Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Chemical Physics xxx (2010) xxx–xxx

Contents lists available at ScienceDirect

Chemical Physics
journal homepage: www.elsevier.com/locate/chemphys

Improvement of the efficiency of thiophene-bridged compounds for dye-sensitized


solar cells
Julien Preat ⇑, Denis Jacquemin, Catherine Michaux, Eric A. Perpète
Unité de Chimie Physique Théorique et Structurale, Facultés Universitaires Notre-Dame de la Paix, rue de Bruxelles, 61, 5000 Namur, Belgium

a r t i c l e i n f o a b s t r a c t

Article history: A quantum-chemical study is conducted in order to provide UV/Vis absorption spectra (with a ±0.20 eV
Received 28 April 2010 accuracy) and oxidation potentials (±0.50 eV accuracy) of a series of conjugated metal-free organic dyes
In final form 2 August 2010 containing triphenylamine (TPA) and thiophene (TH) moieties. These compounds have recently been
Available online xxxx
developed for dye sensitized solar cells (DSSCs), and are here compared to the tetrahydroquinoline
(THQ) class of dyes. Our theoretical results reveal that TPAs provide a larger DGinject. variability than
Keywords: THQ dyes, and we have therefore chosen to optimize the former structures. Our procedure made it pos-
Triphenylamine dyes
sible to get insights into the geometrical and electronic structures of the dyes, and to unravel the struc-
Tetrahydroquinoline dyes
Solar cell sensitizer
tural modifications needed to optimize the properties of TPA-based DSSCs. In particular, we propose ways
Electron injection to improve the electron injection process, as well as the light harvesting efficiency (LHE) of the dyes. On
Light harvesting abilities this purpose, we considered a large set of original compounds, and starting from the TPA structure, were
TDDFT shown to increase the efficiency of the dye: (i) the 18-OH,-COOH, 13,15-diOMe, 1a,1b-diCN functionali-
zation of TPA-2; (ii) the 1a,1b-diCN, 14,15-diOMe,17-CN,18-H,-COOH functionalization of TPA-1, these
specific groups inducing a strongly exergonic free enthalpy of injection; (iii) the 18-diCOOH substitution
of TPA-2 improves the LHE without suffering a deterioration of the exergonic character of the free
enthalpy of injection. Moreover, the molecular topology analysis demonstrates that, due to the lost of
coplanarity between the anchoring and the bridging unit, the positive charge is not directly brought in
contact with the TiO2 surface, consequently limiting the recombination reaction.
Ó 2010 Elsevier B.V. All rights reserved.

1. Introduction chored molecular sensitizer, and a redox electrolyte (the I =I 3


couple) [9–13].
The present energetic and environmental crisis has stimulated The Ru complexes photosensitizers show a solar energy-to-
the interest in the design of renewable energy sources. Indeed, electricity conversion efficiency of 10% in average. Nevertheless,
with the foreseen 1013 W of new power needed for the coming an increasing interest for purely organic DSSCs as substitutes for
50 years, the carbon oxide (mainly CO2) concentration resulting Ru complexes raised in recent years due to their key advantages,
from the fossil fuel-based energy generation will exceed the cur- e.g. a high molar extinction coefficient, a simple and relatively
rent level (400 ppm) that has to be maintained. In this frame- inexpensive preparation processes, a more straightforward compli-
work, solar photovoltaic devices are likely to be leading ance with environmental rules [14]. Moreover, several solid-state
technologies in a promising ‘‘low-carbon level” future [1–8]. DSSCs based on organic dyes appear to have equivalent perfor-
In order to greatly increase the penetration of photovoltaic de- mances than inorganic complexes, suggesting promising commer-
vices into the global ‘‘green” energy markets one should design cial applications [15]. Therefore, metal-free dyes like coumarin
compounds that can offer low-cost per kilowatt whilst presenting [16,17], merocyanine [18], indoline [19], xanthene [20], hemi-cya-
a large light-to-electricity conversion yield. In this context, the nine [21], hydroquinones [22,23], perylene and fluorene [24,25]
dye-sensitized solar cells (DSSCs) certainly appear as one of the have been tested in this framework.
most promising materials for converting the solar energy. Conse- The molecular architecture of most organic sensitizers includes
quently DSSCS received significant attention as low-cost alterna- a donor (D), a bridge (B, typically a p spacer), and an acceptor (A),
tives to conventional semiconductor (SC) photovoltaic devices. which are usually combined following a D–p–A (or D–B–A) rod-
These DSSCs are composed of a wide band gap semiconductor (typ- like configuration (Scheme 1) to maximize the efficiency of the
ically TiO2) deposited on a translucent conducting substrate, an an- photoinduced intramolecular charge transfer (ICT). Generally, the
critical factors governing the sensitization are: (i) the excited
⇑ Corresponding author. state’s redox potential has to match the energy of the conduction
E-mail address: julien.preat@fundp.ac.be (J. Preat). band (CB) of the semiconductor; (ii) the highest occupied

0301-0104/$ - see front matter Ó 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.chemphys.2010.08.001

Please cite this article in press as: J. Preat et al., Chem. Phys. (2010), doi:10.1016/j.chemphys.2010.08.001
2 J. Preat et al. / Chemical Physics xxx (2010) xxx–xxx

tion spectra of large molecular species, such as the DSSCs dyes [35–
HOOC CN 38]. For such calculations, one of the most popular schemes re-
mains the time-dependent density functional theory (TDDFT) as
it provides results that qualitatively agree with experimental data
N for a reasonable computational effort, especially when hybrid func-
tionals are used [39–54]. Nevertheless, it is known that ‘‘cyanine-
like” structures such as triphenylamine derivatives may be prob-
lematic to study with DFT [55–58].
S The present contribution is organized as follows. In Section 2 we
HOOC CN detail the computational level and the procedure that have been
N used to compute all important parameters. In Section 3, we present
a specific way to improve the dye sensitizer properties. More pre-
cisely, in this section, we compare the theoretical and experimen-
tal key parameters that have been used to estimate the free energy
S
change related to the injection. Our procedure has also been ex-
tended to the study of the dye linked to the semiconductor (an-
Donors Bridges Acceptors ode). The insights harvested in this section have been used to
Scheme 1. Sketch of typical D-B-A compounds used as dye sensitizer. In this bring out the optimal structural modifications. We conclude this
scheme, the donors are THQ (top) and TPA (bottom), the bridge is constituted of work by some future prospects and challenges.
phenyl or thiophene ring(s), and the acceptor is a phenyl- or a thiophene-
cyanoacetic acid.
2. Methodology

molecular orbital (HOMO) that must fit the iodine/iodide redox po-


2.1. Operation cell and the key parameters
tential, and the lowest unoccupied molecular orbital (LUMO) that
has to be higher in energy than the conduction band edge of the
We present in Fig. 1 a schematic representation of the operating
semiconductor; (iii) the light harvesting ability of the dye should
principles of the DSSC. The ‘‘heart” of the system is constituted by a
be large enough so to reach a substantial photocurrent response;
mesoporous oxide layer generally composed of TiO2 (anatase). A
(iv) the conjugation across the donor and anchoring group has to
layer of the sensitizer dye is grafted on this material and the pho-
be maximized, as it determines the charge transfer (CT) amplitude;
toexcitation of the dye results in the injection of an electron from
and (v) the electronic coupling strength between dye’s LUMO and
the donor orbital (LUMO) of the excited dye to the CB of the oxide
the semiconductor CB. This coupling guides the efficiency of the
semiconductor film. Unfortunately, the injected electron can
electron injection from the dye onto the semiconductor surface.
recombine with the oxidized sensitizer dyes (recombination). This
In practice, the major factors leading to the low conversion effi-
punitive reaction is in competition with the regeneration of the
ciency are the formation of dye aggregates, and the charge recom-
oxidized dyes by the redox mediator which is usually an organic
bination between the CB electrons of the surface and the positively
solvent containing redox system, such as the iodide/triiodide cou-
charged dye (or the electrolyte) [26–28].
ple (regeneration). The remaining electron can be transported (dif-
To improve the metal-free dyes used in DSSCs, appropriate DBA
fusion) in the semiconductor film as the conducting electrons
systems with adequate properties have to be designed. Recently, it
(CEs). These CEs can also react with the redox mediator molecules
has been reported that triphenylamine-like (TPA) [29] derivatives
or with molecules in solution, before reaching the counter elec-
and the cyanoactetic acid moiety are patterns of choice as D and
trode (leak reaction) [59]. Moreover, the metal-free organic com-
A (also anchoring) moieties, respectively [26,30]. Indeed, TPA is ex-
pounds used in DSSCs are certainly not shape-persistent: in
pected to strongly confine the cationic charge from the semicon-
principle they might undergo fluorescence that can obviously inhi-
ductor surface, therefore hampering the recombination.
bit the injection step. Nevertheless, it has been recently shown that
Additionally TPA features a steric hindrance that could prevent
the time scale for a surface-anchored TPA injected electron to enter
unfavorable aggregation of the dye at the semiconductor surface
in the conduction band of TiO2 from the excited state is on the
[30]. The same hold for similar molecules such as tetrahydroquin-
oline (THQ) dyes: they show valuable performances for DSSC [31].
The bridging groups steer the light absorption regions of the E (in eV)
DSSCs [26,32,22] and subsequently the scale of the electron injec-
tion from the excited state of the dye to the semiconductor surface. CE

