A fullerene alloy based photovoltaic blend with a glass transition temperature above 200 (cid:1) C †

Organic solar cells with a high degree of thermal stability require bulk-heterojunction blends that feature a high glass transition, which must occur considerably above the temperatures encountered during device fabrication and operation. Here, we demonstrate for the ﬁ rst time a polymer : fullerene blend with a glass transition temperature above 200 (cid:1) C, which we determine by plasmonic nanospectroscopy. We achieve this strong tendency for glass formation through the use of an alloy of neat, unsubstituted C 60 and C 70 , which we combine with the ﬂ uorothieno-benzodithiophene copolymer PTB7. A stable photovoltaic performance of PTB7 : C 60 : C 70 ternary blends is preserved despite annealing the active layer at up to 180 (cid:1) C, which coincides with the onset of the glass transition. Rapid deterioration of the power conversion e ﬃ ciency from initially above 5% only occurs upon exceeding the glass transition temperature of 224 (cid:1) C of the ternary blend.


Introduction
Organic photovoltaics receives tremendous interest as an alternative solar cell technology because of its compatibility with low-cost manufacturing through roll-to-roll coating and printing techniques.The active layer material is typically designed according to the bulk-heterojunction concept, i.e. an intimate blend of an electron donor and an acceptor.The most widely explored type of acceptor is fullerenes, which together with polymeric donors can give rise to power conversion efficiencies above 11% in the case of lab-scale devices 1 and 7.5% in the case of small modules. 2 One important requirement for both high-throughput manufacturing and long-term use of polymer : fullerene bulkheterojunctions is excellent thermal stability.The material must be able to withstand elevated processing temperatures because the coating speed is limited by the rate of solvent removal, which can be accelerated by heating.The choice of substrate determines the highest possible processing temperature, e.g.140 C in the case of poly(ethylene terephthalate) (PET) foil. 3,4Moreover, during operation the solar cell must be able to handle temperatures up to 85 C, as required by industry standards. 5r the majority of donor : acceptor blends, the optimal nanostructure, which leads to the highest photovoltaic performance, is located far away from thermal equilibrium.To prevent reorganisation of the nanostructure upon heating, it is necessary to select blends that are characterised by a high glass transition temperature T g . 68][9] A nely mixed blend typically displays a single T g and its nanostructure remains frozen in, as long as the blend remains far below this critical temperature.As such the glass transition should be considered as a kinetic phenomenon that represents a nominal temperature below which relaxation of the donor polymer and diffusion of the fullerene acceptor are strongly slowed down but not prevented. 6apid solidication from solution can lead to nanostructures where a relatively large amount of free volume is trapped below T g (cf.ref. 10: the thickness of a spin-coated uorenebenzothiadiazole copolymer thin lm decreases by more than one percent when heated above T g ).Heating below but sufficiently close to T g , which can occur during the manufacture and operation of a solar cell, is likely to result in gradual rearrangement of donor and acceptor molecules.For instance, we have shown that such sub-T g annealing can occur in blends based on the thiophene-quinoxaline copolymer TQ1, with local structural changes (inferred from photoluminescence and UV-vis spectroscopy) taking place as much as 70 C below the nominal T g $ 110 C. 11 Thus, it would be desirable to identify donor : acceptor blends that feature a T g , which lies signicantly above the maximum processing temperature, e.g.140 C for PET foil.
Substituted fullerenes such as PC 61 BM and PC 71 BM, which are most commonly used, feature a T g of about 110-130 C (ref. 10, 12 and 13) and 160 C, 14,15 respectively.When mixed with common donor polymers, blends with a glass transition of typically around 100 C are obtained, 6 which may not be sufficiently high to ensure complete thermal stability during processing.9][20][21][22] Other advantages of fullerene alloys are cost-reduction and benecial mechanical properties. 23Angmo et al. have shown that such alloys are fully compatible with slot-die coating and may enhance the solar cell efficiency. 24Recently, solar cells based on a PC 61 BM : PC 71 BM alloy with an efficiency above 10% and good thermal stability at 130 C have been demonstrated. 20e have recently explored the use of neat fullerene mixtures comprising C 60 and C 70 in ternary blends with either TQ1 or the uorothieno-benzodithiophene copolymer PTB7 (Fig. 1), 25,26 which promises a considerable reduction in the energy footprint of the acceptor material. 27In particular, we found that the C 60 : C 70 mixture displays a signicantly enhanced solubility in a wide range of organic solvents, which enabled the preparation of highly reproducible eld-effect transistors with mobilities of 1 cm 2 V À1 s À1 and solar cells with a power conversion efficiency of 6%. 26 Moreover, the photovoltaic performance of the PTB7 : C 60 : C 70 ternary blend was unaffected by annealing at 100 C, which suggests that the use of C 60 : C 70 alloys gives rise to a high glass transition temperature.We rationalized the enhanced solubility of C 60 : C 70 mixtures and their tendency for glass formation by the increase in congurational entropy upon mixing. 26ere, we study the photovoltaic performance of ternary PTB7 : C 60 : C 70 blends exposed to extreme annealing temperatures up to 300 C. To detect the glass transition temperature we employ plasmonic nanospectroscopy, 28,29 a technique based on localised surface plasmon resonance (LSPR), which is particularly sensitive in detecting phase transitions in organic thin lms.We have recently shown that this method is valid for organic photovoltaic blends. 30We nd that photovoltaic devices based on PTB7 : C 60 : C 70 display excellent thermal stability with a glass transition temperature at T g $ 224 C and a thermally stable photovoltaic performance up to $180 C.

