Highly functionalizable penta-coordinate iron hydrogen production catalysts with low overpotentials

Article abstract of DOI:10.1039/c5dt03744a

Penta-coordinate iron carbonyl complexes that are built around the rigid Fe(PNP) motif (PNP = (C₆H₅)₂PN(R)P(C₆H₅)₂) are synthesized and structurally and spectroscopically characterized. The

Full text of DOI:10.1039/c5dt03744a

View Article Online
View Journal
Dalton
Transactions
Accepted Manuscript
This article can be cited before page numbers have been issued, to do this please use: S. Eady, T. Breault,
L. Thompson and N. Lehnert, Dalton Trans., 2015, DOI: 10.1039/C5DT03744A.
This is an Accepted Manuscript, which has been through the
Royal Society of Chemistry peer review process and has been
accepted for publication.
Accepted Manuscripts are published online shortly after
acceptance, before technical editing, formatting and proof reading.
Using this free service, authors can make their results available
to the community, in citable form, before we publish the edited
article. We will replace this Accepted Manuscript with the edited
and formatted Advance Article as soon as it is available.
You can find more information about Accepted Manuscripts in the
Information for Authors.
Please note that technical editing may introduce minor changes
to the text and/or graphics, which may alter content. The journal’s
standard Terms & Conditions and the Ethical guidelines still
apply. In no event shall the Royal Society of Chemistry be held
responsible for any errors or omissions in this Accepted Manuscript
or any consequences arising from the use of any information it
contains.
www.rsc.org/dalton

Page 1 of 36
Dalton Transactions
View Article Online
DOI: 10.1039/C5DT03744A
Highly Functionalizable PentaꢀCoordinate Iron Hydrogen
Production Catalysts with Low Overpotentials
Shawn C. Eady†, Tanya Breault‡, Levi Thompson‡, Nicolai Lehnert†*
†Department of Chemistry, University of Michigan, 930 North University, Ann Arbor, Michigan
48109ꢀ1055, United States
‡Department of Chemical Engineering, University of Michigan, 1221 Beal Avenue, Ann Arbor,
Michigan 48109ꢀ2102, United States
KEYWORDS: Sustainable Chemistry, Electrocatalysis, Hydrogen Fuel
ABSTRACT Pentaꢀcoordinate iron carbonyl complexes that are built around the rigid Fe(PNP)
motif (PNP = (C6H5)2PN(R)P(C6H5)2) are synthesized and structurally and spectroscopically
characterized. These complexes allow for facile customization of the secondary ligand sphere
with various types of linkers and functional groups. The new [Fe(S2C6H4)(PNP)(CO)] complexes
show dihydrogen production electrocatalysis at overpotentials of 0.09ꢀ0.21 V vs. Pt under mildly
acidic conditions (and activities from 0.28 to 3.51 sꢀ1). The most active compound exhibits a
turnover frequency (TOF) of 3.51 sꢀ1 at an overpotential of only 0.15 V vs. Pt. Trends in activity
are further analyzed. It is found that (a) a decrease in overpotential correlates strongly with the
electron withdrawing strength of the PNP amine substituent, and (b) aliphatic substituents give
comparatively higher TOFs than aromatic ones. An ECꢀtype mechanism is shown to proceed by
initial reduction of the FeII metal center and proposed formation of an FeIII hydride intermediate.
Analogous cobalt compounds were synthesized and characterized, but were found to have low
1

Dalton Transactions
Page 2 of 36
View Article Online
DOI: 10.1039/C5DT03744A
stability in acidic media under turnover conditions, and hence, are unsuitable as catalysts for
proton reduction.
Introduction
The search for sustainable dihydrogen production systems has been an ongoing focus of global
research efforts for nearly half a century.1ꢀ3 Incorporation of nonꢀprecious metal catalysts to yield
efficient dihydrogen production manifolds in aqueous systems is an everꢀgrowing area in the
field of electrocatalysis.4,5 Major discoveries in proton reduction electrochemistry have been
made in the past few decades, which suggests that nonꢀprecious metals such as iron, nickel and
cobalt with selected ligand manifolds and the correct environment may provide activities and
stabilities competitive to platinum and similar rare metal catalysts.6ꢀ23 11,24ꢀ41
A major challenge in the design of industrially viable electrocatalysts is minimizing the wasted
input energy, i.e. overpotential, required for catalyst activation. A number of ironꢀ and nickelꢀ
centered electrocatalysts for dihydrogen production with incredible activities (>10,000 sꢀ1) have
been reported in the literature; however, many of these are only active in strongly acidic
conditions or have moderate to high overpotentials, typically in excess of 200 mV.6,42 It is
therefore the focus of this study to investigate catalytic systems capable of functioning in weakly
acidic media with minimal overpotentials.
In addition, to transition from basic scientific research in catalyst development to applicationꢀ
driven research, easy access to a variety of functionalities in the ligand periphery of a catalyst is
desirable.43ꢀ45 One such application is the heterogenization of catalysts via immobilization on
electrode surfaces, especially in lightꢀdriven proton reduction systems that couple water
2

Page 3 of 36
Dalton Transactions
View Article Online
DOI: 10.1039/C5DT03744A
oxidation with proton reduction for efficient water splitting.12,29,41,44ꢀ51 Realization of such
systems is met with formidable obstacles, including a means of providing a versatile ligand
system that will not impact catalyst ‘core’ structure and function upon derivatization and surface
immobilization. Thus it can be reasoned that the development of suitable strategies and ligand
systems for catalyst immobilization is as important as the catalyst design itself, and having a
catalyst system with an easily customizable ligand periphery is critical for future applications.
Previously, pentaꢀcoordinate iron catalysts that are active for proton reduction have been
reported by Ott and coworkers.37 These compounds display impressive activity (TOF > 500 sꢀ1)
with modest overpotentials (0.17ꢀ0.24 V vs. Pt for acetic acid concentrations of 0.1ꢀ0.5 M,
respectively) and were obtained in high yields in a single synthetic step from diphosphine and
dithiolate ligands under atmospheric CO pressure. However, these complexes utilize a
conformationally flexible diphosphine ligand, which, when locked into a fixed conformation in
the surfaceꢀbound state, might become less active.44 Herein we report catalytically active
derivatives with a much more rigid, conjugated ‘PNP’ ligand structure, leading to iron
complexes that contain a very rigid, FeꢀPꢀNꢀP fourꢀmembered ring. Several derivatives of these
complexes are presented and are characterized structurally as well as spectroscopically.
Electrochemical properties of the complexes are further investigated, in addition to their
activities for proton reduction. These systems provide a versatile, yet simple ‘toolbox’ for ligand
derivatization which can be tailored to a variety of applications.
Scheme 1. Preparation of pentaꢀcoordinate iron dihydrogen production catalysts.
Experimental Section
3

Dalton Transactions
Page 4 of 36
View Article Online
DOI: 10.1039/C5DT03744A
General Procedures
Unless otherwise stated, all syntheses were performed under a dinitrogen atmosphere with
distilled and degassed solvents.
Materials. Isopropylamine (99%), triethylamine (>99%), 1ꢀaminoꢀ3ꢀaminobutyne (95%), 1ꢀ
pyrenemethylamine hydrochloride (95%), 1ꢀpyrenebutanol (99%), aniline (99%), 4ꢀbromoaniline
(97%), 4ꢀfluoroaniline (99%), 4ꢀfluorobenzylamine (97%), 4ꢀphenylbutylamine (97%), benzeneꢀ
1,2ꢀdithiol (97%), tolueneꢀ3,4ꢀdithiol (>97%), and 3,6ꢀdichloroꢀ1,2ꢀbenzenedithiol (95%), and
Fe(II) sulfate heptahydrate (>99%) were all purchased from Sigma Aldrich and used as received.
Methylene chloride, acetonitrile and hexane were all purchased from Fisher (ACS grade),
distilled over calcium hydride, and degassed with a dinitrogen purge before use. Methanol was
purchased from Fisher (ACS grade), distilled over magnesium sulfate (Fisher), and degassed
with dinitrogen prior to use.
1
31
19
Physical Measurements. H NMR spectra at 500 MHz, P NMR spectra at 202 and F NMR
1
spectra at 348 MHz were obtained on a Bruker DMX 400 spectrometer. All H chemical shifts
were measured relative to residual protons in the lock solvents and are referenced to Me4Si (0.00
ppm). 31P NMR chemical shifts are referenced to the corresponding proton experiment
31
19
conducted directly prior to the P measurement. F chemical shifts were referenced to a CCl3F
internal standard set to 0 ppm. All mass spectra collected using electrospray ionization (ESI)
techniques were recorded on a Micromass Autospec Spectrometer Premier. Solid state infrared
(IR) spectra were recorded on a Perkin Elmer Spectrum 100 FTꢀIR spectrometer by embedding
the compounds in a KBr matrix. Solution IR studies were conducted in acetonitrile (purified as
described above) and measurements were performed between NaCl plates. Elemental analyses
were performed by Atlantic Microlab, Inc, Norcross, GA.
4

