Abnormal strong burn-in degradation of highly efficient polymer solar cells caused by spinodal donor-acceptor demixing

The performance of organic solar cells is determined by the delicate, meticulously optimized bulk-heterojunction microstructure, which consists of finely mixed and relatively separated donor/acceptor regions. Here we demonstrate an abnormal strong burn-in degradation in highly efficient polymer solar cells caused by spinodal demixing of the donor and acceptor phases, which dramatically reduces charge generation and can be attributed to the inherently low miscibility of both materials. Even though the microstructure can be kinetically tuned for achieving high-performance, the inherently low miscibility of donor and acceptor leads to spontaneous phase separation in the solid state, even at room temperature and in the dark. A theoretical calculation of the molecular parameters and construction of the spinodal phase diagrams highlight molecular incompatibilities between the donor and acceptor as a dominant mechanism for burn-in degradation, which is to date the major short-time loss reducing the performance and stability of organic solar cells.

The authors report an apparently abnormal burn-in phase in a novel bulk heterojunction organic photovoltaic (OPV). They propose that this process is caused by a thermodynamically unstable active layer nanostructure. Specifically, under ambient conditions the conjugated copolymer and fullerene undergo excessive demixing over several days via spinodal decomposition. As a consequence, the interfacial volume between both materials is reduced and the efficiency of free charge generation suffers. This lowers the achievable photocurrent density of the solar cell by ~ 30 -40% from its initial value. The authors claim that such a process, which has not been reported to date, must be taken into account when designing organic photovoltaic cells that offer high power conversion efficiency (PCE) and long-term stability.
The research is certainly topical and has the potential to generate strong impact in the field. Given the complex nature of failure modes in OPVs I am willing to accept that unstable active layer nanostructures will eventually contribute to a reduced PCE over time. Unfortunately there is insufficient evidence in the manuscript to support such a specific claim over the timescales investigated. I am also surprised by the decision to fabricate the OPVs in ambient conditions prior to ageing as this will inevitably introduce additional factors into the stability investigation. Given the well known sensitivity of organic semiconductors to water and oxygen (e.g. doping, accelerated trap formation and photobleaching), fabricating OPVs in air is likely to cause unnecessary damage to the semiconductors. How do the authors know the results obtained aren't simply due to air exposure during device fabrication?
To simplify the experiment the effects of an unstable active layer nanostructure (if genuine) must be isolated from environmental factors. Control solar cells should be prepared under standard glovebox conditions, encapsulated, measured, and aged under dry N2/dark conditions prior to the second measurement. If the PCE changes by a significant amount this can be studied using standard techniques. In its present format, the study is simply unconvincing and contains too much uncertainty to confidently ascribe the changes in OPV PCE to a single degradation mechanism.
Other aspects of the study that should be revised include: 1 -Confirming a modified active layer post-ageing. The authors should provide direct structural evidence of this. If the relative volume fractions and electron densities of amorphous and mixedamorphous domains in the sample changes this should be resolvable in a small angle X-ray scattering experiment (see DOI: 10.1021/ma2007706 and DOI: 10.1002/aenm.201301377 for example data) or similar.
2 -Semiconductor properties related to a modified active layer nanostructure. The uv-vis data in the supporting information (S.I.) should be presented on an absolute rather than a normalised scale so that the reader can see if any loss in absorption has taken place. The parameters used to model the CT region of the EQE spectrum should be stated and discussed. It would be interesting to see how much of a change ageing induces in the CT absorbance band width as this is typically seen as a measure of disorder at the donor:acceptor interface.
3 -Ageing under continuous illumination. The spectrum of the LED used in these experiments should be provided in the S.I. The glass transition temperature (Tg) of the PCE11:PCBM thin-film should be measured and discussed in the context of the 350K ageing data (is the rapid drop in PCE a result of being at Tg or above it?). The moisture content of the N2 atmosphere used during certain ageing experiments should be stated. 4 -In the methods section, the following details are missing: the relative humidity during OPV active layer deposition and the amount of time spent by the substrate at 100C (i.e. during spin coating). 5 -In the introduction, there are limited references on the stability of state-of-the-art OPVs, let alone on older systems (e.g. P3HT:PCBM, PCDTBT:PC70BM). These highly relevant studies should be referenced to help put the 'abnormal strong burn-in' into context. On line 132, the following papers should be cited to help the reader appreciate that diidooctane (DIO) may negatively affect the performance/stability figure of merit for a wide range of OPVs, not just those prepared under ambient conditions: DOI: 10.1021/jp510996w, doi:10.1016/j.orgel.2015.12.024.
