Rationalization of dye uptake on titania slides for dye-sensitized solar cells by a combined chemometric and structural approach

Article abstract of DOI:10.1002/cssc.201402194

A model photosensitizer (D5) for application in dye-sensitized solar cells has been studied by a combination of XRD, theoretical calculations, and spectroscopic/chemometric methods. The conformational stability and flexibility of D5 and molecular interactions between adjacent molecules were characterized to obtain the driving forces that govern D5 uptake and grafting and to infer the most likely arrangement of the molecules on the surface of TiO₂. A spectroscopic/chemometric approach was then used to yield information about the correlations between three variables that govern the uptake itself: D5 concentration, dispersant (chenodeoxycholic acid; CDCA) concentration, and contact time. The obtained regression model shows that large uptakes can be obtained at high D5 concentrations in the presence of CDCA with a long contact time, or in absence of CDCA if the contact time is short, which suggests how dye uptake and photovoltaic device preparation can be optimized.

Full text of DOI:10.1002/cssc.201402194

CHEMSUSCHEM
FULL PAPERS
DOI: 10.1002/cssc.201402194
Rationalization of Dye Uptake on Titania Slides for Dye-
Sensitized Solar Cells by a Combined Chemometric and
Structural Approach
Valentina Gianotti,[a] Giada Favaro,[a] Luca Bonandini,[b, c] Luca Palin,[a] Gianluca Croce,[a]
Enrico Boccaleri,[a] Emma Artuso,[b] Wouter van Beek,[d] Claudia Barolo,*[b] and
Marco Milanesio*[a]
This work is dedicated to Professor Davide Viterbo on the occasion of his 75th birthday.
A model photosensitizer (D5) for application in dye-sensitized
solar cells has been studied by a combination of XRD, theoreti-
cal calculations, and spectroscopic/chemometric methods. The
conformational stability and flexibility of D5 and molecular in-
teractions between adjacent molecules were characterized to
obtain the driving forces that govern D5 uptake and grafting
and to infer the most likely arrangement of the molecules on
the surface of TiO2. A spectroscopic/chemometric approach
was then used to yield information about the correlations be-
tween three variables that govern the uptake itself: D5 con-
centration, dispersant (chenodeoxycholic acid; CDCA) concen-
tration, and contact time. The obtained regression model
shows that large uptakes can be obtained at high D5 concen-
trations in the presence of CDCA with a long contact time, or
in absence of CDCA if the contact time is short, which suggests
how dye uptake and photovoltaic device preparation can be
optimized.
Introduction
Photovoltaic (PV) cells based on organic semiconductors and/
or organic light harvesters are potentially extremely inexpen-
sive but their efficiency and stability are still limited compared
to inorganic crystalline solar cells. Among them, dye-sensitized
solar cells (DSC) represent a promising and emerging technolo-
gy.[1] These cells mimic the energy conversion mechanism of
photosynthesis as light is absorbed by an antenna compound
(chlorophyll in photosynthesis, a dye in DSC), then an excited
electron is produced and captured by a complex system
(photo systems I and II in photosynthesis and nanostructured
titanium oxide, tin oxide, and an inorganic electrolyte in DSC),
which exploits the energy to obtain valuable products (i.e.,
chemical energy in the form of sugar in photosynthesis and
electric current in DSC).
The chemical properties of the cell components must be de-
signed and tuned carefully to optimize the yield of PV cells.
Presently, the main issues that still limit their technological ap-
plications[2] are: 1) the achievement of a reasonable conversion
efficiency of the DSC modules (10% for opaque, 5–6% for
transparent),[3] 2) maintaining these yields over the time
needed for a cell working in real conditions (~20 years), and
3) the achievement of reproducible results (ꢀ3–5% differences
between modules). To fulfill these objectives, a detailed molec-
ular-level knowledge of the DSC components is of paramount
importance.
[a] Dr. V. Gianotti, G. Favaro, Dr. L. Palin, Dr. G. Croce, Dr. E. Boccaleri,
Dr. M. Milanesio
Although a great deal of research has been performed to
design more efficient photosensitizers,[4] only recently have ef-
forts been made to understand,[5,6] model,[7–10] and control[11]
the interactions that play a major role in dye uptake. Moreover,
the dye loading amount can be tuned by changing the bath
solvent,[5] which has an important effect on the cell efficiency.
Literature data highlight the importance of chenodeoxycholic
acid (CDCA) as a coadsorbent to control dye aggregation and
electron injection[12] and to improve the performance.[13] How-
ever, a rationalization of the effect of chemical parameters that
affect dye uptake, in relation with chemical forces that govern
molecular interactions is still lacking. A polyene-diphenylaniline
dye (D5) proposed by Hagberg et al.,[14] has been used in
a case study for the rationalization of the uptake conditions in
metal-free dyes. This simple molecule[7,15] can be considered as
a model for the widespread class of donor–p–acceptor (d-p-A)
Dipartimento di Scienze e Innovazione Tecnologica
Universitꢀ del Piemonte Orientale “A. Avogadro”
Viale T. Michel 11, 15121 Alessandria (Italy)
E-mail: marco.milanesio@unipmn.it
[b] L. Bonandini, Dr. E. Artuso, Dr. C. Barolo
Dipartimento di Chimica, NIS Interdepartmental Centre
Universitꢀ di Torino
Via Pietro Giuria 7, 10125 Turin (Italy)
E-mail: claudia.barolo@unito.it
[c] L. Bonandini
DYEPOWER
Viale Castro Pretorio 122, 00185 Rome (Italy)
[d] Dr. W. van Beek
Swiss-Norwegian Beamlines at ESRF
BP 220, Grenoble 38043 (France)
Supporting Information for this article is available on the WWW under
http://dx.doi.org/10.1002/cssc.201402194.
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2014, 7, 1 – 15 1
ÞÞ
These are not the final page numbers!

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
dyes. To date, this class holds the efficiency record for metal-
free dyes.[16] In the last decade, various organic functional
groups have been combined to generate d-p-A structures.
One of the schemes employed most commonly is: an aryl
amine group as electron donor, a thiophene unit as the p
linker, and a cyanoacrylic acid moiety as the electron acceptor/
anchoring group (which are all present in D5).
tion X-ray source. We report on the crystal and molecular
structure of D5 and two related compounds (4 and 6;
Scheme 1) by powder and single-crystal XRD. The analysis of
their packing features allowed us to understand the molecular
forces that govern their intra- and intermolecular interactions.
However, crystallographic studies cannot give direct informa-
tion on the behavior of D5 on TiO2. To shed light on the corre-
lations between the main parameters that govern dye uptake,
a chemometric-driven UV/Vis spectroscopic study was de-
signed and performed. UV/Vis spectroscopy was used recently
by Dell’Orto et al. to assess the kinetics of the absorption of
the N719 dye onto TiO2.[11] We chose to exploit a quantitative
chemometric approach because it allows the maximization of
the information content in the least number of experi-
ments.[25–27] To date, the optimization of the experimental con-
ditions of dye uptake has been performed mainly by trial and
error or at best by “one variable at a time” (OVAT) methods.
Only very recently was the chemometric approach proposed in
the field of DSC by some of us.[28] The present work aims to in-
vestigate both the molecular structure and the dye uptake in
a synergic way and represents the first part of a larger project
that we are undertaking with the purpose to understand, at
the molecular level, the mechanisms involved in DSC function
with the ultimate aim to improve their yields and stability by
optimizing the preparation methods of the cell itself.
Structural and crystallographic studies of organic com-
pounds and molecular complexes allow the assessment of the
possible intra- and intermolecular interactions, which are of
paramount importance for the functionality of the materials
under working conditions.[17–19] Very few structural studies are
found in the field of DSC because of the complications of dye
crystallography, which are mainly because of the difficulty to
obtain suitable single crystals. Ru-based dye compounds[20] are
less difficult to crystallize and represent the majority of the X-
ray crystal structures related to DSC, whereas only a few crystal
structures of compounds related to d-p-A sensitizers are avail-
able in the Cambridge Crystallographic Data Centre (CCDC).[21]
Relevant structures in the database are: 1) a molecule that con-
tains diphenylaminophenyl and carboxylic moieties, that is, D5
without the vinyl thiophene linker;[22] and 2) two molecules
that contain the diphenylaminophenyl moiety and a thiophene
linker.[23] The electronic and molecular surface structure of the
functional dye-sensitized interface has also been studied in
detail for D5 by a combination of core-level spectroscopy, va-
lence-level spectroscopy, X-ray absorption spectroscopy, and
resonant photoemission spectroscopy.[24]
Results and Discussion
In this study, we intend to shed light on the dye-uptake pro-
cess by combining design of experiment (DoE) assisted UV/Vis
spectroscopy with a structural investigation. Our aim is to un-
derstand the mechanism of dye dispersion and bonding to the
TiO2 surface with the long-term goal to understand their influ-
ence on the cell macroscopic behavior. In the present study,
dye-related crystallography problems have been overcome by
exploiting advanced powder XRD methods that use different
high-resolution detectors with a high-flux synchrotron-radia-
D5 synthesis
The synthesis of 4 and D5 was performed with a slight modifi-
cation of literature procedures[14,29] by starting from commer-
cial 4-(N,N-diphenylamino)benzaldehyde (1; Scheme 1). The
first step of our synthetic pathway is a simple Wittig reaction[30]
to obtain alkene 2, which was used as a substrate for a subse-
quent Pd-catalyzed decarbonylative Heck reaction[31] to give in-
termediate 3. Subsequently, lithiation of 3 with n-butyllithium
Scheme 1. Synthesis of D5 dye. Reagents and conditions: i) Ph3PCH3Br, tBuOK, THF, RT, 8 h (76%); ii) Thiophene-2-carboxylic acid, PdCl2, LiCl, di-tert-butyl dicar-
bonate, g-picoline, N-methylpyrrolidone, 1208C, 16 h (70%); iii) DMF, BuLi, THF, ꢁ788C to RT (39%); iv) Cyanoacrylic acid, piperidine, (72%); v) MeOH, H2SO4,
reflux, 24 h; vi) Cyanoacrylic methyl ester, piperidine, ACN.
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 15
&
2
&
ÞÞ
These are not the final page numbers!

