Long-Lived Photoinduced Charge Separation and Redox-Type Photochromism on Mesoporous Oxide Films Sensitized by Molecular Dyads

Article abstract of DOI:10.1021/ja981742j

The photoinduced charge separation in three different assemblies composed of an electron donor D and a chromophore sensitizer S adsorbed on nanocrystalline TiO₂ films (D-S|TiO₂) was investigated. In all of the systems, the sensitizer was a ruthenium(II) bis-terpyridine complex anchored to the semiconductor surface by a phosphonate group. In two of the assemblies, the donor was a 4-(N,N-di-p-anisylamino) phenyl group linked to the 4′ position of the terpyridine, either directly (dyad D1-S) or via a benzyl ether interlocking group (dyad D2-S). In the third system, the sensitizer and the donor (3-(4-(N,N-di-p-anisylamino)phenoxy)-propyl-1-phosphonate) were coadsorbed on the surface ((D3+S)|TiO₂). Laser flash photolysis showed that the photoinduced charge separation process follows the sequence D-S*|TiO₂ ¹→ D-S⁺|(e^-)TiO₂ ²→ D⁺-S|(e^-)TiO₂ ³→ D-S|TiO₂ Resonance Raman spectroscopy indicates that in the excited assemblies D2-S*|TiO₂ and (D3+S*)|TiO₂, one electron is promoted from the metal center to the terpyridine ligand linked to the semiconductor, whereas in the system D1-S*|TiO₂ the excited electron is located on the ligand linked to the donor. The quantum yield of charge separation (steps 1 and 2) was found to be close to unity for the two former assemblies but only 60% for the latter one. In all three cases, the electron injection was very fast (<1 ns), and the hole transfer to the donor was fast (10-20 ns). The half-lifetime of the charge separated state (step 3) was 3 μs for (D3⁺+S)|(e^-)TiO₂, as in the model system S⁺|(e^-)TiO₂; it was 30 μs in D1⁺-S|(e^-)TiO₂ and 300 μs in D2⁺-S|(e^-)TiO₂. Electrodes made of any of the surface-confined dyads on conducting glass display a strong redox-type photochromism. When a positive potential (+0.5 V vs NHE) is applied to the electrode, charge recombination (step 3) is blocked. As a result, the visible absorption spectrum of the electrode changes, due to the appearance of the absorption feature of the oxidized donor (λ_max = 730 nm). Return to the reduced state is achieved by electron injection through the conduction band of the TiO₂ under forward bias (-0.5 V). None of the assemblies D1-S|TiO₂ and D2-S|TiO₂ gave better photovoltaic performances than the model system S|TiO₂. This was attributed in the first case to the low injection efficiency and, in the second case, to an additional short-circuiting pathway constituted by the charge percolation inside the molecular monolayer and to the underlying conducting glass, as previously observed with monolayers of the donor D3 (Bonhote, P.; Gogniat, E.; Tingry, S.; Barbe, C.; Vlachopoulos, N.; Lenzmann, F.; Comte, P.; Graetzel, M. J. Phys. Chem. B 1998, 102, 1498-1507).

Full text of DOI:10.1021/ja981742j

1324
J. Am. Chem. Soc. 1999, 121, 1324-1336
Long-Lived Photoinduced Charge Separation and Redox-Type
Photochromism on Mesoporous Oxide Films Sensitized by Molecular
Dyads
Pierre Bonhoˆte,* Jacques-E. Moser, Robin Humphry-Baker, Nicolas Vlachopoulos,
Shaik M. Zakeeruddin, Lorenz Walder,§ and Michael Gra1tzel
Contribution from the Laboratoire de Photonique et Interfaces, De´partement de Chimie,
Ecole Polytechnique Fe´de´rale de Lausanne, CH-1015 Lausanne, Switzerland
ReceiVed May 19, 1998
Abstract: The photoinduced charge separation in three different assemblies composed of an electron donor D
and a chromophore sensitizer S adsorbed on nanocrystalline TiO2 films (D-S|TiO2) was investigated. In all
of the systems, the sensitizer was a ruthenium(II) bis-terpyridine complex anchored to the semiconductor surface
by a phosphonate group. In two of the assemblies, the donor was a 4-(N,N-di-p-anisylamino) phenyl group
linked to the 4′ position of the terpyridine, either directly (dyad D1-S) or via a benzyl ether interlocking
group (dyad D2-S). In the third system, the sensitizer and the donor (3-(4-(N,N-di-p-anisylamino)phenoxy)-
propyl-1-phosphonate) were coadsorbed on the surface ((D3+S)|TiO2). Laser flash photolysis showed that the
1
2
photoinduced charge separation process follows the sequence D-S*|TiO2 f D-S+|(e-)TiO2
f
3
D+-S|(e-)TiO2 f D-S|TiO2 Resonance Raman spectroscopy indicates that in the excited assemblies
D2-S*|TiO2 and (D3+S*)|TiO2, one electron is promoted from the metal center to the terpyridine ligand
linked to the semiconductor, whereas in the system D1-S*|TiO2 the excited electron is located on the ligand
linked to the donor. The quantum yield of charge separation (steps 1 and 2) was found to be close to unity for
the two former assemblies but only 60% for the latter one. In all three cases, the electron injection was very
fast (<1 ns), and the hole transfer to the donor was fast (10-20 ns). The half-lifetime of the charge separated
state (step 3) was 3 µs for (D3++S)|(e-)TiO2, as in the model system S+|(e-)TiO2; it was 30 µs in
D1+-S|(e-)TiO2 and 300 µs in D2+-S|(e-)TiO2. Electrodes made of any of the surface-confined dyads on
conducting glass display a strong redox-type photochromism. When a positive potential (+0.5 V vs NHE) is
applied to the electrode, charge recombination (step 3) is blocked. As a result, the visible absorption spectrum
of the electrode changes, due to the appearance of the absorption feature of the oxidized donor (λmax ) 730
nm). Return to the reduced state is achieved by electron injection through the conduction band of the TiO2
under forward bias (-0.5 V). None of the assemblies D1-S|TiO2 and D2-S|TiO2 gave better photovoltaic
performances than the model system S|TiO2. This was attributed in the first case to the low injection efficiency
and, in the second case, to an additional short-circuiting pathway constituted by the charge percolation inside
the molecular monolayer and to the underlying conducting glass, as previously observed with monolayers of
the donor D3 (Bonhoˆte, P.; Gogniat, E.; Tingry, S.; Barbe´, C.; Vlachopoulos, N.; Lenzmann, F.; Comte, P.;
Gra¨tzel, M. J. Phys. Chem. B 1998, 102, 1498-1507).
Introduction
centers and therefore incorporated porphyrins as the chro-
mophore.2 On the other hand, charge separation at chro-
mophore-semiconductor interfaces (heterodyad S|A) exploited
on films of nanocrystalline metal oxides has allowed the
development of efficient dye-sensitized nanocrystalline solar
cells.3 Combining the intramolecular with the interfacial light-
induced charge separation strategies is expected to increase the
light-to-electricity conversion yield of the photovoltaic systems.
In the dye-sensitized nanocrystalline solar cells, the sensitizer
Understanding light-induced charge separation in natural or
artificial molecular systems and designing such systems able
to efficiently convert light into chemical or electrical energy is
a fascinating challenge for contemporary chemistry. Up to now,
charge separation resulting from light absorption by molecular
chromophores has been studied, exploiting the following two
parallel strategies: on one hand, numerous molecular assemblies
of donors (D), chromophores (S), and acceptors (A) in many
variations have been synthesized, forming dyads, triads, tetrads,
or even pentads, to achieve charge separation in solution or in
monolayers.1,2 In most cases, the studied supramolecular entities
were aiming at mimicking the natural photosynthetic reaction
(2) (a) Osuka, A.; Marumo, S.; Mataga, N.; Taniguchi, S.; Okada, T.;
Yamazaki, I.; Nishimura, Y.; Ohno, T.; Nozaki, K. J. Am. Chem. Soc. 1996,
118, 155-168 (b) Wiederrecht, G. P.; Niemczyk, M. P.; Svec, W. A.;
Wasielewski, M. R. J. Am. Chem. Soc. 1996, 118, 81-88 (c) Wasielewski,
M. R. Chem. ReV. 1992, 92, 435. (d) Connolly, J. S.; Bolton, J. R. In
Photoinduced Electron Transfer, Part D; Fox, M. A., Chanon M., Eds;
Elsevier: Amsterdam, 1988; p 303. (e) ibid. Part A; p 161. (e) Harrimann,
A.; Odobel, F.; Sauvage, J.-P. J. Am. Chem. Soc. 1994, 116, 5481-5482.
(f) Collin, J.-P.; Harrimann, A.; Heitz, V.; Odobel, F.; Sauvage, J.-P. J.
Am. Chem. Soc. 1994, 116, 5679-5690.
§ Present address: Institut fu¨r Chemie, Barbarastrasse 7, 49069 Osnabru¨ck,
Germany.
(1) (a) Balzani, V.; Scandola, F. Supramolecular Photochemistry; Ellis
Horwood: New York, 1991. (b) Lehn, J.-M. Angew. Chem. 1990, 102,
1347 and references cited in ref 6.
10.1021/ja981742j CCC: $18.00 © 1999 American Chemical Society
Published on Web 01/27/1999