Although I do not recommend publication at this time, the research should be developed as short term stability in OPVs is a pressing challenge and strategies to solve this are sorely lacking. Thoughtful and careful redesign of the core experiments should go some way to help address this.
Reviewer #2 (Remarks to the Author): The manuscript 'Abnormal strong burn-in degradation of highly efficient polymer solar cells caused by spinodal donor-acceptor demixing' provides striking evidence of the fast degradation the highefficiency OPV blends based on PCE11 as donor material and fullerenes such as PCBM-C60 undergo. The manuscript clearly demonstrates that one should not only seek to obtain high efficiencies and highlights unambiguously what issues in materials-and device design need to be overcome if OPVs become a commercially viable technolgy. Already for this message, delivered in such a convincing way, I recommend the manuscript for publication in Nature Communications. A broad audience has to receive this information.
The strength of the message of the present manuscripts stems from the very conclusive device data, supported by EL (PL) data of blends and neat (not pristine!) materials. The emission data clearly indicates phase separation, as suggested by the authors. Some other parts are, however, based on data that are difficult to interpret (or should maybe be interpreted more carefully). This would make the manuscript more accessible for a broad audience. I have no doubts that the authors can do so and I suggest to give them this chance; I would like to see this manuscript be published in Nat. Comms. It deserves it! Points to address: 1) UV-vis data: the authors indirectly use the Spano model and attempt to deduce information on the structural order from their UV-vis data. A) References to work by Spano and co-workers should be provided (more on reference below). B) The authors may read some of the recent work by Spano. The spectra their blends and neat PCE11 display are more typical for J-like aggregates. Hence, the fact that the spectra do no change, only implies that the electronic coupling between chains stays weak and there is a strong coupling along the polymer backbones. This may be due to the more rigid backbone of PCE 11 compared to P3HT; although it cannot drastically more rigid as the melting temperatures are comparable. In P3HT the 0-0 transition often changes as it is usually more H-like with strong electronic coupling perpendicular to the backbone. Hence, generally, a more H-like behaviour infers better electronic coupling along the pi-stack and, as a consequence, can be used as an INDICATION of improved order at least on short length scales. It not necessarily affects the microscopic range, as suggested on page 9. I suggest the authors have a close look at this section.
2) X-ray data. PCE11 features rather strong reflections; hence, changes on smaller length scales may not be noted even in a log intensity plot. Indeed, I am not sure whether the fact that they see no changes really implies no change in molecular order or whether these changes are hidden in the 'amorphous' background. Because of the reasoning on the interpretation of the presented UVvis data, there still could be changes on short ranges. I would combine here some of the DSC data the authors must have already, I am sure. Some details need to be given,: A) How were the DSCs measured on films prepared from solution or powders that simply were mixed? I may have missed the details; I have doubled checked the SI for it. In case the DSCs were made from cast films, it is important to use the first heating scan as compatibilisation effects through the solvent can lead to another phase behaviour (the 2nd heating provides information for melt processed material, hence, is not that relevant; it can provide information on degradation, solvent compatibilisation effects when compared to the 1st scan, etc.). For obtaining information on the molecular order, the authors can compare the enthalpy of fusion for the various blends (1st heating) normalised to the polymer fraction. I certainly would also show the data for compositions in the 1:1 range (see also below). In addition, with the neat polymer featuring such a small supercooling and well defined crystallisation peak in the 1st cooling in the neat polymer, I would suggest to analyse what occurs with this feature upon blending and casting for all the blends. Does the supercooling increase (which means ordering starts to be hindered). Is its enthalpy getting reduced, etc.?
