Variation in Surface Ionization Potentials of ... - Semantic Scholar
Variation in Surface Ionization Potentials of Pristine and Hydrated BiVO4 Rachel Crespo-Otero1* and Aron Walsh2 1
School of Biological and Chemical Sciences, Queen Mary University, London E1 4NS, UK 2
Department of Chemistry, University of Bath, Claverton Down, Bath BA2 7AY, UK SUPPORTING INFORMATION
Table S1. Surface energies of BiVO4 computed as πΎ = Slab models n 1 2 3 4
(πΈπ πππβππΈππ’ππ) 2A
.
πΎ (J/m2) PBEsol PBE 0.303 0.202 0.306 0.206 0.303 0.205 0.300 0.202
Table S2. Electronic band gaps and ionization potentials (IP) for the surface slabs and bulk crystal of BiVO4. The bulk values were determined by aligning the vacuum level of the slabs to the bulk eigenvalues via the O 1s core levels. All values are in EV. Slab model s n 1 2 3 4
Band gaps
Surface IP
PBE
PBEsol
PBE
PBEsol
2.08 2.04 2.04 2.04
2.06 2.02 2.01 2.02
7.18 7.19 7.18 7.19
7.23 7.24 7.24 7.24
Bulk IP PBE 0.084 0.048 0.053 0.054
PBEsol -0.074 0.055 0.059 0.060
1
Table S3. Calculated properties for 10 snapshots obtained from the 300 K dynamics of a water monolayer on the surface of BiVO4 (PBE functional). All values are in eV. Snapshots
Gap
IP
1
1.94
7.29
6.09
1.21
2.85
2
1.94
7.26
6.21
1.04
2.81
3
1.86
7.24
6.07
1.17
2.79
4
1.96
7.26
6.80
0.45
2.81
5
1.76
7.23
6.24
0.98
2.78
6
2.00
7.20
6.35
0.84
2.74
7
1.97
7.25
6.48
0.78
2.80
8
1.80
7.18
5.97
1.22
2.74
9
1.92
7.29
6.47
0.82
2.84
10
1.83
7.32
6.23
1.08
2.87
Average
1.90
7.25
6.29
0.96
2.80
BiVO4
IP
BiVO4 /H 2O
BiVO4 /H 2O
We
BiVO4
-We
e0UVBM
COMPUTATIONAL METHODS AND MODELS Periodic electronic structure calculations were carried out using the VASP code.27 The PBEsol,28 PBE,29 HSE0630 exchange-correlation functionals were used with projectoraugmented pseudopotentials. Bulk calculations were performed with a converged 6Γ4Γ6 k-point mesh and an energy cut-off of 700 eV for the plane waves. The bulk cell parameters were optimised separately with the PBEsol and PBE functionals and held fixed in the surface calculation. The cell parameters and the interatomic distances were in good agreement with the experimental data and previous computational models.25
2
The most stable termination of the (010) surface was modelled using slabs of different sizes from 1 to 4 using a k-grid of 6Γ4Γ1. A vacuum of 15 Γ was used to separate the periodic images in the z direction. The smallest slab has 4 layers of metal oxide. The chosen parameters were well converged with respect to the total energies and the associated surface energies. The atomic coordinates in the surface slabs were completely relaxed.
To analyse the effect of water absorption on the energy alignments, the smallest slab interacting with one water molecule was used. These asymmetric slabs were computed using electrostatic corrections (monopole/dipole, quadrupole corrections IDIPOL and LDIPOL keywords in the VASP code) for the energy and potential corrections in the direction perpendicular to the surface. The number of water molecules (14 molecules per unit cell) required to reproduce the density of liquid water was used to fill a vacuum of 15 Γ . This orthogonal cell was previously equilibrated with Tinker and the TIP3P force field at 298 K.31 To compute the ionization potentials using this slab model, an extra vacuum of 15 Γ was added and the position of the atoms re-optimized. This model is called frozen-liquid in the text.
