Tuning of Catalytic Activity by Thermoelectric Materials for Carbon Dioxide Hydrogenation

An innovative use of a thermoelectric material (BiCuSeO) as a support and promoter of catalysis for CO2 hydrogenation is reported here. It is proposed that the capability of thermoelectric materials to shift the Fermi level and work function of a catalyst lead to an exponential increase of catalytic activity for catalyst particles deposited on its surface. Experimental results show that the CO2 conversion and CO selectivity are increased significantly by a thermoelectric Seebeck voltage. This suggests that the thermoelectric effect can not only increase the reaction rate but also change chemical equilibrium, which leads to the change of thermodynamic equilibrium for the conversion of CO2 in its hydrogenation reactions. It is also shown that this thermoelectric promotion of catalysis enables BiCuSeO oxide itself to have a high catalytic activity for CO2 hydrogenation. The generic nature of the mechanism suggests the possibility that many catalytic chemical reactions can be tuned in situ to achieve much higher reaction rates, or at lower temperatures, or have better desired selectivity through changing the backside temperature of the thermoelectric support.


Introduction
Thermoelectric (TE) materials have recently attracted widespread interest in research because they can convert a temperature difference directly into an electrical voltage via the Seebeck effect, S = −V/ΔT, where V is the voltage between the two ends of the TE material and ΔT the temperature difference, S is the Seebeck coefficient. The performance of a TE material is ranked by its figure of merit ZT = S 2 σT/κ, where σ is the electrical conductivity, κ the thermal conductivity, and thin film and highly dispersed (nanoscale particles) metal catalysts, by using TE materials as a catalyst support for CO 2 hydrogenation. Furthermore, we show that this profound promotional effect on catalytic activity by the TE effect also enables the oxide TE material itself to possess high catalytic activity for CO 2 hydrogenation.
The concentration of carbon dioxide in the atmosphere has risen from ≈280 ppm before the industrial revolution to ≈400 ppm in 2013 and is projected to be ≈500 ppm by 2050. [6] This contributes to the increase in global temperature and climate changes due to the "greenhouse effect." Hence there are extensive efforts to reduce CO 2 emissions around the world. Generally speaking, there are three strategies to achieve this: (i) reduce CO 2 production, (ii) storage, and (iii) usage. The first two options, which involve improving energy efficiency, switching to renewable energy, and CO 2 capture and sequestration, have been the major focus in the past. The third strategy, i.e., using CO 2 as a feedstock for making useful chemicals, is regarded as the most feasible and effective solution to our carbon conundrum. [7] The CO 2 hydrogenation may undergo two main processes, the first is the reverse water-gas shift (RWGS) reaction (Equation (1)) and the other leads to the formation of hydrocarbons (Equation (2) For x = 1, y = 4, and z = 0 (i.e., the inlet gas ratio CO 2 /H 2 = 1:4) one has the methanation reaction. RWGS reaction is one of the most established reactions to convert CO 2 into CO and H 2 O, from which liquid hydrocarbons conversion via Fischer-Tropsch synthesis can be achieved.

Theoretical Consideration
NEMCA, which involves a reversible change of catalytic properties of metal catalysts deposited on solid electrolytes, can be obtained by applying a small external electric current or voltage. NEMCA is due to the back spillover of ionic species from electrolytes to form a double layer at the catalyst surface, which leads to a change of the work function and chemisorption properties of the catalyst. [3,8] The effective change of surface work function leads to an exponential change of chemical reaction, [9] i.e., where r is the new reaction rate, r o is the open-circuit reaction rate, k b is the Boltzmann constant, α is an empirically determined constants, and Δφ is the change of work function due to the applied external voltage. Under certain conditions, Δφ is linearly proportional to the non-Ohm drop of external potential. [3,4,9] The basic idea of this work is to use a thermoelectric material to change the effective work function of catalyst particles to achieve a significant increase of the catalytic activity. First, consider the change of the Fermi level for TE material BCSO when there is a temperature change (Figure 1). BCSO is a p-type TE material, holes at the hot side diffuse into the cold side upon heating, forming an internal electrical field. Once equilibrium is reached the Fermi level (also called electrochemical potential) at the hot side ε F,h is higher than at the beginning and cold side ε F,c and Δε F = ε F,h − ε F,c = -eV, here −e is the charge of an electron and V is the Seebeck voltage.
