- Email: [email protected]
Accurate in-operando study of molten carbonate fuel cell degradation processes -part I: Physiochemical processes individuation
Electrochimica Acta 291 (2018) 343e352
Contents lists available at ScienceDirect
Electrochimica Acta journal homepage: www.elsevier.com/locate/electacta
Accurate in-operando study of molten carbonate fuel cell degradation processes -part I: Physiochemical processes individuation ~ oz a, S.J. McPhail a, F. Santoni a, b, *, M. Della Pietra a, D. Pumiglia a, C. Boigues Mun V. Cigolotti c, S.W. Nam d, M.G. Kang d, S.P. Yoon d a
DTE-PCU-SPCT, ENEA C.R. Casaccia, Via Anguillarese 301, 00123, Rome, Italy Department of Science and Technology, DIST, University Parthenope, Centro Direzionale Isola C4, Naples, 80143, Italy ENEA, Italian National Agency for New Technologies, Energy and the Environment, Portici Research Centre, Piazzale Enrico Fermi 1, 80055, Naples, Italy d Fuel Cell Research Center, KIST - Korea Institute of Science and Technology, 5, Hwarang-ro 14-gil, Seongbuk-gu, Seoul, South Korea b c
a r t i c l e i n f o
a b s t r a c t
Article history: Received 26 February 2018 Received in revised form 2 August 2018 Accepted 14 August 2018 Available online 21 September 2018
This work has the difficult task to deeply study the electrochemical processes that occur inside a 100 cm2 of Molten Carbonate Fuel Cells (MCFC) impedance spectra using the high resolution of Distribution of Relaxation Time (DRT) method. Using this method, it is possible to shed light on the different physicochemical processes occurring within these cells, identifying the characteristic relaxation times by means of an appropriate experimental campaign where temperature and gas compositions in anode and cathode were varied one at a time. The quality of the recorded spectra was verified by Kramers-Kronig relation before applying DRT calculations. In this work, five distinct and separated peaks with different time constants ranging from 0.01 to 500 Hz were identified and associated with physiochemical processes of the cell. Three peaks at high frequency represent the charge transfer processes in anode and cathode active sites. The other two, located at low frequency, are associated with the gas diffusion in the electrodes and to the gas conversion process. This study represents the first application of the DRT approach to this technology allowing to understand the physicochemical origin of the individual polarization processes controlling the cell performance and the degradation. The analysis of degradation processes using the DRT method and the physiochemical processes identification presented in this paper will be shown in part II of this work. © 2018 Elsevier Ltd. All rights reserved.
Keywords: Distribution of relaxation time Molten carbonate fuel cell Polarization losses Deconvolution of impedance spectra Degradation processes
1. Introduction Molten carbonate fuel cells (MCFCs) are promising power generation devices enabling high-efficiency co-generation of electricity and heat with minimal environmental impact [1]. The main advantage in the development and the use of MCFC technology on a large scale is the exploitation of its operational flexibility. In fact, the high operating temperature (approximately 923 K) allows MCFCs to use CO (which is a contaminant for low temperature fuel cells) and low weight hydrocarbons as fuels. These fuels can be converted into hydrogen via water gas shift (WGS) and steam reforming (SR) reactions respectively, and this can take place directly in the cells (internal reforming) or in a separate reformer (external reforming) using the heat generated by the operating stack(s) [2,3].
* Corresponding author. DTE-PCU-SPCT, ENEA C.R. Casaccia, Via Anguillarese 301, 00123 Rome, Italy. E-mail address: [email protected] (F. Santoni). https://doi.org/10.1016/j.electacta.2018.08.100 0013-4686/© 2018 Elsevier Ltd. All rights reserved.
In addition, this technology can be exploited as CO2 concentrator thanks to its intrinsic operating mechanism. In fact, the electrochemical reactions taking place within the MCFC involve the migration of carbonate ions from the cathode to the anode, that corresponds to a depletion of CO2 in the oxidant fed gas and the subsequent enrichment of CO2 in the anode exhaust gas. During this process, the MCFC has the unique ability to operate as an integrated CO2 concentrator and power generation unit, allowing to integrate it with conventional power generation in combination with carbon capture and storage (CCS) technology for the removal of the CO2 from combustion flue gas [4,5]. These practical advantages explain the power generation market interest in MCFCs, but being a commercially embryonic technology compared to the conventional power generation systems, the resources needed to bring this technology into a commercial breakthrough necessary for a fast market penetration are still significant. Primarily, from an economic point of view, the costs of fuel cells per kW is still considerably higher than conventional power
344
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
plants. In addition, an important requirement for commercial applications is the long-term stability. In fact, for stationary applications, agreed target lifespan is generally more than 40,000 h. The two main causes of performance degradation throughout cell lifetime are the loss of electrolyte and dissolution of the NiO in the cathode. Initially, the electrolyte is consumed by corrosion reactions with metal hardware in the fuel cell. Over a longer time, and at a more constant rate, the electrolyte is primarily lost by vaporization into the fuel gas. Although a careful electrolyte management with improved pore structure to maintain electrolyte within the electrodes has reduced the electrolyte loss due to vaporization, this mechanism remains the main cause of performance degradation in atmospheric systems [6]. The cathode is made of NiO, which is slightly soluble in the electrolyte (Eq. (1)). Dissolved nickel ions diffuse into the electrolyte, and precipitate as metallic nickel forming electronically conducting pathways between the anode and cathode.
NiO þ CO2 /Ni2þ þ CO2 3
[1]
The solubility of NiO is managed by a combination of electrolyte composition, gas composition, pressure and operating temperature in the electrolyte. It was early identified as a major obstacle to an increased lifetime of the fuel cell, as it threatens the integrity of the fuel cell by causing short-circuiting [7e10]. The identification of these degradation mechanisms is possible with a post-mortem analysis, underlining the difficulty to recognize them in-operando. Although many studies have been conducted to reduce the degradation mechanisms that lead to cell failure [11e14], progress for a further performance improvement is partly constrained by an incomplete understanding of the physicochemical origin of the individual polarization processes controlling the cell performance. Attending to these considerations, the scientific and industrial communities should work hand-in-hand to enhance the production of a high-performing, and robust products, developing advanced analysis tools and techniques to fully understand the degradation phenomena. In this scenario, Electrochemical Impedance Spectroscopy (EIS) is proposed as a powerful tool to assess the MCFC performance [15]. EIS can be recorded without altering the working conditions of the fuel cell and can thus be used to determine the impact of different parameters (e.g. reactants stoichiometry, current density, temperature) during cell operation. In literature, many works exploit this task, from the pioneering works of Selman et al. [16] or Nishina et al. [17] until to the recent works of C. -G. Lee [18,19], demonstrating the validity of this technique as an in-situ method for characterizing rate-limiting processes in MCFCs porous electrodes. Even though EIS is a very sensitive technique, it can hardly distinguish two processes that differ in their characteristic frequency by less than two decades, thus resulting deeply convoluted in the EIS spectrum and consequently in a not clear image of the real electrochemical processes that occurring the cell. Impedance spectra can be interpreted using an equivalent circuit model (ECM) [20]. In this approach, physical phenomena, taking place in the fuel cell during operation, are represented by equivalent electrical elements. The value of such elements is obtained by the fitting of the measured impedance data by means of mathematical algorithms. An appropriate use of ECM to fit impedance data relies on a priori knowledge about the system since it requires a suitable model structure that represents the physical processes occurring in the cell. Conversely, if there is a little, or no a priori knowledge about the system, the choice of an appropriate
model structure can become a difficult task, leading often to a severe ambiguity of the adopted model [15]. In order to overcome this issue, an alternative approach for analysing impedance spectra is the Distribution of Relaxation Times (DRT) method. Originally introduced for the impedance analysis of solid oxide fuel cells [21e23] and lithium-ion batteries [24], DRT can be used for process identification in complex electrochemical systems, where resistive-capacitive (R//C) features are dominant [22]. The individual processes are separated on the basis of their typical time constants, derived from the associated (R//C) elements, allowing to bypass ambiguous EC modelling and to analyse spectra without pre-assumptions. Moreover, DRT analysis is a useful instrument to investigate the degradation phenomena during a long-term test. In fact, as demonstrated in different studies on SOFCs [25,26], the use of DRT analysis combined with a valid ECM gives the possibility to observe qualitatively and quantitatively the behaviour of individual polarization processes in operando. In this field, accelerated lifetime tests can become a powerful tool to reduce significantly the experimental efforts for long-term tests [27]. A direct application of the existing monitoring methodologies is unfeasible as the lifetime and degradation rate of a fuel cell are governed by several concurring processes. A deconvolution of the different degradation mechanisms by EIS coupled with DRT can provide the required information about the different ageing phenomena in the cell. The global aim of this work is to analyse the degradation phenomena in 100 cm2 class MCFCs, during an endurance test. The chosen tools are the EIS coupled with DRT method allowing to monitor the individual polarization processes controlling the cell performance in operando, and microstructural characterization techniques used for pre- and post-mortem analysis. Since no other study is present in the literature about DRT applied to MCFCs, an identification of the single processes was necessary. For this reason, in this paper is reported the first part of the work, regarding the first findings on the deconvolution of MCFC impedance spectra and the identification of t the individual polarization processes using an appropriate experimental campaign where temperature and gas compositions in anode and cathode were varied one at a time. Whereas the results obtained from the endurance tests will be discussed in the second part. 2. Experimental 2.1. Equipment setup The experimental data produced to support this activity were carried out at the Fuel Cell Research Center laboratories of KIST in Seoul, South Korea, using planar MCFC single cells with an active area of 100 cm2. The characteristics of the single cell components used are summarized in Table 1. The electrolyte layers are alternated with the matrix layers and then assembled between the anode and cathode. The obtained fuel cell is sandwiched between the two current collectors and included in the anodic and cathodic frames (see Fig. 1a), where fuel and oxidant stream were in concurrent flow. The cell was held in a heating block under a mechanic load of 2 kgcm2, obtained with a hydraulic piston as shown in Fig. 1b. The anode and cathode gas compositions, fed to the cells, were measured and controlled by Brooks 5850E Digital Mass Flow Controllers and a thermo-controlled bubbler is used to adjust the required water content in the fuel. The temperature of the anodic inlet pipeline is controlled through heating tapes from the water injection point up to the
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
345
Table 1 Specifications of 100 cm2class MCFC components. Single cell components
Specification
Cell frame
Size [cm x cm] Material Size [cm x cm] Thickness [mm] Current collector Material Porosity Fuel gas (mol ratio) Total flow rate [ml/min] Size [cm x cm] Thickness [mm] Current collector Material Porosity Oxidant gas (mol ratio) Total flow rate [ml/min] Material Mole ratio Material Thickness [mm]
Anode
Cathode
Electrolyte Matrix
13 13 AISI 316L 11 11 0.7 Pure Ni Ni þ 5%wt Al 55e60% Air: CO2 ¼ 0.7:0.3 951 10 10 0.7 AISI 316L Lithiated NiO 60e65% H2: CO2: H2O ¼ 0.72:0.18:0.1 396 Li2CO3/K2CO3 68:32 g-LiAlO2 1.2
Fig. 1. Experimental set-up for the MCFC single cell: (a) Schematic illustration of the single cell assembling (b) Single cell apparatus pressurized inside the oven, and (c) single cell MCFC test equipment.
furnace entry, in order to prevent water condensation inside the inlet pipes and pre-heat the gas mixture. Before the single cell start-up, a pre-treatment was carried out to remove the organic binder by thermal decomposition for 4 days in air at the temperature ranging from 298 K to 723 K and then for 3 days in CO2 at the temperature ranging from 723 K to 923 K. Under the latter conditions, which were very critical for the electrolyte melting region, CO2 was passed through the system at a low flow rate to maintain the distribution of the electrolyte throughout the pores of the matrix, of the cathode, and of the anode. Moreover, this procedure prevented evaporation of the electrolyte. After pre-
treatment, the temperature of the gas in the MCFC was maintained at 923 K for 180 h, using KIST normal gas condition reported in Table 1. The temperature in the MCFC was maintained at the reference temperature of 923 K, or at a different one required for a given test, thanks to the furnace and the heating plates holding the single cell, so that the cell plane temperature could be considered uniform. An electrical loader (ELS300Z, ELTO DC Electronics Co.) was used to impose the external current and to collect the cell voltage; while the nitrogen crossover at the anode outlet was monitored by using gas chromatography (789 A, Agilent Tech.)
