Discovery and high-throughput screening of heteroleptic iridium complexes for photoinduced hydrogen production

Article abstract of DOI:10.1021/ja0427101

The catalytic process of photoinduced hydrogen generation via the reduction of water has been investigated. The use of parallel synthetic techniques has facilitated the synthesis of a 32 member library of heteroleptic iridium complexes that was screened, using high-throughput photophysical techniques, to identify six potential photosensitizers for use in catalytic photoinduced hydrogen production. A Pd/Ni thin film hydrogen selective sensor allowed for rapid quantification of hydrogen produced via illumination of aqueous systems of the photosensitizer, tris(2,2′-dipyridyl)dichlorocobalt ([Co(bpy) ₃]Cl₂), and triethanolamine (a sacrificial reductant) with ultra-bright light emitting diodes (LEDs). The use of an 8-well parallel photoreactor expedited the investigation of the hydrogen evolution process and facilitated mechanistic studies. All six compounds investigated produced considerably more hydrogen than commonly utilized photosensitizers and had relative quantum efficiencies of hydrogen production up to 37 times greater than that of Ru(bpy)₃²⁺.

Full text of DOI:10.1021/ja0427101

Published on Web 04/28/2005
Discovery and High-Throughput Screening of Heteroleptic
Iridium Complexes for Photoinduced Hydrogen Production
Jonas I. Goldsmith, William R. Hudson, Michael S. Lowry,
Timothy H. Anderson, and Stefan Bernhard*
Contribution from the Department of Chemistry, Princeton UniVersity,
Princeton, New Jersey 08544
Received December 3, 2004; E-mail: sbernhar@princeton.edu
Abstract: The catalytic process of photoinduced hydrogen generation via the reduction of water has been
investigated. The use of parallel synthetic techniques has facilitated the synthesis of a 32 member library
of heteroleptic iridium complexes that was screened, using high-throughput photophysical techniques, to
identify six potential photosensitizers for use in catalytic photoinduced hydrogen production. A Pd/Ni thin
film hydrogen selective sensor allowed for rapid quantification of hydrogen produced via illumination of
aqueous systems of the photosensitizer, tris(2,2′-dipyridyl)dichlorocobalt ([Co(bpy)3]Cl2), and triethanolamine
(a sacrificial reductant) with ultra-bright light emitting diodes (LEDs). The use of an 8-well parallel photoreactor
expedited the investigation of the hydrogen evolution process and facilitated mechanistic studies. All six
compounds investigated produced considerably more hydrogen than commonly utilized photosensitizers
and had relative quantum efficiencies of hydrogen production up to 37 times greater than that of
2+
Ru(bpy)3
.
photosensitizer Ru(bpy)32+, the electron relay methyl viologen,
and a platinum catalyst.4-7 Others have been successful employ-
Introduction
The challenge of converting solar radiation into conveniently
usable forms of energy is one that has been considered by many
researchers. Simple solar devices focus or collect sunlight,
harnessing radiation in order to heat dwellings or drive turbines.
Photovoltaic cells can convert solar radiation directly to electrical
current, but their high cost makes them economically unviable
at present. A promising alternative, on which a large amount
of research has also been conducted, is the development of
catalytic systems that can harness the energy in sunlight and
use that energy to split water into hydrogen and oxygen. These
systems hold much promise since they allow solar energy to
be converted to hydrogen, a fuel with a high gravimetric energy
density and clean combustion products. While storage and
distribution remain significant factors impeding the implementa-
tion of a “hydrogen economy”, the greater challenge is to devise
a method of hydrogen production that does not depend on the
use of nonrenewable resources.
Catalytic systems that have been shown to successfully effect
photoinduced hydrogen production1-3 typically use transition
metal complexes as photosensitizers to harvest solar radiation
in conjunction with an electron relay which quenches the excited
photosensitizer through electron transfer, creating a reactive
species which will then reduce protons to diatomic hydrogen
gas. Examples of such systems include those based on the
ing Rh(bpy)3 , ,
2+ 8-10 Co(bpy)32+ 11-14 or other cobalt complexes15
in place of the methyl viologen. Systems with a variety of other
photosensitizers have also demonstrated the photoproduction of
hydrogen.16-18 These systems are extremely complex, often
including a dispersed metal catalyst as well as a sacrificial
reductant in cases where the reduction of water to hydrogen
takes place without the concurrent oxidation to form oxygen.
Despite its promising origins in the late 1970s and early 1980s,
this field has been somewhat dormant in the past two decades.
Due to significant advances throughout the past 20 years, both
in analytical techniques and methodology as well as in transition
(4) Kalyanasundaram, K.; Kiwi, J.; Gratzel, M. HelV. Chim. Acta 1978, 61,
2720-2730.
(5) Moradpour, A.; Amouyal, E.; Keller, P.; Kagan, H. NouV. J. Chim. 1978,
2, 547-549.
(6) Kiwi, J.; Gratzel, M. Nature 1979, 281, 657-658.
(7) Johansen, O.; Launikonis, A.; Loder, J. W.; Mau, A. W. H.; Sasse, W. H.
F.; Swift, J. D.; Wells, D. Aust. J. Chem. 1981, 34, 981-991.
(8) Lehn, J. M.; Sauvage, J. P. NouV. J. Chim. 1977, 1, 449-451.
(9) Kirch, M.; Lehn, J. M.; Sauvage, J. P. HelV. Chim. Acta 1979, 62, 1345-
1384.
(10) Chan, S.-F.; Chou, M.; Creutz, C.; Matsubara, T.; Sutin, N. J. Am. Chem.
Soc. 1981, 103, 369-379.
