Water Splitting on Multifaceted SrTiO₃ Nanocrystals: Computational Study

Abstract

Recent experimental findings suggest that strontium titanate SrTiO₃ (STO) photocatalytic activity for water splitting could be improved by creating multifaceted nanoparticles. To understand the underlying mechanisms and energetics, the model for faceted nanoparticles was created. The multifaceted nanoparticles’ surface is considered by us as a combination of flat and “stepped” facets. Ab initio calculations of the adsorption of water and oxygen evolution reaction (OER) intermediates were performed. Our findings suggest that the “slope” part of the step showed a natural similarity to the flat surface, whereas the “ridge” part exhibited significantly different adsorption configurations. On the “slope” region, both molecular and dissociative adsorption modes were possible, whereas on the “ridge”, only dissociative adsorption was observed. Water adsorption energies on the “ridge” ( $-1.50$ eV) were significantly higher than on the “slope” ( $-0.76$ eV molecular; $-0.83$ eV dissociative) or flat surface ( $-0.79$ eV molecular; $-1.09$ eV dissociative).

1. Introduction

Strontium titanate SrTiO₃ (STO) is a well-known material for water splitting [1,2,3,4,5,6,7,8,9,10,11,12]. The process of water adsorption and dissociation was studied in detail [13,14,15]. The effects of doping are investigated in [16]. Recent developments in nanocrystal synthesis offered materials with enhanced charge separation achieved by heterojunction [17,18], mesocrystallinity [19], or the exposed anisotropic facets [20,21]. Nanoparticles synthesized by Takata et al. [20] were made from STO doped by aluminium and photodeposited cocatalysts Rh/Cr₂O₃ and CoOOH, and demonstrated a quantum efficiency of up to 96% in the range of 350 to 360 nm. Synthesized six and eighteen-facet STO nanocrystals, as described in [21], demonstrated high catalytic activity in water splitting. When doped by Pt and Co₃O₄ on particular facets, these nanoparticles exhibited even higher performance. Such an improvement is attributed to the unique properties of anisotropic facets of the particles.

One of the key properties of a high-performance water splitting material is a low charge recombination rate. Adsorption of water and oxygen evolution reaction (OER) intermediates on stepped surfaces is expected to be qualitatively different than that on flat surfaces. Featuring surfaces of different orientations, the 18-facet nanoparticle provides a natural platform for efficient charge separation. The six-facet nanoparticle is essentially a cube with the {0 0 1} faces. Its edges, however, can be considered as a different reaction area from the {0 0 1} flat parts. Currently, to the best of our knowledge, the structure of the surface of these nanoparticles is described only at the nanoscale. To reveal the properties of different reaction areas of multi-faceted nanoparticles, an atomistic model has to be designed and tested.

In 18-facet STO nanoparticle, the {0 0 1} facets are combined with the facets parallel to the {1 1 0} crystallographic plane. Although for the real material, the surfaces {0 0 1} and {1 0 0} are not equivalent [22], for the present study, the distortion of the perovskite structure is irrelevant and the orientation of the surfaces are given relative to the cubic phase. As it was shown in [23], the ideal polar {1 1 0} surface is unstable. Its stabilization can be achieved by forming steps of the more stable {0 0 1} orientation [24].

In the present study, we propose an atomistic model of the {0 0 1} stepped surface, which is relevant to both six as well as eighteen-facet STO nanoparticles. On this surface, we simulate OER, as suggested by Nørskov [25]. The four-step reaction includes adsorption of H₂O, HO^*, O^*, and HOO^* species. The results will be used to perform thermodynamic simulations of the OER to obtain over-potential ${\eta }^{\mathrm{OER}}$ values.

An extensive investigation of OER on the flat STO surface was performed by Cui et al. [14]. For the flat {0 0 1}, the STO surface of the over-potential value of $0.66$ V was obtained.

2. Methods and Computational Details

We performed density functional theory (DFT) calculations using the Vienna Ab initio Simulation Package (VASP) [26,27,28,29]. Computational details are listed in

Table 1. Computational details.

