Atomistic simulations of thiol-terminated modifiers for hybrid photovoltaic interfaces

TSF-32601; No of Pages 5 Thin Solid Films xxx (2013) xxx–xxx

Contents lists available at ScienceDirect

Thin Solid Films journal homepage: www.elsevier.com/locate/tsf

Atomistic simulations of thiol-terminated modifiers for hybrid photovoltaic interfaces G. Malloci a, A. Petrozza b, A. Mattoni a,⁎ a b

Istituto Officina dei Materiali (CNR-IOM), Unità di Cagliari, Cittadella Universitaria, I-09042 Monserrato (CA), Italy Center for Nano Science and Technology @Polimi, Istituto Italiano di Tecnologia, Via Pascoli 70/3, I-20133 Milano, Italy

a r t i c l e

i n f o

Available online xxxx Keywords: Hybrid polymer/TiO2 solar cells Interface modifiers Self-assembled monolayers Pyridine derivatives Ti–N (Ti–O) anchoring Thiol groups Density functional theory calculations

a b s t r a c t Small aromatic molecules such as benzene or pyridine derivatives are often used as interface modifiers (IMs) at polymer/metal oxide hybrid interfaces. We performed a theoretical investigation on prototypical thiolterminated IMs aimed at improving the photovoltaic performances of poly(3-hexylthiophene)/TiO2 devices. By means of first-principles calculations in the framework of the density functional theory we investigate 3furanthiol (3FT), 4-mercaptobenzoicacid (4MB), and 6-isoquinolinethiol (6QT) molecules. We discuss the role of these molecules as modifiers alternative to 4-mercaptopyridine (4MP) which has recently shown to induce a large improvement in the overall power conversion efficiency of mesoporous films of TiO2 infiltrated by poly(3-hexylthiophene). The IMs investigated are expected to keep the beneficial features of 4MP giving at the same time the possibility to further tune the interlayer properties (e.g., its thickness, stability, and density). Dense interlayers of 6QT turn out to be slightly unstable since the titania substrate induces a compressive strain in the molecular film. On the contrary, we predict very stable films for both 3FT and 4MB molecules, which makes them interesting candidates for future experimental investigations. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Despite very promising in principle, hybrid polymer/metal oxide solar cells have relatively low photoconversion efficiencies, typically below 1% [1]. The reason why they are so inefficient is still not clearly understood [2]. Different strategies have been adopted over the past few years to overcome such limits. Among them, the use of molecular interface modifiers (IMs) has proven to be an effective way to improve the performances of functional interfaces for photovoltaics (PV) [3]. The main motivations addressed in the literature for using IMs in the development of hybrid PV devices are [3–11]: increasing the polymer/metal oxide interaction; enhancing charge transport; adjusting the work function of the substrate by introducing molecular dipoles; contributing to light harvesting; and reducing charge carriers recombination. The pyridine molecule and its derivatives have been largely employed to optimize the performances of hybrid solar cells [4,5,7,11]. A recent work on mesoporous films of TiO2 infiltrated by poly(3-hexylthiophene) (P3HT) [12] shows that the use as IM of the simple optically inactive and inexpensive 4-mercaptopyridine molecule (4MP, C5H5NS) is able to enhance the photocurrent of the cell, inducing a large improvement in the overall power conversion efficiency for this class of devices. The 4MP molecule, consisting of a pyridine aromatic ring with a thiol group in the position opposite to the nitrogen atom (see Fig. 1), is known to form wellordered self-assembled monolayers on silver [13] and gold [14] surfaces. ⁎ Corresponding author. Tel.: +39 070 6754868; fax: +39 070 6754892. E-mail address: [email protected] (A. Mattoni).

