Kinetics of electron transfer reactions of H2-evolving cobalt diglyoxime catalysts

Article abstract of DOI:10.1021/ja9080259

Co-diglyoxime complexes catalyze H₂ evolution from protic solutions at modest overpotentials. Upon reduction to Co^I, a Co ^III-hydride is formed by reaction with a proton donor. Two pathways for H₂ production are analyzed: one is a heterolytic route involving protonation of the hydride to release H₂ and generate Co ^III; the other is a homoytic pathway requiring association of two Co^III-hydrides. Rate constants and reorganization parameters were estimated from analyses of laser flash-quench kinetics experiments (Co ^III-Co^II self-exchange k = 9.5 × 10^-8 - 2.6 × 10^-5 M^-1 s^-1; λ = 3.9 (±0.3) eV: Co^II-Co^I self-exchange k = 1.2 (±0.5) × 10⁵ M^-1 s^-1; λ = 1.4 (±0.05) eV). Examination of both the barriers and driving forces associated with the two pathways indicates that the homolytic reaction (Co ^IIIH + Co^IIIH → 2 Co^II + H₂) is favored over the route that goes through a Co^III intermediate (Co^IIIH + H⁺ → Co^III + H₂).

Full text of DOI:10.1021/ja9080259

Published on Web 12/31/2009
Kinetics of Electron Transfer Reactions of H2-Evolving Cobalt
Diglyoxime Catalysts
Jillian L. Dempsey, Jay R. Winkler,* and Harry B. Gray*
Beckman Institute, California Institute of Technology, Pasadena, California 91125
Received September 21, 2009; E-mail: winklerj@caltech.edu; hbgray@caltech.edu
Abstract: Co-diglyoxime complexes catalyze H2 evolution from protic solutions at modest overpotentials.
Upon reduction to CoI, a CoIII-hydride is formed by reaction with a proton donor. Two pathways for H2
production are analyzed: one is a heterolytic route involving protonation of the hydride to release H2 and
generate CoIII; the other is a homoytic pathway requiring association of two CoIII-hydrides. Rate constants
and reorganization parameters were estimated from analyses of laser flash-quench kinetics experiments
(CoIII-CoII self-exchange k ) 9.5 × 10-8 - 2.6 × 10-5 M-1 s-1; λ ) 3.9 ((0.3) eV: CoII-CoI self-exchange
k ) 1.2 ((0.5) × 105 M-1 s-1; λ ) 1.4 ((0.05) eV). Examination of both the barriers and driving forces
associated with the two pathways indicates that the homolytic reaction (CoIIIH + CoIIIH f 2 CoII + H2) is
favored over the route that goes through a CoIII intermediate (CoIIIH + H+ f CoIII + H2).
1. Introduction
Energy-efficient solar water splitting requires a catalyst that
can use photochemically generated reducing equivalents to
produce molecular hydrogen from protic solutions with high
turnover frequency.1,2 Practical solar hydrogen production
depends on minimizing the release of free energy, yet most
molecular H2-evolution catalysts operate with substantial over-
potentials (-0.9 to -1.8 V vs SHE).3 We are investigating the
kinetics and thermodynamics of active CoII catalysts in order
to elucidate the barriers associated with elementary steps
involved in H2 evolution.
A CoIII-hydride (CoIIIH) formed in the reaction between a
CoI complex and a proton donor (HA) likely is a key intermedi-
ate in the catalytic cycle.4 Production of H2 from CoIIIH can
occur via homolysis (Figure 1, red path) or heterolysis (Figure
1, blue path) of the CoIII-H bond.5,6 Efficient catalysis by the
heterolytic pathway demands that the CoIII product of CoIII-H
bond heterolysis be converted rapidly to CoII in the pool of
reducing equivalents.
Figure 1. Homolytic (red) and heterolytic (blue) reaction pathways for
catalysis of H2 evolution by CoII complexes. Reducing equivalents (D)
deliver electrons to CoII complexes, and the resulting CoI species are
protonated to form CoIIIH intermediates. In the homolytic pathway, two
CoIIIH species undergo bimolecular reductive elimination to produce H2.
In the heterolytic pathway, protonation of CoIIIH is responsible for H2
release, and the CoIII species is reduced back to CoII by D.
employed a laser flash-quench method9,10 to trigger the
reduction of CoII ([Co(dpgBF2)2(CH3CN)2]) to CoI ([Co-
(dpgBF2)2(CH3CN)]-), and time-resolved spectroscopy to iden-
tify and monitor the short-lived species. This method permits
observation of discrete intermediates on time scales appreciably
shorter than those accessible by conventional stopped-flow
spectroscopy. Analyses of these flash-quench kinetics data have
revealed the barriers for key elementary steps in the catalytic
cycle leading to the formation of H2.
Difluoroboryl-bridged CoII-diglyoxime complexes ([Co-
(dpgBF2)2(CH3CN)2], [Co(dmgBF2)2(CH3CN)2]: dpg ) difluo-
roboryl-diphenylglyoxime, dmg ) difluoroboryl-dimethylgly-
oxime)7 are among a handful of catalysts8 that operate at
relatively high rates with modest overpotentials. We have
(1) Esswein, A. J.; Nocera, D. G. Chem. ReV. 2007, 107, 4022–4047.