As no external charges exist, the change of the work function at the surface is the inverse change of the Fermi level, i.e., Δφ = −Δε F , so Δφ = eV at the hot surface T h . If a metal particle is deposited on the TE material, at the hot surface, its Fermi level ε F,m must be the same as the Fermi level of the TE at the surface, i.e., ε F,m = ε F,h ( Figure 1). Therefore, Δε F,m = Δε F and the change of work function Δφ m = eV is also true for metal particles supported on the TE material.
Apply the generalized dependence of catalytic rate on catalyst work function Equation (3), then we have Here γ is a constant, to be determined by experiment. The introduction of a minus sign makes −eV the extra energy of an electron at the surface due to the Seebeck voltage V. Combining Equation (4) and the definition of Seebeck coefficient S gives And Ln / / / ,at cold side Equations (4)-(6) link catalytic activity with Seebeck voltage and temperature for a metallic catalyst supported on a TE material. Figure 2 shows the temperature dependence of (a) Seebeck coefficient and electrical conductivity, (b) thermal conductivity, and (c) power factor and dimensionless figure of merit ZT for the BCSO pellets after spark plasma sintering (SPS). The Seebeck coefficient was highest at room temperature with a value 516 µV K −1 , then decreased with increasing of temperature, and reached 328 µV K −1 at 764 K. The electrical conductivity σ decreased with increasing temperature from room temperature to about 460 K, then increased with further increasing temperature, and reached its highest value of 18.8 S cm −1 at 764 K. The thermal conductivity k was found to decrease continuously with temperature, being 0.84 W m −1 K −1 at 315 K and 0.42 W m −1 K −1 at 764 K; these values are very low even for TE materials. The highest power factor (S 2 σ) of 230 µW m −1 K −2 was obtained at ≈665 K. ZT values were found to increase with increasing temperature and reached 0.37 at ≈764 K. The Seebeck coefficient and electrical conductivity of the samples had lower values but simi lar trends with temperature as BCSO prepared using self-propagating high-temperature synthesis (SHS), [10] but higher than those of BCSO prepared using solid state reaction (SSR). [11] The thermal conductivity values were lower than both SHS and SSR prepared BCSOs. [10,11] As a result, our BCSO showed similar ZT values (0.36 at 665 K and 0.37 at 764 K) as SHS (0.33 at 675 K and 0.49 at 775 K), [10] but higher than SSR (0.09 at 725 K and 0.15 at 775 K) samples. [11]

Thin Film and Nanoparticle Catalysts on BiCuSeO
The surface microstructures of the catalysts on BCSO were investigated using scanning electron microscopy (SEM).  (15), Pt(80), and Pt(NP) refer to the nominal Pt film thicknesses of 15, 80 nm, and as nanoparticles, respectively. The surface of BCSO was relatively smooth. Many Pt particles (indicated by arrow heads) could be observed on the grains of Pt(80)/BCSO and some Pt particles could be seen on the grains of Pt(15)/BCSO as well (indicated by arrow heads). Pt(NP)/ BCSO had smaller grains and more voids. This was probably due to its lower sintering temperature (823 K), compared with 923 K used for the other three samples. X-ray diffraction pattern (XRD) patterns for all of the four samples are shown in Figure 3e, indicating that the BCSO in every sample was almost a single phase (PDF#45-0296) with the ZrSiCuAs structure. No second phase was observed in BCSO, but a small peak at 2θ = 27.2 suggested some Bi 2 O 3 was present as a second phase in the other three samples. Also, the Pt(111) peak was apparent for Pt(80)/BCSO, but could not be seen in the XRD patterns for Pt(15)/BCSO and Pt(NP)/BCSO, indicating that the Pt particle sizes in the latter two samples were too small to be detected by XRD, probably less than 20 nm in diameters.

