and constant cell voltage of 0.6V, a drop of 16.7% was observed within ..... K. W. Jacobsen, Surf. ... C. H. Bartholomew, P. K. Agrawal and J. R. Katzer, Adv. Catal.
Chapter 65 Sulfur poisoning on Ni catalyst and anodes J. Bøgild Hansen and J. Rostrup-Nielsen Haldor Topsøe A/S, Lyngby, Denmark
1 INTRODUCTION One of the big advantages of solid oxide fuel cells (SOFCs) power plants is the flexibility reflected by a wide variety of possible fuels ranging from hydrogen, natural gas or biogases, liquid hydrocarbons or oxygenates to coal-derived synthesis gases. However, many of these fuels contain sulfur in different forms and concentrations. For larger installations, it is normally advantageous to remove the sulfur thoroughly. This can be achieved by use of commercial catalysts used in the petrochemical industry or catalysts specially developed for SOFC applications, for which the pressure is low and pure hydrogen is not normally available. The desulfurization is mainly carried out in order to protect downstream fuel-processing catalysts used to condition the gas for optimum use in the SOFC stack. The fuel processing (FPS) catalysts are very often based on Ni. The SOFC stacks are then effectively protected against sulfur poisoning by the desulfurization catalysts and the fact that the FPS catalysts adsorb the last traces of sulfur. Malfunctions or upsets of the plant operation can, however, lead to breakthroughs of sulfur to the stacks. For some fuels, for instance diesel, it can be very difficult or uneconomical to carry out desulfurization. It is thus of interest to know the impact of sulfur on the most widely used anodes in SOFCs based on nickel, which, as shown both in practical experiments as well by theoretical work, remains the best available electrocatalyst amongst the metals.[1] There are also indications in literature that moderate contents of sulfur can be tolerated on nickel anodes operating on hydrogen. This would eliminate the need for desulfurization of gases with a low methane content. If the gas
contains appreciable amounts of methane, it is, however, still imperative to remove the sulfur quantitatively before the stacks, because the internal reforming reaction catalyzed by nickel is poisoned by sulfur.[2, 3] The development of sulfur-tolerant anodes – not based on nickel – has of course attracted a lot of attention, but this is outside the scope of this article. The direct use of hydrogen sulfide as a fuel for SOFCs has also been investigated and is even possible with Ni–YSZ anodes.[4]
2 THERMODYNAMICS 2.1 Gas-phase thermodynamics The most important gas-phase reactions that are of interest in fuel processing and anode operation in an SOFC plant are CO + H2 O = CO2 + H2
(1)
CH4 + H2 O = CO + 3H2
(2)
CO2 + H2 S = COS + H2 O
(3)
H2 + CH3 S = CH4 + H2 S
(4)
2H2 O + H2 S = SO2 + 3H2
(5)
H2 S + 1.5O2 = SO2 + H2 O
(6)
H2 + 0.5O2 = H2 O
(7)
The equilibrium constants are given in Table 1.
Handbook of Fuel Cells – Fundamentals, Technology and Applications. Edited by Wolf Vielstich, Harumi Yokokawa, Hubert A. Gasteiger. Volume 6: Advances in Electocatalysis, Materials, Diagnostics and Durability. 2009 John Wiley & Sons, Ltd. ISBN: 978-0-470-72311-1.