(2) (a) Du, P.; Schneider, J.; Luo, G.; Brennessel, W. W.; Eisenberg, R.
Inorg. Chem. 2009, 48, 4952–4962. (b) Lazarides, T.; McCormick,
T.; Du, P.; Luo, G.; Lindley, B.; Eisenberg, R. J. Am. Chem. Soc.
2009, 131, 9192–9194. (c) Fihri, A.; Artero, V.; Pereira, A.; Fontecave,
M. Dalton Trans. 2008, 5567–5569. (d) Probst, B.; Kolano, C.; Hamm,
P.; Alberto, R. Inorg. Chem. 2009, 48, 1836–1843.
(7) Hu, X.; Cossairt, B. M.; Brunschwig, B. S.; Lewis, N. S.; Peters, J. C.
Chem. Commun. 2005, 4723–4725.
(8) (a) Appel, A. M.; DuBois, D. L.; DuBois, M. R. J. Am. Chem. Soc.
2005, 127, 12717–12726. (b) Wilson, A. D.; Newell, R. H.; McNevin,
M. J.; Muckerman, J. T.; DuBois, M. R.; DuBois, D. L. J. Am. Chem.
Soc. 2006, 128, 358–366. (c) Razavet, M.; Artero, V.; Fontecave, M.
Inorg. Chem. 2005, 44, 4786–4795.
(3) General reviews on transition metal hydrogen evolution catalysts: (a)
Artero, V.; Fontecave, M. Coord. Chem. ReV. 2005, 249, 1518–1535.
(b) Koelle, U. New. J. Chem. 1992, 16, 157–169.
(4) Connolly, P.; Espenson, J. H. Inorg. Chem. 1986, 25, 2684–2688.
(5) Baffert, C.; Artero, V.; Fontecave, M. Inorg. Chem. 2007, 46, 1817–
1824.
(9) Ogata, T.; Yanagida, S.; Brunschwig, B. S.; Fujita, E. J. Am. Chem.
Soc. 1995, 117, 6708–6716.
(10) Bjerrum, M. J.; Casimiro, D. R.; Chang, I.-J.; Di Bilio, A. J.; Gray,
H. B.; Hill, M. G.; Langen, R.; Mines, G. A.; Skov, L. K.; Winkler,
J. R.; Wuttke, D. S. J. Bioenerg. Biomembr. 1995, 27, 295–302.
(6) Hu, X.; Brunschwig, B. S.; Peters, J. C. J. Am. Chem. Soc. 2007,
129, 8988–8998.
9
1060 J. AM. CHEM. SOC. 2010, 132, 1060–1065
10.1021/ja9080259  2010 American Chemical Society

Electron Transfer Reactions of Cobalt Diglyoximes
A R T I C L E S
Scheme 1
100, 200, and 500 µs, and fits were overlaid with the difference
spectra (Figure S2, Supporting Information). Small deviations
in the fits were observed in the 400-500 nm region and
Figure 2. Transient difference spectra 20 µM [Ru(bpy)3][PF6]2, 3.5 mM
[MV][PF6]2, 0.1 M NBu4PF6, and 100 µM CoII in CH3CN at selected time
delays after laser excitation. λex ) 355 nm.
3+
attributed to the approximation made for εCo (λ).
A kinetics model (Scheme 1) was used to describe the
photochemical electron transfer (ET) reaction scheme. The decay
of the excited state of [Ru(bpy)3]2+ is described by rate constant
k1, and measured via time-resolved luminescence. Bimolecular
oxidative quenching by MV2+ is described by k2, and the charge
recombination by k3. Stern-Volmer quenching was used to
determine k2; the value was consistent with previous reports.11
The single-wavelength transient absorption (λex ) 480 nm, λobs
) 730 nm) of a sample containing 20 µM [Ru(bpy)3][PF6]2,
0.1 M NBu4PF6, 3.5 mM [MV][PF6]2, was modeled with second-
order kinetics to obtain a value for k3 that was consistent with
literature reports.12
Electron transfer from MV•+ to CoII is described by k4.
[Ru(bpy)3]3+ oxidations of CoI and CoII are identified with rate
constants k5 and k6, respectively. Charge recombination reactions
between CoIII and either MV•+ or CoI are described by k7 and
k8, respectively. Kinetics simulations were performed with
differential equations (Scheme S1, Supporting Information)
describing the concentrations of all species involved in these
ET reactions. Inputs to solve the initial value problem included
a vector describing initial concentrations of reactants and
products, a time-span vector (10-6 to 10-1 s), and a vector of
rate constants k1-k8. Profiles of concentration changes versus
time were plotted for each individual species. These concentra-
tion profiles were translated to simulated transient difference
spectra by summing the products of a concentration profile
multiplied by an estimated absorption coefficient at 730 nm for
each species (Table S1, Supporting Information). The rate
constants (k4, k5, k6, k7, and k8) were iteratively adjusted until a
satisfactory simulation of the TA kinetics was obtained (Figure
3). Simulations were repeated for six different sample measure-
ments containing variable amounts of CoII (63-157 µM) and
were iterated to obtain a self-consistent set of rate constants.
