Borohydride electro-oxidation on metal electrodes: structure, composition and solvent effects from DFT

An efficient borohydride (BH₄⁻) electro-oxidation in alkaline media facilitates rapid development of direct borohydride fuel cells (DBFC) as power sources for portable devices and other applications.^1–7  While there have been reviews on the experimental and theoretical studies of BH₄⁻ electro-oxidation,^2,3,8  there has been no reports that consolidate, compare systematically and update electro-oxidation of borohydride on vast structures and composition of metals and metal alloys. Such focused review is very insightful in terms of formation of new ideas for design and development of better catalysts. Theoretical investigations are mostly conducted using density functional theory (DFT). Beyond-DFT has also been used and will also be discussed.

One of the parameters in evaluating the effectiveness of an anode catalyst is the resulting initial oxidative adsorption structure of BH₄⁻. The theoretical anodic reaction of direct BH₄⁻ oxidation in a DBFC can be described as:

Furthermore, depending on the anode catalyst, pH, temperature and residence time on the anode, a thermo-catalytic hydrolysis reaction can also occur, generating H_2,(g):

The in situ H_2(g) evolution complicates the anode operation by generating a mixed potential and gas shielding the anode surface, thereby hampering the BH₄⁻ mass transfer and adsorption and lowering the effective ionic conductivity in the anode compartment. Ideally, an effective electro-catalyst should promote reaction (1) over reaction (2). Promoting (1) implies a more controlled interaction of BH₄⁻ and H₂O species with the anode catalyst. To discuss this in more detail, the initial step of oxidation is considered without solvation:

and with solvation effects:

Eqn (3) indicates that a borohydride ion “sees” a metal site, adsorbs and releases electron due to the potential difference. There are four ways by which the ion adsorbs depending on the catalyst: as a molecule, y= 4 (BH₄*) or as a dissociated structure, y=1–3 (i.e. BH*+3H*, BH₂*+2H*, BH₃*+H*). Same is the case for eqn (4), however the formation of either of these structures depends on the catalyst and the interaction with water (or solvent). It can be noted that a molecular adsorption structure (BH₄*) can inhibit fast formation of H₂ gas making the catalyst selective for direct pathway (1), on the other hand, an almost completely dissociated BH₄⁻ (BH*+3H*) can promote H₂ gas evolution and hence leading to reaction pathway (2), and the partially dissociated structures (BH₂*+2H*, BH₃*+H*) can be somewhere in between. Moreover, the strength of the adsorption of the borohydride especially for the molecularly adsorbed and partially adsorbed ones can impact shifts in the overpotentials (e.g. weak adsorption→can contribute to higher overpotential; strong adsorption→can contribute to lower overpotential). In this article, the initial oxidative adsorption structures and energies of borohydride with and without solvation effects obtained by DFT across many structures of metal anode catalysts are systematically compared.

A borohydride is a trigonal molecule (Fig. 1 inset) with a calculated B–H bond distance of 1.25 Å and H–B–H angle of 109°, in excellent agreement with experiment.⁹ 

Early theoretical studies on borohydride electro-oxidation tackled the adsorption of borohydride via eqn (3) on the (111) facet of 3d, 4d and 5d metals.^10–13  The (111) surface of the metal is shown in Fig. 1, along with the adsorption site (∗) labelled as bridge, hcp-hollow, fcc-hollow or top. The adsorption site is defined as the location of the boron atom, which is the center of the trigonal molecule. Table 1 lists the adsorption structure and energies depending on the transition metal.

The adsorption energies are referenced to isolated metal surface and the gas phase borohydride ion. Generally, there is an increasing trend on the adsorption energies as one goes from left to right within the same period of the periodic table. For the 3d transition metals, stronger adsorption leads to partially dissociated structures. However, the adsorption structures for 4d and 5d metals do not really correlate well with the adsorption energies. That is, a strong adsorption does not imply dissociated structures. For instance, in osmium (Os), the adsorption energy is strong but a molecular adsorption structure is noted. The adsorption structures for 4d and 5d transition metals are depicted in Fig. 2 and for the 3d metals, the structures are shown in Fig. 3.