3) The above is important as I think the authors refer to spinodal donor-acceptor demixing as they like to suggest that demixing of the two materials occurs only in the amorphous phase. This is not that clearly spelled out in the manuscript -if the authors want to stick to this hypothesis. In my opinion, it may be, but it may not; based on the data provided I am not convinced. It still could be order-induced phase separation -just molecular ordering on smaller length scales than e.g. accessible with X-ray diffraction. I think it will be difficult to obtain data that solves this issue fully. I am not sure if it is needed at this stage. The burn-in is clear and I fully agree that it originates from phase separation. I personally would rephrase this part and may take out the word 'spinodal': i.e. I would suggest to simply go for 'Abormal strong burn-in degradation of highly efficient polymer solar cells caused by donor-acceptor blends.' I know some calculations are presented supporting the spinodal idea, however, I wonder how much the graph presented in Figure 4 changes when slightly different parameters are used? Some additional comments: 1) The thermal analysis data presented in Fig S7b indicates that PCE11:PCBM features a eutectic phase diagramme as the P3HT:PCBM binary does, with a eutectic point between 85 and 90 % PCE11 (so, at a much higher donor content than for P3HT:PCBM system). This phase behaviour implies that both melting points are depressed; the one of PCE11 seems more strongly depressed as the eutectic point is so much shifted towards the polymer rich side. Eutectic temperatures is around 250 C, hence the shoulder on the higher-temperature side for the 85% blend indicates that this feature is the PCBM melting -i.e. in my opinion also the PCBM melting is depressed. Clearly, having some more data for the compositions between 15% to 85 % will assist to identify correctly the eutectic composition. This will have significance whether it is or not spinodal decomposition.
2) I do not want to suggest even more experiments, but out of curiosity, have the authors performed measurements with ICBA with a higher Tg and it seems somewhat better miscibility with P3HT?
3) Finally, I think the authors should include some important references: on mixing, initial work from the Ade/McNeill groups, Treat/Chabinyc et al., Westacott/Stingelin, Russell group at UMASS, the group of Rasmus Schroeder in Germany; on spinodal decomposition of OPV blends: Steiner/Friend; phase diagrammes: Nelson/Stingelin, used also by Hadziiaonnou. UV-vis: Work by Spano et al.
In summary, I think the present work is intriguing and a broad audience will be interested in it and will learn from it. Hence, I strongly suggest publication -after issues pointed to above have been clarified or addressed.

Reviewer #3 (Remarks to the Author):
The author demonstrated strong burn-in degradation in highly-efficient state-of-the-art OSCs induced by spinodal demixing of the donor and acceptor phases, which dramatically reduces charge generation and can be attributed to the inherently low miscibility of both materials. In my opinion, the topic of "stability of OSCs" is very important but the novelty of this paper is not enough for Nature Communications. I think Scientific Reports is a good choice for this paper. Some comments are shown as below: 1. We already know the reason of this degradation (metastable state of morphology), but we don't know how to overcome it. That's the point! 2. There is only one high-efficiency polymer/fullerene pair discussed in this paper, it's not enough to make a strong conclusion based on one case. 3. After the degradation of OSCs, the VOC always higher. Please add some discussion.

Comments to the Author
The authors report an apparently abnormal burn-in phase in a novel bulk heterojunction organic photovoltaic (OPV). They propose that this process is caused by a thermodynamically unstable active layer nanostructure. Specifically, under ambient conditions the conjugated copolymer and fullerene undergo excessive demixing over several days via spinodal decomposition. As a consequence, the interfacial volume between both materials is reduced and the efficiency of free charge generation suffers. This lowers the achievable photocurrent density of the solar cell by ~ 30 -40% from its initial value. The authors claim that such a process, which has not been reported to date, must be taken into account when designing organic photovoltaic cells that offer high power conversion efficiency (PCE) and long-term stability.
The research is certainly topical and has the potential to generate strong impact in the field.
Given the complex nature of failure modes in OPVs I am willing to accept that unstable active layer nanostructures will eventually contribute to a reduced PCE over time.
