The crystalline nanocluster phase as a medium for structural and spectroscopic studies of light absorption of photosensitizer dyes on semiconductor surfaces

Article abstract of DOI:10.1021/ja909600w

The crystalline nanocluster phase, in which nanoscale metal oxide clusters are self-assembled in three-dimensional periodic arrays, is described. The crystalline assembly of nanoparticles functionalized with technologically relevant ligands offers the opportunity to obtain unambiguous structural information that can be combined with theoretical calculations based on the known geometry and used to interpret spectroscopic and other information. A series of Tl/O clusters up to ～2.0 nm in diameter have been synthesized and functionalized with the adsorbents catechol and isonicotinic acid. Whereas the isonicotinate is always adsorbed in a bridging monodentate mode, four different adsorption modes of catechol have been identified. The particles show a significantly larger variation of the Tl-O distances than observed in the known TiO₂ phases and exhibit both sevenfold overcoordination and five- and fourfold undercoordination of the Tl atoms. Theoretical calculations show only a moderate dependence of the catecholate net charge on the geometry of adsorption. All of the catechol-functionalized clusters have a deep-red color due to penetration of the highest occupied catechol levels into the band gap of the Tl/O particles. Spectroscopic measurements of the band gap of the Tl ₁₇ cluster are in good agreement with the theoretical values and show a blue shift of ～0,22 eV relative to those reported for anatase nanoparticles.

Full text of DOI:10.1021/ja909600w

Published on Web 02/10/2010
The Crystalline Nanocluster Phase as a Medium for Structural
and Spectroscopic Studies of Light Absorption of
Photosensitizer Dyes on Semiconductor Surfaces
Jason B. Benedict* and Philip Coppens*
Department of Chemistry, UniVersity at Buffalo, State UniVersity of New York, Buffalo,
New York 14260-3000
Received November 11, 2009; E-mail: coppens@buffalo.edu
Abstract: The crystalline nanocluster phase, in which nanoscale metal oxide clusters are self-assembled
in three-dimensional periodic arrays, is described. The crystalline assembly of nanoparticles functionalized
with technologically relevant ligands offers the opportunity to obtain unambiguous structural information
that can be combined with theoretical calculations based on the known geometry and used to interpret
spectroscopic and other information. A series of Ti/O clusters up to ∼2.0 nm in diameter have been
synthesized and functionalized with the adsorbents catechol and isonicotinic acid. Whereas the isonicotinate
is always adsorbed in a bridging monodentate mode, four different adsorption modes of catechol have
been identified. The particles show a significantly larger variation of the Ti-O distances than observed in
the known TiO2 phases and exhibit both sevenfold overcoordination and five- and fourfold undercoordination
of the Ti atoms. Theoretical calculations show only a moderate dependence of the catecholate net charge
on the geometry of adsorption. All of the catechol-functionalized clusters have a deep-red color due to
penetration of the highest occupied catechol levels into the band gap of the Ti/O particles. Spectroscopic
measurements of the band gap of the Ti17 cluster are in good agreement with the theoretical values and
show a blue shift of ∼0.22 eV relative to those reported for anatase nanoparticles.
and examined their catalytic activities in esterification reactions.7
Sanchez and co-workers synthesized the cluster [Ti17O24(OPri)20];8
Introduction
this species and the clusters [Ti18O28H(OBut)17]) and
([Ti18O22(OBun)26(acac)2]9 are the largest Ti/O clusters reported
in the literature.
Semiconductor surfaces are key components of molecule-
based photovoltaic cells, and TiO2 surfaces in particular have
therefore received extensive attention.1 A series of theoretical
calculations on titanium oxide particle/molecular adsorbent
complexes have been reported,2-5 but detailed diffraction
information on the structure of the photosensitizer-semiconductor
surface is not available. It has not been generally recognized
that a number of crystalline materials are known in which
nanoscale metal oxide clusters are aligned in periodic arrays
(see Figure 1), the structure of which can be determined
accurately by diffraction methods. TiO2 cluster phases have been
studied because of their catalytic properties,6 such as in the
photoreduction of CO2 to methane and methanol and in
esterification and polymerization reactions, as well as their use
as precursors in the synthesis of metal oxide ceramics. A number
of structure determinations have been reported. Corden and co-
workers synthesized di-, tri-, and pentanuclear titanium species
The work described here was specifically aimed at examining
the properties of the nanoclusters as models for TiO2 surfaces
across which electron injection takes place in photovoltaic cells.
After a description of the synthetic methods, we discuss the
experimentally determined structures, their comparison with
known TiO2 phases, the geometry of adsorption of dye
molecules, their spectroscopic properties, and the nature of the
bonding as revealed by the experimental results and parallel
theoretical calculations based on the experimental geometry.
Experimental Section
Synthetic Strategies. Two synthetic strategies were employed.