Results and discussion
In a rst set of experiments we employed plasmonic nanospectroscopy to determine the glass transition temperature of the here investigated PTB7 : C 60 : C 70 ternary blend with a 2 : 1 : 1 stoichiometry.This technique exploits LSPR, i.e. photon-driven electron oscillation in metallic nanoparticles positioned on a planar surface, which gives rise to locally enhanced electric elds that are highly sensitive to changes in the surrounding medium. 31Embedding the nanoparticles in an organic thin lm permits us to monitor minute changes in refractive index upon heating due to expansion, which results in a shi in the peak wavelength l peak of the plasmonic resonance by Dl peak ¼ l peak (T) À l peak (RT).Any phase transition is accompanied by an abrupt change in the linear expansion coefficient, and thus manifests itself as a change in the rate by which Dl peak varies with temperature, i.e. d(Dl peak )/dT (cf.ref. 30

for details).
We deposited PTB7 : C 60 : C 70 thin lms with a thickness of about 100 nm on a suitable plasmonic sensor chip (see Experimental).As active nanoantennas in our sensor, we chose Au nanodisks with a diameter of 170 nm and height of 20 nm, which give rise to a plasmonic resonance at l peak $ 920 nm.As a result, we were able to avoid overlap with the absorption band of the ternary blend (Fig. 2a).During the rst heating scan from 60 to 250 C we observe a continuous blue shi of Dl peak with a clear change in slope d(Dl peak )/dT around 224 AE 1 C, which we identify as the glass transition temperature of the ternary blend (Fig. 2b).We note that d(Dl peak )/dT starts to change before this nominal T g $ 224 C is reached (with an onset around 180 C, see ESI, Fig. S1 †), which indicates that structural changes can occur at considerably lower temperatures.We also carried out plasmonic nanospectroscopy on neat PTB7 but did not observe any change in d(Dl peak )/dT (ESI, Fig. S2 †).Differential scanning calorimetry (DSC) of neat PTB7 reveals a shallow exotherm between 140 and 210 C in second heating thermograms, which persists in the ternary blend (ESI, Fig. S3 †).We conclude that the polymer is able to undergo structural reorganisation in this temperature range, which suggests that its glass transition temperature may be found below 140 C.
The plasmonic signal recorded during subsequent heating scans differs from the rst heating scan.We observe a smaller change in Dl peak and a linear region that persists up to above 200 C (ESI, Fig. S4 †), which suggests a higher onset of the glass transition temperature as compared to the rst heating scan.We attribute this behaviour to the irreversible crystallisation that occurs when the ternary blend is heated above its T g , as investigated in detail below.Moreover, thermogravimetric analysis (TGA) indicates a weight loss of about 2% for PTB7 : C 60 : C 70 between 80 and 160 C (ESI, Fig. S3 †), which we associate with trapped solvent (note that we observe a similar weight loss for C 60 : C 70 but none for PTB7).The tendency of fullerenes to trap chlorinated solvents is well documented and This journal is © The Royal Society of Chemistry 2017 may contribute to the observed lower T g that we observe for the rst heating scan.
8][9]32 We used transmission electron microscopy (TEM) and selected area electron diffraction (SAED) to examine the nanostructure of spincoated 2 : 1 : 1 PTB7 : C 60 : C 70 thin lms aer annealing up to 240 C. In TEM bright eld images no distinct, phase-separated domains can be resolved up to an annealing temperature T anneal $ 180 C. The corresponding SAED patterns only reveal an amorphous halo, which indicates that the initially ne nanostructure obtained through spin-coating is preserved (Fig. 3).Instead, annealing at 190 C for 10 min resulted in the appearance of micrometre-sized single-crystal-like entities in TEM images, as well as sharp diffraction spots in the corresponding SAED patterns.These crystallites were surrounded by a featureless matrix (see ESI, Fig. S5 † for atomic force microscopy).Annealing at higher temperatures resulted in an increase in both the density and size of crystals.For instance, T anneal $ 240 C gave rise to a dense coverage of 2 to 3 mm large crystals that were surrounded  by a bright halo (Fig. 3).This depletion region is commonly observed around fullerene crystals that have grown in polymer-: fullerene bulk-heterojunction blends, and arises because crystal growth has consumed the fullerene acceptor from the surrounding blend. 7,8,32,33The halo appears brighter than the surrounding lm because of the higher electron density of fullerenes as compared to the remaining PTB7-rich material.We therefore conclude that the observed crystallites are composed of the fullerene material.SAED revealed distinct diffraction patterns, which conrm that the observed entities are single crystals.