Supercell calculations of the monolayer and frozen-liquid models were performed by expanding the original monoclinic structure according the following matrix transformation:
Γ¦ 4 0 0 ΓΆ Γ§ Γ· Γ§ 2 2 0 Γ· Γ§ 0 0 1 Γ· Γ¨ ΓΈ
3
A slab of base 14.47Γ14.47 Γ 2, and c = 11.58 Γ was built and a vacuum of 15 Γ or 25 Γ was added. These calculations were performed using a cut-off of 700 eV for the for the plane waves using the gamma point version of VASP code. The energetic and electrostatic potentials of the supercell calculations and the periodic equivalent cells were in very good agreement.
The monolayer model with a vacuum of 25 Γ was used to explore the dynamics of the water in contact with water relaxing water and the surface with the PBE functional and a cut-off of 400 eV for the plane-waves. The energy differences obtained between different steps are similar with both used cut-off values. The system was equilibrated for 10 ps and production for 20 ps. The temperature was kept constant at 298 K using the Nose-Hoover thermostat scaling every 40 fs. Hydrogen atoms were computed as deuterium, and the time step was selected as 1 fs. Ten random snapshots were chosen to evaluate the ionization potentials, the energies were recomputed with the 700 eV a cut-off for the plane-waves. The vacuum levels were determined using the Macrodensity program.32
To evaluate the ionization potentials of the (010) surface in contact with water, we considered the previous static and dynamics models. These calculations were performed using an electrostatic correction to eliminate any dipole moment due to the asymmetric slab model. One distinctive feature of the electrostatic potential of such slab models is the presence of two plateaus (vacuum levels) associated to distinct surfaces: the BiVO4 surface and the BiVO4/H2O surface. In addition to ionization potentials and work functions, the work to transport an electron from the semiconductor to the solution can be obtained directly from this model:
4
BiVO4 /H 2O
We
BiVO4
-We
= e0 (V
BiVO4
-V ) H2O
To obtain the electrochemical potentials with reference to the hydrogen electrode, the potentials were adjusted according to:22 e0UVBM = IPBiVO +WH + - D f GHg ,0+ 4
WH + is the work function of H+ in aqueous medium and the free Gibbs energy of
formation of H+ in vacuum, we considered here the values of 11.36 eV and 15.81 eV, respectively.
5
School of Biological and Chemical Sciences, Queen Mary University, London E1 4NS, UK 2
Department of Chemistry, University of Bath, Claverton Down, Bath BA2 7AY, UK SUPPORTING INFORMATION
Table S1. Surface energies of BiVO4 computed as πΎ = Slab models n 1 2 3 4
(πΈπ πππβππΈππ’ππ) 2A
.
πΎ (J/m2) PBEsol PBE 0.303 0.202 0.306 0.206 0.303 0.205 0.300 0.202
Table S2. Electronic band gaps and ionization potentials (IP) for the surface slabs and bulk crystal of BiVO4. The bulk values were determined by aligning the vacuum level of the slabs to the bulk eigenvalues via the O 1s core levels. All values are in EV. Slab model s n 1 2 3 4
Band gaps
Surface IP
PBE
PBEsol
PBE
PBEsol
2.08 2.04 2.04 2.04
2.06 2.02 2.01 2.02
7.18 7.19 7.18 7.19
7.23 7.24 7.24 7.24
Bulk IP PBE 0.084 0.048 0.053 0.054
PBEsol -0.074 0.055 0.059 0.060
1
Table S3. Calculated properties for 10 snapshots obtained from the 300 K dynamics of a water monolayer on the surface of BiVO4 (PBE functional). All values are in eV. Snapshots
Gap
IP
1
1.94
7.29
6.09
1.21
2.85
2
1.94
7.26
6.21
1.04
2.81
3
1.86
7.24
6.07
1.17
2.79
4
1.96
7.26
6.80
0.45
2.81
5
1.76
7.23
6.24
0.98
2.78
6
2.00
7.20
6.35
0.84
2.74
7
1.97
7.25
6.48
0.78
2.80
8
1.80
7.18
5.97
1.22
2.74
9
1.92
7.29
6.47
0.82
2.84
10
1.83
7.32
6.23
1.08
2.87
Average
1.90
7.25
6.29
0.96
2.80
BiVO4
IP
BiVO4 /H 2O
BiVO4 /H 2O
We
BiVO4
-We
e0UVBM
COMPUTATIONAL METHODS AND MODELS Periodic electronic structure calculations were carried out using the VASP code.27 The PBEsol,28 PBE,29 HSE0630 exchange-correlation functionals were used with projectoraugmented pseudopotentials. Bulk calculations were performed with a converged 6Γ4Γ6 k-point mesh and an energy cut-off of 700 eV for the plane waves. The bulk cell parameters were optimised separately with the PBEsol and PBE functionals and held fixed in the surface calculation. The cell parameters and the interatomic distances were in good agreement with the experimental data and previous computational models.25
2
The most stable termination of the (010) surface was modelled using slabs of different sizes from 1 to 4 using a k-grid of 6Γ4Γ1. A vacuum of 15 Γ was used to separate the periodic images in the z direction. The smallest slab has 4 layers of metal oxide. The chosen parameters were well converged with respect to the total energies and the associated surface energies. The atomic coordinates in the surface slabs were completely relaxed.