Unfortunately there is insufficient evidence in the manuscript to support such a specific claim over the timescales investigated. I am also surprised by the decision to fabricate the OPVs in ambient conditions prior to ageing as this will inevitably introduce additional factors into the stability investigation. Given the well known sensitivity of organic semiconductors to water and oxygen (e.g. doping, accelerated trap formation and photobleaching), fabricating OPVs in air is likely to cause unnecessary damage to the semiconductors. How do the authors know the results obtained aren't simply due to air exposure during device fabrication?
To simplify the experiment the effects of an unstable active layer nanostructure (if genuine) must be isolated from environmental factors. Control solar cells should be prepared under standard glovebox conditions, encapsulated, measured, and aged under dry N 2 /dark conditions prior to the second measurement. If the PCE changes by a significant amount this can be studied using standard techniques. In its present format, the study is simply unconvincing and contains too much uncertainty to confidently ascribe the changes in OPV PCE to a single degradation mechanism.
-We greatly thank the reviewer for the very positive comments on our work, and fully agree with the reviewer on this statement. Actually, before starting with the air-processed PCE11:PCBM solar cells, we firstly optimized the PCE11:PCBM solar cells in a glove box without exposure to the air, and characterized their stability under various conditions.
The statistic photovoltaic parameters of the optimized PCE11:PCBM fabricated by spincoating in N 2 atmosphere are summarized in Figure R1a. As shown in the Figure R1b, the solar cells without exposure to the air showed the same strong burn-in loss under continuous one sun illumination. Moreover, similar drop in J SC was also observed for these devices stored in a glove box at room temperature in the dark. According to the experimental data we can conclude that the abnormal strong burn-in loss observed for the PCE11:PCBM system is not affected by the ambient conditions during device fabrication. Other aspects of the study that should be revised include: 1 -Confirming a resolvable in a small angle X-ray scattering experiment (see DOI: 10.1021/ma2007706 and DOI: 10.1002/aenm.201301377 for example data) or similar.
-We thank the reviewer for pointing out the pioneer works published in literature. The related references were added to the manuscript. As suggested by the reviewer, we performed the GISAXS measurements on the fresh and aged PCE11:PCBM sample to analyze the morphology change especially in the amorphous mixed region.    2 -Semiconductor properties related to a modified active layer nanostructure. The uv-vis data in the supporting information (S.I.) should be presented on an absolute rather than a normalised scale so that the reader can see if any loss in absorption has taken place. The parameters used to model the CT region of the EQE spectrum should be stated and discussed.
It would be interesting to see how much of a change ageing induces in the CT absorbance band width as this is typically seen as a measure of disorder at the donor:acceptor interface.
-The uv-vis spectra of fresh and aged PCE11:PCBM samples were taken from the entire device without top electrode, and normalized only to correct the baseline (not normalized from 0 to 1). As shown in the following figure, no degradation or loss was observed from the absorption spectra. The title of y-axis in Figure R5 was corrected to Absorbance (a.u.).  where A is the amplitude, xc is the centroid and w is the width of the fitting. The fitting parameters used to model the CT region of the PCE11:PCBM blend are summarized in Table R2. The value of x c was taken from the peak maximum of EL spectra, which is located at ~1.39 eV, as shown in Figure R6.  We fully agree with the reviewer that the change in CT absorbance band width is related to the change of disorder at the donor-acceptor interfaces. However, the aged BHJ sample is difficult to deconvolute as the FTPS characteristics of the aged PCE11:PCBM sample does not show a clear CT feature, in comparison to the fresh sample. Moreover, the peak maximum of the EL spectrum remained unchanged at ~1.39 eV for aged PCE11:PCBM, which might be due to the strongly overlapped EL emission with the neat PCE11.
Therefore, although E c for aged PCE11:PCBM was fixed to 1.39 eV, the real width of CT absorbance cannot be determined precisely. We summarize the findings that the rather small changes in the CT population are quite small compared to the dramatic increase in the pristine polymer emission. We suggest that CT spectroscopy is probably not ideal to characterize PCE11:PCBM blends as the position of the CT absorption is very close to the polymer's bandgap.  As suggested by the reviewer, temperature modulated differential scanning calorimetry    4 -In the methods section, the following details are missing: the relative humidity during OPV active layer deposition and the amount of time spent by the substrate at 100C (i.e. during spin coating).