The reasons of the beneficial effects of 4MP on the photovoltaic performances of TiO2/P3HT hybrid solar cells have been extensively investigated both experimentally and theoretically [12,15]. A comparative analysis on different pyridine-based modifiers has shown that the good performances of 4MP are most likely due to the spontaneous self-assembling of IM molecules on titania and to the formation of a stable and ordered interlayer; this molecular interlayer is bound to the substrate through Ti\N bonds and exposes thiol groups to the polymer phase [12,15]. Xray photoemission spectroscopy results on the interaction between 4MP and the anatase TiO2 surface reveal, indeed, that the covalent interaction between the N atom and the under-coordinated Ti surface is the dominant bonding type, though a substantial amount of 4MP molecules binds via hydrogen mediation [16]. The simulations, performed in the ideal conditions of defect-free surface and completely saturated IMs, reveal a crystalline 4MP monolayer with a zig-zag motif in which, in principle, all of the available TiO2 adsorption sites are passivated. Notably, it has been found that small details in the structure of the single molecule imply very large differences in the morphology of the corresponding selfassembled monolayer [15]. For example, by modifying the position of the thiol group from 4 to 2 in the pyridine ring (by considering 2mercaptopyridine instead of 4-mercaptopyridine), the corresponding interlayer is not anymore crystalline but disordered. This morphological difference is associated to the loss of the PV response of the corresponding interface [12]. On the other hand, the thiol group of 4MP in a position opposite to the nitrogen atom has two beneficial effects: it contributes to stabilize the monolayer and it favors the formation of a dense and ordered assembling. In addition, the thiol-terminated interlayer gives

0040-6090/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.tsf.2013.09.043

Please cite this article as: G. Malloci, et al., Thin Solid Films (2013), http://dx.doi.org/10.1016/j.tsf.2013.09.043

2

G. Malloci et al. / Thin Solid Films xxx (2013) xxx–xxx

isolated and self-assembled bound molecules. We discuss the candidates that best satisfy the above technological requirements and, as such, are possible candidates for further experimental investigations. More specifically, we propose three prototypical small aromatic molecules alternative to 4MP: 3-furanthiol (3FT, C4H4OS), in which the sixmembered ring of pyridine is replaced by the five-membered ring of furan, 4-mercaptobenzoicacid (4MB, C7H6O2S), differing from 4MP for the presence of a carboxyl group in place of the nitrogen atom, and 6isoquinolinethiol (6QT, C9H7NS) with an additional benzene ring fused to the pyridine ring. This latter molecule, in particular, is similar to 2naphthalenethiol, a small conductive molecule composed of two consecutive benzene rings with a thiol side group, which has been recently used to enhance charge transport in hybrid polymer/ZnO nanorod solar cells [17]. 6QT belongs to the family of thiol-functionalized benzopyridines and enables to investigate the effect of interlayer thickness by varying the number of benzene rings. The interface modifiers considered in the present work are reported in Fig. 1. 2. Computational details

Fig. 1. Stick and ball representations of the thiol terminated interface modifiers considered: 4-mercaptopyridine (4MP, reported for comparison), 3-furanthiol (3FT), 4-mercaptobenzoicacid (4MB), and 6-isoquinolinethiol (6QT).

good compatibility with the polymer by increasing its diffusivity and eventually yielding a larger polymer/substrate interface area, as measured by positron annihilation spectroscopy [12,15]. In the present work, by a comparative theoretical analysis, we discuss alternative IMs that, in principle, are expected to keep the beneficial properties of 4MP while extending the possibility of tuning the interlayer properties (e.g., its thickness, stability, and density) to further optimize the PV performances of polymer/metal oxide devices. We identify the requirements that must be fulfilled by a good IM candidate: (i) It must be a small molecule with aromatic and donor characters. The aromatic character is expected to favor assembling and layer stabilization by π → π interactions. The donor character should favor the electron transfer to the substrate which is expected to adjust the work function of the metal oxide thus increasing the open circuit voltage of the cell. (ii) It must efficiently bind to the surface (binding energy of the order of 1 eV per molecule) in a configuration perpendicular to the substrate and (iii) it must show a spontaneous assembling into a dense and crystalline monolayer. The formation of a dense interlayer driven by an attractive molecule–molecule interaction makes it possible to modify the polymer–substrate interaction over a sizable portion of the hybrid interface. In particular, the crystalline structure of the interlayer should preserve the efficient electron injection by the polymer and avoid the larger resistance of a disordered molecular interlayer (as observed in the case of the 2-mercaptopyridine interlayer [12,15]). (iv) It must provide a smooth electrostatic surface in order to induce a good polymer diffusivity on the substrate. In particular, thiol-terminated IMs have the potential to keep the same features observed in the case of 4MP, namely a larger polymer mobility on the surface and a tendency to spontaneously form ordered interlayers. In addition to the requirements (i)–(iv) described above it would be highly desirable to identify IMs giving rise to interlayers of different thickness, thus allowing for a tuning of the polymer-metal oxide separation in order to optimize charge injection and suppress back recombination. By means of a predictive theoretical investigation using firstprinciples calculations at the density functional theory level, we show that it is possible to identify suitable candidates. For the selected molecules we calculate the structure and the electronic properties of the

