Formic Acid Electroreduction Pathways on (111) Metal Surfaces

Abstract

Formic acid (HCOOH) electroreduction is poorly studied, even though it holds promise for energy conversion and storage applications. We use density functional theory to investigate the most likely products after four HCOOH electroreduction steps on Cu(111), Au(111), Ag(111), Zn(111), Pt(111), Pd(111), and Ru(111). On Cu(111), the formation of H₂, CO (and/or CO‐derived C₂ products), CH₃OH, and CH₄ is thermodynamically allowed. The Cu(111) surface has low selectivity because HCOOH reduction intermediates can adsorb via both O and C atoms. Experimentally, formic acid reduction on Cu produces H₂, C₂ products, and CH₃OH, but very little CH₄. Interestingly, CO/CO₂ reduction on Cu produces CH₄ rather than CH₃OH. The CO/CO₂ reduction pathway can proceed via the *COH intermediate, whereas HCOOH reduction can only proceed via the *CHO or CH₂O*OH intermediates, possibly explaining the different product distribution. Au(111) is the most promising catalyst with high suggested selectivity to methanol and low hydrogen evolution rates. Ag(111) could be selective to methanol, but the first reduction step is very costly, so the reaction rates will be low. Zn(111), most likely, reduces HCOOH to CH₄, whereas Pt(111), Pd(111), and Ru(111) most likely produce CO poisoning the catalyst surfaces.

We explore HCOOH electroreduction pathways on seven (111) surfaces using density functional theory and experimental literature insight. Cu(111) mainly produces H₂, CH₃OH, and C₂ products in accordance with experiments. The most likely outcome is CH₃OH on Au(111), CH₃OH with low rate on Ag(111), CH₄ on Zn(111), and CO that poisons the catalyst on Pt(111), Pd(111), and Ru(111).

1. Introduction

Formic acid electroreduction has been studied surprisingly little, especially compared to glycerol oxidation^{[ 1 ]} and CO/CO₂ electroreduction,^{[ 2 ]} which have been studied hundreds and thousands of times, respectively. However, formic acid electroreduction could be valuable in energy conversion and storage applications. For example, formic acid can be produced in large amounts by biomass electro‐oxidation.^{[ 3} , ⁴ , ⁵ , ⁶ , ^{7 ]} Electro‐oxidation has to be accompanied by an electroreduction reaction, which then, conveniently, could be the electroreduction of the produced formic acid.^{[ 8 ]} We imagine that such an electrochemical setup would both protect the formic acid from further oxidation to CO₂ and, with the right catalyst, convert the formic acid to other desirable organic molecules, such as formaldehyde or methanol. We show a schematic of the envisioned glycerol oxidation‐formic acid reduction electrochemical setup in Figure S1, Supporting Information. The electrochemistry could, for instance, follow: oxidation of three glycerol molecules to seven formic acid and two CO₂ molecules, hereby producing the required amount of H⁺ + e⁻ to reduce the seven formic acid molecules to methanol. This would eliminate the need for water electrolysis and the sluggish oxygen evolution reaction.

We have collected experimental formic acid electroreduction studies from the literature and summarize the product selectivity on pure metal electro‐catalysts in Table  1 . We include an extended table (Table S1, Supporting Information) with additional experimental data covering ref.^{[ 9} , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ^{15 ]} The reported formic acid electroreduction products are methanol (CH₃OH), ethanol (CH₃CH₂OH), acetaldehyde (CH₃CHO), isopropanol (CH₃CHOHCH₃), methane (CH₄), and ethane (C₂H₆). Among the possible products, CH₃OH is the main product with high selectivity for formic acid electroreduction on Cu and Sn pure metals (Table 1). Additionally, Cr might also have high selectivity for CH₃OH, but Cr decomposes, producing large amounts of H₂, which makes it difficult to assess the CH₃OH selectivity (Table S1, Supporting Information, and ref. [15]) Selective HCOOH reduction to CH₃OH is beneficial, since methanol is more energy dense than HCOOH (the enthalpy of combustion at room temperature is 211.5 kJ mol⁻¹ for formic acid and 638.1 kJ/mol for methanol)^{[ 16 ]} and methanol is an important feedstock chemical for the chemical industry.^{[ 17 ]} Additionally, the study^{[ 18 ]} uses oxidized titanium to reduce formic acid and produce CH₃OH, CH₄, and CH₂O with Faradaic efficiencies of 12.6%, 1.4%, and 0.1%, respectively, at −1.0 V vs the reversible hydrogen electrode (RHE).

