrolyte).2 The processes leading to a mechanical damage and the resulting changes in the electrical properties during redox cycles are discussed in detail in this article. When anodes are operated with hydrocarbons, i.e., at low oxygen partial pressures (cf. region I, Figure 1), there is the risk of damage ranging from (i) reversible damage by occupancy of the catalyst surface with various deposits, like coke or tar molecules, to (ii) irreversible damage to the Ni catalyst by carbon whisker growth.3

Moreover, the kinetics of methane conversion under various operating conditions, the carbon formation in the presence of higher hydrocarbons (such as acetylene), and the analysis of electrochemical oxidation and gas diffusion in stable operations with hydrogen or biomass-derived gas are also treated in this article.

1.1 Microstructure of Ni–YSZ Cermets

The Ni/YSZ cermet of an anode-supported cell (ASC) SOFC consists of two (or more) layers with graded functionality as to the Ni:YSZ ratio, porosity, and particle size of the components, as shown in Figure 2. The electrochemically active triple-phase boundary (TPB) follows the solid electrolyte (layer I) and is composed of submicron-sized Ni and YSZ grains. In this layer that can be fairly thin (5–15 µm, depending on the operating temperature4), electrochemical oxidation of H2 to H2O (and of CO to CO2) takes place. Several well-distributed large YSZ particles improve the adhesion to the electrolyte and are able to slow down or inhibit an agglomeration of the fine-grained nickel at high current loads.5-7 The charge transport from the active TPB to the interconnects calls for a modification of the subsequent layer (layer II). A smaller resistance can be achieved by a higher volume fraction as well as a larger grain size of Ni. The mean porosity can be increased, e.g., by coarser YSZ particles, thus facilitating the transport of educts and products. At high current densities or high fuel utilizations, respectively, this yields lower values of gas diffusion polarization as well as a decreased nickel oxidation. Layer III can, in turn, be modified both for the catalytic conversion of hydrocarbons and for the contact with the interconnects.

thumbnail image

Figure 2. Multilayer anode substrate with different functional layers in regions I, II, and III. Variations in Ni/YSZ ratio, porosity, and particle size result in a change of relevant properties

2 Experimental

  1. Top of page
  2. Introduction
  3. Experimental
  4. Conversion of Hydrocarbon Fuels
  5. Carbon Formation and Degradation Effects
  6. Redox Durability
  7. Summary and Conclusions
  8. Acknowledgment
  9. References

Apart from literature data, the results presented in this contribution include measurements performed on anode-supported half-cells and single cells. The anode substrate consisted of Ni/8YSZ with different thicknesses (1000 and 1500 µm).8 Onto these, an 8YSZ thin film electrolyte was sintered with a thickness of 10 µm. The cathode (approximately 40 µm) was made of LSM (La0.65Sr0.3MnO3 − δ)/8YSZ or LSCF (La0.58Sr0.4Co0.2Fe0.8O3 − δ). The formation of insulating secondary phases (lanthanum and strontium zirconates) at the LSCF cathode/electrolyte interface was prevented by a screen-printed ceria (CGO: Ce0.8Gd0.2O2 − δ) interlayer. Detailed information on preparation and properties of these cells is given in Refs. 9-11.

The electrical measurements presented here were carried out on ASCs with a substrate thickness of 1500 µm (substrate area 5 × 5 cm2) and a mixed-conducting LSCF cathode (cathode area 1 cm2, thickness 40 µm). Details of the measurement setup for electrical impedance analysis and about the data evaluation by means of the distribution function of relaxation times (DRT) can be found in Electrochemical impedance spectroscopy and Ref. 12.

The measurement setup and measuring conditions for the investigations related to carbon deposition and the determination of kinetic data for reformate operation are described in detail in Refs. 13, 14 respectively. The results presented here were obtained on ASCs with an anode substrate thickness of 1000 µm (substrate area 5 × 5 cm2) and an LSCF cathode (cathode area 16 cm2, thickness 40 µm).

Half-cells (anode substrate + thin film electrolyte with variable anode substrate thickness t (2700 µm > t > 300 µm)) were exposed to various redox procedures (for details, see corresponding subchapters). Electrochemical performance degradation was determined by I/V measurements during subsequent redox cycling of ASCs (substrate thickness of 1500 µm, substrate area 5 × 5 cm2) with LSM/YSZ cathode (cathode area 1 cm2, thickness 40 µm). For comparison, electrolyte-supported cells (ESCs) from Indec have been tested with the same redox procedure.

3 Conversion of Hydrocarbon Fuels

  1. Top of page
  2. Introduction
  3. Experimental
  4. Conversion of Hydrocarbon Fuels
  5. Carbon Formation and Degradation Effects
  6. Redox Durability
  7. Summary and Conclusions
  8. Acknowledgment
  9. References

For SOFC operation with hydrocarbon fuels, two important conditions have to be fulfilled:

  • Stability, i.e., prevention of carbon formation and deposition, permanent blocking of electrochemically active catalyst surface by adsorption of higher hydrocarbons or sulfurous components, or chemical reactions with the Ni catalyst, e.g., resulting in Ni3C or NiS. The topic of reducing instable hydrocarbons by external reforming has already been addressed in Volume 3 of this series (see Steam reforming, ATR, partial oxidation: catalysts and reaction engineering).

  • Efficiency, i.e., development of a system with a lesser degree of complexity and direct utilization of hydrocarbons in a stable operating point.

An introduction to internal reforming/direct oxidation of hydrocarbons can be found in Volumes 2 and 4 of this series (see Internal reforming; Oxidation reactions in high-temperature fuel cells; Direct hydrocarbon SOFCs). In the present article, reforming and kinetics of this reaction are discussed in detail for stable operating conditions, with focus on methane, the thermodynamically most stable hydrocarbon, as key component for hydrocarbon conversion. Predominantly, the conversion of hydrocarbons proceeds via the reforming reaction, as exemplified for methane:

  • mathml alt image(1)

Theoretically, hydrocarbons can also be electrochemically oxidized, but the kinetics of this reaction is very slow compared to the kinetics of the reforming reaction on a Ni anode.15

In parallel to reaction 1, the water–gas shift reaction proceeds simultaneously:

  • mathml alt image(2)

3.1 Experimental Analysis of Internal Reforming

Internal reforming on SOFC Ni/YSZ standard cermets was studied in several papers;14, 16-22 detailed reviews on the conversion of hydrocarbons can be found in Refs. 15, 21, 22. The gas composition along the anode gas channel or at the outlet of single cells was investigated by several groups by (in situ) gas analyses.16, 17, 19, 20 Rate expressions for the reforming reaction on Ni/YSZ are given in Refs. 22-25 The effect of addition of promoters like alkaline earth metal oxides, precious metals, and ceria-based materials on the reforming reaction rate was studied in Refs. 19, 20, 26, 27. In our own measurements, the methane conversion was monitored along the gas channel in an ASC by in situ gas analysis while varying the temperature from 600 to 850 °C and the methane mole fraction in the hydrogen fuel stepwise from 0 to 20%. Furthermore, the effect of the gas components CO, CO2, H2, and H2O on the methane conversion was studied at 600 °C. The fuel gas inlet flow velocity was set to a value of v = 1 m s−1, corresponding to a residence time of τ = 0.04 s alongside the anode (channel length l = 4 cm). The S/C ratio (steam-to-carbon ratio), defined as the mole ratio between the amount of steam and the amount of methane, was varied from S/C = 1–6. The mole flow at the anode gas outlet was calculated with the help of the nitrogen mole fraction, which was used as internal standard.