Density functional theory (DFT, [18]) calculations have been performed using a plane-wave pseudopotential approach as implemented in the QUANTUM-ESPRESSO package [19]. We used the Perdew– Burke–Ernzerhof exchange-correlation functional [20] and ultrasoft pseudopotentials [21]. Dispersion corrections have been included according to Grimme's parametrization [22,23]. Satisfactorily converged results have been obtained by using kinetic energy cutoffs of 25 Ry for wavefunctions and of 200 Ry for charge density. The k-point sampling of the Brillouin zone was limited to the Γ point. For geometry optimizations we used the Broyden–Fletcher–Goldfarb–Shanno algorithm, with threshold values of 0.0026 eV/Å and 1.4 × 10−4 eV for residual forces and energy variation, respectively. The anatase (101) surface was modeled with a periodically repeated slab cut from the bulk. The optimized bulk lattice parameters (a,b,d = 3.79, 9.74, 2.01 Å) are consistent with previous calculations (a,b,d = 3.786, 9.737, 2.002 Å [24,25]) and agree with available experimental data (a,b,d = 3.782, 9.502, 1.979 Å [26]). To simulate IMs interacting with the anatase (101) surface we used supercells containing four atomic layers (72 atoms, 5.9 Å thick) separated h i by a vacuum of ~15 Å. The surface has dimensions 2 × 3 in the 101 and [010] directions, respectively, corresponding to a surface area of 11.37 × 10.45 Å. During geometry optimizations the atoms in the bottom layer were fixed to their bulk positions. Test calculations with a TiO2 surface of six atomic layers (108 atoms, 9.4 Å thick) did not show any significant difference. To estimate the optical gap of the molecules we performed time-dependent DFT calculations [27] using the TURBOMOLE code [28]; we used a Gaussian atomic orbital basis set of split valence triple-ζ quality (TZVP) and the hybrid exchange-correlation functional B3LYP, which has proven to give a good description of the optical gap of small polyaromatic hydrocarbons [29]. The same code has been used to compute the dipole moment of each molecule. Molecular graphics have been generated by using the XCRYSDEN [30] and VMD [31] packages. 3. Results and discussion We first calculated the electronic properties of the isolated molecules by means of B3LYP time-dependent DFT calculations. As expected, all of the molecules considered are optically transparent, the first optically active transition falling at about 4.9, 5.0, 4.5, 4.1 eV for 4MP, 3FT, 4MB, and 6QT, respectively. Note that the same computational scheme gives ~4.8 eV for the optical gap of pyridine, which is in reasonable agreement with the experimental value of about 4.6 eV (see compilation in Ref. [32]). To compare the different IMs we first calculated the minimumenergy configuration of the isolated molecule adsorbed on the ideal TiO2 anatase (101) surface (see Fig. 2). All of the molecules bind to a

Please cite this article as: G. Malloci, et al., Thin Solid Films (2013), http://dx.doi.org/10.1016/j.tsf.2013.09.043