We herein report the design of new molecules derived from several Diffusion Injection
organic dyes recently synthesized by Li et al. [14], Chang and Chow dye*
[33], and Chen et al. [22]. These dyes use TPA or THQ as donor, phe- E OX
Photoexcitation
nyl- or thiophene-cyanoacetic acid moieties as acceptor and by
thiophenes and/or phenylenes as bridging units. CB
The goal of this investigation is to gain insights into the geomet-
Pt
Leak reaction
4.0
rical and electronic structure of these systems, and to bring out the Fluorescence
adequate structural modifications to optimize the properties of the I3 - / I-
TPA-based DSSCs. Consequently, we focus on the free energy of the
electron injection onto the TiO2 substrate and on the light harvest-
ing abilities of the dyes, for which a TDDFT-based procedure is able VB Recombination Dye Regeneration

to deliver a qualitatively correct description, as demonstrated in


our recent methodological investigations [34].
TiO2 dye
E OX
6.0
Nowadays, ab initio approaches have become efficient and
attractive tools for the interpretation of the experimental data. Fig. 1. Principle of operation of a DSSC. In this scheme, CB and VB are the
The progresses in CPU resources now allow to compute the absorp- conduction and valence bands of the TiO2, respectively.

Please cite this article in press as: J. Preat et al., Chem. Phys. (2010), doi:10.1016/j.chemphys.2010.08.001
J. Preat et al. / Chemical Physics xxx (2010) xxx–xxx 3

order of femtoseconds to picoseconds, which is much faster than


the fluorescence time scale for this kind of compounds (nanosec-
ond order) [60]. In other words, the emission is certainly not a fac-
tor that would kill the TPAs potential in the present case. Finally,
the remaining CEs flow though the external circuit and the iodide
is regenerated by the reduction of triiodide at the counter electrode
(of platinum, Fig. 1).

2.2. Computational level

All calculations have been performed with Gaussian 03 [61], fol-


lowing a two-step procedure: (i) the optimization of the ground-
state geometry with DFT with a tight threshold that corresponds
to a rms (residual mean square) force smaller than 105 a.u. for Scheme 2. Sketch of the TPA-R with the numbering of the substitution positions.
the optimal geometry, and (ii) the determination of the vertical
electronic excitation energies by means of TDDFT. For the first step,
we have used the popular three-parameter B3LYP functional [62], description of the CT states can be achieved when a large fraction
in which the exchange is a combination of 20% Hartree–Fock (ex- of HF exchange is used. While the optimal fraction of exchange re-
act) exchange, Slater functional, and Becke’s generalized gradient mains system-dependent, it appears that using a fraction P0.50
approximation (GGA) correction [63], whereas the correlation part would result in a qualitatively correct physical description, at the
combines local and Lee–Yang–Parr (LYP) functionals [64]. price of an overestimation of both the excitation energies and
We compare in Table 1 the structural parameters (bonds, angles the oscillator strengths [70]. These conclusions therefore confirm
and dihedral) related to the electron diffraction analysis of the selection of BHandHLYP, a 50% hybrid, for the UV/Vis spectra
NPh3[65] with the data calculated for TPA-R (see Scheme 2). The evaluation. As an additional test, we found that the range-sepa-
results listed in this table show a good agreement between both rated CAM-B3LYP [47] functional provides kmax that are almost
approaches with an average error limited to 0.07 Å. This justifies identical (typical deviation of 0.05 eV) to the BHandHLYP [34] for
the choice of B3LYP for the geometry optimization. Moreover, we TPA test structures. This confirms the validity and reliability of
have performed a TDDFT [66] on PBE0 [67] and B3LYP ground-state the latter functional in our case.
geometries. The resulting electronic excitations present similar We have selected the 6-31G(d,p) [71] basis set (BS) for the
energies and oscillator strength f: 3.01 eV vs. 3.01 eV and 1.35 vs. ground-state optimizations and 6-311G+(2d,2p) [71] for the TD-
1.36, respectively. This confirms that the form of the hybrid func- DFT calculations. These BS have been shown to return converged
tional used to optimize the geometry has a very limited impact kmax for a series of TPA [34].
on the excitation energies. All the theoretical kmax reported in the following correspond to
For the TDDFT calculations, we used the BHandHLYP hybrid the first dipole-allowed singlet excited state from the ground state.
[68,63,64]. In a recent work dealing with the relationship between The iodine/iodide couple is used as regenerator in DCCS, implying
spurious (i.e. charge transfer) excited states and the amount of ex- that the solar cells work in solvent phase. This is why the UV/Vis
act exchange, Tretiak and Magyar [69] have examined the perfor- experimental data for triphenylamine-based dyes are reported in
mance of various TDDFT functionals, for a series of polar donor– solution. Therefore the polarizable continuum model (PCM)
acceptor systems, including the (E)-4-(4-(methylsufonyl) styryl)- [72,73,42,74] is used for evaluating bulk solvent effects at all
N,N-diphenylbenzenamine, which is similar to TPA-1. For the con- stages. In PCM, one divides the problem into a solute part (the
sidered systems, the authors have demonstrated that a good dye) lying inside a cavity, and a solvent part. We have selected
the so-called non-equilibrium PCM solutions, and we refer the
reader to Ref. [42] for extensive details about this procedure.
Table 1
According to the experimental set up, we retained tetrahydrofuran
Comparison between the experimental [65] gas-phase structure of
the TPA moiety (see Scheme 2) and their theoretical counterpart.
(THF), dichloromethane, ethanol (EtOH), and dimethylformamide
(DMF, as this compound is not defined in the Gaussian 03 package,
TPA (NPh3)
we used DMSO and fixed the relative dielectric constant to
Parameters Experiment Theory 36.70 [75]).
Bonds in Å
N-4b 1.418 1.429
2.3. Excited state properties
4b-3b 1.404 1.403
3b-2b 1.396 1.394
2b-1b 1.397 1.396 We propose to establish a reliable theoretical scheme to evalu-
1b-6b 1.397 1.396 ate the dye’s excited state oxidation potential, and quantify the
6b-5b 1.396 1.394
electron injection onto a titanium dioxide (TiO2) surface. The free
5b-4b 1.404 1.403
C-Hmean 1.123 1.085
energy change (in electron volts, eV) for the electron injection
can be expressed as [76],
Angles in degrees
N-4b-3b 123.4 120.0 

N-4b-5b 117.5 120.0 DGinject: ¼ Edye SC


OX  ECB ð1Þ
3b-4b-5b 119.1 119.4 
4b-3b-2b 120.3 120.1 where Edye
OX is the oxidation potential of the dye in the excited state,
SC
3b-2b-1b 120.6 120.5 and ECB is the reduction potential of the semiconductor conduction
2b-1b-6b 119.2 119.5 band. It is often difficult to accurately determine ESC
CB experimentally,
1b-6b-5b 120.7 120.5
because it is sensitive to the conditions, e.g. the pH of the solution.
6b-5b-4b 120.2 120.1
There we use ESCCB ¼ 4:0 eV for TiO2 [77], an experimental value cor-
Dihedrals in degrees
responding to conditions where the semiconductor is in contact
3b-4b-N-4a 45.2 48.0
with aqueous redox electrolytes of fixed pH 7.0 [76,78].

Please cite this article in press as: J. Preat et al., Chem. Phys. (2010), doi:10.1016/j.chemphys.2010.08.001
4 J. Preat et al. / Chemical Physics xxx (2010) xxx–xxx


Two models can be used for the evaluation of Edye OX [79,80]. The G
inject. dye* SC
= E OX - E CB
first implies that the electron injection occurs from the unrelaxed E (in
n eV) E
dye* dye
= E OX -
(1)
OX max
excited state. For this reaction path, the excited state oxidation po-
tential can be extracted from the redox potential of the ground
state Edye
OX and the vertical transition energy, that is the photoin-
duced intramolecular CT (ICT) that is related to kð1Þmax , according to
[76]:
dye*
E OX

Edye dye ð1Þ
OX ¼ EOX  kmax ð2Þ CB G
inject.
QS1

In the second model, one considers that electron injection occurs E CB


SC

after relaxation. Given this condition, Edye
OX is expressed as [76,81]: (1)
m ax

Edye
OX ¼ Edye
OX  Edye
00 ð3Þ

where Edye VB
00 is the adiabatic transition energy between the ground
state and the excited state corresponding to the ICT. Though elec- dye

tron injection from unrelaxed excited states has been observed for
TiO2 QS0
E OX
Q
TiO2 [82] and SnO2 surfaces [83,84], the relative contribution of
an ultrafast injection path remains unclear, and most experimental Fig. 2. Schematic representation of the key parameters evaluated in this work.
groups commonly assume that electron injection dominantly oc-
curs after relaxation [76,81]. In our previous investigation, we found
2.4. Summary of the model
that the absolute difference between the relaxed and unrelaxed
DGinject. is constant and is of the same order of magnitude as the
dye Fig. 2 clearly depicts the different parameters that are involved
Edye
OX and EOX mean average error (MAE). Consequently, the unre-
in the injection process and that are calculated as follow:
laxed pathway has been chosen for the DGinject. evaluation [34].
To calculate the 0–0 ‘‘absorption” line, we need both the S0 (sin-
1. Edye ð1Þ dye
OX as well as kmax are directly computed, that is EOX has been
glet ground state) and the S1 (first singlet excited state) equilib-
evaluated at the PCM-B3LYP/UB3LYP/6-31G(d,p) level whereas
rium geometries. More precisely, the 0–0 transition energy is
the kð1Þ
max related to the ICT HOMO ? LUMO transition is com-
calculated as:
puted at the PCM-TDBHandHLYP/6-311G+(2d,2p) level.