Table 1.

Experimental product selectivities (Faradaic efficiency, FE%) of formic acid electroreduction on different metal catalysts and reaction conditions.

Catalyst/FE%	CH3OH	CH3CH2OH	CH3CHO	CH4	C2H6	H2	U(VvsRHE )	pH	Electrolyte/Temperature	Ref
Sn	99	–	–	–	–		−0.22–−0.26	3.8	0.1M NaHCO3 + 0.25M HCOOH, 25 °C	[10]
	95	–	–	–	–		−0.49	≈3.8	0.5M HCOOH + 0.5M HCOONa, 23 °C	[14]
Pb	11.9	–	–	–	–		−0.6–−0.7	≈1	0.1M HClO4 + 0.1M HCOOH 25 °C	[10]
Cua)	27.6 × 2	–	25.8	–	–	10	−0.22	≈0.7	85% w/w H3PO4 + 1% HCOOH, 70 °C	[13]
	15.2 × 2	3.1	9.2	–	–	50	−0.27
	11.5 × 2	5.6	7.5	0.5	–	62	−0.32
	3.1	7.9	2.4	1.3	0.6	79	−0.52
	1.1	4.1	1.2	1.5	0.7	89	−0.62
	0.2	1.3	0.1	1.7	0.9	93	−0.82

^a)50% percent of produced CH₃OH is consumed by reaction with H₃PO₄ and HCOOH, hence the “×2”.

With the exception of Cu, formic acid reduction has not been investigated on many of the frequently used electro‐catalysts, e.g., Au, Ag, Pt, Pd, Ru, and Zn. We, therefore, systematically model possible products after four formic acid electroreduction steps on seven different metal surfaces, i.e., Cu(111), Au(111), Ag(111), Zn(111), Pt(111), Pd(111), and Ru(111). We first discuss Cu(111), where the modeling suggests that many products are thermodynamically allowed, namely H₂, CO (and the CO‐derived C₂ products),^{[ 2} , ¹⁹ , ^{20 ]} CH₃OH, and CH₄. Experimentally, the main products are H₂, CH₃OH, and CH₃CHO, while CH₄ is only a minor product (FE% < 2%). In comparison with Cu(111), modeling indicates that Au(111) yields CH₃OH at lower potentials and with higher selectivity. Ag(111) yields H₂ and CH₃OH at a low rate. Zn(111) favors CH₄, while Pt(111), Pd(111), and Ru(111) strongly favors CO by dissociating *CHO into *CO and H*. The *CO, in turn, likely poisons the Pt(111), Pd(111), and Ru(111) surfaces.

2. Computational Details

We use density functional theory (DFT) to calculate the stability of possible formic acid reduction intermediates and products. The computational settings are akin to our recent study on glycerol electro‐oxidation.^{[ 3 ]} We model the intermediates on 4 × 4 × 4 face‐centered cubic (FCC) metal slabs with (111) surface facet and the two bottom layers fixed in bulk positions. We note that Zn and Ru crystallize in a hexagonal close‐packed structure, but retain the FCC structure of the other metals for simplicity. We separate the pure metal slabs (periodic images) by 17 Å of vacuum to avoid interlayer interactions. We used the atomic simulation environment program^{[ 21 ]} to set up atomic structures and the GPAW program^{[ 22} , ²³ , ^{24 ]} to conduct DFT calculations. We used the revised Perdew−Burke−Ernzerhof (RPBE) exchange‐correlation functional,^{[ 25 ]} 400 eV energy cutoff for plane‐wave basis sets and 4 × 4 × 1 Monkhorst‐Pack k‐point sampling, and relaxed structures to a maximum force on each atom of 0.1 eV/Å.

For each of the four sequential formic acid electroreduction steps, we consider two types of reduction reactions. The first reaction type is hydrogenation (equation 1), where one hydrogen is added to the adsorbate (C₁H _y O _x ), resulting in a more reduced formic acid intermediate (C₁H _y+1O _x ). The added H originates from H⁺(aq) in the electrolyte and e⁻ from the potentiostat.

$${\mathrm{C}}_1{\mathrm{H}}_y{\mathrm{O}}_x+{\text{\hspace{0.17em}H}}^{+}(aq)+{\text{\hspace{0.17em}e}}^{-}\hspace{0.17em}\to {\mathrm{C}}_1{\mathrm{H}}_{y+1}{\mathrm{O}}_x$$ (1)

The second reaction type is OH elimination, where one OH group is hydrogenated to form a liquid water molecule and generate a more reduced intermediate (C₁H _y−1O _x−1).