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
followed by the addition of DMF yielded the corresponding al-
dehyde 4. The electron-withdrawing group is inserted into the
structure by a Knoevenagel reaction between aldehyde 4 and
cyanoacetic acid in the presence of piperidine.
First-principles calculations on stable minima
In Ref. [9] only one isomer of D5 (named D5-2a in Table 1) is
accepted and used generally.[33] As we lack a single-crystal
structure of D5 and as it is almost impossible to investigate
the structure of D5 on TiO2 directly and precisely in solution
during the grafting process, accurate first-principles theoretical
calculations, combined with an experimental structural study
of D5 and its parent compounds, have been performed and re-
ported in detail in a separate paper, together with all the strat-
egies and tricks used for structure solution.[34]
D5 was then converted into its corresponding methyl ester
(6) to verify the configuration of the 2-cyano-3-(thiophen-2-yl)
acrylic acid moiety. Compound 6 was also obtained, in the
same configuration, through the classical Knoevenagel reaction
directly from 4.
In this paper, the possible conformational changes were in-
vestigated by considering the three degrees of freedom (f1,
f2, and f3; Figure 1), which can assume either the s-cis or the
s-trans conformations as E/Z isomerization is already estab-
lished. The conformational changes around f4, f5, and f6 are
less important because of the symmetry of the phenyl groups.
However, for an exhaustive search they were also considered,
but only the more stable conformations of D5 and related
compounds are reported (Table S2).
Computational study of D5 conformational flexibility and
freedom
The chemical formula of D5 (Figure 1) suggests that this mole-
cule should be rather rigid and planar because of the conjuga-
tion between aromatic moieties (thiophene and benzene
Among the possible theoretical conformers, four of them
(namely, D5-1a, D5-1b, D5-2a, and D5-2b) possess stable
energy minima below 1.5 kcalmolꢁ1. According to the Boltz-
mann distribution, they are most probably the dominant ones
for D5 in solution and on the TiO2 surface (i.e., in the relevant
cases).
The geometries of the four isomers, after geometric optimi-
zation by first-principles calculations at the B3LYP/6-31G(d,p)
and B3LYP/6-311+G(2d,2p) levels of theory, are depicted in
Table 1. The 6-31G(d,p) basis set was used to obtain a first fast
geometry optimization and screening of possible stable con-
formations, whereas the 6-311+G(2d,2p) basis set was manda-
tory to obtain a careful description of the molecular geome-
tries of the conformers and of their relative stabilities. Confor-
mer D5-1a is the most stable and also the most prevalent in X-
ray crystal structures (see below) and thus can be considered
the prevalent structure at equilibrium. The structure reported
most commonly (D5-2a)[14,16] and the other two conformers
are less stable by less than 0.5 kcalmolꢁ1 and can be present
at lower concentrations in solution, according to the Boltz-
mann distribution, and at nonequilibrium conditions. Notably,
if D5 first approaches and is then linked to the TiO2 surface,
the carboxyl group is deprotonated and the conformational
degree of freedom around f1 becomes irrelevant because the
COOꢁ moiety is symmetric for a 1808 rotation. Moreover, in de-
protonated D5 the energy differences are even smaller
(Table 2).
Figure 1. Degrees of freedom of D5 molecule: f1, f2, and f3 refer to the O=
CꢁC1=C2, C1=C2ꢁC3=C4, and C5=C6ꢁC7=C8 torsion angles, respectively.
rings) through a C=C bond. In addition, the cyanoacetic group
is planar and connected to the thiophene by a double bond.
The only nonrigid part is the 3-phenylamino moiety, which is
nonplanar and the terminal phenyl groups are free to rotate
and adopt different conformations. The crystal structures of 4
and 6 from single-crystal data gave a clear picture of the ste-
reochemistry around the C=C bond isomerization and con-
firmed the expected E isomer (see below). In addition, NMR
spectroscopy and chromatography experiments (Section S1 in
the Supporting Information) confirmed that only one isomer is
present in solution. Conversely, a rather rich conformational
variability can arise because of the rotation around the single
bonds, as discussed below. In principle, a reliable indication of
the stable conformation of D5 in the solid state could be
gained from single-crystal XRD data, but the same indication
about the dye in solution or bonded to the TiO2 surface can
only be obtained by a combination of experimental XRD data
and computational analysis by first-principles calculations. As
we failed in all of our attempts to grow single crystals of D5
because of the well-known difficulty to crystallize bulky carbox-
ylic acids, in analogy to that observed for fatty acids,[32] high-
resolution X-ray powder diffraction (XRPD) was used. The limit-
ed resolution of XRPD of organic molecules rendered the dis-
crimination between the isomers rather difficult and, therefore,
an accurate conformational analysis was needed.
Therefore, the theoretical calculations suggest that conform-
ers D5-1a, D5-1b, D5-2a, and D5-2b are the most probable.
Three out of these four conformations were observed experi-
mentally in the X-ray crystal structures (see discussion below),
in which crystal packing forces play a relevant role in the selec-
tion of less stable conformers. The compromise between the
intrinsic thermodynamic stability of the isolated conformers
and the effect of intermolecular interactions in the solid-state
is well known and is observed if theoretical calculations are
compared with X-ray structures.[35] Of course, these conclusions
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 15
&
3
&
ÞÞ
These are not the final page numbers!

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
Energy barriers between energy
minima as a function of torsion
angles
Table 1. Geometric features of the most stable conformers (within 1.5 kcalmolꢁ1) after geometric optimization
using B3LYP functional.
Conformer
f1
[8]
f2
[8]
f3
[8]
Relative stability [kcalmolꢁ1
]
Molecular
structure
6-31G(d,p)
6-311+G(2d,2p)
The rotation barriers between
the four conformers were ex-
plored by relaxed potential
energy scans (R-PES) around the
two SꢁCꢁC=C torsion angles (f2
and f3). To explore this energy
surface with an acceptable com-
puting time, initially two 1D R-
PES scans were performed at the
same DFT level (B3LYP/6-
31G(d,p)), and the data are re-
ported in Figure 2a. Then a 2D
R-PES at the less-demanding HF/
3-21G level of calculation was
performed to explore the two
torsions, which produced the 3D
plot shown in Figure 2b. The
basis sets used for geometry op-
timizations would be too time-
consuming and unaffordable for
such an extended PES. However
HF-3-21G still gave acceptable
geometries and energy differen-
ces if the energy minima are
compared to the results of the
more extended basis sets.
D5-1a
D5-1b
0.1
180
180
0.00
0.00
0.73
0.0
0.0
0.0
180.
ꢁ1.1
180
0.95
0.38
1.29
D5-2a
0.1
0.34
0.98
D5-2b
0.1
ꢁ1.8
4
—
ꢁ178
ꢁ177
—
ꢁ179
4.7
D5
ꢁ9.3
0.6
ꢁ173
ꢁ168
179
ꢁ169
The observed minima con-
firmed that the s-cis conforma-
tion of D5-1a is favored and that
the other three conformers (D5-
1b, D5-2a, D5-2b) are the
unique stable minima, in agree-
ment with the first-principles
DFT calculations. The R-PES indi-
cated that rotation around f2 is
easier than that around f3.
Moreover there are no other
minima and there are no sterical-
ly forbidden regions that hinder
the rotation. The heights of the
barrier (ꢂ4 and 10 kcalmolꢁ1
around f2 and f3) suggest that
during D5 manipulation, for
both the DoE-assisted uptake ex-
periment used here and, in gen-
eral, for DSC cell preparation, ro-
tation around this single bond is
possible, which is also indicated
by the fact that 4, D5, and 6
show different conformations in
ꢁ175
6
ꢁ0.9
166
Table 2. Relative stabilities after geometric optimization of the more stable conformers for the four models
employed in the theoretical calculations.
Model
Relative stabilities [kcalmolꢁ1
deprotonated D5ꢁ
]
4
6
6-31G(d,p)
6-311+G(2d,2p)
6-31G(d,p)
6-311+G(2d,2p)
6-31G(d,p)
6-311+G(2d,2p)
D5-1a
D5-1b
D5-2a
D5-2b
0.00
0.98
1.55
2.46
0.00
0.75
1.41
2.05
0.00
1.24
ꢁ0.59
0.79
0.00
1.11
ꢁ0.51
0.66
0.00
0.97
0.36
1.28
0.00
0.77
0.31
0.99
do not take into account the energy barriers for rotation
around the CꢁC bonds, the investigation of which is discussed
in the next paragraph.
their crystal structures. This conformational flexibility is proba-
bly important in driving the D5 uptake on TiO2 as well as the
final arrangement of D5 molecules on its surface. The absolute
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 15
&
4
&
ÞÞ
These are not the final page numbers!