346
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
Electrochemical impedance spectroscopy (EIS) measurements were carried out to analyse the frequency response of the cells tested, within the frequency range 10 kHz - 0.01 Hz, using a frequency response analyser (FRA 1255B, Solartron Co.) coupled with an electrochemical dielectric interface (EI, 1287 Solartron Co.). All impedance measurements performed within this work were carried out under open-circuit conditions. ZView® software for Windows from Scribner Associates Inc. was used in order to fit the empirical data and to calculate the values of resistances associated with each physiochemical process.
2.2. Distribution of relaxation times (DRT) method: deconvolution of signals MCFCs are complex electrochemical devices, in which many processes occur at the same time, each process being characterized by its typical relaxation time. When two (or more) processes have a similar relaxation time, it is hard to separate and distinguish their contributions to the overall impedance of the cell, since the EIS spectra can clearly distinguish processes separated at least by two decades in the frequency domain. Furthermore, it is hard to identify small contributions, these being totally covered by processes with large polarization losses. An instrumental tool to overcome this issue may be the analysis of the DRT functions [21,23,28]. The DRT-method uses the fact, that every impedance function that obeys to the Kramers-Kronig relations can be represented as a differential sum of infinitesimal small RC-elements [22]. This RC circuit is characterized by its time constant t and can be associated with an electrochemical process. The relation between the impedance spectrum Z(u) and the distribution of relaxation times t, which is for sake of simplicity often displayed as a distribution g(t) vs. the relaxation frequencies, is given by Eq. (2):
ZðuÞ ¼ R0 þ Zpol ðuÞ ¼ R0 þ Rpol
∞ ð
0
gðtÞ dt 1 þ jut
[2]
where Zpol(u) represents the overall polarization resistance of the fuel cell and R0 the purely ohmic one. The time constant of a single RC-element is designated by t ¼ RC, the fraction of the overall polarization resistance with relaxation times between t and t þ dt gðtÞ is indicated by the term. 1þj ut
This relation is valid for any kind of system that fulfils the fundamental demands of impedance spectroscopy linearity, causality and time invariance, and is applied a discrete Fourier transformation to compute the DRT from the imaginary part of Z(u). An appropriate extrapolation and filtering of the data are required to suppress numerical error amplification. Thus, the method becomes unsuitable for impedance spectra exhibiting greater errors or inductive artefacts at higher frequencies [29]. The mathematical problem with this approach arises from the inversion of Eq. (2), which is necessary in order to extract g(t) from the measured impedance data Zpol. This problem is known to be an ill-posed problem and requires special methods to be solved in order to avoid false peaks and oscillations. Even if there are numerous methods which can solve this kind of problem [30,31], the most renowned and reliable one is the Tikhonov regularization algorithm which uses a self-consistent regularization parameter l, which determines the smoothness of the DRT. In the case of excellent data quality or synthetic data; a small regularization parameter can be used, for noisy data the regularization parameter has to be increased to suppress false peaks and oscillations in the DRT. Further details about the
regularization can be found in Tikhonov and Groetsch works [31,32]. The usage of this tool gives access to information that until now, for the limited resolution of the Impedance Spectra, was not possible to obtain. In fact, as shown in Fig. 2, from the comparison between the impedance spectrum and the correspondent DRT, it is immediately visible that the raw EIS response has a lower resolution, in fact as just mentioned above it is difficult to clearly visualize the contributions of single physicochemical processes to the total polarization resistance. Moreover, the information obtainable from the analysis of EIS spectra provides a quantitative measurement of the polarization losses, but nothing can be said, for instance, about the nature of the process and where it takes place, anode or cathode. This impedes a profound and complete comprehension of the evolution of these processes and their dependency to the operating conditions of the cell. On the other hand, it is evident that the higher resolution of the DRT allows a clear identification of 5 peaks/ processes with a characteristic frequency domain. The nature of these peaks has been assessed by means of a systematic experimental campaign to investigate the parameter dependence. This allows identifying all the processes, ascribing them to the anode or cathode. For the DRT calculation, an in-house software running in MATLAB® has been used. The DRT method becomes unsuitable for impedance spectra exhibiting greater errors or inductive artefacts at higher frequencies. For this reason, the measurement data quality is crucial. The quality and amount of information that can be extracted from impedance data are implicitly connected to the noise-level and the compliance of the measured curve with the principles of causality, linearity, and stability. A well-established method used to assess the consistency and quality of measured impedance spectra is the Kramers-Kronig validation [15]. In this work, the Kramers-Kronig transformation rules were applied to every impedance data set measured throughout our experimental activity by using the “KK test for Windows” software [33,34]. 2.3. Experimental procedure In order to elucidate the nature of each peak obtained with the DRT method, the cells were operated under a number of different predefined conditions, starting from a reference condition (indicated in bold in the first line) and varying each time one of the following parameters: temperature, anodic gas composition and cathodic gas composition reported in Table 2. For each condition, an
Fig. 2. Imaginary component of the cell impedance and the DRT distribution function (l ¼ 0.1) when the MCFC cell is operating at 923 K under normal KIST conditions.
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
347
Table 2 Test matrix containing the molar fractions of the gas species in anode and cathode used for the DRT analysis. Anode Gas
Anodic pH2 variations (atm)
Anodic pH2O variations (atm)
Anodic pCO variations (atm)
Cathodic pO2 variations (atm)
Cathodic pCO2 variations (atm)
Temperature variations (K)
Cathode Gas
Test
pH2
pCO2
pH2O
pN2
pCO
pO2
pN2
pCO2
I II III IV V I II III IV V I II III IV V I II III IV I II III IV V VI I-853 II-873 III-893 IV-923 V-953
0.72 0.50 0.20 0.10 0.05 0.5 0.5 0.5 0.5 0.5 0.1 0.1 0.1 0.1 0.1 0.72 0.72 0.72 0.72 0.72 0.72 0.72 0.72 0.72 0.72 0.72 0.72 0.72 0.72 0.72
0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18
0.10 0.10 0.10 0.10 0.10 0 0.10 0.15 0.20 0.32 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10
0 0.22 0.52 0.62 0.67 0.32 0.22 0.17 0.12 0 0.22 0.32 0.42 0.52 0.57 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0.40 0.30 0.20 0.10 0.05 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.09 0.06 0.03 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15
0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.61 0.64 0.67 0.45 0.55 0.62 0.70 0.75 0.80 0.55 0.55 0.55 0.55 0.55
0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.40 0.30 0.23 0.15 0.10 0.05 0.30 0.30 0.30 0.30 0.30
EIS measurement was recorded, and the DRT function was calculated from the experimental data acquired. The anode and cathode gas flow rates are maintained at a constant value of 396 ml/min and 951 ml/min respectively during all experiments. 3. Results and discussion 3.1. Validity of measurement data As introduced in Section 2.2, application of DRT necessitates an outstanding measurement quality. Therefore, every measured spectrum presented in the following evaluation has been validated by Kramers-Kronig relations before applying DRT calculations. Fig. 3a and b shows a KramerseKronig validation for a typical impedance spectrum obtained during the experimental campaign and the corresponding Kramers-Kronig residuals. The residuals are less than 0.5% of the whole frequency range. These considerably small residuals are taken as a confirmation of Kramers-Kronig validity indicating that the experimental data impeccably agree with the expected theoretical results. 3.2. Preliminary electrochemical performance comparison The experimental procedure employed for the identification of the different operating processes was carried out on three distinct MCFC single cells; however, the tested cells, described in Table 1, came from the same batch, thus ensuring (as will be seen) a substantial homogeneity. Nevertheless, a preliminary performance characterization was performed on each cell in order to confirm the comparability of the tested samples and of the results obtained from the subsequent experiments. The comparison was made using the normal KIST condition reporting in Table 1 at the temperature of 923 K. The polarization curve reaching a utilization factor of 0.4 both for anode
and cathode, corresponding to 150 mAcm2 of current density. As depicted in Fig. 4, the polarization curves (Fig. 4a) and EIS spectra (Fig. 4b) show that the overall performances of the tested cells are nearly identical. This similarity on the performance is shown also in the corresponding DRT spectra (Fig. 4c) showing the same peaks in the same position, concluding that the tested cells can be considered comparable for the objectives of the experimental campaign in which they have been employed. 3.3. Processes individuation Fig. 5a shows a series of impedance spectra recorded under different fuel compositions, where the partial pressure of hydrogen was varied between 0.72 and 0.05 atm. In Nyquist representation, the impedance spectra show two well-separated arcs. The low-frequency arc indicates a pronounced increase in resistance with decreasing of the hydrogen quantity while a lessmarked dependence is visible for the higher frequencies of the impedance. The DRT calculated from this spectrum is shown in Fig. 5b. Obviously, the visual inspection of the DRT gives considerably more detailed insight into the system than the raw impedance plot. Five distinct and separated peaks with different time constants ranging from 0.01 to 500 Hz are identified by the DRTs. In accordance with the EIS spectrum, only a slight dependence is shown for the peaks located at high frequency, P1an and P3cat, whereas the most influenced peaks are located at low frequencies, P4diff and P5an. The two low frequency processes are the imprints of concentration polarization effects [35,36] and could be considered the deconvolution of the low frequency circle associated to the anodic gas-phase mass transfer processes as has already been suggested by Vogel and Bregoli [37], Yuh and Selman [38], and Lee [39]. As evident in Fig. 5b, these two peaks present an inverse
348
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
Fig. 3. (a) Nyquist plot recorded at 923 K reporting the comparison between the measured EIS obtained for the third composition of pCO2 variation, described in Table IV, and the fitting result obtained with the KKtest (b) computational residuals of Kramers-Kronig reconstruction as a function of frequency.
Fig. 4. Polarization curves (a) EIS spectra (b) and DRT spectra(c) of single cell tested employed for the process characterization experiments.
behaviour in relation to the stepwise hydrogen increment. The behaviour of P4diff, (decreases with the increment of the hydrogen) probably reflects the gas-phase diffusion at the anode side. Whereas the P5an seems to be associated with a gas conversion process due to the water gas shift reaction (WGS) represented in Eq. (3). In fact, the increment of the H2 concentration in the fuel (product of the WGS reaction) inhibits the reaction, provoking the increment of the resistance associated with this peak.