The Thermoelectric and Reduced Thermoelectric (RTE) Effect Conditions
The schematic diagram of the single chamber reactor which combines TE effect with catalytic chemical reaction is shown in Figure 4a and Figure S1  reaction chamber was placed on top of a hot-plate to create a large temperature difference (≈200-300 K when T h > 500 K) between the bottom floor of the chamber and the hot surface T h of the sample (Figure 4b). A large temperature gradient in the chamber can induce strong convection along the vertical direction, which can bring in the reactants and remove the products quickly from the reaction surface T h . Figure 3a-d shows that the samples were not porous; hence, there was no porediffusion limitation. For these reasons, it was assumed that there was no mass transportation limitation, and the intrinsic chemical reaction was the rate limiting step for all the reactions investigated. Disc samples with a diameter of 20 mm and thickness of 2 mm were tested for catalytic activity as represented by CO 2 hydrogenation conversion X (%) at different temperatures (CO 2 conversion X is proportional to the CO 2 reaction rate r if the backward reaction of Equation (1) is ignored, this will be discussed later). Catalytic activities are then compared between the TE and RTE conditions, at the same front (hot) surface temperature. Under normal TE conditions, the backside of the disc was in contact with water cooled stainless steel cap (with a thin mica sheet in between for electrical insulation), so its temperature was never higher than 373 K. A large temperature gradient across the disc thickness was created when the front surface reached a high temperature. Under RTE conditions, the backside of the disc was not in contact with the cooled cap, so the temperature gradient across the disc thickness was much smaller. At a particular hot-plate temperature, after reaching thermal equilibrium, the bottom surface of the disc sample was stabilized at a temperature T h , while the top surface was at a temperature T c . Hence, for the same sample at the same temperature T h under TE and RTE conditions, the only differences was that the top surface temperature T c was different, which led to a different Seebeck voltage across the sample. The Seebeck voltage V between the surfaces T c and T h was monitored continuously during the whole period of the experiment ( surface T c (nominal surface area 100 π mm 2 ), and the side wall of the disc (nominal surface area 40 π mm 2 ) sample. From Tables S1 and S2 (Supporting Information), it can be seen that when T h was below 403 K no CO 2 conversion was obtained, and T c was never higher than 331 K. Especially at high temperatures, T h was much higher than T c , and the temperature of the side wall was between T c and T h . For these reasons we assume that for all of the samples, the measured CO 2 conversion rate was contributed from the hot surface T h only, and the contributions from the cold surface T c and the side wall of the disc sample were negligible.

Higher Catalytic Activity at the Same Temperature under TE than RTE Conditions
The reaction products observed were only CO and CH 4 , with the vast majority (>90%) being CO (Figure 5a). Figure 5b shows the CO 2 conversion as a function of temperature   the TE conditions, much lower than 553 K when it was first measured under the RTE conditions (0.2%). It is plausible to assume that relative to conditions without any TE effect, the promotional effect should be even higher. It is worthy to point out that similar experiments were repeated at least once and the results were reproducible (the same samples were used for oxidized ethylene to form CO 2 and H 2 O, and a repeatable and similar thermoelectric promotional effect was observed; these are the subjects for a separate publication), this ruled out the possible explanation that the conversion difference between the TE and RTE was due to the catalyst particles aggregation at the surface. Another observation was that the BCSO TE sample, without any Pt catalyst, was also catalytically active for CO 2 hydrogenation ( Figure 5b). This may not be a total surprise, as BCSO is electrically conductive, and other conductive oxides have been found to be good catalysts. [12] Moreover, Cu and CuO catalysts are widely used for CO 2 hydrogenation, [13] so the CuO containing BCSO itself could have very low catalytic activity even without thermoelectric promotion. The first measured CO 2 conversion (0.3%) was at 493 K under TE conditions and 633 K under RTE conditions (0.8%). As for Pt(80)/BCSO, at the same temperature T h the CO 2 conversion under TE conditions was much higher than under RTE conditions. At 698 K, the CO 2 conversion was 20.8% under TE conditions compared to 3.4% under RTE conditions. Figure 5c plots Ln(X) against −eV/k b T h for all of the cases (for the p-type BCSO, V was negative and the term −eV was positive). It can be seen that a very good linear relationship existed between Ln(X) and −eV/k b T h for each case.