The relationship between the adsorption structure and energies with the different facets of metal surfaces is investigated in Os.¹⁴  The facets considered, the adsorption structures and the adsorption energies are given in Table 2. Figures 4, 5 and 6 depict the (0001), (101̄0) and (11̄00) facets of Os and the adsorption structures, respectively. (0001) is the basal plane of the Os hcp bulk lattice, while the (101̄0) and (11̄00) are the prism planes. It is well known that (0001) (Fig. 4) is akin to the fcc (111) surface except for the stacking of the atomic layers. On the other hand, (101̄0) and (11̄00) have characteristic rows and trenches (Figs. 5–6). These two surfaces differ mainly in the position of the trench atoms. For instance, in the (101̄0), a trench atom is directly under the fourfold hollow site (h1) while in the (11̄00), it is under the bridge site (b2).

Based on the Table 2, it can be noted that the adsorption energy increases depending on how “open” the facets are. The (101̄0) and (11̄00) facets contain rows and trenches that expose the second layer atoms (hence “open”), thereby facilitating greater conformation and interaction with the borohydride than the (0001), which does not have these structures. The “b” distance, which is the distance between the parallel rows, is greater in (11̄00) than in the (101̄0), making the former more “open”.

Thus, the adsorption energy follows this order: (0001)<(101̄0)<(11̄00). In terms of the adsorption structure, it is observed that within the same kind of metal, the structure can be correlated with the order of the adsorption energies: BH₄*<BH₂*+2H*<BH*+3H*. For (0001), the lateral and vertical interlayer distances are generally not changed with respect to the relevant atoms’ distance from the adsorption site, in contrast to the that of (101̄0) and (11̄00). The adsorption of the molecule on the open surfaces involves binding or interaction with the atoms in the second layer as can be seen in Figs. 5 and 6.

Because of the development of anode catalysts that are both selective and cost-effective, the metal alloys have been studied theoretically.^15,16  In ref. 15, metal alloys based on Au, which are of Au₃M type (M is a 3d transition metal=Cr, Mn, Fe, Co and Ni), are investigated. Table 3 lists the adsorption structure and the adsorption energies on Au₃M (111) surfaces. The initial oxidative adsorption structures are depicted in Fig. 7.

It can be noted that for Au, alloying does not change the adsorption structure of borohydride. It remains molecular with a slight elongation of three B–H bonds. The molecule prefers the M site for the adsorption and the adsorption energy increases in the following order: Au<Au₃Ni<Au₃Mn<Au₃Cr<Au₃Fe<Au₃Co. The mechanism of the alloying effect is explained using the density of states (DOS) of the metal and the molecule as shown in Fig. 8. Without the M atom, in this case Fe, the DOS of the Au is fully occupied (Fig. 8(a)). Upon alloying with Fe, a spin-down DOS forms at and above the Fermi level (E_F). The local DOS shown in Fig. 8(b) indicates that these states originate from the Fe, explaining the preference of the borohydride on the M site. Figure 8(c) further clarifies the bonding interactions via partial charge density plots corresponding to DOS peaks near the E_F.

Other metal alloys have also been considered for instance, Pd–Ir.¹⁶  In this case, the alloying component is not a 3d transition metal but a 5d. Pd is a 4d metal. Two types of composition are studied: Pd₂Ir₁(111) and Pd₂Ir₂(111). The adsorption structure and energies are shown in Table 4. Figure 9 depicts the preferred adsorption structures.

Alloying the Pd with Ir, increases the free energy of the adsorption (i.e. more positive shift in the free energy). Such energy also increases upon increase in the composition of the alloying metal. The structures on the Pd–Ir alloys are shown in Fig. 9. Although both structures are molecular, the orientation of the B–H bonds and the position of the center B atom are different. For Pd₂Ir₁(111), the boron is at the bridge site with two H atoms spanning towards Pd top sites and the other two H atoms towards the Ir top sites. Thus, the molecule conforms with the surface via two H atoms. In the Pd₂Ir₂(111) (greater Ir composition), the molecule conforms with the surface via three H atoms spanning towards two Pd top sites and one Ir top site. This is the typical conformation of the borohydride on some 4d and 5d metals.