$${\mathrm{C}}_1{\mathrm{H}}_y{\mathrm{O}}_x+{\text{\hspace{0.17em}H}}^{+}(aq)+{\text{\hspace{0.17em}e}}^{-} \to {\mathrm{C}}_1{\mathrm{H}}_{y-1}{\mathrm{O}}_{x-1}+{\mathrm{H}}_2\mathrm{O}(l)$$ (2)

We use the computational hydrogen electrode approach^{[ 26} , ^{27 ]} (equation 3) to connect the energy of H⁺(aq) and e⁻ to the free energy of gas‐phase H₂ and electrostatic potential at RHE scale (U _vsRHE ).

$$\text{ΔG}({\mathrm{H}}^{+}(aq)+{\text{\hspace{0.17em}e}}^{-})=\frac{1}{2}\text{\hspace{0.17em}ΔG}({\mathrm{H}}_2)-{\text{\hspace{0.17em}eU}}_{vsRHE}$$ (3)

In the Results and discussion section, we set U _vsRHE  = 0 V, but keep in mind that reaction steps with a free energy cost become downhill if we set eU _vsRHE negative enough to counteract the cost. The DFT energy differences for intermediates (ΔE_intermediate) are given in equation 4.

$${\text{ΔE}}_{\text{intermediate}}{\text{\hspace{0.17em}=\hspace{0.17em}E}}_{\text{slab+intermediate}}{\text{\hspace{0.17em}‐\hspace{0.17em}E}}_{\text{slab}}{\text{\hspace{0.17em}‐\hspace{0.17em}E}}_{\text{formic\hspace{0.17em}acid}}\text{\hspace{0.17em}‐\hspace{0.17em}}\frac{1}{2}{\text{(n}}_{\text{eq1}}{\text{+m}}_{\text{eq2}}{\text{)E}}_{{\mathrm{H}}_2}{\text{\hspace{0.17em}+\hspace{0.17em}m}}_{\text{eq2}}{\mathrm{E}}_{{\mathrm{H}}_2\mathrm{O}}$$ (4)

In equation 4, ${\mathrm{E}}_{\text{slab+intermediate}}$ and ${\text{\hspace{0.17em}E}}_{\text{slab}}$ are DFT energies of the slab with and without adsorbed intermediates, respectively, while ${\mathrm{E}}_{\text{formic\hspace{0.17em}acid}}$, ${\mathrm{E}}_{{\mathrm{H}}_2}$, and ${\mathrm{E}}_{{\mathrm{H}}_2\mathrm{O}}$ are the DFT energies of gas‐phase formic acid, gas‐phase H₂, and gas‐phase H₂O, respectively. The values, ${\mathrm{n}}_{\text{eq1}}$ and ${\mathrm{m}}_{\text{eq2}}$ are the number of equation 1 and equation 2 reaction steps required to convert formic acid to the intermediate in question.

In order to convert DFT energy differences to Gibbs free energy differences, we need to account for entropy differences, zero‐point energy differences, and heating enthalpy differences. These corrections have been calculated for COOH* reduction to CH₃O* on Cu(211),^{[ 28 ]} which is very similar to what we study here. The reduction of COOH* to CH₃O* evolves three equation 1 reduction steps (with an average correction of ${\text{ΔG}}_{\text{eq\hspace{0.17em}1}}^{\text{corr}}\text{=\hspace{0.17em}+0}\text{.32\hspace{0.17em}eV}$), and one equation 2 reduction step (with a correction of ${\text{ΔG}}_{\text{eq\hspace{0.17em}2}}^{\text{corr}}\text{=\hspace{0.17em}‐0}\text{.38\hspace{0.17em}eV}$). See Table S2‐S3, Supporting Information for additional details. We choose to use these correction values for every equation 1 and equation 2 reduction reaction step that we consider in this study. The Gibbs free energy differences for the possible formic acid reduction intermediates are therefore given by equation 5.

$${\text{ΔG}}_{\text{intermediate}}{\text{\hspace{0.17em}=\hspace{0.17em}ΔE}}_{\text{intermediate}}{\text{\hspace{0.17em}+\hspace{0.17em}n}}_{\text{eq1}}{\text{ΔG}}_{\text{eq\hspace{0.17em}1}}^{\text{corr}}{\text{\hspace{0.17em}+\hspace{0.17em}m}}_{\text{eq2}}{\text{ΔG}}_{\text{eq\hspace{0.17em}2}}^{\text{corr}}$$ (5)