Page 5 of 36
Dalton Transactions
View Article Online
DOI: 10.1039/C5DT03744A
Electrochemistry. Cyclic voltammetry (CV) was performed in a 3ꢀelectrode cell under an argon
atmosphere with a glassy carbon working electrode (A = 0.031 cm2), platinum counter electrode,
and Ag/AgNO3 (0.01 M, MeCN) reference electrode. Acetonitrile used for electrochemical
measurements was purchased from Fisher (ACS grade), distilled over calcium hydride, and
degassed with a dinitrogen purge before use. Ferrocene (Fc, 99%) used as an internal standard
was purchased from Sigma and recrystallized from hexane. All potentials are reported vs. the
Fc0/+ redox couple. Bulk electrolysis studies were performed in a twoꢀcompartment cell
separated by a glass frit. The working and counter electrodes were carbon felt, and the reference
electrode was Ag/AgNO3 (0.01 M, MeCN). Ferrocene was used in the counter electrode
compartment as a sacrificial reductant. Working electrodes for all experiments were polished
with alumina and diamond polish (1 µM), followed by electrochemical stripping in a 0.1 M
(TBA)PF6 acetonitrile solution (500 scans, 1 V/s, 0 to ꢀ2 V vs. Fc0/+).
Overpotential Determination. Overpotential was determined by the exact method used by Ott
and coworkers,37 and is given relative to the corresponding EPt1/2 obtained on a freshly polished
Pt electrode under exactly the same conditions. The data reported in ref. 1 are directly relevant to
the complexes reported here since our electrochemical experiments utilized the same solvent and
acids (in nearly identical concentration ranges), as well as overall similar catalysts. Specifically,
since all catalysts reported here reach activity saturation within roughly 100 equivalents of acetic
acid, using 1 mM catalyst concentrations (i.e. acid concentrations do not exceed 0.2 M), the halfꢀ
wave potential (Ecat) observed for platinum at [AcOH] of 0.2 M by Ott and coworkers, ꢀ1.48 V
vs. Fc0/+, was used to calibrate overpotentials for all catalysts reported here.
Gas Chromatography. Gas chromatography (GC) experiments in combination with bulk
electrolysis studies for the measurement of dihydrogen gas were performed with a SRI 8610C
5

Dalton Transactions
Page 6 of 36
View Article Online
DOI: 10.1039/C5DT03744A
gas chromatograph equipped with a thermal conductivity detector (TCD) and a flame ionization
detector (FID), as well as MS13X (6’) and HayesepꢀD (6’) columns using dinitrogen (99.999%
pure) as the electrochemical cell carrier gas with a flow rate of 3 mL/min. For the bulk
electrolysis experiments, injections into the detector were made every ten minutes and
dihydrogen evolution volumes were quantified for these times. To calculate the amount of
hydrogen produced between the 10 minutes intervals, the rate of hydrogen produced per minute
at the two intervals was averaged ((Rx min + Rx+10 min))/2), and this average rate was used for an
estimate of the produced H2 volume over the ten minute interval (Vinterval = Rxꢀ(x+10) avg (V/min) *
10 min). Cumulative volumes were obtained by taking the sum of the aforementioned volumes
from each time span. Faradaic efficiency was calculated as the molar sum of the evolved
dihydrogen divided by half of the total electron charge passed in the same time frame.
XꢀRay Crystallography. All structural data was collected on a Rigaku AFC10K Saturn 944+
CCDꢀbased Xꢀray diffractometer equipped with a low temperature device and a Micromaxꢀ
007HF Cuꢀtarget microꢀfocus rotating anode (λ = 1.54187 Å) operated at 1.2 kW power (40 kV,
30 mA). The Xꢀray intensities were measured at 85(1) K with the detector placed at a distance
of 42.00 mm from the crystal. Analysis of the data showed negligible decay during data
collection; the data were processed with CrystalClear 2.0 and corrected for absorption. The
structures were solved and refined with the Bruker SHELXTL (version 2008/4) software
package. Details about the structure determinations are provided in the Supporting Information.
Computational Methods. All geometry optimizations and frequency calculations were
performed with the Gaussian 09 program package52 with the BP86 functional and the TZVP
basis set (as implemented in Gaussian 09).
6

Page 7 of 36
Dalton Transactions
View Article Online
DOI: 10.1039/C5DT03744A
General Preparation of diphosphine amine ligands:
Our method is a modification of a similar diphosphine synthesis reported by Imhoff et. al:53
Chlorodiphenylphopshine (1.08 mL, 2 mmol) was added dropwise to a stirring solution of the
selected primary amine (1 mmol) and triethylamine (1.4 mL, 10 mmol) in methylene chloride,
causing a fine white precipitate to form. The mixture was allowed to stir overnight, after which
time solvent was removed in vacuo. The solids were thoroughly washed with methanol (5x, 20
mL) and the filtered white product was dried by vacuum. The product was recrystallized from a
dichloromethane/hexane solution at room temperature.
(C6H5)2PN(iPr)P(C6H5)2
Recrystallization afforded the product as a white solid with a yield of 89%. 1HꢀNMR (400 MHz,
i
i
CDCl3): δH = 7.23ꢀ7.42 (m, 20H, (P(C6H5)2)2), 3.2 (m, 1H, Pr), 0.63 (d, 6H, Pr) ppm. 31P{1H}ꢀ
NMR (202 MHz, CDCl3): δP = 40 ppm (broad).
(C6H5)2PN(C6H5)P(C6H5)2
Recrystallization afforded the product as a white solid with a yield of 85%. 1HꢀNMR (400 MHz,
CD2Cl2): δH = 7.23ꢀ7.4 (m, 20H, (P(C6H5)2)2), 6.94 (m, 3H, NC6H5), 6.64 (d, 2H, NC6H5) ppm.
31P{1H}ꢀNMR (202 MHz, CD2Cl2): δP = 68.5 ppm.
(C6H5)2PN(pꢀC6H4Br)P(C6H5)2
Recrystallization afforded the product as a white solid with a yield of 80%. 1HꢀNMR (500 MHz,
CD2Cl2): δH = 7.2ꢀ7.55 (m, 20H, (P(C6H5)2)2), 6.74 (d, 2H, NC6H4Br), 6.5 (d, 2H, NC6H4Br)
ppm. 31P{1H}ꢀNMR (202 MHz, CD2Cl2): δP = 69.2 ppm.
(C6H5)2PN(pꢀC6H4F)P(C6H5)2
Recrystallization afforded the product as a white solid with a yield of 87%.1HꢀNMR (500 MHz,
CD2Cl2): δH = 7.26ꢀ7.53 (m, 20H, (P(C6H5)2)2), 6.58 (m, 2H, NC6H4F) 6.45 (m, 2H, NC6H4F)
ppm. 31P{1H}ꢀNMR (202 MHz, CD2Cl2): δP = 70.3 ppm. 19FꢀNMR (348 MHz, CD2Cl2): δF =
ꢀ117.4 ppm.
(C6H5)2PN(CH2(pꢀC6H4F))P(C6H5)2
7

