This document is the Accepted Manuscript version of a Published Work that appeared in final form in Journal of the American Chemical Society, copyright © American Chemical Society after peer review and technical editing by the publisher. To access the final edited and published work see https://pubs.acs.org/doi/pdf/10.1021/jacs.8b06303 Electrochemically Driven Water-Oxidation Catalysis Beginning with Six Exemplary Cobalt Polyoxometalates: Is It Molecular, Homogeneous Catalysis or Electrode-Bound, Heterogeneous CoOx Catalysis? Scott J. Folkman, Joaquin Soriano-Lopez, José Ramoń Galan-Mascaros, Richard G. Finke Chemistry Department, Colorado State University, Fort Collins, Colorado 80523, United States Institute of Chemical Research of Catalonia (ICIQ), The Barcelona Institute of Science and Technology (BIST), Av Països Catalans 16, E-43007 Tarragona, Spain * Supporting Information ABSTRACT: A series of six exemplary cobalt-polyoxometalate (Co-POM) precatalysts have been examined to determine if they are molecular wateroxidation catalysts (WOCatalysts) or if, instead, they actually form heterogeneous, electrode-bound CoOx as the true WOCatalyst under electrochemically driven water-oxidation catalysis (WOCatalysis) conditions. Specifically, WOCatalysis derived from the following six Co-POMs has been examined at pH 5.8, 8.0, and 9.0: [Co4(H2O)2(PW9O34)2]10− (Co4P2W18), [Co 9(H 2O) 6(OH) 3(H PO4) 2(PW9 O 34) 3]16− ( Co9 P5 W27 ), [ ββCo4(H2O)2(P2W15O56)2]16− (Co4P4W30), [Co(H2O)PW11O39]5− (CoPW11), [α1-Co(H2O)P2W17O61]8− (α1-CoP2W17), and [α2-Co(H2O)P2W17O61]8− (α2-CoP2W17). The amount of Co(II)aq in 500 μM solutions of each Co-POM was measured after 3 h of aging as well as from t = 0 for pH = 5.8 and 8.0 by μM sensitive Co(II)aq-induced 31P NMR line broadening and at pH = 9.0 by cathodic stripping. The amount of detectable Co(II)aq after 3 h for the six Co-POMs ranges from ∼0.25 to ∼90% of the total cobalt initially present in the Co-POM. For 12 out of 18 total Co-POM and different pH cases, the amount Co(II)aq detected after 3 h forms heterogeneous CoOx able to account for ≥100% of the observed WOCatalysis activity. However, under 0.1 M NaPi, pH 5.8 conditions for CoPW11 and α1-CoP2W17 where ∼1.5% and 0.25% Co(II)aq is detectable, the measured Co(II)aq cannot account for the observed WOCatalysis. The implication is that these two Co-POMs are primarily molecular, Co-POM-based, WOCatalysts under electrochemically driven, pH 5.8, phosphate-buffer conditions. Even for the single most stable Co-POM, α1-CoP2W17, CoOx is still an estimated ∼76× faster WOCatalyst at pH = 5.8 and an estimated ∼740× faster WOCatalyst at pH = 8. ■ INTRODUCTION Meeting the growing global energy demand requires the development of new technologies and energy-storage schemes.1,2 Electrocatalytic water splitting is one widely discussed scheme for generating hydrogen as a renewable fuel.2 The bottleneck of the needed electrocatalytic water splitting is the anodic half reaction, catalytic water oxidation. As such, there has been a tremendous interest in, and resultant publication on, the development and screening of water oxidation catalysts (WOCatalysts) (a SciFinder search of “water oxidation” yields 6550 hits, while “water oxidation catalysis” yields 281 references since 2000 and as of March 2018).3−16 The identification of the kinetically dominant WOCatalyst-the primary focus of the present study-is directly relevant to the rational development of selective, active, and long-lived WOCatalysts. Polyoxometalates (POMs), in particular, cobalt-based polyoxometalates (Co-POMs), have attracted huge interest in the WOCatalysis area6,7,9−17 since Hill and co-workers initial report focussing on Co-POMs.9 POMs are discrete metal oxide compounds that can be readily synthesized on the gram to kilogram or larger scale via self-assembly. POMs are typically composed of high-valent (and therefore oxidatively stable) elements such as W(VI), P(V), Mo(VI), and V(V). Interest in POMs for WOCatalysis comes from the fact that POMs are known to incorporate redox-active metal centers such as cobalt and ruthenium, both of which are active toward 18−20 WOCatalysis. However, no known Co-POM is 100% hydrolytically stable over a wide range of pH values. The few Co-POMs that have had their Co(II) binding constants measured show that those Co(II) binding constants are in the micromolar range. 21,22 The micromolar amount of Co(II)aq that is leached when the CoPOMs are aged in buffered solutions can then deposit onto anodes during controlled potential WOCatalysis, in turn creating a well-known heterogeneous CoOx film8 as the active, electrochemically driven WOCatalyst. Such CoOx films have Scheme 1. List of Six Alternative Hypotheses for the Kinetically Dominant WOCatalyst under a Specific Set of Conditions been shown to account quantitatively for all of the observed electrocatalytically driven WOCatalysis current in the case of [Co4(H2O)2(PW9O34)2]10− (Co4P2W18) in 0.1 M sodium phosphate pH = 8.0 buffer and also for [Co4(H2O)2(VW9O34)2]10− (Co4V2W18) in 0.1 M sodium phosphate pH = 8.0 and 5.8 buffers, as well as 0.1 M sodium borate pH = 9.0 buffer.21,23−27 Our 2014 review entitled “Distinguishing Homogeneous from Heterogeneous Water Oxidation Catalysis When Beginning with Polyoxometalates” highlights the issues in, as well as preferred techniques for, distinguishing between homogeneous and heterogeneous WOCatalysis when beginning with POMs.25 The main findings of that review include that (i) multiple complementary methods are necessary en route to determining the Co-POM speciation, stability, and ultimately the identity of the true WOCatalyst;17,25,27 (ii) the amount of redox active metal such as Co(II)aq that is leached into solution (or present as a countercation impurity, as discovered herein) needs to be determined quantitatively; (iii) one needs to perform control experiments examining authentic heterogeneous CoOx self-assembled from Co(II)aq under the catalytic reaction conditions; (iv) the contribution to catalysis of heterogeneous CoOx or other metal oxides must then be quantified; and overall, (v) the stability of each POM is dependent upon the unique POM structure, the structural metals (e.g., W, Mo), the heteroatoms (e.g., P, Si, others), the redox-active metal (e.g., Co, Ru), and the reaction conditions, notably the pH, buffer type, and buffer concentration. Additionally, the true WOCatalyst is often dependent on the method of oxidation (e.g., chemical, photochemical, or electrochemical). Unfortunately, of the many studies using Co-POMs or other M-POMs (M = catalytically active metal) employed as water oxidation precatalysts, very few publications conduct the necessary experiments to provide compelling evidence for or against homogeneous molecular vs heterogeneous metal oxide WOCatalysis. There are important exceptions,7,13,17,23,27,28 that are discussed where relevant in the sections that follow. Other studies that use POMs for WOCatalysis, but which are not specifically treated in the main text of the present contribution, are summarized for the interested reader in Table S1 of the Supporting Information. In short, if and when Co-POMs can serve as molecular WOCatalysts has remained controversial. [Co4(H2O)2 (PW9O34)2]10−: A Prototype Co-POM WOPrecatalyst. The early prototype of a Co-POM WOCatalysis precatalyst system is [Co4(H2O)2 (PW9O34)2]10− (Co4P2W18) in 0.1 M sodium phosphate pH = 8.0.9,23,24 Previous work has shown that, after 3 h of aging in 0.1 M NaPi solution, 500 μM [Co4(H2O)2(PW9O34)2]10− dissociates a mere 58 μM Co(II) corresponding to just 4.3% decomposition (assuming the loss of a single Co(II) from the parent Co-POM).23 That 58 μM Co(II)aq forms a highly catalytically active heterogeneous CoOx films on tin-doped indium oxide (ITO) or glassy carbon electrodes under constant potential electrolysis. 23 The resultant CoOx film accounts for 100 ± 12% of the WOCatalysis current under the 0.1 M NaPi buffer and electrochemically driven WOCatalysis conditions.23 However, and in experiments designed to deliberately favor molecular WOCatalysis by Co 4 P 2 W 18 , when 2.5 μM [Co4(H2O)2 (PW9O34)2]10− is dissolved in NaPi pH 8.0 or 5.8 with ≥600 mV overpotential, the detected amount of Co(II) cannot account for the observed WOCatalysis current under the stated conditions-evidence that CoOx is not the dominant catalyst under those only modestly different conditions.24 The now classic Co4P2W18 system is a good example of how seemingly small changes in conditions can alter the kinetically dominant form of the Co-POM-derived WOCatalyst. A second important example of a system where the formation of CoOx from a Co-POM has been carefully examined is a 2012 Inorg. Chem. publication13 in which the CoPOM [ Co9(H 2O) 6(OH) 3(HPO 4) 2(PW9O34) 3] 16− (Co9P5W27) was shown to form CoOx under controlled potential electrolysis.13 Addition of bipyridine to starting solutions of Co9P5W27 chelates leached Co(II)aq and prevents the formation of CoOx under electrocatalytic conditions.13 WOCatalysis current was still observed in the presence of bipyridine, consistent with molecular Co9P5W27 being a true, electrochemically driven, homogeneous WOCatalyst, albeit one with only ∼2% of the WOCatalysis current of CoOx formed in the absence of bipyridine.13 This is another, important conclusion from prior studies: when molecular WOCatalysis from Co-POMs is seen, that activity (at least to date) is often only 1/2−1/11th that of the activity of CoOx examined under identical conditions.23−27 Identifying the kinetically dominant WOCatalyst from a molecular precatalyst is often difficult,9,13,23−27 especially in cases where as much as >95−99% of the initial POM remains intact under the reaction conditions. Only the scientific method of multiple alternative hypotheses is able to provide Figure 1. Polyhedral representations of the structure of the Co-POMs: (a) Co4P2W18; (b) Co9P5W27; (c) Co4P4W30; (d) CoPW11; (e) α1CoP2W17; (f) α2-CoP2W17. Blue octahedra represent WO6, orange tetrahedra represent PO4, and red spheres are Co(II). The coordination site on the Co atoms typically binds H2O and is where WOCatalysis is generally postulated to occur if the Co-POMs are indeed homogeneous, molecular WOCatalysts. convincing, compelling evidence for the kinetically dominant, “true” WOCatalyst.25,29 Scheme 1 presents six alternative hypotheses for the true catalyst when beginning with molecular, M-POM precatalysts (M = metal such as Co, Ru). The first hypothesis is that the precatalyst remains intact and is a homogeneous WOCatalyst, as the evidence strongly s u p p o r t s f o r t h e R u 4 - P O M , Cs 1 0 [ R u 4 ( μ - O ) 4 ( μ OH)2(H2O)4(γ-SiW10O36)2].7 A second hypothesis is that there is insidious Co(II) (e.g., present as a countercation from the synthesis) which then forms heterogeneous CoOx as the dominant catalyst; Co4O4 cubanes