2.2. Rate Constants and Reorganization Energies. The rate
constants for electron-exchange reactions of CoIII and CoII as
well as CoII and CoI can be estimated using the Marcus cross
relation,13 k12 ) (k11k22K12f12)1/2, where k12 describes the rate
constant of an ET cross reaction, k11 and k22 are the self-
2. Results
2.1. Photochemistry. The photochemical method utilizes
methyl viologen (MV2+) to quench pulsed-laser excited [Ru-
(bpy)3]2+ (bpy ) 2,2′-bipyridine). The reduced viologen species,
MV•+, delivers an electron to CoII, producing anionic CoI and
[Ru(bpy)3]3+ The transient absorption (TA) spectrum recorded
250 ns after excitation is dominated by spectral features
characteristic of both [Ru(bpy)3]3+, bleaching at 450 nm, and
MV•+, absorbing at 390 and 605 nm (Figure 2). Between 20
and 100 µs, the shift in the red absorption feature to 686 nm
with concomitant loss of the 390 nm absorbance matches the
spectroscopic signature of CoI (see Figure S1, Supporting
Information). During the time interval 500 µs to 50 ms, CoI
reacts with two oxidants in solution, [Ru(bpy)3]3+ and CoIII
([Co(dpgBF2)2(CH3CN)2]+), the latter produced transiently via
[Ru(bpy)3]3+ oxidation of CoII.
The transient difference spectra were fit with linear combina-
tions of the following three molar difference spectra:
∆ε1(λ) ) [εRu3+(λ) + εMV•+(λ) - εRu2+(λ) - εMV2+(λ)]
∆ε2(λ) ) [εRu3+(λ) + εCo+(λ) - εRu2+(λ) - εCo2+(λ)]
∆ε3(λ) ) [εCo3+(λ) + εCo+(λ) - 2εCo2+(λ)]
where ∆ε1(λ), ∆ε2(λ), and ∆ε3(λ) describe differences in molar
extinction coefficients between charge transfer pairs. Vectors
(∆A) describing the time-resolved difference spectra at discrete
wavelengths (λ, 400-800 nm) were generated for several delay
times, t. A matrix, ∆ε, was formulated to describe the molar
difference spectra described above, ∆ε ) [∆ε1(λ), ∆ε2(λ),
∆ε3(λ)]. An approximation for εCo (λ) was made (εCo (λ )
400-800 nm) ) 0), as CoIII is not isolable. Note, however,
that the absorption profile of the CoIIIMe species,
Me-Co(dpgBF2)2L, only has a weak feature at 406 nm and a
tail into the visible region. As the transient difference spectra
should be a linear combination of ∆εi (i ) 1,2,3), the expression
∆A ) ∆ε × r was solved for r (r ) ∆ε-1 × ∆A), where r
is a column vector reflecting the contributions of the three molar
difference spectra to the experimental spectrum. Fits of the
3+
3+
(11) Chiorboli, C.; Indelli, M. T.; Scandola, M. A. R.; Scandola, F. J. Phys.
Chem. 1988, 92, 156–163.
(12) Clark, C. D.; Hoffman, M. Z. J. Phys. Chem. 1996, 100, 7526–7532.
(13) Marcus, R. A.; Sutin, N. Biochem. Biophys. Acta 1985, 811, 265–
322.
transient spectra were generated using the expression ∆Afit
∆ε × r. The process was repeated for time delays of 20, 50,
)
9
J. AM. CHEM. SOC. VOL. 132, NO. 3, 2010 1061

A R T I C L E S
Dempsey et al.
HA and CoIIIH. With sufficiently powerful reductants and proton
sources, these steps are exergonic.
In the homolytic reaction pathway, the driving force for the
elementary step for H2 evolution is -∆GoH , the difference
2
between ∆GoR and ∆Go4. The driving force for the elementary
H2 formation step in the heterolytic proton reduction pathway
(∆GoH ) depends on the average of Eo(CoIII/II) and Eo(CoII/I),
2a
relative to the proton reduction potential, and the pKa difference
of HA and CoIIIH. To estimate a lower limit for ∆GoH , we
2a
assume pKa(HA) ) pKa(CoIIIH); thus, ∆GHo > F[Eo
+
II
I
Co /Co
2a
Eo
- 2EoHA/H ].