(11) Brown, G. M.; Brunschwig, B. S.; Creutz, C.; Endicott, J. F.; Sutin, N. J.
Am. Chem. Soc. 1979, 101, 1298-1300.
(12) Krishnan, C. V.; Sutin, N. J. Am. Chem. Soc. 1981, 103, 2141-2142.
(13) Ziessel, R.; Hawecker, J.; Lehn, J. M. HelV. Chim. Acta 1986, 69, 1065-
1084.
(14) Krishnan, C. V.; Brunschwig, B. S.; Creutz, C.; Sutin, N. J. Am. Chem.
Soc. 1985, 107, 2005-2015.
(1) Darwent, J. R. In Photogeneration of Hydrogen; Harriman, A., West, M.
A., Eds.; Academic Press: London, 1982; pp 23-38.
(15) Hawecker, J.; Lehn, J. M.; Ziessel, R. NouV. J. Chim. 1983, 7, 271-277.
(16) Krasna, A. I. Photochem. Photobiol. 1979, 29, 267-276.
(17) Johansen, O.; Mau, A. W. H.; Sasse, W. H. F. Chem. Phys. Lett. 1983, 94,
107-112.
(2) Kiwi, J.; Borgarello, E.; Pelizzatti, E.; Visca, M.; Gratzel, M. In Photo-
generation of Hydrogen; Harriman, A., West, M. A., Eds.; Academic
Press: London, 1982; pp 119-146.
(3) Amouyal, E. In Homogeneous Photocatalysis; Chanon, M., Ed.; John Wiley
& Sons: Chichester, U.K., 1997; pp 263-308.
(18) Ballardini, R.; Juris, A.; Varani, G.; Balzani, V. NouV. J. Chim. 1980, 4,
563-564.
9
7502
J. AM. CHEM. SOC. 2005, 127, 7502-7510
10.1021/ja0427101 CCC: $30.25 © 2005 American Chemical Society

Heteroleptic Iridium Complexes for Photoinduced H Production
A R T I C L E S
mmol) and cobaltous chloride hexahydrate (0.95 g, 4.0 mmol) at reflux
for 12 h in 50 mL of absolute ethanol. The mixture was cooled and
added to 300 mL of diethyl ether. The resulting precipitate was collected
by filtration and dried in vacuo to yield 1.9 g [Co(bpy)3]Cl2 (79% yield).
Combinatorial synthesis of iridium complexes was carried out as
described previously,25 and the emission, excited-state lifetime, and
absorption measurements were performed exactly according to the
protocol described in detail.25 Lifetimes were measured by exciting at
337 nm with an N2 laser (Laser Science, Inc. VSL-337LRF, 10 ns
pulse), and the emission quantum yield (Φem) for each sample was
calculated according to Φs ) Φr(Is/Ir)(Ar/As), where the reference
complex (Φr ) 6.22%) was [Ir(ppy)2(bpy)]Cl for the parallel synthesis
products and the analogous PF6- salt for the control products. To utilize
these iridium complexes in photoinduced hydrogen production systems,
as well as to validate our parallel synthetic methodology, it was
necessary to synthesize large (ca. 50 mg) quantities by traditional
methods, as described in our previous work.25 The identity of each
compound was confirmed by 1H and 13C NMR (Varian Inova-500
spectrometer) and electrospray ionization mass spectrometry (Hewlett-
Packard 5898B MS Engine). All spectral information can be found in
the Supporting Information.
metal chemistry, it seems appropriate to reconsider this area of
research.
Since Merrifield’s pioneering work19 in 1986, combinatorial
techniques20-22 have been widely adopted in areas including
drug discovery23 and materials synthesis.24 These techniques
are applicable to complex problems with many degrees of
freedom, and we have previously used combinatorial synthesis
in concert with high-throughput screening techniques to greatly
accelerate the discovery of luminophoric ionic iridium com-
plexes.25 Not all of the compounds synthesized were suitable
for the OLED applications that were being pursued, but we
recognized that some of these materials could potentially serve
as excellent photosensitizers in homogeneous catalytic photo-
induced hydrogen production systems.
Recently, combinatorial techniques have been applied to the
study of photoinduced oxygen production via the oxidation of
water with colloidal transition metal particles.26 Here, we report
on the successful implementation of combinatorial synthesis and
high-throughput parallel screening techniques for the discovery
of novel photosensitizers that catalyze photoinduced water
reduction to yield hydrogen gas. In addition to identifying
promising photosensitizers, the parallel high-throughput analyti-
cal techniques utilized have allowed considerable insight into
the mechanisms governing the complex process of photoinduced
hydrogen production.
For photoinduced hydrogen production, photosensitizers with
large molar extinction coefficients (to efficiently collect radiant
energy) and relatively long excited-state lifetimes (to allow
electron-transfer quenching to occur) would be desirable. Using
the combinatorial techniques that we have applied to OLED
discovery,25 a library of 32 heteroleptic iridium complexes, each
with two cyclometalating ligands and one diimine ligand, was
synthesized. The photophysical parameters (excited-state life-
time, emission maximum, molar absorptivity, and quantum
efficiency) were measured and used to guide the traditional
synthesis of a six-compound sublibrary. Each compound in this
sublibrary was then used as a photosensitizer in a standard
photoinduced hydrogen production scheme. We utilized a
hydrogen selective Ni/Pd thin film hydrogen detector to quantify
hydrogen production in each system, allowing simple, reproduc-
ible, high-throughput measurements to be performed. We
demonstrated that all six of the iridium complexes investigated
were more effective photosensitizers than the ruthenium com-
plexes typically used.