CO þ H2 O#CO2 þ H2
[3]
The analysis of the H2O partial pressure variations and the CO partial pressure variations are represented in Fig. 5c and d respectively. For CO variations was used also a small amount of H2 (0.10 atm) to avoid an excessive stress on the cell, as reported in Table 2. From both the graphs, it is clearly visible how the processes
P4diff and P5an decrease with the increment of H2O and CO, whereas P2cat does not follow a linear behaviour or is not influenced by these anodic compositions. About the other high frequency peaks, it is important to note that P3cat shows only a minimal dependence from the H2 variations respect to P1an (Fig. 5b). Moreover, it does not present any linear dependence from CO and H2O variations (Fig. 5c and d), while. P1an presents a slight dependence also from the H2O variations and its behaviour is very similar, both for intensity and movement of the peak, to those obtained under H2 variations. The reduction of P5an due to the increment of CO and H2O is because they are directly involved as reagents of the WGS reaction, and their increment reduces the resistance associated with this process. These results are coherent with the hypothesis that the P5an is associated to the WGS reaction. As shown in Fig. 5c the anodic diffusion peak P4diff is strictly
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
349
Fig. 5. Nyquist plot of MCFC single cell under stepwise increase of hydrogen partial pressure(a) and DRT spectra for the anodic partial pressure variations: hydrogen variations(b), CO variations (c) and H2O variations (d).
linked to the conversion of CO in H2 by means of the WGS reaction, for this reason, the two peaks are in this case very similar behaviour. The DRT analysis obtained from the cathodic composition variations is reported in Fig. 6a and b. For the O2 partial pressure variations, four different compositions varying the O2 content from 0.15 to 0.03 atm were tested. In this graph, the process P4diff shows a significant dependency on the O2 content increasing with its decrement, while the process P3cat presents the same behaviour but with a minor dependency. The other three processes present in the spectrum (P1an, P2cat, P5an) are not affected or do not present a linear behaviour with the O2 variations. In Fig. 6b, is reported the DRT analysis obtained for the CO2 partial pressure variations. A similar result to O2 variations has been obtained, in fact, the principal process involved in this analysis is
the P4diff. It is relevant to notice that this process is related to the cathodic gas phase diffusion, showing a strong dependence on the cathodic compositions changes. Moreover, in this analysis is shown also a linear dependence of the peak P2cat from the increment of CO2 in the gas mixture and a slight dependence of P3cat that shows a contrary behaviour. Comparing anodic and cathodic composition variation results, it is possible to see in both cases the relevance of the process P4diff. It is possible that P4diff is composed of two processes, respectively related to the cathodic and anodic diffusion. In fact, it has to be taken into account that if the characteristic frequency of two adjacent processes differs of less than a decade in frequency, two peaks may overlap. A possible explanation of the similar frequency of the diffusion processes above mentioned is the comparable thickness and porosity of the two electrodes as reported in Table 1.
Fig. 6. DRT spectra for the cathodic partial pressure variations: O2 variations (a), CO2 variations (b).
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
dependence with the CO2 partial pressure variations and could be associated with the reaction of CO2 with oxide ions to form carbonate (Eq. (5)). 2 O 2 þ e /O2
[4]
2 O2 2 þ 2CO2 þ 2e /2CO3
[5]
In recent works carried out by Czelej et al. [44,45] the mechanism of the electrode reaction at triple phase boundary in MCFC cathode is proposed based on atomic-scale simulations, and the oxygen reduction is shown as the rate determining step. Moreover, in these studies has been elucidated the prominent role of Oterminated octopolar NiO (111) in the cathodic transformation of carbon dioxide to carbonate anion. The results indicate that Eq. (5) on O-terminated octopolar NiO (111) is facile under the MCFC operation conditions and follows Mars-van Krevelen (MvK) and Eley-Rideal (ER) sequential mechanisms. A less-marked dependence is shown for the processes located at low frequency. In particular, the resistance associated with P4diff decreases with the temperature increment. In case of MCFC, the pore size at the electrodes (~9 mm at the cathode and ~3 mm at the anode) are larger than the SOFC’s electrodes (~sub-mm). In general, in the pore diffusion of gas molecules, Knudsen diffusion is domi1
nant in the pore size of below sub-mm, which shows Rgd fT 2 . However, viscous flow is dominant in the large pores of MCFC 1
electrodes, which shows Rgd fT 2 . Therefore, the diffusion of gas molecules through the electrodes is promoted by the increment of the temperature. Concerning the process P5an, it slightly increases when the temperature rises. This is due to the change of the equilibrium constant of the WGS reaction, that decreases from 2,82 at 853 K to 1,60 at 953 K, causing a reduction of the reaction products and consequently an increase of the resistance [46]. An accurate quantitative relation of the WGS to the resistance based on a physically meaningful equivalent circuit model (ECM) built-up on the base of knowledge given by DRT analysis will be presented in the second part of the work. =
Moreover, since the electrodes in the MCFCs are supposed to be covered with thin electrolyte films estimated as much less than 1 mm [40], the P4diff represents not only a mass transfer process through the gas phase but also the one due to the liquid phase. In Fig. 7a has been reported the EIS measurements obtained at five different temperatures from 853 K to 953 K, using the composition reported in Table 2. In accordance with Yuh and Selman [38], the cell temperature has a strong effect on the highfrequency loop, in fact, as the temperature rises the highfrequency arc shrinks, while the low-frequency loop is always quite the same size. In accordance with the EIS spectra, in the related DRT analysis (Fig. 7b) is evident how the processes located at high-frequency P1an, P2cat and P3cat are all characterized by a pronounced thermal activation. Although the gas composition variations in some cases (H2 variations, O2 variations and CO2 variations) influence this high-frequency peaks, it seems noticeable how the temperature is the predominant operating parameter affecting these processes. In fact, the intensity of the peak P3cat, increases of 300% (from approximately 0.06 ohmcm2/s at 853 K to 0.02 ohmcm2/s at 953 K), P2cat increases of 200% (from approximately 0.03 ohmcm2/s at 853 K to 0.06 ohmcm2/s at 953 K) and P1an increases more than 100% (from approximately 0.06 ohmcm2/s at 853 K to 0.13 ohmcm2/ s at 953 K). Their pronounced thermal activation and their characteristic frequency suggest that these peaks are related to the charge transfer mechanisms in the electrodes. P1an., showing also a slight dependence from the amount of H2 present in the fuel, could be associated with the H2 oxidation at the anode, whereas P2cat and P3cat are ascribable at the cathode processes being influenced by the cathodic composition variations. In particular, P3cat is influenced by the partial pressure of oxygen and CO2 (slightly influenced), for this reason, it could be associated with the O2 reduction (Eq. (4)). In fact, in literature, two main possible paths for the O2 evolution are described [41]: peroxide and superoxide paths. As suggested by Hu [42] and Appleby [43], both of them are limited by the diffusion of the oxygen ions and involve a resistance that increases when the CO2 concentration grows or the O2 concentration decreases. Concerning to the process P2cat, it shows a linear
=
350
Fig. 7. Series of impedance spectra (a) and corresponding DRT spectra obtained at different operating temperature from 853 K to 953 K (b).
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
351
Table 3 Resume of the physicochemical processes occurring in the MCFC and their corresponding peaks in the DRT spectra. Process
P1an P2cat P3cat P4diff P5an
Frequency (Hz)
500…100 100…50 50…5 5…0.5 0.5…0.03
Dependency
Electrode
pH2
pH2O
pCO
Temperature
pCO2
pO2
Low no no High High
Low no no High Medium
no no no High High
High High High Medium Low
no Low no High no
no no Low High no
Table 3 gives an overview of the processes identified by the DRT analysis together with their characteristic frequency range, gas partial pressure and temperature dependency.
4. Conclusions In this study, the distribution of relaxation times was applied successfully to analyse impedance spectra of 100 cm2 Molten Carbonate Fuel Cells. To the authors ‘knowledge, this study represents the first application of the DRT approach to the MCFCs. It is evident, that DRT gives considerably more detailed insight into the system than the raw impedance plot. In fact, five distinct and separated peaks with different time constants ranging from 0.01 Hz to 500 Hz are identified. Three distinct MCFC single cells were tested, and a preliminary performance characterization was performed on each cell, in order to confirm the comparability of the tested samples and of the results obtained from the subsequent experiments. To investigate the parameter dependence of each single polarization process, a series of impedance measurements were carried out, changing only one cell parameter at a time (cathodic gas composition variations, anodic gas composition variations and temperature). The spectra obtained were validated by Kramers-Kronig relations before applying DRT calculations, reporting an error of less 0.5% of the whole frequency range. Investigating the parameter dependence, the physical/chemical origins of these processes can unambiguously be determined, as follows: C P1an (500-100 Hz) Anodic charge transfer process: electrodeoxidation of H2. C P2cat (100-50 Hz) Cathodic charge transfer process: carbonate formation. C P3cat (50-5 Hz) Cathodic charge transfer process: O2 reduction. C P4diff (5e0.5 Hz) Anodic and Cathodic Gas-phase and Liquidphase Mass transfer process: Diffusion. C P5an (0.5e0.03 Hz) Anodic Mass transfer process: gas conversion process due to the Water Gas Shift reaction. The possibility to identify the polarization losses based on their respective time constants is an interesting instrument as the base on which generate comprehensive and robust equivalent circuit models (ECM) and develop physical models for the prediction of the performance degradation in MCFC cells. The possibility to identify the individual polarization losses based on their respective time constants using this strategy (EISDRT) gives access to electrochemical information that till now, for the limited resolution of the Impedance Spectra, was not possible to obtain. This strategy gives the base to generate comprehensive and robust equivalent circuit models to obtain a quantitative value of the resistance associated with each process. In the second part of this work, this tool will be applied to study the degradation processes in operando during an endurance test.