2
Part 5: Performance degradation Table 1. Equilibrium constants (standard state, 1 bar, 298 K). T (K) 773.15 823.15 873.15 923.15 973.15 1023.15 1073.15 1123.15 1173.15 1223.15 1273.15
2.2
1
2
3
4
5
6
7
5.00E+00 3.53E+00 2.61E+00 2.00E+00 1.58E+00 1.28E+00 1.06E+00 8.95E–01 7.69E–01 6.71E–01 5.92E–01
9.94E–03 8.15E–02 5.29E–01 2.82E+00 1.27E+01 4.97E+01 1.71E+02 5.30E+02 1.49E+03 3.87E+03 9.30E+03
1.39E–02 1.85E–02 2.36E–02 2.95E–02 3.59E–02 4.29E–02 5.03E–02 5.82E–02 6.65E–02 7.51E–02 8.39E–02
1.04E+05 4.96E+04 2.57E+04 1.43E+04 8.41E+03 5.20E+03 3.36E+03 2.25E+03 1.56E+03 1.11E+03 8.14E+02
2.77E–11 2.21E–10 1.41E–09 7.36E–09 3.27E–08 1.26E–07 4.33E–07 1.33E–06 3.74E–06 9.67E–06 2.33E–05
9.48E+30 7.00E+28 9.07E+26 1.88E+25 5.79E+23 2.51E+22 1.45E+21 1.09E+20 1.01E+19 1.14E+18 1.53E+17
6.99E+13 6.81E+12 8.64E+11 1.37E+11 2.61E+10 5.83E+09 1.50E+09 4.34E+08 1.39E+08 4.91E+07 1.87E+07
for example the anode is exposed to fuel gas during heating up or cooling down. The use of more active cathodes would also open up for the use of lower operating temperatures where the sulfur poisoning problems will become much more pronounced.
Bulk reactions
It is obviously important to know what the thermodynamically stable state is for the solid components of the catalyst and anodes under the relevant temperatures, pressures, and gas compositions. The most important reactions of relevance for this article are those for the Ni–O–S system. Lohsoontorn et al. have calculated the phase diagram (see Figure 1) for the system at different temperatures.[5] It is only in extreme cases that the sulfur concentration is so high that bulk transformation of nickel will take place at normal SOFC operating temperatures. As an example, Rosenqvist’s investigation of the thermodynamics of nickel sulfide phases[6] indicates that the formation of a bulk-phase sulfide (Ni3 S2 ) at temperatures around 500–700 ◦ C requires a ratio p(H2 S)/p(H2 ) of the order of 10−3 . This ratio is about 100–1000 times above what would normally cause poisoning at those temperatures, so it is very important to also consider the chemisorption of sulfur on the surfaces, which is discussed in the next section. As illustrated earlier, care should be exercised if lower than normal operating temperatures are considered, when
2.3 Chemisorption of sulfur on nickel Under reforming conditions, all sulfur compounds are converted to hydrogen sulfide, which is chemisorbed on the metal surface (M): H2 S + Me = Me–S + H2
(8)
This takes place at H2 S/H2 ratios far below those required for formation of bulk sulfides.[7] Nickel is the most sensitive metal. Hydrogen sulfide chemisorbs dissociatively on nickel. Stable saturation uptakes of sulfur are observed at values of p(H2 S)/p(H2 ) from 1–10 × 10−6 up to 100–1000 × 10−6 above which bulk sulfide is formed. Today, the structure
5 0
NiS2
1073 K Ni3S4
5 673 K
NiSO4
0
NiS2
Ni3S2
log pS2 (bar)
log pS2 (bar)
NiS −5
1000 ppm H2S
−10 −15
NiO
Ni 97% H , 3% H O 2
2
10% H2, 90% H2O 1 ppm H2S
−5
NiS0.84 Ni3S2
−10
NiSO4
1000 ppm H2S
−15 −20
−20 −25 −40 −35 −30 −25 −20 −15 −10 log pO2 (bar)
−5
NiO 97% H2, 3% H2O
10% H2, 90% H2O 1 ppm H S
2 Ni −25 −40 −35 −30 −25 −20 −15 −10 −5
0
0
log pO2 (bar)
Figure 1. Phase diagrams for the Ni–O–S system at 1073 and 673 K. [Reproduced from P. Lohsoontorn, D. J. L. Brett and N. P. Brandon, J. Power Sources, 175, 60 (2008).]