In the first, new clusters were synthesized using a monomeric
titanium alkoxide precursor either with or without a photosensitizer
dye such as catechol (cat-H2, pyrocatechol) or pyridyl-4-carboxylic
acid (isonicotinic acid, INA) added to the reaction mixture.
Progressively larger clusters with nTi ) 2, 3, 4, 6, and 17 Ti atoms
were obtained by increasing the temperature at which the reaction
(1) Diebold, U. Surf. Sci. Rep. 2003, 48, 53–229.
(2) Duncan, W. R.; Prezhdo, O. V. Annu. ReV. Phys. Chem. 2007, 58,
143–184.
(3) Persson, P.; Ojamae, L. Chem. Phys. Lett. 2000, 321, 302–308. Persson,
P.; Lundqvist, M. J. J. Phys. Chem. B 2005, 109, 11918–11924.
(4) Lundqvist, M. J.; Nilsing, M.; Lunell, S.; Åkermark, B.; Persson, P.
J. Phys. Chem. B 2006, 110, 20513–20525.
(7) Corden, J. P.; Errington, W.; Moore, P.; Partridge, M. G.; Wallbridge,
M. G. H. Dalton Trans. 2004, 1846–1851.
(8) Steunou, N.; Kickelbick, G.; Boubekeur, K.; Sanchez, C. J. Chem.
Soc., Dalton Trans. 1999, 3653–3655. Steunou, N.; Portal, R.; Sanchez,
C. High Pressure Res. 2001, 63–70.
(5) Lundqvist, M. J.; Nilsing, M.; Persson, P.; Lunell, S. Int. J. Quantum
Chem. 2006, 106, 3214–3234.
(9) Campana, C. F.; Chen, Y.; Day, V. W.; Klemperer, W. G.; Sparks,
R. A. J. Chem. Soc., Dalton Trans. 1996, 691–702. Toledano, P.; In,
M.; Sanchez, C. C. R. Acad. Sci. Paris, Ser. II 1991, 313, 1247.
(6) Li, G.; Dimitrijevic, N. M.; Chen, L.; Nichols, J. M.; Rajh, T.; Gray,
K. A. J. Am. Chem. Soc. 2008, 130, 5402–5403.
9
2938 J. AM. CHEM. SOC. 2010, 132, 2938–2944
10.1021/ja909600w  2010 American Chemical Society

Studies of Photosensitizer Dyes on Semiconductor Surfaces
A R T I C L E S
Figure 1. Periodic arrangement of the nanoclusters in crystals of (left) Ti6cat6 and (right) Ti17, with the structures of individual molecules (insets). Ti,
purple; O, red; C, gray; N, blue.
was carried out. In the second approach, larger bare titanium
alkoxide clusters with nTi ) 17 were reacted with the selected
adsorbent molecules.
Ti6(µ3-O)6(INA)6(OPri)6 [Ti6INA6]. The procedure was identical
to that for Ti3INA3, except that the reaction was carried out at 100
°C.
Ti6(µ4-O)(cat)6(OPri)10 ·toluene [Ti6cat6]. To a 250 mL Schlenk
flask containing a magnetic stir bar were added pyrocatechol (0.400
g, 3.63 mmol) and toluene (30 mL). The flask was then sealed with
a rubber septum, pumped and purged with argon (three cycles),
and placed in a room-temperature oil bath with stirring. The initially
colorless solution became dark-red when Ti(OPri)4 (1.0 mL, 3.63
mmol) was added dropwise via syringe. After the solution was
heated to 100 °C for 48 h, the solvent was removed in vacuo. The
flask was then transferred to the drybox, where the sticky red residue
was redissolved in toluene. Slow evaporation of the solvent yielded
red crystals of Ti6cat6 suitable for structure determination.
Ti17(µ4-O)4(µ3-O)16(µ2-O)4(OPri)20 [Ti17].8 In air, Ti(OPri)4 (8.4
mL, 29 mmol) was quickly added via syringe to acetic acid (2.1
mL, 37 mmol) in a Teflon-lined Parr bomb with a 23 mL capacity.
The bomb was sealed, placed in a 150 °C oven for 5 days, and
then cooled to room temperature, after which the bomb was opened
and the solution was quickly transferred by pipet to a Schlenk flask.
Colorless crystals of Ti17 precipitated immediately. The atmosphere
was replaced with Ar in several pump-purge cycles. This
compound was identical to the one described previously by Sanchez
and co-workers.8
Chemicals. All of the reagents and solvents were purchased from
commercial sources and used without further purification: Ti(OPri)4
(Sigma-Aldrich), toluene (EMD), hexane (Aldrich), heptane (J. T.
Baker), benzene (Aldrich), isopropyl alcohol (Fisher), pyrocatechol
(Eastman Organic Chemicals), and isonicotinic acid (Alfa Aesar).
Syntheses. Reactions and recrystallizations of the titanium
compounds were carried out under an argon atmosphere using
standard drybox or Schlenk techniques.