Hydrogen Evolution. Samples for photoinduced hydrogen produc-
tion were prepared in 40 mL screw-cap glass vials (VWR) with silicone/
PTFE septa. Each sample was made up to a volume of 20 mL in 1:1
water:acetonitrile. Samples typically contained 1-6 µmol of the
photosensitizer, 50 µmol of [Co(bpy)3]Cl2 (the electron relay), 5.4 mmol
(230 mg) of LiCl, and 11.3 mmol (1.5 mL) of triethanolamine (TEOA)
(the sacrificial reductant); 0.4 mL of 37% HCl was added to each
solution to adjust the pH downward. Sample vials were capped and
deoxygenated by bubbling nitrogen through them for 15 min. Then
the samples were degassed under vacuum, through the septum on the
vial, at ambient temperature for 15 min to remove all of the dissolved
nitrogen. All hydrogen production was carried out with the vials still
under vacuum. The vials were then placed in a home-built eight-
compartment sample holder and illuminated from below using ultra-
bright light emitting diodes (Luxeon V Dental Blue, Future Electronics)
(see Figure 1).
These LEDs were chosen because, unlike the xenon lamps typically
used in similar experiments, they do not emit any light in the infrared
or ultraviolet portions of the spectrum. Additionally, they are modular,
allowing for versatility in experimental design, and they are capable
of delivering consistently reproducible illumination over extremely long
times (lifetime > 50 000 h). These LEDs have a luminous output of
500 mW ( 10% at 465 nm with a 20 nm fwhm and are driven at a
700 mA of current using a Xitanium Driver (Advance Transformer
Company). While the output from these LEDs does not cover the entire
solar spectrum, the robust, reproducible versatility that is gained from
their use is extremely desirable. To allow the screening of photosen-
sitizers that absorb strongly at other wavelengths, future work will
incorporate the use of green, red, and white LEDs into our scheme.
Each LED is cooled with a Wakefield Engineering 628 heat sink, and
the whole apparatus (vials, holder, LEDs) is placed on an orbital shaker
(Labline 3540) and shaken and illuminated overnight. After illumination,
the sample vials were then backfilled with water in order to bring them
to ambient pressure, and 1.0 mL of the bubble of gas trapped in the
vial was sampled using a Hamilton SampleLock syringe. This gas
sample was injected into a home-built sample chamber in which a
hydrogen sensor (H2Scan, RobustHydrogenSensor) was mounted. The
sensor was interfaced to a PC, and data were collected using an interface
designed in Labview. Prior to each sample injection, the sample
chamber was purged with dry nitrogen until the output of the hydrogen
sensor returned to its baseline reading (ca. 2 min). Typically, readings
were taken 90 s after sample injection to allow the system to equilibrate.
Daily calibrations were performed to ensure the accuracy of the sensor.
Experimental Section
Synthesis. Ligands were purchased from Aldrich or synthesized as
described previously.25 The synthesis and characterization of the
previously unreported ligands, 5-methyl-2-(2,4-difluoro)phenylpyridine
(F2-mppy) and 5-methyl-2-(2,4-dichloro)phenylpyridine (Cl2-mppy), are
described in the Supporting Information. Solvents and other reagents
were purchased from Aldrich and used without further purification.
[Co(bpy)3]Cl2 was synthesized by heating 2,2′-dipyridyl (2.0 g, 12.8
(19) Merrifield, B. Science 1986, 232, 341-347.
(20) Lebl, M. J. Comb. Chem. 1999, 1, 3-24.
(21) Kassel, D. B. Chem. ReV. 2001, 101, 255-267.
(22) Balkenhohl, F.; von dem Bussche-Hunnefeld, C.; Lansky, A.; Zechel, C.
Angew. Chem., Int. Ed. 1996, 35, 2288-2337.
(23) Houghten, R. A.; Pinilla, C.; Blondelle, S. E.; Appel, J. R.; Dooley, C. T.;
Cuervo, J. H. Nature 1991, 354, 84-86.
(24) Cawse, J. N. Acc. Chem. Res. 2001, 34, 213-221.
(25) Lowry, M. S.; Hudson, W. R.; Pascal, R. A., Jr.; Bernhard, S. J. Am. Chem.
Soc. 2004, 126, 14129-14135.
Quenching and Electrochemical Studies. Samples for quenching
studies were prepared in 1:1 water:acetonitrile with the appropriate
(26) Morris, N. D.; Mallouk, T. E. J. Am. Chem. Soc. 2002, 124, 11114-11121.
9
J. AM. CHEM. SOC. VOL. 127, NO. 20, 2005 7503

A R T I C L E S
Goldsmith et al.
Figure 1. Eight-well photoreactor for parallel photoinduced hydrogen generation experiments.
quencher concentration and a constant photosensitizer concentration
of 300 µM. Samples were placed in screw-top quartz cuvettes
(Spectrocell) with silicone/TFE septa and degassed with nitrogen for
10 min before measurements were performed. Luminescence lifetimes
were measured as described above.
The combinatorial products include 20 novel complexes and
12 others that can be appended to an existing library of 100
iridium products previously reported by this group.25 The novel
cyclometalating and neutral ligand combinations were chosen
to produce a range of emission maxima, molar absorptivities,
and excited-state lifetimes. As seen in Figure 3A, the products
of this combinatorial approach display exceptional color ver-
satility ranging from emission in the blue (2.59 eV, 477 nm) to
yellow-orange (2.08 eV, 593 nm) regions of the visible
spectrum. Similarly, a range of molar extinction coefficients
(ꢀ, 465 nm) from 384 to 3522 M-1‚cm-1 was observed (Figure
3B), and the excited-state lifetimes (τ) varied by nearly 2 orders
of magnitude from 384 to 13 400 ns (Figure 3C).
