Light-induced degradation of fullerenes in organic solar cells: a case study on TQ1:PC 71 BM † Royal Society of

The stability of organic solar cells (OSCs) is critical for practical applications of this emerging technology. Unfortunately, in spite of intensive investigations, the degradation mechanisms in OSCs have not been clearly understood yet. In this report, we employ a range of spectroscopic and transport measurements, coupled with drift-di ﬀ usion modelling, to investigate the light-induced degradation mechanisms of fullerene-based OSCs. We ﬁ nd that trap states formed in the fullerene phase under illumination play a critical role in the degradation of the open-circuit voltage ( V OC ) in OSCs. Our results indicate that the degradation is intrinsic to the fullerenes in OSCs and that alternative acceptor materials are desired for the development of stable OSCs.


Introduction
Organic solar cells (OSCs) have attracted intensive attention during the past few decades. Very recently, the power conversion efficiency (PCE) of OSCs has reached a high value of above 14%, 1,2 making this technology commercially relevant. However, the stability of OSCs remains an issue for the practical applications of OSCs.
The degradation of OSCs can be caused by different factors, including moisture, oxygen, 3-6 heat, 7-10 and light. [11][12][13][14] Among these different factors, to enhance the stability of OSCs under light is critically important for OSCs, as OSCs are inevitably illuminated during the operation. Light exposure can affect the photovoltaic parameters of OSCs in different ways. For instance, in some material systems mainly the short-circuit current (J SC ) and ll factor (FF) were affected by light exposure, 15 while in some other systems mainly the open-circuit voltage (V OC ) was affected. 16,17 Different mechanisms were found in light-induced degradation processes. For instance, new sub-band gap trap states were identied in polymer donor materials aer exposure to light, and the light-induced new traps can broaden the density of states, decreasing the V OC . 16,17 In addition, the ultraviolet (UV) region of the spectrum can trigger the photochemical reaction between the polymer donor materials and the fullerene, resulting in irreversible degradation of the active layer. 18 For fullerene based OSCs, the fullerene can be dimerized when exposed to light in an inert atmosphere. [19][20][21] The dimerization of the fullerene could result in a signicantly reduced carrier mobility or morphological changes of the active layer, leading to decreased J SC and FF. [19][20][21] In this paper, we investigate the light-induced degradation mechanisms in a classical fullerene-based OSC blend solar cell, i.e. poly[2,3-bis-(3-octyloxyphenyl)quinoxaline-5,8-diyl-altthiophene-2,5-diyl] (TQ1): [6,6]-phenyl-C71-butyric acid methyl ester (PC 71 BM) (as shown in Fig. 1a). We nd that the V OC signicantly decreases during the light-induced degradation process in both conventional and inverted device structures. We attribute the decrease of V OC to non-radiative recombination, as identied by highly sensitive Fourier Transform Photocurrent Spectroscopy (FTPS), electroluminescence (EL) and external quantum efficiency of EL (EQE EL ) measurements. We combine temperature-dependent charge transport measurements and Extended Gaussian Disorder Modelling (EGDM) and nd that the Gaussian disorder width of the electron states increases during the photo-degradation, while that of the holes remains the same. Fourier transform infrared spectroscopy (FTIR) measurements reveal that the fullerenes are oxidized during the degradation process, resulting in electron traps. These different measurements present a consistent picture that the photoinduced oxidation of fullerenes plays a critical role in the degradation of this system, resulting in electron traps and additional non-radiative recombination, which decreases the V OC .