Ti2(cat)2(OPri)4(HOPri)2 [Ti2cat2]. Recrystallization of 2 mL
of the sticky residue formed in the Ti6cat6 reaction (see below)
from isopropanol yielded several small pale-red crystals of Ti2cat2
suitable for structure determination.
Ti2(cat)2(cat-H)4 [Ti2cat6]. To a 250 mL Schlenk flask containing
a magnetic stir bar were added pyrocatechol (1.20 g, 10.9 mmol)
and toluene (30 mL). The flask was then sealed with a rubber
septum, pumped and purged with argon (three cycles), and placed
in a room-temperature oil bath with stirring. The initially colorless
solution turned dark-red when Ti(OPri)4 (1.0 mL, 3.63 mmol) was
added dropwise via syringe. After the solution was heated to 100
°C for 48 h, approximately two-thirds of the solvent was removed
in vacuo. The flask was then transferred to the drybox, after which
the remainder of the solvent was removed by slow evaporation,
yielding a red powder and red crystals of Ti2cat6 suitable for
structure determination.
Ti17(µ4-O)4(µ3-O)16(µ2-O)4(INA)4(OPri)16 [Ti17INA4]. INA (2.9
mg, 23.6 µmol) was partially dissolved in 3 mL of toluene in a
small glass vial. To this was added dropwise a solution of Ti17
(14.2 mg, 5.97 µmol) in 3 mL of toluene. Slow evaporation of the
dark-red solution yielded colorless crystals of Ti17INA4 suitable
for structure determination.
Ti3(µ3-O)(INA)3(OPri)7. [Ti3INA3]. To a 250 mL Schlenk flask
containing a magnetic stir bar were added INA (0.447 g, 3.63 mmol)
and toluene (30 mL). The flask was then sealed with a rubber
septum, pumped and purged with argon (three cycles), and placed
in a room-temperature oil bath with stirring. The initially cloudy
solution clarified when Ti(OPri)4 (1.0 mL, 3.63 mmol) was added
dropwise via syringe. After the solution was heated to 60 °C for
12 h, ∼60% of the solvent was removed in vacuo. The flask was
then transferred to the drybox, where the pale-yellow solution was
allowed to slowly evaporate, yielding colorless crystals of Ti3INA3
suitable for structure determination.
Ti4(µ4-O)(µ2-O)(INA)2(OEt)10 [Ti4INA2]. To a 250 mL Schlenk
flask containing a magnetic stir bar were added INA (0.447 g, 3.63
mmol) and ethanol (30 mL). The flask was then sealed with a rubber
septum, pumped and purged with argon (three cycles), and placed
in a room-temperature oil bath with stirring. The initially cloudy
solution clarified when Ti(OPri)4 (1.0 mL, 3.63 mmol) was added
dropwise via syringe. After the solution was heated to 60 °C for
12 h, the solvent was removed in vacuo. The flask was then
transferred to the drybox, where 10 mL of toluene was added to
the residue. This solution was filtered. Following slow evaporation
of the solvent, colorless crystals of Ti4INA2 suitable for structure
determination were obtained.
Ti17(µ4-O)4(µ3-O)16(µ2-O)4(cat)4(OPri)16 [Ti17cat4]. In a small
glass vial, Ti17 (18.4 mg, 7.73 µmol) was dissolved in 3 mL of
toluene. To this was added dropwise a solution of pyrocatechol
(3.4 mg, 30.8 µmol) in 3 mL of toluene. Slow evaporation of the
dark-red solution yielded red crystals of Ti17cat4 suitable for
structure determination.
Data Collection. X-ray diffraction data were collected on a
Bruker SMART APEX2 CCD diffractometer installed at a rotating
anode source (Mo KR, λ ) 0.71073 Å) and equipped with an
Oxford Cryosystems nitrogen gas flow apparatus. In general, data
were collected at 90 K, but in a few cases, higher temperatures
were required because of crystal disintegration upon cooling. The
oscillation method (ω scans, 180° per scan, 0.5° per frame) was
used for data collection. The crystallographic data are summarized
in Table S1 in the Supporting Information.
Results and Discussion
Structure. Figure 1 illustrates two examples of the periodic
nature of the cluster phases, while the polyhedral building blocks
of the small clusters with adsorbed catechol and INA adsorbents
are shown in Figure 2.
9
J. AM. CHEM. SOC. VOL. 132, NO. 9, 2010 2939

A R T I C L E S
Benedict and Coppens
Figure 2. Views of the structures of small isolated titanium complexes with catechol and isonicotinate adsorbents, showing the polyhedral building blocks
(isopropyl groups have been omitted for clarity).