Cyclic voltammetry was carried out using a EG&G PAR 273A
potentiostat/galvanostat at a potential sweep rate of 100 mV/s. A
homemade platinum disk electrode served as the working electrode, a
coiled platinum wire was used as the counter electrode, and a silver
wire was used as a pseudo-reference electrode. Electrochemical
measurements were carried out in a home-built one-compartment cell
at approximately 1 mM photosensitizer concentration in 0.1 M tetra-
n-butylammonium hexafluorophosphate (TBAH) (Fluka, electrochemi-
cal grade) in acetonitrile (Fluka, >99.5% over molecular sieves).
Solutions were bubbled with nitrogen for 10 min prior to each
measurement to ensure that they were oxygen-free. Ferrocene (Aldrich)
was added as an internal standard, and all potentials were subsequently
referenced to saturated calomel electrode (SCE).
Therefore, the combinatorial approach allowed us to obtain
a collection of products exhibiting diverse photophysical proper-
ties with a relatively modest expenditure of time and resources.
To validate this approach and confirm the identity of the
combinatorial products, we also synthesized six products by
traditional methods; the agreement between the photophysical
properties of the traditionally prepared species and the analogous
combinatorial products can be seen in Table 1. The six control
compounds were selected for use as photosensitizers as dis-
cussed below, but in their capacity as controls, they effectively
confirm the identity of the combinatorial products and the
validity of the measurements. Furthermore, the ability of the
combinatorial technique to provide accurate and reproducible
results can also be demonstrated by adherence to the energy
gap law.27 As seen in Figure 4A, the agreement between the
new library of 32 combinatorial products and the six traditionally
prepared species provides convincing evidence that the com-
binatorial approach is indeed valid. As an indication of the
reproducibility of measurements across multiple combinatorial
Results/Discussion
Photophysical Measurements. A library of 32 complexes
of the type [Ir(C∧N)2(N∧N)]Cl was prepared by coordinating
two equivalent cyclometalating (C∧N) ligands and one neutral
ligand (N∧N) to an iridium(III) metal center using a parallel
synthetic procedure described in our previous work.25 The 32
products comprise the exhaustive set of combinations of four
cyclometalating ligands [C∧N ) 2-phenylpyridine (ppy), 5-meth-
yl-2-(4-fluoro)phenylpyridine (F-mppy), 5-methyl-2-(2,4-di-
fluoro)phenylpyridine (F2-mppy), 5-methyl-2-(2,4-dichloro)-
phenylpyridine (Cl2-mppy)] and eight neutral ligands [N∧N )
2,2′-dipyridyl (bpy), 4,4′-di-tert-butyl-2,2′-dipyridyl (dtbbpy),
4,4′-dimethyl-2,2′-dipyridyl (dmbpy), 4,4′-di-isopropyl-2,2′-
dipyridyl (dipbpy), 1,10-phenanthroline (phen), 3,4,7,8-tetra-
methyl-1,10-phenanthroline (Me4-phen), 4,4′-diphenyl-2,2′-di-
pyridyl (dphphen), cis-1,2-bis(diphenylphosphino)ethylene (dppe)].
The structure of these ligands can be seen in Figure 2.
(27) Kalyanasundaram, K. Photochemistry of Polypyridine and Porphyrin
Complexes; Academic Press: San Diego, CA, 1992.
9
7504 J. AM. CHEM. SOC. VOL. 127, NO. 20, 2005

Heteroleptic Iridium Complexes for Photoinduced H Production
A R T I C L E S
Figure 2. Four cyclometalating (C∧N) and eight neutral (N∧N) ligands employed in combinatorial synthesis.
attempts, the first library of 100 compounds previously re-
ported25 is also compared to the new library in Figure 4B. The
two data sets demonstrate consistent adherence to the negative
slope expected from the energy gap law.
vacuum conditions, the gas produced can reasonably be
considered to be effectively composed entirely of hydrogen.
Drawing on previous studies,3,10,13 the mechanism that we
consider to be in effect is as follows:
From the combinatorial products, a six-member sublibrary
of complexes {[Ir(C∧N)2(N∧N)]PF6, where C∧N ) ppy, F-mppy,
and N∧N ) bpy, phen, dphphen} was traditionally prepared
for photoinduced hydrogen production. The six species tested
were selected for their absorption and excited-state lifetime
properties. Three N∧N ligands were chosen primarily because
of their tendency to form complexes with large molar extinction
coefficients (Figure 3B); presumably, complexes of this type
would be desirable for their ability to efficiently capture radiant
energy. The choice of C∧N ligands, by contrast, reflects the
difference in excited-state lifetime among complexes that differ
in only their C∧N ligand, specifically those containing ppy
versus F-mppy. Because the absorption properties remain
relatively unchanged upon this substitution, ppy and F-mppy
were chosen to examine the effect of a longer lived excited state,
exhibited by the latter, on solar energy conversion.