Anode Cathode Cathode Anode and Cathode Anode
Acknowledgement This research was supported by the Global Research Laboratory Program through the National Research Foundation of Korea funded by the Ministry of Science, ICT and Future Planning of Republic of Korea (NRF-2009-00406) and by the Korea Institute of Energy Technology Evaluation and Planning (KETEP), funded by the Ministry of Trade, Industry and Energy, (No. 20153030040930). References [1] R. O’Hyare, S.W. Cha, W. Colella, F.B. Prinz, Fuel Cells Fundamentals, third ed., John Wiley & Sons, New York, 2016 https://doi.org/10.1002/9781119191766. [2] L. Barelli, G. Bidini, S. Campanari, G. Discepoli, M. Spinelli, Performance assessment of natural gas and biogas fueled molten carbonate fuel cells in carbon capture configuration, J. Power Sources 320 (2016) 332e342, https:// doi.org/10.1016/j.jpowsour.2016.04.071 0378e7753. [3] V. Chiodo, G. Zafarana, S. Maisano, S. Freni, A. Galvagno, F. Urbani, Molten carbonate fuel cell system fed with biofuels for electricity production, Int. J. Hydrogen Energy 41 (2016) 18815e18821, https://doi.org/10.1016/ j.ijhydene.2016.05.119. [4] G. Discepoli, G. Cinti, U. Desideri, D. Penchini, S. Proietti, Carbon capture with molten carbonate fuel cells: experimental tests and fuel cell performance assessment, Int. J. Greenh. Gas Con 9 (2012) 372e384, https://doi.org/10.1016/ j.ijggc.2012.05.002. [5] M. Brenna, F. Foiadelli, G. Manzolini, Grid connection of MCFC applied to power plant with CO2 capture, Int. J. Elec. Power 53 (2013) 980e986, https:// doi.org/10.1016/j.ijepes.2013.06.016. [6] J. R. Selman, Research, development, and demonstration of molten carbonate fuel cell systems, in: L. J. M. J. Blomen. M. N. Mugerwa (eds.), Fuel Cell Systems, Ch 9 Springer, 1993, Massachusetts 345-463. Doi: 10.1007/9781e4899e2424e2427 [7] J.D. Doyon, T. Gilbert, G. Davies, L. Paetsch, NiO solubility in mixed alkali/ alkaline earth carbonates, J. Electrochem. Soc. 134 (1987) 3035e3038, https:// doi.org/10.1149/1.2100335. [8] K. Ota, S. Mitsushima, S. Kato, S. Asano, H. Yoshitake, N. Kamiya, Solubilities of nickel- oxide in molten carbonate, J. Electrochem.Soc. 139 (1980) 667e671, https://doi.org/10.1149/1.2069282. [9] S. Yoshioka, H. Urushibata, Simulation of cathode dissolution and shorting for molten carbonate fuel cells, J. Electrochem.Soc. 144 (1997) 815e822, https:// doi.org/10.1149/1.1837494. [10] H.R. Kunz, J.W. Pandolfo, Effects of nickel oxide cathode dissolution on molten carbonate fuel cell life, J.Electrochem. Soc. 139 (1992) 1549e1556, https:// doi.org/10.1149/1.2069453. [11] A.F. Sammells, S.B. Nicholson, P.G.P. Ang, Development of sulfur-tolerant components for the molten carbonate fuel cell, J. Electrochem. Soc. 127 (1980) 350e357, https://doi.org/10.1149/1.2129669. [12] H. Devianto, S.P. Yoon, S.W. Nam, J. Han, T.-H. Lim, The effect of a ceria coating on the H2S tolerance of a molten carbonate fuel cell, J. Power Sources 159 (2006) 1147e1152, https://doi.org/10.1016/j.jpowsour.2005.11.092. [13] H. Devianto, E. Simonetti, S.J. McPhail, F. Zaza, V. Cigolotti, C. Paoletti, A. Moreno, A. La Barbera, I. Luisetto, Electrochemical impedance study of the poisoning behaviour of Ni-based anodes at low concentrations of H2S in an MCFC, Int. J. Hydrogen Energy 37 (2012) 19312e19318, https://doi.org/ 10.1016/j.ijhydene.2012.03.110. [14] A. Wijayasinghe, B. Bergman, C. Lagergren, LiFeO2 LiCoO2 NiO cathodes for molten carbonate fuel cells, J. Electrochem. Soc. 150 (2003) A558eA564, https://doi.org/10.1149/1.1562592. [15] M.E. Orazem, B. Tribollet, Electrochemical Impedance Spectroscopy, Wiley, Hoboken, N.J, 2008, https://doi.org/10.1002/9781119363682. [16] J.R. Selman, Y.P. Lin, Application of ac impedance in fuel cell research and development, Electrochim. Acta 38 (1993) 2063e2073, https://doi.org/ 10.1016/0013-4686(93)80341-V. [17] T. Nishina, M. Takahashi, I. Uchida, Gas electrode reactions in molten carbonate media IV. Electrode kinetics and mechanism of hydrogen oxidation in (Liþ K)CO3 eutectic, J. Electrochem. Soc. 137 (1990) 1112e1121, https:// doi.org/10.1149/1.2086612.
352
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
[18] C.-G. Lee, Influence of temperature on the anode reaction in a molten carbonate fuel cell, J. Electroanal. Chem. 785 (2017) 152e158, https://doi.org/ 10.1016/j.jelechem.2016.12.032. [19] C.-G. Lee, Analysis of impedance in a molten carbonate fuel cell, J. Electroanal. Chem. 776 (2016) 162e169, https://doi.org/10.1016/j.jelechem.2016.07.005. [20] B.A. Boukamp, A Nonlinear Least Squares Fit procedure for analysis of immittance data of electrochemical systems, Solid State Ionics 20 (1986) 31e44, https://doi.org/10.1016/0167-2738(86)90031-7. e, Evaluation and modeling of the [21] A. Leonide, V. Sonn, A. Weber, E. Ivers-Tiffe cell resistance in anode-supported solid oxide fuel cells, J. Electrochem. Soc. 155 (2008) B36eB41, https://doi.org/10.1149/1.2801372. e, Deconvolution [22] H. Schichlein, A.C. Müller, M. Voigts, A. Krügel, E. Ivers-Tiffe of electrochemical impedance spectra for the identification of electrode reaction mechanisms in solid oxide fuel cells, J. Appl. Electrochem. 32 (2002) 875e882, https://doi.org/10.1023/A:1020599525160. ~ oz, D. Pumiglia, S.J. McPhail, D. Montinaro, G. Comodi, [23] C. Boigues Mun G. Santori, M. Carlini, F. Polonara, More accurate macro-models of solid oxide fuel cells through electrochemical and microstructural parameter estimation Part I: Experimentation, J. Power Sources 294 (2015) 658e668, https:// doi.org/10.1016/j.jpowsour.2015.06.118. e, Studies on [24] J.P. Schmidt, T. Chrobak, M. Ender, J. Illig, D. Klotz, E. Ivers-Tiffe LiFePO4 as cathode material using impedance spectroscopy, J. Power Sources 196 (2011) 5342e5348, https://doi.org/10.1016/j.jpowsour.2010.09.121. [25] Y. Kikuchi, J. Matsuda, Y. Tachikawaa, Y. Shiratoria, S. Taniguchib, K. Sasaki, Degradation of SOFCs by various impurities: impedance spectroscopy and microstructural analysis, ECS Trans. 78 (2017) 1253e1260, https://doi.org/ 10.1149/07801.1253ecst. ~ oz, D. Pumiglia, F. Santoni, S.J. McPhail, G. Comodia, M. Carlini, [26] C. Boigues-Mun Performance degradation prediction of a low-temperature SOFC via impedance spectroscopy and CFD modelling, ECS Trans. 68 (2015) 2227e2235, https://doi.org/10.1149/06801.2227ecst. [27] M. Della Pietra, S.J. McPhail, S. Prabhakar, U. Desideri, S.W. Nam, V. Cigolotti, Accelerated test for MCFC button cells: first findings, Int. J. Hydrogen Energy 41 (2016) 18807e18814, https://doi.org/10.1016/j.ijhydene.2016.07.021. e, Hydrogen oxidation kinetics in [28] A. Kromp, A. Leonide, A. Weber, E. Ivers-Tiffe reformate-fuelled anode supported SOFC, ECS Trans. 35 (2011) 665e678, https://doi.org/10.1149/1.3570045. e, A. Weber, Evaluation of electrochemical impedance spectra by [29] E. Ivers-Tiffe the distribution of relaxation times, J. Ceram. Soc. Jpn. 125 (2017) 193e201, https://doi.org/10.2109/jcersj2.16267. [30] J. Weese, A reliable and fast method for the solution of Fredholm integral equations of the first kind based on Tikhonov regularization, Comput. Phys. Commun. 69 (1992) 99e111, https://doi.org/10.16/0010-4655(92)90132-I. [31] A.N. Tikhonov, A.V. Goncharsky, V.V. Stepanov, A.G. Yagola, Numerical Methods for the Solution of Ill-Posed Problems, Kluwer Academic Publishers, 1995, https://doi.org/10.1007/978-94-015-8480-7. [32] C.W. Groetsch, The Theory of Tikhonov Regularization for Fredholm Equations
for the First Kind, Pitman, London, (1984) ISBN: 0-273-08642-1 [33] B.A. Boukamp, Electrochemical impedance spectroscopy in solid state ionics; recent advances, Solid State Ionics 169 (2004) 65e73, https://doi.org/10.1016/ j.ssi.2003.07.002. [34] B.A. Boukamp, A linear Kronig-Kramers transform test for immittance data validation, J. Electrochem. Soc. 142 (1995) 1885e1894, https://doi.org/ 10.1149/1.2044210. [35] W.G. Bessler, Gas concentration impedance of solid oxide fuel cell anodes I. Stagnation point flow geometry, J. Electrochem. Soc. 153 (2006) A1492eA1504, https://doi.org/10.1149/1.2205150. e, Experimental and modeling [36] S. Gewiesa, W.G. Bessler, V. Sonnb, E. Ivers-Tiffe study of the impedance of Ni/YSZ cermet anodes, ECS Trans. 7 (2007) 1573e1582, https://doi.org/10.1149/1.2729264. [37] W.M. Vogel, L.J. Bregoli, S.W. Smith, Electrochemical oxidation of H2 and CO in fused alkali metal carbonates, J. Electrochem. Soc. 127 (1980) 833e839, https://doi.org/10.1149/1.2129766. [38] C.Y. Yuh, J.R. Selman, The polarization of molten carbonate fuel cell electrodes: II. Characterization by AC impedance and response to current interruption, J. Electrochem. Soc. 138 (1991) 3649e3656, https://doi.org/10.1149/ 1.2085474. [39] C.-G. Lee, J.-Y. Hwang, M. Oh, D.-H. Kim, H.C. Li, Overpotential analysis with various anode gas compositions in a molten carbonate fuel cell, J. Power Sources 179 (2008) 467e473, https://doi.org/10.1016/ j.jpowsour.2007.12.125. [40] C.G. Lee, B.S. Kang, H.K. Seo, H.C. Lim, Effect of gas-phase transport in molten carbonate fuel cell, J. Electroanal. Chem. 540 (2003) 169e188, https://doi.org/ 10.1016/S0022-0728(02)01304-9. [41] B. Huang, X. Ye, S. Wang, Q. Yu, H. Nie, Q. Hu, J. Shi, K. Hu, T. Wen, Electrochemical performance of Y2O3/NiO cathode in the molten Li0.62/K0.38 carbonates eutectics, Mater. Res. Bull. 41 (2006) 1935e1948, https://doi.org/ 10.1016/j.materresbull.2006.03.007. [42] L. Hu, I. Rexed, G. Lindbergh, C. Lagergren, Electrochemical performance of reversible molten carbonate fuel cells, Int. J. Hydrogen Energy 39 (2014) 12323e12329, https://doi.org/10.1016/j.ijhydene.2014.02.144. [43] A.J. Appleby, S.B. Nicolson, Reduction of oxygen in Lithium-Potassium carbonate melt, J. Electroanal. Chem. 112 (1980) 71e76, https://doi.org/10.1016/ S0022-0728(80)80008-8. [44] K. Czelej, K. Cwieka, J.C. Colmenares, K.J. Kurzydlowski, Catalytic activity of NiO cathode in molten carbonate fuel cells, Appl. Catal., B 222 (2018) 73e75, https://doi.org/10.1016/j.apcatb.2017.10.003. [45] K. Czelej, K. Cwieka, J.C. Colmenares, K.J. Kurzydlowski, Atomistic insight into the electrode reaction mechanism of the cathode in molten carbonate fuel cells, J. Mater. Chem. 5 (26) (2017) 13763e13768, https://doi.org/10.1039/ c7ta02011b. [46] R.J.B. Smith, M. Loganathan, M.S. Shantha, A review of the water gas shift reaction kinetics, Int. J. Chem. React. Eng. 8 (2010), https://doi.org/10.2202/ 1542-6580.2238. Review R4.