In Figure 3, the methane conversion at the outlet of the cell (l = 4 cm) is shown, determined by gas chromatography at T = 750 °C and 850 °C as a function of the S/C ratio. The methane conversion inline image is calculated according to

  • mathml alt image(3)

inline image and inline image denote the mole flow of methane at the cell in- and outlet, respectively. At 850 °C, the methane conversion takes on a value of 93%, independent of the varied S/C ratio, thus not quite attaining the (theoretical) equilibrium values of 95% (for S/C = 1) and 99% (for S/C = 3). At 750 °C, a methane conversion of 80–82% is determined, which is also lower than the corresponding equilibrium values of 93% (for S/C = 1) and 98% (for S/C = 3). At 850 °C, the measured methane conversion does not depend on the amount of methane at the cell inlet, whereas at 750 °C the methane conversion slightly decreases with increasing amount of methane in the feed, as depicted in Figure 4.

thumbnail image

Figure 3. Methane conversion at anode gas outlet in dependence of the S/C ratio at yinline image, inline image, yCO = 0.15, inline image, balance: N2 for two different temperatures

thumbnail image

Figure 4. Methane conversion at anode gas outlet in dependence of the methane partial pressure at inline image, inline image, yCO = 0.15, inline image, balance: N2 for two different temperatures

For the reforming reaction (equation 1), the equilibrium is shifted toward the formation of methane and steam in the cell at lower temperatures and small CH4 and H2O concentrations. This can be seen in Figure 5, where the methane mole fraction at the inlet was set to values of 1 and 0.1% at 750 and 650 °C, respectively. At 750 °C, the methane supplied was converted along the cell; in contrast, at 650 °C and at a far lower amount of methane supplied, negative conversion is observed, implying that methane is formed within the cell. As can be seen from Figure 6, the more CO and H2 are present, the more methane is formed. By increasing the steam mole fraction, the equilibrium of the reforming reaction (equation 1) is shifted toward the products CO and H2, resulting in a decrease of the methane mole fraction at the gas outlet. By increasing the methane fraction at the inlet from 0.1 to 2%, the conversion approaches zero as a result of a convergence toward equilibrium composition for the operating conditions considered. However, these measurements carried out with a conversion up to 93% at the cell outlet do not permit any conclusions regarding the local gas composition and the kinetics of reforming. For such statements, the development (and verification) of a reaction model is required.

thumbnail image

Figure 5. Methane conversion along the anode for inline image, inline image, yCO = 0.15, inline image, balance: N2 for two different temperatures

thumbnail image

Figure 6. Methane Conversion at anode gas outlet in dependence of the gas composition at 600 °C. Variation begins with standard gas composition inline image, inline image, inline image, yCO = 0.15, inline image, balance: N2. Symbols account for measured values, lines for fitted values (see Section 3.2)

3.2 Modeling of Methane Conversion

For Ni/YSZ cermets, several kinetics for the reforming reaction were determined in literature (see Table 1). It must be taken into account that composition and microstructure of the anodes investigated varied over a wide range and the operating parameters (temperature range and gas composition) at which the rate of the reforming reaction was determined also differed. The same holds for the mathematical equations and the values of the parameters employed.

Table 1. Kinetics for Methane Conversion on Ni/YSZ Cermet Anodes (Literature Overview)
AuthorRate of reforming reactionnT( °C)EA (kJ mol−1)Anode thickness (µm)Porosity (%)Anode composition (wt%)
Achenbach25inline image1 800–1000821.41034022% Ni/78% ZrO2
Lee24inline image1 800–100098.5Crushed cermet3860% Ni/40% YSZ
    74.6 5570% Ni/30% YSZ
Ahmed23inline image0.85 854–9079550??
Dicks16inline image1 700–1000135Crushed cermet40–4555% Ni/45% YSZ
This workinline image1 600–85030–611 × 1034051% Ni/49% YSZ

Achenbach, Lee, and Ahmed found that the dependence of the reforming rate on the partial pressures of methane and steam can be described by a power law, whereas Dicks et al. employed a Langmuir–Hinshelwood expression that includes the influence of adsorption and desorption effects on the anode surface. In all cases, the reaction order of methane n is between 0.85 and 1. However, Achenbach found a reaction order of zero for steam, whereas Ahmed and Lee obtained negative values for this parameter. In both of the latter studies, the S/C ratio was, however, always over stoichiometric with respect to the reforming reaction (S/C >1). This goes along with a large amount of excess steam that possibly adsorbs on the anode, blocking catalytically active sites for methane reforming. This can lead to a decrease of the reforming rate with increasing steam partial pressure and would explain the negative reaction order of steam. Dicks et al. account for this effect by employing a rate expression of the Langmuir–Hinshelwood type with the steam partial pressure in the denominator. The values for the parameters of the activation energies significantly differ, because they strongly depend on the properties of the anode, i.e., exact composition, porosity, thickness, and the number of active Ni sites.

Especially for thick anode cermets with low porosity, diffusion can limit the effective reaction rate. In these cases, the total activation energy of the formal reaction rate is lower than for the intrinsic case (see mathematical analysis of this effect below) because the temperature dependence of diffusion is not as pronounced as the temperature dependence of the intrinsic methane reforming rate.

In the following, a mathematical model is presented for the reforming reaction on the Ni/YSZ anode described above. This model allows a quantitative assessment of the effective catalytic activity of the anode and takes into consideration both the reforming reaction (equation 1) and the water–gas shift reaction (equation 2). Steady state and ideal gas behavior are assumed. Diffusive gas transport in the anode gas channel over the width of the channel is neglected because the channel is very narrow (channel width 1.5 mm).

The influence of axial dispersion can be estimated by the Bodenstein number Bo, which is defined as the ratio of convective transport to diffusive transport:28

  • mathml alt image(4)

with the flow velocity v, the axial dispersion coefficient Dax,28 and the channel length l. Axial dispersion was not taken into account in the model, because the Bodenstein number is always higher than 70. The temperature is considered to be constant in the cell and energy balances are not included in this version of the model. Atmospheric pressure is assumed to prevail constantly throughout the gas channel.

The following assumptions reduce the number of unknown variables to 13:

  • The flow velocity v in x direction that changes because of the increase in mole flow by the steam reforming reaction (equation 1).

  • The partial pressures of the components in the gas channel pi, and on the external anode surface pi, s, with i = CH4, H2O, CO, H2, CO2, and N2.

All these variables are a function of the coordinate x in flow direction.

With the above-mentioned assumptions, the mole balance for each species can be written as

  • mathml alt image(5)

with i = CH4, H2O, H2, CO, and CO2, N2;df denotes the height of the anode gas channel, which is set to 1.5 mm. βi denotes the mass transfer coefficient for the mass transfer from the channel to the anode surface.29, 30

In addition, Dalton's law is applied:

  • mathml alt image(6)

The partial pressures pi, s of the components on the anode surface are calculated by species balances around the anode surface for each position x:

  • mathml alt image(7)

υi, r and υi, r are the stoichiometric coefficients for component i in the reforming reaction and in the water–gas shift reaction, respectively.

rs is the area-specific conversion rate of the shift reaction calculated from the equilibrium condition (see Ref. 14 for more details).

For the reforming reaction, a kinetic expression of the type

  • mathml alt image(8)

is used with

  • mathml alt image(9)

rr is the reforming rate, which is related to the external anode surface, k0 a constant, and EA the activation energy of the reforming reaction. R denotes the universal gas constant and T the temperature. Kref is the equilibrium constant of the reforming reaction.

