ARTICLES PUBLISHED ONLINE: 11 OCTOBER 2009 | DOI: 10.1038/NMAT2548 On the origin of the open-circuit voltage of polymer–fullerene solar cells Koen Vandewal 1 * , Kristofer Tvingstedt 2 , Abay Gadisa 1 , Olle Inganäs 2 and Jean V. Manca 1 The increasing amount of research on solution-processable, organic donor–acceptor bulk heterojunction photovoltaic systems, based on blends of conjugated polymers and fullerenes has resulted in devices with an overall power-conversion efficiency of 6%. For the best devices, absorbed photon-to-electron quantum efficiencies approaching 100% have been shown. Besides the produced current, the overall efficiency depends critically on the generated photovoltage. Therefore, understanding and optimization of the open-circuit voltage (V oc ) of organic solar cells is of high importance. Here, we demonstrate that charge-transfer absorption and emission are shown to be related to each other and V oc in accordance with the assumptions of the detailed balance and quasi-equilibrium theory. We underline the importance of the weak ground-state interaction between the polymer and the fullerene and we confirm that V oc is determined by the formation of these states. Our work further suggests alternative pathways to improve V oc of donor–acceptor devices. T he most successful solution-processable organic solar cells use a C 60 or C 70 fullerene derivative as an electron acceptor blended with a conjugated polymer 1–3 . In the field, attempts have been made to derive upper limits for the efficiency of this type of polymer–fullerene photovoltaic device, albeit with empirical arguments related to the details of the origin of the open-circuit voltage 3–5 (V oc ). However, as energy is converted from one form (radiation) to another (electrical), fundamental losses should be taken into account and it should be possible to derive an upper limit for V oc , purely on the basis of thermodynamic considerations. For single absorber materials, this fundamental question was answered in 1961 in a seminal paper by Shockley and Queisser 6 . Their analysis was based on the detailed balance of absorption and emission events from the solar cell, a ‘grey’ body at the surface of the Earth, illuminated by the Sun, a black body of much higher temperature. This allowed the derivation of an expression for V oc as a function of the material’s bandgap. It was found that V oc is maximal for the ideal case in which the charges can recombine only radiatively. According to this reasoning, it is clear that the V oc of polymer– fullerene devices has not reached its thermodynamic maximal value yet. This value would be reached if the only recombination mechanism at open-circuit conditions is a radiative one 6 . As a result of the severe luminescence quenching in material blends yielding a substantial charge generation, it is clear that radiative recombination is just a small fraction of the total recombination, and a reduction of the maximum obtainable V oc is expected. In fact, no correlations of V oc with the optical gap of any of the blend constituents, as predicted by Shockley and Queisser 6 , are observed. Instead, V oc is found to scale with the difference between the highest occupied molecular orbital energy of the donor and the lowest unoccupied molecular orbital energy of the fullerene acceptor 4,7 . This leads to the conclusion that in this type of solar cell, the V oc is determined by recombination at the donor/acceptor interface 8–12 . Recently, for some polymer–fullerene blends, radiative interface recombination was observed. The presence of a weak emission signal, redshifted compared to the pure components, was detected in the photoluminescence and electroluminescence spectra and 1 IMEC-IMOMEC, vzw, Institute for Materials Research, Hasselt University, Wetenschapspark 1, 3590 Diepenbeek, Belgium, 2 Biomolecular and Organic Electronics, Center of Organic Electronics (COE), Department of Physics, Chemistry and Biology, Linköping University, 58183 Linköping, Sweden. *e-mail: koen.vandewal@uhasselt.be. was assigned to the emission of interface electron–hole pairs or charge-transfer excitons 13–17 . The signature of this emitting state is also present in the absorption spectrum as a new, weak subgap absorption band in several polymer–fullerene blends used for photovoltaic applications 18–20 . Such absorption bands are typical for the formation of a ground-state charge-transfer complex (CTC) between the polymer and the fullerene. Furthermore, good correlations between the open-circuit voltage and the spectral position of the charge-transfer absorption 20 , photoluminescence 15 or electroluminescence 17 could be made. Here, we show that the electroluminescence and photovoltaic external quantum efficiency spectra in the low-energy, charge- transfer region are related to each other as predicted by the detailed balance approach. Furthermore, it is shown that at V oc , the photocurrent generated by the absorption of sunlight balances with the recombination current, resulting in emission of photons by the excited CTCs. This confirms previous suggestions 10,15,20 , that V oc is determined by the CTC formation between the polymer and the fullerene. To validate the generality of the detailed balance treatment for polymer–fullerene solar cells, blends of five different donor polymers and two fullerene derivatives, that is, [6,6]-phenyl C61 butyric acid methyl ester (PC 61 BM) and [6,6]-phenyl C71 butyric acid methyl ester (PC 71 BM), were investigated. The polymers belong to different conjugated polymer material families, comprising different conjugated backbones. These conjugated polymers are representative of the donor polymers used in polymer–fullerene solar cells explored in the community at present. Their chemical structures are shown in Fig. 1. Devices based on poly[2-methoxy-5-(30,70-dimethyloctyloxy)- 1,4-phenylene vinylene] (MDMO-PPV) and poly[2,7-(9-di-octyl- fluorene)-alt-5,5-(4 0 ,7 0 -di-2-thienyl-2 0 ,1 0 ,3 0 benzothiadiazole)] (APFO3) were prepared using different polymer/fullerene sto- ichiometries. Optimal devices were obtained using a 1:4 poly- mer/fullerene weight ratio, resulting in a power conversion efficiency of ∼2% and ∼2.5% respectively. At a lower fullerene content, the photogenerated current becomes lower and the NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials 1 © 2009 Macmillan Publishers Limited. All rights reserved.