Azulene – Thiophene – Cyanoacrylic acid dyes with donor-π-acceptor structures. Synthesis, characterisation and evaluation in dye-sensitized solar cells

Article abstract of DOI:10.1016/j.tet.2018.04.043

We report the synthesis of five new azulene containing dyes, having D-π-A type structures. These dyes are synthesised using a sulfonium salt cross-coupling reaction. The dyes have been evaluated spectroscopically, electrochemically, crystallographically, and as sensitizers in dye-sensitized solar cells. We propose a rationale for the dyes’ spectroscopic properties and performance in cells, based on conformational data derived from their crystal structures.

Full text of DOI:10.1016/j.tet.2018.04.043

Accepted Manuscript
Azulene – Thiophene – Cyanoacrylic acid dyes with donor-π-acceptor structures.
Synthesis, characterisation and evaluation in dye-sensitized solar cells
Paul Cowper, Adam Pockett, Gabriele Kociok-Köhn, Petra J. Cameron, Simon E.
Lewis
PII:
S0040-4020(18)30432-0
10.1016/j.tet.2018.04.043
TET 29458
DOI:
Reference:
To appear in: Tetrahedron
Received Date: 8 March 2018
Revised Date: 12 April 2018
Accepted Date: 16 April 2018
Please cite this article as: Cowper P, Pockett A, Kociok-Köhn G, Cameron PJ, Lewis SE, Azulene –
Thiophene – Cyanoacrylic acid dyes with donor-π-acceptor structures. Synthesis, characterisation and
evaluation in dye-sensitized solar cells, Tetrahedron (2018), doi: 10.1016/j.tet.2018.04.043.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT
Azulene – thiophene – cyanoacrylic acid dyes with donor-π-acceptor
structures. Synthesis, characterisation and evaluation in dye-sensitized solar
cells
a
b
c
a,b
Paul Cowper *, Adam Pockett , Gabriele Kociok-Köhn , Petra J. Cameron and Simon E.
a,b
Lewis *
a
Department of Chemistry, University of Bath, Bath, BA2 7AY, UK. Tel: +44 (0)1225 386568. E-mail:
p.cowper@bath.ac.uk; s.e.lewis@bath.ac.uk
b
Centre for Doctoral Training in Sustainable Chemical Technologies, University of Bath, Bath, BA2 7AY, UK.
c
Chemical Characterization and Analysis Facility, University of Bath, Bath, BA2 7AY, UK.
Graphical Abstract
Abstract
We report the synthesis of five new azulene containing dyes, having D-π-A type structures. These
dyes are synthesised using a sulfonium salt cross-coupling reaction. The dyes have been evaluated
spectroscopically, electrochemically, crystallographically, and as sensitizers in dye-sensitized solar
cells. We propose a rationale for the dyes’ spectroscopic properties and performance in cells, based
on conformational data derived from their crystal structures.
1
. Introduction
1-2
Since the earliest reports on dye-sensitized solar cells (DSSCs) over 25 years ago, much effort has
been expended in the optimisation of these devices, in pursuit of higher photoelectric conversion
efficiencies (PCEs, η). Central to this effort has been the design, synthesis and evaluation of dyes
with enhanced properties in terms of photon capture, electron injection, dye adhesion and device
3
-4
5
lifetimes, etc. Dyes consisting of ruthenium polypyridyl complexes such as N719 achieved notably
high PCEs, but the high cost of ruthenium is an impediment to the commercialisation of cells based
1

ACCEPTED MANUSCRIPT
on such dyes. In contrast, DSSCs that use wholly organic dye molecules can avoid the use of
expensive metals. Design principles for high efficiency organic dyes have been established, with the
6-7
donor–π linker–acceptor–anchor (D-π-A) assembly proving to be highly successful. Of the various
8-10
known electron donor motifs, arylamines have proven especially effective. Regarding the π linker
motif, a wide variety have been explored, mostly heterocycles, amongst which thiophenes,
oligothiophenes and annulated thiophenes have all proven often to influence positively the overall
11-16
cell efficiency.
An anchoring group is required to immobilise the dye molecule on the metal oxide
surface; a variety of different anchor groups have been investigated, including phosphonic acids and
17,18
carboxylic acids and their derivatives (esters, acid chlorides, carboxylate salts or amides).
Carboxylic acid and cyanoacrylic acid groups are most frequently used for DSSC dyes. More
specifically, because of its excellent electron withdrawing characteristics, cyanoacrylic acid is used in
most D-π-A organic dyes, where it fulfils the dual role of acceptor and anchor. In contrast, simple
carboxylic acid anchors are usually used for transition metal complexes where metal-to-ligand-
charge-transfer (MLCT) predominates.
Figure 1: Structures of azulene and naphthalene; their HOMOs and LUMOs.
The aims of the present study were to synthesise organic dyes for DSSCs that comprise an
azulene ring as the donor motif, to characterise these dyes using a variety of techniques, and to
evaluate their performance in solar cells. Azulene (1) is a bicyclic, nonbenzenoid aromatic
hydrocarbon and is an isomer of naphthalene (2). Azulenes have long been of interest to chemists
2

