Deconstructing the Enhancing Effect on CO2 Activation in the Electric Double Layer with EVB Dynamic Reaction Modeling

Xiaoyu Chen and Mårten S. G. Ahlquist*

Cite This:https://dx.doi.org/10.1021/acs.jpcc.0c05974 Read Online

ACCESS

Metrics & More Article Recommendations

*

^sı Supporting Information

ABSTRACT: The reactivities of the same molecular electrocatalyst under homogeneous and heterogeneous conditions can be dramatically di fferent, highlighting that the reaction environment plays an important role in catalysis.

For catalysis on solid electrodes, reactions take place in the electric double layer (EDL), where a strong electric field is experienced. In this work, empirical valence bond molecular dynamics (EVB-MD) was used to explore CO

₂

binding in the EDL.

It allows explicit descriptions of the solvent, electrolyte, catalyst −reactant, and the electrode surface material, as well as an unbiased description of the applied electric field. The strong local electric field concentrates cations, which in turn stabilizes the bound CO

₂

. Furthermore, controlled computational experiments suggest that neither the electric field nor the cations alone can produce significant stabilization, but that the combination leads to a dramatic stabilization of the CO

₂

bound state.

■ INTRODUCTION

It is well known that homogeneous catalysts o ffer the possibility of molecular engineering, making high activity and selectivity

¹⁻³

possible. However, the general use of organic solvents in homogeneous catalysis is contrary to the concept of green chemistry and can be expensive at a larger scale, without e fficient recycling. Heterogeneous catalysts, such as transition metal oxides

⁴

and metal −organic frameworks (MOF),

⁵

make catalysis in water possible, but the fine tuning of their chemical properties is more complicated, which limits their applications.

Carbon materials with large surface areas, such as carbon nanotubes (CNTs),

⁶

graphene,

⁷

and even carbon black,

⁸

act as good supporting materials for molecular catalysts, enabling them to function in reaction media such as water, which is otherwise impossible. Due to their high electrical conductivity, such catalyst/support combination has been found to create excellent electrocatalytic materials.

^9,10

The easiest strategy for immobilization is physical adsorption, which relies on the π−π contact and/or electrostatic interactions between the catalyst and the electrode, either with or without an extra supporting material.

¹¹

Several studies

^6,12,13

demonstrate the same trend;

when the catalysts are immobilized on a carbon supporting material,

¹⁴

such as CNTs or a graphene nano fiber, and hence react in a heterogeneous fashion, their performances, in terms of turnover frequency (TOF), increased dramatically. In speci fic, it was reported in 2017 that simply adhering Co(TPP) on single-walled CNTs (sw-CNTs) can increase the TOF by a factor of 300.

¹²

The performance of a similar catalyst, heterogeneous CoPc/graphitic carbon, was more than 3 orders of magnitude more active than molecular CoPc under homogeneous conditions.

¹⁵

These experimental observations suggest that the presence of a carbon supporting material, as well as the change in the reaction medium, leads to a dramatic improvement in reactivity,

¹¹

while the di fference in reactivity caused by carbon supporting materials, varying from sw-CNT

¹²

to graphene,

¹⁵

is not as signi ficant.

Inspired by Hu et al. ’s

¹²

work on Co(TPP), where they observed a signi ficant change in performance between homogeneous catalysis in N,N-dimethylformamide (DMF) and heterogeneous catalysis in water, we conducted a theoretical study

¹⁶

on the reaction mechanism and found that water facilitates the binding of CO

₂

to Co

^I

(TPP)

⁻

, as well as the dissociation of hydroxide in a later step. For the CO

₂

binding step, it is more favored in water over DMF, with an activation free energy di fference of 2.8 kcal/mol. CO

2

binding together with proton-assisted reduction and hydroxide dissociation involves states and transition states that are all higher than the initial state in free energy. In the catalytic cycle, the C −O dissociation at Co

^I

(TPP) −COOH is the rate- determining step

¹⁶

(Figure 1). Since all the three steps contribute to the activation energy, an easier CO

₂

binding could lead to a lowered activation energy if the other steps remain unchanged.

Received: June 30, 2020 Revised: September 16, 2020 Published: September 21, 2020

© XXXX American Chemical Society A

https://dx.doi.org/10.1021/acs.jpcc.0c05974 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Downloaded via 85.227.128.122 on October 5, 2020 at 22:53:44 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

We also found that the catalyst −nanotube interaction was important: a stronger tendency for aggregation was found for smaller nanotubes due to the better catalyst −catalyst π−π interactions as compared to the catalyst −smaller CNT interactions. Hence, CNTs ’ curvature affects the aggregation tendency and therefore explains the reverse loading effect reported by Hu et al. Consequently, flatter nanotubes are preferred since they reduce the tendency for aggregation, leaving more catalysts accessible for CO

₂

binding.

