High energy and quantum efficiency in photoinduced charge separation

Article abstract of DOI:10.1021/ja0665500

Supramolecular triad assemblies consisting of a central trisbipyridine ruthenium(II) chromophore (C²⁺) with one or more appended phenothiazine electron donors (D) and a diquat-type electron acceptor (A ²⁺) have been shown to form long-lived photoinduced charge separated states (CSS) with unusually and consistently high quantum efficiency. Up to now, there has been no understanding for why these large efficiencies (often close to unity) are achieved across this entire class of triads when other, seemingly similar systems are often much less efficient. In the present study, using a bimolecular system consisting of a chromophore-acceptor diad (C ²⁺-A²⁺) and an N-methylphenothiazine donor, we demonstrate that a ground-state association exists between the RuL₃²⁺ and the phenothiazine prior to photoexcitation. It is this association process that is responsible for the efficient CSS formation in the bimolecular system and, by inference, also must be an essential factor in the fully intramolecular process occurring with the D-C²⁺-A²⁺ triad analogues.

Full text of DOI:10.1021/ja0665500

Published on Web 12/22/2006
High Energy and Quantum Efficiency in Photoinduced Charge
Separation
John M. Weber, Matthew T. Rawls, Valerie J. MacKenzie, Bradford R. Limoges, and
C. Michael Elliott*
Contribution from the Department of Chemistry, Colorado State UniVersity, Fort Collins,
Colorado 80523
Received September 11, 2006; E-mail: elliott@lamar.colostate.edu
Abstract: Supramolecular triad assemblies consisting of a central trisbipyridine ruthenium(II) chromophore
(C2+) with one or more appended phenothiazine electron donors (D) and a diquat-type electron acceptor
(A2+) have been shown to form long-lived photoinduced charge separated states (CSS) with unusually
and consistently high quantum efficiency. Up to now, there has been no understanding for why these large
efficiencies (often close to unity) are achieved across this entire class of triads when other, seemingly
similar systems are often much less efficient. In the present study, using a bimolecular system consisting
of a chromophore-acceptor diad (C2+-A2+) and an N-methylphenothiazine donor, we demonstrate that a
2+
ground-state association exists between the RuL3 and the phenothiazine prior to photoexcitation. It is
this association process that is responsible for the efficient CSS formation in the bimolecular system and,
by inference, also must be an essential factor in the fully intramolecular process occurring with the D-C2+
-
A2+ triad analogues.
Introduction
various functional analogues (which may or may not also be
structural analogues) of the photosynthetic reaction center: for
example, a light absorbing chromophore (C) covalently linked
to an electron acceptor (A) and/or an electron donor (D).
Efficiently mimicking the first step in the photosynthetic process
is not difficult. There have been many systems reported over
the years in which a photoexcited chromophore, attached to an
acceptor (or a donor), undergoes electron-transfer quenching
with essentially 100% efficiency. The tricky bit is efficiently
affecting the second electron transfer which actually separates
the charge. Here, success with synthetic mimics has been much
less common because of the inherent competition between the
productive, second charge separation step and the parasitic back
electron transfer, wherein the latter usually wins. For example,
it is uncommon for such systems to form multistep charge
separated states (CSS) with quantum efficiencies, ΦCSS, greater
than ca. 25-30%, and frequently it is much lower. There are
few examples outside the systems exemplified by the structure
in Figure 1 (vide infra) where ΦCSS is very large (e.g., >90%),
but these systems are exceptions rather than the rule.10
In photosynthetic reaction centers, nature has designed a
molecular apparatus capable of separating charge with unsur-
passed efficiency. In bacterial systems, where we have good
structural data, the process initiates when a photon is absorbed
by the so-called “special pair” of chlorophylls.1,2 Within a few
picoseconds, the resulting excited state transfers an electron with
unity efficiency to an adjacent pheophytin electron acceptor,
reducing it and leaving a hole on the special pair.3 It is at this
stage where arguably the most remarkable step in the entire
photosynthetic process occurs. The photoinjected electron on
the pheophytin has two choices. It can recombine with the hole
on the special pair, or it can transfer to an adjacent quinone
molecule that serves as a secondary electron acceptor. What
makes this second step so surprising is that essentially 100%
of the time the electron transfers to the quinone and does not
recombine with the hole on the special pair, despite the much
larger thermodynamic driving force for the latter process and
roughly similar electron-transfer distances. Subsequently, the
electron continues to cascade down the series of electron
acceptors until eventually it and the hole are “harvested” as
chemical energy.
A significant portion of our efforts in this area has focused
on so-called “triad” assemblies incorporating an electron ac-
ceptor (A), an electron donor (D), and a light absorbing
Over the last several decades, various groups, including our
own, have attempted to provide functional mimics of the first
several steps in this photosynthetic process.4-9 A common
approach is to synthesize supramolecular assemblies that contain
(5) Schanze, K. S.; Walters, K. A. In Organic and Inorganic Photochemistry;
Ramamurthy, V., Schanze, K. S., Eds.; Marcel Dekker: New York, 1998;
Vol. 2, pp 75-127.
(6) Armaroli, N. Photochem. Photobiol. Sci. 2003, 2, 73-87.
(7) Scandola, F.; Chiorboli, C.; Indelli, M. T.; Rampi, M. T. In Biological and
Artificial Supermolecular Systems; Wiley: Weinheim, 2003; Vol. 3, pp
337-403.
(1) Deisenhofer, J.; Michel, H. Annu. ReV. Cell Biol. 1991, 7, 1-23.
(2) Deisenhofer, J.; Michel, H. Annu. ReV. Biophys. Biophys. Chem. 1991, 20,
247-266.
(3) Greenfield, S. R.; Wasielewski, M. R. Photosynth. Res. 1996, 48, 83-97.
(4) Kalyanasundaram, K. Photochemistry of Polypyridine and Porphyrin
Complexes; Academic Press: San Diego, CA, 1992.
(8) Gust, D.; Moore, T. A.; Moore, A. L. Acc. Chem. Res. 2001, 34, 40-48.
(9) Gust, D.; Moore, T. A. Porphyrin Handb. 2000, 8, 153-190.