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
Figure 2. a) Energy profiles for rotation around f2 (black) and f3 (red) from B3LYP/6-31G(d,p) calculations, all geometries were fully optimized except for the
imposed values of the f2 and f3 torsion angles; b) 3D plot of energies obtained from the 2D R-PES.
minimum of the calculations (conformer 1) shows s-trans con-
formation for both f2 and f3 torsion angles, whereas the con-
formation usually considered in the literature (conformer 2)
shows s-cis and s-trans conformations for f2 and f3, respective-
ly. Both conformations are very close in energy and, therefore,
accessible at room temperature. The higher stability of the s-
cis conformation suggested by first-principles calculations was
confirmed by looking at the distribution of s-cis and s-trans
conformations in structures containing the vinyl thiophenic
moiety in the CCDC database.[21] This search (Figure S4) con-
firmed the prevalence of the s-trans conformation.
problem, on one hand, high-resolution XRPD of excellent spa-
tial and reciprocal space resolution were collected by using
a Pilatus 2M detector and, on the other hand, data from theo-
retical calculations and a priori information from the single-
crystal structures of 4 and 6 were exploited. The crystallization
of these two parent compounds was straightforward, and their
single-crystal structures could give direct and accurate indica-
tions about the more stable conformation in the solid state.
Notably, both structures 4 and 6 showed that the most stable
conformations coincide with those of first-principles calcula-
tions. This information helped the interpretation of the XRPD
from D5. Initially, the single-crystal structure of 4 was solved to
determine the torsion angle f3 and the common geometric
features of these compounds, that is, the planarity of the thio-
phene group and the geometry of the diphenylamino chromo-
phore. Then 6 (a crystalline derivative of D5) was prepared to
obtain experimental data from single-crystal XRD data to shed
light on the conformational features of the 2-cyanoacrylic
moiety, that is, on the conformation around the f1 and f2 tor-
sion angles. In the following sections the relevant features of
the three structures are discussed, and crystal data are report-
ed in the Supporting Information.
Crystal structures of D5, its precursor 4, and its methyl ester
6
The conformational flexibility suggested by first-principles cal-
culations was confirmed by experimental crystallographic data
of 4 and 6. Despite many different attempts, it was not possi-
ble to grow a single crystal of D5 and, as explained before, its
structure was investigated by high-resolution synchrotron-radi-
ation XRPD.[36] The XRPD study was performed under ambient
conditions as an experiment at 100 K (at which better data
could be collected in principle) a new phase appeared with
the formation of a mixture that was impossible to index. For
consistency, the data of compounds 4 and 6 were also mea-
sured at room temperature. As a result of the well-known limi-
tations of the accuracy of the structures solved by XRPD data,
the identification of the correct conformations around the f1,
f2, and f3 torsion angles without a priori information is a diffi-
cult or even impossible task. The conformations depicted in
Figure 1 have very small electron-density differences. Even the
high-quality data recorded with the 1D analyzer detector
(BM1B) and the 2D MAR Image Plate (BM1A) at the Swiss-Nor-
wegian Beamline (SNBL) were not sufficient for a successful
structure solution as detailed elsewhere.[34] To overcome the
Single-crystal structures of 4 and 6
¯
Compound 4 crystallizes in the P1 space group, and the asym-
metric unit contains two molecules arranged in parallel along
their elongation axis in which one is rotated by ꢂ908 with re-
spect to the other to form T-shaped interactions between the
aromatic conjugated moieties (Figure 3). The two molecules
show two different conformations (s-cis, s-trans) for the f3 tor-
sion angle (Table 1), which confirms the possibility that there is
more than one stable conformation as suggested by theoreti-
cal calculations. Conversely, f2 shows an s-trans conformation
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 15
&
5
&
ÞÞ
These are not the final page numbers!

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
Structure of D5 from XRPD data
The unit cell of D5 can contain four molecules and, given the
¯
P1 symmetry with only the inversion center as the symmetry
element, two molecules must be present in the asymmetric
unit. To solve the structure without biasing the search and to
explore the structure solution hypersurface under the best
conditions, all four stable isomers D5-1a, D5-1b, D5-2a, and
D5-2b were used as starting points for the real space structure
solution of D5, and two molecules with different conforma-
tions were also combined as observed for 4. If the simulated
annealing searches are extended sufficiently, in terms of tem-
perature and time, the results converged to a solution with f2
close to 08, and acceptable solutions were obtained with f3
close to both 08 and 1808. In other words, two possible correct
structure solutions are suggested by simulated annealing. The
first has two molecules with the conformation D5-1a and small
differences in the planarity of the vinyl thiophene moiety and
in the arrangement of the three phenylamino groups, and the
second has two different conformations, D5-1a and D5-1b, as
observed for 4. Conformations D5-2a and D5-2b do not
appear in any possible stable solutions among the R values
ranked highly. The best fit of the XRPD data (Figure S3) was
obtained for the first arrangement with two D5-1a molecules
(the absolute minimum of the first-principles calculations) in
the asymmetric unit (Table 1).
Figure 3. Crystal packing of a) 4 and b) 6 that highlight the hydrophobic
clustering of the phenyl groups in both cases.
in both molecules. A short S···O intramolecular contact (dmean
3.04(8) ꢁ) is observed.
=
The packing driving forces are head-to-tail H bonds between
the COOH moieties of adjacent D5 molecules (Figure 4). Hydro-
With regard to the triphenylamine group, the N atom is very
close to sp2 hybridization as the three CꢁNꢁC angles are be-
tween 118 and 1228 and the torsion angle that defines the pi-
ramidality of N (i.e., that obtained by the N atom and the
three C atoms bonded to the N atom) is very close to 08, as ex-
pected for a perfect sp2 hybridization. The two terminal phenyl
groups are not coplanar to minimize their reciprocal steric hin-
drance. The remaining part of the molecule is planar with devi-
ations smaller than 48 in all torsion angles and f1 and f2. Final-
ly, hydrophobic inter- and intramolecular interactions between
the phenyl groups of the dibenzylaminic moieties are present.
Compound 6 crystallizes in the P21/c space group with one
molecule in the asymmetric unit. The triphenylamine and thio-
phene moieties have geometries similar to those of 4, and the
cyanoacetic group is coplanar to the rest of the molecule. The
most relevant information is given by the conformational ar-
rangement around f2 and f3, which are both close to 1808
with a s-trans conformation within 68 (Table 1) as in the abso-
lute minimum of the first-principles calculations. A short S···N
intramolecular contact (d=3.26(7) ꢁ) is observed. The crystal
packing of 6 is driven exclusively by short-contact interactions
(less than the sum of the van der Waals radii) because no H
bonding is possible. Hydrophobic interactions between the
phenyl moieties are observed, similar to those in 4. Further-
more, the molecular packing reveals that the short-contact net-
work is formed by the intermolecular interaction between the
triphenylamine and cyanoacetic groups of adjacent molecules.
Figure 4. Crystal packing of D5 that shows H bonds and hydrophobic inter-
actions between aromatic groups.
phobic clustering of phenyl groups is observed as in 4 and 6.
It can be concluded that the phenyl–phenyl interactions are
the common feature of all three solved crystal structures and
must be important also if D5 is bound to TiO2 nanoparticles.
Apart from these common features, it is surprising that the
crystal packing of D5 and its precursor are different. Although
parallel p–p stacking interactions are observed in D5, in 4 the
molecules are pillared perpendicularly with T-shaped interac-
tions (Table S4). Compound 6 shows a completely different
packing without stacking of the planar aromatic moieties, but
the phenyl–phenyl interactions are still observed. The torsion
angles f2 and f3 are almost planar (Table 1) but with devia-
tions, within 108, larger than those suggested by theoretical
calculations (in which deviations from 0 and 1808 are within
28; Table S4). Such freedom is confirmed by calculated PES,
and a rather flat energy trend is observed between ꢁ20 and
+208 and ꢁ160 and +1608 (Figure 2a), and by experimental
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 15
&
6
&
ÞÞ
These are not the final page numbers!

