Article pubs.acs.org/JPCC
Hydride Transfer at the GaP(110)/Solution Interface: Mechanistic Implications for CO2 Reduction Catalyzed by Pyridine Published as part of The Journal of Physical Chemistry virtual special issue “Veronica Vaida Festschrift”. Martina Lessio,† Johannes M. Dieterich,‡ and Emily A. Carter*,§ †
Department of Chemistry, Princeton University, Princeton, New Jersey 08544-1009, United States Department of Mechanical and Aerospace Engineering, Princeton University, Princeton, New Jersey 08544-5263, United States § School of Engineering and Applied Science, Princeton University, Princeton, New Jersey 08544-5263, United States ‡
S Supporting Information *
ABSTRACT: Surface hydrides (H−*s) play a crucial role in one of the heterogeneous mechanisms proposed for pyridine-catalyzed CO2 reduction on p-GaP electrodes. In this mechanism, H−* is transferred to adsorbed pyridine (Py*) concomitant with aqueous proton addition to form the active catalyst adsorbed dihydropyridine (DHP*), which in turn transfers hydride to CO2 and leads to its reduction. In this contribution, we test the validity of these hypothesized hydride transfers, determining whether or not H−* can participate in the mechanism of CO2 reduction on p-GaP electrodes. To this end, we use our previously developed cluster models with hybrid density functional theory and a mixed implicit−explicit solvation approach to calculate the thermodynamic hydricity of relevant species involved in the proposed mechanism. Overall, the proposed heterogeneous mechanism is supported by the computed thermodynamic hydricities. However, computed reaction and activation energies for H−* transfer from the surface reveal that H−* cannot participate in CO2 reduction on p-GaP electrodes because of a high kinetic barrier to both formation of DHP* and direct CO2 reduction via H−* transfer. We thus conclude that an intermediate whose formation does not require H−* transfer must play the role of the active catalyst in this system. Specifically, our computed thermodynamic hydricities suggest that a recently proposed 2-PyH−* intermediate, formed via two-electron reduction and protonation of Py*, is a plausible candidate for the active catalyst in this system. Several reaction mechanisms, for both Pt3,7 and p-GaP8−10 electrodes, have been proposed since these experimental findings were first reported. Keith and Carter proposed a heterogeneous mechanism for CO2 reduction catalyzed by an adsorbed Py-derived species on p-GaP electrodes.8,9 In this mechanism, adsorbed pyridine (Py*) reacts with a surface hydride (H−*) and a proton from solution (H+sol) to form adsorbed dihydropyridine (DHP*); DHP* then acts as the active intermediate and reduces CO2 to formic acid (HCOOH) by transferring both a hydride and proton to the former. In a more recent study, Lessio and Carter suggested that Py* and adsorbed hydrogen atoms (H*) can be generated by a oneelectron reduction of solvated pyridinium (PyH+sol),11 which is naturally present in the acidified solution used in the experiments. H* can be reduced then to H−* by a second electron provided by the negatively biased photocathode. The overall mechanism proposed by Carter and co-workers is reported in Figure 1. Note that while two isomers exist for DHP (1,2-(ortho)-dihydropyridine (o-DHP) and 1,4-(para)-
1. INTRODUCTION The development of renewable and clean energy sources is one of our generation’s top priorities, given inevitable fossil fuel shortages in the future and the increased concentration of atmospheric CO2 directly impacting Earth’s climate. The use of a renewable energy source to help convert CO2 and H2O into liquid fuels would provide simultaneously a sustainable source of fuels and a way to mitigate our contribution to CO2 in the atmosphere. Bocarsly and co-workers have used a p-GaP photocathode under visible light illumination in contact with an acidified aqueous solution containing pyridine (Py) to reduce CO2 to methanol with excellent selectivity and low applied potentials.1 They observed nearly 100% faradaic efficiency toward methanol at underpotentials more than 300 mV below the thermodynamic potential for CO2 reduction to methanol. Similar findings were observed also when other electrode materials (Pd,2 Pt,3 CdTe,4 and CuInS25,6) were used in contact with an acidified aqueous solution containing Py. These findings confirm that Py, or a species derived from it, is needed to obtain good catalytic performance. Unravelling the reaction mechanism behind these experimental observations could contribute to developing an efficient technology to convert CO2 into liquid fuels using solar power. © XXXX American Chemical Society
Received: May 24, 2017 Revised: July 12, 2017 Published: July 13, 2017 A
DOI: 10.1021/acs.jpcc.7b05052 J. Phys. Chem. C XXXX, XXX, XXX−XXX
Article
The Journal of Physical Chemistry C
Figure 1. Proposed mechanism for DHP-catalyzed CO2 reduction on p-GaP electrodes by Carter and co-workers.
(−0.85 V vs SCE)11 suggest that these processes can be promoted by photoexcited electrons, on the basis of the computed values of solvated GaP(110) and GaP(111) conduction band minima (−1.76 and −1.68 V vs SCE, respectively).14 In contrast, PyH+sol reduction to the solvated pyridinyl radical (computed reduction potentials range from −1.31 to −1.58 V vs SCE),7,18−20 which is a mechanistic step required in both mechanisms proposed by Bocarsly3 and Musgrave,10 is less likely to occur. Finally, recent studies support Keith and Carter’s hypothesis that H−*s may play a crucial role in this mechanism by contributing to the formation of the adsorbed catalyst o-DHP*. Computational and experimental studies have both shown that water favorably dissociates on the GaP(110) surface leading to the formation of adsorbed protons with hydride character.21,22 Furthermore, computed acidity constants revealed that adsorbed protons on the surface are extremely stable and are potentially available to be reduced to H−*, which might then react with Py* as illustrated in the mechanism in Figure 1.12 Although the aforementioned experimental and computational results support the heterogeneous mechanism proposed by Carter and co-workers (Figure 1), this mechanism has not been fully validated yet. In this work, we use cluster models with hybrid density functional theory (DFT) to investigate the energetics of the hydride transfer (HT) steps involved in this mechanism. Using computed thermodynamic hydricities (ΔG hyd s), we can determine whether these HTs are thermodynamically viable. Furthermore, we draw important conclusions regarding whether H−*s and o-DHP* formed via H−* transfer can participate in the mechanism of Py-catalyzed CO2 reduction on p-GaP photoelectrodes based on computed reaction free energies (ΔGs) and activation energies (ΔG⧧s).