Sulfur poisoning on Ni catalyst and anodes 3
Ni(110)
1.1
p (5 × 2)
1.0 0.9
12 L H2S 2
θS
51 × 54 Å
H 2S : H 2 46 ppm 25 14 7 ppm
0.8 0.7 0.6
0.03 0.1 ppm 0.01 0.003 0.001 ppm
0.5
H2S + Nisurface
0.4 700
S–Nisurface + H2
800
900
1000
1100
1200
1300
Temperature (K)
Figure 2. Chemisorption of sulfur on nickel surface. structures observed by scanning tunneling microscosopy (STM) (Besenbacher et al.[8–10] ). [Reproduced with permission from the authors.]
Ni area from the H2 chemisorption (m2 g −1)
of the surface sulfide can be seen by scanning tunneling microscopy[8] as shown in Figure 2. The surface layer is built up by islands of nickel and sulfur atoms in a two-dimensional “crystalline” structure with fixed distances between the atoms. This is in contrast to the classical description with random distribution of sulfur atoms on the nickel surface. This was well described by early low energy electron diffraction (LEED) and Auger experiments[11, 12] and it was shown that the monolayer of sulfur (S/Ni = 0.5 – sulfur capacity = 440 ppm S/m2 Ni) correlated with the surface area determined by hydrogen chemisorption as illustrated in Figure 3. The saturation layer has been described earlier. Below a certain p(H2 S)/p(H2 ), the saturation layers become unstable and the equilibrium coverage is dependent on the p(H2 S)/p(H2 ) and temperature. This coverage can be described in terms of an adsorption isotherm and the isosteric heat of chemisorption. A number of recent investigations have found the isosteric heat of adsorption to be approximately 155 kJ mol−1 ,
20 18 16 14
Figure 4. Isobars for chemisorption of hydrogen sulfide on nickel catalysts. [Reproduced from Ref. 13. Elsevier, 1981.]
decreasing slowly at high coverages. This value should be compared to the heat of formation for bulk sulfide (Ni3 , S2 ) of approximately 75 kJ mol−1 of S. This demonstrates that the adsorbed sulfur is strongly bound to the surface of nickel and that the “two-dimensional sulfide” can be stable at conditions where bulk sulfide does not exist. Attempts to correlate the data with a Langmuir-like isotherm have not been successful. This failure is not surprising in view of the mechanism of the adsorption, violating the assumptions for the Langmuir isotherm. However, the following Temkin-like isotherm proposed by Alstrup et al.[13] has been used to correlate data obtained under very different conditions as shown in Figure 4: Y = exp H00 (1 − aθS )/RT − S 0 /R
Equation (9) implies that the entropy of adsorption is independent of θS , which is in accordance with a “bulk”like behavior of the chemisorbed layer (i.e., a “twodimensional” sulfide). The isobars in Figure 4 show straight lines in the θS versus T plot. The deviations at low temperatures may be explained by adsorption on the support.[13] The plotted lines correspond to the constants – H00 = 280 kJ mol−1 , S 0 = −19 J mol−1 K−1 , and a = 0.69 – which result in the equation (10) θS = 1.45 − 9.53 × 10−5 T + 4.17 × 10−5 T p(H2 S) ln p(H2 )
12 10 8 6 4 2 0 0
5
10
15
20
25
Ni area from the sulfur capacity (m2 g−1)
Figure 3. Nickel surface areas of various catalysts calculated from hydrogen chemisorption data compared with the nickel surface areas determined by sulfur chemisorption. The data are fitted to a straight line through the origin. The slope of the line is 0.98.
(9)
(10)
This expression is not valid for θS close to zero and close to 1.0. θS = 0.5 at 500 ◦ C corresponds to H2 S/H2 = 1.6 × 10−12 . It is evident that a nickel-based prereformer will remove any trace of sulfur. In practice, it would take a very long time to arrive at the equilibrium coverage. Wang and Liu[14] have carried out density function theory (DFT) calculations with the appropriate thermodynamic corrections for the gas-phase species both for the chemisorption equilibrium and the bulk reaction. Two
4
Part 5: Performance degradation
100
1800
3 SULFUR POISONING OF REFORMING REACTIONS
H2S concentration (ppm) 10 1 0.1
1400 1200 800
1000
650
800 600
0.72 0.63 0.75 0.92 0.98 1.12
1.21 1.35
Temperature (°C)
Temperature (K)
1600
400 3
4
5
6 7 8 log (PH /PH S) 2
9
10
11
2
Figure 5. Phase diagram for H2 S and Ni. [Reproduced from Ref. 14. Elsevier, 2007.]