-The solar cells studied in this work were fabricated by doctor blading under ambient conditions in a clean room. The relative humidity in our clean room was well controlled in the range of 40-45%, and the temperature was controlled to ~22°C. The substrate was kept at 100°C prior to deposition of the active layer, and was immediately removed from the hot plate when the active layer was dried. The active layer was dried at 100°C less than 5 s. The solar cells treated with hot solvent vapor annealing, as shown in Figure   XXX, were kept in a DCB atmosphere at 100°C for 20 min to intentionally create phase separation. Hence, generally, a more H-like behaviour infers better electronic coupling along the pistack and, as a consequence, can be used as an INDICATION of improved order at least on short length scales. It not necessarily affects the microscopic range, as suggested on page 9. I suggest the authors have a close look at this section.

-
-We greatly thank the reviewer for the helpful suggestion and comments. The related works published by Spano et al. were added to the manuscript as Ref 50, 51. We fully agree that the backbone of PCE11 is more rigid than that of P3HT, and the crystallinity of neat PCE11 is also higher than that of P3HT when comparing the enthalpy change of the melting peak. The characterization results shown in this work directly or indirectly imply that the drastic reduction in J SC for aged PCE11:PCBM solar cells is ascribed to the demixing of the donor and acceptor phases, even at room temperature in the dark. This is one of the core messages delivered by this contribution. The UV-vis data demonstrated in this work on the one hand revealed that polymer degradation does not occur which may also lead to the reduction in J SC ; on the other hand confirmed along with the water contact angle measurement that the macroscopic morphology of BHJ film did not change. Together with the EL spectra, the FTPS data, the GIWAXS and the supplemented GISAXS data, we can come to a very clear conclusion that the drastic reduction in J SC for aged PCE11:PCBM solar cells is indeed due to the abnormal unstable mixed donor-acceptor amorphous regions.
2) X-ray data. PCE11 features rather strong reflections; hence, changes on smaller length scales may not be noted even in a log intensity plot. Indeed, I am not sure whether the fact that they see no changes really implies no change in molecular order or whether these changes are hidden in the 'amorphous' background. Because of the reasoning on the interpretation of the presented UV-vis data, there still could be changes on short ranges. I would combine here some of the DSC data the authors must have already, I am sure.
Some details need to be given,: A) How were the DSCs measured on films prepared from solution or powders that simply were mixed? I may have missed the details; I have doubled checked the SI for it. In case the DSCs were made from cast films, it is important to use the first heating scan as compatibilisation effects through the solvent can lead to another phase behaviour (the 2nd heating provides information for melt processed material, hence, is not that relevant; it can provide information on degradation, solvent compatibilisation effects when compared to the 1st scan, etc.). For obtaining information on the molecular order, the authors can compare the enthalpy of fusion for the various blends (1st heating) normalised to the polymer fraction. I certainly would also show the data for compositions in the 1:1 range (see also below).
-To address the reviewer's concerns, GISAXS measurements were carried out on the fresh and aged PCE11:PCBM samples. The fresh and aged neat PCE11 samples were measured as references. As depicted in Figures R2 and R3, the GISAXS data on fresh and aged PCE11:PCBM samples clearly confirm the demixing of the amorphous donor and acceptor phases. Figure R4 shows that fresh and aged samples have a peak position corresponding to a domain size of about 44 and 77 nm, respectively. This model-independent analysis is self-consistent with the fitting of GISAXS profiles using the combination of poly-dispersed spheres having a Schultz size distribution with the hard-sphere interaction between PCBM clusters and Debye−Anderson−Brumberger (DAB) model, see Table R1. It has to be underlined that PCE11 strongly aggregates in solution at temperatures lower than 70°C. Different from the DSC sample preparation report previously, 4 Table R3.