PS + hν f PS*
(1)
(2)
(3)
PS* + ER f PS+ + ER-
ER- + H+ f ER + 1/2H2
PS+ + TEOA f PS + TEOA+
(4)
where PS is the photosensitizer molecule and ER is the electron
relay (Co(bpy)32+). It should be noted that it takes two cycles
through the pathway shown above to generate 1 equiv of
hydrogen gas. In certain systems (notably those where an
ascorbate buffer is used as the reaction medium), it has been
shown that the initial reaction is a reductive (not an oxidative)
quenching of the excited photosensitizer by ascorbate.12,28 By
analogy, it might be suspected that TEOA could reductively
quench the excited photosensitizer, but Sutin and co-workers
showed, for Ru(II)/Co(II)/TEOA systems, that TEOA does not
quench the photosensitizer.13 The ability of TEOA to serve as
a two-electron sacrificial reductant has also been postulated,9
and such a process, where one TEOA molecule can reduce a
photosensitizer as well as an electron relay molecule, may be
occurring in our system. The only consequence of such a
pathway, however, would be the reporting of an inflated turnover
Hydrogen Production. In a typical experiment, hydrogen
evolution commenced several minutes after beginning the
illumination of the sample. Bubbles (ca. 3-5 mm in diameter)
began to rise from within the solution at a vigorous rate, and
this continued for 30-60 min. After this time, the bubble
production slowed; after overnight illumination, no more bubbles
were evident. In some cases, the initially transparent solution
became darker when exposed to light. We believe this darkening
is due to the generation of a reduced cobalt species as described
in the literature,8 and this will be discussed further below.
Analysis of the gas produced typically indicated that it was more
than 80% H2 and in some cases greater than 95% H2. Since the
atmosphere inside our sample vials contains water and aceto-
nitrile vapor, as well as some air due to leakage and imperfect
2+
number for the Co(bpy)3 electron relay.
Table 2 shows the results of photoinduced hydrogen produc-
tion experiments: the amount of hydrogen produced as well as
the number of turnovers that the photosensitizer and electron
2+
relay underwent during the course of the experiment. Ru(bpy)3
(28) Creutz, C.; Sutin, N.; Brunschwig, B. S. J. Am. Chem. Soc. 1979, 101,
1297-1298.
9
J. AM. CHEM. SOC. VOL. 127, NO. 20, 2005 7505

A R T I C L E S
Goldsmith et al.
Table 1. A Comparison of Photophysical Properties between the
Six Combinatorial Products and Their Corresponding Controls
1
-1
sample complex
Cl-
PF6
[Ir(ppy)2(bpy)]+1
λ ) 585 nm
Φ ) 6.22%a
τ ) 384 ns
Φ ) 6.22%a
τ ) 390 ns
[Ir(ppy)2(phen)]+1
λ ) 579 nm
Φ ) 13.03%
τ ) 786 ns
Φ ) 11.86%
τ ) 783 ns
[Ir(ppy)2(dphphen)]+1
λ ) 587 nm
Φ ) 10.25%
τ ) 935 ns
Φ ) 26.85%
τ ) 978 ns
Φ ) 37.10%
τ ) 1397 ns
Φ ) 25.17%
τ ) 1645 ns
Φ ) 18.18%
τ ) 1033 ns
Φ ) 21.98%
τ ) 1049 ns
Φ ) 33.06%
τ ) 1636 ns
Φ ) 32.23%
τ ) 1924 ns
[Ir(F-mppy)2(bpy)]+1
λ ) 558 nm
[Ir(F-mppy)2(phen)]+1
λ ) 550 nm
[Ir(F-mppy)2(dphphen)]+1
λ ) 556 nm
a Standard values.
Figure 4. The energy gap law dictates a linear relationship between the
natural log of the nonradiative decay rate constant (knr) and the emission
energy. (A) The agreement between the complete library of 32 combinatorial
products and the six controls products confirms the validity of the
combinatorial technique and the accuracy of the photophysical measure-
ments. (B) The correspondence between the first library previously reported
by this group and the new library indicates that the combinatorial technique
is accurate and reproducible.
2+
and Ru(4,7-dimethyl-1,10 phenanthroline)32+ (Ru(dmphen)3
)
are commonly used photosensitizers that have been shown to
be among the most effective available.3 It is clear from our
results that all six heteroleptic iridium complexes are consider-
ably more effective than current transition metal based photo-
sensitizer benchmark compounds. While no glaring trends can
be determined from the data in Table 2, it does appear that the
(F-mppy)-containing compounds, with longer excited-state
lifetimes than the corresponding (ppy)-containing compounds,
Figure 3. Three-dimensional plots of the emission energy (A), molar
absorptivity at 465 nm (the emission maximum of the LEDs) (B), and the
log of the excited-state lifetime (C) depict the diversity of photophysical
properties exhibited by the combinatorial library. The six compounds chosen
for use as photosensitizers in the hydrogen production reaction are designated
with arrows.
9
7506 J. AM. CHEM. SOC. VOL. 127, NO. 20, 2005

Heteroleptic Iridium Complexes for Photoinduced H Production
A R T I C L E S
Table 2. Photoinduced Hydrogen Productiona
H
2 evolved
PS
ER
photosensitizer
(µmol)
turnovers
turnovers
2+
Ru(bpy)3
50
100
580
800
860
840
920
860
860
2
12
16
17
17
18
17
17
2+
Ru(dmphen)3
290
400
430
420
460
430
430
Ir(ppy)2(bpy)+
Ir(ppy)2(phen)+
Ir(ppy)2(dphphen)+
Ir(F-mppy)2(bpy)+
Ir(F-mppy)2(phen)+
Ir(F-mppy)2(dphphen)+
a All samples have a total volume of 20 mL and contain 50 µM
2+
photosensitizer, 2.5 mM Co(bpy)3 (electron relay), 0.57 M TEOA, and
0.27 M LiCl in 50:50 water:acetonitrile. The pH was adjusted by the addition
of 0.4 mL of 12 M HCl.
produce slightly more hydrogen. It is also important to note
that the excited-state lifetimes of all of the iridium complexes
studied cover a range similar to that spanned by the two standard
ruthenium complexes.29 It is quite interesting that all of the
iridium-based photosensitizers lead to the production of roughly
the same amount of hydrogen. This perhaps points to the
existence of some process that limits one or more in the series
of reactions that leads to hydrogen production.