dihydropyridine (p-DHP)), only the isomer o-DHP is illustrated in Figure 1 and will be discussed in the rest of this paper for simplicity. A similar choice was made in previous work on the basis of the fact that the two isomers have similar thermodynamic properties, and o-DHP is more likely to form on the basis of the proposed mechanism.11,12 The mechanism proposed by Carter and co-workers is supported by several experimental observations. First of all, the electrode dependence of selectivity and applied potential observed in the experiments1−6 points to a heterogeneous mechanism involving species directly adsorbed on the surface. A homogeneous mechanism, such as the one more recently proposed by Musgrave and co-workers10 and the one originally proposed by Bocarsly and co-workers,3 can explain these experimental observations only if the sole role of the electrode is to provide photoexcited electrons with enough energy for the reaction to proceed. However, we recently found that the GaP(110) and the GaP(111) surfaces, which were observed to display different activities toward CO2 reduction,13 provide photoexcited electrons with very similar energy.14 We thus concluded that only a heterogeneous mechanism in which intermediates directly interact with the electrode surface can explain these observations. This conclusion is supported by calculated differences between various intermediates’ adsorption energies on the two surfaces.14 The mechanism in Figure 1 may also explain the observed selectivity toward CO2 reduction versus hydrogen evolution:1 H* is reduced to H−* and reacts with Py* rather than reacting with H+sol to form H2. The catalyst DHP* derived from Py* could explain the observed underpotential that was not observed in the absence of Py.15,16 Experiments also suggest that the acidic environment, and more specifically PyH+sol, plays a role in the catalytic mechanism. In fact, reduction phenomena were only observed under acidic conditions,2 and in the p-GaP experiments the pH of the aqueous solution was set equal to the pKa of PyH+sol (5.2), meaning that 50% of the Py molecules in solution are protonated.1 The need for an acidic solution thus suggests that PyH+sol must play an important role. In the proposed mechanism by Carter and co-workers (Figure 1), PyH+sol has a specific role as the precursor of the reactants necessary for active catalyst formation. In contrast, in the mechanism proposed by Batista and co-workers for Pt electrodes, the only other heterogeneous mechanism published to date, PyH+sol does not have a specific role as it simply acts as a proton donor,7 which any other Brønsted acid could do, despite other acids not catalyzing this reaction. Computed thermodynamic properties such as adsorption free energies, reduction potentials, and acidity constants also support a heterogeneous mechanism involving adsorption of key intermediates. Both Py and o-DHP were found to favorably adsorb on the electrode surface, while CO2 does not.17 These findings support the proposed mechanism in Figure 1, suggesting that an adsorbed cocatalyst is needed to shuttle electrons and protons from the surface to CO2, given that the latter does not adsorb. Moderate computed reduction potentials for Py* reduction to o-DHP* (−0.71 V vs SCE)8,9 and PyH+sol reduction to Py* and H*
2. COMPUTATIONAL APPROACH All calculations were carried out using DFT with the B3LYP23−25 exchange-correlation functional and using the computational chemistry software Orca (version 3.0.3).26 We optimized geometries using a Stuttgart effective core potential (ECP)27,28 for the Ga atoms (ECP28MWB, with 28 electrons replaced by the MWB core potential and a double-ζ valence basis set used for the three remaining valence electrons) and a Pople 6-31G** basis set29,30 for all of the other atoms. We computed single-point energies for the optimized geometries using the same ECP and valence basis set for the Ga atoms and an aug-cc-pVDZ basis set31 for all of the other atoms. The effect of including d polarization functions on the Ga atoms was tested in previous work, and no significant differences in adsorption energies were found.17 Such polarization functions therefore were not used here. We applied Grimme’s D2 dispersion correction32 in all calculations to improve the description of the interaction between adsorbates and the surface. We simulated the effect of the aqueous environment using the continuum Solvation Model based on solute electron Density (SMD).33 This solvation model saves computational time when computing reaction free energies in solution that are needed to calculate several thermodynamic properties (vide B
DOI: 10.1021/acs.jpcc.7b05052 J. Phys. Chem. C XXXX, XXX, XXX−XXX
Article
The Journal of Physical Chemistry C
solution can be directly computed instead of having to use the more time-consuming thermodynamic cycles (as they require more calculations) without repercussion on the accuracy of the results.38,39 The use of the direct approach was also validated in our recent study where we computed the acidity constants of many species relevant to the system under study using thermodynamic cycles as well as the direct approach. We found no significant differences between the two sets of results,12 based on which we chose to employ the direct approach in this work. We used the following expression to calculate ΔGhyd for a generic species XH dissociating a hydride via the reaction XH(aq) → X+(aq) + H−(aq):
infra). In addition, we included a full monolayer of halfdissociated water molecules adsorbed on the cluster surface to better capture solvation effects on the energetics of surface reactions. This adsorption configuration, which had already been employed in our previous work,12 was found to be the most thermodynamically stable and was experimentally observed in a previous combined computational−experimental study.22 We used different numbers of explicit water molecules to calculate ΔGhyds of species in solution, depending on the species under investigation. The number of explicit water molecules used for each species and more details on explicit solvation are provided in sections 1 and 2 of the Supporting Information (SI). We performed a frequency analysis to verify each minimumenergy structure (no imaginary frequency expected) and transition-state (TS) structure (only one imaginary frequency corresponding to the correct reaction coordinate expected). TSs were optimized with the eigenvector-following method as implemented in Orca.34 We further verified the optimized TS structures by relaxing along both directions of the eigenvector corresponding to the imaginary frequency and observing that the structures thus obtained corresponded to the expected initial state (IS) and final state (FS). The computed frequencies were then used to calculate thermochemical corrections at room temperature (298.15 K) via the ideal gas, rigid rotor, and harmonic oscillator approximations. However, only vibrational contributions were considered in free energy calculations involving the GaP cluster because the latter is used to simulate a crystal surface, which has no translational or rotational degrees of freedom. In previous studies,8,9,11,12,14,17,21,22,35 we selected the (110) surface for modeling the p-GaP photocathode because it is the most thermodynamically stable surface of GaP.36 Here, we chose to simulate the GaP(110) surface employing the same cluster model used in our previous work,11,12 which was built starting from a periodic slab and following a procedure we previously established.17 This procedure was validated in our previous study where we benchmarked adsorption energies computed using the GaP(110) cluster model against adsorption energies computed using a periodic slab model of the GaP(110) surface.17 In the present work, we found very small adsorption energy differences (1.7 kcal/mol at most) when simulating adsorption of relevant species (H2O, Py, and CO2) on two different Ga sites of the cluster model. Therefore, we do not expect the chosen adsorption site to affect substantially our predictions. The model consists of 24 Ga atoms, 24 P atoms, and 40 H atoms. The H atoms passivate the covalent dangling bonds left on the bare cluster upon cleavage from the periodic slab. The H atoms are only covering the bottom and side surfaces of the cluster, while the top surface remains free of H atoms to be able to simulate adsorption and surface reactions. The cluster model was used to compute ΔGhyds of H−*s and other adsorbed species, as well as to model H−* transfer reactions. ΔGhyds were computed using an approach similar to the one reported by Marjolin and Keith.37 In this approach, ΔGhyd of a species is simply defined as the reaction free energy in solution for the dissociation of a hydride from this species. Marjolin and Keith computed ΔGhyd using a thermodynamic cycle, which requires the calculation of the reaction free energy in the gas phase and the solvation energy of each species involved in the dissociation equilibrium. However, recent studies have shown that, when using the SMD model, reaction free energies in
ΔG hyd = Gaq (X +) + Gaq (H−) − Gaq (XH)
(1)
The solution free energies of species X+ and XH (Gaq(X+) and Gaq (XH)) were directly calculated by modeling X+ and XH with our computational approach. The free energy of a solvated hydride (Gaq(H−)) was calculated using an expression similar to the one employed by Marjolin and Keith in their work:37 Gaq (H−) = Gg (H 2) − Gaq (H+) + ΔGaq,hetero
(2)
Gaq(H−) thus was obtained by subtracting the free energy of a solvated proton (Gaq(H+)) and adding the energy for the heterolytic dissociation of H2 in aqueous solution (ΔGaq,hetero) to the free energy of H2 in the gas phase (Gg(H2)). Gg(H2) was calculated by modeling a H2 molecule in the gas phase with our computational approach given above. For Gaq(H+), we used the empirical value of −270.3 kcal/mol, which is given by the sum of the empirical gas phase free energy of a proton (−6.3 kcal/ mol), the proton solvation energy (−265.9 kcal/mol), and the energy correction due to the standard state change from gas phase to solution (+1.89 kcal/mol).40 Finally, for ΔGaq,hetero, we used the value recently recommended by Appel and co-workers (34.2 kcal/mol)41 instead of the value employed by Marjolin and Keith (42.1 kcal/mol, which was first reported by Pearson42). Note that our goal in this work is not to reproduce or predict experimental hydricities, but to establish relative hydride donor and acceptor abilities for a series of species relevant to the proposed mechanism. We therefore could have chosen either value of ΔGaq,hetero; however, we chose to use the value recently recommended by Appel and co-workers, which was accurately determined on the basis of values derived from rigorous treatment of experimental data and the use of consistent standard states (i.e., 1 atm for gases and 1 M for species in solution).41 By implementing this approach for ΔGhyd calculations and the value of ΔGaq,hetero recommended by Appel and coworkers,41 ΔGhyd of formate in aqueous solution (HCOO−sol) was found to be equal to 28.4 kcal/mol when using only implicit solvation and 30.6 kcal/mol when using a mixed implicit−explicit solvation approach (see section 2 of the SI for more details on this approach). These values are qualitatively consistent with the value derived from experimental data by Appel and co-workers (24.1 kcal/mol).41 On the basis of our definition of ΔGhyd, the increase in computed HCOO−sol ΔGhyd when including explicit water molecules suggests that the dissociation of a hydride and formation of CO2 is less favored under this condition. This observation can be readily explained when considering that HCOO−, a negatively charged and polar molecule, interacts more strongly than CO2 with the solvating water molecules. C
DOI: 10.1021/acs.jpcc.7b05052 J. Phys. Chem. C XXXX, XXX, XXX−XXX
Article
The Journal of Physical Chemistry C The ΔGs and ΔG⧧s discussed in the main text were computed using a “supermolecule (SM) approach” in which the energy of the IS for a certain reaction is calculated by simulating all reactant species in the same calculation in contact with each other. We also tested a “separate reactants (SR) approach” in which the energy of each individual reactant species is calculated in a separate calculation. We found that the two approaches gave reasonably similar results, with ΔG and ΔG⧧ differences of ∼4 kcal/mol at most. The two approaches and their comparison are further discussed in the Results and Discussion sections and in the SI.