Results and discussion
We employ a conventional device structure, where the active layer is sandwiched between ITO/PEDOT:PSS and LiF/Al. As shown in Fig. 1b, the V OC of the device stored under illumination (without load) shows a signicant decrease, while the one of the device stored in the dark stays constant. The V OC degradation can be due to either the interlayers or the active layer. 22,23 In order to distinguish between the contribution from the interlayers and that from the active layer, we perform a control experiment on an inverted device, where the active layer is sandwiched between two different interlayers (ITO/ZnO and MoO x /Al). Similar to that observed in the conventional structure, the V OC in the inverted structure has also decreased upon light exposure for 72 h (Fig. S1d †). In terms of the J SC and FF, their degradation behaviour is obviously affected by the device structures (and hence the interlayers) ( Fig. S1 and S2 †), meaning that the effects of the interlayers and active layer cannot be easily separated in this case. 23 The change in V OC , on the other hand, is the parameter that contains unique information about the light-induced degradation of the active layer, which is the focus of this work.
In order to understand the mechanisms of light-induced V OC degradation, we quantify the V OC . In the bulk-heterojunction (BHJ) structured OSCs, the V OC can be described using the following equations: 24 and DV non-rad where E CT is the energy of charge-transfer (CT) states, k is Boltzmann's constant, T is the absolute temperature, l is the reorganization energy related to the CT absorption process, q is the elementary charge, c is the speed of light in vacuum, h is Planck's constant and f is a prefactor. As shown in eqn (1), the V OC consists of three components, including the energy of CT states, radiative recombination energy loss and non-radiative recombination energy loss. The rst two parts can be determined by tting the EQE PV or EL spectrum in the CT state regime and the last part correlates to the external efficiency of the EL (EQE EL ) of the device. 24,25 We investigate all three items in eqn (1) to determine the origin of the V OC degradation under illumination. The absorption of the CT states is very weak because of the weak overlap between the highest occupied molecular orbital (HOMO) of the donor and the lowest unoccupied molecular orbital (LUMO) of the acceptor. Herein we employ a sensitive photocurrent spectroscopy method FTPS to measure the EQE PV in the CT state region. As shown in Fig. 2a, there is a region with a swollen FTPS signal at low energy, corresponding to the photocurrent due to CT state absorption. [24][25][26] The shape of the EQE PV curve in the CT state region shows no change aer 72 h illumination, indicating that there is no change in either the E CT or DV rad OC . In order to further conrm the information we obtained from the FTPS measurements, we measure the EL spectrum of the devices as well. As shown in Fig. 2b, the EL spectra of CT states (Fig. S3 †) show no changes aer 72 h illumination, consistent with the FTPS results.
The EQE EL results are depicted in Fig. 2c. Compared with the device stored in the dark, the EQE EL of the device aged under light decreases by one order of magnitude, corresponding to an increase of a non-radiative V OC loss by $60 meV according to eqn (3). This result matches well with the overall V OC decrease aer 72 h illumination (Table S2 †). We also perform EQE EL measurements on the inverted devices and obtain similar results ( Fig. S4 and Table S4 †), further conrming that the degradation originates from the active layer, rather than the interlayers. From this quantication of the light-induced V OC losses, we can conclude that the light-induced V OC loss stems from the increased non-radiative recombination losses in the active layer.
A possible reason for the increased non-radiative V OC losses is the formation of additional trap states induced by illumination, leading to trap-assisted non-radiative recombination. A direct approach to investigate the traps is to perform temperature-dependent mobility measurements. 27-29 Therefore, we measure the J-V curves of the hole-only (HO) and electrononly (EO) devices based on blends at different temperatures. As shown in Fig. 3a, the electron transport of the blend devices decreases aer 72 h illumination, meaning that degradation generates an increasing amount of electron traps. We can further quantify the disorder values in the devices before and aer degradation by using EGDM. The details of this method are shown in the ESI of ref. 12. We nd that the disorder value, which is denoted as s, increases from 46 meV to 50 meV upon degradation ( Fig. 3e and f). Surprisingly, the hole transport shows an enhancement aer 72 h of illumination (Fig. 3d), the reason of which will be discussed later on.
In order to further conrm the light-induced electron traps and understand the unusual enhancement of the hole transport, we measure the single-carrier devices of the pristine materials ( Fig. 3e and f). Indeed, consistent with the EO device of the blend, the EO device of PC 71 BM shows decreased transport, conrming light-induced electron traps. In contrast to the enhancement of the hole transport in the blend, the hole transport in the pristine TQ1 device shows no difference aer 72 h illumination.
We further investigate the degradation processes at the molecular level with polarization-modulation infrared reection-absorption spectroscopy (PM-IRRAS). As shown in Fig. 4a, a new band appears at $1776 cm À1 , near the PC 71 BM methyl ester carbonyl peak at 1732 cm À1 , during the illumination in the TQ1:PC 71 BM blend lm. A similar band at $1776 cm À1 also appears in the pure PC 71 BM lm (Fig. 4b), but not in the pure TQ1 lm (Fig. 4c), indicating that the new vibrational peak is due to the oxidation of PC 71 BM. The fact that pure TQ1 shows no light-induced vibrational band is consistent with the charge transport result in Fig. 3f, where no change of the hole transport is observed in the HO device of TQ1. We propose that the formation of the band at $1776 cm À1 is due to the oxidation of carbon elements resulting in the formation of carbonyl groups (C]O) in PC 71 BM. The origin of oxygen could be from the oxygen in the glovebox (5-10 ppm oxygen level) and/ or small amounts of oxygen remaining in the materials before they are transferred into the glovebox (the solid materials are stored outside in a fridge). It is now becoming clear why we observe abnormal enhancement of the hole transport in the blend lm, as shown in Fig. 3d. We suppose that in the blend lm, the oxygen in TQ1 tends to react with the PC 71 BM due to the scavenger properties of fullerene, 30,31 helping to de-dope the polymer and thus improving the hole transport. 32,33 This process was previously demonstrated to be facilitated by illumination. 20 In order to further demonstrate the oxidization of the materials, we intentionally expose the lms in air under light  exposure, and then perform the FTIR measurements. As shown in Fig. S5a-c, † we observe that all the lms exposed to air show a new vibration signal, corresponding to the carbonyl group. This vibration signal now emerges much more quickly, compared with the experiments performed in the glovebox, as a result of the large amounts of oxygen available in air. We notice that TQ1 in this case also shows an oxidation signal. The lack of the oxidation signal of the TQ1 lm in the glove-box might be due to the fact that compared with PC 71 BM, the TQ1 is less reactive with oxygen. 34 Based on these results, we conclude that the newly appearing carbonyl vibrational band in Fig. 4a comes from the oxidation of PC 71 BM during the illumination.