Comparison of the Cluster Structures with the Pure TiO2
Polymorphs. The three most common crystalline phases of
titanium dioxide are rutile (P42/mnm, a ) b ) 4.584 Å, c )
2.953 Å),10 anatase (I41/amd, a ) b ) 3.782 Å, c ) 9.502 Å),11
and brookite (Pbca, a ) 5.436 Å, b ) 9.166 Å, c ) 5.135
Å).12 In all three structures, titanium is coordinated to six oxygen
atoms, forming a distorted octahedron. In the structure of rutile,
the octahedron is elongated, with axial and equatorial bond
lengths of 1.988 and 1.944 Å, respectively, and shares two
opposing equatorial edges with neighbors. In anatase, the
corresponding Ti-O bond lengths are similar at 1.980 and 1.934
Å; however, the octahedra share two pairs of two adjacent edges
with neighboring octahedra. The brookite structure is strongly
distorted, with the lengths of bonds trans to each other equal to
2.040 and 1.865 Å, 1.993 and 1.919 Å, and 1.993 and 1.946 Å,
with corresponding distortions in the bond angles. Three of the
12 edges of each octahedron are shared between two octahedra,
with two of the edges adjacent to each other. In all three
structures, the individual oxygen atoms are coordinated to three
different titanium atoms (µ3-O).
The Ti/O cores of the nanoclusters share some structural
similarities with the pure-phase titanium dioxide crystal struc-
tures, such as common (but not exclusive) sixfold coordination
of the titanium atoms and the presence of µ3-O bridges.
However, the clusters also possess structural features absent in
the bulk pure-phase crystals, such as undercoordination and
overcoordination of titanium (the latter found in Ti6cat6), which
are likely typical for the surfaces of the TiO2 phases. Li et al.6
attributed the increased reactivity of nanocluster crystals to the
existence of tetrahedral Ti4+ sites, described as “catalytic hot
spots”. In the clusters examined here, four-, five-, and sevenfold
coordination of titanium occurs in addition to the more common
sixfold coordination. Not surprisingly, the clusters show a much
larger variation of the Ti-O bond lengths and octahedral edge
sharing than the pure titanium dioxide phases. The Ti-O bond
lengths vary between 1.75 and 2.30 Å, compared with values
of 1.934 and 1.980 Å in anatase, for example. As described
above, the variation is larger in brookite, which has a much
less regular structure, but is still less than that observed in the
clusters. Details of the octahedral packing are dependent on the
adsorbed species, as is clear from a comparison of the catechol
and isonicotinate adducts of the nTi ) 6 clusters (Figure 2). The
distorted octahedra of the nTi ) 2, 3, 4, and 6 (INA- only)
clusters lack clear axial and equatorial directions in the octahedra
and most strongly resemble the distorted structure of brookite.
The Ti17 core has a four-coordinate titanium at its center that
is surrounded by 12 six-coordinate and four five-coordinate
titanium centers (the surrounding titanium polyhedra are colored
blue and green, respectively, in the left column of Figure 3).
Upon reaction of Ti17 with catechol and INA, the five-coordinate
centers are converted to six-coordinate centers (Figure 3). In
the reaction with catechol, isopropoxide groups attached to five-
coordinate Ti atoms are exchanged for bidentate catechols,
forming six-coordinate Ti atoms, in agreement with the change
in coordination upon binding of ascorbic acid observed in
XANES studies.13 In the reaction with INA, one oxygen of the
isonicotinate anion binds to the five-coordinate site, resulting
in sixfold coordination, and the remaining oxygen displaces an
isopropoxide group on a neighboring titanium. Following
functionalization, both Ti17 clusters still possess a four-
coordinate titanium core, but the 16 surrounding titanium atoms
now all exhibit sixfold coordination. Unlike the products
obtained in the synthesis of the nTi ) 6 clusters in the presence
of INA and catechol, variation of the ligand of the Ti17 cluster
preserves the TiO core, which is identical for the two adsorbents.
The diameter of the metal oxide core of the Ti17 cluster is
∼1.2 nm and increases to ∼1.5 nm when all of the isopropoxide
groups are included. The diameters of the clusters functionalized
with catechol and INA are 2.0 and 1.8 nm, respectively.
(10) Baur, W. H.; Khan, A. A. Acta Crystallogr. 1971, B27, 2133–2139.
(11) Horn, M.; Schwertfeger, C. F.; Meagher, E. P. Z. Kristallogr. 1972,
136, 273–281.
(13) Rajh, T.; Nedeljkovic, J. M.; Chen, L. X.; Poluektov, O.; Thurnauer,
(12) Baur, W. H. Acta Crystallogr. 1961, A14, 214–216.
M. C. J. Phys. Chem. B 1999, 103, 3515–3519.
9
2940 J. AM. CHEM. SOC. VOL. 132, NO. 9, 2010

Studies of Photosensitizer Dyes on Semiconductor Surfaces
A R T I C L E S
Figure 3. View parallel (upper row) and perpendicular (lower row) to the crystallographic or local S4 axis of the semiconductor core of the bare Ti17 cluster
(left) and the clusters functionalized with catechol (center) and INA (right). Isopropyl groups and hydrogen atoms have been omitted for clarity. Both ligands
always attach to five-coordinate titanium atoms of the bare complex, resulting in the increase to sixfold coordination of the titanium atom (highlighted in
green).