When studying electrocatalysis under heterogeneous con- ditions with a hydrophobic supporting material (Figure 2), three key e ffects need to be addressed: (1) the electric field e ffectsince the reactions take place on the electrode surface in the electric double layer (EDL), the catalyst experiences a much stronger electric field than in solution; (2) the surface e ffectthe catalyst now adheres to a hydrophobic surface, which could a ffect the properties of the catalyst and all of the intermediates involved in the catalytic cycle; (3) the solvent e ffectsince it is no longer a requirement to dissolve the catalyst, organic solvents, such as DMF or dimethyl sulfoxide (DMSO), can be changed to the more environmentally friendly, and more polar water, which also forms stronger hydrogen bonds. While the solvent polarity can be readily modeled with reasonable accuracy using standard DFT methods with an implicit solvation model and possibly a few explicit solvent molecules, the e ffects of (1) the hydrophobic surface −solvent interface on the reactions and the (2) electric

fields are challenging to understand, not to mention the combined e ffect of all the three factors mentioned earlier.

We reasoned that all the three e ffects could have significant impacts and we wanted to take the full reaction environment into account in an explicit manner without making too many presumptions. To address these issues, we constructed a molecular dynamics-based empirical valence bond (MD- EVB)

¹⁷⁻¹⁹

model, which allows us to calculate the free-energy pro files for the bond formation processes, at the same time considering all of the intermolecular interactions explicitly throughout the reaction. The original EVB formulation

¹⁷⁻¹⁹

was developed by Warshel and co-workers in the 1980s, since then it was proved to be a very e fficient method

^20,21

for exploring the catalytic free-energy landscape for both enzyme catalysis

^22,23

and recently organometallics.

^24,25

The method is based on force field descriptions of a reactant state (blue) Hamiltonian (H

₁

) and a product state (green) Hamiltonian (H

₂

). The (to-be- fitted) off-diagonal term (H

12

) is a phase-independent

²⁶

coupling between the intersection of the reactant state Hamiltonian (H

₁

) and the product state Hamiltonian (H

₂

) and the actual transition state (TS) (Figure 3). Therefore, with well-parameterized force fields and H

12

, the EVB model can provide a QM description of the reaction system within a valence bond framework.

¹⁹

In our case, since the atomic charges and force field parameters describing the Co −CO

2

bond as well as relevant angles/dihedrals are di fferent in the two force fields, our EVB model essentially

Figure 1.Schematic catalytic cycle with the CO2binding step being exergonic.

Figure 2.Snapshots of the Co^I(TPP)⁻system investigated in this study: (a) reactant state and (b) product state, both at a graphene−water interface.

https://dx.doi.org/10.1021/acs.jpcc.0c05974 J. Phys. Chem. C XXXX, XXX, XXX−XXX B

charge redistribution along the reaction.

The all-atom MD approach of EVB provides an explicit picture of ion redistribution when an electric field is applied with only classical mechanics as preassumption, di ffering our approach from previous studies where ions were either studied implicitly

²⁷⁻²⁹

or were almost immobile.

³⁰

Moreover, once the EVB reaction is calibrated to a reference system (i.e., CO

₂

binding in water), it is then possible to study the same reaction in more complex (e.g., CO

₂

binding in water on CNTs) or rather di fferent environments, (e.g., other reaction media, which is otherwise even more challenging), as the EVB parameters are only mildly a ffected by the environment.

²⁶

Here, we present an EVB study of the CO

₂

binding process for a Co(TPP) −CNT system at the once reduced state, which serves as an example to understand the roles that the before- mentioned three factors played in reactions at the interfaces.

■ ^METHODS

In the EVB approach, the reactant and product state Hamiltonians are approximated by classical MD force field descriptions. Unlike DFT calculations, in EVB-MD simu- lations, the CO

₂

molecule to be bonded cannot be in finitely far away from the catalyst at the reactant state (e.g., unbonded state). Therefore, when we parameterized the force field for the EVB reactant state, a DFT optimization was performed with one CO

₂

at an intermediate distance away from the catalyst such that there were only nonbonded interactions between them. The optimized structure was associated with a Co −CO

2

distance of 3.5 Å. The EVB reactant state is 3.3 kcal/mol higher in Gibbs free energy than the DFT reactant state (Figure 4a), where there is no interaction between CO

₂

and the catalyst.

Force field parameters, which are based on Amber’s GAFF,

^31,32

are taken from a previous study on the Fe(porphyrin) unit in a heme system

³³

while we calculated new atomic partial charges by the electrostatic potential (ESP)

−

− =

H E H

H¹ H¹² E 0

12 2 (1)

Hence, the ground-state energy can be expressed as

= + − − +

E 1 H H H H H

2( ) 1

2 ( ) 4

1 2 1 2 2

122

(2)

H

₁

and H

₂

can be obtained from free-energy perturbation (FEP) calculations as long as we have a good force field description of the reactant and product states. H

₁₂

describes the coupling between the intersection of the reactant state Hamiltonian (H

₁

) and the product state Hamiltonian (H

₂

) and the actual transition state (TS).

The Gibbs free energy to move from the reactant state to the product state follows

∑

δ λ λ

ε

Δ → = →

= − [⟨ − − ⟩ ]

=

−

G H H G +

k T k T E

( ) ( )

1

exp( ( ))

m n

m m

m m

1 2

0 1

1

b b (3)

Parameter λ moves the system slowly from the H

1

state to the H

₂

state to ensure su fficient sampling, such that

ε_m=λH₁+(1−λ)H_{2 (4)}

where ε

m

is known as the mapping potential. It keeps the system around a given energy state.

