Sublimed C60 for efficient and repeatable perovskite-based solar cells

Thermally evaporated C60 is a near-ubiquitous electron transport layer in state-of-the-art p–i–n perovskite-based solar cells. As perovskite photovoltaic technologies are moving toward industrialization, batch-to-batch reproducibility of device performances becomes crucial. Here, we show that commercial as-received (99.75% pure) C60 source materials may coalesce during repeated thermal evaporation processes, jeopardizing such reproducibility. We find that the coalescence is due to oxygen present in the initial source powder and leads to the formation of deep states within the perovskite bandgap, resulting in a systematic decrease in solar cell performance. However, further purification (through sublimation) of the C60 to 99.95% before evaporation is found to hinder coalescence, with the associated solar cell performances being fully reproducible after repeated processing. We verify the universality of this behavior on perovskite/silicon tandem solar cells by demonstrating their open-circuit voltages and fill factors to remain at 1950 mV and 81% respectively, over eight repeated processes using the same sublimed C60 source material. Notably, one of these cells achieved a certified power conversion efficiency of 30.9%. These findings provide insights crucial for the advancement of perovskite photovoltaic technologies towards scaled production with high process yield.

1.The available evidence fully demonstrates that the C60 powder collected from the crucible after the 8th cycle produces a large change, but there is no direct evidence for the difference of C60 evaporated onto the perovskite from different batches.Direct analysis of the changes in the evaporated C60 rather than the changes in the composifion of the residual C60 can befter help us understand this phenomenon.
2. What is the relafionship between material purity and the electronic defects?What kinds of impurifies will cause deep level electronic trap states and how these coalesced C60 produce these trap states?Does this caused by energy levels or other factors?
3. On the other hand, for as-received C60, is there any difference in the morphology and thickness of C60 layers in different batches?
4. If the evaporafion equipment is integrated in the glove box and there is no oxygen introducfion between different batches, is there a problem with the repeatability of the device?Similarly, what would happen if each batch was cooled to room temperature before oxygen was introduced.Can this solve the problem of oxygen-induced coalescence? 5.Although the efficiency of tandem solar cells is very high, the efficiency of single-juncfion perovskite solar cells is relafively low, especially the open circuit voltage, even after introduce CaF2 as interlayer in figure s11.How can such inefficient single-juncfion devices achieve efficient tandem devices?6.In figure 1d, the HTL is NiO/MeO-2PACz.However, in Perovskite/silicon tandem solar cells fabricafion, the HTL is 2PACz.Is there any difference for these two hole-selecfive materials?
Reviewer #3 (Remarks to the Author): This paper aims to elucidate the effects of repeated C60 evaporafion on single-juncfion and silicon/perovskite tandem solar cell devices, with a parficular focus on the high efficiency (31%) achieved in the silicon/perovskite tandem configurafion.The study appears to have meficulously conducted film analyses and calculafions concerning the reuse of C60, as depicted in Fig. 2 and Fig. 3.If the conclusions of this study prove valid, they could offer invaluable insights to numerous research groups employing C60 and industrial insfitufions endeavoring to commercialize perovskite single-juncfion or tandem devices with a p-i-n configurafion.However, a substanfial set of quesfions and concerns arise concerning the primary aspect of this paper -the device data intended to substanfiate the main claims of the authors.Addressing these concerns is imperafive for mandatory revision, as outlined below, before I could recommend the publicafion of this manuscript in Nature Communicafions.
Firstly, the pivotal data that supports the author's conclusions, as presented in Fig. 1, appears insufficient to robustly validate their claims.If I understand correctly, for single-juncfion data, the authors fabricated the half-device (ITO/HTL/Perovskite) within the same batch.Subsequently, they evaporated C60 for the first set of substrates, stored them in a load lock (under low-vacuum condifions), evaporated C60 for the second set of substrates from the same batch as the first C60 devices set, similarly stored them in a load lock, and repeated this process up to the eighth C60 set before finally introducing BCP/Ag and conducfing the experiment.This approach seems appropriate for assessing the influence of the number of C60 deposifions.However, at this juncture, I'm curious that how many fimes the authors repeated this experiment, and the specifics of the experimental design.To ensure that the repeated use of C60 genuinely affects device performance, it is crucial for the authors to replicate this process in separate experiments at least 3-4 fimes (not for 3-4 substrates at one fime) using the same batches of NiOx/SAM/Perovskite layers.Explicitly detailing the steps taken to ensure controlled condifions that exclusively account for the impact of C60 reuse, while excluding other potenfial performance variafions, is essenfial to convince the readers.
Addifionally, measures to mifigate variables stemming from the fime spent in the load lock need to be established (e.g., comparing substrates stored without using the load lock and substrates stored in the load lock for the same durafion as the deposifion fimes for the 1st to 8th C60 experiments, followed by experiments using fresh C60 to verify consistency).