ACCEPTED MANUSCRIPT
due to their striking colours, interesting chemistry and unusual electronic properties. Azulene
19-25
26
derivatives may be used as advanced materials for optoelectronic
and electrochromic devices,
27-29
30
31
charge-transport,
nonlinear optics and chemical sensing; some recent reviews serve to
32-33
summarise these areas.
Some azulene derivatives have also been shown to have anti-
34-35
inflammatory,
36-38
39-42
43
antiulcer,
anticancer
and anti-HIV properties. Azulene is a non-alternant
hydrocarbon, unlike naphthalene, which leads to HOMO nodes at C2 and C6 and LUMO (S ) nodes at
1
C1 and C3 (Figure 1), causing relatively small repulsion between the electrons occupying these
4
4
orbitals, meaning a low S – S transition energy, in the visible region. With naphthalene 2, the
0
1
coefficient magnitude at different carbons is the same for the HOMO and LUMO, giving a higher
degree of overlap, and the electrons occupy the same space with consequent higher repulsion and
higher transition energy in the UV. Therefore naphthalene is colourless but azulene 1, because the
S – S transition corresponds to the green and red regions of the spectrum, is a deep blue colour.
0
1
The absorption spectra for substituted azulenes are very sensitive to the electronic nature of the ring
45
substituents. Hence, by careful design of the substituents, the entire visible (and IR) spectrum is
accessible. This ability to tune the chromophore’s absorption spectrum makes azulene an attractive
motif for exploitation in DSSCs.
In the context of solar cells, sporadic reports of azulene derivatives have appeared in various
contexts. Emrick and co-workers employed copolymers of 2-substituted azulene with zwitterionic
46
methacrylates in bulk-heterojunction (BHJ) solar cells. Imahori and co-workers evaluated
47
copolymers of 1,3-disubstituted azulene and diketopyrrolopyrroles or benzothiadiazoles in BHJ
solar cells, shortly after which and Liu, Zhang and co-workers reported two similar azulene
copolymers, as well as a further example possessing distinct connectivity (a 4,7-disubstituted
48
azulene). Most recently, Gao and co-workers reported a copolymer incorporating a 6,6’-
49
disubstituted 2,2’-biazulenyl. Aside from polymer BHJ solar cells, Huang and co-workers recently
50
reported the use of an azulene-squaraine small molecule in an organic photovoltaic (OPV) device.
Torres, Guldi and co-workers reported an OPV device comprising azulenocyanines immobilised on
3

ACCEPTED MANUSCRIPT
51
carbon nanotubes. In the area of perovskite solar cells, an azulene-containing hole-transport
material has been shown by Wakamiya, Saeki, Scott, Murata and co-workers to exhibit performance
52
superior to that of Spiro-OMeTAD, the most commonly used hole-transport material. There have
been reports of electron injection from an azulene into the conduction band of titania, from
53,54
Piotrowiak, Galoppini and co-workers
(either employing azulene itself, sequestered in a
hemicarceplex, or using azulenes bearing methyl ester anchoring groups), but in no instance was
construction of a DSSC attempted.
To our knowledge, the peer-reviewed literature contains only two reports of azulene
containing dyes being used in DSSCs. In 2007, Cheng, Wang, Zhang and co-workers reported the
synthesis of dyes 3-6 (Figure 2), comprising either an azulene attached directly to a cyanoacrylic acid
5
5
(
3, 5), or with a vinylene linker between the azulene and the anchor motif (4, 6). They found the
presence of the vinylene linker led to a bathochromic shift of λmax and an increase in η with respect
to the dyes lacking this motif, due to increases in both JSC and V . Additionally, they observed an
OC
increase in η for alkylated azulenes 5 and 6 compared to their respective non-alkylated counterparts
3
and 4. The authors ascribe this latter effect to the alkyl groups disfavouring dye aggregation on the
titania surface. In 2014, Hanaya and co-workers reported the evaluation of phenylazulene 7 (Figure
5
6
2
) as a dye in a DSSC. The focus of this study was the evaluation of an alternative redox mediator,
–
–
–
–
–
–
Br /Br , in place of the more commonly used I /I system, with the authors finding that the Br /Br
3
3
3
system afforded a greater VOC and hence η. The patent literature also contains azulene dyes for
57,58
DSSCs. Two patents filed by TDK
describe azulenes 8-17 (Figure 2). The emphasis of the patent is
on the fact that these dyes are all more resistant to photobleaching than a comparator ruthenium-
59
based dye; no further data are given. A patent from Solvay reports azulenocyanines of type 18
5
1
(
Figure 2), which differ from those reported previously by Torres, Guldi and co-workers in that the
azulenocyanines bear anchoring groups for the specific purpose of attachment to titania and use in a
DSSC. (Azulenes are also described elsewhere in the patent literature in the broader context of
4

ACCEPTED MANUSCRIPT
60,61
photovoltaics. ) Finally it should be noted that prospective azulene-based dyes for DSSCs have
62
been considered theoretically.
Figure 2: Previously reported azulene-containing dyes used in DSSCs.
5

ACCEPTED MANUSCRIPT
In the present study, we undertook two rounds of dye design, synthesis and characterisation. In the
first instance we targeted dyes of the D-π-A design, with a 1-substituted azulene as the donor motif
(
as per the majority of examples in Figure 2), and a cyanoacrylic acid as the acceptor and anchor
motif. For the π-linker, we opted to employ a thiophene ring, whose ability to induce bathochromic
11-
shifts of absorption maxima and to afford increased extinction coefficients has been documented.
16
Specifically we targeted azulene-thiophene-cyanoacrylic acid (“Az-tcaa”) 19 and 4,6,8-
trimethylazulene-thiophene-cyanoacrylic acid (“TMAz-tcaa”) 20 (Figure 3). The rationale for
targeting 20 was that the methyl substituents may disfavour dye aggregation on the titania surface,
55
analogous with the reason proposed previously for the increased performance of 6 with respect to
4
. Retrosynthetically, it was envisaged that the targets could be accessed by a Knoevenagel
condensation of aldehyde 22 and cyanoacetic acid. Aldehyde 22 would be accessed in turn using
cross-coupling methodology, from an appropriately functionalised azulene 23 and a thiophene-2-
carbaldehyde metallated at the 5-position, 24. The C1 and C3 positions of azulene are the most
nucleophilic, so a haloazulene of type 23 would be expected to be accessible from the parent
azulene (1 or 25) through an S Ar halogenation.
E
6