Dalton Transactions
Page 8 of 36
View Article Online
DOI: 10.1039/C5DT03744A
Recrystallization afforded the product as a white solid with a yield of 90%. 1HꢀNMR (500 MHz,
CD2Cl2): δH = 7.51ꢀ7.91 (m, 20H, (P(C6H5)2)2), 6.6 (m, 2H, N(CH2)C6H4F), 6.4 (m, 2H,
31
N(CH2)C6H4F), 4.0 (m, 2H, N(CH2)C6H4F) ppm. P{1H}ꢀNMR (202 MHz, CD2Cl2): δP = 72.5
ppm. 19Fꢀ NMR (348 MHz, CD2Cl2): δF = ꢀ112.4 ppm.
(C6H5)2PN((CH2)4C6H5)P(C6H5)2
1
Recrystallization afforded the product as a colorless oil with a yield of 78%. HꢀNMR (500
MHz, CD2Cl2): δH = 7.0ꢀ8.2 (m, 25H, (P(C6H5)2)2 and N((CH2)4C6H5), 2.98 (m, 2H,
N((CH2)4C6H5), 2.6 (m, 2H, N((CH2)4C6H5), 1.54 (m, 2H, N((CH2)4C6H5), 1.13 (m, 2H,
N((CH2)4C6H5) ppm. 31P{1H}ꢀNMR (202 MHz, CD2Cl2): δP = 42.9 ppm.
(C6H5)2PN((CH2)C14H9)P(C6H5)2
Recrystallization afforded the product as a white solid with a yield of 86%. 1H NMR (400 MHz,
CD2Cl2): δH = 7.47ꢀ8.75 (m, 9H, N(CH2)C14H9), 7.2ꢀ7.4 (m, 20H, (P(C6H5)2)2), 5.1 (m, 2H,
(N(CH2)C14H9)) ppm. 31P{1H}ꢀNMR (202 MHz, CD2Cl2): δP = 58.9 ppm.
(C6H5)2PN(pꢀC6H4C≡CH)P(C6H5)2
Recrystallization afforded the product as a beige solid with a yield of 60%. 1HꢀNMR (500 MHz,
CD2Cl2): δH = 7.24ꢀ7.39 (m, 20H, (P(C6H5)2)2), 7.0 (d, 2H, (pꢀHC≡CC6H4)), 6.5 (d, 2H, (pꢀ
HC≡CC6H4)), 3.1 (s, 1H, (pꢀHC≡CC6H4)) ppm. 31P{1H}ꢀNMR (202 MHz, CD2Cl2): δP = 64.3
ppm.
(C6H5)2PN(3ꢀbutyne)P(C6H5)2
Recrystallization afforded the product as a white solid with a yield of 76%. 1H NMR (400 MHz,
CD2Cl2): δH = 7.2ꢀ7.39 (m, 20H, (P(C6H5)2)2), 1.93 (m, 2H, N((CH2)2C≡CH), 1.89 (s, 1H,
31
N((CH2)2C≡CH), 1.51 (m, 2H, N((CH2)2C≡CH) ppm. P{1H} NMR (170 MHz, CD2Cl2): δP =
42.9 ppm.
General preparation of pentaꢀcoordinate iron and cobalt complexes:
Metal compounds were prepared following a procedure initially reported by Takács et al. (see
Scheme 1 above):54 In a large vial, 1,2ꢀbenzenedithiol (0.14 g, 1 mmol) and sodium methoxide
(0.11 g, 2 mmol) were dissolved in 10 mL of methanol. In a Schlenk flask the selected
diphosphine ligand (1 mmol) and ferrous sulfate heptahydrate (0.280 g, 1 mmol) were combined
8

Page 9 of 36
Dalton Transactions
View Article Online
DOI: 10.1039/C5DT03744A
in methanol (30 mL). The Schlenk flask was charged with 1 atm CO pressure, and the
benzenedithiol solution was added dropwise via addition funnel to the mixture while stirring.
Addition of this solution caused an immediate change in color from light yellow to reddishꢀ
brown. The mixture was allowed to stir under CO pressure for 5 hours, after which time the
solvent was removed in vacuo. The resulting solid was washed with methylene chloride and
filtered, the filtrate being collected and reduced in vacuo to a dark redꢀblack solid. Flash column
chromatography of the crude product over neutral silica gel in a 1:1 dichloromethane/hexane
1
solvent mixture yielded complexes determined to be pure by 31P and H NMR in most cases
(exceptions are noted below).
[Fe(S2C6H4)((C6H5)2PN(iPr)P(C6H5)2)CO] (1)
Column chromatography of the product afforded a reddishꢀbrown solid of 1 in 60% yield. Single
crystals suitable for Xꢀray diffraction were obtained by recrystallization in CH2Cl2/hexane
layered solutions with slow evaporation at ꢀ32ºC. EA calc (x0.75 CH2Cl2): C 58.35, H 4.55, N
1
1.95; found: C 59.42, H 4.77, N 2.09. HꢀNMR (400 MHz, CDCl3): δH = 7.13ꢀ8.2 (m, 24H,
i
i
31
(P(C6H5)2)2) and S2C6H4), 3.47 (m, 1H, Pr), 0.62 (d, 6H, Pr) ppm. P{1H}ꢀNMR (202 MHz,
CDCl3): δ = 105.4 ppm; IR (KBr): νmax/cmꢀ1 1932 ν(CꢀO). APSI mass spectrum (positive mode):
m/z = 623.9 (M – CO + H)+.
[Fe(S2C6H4)((C6H5)2PN(C6H5)P(C6H5)2)CO] (2)
Column chromatography of the product afforded a reddishꢀbrown solid of 2 in 65% yield.
Recrystallization of this product in CH2Cl2/hexane layered solutions was required to remove
residual impurities and provide pure compound. EA calc: C 64.82, H 4.26, N 2.04; found: C
1
64.25, H 5.05, N 1.93. HꢀNMR (400 MHz, CD2Cl2): δH = 7.05ꢀ8.16 (m, 24H, (P(C6H5)2)2) and
S2C6H4), 7 (m, 2H, N(C6H5)), 6.6 (m, 2H, N(C6H5)) ppm. 31P{1H}ꢀNMR (170 MHz, CD2Cl2): δP
= 112.7 ppm. IR (KBr): νmax/cmꢀ1 1935 ν(CꢀO).
[Fe(S2C6H4)((C6H5)2PN(pꢀC6H4Br)P(C6H5)2)CO] (3)
Column chromatography of the product afforded a reddishꢀbrown solid of 3 in 40% yield.
Recrystallization of this product in CH2Cl2/hexane layered solutions was required to remove
1
residual impurities and provide pure compound. HꢀNMR (500 MHz, CD2Cl2): δH = 7.15ꢀ8.18
(m, 24H, (P(C6H5)2)2) and S2C6H4), 7.1 (m, 2H, N(pꢀBrC6H5)), 6.4 (m, 2H, N(pꢀBrC6H5)) ppm.
31P{1H}ꢀNMR (202 MHz, CD2Cl2): δP = 113.5 ppm. IR (KBr): νmax/cmꢀ1 1934 ν(CꢀO).
9

