Methane Mono-Oxidation Electrocatalysis by Palladium and Platinum Salts

ABC

METHANE MONO-OXIDATION ELECTROCATALYSIS BY PALLADIUM AND PLATINUM SALTS

R. Soyoung Kim

B.S. and M.S., Chemistry Seoul National University, 2014

SUBMITTED TO THE DEPARTMENT OF CHEMISTRY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY IN CHEMISTRY

AT THE

MASSACHUSETTS INSTITUTE OF TECHNOLOGY

MAY 2020

Signature of Author: ______Department of Chemistry May 8, 2020

Certified by: ______Yogesh Surendranath Paul M. Cook Career Development Associate Professor Thesis Supervisor

Accepted by: ______Robert W. Field Haslam and Dewey Professor of Chemistry Chairman, Departmental Committee on Graduate Students Title page

Signature page

This doctoral thesis has been examined by a Committee of the Department of Chemistry as follows:

Professor Mircea Dincă ______Department of Chemistry Thesis Committee Chairman

Professor Yogesh Surendranath ______Department of Chemistry Thesis Supervisor

Professor Christopher Cummins ______Department of Chemistry Committee Member

Abstract

METHANE MONO-OXIDATION ELECTROCATALYSIS BY PALLADIUM AND PLATINUM SALTS

R. SOYOUNG KIM

SUBMITTED TO THE DEPARTMENT OF CHEMISTRY ON MAY 8, 2020 IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY IN CHEMISTRY

Abstract

Selective oxidation of methane to methanol would enable better utilization of natural gas resources. Many homogeneous metal ions activate methane under mild conditions, but turning this reactivity into catalysis requires a viable oxidation step. Electrochemistry offers unique advantages in this regard, and this thesis demonstrates two mechanistically distinct approaches for methane functionalization electrocatalysis. Following the first approach, a novel high-valent Pd complex with exceptional methane functionalization reaction rates is electrochemically generated in fuming sulfuric acid. We present a structural model of this complex as a PdIII dimer with a Pd–Pd bond and a 5-fold O-atom sulfate/bisulfate coordination environment at each Pd atom. We also discover, using EPR spectroscopy, a mixed-valent II,III Pd2 complex in the electrochemical oxidation sequence. From these and redox potential measurements, II III a comprehensive thermodynamic landscape for the oxidation of Pd to Pd 2 emerge for the first time, and III the critical role of M–M and M–L bonding in driving the electrochemical self-assembly of Pd 2 is exposed. Building on these structural studies, we arrive at a mechanistic model for methane functionalization by III Pd 2 that simultaneously yields methyl bisulfate (MBS) and methanesulfonic acid (MSA). Rate-limiting H III atom abstraction by Pd 2 and product bifurcation from the methyl radical intermediate is proposed based on experimentally determined rate laws and observations with radical scavengers and initiators. DFT calculations likewise support a shared outer-sphere proton-coupled electron transfer (PCET) reaction for the generation of both products. Following the second approach for methane functionalization electrocatalysis, we establish an electrochemical solution to the long-standing oxidant problem of Shilov’s PtII catalyst. Inner-sphere electron transfer facilitates the electrochemical oxidation of PtII to PtIV on Cl-adsorbed platinum electrodes without concomitant methanol oxidation. The favorable catalytic property of this electrode is exploited for the continuous regeneration of the PtIV oxidant during PtII-catalyzed methane functionalization. The critical PtII/IV ratio is maintained via dynamic modulation of the electric current and in situ monitoring of the solution redox potential. Thereby, we show stable and sustained turnover of Shilov’s catalyst for the first time.

Thesis Supervisor: Yogesh Surendranath Title: Paul M. Cook Career Development Associate Professor

Table of Contents

Title page ...... 1 Signature page ...... 3 Abstract ...... 5 Table of Contents ...... 7 Table of Figures ...... 10 Table of Schemes ...... 17 Table of Tables ...... 18 List of Abbreviations ...... 20 1. Introduction ...... 21 1.1. Mild and Selective Oxidation of Methane to Methanol ...... 21 1.2. Organometallic C–H Activation for Methane Functionalization ...... 22 1.2.1. Categories of Organometallic C–H Activation ...... 22 1.2.2. Challenges Involving the Oxidation Step ...... 25 1.3. Electrochemical Methane Functionalization Approaches ...... 26 1.3.1. Potential Advantages of Methane Functionalization Catalysis by Electrochemical Oxidation ...... 26 1.3.2. Mechanism-based Adaptation of Electrochemical Oxidation for Organometallic Methane Functionalization Catalysis ...... 27 1.4. Layout of the Thesis ...... 30 1.5. Summary and Prospectus ...... 30 1.6. References ...... 32 III 2. Structure of Pd 2 and Its Mechanism of Formation via Electrochemical Oxidation . 36 2.1. Introduction ...... 36 III 1 2.1.1. Electro-generated Pd 2 in sulfuric acid ...... 36 III 2.1.2. The need for elucidation of the structure of Pd 2 and its formation mechanism ...... 38 2.2. Results and Discussions ...... 39 III 2.2.1. Structure of Pd 2 ...... 39 II,III 2.2.2. Identification and Structural Assignment of a Pd2 Intermediate ...... 44 III 2.2.3. Structural and Thermochemical Basis for Electrochemical Pd 2 Formation ...... 46 2.3. Conclusions ...... 48 2.4. Methods and Additional Information...... 49 2.4.1. Chemicals, Materials and General Remarks ...... 49

2.4.2. Preparation of samples for X-ray absorption and Raman spectroscopy ...... 50 2.4.3. Preparation of samples for EPR spectroscopy ...... 53 II III 2.4.4. Determination of [Pd ] and [Pd 2] from UV–Vis spectroscopy ...... 55 2.4.5. X-ray Absorption Spectroscopy ...... 57 2.4.6. Raman Spectroscopy ...... 61 2.4.7. Electron Paramagnetic Resonance (EPR) spectroscopy ...... 63 2.4.8. Determination of Thermodynamic Quantities ...... 64 III 2.4.9. Computational Modeling of Pd 2 ...... 69 2.5. References ...... 71 III 3. Reaction Mechanism of Rapid and Selective Methane Functionalization by Pd 2 .... 75 3.1. Introduction ...... 75 3.2. Results and Discussions ...... 76 3.2.1. Experimental Observations ...... 76 3.2.2. Evaluation of mechanistic models with DFT calculations ...... 85 3.2.3. Discussions...... 93 3.3. Conclusions ...... 95 3.4. Methods and Additional Information...... 96 3.4.1. General methods ...... 96 3.4.2. NMR spectroscopy ...... 100 3.4.3. In situ NMR for reaction rate measurements ...... 104 3.4.4. Ex situ measurements for extraction of rate constants ...... 107 3.4.5. Derivation of rate laws ...... 111 3.4.6. Peroxydisulfate-initiated methane oxidation ...... 114 3.4.7. DFT calculation ...... 116 3.5. References ...... 120 4. Electrochemical Oxidation of Platinum Salts for Continuous Methane Hydroxylation Catalysis in Dilute Aqueous Acid ...... 123 4.1. Introduction ...... 123 4.2. Results and Discussions ...... 125 4.2.1. Identification of a suitable electrode for PtII-catalyzed Electrochemical Methane Oxidation Reaction (EMOR) ...... 125 4.2.2. Sustained methane oxidation catalysis via dynamic electrochemical control of the PtII:PtIV ratio ...... 128 4.2.3. Analysis of methane oxidation products from the EMOR reactor...... 132 4.2.4. Outlook for practical methane oxidation ...... 133

4.3. Conclusions ...... 134 4.4. Methods and Additional Information...... 134 4.4.1. Materials and methods ...... 134 4.4.2. Evaluation of PtII electro-oxidation ...... 144 4.4.3. Faradaic efficiency measurements ...... 151 – II 4.4.4. Effect of [H2SO4] and [Cl ] on the catalytic C-H oxidation activity of Pt ...... 153 4.4.5. Mitigation of Pt0 formation in the EMOR reactor ...... 154 4.4.6. Additional information on the EMOR reactor ...... 157 4.4.7. Simulation of reactions in the EMOR reactor ...... 160 4.5. References ...... 164 Acknowledgements ...... 167

Table of Figures

Figure 2.1. Investigation of PdII oxidation in concentrated (95–98%) sulfuric acid at room temperature by CV. Arrows indicate the potential of scan initiation and direction of the scan. (a) ~25 mM of PdSO4, 50 mV/s. (b) ~24 mM of PdSO4, varying scan rates. (c) Proposed mononuclear and binuclear ECE mechanisms. (d) Return scans of CVs (200 mV/s) recorded in four concentrations of PdSO4 depicting the integrated charges, Q1 and Q2, of the back-reduction waves. Reproduced from Ref. X with permission from ACS...... 37 1 III II Figure 2.2. (a) H NMR of the reaction mixture after treating a (black) 4.2 mM Pd 2 and (red) 8.4 mM Pd solution in 20% SO3/H2SO4 with 500 psi of CH4 at 100 ̊C for 20 min. (b) Methane oxidation III reactions of Pd 2 based on the observed stoichiometry for the two products...... 38 + 1 Figure 2.3. NH4 peaks in the H NMR spectra for Evans method magnetic susceptibility measurements. ~5 mM of ammonium sulfate ((NH4)2SO4) was used as a paramagnetic shift reference compound. III II II Blue: Pd 2 (post-electrolysis) solution; Red: Pd (pre-electrolysis) solution; Black: the Pd and III II Pd 2 solutions in co-axial inner (3 mm dia.) and outer (5 mm dia.) tubes. To, Spectra of Pd and III Pd 2 solutions were independently obtained to exclude the possibility that the observed peak shape results from the overlap of two closely-spaced peaks. All three spectra display similar linewidths, indicating a perfect overlap of the peaks in the coaxial double-chamber tube and no III paramagnetic shift by Pd 2...... 38 Figure 2.4. Pd K-edge X-ray absorption spectra of 1-hc and 2-hc: (a) XANES; (b) EXAFS showing the real (solid line) and imaginary (dashed line) components; (c) 1st derivative of the XANES; the lc samples showed essentially identical results (Figure 2.13–Figure 2.16). (d) Pt K-edge EXAFS of III Pt 2 in the solid state...... 41 III Figure 2.5. Raman spectra of (a) fuming H2SO4; (b) Pt 2 in fuming H2SO4, with or without (NH4)2SO4, and aqueous solutions; (c) 1 and 2 in fuming H2SO4 with or without (NH4)2SO4...... 42 III − Figure 2.6. DFT-optimized structures of Pd 2 with six HSO4 ligands with four, two, and zero bridging bisulfates. See 2.4.9 for computational details and other isomers that were calculated. White: H, red: O, yellow: S, light grey: Pt, dark grey: Pd...... 44 Figure 2.7. (a) Background-corrected X-band EPR spectrum of 2 at 60 K. (b) (Black squares) EPR- II measured spin concentrations versus ox.%. Total Pd concentration was 9.3 mM. Cu SO4 dissolved in the same medium was used as a spin quantification standard (see 2.4.7 for details). (Red line) II,III Calculated [Pd2 ] from a least-squares fitting of equation 7 to the EPR-measured spin concentrations...... 45 Figure 2.8. Cyclic voltammograms of (a) 1-hc and (b) 1-lc, for which the PdII concentration was 47 mM and 9 mM, respectively. The similarity in current density despite the higher PdII concentration of 1-hc implies much slower mass transport for the more viscous 1-hc sample. For both samples, the characteristic hysteresis (anodic current larger on the return scan) can be seen, which implies an ECE mechanism (sequential electron transfer-chemical reaction-electron transfer) and formation of the high-valent Pd species.1...... 51

Figure 2.9. UV–Vis spectra of 2-lc diluted in concentrated H2SO4 and recorded over time...... 52

Figure 2.10. UV–Vis spectra of 2-hc and 2-lc diluted in concentrated H2SO4. The spectra are normalized by the absorbance of the 230 nm peak...... 53

II Figure 2.11. (a) Beer’s plot of Pd in 95–98% H2SO4. The y-axis shows background subtracted absorbances measured in a 1 mm-pathlength cuvette at 230 nm. (b) UV–Vis extinction spectra of PdII (black) III and Pd 2 (red). The extinction coefficient is given as per Pd, not per dimer...... 56 Figure 2.12. Unflattened versions of the normalized XAS spectra (solid lines) and normalization II background (dashed lines) of 1-lc, 1-hc, and solid Pd SO4...... 59 Figure 2.13. Comparison of the XANES of hc (red) and lc (black) samples of (a) 1 and (b) 2. With 1, the II spectrum of solid Pd SO4 is also overlaid (light blue)...... 59 Figure 2.14. Comparison of the first derivative of XANES of hc (red) and lc (black) samples of (a) 1 and II (b) 2. With 1, the spectrum of solid Pd SO4 is also overlaid (light blue)...... 60 Figure 2.15. Comparison of the k2-weighted k-space EXAFS of hc (red) and lc (black) samples of (a) 1 and II (b) 2. With 1, the spectrum of solid Pd SO4 is also overlaid (light blue)...... 60 Figure 2.16. Comparison of the R-space (Fourier-transformed) EXAFS of hc (red) and lc (black) samples II of (a) 1 and (b) 2. With 1, the spectrum of solid Pd SO4 is also overlaid (light blue)...... 60 III Figure 2.17. Time-dependent evolution of the Raman spectra of Pt 2 in 1 M H2SO4 after adding 20 mM of III NaCl to 10 mM of Pt 2. With time, peak a remains relatively unchanged, while peak b diminishes and peak c grows in magnitude. Therefore, peak b and c are assigned to vibrational modes in the − III original and the Cl - the original Pt 2 complex, respectively. From the literature, we know that Cl− substitution occurs at the axial position. Along with the low wavenumber of these peaks, we assign peaks b and c to a Pt–Pt vibration...... 62 III Figure 2.18. Perpendicular (red) and parallel (blue)-polarized Raman spectra of Pt 2 (top) and 2-hc (bottom)...... 62 Figure 2.19. Background-uncorrected EPR spectra of (a) commercial fuming sulfuric acid, (b) clean fuming II II sulfuric acid obtained by distillation of SO3, (c) the Cu spin quantitation standard, (d) 1 or Pd , and (e, f) Pd solutions at ox.% = 48% and 95%. The spectrum (f) can be seen larger in Figure 2.7a with background subtraction. All spectra were acquired at 60 K, 0.05024 mW...... 64 II II,III Figure 2.20. Double-integrated EPR signal intensity of Cu and Pd2 at (a, b) various microwave power levels at 60 K and (c) various temperatures at 0.05024 mW (y-axis values for the two metals are normalized to the value at 60 K). The blue lines show the value of microwave power and temperature (0.05024 mW, 60 K) that were selected for the spin quantitation experiments in this study...... 64

Figure 2.21. Estimation of E1. (a) Cyclic voltammograms on a Pt electrode recorded at various scan rates in a dilute (0.6 mM) solution of PdII. The current density is scaled by the square root of scan rate to match the magnitude of the current recorded at different scan rates. (b) Background-subtracted cathodic peak currents plotted as a function of scan rate. The red dotted line is a guide to the eye. (c) CVs in a dilute (0.6 mM) solution of PdII on Pt and FTO electrodes. The current density is scaled by the square root of scan rate and approximate position of the midpoint potential is shown with dotted lines. A low scan rate was adopted for the FTO CV because of the sluggish electron transfer kinetics that give very broad peaks at faster scan rates...... 66 II,III III Figure 2.22. Concentrations of the Pd2 and Pd intermediates calculated using ΔGcomp = 0.15, E1 = 1.69 V, and E2 = 1.49 V...... 67

Figure 2.23. Measurement of E4 for the estimation of E2. The open-circuit potential measurements at each II III ox.% are converted to redox potentials for the Pd /Pd 2 couple using the Nernst equation...... 68 Figure 2.24. Simulated CVs of PdII oxidation. The currents are scaled to match the oxidation peak current. When PdII is oxidized after dimerization, the ratio of the two cathodic peaks does not change. . 68

Figure 2.25. Geometry-optimized structures for the seven isomers that were calculated...... 70 Figure 2.26. Simulated Raman spectra (excitation wavelength = 532 nm) for the seven isomers computed III to model Pd 2...... 71 Figure 3.1. A typical concentration-time plot from real time NMR reaction monitoring of methane III oxidation by Pd 2...... 77 Figure 3.2. The rate of methane oxidation to methyl bisulfate (MBS) and methanesulfonic acid (MSA) at III 50 ̊C measured as a function of the initial concentrations of (a) CH4, (b) Pd 2 and (c) SO3...... 78 Figure 3.3. Arrhenius plots for MBS and MSA formation...... 80

Figure 3.4. Real time concentration-time traces of CH4 oxidation at 50 ̊C recorded (solid symbols) without and (hollow symbols) with O2 co-addition. The solution used in the two experiments contained III 3.3 mM Pd 2 and 8% SO3. An equal volume of CH4 was added to the high-pressure NMR tube for the two experiments. The volume of co-added O2 was equal to that of CH4. The reason why the concentration of CH4 was slightly higher when O2 was co-added is presumably because the NMR tube headspace was not purged before the addition of O2 as the second gas...... 81 Figure 3.5. Real time concentration-time traces obtained with (solid symbols) low and (hollow symbols) II,III II,III II high concentrations of Pd2 . The high [Pd2 ] sample was prepared by adding a Pd solution to III a Pd 2 solution and equilibrating overnight. The slightly higher rate of MBS formation in the high II,III III [Pd2 ] case is due to a slightly higher concentration of the Pd 2 complex in this sample. The exact concentrations of the different Pd species in the solutions and the extracted values of kobs for each experiment are given in Table 3.5...... 81

Figure 3.6. (a) Real time concentration-time traces of CH4 oxidation at 50 ̊C initiated by 10 mM of peroxydisulfate (K2S2O8) in 7.5% SO3, (solid symbols) without and (hollow symbols) with O2 co- addition. The amounts of CH4 and O2 added to the high-pressure NMR tube were the same as in Figure 3.4. (b) (open squares) Experimental and (cross symbols) simulated rates of MSA –1 formation at different concentrations of CH4 and SO3. The rates were simulated with krp1 = 1 M –1 –1 –1 • • s , krp2 = 1500 M s , and [CH3 ] + [CH3SO3 ] = 3.55 μM...... 83 Figure 3.7. The reaction mechanism proposed on the basis of experimental results for methane oxidation III by Pd 2 in fuming sulfuric acid. RLS denotes the rate-limiting step. Abbreviations for each step stand for H-abstraction (ha), radical recombination (rrc), reductive elimination (rel), and radical propagation (rp). Reactants are shown in black, intermediates in green, and byproducts in grey...... 84 III 2 Figure 3.8. Computationally modeled Pd 2(HSO4)6 isomers and their computed free energies. (a) cis-κ - μ2, (b) trans-κ2-μ2, (c) paddlewheel, (d) unbridged. The unbridged complex showed spontaneous Pd–Pd bond cleavage when modeled as a triplet. The free energies are values relative to the most stable isomer (conformer #3, trans-κ2-μ2)...... 86 Figure 3.9. Ligand dissociation pathways with reactant and products at different charge states...... 87 Figure 3.10. A stepwise mechanism that features homolytic ligand dissociation and rate-limiting H abstraction by the bisulfate radical. Abbreviations for each step stand for ligand dissociation (ld), H-abstraction (ha), radical recombination (rrc), reductive elimination (rel), and radical propagation (rp). Reactants are shown in black, intermediates in green, and byproducts in grey...... 88 Figure 3.11. BLYP/SDD,6-311++G(d,p)/SMD-water//BLYP/SDD,6-31+G(d) computed reaction pathways • II,III for H atom abstraction from CH4 and recombination of CH3 with Pd2 . Numbers indicate free energies in kcal/mol for the most stable conformer. Calculations were done at 323 K to match the experimental conditions...... 91

Figure 3.12. Ionic/organometallic C–H activation pathways (i.e., concerted C–H cleavage/Pd–C bond formation) for which transition states were calculated...... 93 Figure 3.13. Optimized structures for the transition state for the pathway shown in Figure 3.12a. ∆G‡ = 35.2 kcal/mol (left), ∆G‡ = 40.3 kcal/mol (right)...... 93 Figure 3.14. Ratios of integration area measured using different delay times for determining proper relaxation delay for accurate quantitation of methane. AQ: acquisition time, D1: relaxation delay. Number of scans = 4 and 8 for 1H and 2H, respectively. Based on these data, AQ + D1 = 40 s was chosen for actual measurements for both 1H and 2H...... 101 Figure 3.15. Representative 1H and 2H NMR spectra from methane oxidation experiments, shown along with baselines and integration regions...... 102 Figure 3.16. Temperature of the NMR tube after insertion assessed with ethylene glycol...... 104 Figure 3.17. (Solid, faint symbols) Raw experimental concentration-time traces were (hollow symbols) smoothed by normalizing the total methyl concentration...... 105 Figure 3.18. (Lines) COPASI simulation of the methane oxidation reactions fitted to (symbols) experimental concentration-time traces...... 108 Figure 3.19. Eyring plot and activation parameters derived from rate measurements at 40, 50, 60 and 72 ̊C. Each data point corresponds to an average of three or more measurements...... 109 1 Figure 3.20. H NMR spectra of quenched NMR tube reactions with (red) CD4 and (green) CH4. Reaction 3 conditions: 50 ̊C, 435 seconds (CH4) or 1344 seconds (CD4). Concentrations: 1.1 mM d -MBS, 3 3.1 mM d -MSA, 12.0 mM CD4; 1.1 mM MBS, 4.3 mM MSA, 9.7 mM CH4...... 111 Figure 3.21. The rate of peroxydisulfate-initiated methane oxidation to methanesulfonic acid at 50 ̊C measured as a function of the initial concentrations of (a) CH4, (b) K2S2O8, and (c) SO3...... 114 Figure 3.22. Simulation of the expression for [MSA] versus time from peroxydisulfate-initiated methane sulfonation...... 116

Figure 3.23. Optimized structures for case B-1; ∆GInt2 = 15.0 kcal/mol (left), ∆GP = 7.4 kcal/mol (right)...... 119

Figure 3.24. Optimized structures for case B-3; ∆GInt2 = 16.9 kcal/mol (left), methyl-equatorial: ∆GP = 1.6 kcal/mol (middle), methyl-axial: ∆GP = –2.7 kcal/mol (right)...... 120

Figure 4.1. (a) Cyclic voltammograms obtained on a Pt disk electrode at room temperature in 0.5 M H2SO4; II II (black) background, (blue) 1 mM K2Pt Cl4, and (red) 1 mM K2Pt Cl4 with 10 mM NaCl. (b) Cyclic voltammograms obtained on a Pt wire electrode in 10 mM NaCl, 0.5 M H2SO4; (black) II II background and (blue) 10 mM K2Pt Cl4 at room temperature, and (red) 10 mM K2Pt Cl4 at 130 ˚C. (c) Tafel plot at 130 ˚C for PtII electro-oxidation. The solution contained 5 mM each of II IV K2Pt Cl4 and Na2Pt Cl6 in 10 mM NaCl, 0.5 M H2SO4. Eeq (= 0.829 V vs SHE) was obtained from the open-circuit potential. (d) Cyclic voltammograms obtained on a Pt wire electrode in 10 mM NaCl, 0.5 M H2SO4 at 130 ˚C; (black) background, (blue) 30 mM CH3OH without the 10 mM –1 NaCl, and (red) 30 mM CH3OH. All scan rates = 100 mV s ...... 126 Figure 4.2. High-pressure, three-electrode, two-compartment electrochemical cell for EMOR. WE: Pt foil working electrode, RE: Ag/AgCl reference electrode, CE: Pt mesh counter electrode. 1: Glass cell, 2: working solution containing the Pt ions, 3: fritted tubes for housing the RE, 4: PTFE stir bar, 5: H+-conducting membrane separating the counter compartment, 6: PTFE body holding the IV membrane stack, 7: counter compartment solution containing (V O)(SO4) as the sacrificial electron acceptor...... 129

Figure 4.3. Representative electrochemical data recorded during an EMOR trial (the 10.5 h-long trial in Table 4.1). The open-circuit potential (EOCP) reading at approx. 1 h time intervals (bottom, black triangles) were used to calculate the PtII% in the solution (top, black squares). This was in turn used to determine how much current to pass (top, red line), and the electrode potential during the electrolysis (ECP) was recorded (bottom, blue line)...... 130 Figure 4.4. (a) Amounts of methane oxidation products from EMOR versus reaction time. Each point represents a different trial in Table 4.1, and the product concentrations were normalized by iave of each trial (see 4.4.6.4 for explanation). The lines represent fitting with the (b) set of suggested reactions...... 133

Figure 4.5. Cyclic voltammogram (CV) of Pt disk electrode in 0.5 M H2SO4. The H UPD region and oxide region are marked according to conventional understanding.24 The blue shading represents the area integrated for electrochemically active surface area determination. Scan rate = 100 mV s–1. ... 136 Figure 4.6. Assessments performed to ensure proper operation of the high-temperature electrochemical cell. (a) Current-time trace from polarization of the electrode at 1.06 V vs SHE at room temperature. Stirring was turned on ~5 sec after the start of electrolysis and the stir rate was increased slowly to 200 rpm. (b) OCP registered at the electrode during heating. The initial rapid decrease in the OCP is because of relaxation from the previous polarization during the stir bar fidelity check shown in (a)...... 136

Figure 4.7. CVs of the Pt foil working electrode in 0.5 M H2SO4 (black) before and (red) after reactor operation for electrochemical methane oxidation reaction (EMOR). Scan rates = 100 mV s–1. 138

Figure 4.8. CVs of a Pt wire electrode in 10 mM NaCl, 0.5 M H2SO4 at (black) room temperature and (red) 130 ˚C. The potentials at both temperatures were converted to the SHE scale by adding 0.307 V to the recorded value. Scan rate = 100 mV s−1...... 139 Figure 4.9. (a) The polybenzimidazole (PBI) membrane appearance changes from the left to right after II IV heating in the presence of 3 mM K2Pt Cl4 and 7 mM Na2Pt Cl6 in 10 mM NaCl, 0.5 M H2SO4 for 19 h at 130 ˚C. (b) The five PBI layers (see Figure 4.10) after EMOR reactor operation. Blackened areas show oxidative degradation and Pt0 deposition. The periphery, where PTFE gaskets were placed, is clear because it was not exposed to the solution. While the 1st layer (closest to the working solution) showed significant blackening and degradation, the 2nd layer showed drastically reduced blackening, and the last layer showed almost no sign of degradation. Incidentally, the color of the membranes is darker than the pristine membrane shown in (a) because of the activation pretreatment...... 140 Figure 4.10. Counter compartment design for the EMOR reactor...... 140 Figure 4.11. Calibration curves for Pt ion quantitation by UV–vis. (a) Absorbance at 404 nm of (black) K2PtCl4 and (red) Na2PtCl6 diluted in 1 M SnCl2, 3 M HCl. (b, c) Absorbance at 262 nm of K2PtCl4 and Na2PtCl6 diluted in 1 M HCl following irradiation with a UV lamp...... 141 Figure 4.12. Representative baseline-corrected (Whittaker smoother) spectrum of the working solution from an EMOR trial (entry 3 in Table 4.1). The “wet” pulse sequence was employed for solvent suppression. The spectrum is referenced to the acetic acid peak at 2.0 ppm...... 143

Figure 4.13. Calibration curves for quantitation of (a) CO2 and (b) CH3Cl by gas chromatography. See the text for details...... 144 Figure 4.14. Investigation of PtII oxidation on a glassy carbon (GC) electrode using a solution of 2 mM II II K2Pt Cl4 in 10 mM of NaCl, 0.5 M H2SO4. (a) CVs with and without Pt . (b) Chronoamperometric (CA) traces at different applied potentials. (c) Series of CVs obtained on the same GC electrode. The 4th cycle was recorded after 220 min of chronoamperometry at 1.14 V that deactivated the electrode. (d) Overlay of PtII oxidation CVs on GC and Pt electrodes. [PtII] = 2 mM for GC and 1 14

mM for Pt. The currents were normalized to geometric surface areas. Scan rates = 100 mV s–1...... 145 II Figure 4.15. Bulk electrolysis of a solution of 2 mM K2Pt Cl4 and 10 mM NaCl in 0.1 M H2SO4 on a graphite felt electrode. The solution was sampled periodically to measure the PtII and PtIV concentrations by UV–Vis spectroscopy...... 146 Figure 4.16. (Red) A portion of the current trace during EMOR reactor operation using a graphite felt working electrode. The solution contained 0.4 mM of PtII, 1.4 mM of PtIV and 10 mM NaCl in 0.5 II M H2SO4 and was 12.3 mL in volume. Concentration of Pt was kept to a minimum for this preliminary trial because of the low efficiency of the electrode for oxidizing PtII. (Blue, dotted) Cumulative charge passed during the experiment. Because a lot of negative charge flowed during the cathodic polarization for electrode regeneration, it was difficult to pass a net positive charge over time; i.e., Q rises and falls over the course of each potential step, but shows no long-term increase. The post-reaction PtII concentration was indicative of little PtII oxidation over the course of the electrolysis. For reference, the complete oxidation of the PtII ions in the solution would require 474 mC of charge passed...... 147 II II Figure 4.17. Investigation of Pt oxidation on an FTO electrode. The solution contained 1 mM K2Pt Cl4 II and 100 mM HCl in 0.5 M H2SO4 . The CVs of Pt on FTO were acquired in the order from red II to pink to brown. (Dotted blue) CV of Pt in 0.5 M H2SO4 obtained on Pt electrode is overlaid for comparison. FTO CVs were normalized by the geometric surface area, and the Pt CV was normalized by the surface area measured by H UPD. Scan rates = 100 mV s–1...... 148

Figure 4.18. Series of superimposed cyclic voltammograms obtained on a Pt electrode in 0.1 M H2SO4 with successive additions of Cl– ion from 10–7 to 10–5 M. Arrows show directions of change of curves with increasing [Cl–]. Dashed curve corresponds to [Cl–] >10–4.5 M. Potentials are vs the reversible – hydrogen electrode (RHE), which is –0.059 V vs SHE in 0.1 M H2SO4 (pKa of HSO4 = 1.99). –1 26 Scan rate = 60 mV s ; VA = 1.375 V; T = 298 K. Reproduced from ref. with permission from The Royal Society of Chemistry...... 149 Figure 4.19. Variation of the quantities of adsorbed Cl– with potential. The solution consisted of 1 × 10–3 N NaCl and 1 × 10–3 N HCl. Potentials are vs NHE. Replotted from ref. 27 with permission from the editorial staff of Russian Chemical Reviews...... 149

Figure 4.20. (a) Cyclic voltammograms obtained on a Pt disk electrode in N2-purged solutions containing II 0, 1, and 10 mM K2Pt Cl4 in 10 mM NaCl, 0.5 M H2SO4 electrolyte at RT. Scan rates = 100 mV s–1. The CV of 10 mM PtII is plotted at 5-fold reduced current density to match the vertical scale for easier comparison. (b) Chronoamperometric traces obtained on a Pt wire electrode at 130 ˚C II in a stirred solution of 10 mM of K2Pt Cl4 in 10 mM NaCl, 0.5 M H2SO4 electrolyte. The current densities are normalized by the electrochemically active surface area determined by integration of the H UPD wave on Pt...... 150 Figure 4.21. Stepped-potential chronoamperometry on a Pt wire electrode at 130 °C. This raw data was used to construct the Tafel plot for PtII oxidation (Figure 4.1c). The stirred solution contained 5 II IV mM of Pt and 5 mM of Pt in 10 mM NaCl, 0.5 M H2SO4...... 151 Figure 4.22. Measurement of (a, c) methane functionalization and (b, d) methanol oxidation activities under – II different concentrations of (a, b) H2SO4 and (c, d) Cl . Catalyst and oxidant loadings were [Pt ] = 3 mM and [PtIV] = 7 mM. Solutions were heated to 130 ˚C for 1.5 h for (a, c) methane functionalization and 3 h for (b, d) methanol oxidation tests. Error bars correspond to standard errors from ≥3 independent measurements...... 153 0 II IV − II (m−2)− Figure 4.23. (a) Experimentally determined Pt -Pt -Pt -Cl equilibria for the reaction, 2 [Pt Clm] ⇄ 0 IV (n−4)− − 34 Pt + [Pt Cln] + (2m–n) Cl . Reproduced from ref. with permission from Elsevier. The

equilibrium constant for this reaction, K, corresponds to the y-intercept of the given plot according to the equation Y = (2m–n) X – log K, where X and Y denote the x- and y- axis values, respectively. The reason why the slope of the data depends on [Cl−] is evident from this equation; at high [Cl−], m~4 and n~6 so that (2m−n)~2, but at low [Cl−], both m and n decrease and (2m−n) deviates from 2. (b) Overlay of a few selected solution compositions on the experimental equilibrium curve. Blue: [PtII] = 3 mM, [PtIV] = 7 mM, [Cl−] = 10 mM. Green: [PtII] = 15 mM, [PtIV] = 35 mM, [Cl−] = 50 mM. Red: [PtII] = 50 mM, [PtIV] = 500 mM, [Cl−] = 200 mM...... 156 Figure 4.24. The glass cell after a 29 h reactor operation. Arrows point to adventitious Pt0 deposits. .... 157 Figure 4.25. The total amount of methane oxidation products from the four EMOR trials (Table 4.1) plotted against (left) the amount of charge passed and (right) the reaction time. The product moles were calculated in a way that counts the total number of oxidation events required to generate each product (μmolTotalProduct = μmolCH3OH + μmolCH3Cl + 2*μmolCH2(OH)2 + 3*μmolHCOOH + 4*μmolCO2). When the total product amount is plotted against the total charge passed (left) in each run, a linear correlation is observed. When the total product is plotted against reaction time (right, hollow black squares), the trend line exhibits slight deviations from a straight line because each EMOR trial had a slight variation in the average current and total charge passed. To account for this variation, we divided the product sum for each trial by the average current of each trial (iave, italicized numbers in the plot) recovering a linear plot (right, hollow red circles; also Figure 4.4a)...... 159

Table of Schemes

Scheme 1.1. Categories of known organometallic C–H activations.15 ...... 23 Scheme 1.2. Catalytic cycles for methane-to-methanol functionalization using homogeneous catalysts. M denotes a metal ion and Y denotes –OH or a hydrolysable functional group, e.g., –OSO3H or – O2CCF3. The top cycle represents electrophilic C–H activation by high-valent metal complexes. The bottom cycle represents reversible C–H activation followed by oxidation of the transient metal-methyl intermediate...... 27 II III III II,III Scheme 2.1. Summary of the reactions between Pd , Pd 2, Pd and Pd2 ...... 46 Scheme 2.2. A qualitative orbital energy diagram showing all Pd’s in the II oxidation state. Only the MOs of Pd dz2 parenthood are shown for Pd–Pd and L–Pd–Pd–L...... 48 Scheme 4.1. The catalytic cycle for the functionalization of methane by aqueous Pt salts (Shilov’s catalyst) and different strategies to overcome the stoichiometric use of PtIV...... 124

Table of Tables

Table 2.1. Summary of XAS results...... 41

Table 2.2. Products from the reaction of 2-hc with methane. 500 psi CH4, 100 ˚C, 40 min...... 53

II III –1 –1 Table 2.3. Molar extinction coefficients, per Pd atom, for Pd and Pd 2 in M cm ...... 56

1 (x–2) Table 3.1. Reaction free energies for the dissociation of an axial κ -HxSO4 ligand. Reactants and products have the trans-κ2-μ2 geometry for the bridging ligands in all three cases. Neutral: III III + III – [Pd 2(HSO4)6]; Cation: [Pd 2(H2SO4)(HSO4)5] ; Anion: [Pd 2(HSO4)5(SO4)] ...... 87 Table 3.2. Reaction free energies (kcal/mol) of the four possible C–H cleavage reactions following III homolytic ligand dissociation from Pd 2...... 88 III Table 3.3. Pd 2 samples for reaction order studies...... 97 + Table 3.4. Validation of using NH4 as an internal NMR integration standard...... 102 Table 3.5. Observed rate constants for MBS and MSA generation from solutions containing different II,III concentrations of Pd2 . The rate constants were obtained from three independent reactions. 107

Table 3.6. TOF at 140 ̊C calculated by extrapolation from kMBS at 50 ̊C...... 110 Table 3.7. Free energy for each configurational isomer relative to the isomer with minimum free energy for the optimized species in the reaction pathways A and B. The unbridged isomer of some species was not obtained since the Pd-Pd bond falls apart during optimization. Free energies are reported in kcal/mol...... 117 Table 3.8. Absolute values of the free energy difference between two possible spin states for each species in the reaction pathways A and B. The unbridged isomer of some species was not obtained since the Pd-Pd bond falls apart during optimization. Also, the triplet state of cis-B-TS was not obtained. Free energies are reported in kcal/mol...... 117 Table 3.9. Free energy for each protonation tautomer relative to the tautomer with minimum free energy 1 2 2 III ([Pd2(κ -HSO4)2(κ -HSO4)2(μ -HSO4)2]) for the trans-Pd 2 complex. Relative energies are reported in kcal/mol...... 117 Table 3.10. Relative free energies for optimized isomers for the ligand dissociation pathways with reactant and products at different charge states. Energies are reported in kcal/mol...... 118 II IV Table 4.1. Results of EMOR trials at T=130 ℃ and PCH4= 675 psi. Initial [Pt ] and [Pt ] in the working solution were 3 mM and 7 mM, respectively, and the solution volume was 23 mL. The electrochemically active surface area of the Pt working electrode was 10.3 cm2...... 131 Table 4.2. Estimated Faradaic efficiencies of different EMOR trials...... 143 Table 4.3. Results of bulk electrolysis of PtII to PtIV at 130 ˚C with stirring at 200 rpm. The solution initially II IV contained 5 mM of K2Pt Cl4, 5 mM Na2Pt Cl6, and 10 mM NaCl in 0.5 M H2SO4 (initial amount of PtII = 110–115 μmol)...... 152 Table 4.4. Results of heating PtII/IV solutions in sealed ampules. Solutions contained combinations of NaCl, II IV K2Pt Cl4, and Na2Pt Cl6 in 0.5 M H2SO4. T = 130 ˚C. Ampules for entries 3 and 4 also contained a few mg of Pt0 particles...... 155 Table 4.5. Amount of Pt0 deposition from reactor operations of varying time duration...... 156

Table 4.6. EMOR reactor results from two trials where the run duration was identical (10.5 h) but the concentrations of PtII, PtIV and Cl– differed by a factor of 5...... 160 Table 4.7. Apparent rate constants from fitting experimental data with the mechanism in Figure 4.4b. 161

Table 4.8. Experimentally determined relative rates of C–H oxidation of CH4 and CH3OH in the literature and this work...... 162 Table 4.9. Evaluation of PtII-catalyzed C–H oxidation of various substrates at 130 ˚C. The test solutions II IV contained 3 mM Pt and 7 mM Pt in 10 mM NaCl, 0.5 M H2SO4...... 163 II IV Table 4.10. Concentrations of CH4 and CH3OH before and after reaction with 3 mM Pt and 7 mM Pt in 10 mM NaCl, 0.5 M H2SO4 at 130 ˚C...... 163

List of Abbreviations

CV cyclic voltammetry, or cyclic voltammogram DFT density functional theory

Ea Arrhenius activation energy ECE sequential electrochemical-chemical-electrochemical reactions

Em midpoint potential (average of cathodic and anodic peak potentials in a CV) EMOR electrochemical methane oxidation reaction EPR electron paramagnetic resonance ET electron transfer EXAFS extended X-ray absorption fine structure FE faradaic efficiency FTO fluorine-doped tin oxide ICP-MS inductively coupled plasma mass spectrometry KIE kinetic isotope effect

MBS methyl bisulfate (CH3OSO3H)

MSA methanesulfonic acid (CH3SO3H) NHE normal hydrogen electrode NMR nuclear magnetic resonance OCP open-circuit potential PCET proton-coupled electron transfer PTFE polytetrafluoroethylene (“Teflon”) SHE standard hydrogen electrode

SSE saturated silver/silver sulfate electrode (Ag2SO4/Ag in 95–98% H2SO4) TOF turnover frequency TON turnover number XANES X-ray absorption near edge structure XAS X-ray absorption spectroscopy

1. Introduction

1.1. Mild and Selective Oxidation of Methane to Methanol

Methane, the main component of natural gas, is a valuable carbon resource that is abundant yet underutilized. Because of its low boiling point, compression, storage, and transportation of methane require expensive infrastructures that operate with an economy of scale. Thus, technologies for converting methane to value-added liquid chemicals such as methanol would enable more efficient utilization of this non- renewable resource. Current methane valorization technologies rely on the heterogeneously catalyzed steam reforming process that produces H2 and CO through an endothermic, energy-intensive reaction. For end products such as ethylene or methanol, such an indirect route is not the most ideal. Moreover, the high temperatures and pressures under which steam-methane reforming operates demand large, capital-intensive facilities for economic viability, similarly to the physical handling of methane. The need for heavy infrastructures of current technologies has limited their deployment at remote and stranded methane sources.1 Consequently, spontaneously released natural gas at oil wells is flared at massive scales,

2,3 contributing to atmospheric CO2 as well as wasting the valuable carbon resource.

The development of mild temperature, direct methane functionalization processes that can operate portably is expected to stem flaring as well as expand the versatility of methane as a chemical feedstock.4,5 Particularly attractive is the direct and thermodynamically favorable oxidative mono-functionalization of methane, such as that shown in equation 1.1:

CH + 0.5 O → CHOH … 1.1 The enzyme methane monooxygenase catalyzes this reaction, suggesting that in principle, it is possible to carry out this reaction under mild conditions. The use of O2 as the terminal oxidant is important, as costly chemical oxidants are impractical to use in the generation of a bulk commodity chemical such as methanol. However, the greater propensity of methanol to be further oxidized poses a great selectivity challenge in catalyst design. The electron-rich hydroxyl group in methanol polarizes its C–H bonds for electrophilic attack and/or serves as a binding site for catalysts. The bond dissociation free energy (BDFE) of the C–H bond is also weaker for methanol than for methane by ~10 kcal/mol. The selectivity challenge is only aggravated by the extreme chemical inertness of the nonpolar, symmetric methane molecule, as high

21 temperatures and aggressive reagents are usually required to activate methane. Due to this dual challenge of activity and selectivity, a suitable catalyst system for the highly appealing direct methane-to-methanol reaction at mild temperatures has yet to be achieved and developed.

Metal ions play a central role in catalyzing difficult reactions, and good understanding and exploitation of their properties is critical to achieving selective methane functionalization catalysis. For example, enzymes that selectively oxidize methane to methanol feature Fe and Cu ions in the active site. Although a subject of ongoing studies, particularly for the Cu-containing enzyme, it is widely accepted that the diiron active site in soluble methane monooxygenase cleaves the methane C–H bond by H atom abstraction from an O atom bound to high-valent Fe.6 The methyl radical then rapidly rebounds to the hydroxyl group to form methanol. Mimicking these enzymatic active sites, Fe- and Cu-containing zeolites and MOFs demonstrated selective oxidation of methane to methanol in a stoichiometric manner.7,8 Some DFT studies of the reaction mechanism imply H atom abstraction and methyl radical rebound similar to that of the Fe enzyme. Importantly, such a radical-based mechanism is unfavorable to selective methane oxidation in the solution phase where substrates and reactive intermediates may freely diffuse and react indiscriminately.9 The enzyme uses a complex gating mechanism to orchestrate mutually incompatible substrate activation events and bar methanol from accessing the active site. For the heterogeneous materials, an energy-intensive chemical looping procedure separates the methane activation step and catalyst regeneration step; switching to continuous, catalytic methane oxidation with O2, albeit an impressive feat, resulted in reduced selectivity.10 Generally speaking, H atom abstraction is the key step in most oxidations by high-valent metal oxos, which may be difficult to carry out selectively in the solution phase.11 Alternatively, instead of cleaving the C–H bond indirectly by abstracting an H atom with the ligand, some metal centers directly interact with methane to form metal-carbon bonded organometallic intermediates and avoid radical routes. Chloro-aquo complexes of PtII in dilute aqueous acids, also known as Shilov’s catalyst, first had this reactivity observed with methane at a mild temperature of ~100 ̊C. Moreover, the catalytic oxidation of methane to methanol was demonstrated.12 The ~1:1 selectivity for methane oxidation over methanol oxidation, while modest in an absolute sense, is greater than selectivities observed from H atom abstraction, which usually follows the order of C–H bond strength. Referred to as organometallic C–H activation,13 the direct interaction of metal ions with inert sp3 C–H bonds has since been demonstrated and studied in many more instances.14,15

1.2. Organometallic C–H Activation for Methane Functionalization

1.2.1. Categories of Organometallic C–H Activation

Based on the nature of the active agent that effects C–H cleavage, organometallic methane activation can be broadly classified into five categories, following Bercaw and Labinger (Scheme 1.1).15

Scheme 1.1. Categories of known organometallic C–H activations.15

First, the C–H bond of methane can undergo oxidative addition to electron-rich, low-valent metal ions to yield methyl hydride complexes. Typical examples include cyclopentadienyl complexes of late transition metals such as Ir and Rh. The reaction is initiated by coordination of CH4 to the metal, whose open coordination site is often obtained in situ from a saturated precursor by thermal/photochemical liberation of ligands. Though a subject of debate, methane activation by Shilov’s PtII catalyst is also thought to go through an oxidative addition pathway that is concomitant with, or rapidly followed by, deprotonation.16 The PtIV methyl hydride intermediate has been observed in studies of model complexes in organic solvents.17 The intimate electronic interaction between the metal and the substrate generally confers good selectivity for methane to complexes in this category.18 A mechanistically different but related example

II III III is the Rh porphyrin dimer, which splits into two upon activating methane into Rh -CH3 and Rh -H. The stringent steric requirements make this metalloradical activation highly selective for methane. Unfortunately, except for Shilov’s catalyst, which arguably belongs to the oxidative addition category, these electron-rich metal complexes are usually incompatible with oxidizing and/or protic environments that are usually required for oxidative functionalization of methane to methanol.

Sigma-bond metathesis and 1,2-addition involve the addition of the methane C–H bond across an M–L bond, a metal-ligand single bond for the former and double bond for the latter. The metal is typically a lanthanide or early transition metal of zero d electron count with an alkyl group to be liberated. Sigma- bond metathesis, in effect, exchanges this alkyl group with the methyl group of methane. Similarly, the reactive metal-nonmetal double bond to which methane adds across in 1,2-addition is usually generated from the loss of an alkane from a metal alkyl precursor. Therefore, while interesting in themselves, these complexes would be largely irrelevant for methane-to-methanol catalysis, because the active metal alkyl complex cannot be catalytically regenerated from CH4 and O2. Parenthetically, sometimes the term “sigma- bond metathesis” is used more broadly to denote generic bond metatheses between M–L and C–H.

The last category of organometallic methane activation is electrophilic activation, which is a net replacement of one of the methane protons with an electrophilic, high-valent metal ion and one of the metal

– ligands with CH3 . The resultant high-valent methyl complex usually undergoes rapid reductive elimination to generate the functionalized product, and oxidation of the reduced metal ion to the high-valent state completes the catalytic cycle. Facile functionalization from the methyl intermediate and compatibility with oxidizing conditions make this type of C–H activation generally amenable to methane-to-methanol catalysis. However, electrophilic metal ions would also be more reactive towards methanol than methane, as the former is more electron-rich and can bind to the metal via the hydroxyl group. Additionally, electrophilic metal centers would not easily lend an open coordination site to methane in the presence of even mild nucleophiles such as water.