Software	VASP 6 [27,28,29]
Exchange-correlation functional	GGA-PBE [31]
Pseudopotentials	Ultra Soft [32,33] potentials using the Projector Augmented Wave (PAW) method [34,35]
Smearing	Gaussian smearing
Ti-valence configuration	$3p^{6}3d^{2}4s^2$, valence 10, energy cutoff 222 eV, generated 07.09.2000
Sr-valence configuration	$4s^{2}4p^{6}5s^2$, valence 10, energy cutoff 229 eV, generated 07.09.2000
O-valence configuration	$2s^{2}2p^4$, valence 6, energy cutoff 400 eV, generated 08.04.2002
H-valence configuration	$1s^1$, valence 1, energy cutoff 250 eV, generated 15.06.2001
Spin polarization	Non-spin polarized calculation
Plane wave basis set cut-off	520 eV
Flat surface geometry (Figure 1)	$2a_0\times 4a_0$ surface cell, seven layers-thick, 20 Å vacuum gap, 144 atoms
Stepped surface geometry (Figure 2)	$2(2a_0\times 2a_0)$, $2\sqrt{2}{a}_0$ thickness, 10 Å vacuum gap, 104 atoms
Flat surface k-point mesh	$4\times 2\times 2$ Monkhorst-Pack [36]
Stepped surface k-point mesh	$4\times 4\times 2$ Monkhorst-Pack [36]

. Relaxed rhombohedral SrTiO₃ phase ($R\overline{3}c$) with the optimized lattice constant $a_0$ of $3.92$ Å was used. We compared flat surface and stepped surface, as shown in Figure 1 and Figure 2, respectively. Adsorbate was placed on both terminations of the slab to neutralize the electric dipole moment.

Adsorption energy $E_{\mathrm{ads},x}$ for the configuration x was calculated by Equation (1), where $E_x$ is the total energy of the configuration with the adsorbate, $E_{\mathrm{surf}}$ is the total energy of the corresponding surface without the adsorbate, $E_{{\mathrm{H}}_2\mathrm{O}}$ is the total energy of water, $E_{H_2}$ is the total energy of molecular hydrogen, and coefficients $c_1$ and $c_2$ are determined so that the total number of particles on the left hand side of the equation is zero. Factor $\frac{1}{2}$ is a result of the adsorbate being placed on both terminations of the slab. Lower adsorption energy corresponds to the stronger binding.

All figures were created in the VESTA visualization system [30].

$$E_{\mathrm{ads},x}=\frac{1}{2}(E_x-(E_{\mathrm{surf}}+c_1{E}_{{\mathrm{H}}_2\mathrm{O}}+c_2{E}_{\mathrm{H}2}))$$

(1)

3. Results and Discussion

3.1. Water Adsorption

It is important to understand how water adsorption on stepped surfaces distinguishes from that on flat surfaces. On flat surfaces, the most preferable water adsorption site is atop titanium. The stepped model of faceted surfaces features three regions: the ridge, slope, and gully, marked on Figure 2.

On flat surfaces, two adsorption modes are possible: molecular (Figure 3a; $E_{\mathrm{ads}}=-0.79$ $\mathrm{eV}$) and dissociative (Figure 3b; $E_{\mathrm{ads}}=-1.09$ $\mathrm{eV}$), with dissociative being more energetically favorable, which is in agreement with Reference [13]. In [13], it was demonstrated that there is no significant transition barrier ($0.09$ eV) between the two adsorption modes. On the stepped surface, the situation is more complex and each adsorption region should be discussed separately.

On the Slope region, there are several possible adsorption configurations. The most energetically favorable one is dissociative adsorption along Slope (Figure 4d; $E_{\mathrm{ads}}=-0.83$ $\mathrm{eV}$), followed closely by molecular adsorption (Figure 4b,c with $E_{\mathrm{ads}}$ of $-0.76$ $\mathrm{eV}$ and $-0.71$ $\mathrm{eV}$, respectively). Molecular adsorption energy on the slope was loosely dependent on the orientation of water molecules. Dissociation towards the gully or ridge was less favorable (Figure 4e,f with $E_{\mathrm{ads}}$ of $-0.62$ $\mathrm{eV}$ and $-0.41$ $\mathrm{eV}$, respectively). Although, similarly to flat surfaces, one of the possible dissociative configurations on the slope was more favorable than molecular ones, the difference between energies was not as large, thus it cannot be unequivocally concluded, wherein the adsorption mode dominates.