Obtaining ΔG as a function of λ is not sufficient to evaluate activation energy, which re flects the probability of being at the transition state on the ground-state surface.

¹⁹

The correspond- ing formulism de fines the reaction coordinate as the energy di fference between the reactant (H

1

) and product (H

₂

) states

Figure 4.Illustration of (a) how EVB states relate with DFT-calculated energies with a dielectric continuum water model and the (b) calibrated EVB model. Unit: kcal/mol.

https://dx.doi.org/10.1021/acs.jpcc.0c05974 J. Phys. Chem. C XXXX, XXX, XXX−XXX C

α λ δ

− = Δ −

⟨ ′ − × ^{− ′ −ε ′} ⟩

G X G RT

X X

( ) ( )

ln ( ) e

m

E X X RT m ( ) _m( )/

(5)

where α is a constant describing the energy of formation (of the product), which is not covered in force field parameters.

The two reaction-speci fic parameters (H

12

and α) need to be fitted to complete the EVB model. After fitting the Co

^I

(TPP) ···

CO

₂⁻

/Co

^I

(TPP) −CO

2−

reaction to the DFT results (Figure 4a), H

₁₂

and α were determined to be 8.0 kJ

²

/mol

²

and −690.5 kJ/mol, respectively. In Figure 4b, we can see that the calibrated EVB energy surface has the right key characters, namely, ΔG

^‡

and ΔG, compared to the DFT results.

In our simulations, a graphene sheet 36 × 30 Å

²

in size was used to represent the mw-CNTs with small curvatures. A position restraint was applied throughout all the simulations to mimic the fact that it was fixed on the electrode surface. The Co(TPP) ···CO

2

system, solvent molecules, and electrolyte ions (5K

⁺

and 4Cl

⁻

, 4KCl to give an electrolyte concentration of 0.1 M, the extra K

⁺

to compensate the negative charge on Co

^I

(TPP)

⁻

to a fford a neutral system) were free to move in the simulation box (43.92 × 43.92 × 43.92 Å

³

).

For each FEP process, the reaction coordinate was divided into 68 windows and each window was sampled for 200 ps (with the exception of the simulation in DMF, where a 1 ns simulation was performed) at a timestep of 0.0001 ps so that the whole process was sampled for 13.8 ns. Data acquired were then fed into our calibrated EVB model to plot the free-energy pro files.

■ RESULTS AND DISCUSSION

Electric Field E ffect. When a bias potential is applied, electrolyte ions in solution rearrange to screen the surface charge on the electrode, forming the so-called electrical double layer (EDL), which is rich in one of the counter ions. The negatively charged, bonded catalyst-CO

₂

complex could potentially be stabilized by the high cation concentration near the cathode. Either Poisson −Boltzmann

²⁹

or Possion − Nernest −Planck

³⁴

type of model gives a good analytical description of the EDL, and hence provides a concentration pro file of the electrolyte cations, as well as many other species in the system, as a function of the distance from the electrode surface. Modeling the EDL by DFT-based methods and hence

Figure 5.EVB profiles of CO2binding in (a) water and (b) 0.1 M KCl(aq) and with the electricfield strength being −0.4 V/nm. Note the different scales on the y-axis.

Figure 6.Normalized ion distribution as a function of the distance from the graphene electrode surface in the presence (a, b) and the absence (c, d) of an electricfield.

https://dx.doi.org/10.1021/acs.jpcc.0c05974 J. Phys. Chem. C XXXX, XXX, XXX−XXX D

calculating the e ffect that an electric field brings to the reaction profile is, however, not straightforward.

Most DFT studies used implicit solvent models and relied on immobilized cations at the interface to create a biased potential. These studies suggest a lowered activation energy toward CO

₂

binding.

^35,36

Ab initio MD studies of such systems (i.e., implicit solvent models with surface cations) have been reported,

³⁷

con firming that the present adsorbed cations indeed interact with the terminal oxygens on activated catalyst-CO

₂

complexes in a constructive fashion. However, very few DFT-based studies took an explicit solvent into account due to the complexity of the system and hence high computational cost,

^38,39

while the classical MD is not applicable since it does not allow bond formation. Specifically, ab initio MD was used to study the electrode −electrolyte interface in a few reports,

^40,41

but the systems considered remained very small with the simulation length usually below 10 ps. In the present work, we use EVB-MD to simulate CO

₂

binding on a graphene electrode surface with an explicit description of the system by allowing ion migration and hence redistribution in the electric field throughout the reaction.

Since EVB is based on classical MD descriptions, it is computationally much less intense compared to DFT and hence allows us to explore larger systems and much longer timescales (ns to μs), providing a more realistic description of the system. Moreover, the classical description allows us to decompose the e ffects as specific terms can be readily turned on or o ff.