This system of equations was implemented in Matlab®, where equation 5 was discretized and solved by the Euler method. The values of k0, EA, n, and m were obtained by fitting the simulated values to the measurement results shown above. Table 2 lists the values determined for the parameters. In Figure 7, the values determined for the reaction constant k in the Arrhenius form are given. The curve shows a sharp bend above T = 750 °C, resulting in a lower value for the activation energy in the temperature range between 750 and 850 °C, as compared to its value between 600 and 750 °C. Possible reasons for the occurrence of different activation energies are discussed below.

thumbnail image

Figure 7. Arrhenius plot of the reaction constant k for the reforming reaction. Symbols account for measured values, lines for fitted values

Table 2. Values of the Parameters k0, EA, n, and m in Equations 8 and 9, Respectively
Temperature ( °C)k0 (mol s−1· m−2 · bar−1)EA (kJ mol−1)nm
600–750 1483 61 1 0
750–850 38 30 1 0

3.3 Modeling Validation

Figure 6 shows a comparison of the results from simulations and measurements at 600 °C. At this temperature, the model is able to describe the dependency on gas composition very well. The temperature dependence of the methane conversion is depicted in Figure 8 for two different methane mole fractions at the inlet. Here, too, the measured values are in good agreement with the results of the simulation. The most significant deviation between simulation and measurement is obtained for a low conversion at 700 °C. In this case, the value of the equilibrium constant, which varies by up to 7% depending on the method of its calculation, has a significant influence on the simulation results. Moreover, the experimental error is largest for low methane values. Another cause of this deviation could be due to diffusion effects in the anode substrate. Since the mean difference between simulation and measurement amounts to only 7%, the temperature dependency can be simulated over a wide range within a reasonable accuracy.

thumbnail image

Figure 8. CH4 conversion as a function of temperature given for two methane mole fractions at anode gas inlet. Symbols account for measured values, lines for fitted values

So far, only simulated results for methane conversion at the cell outlet were considered. Now, the spatial dependence and the dependence on the flow velocity are determined by using the residence time τ (according to equation 10) and taking into account the local mean flow velocity v̄(x) (which is v averaged over x).

  • mathml alt image(10)

The simultaneous description of the methane conversion as a function of temperature succeeds by introducing the Dammköhler number DaI29:

  • mathml alt image(11)

Here pinline image denotes the partial pressure of methane at the reactor inlet. As presented in Figure 9, the simulated results fit well the measured values over a wide range of operating conditions. Since all plots that have been presented show good correlations between simulated and measured values, the model is validated for the investigated parameter range.

thumbnail image

Figure 9. CH4 conversion as a function of the Damköhler number DaI. Comparison of simulation results with measured values. T = 600–850 °C, inline image, inline image, yCO = 0.15, inline image, inline image, balance: N2, v = 1 m s−1, x = 1–4 cm

As a conclusion of the results presented on kinetic modeling, the kinetic expression employed seems to be appropriate to describe the rate of the methane reforming reaction on a Ni/YSZ anode substrate at the operating conditions investigated. The reaction order is 1 for methane and 0 for steam, wherein steam partial pressure was varied between 0.02 and 0.3 bar (corresponding to a variation of the S/C ratio from 1 to 6). The presence of steam thus shows no influence on kinetics over a wide S/C range, yet it influences the position of equilibrium and therefore the driving force of the chemical reaction.

The apparent activation energies of 30 kJ mol−1 in the range of 750–850 °C and 61 kJ mol−1 in the range of 600–750 °C given in Table 2 are much lower than the literature values cited in Table 1. These considerable differences can be a result of the higher catalytic activity of the present Ni/YSZ anode substrate with a thickness of 1000 µm investigated here or of a limitation of the reaction rate due to diffusion in the pores in the present cermet anode.

The influence of pore diffusion can be estimated by the Weisz–Prater criterion31 with an effective pore diffusion coefficient Deff and the thickness of the anode d.

  • mathml alt image(12)

If this condition is fulfilled, pore diffusion can be neglected. We obtained values between 0.1 and 1000, depending on methane inlet mole fraction and temperature. Therefore, pore diffusion has to be taken into consideration in any case. Especially at temperatures above 750 °C the activation energy is decreased by diffusion processes in the anode substrate. Nevertheless, using the apparent activation energies of 30–61 kJ mol−1, the simulation of the methane conversion on the Ni/YSZ anode is in good accordance with the measured values.

4 Carbon Formation and Degradation Effects

  1. Top of page
  2. Introduction
  3. Experimental
  4. Conversion of Hydrocarbon Fuels
  5. Carbon Formation and Degradation Effects
  6. Redox Durability
  7. Summary and Conclusions
  8. Acknowledgment
  9. References

In the presence of hydrocarbons and carbon monoxide, carbon can be formed and deposited according to the Boudouard reaction:

  • mathml alt image(13)

or via cracking of higher hydrocarbons:

  • mathml alt image(14)

Furthermore, reaction between CO and H2 can result in carbon formation:

  • mathml alt image(15)

Under SOFC operating conditions, several investigations have been conducted with hydrocarbons higher than methane. These studies include methanol,32, 33 ethanol,33, 34 ethane,35-37 ethylene,35 propane,38-44 butane, 36, 41, 45-50 iso-octane,51-53 n-decane,41, 47 toluene,41, 47 dodecane,54 dimethyl ether (DME),34 methanoic acid,34 glycerol,34 dimethoxymethane,34 and n-methyl- methanamide.34

4.1 Mechanisms of Carbon Formation

In the following, the mechanisms of carbon formation are discussed, with respect to their dependency on temperature and gas composition.

Carbon deposition under thermodynamically feasible conditions has been dealt with in Refs. 34, 36, 37, 46, 47, 50. The formation of such deposits can be initiated by homogeneous reactions in the gas phase,50, 55 or by catalyzed reactions on the surfaces of metals like Ni and Fe.3 The latter is especially significant in the case of Ni-containing anodes. Here, three different mechanisms leading to carbonaceous deposits (coke) can be distinguished:3

  1. Hydrocarbons are adsorbed on the Ni surface and form a nonreactive film that can encapsulate the Ni particles at temperatures lower than 500 °C. The catalytic Ni surface is thus deactivated by this process.

  2. At temperatures higher than 600 °C, pyrolysis of hydrocarbons can occur on the Ni surface, which also leads to a deactivation of the catalytic Ni surface.

  3. At temperatures higher than 450 °C, diffusion of adsorbed carbon from the surface into the Ni particle takes place, followed by a nucleation process. The carbon then starts to grow in a fibrous (whisker-like) structure out of the Ni particle, thereby lifting parts of the Ni particles away and thus irreversibly breaking apart the anode microstructure. In extreme cases, this process may ultimately lead to a complete destruction of the anode.

4.2 Experimental Analysis of Carbon Formation

The formation of carbonaceous deposits in SOFC operation with several hydrocarbon fuels (n-butane, naphta, toluene, and n-decane) was observed by46, 47 The authors ascribed these deposit formations to gas-phase reactions to the fact that the anode materials under investigation (Cu/CeO2/YSZ) do not catalyze coke formation. At temperatures above 700 °C, the blocking of the pores and the resulting inhibition of the fuel supply to the electrochemically active sites lead to a strong decrease in power density within a few hours, as could be shown by comparison of I/V curves before and after operation with hydrocarbons.47 The deposits could be removed by oxidation with steam at temperatures higher than 800 °C (in the reverse of reaction 15) and at 627 °C with oxygen to CO2, thereby restoring the initial values of power density.