On the Ridge region, only one configuration was observed: water dissociation accompanied by spontaneous oxygen vacancy formation (Figure 4a). It also had the lowest adsorption energy (strongest adsorption) of $-1.50$ eV among all the tested configurations. The investigation of the question regarding whether the ridge breaks down irreversibly or whether the vacancy is healed during subsequent water adsorption is out of scope of this paper.

On the Gully region, only one dissociative configuration was found, but only hydrogen was adsorbed, while oxygen was on the slope. Moreover, this configuration had a relatively weak binding (Figure 4g, $E_{\mathrm{ads}}=-0.49$ $\mathrm{eV}$), hence we did not investigate it further.

3.2. Oxygen Evolution Reaction (OER) Intermediates

To perform thermodynamic simulations to estimate STO photo-catalytic activity, it is necessary to compute adsorption energies for oxygen evolution reaction (OER) intermediates: HO^*, O^*, and HOO^*, where the star * denotes the active adsorption site. All energies are compiled in

Table 2. Adsorption energies of oxygen evolution reaction (OER) intermediates on different types of surfaces.

Surface Type	HO*, eV	O*, eV	HOO*, eV
Flat surface	0.77 (Figure 5a)	2.87 (Figure 5b)/ 3.52 (Figure 5c)	3.64 (Figure 5d)/ 4.24 (Figure 5e)
Slope	1.35 (Figure 6a)	2.93 (Figure 6b)/ 4.11 (Figure 6c)	3.48 (Figure 6d)/ 4.52 (Figure 6e)
Ridge	1.24 (Figure 7a)	2.40 (Figure 7b)	3.13 (Figure 7c)

.

Results for the flat surface are shown in Figure 5. HO^* had only one possible configuration, as shown in Figure 5a. There were two possible configurations for O^*: where oxygen from the adsorbate bonded to surface oxygen, denoted as O^*_(Osurf) (Figure 5b), and where oxygen was atop titanium, denoted as O^* (Figure 5c). The O^*_(Osurf) adsorption energy was much lower ($2.87$ eV versus $3.52$ eV), thus it was more energetically favorable. The HOO^* adsorbate also had two configurations: one where hydrogen bonded to surface oxygen (Figure 5d), denoted as H_(Osurf)OO^*, and the other where hydrogen bonded to the adsorbate’s oxygen (Figure 5e), denoted as HOO^*. In [14], only the second configuration was mentioned, although its adsorption energy was significantly higher than that of the H_(Osurf)OO^* configuration: $4.24$ eV versus $3.64$ eV.

The adsorption of intermediates on the slope (Figure 6) was similar to that of the flat surface: only one HO^* configuration (Figure 6a), O^*_(Osurf) (Figure 6b), and O^* (Figure 6c), and two HOO^* configurations, namely H_(Osurf)OO^* (Figure 6d) and HOO^* (Figure 6e) were observed. O^*_(Osurf) was more energetically favorable than O^* ($2.93$ eV versus $4.11$ eV) and H_(Osurf)OO^* was more favorable than HOO^* ($3.48$ eV versus $4.52$ eV), analogous to the flat surface. HO^* on the slope had higher adsorption energy than HO^* on the flat surface ($1.35$ eV versus $0.77$ eV).

Results for the intermediates on the ridge region are shown in Figure 7. For each intermediate, only one adsorption configuration was observed. HO^* (Figure 7a) and HOO^* (Figure 7c), similarly to the water adsorption, were accompanied by spontaneous oxygen vacancy formation, while O^* bonded between surface oxygen and titanium (Figure 7b).

4. Conclusions

We have performed a detailed investigation of water adsorption and oxygen evolution reaction (OER) intermediate adsorption on strontium titanate SrTiO₃ flat and stepped surfaces. In contrast to the flat surface, the stepped surface, a significant part of which comprises the ridge region, demonstrated high adsorption energies as well as pronounced structural transformations caused by the adsorbate. Our findings suggest that:

• The ridge region permits dissociative water adsorption only, accompanied by spontaneous formation of oxygen vacancy;
• Results for the flat surface are in agreement with other computational studies [13];
• On the slope region, both molecular and dissociative adsorption modes are possible;
• Adsorption of both water and its intermediates on the slope region is similar to that on flat surfaces;
• Except for atomic hydrogen, no adsorption was observed on the gully region; and
• There are different adsorption configurations of OER intermediates possible on flat surfaces and slope regions.