G. Malloci et al. / Thin Solid Films xxx (2013) xxx–xxx

5-fold coordinated titanium atom of the substrate. We found the formation of an effective Ti\O (Ti\N) bond of ~0.2 nm in the case of 3FT and 4MB (6QT). In the case of 4MB we have considered only the molecular monodentate adsorption mode in which the carboxylic group interacts by hydrogen bonding with a surface O atom that is not directly bound to the Ti atom anchoring the molecule [33]. The bond-distances and the adsorption energies Eads, computed by subtracting the energies of the IM and the surface from that of the interacting system, are reported in Table 1. By definition, a positive Eads corresponds to a stable adsorption on the substrate. The adsorption energies are comparable for 4MP, 4MB, and 6QT, ranging between 1.2 and 1.3 eV, while Eads reduces to about 0.7 eV for 3FT. The 3FT molecule is also found to be almost perpendicular to the substrate as in the case of 4MP [15] and THF [34]. The 4MB and 6QT molecules tend to tilt as a consequence of the long-range interactions with the substrate induced by their larger molecular size. A similar effect has been previously observed in the case of 4-tertbutylpyridine [15]. In Table 1 we report also the net charge transfer Δq from the molecule to the substrate as computed by evaluating the partial Löwdin charges before and after chemisorption. While the absolute value of Δq is known to depend on the method used to evaluate the atomic partial charges, and more accurate analyses have been developed (see e.g., Ref. [33]), its use here for comparative purposes is justified. The value of Δq found for 6QT is similar to the one computed for 4MP, as expected from the similar nature of the Ti\N bond involving the active lone pair of the nitrogen atom [15]. On the contrary, Δq is considerably reduced in the case of the Ti\O bond of 4MB, and even more for 3FT. To quantify the effect of molecular adsorption on the electronic properties we calculated the density of states (DOS) for the bare surface, the isolated molecules, and the interacting systems. The computed DOS are reported in Fig. 3. As shown in the figure, our calculations lead to an underestimation of the molecule and semiconductor energy gaps; as an example, the computed TiO2 bandgap is ~2.6 eV to be compared with an experimental value of 3.2 eV. This limit, which is a well known error of DFT due to the discontinuity of the exchange-correlation employed [35], does not affect our comparative analysis. In agreement with the molecule–substrate charge transfer discussed above, as shown by the projected DOS reported in Fig. 3, the highest-occupied and lowestunoccupied molecular orbital levels of 3FT and 4MB are only slightly perturbed after anchoring to the substrate, while those of 6QT and 4MP are lowered by a significant amount. At the same time, the interaction between the molecules and the substrate affects the metal oxide energy levels, which has been explained in terms of molecular dipoles [33,36,38,39,40]. The upward shift of the TiO2 band edge ΔV (about 0.16, 0.33, 0.40, and 0.45 eV) exhibits an almost linear correlation with the absolute value of the molecular dipoles (0.3, 1.4, 2.2, and 2.1 D) for 3FT, 4, 6QT, and 4MP, respectively (see Table 1), calculated

3

Table 1 Bond distance d and adsorption energy to the TiO2 metal oxide substrate for the different thiol-terminated IMs considered. Δq is an estimate of the charge transfer from the molecule to the substrate as computed by evaluating the partial Löwdin charges before and after chemisorption. ΔV is the upward shift of the metal oxide work function following the interaction with the IMs (see Fig. 2). μ is the dipole moment of the isolated molecule, computed at the bound configuration. Ebind is the binding energy of the self-assembled monolayer as defined in the text. Molecule

4MP 3FT 4MB 6QT

d

Eads

Δq

ΔV

μ

Ebind

(Å)

(eV)

(e)

(eV)

(D)

(eV)