adsorption on P surface sites; the resulting adsorbed protons have hydridic character as they draw electron density away from the P lone pairs.21,22 These results suggest that the main source of adsorbed hydrogen at the GaP(110)/H2O interface is from heterolytically dissociated water. We expect that the applied negative bias and the photoexcited electrons generated under illumination allow for further negative charging of these adsorbed protons. In contrast, the other adsorbed species formed as a result of water dissociation and adsorption (i.e., OH− and H2O) are closed-shell and will not accept further negative charge. Thus, adsorbed protons might take on sufficient negative charge to become hydridic. Given the potential variable charge of these adsorbed protons, it is important to consider the full range of possible reactivity, namely, both acidity and hydricity. We recently computed their acidity and found that it is very low under typical experimental conditions.12 These results motivated investigation of their hydricity here. On the basis of the principle of microscopic reversibility, a proton and two electrons (i.e., a hydride) must be added to our cluster model in order to simulate the reactant of a surface hydride dissociation reaction. We also include explicit solvation of the surface to improve representation of the semiconductor− water interface (see section 1 of the SI for more details on this solvation approach). Because the half-dissociated water adlayer already contains adsorbed protons, we only need to add two electrons to create the conditions for an adsorbed hydride to form. Explicit solvation of the surface previously was found to strongly increase the stability of the adsorbed protons;12 a similar effect might be observed for H−*s. We also tested the effect of adding a variable number of extra electrons to the system to “charge” H−*. The effect of the surface solvation approach (implicit versus mixed implicit−explicit) and the effect of adding extra electrons (zero, one, and two) to the system on the computed H−* ΔGhyd are summarized in Figure 2. The addition of extra electrons causes a large decrease of
3. RESULTS 3.1. Mechanistic Insights from Computed Thermodynamic Hydricities. In this section, we assess whether the hypothesized HTs in the heterogeneous mechanism of Figure 1 are thermodynamically feasible by comparing computed ΔGhyds of relevant species. ΔGhyd is a quantitative measure of the thermodynamic driving force for a certain species to dissociate a hydride: the lower ΔGhyd, the higher the driving force to dissociate a hydride. Relative ΔGhyds therefore can be used to investigate many important mechanistic aspects. First, they can reveal whether it would be thermodynamically feasible for a certain species to donate a hydride to another species. Second, for reaction steps involving both hydride and proton transfers, they can help us understand whether the hydride and proton transfers are likely to occur concertedly or sequentially and, if sequentially, in what order. Third, ΔGhyds can be used to study how the hydride donor/acceptor ability of a certain species is affected by phenomena such as surface adsorption. In the following, ΔGhyds are used to investigate all of these aspects for the two HTs hypothesized in the proposed mechanism in Figure 1. In the first HT, H−* is transferred to Py* along with H+sol resulting in o-DHP* formation. We can assess the feasibility of this first HT by comparing H−* ΔGhyd and o-DHP* ΔGhyd. However, such a comparison is only useful to determine whether o-DHP* can favorably form either via a concerted proton-coupled hydride transfer (PCHT) or with H+sol transfer first followed by H−* transfer. We also computed ΔGhyd of adsorbed, deprotonated o-DHP (2-PyH−*) to explore the possibility of o-DHP* forming via H−* transfer first followed by H+sol transfer. Finally, we computed ΔGhyd of o-DHP in solution (o-DHPsol) in order to answer a question raised by Musgrave and co-workers, which is important for the feasibility of this proposed heterogeneous mechanism:43 Does adsorption of o-DHP cause its hydride donor ability to decrease in such a way that it can no longer transfer a hydride to CO2? The second HT is from o-DHP* to CO2 and leads to either HCOO− or HCOOH formation depending on the pH at the interface; if the pH at the interface is lower (higher) than HCOOH pKa, then HCOOH (HCOO−) will form. We also computed ΔGhyd of both HCOOH and HCOO− since the exact pH at the interface is unknown. HCOO− favorably adsorbs on the GaP(110) surface, while HCOOH adsorption is thermoneutral. 12 Thus, ΔG hyd s of adsorbed formate (HCOO−*), adsorbed formic acid (HCOOH*), and formic acid in solution (HCOOHsol) were calculated. HCOO−sol ΔGhyd was calculated for comparison to experiments (see the Computational Approach section). We first computed H−* ΔGhyd by developing a physically sensible model for H−*. We established previously that dissociative adsorption of water on GaP(110) leads to proton
Figure 2. Effect of solvation approach and addition of extra electrons on the computed thermodynamic hydricity of a surface hydride (H−* ΔGhyd).
