Identification of Active Surface Species in Molten Carbonates Using in situ Raman Spectroscopy

Here we report the results of a study on active surface species of a pristine and modified (Li-Na)₂CO₃ eutectic using in situ Raman spectroscopy technique. The effects of gas compositions, temperature, time, and alkaline earth have been systematically studied. The species of CO₄^2–, HCO₄^–, and C₂O₅^2– are identified as the three major active species on the surface of (Li-Na)₂CO₃ eutectic by a combined Raman spectroscopy and theoretical density functional theory calculations. The results further reveal that CO₄^2–, HCO₄^–, and C₂O₅^2– are preferably formed in the presence of O₂, H₂O, and high CO₂ concentration. With the addition of Ba to the pristine (Li-Na)₂CO₃ eutectic, the Raman CO₄^2–/HCO₄^– shifts become more pronounced.

Introduction

Molten carbonate fuel cells (MCFCs) are a class of energy-efficient and low-emission power generators (Morita et al., 2002; Watanabe et al., 2006; Kawase, 2017). One of their unique features is the use of CO₂ and O₂ as the cathode gas. The mixture of CO₂ and O₂ is also a major component in flue gas produced by either coal-fired or natural-gas (NG)-fired power plants. A natural and logical question is if the flue gas can be directly utilized as the oxidant by MCFCs to produce additional power while mitigating the CO₂ emission.

The concentration of CO₂ emitted by coal-fired power plants is more than twice that emitted by NG-fired power plants, e.g., 10 vs. 4%. The current benchmark MCFCs can operate normally with a CO₂ concentration higher than 10%; at lower CO₂ concentrations, their performance is significantly limited by diffusion-related concentration polarization. To generate power from MCFCs using low CO₂ concentration flue gas produced from NG-fired power plants, the existing MCFC cell/stack technology needs to be further improved to minimize the concentration polarization and meet the performance requirements. In addition, there is a significant amount of H₂O, e.g., 10%, in NG flue gas. The effect of H₂O on the MCFC performance is not well understood from early studies.

A recent result has confirmed that MCFCs exhibit a high concentration-polarization at low CO₂ concentrations and surprisingly enhanced performance in the presence of H₂O (Rosen et al., 2020). An electrochemical model based on active species peroxide-ion (O₂^2–) and superoxide (O₂^–) has been proposed with satisfactory agreement with the experimental data (Cassir et al., 1993). Indeed, an independent Raman spectroscopy study has confirmed the existence of O₂^2– on the surface of a Li₂/K₂CO₃ eutectic (Chen et al., 2004).

A parallel conduction mechanism is also proposed for OH^– as a working ion when H₂O is present in the oxidant feed. From the water mass collected at both cathode and anode sides, it appears to suggest that the conduction mechanism of OH^– in a MC matrix is real (Rosen et al., 2020). However, more experimental evidence related to active species to support water transport is needed.

Raman spectroscopy is a very sensitive technique to detect vibrational modes of C-O-C, C-O, and C = O bonds that are abundant in C-containing species such as CO₃^2– and CO₄^2– and are of interest to this research (Itoh et al., 2004; Mendoza et al., 2004). We have previously used Raman spectroscopy to probe the surface chemistry of an Ag-MC (Molten Carbonate) membrane operated under a flue gas condition containing O₂ and CO₂ to facilitate our understanding of why enhanced oxygen permeation was observed (Tong et al., 2016). We concluded from both experimental and theoretical data that LiCO₄^– was the active species on the Ag-MC membrane surface under simulated flue gas conditions and the extra oxygen in CO₄^2– ligand relative to CO₃^2– is the reason for the increased oxygen transport. The in situ Raman spectra of the Ag-MC membrane were collected by a LabRam/HR confocal Raman system (LabRam Invers, Horiba Jobin-Yvon) equipped with a 632.8 nm He-Ne laser and hot stage (Linkam TS1500, 0–1,500°C). From this work, we were able to identify that CO₄^2– is an active species on the surface of an atomic layer deposition Al₂O₃-coated Ag-MC composite membrane and explain the phenomenon of the enhanced oxygen flux based on the percarbonate mechanism. In a similar work (Mendoza et al., 2004), we were also able to identify the active surface species as C₂O₅^2– with in situ Raman spectroscopy when a MC melt was exposed to a pure CO₂ atmosphere.

