Computational Investigation of Dye–Iodine Interactions in Organic Dye

Feb 11, 2012 - DFT, MP2, and Car−Parrinello molecular dynamics simulations, aimed to model ... donation from the electrolyte to the oxidized dye, wh...
1 downloads 0 Views 4MB Size
Article pubs.acs.org/JPCC

Computational Investigation of Dye−Iodine Interactions in Organic Dye-Sensitized Solar Cells Mariachiara Pastore,* Edoardo Mosconi, and Filippo De Angelis* Istituto CNR di Scienze e Tecnologie Molecolari c/o Dipartimento di Chimica, Università di Perugia, via Elce di Sotto 8, I-06123, Perugia, Italy S Supporting Information *

ABSTRACT: We present a computational investigation, based on DFT, MP2, and Car−Parrinello molecular dynamics simulations, aimed to model the dye−iodine interactions in organic dyesensitized solar cells. We evaluate the binding of I2 and I3− and Li+ to the various electron-donating sites in organic sensitizers considering two coumarin dyes differing by the number of thiophene rings in the bridge (NKX2587 vs NKX2697) and one carbazole dye (MK3), which have been the subject of a detailed experimental investigation (J. Am. Chem Soc. 2008, 130, 17874− 17881). We find that oxygen atoms are the preferred binding sites for I2, while I3− tends to interact with the π system of the coumarin donor unit. Our results suggest that the increase of the distance of the carbonyl oxygen from the titania going from NKX2587 to NKX2697 could explain the differences in the lifetime values measured for NKX2587-sensitized solar cells compared to those employing the longer NKX2697 homologues. The shorter lifetime measured for MK3 compared to NKX2687, which have the same π-spacer and acceptor units, is instead attributed to the twisting of the dye structure in MK3, which possibly forms less compact dye layers on titania. The effect of the dye structure and Li+ coordination to the dyes on the titania conduction band is also examined.

1. INTRODUCTION Dye-sensitized solar cells (DSCs)1−3 are among the most promising solar technologies, which offer a potentially high efficiency and a considerably lower fabrication cost with respect to traditional silicon-based photovoltaics. Up to now, significant advances have been achieved in terms of both materials optimization and understanding of the key processes governing the cell efficiency.4,5 In DSC devices, a dye sensitizer, adsorbed on a mesoporous oxide layer composed of nanometer-sized particles, upon light irradiation, injects the photoexcited electrons into the manifold of unoccupied states of the semiconductor, generically indicated as “conduction band” (CB). The dye ground state is then restored by electron donation from the electrolyte to the oxidized dye, which is in turn regenerated by its reduction at the counter electrode, closing the electrical circuit (Figure 1). The overall performance of the DSC is strongly determined by the efficiency of the various forward (desired) electron transfer and charge transport processes (Figure 1), which in turn are dependent, among other factors, on the optical,6,7 photoelectrochemical,8,9 and structural properties of the dye sensitizer.10−16 A potentially high-efficient sensitizer should have a wide absorption in the UV and NIR spectrum associated to a long-lived charge-separated excited state, possibly strongly coupled to the oxide CB states; ground and excited state oxidation potentials which properly match the electrolyte and © 2012 American Chemical Society

Figure 1. Scheme of forward (green lines) and back (dotted black lines) electronic processes in DSCs.

TiO2 CB, respectively (Figure 1); and a molecular structure designed to prevent detrimental aggregation phenomena but Received: January 4, 2012 Revised: February 7, 2012 Published: February 11, 2012 5965

dx.doi.org/10.1021/jp300095z | J. Phys. Chem. C 2012, 116, 5965−5973

The Journal of Physical Chemistry C

Article

regeneration process, while experimental evidence of stable Ru-dyeI2 adducts has been reported by Tuikka et al.,49 which isolated N3I2 crystals where iodine is bonded to both thiocyanate groups of the dye. Regarding the recombination with the electrolyte, O’Regan et al.10 found for a Ru-dye that the replacement in the ligands of two oxygen atoms with two sulfur atoms caused a 2-fold increase in the recombination rate and, on the basis of literature data, ascribed such a difference to a stronger tendency of ethylthioether compared to ethylether to bind I2. An extensive investigation on the effects of dye molecular structure and electrolyte composition on the recombination dynamics has been reported by Miyashita et al.,13 who examined the photovoltaic performances of a series of organic- and Rubased DSCs, consistently finding lower electron lifetime values for the metal-free based DSCs compared to those employing Ru dyes. The lower Voc values were principally interpreted as arising from an increased I3− concentration in the proximity of the oxide surface, yielding higher recombination rates. Interestingly, among the organic sensitizers, those having alkyl chains and larger molecular size showed longer lifetimes, possibly due to the formation of a compact aggregate layer able to protect the TiO2 surface from the I3− approach.9 In a recent work,15 we have shown that in the case of two organic dyes containing, respectively, one and two ethylenedioxythiophene (EDOT) moieties the O−I2 interaction is slightly stronger than the S−I2 one and that, in any case, both sites might effectively bind iodine, thus accounting for the accelerated recombination rate measured for the dye bearing two EDOT groups. More recently, the formation of organic dye−I2 or −I3− complexes has also been reported by Zhang et al.20 and Bai et al.,19 which demonstrated the benefit of using iodine-free electrolytes in organic dye sensitized solar cells. Following the computational strategy developed in ref 15, based on a mixed DFT/MP2 approach, here we investigate the dye/iodine interaction by modeling the binding of I2 and I3− to the various electron-donating sites in the dye molecules, for a representative group of sensitizers, selected among those examined by Miyashita et al.13 In particular, we considered two coumarin dyes (NKX2587 and NKX2697) differing by the number of thiophene rings in the bridge (one and three, respectively) and one carbazole dye (MK3), having the same number of thiophene units of NKX2697 but a different donor unit. Such a choice allowed us to investigate the interaction of iodine with a different class of dyes (coumarin vs carbazole) as well as the effect of the molecular size (NKX2587 vs NKX2697). We also model the interaction between the dye (the carbonylic oxygen of the coumarin dye) and the Li+ cation, as well as the formation of dye−Li+−I3− complexes. Upon establishing the preferred binding sites and binding geometries for the I2 and Li+ species, we performed DFT geometry optimization and Car−Parrinello molecular dynamic simulations for the NKX2587, NKX2697, MK3, NKX2587−I2, and NKX2587−Li+ systems grafted onto a TiO2 nanoparticle surface, discussing the effect of the iodine and lithium ion on the recombination reaction and conduction band edge shift, respectively.