Figure 4. Different binding modes of the catecholate ion: (left) Ti2cat6 showing monodentate, bidentate, and µ2-(O,O′,O′) binding; (right) fragment of
Ti6cat6 showing µ3-(O,O′,O,O′) binding.
Geometry of Adsorption. Previous theoretical calculations
have shown that the manner in which a dye sensitizer binds to
the substrate significantly affects the electronic properties of
the system.14 The prototypical sensitizer in photovoltaic devices
is cis-dithiocyanatobis(2,2′-bipyridyl-4,4′-dicarboxylic acid)ru-
thenium(II) (N3).15 There is a general consensus that it binds
to the semiconductor substrate through the bis(isonicotinic acid)
group, but geometric details of this interaction are not available.
In the current work, we examined two adsorbents, catechol and
INA, the latter of which is part of the adhering group of N3.
Catechol. While 19 crystal structures containing catechol
bound to titanium are currently listed in the Cambridge
Structural Database (CSD),16 only four neutral titanium cat-
echolate structures have been reported.17,18 On the basis of a
survey of the CSD, Gigant et al.17 identified eight different
coordination modes of catechol, several of which occur in the
Ti/O clusters described here. In the Ti2cat2 structure, which is
a new polymorph of a previously reported structure, the two
catechol dianions bind in a bridging chelate µ2-(O,O′,O′) mode.
Ti2cat6 has the largest catechol-to-titanium ratio (3:1) reported
for a neutral titanium catecholate to date. The three crystallo-
graphically unique catechol ligands, one dianion and two anions,
illustrate the variety of ligand binding to TiO clusters. They
bind in a singly bridging chelate µ2-(O,O′,O′) mode, a mono-
dentate mode, and a bidentate mode, respectively, as illustrated
in Figure 4. The six unique catechol dianion ligands found in
Ti6cat6 are related by a pseudo-C2 axis, but only two unique
modes of binding to the metal centers are present: singly
bridging chelate µ2-(O,O′,O′) (two ligands) and doubly bridging
chelate µ3-(O,O′,O,O′) (four ligands).
Isonicotinic Acid. The reactions between INA and the titanium
alkoxides yield more a more consistent binding motif. In all
four structures reported herein, INA is deprotonated to give the
isonicotinate anion in the single-crystalline phase. It always
exhibits monodentate binding of each of the oxygens to a
titanium atom, so each ligand bridges two titanium atoms. While
no examples of titanium compounds containing INA or its anion
could be found in the CSD, four crystal structures of titanium
alkoxides containing the related benzoate anion are present. In
these structures, the benzoate anion, like the isonicotinate anion,
always bridges two titanium centers. It is worth noting that the
structure of Ti6(µ3-O)(µ2-benzoato-O,O′)-(2,2-dimethyl-PrO),19
despite possessing neopentoxide and benzoate instead of iso-
propoxide and isonicotinate ligands, has the same space group
(14) Redfern, P. C.; Zapol, P.; Curtiss, L. A.; Rajh, T.; Thurnauer, M. C.
J. Phys. Chem. B 2003, 107, 11419–11427.
(15) Gra¨tzel, M. Inorg. Chem. 2005, 44, 6841–6851.
(16) Allen, F. H. Acta Crystallogr. 2002, B58, 380–388.
(17) Gigant, K.; Rammal, A.; Henry, M. J. Am. Chem. Soc. 2001, 123,
11632–11637.
(18) Davidson, M. G.; Jones, M. D.; Lunn, M. D.; Mahon, M. F. Inorg.
Chem. 2006, 45, 2282–2287.
9
J. AM. CHEM. SOC. VOL. 132, NO. 9, 2010 2941

A R T I C L E S
Benedict and Coppens
Figure 5. Absorption spectra of Ti17O24OPri20 (Ti17) in CH2Cl2 at the
following concentrations (right to left): 1.3, 0.63, 0.31, 0.16, 7.8 × 10-2
,
3.9 × 10-2, 2.0 × 10-2, 9.8 × 10-3, and 4.9 × 10-3 g/L.
j
(R3) and molecular packing as Ti6INA6 with identical molecular
connectivity of the adsorbents.
Spectroscopic Measurement of the Band Gaps of the
Clusters. As described by Adachi,20 for optical transitions near
the absorption edge, the absorption coefficient R (in cm-1) is
given by the following expression:
Figure 6. Plots of R (in cm-1) vs photon energy (eV) and corresponding
fits to eq 1 for Ti17 in CH2Cl2 with loading densities of (top) 1.3 and (bottom)
4.9 × 10-3 g/L. The top panel shows data (O), the n ) 3 fit (solid line),
and the n ) 2 fit (dashed line); data (O) and the n ) 0.5 fit (solid line) are
illustrated in the bottom panel.