Contents lists available at ScienceDirect
Electrochimica Acta journal homepage: www.elsevier.com/locate/electacta
Accurate in-operando study of molten carbonate fuel cell degradation processes -part I: Physiochemical processes individuation ~ oz a, S.J. McPhail a, F. Santoni a, b, *, M. Della Pietra a, D. Pumiglia a, C. Boigues Mun V. Cigolotti c, S.W. Nam d, M.G. Kang d, S.P. Yoon d a
DTE-PCU-SPCT, ENEA C.R. Casaccia, Via Anguillarese 301, 00123, Rome, Italy Department of Science and Technology, DIST, University Parthenope, Centro Direzionale Isola C4, Naples, 80143, Italy ENEA, Italian National Agency for New Technologies, Energy and the Environment, Portici Research Centre, Piazzale Enrico Fermi 1, 80055, Naples, Italy d Fuel Cell Research Center, KIST - Korea Institute of Science and Technology, 5, Hwarang-ro 14-gil, Seongbuk-gu, Seoul, South Korea b c
a r t i c l e i n f o
a b s t r a c t
Article history: Received 26 February 2018 Received in revised form 2 August 2018 Accepted 14 August 2018 Available online 21 September 2018
This work has the difficult task to deeply study the electrochemical processes that occur inside a 100 cm2 of Molten Carbonate Fuel Cells (MCFC) impedance spectra using the high resolution of Distribution of Relaxation Time (DRT) method. Using this method, it is possible to shed light on the different physicochemical processes occurring within these cells, identifying the characteristic relaxation times by means of an appropriate experimental campaign where temperature and gas compositions in anode and cathode were varied one at a time. The quality of the recorded spectra was verified by Kramers-Kronig relation before applying DRT calculations. In this work, five distinct and separated peaks with different time constants ranging from 0.01 to 500 Hz were identified and associated with physiochemical processes of the cell. Three peaks at high frequency represent the charge transfer processes in anode and cathode active sites. The other two, located at low frequency, are associated with the gas diffusion in the electrodes and to the gas conversion process. This study represents the first application of the DRT approach to this technology allowing to understand the physicochemical origin of the individual polarization processes controlling the cell performance and the degradation. The analysis of degradation processes using the DRT method and the physiochemical processes identification presented in this paper will be shown in part II of this work. © 2018 Elsevier Ltd. All rights reserved.
Keywords: Distribution of relaxation time Molten carbonate fuel cell Polarization losses Deconvolution of impedance spectra Degradation processes
1. Introduction Molten carbonate fuel cells (MCFCs) are promising power generation devices enabling high-efficiency co-generation of electricity and heat with minimal environmental impact [1]. The main advantage in the development and the use of MCFC technology on a large scale is the exploitation of its operational flexibility. In fact, the high operating temperature (approximately 923 K) allows MCFCs to use CO (which is a contaminant for low temperature fuel cells) and low weight hydrocarbons as fuels. These fuels can be converted into hydrogen via water gas shift (WGS) and steam reforming (SR) reactions respectively, and this can take place directly in the cells (internal reforming) or in a separate reformer (external reforming) using the heat generated by the operating stack(s) [2,3].
* Corresponding author. DTE-PCU-SPCT, ENEA C.R. Casaccia, Via Anguillarese 301, 00123 Rome, Italy. E-mail address: [email protected] (F. Santoni). https://doi.org/10.1016/j.electacta.2018.08.100 0013-4686/© 2018 Elsevier Ltd. All rights reserved.
In addition, this technology can be exploited as CO2 concentrator thanks to its intrinsic operating mechanism. In fact, the electrochemical reactions taking place within the MCFC involve the migration of carbonate ions from the cathode to the anode, that corresponds to a depletion of CO2 in the oxidant fed gas and the subsequent enrichment of CO2 in the anode exhaust gas. During this process, the MCFC has the unique ability to operate as an integrated CO2 concentrator and power generation unit, allowing to integrate it with conventional power generation in combination with carbon capture and storage (CCS) technology for the removal of the CO2 from combustion flue gas [4,5]. These practical advantages explain the power generation market interest in MCFCs, but being a commercially embryonic technology compared to the conventional power generation systems, the resources needed to bring this technology into a commercial breakthrough necessary for a fast market penetration are still significant. Primarily, from an economic point of view, the costs of fuel cells per kW is still considerably higher than conventional power
344
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
plants. In addition, an important requirement for commercial applications is the long-term stability. In fact, for stationary applications, agreed target lifespan is generally more than 40,000 h. The two main causes of performance degradation throughout cell lifetime are the loss of electrolyte and dissolution of the NiO in the cathode. Initially, the electrolyte is consumed by corrosion reactions with metal hardware in the fuel cell. Over a longer time, and at a more constant rate, the electrolyte is primarily lost by vaporization into the fuel gas. Although a careful electrolyte management with improved pore structure to maintain electrolyte within the electrodes has reduced the electrolyte loss due to vaporization, this mechanism remains the main cause of performance degradation in atmospheric systems [6]. The cathode is made of NiO, which is slightly soluble in the electrolyte (Eq. (1)). Dissolved nickel ions diffuse into the electrolyte, and precipitate as metallic nickel forming electronically conducting pathways between the anode and cathode.
NiO þ CO2 /Ni2þ þ CO2 3
[1]
The solubility of NiO is managed by a combination of electrolyte composition, gas composition, pressure and operating temperature in the electrolyte. It was early identified as a major obstacle to an increased lifetime of the fuel cell, as it threatens the integrity of the fuel cell by causing short-circuiting [7e10]. The identification of these degradation mechanisms is possible with a post-mortem analysis, underlining the difficulty to recognize them in-operando. Although many studies have been conducted to reduce the degradation mechanisms that lead to cell failure [11e14], progress for a further performance improvement is partly constrained by an incomplete understanding of the physicochemical origin of the individual polarization processes controlling the cell performance. Attending to these considerations, the scientific and industrial communities should work hand-in-hand to enhance the production of a high-performing, and robust products, developing advanced analysis tools and techniques to fully understand the degradation phenomena. In this scenario, Electrochemical Impedance Spectroscopy (EIS) is proposed as a powerful tool to assess the MCFC performance [15]. EIS can be recorded without altering the working conditions of the fuel cell and can thus be used to determine the impact of different parameters (e.g. reactants stoichiometry, current density, temperature) during cell operation. In literature, many works exploit this task, from the pioneering works of Selman et al. [16] or Nishina et al. [17] until to the recent works of C. -G. Lee [18,19], demonstrating the validity of this technique as an in-situ method for characterizing rate-limiting processes in MCFCs porous electrodes. Even though EIS is a very sensitive technique, it can hardly distinguish two processes that differ in their characteristic frequency by less than two decades, thus resulting deeply convoluted in the EIS spectrum and consequently in a not clear image of the real electrochemical processes that occurring the cell. Impedance spectra can be interpreted using an equivalent circuit model (ECM) [20]. In this approach, physical phenomena, taking place in the fuel cell during operation, are represented by equivalent electrical elements. The value of such elements is obtained by the fitting of the measured impedance data by means of mathematical algorithms. An appropriate use of ECM to fit impedance data relies on a priori knowledge about the system since it requires a suitable model structure that represents the physical processes occurring in the cell. Conversely, if there is a little, or no a priori knowledge about the system, the choice of an appropriate
model structure can become a difficult task, leading often to a severe ambiguity of the adopted model [15]. In order to overcome this issue, an alternative approach for analysing impedance spectra is the Distribution of Relaxation Times (DRT) method. Originally introduced for the impedance analysis of solid oxide fuel cells [21e23] and lithium-ion batteries [24], DRT can be used for process identification in complex electrochemical systems, where resistive-capacitive (R//C) features are dominant [22]. The individual processes are separated on the basis of their typical time constants, derived from the associated (R//C) elements, allowing to bypass ambiguous EC modelling and to analyse spectra without pre-assumptions. Moreover, DRT analysis is a useful instrument to investigate the degradation phenomena during a long-term test. In fact, as demonstrated in different studies on SOFCs [25,26], the use of DRT analysis combined with a valid ECM gives the possibility to observe qualitatively and quantitatively the behaviour of individual polarization processes in operando. In this field, accelerated lifetime tests can become a powerful tool to reduce significantly the experimental efforts for long-term tests [27]. A direct application of the existing monitoring methodologies is unfeasible as the lifetime and degradation rate of a fuel cell are governed by several concurring processes. A deconvolution of the different degradation mechanisms by EIS coupled with DRT can provide the required information about the different ageing phenomena in the cell. The global aim of this work is to analyse the degradation phenomena in 100 cm2 class MCFCs, during an endurance test. The chosen tools are the EIS coupled with DRT method allowing to monitor the individual polarization processes controlling the cell performance in operando, and microstructural characterization techniques used for pre- and post-mortem analysis. Since no other study is present in the literature about DRT applied to MCFCs, an identification of the single processes was necessary. For this reason, in this paper is reported the first part of the work, regarding the first findings on the deconvolution of MCFC impedance spectra and the identification of t the individual polarization processes using an appropriate experimental campaign where temperature and gas compositions in anode and cathode were varied one at a time. Whereas the results obtained from the endurance tests will be discussed in the second part. 2. Experimental 2.1. Equipment setup The experimental data produced to support this activity were carried out at the Fuel Cell Research Center laboratories of KIST in Seoul, South Korea, using planar MCFC single cells with an active area of 100 cm2. The characteristics of the single cell components used are summarized in Table 1. The electrolyte layers are alternated with the matrix layers and then assembled between the anode and cathode. The obtained fuel cell is sandwiched between the two current collectors and included in the anodic and cathodic frames (see Fig. 1a), where fuel and oxidant stream were in concurrent flow. The cell was held in a heating block under a mechanic load of 2 kgcm2, obtained with a hydraulic piston as shown in Fig. 1b. The anode and cathode gas compositions, fed to the cells, were measured and controlled by Brooks 5850E Digital Mass Flow Controllers and a thermo-controlled bubbler is used to adjust the required water content in the fuel. The temperature of the anodic inlet pipeline is controlled through heating tapes from the water injection point up to the
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
345
Table 1 Specifications of 100 cm2class MCFC components. Single cell components
Specification
Cell frame
Size [cm x cm] Material Size [cm x cm] Thickness [mm] Current collector Material Porosity Fuel gas (mol ratio) Total flow rate [ml/min] Size [cm x cm] Thickness [mm] Current collector Material Porosity Oxidant gas (mol ratio) Total flow rate [ml/min] Material Mole ratio Material Thickness [mm]
Anode
Cathode
Electrolyte Matrix
13 13 AISI 316L 11 11 0.7 Pure Ni Ni þ 5%wt Al 55e60% Air: CO2 ¼ 0.7:0.3 951 10 10 0.7 AISI 316L Lithiated NiO 60e65% H2: CO2: H2O ¼ 0.72:0.18:0.1 396 Li2CO3/K2CO3 68:32 g-LiAlO2 1.2
Fig. 1. Experimental set-up for the MCFC single cell: (a) Schematic illustration of the single cell assembling (b) Single cell apparatus pressurized inside the oven, and (c) single cell MCFC test equipment.
furnace entry, in order to prevent water condensation inside the inlet pipes and pre-heat the gas mixture. Before the single cell start-up, a pre-treatment was carried out to remove the organic binder by thermal decomposition for 4 days in air at the temperature ranging from 298 K to 723 K and then for 3 days in CO2 at the temperature ranging from 723 K to 923 K. Under the latter conditions, which were very critical for the electrolyte melting region, CO2 was passed through the system at a low flow rate to maintain the distribution of the electrolyte throughout the pores of the matrix, of the cathode, and of the anode. Moreover, this procedure prevented evaporation of the electrolyte. After pre-
treatment, the temperature of the gas in the MCFC was maintained at 923 K for 180 h, using KIST normal gas condition reported in Table 1. The temperature in the MCFC was maintained at the reference temperature of 923 K, or at a different one required for a given test, thanks to the furnace and the heating plates holding the single cell, so that the cell plane temperature could be considered uniform. An electrical loader (ELS300Z, ELTO DC Electronics Co.) was used to impose the external current and to collect the cell voltage; while the nitrogen crossover at the anode outlet was monitored by using gas chromatography (789 A, Agilent Tech.)