being a case in point.28 A third hypothesis is that the precatalyst (Co-POM) is hydrolytically unstable and leaches Co(II)aq into solution which then forms heterogeneous CoOx as the WOCatalyst. Such leaching of Co(II)aq and then the formation of CoOx is observed for both Co4P2W18 and Co4V2W18, as already noted.23,27 A fourth alternative hypothesis is that electrodebound Co-POM serves as a direct precursor to CoOx on the electrode without yielding solution-deteactable Co(II)aq. A fifth, quite reasonable hypothesis is that a fragment of the original Co-POM, POM-stabilized CoOx nanoparticles, or perhaps some other presently unidentified species is actually the true catalyst. Lastly, it is always possible that more than one of the five hypotheses listed might be occurring simultaneously, as was the case with the formation of CoOx from Co9P5W27 where WOCatalysis activity is still observed when Co(II)aq is removed by chelation with bipyridine (vide supra).13 Focus of the Present Studies. The focus of the current study is to establish the stability, speciation, and kinetically dominant WOCatalysts from the six exemplary Co-POMs shown in Figure 1. Three buffer and pH 5.8, 8.0, and 9.0 conditions have been selected because they are the most common buffers in the Co-POM WOCatalysis literature, and because they examine a case favoring Co-POM stability (pH = 5.8) and a case at higher pH where the thermodynamics of water oxidation are favored in the same buffer (pH = 8.0). These exemplary Co-POMs allow examination of the observed WOCatalysis as a function of varied Co(II) coordination environments (e.g., single vs multiple redox centers) and as a function of different Co(II) binding sites. The six Co-POMs chosen for study are the prototype [Co4(H2O)2(PW9O34)2]10− (Co4P2W18) (because it is relatively well-studied9,17,23,24 and, therefore, serves as a benchmark system for controls and comparisons); [Co9 (H2 O) 6 (OH) 3 (HPO 4 )2(PW9 O 34 )3 ] 16− (Co9P5W27), which has been reported to exhibit homogeneous WOCatalysis under electrocatalytically driven conditions (vida supra) and which shows very interesting, high WOCatalysis activity (η = 189 mV at 1 mA/cm2) as an insoluble Ba2+ salt embedded within amorphous carbon paste;13,30−33 and [ββCo4(H2O)2(P2W15O56)2]16− (Co4P4W30), selected because its Co centers are isostructural with Co4P2W18, yet this Co-POM was previously reported, surprisingly, as not exhibiting WOCatalysis using Ru(bpy) 33+ as the oxidant9,34,35 even though its close congener, Co4P2W18, does.9 The final three of the six Co-POMs are single Co-containing [Co(H2O)PW11O39]5− (CoPW11), which has been shown to form CoOx under electrocatalytic conditions in pH 7 phosphate buffer solutions,9,36−40 yet is reported to not exhibit WOCatalysis activity using Ru(bpy) 33+ as the chemical oxidant;9 and [α1Co(H2O)P2W17O61]8− (α1-CoP2W17) and [α 2-Co(H2O)P2W17O61]8− (α2-CoP2W17), two isomeric, single-cobalt CoPOMs21,22,41,42 chosen because they have literature precedent43 as WOPrecatalysts and because they therefore allow insights into the role of different Co(II)-to-POM binding sites and structures on the resultant WOCatalysis and kinetically dominant WOCatalyst. Meriting mention here is that the dicobalt(IV)-μ-oxo dimer of α2-CoP2W17 [(α2- CoIVP2W17O61)2O]14− (formed from α2CoP2W17 using ozone as the oxidant and as an inner-sphere oxo transfer reagent) has been shown to generate O2 from water in ∼95% yield, according to eq 1.44 However, it is not currently known if [(α2- CoIVP2W17O61)2O]14− can form from [α2-CoII(H2O)P2W17O61]8− under electrochemical oxidation. If formation of the μ-oxo dimer did occur, then one might expect to observe homogeneous WOCatalysis from α2CoP2W17. [(α 2‐CoIV P2 W O61)2 O]14− + 3H 2O 17 → 2[α2‐Co II(OH 2 )P2 W O61]8− + O2 + 2H + 17 (1) Choice of Reaction Conditions and Key Experimental Methodologies. The conditions chosen to examine the CoPOMs in Figure 1 include sodium phosphate buffer (NaPi) at both pH 5.8 (favoring the stability of the Co-POMs) and 8.0 (favoring the thermodynamics of water oxidation). We also used sodium borate buffer (NaB) at pH 9.0 to compare the effect of buffer since Co-POMs tend to be more stable in NaB buffer17 and because NaPi can, at least in principle, drive the decomposition of Co-POMs due to the formation of insoluble Co 3(PO 4) 2 (K sp ≈ 10 −35).45 Similar to our previous publications, we aged the Co-POMs in each respective buffer for 3 h as a relatively minimal solution lifetime.23,27 Note that 3 h aging is at most a minimum test of the stability of the Co-POMs because any truly useful WOCatalyst will need to be active for perhaps 103−4 h or more of WOC (thereby for lifetimes that may approach an estimated >109 total turnovers),27 so that even if the turnover frequency was among the highest reported for a Co-POM (i.e., 200 s−1),16 then any molecular Co-POM WOCatalyst would still need to be active for >140 h-meaning that our 3 h test is only 2% of the required catalytic lifetime. However, and importantly, we also examine the amount of Co(II)aq formation at t ≈ 0 and as a function of time by 31P NMR in what follows. In order to quantify the amount of Co(II) that dissociates from the complexes, Co(II)aq-induced 31P NMR line broadening of the P atom in the phosphate buffer is used.27,28,46−48 Adsorptive cathodic stripping is then used in what follows as a secondary method to quantify the Co(II)aq in NaPi and the primary method to quantify the Co(II)aq leached from the CoPOMs in NaB.23,27,49 Once the stability of each Co-POM was established under a given set of conditions, controlled potential electrolysis was conducted at 1.1 V vs Ag/AgCl in the aged CoPOM solutions and the amount of O2 produced was compared with the amount of O2 produced from an equivalent amount of Co(II)aq that was detected. Note that 1.1 V was chosen because glassy carbon is known to oxidize to CO2 at ≥1.2 V vs Ag/AgCl.28 As such, each buffer condition has a different applied over potential of η = 410 mV at pH = 5.8, η = 540 mV at pH = 8.0, and η = 600 mV at pH = 9.0. Next, controlled potential electrolysis is conducted to allow film accumulation, followed by cyclic voltammetry in the original Co-POM solution. The electrodes were then removed and rinsed followed by cyclic voltammetry of the working electrode in a fresh, buffer-only solution, thereby obtaining the CV of any deposited film. The deposited films in what follows are also characterized by scanning electron microscopy (SEM) and Xray photoemission spectroscopy (XPS). The sum of these experiments is then used collectively to provide evidence for the kinetically dominant WOCatalyst under a stated set of conditions. Finally, a historical note is perhaps of some interest: we never started out to pursue the “what is the true catalyst?” question in the WOCatalysis area and despite our background with this question in the area of hydrogenation catalysis with low-valent metal nanoparticles.50 Instead, this key question quickly found us in the area of Co-POMs as WOPrecatalysts. Our first experiments used Co4P2W18 as a WOPrecatalyst in our OPV-driven WOC half-cell,51 the Co-POM Co4P2W18 being “close to our intellectual hearts” since we discovered the rational synthesis of and Co4P2W18, Co4P4W30, and the other members of this class of M4-containing POMs in 1981.52 The very f irst experiments with Co4P2W18 provided evidence that an electrode-bound catalyst, the same color as Nocera’s CoOx/Pi catalyst8 that we had been examining, had formed on the ITO anode f rom the Co4P2W18 precatalyst.23 The findings quickly followed that the Co4P2W18 POM leached Co(II) into solution from just 4.3% decomposition over 3 h, and that the resultant 58 μm Co(II) formed electrode-bound CoOx that accounted for 100 ± 12% of the observed, electrochemically driven, WOCatalysis current.23 A similar situation occurred for the V-based congener, Co4V2W18: we were intrigued by the claim of 100% hydrolytic stability, and 200-fold higher catalytic activity compared to the P-congener.16 Yet when we prepared Co4V2W18 by the literature route and tried to purify it to the 51 V NMR resonance assigned in the literature to Co4V2W18, the resultant, different color POM contained only ∼1 Co per V2W18O6818− unit-yet had the same 51V NMR resonance ascribed to “Co4V2W18”.53 The 100% hydrolytic instability of Co4V2W18, its decomposition to Co(II) and formation of electrode-bound CoOx/Pi that carries 100% of the observed WOC within experimental error, as well as assignment of the observed 51V NMR resonance to the impurity V2W4O 6− 19 followed after considerable effort.27,53 In short, the “what is the true catalyst?” question raised its omnipresent head each and every time we tried to build off the literature of Co-POMs as WO(Pre)Catalysts. That observation is, actually, not surprising, at least in hindsight: the identity of the true catalyst in any and all catalytic reactions is an important, often overlooked, typically challenging, critical question in catalysis. Reflection makes the latter claim obvious once one realizes that all catalytic properties of interest derive from the precise composition and nature of the actual catalyst, including the: catalytic activity, selectivity, lifetime, poisoning, reisolation, and catalyst regeneration. The “what is the catalyst?” question, and the associated “Is it homogeneous or heterogeneous catalysis?” question, had not been fully raised nor critically addressed for cobalt or other POM-based WOCatalysts before our 2011 study that has (as of July 2018) over 246 citations.23 The present work brings to completion our studies of the kinetically dominant, “true” catalyst(s) derived from exemplary Co-POMs in buffer solutions under electrochemically driven and the other stated, specific WOCatalysis conditions-conditions that matter greatly, vide infra. It is hoped that the WOCatalysis community can use methods and approach herein to provide evidence for the kinetically dominant WOCatalyst as a critical part of their own WOCatalysis studies. ■ EXPERIMENTAL SECTION General Considerations. All reagents used were the highest purity available and were used without further purification. Water (18 MΩ) was obtained from an in house Barnstead Nanopure filtration system. FT-IR were collected using a ThermoScientific Nicolet iS50 FT-IR spectrometer in transmission mode using KBr pellets containing approximately 2 wt % of the analyte. Thermogravimetric analysis (TGA) was performed using a TA Instruments TGA 2950 with a 5 °C/min ramp rate to 500 °C on a platinum sample pan. TGA was used to determine the waters of hydration because water is the only volatile component of the Co-POMs at ≤500 °C. 31P NMR was collected using either an Agilent (Varian) 400 MHz NMR or an Agilent Inova 500 MHz NMR; the spectral ranges and pulse sequences were optimized for the resonance of the 31P atom of interest. Elemental