nickel surfaces, Ni(100) and Ni(111) have been considered in the calculations and four different active sites for the chemisorption. A 32 Ni super cell has been used for the bulk calculations. For the surfaces, four-layer Ni(100) and Ni(111) slabs with 32 and 48 Ni atoms, respectively, have been taken into consideration followed by 1-nm vacuum space. Galea et al. has carried out similar calculations.[15] Based on these calculations, a phase diagram (see Figure 5) for both the bulk and chemisorption equilibria between H2 S and Ni has been constructed. The white region represents clean Ni, the blue region adsorbed S on Ni, and the yellow bulk Ni3 S2 . The black line represents the data of Rosenqvist.[6] The red line is from the DFT calculations of Wang.[14] The black triangles are the chemisorption results at different surface coverages and, finally, the red dots are from the SOFC experiments of Matsuzaki and Yasuda.[16] Wang and Liu calculated the variation of adsorption energy as a function of surface coverage, which varies from −2.64 eV at 1/16θ to −1.40 eV at 10/16θ. The effect of an electrical field near the Ni surface, as may be present near the electrolyte in a working SOFC, has also been calculated. An external field of 5 V nm−1 would induce a change in the adsorption energy of less than 0.20 eV corresponding to a shift in the adsorption phase boundaries in Figure 5 by less than 50 K. There have been several attempts to characterize the surface Ni–S species and one of the most promising techniques are Raman spectroscopy + LEED/Auger.[17–22] The studies should, however, be very carefully executed and preferably be done in situ as Ni3 S2 will undergo phase transformation if the sample is cooled down. Moreover, it can very easily be oxidized during handling.[20]
Group VIII metals are susceptible to sulfur poisoning[23] and sulfur must be removed from the feed stream. Natural gas may be cleaned over zinc oxide (sometimes promoted with Cu) or active carbon. Liquid hydrocarbons require hydrodesulfurization over CoMo catalyst. Poisoning effects are often correlated with the poison concentration in the feed stream, which, of course, is the important parameter in practical operation; however, in a more detailed analysis this approach can hardly be justified other than for isothermal tests in gradientless reactors. The adsorption equilibrium depends on the temperature and the composition of the gas phase, which varies through the reactor as well as within the single catalyst pellet. Therefore, it appears more rational to correlate the deactivation with the amount of poison present on the catalyst rather than with the poison concentration in the feed stream. However, the correlation between sulfur in the feed and in the catalyst may be complex as illustrated above.
3.1 Regeneration In principle, it should be expected that reaction (8) is influenced by the competing chemisorption reaction ∗ + H2 O = ∗ – O + H2
(11)
However, it was shown that the sulfur uptake was not affected by the presence of steam in the fuel up to the H2 O/H2 ratio of 10. These observations were confirmed by STM experiments illustrating how H2 S dispenses adsorbed oxygen.[8] This result may be surprising as the heat of chemisorption of oxygen is high (ca 430 kJ mol−1 of O2 ). However, this value is less than the heat of chemisorption for sulfur, which 0 is about 480 kJ mol−1 of S2 (as calculated from Hads for H2 S). The chemisorption process is reversible and, in principle, sulfur should be removed in a hydrogen stream. However, the driving force is very small and decreases with decreasing θS , and the elution process is subject to diffusion restrictions. Sulfur may be removed by oxidation under controlled conditions. No improvement is observed as expected from the chemisorption experiments when regenerating reforming catalysts with steam with a slight increase in the pH2 O /pH2 ratio., However, at pH2 O /pH2 above 150–250, a significant change of the degree of regeneration of the catalyst is obtained. Increased sulfur removal is achieved at a
Sulfur poisoning on Ni catalyst and anodes 5
Saturation 0.3 ppm
1
0.15
0.6
0.075
youtlet /yinlet
0.5
0.030
0.4
0.015 0.0075 ppm
0.2
0 0
1000 2000 3000 4000 Distance along the shown path in the pellet (µm)
Figure 6. Distribution of sulfur in a catalyst pellet. The profile indicates that the gas flows equally well through the seven holes in the pellet and around the pellet. [Reproduced from Ref. 24. Elsevier, 2006.]