These prior works demonstrate the versatility and capability of in situ Raman spectroscopy in probing active species on the surface of MC. Therefore, in this study, the in situ Raman spectroscopy technology was selected to identify the species of molten carbonate exposed to varies atmospheres at MCFC operating temperatures.

Experimental

Sample Preparation

The molten carbonate compositions under this study were Li₂CO₃-Na₂CO₃ eutectic with a ratio of 52:48 mol% (MC), and 10 mol%-SrCO₃ added MC (Sr-MC) and 10 mol%-BaCO₃ added MC (Ba-MC). The reason for selecting Li₂CO₃-Na₂CO₃ eutectic was primarily based on the consideration of its higher ionic conductivity than other alkali carbonates eutectics. To prevent MC overflow at high temperatures, an Ag porous matrix was used to contain all MCs. The Ag porous matrix was prepared in the same way as in our previous study (Zhang et al., 2016). Briefly, Ag powders (99.9% metal basis, Alfa Aesar) were mixed with carbon black as a pore former in a volume ratio of 1:1 in ethanol. The dried power mixture was then pressed into pellets under 70 MPa, followed by sintering at 650°C for 2 h in air to remove the carbon pore former and achieve good mechanical strength. Then the porous Ag pellet was soaked into a MC at 650°C for 2 h in air, forming Ag/MC composite.

Characterization

The Raman system used for this study is a Horiba Jobin-Yvon LabRAM HR800. It features a focal length of 800 mm, spectral length of 100–4,000 cm^–1 with a spectral resolution up to 1.5 cm^–1, accuracy of wavenumber of 1 cm^–1 and spatial resolution up to 1 μm. We use the following operating parameters for Raman spectroscopy study on the surface of a MC:

• Wavelength: 300–1,800 cm^–1

• Light source: 632 nm red

• Acquisition time: 20 s

• Accumulation: 1

• Hole: 500

• Filter: 100%.

The system is also adaptable to high-temperature subsystem for in situ measurements. The high temperature unit (or herein called hot stage) is model Linkam TS1500. The hot-stage is connected to a water-cooling system driven by a water circulation pump. The actual temperature was calibrated with the melting point of prepared MC (490°C), and temperature effect was studied at 575 and 625°C. Since the atmosphere effect is a focus of the study, we systematically varied the gas compositions over the MC at a temperature. Table 1 lists all the gas mixtures studied. The total gas flow was fixed at 50 sccm for all measurements.

TABLE 1

Table 1. Gas compositions tested under this project.

DFT Calculations

The equilibrium geometries of species were fully optimized at the B3LYP/6-311G^∗ level using the Gaussian 09 program package (Hehre, 2002; Frisch et al., 2009). To better understand the vibrational modes and figure out fingerprints of these species, Raman spectra were simulated using the B3LYP/6-311G^∗ basis sets under the harmonic approximation at the same level of geometry optimization. Vibrational analysis showed that all the structures were at minima in the potential energy surface (no imaginary frequencies). Based on the Raman activity obtained from density functional theory (DFT) calculations, the Raman intensity was determined by the GaussSum software (O’Boyle et al., 2008).