An electric field strength of −0.4 V/nm was estimated according to the linearized Poisson −Boltzmann equation with an electrode potential of −1.1 V (vs SHE) and a Debye length of 0.96 nm, which corresponds to a 0.1 M KCl solution, which is typically used in experiments.

^42,43

We also tested a very strong electric field (−4.0 V/nm) but it resulted in a large enough force to move the catalyst into the solution phase, which could be one of the deactivation paths.

A dramatic change in the free-energy pro file can be observed (Figure 5) when both the electrolyte and the electric field are presented. Not only is the activation energy ( ΔG

^‡

) lowered from 9.1 kcal/mol (Figure 5a) to 4.7 kcal/mol (Figure 5b), but the nature of the reaction changes from endergonic to exergonic. The presence of the CNT supporting material has a minor improvement for CO

₂

binding (Figure 5a), which we shall discuss in the next section.

A higher cation concentration was observed near the electrode surface and a higher anion concentration further away from the electrode (Figure 6a,b), when an electric field is applied, while such variation is not seen without an electric field ( Figure 6c,d). The higher local cation concentration likely

stabilizes the intermediate through Coulombic interactions and hence produces the “cation effect”,

29,35,39,44

which states that the presence of cations enhances the reactivity for electro- chemical CO

₂

reduction. CO

₂

binding is usually an endergonic step that contributes to the overall activation energy for molecular electrocatalysts.

^45,46

By lowering the activation energy and reaction free energy for this step, the energy for the rate-determining transition state could be subsequently lowered (Figure 1) and hence an enhancement in the reactivity was observed, given that the surface does not alter the following steps. The surface could of course a ffect these steps but that is beyond the scope of this report.

Moreover, a closer examination of the cation distribution illustrates a higher K

⁺

concentration around CO

₂

in the presence of an electric field ( Figure 7) at both the reactant and the product states. In the reactant state, a peak at ca. 5 Å can be observed in the presence of an electric field, while no signi ficant features were observed otherwise ( Figure 7a). In the product state, an additional peak at 3.7 Å was also observed in the presence of an electric field. Although these two peaks were also observed in the absence of the electric field in the product state, the probability densities are much lower (Figure 7b).

Such phenomena were not reported earlier since the redistribution of ions can only be observed by explicit models and with long simulations (50 ns in our case).

In molecular dynamics, the K

⁺

···CO

2

interactions are de fined by the sum of Coulombic and Lennard-Jones (LJ) interactions. A higher local K

⁺

concentration in the presence of an electric field provides a significant stabilization effect throughout the catalyst-to-CO

₂

charge transfer process (Figure 8).

The same analysis should, in principle, apply to protons as well, which play a crucial role in the CO

₂

to CO reduction.

^16,47

We have not modeled protons, however, due to their nonclassical di ffusion behavior. Nonetheless, the electric field−proton interactions lead to a lowered pH

³⁴

at the

Figure 7.Smoothed radial distribution functions (RDF) of the K ions and the CO2carbon recorded for a 50 ns simulation at the (a) reactant state and the (b) product state. The Savitzky−Golay filter was applied to the original RDFs (Figures S2 and S3) to smooth the curves.

Figure 8. K⁺···CO2 interactions as a function of the reaction coordinate.

https://dx.doi.org/10.1021/acs.jpcc.0c05974 J. Phys. Chem. C XXXX, XXX, XXX−XXX E

electrode surface where catalysis takes place, making both protonation and proton-assisted reduction, two important steps in CO

₂

reduction, much easier than in the bulk solution.

Just like cations, the electric field itself could also interact with the reactant and product states di fferently and thereby lead to stabilization as the reaction proceeds. We will decompose the two e ffects below to understand their relative magnitudes. In the presence of either a KCl electrolyte (Figure S7) or the electric field ( Figure S8), a minor, almost negligible, improvement in both the activation free energy and the reaction free energy was observed. This is possibly contributed by a slightly improved electrostatic interaction between the product state and the surroundings. Therefore, only the combination of both conditions can build up a high local cation concentration (Figures 6 and 7) and induce the “cation e ffects” observed for catalysis.

Surface E ffect. The presence of a carbon supporting material e ffectively blocks the access of the solvent to the catalyst from beneath, especially in the case when a larger, flatter mw-CNT is used. As a consequence, both the activation free energy ( ΔG

^‡

) and the reaction free energy ( ΔG) are lowered (Figure 5a).

The explanation lies in the change in the solvation stabilization energy throughout the charge transfer process.

In MD simulations, solvation stabilization energy is de fined as the sum of electrostatic and vdW interactions between the reactants and the solvent molecules. As the catalyst becomes less charged during the charge transfer process, its solvation energy becomes less negative, as the electrostatic interactions between the catalyst and water molecules weaken. A larger decrease in solvation stabilization was observed for molecular Co(TPP) as compared to the Co(TPP)/CNT system (Figure 9) since both sides of the catalysts are accessible by water.