B(hν - Eg)n
The experimental and fitted absorption curves of Ti17 shown
in Figure 6 bear a striking resemblence to the curves of colloidal
anatase particles with diameters of 2.1, 13.3, and 26.7 nm, as
measured by Serpone et al.,21 which showed a band gap of Eg
) 4.03-4.04 eV. The band gap of the strongest allowed indirect
transition was found to be equal to 3.21 eV. The authors
concluded that no size quantization occurred for particles with
diameters greater than 2.0 nm. The allowed direct transition
across the band gap of ∼4.25 eV in Ti17 is blue-shifted by ∼0.22
eV relative to that of the anatase nanoparticles.
R )
(1)
hν
where B is the absorption constant for the transition, Eg (in eV)
is the energy of the band gap, and hν (in eV) is the photon
energy. The exponent n characterizes the nature of the transition.
For allowed direct and indirect transitions (the latter of which
also involve momentum transfer), n ) 0.5 and 2, respectively.
For forbidden direct and indirect transitions, n ) 1.5 and 3,
respectively.
Analysis of the absorption coefficient at the onset of absorp-
tion for different loading densities has been used to determine
the band gap of colloidal anatase nanoparticles,21 but the analysis
was limited to allowed direct or indirect transitions. Absorption
spectra of Ti17 in methylene chloride at concentrations from
4.9 × 10-3 to 1.3 g/L are shown in Figure 5. An increase in
baseline absorbance at loading densities greater than 1.3 g/L
(not shown in Figure 5) is attributed to scattering from
aggregates in solution. At concentrations up to 1.3 g/L, the
absorption changes exhibited Beer-Lambert behavior at wave-
lengths for which the absorbance was below 1.5, indicating that
aggregation did not occur at these loading densities.
While our data suggest that the low-energy indirect transition
is forbidden rather than allowed, analysis with either function
leads to an indirect band gap energy that is ∼0.17 eV larger
than the anatase value of Serpone et al.21
Theoretical Calculations. Extensive calculations on optimized
TiO2 nanoparticles have been reported by Persson and
co-workers.5,22-24 Density functional theory (DFT) calculations
on the experimental structures of the nanoclusters were per-
formed with the Gaussian03 package25 using the WebMO
interface. GaussView26 was used for visualization and further
analysis of the computational results. The hybrid exchange
correlation functional B3LYP was used for all of the calculations
in the present study. The crystallographic symmetries of the
complexes were maintained in all of the calculations. Because
of the disorder of the isopropoxide groups in nearly every
structure examined, geometry optimizations were performed
starting with the crystal structure geometry after removal of a
At the lowest concentration examined (4.9 × 10-3 g/L), the
absorption coefficient between 4.4 and 5.0 eV follows a square-
root dependence (n ) 0.5), indicating that this is an allowed
direct transition corresponding to a band gap of 4.26 eV (R2 )
1.00). At high loading densities (1.3 g/L), the absorption
coefficient at the onset of absorption is best fit by n ) 3,
suggesting that the lowest-energy transition is a forbidden
indirect transition with a band gap of 3.36 eV (R2 ) 0.99).
Fitting the high loading density data using n ) 2 yielded a nearly
identical band gap (3.40 eV), but R2 was reduced to 0.96.
(21) Serpone, N.; Lawless, D.; Khairutdinov, R. J. Phys. Chem. 1995, 99,
16646–16654.
(22) Nilsing, M.; Lunell, S.; Persson, P.; Ojama¨e, L. Surf. Sci. 2005, 582,
49–60.
(23) Persson, P.; Lundqvist, M. J.; Ernstorfer, R.; Goddard, W. A.; Willig,
F. J. Chem. Theory Comput. 2006, 2, 441–451.
(24) Persson, P.; Gebhardt, J. C. M.; Lunell, S. J. Phys. Chem. B 2003,
107, 3336–3339.
(19) Ammala, P. S.; Batten, S. R.; Kepert, C. M.; Spiccia, L.; van den
Bergen, A. M.; West, B. O. Inorg. Chim. Acta 2003, 353, 75; CCDC
entry 187335.
(25) Frisch, M. J.; et al. Gaussian 03, revision C.02; Gaussian, Inc.:
Wallingford, CT, 2004.
(20) Adachi, S. Optical Properties of Crystalline and Amorphous Semi-
conductors: Materials and Fundamental Principles; Kluwer: Norwell,
MA, 1999; p 280.
(26) Moser, J.; Punchihewa, S.; Infelta, P. P.; Gra¨tzel, M. Langmuir 1991,
7, 3012–3018.