Oxide Films Sensitized by Molecular Dyads
J. Am. Chem. Soc., Vol. 121, No. 6, 1999 1325
is grafted by a suitable anchoring group (carboxylate,3c
phosphonate,3e,f salicylate,4 acetylacetonate5) on the surface of
a semiconducting mesoporous oxide. Light excites the dye
(S|TiO2 f S*|TiO2), which then injects an electron into the
conduction band of the semiconductor (S*|TiO2 f S+|(e-cb)TiO2).
The reduced state S is regenerated by a reversible redox mediator
M present in the cell electrolyte. The maximal voltage delivered
by the device corresponds to the difference between the quasi-
Fermi level of the electrons in the semiconductor and the redox
potential of the mediator. A higher cell voltage can hence be
obtained either by lowering the quasi-Fermi level or by raising
the redox potential of the M/M+ couple. Under open circuit,
for a given semiconductor, the quasi-Fermi level in the cell
depends on the electron concentration in the conduction band.
This concentration relaxes to a steady state where the electron
injection flux equals the electron escape flux. The escape flux
is composed of two contributions: the electron-sensitizer
recombination flux (S+|(e-cb)TiO2 f S|TiO2) and the electron
the above-described expectation. The dyad studied by Collin et
al. (daap-terpy)Ru(ttp)2+, where daap-terpy is 4′-[(N,N-di-p-
anisylamino) phenyl)]-2,2′:6′,2′′-terpyridine and ttp is 4′-p-tolyl-
2,2′:6′,2′′-terpyridine, appeared thus to be an excellent model
compound.6 The excited ruthenium-terpyridine chromophore
is known to efficiently inject electrons into TiO2 3e,f and not to
undergo dissociation as do the excited ruthenium-pyridine
complexes. 7 The triarylamine subunit chosen as donor moiety
8
forms very stable cation radicals. The phosphonate group in
3e,f
dyad 1 allows it to be firmly anchored at the TiO2 surface.
A second dyad (2) was synthesized in which the distance
between the donor and the sensitizer is larger. Finally, the
phosphonated triarylamine 3 was also prepared in order to study
bimolecular heterotriads formed by surface coadsorption of such
a donor and a ruthenium-terpyridine sensitizer (4, 5). We report
here on the synthesis, electrochemistry, electronic absorption,
and emission spectroscopy, resonance-Raman spectroscopy of
dyads 1 and 2 and of the separated redox system composed of
3 and a sensitizer, as well as on photoelectrochemistry and
photodynamics of the related heterotriads formed by adsorption
of these assemblies on TiO2. All the heterotriads 1|TiO2, 2|TiO2,
and 3+S|TiO2 display redox-type photochromism. To our
knowledge, the only related phenomenon hitherto described is
the UV light-induced photochromism of Prussian blue on flat
TiO2.9
leak to the mediator in the contacting electrolyte (M+
+
(e-cb)TiO2 f M + TiO2). Lowering the quasi-Fermi level by
reducing the former flux requires a lengthening of the lifetime
of the charge-separated state S+|(e-cb)TiO2. If alternatively, an
improved solar cell voltage is to be achieved by raising the redox
potential of the mediator, the same requirement applies. In fact,
the higher redox potential results in a reduced driving force for
the regeneration of the dye (S+ + M f S + M+) which will be
accompanied, in the normal Marcus region, by a reduced rate.
If this rate decreases, the competitive recombination flux
increases, unless, again, the lifetime of the charge-separated state
S+|(e-cb)TiO2 is prolonged. For a longer lifetime, it is obviously
necessary to remove the hole on S+ away from the semiconduc-
tor surface. This can be achieved by replacing the simple
chromophoric sensitizer S by a dyad sensitizer S-D in which
D is an electron donor possessing a redox potential lower than
that of the sensitizer but higher than that of the redox mediator.
Consequently, after electron injection, the hole will be trans-
ferred from S+ toward D (D-S+|(e-cb)TiO2 f D+-S|(e-cb)-
TiO2). The recombination D+-S|(e-cb)TiO2 f D-S|TiO2 is
expected to be slower than the recombination S+|(e-cb)TiO2 f
S|TiO2, because of the exponential decay of the electron-transfer
rate kex(r) with the distance r, according to eq 1
Shortly before we reported preliminary results concerning
11
dyad 1,10 a paper by Argazzi et al. was published in which
the authors described the performance of a nanocrystalline solar
cell incorporating a molecular dyad based on ruthenium
bipyridine as a sensitizer and phenothiazine as a donor. The
aim was the same as ours, i.e., lengthening the lifetime of the
charge-separated state S+|(e-cb)TiO2 in order to lower the quasi-
Fermi level and to gain potential, and this appears to have been
achieved. No photochromism was observed, owing to the
overlap of the visible absorption bands of oxidized phenothiazine
and of the sensitizer.
Experimental Section
Synthesis. Dry solvents for synthesis were the absolute, oVer
molecular sieVe quality from Fluka. Water content was measured before
use by Karl-Fisher automatic titration using a Metrohm 684 KF
coulometer. tert-Butyl methyl ether is abbreviated TBME. Silicagel
was Silicagel 60 from Fluka. Melting points were recorded on a Bu¨chi
530 apparatus, at 1 °C/min heating rate. NMR spectra were obtained
on a Bruker Spectrospin 200 MHz spectrometer. Proton chemical shifts
are given versus the internal standard Me4Si or, in D2O, versus DSS
(sodium 3-trimethylsilyl 1-propanesulfonate-1,1,2,2,3,3-d6). Diazabi-
cyclooctane (DABCO) (δ ) 2.86, s) was added to some of the
triarylamine solutions in CDCl3 to prevent oxidation by air. 31P chemical
shifts are given versus 85% aqueous H3PO4 as an external standard.
FAB-MS spectra were obtained at the Institut de Pharmacognosie et
Phytochimie of the University of Lausanne, by using a Finnigan MAT
kex(r) ) k0ex exp(-â(r - r0))
(1)
in which k0 is the rate at the closest distance r0, and â is the
ex
damping factor of the electron transfer. The whole system
D-S|TiO2 constituted by the dyad adsorbed on the semiconduc-
tor, considered as an acceptor, will be referred to as heterotriad.
For the dyad, we first chose a structure that was already
known to display light-induced charge separation in solution
and that was moreover linear and rigid, affording the best
chances to obtain on the TiO2 surface a heterotriad fulfilling
(6) Collin, J. P.; Guillerez, S.; Sauvage, J. P.; Barigelletti, F.; De Cola,
L.; Flamigni, L.; Balzani, V. Inorg. Chem., 1991, 30, 4230-4238.
(7) Coe, B. J., Friesen, D. A., Thompson, D. W., Meyer, T. J. Inorg.
Chem. 1996, 35, 4575-4584 and refs 39-41 therein.
(8) (a) Dapperheld, S.; Steckhan, E.; Grosse Brinkhaus, K. H.; Esch, T.
Chem. Ber. 1991, 124, 2557-2567. (b) Steckhan, E. Top. Curr. Chem. 1987,
142, 3-69.
(9) DeBerry, D. W.; Viehbeck, A. J. Electrochem. Soc. 1983, 130, 249-
251.
(10) Bonhoˆte, P.; Moser, J. E.; Vlachopoulos, N.; Walder, L.; Zakeer-
uddin, S. M.; Humphry-Baker, R.; Pe´chy. P.; Gra¨tzel, M. J. Chem. Soc.,
Chem. Commun. 1996, 1163-1164.
(11) (a) Argazzi, R.; Bignozzi, C. A.; Heimer, T. A.; Castellano, F. N.;
Meyer, G. J. J. Am. Chem. Soc. 1995, 117, 11815-11816. The full paper
was published recently: (b) Argazzi, R.; Bignozzi, C. A.; Heimer, T. A.;
Castellano, F. N.; Meyer, G. J. J. Phys. Chem. B 1997, 101, 2591-2597.
(3) (a) Desilverstro, J.; Gra¨tzel, M.; Kavan, L.; Moser, J.; Augustynski,
J. J. Am. Chem. Soc. 1985, 107, 2988 (b) O’Reagan, B.; Gra¨tzel, M. Nature
1991, 353, 737-739. (c) Nazeeruddin, M. K.; Liska, P.; Vlachopoulos, N.;
Gra¨tzel, M. HelV. Chim. Acta 1990, 73, 1788-1803. (d) Nazeeruddin, M.
K.; Kay, A.; Rodicio, J.; Humphry-Baker, R.; Mu¨ller, E.; Liska, P.;
Vlachopoulos, N.; Gra¨tzel, M. J. Am. Chem. Soc., 1993, 115, 6382-6390.
(e) Pe´chy, P.; Rotzinger, F. P.; Nazeeruddin, M. K.; Kohle, O.; Zakeeruddin,
S. M.; Humphry-Baker, R.; Gra¨tzel, M. J. Chem. Soc., Chem. Commun.
1995, 65-66. (f) Zakeeruddin, S. M.; Nazeeruddin, M. K.; Pe´chy, P.;
Rotzinger, F. P.; Humphry-Baker, R.; Kalyanasundaram, K.; Gra¨tzel, M.;
Schklover, V.; Haibach, T. Inorg. Chem. 1997, 36, 5937-5946.
(4) Campus, F.; Bonhoˆte, P.; Gra¨tzel, M.; Heinen, S.; Walder, L. Sol.
Energy Mater. 1999, 56, 281-297.
(5) Heimer, T. A.; D’Archangelis, S. T.; Farzad, F.; Stipkala, J. M.;
Meyer, G. J. Inorg. Chem. 1996, 35, 5319-5324.