analyses were obtained from Galbraith Laboratories in Knoxville, TN. Electrochemically driven WOCatalysis experiments were conducted in 0.1 M sodium phosphate buffer (NaPi) either at pH 5.8 or 8.0 or in 0.1 M sodium borate (NaB) pH 9.0.54 All stability, electrochemistry, and WOCatalysis experiments were conducted with a 500 μM Co-POM concentration, chosen because the stability of the complexes can be difficult to quantify, and hence, employing this higher, 500 μM concentration facilitates detection of decomposition byproducts by 31P NMR, for example (vide infra). All of the electrochemistry was performed using a CH Instruments CHI630D with a three-electrode setup. All potentials are referenced to Ag/AgCl, with a platinum wire as the counter electrode and glassy carbon either 1.0 or 0.071 cm2 as the working electrode. SEM was conducted on a JEOL JSM-6500F microscope with magnification from 1000 to 20000. XPS was conducted on a PE-5800 X-ray photoelectron spectrometer; full scans were collected on deposited films as well as high-resolution scans for individual elements. Syntheses of the Co-POMs were conducted according to literature methods and characterized via FT-IR, 31P NMR, TGA, and elemental analysis. The procedures followed and resulting characterization data are presented in the Supporting Information for the interested reader (Figures S1−S8).9,13,21,23,30,31,34−43,52 Characterization of the CoPOMs was consistent with prior literature and are isomerically pure samples, with the exception of K8[α1-Co(H2O)P2W17O61], which contains a presently inseparable 5% impurity of the isomeric K8[α2Co(H2O)P2W17O61]. Stability of the Co-POMs in Buffered Solutions. Stability of the CoPOMs Determined by Co(II)-Induced 31 P NMR Line Broadening. The well-established method of Co(II)aq-induced 31P NMR line broadening of the sodium phosphate buffer, first observed by Klanberg and Dodgen46 and used later by Nocera and others to quantify aqueous Co(II) leached out of CoOx film or molecular Co complexes,27,28,47,48 was used to detect the amount of Co(II)aq present in NaPi-buffered solutions for each Co-POM. This 31P NMR technique is powerful because it is selective toward Co(II)aq (i.e., and insensitive to Co(II) within a Co-POM) while also having a detection limit of ∼2 μM Co(II)aq.26 Further precedent for this 31P NMR methodology is its recent use to quantify the amount of Co(II) leached from [Co4V2W18O68]10− as well as [Co4P2W18O68]10−, results which demonstrate that the 31P method agrees with cathodic stripping determinations of Co(II) to within ±5% for both [Co4V2W18O68]10− and [Co4P2W18O68]10− in 0.1 M NaPi pH = 8.0.27 We followed the same general procedure outlined in our 2017 paper27 for the 31P NMR determinations of Co(II)aq, except the CoPOM concentrations employed herein are 500 μM. (The lower concentration of 5 μM Co-POM used in our 2016 paper was chosen because [Co4V2W18O68]10− decomposes 100%, resulting in Co(II)aq concentrations too high to measure reliably at more than 5 μM of that particular Co-POM). First, a calibration curve was developed using Co(NO3)2 as an authentic source of Co 2+aq for the line broadening experiments in both pH 5.8 and 8.0 NaPi (as 100 mM solutions in 25% D2O, Figure S9 in the Supporting Information). Next, the appropriate amount of Co-POM was weighed in a 1 dram vial. To prepare 2 mL of a 500 μM solution, 1 μmol of each POM is required; therefore the following masses of each indicated Co-POM were used: Co9P5W27, 8.97 mg; Co4P4W30, 8.77 mg; CoPW11, 3.20 mg; α1CoP2W17, 4.86 mg; α2-CoP2W17, 4.82 mg. Next, 1 mL of 200 mM NaPi (pH 5.8 or 8.0), 500 μL D2O, and 500 μL water were added to the Co-POM powder in the 1 dram vial, yielding 2 mL of a solution with 500 μM Co-POM, 100 mM NaPi, and 25% D2O. The timer was started immediately upon addition of the buffer solution to the solid Co-POM. A 1 mL aliquot was then transferred into a 5 mm OD NMR tube which was then inserted into the NMR. 31P NMR was then collected on the sample without shimming and under conditions identical to those used for the calibration curve. A 500 MHz Varian NMR spectrometer was used at 25 °C with scans from +64.9 to −64.9 ppm, a 45° pulse angle, a 1.000 s relaxation delay, and a 0.624 s acquisition time. The peak width of the 31P NMR peaks were determined using the instrument’s VNMRJ software after phase correction. To confirm the line broadening is caused almost completely by Co(II)aq and not by the Co(II) present within the intact Co-POM, we conducted the same experiments as above except in the presence of 92 μM EDTA to complex any free Co(II) (i.e., an amount of EDTA in 1.2−10-fold excess of the Co(II)aq detected by the initial 31P NMR experiment for Co4P4W30 and α1-CoP2W17 for example). Any residual line broadening over that original 31P NMR was then assigned to the intact Co-POM, an amount that ranged from just 2 to 8 Hz, so only between 1.3 and 6 μM Co(II)aq for Co4P2W18 and Co4P4W30 in 0.1 M NaPi pH = 8.0. This in turn means that the contribution from the intact Co-POMs to the observed 31P NMR line broadening is at most only 8% of the Co(II)aq detected for Co4P4W30 in 0.1 M NaPi pH = 8.0. The residual line broadening from the added EDTA experiment was subtracted from the raw fwhm values for the particular Co-POM being examined before the fwhm values were fit to the calibration curve to calculate the final Co(II)aq concentration. For the Co(II)aq values that were close to the detection limit of the initial calibration curve (e.g., α1-CoP2W17 at pH = 5.8 where the detected Co(II)aq was 2.9 ± 3 μM) an additional calibration curve was generated that was able to more precisely determine the Co(II)aq and with a lower <0.5 μM detection limit (Figure S9 of the Supporting Information). Stability of the Co-POMs As Determined by Cathodic Adsorptive Stripping As a Second Technique. The reliability of the 31P NMR technique for the quantitation of Co(II)aq has been demonstrated for both Co4P 2W18 and Co4V2W 18 . 23,27 However, we wanted to determine the amount of Co(II)aq present in the 500 μM Co-POM solutions in pH 9.0 NaB after 3 h of aging (i.e., and under conditions where no Pi is available for the use of the 31P NMR method). Therefore, and as before,23,27 an adsorptive cathodic stripping method was employed that quantifies Co(II)aq by adsorption of the neutral cobalt dimethylglyoxime (DMG) complex on a bismuth electrode and subsequent reductive stripping.23,27,49 Electrode Preparation. The Bi film electrode was prepared using a method adapted from previous studies.23,27,49 First, a clean glassy carbon electrode (3 mm diameter), a Ag/AgCl reference electrode, and a Pt wire counter electrode were placed into an aqueous solution containing 0.02 M Bi(NO3)3, 0.5 M LiBr, and 1 M HCl. Then constant potential electrolysis was conducted at −0.25 V until 10 mC of charge had accumulated (∼45 s). The electrodes were then removed and rinsed gently with water prior to being placed into the analyte solution containing either Co(NO3)2 for the calibration curve or the aged Co-POM solutions. Calibration Curve. A calibration curve was developed using Co(NO3)2 as an authentic source of Co(II)aq, with concentrations ranging from 1.0 to 50 μM Co(II)aq in NaPi pH 8.0 and NaB pH 9.0 (Figure S10 in the Supporting Information). Using freshly plated Bi films, the electrodes were placed into a 1 dram vial containing a buffered solution (either 0.1 M NaPi pH 8.0 or 0.1 M NaB pH 9.0) that contained the desired Co(NO3)2 concentration and 100 μM DMG. The solution was stirred for 3 s and allowed to reach stillness, and then the CoDMG2 was adsorbed by applying −1.3 V to the Bi film electrode for 15 s. The solution was again stirred for 3 s and allowed to settle before differential pulse voltammetry (DPV) from −0.7 to −1.3 V using a 0.1 s pulse width, 50 mV amplitude, and a 0.0167 s sampling width. The height of the DPV waves were measured from the background using the CH Instruments software, and plotted against the known Co(II)aq concentration for the calibration curves (Figure S10 in the Supporting Information). Worth noting is that the use of pH = 8.0 to 9.0 buffer is essential because at pH = 5.8 the adsorptive cathodic stripping is not responsive to the Co(II)aq concentration, likely because the DMG must be deprotonated by pH > 5.8 to form Co(DMG)2 that is an intermediate in the Co-stripping on the Bi film. Aging of the Co-POMs and Cathodic Stripping. First, 500 μM solutions of the Co-POMs were prepared by weighing an appropriate amount of the solid Co-POM material into a 1 dram vial and then adding 2 mL of either 0.1 M NaPi pH 8.0 or NaB pH 9.0. The solutions were then aged 3 h before an aliquot, typically 200 μL, was used in the same analyte solution as the calibration curve. (While as noted the aliquot was typically 200 μL, the actual microliter volume of the aliquot was adjusted such that the detected Co(II) aq concentration was within the calibration curve’s linear range of 1− 10 μM, as explained in greater detail below.) Because DMG binding of Co(II) could, in principle, shift the Co-POM dissociative equilibrium yielding a larger Co(II)aq concentration than without DMG, the time between aliquot addition and cathodic stripping was kept to a minimum (<1 min). The Co(DMG)2 deposition and the DPV were conducted in the same manner as for the calibration curve above. The peak height of the DPV was fit to the calibration curves (Figure S10 of the Supporting Information), and the results were used to calculate the Co(II)aq concentration in the analyte solutions. The Co(II)aq concentration in the original solution was determined by taking into account the 1:10 dilution from the original solution to the analyte solution. For cases where the measured Co(II)aq was not within the linear range of the calibration curve, the dilution factor from the original to the analyte solution was adjusted so that the detected Co(II)aq concentration was within the range of the linear calibration curve. For example, the Co(II)aq detected from a 1:10 dilution of CoPW11 is ≫10 μM and therefore outside the linear range of the calibration curve. Instead, a 20 μL aliquot of the aged CoPW11 was used (a 1:100 dilution) and the Co(II)aq concentration in the diluted solution was determined to be 4.4 ± 0.5 μM, meaning that the actual Co(II)aq concentration in the original, undiluted CoPW11 solution was 100-fold larger, specifically 440 ± 50 μM. Electrocatalytically Driven Water Oxidation Catalysis Beginning with the Co-POMs. Electrolysis Using the Co-POMs in Buffered Solutions in Comparison with Co(II)aq. From the 31P NMR and cathodic stripping studies, the amount of Co(II)aq that dissociates into buffered solution after 3 h is known. Comparing the observed activity of the aged Co-POM solutions with solutions containing authentic Co(II)aq tests if the WOCatalysis activity can be accounted for by the dissociated