Critical to overcoming these problems was the use of strongly acidic media such as concentrated/fuming sulfuric acid and trifluoroacetic acid. Being labile, poor nucleophiles themselves and aggressively protonating any nucleophiles in the medium, strong acid solvents minimize catalyst poisoning and enhance the electrophilicity of the active high-valent metal species.19 Moreover, the strong acid derivatizes methanol to the methyl ester of its conjugate base, which exerts a strong electron-withdrawing effect at the methyl group and drastically reduces its reactivity towards the electrophilic catalyst.20 This strategy, which —instead of changing the catalyst itself— exploits the inherent selectivity of electrophilic metal ions for electron-rich substrates by reversing the polarity of the product, was so effective in enhancing the selectivity of the reaction that it has been referred to as “product protection.” Subsequent hydrolysis of the methyl ester yields methanol as the final product. Using HgII, Periana first demonstrated this strategy to show the catalytic conversion of methane to methyl bisulfate at 85% selectivity at 50% conversion in weakly fuming sulfuric acid, at a relatively mild temperature of 180 ̊C.21 Such reactivity was also found in other metal ions, particularly in heavy late-transition or post-transition metal ions that can make strong M– C bonds by virtue of their polarizability (“softness”) and favorable energetic alignment of their frontier orbitals with those of carbon.20

For these electrophilic C–H activation reactions in strong acids, selective functionalization and predictable trends in reactivity have been observed. In a seminal study, ethane oxidation by TlIII in trifluoroacetic acid was shown to be highly selective to monohydroxylation of the –CH3 group as opposed

22 to C–C bond cleavage or overoxidation to CO2. Several pieces of evidence, e.g., insensitivity to O2

III addition and facile reductive elimination from Tl -CH3, strongly support a non-radical mechanism. The study also demonstrated, among isoelectronic d10 metal ions HgII, TlIII, and PbIV, a straightforward correlation between increasing electrophilicity and the rate of ethane oxidation. On the other hand, when various metal ions were reacted with methane and ethane in fuming sulfuric acid, two products attributed

24 to non-radical and radical pathways were observed. The product ratio depended on the identity of the metal, and metals such as Pd and Pt mostly gave the non-radical product.23,24 Methane oxidation by homogeneous metal ions has been extensively reviewed with a partly mechanism-based classification.25 While rigorous studies of the reaction mechanism are not always available, organometallic, electrophilic methane activation is implicated in many cases.

1.2.2. Challenges Involving the Oxidation Step

While C–H activation at the metal ion is critical to the rate and selectivity of methane functionalization, an important and related consideration is the oxidation step that is required for methane- to-methanol catalysis. As shown above, incompatibility of the oxidation step with electron-rich metals have precluded their use for methane-to-methanol catalysis in spite of their selective and facile C–H activation reactivity, although recently catalytic methanol generation was demonstrated with RhII porphyrin dimers by carefully controlling oxidant delivery using a nanostructured electrode (see below).26 Even for catalysts that are compatible with oxidizing conditions, economic considerations constrain the stoichiometric oxidant to

25 O2 or O2-regenerable oxidants, which adds additional challenges to catalyst design. For example, one of the greatest long-standing drawbacks of Shilov’s PtII catalyst was the need for PtIV salts as the stoichiometric

27 IV oxidant. Efforts to replace Pt with O2 or O2-regenerable oxidants have met only with partial successes that would not support stable and sustained catalysis; particularly, irreversible Pt0 metal precipitation seemed perpetually pernicious.27–29 Periana’s PtII catalyst in fuming sulfuric acid overcame both the requirement for PtIV stoichiometric oxidant and Pt0 precipitation by using sulfuric acid as the stoichiometric oxidant and ligating the Pt ion with the Brønsted-basic bipyrimidine ligand that was stable in the caustic medium so that reduced Pt species will not aggregate.30 Sulfuric acid was deemed air-regenerable by the reaction SO2 + O2 → SO3. However, oxidation was found to be slower than C–H activation, limiting the overall rate of catalysis. More importantly, the high heat of hydration and high boiling point of sulfuric acid make it an unfavorable medium for product separation; specifically, the methyl ester has to be hydrolyzed to methanol by the addition of water, and re-concentration of the diluted sulfuric acid is energetically highly demanding.31 Still, its air-regenerable oxidizing property makes it the medium of choice for demonstrating catalytic methane oxidation, even when similar C–H activation rates and selectivities may be observed in other strong acid media that are more amenable to product separation.21 Some researchers have tried to

32 employ O2 as the oxidant in the same pot by adding redox mediators such as polyoxometalates or

33 benzoquinone in combination with NO2, but these conditions yielded overoxidized products or CO2 from

34 solvent decomposition. In conclusion, stoichiometric oxidation using O2 as the terminal oxidant presents an additional challenge in the design of homogeneous methane-to-methanol catalysts.

1.3. Electrochemical Methane Functionalization Approaches

1.3.1. Potential Advantages of Methane Functionalization Catalysis by Electrochemical Oxidation

Achieving redox transformation with electricity instead of chemical reagents is attractive in many ways.35 First, the application of high driving forces allows the generation of reactive species at low temperatures, which offers advantages for selectivity compared to thermal reactions. The driving force is also tunable in time and space, conferring a degree of control to the chemist that is not possible in non- electrochemical systems. From a practical perspective, replacement of stoichiometric reducing and/or oxidizing equivalents with electrical charge contributes to atom economy, and sometimes to step economy as well.36 Appropriate pairing of oxidation and reduction half reactions double the benefit; for example, methane oxidation reaction may be paired with proton reduction or O2 reduction to either generate H2 as a useful side product or reduce the overall cell potential. With these advantages and the rising availability of cheap, renewable electricity, researchers are increasingly turning to electrochemical methods.37

However, among the vast literature for selective methane oxidation, reports of electrocatalysis are sparse.38 Electrochemical oxidation of methane has been traditionally explored in the context of fuel cells

39 for energy generation by full oxidation to CO2. While certain metal oxide anodes decorated with catalytically active metals show promising results for partial oxidation of methane,40,41 sometimes in combination with light irradiation,42 the scaling relationship between the free energy of surface-bound intermediates and rates of competing reactions makes selective methane oxidation challenging on solid electrodes.43 Importantly, very limited studies have been done with homogeneous catalysts that activate methane via the organometallic pathway described above. Attempts to utilize Shilov’s catalyst suffered from eventual catalyst loss as Pt0 and poor performance with methane as a substrate,44,45 while immobilization of PtII-bipyridine complexes on porous carbon electrodes resulted in full oxidation of

46 methane to CO2. As for electrophilic catalysts in strong acid media, presumably due to the caustic nature of the solvent, no examples could be found before our group’s pioneering work.47 Very recently, an elegant study came out that employs the unique reactivity of the RhII-porphyrin dimer. The low-valent catalyst’s

II incompatibility with O2 was overcome by generating the active Rh 2 species by electrochemical reduction

III II of the Rh precursor at the electrode, which also reduced O2 before it could oxidize the Rh 2 catalyst. The narrow nanowire array geometry of the electrode played a key role in maintaining a low O2 concentration at the site of catalyst generation, and a turnover number (TON) up to 52,000 over 24 h was achieved, based

26 on active catalyst concentration. Although this study derives the oxidizing equivalents from O2 and is therefore, strictly speaking, not an example of electrocatalytic methane oxidation, it exemplifies the

26 potential of applying electrochemistry to homogeneous catalysts with a solid understanding of their reaction mechanism.

1.3.2. Mechanism-based Adaptation of Electrochemical Oxidation for Organometallic Methane Functionalization Catalysis

Motivated by this state of the matter, we explored selective methane oxidation electrocatalysis using homogeneous metal ion catalysts. To be general in our approach, systems that are compatible with oxidizing conditions were studied. Existing understanding of their reaction mechanism, depicted in Scheme 1.2, led to two different approaches.

Scheme 1.2. Catalytic cycles for methane-to-methanol functionalization using homogeneous catalysts. M denotes a metal ion and Y denotes –OH or a hydrolysable functional group, e.g., –OSO3H or –O2CCF3. The top cycle represents electrophilic C–H activation by high-valent metal complexes. The bottom cycle represents reversible C–H activation followed by oxidation of the transient metal-methyl intermediate.

1.3.2.1. Approach 1: Electrochemical generation of high-valent metal ions For methane functionalization by the top cycle in Scheme 1.2, an oxidant must regenerate the active, high-valent state of the catalyst. Additionally, provided that solvent coordination is not strong, more electrophilic metal ions are expected to show a higher reaction rate, as demonstrated by the aforementioned study of HgII, TlIII, and PbIV in trifluoroacetic acid.22 Metal ion electrophilicity is correlated with the

(n+2)/n thermodynamic redox potential for M , and the requirement of O2 being the terminal oxidant puts an upper limit to the redox potential of M(n+2)/n accessible by chemical oxidation. The input of electrical energy overcomes this limitation and allows the generation of more electrophilic metal ions that may show higher methane activation rates. The lack of precedents for electrochemical oxidation for methane activation by homogeneous electrophilic catalysts is presumably due to the strongly acidic solvent, along with the application of high potentials, being detrimental to most electrodes. Indeed, we observed that under anodic polarization in sulfuric acid, even Pt electrodes corrode at the high temperatures typical for methane oxidation catalysis. However, under the same conditions, fluorine-doped tin oxide (FTO) electrodes

27 uniquely resisted degradation, even at an elevated temperature of 200 ̊C and over long periods of time. The

– inherent electrical conductivity of sulfuric acid was sufficiently high owing to autoionization into HSO4

+ 48 and H3SO4 , conveniently obviating the need for supporting electrolytes. With the acid- and oxidation- resistant FTO electrode, various metal ions could be tested for their ability to undergo electrochemical oxidation and activate methane.

Our preliminary results point to challenges and opportunities in the electrochemical generation of high-valent metal ions in strong acids. NiII, CuII, and RhIII were either insoluble or showed no sign of electrochemical oxidation on FTO within the potential window of the sulfuric acid solvent. PbII was oxidized to PbIV but did not activate methane. Oxidation of metallic Au, AgI, and CoII each led to a high- valent species that showed the desired reactivity with methane to yield methyl bisulfate, in line with previous reports of stoichiometric reactivity of these metals. Further investigation with these metals was hindered, however, because electro-generated AuIII precipitated to insoluble and electrically disconnected Au0 particles during methane oxidation catalysis, while AgI and CoII oxidation was accompanied by substantial side reactions such as solvent oxidation49 and product over-oxidation. We envision that modification of the metal ions with acid-stable ligands as well as a survey of other metals would lead to new potent methane functionalization catalysts.

In addition to accessing a wide range of electrophilic metal ion catalysts, electrochemical oxidation allows exploring various non-oxidizing strong acid media, as the solvent does not need to act as an oxidant any more. Another strong acid medium popularly used for electrophilic methane functionalization catalysis, trifluoroacetic acid, was non-conductive on its own but became conductive when trifluoroacetate salts were added as supporting electrolyte. While there are drawbacks such as low electrical conductivity and modest anodic potential limit due to the oxidative instability of this solvent

• towards releasing CF3 radicals and CO2, potentially easier separation of the methanol product after hydrolysis makes it attractive. Importantly, with electrochemical oxidation, the reaction medium can be chosen or engineered to optimize properties such as cost of product separation, solvent recycling, methane solubility, and reactor design constraints. Therefore, the electrochemical oxidation strategy opens up a wide space of exploration for selective methane oxidation by electrophilic homogeneous catalysts.

II Conspicuously, before this wide catalyst design space could be explored, Pd SO4 in concentrated or fuming sulfuric acid was found to undergo electrochemical oxidation to a previously unknown high- valent Pd species. The metastable complex selectively mono-functionalized methane to methyl bisulfate

–1 (CH3OSO3H) and methanesulfonic acid (CH3SO3H). The reaction rate, 2000 h at 140 ̊C under 500 psi of methane, was higher than any other electrophilic metal ion catalyst in sulfuric acid known to date. The electrochemical data acquired at different scan rates and Pd ion concentrations indicated that the high-valent

28 complex was formed via three sequential reactions: an electron transfer, a dimerization of two Pd ions, and another electron transfer. These and other pieces of data indicated the formation of a PdIII dimer, denoted

III 47 as Pd 2. This unexpected discovery highlights the potential of electrochemical oxidation for generating reactive high-valent complexes.

1.3.2.2. Approach 2: Mediated oxidation of the transient methyl complex

For methane functionalization by the bottom cycle in Scheme 1.2 where C–H activation is reversible, an oxidant must efficiently oxidize the transient methyl intermediate while leaving the low- valent active catalyst intact. The high-valent form of the catalyst in this case is inactive towards methane due to coordinative saturation and sluggish ligand exchange. Therefore, unlike in Approach 1, increasing the oxidizing driving force would not necessarily lead to faster catalysis. The oxidation of the methyl complex must be rapid, however, as it is a transient intermediate that can revert to the catalyst’s initial state and a free methane molecule. Thus, this mechanism requires selective yet fast oxidation of a low- concentration intermediate. Shilov’s catalyst, which follows this catalytic cycle, employs expensive PtIV

II ions as the stoichiometric oxidant as they can satisfy this difficult requirement, rapidly oxidizing Pt -CH3 without consuming inorganic PtII.

The direct electrochemical oxidation of a transient, low-concentration intermediate would be even more difficult than chemical oxidation because of the spatial confinement of electrochemical reactions to the two-dimensional electrode surface. However, indirect, mediated oxidation with an efficient oxidant such as the PtIV ions in Shilov’s catalyst would effectively expand the electrode’s reach to the three-dimensional solution phase. Periana’s PtII-bipyrimidine catalyst, which is an adaptation of Shilov’s catalyst in fuming sulfuric acid, was also found to involve the sulfuric acid oxidant oxidizing PtII to PtIV that subsequently

II II 50 oxidizes the Pt -CH3 intermediate, rather than the direct oxidation of Pt -CH3 by sulfuric acid. Importantly, such a mediation scheme must maintain the redox balance in the solution to preserve the low-valent active species while driving oxidative catalysis. Periana’s system maintains this redox balance by virtue of the slow oxidation of PtII to PtIV, which is the rate-limiting step of the overall reaction. Although the slow oxidation rate may be a drawback for the overall rate of catalysis, stable catalysis would have been challenging if oxidation were not rate-limiting, as other attempts to mediate turnover of Shilov’s catalyst with chemical oxidants have shown.29 In electrochemically mediated oxidation, however, the redox balance can be maintained by the precisely controlled delivery of oxidizing equivalents that is uniquely possible with electrochemistry. Additionally, the redox balance in the solution may be probed in real-time by electrochemical potential measurements.

1.4. Layout of the Thesis

Following up on the above two approaches for electrocatalytic methane functionalization, this

III thesis presents detailed structural and mechanistic studies of the newly discovered Pd 2 complex and the first demonstration of stable and continuous turnover of Shilov’s catalyst. In Chapter 2, a structural model of the potent high-valent Pd complex is assembled from X-ray absorption and Raman spectroscopic studies, whose key feature is the definitive presence of a Pd–Pd bond. Insights into the oxidation-induced self- assembly of this Pd–Pd bonded complex are gleaned from a thermochemical analysis with EPR and electrochemical data. Having established a structural foundation for understanding its rapid methane functionalization reactivity, experimental and computational mechanistic studies are covered in Chapter 3. The mechanistic model, which is consistent with all available experimental data and calculated to be energetically feasible, unexpectedly proposes H atom abstraction rather than electrophilic C–H activation. In this model, a common methyl radical intermediate bifurcates to the two products, methyl bisulfate and methanesulfonic acid. The unusual mechanism likely originates from the very high redox potential of the electro-generated complex. In Chapter 4, we apply mediated electrochemical oxidation to the well-known Shilov’s PtII catalyst to achieve stable and continuous electrocatalytic methane functionalization. The work highlights the previously understated importance of electrode surface adsorbates for facilitating electron transfer and the power of real-time control over the oxidation rate for maintaining the solution redox balance. Altogether, this thesis illustrates electrochemical approaches to achieving a difficult catalytic transformation.

1.5. Summary and Prospectus

In summary, we applied electrochemical oxidation to methane oxidation catalysis by metal ions in homogeneous liquids in order to overcome the challenges of, and even surpass what is possible with, stoichiometric oxidizing reagents. Recognizing that methane activation at a metal ion may proceed from either the high-valent state or the low-valent state, different strategic approaches were taken.

First, targeting enhanced electrophilicity for enhanced reactivity towards methane, we generated a highly oxidizing PdIII dimer from the electrochemical oxidation of PdII at high applied potentials. The initial product of the electrochemical oxidation, a monomeric PdIII ion, underwent spontaneous dimerization and further oxidation. The sulfuric acid medium, providing abundant weak-field ligands that axially ligated to the incipient dimer, facilitated the dimerization reaction and the second oxidation. Thus stabilized, the high-valent PdIII state could persist in the solution until it encountered and reacted with methane. This reaction occurred via H atom abstraction rather than electrophilic substitution; enabled by the high

30 oxidation potential, this outer-sphere proton-coupled electron transfer (PCET) pathway may be the key to

III the exceptionally high reaction rate and low activation barrier that the Pd 2 complex exhibited towards the

III extremely inert and non-coordinating substrate. In spite of the high reactivity of the Pd 2 complex, selective oxidation of methane was achieved by the spontaneous derivatization of the methanol product into the

II,III electron-deficient ester form in the sulfuric acid medium. Additionally, for the mixed-valent Pd2 intermediate generated after the H atom abstraction, a protective role was implicated from its ability to capture the O2-sensitive methyl radicals.

Second, to achieve stable and continuous oxidative catalysis for the case in which the metal ion is active in the less oxidized, low-valent state, we electrochemically monitored and controlled the solution redox balance in real-time. In the specific catalytic system we chose to study, the catalytically inactive high- valent ion, PtIV, functioned as an efficient oxidant for rapidly oxidizing the PtII-methyl intermediate, transiently generated from the reversible methane C–H activation at the catalytically active low-valent ion, PtII. The high-valent PtIV ion was also required for suppressing the disproportionation of PtII that produced Pt metal precipitates so that carefully maintaining sufficient concentrations of both the low- and high-valent states of the Pt ions was crucial. However, under constant current, even a slight discrepancy between the rate of methane oxidation catalysis and the rate of oxidation would be amplified over time, because the former reaction both produces and is accelerated by the low-valent ion. This necessitated the constant real- time modulation of the rate of oxidation, which could be achieved by measuring the instantaneous ratio of the low- and high-valent states from the solution redox potential and adjusting the current accordingly. Importantly, the success of this strategy hinged on the electrode’s ability to achieve facile redox interconversion between PtII and PtIV while remaining inert to the catalytic reaction product. In our case, these requirements were satisfied by the adsorption of Cl– ions in the electrolyte to the surface of the electrode that critically altered its electrocatalytic properties.

These results emphasize the unique yet under-appreciated advantages that the electrochemical, as opposed to chemical, delivery of redox equivalents offers for efficient catalysis. It is widely recognized that electrochemistry contributes to atom economy by replacing stoichiometric reagents with electricity, and the increasing availability of renewable electricity and the need for environmentally friendly technologies are fomenting a renewed interest in electrochemistry for the catalysis of a broad range of chemical transformations beyond those in sensors, batteries, and fuel cells. However, as our studies demonstrate, the electrode can do much more than just supplying or absorbing electrons. It can generate reactive species charged with high electrochemical potential energies to rapidly react even with methane. The electrode is also uniquely capable of reporting the solution redox balance and adjusting the flux of redox equivalents in

31 real-time, digitally. These powerful advantages of electrochemistry are not broadly appreciated yet in the context of catalytic transformations outside the traditional realms of electrochemistry.

Our studies also highlight the importance of catalytically inactive components in the electrolyte and their interaction with the catalytically active species as well as the electrode. The sulfuric acid medium, while serving as an electrically conductive solvent, was crucial to the formation of the dimeric PdIII complex from the electrochemical oxidation of monomeric PdII ions. It also played a synergistic role with the electrophilic catalyst for selective oxidations. In PtII-catalyzed methane oxidation, the Cl– ions were essential for the facile electrochemical oxidation of PtIV ions that enabled the electrochemical modulation of the PtII/PtIV redox balance. Notably, the Cl– ions performed this function by adsorbing to the electrode surface and modifying its electron transfer reactivity.

Projecting into the future, the desire and need for quantitative control over the driving force and rate of chemical reactions are likely to only increase. This thesis work, conducted in the specific context of selective methane oxidation catalysis by Pd and Pt ions in homogeneous solutions, aims to demonstrate how electrochemistry can vitally contribute to this pursuit.

1.6. References

(1) Holmen, A. Direct Conversion of Methane to Fuels and Chemicals. Catal. Today 2009, 142 (1–2), 2–8. (2) Promoppatum, P.; Viswanathan, V. Identifying Material and Device Targets for a Flare Gas Recovery System Utilizing Electrochemical Conversion of Methane to Methanol. ACS Sustain. Chem. Eng. 2016, 4 (3), 1736–1745. (3) Bank, W. Global Gas Flaring Reduction Partnership (GGFR) http://www.worldbank.org/en/programs/gasflaringreduction#7 (accessed Jan 24, 2018). (4) Wogan, T. Methane to methanol catalyst could end gas flaring www.chemistryworld.com/news/methane-to-methanol-catalyst-could-end-gas- flaring/3007247.article (accessed Feb 1, 2019). (5) Ravi, M.; Ranocchiari, M.; van Bokhoven, J. A. The Direct Catalytic Oxidation of Methane to Methanol—A Critical Assessment. Angew. Chemie - Int. Ed. 2017, 56 (52), 16464–16483. (6) Wang, V. C. C.; Maji, S.; Chen, P. P. Y.; Lee, H. K.; Yu, S. S. F.; Chan, S. I. Alkane Oxidation: Methane Monooxygenases, Related Enzymes, and Their Biomimetics. Chem. Rev. 2017, 117 (13), 8574–8621. (7) Sushkevich, V. L.; Palagin, D.; Ranocchiari, M.; van Bokhoven, J. A. Selective Anaerobic Oxidation of Methane Enables Direct Synthesis of Methanol. Science (80-. ). 2017, 356 (6337), 523–527. (8) Zheng, J.; Ye, J.; Ortuño, M. A.; Fulton, J. L.; Gutiérrez, O. Y.; Camaioni, D. M.; Motkuri, R. K.; Li, Z.; Webber, T. E.; Mehdi, B. L.; et al. Selective Methane Oxidation to Methanol on Cu-Oxo Dimers Stabilized by Zirconia Nodes of an NU-1000 Metal-Organic Framework. J. Am. Chem. Soc. 2019,

141 (23), 9292–9304. (9) Dinh, K. T.; Sullivan, M. M.; Serna, P.; Meyer, R. J.; Dincă, M.; Román-Leshkov, Y. Viewpoint on the Partial Oxidation of Methane to Methanol Using Cu- and Fe-Exchanged Zeolites. ACS Catal. 2018, 8 (9), 8306–8313. (10) Narsimhan, K.; Iyoki, K.; Dinh, K.; Román-Leshkov, Y. Catalytic Oxidation of Methane into Methanol over Copper-Exchanged Zeolites with Oxygen at Low Temperature. ACS Cent. Sci. 2016, 2 (6), 424–429. (11) Shilov, A. E.; Shul’pin, G. B. Activation and Catalytic Reactions of Saturated Hydrocarbons in the Presence of Metal Complexes; Catalysis by Metal Complexes; Kluwer Academic Publishers: Dordrecht, 2002; Vol. 21. (12) Goldshleger, N. F.; Eskova, V. V.; Shilov, A. E.; Shteinman, A. A. Reactions of Alkanes in Solutions of Chloride Complexes of Platinum. Russ. J. Phys. Chem. 1972, 46 (5), 785–786. (13) Cundari, T. R. Methane Manifesto: A Theorist’s Perspective on Catalytic Light Alkane Functionalization. Comments Inorg. Chem. 2017, 37 (5), 219–237. (14) Crabtree, R. H. Alkane C-H Activation and Functionalization with Homogeneous Transition Metal Catalysts: A Century of Progress - A New Millennium in Prospect. J. Chem. Soc. Dalt. Trans. 2001, No. 17, 2437–2450. (15) Labinger, J. A.; Bercaw, J. E. Understanding and Exploiting C–H Bond Activation. Nature 2002, 417 (6888), 507–514. (16) Shteinman, А. А. Activation and Selective Oxy-Functionalization of Alkanes with Metal Complexes: Shilov Reaction and Some New Aspects. J. Mol. Catal. A Chem. 2017, 426, 305–315. (17) Stahl, S. S.; Labinger, J. A.; Bercaw, J. E. Exploring the Mechanism of Aqueous C-H Activation by Pt(II) through Model Chemistry: Evidence for the Intermediacy of Alkylhydridoplatinum(IV) and Alkane σ-Adducts. J. Am. Chem. Soc. 1996, 118 (25), 5961–5976. (18) Cook, A. K.; Schimler, S. D.; Matzger, A. J.; Sanford, M. S. Catalyst-Controlled Selectivity in the C-H Borylation of Methane and Ethane. Science (80-. ). 2016, 351 (6280), 1421–1424. (19) Ahlquist, M.; Periana, R. A.; Goddard, W. A. C-H Activation in Strongly Acidic Media. the Co- Catalytic Effect of the Reaction Medium. Chem. Commun. 2009, No. 17, 2373–2375. (20) Ahlquist, M.; Nielsen, R. J.; Periana, R. A.; Goddard, W. A.; Goddard III, W. A. Product Protection, the Key to Developing High Performance Methane Selective Oxidation Catalysts. J. Am. Chem. Soc. 2009, 131 (47), 17110–17115. (21) Periana, R. A.; Taube, D. J.; Evitt, E. R.; Löffler, D. G.; Wentrcek, P. R.; Voss, G.; Masuda, T. A Mercury-Catalyzed, High-Yield System for the Oxidation of Methane to Methanol. Science (80-. ). 1993, 259 (5093), 340–343. (22) Hashiguchi, B. G.; Konnick, M. M.; Bischof, S. M.; Gustafson, S. J.; Devarajan, D.; Gunsalus, N.; Ess, D. H.; Periana, R. A. Main-Group Compounds Selectively Oxidize Mixtures of Methane, Ethane, and Propane to Alcohol Esters. Science (80-. ). 2014, 343 (6176), 1232–1237. (23) Basickes, N.; Hogan, T. E.; Sen, A. Radical-Initiated Functionalization of Methane and Ethane in Fuming Sulfuric Acid. J. Am. Chem. Soc. 1996, 118 (51), 13111–13112. (24) Mukhopadhyay, S.; Bell, A. T. Catalyzed Sulfonation of Methane to Methanesulfonic Acid. J. Mol. Catal. A Chem. 2004, 211 (1–2), 59–65. (25) Gunsalus, N. J.; Koppaka, A.; Park, S. H.; Bischof, S. M.; Hashiguchi, B. G.; Periana, R. A. Homogeneous Functionalization of Methane. Chem. Rev. 2017, 117 (13), 8521–8573. 33

(26) Natinsky, B. S.; Lu, S.; Copeland, E. D.; Quintana, J. C.; Liu, C. Solution Catalytic Cycle of Incompatible Steps for Ambient Air Oxidation of Methane to Methanol. ACS Cent. Sci. 2019. (27) Labinger, J. A.; Bercaw, J. E. Mechanistic Studies on the Shilov System: A Retrospective. J. Organomet. Chem. 2015, 793, 47–53. (28) Bar-Nahum, I.; Khenkin, A. M.; Neumann, R. Mild, Aqueous, Aerobic, Catalytic Oxidation of Methane to Methanol and Acetaldehyde Catalyzed by a Supported Bipyrimidinylplatinum- Polyoxometalate Hybrid Compound. J. Am. Chem. Soc. 2004, 126 (33), 10236–10237. (29) DeVries, N.; Roe, D. C.; Thorn, D. L. Catalytic Hydroxylation Using Chloroplatinum Compounds. J. Mol. Catal. A Chem. 2002, 189 (1), 17–22. (30) Periana, R. A.; Taube, D. J.; Gamble, S.; Taube, H.; Satoh, T.; Fujii, H. Platinum Catalysts for the High-Yield Oxidation of Methane to a Methanol Derivative. Science (80-. ). 1998, 280 (5363), 560– 564. (31) Michalkiewicz, B. Assessment of the Possibility of the Methane to Methanol Transformation. Polish J. Chem. Technol. 2008, 10 (2), 20–26. (32) Yuan, J.; Liu, L.; Wang, L.; Hao, C. Partial Oxidation of Methane with the Catalysis of Palladium(II) and Molybdovanadophosphoric Acid Using Molecular Oxygen as the Oxidant. Catal. Letters 2013, 143 (1), 126–129. (33) An, Z.; Pan, X.; Liu, X.; Han, X.; Bao, X. Combined Redox Couples for Catalytic Oxidation of Methane by Dioxygen at Low Temperatures. J. Am. Chem. Soc. 2006, 128 (50), 16028–16029. (34) Vargaftik, M. N.; Stolarov, I. P.; Moiseev, I. I. Highly Selective Partial Oxidation of Methane to Methyl Trifluoroacetate. J. Chem. Soc. Chem. Commun. 1990, No. 15, 1049–1050. (35) Kärkäs, M. D. Electrochemical Strategies for C–H Functionalization and C–N Bond Formation. Chem. Soc. Rev. 2018, 47 (15), 5786–5865. (36) Meyer, T. H.; Finger, L. H.; Gandeepan, P.; Ackermann, L. Resource Economy by Metallaelectrocatalysis: Merging Electrochemistry and C H Activation. Trends Chem. 2019, 1 (1), 63–76. (37) Yan, M.; Kawamata, Y.; Baran, P. S. Synthetic Organic Electrochemical Methods Since 2000: On the Verge of a Renaissance. Chem. Rev. 2017, 117 (21), 13230–13319. (38) Xie, S.; Lin, S.; Zhang, Q.; Tian, Z.; Wang, Y. Selective Electrocatalytic Conversion of Methane to Fuels and Chemicals. J. Energy Chem. 2018, 27 (6), 1629–1636. (39) Stoukides, M. Electrochemical Studies of Methane Activation. J. Appl. Electrochem. 1995, 25 (10). (40) Lee, B.; Hibino, T. Efficient and Selective Formation of Methanol from Methane in a Fuel Cell- Type Reactor. J. Catal. 2011, 279 (2), 233–240. (41) Che, F.; Ha, S.; McEwen, J. S. Catalytic Reaction Rates Controlled by Metal Oxidation State: C−H Bond Cleavage in Methane over Nickel-Based Catalysts. Angew. Chemie - Int. Ed. 2017, 56 (13), 3557–3561. (42) Li, W.; He, D.; Hu, G.; Li, X.; Banerjee, G.; Li, J.; Lee, S. H.; Dong, Q.; Gao, T.; Brudvig, G. W.; et al. Selective CO Production by Photoelectrochemical Methane Oxidation on TiO2. ACS Cent. Sci. 2018, 4 (5), 631–637. (43) Arnarson, L.; Schmidt, P. S.; Pandey, M.; Bagger, A.; Thygesen, K. S.; Stephens, I. E. L.; Rossmeisl, J. Fundamental Limitation of Electrocatalytic Methane Conversion to Methanol. Phys. Chem. Chem. Phys. 2018, 20 (16), 11152–11159.

(44) Freund, M. S.; Labinger, J. A.; Lewis, N. S.; Bercaw, J. E. Electrocatalytic Functionalization of Alkanes Using Aqueous Platinum Salts. J. Mol. Catal. 1994, 87 (1), L11–L15. (45) Liu, S. F.; Nusrat, F. Electrocatalytic Shilov Chemistry for the Oxidation of Aliphatic Groups. Mol. Catal. 2019, 463, 16–19. (46) Joglekar, M.; Nguyen, V.; Pylypenko, S.; Ngo, C.; Li, Q.; O’Reilly, M. E.; Gray, T. S.; Hubbard, W. A.; Gunnoe, T. B.; Herring, A. M.; et al. Organometallic Complexes Anchored to Conductive Carbon for Electrocatalytic Oxidation of Methane at Low Temperature. J. Am. Chem. Soc. 2016, 138 (1), 116–125. (47) O’Reilly, M. E.; Kim, R. S.; Oh, S.; Surendranath, Y. Catalytic Methane Monofunctionalization by an Electrogenerated High-Valent Pd Intermediate. ACS Cent. Sci. 2017, 3 (11), 1174–1179. (48) Cotton, F. A.; Wilkinson, G.; Murillo, C. A.; Bochmann, M. Advanced Inorganic Chemistry, 6th ed.; 1999. (49) Połczyński, P.; Jurczakowski, R.; Grochala, W. Strong and Long-Lived Free-Radical Oxidizer Based on Silver(II). Mechanism of Ag(I) Electrooxidation in Concentrated H2SO4. J. Phys. Chem. C 2013, 117 (40), 20689–20696. (50) Mironov, O. a.; Bischof, S. M.; Konnick, M. M.; Hashiguchi, B. G.; Ziatdinov, V. R.; Goddard, W. a.; Ahlquist, M.; Periana, R. a.; Goddard III, W. A.; Ahlquist, M.; et al. Using Reduced Catalysts for Oxidation Reactions: Mechanistic Studies of the “Periana-Catalytica” System for CH4 Oxidation. J. Am. Chem. Soc. 2013, 135 (39), 14644–14658.

III 2. Structure of Pd 2 and Its Mechanism of Formation via Electrochemical Oxidation

Parts of this chapter have been adapted and reprinted with permission from O’Reilly, M. E.; Kim, R. S.; Oh, S.; Surendranath, Y. Catalytic Methane Monofunctionalization by an Electrogenerated High-Valent Pd Intermediate. ACS Cent. Sci. 2017, 3 (11), 1174–1179.

The chapter contains contributions from collaborators:

(X-ray absorption spectroscopy) Dr. Evan C. Wegener, Prof. Jeffrey T. Miller

(DFT computation) Min Chieh Yang, Prof. Christopher H. Hendon

2.1. Introduction

III 1 2.1.1. Electro-generated Pd 2 in sulfuric acid

As introduced in Chapter 1, we discovered that the electrochemical oxidation of PdII sulfate in concentrated or fuming sulfuric acid results in the formation of a dinuclear PdIII complex. The cyclic voltammogram (CV) with a characteristic hysteresis (Figure 2.1a) is diagnostic of an ECE mechanism, in which a chemical step (C) occurs between two distinct electron transfer steps (E), with the second electron transfer occurring at a lower potential than the first. When the rate of the C step is comparable to the scan rate, larger anodic current flows in the return scan than the forward scan because the second E step, while thermodynamically favorable, can occur only after the C step following the first E step generates the more easily oxidized intermediate. A fluorine-doped tin oxide (FTO) electrode and a Pt wire were employed as the working electrode and the pseudo-reference electrode, respectively. The pseudo-reference electrode was externally calibrated to the Ag2SO4/Ag redox couple in saturated Ag2SO4 in sulfuric acid (SSE), whose potential is reported to be 0.815 V vs NHE.2 CVs obtained at varying scan rates showed the disappearance of the hysteresis at faster scan rates and growth of another reduction peak at ~1.4 V vs SSE, which corresponds to the reverse of the first oxidation step and indicates that the C step is outcompeted by this cathodic reaction during fast scans (Figure 2.1b). Based on the oxidation states available for Pd and the ECE sequence, two mechanisms were proposed with different molecularity of the C step (Figure 2.1c). The

36 dependence of the ratio of two reduction peaks on PdII concentration supported the binuclear mechanism (Figure 2.1d). Overall, the voltammetric data implied the formation of a dinuclear PdIII species, denoted as

III Pd 2.

III The putative Pd 2 complex could be generated via bulk electrolysis at room temperature for 2

III days. When a fuming sulfuric acid solution of Pd 2 was reacted with methane at 100 ̊C, two monofunctionalized products resulted: methyl bisulfate (CH3OSO3H, MBS) and methanesulfonic acid

III II (CH3SO3H, MSA) (Figure 2.2a). The reaction mixture showed full reduction of Pd 2 to Pd without any Pd0 formation. Quantitation of products showed that one equivalent of methyl bisulfate was formed per one

III equivalent of Pd 2, in agreement with the oxidation state assigned to the high-valent Pd complex. On the other hand, methanesulfonic acid was generated in superstoichiometric amounts, suggesting a catalytic role

III of Pd 2 in the formation of the latter product (Figure 2.2b).

III Although the Pd atoms in the Pd 2 complex has an odd number of d electrons, paramagnetic susceptibility measured by Evans method indicated that the complex is diamagnetic (Figure 2.3). Based on the NMR resolution (0.001 ppm) and the concentration of Pd (10 mM), we conclude that the magnetic moment per Pd ion in the sample is less than 0.24 μB. Using the same preparation method and 10 mM of

II Ni SO4 in H2SO4, we observed a magnetic moment of 2.9 μB, close to the theoretical value of 2.8 μB. The

III III lack of paramagnetic shift indicates that the unpaired electrons on each Pd centers are paired in the Pd 2 complex, which can result from Pd–Pd bond formation, antiferromagnetic coupling through a bridging

3 II,IV III,III 4,5 ligand, or asymmetric valency (i.e. Pd2 instead of Pd2 ).

1 III II Figure 2.2. (a) H NMR of the reaction mixture after treating a (black) 4.2 mM Pd 2 and (red) 8.4 mM Pd solution in 20% SO3/H2SO4 with 500 psi of CH4 at 100 ̊C for 20 min. (b) Methane oxidation reactions of III Pd 2 based on the observed stoichiometry for the two products.

+ 1 Figure 2.3. NH4 peaks in the H NMR spectra for Evans method magnetic susceptibility measurements. ~5 mM of ammonium sulfate ((NH4)2SO4) was used as a paramagnetic shift reference compound. Blue: III II II III Pd 2 (post-electrolysis) solution; Red: Pd (pre-electrolysis) solution; Black: the Pd and Pd 2 solutions in II III co-axial inner (3 mm dia.) and outer (5 mm dia.) tubes. To, Spectra of Pd and Pd 2 solutions were independently obtained to exclude the possibility that the observed peak shape results from the overlap of two closely-spaced peaks. All three spectra display similar linewidths, indicating a perfect overlap of the III peaks in the coaxial double-chamber tube and no paramagnetic shift by Pd 2.

III 2.1.2. The need for elucidation of the structure of Pd 2 and its formation mechanism

The foregoing studies established the ECE sequence for PdII oxidation and the nuclearity and average Pd oxidation state of the electro-generated high-valent Pd complex. However, they provided

III negligible insight into the molecular structure of Pd 2 or the intermediates involved in its electrochemical

II III generation from mononuclear Pd . In particular, the nature of the metal-metal interaction in Pd 2 remains

III 6,7 unknown. Whereas many Pd 2 complexes contain direct Pd–Pd bonds, the Pd centers can also be linked

3,8 III through one or more bridging ligands. Furthermore, known metal-metal bonded Pd 2 complexes are

II generally obtained via oxidation of co-facially oriented, ligand-bridged, dinuclear Pd 2 complexes in which

III the Pd centers are predisposed towards facile M–M bond formation. Additionally, prior Pd 2 complexes

III are generally formed using two-electron chemical oxidants. In contrast, our Pd 2 complex is generated via sequential one-electron electrochemical oxidation from a simple mono-nuclear PdII(sulfate) complex,

III leaving open the critical questions of whether an M–M bond exists in the Pd 2 species and what role it plays in fostering the unique ECE electrochemical oxidation mechanism. Given the key role of M–M bonding in high valent Pd oxidation catalysis,9 addressing these structural and mechanistic knowledge gaps is critical for the rational design of Pd-mediated electrochemical C–H functionalization.

III Herein, we establish the core structure of the electrochemically generated Pd 2 and provide a

III structural basis for its unique mechanism of formation. Since the Pd 2 complex cannot be isolated from the sulfuric acid medium, we combine X-ray absorption and Raman spectroscopies to establish that it contains a Pd–Pd bond with each Pd atom coordinated by 5 O atoms. Against this backdrop, we use EPR spectroscopy to identify a mixed-valent intermediate in the ECE reaction sequence and combine this data with electrochemical measurements to map the thermodynamic landscape that drives the dimerization of the two Pd centers. Analysis of our results and previous electrochemical studies in the literature reveal the importance of the Pd–Pd bond and axial ligand coordination for enabling the electrochemical oxidation and

II III dimerization of Pd to Pd 2. These insights provide a foundation for understanding the unusual reactivity

III 10 of Pd 2 and for electrochemically generating new high-valent Pd complexes that may enable challenging C–H functionalization reactions.

2.2. Results and Discussions

III 2.2.1. Structure of Pd 2

2.2.1.1. Sample preparation

III II The Pd 2 sample for spectroscopic investigation was generated by bulk electrolysis of Pd SO4, in fuming H2SO4 containing 18–24% SO3 by weight. We designate the pre- and post-electrolysis samples as 1 and 2, respectively. 2 displays a strong absorbance at 300 nm,1 which allowed us to monitor the progress of the electrolysis by UV–Vis spectroscopy. The presence of SO3 suppressed the spontaneous reduction of

– the high-valent species which presumably occurs via solvent oxidation (H2SO4 → ½ O2 + SO3 + 2 e ). The

II solubility of Pd SO4 in fuming sulfuric acid was limited to ca. 10 mM; in our effort to increase the signal-

II to-noise ratio of our measurements, we found that addition of 1.4 M of (NH4)2SO4 increased Pd solubility to ca. 50 mM.11 Voltammetric, spectroscopic and methane reactivity studies all indicated that electrolysis

III in the presence of (NH4)2SO4 generates the same Pd 2 complex (see 2.4.2.7). Thus, low and high

39 concentration samples of 1 and 2, designated as 1/2-lc and 1/2-hc, were prepared in the absence and presence of (NH4)2SO4, respectively, for further analysis.

2.2.1.2. X-ray Absorption Spectroscopy X-ray absorption near edge structure (XANES) spectra support the formation of a high-valent species upon electrooxidation of 1 (Figure 2.4a). Consistent with an increase in the oxidation state, the XANES spectrum of 2 displays a rising edge inflection point that is 7.4 eV higher than 1 (Table 2.1) for both lc and hc samples. Repeated measurements on the same sample did not show any shift in the edge energy, indicating that the complex was not subject to X-ray photodegradation over the timescale of the measurement. Importantly, the XANES spectrum of 2 displays a smooth-rising edge. This implies uniform Pd oxidation states and argues against the presence of a mixed-valent dinuclear species such as PdII–PdIV, which has been demonstrated in the presence of disparate apical ligand environments.4,5

Extended X-ray absorption fine structure (EXAFS) spectra indicate that the PdIII centers are each ligated by 5 oxygen donor ligands. The Fourier-transformed EXAFS of 1 and 2 in R-space both display a prominent peak at 1.5 Å (phase uncorrected distance) arising from the first nearest neighbor scattering

(x−2) interactions (Figure 2.4b). Since the only ligand present in the system is HxSO4 , this peak was isolated and fitted with oxygen scatterers (Table 2.1). For 1, the Pd–O coordination number of 4 and average bond

II 12 distance of 2.01 Å were in line with the known structure of square-planar Pd SO4. Similar fitting of the

III Pd 2 species, 2, revealed an increase in the number of coordinated O atoms to 5 and a slight contraction of the average Pd–O bond distance to 1.99 Å. Known structurally characterized PdIII complexes usually show a distorted pseudo-octahedral coordination,13 and the change in edge shape, particularly evident in the first derivative of the spectra (Figure 2.4c), suggests a transition from square-planar to octahedral coordination of the Pd center upon electrooxidation (see 2.4.5.3 for a detailed explanation).14–17 Moreover, the diamagnetism of 21 implies electronic coupling between the two d7 PdIII centers. These observations led us to reason that there should be a Pd as the sixth coordinating atom. Notably, this result rules out the formation of polynuclear 1-D chains of PdIII centers.18

Figure 2.4. Pd K-edge X-ray absorption spectra of 1-hc and 2-hc: (a) XANES; (b) EXAFS showing the real (solid line) and imaginary (dashed line) components; (c) 1st derivative of the XANES; the lc samples III showed essentially identical results (Figure 2.13–Figure 2.16). (d) Pt K-edge EXAFS of Pt 2 in the solid state.

Table 2.1. Summary of XAS results.

2 3 2 Sample Edge E (keV) CNPd-O R (Å) σ (×10 Å ) E0 (eV) 1, solid 24.3550 4.0 2.01 1.4 1.2 1-lc 24.3550 4.0 2.00 3.0 –0.3 1-hc 24.3550 4.0 2.01 1.2 –0.2 2-lc 24.3624 5.0 2.00 2.7 2.6 2-hc 24.3624 5.2 1.98 0.9 2.8

III Pt 2, solid 11.5660 5.0 1.98 3.2 –2.2 In principle, the EXAFS of 2 should contain a contribution from Pd–Pd scattering, as has been shown for other M–M bonded compounds.19–23 However, while detection of metal-metal interactions by EXAFS is possible, reliable assignment depends on the nature of the solvent, the strength of the scattering, and the overlapping scattering from atoms at longer distances.24,25 Here, we cannot conclusively fit the Pd– Pd scattering peak due to the weak signal beyond the strong first peak and possible interference from multiple scattering paths from sulfurs and oxygens. In an attempt to aid our assignment, we prepared the

III III III well-known paddlewheel Pt dimer with sulfate ligands, K2[Pt 2(SO4)4(H2O)2] (abbreviated as Pt 2),

26 III which features a Pt–Pt bond. The EXAFS of Pt 2 (Figure 2.4d) was similar to that of 2, and consistent with 5 coordinated oxygens (Table 2.1). The relatively weak features in the higher shell, however, again

41 prevented reliable fitting of the Pt–Pt vector. Since Pd is lighter than Pt, Pd–Pd scattering is weaker and more difficult to detect by EXAFS.27 While this second inconclusive measurement of M–M scattering is not proof of the existence of our proposed Pd–Pd interaction, it suggests that EXAFS alone cannot be used to establish the nature of the metal-metal connectivity in this system.

2.2.1.3. Raman Spectroscopy Raman spectra of M–M single bonds are documented for a variety of dinuclear metal

28,29 III 30 complexes including Pt 2. Fortunately, the high mass and relatively weak force constant of M–M single bonds make their vibrations appear in the 100–300 cm–1 region, where the spectrum of the fuming

H2SO4 solvent is relatively featureless (Figure 2.5a). To confirm our ability to observe the M–M vibration,

III we acquired the Raman spectra of Pt 2 (Figure 2.5b). In fuming H2SO4, with or without (NH4)2SO4, we observed a peak at 227 cm−1 (Figure 2.5b, orange & red). A similar peak appears at 237 cm−1 in 1 M

− aqueous H2SO4 (Figure 2.5b, blue). Importantly, upon addition of Cl , this peak diminishes and is replaced by a new peak at 209 cm−1 (Figure 2.5b, green; Figure 2.17 shows time-dependent evolution of the spectrum). Since Cl− is known to substitute for axial ligands and bind trans to the Pt–Pt bond,30,31 this observation strongly supports the assignment of these peaks to a Pt–Pt vibration.

III Figure 2.5. Raman spectra of (a) fuming H2SO4; (b) Pt 2 in fuming H2SO4, with or without (NH4)2SO4, and aqueous solutions; (c) 1 and 2 in fuming H2SO4 with or without (NH4)2SO4.

Encouraged by this result, we collected the Raman spectra of 1 and 2 (Figure 2.5c). Expectedly, 1 is featureless in the 100–300 cm–1 region (Figure 2.5c, purple). Contrastingly, both 2-hc and 2-lc show a new low energy peak centered at 268 cm–1 (Figure 2.5c, orange & red), the magnitude of which is much larger in the higher Pd concentration sample. This feature is higher in energy than the 227 cm–1 peak

III observed in the spectrum of Pt 2, consistent with the lower atomic mass of Pd. Moreover, polarized Raman measurements gave a low and identical depolarization ratio of ca. 0.4 for both the 268 cm–1 band of 2-hc

–1 III and the 227 cm peak of Pt 2 (Figure 2.18). The low depolarization ratio is consistent with a totally symmetric vibration that is expected for an M–M vibration, further supporting the assignment of the 227

–1 –1 III III 32 –1 cm and 268 cm peaks to M–M stretches in Pt 2 and Pd 2, respectively. Notably, the 268 cm peak of

−1 III 2 is much broader than the 227 cm peak of Pt 2 in the same medium. This may be due to a more labile

III III and dynamic coordination environment of Pd 2 compared to Pt 2. Additionally, the spectrum of 2-hc displays a distorted and attenuated solvent peak at ca. 325 cm–1. We speculate that this change in solvent

III modes may arise from changes in hydrogen bonding caused by the relatively high concentration of Pd 2 species in the presence of high salt concentration (1.4 M). Together, these Raman data provide positive

III evidence for the presence of a Pd–Pd bond in Pd 2.