1326 J. Am. Chem. Soc., Vol. 121, No. 6, 1999
Bonhoˆte et al.
TSQ 700 mass spectrometer, from a nitrobenzyl alcohol matrix, with
positive ionization.
filtered, brought to pH ) 9.7 by addition of Na2CO3, and evaporated
to dryness in a Rotavapor. The resulting solid was extracted with ethanol
and methanol, the solvents were evaporated, leaving 134 mg of white
solid containing about 47% NaBr. 1H NMR (δ in D2O) 8.70 (d, 2H, J
) 5 Hz), 8.38, 8.33, and 8.29 (2 overlapping d, 4H, J ) 11 and 8 Hz),
8.07 (td, 2H, J ) 8 and 1.5 Hz), 7.57 (dd, 2H, J ) 5 and 7 Hz).
Di-p-anisyl-p-bromophenylamine (7). The procedure was derived
12
from that of Bacon et al. Under Ar, in a 500-mL round-bottomed
flask equipped with reflux condenser and magnetic stirrer, sodium (4.03
g, 0.175 mol) freshly cut into small pieces was added to dry methanol
(70 mL). After complete dissolution, the solution was concentrated by
distillation of 30 mL methanol. sym-Collidine (Fluka, puriss., 220 mL)
dried over molecular sieve, copper(I) iodide (Fluka, 13.3 g, 0.070 mol),
and the trisbromophenylamine 6 (Aldrich, 33.74 g, 0.070 mol) were
added, and the mixture was heated to reflux with vigorous stirring for
2 h. It was then filtered while still hot, and the solvents were removed
under 12 mbar in a Rotavapor. The solid residue was extracted by
TBME (250 mL). The extract was washed by 1 M aqueous HCl (200
mL) and twice by water (200 mL). The organic phase was dried over
MgSO4 and the solvent was distilled off to leave 24.57 g of a pale
yellow oil that was charged onto a silica gel column (850 g). Elution
was done with heptane (15 L) and then heptane/TBME 99:1. p-Anisyl-
di-p-bromophenylamine eluted first, closely followed by 7. The
trianisylamine 8 was eluted last. The fractions containing mixed mono-
and dibromo compounds were purified over another 400 g of silica
gel. Finally, the products were isolated by crystallization from heptane.
Yield: 6.77 g (25%) 7 as a yellowish powder and 5.74 g (24%) 8 as
white needles (mp 95 °C). 1H NMR of 7 (δ in CDCl3) 7.23 (dt, 2H, J
) 9 and 2.2 Hz), 7.02 (dt, 4H, J ) 9 and 2.2 Hz), 6.81 (dt, 4H, J )
9 and 2.2 Hz), 6.78 (dt, 2H, J ) 9 and 2.2 Hz); mp 98-99 °C.
4-(N,N-Di-p-anisylamino)benzaldehyde (9). The procedure was
derived from that of Olah et al. 13 Under Ar, 7 (3.84 g, 10 mmol) was
dissolved in dry THF (50 mL). The solution was cooled to -70 °C
and, with stirring, tert-butyllithium (Aldrich, 1.5 M in pentane, 7.3 mL,
11 mmol) was slowly added by syringe, keeping the temperature below
-65 °C. After 45 min, a solution of N-formylmorpholine (Fluka, 1.73
g, 15 mmol) in dry THF (10 mL), dried over molecular sieve, was
added in one portion. The mixture was allowed to warm to room
temperature After 12 h, the reaction was quenched by 1 M aqueous
HCl (15 mL), and extraction was carried out by TBME (100 mL), the
organic phase being washed sequentially by saturated aqueous NaHCO3
and water (2 × 50 mL each). The ethereal extract was dried over Na2-
SO4. After solvent removal, the remaining 3.8 g of yellow oil was
purified by chromatography over silica gel (150 g), eluting by heptane/
TBME, 99:1 to 95:5. The fractions containing the aldehyde were yellow
with blue fluorescence in UV light. Yield: 3.00 g (84%) of a yellow
LRuterpyPO3 (1). LRuCl3 (74.3 mg, 0.10 mmol), 11 (67 mg, 0.10
mmol, containing 47% NaBr), triethylamine (3 mL), and ethanol (2
mL) were mixed in N-methylpyrrolidone (20 mL). The solution was
stirred at 110 °C for 24 h. The solvent was distilled off in a Rotavapor,
and the residue was dissolved in acetonitrile/water, 10:1, and purified
by chromatography over silica gel, eluting with acetonitrile/water, 2:1.
The bright-red fractions having Rf ) 0.1 in TLC on silica gel (CH3-
CN/aqueous 0.1 M KNO3, 2:1) and the absorption spectrum given in
Table 2 were collected. The solvent was distilled off, and the orange
solid which contained some silica from the column was dissolved in 5
mL ethanol, whereby it was separated from the SiO2. Evaporation of
the solvent afforded 30.0 mg 1 (32%): 1H NMR (δ in CD3OD) 9.14
(s, 2H), overlapping with 9.14 (d, 2H, J ) 10 Hz), 8.80 (d, 2H, J ) 8
Hz), 8.63 (d, 2H, J ) 8 Hz), 8.10 (d, 2H, J ) 9 Hz), 8.00-7.90 (m,
4H), 7.47 (t, 4H, J ) 7 Hz), 7.27-7.17 (m, 8H), 7.11 (d, 2H, J ) 9
Hz), 6.97 (d, 2H, J ) 9 Hz), 3.84 (s, 6H); FAB-MS (nitrobenzyl alcohol
matrix) m/z 950.6 (calcd 949.2); UV-vis, see Table 2.
Tris(p-anisyl)amine (8). A procedure similar to the one for 7 was
followed, with sodium (5.52 g, 0.24 mol), methanol (40 mL), CuI (7.60
g, 0.040 mol), the tris-bromophenylamine 6 (19.82 g, 0.040 mmol),
and sym-collidine (250 mL). Reflux was maintained for 14 h. A first
pure fraction of the product was obtained by two crystallizations from
heptane (7.92 g). Distillation of the solvent from the supernatant solution
afforded an oil that was shown by 1H NMR to contain equimolar
amounts of 8 and 7. Chromatography over silica gel (130 g) gave
another 1.43 g 8 as well as 1.20 g (8%) 7: total yield of 8, 9.35 g
1
(70%); H NMR (δ in CDCl3) 6.96 (dt, 6H, J ) 9 and 2.2 Hz), 6.77
(dt, 6H, J ) 9 and 2.2 Hz), 3.77 (s, 9H); mp 95 °C.
4-(N,N-di-p-anisylamino)phenol (12). 8 (6.70 g, 20 mmol) was
dissolved in a mixture of acetic acid (42 mL), 48% aqueous HBr (30
mL), and water (5 mL). The mixture was refluxed, and the reaction
was monitored by TLC on silica gel, with TBME/heptane, 1:1, as eluent.
Rf are the following: 0.66 (8); 0.50 (12); 0.32 (bis-p-hydroxyphenyl-
p-anisylamine); 0.20 (nitrilotriphenol). After 20 min, the mixture was
cooled to room temperature and concentrated in a Rotavapor under 12
mbar to 10 mL; TBME (20 mL) was added and washed twice by water
(20 mL). The organic phase was dried over MgSO4, and the solvent
was distilled off. The remaining oil was chromatographed over silica
gel (270 g), using first heptane and then heptane/TBME, 9:1 to 7:3 as
eluents. 8 was eluted first, followed after several liters of eluent by 12.
Distillation of the solvent in a Rotavapor, finally under 0.1 mbar at 50
°C for 1 h, affords 2.75 g of a waxy solid that was shown by NMR to
still contain 33% molar fraction of TBME. Effective yield: 38%. A
0.10 M solution in dry THF was immediately prepared for use in the
next syntheses. 1H NMR (δ in CDCl3 + DABCO) 6.98-6.65 (m, 12H),
3.77 (s, 6H), 3.22 (s, TBME), 1.19 (s, TBME).
1
oil. H NMR (δ in CDCl3) 9.75 (s, 1H), 7.62 (dt, 2H, J ) 9 and 2.2
Hz), 7.12 (dt, 4H, J ) 9 and 2.2 Hz), 6.88 (dt, 4H, J ) 9 and 2.2 Hz),
6.84 (dt, 2H, J ) 9 and 2.2 Hz), 3.81 (s, 6H).
4′-[4-(N,N-Di-p-anisylamino)phenyl]-2,2′:6′,2′′-terpyridine (L, 10).
This was prepared according to the literature procedure, 6,14 with 9 (3.18
g, 9.55 mmol), 2-acetylpyridine (Fluka, 2.31 g, 19.3 mmol), ammonium
acetate (11.5 g) in acetamide (18 g). The black glassy product obtained
at the end of the reaction was not treated with HBr in acetic acid, as in
the tolylterpyridine synthesis, since this procedure does not lead to
precipitation of the hydrobromide. Chromatography on silica gel (120
g), eluting with TBME/heptane/triethylamine, 85:14.5:0.5, afforded 10
as the second fraction, characterized by Rf ) 0.52 in TLC on silica gel
(TBME/heptane/triethylamine, 80:19:1). Fractions containing 10 mixed
with other products were rechromatographed. Final yield 0.798 g (16%).
1H NMR (δ in CDCl3) identical to previously reported spectrum;6 mp
201-203 °C.
4′-p-tolyl-2,2′:6′,2′′-terpyridine (13) and 4′-(p-bromomethyl-
phenyl)-2,2′:6′,2′′-terpyridine (14). The compounds were prepared
following the previously described procedure.6
4′-(4-(N,N-di-p-anisylamino)phenoxymethylphenyl)-2,2′:6′,2′′-
terpyridine (L′ 15). 12 (0.10 M) in dry THF (5.0 mL, 0.50 mmol)
and potassium tert-butoxide (1.0 M) in dry THF (0.5 mL, 0.50 mmol)
were mixed under Ar. After 15 min, a solution of 14 (0.201 g, 0.50
mmol) in 8 mL of dry THF was added under stirring. The solution
was refluxed for 24 h; meanwhile a precipitate of KBr was formed.
The solvent was removed in a Rotavapor and the solid extracted by
TBME (50 mL) and water (20 mL). The organic phase was washed
with water and dried over Na2SO4. The waxy solid remaining after
solvent distillation was chromatographed over silica gel (11 g), eluting
first by heptane/triethylamine, 99:1, and then by heptane/TBME, 100:0
LRuCl3. This was prepared according to the usual procedure.6,15
After filtration of the product (yield: 80%), addition of aqueous KPF6
to the filtrate caused the precipitation of RuL2(PF6)2 (18) (yield: 16%).
Sodium 2,2′:6′,2′′-terpyridine-4′-phosphonate (terpyPO3Na2, 11).
4′-Diethyl phosphono-2,2′:6′,2′′-terpyridine3f (73.8 mg, 0.20 mmol) was
refluxed for 15 h in 3 M aqueous HBr (20 mL). The solvent was
evaporated to dryness. The residue was dissolved in H2O (20 mL),
(12) Bacon, R. G. R.; Rennison, S. C. J. Chem. Soc. C, 1969, 312-315.
(13) Olah, G. A.; Ohannesian, L.; Arvanaghi, M. J. Org. Chem. 1984,
49, 3856-3857.
(14) Spahni, W.; Calzaferri, G. HelV. Chim. Acta 1984, 67, 450-454.
(15) Sullivan, B. P.; Calvert, J. M.; Meyer, T. J. Inorg. Chem. 1980, 19,
1404-1407.
1
to 50:50. Yield: 154 mg (48%). H NMR (δ in CDCl3) 8.75 (s, 2H)
overlapping in part with 8.72 (dd, 2H, J ) 5 and 1 Hz), 8.68 (dt, 2H,
J ) 9 and 1 Hz), 7.93 (d, 2H, J ) 8 Hz), 7.89 (td, 2H, J ) 8 and 2
Hz), 7.57 (d, 2H, J ) 8 Hz), 7.36 (ddd, 2H, J ) 8, 5 and 1 Hz), 6.99