Results and Discussion

Comparison of Raman Spectra in Dry Gas

Figure 1 compares the Raman spectra at 575°C in pure N₂ (gas #1), 2% CO₂ (gas #2), and 10% CO₂ (gas #3) atmospheres. The lower intensity of the Raman spectra in 10% CO₂ is due to the presence of Ag in the Ag-MC sample used to prevent the overflow of the MC. In the following studies, all Raman spectra were collected with the Ag-MC sample. As shown in the Figure 1, the main band at 1,072 cm^–1 is assigned to symmetric stretching (v₁) of O-C-O, where the bands at 707, 885, 1,391, 1,500, and 1,762 cm^–1 correspond to the in-plane bending (v₄) of C-O, out-of-plane bending (v₂) of O-C-O, doubly degenerate asymmetric stretching (v₃) of O-C-O, and the overtone of the out-of-plane bending mode (2v₂) vibrations, respectively (Bates et al., 1972; Chen et al., 2002; Zhang et al., 2013). Clearly, there is no difference in these Raman shifts, implying that dry N₂ and CO₂ conditions do not invoke new species other than CO₃^2–. Similar phenomenon was also observed in our previous study, where there is no new peak in the dry gas conditions with Ag-MC sample (Tong et al., 2016).

FIGURE 1

Figure 1. The Raman spectra of the MC/Ag-MC samples at 575°C in dry gas atmospheres with different CO₂ concentration. (N₂: gas #1, 2% CO₂: gas #2, 10% CO₂: gas #3).

The temperature and time effect on the Raman spectra of the Ag-MC samples in dry 10% CO₂ atmosphere (gas #3) are compared in Figure 2. Figures 2A,B show the Raman spectra of the Ag-MC sample at 575 and 625°C, respectively. These results confirm that there is no difference in spectrum in dry gas at both temperatures.

FIGURE 2

Figure 2. The effects of temperature and time on Raman spectra of the Ag-MC sample in dry 10% CO₂ gas (gas #3). (A) 575°C and (B) 625°C.

CO₂ Concentration Effect in Wet Gas

To investigate why steam promotes the MCFC performance, we also conducted the Raman spectroscopy of the Ag-MC sample in wet gas (10% H₂O) with different CO₂ concentrations. For an easy comparison, the Raman spectra in dry N₂ is also shown in Figure 3. Some new shift between 400 and 607 cm^–1 appear in wet CO₂ atmospheres. To identify these new shifts, DFT calculations were performed. From the DFT calculations, the weak but new shifts at 409 and 567 cm^–1 seem to relate to HCO₄^–, while the weak but new shifts at 607 cm^–1 can be assigned to CO₄^2–. These two species are observed in CO₂ concentration gases from 2 to 45% CO₂, while at high CO₂-concentraiton, e.g., 90% CO₂, C₂O₅^2– species appears at 455 cm^–1. The observation of C₂O₅^2– species at high CO₂ concentration is reasonable given the fact that CO₂ dissolution into MC to form C₂O₅^2– by CO₂+CO₃^2– = C₂O₅^2– (Claes et al., 1996).

FIGURE 3

Figure 3. Raman spectra, along with DFT calculations, of the Ag-MC sample collected at 625°C in different CO₂ concentration gas conditions: dry N₂ (gas #1), 2% CO₂ (gas #4), 9% CO₂ (gas #5), 45% CO₂ (gas #6), 90% CO₂ (gas #7).

When the gas atmosphere changed from dry N₂ to wet 2% CO₂, the weak but new bands at 567 and 607 cm^–1 seem to emerge, which correspond to HCO₄^– and CO₄^2–, respectively. (Note: the wet 2% CO₂ gas also contains 9% O₂, see Table 1, gas #4). Then the intensity of the new bands slightly increased by increasing CO₂ concentration to 9 and 45%. The formation mechanisms for the observed HCO₄^– and CO₄^2– are (Zhang et al., 2014; Tong et al., 2016):

${\text{H}}_2\text{O}+2\mathrm{C}{\mathrm{O}}_2+3/2{\mathrm{O}}_2+2{\mathrm{e}}^{-}=2\mathrm{H}\mathrm{C}{\mathrm{O}}_4^{-}(1)$

${\text{CO}}_2+3/2{\mathrm{O}}_2+{\text{CO}}_3^{2-}+2{\mathrm{e}}^{-}=2\mathrm{C}{\mathrm{O}}_4^{2-}(2)$

From these reactions, H₂O is essential to form HCO₄^–, but not for CO₄^2– species. However, it is interesting to see that CO₄^2– appears in wet flue gas but not in dry flue gas (i.e., ∼10% CO₂-10% O₂), see Figure 2B. A possible reason is that the steam facilitates the dissolution of O₂ into MC. From Figure 3, HCO₄^– and CO₄^2– bands disappear and C₂O₅^2– band appear with further increasing CO₂ concentration to 90% (no O₂), which is consistent with our previous study (Zhang et al., 2013).