III 2.2.1.4. Structural Model of Pd 2

III In combination, the above studies allow us to assemble a structural model for the Pd 2 species. X- ray absorption spectroscopy indicates 5-fold coordination by oxygen atoms and an octahedral geometry at each Pd, and Raman spectroscopy strongly supports the presence of a metal-metal vibration mode. This

III III allows us to conclude that the structure of our Pd 2 complex consists of a (Pd O5)2 core that is analogous to the known PtIII sulfate dimer.26 These experimental observations, though, do not yield information about

III III the exact ligand geometry of Pd 2. Pt 2 in the solid state is ligated by sulfates in a four-fold bridging paddlewheel structure and the relatively narrow Raman band for Pt–Pt vibration is retained across various solvents (Figure 2.5b), suggesting that this paddlewheel structure persists in solution. In contrast with the

III relatively narrow Raman peak of Pt 2, 2 displays an extremely broad Raman signal, suggesting that a simple

III III paddlewheel ligation structure for Pd 2 is unlikely. Initial computational modeling of Pd 2 revealed a number of viable conformers and protonation isomers (Figure 2.6b and Figure 2.25), but poor agreement between the experimental and calculated Raman spectra was found (Figure 2.26). In Chapter 3, we used

III free energies as the basis for refining DFT models for the Pd 2 complex. Furthermore, a larger number of isomers and conformations were studied.

III − Figure 2.6. DFT-optimized structures of Pd 2 with six HSO4 ligands with four, two, and zero bridging bisulfates. See 2.4.9 for computational details and other isomers that were calculated. White: H, red: O, yellow: S, light grey: Pt, dark grey: Pd.

II,III 2.2.2. Identification and Structural Assignment of a Pd2 Intermediate

2.2.2.1. Detection and assignment of an EPR signal

III Our previous electrochemical data pointed to an ECE mechanism for the formation of Pd 2 that is detailed below (equations 2.1–2.3; E1 and E2 represent standard reduction potentials and ΔGdim,het stands for the free energy of heterodimerization). This sequence invokes two odd-electron Pd species, PdIII and

II,III II,III Pd2 , as putative intermediates. We stress that Pd2 is a formal notation that does not imply electron localization. Although our bulk-electrolyzed Pd solution, 2, is diamagnetic by Evans’ method analysis,1 this solution none-the-less revealed a weak EPR signal at low temperatures that is absent in 1 (Figure 2.7a). The high anisotropy of this EPR signal indicates that the unpaired electron is metal-based,33 providing positive evidence for an EPR-active Pd minor component in 2.

E Pd ⇄ Pd + e 퐸 … 2.1 , C Pd + Pd ⇄ Pd ∆퐺, … 2.2 , E Pd ⇄ Pd + e 퐸 … 2.3 III II,III In order to determine which radical intermediate, Pd or Pd2 , gives rise to the observed EPR signal, we prepared a series of solutions containing the same total Pd ion concentration but with varying

II III ratios of Pd and Pd 2 and measured the spin concentrations (see 2.4.3 for details of sample preparation

[ ] and data collection). The measured spin concentrations are plotted vs ox.%≡ in Figure 2.7b. [ ][ ] III III III Since Pd arises from fragmentation of Pd 2 (equation 2.4), [Pd ] is expected to increase monotonically

II,III with ox.% (equation 2.5). In contrast, [Pd2 ] should display a maximum at intermediate levels of oxidation

II III because it arises from the comproportionation of Pd and Pd 2 (equations 2.6 and 2.7). Thus, the observed parabolic trend (Figure 2.7b, black squares) supports assignment of this signal to the mixed-valent dimer,

II,III III Pd2 , rather than the monomeric Pd intermediate.

2 Pd ⇄ Pd ∆퐺, … 2.4

[Pd ] [Pd](ox. %) [Pd] = = … 2.5 퐾, 2퐾,

, 2 Pd + Pd ⇄ 2 Pd ∆퐺 … 2.6

퐾[Pd] Pd , = 퐾 [Pd][Pd ] = (1 − ox. %)(ox. %) … 2.7 2

Figure 2.7. (a) Background-corrected X-band EPR spectrum of 2 at 60 K. (b) (Black squares) EPR- II measured spin concentrations versus ox.%. Total Pd concentration was 9.3 mM. Cu SO4 dissolved in the same medium was used as a spin quantification standard (see 2.4.7 for details). (Red line) Calculated II,III [Pd2 ] from a least-squares fitting of equation 7 to the EPR-measured spin concentrations.

The equilibrium concentration data in Figure 2.7b yields the thermodynamic stability of this

II III mixed-valent intermediate relative to Pd and Pd 2. Fitting equation 2.7 to the data (Figure 2.7b, red line),

−1 –1 we obtained Kcomp = 0.78±0.11 M and ΔGcomp = 0.15±0.08 kcal mol . While the standard comproportionation free energy is only slightly positive, due to the low total Pd concentration of our

II,III II III solutions (<10 mM), the Pd2 in our samples is driven towards disproportionation to 2 Pd + Pd 2.

II,III Together, this analysis establishes that the EPR signal in 2 arises from a Pd2 species formed from minor-

III II equilibrium comproportionation of Pd 2 and residual Pd .

II,III 2.2.2.2. Structure of Pd2

III II,III To be consistent with the structural model for Pd 2 put forward in the previous section, the Pd2 complex is expected to have an incipient Pd–Pd bond with a bond order of 0.5. The unpaired electron would

34,35 reside in the M–M σ* orbital formed from the overlap of the dz2 orbitals along the Pd–Pd bond axis.

33 Indeed, the axial EPR signal with gx,y > gz is consistent with the dz2 character of the SOMO and with the

II,III 34,36,37 spectra of other mixed-valent Pd2 complexes in the literature. Furthermore, the g-tensor anisotropy of our dimer is exceptionally high, gx,y = 2.33 and gz = 2.01; the spread in g components is the greatest

II,III 34,36,37 among those of reported Pd2 complexes. This large anisotropy indicates low metal-ligand

(x−2) 38 covalency, as would be expected for the hard HxSO4 ligands.

Electrochemical data are also consistent with a formally mixed-valent dimer with a half-order Pd– Pd bond. Prior voltammetric data recorded on FTO electrodes (Figure 2.8) was insufficient for the quantitative determination of E1 and E2 due to the convolution of slow ET kinetics and an unknown rate constant for the chemical step in the ECE mechanism. Using a combination of a Pt electrode with faster ET kinetics and low PdII concentrations to slow the dimerization C step, we were able to observe a chemically reversible PdII/III oxidation wave at a midpoint potential of 1.69 V vs SSE (Figure 2.21). Because the voltammetric features indicated adsorption of PdIII on the Pt electrode surface, this is a lower limit for the value of E1 (see 2.4.8.2 for details). As for E2, it is equal to the standard redox potential for the overall

II III Pd /Pd 2 redox process minus ΔGcomp/2F (see 2.4.8.1 for derivation); therefore, we used open circuit

II III measurements of mixed solutions of Pd and Pd 2 (Figure 2.23) to extract E2 = 1.49 V. This analysis

II,III II establishes that Pd2 is easier to oxidize by >200 mV than Pd , and is consistent with a M−M σ* SOMO

II,III II in Pd2 that is significantly higher in energy than the native dz2 orbital in square-planar Pd . Together, the

II, III EPR and electrochemical data support the intermediacy of an S=½ Pd2 complex, containing a half-order

III Pd–Pd bond, formed en route to Pd 2.

III 2.2.3. Structural and Thermochemical Basis for Electrochemical Pd 2 Formation

The foregoing structural and spectroscopic insights provide a basis for understanding the unusual

III II ECE mechanism that forms Pd 2 by electrochemical oxidation of Pd .

2.2.3.1. Driving force for dimerization

III The viability of the ECE mechanism and the formation of Pd 2 require thermodynamically favorable dimerization reactions (equations 2.2 and 2.4). The foregoing ΔGcomp, E1 and E2 values allow us

II,III III to compute free energies of dimerization for both Pd2 and Pd 2. In Scheme 2.1, we summarize the set of equilibria describing the Pd system: E1, ΔGdim,het and E2 are as defined in equations 2.1–2.3, and ΔGdim,hom and ΔGcomp are in equations 2.4 and 2.6.

II III III II,III Scheme 2.1. Summary of the reactions between Pd , Pd 2, Pd and Pd2 .

Hess’s law provides

∆퐺, = ∆퐺 + 퐹(퐸 − 퐸) … 2.8

∆퐺, = ∆퐺 + 2퐹(퐸 − 퐸) … 2.9 –1 II,III (see 2.4.8.1 for derivations). From ΔGcomp = 0.15 kcal mol and E2 – E1 < –0.20 V, we compute that Pd2

III and Pd 2 are stabilized relative to their constituent monomers by more than 4.5 and 9.1 kcal/mol, respectively. These exergonic dimerization free-energies are consistent with the occurrence of the ECE sequence, in which the C step is a spontaneous heterodimerization reaction. Although homodimerization is even more strongly favored than heterodimerization, analysis of the hysteresis in the cyclic voltammogram for PdII oxidation establishes heterodimerization as the dominant pathway. Comparison of the inherent kinetics (i.e., rate constants) of hetero- versus homodimerization is obscured by the dynamically changing population of PdIII and PdII at the electrode surface during the potential sweep. The calculated dimerization free energies highlight the high preference for dimerizing of Pd centers in this system that serves to promote

III ECE oxidation and Pd 2 formation.

III 2.2.3.2. M–M and M–L bonding drive ECE formation of Pd 2 The EXAFS data above establishes that the overall ECE sequence transforms a four-coordinate

II III square planar Pd ion into a Pd 2 species with octahedral coordination at each Pd center. Thus, the ECE sequence involves not only the M–M bond formation, but also the binding of two axial sulfate ligands, both of which would contribute to the dimerization free energies for PdII + PdIII and PdIII + PdIII. According to equations 2.8 and 2.9, for any given ΔGcomp, both ΔGdim,het and ΔGdim,hom scale proportionally with the degree of potential inversion, |E2 – E1|. Thus, the >200 mV potential inversion that we observed contributes to an increased favorability of dimerization. Since the sulfuric acid medium contains an abundance of ligating ions, we cannot directly measure the contribution of M–L bonding relative to M–M bonding in the observed potential inversion and dimerization free-energies. However, it is noteworthy that electrochemical oxidation

II of ligand-bridged Pd 2 precursors in non-coordinating electrolyte media typically proceeds via sequential oxidation without potential inversion.37,39–41 The addition of two-equivalents of Cl− transformed the two sequential 1-e– waves in the CV to a single 2-e– wave, indicating potential inversion and facile oxidation to

III 42,43 II III Pd 2. Likewise, Pd to Pd 2 oxidation is facile with chemical oxidants (e.g., PhICl2, PhI(OAc)2) which simultaneously transfer holes and ligands to the Pd centers. These observations suggest that, in addition to M–M bonding, axial ligation is also a critical factor for driving potential inversion, Pd dimerization, and facile oxidation.

The combined roles of M–M and M–L bonding in causing dimerization and potential inversion can be rationalized using the qualitative orbital diagram in Scheme 2.2 that illustrates the change in the energies of Pd dz2 orbitals with decreased M−M and M−L (L: axial ligand) distances. Without a significant

47 change in structure, electrostatic considerations dictate that each successive redox event of a given compound is more difficult than the previous. Thus, a potential inversion requires that bonding-induced

II changes in orbital energetics substantially overcome this electrostatic effect. Ligand-bridged Pd 2 complexes in the literature, whose electrochemistry does not exhibit potential inversion, display some M– M bonding character prior to oxidation.44 Thus, when these complexes are oxidized in non-coordinating media, both electrons are sourced from the M−M σ* orbital (green box) with only changes in the M–M bond distance contributing to attenuating the potential separation between E1 and E2. In contrast, in our

II system, the first electron is removed from the dz2 orbital of monomeric square-planar Pd (red box; see

II II,III 2.4.8.4 for evidence that Pd is monomeric), whereas the second electron is sourced from a putative Pd2 intermediate (blue box) that is coordinatively saturated. Thus, the combined effect of axial ligand binding and M–M bond formation that occurs as part of the C step drives up the energy of the dz2 orbital, leads to potential inversion, and enables the overall ECE oxidation. This analysis highlights the important role of the sulfuric acid media in facilitating rapid ligand binding towards electrochemical generation of a potent

III Pd dimer for methane C–H functionalization.

Scheme 2.2. A qualitative orbital energy diagram showing all Pd’s in the II oxidation state. Only the MOs of Pd dz2 parenthood are shown for Pd–Pd and L–Pd–Pd–L.

2.3. Conclusions

Using a combination of spectroscopic techniques, we have assembled structural models of the key high-valent Pd2 intermediates involved in rapid electrochemical methane functionalization and quantified the thermodynamics of spontaneous dimerization. X-ray absorption and Raman spectroscopies indicate that

48 the product of electrochemical PdII oxidation is a PdIII dimer with a Pd–Pd bond and a 5-fold O-atom

(x–2) coordination by HxSO4 at each Pd center. EPR spectroscopy establishes the presence of a minor-

II,III II equilibrium mixed-valent Pd2 intermediate in the ECE mechanism for Pd oxidation. The measured comproportionation free energies and redox potentials for key species in the oxidation sequence enabled quantification of negative free energies for Pd dimerization. The favorable dimerization free energy and large potential inversion arise from the combined effect of M–M and axial M–L bonding interactions that form during the C step of the ECE oxidation sequence. Together, these studies establish the core structure

III of Pd 2 as well as a structural basis for its unusual ECE formation mechanism, thereby, enabling the design of new electrocatalytic oxidation sequences mediated by high-valent Pd complexes.

2.4. Methods and Additional Information

2.4.1. Chemicals, Materials and General Remarks

PdSO4∙2H2O (Strem Chemicals), (NH4)2SO4 (99.999%, Alfa Aesar), concentrated sulfuric acid

(95-98%, EMD Millipore), fuming sulfuric acid (18–24% SO3, Alfa Aesar; 65% SO3, Sigma-Aldrich),

K2Pt(NO2)4 (Pt 42.6% min., Alfa Aesar), CuSO4∙5H2O (99.999%, Strem Chemicals), and Fluorine-doped tin oxide (FTO) coated glass slides (Sigma Aldrich, TEC 7, surface resistivity ~7 Ω/sq) were purchased from the respective vendors. Water was reagent grade (Millipore Type 1, 18.2 MΩ cm resistivities). For the

II preparation of Pd SO4 solutions, 20 mL reaction vials (Chemglass, CG-4904-01) with an additional PTFE tape layer (Slic-tite PTFE thread tape, 1” wide) inside the PTFE-lined septum cap were used. All concentrated or fuming sulfuric acid samples were contacted with glass or PTFE only except for the electrodes.

Electrochemistry was performed with a Biologic VMP3 potentiostat. For electrical connections, Pt wire was used because of its corrosion resistance. Two strands of thin Pt wire (Alfa Aesar 00263; ≥99.9% (metals basis), dia. 0.127 mm) were twisted together to confer robustness and flexibility. A Pt wire pseudo-

1 reference electrode or a saturated Ag2SO4/Ag reference electrode (SSE) was used for potential measurements, and all potentials in this work are shown against the latter.

UV–Vis measurements were performed on an Agilent Cary 50 instrument using a quartz cuvette of 1 mm pathlength. All UV–Vis spectra were acquired immediately after diluting to ~5 vol% in 95–98%

H2SO4 (typically measured background-subtracted absorbance = 0.1 – 0.6). Fuming H2SO4 could not be

II III used as the diluent as it absorbed in the UV region. Pd shows a strong peak at 230 nm whereas Pd 2 shows two strong peaks at 230 and 300 nm. The exact λmax blue-shifted with increasing ox.%, but since the shift

49 was <5 nm, for the sake of simplicity, we refer to the two peaks as 230 nm and 300 nm peaks. Details of quantitative analysis by UV–Vis are described below.

Details of X-ray absorption, Raman and EPR spectroscopies are also below in their respective sections.

2.4.2. Preparation of samples for X-ray absorption and Raman spectroscopy

2.4.2.1. Preparation of 1-lc

PdSO4∙2H2O was dissolved in fuming H2SO4 (18–24% SO3) at 9 mM by vigorous stirring for 1−2 h at 130 ˚C until the solution was clear. The solution was prepared in a 20 mL glass reaction vial with a PTFE-lined cap with an additional layer of PTFE tape and a PTFE-coated stir bar.

2.4.2.2. Preparation of 2-lc As reported previously,1 1-lc was subjected to ~2 days of bulk electrolysis with an FTO working electrode, a Pt wire pseudo-reference electrode, and a Pt mesh counter electrode. Degree of conversion was assessed by UV−Vis spectroscopy by diluting an aliquot in concentrated H2SO4 (95–98%; non-fuming) (Figure 2.11b; see below for more technical details of UV−Vis measurements).1

2.4.2.3. Preparation of 1-hc

1-hc was prepared similarly as 1-lc with the addition of (NH4)2SO4 and a larger amount of PdSO4.

In a representative preparation, 104 mg of PdSO4∙2H2O and 1.7 g of (NH4)2SO4 were added to 8 mL of 20% fuming H2SO4. The mixture was dissolved by vigorous stirring for 1−2 h at 130 ˚C in a 20 mL glass reaction vial with a PTFE-lined cap with an additional layer of PTFE tape. The solution volume increased to 9.3 mL because the density of the (NH4)2SO4–fuming H2SO4 mixture was slightly smaller than that of pure 18–24%

SO3 fuming sulfuric acid (1.83 vs 1.92 g/mL). The mixture was metastable and over long periods of time

(NH4)2SO4 crystals precipitated out. These precipitates could be redissolved by heating and stirring.

2.4.2.4. Preparation of 2-hc 1-hc was subjected to bulk electrolysis at room temperature as 2-lc, but procedures were slightly modified due to the longer time duration required for the electrolysis. Presumably due to the high viscosity of the solvent and the slow diffusion of Pd ions, less current could be passed with 1-hc than with 1-lc (Figure 2.8). In a two-electrode configuration with an FTO working electrode and a Pt mesh counter electrode (identical to the cell for 2-lc except for the reference electrode), a constant current density of 80 μA/cm2 was applied for 4.5 days under constant stirring. The reference electrode was omitted for fear of potential drift of the Pt pseudo-reference electrode. In the absence of the reference electrode, galvanostatic

50 electrolysis was chosen to better control the reaction. The potential never exceeded 2.5 V. As in the preparation of 2-lc, the electrolysis was terminated after checking the UV−Vis spectra of diluted aliquots.

Figure 2.8. Cyclic voltammograms of (a) 1-hc and (b) 1-lc, for which the PdII concentration was 47 mM and 9 mM, respectively. The similarity in current density despite the higher PdII concentration of 1-hc implies much slower mass transport for the more viscous 1-hc sample. For both samples, the characteristic hysteresis (anodic current larger on the return scan) can be seen, which implies an ECE mechanism (sequential electron transfer-chemical reaction-electron transfer) and formation of the high-valent Pd species.1

III 2.4.2.5. Preparation of Pt 2

III 26 Pt 2 was prepared according to a literature procedure for preparing K2[Pt2(SO4)4(H2O)2]. 1.0 g of K2Pt(NO2)4 was mixed with 10 mL of ~9 M aqueous H2SO4 in a round bottom flask attached to a reflux condenser, and heated to 105 °C under N2 atmosphere in an oil bath. The solution, initially deep blue, transforms to green when heated. Subsequently, in the course of 2−5 h, the solution turns brown and yellow product precipitates from the solution. The amount of O2 was critical to the success of this reaction. While exposure to ambient air destroys the reaction, the reaction proceeded extremely slowly under a rigorous N2

45 atmosphere. To allow this subtle control of O2 amount, a venting needle was inserted to a rubber septum capping the system. After a sufficient amount of the yellow precipitates formed, the reaction flask was removed from the oil bath and diluted with 10−20 mL of water. The solution was filtered through a porosity

D glass filter funnel and rinsed with 5−10 mL of water. The dark yellow filtrate was heated for ~3 h more and filtered again. The filtered precipitates were rinsed with 5−10 mL each of water, acetone, and ether and dried in vacuo to yield a yellow powder in 43% yield.

III 2.4.2.6. Stability of the Pd 2 species

III II Pd 2 is a metastable species that spontaneously reduces back to Pd , presumably by oxidizing water in the sulfuric acid medium. This necessitated taking UV–Vis spectra immediately after diluting an

III aliquot for assessing Pd 2 concentration in samples (Figure 2.9). We have found that this decay is significantly slowed in the presence of SO3, which is the anhydrous form of H2SO4 (=H2O + SO3). In order

III to ensure that Pd 2 persisted throughout various measurements, its stability was assessed through UV–Vis

III spectroscopy. For 2-lc and 2-hc kept in a tightly closed vial, the Pd 2 concentration assessed by UV–Vis spectroscopy decreased by only 11% and 7% after 8 days at room temperature, respectively.

Figure 2.9. UV–Vis spectra of 2-lc diluted in concentrated H2SO4 and recorded over time.

2.4.2.7. Verification that 2-lc and 2-hc samples are chemically identical

III In order to verify that the Pd 2 species prepared at different concentrations with and without

(NH4)2SO4 (2-hc and 2-lc) were chemically identical, we conducted the following studies.

First, cyclic voltammograms (CV) of PdSO4 in fuming sulfuric acid with and without (NH4)2SO4

II III 1 (1-hc and 1-lc) were very similar, displaying key features associated Pd oxidation to Pd 2 (Figure 2.8). In brief, the characteristic hysteresis (anodic current larger on the return scan) can be seen in both samples, which implies an ECE mechanism (sequential electron transfer-chemical reaction-electron transfer) and formation of the high-valent Pd species: PdII is first oxidized to PdIII (E), then dimerizes with another PdII

III (C), and finally oxidized once more (E) to generate a Pd 2 species. The hysteresis is most pronounced at scan rates where the E and C steps have comparable kinetics. Differences in CV shapes in the four panels of Figure 2.8, therefore, arise from differences in overall Pd concentration and solution viscosity that change the rates of E and C steps. The four CVs together indicate that PdII oxidation proceeds through the same mechanism in 1-lc and 1-hc. 52

Second, diluted aliquots of 2-hc and 2-lc showed identical UV−Vis spectra (Figure 2.10).

Figure 2.10. UV–Vis spectra of 2-hc and 2-lc diluted in concentrated H2SO4. The spectra are normalized by the absorbance of the 230 nm peak.

Third, when 2-hc was subjected to 500 psi of methane at 100 ˚C for 40 min., stoichiometric

III amounts of methyl bisulfate (CH3OSO3H) was generated (Table 2.2), as observed with Pd 2 in the absence

1 of (NH4)2SO4. Interestingly, compared to 2-lc, 2-hc showed significantly suppressed formation of methanesulfonic acid (CH3SO3H), which is a non-stoichiometric product formed alongside methyl bisulfate.

Since the rate of methanesulfonic acid formation is correlated with the concentration of SO3 (Chapter 3),

− − we postulate that the high concentration of added sulfate salt reacts with free SO3 (SO3 + HSO4 ⇄ HS2O7 ) to lower the concentration of free SO3 and suppress CH3SO3H formation. Alternatively, the increased basicity may have impacted the methane oxidation reaction.

Lastly, the X-ray absorption spectra of 1 and 2 in the lc and hc samples looked identical (Figure 2.13–Figure 2.16).

Table 2.2. Products from the reaction of 2-hc with methane. 500 psi CH4, 100 ˚C, 40 min.

Concentrations [Pd] 45 mM a [CH3OSO3H] 23 mM a [CH3SO3H] 1.4 mM aQuantified by the addition of 5 μL of acetic acid standard into 1 mL of the reaction solution.

2.4.3. Preparation of samples for EPR spectroscopy

2.4.3.1. Solvent purification

(Caution! SO3 is an extremely corrosive gas with a high vapor pressure. The distillation procedure involves slight pressure build-up inside the apparatus. Even with tight sealing with PTFE tape, small leakage of SO3 through the joints may occur. Never heat solid SO3 (without the addition of 95–98% sulfuric

53 acid) in a closed glass vessel as it can result in a sudden build-up of pressure and shattering of the glass vessel.) For EPR spectroscopy, high levels of EPR-active transition metal impurities in the commercial fuming sulfuric acid (Alfa Aesar 45543, lot no. Z22C001; Fe 12.035 ppm) led us to prepare clean fuming sulfuric acid without these impurities. We mixed 95–98% sulfuric acid (EMD Millipore SX1244, lot no.

58346; <5 ppb each for all specific metal impurities tested and <1 ppm for total heavy metals) and SO3 obtained from distillation. 100% SO3 purchased in a lecture bottle contained unknown stabilizers that

III seemed to prevent the formation of the Pd dimer. SO3 was distilled from 65% SO3 oleum using a conventional glass distillation apparatus (two round bottom flasks for sending/receiving and a water- jacketed distillation arm) with joints sealed by PTFE tape. SO3 started to distill out at ~80 ̊C, and the temperature was increased over time as SO3 concentration decreased in the sending flask. The distillation arm temperature had to be controlled carefully with cooling water and a heat gun because the boiling point and melting point of SO3 are very close to each other such that solid SO3 can clog up the distillation arm if it is too cold. The receiving flask was submerged in an ice bath. After finishing the collection of solid SO3,

95–98% sulfuric acid was added carefully and stirred until all solid SO3 dissolved. Some polymeric forms of SO3 caused by the presence of trace water took a long time to dissolve, and sometimes required heating the solution up to 60–70 ̊C. The clean oleum prepared this way contains SO2 because SO2 is present in the commercial 65% oleum and distills alongside SO3, and their presence was confirmed by UV–Vis spectroscopy. However, they disappeared after bulk electrolysis, indicating that they were oxidized to SO3.

PdII solutions prepared in the clean fuming sulfuric acid gave identical cyclic voltammograms

III characteristic of Pd 2 formation. UV–Vis spectroscopy and methane oxidation reactivity of bulk- electrolyzed solutions were also identical to what was observed in the regular solvent. The thermal stability

III of Pd 2 in the clean solvent was significantly enhanced compared to that in the regular solvent.

2.4.3.2. Preparation of variable ox.% samples Solutions of variable ox.% (different ratios of II and III oxidation states of Pd; i.e., different ratios

II III [ ] of [Pd ] and [Pd 2]; ox.%≡ ) were prepared either by carrying out partial electrolysis or by [ ][ ] III II III mixing fully bulk-electrolyzed Pd 2 solutions with Pd solutions prepared from heat-quenching Pd 2

III II II solutions (so as to reduce Pd 2 to Pd ). Pd solution that was not bulk-electrolyzed contained substantial

III amounts of SO2 co-distilled during solvent purification, and caused the reduction of Pd 2 upon mixing. At least 3 h was given after sample preparation before freezing the samples. Samples frozen soon after the termination of electrolysis or mixing showed a large deviation of the spin concentration from a consistent trend. Waiting longer (up to overnight) made little difference.

For these solutions, the concentration of PdIII will monotonically increase with increasing ox.%,

II,III while that of Pd2 will show a maximum at an intermediate ox.%. The latter is what we observe

II,III experimentally, and the concentration of Pd2 calculated according to the comproportionation equilibrium (equation 2.6 and 2.7) showed a reasonably good fit with experimental values (Figure 2.7b). The deviation likely arises from the (i) approximation used in the calculation as well as the (ii) uncertainty in the experimental determination of ox.%. To describe in detail,

II,III II III (1) In the calculation of [Pd2 ] using equation 2.7, [Pd ] and [Pd 2] were calculated as

II,III [Pd]total⨯(1 – ox.%) and 0.5⨯[Pd]total⨯ox.%, respectively. However, this assumes [Pd2 ] is

II III II III negligible compared to [Pd ] + 2 [Pd 2], so that [Pd ] + 2 [Pd 2] = [Pd]total. The actually

II,III measured concentration of [Pd2 ] was up to 0.2 mM, which corresponds to 2.2% of [Pd]total = 9.3 mM.

II III (2) Likewise, experimental determination of [Pd ] and [Pd 2] rely on UV–Vis measurements that

II,III II III ignore the contribution of Pd2 . The extinction coefficients for Pd and Pd 2 were

II III determined from the spectra at the limit of Pd = 100% and Pd 2 ~ 100% (see 2.4.4), and ox.% is determined from interpolation. There are also uncertainties in the extinction coefficient for

III III Pd 2 due to the difficulty of achieving absolutely 100% of Pd 2, as noted earlier.

II III 2.4.4. Determination of [Pd ] and [Pd 2] from UV–Vis spectroscopy

II III Although the UV–Vis spectra and extinction coefficients of Pd and Pd 2 have been reported in our previous work,1 we refined our measurements and updated the values in this work for more accurate quantitation, particularly for the quantitative EPR measurements at variable ox.% (Figure 2.7b).

First, for the extinction coefficient of PdII at 230 nm, we determined the Pd concentration accurately by precise dilution of a stock PdII solution whose concentration was determined by ICP–MS. To ensure complete dissolution of PdII ions in the stock solution, the stock solution was prepared at a relatively low concentration (~3 mM) and centrifuged to remove any solid residues. Precise dilution of the stock solution was achieved by measuring the weights of an empty vial, the vial after addition of 95–98% H2SO4, and the vial after addition of the PdII stock solution and calculating the dilution ratio by weight. The resultant Beer’s plot is shown in Figure 2.11a.

III Second, for the extinction coefficient of Pd 2, we could not construct an accurate Beer’s plot as

II III with Pd because the spectra of Pd 2 had to be measured immediately following dilution. The spontaneous

III II decay of Pd 2 back to Pd was greatly accelerated after dilution (Figure 2.9), and the protocol for precise

III dilution described above prevented immediate measurement of the spectrum. Instead, a spectrum of Pd 2

III was measured immediately following dilution to obtain a spectrum of Pd 2 at an unknown concentration. 55

III II Then, this solution was heated to 80–100 ̊C until all Pd 2 was quenched to return to Pd ions by presumably oxidizing residual water in the solvent. The spectrum of the quenched solution, with the extinction

II III coefficient of Pd , gave the total Pd concentration. From this, the extinction coefficient of Pd 2 could be obtained, and the molar extinction plot is shown in Figure 2.11b. Extinction coefficients are provided in Table X.

II Figure 2.11. (a) Beer’s plot of Pd in 95–98% H2SO4. The y-axis shows background subtracted absorbances II III measured in a 1 mm-pathlength cuvette at 230 nm. (b) UV–Vis extinction spectra of Pd (black) and Pd 2 (red). The extinction coefficient is given as per Pd, not per dimer.

II III –1 –1 Table 2.3. Molar extinction coefficients, per Pd atom, for Pd and Pd 2 in M cm .

ε230nm ε300nm PdII 11016 69

III Pd (2) 13941 15507

[ ] In order to determine the ox.%(≡ ) of a mixed solution, the following equation was [ ][ ] employed, where A denotes the background subtracted sample absorbance at a given wavelength:

퐴 휀 −휀 , 퐴 , Pd%= 퐴 휀, −휀, − 휀, −휀, 퐴

This equation was derived from the following relationships:

퐴 = [Pd]{휀,(1−ox.%) +휀,(ox. %)}

II III Incidentally, we note that achieving absolutely 100% conversion of Pd to Pd 2 by bulk electrolysis is difficult. It was especially more difficult in the initial phase of our studies when we used

III fuming sulfuric acid containing transition metal impurities that decreased the lifetime of Pd 2 and an electrochemical cell design that had inferior protection from ambient moisture. Rigorously speaking, samples of 2-lc and 2-hc used for X-ray absorption and Raman spectroscopies were prepared at ox.% > 90%, though this does not affect the conclusion of our work.

2.4.5. X-ray Absorption Spectroscopy

2.4.5.1. Methods

Pd K (24.350 keV) and Pt LIII (11.564 keV) edge X-ray absorption measurements were performed on the bending magnet beamline of the Materials Research Collaborative Access Team (MRCAT) at the Advanced Photon Source, Argonne National Laboratory. Measurements were acquired in step-scan transmission mode, and data were collected in three regions: a pre-edge region, the XANES region, and the EXAFS region (see detailed information below). A reference Pd foil spectrum was collected simultaneously with each sample for energy calibrations.

Solid samples were diluted with boron nitride powder and pressed into self-supporting wafers in a cylindrical sample holder containing six wells and placed in a quartz reactor tube (1-in. ID, 10-in. length) sealed with Kapton windows by two Ultra-Torr fittings. Liquid samples prepared as described above were shipped frozen in dry ice, thawed on site, and then loaded into a custom-made PTFE cell with 1 mm-thick PTFE windows on each side.

XAS spectra were analyzed using the Demeter software suite.46 Spectra were normalized with linear and cubic fits through the pre-edge and post-edge regions, respectively. This method generated the XANES spectra shown in Figure 2.13. The EXAFS were extracted by performing cubic spline fits of normalized spectra from 0.5 to 12.9 Å−1 for all samples. The k2-weighted Fourier transform of the entire 0.5 to 12.9 Å−1 spectral range generated the EXAFS spectra shown in Figure 2.16. Coordination parameters were obtained by simultaneous fits in R-space (ΔR = 1.13 – 1.91 Å) of the magnitude of the Fourier transformed k1, k2, and k3-weighted EXAFS. Theoretical phase shift and back-scattering amplitude fitting

47 2 functions were calculated using FEFF. Amplitude reduction factors (S0 ) for Pd (0.84) and Pt (0.78) were determined from the metal foils.

Acquisition details:

Pd K Edge: Pre-edge region (−250 to −50 eV, step size = 10 eV, dwell time = 0.25 s; −50 to −40 eV, step size = 5 eV, dwell time = 0.25 s), XANES region (−40 to 30 eV, step size = 0.5 eV, dwell time = 0.25 s) and EXAFS region (0 to 8 Å−1, step size = 0.05 Å−1, dwell time = 0.5 s; 8 to 12 Å−1, step size = 0.05 Å−1, dwell time = 1.0 s; 12 to 13 Å−1, step size = 0.05 Å−1, dwell time = 1.5 s)

Pt LIII Edge: Pre-edge region (−200 to −50 eV, step size = 10 eV, dwell time = 0.2 s; −50 to −10 eV, step size = 5 eV, dwell time = 0.2 s), XANES region (−10 to 30 eV, step size = 0.5 eV, dwell time = 0.3 s) and EXAFS region (0 to 8 Å−1, step size = 0.05 Å−1, dwell time = 0.5 s; 8 to 12 Å−1, step size = 0.05 Å−1, dwell time = 0.5 s; 12 to 13 Å−1, step size = 0.05 Å−1, dwell time = 0.5 s)

2.4.5.2. Comparison of hc and lc samples The low concentration of Pd in samples 1-lc and 2-lc resulted in low transmission edge-steps (i.e., the increase in absorption due to the Pd K edge) of ~0.05. Normalization of the spectra with such low edge- steps is affected by the non-linearity of the background absorption by non-Pd species in the beam path (e.g., cell walls, solvent, air). The curvature (i.e., non-linearity) of the raw absorption spectrum over the measured energy range is large enough in relation to the edge-step that the linear fit of the pre-edge region, which is subtracted from a polynomial fit of the post-edge region for normalization, does not adequately capture this background absorption. When comparing the lc and hc samples, the differences in normalization cause apparent differences in the XANES (increased intensity before the edge and decreased white line intensity for lc relative to hc; Figure 2.13). Furthermore, this deviation from approximate parallelism of the pre and post-edge fits resulted in the EXAFS of the un-flattened normalized spectra to slope upwards from unity at the edge (Figure 2.12). The upward slope causes an overestimation of the atomic absorption magnitude, with a larger effect at higher k, when performing the cubic-spline fit for extraction of the EXAFS. This leads to an apparent k-dependent suppression in the EXAFS magnitude (Figure 2.15). In accordance with

II this effect, when fitting the EXAFS for Pd SO4, whose Pd–O coordination number is known as 4, a larger

II Debye-Waller factor was required for the spectrum of 1-lc compared to those of 1-hc or solid Pd SO4. Therefore, when fitting samples 2-hc and 2-lc, it was assumed there would be approximately the same concentration effect on the EXAFS as was observed in samples 1-hc and 1-lc. For consistency between samples, the difference in the Debye-Waller factor obtained from the fits of the pre-catalysts was fixed when fitting samples 2-hc and 2-lc. Because of the difference in the Debye-Waller factor, the Pd–O coordination numbers obtained from the hc and lc spectra were similar in spite of the difference in the apparent peak height of the Pd–O scatter in the Fourier-transformed EXAFS (Figure 2.16).

2.4.5.3. Edge shape analysis of 1 and 2 The effects of coordination geometry on K-edge XANES have been well-studied for the 3d transition metals.14,15,48 Changes in edge energies have been associated with oxidation state and metal-ligand bond covalency49 while the intensity of edge and pre-edge features have been related to coordination geometry.48 Similar trends should be observed in the K-edge XANES of the 4d metals, since these spectra also arise from the excitation of a 1s electron to valence p-states, albeit with decreased resolution due to the increased lifetime broadening. The shoulder observed in the edge of compound 1 (Figure 2.13a; better seen

58 in Figure 2.14a) likely arises from the transition of a 1s electron to a non-bonding 5pz orbital, similar to what has been described for 3d metal compounds with square-planar geometries.50 The smooth edge structure of compound 2 (Figure 2.13b; better seen in Figure 2.14b) suggests the degeneracy of p orbitals caused by a change in the geometry from square-planar to octahedral.

Figure 2.12. Unflattened versions of the normalized XAS spectra (solid lines) and normalization II background (dashed lines) of 1-lc, 1-hc, and solid Pd SO4.

Figure 2.13. Comparison of the XANES of hc (red) and lc (black) samples of (a) 1 and (b) 2. With 1, the II spectrum of solid Pd SO4 is also overlaid (light blue).

Figure 2.14. Comparison of the first derivative of XANES of hc (red) and lc (black) samples of (a) 1 and II (b) 2. With 1, the spectrum of solid Pd SO4 is also overlaid (light blue).

Figure 2.15. Comparison of the k2-weighted k-space EXAFS of hc (red) and lc (black) samples of (a) 1 and II (b) 2. With 1, the spectrum of solid Pd SO4 is also overlaid (light blue).

Figure 2.16. Comparison of the R-space (Fourier-transformed) EXAFS of hc (red) and lc (black) samples II of (a) 1 and (b) 2. With 1, the spectrum of solid Pd SO4 is also overlaid (light blue).

2.4.6. Raman Spectroscopy

2.4.6.1. Methods Raman spectra were acquired on a Horiba Jobin-Yvon LabRAM Raman Confocal Microscope using a 532 nm excitation laser and 10× objective lens. A notch filter was used to remove Rayleigh scattering, and the spectrometer was calibrated with a Si crystal each time before use. An 1800 grooves/mm grating was used with a hole size of 500 μm and a slit width of 100 μm. The liquid sample was contained in a 1 mm-pathlength quartz cuvette and sealed with PTFE tape. The resultant raw spectra, after manual removal of random spikes from instrument noise, were plotted as-is without background correction. The background spectra of the solvent (Figure 2.5a) are presented alongside the sample spectra. For polarization measurements (Figure 2.18, also see below), a plane-polarizing analyzer filter was inserted before the detector at two different angles, one being perpendicular to the linearly polarized incident laser and the other being parallel to the incident laser.

2.4.6.2. Determination of Raman mode symmetry from depolarization ratios Depolarization ratio (ρ) of a Raman scattering peak reflects the symmetry of the vibration mode. An M–M vibration in a symmetric metal dimer would be totally symmetric. ρ is defined as51

퐼 ρ= 퐼∥

where 퐼 denotes the intensity of scattered light polarized perpendicular to the incident light

(which is itself linearly polarized), and 퐼∥ denotes the intensity of scattered light polarized parallel to the incident light. Theoretically, ρ should be 0.75 for non-totally symmetric vibrations and below 0.75 for totally symmetric vibrations.51 While a quantitative assessment of ρ requires careful calibration of the instrument,51 a qualitative determination of the symmetry of a certain vibration may be achieved from a simple comparison of relative magnitudes.32 Since ρ is either 0.75 or below 0.75 depending on the symmetry of the vibration mode, a peak whose value of ρ is significantly smaller than that of other peaks is indicative of a value below 0.75 and therefore corresponds to a totally symmetric vibration.

III Based on the above, the raw perpendicular- and parallel-polarized Raman spectra of Pt 2 and 2 were scaled so that the peak arising from the solvent would display a ρ = 0.75 (Figure 2.18; see Figure

III –1 2.5a for the solvent spectra). The Raman peak of Pt 2 at 227 cm was already assigned to a Pt–Pt vibration based on axial ligand substitution in an aqueous solution (explained in 2.2.1.3; also Figure 2.17). Following

–1 III the relative scaling of spectra, the value of ρ for the 227 cm peak of Pt 2 was calculated to be ca. 0.4, in agreement with the assignment to an M–M vibration. The Raman peak at 268 cm–1 of 2 also displayed a ρ

≈ 0.4, lending support to the hypothesis that the 268 cm–1 peak of 2 likewise arises from a symmetric M– M stretch.

III Figure 2.17. Time-dependent evolution of the Raman spectra of Pt 2 in 1 M H2SO4 after adding 20 mM of III NaCl to 10 mM of Pt 2. With time, peak a remains relatively unchanged, while peak b diminishes and peak c grows in magnitude. Therefore, peak b and c are assigned to vibrational modes in the original and the Cl−- III − the original Pt 2 complex, respectively. From the literature, we know that Cl substitution occurs at the axial position. Along with the low wavenumber of these peaks, we assign peaks b and c to a Pt–Pt vibration.

III Figure 2.18. Perpendicular (red) and parallel (blue)-polarized Raman spectra of Pt 2 (top) and 2-hc (bottom).

2.4.7. Electron Paramagnetic Resonance (EPR) spectroscopy

2.4.7.1. Methods EPR spectra were acquired on a Bruker EMX-Plus spectrometer in X-band, perpendicular mode. Samples were frozen in a 90:10 ethylene glycol:ethanol cooling bath, and subsequently transferred to liquid

N2 before inserting into the spectrometer. To avoid breaking the quartz wall of the tubes due to volume expansion upon freezing, the EPR sample tubes were immersed in the cooling bath and put into a freezer at ca. –25 ̊C, in which both the sulfuric acid solution and the cooling bath froze. The tubes could be pulled out after a little bit of thawing thanks to the low friction between glass and the alcoholic ice. Because of inherent transition metal impurities in the commercially available fuming sulfuric acid, the solvent for EPR measurements was prepared by mixing high-purity concentrated sulfuric acid with distilled SO3 (details above). Background EPR spectrum of the solvent is shown in Figure 2.19. Cyclic voltammetry as well as UV–Vis spectroscopy and methane oxidation reactivity of the bulk-electrolyzed solution were identical in the clean solvent as in the regular solvent.

2.4.7.2. Quantitation

II For quantitative EPR measurements, Cu SO4 was dissolved in the same fuming sulfuric acid medium used for Pd sample preparations (spectrum shown in Figure 2.19). To diagnose saturation effects, variable power and variable temperature (20–100 K) measurements were also conducted (Figure 2.20). If not specified otherwise, all spectra were acquired at 0.05024 mW and 60 K. Double integration was performed with polynomial background correction with the EPR acquisition software, Xenon (Bruker). To minimize baseline drift during the integration, the integration region for each spectrum was selected so that the single integral value is close to zero. The double integral value of the CuII standard was divided by the concentration of Cu determined by ICP–MS, and this value was used for converting double integrals into concentrations for our Pd-based radical, which is valid because it is also a S=½ species.52,53

Figure 2.19. Background-uncorrected EPR spectra of (a) commercial fuming sulfuric acid, (b) clean fuming II II sulfuric acid obtained by distillation of SO3, (c) the Cu spin quantitation standard, (d) 1 or Pd , and (e, f) Pd solutions at ox.% = 48% and 95%. The spectrum (f) can be seen larger in Figure 2.7a with background subtraction. All spectra were acquired at 60 K, 0.05024 mW.

II II,III Figure 2.20. Double-integrated EPR signal intensity of Cu and Pd2 at (a, b) various microwave power levels at 60 K and (c) various temperatures at 0.05024 mW (y-axis values for the two metals are normalized to the value at 60 K). The blue lines show the value of microwave power and temperature (0.05024 mW, 60 K) that were selected for the spin quantitation experiments in this study.

2.4.8. Determination of Thermodynamic Quantities

2.4.8.1. Mathematical derivation of relationships between the quantities In this section, we show how the different thermodynamic quantities (redox potentials and ΔG values) are related to each other. In short, all thermodynamic quantities can be determined from E1, E2 and

ΔGcomp, which were experimentally estimated in this work. (note: we omit the superscript 0 from all redox potentials and Gibbs free energies of reaction for simplicity.)

To write the relevant equations (see Scheme 2.1, for a summary diagram),

Pd + e ⇄ Pd ∆퐺 = −퐹퐸 … 2.10 , Pd + Pd ⇄ Pd ∆퐺, … 2.11 , Pd + e ⇄ Pd ∆퐺 = −퐹퐸 … 2.12 Subtracting equation 2.10 from equation 2.12,

, Pd + Pd ⇄ Pd + Pd ∆퐺 = −퐹(퐸 − 퐸) … 2.13 Adding equation 2.11 and equation 2.13,

, 2 Pd + Pd ⇄ 2 Pd ∆퐺 = ∆퐺, + ∆퐺 … 2.14

→ ∆퐺, = ∆퐺 + 퐹(퐸 − 퐸) … 2.15 Subtracting equation 2.13 from equation 2.11,

2 Pd ⇄ Pd ∆퐺, = ∆퐺, + 퐹(퐸 − 퐸) … 2.16

→ ∆퐺, = ∆퐺 + 2퐹(퐸 − 퐸) … 2.17 Additionally, we can obtain the redox potentials for other pairs of Pd species:

, Pd + e ⇄ 2 Pd ∆퐺 = −퐹퐸 … 2.18

∆퐺 = ∆퐺 − ∆퐺, … 2.19 Pd + 2 e ⇄ 2 Pd ∆퐺 = −2퐹퐸 … 2.20

∆퐺 = ∆퐺 + ∆퐺 − ∆퐺, … 2.21

The value of E4 is important as it will determine the open-circuit potential of a Pd solution at a given ox.% according to the Nernst equation. Substituting equation 2.15 into equation 2.21,

∆퐺 = ∆퐺 + ∆퐺 − ∆퐺 + ∆퐺 − ∆퐺 = 2∆퐺 − ∆퐺 … 2.22

∆퐺 퐸 = 퐸 + … 2.23 2퐹

The Nernst equation used for obtaining E4 from the measured open circuit potential (EOCP):

푅푇 [Pd ] 퐸 = 퐸 + ln ( ) … 2.24 2퐹 [Pd] Finally, in the calculation of the thermodynamic quantities, the standard error was obtained by propagating the error in each constituent term.

2.4.8.2. Estimation of E1

The value of E1 was estimated from cyclic voltammograms (CVs) recorded in a dilute solution of PdII (0.6 mM) in fuming sulfuric acid (Figure 2.21a). A low Pd concentration was used in order to minimize convolution from the follow-up dimerization reaction. Due to the slow electron transfer kinetics on FTO electrodes, a Pt wire electrode (0.314 cm2) was used as the working electrode. The CVs at varying scan rates showed a reversible redox feature around ~1.7 V vs SSE, but with decreasing scan rate, the magnitude of the cathodic return wave relative to the anodic wave decreased. This is due to the competition of the 65

II III II,III III heterodimerization reaction (i.e., Pd + Pd → Pd2 ) with the electrochemical reduction of Pd that becomes more significant as the scan rate is decreased. Indeed, at lower scan rates, a second reduction peak

III that corresponds to Pd 2 reduction grows in at ~1.2 V vs SSE. Since a chemical reaction that accompanies electron transfer can shift the apparent redox potential (midpoint potential Em: the average of the anodic

0 and cathodic peak potentials, ≈E ) from the true redox potential, we determined Em of the CV obtained at the highest scan rate, 2 V s−1, as 1.69 V vs SSE.

III This value, however, is a lower limit for E1 because our data imply adsorption of Pd ions on the Pt electrode surface that negatively shifts the oxidation potential. We give two reasons: first, the cathodic peak shape resembles that of a surface-adsorbed species, not a freely diffusing solution phase species (Figure 2.21a). The cathodic peak currents at different scan rates also show a linear correlation with the scan rate, rather than the square root of the scan rate, as would be expected for a surface-adsorbed species (Figure 2.21b). The deviation from a straight line at lower scan rates is due to the convoluting follow-up dimerization reaction, and plotting the peak currents against the square root of scan rate results in a much larger deviation from linearity. Second, the CV on an FTO electrode shows an Em that is clearly higher than that observed on a Pt electrode (Figure 2.21c). Although the midpoint potential does not exactly equal the thermodynamic redox potential due to, e.g., asymmetric transfer coefficients, the large positive shift of the FTO CV compared to that of Pt CV strongly implies different redox potentials for the PdII/III couple on Pt and FTO. Preferential adsorption of PdIII ions over PdII ions on the Pt electrode would give rise to a lower

II/III potential for Pd oxidation and explain the observed Em difference. Since the broad peaks on an FTO electrode, along with the convolution from follow-up chemical reactions, make it difficult to measure Em, we report a lower limit for E1 based on our measurements on a Pt electrode.

