Carolyn N.
Virca
and
Theresa M.
McCormick
*
Portland State University, College of Liberal Arts & Sciences, Department of Chemistry, Post Office Box 751 CHEM, Portland, Oregon 97207, USA. E-mail: t.m.mccormick@pdx.edu
First published on 14th July 2015
Nickel pyridine 2-thiolate (Ni(PyS)3−) has shown good stability and activity as a H2 generation catalyst for use in solar energy storage. The experimentally proposed catalytic pathway is explored using DFT calculations. Free energy changes along the reaction coordinate, spin states, localization of charge and geometry of the intermediates were explored. Calculations were performed using Gaussian 09 with a B3P86/6-31+G(d) basis set and a CPCM water solvation model. Of particular interest were our findings that the first reduction occurs at the nickel rather than through non-innocent ligands and that water coordination is not favourable although protonation of the pyridyl nitrogen causes dechelation. Sequential and concerted proton coupled electron transfer were considered in the formation of the hydride.
The reaction pathway for this catalyst has been proposed from experimental observations (Fig. 1).6 In the proposed reaction pathway, Ni(PyS)3− (1−) undergoes protonation at a pyridyl nitrogen to form 2N, followed by reduction (2N−), then proton coupled electron transfer (PCET) to make the hydride, 3, and finally release of H2. In this study we examine electronic and structural properties of the reaction intermediates using DFT calculations.
A recent review highlighted several advancements towards using computations in rational design of water-reduction catalysts.18 Computational methods have previously been used to enhance knowledge of hydrogen evolution catalytic cycles in a similar manner,12,15,19–25 including those of Co glyoxime26–28 and Ni P2N229–32 type catalysts. The objective of this paper is to use DFT methods to probe the intermediates in the proton reduction pathway of Ni(PyS)3−. DFT can provide information on reaction intermediates that are difficult to probe experimentally, e.g. spin state, localization of charges and structure. Also identification of high-energy intermediates could help target catalyst modification to increase catalytic turnover frequency.
To this end, we have determined the energies of key intermediates through frequency calculations, identified a potential transition state and generated spin density maps for intermediates in various spin states. Geometries were determined through optimization calculations performed on possible intermediates of the reaction pathway. Optimized structures allowed us to explore the possibility of solvent coordination at a vacant site on the nickel during the catalytic cycle.
Crystal structure | B3LYP | B3LYP | B3LYP | B3P86 | B3P86 | B3P86 | |
---|---|---|---|---|---|---|---|
Ethanol CPCM | Gas | Water CPCM | Ethanol CPCM | Gas | Water CPCM | ||
Bond | Length | Length | Length | Length | Length | Length | Length |
Ni–N1 | 2.034 Å | 2.069 Å | 2.064 Å | 2.090 Å | 2.063 Å | 2.071 Å | 2.062 Å |
Ni–N2 | 2.041 Å | 2.102 Å | 2.101 Å | 2.065 Å | 2.041 Å | 2.044 Å | 2.044 Å |
Ni–N3 | 2.081 Å | 2.063 Å | 2.071 Å | 2.070 Å | 2.044 Å | 2.038 Å | 2.041 Å |
Ni–S1 | 2.541 Å | 2.565 Å | 2.609 Å | 2.603 Å | 2.549 Å | 2.554 Å | 2.549 Å |
Ni–S2 | 2.526 Å | 2.608 Å | 2.604 Å | 2.596 Å | 2.560 Å | 2.548 Å | 2.563 Å |
Ni–S3 | 2.518 Å | 2.607 Å | 2.564 Å | 2.613 Å | 2.545 Å | 2.514 Å | 2.545 Å |
We used the catalytic cycle laid out in the literature as a starting point for selecting intermediates of interest.6 Structures of other isomers, and alternative reaction paths were also considered. Structures of these intermediates were optimized with DFT to determine specific complex geometries, spin states, and total energies. Spin density maps of doublet, triplet and quartet states were generated using the cubegen utility in Gaussian 09.