Furthermore, a major factor contribufing to the lack of persuasiveness in the device data is the behavior observed in the Fig. 1 graph.A drop is evident between the first and second data points, followed by almost consistent values between the 2nd and 8th data points, with a subtle decrease at the end.This paftern does not match well with the QFLS data in Figure 2. Efficiency remains nearly constant between the 2nd and 7th data points (excluding the 1st and 8th).It is imperafive to elucidate whether the divergent efficiencies at the 1st and 8th data points are coincidental or genuinely aftributed to the reuse of C60.The tandem data in Fig. 1 does not exhibit this discrepancy as prominently.The J-V curves in the tandem data are almost superimposed, making it challenging for readers to accept the efficiency drop posited by the authors.While a numerical efficiency drop of 0.6% presented, it's a small value; it might be more compelling to represent this reducfion through the distribufion illustrated in the supplementary materials.
I acknowledge that repeafing the enfire experiment for all condifions is a considerable undertaking.Therefore, I recommend focusing on replicafing 2-3 types of devices: 1) using fresh C60, and 2) applying 8 cycles of heafing and cooling for C60(w/o fabricafing device), followed by fabricafing the device repeatedly using the same batch of perovskite layers, while verifying the absence of any load lock storage influence.Addifionally, it could be beneficial to introduce one more condifion in the middle, involving 4 cycles of heafing and cooling, if feasible for the authors.This type of data could be stafisfically integrated into Figure 1.If the authors' claims hold true and similar results are achieved, the disparifies could be more explicitly displayed on the Fig. 1 graph, featuring an ample number of data points.Moreover, with the careful specificafion of the controlled condifions menfioned earlier, readers are likely to find greater resonance with the authors' asserfions alongside the improved trends demonstrated by the enhanced Fig. 1 graph.

Reviewer #1
This manuscript reports the influence of C60 aggregation on perovskite solar cells' performance and the reproducibility of p-i-n device configuration useful for silicon/perovskite tandem solar cells.The findings highlight the evolution of C60 thin films' electronic quality through repeated evaporation processes and its impact on device performance.By using purified C60, it's possible to mitigate the negative effects and maintain device performance for perovskite-based photovoltaic technologies.The study contributes to a better understanding of the role of C60 thin films in electronic devices and offers practical insights for optimizing their performance.The findings are in the interest of Nature Communications' readership, and we recommend publication after addressing the following queries.
We thank the reviewer for the positive feedback and recommendation for publication.We appreciate the reviewer's assessment of the paper's significance and the clarity of the presented mechanism.In our response letter, we carefully considered the reviewer's suggested changes.
(1) The authors report purified (sublimed) C60 avoids aggregation issues, and device performance remains unaffected even after repeated deposition cycles.What are the impurities that are removed by sublimation?
In our initial submission, we provided in Figure S14 (has been updated in the Supplementary Information as Figure S17) the high-performance liquid chromatography (HPLC) chart which confirms that as-received C60 has a purity of 99.75-99.8% and includes C60 oxide, C70, and C60 dimers as well as unidentified impurities eluting before C60.While not definitively confirmed, such relatively early elution while absorbing at 330 nm could point to unidentified impurities being fullerene fragments or fullerene adducts.After sublimation, the purity of sublimed C60 increased to 99.95% (as shown in Figure S18) mainly by removing the unidentified impurities from the powder.We updated Figure S17 and S18 with better integrations.
(2) The electron mobilities (µe) of the 1st deposition (0.083 cm 2 /V s) are higher than the 8th deposition (0.034 cm 2 /V s), implying the structural changes of the C60 layer, which is important to characterize by spectroscopic techniques.
Thanks for raising this point to support our manuscript.This question motivated us to do more analysis on thin films, other than powders.To respond to the reviewer's question, we carried out a new MALDI-TOF mass spectroscopy analysis for newly prepared fresh (1st cycle) and thermally cycled (>8 cycles) as-received (non-sublimed C60) C60 thin films.For this, we deposited layers on large clean glass substrates and scratched them with clean blades, and we performed the measurements on collected powders.
As shown in Figure R1, the MALDI-TOF analysis indicates the presence of the peaks at 1322, 1345, 1369, 1392, and 1417 m/z in both C60 thin films with different peak intensities, which are assigned to C110, C112, C114, C116, and C118, respectively.Importantly, the peaks of thermally cycled thin film are stronger than that of fresh thin film, which confirms the concentration of the higher molecular weight fullerenes in thermally cycled thin film is higher than that of fresh thin film.
Interestingly, the fresh C60 thin film lacked the peak at 1441 m/z for C120, which appeared in the thermally cycled thin film.Figure R1 has been added to the main text as Figure 2f.(3) The presence of an emission shoulder in the uncontaminated sample, particularly around 800 nm, becomes more pronounced in the emission peak of thermally cycled powders at 835 nm.This is in addition to the primary peak for pure C60 situated at 742 nm, underscoring that even the initial sample exhibits certain issues.