Additionally, if E1 is equal to 1.69 V, the lower limit value measured on a Pt electrode, the

III II,III concentration of Pd is calculated to be as high as the concentration of Pd2 at high ox.% (Figure 2.22), which does not agree with EPR observations. The EPR signal of the monomeric PdIII ion in sulfuric acid, which is expected to have a symmetric coordination environment with a small Jahn-Teller distortion, would

II,III 38 be less anisotropic than Pd2 . The similarity of the EPR spectra at ox.% = 95% and 48% (Figure 2.19) indicate the presence of a single EPR-active species with an anisotropic EPR signal at all ox.%.

II,III III Figure 2.22. Concentrations of the Pd2 and Pd intermediates calculated using ΔGcomp = 0.15, E1 = 1.69 V, and E2 = 1.49 V.

2.4.8.3. Estimation of E2

As shown in equation 2.23, E2 may be calculated from E4, which may be measured from the open-

II III circuit potential (OCP) of an electrode equilibrated with a solution containing both Pd and Pd 2. A platinum working electrode was immersed solutions of different ox.%, and its potential was measured against SSE after stabilization (~5 min). The ox.% of each solution was measured by UV–Vis spectroscopy immediately following the OCP measurement. Then, E4 was obtained from equation 2.24. Measurements of E4 according to this method across different ox.% shows a consistent value at 1.492±0.006 V vs SSE

(Figure 2.23). Therefore, using equation 2.23, E2 = 1.489±0.006 V vs SSE.

2.4.8.4. Ruling out pre-dimerization of PdII Since sulfate is a bidentate ligand and PdII is known to exist as dimers and trimers in acetic acid,54 it may be argued that PdII in sulfuric acid is in equilibrium with dimers or multimers prior to oxidation. In

II,III * that case, the ECE mechanism may be written as below (equations 2.25–2.28). The intermediate (Pd2 )

II,III refers to a species that undergoes chemical conversion to another form of Pd2 to be oxidized further. Importantly, since dimerization occurs before the ECE sequence, the chemical step (equation 2.27) is no longer bimolecular, but unimolecular. As such, changing the concentration of PdII does not impact the rate

II of the C step relative to the E steps, but only the effective concentration of the starting material, Pd 2. CV simulations (details below), accordingly, show that the ratio between the first and second reduction peaks in the cathodic scan does not vary with [PdII] (Figure 2.24), in contradiction to our experimental CVs collected with variable [PdII].1

Pre-dimerization 2 Pd ⇄ Pd … 2.25 E , ∗ Pd ⇄ Pd + e … 2.26 C , ∗ , Pd ⇄ Pd … 2.27

, E Pd + e ⇄ Pd … 2.28

Figure 2.24. Simulated CVs of PdII oxidation. The currents are scaled to match the oxidation peak current. When PdII is oxidized after dimerization, the ratio of the two cathodic peaks does not change.

*Details of the CV simulation:

We carried out a finite element simulation using the CV simulation module in DigiElch by ElchSoft. All electron transfer steps are modeled with Butler-Volmer kinetics. Monomeric Pd species and dimeric Pd species were assigned diffusion coefficients of 1⨯10–7 cm2 s–1 and 5⨯10–8 cm2 s–1, respectively.

- The mechanism for Figure 2.24a: ECE mechanism starting from a PdII monomer:

0 Reaction E or Keq α ks or kf Pd + e ⇄ Pd 1.690 V 0.75 1⨯10–5 cm s–1

, –1 –1 Pd + Pd ⇄ Pd 2121.2 N/A 200 M s , –6 –1 Pd + e ⇄ Pd 1.489 V 0.25 1⨯10 cm s - The mechanism for Figure 2.24b: ECE mechanism starting from a PdII dimer:

0 Reaction E or Keq α ks or kf

4 –1 –1 2 Pd ⇄ Pd 1000 N/A 1⨯10 M s , ∗ 1.690 V 0.75 1⨯10–5 cm s–1 Pd + e ⇄ Pd , ∗ , 1 N/A 1 s–1 Pd ⇄ Pd

, –6 –1 Pd + e ⇄ Pd 1.489 V 0.25 1⨯10 cm s - Simulation parameters:

Electrode geometry = planar, 1 cm2 Diffusion = semi-infinite 1D Pre-equilibrium = Smart

Potential steps = 0.005 V Ru and Cdl = both 0 Temperature = 295.2 K Noise level = 0% Gauss-Newton Iterations = 2 Exp. Factor (x-grid) = 0.5 Truncation Error = 10–5 Xmax/sqrt(Dt) = 6 Local FEM error level = 0.02

III 2.4.9. Computational Modeling of Pd 2

III − We constructed a series of plausible structures for Pd 2 with HSO4 and H2SO4 ligands and computed their Raman-active vibration modes and their intensities. Three different protonation states were

− − − selected (total charge q = 0, 2, and 4 for 6 HSO4 /0 H2SO4, 4 HSO4 /2 H2SO4, and 2 HSO4 /4 H2SO4 ligands), and two or three isomers with different numbers of bridging ligands (µ4, µ2, µ0) were calculated for each protonation state (Figure 2.25).

Initial molecular models were constructed using the molecular editor and visualizer software, Avogadro Version 1.2.0.55 The extensive hydrogen bonding interaction in the sulfuric acid solvent kept us from including explicit solvation due to computational expense. Calculations were therefore carried out in the aqueous phase with implicit solvation using a polarizable continuum model. Geometry optimizations and vibrational frequency calculations were performed using density functional theory (DFT) as implemented in Gaussian09.56 Geometry optimizations were performed using a “tight” convergence criteria, converging to an RMS force of 1  10−5 hartrees/bohr. The optimized structures are shown in Figure 2.25. Force constants and Raman vibrational frequencies were computed with modified two-electron integrals and derivatives, specifying a super-fine integration grid for the numerical integration. Calculated vibrational frequencies were scaled by a factor of 0.9862.57 For all calculations, the electron exchange and correlation 69 were described with the GGA functional PBE.58 Electronic wave functions were constructed with a LANL2DZ basis, a double zeta basis set. Raman scattering intensities (Figure 2.26, vertical bars) were obtained with a 532 nm simulated excitation source as implemented in GaussSum3.59 Simulated Raman spectra (Figure 2.26, curvy lines) were obtained by applying Gaussian lineshape and smearing to the individual scattering intensities and combining them.

As can be seen in Figure 2.26, poor agreement between the experimental and calculated Raman

III spectra was found. In Chapter 3, we used free energies as the basis for refining DFT models for the Pd 2 complex. Furthermore, a larger number of isomers and conformations were studied.

Figure 2.25. Geometry-optimized structures for the seven isomers that were calculated.

Figure 2.26. Simulated Raman spectra (excitation wavelength = 532 nm) for the seven isomers computed III to model Pd 2.

2.5. References

(1) O’Reilly, M. E.; Kim, R. S.; Oh, S.; Surendranath, Y. Catalytic Methane Monofunctionalization by an Electrogenerated High-Valent Pd Intermediate. ACS Cent. Sci. 2017, 3 (11), 1174–1179. (2) Połczyński, P.; Jurczakowski, R.; Grochala, W. Stabilization and Strong Oxidizing Properties of Ag(II) in a Fluorine-Free Solvent. Chem. Commun. 2013, 49 (68), 7480–7482. (3) Khusnutdinova, J. R.; Rath, N. P.; Mirica, L. M. Dinuclear Palladium(III) Complexes with a Single Unsupported Bridging Halide Ligand: Reversible Formation from Mononuclear Palladium(II) or Palladium(IV) Precursors. Angew. Chemie - Int. Ed. 2011, 50 (24), 5532–5536. (4) Kornecki, K. P.; Berry, J. F.; Powers, D. C.; Ritter, T. Metal-Metal Bond-Containing Complexes as Catalysts for C-H Functionalization. In Progress in Inorganic Chemistry Volume 58; Karlin, K. D.,

Ed.; John Wiley & Sons, Inc., 2014; Vol. 58, pp 225–302. (5) Ariafard, A.; Hyland, C. J. T.; Canty, A. J.; Sharma, M.; Brookes, N. J.; Yates, B. F. Ligand Effects in Bimetallic High Oxidation State Palladium Systems. Inorg. Chem. 2010, 49 (23), 11249–11253. (6) Mirica, L. M.; Khusnutdinova, J. R. Structure and Electronic Properties of Pd(III) Complexes. Coord. Chem. Rev. 2013, 257 (2), 299–314. (7) Lu, E.; Liddle, S. T. Group 10 Metal-Metal Bonds. In Molecular Metal-Metal Bonds; Wiley-VCH Verlag GmbH & Co. KGaA: Weinheim, Germany, 2015; pp 325–395. (8) Eitel, S. H.; Bauer, M.; Schweinfurth, D.; Deibel, N.; Sarkar, B.; Kelm, H.; Krüger, H. J.; Frey, W.; Peters, R. Paramagnetic Palladacycles with Pd III Centers Are Highly Active Catalysts for Asymmetric Aza-Claisen Rearrangements. J. Am. Chem. Soc. 2012, 134 (10), 4683–4693. (9) Powers, I. G.; Uyeda, C. Metal-Metal Bonds in Catalysis. ACS Catal. 2017, 7 (2), 936–958. (10) Zhang, B.; Yan, X.; Guo, S. Synthesis of Well-Defined High-Valent Palladium Complexes by Oxidation of Their Palladium(II) Precursors. Chem. - Eur. J. 2020, No. Ii. (11) Gruzensky, P. M. Crystallization of Anhydrous Copper Sulfate from Sulfuric Acid - Ammonium Sulfate Mixtures. J. Res. Natl. Bur. Stand. Sect. A Phys. Chem. 1964, 68A (3), 313. (12) Dahmen, T.; Rittner, P.; Böger-Seidl, S.; Gruehn, R. Beiträge Zum Thermischen Verhalten von Sulfaten XIV. Zum Thermischen Verhalten von PdSO4 · 2H2O Und PdSO4 · 0,75H2O Sowie Zur Struktur von M-PdSO4. J. Alloys Compd. 1994, 216 (1), 11–19. (13) Khusnutdinova, J. R.; Mirica, L. M. CHAPTER 5. Organometallic Pd III Complexes in C–C and C–Heteroatom Bond Formation Reactions. In C–H and C–X Bond Functionalization: Transition Metal Mediation; Ribas, X., Ed.; The Royal Society of Chemistry, 2013; pp 122–158. (14) Colpas, G. J.; Maroney, M. J.; Bagyinka, C.; Kumar, M.; Willis, W. S.; Suib, S. L.; Mascharak, P. K.; Baidya, N. X-Ray Spectroscopic Studies of Nickel Complexes, with Application to the Structure of Nickel Sites in Hydrogenases. Inorg. Chem. 1991, 30 (5), 920–928. (15) Cramer, S. P.; Eidsness, M. K.; Pan, W. H.; Morton, T. A.; Ragsdale, S. W.; DerVartanian, D. V.; Ljungdahl, L. G.; Scott, R. A. X-Ray Absorption Spectroscopic Evidence for a Unique Nickel Site in Clostridium Thermoaceticum Carbon Monoxide Dehydrogenase. Inorg. Chem. 1987, 26 (15), 2477–2479. (16) Getsoian, A. “Bean”; Das, U.; Camacho-Bunquin, J.; Zhang, G.; Gallagher, J. R.; Hu, B.; Cheah, S.; Schaidle, J. A.; Ruddy, D. A.; Hensley, J. E.; et al. Organometallic Model Complexes Elucidate the Active Gallium Species in Alkane Dehydrogenation Catalysts Based on Ligand Effects in Ga K- Edge XANES. Catal. Sci. Technol. 2016, 6 (16), 6339–6353. (17) Zhang, G.; Li, J.; Deng, Y.; Miller, J. T.; Kropf, a J.; Bunel, E. E.; Lei, A. Structure-Kinetic Relationship Study of Organozinc Reagents. Chem. Commun. (Camb). 2014, 50 (63), 8709–8711. (18) Campbell, M. G.; Powers, D. C.; Raynaud, J.; Graham, M. J.; Xie, P.; Lee, E.; Ritter, T. Synthesis and Structure of Solution-Stable One-Dimensional Palladium Wires. Nat. Chem. 2011, 3 (12), 949– 953. (19) Blackburn, N. J.; Barr, M. E.; Woodruff, W. H.; van der Ooost, J.; de Vries, S. Metal-Metal Bonding in Biology: EXAFS Evidence for a 2.5 Å Copper-Copper Bond in the CUA Center of Cytochrome Oxidase. Biochemistry 1994, 33 (34), 10401–10407. (20) Jalilehvand, F.; Maliarik, M.; Mink, J. J.; Sandström, M.; Ilyukhin, A.; Glaser, J. Structure Studies 4- of Dimeric [Pt2(CN)10] Pentacyanoplatinum(III) and Monomeric Pentacyanoplatinum(IV) Complexes by EXAFS, Vibrational Spectroscopy, and X-Ray Crystallography. J. Phys. Chem. A

2002, 106 (14), 3501–3516. (21) Kornecki, K. P.; Briones, J. F.; Boyarskikh, V.; Fullilove, F.; Autschbach, J.; Schrote, K. E.; Lancaster, K. M.; Davies, H. M. L.; Berry, J. F. Direct Spectroscopic Characterization of a Transitory Dirhodium Donor-Acceptor Carbene Complex. Science (80-. ). 2013, 342 (6156), 351–354. (22) Tanase, T.; Kawahara, K.; Ukaji, H.; Kobayashi, K.; Yamazaki, H.; Yamamoto, Y. Electrochemical Preparation and Characterization of Binuclear Palladium(I) Complexes Containing Aromatic Isocyanide and Chelating Diphosphine Ligands. Inorg. Chem. 1993, 32 (17), 3682–3688. (23) Fernández, E. J.; Hardacre, C.; Laguna, A.; Cristina Lagunas, M.; López-de-Luzuriaga, J. M.; Monge, M.; Montiel, M.; Elena Olmos, M.; Puelles, R. C.; Sánchez-Forcada, E. Multiple Evidence for Gold(I)-Siher(I) Interactions in Solution. Chem. - Eur. J. 2009, 15 (25), 6222–6233. (24) Asakura, K.; Kitamura-Bando, K.; Isobe, K.; Arakawa, H.; Iwasawa, Y. Metal-Assisted CO Insertion Reaction on a New Surface Rhodium Dimer Catalyst Observed by an in Situ EXAFS Technique. J. Am. Chem. Soc. 1990, 112 (8), 3242–3244. (25) Askham, F. R.; Stanley, G. G.; Marques, E. C. A New Type of Transition-Metal Dimer Based on a 2+ Hexaphosphine Ligand System: Co2(CO)4(eHTP) (eHTP = (Et2PCH2CH2)2PCH2P(CH2CH2PEt2)2). J. Am. Chem. Soc. 1985, 107 (25), 7423–7431.

(26) Muraveiskaya, G. S.; Orlova, V. S.; Evstaf’eva, O. N. Reaction of K2Pt(NO2)4 with H2SO4 and the Preparation of Sulphato-Compounds of Platinum. Russ. J. Inorg. Chem. 1974, 19 (4), 1030-1035 (561-566). (27) Chen, M.; Wei, C.; Tao, J.; Wu, X.; Huang, N.; Zhang, G.; Li, L. Supramolecular Polymers Self- Assembled from Trans-Bis(Pyridine) Dichloropalladium(II) and Platinum(II) Complexes. Chem. - Eur. J. 2014, 20 (10), 2812–2818. (28) Gager, H. M.; Lewis, J.; Ware, M. J.; Lewis, L.; Ware, M. J. Metal–Metal Stretching Frequencies in Raman Spectra. Chem. Commun. (London) 1966, 0 (17), 616–617. (29) Miskowski, V. M.; Smith, T. P.; Loehr, T. M.; Gray, H. B. Properties of Metal-Metal Single Bonds. Vibrational and Electronic Spectra of Binuclear Rhodium(II) and Iridium(II) Isocyanide Complexes with Comparisons to Decacarbonyldimanganese [Mn2(CO)10]. J. Am. Chem. Soc. 1985, 107 (26), 7925–7934. (30) Newman, R. A.; Martin, D. S.; Dallinger, R. F.; Woodruff, W. H.; Stiegman, A. E.; Che, C. M.; Schaefer, W. P.; Miskowski, V. M.; Gray, H. B. Vibrational and Electronic Spectra of Tetrasulfatodiplatinate(2-) Complexes. Inorg. Chem. 1991, 30 (24), 4647–4654. (31) Orlova, V. S.; Muraveiskaya, G. S.; Evstaf’eva, O. N. Preparation of the Sulphato-Coplexes of Platinum Mn[Pt2(SO4)4X2].H2O and Mn[Pt2(SO4)4XY].H2O. Russ. J. Inorg. Chem. 1975, 20, 1340– 1346. (32) Miskowski, V. M.; Dallinger, R. F.; Christoph, G. G.; Morris, D. E.; Woodruff, W. H.; Spies, G. H.; Woodruff, W. H. Assignment of the Rhodium-Rhodium Stretching Frequency in Rh2(O2CCH3)4L2 Complexes and the Crystal and Molecular Structure of [C(NH2)3]2[Rh(O2CCH3)4Cl2]. Relationship between Vibrational Spectra and Structure. Inorg. Chem. 1987, 26 (13), 2127–2132. (33) Drago, R. S. Physical Methods in Inorganic Chemistry, 2nd ed.; Surfside Scientific Publishers, 1992. (34) Yao, C. L.; He, L. P.; Korp, J. D.; Bear, J. L. Dipalladium Complexes with N,N′- Diphenylbenzamidine Bridging and Chelating Ligands. Synthesis and Structural and Electrochemical Studies. Inorg. Chem. 1988, 27 (24), 4389–4395. (35) Powers, D. C.; Ritter, T. Bimetallic Redox Synergy in Oxidative Palladium Catalysis. Acc. Chem. Res. 2012, 45 (6), 840–850. 73

(36) Berry, J. F.; Bill, E.; Bothe, E.; Cotton, F. A.; Dalal, N. S.; Ibragimov, S. A.; Kaur, N.; Liu, C. Y.; 5+ Murillo, C. A.; Nellutla, S.; et al. A Fractional Bond Order of 1/2 in Pd2 -Formamidinate Species; the Value of Very High-Field EPR Spectra. J. Am. Chem. Soc. 2007, 129 (5), 1393–1401. (37) Dudkina, Y. B.; Kholin, K. V.; Gryaznova, T. V.; Islamov, D. R.; Kataeva, O. N.; Rizvanov, I. K.; Levitskaya, A. I.; Fominykh, O. D.; Balakina, M. Y.; Sinyashin, O. G.; et al. Redox Trends in Cyclometalated Palladium(II) Complexes. Dalt. Trans. 2017, 46 (1), 165–177. (38) McAuley, A.; Whitcombe, T. W. Bis(1,4,7-Triazacyclononane)Palladium(III): Characterization and Reactions of an Unusually Stable Monomeric Palladium(III) Ion. Inorg. Chem. 1988, 27 (18), 3090– 3099. (39) Powers, D. C.; Geibel, M. A. L.; Klein, J. E. M. N.; Ritter, T. Bimetallic Palladium Catalysis: Direct Observation of Pd(III)-Pd(III) Intermediates. J. Am. Chem. Soc. 2009, 131 (47), 17050–17051. (40) Dudkina, Y. B.; Mikhaylov, D. Y.; Gryaznova, T. V.; Tufatullin, A. I.; Kataeva, O. N.; Vicic, D. A.; Budnikova, Y. H. Electrochemical Ortho Functionalization of 2-Phenylpyridine with Perfluorocarboxylic Acids Catalyzed by Palladium in Higher Oxidation States. Organometallics 2013, 32 (17), 4785–4792. (41) Nguyen, B. N.; Adrio, L. A.; Albrecht, T.; White, A. J. P.; Newton, M. A.; Nachtegaal, M.; Figueroa, S. J. A.; Hii, K. K. Electronic Structures of Cyclometalated Palladium Complexes in the Higher Oxidation States. Dalt. Trans. 2015, 44 (37), 16586–16591. (42) Durrell, A. C.; Jackson, M. N.; Hazari, N.; Gray, H. B. Making Carbon-Chlorine Bonds by Dipalladium Electrocatalysis. Eur. J. Inorg. Chem. 2013, 2013 (7), 1134–1137. (43) Koshevoy, I. O.; Lahuerta, P.; Sanaú, M.; Ubeda, M. A.; Doménech, A. Dinuclear Palladium(II) Compounds with Bridging Cyclometalated Phosphines. Synthesis, Crystal Structure and Electrochemical Study. J. Chem. Soc. Dalt. Trans. 2006, No. 46, 5536–5541. (44) Bercaw, J. E.; Durrell, A. C.; Gray, H. B.; Green, J. C.; Hazari, N.; Labinger, J. A.; Winkler, J. R. Electronic Structures of Pd(II) Dimers. Inorg. Chem. 2010, 49 (4), 1801–1810. (45) Appleton, T. G.; Barnham, K. J.; Byriel, K. A.; Hall, J. R.; Kennard, C. H. L.; Mathieson, M. T.; Penman, K. G. Reactions of Nitroplatinum Complexes. 2. Reactions of K2[Pt(NO2)4] and Related Complexes with Aqueous Acids (CH3CO2H, HClO4, CF3SO3H, HNO3, and H2SO4): Pathways to Platinum(III) Complexes with Acetate Bridges. Crystal Structure of K2[{Pt(NO2)2(µ-CH3CO2)}. Inorg. Chem. 1995, 34 (24), 6040–6052. (46) Ravel, B.; Newville, M. ATHENA, ARTEMIS, HEPHAESTUS: Data Analysis for X-Ray Absorption Spectroscopy Using IFEFFIT. J. Synchrotron Radiat. 2005, 12, 537–541. (47) Rehr, J. J.; Mustre de Leon, J.; Zabinsky, S. I.; Albers, R. C. Theoretical X-Ray Absorption Fine Structure Standards. J. Am. Chem. Soc. 1991, 113 (14), 5135–5140. (48) Kau, L. shan; Spira-Solomon, D. J.; Penner-Hahn, J. E.; Hodgson, K. O.; Solomon, E. I. X-Ray Absorption Edge Determination of the Oxidation State and Coordination Number of Copper. Application to the Type 3 Site in Rhus Vernicifera Laccase and Its Reaction with Oxygen. J. Am. Chem. Soc. 1987, 109 (21), 6433–6442. (49) Shulman, G. R.; Yafet, Y.; Eisenberger, P.; Blumberg, W. E. Observations and Interpretation of X- Ray Absorption Edges in Iron Compounds and Proteins. Proc. Natl. Acad. Sci. 1976, 73 (5), 1384– 1388. (50) Smith, T. A.; Penner-Hahn, J. E.; Berding, M. A.; Doniach, S.; Hodgson, K. O. Polarized X-Ray Absorption Edge Spectroscopy of Single-Crystal Copper(II) Complexes. J. Am. Chem. Soc. 1985, 107 (21), 5945–5955.

(51) Allemand, C. D. Depolarization Ratio Measurements in Raman Spectrometry. Appl. Spectrosc. 1970, 24 (3), 348–353. (52) Eaton, G. R.; Eaton, S. S.; Barr, D. P.; Weber, R. T. Quantitative EPR; Springer Vienna: Vienna, 2010. (53) Duin, E. C. Introduction to Biomolecular Electron Paramagnetic Resonance Theory, Chapter 2. (54) Pandey, R. N.; Henry, P. M. Interaction of Palladium(II) Acetate with Sodium and Lithium Acetate in Acetic Acid. Can. J. Chem. 1974, 52 (8), 1241–1247. (55) Hanwell, M. D.; Curtis, D. E.; Lonie, D. C.; Vandermeerschd, T.; Zurek, E.; Hutchison, G. R. Avogadro: An Advanced Semantic Chemical Editor, Visualization, and Analysis Platform. J. Cheminform. 2012, 4 (8). (56) Frisch, M. J.; Trucks, G. W. .; Schlegel, H. B. .; Scuseria, G. E. .; Robb, M. A. .; Cheeseman, J. R.; Scalmani, G. .; Barone, V. .; Mennucci, B. .; Petersson, G. A.; Nakatsuji, H. .; Caricato, M. .; et al. Gaussian 09, Revision E. 01; Gaussian; 2009. (57) Johnson III, R. D. NIST Computational Chemistry Comparison and Benchmark Database. NIST Stand. Ref. Database Number 101 2016, http://cccbdb.nist.gov/. (58) Quintal, M. M.; Karton, A.; Iron, M. A.; Boese, A. D.; Martin, J. M. L. Benchmark Study of DFT Functionals for Late-Transition-Metal Reactions †. J. Phys. Chem. A 2006, 110 (2), 709–716. (59) O’boyle, N. M.; Tenderholt, A. L.; Langner, K. M. Cclib: A Library for Package-Independent Computational Chemistry Algorithms. J. Comput. Chem. 2008, 29 (5), 839–845.

3. Reaction Mechanism of Rapid and Selective Methane III Functionalization by Pd 2

Parts of this chapter contain contributions from collaborators:

(DFT computation) Dr. Azadeh Nazemi, Prof. Thomas R. Cundari

3.1. Introduction

The functionalization of unactivated C–H bonds is an attractive route for the preparation of valuable chemicals from abundant hydrocarbon feedstocks.1,2 Molecular Pd complexes are potent agents for C–H functionalization catalysis and high-valent Pd intermediates in +3 or +4 formal oxidation states have been shown to complement the reactivity of the more traditional Pd0/II catalysis, particularly for the installment of nucleophilic functional groups.3–10 Generally, C–H activation at PdII is followed by oxidation

IV III to an organometallic Pd or Pd 2 that undergoes reductive elimination. This activation-oxidation- functionalization sequence has enabled the efficient installment of various nucleophiles at easily activated sp2 C–H bonds. However, functionalization of C–H bonds that are less reactive towards PdII, particularly sp3 C–H bonds, remains challenging, requiring carefully chosen directing groups along with steric pressure.11,12 This challenge may be addressed by designing mechanistic sequences in which C–H activation occurs from a high-valent Pd center.13,14 For example, a distinct selectivity profile was observed with C–H activation at PdIV compared to PdII,15 and a PdIII dimer intermediate was implicated in the aerobic α- hydroxylation of carbonyl compounds.16 Such examples are rare, however, because many chemical oxidants cannot oxidize inorganic PdII, and the coordinative saturation of high-valent Pd centers may impede substrate coordination. Overcoming these obstacles could enable oxidation-first sequences that expand the scope of Pd-mediated C–H functionalization.

Using electrochemistry, we have shown that this oxidation-first sequence is effective for the functionalization of the inert C–H bonds of methane.14 Unlike chemical oxidants, electrochemistry affords

17,18 II access to high and tunable driving forces for oxidation. Leveraging this quality, we showed that Pd SO4 in concentrated or fuming sulfuric acid could be electrochemically oxidized to a novel high-valent Pd species that rapidly functionalizes methane.14 The rarity of C–H activation at a high-valent Pd species, combined with the exceptional reaction rates and parallel formation of stoichiometric and superstoichiometric methane functionalization products, motivated detailed inquiries. In Chapter 2, through spectroscopic studies, we arrived at a structural model for the high-valent Pd species as a Pd–Pd bonded complex in an all-oxidic octahedral ligand field, and elucidated the mechanism by which it is formed by electrochemical oxidation. In this chapter, we present the results of our mechanistic investigation

III of Pd 2 with methane by a combination of NMR kinetic studies and DFT computation.

3.2. Results and Discussions

3.2.1. Experimental Observations

3.2.1.1. Rate laws from initial rate measurements In order to obtain rate laws for MBS and MSA generation, we measured initial rates and

III 14 determined reaction orders in all reactants, CH4, Pd 2, and SO3. In our previous work, we measured the rate of MBS generation using electrochemical methods. However, as the method relies on the detection of redox events at an electrode, it could not measure the rate of MSA generation, which is net redox neutral

III II (CH4 + SO3  CH3SO3H) and does not accompany the reduction of Pd 2 to Pd . We therefore turned to

NMR spectroscopy and recorded the concentrations of methane, MBS, and MSA in real time (Figure 3.1) by carrying out the reaction inside a heavy-walled NMR tube that could be pressurized with methane (see 3.4.3 for details of the experiment). Usually, methane oxidation catalysis requires high pressure and

III temperature and is therefore not amenable to real time NMR measurements. The fast reaction rates of Pd 2 at a mild pressure/temperature enabled our measurements using standard apparatus and instruments.

Importantly, the total methyl concentration (i.e., [CH4] + [MBS] + [MSA]) was constant over time, indicating no overoxidation of either MBS or MSA occurs (Figure 3.1, green). We also confirmed that the

III spontaneous decay of Pd 2 would be negligible in our analyses (see 3.4.2.2). The concentration-time traces were first fitted to generic exponential equations, without any a priori assumptions about the mechanism. [] Plotting the initial rates ( ≡풓,) thus obtained versus the initial CH4 concentration ([CH4]i), we derived reaction orders of MBS and MSA in CH4 as first-order for both (Figure 3.2a). While the slope of log rMBS,i vs log [CH4]i deviates slightly from 1, mass conservation dictates that rMBS,i = –(rCH4,i + rMSA,i); therefore, the clear first-order dependence of methane consumption and MSA generation rates on [CH4] supports the first-order dependence of MBS generation rate on [CH4]. This is also in agreement with our previous conclusion from electrochemical measurements.14 Therefore, NMR spectroscopy yielded real time concentration-time information from which initial rates and methane reaction orders could be determined without any assumptions about the reaction mechanism.

Figure 3.1. A typical concentration-time plot from real time NMR reaction monitoring of methane III oxidation by Pd 2.

III Reaction orders in Pd 2 and SO3 were determined by carrying out real time NMR measurements

III III under different initial concentrations of Pd 2 and SO3 ([Pd 2]i and [SO3]i). Cognizant of the low concentration of MBS and the large error associated with the fitted parameters, we sought to reduce the number of fitting parameters by utilizing the fact that both MBS and MSA formation rates are first-order in methane. Following the derivation described in 3.4.4.2, we performed a global fit comprised of only two

77 fitting parameters to the three concentration-time traces of each experiment. The two parameters are the [] observed rate constants for MBS and MSA generation, kMBS,obs and kMSA,obs, from = rMBS =

[] III kMBS,obs[CH4] and = rMSA = kMSA,obs[CH4]. Plots of log kMBS,obs and log kMSA,obs versus log [Pd 2]i and III log [SO3]i are shown in Figure 3.2b,c. Pd 2 and SO3 affected the two product formation rates very

III differently. MBS formation rate (rMBS) and [Pd 2] was linearly correlated while rMSA showed a near zero-

III order dependence on [Pd 2]. On the other hand, increasing [SO3] facilitated MSA formation, but minimally affected or even slightly suppressed MBS formation. To obtain the rate laws, we rounded the slopes in

III Figure 3.2b and c to the nearest integer values and assigned, for MBS, first-order in Pd 2 and zero-order

III in SO3, while for MSA, zero-order in Pd 2 and first-order in SO3. The slope of log kMSA,obs vs log [SO3]i deviates significantly from 1, but even so, we assigned the SO3 reaction order of MSA as 1 rather than ½ considering that the net reaction for MSA formation is CH4 + SO3 → CH3SO3H. We postulate that the nominal concentration of SO3 may differ considerably from the actual activity of SO3, as SO3 in fuming

19 sulfuric acid exists in complex equilibria involving disulfuric acid (H2S2O7) that shift with [SO3], although our other results imply additional subtleties that are beyond the scope of our studies here (see 3.2.1.5).

Together with the CH4 reaction order, we obtained the following rate laws:

푑[MBS] = 풓 = 푘 [CH ][Pd ] … 3.1 푑푡 푑[MSA] = 풓 = 푘 [CH ][SO ] … 3.2 푑푡 III Thus, reaction orders for methane monofunctionalization by Pd 2 were determined for all three reactants, and experimental rate laws were formulated.

Figure 3.2. The rate of methane oxidation to methyl bisulfate (MBS) and methanesulfonic acid (MSA) at III 50 ̊C measured as a function of the initial concentrations of (a) CH4, (b) Pd 2 and (c) SO3.

3.2.1.2. KIE and activation barriers Carrying out the reaction in NMR tubes, we also determined kinetic isotope effects (KIE) and activation barriers for MBS and MSA formation. Rate constants were obtained with CD4 at 50 ̊C and with

CH4 at variable temperatures from 40 to 72 ̊C. These rate constants could not be determined by real time

78 reaction monitoring as we did in the preceding section; for the reaction of CD4, the sensitivity of our NMR instrument for deuterium was too low for rapid measurements, and for the reaction at high temperatures, the reaction was too fast compared to the relaxation delay required for accurate quantitation such that not enough data points could be acquired during the reaction. Instead, rate constants were determined by inputting initial and final concentrations of CH4, MBS, and MSA into a microkinetic model constructed using the foregoing rate laws. The reaction was carried out in a sealed NMR tube immersed in a heated oil bath for a set amount of time, after which the NMR tube was rapidly cooled down and the final concentrations of CH4, MBS, and MSA were determined. Initial CH4 concentration was obtained as their sum total, since the total methyl group concentration is conserved (Figure 3.1). To analyze this data, although we do not know the true reaction mechanism yet, a phenomenological mechanism that reflects the

III measured reaction orders and stoichiometries (i.e., one equivalent of Pd 2 is reduced for every equivalent

III 14 of MBS generated, but MSA does not consume Pd 2) were set up:

… 3.3 CH + Pd ⎯⎯ CHOSOH + 2 Pd … 3.4 CH + SO ⎯⎯ CHSOH Simulation of this mechanism with COPASI showed good agreement with the experimental concentration-time traces (Figure 3.18). Therefore, concentrations of CH4, MBS, and MSA measured at time t, along with their sum (equivalent to the initial CH4 concentration), were fitted with this mechanism using the COPASI software to give the rate constants for MBS and MSA formation (see 3.4.4 for further details).

Comparing the rate constants independently measured for CH4 and CD4 at 50 ̊C, KIEs of 4.1(±0.6) and 4.3(±0.4) were obtained for MBS and MSA, respectively. Therefore, C–H cleavage is rate-limiting for both reactions. It was found that there is no H/D exchange, either in the product or the starting material, which is consistent with rate-limiting C–H cleavage followed by fast subsequent reactions (Figure 3.20). The similarity of the KIE values also suggest that the rate-limiting C–H cleavage step for the two reactions may be mechanistically similar or even shared.

Rate constants measured at different temperatures gave relatively low Arrhenius activation energies of 21.3(±0.2) and 26.3(±0.4) kcal/mol for MBS and MSA, respectively (Figure 3.3). This Ea for MBS is lower than the value measured electrochemically at higher temperatures (80–140 ̊C), 25.9(±2.6) kcal/mol.14 We consider the value obtained in the current study to be more accurate, as the electrochemical method is indirect. Comparison of the rates measured at high temperatures in our previous work and those at low temperatures in this study requires knowledge of CH4 concentration under the conditions of our

20 20,21 previous study, which is unavailable. An estimate of [CH4] based on available literature leads to better agreement between the low- and high-temperature rate constants using the Ea value measured in this work 79

(see 3.4.4.3 for details). Incidentally, the higher Ea for MSA generation compared to MBS gave increased selectivity for MSA at increasing temperatures. Altogether, analyzing reactant and product concentrations with a microkinetic model derived from experimental rate laws, we obtained KIEs and Arrhenius activation

III energies for the functionalization of methane to MBS and MSA by Pd 2.

Figure 3.3. Arrhenius plots for MBS and MSA formation.

II,III 3.2.1.3. Suppression of methanesulfonic acid generation by O2 and Pd2 Based on studies using various radical initiators, the generation of methanesulfonic acid in fuming sulfuric acid has been attributed to a radical chain reaction. The following sequence of reactions have been proposed:22

∙ ∙ … 3.5 CH + SO ⎯ CHSO ∙ ∙ … 3.6 CHSO + CH ⎯ CHSOH + CH (the subscripts “rp” in the rate constants stand for radical propagation). To experimentally probe the radical

III nature of MSA generation from the reaction of Pd 2 with methane, O2 was added as a radical scavenger that is stable in the fuming sulfuric acid medium. Co-addition of O2 (Figure 3.4, hollow symbols) resulted in suppression of MSA generation, implying the involvement of radical intermediates for MSA generation. Additionally, the decrease in MSA generation was accompanied by an increase in the formation of MBS, for which we give a tentative explanation in a later discussion (3.2.3).

In order to further investigate the radical nature of MSA generation, we studied the effect of

II,III increasing the concentration of the Pd2 intermediate. As described in Chapter 2, the mixed-valent dimer

II III III exists in a minor equilibrium with Pd and Pd 2, on the order of 1–100 μM, for the Pd 2 samples used in this study. The S = ½ complex clearly inhibited MSA generation (Figure 3.5 and Table 3.5), indicating that

II,III it is an effective scavenger for the radical intermediates involved in MSA formation. Unlike O2, the Pd2 complex did not influence the rate of MBS formation.

Figure 3.5. Real time concentration-time traces obtained with (solid symbols) low and (hollow symbols) II,III II,III II III high concentrations of Pd2 . The high [Pd2 ] sample was prepared by adding a Pd solution to a Pd 2 II,III solution and equilibrating overnight. The slightly higher rate of MBS formation in the high [Pd2 ] case is III due to a slightly higher concentration of the Pd 2 complex in this sample. The exact concentrations of the different Pd species in the solutions and the extracted values of kobs for each experiment are given in Table 3.5.

3.2.1.4. Relative concentrations of the propagating radicals: MSA generation initiated by peroxydisulfate

III The previous section clearly demonstrates that MSA in the reaction of Pd 2 with methane is formed via radical propagation reactions, presumably those in equations 3.5 and 3.6 that are commonly proposed for other radical initiators in fuming sulfuric acid. Surprisingly, although widely accepted in the literature, little was known about the kinetics of these reactions.

III Using similar reaction conditions and methods utilized in the study of Pd 2, we monitored the reaction of methane with potassium peroxydisulfate (K2S2O8; although it will be protonated in the sulfuric acid solution, it will be referred to as peroxydisulfate in this work), a classic radical initiator.22,23 As the solid symbols in Figure 3.6a shows, only MSA was generated. The total methyl concentration (i.e., [CH4] + [MSA]) was preserved, indicating that the radical chain reaction does not involve unselective oxidation, at least under the mild temperature (50 ̊C) employed. Importantly, the concentration-time trace looked very

III different from that of methane oxidation by Pd 2; whereas exponential curves were observed for methane

III oxidation and MSA generation by Pd 2 due to the first-order dependence of the rate on [CH4], methane oxidation initiated by peroxydisulfate showed a linear trace after a slight induction period. This indicates that the rate of MSA formation remains constant even as methane is being consumed, implying that MSA formation rate is independent of [CH4] (see below and 3.4.6 for reaction orders in all reactants and a simulation of the concentration-time trace).

When O2 was added (Figure 3.6a, hollow symbols), suppression of MSA generation was observed,

III as expected. The degree of suppression, however, was much greater than with the Pd 2 complex. Moreover, the total methyl concentration (Figure 3.6a, green symbols) during peroxydisulfate-initiated methane functionalization showed a slight decrease when O2 was added, suggesting a mild overoxidation of MSA or other side reactions.

Importantly, analysis of the variation of MSA formation rate upon variation of [CH4] and [SO3] yielded the relative magnitude of the rate constants for the two radical propagation reactions, equations 3.5 and 3.6. Rates of MSA generation under different [CH4] and [SO3], obtained from the linear portion of concentration-time traces (see 3.4.6 for details), showed ~zero- and first-order dependence on [CH4] and

[SO3], respectively (Figure 3.6b, open squares). Then, assuming slow initiator decomposition and a steady-

• • state condition so that the total concentration of propagating radicals is constant (i.e., [CH3 ] + [CH3SO3 ] = constant), the rate of MSA generation was simulated and roughly fitted to the experimental data. The relatively good fit shown in Figure 3.6b (cross symbols) could be achieved only when the rate constant for

• reaction 3.5, the addition of SO3 to CH3 , was set much lower than the rate constant for reaction 3.6, the

• abstraction of H atom from methane by the CH3SO3 radical. This was surprising, as it means that H atom

82 abstraction from the supposedly inert methane molecule is faster than the reaction of the supposedly reactive methyl radical. Moreover, reaction 3.5 has been computed to be barrierless, both in the literature24 and in our own hands. However, we rationalize the discrepancy between the computational result and the experimental result with the fact that SO3 does not exist as free molecules in the relatively dilute oleum

19 solvent that we use in our studies. Our DFT computation of reaction 3.5 with H2S2O7, which is the major form of SO3 in fuming sulfuric acid, showed a reasonably high activation free energy of 24.0 kcal/mol. In

• conclusion, our rate measurements and analysis indicate that H atom abstraction by CH3SO3 from methane

• is very fast and the rate of radical propagation is limited by the addition of SO3 to CH3 , not the C–H cleavage of methane.

III 3.2.1.5. Mechanistic models for oxidation of methane by Pd 2

From the concentration of CH4 and SO3 under our reaction conditions and the relative magnitude of the rate constants for reactions 3.5 and 3.6 obtained from the foregoing analysis, we estimate the steady-

• • state concentration of CH3 to be higher than that of CH3SO3 by about a factor of 10 (see 3.4.6 for details).

II,III Therefore, if the two propagating radicals are scavenged at similar rates, O2 and the Pd2 complex would

• II,III • III be mostly reacting with CH3 . Importantly, the recombination of Pd2 with CH3 will produce a CH3Pd 2 intermediate, which is expected to undergo a facile reductive elimination to form MBS according to insight from the literature.6

Combining the foregoing experimental observations, we assemble the mechanistic model shown

III • in Figure 3.7 for methane oxidation by the Pd 2 complex. The generation of CH3 as a common intermediate for MBS and MSA formation nicely explains the simultaneous formation of MBS as a

• stoichiometric product and MSA as a superstoichiometric co-product alongside MBS. If CH3 undergoes

II,III radical recombination with Pd2 , a Pd-methyl complex results. The reductive elimination of MBS from

III ‡ CH3Pd 2 was calculated to be barrierless (∆G = –0.7 kcal/mol; see the next section for details of our

III • computational modeling of the Pd 2 complex). On the other hand, if CH3 is captured by SO3, it enters the radical propagation sequence mentioned above (3.2.1.3). Such product bifurcation from the methyl radical intermediate explains the slight suppression of rMBS by SO3 (Figure 3.2c) as well as the inhibitory effect of

II,III the Pd2 complex on MSA formation (Figure 3.5). Furthermore, the common rate-limiting C–H activation step for both MBS and MSA explains the similar and high KIE values for the two products. The rate laws derived from this mechanism (see 3.4.5 for details of the mathematical derivation), equations 3.7 and 3.8,

III are also fully consistent with the observed reaction orders of both products on CH4, Pd 2, and SO3. We do not know the exact origin of the experimental fractional reaction order of ~0.5 for SO3; since the rate of peroxydisulfate-initiated MSA formation shows a clear first-order dependence on [SO3] (Figure 3.21), it would not be solely due to the complex molecular identity of SO3 in fuming sulfuric acid. One possible explanation for the lower reaction order is that the methyl radical generated from H atom abstraction by

III Pd 2 is surrounded by a solvent cage that affects the rate of SO3 addition in a way that distorts the observed reaction order. All in all, the forward dependence is consistent with equation 3.8, and we were unable to formulate a mechanistic model that would predict a half-order for SO3 while being consistent with all the

III other experimental observations. Therefore, we propose that the dual reactivity of Pd 2 towards methane for the parallel generation of MBS and MSA originates from a common methyl radical intermediate formed

III from a rate-limiting H atom abstraction reaction by Pd 2.

풓 = 푘[CH][Pd ] … 3.7 푘푘 [Pd ] 풓 = [CH][SO] , … 3.8 푘 [Pd ]

Figure 3.7. The reaction mechanism proposed on the basis of experimental results for methane oxidation III by Pd 2 in fuming sulfuric acid. RLS denotes the rate-limiting step. Abbreviations for each step stand for H-abstraction (ha), radical recombination (rrc), reductive elimination (rel), and radical propagation (rp). Reactants are shown in black, intermediates in green, and byproducts in grey. 84

While the above mechanistic model is the simplest mechanism that is consistent with all the experimental observations, an ionic pathway for MBS that is independent of MSA generation cannot be ruled out:

( ) Pd + CH ⎯⎯⎯⎯⎯⎯⎯ CH Pd ⎯ CHPd → CHOSOH + 2 Pd … 3.9 III In fact, the H atom abstraction reactivity of Pd 2 was not precedented in the literature of Pd–Pd bonded dimeric PdIII complexes, although there are very few examples of C–H activation from these

8,16 III complexes to begin with. Therefore, further investigation of the reaction of sulfate-ligated Pd 2 with methane, particularly the critical C–H activation step, was carried out with DFT calculations.

3.2.2. Evaluation of mechanistic models with DFT calculations

III 3.2.2.1. Computational modeling of Pd 2

III To further probe the rate-determining C–H activation step by DFT calculations, the Pd 2 complex was modeled based on structural information from X-ray absorption and Raman spectroscopies,14 as well as reported bimetallic complexes in the literature.25–30 The two PdIII centers are connected through a metal-

(x–2) metal single bond and each ligated by five oxygen atoms from HxSO4 ligands coming from the fuming sulfuric acid solvent. This surrounding medium would exhibit extensive hydrogen bonding, protonation/deprotonation and ligand exchange reactions with the complex, but the prohibitive computational cost associated with modeling all of these interactions forced us to opt for implicit solvation based on the SMD method using a highly polar medium (i.e., water) to mimic fuming sulfuric acid. While

(x–2) III numerous protonation states and numbers of HxSO4 ligands attached to Pd 2 are possible, computational

– investigations were focused on the neutral complex with six bisulfate (HSO4 ) ligands unless it was deemed

III informative to explore other isomers (see below). For Pd 2(HSO4)6, 36 stereoisomers were calculated having four different ligation geometries: the bisulfate ligands were attached to the Pd atoms in a monodentate (κ1), bidentate (κ2), or bridging (μ2) fashion, resulting in (a) cis-κ2-μ2, (b) trans-κ2-μ2, (c) paddlewheel, and (d) unbridged geometries (Figure 3.8). Study of possible tautomers shows that the symmetrically protonated hexa-bisulfate tautomer is the most stable (Table 3.9). All conformers had a lower computed free energy for the singlet (orange) versus triplet (green) spin state, consistent with the

III 2 2 experimentally observed diamagnetism of Pd 2. Among the different geometries, the trans-κ -μ geometry was more stable than the others. Later calculations of reaction intermediates and transition states also

2 2 III showed the trans-κ -μ geometry to be the most favorable (see 3.4.7.2). In sum, we modeled the Pd 2

III complex as neutral Pd 2(HSO4)6 and found the most stable geometrical isomer to have two bridging bisulfate ligands that are trans to each other.

III 2 Figure 3.8. Computationally modeled Pd 2(HSO4)6 isomers and their computed free energies. (a) cis-κ - μ2, (b) trans-κ2-μ2, (c) paddlewheel, (d) unbridged. The unbridged complex showed spontaneous Pd–Pd bond cleavage when modeled as a triplet. The free energies are values relative to the most stable isomer (conformer #3, trans-κ2-μ2).