The total energy of each intermediate was obtained from the total thermal and electronic energy from the frequency calculation (see ESI†). The ΔG° for each reaction at standard state was determined using eqn (1). Gproducts and Greactants is the sum of the total internal energy of the reaction components.
ΔG°rxn = Gproducts − Greactants | (1) |
The reduction potentials were determined using isodesmic reactions with the reduction of 2N to 2N− as the reference reaction (E°ref = −1.18 V vs. SCE).6 Using the experimentally reported reduction of the catalyst results in the calculated reduction potential of 2N being accurate by design. This method avoids issues of experimental reduction potentials reported in various solvents. The use of isodesmic reactions and experimentally determined reduction potentials as references eliminates systematic computation errors and results in a reported accuracy of the method of ca. 0.1 V.21 Using the free energy change of the reaction below, the reduction potentials were determined.
Ox + 2N− → Red + 2N | (2) |
Ox and Red are the oxidized and reduced intermediates of interest. The ΔG° for these isodesmic reactions was used to calculate E° of the desired reaction using the eqn (3) where F is Faraday's constant. All reduction potentials are reported vs. SCE to match experiment.
E° = −ΔG°/F + E°ref | (3) |
Since protonation at the ligand involves breaking the Ni–S or Ni–N bond, using an isodesmic reaction as a reference for determining the reaction pKa becomes problematic. The reported error for pKa calculations with B3P86 level of theory is 2.6 units if no isodesmic reaction is used to correct for systematic errors.41 To determine pKa's for the possible protonation sites and to calculate G the value of −264 kcal mol−1 for water solvated H+ was used.21 The pKa was calculated using eqn (4), using ΔG for the corresponding reaction below where B is the intermediate being protonated (see ESI† for more details).
B + H+ → BH pKa = −ΔG/RTln10 | (4) |
Energy coordinate diagrams were generated from the free energy change of each step in the catalytic cycle with the reduction potentials referenced either to SCE or −1.3 V vs. SCE and a pH of 12 as indicated in the figure captions. These were chosen to indicate the reaction coordinates at experimental conditions. The reaction was optimal at pH 12 and −1.3 V is the reduction potential of reduced fluorescein dye. The ΔG° values reported in the text are standard state, and the values in the figures are referenced to these experimental values. Any interaction with the sacrificial electron donor triethylamine, or the decomposition products of triethylamine were not considered in this study.
The structure of the transition state was determined using the QST2 method with compound 3 and 1−/H2 as the input geometries. The frequency calculation for the transition state structure showed one negative frequency. The normal mode for the negative frequency was identified as corresponding to the H–H bond stretch, suggesting that it is the correct transition state for the formation of H2.
Protonation can occur at either the sulphur or pyridyl nitrogen; resulting in dechelation of the ligand. Protonation of the pyridyl nitrogen is proposed by Eisenberg et al. A crystal structure of a similar compound, Ni(H-PyS)4,6 shows protonation of the pyridyl nitrogen in a square planar compound with only remaining sulphur coordination. However, for a similar complex with a tris-phosphine chelating ligand and a PyS ligand, it has been suggested that protonation occurs at the sulphur without detachment from the Ni, followed by proton migration to the detached pyridyl nitrogen.42 To probe the preferred protonation site we optimized structures with one protonated sulphur, both attached (ring closed) 2Sc and detached (ring open) 2So from the Ni, and the protonated pyridyl N (ring open), 2N. The open and closed structures of the protonated sulphur optimize to the same open structure, 2s (Fig. 2). Dechelation of the protonated ligand results in a square pyramidal structure. The meridinal geometry of 1− results in two distinct pyridyl nitrogen sites. Protonation to form the cis isomer is statistically more likely, and slightly lower in energy; as such we only consider the cis geometry.