Co(II)aq or, alternatively, if WOCatalysis by the Co-POM itself is indicated. Hence, we conducted bulk electrolysis using a 1 cm2 working electrode in buffered solutions that either contained a 500 μM Co-POM solution that had aged 3 h or an amount of authentic Co(II)aq that matched the measured Co(II)aq after 3 h, as determined by 31P NMR or cathodic stripping. Electrolysis was conducted in the same manner as previous studies using Co4V2W18 as a WOPrecatalyst.27 Briefly, the experiments were conducted in a custom built U-cell with a medium fritted glass filter separating the working and counter electrodes. The working compartment was sealed using a Teflon lid pierced to accommodate the working electrode, the reference electrode, and the O2 detection sensor (NeoFox; FOSPOR-R probe), all in a 6 mL, argon-purged solution. The headspace was kept to a minimum in order to diminish O2 equilibration with the headspace which otherwise results in an excessively systematically low measured O2 concentration and low apparent faradaic efficiency. The O2 sensor was calibrated using a 2point calibration curve consisting of air-saturated DI water (∼220 μM at 22 °C, for a typical barometric pressure of 0.84 atm for Fort Collins, CO), and O2-free solutions were generated by addition of excess sodium sulfite to the solution. Electrolysis was conducted at 1.1 V for 5 min with stirring at ∼600 rpm. The final faradaic efficiency was determined by comparing the final O2 concentration to the O2 concentration expected from the total charge passed during the experiment (i.e., 4 e− passed per 1O2 produced). Electrochemical and Morphological Characterization of the Films Electrodeposited from the Co-POM Solutions. Deposition and Cyclic Voltammograms of CoOx Films. Previous work has documented the value of controlled potential electrolysis and subsequent analysis of deposited films from Co-POMs.23,27 As such, controls were conducted in a similar manner in which constant potential electrolysis was conducted at 1.1 V on a glassy carbon electrode for 5−30 min to allow sufficient accumulation of an electrodeposited film to be visible to the naked eye. After electrolysis, cyclic voltammetry was conducted on the film in the same Co-POM solution. The electrodes were subsequently removed from the original Co-POM solution, rinsed with water, and placed into a buffer-only solution. Cyclic voltammetry was then conducted on the electrodeposited film in the buffer-only solution-this allows comparison of the observed WOCatalysis activity from the deposited film to that of the starting Co-POM solution. Electrolysis was then conducted on the deposited film in the buffer-only solution under otherwise identical conditions to the Co-POM solution. To test the hypothesis that CoOx forms from Co(II)aq, and not directly from Co-POM bound to the electrode surface, EDTA was added at a concentration 10 times the measured Co(II)aq. Constant potential electrolysis at 1.1 V was then conducted. Controls with Co(NO3)2 and EDTA present demonstrate that no film is deposited from the Co·EDTA complex. This, in turn, means that if a film is observed from any Co-POM solution containing 10 equiv of EDTA/ Co(II)aq, then that film would have to be formed from some route not involving freely diffusing Co(II)aq, for example, conceivably directly from Co-POM adsorbed on the electrode. Morphological and Compositional Analysis of the Deposited Films. The electrodeposited films were examined by XPS and SEM to quantify elements in the surface of the film, and to capture morphological features, respectively. The films were deposited on glassy carbon (1 cm2) at 1.1 V for 30 min from Co-POM solutions in 0.1 M NaPi pH 5.8 and 8.0 as well as 0.1 M NaB pH 9.0. The electrodes were then removed from solution and allowed to air-dry on the bench before being placed into a desiccator overnight. XPS was conducted on a PE-5800 X-ray photoelectron spectrometer; survey scans were collected from 10 to 1100 eV with 1.6 eV/step and 187.85 eV pass energy. High resolution scans were collected for each element detected from the survey (such that sufficient background was included with 0.1 eV/step and 23.5 eV pass energy). SEM was conducted on a JEOL JSM 6500F scanning electron microscope. Images were collected from 1000× to 20000× magnification to demonstrate the homogeneity of the film as well as to visualize morphological details. ■ RESULTS AND DISCUSSION Stability of the Co-POMs Assayed by Co(II)aq-Induced P NMR Line Broadening. Quantitative knowledge of the stability of any precatalyst under a given set of conditions is crucial to understanding the kinetically dominant, most active form of the catalyst.23,25−27 Using the Co(II)aq-induced, 31P NMR line-broadening experiments first developed by Klanberg and Dodgen46 and then Nocera and co-workers,28,47 the amount of Co(II)aq present as a function of time for each CoPOM was measured in NaPi pH 5.8 and 8.0. The Co(II)aq vs time traces for selected Co-POMs are shown in Figure 2 and 31 Figure 2. Co(II)aq concentration vs time determined by Co(II)aq induced line broadening in 0.1 M NaPi (pH 5.8, red and pH 8.0, blue) for 500 μM solutions of (a) Co4P2W18 (adapted with permission from ref 27, Copyright 2017 American Chemical Society); (b) Co9P5W27; and (c) Co4P4W30. The value for each Co(II)aq concentration was determined by fitting the observed 31P NMR line widths of the NaPi to the calibration curve generated with authentic Co(NO3)2. The percent of total cobalt refers to the percent of cobalt that is detected in solution compared to the total Co(II) present initially in the specific Co-POM. Error bars are the standard deviation from three repeat experiments. The lines between points have been added to guide the eye and, hence, are not fits to any specific equation. The Co(II)aq vs time plots for the other Co-POMs are shown in Figure S11 of the Supporting Information. Figure S11 of the Supporting Information. The percent of total Co(II) in the Co-POM solution that is present as aqueous Co(II)aq after 3 h of aging is presented in Figure 3 and Table 1. All of the Co-POMs examined showed some detectable Co(II)aq over 3 h in NaPi buf fer ranging from ∼0.25 to 50% (in NaPi buffer) of the total Co(II) present in the given Co-POM solution, the exact percentage depending on the Co-POM and the precise pH and buffering conditions, vide infra. Three of the Co-POMs examined, specifically Co4P2W18, Co9P5W27, and CoPW11, show increasing concentration of Co(II) leached into solution over 3 h at pH = 8.0 and 5.8, Figure 2 and Figure S11 of the Supporting Information. For these cases, the detected, increasing Co(II)aq is most simply attributed to (continued) dissociation of Co(II) from the CoPOM precatalyst. One interesting point to note is that while Co4P2W18 is more stable at pH = 5.8, Co9P5W27 is more stable at pH = 8.0. This is consistent with the fact that a mixture of Co4P2W18 and Co9P5W27 is obtained from reactions of HPO42−, Co(II), and WO42−,30 with Co9P5W27 being more prevalent at the more basic pH > 7.31 Restated, this evidence suggests unsurprisingly that individual Co-POMs tend to be more stable in the pH range where they are synthesized. Leaching of Co(II)aq from the complex is consistent with hypothesis #3 from Scheme 1 for the above three Co-POMs. The other three Co-POMs, Co4P4W30, α1-CoP2W17, and α2-CoP2W17, show detectable, 0.25(±0.06)−3.9(±0.1)% but relatively flat Co(II)aq over 3 h at pH 5.8 and 8.0 (with the exception of α2-CoP2W17 at pH 8.0, vide infra). Note that all of the Co-POMs have non-zero amounts of Co(II)aq detected that are well above the detection limit (∼2 μM generally, but ∼0.5 μM for α1-CoP2W17 at pH = 5.8 using our third, more precise, lowest [Co(II)aq] calibration curve described in the Experimental section and Figure S9 of the Supporting Information, which focuses on the lower concentrations of 0.5, 1, 5, and 20 μM Co(II)aq). Three repetitions of each of these lower [Co(II)aq] were conducted using α1-CoP2W17 at pH = 5.8, with the key result that the detected amount of Co(II)aq for α1-CoP2W17 at pH = 5.8 is 1.2 ± 0.3 μM. In short, the detected Co(II)aq for α1-CoP2W17 at pH = 5.8 is also experimentally non-zero, as well as relatively flat. A flat Co(II)aq vs time dependence implies either: (i) that rapid Co(II)aq dissociation from the Co-POM to reach equilibrium quickly has occurred, or (ii) that the Co(II)aq is present as a countercation to the Co-POM from the synthesis (or, conceivably (iii) a combination of (i) and (ii)). If the Co(II)aq is, in fact, present as a countercation, then one might expect to observe a high Co(II) weight percent (wt %) in the elemental analysis. As a specific example, the weight percent of Co by elemental analysis for Na 16 [ββ-Co 4 (H 2 O) 2 (P 2 W 15 O 56 ) 2 ]·39H 2 O (Co4P4W30) is the same within experimental error, with a found Co wt % of 2.62% vs the expected 2.69%. Furthermore, the molar amount of Co(II) present in the Co4P4W30 solutions (14−16% mol of Co(II)/mol Co-POM) is not distinguishable if one assumes an error of ±0.4 absolute wt %. Indeed, the expected wt % cobalt would change from 2.69% for the ele mental for mula of the pure Na16ββ [Co4(H2O)2(P2W15O56)]·39H2O to 2.81% for the hypothetical case where 16 mol % of Co(II)/ Co4P4W30 as a countercation was present for a (hypothetical) elemental formula of Na15.68Co0.16[ββ-Co4(H2O)2(P2W15O56)2]·39H2O, a difference of only 0.11 wt %. In short, a publishable (±0.4% Figure 3. Percent of total cobalt that is present as Co(II)aq after 3 h of aging in solution for 500 μM solutions of each Co-POM in 0.1 M NaPi, pH = 5.8 (red) and pH = 8.0 (blue), as well as in 0.1 M NaB pH = 9.0 (gray). Decomposition data for Co4V2W18 has been adapted with permission from ref 26 (Copyright 2017 American Chemical Society) for comparison with the other Co-POMs, albeit with a 5 μM Co-POM concentration under otherwise identical conditions. The lower concentration of Co4V2W18 had to be used because Co4V2W18 is so unstable that, at 500 μM, the Co(II)aq detected is otherwise above the linear range of the calibration curve. absolute wt %) elemental analysis is not sufficient evidence to disprove Co(II) impurities as counter cations present in Co4P4W30 nor, by analogy, more generally in other Co-POMs. To provide evidence for or against Co(II)aq being present as a countercation vs the rapid dissociation of Co(II) from Co4P4W30 to an equilibrium value, we conducted 31P NMR control experiments by adding 1 equiv of EDTA/Co(II)aq to the Co4P4W30 solutions and, then, repeated the 31P NMR linebroadening experiment, Figure 4. The results of that experiment show that addition of 1 equiv of EDTA/Co(II)aq lowers-but does not remove all-of the detected Co(II)aq (black dashed line, Figure 4). Furthermore, an important observation is that the Co(II)aq concentration does not immediately return to the higher, 60−80 μM value, thereby