3.2.2.2. Ligand dissociation: Homolytic dissociation is favored over heterolytic dissociation

III Since Pd 2 is coordinatively saturated, the C–H activation reaction was initially modeled as a two- step process where rate-limiting C–H cleavage is preceded by ligand dissociation to expose the electrophilic PdIII center. The experimentally observed high KIE of ~4 demands that the ligand dissociation step is not rate-limiting. With this, the relatively low activation energies of ~20 kcal/mol places an upper limit on the energetics of the ligand dissociation step. It was hypothesized that the dissociation of a weak-field ligand like sulfate/bisulfate in the highly ionic and acidic solvent environment will be facile. Moreover, calculated structures showed lengthening of axial Pd–O bonds (2.25±0.04) by 0.15 Å on average compared to equatorial Pd–O bonds (2.10±0.03) due to the trans effect of the Pd–Pd bond. Subsequent C–H activation at the exposed PdIII center would be consistent with reported examples of electrophilic C–H activation by

31 – III high-valent metals. Contrary to expectations, dissociation of an axial HSO4 from the neutral Pd 2(HSO4)6 complex was significantly uphill in free energy, ΔGrxn (reaction free energy) = 16.4 kcal/mol. To account for the effect of electrostatic charge generation upon ligand heterolysis, the same reaction was then

III + III – computed with [Pd 2(H2SO4)(HSO4)5] and [Pd 2(HSO4)5(SO4)] (Figure 3.9 and Table 3.1). This led to

– lower ΔGrxn’s, especially for the positive complex from which an H2SO4 molecule, rather than HSO4 , dissociated. However, protonation of a Pd-bound bisulfate may have a significant energetic penalty. Therefore, we were unable to support or disprove the hypothetical heterolytic ligand dissociation mechanism by comparing calculated reaction free energies with experimental results due to the difficulty of accurately modeling the solvation environment.

Figure 3.9. Ligand dissociation pathways with reactant and products at different charge states.

1 (x–2) Table 3.1. Reaction free energies for the dissociation of an axial κ -HxSO4 ligand. Reactants and 2 2 III products have the trans-κ -μ geometry for the bridging ligands in all three cases. Neutral: [Pd 2(HSO4)6]; III + III – Cation: [Pd 2(H2SO4)(HSO4)5] ; Anion: [Pd 2(HSO4)5(SO4)] .

Neutral Cation Anion ∆G Het 16.4 –3.0 7.4 (kcal/mol) ∆G Hom 6.7 –9.4 1.5 (kcal/mol)

Reaction free energies of ligand dissociation by Pd–O homolysis were, however, calculated to be more energetically favorable than Pd–O heterolysis in all three cases (Table 3.1). Even for the cationic complex, where the charge states of the reactant and products were matched between the two pathways, homolysis was more favorable by ΔΔGrxn ≈ 6 kcal/mol. For the neutral complex, using the most polar SMD solvation model (N-methylformamide-mixture, ε = 181.56) did not significantly reduce the ΔGrxn difference.

III Thus, electrostatics do not seem to be a major factor that determines the preference of Pd 2 for homolytic rather than heterolytic ligand dissociation. Other geometrical isomers also favored homolysis over heterolysis. They showed higher free energies for the reactant and products of both the heterolytic and homolytic pathways compared to the trans-κ2-μ2 isomer (Table 3.10), indicating that the ligand dissociation reaction will not induce rearrangement of the remaining ligands. While we cannot compare the absolute values of the calculated ΔGrxn’s for ligand dissociation with its experimentally suggested upper limit due to the aforementioned difficulty of modeling solvation, our results suggest that if ligand dissociation were to occur, it would be homolytic rather than heterolytic.

3.2.2.3. Initial mechanistic model: H atom abstraction by a free bisulfate radical To model the C–H bond cleavage step following ligand dissociation, four radical abstraction

II,III pathways were modeled: abstraction of either an H atom or a methyl radical by either the Pd2 complex or the free bisulfate radical (Table 3.2). The bisulfate radical was found to be a much more effective radical

II,III abstracting agent than the Pd2 complex, and H atom abstraction was easier than methyl radical

• abstraction. The free energy barrier for H atom abstraction by HSO4 is ~9.5 kcal/mol, which is highly favorable and qualitatively consistent with the experimental activation barrier when combined with a ligand

III homolysis free energy. These results led us to propose that C–H activation by Pd 2 proceeds through ligand homolysis to reversibly generate a reactive bisulfate radical that undergoes rate-limiting H abstraction with

II,III methane to generate a Pd2 complex and a methyl radical (equations 3.10 and 3.11). The proposed mechanism as a whole is depicted in Figure 3.10.

, ∙ Pd ⇄ Pd + HSO … 3.10 ∙ ∙ HSO + CH ⇄ HSO + CH … 3.11 Table 3.2. Reaction free energies (kcal/mol) of the four possible C–H cleavage reactions following III homolytic ligand dissociation from Pd 2.

Abstracted species • • H CH3 Abstracting agent

II,III Pd2 52.5 66.9

• HSO4 0.04 30.0

Figure 3.10. A stepwise mechanism that features homolytic ligand dissociation and rate-limiting H abstraction by the bisulfate radical. Abbreviations for each step stand for ligand dissociation (ld), H- abstraction (ha), radical recombination (rrc), reductive elimination (rel), and radical propagation (rp). Reactants are shown in black, intermediates in green, and byproducts in grey.

The suggested H abstraction reaction mechanism involving the generation of a free bisulfate radical, however, was at odds with the experimentally observed rate laws. Specifically, it was observed that

III MBS generation is first-order in both [Pd 2] and [CH4]. From Figure 3.10 under steady state approximation (see 3.4.5 for details of the mathematical derivation),

∙ 풓 = 푘[CH][HSO] … 3.12 ∙ 푘[Pd ] [HSO] = , … 3.13 푘[CH] + 푘Pd

Combining the two equations,

푘푘[CH][Pd ] 풓 = , … 3.14 푘[CH] + 푘Pd Equation 3.14 can be simplified by considering two extreme cases. First, if ligand dissociation is

II,III [][ ] fast and reversible, k–ld[Pd2 ] >> kha[CH4] and 풓 ≅ , . This is inconsistent with the

III observation that MBS generation is first-order in [Pd 2]. It is also inconsistent with Figure 3.5, which

II,III shows that Pd2 suppresses MSA generation but does not influence the rate of MBS generation. On the

• other hand, if ligand homolysis is significantly uphill and lifetime of the HSO4 radical is very short, it may

II,III be that k–ld[Pd2 ] << kha[CH4]. The KIE alone is not sufficient to rule out this possibility since it only dictates that the transition state for H abstraction be higher than the transition state for ligand dissociation.

III However, this scenario leads to the reduction of equation 3.14 to rMBS ≈ kld[Pd 2], which is inconsistent with the linear dependence of rMBS on methane concentration. Changing the relative magnitude of the terms

II,III k–ld[Pd2 ] and kha[CH4] in equation 3.14 will only alleviate the discrepancy with the experimental reaction

III order for Pd 2 or CH4 at the cost of increasing the discrepancy for the other. Similarly, the derived rate law for MSA generation disagreed with the experimentally observed one (see 3.4.5.1). In sum, H atom

III abstraction by Pd 2 would not proceed from a free bisulfate radical that has dissociated from the complex.

3.2.2.4. Convergence with the experimentally proposed mechanistic model: H atom abstraction by

III a bound ligand on Pd 2 The foregoing discrepancy between our initial mechanistic model and experimental results could

III be reconciled by removing the ligand dissociation step and considering H atom abstraction by the Pd 2

III complex as-is, as proposed earlier in Figure 3.7. Notably, since Pd 2 is coordinatively saturated, not the Pd

III atom but one of the ligands bound to the Pd 2 complex would participate in the abstraction of an H atom. The reaction was modeled with DFT to examine its energetic feasibility (Figure 3.11). Specifically, the following reaction was calculated:

, ∙ Pd (HSO) + CH → Pd (HSO) + CH + HSO … 3.15 III where one of the bound bisulfate ligands of Pd 2(HSO4)6 (Figure 3.11, R) abstracts the methane H atom

II,III and dissociates from the complex as H2SO4. If it remains bound, the resulting Pd2 complex was calculated to have a high free energy because of the weak binding strength of neutral H2SO4 (see 3.4.7.3

– 1 2 for additional reaction pathways that were calculated). Since the HSO4 ligand can be either κ or κ , we calculated pathways for each being the H atom abstracting center: pathway A is the case where a κ1 ligand abstracts H, and pathway B is the case where a κ2 ligand first converts to κ1 by partially dissociating from Pd (B-Int1). Put another way, the transition state involved either a 6,6-coordinate (A-TS) or a 5,6-

III 1 coordinate Pd 2 complex (B-TS). For pathway B, following H abstraction and dissociation of H2SO4, a κ ligand was rearranged to κ2 so that the Pd coordination number will be the same as pathway A; without this

II,III rearrangement, the 4,6-coordinate Pd2 that resulted was 4.6 kcal/mol higher in free energy (see 3.4.7.3).

II,III The 5,6-coordinate Pd2 after H atom abstraction (A-Int2 and B-Int2) undergoes recombination with

• III CH3 to yield the CH3Pd 2 complex (A-P and B-P). The calculated free energies of P indicated that the modeled reaction pathways A and B are both thermodynamically favorable, by 19.5 and 17.7 kcal/mol each. For all species in the two pathways, the trans-κ2-μ2 isomer showed the lowest free energies among the different geometrical isomers (see 3.4.7.2). Notably, therefore, our results suggest no ligand rearrangement during the C–H activation reaction. The two reaction pathways, along with the optimized structures and their free energies, are presented in Figure 3.11.

For all Pd complexes in the two reaction pathways, all possible spin states were also examined (see 3.4.7.2 for the full details). The reactant (R) and product (A-P and B-P) were determined to have singlet multiplicities. For transition state (A-TS and B-TS), triplet and open-shell singlet states were favored over closed-shell singlet by ~9.0 kcal/mol for A-TS and 12.0 kcal/mol for B-TS, respectively, reflecting the radical character of the reaction predicted from the foregoing experimental and computational results. The two open-shell states had very similar geometries and were very close in energy; triplet was favored relative to open-shell singlet only by 0.7 and 0.6 kcal/mol for A-TS and B-TS, respectively. Combined with the fact that Pd is a relatively heavy element, the rate of flipping spins at the TS is expected to be fast and not to impact the overall reaction rate significantly. As for the intermediate with partial ligand dissociation prior to H abstraction (B-Int1), closed-shell singlet was the most stable, but only by 1.5

II,III kcal/mol compared to the triplet state. The Pd2 intermediates after H abstraction (A-Int2 and B-Int2) had doublet multiplicities as expected. Overall, free energies of the different spin states calculated for both pathways were consistent with the proposed mechanism.

Focusing on the transition states (A-TS and B-TS), calculated activation free energies and kinetic isotope effects showed good agreement with experimental results for both pathways A and B. Several conformations of the TS with different positions of methane relative to Pd were optimized to identify the most stable structure. The free energies for all the calculated TS conformations were very close and not impacted significantly by the position of the methane molecule; for example, the free energy difference between the two conformations of A-TS in Figure 3.11 is ~0.2 kcal/mol. The resultant activation free energies of pathways A and B calculated from the most stable TS conformations were 27.3 kcal/mol and 25.1 kcal/mol, respectively, which are different by only 2.2 kcal/mol. Moreover, these values are in good

‡ agreement with the experimentally measured value of ΔG ha = 20.7(±0.3) kcal/mol (see 3.4.4.3 for the Eyring analysis). Since the rate of MBS generation is equivalent to the rate of C–H activation under steady-

90 state approximation (equation 3.7), the values of kMBS measured at variable temperatures were treated with transition state theory to give the value of ΔG‡ for C–H activation or, equivalently, H atom abstraction by

III Pd 2. KIEs calculated for CD4 versus CH4 likewise showed good agreement with experimental results. Computational KIEs were obtained with a Python version of Kinisot program32 which uses Gaussian output files of the reactants and transition states. From analysis of the computed vibrational frequencies at the TS using the Bigeleisen-Mayer equation,33 pathways A and B gave KIEs of 5.9 and 7.8, respectively. The substantial KIE is qualitatively consistent with the experimental value, 4.1(±0.6). Therefore, transition states consistent with experimental data could be located for both pathways in Figure 3.11, lending support for our DFT model of reaction 3.15.

Figure 3.11. BLYP/SDD,6-311++G(d,p)/SMD-water//BLYP/SDD,6-31+G(d) computed reaction pathways • II,III for H atom abstraction from CH4 and recombination of CH3 with Pd2 . Numbers indicate free energies

91 in kcal/mol for the most stable conformer. Calculations were done at 323 K to match the experimental conditions.

3.2.2.5. Further investigation of ionic/organometallic C–H activation The ionic pathway for C–H activation, proposed in equation 3.9, is unlikely if heterolytic ligand dissociation from the PdIII center has a high barrier. However, the results in Table 3.1 suggest that the calculated ligand dissociation free energies have a large uncertainty due to our inability to properly model solvation and that it would be unwise to rule out the ionic pathway solely on the basis of these values. For example, the highly ionic sulfuric acid solvent with extensive hydrogen bonding may assist the heterolytic

III dissociation of a bisulfate anion from Pd 2 and dramatically lower the free energy for ligand dissociation.

Therefore, we further investigated the viability of the ionic pathway by computing the C–H activation reaction that would follow ligand dissociation (i.e., the second reaction in equation 3.9). Shown in Figure 3.12 are several pathways of organometallic C–H activation following partial ligand dissociation calculated for locating transition states. The starting complex for pathways shown in Figure 3.12a and b was identical to B-Int1 in Figure 3.11, as different localization of charge cannot be distinguished in the computational input structure. The free energy of the transition state for this pathway, regardless of the position of the incoming methyl group, was much higher than that for H atom abstraction. The transition state in which the methyl group is located in the axial position was 35.2 kcal/mol above the starting complex, and the transition state in which the methyl group is located in the equatorial position had a free energy of 40.3 kcal/mol (Figure 3.13). Both of these ΔG‡ values are at least 10 kcal/mol higher than that for H atom abstraction, which was 25.1 kcal/mol. As for the pathways shown in Figure 3.12b, we failed to locate a viable transition state for both axial and equatorial methyl binding. In fact, we observed a tendency of the methyl group to move away from the metal, resembling the stepwise H atom abstraction/methyl radical recombination pathway calculated earlier. The pathway shown in Figure 3.12c, starts from a tautomeric

III 2– form of the neutral Pd 2 complex, and C–H activation is assisted by a more basic SO4 ligand. Even for this pathway, the activation free energies were much higher than the radical pathway, 36.2 kcal/mol and 33.7 kcal/mol for axial and equatorial methyl binding, respectively. While we are continuing our computational efforts to locate a lower transition state for the ionic pathway, from these and other results

III presented here, we conclude that Pd 2 likely favors a radical pathway for C–H activation of methane.

Figure 3.12. Ionic/organometallic C–H activation pathways (i.e., concerted C–H cleavage/Pd–C bond formation) for which transition states were calculated.

Figure 3.13. Optimized structures for the transition state for the pathway shown in Figure 3.12a. ∆G‡ = 35.2 kcal/mol (left), ∆G‡ = 40.3 kcal/mol (right).

3.2.3. Discussions

3.2.3.1. Nature of the H atom abstraction reaction

III The foregoing results show that the Pd 2 complex in fuming sulfuric acid reacts with methane by H atom abstraction. This reactivity may be understood as a concerted proton-coupled electron transfer

II,III (PCET) reaction. Since H atom abstraction from methane yields the Pd2 mixed-valent complex according to Figure 3.7, the electron from the methane molecule may be going to a metal-based orbital, most likely the σ* antibonding orbital of the Pd–Pd bond. On the other hand, the proton from the methane molecule is probably picked up by the surrounding solvent molecules or the sulfate ligands on the Pd complex. We

III postulate that this reaction is driven by the high redox potential of Pd 2 that facilitates outer sphere electron

III transfer (OSET) from methane to Pd 2. While a sequential electron transfer-proton transfer (ET-PT)

III pathway is energetically unfavorable (ET intermediate is 40.5 kcal/mol above the starting Pd 2 complex), the asynchronicity factor for the reaction was calculated to be ~2000 mV,34 which means that the transition state has a larger electron transfer character than proton transfer. Therefore, the outer-sphere PCET pathway

III for methane activation by Pd 2 may be rationalized on the basis of the high oxidation potential of electro-

III generated Pd 2.

III 3.2.3.2. The selectivity and rate of methane functionalization by Pd 2 It has been shown previously that various metal ions yield MBS and MSA from methane in fuming sulfuric acid with varying rates and selectivities.35 While it was recognized that MSA is formed from radical intermediates, the molecular mechanism of methane oxidation and sulfonation in fuming sulfuric acid has not been well understood. For example, it has long been debated whether HgII ions react with methane by C–H activation or H atom abstraction.22,36 Therefore, our ability to predict and control the reaction of methane in fuming sulfuric acid by different catalysts and additives has been severely limited, in spite of the efficient and selective methane functionalization that may be possible in this superacid medium.37 Our detailed mechanistic studies reveal several insights towards this goal.

On selectivity, it is notable that selective mono-functionalization of methane is possible even if radical intermediates are involved. The selectivity arises from the polarity effect; that is, electrophiles react with nucleophilic radicals faster than with electrophilic radicals, and vice versa.31 We attribute the high

III selectivity observed from the reaction of Pd 2 with methane to this polarity effect. However, while it is often argued that H atom abstraction from MBS and MSA by the electrophilic methanesulfonyl radical

• (CH3SO3 ) is suppressed because of the electron-withdrawing effect of the –OSO3H and –SO3H groups, previous computational studies show that the activation barrier for H atom abstraction is not significantly influenced by the presence of these electron-withdrawing substituents.36 We think that the actual reaction

• • that is suppressed may be SO3 addition to CH2OSO3H or CH2SO3H radicals, as both SO3 and these radicals are electrophilic. Also, in general, these electrophilic radicals will be stable in the superacid medium that is

• devoid of nucleophiles except for methane that is added. Moreover, SO3 addition to CH3 is already the slower reaction among the two radical propagation reactions, presumably due to the conjugation of SO3 with the sulfuric acid solvent to exist as H2S2O7 or other extended forms (see 3.2.1.4).

II,III Another factor that may contribute to selectivity is radical capture by the Pd2 complex. We

III observed weaker inhibitory effect of O2 on the MSA formation rate for Pd 2-initiated case compared to the peroxydisulfate-initiated case. Furthermore, the decrease in MSA formation was nearly quantitatively compensated by the increase in MBS formation. On the other hand, peroxydisulfate-initiated methane sulfonation showed not only dramatic suppression of MSA generation but also slight overoxidation in the 94

II,III presence of O2. We tentatively attribute these differences to the ability of Pd2 to capture radical

• II,III intermediates. First, since CH3 capture by Pd2 results in the formation of a functionalized product, MBS,

II,III the Pd2 plays a protective role against the deleterious radical scavenger O2, resulting in weaker

III suppression of MSA generation when initiated by Pd 2 compared to when initiated by peroxydisulfate.

Similarly, to explain the compensatory increase in MBS production under O2 environment for methane

III oxidation by Pd 2, we postulate that the partially oxidized methyl intermediate from reaction with O2 is

II,III captured by Pd2 and prevented from further oxidation.

On reaction rate, our results imply that an H atom abstraction pathway, as opposed to an ionic/organometallic pathway, tends to exhibit higher reaction rates for the activation of methane C–H bond. While we do not have a definite evidence to rule out the ionic/organometallic pathway altogether, our experimental and computational results support a single radical pathway for methane functionalization by

III • Pd 2 to MBS and MSA. Most notably, the rate of H abstraction by CH3SO3 was found to be faster than the

• rate of SO3 addition to CH3 , supporting the notion that H atom abstraction from methane can be very fast, even though it involves the cleavage of the strongest C–H bond among sp3 C–H bonds. Additionally, organometallic C–H activation involving metal-carbon bond formation would require a greater structural organization at the transition state compared to outer-sphere PCET. Considering the highly ionic and hydrogen-bonded solvent medium, the large reorganization energy associated with the ionic pathway may have contributed to the preference for the observed H atom abstraction pathway.

3.3. Conclusions

III We examined the reaction mechanism of methane functionalization by Pd 2 is through experimental and computational work. Reaction rates were measured with NMR spectroscopy to establish rate laws, obtain kinetic isotope effects (KIE) and measure activation barriers for the two products, methyl bisulfate (MBS) and methanesulfonic acid (MSA). Involvement of radical intermediates for the formation

II,III of MSA was diagnosed from its suppression by O2, and the mixed-valent Pd2 complex was shown to also behave as a radical scavenger. From kinetic studies of peroxydisulfate-initiated methane sulfonation,

• we concluded that CH3 is the major radical species in the solution. The information led us to propose a mechanistic model that features product bifurcation from a common methyl radical intermediate that is

III produced from rate-limiting H atom abstraction by the Pd 2 complex. This surprising radical-based C–H activation is supported by DFT computations that reveal a preference of the complex for homolytic ligand dissociation over heterolytic, and an energetically viable pathway for the proposed H atom abstraction and methyl radical rebound reactions. Altogether, our detailed mechanistic investigation of the dual reactivity

III of Pd 2 in fuming sulfuric acid shows that H atom abstraction or outer-sphere PCET pathways may lead to both net two-electron oxidation and radical chain sulfonation at exceptionally high reaction rates.

3.4. Methods and Additional Information

3.4.1. General methods

3.4.1.1. Chemicals and materials

In addition to the chemicals and materials mentioned in Chapter 2, (NH4)2SO4 (99.999%, Alfa

Aesar), potassium hydrogen phthalate (≥99.95%, Sigma-Aldrich), CH4 (Ultra High Purity grade, Airgas),

O2 (Industrial grade, Airgas) and CD4 (99 atom% D, Aldrich) were used as received. Clean fuming sulfuric acid was prepared from 65% SO3 fuming sulfuric acid (Sigma-Aldrich), and all concentrated or fuming sulfuric acid samples were contacted with glass or PTFE only except for the FTO and Pt electrodes during electrolysis, as described in Chapter 2.

Caution! Concentrated sulfuric acid, fuming sulfuric acid and SO3 are extremely corrosive.

Addition of water incurs a violent exothermic reaction. Fuming sulfuric acid and SO3 have high vapor pressures. Handle them with caution inside a fume hood.

III 3.4.1.2. Preparation of Pd 2 solutions

III II Fuming sulfuric acid solutions of Pd 2 were prepared by bulk electrolysis of Pd solutions following procedures described in Chapter 2 but using an improved electrochemical cell design. FTO was used as the working electrode, and a Pt wire and a Pt mesh, each encased in a glass tube with a fritted tip, were used as the reference and counter electrodes, respectively. All three electrodes were connected, at the top, to thin and flexible Pt wires (prepared from twisting together two thin wires of 0.127 mm dia.) as electrical connections that resist corrosion from the SO3 vapor inside the cell. The electrochemical cell was a tall 40 mL vial (28 mm dia.) charged with a spinfin PTFE stir bar. After adding the PdII solution and electrodes, a custom-made PTFE plug with grooves to accommodate the thin Pt wires was snugly fitted to the mouth of the vial and wrapped with PTFE tape and parafilm. It was to maintain the SO3 concentration during electrolysis that took ~2 days by minimizing SO3 evaporation and absorption of ambient water.

Measurement of SO3% before and after electrolysis showed minimal changes of 0–3%. After confirming

[ ] the completion of electrolysis by UV–Vis spectroscopy ( ≥ 93%), the solution was collected [ ][ ] into several 4 mL glass vials (with caps lined with PTFE septum; CG 4904-06, Chemglass). The solutions were immediately transferred into a fridge at –26 ̊C and kept frozen until use. No more than 1 mL of solution

96 was added to each vial, as larger volumes of sample could break the vial upon volume expansion from freezing. Once a vial was thawed and opened, it was not reused for kinetic studies.

To each solution prior to electrolysis, 3–7 mM of ammonium sulfate ((NH4)2SO4) was added as

III an internal standard for NMR concentration measurements. Given the highly oxidizing nature of Pd 2, we searched for a chemically inert compound for use as our 1H NMR integration standard. We verified that

III ammonium ions are not oxidized by the solvent, Pd 2, or the polarized electrode during bulk electrolysis of PdII. We also verified that possible exchange of the ammonium protons with the solvent does not affect the concentration measurements, presumably because it is slow compared to the timescale of NMR measurements (see 3.4.2.1).

III II For the determination of Pd 2 and SO3 orders, solutions containing different concentrations of Pd

II III and SO3 were prepared and subjected to bulk electrolysis (Table 3.3). Measurements of [Pd ] and [Pd 2] were carried out by UV–Vis spectroscopy as described in Chapter 2. [SO3] was measured from titration as

III III described in 3.4.1.3. For the determination of CH4 order, the same Pd 2 solution (Pd 2] = 2.5 mM, [SO3] = 2.6 M) was pressurized with different amounts of methane.

III Table 3.3. Pd 2 samples for reaction order studies.

III Samples for [Pd 2] variation Samples for [SO3] variation

III ([SO3] = 2.6 M) ([Pd 2] = 2.5 mM) Conc. 1 0.7 1.3 Conc. 2 1.2 2.6 Conc. 3 2.5 4.5 Conc. 4 4.3 –

3.4.1.3. Determination of [SO3]

Determination of SO3%

Fuming sulfuric acid solutions containing excess SO3 is typically described by SO3%, i.e., the weight% of free SO3. Following the acid-base titration procedure described here, SO3% could be measured within 1%. For example, four consecutive measurements of the same sample gave 9.19%, 8.55%, 9.15%, and 8.46%.

To a pre-weighed 20 mL reaction vial (with open-top caps lined with PTFE septum; CG-4904-01, Chemglass), ~8 drops of fuming sulfuric acid sample solution was added with a disposable glass pipette.

After immediately capping the vial to minimize the loss of SO3 vapor, the vial was weighed again to obtain the weight of the acid. ~5 mL of water was then quickly injected into the tightly capped vial with a needle

97 and syringe, and the vial was thoroughly agitated until the SO3 fumes inside the vial cleared away. The needle and the syringe, attached to the vial with its plunger fully pushed in, was kept that way during agitation of the vial by holding them firmly together. Afterwards, just the plunger was removed by pulling it out, and 2–3 mL of water was added to rinse the inside of the syringe. Then, the plunger was inserted back in to the syringe body to inject the rinsate into the vial. To remove the needle and syringe from the vial, we first pulled out the plunger and then removed the needle and syringe body to minimize liquid leakage through the septum hole while the needle is being pulled out.

To this sample vial with the punctured cap, a pea-shaped PTFE stir bar was added. After weighing the vial, a standardized NaOH solution was added until equivalence point was reached. A standardized HCl solution was added when too much NaOH was added by mistake. Methyl red (2-(N,N-dimethyl-4- aminophenyl) azobenzenecarboxylic acid, pKa = 5.1) was used as the indicator because its pKa was lower

– + than that of CO2 (6.3 for CO2 + H2O ⇄ HCO3 + H ). Using methyl red, consistent results could be achieved even when the NaOH solution absorbed atmospheric CO2 over time. The drastic pH change that accompanies titration of a strong acid with a strong base allows flexibility in the choice of indicator pKa. The amounts of NaOH and HCl added were obtained by vial weight measurements. The NaOH solution was standardized by titration of a solution of potassium hydrogen phthalate (KHP), which is a primary titration standard, with phenolphthalein (3,3-Bis(4-hydroxyphenyl)-2-benzofuran-1(3H)-one) as the indicator. The HCl solution was then standardized by titration with the standardized NaOH solution. Subsequent batches of NaOH solutions were standardized with the standardized HCl solution. Concentrations of NaOH and HCl solutions used were ~0.4 and ~0.2 m (=mol/kg), respectively.

Critical to the reliability of the titration results were accurate weight measurements. To minimize the influence of static electricity, the vials were wrapped with an aluminum foil and moved in and out of the weighing area using a pair of large metal tweezers. A typical analytical balance (Mettler Toledo) that reads up to 0.1 mg was used.

To calculate SO3%, the following equation was used to first obtain H2SO4% with the following equation:

98.08 g/mol × (mol − mol ) 2 HSO% = … 3.16 g

98.08 g/mol is the molecular weight of H2SO4, molNaOH and molHCl are the weight of NaOH and

HCl solutions multiplied by their concentrations, and gsample is the weight of the fuming sulfuric acid sample.

The conversion of H2SO4% to SO3% is explained below.

Definition of H2SO4%, SO3%, [SO3], and conversion between these units

H2SO4 is the hydrated form of SO3 (= H2O + SO3), and H2SO4% denotes the weight% of H2SO4 units in a given solution. Therefore,

( ) 80 + 18 g/mol × 푛 HSO% = … 3.17 80 g/mol × 푛 + 18 g/mol × 푛

where 80 and 18 are the molar mass of SO3 and H2O, and nSO3 and nH2O stand for the total number of moles of SO3 and H2O, respectively. When SO3 is in excess, the equation assumes that all SO3 were hydrated so that H2SO4% exceeds 100%. When water is in excess, H2SO4% is less than 100%.

Writing SO3% in terms of nSO3 and nH2O,

80 g/mol × (푛 − 푛) SO% = … 3.18 80 g/mol × 푛 + 18 g/mol × 푛

Therefore,

80 SO % = × (H SO % − 1) 18 … 3.19

The nominal molar concentration of free SO3, [SO3], is simply:

SO % [SO ] = × 푑 80 g/mol … 3.20

where d is the density of the solution. The value 1.92 kg/L was used for d.

3.4.1.4. Preparation of NMR samples for reaction rate measurements For carrying out methane oxidation in NMR tubes, heavy-walled NMR tubes with PTFE valves

III (S-5-500-HW-7, Norell) were charged with solutions of electrogenerated Pd 2, pressurized with methane, and thoroughly agitated to dissolve methane into the liquid phase. The solution height was 5 cm (190 μL). These preparations were done at room temperature in ambient. The o-ring on the PTFE valve of the heavy- walled NMR tube wore out over time and required replacement (1 mm wide, 4.5 mm ID).

Two methods of pressurizing the tube were used. First, the tube was directly connected to a controlled-pressure methane outlet. After >5 times of pressurizing and venting cycles, the valve was closed

III and detached from the methane outlet. Second, a tube charged with Pd 2 solution was connected to a syringe charged with a defined volume of the desired gas. The syringe was connected to the valve of the tube via a Luer-to-Swage stainless steel adaptor. The valve attached to the tube was closed tightly before cooling; otherwise, the tube cracked. After the tube was immersed in liquid nitrogen and cooled down, the valve was opened, allowing the gas in the syringe to be sucked in. The valve was then closed tightly again and the tube was removed from liquid nitrogen and allowed to warm up slowly. The second method using

99 liquid nitrogen was employed in the determination of KIE and in the O2 co-addition experiment, as our lecture bottle of CD4 did not have sufficient outlet pressure and two different gases had to be added to the tube. Though the choice of pressurization method did not significantly affect the measured rates, we used the same pressurization method when comparing two reactions side-by-side (e.g. CD4 versus CH4, O2 added versus no-O2).

After pressurization, dissolution of methane in the gas phase into the liquid phase was achieved by a careful tapping motion that created bubbles inside the tube. Sonication was ineffective. We confirmed equilibration of methane in the gas phase with the solution phase after ~5 minutes of agitation.

3.4.2. NMR spectroscopy

3.4.2.1. Technical details of acquisition Instrumentation 1H NMR was acquired either on a Bruker Avance Neo spectrometer operating at 500.18 MHz or a Bruker Avance-III HD Nanobay spectrometer operating at 400.13 MHz equipped with a 5 mm liquid- nitrogen cooled Prodigy broad band observe (BBO) cryoprobe. 2H NMRs were always acquired on the latter instrument, using the lock channel as the observe probe.

Since the samples were prepared in protic sulfuric acid, all acquisitions (both 1H and 2H) were done without locking. Comparison with samples prepared with ~20wt.% D2SO4 confirmed that locking did not affect the NMR measurements. Instrument drift in terms of measured chemical shifts was unnoticeable in the duration of our acquisitions (~2 h at maximum). Automatic gradient shimming was performed with the 1H signal.

Acquisition and data processing details For 1H spectrum, excitation sculpting (pulse program in Bruker: zgesgp) was used to suppress the solvent peak that appeared at 10–11 ppm (referenced with respect to methane at ~0 ppm). Pulse width was calibrated each day before the NMR experiment for reliable quantitation. The pulse width in concentrated or fuming sulfuric acid was 1.2–1.6 times longer than that in organic solvents. For 2H spectrum, a simple 45˚ pulse was applied. Relaxation delay time was determined from measurements of the integration area of the methane peak relative to that of MSA (for 1H NMR) or d4-acetic acid (for 2H NMR) at different delay times (Figure 3.14). Methane shows the slowest nuclear spin relaxation among other compounds of interest because its high symmetry and lack of a dipole moment slow down the spin relaxation; therefore, insufficient delay times give lower integration ratios. From the data in Figure 3.14, we chose acquisition time = 3 s and relaxation delay = 37 s for 1H and acquisition time = 2 s and relaxation delay = 38 s for 2H NMR measurements. The relaxation delays are particularly long because our fuming sulfuric acid medium

100 is highly viscous and contains very low levels of paramagnetic transition metal impurities. Also, relaxation delays were measured at 50 ̊C (the temperature at which real time measurements were performed) as well as at room temperature (Figure 3.14) because nuclear spin relaxation is generally slowed down with a decrease in solvent viscosity at elevated temperatures.

Figure 3.14. Ratios of integration area measured using different delay times for determining proper relaxation delay for accurate quantitation of methane. AQ: acquisition time, D1: relaxation delay. Number of scans = 4 and 8 for 1H and 2H, respectively. Based on these data, AQ + D1 = 40 s was chosen for actual measurements for both 1H and 2H.

Data processing was performed with the software MestReNova. For 1H NMR, phase correction and baseline correction by Whittaker smoother at 100 Hz was applied. For 2H NMR spectra, which had a noisier baseline, baseline correction by Whittaker smoother with an automatically selected frequency was used. Reliability of NMR integration for 2H NMR was tested by preparing a sample with known

6 3 4 concentrations of d -DMSO, d -MBS, and d -CH4 (estimated from [CH4] prepared identically). The

3 4 measured concentrations of d -MBS and d -CH4 were –5% and +11% compared to the known concentrations. Representative spectra from 1H and 2H NMR with baselines are shown in Figure 3.15.

A small unassignable peak, originating from the fuming sulfuric acid solvent, always appeared at ~0.8 ppm in the 1H NMR spectrum. The concentration of this impurity by number of protons was <1 mM. It was not affected by the reaction and we were unable to remove it or identify it.

101

Figure 3.15. Representative 1H and 2H NMR spectra from methane oxidation experiments, shown along with baselines and integration regions.

+ 1 Validity of using NH4 ions as an internal H NMR integration standard

+ Concentrations of methane, MBS, and MSA were obtained using NH4 ions as an internal integration standard, which were added to the Pd solutions prior to bulk electrolysis (see 3.4.1.2). Since the

+ protons in the NH4 ions can exchange with the solvent protons whose NMR peak is suppressed, we

+ investigated whether the NMR signal from NH4 is quantitative. Two solutions of fuming sulfuric acid containing different amounts of ammonium sulfate and methanesulfonic acid (MSA) were prepared from stock solutions (~0.5 M) of the two compounds in fuming sulfuric acid. Weight measurements of the sample

+ weight vials during sample preparation yielded mole ratios of NH4 and MSA in these samples (R ). The ratios were compared with the ratios of their NMR integration area (RNMR). Very close correlation between the

+ values from two independently prepared samples (Table 3.4) give us confidence that NH4 is a faithful

+ + NMR integration standard under our measurement conditions; presumably, H exchange between the NH4

– 2– ions and basic species in the solvent (e.g., HSO4 and SO4 ions) is slow enough thanks to the large pKa difference.

+ Table 3.4. Validation of using NH4 as an internal NMR integration standard.

Sample 1 Sample 2

Rweight 0.090 0.460 RNMR 0.059 0.299 RNMR/Rweight 0.647 0.650

3.4.2.2. Control studies for rate measurements with NMR For reliable rate measurements, we checked for the presence of possible side reactions or processes, i.e., those other than the generation of MBS and MSA from methane. 102

Exchange of dissolved methane with the gas phase Fast exchange of methane between the gas and liquid phases would convolute our measurements because methane can either (i) be replenished during the reaction from the gas phase, as it is being consumed by the reaction, or (ii) evaporate, given that methane is dissolved into the liquid phase at room temperature but reaction is carried out at elevated temperatures. However, we confirmed that measurable exchange of methane between the gas and liquid phases occurs in either direction inside the constricted NMR tube even with heating for a prolonged time. A tube containing fuming sulfuric acid was pressurized with methane but not agitated to dissolve in the gas. After heating for 1 h at 110 ̊C, no methane was detected by NMR. Subsequently, the tube was mechanically agitated to equilibrate the gas and liquid phases at room temperature and heated for 2 h at 110 ̊C. The concentration of methane was 13.5 and 13.9 mM before and after the heating procedure. Therefore, we conclude that any change in methane concentration in our NMR

III tubes must arise from its reaction with Pd 2; this observation is different from our previous study using aqueous solutions38 and we attribute it to the much higher viscosity of fuming sulfuric acid.

III Spontaneous decay of Pd 2 in the absence of substrate

III II The highly oxidizing Pd 2 complex undergoes spontaneous reduction to Pd upon heating,

– presumably by oxidizing the sulfuric acid solvent (H2SO4 → ½ O2 + SO3 + 2 e ). The decay is greatly

III accelerated in non-fuming sulfuric acid. The stability of Pd 2 is also critically lowered by the presence of transition metal impurities, which are abundant in commercial grade fuming sulfuric acid. These impurities were removed in the present study (see 3.4.1.1).

III We found that the rate of spontaneous decay of Pd 2 was slow enough to be ignored in our rate

III measurements. [Pd 2] decreased by 10% after 20 minutes at 50 ̊C and by 16% after 46 seconds at 80 ̊C. From this, first-order decay rate constants at different temperatures were estimated. For real time NMR reaction monitoring at 50 C̊ (section 3.4.3), only the first <400 seconds of data were analyzed for the

III extraction of initial rates, during which Pd 2 is predicted to spontaneously decay by ~3%. For ex situ rate

III measurements (section 3.4.3.3) at the highest temperature, Pd 2 is predicted to have spontaneously decayed by ~7% or less when the reaction was quenched (26–47 seconds at 72 ̊C), and at the lowest temperature, by

~2% (1070–1160 seconds at 40 ̊C). For ex situ rate measurement with CD4 at 50 ̊C, longer reaction time

III (1190–1410 seconds) was required, but only 10–12% of Pd 2 decay was predicted by the time of quenching

III the reaction. Therefore, we did not take the Pd 2 decay reaction into account in our kinetic analyses.

Further reaction of products Overoxidation of MBS and MSA was ruled out on the basis of the observed conservation of total carbon concentration (Figure 3.1). Conversion of MSA to MBS by SO3 (CH3SO3H + SO3 → CH3OSO3H

103

22 + SO2) would also be too slow under the conditions employed in this study. Therefore, no further reaction of products was considered.

3.4.3. In situ NMR for reaction rate measurements

3.4.3.1. Experimental procedure First, temperature of the NMR spectrometer was set to 50 ˚C with a “dummy” NMR tube charged with the same amount of fuming sulfuric acid solution as the actual sample. When the spectrometer reached the set temperature, probe tuning/shimming/pulse width calibration were performed. After setting up acquisition parameters, the sample tube was inserted and spectra acquisition was initiated. Consecutive single-scan NMR spectra were automatically acquired with pre-acquisition delay (pad) set to match the required relaxation delay time. Measurements with ethylene glycol, a temperature standard whose chemical shift values report the temperature, showed that the tube temperature reaches 45 ˚C in ~10 s and 50 ˚C in ~30 s after insertion (Figure 3.16).

Figure 3.16. Temperature of the NMR tube after insertion assessed with ethylene glycol.

3.4.3.2. Data processing The NMR spectra were processed as described in 3.4.2.1 to produce concentration-time traces (Figure 3.1). These plots were further processed by smoothing and the initial portions were fitted to

III mathematical functions using OriginPro 2020. Since one molecule of Pd 2 is consumed for every molecule

III of MBS produced, we only used the data up to the point when [MBS] is 20% or less of the initial [Pd 2]. At first, we fitted the reaction traces with generic exponential functions, and found that the reaction order in CH4 was first-order for both MBS and MSA. Then, with this knowledge, we derived analytical expressions for the reaction rates that allowed us to reduce the number of fitting parameters for obtaining

III Pd 2 and SO3 orders.

Smoothing of the raw data

104

The raw concentration-time data (Figure 3.1) was smoothed by normalizing the total carbon

total total concentration, [C ]. As mentioned earlier, [C ] is the sum of [MBS], [MSA], and [CH4] and was constant throughout the reaction time. Therefore, we obtained the average of [Ctotal] over the entire reaction time

total total total (=[C ]ave), and multiplied [C ]ave/[C ]t to the MBS, MSA, and CH4 concentrations at each time point

total total ([C ]t is the value of [C ] at time = t). The concentration-time plot before and after this smoothing procedure is shown in Figure 3.17.

Figure 3.17. (Solid, faint symbols) Raw experimental concentration-time traces were (hollow symbols) smoothed by normalizing the total methyl concentration.

CH4 order: fitting with generic exponential functions A generic exponential function of the form y=퐴e⁄ +퐶 (the “ExpDec1” function in

OriginPro) was used to individually fit the concentration-time traces of MBS, MSA, and CH4. Then, from

⁄ [] =− e , the initial rate ( = ≡ rX,i) is − . Absolute values of the initial rates were plotted against initial methane concentration (i.e., total methyl concentration) as a log-log plot to give Figure 3.2. The error bars represent the standard errors obtained from the fitting (corrected for taking logarithm; i.e. if y = log x, then SEy = (SEx/x)/(ln 10)).

III Pd 2 and SO3 orders: fitting with two global parameters

With the knowledge that both MBS and MSA are first-order in CH4, and that ([MBS] + [MSA] +

[CH4]) is constant at all times, we can write the following expressions:

푑[MBS] = 푘 [CH ] … 3.21 푑푡 푑[MSA] = 푘 [CH ] … 3.22 푑푡

푑[CH] = −(푘 + 푘 )[CH ] … 3.23 푑푡

105

III where k1 and k2 are the observed rate constants that may include Pd 2 and SO3 terms. From equation 3.23,

[CH] = [CH] exp(−(푘 + 푘)푡) … 3.24

where [CH4]0 is the initial concentration of methane. Substituting equation 3.24 into equation 3.21 and integrating,

푑[MBS] = 푘 [CH ] exp(−(푘 + 푘 )푡) … 3.25 푑푡

푘[CH] [MBS] = {1 − exp(−(푘 + 푘)푡)} … 3.26 푘 + 푘

Analogously,

푘[CH] [MSA] = {1 − exp(−(푘 + 푘)푡)} … 3.27 푘 + 푘

Therefore, MBS, MSA, and CH4 concentrations can be described using only two unknown parameters that are shared, k1 and k2. Performing a global fit in OriginPro, we obtained their values for each set of concentration-time traces. These rate constants are not the true rate constants but the observed rate

III constants for a given Pd 2 and SO3 concentration (i.e., rate constants for a pseudo-first order reaction in

III III CH4). They are plotted as a log-log plot against Pd 2 or SO3 concentrations in Figure 3.2. For each Pd 2

III sample at a given Pd 2 and SO3 concentrations, three independent experiments were performed. The error bars come from the larger of either the standard deviation among the independent experiments (SE=SD(a, b, c)/√3, where SE and SD stand for standard error and standard deviation) or the standard errors of fitting

2 2 2 from each experiment (SE=√(SEa + SEb + SEc )/3). To plot the standard error on the log-log graph, the above-mentioned formula was used.

II,III III 3.4.3.3. The effect of Pd2 on methane oxidation by Pd 2

II,III In order to diagnose the effect of the mixed-valent dimer Pd2 on the kinetics of methane

III III III oxidation by Pd 2, two Pd 2 solutions (10% SO3) were prepared at similar concentrations of Pd 2 but

II,III II,III II III different concentrations of Pd2 . Since Pd2 is formed from comproportionation between Pd and Pd 2,

II,III II III the solution with high [Pd2 ] was prepared by adding a Pd solution to a Pd 2 solution and letting the

II III mixed solution equilibrate overnight at room temperature. These mixed solutions of Pd and Pd 2 were prepared in the same way as the variable ox.% EPR samples in Chapter 2. The reaction with methane was monitored by NMR and the observed rate constants for MBS and MSA generation were extracted using the

III procedure described above (3.4.3.2, fitting with two global parameters for obtaining Pd 2 and SO3 orders).

106

The recorded concentration-time trace is presented in Figure 3.5, and the concentrations of the Pd solutions and the measured values of kobs are shown in Table 3.5.

Table 3.5. Observed rate constants for MBS and MSA generation from solutions containing different II,III concentrations of Pd2 . The rate constants were obtained from three independent reactions.

III II II,III a –1 –1 [Pd 2] (mM) [Pd ] (mM) [Pd2 ] (mM) kMBS,obs (s ) kMSA,obs (s ) Solution 1 1.2 0.1 ~0.002 4.76(±0.43)⨯10–5 5.38(±0.31)⨯10–4 Solution 2 1.5 5.6 ~0.2 6.28(±1.45)⨯10–5 2.48(±0.15)⨯10–4 Ratio of observed rate constants Solution2 Solution1 1.32 0.46 (kX,obs / kX,obs ):

a II,III Pd2 concentrations are approximate because they are calculated from the comproportionation equilibrium II,III constant, which was obtained from fitting to EPR-quantified spin concentrations. The EPR-measured [Pd2 ], however, was significantly larger than the fitting curve at high PdIII% (data and possible explanations for this II,III deviation are given in Chapter 2). Therefore, the actual Pd2 concentration in the solution may have been significantly higher than the values shown here.

3.4.4. Ex situ measurements for extraction of rate constants

3.4.4.1. Experimental procedure Measurements of rate constants for MBS and MSA formation were performed from ex situ concentration measurements for the determination of KIEs and activation barriers. NMR tubes charged with

III Pd 2 solution and methane were prepared as described in 3.4.1.4 and immersed into oil baths set to desired temperatures. The oil baths were pre-heated with stirring on an IKA hot plate (C-MAG HS 7) and stabilized at the set temperature prior to the preparation of NMR samples. After the desired amount of time, the tubes were transferred to an ice bath. Concentrations were measured by NMR right after the reaction in order to minimize background reaction at room temperature.

III All ex situ measurements were performed with the same batch of Pd 2 solution containing 3.3

III mM of Pd 2 and 1.9 M of SO3. The reaction was quenched in the initial phase when ~5–7 % of methane was converted to MBS.

3.4.4.2. Microkinetic modeling for rate constant extraction Methane oxidation to MBS and MSA was simulated using the free software COPASI 4.25 with a phenomenological mechanism comprised of these two reaction equations:

CH + Pd ⎯⎯ CHOSOH + 2 Pd … 3.28

CH + SO ⎯⎯ CHSOH … 3.29

Simulation with this model showed a good fit with the experimental data (Figure 3.18). In order to obtain rate constants from ex situ concentration measurements, fitting was performed using the

107 concentrations of MBS, MSA, and methane at two time points, initial and final. The final concentrations were measured as described above. The initial concentrations of MBS and MSA were zero, and that of methane was the sum of final MBS, MSA, and methane concentrations. Since only two time points were used, the error associated with the COPASI-fitted parameters, i.e. the uncertainty of the fitting, were very small. Three or more experiments were performed for reactions at each temperature and five experiments were performed for reaction of CD4. The reported standard error is (standard deviation)/√N, where N is the number of measurements.

Figure 3.18. (Lines) COPASI simulation of the methane oxidation reactions fitted to (symbols) experimental concentration-time traces.

Comparison with the in situ method (3.4.3) To ensure the validity of the fitting method, an in situ time trace was analyzed by this ex situ rate constant extraction procedure. A single time point was picked from the concentration-time trace, and the MBS, MSA and methane concentrations at this time point were fitted to the microkinetic model as described above. Repeating this for five different time points, the rate constants were obtained as kMBS =

–2 –1 –1 –4 –1 –1 5.26(±0.23)⨯10 M s and kMSA = 3.22(±0.06)⨯10 M s . These compare well with the rate constants obtained from fitting exponential functions to the initial portion of the traces and then dividing the observed

III rate constants by Pd 2 and SO3 concentrations to convert them to actual rate constants: kMBS =

–2 –1 –1 –4 –1 –1 5.21(±0.25)⨯10 M s and kMSA = 3.38(±0.05)⨯10 M s . Therefore, the rate constant extraction procedure using the final concentrations and fitting with COPASI was found to be reliable.