E00 ¼ kð1Þ reorg: 2. DGinject. and Edye


OX have been indirectly evaluated using Eqs. (1)
max  ES1 ð4Þ
and (2), respectively.
where
We want to underline that this procedure has already been pro-
Ereorg:
S1 ¼ ES1 ðQ S0 Þ  ES1 ðQ S1 Þ ð5Þ ven efficient for Ru based DSSCs [81], allows a fast and reliable
evaluation of the DGinject., as the excited state geometry optimiza-
and
tions are useless. Eventually, the possible origins of the major dis-
ES1 ðQ S0 Þ ¼ ES0 ðQ S0 Þ þ DES1 ð6Þ crepancies between theory and experiment can be considered: (i)
PCM does not explicitly consider solute–solvent specific interac-
Q S0 and Q S1 are the equilibrium geometries of the S0 and S1 states, tions; (ii) the vibronic effects are not taken into account [87], more
respectively. ES1 ðQ S0 Þ and ES1 ðQ S1 Þ denote the internal energies for particularly, for compounds containing a bithiophene unit, a dis-
the S1 state calculated at Q S0 and Q S1 , respectively, whereas DES1 tortion of the bridging chain (low-frequency bending modes) can
is the S0 ? S1 excitation energy. As geometry optimization in sol- occur in solution though in our calculations, the bithiophene and
vent phase of the excited singlet states at the TDDFT level is not trithiophene are only considered in their planar conformation;
available in QM programs available to us, we cannot determine (iii) the CT transition (for which one notices an important change
Q S1 directly. However, Cave et al. calculated the fluorescence spec- of the dipolar moment between the ground and excited state) in
tra of coumarin derivatives at the DFT level and propose to use the solution could be more accurately described by using the state-
geometry of the T1 state as an estimate of the S1 equilibrium geom- specific PCM [88]; (iv) Eq. (2) is only valid if the entropy change
etry [85,86]. Indeed, since both the S1 and T1 states are strongly associated with the light absorption process is negligible, (v) we
characterized by a HOMO ? LUMO contribution, it is reasonable have only considered the unrelaxed path of injection for comput-
to postulate that their equilibrium geometries are indeed similar. ing the DGinject. [34].
Therefore, the geometry optimization for the S0(T1) is performed
at the (U) B3LYP/6-31G(d,p) level, and for the sake of consistency,
3. Results
DES1 has been calculated using the same functional with the 6-
311G+(2d,2p) BS. For TPA-R of Scheme 2 this leads to a DES1 that
A DSSC comprises four major components: (i) the dye, (ii) the re-
values 2.45 eV and a Ereorg:
S1 that has been calculated at 0.72 eV using
dox shuttle, (iii) the SC photoanode and (iv) the cathode (Scheme 1).
Eq. (7):
In order to obtain significant improvements it is necessary to alter
Ereorg:
S1 ¼ ES0 ðQ S0 Þ  ES1 ðQ S1 Þ þ DES1 ð7Þ at least one of these four key contributors and we focus here on
ways to improve the dye sensitizer properties. In this section, we
In other words, it turns out that this reorganization energy remains present a confrontation between the theoretical and experimental
dye
quite small comparatively, i.e. Ereorg:
S1 only amounts to 20% of the ICT parameters ðkð1Þ dye
max ; EOX , and EOX Þ that have been used to estimate
ð1Þ
excitation energy ðkmax values 3.01 eV), in good agreement with the the free energy change related to the injection. We have also con-
results presented in our previous work [34]. sidered the impact of the bonding of the dye to the semiconductor.
Considering that the relaxed path would imply unaffordable As suggested by Fig. 2, we must raise the excitation energy of
CPU resources at this stage, we strive for the unrelaxed path of the dye without altering significantly its Edye OX Þ in order to reach a
injection for computing the DGinject. of novel dyes [34]. more exergonic injection reaction.

Please cite this article in press as: J. Preat et al., Chem. Phys. (2010), doi:10.1016/j.chemphys.2010.08.001
J. Preat et al. / Chemical Physics xxx (2010) xxx–xxx 5

3.1. UV/Vis spectra to-electricity conversion process of these DSSCs. For the THQ-
based dyes, TDDFT predicts only one allowed excitation with an
For the TPA derivatives, our previous investigation has shown important oscillator strength (P1.0) in the visible domain. This
that two allowed excited states characterized by large transition transition corresponds to a HOMO to LUMO electron promotion,
probabilities appear in the UV/Vis region [34]. This is in complete and explains the strong band centered between 400 and 500 nm.
agreement with experimental findings, since measurements show The molecular orbital analysis (MOA) [89] of THQ-2 (Fig. 3) con-
an intense first absorption band ½kð1Þmax  in the 400–600 nm region, firms the highly-delocalized character of the frontier orbitals.
and a second strong absorption band ½kð2Þ max  of close to 300 nm. Table 2 presents a comparison between simulated and experi-
These transitions are associated to excitations predominantly mental kmax. We consider a panel of 12 representative compounds
implying three molecular orbitals: HOMO-1, HOMO and LUMO: of the DPA/TPA families (see Schemes 3 and 4) [90,33,14], includ-
HOMO ? LUMO for kð1Þ max and HOMO-1 ? LUMO for kmax .
ð2Þ
ing four NDPA derivatives (in which a phenyl of the donor group
ð1Þ
kmax corresponds to an ICT between the TPA donor and the is replaced by a naphtene group). For the THQ family, 9 dyes have
cyanoacetic acid acceptor end group, whereas kð2Þ max (HOMO- been treated. For the majority of them, the bridging group is con-
1 ? LUMO) is assigned to a standard p ? p* transition. Therefore, stituted by ethylene and thiopene groups, except for THQ-9, rely-
kð1Þ
max witnesses the electronic excitation that activates the energy- ing on a thieno-thiophene moiety (Scheme 5).
TDDFT systematically underestimates the kð1Þ max and the TPA
compounds listed in Table 2 can be classified into two phenomeno-
logical groups: (i) the TPA-4 ? -7 derivatives, with TDDFT devia-
tions close to 0.40 eV, (ii) other dyes undergoing small errors
(0.10–0.30 eV). For the THQ series, the MAE are even larger. For
THQ-9, BHandHLYP overestimates the excitation energy by
0.70 eV and we have tested the popular PBE0 [67] hybrid shown
to be optimal in average [47]. The comparison between the TDB-
HandHLYP and TDPBE0 results shows that: (i) BHandHLYP pro-
vides two excitations with a f close to 1.0 (at 3.46 and 3.71 eV
with similar f of 0.8573 and 0.9015, respectively) which are evalu-
ated with PBE0 at 2.96 eV (f = 0.9759) and at 3.23 eV (f = 0.7153),
and (ii) PBE0 indeed provides a kTHQ
max
-9
in better agreement with
the experimental value (absolute deviation of 0.17 eV).
Note that, aside from the approximate nature of the available
functionals, a second source of error originates in the imperfect
(and single) ground-state geometries selected. For example, the
underestimated excitation energies of TPA-4 ? -7 may be related
to the floppy vinylene link presents in their backbone. This bridg-
Fig. 3. Representation of the THQ-2 HOMO (top) and LUMO (bottom). They have ing unit may present a variety of torsional conformations experi-
been obtained at the TDB3LYP/6-31G(d,p)//B3LYP/6-31G(d,p) level with a constant mentally, whereas it is perfectly planar in theory. This could
threshold of 0.05 jej. explain the observed red shifts. In contrast, the backbones of

Table 2
ð1Þ
kmax (in nm) provided by PCM-TDBHandHLYP//6-311+G(2d,2p), EdyeOX (in eV) that are obtained at the PCM-B3LYP/6-31G(d,p) level, and the resulting DG
inject.
(in
eV). We also provide the corresponding MAE (in eV) for the two series of dyes.

Compounds Edye
ð1Þ
kmax DGinject. Solvent(s)a Ref.
OX

Theory Exp. Theory Exp. Theory Exp.