H−* ΔGhyd (∼40 kcal/mol decrease when one electron is added and ∼75 kcal/mol decrease when two electrons are added with the mixed solvation approach), meaning that H−* dissociation becomes much more favorable under these conditions. Such a result is expected, because H−* dissociation requires not only an adsorbed proton but also two electrons to dissociate from the surface and any number of electrons less than that would require ionization of the lattice to allow a hydride ion to dissociate. Explicit solvation also causes H−* ΔGhyd to decrease (∼12 kcal/mol decrease when no extra D
DOI: 10.1021/acs.jpcc.7b05052 J. Phys. Chem. C XXXX, XXX, XXX−XXX
Article
The Journal of Physical Chemistry C
HCOOH ΔGhyds to determine whether or not adsorption actually prevents o-DHP from being a good hydride donor for CO2. If the pH at the interface is higher than HCOOH pKa (3.8), then mostly HCOO−* is expected to form upon H−* transfer to CO2. The HCOO−* ΔGhyd was computed using both the “SM approach” and the “separate products approach” because the product of hydride dissociation (CO2) does not favorably adsorb on the surface. The two schemes also give similar results in this case with only ∼4 kcal/mol difference. HCOO−* ΔGhyd is relatively high (43.8 and 39.5 kcal/mol, depending on the approach), suggesting that there is a large driving force for transferring H−* or a hydride from o-DHP* to CO2. Note that adsorption causes a non-negligible increase in HCOO− ΔGhyd (by 8.9 and 13.2 kcal/mol, depending on the approach) due to the large stabilization effect of adsorption on HCOO−. If the pH at the interface is lower than the HCOOH pKa (3.8), then both HCOOHsol and HCOOH* may form. The HCOOH ΔGhyd is much higher (by ∼43 kcal/mol) in solution than when it is adsorbed due to the formation of COOH+, a very unstable species especially in solution, upon hydride dissociation. Overall, whether HCOO−*, HCOOHsol, or HCOOH* is the CO2-reduction product, ΔGhyd for these species is much higher than ΔGhyd for all of the other species considered in this study, thus suggesting that reduction of CO2 by HT is a favorable process. Overall, our computed ΔGhyds indicate that the two HTs hypothesized in the proposed mechanism in Figure 1 are thermodynamically favored. Depending on the scheme used for computing o-DHP* ΔGhyd and HCOO−* ΔGhyd, there is a thermodynamic driving force of at least 21.5 kcal/mol for transferring H−* to Py* in a PCHT step or to PyH+sol (thus forming o-DHP*) and at least 16.7 kcal/mol for transferring a hydride from o-DHP* to CO2. Other important aspects emerge when comparing ΔGhyd values. First, while o-DHP ΔGhyd increases upon adsorption (by more than 13 kcal/mol), meaning that its hydride donor ability decreases, o-DHP* can still favorably donate a hydride to CO2 (16.7 kcal/mol minimum driving force). Second, 2-PyH−* is clearly a better hydride donor than o-DHP* given its much lower ΔGhyd (by at least 23.3 kcal/mol). However, the nearly equivalent 2-PyH−* ΔGhyd and H−* ΔGhyd suggest that there exists no strong driving force to form 2-PyH−* by H−* transfer to Py*, and the previously computed pKa for o-DHP* (13.4)12 suggests that 2PyH−* will likely be protonated subsequently if it forms at all. On the other hand, if 2-PyH−* could be formed via a more favorable pathway and had a long enough lifetime before being protonated, then it could potentially be a better catalyst for CO2 reduction via HT than o-DHP*. We had proposed another possible mechanism in which an adsorbed 2-pyridinyl radical (2-PyH•*) could be the active catalyst.14,44 In this mechanism, CO2 is reduced by an electron shuttled from the surface through 2-PyH•* along with an H atom transferred directly from 2-PyH•*. Subsequent recent work suggests that a stable 2-PyH−* intermediate might form instead (via two photoexcited electrons transferring from the electrode and a solvated proton) and then transfer a hydride to CO2.45 We are currently investigating this possibility on the GaP(110) surface, as we have recently found that 2-PyH−* formation via transfer of photoexcited electrons and a proton from solution is thermodynamically favored on GaP(111), CdTe(111), and CuInS2(112) surfaces.45 Third, given that a thermodynamic driving force of at least 38.2 kcal/mol exists for transferring
electrons are added to the system) due to the extra electron density donated into the surface by H2O* and OH−*. However, explicit solvation no longer has an effect once extra electrons are added to the system because the former does not provide as much additional electron density as added extra electrons do. Overall, we see that H−* ΔGhyd calculated with the mixed solvation approach and two extra electrons added to the system (1.3 kcal/mol) is appropriate for comparison to other computed ΔGhyds. We believe that this is a better representation of H−* as the aqueous environment is properly modeled and microscopic reversibility demands two additional electrons and an adsorbed proton (provided by heterolytically dissociated water) in the model of a hydride. Section 3 of the SI presents further tests to validate this model. Next, we compare H−* ΔGhyd to o-DHP* ΔGhyd and 2PyH−* ΔGhyd to investigate the mechanism of o-DHP* formation via H−* transfer. Note that PyH+sol, which does not favorably adsorb on the surface, is formed when o-DHP* dissociates a hydride. We therefore tested two approaches for computing o-DHP* ΔGhyd: an “SM approach,” in which PyH+sol is modeled in contact with the GaP cluster, and a “separate products approach,” in which the free energies of PyH+sol and the explicitly solvated GaP cluster are computed in two separate calculations. The two approaches give similar results for o-DHP* ΔGhyd (2.4 kcal/mol difference). These results and the other computed ΔGhyds discussed in the following are summarized in Table 1. 2-PyH−* ΔGhyd is very Table 1. Computed Thermodynamic Hydricities (ΔGhyds) in kcal/mol of Relevant Species Based on the Proposed Mechanism Reported in Figure 1a species −
2-PyH * surface hydride o-DHPsol o-DHP* HCOO−sol HCOO−* HCOOH* HCOOHsol
ΔGhyd (kcal/mol) −0.5 1.3 9.6 25.2,b 22.8c 30.6 43.8,b 39.5c 46.6 89.4
* indicates adsorbed species. bValue computed using “supermolecule approach”. cValue computed using “separate products approach”. All ΔGhyd values were computed with the mixed solvation approach except for HCOOHsol ΔGhyd, which was computed with the implicit solvation approach. See sections 1 and 2 of the SI for more details on these solvation approaches. a
similar to H−* ΔGhyd (1.8 kcal/mol difference) and is more than 20 kcal/mol lower than o-DHP* ΔGhyd. These results suggest that there is no thermodynamic driving force to form 2PyH−* via H−* transfer to Py* and that (unsurprisingly) PyH+sol is a much stronger hydride acceptor than Py*. o-DHP* therefore is more likely to form via H−* transfer to PyH+sol or via a concerted PCHT to Py* and is unlikely to form via a 2PyH−* intermediate produced from H−*. In the next section, we will consider both the reaction and the activation energies of these reaction steps to more accurately establish the most favorable pathway for o-DHP* formation via H−* transfer. oDHP* ΔGhyd is more than 13 kcal/mol higher than o-DHPsol ΔGhyd, showing that adsorption indeed reduces o-DHP hydride donor ability as suggested by Musgrave and co-workers.43 However, we need to compare o-DHP* ΔGhyd to HCOO− and E
DOI: 10.1021/acs.jpcc.7b05052 J. Phys. Chem. C XXXX, XXX, XXX−XXX
Article
The Journal of Physical Chemistry C
Figure 3. Computed two-dimensional potential energy surface for the Py* + H−* + H+sol→ o-DHP* reaction. The N−H distance corresponds to the distance between the Py N atom and one of the three protons belonging to H3O+. The C−H distance corresponds to the distance between the Py C atom in the ortho position and H−*. Exact N−H and C−H bond lengths that have been scanned to generate this plot are illustrated in the figure inset. Ga atoms are represented in blue, P atoms in green, C atoms in purple, N atoms in light blue, O atoms in red, and H atoms in off-white. Reaction steps discussed in the main text are indicated by black arrows and labeled with capital letters A, B, C, D, and E.
