Designing intrinsically photostable low band gap polymers: a smart tool combining EPR spectroscopy and DFT calculations

Université de Pau et des Pays de l'Adou Président Angot, 64053 Pau, France. E-mail Université Blaise Pascal, Institut de Photochimie, BP 10448, F-63000 Clermont-F Université Clermont Auvergne, SIGMA Cle Ferrand, BP 10448, F-63000 Clermont-Ferra CNRS, UMR 6296, ICCF, Équipe Photochim Chemical Engineering and Applied Chemistr UK Aston Materials Centre, Aston University, B CNRS, IPREM (EPCP, CNRS-UMR 5254), 64 Université Grenoble Alpes, INAC-SPRAM, F CNRS, INAC-SPRAM, F-38000 Grenoble, Fr CEA, INAC-SPRAM, F-38000 Grenoble, Fran † Electronic supplementary informa 10.1039/c6ta05455b Cite this: J. Mater. Chem. A, 2016, 4, 15647


Introduction
Incorporating stable materials is a prerequisite in creating long-lasting, stable devices.This is particularly pertinent for materials used to make organic solar cells because polymers used in the active layer are generally unstable under light exposure.Only very few photo-induced defects formed within the polymer are required to drastically impair solar cell performance, since such defects deleteriously affect the charge generation process. 1 Besides photovoltaic conversion, other photophysical or photochemical processes can occur in the excited state.Among these latter processes, rearrangement, chain scission, and crosslinking all result in modication of the chemical structure of the polymer (see Scheme S1 in ESI †). 2 Identifying the photodegradation mechanism(s) occurring is not easy for polymers with simple structures and even more troublesome for low band gap (LBG) polymers, which are now extensively used in efficient organic solar cells. 3,4Nevertheless, it is essential to design improved, efficient and stable LBG polymer structures to enable the fabrication of stable organic solar cells.The conjugated skeleton of LBG polymers is based on the alternation of donor and acceptor units on which alkyl side chains can be attached, where the attachments themselves may contain different types of heteroatoms.The alkyl side chains can be linear or branched, and be attached to the donor and/or acceptor units of the polymer.Based on these multiple combinations, countless types of LBG polymers can be synthesized. 5,6ne of the main problems in addressing photodegradation is that there are no sufficiently sensitive tools that are able to detect at low concentrations both small structural changes in the polymer (at the molecular scale) and the primary species at the origin of the photochemical process.In fact, a dramatic decrease in organic solar cell efficiency is generally associated with only very slight changes in the 'usual' spectrometric parameters (Raman, UV-visible, XPS.) of the polymer. 7Furthermore, such spectroscopies cannot reveal the primary species at the origin of degradation.They only allow detection of the structural/chemical changes due to subsequent sequential steps of the degradation process.Thus, they do not give information on the mechanism behind such processes. 8It is therefore vital to develop a technique that is able to unambiguously detect, and identify, the primary species involved in photochemical processes.Electronic Paramagnetic Resonance (EPR) combines simple implementation of global tracing methods (a few milligrams of the macroscopic sample deposited on the same substrate, under the same conditions as those in real organic solar cells, are directly observed) with very high sensitivity. 9In addition, the more sophisticated 'EPR spin trapping method' allows the detection of primary unstable free radicals resulting from direct (intrinsic) photochemical processes.Density Functional Theory (DFT) calculations allow one to identify weak points in a chemical structure in terms of their photochemical reactivity (via the binding energies of different bonds).Herein, we exploit the remarkable synergy of EPR spectroscopy and DFT modelling to obtain information on photochemical mechanisms, 10,11 thus allowing the design of push-pull polymers adapted to long-life organic solar cells without, most importantly, the necessity of testing them in real devices.
In the work herein, the more sophisticated 'Spin Trap Method' has been used in place of 'basic' EPR.The 'Spin Trap Method' consists of detecting radicals of short life-time (in this case the radicals eventually formed during the polymer photodegradation process) by trapping them with a diamagnetic molecule to produce a stable paramagnetic adduct; it was developed in parallel in 1968 by Janzen 12,13 and Lagercrantz. 14his method has been used recently to study the formation of the oxygen superoxide anion radical, which is trapped at the very beginning of illumination and assumed to be responsible for the degradation of poly(3-hexylthiophene) (P3HT) in chlorobenzene in the presence of oxygen. 15In this study, the authors did not report radical trapping arising from polymer degradation.On closer inspection of their data, and in agreement with the authors, only OHc radicals were trapped at the beginning of the experiment.However, aer a few minutes of irradiation, a dramatic change was observed in the EPR spectrum.We ascribe this to the trapping of other radicals, such as those arising from degradation of the polymer.Therefore, the present study focuses on what happens aer 15 minutes of illumination by solar simulated light in order to detect eventual radicals arising from the intrinsic polymer photochemical processes with the aim of further probing polymer photo-degradation mechanisms.