9
2942 J. AM. CHEM. SOC. VOL. 132, NO. 9, 2010

Studies of Photosensitizer Dyes on Semiconductor Surfaces
A R T I C L E S
Table 1. Calculated HOMO-LUMO Gaps ∆E (eV) of Ligands and Titanium Clusters
catechol
INA
Ti2cat2
Ti2cat6
Ti6cat6
Ti3INA3
Ti4INA2
Ti6INA6
Ti17INA4
Ti17
symmetry
ELUMO
EHOMO
∆E
C1
C1
C1
Ci
C1
C1
C2
Ci
S4
C1
-0.29
-6.02
5.73
-2.48
-7.25
4.77
-1.51
-5.29
3.78
-1.90
-5.41
3.51
-1.82
-5.23
3.41
-2.30
-6.93
4.62
-1.99
-6.71
4.72
-2.74
-7.13
4.39
-2.41
-7.02
4.61
-1.93
-6.64
4.71
minor disordered component. In the optimized structures, the
crystal structure geometry was essentially preserved. The
inability to reliably determine the positions of the hydrogen
atoms by geometry optimization of the Ti17cat4 structure
prevented its inclusion in the calculations. Final calculations
for all of the complexes were completed using the LANL2DZ
double-ꢀ basis set with an ultrafine integration grid. The
normalized sum of the squares of the orbital coefficients was
used for partitioning the molecular orbitals between the cluster
and adsorbates.5
Band Structure. The HOMO and LUMO of unsubstituted
Ti17 have energies of -6.64 and -1.93 eV, respectively (last
column of Table 1), corresponding to a HOMO-LUMO gap
(∆E ) ELUMO - EHOMO) of 4.71 eV. B3LYP/VDZ calculations
by Lundqvist et al.5 yielded the slightly larger value of 4.98
eV for an nTi ) 16 TiO2 cluster optimized starting from the
structure of anatase. In clusters with adsorbed INA, the
HOMO-LUMO gap is only slightly smaller than in Ti17 and
varies between 4.4 and 4.7 eV. There is no correlation between
cluster size and the calculated band gaps in the limited number
of clusters examined in the current study.
Figure 7. DOS plots for Ti17 and smaller clusters containing catechol,
illustrating the band gap and evolution of quasi-valence and -conduction
bands with increasing cluster size.
The energy of the HOMO-LUMO gap in the titanium
clusters with adsorbed catechol (3.41-3.78 eV) is nearly 1 eV
smaller than those in Ti17 and the titanium complexes containing
INA. In the Ti/catechol complexes, the HOMO, HOMO-1, and
HOMO-2 orbitals are localized almost exclusively on the
catechol ligand (85-98%), whereas the LUMO, LUMO+1, and
LUMO+2 orbitals consist largely of titanium d orbitals
(68-72%), indicating that the HOMO-to-LUMO and other low-
lying transitions involve catechol-to-cluster charge transfer, in
agreement with earlier theoretical calculations of catechol
adsorbed on TiO2.14 This smaller HOMO-LUMO gap in the
Ti/catechol compounds is responsible for their deep-red color.
Such coloration is characteristic of catechol adsorbed on anatase
nanoparticles, as first reported by Moser and Gra¨tzel26 and
subsequently by numerous others.13,27,28 The dark-red color
observed in all four Ti/catechol complexes presented herein
suggests that the number of titanium atoms, or cluster size, has
little impact on the size of ∆E, in agreement with the results of
our calculations.
On the other hand, the character of the lowest-energy
electronic transition in the Ti/INA complexes varies from cluster
to cluster. In the nTi ) 3 and 6 clusters with INA, the HOMOs
consist of 95 and 98% INA orbitals, respectively. For the
Ti3INA3 cluster, the LUMO is 50% INA- and 41% cluster-
centered, so the HOMO-LUMO transition for this compound
is a mixture of ligand-to-ligand and ligand-to-cluster charge
transfer. The transition is similar for the Ti6INA6 cluster, as
the LUMO is 64% INA- and 33% cluster-centered. In the
Ti4INA2 complex, the HOMO is 22% INA- and 30% cluster-
centered, with the remainder of the orbital residing on the
isopropoxide groups; the LUMO is 72% INA- and 22% cluster-
Figure 8. DOS plots of Ti17 and smaller clusters containing INA,
illustrating the band gap and evolution of quasi-valence and -conduction
bands with increasing cluster size.
centered. INA contributions are almost completely absent in
the Ti17INA4 HOMO. They account for only 1% of the orbital,
whereas 61% of the HOMO is centered on the Ti/O core, with
the remainder composed of isopropoxide basis functions. The
LUMO is split almost evenly between INA (46%) and the core
(52%).
While the cluster size seems to have little or no effect on the
HOMO-LUMO gap of the titanium cluster, an increase in the
size of the cluster is accompanied by the formation of quasi-
continuous conduction and valence bands. Density of states
(DOS) plots for Ti17 (Figures 7 and 8) reveal the similarity of
the pseudoconduction and -valence bands to those of similarly
sized anatase particles.5
The band gap of ∼4.7 eV calculated for nonfunctionalized
Ti17 is larger than our spectroscopic result of ∼4.25 eV. It is
well-known that the HOMO-LUMO gap obtained using DFT
methods exceeds values calculated using time-dependent DFT.5
(27) Wang, Y.; Hang, K.; Anderson, N. A.; Lian, T. J. Phys. Chem. B
2003, 107, 9434–9440.