Oxide Films Sensitized by Molecular Dyads
J. Am. Chem. Soc., Vol. 121, No. 6, 1999 1327
(dt, 6H, J ) 9 and 2.2 Hz), 6.87 (dt, 2H, J ) 9 and 2.2 Hz), 6.79 (dt,
2H, J ) 9 and 2.2 Hz), 5.11 (s, 2H), 3.78 (s, 6H); mp 161-163 °C.
L′RuCl3. The compound was prepared according to the usual
procedure.6,15 Yield: 86%.
0.90 g of crude product that was purified by chromatography over silica
gel (30 g). Eluting with TBME/heptane, 33:67 to 100:0, and finally by
ethyl acetate gave 0.578 g of an oil that was shown by H NMR to
1
contain 10-20% 17. This starting material was evaporated under 0.05
mbar at 100 °C for 1 h, leaving 0.546 g of pure ethyl ester of 3 (73%).
This product (1.01 mmol) was diluted under Ar with dry CH2Cl2 (25
mL), trimethylsilyl bromide (Fluka, 1.84 g, 12 mmol) was added, and
the mixture was stirred for 15 h at room temperature. The solvent and
excess TMSBr were evaporated off under Ar in a Rotavapor. Fresh
CH2Cl2 (20 mL) was added, and the solution was washed with water
(2 × 20 mL). The organic phase was mixed with an equal volume of
water, and the biphasic system was slowly neutralized by 0.1 M aqueous
NaOH with gentle stirring to prevent emulsification. Water was removed
from the aqueous phase in a Rotavapor, and the remaining solid was
extracted with THF (20 mL). Precipitation of 3 was accomplished by
addition of TBME. Yield: 0.376 g white solid (71% from the ethyl
ester). 1H NMR (δ in D2O) diethylester of 3: 6.98-6.92 (m, 6H), 6.81-
6.73 (m, 6H), 4.11 (quint. d, 4H J ) 7 and 1.7 Hz), 3.97 (t, 2H, J )
7 Hz), 3.78 (s, 6H), 2.15-1.85 (m, 4H), 1.32 (t, 6H, J ) 7 Hz); 3:
6.92-6.75 (m, 12H), 3.96 (t, 2H, J ) 7 Hz), 3.75 (s, 6H), 2.15-2.00
(m, 2H), 1.68-1.51 (m, 2H). Anal. Calcd. for 3‚H2O + 3.4% NaHCO3
found (calcd): C 53.21 (53.38), H 5.06 (5.15), N 2.67 (2.30), O 23.45
(23.62), Na 9.70 (9.71).
L′Ru(terpyPO3) (2). The compound was prepared by a procedure
similar to the one for 1. Elution was done with acetonitrile/water, 3:1.
To remove the silica gel dissolved from the column, the product
obtained after elimination of the solvent was redissolved in acetonitrile/
ethanol, 1:1, and the insoluble SiO2 was filtered off. 2 was isolated by
1
precipitation from the filtrate by addition of TBME. Yield: 25%. H
NMR (δ in CD3OD) 9.17 (s, 2H), 9.10 (d, 2H, J ) 11 Hz), 8.78 (d,
2H, J ) 8 Hz), 8.65 (d, 2H, J ) 8 Hz), 8.28 (d, 2H, J ) 8 Hz), 7.97
(t, 4H, J ) 8 Hz), 7.83 (d, 2H, J ) 8 Hz), 7.47 (d, 2H, J ) 6 Hz), 7.43
(d, 2H, J ) 6 Hz), 7.23 (t, 4H, J ) 7 Hz), 6.97 (s, 4H), 6.95 (d, 4H,
J ) 9 Hz), 6.84 (d, 4H, J ) 9 Hz), 5.27 (s, 2H), 3.77 (s, 6H); FAB-
MS (nitrobenzyl alcohol matrix) m/z 1056.7 (calcd 1055.2); UV-vis:
see Table 2.
(Et2O3Pterpy)RuCl3. The compound was prepared according to the
usual procedure15 from RuCl3‚3H2O and 4′-diethylphosphono-2,2′:6′,2′′-
terpyridine (Et2O3Pterpy). Yield: 82%.
(Me2bipy)Ru(NCS)(terpyPO3H) (4). The synthesis of this complex
was reported elsewhere.3f
Me3terpyRuterpyPO3H(ClO4) (5). Et2O3terpyRuCl3 (0.10 g, 0.18
mmol) and 4,4′,4′′-trimethyl-2,2′:6′,2′′-terpyridine (Me3terpy, 60 mg,
0.22 mmol) were reacted for 4 h in refluxing DMF (30 mL). The
solution was cooled to room temperature, filtered, and concentrated in
a Rotavapor. Diethyl ether was added until precipitation of the complex.
1H NMR showed partial de-ethylation of the phosphonate. The
hydrolysis was completed by refluxing for 10 h in 4 M aqueous HCl.
The solvent was removed in a Rotavapor. The dry solid was redissolved
in water, and the pH was raised to 6 by addition of 0.01 M aqueous
NaOH and then lowered to 2.5 by addition of HClO4, whereby 5
precipitated. Yield: 60 mg (40%). 1H NMR (δ in DMSO-d6) 9.17 (d,
2H), 8.94 (s, 2H), 8.83 (d, 2H), 8.65 (d, 2H), 7.97 (d, 2H), 7.40 (d,
2H), 7.27 (t, 2H), 7.16 (d, 2H), 7.10 (d, 2H), 2.90 (s, 3H), 2.40 (s,
6H); 31P NMR (δ in DMSO-d6) 3.10; UV-vis: see Table 2. An X-ray
crystal structure of the compound was obtained and will be published
elsewhere.
Ru(terpyPO3H2)2Cl2 (16). 4′-Diethylphosphono-2,2′:6′,2′′-terpyri-
dine (0.12 g, 0.32 mmol) and RuCl3‚3H2O (0.040 g, 0.15 mmol) were
reacted for 6 h in refluxing DMF (25 mL). The solution was filtered at
room temperature and concentrated in a Rotavapor, and diethyl ether
was added until precipitation of the complex. Hydrolysis was carried
out as for 5. The precipitate formed after cooling at room temperature
of the HCl solution was isolated by filtration. Yield: 60 mg (49%). 1H
NMR (δ in 0.02 M NaOD/D2O) 8.93 (d, 2H), 8.53 (d, 2H), 7.87 (t,
2H), 7.40 (d, 2H), 7.09 (t, 2H); 31P NMR (δ in 0.02 M NaOD/D2O)
7.23; UV-vis: see Table 2. Anal. Calcd. for 16‚5H2O found (calcd):
C 40.6 (42.4), H 9.5 (9.5), N 3.8 (3.9).
Sodium 3-(4-phenyl-phenoxy)propyl-1-phosphonate (19). A pro-
cedure similar to the one for 3 was used. From 4-hydroxybiphenyl
(Fluka, 0.510 g, 3.0 mmol), potassium tert-pentylate (1.7 M) in toluene
(Fluka, 1.76 mL, 3.0 mmol) and 17 (0.777 g, 3.0 mmol), we obtained
after chromatography 0.686 g of diethylester of 19 (Rf ) 0.19 on silica
gel, with TBME). For hydrolysis, this ester was dissolved in dichlo-
romethane (20 mL) and trimethylsilyl bromide (1.90 M in CH2Cl2, 12.6
mL, 24.0 mmol) was added. After 15 h stirring at room temperature,
the solvent and excess TMSBr were distilled off. Dichloromethane and
water (20 mL each) were added, and the mixture was stirred. After 2
min, a white solid formed. Some more product was precipitated by
addition of heptane (10 mL). The product was filtered, washed with
water, and dried (yield: 493 mg). To a suspension of this solid in water
(20 mL), 1.00 M aqueous NaOH (3.37 mL) was added to dissolve the
solid. The solution was filtered, and the water was distilled off from
the filtrate, leaving a white solid that was dissolved in methanol/water
(9:1) and precipitated by slow addition of ethanol under stirring.
Yield: 428 mg (42%). 1H NMR (δ in D2O) 7.47-7.42 (m, 4H), 7.33-
7.17 (m, 3H), 6.91 (dt, 2H, J ) 9 and 2 Hz), 3.92 (t, 2H, J ) 7 Hz),
1.84-1.68 (m, 2H), 1.40-1.23 (m, 2H). Anal. Calcd. for 19‚1.2H2O
found (calcd): C 50.8 (50.4), H 5.1 (4.9), O 23.26 (23.25).
Preparation and Characterization of the Nanocrystalline Layers.
The procedure was carried out as previously described.16,17 The thickness
of the layer was 4.5 to 5 µm, affording a roughness factor F ) 450-
500.17 To remove water and organic material adsorbed during storage,
the electrodes were heated to 350 °C for 30 min in an air stream.
Derivatization was done by immersing the still hot (60-80 °C) electrode
in a sub-millimolar solution of the relevant compound in absolute
ethanol. After 12 h at room temperature, the electrodes were rinsed
with absolute ethanol, dried briefly (1 min) at 100 °C, and used directly
or stored in cyclohexane. At monolayer coverage, the very large surface
area offered by the porous solid allowed adsorption of a considerable
Diethyl 3-bromopropylphosphonate (17). In a 50-mL two-necked
round-bottomed flask equipped with a reflux condenser on one neck
and a distillation head on the other, a mixture of 1,3-dibromopropane
(Fluka, 20.2 g, 0.10 mol) and triethyl phosphite (Fluka, 4.15 g, 0.025
mol) was heated to 130 °C with stirring. Bromoethane distilled during
30 min. Heating was continued for another 90 min. The excess
dibromopropane was removed under 12 mbar in a Rotavapor. 1H NMR
shows no impurity in the crude product. It was however distilled over
a 5-cm Vigreux column, under 0.1 mbar. 17 was collected at 57.5 °C.
amount of the dyes, thus leading to high optical densities (A430
g
nm
0.5). It was shown previously that a given ratio of the two phosphonates
3 and 4 (or 5) in solution adsorbed in the same ratio on the TiO2
surface.17
Instrumentation. Cyclic voltammetry, electrolyses, and photoelec-
trochromic experiments were carried out with a PC-controlled Eco-
chemie model Autolab P20 potentiostat. In all of the experiments, the
reference electrode was Ag/Ag+-saturated NaCl (+0.197 V vs NHE),
and the counter electrode was glassy carbon. Both were separated by
1
Yield, 3.80 g (59%). H NMR (δ in CDCl3) 4.10 (t, 4H, J ) 7 Hz),
3.48 (t, 2H, J ) 6 Hz), 2.23-2.06 (m, 2H), 1.98-1.81 (m, 2H), 1.34
(t, 6H, J ) 7 Hz); density at 20 °C: 1.348 g/cm3.
Sodium 3-(4-(N,N-di-p-anisylamino)phenoxy)-propyl-1-phospho-
nate (3). Under Ar, 12 (0.10 M) in dry THF (15.0 mL, 1.50 mmol)
and potassium tert-butoxide (0.95 M) in dry THF (1.64 mL, 1.50 mmol)
were mixed. After 15 min, a solution of 17 (0.390 g, 1.50 mmol) was
added under stirring, and the solution was refluxed. After 5 min, a
precipitate of KBr appeared. Monitoring by TLC on silica gel (eluent:
TBME) showed that after 4 h, the reaction did not progress any more
(12: Rf ) 0.59; 3: Rf ) 0.22) The mixture was diluted with TBME
(30 mL) and washed twice with water (30 mL). The organic phase
was dried over Na2SO4. Removal of the solvent in a Rotavapor afforded
(16) (a) O’Regan, B.; Moser, J.; Anderson, M.; Gra¨tzel, M. J. Phys.
Chem. 1990, 94, 8720. (b) Barbe´, C.; Arendse F.; Comte, P.; Jirousek, M.;
Lenzmann, F.; Shklover, V.; Gra¨tzel, M. J. Am. Ceram. Soc. 1997, 80,
3157-3171.
(17) Bonhoˆte, P.; Gogniat, E.; Tingry, S.; Barbe´, C.; Vlachopoulos, N.;
Lenzmann, F.; Comte, P.; Gra¨tzel, M. J. Phys. Chem. B 1998, 102, 1498-
1507.