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
data in the literature (Figure S4). The variety and richness of
the molecular interactions and deviation from planarity high-
lighted by crystallographic and computational studies, which
probably also occurs also in solution and after adsorption on
the TiO2 surface, suggest that dye uptake is a complex phe-
nomenon that requires a quantitative study for optimization in
both material use and cell performance.
(not less than ꢂ0.1%, as estimated from molar extinction coef-
ficient and the concentration of D5). To find the correct ranges
of D5 concentrations and P25 amounts, we assumed the
chemical formula of TiO2 (anatase phase, density 4.23 gcmꢁ3),
a spherical shape of the particles with an average diameter of
ꢂ20–25 nm (confirmed by Scherrer particle size analysis from
grazing-incidence XRPD data collected from TiO2-covered
slides; Figure S1), the weight of one sphere of P25 is 3.1906ꢂ
10ꢁ17 g, and the surface available for the sorption per mg of
P25 is 6.14ꢂ1016 nm2. Moreover, as evaluated with
MOLDRAW,[40] a molecule of D5 bound to the sphere by the cy-
anoacetic group in which the diphenyl amino moiety forms
the outer part covers approximately 0.5 nm2.
Chemometric study of D5 uptake
A full factorial design (FFD) was used that took into account all
the parameters involved in the dye uptake to maximize the in-
formation with the minimum number of experiments per-
formed.[37] A DoE approach allows us to evaluate the effect of
three variables on two series of samples (TiO2 powder and
slides) in triplicate by performing only 22 experiments. If we
used a standard 3D experimental grid, 162 experiments (33 ꢂ
3ꢂ2) would be necessary.
The results obtained from these preliminary uptake experi-
ments are reported in Table 3. In each experiment, P25 was
put in contact with the D5 solution for 16 h at 258C in a dark
Table 3. D5 uptake on P25 powders.
Preliminary experiments with P25 powder were performed
to estimate the correct range of D5 concentrations for the
uptake study. Then a first FFD was performed using different
amounts of P25 powder put in contact with D5 solutions at
different concentrations with and without the presence of
a dispersant (CDCA) to evaluate the effects of these parame-
ters in a model system. Finally, an FFD experiment applied to
a more complex and realistic model on standard TiO2 slides
was performed to explore the mutual influence of three pa-
rameters (D5 concentration, CDCA concentration, and soaking
time (t)) on D5 uptake in real systems.
Experiment
D5 concentration
[mmolLꢁ1
Amount of P25
[mg]
]
1
2
3
4
5
6
7
0.1
0.5
0.1
0.5
0.3
0.3
0.3
1.0
1.0
5.0
5.0
2.5
2.5
2.5
A simple, fast, and nondestructive method for dye-uptake
evaluation is UV/Vis spectroscopy, which can measure indirect-
ly but precisely the amount of dye extracted from solution by
both TiO2 powders and slides.[38,39] The direct evaluation of the
amount of grafted dye can be undertaken by disruptive and
more time-consuming methods such as UV/Vis spectroscopy
after dye desorption (with the implicit risk of partial degrada-
tion of the dye) or thermogravimetric analysis (TGA; which is
less selective and precise and has the drawback of that it is
not applicable to standard electrodes) on powdered TiO2 sam-
ples. For these reasons, the indirect UV/Vis method was select-
ed for extensive FFD studies on real samples, and TGA meas-
urements were used to validate some relevant uptake condi-
tions on P25 powders.
bottle to protect the solution from the light. The results are ex-
pressed as the number of D5 molecules (molec) retained
under batch conditions by 1.0 mg of P25 [molecmgꢁ1]. In the
planned experiments, the maximum and minimum levels of
the variation ranges of D5 and P25 (experiments 1–4) were se-
lected; moreover, to evaluate the experimental error associated
with the method, three replicates were performed with both
variables fixed to the values that correspond to the center of
the ranges (experiments 5–7). The evaluated standard devia-
tion associated with the methodology was 4.77ꢂ
1016 molecmgꢁ1, and the measured differences in the quantity
of grafted D5 are, therefore, statistically significant.
The obtained data confirmed that the sorption is affected by
both variables with a positive correlation (the amount of graft-
ed D5 increases as the amounts of both D5 and P25 increase)
and allowed us to optimize the experimental procedure.[41] No-
tably, the recorded UV/Vis spectra of the concentrated D5 solu-
tions show that the wavelength of the absorption maximum
has a bathochromic shift of ꢂ20 nm, which is probably caused
by the attractive intermolecular interactions highlighted by
XRD analysis that are more likely to occur in concentrated solu-
tions. To avoid these aggregation processes, the dispersant
CDCA must be included in the real uptake experiments on
TiO2 powders and slides.
Preliminary evaluations
UV/Vis spectra of D5 ethanol solutions at different concentra-
tions were measured to obtain a mean value of the molar ex-
tinction coefficient (e) of 35530 cmꢁ1 at l=448 nm as the only
reported value was measured in acetonitrile.[15] Notably, the
maximum number of D5 molecules that can be grafted theo-
retically must not exceed the physical sorption limit. To stay
below this limit, suitable amounts of D5 and P25 for the ad-
sorption should range from 1.0–5.0ꢂ10ꢁ4 m in 10 mL of D5 so-
lution if amounts of P25 of 1.0–5.0 mg are used. The indirect
evaluation of dye uptake by UV/Vis measurements requires
that the process consumes a significant mole fraction of dye
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 15
&
7
&
ÞÞ
These are not the final page numbers!

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
D5 sorption on P25 powder
hamper high uptakes. Nevertheless, for the comprehension of
the system, the effects of the interaction of the factors, if rele-
vant, must be considered. The interaction of the factors allows
us to describe the simultaneous effects that the factors exert
on the system in either a synergic or antagonistic way. In our
case, only the interaction between the concentration of D5
and t is relevant; a graphical method based on a two-way
table is the best approach to highlight their mutual interac-
tion.
23 FFD
P25 powder was used as the simplest possible model system.
A FFD was planned to investigate the effect of D5 concentra-
tion, contact time (t), and CDCA concentration.[7] To investigate
the principal and the interaction effects of the three variables,
an FFD 23 plan was performed, and the eight required experi-
ments (experiments 1–8) are reported in Table 4; moreover,
Two-way Table 5 was built by averaging the response of
each couple of combinations with the same values of the two
Table 4. Experimental data from 23 FFD on TiO2 powders. + and ꢁ repre-
sent the highest and the lowest values of the variables.
Table 5. Two-way table that illustrates the D5*t two-factor interaction.
Experiment D5 t CDCA D5 conc.
t
CDCA
conc.
D5 uptake
D5 uptake [1016 molecmgꢁ1
]
[mmolLꢁ1] [h] [mmolLꢁ1
]
[1016 molecmgꢁ1
]
D5 concentration
[mmolLꢁ1
0.40
0.04
6.8
6.9
12.7
5.9
1
2
3
4
5
6
7
8
9
ꢁ
+
ꢁ
+
ꢁ
+
ꢁ
+
0
ꢁ ꢁ
ꢁ ꢁ
+ ꢁ
+ ꢁ
ꢁ +
ꢁ +
+ +
+ +
0.04
0.40
0.04
0.40
0.04
0.40
0.04
0.40
0.22
0.22
0.22
4.0 0.0
4.0 0.0
28.0 0.0
28.0 0.0
4.0 16.0
8.1
8.4
7.3
13.5
5.7
5.2
4.4
12.0
8.0
8.2
8.2
]
4.0
28.0
t [h]
4.0 16.0
28.0 16.0
28.0 16.0
16.0 8.0
16.0 8.0
16.0 8.0
variables: the D5 concentration values are the rows and the t
values are the columns, so the bottom left quadrant represents
the experiments characterized by the lowest D5 value (ꢁ) and
the lowest t value (ꢁ); as there are two experiments with
these values (experiments 1 and 5) the average of the respons-
es given by two experiments is reported in the table.
0
0
0
0
0
0
10
11
0
0
In the bold central cell, on moving from left to right, the D5
concentration is kept constant (and t goes from the lowest to
the highest value), conversely, on moving from bottom to top,
t remains constant and the D5 concentration increases.
On one hand, D5 and t show a synergic effect as the largest
values of dye uptake were obtained, as expected, if high con-
centrations of D5 were put in contact for a long time (top
right corner). Under these conditions, the synergistic effect is
dominant with respect to the CDCA concentration as no rele-
vant variations in the amount of dye uptake were recorded in
the experiments with or without the dispersing agent (experi-
ments 4 and 8 in Table 4). On the other hand, both intermedi-
ate conditions (long soaking time and low D5 concentration or
short soaking time and high D5 concentration) reduce D5
uptake with respect to low D5 concentration and short t
(bottom left corner).
three replicates of the central experiment (experiments 9–11)
were performed at the beginning, in the middle, and at the
end of the FFD to confirm the repeatability and to estimate
the experimental error.
The results reported in Table 4 were used to calculate an or-
dinary least-squares (OLS) regression model that relates the ex-
perimental result y, that is, the amount of grafted D5, to the
experimental factors (D5 and CDCA concentrations and contact
time) and their interactions. The significant effects were evalu-
ated by Student’s t test in which each regression coefficient
was compared with the standard error multiplied by the ap-
propriate Student’s t value of 2.92 (a=95%, three degrees of
freedom).
The following OLS model was obtained for D5 uptake by
powdered TiO2 [1016 molecmgꢁ1], that is, Yp in Equation (1):
In these experiments the result of the F test[42] for the pres-
ence of the quadratic effect was negative, so additional experi-
ments to evaluate further variable levels, besides the three
chosen ones, would not add any new information about the
studied system.
Yp ¼ 8:11þ1:69 D5þ1:24 tꢁ1:25 CDCAþ1:73 D5 t
ð1Þ
This is satisfactory as the R2 value is 0.9914 and the observed
and predicted values are in good agreement (Figure S5A).
The OLS model indicates the relevant factors and their ef-
fects on the amount of adsorbed D5. The higher the value of
the coefficient in each term, the more important the effect of
the factor on the response. + or ꢁ indicate an increase or de-
crease of the D5 uptake if the considered factor is increased.
All the principal factors are relevant from statistical analysis of
experimental data: the concentration of D5 and t are both as-
sociated with a positive effect, whereas the concentration of
CDCA has a negative effect, that is, high CDCA concentrations
The adopted method can measure precisely but indirectly
the amount of dye extracted from the solution by TiO2 pow-
ders and slides. The method was validated by evaluating the
quantity of grafted dye by TGA measurements, which have the
advantage of directly detecting the amounts taken up but
have the drawback that they are less precise and not applica-
ble to slides used in technological applications.
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 15
&
8
&
ÞÞ
These are not the final page numbers!