ACCEPTED MANUSCRIPT
st
Figure 3: Targeted 1 generation dyes and retrosynthesis.
Of the various cross-coupling methodologies that could be used to synthesise 22, we opted to
employ a Suzuki–Miyaura reaction, since there is precedent for such couplings using 1-haloazulenes
6
3
64
as the substrate, in conjunction with bulky monodentate phosphine ligands. The required
nucleophilic reaction partner of type 24 was therefore (5-formylthiophen-2-yl)boronic acid (29). This
65-66
was synthesised through a variant of reported procedures,
as shown in Scheme 1a. We found
that 29 could be conveniently purified by recrystallization from water, although the polymorph so
formed was shown by x-ray crystallography to be a monohydrate (see ESI). This water of
crystallisation could subsequently be removed simply by drying the ethereal solution of 29•H O over
2
MgSO4.
Scheme 1: Synthesis of cross-coupling partners and first Suzuki–Miyaura coupling.
An azulene cross-coupling partner was prepared by direct halogenation of azulene (1) with N-
67
chlorosuccinimide to give 1-chloroazulene (30). Clean monohalogenation of azulene through an
SEAr process is not possible, since the resultant 1-haloazulene can readily undergo a second
substitution. Thus, even with only 1 equivalent of NCS, formation of a small quantity of 1,3-
dichloroazulene (31) is unavoidable (Scheme 1b). Haloazulenes are also rather unstable, but 30 is
sufficiently stable to be separable from 31 by chromatography. With both coupling partners in hand,
64
we attempted their union using reaction conditions described by Buchwald. In the first instance
7

ACCEPTED MANUSCRIPT
(
Scheme 1c), however, thienyl azulene 32 was produced in only 6% yield. We ascribe this low yield
primarily to instability of boronic acid 29 to the reaction conditions, as we observed significant
quantities of protodeborylation product (thiophene-2-carbaldehyde) being formed. Decomposition
of 30 may also have contributed to the low yield. Use of the bromo analogue of 30 was considered
but not attempted, since although it would be more reactive, it is appreciably more unstable.
Attempts at reaction optimisation (variation of solvent, reaction stoichiometry, etc.) afforded only
marginal improvements in yield.
A significant improvement in the cross-coupling yield was achieved by varying the
nucleophilic coupling partner, employing a pinacolborane group in place of the boronic acid, and
with the aldehyde protected as an acetal. Thus, coupling of 30 with dioxaborolane 33 gave protected
thienylazulene 34 (which was not purified); immediate deprotection by transketalisation in acetone
gave desired aldehyde 32 in a much-improved 54% yield over 2 steps (Scheme 2a). The analogous
reaction sequence with 4,6,8-trimethylazulenyl halide 35 gave the corresponding aldehyde 37 in
5
9% (Scheme 2b). Finally, the yields were improved still further by the use of azulenesulfonium salts
3
8-39 as pseudohalides in the cross-coupling (Scheme 2c), methodology recently developed by us to
68
address the issue of the instability of azulenyl halides.
Scheme 2: Improvements to the cross-coupling
8

ACCEPTED MANUSCRIPT
With the required aldehydes 32 and 37 in hand, we carried out the Knoevenagel condensation to
furnish the desired dyes 19 and 20 respectively, as shown in Scheme 3. Literature dye 3 was also
55,69
synthesised
for spectroscopic comparison with 19.
Scheme 3: Knoevenagel condensations.
Absorption spectra for 3, 19 and 20 were acquired in acetonitrile (see ESI), and key spectroscopic
data are shown in Table 1. The spectra have similar features, such as an absorption in the 400-500
nm region, a region where corresponding the unsubstituted parent azulene compounds do not
absorb. When comparing literature dye 3 with our dye 19 incorporating a thiophene π-linker, a
bathochromic shift of λmax of 50 nm is evident.
Dye
λmax
nm
ε
λonset
nm
E (opt)
g
ꢀ1
ꢀ1
L mol cm
eV
4
Az-1-caa (3)
423
473
452
3.952 x 10
3.026 x 10
1.914 x 10
460
547
547
2.70
2.27
2.27
4
Az-1-tcaa (19)
TMAz-1-tcaa (20)
4
Table 1. UV/Vis data for 1st generation dyes
Cyclic voltammetry (CV) was then used to determine the oxidation potentials of the three azulene
based dyes: Az-1-caa (3), Az-1-tcaa (19), and TMAz-1-tcaa (20) (Figures 4-6). All CVs were run in dry
THF with 0.1 M tetrabutylammonium hexafluorophosphate (Bu NPF ) supporting electrolyte at a
4
6
-
1
scan rate of 100 mV s (see ESI for full details). Three scans were taken for each run. The
voltammograms show plots for the first and third scans only. All the dyes exhibited at least one
oxidation peak; however the oxidations were all irreversible so it was not possible to calculate the E½
9

ACCEPTED MANUSCRIPT
+
potentials. The peak oxidation potentials vs. Fc/Fc are summarised below (Table 2). The position of
the HOMO is estimated from the onset of the oxidation wave in the irreversible scans.
Dye
Epa1
mV
Ipa1
ꢁA
Epa2
mV
Ipa2
ꢁA
Eox
HOMO
mV
V vs NHE
Az-1-caa (3)
837
455
455
18.04
15.79
9.22
-
-
-
-
670
310
320
1.30
0.94
Az-1-tcaa (19)
TMAz-1-tcaa (20)
566
8.88
0.95
+
Table 2. First generation dyes’ electrochemical data in THF (E/V vs. Fc/Fc )
Figure 4. Voltammogram for Az-1-caa (3), 0.5 mM in THF. Supporting electrolyte 0.1 M Bu NPF . Scan rate 100
4
6
–
1
mV s .
1
0

ACCEPTED MANUSCRIPT
Figure 5. Voltammogram for Az-1-tcaa (19), 0.5 mM in THF. Supporting electrolyte 0.1 M Bu NPF . Scan rate
4
6
–
1
1
00 mV s .
Figure 6. Voltammogram for TMAz-1-tcaa (20), 0.5 mM in THF. Supporting electrolyte 0.1 M Bu NPF . Scan
4
6
–
1
rate 100 mV s .
1
1