TPA-1 4.95 5.25 2.91 2.97 1.96 1.72 THF [33]
TPA-2 4.88 5.19 2.78 2.90 1.90 1.71 THF [33]
TPA-3 4.89 5.19 2.54 2.69 1.65 1.50 THF [33]
TPA-4 4.71 5.50 2.44 2.84 1.73 1.34 DMF [14]
TPA-5 4.58 5.45 2.30 2.71 1.72 1.17 DMF [14]
TPA-6 4.52 5.44 2.21 2.62 1.69 1.18 DMF [14]
TPA-7 4.62 5.49 2.42 2.80 1.80 1.31 DMF [14]
DPA 4.73 4.98 2.36 2.65 1.63 1.67 THF [33]
NDPA-1 4.91 5.22 2.90 2.94 1.99 1.72 THF [33]
NDPA-2 4.91 5.22 2.77 2.94 1.86 1.72 THF [33]
NDPA-3 4.87 5.21 2.53 2.69 1.66 1.48 THF [33]
NPA 4.72 5.13 2.35 2.58 1.63 1.45 THF [33]
MAE 0.50 0.24 0.28
THQ-1 4.83 5.35 2.77 2.81 1.94 1.46 DMF/EtOH [22]
THQ-2 4.66 5.21 2.47 2.65 1.81 1.44 DMF/EtOH [22]
THQ-3 4.63 5.26 2.48 2.68 1.85 1.42 DMF/EtOH [22]
THQ-4 4.50 5.13 2.28 2.65 1.78 1.52 DMF/EtOH [22]
THQ-5 4.52 5.21 2.33 2.72 1.81 1.51 DMF/EtOH [22]
THQ-6 4.41 5.11 2.19 2.61 1.78 1.50 DMF/EtOH [22]
THQ-7 3.84 5.12 1.84 2.52 2.00 1.40 DMF/EtOH [22]
THQ-8 4.70 5.37 2.48 2.64 1.78 1.27 DMF/EtOH [22]
THQ-9 4.61 5.28 3.46 2.79 2.85 1.51 DMF/EtOH [22]
MAE 0.70 0.35 0.51
a
For the THQ series, the Edye
OX is experimentally evaluated in DMF solution whereas the UV/Vis spectra have been taken in EtOH.

Please cite this article in press as: J. Preat et al., Chem. Phys. (2010), doi:10.1016/j.chemphys.2010.08.001
6 J. Preat et al. / Chemical Physics xxx (2010) xxx–xxx

Scheme 3. Sketch of the NDPA, NPA, TPA and DPA derivatives listed in Table 1.

Scheme 4. Sketch of TPA-4 to -7 derivatives listed in Table 1.

TPA-1 ? -3 are more rigid and the calculated excitation energies [93–95]. The analysis of the Figs. 4 and 5 leads to the following
are in better agreement with experiment. This trend is also notice- conclusions: (i) the p conjugated bridge in TPA-7 is coplanar to
able for THQ-7: 2 vinylene links, 0.7 eV red shift. the D/A moieties whereas for NPA-1 the two bridging phenyls un-
Despite these limitations, the BHandHLYP deviations remain in dergo a 30° out-of-plane distortion, therefore altering the conjuga-
the line of recently published PCM-TDDFT studies for triphenyl- tion of the acceptor–donor system, (ii) the MOs clearly indicate
methane dyes similar to TPA [91,92], as well as for other structures that the kNDPA-1
max corresponds to a CT process.

Please cite this article in press as: J. Preat et al., Chem. Phys. (2010), doi:10.1016/j.chemphys.2010.08.001
J. Preat et al. / Chemical Physics xxx (2010) xxx–xxx 7

NC

COOH

ab
THQ-1 01
THQ-2 11
THQ-3 02
THQ-4 12
THQ-5 03
THQ-6 13
NC

COOH

Fig. 4. 3-D structures of TPA-7 (top) and NPA-1 (bottom) derivatives.


THQ-7
NC
(TAP-1) to 4.88 eV (TPA-2). Also, in the series TPA-4 ? 5 ? 6 ? 7,
COOH
our theoretical results (4.71 eV ? 4.58 eV ? 4.52 eV ? 4.62 eV)
are in qualitative agreement to the measured trends (5.50 eV ?
S 5.45 eV ? 5.44 eV ? 5.49 eV). For the THQ-1 ? 2 ? 3 ? 4 ?
5 ? 6 series, the cathodic shift observed when the bridge length
N increases is also correctly reproduced by the theory, as the exper-
imental evolution, 4.83 eV ? 4.66 eV ? 4.63 eV ? 4.50 eV ?
Ph 4.52 eV ? 4.41 eV, is reasonably reproduced: 5.35 eV ? 5.21 eV ?
THQ-8 5.26 eV ? 5.13 eV ? 5.21 eV ? 5.11 eV. On the contrary, our
methodology is inadequate for THQ-7: the oxidation potential of
NC
this compound is dramatically underestimated by DFT. This spe-
S COOH cific error could be related to the free rotation around the bond
linking the ethylene units and the hydroquinoline moiety that
might be activated. Of course, while the Edye
 OX displacements
S
DEdye
OX are quantitatively well-reproduced, the absolute Edye
OX val-
S ues are only accurate to ±0.50 eV, and ±0.70 eV for TPAs and THQs,
respectively. Nevertheless, since DEdye
OX remains the truly crucial

N parameter, our procedure can safely be used to optimize new


H structures for DSSC applications.
THQ-9
Scheme 5. Sketch of the THQ derivatives with the numbering of the substitution 3.3. Electron injection
positions.
This section concerns one of the most important part of the pro-
cedure, that is the analysis of the free energy change related to the
3.2. Oxidation potential electron injection. Our procedure provides values of DGinject. that
qualitatively agree with experiment, with average errors of

In Table 2, we provide the DGinject., as well as Edye dye
OX and EOX (in 0.28 eV for TPAs and 0.51 eV for THQs, and the dependence upon
eV) for the TPA and THQ series. For the record, the impact of the molecular systems is quantitatively well-reproduced by the theory.
thermal contribution on the free energy change ðDGdye OX Þ has been In most cases, the underestimation of both Edye OX and the excitation
calculated for TPA-2 by performing a vibrational analysis at the energy leads to an overestimation of the ‘‘exergonic character” of
PCM-B3LYP/6-31G(d,p), and it turns out to be negligible (DGdye OX the injection free energy change. Note that for DPA, the accurate
amounts 4.83 eV vs. 4.88 eV for the internal energy). In addition, evaluation of the DGinject. (deviation of 0.03 eV) can be directly re-
the zero-point correction could also be neglected, well in the line lated the quality of Edye
OX (0.15 eV error). On the contrary, for THQ-
of our previous investigation [34]. 7, the overestimation of the free enthalpy variation parallels the
The results listed in Table 2 demonstrate that the cathodic dis- important underestimation of both the oxidation potential and
placements of Edye
OX are in good agreement with the experimental kmax [96].
trends. For instance, in the series TPA-1 ? 2 ? 3, lab measure- With respect to DGinject., we pick out four compounds in the TPA
ments show a slight cathodic shift (0.06 eV between TPA-1 and set, exhibiting the most exergonic character, mainly TPA-1, NDPA-
-2) that is perfectly reproduced by the theory: from 4.95 eV 1, TPA-2 and NDPA-2, for which the experimental DGinject. values

Please cite this article in press as: J. Preat et al., Chem. Phys. (2010), doi:10.1016/j.chemphys.2010.08.001
8 J. Preat et al. / Chemical Physics xxx (2010) xxx–xxx

Fig. 5. Representation of the NDPA-1 (left) and TPA-7 (right) HOMO (top) and LUMO (bottom). They have been obtained at the TDB3LYP/6-31G(d,p)//B3LYP/6-31G(d,p) level
with a constant threshold of 0.05 jej.