pH2 O /pH2 , which is close to the equilibrium constant for the oxidation of the catalyst. The exit gas shows the presence of sulfur dioxide as well as hydrogen sulfide, which may indicate the following reaction patterns: Ni − S + H2 O ←−→ NiO + H2 S
(12)
H2 S + 2H2 O ←−→ SO2 + 3H2
(13)
The chemisorption of hydrogen sulfide on nickel is very rapid with a sticking coefficient of close to 1.0 for less than 70% of a full monolayer. The sulfur uptake is strongly limited by diffusion[24, 25] as illustrated by the sulfur profiles in Figure 6. The pore diffusion restrictions in the sulfur adsorption in a single pellet have a complex influence on the transient sulfur profiles in the reactor; a mathematical model is required to evaluate the time for full saturation and the breakthrough curves of sulfur more exactly. Figure 7 shows calculated breakthrough curves in a steam reformer of a naphtha-based ammonia plant. The relative concentration of hydrogen sulfide at the tube exit is plotted as a function of time. It is seen that the breakthrough occurs immediately for sulfur contents higher than 0.1 wt% ppm in the naphtha, indicating the serious effect of a sulfur peak in the reformer feed on a downstream catalyst.
3.2 Impact of sulfur on reforming reactions It was shown that the reforming rate depended on the sulfur coverage to the third power (1 − θS )3 as illustrated in Figure 8.[26] Steam reforming is possible at high temperatures with high coverages of sulfur, but it requires high tube-wall temperatures in fired reformers, and the risk of carbon formation from cracking of higher hydrocarbons has to be taken into account.
0
2000
4000 6000 Time (h)
8000
10 000
Figure 7. Breakthrough curves for hydrogen sulfide. [Reproduced from J. R. Rostrup-Nielsen, Progress in Catalyst Deactivation, 219 (1982), with kind permission of Springer Science and Business Media.]
Sulfur-free test Sulfur passivated catalyst
10 000
Total Ni area (mol/(m2 Ni)/h
S content, (arb unit)
0.8
r sp /(1 − θs)3 1000
r 0sp 100
r sp /(1 − θs) 10
900 1 0.8
r sp 700 1.0
500 1.2
400 °C 1.4 103/T (K−1)
Figure 8. Sulfur passivation and specific reforming activity at 1 bar. Ni/Al2 O3 catalyst. Sulfur-passivated catalyst: H2 O/CH4 = 1, H2 O/H2 = 5, H2 S/H2 = 2.8 × 10−5 ; sulfur-free o = specific activity catalyst: H2 O/CH4 = 0.94, H2 O/H2 = 2.5; rsp of nonpoisoned catalyst; rsp = specific activity reformed to total nickel area. [Reproduced from J. R. Rostrup-Nielsen, J. Catalysis 85, 31 (1984).]
Carbon may be formed by three different mechanisms: “whisker” carbon, gum formation, and pyrolytic carbon. Whisker carbon is formed by dissociation of hydrocarbons or carbon monoxide on the nickel surface. The nucleation is blocked by sulfur poisoning. The higher hydrocarbons may lead to carbon formation by all three mechanisms. Thermodynamics predicts carbon formation as long as the higher hydrocarbons are present. At low temperatures, adsorbed hydrocarbons may accumulate on the nickel surface and slowly be transformed into a polymer film (“gum”) blocking the nickel surface.