Figure 5. Relative quantum yield of hydrogen for six novel iridium
photosensitizers and two standard ruthenium photosensitizers. Relative
2+
quantum yield for Ru(bpy)3 set to 1.
Table 3. Quenching Kinetics
2+
k
q Co(bpy)3
k
q TEOA
1
1
photosensitizer
(M-1 s-
)
(M-1 s-
)
As a test of our combinatorial methodology, samples from
the parallel synthesis described above (chloride salts in ethylene
glycol) were also used for photoinduced hydrogen generation
2+
Ru(bpy)3
2.3 × 107
1.3 × 108
4.8 × 108
6.9 × 108
6.3 × 108
4.5 × 108
6.0 × 108
4.5 × 108
no quenching
no quenching
5.4 × 106
7.7 × 106
1.1 × 107
2.7 × 107
3.4 × 107
3.3 × 107
2+
Ru(dmphen)3
Ir(ppy)2(bpy)+
Ir(ppy)2(phen)+
by adding the sample in its entirety to the standard H2O/CH3-
Ir(ppy)2(dphphen)+
Ir(F-mppy)2(bpy)+
Ir(F-mppy)2(phen)+
Ir(F-mppy)2(dphphen)+
2+
CN/TEOA/LiCl/Co(bpy)3
reaction mixture. In all cases,
hydrogen was produced, but these results cannot be directly
compared to those shown in Table 2 due to the presence of the
ethylene glycol solvent and differences in photosensitizer
concentration. However, the same general trends seen in Table
2 for traditionally synthesized photosensitizers were observed
when the parallel-synthesized photosensitizers were utilized for
hydrogen production.
constant for dynamic quenching.30 As can be seen in Figure
6A-F, all plots of (τ0/τ) versus concentration are extremely
linear (R2 > 0.99 in all cases).
It is notable that neither Ru(bpy)32+ nor Ru(dmphen)32+ are
quenched at all by TEOA, which is in agreement with the
literature.9,13 The values of kq for each compound with each of
the two quenchers can be seen in Table 3. Somewhat surpris-
ingly, all of the heteroleptic iridium complexes under investiga-
The two ruthenium-based photosensitizers mentioned above
have much larger molar extinction coefficients than any of the
iridium complexes utilized, and in order to more accurately rate
the energy harvesting and conversion properties of our iridium
complexes, it is necessary to consider this fact. Using the
measured extinction coefficients and the photosensitizer con-
centration, the fraction of incident radiation captured can be
determined and used to calculate a relative quantum yield of
hydrogen, that is, the amount of hydrogen produced per amount
of light energy collected by the photosensitizer. These results
can be seen in Figure 5, in which the relative quantum yield of
2+
tion were quenched by both Co(bpy)3 and TEOA, although
the rate for quenching by TEOA is approximately 1 order of
2+
magnitude slower than the Co(bpy)3 quenching rate. It is
noteworthy that the iridium photosensitizers are quenched more
strongly than their ruthenium-based counterparts by the cobalt
electron relay, which correlates with the increased hydrogen
production observed.
2+
hydrogen for Ru(bpy)3 has been set to unity.
This figure demonstrates that these newly synthesized iridium
complexes are generally at least 20 times more efficient than
Electrochemical analyses were performed in order to measure
the redox potentials of all eight photosensitizers used, and in
conjunction with the spectroscopic data shown in Figure 3A,
the excited-state redox potentials were determined.27 The
ground-state redox potentials are shown in Table 4, and the
2+
Ru(bpy)32+ and 3-6 times more efficient than Ru(dmphen)3
at converting visible light into hydrogen.
Quenching and Electrochemical Studies. To gain a more
complete understanding of the complex behavior of these novel
photosensitizers, the dynamic quenching of each of the eight
photosensitizers utilized, with both Co(bpy)32+ and TEOA acting
as quenchers, was investigated. The luminescence lifetime of
each photosensitizer was measured without quencher and in the
presence of varying quencher concentrations, and the lifetime
ratio (τ0/τ) was determined and used to calculate kq, the rate
excited-state redox potentials are tabulated in Table 5. Ru-
2+
(dmphen)3
and all six iridium photosensitizers are all con-
siderably more powerful reducing agents than Ru(bpy)32+, which
again is associated with the increased hydrogen production seen
with these compounds. The three iridium complexes with
F-mppy ligands are all stronger oxidizers than the corresponding
iridium ppy complexes, which goes hand in hand with the
(29) Juris, A.; Balzani, V.; Barigelletti, F.; Campagna, S.; Belser, P.; von
(30) Lakowicz, J. R. Principles of Fluorescence Spectroscopy; Plenum Press:
Zelewsky, A. Coord. Chem. ReV. 1988, 84, 85-277.
New York, 1983.
9
J. AM. CHEM. SOC. VOL. 127, NO. 20, 2005 7507

A R T I C L E S
Goldsmith et al.