Promotion of CO 2 Conversion by Thermoelectric Effect
To further investigate the relationships between the temperature, Seebeck voltage, and catalytic activity, the CO 2 reduction reactions were studied for different samples under different inlet gas compositions, all under TE conditions. Figure 6a displays the measured thermoelectric voltage as a function of the temperature difference ΔT across the sample thickness for four samples, namely Pt(80)/BCSO, Pt(15)/BCSO, Pt(NP)/ BCSO, and bare BCSO, at the inlet gas ratios of CO 2 /H 2 = 1:1 and 1:4. All of the four samples were weighted as 5.8 g. All of the samples had zero voltages when their bottom and top surfaces were at the same (room) temperature. The measured voltage for each sample increased linearly with temperature difference. The linear gradient for Pt(80)/BCSO was 319 µV K −1 for ΔT < 200 K, and then decreased with increasing ΔT. The gradients for BCSO and Pt(15)/BCSO were similar and did not change with the change of the inlet gas compositions. These are typical values for Seebeck coefficient of BCSO. [2] Note that the Seebeck coefficient of 319 µV K −1 here is lower than the values reported in Figure 2a for SPS processed material. The main reason for this was that the BCSO used for the above catalysis experiments was not densified by SPS but using conventional sintering, so an inferior crystallinity and density of this sample was expected, which led to a smaller Seebeck coefficient. Another reason was that the Seebeck coefficient obtained through the linear gradient here was the average value over a large temperature range, while those in Figure 2a were obtained by changing the temperature over a smaller range (≈50 K over a 13 mm long sample), and generally speaking, the Seebeck coefficient is temperature dependent. The gradient for the Pt(NP)/BCSO was much lower, at about 136 µV K −1 , again, it also kept the same value when the inlet gases ratio was changed from 1:1 to 1:4. This much lower Seebeck coefficient was due to the fact that this sample was sintered at a much lower temperature (823 K as compared to 923 K for other BCSO), and there were still some second phases such as Bi 2 O 3 and void in the sample (Figure 3d). These results demonstrate that the measured voltage was determined by the intrinsic thermoelectric properties of the sample and the temperature difference only, and was not affected by the gas compositions. Again, the reaction products were found to be CO and CH 4 , with the majority (>80%) being CO. Higher H 2 concentration in the inlet gases led to lower CO selectivity. The temperature dependences of CO selectivity for six cases are shown in Figure 6b, while the other two, BCSO 1:1 TE and Pt(80)/ BCSO 1:1 TE, are shown in Figure 5a. Generally speaking, at T > 600 K, the CO selectivity increased with temperature and voltage. Figure 6c shows the CO 2 conversion as a function of the hot-surface temperature T h for different samples at the inlet gas ratios of CO 2 /H 2 = 1:1 or 1:4. All of the samples showed a similar trend, i.e., the conversion increased with temperature. It can be seen that for the same sample, higher H 2 concentration leads to higher CO 2 conversion. Pt(80)/BCSO reached 48.4% conversion at 656 K, the highest for all of the samples, indicating that the Pt surface had the highest catalytic activity. Remarkably, even without any Pt catalyst, the TE material BCSO (CO 2 /H 2 = 1:4) itself reached a conversion of 41.2% at 703 K.