The Effect of Alkaline Earth-Addition to MC

To further confirm that the formation of HCO₄^– and CO₄^2– is highly influenced by the solubility of molecular oxygen, and therefore, the basicity in MCs. We measured the Raman spectra of alkaline earth-addition MC, since it has been reported that the alkaline earth metals can enhance the oxygen solubility of the molten carbonate (Scaccia and Frangini, 2009). At first, the Raman spectra of Ba- and Sr-added MC under in 45% CO₂-10% O₂-10% H₂O-N₂ atmosphere (gas #6) were collected and are shown in Figure 4. It seems that Sr-doped MC sample show similar new bands with the pristine MC sample. However, the intensities of HCO₄^– and CO₄^2– bands of Ba-added MC samples are increased appreciably. In addition, a couple of more pronounced new peaks at ∼360 and 770 cm^–1 are observed of the Ba-added sample. Compared to DFT-calculated Raman bands, it is determined that the new shift is related to CO₄^2–. These results further suggest that high oxygen solubility promotes the formation of CO₄^–.

FIGURE 4

Figure 4. Comparison of Raman spectra of three types of Li-Na MC (Ba-MC, Sr-MC, and MC) exposed to 45% CO₂-10% O₂-10% H₂O-N₂ atmosphere at 625°C.

Since both the H₂O and alkaline earth metals increase the oxygen solubility of MCs, we also conducted Raman study of the Ba-MC sample under dry gas condition. Figure 5 shows that the Ba-MC sample does exhibit new surface species under CO₂-free, H₂O-free and wet gas conditions. Similar Raman spectra were obtained under the dry flue gas and wet 45% CO₂ atmosphere, confirming that the oxygen solubility highly influences the formation of CO₄^2– and HCO₄^–. Figure 5 also shows significant appearance of new shifts at 403 and 803 cm^–1 under dry 10% O₂ condition. DFT calculations suggest that it is related to BaCO₃. This finding implies that Ba-MC is unstable under dry 10% O₂-N₂, precipitating out BaCO₃ on the MC surface. However, in the presence of both H₂O and CO₂, BaCO₃ re-dissolves back into MC, but with a pronounced shift at ∼600 cm^–1 relating to species HCO₄^– and CO₄^2–. In addition, the new shift at 360 cm^–1 is attributed to CO₄^2– according to the DFT calculation, further indicating that the Ba-added MC enhances oxygen solubility. For convenience, we show in Figure 6 a schematic of the BaCO₃ precipitation and re-dissolution mechanisms in dry O₂ and CO₂-H₂O-O₂-containing gas mixtures, respectively.

FIGURE 5

Figure 5. Raman spectra of Ba-added MC collected under different gas conditions at 625°C.

FIGURE 6

Figure 6. Surface species of Ba-added MC exposed to different gas mixtures.

Overall, based on the above Raman spectroscopic results, we propose the following mechanisms for CO₂ and H₂O transport in MCFCs with a high-basicity MC as the electrolyte, H₂ as the anode gas and CO₂/H₂O/O₂ as the cathode gas:

At Cathode:

$\mathrm{Main}\mathrm{reaction}:{\mathrm{CO}}_2+1/2{\mathrm{O}}_2+2{\mathrm{e}}^{-}={\text{CO}}_3^{2-}(3)$

$\mathrm{Parallel}\mathrm{reaction}(1):{\mathrm{H}}_2\text{O}+2\mathrm{C}{\mathrm{O}}_2+3/2{\mathrm{O}}_2+2{\mathrm{e}}^{-}=2\mathrm{H}\mathrm{C}{\mathrm{O}}_4^{-}(\text{intermediate})(4)$

${\text{HCO}}_4^{-}+2{\mathrm{e}}^{-}={\text{OH}}^{-}+{\text{CO}}_3^{2-}(5)$

At Anode:

$\mathrm{Main}\mathrm{reaction}:{\mathrm{CO}}_3^{2-}+{\text{H}}_2={\text{H}}_2\text{O}+{\text{CO}}_2+2{\mathrm{e}}^{-}(6)$

$\mathrm{Parallel}\mathrm{reaction}(1):\mathrm{\hspace{0.33em}2}{\mathrm{OH}}^{-}+{\text{H}}_2=2{\mathrm{H}}_2\text{O}+2{\mathrm{e}}^{-}(7)$

Since HCO₄^– is virtually an intermediary species, the experimentally observed H₂O and CO₂ transport in MCFC are carried out by OH^– and CO₃^2–.

Conclusion

In summary, the surface chemistry of pristine and Ba- or Sr-modified Li-Na eutectic under seven gas mixtures has been studied by in situ Raman spectroscopy. Compared to the pristine MC under dry N₂, all the CO₂-containing dry gas conditions do not change Raman spectra of the MC. Under wet gas conditions, relatively weak new shifts appear in 300–700 cm^–1 and their intensities increase with CO₂ concentration. Assisted by DFT calculations, the new shifts are identified to be CO₄^2–/HCO₄^– at low CO₂ concentrations and C₂O₅^2– at high CO₂ concentrations. With the addition of Ba to the pristine MC, the CO₄^2–/HCO₄^– shifts become more pronounced due to enhanced oxygen solubility in MC. It was also discovered that BaCO₃ can precipitate out under CO₂-free dry gas environment. Based on the found surface intermediate species, a mechanism concerning H₂O and CO₂ co-transport is proposed.

Data Availability Statement

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.

Author Contributions

PZ conducted the experiment. TW performed the DFT calculations. KH conceived the idea. All authors contributed to the article and approved the submitted version.

Funding

Financial supports from National Science Foundation (Award # 1924095) and ExxonMobil are greatly appreciated.

Conflict of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Acknowledgments

The authors would like to acknowledge constructive discussion with Timothy A. Barckholtz, Gabor Kiss, and David Perkins at ExxonMobil.

References

Bates, J. B., Brooker, M. H., Quist, A. S., and Boyd, G. E. (1972). Raman spectra of molten alkali metal carbonates. J. Phys. Chem. 76, 1565–1571. doi: 10.1021/j100655a013

CrossRef Full Text | Google Scholar

Cassir, M., Moutiers, G., and Devynck, J. (1993). Stability and characterization of oxygen species in alkali molten carbonate: a thermodynamic and electrochemical approach. J. Electrochem. Soc. 140:3114. doi: 10.1149/1.2220995

CrossRef Full Text | Google Scholar

Chen, L., Lin, C., Zuo, J., Song, L., and Huang, C. (2004). First spectroscopic observation of peroxocarbonate/peroxodicarbonate in molten carbonate. J. Phys. Chem. B 108, 7553–7556. doi: 10.1021/jp035749l

CrossRef Full Text | Google Scholar

Chen, L.-J., Cheng, X., Lin, C.-J., and Huang, C.-M. (2002). In-situ Raman spectroscopic studies on the oxide species in molten Li/K₂CO₃. Electrochim. Acta 47, 1475–1480. doi: 10.1016/s0013-4686(01)00872-6

CrossRef Full Text | Google Scholar

Claes, P., Thirion, B., and Glibert, J. (1996). Solubility of CO₂ in the molten Na₂CO₃-K₂CO₃ (42 mol%) eutectic mixture at 800 ^oC. Electrochim. Acta 41, 141–146. doi: 10.1016/0013-4686(95)00278-m

CrossRef Full Text | Google Scholar

Frisch, M. J. E. A., Trucks, G. W., Schlegel, H. B., Scuseria, G. E., Robb, M. A., Cheeseman, J. R., et al. (2009). Gaussian 09, Revision d. 01.Wallingford CT: Gaussian. Inc, 201.

Google Scholar

Hehre, W. J. (2002). Ab initio molecular orbital theory. Acc. Chem. Res. 9, 399–406.