These free energy corrections give a reasonably accurate overall reaction free energy at standard conditions for converting formic acid to methanol (HCOOH(aq) + 2H₂(g) → CH₃OH(aq) + H₂O(l)), namely −0.19 eV compared to the tabulated value of −0.42 eV.^{[ 29 ]} The reaction free energy errors can be further reduced by correcting systematic functional group DFT errors, i.e., for CO, —(C = O)O—, —C = O—, and —OH.^{[ 30} , ^{31 ]} We do not use such corrections in the main paper, but add them in Table S15‐S16, Supporting Information, and Figure S2‐S6, Supporting Information. The corrections stabilize CH₂O*OH compared to *CHO, but, for both intermediates, the subsequent intermediate is CH₂O; hence, our conclusions are unaffected.

In our recent study of glycerol electro‐oxidation, we considered the reverse of equation 1 and equation 2 as the possible oxidation reactions and used the same literature conversion between COOH* and CH₃O* to get free energy corrections.

Our calculations do not include the aqueous electrolyte. We therefore miss two important effects, namely 1) solvation energy differences between the formic acid molecule and subsequent adsorbed reaction intermediates, and 2) the free energy cost to displace water molecules and gain access to the surface (i.e., competitive adsorption). Solvation energies are generally large, but also computationally expensive to accurately account for.^{[ 32 ]} However, formic acid and many of the intermediates have similar functional groups and therefore similar interactions with the electrolyte, so we expect some error cancellation in the solvation energy differences between the different investigated species. Competitive adsorption can also result in large energy costs when species adsorb to the surface; however, the coverage of water molecules on (111) surfaces is relatively low, so only a little water needs to be displaced compared to surfaces with more undercoordinated metal atoms and higher water coverage.^{[ 33 ]}

The corrections used to convert from DFT reaction energies to reaction free energies include vibrational and rotational entropy of adsorbed intermediates and all entropy contributions from gas‐phase molecules. It does not include the coverage and concentration‐related configurational entropies. We assume the coverage is similar for formic acid and all intermediates, such that the coverage‐related entropies cancel out. The only remaining issue is in the first reduction step, where HCOOH is transferred from aqueous solution to the surface. One could imagine that adsorption lowers the entropy, but this is somewhat misleading, because competitive adsorption could displace water molecules from the surface, which then increases the entropy. In the end, we chose not to account for any entropy change of moving HCOOH from solution to the surface.

On each (111) metal surface, we construct the possible formic acid electro‐reaction pathways by going through the most stable intermediate at every reduction step, except for some discussed and rationalized cases. The free energies and structure illustrations of all the analyzed intermediates can be found in Table S4‐S10, Supporting Information. The DFT calculations and Python scripts are accessible online at https://erda.ku.dk/archives/2eb37aa2e5099fc29f9e1a04d10220d8/published‐archive.html.

3. Results and Discussion

3.1. Formic Acid Reduction on Cu(111)

Cu is one of the most experimentally studied formic acid electroreduction catalysts, so we first consider the possible formic acid electroreduction pathways on Cu(111) (Figure  1 ). In the first reduction step, both addition of an H atom (via reaction equation 1) or removal of an OH group (via reaction equation 2) leaves formic acid with a dangling bond, such that it can chemisorb to the Cu(111) surface, either as CH₂O*OH (ΔG_intermediate = 0.60 eV) or *CHO (ΔG_intermediate = 0.54 eV) (Figure 1). These intermediates have a larger ΔG_intermediate than the free energy of H* adsorption, ΔG_H* = 0.27 eV, so it's unlikely that formic acid electroreduction can occur on Cu(111) without the competing hydrogen evolution reaction. And indeed, hydrogen evolution is observed in the formic acid electroreduction experiments on Cu.^{[ 13 ]} The similar stability of CH₂O*OH and *CHO means that they are accessible at similar negative potentials, namely at U _vsRHE  = −0.60 V for CH₂O*OH and at−0.54 V for *CHO, respectively. Subsequently, the *CHO intermediate may undergo a purely chemical dissociation reaction to H* + *CO, which lowers ΔG_intermediate from 0.54 eV to−0.01 eV. Experiments have found that formic acid reduction on Cu leads to both CH₃OH and C₂ products (selectivity follows CH₃OH ≥ CH₃CHO > CH₃CH₂OH > C₂H₆).¹³ CO electroreduction on Cu produces significant amount of C₂ products, but generally not CH₃OH,^{[ 19} , ²⁰ , ^{34 ]} We, therefore, suggest that some of the formic acid is converted to H* + *CO and then to C₂ products, while the remaining formic acid exist as CH₂O*OH or *CHO after the first reduction step, even though H* + *CO is thermodynamically preferred.