6
Part 5: Performance degradation microreactor. Different characterization techniques were also applied on the fresh and used samples. The conclusions were that the reactions involving water (steam reforming and water–gas shift) were more strongly affected by sulfur than were the CO or the CH4 oxidations. The reverse water–gas shift was not affected either. The YSZ showed no catalytic activity for the steam methane reforming reaction and only minor activity for the shift or methane oxidation reactions. The latter two reactions were not affected by sulfur on YSZ.
530 520
Temperature (°C)
510 500 490 Diesel – 37 h Diesel – 304 h Jet fuel – 38 h Jet fuel – 303 h
480 470 460 0
0.5
1.0
Relative axial distance
Figure 9. Adiabatic prereforming of “logistic fuels”. Temperature profiles with Topsoe RKNGR catalyst. [Reproduced from J. R. Rostrup-Nielsen, I. Dybkjær and T. S. Christensen, Stud. Surf. Sci. Catal., 113, 2 (1998).]
Figure 9 shows an example from a prereformer operating on heavy feedstock.[27] With kerosene (jet fuel), the catalyst is gradually poisoned by sulfur, resulting in a shifting of the temperature profile. When the process is operated with gas oil (diesel), the sulfur poisoning is accompanied by poisoning resulting from “gum” formation. At high temperatures, ethylene from the pyrolysis of higher hydrocarbons may lead to pyrolytic coke, which may encapsulate the catalyst pellets. The impacts of sulfur on catalytic partial oxidation (CPO) catalysts are more benign. In the presence of oxygen, sulfur is oxidized to SO2 , but nickel can also be oxidized, which is not the case for rhodium, a typical catalyst for CPO. It means that rhodium remains active as long as oxygen is present. After depletion of oxygen, SO2 is reduced to H2 S, which is then chemisorbed on the catalyst and the downstream catalysts and anodes. The task is then reduced to remove hydrogen sulfide from the product gas. This sequence is reflected by CPO experiments. In presence of sulfur in the feedgas, the reaction temperature increases, because the endothermic steam-reforming reaction, which normally proceeds after depletion of oxygen, is poisoned. The sulfur resistance of the CPO process is, in particular, an advantage when using heavy fuels (diesel) as feed. These fuels contain sulfur compounds that are difficult to remove at low pressures.
4
IMPACT OF SULFUR ON OTHER ANODE REACTIONS
Kuhn et al.[28] carried out a series of temperature programmed reaction studies on Ni–YSZ, sulfided Ni–YSZ, and yttria stabilized zirconia (YSZ) alone in an isothermal
5 IMPACT OF SULFUR ON ELECTROCHEMICAL REACTIONS There is general consensus that • the deactivation by H2 S increases by lowering temperature; • the effect levels off as the H2 S concentration increases; and • the impact of sulfur is less at a high current density. There is also general consensus that sulfur has an immediate impact on the electrochemical performance of Ni anodes but there is, however, no agreement on the degree of recovery achievable after sulfur is removed from the feed stream again. The long-term effects are not known. Ideally, the impact of sulfur should be reported as the reduction in the electrochemical activity of the anode itself, which is normally manifested in an increase only in the polarization resistance. The deactivation of the anode, however, obviously also changes the operating conditions of the cathode and for complete stacks, the temperature profile will be affected and will have to be taken into account. Impedance spectroscopy can unravel some of these issues and has been widely employed.
5.1 H2 /N2 mixtures One of the first detailed investigations of Ni–YSZ susceptibility to sulfur as a function of temperature and concentration was reported by Matsuzaki and Yasuda from Tokyo Gas.[16] They varied the H2 S concentration from 0.02 to 15 ppm and used temperatures of 1023, 1173, and 1273 K. The polarization resistance and the overvoltage started to increase when the H2 S concentration exceeded 0.05, 0.5, and 2 ppm at 1023, 1173, and 1273 K, respectively. The time to attain steady state was found to be almost independent of the sulfur concentration and only to depend on temperature. The anode recovered completely the original performance when pure fuel was used again. Measurements were also performed keeping the concentration of H2 S constant with varying steam partial pressures or changing the
Sulfur poisoning on Ni catalyst and anodes 7 24
24 22
Cell performance drop (%)
22
50 ppm 800 °C, 0.8 V
20
20 18
18 700 °C, 0.7 V
16
16 14
14 800 °C, 0.7 V
12 10
900 °C, 0.7 V
8
50 ppm
12
800 °C, 0.6 V
10 8
6
6
4
4
2
2
0 0
1
2
3
4 5 6 7 8 9 10 49 H2S concentration in H2 (ppm)
50
0 51
Figure 10. Percentage drop in current density as a function of H2 S concentration and temperature. [S. Zha, Z. Cheng, and M. Liu, J Electrochem Soc 154, B201(2007). Reproduced by permission of ECS - The Electrochemical Society.]