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
Thermogravimetric analysis
D5 sorption on commercial TiO2 slides
TGA was performed in air on D5, CDCA, and pure TiO2 powder
as reference materials (Figure S6A) and on D5-sensitized TiO2
powders with and without CDCA under the same conditions of
experiments 4 (+ + ꢁ) and 6 (+ ꢁ +) (Table 2), which corre-
spond to the highest uptake in the absence and presence of
CDCA, respectively (Figure S6B). Full TGA data plots and com-
ments are available in the Supporting Information. UV/Vis
spectroscopic indirect determination indicated an uptake of
1.0ꢂ1017 molecules of D5 for 3.0 mg of P25. If we take into ac-
count the molecular weight of D5, this corresponds to an ex-
pected weight loss of 2.7%. TGA data indicated a weight loss
between 3 and 6% for the two analyzed samples, which de-
pended on the adsorption conditions. These values are in
agreement with the previous determination (same order of
magnitude as the UV/Vis measurements) and confirm that the
dye that has left the solution, detected precisely by UV/Vis
measurements, was actually grafted on the TiO2 powder.
The TGA profiles of P25 and D5-sensitized P25 are signifi-
cantly different (Figure S6). The first consideration, which con-
firms a chemical interaction between the dye and the sub-
strate, is the clear difference in the thermal degradation profile
shown by pure D5 with respect to the D5-sensitized sample.
This suggests that the effect of the contact does not originate
a physical mixture, but instead, a system with strong interfacial
interactions, which is able to modify the thermal degradation
profile significantly.
After the successful experiment on P25 powders, an analogous
FFD 23 plan was performed on commercial TiO2 slides to inves-
tigate the principal and the interaction effects of the three var-
iables under real working conditions. Each experiment consist-
ed of a sorption test in which the TiO2 slides were immersed in
10.0 mL of a solution that contained D5 and CDCA at different
concentrations for the contact times required by the experi-
mental plan. Three replicates of the central experiment (experi-
ments 9–11) were performed at the beginning, in the middle,
and at the end of the FFD to confirm the repeatability of the
analysis and to estimate the experimental error. The eight re-
quired experiments (experiments 1–8) plus three repetitions of
the central point are reported in Table 6, in which uptake is ex-
Table 6. Experimental data from 23 FFD on TiO2-covered slides
Experiment
t
D5 CDCA t
[h] D5
Concentration [mmolLꢁ1] D5 uptake
CDCA
[1019 moleccmꢁ3
]
1
2
3
4
5
6
7
8
9
ꢁ ꢁ
+ ꢁ
ꢁ +
+ +
ꢁ ꢁ
+ ꢁ
ꢁ +
+ +
ꢁ
ꢁ
ꢁ
ꢁ
+
+
+
+
0
8.0 0.05
24.0 0.05
8.0 0.50
24.0 0.50
8.0 0.05
24.0 0.05
8.0 0.50
24.0 0.50
16.0 0.27
16.0 0.27
16.0 0.27
0.0
0.0
0.0
4.0
3.5
56.3
35.7
5.3
0.0
16.0
16.0
16.0
16.0
8.0
6.3
51.3
60.8
29.1
28.9
29.2
0
0
0
0
0
0
10
11
0
0
8.0
8.0
Although pure P25 shows a total weight loss of 1.4%, with
a significant contribution caused by the desorption of physi-
sorbed water, the D5-sensitized sample has a lower weight res-
idue because of the decomposition of the organic dye, which
leads to a final weight loss of 7.4%.
pressed, different from experiments with TiO2 powders in
which the number of molecules taken up in a unit volume of
TiO2 [moleccmꢁ3] is given. This unit was chosen because is
more direct from a technological application viewpoint and
because an accurate evaluation of the weight of TiO2 film on
slides is impossible. Uptake values in this unit can be calculat-
ed if we know the thickness of the slides (6.5ꢀ0.4 mm), which
is homogeneous within the experimental error (Experimental
Section). This homogeneity allows us to compare data from
different slides and to transform the amount of grafted mole-
cules from moleccmꢁ2 of a slide (the quantity used for techno-
logical applications) to moleccmꢁ3 of TiO2 (used in the TiO2
slide experiments) and molecmgꢁ1 of TiO2 used in the powder
experiments. Notably, the large apparent differences in the
data presented in Tables 4 and 6 (three orders of magnitude) is
because of the different measurement units.
To quantify the amount of dye in the sample, a weight loss
contribution of P25 similar to that of the neat material (1.4%)
should be assumed and subtracted from 7.4% to obtain 6.0%.
As D5 degrades only for the 95.9% of its initial weight, the
neat weight loss caused by D5 can be estimated to be ꢂ6.3%
of the total weight of the sample.
The estimated value of adsorbed D5 molecules per gram of
P25 from these measurements is 9.0ꢂ1019 molecgꢁ1. This
result is consistent with the indirect UV/Vis measurements and
demonstrates that the main mechanism of dye removal from
the contact solution is caused by adsorption onto P25.
In the presence of CDCA during the dye uptake, the TGA
profile shows a higher weight loss and a slight shift of all the
degradation processes to lower temperatures. A shift of the
onset of the degradation process (which appears above
2008C) is anticipated of ꢂ208C with respect to the D5-P25
system (observed at 2208C), and the maximum of the degrada-
tion rate is anticipated of ꢂ98C. Compared with D5-sensitized
P25, the additional weight difference (0.68%) suggests the
presence of CDCA cografted with D5 in the sample. An ap-
proximate 1:10 ratio between CDCA and D5 on the TiO2 sur-
face can be estimated from this experiment.
From these results, the following OLS model was obtained
for D5 uptake by TiO2 slides (1019 moleccmꢁ3), that is, Ys in
Equation (2):
Ys ¼28:2þ23:1 D5þ3:0 CDCAþ3:96 t CDCA
ð2Þ
þ2:0 D5ꢁCDCAþ3:6 t D5 CDCA
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 15
&
9
&
ÞÞ
These are not the final page numbers!

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
This yielded satisfactory results (R2 =0.9850), and the ob-
served and predicted values are in good agreement (Fig-
ure S5B).
D5 and CDCA are the only principal relevant factors and are
both associated with a positive effect, contrary to TiO2 powder
FFD [Eq. (1)] in which CDCA has a negative effect. Moreover,
the interactions of two and three factors are relevant. In this
case, the information contained in the three-factor interaction
can be extracted efficiently and shown by a graphical method
that considers the three possible two-way tables constructed
from the variation of the experimental response if a couple of
factors is varied each time and the third factor is constant.
From the data shown in Table 7, it is clear that if the D5 con-
centration is high (left), it is possible to obtain a large D5
uptake in many different situations (i.e., at short contact times
Figure 5. Surface plot that illustrates the D5*CDCA*t three-factor interaction
obtained for D5 at fixed high concentration constructed from the two-way
tables. The other two interactions (for fixed values of CDCA concentration
and t) are shown in Figure S7.
Table 7. Two-way table that illustrates the D5*CDCA*t three-factor inter-
action: here only the table obtained for D5 fixed at high and low values
is presented, the other two tables (for fixed values of CDCA and t) are
shown in Figure S7.
easily accessible to D5. At the same time, at a concentration of
the order of 10ꢁ4 m, D5 molecules are diluted, and aggregation
effects are limited either in the presence or in absence of
CDCA. On the contrary, on slide samples, if D5 molecules ap-
proach the TiO2 slide surface and start to penetrate into the in-
terstitial space among the nanoparticles of the TiO2 layer, the
available space is much smaller, and diffusion becomes a limit-
ing step with a compromise between kinetic and thermody-
namic effects (as discussed with respect to the data shown in
Table 7). D5 molecules are then constrained close to each
other, which thus facilitates the formation of dimers and ag-
gregates induced by the interactions unraveled by XRD and
calculations. In this situation, the action of CDCA as a disper-
sant becomes important to optimize the dye uptake, especially
at high concentrations and for long soaking times, as both fac-
tors are able to induce aggregation. An understanding the
driving forces of the aggregation processes by analysis of the
structural and molecular interactions is then of paramount im-
portance for the interpretation of the dye-uptake results,
which is summarized in the following section.
D5 uptake [1019 moleccmꢁ3
at high values at low values
]
CDCA concentration
[mmolLꢁ1
16.0
0.0
51.3
56.3
60.8
35.7
5.3
4.
6.3
3.5
]
8.0
24.0
8.0
24.0
t [h]
in the absence of CDCA, 56.3ꢂ1019 moleccmꢁ3, or even in the
presence of a high concentration CDCA if the contact time is
long, 60.8ꢂ1019 moleccmꢁ3). This behavior is caused by the
polydispersity of the TiO2 substrate, which presents a distribu-
tion of adsorption sites. At shorter times and without CDCA, ki-
netic effects prevail and less stable, more accessible sites are
saturated. Conversely, at longer times thermodynamic equilibri-
um is reached, which saturates stable sites with a partial
bleaching of less stable sites.
At low values of D5 (Table 7, right) the best result was ob-
tained if the values of the concentration of CDCA and t are
high, but the grafted amounts are very small and comparable
to the experimental error, so the recorded variations cannot be
considered as statistically significant; the same considerations
can be taken into account for the other two-way tables (Fig-
ure S7). In addition, the evaluation of the second-order effect,
with the addition of further variable levels to be investigated,
was not required as the result of the F test for the presence of
the quadratic effect was negative.[42]
Discussion of the combined approach
The structures obtained from XRD show the most stable con-
formations suggested by the calculations on the isolated mole-
cules. This implies that the intermolecular interactions that dic-
tate the crystal packing are not strong enough to vary the
thermodynamically stable conformations. Several weak interac-
tions, besides the expected H bonds, were observed in the
three structures, all of which showed close contacts between
the phenyl moieties. The D5 precursor 4 showed T-shaped in-
teractions between thiophene groups, whereas D5 showed
parallel packing of the aromatic moieties. These two kinds of
stacking are also probable on the TiO2 surface with no definite
preference for either arrangement. It can be inferred that the
same interactions must play an important role and induce ag-
gregation of D5-related molecules in solution and on the TiO2
The higher uptake conditions in the most interesting cases
of high D5 concentrations can also be seen clearly in Figure 5,
which provides a 3D visualization of the data shown in
Table 7.
The different effect of the concentration of CDCA on TiO2
slides and powder can be explained by the different diffusion
conditions and available space for the D5 molecules in the two
samples. In the powder sample, the TiO2 nanoparticles are sus-
pended in the solvent with no diffusion limitations and are
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 15
&10
&
ÞÞ
These are not the final page numbers!