To probe the impact of annealing on the local makeup of the ternary blend in more detail, we carried out photoluminescence (PL) spectroscopy.The PL emission of neat PTB7 is strongly quenched by a factor of 130 upon addition of the fullerene alloy (comparison of neat PTB7 and the as-cast ternary blend at room temperature).We observe no clear trend in PL quenching efficiency upon annealing the blend up to T anneal $ 300 C (ESI, Fig. S6 †), which indicates that the conditions for exciton PL quenching and exciton dissociation are not signicantly altered.We conclude that despite the removal of the fullerene material from part of the lm (cf.depletion regions in TEM images) a sufficiently large fraction of the acceptor remains present to effectively quench PL emission from neat PTB7.
In a further set of experiments we compared the thermal behaviour of the PTB7 : C 60 : C 70 ternary blend with its photovoltaic performance.To this end we prepared solar cells with a ternary blend active layer that we annealed during device fabrication (see the inset in Fig. 4b for device architecture).Active layers with a thickness of 100-110 nm were spin-coated onto the ITO/PEDOT : PSS anode, followed by annealing in a dark, nitrogen lled glovebox for 10 min at T anneal ranging from $RT to 300 C. The top LiF/Al electrode was deposited aer annealing.Devices were not encapsulated and measured in an ambient environment shortly aer fabrication.
We nd that for annealing at T anneal # 180 C the photovoltaic performance is unaffected (Fig. 4).Averaging over all solar cells annealed at up to 180 C yields a short-circuit current density J sc $ 12.8 AE 0.6 mA cm À2 , an open-circuit voltage V oc $ 0.67 AE 0.01 V, a ll factor FF $ 0.62 AE 0.02 and a power conversion efficiency of PCE $ 5.3 AE 0.3%, with champion devices reaching 6%.We note that the here observed thermal stability is in excellent agreement with the onset of the glass transition that we observed by plasmonic nanospectroscopy (cf.Fig. 2b).At higher annealing temperatures both the J sc and FF start to decrease, rst only slightly for 180 C < T anneal # 220 C, and then rapidly for T anneal > 220 C. Evidently, the T g $ 224 C of the ternary blend coincides with the annealing temperature at which a rapid loss in photovoltaic performance is observed.
Annealing at 260 C results in a near complete loss in J sc from initially 12.8 to only 2 mA cm À2 .In contrast, no signicant change in V oc has occurred.PL and electroluminescence (EL) spectra of thin lms annealed up to T anneal $ 260 C appear unchanged in shape and energetic position (ESI, Fig. S6 †), which indicates that the energy of the charge-transfer state is unaffected by aggressive thermal treatment.The energy of the charge-transfer state is directly related to the generated photovoltage, [34][35][36] which is in agreement with the observed invariance in V oc (Fig. 4c).We conclude that the properties of the donor/ acceptor interface, which dene the energetic position and the width of the EL emission, are unaffected upon annealing up to 260 C. Since both EL and PL measurements indicate that the nanoscale phase separation is not signicantly altered upon annealing, it is reasonable to assume that charge carrier transport through the bulk of the lm is not affected.External quantum efficiency (EQE) spectra corroborate this picture.The shape of EQE spectra is comparable up to T anneal $ 260 C (Fig. 5), which suggests that the conditions for charge generation, related to the nanoscale phase separation of the photovoltaic blend, are not signicantly affected upon thermal treatment.Instead, the decrease in device performance upon annealing at T anneal > 180 C likely arises due to a decrease in collection efficiency of photogenerated charges, which we tentatively assign to degradation of the interface of the active layer with the anode and/or cathode.
We carried out a nal experiment to explore whether the here investigated use of neat fullerene mixtures is a general approach towards bulk heterojunction blends with a high glass transition temperature.We chose to study a ternary blend of C 60 : C 70 and the thiophene-quinoxaline copolymer TQ1 (ESI, Fig. S7 †).Plasmonic nanospectroscopy of a spin-coated 2 : 1 : 1 TQ1 : C 60 : C 70 thin lm reveals a T g $ 141 C, which is considerably higher than the glass transition temperatures of neat TQ1 (ref.).Similar to the ternary blend based on PTB7, we also nd for TQ1 : C 60 : C 70 that the T g shis to an even higher temperature for the second heating scan.Evidently, glassy bulk-heterojunctions with a high degree of thermal stability can also be achieved with other ternary blends of a donor plus a neat fullerene alloy.