Combining the results as shown in Figures 5c and 6d, we can summarize the observed relationship in Equation (7) Ln / / Here, X 0 is the conversion rate when V equals zero, i.e., when T c = T h . For the p-type TE material BSCO, V at T h surface is negative, so − γeV is positive and the conversion rate at the hot side T h could be much higher with a TE voltage than without, we call this thermoelectric promotion of catalysis (TEPOC), or thermoelectrocatalysis as the TE material itself can be catalytic active. Take an experimental data point for Pt(80)/BCSO @ 1:4, T h = 656 K, V = −86 mV, and γ = 7.15, so Ln(X/X 0 ) = −γeV/k b T h = 10.88, and X/X 0 = 53103. This means that at 656 K, the conversion with a Seebeck voltage of −86 mV was more than 53 thousand times higher than without a Seebeck voltage. Equation (4) can lead to Equation (7), and vice versa, if the conversion X is proportional to the reaction rate r. This requires the conversion X to be much lower than the thermal equilibrium conversion (TEC) of the reactions in Equation (1) and Equation (2), so that the backward reactions were negligible. The TECs at 673 K for CO 2 conversion in RWGS reaction without methanation (to CO only) are about 22% and 42% for an inlet gas ratio CO 2 /H 2 = 1:1 and 1:4 respectively; with methanation, the corresponding values are about 23% and 80%, respectively. [14,15] Under RTE conditions, the conversion rate was very low; hence, the CO 2 conversion on both Pt(80)/BCSO and BCSO was far away from the TEC, so it is safe to assume that the backward water-gas shift reaction can be ignored and the CO 2 conversion X was linearly proportional to the reaction rate r. So for the two cases under RTE conditions, the experimental results confirmed the prediction of Equation (4).
The rate of chemical reactions usually follows the Arrhenius law, so r 0 = k 0 exp(−E a /k b T h ), here k 0 is a constant, E a the activation energy of the reaction. Equations (4) and (7) apparently suggest that the activation energy is reduced by −γeV, i.e., E′ a = E a + γeV, E′ a is the new activation energy when there is a TE voltage V (a negative value for our case, as the reaction take place at the hot side of a p-type TE material). Figures 5b and 6c show that at high temperatures around 673 K, when there was a high Seebeck voltage, the CO 2 conversion rate reached TEC (22.9% for Pt(80)/BCSO @ 1:1 TE at 673 K), or just slightly below TEC (17.6% for BCSO @ 1:1 TE at 673 K, 37.2% for Pt(15)/BCSO @ 1:4 at 678 K, 36.3% for BCSO 1:4 at 683 K), even above the TEC (48.4% for Pt(80)/ BCSO 1:4 at 656 K) without methanation. For the purpose of comparison, the TEC values for CO 2 conversion in RWGS reactions with and without methanation at CO 2 /H 2 ratios 1/1 and 1/4 at 673 K are also presented in Figure 6c. To the best of our knowledge, 48.4% is the highest reported CO 2 conversion to CO (with 100% CO selectivity) at atmosphere pressure below 673 K with an inlet gas ratio CO 2 /H 2 no larger than 4. [13,15,16] How can the CO 2 conversion to CO exceeds the TEC at 673 K? It can be seen from Figure 6b that at temperatures T h > 678 K, the CO selectivity was >95% for all the samples. Notably for Pt(80)/BCSO @ 1:4, the CO selectivity was 100% at 656 K. The CO selectivities observed at these temperatures were also much higher than the predicted values under the consideration of thermal equilibrium. [13,15,16] These results indicate that the Seebeck voltage promoted the conversion to CO and forward reaction in Equation (1), hence changed the TEC. This agrees with the observation that an electric field (via the NEMCA mechanism) shifted the chemical equilibrium, increased the RWGS reaction, and decreased the (backward) water-gas shift reaction. [15] With the assistance of an electric voltage of 1.6 kV, CO 2 conversion to CO on a Pt/La-ZrO 2 catalyst reached 40.6% with an inlet gas ratio CO 2 /H 2 = 1:1 at 648 K, much higher than the TEC of about 20% without electric field at the same temperature. [15] This shifted the chemical equilibrium and TEC by an electrochemical energy of −eV and may also be the reason why we observed the linear relationships in Figures 5c and 6d. Strictly speaking, if the conversion rate is close to the TEC and the backward water-gas-shift reaction cannot be ignored, and the conversion X is not determined by the reaction rate, and Equation (4) cannot lead to Equation (7). Nevertheless, a very good linear relationship between Ln(X) and −eV/k b T h was observed for all the cases investigated. The most plausible explanation is that the Seebeck voltage V (or electrochemical energy −eV) shifted the reactions in Equation (1) toward the forward reaction, i.e., the RWGS against the backward water-gas shift reaction. Hence, the achieved conversion rate was still far away from the new chemical equilibrium and Equation (7) can still be explained by Equation (4).