(10) Bahr, J. L.; Kuciauskas, D.; Liddell, P. A.; Moore, A. L.; Moore, T. A.;
Gust, D. Photochem. Photobiol. 2000, 72, 598-611.
9
10.1021/ja0665500 CCC: $37.00 © 2007 American Chemical Society
J. AM. CHEM. SOC. 2007, 129, 313-320
313

A R T I C L E S
Weber et al.
λ
max, yet both systems have essentially the same ΦCSS (close to
unity). For comparison, the group at the Center for Study of
Early Events in Photosynthesis at Arizona State University has
studied D-C-A triads containing carotenoid donors, porphyrin
chromophores, and C60 acceptors. Within this family of triads,
ΦCSS can be changed from 100 to 22% by merely altering the
type of substituents on the porphyrin chromophore periphery
without even affecting the amount of stored energy.10,17
The purpose of this article is to develop an explanation for
why the particular assemblies exemplified in Figure 1 are so
uniformly efficient at forming CSS while other putatively similar
photoinduced charge separation systems are not. Before em-
barking on the discussion of the present results, it is necessary
to review a few important general facts established in previous
studies about this family of triad assemblies. The excited-state
form of the chromophore is a 3MLCT that is highly luminescent
in the absence of any attached acceptors or donors. The solution
emission lifetime, τem, is ca. 600 to 900 ns dependent upon the
solvent. Diad assemblies of both C2+-A2+ and C2+-D have
been prepared and their emission lifetimes examined to establish
the respective quenching rates. As the driving force is changed
and numbers of methylenes in the chain linking C2+ and A2+
are varied from 2 to 6, the oxidative electron-transfer quenching
rates for C2+-A2+ diads vary from about 1 × 1010 to 1 × 107
Figure 1. Representative donor-chromophore-acceptor triad complex that
undergoes efficient photoinduced charge separated state formation.
chromophore (C) represented by the structure shown in Figure
1. The components of these triads typically consist of a central
2+
trisbipyridine ruthenium(II) chromophore (RuL3 ) C2+
)
covalently linked to one or more phenothiazine donors (PTZ )
D) and to an N,N′-diquaternary-2,2′-bipyridinium electron
acceptor (DQ2+ ) A2+).11-14 The linkages between the chro-
mophore and the D and A2+ moieties are typically polymeth-
ylene chains emanating from the periphery of bipyridine ligands.
Thus, the saturated alkyl chains bridging the molecular com-
ponents are both nominally electrically insulating and flexible.
The molecule shown in Figure 1 is one example of roughly 50
similar triad assemblies that we have prepared and studied.11-14
Considering the many types of light-induced charge separation
systems that exist, there is no a priori reason one should expect
triads exemplified by Figure 1 to be uniformly efficient at charge
separation; yet, they are. Typically, all D-C2+-A2+ systems
of this family undergo photoinduced CSS formation with very
high quantum efficiencies, often approaching unity.15,16 Ad-
ditionally, counter to what Marcus Theory predicts, there seems
to be little correlation between the thermodynamic driving force
for the various electron-transfer steps and ΦCSS. To illustrate
this point, consider the D-C2+-A2+ triad shown in Figure 1
but with two methylenes in the chain connecting the quaternary
nitrogens of the acceptor (rather than three). The photoinduced
CSS formed from this triad stores ca. 1.13 eV of energy, about
56% of the triplet metal-to-ligand charge-transfer excited state
(3MLCT) energy and 42% of the incident photon energy at λmax
of the chromophore absorption (460 nm). Altering the diquat
acceptor such that a four-carbon chain links the two quaternary
nitrogens increases the energy stored in the CSS to 1.47 eV or
72% 3MLCT energy and 55% of the incident photon energy at
s-1 12,14,18-20
In contrast, as the number of methylenes in the
.
chain linking C2+ and D is varied from 1 to 9, the reductive
quenching rates for C2+-D vary only from ca. 5 × 106 to 1 ×
106 s-1 13,21
Thus, the slowest oxidative quenching is at least 2
.
times faster than the fastest reductive quenching. More impor-
tantly, when the emission decay rate of any D-C2+-A2+ triad
is compared with that of the corresponding C2+-A2+ diad (i.e.,
with the same A2+-containing ligand), the relatiVe rates never
differ by more than a factor of 2 even though the absolute decay
rates change by more than a factor of ca. 103 (over the entire
collection of compounds). The conclusion that must be drawn
from these data is that the initial quenching event in all D-C2+
-
A2+ triads of this family is always oxidative quenching and D
is, at this point, not involved:
D-C2+*-A2+ f D-C3+-A•+
Another relevant fact from these earlier studies is that, in the
time-resolved absorption spectra of the C2+-A2+ diads, there
is never any detectable amount of A•+ present.12 One must
conclude then, that the back electron transfer rate for
C3+-A•+ f C2+-A2+
must be significantly faster than the oxidative quenching rate.
At this juncture, the stage has been set to propose two
alternate hypotheses for the origin of the consistently large ΦCSS
values of these triads:
(11) Danielson, E.; Elliott, C. M.; Merkert, J. W.; Meyer, T. J. J. Am. Chem.
Soc. 1987, 109, 2519-2520.
(12) Cooley, L. F.; Larson, S. L.; Elliott, C. M.; Kelley, D. F. J. Phys. Chem.
1991, 95, 10694-10700.
(17) Kodis, G.; Liddell, P. A.; Moore, A. L.; Moore, T. A.; Gust, D. J. Phys.
Org. Chem. 2004, 17, 724-734.
(13) Larson, S. L. Ph.D. Dissertation, Colorado State University, Fort Collins,
CO, 1994.
(18) Schmehl, R. H.; Ryu, C. K.; Elliott, C. M.; Headford, C. L. E.; Ferrere, S.
AdV. Chem. Ser. 1990, 226, 211-223.
(14) Larson, S. L.; Elliott, C. M.; Kelley, D. F. J. Phys. Chem. 1995, 99, 6530-
6539.
(19) Cooley, L. F.; Headford, C. E. L.; Elliott, C. M.; Kelley, D. F. J. Am.
Chem. Soc. 1988, 110, 6673-6682.