Table 4. Redox Properties of Photosensitizersa
Table 6. Effect of Photosensitizer Concentration on Photoinduced
Hydrogen Productiona
E0′ Mn+/M(n
∆Ep
E0′ L/L-
∆Ep
+1)+b
photosensitizer
photosensitizer
(V vs SCE)
(mV)
(V vs SCE)
(mV)
concentration
M)
H
2 evolved
photosensitizer
turnovers
2+
Ru(bpy)3
+1.26
+1.11
+1.25
+1.24
+1.23
+1.38
+1.36
+1.36
80
60
65
65
75
75
60
75
-1.36
-1.53
-1.42
-1.42
-1.38
-1.39
-1.39
-1.35
75
65
70
80
70
60
80
70
(
µ
(µ
mol)
2+
Ru(dmphen)3
Ir(ppy)2(bpy)+
0.50
0.99
2.5
4.9
7.4
15
17
37
74
296
320
43
83
9200
8900
6900
4500
4200
2300
2400
1100
540
Ir(ppy)2(phen)+
Ir(ppy)2(dphphen)+
Ir(F-mppy)2(bpy)+
Ir(F-mppy)2(phen)+
Ir(F-mppy)2(dphphen)+
160
210
300
320
390
400
377
383
401
a Cyclic voltammetry was carried out at 100 mV/s in 0.1 M TBAH/
CH3CN at a Pt working electrode with a Pt counter electrode and Ag wire
pseudo-reference. Ferrocene was used as an internal standard, and potentials
are reported with respect to a saturated calomel (SCE) electrode. b For
ruthenium complexes, n ) 2; for iridium complexes, n ) 3.
140
130
a All samples have a total volume of 20 mL and contain the specified
amount of Ru(dmphen)3 (the photosensitizer), 2.5 mM Co(bpy)3
Table 5. Excited-State Redox Properties of Photosensitizers
2+
2+
(electron relay), 0.57 M TEOA, and 0.27 M LiCl in 50:50 water:acetonitrile.
The pH was adjusted by the addition of 0.4 mL of 12 M HCl.
emission
+
1)
+a
-1)+a
max
(eV)
E0′ *Mn+/M(n
E0′ *Mn+/M(n
photosensitizer
(V vs SCE)
(V vs SCE)
led us to question whether we were actually probing the
performance limits of the photosensitizers. By decreasing the
2+
Ru(bpy)3
2.04
2.07
2.10
2.12
2.08
2.20
2.22
2.20
-0.78
-0.96
-0.85
-0.88
-0.85
-0.82
-0.86
-0.84
+0.68
+0.54
+0.68
+0.70
+0.70
+0.81
+0.83
+0.85
2+
Ru(dmphen)3
2+
concentration of Ru(dmphen)3 used while keeping the con-
Ir(ppy)2(bpy)+
Ir(ppy)2(phen)+
centration of electron relay constant, it was possible to establish
an upper limit to the number of turnovers of the photosensitizer.
The results from this experiment (the quantity of hydrogen
evolved and the number of turnovers for the photosensitizer)
can be seen in Table 6. Even at submicromolar photosensitizer
concentrations, hydrogen was still produced. It appears that the
Ru(dmphen)32+ photosensitizer is capable of approximately 9000
turnovers. As will be discussed below, the photosensitizer
appears not to be the limiting factor in the photoinduced
production of hydrogen. In Figure 7, a plot of hydrogen
production versus photosensitizer concentration can be seen.
When the photosensitizer concentration is low, the amount
of hydrogen produced increases substantially with increasing
photosensitizer concentration. However, as the concentration is
increased, it appears that the amount of hydrogen produced
reaches a plateau. We ascribe this behavior to the optical
properties of the photosensitizer; if enough chromophore is
present, all the light from the LED light sources will be absorbed
and additional photosensitizer will have no effect. The amount
of hydrogen evolved should be proportional to the amount of
light energy collected and can be given by eq 5
Ir(ppy)2(dphphen)+
Ir(F-mppy)2(bpy)+
Ir(F-mppy)2(phen)+
Ir(F-mppy)2(dphphen)+
a For ruthenium complexes, n ) 2; for iridium complexes, n ) 3.
observation (see Table 3) that the three F-mppy iridium
compounds are more strongly quenched by TEOA (the sacri-
ficial reductant which gets oxidized) than their ppy counterparts.
It is plausible that the increased hydrogen production seen in
the F-mppy compounds relative to the ppy compounds is due
to the increased efficiency of a secondary pathway to hydrogen
production via quenching of the excited photosensitizer by
TEOA. However, from our work to date, a straightforward
correlation between either quenching or electrochemical
parameters and hydrogen production is not evident. This
complexity lends credence to our decision to approach the
investigation of this system with a parallel methodology.
Mechanistic Studies. The lack of clear correlations between
the photophysical properties of our iridium photosensitizers and
their performance in terms of hydrogen production led us to
conduct several studies to try to elucidate information about
the mechanism of the obviously complex catalytic process
occurring. These studies were facilitated through the use of the
high-throughput photophysical screening protocols described
previously as well as by the multiwell apparatus and high-
throughput hydrogen sensor described above.
One could easily envision a scenario in which hydrogen
production is limited by the catalytic turnover of the photosen-
sitizer molecule. Alternatively, hydrogen generation might
instead be controlled by the catalytic limitations of the electron
relay. The goal of our initial mechanistic study was to investigate
the catalytic properties of the photosensitizer, specifically how
many times the molecule could “turn over”. For this investiga-
tion, we employed the “standard” sensitizer, Ru(dmphen)3
but ongoing research will extend this work to the iridium
photosensitizers that have been discovered. The results shown
in Table 2 indicate approximately the same number (ca. 800-
900) of turnovers for most of the photosensitizers used, which
H2 ) m(1 - 10-ꢀLc
)
(5)
where m is a proportionality constant, ꢀ is the molar extinction
coefficient of the chromophore, L is the path length of the system
(approximately 3.5 cm from the bottom of the sample vial to
the top of the solution), and c is the concentration of the
photosensitizer. The solid line in Figure 7 is a fit of the data to
eq 5, and it is clear that this model describes the data quite
well, with an R2 value of 0.9792.