1328 J. Am. Chem. Soc., Vol. 121, No. 6, 1999
Bonhoˆte et al.
Scheme 1a
a salt bridge from the solution in contact with the working electrode.
For photoelectrochromic measurements, the working electrodes were
4.8 µm TiO2 on 1 cm2 SnO2, derivatizated in 0.5 mM solutions in EtOH
for 12-15 h. The spectrophotometric measurements were effected with
a HP8453 diode array spectrophotometer from Hewlett-Packard, in a
1 × 1 cm spectroelectrochemical cell. Electrolyses were carried out
on carbon felt. Illumination for the photochromic measurements was
done with a halogen source, through two optical fibers, using a Highlight
3001 device from Olympus, equipped with a UV filter (cutoff 420 nm).
Light intensity on the electrodes was 1 ( 0.2 sun (AM 1.5).
The resonance Raman (RR) spectra were obtained using a coherent
INNOVA 200K Kr+ laser source and a Spex 1877 Triplemate with
N2-cooled CCD-1024 detection. All of the samples were prepared as 1
mM aqueous solutions stored in 1-mm i.d. glass capillaries. The sample
were irradiated in a 90° geometry.
Emission spectra were recorded in a Spex Fluorolog 112 using a
90° optical geometry. The emitted light was detected with a Hamamatsu
R2658 photomultiplier operated in single photon counting mode. The
emission spectra were photometrically corrected using a calibrated
200-W tungsten lamp as reference source. The samples were purged
with N2 and the absorption spectra measured prior to each emission
measurement. A dilute solution of quinine sulfate was used as a quantum
yield standard.18
The kinetics of electron-transfer processes were examined by laser
flash photolysis using dye-sensitized porous transparent TiO2 films
deposited on glass substrates. The surface of the sample was usually
covered by a film of solvent, and protected by a thin microscope
coverglass. Pulsed laser excitation was applied using a GWU BBO-
C355 broad-band optical parametric oscillator pumped by a Continuum
Powerlite 7030 frequency-tripled Q-switched Nd:YAG laser (λ ) 355
nm, repetition rate 30 Hz). The output of the OPO (pulse width at half-
height ca. 5 ns) was typically tuned at λ ) 430 or 480 nm and attenuated
to 1.5 mJ/pulse. The beam was expanded by a plano-concave lens to
irradiate a large cross-section (approximately 1 cm2) of the sample,
whose surface was kept at a 45° angle to the excitation and to the
analyzing light. The probe beam, produced by a CW 450-W Xe-arc
lamp, was passed through various optics and filters, the sample, and a
monochromator prior to being detected by a Hamamatsu R928
photomultiplier tube, only three dynodes of which were used. A
Tektronix DSA 602A 1-GHz digital signal analyzer was employed to
record the time course of the optical absorbance changes induced by
pulsed laser excitation of the films. Satisfactory S/N ratios exceeding
20 were typically obtained by averaging over 10-200 laser shots.
Kinetic data were finally analyzed on an Apple Power Macintosh
personal computer using WaveMetrics Igor Pro software.
The photovoltaic performance of solar cells incorporating either 1,
2, or 5 as photosensitizers was characterized using two complementary
techniques. The first method measures the current-voltage characteristic
of a sandwich-type cell with platinized SnO2 counter electrode under
different light intensities up to AM 1.5 (or 1 Sun). The simulated solar
source was a 450-W Xe lamp whose spectral output matched the AM
1.5 solar spectrum in the region of 350 to 750 nm (mismatch < 2%).
Different light intensities were adjusted with neutral wire mesh
attenuators. The applied potential and measured cell current was
performed using a Keithley model 2400 digital source meter.16b The
fully automated experiment was run using WaveMetrics Igor Pro. The
spectral response of the cell was determined by measuring the incident
photon-to-current conversion efficiency (IPCE) as a function of the
wavelength16b
a 2.5 equiv of MeONa, CuI/collidine-MeOH 2 h reflux (25%); (b)
tert-BuLi/THF, -70 °C 45 min; N-formylmorpholine/THF, 12 h at rt
(84%); (c) 2-acetylpyridine, NH4CH3COO, CH3CONH2 (16%); (d)
RuCl3‚3H2O/EtOH (80%). (e) terpyPO3Na2/NMP-NEt3-EtOH 24 h 110
°C (47%).
intermediate dianisylaminobenzaldehyde 9 from the com-
mercially available tris-bromophenylamine 6 rather than from
N,N′-di-p-anisylphenylamine. The copper (I)-catalyzed substitu-
tion of the halogen atoms by methoxy groups12 in 6 can be well
controlled by the amount of added alkoxide and reaction time.
Chromatography allows the separation of the mono-, di-, and
trialkoxilated products. The lithiated derivative of 7 was
efficiently formylated by N-formylmorpholine13 to give 9. The
building of the terpyridine moiety on the formyl group was done
as previously described6 to afford the ligand L (10) which
reacted smoothly with RuCl3 to form the complex LRuCl3,15
together with a small amount of the bis-complex LRu(II)L (18).
The last step of the synthesis involves the displacement of three
chloride ions of LRuCl3 by the ligand 10, using triethylamine
as a reducing agent.
Dyad 2 was prepared according to Scheme 2. Partial acidic
hydrolysis of the trianisylamine 8 followed by chromatographic
separation afforded the monophenol 12, a key intermediate for
all the syntheses of functionalized tris-alcoxyphenylamines.
Bromotolylterpyridine 14 was prepared according to the previ-
ously described procedure6 and reacted with the phenolate of
12, to give the ligand L′ (15). Formation of the final ruthenium
complex 2 was achieved as for dyad 1.
The phosphonated triarylamine 3 was obtained by nucleo-
philic substitution between the phenolate of 12 and diethyl
3-bromopropylphosphonate 17, followed by de-ethylation with
trimethylsilyl bromide.19 17 was itself made by Arbusow
IPCE(λ) ) 1240(Isc/λΦ)
(2)
where λ is the wavelength (nm), Isc is the current at short circuit
(mA/cm2) and Φ is the incident radiative flux (W/m2).
reaction of 1,3-dibromopropane with triethyl phosphite.20
The 1H NMR chemical shifts recorded for L, L′, terpyPO3
2-
,
Results and Discussion
and for the complexes 1 and 2 were attributed considering the
previously reported assignments (Figure 1),2f,21 as well as the
Synthesis. Synthesis of dyad 1 was carried out as depicted
in Scheme 1. It followed in part the procedure established by
Collin et al.6 except that we found it easier to prepare the
(19) Katz, H. E.; Bent, S. F.; Wilson, W. L.; Schilling, M. L.; Ungashe,
S. B. J. Am. Chem. Soc. 1991, 116, 6631-6635.
(20) Arbusow, B. A. Pure Appl. Chem. 1964, 9, 307-335. Katz, H. E.;
Wilson, W. L.; Scheller, G. J. Am. Chem. Soc. 1994, 116, 6636-6640.
(18) Parker, C. A.; Rees, W. T. Analyst (London) 1960, 85, 587.

Oxide Films Sensitized by Molecular Dyads
J. Am. Chem. Soc., Vol. 121, No. 6, 1999 1329
Scheme 2a
Figure 1. 1H NMR chemical shifts (δ in ppm versus Me4Si)
assignments for L, terpyPO32-, and L′, in the isolated molecules (top
2-
value) and in the complexes 1 and 2 (bottom value; for terpyPO3
left, δ in 1 and right, δ in 2).
:
observed multiplicities and coupling constants of the signals.
In the free ligands L, the strong electronic coupling between
a 3 equiv of MeONa, CuI/collidine-MeOH, 14 h reflux (70%); (b)
concentrated HBr/AcOH reflux 20 min (33%); (c) CH3COONH4/
CH3CONH2 reflux 4 h, Fe2+ complexation, H2O2/NaOH (23%); (d)
NBS, (Ph-COO)2/benzene, 18 h at reflux (85%); (e) tert-BuOK/THF
(48%); (f) RuCl3‚3H2O/EtOH (86%); (g) terpyPO3Na2/NMP-NEt3-EtOH
24 h 110 °C (25%).
dyad 2, the latter potential could not be measured, since the
irreversible second oxidation of the triarylamine moiety (E1/2
) 1.39 V for 8) occurs in the same range. The value should be
similar to those measured for complexes 1, 5, and 18. In 18,
the return wave of the second terpy reduction forms a spike,
denoting that the bis-reduced complex precipitated on the
electrode. For 1 and 2, only the first one-electron reduction was
clearly observed. The redox potentials of the triarylamine part
of 2 and 3 are very similar to that measured for the free
trianisylamine (8) (E1/2 ) 0.79 V). Replacement of one alkoxy
group of 8 by the cationic Ru(II)-terpyridine group raises the
oxidation potential of the donor part in dyads 1 and 18.
Symmetrically, the strong coupling with the electron-donating
dianisylaminophenyl group lowers the reduction potential of the
terpyridine ligand of these complexes by comparison with 2.
From the values of Table 1, the driving forces for the charge-
the electron donor (triarylamine) and the electron acceptor
(terpyridine) shows itself by a downfield shift of the protons in
the former moiety and by an upfield shift in the latter, compared
to L′, where this coupling is much weaker. In 1, the inductive
effect of the central ion extends into the triarylamine, which is
not the case in 2, due again to the difference in the donor-
chromophore couplings.
Electrochemistry. The electrochemical properties of the
investigated systems are listed in Table 1. In accordance with
previous work6,22 the redox potentials observed close to -1 V
were attributed to one-electron reduction of the terpyridine
ligands and that near 1.5 V to metal-centered oxidation. For
(21) Tiecco, M.; Testaferri, L.; Tingoli, D.; Chianelli, D.; Montanucci,
S. Synthesis 1984, 736-738.
(22) Juris, A.; Balzani, V.; Barigelletti, F.; Campagna, S.; Belser, P.;
Von Zelewsky, A. Coord. Chem. ReV. 1988, 84, 85-277.

1330 J. Am. Chem. Soc., Vol. 121, No. 6, 1999
Bonhoˆte et al.
Table 1. Electrochemical Redox Potentials, E1/2 in V vs NHE, of
Compounds 1-18a
0/-
-/2-
compd
Ru3+/2+
(terpy)2
(terpy)2
D+/0
1b
18c
2d
1.50
1.57
(1.5)
-1.09
-1.04
-0.94
0.95
0.98 (2e-)
0.85
-1.23
3b,e
4f
0.80
1.11
1.36
-1.12
-1.05
5f
-1.31
a Reference electrode: Ag/AgCl-saturated NaCl (+0.197 V vs
NHE); supporting electrolyte: 0.1 M NBu4+ CF3SO3-; all one-electron
processes, except where specified. b In propylene carbonate. c In ac-
etonitrile. d In acetonitrile/DMSO, 9:1. e Measured in the surface-
confined state, as a monolayer on a TiO2 nanocrystalline electrode.17 f
in DMSO.
Table 2. Ground-State Absorption Maxima: λmax/nm (ꢀ/103 M-1
cm-1
compd
LC
LC
MLCT
D+ c
1a
1+ b
18a
2a
274 (45.6)
311 (46.0)
500 (23.0)
486 (19)
516 (51.0)
486 (21.6)
486 (23)
Figure 2. Absorption and emission spectra of 1 (a) and 2 (b) 20 µM
in ethanol.
750 (13)
277 (66.0)
310 (87.1)
2+ a
3+ a
4d
720 (22)
720 (24)
Resonance Raman (RR) spectroscopy was used to elucidate
the localization of the electron density upon excitation. Charge
transfer from the metal center to a ligand (MLCT) gives rise to
resonance enhancements of the symmetric stretching A1 modes
of that ligand.23 For a broad visible absorption spectrum as seen
for the heterolyptic complexes 1 and 2 several overlapping
MLCT transitions are possible, dictated by the energy of the
π* level of the implicated ligand. The relative contribution of
each transition can be ascertained by examining the resonant
enhancement of the vibrational modes for the different ligands
as a function of the probing Raman wavelength. Other groups
have demonstrated the utility of this method to map out the
MLCT chromophores.24
By selecting a wavelength at 530.9 nm, just to the red of the
maximum at 504 nm in aqueous 1 (see Figure 2), RR shows a
spectrum whose fingerprint in terms of peak position is
analogous to that of the ligand L (Figure 3). At shorter
wavelengths, higher energy electronic states are probed, and
the RR spectrum becomes more complex. The set of peaks
ascribed to L diminish in intensity, while a new set of peaks
appear. At 415.4 nm, the RR spectrum resembles the terpy-
PO3 ligand fingerprint. It is thus clear that, on the RR time scale
(<10-13 s), the lowest energy MLCT corresponds to the Ru f
L transition. When 1 is adsorbed onto the nanocrystalline TiO2
film (1|TiO2), the RR clearly shows the same MLCT, even
further to the red (568.2 nm) due to less experimental inter-
ference from the natural emission of the dye. The terpyPO3
ligand vibrational peaks are also weakly observed at low
energies underlying the ligand L modes but become dominant
at higher energies.
280 (35.0)
276 (37.2)
272 (34.2)
320 (26.1)
312 (51.8)
310 (39.2)
498 (8.5)
482 (15.3)
480 (12.5)
5d
16e
a In acetonitrile. b In acetonitrile/propylene carbonate, 96:4. c Ob-
tained by electrochemical oxidation in a spectrophotometric cell (see
Experimental Section). d In ethanol/DMSO, 9:1. e In water (pH ) 9).
transfer D-S+|(e-cb)TiO2 f D+-S|(e-cb)TiO2 can be estimated
to be 0.55 V for 1, 0.65 V for 2, 0.31 V for 3 + 4, and 0.56 V
for 3 + 5.
Absorption, Emission and Resonance Raman Spectros-
copy. The UV-vis absorption spectra of the studied compounds
are summarized in Table 2. The UV bands are attributed to the
ligand-centered (LC) transitions, and the visible bands cor-
respond to the well-known metal-to-ligand charge transfer
(MLCT). The oxidized triarylamine moieties are characterized
by an intense absorption in the red, due to the π f π transition
in the radical cation.
The substantial bathochromic-hyperchromic effect induced
by the triarylamine conjugated to the terpyridine ligand is
obvious in the series including complex 5, dyad 1, and complex
18. The effect is attributed rather to an energy-raising of the
metal-centered HOMO of the complex due to the electron-
donating character of the triarylamine, than to a lowering of
the LUMO (π*) of the ligand due to the extension of the
conjugation. In fact, if the conjugated triarylamine would only
lower the π* of the ligand, the absorption of 18 would not be
red-shifted compared to 1. When the electron-donation is
suppressed by electrochemical oxidation of the triarylamine
moiety (1+), weakening and blue-shifting of the MLCT band
is observed. In dyad 2, the absence of π-conjugation between
the triarylamine and the Ru-terpyridine chromophore results
in a MLCT absorption at shorter wavelength than for 1 and to
the insensitivity of this band to the oxidation the triarylamine
(2+).
For complex 2, the RR vibrational spectrum strikingly
resembles the spectrum of 16 obtained at 530.9 nm excitation.
In that case, the lowest energy MLCT corresponds to the Ru
f terpyPO3 transition.
Charge-Transfer Dynamics. The transient difference signals
of compounds 1, 2, 5, and 3+5 adsorbed on nanocrystalline
TiO2 films on glass were recorded in different media.
In air, the transient difference spectra obtained upon excitation
of heterodyad 5|TiO2 and heterotriad 1|TiO2, respectively, were
identical. In both cases, excitation at λ ) 430 nm resulted in a
Complexes 1 and 2 show the typical emission spectra of
ruthenium complexes with respective emission maxima at 686
( 5 nm and 652 ( 5 nm in ethanol (Figure 2). This clearly
reflects the energetics displayed by the corresponding MLCT
absorption peaks at 505 and 486 nm. Consistent with other
complexes3f having the terpyridyl ligand, the emission quantum
(23) Strommen, D. P.; Mallick, P. K.; Danzer, G. D.; Lumpkin, R. S.;
Kincaid, J. R. J. Phys. Chem. 1990, 94, 1357-1366.
(24) Yabe, T.; Orman, L. K.; Anderson, D. R.; Yu, S.-C.; Xu, X.;
Hopkins, J. B. J. Phys. Chem. 1990, 94, 7128-7132.
yield is low at 8 × 10-6
.