1.5% H2 O, 48.5% N2 can then be used to calculate θS from equation (10). In Figure 11, the cell performance drop in percentage is plotted versus θS . A very good correlation of the performance loss in percentage, Pl, of the cells is obtained by equation (14): P l = k ∗ (θS − θmin ) = 53.8θS − 32.2
(14)
with R 2 = 0.985. Considering the wide span in temperatures and hydrogen sulfide concentration, this is convincing evidence that the performance drop is strongly related to sulfur coverage of the nickel in the anode. In the experiments of Matsuzaki and Yasuda,[16] the polarization resistance and the overvoltage started to increase when the H2 S concentration exceeded 0.05, 0.5, and 2 ppm
Cell performance drop (%)
concentration of H2 S with constant steam content. It was concluded that the deterioration depended on the concentration of H2 S and not on the partial pressure of S2 , which, at equilibrium, would vary with the steam content. From a very comprehensive study, Zha et al.[29] report on the influence of H2 S concentration, cell voltage, time of exposure, and temperature. With 50 ppm H2 S at 800 ◦ C and constant cell voltage of 0.6 V, a drop of 16.7% was observed within minutes of the introduction of sulfur, followed by a slow, continuous further drop up to 4% during the next 120 h. When clean hydrogen/nitrogen was introduced, the performance improved immediately followed by a gradual recovery, which was not complete. The cell stabilized at 96% of the original performance. Independent experiments with clean fuel had shown that the cells were completely stable at 800 ◦ C after an initial conditioning at 900 ◦ C. The same picture was repeated with 2 ppm H2 S, but the degradation was only 12.7%. Recovery was quicker and more complete with only 1% loss in performance. The effect of shortterm (6 min) exposure of the anode to 2 ppm H2 S could be completely recovered. This means that an SOFC stack will not be permanently damaged by a spike in the anode feeds due to upsets. The immediate impact of H2 S as a function of concentration and temperature is illustrated in Figure 10. Even 0.18 ppm H2 S caused a drop in performance at all temperatures. Hansen[30] have analyzed the data of Zha et al.[29] There is no mention in Ref. [29] of the exact active area of the SOFC cells used, but assuming it to be 0.25 cm2 as in other work by the group, it can be calculated that the fuel utilization is below 6% even at the highest power densities of 0.52 A cm−2 with the feed flow given as 20 ml min−1 . The feed gas compositions given: 50% H2 ,
24 22 20 18 16 14 12 10 8 6 4 2 0 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 θs
1
1.1
Figure 11. Cell performance loss as function of θS . Measurements from Ref. [29]. Potentiostatic at 0.7 V; diamonds 900 ◦ C, squares 800 ◦ C, and triangles 700 ◦ C; cell performance loss = 53.8 θS – 32.2; R 2 = 0.985. [S. Zha, Z. Cheng, and M. Liu, J Electrochem Soc 154, B201(2007). Reproduced by permission of ECS - The Electrochemical Society.]