Dalton Transactions
Page 10 of 36
View Article Online
DOI: 10.1039/C5DT03744A
[Fe(S2C6H4)((C6H5)2PN(pꢀC6H4F)P(C6H5)2)CO] (4)
Column chromatography of the product afforded a reddishꢀbrown solid of 4 in 60% yield. EA
calc: C 63.17, H 4.01, N 1.99; found: C 62.19, H 4.22, N 1.95. 1HꢀNMR (500 MHz, CD2Cl2):
δH = 7.1ꢀ8.2 (m, 24H, (P(C6H5)2)2) and S2C6H4), 6.74 (m, 2H, N(pꢀFC6H5)), 6.5 (m, 2H, N(pꢀ
FC6H5)) ppm. 31P{1H}ꢀNMR (202 MHz, CD2Cl2): δP = 113.5 ppm. 19Fꢀ NMR (348 MHz,
CD2Cl2): δF = ꢀ114.2 ppm. IR (KBr): νmax/cmꢀ1 1935 ν(CꢀO).
[Fe(S2C6H4)((C6H5)2PN(CH2(pꢀC6H4F))P(C6H5)2)CO] (5)
Column chromatography of the product afforded a reddishꢀbrown solid of 5 in 55% yield. Single
crystals suitable for Xꢀray diffraction were obtained by recrystallization in CH2Cl2/hexane
layered solutions with slow evaporation at ꢀ32ºC. EA calc: C 63.61, H 4.21, N 1.95; found: C
1
63.19, H 4.18, N 1.97. HꢀNMR (500 MHz, CD2Cl2): δH = 7.13ꢀ8.55 (m, 24H, (P(C6H5)2)2) and
S2C6H4), 6.55 (m, 2, N(CH2)C6H4F), 6.45 (m, 2, N(CH2)C6H4F), 4.5 (m, 2, N(CH2)C6H4F);
19
31P{1H}ꢀNMR (202 MHz, CD2Cl2): δP = 111.4 ppm. FꢀNMR (348 MHz, CD2Cl2): δF = ꢀ115.1
ppm. IR (KBr): νmax/cmꢀ1 1931 ν(CꢀO).
[Fe(S2C6H4)((C6H5)2PN((CH2)4C6H5)P(C6H5)2)CO] (6)
Column chromatography of the product afforded a reddishꢀbrown solid of 6 in 67% yield.1Hꢀ
NMR (500 MHz, CD2Cl2): δH = 7.1ꢀ8.1 (m, 27H, (P(C6H5)2)2), N((CH2)4C6H5) and S2C6H4),
6.88 (d, 2H, N((CH2)4C6H5)), 2.94 (m, 2H, N((CH2)4C6H5)), 2.3 (d, 2H, N((CH2)4C6H5)), 2.15
31
(d, 2H, N((CH2)4C6H5)), 1.19 (d, 2H, N((CH2)4C6H5)) ppm. P{1H}ꢀNMR (202 MHz, CDCl3):
δP = 109.4 ppm. IR (KBr): νmax/cmꢀ1 1932 ν(CꢀO).
[Fe(S2C6H4)((C6H5)2PN(CH2C14H10)P(C6H5)2)CO] (7)
1
Column chromatography of the product afforded a reddishꢀbrown solid of 7 in 63% yield. Hꢀ
NMR (400 MHz, CD2Cl2): δH = 6.8ꢀ8.2 (m, 33H, N(CH2)C14H9), (P(C6H5)2)2) and S2C6H4), 5.3
(m, 2H, N(CH2)C14H9) ppm. 31P{1H}ꢀNMR (202 MHz, CD2Cl2): δP = 110.8 ppm. IR (KBr):
νmax/cmꢀ1 1931 ν(CꢀO).
[Fe(S2C6H4)((C6H5)2PN(pꢀC6H4C≡CH)P(C6H5)2)CO] (8)
1
Column chromatography of the product afforded a reddishꢀbrown solid of 8 in 66% yield. Hꢀ
NMR (500 MHz, CD2Cl2): δH = 7.0ꢀ8.2 (m, 24H, (P(C6H5)2)2) and S2C6H4), 6.9 (d, 2H, N(pꢀ
HC≡CC6H4), 6.4 (d, 2H, N(pꢀHC≡CC6H4), 3.2 (d, 1H, N(pꢀHC≡CC6H4) ppm. 31P{1H}ꢀNMR
(202 MHz, CD2Cl2): δP = 113.4 ppm. IR (KBr): νmax/cmꢀ1 1939 ν(CꢀO).
[Fe(S2C6H4)((C6H5)2PN((CH2)2C≡CH)P(C6H5)2)CO] (9)
1
Column chromatography of the product afforded a reddishꢀbrown solid of 9 in 45% yield. Hꢀ
NMR (400 MHz, CDCl3): δH = 7.12ꢀ8.1 (m, 24H, (P(C6H5)2)2) and S2C6H4), 1.93 (m, 2H,
10

Page 11 of 36
Dalton Transactions
View Article Online
DOI: 10.1039/C5DT03744A
31
N((CH2)2C≡CH), 1.89 (s, 1H, N((CH2)2C≡CH), 1.51 (m, 2H, N((CH2)2C≡CH) ppm. P{1H}ꢀ
NMR (202 MHz, CD2Cl2): δP = 109.2 ppm; IR (KBr): νmax/cmꢀ1 1927 ν(CꢀO).
[Co(S2C6H4)((C6H5)2PN(iPr)P(C6H5)2)CO] (10)
A reddishꢀbrown solid with high luster was collected with a crude yield of 70%. Single reddishꢀ
brown crystals confirmed as pure product and suitable for Xꢀray diffraction were obtained by
recrystallization in CH2Cl2/hexane layered solution with slow evaporation at room temperature.
31P{1H} NMR (202 MHz, CD2Cl2): δP = 80.9 ppm; IR (KBr): νmax/cmꢀ1 1988 ν(CꢀO). APSI mass
spectrum (positive mode): m/z = 627.0 (M – CO + H)+.
Results
1. Characterization
In all cases, dropwise addition of the deprotonated dithiolate ligand in methanol to a
suspension of ferrous sulfate and the respective PNP diphosphines in methanol under a carbon
monoxide atmosphere led to a distinct color change of the solution to deep reddishꢀbrown.
Removal of solvent after 5 hours and subsequent extraction of the products into methylene
chloride provided the corresponding iron species in crude yields of 60ꢀ90%.54 Flash
chromatography over neutral silica generally gave the compounds in high purity as assessed by
1H, 31P and 19F NMR (Figures S1ꢀ3), infrared spectroscopy (IR) (Figure S4), and mass
spectrometry (MS). IR spectra of the iron complexes consistently show a single ν(CꢀO) stretch at
frequencies of 1927ꢀ1939 cmꢀ1, indicative of species containing a single carbonyl ligand. Two
intense bands are also present at energies of 1434 and 694 cmꢀ1 that are characteristic of the
phenyl groups of the diphosphine ligand, and that vary by a minimal amount (~1ꢀ2 cmꢀ1) with
11

Dalton Transactions
Page 12 of 36
View Article Online
DOI: 10.1039/C5DT03744A
Table 1. Pentaꢀcoordinate iron compounds prepared here and selected properties
R=
31
19
k
ν(CꢀO)
*
◊
†
cat
ꢀ1
P NMR
F NMR
ꢀ1
i /i
#
‡
ꢀ1 ꢀ1
E
(V)
cat
p
TOF (s )
η (V)
ν
(cm )
δ (ppm)
δ (ppm)
1/2
max
P
F
(M s )
1
2
3
4
1932
105.4
112.7
113.5
113.5
N/A
N/A
-1.69
-1.62
-1.57
-1.61
0.21
0.14
0.09
0.13
2.7
2.9
1.38
1.58
0.96
1.28
120
235
144
301
1935
1934
1935
N/A
2.26
2.61
-120
5
6
1931
1932
111.4
109.4
-115
N/A
-1.66
-1.63
0.18
0.15
4.28
4.32
3.46
3.51
1243
1375
C4H8
7
1931
110.8
N/A
-1.68
0.20
2.45
1.13
197
8
9
1939
1927
113.4
109.2
N/A
N/A
-1.6
0.12
0.15
1.22
2.61
0.28
1.28
8
-1.63
109
* Potentials are reported versus a ferrocene/ferrocenium internal standard (0.64 V vs. NHE)
◊ Overpotentials reported are lowerꢀend estimates vs. Pt with 0.1 M acetic acid in acetonitrile.
† Turnover frequencies were calculated using equation 1 at the E1/2 potential for the corresponding complex
with an [AcOH] of 0.1 M and a catalyst concentration of 1 mM.
‡ Bimolecular rate constants calculated from the slope of the icat vs. [ACOH]1/2 plot.
substitution at the amine position. The 31P NMR spectra of the iron complexes show a single
sharp resonance at 105ꢀ113 ppm depending on the substituents at the PNP amine group (Table
1). The molecular structures of 1 and 5 have been confirmed by Xꢀray crystallography, showing
pentaꢀcoordinate complexes with distorted squareꢀpyramidal geometries and an axial carbonyl
ligand (Figure 1). The FeꢀP and FeꢀS bond lengths observed for 1 and 5 are nearly identical to
the ‘PCNCP’ ligated iron complex, [Fe(S2C6H4)((C6H5)2PCH2N(dep)CH2P(C6H5)2)(CO)],
reported by Ott and coworkers, as well as to the P(FeCp2)P ligated iron complex,
[Fe(S2C6H4)((C6H5)2P(FeCp2) P(C6H5)2)(CO)], reported by Jones and coworkers (Table 2).28,37
12