Oxide Films Sensitized by Molecular Dyads
J. Am. Chem. Soc., Vol. 121, No. 6, 1999 1331
1, which is thermodynamically very favorable in propylene
carbonate (Table 1), does not take place in air. The medium
reorganization is not expected to play an important role in
defining the potential of the S+/S couple in the complex. The
absence of solvent then apparently affects the oxidation potential
of the amine by increasing its value by more than 0.5 V.
In propylene carbonate (dielectric constant ꢀ ) 64), the
transient signal of heterodyad 5|TiO2 was the same as that in
air. For heterotriad 1|TiO2, on the contrary, the ground-state
absorption of the dye recovered much faster: more than half
of the ground-state absorption was regenerated in less than 20
ns (Figure 4a). Simultaneously, a pronounced absorption signal
appeared within the laser pulse with a maximum at 750 nm,
corresponding to the spectral feature of D+. Quantitative
comparison of the initial ground-state bleaching (ꢀ(S, 500 nm)
) 2.3 × 104 M-1 cm-1) with the D+ build-up signal (ꢀ(D+,
620 nm) ) 3.5 × 103 M-1 cm-1) shows that 60% of the excited
heterotriads underwent intramolecular electron-transfer accord-
ing to parts a and b of eq 5.
D-S*|TiO2 f D-S+|(e-) TiO2
D-S+|(e-) TiO2 f D+-S|(e-) TiO2
D-S+|(e-) TiO2 f D-S|TiO2
(5a)
(5b)
(5c)
Given the long lifetime of the S+|(e-)TiO2 state in 5|TiO2 and
the fast appearance of the absorption of D+ within the excitation
pulse (eq 5b), the quantum yield for the formation of the radical
cation is not likely to be controlled by kinetic competition with
charge recombination (eq 4). Oxidation of the donor moiety
would therefore be expected to take place quantitatively. For
that reason, we cannot formulate a convincing hypothesis for
the origin of the low yield of charge separation within the frame
of this mechanism. Alternatively to process 5a and 5b, charge
separation can take place through reductive quenching of S*
by D before charge injection into the solid occurs (eq 6).
Figure 3. Resonance Raman spectra of 1, using (a) 530.9 nm or (b)
415.4 nm laser excitation, of 2 (c) using 482.5 nm excitation, and of
16 (d) using 530.9 nm excitation.
bleaching of the ground-state absorption of the dye, monitored
at 500 nm, concomitantly with the laser pulse. Recovery
occurred in the microseconds time scale (τ1/2 ) 3 µs). The
excited-state lifetime of the Ru-terpyridine chromophore being
less than 10 ns, the slow recovery of the ground-state absorption
is thus interpreted as resulting of a charge injection from the
dye MLCT excited state (S*) into the conduction band of the
semiconductor (eq 3), followed by the recapture of injected
electrons by oxidized dye (S+) at the surface (eq 4).
D-S*|TiO2 f D+-S-|TiO2 f D+-S|(e-) TiO2 (6)
This mechanism cannot be experimentally distinguished from
the previous one in this case since they both take place within
the laser pulse and lead to the same final products. However,
the moderate yield of charge separation could be explicable on
the basis of this second mechanism. Resonance Raman spec-
troscopy showed that the exicted electron is located on terpy
of the ligand L, close to the donor. Therefore, the intramolecular
charge recombination in the D+-S- system could possibly be
fast enough to compete efficiently with the electron injection
into the semiconductor.
For heterotriad 2|TiO2 in propylene carbonate, the transient
bleaching signal of the ground state of S measured at 500 nm
displayed 80% recovery of initial absorption within 10 ns,
followed by a second kinetic step that led to complete
regeneration 100 ns after the laser pulse. Simultaneously, the
absorption of D+ appeared and was monitored at 620 nm. Both
signals displayed identical kinetic features with the components
k1 ) 108 s-1 (80%) and k2 ) 3 × 107 s-1 (20%) (Figure 4a).
Using extinction coefficients ꢀ(S, 500 nm) ) 1.5 × 104 M-1
cm-1 and ꢀ(D+, 620 nm) ) 6.6 × 103 M-1 cm-1, a unit quantum
yield for processes 5a and b is derived. This high efficiency is
attributed to the fact that, unlike in 1, the lowest π* orbital of
S*|TiO2 f S+|(e-) TiO2
S+|(e-) TiO2 f S|TiO2
(3)
(4)
The back electron transfer (eq 4) was found for similar redox
sensitizer systems to occur in the microsecond time domain.
11b,25 This slow recombination could be rationalized by consid-
ering the large driving force (∆G ) -1.5 eV) for the process,
which combined with very small reorganization energies (λ <
0.1 eV) characterizing the Ru(II)/Ru(III) exchange in poly-
pyridine complexes,30 places it deep in the Marcus inverted
region. Oxidation of the triarylamine moiety D by S+ in dyad
(25) O’Regan, B.; Moser, J.; Anderson, M.; Gra¨tzel, M. J. Phys. Chem.
1990, 94, 8720. Moser, J. E.; Gra¨tzel, M. Chem. Phys. 1993, 176, 493-
500.
(26) Tachibana, Y.; Moser, J. E.; Gra¨tzel, M.; Klug, D. R.; Durrant, J.
R. J. Phys. Chem. 1996, 100, 20056-20062.
(27) Rothenberger, G.; Moser, J.; Gra¨tzel, M.; Serpone, N.; Sharma, D.
K. J. Am. Chem. Soc. 1985, 107, 8054-8059.
(29) Pleskov, Y. V. Solar Energy ConVersion; Springer: Berlin, 1990;
pp 8-11.
(30) Sutin, N. Prog. Inorg. Chem. 1983, 30, 441-498.
(28) Bonhoˆte, P.; Dias, A. P.; Papageorgiou, N.; Kalyanasundaram, K.;
Gra¨tzel, M.; Armand, M. Inorg. Chem. 1996, 35, 1168-1178.