We fully agree with the reviewer that the enthalpy change of PCE11:PCBM blends would give valuable information on the molecular order. However, it might be a bit difficult to explore much information from the PCE11:PCBM blends, as the melting peak of PCE11 is strongly overlapped with that of PCBM at ~281°C. As mentioned above, the melting peak at ~281°C (1. heating scan) represents the melting process of both PCE11 and PCBM crystallites. We could distinguish from the 2 nd heating scan that the melting peak at ~285°C consisted of PCBM crystallites, and the one at ~254°C mainly of the PCE11:PCBM mixture crystallites. As the melting temperature of PCE11 crystallites is overlapped with that of PCBM crystallites, the melting peak at 281°C for PCE11:PCBM 1:1 blend cannot be deconvoluted to extract more information. Although detailed analysis of the thermal behaviour of the PCE11:PCBM blends from DSC measurements is already out of the scope of this manuscript, the aforementioned discussion along with Figure R10 and Table R3 was added to the Supplementary Information. Figure R10 thermal bahavior of PCE11, PCBM and PCE11:PCBM 1:1 blend measured from DSC heating and cooling scans. Scale bar: 0.5 W/g. Table R3 Enthalpy change of PCE11, PCBM and PCE11:PCBM 1:1 blend. (1) Melting peak at low temperature. (2) Melting peak at high temperature.     3) The above is important as I think the authors refer to spinodal donor-acceptor demixing as they like to suggest that demixing of the two materials occurs only in the amorphous phase. This is not that clearly spelled out in the manuscript -if the authors want to stick to this hypothesis. In my opinion, it may be, but it may not; based on the data provided I am not convinced. It still could be order-induced phase separation -just molecular ordering on smaller length scales than e.g. accessible with X-ray diffraction. I think it will be difficult to obtain data that solves this issue fully. I am not sure if it is needed at this stage. The burn-in is clear and I fully agree that it originates from phase separation. I personally would rephrase this part and may take out the word 'spinodal': i.e. I would suggest to simply go for 'Abormal Moreover, to clearly present and highlight the significance of the work on theoretical calculation, the work flow of the calculation process is schematic illustrated in Figure   R12. Starting from the Molecular Structure, the required information can be step by step calculated for predicting the miscibility of two components. The calculated molecular parameters are in great accordance with the experimental values as well as the values reported in literature. It is worthwhile to again underline that this work not only delivers the information on the demixing of donor and acceptor phases in PCE11:PCBM system, but also demonstrates an elaborated protocol on characterizing, analyzing and predicting the phase behavior and the miscibility of donor and acceptor materials. This protocol, which is a very powerful tool for design and development of next generation OPV systems with promising stability and reliability, is definitely of great significance and interest to the community. This phase behaviour implies that both melting points are depressed; the one of PCE11 seems more strongly depressed as the eutectic point is so much shifted towards the polymer rich side. Eutectic temperatures is around 250 C, hence the shoulder on the higher-temperature side for the 85% blend indicates that this feature is the PCBM melting -i.e. in my opinion also the PCBM melting is depressed. Clearly, having some more data for the compositions between 15% to 85 % will assist to identify correctly the eutectic composition. This will have significance whether it is or not spinodal decomposition.

PCE11
-We greatly thank the reviewer for the helpful comments. The required DSC heating scans of PCE11:PCBM blends are summarized in Figure R13a. According to the DSC data, we fully agree with the reviewer that the melting point depression was observed for both PCE11 and PCBM crystallites. The crystallites of PCE11 are more strongly depressed than that of PCBM. The eutectic point was found at ~253°C, and can be already resolved from the blends with 85-90% PCE11, indicating that the thermal behavior of PCE11 crystallites were easily influenced by adding small amount of PCBM. The melting peak of PCBM crystallites can still be detected for the blends with up to 60% PCE11. This is significantly different from the P3HT:PCBM system previously reported by us. 4 As shown in Figure R13b, the melting peak of PCBM crystallites can only be detected for the blends with 0-30% P3HT. The difference in thermal behavior of PCE11:PCBM and P3HT.PCBM reveals that P3HT is more miscible with PCBM than PCE11, which is in excellent agreement with the findings demonstrated in this work. 2) I do not want to suggest even more experiments, but out of curiosity, have the authors performed measurements with ICBA with a higher Tg and it seems somewhat better miscibility with P3HT?