Page 13 of 36
Dalton Transactions
View Article Online
DOI: 10.1039/C5DT03744A
Figure 1. Crystal structures of [Fe(S2C6H4)((C6H5)2PN(iPr)P(C6H5)2)(CO)] (1) (left) and
[Fe(S2C6H4)((C6H5)2PN(pꢀfluorobenzyl)P(C6H5)2)(CO)] (5) (right) with ellipsoids shown at 50% probability.
Hydrogen atoms are omitted for clarity. Selected bond distances [Å] and angles [degrees]: (1) Fe1ꢀP1
2.203(15), Fe1ꢀP2 2.2(13), Fe1ꢀS1 2.161(13), Fe1ꢀS2 2.189(14), P1ꢀN1 1.71(4), P2ꢀN1 1.703(4), P1ꢀꢀꢀP2
2.577(17), S1ꢀꢀꢀS2 3.082, S1ꢀFe1ꢀS2 90.25(5), P1ꢀFe1ꢀP2 71.65(5), FeꢀPꢀNꢀP torsion 6.74. (5) Fe1ꢀP1
2.215(6), Fe1ꢀP2 2.144(6), Fe1ꢀS1 2.166(6), Fe1ꢀS2 2.21(6), P1ꢀN1 1.718(16), P2ꢀN1 1.708(17), P1ꢀꢀꢀP2
2.583(7), S1ꢀꢀꢀS2 3.096, S1ꢀFe1ꢀS2 90.03(2), P1ꢀFe1ꢀP2 72.65(2), FeꢀPꢀNꢀP torsion 8.19.
The SꢀFeꢀS bond angles also show strong similarities across these compounds, as expected for
the identical dithiolate ligand. In contrast, the PꢀFeꢀP bond angle is significantly larger (by over
15º) in the complex reported by Jones, and even more so (over 17º) in Ott’s compound. Most
notably, the torsion angles of the FeꢀPꢀNꢀP ring in the complexes reported here are very small (<
9º), showing a nearly planar ring structure. In comparison, the FeꢀPꢀNꢀP unit of Ott’s complex
has a torsion angle of 14º, and shows a distinctly nonꢀplanar chair conformation for the whole
FeꢀPꢀCꢀNꢀCꢀP unit (Figure S6). Jones’ compound exhibits a similarly large torsion angle of 18º
for its FeꢀPꢀFeꢀP unit. These differences highlight the more rigid, conjugated structure of the
FePNP ring in the complexes reported here. Comparison of CꢀO stretching frequencies shows
strong similarity for 1 and 5 (ν(CꢀO)MeCN = 1933 and 1931 cmꢀ1, respectively) compared to the
compound reported by Ott and coworkers (ν(CꢀO)MeCN = 1932 cmꢀ1), while both sets of
13

Dalton Transactions
Page 14 of 36
View Article Online
DOI: 10.1039/C5DT03744A
Table 2. Comparison of selected bond lengths [Å] and angles [º] for selected pentaꢀcoordinate iron compounds.
MPNP
torsion
τ
Complex (#)
MꢀS1
MꢀS2
MꢀP1
MꢀP2
PꢀMꢀP
SꢀMꢀS
(0ꢀ1)
1
5
2.161(13)
2.166(6)
2.194(9)
2.189(14)
2.21(6)
2.203(15)
2.215(6)
2.188(9)
2.2(13)
71.65(5)
72.65(2)
72.13(3)
90.25(5)
90.03(2)
91.16(3)
6.74
0.26
0.73
0.03
2.144(6)
2.187(8)
8.19
3.77
10
2.202(8)
Fe(S C H )((C H ) PCH N
2
6
4
6
5 2
2
1*
2.187
2.21
2.2
2.187
2.17
2.208
2.14
2.212
2.17
89.59
80.5
89.32
88.7
14.01
29/58
0.13
0.22
0.72
(dep)CH P(C H ) )(CO)
2
6
5 2
Fe(S C H )((C H ) P(CH
2
2
6
4
6
5 2
Ph
41*
N CH ) P(C H ) )(CO)
2 2
6
5 2
Fe(S C H )((C H ) P
2
6
4
6
5 2
18.2
(FePFeP)
30*
2.177
2.222
2.225
87.49
89.31
(FeCp ) P(C H ) )(CO)
2
6
5 2
*dep: 1,1ꢀdiethoxypropane; FeCp2: bis(cyclopentadienyl)iron or ferrocene; NPh: N(C6H5).
compounds show a somewhat higher ν(CꢀO) than that reported by Jones and coworkers (ν(Cꢀ
O)DCM = 1915 cmꢀ1). Addison’s tau (τ) values (defined as τ = (β − α)/60)55 were determined for
the pentaꢀcoordinate iron complexes as a measure of distortion from an ideal squareꢀpyramidal
geometry (τ = 0) or an ideal trigonalꢀbipyramidal geometry (τ = 1) with S1ꢀFe1ꢀP2 constituting α
and S2ꢀFe1ꢀP1 constituting β for τ = (β – α)/60. Tau values for compounds 1 and 5 are 0.26 and
0.73 respectively, showing a substantial change in solid state geometry despite a minimal change
in the overall PNP structure. While the compounds reported by Ott and coworkers have a squareꢀ
pyramidal structure more closely related to 1, the ferrocene derivatives reported by Jones and
coworkers are observed to have a nearly identical tau value to 5.
Cyclic voltammograms of the pentaꢀcoordinate iron compounds in acetonitrile solution show
reversible redox waves, assigned to the FeII/I couple, at potentials ranging from ꢀ1.57 to ꢀ1.69 V
versus Fc0/+ (Figure 2). All compounds are observed to have ipa/ipc separations of approximately
14

Page 15 of 36
Dalton Transactions
View Article Online
DOI: 10.1039/C5DT03744A
0.07 V, as is observed for the ferrocene internal standard under identical conditions. For most
compounds, the ipa/ipc ratio consistently gave values of 1 ( 0.1) at all scan rates, suggesting
±
chemical reversibility with no decomposition occurring after reduction on the voltammogram
time scale (Figure S11). Complexes 2 and 8 have disproportionately smaller anodic peaks
compared to cathodic peaks (ipa/ipc < 0.9). For several of the compounds, a small decrease in the
ipa/ipc ratio with increasing scan rates is observed, most noticeably for 2 and 8. As it is more
significant at higher scan rates, this decrease in ipa/ipc is likely due to a somewhat slow geometric
reorganization step for these compounds (see DFT results below), resulting in a mixture of
conformations with slightly offset oxidation potentials and a slightly lower ia overall at higher
scan rates. With the exceptions of these complexes, all ic and ia values are seen to increase
proportionally to [scan rate]1/2 (Figure S11), and no significant change in peak potentials (Epa and
Epc) is observed upon varying the scan rate.
2. Electrocatalytic Studies
15

Dalton Transactions
Page 16 of 36
View Article Online
DOI: 10.1039/C5DT03744A
Upon addition of acetic acid (pKaMeCN = 22.3), the cathodic waves of the iron species increased
approximately linearly with acid concentration, which simultaneously occurred with the
subsiding of the corresponding anodic waves. This behavior is indicative of electrocatalytic
proton reduction from acetic acid, and consistently gave halfꢀmaximum wave potential (Ecat)
values that directly coincide with the E1/2 of each iron species. Background reduction of acetic
acid protons by glassy carbon at these potentials and acid concentrations is negligible as shown
in Figure 2. Importantly, activity occurs at potentials that correspond to only very slight
overpotentials for proton reduction, as compared to the same process (acetic acid reduction,
0.1M) by platinum under identical conditions (see Experimental Section).37 A halfꢀwave
potential (Ecat) of only ꢀ1.57 V vs. Fc0/+ is displayed by the compound with the 4ꢀbromophenyl
substituent at the PNP amine (3), corresponding to an overpotential of only 0.09 V vs. platinum,
the lowest for all of the compounds reported here. Discussion of overpotential with respect to the
thermodynamically reversible reduction of acetic acid is provided in the SI, Section 5. The
Figure 2. Left: Cyclic voltammetry of a 2mM solution of 1 in acetonitrile at various scan rates Right: Cyclic
voltammetry of 1 at a scan rate of 100 mV/s with the addition of increasing equivalents of acetic acid. Solutions
were 0.1 M (TBA)PF6 and ferrocene was used as an internal standard. The working electrode was a glassy carbon
disc, the reference was a nonꢀaqueous Ag/AgNO3 (0.01M) electrode, and the auxiliary electrode was a platinum
disc.
16