ACCEPTED MANUSCRIPT
The dye with three methyl groups on the azulene (20) has a similar oxidation potential to the
unsubstituted analogue 19. Furthermore, the literature dye without a thiophene linker (3) has a
significantly elevated oxidation potential compared with the dyes with a thiophene linking unit (19
and 20). This suggests that the thiophene linker makes it easier to extract an electron from the dye
molecule and that the resulting cation should be more stable. Also of note, 20 exhibited two
oxidation waves, neither being reversible (Figure 6). This means that two electrons can be removed
from this molecule, resulting in a dication. Crucially, all of the voltammograms show that breakdown
products are generated following oxidation of the dyes. This can be seen by comparing the traces of
scan 1 (black line) and scan 3 (red line) in each voltammogram. With scan 1, there is no sign of any
oxidation before the main oxidation peak begins, e.g. at approx. 0.3 V in Figure 6. On subsequent
scans, there is evidence of oxidation processes occurring at potentials lower than the main oxidation
+
peak, e.g. in Figure 6 two minor oxidation peaks can be seen at -0.049 V and 0.263 V vs Fc/Fc .
These minor oxidation waves appear to be at least partially reversible since they have corresponding
reduction waves. Since the minor oxidation waves did not occur in the first scan, they must be due
to species generated from the oxidised dye, which is a radical cation. Such a radical cation would be
highly reactive and may react with other molecules present e.g. solvent, other dye molecules
(
hydrogen abstraction) or other radicals (combination).
The cyclic voltammetry data discussed above suggest these dyes may have a limited lifetime
in a DSSC. We were mindful of the fact that a characteristic of successful ruthenium polypyridyl dyes
70
is that they are known to exhibit reversible electrochemistry. Therefore, the objective of the
second phase of this work was to synthesise dyes which exhibited reversible electrochemistry, which
through modifications to Az-1-tcaa (19) and TMAz-1-tcaa (20) were rationally designed to stabilise
their oxidised states.
A literature search for azulene-containing compounds with reported electrochemical data,
and that exhibit reversible oxidation waves, has suggested that both the azulene C-1 and C-3
1
2

ACCEPTED MANUSCRIPT
positions must be substituted, preferably with substituents with steric bulk e.g. tert-butyl groups.
Gerson et al. have reported that only azulenyl radical cations substituted at both the C-1 and C-3
positions were sufficiently persistent to be characterised by EPR spectroscopy, and tert-butyl groups
71
were necessary to enable ENDOR and TRIPLE resonance studies. The radical cations of 1,3-di-tert-
butylazulenes 40 (Figure 7) persisted for 20-40 minutes at 273-283 K. The authors offered no
explanation as to why the tert-butyl groups should confer stability to the azulene radical cation. 1,3-
Dimethylazulene radical cations were more stable than those of unsubstituted azulene, but were not
so long-lived as the 1,3-di-tert-butylazulene ones. It is postulated that the tert-butyl groups shield
the radical unpaired electron purely through their steric bulk, preventing the close approach of
species that would otherwise be liable to react with the radical. Shoji, Ito, Morita and co-workers
72
have reported that 1,3-bis(methylthio)azulenes 41 also exhibit reversible electrochemistry. The
methylthio groups may also stabilise the corresponding radical cations through steric shielding;
conjugation with the sulfur atoms may also be important.
Figure 7: Reported classes of azulenes exhibiting reversible electrochemistry.
We resynthesized 1,3-di-tert-butylazulene and 1,3-bis(methylthio)azulene and verified that they
both do exhibit reversible electrochemistry (ESI, Figures S4-S5). An azulenyl nitrone reported by
73
Becker, Echegoyen and co-workers to exhibit reversible electrochemistry was also resynthesized.
While this did not display evidence of decomposition products, nor did it display fully reversible
electrochemistry in our hands, as the amplitude of the oxidation wave decreased in the second and
subsequent scans (ESI, Figures S6-S7). These literature examples of azulenes that exhibit reversible
electrochemistry inspired a second generation of dye designs, incorporating both the azulene-
thiophene-cyanoacrylic acid D-π-A motif in dyes 19 and 20, and also the C3 substituents of 40 and
4
1. It was hoped that such dyes would display fully reversible electrochemistry, whilst at the same
1
3

ACCEPTED MANUSCRIPT
time retaining the D-π-A motif and hence favouring electron injection from the excited state. The
second generation target dyes 42-44 are shown in Figure 8.
Figure 8: Second generation azulene D-π-A dye targets.
Of the dye targets 42-44, we first sought to synthesise 42, bearing a tert-butyl group. 1,3-Di-tert-
butylazulene (40, R = R2 = H) can be synthesised directly from azulene by means of a double Friedel–
74
Crafts alkylation, using excess tert-butanol and HBF4. However, attempts to use this procedure to
introduce a tert-butyl group directly onto the C3 position of aldehyde 32 were unsuccessful, most
likely because the functionality present in this precursor to dye 19 was incompatible with the harsh
Friedel–Crafts reaction conditions. Instead, therefore, we opted to synthesise 42 via a reaction
sequence analogous to that for 19, but starting with 1-tert-butylazulene 45 (i.e. a synthesis with the
tert-butyl group present from the beginning). However, accessing pure 1-tert-butylazulene (45) is
not straightforward. Treatment of azulene with only one equivalent of tert-butanol and HBF4
2
unavoidably leads to over-substitution, i.e. formation of some 1,3-di-tert-butylazulene (40, R = R =
H) in addition to the desired 1-tert-butylazulene (45); in our hands these two products proved to be
difficult to separate by chromatography. To overcome this problem, we devised a new synthesis of
2
1
-tert-butylazulene. It is known that 1,3-di-tert-butylazulene (40, R = R = H) undergoes ipso
75
substitution upon exposure to Vilsmeier conditions, giving 1-formyl-3-tert-butylazulene.
1
4