(15) Klumpp, T.; Linsenmann, M.; Larson, S. L.; Limoges, B. R.; Buerssner,
D.; Krissinel, E. B.; Elliott, C. M.; Steiner, U. E. J. Am. Chem. Soc. 1999,
121, 4092.
(20) Ryu, C. K.; Wang, R.; Schmehl, R. H.; Ferrere, S.; Ludwikow, M.; Merkert,
J. W.; Headford, C. E. L.; Elliott, C. M. J. Am. Chem. Soc. 1992, 114,
430-438.
(16) Klumpp, T.; Linsenmann, M.; Larson, S. L.; Limoges, B. R.; Buerssner,
D.; Krissinel, E. B.; Elliott, C. M.; Steiner, U. E. J. Am. Chem. Soc. 1999,
121, 1076-1087.
(21) Larson, S. L.; Elliott, C. M.; Kelley, D. F. Inorg. Chem. 1996, 35, 2070-
2076.
9
314 J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007

High Energy and Quantum Efficiency in Charge Separation
A R T I C L E S
lized from ethyl acetate. Lithium diisopropylamid (LDA) was prepared
by the addition of a heptane solution of n-butyllithium to diisopropyl-
amine in tetrahydrofuran (THF). THF was dried by refluxing under
nitrogen over sodium/benzophenone.
4-(3-Hydroxypropyl)-4′-methyl-2,2′-bipyridine, (I). Under nitro-
gen, 0.95 equiv of LDA in THF was added to 10.0 g of DMB which
had been dissolved in 200 mL of THF and cooled to -78 °C. This
solution was stirred for 3 h. A 4-fold excess of ethylene oxide was
then added via cannula. The reaction mixture was allowed to warm to
room temperature and stirred for 18 h before quenching by addition of
100 mL of water. The THF was removed by rotary evaporation, and
the product, 4-(3-hydroxypropyl)-4′-methyl-2,2′-bipyridine, was ex-
tracted into methylene chloride (100 mL; 3×) from the aqueous slurry.
The methylene chloride fractions were combined and brought to dryness
by rotary evaporation. The solid residue was redissolved in ca. 150
mL of ethyl acetate and cooled to 0 °C, whereupon much of the
unreacted DMB precipitated from solution while the product remained
dissolved. Ethyl acetate was removed by rotary evaporation at room
temperature, yielding the product contaminated with DMB. This crude
product was purified by “flash” silica gel chromatography, eluting first
with 10% ethyl acetate in methylene chloride until all DMB was off
the column. The product was then eluted from the column by gradually
increasing the mobile phase polarity by changing the ethyl acetate/
methylene chloride ratio. After rotary evaporation of solvent, the
remaining white solid, I, was vacuum-dried overnight to yield 5.1 g of
product (41%); 1H NMR (CDCl3) δ -CH2- 1.9 (quintet), CH3 2.4 (s),
-CH2- 2.8 (t), -CH2OH 3.7 (t), aromatic 7.2, 8.3, 8.6.
Figure 2. C2+-A2+ diad, VI, used in bimolecular quenching experiments
with N-MePTZ donor.
1. The rate of intramolecular electron transfer between D and
C
3+ is inherently and significantly faster than that between C3+
and A•+, and there is nothing else unusual about these triads.
2. The rate of intramolecular electron transfer between D and
C
3+ is abnormally fast as a result of some type of ground-state
association involving D prior to photoexcitation.
To test these hypotheses, it is necessary to determine the effect
on ΦCSS of varying the relative concentrations of D and C.
Obviously, this cannot be done in an entirely intramolecular
system. Thus, in what follows, results are presented for the CSS
process in a bimolecular system consisting of an N-methylphe-
nothiazine (N-MePTZ) donor, D, and a C2+-A2+ diad. Figure
2 shows the C2+-A2+ moiety, VI, chosen for the present study.
The C2+ and A2+ components differ somewhat from those
comprising the D-C2+-A2+ assemblies represented in Figure
1. First, the electron acceptor is a “paraquat-type” moiety rather
than a “diquat” (i.e., it is based on an N,N′-dialkyl-4,4′-
bipyridinium instead of an N,N′-dialkyl-2,2′-bipyridinium).22
Second, the two “remote” ligands on the ruthenium (i.e., the
ones not having an appended paraquat) are 1,10-phenanthroline
rather than 2,2′-bipyridine. Phenanthroline was chosen over
bipyridine for reasons that will be considered later but, suffice
it so say that, from the ruthenium’s perspective, the electronic
difference between phenanthroline and bipyridine is fairly minor
4-(3-Bromopropyl)-4′-methyl-2,2′-bipyridine, (II). 4-(3-Hydroxy-
propyl)-4′-methyl-2,2′-bipyridine was converted to II by refluxing for
12 h in a 1:1 mixture of aqueous HBr (48%) and acetic acid. The
reaction mixture was cooled, and the pH was brought into the range of
5-6 by addition of aqueous sodium bicarbonate. The product was
extracted into methylene chloride, which was then dried over magne-
sium sulfate and filtered. The solvent was removed by rotary evapora-
1
tion at room temperature, yielding a red oil (88%); H NMR (CDCl3)
δ -CH2- 2.2 (quintet), CH3 2.4 (s), -CH2- 2.8 (t), -CH2Br 3.4 (t),
aromatic 7.2, 8.3, 8.6.
[1-Methyl-4,4′-bipyridinium](PF6), (III). 4,4′-Bipyridine was stirred
with a 2-fold excess of iodomethane in ether for 24 h in the dark. The
resulting yellow precipitate, [1-methyl-4,4′-bipyridinium] iodide, was
filtered and washed with cold ether. Dissolving the iodide salt in water
and adding a saturated aqueous solution of ammonium hexafluoro-
phosphate dropwise yielded the hexafluorophosphate salt as a white
precipitate. The precipitate, III, was washed with cold water and
1
vacuum-dried (28%); H NMR (CD3CN) δ CH3 4.4 (s), aromatic 7.8,
2+
2+
(i.e., Ru(bpy)3 and Ru(phen)3 have very similar visible
spectra and redox properties).23 Regardless of the exact identity
of the acceptor (i.e., whether diquat or paraquat), the first
electron-transfer event following photoexcitation is oxidative
quenching.14,19,20 Paraquat was chosen as the acceptor on the
basis of the recombination kinetics of this initial electron-transfer
quenching product (vide infra).18,22
8.4, 8.8, 8.9.