2.30 2.39 2.08 2.28

1.24 0.65 1.24 1.31

0.11 0.03 0.05 0.12

0.45 0.16 0.33 0.40

2.14 0.30 1.39 2.24

0.09 0.20 0.38 −0.09

for the isolated molecules at the bound configuration. At variance with the other molecules investigated, the dipole of 3FT points towards the substrate, a consequence of the asymmetric position of the thiol group with respect to the oxygen atom in the pentagonal ring (see Fig. 1). To verify the stability of our results we performed test calculations for a single 4MP molecule bound to a larger TiO2 surface (22.74 × 20.90 Å). As expected we did not find any significant difference concerning the molecular energy levels shift, the adsorption energy Eads and the net charge transfer Δq; on the contrary, we found a considerable reduction of the TiO2 band edge shift ΔV (~ 0.15 eV vs. 0.45 eV) which was also expected for a smaller degree of molecular coverage. The values reported in Table 1 for ΔV, therefore, must be regarded only in comparative terms. From the above results we found that the IMs considered should give rise to an increase in VOC moving from 3FT to 4MB and 6QT. The effect is less pronounced in the Ti\O anchoring cases (3FT and 4MB molecules) while it is larger for the Ti\N anchoring 6QT molecule, whose associated ΔV is similar to that computed for 4MP. To maximize the increase in solar cell efficiency by using IMs, ideally one would like to improve simultaneously the device's open circuit voltage VOC and the generated photocurrent JSC. In the specific case of 4MP an increase in both VOC and JSC has been reported and attributed to the possible formation of a self-assembled 4MP crystalline interlayer [12,15]. We therefore extended the analysis of the single molecules 3FT, 4MB, and 6QT, to the case of self-assembled monolayers and compared the results with those previously found for 4MP. To this aim, we replicated the molecules (in their bound configuration) over all possible adsorption sites of the surface starting from a guess geometry similar to the zig-zag pattern observed for 4MP [12,15]. Model potential molecular dynamics simulations performed for pyridine and 4-mercaptopyridine starting from different configurations of the self-assembled monolayer shows in all cases the spontaneous formation of the above zig-zag pattern which is the most

Fig. 2. Top and side views of the minimum-energy configurations for the thiol-terminated IMs considered, as computed by DFT.

Please cite this article as: G. Malloci, et al., Thin Solid Films (2013), http://dx.doi.org/10.1016/j.tsf.2013.09.043

4

G. Malloci et al. / Thin Solid Films xxx (2013) xxx–xxx

PDOS (arbitrary units)

4MP 6QT 4MB 3FT 7

8

9

10

11

12

13

14

Energy (eV) Fig. 3. Projected density of states (PDOS) of the IM/TiO2 systems (continuous lines for interacting IM-TiO2, dashed lines for the separated components); the electronic Eigenvalues have been aligned by using the 1 s level of a He atom inserted as a reference in all the supercells [37].

energetically favored and remains stable at room temperature [15]. In the ideal regime in which all 5-fold Ti atoms are passivated the IM density is as large as 8.4 μmol m−2, a value comparable with typical dye loadings [41]. All the zig-zag assembled layers were fully optimized at the DFT level making it possible to compare the relative energetic stability of the monolayers. An example of relaxed zig-zag layer is reported in Fig. 4 for the case of the 6QT molecule. We calculate the binding energy Ebind of the film as the difference between the adsorption energy of the monolayer Eml and that of the same number of single bound molecules Eads: Ebind = Eml − N × Eads, where N is the number of molecules in the monolayer. By definition, this quantity is positive when the monolayer is more stable with respect to the single bound molecules, as a result of an attractive interaction between neighboring molecules. The values computed for Ebind are reported in Table 1 and compared with the value previously computed for 4MP [15]. As shown in the table, the stability of the film is strongly dependent on the nature of the molecule. Dense interlayers of 6QT turn out to be slightly unstable (Ebind ~ −0.1). On the contrary, both 3FT and 4MB give very stable films, Ebind ~ 0.2, and ~0.4 eV, respectively. This result makes these molecules interesting candidates for future experimental investigations. In particular, from the point of view of the expected device performance we predict that both IMs should give rise to i) an increased stability, coming from the interlayer stability; ii) an increased VOC, driven by the interaction of the molecules with the substrate that upshifts the TiO2 band edge (see Table 1); and iii) an increased JSC due to the better compatibility between the polymer and the metal oxide.