The total change in free energy consists of the change in reactant −reactant interactions plus the change in reactant−

environment interactions. If we assume the reactant −reactant interactions remain unchanged for the same chemical reaction as in the EVB theory,

¹⁸

a smaller decrease in solvation stabilization (Figure 9, blue) results in both a lowered activation free energy and a lowered reaction free energy (Figure 5a). Figure 9 shows the change in the solvation energy

(Figure 9, black) throughout CO

₂

binding. The total solvation energy was divided into the solvation energy experienced by Co(TPP) (Figure 9, blue) and CO

₂

(Figure 9, red) to illustrate our point (i.e., Co(TPP) in the Co(TPP)/CNT system experiences less destabilization through the charge transfer).

= +

_ _ _

E_{solv total} Esolv Co(TPP) Esolv CO2

It shows an extra −1.5 kcal/mol in solvation stabilization for the Co(TPP)/CNT system as a result of CO

₂

binding. This value is very close to the EVB calculated stabilization in ΔG ( −1.3 kcal/mol) in the presence of the graphene surface, adding credibility to our explanations.

Solvent E ffect. Previous DFT studies, both by us and by other groups, suggest that a polar solvent is essential for the charge transfer process during CO

₂

binding.

^48,49

It is well known that water as a polar solvent stabilizes charged species better than its less-polar counterparts, and its hydrogen bonding ability stabilizes the charge accumulation on the CO

₂

oxygen atoms. When the less-polar and nonhydrogen bond-donating solvent DMF is used, there is a smaller change in solvation energy throughout the CO

₂

binding due to the much weaker charge−charge interactions. Despite a lesser degree of destabilization for the catalyst, CO

₂

is also less stabilized once it becomes charged and the latter effect dominates (Figure S12). The activation free energy as well as the reaction free energy calculated by EVB match well with our DFT calculations using a polarizable continuum model (PCM), which reconfirms the reliability of our fitted EVB model (Figure 10). A higher activation energy translates to a

lower reactivity. Even though the reaction steps after CO

₂

binding are rate determining due to the scarcity of protons,

¹⁶

the fact that CO

₂

binding is less favored in DMF still contributes to the lowered reactivity in DMF, as it does contribute to the overall activation energy when the reaction is endergonic (Figure 1).

One may expect that solvents (e.g., methanol) that are not as polar as water but are still able to form hydrogen bonds with the oxygen atoms on CO

₂

are superior to both water and DMF. In other words, it could stabilize CO

₂

after the charge transfer process without lowering the reactivity of the catalyst as in the case of water (Figure 9). However, our EVB results reveal that methanol ranks in between water and DMF, in terms of both activation free energy and reaction free energy

Figure 9.Change in the solvation energy for Co(TPP)···CO2(black),

Co(TPP) (blue), and CO2 (red) throughout the charge transfer process.

Figure 10.EVB profile of CO2binding in DMF (black) with the DFT curve (fitted to the calculated ΔG^‡andΔG, red) as reference.

https://dx.doi.org/10.1021/acs.jpcc.0c05974 J. Phys. Chem. C XXXX, XXX, XXX−XXX F

activation energy (9.8 vs 10.9 kcal/mol) and reaction free energy (3.6 vs 6.3 kcal/mol) are lower in water than in methanol. These values do not fit as well with the DFT data obtained using a PCM model (Figure 11) as in the case of water and DMF. However, one should note that PCM models ignore the explicit solute −solvent hydrogen bonds, which could be the reason for the overestimation of ΔG

^‡

and ΔG, highlighting the advantage of our full explicit solvent description.

Calculations in different reaction media and the analysis of solvation free energy reveal that the most important function of the solvent is to stabilize CO

₂

in the product state.

Furthermore, an ideal solvent for CO

₂

binding is one that can stabilize CO

₂

while it does not significantly destabilize the catalyst. The presence of a hydrophobic supporting material cleverly satis fies both criteria by limiting the catalyst’s solvent accessible surface area while the local environment of CO

₂

is hardly a ffected.

■ CONCLUSIONS

EVB-MD allowed us to decompose the e ffects of the heterogeneous conditions to fully understand the CO

₂

binding process in the electric double layer. We found three major contributors to the superior performance of heterogeneous catalysts as compared to their homogeneous counterparts using Co(TPP) and Co(TPP)/CNT as an example: (1) higher cation concentration in the EDL has a very strong constructive e ffect for CO

2

binding and likely also for other endergonic steps in catalysis. The field pulls the cations closer to the electrode surface and enables them to interact with CO

₂

throughout the charge transfer process and hence stabilizes the system. Our EVB calculations with only the electric field and only the electrolyte con firm that such stabilization can only be achieved when both conditions are ful filled; (2) the immobilization of the catalyst allows for reactions in water even if the catalyst is not soluble. Furthermore, water as a solvent facilitates charge separation, which is much less signi ficant in the polar organic solvents that we tested (DMF and methanol); (3) the ideal solvent for catalyst-to-CO

₂

charge transfer is one that can stabilize the charged CO

₂

well but not

are very di fferent from homogeneous solution-phase reactions.

Therefore, DFT calculations with implicit solvation as in standard practice may not give the correct results for a supported molecular catalyst reacting at the interfaces.

■ ASSOCIATED CONTENT

*

^sı Supporting Information

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jpcc.0c05974.