To consider the effect of the water as the solvent, the Gibbs free energy, ΔG_ads, for the initial oxidative adsorption of the BH₄⁻ based on eqn (4) is derived to draw the thermodynamics of the reaction. There are two ways by which this can be done, the vacuum slab method^17,18  and the double-reference method.^19,20  Mainly, the difference between the two methods is the inclusion of water molecules in the model (or supercell). In the former, as the name suggests, there are no water molecules that interact with the borohydride within the supercell and the effect of solvent is integrated via the free energies of H₂O adsorption (H₂O*) and the aqueous H₂O_aq (please see eqn (4)) alone, which are computed in separate supercells. The latter on the otherhand, includes water molecules within the supercell and so the effect of solvent is integrated via the usual free energies of H₂O adsorption (H₂O*) and the aqueous H₂O_aq as in the above method plus the borohydride–water interactions. The reader is suggested to refer to the given sources above for details. Figure 10 best captures the solvent effects using the double reference method on Au and Pt.¹² 

The re-orientation of the adsorbed borohydride occurs on Au. Two B–H bonds, instead of one, point away from the metal surface. There is negligible change in the orientation of the borohydride on Pt. It can be noted that for molecularly adsorbed structure (BH₄*), the effect of the interaction with water molecules in the orientation of borohydride can be observed. In terms of energetics (ΔG_ads), Fig. 11 shows that for Pt, the ΔG_ads as a function of electrode potential, V (SHE) is steeper when the vacuum slab method is used (dashed line) than when the double reference method is used (solid line). This is attributed to the dipole moment or polarizability effects.²¹  Basically, there is not much difference between the two methods for Au. Now, when the energies of the adsorption of borohydride in unsolvated (Table 1 of Section 1) and the solvated (Fig. 11) models of Au and Pt are compared, it can be noted that the trends are actually well-preserved.

As discussed above the adsorption of water (H₂O*) on the metal surface can bring about shifts in ΔG_ads. The study of the differences in the H₂O* structure on various facets of metal surfaces is worthwhile. Due to its interesting interaction of borohydride with Os as discussed previously, the H₂O* structure is obtained on this metal. The (0001), (101̄0) and (11̄00) surfaces of Os are used. Extensive adsorption configurations are evaluated, that is using the flat, up and down orientation of the molecule as well as its translation and rotation on the surface (please see Fig. 12). For instance, on (0001), for each water orientation (up and down), the molecule is rotated in-plane such that the O–H bond “parallel” to the surface can point towards different directions. Also, the O atom position is shifted to other high symmetric sites (i.e. translation). Thus, each adsorption configuration can be identified by X1–X2–X3, where X1 indicates the molecule orientation (flat (f), one H down (d) or one H up (u)), X2 is the position of the oxygen atom (hollow (h), top (t), bridge (b)) and X3 is the direction where the H expands as a result of planar rotation (towards the bridge (b) or hollow site (h)). Hence, f-t-b means a flat orientation with O atom on top site and the O–H1 bond (see Fig. 12(a) for the H1 atom) pointing towards the bridge. A total of 24 adsorption configurations on (0001) facet are considered. The adsorption energy per configuration is calculated based on the following:

where E_T, E_g and E_s are the total energy of adsorbed system, isolated molecule and the metal surface (or slab), respectively. Table 5 lists the adsorption energies on (0001).