-We fully agree on this useful suggestion. We tried to fabricate and optimize the solar cells based on PCE11:ICBA. However, the PCE11:ICBA did not exhibit satisfied performance compared to the PCE11:PCBM control cells, as shown in Figure R14.
Although reasonable high V OC was obtained for the PCE11:ICBA due to the preferable energetic levels, the significantly low J SC and FF have to be attributed to the insufficient charge carrier dissociation and elevated charge recombination, which is strongly related to the morphological properties of the PCE11:ICBA blend.   the Table R5.
where χ 1,2 is the polymer-fullerene interaction parameter, v 0 is the molar volume of the lattice site in the Flory-Huggins model. The entropic contribution is usually between 10 −6 and 10 −2 , which is smaller in magnitude than the enthalpic contribution given for polymer solvent (approximately 0.34). [9][10][11] Therefore, the entropic contribution was not taken into account for calculation.   In summary, I think the present work is intriguing and a broad audience will be interested in it and will learn from it. Hence, I strongly suggest publication -after issues pointed to above have been clarified or addressed.

Referee: 3 Comments to the Author
The author demonstrated strong burn-in degradation in highly-efficient state-of-the-art OSCs induced by spinodal demixing of the donor and acceptor phases, which dramatically reduces charge generation and can be attributed to the inherently low miscibility of both materials. In my opinion, the topic of "stability of OSCs" is very important but the novelty of this paper is not enough for Nature Communications. I think Scientific Reports is a good choice for this paper. Some comments are shown as below: 1. We already know the reason of this degradation (metastable state of morphology), but we don't know how to overcome it. That's the point! -We'd like to thank the reviewer for time and effort spent in evaluating our work.
However, we politely disagree with the reviewer´s opinion that this manuscript lacks in novelty or impact for publication in Nature Communications. As follows, we summarize in short why we believe that this work is indeed of great significance and interests to the community and the broad readership of Nature Communications.
We demonstrate in this work that the abnormal strong burn-in loss of PCE11:PCBM solar cells, which occurred even at room temperature in the dark, can be ascribed to a spontaneous spinodal demixing of the amorphous donor and acceptor phases.
Systematic investigations have been carried out to directly and indirectly verify the microstructure changes in the PCE11:PCBM BHJ morphology. Based on the elaborated experimental data and analysis summarized in the manuscript, we can finally concluded that donor-acceptor demixing in the amorphous regions is indeed caused by the inherent low miscibility of the two components, and is therefore inevitable for such highly efficient OPV system. This information is, for the first time, clearly demonstrated for the current generation of state-of-the-art OPV systems and is of great importance to the research community.
Moreover, as the abnormal burn-in loss is induced by the inherently low miscibility of the donor and acceptor materials, design and development of novel organic materials with promising efficiency and proper miscibility will be the most elegant way to solve these microstructure meta-stability issues. In this work, we demonstrate for the first time a systematic procedure allowing to predict the stability of a BHJ microstructure based on a combined experimental / theoretical approach. This procedure is not only a powerful tool for designing and developing the next generation OPV material systems with superior stability and reliability, but also very useful to understand the phasebehaviour, limits and potential of all kinds of organic electronic devices, such as organic photodetectors, organic light emitting diodes, etc.
2. There is only one high-efficiency polymer/fullerene pair discussed in this paper, it's not enough to make a strong conclusion based on one case.
-We thank the reviewer for the comment, but politely disagree with the reviewer's opinion that the conclusion of our work is not strong enough. This work demonstrates on the abnormal strong burn-in loss observed for PCE11:PCBM system. The drastic reduction in J SC of highly efficient solar cells, which occurred even at room temperature in the dark, is ascribed to the demixing of the amorphous donor and acceptor phases. As this conclusion was made based on systematic investigations, including temperature-dependent J-V characterization, EL, FTPS, GIWAXS and GISAXS data, we believe that the messages and conclusion delivered by this work are very strong and rigid.