Conclusions
We have establishedusing plasmonic nanospectroscopythat a ternary blend of PTB7 and a C 60 : C 70 fullerene alloy with a 2 : 1 : 1 stoichiometry displays an exceptionally high glass transition temperature of T g $ 224 C, which is considerably higher than any other value reported to date for a donor : acceptor bulk-heterojunction.Electron microscopy conrmed the tendency of the ternary blend to form glassy, amorphous thin lms with a homogeneous nanostructure.The photovoltaic performance of the corresponding solar cell active layers remained unaltered upon annealing at temperatures up to T anneal $ 180 C, which was found to be in excellent agreement with the onset of the blend T g .Signicant loss in photocurrent, but not photovoltage, coincided with the glass transition, giving rise to a continuous drop in power conversion efficiency for T anneal > 180 C from initially more than 5%.We argue that true thermal stability of a photovoltaic blend can only be achieved if systems with a T g signicantly above the solar cell processing and operating temperatures are selected.

Plasmonic nanospectroscopy
T g measurements were carried out by plasmonic nanospectroscopy for which nanoplasmonic chips consisting of arrays of Au nanodisks of 170 nm and 20 nm of diameter and height, respectively, were employed.Details of the fabrication have been described elsewhere. 29,30The chips were mounted in an insulated quartz tube gas ow reactor system with optical access (Insplorion X1, Insplorion AB, Göteborg, Sweden) connected to mass ow controllers (Bronkhorst) regulating the composition and total ow rate with a constant pressure of 1 atm.The chips were illuminated using a bre-coupled halogen lamp (AvaLight-Hal-S, Avantes) while the wavelength-resolved extinction spectra (400-1100 nm) were continuously recorded by using a bre-coupled xed grating spectrometer (AvaSpec-1024, Avantes).The working temperature (including the ramping rate) was set by the heating coil that is connected to the Eurotherm controller.The chip temperature was monitored via a thermocouple in direct contact with the surface of the chip.The sequence of T g measurements in this work is as follows: (1) the chip was heated to 80 C and dwelled for 1.5 h to remove any remaining solvents, (2) the temperature was reduced to 60 C, and (3) heating from 60 to 250 C at a rate of 5 C min À1 .All of the processes described were performed under 50 mL min À1 constant ow of Ar.The shi of the LSPR peak Dl peak was obtained by tting the measured extinction spectra with a Lorentzian in the AE100 nm wavelength range around the LSPR peak maximum.

Thermal analysis
Dynamic Scanning Calorimetry (DSC) was performed under nitrogen with a Mettler Toledo DSC 2 equipped with a Huber TC 125-MT intracooler at a scan rate of 20 C min À1 .Thermal Gravimetric Analysis (TGA) was performed under nitrogen at a scan rate of 10 C min À1 with a Mettler Toledo TGA/DSC 3+.Atomic force microscopy (AFM) AFM was performed in intermittent contact mode using an Agilent 5500 system in air.