possibly allowing the formation of a sufficiently compact dye layer capable of insulating the semiconductor surface from the contact with oxidized species in the electrolyte. Although the highest-performance DSC devices based on the liquid I−/I3− electrolyte17 are still those employing Ru(II)based dyes, such as N3, N719, or C106,2,18 a number of promising push−pull metal-free dyes,19,20 allowing simpler and lower cost preparation processes as well as larger structural variety, have been developed, with overall efficiencies approaching 10%.21,22 Moreover, in conjunction with ferrocene or cobalt-based electrolytes, organic dyes have been shown to clearly outperform Ru(II)-based dyes.19,23−26 Recently, a cocktail of organic dyes with metal porphyrins in conjunction with a cobalt electrolyte has achieved an impressive 13% efficiency.27 The limitation in the efficiency of organic sensitizers, despite their generally high extinction coefficients, has been attributed to both charge recombination of the injected electrons with the oxidized dye or electrolyte and the formation of dye aggregates on the TiO2 surface.12−14,28−32 While dye aggregation can be controlled to some extent by the use of an antiaggregation coadsorbent, typically chenodeoxycholic acid (CDCA),33 the effect of the dye molecular structure on the recombination kinetics, is more entangled and has been investigated for both metal-based10,11,34,35 and metal-free dyes.12,13,15,19,36−39 The recombination of the injected electrons with the oxidized dyes follows a nonadiabatic regime40 and is dependent on the electron density in the semiconductor conduction band/ trap states, on the electrolyte composition,41,42 as well as on the spatial separation of the highest occupied molecular orbital (HOMO) of the dye and the TiO2 surface.43,44 On the other hand, a full understanding of the recombination mechanism involving the I−/I3− electrolyte has still to be reached since, for instance, it is not definitely established if the recombination takes place with either I2 or I3−17 as well as if, and to which extent, the surface trap states participate in the recombination reaction.45,46 The kinetics of this recombination process directly influence the charge density into the semiconductor and thus the open-circuit voltage of the cell, Voc, which is the difference between the quasi-Fermi level of the semiconductor under illumination and the redox potential of the electrolyte. By definition, at open circuit the flux of the injected electrons equals that of the recombining electrons, fixing the Fermi level and thus the output voltage of the device. The Fermi level is determined by the CB edge potential and by the electron density into the TiO2, n, which is in turn proportional to the rate of the electron injection and to the electron lifetime, τ, namely, the average time spent by the injected electrons into the TiO2 CB; low electron lifetime values, due to high recombination rates, therefore are related to low V oc values.10,11,13,36,47 Despite the accepted knowledge that the adsorbed dyes act as an insulating layer, keeping the reduced electrolyte far from the direct contact with the oxide surface,13,36,42,48 some authors10,13,15,19,49 suggested the idea that particular atoms or chemical groups can also provide binding sites for I2 (I3−), increasing its concentration close to the TiO2 surface and thus accelerating the recombination and/or regeneration processes. Recently a number of experimental and computational studies have been focused on the interactions between the oxidized dye and iodide (I−) and their implications in the dye regeneration mechanism.49−55 Computational investigations50−52,55 have contributed to the understanding of the oxidized dye

2. COMPUTATIONAL DETAILS The computational strategy employed to evaluate the stabilization energies of the various dye−I2 adducts relies on a mixed DFT/MP2 approach: the geometry optimization was carried out by B3LYP in the gas phase; then considering the 5966

dx.doi.org/10.1021/jp300095z | J. Phys. Chem. C 2012, 116, 5965−5973

The Journal of Physical Chemistry C

Article

Figure 2. Optimized molecular structures of the NKX2587−I2 and NKX2587−I3− adducts. Iodine atoms are in purple, carbon in green, sulfur in yellow, nitrogen in blue, oxygen in red, and hydrogen in white.

Figure 3. Optimized molecular structures of the NKX2697−I2 adducts. Iodine atoms are in purple, carbon in green, sulfur in yellow, nitrogen in blue, oxygen in red, and hydrogen in white.

bound to the carbonylic oxygen of the coumarin group (Figure 5a) and with the triiodide interacting with the π system of the coumarin moiety while Li+ binds to the carbonylic oxygen of the dye (Figure 5b), which is the I3− binding geometry in the absence of the Li+ cation. To have an estimate of the dye-bonded I2 from the oxide surface, considering the preferred bidentate bridging anchoring mode,15 we carried out full geometry optimizations of the three dyes onto a (TiO2)38 anatase slab exposing the majority (101) surfaces, and for NKX2587 we also optimized the O−I2 adduct structure adsorbed onto the (TiO2)38 cluster. The effect of the Li+ cation on the TiO2 CB energetics was examined by comparing the density of states profiles in the case of the two coumarin dyes attached to the (TiO2)38 cluster, NKX2587 and NKX2697, with or without the lithium ion. The optimizations were carried out in the gas phase with the ADF program package61 employing the Becke−Perdew exchange-correlation functional62,63 with a TZVP(DZVP) basis set for Ti and I (H, C, N, O, S, Li). For the NKX2587 and NKX2697, we also analyzed the Density of States (DOS) profiles calculated at the B3LYP/6-311G** level of theory in water solution, considering a window of 15 virtual orbitals. By calculating for each unoccupied orbital the projected density of states on the dye and on the TiO2, we thus obtained partial DOS profiles, plotted using a gaussian broadening with σ = 0.20 eV and intensities