ACCEPTED MANUSCRIPT
Furthermore, there are reported examples of 1-formylazulenes undergoing deformylation upon
76,77
treatment with pyrrole.
Exploiting these literature reports, we were able to synthesise pure 1-
tert-butylazulene (45) as shown in Scheme 4. Double Friedel–Crafts alkylation of azulene gives 1,3-
di-tert-butylazulene, which then undergoes substitution of a tert-butyl group for a formyl group to
74,75
give 46, as per the literature procedures.
Application of the pyrrole-mediated azulene
deformylation protocol to 46 has not been reported previously, but in the event proceeded
smoothly to give the desired 45.
Scheme 4: New synthesis of 1-tert-butylazulene.
With a reliable route to 45 established, we utilised our sulfonium salt cross-coupling methodology to
elaborate this towards dye target 42. Treatment of 45 with tetramethylenesulfoxide 47 and
trifluoroacetic anhydride led to azulenesulfonium salt 48 (after anion exchange). This in turn
underwent Suzuki–Miyaura cross-coupling followed by acetal deprotection to give aldehyde 50 (the
68
tert-butyl analogue of 32), as per our reported procedure (Scheme 5).
Scheme 5: Synthesis of aldehyde 50, precursor to dye 42.
To synthesise methylthio-substituted dyes 43-44, we were able to employ a more convergent route,
introducing the desired –SMe substituent directly onto the azulene 3-position of thienyl aldehydes
1
5

ACCEPTED MANUSCRIPT
3
2 and 37, that we had prepared previously. According to the procedure of Shoji, Morita and co-
workers, exposure of an azulene to a sulfoxide and an activating agent leads to formation of an
78
azulenesulfonium salt. In many instances, trifluoroacetic anhydride (TFAA) will suffice as the
79
activating agent, but for more electron-poor azulenes (or unactivated arenes ),
trifluoromethanesulfonic anhydride (Tf O) is used. Thus, as shown in Scheme 6, the most
2
nucleophilic position in azulenes 32 and 37 is still the azulene C3 position, regardless of the presence
of the thiophene, and it is this position that reacts with the activated sulfoxide to give
azulenesulfonium salts 51-52. Whereas we have exploited azulenesulfonium salts for cross coupling
(
c.f. Schemes 2 and 5), Shoji, Morita and co-workers have demonstrated that they may be readily
dealkylated to give the corresponding thioethers. In our case, 51 and 52 were not purified, but were
instead directly subjected to demethylation conditions to furnish the desired aldehydes 53 and 54.
Scheme 6: Direct late-stage C3-functionalisation of thienyl aldehydes.
With aldehydes 50, 53 and 54 in hand, access to target dyes 42-44 required only the final
Knoevenagel condensation with cyanoacrylic acid. Application of the Knoevenagel conditions
employed for the first-generation dyes (catalytic piperidine, reflux in acetonitrile) did lead to
formation of the target dyes, but the presence of appreciable amounts of residual starting material
and other products hampered their purification. The first generation dyes were purified in a
straightforward fashion by trituration with dichloromethane, followed by recrystallization, but this
protocol was not applicable to the second generation dyes – their greater solubility in
dichloromethane precluded effective trituration, and direct recrystallization of the crude was not
successful. We therefore sought to improve the Knoevenagel condensation to drive the reaction to
1
6

ACCEPTED MANUSCRIPT
completion and reduce the proportion of impurities present in the crude that were interfering with
recrystallization. Aside from piperidine, many other promoters of the Knoevenagel reaction have
been reported in the literature. Of various literature procedures we tried, use of a stoichiometric
80-81
quantity of 6-aminocaproic acid in place of piperidine was most beneficial.
Thus, after
optimisation, we found that treatment of 50 with 1 equivalent of 6-aminocaproic acid and 1.5
equivalents of cyanoacrylic acid, in ethanol at 50 °C, led to consumption of starting material by TLC
after 3.5 hours. Addition of 1M H SO
then gave a tar, which was dissolved in dichloromethane
4(aq)
2
and washed with water; removal of solvent was followed by a successful recrystallization to give
target dye 42 in 49% yield. The same procedure, when applied to 53 and 54 gave a precipitate upon
addition of acid, which could simply be isolated by filtration and recrystallized to provide 43 (67%
yield) and 44 (63% yield) respectively, as shown in Scheme 7.
Scheme 7: Final Knoevenagel condensations to produce second generation dyes.
With the target second generation dyes 42-44 in hand, these were characterised by the same
techniques employed for 3, 19 and 20. UV-Vis spectra were acquired at 0.01, 0.03 and 0.05 mM
concentrations. During the dye synthesis and purification steps, it had been noted that the colour of
the dyes in solution differed slightly depending on the solvent used (solvatochromism). When in
dichloromethane solution, for instance, the dyes appeared to be somewhat bluer than when
acetonitrile was used. Spectra were therefore acquired in three solvents: acetonitrile (MeCN),
tetrahydrofuran (THF) and dichloromethane (DCM) (see ESI, Figures S8-S19.). The UV-Vis data for
nd
the 2 generation dyes are presented in Table 3.
1
7

ACCEPTED MANUSCRIPT
Dye
Solvent
MeCN
THF
λmax
nm
ε (λmax
)
λonset
nm
Eg (opt)
eV
ꢀ1
ꢀ1
L mol cm
4
tBuAz-1-tcaa
A: 488
B: 351
4.12 x 10
567
556
586
515
563
592
597
585
628
2.19
2.23
2.12
2.41
2.20
2.09
2.08
2.12
1.97
4
(
42)
2.06 x 10
4
A: 487
B: 353
3.57 x 10
4
1.57 x 10
4
DCM
A: 512
B: 358
3.97 x 10
4
1.68 x 10
4
MeTAz-1-tcaa
43)
MeCN
THF
A: 436
B: 357
1.99 x 10
4
(
1.57 x 10
4
A: 478
B: 357
2.57 x 10
4
1.52 x 10
4
DCM
A: 497
B: 363
2.56 x 10
4
1.43 x 10
4
MeTTMAz-1-tcaa
44)
MeCN
THF
A: 482
B: 367
1.15 x 10
4
(
2.47 x 10
4
A: 475
B: 362
1.15 x 10
4
2.63 x 10
4
DCM
A: 506
B: 377
1.13 x 10
4
2.62 x 10
nd
Table 3. UV/Vis data for 2 generation dyes. Extinction coefficients quoted are mean
values of the coefficients calculated for the three dye concentrations.
nd
The three 2 generation dyes all have two principal absorption bands in the region of interest, one
in the visible region with λmax ≈ 430 – 520 nm (band A) and the other in the near UV with λmax ≈ 350 –
3
80 nm (band B). Dye MeTTMAz-1-tcaa (44) has an additional absorption band with λmax ≈ 320 nm.
The band A absorption is assumed to be due to charge transfer between the thiophene linker and
the azulene moiety. With dyes tBuAz-1-tcaa (42) and MeTAz-1-tcaa (43) (i.e. those with an
unsubstituted azulene 7-membered ring), the absorption band A is the more intense of the two.
However, with dye MeTTMAz-1-tcaa (44), the situation is reversed, with the absorption band A
1
8