Conclusion
We investigate the light-induced degradation of the devices based on TQ1:PC 71 BM and nd that the oxidation of PC 71 BM (as revealed by FTIR results) plays a key role in the degradation of the device parameters, especially the V OC . The oxidation of PC 71 BM results in electron traps, leading to enhanced nonradiative recombination loss of the V OC , which we have quan-tied by measuring the electroluminescence quantum efficiency of the devices. We also quantied the energetic disorder induced by the electron trap states due to the oxidation of PC 71 BM. Our work indicates that fullerene acceptors are detrimental to the light stability of devices, and alternative acceptor materials are required to develop stable OSCs. Our conclusion is in line with the recent trend towards the development of OSCs based on non-fullerene acceptors. [35][36][37][38] Experimental section

Device preparation
All the solar cells are fabricated on an indium tin oxide (ITO) coated glass substrate and a 90 nm aluminium layer is used as the top contact. The ITO glass is cleaned in the TL1 process (boiled for 20 min in the mixed solution of hydrogen peroxide (H 2 O 2 ) and ammonia (NH 3 )). A structure of ITO/PEDOT:PSS/ active layer/LiF/Al is employed for the conventional structure device and a structure of ITO/ZnO/active layer/MoO x /Al is employed for the inverted structure device. A structure of ITO/ PEDOT:PSS/active layer/MoO x /Al is employed for the HO devices and a structure of ITO/ZnO/active layer/LiF/Al is employed for the EO devices. The thickness of the TQ1:PC 71 BM blend active layer is $90 nm. The thickness of the pure TQ1 and PC 71 BM active layers is $80 nm and $110 nm, respectively.

Stability measurements
All the devices studied in our experiments are encapsulated with a coverslip which is UV-glued on the surface of the device. The devices in the glovebox (lled with nitrogen, #1% ppm moisture and 5-10% ppm oxygen) are divided into two groups, with one group stored in the dark and the other group stored under the illumination of a white LED. The LED spectrum is shown in Fig. S6 † and the intensity is calibrated to be $50 mW cm À2 with a standard Si solar cell. J-V measurements of the devices are performed outside the glovebox. An AM1.5G sun simulator purchased from Oriel is used for the J-V measurements.

FTPS measurements
FTPS measurements are performed on a Vertex 70 from Bruker optics, equipped with a QTH lamp, quartz beam-splitter and external detector option. A low noise current amplier (SR570) is used to amplify the photocurrent produced by the illumination of FTIR on the devices. The output voltage of the amplier is fed back to the external port of the FTIR, in order to be able to employ the FTIR soware to collect the photocurrent spectrum.

EL measurements
A Keithley 2400 is used for supplying voltage/current to the devices. An Andor spectrometer (Shamrock sr-303i-B, coupled to a Newton EMCCD detector) is used to record the EL spectra.

EQE EL measurements
The EQE EL measurements are performed on a home built setup comprising a Hamamatsu silicon photodiode 1010B, a Keithley 2400 for supplying voltages and recording the injected current, and a Keithley 485 for measuring the emitted light intensity.

FTIR measurements
Polarization-modulation infrared reection-absorption spectroscopy (PM-IRRAS) measurements are carried out by means of a Bruker PMA 50 with a LN2-cooled mercury cadmium telluride (MCT) detector at 86 grazing angle, using a Bruker VERTEX70 as a light source. A photo-elastic modulator (Hinds: ZnSe 50 kHz) is used to modulate the polarization of light. Spectra containing PC 71 BM are normalized to the fullerene peak near 1430 cm À1 , and the TQ1 spectra normalized to the ring vibration at 1465 cm À1 .

Conflicts of interest
There are no conicts to declare.