CoIII/CoII
2
The barriers for electron transfer from D to CoII (∆GEqT1 and
D to CoIII (∆GEqT2) depend both on the driving forces and
reorganization parameters. We assume Eo(D+/0) ) Eo(CoII/I
)
and a range of self-exchange reorganization energies for the
D+/0 couple (λ11 ) 0-1.5 eV). We take λ22(CoII/I) ) 1.4 eV
and λ22(CoIII/II) ) 3.9 eV, as calculated above, and the
reorganization energy for the cross reactions as λ12 ) 1/2(λ11
+ λ22). The activation free-energy change for reduction of CoII is
∆Gq(CoII/I) ) λ12(CoII/I)/4 (4-8 kcal mol-1) and that for reduction
of CoIII is ∆Gq(CoIII/II) ) [-0.5 + λ12(CoIII/II)]2/(4λ12(CoIII/II))
(6-10 kcal mol-1, -∆Go for reduction of CoIII is -0.5 eV). The
relative reaction rates for these two ET processes are given by
k(CoII/I)/k(CoIII/II) ) exp([∆Gq(CoIII/II) - ∆Gq(CoII/I)]/RT) ) 35
at 295 K.
Figure 3. Kinetics trace of 20 µM [Ru(bpy)3][PF6]2, 3.5 mM [MV][PF6]2,
0.1 M NBu4PF6, 125 µM CoII, and the simulated fit in logarithmic time. λex
)
480 nm, λobs ) 730 nm. ET rate constants used in the simulation: k1, 1.04 ×
106 s-1; k2, 1.58 × 109 M-1 s-1; k3, 5.7 × 109 M-1 s-1; k4, 2.0 × 108 M-1 s-1
108 M-1 s-1. Inset: kinetics and fit plotted on a linear time axis.
;
k5, 9.8 × 109 M-1 s-1; k6, 5.2 × 107 M-1 s-1; k7, 6.0 × 106 M-1 s-1; k8, 4.4 ×
exchange reaction rate constants, K12 is the equilibrium constant
for the cross reaction, f12 is a known function of these
parameters, ln f12 )((ln K12)2)/(4 ln ((k11k22)/Z2)), and Z is the
collision frequency, estimated to be 1011 M-1 s-1
.
To solve for the self-exchange rate constants of the cobalt
redox pairs, the specific rate for [Ru(bpy)3]3+/2+ self-exchange
was taken from literature data. There are two generally accepted
values for this rate constant; 2.0 × 109 M-1 s-1 at 25 °C in 1
M HClO4(aq) reported by Meyer and co-workers14 and 8.3 ×
106 M-1 s-1 at 25 °C in CH3CN obtained by Chan and Wahl.15
Temperature dependence measurements allow us to extrapolate
the latter value to 7.3 × 106 M-1 s-1 at 20 °C. Using these rate
constants as limiting values, we estimated ranges for the self-
exchange rate constants for CoIII-CoII and CoII-CoI.
3. Discussion
TA spectroscopy demonstrates that MV2+ oxidatively quenches
pulsed-laser excited [Ru(bpy)3]2+ to form MV•+, which delivers
an electron to CoII, producing anionic CoI and [Ru(bpy)3]3+
.
[Ru(bpy)3]3+ oxidizes both CoI and CoII, and the CoIII, although
formed irreversibly on the electrochemical time scale (possibly
owing to changes in ligation),6 is reduced back to CoII by the
remaining reductants in solution, CoI and MV•+. Global analysis
supports these assignments, as the TA spectra can be modeled
as linear combinations of steady-state absorption spectra of the
starting compounds and the transiently formed ET species.
Kinetics traces (Figure 3) measured at 730 nm provide informa-
tion about ET in the system. The rate law for the coupled series
of ET reactions (Schemes 1 and S1, Supporting Information)
was solved numerically to produce time-dependent concentration
profiles for all species which, with molar extinction coefficients
and selected independently determined rate constants, allowed
us to simulate the observed kinetics and extract elementary rate
constants. The specific rate for CoII reduction by MV•+ is 2.0
× 108 M-1 s-1 and those for CoI oxidations by [Ru(bpy)3]3+
The cross reaction of [Ru(bpy)3]3+ + MV•+ f [Ru(bpy)3]2+
+ MV2+ (k3) gave self-exchange rate constants for MV2+/•+ (6.2
× 102 and 1.7 × 105 M-1 s-1) that were 3-6 orders of
magnitude smaller than the literature value (5.4 × 108 M-1
s-1).16 The deviation can be attributed to the fact that the Marcus
cross relation does not take into account inverted driving force
effects or diffusion limits. For later calculations, k11 for
MV2+/•+ was assigned the literature value.16 The self-exchange
rate constants calculated for CoIII-CoII and CoII-CoI are set out
in Table 1. The reorganization parameter, λ, is estimated from self-
exchange rate constants via k11 ) Z e((-λ )/(4RT)), where R is the
11
molar gas constant, and T is temperature.
2.3. Thermodynamic Analysis. The driving forces (-∆Go)
for the elementary steps associated with H2 evolution were
calculated as described in Figure 4 and Table S3, Supporting
Information. The driving force for proton reduction (-∆GoR) is
the difference between the standard potential for hydrogen
evolution by proton donor HA17 (Eo(HA/H2)) and the reduction
potential of the reducing equivalent, D.
and CoIII are 9.8 × 109 M-1 s-1 and 4.4 × 108 M-1 s-1
,
respectively. Rate constants for two other ET reactions were
extracted from the simulations: CoII oxidation by [Ru(bpy)3]3+
(5.2 × 107 M-1 s-1); and CoIII reduction by MV•+ (6.0 × 106
M-1 s-1).