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
surface. These aggregation forces can explain the well-known
dispersion problems shown by D5 and by D5 on TiO2 surfaces,
that is, self absorption and lateral charge transfer between dif-
ferent dye molecules, with reduction of the injection yields.
The information on the energy barriers suggests that prepa-
ration and soaking conditions allow the coexistence (in solu-
tion and on the surface) of a variety of conformers that are dif-
ferent from the most stable ones. Moreover, the electron injec-
tion yield during DSC function can be conformer-dependent in
principle, as different conformations show different planarity
and the electronic conjugation along the D5 framework is
modified. A computational study of the excited state struc-
tures, which takes into account both flexibility and conforma-
tional freedom, might provide more insight into the “real
world”.
available conformational landscape. These data are fundamen-
tal for us to better understand the role of chenodeoxycholic
acid (CDCA) and the optimized uptake conditions of D5 on
a TiO2 slides under working conditions. The ability of CDCA to
modify dye uptake (DoE), that is, that to hinder phenyl–phenyl
intermolecular contact and contrast the T-shaped and parallel
stacking (XRD) by intercalating on TiO2 (TGA) between adjacent
D5 molecules, is clarified by the quantitative measurement
(UV/Vis) of the parameters involved in dye uptake.
The DoE-assisted spectroscopic investigation was applied to
evaluate the dye uptake in DSC. The successful interpretation
of the obtained model, performed by complementary charac-
terization techniques, allowed us to propose the presented
UV/Vis–DoE approach as the simplest, fastest, most reproduci-
ble, and sensitive method that can be applied widely to under-
stand and optimize the uptake of any kind of dye.
XRD analysis and calculations gave interesting indications of
the structure of D5 in different situations but could not evalu-
ate the importance of the dispersant (CDCA) and the influence
of the soaking time, concentration of reagents, and the physi-
cal form of TiO2 (powder or slide) on the dye-uptake mecha-
nism. The spectroscopy measurements, aided by a chemomet-
ric approach to reduce the number of experiments and investi-
gate the interactions between the various parameters that in-
fluence dye uptake, allowed us to look into some of the issues
in which XRD could not provide an insight.
Experimental Section
Materials
TiO2 (Degussa P25, purity 99.5%; Germany), ethanol (purity 99.8%),
and CDCA (>97%) were purchased from Sigma–Aldrich (Milwau-
kee, WI, USA). Glass slides covered with TiO2 were purchased from
DyeSol Italia (Rome, Italy). The D5 stock solution was prepared at
5.0ꢂ10ꢁ4 m by dissolving 0.0445 g of D5 in 200.0 mL of ethanol;
working solutions at different concentrations were obtained by di-
lution of the stock solution with ethanol.
The FFD indicated, initially, that TiO2 powder and slides
behave differently (the role of CDCA is relevant only on TiO2
slides; Table 7), likely because of the greater importance of dif-
fusion problems in the solid sintered thin film. Therefore, of
the studied models, the reference model must comprise TiO2
slides. In this case, time and CDCA concentration are antago-
nists, therefore, the presence of CDCA allows a large dye
uptake only with long soaking times, whereas good uptakes
can be obtained at short soaking times without CDCA. These
two situations both allow large uptakes, but with long soaking
times in the presence of CDCA a uniform sparse TiO2 loading is
obtained, as suggested by the high injection yields,[7,14] where-
as with a short soaking time and no CDCA, D5 aggregation
and island formation probably occurs on the TiO2 surface.
Synthesis
Full details on the synthesis of 3, 4, D5, and 6 are available in Sec-
tion S1 in the Supporting Information.
Instrumentation
1H and 13C NMR spectra were recorded by using an Avance-200 in-
strument (Bruker, Milan, Italy) at 200 and 50 MHz, respectively. ESI-
MS spectra were recorded by using a LCQ Deca XP plus spectrom-
eter (ThermoElectron Corporation, Rodano, MI, Italy) as detailed in
the Supporting Information (Section S1). UV/Vis data were collect-
ed by using a UV/Vis Lambda 900 spectrophotometer (PerkinElmer,
Monza, MI, Italy). TGA measurements were performed by using
a TGA/DTA LF1100/851e instrument equipped with Store Software
(Mettler Toledo, Novate Milanese, MI, Italy) under the following
standard conditions: equilibration step at 608C for 30 min followed
by a ramp at 108Cminꢁ1 rate up to 8008C. Measurements were
collected under an air flow. XRPD measurements to analyze TiO2
particle size were performed by using a ThermoARL powder dif-
fractometer XTRA, and the details are given in the Supporting In-
formation.
Conclusions
First-principles calculations, XRD, UV/Vis spectroscopy, ther-
mogravimetric analysis (TGA), and design of experiment (DoE)
techniques have been used in a synergic way to shed light on
the destiny of dye molecules before, during, and after the
grafting process on TiO2 electrodes. Key components of dye-
sensitized solar cells (DSC), that is, dye and TiO2, have been
studied from the viewpoints of the molecular structure and of
the dye-uptake mechanism using the well-known D5 molecule
as a case study. This combined characterization approach pro-
vided detailed information on the molecular interactions,
stable conformations, and flexibility of the dye molecules.
Relaxed potential energy scan calculations, in addition to facili-
tating structure solutions by powder XRD, suggested that dyes
can exploit their conformational flexibility to optimize grafting
and packing on the TiO2 surface with a wider than expected
Single-crystal diffraction data were collected by using an Oxford
Xcalibur CCD area detector diffractometer with graphite mono-
chromator and MoKa (l=0.71069 ꢁ) radiation. Data reduction and
absorption corrections were performed using CrysAlisPRO
171.34.44 (Agilent Technologies, Cernusco, MI, Italy). Single-crystal
structure solution was performed by direct methods using
SIR2011[43] and refinement with full-matrix least-squares employing
SHELX-97.[44] H atoms were generated in calculated positions by
SHELX-97. Single crystals of 4 and 6 suitable for X-ray analysis were
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 15
&
11
&
ÞÞ
These are not the final page numbers!