346
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
Electrochemical impedance spectroscopy (EIS) measurements were carried out to analyse the frequency response of the cells tested, within the frequency range 10 kHz - 0.01 Hz, using a frequency response analyser (FRA 1255B, Solartron Co.) coupled with an electrochemical dielectric interface (EI, 1287 Solartron Co.). All impedance measurements performed within this work were carried out under open-circuit conditions. ZView® software for Windows from Scribner Associates Inc. was used in order to fit the empirical data and to calculate the values of resistances associated with each physiochemical process.
2.2. Distribution of relaxation times (DRT) method: deconvolution of signals MCFCs are complex electrochemical devices, in which many processes occur at the same time, each process being characterized by its typical relaxation time. When two (or more) processes have a similar relaxation time, it is hard to separate and distinguish their contributions to the overall impedance of the cell, since the EIS spectra can clearly distinguish processes separated at least by two decades in the frequency domain. Furthermore, it is hard to identify small contributions, these being totally covered by processes with large polarization losses. An instrumental tool to overcome this issue may be the analysis of the DRT functions [21,23,28]. The DRT-method uses the fact, that every impedance function that obeys to the Kramers-Kronig relations can be represented as a differential sum of infinitesimal small RC-elements [22]. This RC circuit is characterized by its time constant t and can be associated with an electrochemical process. The relation between the impedance spectrum Z(u) and the distribution of relaxation times t, which is for sake of simplicity often displayed as a distribution g(t) vs. the relaxation frequencies, is given by Eq. (2):
ZðuÞ ¼ R0 þ Zpol ðuÞ ¼ R0 þ Rpol
∞ ð
0
gðtÞ dt 1 þ jut
[2]
where Zpol(u) represents the overall polarization resistance of the fuel cell and R0 the purely ohmic one. The time constant of a single RC-element is designated by t ¼ RC, the fraction of the overall polarization resistance with relaxation times between t and t þ dt gðtÞ is indicated by the term. 1þj ut
This relation is valid for any kind of system that fulfils the fundamental demands of impedance spectroscopy linearity, causality and time invariance, and is applied a discrete Fourier transformation to compute the DRT from the imaginary part of Z(u). An appropriate extrapolation and filtering of the data are required to suppress numerical error amplification. Thus, the method becomes unsuitable for impedance spectra exhibiting greater errors or inductive artefacts at higher frequencies [29]. The mathematical problem with this approach arises from the inversion of Eq. (2), which is necessary in order to extract g(t) from the measured impedance data Zpol. This problem is known to be an ill-posed problem and requires special methods to be solved in order to avoid false peaks and oscillations. Even if there are numerous methods which can solve this kind of problem [30,31], the most renowned and reliable one is the Tikhonov regularization algorithm which uses a self-consistent regularization parameter l, which determines the smoothness of the DRT. In the case of excellent data quality or synthetic data; a small regularization parameter can be used, for noisy data the regularization parameter has to be increased to suppress false peaks and oscillations in the DRT. Further details about the
regularization can be found in Tikhonov and Groetsch works [31,32]. The usage of this tool gives access to information that until now, for the limited resolution of the Impedance Spectra, was not possible to obtain. In fact, as shown in Fig. 2, from the comparison between the impedance spectrum and the correspondent DRT, it is immediately visible that the raw EIS response has a lower resolution, in fact as just mentioned above it is difficult to clearly visualize the contributions of single physicochemical processes to the total polarization resistance. Moreover, the information obtainable from the analysis of EIS spectra provides a quantitative measurement of the polarization losses, but nothing can be said, for instance, about the nature of the process and where it takes place, anode or cathode. This impedes a profound and complete comprehension of the evolution of these processes and their dependency to the operating conditions of the cell. On the other hand, it is evident that the higher resolution of the DRT allows a clear identification of 5 peaks/ processes with a characteristic frequency domain. The nature of these peaks has been assessed by means of a systematic experimental campaign to investigate the parameter dependence. This allows identifying all the processes, ascribing them to the anode or cathode. For the DRT calculation, an in-house software running in MATLAB® has been used. The DRT method becomes unsuitable for impedance spectra exhibiting greater errors or inductive artefacts at higher frequencies. For this reason, the measurement data quality is crucial. The quality and amount of information that can be extracted from impedance data are implicitly connected to the noise-level and the compliance of the measured curve with the principles of causality, linearity, and stability. A well-established method used to assess the consistency and quality of measured impedance spectra is the Kramers-Kronig validation [15]. In this work, the Kramers-Kronig transformation rules were applied to every impedance data set measured throughout our experimental activity by using the “KK test for Windows” software [33,34]. 2.3. Experimental procedure In order to elucidate the nature of each peak obtained with the DRT method, the cells were operated under a number of different predefined conditions, starting from a reference condition (indicated in bold in the first line) and varying each time one of the following parameters: temperature, anodic gas composition and cathodic gas composition reported in Table 2. For each condition, an
Fig. 2. Imaginary component of the cell impedance and the DRT distribution function (l ¼ 0.1) when the MCFC cell is operating at 923 K under normal KIST conditions.
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
347
Table 2 Test matrix containing the molar fractions of the gas species in anode and cathode used for the DRT analysis. Anode Gas
Anodic pH2 variations (atm)
Anodic pH2O variations (atm)
Anodic pCO variations (atm)
Cathodic pO2 variations (atm)
Cathodic pCO2 variations (atm)
Temperature variations (K)
Cathode Gas
Test
pH2
pCO2
pH2O
pN2
pCO
pO2
pN2
pCO2
I II III IV V I II III IV V I II III IV V I II III IV I II III IV V VI I-853 II-873 III-893 IV-923 V-953
0.72 0.50 0.20 0.10 0.05 0.5 0.5 0.5 0.5 0.5 0.1 0.1 0.1 0.1 0.1 0.72 0.72 0.72 0.72 0.72 0.72 0.72 0.72 0.72 0.72 0.72 0.72 0.72 0.72 0.72
0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18
0.10 0.10 0.10 0.10 0.10 0 0.10 0.15 0.20 0.32 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10
0 0.22 0.52 0.62 0.67 0.32 0.22 0.17 0.12 0 0.22 0.32 0.42 0.52 0.57 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0.40 0.30 0.20 0.10 0.05 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.09 0.06 0.03 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15
0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.61 0.64 0.67 0.45 0.55 0.62 0.70 0.75 0.80 0.55 0.55 0.55 0.55 0.55
0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.30 0.40 0.30 0.23 0.15 0.10 0.05 0.30 0.30 0.30 0.30 0.30
EIS measurement was recorded, and the DRT function was calculated from the experimental data acquired. The anode and cathode gas flow rates are maintained at a constant value of 396 ml/min and 951 ml/min respectively during all experiments. 3. Results and discussion 3.1. Validity of measurement data As introduced in Section 2.2, application of DRT necessitates an outstanding measurement quality. Therefore, every measured spectrum presented in the following evaluation has been validated by Kramers-Kronig relations before applying DRT calculations. Fig. 3a and b shows a KramerseKronig validation for a typical impedance spectrum obtained during the experimental campaign and the corresponding Kramers-Kronig residuals. The residuals are less than 0.5% of the whole frequency range. These considerably small residuals are taken as a confirmation of Kramers-Kronig validity indicating that the experimental data impeccably agree with the expected theoretical results. 3.2. Preliminary electrochemical performance comparison The experimental procedure employed for the identification of the different operating processes was carried out on three distinct MCFC single cells; however, the tested cells, described in Table 1, came from the same batch, thus ensuring (as will be seen) a substantial homogeneity. Nevertheless, a preliminary performance characterization was performed on each cell in order to confirm the comparability of the tested samples and of the results obtained from the subsequent experiments. The comparison was made using the normal KIST condition reporting in Table 1 at the temperature of 923 K. The polarization curve reaching a utilization factor of 0.4 both for anode
and cathode, corresponding to 150 mAcm2 of current density. As depicted in Fig. 4, the polarization curves (Fig. 4a) and EIS spectra (Fig. 4b) show that the overall performances of the tested cells are nearly identical. This similarity on the performance is shown also in the corresponding DRT spectra (Fig. 4c) showing the same peaks in the same position, concluding that the tested cells can be considered comparable for the objectives of the experimental campaign in which they have been employed. 3.3. Processes individuation Fig. 5a shows a series of impedance spectra recorded under different fuel compositions, where the partial pressure of hydrogen was varied between 0.72 and 0.05 atm. In Nyquist representation, the impedance spectra show two well-separated arcs. The low-frequency arc indicates a pronounced increase in resistance with decreasing of the hydrogen quantity while a lessmarked dependence is visible for the higher frequencies of the impedance. The DRT calculated from this spectrum is shown in Fig. 5b. Obviously, the visual inspection of the DRT gives considerably more detailed insight into the system than the raw impedance plot. Five distinct and separated peaks with different time constants ranging from 0.01 to 500 Hz are identified by the DRTs. In accordance with the EIS spectrum, only a slight dependence is shown for the peaks located at high frequency, P1an and P3cat, whereas the most influenced peaks are located at low frequencies, P4diff and P5an. The two low frequency processes are the imprints of concentration polarization effects [35,36] and could be considered the deconvolution of the low frequency circle associated to the anodic gas-phase mass transfer processes as has already been suggested by Vogel and Bregoli [37], Yuh and Selman [38], and Lee [39]. As evident in Fig. 5b, these two peaks present an inverse
348
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
Fig. 3. (a) Nyquist plot recorded at 923 K reporting the comparison between the measured EIS obtained for the third composition of pCO2 variation, described in Table IV, and the fitting result obtained with the KKtest (b) computational residuals of Kramers-Kronig reconstruction as a function of frequency.
Fig. 4. Polarization curves (a) EIS spectra (b) and DRT spectra(c) of single cell tested employed for the process characterization experiments.
behaviour in relation to the stepwise hydrogen increment. The behaviour of P4diff, (decreases with the increment of the hydrogen) probably reflects the gas-phase diffusion at the anode side. Whereas the P5an seems to be associated with a gas conversion process due to the water gas shift reaction (WGS) represented in Eq. (3). In fact, the increment of the H2 concentration in the fuel (product of the WGS reaction) inhibits the reaction, provoking the increment of the resistance associated with this peak.