Sample preparation.
Films on KBr substrates were prepared via doctor blading (Erichsen Coatmaster 809 MC) from a 0.9% (w/w) polymer solution in para-xylene : orthodichlorobenzene (7 : 1).Films on PET/ITO substrates were dropcast from a 1% polymer solution (w/w) in ortho-dichlorobenzene.For the purpose of the EPR experiments, polymer lms on PET/ITO were introduced in EPR Quartz tubes (Ø ¼ 3 mm) and wetted by a solution of DMPO in ethanol (50 g L À1 ).
2.2.2 UV-vis and IR monitoring before and aer irradiation.UV-vis absorption spectra were obtained using a Shimadzu UV-2600 spectrophotometer with an integration sphere.Infrared transmission spectra were recorded by a Nicolet 760 Magna spectrophotometer purged with dry air (32 acquisitions with 4 cm À1 resolution).Irradiation of deposits on KBr substrates was performed in a Suntest CPS/XLS Atlas device, provided with a xenon lamp from ATLAS (NXE1700) congured at 750 W m À2 in the UV-visible domain.A lter cuts off IR irradiation and UV photons below 300 nm.A cryostat maintained the Black Standard Temperature (BST) at 60 C, corresponding to a chamber temperature of 35 C. Samples irradiated in the absence of oxygen were encapsulated in borosilicate tubes under secondary vacuum (10 À6 Pa).
2.2.3 EPR analysis before and aer illumination.EPR spectra were recorded on an EMX Brucker spectrometer operating at 9 GHz at room temperature.They were simulated with free winsim2002 soware. 16EPR tubes containing polymer lms deposited on PET/ITO were placed in a xed position located 5 cm below the device producing the photons from the solar simulator (so that the center of the simulator corresponded to the center of the sample).The solar simulator was a New Port Oriel, Sol 3A class 3A solar simulator 94023A.Irradiation of the sample was made according to the set-up depicted in Fig. 2.
In the conguration shown in Fig. 2, the sample was exposed to 15 min AM 1.5 light soaking and the substrate was rstly irradiated before the deposited polymer was wetted by the DMPO solution in order to minimize photo-degradation of the DMPO.EPR spectra were recorded under Ar atmosphere aer air exchange inside a glovebox in order to avoid broadening of the line widths due to the presence of oxygen.
2.2.4 Computational details.The photostability of the polymer was evaluated via calculation of the bond dissociation energies of selected bonds in the lateral chains of a DTSTzTz monomer, namely the C-H bonds in both lateral chains (labels a-j in Fig. 1) and the Si-C bond or C-C bonds involved in the attachment of the side chains to the polymer.This was calculated within B3LYP/6-31G** level of theory [17][18][19][20][21][22] using UKS wavefunctions whenever needed.The general system studied is depicted in Fig. 1.The procedure used to evaluate the lability of these hydrogen atoms has been described elsewhere. 23The g-factors were calculated on the B3LYP/6-31G** optimized geometries but using the B3LYP/IGLO-III 24 level of theory using an UKS wave-function.All calculations were performed on ORCA 3.0.3soware. 25Results and discussion