ostensibly ruling out a fast, initial release of Co(II)aq to reach an equilibrium level at either pH of 8.0 or 5.8. Addition of an excess, 100 μM amount of EDTA does remove all of the observed Co(II)aq, which then remains at zero and hence constant within experimental error over the 3 h experiment (black solid line, Figure 4). In short, the data suggest that the Co(II)aq being detected is present initially at a countercation attached tightly to the highly negatively charged, [ββCo4(H2O)2(P2W15O56)2]16− polyoxopolyanion and, therefore, not available to contribute to the phosphate line broadening to any great extent. Such tight-ion pairing between a dicationic Co(II ) 2+ and t he 16 minu s PO M, [ ββ Co4(H2O)2(P2W15O56)2]16−, even in water is not unreasonable nor unexpected. The evidence provided above demonstrates that there is an EDTA-removable amount of additional 31P NMR line broadening in the Co4P4W30 system, consistent with an additional amount of tight ion paired Co(II) attached to the [ββ-Co4(H2O)2(P2W15O56)2]16−. It is therefore reasonable to sum the observed Co(II)aq in the absence of EDTA with the observed Co(II)aq seen upon the addition of 1 equiv of EDTA to give the total apparent Co(II)aq as shown in Figure 4. Specifically, one can calculate that in pH = 5.8 buffer, the total Co(II)aq value = 62(±1) + 19(±2) = 81(±2) μM (i.e., the solid red line plus the dashed black line yields the dashed red line in Figure 4), while in pH = 8.0 the total Co(II)aq= 78(±2) + 10(±3) = 88(±4) (i.e., the solid blue line plus the dashed black line yields the dashed blue line in Figure 4). Averaging the pH 5.8 and 8.0 data yields a Co(II)aq value of 85(±4) μM as an estimate of the amount of Co(II)aq present as a countercation from the synthesis in Co4P4W30. The systematic difference of the measured Co(II)aq in pH 5.8 vs 8.0 of 62(±1) vs 78(±2) μM, respectively, is discussed in the Supporting Information for the interested reader. The observation of Co(II) as a countercation is an important finding for at least two reasons, the first of which is because it provides evidence for hypothesis #2 from Scheme 1, where Co(II) is present as a normally undetected impurity in the postsynthesis Co4P4W30.28 Second, the results in Figure 4 are significant as they imply that the presence of dication impurities in the syntheses of highly charged POMs is very likely a little recognized, but more general, phenomenon in polyoxometalate and other polyanionic self-assembly syntheses. Figure 4. Plots of Co(II)aq concentration vs time for a 500 μM solution of Co4P4W30 in 0.1 M NaPi (pH 5.8, left and pH 8.0, right). The red and blue lines are for Co4P4W30 in the absence of any added EDTA (i.e., the same as Figure 2), the dashed black lines are for experiments where 1 equiv of EDTA/Co(II)aq has been added (60 and 80 μM for pH 5.8 and 8.0 ,respectively), and solid black lines represent the addition of excess EDTA (100 μM). The dashed red and blue lines represent the true Co(II)aq concentration (i.e., the sum of the solid colored line with the dashed black line for each pH condition. Because of the intrinsically high molecular weight of large POM anions, low levels of countercation impurities are difficult to detect via standard elemental analysis methods such as ICP-OES (vide supra). This highlights the power of the Co(II)aq-induced 31P NMR line-broadening technique because it has high selectivity toward Co(II)aq with a detection limit of ∼2 μM Co(II)aq, which in turn corresponds to ∼0.4 mol % regardless of the molar mass of the Co-POM. Future research using Co-POMs for WOCatalysis should use 31P NMR line broadening to quantify Co(II)aq because it is likely present in at least some as-synthesized Co-POMs. However, the Co(II)aq-induced 31P NMR methods herein can now be used on Co-POMs that are, for example, not run down ionexchange columns or not exposed to multiple recrystallizations from, say, Na+, K+, or other desired cation-containing recrystallization solutions. 31 P NMR Line-Broadening Data for the Relatively Stable Co-POMs, α1-CoP2W17, and α2-CoP2W17. For the case of α1-CoP2W17 and α2-CoP2W17 at pH = 5.8 and 8.0 and because these Co-POMs appear to be relatively “stable” in initial Co(II)aq detection experiments, we conducted 31P NMR experiments over a longer time scale, 7−10 h, Figure S12 of the Supporting Information. These longer time scale experiments show that at pH = 5.8 little change in the Co(II)aq beyond experimental error is observed. Addition of excess EDTA (92 μM) to α1-CoP2W17 and α2-CoP2W17 at pH = 5.8 returns the 31P NMR line width of NaPi to its natural width of ∼2 Hz. Overall, the results teach that α1-CoP2W17 and α2CoP2W17 contain from ∼0.25% to ∼1.5% of their Co(II) in solution, a level of Co(II) that could readily be explained by either a low level of Co(II) countercation impurity or Co(II) leaching (or a combination of these two). The bottom line is clear, however: detecting Co(II) that leads to CoOx or other possible catalyst species derived from the parent Co-POM is a ≤ μM detection problem. As for the pH = 8.0 experiments, observing the Co(II)aq concentration from α1-CoP2W17 over longer time-scales (10 h) at pH = 8.0 demonstrates that the Co(II)aq concentration increases at a slow rate without plateauing-even after 10 h. This indicates that α1-CoP2W17 is unstable at pH = 8.0 and dissociating Co(II)aq, Figure S12. Intriguingly, the Co(II)aq concentration from α2-CoP2W17 actually decreases over time in the pH 8.0 solution (Figures S11 and S12 of the Supporting Information). Possible explanations for this interesting observation, notably the possible consumption of Co(II) by the conceivable formation of Co4P4W30, are discussed in the Supporting Information for the interested reader.55,56 To summarize the Co(II)aq-induced 31P NMR line-broadening experiments, all of the Co-POMs examined show nonzero detectable amounts of Co(II)aq under the buf fer conditions specif ied. The amount of Co(II)aq released into solution ranges from ∼0.25% to 50% of the total cobalt in 0.1 M NaPi pH = 5.8 and 8.0. Furthermore, due to the large molecular mass of the Co-POMs, cobalt elemental analysis is insufficient to quantify Co(II) present as a countercation and at the low levels that can matter for WOCatalysis by electrode bound and formed CoOx. However, Co(II)aq -induced line broadening of the 31P NMR peak of NaPi is a much more useful, powerful, and relatively direct technique to quantify the amount of Co(II)aq either leached into solution, or present initially as a Co(II) counterion impurity from syntheses employing Co(II). Stability of the Co-POMs-cathodic Stripping. Because 11B is a quadrupolar nucleus with relative receptivity of 0.165 compared to 1H, and perhaps also because borate buffer has a complex speciation (especially near its pKa, with at least 5 boron species being present),54 Co(II)aq-induced 11B NMR line broadening is unknown at present. Hence, to measure the amount of Co(II)aq that leaches from the Co-POMs after 3 h of aging in 0.1 M NaB pH 9.0 buffer, cathodic stripping was employed as the most convenient, sensitive, and selective method presently available for the NaB buffer systems. The results of the cathodic stripping studies are summarized in Figure 3 and Table 1. The amount of Co(II)aq detected for the six prototype Co-POMs by 31P NMR at pH 5.8 and 8.0 are also summarized in Table 1 for comparison. The amount of Co(II)aq detected by cathodic stripping for the 0.1 M NaPi pH = 8.0 conditions proved to be the same within experimental error to the Co(II)aq detected by 31P NMR (the error bars are much larger for cathodic stripping, that method often Table 1. Comparison of the Leached Co(II)aq (μM) after 3 h of Solution Aging from 500 μM Co-POM Solutions under the Three Buffer Conditions (Values Shown in Bold in Parentheses are the Percent of Cobalt That Has Dissociated from the Co-POM Compared to the Total Cobalt Present Initially in the Co-POM)a [Co(II)aq] by 31P NMR, μM (% Co(II) after 3 h) polyoxometalate Co4P2W18 Co9P5W27 Co4P4W30 CoPW11 α1-CoP2W17 α2-CoP2W17 0.1 M NaPi pH 5.8 [data range] 11 ± 3 (0.5 ± 0.2%) [8−15] 75 ± 2 (1.7 ± 0.1%) [73−77] 62 ± 3 (3.1 ± 0.4%) [59−66] 6± 3 (1.3 ± 0.6%) [3−9] 1.2 ± 0.3 (0.25 ± 0.06%)b [0.8−1.4]b 7.7 ± 3 (1.5 ± 0.6%) [4−11] [Co(II)aq] by cathodic stripping, μM (% Co(II) after 3 h) 0.1 M NaPi pH 0.1 M NaB pH 9.0 [data range] 8.0 [data range] 55 ± 3 (2.8 ± 0.3%) [52−58] 37 ± 2 (0.8 ± 0.1%) [35−39] 79 ± 3 (3.9 ± 0.1%) [77−82] 247 ± 3 (50 ± 5%) [245−250] 6± 3 (1.2 ± 0.6%) [3−9] 10 ± 3 (1.9 ± 0.6%) [7−12] 44 ± 5 (2.2 ± 0.3%) [38−49] 44 ± 5 (1.0 ± 0.1%) [39−50] 170 ± 20 (9 ± 1%) [150−192] 440 ± 50 (90 ± 10%) [390−490] 33 ± 5 (6.6 ± 0.6%) [29−38] 97 ± 9 (19 ± 2%) [88−106] a The Co(II)aq values in 0.1 M NaPi at pH 5.8 and 8.0 were determined using Co(II)aq-induced line broadening 31P NMR. The Co(II)aq values in 0.1 M NaB pH 9.0 were determined using cathodic stripping. bValues obtained for α 1-CoP 2W 17 are with the third, more precise, lower concentration Co(II)aq calibration curve described in the Experimental Section, a calibration curve designed and conducted specifically for this lowest detected Co(II)aq value. complicated by W reduction waves in the differential pulse voltammetry). The results in Table 1 further demonstrate that all of the CoPOMs show some detectable Co(II)aq under any of the conditions examined, ranging from ∼0.25% to now ∼90% of the total cobalt present initially in the Co-POMs in the more basic, pH = 9.0 solution. Additionally, clear solution pHdependent trends are apparent for each Co-POM, Table 1. For example, after 3 h the relatively stable CoPW11 dissociates just 1.3(±0.6)% of its Co(II) in pH 5.8, but dissociates 50(±5) and 90(±10)% of its Co(II) in pH 8.0 and 9.0 solution, respectively. The pH stability of CoPW11 makes sense considering that the synthesis of CoPW11 relies on the partial degradation of the parent PW12O403− Keggin ion at pH ∼ 538 (the parent PW12O403− itself being prepared using concentrated HCl36). Hence, CoPW11 is more stable at the mildly acidic pH 5.8 NaPi buffer employed and then is as expected to be less stable at the higher pH 8−9 values. Overall, our results reiterate an undeniable fact about CoPOMs, namely that Co-POM precatalysts cannot be generally described as 100% “stable” over time under a variety of common buffer and aqueous,25 WOCatalysis and pH conditions, at least as judged by whether or not Co(II)aq is detectable at the ∼0.25% or higher, μM level. Instead, each of Co4P2W18, Co 9 P5 W 27 , Co 4 P4 W 30, CoPW11 , α 1 -CoP 2 W 17 , and α2 CoP2W17 show somewhere between the limits seen of ∼0.25−∼50% detectable Co(II)aq in 0.1 M, NaPi pH = 5.8 or 8.0 and up to ∼90% Co(II) leaching in NaB pH = 9.0 buffer solutions. The percentage of the WOCatalysis observed that can, therefore, be attributed to CoOx formed from even those trace levels of Co(II)aq has