H−* directly to CO2, one could ask why o-DHP* would be needed to mediate this HT. Previously computed adsorption energies17 suggest that CO2 might not have access to H−*, because other species (such as H2O, Py, and o-DHP) adsorb more strongly than CO2; under these conditions, o-DHP* might facilitate the transfer of H−* to CO2. We conclude on the basis of all these considerations that the DHP*-mediated mechanism (Figure 1) cannot be excluded by our computed ΔGhyds. In the next section, we move on to investigate kinetic aspects of this mechanism. 3.2. Hydride Transfer Reaction Modeling. A full validation of any mechanism requires assessment of both thermodynamic and kinetic aspects. In the previous section, we used ΔGhyds to evaluate the thermodynamics of the full mechanism illustrated in Figure 1. We now move on to investigate the kinetics, focusing on validating two crucial hypotheses this mechanism relies upon: o-DHP* forms via H−* transfer and acts as a hydride shuttle for CO2 reduction to HCOO−. The first hypothesis would be validated if ΔG⧧ for oDHP* formation via H−* transfer is low enough that it can be overcome under experimental conditions. The second hypothesis (i.e., the need of o-DHP* to shuttle H−* from the surface to CO2) would be strongly supported if ΔG⧧ for o-DHP* formation via H−* transfer was lower than ΔG⧧ for directly transferring H−* to CO2. In contrast, if ΔG⧧ for o-DHP* formation via H−* transfer was higher than ΔG⧧ for CO2 reduction via direct H−* transfer, then there would be no point in forming o-DHP* to catalyze the first step of CO2 reduction (i.e., reduction to HCOO−). However, o-DHP* could still be formed and catalyze later steps of CO2 reduction that might have higher ΔG⧧s. Here, we further investigate this aspect by computing ΔG and ΔG⧧ for CO2 reduction to HCOO−* via direct H−* transfer and for o-DHP* formation via H−* transfer. 3.2.1. o-DHP* Formation via Surface Hydride Transfer. In the DHP*-mediated mechanism (Figure 1), we hypothesized that o-DHP* forms via H−* transfer and H+sol transfer to Py*. To further investigate the thermodynamic and kinetic feasibility of these hypothesized reaction steps, we computed a two-
dimensional potential energy surface (PES) for the reaction Py* + H+sol + H−* → o-DHP* (Figure 3). H+sol was modeled as an Eigen cation (i.e., hydronium solvated by three H2O molecules) based on our previous pKa results showing that this model correctly captures the acidity of a proton in solution within our computational approach.12 To form the IS (Py* + H+sol + H−*), the Eigen cation was placed above the explicitly solvated cluster surface with one adsorbed water molecule (H2O*) removed and replaced by Py*. Replacing H2O* with Py* is justified by our previously computed adsorption energies, which show that Py adsorbs more strongly than H2O.11,12 Furthermore, two extra electrons were added to the explicitly solvated cluster model, which already contains an adsorbed proton, to simulate H−*, as dictated by microscopic reversibility (vide supra). This is overall a rather complicated IS for which many different local minima are possible, and a normal optimization algorithm will be unlikely to find the global minimum. We thus extended the capability of the global optimization code OGOLEM46 and used it to identify the global minimum. Sections 4 and 5 of the SI respectively report more details of the global optimization procedure and the PES computation. The computed PES (Figure 3) was used to compare the approximate thermodynamics and kinetics of the three possible mechanisms for the reaction Py* + H+sol + H−* → o-DHP*: a two-step mechanism with H−* transfer to form a 2-PyH−* intermediate followed by H+sol transfer (step A + step B in Figure 3), a concerted mechanism (step C), and a two-step mechanism with H+sol transfer to form a PyH+sol intermediate followed by H−* transfer (step D + step E). The concerted mechanism (step C) is thermodynamically favored by more than 20 kcal/mol compared to either of the preliminary steps of the two-step mechanisms. However, step C has a higher ΔG⧧ than the two-step mechanisms. Thus, we will focus on comparing the approximate energetics derived from the PES for mechanisms A + B and D + E. In fact, there is no need to consider steps B and D in our analysis. Step B is the protonation of 2-PyH−* and has a very small or nonexistent F
DOI: 10.1021/acs.jpcc.7b05052 J. Phys. Chem. C XXXX, XXX, XXX−XXX
Article
The Journal of Physical Chemistry C barrier, in agreement with the large pKa for o-DHP* (13.4) we previously computed.12 Step D is the protonation of Py* to produce PyH+sol, which is present anyway in large concentrations under experimental conditions given the solution’s acidic pH and so does not need to be formed via this pathway; thus, the energetics of PyH+sol formation via step D are not relevant for studying the energetics of o-DHP* formation. We therefore only compare the energetics of steps A and E. Step A is thermoneutral or only slightly thermodynamically favored; in contrast, step E is much more thermodynamically favored (by at least ∼25 kcal/mol). These results are in full agreement with our computed ΔGhyds, showing that there is a much larger driving force for transferring H−* to PyH+sol than to Py*. ΔG⧧s for steps A and E appear to be in the same range. Thus, the kinetics of these steps will be investigated and compared in more depth in the following. Note that we also computed the two-dimensional PES for transferring H−* from the surface and the closest H+sol belonging to a solvating H2O molecule rather than H+sol directly from H3O+ (Figure S5 of the SI). This PES gave a very similar energy landscape to the PES in Figure 3, thus confirming the conclusions we have drawn above: concerted step C is the most kinetically hindered, ΔG⧧s for steps A and E appear to be similar, and transferring any H+sol in the vicinity of 2-PyH−* is barrierless. Step E not only is an alternative pathway for o-DHP* formation with respect to the one originally proposed (Py* + H+sol + H−* → o-DHP*)8,9 but also represents a new pathway for PyH+sol reduction that has not been proposed so far. Thus, before computing the energetics of step E and comparing it to the energetics of step A, we must first assess whether PyH+sol will actually be available to react via step E instead of reacting via alternative, more favorable reduction pathways. To this end, we computed the reduction potential for reducing PyH+sol to oDHP* (via H+sol and two electron transfers, possibly as an H−*) with our mixed implicit−explicit solvation approach for the cluster surface and compared it to the reduction potentials of previously identified PyH+sol reduction pathways (Table 2).
solvation to be able to make truly direct comparisons. We found that the values changed by 0.06 V at most, preserving the same trend mentioned above, namely, that PyH+sol + 1H+sol + 2e− → o-DHP* remains the most favorable pathway for PyH+ reduction. All the newly reported reduction potentials were computed with the same method employed in our previous work (see detailed explanation in the SI of ref 11). We note that this method was validated in our group’s first study on this system by computing the reduction potentials for a series of substituted pyridinium species and comparing them to the experimental values.18 The calculated reduction potentials agree to within 0.2 V of experiment. Furthermore, as discussed in that study and in references therein, the uncertainty in reduction potentials computed in this manner is expected to be less than ∼0.3 V. As discussed in the Introduction, PyH+sol is present in significant concentration in the experimental system, and experimental observations suggest that it is essential for the catalysis. In the proposed mechanism in Figure 1, PyH+sol is the source of Py* and H*, both of which are needed to form the active catalyst o-DHP*. Here, we propose an alternative role for PyH+sol that may better explain experimental observations: PyH+sol might react with H−* to directly form the proposed active catalyst o-DHP*. Next, we compare ΔG and ΔG⧧ for forming o-DHP* via either step A or E computed using the “SM approach”, as reported in Table 3. The IS, TS, and FS used to calculate these
Table 2. Computed Reduction Potentials (E0s) at pH 5.2 for Possible PyH+sol Reduction Pathwaysa
Similar results obtained with the “separate reactants approach” are reported in Table S2 of the SI. * indicates adsorbed species. Subscript “sol” indicates species in solution. CO2*perp indicates that the reactant CO2 interacts with one Ga surface site in the perpendicular orientation. CO2*parall indicates that the reactant CO2 interacts with two Ga surface sites in the parallel orientation. See the main text and section 10 of the SI for more details on these orientations.
reaction PyH+sol PyH+sol PyH+sol PyH+sol
+ + + +
1H+sol + 2e− → o-DHP* 1e− → Py* + H* 1e− → 2-PyH•* 1e− → 1-PyH•
Table 3. Reaction Free Energies (ΔGs) and Activation Energies (ΔG⧧s) for Relevant Reactions Discussed in the Main Text Computed Using the “Supermolecule Approach”a reaction −
−
Py* + H * → 2-PyH * PyH+sol + H−* → o-DHP* CO2,sol + H−* → HCOO−sol CO2*,perp + H−* → HCOO−* CO2*parall + H−* → HCOO−*
ΔG (kcal/mol)
ΔG⧧ (kcal/mol)
+0.7 −19.9 −27.8 −37.5 −42.4
+45.3 +32.1 +32.2 +31.3 +27.0
a
E0 (V vs SCE) at pH = 5.2 −0.73 (−0.79) −0.91 (−0.85) −1.29 (−1.31) −1.44
* indicates adsorbed species. Subscript “sol” indicates species in solution. E0s for heterogeneous reduction pathways were computed using our mixed implicit−explicit solvation approach for the cluster surface, and E0s obtained with only implicit solvation are reported in parentheses. PyH+sol + 1e− → Py* + H* E0 and PyH+sol + 1e− → 2PyH•* E0 computed with only implicit solvation are taken from refs 11 and 14, respectively. PyH+sol + 1e− → 1-PyH• E0 (taken from ref 18) was computed with an implicit solvation approach and one explicit water molecule. a
values are shown in Figures 4 and 5, and are further discussed in sections 7 and 8 of the SI. The computed ΔGs reveal that step A is nearly thermoneutral (ΔG = +0.7 kcal/mol), while step E is thermodynamically favored (ΔG = −19.9 kcal/mol), consistent with the computed ΔGhyds and the computed PES. The ΔG estimates that one can derive from ΔGhyd differences (e.g., difference between H−* ΔGhyd and 2-PyH−* ΔGhyd to estimate step A ΔG) slightly deviate from the computed ΔGs reported in Table 3. This is because ΔGhyd differences reveal how much energy is lost or gained by taking a hydride from a species (e.g., H−*) and giving it to another species (e.g., Py*) when these two species are not in contact. In contrast, ΔGs reported in Table 3 were computed by simulating the HT between the two explicitly interacting species and should be more reliable. The computed ΔG⧧s (+45.3 kcal/mol for step A and +32.1 kcal/mol for step E) suggest that o-DHP* formation via H−* transfer to PyH+sol is not only more thermodynamically
On the basis of the computed reduction potentials, PyH+sol + 1H+sol + 2e− → o-DHP* is the most favorable PyH+sol reduction pathway. Note that the reduction potentials for the alternative heterogeneous PyH+sol reduction pathways were computed with only implicit solvation in previous work.11,14 We therefore recomputed these reduction potentials with our mixed solvation approach, and we also computed the reduction potential for reducing PyH+sol to o-DHP* with only implicit G
DOI: 10.1021/acs.jpcc.7b05052 J. Phys. Chem. C XXXX, XXX, XXX−XXX
Article
The Journal of Physical Chemistry C
Figure 4. Side view of the initial state (a), transition state (b), and final state (c) used to calculate the energetics of the Py* + H−* → 2-PyH−* reaction. Ga atoms are represented in blue, P atoms in green, C atoms in purple, N atoms in light blue, O atoms in red, and H atoms in off-white.