expectedly weak dye−I2 interactions, MP2 single point energy evaluations on the DFT-optimized geometries were performed both in vacuo and in acetonitrile, employing the conductor-like polarizable continuum model (C-PCM).56 The consistency of this combined DFT/MP2 procedure was checked in ref 15 by repeating the geometry optimizations of selected dye−I2 adducts at the MP2 level, with resulting differences in the stabilization energies within 0.4 kcal/mol. A 6-311G** basis set was employed for both geometry optimization and subsequent single point calculations, which were carried out with the Gaussian 03 package of programs.57 To test the applicability of DFT methodologies expressly developed to describe weak intermolecular interactions, we added the D3 correction by Grimme and co-workers58,59 to the B3LYP energies in the gas phase and further we carried out calculations with the M06 functional by Zhao and Truhlar.60 We calculated the binding energies with iodine for different sites in the molecules, namely, the cyano group (CN−I2 adduct), the sulfur of the thiophene rings (Sn−I2 adducts), and, for the coumarin dyes, the carbonyl oxygen (O−I2 adduct). The formation and the stability of dye− I3− complexes in the presence or not of the Li+ cation was evaluated by performing geometry optimizations and subsequent single point MP2 calculations in acetonitrile solution of the NKX2587−Li+−I3− adduct in two different geometries, namely, with the I3− ion directly interacting with the Li+ cation 5967

dx.doi.org/10.1021/jp300095z | J. Phys. Chem. C 2012, 116, 5965−5973

The Journal of Physical Chemistry C

Article

Figure 4. Optimized molecular structures of the MK3−I2 adducts. Iodine atoms are in purple, carbon in green, sulfur in yellow, nitrogen in blue, oxygen in red, and hydrogen in white.

Table 1. Calculated B3LYP, B3LYP-D3, M06, and MP2 Binding Energies (kcal/mol) in Vacuo and in Acetonitrile Solution for the Various Dye−I2 Adducts and for the NKX2587−I3− Complexa Est (kcal/mol) B3LYP-D3

M06

NKX2587

Vac

B3LYP ACN

Vac

Vac

Vac

MP2 ACN

O−I2 S1I2 CN−I2 NKX2587−I3− NKX2697 O−I2 S1−I2 S2−I2 S3−I2 CN−I2 MK3 S1−I2 S2−I2 S3−I2 CN−I2

−5.8 −1.5 −5.0 −12.6

−3.6 (−2.2) −0.4 (+2.4) −2.3 (−1.9) +0.9

−9.5 −6.3 −7.2 −18.6

−8.8 −4.7 −7.2 −18.0

−6.8 −3.3 −4.9 −14.7

−4.3 (−2.9) −1.9 (+0.9) −2.1 (−1.7) −2.5

I−O 2.84 I−S 3.41 I−N 2.93

−6.4 −2.0 −1.2 −0.4 −5.1

−4.2 −1.3 −0.7 +0.3 −2.2

(−2.8) (+1.8) (+2.1) (+3.6) (−1.8)

−10.1 −7.0 −6.0 −3.1 −7.2

−8.4 −4.6 −4.0 −1.5 −6.4

−7.4 −5.3 −4.7 −2.5 −5.6

−5.2 −4.6 −4.2 −1.8 −3.4

(−3.8) (−1.6) (−1.3) (+1.5) (−3.0)

I−O 2.78 I−S1 3.38 I−S2 3.48 I−S3 4.23 I−N 2.90

−1.6 −1.2 −1.0 −5.0

−0.1 −0.7 −0.4 −2.3

(+3.3) (+2.1) (+2.0) (−1.8)

−7.2 −6.0 −5.4 −7.3

−4.8 −3.8 −3.0 −6.4

−5.2 −4.3 −3.3 −5.2

−3.8 −3.7 −2.7 −2.7

(−0.3) (−0.9) (−0.3) (−2.3)

S1−I 3.41 S2−I 3.47 S3−I 3.50 CN−I 2.92

I−X distance (Å)

a

The relative values without the nonelectrostatic terms are reported within parentheses. In the last column, the predicted bond distances (in Å) are also listed.

to perform a rapid nuclear thermalization and to maintain the adiabaticity throughout the simulation.

given by the calculated percentage weights for the dye and for the TiO2 cluster. For the NKX2587@(TiO2)82 system dye, we also investigated the dynamical aspects by means of Car−Parrinello molecular dynamics,64 as implemented in the QuantumEspresso package,65 employing the GGA-PBE exchangecorrelation functional66 in combination with a plane wave basis set and ultrasoft pseudopotentials.67 Plane wave basis set cutoffs set for the smooth part of the wave functions and the augmented density are 25 and 200 Ry, respectively. The dimensions of the simulation supercells have been defined by adding 7 Å of vacuum to the largest nanoparticle dimension in each direction. Molecular dynamics simulations have been carried out, with an integration time step of 5 au, for a total simulation time of ca. 3 ps. The fictitious mass used for the electronic degrees of freedom is 500 au, and we set the masses to the real masses for the atoms except for the H atom for which we have set a mass value of 2 amu. This setup allows us

3. RESULTS AND DISCUSSION 3.1. Stability and Geometries of Dye−I2/I3− Adducts. Let us start our investigation by discussing the stabilization energies of the relative dye−I2 complexes and their geometry for the stand-alone dyes. The optimized molecular structures of the various adducts of NKX2587, NKX2697, and MK3 with I2 are displayed in Figures 2, 3, and 4 respectively, while the corresponding binding energies as well as the I−X (with X = O, S and N) distances are listed in Table1; Figure 2 also displays the optimized geometry of the NKX2587−I3− complex. As shown in Figure 2, I2 binds to sulfur, oxygen, and nitrogen atoms with a different geometrical arrangement: it is coplanar to the molecule in the case of N and O with C−N−I and C− O−I angles of ca. 180° and 120°, while it is perpendicular to the dye plane when binding to sulfur; on the other hand, I3− essentially interacts with the π system of the coumarin moiety. 5968