Figure 1.

Free energies for possible formic acid electroreduction pathways on Cu(111) at U _vsRHE  = 0 V. We, additionally, propose that *CHO can dissociate to H* + *CO and that the formed CO is converted to C₂ products. The CH₂O*OH and the *CHO that do not decompose is reduced all the way to CH₃OH and/or CH₄.

CO electroreduction on Cu has been extensively modeled elsewhere,^{[ 2} , ^{19 ]} so for the second reduction step, we only consider reduction of the second most stable *CHO and third most stable CH₂O*OH intermediates. There are four possible reduction intermediates, of which the most stable is physisorbed formaldehyde (CH₂O) with ΔG_intermediate = 0.21 eV. The CH₂O intermediate is accessible from both *CHO (by H addition via reaction equation 1) and CH₂O*OH (by OH removal via reaction equation 2). The second most stable intermediate is physisorbed CH₂(OH)₂ with ΔG_intermediate = 0.53 eV.

In the third reduction step, CH₂O is reduced to CH₃O* via reaction equation 1, with ΔG_intermediate = −0.15 eV. The only other intermediate is *CH₂OH with ΔG_intermediate = 0.86 eV. The free energy favored conversion of CH₂O to CH₃O* is, likely, the reason why CH₂O is not observed as a product in formic acid reduction on Cu.^{[ 10} , ¹³ , ^{15 ]}

Finally, in the fourth reduction step, CH₃O* is reduced to CH₃OH or CH₄+O* with ΔG of −0.24 eV and −0.40 eV, respectively. Since both CH₃OH and CH₄+O* formation is downhill, especially at the ≈−0.6 V vs RHE required in the first reduction step, we cannot conclude which is the most likely product. In formic acid reduction experiments on both Cu¹³ and Cu_0.88Sn_0.06Pb_0.06 ^{[ 12 ]} (see further details in Table S1, Supporting Information), more CH₃OH is produced than CH₄, which indicates that CH₄+O* formation is kinetically hindered. Additionally, Hori et al. measured CH₂O electroreduction on Cu at−0.7 V vs RHE and found that CH₂O is mainly converted to CH₃OH (10.1 FE%) rather than CH₄ (0.7 FE%), although most of the electrons produce H₂ (89.8 FE%).^{[ 35 ]} This fits with our finding that both CH₃OH and CH₄ are accessible from CH₂O. On the other hand, the experiments in ref^{[ 36 ]} found that CH₂O electro‐reaction on Cu, Au, and Ag, almost exclusively makes CH₃OH with very little CH₂O becoming CH₄.

Overall, the Cu(111) surface can bind intermediates via both the C atom and an O atom, so many distinct intermediates are accessible from a free energy perspective. This, in turn, means that the possible formic acid electroreduction pathways on Cu(111) are very branched and that formic acid reduction on Cu isn't very selective. Moreover, it is interesting that both formic acid and CH₂O electro reduction produce CH₃OH, whereas CO and CO₂ electro reduction produce (mainly) CH₄ and C₂ products, but not CH₃OH.^{[ 2} , ¹⁹ , ^{20 ]} This suggests that two different pathways are at play. There is an ongoing discussion^{[ 37 ]} whether CO electroreduction toward C₁ products goes through the *CHO^{[ 38} , ^{39 ]} or *COH^{[ 28} , ⁴⁰ , ^{41 ]} intermediate. The *COH intermediate is not possible in formic acid reduction, so one way to explain the different product distributions is that CO electroreduction results in *COH rather than *CHO, and that *COH is part of the pathway toward CH₄.

3.2. Formic Acid Reduction on Au(111)

Next, we consider formic acid electroreduction on Au(111) (Figure  2 ). In the first reduction step, the most stable intermediate (out of four possibilities) is *CHO formation via reaction equation 1, with ΔG_intermediate = 0.47 eV. The first reduction step is less costly than the one on Cu(111), meaning that it is easier to chemisorb HCOOH on Au(111) than on Cu(111). Furthermore, it is unlikely that *CHO dissociates to H* + *CO on Au(111), since H* + *CO is less stable (ΔG_intermediate = 1.01 eV).

Figure 2.

Free energies for the possible formic acid electroreduction pathway on Au(111) at U _vsRHE  = 0 V. The reaction will likely produce CH₃OH with high selectivity, if it does not terminate at CH₂O formation.