to be carefully examined to answer the question of if the observed WOCatalysis is by the intact, molecular Co-POM or the often low-level amount of, however, high activity CoOx formed by even trace levels of Co(II)aq. WOCatalysis Activity: Confirming the Anodic Current Is Due to Water Oxidation. To ensure that the anodic current being observed is from water oxidation, and not some other process such as oxidation of the glassy carbon electrode (which has been observed in potentials greater than +1.4 V vs Ag/AgCl),28 we quantified the O2 produced under standard conditions of 500 μM Co-POM aged 3 h or Co(NO3)2 (6− 500 μM), 0.1 M NaPi pH = 5.8 or 8.0, and NaB pH = 9.0 and at 1.1 V vs Ag/AgCl for 5 min. The theoretical O2 yield for each electrolysis experiment was calculated by dividing the total charge passed in coulombs (determined by integrating the current over time) by the charge of an electron (1.602 × 10−19 C/e−) and using the stoichiometry of 4 e− passed per each 1O2 produced. The O2 concentration was monitored using an Ocean Optics NEOFOX O2-detection probe. By dividing the measured O2 yield at the end of the reaction by the theoretical O2 yield, the Faradaic efficiency of the reaction was also determined. The observed Faradaic efficiency ranged f rom 80 to 100% in all cases. Additionally, a ca. 8% decline in the detected O2 concentration over a ∼ 1 min period after the electrolysis is stopped is almost surely due to O2 equilibration with the reaction vessel’s (minimized) headspace or possibly some escape from the electrochemical cell. In short, the Faradaic efficiency of O2 production is at least ≥80−100%, and because of this, the anodic current can be used as a semiquantitative metric to compare WOCatalysis activity of the Co-POMs and authentic CoOx (i.e., and to within a ± < 20% error), more than sufficient for any of the conclusions reached in the present work. WOCatalysis Activity: O2 Evolution from Co-POMs in Comparison with the Amount of Co(II)aq Released. Constant potential electrolysis was conducted on 3 h aged 500 μM solutions of the Co-POMs and Co(NO3)2 in each of the buffer conditions. The Co(NO3)2 concentrations chosen to compare with each Co-POM were based upon the amount of Co(II)aq that was detected in each buffer condition, Table 1, vide supra. The O2 produced by each Co-POM is summarized in Table S2 of the Supporting Information. The amount of WOCatalysis activity that can be attributed to Co(II)aq is shown in Table 2, in which the O2 yield from Co(II)aq is divided by the O2 yield from the Co-POM (eq 2). A value of 100% (or more) means that all of the catalysis can quantitatively accounted for by Co(II)aq. For example, the percentage of WOCatalysis activity that can be attributed to Co(II)aq for Co4P2W18 in NaPi pH = 8.0 is 150 ± 50%. Such values near or >100% mean that the Co(II)aq present is able to account for all of the WOCatalysis under those specific conditions. % of WOCatalysis contributable to Co(II)aq = O 2 yield from Co(II)aq O2 yield from Co‐POM × 100 (2) Values significantly above 100% (e.g., for α2-CoP2W17 at pH 9.0, 800 ± 300%, Table 2) indicate that the equivalent amount Table 2. Percent of WOCatalysis Activity That Can Be Accounted for by Co(II)aq for the Co-POMs under Each Buffer Conditiona buffer system polyoxometalate Co4P2W18 Co9P5W27 Co4P4W30 CoPW11 α1-CoP2W17 α2-CoP2W17 0.1 M NaPi pH 5.8 60 ± 70 ± 60 ± 20 ± 16 ± 60 ± 30% 60% 40% 20% 6%b 60% 0.1 M NaPi pH 8.0 150 ± 96 ± 140 ± 180 ± 90 ± 90 ± 50%% 24 70% 40% 30% 50% 0.1 M NaB pH 9.0 400 ± 300 ± 140 ± 100 ± 350 ± 800 ± 200% 200% 70% 40% 40% 300% a The Co-POMs (500 μM) were aged for 3 h under each buffer condition. Electrolysis was then conducted at 1.1 V vs Ag/AgCl. The O2 yield (μmol) was determined as described in the text and is listed in Table S2 of the Supporting Information. To compare with the amount of Co(II)aq that is leached, Co(NO3)2 was used in the concentrations determined and the amounts are summarized in Table 1. The amount of O2 produced from the Co(II)aq was divided by the amount of O2 produced from the Co-POMs to determine the percent of WOCatalysis activity that can be accounted for by the Co(II)aq present. bData obtained using the third, lower Co(II)aq calibration curve described in the Experimental Section and designed specifically for the α1-CoP2W17 POM system. of Co(II)aq that is detected after 3 h has a greater WOCatalysis activity than the films generated from Co-POMs. The ≫100% values are interesting, and suggest several possible interpretations, including: (i) that the Co-POM somehow poisons CoOx films made from Co(II)aq in the presence of Co-POMs (indeed, evidence of W incorporation is demonstrated in some Co-POM derived films, vide infra); (ii) that the NO3− somehow enhances the catalysis in CoOx made from Co(NO3)2; (iii) that the Co(II)aq values determined by 31P NMR or cathodic stripping are somewhat higher than the true Co(II)aq values; or possibly (iv) that the film formation (and, for example, the surface area and number of active sites) is affected by the pH54 or the presence of POMs, which in turn affects the observed WOCatalysis. However, the most obvious, and most important, conclusion is (v) that CoOx formed from at least CoII(NO3)2 is a better WOCatalyst at pH 8 and 9 than any of the Co-POMs tested. Values ≪100% are also of considerable interest because they are consistent with molecular, homogeneous Co-POM WOCatalysis or conceivably consistent with some other, presently unknown, ostensibly homogeneous catalyst derived from the Co-POM (alternative hypothesis #5 from Scheme 1). However, considering that both CoPW11 and α1-CoP2W17 are stable in NaPi pH = 5.8 (decomposing by 1.3% and 0.25%, respectively, at this pH), and given that the decomposition byproduct detected is Co(II)aq and (by mass balance) the lacunary PW11O397− and [α1-P2W17O61]10− (which do not contain oxidizable metals that can serve as at least facile WOCatalysts), the simplest (Ockham’s razor) interpretation of the ≪100% data is that the intact CoPW11 and α1-CoP2W17 are the dominant WOCatalysts under those specific pH 5.8 NaPi conditions. For example, the percentage of WOCatalysis activity that can be attributed to Co(II)aq for CoPW11 and α1CoP2W17 in NaPi pH = 5.8 is 20(±20)% and 16(±6%), respectively. These data, in turn, imply that intact CoPW11 and α1-CoP2W17 are the dominant electrochemically driven WOCatalyst at pH = 5.8 for 80(±20)% and 84(±6)% of the observed current. In short, the ≪100% data in Table 2 are consistent with if not strongly supportive of the interpretation that the most stable Co-POMs examined, CoPW11 and α1CoP2W17, are serving as electrochemically driven, molecular WOCatalysts, a previously unavailable, important conclusion given the controversy about when and where Co-POMs can be molecular WOCatalysts. Looking more broadly at Table 2, there are several overarching trends in the data and even given the inherently large error bars in Table 2 (that derive from having to detect mere micromolar levels of Co(II)aq as discussed more in the Supporting Information): at lower pH the Co-POMs account for a greater amount of the WOCatalysis. At higher pH the WOCatalysis current from Co(II)aq becomes increasingly prevalent, with Co(II) accounting for ≥100% of the observed WOCatalysis activity. This pH trend in Co(II)aq contribution to WOCatalysis activity makes sense considering that the CoPOMs examined are often (although not always) more stable at the lower pH, for example, CoPW11 decomposes by only 1.3(±0.6)% at pH 5.8 but decomposes by 50(±5)% and 90(±10)% at pH 8.0 and 9.0, respectively. Hence, unsurprisingly, the Co-POMs examined are more likely to be intact WOCatalyst under conditions where they are demonstrably more stable, pH values closer to the pHs at which they form and are synthesized. Also worth noting here is that the CoOx catalyst is also affected by pH as previously reported,57 with CoOx being more active at higher pH, and not being stable below pH = 3.5.57 Greater WOCatalysis Activity of CoOx Compared to That of the Most Stable Co-POMs. Lastly, although our evidence supports CoPW11 and α1-CoP2W17 as homogeneous WOCatalysts at pH = 5.8, a critical point is that the CoOx formed f rom the equivalent amount of Co(II)aq is an estimated ∼35−150-fold faster WOCatalyst at pH = 5.8 than is the corresponding homogeneous Co-POM (as detailed in the Supporting Information). Even using the ranges and error bars on the data in Tables 1 and 2 to bias the estimate as much as possible in favor of the Co-POM as the catalyst (and then also for the single most stable Co-POM examined, α1CoP2W17) still yields the insight that CoOx formed from the released Co(II)aq is at least 35-fold more active than α1CoP2W17 (see the Supporting Information for details). If one does this same calculation for, again, the most stable α1-CoP2W17 but now at pH = 8, the CoOx is at least 80-fold more active if (and if one again biases the calculation as much as the data allow in favor of Co-POM-based catalysis; see the Supporting Information for details of these estimates), and likely ∼740-fold more active at pH = 8 (see the Supporting Information for the detailed calculations). To summarize, comparing the WOCatalysis activity of the 3 h aged Co-POMs with the amount of detected Co(II)aq reveals that at pH = 8.0 in 0.1 M NaPi and pH = 9.0 in 0.1 M NaB all of the six exemplary Co-POMs examined give rise to heterogeneous CoOx as the dominant WOCatalyst. However, at pH = 5.8 in 0.1 M NaPi and under electrochemically driven WOCatalysis conditions, the evidence strongly suggests that CoPW11 and α1-CoP2W17, and perhaps also Co4P2W18 and α2-CoP2W17, can serve as homogeneous, molecular WOCatalysts, albeit with CoOx being ∼35−150× faster at pH = 5.8 and, most likely, ∼740× faster at pH = 8 (numbers that can be refined using the methods herein if others require greater precision than reported). One key, unequivocal conclusion from the present studies is clear, however: CoOx is at least a ≥ 10-fold more active WOCatalys under electrochemical Figure 5. Selected CVs of electrodes after 5 min controlled potential electrolysis in the original Co-POM solution (red) and once the electrodes were removed, rinsed, and replaced into a fresh, buffer-only solution (blue): (a) Co4P2W18 in 0.1 M NaPi pH 8.0; (b) Co4P2W18 in 0.1 M NaPi pH 8.0 with 120 μM EDTA (2 equiv/Co(II)aq); (c) α1-CoP2W17 in 0.1 M NaPi pH 5.8; (d) α2-CoP2W17 in 0.1 M NaPi pH 8.0. The remainder of the CVs are shown in the Supporting Information. conditions than any of the Co-POMs examined to date for any of the pHs (5.8, 8.0. 