dx.doi.org/10.1021/jp300095z | J. Phys. Chem. C 2012, 116, 5965−5973

The Journal of Physical Chemistry C

Article

As shown in Table 1, the B3LYP binding energies are overall slightly underestimated with respect to the MP2 ones, where a more accurate description of weak intermolecular interaction is expected; then, at both levels of calculation the stabilization energies in the gas phase are larger than those obtained in the polar acetonitrile solvent, due to the partial screening of the electrostatic interactions. When B3LYP is corrected with the dispersion contribution58,59 (B3LYP-D3) as well as when the M06 functional60 is employed, larger binding energies compared to the MP2 ones are obtained in the gas phase. Interestingly, at all levels of calculations, the relative stability among oxygen, sulfur, and −CN complexes is the same (see Table 1). The present results thus demonstrate the reliability of such DFT approaches in describing weak intermolecular interactions and their applicability to larger systems, for which MP2 calculations become unfeasible. As previously observed,15 the interaction of I2 with the oxygen lone pair electrons is stronger than that computed with the sulfur lone pairs: in ref 15 we obtained gas phase binding energies of −6.9 and −5.1 kcal/mol for the oxygen and sulfur atoms of the ethylenedioxythiophene (EDOT) group, respectively, and −4.5 kcal/mol for the sulfur of the thiophene ring. The data reported in Table 1 show a comparable and even more pronounced difference in the I2−O and I2−S interactions as well as a sizable tendency to form CN−I2 adducts. Considering the MP2 data in solution as the reference ones, we calculate for NKX2587 stabilization energies of −4.3, −1.9, and −2.1 kcal/mol with bond distances of 2.84, 3.41, and 2.93 Å for the O−I2, S1−I2, and CN−I2 adducts, respectively. Interestingly, moving to the longer NKX2697 coumarin dye, one observes a slightly increased O−I2 interaction with a binding energy of −5.2 kcal/mol and a distance of 2.78 Å, whereas the strength of the S−I2 bond decreases going from S1 to S3, i.e., going toward the acceptor group. A stabilization energy of −3.5 kcal/mol is also computed for the CN−I2 adduct, with a N−I distance of 2.90 Å. A similar trend is found for the carbazole MK3 dye, where the S−I2 interaction (distance) becomes weaker (longer) as the distance from the donor unit increases: −3.8, −3.7, and −2.7 kcal/mol with S−I distances of 3.41, 3.47, and 3.50 Å, respectively. Finally, the CN−I2 adduct is predicted to be stable by −2.73 kcal/mol with a N−I distance of 2.92 Å. We also investigated the possible formation of NKX2587−I3− complexes, finding a slightly favorable interaction (−2.5 kcal/mol at the MP2 level in acetonitrile) between the triiodide and the π-system of the coumarin (see optimized structure in Figure 3 and stability data in Table 1). As reported by Furube et al.68 for a coumarin derivative, electron-rich sites in the dye can bind Li+ ions and possibly attract negative I3− ions, which would therefore result to be close to the TiO2 surface. To estimate the effect of Li+ ions on the stability of the dye−I3− complex, we optimized, in acetonitrile solution at the B3LYP/6311G** level of theory, the structure of the NKX2587−I3− adduct with a Li+ ion bound to the carbonyl oxygen of the coumarin group (Figure 5). We examined two different coordination geometries of I3−, with (a) I3− directly interacting with the Li+ cation bound to the carbonylic oxygen of the coumarin group and (b) the triiodide interacting with the π system of the coumarin moiety while Li+ binds to the carbonylic oxygen of the dye, roughly corresponding to the binding geometry of I3− in the absence of the Li+ cation. The calculated Li+−O distances are 1.93 and 1.90 Å when I3− interacts with the π system of the donor group

Figure 5. Optimized molecular structures of the NKX2587−I3−−Li+ complexes in two different geometries (a) and (b). Iodine atoms are in purple, carbon in green, sulfur in yellow, nitrogen in blue, oxygen in red, hydrogen in white. and lithium in gray.

and with Li+, respectively, indicating a modest interaction between triiodide and dye-bound lithium ions, which are in fact predicted to lie at a distance of ca. 5−6 Å. The presence of the Li+ cation bound to the dye slightly affects the interaction between I3− and the coumarin group (Figure 5b), reducing from 4.98 to 4.71 Å the distance between the I atom (in the middle) and the N atom of the coumarin group. We found at both the B3LYP and MP2 level of calculation in solution the (b) adduct of Figure 5 to be more stable than (a) by 2.3 and 3.6 kcal/mol, with a calculated MP2 binding energy with respect to the NKX2587−Li+ and −I3− isolated species of 3.4 kcal/mol. Therefore, the presence of the lithium ion bound to the electron-rich carbonyl oxygen does not have a relevant role in the formation of dye−I3− complexes: the preferred interaction remains that established with the π system of the coumarin moiety, and thus it turns out to be nonspecific within the class of coumarin dyes. Summarizing, the present data are fully in line with previous findings,15 indicating the oxygen atoms as the preferred binding sites for I2. The formation of dye−I2 or dye−I3− complexes in proximity of the TiO2 surface can play a crucial role in determining the short lifetimes measured for certain organic dyes containing oxygen atoms in the donor and bridge units.15 3.2. Adsorption on TiO2, Electronic Structure, and Recombination Kinetics. Upon establishing the affinity of sulfur, oxygen, and nitrogen atoms in the stand-alone dyes to bind I2, now we move to model the dye/semiconductor interface, discussing the dye adsorption onto TiO2, the effective distances of the iodine from the TiO2 surface, and the possible effect of dye-bound Li+ cations on the energetics of the TiO2 conduction band. As displayed in Figures 6 and 7 and in line with previous results obtained for analogous dyes bearing a cyanoacrylic anchoring group,15,69−73 the dyes have a bidentate coordination to the TiO2 surface. The Car−Parrinello dynamics, carried out for a total simulation time of 3.2 ps (movie supplied in Supporting Information) shows that the NKX2587 dye anchored to the (TiO2)82 cluster (Figure 6B) remains on average perpendicular to the TiO2 surface, without significant geometrical distortions with respect to the optimized adsorption geometry at 0 K and hence without a significant change in the occupied volume as well as in the average distances of the possible I2 binding sites from the surface 5969

dx.doi.org/10.1021/jp300095z | J. Phys. Chem. C 2012, 116, 5965−5973

The Journal of Physical Chemistry C

Article

Figure 6. (A) Distance (in Å) of the NKX2587 carbonyl oxygen from the TiO2 surface against the simulation time for the NKX2587@(TiO2)82 system; the average distance is reported in black. (B) Molecular structure of the NKX2587@(TiO2)82 system.