ACCEPTED MANUSCRIPT
being less intense than the absorption band B. Indeed, absorption band A for MeTTMAz-1-tcaa (44)
is only around half the intensity of absorption band A for MeTAz-1-tcaa (43). This is an unexpected
result at first glance, since the only difference between these two dyes is the presence of the
additional methyl groups in 44, which do not extend the chromophore. Upon reinspection of the UV-
vis data for the first generation dyes, the same effect can be discerned. The visible-region absorption
band for TMAz-1-tcaa (20) is significantly less intense than for Az-1-tcaa (19). Once again, the only
difference in structure is the presence of the additional methyl groups in 20.
A possible explanation for the phenomena described above may be found in x-ray
crystallographic data acquired for the dyes. Crystals suitable for x-ray diffraction were grown for
dyes 20, 42 and 43. The structures are depicted in Figures 9-11.
Figure 9. (A) ORTEP structure of TMAz-1-tcaa (20) and (B) rotated through 90° in the horizontal plane
to show the "end elevation" view. Ellipsoids are represented at 50% probability. H atoms are shown
as spheres of arbitrary radius. CCDC #1815403.
1
9

ACCEPTED MANUSCRIPT
Figure 10. (A) ORTEP structure of tBuAz-1-tcaa (42) and (B) rotated through 90° in the horizontal
plane to show the "end elevation" view. Ellipsoids are represented at 50% probability. H atoms are
shown as spheres of arbitrary radius. CCDC #1815402.
Figure 11. (A) ORTEP structure of MeTAz-1-tcaa (43) and (B) rotated through 90° in the horizontal
plane to show the "end elevation" view. Ellipsoids are represented at 50% probability. H atoms are
shown as spheres of arbitrary radius. CCDC #1815401.
In the case of MeTAz-1-tcaa (43), the structure is essentially planar (Figure 11), with a torsion angle
between the azulene and thiophene rings (defined as Az(C2)-Az(C1)-Thio(C2)-Thio(C3)) of θ = 0.9
(
4)°. As such, it seems optimal orbital overlap is achievable in the solid state (and presumably also in
solution) and the two rings are fully conjugated. In the case of tBuAz-1-tcaa (42), the solid-state
structure exhibits some distortion away from planarity (Figure 10), with the analogous torsion angle
being θ = 26.8 (3)°. However, most strikingly of all, in the case of TMAz-1-tcaa (20) the structure is
far from planar (Figure 9), with the analogous torsion angle being θ = 122.2 (3)°, i.e. the azulene and
thiophene rings are closer to being perpendicular than they are to being coplanar. Rather than this
being due to any crystal packing effects, we propose this is an unavoidable consequence of the
presence of the methyl groups on the azulene 7-membered ring. Thus, if the azulene and thiophene
rings were coplanar, the azulene 8-position methyl substituent would be significantly occluded with
the thiophene C3 hydrogen. Relief of steric strain therefore necessitates that the thiophene ring
rotates out of planarity with the azulene, which will impair the conjugation of the two rings. Since
2
0

ACCEPTED MANUSCRIPT
the visible region absorption band is proposed to arise from charge transfer between the azulene
and thiophene rings for these dyes, the reduced π-orbital overlap between these two motifs may
therefore explain the relatively lower extinction coefficient observed for TMAz-1-tcaa (20) in
comparison with the analogue Az-1-tcaa (19) which lacks the methyl groups. In the case of
MeTTMAz-1-tcaa (44), although we were not able to obtain an x-ray crystal structure, we can
nevertheless propose that the same effect is in operation, leading to the less intense absorption in
the visible region with respect to MeTAz-1-tcaa (43) that lacks the methyl groups (and which can be
seen in Figure 11 to be fully planar).
nd
Cyclic voltammograms (CVs) were acquired for the three 2 generation dyes 42-44. All three
dyes show quasi-reversible redox chemistry (Figures 12-14). The radical cations produced by dye
molecule oxidation clearly survive to be reduced back to the dye molecules as evidenced by the
reduction waves in the reverse sweep which correspond to the oxidation waves in the forward
sweep of the CVs. There is no sign of any decomposition products evident in the second and
subsequent scans. For all three dyes, the second and subsequent CV scans are almost exactly super-
imposed (scan 1 and scan 3 are plotted for each dye), which indicates that there is little or no
depletion of the dye species in the vicinity of the working electrode. This is in contrast with the first
generation dye analogues (Figures 4-6), where the magnitude of the oxidation waves diminish with
each successive scan indicating that the dyes are being consumed.
2
1

ACCEPTED MANUSCRIPT
Figure 12. Voltammogram for tBuAz-1-tcaa (42), 0.5 mM in THF. Supporting electrolyte 0.1 M Bu NPF . Scan
4
6
–
1
rate 100 mV s .
Figure 13. Voltammogram for MeTAz-1-tcaa (43), 0.5 mM in THF. Supporting electrolyte 0.1 M Bu NPF . Scan
4
6
–
1
rate 100 mV s .
2
2