[1-(3-(4′-Methyl-2,2′-bipyridin-4-yl)propyl)-1′-methyl-4,4′-bipy-
ridinediium](PF6)2, (IV). Quantities of 0.5 g of 4-(3-bromopropyl)-
4′-methyl-2,2′-bipyridine and 3.3 g (6 equiv) of [1-methyl-4,4′-
bipyridinium](PF6), III, were refluxed in 150 mL of butyronitrile under
nitrogen, excluding light, for 24 h. The yellow precipitate was
metathesized to the hexafluorophosphate salt by the procedure described
earlier. Yield of 0.54 g (46%); 1H NMR (CD3CN) δ -CH2-, CH3 2.4
(quintet + s), -CH2- 2.9 (t), N+CH3 4.4 (s), N+CH2 4.7 (t), aromatic
7.2-8.9.
The choice of donor is straightforward. The redox potentials
of N-alkylphenothiazines are essentially insensitive to the
identity of the alkyl group. A methyl group is the least sterically
demanding, and therefore N-MePTZ was employed as the donor.
Ru(1,10-phenanthroline)2Cl2, (V). A quantitiy of 1.0 g of Ru-
24
(DMSO)4Cl2 was dissolved in 25 mL of dimethylformamide which
was previously dried over alumina. This solution was saturated with
lithium chloride before adding 0.82 g (2 equiv) of 1,10-phenanthroline.
The reaction mixture was warmed to just below reflux and stirred for
1 h. Additional lithium chloride was then added, and the temperature
increased to full reflux. After 75 min of being refluxed, the solution
was cooled to room temperature and 60 mL of water was added. The
Experimental Section
Syntheses. Unless otherwise stated, all chemicals were obtained from
Aldrich and used without further purification. 4,4′-Dimethyl-2,2′-
bipyridine (DMB) was purchased from Reilley Industries and recrystal-
(22) Yonemoto, E. H.; Riley, R. L.; Kim, Y. I.; Atherton, S. J.; Schmehl, R.
H.; Mallouk, T. E. J. Am. Chem. Soc. 1992, 114, 8081-8087.
(23) Balzani, V.; Juris, A.; Barigelletti, F.; Campagna, S.; Belser, P.; von
Zelewsky, A. Coord. Chem. ReV. 1988, 84, 85-277.
(24) Wilkinson, G.; Evans, I. P.; Spencer, A. J. Chem. Soc., Dalton Trans. 1973,
204-209.
9
J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007 315

A R T I C L E S
Weber et al.
Scheme 1. Jablonsky Energy Level Diagram for Collisional
precipitated crude product, V, was filtered and vacuum-dried (17%)
and used in subsequent reactions without further purification; 1H NMR
(CD3CN) δ CH3 2.5 (s), aromatic 7.1-8.7.
(Left-Hand Branch) and Preassociation Mechanisms (Right-Hand
Branch) for CSS Formation in a C2+-A2+/D Bimolecular System
[Ru(1,10-phenanthroline)2(4-(3-(1′-methyl-4,4′-bipyridinediium-
1-yl)-propyl)-4′-methyl-2,2′-bipyridine)](PF6)4, (VI). Quantities of
79 mg of Ru(phen)2Cl2 and 100 mg (1 equiv) of [1-(3-(4′-methyl-2,2′-
bipyridin-4-yl)propyl)-1′-methyl-4,4′-bipyrinediium](PF6)2, IV, were
refluxed in methanol for 18 h. The reaction vessel was kept in the
dark under nitrogen for the duration of the reaction. Upon being cooled
to room temperature, a precipitate formed that was filtered off and
discarded. A solution of methanol saturated with ammonium hexafluo-
rophosphate was added dropwise to precipitate the crude product, VI,
as the hexafluorophosphate salt. “Flash” silica gel chromatography was
used to isolate the product. The “full strength” elution solvent was 50:
40:10 water/acetonitrile/saturated aqueous KNO3. A gradient elution
with the mobile phase was required where the full strength elution
solvent was diluted at each stage with 1:1 acetonitrile/water. The crude
product was dissolved in neat acetonitrile and loaded on to ap-
proximately 5 in. of flash silica gel in a 1-in. diameter column. The
column was eluted with progressively less dilute solvent starting with
a 20-fold dilution and ending with “full strength”. Fractions containing
only the desired product (as determined by TLC) were combined, and
the acetonitrile was removed by rotary evaporation at just above room
temperature. The product was metathesized to the hexafluorophosphate
salt, and the resulting precipitate was washed with water. Yield of 66
passing through a mechanical chopper operated at 30 Hz with slits such
that light was transmitted about 3% of the time. For the Opotek OPA
system, the chopper wheel was operated at 20 Hz. The input resistance
of the oscilloscope was either 50 or 100 Ω. The photodiode output
from the chopper was used as the trigger signal for the Nd:YAG laser.
A Hamamatsu model R2496 photomultiplier tube (PMT) was employed
to monitor the intensity of the probe beam. A Tektronix TDS 620B
digital oscilloscope recorded the current output from the PMT for the
first 500 ns following excitation by the pump beam. The oscilloscope
was triggered with a Thorlabs DET210 fast photodiode. The results
from 200 pulses were averaged. Differential absorbance due to the
reduced viologen acceptor was calculated using Beer’s Law.
Emission Lifetime Measurements. Emission lifetimes of VI in
dichloromethane were obtained employing time correlation single
photon counting as described by O’Connor and Phillips.25 A Ti:sapphire
regenerative amplifier (Coherent RegA) produced pulses centered at
866 nm with a 200 kHz repetition rate and 650 µJ/pulse. The light was
frequency doubled to promote the MLCT transition. The subsequent
fluorescence relaxation was collected through a subtractive double
monochromator. Single photons impinging on a microchannel plate
detector (Hamamatsu, MCP-PMT) were detected and digitized using
an analog-to-digital converter (Oxford ADC) and served as the start
pulse for a time-to-amplitude converter (Oxford, TAC). A small fraction
of the excitation light was split off and directed to a fast photodiode
(ThorLabs). This served as the stop pulse for the TAC. Signals from
the TAC were fed into a multichannel analyzer (Oxford) whose output
was monitored by a computer. Decay traces were collected until a
sufficient signal-to-noise ratio was achieved to allow adequate data
analysis.