Fig. 2 Optimized structures of 1−, 2N, 2S and 12−. Grey, carbon; blue, nitrogen; yellow, sulphur; white, hydrogen; light blue, nickel. |
Using the energy of a solvated proton, the pKa for the species protonated at the sulphur and nitrogen were calculated to be 3.2 and 9.9 respectively (eqn (5) and (6) in standard state). The experimentally determined pKa of the catalyst is 12.1.11 The calculated pKa of 9.9 for the species protonated at the pyridyl nitrogen is within the 2.6 unit error typical for this method.21 The photocatalytic experimental reaction conditions report an optimal pH of 12. Protonation at the sulphur under these conditions is less favoured than protonation at the nitrogen and unlikely to occur (Fig. 3).
1− + H+ → 2S ΔG° = −4.31 kcal mol−1, pKa 3.2 | (5) |
1− + H+ → 2N ΔG° = −13.5 kcal mol−1, pKa 9.9 | (6) |
Fig. 3 Energy coordinate diagram vs. SCE at pH 12 of: Green: reduction of 1−; red: protonation at S followed by reduction; blue: protonation at pyridyl N followed by reduction. |
The calculated pKa for the species protonated at the pyridyl nitrogen supports that the catalytic cycle starts with protonation and subsequent reduction to form complex 2N−.
2N + H2O → 4 ΔG° = +6.42 kcal mol−1 | (7) |
2N− + H2O → 4− ΔG° = +34.96 kcal mol−1 | (8) |
After reduction the catalyst is still five coordinate. Likewise, water coordination after reduction is thermodynamically unfavourable with ΔG = +34.96 kcal mol−1 (eqn (8)). Calculations indicate that although there is a vacant coordination site on complex 2N and 2N−, water coordination to make 4 and 4− is not energetically favourable and is unlikely to occur.
However, coordination of a hydroxide to 2N to get 5 ion is exothermic, ΔG° = −11.38 kcal mol−1. The reduction potential of 5 is −1.7 V, outside the potential available under experimental conditions indicating that either reduction of 2N is kinetically favoured over hydroxide coordination or hydroxide coordination is reversible. This intermediate is unlikely to contribute to the catalytic cycle and is not considered further.
Intermediate 2N− could be either a doublet with an unpaired electron on the metal, or a quartet with one unpaired electron on a ligand and two on the metal. The geometry was optimized for the complex in both spin states. Many redox catalysts operate through redox active (non-innocent) ligands.45,46 Spin localization can be difficult to determine experimentally for intermediates not long-lived enough for electron paramagnetic resonance (EPR) experiments, however localization of spin density can be determined through computationally generated spin density maps.
Spin density maps were generated to localize the electron density of the unpaired electrons (Fig. 4). The geometry optimizations of 2N− as a doublet and quartet show minimal change in structure. However, from the spin density maps it is evident that the electron density of the doublet of complex 2N− is heavily localized on the central nickel atom. The quartet spin has electron density more delocalized across the ligands, predominantly on the protonated pyridyl ligand. The doublet zero-point energy is lower by 87.2 kcal mol−1 than the quartet. The lower energy of the doublet state indicates that reduction of 2N results in a doublet state with electron density on the Ni centre, not the ligand.
Fig. 5 PCET alternative intermediates, reduction, 6; protonation, 7; and reduction with concerted proton migration, 8. |
To begin, we consider reduction of 2N− to yield 6. The calculated reduction potentials for 2N− is −2.1 V. Under experimental photocatalytic conditions, electron transfer is thought to occur through a reductive quenching mechanism such that the electrons would come from reduced dye, fluorescein, at a potential of −1.3 V.11 This would not have enough potential to reduce 2N−. Due to the high reduction potential for 2N−, we no longer consider option A.
Option B, proton transfer from the pyridyl nitrogen to the metal centre with 2N− reduction, results in a hydride intermediate, 8 that is −42.14 kcal mol−1 more stable than complex 6. Concerted proton transfer to the Ni and reduction has a calculated potential of −1.8 V. Protonation of this complex is unfavourable by 30.49 kcal mol−1. Attempts to model an intermolecular PCET reaction with an explicit water molecule did not optimize to a protonated structure.