Google Scholar

Itoh, T., Abe, K., Dokko, K., Mohamedi, M., Uchida, I., and Kasuya, A. (2004). In situ raman spectroelectrochemistry of oxygen species on gold electrodes in high temperature molten carbonate melts. J. Electrochem. Soc. 151:A2042.

Google Scholar

Kawase, M. (2017). Durability and robustness of tubular molten carbonate fuel cells. J. Power Sources 371, 106–111. doi: 10.1016/j.jpowsour.2017.10.024

CrossRef Full Text | Google Scholar

Mendoza, L., Baddour-Hadjean, R., Cassir, M., and Pereira-Ramos, J. P. (2004). Raman evidence of the formation of LT-LiCoO₂ thin layers on NiO in molten carbonate at 650 C. Appl. Surf. Sci. 225, 356–361. doi: 10.1016/j.apsusc.2003.10.026

CrossRef Full Text | Google Scholar

Morita, H., Komoda, M., Mugikura, Y., Izaki, Y., Watanabe, T., Masuda, Y., et al. (2002). Performance analysis of molten carbonate fuel cell using a Li/Na electrolyte. J. Power Sources 112, 509–518. doi: 10.1016/s0378-7753(02)00468-8

CrossRef Full Text | Google Scholar

O’Boyle, N. M., Tenderholt, A. L., and Langner, K. M. (2008). cclib: a library for package-independent computational chemistry algorithms. J. Comput. Chem. 29, 839–845. doi: 10.1002/jcc.20823

PubMed Abstract | CrossRef Full Text | Google Scholar

Rosen, J., Geary, T., Hilmi, A., Blanco-Gutierrez, R., Yuh, C.-Y., Pereira, C. S., et al. (2020). Molten carbonate fuel cell performance for CO₂ capture from natural gas combined cycle flue gas. J. Electrochem. Soc. 167:064505. doi: 10.1149/1945-7111/ab7a9f

CrossRef Full Text | Google Scholar

Scaccia, S., and Frangini, S. (2009). Effect of Ba and Ca additions on the oxygen solubility properties of a (70/30) mol% Li₂CO₃/Na₂CO₃ carbonate melt. J. Mol. Liq. 146, 39–43. doi: 10.1016/j.molliq.2009.01.011

CrossRef Full Text | Google Scholar

Tong, J., Lei, X., Fang, J., Han, M., and Huang, K. (2016). Remarkable O₂ permeation through a mixed conducting carbon capture membrane functionalized by atomic layer deposition. J. Mater. Chem. A. 4, 1828–1837. doi: 10.1039/c5ta10105k

CrossRef Full Text | Google Scholar

Watanabe, T., Izaki, Y., Mugikura, Y., Morita, H., Yoshikawa, M., Kawase, M., et al. (2006). Applicability of molten carbonate fuel cells to various fuels. J. Power Sources 160, 868–871.

Google Scholar

Zhang, L., Huang, X., Qin, C., Brinkman, K., Gong, Y., Wang, S., et al. (2013). First spectroscopic identification of pyrocarbonate for high CO₂ flux membranes containing highly interconnected three dimensional ionic channels. Phys. Chem. Chem. Phys. 15, 13147–13152. doi: 10.1039/c3cp52362d

PubMed Abstract | CrossRef Full Text | Google Scholar

Zhang, L., Tong, J., Gong, Y., Han, M., Wang, S., and Huang, K. (2014). Fast electrochemical CO₂ transport through a dense metal-carbonate membrane: a new mechanistic insight. J. Membr. Sci. 468, 373–379. doi: 10.1016/j.memsci.2014.06.028

CrossRef Full Text | Google Scholar

Zhang, P., Tong, J., Jee, Y., and Huang, K. (2016). Stabilizing a high-temperature electrochemical silver-carbonate CO₂ capture membrane by atomic layer deposition of a ZrO₂ overcoat. Chem. Commun. 52, 9817–9820. doi: 10.1039/c6cc04501d

PubMed Abstract | CrossRef Full Text | Google Scholar