Figure 7. Optimized molecular structures of the (a) O−I2 adduct of NKX2587@(TiO2)38; (b) NKX2587−Li+@(TiO2)38; (c) NKX2697− Li+@(TiO2)38, and (d) MK3@(TiO2)38 systems.

compared to a “static model” derived from local geometry optimizations. In fact, as shown in Figure 6A, the variation of the distance of the carbonyl oxygen from the surface during simulation time of ca. 3 ps is about 0.6 Å, going from a minimum value of 12.2 to a maximum of 11.6 Å, with an average value of 11.88 Å. This is not surprising considering the relatively small size and the rather rigid structure of the NKX2587 dye. We also optimized the structure of the O−I2 adduct of NKX2587@(TiO2)38, and as displayed in Figure 7a, this structure shows negligible geometrical rearrangements of the O−I2 configuration due to the presence of the substrate: the O−I distance is predicted to be 2.85 Å compared to the value of 2.84 Å reported in Table 1; the C−I−O angle increases from 120° to ca. 150°, and the distance from the surface is about 10 Å, as predicted by the Car−Parrinello dynamics simulation. Therefore, a reasonable approximation of the average distance of the bonded I2 from the surface oxide can be obtained by taking the distances from the TiO2 cluster of the various I2 binding sites in the dye@(TiO2)38 optimized structures at 0 K (Figure 7a). The distance of the I2 eventually

bonded by the CN group is about 5 Å, while the carbonyl oxygens, which are the predominant binding sites, lie at ca. 11 and 17 Å from the titania in NKX2587@TiO2 and NKX2696@ TiO2, respectively (a, b, and c in Figure 7). The sulfur atoms in NKX2697 are at about 8, 11, and 14 Å from the TiO2, while for the more distorted MK3 dye they turn out to be slightly closer to the surface (7, 10, and 13 Å, respectively). As discussed by Miyashita et al.,13 measuring the electron lifetime at high [I−] concentration, where the dye cation is fast regenerated and the back recombination with the cation is thought to be negligible, gives an indication of the tendency of differently sensitized cells to be liable to accelerated recombination with the electrolyte. The measured lifetime at high [I−] concentration13 showed that the NKX2587 sensitized cells have shorter lifetime than those based on NKX2697, but their lifetime was comparable to that obtained for MK3. Interestingly, the difference in the lifetimes of the two coumarin dyes at low LiI concentration is even larger, suggesting a faster rate of the back recombination with the dye cation for NKX2587, possibly related to differences in the reorganization 5970

dx.doi.org/10.1021/jp300095z | J. Phys. Chem. C 2012, 116, 5965−5973

The Journal of Physical Chemistry C

Article

Figure 8. Plots of projected density of states (TiO2 cluster contribution) for NKX2587@(TiO2)38 (left side) and NKX2697@(TiO2)38 (right side) with (blue line) and without (black line) the Li+ bound to the carbonylic oxygen of the coumarin group.

energy, λ, as well as to a nonefficient charge displacement due to the small donor−acceptor separation. The increasing distance of the carbonyl oxygen from the titania going from NKX2587@TiO2 (≈11 Å) to NKX2697 (≈17 Å) and hence the increased I2 concentration in proximity of the oxide surface might account for the differences in the lifetime values measured in ref 13 for NKX2587-sensitized solar cells compared to those employing the longer NKX2697 homologues. Anyway, a model solely based on the distance of the I2 binding atoms from the surface cannot provide an explanation for the considerably shorter electron lifetime measured for MK3 with respect to that obtained with NKX2697 since the dyes have the same number of thiophene rings and most importantly the former does not contain I2binding oxygen atoms. As shown by the shorter lifetimes measured at lower dye loading,13 the formation of a compact dye layer can effectively protect the TiO2 surface from the contact with the electrolyte, preventing the recombination processes; this blocking effect is enhanced by the introduction of hydrophobic alkyl chains in the dye structure as well as by the tendency of the dye to aggregate upon adsorption on the TiO2. As we have previously discussed, while the coumarin dyes have a planar structure, in the carbazole MK3 dye, the donor and the bridge units turn out to be distorted by about 30 degrees. One can therefore expect a more compact packing for the planar NKX2697, favored by π−π stacking interactions, with respect to MK3, for which a less dense adsorbate layer can be hypothesized. The marked difference in the electron lifetimes might therefore be partially attributed to an accelerated recombination with the electrolyte, which in MK3-sensitized cells can more easily permeate to the oxide surface. Finally we discuss the effect of the Li+ coordination to the carbonyl oxygen on the Density of States (DOS) profiles for the NKX2587 and NKX2697 coumarin sensitizers (Figure 8); here we specifically refer to the partial TiO2 DOS. The influence of various cations upon the photovoltaic performance of DSCs has been widely investigated in the last years,42,74−77 and in particular, due to its small size, Li+ is supposed to intercalate and possibly strongly adsorb to the oxide surface, resulting in a positive shift of the TiO2 conduction band and

hence in a drastic drop in the open-circuit voltage of the cell. As we have discussed in the previous section, the carbonyl oxygen of the coumarin unit can effectively coordinate Li+, although its presence does not enhance the tendency of the dye to bind I3−, which preferably interacts with the π system of the coumarin. The ability of specific atoms or groups of atoms to bind Li+ has been previously reported by Kuang et al.,77 which proposed the synthesis of an “ion coordinating” sensitizer able to separate the site for Li+ adsorption from the TiO2 surface and hence to reduce the drastic Voc drop associated to Li+ interaction with the semiconductor. Comparing the DOS of the TiO2 sensitized by the two coumarin dyes (Figure 8), one can notice that in the shorter NKX2587 molecule, where the Li+ is bound at a distance from the oxide surface of about 10−11 Å, a slight CB upshift of ca. 60 meV is obtained (blue line) with respect to the DOS curve calculated without the Li+ (black line); this effect is negligible for the longer NKX2697 coumarin, where the Li+ coordinating oxygen is at about 17 Å from the TiO2 surface and its effect on the TiO2 DOS is effectively screened out by the polar solvent environment. This observation is consistent with the conduction band energy upshift of ca. 40 mV observed for NKX2587-sensitized TiO2 in ref 13. We notice, however, that these results must be taken with some care considering that the CB shifts are also linearly dependent on the dye coverage, which is obviously not taken into account here.