ACCEPTED MANUSCRIPT
Figure 14. Voltammogram for MeTTMAz-1-tcaa (44), 0.5 mM in THF. Supporting electrolyte 0.1 M Bu NPF .
4
6
–
1
Scan rate 100 mV s .
Both of the radical shielding groups therefore appear to do a good job of protecting the radical
cation formed by abstracting an electron from the dye molecules. The CVs of the dyes containing a
methylthio group (MeTAz-1-tcaa 43 and MeTTMAz-1-tcaa 44) are very similar; the three methyl
groups attached to the 7-membered ring increase the dye oxidation potential by approximately 30
mV. The difference between the methylthio and tert-butyl shielding groups is more striking; the
shift in oxidation potential between MeTAz-1-tcaa (43) and tBuAz-1-tcaa (42) is over 250 mV,
showing that the dye HOMO energy level can be ‘tuned’ to a certain extent by changing the shielding
group. Electrochemical data extracted from the cyclic voltammograms are presented in Table 4. In
the ideal case of a reversible one electron redox couple, the difference between the peak cathodic
and anodic potentials (Epc and E ), ΔE is approximately 59 mV (at 25 °C) and the ratio of the peak
pa
p
8
2
anodic and cathodic currents (Ipa and I ) is unity. The three dyes depart slightly from this ideal; ΔEp
pc
=
131 mV for all three dyes and I /I varies from 0.92 to 0.94. It is interesting to note that the
pc pa
values for ΔE are identical for each dye, suggesting that the slower than ideal electrode dynamics
p
2
3

ACCEPTED MANUSCRIPT
may be due to a different factor, e.g. solvent or electrolyte, which were constant for each
experiment.
0
Dye
Epa
Epc
∆Ep
E
Ipc
Ipa
Eox mV
½
Ipa
µ
A
Ipc µA
mV
mV
mV
mV
534
287
312
tBuAz-1-tcaa
599
352
378
468
221
247
131
131
131
9.63
8.45
7.38
9.10
7.72
6.95
0.92
0.92
0.94
470
210
240
+
(
42)
MeTAz-1-
tcaa (43)
MeTTMAz-1-
tcaa (44)
Table 4. Second generation dyes’ electrochemical data in THF (E/V vs. Fc/Fc )
Energy level data for the first and second generation dyes are summarised in Table 5. The
0
+
ground state (D /D ) potentials and band gap (E ) values are taken from the electrochemistry and
g
0
+
UV/Vis spectrometry results (see above). Subtracting E from D /D gives the ground state HOMO
g
level. All of the dyes satisfy the energy level requirements to function as DSSC sensitizers. The redox
-
-
3
potential of the ground state is 0.44 to 0.70 V below that of the I /I couple used in the electrolyte.
0
D
Eg (opt)
eV
HOMO
Dye
V vs NHE
+
D
V vs NHE
Az-1-tcaa (19)
0.94
2.27
2.27
2.23
2.20
2.12
-1.33
-1.32
-1.13
-1.36
-1.25
TMAz-1-tcaa (20)
tBuAz-1-tcaa (42)
MeTAz-1-tcaa (43)
MeTTMAz-1-tcaa (44)
0.95
1.1
0.84
0.87
Table 5. First and second generation dyes’ energy levels.
2
4

ACCEPTED MANUSCRIPT
Construction and evaluation of DSSCs using the first and second generation dyes was then
2
undertaken. The iodide/triiodide redox couple was used, with a 1 cm cell area. Initial experiments
83-84
85-86
with 44 indicated that addition of lithium iodide
and chenodeoxycholic acid (CA)
to the
electrolyte were both beneficial. Thus, for comparison of all the dyes, the electrolyte composition
shown in Table 6 was used. Cells constructed concurrently using dye N719 were also evaluated as a
5
reference for comparison.
Component
Concentration
0.05 M
Iodine
1-methyl-3-propyl-imidazonium iodide (ionic liquid)
0.60 M
Lithium Iodide
0.10 M
Chenodeoxycholic acid
0.10 M
Table 6. DSSC electrolyte formulation in acetonitrile.
The DSSC evaluation results are presented in Table 7 (and graphically in Figures S20-S23, see ESI).
The best result was achieved with the reference dye N719; of our dyes, the highest mean efficiency
of 1.20% was achieved with the first generation dye TMAz-1-tcaa (20). The next highest performing
cells were based on the second generation dye MeTTMAz-1-tcaa (44). It is interesting to note that,
contrary to expectation, the best performing dyes were the ones based on 4,6,8-trimethylazulene
(
25) and therefore had a methyl group in the seven-membered ring that causes a large torsional
angle between the plane of the azulene rings and the thiophene ring of the linker group (see above).
This was expected to disrupt conjugation between the azulene and thiophene rings and therefore
restrict charge transfer between them too, limiting DSSC performance. However, another factor that
detrimentally influences DSSC efficiency is re-combination of electrons injected into the TiO with
2
the oxidised dye species. The high torsional angle of the dyes based on 4,6,8-trimethylazulene may
conceivably be beneficial in this respect, helping to block charge transfer from TiO to the oxidised
2
azulene moiety, resulting in improved efficiency.
2
5