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
obtained by the slow cooling of a saturated hot ethanol solution.
Attempts to grow D5 crystals from different solvents and different
temperature conditions only yielded small micron-sized crystals,
and an XRPD experiment had to be performed instead using the
microcrystals grown in acetonitrile. Relevant crystal data are report-
ed in the Supporting Information. XRPD experiments were per-
formed at the ESRF in Grenoble on the BM1A and BM1B beamlines
by using a high-resolution powder diffraction instrument (for in-
dexing) and a Pilatus 2M detector[45] placed at a distance of
120 cm at two different heights with respect to the incoming X-ray
beam to obtain a low and high 2q angular range (used for struc-
ture solution). The Pilatus XRPD patterns were collected by using
radiation with l=0.7040(1) ꢁ. The calibration was performed using
the lattice parameters of the NIST lanthanum hexaboride (LaB6)
standard (SRM 660b; nominal a=4.15695(6) ꢁ at RT). The crystal
structure was solved from XRPD data by simulated annealing using
the low-angle dataset only by EXPO2011 software.[46] The two
powder patterns, at low and high 2q range, were refined together
by the Rietveld method using TOPAS software.[47] Full details of the
crystallographic measurements are reported in the Supporting In-
formation.
The solution was collected, filtered, and analyzed by UV/Vis spec-
troscopy (447.9 nm) to determine the amount of dye still present
in solution. All the solutions were maintained in the dark.
After the sorption experiments, the P25 powder and TiO2 slides
were washed with ethanol (2ꢂ10.00 mL); the aliquots were then
recovered, centrifuged, filtered, and analyzed by UV/Vis spectrosco-
py under the same conditions of the sorption experiments.
Chemometric analysis
FFD, regression models, and all graphical representations were per-
formed by using Statistica 7.1 (Statsoft Inc., USA) and Excel 2013
(Microsoft Corporation, USA).
Acknowledgements
We would like to thank A. Coelho (Coelho Software Brisbane,
Australia) for his advice and especially for the use of constraints
in the TOPAS program. Prof. Davide Viterbo is acknowledged for
useful suggestions and discussion during data analysis and
manuscript preparation. ESRF and SNBL are acknowledged for
beam time. D. Chernyshov (SNBL, Grenoble, FR) is acknowledged
for support in data collection and reduction of XRPD Pilatus 2M
data. Fondazione CRC (grant n. 2008–1581) is gratefully acknowl-
edged for funding. L.P. acknowledges financial contributions from
the MIUR project “Multidisciplinary modeling of the structure of
layered materials” funded as FIRB in 2012, (code RBFR10CWDA)
for funding his bursary. C.B. and V.G. gratefully acknowledge fi-
nancial support by DSSCX project (PRIN 2010–2011, 20104XET32)
from Ministero dell’Istruzione, dell’Universitꢀ e della Ricerca.
Finally, G.F. thanks the “Scuola di Alta Formazione” of the Univer-
sity of Piemonte Orientale (Italy) and Compagnia di San Paolo
for funding her PhD bursary.
Theoretical calculations
The structural models of D5 were obtained by first-principles DFT
calculations employing G03[48] software. A careful analysis of stable
energy minima and of the energy barriers that separate them was
performed by using the B3LYP[49] functional and different basis
sets.
Determination of D5 uptake
Sorption experiments were performed by adding, under static con-
ditions, appropriate amounts of D5 to the selected amounts of P25
powder for each experiment. The systems were stirred magnetical-
ly for 16 h; then a portion of the solution (1.00 mL) was collected,
which was centrifuged twice at 268C and 3000 rpm for 15.0 min
and filtered through a 0.20 mm polypropylene membrane (VWR In-
ternational, West Chester, PA, USA). UV/Vis analysis was performed
at 447.9 nm for the determination of the amount of dye still pres-
ent in solution. All solutions were maintained in the dark.
Keywords: density functional calculations · dyes/pigments ·
titanium · UV/Vis spectroscopy · X-ray diffraction
[1] a) M. Grꢃtzel, Acc. Chem. Res. 2009, 42, 1788–1798; b) A. Hagfeldt, G.
Boschloo, L. Sun, L. Kloo, H. Pettersson, Chem. Rev. 2010, 110, 6595–
6663; c) J. Burschka, N. Pellet, S. J. Moon, R. Humphry-Baker, P. Gao,
M. K. Nazeeruddin, M. Grꢃtzel, Nature 2013, 499, 316–319.
[2] H. S. Jung, J. K. Lee, J. Phys. Chem. Lett. 2013, 4, 1682–1693.
[3] R. Y. Ogura, S. Nakane, M. Morooka, M. Orihashi, Y. Suzuki, K. Noda, APL
94 2009, 073308.
[4] For comprehensive reviews, see: a) H. Imahori, T. Umeyama, S. Ito, Acc.
Chem. Res. 2009, 42, 1809–1818; b) A. Mishra, M. K. R. Fischer, P.
Bauerle, Angew. Chem. Int. Ed. 2009, 48, 2474–2499; Angew. Chem.
2009, 121, 2510–2536; c) J. N. Clifford, M. Planells, E. Palomares, J.
Mater. Chem. 2012, 22, 24195–24201; d) J. Park, G. Viscardi, C. Barolo,
N. Barbero, Chimia 2013, 67, 129–135; for some selected recent exam-
ples, see: e) A. Yella, H.-W. Lee, H. N. Tsao, C. Yi, A. K. Chandiran, M. K.
Nazeeruddin, E. W.-G. Diau, C. Y. Yeh, S. M. Zakeeruddin, M. Grꢃtzel, Sci-
ence 2011, 334, 629–634; f) J. Park, C. Barolo, F. Sauvage, N. Barbero, C.
Benzi, P. Quagliotto, S. Coluccia, D. Di Censo, M. Grꢃtzel, Md. K. Nazeer-
uddin, G. Viscardi, Chem. Commun. 2012, 48, 2782–2784; g) C. Barolo,
J-H Yum, E. Artuso, N. Barbero, D. Di Censo, M. G. Lobello, S. Fantacci, F.
De Angelis, M. Grꢃtzel, Md. K. Nazeeruddin, G. Viscardi, ChemSusChem
2013, 6, 2170–2180.
The particle size and film thickness of the transparent TiO2-covered
glass (TiO2 slides), purchased from Dyesol, were characterized by
XRPD and UV/Vis-NIR spectroscopy, respectively (Figures S1 and
S2). A particle size of ꢂ25 nm and a thickness of the TiO2 film of
(6.5ꢀ4) mm were determined. These values were used to calculate
the amount of available grafting sites and to estimate the amount
of D5 that can be adsorbed by a single slide for a better design of
the preliminary experimental plan. The TiO2 thickness was con-
firmed by NIR spectroscopy (Figure S2 and Table S1) by analysis of
the absorption interference fringes of the TiO2 slides, generated by
the similarity of the radiation wavelength and the TiO2 thickness.[50]
The transparent TiO2 slides were immersed under static conditions
in a beaker in different solutions (10.00 mL) that contained differ-
ent amounts of D5 and CDCA for the contact times dictated by
the DoE. Usually, concentrations and contact times are optimized
by trial and error. Typical literature[7–14] conditions are 1–0.1 mm for
D5, 10 mm for CDCA, and 16 h contact time (overnight). Their
values were chosen for the DoE to explore the variable space and
find the optimal soaking conditions.
[5] N. Cai, R. Li, Y. Wang, M. Zhang, P. Wang, Energy Environ. Sci. 2013, 6,
139–147.
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 15
&12
&
ÞÞ
These are not the final page numbers!