(28) Creutz, C.; Chou, M. H. Inorg. Chem. 2008, 47, 3509–3514.
9
J. AM. CHEM. SOC. VOL. 132, NO. 9, 2010 2943

A R T I C L E S
Benedict and Coppens
Figure 9. Structure of the Ti17INA4 complex as optimized using the B3LYP/LANL2DZ method. Independent atoms of the Ti/O core shown in the bond
length (upper-right, in Å) and Mayer bond order (lower-right) diagrams are labeled and colored green in the diagram at the left. In the diagrams on the right,
the green coloring of the selected atoms has been removed, and the Ti and O atoms are colored silver and red, respectively.
(Figure 4). Unlike earlier studies,29 the calculations showed no
evidence for a significant enhancement of the dipole moment
Table 2. Net Charges (e) on Free Catecholate and the Adsorbed
Catecholate in Ti6cat6
bonding
net charge
charges on oxygen atoms
upon adsorption.
catechol moleculea
µ2-(O,O′,O′)
-1.67
-1.67
-1.72
-1.58
-1.58
-1.78
-1.78
-0.52, -0.47
-0.62, -0.70
-0.63, -0.72
-0.63, -0.55
-0.54, -0.63
-0.69, -0.70
-0.70, -0.71
Conclusions
While research involving wide-band-gap semiconductor
particles has attracted considerable attention, the lack of atomic-
resolution structures of the nanoparticles or adsorbed dye
sensitizers introduces ambiguities into the structure/function
relationships. The crystalline phase of nanoparticles function-
alized with technologically relevant ligands offers the op-
portunity to obtain unambiguous structural information. The
particles can be synthesized and subsequently self-assembled
into single crystals suitable for structure determination, spec-
troscopic characterization, and theoretical analysis. The UV-vis
spectra of the Ti17 nanoparticles presented herein are remarkably
similar to those of semiconductor nanoparticles currently
employed in photovoltaic research. The band gaps of the Ti17
cluster derived from the spectra and the theoretical calculations
agree within a few tens of electron volts with that measured
for anatase. The Ti-O bond lengths in the clusters show a larger
range than those in the known TiO2 phases, indicating the
importance of surface effects. The catecholate adsorbent exhibits
four different modes of binding to the clusters. The effect of
such differences on physical properties merits further attention.
µ3-(O,O′,O,O′)
a Hydroxyl hydrogens were omitted in the summation of the atomic
charges.
The DOS of clusters containing catechol do not display
prominent conduction and valence bands and are instead
primarily composed of discrete energy levels. This is not
surprising, as the levels between -7 and -5 eV are orbitals of
catechol ligands, which are energetically isolated from the
remaining cluster orbitals. In the INA clusters, even the smallest,
Ti3INA3, exhibits a modest conduction band with a significant
density of states over a wide energy range. The DOS increases
as the number of titanium atoms gets larger. Ti6INA6 is the
smallest cluster for which a valence-band-type DOS is observed.
It grows into a well-formed band in the case of the Ti17INA4
cluster.
Nature of the Bonding in the Clusters. Analysis of the Mayer
bond orders in the clusters shows a significant variation in Ti-O
bonding, in agreement with the considerable spread in the Ti-O
bond lengths discussed above. The results for Ti17INA4 are
shown in Figure 9. The theoretical and experimental bond
lengths agree within ∼0.05 Å (Table S2 in the Supporting
Information). Several of the longer bond lengths are overesti-
mated by the calculations. Weak covalency is also found for
the Ti-Ti interactions. The covalent contributions to the
cluster’s stability are clearly considerable. The calculations do
not show significant charge transfer from the catechol into the
cluster in the ground state, as shown for Ti6cat6 in Table 2.
The negative charges on the catecholate adsorbent are similar
to those in the isolated cation but larger for the µ3-(O,O′,O,O′)
bonding mode. The oxygen atoms are more negative in this
bonding mode, in which each oxygen is linked to two Ti atoms
Acknowledgment. This work was funded by the Division of
Chemical Sciences, Geosciences, and Biosciences, Office of Basic
Energy Sciences, of the U.S. Department of Energy through Grant
DE-FG02-02ER15372. Technical assistance by Renata Freindorf
is greatly appreciated.
Supporting Information Available: Complete ref 25, Tables
S1 and S2, and crystallographic data for the clusters (CIF). This
material is available free of charge via the Internet at http://
pubs.acs.org.
JA909600W
(29) Rajh, T.; Chen, L. X.; Lukas, K.; Liu, T.; Thurnauer, M. C.; Tiede,
D. M. J. Phys. Chem. B 2002, 106, 10543–10552.
9
2944 J. AM. CHEM. SOC. VOL. 132, NO. 9, 2010