are about 1.70 eV. Concerning the THQ series, all compounds Before ranking the compounds, it has been checked that all the
present similar properties and no significant outlier emerges. compounds listed in Table 3 show a higher oxidation potential
than the I =I 3 redox potential (4.8 eV). Note that we did not ex-
3.4. Effects of chemical modifications clude TPC-13, TPC-14, nor TPA-15, even if their Edye OX are quite close
to the I =I3 redox potential (4.79, 4.77 and 4.68, respectively). In-
In the present section, we propose structural modifications deed the difference between Edye  
OX and the I =I3 redox potential is
improving the electron injection efficiency of the TPA-based DSSCs, smaller than the theoretical results accuracy.
and more precisely TPA-1 and -2, since they provide a larger DGin- With respect to the relative free energies of injection
ject
. variability. (DGinject:
r ¼ DGinject: inject:
dye =DGTPA-2 ) and the relative LHE (RLHE, is obtained
Of course, all modifications are theoretically possible, and a using Eq. (8) with the ratio fdye/fTPA-2 replacing f), the compounds
large panel of new structures can be tested. Nevertheless, we im- listed in Table 3 show the following trends: (i) TPA-10, -13, -14,
pose that all dyes include at least a terminal -COOH moiety on and -15 have a LHE significantly superior to TPA-2 (i.e. >0.9200),
the acceptor side, as this group is necessary to link the dye to the whereas TPA-16 (18-diCOOH-TPA-2) shows an improved LHE
semiconductor surface [30]. Next we applied three criteria: (i) (RLHE of 0.9046) without significant deterioration of the exergonic
the free energy of injection DGinject. in TiO2 has to be smaller than character of the free enthalpy of injection; (ii) for the set of the
1.84 eV, the referential value calculated for TPC-1, as the larger compounds listed in the two tables, it is impossible to isolate
DGinject., the faster the electron injection from the valence excited one structure showing a huge improvement of both LHE and injec-
state [80,97]; (ii) the oxidation potential of the dyes must be more tion driving force; (iii) for a large majority of dyes, the chemical
positive than the I =I3 redox couple (4.8 eV ± 0.1 eV) [98], ensuring modifications lead to a slight improvement of RLHE and a deterio-
that there is enough driving force for a fast and efficient regenera- ration of the DGinject. with respect to TPA-2; (iv) TPA-11, TPA-18,
tion of the dye cation radical; and (iii) the light harvesting effi- -21 and -22 are characterized by the worst LHE and DGinject:
r param-
ciency (LHE) of the dye has to be as large as possible in order to eters; (v) a cyano group grafted on the TPA-2 or -10 acceptor site in
maximize the photocurrent response. Here, LHE is expressed as position 17 (TPA-18) significantly deteriorates the free energy of
[99]: injection; (vi) the TPA-17 and -20 have DGinject. < 2.00 eV; (vii)
for the series TPA-13 to -15 (Scheme 6), the RLHE factor evolves
LHE ¼ 1  10A ¼ 1  10f ð8Þ
with the bridge length in the order TPA-14 (0.9291) ’ TPA-13
where A(f) is the absorption (oscillator strength) of the dye associ- (0.9478) < TPA-15 (0.9609); (viii) on the other hand, by adding
ated to kð1Þ
max . It is known that TDDFT is less accurate for the evalua- one (TPA-13 and -14) or two (TPA-15) ethylene moieties on each
tion of transition probabilities than for transition energies. The LHE side of the central thiophene bridging group of TPA-2, one signifi-
criterion has therefore been underweighted in our classification, as cantly deteriorates the DGinject:
r factor. This is explained by the
our assessments of DGinject. and Edye
OX are probably more reliable, even important decrease (0.2 eV for TPA-13 and -14 and 0.40 eV for
if the dependence of the experimental extinction coefficient with TPA-15) of the excitation energies when ethylene subunits are
respect to auxochromic effects can be qualitatively reproduced for added; (ix) for the same series, the oscillator strength is also
triphenylmethane derivatives [91]. strongly affected. For TPC-14, this modification is probably related
These three criteria have been used to set up an efficiency rank- to the gain of coplanarity of the bridging group. Indeed, TPA-2
ing of the compounds listed in Tables 3 and 4 in which we provide shows a 30° twist, altering the conjugation of the acceptor–donor
the relevant parameters for a set of more than 10 extra species. system, whereas TPA-14 is perfectly coplanar.

Please cite this article in press as: J. Preat et al., Chem. Phys. (2010), doi:10.1016/j.chemphys.2010.08.001
J. Preat et al. / Chemical Physics xxx (2010) xxx–xxx 9

Table 3
dye
Estimated (relative) DGinject: ðDGinject:
r Þ; Edye ð1Þ
OX and EOX (in eV) for a series of new structures. For each compound, we also provide the kmax (in eV), the corresponding oscillator strength
(f), and the relative light harvesting efficiency (RLHE). The theoretical parameters presented are obtained at the restricted and unrestricted levels using the B3LYP functional
ð1Þ
combined with the 6-31G(d,p) basis set. The kmax s (in eV) are obtained at the TDBHandHLYP/6-311+G(2d,2p) level. For both the ground and excited states, the solvation (THF) has
been taken into account by using the PCM model. Note that all the electrochemical parameters obtained for the new structures are compared to the TPA-2 counterparts.

DGinject.

Compounds Edye Edye
ð1Þ
kmax f RLHE DGinject:
OX OX r

TPA-2 1.90 4.88 2.10 2.78 1.7088 0.9000 1.00


1a,1b-diBr-TPA-2 TPA-8 1.14 5.68 2.86 2.82 1.8139 0.9132 0.60
1a,1b-diCl-TPA-2 TPA-9 1.79 5.03 2.21 2.82 1.8013 0.9117 0.94
1a,1b-diCN-TPA-2 TPA-10 1.69 5.26 2.31 2.95 1.8924 0.9219 0.89
1a,1b-diCN,17-CN-TPA-2 TPA-11 0.96 5.60 3.04 2.56 1.2079 0.8036 0.51
1a,1b-diF-TPA-2 TPA-12 1.82 4.94 2.18 2.76 1.7680 0.9077 0.96
18-diCOOH-TPA-2 TPA-13 1.76 4.79 2.24 2.55 2.1908 0.9478 0.93
TPA-14 1.74 4.77 2.26 2.51 1.9764 0.9291 0.92
TPA-15 1.70 4.68 2.30 2.38 2.4050 0.9609 0.89
TPA-16 1.87 4.87 2.13 2.74 1.7435 0.9046 0.98
TPA-17a 2.87 4.52 1.13 3.39 1.6731 0.9791 1.51
TPA-18b 0.85 5.98 3.15 2.83 1.1303 0.7820 0.45
TPA-19c 1.91 5.38 2.09 3.29 1.2783 0.7481 1.01
a
1a,1b-diCN,13,15-diOMe,18-OH,-COOH-TPA-2 (see Scheme 7).
b
1a,1b-diCN,13,15-diOMe,17-CN-TPA-2 (see Scheme 7).
c
1a,1b-diCN,13,15-diOMe,17-CN,18-H,-COOH-TPA-2.

Table 4 From the whole original set of new potential structures listed in
dye
Estimated (relative) DGinject: ðDGinject:
r Þ; Edye
OX and EOX (in eV) for a altered structures of Tables 3 and 4, only two dyes passed out our test grid of ‘‘best pho-
ð1Þ
TPA-1 and 3. For each compound, we also provide the kmax (in eV), the corresponding tovoltaic properties”: TPA-20 and -17. Indeed, their DGinject: > 1:00
r
oscillator strength (f), and the relative light harvesting efficiency (RLHE). The
theoretical parameters presented are obtained at the restricted and unrestricted
(1.12 and 1.51, respectively), a fact related to the important varia-
levels using the B3LYP functional combined with the 6-31G(d,p) basis set. The kmax ð1Þ
s tion of the kð1Þ
max (for instance, when going from TPA-2 to -17,
(in eV) are obtained at the TDBHandHLYP/6-311+G(2d,2p) level. For both the ground Dkð1Þ
max ¼ 0:41 eVÞ, and the lowering of the EdyeOX variation (0.36 and
and excited states, the solvation (THF) has been taken into account by using the PCM 0.07 eV for TPA-17 and -20, respectively).
model. Note that all the electrochemical parameters obtained for the new structures
For TPA-19, the increase of the excitation energy (0.41 eV) is
are compared to the TPA-2 counterparts.
quasi identical to the Edye
OX variation (+0.40 eV), therefore inducing
DGinject. a slight improvement of the DGinject:

Compounds Edye Edye
ð1Þ
kmax f RLHE DGinject: (0.01) but unfortunately, an
OX OX r r

TPA-2 1.90 4.88 2.10 2.78 1.7088 0.9000 1.00 important deterioration of the RLHE factor (0.1519). fTPA-19 (as
TPA-20a 2.36 4.81 1.64 3.17 1.0023 0.7409 1.12 well as fTPA-20) are very low (1.0023 and 1.2783) and this can be
TPA-21b 1.47 5.43 2.53 2.90 1.4447 0.8573 0.77 easily explained by the followings: (i) for TPA-20, his oscillator
TPA-22c 1.76 5.08 2.24 2.84 1.2018 0.7033 0.93 strength decreases originates in a loss of coplanarity of the bridg-
a
1a,1b-diCN,14,15-diOMe,17-CN,18-H,-COOH-TPA-1. ing group; (ii) for TPA-19, the B group is coplanar to D and the
b
1a,1b-diCN,9,11-diOMe,17-CN,18-H,-COOH-TPA-1. smaller f is probably related to the expected smaller electron
c
1a,1b-diCN,8,9,12,13-tetraOMe,17-CN,18-H,-COOH-TPA-3. mobility in thiophene than in phenyl rings. Moreover, Fig. 6 shows
that the LUMO of TPA-19 and -20 are centered on the anchoring
moiety and would therefore favor the electron injection from dye
3.5. Optimal structures to the semiconductor. Furthermore, the evaluation of the atomic
charge carried by the nitrogen qN in jej) at the TPA level confirms
Firstly, we would like to underline that whilst the absolute Edye
OX
the DSSC abilities of TPA-2-like structures.
values present an accuracy that is not optimal, the experimental As underlined in the introduction, the cationic TPA moiety con-
dye
EOX shifts are quantitatively well-reproduced by the procedure. centrates the positive charge far away from the semiconductor sur-
Consequently, in this section we have decided to select the optimal face after injection, and efficiently restricts the recombination
compounds following relative criteria: DEdye
OX remains the truly cru-
process. To evaluate this effect we used the PCM(THF)-B3LYP/6-
cial parameter. 31G(d,p) charges derived from the electrostatic potential (so-called

HOOC
S
CN

ab
TPA-13 01
TPA-14 10
TPA-15 11

Scheme 6. Sketch of the TPA-13, -14 and TPA-15 derivatives.