Under similar experimental conditions, in the case of Ni-containing anodes, carbon whiskers would form. Their formation kinetics strongly depends on the kind of hydrocarbon utilized.3 The formation rate increases with the unsaturated character of the hydrocarbon involved. Acetylene and ethylene are known for a very high affinity to coke and soot formation,56, 57 Kikuchi et al. reported severe degradation of Ni/YSZ anodes operated with ethane and ethylene at S/C = 3.5 after few hours' operation.35

In Ref. 13 the effect of coking was studied at 650 °C in the presence of an acetylene-containing reformate. The anode substrate and functional layer of the ASC under investigation consisted of a Ni/YSZ cermet described in the experimental part. Figure 10 illustrates the electrochemical performance of ASCs during galvanostatic operation with and without acetylene. After 4 h of operation with acetylene, a dramatic increase in the degradation rate can be observed (ΔUt > 1 mV h−1). After 10 h of operation with acetylene, black and gray deposits on the anode substrate surface are macroscopically visible as well as structural changes in the anode volume. A scanning electron microscope (SEM)/energy dispersive X-ray analysis (EDX analysis) reveals that

  • on the surface, the black deposits are coke (Figure 11);

  • on the surface, the gray deposits consist of Ni as well as YSZ, probably initiated by the formation of carbon whiskers and subsequent stepwise removal of the nickel (and YSZ) phase; and

  • in the volume, carbon whiskers have grown out of the Ni particles.

thumbnail image

Figure 10. Electrochemical performance of an ASC at inline image%, T = 650 °C, with inline image%, yCO = 22%, inline image%, inline image%, inline image%, balance: N2 and j = 250 mA cm−2

thumbnail image

Figure 11. Deposits on the surface of an anode substrate after 10 h operation of a single cell with inline image%, yCO = 22%, inline image%, inline image%, inline image%, balance: N2 at 650 °C and j = 300 mA cm−2

This irreversible structural change within the Ni/YSZ anode is caused by the presence of the unsaturated hydrocarbon acetylene; operation with a reformate that does not contain any acetylene leads to no detectable decrease in the electrochemical performance after 10 h, as demonstrated in Figure 10. The formation of carbon whiskers is initiated by decomposition of the hydrocarbons on the Ni catalyst surface; accumulation and growth or back-formation are determined by the reaction rates of hydrocarbon decomposition and subsequent gasification with steam. With higher hydrocarbons such as acetylene the reaction rate of decomposition can proceed faster than the gasification rate, thus promoting the formation of carbon whiskers.

4.2.1 Cell Degradation Effects in Biomass-Derived Gas

Depending on the composition of the fuel gas, coke formation and resulting degradation of the cell performance can also be observed even if the formation of a carbon phase is thermodynamically not possible.35 This could be shown for an operation with biogas derived from corn silage.

The gas composition of this biomass-derived gas was analyzed (cf. Table 3) and, from a thermodynamical point of view, cannot account for any carbon formation. An ASC was operated for 100 h in H2 (5.5% H2O) and subsequently for 14 h at an S/C ratio of 4 and an operating temperature of 793 °C at open cell voltage (OCV) with this biomass-derived gas (Figure 12). Clearly, the moderate degradation behavior of the anode polarization during hydrogen operation suddenly changes when biomass-derived gas is supplied to the anode. Compared to the operation in H2 (5.5% H2O), the polarization resistance strongly increased by using the biomass-derived gas (Figure 13a). A DRT analysis (Figure 13b) furnished proof that within 14 h both the losses caused by the electrochemical conversion in the anode functional layer (R2A, R3A) and those caused by the gas diffusion in the anode substrate (R1A) increased by 9.5 and 6.5%, respectively.

thumbnail image

Figure 12. Schematic measurement sequence followed during degradation experiment with a change in gas composition from H2 to biomass-derived gas. Arrows depict electrochemical impedance spectroscopy (EIS) measurements

thumbnail image

Figure 13. (a) Impedance curves and (b) corresponding distribution of relaxation times obtained at three different times (t0, t1, and t2) during degradation experiment. (inline image atm (balance H2), T = 800 °C, cell ID Z2_152, op. = operation)

Table 3. Composition of the Biomass-Derived Gas Supplied to the Cell
Component Volume fraction (vol%)
Methane 17.96 Hydrocarbons 25%
Sum C2 4.34
Propane 2.04
Butane 0.10
iso-Butane 0.04
Pentane 0.63
iso-Pentane 0.007
Hexane 0.157
Benzol 0.0012
>C7 0.0803
Toluol 0.0013
H2 72 Permanent gases 75%
CO2 0.07
CO 0.69
N2 1.76
O2 0.19

The degradation rate of the cathodic losses (R2C) expectedly remains unaltered during the biomass-derived fuel gas operation (cf. Figure 14). The increased degradation rate of the resistances R2A, R3A can be attributed to damaging of the Ni/YSZ anode structure in the electrochemically active part of the ASC. As explained above, these damages are most likely a result of the formation of carbon whiskers and subsequent stepwise removal of the nickel phase, thus leading to a reduction in the number of active triple-phase points. The enhanced degradation rate observed for the gas diffusion part (R1A) can be explained by carbon deposits, leading to a blocking of the pores of the anode substrate. Furthermore, it cannot be ruled out that the Ni particles removed from the matrix may lead to an additional reduction of the total pore volume.

thumbnail image

Figure 14. Degradation behavior of the polarization resistances at the anode (R1A, R2A + R3A) during hydrogen and biomass-derived gas operation

The aforementioned separation of degradation into cathode and anode contributions of an ASC is only possible if the number, position, and physical origin of the polarization processes in the impedance spectrum are exactly known. The values of the polarization resistances discussed above were determined by means of a complex nonlinear least squares (CNLS) fit of the equivalent circuit depicted in Figure 15(a) to the impedance curves obtained from measurements. Figure 15(b) shows the CNLS adjustment between the model and the imaginary part of the measured impedance curve, exemplified for the case of the impedance measurement subsequently shown in Figure 16(a). The resulting relative errors (residuals) are plotted in Figure 15(c) on a logarithmic frequency scale. The high quality of this CNLS fit confirms the plausibility of the proposed equivalent circuit. The equivalent circuit model was developed by preidentification of the impedance response by calculating and analyzing the corresponding DRT and has been presented in Ref. 12. The higher resolution of the DRT renders possible the identification of losses with characteristic frequencies separated by only half a decade as demonstrated by the comparison of the impedance spectrum (Figure 16a) and the corresponding DRTs (Figure 16b) of an ASC. Unlike the impedance curve where the individual polarization processes overlap, at least five processes (P1C, P1A, P2C, P2A, and P3A) can be clearly distinguished in the calculated DRTs. The processes P1C, P2A, and P3A can be modeled by simple RQ elements and the cathodic process P2C is modeled by a Gerischer element. Particular attention should be paid to process P1A. In the equivalent circuit, this is modeled by a generalized finite length Warburg element (G-FLW). Characteristic of this process are its two peaks in the DRT at around 10 and 100 Hz.12 Depending on the operating conditions, this fact makes it difficult to identify the cathodic process P2C as this has a maximum in a frequency range similar to that of the smaller P1A peak, i.e., at around 100 Hz.

thumbnail image

Figure 15. (a) Equivalent circuit model used for the CNLS fit of impedance data. (b) CNLS fit of the imaginary part of the impedance spectra shown in Figure 16. (c) Residual pattern of the fit

thumbnail image

Figure 16. (a) Impedance spectra of an anode-supported cell with LSCF cathode recorded at T = 800 °C, inline image atm, inline image atm and (b) corresponding distribution function of relaxation times (DRT). Unlike the Nyquist plot, at least five processes are visible in the distribution curve

The physical origin of the individual polarization processes could be narrowed down by varying the operation parameters and analyzing the resulting DRTs. In order to analyze the dependency of each anodic process on the partial pressure of water inline image in the fuel gas, the H2O content was varied stepwise between 4.88 and 62.5%. Air was used as the cathode gas. Figure 17(a) shows the DRTs computed from the impedance spectra recorded at different inline image values. All three anodic losses (P1A, P2A, P3A) exhibit a more or less pronounced dependency on the water content in the fuel gas, process P3A showing the least effect. On the other hand, Figure 17(b) shows the influence of the oxygen partial pressure inline image at the cathode on the DRTs of the cell. Clearly, an increasing oxygen depletion gives rise to a peak (P1C) below 10 Hz, which is not observed for higher inline image values. Furthermore, a second peak appears in the frequency range between 100 and 10 Hz, overlapping with the two peaks caused by process P1A. These results confirm the assumption that the cathodic polarization losses can be completely described by the two processes P1C and P2C. At the same time, this analysis clearly indicates that the processes P1A, P2A, and P3A are ascribable to the anodic losses as they are not affected by the inline image variation.

thumbnail image

Figure 17. (a) Series of distribution curves at 4 different inline image, inline image atm (air), T = 757 °C. (b) Series of distribution curves at four different pinline image, pinline image atm, balance : H2, and T = 800 °C)

The processes P1A and P1C exhibit a very low thermal activation behavior (a negligibly small (even negative) activation energy), whereas P2C, P2A, and P3A are characterized by a pronounced thermal activation.