Since the 4MP and 6QT are closely related, in order to understand the different stabilities of their corresponding interlayers we performed an additional in-depth analysis. First, we observed that the distribution −2 of adsorption sites on the titania substrate is σ = (πrsub , with an 0 ) sub intermolecular distance r0 = 2.51 Å. Obviously, this molecular density value does not necessarily corresponds to the optimal density of the interlayers. This latter is estimated for the two molecules 4MP and 6QT by considering the corresponding crystalline layer (not bound to the substrate) and by calculating the cohesive energy as a function of the intermolecular distance r during a homogenous in-plane scaling. The calculated data, reported in Fig. 5, can be fitted by the universal energy relation −s

uðsÞ ¼ −E0 ð1 þ sÞe

 ;

s¼

 r 1 ; −1 r0 μ

that describes the energy dependence on strain for a wide class of cohesive materials by means of only three parameters: the cohesive energy E0, the equilibrium distance r0, and a third quantity μ related to the inplane stiffness of the molecular crystals [42]. We found that the optimal intermolecular distance is larger for 6QT than for 4MP (r4MP = 0 2.55 Å b r6QT = 2.70 Å) and, more importantly, we find that r4MP better 0 0 matches the density of titania surface rsub = 2.51 Å (vertical dashed 0 line in Fig. 5). In other words, the 4MP arrangement at rsub is energeti0 cally favored since it does not induce any strain in the molecular film   du j ≃0 at variance with 6QT for which du j b0 (compressive strain). dr r 0 dr r 0 Furthermore, we found a larger cohesive energy for 6QT (E4MP = 0 0.34 eV b E6QT = 0.47 eV) which was expected from the larger mole0 cule–molecule interaction due to the additional benzene ring of 6QT. Finally, we obtained μ ~ 0.11 for both layers, consistently with the same origin of the intermolecular interaction. According to the above analysis, stable 6QT interlayers are compatible with lower molecular densities but a detailed study of their possible assembling patterns is beyond the scopes of the present investigation. 4. Concluding remarks In conclusion, we demonstrate that atomistic calculations can be used to predict the functionality of modified interfaces. Driven by recent results on the 4-mercaptopyridine molecule as interface modifier of the hybrid P3HT/TiO2 interface, we performed a theoretical investigation on alternative prototypical thiol-terminated IMs aimed at further improving the photovoltaic performances of hybrid polymer/metal oxide devices. By means of first-principle density functional theory calculations

Fig. 4. Top and side view of the minimum-energy configuration of the self-assembled monolayer of 6QT, as computed by DFT.

Please cite this article as: G. Malloci, et al., Thin Solid Films (2013), http://dx.doi.org/10.1016/j.tsf.2013.09.043

G. Malloci et al. / Thin Solid Films xxx (2013) xxx–xxx

cohesive energy (eV)

0

5

References

−0.2

[1] [2] [3] [4] [5]

4MP 6QT

[6]

−0.4

[7]

2.5

3

3.5

4

effective intermolecular distance (Å) Fig. 5. Cohesive energy as a function of the intermolecular distance for the 4MP and 6QT crystalline layers during a homogenous in-plane scaling. The vertical dashed line represents the intermolecular distance rsub = 2.51 Å corresponding to the adsorption sites on 0 the (101) titania surface. The continuous lines fit to the data obtained using the universal energy relation (see text).

[8] [9] [10] [11] [12]

[13] [14]

we computed the electronic structure and adsorption energy of each IMs, both singly-bound and self-assembled on a TiO2 (101) anatase surface. In particular, we considered 3-furanthiol (3FT), 4-mercaptobenzoicacid (4MB), and 6-isoquinolinethiol (6QT) by discussing their role as possible IMs alternative to 4MP. These IMs are expected to keep the beneficial features of 4MP giving at the same time the possibility to further tune the interlayer properties (e.g., its thickness, stability, and density). Overall, we found that the stability of the film is strongly dependent on the nature of the molecule. Interestingly, the comparative analysis of the energy dependence on strain for 4MP and 6QT crystalline layers reveals that the 6QT arrangement at the intermolecular distance of the TiO2 (101) substrate is energetically unfavored since it induces a compressive strain in the molecular film. However, stable 6QT interlayers are expected at lower molecular densities and a detailed study of their possible assembling patterns deserves additional work. On the contrary, high density stable films are predicted for both 3FT and 4MB molecules. While additional work is needed to simulate non-perfect self-assembled monolayers, our results for the ideal regime in which all possible adsorption sites are passivated still make these molecules interesting candidates for future experimental investigations. Both IMs, in particular, are expected to improve the performance of actual devices in terms of increased stability, open-circuit voltage, and extracted photocurrent.