4. CONCLUSIONS We have presented a computational investigation based upon DFT and MP2 calculations aimed to elucidate the possible formation of dye−I2 complexes and their implications in the recombination processes of the injected electrons with the electrolyte. We have examined the dye/electrolyte interaction, for a group of three metal free sensitizers, namely, two coumarin and one carbazole dyes, selected among those investigated in a recent experimental work by Miyashita et al.13 The stability analysis of the various dye/I2 adducts in the gas phase and in acetonitrile solution, in line with the findings of ref 15, indicates that oxygen atoms are the preferred binding sites for I2, while I3− tends to interact with the π system of the donor unit, thus giving not-specific interactions. We also found stable coordination of Li+ to the carbonylic oxygen of the coumarin 5971

dx.doi.org/10.1021/jp300095z | J. Phys. Chem. C 2012, 116, 5965−5973

The Journal of Physical Chemistry C

Article

group, but our calculations showed that the Li+ binding does not attract I3− anions, which despite the presence of Li+ preferably interacts with the π system of the coumarin moiety. By performing full geometry optimizations for the adsorbed dyes at 0 K and a Car−Parrinello dynamics simulation for the NKX2587 coumarin dye, we suggest that the increasing of the distance of the carbonyl oxygen from the titania going from NKX2587@TiO2 (ca. 11 Å) to NKX2697 (ca. 17 Å) could be the reason for the differences in the lifetime values measured in ref 13 for NKX2587-sensitized solar cells compared to those employing the longer NKX2697 homologues. The same reason might explain the small conduction band upshift in TiO2 sensitized by NKX2587@TiO2 when Li+ is coordinated to the carbonyl oxygen of the coumarin group. Then the surprisingly short lifetime measured for the carbazole MK3 dye, having the same number of thiophene rings as NKX2697 but a different donor unit, has been attributed to its nonplanar molecular structure, which implies a less compact dye layer and hence a larger accessibility to the oxide surface for the electrolyte. Overall, our results provide a rationale for the subtle differences observed in the photovoltaic performances of DSCs employing the considered dyes, paving the way to the design of new and more efficient organic dye sensitizers.