Transmission electron microscopy (TEM)
Samples were prepared by spin-coating thin lms on poly(3,4ethylene dioxythiophene) : poly(styrene sulfonate) PEDOT : PSS, followed by oating off lms in water and nally collection with TEM copper mesh grids.TEM images were recorded with a G 2 T20 Tecnai instrument operated at an acceleration voltage of 200 kV.

Photoluminescence (PL) spectroscopy
PL spectra of thin lms were recorded using an Oriel liquid light guide and a SR 303i spectrograph coupled to a Newton EMCCD silicon detector.The lms were excited using a blue PMM-208G-VT laser pump (4 mW cm À2 ) with a wavelength of 532 nm.PL data were scaled by sample absorption, and measured with a Perkin-Elmer Lambda 900 spectrophotometer equipped with an integrating sphere.

Photovoltaic devices
Photovoltaic devices were fabricated in standard device geometry, i.e. glass/ITO/PEDOT : PSS/active layer/LiF/Al (cf.inset Fig. 4).ITO-patterned glass substrates were oxygen-plasma treated for 1 min prior to the deposition of the PEDOT : PSS electrode (Heraeus, Clevios P VP Al 4083, annealed at 120 C for 15 min aer spin-coating, thickness $ 40 nm).PTB7 : fullerene active layers were prepared by (1) heating of fullerene solutions for two days at 27 C, (2) followed by addition of PTB7, (3) heating for about 2 h at 80 C, and then (4) spin-coating of the obtained o-DCB solutions, (5) followed by annealing of the active layer in a nitrogen-lled glovebox (before deposition of the top electrode) for 10 minutes at different temperatures (RT to 300 C).The thickness of the active layers was around 100-110 nm as measured with a Dektak 150 surface proler (estimated error AE7 nm).A LiF layer (thickness $ 6 Å) and aluminium top electrodes (thickness $ 90 nm) were deposited via thermal evaporation under vacuum (below 4 Â 10 À6 mbar).J-V curves were recorded with a Keithley 2400 Source Meter under AM 1.5G illumination with an intensity of 100 mW cm À2 from a solar simulator (Model SS50A, Photo Emission Tech., Inc.).The light source used was a 180 watt xenon arc lamp solar simulator (Photo Emission Tech.).The intensity was calibrated using a standard silicon photodiode calibrated at the Energy Research Centre of the Netherlands (ECN).The active area of the solar cells was determined with an optical microscope.

Electroluminescence (EL) spectroscopy
EL spectra of solar cells were recorded using an Oriel liquid light guide and a Shamrock SR 303i spectrograph coupled to a Newton EMCCD silicon detector.Samples annealed at higher temperatures required a considerably larger positive bias for a comparable current density to be achieved: 0.84 V and 3 V for RT and 240-260 C devices, respectively.EL spectra were collected at a current density of 21.7 mA cm À2 , 15.2 mA cm À2 and 10.9 mA cm À2 for RT, 240 C and 260 C devices, respectively.

External quantum efficiency (EQE)
EQE spectra of photovoltaic devices were recorded with a homebuilt setup using a Newport Merlin lock-in amplier.Devices were illuminated with chopped monochromatic light through the transparent ITO electrode.Measured EQE spectra were scaled so that the estimated short-circuit current density from the EQE measurement matched the short-circuit current density of the corresponding J-V curve.

Fig. 2
Fig. 2 (a) Optical extinction of a 2 : 1 : 1 PTB7 : C 60 : C 70 thin film (blue) and the localised surface plasmon resonance (LSPR) of the embedded Au nanodisks (red).(b) Plasmonic nanospectroscopy: first heating scan to monitor the shift in the LSPR peak Dl peak during heating from 60 to 250 C (red); the intersection of the straight line fits (dashed) indicates a thermal transition, which we interpret as the glass transition temperature T g $ 224 C.

Fig. 4
Fig. 4 (a) Representative current-voltage J-V characteristics of 2 : 1 : 1 PTB7 : C 60 : C 70 devices comprising active layers that were thermally treated at T anneal , and dark current of an as-cast device (dashed).(b) J sc and FF, (c) MPP and V oc of 2 : 1 : 1 PTB7 : C 60 : C 70 devices as a function of T anneal ; each data point corresponds to a measured device, error bars indicate the standard deviation of 3-6 devices on the same substrate (solid lines are a guide to the eye); inset: solar cell device architecture.RT ¼ room temperature.