In summary, Leonide et al.12 conclude that P1A occurs owing to an inhibited gas diffusion within the Ni/YSZ anode substrate. The high-frequency processes P2A (1000 Hz) und P3A (10 kHz) are related to the charge transfer resistance during electrochemical oxidation of hydrogen (or the reformate) and the ionic transport in the Ni/YSZ anode structure. The two processes P1C and P2C are cathodic processes: P1C characterizes the gas diffusion losses within the pores of the cathode (and is, thus, negligible in air), whereas the faster process P2C is inherently electrochemical, accounting for the losses resulting from oxygen incorporation and oxygen ion transport within the cathode (LSCF).

4.2.2 Influence of Temperature

Since the cracking reaction is endothermic, this reaction is thermodynamically favored at higher temperatures. Therefore, with dry or slightly humidified hydrocarbons the amount of carbonaceous deposits that are formed by cracking increases with increasing temperature and hence cell degradation is more pronounced at higher temperatures under these conditions,37, 45, 47 Yamaji et al.37 have shown for SOFC operation with humidified ethane on a Ni/ScSZ anode that cell degradation could be significantly decreased at a temperature of 550 °C and a fuel utilization of 55% as compared to operation at 650 °C.

When the concentration of oxygen-containing compounds like steam, carbon monoxide, or carbon dioxide in the feed gas is increased, the temperature dependence of deposit formation is inverted and the amount of coke increases with decreasing temperature. This is shown in Figure 18 for a simulated reformate from liquid hydrocarbons.13 The coke yield, which is plotted in Figure 18, is defined as the molar amount of deposited carbon divided by the molar amount of carbon contained in the feed compounds during the operation time. The molar amount of the deposited carbon was determined by gasification with steam, as described in Ref. 13. In both cases, in thermodynamic equilibrium and in the measurements, the coke yield increases with decreasing temperature. At a temperature of 750 °C and above, no graphite is produced in thermodynamic equilibrium (Figure 18b). In the measurements, the limiting temperature for coke formation at an acetylene volume fraction of 0.1% is 850 °C. At a temperature of 700 °C and below, the coke yield was much lower in the experiments than in thermodynamic equilibrium. This indicates that the slow reaction rates noticeably inhibit the coke formation.

thumbnail image

Figure 18. Comparison of coke yields obtained in experiments and by equilibrium calculations with inline image%, yCO = 22%, inline image%, inline image%, inline image%, balance: N2 for an operation time of tc = 2 h. Components that were considered in the equilibrium calculations are C, CO, CO2, CH4, N2, H2O, and H2

The coke amount significantly influences the cell degradation. For the acetylene-containing reformate described above, Table 4 summarizes the duration after which severe degradation (when the degradation rate is ΔUt > 1 mV h−1) begins in the temperature range from 650 to 850 °C. Over a period of 8 h, degradation only occurs at a temperature as low as 650 °C. Since this temperature appeared to be most critical for coke formation in the investigated temperature range, the operation time was extended up to 50 h, but at lower acetylene concentrations. By reducing the acetylene concentration to one half of its initial value, the period until severe degradation appears could be extended by a factor of 3. When no acetylene is present in the applied reformate gas mixture, the cell does not degrade over 50 h of operation.

Table 4. Electrochemical Degradation During Operation with Acetylene-Containing Reformate with vCO = 22%, inline image%, inline image%, inline image%, Balance: N2
TemperatureVolume fraction inline image (%)Duration of measurementTime until severe degradation begins
850 °C0.05%8 h
 0.10%8 h
 0.25%8 h
750 °C0.05%8 h
 0.10%8 h
 0.25%8 h
650 °C0.05%8 h
 0.10%8 h4 h
 

0%

50 h
 0.05%24 h12.8 h
4.2.3 Influence of Gas Composition

The results discussed above have shown that coking on Ni anodes must be avoided because whisker-like coke can be formed at temperatures above 450 °C, resulting in an irreversible damage of the anode microstructure. In the following, the influence of gas composition, explicitly the concentration of compounds with high molecular oxygen content like steam, carbon dioxide, or air, is discussed. Saunders et al.51 found for i-octane by temperature-programmed oxidation (TPO) measurements that at a ratio of S/C = 3 the amount of carbonaceous deposits can be reduced to 1% of the value that was obtained with dry i-octane. In Ref. 13 the gas composition was varied for a simulated reformate from the partial oxidation (POX) of liquid hydrocarbons at 650 °C (Figures 19, 20). Acetylene has the highest influence on the coke mass. H2O and CO2 reduce the coke formation linearly, with H2O being twice as effective as CO2. The increase in coke mass mc with increasing H2 content was explained by Grabke et al.58, 59 Hydrogen was found to react with adsorbed oxygen on a metal (Fe, Ni) surface. If more hydrogen is present, less of the Ni surface is covered with adsorbed oxygen, and coking (by dissociative adsorption of CO) can proceed faster in this case. As can be seen from Figure 19, coking cannot be totally avoided under the conditions investigated. In order to avoid coking totally in the presence of acetylene with a mole fraction of inline image% or higher, the temperature has to be increased (as shown above).

thumbnail image

Figure 19. Effect of CO2, H2O, and C2H2 mole fractions on coke formation at 650 °C. Gas composition at normalized coke mass inline image%, yCO = 22%, inline image%, inline image%, inline image%, inline image%, coking time: 0.5 h

thumbnail image

Figure 20. Effect of CO and H2 mole fractions on coke formation at 650 °C. Gas composition at normalized coke mass inline image%, yCO = 22%, inline image%, inline image%, inline image, inline image%, coking time: 0.5 h

4.3 Substitute Compounds for Ni/YSZ Anodes

The major disadvantage of Ni-based anodes in combination with hydrocarbon fuels is their susceptibility to the formation of carbon whiskers, which can cause serious damage in the anode. This problem can be avoided by an increase of the gasification reaction of coke with steam, hence preventing an accumulation of C deposits. Compared to Ni, materials like ceria, CaO, and MgO promote the adsorption of steam.47, 55 When these materials are added to the anode, a decrease in coke formation, as compared to a pure Ni/YSZ anode, or even a complete prevention have been reported.35, 42, 53 A similar effect could also be achieved through the addition of precious metals (Pt, Ru), either by a finely dispersed impregnation of the Ni/YSZ anode35 or by application of a Ru-containing catalyst layer on the anode surface.52

However, a worthwhile alternative would be the complete replacement of nickel by other catalyst materials that prove to be resistant to carburization. Several investigations show that stable operation with hydrocarbon fuels can be achieved over several hours using Cu-based anodes with YSZ or ceria-based materials as ceramic phase.41, 46, 47, 49 However, in the long term, coke that is formed in the gas phase can fill the pores of the anode and block the fuel supply to the electrochemically active sites. Therefore, even if anode materials like Cu are used, which do not catalyze coke formation, the thermodynamic limits for carbon formation have to be considered.