The specific rates of CoIII-CoII and CoII-CoI electron self-
exchange reactions can be estimated using the Marcus cross
relation.13 Taking the rate constants for reductions of [Ru-
(bpy)3]3+ by CoII and CoIII by MV•+, we estimate that the
CoIII-CoII rate constant is between 9.5 × 10-8 and 2.6 × 10-5
M-1 s-1. This value is lower than that estimated for
[CoIII/II(dmgBF2)2(H2O)2]+/0 in aqueous solution (1.7 × 10-4
The driving force for reduction of CoII by reductant D is
described by -∆G1o and -(∆G2o - ∆Go1). The driving force for
protonation of CoI (-(∆G2oa - ∆Go1), -(∆G3o - ∆Go2), and
-(∆Go4 - ∆G3o)) is proportional to the pKa difference between
(14) Young, R. C.; Keene, F. R.; Meyer, T. J. J. Am. Chem. Soc. 1977, 99,
2468–2473.
(15) Chan, M.-S.; Wahl, A. C. J. Phys. Chem. 1978, 82, 2542–2549.
(16) Rieger, A. L.; Rieger, P. H. J. Phys. Chem. 1984, 88, 5845–5851.
(17) Felton, G. A. N.; Glass, R. S.; Lichtenberger, D. L.; Evans, D. H.
Inorg. Chem. 2006, 45, 9181–9184.
9
1062 J. AM. CHEM. SOC. VOL. 132, NO. 3, 2010

Electron Transfer Reactions of Cobalt Diglyoximes
A R T I C L E S
Table 1. Reaction Parameters for Co(dpgBF2)2(CH3CN)2
cross
reaction
k11 reaction
k22 reaction
∆E (V)a
K12
k12 (M-1 s-1
)
k11 (M-1 s-1
)
k22 (calcd) (M-1 s-1
)
λ (eV)
b
c
d
3+/2+
6
Ru(bpy)3
MV2+/·+
Ru(bpy)3
MV2+/·+
CoIII/CoII
0.97
4.74 × 1016
5.2 × 107
6.0 × 106
5.7 × 109
2.0 × 108
2.0 × 109
7.3 × 106
5.4 × 108
9.5 × 10-8
2.6 × 10-5
9.3 × 10-7
AVg.
4.19
3.60
3.97
3.9 ( 0.3
1.91
7
3
4
CoIII/CoII
MV2+/·+
CoII/CoI
0.76
1.73
0.18
1.16 × 1013
5.52 × 1029
1.24 × 103
b
c
d
2.0 × 109
7.3 × 106
5.4 × 108
6.2 × 102
3+/2+
1.7 × 105
1.34
e
1.2 × 105
1.38f
e
a See Supporting Information. b Reference 14. c Reference 15. d Reference 16 Estimated uncertainty ( 0.5 M-1 s-1
.
f Estimated uncertainty ( 0.05
eV.
Figure 4. Thermodynamic analysis of the homolytic and heterolytic pathways for hydrogen evolution. D is a reductant. Values for barriers and driving
forces are given in Table S3, Supporting Information.
M-1 s-1 to 8.7 × 10-3 M-1 s-1).18 Corresponding self-exchange
rate constants for [CoIII/II(N4)(H2O)2]3+/2+ complexes with
equatorial tetra-aza macrocylic ligands (4.5 × 10-5 M-1 s-1 to
5.0 × 10-2 M-1 s-1) correlate with structural changes, particu-
larly variations in axial Co-OH2 bond lengths.19 The rate
constant estimated for CoIII-CoII is near the low end of the
range for low-spin CoIII/II complexes in aqueous solution.
18 M-1 s-1; [CoII/I(bpy)3]2+/+, ∼109 M-1 s-1).20 The reactivity
differences are attributable to the substantial nuclear configu-
ration changes that accompany [CoIII/II(bpy)3]3+/2+ exchange (∆d0
for Co-N bonds ) 0.19 Å), and the modest rearrangements
upon [CoII(bpy)3]2+ reduction to [CoI(bpy)3]+ (∆d0 ) -0.02
Å). It is interesting to note that a pronounced change is found
in the solid-state structures of CoII and CoI: an axial CH3CN
ligand is lost upon reduction, accompanied by a 0.268 Å
shortening of the remaining nitrile bond, 0.27 Å displacement
of the CoI ion, and a 0.035 Å reduction of the Co-Nimine
distance.6 These considerable structural changes are at variance
with the large CoII-CoI electron self-exchange rate constants
extracted using the cross relation. The discrepancy may be an
indication that electron transfer and the large structural rear-
rangements occur in separate elementary reaction steps. The
large nuclear rearrangements expected for oxidation of CoII to
CoIII, however, are reflected in the CoIII-CoII self-exchange
rate constant.18
Using the rate constants for CoII reduction by MV•+, we can
estimate the CoII-CoI self-exchange rate constant to be 1.2
((0.5) × 105 M-1 s-1. While we determined the rate constant
for CoI oxidation by [Ru(bpy)3]3+, the large driving force for
this reaction (1.55 eV) places it in the Marcus inverted region
where calculation of self-exchange rate constants is unreliable.