1
mg (30%); H NMR (CD3CN) δ -CH2- 2.4 (quintet), CH3 2.5 (s),
-CH2- 2.9 (t), N+CH3 4.4 (s), N+CH2 4.7 (t), aromatic 7.1-8.9; GC/
MS (ES+) m/z calcd 1279.15, found 1279.00.
[Ru(1,10-phenanthroline)2(4,4′-dimethyl-2,2′-bipyridine)]-
(PF6)2, (VII). Quantities of 100 mg of V and 34 mg of DMB were
dissolved in 25 mL of ethanol and brought to reflux for 20 h. Ethanol
(15 mL) was removed by rotary evaporation before 30 mL of water
was added. The product was metathesized to the hexafluorophosphate
salt as described and recrystallized twice from methanol. Yield of 26
1
mg (25%); H NMR (CD3CN) δ CH3 2.5 (s), aromatic 7.1-8.7; GC/
MS (ES+) m/z calcd 791.14, found 791.07.
Preparation of Samples. Sample solutions were prepared in a dark
room and kept from light until frozen for degassing. A stock solution
of VI in dichloromethane (Aldrich, Spectro Grade) was prepared by
adding solvent to a small amount of the solid complex until the
absorbance at the maximum of the MLCT band monitored at 450 nm
in a 1-cm cell was approximately 0.15 (ca. 10-5 M). The appropriate
amount of solid N-MePTZ to obtain each concentration desired was
weighed into vials, and 3 mL of stock solution of VI was added to
each vial. Samples were then transferred to an optical cell constructed
with a sidearm for freeze-pump-thaw (FPT) degassing, a Teflon valve,
and a fitting for attachment to a vacuum line. Solution in the cell was
transferred to the sidearm and frozen in liquid nitrogen. Each sample
was FPT degassed several times until there was no evidence of
outgassing upon thawing.
Transient Absorption Measurements. The initial amount of CSS
formed was determined by transient absorption spectroscopy. The 355-
nm tripled output of a Spectra-Physics Lab-190 Nd:YAG laser
producing 8-ns pulses at 30 Hz with a power of approximately 2.0 W
was used to pump Coumarin 450 dye in a Spectra-Physics PDL-3 dye
laser. The dye laser output at 450 nm was used to excite the sample.
Typical output powers were approximately 70 mW. Alternately, an
Opotek OPA was used to provide the 450-nm excitation beam at powers
of 50-55 mW. An Oriel 75 W xenon arc lamp was employed as a
continuous probe source.
Solution Viscosity Determination. Solution viscosities were mea-
sured using Cannon-Ubbelohde type viscometers from Cannon Instru-
ment Company.
Results and Discussion
The first goal in this study is to establish which of the two
hypotheses best explains the observed CSS formation in
D-C2+-A2+ assemblies. In the form of a Jablonsky diagram,
Scheme 1 allows simultaneous examination of both of the
limiting mechanistic possibilities.
The bottom center of Scheme 1 depicts a ground-state
association equilibrium between the donor, D, and the C2+
-
A2+ diad. For the sake of discussion, this association is
represented in the scheme as if it were occurring between D
and C2+; however, the nature of any association has yet to be
established and this issue will be revisited subsequently. In
considering the two limiting cases represented by the left- and
right-hand branches of Scheme 1, we assume that the CSS
The reduced viologen acceptor radical has an absorbance maximum
at 397 nm. Prior to passing the probe light through the sample, this
wavelength was isolated from the xenon lamp’s continuous output using
a Johnson and Johnson monochromator. Subsequent to passing through
the sample cell, the probe beam was collimated and then focused onto
the entrance slit of a Jarrell Ash model 82-410 monochromator after
(25) O’Connor, D. V.; Phillips, D. Time-Correlated Single-Photon Counting;
Academic Press: London, 1984.
9
316 J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007

High Energy and Quantum Efficiency in Charge Separation
A R T I C L E S
confidence that the data in Figure 3A truly follow the functional
form expressed in eq 3 (vide infra). In dichloromethane, it is
possible to reach D concentrations yielding ∆A°397 values much
closer to the maximum predicted from eq 3. As with acetonitrile,
the fit at lower [D] to eq 3 is excellent, but above ca. 20 mM
there is some scatter.27 However, if only the data below 20 mM
are used in the fit to eq 3, almost exactly the same values of a
and b result (i.e., the fit constants) as when the entire data set
is used.
In a mechanism involving only bimolecular diffusional
encounters between the N-MePTZ donor and C3+-A•+ (the left-
hand side of Scheme 1), the resulting relationship connecting
[CSS] and [D] takes the form:
[CSS]0
[CA]i
(kcs/kbet)[D]
)
(1)
1 + (kcs/kbet)[D]
where [CA]i is the total concentration of the C2+-A2+ diad,
[CSS]0 is the concentration of CSS at t ) 0, kcs is the
bimolecular rate constant for forming the charge separated state,
and kbet is the rate of geminate recombination between C3+
A•+ (Scheme 1).
-
If, on the other hand, ground-state association between D and
C
2+-A2+ was the sole mechanism responsible for CSS forma-
tion (the right-hand side of Scheme 1), the following relationship
results:
[CSS]0
[CA]i
(kcs/kbet)
Keq[D]
)
‚
(2)
1 + Keq[D]
1 + (kcs/kbet)
Figure 3. ∆A°397 (proportional to the amount of charge separated state
initially formed) plotted vs [N-MePTZ] in acetonitrile (A) and dichlo-
romethane (B) solvents. The right-hand y axis is the amount of CSS formed
expressed as a percentage of CSSmax as predicted from eq 3 (see text for
explanation).
where Keq is the complexation equilibrium constant. Again, in
deriving eq 2, it is assumed that an equilibrium association
between D and C2+-A2+ is established prior to excitation and
that CSS is only formed from that part of the total population
of C2+-A2+ that is associated with D in the ground state.