In the second reduction step, *CHO is reduced via reaction equation 1 to physisorbed CH₂O with ΔG_intermediate = 0.20 eV. The other possible intermediate, *CHOH, is much less stable with ΔG_intermediate = 1.33 eV. The third reduction step, which converts CH₂O to *CH₂OH via reaction equation 1, is somewhat costly (*CH₂OH has ΔG_intermediate = 0.54 eV), and this makes it possible that formic acid electroreduction can terminate with CH₂O formation on Au. Finally, *CH₂OH is reduced to physisorbed CH₃OH (ΔG_intermediate = −0.17 eV) via reaction equation 1 in the fourth reduction step. The other possible intermediate, *CH₂, is much less stable with ΔG_intermediate = 0.66 eV. On Au(111), we therefore predict that CH₂O is selectively reduced to CH₃OH rather than CH₄, which agrees with experiments for CH₂O electroreduction on Au.^{[ 36 ]}

Overall, the intermediates that adsorb to the Au(111) surface via the C atom are much more stable than intermediates that adsorb via an O atom, so, opposite to Cu(111), the possible electroreduction pathway on Au(111) is not branched. Therefore, formic acid electroreduction should selectively produce CH₃OH if it does not terminate with CH₂O formation. Furthermore, the free energy of H* adsorption (ΔG_H* = 0.67 eV) is larger than any of the formic acid reduction steps, so hydrogen evolution should not compete with formic acid electroreduction. Given the possibility of selective formic acid electroreduction and expected low hydrogen evolution rates, Au(111) is the most promising catalyst of the ones that we consider.

Since Au(111) is a promising catalyst, it is interesting to test whether other Au facets also have a preference for adsorption via the C* atom and low energy cost for reduction. We, therefore, consider the first HCOOH reduction step on Au(100) and Au(211) step edges and find that *CHO is more stable than H₂CO*OH and H* + *CO, similar to Au(111) (Table S12, S13, Supporting Information). Furthermore, the energy cost is reduced in line with Au(100) and Au(211) step edges, having lower coordination number and higher reactivity than Au(111). Accordingly, we expect that other Au facets require less negative potential for HCOOH reduction, but result in the same preference for CH₂O and/or CH₃OH.

We note that formic acid reduction on oxidized titanium^{[ 18 ]} produces both CH₂O (FE 0.1%) and CH₃OH (12.6%). If oxidized titanium also has CH₂O as an intermediate toward CH₃OH, the strong preference for the latter indicates that it is difficult to terminate the reduction at CH₂O.

3.3. Formic Acid Reduction on Ag(111)

Formic acid adsorbs as *CHO (ΔG_intermediate = 0.82 eV) in the first reduction step on Ag(111) (Figure  3 ). The CH₂O*OH intermediate is less stable with ΔG_intermediate = 1.02 eV (Table S6, Supporting Information). The *CHO formation on Ag(111) is the most energy‐costly first reduction step among the (111) metal surfaces that we have examined. The step is also more costly than H* adsorption on Ag(111), ΔG_H* = 0.66 eV, and the onset of hydrogen evolution happens already around−0.4 V vs RHE on Ag(111).^{[ 42 ]} We therefore expect formic acid electroreduction on Ag(111) to be very slow and unlikely to happen without hydrogen evolution. The Ag(111) surface may even be poisoned by H* at the required−0.82 V vs RHE.

Figure 3.

Free energies for possible formic acid electroreduction pathways on Ag(111) at U _vsRHE  = 0 V. The *CHO formation is the most energy‐costly first reduction step among the examined (111) metal surfaces and more costly than H* adsorption on Ag(111), given by ΔG_H*.

To the extend that *CHO is formed, it can dissociate and form H* + *CO, similarly to *CHO on Cu(111). However, the H* + *CO is only 0.1 eV more stable than *CHO, so the driving force for the dissociation is much smaller than on Cu(111), and we expect that less CO will be formed on Ag(111). Ag is known as a selective CO₂ to CO electro‐catalyst,^{[ 43 ]} so any formed CO molecules will likely not get further reduced on Ag(111).