Please cite this article in press as: J. Preat et al., Chem. Phys. (2010), doi:10.1016/j.chemphys.2010.08.001
10 J. Preat et al. / Chemical Physics xxx (2010) xxx–xxx

in contact with the TiO2 surface, subsequently inhibiting the


recombination reaction.

3.6. Impact of the semiconductor

In this section we perform a theoretical investigation of the dye


linked to the semiconductor. As a first approximation, we consider
that the dye is bounded to one or two Ti atoms and a set of water
and hydroxyl ligands are added to make sure that the whole com-
plex is neutral (Fig. 7). This model is similar to the one of Peng and
coworkers [101].
Experiment have shown that for dyes like TPA-23, a bidentate
bridging (2 Ti) or a bidentate chelating (1 Ti) structure is likely to
happen for most anchored dyes [90]. The results stemming from

D D

O O O O
HO OH
OH2 Ti OH HO Ti Ti H2O
HO OH O
HO OH
H2O OH2
Scheme 7. Sketch of the TPA-1 and -2 derivatives with the numbering of the
substitution positions.

MK or Merz–Singh–Kollman charges [100]). For TPA-2, the varia-


tion of the charge (DqN) between the neutral qNdye ¼ 0:54jejÞ and
cationic species ðqNdyeþ ¼ 0:41jejÞ amounts to 0.13jej, whereas for N
TPA-17 (for which a larger DGinject:
r has been calculated), the varia-
tion is only limited to 0.06jej, the positive charge being completely D=
diluted in the conjugated bridging group. Of course, several exper-
imental factors might affect the recombination. However for TPA-
CN
17, since the coplanarity between the -CHCN-CHCOOH anchoring
group and its bridging unit is broken (60° out-of-plane distor- Fig. 7. Bidentate chelating (left, 1 Ti) and bidendate binding (right, 2 Ti) modes for
tion), it seems likely that the positive charge may not be directly TPA-23.

Fig. 6. Representation of the TPA-19 (left) and TPA-20 (right) HOMO (top) and LUMO (bottom). They have been obtained at the TDB3LYP/6-31G(d,p)//B3LYP/6-31G(d,p) level
with a constant threshold of 0.05 jej.

Please cite this article in press as: J. Preat et al., Chem. Phys. (2010), doi:10.1016/j.chemphys.2010.08.001
J. Preat et al. / Chemical Physics xxx (2010) xxx–xxx 11

Table 5

Estimated DGinject: ; Edye
OX and its experimental value ðEOX
dyeExp:
Þ; ðEdye ð1Þ
OX and kmax (in eV) as well as the oscillator strength (f) and the related LHE parameter for TPA-23 upon three
conditions: (i) free in solution, (ii) bounded to Ti (OH)3H2O (1 Ti model) and (iii) to Ti2O (OH)5(H2O)3 (2 Ti model). We also provide the variation (D) between the Free and the 1 Ti
and 2Ti models. a

DGinject.

Edye Edye-Exp:90 [90] Edye
ð1Þ
kmax f LHE
OX OX OX

Free 1.14 4.89 5.47 2.86 2.03 1.14 0.93


Ti(OH)3H2O 0.99 5.06 – 3.01 2.05 1.21 0.94
Ti2O(OH)5(H2O)3 1.01 5.06 – 2.99 2.03 1.23 0.94
D1Ti 0.15 0.17 – 0.15 0.02 0.07 0.01
D2Ti 0.13 0.17 – 0.13 0.0 0.08 0.01
ð1Þ
a
The Edye
OX s presented here are obtained at the restricted and unrestricted levels using the B3LYP functional combined with the 6–31G(d,p) basis set. The kmax s are obtained
at the TDB3LYP/6-311+G(2d,2p). For both the ground and excited states, the solvation has been taken into account by using the PCM model. In order to get more details about
the procedure, we refer the reader to our previous methodological investigations [34]. The selected solvent is dichloromethane [90].

the extended procedure is organized as the following: firstly, we abilities of a panel of dyes have been investigated. Starting with
evaluate the SC effects on the key parameters ðkmax ; Edye OX and the TPA-1 and -2 structures, the following modifications help to

Edye
OX Þ required to estimate the free energy change related to the improve the properties of the DSSC: (i) the 18-OH,-COOH, 13,15-
injection; secondly, we discuss the impact of the SC on the DGinject. diOMe, 1a,1b-diCN functionalization of TPA-2 (that is TPC-17);
parameter. (ii) the displacement of the terminal -CN group from the position
All parameters listed in Table 5 have been evaluated using both 18 to 17, the substitution of the hydrogen atoms by two -OMe
the 1 Ti and 2 Ti models and we provide the estimated variations of functions in positions 14 and 15, as well as the grafting of two -
dye dye inject.
kð1Þ
max ; EOX ; EOX and DG (in eV) when going from the free mole- CN groups in positions 1a and 1b on the TPA-1 moiety (leading
cules in solution to the 1 Ti and 2 Ti models. From results listed in to TPC-20). These changes provide a highly exergonic free enthalpy
this table we can conclude that: (i) the differences between 1 Ti of injection (2.87 eV ± 0.50 eV and 2.36 eV ± 0.50 eV, for TPA-17
and 2 Ti parameters are negligible and the first approach is ade- and -20, respectively, compared to 1.96 and 1.90 eV ± 0.50 eV
dye
quate to describe the SC effects; (ii) both kð1Þ max and EOX are affected for TPA-1 and -2).
by the titanium complex; (iii) when the dye is anchored on the In addition, an improved LHE parameter was predicted for TPA-
TiO2 surface, the experimental absorption spectra are blue-shifted 16 (18-diCOOH-TPA-2, RLHE of 0.9046) without a significant dete-
by a 0.1 eV compared to the solvated case [90] and this blue-shift rioration of the exergonic character of the electron injection.
is partially reproduced by the theory, tough within the selected Examination of the atomic charges and molecular orbital topol-
models, Dkð1Þ max is strongly underestimated; (iv) the effect of the SC ogy demonstrates that the positive charge is completely diluted in
is more important for Edye OX for which we notice an anodic displace- the conjugated bridging group. However, since the coplanarity be-
ment (0.2 eV) when going from the free dye to the complexed tween the -CHCN-CHCOOH anchoring part and the bridging unit is
structure; (v) the 1 Ti model provides Edye OX in better agreement with broken, the positive charge is not directly in contact with the TiO2
the experiment value than the free dye approximation (the error is surface and the recombination reaction should therefore be
0.41 eV instead of 0.50 eV) [34]; (vi) with the bidendate chelating inhibited.
dye
scheme (1 Ti model), one observes that Dkð1Þ max < DEOX . This results An investigation of the dye linked to the semiconductor as also
inject.
in a decrease of DG to 0.15 eV, which suggests that the QM been performed using a dye bounded to one or two titanium
approach dealing with free dyes in solution tends to slightly over- atoms. This study allowed to conclude that, on one hand, when
estimate the free enthalpy of the electron injection. the dye is attached on the TiO2 surface, its experimental kmax is
blue-shifted, and, on the other hand, the effect of the SC is more
important for EdyeOX for which we notice an anodic displacement
4. Concluding remarks (0.2 eV). Additionally this investigation suggests that the QM ap-
proach dealing with free dyes in solution tends to slightly overes-
In the present work, we managed to (i) gain insights into the timate the free enthalpy of the electron injection.
geometrical and electronic structures of thiophene-bridged organic To finish, let us state that the absolute computational accuracy
dyes; and (ii) to bring out the adequate structural modifications can be improved by solving the two principal computational road-
optimizing the properties of the TPA and THQ-based DSSCs, using blocks which are: (i) the TDDFT framework delivers a too restrictive
phenyl and thiophene spacers. It is clear that, in agreement with description of the electronic structure of TPAs. This could be fixed by
experimental trends, the BHandHLYP hybrid is a reasonable com- using more refined methods like CAS-SCF (or CAS-PT2) [102,56] or
promise functional for the TDDFT calculations. Using this scheme, MR-CI [103], allowing a multideterminental approach; (ii) the
the mean average error lies in the 0.200.30 eV domain. This sat- long-range corrected functionals could reduce the TDDFT errors
isfying achievement results in part from the selection of quite ex- though apparently not for cyanine systems [48–50,58,51,52,47].
tended basis sets and from the explicit consideration of bulk Nevertheless, the use of conventional DFT approaches to compute
solvent effects. auxochromic shifts of CT dyes are not uncommon in literature
Admittedly, the absolute accuracy for TPAs Edye OX is not optimal [53,54,104,57].
and the Edye
OX absolute deviation should be reduced, but the theoret- Similarly, we want to underline the recent work by Hagfeldt
ical procedure quantitatively reproduces the experimental dis- et al., who report the synthesis of a novel dye containing two
placements (e.g. DEdye OX Þ. Consequently, the ‘‘best” compounds TPA units connected by a vinyl group, linked to a rhodanine-3-ace-
selection as been performed following relative criteria. Indeed, tic acid moiety as electron acceptor. This dye has been synthesized
since DGinject. remains the truly crucial parameter, our procedure as a reference to study the intramolecular energy transfer (IET)
can safely be used to optimize new structures for DSSC applica- and charge transfer (ICT) processes and the results suggest that
tions. This investigation shows that TPAs provide a larger DGinject. both processes show a positive effect on the performance of DSSCs
variability (than THQs), and TPA-1 and -2 cores were therefore se- [60].
lected for further optimizations. In that context, the electron injec- Undoubtedly, the close future of organic solar cells relies on
tion efficiency into the TiO2 surface and the light harvesting their economic potential that depends upon several critical factors