9.0) and buffers examined. Electrochemical Characterization of the Deposited Films. Previous studies have shown that electrode-bound heterogeneous CoOx formed from aged Co-POM solutions is active toward WOCatalysis.23,27 Additionally, such CoOx films remain active when the working electrode is removed from the original Co-POM solution and placed in a fresh, buffer-only solution,23,27 thereby providing a way to characterize what amount of the WOCatalysis current detected is attributable to the film. Controls similar to those performed before23,27 were therefore conducted in which controlled potential electrolysis (5−30 min) was conducted in 500 μM solutions of Co-POM that had been aged 3 h. Cyclic voltammetry (CV) was then conducted first in the original Co-POM solution. The electrodes were subsequently removed, rinsed gently with water, replaced into a fresh, buffer-only solution, and a second CV was obtained. The resultant before and after CVs for selected Co-POMs are shown in Figure 5; the rest of the CVs for the Co-POMs and additional CV experiments are provided in Figure S13 Supporting Information. Figure 5a is a control demonstrating that the previously reported, known23,24 catalytically active film from Co4P2W18 can be reproducibly formed as part of the present studies from a 500 μM solution of Co4P2W18 in 0.1 M NaPi at pH 8.0 and after 3 h aging. Figure 5b is a second control that tests the possibility raised previously24 (but heretofore not tested) that CoOx might directly form from Co-POMs as well as from Co(II)aq at sufficiently oxidizing potentials. Hence, the experiment reported in Figure 5b also contains 500 μM Co4P2W18 in 0.1 M NaPi at pH 8.0 that has aged 3 h, but now has been spiked after aging with 120 μM EDTA to chelate the free ∼60 μM Co(II)aq known to be formed. Almost all of the WOCatalysis activity is diminished, and no significant film is formed, implying that Co4P2W18 does not serve as a direct precursor to CoOx at pH 8.0, thereby disproving hypothesis #4 from Scheme 1. The CVs shown in Figure 5c,d present the CVs after electrolysis in the original buffer solution and then in a bufferonly solution for α1-CoP2W17 in 0.1 M NaPi at pH 5.8 and α2CoP2W17 in 0.1 M NaPi at pH 8.0, respectively (both after 3 h of solution aging). The significantly higher current and unique CV features of the original Co-POM solution, vs those for the rinsed electrode replaced into buffer-only solution CV, provide additional evidence for a solution-based species having a role in the observed WOCatalysis for α1-CoP2W17, α2-CoP2W17, and CoPW11. The Ockham’s razor based hypothesis is that, under conditions where a Co-POM such as α1-CoP2W17 in 0.1 M NaPi at pH 5.8 is relatively stable (less than 2% detectable Figure 6. SEM micrograph (left) and XPS (right) of electrodes after 30 min bulk electrolysis from a 3 h aged solution of 500 μM α2-CoP2W17 in 0.1 M NaPi pH 8.0. The globular nature of the film is similar to previously observed films from Co(II) or Co-POMs.8,23,26 The i vs t curve for the film deposition is presented in Figure S14 of the Supporting Information. Co(II)aq), the α1-CoP 2W 17 is serving as a molecular, homogeneous WOCatalysts-albeit one with ≥10× lower WOCatalysis current than the CoOx films formed from the less stable Co-POMs (Table S2). In summary, electrolysis and CV of the electrodes in the electrolyzed solutions (red traces in Figure 5 and Figure S13 of the Supporting Information) followed by electrolysis in bufferonly solutions (blue traces in Figure 5 and Figure S13 of the Supporting Information) helps illuminate whether the active catalyst is a solution-based species or an electrode bound species. The results are in good agreement with the percent WOCatalysis activity from the previous section. For example, at pH 5.8 the percent WOCatalysis evidence suggests that Co4P2W18, CoPW11, α1-CoP2W17, and α2-CoP2W17 can serve as molecular, homogeneous WOCatalyst. The CVs for those Co-POMs in pH 5.8 provide additional evidence in support of a solution-based WOCatalyst (Figure 5 and Figure S13 of the Supporting Information). Other Co-POMs that show evidence of a solution-based WOCatalyst in NaPi at pH = 8.0 are Co9P5W27, α1-CoP2W17, and α2-CoP2W17, whereas in NaB pH = 9.0 only α1-CoP2W17 exhibits evidence of a solutionbased WOCatalyst (Figure 5 and Figure S13 of the Supporting Information). Note that although at pH = 9.0 the CVs of Co9P5W27, α1-CoP2W17, and α2-CoP2W17 at pH = 8.0 and α1CoP2W17 provide evidence of a solution-based WOCatalyst, the results in Table 2 provide evidence that under those at pH = 9.0 conditions, CoOx is still the dominant WOCatalyst. Morphological and Compositional Characterization of Deposited Films. Most of the Co-POMs showed an increase in WOCatalysis activity for longer electrolysis times, which is characteristic of CoOx film deposition (Figure S14 of the Supporting Information).8,23,27 Hence, we conducted electrolysis for 30 min to allow film accumulation and then dried the films for SEM and XPS characterization. Figure 6 shows a typical electrode-bound film of globular particles that are formed from 3 h aged solutions of 500 μM α2-CoP2W17 in 0.1 M NaPi pH 8.0. The XPS of the film from α2-CoP 2W 17 contains carbon (from the glassy carbon substrate), oxygen, cobalt, sodium, phosphorus, and tungsten, Figure 6 (right). The Co/W atom ratio from the highresolution XPS scans was determined to be 2.1:1.3, whereas the Co/W ratio in the structure is 1:17, meaning that although W incorporation does occur, the original Co-POM is not a major component. This experiment was reproduced twice and similar XPS spectra were obtained, demonstrating reproducible W incorporation into CoOx films produced from 500 μM α2CoP2W17 in 0.1 M NaPi pH 8.0. Relevant here is that W incorporation into CoOx films formed in the presence of aqueous Na2WO4 are both known and exhibit different WOC activity than seen for pure CoOx films formed from just Co(II) in the absence of W.58 The presence of tungsten in films formed from α2-CoP2W17 therefore differs from CoOx films that form from Co4P2W18 and Co4V2W18 that do not contain tungsten23,27 and from films formed from just Co(II). In short, using Co-POMs as a precatalyst for films that make Wcontaining CoOx films involves two (unintended) leached elements of the original Co-POM. Next, 30 min electrolysis was conducted on 3 h aged solutions of 500 μM Co 4P 2W18 in 0.1 M NaPi pH 8.0 with 10 equiv of EDTA/Co(II)aq added after 3 h aging, but prior to electrolysis. The SEM and XPS of that particular electrode is presented in Figure 7 and confirms that heterogeneous CoOx does not form in the presence of excess EDTA from 3 h aged solutions of 500 μM Co4P2W18 in 0.1 M NaPi pH 8.0. This finding provides further evidence that the Co-POM cannot form CoOx directly from, for example, putative electrodebound Co-POM. Instead, the CoOx film observed when starting with the Co4P2W18 precatalyst is formed by Co4P2W18 releasing Co(II)aq, consistent with hypothesis #3 (i.e., and not #4) from Scheme 1, vide supra. Additional CV experiments using 3 h aged 500 μM α1CoP2W17 in 0.1 M NaPi pH 5.8 are discussed in the Supporting Information (Figures S15 and S16). The main results from those experiments using this more stable Co-POM is that although catalytic current increases, an electrode bound film is not formed from the bulk electrolysis of the Co-POM solution. Figure 7. SEM micrograph (left) and XPS (right) of electrodes after 30 min bulk electrolysis from a 3 h aged solution of 500 μM Co4P2W18 in 0.1 M NaPi pH 8.0 with 600 μM EDTA (10 equiv/Co(II)aq). The i vs t curve for the film deposition is presented in Figure S14 of the Supporting Information. To summarize the experiments on the electrochemical and morphological characterization of the deposited films, under conditions where the Co-POMs show >2% detectable Co(II)aq, CoOx is formed and that film accounts quantitatively for the observed WOCatalysis (Table 2, Figure 3, and Figures S13 and S16 of the Supporting Information). However, under conditions where the Co-POMs are more stable (<2% detectable Co(II)aq) such as with α1-CoP2W17, no detectable electrode-bound CoOx is seen. Rather, a solution-based species is responsible for the observed WOCatalysis current (Table 2, Figures 5, and Figures S13 and S15 of the Supporting Information), again and ostensibly the starting Co-POM at the Ockham’s razor level of interpretation. Lastly, addition of a 10fold excess of EDTA (vs the amount of free Co(II)aq detected) prevents the formation of CoOx, at least with 3 h aged solution of 500 μM Co4P2W18 in 0.1 M NaPi pH 8.0 (Figure 7). This, too, is evidence that CoOx is formed from Co(II)aq and not from intact, electrode-bound Co-POM. ■ SUMMARY AND CONCLUSIONS The present study details the broadest and most quantitative, micromolar-level examination to date of the stability and electrochemically driven WOCatalysis from Co-POM precatalysts. Six exemplary Co-POMs [Co4(H2O)2 (PW9O34)2]10− (Co4P2W18), [β,β-Co4(H2 O)2(P2W15 O56)2]16− (Co 4P4 W30), [Co9(H2O)6(OH)3(HPO4)2(PW9O34)3]16− (Co9P5W27), [Co(H2O)PW 11O39]5− (CoPW 11), [α 1-Co(H2 O)P2 W 17 O 61] 8− ( α 1 -CoP 2 W 17 ), and [ α 2 -Co(H 2 O)P 2 W 1 7 O 6 1 ] 8 − ( α 2 CoP2W17) were synthesized, their structural integrity established, and then their stability and electrochemically driven WOCatalysis examined under pH 5.8, 8.0, and 9.0 buffer conditions chosen from the literature. Importantly, the amount of Co(II)aq leached from the Co-POMs into solution was quantified directly, at the μM-level, using Co(II)aq-induced line broadening of the 31P NMR resonance of phosphate buffer at pH 5.8 and 8.0, and by cathodic stripping in the case of pH 9.0 borate buffer. The WOCatalysis activity derived from the CoPOM precatalysts was then compared with the WOCatalysis activity of the equivalent amount of Co(II)aq present in solution from each of the Co-POMs. The main conclusions from this study are the following: • Significantly, Co(II)aq at the micromolar or higher level was detected for every Co-POM under each set of pH and buffer conditions. The amount of detectable Co(II)aq as a percentage of the total cobalt present in each Co-POM varies from ∼0.25% to 50% (and up to 90% in borate buffer at pH 9) of the total Co(II) after 3 h in solution, the precise amount being unique to the POM structure/Co(II) binding site and notably the pH and buffer, higher pH values and phosphate buffer in general leading to higher levels of Co(II)aq (Figure 2 and Table 1, vide supra). • In the case of Co-POMs with high anionic charge such as [β,β-Co 4 (H 2 O) 2 (P 2 W 15 O 56 ) 2 ] 16 − (Co 4 P 4 W 30 ), Co(II) can be present as a countercation impurity-a likely more general phenomenon for Mn+ ions used in the synthesis of Mn+-POMs. While ion-exchange resin, recrystallization from counterion-controlled solutions, or other countercation control efforts may be able to remove such countercation impurities, that remains to be demonstrated by ≤ micromolar-sensitive methods such as those employed herein. • In 12 out of the 18 Co-POM cases at pH 8.0 and 9.0 in Table 2, the amount of heterogeneous CoOx generated from the detected Co(II)aq accounts for ≥100% of