Figure 5. Side view of the initial state (a), transition state (b), and final state (c) used to calculate the energetics of the PyH+ + H−* → o-DHP* reaction. Ga atoms are represented in blue, P atoms in green, C atoms in purple, N atoms in light blue, O atoms in red, and H atoms in off-white.
increase in ΔG⧧ and ΔG with no changes to our conclusions (see Table S2 and section 9 of the SI for a more detailed comparison of the two approaches). In summary, step E (i.e., PyH+sol + H−* → o-DHP*) is the most favorable pathway for o-DHP* formation from H−* based on our computed PES and ΔG⧧s; however, o-DHP* will be unlikely to form via this pathway unless its ΔG⧧ can be lowered under applied potential or overcome by photoexcitation. Independent from whether this ΔG⧧ can be lowered, we must compare it next to ΔG⧧ for transferring H−* directly to CO2 in order to understand whether the formation of o-DHP* via H−* transfer can be a step in the mechanism of CO2 reduction on p-GaP electrodes. 3.2.2. CO2 Reduction via Surface Hydride Transfer. To model CO2 reduction to HCOO− via H−* transfer, we placed a CO2 molecule above the explicitly solvated cluster surface. Furthermore, we added two extra electrons to the explicitly solvated cluster model, which already contains an adsorbed proton, in order to simulate H−*, as dictated by microscopic reversibility (vide supra). CO2 does not favorably adsorb, while the product HCOO− adsorbs more favorably than H2O via two dative bonds between its oxygen atoms and two Ga surface sites.12 We thus modeled the CO2 + H−* → HCOO− reaction on the explicitly solvated cluster surface with and without free adsorption sites to get a more accurate estimate of its
favored but also more kinetically favored than o-DHP* formation via a 2-PyH−* intermediate (formed by H−* transfer to Py*). Note that the computed PES (Figure 3) suggests that step A and step E have similar barriers in the range of 20 kcal/ mol, which is different from the trend revealed by the computed ΔG⧧s. This difference is attributable to the optimization approach for the structures used to compute the PES. First, these structures were optimized with looser convergence criteria to reduce computational cost; in contrast, the structures used to calculate ΔGs and ΔG⧧s were optimized with tighter energy criteria, thus leading to lower energy minima, which were confirmed with a vibrational frequency analysis. Second, all of these structures were optimized with a normal optimization algorithm; in contrast, when computing ΔG⧧s, we employed a TS optimization algorithm to identify TS structures, which were confirmed with a vibrational frequency analysis. Furthermore, using properly optimized TS structures also leads to a ΔG⧧ trend consistent with the ΔGs, in agreement with Hammond’s postulate: step A, which is nearly thermoneutral, has a higher barrier than step E, which is thermodynamically favored. However, ΔG⧧ for step E is still very high and suggests that this process will be kinetically hindered. Using the “SR approach” to calculate step E energetics resulted in a small H
DOI: 10.1021/acs.jpcc.7b05052 J. Phys. Chem. C XXXX, XXX, XXX−XXX
Article
The Journal of Physical Chemistry C
Figure 6. Side view of the initial state (a), transition state (b), and final state (c) used to calculate the energetics of the CO2 + H−* → HCOO−sol reaction on the fully solvated cluster surface. Ga atoms are represented in blue, P atoms in green, C atoms in purple, N atoms in light blue, O atoms in red, and H atoms in off-white.
the computed ΔGs are attributable to the fact that the latter is computed by simulating the HT between two interacting species while ΔGhyds are computed for the isolated species. ΔG is especially favorable (∼40 kcal/mol depending on CO2 orientation) with two free Ga adsorption sites due to the large adsorption energy of the product HCOO−*. The ΔG⧧ values are all in the same range (∼30 kcal/mol) as ΔG⧧ computed for transferring H−* to PyH+sol. We found a slightly lower ΔG⧧ (27.0 kcal/mol) when CO2 is allowed to interact with two Ga surface sites in the parallel orientation. However, this orientation is less favored, and CO2 would not actually have access to Ga surface sites in an aqueous environment. Also, in this case, using the “separate-reactants approach” resulted in both ΔGs and ΔG⧧s being slightly less favorable but preserved the overall trend: except for Py* + H−* → 2-PyH−*, all investigated reactions are thermodynamically favored and have ΔG⧧s all in the same range (∼30 kcal/mol).
energetics. IS, TS, and FS for the reaction modeled without free adsorption sites are illustrated in Figure 6. IS, TS, and FS for the reaction modeled with free adsorption sites are illustrated and discussed in section 10 of the SI. We removed one H2O* as well as the adsorbed hydroxide (OH−*) hydrogen-bonded to it to free up the two surface sites needed for HCOO− adsorption. Note that while OH− adsorbs more strongly than HCOO−, OH−* is only stable when forming a hydrogen-bond with a nearby H2O*; if the nearby H2O* is removed (to free the first adsorption site needed for HCOO− adsorption), then OH−* will get protonated. The newly formed H2O* thus can be removed to free up the second adsorption site needed for HCOO− adsorption. Furthermore, we also removed the proton adsorbed next to OH−* to remove an overall neutral species. Note that modeling the reaction with and without free adsorption sites also would be required to study the PyH+sol + H−* → o-DHP* reaction discussed in the previous section, given that PyH+sol does not favorably adsorb while o-DHP adsorbs more strongly than H2O. However, PyH+sol is bulkier than CO2 and thus will not be able to approach free adsorption sites because of the steric hindrance from the surrounding H2O* and OH−*. ΔG and ΔG⧧ values computed for the CO2 + H−* → HCOO− reaction (with and without free adsorption sites) using the “SM approach” are reported in Table 3, where they are compared to the energetics for the Py* + H−* → 2-PyH−* and PyH+sol + H−* → o-DHP* reactions. Note that two possible orientations (parallel and perpendicular) for CO2 interacting with the two free Ga adsorption sites were considered, since we found that they lead to different TSs and FSs. In the parallel orientation, both CO2 oxygen atoms interact with the surface and CO2 lies along a Ga row. In the perpendicular orientation, only one CO2 oxygen atom interacts with the surface, and CO2 lies perpendicular to the Ga rows. The two CO2 orientations and the respective TSs and FSs are discussed in section 10 of the SI. The perpendicular orientation is slightly energetically favored (by 1.6 kcal/mol). The computed ΔGs are consistent with the computed ΔGhyds: H−* transfer to CO2 is more thermodynamically favored than H−* transfer to PyH+sol. As mentioned earlier, quantitative differences between the ΔG estimates derived from ΔGhyds and