In the second reduction step, *CHO is reduced to CH₂O, which lowers ΔG_intermediate to 0.20 eV. The CH₂O can be further reduced, first to CH₃O* (ΔG_intermediate = 0.28 eV) and subsequently to CH₃OH (ΔG_intermediate = −0.19 eV). The alternative intermediates to the ones included in Figure 3, are all much less stable (Table S6, Supporting Information). This means that, if *CHO can be formed, the subsequent reduction to CH₃OH should be very selective. Again, experiments show that CH₂O is reduced to CH₃OH with very little CH₄ formation.^{[ 36 ]}

3.4. Formic Acid Reduction on Zn(111)

Among the metals we study, Zn is the least noble and forms the most stable oxide.^{[ 29 ]} The electrostatic potential must therefore be quite low (−0.9 V at pH = 0 according to the Zn Pourbaix diagram)^{[ 44 ]} to protect the Zn electrode against oxidation and corrosion.

On Zn(111) (Figure  4 ), formic acid adsorbs via reaction equation 1 as CH₂O*OH with ΔG_intermediate = 0.30 eV; however, because the electrode must be kept at a much lower potential, hydrogen adsorption (ΔG_H* = 0.75 eV) and hydrogen evolution are likely competing reactions. In the second reduction step, CH₂O*OH is reduced to either O*+CH₃OH with ΔG_intermediate = −0.02 eV or CH₂O with ΔG_intermediate = 0.17 eV. Both intermediates are downhill, especially at negative potentials, so we cannot credibly distinguish between the two pathways. The CH₃OH and CH₂O are desirable products, but, unfortunately, the reduction reaction likely continues with two more downhill reduction steps. In the third step, O* + CH₃OH is reduced to O* + *CH₃ (ΔG_intermediate = ‐0.36 eV) via reaction equaiton 2, while CH₂O is reduced to CH₃O* (ΔG_intermediate = ‐0.47 eV) via reaction equation 1. In the fourth step, both O* + *CH₃ and CH₃O* are reduced to CH₄ + O*, further lowering the free energy to ΔG_intermediate = −1.25 eV. We therefore predict that formic acid is reduced all the way to CH₄ on metallic Zn(111). The pathway is completed by converting O* to *OH and *OH to H₂O. ΔG_intermediate is −1.19 eV for CH₄ + *OH and −1.49 eV for CH₄ + H₂O, so O* conversion to *OH has a small energy cost, while *OH conversion to H₂O is downhill in energy.

Figure 4.

Free energies for possible formic acid electroreduction pathways on Zn(111) at U _vsRHE  = 0 V. After the first step, all subsequent steps are downhill in free energy, and we predict that formic acid is reduced all the way to CH₄.

3.5. Formic Acid Reduction on Pt(111), Pd(111), Ru(111)

Formic acid reduction behaves similarly on Pt(111) (Figure  5a), Pd(111) (Figure 5b), and Ru(111) (Figure 5c), so we discuss these three surfaces together. On these surfaces, formic acid is reduced to *CHO with a downhill energy even at 0 V vs RHE. However, the *CHO can further dissociate to H* + *CO and lower the free energy by between−0.6 and−1.3 eV. We believe that formic acid reduction will follow this pathway. Furthermore, none of these metals are good CO reduction catalysts, since they tend to strongly adsorb CO, such that they are poisoned by the CO molecules.^{[ 45} , ⁴⁶ , ^{47 ]} We calculate the CO desorption energies from H*+*CO to be 1.19 eV, 1.50 eV, and 1.60 eV for Pt(111), Pd(111), and Ru(111), respectively. In addition, we investigated the first HCOOH reduction step on Pt(100) and Pt(211), but similar to Au, changing facets only results in increased reactivity (Table S12, S14, Supporting Information). Hence, H* + *CO is still the most stable intermediate and adsorbs even strongly on Pt(100) and Pt(211) compared to Pt(111). It is therefore unlikely that Pt, Pd, and Ru are good catalysts for formic acid reduction.

Figure 5.

Free energies for the most likely formic acid electroreduction pathways on a) Pt(111), b) Pd(111), and c) Ru(111) at U _vsRHE  = 0 V. The *CHO, formed in the first step, likely dissociates to H* + *CO, which probably leads to CO poisoning of the surfaces.

3.6. Summary and Comparison of Formic Acid Reduction on the Different Surfaces

Figure  6 summarizes the pathways (a) and predicted products (b) for formic acid reduction on the different (111) metal surfaces. In the first reduction step, the *CHO intermediate is preferred on Au, Ag, Pt, Pd, and Ru, whereas both *CHO and CH₂O*OH are energetically accessible on Cu, and CH₂O*OH is preferred on Zn. The *CHO intermediate can dissociate (non‐electrochemically) to H* + *O on Cu, Pt, Pd, and Ru, resulting in a large energy stabilization, and on Ag with a small energy stabilization. This dissociation is unwanted because it either leads to CO poisoning of the surface or lower selectivity for desirable products, e.g., CH₂O or CH₃OH. The lower selectivity is observed experimentally for formic acid reduction on Cu, while CO poisoning likely occurs on Pt, Pd, and Ru.