Discussion
Referring to Figure 6a,c,d, all the samples, either bare BSCO or BCSO with a continuous Pt thin film Pt(80)/BCSO, or BCSO with discontinuous Pt nanoparticles Pt(15)/BCSO and Pt(NP)/ BCSO, showed similar CO 2 conversion dependence with the temperature T h and Seebeck voltage V (Figures S4 and S5, Supporting Information). The four samples with similar Seebeck voltage at a particular temperature, i.e., BCSO @1:4, Pt(15) BCSO @1:1, Pt(15)/BCSO @1:4, and Pt(80)/BCSO @1:4, also had similar Ln(X) ∼ −eV/k b T h relationships. The sample Pt(NP)/ BCSO @ 1:1 and 1:4 had the lowest Seebeck voltage and also had a similar Ln(X) ∼ −eV/k b T h relationship. This suggests that the Seebeck voltage, not specific surface property, was the most important factor in determining the catalytic activity. This also agrees with the observation that the CO 2 conversion stronger dependence on the effect of the electric field than the nature of the catalyst. [15] These results also agree with the observations in NEMCA of CO 2 hydrogenation in that a negative (reduced) potential increased the selectivity and reaction rate to CO, and a positive (increased) potential increased the selectivity and reaction rate to CH 4 . [13,17] From the above discussion, all the above observed results can be explained by Equation (4), i.e., the change of work function lead to the promotion of catalytic activity. This mechanism based on the change of work function through the in situ and controlled TE effect suggests that TEPOC is an effective mechanism for any metallic catalysts, regardless of their properties such as particle size or total amount of the metal. This is because whatever the particle size or chemisorption property, the Fermi level of the metallic particle will be the same as that of the surface TE materials supporting them. The total amount of the metal particles, indeed any second phase materials, will affect the TE properties such as Seebeck coefficient and electrical conductivity, as the whole system can be regarded as a TE composite. This is because all of the samples, with or without metal Pt, are just thermoelectric materials with a different Seebeck coefficient. Of course, the metal particle surface and TE surface may have different adsorption properties, which may lead to different catalytic properties.
Since the TE effect can be realized independently of chemical reactions, its modification to the catalytic activity can be in situ under operational conditions, and controlled through the control of the backside temperature, e.g., changing the water cooling to liquid nitrogen cooling. For n-type TE materials, the Fermi level at the cold side is higher than at the hot side, but the relationship ε F,h − ε F,c = −eV is still valid, so is Δφ = eV, but V is now positive.
The significant promotional effect of the TE effect when there is a large Seebeck voltage can be understood from the energy point of view. −eV/k b T h can be regarded as the ratio between the extra electrochemical energy induced by TE effect and the thermal energy of an electron at the reaction surface. At 300 K, the thermal energy k b T is 25.9 meV. So, 104 mV of Seebeck voltage gives 104 meV extra electrochemical energy to an electron at the Fermi level, which is equivalent to the thermal energy of an electron at 1200 K, but a 104 mV Seebeck voltage can be generated by a temperature difference of 347 K by a TE material (such as BCSO) with an average Seebeck coefficient of 300 µV K −1 . So, TE effect is a very efficient way to enhance the electrochemical energy of an electron at the reaction surface.