Additionally, kbet and kcs are, respectively, first-order rate
constants of geminate recombination and charge separation
relevant to the associated complex. In this instance, kbet may or
may not be the same as kbet in the absence of D/C2+-A2+
association (i.e., as in eq 1).
In either of the above scenarios, the governing relation
between [CSS]0 and [D] has the same functional (i.e., math-
ematical) form. Also, since [CSS]0 is directly proportional to
∆A°397, both expressions can be written in the general form:
exclusively results through either a diffusional mechanism or a
preassociation, not both.26
Transient spectroscopy was utilized to monitor the absorbance
(397 nm) of the reduced viologen radical (A•+) formed upon
photoexcitation (450 nm) of a C2+-A2+ diad as a function of
[N-MePTZ]. First, in the absence of N-MePTZ there is no signal
for reduced viologen. Thus, it can be assumed that the only
viologen radical observed comes from CSS production. In other
words, it is only when the donor reduces the oxidized chro-
mophore prior to geminate recombination that the CSS (and
thus reduced viologen) is observed. The lifetime of the bimo-
lecular CSS is on the order of tens of microseconds. Therefore,
the change in absorbance at t f 0 (i.e., ∆A°397) is directly
proportional to the amount of CSS initially formed.
The results obtained for CSS formation in acetonitrile and
dichloromethane solvents are shown in Figure 3A,B, respec-
tively, along with fits of the data to eq 3 (vide infra). While the
range of [D] available in acetonitrile only permits about 40%
of the predicted (from eq 3) maximum ∆A°397 (and, thus the
maximum [CSS]) to be accessed, the quality of the fit is very
good. The value of R2 is 0.9996, which gives us considerable
∆A°397 ) a{[D]/(b + [D])}
(3)
Thus, the only differentiating factors in the mathematical
relations governing the two limiting mechanisms are the
meaning of the constants. In both cases, the constant a depends
on, among other things, experimental parameters such as the
laser power and overlap of the pump and probe beams. For the
sake of the work presented here, a can be considered to be an
arbitrary constant in both models. In contrast, b is independent
of similar experimental parameters (as long as they are constant).
In the case of a simple bimolecular collisional reaction,
(26) It is also assumed that the extinction coefficient at 450 nm is the same for
free C2+-A2+ and D/C2+-A2+. This assumption is necessary to assure
that the same relative fractions C2+-A2+ and D/C2+-A2+ are excited as
exist in the equilibrium solution. The validity of this assumption can be
defended on the basis of the very minor change observed in the MLCT
spectrum upon addition of N-MePTZ (see Figure 6).
(27) At the higher [D] there is moderate absorbance of N-MePTZ at 397 nm
that lowers the signal-to-noise of the measurement and likely contributes
to the scatter in ∆A397. In comparing the data in Figure 3A,B, it is
noteworthy that at a given [N-MePTZ], the [CSS] produced relative to the
maximum amount predicted, [CSS]max, differs significantly for the two
solvents (see also Supporting Information).
9
J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007 317

A R T I C L E S
Weber et al.
b ) kbet/kcs
(4)
(5)
In the case of an equilibrium preassociation,
-1
b ) Keq
Mallouk et al. previously reported values for both ket and kbet
obtained in acetonitrile for a C2+-A2+ complex very similar
to VI.28 Values of the forward electron-transfer rate constant,
ket, for Mallouk’s complex and for VI (in dichloromethane) are
similar (1.9 × 109 and 8.2 × 1010 s-1, respectively). It is
reasonable to assume, therefore, that Mallouk et al.’s value of
kbet (6.5 × 109 s-1) is a fair lower-limit estimate for the
analogous process in VI. Assuming the bimolecular collisional
mechanism and using Mallouk’s value for kbet and the value of
b obtained from the fit to eq 3, a lower limit of 3.9 × 1010 M-1
s-1 is obtained for kcs. Using the standard relation29
Figure 4. ∆A°397 (proportional to the amount of charge separated state
initially formed) in the presence of 50 mM N-MePTZ plotted vs solution
viscosity. The error bars represent one standard deviation.
effectively rules out any CSS formation mechanism whereby
simple collisional encounters between D and C3+-A•+ are
involved (i.e., hypothesis 1). If the process is not collisional,
then the only reasonable possibility remaining is the mechanism
represented in the right-hand side of Scheme 1 and by hypothesis
2. However, none of the results considered thus far addresses
the nature of the preassociation ground state. To shed light on
this issue, other data must be considered.
8RT
3η
kdiff
)
(6)
we calculated an estimate for the diffusion-controlled limiting
rate constant of CSS formation in acetonitrile to be 2 × 1010
M-1 s-1, where η is the viscosity of acetonitrile at room
temperature. The lower limit for kcs obtained from the fit to eq
3 is thus greater than the estimated diffusion controlled rate
constant by a factor of ca. 2. The same calculations performed
on the dichloromethane data yield a lower-limit value for kcs of
2.1 × 1011 M-1 s-1, and an estimated diffusion-controlled rate
constant is 1.7 × 1010 M-1 s-1 (i.e., kcs is too large by an order
of magnitude). Both calculations depend on the assumption that
In looking into the nature of the association, it is helpful at
this point to construct two additional hypotheses:
2a. The association between D and C2+-A2+ leading to CSS
formation is between D and C2+ (D/C2+), most likely π-π van
der Waals in nature.
2b. The association between D and C2+-A2+ leading to CSS
formation is between D and A2+ (D/A2+), possibly both charge
transfer and π-π van der Waals in nature.
Mallouk et al.’s value is a reasonable lower-limit estimate for
Distinguishing between these hypotheses has proven more
challenging than simply demonstrating the participation of the
preassociation in CSS formation. While there is literature
precedent for both D/A2+- and D/C-type associations (at least
in loosely related systems),33,34 there is no single piece of data
that unambiguously demonstrates the nature of the association;
rather, consideration of a collection of data from several
experiments is necessary to develop a case. First, spectral
evidence for any association between the two pairs of moieties
(i.e., between D/A2+ and D/C2+) was examined. Neither D nor
30-32
kbet
.