[15] [16] [17] [18] [19]

[20] [21] [22] [23] [24] [25] [26] [27] [28]

[29] [30] [31] [32] [33] [34]

Acknowledgments This work has been funded by Italian Institute of Technology (IIT) under Project SEED “POLYPHEMO” and Platform “Computation”, by Regione Autonoma della Sardegna under L.R. 7/2007 CRP-249078 and CRP-18013, by MIUR Under PON 2007-2013 (Project NETERGIT), and by Consiglio Nazionale delle Ricerche (Progetto Premialità RADIUS). We acknowledge computational support by CINECA through ISCRA Initiative (Project TERNARY).

[35] [36] [37] [38] [39] [40] [41] [42]

M. Wright, A. Uddin, Solar Energy Materials & Solar Cells 107 (2012) 87. J. Bouclé, J. Ackermann, Polymer International 61 (2012) 355. C. Goh, S.R. Scully, M.D. McGehee, J. Appl. Phys. 101 (2007) 114503. Y.Y. Lin, T.H. Chu, C.W. Chen, W.F. Su, Appl. Phys. Lett. 92 (2008) 053312. Y.Y. Lin, T.H. Chu, S.S. Li, C.H. Chuang, C.H. Chang, W.F. Su, C.P. Chang, M.W. Chu, C.W. Chen, J. Am. Chem. Soc. 131 (2009) 3644. J.A. Chang, J.H. Rhee, S.H. Im, Y.H. Lee, H.J. Kim, S.I. Seok, M.K. Nazeeruddin, M. Grätzel, Nano Lett. 10 (2010) 2609. Y.-C. Huang, J.-H. Hsu, Y.-C. Liao, W.-C. Yen, S.-S. Li, S.-T. Lin, C.-W. Chen, W.-F.J. Su, Mater. Chem. 21 (2011) 4450. J. Weickert, R.B. Dunbar, H.C. Hesse, W. Wiedemann, L. Schmidt-Mende, Adv. Mater. 23 (2011) 1810. J. Weickert, F. Auras, T. Bein, L. Schmidt-Mende, J. Phys. Chem. C 115 (2011) 15081. J. Yu, T.L. Shen, W.H. Weng, Y.C. Huang, C.I. Huang, W.F. Su, S.P. Rwei, K.C. Ho, L. Wang, Adv. Energy Mater. 2 (2012) 245. J.-F. Lin, G.-Y. Tu, C.-C. Ho, C.-Y. Chang, W.-C. Yen, S.-H. Hsu, Y.-F. Chen, W.-F. Su, ACS Appl. Mater. Interfaces 5 (2013) 1009. E.V. Canesi, M. Binda, A. Abate, S. Guarnera, L. Moretti, V. D'Innocenzo, R.S. Santosh Kumar, C. Bertarelli, A. Abrusci, H. Snaith, A. Calloni, A. Brambilla, F. Ciccacci, S. Aghion, F. Moia, R. Ferragut, C. Melis, G. Malloci, A. Mattoni, G. Lanzani, A. Petrozza, Energy Environ. Sci. 5 (2012) 9068. W.H. Do, C.J. Lee, D.Y. Kim, M.J. Jung, J. Ind. Eng. Chem. 18 (2012) 2141. E.A. Ramírez, E. Cortés, A.A. Rubert, P. Carro, G. Benítez, M.E. Vela, R.C. Salvarezza, Langmuir 28 (2012) 6839. G. Malloci, M. Binda, A. Petrozza, A. Mattoni, J. Phys. Chem. C 117 (2013) 13894. A. Calloni, A. Brambilla, G. Berti, G. Bussetti, E.V. Canesi, M. Binda, A. Petrozza, M. Finazzi, F. Ciccacci, L. Duò, Langmuir 29 (2013) 8302. Jian-Yu Chen, Fang-Chi Hsu, Yun-Ming Sung, Yang-Fang Chen, J. Mater. Chem. 22 (2012) 15726. W. Kohn, Rev. Mod. Phys. 71 (1999) 1253. P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, G.L. Chiarotti, M. Cococcioni, I. Dabo, A. Dal Corso, S. de Gironcoli, S. Fabris, G. Fratesi, R. Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, M. Lazzeri, L. Martin-Samos, N. Marzari, F. Mauri, R. Mazzarello, S. Paolini, A. Pasquarello, L. Paulatto, C. Sbraccia, S. Scandolo, G. Sclauzero, A.P. Seitsonen, A. Smogunov, P. Umari, R.M. Wentzcovitch, Journal of Physics: Cond. Matt. 21 (2009) 395502. J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865. D. Vanderbilt, Phys. Rev. B 41 (1990) 7892. S. Grimme, J. Comput. Chem. 27 (2006) 1787. V. Barone, M. Casarin, D. Forrer, M. Pavone, M. Sambi, A. Vittadini, J. Comput. Chem. 20 (2009) 934. M. Lazzeri, A. Vittadini, A. Selloni, Phys. Rev. B 63 (2001) 155409. G. Mattioli, F. Filippone, A. Amore Bonapasta, J. Am. Chem. Soc. 128 (2006) 13772. J.K. Burdett, T. Hughbanks, G.J. Miller, J.W. Richardson, J.V. Smith, J. Am. Chem. Soc. 109 (1987) 3639. M.A.L. Marques, E.K.U. Gross, Ann. Rev. Phys. Chem. 55 (2004) 3425. TURBOMOLE V6.2 2010, a development of University of Karlsruhe and Forschungszentrum Karlsruhe GmbH, TURBOMOLE GmbH, 1989-2007, since 2007; available from http://www.turbomole.com. G. Malloci, G. Cappellini, G. Mulas, A. Mattoni, Chem. Phys. 384 (2011) 19. A. Kokalj, J. Mol. Graphics Mod. 17 (1999) 176. W. Humphrey, A. Dalke, K. Schulten, J. Mol. Graphics 14 (1996) 33. J. Bousquet, R. Fukuda, P. Maitarad, D. Jacquemin, I. Ciofini, C. Adamo, M. Ehara, J. Chem. Theory Comput. 9 (2013) 2368. E. Ronca, M. Pastore, L. Belpassi, F. Tarantelli, F. De Angelis, Energy Environ. Sci. 6 (2013) 183. M.I. Saba, V. Calzia, C. Melis, L. Colombo, A. Mattoni, J. Phys. Chem. C. 116 (23) (2012) 12644–12648. L.J. Sham, M. Schlüter, Phys. Rev. Lett. 51 (1983) 1888. E.M. Johansson, R. Scholin, H. Siegbahn, A. Hagfeldt, H. Rensmo, Chem. Phys. Lett. 515 (2011) 146. G. Mattioli, C. Melis, G. Malloci, F. Filippone, P. Alippi, P. Giannozzi, A. Mattoni, A. Amore Bonapasta, J. Phys. Chem. C 116 (2012) 15439. M.K. Nazeeruddin, A. Kay, I. Rodicio, R. Humphry-Baker, E. Mueller, P. Liska, N. Vlachopoulos, M. Grätzel, J. Am. Chem. Soc. 115 (1993) 6382. H. Kusama, H. Orita, H. Sugihara, Langmuir 24 (2008) 4411. H. Kusama, H. Orita, H. Sugihara, Solar Energy Mater. Solar Cells 92 (2008) 84. J. Krüger, U. Bach, M. Grätzel, Adv. Mater. 12 (2000) 447. A. Mattoni, M. Ippolito, L. Colombo, Phys. Rev. B 76 (2007) 224103.

Please cite this article as: G. Malloci, et al., Thin Solid Films (2013), http://dx.doi.org/10.1016/j.tsf.2013.09.043