5 Redox Durability

  1. Top of page
  2. Introduction
  3. Experimental
  4. Conversion of Hydrocarbon Fuels
  5. Carbon Formation and Degradation Effects
  6. Redox Durability
  7. Summary and Conclusions
  8. Acknowledgment
  9. References

5.1 Reduction and Oxidation Kinetics

The reduction of NiO is considered to be a two-step process.60-65 In the NiO lattice, nickel vacancies exist as Schottky defects V″Ni; to ensure charge compensation, either trivalent nickel cations or defect electrons (holes) are formed,66, 67

  • mathml alt image(16)

The reduction is initiated by chemisorption of H2 in a thin boundary layer (denoted by the subscript (b)) at the NiO surface:62

  • mathml alt image(17)

In the course of chemisorption, the adsorbed hydrogen species generate electrons within the layer:

  • mathml alt image(18)

This inversion leads to a p–n junction between boundary layer and the NiO inner phase. The subsequent induction step has a very small reaction rate63 and is determined by the deletion of Ni vacancies, which in turn gives rise to a critical amount of Ni seed elements at the surface of the NiO grain. The reactions

  • mathml alt image(19)

or

  • mathml alt image(20)

result in an annihilation of nickel vacancies until the chemical potential of nickel in the boundary layer reaches the same value as within the Ni seed. Only then can spontaneous formation of nickel seeds occur on the NiO surface. Their growth was observed with in situ transmission electron microscopy (TEM),68 and;69 the redox behavior of Ni (NiO) was recently studied by means of SEM/FIB (focused ion beam).70

Following the induction step, the reduction rate strongly increases on account of the autocatalytical effect of the Ni seeds.61-63 H+ is transported to the Ni/NiO interface via interstitials in the Ni lattice67 or via small pores.63 Simultaneously, oxygen is transported from the NiO lattice through the nickel layer to the surface where it reacts with gaseous hydrogen.60, 63, 71

The reduction of NiO–YSZ cermets is slowed down compared to the pure NiO phase.72 This slowed-down reduction of NiO in the presence of YSZ is presumably due to a reversed diffusion of oxygen ions out of the YSZ into the adjacent NiO phase, corresponding to an adjustment of the oxygen concentration of YSZ to ambient conditions. According to literature, nickel oxidation begins with the adsorption and dissociation of O2 at the nickel surface:66, 73

  • mathml alt image(21)

During the induction period, a thin continuous oxide layer grows. Generation of defect electrons gives rise to a space charge layer that determines the oxidation rate. At the NiO/gas interface, electrons are consumed in order to form O2− ions. At the same time, Ni vacancies (V″Ni) are transported through the NiO lattice to the Ni/NiO interface. As soon as the NiO layer reduces the space charge layer, further oxidation is diffusion controlled.66, 73, 67 On the one hand, nickel vacancies move through the NiO lattice toward the Ni/NiO interface; on the other hand, oxide ions are transported in the same direction of the NiO lattice, leading to an oxide growth both at the interface NiO/gas phase and at the Ni/NiO interface. The diffusion rate of V″Ni in NiO is, however, higher than that of the O2−. 73

The estimated growth rate at 800 °C is 0.25-µm film thickness of NiO per minute.74 Nickel oxidation at high temperatures is thus a potentially fast process. The activation energy has been determined as 144 kJ mol−1 and the rate constant kp at 800 °C has values in the range of 10−11 and 10−12 cm2 s−1.75-77 Wagner theory predicts that kp is dependent on the oxygen partial pressure inline image.78

Compared to reduction, oxidation proceeds slower owing to lower diffusion rates in NiO. As in the case of reduction kinetics, the oxidation kinetics can also be influenced by additional YSZ. Oxidation occurs faster in Ni–YSZ cermets than in pure Ni powders owing to the fact that by the presence of ion-conducting YSZ an additional transport path for oxygen ions is generated.

5.2 Reoxidation of Anode Substrates

In the course of sequential cyclic redox, both a sustainable electrochemical performance of the TPB electrolyte/Ni catalyst and the mechanical integrity of the thin film electrolyte in ASCs are of vital importance. A great deal of investigations have focused on mixed Ni–YSZ powders74-84 and on Ni–YSZ anode substrates, but there are only sparse studies referring to the electrochemical or mechanical behavior of complete cells (ESC/ASC) or the combination interconnects joining material cell.

Dilatometric measurements performed on anode substrates showed no length change during the first reduction of the precursor NiO to catalytically active Ni, although the reduction of nickel oxide to nickel brings about a decrease in the volume of 41%.79 This result can be attributed to the volumetric stability of the (NiO)–YSZ matrix, which is sintered at fairly high temperatures and is not affected by the first change from an oxygen-rich atmosphere to hydrogen (fuel gas). At cyclic redox, however, an irreversible increase in the length is observed, to an extent that depends on cermet composition and preparation. This length change is caused by the necessary restructuring of Ni to NiO2, 85 which leads to a volume increase of the substrate of nearly 70% during a complete reoxidation.

Fouquet et al.1 studied linear dimensional changes during repeated reoxidation (complete oxidation/reduction) by dilatometry as a function of precursor particle size ratio (NiO/YSZ) and sintering temperature (porosity). No length change during initial reduction, but accumulated expansion on repeated reoxidation and irreversibility on subsequent reduction was found. The least microcracking, corresponding to a linear expansion of 0.1% after two redox cycles (0.2% after three redox cycles), was determined for (i) a cermet structure made from precursors of 65 mol% NiO (particle size 0.5 µm) −35 mol% 8YSZ (particle size 0.2 µm) and (ii) the lowest sintering temperature (1200 °C) investigated. Hence, higher porosity (lower sintering temperature) of the anode substrate was assumed to be advantageous to accommodate Ni [RIGHTWARDS ARROW] NiO volume changes upon reoxidation. Corresponding results have been presented by Malzbender et al.86 Considering thermoelastic boundary conditions, higher porosity was proved to be beneficial for dimensional redox stability of Ni–YSZ substrates and anodes.

Recently, Sarantaridis et al.70 monitored strain in redox-cycled Ni–YSZ cermet substrates (produced by Forschungszentrum Jülich) as a function of degree of oxidation (DoO); this is defined as the ratio between the actual oxygen content adsorbed in the substrate and the oxygen content adsorbed at complete oxidation:

  • mathml alt image(22)

ranging from 0 to 100% at 800 °C. The commonly observed expansion on oxidation and its irreversibility on subsequent reduction were confirmed but were largely dependent on oxidation conditions. Subsequent interrupted oxidation resulted in a total linear expansion in the range of 0.55% (ambient air, no airflow), following a close-to-linear relation with DoO. It was clearly shown that the strain depends on the oxidation procedure giving values up to 0.8% (complete oxidation in flowing air). The oxidation expansion did not show a detectable effect on the integrity of the anode substrate, and redox cycling was accompanied by reversible changes in elastic modulus. It was concluded that oxidation expansion, as the critical parameter for cell failure, was closely related to microstructural changes of the Ni/NiO grains. By highly informative SEM/FIB investigations, it could be shown that formation of closed porosity in NiO (due to intrinsic features of the oxidation mechanism) is the controlling factor for the oxidation expansion.

5.3 Reoxidation of Half-Cells

In general, as shown by several studies,87-90 strain (and cell failure) increases with reoxidation temperature. Redox cycling of half-cells (Ni–YSZ anode substrate and thin film electrolyte (cofired)) leads to tensile stress in the thin film electrolyte during reoxidation of the anode substrate due to different expansion coefficients and different thicknesses of anode and electrolyte.

If the tensile strength is exceeded, cracks are formed and, ultimately, the thin film electrolyte cracks.2, 86, 89, 91-95 Stress, curvature, and cracking behavior are closely related to the DoO.86, 89, 91, 94, 96

Sintering procedure leaves the thin film electrolyte in a state of compression and the NiO–YSZ anode substrate in a state of tension reflecting the difference in thermal expansion coefficients.97 This leads to a curvature (1/r) of the half-cell; the elastic energy represented by the residual stress is conserved during the reduction process of NiO to Ni (relaxation/creep processes are negligible). As a result of the lower stiffness of the Ni/YSZ cermet compared to the NiO/YSZ structure, the curvature of the reduced half-cell is increased.