Page 17 of 36
Dalton Transactions
View Article Online
DOI: 10.1039/C5DT03744A
overpotentials of all compounds are listed in Table 1. It is worth noting that the current increase
for 8 is significantly lower than that for all other compounds, low enough that 8 could be
considered inactive for proton reduction (within error).
The catalytic current increases linearly with catalyst concentration over the investigated range
of catalysts [1.0ꢀ1.75 mM] at acid saturation conditions, indicative of a first order dependence of
the rate on catalyst concentration (Figure S10). With lower acid concentrations, total catalysis is
observed, where activity is limited only by diffusion of substrate (acid) to the electrode and
catalytic activity increases linearly with acid concentration.56 At higher acid concentrations
(exact number of equivalents is dependent on the ligand substitution), catalytic plateau currents
are approximately linear with [AcOH]1/2 (Figure S9) until activity saturation behavior is
observed. Bimolecular rate constants are calculated from the slopes of these plots with equation
1 analogous to the methods used by Ott and coworkers to provide a direct comparison to their
similar catalyst systems:
ꢈ
ꢈ
ꢍ
ꢊ
(1) ꢀꢁ = ꢂꢃꢄꢅꢆꢇ ꢉ ꢋꢌꢄ
√
Here, D is the diffusion coefficient as determined by the Randles Sevcik equation (Equation S3)
to be 1 x 105 (+/ꢀ 0.2) cm2/s, and A is the electrode area (0.031 cm2). Rate constants from
Equation 1 give a range of activities from 8 to 1375 Mꢀ1 sꢀ1 (Table 1).37 At the activity saturation
point where the peak catalytic current (icat) is seen to be independent of acid concentration, rate
estimates for each species can also be obtained from the ratio of the peak catalytic current to the
cathodic current in the absence of acid (equation 2).57
ꢎ
ꢋ
ꢑꢒꢌ
ꢓꢔꢕ
ꢅꢆꢇ
ꢊ
ꢂꢖ
(2)
=
ꢎ
ꢈ.ꢏꢏꢐ
ꢁ
Here, F = Faraday’s constant and ν = scan rate (V/s). This method is used to calculate TOF for
17

Dalton Transactions
Page 18 of 36
View Article Online
DOI: 10.1039/C5DT03744A
the pentaꢀcoordinate species, varying from 0.28 to 3.51 sꢀ1, depending on the ligand substitution
pattern as listed in Table 1 (see Table for TOF measurement conditions).
Bulk electrolysis coupled with gas chromatography (GC) was used to monitor electrocatalytic
dihydrogen production by 5 (1 mM) in acetonitrile solution from 0.04 M acetic acid (Figure 3).
GC confirmed continued catalytic activity for 2 hours to produce a total of 87 micromoles of
dihydrogen with 73% Faradaic efficiency. From this electrolysis data a turnover number (TON)
of 6 has been calculated for 5. After initial activity subsided (t = 2 hours), a second aliquot of
acid was added; however, no significant resurgence in dihydrogen production was observed
beyond control levels, confirming the compound’s activity had not ceased due to acid depletion.
Significant bleaching of the solution had occurred within the initial 2 hours of electrolysis.
Solution IR analysis of the electrolysis solution at this time showed negligible signal intensity at
the carbonyl ligand stretching frequency for 5 (1931 cmꢀ1), indicating the complex had degraded
including loss of CO. 31P NMR confirmed the degradation of 5 with complete loss of the
Figure 3. Left: Gas chromatography dihydrogen production measurement for bulk electrolysis at ꢀ1.7 V vs. Fc0/+
of 5 (1 mM) with acetic acid (40 mM) in a 0.2 M (TBA)PF6 acetonitrile solution. Right: Charge passed during the
electrolysis experiment. 0.1 M (TBA)PF6 and ferrocene was used as an internal standard. The working and
auxiliary electrodes were carbon felt (4 cm2), and the reference was a nonꢀaqueous Ag/AgNO3 (0.01M) electrode.
Ferrocene (0.1 M) was used in the counter compartment as a sacrificial reductant. Arrows indicate addition of
supplemental acid, equivalent to the initial addition.
18

Page 19 of 36
Dalton Transactions
View Article Online
DOI: 10.1039/C5DT03744A
resonance at 111.4 ppm, and decomposition was further evidenced by a change in solution color
from reddishꢀbrown to yellow.
The less than quantitative efficiency is likely due to trace amounts of ferrocenium
hexafluorophosphate (the oxidized form of the sacrificial reductant) leaking through the glass frit
and being reduced at the working electrode. This phenomenon is also observed in the control
experiments, confirming that substantial nonꢀfaradaic charge is being passed in the absence of
catalyst. Additional charge loss could be attributed to a possible reductive decomposition
pathway for these catalyst species.
3. Mechanistic Studies
19

Dalton Transactions
Page 20 of 36
View Article Online
DOI: 10.1039/C5DT03744A
Figure 4. Variation of ν(CꢀO) for [Fe(S2C6H4)((C6H5)2PN(iPr)P(C6H5)2)(CO)] (1) in acetonitrile solution
upon the addition of acetic acid (left) and toluene sulfonic acid (right), monitored by solution IR
spectroscopy.
Monitoring of the CꢀO stretching frequency of 1 in acetonitrile solution by IR spectroscopy
upon the addition of acetic acid (pKaMeCN = 22.3) shows no significant shift of the ν(CꢀO) band
under these conditions (Figure 4). This result rules out chemical alteration of the catalyst by
acetic acid prior to the reduction of the complexes, and is further supported by the fact that the
catalytic onset potential for each iron complex with this acid is nearly identical to its E1/2 value
(FeII/FeI potential). These observations are consistent for all complexes investigated here (1ꢀ9)
that differ in the PNP substitution pattern. The same results were obtained using the significantly
stronger toluenesulfonic acid (TsOH, pKaMeCN = 8.3). These results suggest that the FeII
complexes cannot be protonated by the acids used for the electrocatalytic studies performed here,
and that the initial mechanistic step for our pentaꢀcoordinate iron catalysts is a oneꢀelectron
reduction. To observe intermediates after initial reduction, bulk electrolysis of species 1 was
performed, and the reaction was monitored by solution IR spectroscopy (Figure 5). Upon
applying a potential of ꢀ1.7 vs. Fc+/0, a red shift in the carbonyl ligand stretching frequency of
nearly 100 cmꢀ1 was observed, showing a slightly broader and less intense signal at 1836 cmꢀ1.
20

Page 21 of 36
Dalton Transactions
View Article Online
DOI: 10.1039/C5DT03744A
This shift is indicative of the formation of a reduced FeIꢀCO species, and correlates well with the
reduced [Fe(S2C6H4)((C6H5)2P(CH2NPhCH2)2P(C6H5)2)(CO)] species observed by Ott and
coworkers with a ν(CꢀO) band at 1850 cmꢀ1.40 This species is assumed to be the reactive
intermediate for hydride formation as no further reduction events are evident at this applied
Figure 5. Bulk electrolysis at ꢀ1.8 V vs. Fc+/0 of a 4 mM solution of 1 in acetonitrile with 0.1 M (TBA)PF6
monitored by solution IR spectroscopy. The black trace is the spectrum of the compound prior to passing
charge; the blue trace corresponds to the species formed during bulk electrolysis.
potential (no further carbonyl band shifting is observed with additional electrolysis time). Loss of
the ν(CꢀO) signal at 1836 cmꢀ1 for 1ꢀ was observed only several minutes after the applied
potential was removed, and no substantial recovery of 1 as indicated by the ν(CꢀO) band at 1933
cmꢀ1 was observed, suggesting decomposition by CO loss.
Density Functional Calculations
Given the similarity in overall structure for all [Fe(S2C6H4)(PNP)(CO)] derivatives reported
here, full geometry optimizations of 1 (S = 0) with the BP86 functional and the TZVP basis set
were performed as a representation of all the complexes investigated here. The optimized
structure of 1 overall matched reasonably well with the crystal structure as shown in Table 3,
with an average offset of 0.03 Å for ironꢀligand bond lengths and 0.3 degrees for ironꢀligand
21