The stark difference between the CoIII-CoII and CoII-CoI rates
is similar to that found for [CoII(bpy)3]2+ ([CoIII/II(bpy)3]3+/2+
,
(18) Wangila, G. W.; Jordan, R. B. Inorg. Chim. Acta 2005, 358, 2804–
2812.
(19) Endicott, J. F.; Durham, B.; Glick, M. D.; Anderson, T. J.; Kuszaj,
J. M.; Schmonsees, W. G.; Balakrishnan, K. P. J. Am. Chem. Soc.
1981, 103, 1431–1440.
(20) Szalda, D. J.; Creutz, C.; Mahajan, D.; Sutin, N. Inorg. Chem. 1983,
22, 2372–2379.
9
J. AM. CHEM. SOC. VOL. 132, NO. 3, 2010 1063

A R T I C L E S
Dempsey et al.
as received. Elemental analysis was performed by Columbia
Analytical Services, Tuscon, AZ.
The reorganization parameter, λ, which accounts for both
inner- and outer-sphere rearrangements accompanying an ET
process, can be estimated from self-exchange rates. The
reorganization energy for CoIII-CoII self-exchange is 3.9 ((0.3)
eV (89.9 ((6.9) kcal mol-1), while that for the CoII-CoI
reaction is substantially smaller, about 1.4 ((0.05) eV (34.6
((1.2) kcal mol-1). These calculations emphasize the formidable
energy barriers associated with a hydrogen-evolving mechanism
that requires a CoIII intermediate.
All samples were prepared in high purity (>99.99%) CH3CN
(EMD) in a high-vacuum 1-cm path length fused quartz cell (Starna
Cells) connected to a 10 mL bulb. A typical sample was prepared
as follows. In an inert atmosphere glovebox, a 2 mL solution of
∼126-314 µM Co(dpgBF2)2(MeCN)2 and a 2 mL solution of ∼40
µM [Ru(bpy)3][PF6]2, 7 mM [MV][PF6]2, and 0.2 M NBu4PF6 in
dry, degassed CH3CN were placed into the cell and isolated from
atmosphere and the bulb by a high-vacuum Teflon valve (Kontes).
Co(dpgBF2)2(CH3CN)2 was synthesized under an inert atmo-
sphere with degassed solvents according to a literature method,25
recrystallized in CH3CN, and analyzed by cyclic voltammetry,
absorption measurements, and elemental analysis. The hexafluo-
The large reorganization energy associated with CoIII-CoII
electron transfer implies that the barrier to reduction of CoIII
by D (∆GqET2) is about 2.1 kcal mol-1 higher than that of CoII
(∆GqET1), corresponding to a ∼35-fold lower specific rate. The
largest barriers to catalysis, then, likely will be associated with
the elementary steps that form H2.21 Protonation of CoIIIH leads
to release of H2 and generation of CoIII in the heterolytic
rophosphate salts of [Ru(bpy)3]2+ and methyl viologen ([MV]2+
)
were prepared from the corresponding chloride salts via salt
metathesis with NH4PF6 in water (0 °C).
Na[Co(dpgBF2)2(CH3CN)] was prepared by reaction of
Co(dpgBF2)2(CH3CN)2 with excess sodium mercury amalgam
(0.5% Na) under vacuum in anhydrous acetonitrile, analogous to a
literature preparation.6 The reduced species was transferred over a
frit to a fused 2-mm quartz cuvette; the reduction process was
monitored via UV-vis absorption spectroscopy until the reduction
was complete.
pathway, and the driving force for this reaction (-∆GoH , which
2a
depends on Eo(CoIII/II) (relative to Eo(HA/H2) is in most cases
thermodynamically unfavorable. Thus, the barrier for this
elementary step (∆GqH ) is estimated to be greater than 11 kcal
mol-1, the lower limit2afor∆GoH . Little is known about the rate
2a
of bimolecular reductive elimination of H2 from metal hydrides
(homolytic path). Under the assumption of rapid electron and
proton transfers such that CoII, CoI, and CoIIIH will be present
at their equilibrium concentrations, the rate of H2 formation
(d[H2]/dt) will be proportional to kH [CoIIIH]2 for the homolytic
pathway and to kH [HA][CoIIIH] 2for the heterolytic route.