4. DISCUSSION Overall, the computed reaction energetics for o-DHP* formation via H−* transfer and CO2 reduction via H−* transfer lead to two important conclusions. First, H−*s are unlikely to play a role in the catalytic mechanism. Second, there is no point in forming o-DHP* via H−* transfer to catalyze the first CO2 reduction step. The fact that high ΔG⧧s were computed for transferring H−* to PyH+sol, Py*, and CO2 suggests that H−* transfer away from the surface is kinetically hindered. This finding is in agreement with previous measurements showing that hydride-like adsorbed protons on GaP(110) are strongly bound to the surface.22 The high ΔG⧧ calculated for reducing CO2 via H−* transfer is also consistent with the high overpotentials observed for the photoelectrocatalytic reduction of CO2 on p-GaP electrodes in the absence of Py.15,16 This finding suggests that, while H−*s are unlikely to play a role in the Py-catalyzed CO2 reduction mechanism, they might be the source of protons and electrons in the mechanism at work when Py is not present. Two results strongly suggest that o-DHP* is unlikely to act as a catalyst for CO2 reduction to HCOO− if it has to form via H−* transfer. First, our computed PES and ΔG⧧s show that if I
DOI: 10.1021/acs.jpcc.7b05052 J. Phys. Chem. C XXXX, XXX, XXX−XXX
The Journal of Physical Chemistry C
■
o-DHP* is formed via H−* transfer, then it will most favorably form via H−* transfer to PyH+sol, a species that does not favorably adsorb on the surface. However, we previously argued that CO2 cannot be directly reduced by H−* transfer because it does not favorably adsorb on the surface and thus will have limited access to H−*. On the basis of this same reasoning, PyH+sol should also have limited access to H−*, and we must conclude that o-DHP* is unlikely to form via H−* transfer to PyH+sol. Second, ΔG⧧ for o-DHP* formation via H−* transfer to PyH+sol and ΔG⧧ for CO2 reduction via H−* transfer are very similar. Thus, there is no point in forming o-DHP* to catalyze the first step of CO2 reduction given that these two processes will have a similar energy cost; however, if these ΔG⧧s could somehow be reduced and PyH+sol could access H−*, then o-DHP* might form and catalyze later steps of CO2 reduction with potentially higher barriers. Alternatively, oDHP* might form via a different mechanism with a lower ΔG⧧. Or a different intermediate, whose formation does not require H−* transfer, might act as catalyst in this system. As an example, the recently proposed 2-PyH•* intermediate14 or the newly introduced 2-PyH−* intermediate45 could play this role. Our computed ΔGhyds suggest that 2-PyH−* would be a very strong hydride donor and could be a powerful catalyst for CO2 reduction, assuming that it can be favorably formed via a mechanism other than H−* transfer to Py* and has a long enough lifetime. We are currently investigating alternative mechanisms for 2-PyH−* formation and reaction with CO2 on GaP(110).
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. ORCID
Emily A. Carter: 0000-0001-7330-7554 Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS The authors acknowledge financial support from the Air Force Office of Scientific Research under AFOSR Awards FA9550-101-0572 and FA9550-14-1-0254. The authors thank Dr. Thomas P. Senftle and Ms. Nari L. Baughman for helpful feedback during manuscript preparation.
■
REFERENCES
(1) Barton, E. E.; Rampulla, D. M.; Bocarsly, A. B. Selective SolarDriven Reduction of CO2 to Methanol Using a Catalyzed p-GaP Based Photoelectrochemical Cell. J. Am. Chem. Soc. 2008, 130, 6342−6344. (2) Lin, C.; Seshadri, G.; Bocarsly, A. B. A New Homogeneous Electrocatalyst for the Reduction of Carbon Dioxide to Methanol at Low Overpotential. J. Electroanal. Chem. 1994, 372, 145−150. (3) Cole, E. B.; Lakkaraju, P. S.; Rampulla, D. M.; Morris, A. J.; Abelev, E.; Bocarsly, A. B. Using a One-Electron Shuttle for the Multielectron Reduction of CO2 to Methanol: Kinetic, Mechanistic, and Structural Insights. J. Am. Chem. Soc. 2010, 132, 11539−11551. (4) Jeon, J. H.; Mareeswaran, P. M.; Choi, C. H.; Woo, S. I. Synergism between CdTe Semiconductor and Pyridine − Photoenhanced Electrocatalysis for CO2 Reduction to Formic Acid. RSC Adv. 2014, 4, 3016−3019. (5) Yuan, J.; Hao, C. Solar-Driven Photoelectrochemical Reduction of Carbon Dioxide to Methanol at CuInS2 Thin Film Photocathode. Sol. Energy Mater. Sol. Cells 2013, 108, 170−174. (6) Yuan, J.; Zheng, L.; Hao, C. Role of Pyridine in Photoelectrochemical Reduction of CO2 to Methanol at a CuInS2 Thin Film Electrode. RSC Adv. 2014, 4, 39435−39438. (7) Ertem, M. Z.; Konezny, S. J.; Araujo, C. M.; Batista, V. S. Functional Role of Pyridinium during Aqueous Electrochemical Reduction of CO2 on Pt(111). J. Phys. Chem. Lett. 2013, 4, 745−748. (8) Keith, J. A.; Carter, E. A. Theoretical Insights into Electrochemical CO2 Reduction Mechanisms Catalyzed by Surface Bound Nitrogen Heterocycles. J. Phys. Chem. Lett. 2013, 4, 4058−4063. (9) Keith, J. A.; Carter, E. A. Correction to “Theoretical Insights into Electrochemical CO2 Reduction Mechanisms Catalyzed by SurfaceBound Nitrogen Heterocycles. J. Phys. Chem. Lett. 2015, 6, 568−568. (10) Lim, C.-H.; Holder, A. M.; Hynes, J. T.; Musgrave, C. B. Reduction of CO2 to Methanol Catalyzed by a Biomimetic OrganoHydride Produced from Pyridine. J. Am. Chem. Soc. 2014, 136, 16081− 16095. (11) Lessio, M.; Carter, E. A. What Is the Role of Pyridinium in Pyridine-Catalyzed CO2 Reduction on p-GaP Photocathodes? J. Am. Chem. Soc. 2015, 137, 13248−13251. (12) Lessio, M.; Riplinger, C.; Carter, E. A. Stability of Surface Protons in Pyridine-Catalyzed CO2 Reduction at p-GaP Photoelectrodes. Phys. Chem. Chem. Phys. 2016, 18, 26434−26443. (13) Hu, Y. Solar Fuel Generation on Semiconductors: Photo-Assisted H2 Evolution on a Novel Delafossite AgRhO2 and a p-GaP (111) Surface With {110} Faces Revealed by Etching for CO2 Reduction. Doctoral Thesis, Princeton University, 2015. (14) Lessio, M.; Senftle, T. P.; Carter, E. A. Is the Surface Playing a Role during Pyridine-Catalyzed CO2 Reduction on p-GaP Photoelectrodes? ACS Energy Lett. 2016, 1, 464−468. (15) Halmann, M. Photoelectrochemical Reduction of Aqueous Carbon Dioxide on p-Type Gallium Phosphide in Liquid Junction Solar Cells. Nature 1978, 275, 115−116.
5. SUMMARY AND CONCLUSIONS In this work, we computed hydricities (ΔGhyds) of relevant intermediates and energetics of relevant reaction steps to test a previously proposed mechanism for Py-catalyzed CO2 reduction on p-GaP electrodes. The computed ΔGhyds support this mechanism, which relies on H−* transfer to generate the proposed active catalyst o-DHP*. We found that o-DHP* formation via concerted H+sol and H−* transfers to Py* is less energetically favored than o-DHP* formation via H−* to Py* (followed by spontaneous protonation of 2-PyH−*) or H−* transfer to PyH+sol. However, the overall high ΔGs calculated for transferring H−*s to either Py* (to form 2-PyH−*), PyH+sol (to form o-DHP*), or CO2 (to form HCOO−) suggest that H−*s are unlikely to participate in this chemistry as their transfer away from the surface is severely kinetically hindered. This result also implies that the proposed active catalyst oDHP* is either formed via a more kinetically favored mechanism yet to be identified or, if formed at all via H−* transfer, that it plays a role only in later steps of the CO2 reduction mechanism. Alternatively, another Py-derived intermediate whose formation does not require H−* transfer might be the active catalyst in this system; the recently proposed 2PyH−* species formed from two photoexcited electrons, Py*, and a solvated proton is one we continue to investigate in this regard.