Dalton Transactions
Page 22 of 36
View Article Online
DOI: 10.1039/C5DT03744A
bond angles. In contrast, both the FePNP torsion and τ values are somewhat underestimated,
with a predicted torsion angle that is 4 degrees lower than that in the crystal structure and an
estimated τ value of 0.01. Frequency calculations of 1 produce a single carbonyl stretch with
ν(CꢀO) of 1951 cmꢀ1, fairly close to the experimental value of 1933 cmꢀ1.
Calculations on the reduced FeI species 1− (S = 1/2) show a relatively small change in bond
lengths compared to 1, with an increase of 0.06 and 0.08 Å for FeꢀS1 and FeꢀS2, respectively,
and a decrease of only 0.02 Å for FeꢀP1. The P1ꢀFe1ꢀP2 bond angle has a similar change of
approximately 2 degrees. Far more substantial changes in FePNP torsion angle and τ value are
predicted upon reduction, with a 4 degree increase in torsion angle and an increase of τ from 0.01
to 0.7. This represents a substantial change in complex geometry upon reduction from an
essentially ideal square pyramidal structure to a more closely trigonal bipyramidal geometry.
Frequency calculations of 1− produce a single carbonyl stretch with ν(CꢀO) of 1886 cmꢀ1,
substantially higher than the ν(CꢀO) value of 1836 cmꢀ1 observed experimentally upon reduction
Table 3. Select bond lengths [Å] and angles [º] for pentaꢀcoordinate iron species 1 as predicted by DFT calculations
and comparison to crystal structure values. Calculations were performed with BP86/TZVP.
τ
FePNP
torsion
Structure source
FeꢀS1
FeꢀS2
FeꢀP1
FeꢀP2
PꢀFeꢀP
SꢀFeꢀS
(0-1)
Crystal structure of 1
DFT structure of 1
2.16
2.189
2.198
2.286
2.226
2.272
2.203
2.226
2.206
2.152
2.183
2.2
72.66
73.02
72.93
73.35
73.17
90.03
90.28
88.38
91.06
88.73
6.74
0.26
0.01
0.7
2.198
2.255
2.216
2.272
2.238
2.24
2.71
6.26
0.01
2.94
ꢀ
DFT structure of 1
ꢀ
2.164
2.194
0.19
0.03
DFT structure of 1 ꢀ (CO)
ꢀ
DFT structure of 1 (MeCN)
of 1.
22

Page 23 of 36
Dalton Transactions
View Article Online
DOI: 10.1039/C5DT03744A
Loss of the ν(CꢀO) signal for 1− after several minutes, as evidenced by IR spectroscopy,
prompted DFT investigations of the reduced structure after the loss of CO (1− ꢀ (CO), S = 1/2).
The ꢁE for CO loss from the lowꢀspin FeI complex was calculated to be unfavorable by 45.6
kcal/mol, a substantial value for a process seen to be experimentally credible. While this value
supports the slow timeframe of CO loss after reduction observed by IR spectroscopy, it likely
also suggests an alternative chemical transformation prior to CO loss to promote this reaction
and avoid such an energetically unfavorable pathway as direct loss from the reduced, fiveꢀ
coordinate complex.
Cobalt Analogs
23

Dalton Transactions
Page 24 of 36
View Article Online
DOI: 10.1039/C5DT03744A
Figure 6. Crystal structure of [Co(S2C6H4)((C6H5)2PN(iPr)P(C6H5)2)(CO)] (10) with ellipsoids shown at 50%
probability. Hydrogen atoms are omitted for clarity. Selected bond distances [Å] and angles [degrees]: Co1ꢀP1
2.188(9), Co1ꢀP2 2.187(8), Co1ꢀS1 2.194(9), Co1ꢀS2 2.202(8), P1ꢀN1 1.695(2), P2ꢀN1 1.708(2), P1ꢀꢀꢀP2
2.576(10), S1ꢀꢀꢀS2 3.140, S1ꢀCo1ꢀS2 91.16(3), P1ꢀCo1ꢀP2 72.13(3), CoꢀPꢀNꢀP torsion 3.77
The cobalt compound analogous to 1 was prepared following the same synthetic procedure as in
the case of iron, but using a cobaltous sulfate reagent as the metal source. IR spectroscopic
analysis of this compound (crude product) shows a single ν(CꢀO) band at 1988 cmꢀ1 (Figure S5),
suggesting the corresponding cobalt pentaꢀcoordinate compound had been formed with an
analogous ligand environment as in the case of 1. Mass spectrometry confirmed the presence of a
species with m/z of 627, corresponding to the cobalt dithiolate diphosphine complex after the
loss of the carbonyl ligand (as was seen by mass spectrometry analysis for the iron analog).
Single crystals of 10 suitable for diffraction were afforded with some difficulty, and analysis
confirmed the expected fiveꢀcoordinate structure displaying a square pyramidal geometry with
an axial carbonyl ligand, completely analogous to the iron complex (Figure 6).
Given the difficulty in obtaining pure product for 10, electrochemical analysis was performed
on the crude product. Cyclic voltammetry of this mixture showed a main redox feature at ꢀ1.57 V
vs. Fc0/+, while a minor feature attributed to an impurity was evident with an E1/2 of ꢀ1.32 V
(Figure S8). Upon the addition of tributylammonium hexafluorophosphate (Bu3NH PF6), an
increasing cathodic current was observed at ꢀ1.63 V, corresponding to the E1/2 of the more
24

Page 25 of 36
Dalton Transactions
View Article Online
DOI: 10.1039/C5DT03744A
Figure 7. Cyclic voltammetry of 10 (2 mM) at a scan rate of 100 mV/s with the addition of increasing equivalents
of (Bu3NH)BF4. Solutions contained 0.1 M (TBA)PF6 as a supporting electrolyte and ferrocene was used as an
internal standard. The working electrode was a glassy carbon disc, the reference was a nonꢀaqueous Ag/AgNO3
(0.01M) electrode, and the auxiliary electrode was a platinum disc electrode.
prominent redox feature in the CV of crude 10 (Figure 7). Increasing acid concentrations
showed a linear increase in catalytic current; however, multiple features were apparent upon
close analysis of the catalytic wave, suggesting the presence of multiple active species in the
crude product or the availability of multiple mechanistic pathways at more negative potentials. A
substantial cathodic shift of the Ecat was also evident with increasing acid concentration,
indicating the species or mechanistic pathway giving rise to the more anodic portion of the wave
was being altered with additional acid. Through rigorous chromatography and mechanistic
studies (detailed in SI section 4), it was determined that the component of the crude product
responsible for the sustained current response at high acid concentrations was not the pentaꢀ
coordinate cobalt analog, but rather a cobalt bis(dithiolate) byꢀproduct similar to those reported
independently by McNamara et al.17
Discussion
25