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
[6] a) B. O’Regan, L. Xiao, T. Ghaddar, Energy Environ. Sci. 2012, 5, 7203–
7215; b) V. Johansson, L. Ellis-Gibbings, T. Clarke, M. Gorlov, G. G. Ander-
sson, L. Kloo, Phys. Chem. Chem. Phys. 2014, 16, 711–718.
[7] D. P. Hagberg, J.-H. Yum, H. J. Lee, F. De Angelis, T. Marinado, K. M. Karls-
son, R. Humphry-Baker, L. Sun, A. Hagfeldt, M. Grꢃtzel, Md. K. Nazeerud-
din, J. Am. Chem. Soc. 2008, 130, 6259–6266.
[8] E. Ronca, M. Pastore, L. Belpassi, F. Tarantelli, F. De Angelis, Energy Envi-
ron. Sci. 2013, 6, 183–193.
[9] M. Pastore, E. Mosconi, F. De Angelis, M. Grꢃtzel, J. Phys. Chem. C 2010,
114, 7205–7212.
[32] J. W. Hagemann in Crystallization and Polymorphism of Fats and Fatty
Acids (Eds.: N. Garti, K. Sato), Marcel Dekker Inc, New York and Basel,
1988, pp. 9–96.
[33] M. P. Balanay, S.-M. Kim, M. J. Lee, S. H. Lee, D. H. Kim, Bull. Korean
Chem. Soc. 2009, 30, 2077–2082.
[34] L. Palin, C. Barolo, D. Chernyshov, D. Viterbo, A. G. Moliterni, G. Croce,
W. van Beek, M. Milanesio, Data collection and analysis strategies for
structure solution of complex organic structures, in preparation for sub-
mission to J. Appl. Cryst.
[35] M. Milanesio, P. Ugliengo, D. Viterbo, G. Appendino, J. Med. Chem.
1999, 42, 291–299.
[36] The crystal structures of D5, 4, and 6 were submitted to the CCDC data
centre with submission codes CCDC 953631, 895122, and 895123, re-
spectively. These data can be obtained free of charge from The Cam-
bridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data_re-
quest/cif.
[10] F. Labat, T. Le Bahers, I. Ciofini, C. Adamo, Acc. Chem. Res. 2012, 45,
1268–1277.
[11] E. Dell’Orto, L. Raimondo, A. Sassella, A. Abbotto, J. Mater. Chem. 2012,
22, 11364–11369.
[12] H.-P. Lu, C.-Y. Tsai, W.-N. Yen, C.-P. Hsieh, C.-W. Lee, C.-Y. Yeh, E. W.-G.
Diau, J. Phys. Chem. 2009, 113, 20990–20997.
[37] DoE is based on model systems of different complexity to evaluate and
control the investigated variables in the appropriate variation ranges.
Generally, the exploration of the experimental domain starts with
a two-level FFD, which allows the effects of the principal factors and of
their interactions on the investigated response to be studied. The
number of experiments required is 2p, in which p is the number of in-
vestigated factors. These experiments correspond to all the possible
combinations of the two levels (usually indicated by + and ꢁ) of the
considered factors; then, if necessary, other experiments are added to
study the second-order effects of the investigated factors.
[38] V. Gianotti, M. Benzi, G. Croce, P. Frascarolo, F. Gosetti, E. Mazzucco, M.
Bottaro, M. C. Gennaro, Chemosphere 2008, 73, 1731–1736.
[39] S. Polati, F. Gosetti, V. Gianotti, M. C. Gennaro, J. Environ. Sci. Health Part
B 2006, 41, 765–779.
[40] a) P. Ugliengo, D. Viterbo, G. Chiari, Z. Kristallogr. 1993, 207, 9–23;
b) http:/www.moldraw.unito.it.
[41] To verify if the D5 molecules were grafted effectively on the P25
powder, two washing procedures were tested. In the first, the P25
powder that remained after centrifugation in experiments 6 and 7 was
contacted with 10.0 mL ethanol and left to stand overnight. The solu-
tion was centrifuged (3000 rpm, 268C, 15 min), and the powder was
again contacted with 10.0 mL ethanol for 10 min. The solution was cen-
trifuged, and the supernatant was combined with that obtained from
the first centrifugation and analyzed by UV/Vis. The second method dif-
fered only in the contact time of the first ethanol aliquot, which was
10 min. The number of molecules removed by the two treatments was
similar and of approximately the same order of magnitude as the esti-
mated standard deviation, which indicates that the molecules of D5 are
tightly bonded to the surface of TiO2 and not only physisorbed and/or
stacked on the surface in weakly bound multilayers.
[42] A. I. Khuri, J. A. Cornell, Response Surface, Design and Analysis, Marcel
Dekker, New York, 1987.
[43] M. C. Burla, R. Caliandro, M. Camalli, G. L. Cascarano, C. Giacovazzo, M.
Mallamo, A. Mazzone, G. Polidori, R. Spagna, J. Appl. Crystallogr. 2012,
45, 357–361.
[44] G. M. Sheldrick, Acta Crystallogr. Sect. A 2008, 64, 112–122.
[45] B. Henrich, A. Bergamaschi, C. Broennimann, R. Dinapoli, E. F. Eikenberry,
I. Johnson, M. Kobas, P. Kraft, A. Mozzanica, B. Schmitt, Nucl. Instrum.
Methods Phys. Res. Sect. A 2009, 607, 247–249.
[46] a) EXPO2011, version 1.12.09a, http://www.ba.cnr.it; b) A. Altomare,
M. C. Burla, G. Cascarano, C. Giacovazzo, A. Guagliardi, A. G. G. Moliterni,
G. Polidori, J. Appl. Crystallogr. 1995, 28, 842–846; c) A. Altomare, G.
Cascarano, C. Giacovazzo, A. Guagliardi, M. C. Burla, G. Polidori, M. Ca-
malli, J. Appl. Crystallogr. 1994, 27, 435–436.
[13] J. Li, W. J. Wu, J. B. Yang, J. Tang, Y. T. Long, J. L. Hua, Science China
Chem. 2011, 54, 699–706.
[14] D. P. Hagberg, T. Edvinsson, T. Marinado, G. Boshloo, A. Hagfeldt, L. Sun,
Chem. Commun. 2006, 2245–2247.
[15] G. Boschloo, T. Marinado, K. Nonomura, T. Edvinsson, A. G. Agrios, D. P.
Hagberg, L. Sun, M. Quintana, C. S. Karthikeyan, M. Thelakkat, A. Hag-
feldt, Thin Solid Films 2008, 516, 7214–7217.
[16] a) W. Zeng, Y. Cao, Y. Bai, Y. Wang, Y. Shi, M. Zhang, F. Wang, C. Pan, P.
Wang, Chem. Mater. 2010, 22, 1915–1925; b) H. N. Tsao, J. Burschka, C.
Yi, F. Kessler, M. K. Nazeeruddin, M. Grꢃtzel, Energy Environ. Sci. 2011, 4,
4921–4924; c) H. N. Tsao, C. Yi, T. Moehl, J.-H. Yum, S. M. Zakeeruddin,
M. K. Nazeeruddin, M. Grꢃtzel, ChemSusChem 2011, 4, 591–594; d) M.
Xu, M. Zhang, M. Pastore, R. Li, F. De Angelis, P. Wang, Chem. Sci. 2012,
3, 976–983.
[17] R. Gobetto, C. Nervi, B. Romanin, L. Salassa, M. Milanesio, G. Croce, Or-
ganometallics 2003, 22, 4012–4019.
[18] A. Arrais, E. Diana, R. Gobetto, M. Milanesio, D. Viterbo, P. L. Stanghellini,
Eur. J. Inorg. Chem. 2003, 1186–1192.
[19] A. Arrais, E. Boccaleri, G. Croce, M. Milanesio, R. Orlando, E. Diana, Crys-
tEngComm 2003, 5, 388–394.
[20] R. Gobetto, G. Caputo, C. Garino, S. Ghiani, C. Nervi, L. Salassa, E. Rosen-
berg, J. B. A. Ross, G. Viscardi, G. Martra, I. Miletto, M. Milanesio, Eur. J.
Inorg. Chem. 2006, 2839–2849.
[21] F. H. Allen, Acta Crystallogr. Sect. B 2002, 58, 380–388.
[22] a) S. P. Anthony, C. Delaney, S. Varughese, L. Wang, S. M. Draper, Crys-
tEngComm 2011, 13, 6706–6711; b) S. P. Anthony, S. Varughese, S. M.
Draper, Chem. Commun. 2009, 7500–7502; c) S. P. Anthony, S. Varugh-
ese, S. M. Draper, J. Phys. Org. Chem. 2010, 23, 1074–1079.
[23] a) Y.-T. Li, C.-L. Chen, Y.-Y. Hs, H.-C. Hsu, Y. Chi, B.-S. Chen, W.-H. Liu, C.-H.
Lai, T.-Y. Lin, P.-T. Chou, Tetrahedron 2010, 66, 4223–4234; b) S. Zheng,
S. Barlow, C. Risko, T. L. Kinnibrugh, V. N. Khrustalev, S. C. Jones, M. Yu.
Antipin, N. M. Tucker, T. V. Timofeeva, V. Coropceanu, J.-L. Bredas, S. R.
Marder, J. Am. Chem. Soc. 2006, 128, 1812–1817; c) F.-R. Dai, W.-J. Wu,
Q.-W. Wang, H. Tian, W.-Y. Wong, Dalton Trans. 2011, 40, 2314–2323.
[24] E. M. J. Johansson, T. Edvinsson, M. Odelius, D. P. Hagberg, L. Sun, A.
Hagfeldt, H. Siegbahn, H. Rensmo, J. Phys. Chem. C 2007, 111, 8580–
8586.
[25] G. E. P. Box, W. G. Hunter, J. S. Hunter, Statistics for Experimenters, Wiley,
New York, 1978.
[26] R. Carlson, Design and Optimisation in Organic Synthesis, Elsevier, Am-
sterdam, 1992.
[27] S. N. Deming, S. L. Morgan, Experimental Design: A Chemometric Ap-
proach, Elsevier, Amsterdam, London, New York, Tokyo, 1993.
[28] a) F. Bella, D. Pugliese, J. R. Nair, A. Sacco, S. Bianco, C. Gerbaldi, C.
Barolo, R. Bongiovanni, Phys. Chem. Chem. Phys. 2013, 15, 3706–3711;
b) D. Pugliese, F. Bella, V. Cauda, A. Lamberti, A. Sacco, E. Tresso, S.
Bianco, ACS Appl. Mater. Interfaces 2013, 5, 11288–11295.
[29] Y. J. Cheng, J. Luo, S. Hau, D. H. Bale, T.-D. Kim, Z. Shi, D.-B. Lao, N. M.
Tucker, Y. Tian, L. R. Dalton, P. J. Reid, A. K. Y. Jen, Chem. Mater. 2007, 19,
1154–1163.
[47] A. A. Coelho, J. Appl. Crystallogr. 2005, 38, 455–461; A. A. Coelho, J.
Appl. Crystallogr. 2003, 36, 86–95.
[48] Gaussian 03, Revision C.02, M. J. Frisch, G. W. Trucks, H. B. Schlegel,
G. E. Scuseria, M. A. Robb, J. R. Cheeseman, J. A. Montgomery Jr., T.
Vreven, K. N. Kudin, J. C. Burant, J. M. Millam, S. S. Iyengar, J. Tomasi, V.
Barone, B. Mennucci, M. Cossi, G. Scalmani, N. Rega, G. A. Petersson, H.
Nakatsuji, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M.
Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, M. Klene, X. Li, J. E.
Knox, H. P. Hratchian, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R.
Gomperts, R. E. Stratmann, O. Yazyev, A. J. Austin, R. Cammi, C. Pomelli,
J. W. Ochterski, P. Y. Ayala, K. Morokuma, G. A. Voth, P. Salvador, J. J. Dan-
[30] H. Xia, J. He, B. Xu, S. Wen, Y. Li, W. Tian, Tetrahedron 2008, 64, 5736–
5742.
[31] L. J. Gooßen, J. Paetzol, L. Winkel, Synlett 2002, 10, 1721–1723.
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 15
&13
&
ÞÞ
These are not the final page numbers!

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
nenberg, V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. Strain, O.
Farkas, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V.
Ortiz, Q. Cui, A. G. Baboul, S. Clifford, J. Cioslowski, B. B. Stefanov, G. Liu,
A. Liashenko, P. Piskorz, I. Komaromi, R. L. Martin, D. J. Fox, T. Keith,
M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W.
Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez, J. A. Pople, Gaussi-
an, Inc., Wallingford CT, 2004.
[50] a) M. Caglar, Y. Caglar, S. Ilican, J. Optoelectr. Adv. Mater. 2006, 8, 1410–
1413; b) S. Ilican, M. Caglar, Y. Caglar, Mater. Sci. Poland 2007, 25, 709–
718.
Received: March 18, 2014
[49] a) A. D. Becke, J. Chem. Phys. 1993, 98, 5648–5652; b) C. Lee, W. Yang,
R. G. Parr, Phys. Rev. B 1988, 37, 785–789.
Published online on && &&, 0000
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 15
&14
&
ÞÞ
These are not the final page numbers!

FULL PAPERS
Live and let dye: A polyene-diphenyla-
niline dye (D5) for dye-sensitized solar
cells is studied by a combination of
XRD, theoretical calculations, and spec-
troscopic/chemometric methods. These
data allow us to characterize the driving
forces that govern D5 uptake and graft-
ing, to infer the most likely arrangement
of the D5 molecules on TiO2, and to un-
derstand the aggregation phenomena
suggested by the chemometric study.
V. Gianotti, G. Favaro, L. Bonandini,
L. Palin, G. Croce, E. Boccaleri, E. Artuso,
W. van Beek, C. Barolo,* M. Milanesio*
&& – &&
Rationalization of Dye Uptake on
Titania Slides for Dye-Sensitized Solar
Cells by a Combined Chemometric
and Structural Approach
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 15
&15
&
ÞÞ
These are not the final page numbers!