When the rate constants measured from an ex situ experiment was compared with those measured

–2 –1 – from an in situ experiment, a larger but insignificant difference was found: kMBS = 7.35(±0.10)⨯10 M s

1 –4 –1 –1 and kMSA = 4.30(±0.11)⨯10 M s , measured from five ex situ measurements, are 41% and 27% larger

III than the rate constants obtained from in situ measurements using the same Pd 2 solution, kMBS =

–2 –1 –1 –4 –1 –1 5.21(±0.25)⨯10 M s and kMSA = 3.38(±0.05)⨯10 M s . A possible reason for this difference is a

108 potential difference between the oil bath temperature and the temperature reached inside the NMR

ex instrument. The difference between the logarithm of the in situ and ex situ rate constants (i.e. log (kX situ in situ /kX )), however, is only 0.15 and 0.10 for kMBS and kMSA, respectively. While this difference is larger than the standard errors, it is insignificant compared to the range spanned by the rate constants measured under different conditions in this work. Moreover, comparisons for the purpose of drawing scientific conclusions were only made between rate constants measured using the same method. Therefore, measurements from both methods were considered dependable.

3.4.4.3. Activation parameters from variable temperature measurements Rate constants measured at variable temperatures, plotted earlier as ln k vs T–1 to derive Arrhenius activation barriers (Figure 3.3), are plotted here as ln (k T–1) vs T–1 to derive activation parameters according to the transition state theory. Since activation parameters are thermodynamic properties of the transition state of elementary reactions, the analysis is only valid when the measured rate corresponds to an elementary reaction. The rate of MBS formation corresponds to the rate of the C–H activation step (see 3.4.5), which is an elementary reaction. The rate of MSA formation, however, would be a composite of the rates of the C–H activation and radical propagation steps. Thus, while activation parameters derived from measurements of kMBS correspond to those for the C–H activation reaction, activation parameters from kMSA cannot be assigned definite physical meanings.

Figure 3.19. Eyring plot and activation parameters derived from rate measurements at 40, 50, 60 and 72 ̊C. Each data point corresponds to an average of three or more measurements.

Comparison with the activation barrier determined by electrochemical methods As mentioned in 3.2.1.2, the activation barrier for MBS generation determined in our work at 40– 72 ̊C, 21.3 kcal/mol, was lower than the value determined previously by electrochemical methods at higher

14 –1 –1 temperatures (80–140 ̊C), 26.2 kcal/mol. The value of kMBS measured at 50 ̊C in this work (0.186 M s )

109 can be extrapolated to 140 C̊ using the lower and higher activation barriers to yield 252 M–1 s–1 and 1350

–1 –1 M s , respectively. In order to make a comparison with the experimental kMBS at 140 ̊C, we need the concentration of methane under the measurement conditions in our earlier work14 because the rates were measured as turnover frequency (TOF), which is equal to kMBS multiplied by [CH4]. The TOF measured for

III –1 the Pd 2-mediated electrocatalytic methane-to-MBS reaction at 140 ̊C and 500 psi CH4 was 2000 h . From

20,21 the literature, we estimate [CH4] under these conditions to be 2–20 mM. Table 3.6 shows that the TOF calculated using the lower activation energy agrees much better with the experimental TOF of 2000 h–1.

Table 3.6. TOF at 140 ̊C calculated by extrapolation from kMBS at 50 ̊C.

–1 Assumed Calculated TOF (h ) [CH4] (mM) From Ea = 20.3 kcal/mol From Ea = 26.2 kcal/mol 2 1810 9690 5 4530 24200 20 18100 96900

3.4.4.4. Reaction of CD4

Quantitation of CD4 and its reaction products

+ For CD4, we could not use ammonium ions as an internal standard because the D ions will be

+ III slowly exchanged to H ’s in the protic solvent. Because of the highly oxidizing property of Pd 2, we also wanted to avoid organic internal standards. Therefore, deuterated species were indirectly quantified without an internal standard by pressurizing the tubes with a consistent methane volume (7 mL; see 3.4.1.4 for

3 3 detailed procedures). The initial concentration of CD4, which should equal the sum of d -MBS, d -MSA and CD4 at the end of the reaction, was considered to be the same as that of CH4 pressurized in the same way, which was measured to be 16.5±0.7 mM (11 measurements). Solubility isotope effects exist but are likely negligible for our studies.39

H/D exchange

As shown in Figure 3.20, no H/D exchange was observed between CD4 or its reaction products and the protic sulfuric acid solvent under our conditions.

110

1 Figure 3.20. H NMR spectra of quenched NMR tube reactions with (red) CD4 and (green) CH4. Reaction 3 3 conditions: 50 ̊C, 435 seconds (CH4) or 1344 seconds (CD4). Concentrations: 1.1 mM d -MBS, 3.1 mM d - MSA, 12.0 mM CD4; 1.1 mM MBS, 4.3 mM MSA, 9.7 mM CH4.

3.4.5. Derivation of rate laws

3.4.5.1. Consideration of alternative radical chain termination reactions The mechanisms discussed (Figure 3.10 and Figure 3.7) only show radical chain termination via

• II,III • recombination of CH3 and Pd2 , based on the observation that CH3 is the majority free radical in the solution. However, for the sake of completeness, we also considered radical chain termination by the

• II,III recombination of CH3SO3 and Pd2 in the following sections. After showing the derivation for the case

• II,III where CH3 and Pd2 is the only radical recombination pathway, we show the case that includes the

• II,III reaction CH3SO3 + Pd2 , which yield rate laws for MSA formation that are inconsistent with the

• • experimental rate law. Incidentally, the radical recombination reaction CH3 + CH3SO3 was assumed to be negligible because of the low concentrations of these transient radicals; the second order reaction rates will be extremely low.

3.4.5.2. Ligand homolysis and H abstraction by free bisulfate radical (Figure 3.10)

[] From steady-state approximation, = 0 for all intermediates X. Therefore,

푑[CHPd ] , = 푘 Pd [CH∙ ] − 푘 [CH Pd ] = 0 … 3.30 푑푡

푑[CH∙ ] = 푘 [CH ][HSO∙ ] − 푘 Pd ,[CH∙ ] 푑푡 … 3.31 ∙ ∙ +푘[CH][CHSO] − 푘[SO][CH] = 0

[ ∙ ] 푑 CHSO ∙ ∙ = 푘 [SO ][CH ] − 푘 [CH ][CH SO ] = 0 … 3.32 푑푡

From 3.31 and 3.32,

111

∙ , ∙ 푘[CH][HSO] = 푘Pd [CH] … 3.33

Therefore, from 3.30 and 3.33,

∙ 풓 = 푘[CH][HSO] … 3.34

Again, from steady-state approximation,

∙ 푑[HSO] , = 푘 [Pd ] − 푘 Pd [HSO∙ ] − 푘 [CH ][HSO∙ ] = 0 … 3.35 푑푡

∙ Since C–H cleavage is rate-determining, we assume that 푘[CH][HSO] is much lower than , ∙ 푘[Pd ] and 푘 Pd [HSO]. Therefore,

∙ 푘[Pd ] [HSO] ≈ , … 3.36 푘Pd

Combining 3.34 and 3.36,

푘푘 [Pd ] 풓 = [CH] , … 3.37 푘 [Pd ]

For MSA generation,

∙ ∙ 풓 = 푘[SO][CH] or 푘[CH][CHSO] … 3.38

From 3.33 and 3.36,

[ ][ ∙ ] [ ] ∙ 푘 CH HSO 푘푘 CH [Pd ] [CH] = ≈ 푘 Pd , , … 3.39 푘푘Pd

Therefore,

푘푘푘[CH][SO][Pd ] 풓 ≈ , … 3.40 푘푘Pd

To consider the alternative radical recombination pathway (3.4.5.1), equation 3.32 is now:

[ ∙ ] 푑 CHSO ∙ ∙ , ∙ = 푘 [SO ][CH ] − 푘 [CH ][CH SO ] − 푘 Pd [CH SO ] = 0 … 3.41 푑푡

Therefore, substituting 3.41 into 3.31,

∙ , ∙ , ∙ 푘[CH][HSO] = 푘Pd [CH] + 푘Pd [CHSO] = 0 … 3.42

112

For the derivation, we assume the extreme case where radical chain termination mostly occurs

• II,III • II,III through the recombination of CH3SO3 with Pd2 rather than that of CH3 with Pd2 , i.e. , ∙ , ∙ 푘Pd [CH] ≪ 푘Pd [CHSO]:

∙ , ∙ 푘[CH][HSO] ≈ 푘Pd [CHSO] … 3.43

[ ][ ∙ ] [ ] ∙ 푘 CH HSO 푘푘 CH [Pd ] [CHSO] ≈ = 푘 Pd , , … 3.44 푘푘Pd

From equation 3.38,

푘푘푘[CH] [Pd ] 풓 ≈ , … 3.45 푘푘Pd

The second order dependence on CH4 is not consistent with the clean first-order dependence observed experimentally.

III 3.4.5.3. H atom abstraction by Pd 2 (Figure 3.7)

∙ Equations from the preceding section hold true except for the substitution of the 푘[CH][HSO] , ∙ term by 푘[CH][Pd ] and the irrelevance of the terms 푘[Pd ] and 푘Pd [HSO] . Therefore,

풓 = 푘[CH][Pd ] … 3.46

For MSA,

[ ] ∙ 푘[Pd ] CH [CH] = , … 3.47 푘Pd

푘푘[CH][SO][Pd ] 풓 = , … 3.48 푘Pd

To consider the alternative radical recombination pathway (3.4.5.1), equation 3.42 becomes:

, ∙ , ∙ 푘[CH][Pd ] = 푘Pd [CH] + 푘Pd [CHSO] = 0 … 3.49

Following similar derivations,

푘푘[CH] [Pd ] 풓 ≈ , … 3.50 푘Pd

Again, the second order dependence on CH4 is not consistent with the clean first-order dependence observed experimentally.

113

3.4.6. Peroxydisulfate-initiated methane oxidation

3.4.6.1. Reaction orders The in situ NMR measurement technique described in this work was applied to a preliminary study of the kinetics of peroxydisulfate-initiated methane oxidation. Clean fuming sulfuric acid solutions containing three different concentrations of K2S2O8, CH4, and SO3 were reacted with methane inside NMR tubes at 50 ̊C with real-time concentration measurements. The sole product was methanesulfonic acid and

III total methyl concentration was preserved unless O2 was co-added. As with Pd 2, ammonium sulfate was used as the internal concentration standard.

The reaction time trace shows a slight induction period (Figure 3.4b), presumably due to the slow thermal decomposition of peroxydisulfate that is needed to set off the radical chain reaction. However, for all test solutions, the concentration-time trace becomes nearly linear at 750–1250 s, in accordance with the equation derived below (see Figure 3.22). Reaction rates were extracted as the slope of the concentration- time traces from this linear portion. The logarithm of the rate is plotted versus the logarithm of the initial concentration of K2S2O8, CH4, and SO3 in Figure 3.21.

Figure 3.21. The rate of peroxydisulfate-initiated methane oxidation to methanesulfonic acid at 50 ̊C measured as a function of the initial concentrations of (a) CH4, (b) K2S2O8, and (c) SO3.

3.4.6.2. Simulation of the reaction rates and concentration-time trace

2– •– • Denoting S2O8 = Int2 and SO4 = Int ,

⋅ Int 2 Int … 3.51

⋅ ⋅ Int + CH → CH + HI … 3.52

∙ ∙ CH + SO ⎯ CHSO … 3.53

∙ ∙ CHSO + CH ⎯ CHSOH + CH … 3.54

114

From the steady-state approximation,

∙ ∙ 풓 = 푘[SO][CH] = 푘[CH][CHSO] … 3.55

Assuming reaction 3.52 is rapid compared to the radical propagation reactions,

⋅ ∙ ∙ [Int ] ≪ [CH] 표푟 [CHSO] … 3.56

Applying the steady-state approximation to assume the total radical concentration is constant,

∙ ∙ ⋅ [CH] + [CHSO] = 푐표푛푠푡푎푛푡 = [R ] … 3.57

Now, by combining equations 3.55 and 3.57, the rate of MSA generation may be calculated by solving the following equation:

∙ ⋅ ∙ 풓 = 푘[SO][CH] = 푘[CH]([R ] − [CH]) … 3.58

Thus, the simulated values of rMSA in Figure 3.6 were calculated according to equation 3.58 under the CH4 and SO3 concentrations used in our experiments ([CH4] = 6–22 mM and [SO3] = 0.6–2.7 M). Regardless of the value of [R⋅] and the absolute magnitude of the radical propagation rate constants, the

푘 푘 ratio determined the dependence of rMSA on [CH4] and [SO3]. For ≫1, rMSA showed 푘 푘

푘 a shallow dependence on [CH4] and a linear dependence on [SO3]. For ≪1, rMSA showed a 푘 linear dependence on [CH4] and a shallow dependence on [SO3]. Also,

∙ [CH] 푘[CH] ∙ = … 3.59 [CHSO] 푘[SO]

Therefore, from fitting the simulated rMSA values to experimental data, (Figure 3.6) we could ∙ [] conclude, under our reaction conditions, ∙ is on the order of ~10. []

Furthermore, the concentration-time traces from peroxydisulfate-initiated MSA formation were reproduced with the equation for [MSA] derived below. Considering initiator decomposition (equation 3.51) alone,

⋅ 푑[Int] 1 푑[Int ] = − = −푘 [Int ] … 3.60 푑푡 2 푑푡

[Int] = [Int]푒 … 3.61

[⋅] Integrating the equation for from equations 3.60 and 3.61,

115

⋅ [Int ] = 2[Int](1 − 푒 ) … 3.62

Again assuming reaction 3.52 is rapid compared to the radical propagation reactions, the total ∙ ∙ ⋅ radical concentration is [CH] + [CHSO]. It should equal [Int ] in the absence of radical propagation reactions, if the rate of termination is negligible. Thus,

∙ ∙ [CH] + [CHSO] = 2[Int](1 − 푒 ) … 3.63

∙ ∙ ∙ From [CH] ≫ [CHSO] and 풓 = 푘[SO][CH],

푑[MSA] 풓 = 2푘 [SO ][Int ] 1 − 푒 = … 3.64 푑푡

1 − 푒 [MSA] = 푘[SO][Int] 푡 − … 3.65 푘

Plotting equation with arbitrary values for the constant terms give Figure 3.22.

Figure 3.22. Simulation of the expression for [MSA] versus time from peroxydisulfate-initiated methane sulfonation.

3.4.7. DFT calculation

3.4.7.1. Methods Density functional theory (DFT) was performed using the Gaussian 09 software package40 for geometry optimization, frequency determinations, and single point energy calculations. Visualization of optimized structures were done with Chemcraft 1.8 software. Geometry optimizations were carried out using BLYP functional41,42 along with the Stuttgart ten-electron pseudopotential and associated valence basis set of Preuss and coworkers (SDD)43 for palladium in combination with the split-valence double-zeta 6-31+G(d) basis set for main group elements in the gas phase. Single point energy calculations were conducted in water (ε ~ 78.5) to mimic the very polar nature of fuming sulfuric acid using a continuum solvent model (SMD)44 in conjunction with the BLYP functional with a larger basis set on the main group

116 elements, 6-311++G(d,p), and SDD for palladium. To be consistent with experiments all calculations were done at 323 K. Minima are deﬁned by having zero imaginary vibrational frequencies, whereas transition states have one imaginary vibrational frequency, which is related to the motion of the active hydrogen between the carbon and oxygen. All relevant multiplicities were evaluated for minima and transition states. All free energies are reported in kcal/mol.

3.4.7.2. Free energies of other isomers and tautomers Table 3.7. Free energy for each configurational isomer relative to the isomer with minimum free energy for the optimized species in the reaction pathways A and B. The unbridged isomer of some species was not obtained since the Pd-Pd bond falls apart during optimization. Free energies are reported in kcal/mol.

Species R B-Int1 A-TS B-TS A-Int2 B-Int2 A-P B-P Isomers Trans 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Cis 9.7 0.7 2.9 5.0 2.6 0.9 11.0 12.0 Unbridged 14.9 5.6 - 7.8 - 12.4 10.5 - Paddlewheel 20.8 - 6.0 - 8.3 - 12.3 -

Table 3.8. Absolute values of the free energy difference between two possible spin states for each species in the reaction pathways A and B. The unbridged isomer of some species was not obtained since the Pd-Pd bond falls apart during optimization. Also, the triplet state of cis-B-TS was not obtained. Free energies are reported in kcal/mol.

Species A- B- A- B- RCSS

Table 3.9. Free energy for each protonation tautomer relative to the tautomer with minimum free energy 1 2 2 III ([Pd2(κ -HSO4)2(κ -HSO4)2(μ -HSO4)2]) for the trans-Pd 2 complex. Relative energies are reported in kcal/mol.

a Tautomers ∆Grel

1 2 2 [Pd2(κ -HSO4)2(κ -HSO4)2(μ -HSO4)2] 0.0

1 1 2 2 2 [Pd2(κ -HSO4)(κ -H2SO4)(κ -HSO4)(κ -SO4)(μ -HSO4)2] 9.2

1 1 2 2 2 [Pd2(κ -HSO4)(κ -H2SO4)(κ -HSO4)2(μ -HSO4)(μ -SO4)] 9.6

1 1 2 2 [Pd2(κ -SO4)(κ -H2SO4)(κ -HSO4)2(μ -HSO4)2] 9.9

117

1 2 2 2 [Pd2(κ -HSO4)2(κ -HSO4)2(μ -H2SO4)(μ -SO4)] 12.4

1 2 2 2 [Pd2(κ -SO4)(κ -HSO4)2(μ -HSO4)(μ -H2SO4)] 12.9

a 1 1 2 2 2 1 1 1 The tautomer [Pd2(κ -HSO4)(κ -SO4)(κ -H2SO4)(κ -HSO4)(μ -HSO4)2] converts to [Pd2(κ -HSO4)(κ -SO4)(κ - 2 2 H2SO4)(κ -HSO4)(μ -HSO4)2] upon optimization to have 7.7 kcal/mol. Since the ligation geometry is different (one less coordination number at one of the Pd’s), it is not shown in the table alongside other tautomers.

Table 3.10. Relative free energies for optimized isomers for the ligand dissociation pathways with reactant and products at different charge states. Energies are reported in kcal/mol.

Neutral Complex Initial & Final Initial Complex Heterolytic Product Homolytic Product Complex 6+ 6+ + 5+ Pd (HSO ) [Pd (HSO ) ] Pd (HSO ) Isomers 2 4 6 2 4 5 2 4 5 Trans 0.0 0.0 0.0 Paddlewheel 20.8 4.6 8.2 Cis 9.7 1.6 2.9 Unbridged 14.9 3.0 11.1

Protonated Complex Initial & Final Initial Complex Heterolytic Product Homolytic Product Complex 6+ + 6+ + 6+ + [Pd (HSO ) (H SO )] [Pd (HSO ) ] [Pd (HSO ) (H SO )] Isomers 2 4 5 2 4 2 4 5 2 4 4 2 4 Trans 0.0 0.0 0.0 Paddlewheel 22.5 4.6 27.4 Cis 11.0 1.6 17.6 Unbridged 13.8 3.0 20.9

Deprotonated Complex Initial & Final Initial Complex Heterolytic Product Homolytic Product Complex 6+ - 6+ 5+ - [Pd (HSO ) (SO )] [Pd (HSO ) (SO )] [Pd (HSO ) (SO )] Isomers 2 4 5 4 2 4 4 4 2 4 4 4 Trans 0.0 0.0 0.0 Paddlewheel 13.3 7.7 8.1 Cis 5.7 2.0 4.7 Unbridged 9.7 6.3 10.3

3.4.7.3. Variations from pathway B: different ligand conformation after H abstraction 118

For the intermediates in pathway B after H abstraction, three different situations were evaluated. The tested situations include:

B-1) no ligand rearrangement after TS and considering coordination number of active Pd to be 4 and 5 for B-Int2 and B-P complexes, respectively (Figure 3.23);

B-2) keeping coordination number of B-Int2 and B-P same as the pathway A by ligand rearrangement from κ1 to κ2 (Figure 3.11);

B-3) keeping coordination number of B-Int2 and B-P same as the pathway A by leaving H2SO4 to remain attached to the Pd after HAA (Figure 3.24).

Case B-2 is energetically the most favorable and was shown earlier in 3.2.2.4. For case B-1, the

III reaction free energies for C–H bond activation (i.e. free energy of B-Int2) and CH3Pd 2 formation (i.e. free energy of B-P) are 15.0 kcal/mol and 7.4 kcal/mol, respectively, which are ~8.5 kcal/mol and ~26.9 kcal/mol higher than the corresponding values for case A. The higher reaction free energies for case B-1 compared to case A are probably due the lower coordination number of the palladium atom for B-Int2 and B-P versus the related species on the reaction coordinate for case A (A-Int2 and A-P).

For case B-3, if H2SO4 remains attached to the Pd after abstraction of the hydrogen atom, the intermediate B-Int2 and product B-P will be 16.9 kcal/mol and 1.6 kcal/mol. Although in this case the active Pd for B-Int2 and B-P are 5- and 6-coordinate, respectively, since the H2SO4 ligand is less stable

– than HSO4 ligand, the energies are higher than reaction free energies of A-Int2 and A-P by ~10.0 kcal/mol and ~18.0 kcal/mol, respectively. The energy of the product state was lower by 4.3 kcal/mol when the methyl group was bound to the axial site of the PdIII dimer.

Figure 3.23. Optimized structures for case B-1; ∆GInt2 = 15.0 kcal/mol (left), ∆GP = 7.4 kcal/mol (right).

119

Figure 3.24. Optimized structures for case B-3; ∆GInt2 = 16.9 kcal/mol (left), methyl-equatorial: ∆GP = 1.6 kcal/mol (middle), methyl-axial: ∆GP = –2.7 kcal/mol (right).

3.5. References

(1) He, J.; Wasa, M.; Chan, K. S. L.; Shao, Q.; Yu, J. Q. Palladium-Catalyzed Transformations of Alkyl C-H Bonds. Chem. Rev. 2017, 117 (13), 8754–8786. (2) Goldberg, K. I.; Goldman, A. S. Large-Scale Selective Functionalization of Alkanes. Acc. Chem. Res. 2017, 50 (3), 620–626. (3) Muñiz, K. High-Oxidation-State Palladium Catalysis: New Reactivity for Organic Synthesis. Angew. Chemie - Int. Ed. 2009, 48 (50), 9412–9423. (4) Grice, K. A.; Scheuermann, M. L.; Goldberg, K. I.; Labinger, J. A.; Bercaw, J. E. Higher Oxidation State Organopalladium and Platinum Chemistry: Five-Coordinate Platinum(IV) Complexes; Canty, A. J., Ed.; Topics in Organometallic Chemistry; Springer Berlin Heidelberg: Berlin, Heidelberg, 2011; Vol. 35. (5) Hickman, A. J.; Sanford, M. S. High-Valent Organometallic Copper and Palladium in Catalysis. Nature 2012, 484 (7393), 177–185. (6) Powers, D. C.; Ritter, T. Bimetallic Redox Synergy in Oxidative Palladium Catalysis. Acc. Chem. Res. 2012, 45 (6), 840–850. (7) Vedernikov, A. N. CHAPTER 4. PdII/PdIV Redox Couple Mediated C–X Bond Formation. In C–H and C–X Bond Functionalization: Transition Metal Mediation; Ribas, X., Ed.; The Royal Society of Chemistry, 2013; pp 108–121. (8) Khusnutdinova, J. R.; Mirica, L. M. CHAPTER 5. Organometallic Pd III Complexes in C–C and C–Heteroatom Bond Formation Reactions. In C–H and C–X Bond Functionalization: Transition Metal Mediation; Ribas, X., Ed.; The Royal Society of Chemistry, 2013; pp 122–158. (9) Yin, G.; Mu, X.; Liu, G. Palladium(II)-Catalyzed Oxidative Difunctionalization of Alkenes: Bond Forming at a High-Valent Palladium Center. Acc. Chem. Res. 2016, 49 (11), 2413–2423. (10) Burt, L. K.; Ho, C. C. High Oxidation State Organopalladium(IV) Species: Applications to C–H Activation. Aust. J. Chem. 2018, 71 (10), 753. (11) Desai, L. V.; Hull, K. L.; Sanford, M. S. Palladium-Catalyzed Oxygenation of Unactivated sp3 C−H Bonds. J. Am. Chem. Soc. 2004, 126 (31), 9542–9543.

120

(12) McNally, A.; Haffemayer, B.; Collins, B. S. L.; Gaunt, M. J. Palladium-Catalysed C-H Activation of Aliphatic Amines to Give Strained Nitrogen Heterocycles. Nature 2014, 510 (7503), 129–133. (13) Hashiguchi, B. G.; Konnick, M. M.; Bischof, S. M.; Gustafson, S. J.; Devarajan, D.; Gunsalus, N.; Ess, D. H.; Periana, R. A. Main-Group Compounds Selectively Oxidize Mixtures of Methane, Ethane, and Propane to Alcohol Esters. Science (80-. ). 2014, 343 (6176), 1232–1237. (14) O’Reilly, M. E.; Kim, R. S.; Oh, S.; Surendranath, Y. Catalytic Methane Monofunctionalization by an Electrogenerated High-Valent Pd Intermediate. ACS Cent. Sci. 2017, 3 (11), 1174–1179. (15) Topczewski, J. J.; Sanford, M. S. Carbon–Hydrogen (C–H) Bond Activation at Pd IV: A Frontier in C–H Functionalization Catalysis. Chem. Sci. 2015, 6 (1), 70–76. (16) Chuang, G. J.; Wang, W.; Lee, E.; Ritter, T. A Dinuclear Palladium Catalyst for α-Hydroxylation of Carbonyls with O2. J. Am. Chem. Soc. 2011, 133 (6), 1760–1762. (17) Kärkäs, M. D. Electrochemical Strategies for C–H Functionalization and C–N Bond Formation. Chem. Soc. Rev. 2018, 47 (15), 5786–5865. (18) Ma, C.; Fang, P.; Mei, T. S. Recent Advances in C-H Functionalization Using Electrochemical Transition Metal Catalysis. ACS Catal. 2018, 8 (8), 7179–7189. (19) Liler, M. Reaction Mechanisms in Sulphuric Acid; Academic Press: London, 1971. (20) Zimmermann, T.; Bilke, M.; Soorholtz, M.; Schüth, F. Influence of Catalyst Concentration on Activity and Selectivity in Selective Methane Oxidation with Platinum Compounds in Sulfuric Acid and Oleum. ACS Catal. 2018, 8 (10), 9262–9268. (21) Lutsyk, A. I.; Rudakov, E. S.; Gundilovich, G. G.; Mochalin, V. N. Solubility of Methane, Propane, and n-Hexane in the Water-Sulfuric Acid System at 283-343 K. Russ. J. Phys. Chem. A 1999. (22) Basickes, N.; Hogan, T. E.; Sen, A. Radical-Initiated Functionalization of Methane and Ethane in Fuming Sulfuric Acid. J. Am. Chem. Soc. 1996, 118 (51), 13111–13112. (23) Lin, M.; Sen, A. Oxidation and Oxidative Carbonylation of Methane and Ethane by Hexaoxo-µ- Peroxodisulfate(2–) Ion in Aqueous Medium. A Model for Alkane Oxidation through the Hydrogen- Atom Abstraction Pathway. J. Chem. Soc., Chem. Commun. 1992, No. 12, 892–893. + (24) Roytman, V. A.; Singleton, D. A. Comment on “Activation of Methane to CH3 : A Selective Industrial Route to Methanesulfonic Acid.” Science (80-. ). 2019, 364 (6440), eaax7083. (25) Eitel, S. H.; Bauer, M.; Schweinfurth, D.; Deibel, N.; Sarkar, B.; Kelm, H.; Krüger, H. J.; Frey, W.; Peters, R. Paramagnetic Palladacycles with PdIII Centers Are Highly Active Catalysts for Asymmetric Aza-Claisen Rearrangements. J. Am. Chem. Soc. 2012, 134 (10), 4683–4693. (26) Powers, D. C.; Lee, E.; Ariafard, A.; Sanford, M. S.; Yates, B. F.; Canty, A. J.; Ritter, T. Connecting Binuclear Pd(III) and Mononuclear Pd(IV) Chemistry by Pd-Pd Bond Cleavage. J. Am. Chem. Soc. 2012, 134 (29), 12002–12009. (27) Albéniz, A. C.; Espinet, P. Palladium: Inorganic & Coordination Chemistry. In Encyclopedia of Inorganic Chemistry; John Wiley & Sons, Ltd: Chichester, UK, 2006. (28) Al-Jibori, S. A.; Al-Saraj, E. G. H.; Hollingsworth, N.; Hogarth, G. Palladium(II) Complexes with 2-Acetylamino-5-Mercapto-1,3,4-Thiadiazolate (Amta) Ligands: Molecular Structures of the All 1 2 Trans Dipalladium “Paddlewheel” Pd2(μ-Amta)4 and Pd(κ -Amta)2(κ -Dppe). Polyhedron 2012, 44 (1), 210–214. (29) Gao, E.-J.; Wang, K.-H.; Gu, X.-F.; Yu, Y.; Sun, Y.-G.; Zhang, W.-Z.; Yin, H.-X.; Wu, Q.; Zhu, M.- C.; Yan, X.-M. A Novel Binuclear Palladium Complex with Benzothiazole-2-Thiolate: Synthesis, Crystal Structure and Interaction with DNA. J. Inorg. Biochem. 2007, 101 (10), 1404–1409. 121

(30) Ariafard, A.; Hyland, C. J. T.; Canty, A. J.; Sharma, M.; Brookes, N. J.; Yates, B. F. Ligand Effects in Bimetallic High Oxidation State Palladium Systems. Inorg. Chem. 2010, 49 (23), 11249–11253. (31) Gunsalus, N. J.; Koppaka, A.; Park, S. H.; Bischof, S. M.; Hashiguchi, B. G.; Periana, R. A. Homogeneous Functionalization of Methane. Chem. Rev. 2017, 117 (13), 8521–8573. (32) Paton, R. S. Kinisot: v 1.0.0 public API for Kinisot.py (Version v1.0) http://doi.org/10.5281/zenodo.60082. (33) Bigeleisen, J.; Mayer, M. G. Calculation of Equilibrium Constants for Isotopic Exchange Reactions. J. Chem. Phys. 1947, 15 (5), 261–267. (34) Bím, D.; Maldonado-Domínguez, M.; Rulísek, L.; Srnec, M. Beyond the Classical Thermodynamic Contributions to Hydrogen Atom Abstraction Reactivity. Proc. Natl. Acad. Sci. U. S. A. 2018, 115 (44), E10287–E10294. (35) Mukhopadhyay, S.; Bell, A. T. Catalyzed Sulfonation of Methane to Methanesulfonic Acid. J. Mol. Catal. A Chem. 2004, 211 (1–2), 59–65. (36) Fuller, J. T.; Butler, S.; Devarajan, D.; Jacobs, A.; Hashiguchi, B. G.; Konnick, M. M.; Goddard, W. A.; Gonzales, J.; Periana, R. A.; Ess, D. H. Catalytic Mechanism and Efficiency of Methane Oxidation by Hg(II) in Sulfuric Acid and Comparison to Radical Initiated Conditions. ACS Catal. 2016, 6 (7), 4312–4322. + (37) Díaz-Urrutia, C.; Ott, T. Activation of Methane to CH3 : A Selective Industrial Route to Methanesulfonic Acid. Science (80-. ). 2019, 363 (6433), 1326–1329. (38) Kim, R. S.; Surendranath, Y. Electrochemical Reoxidation Enables Continuous Methane-to- Methanol Catalysis with Aqueous Pt Salts. ACS Cent. Sci. 2019, 5 (7), 1179–1186. (39) Bacsik, Z.; Lopes, J. N. C.; Gomes, M. F. C.; Jancsó, G.; Mink, J.; Pádua, A. A. H. Solubility Isotope Effects in Aqueous Solutions of Methane. J. Chem. Phys. 2002, 116 (24), 10816–10824. (40) Frisch, M. J.; Trucks, G. W. .; Schlegel, H. B. .; Scuseria, G. E. .; Robb, M. A. .; Cheeseman, J. R.; Scalmani, G. .; Barone, V. .; Mennucci, B. .; Petersson, G. A.; Nakatsuji, H. .; Caricato, M. .; et al. Gaussian 09, Revision E. 01; Gaussian; 2009. (41) Becke, A. D. Density-Functional Exchange-Energy Approximation with Correct Asymptotic Behavior. Phys. Rev. A 1988, 38 (6), 3098–3100. (42) Lee, C.; Yang, W.; Parr, R. G. Development of the Colle-Salvetti Correlation-Energy Formula into a Functional of the Electron Density. Phys. Rev. B 1988, 37 (2), 785–789. (43) Bergner, A.; Dolg, M.; Küchle, W.; Stoll, H.; Preuß, H. Ab Initio Energy-Adjusted Pseudopotentials for Elements of Groups 13–17. Mol. Phys. 1993, 80 (6), 1431–1441. (44) Marenich, A. V.; Cramer, C. J.; Truhlar, D. G. Universal Solvation Model Based on Solute Electron Density and on a Continuum Model of the Solvent Defined by the Bulk Dielectric Constant and Atomic Surface Tensions. J. Phys. Chem. B 2009, 113 (18), 6378–6396.

122

4. Electrochemical Oxidation of Platinum Salts for Continuous Methane Hydroxylation Catalysis in Dilute Aqueous Acid

Adapted and reprinted with permission from Kim, R. S.; Surendranath, Y. Electrochemical Reoxidation Enables Continuous Methane-to-Methanol Catalysis with Aqueous Pt Salts. ACS Cent. Sci. 2019, 5 (7), 1179–1186.

4.1. Introduction

Among many homogeneous and heterogeneous systems that have been investigated for methane-

1,2 II II (2–x) to-methanol conversion, simple Pt chloride salts in water, Pt Clx(H2O)(4–x) (denoted collectively as PtII), offer unique advantages.3 The catalytic cycle (Scheme 4.1) is initiated by PtII ions, which carry out

II reversible C-H activation of CH4 to yield a Pt -CH3 intermediate. This intermediate is then oxidized by

IV (4–x) IV IV Pt Clx(H2O)(6–x) (denoted collectively as Pt ) to generate a Pt -CH3 species that undergoes rapid reductive elimination to produce CH3OH or CH3Cl, which can be hydrolyzed to CH3OH. This system has the following advantages: first, the organometallic activation of methane offers superior selectivity for mono-oxidation compared to catalysts that operate via radical intermediates;2,4–6 second, while most homogeneous catalysts that do organometallic activation require impractical2,7 concentrated acid media for boosting the catalytic rate and selectivity,8,9 PtII operates in dilute aqueous acids. Along with the relatively low reaction temperature (> 100 °C), these advantages position PtII chloride salts, often referred to as “Shilov’s catalyst,” as privileged agents for methane-to-methanol conversion under mild conditions.

A critical drawback of Shilov’s catalyst, as originally reported, is its requirement for a stoichiometric PtIV oxidant, which is economically impracticable.3 The key to developing an alternative oxidation strategy for this catalytic system is to achieve precise control over the driving force (thermodynamics) and/or rate (kinetics) of the oxidation reaction. In view of the catalytic cycle, there are two distinct approaches to the problem. First, PtIV may be directly replaced by an alternative oxidant that

II can oxidize the Pt -CH3 intermediate (Scheme 4.1, Strategy A). Success of this strategy requires an oxidant

II II that (i) rapidly oxidizes the fleeting Pt -CH3 intermediate before it can undergo protonation back to Pt +

II IV CH4 and (ii) possesses a low enough redox potential to avoid oxidizing the Pt catalyst to Pt , which is inactive towards CH4. The conflicting requirement for fast rates and low driving force places an inherent 123 constraint on the oxidants that are viable. Second, one may employ PtIV itself, which is an efficient oxidant

II for Pt -CH3, as a redox mediator for the overall reaction (Scheme 4.1, Strategy B). Success of this strategy hinges on carefully matching the rate of PtIV regeneration by PtII oxidation to the rate of its consumption by methane functionalization. Rapid PtII oxidation will progressively deplete the pool of PtII, retarding catalysis, whereas slow oxidation will deplete PtIV and induce irreversible decomposition of the PtII to metallic Pt0 via, inter alia, disproportionation of PtII.3,10 Thus, a viable alternate oxidant must achieve good control over the oxidation driving force and/or rate.

Scheme 4.1. The catalytic cycle for the functionalization of methane by aqueous Pt salts (Shilov’s catalyst) and different strategies to overcome the stoichiometric use of PtIV.

The inherent difficulty of fine-tuning oxidation using chemical reagents, has, presumably, contributed to the limited success in replacing stoichiometric PtIV. Notably, oxidants such as heteropoly

II acids, CuCl2, FeCl3, and Br2 were identified as kinetically competent toward the oxidation of Pt -CH3 (Scheme 4.1, Strategy A).11 These oxidants have achieved PtII-mediated oxidation of methane or other

12–15 aliphatic substrates, and some of them, being air-regenerable, have been employed in concert with O2 to effect overall aerobic methane functionalization. However, none of these studies established long-term

II stability. For example, the combination of CuCl2 and O2 ultimately resulted in complete oxidation of Pt to PtIV,3 highlighting the difficulty of controlling the oxidation driving force. Studies aimed at mediating

II/IV 16 10 turnover via the Pt redox couple (Scheme 4.1, Strategy B) showed that Cl2 and H2O2 are viable oxidants. However, the PtII oxidation rate was not actively modulated and, thus, continuous operation was not demonstrated. Furthermore, neither of these oxidants is air-regenerable or affordable for methanol production. In sum, there exists yet no suitable alternative to stoichiometric PtIV for sustained aqueous PtII- catalyzed methane-to-methanol conversion.

124

We show that electrochemistry affords a unique solution to this problem. Unlike all stoichiometric chemical oxidations, electrochemical oxidation allows for unparalleled control over the rate and driving force for electron transfer. Furthermore, the rate and driving force can be toggled instantaneously for real-

II time, dynamic modulation. While direct electro-oxidation of the fleeting Pt -CH3 intermediate is unfeasible due to the small fraction of reaction solution volume in contact with the electrode surface, electrochemistry is well suited to regenerate PtIV via re-oxidation of PtII (Scheme 4.1, Strategy C). As noted above, the success of this approach relies on maintaining a constant PtII:PtIV ratio; electrochemistry allows for simultaneous measurement and fine tuning of this ratio in real-time. In addition, coupling the methane oxidation half-reaction with an oxygen reducing cathode would render the overall process aerobic.

Despite its attractiveness, there exists a paucity of examples of this approach. One report applied

II electrochemical oxidation in the presence of Pt , heteropoly acids, and O2 to achieve 1.4 turnovers for methanol production, but no information about the mechanism or stability of the system was provided.17 Earlier, a similar scheme was employed to oxidize a non-gaseous test substrate, p-toluenesulfonic acid; while 11 turnovers of the PtII catalyst were attained, deposition of Pt0 was observed with increasing reaction times.18 A particular impediment to the electrochemical turnover of the aqueous PtII catalyst is the general sluggishness of two-electron PtII/IV oxidation at an electrode.19,20 Herein, we combine Pt electrodes that catalyze facile oxidation of PtII to PtIV 21 with in situ modulation of electric current to achieve continuous, steady-state methane oxidation over the course of 30 hours. We observe the generation of methanol and methyl chloride as the principal products with >80% combined selectivity, demonstrating continuous PtII- catalyzed electrochemical methane oxidation.

4.2. Results and Discussions

4.2.1. Identification of a suitable electrode for PtII-catalyzed Electrochemical Methane Oxidation Reaction (EMOR)

4.2.1.1. Electrochemical oxidation of PtII at room temperature The electrochemical mediation scheme put forward above (Scheme 4.1, Strategy C) requires an electrode capable of oxidizing PtII to PtIV. In view of the high PtII/IV oxidation potential (E0 = 0.68 V vs SHE

II 2– IV 2– 22 23 for Pt Cl4 /Pt Cl6 ) and the acidic environment required for stability of the Pt ions, we focused our investigations on carbon, fluorine-doped tin oxide (FTO), and Pt electrodes as possible candidates. Whereas carbon and FTO electrodes displayed progressive deactivation and/or sluggish PtII oxidation kinetics (see

II 4.4.2.3), Pt electrodes showed facile oxidation of Pt at modest potentials. In 0.5 M H2SO4, the Pt electrode

125 displays the typical voltammetric features associated with hydrogen underpotential deposition (H UPD) and oxide formation at low and high potentials, respectively (Figure 4.1a, black; also Figure 4.5).24,25

II 2− Upon addition of 1 mM Pt Cl4 , a reversible wave appears at Ep,a = 1.1 V and Ep,c = 0.8 V (Figure 4.1a, blue). The appearance of this wave is accompanied by a suppression in the background Pt oxide wave,

− II 2− which we ascribe to inhibition by surface-adsorbed Cl that has dissociated from the Pt Cl4 ions (Figure 4.18).26 As sustained methanol production requires Cl– ions (see below), we also examined the voltammetric response of PtII in the presence of 10 mM Cl– (Figure 4.1a, red). Whereas the PtII oxidation wave is largely unaffected by the additional Cl–, the cathodic wave associated with PtIV back-reduction is significantly suppressed. These observations are in line with previous literature on PtII/IV oxidation at Pt electrodes that

126 invokes an inner-sphere electron transfer mechanism involving transfer of a surface-adsorbed Cl– to PtII during the oxidation reaction.21 Indeed, the reported Cl-adsorption isotherm at 2 mM Cl− stretches from 0 to 0.8 V vs SHE (Figure 4.19),27 showing near-saturation at the potential for PtIV reduction. These observations suggests that Cl– surface coverage at this potential may be incomplete at low [Cl–] but complete at 10 mM Cl−. Thus, higher surface coverage of Cl– induced by higher [Cl–] has a negligible impact on PtII oxidation, but the back-reduction of PtIV, which requires Cl– transfer back to the electrode surface, is inhibited (see 4.4.2.3). This inner-sphere mechanism explains why Pt electrodes display superior PtII electro-oxidation kinetics compared to carbon or FTO.

4.2.1.2. Electrochemical oxidation of PtII at 130 ̊C for methane oxidation catalysis Having identified a suitable electrode material, we then investigated PtII/IV electro-oxidation at the elevated temperatures required for methane activation by PtII. These experiments were conducted above the boiling point of water and were, therefore, carried out in a home-built high-pressure electrochemical cell (see 4.4.1.2 and below). As shown in Figure 4.1b, red, high PtII oxidation current flowed at 130 °C; the 5- fold enhancement in current and approximately 100 mV negative shift in Ep,a compared to room temperature reflect faster mass transport and electrode kinetics. The decrease in current at E > 1.1 V is attributed to the formation of surface oxides that inhibit the inner-sphere PtII oxidation. This inhibition is particularly pronounced at high [PtII] and high temperatures (see 4.4.2.3). We also examined the dependence of PtII oxidation current on electrochemical driving force (Figure 4.1c). Keeping the potential below Pt oxide formation, < 1.1 V, the steady-state current increased 10-fold per 104 mV of additional overpotential (η =

E – Eeq). This Tafel slope corresponds to a rate-limiting one-electron transfer with a transfer coefficient of 0.77, in agreement with the aforementioned mechanism.28 These results show that Pt electrodes are capable of facile oxidation of PtII at elevated temperatures.

Pt electrodes were also capable of sustained and efficient PtII/IV oxidation. We carried out bulk electrolyses of a stirred solution at 130 °C by applying a constant potential below 1.1 V. After chronoamperometry at 0.874, 0.924 and 0.974 V for 77, 40 and 17 min, respectively, half of the PtII ions in the initial solution were converted to PtIV ions as determined by UV-Vis analysis. At all three potentials examined, PtIV was generated with 100% Faradaic efficiency (Table 4.3).

4.2.1.3. Electrochemical oxidation of methanol Sustained methane oxidation catalysis will lead to a progressive rise in methanol concentration in the reactor over time. Thus, in addition to supporting facile PtII/IV oxidation, the electrode must be inert towards further oxidation of the CH3OH product. This is a particular concern for Pt, which is the standard

29 electrocatalyst for oxidation of CH3OH to CO2. Indeed, in 0.5 M H2SO4 at 130 ˚C, addition of 30 mM

CH3OH gives rise to the well-known anodic features associated with CH3OH electro-oxidation (Figure

127

30 – 4.1d, blue). Remarkably, upon addition of 10 mM of Cl , this CH3OH oxidation feature is almost completely suppressed (Figure 4.1d, red) over the entire potential range examined. This suppression is ascribed to surface adsorption of Cl– ions.31 Additional control experiment confirmed that the non-

32 electrochemical oxidation of CH3OH catalyzed on metallic Pt is also negligible under our conditions (see 4.4.2.4). These data indicate that, fortuitously, the presence of Cl– serves to simultaneously promote PtII/IV oxidation and suppress surface-catalyzed oxidation of the methanol product. Together, these studies establish that Pt electrodes are suitable for EMOR.

4.2.2. Sustained methane oxidation catalysis via dynamic electrochemical control of the PtII:PtIV ratio

4.2.2.1. Construction of the electrochemical reactor for EMOR The above studies provide the basis for carrying out continuous methane-to-methanol oxidation catalysis via electrochemical regeneration of PtIV (Scheme 4.1, Strategy C). The EMOR was carried out in a home-built high-pressure cell which consisted of a modified Parr reactor with electrical feedthroughs (Figure 4.2; see 4.4.1.2 for full details). The working compartment was charged with 3 mM PtII and 7 mM

IV Pt in 10 mM NaCl, 0.5 M H2SO4 (see 4.4.4 for details of electrolyte optimization). The counter

+ IV compartment, separated by a H -conducting membrane stack, contained 3 M vanadyl sulfate ((V O)(SO4)) as a sacrificial oxidant to be reduced at the cathode. In a practical device, oxygen could be supplied to the cathode, but given the low solubility of O2 and complications of co-pressurizing the cell with O2, we opted to use the vanadyl ion as a surrogate. The highly soluble and fairly inert vanadyl ions enabled examination of long-term electrolysis. This counter reaction prevented H2 evolution, which must be avoided in this

II 0 configuration due to the irreversible reduction of Pt to Pt by H2; however, in a well-engineered system with good gas stream separation, H2 may be deliberately generated as a useful byproduct. The solutions and the cell were purged to remove O2 prior to pressurization with 500 psi of methane. Following heating and temperature stabilization at 130 ˚C, electrolysis was initiated to continuously re-oxidize PtII during methane functionalization catalysis. The electrolysis was carried out with control of the current instead of the potential, which is the preferred method in industrial electrolysis.33

128

Figure 4.2. High-pressure, three-electrode, two-compartment electrochemical cell for EMOR. WE: Pt foil working electrode, RE: Ag/AgCl reference electrode, CE: Pt mesh counter electrode. 1: Glass cell, 2: working solution containing the Pt ions, 3: fritted tubes for housing the RE, 4: PTFE stir bar, 5: H+- conducting membrane separating the counter compartment, 6: PTFE body holding the membrane stack, 7: IV counter compartment solution containing (V O)(SO4) as the sacrificial electron acceptor.

4.2.2.2. Real-time control of the PtII:PtIV ratio during EMOR Careful choice of the applied current is critical for sustained catalysis. In order to maintain a constant PtII:PtIV ratio over the course of the reaction, the rate of PtII oxidation at the electrode must match the rate of methane oxidation catalysis in the solution. A simple mathematical derivation shows that, at a fixed rate of PtII/IV oxidation, any small difference between the two rates will cause the PtII:PtIV ratio to deviate from the initial value exponentially over time (see 4.4.6.2). Thus, the applied current must be constantly readjusted to match the rate of catalysis in order to maintain a steady ratio of PtII:PtIV. To achieve this, we employed the open-circuit potential (OCP) of the working compartment as an in situ probe of the instantaneous PtII:PtIV ratio in solution and adjusted the current (i) accordingly. In our reactor, the PtII and

IV II (2–x) IV (4–x) Pt ions exist in various ligated states (Pt Clx(H2O)(4–x) and Pt Clx(H2O)(6–x) ), each pair of which has different redox potentials. Assuming that [Cl−] is constant, the following modified form of the Nernst equation is derived:

푅푇 [Pt] 푅푇 1 퐸 = 퐸 + ln ; 퐸 = 퐸 + ln … 4.1 2퐹 [Pt] 2퐹 [Cl] where E0’’ and n represents the weighted average of the redox potentials and Cl− stoichiometries, respectively. Thus, using equation 3.17, we can estimate the instantaneous PtII:PtIV ratios potentiometrically. EC can be determined from the initial OCP reading and the known initial PtII:PtIV ratio. 129

Figure 4.3 shows the electrochemical data recorded during a typical EMOR trial with periodic OCP monitoring and adjustment of the current. To aid the interpretation, the PtII:PtIV ratio was converted to the percentage of PtII ions (PtII%), defined as [PtII]/([PtII]+[PtIV]). In a representative reaction, after 1.0 mA of current was passed for 1 h, the PtII% decreased from 30% to 25%. This led us to adjust the current to 0.9 mA, and after another 1 h, the PtII% rose to 29%. This process of quantifying the PtII% in the solution and adjusting the current to maintain a roughly constant PtII% was repeated periodically until the reaction was terminated. Incidentally, while our reactor was too congested to conveniently add a fourth electrode, incorporation of a separate sensing electrode could allow, in principle, for real-time feedback modulation of i.