Option C, formation of 7, is exergonic by 11.96 kcal mol−1, this corresponds to a pKa of 8.8 for the Ni–H on compound 7.
2N− + H+ → 7 ΔG° = −11.96 kcal mol−1, pKa 8.8 | (9) |
Experimentally, the reaction occurs at a pH of 12.11 Even considering a 2.6 pKa unit margin of error, equilibrium likely lies towards having higher concentrations of 2N−, rather than protonated compound 7. Interestingly Eisenberg et al. noted a small reduction wave in the CV around −1.0 V that did not show any catalytic behaviour and that disappeared upon neutralization.6 Our results indicate this peak may correspond to reduction of 7 that has a calculated reduction potential of −1.0 V. At pH 12 protonation of 2N− is only exothermic by 4.41 kcal mol−1. Although only small amounts of 7 may be formed it would be easily reduced, driving this reaction.
Finally, eqn (10), concerted protonation and reduction was calculated to require 2.3 kcal mol−1. In the absence of kinetic data, we hypothesize that the transition from 2N− to 3 will have a lower barrier of activation than the transition from 7 to 3 (Fig. 6).
2N− + e− + H+ → 3 ΔG° = 2.35 kcal mol−1 | (10) |
Fig. 6 Energy coordinate diagram for alternative paths to PCET. Red: 2N− is first protonated (7), then reduced at the Ni centre; green reduced to 6, and proton migration, 8; blue: concerted PCET. |
The ground state of the hydride, 3, and the doubly reduced compound, 6, could be either a singlet state or a triplet state. Knowing where electron density is localized is important for ligand design. The energy of the triplet state for 3 is 61.7 kcal mol−1 lower in energy than the singlet state. The spin density map of 3 in a triplet state shows electron density on both the metal and the hydride, supporting the hydride designation of this complex (Fig. 7). Likewise, the triplet state of 6 is 35.6 kcal mol−1 lower in energy than the singlet. The electron density is located primarily on the protonated ligand (Fig. 7). These electron density maps suggest that unlike the first reduction, which was centred on the nickel, the second reduction is ligand based.
Fig. 7 Spin density map for 3, triplet, left, and 6, triplet, right, showing spin density located heavily on the hydride and protonated PyS ligands respectively. |
From experiment it is still unclear if the formation of the hydride, 3, occurs through sequential or concurrent protonation and reduction. Structures of protonated intermediate 7 and reduced intermediates 6 and 8 were optimized suggesting these may be stable intermediates and sequential PCET may occur. However, the reduction potential of 2N− to make 6 and 8 are −2.1 V and −1.8 V respectively, making formation of these intermediates unlikely (Fig. 9). The protonated compound 7, if present in solution, is likely found in low concentrations, however the low reduction potential would drive this reaction toward the hydride 3. Additionally, electrochemical data indicates the presence of a second reduction at −1.0 V. Our calculated reduction potential of −1.0 V for 7 indicates correct identification of this reduction wave. Taking into consideration the pKa's and reduction potentials a sequential PCET where protonation occurs before reduction is most consistent with experimental results. However, the overall ΔG of 2.35 kcal mol−1 for PCET, where 2N− becomes 3, suggests a low barrier to a concerted mechanism. The structure of the hydride 3 shows the protons in close proximity, requiring very little energy to generate the transition state for H2 elimination to regenerate the initial catalyst 1−.
These calculations suggest two main routes to improve catalyst design, first sterically hindering the space around the nickel could prevent coordination of a hydroxide that forms a complex that cannot continue through the catalytic cycle. Secondly, the high energy steps are protonations, thus increasing the basicity of the ligand may result in a more active catalyst. Also the reduction potentials determined suggest that a less reducing photosensitizer would still provide the potential needed to drive this reaction and could allow for ideal matching of photosensitizer and catalysts. Further studies into this hypothesis are underway.
Footnote |
† Electronic supplementary information (ESI) available: Bond length errors, total energies and coordinates for all reported compounds. See DOI: 10.1039/c5dt02044a |
This journal is © The Royal Society of Chemistry 2015 |