All details of the computational study including force field parametrization, molecular dynamics simulations, and empirical valence bond parameterization; solvation energy analysis and cation distribution as a function of λ (PDF)

Topology file including all parameters used for free- energy perturbation (TXT)

Geometry with consistent atom numbering as the topology file (TXT)

■ AUTHOR INFORMATION

Corresponding Author

Mårten S. G. Ahlquist − Department of Theoretical Chemistry and Biology, School of Engineering Sciences in Chemistry, Biotechnology and Health, KTH Royal Institute of Technology, 10691 Stockholm, Sweden; orcid.org/0000-0002-1553- 4027; Email: ahlqui@kth.se

Author

Xiaoyu Chen − Department of Theoretical Chemistry and Biology, School of Engineering Sciences in Chemistry, Biotechnology and Health, KTH Royal Institute of Technology, 10691 Stockholm, Sweden

Complete contact information is available at:

https://pubs.acs.org/10.1021/acs.jpcc.0c05974

Author Contributions

This paper was written through the contributions of all authors. All authors have given approval to the final version of the paper.

Notes

The authors declare no competing financial interest.

■ ACKNOWLEDGMENTS

The computations were enabled by resources provided by the Swedish National Infrastructure for Computing (SNIC), which is funded by the Swedish Research Council through grant agreement no. 2016-07213, in Linköping (NSC), and the PDC Centre for High Performance Computing (PDC-HPC) in Stockholm under project numbers SNIC 2019/3-284, SNIC 2019/3-6, and SNIC 2020/5-41. We acknowledge NordForsk (no. 85378) for the Nordic University hub NordCO

₂

.

Figure 11.EVB profile of CO2 binding in MeOH (black) with the

DFT curve (fitted to the calculated ΔG^‡andΔG, red) as reference.

https://dx.doi.org/10.1021/acs.jpcc.0c05974 J. Phys. Chem. C XXXX, XXX, XXX−XXX G

M.S.G.A. is supported by the Swedish Research Council (VR) grant number 2018-05396, and the Knut & Alice Wallenberg (KAW) project CATSS (KAW 2016.0072). X.C. acknowledges the China Scholarship Council CSC for financial support.

■ ABBREVIATIONS

EVB, empirical valance bond; MD, molecular dynamics; DFT, density functional theory; EDL, electric double layer

■

(1) Azcarate, I.; Costentin, C.; Robert, M.; Saveant, J.-M. Through-

^REFERENCES

Space Charge Interaction Substituent Effects in Molecular Catalysis Leading to the Design of the Most Efficient Catalyst of CO2to CO Electrochemical Conversion. J. Am. Chem. Soc. 2016, 138, 16639−

16644.

(2) Jiang, Y.; Li, F.; Zhang, B.; Li, X.; Wang, X.; Huang, F.; Sun, L.

Promoting the Activity of Catalysts for the Oxidation of Water with Bridged Dinuclear Ruthenium Compleexes. Angew. Chem., Int. Ed.

2013, 52, 3398−3401.

(3) Johnson, B. A.; Agarwala, H.; White, T. A.; Mijangos, E.; Maji, S.; Ott, S. Judicous Ligand Design in Ruthenium Polypyridyl CO2

Reduction Catalysts to Enhance Reactivity by Steric and Electronic Effects. Chem. - Eur. J. 2016, 22, 14870−14880.

(4) Chen, Y.; Kanan, M. W. Tin Oxide Dependence of the CO₂ Reduction Efficiency on Tin Electrodes and Enhanced Activity for Tin/Tin Oxide Thin-Film Catalysts. J. Am. Chem. Soc. 2012, 134, 1986−1989.

(5) Zhu, W.; Zhang, C.; Li, Q.; Xiong, L.; Chen, R.; Wan, X.; Wang, Z.; Chen, W.; Deng, Z.; Peng, Y. Selective Reduction of CO₂ by Conductive Mof Nanosheets as an Effecient Co-Catalyst under Visible Light Illumination. Appl. Catal., B 2018, 238, 339−345.

(6) Wang, M.; Chen, L.; Lau, T.-C.; Robert, M. A Hybrid Co Quaterpyridine Complex/Carbon Nanotube Catalytic Material for Co₂Reduction in Water. Angew. Chem., Int. Ed. 2018, 57, 7769−7773.

(7) Choi, J.; Wagner, P.; Jalili, R.; Kim, J.; MacFarlane, D. R.;

Wallace, G. G.; Officer, D. L. A Porphyrin/Graphene Framework: A Highly Efficient and Robust Electrocatalyst for Carbon Dioxide Reduction. Adv. Energy Mater. 2018, 8, No. 1801280.

(8) Yin, Z.; et al. Highly Selective Palladium-Copper Bimetallic Electrocatalysts for the Electrochemical Reduction of CO₂ to CO.

Nano Energy 2016, 27, 35−43.

(9) Liu, Z.-Q.; Cheng, H.; Ma, T. Y.; Su, Y.-Z.; Li, N. ZnCo₂O₄ Quantum Dots Anchored on Nitrogen-Doped Carbon Nanotubes as Reversible Oxygen Reduction/Evolution Electrocatalysts. Adv. Mater.