The potential required for electrolysis (ECP, CP=chronopotentiometry) equals the equilibrium electrode potential (OCP) plus the magnitude of overpotential (η) applied. By definition, η is the difference between the applied potential (ECP) and EOCP, as marked with green arrows in Figure 4.3. Over multiple trials, we consistently observed a steady decrease in η during the initial 2–3 hours of each electrolysis, which we attribute to a slow initial electrode activation process. After stabilization of the electrode activity, η was ca. 20–40 mV, at an average current of around 0.9 mA. Normalizing by the electrode surface area,

130 we estimate an average current density of 0.09 mA/cm2. This is in line with the previously obtained Tafel plot (Figure 4.1c) after considering the difference in [PtII] (5 mM in the Tafel plot, approx. 3 mM in the EMOR trials). While the required η in our system is an extrinsic parameter that depends on the reactor configuration (see 4.4.6.3), we emphasize that the fast PtII oxidation kinetics on Pt electrode enables such a low η.

4.2.2.3. Effectiveness of the electrochemical control of the PtII:PtIV ratio Independent quantification of the PtII:PtIV ratio at the end of the EMOR confirmed the power of in situ current modulation. At the end of each reaction, [PtII] and [PtIV] in the working compartment was measured by UV-Vis spectroscopy. Despite a wide variation in reaction time (5–29 h) and consequently turnover number (see below), UV–Vis analysis confirmed that the final PtII% (19–23%) values were all similar (Table 4.1). These values are somewhat lower than the initial PtII% (30%), reflecting our preference to err on the side of lower PtII% to prevent irreversible Pt0 deposition (see below). Interestingly, despite the

II II agreement in final Pt % values, ΔOCP (= OCPlast – OCPfirst), which should reflect the final Pt % according to equation 4.1, was more negative for longer reactions by up to 14 mV. We postulate that this may be due

– to decreasing [Cl ] in the reaction solution as a result of CH3Cl formation. Despite this additional long-term effect, changes in the OCP between constant-current intervals provided a faithful indication of whether the PtII% was increasing or decreasing, allowing for appropriate adjustment of i. Together, these results demonstrate that the PtII% can indeed be maintained over long time durations of catalysis through dynamically-controlled electrochemical oxidation.

II IV Table 4.1. Results of EMOR trials at T=130 ℃ and PCH4= 675 psi. Initial [Pt ] and [Pt ] in the working solution were 3 mM and 7 mM, respectively, and the solution volume was 23 mL. The electrochemically active surface area of the Pt working electrode was 10.3 cm2.

e e approx. TOF a b c Product (μmol (rel. fraction)) approx. TON –1 Time iave ΔOCP Final (h ) (h) (mA) (mV) [PtII]% d CH3OH CH3Cl CH2(OH)2 HCOOH CO2 CH3X Total CH3X Total

60.5 20.1 2.2 0.1 1.1 4.9 1.19 7.9 22% 1.4 1.6 0.29 0.32 (72%) (24%) (3%) (0%) (1%)

93.7 27.9 5.1 1.2 4.4 10.5 0.88 5.7 19% 2.3 2.9 0.21 0.27 (71%) (21%) (4%) (1%) (3%)

205.4 44.8 21.9 2.9 12.2 18.4 1.00 -2.8 22% 4.5 6.3 0.24 0.34 (72%) (16%) (8%) (1%) (4%)

268.0 52.0 36.4 7.2 24.1 29.3 0.91 -6.0 23% 5.8 9.3 0.20 0.32 (69%) (13%) (9%) (2%) (6%)

131

aThe reaction time is the length of time the reactor was at the designated temperature, which spanned from ~80 minutes after the start of heating to the time at which the reactor was removed from the oil bath. b iave was calculated by dividing the total charge passed by the reaction time. c ΔOCP is the difference between the first and last OCP readings (= OCPlast – OCPfirst). dThe hydrated form of formaldehyde, which is the predominant form of formaldehyde in the acidic pH employed. eThe TONs were determined from dividing the moles of product by the average of the initial and final moles of PtII for each reaction. The TOFs were obtained by dividing the TON by the time duration of each reaction. The total II number of turnovers were calculated by assuming that all oxidation reactions were catalyzed by Pt : (μmolCH3OH + μmolCH3Cl + 2*μmolCH2(OH)2 + 3*μmolHCOOH + 4*μmolCO2) was divided by the average μmolPtII to determine total TON. For CH3X-specific turnovers, only (μmolCH3OH + μmolCH3Cl) was divided by μmolPtII.

Careful control of the PtII:PtIV ratio during the reaction is essential for another reason: PtIV ions suppress the irreversible decomposition of PtII to Pt0.3,10 Indeed, at the end of all of our EMOR trials, the bulk reaction solutions contained no visible Pt0 precipitates. Only a few adventitious Pt0 deposits were observed on the reactor surfaces and crevices where mass transport was restricted and replenishment of PtIV was impeded (see 4.4.5.2). One of the Pt0 deposition mechanisms is disproportionation of PtII.10 While the solution composition is thermodynamically inclined to deposit Pt0 (Figure 4.23),34 our results demonstrate that under sufficiently high [PtIV], nucleation of Pt0 may be inhibited (see 4.4.5.1 and Table 4.4). Although an extensive discussion of Pt0 deposition mechanisms is beyond the scope of the current work, these considerations highlight the importance of maintaining a stable PtII:PtIV ratio.

4.2.3. Analysis of methane oxidation products from the EMOR reactor.

Operation of the EMOR reactor using the feedback modulation procedure described above allowed for continuous functionalization of methane (Table 4.1 and Figure 4.4). In all cases, we observe

CH3OH as the majority product in 69–72% yield (Table 4.1). We also observe appreciable quantities of

CH3Cl with a yield that decreases from 24 to 13% as the reaction time increases. Small amounts of overoxidized products (CH2(OH)2, HCOOH and CO2) were observed in less than 20% combined yield.

II Taking these overoxidized products to represent Pt -catalyzed oxidation of CH3OH by 1, 2 and 3- equivalents of PtIV, respectively, the overall Faradaic efficiencies were in excess of 90% in all cases (Table 4.2). The per-PtII turnover numbers could not be rigorously determined due to minor fluctuations in [PtII] over the course of the reaction (see above), but approximate values were calculated from the known initial

II and final Pt amounts. For the longest trial, TON values of 6 and 9 for monofunctionalized products (CH3X

= CH3OH and CH3Cl) and total oxidation events were obtained, respectively (Table 4.1). The TOF for

–1 CH3X, estimated to be 0.2–0.3 h , showed a decreasing trend with increasing reaction time due to the overoxidation of CH3OH. In contrast, the TOF for total oxidation events was relatively constant at ca. 0.3 h–1 for different reaction times. Together, these observations demonstrate that electrochemical re-oxidation effectively sustains PtII-based methane functionalization catalysis.

132

Figure 4.4. (a) Amounts of methane oxidation products from EMOR versus reaction time. Each point represents a different trial in Table 4.1, and the product concentrations were normalized by iave of each trial (see 4.4.6.4 for explanation). The lines represent fitting with the (b) set of suggested reactions.

Combining the four trials in Table 4.1, Figure 4a visualizes the temporal progression of EMOR.

We fit these data to the set of reactions suggested earlier: oxidation of CH4 to CH3OH and CH3Cl, hydrolysis of CH3Cl to CH3OH, and subsequent overoxidation of CH3OH to CH2(OH)2, HCOOH, and CO2 (Figure

4.4b). While the fitted apparent rate constants (Table 4.7) for CH2(OH)2 and HCOOH oxidation show deviation from values separately determined outside the reactor (Table 4.9), the fitted values for CH4 and

CH3OH oxidation are in good agreement with those independent measurements (Table 4.8). Thus, this simple model provides a reasonable description of the methane oxidation processes taking place during EMOR.

All of the EMOR experiments shown in Table 1 were performed with identical reaction solution compositions with a low (3 mM) catalyst concentration. When the concentrations of PtII, PtIV and Cl– were increased, CH3OH and CH3Cl output increased while the fraction of CO2 decreased (see 4.4.6.5 and Table 4.6). These results suggest that there is ample room for optimization of the solution composition to maximize yield and selectivity.

4.2.4. Outlook for practical methane oxidation

Our studies establish that electrochemical oxidation endows Shilov’s catalyst with a sustainable mechanism for turnover and an inherent stability against deactivation through either complete oxidation of PtII to PtIV or Pt0 deposition. However, we acknowledge that the PtII catalyst displays a relatively low reaction rate and moderate selectivity. Our work does not directly address these inherent limitations of the catalyst;3 furthermore, our proof-of-concept reactor was not designed to demonstrate optimal TON, TOF, or selectivity for methanol. However, the EMOR approach developed here opens the door towards a broader exploration of reaction conditions and reactor configurations that may overcome these rate and selectivity

133 limitations. For example, higher temperatures and catalyst concentrations could be employed to enhance the reaction rate, but these conditions would lower the kinetic barrier to deactivation by Pt0 deposition. EMOR can be used to maintain an optimal PtII/IV ratio that is matched to these conditions (e.g. Figure 4.23,

II (2−x) red square) and thereby sustain catalysis at higher volumetric productivity. Additionally, since the Pt Clx catalyst displays modest selectivity for methane vs methanol oxidation (~1:1) (see 4.4.7.2),5 strategies for reactive distillation of the product would be needed to minimize over-oxidation. As opposed to a volatile chemical oxidant that may be released at a similar rate as the methanol product, electrochemical oxidation could allow for independent control of oxidant delivery and product release. While many challenges remain, EMOR offers new opportunities for developing practical Shilov-type systems for methane-to-methanol conversion.

4.3. Conclusions

We have established an electrochemical approach for continuous methane-to-methanol conversion using aqueous PtII catalysts. Cl-adsorbed Pt surfaces were shown to be competent for the inner-sphere two- electron oxidation of PtII to PtIV while inert toward parasitic oxidation of the methanol product. In situ potential measurements and current modulation allowed us to carry out continuous steady-state catalysis by maintaining the PtII:PtIV ratio. While our test reactors were run up to 30 h, further reactor engineering to automatically modulate the current in real-time, enhance solution mixing, and rigorously separate the anode and cathode compartments should allow for extended operation. Moreover, the integration of an oxygen- consuming counter electrode will enable net aerobic methane-to-methanol conversion. While many additional challenges remain in order to realize viable PtII-catalyzed methane conversion,3 we envision that the electrochemical approach developed here will stimulate continued progress toward practical technologies for aerobic methane valorization.

4.4. Methods and Additional Information

4.4.1. Materials and methods

4.4.1.1. Chemicals and Materials Methane (UHP GR 4.0) was purchased from Airgas. All solutions were prepared with ultrapure water (Milli-Q Type 1; resistivity = 18 MΩ cm). Potassium tetrachloroplatinate (K2PtCl4, 99.9% metals basis) was purchased from Strem Chemicals. Sodium hexachloroplatinate hexahydrate (Na2PtCl6·6H2O,

134

31.3% Pt), platinum foil (0.025 mm thick)/mesh/wire (99.9% metals basis), and silver wire (1.0 mm dia., 99.999%) were purchased from Alfa Aesar. Glassy carbon disk (3 mm dia.) and platinum disk (2 mm dia.) electrodes and Hg/Hg2SO4 (in sat. K2SO4; 0.64 V vs SHE) reference electrodes were purchased from CH Instruments. Fluorine-doped tin oxide (FTO) (TEC15, ~7 Ω/sq) was purchased from Hartford Glass Co. Inc. (Hartford City, IN). Ceramic fritted glass tubes for the home-made double-junction Ag/AgCl reference electrode were purchased from Pine Instruments. Nafion 117 (178 μm thick) and Nafion HP (20 μm thick; PTFE-reinforced) were purchased from Ion Power Inc., and polybenzimidazole membranes (55 μm thick) were purchased from PBI Performance Products Inc. The Nafion and PBI membranes were activated by established pretreatment procedures. Briefly, Nafion membranes were heated at 70–80 ˚C, in 3% H2O2 and subsequently in 0.5 M H2SO4, for 30–60 min each, then stored in 0.5 M H2SO4 or water. PBI was soaked in concentrated phosphoric acid (85%) for ~20 min at room temperature, then washed with 0.5 M H2SO4. We found that PBI membranes tend to crumple and shrivel during the activation procedure; holding the membrane between two glass surfaces mitigated this problem. The activated PBI membranes were stored in 0.5 M H2SO4.

4.4.1.2. Electrochemical methods General Electrochemical experiments were performed using a Biologic VMP3 or CHI760E potentiostat. Glassy carbon and platinum disk electrodes were polished successively with 1 μm, 0.3 μm, and 0.05 μm alumina slurries on a soft polishing cloth, with >5 min of sonication in Milli-Q water in between. At room temperature, the counter compartment was separated from the working solution by a Nafion 117 (~180 μm thick) membrane and a Pt mesh was used as the counter electrode. Room temperature cyclic voltammetry and bulk electrolysis were performed under ambient conditions.

All potential values in the manuscript are referenced to the Standard Hydrogen Electrode (SHE). Current values were reported as current densities in most cases, normalized by the surface area of the electrode. For glassy carbon and FTO electrodes, the geometric surface areas were used. For Pt electrodes, the electrochemically active surface area was determined by integrating the hydrogen underpotential deposition (H UPD) region and dividing by the known capacitance for surface-adsorbed H (210 μC cm−2)24 (Figure 4.5).

135

High-temperature electrochemistry The reactor and its operation. A modified Parr reactor as depicted Figure 2 was used for high temperature (130 ˚C) experiments. The Parr reactor was a Series 4760 General Purpose Pressure Vessel (300 mL size, constructed from T316 stainless steel). The reactor head was adapted with a high-pressure fitting holding PTFE-sheathed electrical wires (Conax Technologies, part no. TG-24T(KP)-A4-T), a stainless steel 1/8” needle valve for gas inlet/outlet, and a pressure relief valve for safety. The connection between the electrical wires and electrodes were made by twisting the wires together and the connection was further secured by wrapping thin twisted Ti or Pt wires around the wire junction. After ensuring that the wire junction resistance was ~2–3 Ω, the connection was tightly wrapped with PTFE tape, which may be further heat-sealed with a flame-heated glass pipette.35

Figure 4.6. Assessments performed to ensure proper operation of the high-temperature electrochemical cell. (a) Current-time trace from polarization of the electrode at 1.06 V vs SHE at room temperature. Stirring was turned on ~5 sec after the start of electrolysis and the stir rate was increased slowly to 200 rpm. (b) OCP registered at the electrode during heating. The initial rapid decrease in the OCP is because of relaxation from the previous polarization during the stir bar fidelity check shown in (a).

136

While omitted in the schematic diagram of the reactor (Figure 4.2), the glass cell containing the working solution was actually placed in a larger glass liner that fitted snugly inside the reactor. This was done in order to reduce the working solution volume, and, thereby, facilitate stirring and reduce the quantity of Pt salts employed. A custom-made PTFE piece was placed between the glass liner and the glass cell to fill the void space between the two and hold the working electrode, reference electrode and counter compartments in their respective positions. At the end of a high-temperature experiment, the solution volume decreased from 23 mL to 18–20 mL from evaporation and condensation onto the inner surfaces of the reactor. The condensed droplets were collected separately in our analysis (see below).

To set up the reactor, the working solution (23 mL) and counter solution (3 mL) were first degassed with Ar or N2. After the various parts of the reactor were assembled and the reactor was sealed, the headspace was purged with Ar or N2 by three vacuum-refill cycles. For EMOR, the headspace was filled at room temperature with 500 psi of methane with at least three pressurization-vent cycles.

The solution was constantly stirred at 200 rpm with a spinfin stir bar, which has the advantage of having a relatively stationary footprint. Since the reactor walls prevented visual confirmation of effective stirring, we used the following procedure to ensure convective transport in all reactor trials: after reactor assembly and setup, the electrode was polarized at 1.06 V vs SHE and a chronoamperometry trace was recorded. Then, stirring was turned on and the stir rate was gradually increased to 200 rpm. If the current decayed to a non-zero steady state value indicative of convective mass transport (e.g., Figure 4.6a), we stopped the application of potential and started heating the reactor. If the current showed a smooth Cottrell- like decay towards zero current, diagnostic of a quiescent solution, the reactor was disassembled to reposition the stir bar, and was then reassembled.

After confirming stirring, the reactor was placed in an oil bath and heated to 130 ˚C. The actual

CH4 pressure during reactor operation (130 ˚C) is estimated at 675 psi according to the ideal gas law. During heating, the open-circuit potential (OCP) of the electrode was monitored and showed a steady and reproducible increase (Figure 4.6b). We considered the temperature inside the reactor to have stabilized when the potential reached a plateau (typically ~1 h 20 min after the initiation of heating), at which point we started applying electrochemical bias to reoxidize PtII.

The working electrode (WE). A platinum wire (for data in Figure 4.1) or a platinum foil (for measurements of PtII electro-oxidation faradaic efficiencies and EMOR) was used as the working electrode. They were cleaned before and after each experiment by cycling the potential between 1.47 and –0.06 V vs

SHE in 0.5 M H2SO4 until a reproducible cyclic voltammogram was obtained with characteristic hydrogen underpotential deposition and surface oxide formation features. Generally, little change was observed before and after each experiment (Figure 4.7).

137

Figure 4.7. CVs of the Pt foil working electrode in 0.5 M H2SO4 (black) before and (red) after reactor operation for electrochemical methane oxidation reaction (EMOR). Scan rates = 100 mV s–1.

The reference electrode (RE). For the reference electrode, a home-made double-junction Ag/AgCl reference electrode was used. A clean silver wire (1.0 mm dia., 99.999%) was polished with fine-grit sandpaper and sonicated in 3% HNO3 and Milli-Q water for 10 min each. Subsequently, the silver wire was

−2 galvanostatically oxidized at 10 μA cm for >24 h in 10 mM NaCl, 0.5 M H2SO4. For this electrolysis, a graphite counter electrode was employed and was separated from the working solution by a Nafion 117 membrane. The resulting AgCl-coated wire was encased in a glass tube closed at one end with a ceramic frit, which was encased in another larger glass tube with a ceramic frit tip. The potential of the reference electrode fabricated as such was –0.333 V vs Hg/Hg2SO4, or +0.307 V vs SHE at room temperature. Potentials at high temperature was also converted to the SHE scale by adding 0.307 V. While reference redox potentials can vary with temperature,36 we find that using this conversion value leads to background Pt H UPD wave potentials that coincide between room temperature and 130 ˚C (Figure 4.8). Hence, we decided to ignore temperature-dependent potential shifts in our studies. After operation at high temperature, the reference electrode typically showed a 0–3 mV variation in potential, which was ignored in the analysis of the data. After each EMOR trial, the AgCl-coated Ag wire was treated by application of ~5 μA cm−2 galvanostatic oxidation current in 10 mM NaCl, 0.5 M H2SO4 for ~30 min to increase its longevity.

138

The counter electrode (CE) and counter compartment. The counter electrode was a Pt mesh separated from the working solution with H+-conducting membranes. For long-term EMOR trials, we

+ needed to prevent the reduction of H to H2 at the counter electrode because H2 was found to diffuse into the working solution and reduce the Pt ions to metallic Pt0. Thus, we dissolved 3 M of vanadyl sulfate

IV ((V O)(SO4)) into the counter compartment electrolyte (10 mM NaCl in 0.5 M H2SO4) to function as a surrogate electron acceptor; the blue VIVO2+ ions are reduced to the green VIII ions at potentials more positive than H+ reduction, therefore functioning as the terminal oxidant in our system. With the vanadyl ions, no H2 was detected in the headspace GC analysis.

Due to the high reaction temperatures and the presence of reactive Pt ions, we employed two materials in conjugation to generate a viable H+-conducting composite membrane. The H+-conducting membrane stack consisted of alternating layers of Nafion HP (20 μm thick, PTFE-enhanced) and polybenzimidazole (PBI) membranes; Nafion is chemically stable towards Pt ions, but has a low operating temperature range (up to around 80 ℃). Specifically, the glass transition temperature of Nafion is 110 ℃,37 and at 130 ℃ we observed loss of ionic conductivity for the thicker Nafion 117 and slow electrolyte leakage for the thinner Nafion HP. On the other hand, the PBI membranes retain their performance at high temperature, but contain aryl C-H bonds that can be activated by PtII.38 Indeed, we observed that PBI membranes are discolored upon heating in the presence of Pt ions (Figure 4.9a). To overcome these issues with Nafion and PBI, we alternatingly stacked five pairs of Nafion HP-PBI membranes and added 3–4 layers of Nafion at the side facing the working solution that contains Pt ions. In order to firmly hold the membrane stack, a custom-designed apparatus was used (Figure 4.10). A glass tube (wall thickness 2 mm) with a perforated glass disk on one end was snugly fitted into a PTFE tube with a hole in the bottom. Between the bottom of the glass tube and the PTFE tube, a PTFE gasket, the H+-conducting membrane

139 stack, and another PTFE gasket were placed successively. Then, to seal the edge of the membranes, we pressed the glass tube against the PTFE tube by tightening a PTFE screw top (see Figure 4.10). With this configuration, only the first one or two layers of PBI were damaged (Figure 4.9b), and leakage of the vanadyl ions into the working solution or Pt ions into the counter solution was prevented.

Figure 4.9. (a) The polybenzimidazole (PBI) membrane appearance changes from the left to right after II IV heating in the presence of 3 mM K2Pt Cl4 and 7 mM Na2Pt Cl6 in 10 mM NaCl, 0.5 M H2SO4 for 19 h at 130 ˚C. (b) The five PBI layers (see Figure 4.10) after EMOR reactor operation. Blackened areas show oxidative degradation and Pt0 deposition. The periphery, where PTFE gaskets were placed, is clear because it was not exposed to the solution. While the 1st layer (closest to the working solution) showed significant blackening and degradation, the 2nd layer showed drastically reduced blackening, and the last layer showed almost no sign of degradation. Incidentally, the color of the membranes is darker than the pristine membrane shown in (a) because of the activation pretreatment.

Figure 4.10. Counter compartment design for the EMOR reactor.

4.4.1.3. Quantitation of Pt ions by UV–Vis spectroscopy Quantitation of [PtII] and [PtIV] was performed with UV–Vis spectroscopy (Cary 50, Agilent).

IV 2– II 2– Pt Cl6 ions in aqueous solutions display a strong absorption at 262 nm, where Pt Cl4 has minimal

39 II 2– absorption at this wavelength (see below). Pt Cl4 ions show an absorption maximum at 214 nm, but this

IV 2– peak is often obscured by the strong absorbance of Pt Cl6 at this wavelength in mixed solutions, impeding

140 direct deconvolution of the spectra to the determine [PtII] (however, see the Note below). Instead, the [PtII] concentration was calculated by subtracting the [PtIV] from the total concentration of Pt ions, [Ptn]. [Ptn] was determined by treating each solution with SnCl2.

n II IV Determination of [Pt ]. In the presence of excess SnCl2, both Pt and Pt undergo complexation

II − 40 IV with Sn Cl3 to give a strong orange-red complex that displays an absorption maximum at 404 nm (Pt

II 41 is reduced by SnCl2 to Pt prior to complexation). The standard protocol was to dilute the sample solution in a solution of 1 M SnCl2 in 3 M HCl and let it react for >5 min. For accurate determination of the extinction coefficient at 404 nm, a Beer’s Law plot was constructed with solutions of PtII and PtIV whose Pt

II IV concentrations were determined by ICP-MS. Stock solutions of Pt and Pt were prepared from K2PtCl4 and Na2PtCl6, respectively. These stock solutions were diluted to three different concentrations in the 1 M

SnCl2 in 3 M HCl solution. The background-subtracted absorbance was then plotted against the concentration determined by ICP-MS to obtain the extinction coefficient (Figure 4.11a). The average value 8.1 × 103 cm–1 M–1 was taken as the extinction coefficient for determination of [Ptn].

Figure 4.11. Calibration curves for Pt ion quantitation by UV–vis. (a) Absorbance at 404 nm of (black) K2PtCl4 and (red) Na2PtCl6 diluted in 1 M SnCl2, 3 M HCl. (b, c) Absorbance at 262 nm of K2PtCl4 and Na2PtCl6 diluted in 1 M HCl following irradiation with a UV lamp.

II IV n Determination of [Pt ] and [Pt ]. [Pt ], determined by the method above, and A262nm, the combined absorbance of the two ions at 262 nm (see below), were inputted into the following equations.

II IV εPtIV and εPtII denote the extinction coefficients of Pt and Pt at 262 nm, and d denotes the dilution factor.

IV n [Pt ] = (A262nm/d – εPtIV*[Pt ])/(εPtIV – εPtII)

[PtII] = [Ptn] – [PtIV]

II 2– IV 2– Importantly, Pt Cl4 and Pt Cl6 ions undergo hydrolysis over time, and the species with fewer Cl− ligand exhibit different extinction coefficients. To exclude convolution from Cl– ligand speciation, each sample was diluted in 1 M HCl and irradiated with a 4 W UV lamp (252 or 365 nm) for 5−25 min prior to the measurement of absorbance at 262 nm. This procedure ensures complete anation of the Pt ions.42 To

141 determine εPtIV and εPtII, freshly prepared stock solutions of K2PtCl4 and Na2PtCl6 were serially diluted in 1

4 –1 –1 2 M HCl and UV–vis spectra were recorded (Figure 4.11b,c). εPtIV = 2.4 × 10 cm M and εPtII = 4.0 × 10 cm–1 M–1.

II 2– *Note: During the course of our work, we learned that the second absorption maximum of Pt Cl4 ions at 230 nm, though lower in extinction coefficient (7.2 × 103 cm–1 M–1),43 is suitable for determination

II IV 2– of [Pt ] because absorption by Pt Cl6 ions is at a minimum at this wavelength. An alternative quantitation protocol that uses the absorbance at 230 nm and 262 nm showed identical results to the protocol described

– above that uses the absorbance at 262 nm and the absorbance at 404 nm from the SnCl3 -complex of Pt ions.

4.4.1.4. Determination of methane oxidation products

CH3OH, CH2(OH)2 and HCOOH. These solution-phase products were determined by NMR (Varian 500 MHz or Bruker 500/600 MHz instruments) with various solvent suppression techniques to suppress the H2O peak (presaturation, excitation sculpting, or wet). The sample solution was mixed with a

D2O solution containing acetic acid as an internal standard (caution: prolonged storage of this internal standard solution compromises the measured concentration via slow H/D exchange of the acetic acid methyl protons in D2O), and then adjusted to ~2 M total acid concentration by the addition of 8 M D2SO4. This procedure was employed because the peak position of CH2(OH)2 (hydrated form of formaldehyde, which is the predominant form in 0.5 M H2SO4) was close to that of the solvent water peak; lowering the pH shifted the water peak downfield and allowed us to observe and integrate the CH2(OH)2 peak (representative spectrum in Figure 4.12). We found that the high acid concentration impeded probe tuning; this could be improved by using a 3 mm-diameter NMR tube instead of a 5 mm-diameter tube to reduce the sample volume. The 90˚-pulse width was also manually calibrated for each sample. When only the concentration of CH3OH or HCOOH was of interest (e.g. during electrolyte optimization: see 4.4.4), the addition of 8 M

D2SO4 was omitted. The relaxation time was 15 sec or longer. The spectra were processed in MestReNova with phase and baseline corrections. The high acid concentration also caused poor shimming in the spectra, and thus the integrals of the NMR peaks were calculated using the sum method without any peak fitting.44

For determining methane oxidation products in the reactor, solutions were collected from the working compartment, reference compartment, and droplets condensed on the inner walls. All of them contained some product because the high temperature of the reactor allowed for continuous evaporation and re-precipitation. The high concentration of paramagnetic vanadyl ions in the counter compartment precluded NMR quantification of products in the counter chamber. Nonetheless, our ability to account for

142 the majority of the oxidation charge passed (see 4.4.3 or Table 4.2) indicates negligible losses to the counter chamber.

Figure 4.12. Representative baseline-corrected (Whittaker smoother) spectrum of the working solution from an EMOR trial (entry 3 in Table 4.1). The “wet” pulse sequence was employed for solvent suppression. The spectrum is referenced to the acetic acid peak at 2.0 ppm.

Table 4.2. Estimated Faradaic efficiencies of different EMOR trials.

Faradaic Time (h) Efficiency

4.9 90.6% 10.5 96.4% 18.4 97.8% 29.3 94.0% 10.5 (5× concentrations)a 101.4% aSee 4.4.6.5 for description of this EMOR trial.

CO2 and CH3Cl. These gaseous products were determined by gas chromatography (GC) analysis (SRI instruments, model 8610C) of the reactor headspace gas after the reactor was cooled to room temperature. CO2 was calibrated by serial dilution of a commercial calibration gas (Product no.

X08AR98C33A0000, Airgas) with Ar (Figure 4.13a). CH3Cl was calibrated by diluting CH3Cl gas (Sigma

Aldrich) in a septum-capped, air-filled 1 L flask along with equal amounts of CH4, which was already calibrated as it was a component of the calibration gas (Figure 4.13b). In this way, systematic error was minimized. To account for the amount of CO2 and CH3Cl dissolved in the solution phase, Henry’s constants

–1 –1 at room temperature were taken from the NIST webbook (0.034 and 0.12 mol kg bar for CO2 and CH3Cl, respectively) to calculate the solution concentration of each gas from its partial pressure determined by GC.

143

Figure 4.13. Calibration curves for quantitation of (a) CO2 and (b) CH3Cl by gas chromatography. See the text for details.

4.4.1.5. Estimation of reaction rates of PtII-catalyzed C–H oxidations In order to assess reaction rates of non-electrochemical catalysis by PtII (i.e. PtIV are stoichiometric oxidants and no reoxidation of PtII occurs), solutions of PtII + PtIV were heated in the presence of substrate in heavy-walled NMR tubes or glass ampules.

To measure the rate of methane oxidation, heavy-walled NMR tubes (Norell, item no. S-5-500- HW-7) were charged with solutions of PtII and PtIV, pressurized to 100 psi of methane, and manually agitated for >2 min to allow thorough gas-liquid mixing and dissolve methane. The tubes were placed in a stirred oil bath and heated to 130 ˚C. After a set time (typically ~1.5 h), the tubes were cooled down and the solution was withdrawn and analyzed by NMR.

To measure the oxidation rate of methanol, formaldehyde, and formic acid, solutions of PtII and PtIV containing the substrate were flame-sealed in scored glass ampules (Kimble Chase, 1 mL, item no. 12010L-1), placed in an aluminum heating block with silicone oil, and heated to 130 ˚C.

4.4.2. Evaluation of PtII electro-oxidation

4.4.2.1. Carbon electrodes Carbon is a versatile and inexpensive electrode material, and so we initially evaluated carbon electrodes for EMOR. Initial observations were promising; a glassy carbon electrode shows a clear electrochemical oxidation wave in the presence of PtII ions (Figure 4.14a, wave A). Though the potential was quite high in the first scan (Ep,a = 1.3 V), in subsequent cyclic voltammetry (CV) scans, a new wave at a lower overpotential appeared (wave B; Ep,a = 1.0 V). When a constant potential was applied using a large surface area carbon electrode (e.g. carbon paper or carbon felt), bulk conversion of PtII to PtIV could be

144 achieved, at potentials corresponding to either wave A or B. However, the electrode underwent gradual deactivation at both potentials (Figure 4.14b).

Figure 4.14. Investigation of PtII oxidation on a glassy carbon (GC) electrode using a solution of 2 mM II II K2Pt Cl4 in 10 mM of NaCl, 0.5 M H2SO4. (a) CVs with and without Pt . (b) Chronoamperometric (CA) traces at different applied potentials. (c) Series of CVs obtained on the same GC electrode. The 4th cycle was recorded after 220 min of chronoamperometry at 1.14 V that deactivated the electrode. (d) Overlay of PtII oxidation CVs on GC and Pt electrodes. [PtII] = 2 mM for GC and 1 mM for Pt. The currents were normalized to geometric surface areas. Scan rates = 100 mV s–1.

In order to rationalize these observations, we put forth the following set of hypotheses, based on the series of CVs acquired successively (Figure 4.14c):

(i) Wave A corresponds to the oxidation of PtII to surface-adsorbed PtIV species.45 These PtIV species are reduced to Pt0 in the negative scan of the CV and activates the electrode for PtII oxidation at a lower potential that corresponds to wave B (Figure 4.14c, 1st cycle vs 3rd cycle). In support of this hypothesis, the average of the forward and return waves of B returns a redox potential close to that of PtII/IV couple on a Pt electrode (Figure 4.14d).

(ii) The surface-bound Pt0 species, during constant polarization, undergoes oxidative dissolution over time, deactivating the electrode (Figure 4.14b and Figure 4.14c 4th cycle). The electrode can be regenerated by re-adsorption of the PtIV species and subsequent re-reduction (Figure 4.14c, 4th & 5th cycle).

145

(iii) Based on the observation that PtII is first oxidized to a surface-adsorbed PtIV species,45 we postulate that accumulation of the surface-adsorbed PtIV may passivate the surface and contribute to the current decay at the higher potential of 1.39 V (Figure 4.14b).

In spite of these difficulties, we investigated the carbon electrodes a little further. A test bulk electrolysis of PtII to PtIV at room temperature was conducted with a high surface area carbon electrode (e.g. graphite felt or carbon paper), and bulk conversion of PtII ions to PtIV ions could be achieved (Figure 4.15). However, the Faradaic efficiency (FE) was always ~50%. This implies the presence of parasitic oxidation process that occurs concomitantly with PtII oxidation, which we attribute to oxidative degradation of the carbon electrode. Nonetheless, encouraged by the observations that (i) PtII could be converted to PtIV at the electrode and that (ii) deactivation could be partially reversed upon negative polarization of the electrode to ~0.5 V vs SHE (Figure 4.14c), we tried using carbon electrodes in our EMOR reactors. In brief, we employed a piece of graphite felt as the working electrode, and programmed the potentiostat to apply a cathodic potential for a set amount of time when the anodic current decays below a threshold value, after the reactor was pressurized with methane and heated. Unfortunately, under these conditions we found it difficult to pass a net positive charge and control the rate of PtII oxidation (Figure 4.16). In summary, our results indicate that carbon is not a suitable electrode material for realizing EMOR with PtII reoxidation.

II Figure 4.15. Bulk electrolysis of a solution of 2 mM K2Pt Cl4 and 10 mM NaCl in 0.1 M H2SO4 on a graphite felt electrode. The solution was sampled periodically to measure the PtII and PtIV concentrations by UV–Vis spectroscopy.

146

Figure 4.16. (Red) A portion of the current trace during EMOR reactor operation using a graphite felt II IV working electrode. The solution contained 0.4 mM of Pt , 1.4 mM of Pt and 10 mM NaCl in 0.5 M H2SO4 and was 12.3 mL in volume. Concentration of PtII was kept to a minimum for this preliminary trial because of the low efficiency of the electrode for oxidizing PtII. (Blue, dotted) Cumulative charge passed during the experiment. Because a lot of negative charge flowed during the cathodic polarization for electrode regeneration, it was difficult to pass a net positive charge over time; i.e., Q rises and falls over the course of each potential step, but shows no long-term increase. The post-reaction PtII concentration was indicative of little PtII oxidation over the course of the electrolysis. For reference, the complete oxidation of the PtII ions in the solution would require 474 mC of charge passed.

4.4.2.2. Fluorine-doped tin oxide (FTO) electrodes FTO is a relatively inexpensive electrode material that has been shown to be remarkably robust in highly acidic and oxidizing environments.35 Therefore, we investigated the ability of FTO to effect electrochemical oxidation of PtII to PtIV. As shown in Figure 4.17, we could initially observe a PtII oxidation wave at E ≈ 1.5 V. However, the PtII electro-oxidation activity at this potential quickly decayed upon several cycles (Figure 4.17, red-pink-brown). PtII electro-oxidation was sustained at higher potentials (E > 1.8 V), and the Faradaic efficiency was 86% at E = 1.84 V. Nonetheless, the current density was only 0.15 mA cm– 2 even at this extremely high potential. The CVs shown in Figure 4.17 were collected in the presence of 100 mM of Cl– in an effort to facilitate PtII oxidation; similar results were obtained in the absence of Cl–. In summary, our results indicate that FTO is not a suitable electrode material for realizing EMOR with PtII reoxidation.

147

II II Figure 4.17. Investigation of Pt oxidation on an FTO electrode. The solution contained 1 mM K2Pt Cl4 II and 100 mM HCl in 0.5 M H2SO4 . The CVs of Pt on FTO were acquired in the order from red to pink to II brown. (Dotted blue) CV of Pt in 0.5 M H2SO4 obtained on Pt electrode is overlaid for comparison. FTO CVs were normalized by the geometric surface area, and the Pt CV was normalized by the surface area measured by H UPD. Scan rates = 100 mV s–1.

4.4.2.3. Pt electrodes Additional information for the interpretation of PtII CVs

II II 2– Suppression of background oxide features in the presence of Pt . The Pt Cl4 ion undergoes slow acid hydrolysis in aqueous solutions with a rate constant of 4 × 10–5 s–1.46 Therefore, after 5 min., a freshly

II – prepared 1 mM K2Pt Cl4 solution will have generated 0.01 mM Cl , and after an hour, 0.13 mM. This small concentration of Cl– is sufficient to suppress the formation of a surface O/OH layer on Pt at low potentials (Figure 4.18).26

148

Suppression of PtIV reduction in the presence of 10 mM Cl–. According to the preceding analysis

II 2– – and 4.2.1, in the presence of Pt Cl4 ions, the Pt surface is effectively covered with Cl to suppress oxide formation and catalyze PtII oxidation even without addition of exogenous Cl– ions. However, the PtIV reduction wave was suppressed only upon addition of exogenous Cl– (Figure 4.1a). This is because the surface concentration of adsorbed Cl– depends on the electrode potential – with high potentials favoring Cl– adsorption. The Cl– adsorption isotherm determined using radioactive Cl– (Figure 4.19)27 shows that the isotherm collected at 2 mM Cl– is almost saturated at 0.8 V vs SHE, which is the peak potential at which PtIV reduction was observed (Figure 4.1a). In the absence of exogenous Cl–, this adsorption isotherm will be shifted to higher potentials that allows for suppression of the surface oxide wave and catalysis of PtII oxidation (see above) while still providing available sites for back-reduction of PtIV. Upon addition of 10 mM added Cl–, the isotherm shifts to lower potentials, blocking PtIV reduction.

Figure 4.19. Variation of the quantities of adsorbed Cl– with potential. The solution consisted of 1 × 10–3 N NaCl and 1 × 10–3 N HCl. Potentials are vs NHE. Replotted from ref. 27 with permission from the editorial staff of Russian Chemical Reviews.

Blocking effect of the surface oxide at high potentials While Cl– ions adsorb to Pt electrodes and suppress surface oxide formation, at high potentials (above 1.1 V vs RHE) the electrode surface acquires adsorbed O and/or OH species (Figure 4.18). Since PtII oxidation is facilitated via a Cl-bridged inner-sphere electron transfer mechanism, such an oxide layer suppresses PtII oxidation. However, in practice, we observed such suppression in the CV only at high PtII concentration (Figure 4.20a, blue vs red). We believe this is because at low [PtII], the limiting current density for PtII oxidation (i.e. mass transport-controlled PtII oxidation current) is low and, thus, there are

149 enough open (or Cl-adsorbed) sites on the electrode despite partial coverage by the oxide. At higher [PtII] where the flux of PtII ions coming to the electrode is greater, the sites available for PtII oxidation on the electrode become saturated and we observe suppression in the CV. We also note that at higher temperature, the CV shows more pronounced suppression both because the flux of PtII ions is greater and because the oxide layer forms faster (leading to higher oxide coverage on the electrode surface at the potentials where PtII is oxidized). If we polarize the electrode at 1.11 V, which is above the potential at which oxide formation begins, we observe a progressive passivation of the electrode (Figure 4.20b, red).

Figure 4.20. (a) Cyclic voltammograms obtained on a Pt disk electrode in N2-purged solutions containing II –1 0, 1, and 10 mM K2Pt Cl4 in 10 mM NaCl, 0.5 M H2SO4 electrolyte at RT. Scan rates = 100 mV s . The CV of 10 mM PtII is plotted at 5-fold reduced current density to match the vertical scale for easier comparison. (b) Chronoamperometric traces obtained on a Pt wire electrode at 130 ˚C in a stirred solution II of 10 mM of K2Pt Cl4 in 10 mM NaCl, 0.5 M H2SO4 electrolyte. The current densities are normalized by the electrochemically active surface area determined by integration of the H UPD wave on Pt.

Acquisition of current-overpotential relationship (Tafel plot) The PtII oxidation Tafel plot at 130 ˚C, shown in Figure 4.1c, was obtained from a series of 30 sec chronoamperograms at varying stepped potentials spanning from 0.837 V to 1.057 V in 30–40 mV intervals (Figure 4.21). At each potential, the average current over the last 5 sec was taken as the steady-state value. The overpotentials were calculated by subtracting the observed OCP, 0.829 V.

150

Figure 4.21. Stepped-potential chronoamperometry on a Pt wire electrode at 130 °C. This raw data was used to construct the Tafel plot for PtII oxidation (Figure 4.1c). The stirred solution contained 5 mM of PtII IV and 5 mM of Pt in 10 mM NaCl, 0.5 M H2SO4.

0 4.4.2.4. Assessment of Pt -catalyzed non-electrochemical oxidation of CH3OH Figure 4.1d highlights that Pt electrodes, which are good methanol electro-oxidation catalysts, are passivated towards electrochemical CH3OH oxidation by surface-adsorbed chloride. However, metallic Pt

32 is also known to catalyze the non-electrochemical oxidation of CH3OH in the presence of oxidants. As

II IV 0 Pt and Pt ions possess sufficient oxidizing power for CH3OH oxidation (E = 0.68–0.76 V vs SHE for PtIV/PtII, PtII/Pt0, and PtIV/Pt0),22 the Pt working electrode may effect non-Faradaic oxidation of methanol during EMOR.

II We evaluated the importance of such processes by heating 8 mM of CH3OH with 2 mM Pt and 8

IV – mM Pt in 10 mM Cl , 0.5 M H2SO4 solutions along with pieces of metallic Pt in glass ampules. To account

II for the oxidation of CH3OH catalyzed by Pt alone, control experiments were performed in parallel with ampules that do not contain the metallic Pt pieces. The concentrations of PtII/IV and Cl– ions and the size of the Pt pieces were chosen to match the concentrations and working electrode surface area used in our EMOR reactors. In detail, our ampules contained 0.9 mL of the test solution, whereas our EMOR trials were carried out with 23 mL of working solution and a Pt foil working electrode whose electrochemically active surface area was measured to be 10.3 cm2. Thus, to match the ratio of Pt surface area to solution volume, the surface area of metallic Pt pieces in the ampules had to be 0.4 cm2 (0.4/0.9 = 10.3/23). These were prepared by measuring the electrochemically active surface area of a long Pt wire in 0.5 M H2SO4, then cutting them into pieces that would each expose approximately 0.4 cm2 of surface area. Following 3 h of heating at 130 ˚C, 2.8 ± 0.2 mM CH3OH was oxidized in the presence of Pt metal, whereas the same amount, 2.7 ± 0.4 mM, was oxidized in the absence of metallic Pt. Based on this data, we conclude that Faradaic or non-Faradaic overoxidation of the methanol product is negligible on Pt electrodes during EMOR.

4.4.3. Faradaic efficiency measurements

Faradaic efficiency (FE) is defined by the moles of product of electron transfer divided by the moles of electrons that were passed through the circuit. In all cases the number of moles of electrons passed was calculated by dividing the integrated charge passed by Faraday’s constant.

4.4.3.1. Bulk electrolysis of PtII to PtIV at 130 ˚C

22 or 23 mL of 5 mM of K2PtCl4, 5 mM Na2PtCl6 and 10 mM NaCl in 0.5 M H2SO4 were oxidized with stirring at a Pt foil working electrode. A pure PtII solution was not used because of its tendency towards

151 disproportionation and Pt0 precipitation at elevated temperatures (see 4.4.5). The [PtIV] at the end was measured by UV–Vis spectroscopy (see above) to calculate the μmol of PtIV generated (ΔPtIV). Then FE values were calculated using the follow equation,

FE = 2*ΔPtIV/(μmol of e–)

At the three different potentials that were examined, the Faradaic efficiencies were ~100% (Table 4.3). We note that the small error in this measurement arises from the difficulty of accurately measuring the solution volume, which decreases at the end of the electrolysis due to evaporation.

Table 4.3. Results of bulk electrolysis of PtII to PtIV at 130 ˚C with stirring at 200 rpm. The solution initially II IV II contained 5 mM of K2Pt Cl4, 5 mM Na2Pt Cl6, and 10 mM NaCl in 0.5 M H2SO4 (initial amount of Pt = 110–115 μmol).

Duration E (V vs SHE) e– passed (μmol) ΔPtIV (μmol) FE (%) (min) 0.874 78 111.4 55.7 103% 0.924 40 109 54.5 106% 0.974 17 115 57.5 103%

4.4.3.2. Faradaic efficiency of EMOR reactors In the presence of methane, PtIV in the solution is consumed by PtII–catalyzed oxidation of methane or products generated from methane oxidation. Thus, the overall Faradaic efficiency was calculated by summing the μmoles of each methane oxidation product multiplied by the respective number of oxidizing

IV IV IV equivalents required to generate each product and the change in the amount of Pt ions (ΔPt = Pt final –

IV Pt initial). In particular, the following equation was employed,

IV FE = 2 × (nCH3OH + nCH3Cl + 2 × nCH2(OH)2 + 3 × nHCOOH + 4 × nCO2 + ΔPt )/ne-

where ni denotes the moles of species i. Solutions in the working compartment, in the reference electrode compartment, and droplets condensed on the inner surfaces of the reactor were separately collected and analyzed by NMR to determine the concentrations of CH3OH, CH2(OH)2 and HCOOH. These were multiplied by their respective solution volumes and combined. As noted above, NMR quantitation of the counter compartment solution could not be carried out due to the high concentration of paramagnetic vanadium species. The headspace gas was analyzed for CH3Cl and CO2 (see above).

Because the FE for PtII electro-oxidation is 100%, the FE of EMOR is also expected to be 100%. Indeed, we observe FE values close to 100% (Table 4.2). The missing FE may be accounted for by the products in the counter compartment that were not accounted for or errors that accrue from NMR and GC integrations, solution volume measurement for each compartment, etc. 152

– II 4.4.4. Effect of [H2SO4] and [Cl ] on the catalytic C-H oxidation activity of Pt

In order to select the electrolyte environment for carrying out EMOR, we explored the effect of electrolyte composition on the catalytic activity of PtII for methane oxidation and undesired methanol

II IV oxidation. Solutions containing 3 mM K2Pt Cl4 and 7 mM Na2Pt Cl6 were prepared with different concentrations of acid and Cl− and heated with methane or methanol to assess the catalytic activity for favorable methane oxidation and undesired methanol oxidation.

4.4.4.1. Choice of acid and the effect of its concentration The reaction solution must be sufficiently acidic to prevent polynuclear hydrolysis of the platinum ions at elevated temperatures.23 Conveniently, the requirement for acid environments automatically results in an electrically conductive solution required for electrochemistry. Sulfuric acid electrolytes were chosen because of their high chemical stability, low-cost, and minimal interference with C–H activation by PtII.47 The concentration of sulfuric acid showed a small yet measurable effect on the rate of PtII-catalyzed oxidation of methane to methanol and further oxidation of methanol (Figure 4.22a,b). The rate of methane oxidation decreased monotonically with increasing acid strength. This is expected because protonolysis of

II the Pt –CH3 intermediate competes with the subsequent steps in the catalytic cycle (Scheme 4.1, top).