CO þ H2 O#CO2 þ H2
[3]
The analysis of the H2O partial pressure variations and the CO partial pressure variations are represented in Fig. 5c and d respectively. For CO variations was used also a small amount of H2 (0.10 atm) to avoid an excessive stress on the cell, as reported in Table 2. From both the graphs, it is clearly visible how the processes
P4diff and P5an decrease with the increment of H2O and CO, whereas P2cat does not follow a linear behaviour or is not influenced by these anodic compositions. About the other high frequency peaks, it is important to note that P3cat shows only a minimal dependence from the H2 variations respect to P1an (Fig. 5b). Moreover, it does not present any linear dependence from CO and H2O variations (Fig. 5c and d), while. P1an presents a slight dependence also from the H2O variations and its behaviour is very similar, both for intensity and movement of the peak, to those obtained under H2 variations. The reduction of P5an due to the increment of CO and H2O is because they are directly involved as reagents of the WGS reaction, and their increment reduces the resistance associated with this process. These results are coherent with the hypothesis that the P5an is associated to the WGS reaction. As shown in Fig. 5c the anodic diffusion peak P4diff is strictly
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
349
Fig. 5. Nyquist plot of MCFC single cell under stepwise increase of hydrogen partial pressure(a) and DRT spectra for the anodic partial pressure variations: hydrogen variations(b), CO variations (c) and H2O variations (d).
linked to the conversion of CO in H2 by means of the WGS reaction, for this reason, the two peaks are in this case very similar behaviour. The DRT analysis obtained from the cathodic composition variations is reported in Fig. 6a and b. For the O2 partial pressure variations, four different compositions varying the O2 content from 0.15 to 0.03 atm were tested. In this graph, the process P4diff shows a significant dependency on the O2 content increasing with its decrement, while the process P3cat presents the same behaviour but with a minor dependency. The other three processes present in the spectrum (P1an, P2cat, P5an) are not affected or do not present a linear behaviour with the O2 variations. In Fig. 6b, is reported the DRT analysis obtained for the CO2 partial pressure variations. A similar result to O2 variations has been obtained, in fact, the principal process involved in this analysis is
the P4diff. It is relevant to notice that this process is related to the cathodic gas phase diffusion, showing a strong dependence on the cathodic compositions changes. Moreover, in this analysis is shown also a linear dependence of the peak P2cat from the increment of CO2 in the gas mixture and a slight dependence of P3cat that shows a contrary behaviour. Comparing anodic and cathodic composition variation results, it is possible to see in both cases the relevance of the process P4diff. It is possible that P4diff is composed of two processes, respectively related to the cathodic and anodic diffusion. In fact, it has to be taken into account that if the characteristic frequency of two adjacent processes differs of less than a decade in frequency, two peaks may overlap. A possible explanation of the similar frequency of the diffusion processes above mentioned is the comparable thickness and porosity of the two electrodes as reported in Table 1.
Fig. 6. DRT spectra for the cathodic partial pressure variations: O2 variations (a), CO2 variations (b).
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
dependence with the CO2 partial pressure variations and could be associated with the reaction of CO2 with oxide ions to form carbonate (Eq. (5)). 2 O 2 þ e /O2
[4]
2 O2 2 þ 2CO2 þ 2e /2CO3
[5]
In recent works carried out by Czelej et al. [44,45] the mechanism of the electrode reaction at triple phase boundary in MCFC cathode is proposed based on atomic-scale simulations, and the oxygen reduction is shown as the rate determining step. Moreover, in these studies has been elucidated the prominent role of Oterminated octopolar NiO (111) in the cathodic transformation of carbon dioxide to carbonate anion. The results indicate that Eq. (5) on O-terminated octopolar NiO (111) is facile under the MCFC operation conditions and follows Mars-van Krevelen (MvK) and Eley-Rideal (ER) sequential mechanisms. A less-marked dependence is shown for the processes located at low frequency. In particular, the resistance associated with P4diff decreases with the temperature increment. In case of MCFC, the pore size at the electrodes (~9 mm at the cathode and ~3 mm at the anode) are larger than the SOFC’s electrodes (~sub-mm). In general, in the pore diffusion of gas molecules, Knudsen diffusion is domi1
nant in the pore size of below sub-mm, which shows Rgd fT 2 . However, viscous flow is dominant in the large pores of MCFC 1
electrodes, which shows Rgd fT 2 . Therefore, the diffusion of gas molecules through the electrodes is promoted by the increment of the temperature. Concerning the process P5an, it slightly increases when the temperature rises. This is due to the change of the equilibrium constant of the WGS reaction, that decreases from 2,82 at 853 K to 1,60 at 953 K, causing a reduction of the reaction products and consequently an increase of the resistance [46]. An accurate quantitative relation of the WGS to the resistance based on a physically meaningful equivalent circuit model (ECM) built-up on the base of knowledge given by DRT analysis will be presented in the second part of the work. =
Moreover, since the electrodes in the MCFCs are supposed to be covered with thin electrolyte films estimated as much less than 1 mm [40], the P4diff represents not only a mass transfer process through the gas phase but also the one due to the liquid phase. In Fig. 7a has been reported the EIS measurements obtained at five different temperatures from 853 K to 953 K, using the composition reported in Table 2. In accordance with Yuh and Selman [38], the cell temperature has a strong effect on the highfrequency loop, in fact, as the temperature rises the highfrequency arc shrinks, while the low-frequency loop is always quite the same size. In accordance with the EIS spectra, in the related DRT analysis (Fig. 7b) is evident how the processes located at high-frequency P1an, P2cat and P3cat are all characterized by a pronounced thermal activation. Although the gas composition variations in some cases (H2 variations, O2 variations and CO2 variations) influence this high-frequency peaks, it seems noticeable how the temperature is the predominant operating parameter affecting these processes. In fact, the intensity of the peak P3cat, increases of 300% (from approximately 0.06 ohmcm2/s at 853 K to 0.02 ohmcm2/s at 953 K), P2cat increases of 200% (from approximately 0.03 ohmcm2/s at 853 K to 0.06 ohmcm2/s at 953 K) and P1an increases more than 100% (from approximately 0.06 ohmcm2/s at 853 K to 0.13 ohmcm2/ s at 953 K). Their pronounced thermal activation and their characteristic frequency suggest that these peaks are related to the charge transfer mechanisms in the electrodes. P1an., showing also a slight dependence from the amount of H2 present in the fuel, could be associated with the H2 oxidation at the anode, whereas P2cat and P3cat are ascribable at the cathode processes being influenced by the cathodic composition variations. In particular, P3cat is influenced by the partial pressure of oxygen and CO2 (slightly influenced), for this reason, it could be associated with the O2 reduction (Eq. (4)). In fact, in literature, two main possible paths for the O2 evolution are described [41]: peroxide and superoxide paths. As suggested by Hu [42] and Appleby [43], both of them are limited by the diffusion of the oxygen ions and involve a resistance that increases when the CO2 concentration grows or the O2 concentration decreases. Concerning to the process P2cat, it shows a linear
=
350
Fig. 7. Series of impedance spectra (a) and corresponding DRT spectra obtained at different operating temperature from 853 K to 953 K (b).
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
351
Table 3 Resume of the physicochemical processes occurring in the MCFC and their corresponding peaks in the DRT spectra. Process
P1an P2cat P3cat P4diff P5an
Frequency (Hz)
500…100 100…50 50…5 5…0.5 0.5…0.03
Dependency
Electrode
pH2
pH2O
pCO
Temperature
pCO2
pO2
Low no no High High
Low no no High Medium
no no no High High
High High High Medium Low
no Low no High no
no no Low High no
Table 3 gives an overview of the processes identified by the DRT analysis together with their characteristic frequency range, gas partial pressure and temperature dependency.
4. Conclusions In this study, the distribution of relaxation times was applied successfully to analyse impedance spectra of 100 cm2 Molten Carbonate Fuel Cells. To the authors ‘knowledge, this study represents the first application of the DRT approach to the MCFCs. It is evident, that DRT gives considerably more detailed insight into the system than the raw impedance plot. In fact, five distinct and separated peaks with different time constants ranging from 0.01 Hz to 500 Hz are identified. Three distinct MCFC single cells were tested, and a preliminary performance characterization was performed on each cell, in order to confirm the comparability of the tested samples and of the results obtained from the subsequent experiments. To investigate the parameter dependence of each single polarization process, a series of impedance measurements were carried out, changing only one cell parameter at a time (cathodic gas composition variations, anodic gas composition variations and temperature). The spectra obtained were validated by Kramers-Kronig relations before applying DRT calculations, reporting an error of less 0.5% of the whole frequency range. Investigating the parameter dependence, the physical/chemical origins of these processes can unambiguously be determined, as follows: C P1an (500-100 Hz) Anodic charge transfer process: electrodeoxidation of H2. C P2cat (100-50 Hz) Cathodic charge transfer process: carbonate formation. C P3cat (50-5 Hz) Cathodic charge transfer process: O2 reduction. C P4diff (5e0.5 Hz) Anodic and Cathodic Gas-phase and Liquidphase Mass transfer process: Diffusion. C P5an (0.5e0.03 Hz) Anodic Mass transfer process: gas conversion process due to the Water Gas Shift reaction. The possibility to identify the polarization losses based on their respective time constants is an interesting instrument as the base on which generate comprehensive and robust equivalent circuit models (ECM) and develop physical models for the prediction of the performance degradation in MCFC cells. The possibility to identify the individual polarization losses based on their respective time constants using this strategy (EISDRT) gives access to electrochemical information that till now, for the limited resolution of the Impedance Spectra, was not possible to obtain. This strategy gives the base to generate comprehensive and robust equivalent circuit models to obtain a quantitative value of the resistance associated with each process. In the second part of this work, this tool will be applied to study the degradation processes in operando during an endurance test.