(11) O’Regan, B. C.; López-Duarte, I.; Martínez-Díaz, V.; Forneli, A.; Albero, J.; Morandeira, A.; Palomares, E.; Torres, T.; Durrant, J. R. J. Am. Chem. Soc. 2008, 130 (10), 2906−2907. (12) Wiberg, J.; Marinado, T.; Hagberg, D. P.; Sun, L.; Hagfeldt, A.; Albinsson, B. J. Phys. Chem. C 2009, 113 (9), 3881−3886. (13) Miyashita, M.; Sunahara, K.; Nishikawa, K.; Uemura, Y.; Koumura, N.; Hara, K.; Mori, A.; Abe, T.; Suzuki, E.; Mori, S. J. Am. Chem. Soc. 2008, 130, 17874−17881. (14) Pastore, M.; De Angelis, F. ACS Nano 2010, 4 (1), 556−562. (15) Planells, M.; Pellejà, L.; Clifford, J. N.; Pastore, M.; De Angelis, F.; López, N.; Marder, S. R.; Palomares, E. Energy Environ. Sci. 2011, 4, 1820−1829. (16) Hagberg, D. P.; Yum, J.-H.; Lee, H.; De Angelis, F.; Marinado, T.; Karlsson, K. M.; Humphry-Baker, R.; Sun, L.; Hagfeldt, A.; Grätzel, M.; Nazeeruddin, M. K. J. Am. Chem. Soc. 2008, 130, 6259−6266. (17) Boschloo, G.; Hagfeldt, A. Acc. Chem. Res. 2009, 42 (11), 1819− 1826. (18) Grätzel, M. J. Photochem. Photobiol. A 2004, 164 (1−3), 3−14. (19) Bai, Y.; Zhang, J.; Zhou, D.; Wang, Y.; Zhang, M.; Wang, P. J. Am. Chem. Soc. 2011, 133 (30), 11442−11445. (20) Zhang, M.; Liu, J.; Wang, Y.; Zhou, D.; Wang, P. Chem. Sci. 2011, 2, 1401−1406. (21) Ito, S.; Miura, H.; Uchida, S.; Takata, M.; Sumioka, K.; Liska, P.; Comte, P.; Pechy, P.; Grätzel, M. Chem. Commun. 2008, 41, 5194− 5196. (22) Zeng, W.; Cao, Y.; Bai, Y.; Wang, Y.; Shi, Y.; Zhang, M.; Wang, F.; Pan, C.; Wang, P. Chem. Mater. 2010, 22 (5), 1915−1925. (23) Daeneke, T.; Kwon, T.-H.; Holmes, A. B.; Duffy, N. W.; Bach, U.; Spiccia, L. Nat. Chem. 2011, 3, 213−217. (24) Feldt, S. M.; Gibson, E. A.; Gabrielsson, E.; Sun, L.; Boschloo, G.; Hagfeldt, A. J. Am. Chem. Soc. 2010, 132, 16714−16724. (25) Feldt, S. M.; Cappel, U. B.; Johansson, E. M. J.; Boschloo, G.; Hagfeldt, A. J. Phys. Chem. C 2010, 114, 10551−10558. (26) Cameron, P. J.; Peter, L. M.; Zakeeruddin, S. M.; Grätzel, M. Coord. Chem. Rev. 2004, 248, 1447−1453. (27) Yella, A.; Lee, H.-W.; Tsao, H. N.; Yi, C.; Chandiran, A. K.; Nazeeruddin, M. K.; Diau, E. W.-G.; Yeh, C.-Y.; Grätzel, M. Science 2011, 4, 629−634. (28) Tatay, S.; Haque, S. A.; O’Regan, B.; Durrant, J. R.; Verhees, W. J. H.; Kroon, J. M.; Vidal-Ferran, A.; Gavina, P.; Palomares, E. J. Mater. Chem. 2007, 17 (29), 3037−3044. (29) Sayama, K.; Tsukagoshi, S.; Hara, K.; Ohga, Y.; Shinpou, A.; Abe, Y.; Suga, S.; Arakawa, H. J. Phys. Chem. B 2002, 106 (6), 1363− 1371. (30) Burfeindt, B.; Hannappel, T.; Storck, W.; Willig, F. J. Phys. Chem. 1996, 100 (41), 16463−16465. (31) Liu, D.; Fessenden, R. W.; Hug, G. L.; Kamat, P. V. J. Phys. Chem. B 1997, 101 (14), 2583−2590. (32) Tian, H.; Yang, X.; Chen, R.; Zhang, R.; Hagfeldt, A.; Sun, L. J. Phys. Chem. C 2008, 112, 11023−11033. (33) Kay, A.; Grätzel, M. J. Phys. Chem. 1993, 97 (23), 6272−6277. (34) Liang, Y.; Peng, B.; Chen, J. J. Phys. Chem. C 2010, 114 (24), 10992−10998. (35) Kim, J.-J.; Lim, K.; Choi, H.; Fan, S.; Kang, M.-S.; Gao, G.; Kang, H. S.; Ko, J. Inorg. Chem. 2010, 49 (18), 8351−8357. (36) Koumura, N.; Wang, Z.-S.; Mori, S.; Miyashita, M.; Suzuki, E.; Hara, K. J. Am. Chem. Soc. 2006, 128 (44), 14256−14257. (37) Wang, Z. S.; Koumura, N.; Cui, Y.; Takahashi, M.; Sekiguchi, H.; Mori, A.; Kubo, T.; Furube, A.; Hara, K. Chem. Mater. 2008, 20 (12), 3993−4003. (38) Nishida, J.-i.; Masuko, T.; Cui, Y.; Hara, K.; Shibuya, H.; Ihara, M.; Hosoyama, T.; Goto, R.; Mori, S.; Yamashita, Y. J . Phys. Chem. C 2010, 114 (41), 17920−17925. (39) Barea, E. M.; Zafer, C.; Gultekin, B.; Aydin, B.; Koyuncu, S.; Icli, S.; Santiago, F. F.; Bisquert, J. J. Phys. Chem. C 2010, 114 (46), 19840−19848. (40) Prezhdo, O. V.; Duncan, W. R.; Prezhdo, V. V. Acc. Chem. Res. 2008, 41 (2), 339−348.

ASSOCIATED CONTENT

S Supporting Information *

The movie of the Car−Parrinello dynamics, carried out for a total simulation time of 3.2 ps, of the NKX2587 dye anchored to the (TiO2)82 cluster is provided. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]; [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS We thank FP7-NMP-2009 project 246124 “SANS”, Fondazione Istituto Italiano di Tecnologia, Platform Computation, Project Seed 2009 “HELYOS”, and CNR-EFOR for financial support.



REFERENCES

(1) O’Regan, B.; Grätzel, M. Nature 1991, 353 (6346), 737−740. (2) Nazeeruddin, M. K.; De Angelis, F.; Fantacci, S.; Selloni, A.; Viscardi, G.; Liska, P.; Ito, S.; Takeru, B.; Grätzel, M. J. Am. Chem. Soc. 2005, 127, 16835−16847. (3) Grätzel, M. Acc. Chem. Res. 2009, 42 (11), 1788−1798. (4) Listorti, A.; O’Regan, B.; Durrant, J. R. Chem. Mater. 2011, 23 (15), 3381−3399. (5) Hagfeldt, A.; Boschloo, G.; Sun, L.; Kloo, L.; Pettersson, H. Chem. Rev. 2010, 110 (11), 6595−6663. (6) Pastore, M.; Mosconi, E.; De Angelis, F.; Grätzel, M. J. Phys. Chem. C 2010, 114 (15), 7205−7212. (7) De Angelis, F.; Fantacci, S.; Mosconi, E.; Nazeeruddin, M. K.; Grätzel, M. J . Phys. Chem. C 2011, 115 (17), 8825−8831. (8) Pastore, M.; Fantacci, S.; De Angelis, F. J. Phys. Chem. C 2010, 114 (51), 22742−22750. (9) De Angelis, F.; Fantacci, S.; Selloni, A. Nanotechnology 2008, 19, 424002. (10) O’Regan, B. C.; Walley, K.; Juozapavicius, M.; Anderson, A. Y.; Matar, F.; Ghaddar, T.; Zakeeruddin, S. M.; Klein, C.; Durrant, J. R. J. Am. Chem. Soc. 2009, 131 (10), 3541−3548. 5972