the observed activity-meaning that under those higher pH conditions the kinetically dominant, electrochemically driven WOCatalyst is heterogeneous CoOx. In those cases, using simple Co(II) salts to prepare the resultant, high-activity CoOx would be a far easier, greener, and overall better use of chemicals, time, and synthetic effort. • In terms of catalytic rate, at pH 8.0 and for the single most stable Co-POM, α1-CoP2W17, the CoOx catalyst formed from Co(II)aq is an estimated at least 80-fold if not ∼740-fold more active than any (undetectable) CoPOM based WOC. As an illustrative example, this means that (using the 740-fold value) even ∼0.14% of decomposition of α1-CoP2W17 to Co(II)aq would be able in turn, at pH = 8.0, to carry ≥99% of the catalytic WOCatalysis current. Put in other words, finding the kinetically dominant true WOCatalyst when starting with 500 mM Co-POMs is a μM detection and detective problem. • However, under pH 5.8 conditions where the Co-POMs are generally more stable, the amount of Co(II)aq detected cannot account for the observed WOCatalysis. Specifically, for CoPW11 and α1-CoP2W17 at pH 5.8 where ∼1.3% and ∼0.25% detectable Co(II)aq are seen, respectively, 80(±20%) and 84%(±6%) of the observed WOCatalysis activity can be ascribed (in the Ockham’s razor interpretation) to molecular, Co-POM-based catalysis, Table 2, vide supra. That said, the Co-POMbased WOCatalysis rate is still an estimated ∼35 to 150fold slower than that for an equivalent amount of CoOx for even the most stable Co-POM examined, α1CoP2W17. • In general, our findings confirm and fully support those of prior workers who have concluded that the reaction conditions are important in determining the identity of the kinetically dominant WOCatalyst derived from CoPOMs.17,24−27 We emphasize here that all of our experiments were deliberately conducted under electrochemical conditions; the nature of the true WOCatalysts under chemical or photochemical oxidation (e.g., using Ru(III)bpy33+ or Ru(II)bpy32++ hν) will likely be different under those (different) conditions. That said, the method of multiple alternative hypotheses, particularly those listed in Scheme 1, are expected to prove useful to future researchers striving to determine experimentally the true, kinetically dominant WOCatalyst under their own oxidant, pH, buffer, and other specific conditions. • A summary of additional POMs used in WOCatalysis which are not discussed in the main text, yet merit further study as to the identity of the true catalyst, are presented in Table S1 of the Supporting Information for the interested reader. Finally and overall, the results obtained and presented herein in combination with prior notable work from others in the field of electrocatalytic WOCatalysis7,13,17,23−27 suggest that even more hydrolytically stable Co-POM and other Metal-POM WOCatalysts merit further development. The combined results also illustrate a successful, arguably preferred methodology for distinguishing molecular homogeneous from metal oxide heterogeneous WOCatalysts, even when metal-leaching or countercation contamination is present at just micromolar levels. It is hoped that these efforts will allow even more stable and active Co-POM based WOCatalysts to be developed, studies also hopefully now able to report compelling evidence for or against molecular, Co-POM-based vs heterogeneous, CoOx-based WOCatalysis. ■ ASSOCIATED CONTENT * Supporting Information The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/jacs.8b06303. Tables S1 and S2, Figures S1−S16, and a brief discussion of the error bars in Table 2 (PDF) ■ AUTHOR INFORMATION Corresponding Author *richard.finke@colostate.edu ORCID ́ José Ramon Galan-Mascaros: 0000-0001-7983-9762 Richard G. Finke: 0000-0002-3668-7903 Present Address § Present address: School of Chemistry and CRANN/AMBER Nanoscience Institute, Trinity College, College Green, Dublin 2, Ireland. Notes The authors declare no competing financial interest. ■ ACKNOWLEDGMENTS ■ REFERENCES Funding from NSF Grant Nos. CHE-1361515 and 1664646 is gratefully acknowledged. It is also a pleasure to acknowledge discussions and cooperation with Professor Craig Hill and his research group, discussions and a cooperation which have helped highlight the issues and complexities in establishing the true catalyst in Co-POM precatalyst-based WOCatalysis. Finally, Dr. Jordan Stracke is thanked for early thoughts about this project, and for preparing the sample of Co4P2W18 used herein. (1) Turner, J. A. Science 1999, 285, 687−689. (2) Lewis, N. S.; Nocera, D. G. Proc. Natl. Acad. Sci. U. S. A. 2006, 103, 15729−15735. (3) Karkas, M.; Åkermark, B. Dalton Trans 2016, 45, 14421−14461. (4) Llobet, A. Molecular Water Oxidation Catalysis, 1 ed..; Wiley, 2014. (5) Liu, F.; Concepcion, J. J.; Jurss, J. W.; Cardolaccia, T.; Templeton, J. L.; Meyer, T. J. Inorg. Chem. 2008, 47, 1727−1752. (6) Geletii, Y. V.; Botar, B.; Kögerler, P.; Hillesheim, D. A.; Musaev, D. G.; Hill, C. L. Angew. Chem., Int. Ed. 2008, 47, 3896−3899. (7) Sartorel, A.; Carraro, M.; Scorrano, G.; Zorzi, R. D.; Geremia, S.; McDaniel, N. D.; Bernhard, S.; Bonchio, M. J. Am. Chem. Soc. 2008, 130, 5006−5007. (8) Kanan, M. W.; Nocera, D. G. Science 2008, 321, 1072−1075. (9) Yin, Q.; Tan, J. M.; Besson, C.; Geletii, Y. V.; Musaev, D. G.; Kuznetsov, A. E.; Luo, Z.; Hardcastle, K. I.; Hill, C. L. Science 2010, 328, 342−345. (10) Huang, Z.; Luo, Z.; Geletii, Y. V.; Vickers, J. W.; Yin, Q.; Wu, D.; Hou, Y.; Ding, Y.; Song, J.; Musaev, D. G.; Hill, C. L.; Lian, T. J. Am. Chem. Soc. 2011, 133, 2068−2071. (11) Zhu, G.; Geletii, Y. V.; Kögerler, P.; Schilder, H.; Song, J.; Lense, S.; Zhao, C.; Hardcastle, K. I.; Musaev, D. G.; Hill, C. L. Dalton Trans 2012, 41, 2084. (12) Zhu, G.; Glass, E. N.; Zhao, C.; Lv, H.; Vickers, J. W.; Geletii, Y. V.; Musaev, D. G.; Song, J.; Hill, C. L. Dalton Trans 2012, 41, 13043−13049. (13) Goberna-Ferron, S.; Vigara, L.; Soriano-Lopez, J.; GalanMascaros, J. R. Inorg. Chem. 2012, 51, 11707−11715. (14) Car, P.-E.; Guttentag, M.; Baldridge, K. K.; Alberto, R.; Patzke, G. R. Green Chem. 2012, 14, 1680. (15) Evangelisti, F.; Car, P.-E.; Blacque, O.; Patzke, G. R. Catal. Sci. Technol. 2013, 3, 3117. (16) Lv, H.; Song, J.; Geletii, Y. V.; Vickers, J. W.; Sumliner, J. M.; Musaev, D. G.; Kögerler, P.; Zhuk, P. F.; Bacsa, J.; Zhu, G.; Hill, C. L. J. Am. Chem. Soc. 2014, 136, 9268−9271. (17) Vickers, J. W.; Lv, H.; Sumliner, J. M.; Zhu, G.; Luo, Z.; Musaev, D. G.; Geletii, Y. V.; Hill, C. L. J. Am. Chem. Soc. 2013, 135 (38), 14110−14118. (18) Pope, M. T. Heteropoly and Isopoly Oxometalates; SpringerVerlag, 1983. (19) Yamase, T.; Pope, M. Polyoxometalate Chemistry for NanoComposite Design; Springer Science & Business Media, 2006. (20) Pope, M.; Muller, A. Polyoxometalates: From Platonic Solids to Anti-Retroviral Activity: From Platonic Solids to Anti-Retroviral Activity; Springer Science & Business Media, 1994. (21) Contant, R.; Ciabrini, J.-P. J. Chem. Res. 1982, 50−51. (22) Contant, R. J. Chem. Res. 1984, 120−121. (23) Stracke, J. J.; Finke, R. G. J. Am. Chem. Soc. 2011, 133, 14872− 14875. (24) Stracke, J. J.; Finke, R. G. ACS Catal. 2013, 3, 1209−1219. (25) Stracke, J. J.; Finke, R. G. ACS Catal. 2014, 4, 909−933. (26) Stracke, J. J.; Finke, R. G. ACS Catal. 2014, 4 (1), 79−89. (27) Folkman, S. J.; Finke, R. G. ACS Catal. 2017, 7, 7−16. (28) Ullman, A. M.; Liu, Y.; Huynh, M.; Bediako, D. K.; Wang, H.; Anderson, B. L.; Powers, D. C.; Breen, J. J.; Abruña, H. D.; Nocera, D. G. J. Am. Chem. Soc. 2014, 136 (50), 17681−17688. (29) Platt, J. R. Science 1964, 146 (3642), 347−353. (30) Weakley, T. J. R. J. Chem. Soc., Chem. Commun. 1984, 21, 1406−1407. (31) Galan-Mascaros, J. R.; Gomez-Garcia, C. J.; Borras-Almenar, J. J.; Coronado, E. Adv. Mater. 1994, 6 (3), 221−223. (32) Soriano-Lopez, J.; Goberna-Ferron, S.; Vigara, L.; Carbo,́ J. J.; Poblet, J. M.; Galan-Mascaros, J. R. Inorg. Chem. 2013, 52 (9), 4753− 4755. (33) Blasco-Ahicart, M.; Soriano-Lopez, J.; Carbo,́ J. J.; Poblet, J. M.; Galan-Mascaros, J. R. Nat. Chem. 2018, 10 (1), 24−30. (34) Darensbourg, M. Inorganic Syntheses; John Wiley & Sons, 1998; Vol. 32, pp 175−182. (35) Ruhlmann, L.; Costa-Coquelard, C.; Canny, J.; Thouvenot, R. Eur. J. Inorg. Chem. 2007, 2007, 1493−1500. (36) Bailar, J. C.; Booth, H. S.; Grennert, M. Phosphotungstic Acid. In Inorganic Syntheses; Booth, H. S., Ed.; John Wiley & Sons, Inc., 1939; pp 132−133. (37) Weakley, T. J. R.; Malik, S. A. J. Inorg. Nucl. Chem. 1967, 29, 2935−2944. (38) Tourne,́ C. M.; Tourne,́ G. F.; Malik, S. A.; Weakley, T. J. R. J. Inorg. Nucl. Chem. 1970, 32, 3875−3890. (39) Jorris, T. L.; Kozik, M.; Casan-Pastor, N.; Domaille, P. J.; Finke, R. G.; Miller, W. K.; Baker, L. C. W. J. Am. Chem. Soc. 1987, 109, 7402−7408. (40) Li, J.; Wang, J.; Zhang, L.; Sang, X.; You, W. J. Coord. Chem. 2017, 70, 2950−2957. (41) Contant, R.; Klemperer, W. G.; Yaghi, O. Potassium Octadecatungstodiphosphates(V) and Related Lacunary Compounds. Inorganic Syntheses; Ginsberg, A. P., Ed.; John Wiley & Sons, Inc., 1990; pp 104−111. (42) Samonte, J. L.; Pope, M. T. Can. J. Chem. 2001, 79, 802−808. (43) Lyon, D. K.; Miller, W. K.; Novet, T.; Domaille, P. J.; Evitt, E.; Johnson, D. C.; Finke, R. G. J. Am. Chem. Soc. 1991, 113, 7209−7221. (44) Barats-Damatov, D.; Shimon, L. J. W.; Weiner, L.; Schreiber, R. E.; Jimenez-Lozano, P.; Poblet, J. M.; de Graaf, C.; Neumann, R. Inorg. Chem. 2014, 53 (3), 1779−1787. (45) Haynes, W. M.; Lide, D. R. Handbook of Chemistry and Physics, 92nd ed.; Taylor and Francis Group, 2011. (46) Klanberg, F.; Hunt, J. P.; Dodgen, H. W. Naturwissenschaften 1963, 50 (3), 90−91. (47) Surendranath, Y.; Lutterman, D. A.; Liu, Y.; Nocera, D. G. J. Am. Chem. Soc. 2012, 134 (14), 6326−6336. (48) Bazzan, I.; Volpe, A.; Dolbecq, A.; Natali, M.; Sartorel, A.; Mialane, P.; Bonchio, M. Catal. Today 2017, 290, 39−50. (49) Krolicka, A.; Bobrowski, A.; Kalcher, K.; Mocak, J.; Svancara, I.; Vytras, K. Electroanalysis 2003, 15 (23−24), 1859−1863. (50) Widegren, J. A.; Finke, R. G. J. Mol. Catal. A: Chem. 2003, 198 (1), 317−341. (51) Kirner, J. T.; Stracke, J. J.; Gregg, B. A.; Finke, R. G. ACS Appl. Mater. Interfaces 2014, 6 (16), 13367−13377. (52) Finke, R. G.; Droege, M.; Hutchinson, J. R.; Gansow, O. J. Am. Chem. Soc. 1981, 103 (6), 1587−1589. (53) Folkman, S. J.; Kirner, J. T.; Finke, R. G. Inorg. Chem. 2016, 55 (11), 5343−5355. (54) Baes, C.; Mesmer, R. The Hydrolysis of Cations; John Wiley & Sons, 1976. (55) Finke, R. G.; Droege, M. W.; Domaille, P. J. Inorg. Chem. 1987, 26 (23), 3886−3896. (56) Ruhlmann, L.; Nadjo, L.; Canny, J.; Contant, R.; Thouvenot, R. Eur. J. Inorg. Chem. 2002, 2002, 975−986. (57) Gerken, J. B.; McAlpin, J. G.; Chen, J. Y. C.; Rigsby, M. L.; Casey, W. H.; Britt, R. D.; Stahl, S. S. J. Am. Chem. Soc. 2011, 133 (36), 14431−14442. (58) Zhang, B.; Wu, X.; Li, F.; Yu, F.; Wang, Y.; Sun, L. Chem. Asian J. 2015, 10, 2228−2233.