1332 J. Am. Chem. Soc., Vol. 121, No. 6, 1999
Bonhoˆte et al.
latter step in heterotriad 2|TiO2 compared with that in 1|TiO2
is attributed to a larger D-S distance. The hole transfer remains,
however, much faster than the charge recombination (eq 5c).
The biphasic kinetics observed can be due to the multiplicity
of the conformations of the molecule that result in different D-S
distances or to the existence of intermolecular D f S+ electron
transfers inside the monolayer.
For heterotriad (3+5)|TiO2 (3 and 5 coadsorbed in a 1:1
ratio), the transient spectra recorded upon pulse excitation at
480 nm was very similar as with 2|TiO2. The fast recovery of
the ground-state absorption of the dye within the laser pulse
was mirrored by the immediate appearance of the transient
absorption of 3+ in the red part of the spectrum. They could
not be time-resolved in our experimental setup. This observation
is interpreted in terms of a similar electron-transfer sequence
as in 2|TiO2 (eqs 7 and 8).
(D + S*)|TiO2 f (D + S+)|(e-) TiO2
(D + S+)|(e-) TiO2 f (D+ + S)|(e-) TiO2
(7)
(8)
Comparison of the transient signals (ꢀ(5, 500 nm) ) 104 M-1
cm-1 and ꢀ(3+, 620 nm) ) 7 × 103 M-1 cm-1) indicates that
the quantum yield of these processes is practically unity.
The charge recombination processes 9 between the conduc-
tion-band electrons and the oxidized donor D+ were followed
in propylene carbonate for heterotriads 1|TiO2, 2|TiO2, and
(3+5)|TiO2 by the evolution of the transient absorption at 620
nm (Figure 4b).
D+|(e-) TiO2 f D|TiO2
(9)
The lifetime of the charge separated state (3++5)|(e-)TiO2 is
very similar to that obtained with 5+|(e-)TiO2 alone. In an
inverted regime, the decrease of the driving force for charge
recombination by 0.56 eV and the existence of a large solvent
reorganization energy associated with the D/D+ couple should
indeed compensate in part for the larger distance over which
the charge recombination takes place. The recombination
kinetics of the three heterotriads are very dissimilar, with half-
lifetime (τ1/2) for the D+|(e-) TiO2 state of 3 µs in (3+5)|TiO2,
30 µs in 1|TiO2, and 300 µs in 2|TiO2. These values have to be
compared to the 3 µs half-lifetime of the reference heterodyad
5+|(e-)TiO2. Using known values for the roughness factor of
the nanocrystalline film (F ) 700), its absorbance after dying
(A480 nm ) 0.5), and the incident laser pulse energy (1.5 mJ), the
average number of electrons initially injected per individual TiO2
Figure 4. Transient absorbance signals recorded upon laser irradiation
on nanocrystalline TiO2 films sensitized by adsorption of 1, 2, 3 + 5,
and 5 (heterotriads 1|TiO2, 2|TiO2, (3+5)|TiO2, and heterodyad 5|TiO2).
Pulse width ) 5 ns, pulse energy ) 1.5 mJ. (a) Signals measured in
propylene carbonate at λ ) 500 nm (irradiation at λ ) 430 nm),
corresponding to the temporal evolution of the bleaching of the MLCT
ground-state absorption of the chromophore S: [a] 5|TiO2, [b] 1|TiO2,
and [c] 2|TiO2; signals measured at λ ) 750 nm, corresponding to the
temporal evolution of the absorption of the radical cation of the donor
(D+) moiety of the heterotriads: [d] 1|TiO2 and (e) 2|TiO2. (b) Signals
of D+ measured in propylene carbonate at λ ) 620 nm (irradiation at
λ ) 480 nm) with heterotriads 1|TiO2, 2|TiO2, and 3+5|TiO2,
respectively. The observed decays are due to the recombination of the
D+ radical cation species with electrons injected in the conduction band
of the semiconductor. (c) Signals of D+ measured at λ ) 620 nm
(irradiation at λ ) 480 nm) with 2|TiO2 in two different media, at
complete and half monolayer coverage respectively: [a] propylene
carbonate, 100% surface coverage; [b] same medium, 50% surface
coverage; [c] 1-ethyl-2-methylimidazolium bis(trifluoromethane)sulfon-
imide, 100% surface coverage; [d] same medium, 50% surface
coverage. Absorbance changes were normalized to the same initial
amplitude.
particle (10 nm radius) is estimated to be around x ) 50.
0
Correlation between injected electrons in the solid and D+ on
the surface being unlikely, such a high average initial occupancy
of the semiconductor particles by charge-separated pairs should
allow the use of a second-order rate law to describe the kinetics
of their recombination.27 The differences in the recombination
rates of the different heterotriads can be correlated with the mean
distance separating the amine from the surface of the oxide.
Perpendicular attachment of the molecules on the surface would
imply D-TiO2 distances of 12, 18, and 24 Å for the heterotriads
(3+5)|TiO2, 1|TiO2, and 2|TiO2, respectively (Figure 5). As-
suming a typical damping factor â ) 1.2 Å-1 for “through-
space” electronic coupling between conduction-band electrons
and D+, a decrease of the recombination rate by 1 order of
magnitude is expected to correspond to an increase of the mean
electron-transfer distance by 1.9 Å. The discrepancy between
that value and the 6 Å distance differences expected from
perpendicular anchoring can be explained by the fact that, very
the system belongs to the terpyPO3 ligand which anchors onto
the TiO2 surface, resulting in an efficient electronic coupling
between the excited-state S* and the semiconductor acceptor
levels. The observed rate for hole-transfer process (eq 5b) is
orders of magnitude smaller than for the electron injection (eq
3) from an efficient sensitizer.26 This consideration, combined
with the absence of spectral detection of the D+-S-|TiO2,
allows for the exclusion of process 6 as a possible oxidation
pathway of the amine besides 5a and b. The lower rate for the

Oxide Films Sensitized by Molecular Dyads
J. Am. Chem. Soc., Vol. 121, No. 6, 1999 1333
corresponding to 50% of a full monolayer was acheived by brief
bathing of the TiO2 nanocrystalline films in the sensitizer
solution, until absorbance at 480 nm was 0.25. Upon irradiation
by a laser pulse of fixed energy, this partial covering implies a
50% reduction of the number of charge-separated pairs.
Considering a second-order kinetic law, the rate of recombina-
tion is expected to decrease accordingly. It was observed on
the contrary (Figure 4c) that the slowest kinetic components
measured for 2|TiO2 at surface saturation disappeared under
partial coverage conditions, while τ1/2 was reduced by more than
an order of magnitude. An effect of the same amplitude was
observed with heterotriads 1|TiO2 and (3+5)|TiO2. A more
detailed study of this phenomenon, which is attributed to a
change in the geometry associated with the electron-transfer
process, is currently underway and will be reported elsewhere.
Heterotriad 2|TiO2 represents obviously the most promising
system for long-lived charge separation. In an attempt to further
increase the lifetime of 2+|(e-)TiO2 state, the photodynamics
was studied in the ambient temperature liquid salt 1-ethyl-2-
methylimidazolium bis(trifluoromethylsulfonyl)imide28 which
was shown to behave at the molecular scale like a solvent of
low dielectric constant (ꢀ e 10), where the medium reorganiza-
tion energy should be minimized. Figure 4c shows that the
charge recombination process 9 was indeed slowed significantly,
with a τ1/2 twice as long as that in propylene carbonate. A very
long tail in the kinetic curve, accounting for a few percent of
the initial absorbance, was even found to extend up to several
hundreds of milliseconds (τ1/4 ) 5 ms, τ1/8 ) 120 ms).
Interestingly, carefully degassing the molten salt by successive
freeze-pump-thaw cycles did not affect this long time kinetic
phase, indicating that removal of the conduction-band electrons
by molecular oxygen was not competing with the charge
recombination.
Photochromism. Heterotriads 1|TiO2, 2|TiO2, and (3+S)|TiO2
display redox-type photochromism. When electrodes made of
transparent, nanocrystalline TiO2 film on conducting glass were
derivatized by a monolayer of the diads 1, 2, 3+4, or 3+5,
their absorption spectrum could be changed by the combined
action of visible light and applied potential. If the applied
potential is maintained higher than the conduction-band level
of the semiconductor, but still lower than the redox potential
of the triarylamine moiety, then the electron-hole pairs created
by the photoinduced charge separation process D-S|TiO2 f f
D+-S|(e-cb)TiO2 can no longer recombine. Figure 6 shows the
change in the absorption spectra that occurred under illumination
with white light (intensity 1 ( 0.2 sun (AM 1.5)) of the different
heterotriad-derivatized electrodes polarized at + 0.50 V. The
spectra of the oxidized states were recorded after 10 min
illumination, when a steady-state plateau was reached.
Figure 5. Distances, energy levels, and electron-transfer half-lives in
the heterotriads 1|TiO2, 2|TiO2, and 3+5|TiO2. The distances were
calculated by molecular mechanics using CAChe software, assuming
a perpendicular arrangement of the molecules on the surface. The energy
levels are those obtained and calculated by electrochemistry as well as
by absorption and emission spectroscopy. The half-lives for the electron
transfers are those obtained by laser flash experiment. The value of 3
µs for the S+|(e-cb)TiO2 f S|TiO2 recombination was obtained with
the heterodyad 5|TiO2.
If, after reaching the steady state, the electrical circuit was
opened, the spectrum of the D+-S|TiO2 state remained
unchanged for days, provided the contacting electrolyte was free
of reducing or nucleophilic impurities, for example, the ambient-
temperature liquid salt 1-ethyl-2-methylimidazolium bistrifluo-
romethane sulfonimide.28
Rapid reduction D+-S|(e-cb)TiO2 f D-S|TiO2 to the initial
state was achieved by applying a negative potential (e0 V) to
the electrode in order to pass electrons above the TiO2
conduction band. Even though the system was not yet tested
for long-term stability, it appears to sustain at least 10 light-
induced oxidation-reduction cycles without degradation of the
heterotriads.
likely, a large portion of adsorbed molecules adopt a tilted
conformation on the surface, leading to a broad distribution of
distances for the electron transfer 9, and hence to a complex
kinetic behavior. In fact, attempts to fit the observed temporal
curves demonstrated that they are multiphasic, certainly because
the adsorption geometry of the molecules is not fixed. This
geometry is likely to be dependent on the surface coverage. In
a compact monolayer, perpendicular anchoring should be
favored, whereas molecules widely dispersed on the surface
should be freer to lean over. It was indeed observed for all three
heterotriads that the rate of the recombination reaction 9 depends
drastically on the surface concentration. A surface coverage
For the heterotriad 1|TiO2, the observed behavior follows the
above-described mechanism implying only interfacial electron

1334 J. Am. Chem. Soc., Vol. 121, No. 6, 1999
Bonhoˆte et al.
Figure 7. Proposed electron fluxes in the illuminated heterotriad 2|TiO2
on SnO2. Light-induced excitation of the sensitizer S is followed by
electron injection (a) into the TiO2 and oxidation (b) of the donor D.
Lateral conduction (c) inside the monolayer allows electrons to flow
from the SnO2 (d) to the oxidized donors.
Table 3. Photochromic Ratio R and Oxidized Part of Heterotriad
2|TiO2 in the Steady State under AM 1.5 Illumination at +0.5 V, as
a Function of the Molar Fraction x of the Dyad D-S or Donor 3 on
the Surfacea
heterotriad
underlayer
xa
Rb
oxidized partc
1|TiO2
no
1.0
1.0
0.5
0.2
0.5d
0.5d
0.46
0.53
0.57
0.71
0.52
0.76
0.68
0.55
0.59
0.74
0.18
0.27
2|TiO2
yes
yes
yes
no
2|TiO2
2|TiO2
3+4|TiO2
3+4|TiO2
yes
a x on the surface was assumed to be equal to the molar fraction of
the relevant molecules to the total concentration of phosphonates in
the derivatizing solution. The validity of this assumption was demon-
strated previously.17 b Photochromic ratio between the MLCT and the
D+ absorbances (see Table 2). c By dividing R by the corresponding
ratio obtained for the corresponding oxidized dyad in solution. d Molar
fraction of 3.
was gained from the observation of a percolation threshold when
the surface concentration of the electroactive molecules 3 was
progressively increased in a monolayer of nonelectroactive
coadsorbate. Below a molar fraction x ) 0.5 of 3, the monolayer
is electrochemically entirely inactive but abruptly becomes
active above that limit, with an apparent diffusion coefficient
increasing with x. Heterotriad 2|TiO2, with its freely rotating
triarylamine, is apparently also able to build hole (or electron)-
conducting monolayers. The lateral conductivity efficiently
competes with light-induced interfacial oxidation D-S|TiO2
f f D+-S|(e-cb)TiO2, forming a short-circuit according to
Figure 7, and thus prevents the appearance of photochromism
in the present case. This limitation could be circumvented by
deposition of a “compact” TiO2 layer between the SnO2
conducting glass and the nanocrystalline TiO2 film. Such a layer
was simply built by dipping the electrode in a Ti(OBu)4 solution
followed by hydrolysis in air. Clearly, such a layer is not
completely chemically insulating the conducting glass and
contains cracks. It sufficiently reduces, however, the contact
between the monolayer 2|TiO2 and the SnO2 to impede the
short-circuit reaction, as shown in Figure 6b and Table 3 by
the spectacular photochromism of the considered heterotriad in
this situation. When dyads 2 were diluted on the surface by the
insulating molecules 19, the photochromic ratio and the steady-
state proportion of oxidized 2|TiO2 increased with decreasing
concentration (see Table 3) consistently with a reduction of the
lateral electron transfer caused by the introduction of the
insulating molecules in the monolayer, similarly to what was
observed in the monolayers of 3.
Figure 6. Absorption spectra of the heterotriads D-S|TiO2 under
positive polarization (0.5 V), in the dark (dotted lines) and after 10
min illumination (full lines). Thickness of the nanocrystalline films:
4.8-5 µm, roughness factor 400 < η < 450. Interference pattern in
the spectra are due to the nanocrystalline layer.
transfer. The absorption spectrum after light-induced oxidation
(D+-S|TiO2) resembles that of 1+ in solution (Figure 6a). The
ratio R between the absorbance of D+ at 752 nm (0.491) and
the MLCT absorbance at 486 nm (1.07), hereafter called
“photochromic ratio” is 0.46. For 1+ in solution, the absorbance
ratio for the same wavelengths is 0.68 (see Table 2). Dividing
the former ratio by the latter affords the fraction of the dyad
which was oxidized (D+-S|TiO2) in the steady-state, in this
case 67%.
For the heterotriad 2|TiO2, the situation is more complex. If
the experiment was carried out as with heterotriad 1|TiO2, no
photochromism was observed. Moreover, if the electrode
potential was increased up to 1 V, complete oxidation of 2 to
2+ occurred without illumination, which was manifested by the
electrochromism of the electrode. The same electrochromic
behavior was observed with nanocrystalline layers of TiO2
derivatized by the phosphonated triarylamine 3. The phenom-
enon was attributed to hole injection from the underlying SnO2
to the contacting triarylamine molecules, followed by an efficient
charge transfer inside the densly packed self-assembled mono-
layer of electroactive molecules.17 Support for this hypothesis
The photochromism observed with the system (3+4)|TiO2
shows that a chromophore and an electron donor coadsorbed
on a surface can form a noncovalently linked dyad, as discussed