Ultimately, the dom2ianance of one path over the other during
catalysis depends not just on the relative barrier heights for H2
elimination but also on the relative concentrations of HA and
CoIIIH.22
Methyl viologen cation radical (MV•+) was prepared in situ by
the photolysis (Hg lamp) of a degassed sample containing 10 µM
[Ru(bpy)3][PF6]2, 84 µM [MV][PF6]2, 0.1 M NBu4PF6, and 0.1 M
triethanolamine in CH3CN for 2 h. The resulting absorption
spectrum matched the literature spectrum.26
[Ru(bpy)3]3+ was prepared in situ by oxidizing [Ru(bpy)3]2+ with
ammonium cerium nitrate (CeIV) according to a literature method.27
MeCo(dpgBF2)2(L) (L ) H2O, CH3CN) was prepared by a
method similar to that of Ram et al.28 for MeCo(dmgBF2)2(H2O).
Co(dpgBF2)2(CH3CN)2 (0.505 g, 0.724 mmol) was suspended under
argon in 22 mL of degassed CH3OH in a two-neck 100 mL round-
bottom flask equipped with a septum, vacuum adaptor, and stirbar.
NaOH (0.091 g, 2.3 mmol) was added, and the solution was stirred
until the NaOH dissolved, followed by addition of pyridine (0.058
g, 0.733 mmol). After cooling the suspension to 0 °C, NaBH4 (0.046
g, 1.21 mmol) was added, the reaction was stirred 15 min, and the
suspension turned a blue-gray color. Methyl trifluoromethane-
sulfonate (1 g, 6.1 mmol) was added via syringe, and the suspension
turned yellow-brown over 10 min. The reaction was then exposed
to the atmosphere, at which time 1 mL of pyridine and 40 mL of
H2O were added 5 min apart. After 10 min of stirring, the product
was filtered to yield a yellow-brown solid, and this was washed
with 3 × 40 mL of H2O. Then the product was collected, placed
in a round-bottom flask, suspended in 30 mL of 6 M HClO4 and
stirred for 20 min to remove pyridine. The suspension was then
filtered to yield a yellow-brown solid and washed with 2 × 30 mL
of HClO4 followed by copious amounts of H2O and 30 mL of
hexanes. The yellow-brown solid was collected (315 mg, 65%).
Vapor diffusion of ether into a saturated CH3CN solution of
MeCo(dmgBF2)2(CH3CN) afforded a yellow powder. 1H NMR (300
MHz, CD3CN): δ 7.28-7.48 (m, 20H, Ph), δ 1.87 (s, 3H, CH3)
ppm. HRMS (FAB+), m/z calculated for C31H26B2F4N5O4Co (L )
CH3CN): 689.14. Found: 690.1559 (M + H), 628.7951 (M - C6H5),
565.7864 (M - C6H5 - BF2 + H), 513.9995 (M - C6H5 - 2
BF2), 485.9959 (M - M - C6H5 - 2 BF2 - 2 CH3). Elemental
analysis calculated for C29H25B2F4N4O5Co (L ) H2O): C, 52.29;
H, 3.78; N, 8.41. Found: C, 52.62; H, 3.95; N, 8.31.
4. Conclusions
Thermodynamic analysis of two proton reduction pathways
reveals that the driving force for H2 production in the homolytic
pathway depends on Eo(CoII/I) (relative to Eo(HA/H2)) and the
difference in pKa between CoIIIH and HA. In the heterolytic
pathway, Eo(CoIII/II) is an additional determinant that, in most
cases, renders heterolytic H2 evolution extremely unfavorable.23
Furthermore, the barrier associated with reduction of transiently
generated CoIII is significantly higher than that for CoII.
Although the heterolytic route can dominate at very high acid
concentrations, the relatively high energy barriers and unfavor-
able driving forces are significant. While little is known about
the barrier to the CoIIIH bimolecular reaction, covalently linking
two CoIIIH complexes could substantially increase the rate of
H2 production by decreasing the volume required for diffusional
collisions.
5. Experimental Section
5.1. Reagents. Syntheses of air- and moisture-sensitive com-
pounds were carried out using Schlenk techniques or in a nitrogen
atmosphere glovebox. Solvents for these syntheses were dried by
a standard method24 or over activated sieves followed by passage
over activated alumina. CD3CN was obtained from Cambridge
Isotope Laboratories, Inc. All materials, unless noted, were used
(25) Tovrog, B. S.; Kitko, D. J.; Drago, R. S. J. Am. Chem. Soc. 1976, 98,
5144–5153.
(21) Bhugun, I.; Lexa, D.; Save´ant, J.-M. J. Am. Chem. Soc. 1996, 118,
3982–3983.
(26) Kosower, E. M.; Cotter, J. L. J. Am. Chem. Soc. 1964, 86, 5524–
5527.
(22) Chao, T.-H.; Espenson, J. H. J. Am. Chem. Soc. 1978, 100, 129–133.
(23) Kellett, R. M.; Spiro, T. G. Inorg. Chem. 1985, 24, 2373–2377.
(24) Pangborn, A. B.; Giardello, M. A.; Grubbs, R. H.; Rosen, R. K.;
Timmers, F. J. Organometallics 1996, 15, 1518–1520.
(27) Bryant, G. M.; Fergusson, J. E. Aust. J. Chem. 1971, 24, 275–286.