2016, 28, 3777−3784.

(10) He, Z.; Chen, J.; Liu, D.; Tang, H.; Deng, W.; Kuang, Y.

Deposition and Electrocatalytic Properties of Platinum Nanoparticals on Carbon Nanotubes for Methanol Electroxidation. Mater. Chem.

Phys. 2004, 85, 396−401.

(11) Hu, X.-M.; Pedersen, S. U.; Daasbjerg, K. Supported Molecular Catalysts for the Heterogeneous CO2 Electroreduction. Curr. Opin.

Electrochem. 2019, 15, 148−154.

(12) Hu, X.-M.; Ronne, M. H.; Pedersen, S. U.; Skrydstrup, T. S.;

Daasbjerg, K. Enhanced Catalytic Activity of Cobalt Porphyin in CO2

Electroreduction Upon Immobilization on Carbon Materials. Angew.

Chem., Int. Ed. 2017, 56, 6468−6472.

(13) Kumar, P.; Kumar, A.; Sreedhar, B.; Sain, B.; Ray, S. S.; Jain, S.

L. Cobalt Phthalocyanine Immobilized on Graphene Oxide: An Efficient Visible-Active Catalyst for the Pgotoreduction of Carbon Dioxide. Chem. - Eur. J. 2014, 20, 6154−6161.

(14) Serp, P.; Corrias, M.; Kalck, P. Carbon Nanotubes and Nanofibers in Catalysis. Appl. Catal., A 2003, 253, 337−358.

(15) Lieber, C. M.; Lewis, S. L. Catalytic Reduction of CO₂ at Carbon Electrodes Modified with Cobalt Phthalocyanine. J. Am.

Chem. Soc. 1984, 106, 5033−5034.

(16) Chen, X.; Hu, X.-M.; Daasbjerg, K.; Ahlquist, M. S. G.

Understanding the Enhanced Catalytic CO₂ Reduction Upon

Adhering Cobalt Porphyrin to Carbon Nanotubes and the Inverse Loading Effect. Organometallics 2020, 1634−1641.

(17) Muller, R. P.; Warshel, A. Ab Initio Calculations of Free Energy Barriers for Chemical Reactions in Solution. J. Phys. Chem. A 1995, 99, 17516−17524.

(18) Warshel, A.; Weiss, R. M. An Empirical Valence Bond Approach for Comparing Reactions in Solution and in Enzymes. J.

Am. Chem. Soc. 1980, 102, 6218−6226.

(19) Hwang, J.-K.; King, G.; Creighton, S.; Warshel, A. Simulation of Free Energy Relationships and Dynamics of SN₂ Reactions in Aqueous Solutions. J. Am. Chem. Soc. 1988, 110, 5297−5311.

(20) Kamerlin, S. C. L.; Warshel, A. Multiscale Modelling of Biological Functions. Phys. Chem. Chem. Phys. 2011, 13, 10401−

10411.

(21) Kamerlin, S. C. L.; Warshel, A. At the Dawn of the 21st Century: Is Dynamics the Missing Link for Understanding Enzyme Catalysis. Proteins 2010, 78, 1339−1375.

(22) Socan, J.; Purg, M.; Åqvist, J. Computer Simulations Explain the Anomalous Temperature Optimum in a Cold-Adapted Enzyme.

Nat. Commun. 2020, No. 2644.

(23) Isaksen, G. V.; Åqvist, J.; Brandsdal, B. O. Enzyme Surface Rigidity Tunes the Temperature Dependence of Catalytic Rates. Proc.

Natl. Acad. Sci. U.S.A. 2016, 113, 7822−7827.

(24) Zhan, S.; Mårtensson, D.; Purg, M.; Kamerlin, S. C. L.;

Ahlquist, M. S. G. Capturing the Role of Explicit Solvent in the Dimerization of Ru^v(bda) Water Oxidation Catalysts. Angew. Chem., Int. Ed. 2017, 56, 6962−6965.

(25) Zhan, S.; Ahlquist, M. S. G. Dynamics and Reactions of Molecular Ru Catalysts at Carbon Nanotube - Water Interfaces. J. Am.

Chem. Soc. 2018, 140, 7498−7503.

(26) Hong, G.; Rosta, E.; Warshel, A. Using the Constrained Dft Approach in Generating Diabatic Surfaces and Off Diagonal Empirical Valence Bond Terms for Modelling Reactions in Codensed Phases. J.

Phys. Chem. B 2006, 110, 19570−19574.

(27) Ringe, S.; Morales-Guio, C. G.; Chen, L. D.; Fields, M.;

Jaramillo, T. F.; Hahn, C.; Chan, K. Duoble Layer Charing Driven Carbon Dixoide Adsorption Limits the Rate of Electrochemical Carbon Dioxide Reduction on Gold. Nat. Commun. 2020, No. 33.

(28) Gupta, N.; Gattrell, M.; Macdougall, B. Calculation for the Surface Concentration in the Electrochemical Reduction of CO2 in Khco₃Solutions. J. Appl. Electrochem. 2006, 36, 161−172.