dx.doi.org/10.1021/jp300095z | J. Phys. Chem. C 2012, 116, 5965−5973

The Journal of Physical Chemistry C

Article

(41) O’Regan, B.; Moser, J.; Anderson, M.; Grätzel, M. J. Phys. Chem. 1990, 94, 8720−8726. (42) Nakade, S.; Makimoto, Y.; Kubo, W.; Kitamura, T.; Wada, Y.; Yanagida, S. J. Phys. Chem. B 2005, 109 (8), 3488−3493. (43) Ardo, S.; Meyer, G. J. Chem. Soc. Rev. 2009, 38 (1), 115−164. (44) Haque, S. A.; Handa, S.; Peter, K.; Palomares, E.; Thelakkat, M.; Durrant, J. R. Angew. Chem., Int. Ed. 2005, 44, 5740−5744. (45) Ondersma, J. W.; Hamann, T. W. J. Am. Chem. Soc. 2011, 133 (21), 8264−8271. (46) Bisquert, J.; Zaban, A.; Salvador, P. J. Phys. Chem. B 2002, 106 (34), 8774−8782. (47) Hara, K.; Miyamoto, K.; Abe, Y.; Yanagida, M. J. Phys. Chem. B 2005, 109 (50), 23776−23778. (48) Burke, A.; Ito, S.; Snaith, H. J.; Bach, U.; Kwiatkowski, J.; Grätzel, M. Nano Lett. 2008, 8 (4), 977−981. (49) Tuikka, M.; Hirva, P.; Rissanen, K.; Korppi-Tommola, J.; Haukka, M. Chem. Commun. 2011, 47, 4499−4501. (50) Privalov, T.; Boschloo, G.; Hagfeldt, A.; Svensson, P. H.; Kloo, L. J. Phys. Chem. C 2009, 113 (2), 783−790. (51) Hu, C.-H.; Asaduzzaman, A. M.; Schreckenbach, G. J. Phys. Chem. C 2010, 114 (35), 15165−15173. (52) Schiffmann, F.; VandeVondele, J.; Hutter, J.; Urakawa, A.; Wirz, R.; Baiker, A. Proc. Nat. Acad. Sci. U.S.A. 2010, 107 (11), 4830−4833. (53) Nyhlen, J.; Boschloo, G.; Hagfeldt, A.; Kloo, L.; Privalov, T. Chem. Phys. Chem. 2010, 11 (9), 1858−1862. (54) Kusama, H.; Sugihara, H.; Sayama, K. J. Phys. Chem. C 2011, 115 (5), 2544−2552. (55) Lobello, M. G.; Fantacci, S.; De Angelis, F. J . Phys. Chem. C 2011, 115 (38), 18863−18872. (56) Cossi, M.; Rega, N.; Scalmani, G.; Barone, V. J. Comput. Chem. 2003, 24, 669−681. (57) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Montgomery, J. A.; Vreven, J., T.; Kudin, K. N.; Burant, J. C. et al. Gaussian 03, Revision B05; Gaussian Inc.: Pittsburgh PA, 2003. (58) Grimme, S.; Antony, J.; Ehrlich, S.; Krieg, H. J. Chem. Phys. 2010, 132 (15), 154104. (59) Grimme, S.; Ehrlich, S.; Goerigk, L. J. Comput. Chem. 2011, 32 (7), 1456−1465. (60) Zhao, Y.; Truhlar, D. G. Theor. Chem. Acc. 2008, 120, 215−241. (61) te Velde, G.; Bickelhaupt, F. M.; Baerends, E. J.; Fonseca Guerra, C.; van Gisbergen, S. J. A.; Snijders, J. G.; Ziegler, T. J. Comput. Chem. 2001, 22 (9), 931−967. (62) Becke, A. D. Phys. Rev. A 1988, 38 (6), 3098−3100. (63) Perdew, J. P. Phys. Rev. B 1986, 33, 8822−8824. (64) Car, R.; Parrinello, M. Phys. Rev. Lett. 1985, 55 (22), 2471− 2474. (65) Giannozzi, P.; Baroni, S.; Bonini, N.; Calandra, M.; Car, R.; Cavazzoni, C.; Ceresoli, D.; Chiarotti, G. L.; Cococcioni, M.; Dabo, I.; et al. J. Phys.: Condens. Matter 2009, 21, 395502. (66) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77 (18), 3865−3868. (67) Giannozzi, P.; Angelis, F. D.; Car, R. J. Chem. Phys. 2004, 120 (13), 5903−5915. (68) Furube, A.; Katoh, R.; Hara, K.; Sato, T.; Murata, S.; Arakawa, H.; Tachiya, M. J. Phys. Chem. B 2005, 109 (34), 16406−16414. (69) Pastore, M.; De Angelis, F. Phys. Chem. Chem. Phys. 2011, 14, 920−928. (70) Chen, P.; Yum, J. H.; Angelis, F. D.; Mosconi, E.; Fantacci, S.; Moon, S.-J.; Baker, R. H.; Ko, J.; Nazeeruddin, M. K.; Grätzel, M. Nano Lett. 2009, 9 (6), 2487−2492. (71) Lee, D. H.; Lee, M. J.; Song, H. M.; Song, B. J.; Seo, K. D.; Pastore, M.; Anselmi, C.; Fantacci, S.; De Angelis, F.; Nazeeruddin, M. K.; Gräetzel, M.; Kim, H. K. Dyes Pigm. 2011, 91 (2), 192−198. (72) Rocca, D.; Gebauer, R.; De Angelis, F.; Nazeeruddin, M. K.; Baroni, S. Chem. Phys. Lett. 2009, 475, 49−53. (73) Hara, K.; Kurashige, M.; Dan-oh, Y.; Kasada, C.; Shinpo, A.; Suga, S.; Sayama, K.; Arakawa, K. New J. Chem. 2003, 27, 783−785.

(74) Pelet, S.; Moser, J.-E.; Grätzel, M. J . Phys. Chem. B 2000, 104 (8), 1791−1795. (75) Kambe, S.; Nakade, S.; Kitamura, T.; Wada, Y.; Yanagida, S. J. Phys. Chem. B 2002, 106 (11), 2967−2972. (76) Nakade, S.; Kambe, S.; Kitamura, T.; Wada, Y.; Yanagida, M. J . Phys. Chem. B 2001, 105 (38), 9150−9152. (77) Kuang, D.; Klein, C.; Snaith, H. J.; Moser, J.-E.; HumphryBaker, R.; Comte, P.; Zakeeruddin, S. M.; Grätzel, M. Nano Lett. 2006, 6 (4), 769−773.

5973

dx.doi.org/10.1021/jp300095z | J. Phys. Chem. C 2012, 116, 5965−5973