The most stable H₂O* structure originated from the f-t-h initial configuration with an adsorption energy of −0.484 eV. Figure 13(a) depicts the final adsorption structure. The oxygen atom is on the top site and the O–H bonds are almost parallel to the surface. Both O–H bonds have the same lengths and the O–H–O angle is ∼105.495°. The O–Os distance is 2.302 Å. A small change in the interlayer distance, d₁₂ of ∼0.029 Å of the surface upon H₂O adsorption can be noted. Water monomer adsorption on metal surfaces has been studied in other metals previously.^22–25  Although a direct comparison cannot be made due to the different metals used and the size of the supercell, in general, it can be noted that the O–Os, O–H distances and the H–O–H angle are comparable to those reported in the literature for the fcc (111) metals. For Au(111), the literature reports a slightly weaker binding of water on the surface.²² 

Next, for the (101̄0), Fig. 12(b) shows the symmetric sites. Here, there are two bridge sites (b1, b2) corresponding to the a and b lateral distances. The same three orientations of H₂O, in-plane rotations, and configuration notations in (0001) are also employed in (101̄0). Table 6 gives the adsorption energies. The most stable site originated from the d-b1-h initial adsorption configuration. The adsorption energy is −0.660 eV. Figure 13(b) depicts the final adsorption state. The molecule is in bent configuration with the O–Os axis forming a 23.53° angle with the vertical axis (see Fig. 13(b), bottom panel). One of the O–H bonds (O–H2) points towards the b1 site and is longer than the O–H1 bond. This is because the H2 atom is lower (or closer to the surface) than the H1. This bent structure is due to the “near top” adsorption, in contrast to (0001) where the molecule directly sits on the top site. The O–Os distance in (101̄0) is shorter by 0.093 Å as compared to that of the (0001). We note that this bending of water molecule towards the trench resulted in a stronger adsorption.

Similar trend can be observed for (11̄00). Table 7 lists the adsorption energies and the final state of the molecule is shown in Fig. 13(c). This stable state originated from the d-b1-h configuration with an adsorption energy of −0.944 eV. The water molecule is more bent as compared to the previous two cases. The deviation of the O–Os axis from the vertical axis is 32.27° (see Fig. 13(c), bottom panel) and in the same way as in (101̄0), one O–H bond (i.e. O–H2) is lower (or closer to the surface) than the other (i.e. O–H1). Because of the larger deviation of the molecule from the top site, the O–Os vertical distance is much shorter (1.847 Å) and the O–H2 bond is longer (0.997 Å), all leading to a much stronger adsorption energy. This configuration of the H₂O molecule on (11̄00) is termed “edge site” adsorption. We can recall that the trench atom in (11̄00) is directly under the b1 site, creating a stronger interaction with the H2 atom of the H₂O molecule than that of the (101̄0). Based on the above, we clearly see the role of the trench atom and its contribution to the adsorption of water monomer on Os surfaces. The interaction of H₂O with other metals having the same trenches and rows can be predicted from this study on Os.

Next, what is the effect of the water interaction with Os surface on the initial oxidative adsorption of borohydride? Using eqn (4) and the vacuum slab model, the ΔG_ads as a function of electrode potential, U (SHE) for the different facets is shown in Fig. 14. First, it can be noted that the initial borohydride oxidation is favorable across potentials of interest for all facets, that is, the lines are below 0. The favorability of the initial oxidative adsorption of borohydride on Os surfaces increases in the following order: (0001)<(101̄0)<(11̄00). This suggests that the overpotentials for BH₄⁻ oxidation on the more open surfaces are lower. However, as a result of BH₄⁻ interaction with such surfaces (recall Figs. 5 and 6), larger hydrogen atom coverage can be expected, which in turn can also promote H₂ evolution. The investigation of the competing reactions is worthwhile. Lastly, in Fig. 14, comparison of the ΔG_ads vs. U_SHE with H₂O (thick lines) and without H₂O effects (thin lines), the absence of adsorbed H₂O shifts the Gibbs free energy of the oxidative BH₄⁻ adsorption downward. Thus, without the solvent effects, the adsorption energies obtained for the borohydride on metals are generally higher.