(29) Ringe, S.; Clark, E. L.; Resasco, J.; Walton, A.; Seger, B.; Bell, A.

T.; Chan, K. Understanding Cation Effect in Electrochemical CO2

Reduction. Energy Environ. Sci. 2019, 12, 3001−3014.

(30) Chen, L. D.; U, M.; Chan, K.; Norskov, J. K. Electric Field Effects in Electrochemical CO2 Reduction. ACS Catal. 2016, 6, 7133−7139.

(31) Wang, J.; Wang, W.; Kollman, P. A.; Case, D. A. Automatic Atom Type and Bond Type Perception in Molecular Mechanical Calculations. J. Mol. Graphics Modell. 2006, 25, No. 247260.

(32) Wang, J.; Wolf, R. M.; Caldwell, J. W.; Kollman, P. A.; Case, D.

A. Development and Testing of a General Amber Force Field. J.

Comput. Chem. 2004, 25, 1157−1174.

(33) Shahrokh, K.; Orendt, A.; Yost, G. S.; Cheatham, T. E., III Quantum Mechanically Derived Amber-Compatible Heme Parame- ters for Various States of the Cytochrome P450 Catalytic Cycle. J.

Comput. Chem. 2012, 33, 119−133.

(34) Bohra, D.; Chaudhry, J. H.; Burdyny, T.; Pidko, E. A.; Smith, W. A. Modeling the Electrical Double Layer to Understand the Reaction Environment in a CO₂ Electrocatalytic System. Energy Environ. Sci. 2019, 12, 3380−3389.

(35) Pérez-Gallent, E.; Marcandalli, G.; Figueirdo, M. C.; Callo- Vallejo, F.; Koper, M. T. M. Structure- and Potential-Dependent Cation Effects on CO Reduction at Copper Single-Crystal Electrode.

J. Am. Chem. Soc. 2017, 139, 16412−16419.

(36) Akhade, S. A.; McCrum, I. T.; Janik, M. J. The Impact of Specifically Adsorbed Ions on the Copper Catalyzed Electroreduction of CO₂. J. Electrochem. Soc. 2016, 163, F477−F484.

https://dx.doi.org/10.1021/acs.jpcc.0c05974 J. Phys. Chem. C XXXX, XXX, XXX−XXX H

2

Based Molecular Dynamics Simulation. J. Phys.: Condens. Matter 2014, 26, No. 244108.

(41) Schnur, S.; Gross, A. Properties of Metal-Water Interfaces Studied from First Principles. New J. Phys. 2009, 11, No. 125003.

(42) Ma, T.; Fan, Q.; Tao, H.; Han, Z.; Jia, M.; Gao, Y.; Ma, W.;

Sun, Z. Heterogeneous Electrochemical CO₂ Reduction Using Nonmetallic Carbon-Based Catalysts: Current Status and Future Challenges. Nanotechnology 2017, 28, No. 472001.

(43) Long, C.; Li, X.; Guo, J.; Shi, Y.; Liu, S.; Tang, Z.

Electrochemical Reduction of CO₂over Heterogeneous Catalysts in Aqueous Solution: Recent Progress and Perspectives. Small Methods 2019, 3, No. 1800369.

(44) Sato, S.; Saita, K.; Sekizawa, K.; Maeda, S.; Morikawa, T. Low- Energy Electrocatalytic CO₂Reduction in Water over Mn-Complex Catalyst Electrode Aided by a Nanocarbon Support and K⁺Cations.

ACS Catal. 2018, 8, 4452−4458.

(45) Ariafard, A.; Brookes, N.; Stranger, R.; Yates, B. F. DFT Study on the Mechanism of the Activation and Cleavage of CO₂ by (NHC)CuEPh₃ (E= Si, Ge, Sn). Organometallics 2011, 30, 1340−

1349.

(46) Planas, N.; Ono, T.; Vaquer, L.; P, M.; Benet-Buchholz, J.;

Gagliardi, L.; Cramer, C. J.; Llobet, A. Carbon Dioxide Reduction by Mononuclear Ruthenium Polypyridyl Complexes. Phys. Chem. Chem.

Phys. 2011, 13, 19480−19484.

(47) Miyamoto, K.; Asahi, R. Water Facilitated Electrochemical Reduction of CO₂ on Cobalt-Porphyrin Catalysts. J. Phys. Chem. C 2019, 123, 9944−9948.

(48) Shen, J.; Kolb, M. J.; Gottle, A. J.; Koper, M. T. M. DFT Study on the Mechanism of the Electrochemical Reduction of CO₂ Catalyzed by Cobalt Porphyrins. J. Phys. Chem. C 2016, 120, 15714−15721.

(49) Meshitsuka, S.; Ichikawa, M.; Tamaru, K. Electrocatalysis by Metal Phthalocyanines in the Reduction of Carbon Dioxide. J. Chem.

Soc., Chem. Commun. 1974, 5, 158−159.

https://dx.doi.org/10.1021/acs.jpcc.0c05974 J. Phys. Chem. C XXXX, XXX, XXX−XXX I