(28) Ram, M. S.; Riordan, C. G.; Yap, G. P. A.; Liable-Sands, L.;
Rheingold, A. L.; Marchaj, A.; Norton, J. R. J. Am. Chem. Soc. 1997,
119, 1648–1655.
9
1064 J. AM. CHEM. SOC. VOL. 132, NO. 3, 2010

Electron Transfer Reactions of Cobalt Diglyoximes
A R T I C L E S
5.2. Physical Methods. NMR spectra were recorded using a
Varian Mercury 300 spectrometer. H NMR chemical shifts were
beam was collinear with the probe light. Light intensity was read
by two photodiode arrays (Ocean Optics S1024DW Deep Well
Spectrometer), with scattered excitation light rejected by a 370 nm
long pass filter. The timing synchronization of the laser fire,
flashlamp fire, and photodiode array readout were controlled by a
series of timing circuits triggered by either a Q-switch advance
logic pulse for nanosecond (or a laser lamp sync pulse for
microsecond) lamp measurements. The photodiode readout was
interfaced with a PC via a National Instruments multifunction input/
output card. Measurements were made with and without excitation,
corrected for background light, and corrected for fluorescence when
necessary. Difference spectra were averaged over approximately
500 shots.
All instruments and electronics in these systems were controlled
by software written in LabVIEW (National Instruments). Data
manipulation was performed with MATLAB R2008a (Mathworks,
Inc.) and graphed with Igor Pro 5.01 (Wavemetrics).
5.3. Data Analysis. Global analysis and kinetics simulations
were performed in MATLAB, the latter using the ordinary
differential equation solver ode23s (Rosenbrock method) and an
m-file with the differential equations. [Ru(bpy)3]2+* luminesces
weakly at the wavelength the kinetics traces were measured.
Because of the finite amplifier response for the longer time scales,
the first few microseconds of data were obscured by this emission
process.
1
referenced to residual solvents as determined relative to Me4Si (δ
) 0 ppm). High resolution mass spectra (HRMS) were obtained at
the California Institute of Technology Mass Spectrometry Facility.
UV-visible absorption measurements were carried out using a
Hewlett-Packard 8452 spectrophotometer in 0.2 or 1 cm path length
quartz cuvettes.
Time-resolved spectroscopic measurements were carried out at
the Beckman Institute Laser Resource Center. Laser excitation was
provided by 8-ns pulses from a 10 Hz Q-switched Nd:YAG laser
(Spectra-Physics Quanta-Ray PRO-Series). The third harmonic was
either used directly or to pump an optical parametric oscillator
(OPO, Spectra-Physics Quanta-Ray MOPO-700, tunable in the
visible region).
Probe light for transient absorption kinetics measurements was
provided by a 75-W arc lamp (PTI model A 1010) that could be
operated in continuous wave or pulsed modes. Timing between the
laser and the probe light was controlled by a digital delay generator
(EG&G 9650). Sample excitation (λ ) 480 nm) by the laser beam
was collinear with the probe light; scattered excitation light was
rejected by suitable long pass and short pass filters, and probe
wavelengths were selected for detection by a double monochromator
(Instruments SA DH-10) with 1 mm slits. Transmitted light was
detected with a photomultiplier tube (PMT, Hamamatsu R928). The
PMT current was amplified and recorded with a transient digitizer
(LeCroy 9354A or Tektronix DSA 602). Absorption data were
averaged over at least 500 laser pulses. The data were converted
to units of ∆OD (∆OD ) -log10(I/I0), where I is the time-resolved
probe-light intensity with laser excitation, and I0 is the intensity
without excitation). Samples measured on the microsecond time
scale or faster were stirred and measured using a laser repetition
rate of 10 Hz, while samples measured on a millisecond time scale
were excited with a single shutter-released laser pulse, stirred for
1 s after collecting data, then allowed to sit until the solution settled
(2 s) before the next laser pulse. Data were averaged over
approximately 500 shots.
Acknowledgment. We thank Bruce Brunschwig, Xile Hu, Jay
Labinger, and Jonas Peters for insightful discussions. We also thank
Etsuko Fujita for generous assistance in obtaining in situ absorption
spectra. This work was supported by the NSF Center for Chemical
Innovation (Powering the Planet, CHE-0802907 and CHE-
0947829), the Arnold and Mabel Beckman Foundation, CCSER
(Gordon and Betty Moore Foundation), and the BP MC2 program.
JLD is an NSF Graduate Research Fellow.
Supporting Information Available: Full details of absorption
measurements, global analysis, kinetics modeling, reduction
potentials, and thermodynamic analysis. This information is
available free of charge via the Internet at http://pubs.acs.org.
Probe light for transient absorption spectra was provided by white
light flash lamp sources with either nanosecond or microsecond
durations. Probe light was transported via a fiber optic and split by
a partial reflector. Approximately 70% of the probe light was passed
through the sample, with the remainder directed around the sample
as a reference beam. Sample excitation (λex ) 355 nm) by the laser
JA9080259
9
J. AM. CHEM. SOC. VOL. 132, NO. 3, 2010 1065