However, the rate of methanol oxidation showed a minimum at 0.5 M H2SO4 and increased at higher acid concentrations. Therefore, we selected 0.5 M as the optimal H2SO4 concentration.

Figure 4.22. Measurement of (a, c) methane functionalization and (b, d) methanol oxidation activities under – II different concentrations of (a, b) H2SO4 and (c, d) Cl . Catalyst and oxidant loadings were [Pt ] = 3 mM and [PtIV] = 7 mM. Solutions were heated to 130 ˚C for 1.5 h for (a, c) methane functionalization and 3 h for (b, d) methanol oxidation tests. Error bars correspond to standard errors from ≥3 independent measurements.

153

4.4.4.2. Effect of chloride concentration The effect of [Cl–] on the rate of PtII-catalyzed C–H oxidation was assessed at 0, 10, and 100 mM of added Cl−. Both methane oxidation and methanol oxidation showed some suppression by Cl− (Figure

38,48 II II − 4.22c,d), in accordance with the literature. Because Pt Cl2(H2O)2 is more active than Pt Cl4 or

II − – Pt Cl3(H2O) for the C–H activation step, the reaction is inhibited by Cl , and the magnitude of the inverse reaction order increases with increasing [Cl–]. With the knowledge that Cl– is necessary to suppress methanol oxidation, we selected 10 mM as the optimal Cl– concentration.

Incidentally, we also prepared a sample that contains a “negative” concentration Cl− by preparing

IV II IV the Pt ions from bulk electrolysis of K2Pt Cl4 rather than from Na2Pt Cl6. Surprisingly, this solution actually showed slower production of methanol compared to a solution containing equal concentrations of

II IV IV 2– IV Pt and Pt but Pt Cl6 as the Pt ions (Figure 4.22c).

4.4.5. Mitigation of Pt0 formation in the EMOR reactor

4.4.5.1. Thermodynamics and kinetics of Pt0 formation As mentioned in the introduction, Shilov’s catalyst has the propensity to decompose to Pt0, especially when oxidant is depleted. Two mechanistic possibilities have been put forth: the reduction of PtII to Pt0 by the PtII–R intermediate (equation 4.2):8

Pt Cl + Pt RCl(HO) → Pt RCl(HO) + Pt + (푥 + 푦 − 푧) Cl … 4.2 or the disproportionation of PtII (equation 4.3):10

2 Pt Cl ⇄ Pt Cl + Pt + (2x − y) Cl … 4.3 While it is beyond the scope of this work to parse the mechanistic details, we note our observation that reaction 4.3 is facile when PtIV is depleted. As shown in Table 4.4, entry 1, a PtII solution without any PtIV precipitated out Pt0 in less than 4 h when heated at 130 ˚C, and showed a reduction in [PtII] and an increase in [PtIV] consistent with a disproportionation reaction. In contrast, a solution containing 3 mM PtII, 7 mM PtIV and 10 mM Cl− (the solution composition used in our EMOR trials) did not show any visible Pt0 depositions within the time duration we examined (46 h), and comparison of [PtII] and [PtIV] before and after the heat treatment revealed no change (Table 4.4, entry 2).

Importantly, this lack of Pt0 formation is of kinetic origin, rather than thermodynamic. As shown in Figure 4.23, the solution composition for our EMOR reactor and entry 2 in Table 4.4 favors disproportionation thermodynamically. We attribute the observed lack of Pt0 formation to a kinetic barrier to nucleation. Namely, incipient Pt0 nuclei with high surface energy may undergo facile oxidative dissolution by reaction with PtIV. Based on the classical nucleation theory of La Mer,49 observable Pt0

154 formation will only occur when these Pt0 clusters are able to reach a critical nucleus size from which the additional growth of the particle becomes downhill. In accordance with this hypothesis, when Pt ions are heated in the presence of Pt0 seed particles, some disproportionation occurs (entry 3, Table 4.4; decrease in [PtII] and increase in [PtIV]), presumably due to Pt0 deposition onto the surface of pre-existing Pt0 particles. Incidentally, the Pt electrode in our EMOR reactor does not seem to catalyze this disproportionation reaction (see last entry of Table 4.5) because of the constant anodic polarization that oxidizes the PtII ions to PtIV. While the detailed mechanism of Pt0 nucleation may be more complex,50 our observation indicates that a sufficiently high concentration of PtIV is able to suppress Pt0 formation. Additionally, when the solution composition was set to favor the reverse of this disproportionation, non-electrochemical dissolution of Pt0 to PtII could be observed (entry 4, Table 4.4), which suggests that the PtII catalyst may be regenerated by electrochemically adjusting the PtII:PtIV ratio.

Table 4.4. Results of heating PtII/IV solutions in sealed ampules. Solutions contained combinations of NaCl, II IV K2Pt Cl4, and Na2Pt Cl6 in 0.5 M H2SO4. T = 130 ˚C. Ampules for entries 3 and 4 also contained a few mg of Pt0 particles.

Initial Initial Initial Final Final visible Pt0 Entry Duration (h) Initial Pt0? [Cl–] [PtII] [PtIV] [PtII] [PtIV] precipitation? 1 29 0 10.0 0 No 1.4 4.3 Yes; within <4 h 2 46 10 2.6 6.7 No 2.6 6.7 No n/a (added at the 3 26 10 2.6 6.7 Yes 2.1 6.9 beginning) n/a (added at the 4 46 500 0.0 28.5 Yes 14.4 22.7 beginning)

155

0 II IV − II (m−2)− Figure 4.23. (a) Experimentally determined Pt -Pt -Pt -Cl equilibria for the reaction, 2 [Pt Clm] ⇄ 0 IV (n−4)− − 34 Pt + [Pt Cln] + (2m–n) Cl . Reproduced from ref. with permission from Elsevier. The equilibrium constant for this reaction, K, corresponds to the y-intercept of the given plot according to the equation Y = (2m–n) X – log K, where X and Y denote the x- and y- axis values, respectively. The reason why the slope of the data depends on [Cl−] is evident from this equation; at high [Cl−], m~4 and n~6 so that (2m−n)~2, but at low [Cl−], both m and n decrease and (2m−n) deviates from 2. (b) Overlay of a few selected solution compositions on the experimental equilibrium curve. Blue: [PtII] = 3 mM, [PtIV] = 7 mM, [Cl−] = 10 mM. Green: [PtII] = 15 mM, [PtIV] = 35 mM, [Cl−] = 50 mM. Red: [PtII] = 50 mM, [PtIV] = 500 mM, [Cl−] = 200 mM.

Table 4.5. Amount of Pt0 deposition from reactor operations of varying time duration.

Time (h) Pt loss as Pt0 4.9 3.0% 10.5 4.4% 18.4 6.5% 29.3 13% 10.5 (5× concentrations)a 0.0% aSee 4.4.6.5 for description of this EMOR trial.

4.4.5.2. Observation of Pt0 in the reactor As discussed in 4.2.2 and above, PtII decomposes to Pt0 when the oxidant, PtIV, is depleted. In our reactors where the PtII:PtIV ratio was constantly monitored and controlled, no Pt0 was visible in the well- stirred portion of the working solution. However, we did observe Pt0 deposits in corners and nooks of the reactor that experiences inhibited mass transport. First, we observed specks of grey Pt0 on the upper parts of the glass cell wall where droplets of the reaction solution had splashed (Figure 4.24); isolated from the bulk solution in contact with the electrode, these droplets were depleted of PtIV from the reaction with methane, leading to Pt0 deposition. For long reaction times, grey Pt0 deposited were also observed along the meniscus of the reaction solution (Figure 4.24) and within the narrow crevice between the H+- conducting membranes and the PTFE block, both areas that experienced inhibited convection during the reactor trial. Additionally, as mentioned earlier, Pt0 deposition was found on/inside the H+-conducting membranes (Figure 4.9). We attribute this to the slow leakage of Pt ions through the thin Nafion HP membrane (~20 µm) and their reaction with the polybenzimidazole membrane. We emphasize that there was no visible Pt0 in the well-stirred part of the solution where PtIV replenishment through mass transport was unhindered, demonstrating the importance of maintaining an appropriate PtII:PtIV ratio. In contrast, if a solution of PtII and PtIV is reacted with methane without concomitant electrochemical oxidation to regenerate PtIV, visible Pt0 colloids are observed suspended through the solution.

The amount of Pt that deposited as Pt0 was calculated from the difference in total μmol of PtII and PtIV ions before and after the reaction. Dividing this by the initial μmol of Pt ions, we obtain the % loss of 156

Pt ions (Table 4.5) for each EMOR trial shown in Table 4.1 and the concentration scale-up trial (see 4.4.6.5). The amount of the irreversible Pt0 deposition increases with increasing reactor operation time. The higher concentration trial showed negligible Pt0 loss, which may be due to the higher overall PtIV concentration. Both observations are consistent with Pt0 formation being suppressed by PtIV.

Figure 4.24. The glass cell after a 29 h reactor operation. Arrows point to adventitious Pt0 deposits.

4.4.6. Additional information on the EMOR reactor

4.4.6.1. Control experiment for assessing product oxidation by VO2+ in the counter compartment Evaporation and re-precipitation allow the methane oxidation products to migrate to other parts of the reactor, including the reference and counter compartments. The counter compartment contained a high (3 M) concentration of vanadyl sulfate, which has an oxidation potential high enough to oxidize methanol. To estimate the degree of product oxidation by vanadyl ions, we assembled the cell with a working compartment solution containing Pt-free 0.5 M H2SO4 electrolyte spiked with 4.6 mM of CH3OH (105

μmol total). Analogous to the EMOR trials, the cell was pressurized with CH4 and heated at 130 ˚C. After

37 h, we measured 0.7 μmol of CH2(OH)2 and 2.5 μmol of CO2. From this result, we estimate CH3OH in the working solution is oxidized to CH2(OH)2 in the working solution and CO2 with a yield of 0.017% and

0.065% per hour, respectively. While CH3OH and CH2(OH)2 in the vanadyl solution cannot be quantified (see above), this does not affect our estimation as this was also true for our EMOR trials. Additionally, the aggregate moles of carbon (from CH3OH, CH2(OH)2, and CO2) in the reactor at the end of the experiment was very similar to the initial moles of CH3OH (104.8 vs 105.2 μmol). From this and the results in Table

4.1, we estimate that ~2% of the total CH2(OH)2 and ~10% of the total CO2 measured from the EMOR trials may be attributed to oxidation by vanadyl ions in the counter compartment. As both CH2(OH)2 and

CO2 are minor products in our EMOR trials, we ignored this contribution in our analysis.

4.4.6.2. Mathematical analysis of the relationship between the applied current, catalytic rate, and the PtII:PtIV ratio

157

Given the 100% Faradaic efficiency of PtII/IV oxidation at the Pt electrode, the molar rate of PtII oxidation, rox, is directly proportional to the applied current (i):

푖 풓 = … 4.4 풐풙 2퐹푉 where F is Faraday’s constant and V is the volume of the reaction solution. The denominator contains a factor of 2 to account for the two electrons required for each PtII/IV oxidation reaction. The rate

II of methane oxidation catalysis, rcat, is first-order in [Pt ]:

풓풄풂풕 = 푘[Pt ] … 4.5

where kobs is the observed pseudo-first order rate constant under the CH4 pressure and temperature

IV II conditions employed. For every catalytic turnover, an equivalent of Pt is reduced to Pt , and, thus, rcat has

II II a positive contribution to d[Pt ]/dt. On the other hand, rox has a negative contribution to d[Pt ]/dt. Overall, we obtain:

푑[Pt] = 풓 − 풓 … 4.6 푑t 풄풂풕 풐풙 = 푘[Pt ] − 풓풐풙 … 4.7

For a fixed value of applied current, rox is time-invariant, thus integration yields:

풓 [Pt] = 퐶푒 + 풐풙 … 4.8 푘 Upon solving for the integration constant C using the initial conditions, we obtain:

풓풐풙 풓풐풙 [Pt ] = ([Pt ] − )푒 + … 4.9 푘 푘 II II If rox exactly equals the rate of Pt -catalyzed C–H functionalization (rox = kobs[Pt ]t=0), the time- dependent exponential term in equation 1.1 will go to zero and [PtII] will remain constant over time.

II However, even very small differences between rox and kobs[Pt ]t=0 will result in a non-zero exponential term

II II IV that will cause the [Pt ] and, thus, the Pt :Pt ratio, to rapidly deviate from its initial value over time. If rox

II is constantly re-adjusted to match rcat, however, [Pt ] can be maintained at a steady-state. Therefore, these

II equations highlight the need to constantly modulate the rate of Pt electro-oxidation, rox.

4.4.6.3. The relationship between overpotential (η) and reactor configuration The required η is determined by the current density (j) required for steady-state catalysis, and j equals the required current (i) divided by the electrode area (A). Since i depends on the reactor solution volume (V) and the catalytic rate constant (kobs) (equations 4.4 and 4.7 above), the magnitude of η will also depend on these parameters.

While enlarging A will decrease η and, thus, the electrical energy input, it will also increase

0 electrode cost and may increase the rate of parasitic Pt -catalyzed CH3OH oxidation. We underscore that in

158 our reactors, η was quite small (<50 mV) even when the electrode was sufficiently small so as to observed

II negligible surface-mediated CH3OH oxidation (see 4.4.2.4). Also, because the rate of Pt electro-oxidation at any η is proportional to [PtII], the [PtII] can be increased to increase the overall rate of catalysis without requiring additional overpotential.

4.4.6.4. Explanation for normalization of product concentration by iave in Figure 4a Since electrochemical oxidation provides the oxidizing equivalents, a linear correlation exists between the total methane oxidation products and the charge passed (Figure 4.25, left). If the electric current (i), which is by definition charge passed per time, is constant throughout the EMOR trial, a linear correlation will also arise between methane oxidation products and reaction time. However, because of constant modulation of i with minor fluctuations in [PtII], i is not strictly constant. Indeed, if the product concentration is simply plotted versus reaction time, we observed slight deviations from linearity resulting from small variations in iave between EMOR trials (Figure 4.25, right, black squares). To exclude the effect of these known variations in iave, we normalize the amount of product by the average current of each EMOR trial, leading to a clear linear correlation between the amount of product and reaction time (Figure 4.25, right, red circles). Product amounts in µmol were converted to concentration (mM) by dividing by the reaction solution volume, 23 mL.

Figure 4.25. The total amount of methane oxidation products from the four EMOR trials (Table 4.1) plotted against (left) the amount of charge passed and (right) the reaction time. The product moles were calculated in a way that counts the total number of oxidation events required to generate each product (μmolTotalProduct = μmolCH3OH + μmolCH3Cl + 2*μmolCH2(OH)2 + 3*μmolHCOOH + 4*μmolCO2). When the total product amount is plotted against the total charge passed (left) in each run, a linear correlation is observed. When the total product is plotted against reaction time (right, hollow black squares), the trend line exhibits slight deviations from a straight line because each EMOR trial had a slight variation in the average current and total charge passed. To account for this variation, we divided the product sum for each trial by the average current of each trial (iave, italicized numbers in the plot) recovering a linear plot (right, hollow red circles; also Figure 4.4a).

4.4.6.5. High-concentration EMOR trial

159

All of the EMOR experiments reported in 4.2.2 were performed with identical reaction solution compositions ([PtII] = 3 mM, [PtIV] = 7 mM, [Cl–] = 10 mM). In order to gain further insight into the behavior at high Pt concentrations, we conducted EMOR with a 5-fold increase in concentrations to [PtII] = 15 mM, [PtIV] = 35 mM, and [Cl–] = 50 mM. Data for this trial, along with a trial with regular concentrations and identical operation time, are shown in Table 4.6. We observe the follow trends upon increasing concentrations:

II i) The overall rate of catalysis (rcat) increased with increasing Pt concentration. However, the increase was not 5-fold but rather ~2.5-fold because of the inhibitory effect of Cl– (see 4.4.4.2), which is reflected in the ~2-fold reduction in TON and TOF.

– ii) Due to the higher Cl concentration, more CH3Cl and less CH3OH were formed.

iii) The fraction of CO2 was lower even though the reactor was run for the same amount of time

0 and generated a higher CH3OH concentration. In combination with the fourth observation (negligible Pt deposition), this observation supports the hypothesis set forth in 4.4.7, that further oxidation of CH2(OH)2 and HCOOH inside the EMOR reactor may have been catalyzed by Pt0.

iv) There was almost no loss of Pt ions as Pt0, presumably due to a higher concentration of PtIV (see 4.4.5).

Table 4.6. EMOR reactor results from two trials where the run duration was identical (10.5 h) but the concentrations of PtII, PtIV and Cl– differed by a factor of 5.

d d –1 i b Final PtII Product (μmol (rel. fraction)) approx. TON approx. TOF (h ) [PtII] ave (mA) (%) CH3OH CH3Cl CH2(OH)2 HCOOH CO2 CH3X Total CH3X Total 93.7 27.9 5.1 1.2 4.4 3 0.88 19% 2.7 3.4 0.26 0.32 (71%) (21%) (4%) (1%) (3%) 178.9 131.6 32.6 3.0 4.9 15 2.47 20% 1.2 1.5 0.11 0.15 (51%) (37%) (9%) (1%) (1%)

4.4.7. Simulation of reactions in the EMOR reactor

4.4.7.1. Simulation details The concentrations of various methane oxidation products were calculated numerically with the simple mechanism in Figure 4.4b using Microsoft Excel.

For CH3OH,

푑[CH OH] = 푘 [CH ] − 푘 [CH OH] + 푘 [CH Cl] 푑t

[CHOH] = [CHOH] + 푘[CH]∆푡 − 푘[CHOH]∆푡 + 푘[CHCl]∆푡 160

For CH2(OH)2,

푑[CH (OH) ] = 푘 [CH OH] − 푘 [CH (OH) ] 푑t

[CH(OH)] = [CH(OH)] + 푘[CHOH]∆푡 − 푘[CH(OH)]∆푡

For HCOOH,

푑[HCOOH] = 푘 [CH (OH) ] − 푘 [HCOOH] 푑t

[HCOOH] = [HCOOH] + 푘[CH(OH)]∆푡 − 푘[HCOOH]∆푡

For CO2,

푑[CO ] = 푘 [HCOOH] 푑t

[CO] = [CO] + 푘[HCOOH]∆푡

For CH3Cl,

푑[CH Cl] = 푘 [CH ] − 푘 [CH Cl] 푑t

[CHCl] = [CHCl] + 푘[CH]∆푡 − 푘[CHCl]∆푡

∆t was set to 0.0093 h and further decreasing this value led to no difference in the simulation outputs. Based on a Henry’s law calculation at the temperature and pressure of the reactor, the [CH4] was set to 44 mM.51 We note that this value is an approximation that derives from extrapolation of the Henry’s law constant function to the temperature of the reactor (403 K); the function is validated in the range T =

273–361 K. PCH4 and [CH4] were taken to be constant throughout the EMOR trial because the amount of methane that was converted to products in our EMOR reactors (<400 µmol for the longest EMOR trial) was negligible compared to the amount of methane in the large headspace (~200 mmol). Then, the parameters k1 through k6 were adjusted manually until a good fit with experimental reactor data was achieved. The fitted parameters are given in Table 4.7. We emphasize that the fitted parameters are crude estimates of apparent rate constants because of the low data-to-parameter ratio in the fit and the lack of independent validation of the reaction mechanism (Figure 4.4b). Nonetheless, this model provides a useful approximation of the reactions occurring during EMOR.

Table 4.7. Apparent rate constants from fitting experimental data with the mechanism in Figure 4.4b.

k1 k2 k3 k4 k5 k6 Value (mM–1 h–1) 1.1×10–2 1.9×10–2 8.5×10–2 4.0×10–1 3.9×10–3 5.7×10–2

161

4.4.7.2. Independent determination of relative rate constants outside the EMOR reactor

CH4 vs CH3OH oxidation To the best our knowledge, two literature reports6,52 evaluate the selectivity of aqueous PtII chloride salts for CH4 vs CH3OH oxidation. We note that CH4 vs CH3OH oxidation selectivity may differ from RCH3 vs RCH2OH oxidation selectivity. As these two literature reports reported significantly different selectivities, we performed selectivity measurements as well. The different relative rates are summarized in Table 4.8.

II II Parenthetically, a model Pt complex in trifluoroethanol, (N-N)Pt (CH3)(TFE) (N-N = ArN=C(CH3)–

5 C(CH3)=NAr, TFE = trifluoroethanol), showed relative rates of C–H activation of kCH4/kCH3OH = 0.77.

For our experiment, two identical high-pressure NMR tubes were charged with the same solution

II IV of 3 mM Pt and 7 mM Pt in the 10 mM NaCl, 0.5 M H2SO4 electrolyte. One contained 7.5 mM of CH3OH while the other did not. The tube without CH3OH (Tube 1) was pressurized with 100 psi of CH4, while the tube containing CH3OH (Tube 2) was pressurized with 100 psi of Ar. Another heavy-walled NMR tube was charged with blank electrolyte containing internal standards and pressurized with 100 psi of CH4 (Tube 3). The three heavy-walled NMR tubes were heated together in an oil bath for 1 h and 10 min at 130 ˚C, then quantitated for the amount of CH3OH and compared with the initial CH4 or CH3OH concentration. The initial CH4 concentration in Tube 1 was estimated from Tube 3, which showed [CH4] = 8.6 mM before heating and 5.7 mM after heating due to reduced solubility of methane at elevated temperatures; this range in [CH4] concentrations lead to a corresponding uncertainty range in the conversion percentage. As shown in Table 4.10, from the 5.7–8.6 mM of methane in Tube 1, we observed 1.1 mM of methanol formed, corresponding to 13–19% conversation of the CH4. From Tube 2, of the 7.5 mM of methanol starting material, 1.2 mM was oxidized in net, corresponding to 16% oxidation of the CH3OH. Taking the ratio of the product conversion percentages, we estimate kCH4/kCH3OH to be 0.8–1.2. These results are summarized in Table 4.8 along with other literature values.

CH3OH, CH2(OH)2, HCOOH oxidation Sealed glass ampules containing solutions of PtII, PtIV and the different substrates in 10 mM NaCl,

0.5 M H2SO4 electrolyte were heated at 130 ˚C. The decrease in substrate concentrations for different time durations are compared in Table 4.8.

Table 4.8. Experimentally determined relative rates of C–H oxidation of CH4 and CH3OH in the literature and this work.

T (˚C) PCH4 (atm) PO2 (atm) Duration (h) kCH4/kCH3OH Ref. 105 41, 83 14 >300 0.17a 52

162

120 10 0 1 6 6 130 6.9 0 1.17 0.8–1.2 This work aAuthors mention possibility of Pt0 formation during the reaction.

Table 4.9. Evaluation of PtII-catalyzed C–H oxidation of various substrates at 130 ˚C. The test solutions II IV contained 3 mM Pt and 7 mM Pt in 10 mM NaCl, 0.5 M H2SO4.

Substrate (=S) [S]initial (mM) Duration (h) Δ[S]/[S]initial (%)

CH3OH 11.13 3 44 (ave. of 3 trials) 9.17 3 40 (ave. of 3 trials)

a CH2(OH)2 11.18 3 25 (ave. of 4 trials) HCOOH 10.89 3 9.3 (no Cl–) 8.94 8.5 21

a For CH2(OH)2, some CH3OH was present initially because they were added as a polymerization inhibitor in the concentrated formaldehyde bottle. Δ[CH2(OH)2] was calculated by subtracting Δ[CH3OH] from Δ[CH2(OH)2] in order to account for CH2(OH)2 generated from CH3OH oxidation.

II IV Table 4.10. Concentrations of CH4 and CH3OH before and after reaction with 3 mM Pt and 7 mM Pt in 10 mM NaCl, 0.5 M H2SO4 at 130 ˚C.

Initial Final Reacted amount

Tube 1 [CH4] = 5.7–8.6 mM [CH3OH] = 1.1 mM 1.1 mM (13–19%)

Tube 2 [CH3OH] = 7.5 mM [CH3OH] = 6.3 mM 1.2 mM (16%)

4.4.7.3. Possible explanations for the discrepancy between the rates The rate constants derived from simulation and stoichiometric reactions outside the EMOR reactor are all apparent or observed rate constants (kobs) which are extrinsic values that depend on the reaction conditions employed. While this precludes a direct comparison between the two sets of rate constants, comparison of the ratios of these rate constants can be made.

II Doing this comparison, we find that the selectivity of Pt for CH4 over CH3OH was similar (k1/k2

= 0.6 vs 0.8–1.2 for EMOR-simulated vs non-EMOR estimation), but rates of further oxidation of CH3OH showed greater discrepancies (k2/k3 = 0.2 vs >1 and k3/k4 = 0.2 vs ≫1). We postulate that these differences

0 may point to Pt -catalyzed oxidation of CH2(OH)2 and HCOOH at the electrode surface. While we show that Cl-adsorption effectively suppresses surface-mediated oxidation of CH3OH, surface-mediated oxidation of CH2(OH)2 and HCOOH may be incompletely suppressed. This highlights that the apparent rate constants derived from simulation of the actual reactor may have contributions beyond those resulting exclusively from PtII-catalyzed oxidation.

163

4.5. References

(1) da Silva, M. J. Synthesis of Methanol from Methane: Challenges and Advances on the Multi-Step (Syngas) and One-Step Routes (DMTM). Fuel Process. Technol. 2016, 145, 42–61. (2) Ravi, M.; Ranocchiari, M.; van Bokhoven, J. A. The Direct Catalytic Oxidation of Methane to Methanol—A Critical Assessment. Angew. Chemie - Int. Ed. 2017, 56 (52), 16464–16483. (3) Labinger, J. A.; Bercaw, J. E. Mechanistic Studies on the Shilov System: A Retrospective. J. Organomet. Chem. 2015, 793, 47–53. (4) Latimer, A. A.; Kakekhani, A.; Kulkarni, A. R.; Nørskov, J. K. Direct Methane to Methanol: The Selectivity-Conversion Limit and Design Strategies. ACS Catal. 2018, 8 (8), 6894–6907. (5) Owen, J. S.; Labinger, J. A.; Bercaw, J. E. Kinetics and Mechanism of Methane, Methanol, and Dimethyl Ether C-H Activation with Electrophilic Platinum Complexes. J. Am. Chem. Soc. 2006, 128 (6), 2005–2016. (6) Labinger, J. A.; Bercaw, J. E.; Luinstra, G. A.; Lyon, D. K.; Herring, A. M. Organometallic Methane Activation: Functionalization by Aqueous Platinum Complexes. In Studies in Surface Science and Catalysis; 1994; pp 515–520. (7) Michalkiewicz, B. Assessment of the Possibility of the Methane to Methanol Transformation. Polish J. Chem. Technol. 2008, 10 (2), 20–26. (8) Labinger, J. A. Chapter 2. Alkane Functionalization via Electrophilic Activation. In Catalysis by Metal Complexes: vol.38, Alkane C-H Activation by Single-Site Metal Catalysis; Pérez, P. J., Ed.; Catalysis by Metal Complexes; Springer Netherlands: Dordrecht, 2012; Vol. 38, pp 17–71. (9) Gunsalus, N. J.; Koppaka, A.; Park, S. H.; Bischof, S. M.; Hashiguchi, B. G.; Periana, R. A. Homogeneous Functionalization of Methane. Chem. Rev. 2017, 117 (13), 8521–8573. (10) DeVries, N.; Roe, D. C.; Thorn, D. L. Catalytic Hydroxylation Using Chloroplatinum Compounds. J. Mol. Catal. A Chem. 2002, 189 (1), 17–22. (11) Weinberg, D. R.; Labinger, J. A.; Bercaw, J. E. Competitive Oxidation and Protonation of Aqueous Monomethylplatinum(II) Complexes: A Comparison of Oxidants. Organometallics 2007, 26 (1), 167–172. (12) Lin, M.; Shen, C.; Garcia-zayas, E. A.; Park, U. V; Pennsyl, V.; June, R. V; Sen, A.; Park, U. V; Pennsyl, V.; June, R. V. Catalytic Shilov Chemistry: Platinum Chloride-Catalyzed Oxidation of Terminal Methyl Groups by Dioxygen. J. Am. Chem. Soc. 2001, 123 (5), 1000–1001. (13) Bar-Nahum, I.; Khenkin, A. M.; Neumann, R. Mild, Aqueous, Aerobic, Catalytic Oxidation of Methane to Methanol and Acetaldehyde Catalyzed by a Supported Bipyrimidinylplatinum- Polyoxometalate Hybrid Compound. J. Am. Chem. Soc. 2004, 126 (33), 10236–10237. (14) Kreutz, J. E.; Shukhaev, A.; Du, W.; Druskin, S.; Daugulis, O.; Ismagilov, R. F. Evolution of Catalysts Directed by Genetic Algorithms in a Plug-Based Microfluidic Device Tested with Oxidation of Methane by Oxygen. J. Am. Chem. Soc. 2010, 132 (9), 3128–3132. (15) Lee, M.; Sanford, M. S. Platinum-Catalyzed, Terminal-Selective C(sp3)-H Oxidation of Aliphatic Amines. J. Am. Chem. Soc. 2015, 137 (40), 12796–12799. (16) Horváth, I. T.; Cook, R. A.; Millar, J. M.; Kiss, G. Low-Temperature Methane Chlorination with Aqueous Platinum Chlorides in the Presence of Chlorine. Organometallics 1993, 12 (1), 8–10. (17) Liu, S. F.; Nusrat, F. Electrocatalytic Shilov Chemistry for the Oxidation of Aliphatic Groups. Mol. Catal. 2019, 463, 16–19.

164

(18) Freund, M. S.; Labinger, J. A.; Lewis, N. S.; Bercaw, J. E. Electrocatalytic Functionalization of Alkanes Using Aqueous Platinum Salts. J. Mol. Catal. 1994, 87 (1), L11–L15. (19) Lappin, G. Redox Mechanisms in Inorganic Chemistry; Ellis Horwood: New York, 1994. (20) Jude, H.; Krause Bauer, J. A.; Connick, W. B. An Outer-Sphere Two-Electron Platinum Reagent. J. Am. Chem. Soc. 2003, 125 (12), 3446–3447. (21) Cushing, J. P.; Hubbard, A. T. Study of the Kinetics of Electrochemical Reactions By Thin Layer Voltammetry. II. Electro-Oxidation of Platinum(II) Complexes. Electroanal. Chem. Interfacial Electrochem. 1969, 23, 183–203. (22) Vanysek, P. (5) Thermo, Electro & Solution Chemistry: Electrochemical Series. In CRC Handbook of Physics and Chemistry, 99th ed.; 2018. (23) Elding, L. I. Preparation and Properties of the Tetra-Aquaplatinum(II) Ion in Perchloric Acid Solution. Inorganica Chim. Acta 1976, 20, 65–69. (24) Scortichini, C. L.; Reilley, C. N. Surface Characterization of Pt Electrodes Using Underpotential Deposition of H and Cu. V. Characterization of BD Pt Catalyst Surface. J. Catal. 1983, 79 (1), 138– 146. (25) Jerkiewicz, G.; Vatankhah, G.; Lessard, J.; Soriaga, M. P.; Park, Y. S. Surface-Oxide Growth at Platinum Electrodes in Aqueous H2SO4 Reexamination of Its Mechanism through Combined Cyclic-Voltammetry, Electrochemical Quartz-Crystal Nanobalance, and Auger Electron Spectroscopy Measurements. Electrochim. Acta 2004, 49 (9–10), 1451–1459. (26) Novak, D. M.; Conway, B. E. Competitive Adsorption and State of Charge of Halide Ions in Monolayer Oxide Film Growth Processes at Pt Anodes. J. Chem. Soc. Faraday Trans. 1 Phys. Chem. Condens. Phases 1981, 77 (10), 2341–2359. (27) Balashova, N. A.; Kazarinov, V. E. Study of the Structure of the Electrical Double Layer on Platinum by the Radioactive Tracer Method. Russ. Chem. Rev. 1965, 34 (10), 730–736. (28) Compton, R. G.; Banks, C. E. Understanding Voltammetry, 2nd ed.; World Scientific Publishing: Singapore, 2007. (29) Zhao, X.; Yin, M.; Ma, L.; Liang, L.; Liu, C.; Liao, J.; Lu, T.; Xing, W. Recent Advances in Catalysts for Direct Methanol Fuel Cells. Energy Environ. Sci. 2011, 4 (8), 2736. (30) Chung, D. Y.; Lee, K. J.; Sung, Y. E. Methanol Electro-Oxidation on the Pt Surface: Revisiting the Cyclic Voltammetry Interpretation. J. Phys. Chem. C 2016, 120 (17), 9028–9035. (31) Snell, K. D.; Keenan, A. G. Chloride Inhibition of Ethanol Electrooxidation at a Platinum Electrode in Aqueous Acid Solution. Electrochim. Acta 1981, 26 (9), 1339–1344. (32) Sen, A.; Lin, M.; Kao, L. C.; Hutson, A. C. C-H Activation in Aqueous Medium. The Diverse Roles of Platinum(II) and Metallic Platinum in the Catalytic and Stoichiometric Oxidative Functionalization of Organic Substrates Including Alkanes. J. Am. Chem. Soc. 1992, 114 (16), 6385– 6392. (33) Cardoso, D. S. P.; Šljukić, B.; Santos, D. M. F.; Sequeira, C. A. C. Organic Electrosynthesis: From Laboratorial Practice to Industrial Applications. Org. Process Res. Dev. 2017, 21 (9), 1213–1226. (34) Gammons, C. H. Experimental Investigations of the Hydrothermal Geochemistry of Platinum and Palladium: V. Equilibria between Platinum Metal, Pt(II), and Pt(IV) Chloride Complexes at 25 to 300°C. Geochim. Cosmochim. Acta 1996, 60 (10), 1683–1694. (35) O’Reilly, M. E.; Kim, R. S.; Oh, S.; Surendranath, Y. Catalytic Methane Monofunctionalization by an Electrogenerated High-Valent Pd Intermediate. ACS Cent. Sci. 2017, 3 (11), 1174–1179. 165

(36) Bratsch, S. G. Standard Electrode Potentials and Temperature Coefficients in Water at 298.15 K. J. Phys. Chem. Ref. Data 1989, 18 (1), 1–21. (37) Yadav, R.; Fedkiw, P. S. Analysis of EIS Technique and Nafion 117 Conductivity as a Function of Temperature and Relative Humidity. J. Electrochem. Soc. 2012, 159 (3), B340–B346. (38) Shilov, A. E.; Shul’pin, G. B. Activation and Catalytic Reactions of Saturated Hydrocarbons in the Presence of Metal Complexes; Catalysis by Metal Complexes; Kluwer Academic Publishers: Dordrecht, 2002; Vol. 21. (39) Kirkland, J. J.; Yoe, J. H. Ultraviolet Spectrophotometric Determination of Platinum. Anal. Chim. Acta 1953, 9, 441–445. (40) Ayres, G. H.; Meyer, A. S. Spectrophotometric Study of the Platinum(IV)-Tin(II) Chloride System. Anal. Chem. 1951, 23 (2), 299–304. (41) Young, J. F.; Gillard, R. D.; Wilkinson, G. 992. Complexes of Ruthenium, Rhodium, Iridium, and Platinum with Tin(II) Chloride. J. Chem. Soc. 1964, 8 (5176), 5176. (42) Elding, L. I.; Gustafson, L. Kinetics and Mechanism for Chloride Anation of Some Platinum(IV) Aqua Complexes in the Presence of Platinum(II). Inorganica Chim. Acta 1976, 19 (C), 31–38. (43) Elding, L. I. L. I.; Olsson, L. F. Electronic Absorption Spectra of Square-Planar Chloro-Aqua and Bromo-Aqua Complexes of Palladium(II) and Platinum(II). J. Phys. Chem. 1978, 82 (1), 69–74. (44) Bernstein, M. Getting REALLY fussy about integration resources.mestrelab.com/getting-really- fussy-about-integration/ (accessed Mar 7, 2019). (45) Xu, L.; Li, F.; Dong, S. Electro-Oxidation of a Chloride Complex of Platinum(II) at a Glassy Carbon Electrode. J. Electroanal. Chem. 1995, 383 (1–2), 133–137. (46) Elding, L. I.; Gröning, A. B. The Solvent Path in Square-Planar Substitutions. Kinetics and Mechanism for Reactions of Tetrachloroplatinate(II) Andaquachloroplatinates(II) with Halides, Thiocyanate and Dimethyl Sulfoxide. Inorganica Chim. Acta 1978, 31 (C), 243–250. (47) Shilov, A. E.; Shteinman, A. A. Activation of Saturated Hydrocarbons by Metal Complexes in Solutions. Kinet. Catal. 1977, 18 (5), 924–965. (48) Gol’dshleger, N. F.; Shteinman, A. A. Pt(II) Complexes in Activation of Saturated Hydrocarbons. React. Kinet. Catal. Lett. 1977, 6 (1), 43–50. (49) Mer, V. K. La. Nucleation in Phase Transitions. Ind. Eng. Chem. 1952, 44 (6), 1270–1277. (50) Colombo Ciacchi, L.; Pompe, W.; De Vita, A. Initial Nucleation of Platinum Clusters after Reduction of K2PtCl4 in Aqueous Solution: A First Principles Study. J. Am. Chem. Soc. 2001, 123 (30), 7371– 7380. (51) Yaws, C. L. Yaws’ Handbook of Thermodynamic and Physical Properties of Chemical Compounds https://app.knovel.com/web/toc.v/cid:kpYHTPPCC4/viewerType:toc/ (accessed Dec 12, 2018). (52) Sen, A.; Benvenuto, M. A.; Lin, M.; Hutson, A. C.; Basickes, N. Activation of Methane and Ethane and Their Selective Oxidation to the Alcohols in Protic Media. J. Am. Chem. Soc. 1994, 116 (3), 998–1003.

166

Acknowledgements

In addition to being a journey of great scientific learning, the years in graduate school at MIT was also a time of intense personal growth for me. I owe my achievements and who I am to all those mentioned below and those whom I could not include due to limited space and memory.

Yogi, my advisor, has been a supportive and trustworthy mentor ever since I first met him over the phone interview and through the numerous ups and downs I experienced. His unassuming and quirky manners have always made me feel comfortable around him and encouraged me to be myself. Yet I admire him for the depth and rigor of his thoughts as a scientist, and these qualities of him that I experienced during graduate school will continue to inspire me and push me to be a better scientist in the future. He was not afraid to think and probe deeper until we could get to the bottom of the matter; there are many times that I came to him proudly with my data and thoughts, only to have him surprise me by revealing a new way to think about it or something that I missed. I also benefited significantly from non-scientific conversations that he generously engaged in despite his busy schedule. As he is someone who genuinely cares about the success of his students, I likewise wish him all the best; I am very happy for the official announcement of his tenure that literally just happened as I am writing this.

I am also thankful to my thesis committee and other MIT faculty members. Professor Mircea Dincă, who served as my thesis chair, helped me parse out the objective value of a scientific question from my personal excitement by his calm demeanors, and would surprise me by casually providing valuable feedback and questions. I thank him for his faithful service and supervision. I am especially thankful to Professor Christopher Cummins, who graciously accepted my request to join my thesis committee in my final year. I thank him for the genuine interest he showed in my thesis work and his thoughtful suggestions and feedback. While Professor Daniel Suess did not directly serve on my committee, he kindly answered my EPR questions and helped me with postdoctoral applications, for which I am very grateful. I am also thankful for the wonderful classes I took at MIT, and would like to thank, in addition to those mentioned, Professor Stephen Lippard, Professor Liz Nolan, and Professor Jennifer Rupp.

I thank my colleagues and collaborators for their contribution. Dr. Matthew O’Reilly, with his fearless endeavors, discovered the PdIII dimer in sulfuric acid and spearheaded difficult and dangerous experiments. I thank him for being a pioneer and patiently teaching and working with me, accommodating my different style and personality. Dr. Randall Field at the MIT Energy Initiative helped us prepare meetings with our industrial sponsor and taught me the engineering aspect of chemical processes, and I am grateful for his kindness and sincerity as well as his expertise. Professor Jeffrey Miller and his student Dr. Evan

167

Wegener helped us with X-ray absorption spectroscopy and careful interpretation of the data. I greatly appreciate their scientific integrity and infinite patience with our work. The DFT collaboration with Professor Thomas Cundari and his student Dr. Azadeh Nazemi opened up new directions in our structural and mechanistic investigation of the PdIII dimer. I am deeply thankful for their interest in this project and their courage to tackle such ill-defined and difficult calculations. I also enjoyed our numerous scientific discourses over email and Slack.

Many thanks go to the staff at the Department of Chemistry Instrumentation Facility (DCIF) and others who serve the department. Particularly, Dr. Bruce Adams enabled me to carry out the NMR kinetic experiments with NMR tubes filled with fuming sulfuric acid and pressurized with methane. He took the time to explain various aspects of NMR and answered my many questions as fast as he could. I also thank him for always greeting me warmly and inviting me to serve as a DCIF steward. John Grimes diligently served the DCIF and helped me with the EPR; it is difficult to imagine the DCIF without him. Once he hurried back from the Muddies to save me from a potential catastrophe with the NMR sample changer, for which I am forever grateful. I also thank him for fun memories, including him gulping down a bag of Christmas cookies in one sitting. Dr. Walt Massefski supervised the DCIF with all his mind and heart, overcoming many obstacles and troubleshooting problems over weekends. I learned a lot from his vision and thoughtfulness, and I greatly appreciate his willingness to interact personally with the students. My work would not have been possible without these people. I also thank Joanne Baldini, Michele Harris, Jennifer Weisman, Brian Pretti, and Scott Ide for their dedicated service and warm heart.

Among those in the Surendranath team, Megan Jackson was one of the first people I met, as she was assigned as my mentor by Women In Chemistry (WIC). She then turned out to be in Yogi’s lab, my TA, and a member of the Graduate Christian Fellowship (GCF), which became an essential part of my life at MIT. I am indebted to her for her kind friendship, in which I found a confidante, refuge, wisdom, and lots of laughter. I am very glad that the defense cake she and Corey sent did not get lost. Another dear friend is Bing Yan, with whom I entered the program and Yogi’s lab in 2014. Being with her always makes me feel warm, and I am grateful for her steadfast love. I also thank her for scientific conversations and casual interactions in the lab as deskmates. I was very happy when she found a postdoctoral position at MIT so that I could still see her after her graduation. Seokjoon Oh also entered the program and the group together, and I thank him for his willingness to help every time I would ask something. I thank Anna Wuttig, Sterling Chu, Youngmin Yoon, Joe Falkowski, Shoji Hall, Dimitri Vaughn, and Tomohiro Fukushima for setting up the lab and shaping the unique culture of our group in the beginning. I especially thank Tomo for his mentorship and kindness in my first year.

168

I thank Corey Kaminsky and Travis Marshall-Roth for being smart, fun, and reliable companions going through the program. I thank Jaeyune Ryu for his wise advice about my thesis work and many conversations in Korean over scientific and non-scientific matters. I thank Andrew Licini for his sincere heart and deep conversations, and Jonathan Melville for accompanying me to Argonne National Lab and running the lab twitter. I thank Marcel Schreier for his help with gas chromatography and for the expertise, passion, and openness he brought into the group. I thank William Howland for sharing the lab space and for his gentle spirit, Onyu Jung for taking on many responsibilities in the lab and for our chats in Korean, Jeffrey Rosenberg for new perspectives and dedication to what he believes is right, and Thejas Wesley for his kind help in several things and patient pursuit of scientific truth. I thank An Chu, Bryan Tang, James Toh, Noah Lewis, Deiaa Harraz, and Sophia Weng for the fresh atmosphere they create in the lab, and I am excited for what they will accomplish in the coming years. I really enjoyed working with them and getting to know them as friends. I thank Sasha Alabugin for his open-hearted friendship and mind-stimulating conversations. I also thank Sahr Khan, Alex Murray, Chris Hendon, Ryan Bisbey, Michael Pegis, and Patrick Smith for the diverse personalities and skills that they brought into the group. Particularly, I thank Ryan for his kind spirit that makes the lab a friendlier place, Mike for his probing questions that improved the analysis of my kinetic data, and Patrick for his help with gaining a deeper understanding about the PdIII dimer.

I found my family in Cambridge in the MIT Graduate Christian Fellowship (GCF) community. I thank Megan Jackson, Stephanie Lam, Gerald Pho, Michelle Chang, Heather Beem, Samuel Perli, Frank Yaul, Annie Chen, Peng Shi, Yukkee and Ming-Zher Poh, and Anya Burkart for welcoming me to MIT, loving me in my weakest moment, and being examples of living by faith in graduate school. I thank Linyi Gao, Joshua Bosshardt, Kathleen Davis, Grace Goon, Sam Elder, Clement Li, Daisy Green, AJ Fenton, Olivia Fiebig, Linda Zhong, Randi Williams, Sika Gadzanku, David and Adeline Miculescu, Jin Wu, Brian Yue, Maria Eugenia Inda, Phillip Daniel, Rachel Luo, Johanna Barbour, and Elizabeth Young. Their genuine friendship and the many hours we spent together outside of work, in musical worship, meals, Bible studies, prayers, retreats, and other activities have encouraged and sustained my soul through graduate school, just like peanut butter & jelly sandwich for my body. It would be impossible to recount all the memories and conversations. I thank Kevin Ford for faithfully serving this community and providing perspective and wisdom.

I also thank my Park Street Church friends not mentioned above, Lei Ma, Ashley Sun, Jennifer Wang, and Patrick Cheung for their love and precious memories unique to each friend. I have experienced pure joy in my friendship with them. Annie Chen and Ellen and Jerry Gabrielse have edified and supported me continually with love, wisdom, and prayer through the years, and I am grateful for their generous heart

169 towards me. I thank the prayer community at MIT, including Luke Shimanuki, Mary Thompson, Professor Jing Kong, and Celia Wang, with whom I could navigate the final year of my PhD with greater peace and confidence. I feel so blessed as I look back upon the years and think of these people, I had absolutely no idea when entering MIT.

Additionally, I thank Kelley Danahy for her continued friendship and for reading over this thesis with her sensitive and agile mind. It was so helpful and encouraging. I thank Sara Park, Hyowon Seo, Byungsoo Kwon, Sangyeon Cho, Hyewon Moon, Kyungsuk Jin, Minjoo Chung, Changhae Kim, and Minjung Son for the Korean community which reminded me of home and personal interactions over the years. I thank Mengshan Ye and Arun Sridharan for help with the 3rd year oral exam, and Sahag Voskian for answering my questions about distillation as well as his kind friendship. I am thankful for the Chemistry Student Seminar (CSS) team, to which Anna Wuttig invited me to serve. I thank her and Kathleen White, Timothy Barnum, Michael Geeson, Martin Riu, and recently Bryan Tang and Arun Sridharan for carrying on the great tradition.

My landlady Esther In and landlord Peter Kye literally fed me since my second year, and their loving presence has let me feel at home at home. I thank them for their delicious Korean food and hard work that support many hungry and poor students in Cambridge. Going back farther in time, I thank Professor Taek Dong Chung for introducing the world of electrochemistry to me and mentoring me with an appreciation of who I am as a person. His scientific vision and gracious personality continue to inspire me.

I am infinitely grateful to my family in Korea, especially my parents, who carried me through the hardest time in my life with many tears and bleeding hearts. I am sorry that they cannot celebrate this milestone with me in person due to COVID-19, but hope that this will not diminish their joy and pride. I deeply respect them for their humble, faithful, and truthful way of life. I realize that it has been a great comfort growing up to know that I can turn to them at any time. My enjoyment of science derives in large part from the love of nature and knowledge that my father Sunghyun Kim passed down to me. I thank him for his gentle encouragements and quiet love. I thank my mother, Hyunsook Yang, for her dedication to raising us up and her prayers. She taught me an appreciation of life and people, and instilled in me a constant yearning for spiritual growth. I also thank my brother Joonsik Kim and my sister Gayoung Kim, who bring me joy just by being who they are. All the ways the three of us are similar and different are always so interesting. Last but not least, I thank Hyun-kee Harry Lee, who has quickly become a precious part of my life and has suffered through the final semester and the pandemic with me. It is hard to imagine what it would have been like without him. I am amazed by all the laughter and joy he brings, and love. I also thank him for reading through my thesis and convincing me to use Oxford commas.

170

Above all, I thank my Father in Heaven, who has given me every good, perfect, and undeserved gift in my life. I thank Him for the eternal hope He has given for all people, the hope that makes me live and strive towards my utmost.

171