In Figure 21, the change of curvature (1/r) of a reduced half-cell (2700-µm anode substrate, 10-µm thin film electrolyte) during two subsequent redox cycles is depicted. As exemplified in Figure 22, the curvature 1/r changes its sign during a short transitional period corresponding to a change in stiffness of the cermet, and then takes on positive values again, closely related to the irreversible expansion of Ni during transformation into NiO. After the first cycle, the resulting curvature is larger than its initial value in the reduced state. The subsequent second reoxidation and reduction cycle further increases the curvature of the half-cell. This behavior has been explained by a model proposed in Ref. 98. The time profile of the reaction front, starting at the free surface of the anode substrate, continued through the bulk up to the thin film electrolyte, momentarily changes the sign of the curvature during oxidation (Ni [RIGHTWARDS ARROW] NiO), whereas a short curvature increase is measured at the beginning of each reduction (NiO [RIGHTWARDS ARROW] Ni) (t = 300 min). Figure 23 shows an oxidation front as it occurs during the transitional period from the reduced to the oxidized state in an anode substrate. In the lower part the anode consists of fully reduced Ni in the Ni/YSZ structure (Ni grains appear as bright spots), whereas in the upper part it is composed of NiO/8YSZ (no contrast between NiO and 8YSZ, pores appear as black dots).

thumbnail image

Figure 21. Curvature 1/r at reoxidation and at re-reduction (half- cell 0.27 mm anode,10-µm thin film electrolyte at 800 °C)

thumbnail image

Figure 22. schematic course of curvature 1/r at reoxidation and re-reduction during two redox cycles of a half-cell (see Figure 21)

thumbnail image

Figure 23. SEM image (inlens detector) of a partially reoxidized anode substrate (cross-fractured surface). Upper part is NiO/8YSZ, lower part is Ni/8YSZ; the reoxidation front is clearly visible as a border line between Ni (bright spots) and reoxidized NiO

The transition area with its Ni/NiO core shell structure70 is shown in a high resolution (SEM) image (Figure 24). The bright smooth regions represent the fully reduced Ni (grain size approximately 1 µm) on which smaller NiO grains (approximately 50 nm) have formed. The 8YSZ grains and the pores appear as dark gray and black areas respectively.

thumbnail image

Figure 24. SEM image (inlens detector) of a partially reoxidized anode substrate (cross-fractured surface).taken from the borderline between Ni and reoxidized NiO. Large Ni grains (size 1 µm) are coated by small NiO grains (size 50 nm)

This process does, however, not account for the resulting final curvature that is additionally increased during each further reduction/reoxidation cycle conducted. Presumably, this behavior is related to a decrease of the anode support stiffness resulting from microcracks observed in the anode substrate.86, 98

The formation of cracks in the thin film electrolyte is the dominant failure mechanism, once the tensile residual stress exceeds the tensile strength of the electrolyte.

In Figure 25, SEM images show the observed microstructural changes of the anode-supported half-cell during the subsequent redox cycles given in Figure 21.99 The micrographs were taken from exactly the same location ranging from the cofired state to the status after second reoxidation.

thumbnail image

Figure 25. SEM study of microstructural changes in anode substrate and thin film electrolyte as a consequence of sequential cyclic reduction and oxidation (redox): (a) initial (cofired state), (b) reduced, (c) reoxidized, (d) re-reduced, and (e) re-re-oxidized

The initially dense NiO particles (Figure 25a) shrink upon reduction to Ni (Figure 25b) and increase the porosity of the substrate. After reoxidation (Figure 25c), the NiO appearance changes and occupies a larger volume compared to the initial stage (estimated: 30%). Furthermore, electrolyte fracture and microcracks in the anode cermet are clearly visible. A second redox cycle strengthens all microscopic effects; macroscopically the crack density both in the electrolyte and in the substrate had increased. TEM investigations revealed a substantial porosity in the reoxidized NiO particles,99 which is in good agreement with the findings in Ref. 70.

The residual stress (determined at room temperature by X-ray diffraction (XRD100) in the thin film electrolyte (Ni/8YSZ anode 1500 µm) was determined as 560 MPa in the initial oxidized and as 520 MPa in the initial reduced state, resulting in a large compressive stress. Sequential reoxidation and re-reduction cycles reduce the residual stress by 60% (reoxidized state) and by 30% (reduced state) as reported in Ref. 101. In Figure 26, some of our own results are presented: the residual stress was determined depending on substrate thickness (1000 and 1500 µm), DoO (up to 20%), and temperature (600 and 800 °C, respectively).

thumbnail image

Figure 26. Relationship between residual stress and degree of reoxidation (XRD measurement) for temperatures of 600 °C (a) and 800 °C (b). Dots: 1.0-mm substrate, squares: 1.5-mm substrate

The residual stress in the reoxidized state is in fact smaller than in (or equal to) the reduced state, and the substrate curvature has increased. A decrease of the residual stress in the reoxidized state can effectively only be explained by creep or microcracks within the anode substrate (cf. Figure 25).

5.3.1 Degree of Oxidation

Malzbender et al.86, 94 monitored the typical course of the half-cell (Ni–YSZ anode substrate thickness of 1500 µm coated with an anode functional layer of 7 µm and a thin film electrolyte of 10 µm) curvature during reoxidation in air at 800 °C as a function of DoO (from 8 to 38%), see Figure 27.

thumbnail image

Figure 27. Warpage of half-cells (anode substrate 1500 µm) during reoxidation at 800 °C for 15 min; percentage of DoO ranging from 8 to 38% as indicated in the single photographs (for a better visibility of the warpage the samples are placed on a mirror with the thin-film electrolyte on the bottom side). [Reproduced from Ref. 94. © Wiley VCH, 2007.]

The resulting curvature is directly related to the DoO; furthermore, measurements at 600 and at 800 °C showed a clear dependence on the reoxidation temperature. Reoxidation at 600 °C resulted in (i) NiO particles with a lower inner porosity, (ii) a more homogeneous formation of NiO in the entire volume of the anode (from the substrate surface to the functional layer), and hence, and (iii) a lower curvature of the anode substrate.

In Figure 28, the relationship between DoO, airflow rate, and reoxidation temperature is displayed for a reoxidation time of t = 15 min. At 800 °C, a maximum DoO value of 74% is obtained; first cracks in the electrolyte appear at a DoO of 21%; and in contrast, at 600 °C a maximum DoO of 51% is reached and the electrolyte remains intact. For half-cells with lower substrate thicknesses and a similar Ni/YSZ microstructure, the electrolyte remains intact even at a DoO of 100% (reoxidation temperature 600 °C, substrate thicknesses 1000 and 500 µm) because the smaller anode volume is oxidized more homogeneously and the curvature is reduced owing to the more favorable ratio of anode substrate and electrolyte thickness. In the case of higher reoxidation temperatures, the porosity of the anode substrate should additionally be decreased so as to prevent the formation of cracks in the electrolyte. (However, for high fuel utilization, a smaller porosity of the anode substrate leads to gas diffusion polarization, which lessens the performance.)

thumbnail image

Figure 28. Degree of oxidation as a function of reoxidation temperature (600 and 800 °C) and airflow rate (oxidation time t = 15 min), half-cells (anode substrate 1500 µm)

In Figure 29, the reoxidation behavior of a substrate with 500 µm thickness and varying degree of porosity is shown at 800 °C. For an oxidation time of t = 15 min, a DoO value of 100% is obtained for a substrate with high porosity (the electrolyte features cracks); at a lower porosity, however, the DoO only reaches a value of 25% (no cracks in the electrolyte). This leads to the conclusion that a homogeneous reoxidation in the entire anode volume plays a decisive part in the intactness of the electrolyte. Depending on the reoxidation conditions (gas flow, gas composition, temperature, and duration), a combination of substrate thickness and porosity can be found, which prevents the formation of cracks in the electrolyte.

thumbnail image

Figure 29. Degree of oxidation as a function of anode substrate porosity and airflow rate (oxidation time t = 15 min), half-cells (anode substrate 500 µm)

In practice, it must be considered, though, that reoxidation of the anode substrate within a stack occurs under different conditions. In this case, a failure of the thin film electrolyte or a fracture of the anode substrate may already take place at lower DoO values. This is confirmed by preliminary findings on half-cells (anode substrate 1500 µm) with a metallic interconnector (Crofer22APU). Reoxidation at 800 °C and constraint bending resulted in macroscopic cracks in the electrolyte or the anode substrate at a DoO value of only 10% (as compared to values of 25% in the case of free bending). Cell cracking starts at comparatively low DoO values as a result of (i) minimized or constrained elongation in the x − y direction and/or (ii) constrained cell warping, and/or (iii) inhomogeneities of reoxidation, the latter also caused by the gas distribution within the interconnects.