Results
3.1.1UV-vis and IR spectroscopy.Signicant absorbance decrease in the UV-vis spectra of PDTSTzTz could be detected when exposed to light (AM 1.5) under ambient air conditions even aer very short irradiation times ($2% decrease aer 15 min exposure) (see Fig. 3).This indicates a photobleaching of the polymer resulting from the disruption of the p-conjugation of the backbone (e.g. via breakup or saturation of the backbone).
It is also clear that when irradiation is performed in the absence of oxygen, the effect of photodegradation on the UV-visible absorbance is signicantly reduced compared to irradiation in ambient air, as previously evidenced for MDMO-PPV, P3HT and PCDTBT. 26In parallel, IR spectroscopy revealed a decrease of the polymer bands and the formation of oxidation products upon irradiation in the presence of air, whereas no new bands were observed when samples were irradiated in glass tubes sealed under vacuum, conrming that the tubes were fully degassed (see Fig. 3b).
It is noteworthy that photooxidative processes occurring with irradiation in ambient air, resulting in easily identiable oxygen xation on the chemical structure, does not exclude the involvement of pure photochemical processes that are  unaffected by atmospheric conditions.Furthermore, if an efficient encapsulation barrier may suppress, or at least drastically reduce, oxidation processes, pure photochemical processes have to be regarded as the intrinsic fragility of the polymer.Therefore, characterizing exclusively light-induced degradation processes is crucial if one wishes to develop intrinsically more stable materials, which will then allow signicant improvements in device lifetime to be made.The spectroscopic analysis of PDTSTzTz deposits submitted to long duration irradiations in the total absence of oxygen reveals the presence of intrinsic photochemical processes that lead to some modication of the PDTSTzTz backbone structure.In this respect, previous work devoted to other polymers, such as P3HT and PCDTBT, suggested that irradiation in the absence of oxygen provoked homolytical scission between the side chain and conjugated backbone.Successive saturation of the P3HT conjugated skeleton by the alkyl fragments would induce a reduction of the conjugation length of the p-conjugated system, thus explaining the decrease in absorbance. 27Macroradical recombination was proposed in the case of PCDTBT, with parallel formation of low molecular weight alkyl products. 9Moreover, to our knowledge, scientic literature reports no direct photochemical process which would uniquely involve the building blocks of the conjugated skeleton leading to the disruption of p-conjugation.The side-chain cleavage hypotheses presented in the literature for the degradation of P3HT and PCDTBT cannot, however, be validated for PDTSTzTz uniquely on the basis of UV-vis and IR analyses presented herein.
In order to clarify the mechanisms occurring in the early stages of polymer degradation and investigate the possible sidechain cleavage processes, intermediate short-lived free radicals formed during the irradiation process have been studied herein by performing spin trapping experiments.
3.1.2EPR experimental data.In the present study, the spin trap used was the DMPO nitrone, which results in a stable nitroxide spin adduct; the EPR hyperne couplings of which allow the nature of the trapped radicals to be identied.The spin trapping method using DMPO was applied to study the free radicals formed aer 15 min of irradiation of the polymer wetted by a solution of DMPO in ethanol.The spectra of ethanol with DMPO, with and without irradiation under inert atmosphere, have been recorded; no signals were observed.Similarly, the spectra of the polymer without DMPO, with and without irradiation under Ar, were recorded: again no EPR signals were observed in either case (Fig. 4a presents the EPR spectrum of the polymer, without DMPO).Subsequently, the polymer without DMPO, in the presence of oxygen, with and without illumination was characterized.In dark conditions, no signal was observed, while under illumination, a stable signal exhibiting a single line (Fig. 4b) with a g-factor equal to 2.0023 was recorded.
Accordingly, the spectrum of PDTSTzTz wetted by the DMPO solution and illuminated under argon was then recorded.Fig. 4c presents the recorded spectrum together with its simulation created with a single radical species using three hyperne couplings: An ¼ 15.2 G; AH 1 ¼ 22 G and AH 2 ¼ 0.5 G.These hyperne couplings are very similar to those observed for an alkyl radical of Rc type (for example CH 3 CH 2 c in water gives rise to hyperne couplings of AN ¼ 16.3 G, AH 1 ¼ 23.5 G). 28 ‡ The fact that hyperne couplings observed in our case are slightly lower is attributed to the difference in polarity between water and ethanol.Finally, the spectrum of PDTSTzTz, wetted by the DMPO solution, and irradiated under oxygen was recorded; Fig. 4d shows the experimental spectrum together with its simulation/t.The experimental spectrum is satisfactorily tted with two radicals: the rst radical (of low intensity) gives rise to hyperne couplings of AN ¼ 13.4 G, AH 1 ¼ 7.5 G and AH 2 ¼ 1.8 G (six pairs of lines), while the second radical (of higher intensity) gives rise to only one hyperne coupling of AN ¼ 14 G.The hyperne couplings of the rst radical species are very similar to those observed for an alkoxy radical of the type R 0 Oc (for example CH 3 CH 2 Oc in ethanol exhibits AN ¼ 13.5 G, AH 1 ¼ 7.4 G and AH 2 ¼ 1.7 G). 28 The hyperne AN coupling of the second radical ts well with that of the oxidative degradation product of DMPO. 28Both species present a g-factor in the order of 2.006.The EPR data herein are interpreted as follows: rst, the stable spectrum (g ¼ 2.0023) observed for the polymer alone (without DMPO) when illuminated under oxygen is ascribed to the formation of the cation radical of the polymer associated with the superoxide anion radical, as previously reported for other polymers. 29Second, when irradiation of the polymer with DMPO is performed, the observed Rc signal (Fig. 4b) is attributed to a 'small' alkyl molecule and not to a macroradical.In fact, if this signal would correspond to a polymeric structure, the lines would be broader due to the rotational diffusion correlation time, especially the one located at high external magnetic eld. 30,31Thirdly, when irradiation of the polymer with DMPO is performed under oxygen, the observed R 0 Oc (Fig. 4b) results from the oxidation of alkyl radicals.
3.1.3DFT calculations.Firstly, the g-factor of the polaron of PDTSTzTz was calculated.The determined value (2.0023) was in full agreement with that found by EPR (2.0023).As mentioned above, light absorption is able to provoke direct homolytic scission of weak bonds.Thus, we investigated two situations; one in which the DTS moiety loses a 2-ethyl-hexyl chain via direct homolysis of the Si-C bond and the other in which the thiophene ring loses its lateral alkyl chain (a propyl group in this model, see Fig. 1).The bond dissociation energies were found to be 77.76 and 102.31 kcal mol À1 , respectively.The magnitude of these values indicates that the Si-R can be homolitically broken following the mechanism published for polyvinylcarbazole material. 32,33This is an interesting result, since the delocalization of the so-formed radical can take place partially within the heterocycle in question (as presented in Fig. 5), which considerably lowers the bond energy.It is noteworthy that the energy of UV photons is actually sufficient to break this weak bond (71-86 kcal mol À1 for photons in the range of 330-400 nm).
In addition to the homolytic scissions, irradiation of the sample in the presence of oxygen would also lead to photooxidative processes, commonly starting by abstraction of labile hydrogen atoms of the polymer as reported for PCDTBT 9 and PVK. 32,33Using DFT calculations, the most labile hydrogen atoms of PDTSTzTz have been identied.The bond dissociation energies (BDEs) were calculated for the labelled protons in Fig. 1 and the values are given in Table 1.
From the data shown in Table 1, one can see that the hydrogen atom in the alpha position to the thiophene ring is the most labile one, and can therefore be more easily abstracted from the chain, even more so than the hydrogen on the tertiary carbon of the 2-ethyl-hexyl chain, in line with Santos Silva et al. 23 Furthermore, the spin delocalization over the polymer repeat unit which has lost a H radical in the alphaposition of the thiophene units takes place over a large amount of neighbouring double bonds as presented in Fig. 6.Therefore, although the energy of dissociation of the C-H bond is larger than that of Si-C bond, this radical is more delocalized, thus, more stable.