To further probe this system, another set of experiments was
2+
carried out, varying the amount of Co(bpy)3 used while
keeping a constant concentration of photosensitizer. For these
experiments, a photosensitizer concentration was chosen that
was in the “plateau region” described above. This choice was
made to ensure that there was no variation in the amount of
hydrogen produced due to the amount of photosensitizer present.
The results of these experiments (the quantity of hydrogen
evolved and the number of turnovers for the electron relay) are
2+
,
9
7508 J. AM. CHEM. SOC. VOL. 127, NO. 20, 2005

Heteroleptic Iridium Complexes for Photoinduced H Production
A R T I C L E S
Figure 6. (A-C) Dynamic quenching studies of two ruthenium-based photosensitizers and six novel iridium photosensitizers by TEOA. (D-F) Dynamic
quenching studies of two ruthenium-based photosensitizers and six novel iridium photosensitizers by Co(bpy)32+. All linear fits shown have R2 > 0.99.
Table 7. Effect of Electron Relay Concentration on Photoinduced
Hydrogen Productiona
2+
Co(bpy)3
2
+
concentration
(mM)
H
2 evolved
mol)
Co(bpy)3
(
µ
turnovers
0.17
0.46
0.88
1.25
1.83
2.50
5.0
120
140
210
220
300
270
360
332
74
31
24
18
16
11
7
10.0
3
a All samples have a total volume of 20 mL and contain 300 µM
2+
2+
Ru(dmphen)3 (the photosensitizer), the specified amount of Co(bpy)3
(electron relay), 0.57 M TEOA, and 0.27 M LiCl in 50:50 water:acetonitrile.
The pH was adjusted by the addition of 0.4 mL of 12 M HCl.
2+
2+
Figure 7. Hydrogen evolution versus concentration of Ru(dmphen)3
photosensitizer. In all cases, the concentration of electron relay Co(bpy)3
was 2.5 mM. The inset gives detail at low concentration. Solid line is a fit
concentrations, it reaches a plateau. However, the function of
the electron relay is to react with excited photosensitizer
molecules, not to absorb light energy; hence, increasing the
of data to eq 5 (R2 ) 0.9792).
2+
Co(bpy)3 concentration would not be expected to have the
shown in Table 7. Additionally, a plot of hydrogen production
versus electron relay concentration can be seen in Figure 8.
Superficially, the behavior shown in this plot appears similar
to that for the photosensitizer; at low Co(bpy)32+ concentrations,
the amount of hydrogen produced increases, and at higher
same effect as increasing the photosensitizer concentration. For
the experiments described in Table 7, the concentration of the
electron relay is changed by a factor of 60. However, the amount
of hydrogen produced only changes by a factor of 3. At low
9
J. AM. CHEM. SOC. VOL. 127, NO. 20, 2005 7509

A R T I C L E S
Goldsmith et al.
our methodology, which involves combinatorial synthesis, high-
throughput photophysical screening, parallel photocatalytic
reactions, and the use of a fast, selective hydrogen sensor, is an
efficient and effective pathway for the advancement of science
and technology in this area of research.
The ability to carry out photoinduced hydrogen generation
experiments in parallel and to quickly analyze the gaseous
products has enabled the acquisition of significant mechanistic
understanding of the processes involved. While the photo-
sensitizer is extremely catalytic (at least 9000 turnovers), the
limiting factor in the generation of hydrogen appears to be the
decay of the electron relay Co(bpy)32+, which decomposes after
less than 100 turnovers. Work is currently underway to optimize
the electron relay through the application of the combinatorial,
high-throughput methodology laid out in this paper.
Figure 8. Hydrogen evolution versus concentration of Co(bpy)32+ electron
relay. In all cases, the concentration of the photosensitizer Ru(dmphen)3
was 300 µM. The solid line is a guide to the eye.
2+
Acknowledgment. We thank Nathan Speer and Robert
Pendergrass at H2Scan for making the RobustHydrogenSensor
available, as well as for useful discussions and generous
technical support. We also thank Robert A. Pascal Jr. for the
generous gift of laboratory equipment, and Andrew Bocarsly
for the use of a potentiostat. S.B. acknowledges support from a
Camille and Henry Dreyfus New Faculty Award and a NSF
CAREER award (CHE-0449755). T.H.A. acknowledges support
for summer research through Princeton University’s Partners
in Science Program.
concentrations of Co(bpy)32+, the electron relay is capable of
more than 70 turnovers, while at very high concentrations, it
only turns over several times. We postulate that this behavior
is due to a reaction of the reduced (i.e., very reactive) electron
relay molecules that yields an inactive cobalt species. The
existence of this reaction would have the effect of destroying
the electron relay and shutting off the production of hydrogen,
and evidence of such a process will be detailed in a forthcoming
manuscript.
Conclusions
Supporting Information Available: Procedures for the syn-
thesis of the ligands F2-mppy and Cl2-mppy, as well as for the
preparation of the six iridium complexes examined in this work.
Structural identification via 1H and 13C NMR as well as by ESI-
MS is also included. This material is available free of charge
via the Internet at http://pubs.acs.org.
In this work, we have demonstrated the combinatorial
development of six heteroleptic iridium transition metal com-
plexes that serve as extremely effective photosensitizers for the
generation of hydrogen through the photocatalytic reduction of
water. All six of these photosensitizers perform more efficiently
than the current best photosensitizers. We have also shown that
JA0427101
9
7510 J. AM. CHEM. SOC. VOL. 127, NO. 20, 2005