To analyse the effect of water absorption on the energy alignments, the smallest slab interacting with one water molecule was used. These asymmetric slabs were computed using electrostatic corrections (monopole/dipole, quadrupole corrections IDIPOL and LDIPOL keywords in the VASP code) for the energy and potential corrections in the direction perpendicular to the surface. The number of water molecules (14 molecules per unit cell) required to reproduce the density of liquid water was used to fill a vacuum of 15 Γ . This orthogonal cell was previously equilibrated with Tinker and the TIP3P force field at 298 K.31 To compute the ionization potentials using this slab model, an extra vacuum of 15 Γ was added and the position of the atoms re-optimized. This model is called frozen-liquid in the text.
Supercell calculations of the monolayer and frozen-liquid models were performed by expanding the original monoclinic structure according the following matrix transformation:
Γ¦ 4 0 0 ΓΆ Γ§ Γ· Γ§ 2 2 0 Γ· Γ§ 0 0 1 Γ· Γ¨ ΓΈ
3
A slab of base 14.47Γ14.47 Γ 2, and c = 11.58 Γ was built and a vacuum of 15 Γ or 25 Γ was added. These calculations were performed using a cut-off of 700 eV for the for the plane waves using the gamma point version of VASP code. The energetic and electrostatic potentials of the supercell calculations and the periodic equivalent cells were in very good agreement.
The monolayer model with a vacuum of 25 Γ was used to explore the dynamics of the water in contact with water relaxing water and the surface with the PBE functional and a cut-off of 400 eV for the plane-waves. The energy differences obtained between different steps are similar with both used cut-off values. The system was equilibrated for 10 ps and production for 20 ps. The temperature was kept constant at 298 K using the Nose-Hoover thermostat scaling every 40 fs. Hydrogen atoms were computed as deuterium, and the time step was selected as 1 fs. Ten random snapshots were chosen to evaluate the ionization potentials, the energies were recomputed with the 700 eV a cut-off for the plane-waves. The vacuum levels were determined using the Macrodensity program.32
To evaluate the ionization potentials of the (010) surface in contact with water, we considered the previous static and dynamics models. These calculations were performed using an electrostatic correction to eliminate any dipole moment due to the asymmetric slab model. One distinctive feature of the electrostatic potential of such slab models is the presence of two plateaus (vacuum levels) associated to distinct surfaces: the BiVO4 surface and the BiVO4/H2O surface. In addition to ionization potentials and work functions, the work to transport an electron from the semiconductor to the solution can be obtained directly from this model:
4
BiVO4 /H 2O
We
BiVO4
-We
= e0 (V
BiVO4
-V ) H2O
To obtain the electrochemical potentials with reference to the hydrogen electrode, the potentials were adjusted according to:22 e0UVBM = IPBiVO +WH + - D f GHg ,0+ 4
WH + is the work function of H+ in aqueous medium and the free Gibbs energy of
formation of H+ in vacuum, we considered here the values of 11.36 eV and 15.81 eV, respectively.
5