Please cite this article in press as: J. Preat et al., Chem. Phys. (2010), doi:10.1016/j.chemphys.2010.08.001
12 J. Preat et al. / Chemical Physics xxx (2010) xxx–xxx

like the efficiency, the manufacturing costs, as well as sustainabil- [28] K. Sayama, S. Tsukagochi, K. Hara, Y. Ohga, A. Shinpou, Y. Abe, S. Suga, H.
Arakawa, J. Phys. Chem. B 106 (2002) 1363.
ity, weight, scalability or lifetime. At this point, a couple of condi-
[29] A. Hagfeldt, M. Grätzel, Acc. Chem. Res. 33 (2000) 269.
tioning techniques can be foreseen: solvent and solvent-free [30] Z. Ning, Q. Zhang, W.H.P. Wu, H. Tian, J. Org. Chem. 73 (2008) 3791.
processings. The first process offers a remarkable efficiency and [31] R. Chen, X. Yang, H. Tian, L. Sun, Photochem. Photobiol. A 189 (2007) 295.
are low-cost. However, these cells use volatile solvents in their [32] D.P. Hagberg, T. Marinado, K.M. Karlsson, K. Nonomura, P. Qin, G. Boschloo, T.
Brinck, A. Hagfeldt, L. Sun, J. Org. Chem. 72 (2007) 9550.
electrolytes, which is problematic for outdoor solar panels applica- [33] Y.J. Chang, T.J. Chow, Tetrahedron 13 (2009) 4726.
tions, in views of the need of a robust encapsulation. Therefore, sol- [34] J. Preat, C. Michaux, D. Jacquemin, E.A. Perpète, J. Phys. Chem. C 113 (2009)
vent-free ionic liquids have been looked for to obtain attractive 16821.
[35] F. Labat, I. Ciofini, H.P. Hratchian, M. Frisch, K. Raghavachari, C. Adamo, J. Am.
solutions. Device efficiencies of over 8% were recently achieved Chem. Soc. 131 (2009) 14290.
by using some low-viscosity formulations containing a ternary [36] J.L. Brédas, J.E. Norton, J. Cornil, V. Coropceanu, Acc. Chem. Res. 42 (2009)
melt of imidazolium iodure [105]. This result is of great importance 1691.
[37] D. Jacquemin, E.A. Perpète, I. Ciofini, C. Adamo, Acc. Chem. Res. 42 (2009) 326.
in the perspective of a large-scale real-life applications of DSSCs. [38] V. Barone, A. Polimeno, Chem. Soc. Rev. 36 (2007) 1724.
Nevertheless, whilst such a strategy has been widely applied for [39] C. Jamorski-Jödicke, H.P. Lüthi, J. Am. Chem. Soc. 125 (2002) 252.
Ru organic cells, this process has not yet been transfered to me- [40] C. Jamorski-Jödicke, H.P. Lüthi, J. Chem. Phys. 117 (2002) 4146.
[41] J. Preat, D. Jacquemin, V. Wathelet, J.M. André, E.A. Perpète, J. Phys. Chem. A
tal-free organic solar cells like triphenylamine. In such a context, 110 (2006) 8144.
a fundamental-level understanding of the interactions between [42] M. Cossi, V. Barone, J. Chem. Phys. 115 (2001) 4708.
the dye and the ionic solvent would be warmly welcomed to allow [43] C. Adamo, V. Barone, Chem. Phys. Lett. 330 (2000) 152.
[44] W. Adam, O. Krebs, Chem. Rev. 103 (2003) 4131.
a clear optimization strategy [105].
[45] E.J. Baerends, G. Ricciardi, A. Rosa, S.J.A. van Gisbergen, Coord. Chem. Rev. 230
(2002) 5.
[46] D. Jacquemin, J. Preat, V. Wathelet, M. Fontaine, E.A. Perpète, J. Am. Chem. Soc.
Acknowledgments 128 (2006) 2072.
[47] D. Jacquemin, V. Wathelet, E.A. Perpète, C. Adamo, J. Chem. Theory. Comput. 5
(2009) 2420.
The authors thank the Belgian National Fund for Scientific Re- [48] T. Tawada, T. Tsuneda, S. Yanagisawa, T. Yanai, K. Hirao, J. Chem. Phys. 120
search (FRS-FNRS) for their respective positions. All calculations (2004) 8425.
have been performed on the Interuniversity Scientific Computing [49] M. Kamiya, H. Sekino, T. Tsuneda, K. Hirao, J. Chem. Phys. 122 (2005) 234111.
[50] M. Chiba, T. Tsuneda, K. Hirao, J. Chem. Phys. 124 (2006) 144106.
Facility (ISCF), installed at the Facultés Universitaires Notre-Dame
[51] M.J.G. Peach, P. Benfield, T. Helgaker, D.J. Tozer, J. Chem. Phys. 128 (2008)
de la Paix (Namur, Belgium), for which the authors gratefully 044118.
acknowledge the financial support of the FNRS-FRFC and the ‘‘Lote- [52] M.A. Rohrdanz, J.M. Herbert, J. Chem. Phys. 129 (2008) 034107.
[53] C.A. Bertolino, A.M. Ferrari, C. Barolo, G. Viscardi, S. Caputo, G. Coluccia, Chem.
rie Nationale” for the convention number 2.4578.02 and of the
Phys. 52 (2006) 330.
FUNDP. [54] J.B. Prieto, F.L. Arbeloa, V.M. Martinez, I.L. Arbeloa, Chem. Phys. 13 (2003) 296.
[55] S. Grimme, F. Neese, J. Chem. Phys. 127 (2007) 154116.
[56] M. Schreiber, V. Bub, M.P. Fülscher, Phys. Chem. Chem. Phys. 3 (2001) 3906.
References [57] D. Jacquemin, E.A. Perpète, G. Scalmani, M.J. Frisch, R. Kobayashi, C. Adamo, J.
Chem. Phys. 126 (2007) 144105.
[58] D. Jacquemin, E.A. Perpète, G. Scuseria, I. Ciofini, C. Adamo, J. Chem. Theory.
[1] J.P. Holdren, Science 319 (2008) 424.
Comput. 4 (2008) 123.
[2] M. Parry, O. Canziani, J. Palutikof, P. van der Linden, C. Hanson (Eds.),
[59] T.W. Hamann, R.A. Jensen, A.B.F. Martinson, H. Van Ryswyk, J.T. Hupp, Energy
Intergovernmental Panel on Climate Change (IPCC): Climate Change 2007 –
Environ. Sci. 1 (2008) 66.
Impacts, Adaptation, and Vulnerability, Cambridge University Press,
[60] H. Tian, X. Yang, J. Pan, R. Chen, M. Liu, Q. Zhang, A. Hagfeldt, L. Sun, Adv.
Cambridge, UK, 2007, p. 1.
Funct. Mater. 18 (2008) 3461.
[3] M.I. Hoffert et al., Science 298 (2002) 981.
[61] M.J. Frisch et al., Gaussian, Revision D.02, in: Gaussian, Inc., (Ed.), Gaussian,
[4] M.A. Green, Power to the People: Sunlight to Electricity using Solar Cells,
Inc., Wallingford, CT, USA, 2005.
University of New South Wales Press, Sydney, Australia, 2000. pp. 6–91.
[62] A.D. Becke, J. Chem. Phys. 98 (1993) 1372.
[5] K. Zweibel, J. Mason, V. Fthenakis, Sci. Am. 298 (2008) 64.
[63] A.D. Becke, Phys. Rev. A 38 (1988) 3098.
[6] R.F. Service, Science 309 (2005) 548.
[64] C. Lee, W. Yang, R.G. Parr, Phys. Rev. B 37 (1988) 785.
[7] D.M. Kammen, S. Pacca, Annu. Rev. Environ. Sci. 29 (2004) 301.
[65] V.A. Naumov, S. Samdal, A.V. Naumov, S. Gundersen, H.V. Volden, Russ. J. Gen.
[8] E.A. Alsema, Prog. Photovoltaic 8 (2000) 17.
Chem. 75 (2005) 1956.
[9] T.A. Heimer, E.J. Heilweil, C.A. Bignozzi, G. Meyer, J. Phys. Chem. A 104 (2000)
[66] TDBHandHLYP/6-311G+(2d,2p) with PCM (CH2Cl2).
4256.
[67] C. Adamo, V. Barone, J. Chem. Phys. 110 (1999) 6158.
[10] M.K. Nazeeruddin, Coord. Chem. Rev. 248 (2004) 1161.
[68] A.D. Becke, J. Chem. Phys. 98 (1993) 5648.
[11] P.V. Kamat, M. Haria, S. Hotchandani, J. Phys. Chem. B 108 (2004) 5166.
[69] R.J. Magyar, S. Tretiak, J. Chem. Theory Comput. 3 (2007) 976.
[12] J. Bisquert, D. Cahen, G. Hodes, S. Ruehle, A. Zaban, J. Phys. Chem. B 108 (2004)
[70] J.L. Brédas, D. Beljonne, V. Coropceanu, J. Cornil, J. Chem. Rev. 104 (2004)
8106.
4971.
[13] A. Furube, R. Katoh, T. Yoshihara, K. Hara, S. Murata, H. Arakawa, M. Tachiya, J.
[71] R. Krishnan, J.S. Binkley, R. Seeger, J.A. Pople, J. Chem. Phys. 72 (1980) 650.
Phys. Chem. B 108 (2004) 12588.
[72] C. Amovilli, V. Barone, R. Cammi, E. Cancès, M. Cossi, B. Mennucci, C.S.
[14] G. Li, K.J. Jiang, Y.F. Li, S.L. Li, L.M. Yang, J. Phys. Chem. C 112 (2008) 11591.
Pomelli, J. Tomasi, Adv. Quant. Chem. 32 (1998) 227.
[15] M.K. Nazeeruddin, F. De Angelis, S. Fantacci, A. Selloni, G. Viscardi, P. Liska, S.
[73] J. Tomasi, B. Mennucci, R. Cammi, Chem. Rev. 105 (2005) 2999.
Ito, T. Bessho, M. Grätzel, J. Am. Chem. Soc. 127 (2005) 16835.
[74] For the geometry optimisation process, we turned on the following options (i)
[16] S.Z. Wang, Y. Cui, K. Hara, Y. Dan-Oh, C. Kasada, A. Shinpo, Adv. Mater. 19
Radii = UAKS which selects radii optimized at DFT level of theory for building
(2007) 1138.
the solute cavity and (ii) NoAddSph [61] which avoids the generation of
[17] B.M. Wong, J.G. Codaro, J. Chem. Phys. 129 (2008) 214703.
smoothing spheres on the cavity surface.
[18] K. Sayama, K. Hara, N. Mori, M. Satsuki, S. Suga, S. Tsukagochi, Y. Abe, H.
[75] A. Sen, P.E. Nielsen, Nucleic Acids Res. 35 (2007) 3367.
Sugihara, H. Arakawa, Chem. Commun. (2000) 1173.
[76] R. Katoh, A. Furube, T. Yoshihara, K. Hara, G. Fujihashi, S. Takano, S. Murata, H.
[19] T. Horiuchi, H. Miura, K. Sumioka, S. Uchida, J. Am. Chem. Soc. 126 (2004)
Arakawa, M. Tachiya, J. Phys. Chem. B 108 (2004) 4818.
12218.
[77] J.B. Asbury, Y.Q. Wang, E. Hao, H. Ghosh, T. Lian, Res. Chem. Intermed. 27
[20] K. Hara, T. Horiguchi, T. Kinoshita, K. Sayama, H. Sugihara, H. Arakawa, Sol.
(2001) 393.
Energy Mater. Sol. Cells 64 (2000) 115.
[78] A. Hagfeldt, M. Grätzel, Chem. Rev. 95 (1995) 49.
[21] E. Stathatos, P. Lianos, A. Laschewsky, O. Ouari, P. Van Cleuvenbergen, Chem.
[79] P.F. Barbara, T.J. Meyer, M.A. Ratner, J. Phys. Chem. 100 (1996) 13148.
Mater. 13 (2001) 3888.
[80] D. Matthews, P. Infelta, M. Grätzel, Sol. Energy Mater. Sol. Cells 44 (1996) 119.
[22] R. Chen, X. Yang, H. Tian, X. Wang, A. Hagfeldt, L. Sun, Chem. Mater. 19 (2007)
[81] F. De Angelis, S. Fantacci, A. Selloni, Nanotechnology 19 (2008) 424002.
4007.
[82] G. Benkö, J. Kallioien, J.E.I. Korppi-Tommola, A.P. Yartsev, V. Sundström, J. Am.
[23] C. Baik, D. Kim, M.S. Kang, K. Song, O.K. Sang, J. Ko, Tetrahedron 65 (2009)
Chem. Soc. 124 (2002) 489.
5302.
[83] S. Iwa, K. Hara, S. Murata, R. Katoh, H. Sugihara, H. Arakawa, J. Chem. Phys.
[24] S. Ferrere, A. Zaban, B. Gregg, J. Phys. Chem. B 101 (1997) 4490.
113 (2000) 3366.
[25] S. Ferrere, B. Gregg, New J. Chem. 26 (1997) 1155.
[84] C. Bauer, G. Boschloo, E. Mukhtar, A. Hagfeldt, Int. J. Photochem. 4 (2002) 17.
[26] D. Liu, R.W. Fessenden, G.L. Hug, P.V. Kamat, J. Phys. Chem. B 101 (1997) 2583.
[85] R.J. Cave, E.W. Castner Jr., J. Phys. Chem. A 106 (2002) 12117.
[27] B. Burfeindt, T. Hannappel, W. Storck, F. Willig, J. Phys. Chem. 100 (1996)
[86] R.J. Cave, K. Burke, E.W. Castner Jr., J. Phys. Chem. A 106 (2002) 9294.
16463.