Discussion
The UV-vis absorbance decay of PDTSTzTz lms indicate a higher degree of degradation for samples irradiated in air.Thus, whilst pure photolytic processes can happen both in the presence and absence of oxygen, photooxidative processes are those predominantly leading to the degradation of the polymer.As for photolytic processes (i.e.processes occurring upon irradiation under inert atmosphere), DFT calculations indicate that the Si-R bond can be homolitically broken.Such photochemical homolysis would lead to the formation of two types of radicals: a macro-Sic radical and a cC-R radical, with the latter being unstable as delocalization is not possible (see Fig. 7).This result leads one to assume that the cC-R trapped by DMPO under illumination without oxygen, observed by EPR, is likely to originate from the homolysis of the alkyl chain linked to the Si atom, the so-called R c 2 according to the labelling in Fig. 1; with the calculated g-value for one DMPO adduct being in full agreement with that observed by EPR (g ¼ 2.006).No spectrum corresponding to the macro-Sic radical was observed from the present spin trapping experiments.This is not surprising since, to our knowledge, no adduct of DMPO with Sic has ever been reported in the literature.The so-formed alkyl radicals could saturate the p-conjugated skeleton of the polymer (as suggested for P3HT) 27 and the macroradicals on the Si atom are also susceptible to combination, thus provoking the crosslinking of the irradiated material (as suggested for PCDTBT 9 and PVK) 32,33 .Both processes could be responsible for the change in chemical structure of the polymer leading to a decrease in the UV-vis absorption of the polymer.
Under photooxidative conditions, radicals (such as those created by solely homolytic processes) are able to propagate the chain oxidation process by preferentially abstracting the most labile hydrogen atoms, i.e. those in the alpha position to the thiophene ring according to the data obtained by DFT calculations (Table 1).Irradiation of PDTSTzTz lms in air led to EPR identication of the positive polaron (macro-C + c in Fig. 7) and of alkoxy radicals, in good accordance with these calculations.As depicted in Fig. 7, such alkoxy radicals can stem from the oxidation of the R 1 radical (R 1 Oc) or the aforementioned R c 2 , generated by homolytical scission.Although the EPR spectrum in Fig. 4d has signals ascribed to both R 1 Oc and R 2 Oc (their adducts with DMPO giving similar EPR spectra), evolution and kinetics obtained by UV-vis and IR analysis clearly show that oxidative processes are predominant over intrinsic photochemical processes.Therefore, one can assume that ROc are primarily related to R 1 moieties.
In this work, spin-trapping EPR spectroscopy is shown as a very sensitive tool, even aer very short irradiation times (only 15 min AM 1.5 exposure), to identify the radical species (stable and unstable) resulting from photochemical processes.In combination with DFT calculations, the most likely photochemical processes occurring at these early stages of degradation are summarised in Fig. 7.According to the results herein presented, the absorption of UV-vis light by PDTSTzTz leads to (at least in part) the homolysis of the bond between the Si atom and its side-chain.Interestingly, previous investigations have indicated that replacing the C atom by Si, as a bridging atom to the alkyl-chains, actually enhances the stability of the attached side-chain against photooxidation processes. 11However, according to the study herein presented,  the presence of a Si atom bearing a side-chain does represent the Achilles' heel of the chemical structure of PDTSTzTz; on the one hand, direct degradation can occur at this point under inert conditions via light-induced homolysis of this bond.On the other hand, if oxygen is present, the radicals produced are able to propagate oxidation, thus further degrading the polymer.More generally, irradiation in presence of oxygen is identied as the most destructive for the polymer.DFT calculations demonstrate that the primary steps of this oxidation process occur at the side chain attached to the thiophene ring, characterised by a low C-H binding energy (see Fig. 7).In summary, the work presented identies the combination between EPR and DFT calculations as a smart tool to identify the species at the origin of degradation.Furthermore, our ndings are highly relevant for the organic electronic and photovoltaic communities as for designing polymers with more stable chemical structures.