We don't think the observed abnormal strong burn-in loss needs to be proven using other high-efficiency polymer/fullerene pairs, as this is not the general case for all OPV material systems. As also accepted by the reviewer, the degradation is usually induced by the metastable state of BHJ morphology, which changes upon external stress such as thermal annealing or solvent vapor annealing. However, this work demonstrates an extreme case observed for the state-of-the-art OPV material system, in which the finely-mixed donor-acceptor amorphous regions phase-separated even at room temperature in the dark, owing to the poor inherent miscibility of the two components. This information, which is not only related to OPV technology, but also very useful to understand the limits and potential of all kinds of organic electronic devices, must be delivered to the community, especially to chemists and materials scientists for future design of highly efficient and stable organic materials.
3. After the degradation of OSCs, the V OC always higher. Please add some discussion.
-It is generally accepted that the V OC of OPV devices is directly related to the effective bandgap E eff of BHJ system, which is defined by the following equation 13 : where E CT is the band maximum of the charge transfer absorbance, and w is the width of the Gaussian fitting for the CT absorbance.
According to Vandewal et al., an excellent correlation was found to define the V OC of OPV systems, as shown in the following equation 13 : The effective bandgap E eff is strongly related to the nature of the donor-acceptor interfaces in BHJ blends. By reducing the donor-acceptor interface area, the V OC of solar cells can be correspondingly enhanced. 14  and V OC of solar cells is well known to the OPV community, the discussion on V OC changes for aged PCE11:PCBM samples is beyond the scope of the current work, we didn't add the above discussion to the previous manuscript.
To summarize, we greatly acknowledge the time and effort the 3 rd reviewer spent in evaluating our current work. However, we politely disagree with the reviewer's opinion that our work is not relevant enough for publication in Nature Communications. We hope that our response to all reviewers could also convince reviewer 3 that spinodal demixing in the amorphous phase of polymer-fullerenes blends is indeed a most crucial process determining the long time stability. The origin of the spinodal demixing lies in the low miscibility of the two components, and the direct consequence on the device performance is a fast demixing and J SC burn-in.
Theoretical calculations of the interaction parameters underline the importance of miscibility and suggests to further explore this as a design parameter for developing the next generation of stable and efficient solar cell materials. We hope that this may convince reviewer 3 that this work is of great significance and broad interest to the research community.
Reviewer #1 (Remarks to the Author): Decision: Accept pending minor revisions.
I am broadly satisfied with the revisions to the manuscript. It now makes a much stronger case for the origin of the burn-in behaviour and its implications. A small number of revisions are requested before the manuscript is suitable for publication in Nature Communications.
-We greatly thank the reviewer for the very helpful comments.
1) Solar cell performance metrics for the fresh device batches should be tabulated in the supporting information (S.I.).
-The following information was added to the supplementary information, Table S1. 2) References to '1 sun illumination' should read '1 sun equivalent illumination' because the LED output does not match the AM 1.5 spectrum.
-The description was revised according to the reviewer's suggestion.
3) Lines 166-169. The Tg value of 14.6 o C should be treated with more caution than the authors currently communicate because the m-DSC measurements do not replicate the thermal histories of the blend films used for the solar cell devices. The absence of DIO in the m-DSC samples should also be explicitly noted. Please address this in the main text.
-The description in page 8 was modified as follows: "We attempted to determine the glass transition temperature (T g ) of neat materials and BHJ blends by means of temperature-modulated differential scanning calorimetry (m-DSC).
However, as depicted in Supplementary Fig. 6 4) The EL spectra corresponding to the fresh and aged BHJ samples should also be presented at constant current density (e.g. as a complementary figure in the S.I.). Does the aged BHJ sample have higher electroluminescence external quantum efficiency?
-The EL spectra of fresh and aged PCE11:PCBM samples measured at an external constant current of 50 mA were added to the S.I., Figure S7. The aged BHJ sample shows higher EL external quantum efficiency than the fresh sample, as the singlet emission of PCE11 and PCBM are also involved in the emission spectrum of the aged BHJ sample.