Considering Δφ = eV in TEPOC, note that Equation (4) is similar to the rate equation for NEMCA [3,8,9] , which is Ln(r/r o ) = α(Δφ − Δφ*)/k b T, where r o is the open-circuit reaction rate, α and Δφ* are empirically determined constants, Δφ is the change of work function due to the applied external voltage. Under certain conditions, Δφ is linearly proportional to the non-Ohm drop of external potential, [3,8] so the rate Equation (4) for TEPOC looks exactly the same as the rate equation for NEMCA. However, there are a few important differences between NEMCA and TEPOC. (i) No electrolyte nor external voltage are needed in the TEPOC system, while for NEMCA, an electrical insulating electrolyte layer is crucial otherwise a non-Ohm drop of potential (or ionic current) cannot be established. In fact, the unusually low thermal conductivity of BCSO has been attri buted to its negligible ionic conductivity, so the back spillover of ionic species in BCSO would have been negligible. [18] Also, we did not observe any change of reaction rate when an external voltage (positive or negative) was applied to the Pt(80)/ BCSO or other samples. (ii) Unlike in NEMCA, the catalyst in TEPOC (e.g., Pt) does not need to be continuous, as TE materials are electrically conductive. Highly, separately dispersed catalysts, including nanoparticle catalysts, can be promoted by TEPOC. (iii) The constant α in NEMCA is smaller than unity, but the values for the constant γ in TEPOC have been found to be larger than 1. The fact that γ > 1 in the Equations (4)-(7) for TEPOC indicates that there is an amplification effect when the extra electrochemical energy eV is present during catalytic chemical reactions. The mechanism for this is not clear yet, but we speculate that this is related to the increase of the number of electrons available for catalytic reaction with the increasing temperature, as no change of electron density with temperature should mean γ = 1. (iv) The TE material itself can be used as a catalyst when there is a large Seebeck voltage. (v) Furthermore, more importantly, the mechanism for the change of work function at the catalyst surface in TEPOC is different from that in NEMCA. In NEMCA, the external voltage induces the diffusion of ionic species, which form a double layer on the catalyst surface, and produce a change of the work function. [3,8] Hence, the change of work function Δφ is an indirect consequence of the external voltage V, and the linear relationship between Δφ and V is true only under certain conditions and may be sample dependent. [19] In TEPOC, the relationship between Δφ and Seebeck voltage V is directly linked by the change of Fermi level, not through the formation of a double layer.

Conclusions
The thermoelectric oxide BiCuSeO has been produced using a facile solid state reaction method using B 2 O 3 as a flux agent in air. An innovative use of the thermoelectric material as a catalyst support and promoter has been proposed and investigated through the CO 2 hydrogenation to produce CO and CH 4 . A very high CO 2 conversion of 48.4% to CO with 100% CO selectivity under atmosphere at temperatures below 673 K with the inlet gas ratio CO 2 /H 2 = 1:4 was obtained.
It is proposed that the thermoelectric effect can change the Fermi level and therefore the work function of the electrons in the catalyst particles supported on a thermoelectric material. This change of work function leads to exponential increase of catalytic activity. It was indeed observed in experiments that the catalytic activity of metallic particles supported on the thermoelectric materials, as represented by the CO 2 conversion, was significantly promoted by a Seebeck voltage generated through a temperature difference across the thickness of the thermoelectric support. This thermoelectric promotion of catalysis also enabled the BiCuSeO itself to possess high catalytic activity. It was further confirmed by experimental results that there exists a linear relationship between the logarithm of the catalytic activity, and − eV/k b T, which can be regarded as the ratio of extra electrochemical energy (−eV) induced by thermoelectric effect and thermal energy (k b T) of an electron. This extra electrochemical energy can also change the chemical equilibrium and selectivity of the reaction.
The general nature of the mechanism suggests that thermoelectric promotion of catalysis could be a universal phenomenon.