Part 5: Performance degradation
Table 2. Power loss as function of operating conditions. Galvanostatic masurements. H2 S (ppm)
mA cm−2
Power loss (%)
References
1000
10 10
750
1
10.3 15.6 19.0 56.0 6.5 9.8 11.8
[31]
800
160 250 130 400 250 500 900
Temperature ( ◦ C)
[32] [33]
at 1023, 1173, and 1273 K respectively. Using equation (10) and (14) from above and 79% H2 in the gas, it can then be calculated that effect of sulfur was noticed at a θS of 0.645, 0.64, and 0.645 at 1023, 1173, and 1273 K, respectively, which is in very good accordance with the findings from Figure 11. The trend with respect to sulfur sensitivity and current density can at first sight seem rather confusing. It depends on whether the measurements are carried out in the galvanostatic mode (constant cell current) or potentiostatic mode (constant cell voltage). Examples of the galvanostatic mode are given in Table 2. It can be seen that in all cases that impact of sulfur on power loss increases with increasing current density. This is in contrast to the findings of Zha et al. cited above,[29] but they operated in potentiostatic mode. This apparently contradictory evidence can be understood by modeling the stack as a simple equivalent circuit as pointed out by Cheng et al.[34] When operating on clean fuel, the situation is described by U0 = E0 − I0 Rcell
0
Cheng et al.[34] carried out experiments both in the galvanostatic and the potentiostatic modes at 800 ◦ C with H2 S concentrations from 0.18 to 10 ppm. The relative increase in overall internal cell resistance was calculated from equations (17) and (18) and it was demonstrated that the apparent contradiction concerning the sulfur impact as a function of current density is resolved if the influence of sulfur is correlated with the increase in cell resistance instead of power loss. Hansen[30] also used the data of Cheng et al.[34] and correlated them with the use of equation (10) and an equation similar to equation (14). The results are displayed in Figures 12 and 13 for the galvanostatic and the potentiostatic cases, respectively. Again, it can be observed that the 20 Increase in cell resistance (%)
8
18 16 14 12 10 8 6 4 2 0 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 θs
1
1.1
Figure 12. Increase in cell resistance (%) as function of θS . Measurements. Galvanostatic, diamonds 241 mA cm−2 , squares 409 mA cm−2 . Increase in cell resistance (%) = 61.7θS − 34.8; R 2 = 0.970 − 241 mA cm−2 . Increase in cell resistance (%) = 63.3θS − 39.1. R 2 = 0.989 − 409 mA cm−2 . [Reproduced from Ref. [34]. Elsevier, 2007.]
(15) 20
Rrel
Rcell = Rcell 0
(16)
It is possible after some algebraic manipulation to deduce the following useful equations. For the galvanostatic case,
18 Increase cell resistance (%)
When the cell is exposed to H2 S, the total internal resistance increases to Rcell 0 + Rcell . Introducing the relative decrease in power output Pr and the relative increase in total cell internal resistance,
16 14 12 10 8 6 4 2 0
Rrel =
Rcell U0 = Pr Rcell 0 E 0 − U0
0
(17)
and for the potentiostatic case, Rrel =
Rcell Pr = Rcell 0 1 − Pr
(18)
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 θs
1
1.1 1.2
Figure 13. Increase in cell resistance (%) as function of θS . Measurements. Potentiostatic, diamonds 0.7 V, squares 0.535 V. Increase in cell resistance (%) = 71.5θS − 44.7. R 2 = 0.991 − 0.7 V. Increase in cell resistance (%) = 68.1θS − 45.8. R 2 = 0.88 − 0.535 V. [Reproduced from Ref. [34]. Elsevier, 2007.]
Sulfur poisoning on Ni catalyst and anodes 9 data is correlated very well by equations (10) and (14). The slopes of the curves are very similar, but the intercept with the abscissa, θmin , increases with increasing current density. From 0.564 to 0.618 in the galvanostatic mode and from 0.625 to 0.672 in the potentiostatic mode. It is, of course, even more interesting to study specifically the increase in the anode polarization, area specific resistance (ASR), as a function of H2 S concentration, but this is seldom done. An exception is the result from Ref. [29] where the increase at open circuit voltage (OCV) ascribed to anode polarization is compared to the increase at 0.7 V and initial current density around 260 mA cm−2 . It is evident that the anode interfacial resistance increases by up to 200% even at low (