ACCEPTED MANUSCRIPT
Dye
Voc (V)
Jsc (mA)
FF (%)
η
(%)
η
/
ηN719 (%)
N719
0.549
16.07
55.7
4.96
n/a
Az-1-tcaa
0.309
0.311
3.01
3.15
2.98
3.05
61.0
59.3
54.8
58.4
0.57
0.59
0.50
0.55
11.5
11.9
10.1
11.1
(
19)
0.305
Mean
0.308
TMAz-1-tcaa
0.387
0.378
5.38
5.17
5.07
5.21
60.3
61.0
60.1
60.5
1.27
1.21
1.13
1.20
25.6
24.3
22.8
24.2
(
20)
0.366
Mean
0.377
tBuAz-1-tcaa
0.303
0.310
3.06
2.54
3.11
2.90
61.0
60.7
61.7
61.1
0.57
0.48
0.63
0.56
11.5
9.7
(
42)
0.325
12.7
11.3
Mean
0.313
MeTAz-1-tcaa
0.256
0.254
1.71
1.78
1.80
1.76
52.0
51.1
48.5
50.5
0.23
0.23
0.22
0.23
4.6
4.6
4.4
4.6
(
43)
0.245
Mean
0.252
MeTTMAz-1-tcaa
0.350
0.342
3.38
3.61
3.29
3.43
62.3
60.8
59.1
60.7
0.75
0.76
0.66
0.72
15.1
15.3
13.3
14.5
(
44)
0.335
Mean
0.342
Table 7. DSSC cell parameters, with LiI and CA. FF = Fill factor
Two dyes were selected for a more detailed electrolyte additive analysis: first generation dye TMAz-
-tcaa (20) and second generation dye tBuAz-1-tcaa (42). The three additives to be evaluated in
1
combination were lithium iodide (LiI), chenodeoxycholic acid (CA), and guanidine thiocyanate (GT).
GT has been shown to increase DSSC short circuit current density (J ) and therefore increase
SC
87
efficiency. The DSSC evaluation results for these modified electrolytes are presented in Table 8
and graphically in Figures S24-S27, see ESI). The best DSSC efficiencies for both dyes were obtained
with the combination of both LiI and CA. As before, the first generation dye TMAz-1-tcaa (20)
(
2
6

ACCEPTED MANUSCRIPT
produced the highest efficiency of 1.33% while the second generation dye tBuAz-1-tcaa (42) gave an
efficiency of 0.92%. The addition of guanidine thiocyanate to the DSSC electrolyte actually had a
detrimental effect, reducing DSSC efficiency by 25 – 33%. Rather than increase short circuit current
density, the JSC was severely reduced while open circuit voltage was increased. The addition of LiI
and/or CA to the electrolyte both have a positive influence on cell efficiencies and seem to exhibit a
synergistic effect when both are added together. The LiI and CA acid both influence the DSSC
efficiencies by increasing the closed circuit current density (J ). While both also reduce the open
SC
circuit voltage (V ), the effect on efficiencies is more than offset by the increased J .
OC
SC
2
7

ACCEPTED MANUSCRIPT
Dye
Additive
Voc (V)
Jsc (mA)
FF (%)
η
(%)
TMAz-1-tcaa
Li
0.498
0.503
0.501
1.55
1.40
1.48
71.8
73.0
72.4
0.56
0.51
0.54
(
20)
Mean
CA, Li
0.391
5.72
5.73
5.73
59.3
58.6
59.0
1.33
1.32
1.33
0.390
Mean
0.391
CA, GT
0.544
1.22
1.37
1.30
75.7
75.6
75.7
0.50
0.58
0.54
0.554
Mean
0.549
CA, Li, GT
0.482
2.97
3.00
2.99
69.1
69.5
69.3
0.99
0.99
0.99
0.474
Mean
Li
0.478
tBuAz-1-tcaa
42)
0.479
0.494
0.487
1.11
1.31
1.21
73.6
73.4
73.5
0.39
0.47
0.43
(
Mean
CA, Li
0.370
4.14
4.35
4.25
58.1
58.1
58.1
0.89
0.94
0.92
0.372
Mean
0.371
CA, GT
0.524
0.94
0.88
0.91
74.4
68.1
71.3
0.37
0.30
0.34
0.495
Mean
0.510
CA, Li, GT
0.442
1.89
2.05
1.97
70.7
69.1
69.9
0.59
0.63
0.61
0.447
Mean
0.445
Table 8. DSSC cell parameters – evaluation of electrolyte additives. FF = Fill factor.
3
. Conclusions
In summary, we have prepared five new azulene-containing dyes and evaluated their performance
as sensitizers of titania in dye-sensitized solar cells. The synthesis of these dyes showcases the utility
of the azulenesulfonium salt cross-coupling reaction that we have developed, with the dyes being
accessed in only 4-5 steps from the parent azulene starting materials. The first generation of dyes
2
8

ACCEPTED MANUSCRIPT
(
Az-1-tcaa 19 and TMAz-1-tcaa 20) were shown by cyclic voltammetry to display non-reversible
electrochemistry. This in turn prompted a literature search for azulenes that are known to exhibit
reversible electrochemistry. Based on the results of this search, second generation dyes were
rationally designed to target the desired property of reversible electrochemistry. The second
generation dyes (tBuAz-1-tcaa 42, MeTAz-1-tcaa 43 and MeTTMAz-1-tcaa 44) were synthesised and
it was shown by cyclic voltammetry that they do indeed exhibit reversible electrochemistry.
However, when both the first and second generation dyes were evaluated in dye-sensitized solar
cells, no trend for increased efficiency of the second generation dyes over the first generation dyes
was observed. This implies that stability of the oxidised form of the dye was not the dominant factor
in determining overall cell efficiency; we instead propose a rationale for the observed efficiencies
based on an increased torsional angle between the azulene and thiophene rings induced by the
presence of the additional methyl substituents in dyes 20 and 44 (the two dyes with highest η).
4
. Supporting information
1
13
Synthetic procedures and compound characterisation data. H and C NMR spectra for all novel
compounds. Experimental procedures for cyclic voltammetry and solar cell construction. UV-vis
absorption spectra for dyes 19, 20, 42, 43, and 44. Cyclic voltammograms for previously reported
azulenes. X-ray crystallographic data for 20, 29, 42 and 43.
5
. Acknowledgements
We thank the University of Bath (for a PhD fee waiver to Paul Cowper) and EPSRC (CDT PhD
studentship for Adam Pockett, Grant EP/G03768X/1).
6
1
2
. References
.
.
Tian H, Huang S, Zhou C. Acta Phys-Chim Sin 1988;4:314–19.
O'Regan B, Grätzel M. Nature 1991;353:737-40.
2
9