Please cite this article in press as: J. Preat et al., Chem. Phys. (2010), doi:10.1016/j.chemphys.2010.08.001
J. Preat et al. / Chemical Physics xxx (2010) xxx–xxx 13

[87] P. Carbonniere, T. Lucca, C. Pouchan, N. Rega, V. Barone, J. Comput. Chem. 26 agreement with the experimental value, subsequently giving a theoretical
(2005) 384. DGinject: (2.35 eV) closer to the experimental value (1.51 eV).
[88] R. Improta, V. Barone, G. Scalmani, M.J. Frisch, J. Chem. Phys. 125 (2006) [97] This assumption is valid as long as the injection is restricted to energy levels
054103. close to the conduction band edge, and is combined to a density of acceptor
[89] The MOA has been performed with Molekel (version 5.4.0.8) of the Swiss states constant in the same energy range.
National Supercomputing Centre (CSCS). [98] D. Cahen, G. Hodes, M. Grätzel, J.F. Guillermoles, I. Riess, J. Phys. Chem. B 104
[90] H. Tian, X. Yang, R. Chen, R. Zhang, A. Hagfeldt, L. Sun, J. Phys. Chem. C 112 (2000) 2053.
(2008) 11023. [99] H.S. Nalwa, Handbook of Advanced Electronic and Photonic Materials and
[91] J. Preat, D. Jacquemin, V. Wathelet, J.M. André, E.A. Perpète, Chem. Phys. 335 Devices, Academic, San Diego, CA, 2001. pp. 1–3366.
(2007) 177. [100] B.H. Besler, K.M. Merz, P.A. Kollman, J. Comput. Chem. 11 (1990) 431.
[92] C. Loison, R. Antoine, M. Broyer, J. Dugroud, J. Guthmuller, D. Simon, Chem. [101] B. Peng, S. Yang, L. Li, F. Cheng, J. Chen, J. Chem. Phys. 132 (2009) 034305.
Eur. J. 14 (2008) 7351. [102] K. Andersson, P.-A. Malmqvist, B.O. Roos, J. Chem. Phys. 96 (1992) 1218.
[93] D. Guillaumont, S. Nakamura, Dyes Pigm. 46 (2000) 85. [103] P.J. Bruna, S.D. Peyerimhoff, R. Buenker, J. Chem. Phys. Lett. 72 (1980) 278.
[94] Q. Li, X.X. Zhang, Y. Meng, M.B. Chen, Chin. Chem. Lett. 16 (2005) 693. [104] D. Jacquemin, M. Bouhy, E.A. Perpète, J. Chem. Phys. 126 (2006) 204321.
[95] W. Xu, B. Peng, J. Chen, M. Liang, F. Cai, J. Phys. Chem. C 112 (2008) 874. [105] Y. Bai, Y. Cao, J. Zhang, M. Wang, R. Li, P. Whang, S.M. Zakeeruddin, M. Grätzel,
[96] For THQ-9, the excitation energy is overestimated by BHandHLYP, whereas Nature Mater. 7 (2008) 626.
Edye
OX is underestimated. As previously shown, PBE0 provides a kmax
THQ -9
in better

Please cite this article in press as: J. Preat et al., Chem. Phys. (2010), doi:10.1016/j.chemphys.2010.08.001

You might also like