Anode Cathode Cathode Anode and Cathode Anode
Acknowledgement This research was supported by the Global Research Laboratory Program through the National Research Foundation of Korea funded by the Ministry of Science, ICT and Future Planning of Republic of Korea (NRF-2009-00406) and by the Korea Institute of Energy Technology Evaluation and Planning (KETEP), funded by the Ministry of Trade, Industry and Energy, (No. 20153030040930). References [1] R. O’Hyare, S.W. Cha, W. Colella, F.B. Prinz, Fuel Cells Fundamentals, third ed., John Wiley & Sons, New York, 2016 https://doi.org/10.1002/9781119191766. [2] L. Barelli, G. Bidini, S. Campanari, G. Discepoli, M. Spinelli, Performance assessment of natural gas and biogas fueled molten carbonate fuel cells in carbon capture configuration, J. Power Sources 320 (2016) 332e342, https:// doi.org/10.1016/j.jpowsour.2016.04.071 0378e7753. [3] V. Chiodo, G. Zafarana, S. Maisano, S. Freni, A. Galvagno, F. Urbani, Molten carbonate fuel cell system fed with biofuels for electricity production, Int. J. Hydrogen Energy 41 (2016) 18815e18821, https://doi.org/10.1016/ j.ijhydene.2016.05.119. [4] G. Discepoli, G. Cinti, U. Desideri, D. Penchini, S. Proietti, Carbon capture with molten carbonate fuel cells: experimental tests and fuel cell performance assessment, Int. J. Greenh. Gas Con 9 (2012) 372e384, https://doi.org/10.1016/ j.ijggc.2012.05.002. [5] M. Brenna, F. Foiadelli, G. Manzolini, Grid connection of MCFC applied to power plant with CO2 capture, Int. J. Elec. Power 53 (2013) 980e986, https:// doi.org/10.1016/j.ijepes.2013.06.016. [6] J. R. Selman, Research, development, and demonstration of molten carbonate fuel cell systems, in: L. J. M. J. Blomen. M. N. Mugerwa (eds.), Fuel Cell Systems, Ch 9 Springer, 1993, Massachusetts 345-463. Doi: 10.1007/9781e4899e2424e2427 [7] J.D. Doyon, T. Gilbert, G. Davies, L. Paetsch, NiO solubility in mixed alkali/ alkaline earth carbonates, J. Electrochem. Soc. 134 (1987) 3035e3038, https:// doi.org/10.1149/1.2100335. [8] K. Ota, S. Mitsushima, S. Kato, S. Asano, H. Yoshitake, N. Kamiya, Solubilities of nickel- oxide in molten carbonate, J. Electrochem.Soc. 139 (1980) 667e671, https://doi.org/10.1149/1.2069282. [9] S. Yoshioka, H. Urushibata, Simulation of cathode dissolution and shorting for molten carbonate fuel cells, J. Electrochem.Soc. 144 (1997) 815e822, https:// doi.org/10.1149/1.1837494. [10] H.R. Kunz, J.W. Pandolfo, Effects of nickel oxide cathode dissolution on molten carbonate fuel cell life, J.Electrochem. Soc. 139 (1992) 1549e1556, https:// doi.org/10.1149/1.2069453. [11] A.F. Sammells, S.B. Nicholson, P.G.P. Ang, Development of sulfur-tolerant components for the molten carbonate fuel cell, J. Electrochem. Soc. 127 (1980) 350e357, https://doi.org/10.1149/1.2129669. [12] H. Devianto, S.P. Yoon, S.W. Nam, J. Han, T.-H. Lim, The effect of a ceria coating on the H2S tolerance of a molten carbonate fuel cell, J. Power Sources 159 (2006) 1147e1152, https://doi.org/10.1016/j.jpowsour.2005.11.092. [13] H. Devianto, E. Simonetti, S.J. McPhail, F. Zaza, V. Cigolotti, C. Paoletti, A. Moreno, A. La Barbera, I. Luisetto, Electrochemical impedance study of the poisoning behaviour of Ni-based anodes at low concentrations of H2S in an MCFC, Int. J. Hydrogen Energy 37 (2012) 19312e19318, https://doi.org/ 10.1016/j.ijhydene.2012.03.110. [14] A. Wijayasinghe, B. Bergman, C. Lagergren, LiFeO2 LiCoO2 NiO cathodes for molten carbonate fuel cells, J. Electrochem. Soc. 150 (2003) A558eA564, https://doi.org/10.1149/1.1562592. [15] M.E. Orazem, B. Tribollet, Electrochemical Impedance Spectroscopy, Wiley, Hoboken, N.J, 2008, https://doi.org/10.1002/9781119363682. [16] J.R. Selman, Y.P. Lin, Application of ac impedance in fuel cell research and development, Electrochim. Acta 38 (1993) 2063e2073, https://doi.org/ 10.1016/0013-4686(93)80341-V. [17] T. Nishina, M. Takahashi, I. Uchida, Gas electrode reactions in molten carbonate media IV. Electrode kinetics and mechanism of hydrogen oxidation in (Liþ K)CO3 eutectic, J. Electrochem. Soc. 137 (1990) 1112e1121, https:// doi.org/10.1149/1.2086612.
352
F. Santoni et al. / Electrochimica Acta 291 (2018) 343e352
[18] C.-G. Lee, Influence of temperature on the anode reaction in a molten carbonate fuel cell, J. Electroanal. Chem. 785 (2017) 152e158, https://doi.org/ 10.1016/j.jelechem.2016.12.032. [19] C.-G. Lee, Analysis of impedance in a molten carbonate fuel cell, J. Electroanal. Chem. 776 (2016) 162e169, https://doi.org/10.1016/j.jelechem.2016.07.005. [20] B.A. Boukamp, A Nonlinear Least Squares Fit procedure for analysis of immittance data of electrochemical systems, Solid State Ionics 20 (1986) 31e44, https://doi.org/10.1016/0167-2738(86)90031-7. e, Evaluation and modeling of the [21] A. Leonide, V. Sonn, A. Weber, E. Ivers-Tiffe cell resistance in anode-supported solid oxide fuel cells, J. Electrochem. Soc. 155 (2008) B36eB41, https://doi.org/10.1149/1.2801372. e, Deconvolution [22] H. Schichlein, A.C. Müller, M. Voigts, A. Krügel, E. Ivers-Tiffe of electrochemical impedance spectra for the identification of electrode reaction mechanisms in solid oxide fuel cells, J. Appl. Electrochem. 32 (2002) 875e882, https://doi.org/10.1023/A:1020599525160. ~ oz, D. Pumiglia, S.J. McPhail, D. Montinaro, G. Comodi, [23] C. Boigues Mun G. Santori, M. Carlini, F. Polonara, More accurate macro-models of solid oxide fuel cells through electrochemical and microstructural parameter estimation Part I: Experimentation, J. Power Sources 294 (2015) 658e668, https:// doi.org/10.1016/j.jpowsour.2015.06.118. e, Studies on [24] J.P. Schmidt, T. Chrobak, M. Ender, J. Illig, D. Klotz, E. Ivers-Tiffe LiFePO4 as cathode material using impedance spectroscopy, J. Power Sources 196 (2011) 5342e5348, https://doi.org/10.1016/j.jpowsour.2010.09.121. [25] Y. Kikuchi, J. Matsuda, Y. Tachikawaa, Y. Shiratoria, S. Taniguchib, K. Sasaki, Degradation of SOFCs by various impurities: impedance spectroscopy and microstructural analysis, ECS Trans. 78 (2017) 1253e1260, https://doi.org/ 10.1149/07801.1253ecst. ~ oz, D. Pumiglia, F. Santoni, S.J. McPhail, G. Comodia, M. Carlini, [26] C. Boigues-Mun Performance degradation prediction of a low-temperature SOFC via impedance spectroscopy and CFD modelling, ECS Trans. 68 (2015) 2227e2235, https://doi.org/10.1149/06801.2227ecst. [27] M. Della Pietra, S.J. McPhail, S. Prabhakar, U. Desideri, S.W. Nam, V. Cigolotti, Accelerated test for MCFC button cells: first findings, Int. J. Hydrogen Energy 41 (2016) 18807e18814, https://doi.org/10.1016/j.ijhydene.2016.07.021. e, Hydrogen oxidation kinetics in [28] A. Kromp, A. Leonide, A. Weber, E. Ivers-Tiffe reformate-fuelled anode supported SOFC, ECS Trans. 35 (2011) 665e678, https://doi.org/10.1149/1.3570045. e, A. Weber, Evaluation of electrochemical impedance spectra by [29] E. Ivers-Tiffe the distribution of relaxation times, J. Ceram. Soc. Jpn. 125 (2017) 193e201, https://doi.org/10.2109/jcersj2.16267. [30] J. Weese, A reliable and fast method for the solution of Fredholm integral equations of the first kind based on Tikhonov regularization, Comput. Phys. Commun. 69 (1992) 99e111, https://doi.org/10.16/0010-4655(92)90132-I. [31] A.N. Tikhonov, A.V. Goncharsky, V.V. Stepanov, A.G. Yagola, Numerical Methods for the Solution of Ill-Posed Problems, Kluwer Academic Publishers, 1995, https://doi.org/10.1007/978-94-015-8480-7. [32] C.W. Groetsch, The Theory of Tikhonov Regularization for Fredholm Equations
for the First Kind, Pitman, London, (1984) ISBN: 0-273-08642-1 [33] B.A. Boukamp, Electrochemical impedance spectroscopy in solid state ionics; recent advances, Solid State Ionics 169 (2004) 65e73, https://doi.org/10.1016/ j.ssi.2003.07.002. [34] B.A. Boukamp, A linear Kronig-Kramers transform test for immittance data validation, J. Electrochem. Soc. 142 (1995) 1885e1894, https://doi.org/ 10.1149/1.2044210. [35] W.G. Bessler, Gas concentration impedance of solid oxide fuel cell anodes I. Stagnation point flow geometry, J. Electrochem. Soc. 153 (2006) A1492eA1504, https://doi.org/10.1149/1.2205150. e, Experimental and modeling [36] S. Gewiesa, W.G. Bessler, V. Sonnb, E. Ivers-Tiffe study of the impedance of Ni/YSZ cermet anodes, ECS Trans. 7 (2007) 1573e1582, https://doi.org/10.1149/1.2729264. [37] W.M. Vogel, L.J. Bregoli, S.W. Smith, Electrochemical oxidation of H2 and CO in fused alkali metal carbonates, J. Electrochem. Soc. 127 (1980) 833e839, https://doi.org/10.1149/1.2129766. [38] C.Y. Yuh, J.R. Selman, The polarization of molten carbonate fuel cell electrodes: II. Characterization by AC impedance and response to current interruption, J. Electrochem. Soc. 138 (1991) 3649e3656, https://doi.org/10.1149/ 1.2085474. [39] C.-G. Lee, J.-Y. Hwang, M. Oh, D.-H. Kim, H.C. Li, Overpotential analysis with various anode gas compositions in a molten carbonate fuel cell, J. Power Sources 179 (2008) 467e473, https://doi.org/10.1016/ j.jpowsour.2007.12.125. [40] C.G. Lee, B.S. Kang, H.K. Seo, H.C. Lim, Effect of gas-phase transport in molten carbonate fuel cell, J. Electroanal. Chem. 540 (2003) 169e188, https://doi.org/ 10.1016/S0022-0728(02)01304-9. [41] B. Huang, X. Ye, S. Wang, Q. Yu, H. Nie, Q. Hu, J. Shi, K. Hu, T. Wen, Electrochemical performance of Y2O3/NiO cathode in the molten Li0.62/K0.38 carbonates eutectics, Mater. Res. Bull. 41 (2006) 1935e1948, https://doi.org/ 10.1016/j.materresbull.2006.03.007. [42] L. Hu, I. Rexed, G. Lindbergh, C. Lagergren, Electrochemical performance of reversible molten carbonate fuel cells, Int. J. Hydrogen Energy 39 (2014) 12323e12329, https://doi.org/10.1016/j.ijhydene.2014.02.144. [43] A.J. Appleby, S.B. Nicolson, Reduction of oxygen in Lithium-Potassium carbonate melt, J. Electroanal. Chem. 112 (1980) 71e76, https://doi.org/10.1016/ S0022-0728(80)80008-8. [44] K. Czelej, K. Cwieka, J.C. Colmenares, K.J. Kurzydlowski, Catalytic activity of NiO cathode in molten carbonate fuel cells, Appl. Catal., B 222 (2018) 73e75, https://doi.org/10.1016/j.apcatb.2017.10.003. [45] K. Czelej, K. Cwieka, J.C. Colmenares, K.J. Kurzydlowski, Atomistic insight into the electrode reaction mechanism of the cathode in molten carbonate fuel cells, J. Mater. Chem. 5 (26) (2017) 13763e13768, https://doi.org/10.1039/ c7ta02011b. [46] R.J.B. Smith, M. Loganathan, M.S. Shantha, A review of the water gas shift reaction kinetics, Int. J. Chem. React. Eng. 8 (2010), https://doi.org/10.2202/ 1542-6580.2238. Review R4.