The effect of water on the borohydride electro-oxidation has been investigated on Au₃M(111) alloys, where M=Cr, Mn, Fe, Co and Ni.^26,27  Using the vacuum slab model, the ΔG_ads as a function of electrode potential φ(NHE) is shown in Fig. 15. It is observed that the initial oxidative adsorption of borohydride in Au₃M(111) is favourable at a potential range of −0.60 to −0.32 V vs. NHE. This is 0.17 to 0.45 V lower in potential compared to Au(111). The shift towards more positive potentials follows this order: Au₃Co<Au₃Fe<Au₃Ni<Au₃Mn<Au₃Cr<Au. For pure Au(111), increase in the coverage of borohydride leads to more positive potential shifts. Comparing the order of potential shifts for the Au₃M with that of the order of adsorption energies on the same systems without solvent effects as shown in Table 3 shows that the trends are not necessarily preserved especially for Au₃Ni, Au₃Mn and Au₃Cr. This may be due to the interaction of water with M.

Nanoparticles are different from the surfaces because they contain low coordinated sites such as edges and vertices. The dependence of borohydride adsorption on the size of metal nanoparticle has been studied using Os.¹⁴  In this study, the size ranges from 1.0 nm to 2.0 nm (see Fig. 16). The borohydride prefers the edge-vertex sites combination in the adsorption on all the nanoparticle sizes. When the size increases, the adsorption also increases. This is basically due to the increasing Os–Os lateral distance as the nanoparticle size is increased. Larger interatomic distances suggest less overlap of the d-states of the metal and therefore a higher entire d-band and more enhanced reactivity.

A single nanoparticle can contain several crystal planes or facets. For instance, in the Os-214 in Fig. 16, the (0001), (101̄0) and (11̄00) facets can be found. The dependence of adsorption of borohydride on the different facets of Os-214 is also reported in ref. 14. Please note that the facets in the metal surface and in the nanoparticles are quite different, in that the latter contains low coordinated sites. Thus, the configurations of borohydride can vary even on a single facet. For instance, in Fig. 17, three binding sites for (0001) and (101̄0) and two binding sites for (11̄00) are identified. The adsorption structure of borohydride is depicted in Fig. 17 and the adsorption energies are given in Table 7. For (0001), all the structures are molecular, BH₄* and the strongest adsorption can be found at site (b), which is a combination of edge-vertex Os atoms. In other metal nanoparticles, such reactivity of “combined” low-coordinated atoms can be expected. For the (101̄0), the three different binding sites also produce three types of borohydride structures: (d) BH₂*+2H*, (e) BH₄* and (f) BH₃*+H*. The strongest adsorption is in site (d), where the boron situates at the bridge site formed by two edge-atoms. For the (11̄00) facet, the two binding sites also produces two types of adsorption structures, namely: (g) BH*+3H* and (h) BH₂*+2H*. In this case there is no molecular adsorption. The strongest adsorption is noted in (h), where the B atoms sits in the hollow site. In the Os nanoparticle, the strongest adsorption for all facets is in (101̄0) which is not the same as that of the Os metal surface. First, it can be noted that the adsorbate structure on the (11̄00) facet of Os nanoparticle is BH₂*+2H*, which is different from the counterpart facet in the metal surface, BH*+3H*. The breaking of one H from B atom is restricted by the larger distance between the boron and the next neighboring bridge site where H can settle most favorably (marked by X in Fig. 17). In fact, on the nanoparticle, two hydrogen atoms sit instead on the top sites resultingin the less stable final BH₂*+2H* structure. Thus, the accessibility of the favorable hydrogen adsorption site contributes to the difference in the geometry of the adsorbate between the nanoparticle and the surface.

Using vacuum slab method, the initial oxidative adsorption of borohydride is evaluated for all facets of the Os nanoparticle. Figure 18 shows the comparison of the ΔG_ads with respect to the electrode potential U_SHE among (0001), (101̄0) and (11̄00) facets of Os-214 nanoparticle and (0001) facet of Os surface.