Experimental Section
Thermoelectric Material Preparation: The TE material BCSO was synthesized by solid-state reaction using boron oxide (B 2 O 3 , Alfa Aesar, 99%) as a flux agent in air. During the flux synthesis, the melted B 2 O 3 served as a liquid-seal on the top of the crucible. The obtained product of each sample was then ground to a fine powder. The latter was densified at 150 MPa using a hydraulic press system to form a dense pellet of 20 mm in diameter and 2 mm in thickness. Then, the green pellet was sintered at 923 K for 10 h under an argon atmosphere. Further sintering by SPS was carried out before thermoelectric property measurements, using a HP D 25/1(FCT Systeme GmbH, Frankenblick, Germany).
Preparation of Catalysts: The film catalysts were deposited on BCSO by magnetron sputtering method (Nordiko). The Pt films were prepared using pure Pt (99.99%) as the sputtering target. The thicknesses of the Pt films were ≈80 nm for three min and ≈15 nm for 20 s of sputtering time, respectively named Pt(80)/BCSO and Pt(15)/BCSO. Another platinum nanoparticle sample (Pt(NP)/BCSO) was synthesized using an impregnation method. For this sample, the green pellet was calcined at 823 K for 2 h under argon atmosphere and then reduced under 5% H 2 in Ar at 773 K for 4 h.
The microstructural investigations were carried out using XRD (Siemens 5005) at 40 kV with a Cu Kα source and a scanning electron microscope (Philips, FEI XL30 SFEG).
Thermoelectric Property Measurements: The thermal diffusivity (D) was measured by using laser flash method (LFA-457, Netzsch, Germany) under a continuous argon flow. The total thermal conductivity (κ total ) was calculated by the formula κ = DC p ρ, where ρ was the mass density measured by the Archimedes method, while the specific heat (C p ) was determined using a differential scanning calorimeter instrument. The electrical conductivity and Seebeck coefficient were simultaneously measured (LSR-3/1100, Linseis) in a He atmosphere.
The Reaction Chamber: Chemical reactions were performed in a single chamber reactor. A schematic diagram of the reactor can be seen in Figure 1. The cover plate was cooled with continuous running water. Gold wires (Agar Scientific, Ø 0.2 mm) were used as electrical contacts, and temperatures were measured with K-type thermocouples (Ø 0.25 mm, TC Direct) placed directly on the sample surfaces. The reaction chamber was placed directly onto a high temperature hot plate (HP99YX, Wenesco, Inc.) with a temperature controller. The Seebeck voltage was measured continuously (Figures S1 and S2, Supporting Information) between the bottom surface and the top electrode (Au) using a potentiostat-galvanostat (VersaStat 3F, Princeton Applied Research).
Catalytic Activity Measurement: The catalytic activity measurements of different catalysts were carried out at atmospheric pressure in a continuous flow apparatus equipped with the stainless steel reactor (Figure 1b). The reaction reactants and products were continuously monitored using online gas chromatography (GC8340, CE instruments) and online IR analyzer (G150 CO 2 , Gem Scientific) to quantify the concentration of H 2 , CO, CH 4 , and CO 2 . To monitor the temperature, a K-type thermocouple was attached (and fixed using a high temperature tape) onto the catalyst surface for the T h measurement. Another K-type thermocouple was placed in proximity of the top surface for the T c measurement. Carbon mass balance for all of the experiments was found to be within 6%.
The catalyst activities were investigated with the composition of carbon dioxide and hydrogen at a ratios of CO 2 :H 2 = 1:1, and CO 2 :H 2 = 1:4. All samples were tested at an overall flow rate of 100 mL min −1 .
The conversion of CO 2 and the selectivity of CO and CH 4 were evaluated from the outlet carbon percentage values obtained by the gas analysis. H 2 O vapor was condensed before entering the GC to prevent deterioration of the GC column. CO 2 conversion X CO 2 , and the selectivity of CO and CH 4 were calculated as where y CO 2 , y CO , and y CH 4 were the mol fractions of CO 2 , CO, and CH 4 in the outlet, respectively. × × , where f v = 100 mL min −1 is the volumetric flow rate at the outlet of the reactor and r CO 2 is the CO 2 reaction rate.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.