Conclusions
Most polymers are photo-instable, and the identication of their Achilles' heel is particularly difficult when focusing on complex chemical structures such as LBG polymers.To this end we have demonstrated in the present study that the judicious combination of EPR spectroscopy and DFT calculations is a smart and efficient approach to unambiguously determine radical processes occurring in the early stages of photodegradation of low band gap polymers.Implementing this tool on PDTSTzTz, as a model polymer, we have revealed that it is unstable under light irradiation, as most polymers are, to identify the chemical structure of transient radicals formed during the photodegradation process and consequently to point out the Achilles' heel of the polymer structure.As follow up to this work, DFT modelling could allow one to tune the effect of structural modications (for instance, carbon and germanium-analogue structures) to design a more photostable material.More generally, the smart combined approach demonstrated herein is a way to intuitively expand the tool box of building blocks available to design photostable low band gap polymers, which is a prerequisite for materials which will be submitted to UV-vis light irradiation over the lifetime of the organic solar cell.

Fig. 1
Fig.1Chemical structure of PDTSTzTz, where R 1 is a dodecyl linear chain and R 2 is a 2-ethylhexyl chain.The calculations were performed on a monomer of this structure for which R 1 is a methyl group.The hydrogen atoms used for the determination of the bond dissociation energies have been labelled a-i.

Fig. 2
Fig. 2 Experimental set-up related to sample irradiation for EPR spintrapping analysis.

Fig. 3
Fig. 3 Changes in (a) UV-visible and (b) IR spectra of PDTSTzTz thin films during irradiation in the SUNTEST device under ambient conditions (red) and in the absence of oxygen (blue).

Fig. 4
Fig. 4 EPR spectra of PDTSTzTz not wetted by DMPO solution, and irradiated with the solar simulator (a) under argon and; (b) in the presence of air (g ¼ 2.0023).EPR spectra of PDTSTzTz wetted by DMPO solution and irradiated (c) under argon; (d) in presence of air.Parts (c) and (d) show the experimental spectra (solid blue line) together with their simulations (dotted red line).

Fig. 5
Fig. 5 Spin delocalization over the PDTSTzTz repeat unit which has lost the lateral chain of the Si atom moiety.

Fig. 6
Fig. 6 Spin delocalization over the PDTSTzTz repeat unit which has lost a H radical in the alpha-position of the thiophene units.

Fig. 7
Fig. 7 Schematic representation of the early stages of the light-induced degradation of the PDTSTzTz polymer.