It can be observed that in the nanoparticle, the shifts towards more positive potentials are in this order: (101̄0)<(11̄00)<(0001), making the (101̄0) most active. To compare the (0001) surface in the Os surface and the same facet in the Os nanoparticle, the latter seemed more active. This is depicted in Fig. 18. The surface almost has the same ΔG_ads vs. U relationship as that of the (11̄00) facet of the nanoparticle. Moreover, it can be noted that the initial borohydride oxidation is favorable across potentials of interest for all facets in the Os nanoparticles just as in the case of the Os surface.

Theoretically, the electro-oxidation of borohydride on metal anode catalysts is investigated using density functional theory. Most of the study focuses on the initial oxidative adsorption of borohydride on many catalysts structures and compositions. The adsorption structures and energies predict catalysts’ selectivity. The agreement of these DFT studies with experiment have been discussed previously,⁸  however, comparison of these adsorption properties across many metal catalysts as a function of facets, sizes, alloy composition and solvent effects are still lacking. This review consolidates the studies on such properties and presents comparison among many metal structures and compositions with and without solvent effects. It has been noted that the adsorption energies of borohydride increases as one goes from left to right within the same period of the periodic table, that is, if the same structure (facet) of the metal surface is considered. However, the adsorption structures do not correlate well with adsorption energies. The strongest adsorption in 5d metals is found in Os however, the adsorption structure is still molecular (no dissociated H atoms). 4d transition metal presents the same scenario. For the 3d metals, the strongest adsorption is observed on Cr, with only one H atom dissociated. On the other hand, when different facets are compared within the same metal, the typical correlation between the adsorption energy and the adsorption structure can be observed. For instance, the more open the facet becomes (i.e. (0001)<(101̄0)<(11̄00)), the stronger is the adsorption and the more B–H bonds are dissociated leaving more H atoms on the surface. When the pure 5d metal is alloyed with 3d metal M (=Cr, Mn, Fe, Co and Ni), the adsorption energy increases in the following order: Au<Au₃Ni<Au₃Mn<Au₃Cr<Au₃Fe<Au₃Co. The borohydride remains molecular and prefers to adsorb on the M site. When the composition is changed, for instance, in Pd–Ir case, the greater is the composition of the alloying metal, Ir, the less is the adsorption energy. It can be expected that if this alloying metal is a 3d transition metal, M, the opposite trend can be noted.

The effect of the solvent has also been studied, especially on Au, Pt and Os. For the first two, the double reference method is used. It was observed that the water changes the orientation of borohydride on Au but not on Pt. The free energies of the electro-oxidation of borohydride, however is more affected in Pt than in Au, which is attributed to polarizability effect. For the alloyed Au, the inclusion of solvent effects via vacuum slab method indicates shifts to more positive potentials in the following order: Au<Au₃Ni<Au₃Mn<Au₃Cr<Au₃Fe<Au₃Co. This trend differs slightly from that of the catalysts without the solvent effects. For Os, the effect of surface facets has been studied. It was found that water binds on the surface via the edge site of (11̄00), however, the structure remains molecular for all the facets studied ((0001), (101̄0) and (11̄00)). This suggests that interaction adsorbed borohydride with water molecules is highly likely in borohydride electro-oxidation. On the nanoparticle, when the size is increased, its interaction with the borohydride also increases. Since, a nanoparticle also contains several facets, the interaction of borohydride with the planes of Os nanoparticle is also studied. It was noted that the most active facet is (101̄0), which is different from that of Os metal surface interaction. Nanoparticle presents more constricted adsorption sites than on the surface. The accessibility of hydrogen adsorption site is less favourable in nanoparticle and this contributes to the difference between the nanoparticle and the surface.

M. C. S. Escaño extends gratitude to Ministry of Education, Culture, Sports, Science and Technology, Japan (MEXT)/Japan Science and Technology Agency (JST)/Tenure Track Program for Innovative Research and Japan Society for Promotion of Science (JSPS) Grant-in-aid for Young Scientist B – Grant Number 15K21028 for research funds. The calculations are done using the ACCMS, Kyoto University – access funded by CII, University of Fukui and the High-Performance Computing Cluster Fukui (HPCCF) of Escaño Research Group, University of Fukui.