■
Article
ASSOCIATED CONTENT
* Supporting Information S
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jpcc.7b05052. Further calculation details, additional results, and Cartesian coordinates of the structures discussed in this work (PDF) J
DOI: 10.1021/acs.jpcc.7b05052 J. Phys. Chem. C XXXX, XXX, XXX−XXX
Article
The Journal of Physical Chemistry C (16) Aurian-Blajeni, B.; Halmann, M.; Manassen, J. Electrochemical Measurement on the Photoelectrochemical Reduction of Aqueous Carbon Dioxide on p-Gallium Phosphide and p-Gallium Arsenide Semiconductor Electrodes. Sol. Energy Mater. 1983, 8, 425−440. (17) Keith, J. A.; Muñoz-García, A. B.; Lessio, M.; Carter, E. A. Cluster Models for Studying CO2 Reduction on Semiconductor Photoelectrodes. Top. Catal. 2015, 58, 46−56. (18) Keith, J. A.; Carter, E. A. Theoretical Insights into PyridiniumBased Photoelectrocatalytic Reduction of CO2. J. Am. Chem. Soc. 2012, 134, 7580−7583. (19) Tossell, J. A. Calculation of the Properties of Molecules in the Pyridine Catalyst System for the Photochemical Conversion of CO2 to Methanol. Comput. Theor. Chem. 2011, 977, 123−127. (20) Lim, C.-H.; Holder, A. M.; Musgrave, C. B. Mechanism of Homogeneous Reduction of CO2 by Pyridine: Proton Relay in Aqueous Solvent and Aromatic Stabilization. J. Am. Chem. Soc. 2013, 135, 142−154. (21) Muñoz-García, A. B.; Carter, E. A. Non-Innocent Dissociation of H2O on GaP(110): Implications for Electrochemical Reduction of CO2. J. Am. Chem. Soc. 2012, 134, 13600−13603. (22) Kronawitter, C. X.; Lessio, M.; Zhao, P.; Riplinger, C.; Boscoboinik, A.; Starr, D. E.; Sutter, P.; Carter, E. A.; Koel, B. E. Observation of Surface-Bound Negatively Charged Hydride and Hydroxide on GaP(110) in H2O Environments. J. Phys. Chem. C 2015, 119, 17762−17772. (23) Becke, A. D. Density-Functional Exchange-Energy Approximation with Correct Asymptotic Behavior. Phys. Rev. A: At., Mol., Opt. Phys. 1988, 38, 3098. (24) Lee, C.; Yang, W.; Parr, R. G. Development of the Colle-Salvetti Correlation-Energy Formula into a Functional of the Electron Density. Phys. Rev. B: Condens. Matter Mater. Phys. 1988, 37, 785−789. (25) Becke, A. D. Density Functional Thermochemistry. III. The Role of Exact Exchange. J. Chem. Phys. 1993, 98, 5648−5652. (26) Neese, F. The ORCA Program System. Wiley Interdiscip. Rev. Comput. Mol. Sci. 2012, 2, 73−78. (27) Bergner, A.; Dolg, M.; Küchle, W.; Stoll, H.; Preuß, H. Ab Initio Energy-Adjusted Pseudopotentials for Elements of Groups 13−17. Mol. Phys. 1993, 80, 1431−1441. (28) Leininger, T.; Berning, A.; Nicklass, A.; Stoll, H.; Werner, H.-J.; Flad, H.-J. Spin-Orbit Interaction in Heavy Group 13 Atoms and TlAr. Chem. Phys. 1997, 217, 19−27. (29) Hariharan, P. C.; Pople, J. A. The Influence of Polarization Functions on Molecular Orbital Hydrogenation Energies. Theor. Chim. Acta 1973, 28, 213−222. (30) Francl, M. M.; Pietro, W. J.; Hehre, W. J.; Binkley, J. S.; Gordon, M. S.; DeFrees, D. J.; Pople, J. A. Self-Consistent Molecular Orbital Methods. XXIII. A Polarization-Type Basis Set for Second-Row Elements. J. Chem. Phys. 1982, 77, 3654−3665. (31) Dunning, T. H. Gaussian Basis Sets for Use in Correlated Molecular Calculations. I. The Atoms Boron through Neon and Hydrogen. J. Chem. Phys. 1989, 90, 1007−1023. (32) Grimme, S. Semiempirical GGA-Type Density Functional Constructed with a Long-Range Dispersion Correction. J. Comput. Chem. 2006, 27, 1787−1799. (33) Marenich, A. V.; Cramer, C. J.; Truhlar, D. G. Universal Solvation Model Based on Solute Electron Density and on a Continuum Model of the Solvent Defined by the Bulk Dielectric Constant and Atomic Surface Tensions. J. Phys. Chem. B 2009, 113, 6378−6396. (34) Baker, J. An Algorithm for the Location of Transition States. J. Comput. Chem. 1986, 7, 385−395. (35) Kronawitter, C. X.; Lessio, M.; Zahl, P.; Muñoz-García, A. B.; Sutter, P.; Carter, E. A.; Koel, B. E. Orbital-Resolved Imaging of the Adsorbed State of Pyridine on GaP(110) Identifies Sites Susceptible to Nucleophilic Attack. J. Phys. Chem. C 2015, 119, 28917−28924. (36) Hayashi, K.; Ashizuka, M.; Bradt, R. C.; Hirano, H. Cleavage of Gallium Phosphide. Mater. Lett. 1982, 1, 116−118.
(37) Marjolin, A.; Keith, J. A. Thermodynamic Descriptors for Molecules That Catalyze Efficient CO2 Electroreductions. ACS Catal. 2015, 5, 1123−1130. (38) Ribeiro, R. F.; Marenich, A. V.; Cramer, C. J.; Truhlar, D. G. Use of Solution-Phase Vibrational Frequencies in Continuum Models for the Free Energy of Solvation. J. Phys. Chem. B 2011, 115, 14556− 14562. (39) Ho, J. Are Thermodynamic Cycles Necessary for Continuum Solvent Calculation of pKas and Reduction Potentials? Phys. Chem. Chem. Phys. 2015, 17, 2859−2868. (40) Keith, J. A.; Carter, E. A. Quantum Chemical Benchmarking, Validation, and Prediction of Acidity Constants for Substituted Pyridinium Ions and Pyridinyl Radicals. J. Chem. Theory Comput. 2012, 8, 3187−3206. (41) Connelly, S. J.; Wiedner, E. S.; Appel, A. M. Predicting the Reactivity of Hydride Donors in Water: Thermodynamic Constants for Hydrogen. Dalt. Trans. 2015, 44, 5933−5938. (42) Pearson, R. G. Ionization Potentials and Electron Affinities in Aqueous Solution. J. Am. Chem. Soc. 1986, 108, 6109−6114. (43) Lim, C.-H.; Holder, A. M.; Hynes, J. T.; Musgrave, C. B. Catalytic Reduction of CO2 by Renewable Organohydrides. J. Phys. Chem. Lett. 2015, 6, 5078−5092. (44) Senftle, T. P.; Lessio, M.; Carter, E. A. Interaction of Pyridine and Water with the Reconstructed Surfaces of GaP(111) and CdTe(111) Photo-Electrodes: Implications for CO2 Reduction. Chem. Mater. 2016, 28, 5799−5810. (45) Senftle, T. P.; Lessio, M.; Carter, E. A. Submitted 2017. (46) Dieterich, J. M.; Hartke, B. OGOLEM: Global Cluster Structure Optimisation for Arbitrary Mixtures of Flexible Molecules. A Multiscaling, Object-Oriented Approach. Mol. Phys. 2010, 108, 279− 291.
K
DOI: 10.1021/acs.jpcc.7b05052 J. Phys. Chem. C XXXX, XXX, XXX−XXX