The observations that (1) irrespective of solvent, we
were unable to observe any photoreduced acceptor with VI in
the absence of N-MePTZ and (2) back electron transfer rates
for related compounds are typically not strongly solvent
dependent are facts fully consistent with a kbet > 6.5 × 109 s-1
for VI.
The results that the values of kcs exceed the diffusion-
controlled limit in both solvents are strongly suggestive that
the mechanism of CSS formation is not diffusional; however,
a much stronger argument can be made by examining the
dependence of ΦCSS on solution viscosity. A series of stock
solutions containing ca. 10-5 M C2+-A2+ complex (VI) and
50 mM N-MePTZ were prepared in dichloromethane solvent.
To these solutions, 0-80 mg/mL of polystyrene was added to
increase solution viscosities. A corresponding weight of toluene
was also added such that the (polystyrene + toluene)/solvent
ratio was held constant. The viscosity of each solution was
measured five times, and the averages ranged from 0.46 to 12.1
cP. Figure 4 shows the dependence of ∆A°397 on viscosity.
These data demonstrate conclusively that the amount of
[CSS]0 formed is independent of solution viscosity. Clearly, a
ca. 25-fold increase in viscosity with no change in [CSS]0
A
2+ absorbs light in the visible, but both absorb strongly in the
UV. If any degree of interaction between these species were
occurring in solution, some perturbation in UV spectra of the
two moieties should be evident, especially if there is any charge-
transfer nature to that interaction. Figure 5 shows that the
absorption bands of the donor and acceptor are unperturbed by
mixing the solutions. This is most clearly demonstrated by the
fact that the spectrum resulting from the mathematical sum of
the individual components exactly overlays the spectrum
obtained from mixing the donor and acceptor solutions together
(within experimental error).
In contrast to the data in Figure 5, there is a small but clear
perturbation in the visible spectrum of combined solutions of
D and C2+. Attempts to quantify the intermolecular interaction
via titration are complicated by overlap from the intense MLCT
transitions and the weak nature of the optical signal resulting
from the D/C2+ interaction; nevertheless, the qualitative effect
(28) Yonemoto, E. H.; Saupe, G. B.; Schmehl, R. H.; Hubig, S. M.; Riley, R.
L.; Iverson, B. L.; Mallouk, T. E. J. Am. Chem. Soc. 1994, 116, 4786-
4795.
(29) Atkins, P. W. Physical Chemistry, 5th ed.; W. H. Freeman: New York,
1994.
(30) Ohno, T.; Yoshimura, A.; Prasad, D. R.; Hoffman, M. Z. J. Phys. Chem.
1991, 95, 4723-4728.
(31) Sun, H.; Hoffman, M. Z. J. Photochem. Photobiol., A 1994, 84, 97-99.
(32) Linsenmann, M. Ph.D. Dissertation, University of Konstanz, Konstanz,
Germany, 1997.
(33) Deronzier, A.; Essakalli, M. J. Phys. Chem. 1991, 95, 1737-1742.
(34) McClenaghan, N. D.; Loiseau, F.; Puntoriero, F.; Serroni, S.; Campagna,
S. Chem. Commun. 2001, 2634-2635.
9
318 J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007

High Energy and Quantum Efficiency in Charge Separation
A R T I C L E S
Figure 7. Time-resolved emission decays for a ca. 1 × 10-5 M solution
of VI in oxygen-free dichloromethane.
Time-resolved emission spectra of C2+-A2+ VI as a function
of added N-MePTZ also assist in understanding the nature of
the association between D and C2+-A2+. Figure 7 shows the
emission decay of a ca. 1 × 10-5 M dichloromethane solution
of VI in the presence of 0, 20, and 120 mM N-MePTZ,
respectively. In the absence of any D, the lifetime of the
emission is determined entirely by the oxidative quenching
process. Obvious from the figure, however, is that addition of
Figure 5. Visible spectra in acetonitrile of a 1.5 mM solution of N-MePTZ
(long-dashed curve), 1.5 mM solution of methyl viologen (short-dashed
curve), and a solution that is 1.5 mM in both N-MePTZ and methyl viologen
(solid curve). Superimposed on the solid spectrum is the sum of the two
dashed spectra (i.e., the spectra of the isolated components). The lower
part of the figure is the difference spectrum obtained by subtracting each
spectrum of the individual components from the spectrum of the mixture.
N-MePTZ increases the average emission lifetime of C2+*
2+. From biexponential fits, curves B and C in Figure 7 each
-
A
consist of the same two lifetime components (i.e., 3.8 ( 0.2
and 0.72 ( 0.03 ns). While the lifetimes are within a similar
experimental error, the relative contributions of each are
different. When the data in Figure 7 is compared with that in
Figure 3B, an interesting and relevant correlation emerges. As
stated previously, the right-hand y axis of Figure 3A,B represents
the fraction of CSS formed relative to the theoretical maximum
CSS. That axis also represents the fraction of the total C2+
2+ which is tied up as D/C2+-A2+. Expressed as percentages,
at 20 and 120 mM N-MePTZ, respectively, 40 and 80% of the
2+-A2+ is associated with D. From the fits of the data in
-
A
C
Figure 7, the slow component contributes 45 and 70% to the
total emission at 20 and 120 mM N-MePTZ, respectively. The
correlation suggests that the slow component of the emission
decay arises from the fraction of the C2+-A2+ population
Figure 6. Visible spectra of a ca. 3 × 10-5 M solution of VII in
dichloromethane (solid curve) and the same solution containing 50 mM
N-MePTZ (dashed curve). The dashed curve was corrected for any small
absorption of N-MePTZ by subtracting the spectrum of a 50 mM N-MePTZ
solution in dichloromethane. The lower part of the figure is the difference
spectrum obtained by subtraction of the dashed spectrum from the solid
spectrum.
associated with D. Since conformational flexibility of the C2+
-
A2+ linkage appears to be required for the initial oxidative
quenching,18 were D associated with A2+ (rather than C2+),
much larger changes in emission lifetime should result than are
observed. The difference in hydrodynamic radii between A2+
and D/A2+ should be large and should produce large changes
in the quenching rate. In contrast, an association between C2+
and D might sterically restrict some encounters between A2+
and C2+ but only along a limited fraction of possible approach
trajectories resulting in a more modest change in emission
lifetime (as is observed).