Oxide Films Sensitized by Molecular Dyads
J. Am. Chem. Soc., Vol. 121, No. 6, 1999 1335
Figure 8. Energy levels and electron fluxes in a dye-sensitized solar cell. Left: regenerative cell with redox mediator in solution. Right: setup for
measurement of the photopotential versus a reference eletrode, in the absence of redox mediator. S is the dye, M is the redox mediator, Iinj is the
injection flux, Irec is the recombination flux, IM is the current flux to M+ and I is the net current delivered by the system.
Table 4. Photopotentials VF in Absence of Redox Mediator as
above in the photodynamics section. Here again, the photo-
chromic steady-state appears to establish as a balance between
two opposite fluxes, the light-induced oxidation of 3 and the
lateral conductivity allowing electrons to percolate back from
the SnO2 to oxidized 3.
Well as Open Circuit Photovoltage Isc and Short Circuit Current Isc
in a Regenerative Solar Cell with I-/I3 as a Redox Couplea
-
1|TiO2
2|TiO2
5|TiO2
b
[I-]
VF /Uoc
Isc
VF /Uoc
Isc
VF /Uoc
Isc
b
c
b
c
c
Photovoltaics. The link between photodynamics and photo-
voltaics was established according to the following model,
depicted in Figure 8. Under illumination in open circuit, the
electrons injected into the TiO2 conduction band at the flux Iinj
can either recombine with the oxidized sensitizer with the pseudo
first-order rate constant krec or reduce the oxidized redox
mediator M+ in solution with the second-order rate constant
kM. The steady-state electron density n in the semiconductor is
thus given by eq 10. The quasi-Fermi level F of the semicon-
ductor is varying with the electron density according to eq 11,
where Ec is the conduction band level and Nc is the effective
density of states in the conduction band, a constant value for a
given material at a given temperature.29 The corresponding
photopotential VF of the electrode is obtained by division by
the electronic charge e (eq 12). Combining eqs 10 and 12 affords
the photopotential as a function of the kinetic parameters (eq
13).
0.0
0.5
729
692
0.0
2.6
616
612
0.0
1.2
647
553
0.0
1.8
a Average of two electrodes; 10 µm nanocrystalline TiO2 on
“compact” TiO2 on SnO2 conducting glass; estimated uncertainty: (20
mV. Solvent: propylene carbonate. b In mV vs NHE; estimated
uncertainty: (20 mV. c In mA/cm2.
M tetrabutylammonium triflate). The open-circuit photovoltages
versus the I-/I3- redox couple and the short-circuit photocurrents
Uoc were measured in a regenerative cell, with 1,2-dimethyl-
3-propylimidazolium iodide (0.475 M)/triiodide (0.025 M) as
electrolyte. The results are presented in Table 4.
Heterotriad 1|TiO2 gave an 82-mV higher photopotential than
5|TiO2. This observation can only be in part (58 mV) accounted
for by the 10-fold reduced recombination rate. Another contri-
bution of 10 mV is expected from the 50% higher injection
flux Iinj of the former system in polychromatic light inferred
from the values of Isc in the regenerative cell and due to the
wider absorption spectrum.
Despite a 100-fold lower recombination rate compared with
5|TiO2, expected to show itself by a 116-mV gain, the
heterotriad 2|TiO2 gave the lowest photopotential of the series.
This observation must be related to a supplementary contribution
to Irec represented by the short-circuiting pathway described
above, constituted by the lateral charge percolation from
triarylamine to triarylamine and finally to the SnO2. It was in
fact shown that a monolayer of the model compound 3 on a
5-µm thick nanocrystalline TiO2 film on conducting glass is
able to sustain currents as high as 5 mA/cm2 of electrode at 1.0
V.
In a regenerative solar cell, with the I-/I3- redox couple, the
open circuit photovoltage of 1|TiO2 was now 139 mV higher
than that of 5|TiO2. This increased difference can be attributed
to a reduced electron-triiodide reaction flux (IM). In fact, in a
regenerative cell in open-circuit, recombination is no longer the
only way of escape for the electrons in the conduction band,
since they can also reduce the oxidized mediator in solution.
The measured values of IM in the dark, under 550 mV reverse
bias were 6.8 mA/cm2 with 5|TiO2 but only 4 mA/cm2 with
1|TiO2 and 3.4 mA/cm2 with 2|TiO2, probably as a result of a
restricted access of the triiodide ions to the semiconductor
caused by the presence of the bulky triarylamine groups.
Heterotriad 2|TiO2, which photopotential was 31 mV lower than
5|TiO2, affords in the regenerative cell a 59 mV higher
n ) Iinj/(krec + kM [M+])
F ) Ec - kT ln(Nc/n)
(10)
(11)
(12)
VF ) Vc - (kT/e) ln(Nc/n)
VF ) Vc - (kT/e) ln(Nc (krec + kM [M+])/Iinj
)
(13)
(14)
∆VF ) (kT/e) ln(krecA injB/krecBIinjA)
I
If two systems A|TiO2 and B|TiO2, characterized by krecA
,
I
injA and krecB, IinjB, respectively, are illuminated in an electrolyte
free of redox mediator, then their photopotential difference ∆VF
can be expressed by eq 14. It must be specified that the laser
flash experiment does not afford the pseudo-first-order rate
constants krec but rather rates of the type krec′ ) n′krec, since the
electron density n′ in the semiconductor after the charge injection
induced by the laser flash is not exactly known. This density
will be taken as identical for all of the considered laser
experiments, allowing, consequently, the use of the kinetic
values derived from them in eq 14. To account for the
multiphasic nature of the decays, synthetic rates corresponding
to the first half-lives of the charge-separated states will be
considered.
The photopotentials of the three systems 1|TiO2, 2|TiO2, and
5|TiO2 were measured under full sun illumination (AM 1.5) in
propylene carbonate with an inert supporting electrolyte (0.1

1336 J. Am. Chem. Soc., Vol. 121, No. 6, 1999
Bonhoˆte et al.
Table 5. Photovoltaic Quantum Yield at 550 nm of Heterotriad
2|TiO2
being obtained with the latter system in which that localization
occurs on the ligand bound to the semiconductor surface. Thus,
not only the energy levels of the D and S moieties of the
heterotriad must be adjusted to allow the charge separation, but
also the geometry of the LUMO of S must be considered.
The lifetime of the charge-separated state is controlled by
the distance between the electron donor and the surface of the
semiconductor. This distance could be further increased in new
heterotriads as, even in 2|TiO2, the hole-transfer process (eq
5b) is still much faster than the recombination (eq 5c). A long
and well-defined D-S distance in the molecular dyad is not,
however, a sufficient condition to achieve a long-lived charge
separation in the heterotriad. The orientation of the molecules
in the monolayer plays a crucial role, and care must be taken
that a compact monolayer is obtained, ensuring the largest
distance of charge separation.
xa
absorptionb (%)
IPCEc (%)
quantum yield (%)
1.00
0.80
0.65
0.50
0.20
77
44
30
24
20
37
32
24
24
19
48 ( 2
73 ( 4
81 ( 6
100 ( 8
96 ( 9
a Molar fraction on the surface. See remark in Table 3. b Light
absorption due to 2 at 550 nm, average of 3 points of the electrode.
Estimated nonsystematic error: (1% of absorption. c Incident photon
to current efficiency at 550 nm. Estimated nonsystematic error: (1%
of IPCE. Electrolyte: 1 M LiI, 0.01 M LiI3 in propylene carbonate.
photopotential, very likely because, in addition to a lower IM,
the reduction of 2+|TiO2 by iodide competes with the lateral
electron transfer.
The efficient lateral charge percolation inside a monolayer
of closely packed triarylamine studied in detail in the system
3|TiO217 appears to also occur with 2|TiO2, opening an important
short-circuiting pathway in the photoanode incorporating that
heterotriad, leading to poor photovoltaic performances. There-
fore, in new dyads, donor groups must be chosen which are
not prone to build such conducting assemblies. Alternatively,
the electronic contact between the molecular monolayer and the
underlying conducting substrate must be interrupted by the
intercalation of a compact semiconductor layer between the
nanocrystalline film and that of the substrate. On the way to
improved dye-sensitized solar cells based on heterotriads, the
next step that must be studied is the hole transfer from the
oxidized donor to the redox mediator in solution. The process
should be at least as efficient as that with the sensitizer alone.
The photoelectrochromism of the photoanode made of the
investigated heterotriads constitutes a new phenomenon that
opens the way to a new type of application for the nanocrys-
talline TiO2 films. Self-adapting filters and windows exploiting
that property can be conceived. Information storage supports
based on the same principle could also be feasible, provided
the lateral charge migration inside the monolayer could be
blocked.
To further elucidate the limiting effect of the lateral electron
transfer on the photovoltaic performance of the heterotriad
2|TiO2, the quantum yield of the whole charge separation
process (eq 5a and b) was measured in a mixed monolayer of
2 and 19. That value which can be regarded as the number of
electrons delivered by one molecule in the electrical circuit per
100 absorbed photons, was obtained by dividing the IPCE
recorded at a wavelength (550 nm) where no saturation occurs
(A < 1), by the fraction of absorbed light. It appears clearly
(Table 5) that this yield increased with the surface dilution,
which can be attributed to the concomitant reduction of the
recombination caused by the lateral electron transfer, when the
dye molecules are moved away from each other.
Conclusions
Photoinduced charge separated states reaching half-life times
of 300 µs were achieved by combining intramolecular and
interfacial electron transfer. Spectroelectrochemistry, resonance
Raman spectroscopy, and laser flash photolysis have allowed
detailed understanding of the relevant parameters governing the
charge separation and recombination processes in the considered
heterotriads, opening the way to an improved efficiency of the
light-to-electricity conversion process in the dye-sensitized
nanocrystalline solar cell. The efficiency of the photoinduced
charge separation was shown by comparison between heterot-
riads 1|TiO2 and 2|TiO2 to depend on the localization of the
electron in the excited state of the chromophore, the best yield
Acknowledgment. This work was supported by the Swiss
National Science Foundation (FNRS).
JA981742J