Figure 6.

a) Summary and comparison of formic acid electro reduction pathways on Cu(111), Au(111), Ag(111), Zn(111), Pt(111), Pd(111), and Ru(111). b) Predicted HCOOH electroreduction products on Cu(111), Au(111), Ag(111), Zn(111), Pt(111), Pd(111) and Ru(111).

Electroreduction of the *CHO intermediate on Cu, Au, and Ag exclusively leads to CH₂O, electroreduction of CH₂O*OH on Cu also leads to CH₂O, while electroreduction of CH₂O*OH on Zn might lead to either CH₂O or O*+CH₃OH. Reduction of CH₂O on Au and Ag selectively produces CH₃OH; reduction of CH₂O on Cu can energetically produce both CH₃OH and CH₄, but experiments show that CH₃OH is the main product. Finally, Zn likely reduces the desirable CH₂O and O*+CH₃OH products to the much less desirable CH₄.

Overall, there seems to be an interesting link between the affinity of a surface to adsorb formic acid via the C* atom and the product distribution. Zn strongly favors adsorption via the O* atom and therefore converts any formed CH₃OH to CH₄. Cu adsorbs formic acid equally strongly via the O* and C* atoms and can therefore produce both CH₃OH and CH₄. Au favors adsorption via the C* atom compared to adsorption via O*, so Au strongly favors CH₃OH. Ag also favors adsorption via the C* in the first reduction step, but the adsorption is so weak that it is outcompeted by H* adsorption. Finally, Pt, Pd, and Ru have too strong an affinity for the C* atom, and this leads to C—H bond dissociation and *CO poisoning.

In the case of Cu(111), Au(111), Ag(111), and Zn(111), the first electroreduction step, where formic acid chemisorbs to the surface, is the potential limiting step (the most positive change in ΔG_intermediate). This influences the expected pH dependency. More alkaline conditions lower the concentration of HCOOH by deprotonation to HCOO⁻. However, importantly, we should not analyze HCOO⁻ reduction separately, because equilibrium ensures that the concentration‐dependent reaction free energy for HCOOH → HCOO⁻ + H⁺ is 0. Rather, by considering only the case for HCOOH, we deduce that the lower concentration of HCOOH at higher pH increases the entropy cost of chemisorbing HCOOH on the surface, lowers the coverage of intermediates on the surface, and overall lowers the formic acid electroreduction rate at constant potential vs RHE. This rationalization is tentatively supported by the collected formic acid electroreduction experiments (Table S1, Supporting Information), which are all performed under acidic conditions.

4. Conclusion

In this study, we have explored the possible formic acid electroreduction intermediates and products on Cu(111), Au(111), Ag(111), Zn(111), Pt(111), Pd(111), and Ru(111) catalysts. On Cu(111), formic acid is reduced to CH₃OH, CH₄, or C₂ products, which corresponds well with experiments. CH₃OH is obtained via the *CHO/CH₂O*OH, CH₂O, and CH₃O* intermediates in the first, second, and third electroreduction step, respectively. However, Cu(111) adsorbs H* at a less negative potential, so hydrogen evolution is a competitive reaction. The Cu(111) surface adsorbs intermediates via both C and O atoms, which allows for branching of the possible reaction pathways and consequently low selectivity to specific products. On Au(111), we expect that formic acid is reduced to CH₃OH via *CHO, CH₂O, and *CH₂OH reduction intermediates, at potentials where the competitive hydrogen evolution reaction rate is low. On Ag(111), formic acid can also be reduced to CH₃OH; however, the reduction requires more negative potentials than the hydrogen evolution reaction, and the Ag(111) surface may even be poisoned by H*. On Zn(111), formic acid is likely reduced into methane, while formic acid is converted to H*+*CO on Pt(111), Pd(111), and Ru(111), likely resulting in surface poisoning by *CO. In short, Au(111) is the most promising catalyst for formic acid reduction, with the possibility of being selective to CH₃OH and having low hydrogen evolution rates.

Conflict of Interest

The authors declare no conflict of interest.

Supporting information

Supplementary Material

Data Availability Statement

The DFT calculations and Python scripts are accessible online at https://erda.ku.dk/archives/2eb37aa2e5099fc29f9e1a04d10220d8/published‐archive.html.