of added N-MePTZ on the MLCT band of VII can be clearly
seen in the spectra of Figure 6. Similar evidence of a D/C2+
interaction comes from comparisons among spectra from D-C2+
diads, D-C2+-A2+ triads, and isolated C2+, D, and A2+
moieties. We have noted in earlier publications that, whenever
a complex has incorporated one or more D-containing ligands,
it is never possible to exactly simulate the spectrum by summing
weighted spectra of the respective C2+, D, and A2+ compo-
nents.13,21 Difference spectra from those attempts look quali-
tatively similar to those in Figure 6.35
Finally, if the association of D and C2+-A2+ were between
D and A2+, the formation constant, Keq, should be quite sensitive
to the structure and redox potential of the acceptor. A titration
with N-MePTZ in dichloromethane was carried out on an
analogue of IV in which only the acceptor ligand was changed;
specifically, the paraquat was replaced with the same diquat
ligand as shown in Figure 1. Within experimental error, the
(35) Limoges, B. Ph.D. Dissertation, Colorado State University, Fort Collins,
CO, 2001.
9
J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007 319

A R T I C L E S
Weber et al.
Scheme 2. Depiction of the Phenothiazine Electron Donor Moeity
Associated with the Local Bipyridine Ligand in a Triad Assembly
Such as That in Figure 1
resulting titration is identical to that in Figure 3B (see Supporting
Information; cf. Figures 3B and SM2). It is difficult, indeed, to
rationalize how a Keq for association between D and A2+ could
be the same given such dramatic differences in A2+ structure
(i.e., rigidly twisted versus freely rotating around the π-π bond),
especially in light of the strong solvent dependence demonstrated
in Figure 3A,B.
To summarize, it seems quite clear from these collective
results that CSS formation is dependent on a ground-state
preassociation of the donor with the chromophore. Were there
a significant charge-transfer component to this association
process, differences in redox potentials would predict that a
D/A2+ association should have a significantly more favorable
enthalpy than would a D/C2+ association. If, on the other hand,
the association is driven predominantly by van der Waals
interactions and entropy, there is no a priori reason why one
pairwise association should be favored over the other. The very
minor changes in the absorption spectrum (Figure 6) and in time-
resolved emission spectrum (Figure 7) produced by bimolecular
association of N-MePTZ and IV, and the similarly small
absorption and time-resolved emission differences between
whereby systems might be designed to give efficient photoin-
duced charge separation without automatically sacrificing the
quantity of energy stored in the CSS. Typically, the electronic
coupling between donor and acceptor orbitals does not change
dramatically upon electron transfer; consequently, an efficient
forward electron transfer usually means an efficient recombina-
tion. Therefore, with a multistep photoinduced charge separation
system, increasing ΦCSS often comes with a concomitant cost
in stored energy. For D-C2+-A2+ triads such as that in Figure
1, the flexible linkage allows D/C2+ to associate prior to any
electron transfer, thus enhancing the D/C2+ orbital coupling,
and to turn off the enhanced coupling by disassociation once
the donor is oxidized (because of electrostatic repulsion; see
Supporting Information for a more detailed consideration of this
issue). Consequently, ΦCSS is largely independent of driving
force and, thus, of the energy stored in the CSS.
Considered in light of the present results, these specific triad
assemblies provide the first example of which we are aware
where a change in orbital coupling of this magnitude is induced
by an intramolecular electron transfer, a change large enough
to swamp the thermodynamic factors that usually control the
charge separation kinetics. This suggests that related self-
assembly processes could be intentionally designed and ex-
ploited to effect efficient CSS formation. Of particular impor-
tance is, in principle, that efficient photoinduced charge
separation via such a mechanism should be possible without
the loss in stored energy that most often accompanies efforts
to use driving-force manipulations to improve CSS formation
quantum efficiencies.
C
2+-A2+ diads and D-C2+-A2+ triads,12-14 all suggest that
there is not a large enthalpic contribution for either the inter-
or intramolecular association. The equilibrium may, therefore,
be largely driven by entropy, but that is yet to be determined.
Finally, the question must be addressed of how the results
from the bimolecular system actually inform about behavior in
triad assemblies. First, consideration of molecular models
demonstrates that intramolecular D/C2+ associations are steri-
cally possible without significant strain energy. In the triad
systems, the effective local concentration of D in the vicinity
of the C2+ is much higher than can be achieved in the
bimolecular system. This inherent concentration bias is the
reason for using phenanthroline ligands, with their extended
π-systems, in IV rather than bipyridines. The argument that an
analogous intramolecular D/C2+ association occurs in the triad
assemblies is further supported by the fact that difference spectra
between C2+-A2+ diads and D-C2+-A2+ triads are qualita-
tively similar to the difference spectrum in Figure 6. Last,
D-C2+-A2+ triads incorporating phenoxazine (POZ) rather
than phenothiazine donors form photoinduced CSS with equiva-
1
lently large ΦCSS values. Two-dimensional H-1H ROESY
NMR experiments performed on the POZ analogue of the
D-C2+-A2+ triad shown in Figure 1 show clear evidence of
spin exchange between the POZ and bipyridine protons,
indicating their close average proximity as would be expected
from the D/C2+ associations (see Supporting Information).
Further, no cross-peaks are seen between the diquat acceptor
and any species other than itself.
Acknowledgment. We gratefully acknowledge support of this
work through a grant from the U.S. DOE Office of Science
(DE-FG02-04ER15591). We also thank Dr. Chris Rithner for
assisting with the 2D NMR experiment.
Supporting Information Available: 2D ROESY NMR spec-
tra demonstrating correlation of the donor protons with the
chromophore ligand protons as well as association data in mixed
solvents and an analogous CA2+ with N-MePTZ system. This
material is available free of charge via the Internet at
http://pubs.acs.org.
Conclusion
Scheme 2 represents, in cartoon form, how the D/C2+
association plays the critical role in intramolecular CSS forma-
tion for triads such as that in Figure 1. This mechanism provides
a plausible explanation for the large CSS formation quantum
efficiencies observed uniformly across this class of triads, but
most importantly, it provides insight into a general means
JA0665500
9
320 J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007