Dalton Transactions
Page 26 of 36
View Article Online
DOI: 10.1039/C5DT03744A
Scheme 2. Illustration of the delocalization of the nitrogen lone pair of the PNP ligand, leading to a planar FePNP ring.
In this paper, we report a new series of fiveꢀcoordinate FeIIꢀCO complexes that serve as H2
production catalysts with weak acids at very low overpotentials (relative to the thermodynamic
activation potential for acetic acid). Our pentaꢀcoordinate iron compounds all contain a rigid
Fe(S2C6H4)(PNP) ligand structure and show very similar ν(CꢀO) frequencies, 31P NMR
resonances, and E1/2 values. These results generally suggest that modification of the secondary
ligand sphere of the PNP ligand has a limited effect on the geometrical structure and electron
density at the metal center in the resulting complexes.
The pentaꢀcoordinate iron compound 1 exhibits a distorted squareꢀpyramidal geometry with an
Addison τ value (defined as τ = (β − α)/60)55 of 0.26. The FeꢀPꢀNꢀPꢀ ring is nearly perfectly
planar, confirming the rigid nature of the ligand moiety. This planar geometry around the
nitrogen atom is surprising and is indicative of a conjugated ring structure (see Scheme 2). In
comparison, the structures of the related complexes by Ott and coworkers,
[Fe(S2C6H4)((C6H5)2PCH2N(dep)CH2P(C6H5)2)(CO)],
and
Jones
and
coworkers,
[Fe(S2C6H4)((C6H5)2P(FeCp2)P(C6H5)2)(CO)], show very similar core structures, but nonꢀplanar
rings in the FeꢀPꢀCꢀNꢀCꢀP and FeꢀPꢀCꢀFeꢀCꢀP units, respectively (Figure S6). As identical
dithiolate and CO ligands are used in all three types of compounds, the reduced activity and
lower overpotentials seen in the compounds reported here can be exclusively attributed to the
structural differences from the ‘PNP’ diphosphine ligand used here.28,37
26

Page 27 of 36
Dalton Transactions
View Article Online
DOI: 10.1039/C5DT03744A
Electrochemical analysis of compounds 1ꢀ9 shows that all pentaꢀcoordinate iron complexes
have chemically reversible FeII/FeI couples, with the exception of some irreversible behavior
(ipa/ipc < 1±0.1) observed for complexes 2 and 8 when cycling at higher scan rates (>100 mV/s).
This result indicates that there is a relatively slow geometric reorganization prior to reoxidation
for these complexes. Interestingly, DFT calculations on complex 1 indicate that the
approximately squareꢀplanar Fe(II) complex undergoes a rearrangement to a trigonalꢀ
bipyramidal structure upon reduction. Hence, there might be a more hindered (hence slower)
rearrangement from squareꢀpyramidal to trigonalꢀbipyramidal for 2 and 8 compared to the other
catalysts investigated here, which would explain the somewhat irreducible behavior of these
complexes at higher scan rates. Overpotential values, as determined by comparison to the
potential for proton reduction from acetic acid at a platinum electrode (under identical
conditions), are observed to vary significantly between different amine substitutions for
i
complexes 1ꢀ9. The most notable differences are seen with more electron donating (R = Pr, η =
0.21 – 0.28 V) aliphatic amine substituents and those with more electron withdrawing
substituents (R = pꢀBrC6H4, η = 0.09 – 0.16 V), showing a definitive effect of the PNP ligand
properties on the metal’s redox potential. However, comparison among the complexes with more
similar PNP amine substitutions does not show conclusive trends, such as those for R = C6H4, pꢀ
BrC6H4, and pꢀFC6H4, with lowerꢀestimate η values of 0.14, 0.09, and 0.13 V, respectively. In
comparison, the ‘PCNCP’ꢀtype complexes reported by Ott and coworkers have overpotential
values ranging from 0.17 to 0.25 V for proton reduction from acetic acid (lower end estimates).37
Similarly, the ‘PCFeCP’ꢀtype complex prepared by Jones and coworkers has a reported
overpotential range of 0.17 ꢀ 0.2 V.28 The higher overpotentials seen in the ‘PCNCP’ꢀ and
‘PCFeCP’ꢀtype complexes suggest that the more rigid, conjugated ‘PNP’ꢀtype ligand structure
27

Dalton Transactions
Page 28 of 36
View Article Online
DOI: 10.1039/C5DT03744A
used here is advantageous for lowering the overpotential for proton reduction, and also allows
for a stronger amine substituent effect of the phosphine ligand on the metal’s redox potential.
Variation in electrocatalytic activity for proton reduction is also seen across our series of
compounds, spanning a range of roughly an order of magnitude in TOF with the lowest rate of
0.28 sꢀ1 for 8 and the highest TOF of 3.51 sꢀ1 for 6. Direct comparison of similar substituents
shows substantial differences, such as the nearly 300% increase in TOF from fluorophenyl (4) to
fluorobenzyl (5), in this example with only the addition of a methylene moiety before the phenyl
ring substituent. Similarly, comparing the complexes with the phenyl substituent (2) and the
butylphenyl substituent (6) at the PNP amine shows an increase of more than 200% in TOF (1.58
sꢀ1 for (2) and 3.51 sꢀ1 for (6)), which is again thought to be caused by the aliphatic butyl moiety.
While this effect could be due to the more electronꢀdonating character of the aliphatic substituent
versus the aromatic substituent, the relatively low activity seen in the isopropyl derivative (1,
1.38 sꢀ1) does not follow this trend. Alternatively, this might be better attributed to the steric
flexibility allowed by the alkyl groups and the geometric structures of the FeII complexes, as the
reduction from FeII to FeI very likely causes a change in the geometry of the complexes towards
trigonalꢀbipyramidal (see DFT results above). In this regard, it is interesting to note that complex
1, which is approximately square planar, shows one of the lowest TOF (and kcat) values, whereas
complex 5, which is already close to trigonalꢀbipyramidal in its Fe(II) form, has one of the
highest catalytic activities. This indicates that the structural rearrangement of the complexes after
reduction might be a key factor that determines the electrocatalytic activities of our catalysts. For
comparison,
‘PCFeCP’ꢀtype
complex
reported
by
Jones
and
coworkers,
[Fe(S2C6H4)((C6H5)2P(FeCp2)P(C6H5)2)(CO)], was not seen to reach activity saturation up to a
[AcOH] concentration of 1.6 M in THF, with an estimated TOF of 241 sꢀ1, determined by the
28

Page 29 of 36
Dalton Transactions
View Article Online
DOI: 10.1039/C5DT03744A
same method used here (equation 1; catalyst concentration 0.6 mM) at the peak catalytic
potential of approximately ꢀ1.8 V. Notably, this maximum TOF is nearly two orders of
magnitude larger than that of the most active of our compounds (3.51 sꢀ1 for 6), which reach their
maximum TOF at much lower acid concentrations.28
Bimolecular catalytic rate constants were calculated for 1ꢀ9 based on cyclic voltammetry data
using equation 2 in an analogous manner to Ott and coworkers. These calculated values also vary
widely across the catalyst series by roughly two orders of magnitude, from 8 Mꢀ1 sꢀ1 to 1375 Mꢀ1
sꢀ1 As expected, the general trend in activity across the series of compounds reported here is
.
preserved in both TOF and rate constant estimates. Interestingly, at peak acid concentrations (0.1
M), a TOF of 137 sꢀ1 can be calculated for 6 (equation 1), which is over an order of magnitude
higher than the TOF calculated for 6 by equation 2 (3.51 sꢀ1). It is also of interest that the
estimated TOFs and the bimolecular rate constants do not correlate with the overpotential in
these complexes. The bimolecular rate constants estimated for the corresponding ‘PCNCP’ꢀtype
complex [Fe(S2C6H4)((C6H5)2PCH2N(dep)CH2P(C6H5)2)(CO)] by Ott and coworkers is nearly an
order of magnitude higher (1000 Mꢀ1 sꢀ1) as compared to the average kcat value in our series
(Table 1).37 The exceptions to this trend are compounds 5 and 6, which both show exceptional
kcat values (1243 and 1375 Mꢀ1 sꢀ1, respectively) compared to our other complexes. While these
kcat values exceed those estimated by Ott and coworkers, comparison of the acid titration CV data
shows that our complexes reach activity saturation at much lower acid concentrations (below 0.1
M, with a maximum ic/ip of only about 5), whereas for Ott’s compound no activity saturation is
observed up to an [AcOH] concentration of 0.5 M (resulting in ic/ip of approximately 20, and
TOF = 77 sꢀ1 from equation 1). These results show that the bimolecular rate constants are not an
accurate predictor of maximum catalyst activity, since acidꢀindependent steps in the mechanism
29