5.4 Cell Degradation Effects by Sequential Redox Cycling

The interactions between sequential cyclic redox and electrochemical performance were so far only investigated in a few measurements carried out on whole cells (ASC, ESC). Cassidy et al.2 observed a drop in the cell voltage (OCV decreases from 0.85 to 0.4 V) in redox experiments, which they attributed to a failure of the thin film electrolyte. In Ref. 96, the electrochemical performance upon a complete reoxidation of the anode is at least partly retained at 750 °C. The cell performance after reoxidation under various measuring conditions (DoO, number of cycles, anode layer composition, with or without a gas diffusion blocking layer at the anode substrate surface) was analyzed in Refs. 95, 102. In Figure 30, ESCs (electrolyte 8YSZ 150 µm, Ni/8YSZ anode 30 µm) and ASCs (electrolyte 8YSZ 10 µm, Ni/8YSZ anode substrate 1500 µm) underwent a total of 100 redox cycles. The open cell voltage U0 and the output power P at a cell voltage of 0.7 V were measured before and after each redox cycle (ASC at 600, 700, and 800 °C; ESC only at 800 °C). The reoxidation time was augmented from t = 1 min (cycle no. 1–10 min (cycle no. 51–100).

thumbnail image

Figure 30. Open cell voltage Uo and power output P (0.7 V) as a function of redox cycles (redox cycle no. 1 to 50: t = 1 min airflow, redox cycle no. 51 to 100: t = 10 min airflow); ESCs (a) at 800 °C and ASCs (b) at 600, 700, and 800 °C

A comparison of both cell types yields the following conclusions:

  • ESC: the open cell voltage remains unaltered by redox cycling and the electrolyte is not damaged (as expected). The output power at 800 °C decreases (after cycle no. 5) from its initial value of 300 mW cm−2 down to around 100 mW cm−2 (cycle no. 80) and does not change thereafter.

  • ASC: the open cell voltage remains constant during the 1-min oxidation cycles, i.e., the thin film electrolyte stays intact. During the 10-min oxidation cycles, an influence of the temperature is observed as expected from the DoO. At 800 °C the electrolyte immediately breaks down. The decrease (after 50 cycles at a cycle duration of t = 1 min) determined in the significantly larger power of the ASC depends on the reoxidation temperature: At 800 °C, the power strongly decreases (from 600 to 180 mW cm−2); at 700 °C, there is only a slight decrease (from 400 to 330 mW cm−2); and at 600 °C, no change is detectable (100 mW cm−2). At the cycles with a duration of t = 10 min, there is a change in the anode functional layer also at 700 and 600 °C, but the electrochemical performance is retained at a lower level.

Even after several redox cycles, the ASC shows a conspicuously higher performance than the ESC, especially at lower temperatures. For practical applications (stack) the respective boundary conditions (quantity and distribution of gas, stress distribution) are crucial for an integrity or failure of the thin film electrolyte in an ASC.

6 Summary and Conclusions

  1. Top of page
  2. Introduction
  3. Experimental
  4. Conversion of Hydrocarbon Fuels
  5. Carbon Formation and Degradation Effects
  6. Redox Durability
  7. Summary and Conclusions
  8. Acknowledgment
  9. References

State-of-the-art Ni/YSZ composites applied in ASCs for stationary and mobile SOFC systems were assessed in the following aspects:

  1. Methane Reforming Kinetics

    The methane reforming reaction rate on a Ni/YSZ anode was calculated by kinetic modeling. The reaction order is 1 for methane and 0 for steam. Over a wide S/C ratio (from 1 to 6), the presence of steam is insignificant for the kinetics, but it noticeably influences the position of equilibrium and the driving force of the chemical reaction. The apparent activation energies of 61 kJ mol−1 (600–750 °C) and of 30 kJ mol−1 (750–850 °C) are considerably low, because of the influence of diffusion limitations on the reaction rate in the anode substrate (thickness 1000 µm) investigated. This result underlines the advantage of high porosity substrates for technical application at temperatures above 750 °C.

  2. Carbon Formation in Presence of Hydrocarbons

    System operation with reformate gas (purified of hydrocarbons) does not influence the electrochemical performance compared to a fuel gas composition like H2/H2O. Irreversible structural changes within the Ni/YSZ anode are caused by the presence of hydrocarbons like acetylene. The formation of carbon whiskers is initiated by decomposition of the hydrocarbons on the Ni catalyst surface. The accumulation and growth or back-formation is determined by the reaction rates of the hydrocarbon decomposition and subsequent gasification with steam. With higher hydrocarbons, such as acetylene, the reaction rate of decomposition can proceed faster than the gasification rate, thus promoting the formation of carbon whiskers. The rate of carbon formation and deposition is strongly dependent on temperature. At higher temperatures (and low oxygen partial pressure), the carbon formation rate is controlled by the cracking reaction. By increasing the oxygen content (adding H2O, CO, or CO2 to the gas), the temperature dependence of deposit formation is inverted and the rate of coke formation increases with decreasing temperature. Depending on the composition of the fuel gas, coke formation and resulting degradation of the cell performance can also be observed even if a carbon phase is thermodynamically not possible. This could be shown for an ASC operated with biogas derived from corn silage.

  3. Performance Degradation in Presence of Biogas

    Electrochemical impedance spectroscopy measurements followed by a distribution of relaxation times analysis furnished proof that enhanced degradation in biogas originated both from (i) an increase of gas diffusion polarization (due to carbon deposition in the pore volume) and (ii) an increase of charge transfer resistance during electrochemical oxidation (caused by formation of carbon whiskers with a subsequent reduction of active triple-phase points).

  4. Structural and Performance Degradation During Redox Cycling

    Oxidation expansion is closely related to microstructural changes of the Ni/NiO grains in Ni/YSZ anode substrates and is identified as the critical parameter for tensile stress in the thin film electrolyte followed by cell failure. Redox cycling results in a decrease of the residual stress, explained by creep and/or microcracks within the anode substrate. Stress, curvature, and cracking behavior of ASCs are closely related to the DoO and the temperature. The decrease of substrate thickness and operating temperature results in more homogeneous redox conditions, which seems to be the deciding factor. Depending on reoxidation conditions (gas flow, gas composition, temperature, duration), a combination of substrate thickness and porosity is conceivable, which prevents cell failure. A performance test of ASCs at 600, 700, and 800 °C in combination with 100 redox cycles (reoxidation time was augmented from t = 1 to t = 10 min in air) confirmed that ASCs retain respectable electrochemical performance at 600 and 700 °C.

Acknowledgment

  1. Top of page
  2. Introduction
  3. Experimental
  4. Conversion of Hydrocarbon Fuels
  5. Carbon Formation and Degradation Effects
  6. Redox Durability
  7. Summary and Conclusions
  8. Acknowledgment
  9. References

The authors wish to express their thanks to Manuel Ettler from the Institute of Energy Research at the Forschungszentrum Jülich for his helpful contributions to the redox part of this manuscript. Dr Andre Weber and Dr Stefan Wagner from IWE are gratefully acknowledged for their invaluable and continuous support in preparing this manuscript.

References

  1. Top of page
  2. Introduction
  3. Experimental
  4. Conversion of Hydrocarbon Fuels
  5. Carbon Formation and Degradation Effects
  6. Redox Durability
  7. Summary and Conclusions
  8. Acknowledgment
  9. References