We understand that the reviewer refers to the PL spectra acquired from powders shown in Figure S6b.Indeed, we have a shoulder even for the as-received C60 powders (before repeated deposition), and obviously, we have some other impurities in these samples as reported in the HPLC analysis in Figure S17.
To understand the relation of impurities on the peak located around 835 nm, we now analyzed the sublimed C60 powder via PL, in addition to our previous analysis on asreceived C60 powder.As shown in Figure R2, the sublimed C60 showed a weaker peak at 835 nm, which implies even after purification still the peak exists and that the peak at 835 nm becomes stronger with an increase in the impurity content of C60. Figure R2 has been added to the Supplementary Information as Figure S23b.
In literature, the PL peak at 742 nm is assigned to radiative recombination, while the peak at 835 nm is attributed to phonon replicas, which is an inherent feature of C60 (Solid State Communications, Vol. 98, No. 9, 853-858, 1996).

Reviewer #2
This manuscript mainly reports that low-purity C60 will affect batch repeatability of device performance due to oxygen-induced coalescence during repeated thermal evaporation processes.The coalescence of C60 leads to the formation of deep states, increasing additional nonradiative recombination, thus resulting in a decrease in device performance.After sublimation and purification of C60, the purity has been improved to 99.95%, effectively solving the problem of repeatability, and achieving a certified power conversion efficiency of 30.9% for perovskite-silicon tandem solar cells.This work provides unique insights into the fabrication of C60 layers in perovskite-based devices, so I recommend publishing it in Nature Communication.The following questions are suggested for improving the manuscript.
We extend our gratitude to the reviewer for recognizing the distinctiveness and depth of our contribution to the perovskite-based solar cells field.We appreciate the recommendation for the publication of our work in Nature Communications.With this opportunity, we improved the quality of our article further by addressing the queries of the reviewer.
1.The available evidence fully demonstrates that the C60 powder collected from the crucible after the 8th cycle produces a large change, but there is no direct evidence for the difference of C60 evaporated onto the perovskite from different batches.Direct analysis of the changes in the evaporated C60 rather than the changes in the composition of the residual C60 can better help us understand this phenomenon.
We thank the reviewer for raising this critical point.This question implies the same manner as R1's 2 nd question.We kindly request the reviewer to refer to that section.
2. What is the relationship between material purity and the electronic defects?What kinds of impurities will cause deep level electronic trap states and how these coalesced C60 produce these trap states?Does this caused by energy levels or other factors?
Regarding the purity level of the materials; according to the HPLC analysis, we observed C60 oxide, C70 and C60 dimer, and unknown impurities eluting before C60.These unknown impurities, possibly fullerene fragments and/or fullerene adducts, are expected to have electronic properties that are significantly different from those of C60 and could therefore negatively impact electron transport.
However, we need to keep in mind the effect of the vapor deposition process of C60.The sublimation temperatures of possibly present fullerene fragments can be expected to be significantly lower than that of C60 and therefore such fullerene fragments are unlikely to be deposited on the substrate at the process conditions used here (it might be evaporated during soaking time when the substrate shutter was closed, which means It can present in the films).In that concern, fullerene adducts usually decompose, often reforming C60, under the conditions of vapor deposition (see e.g., Giovane et al., Kinetic Stability of the C60-Cyclopentadiene Diels-Alder Adduct.J. Phys.Chem. 1993,97, 8560-8561.).This explains the very similar device performance in the first vapor-deposition cycle for both as-received and sublimed C60 powders.
From ultraviolet photoelectron spectroscopy (UPS) analysis, which was already given in Figure 2e (main text) in Figure S6a, we have seen that the states close to the highest occupied molecular orbital (HOMO), or in other words valance band maximum (VBM) show some changes after thermal cycling of the as-received powders but we have not seen any additional states close to the band edge.From PDS analysis, we observe that the C60 impurities directly cause nonradiative recombination, which leads to loss in Voc.
With our analysis, we demonstrated that the recombination occurs on the perovskite surface.This means that either the defects are created on the PK/C60 interface or the defects are directly in the C60 layer.The DFT calculations show deep electronic states at the degraded C60/PK interface, which strongly supports the creation of electronic defects by C60 impurities.
3. On the other hand, for as-received C60, is there any difference in the morphology and thickness of C60 layers in different batches?
Upon the reviwer's request, we investigated the surface morphology of fresh and thermally cycled as-received C60 films was investigated by using atomic force microscopy (AFM).There is no significant change in the morphology of fresh and thermally cycled as-received C60 films as shown in Figure R3 a and b, respectively.In addition, the root mean square (RMS) of fresh as-received C60 film is 1.2 nm, while the RMS of thermally cycled as-received C60 film is 1.47 nm.The thickness of C60 film is controlled during the deposition by quartz crystal microbalance (QCM).Therefore, the thickness of fresh and thermally cycled C60 films was controlled to reach 20 nm. Figure R3 has been added to the Supplementary Information as Figure S9. 4. If the evaporation equipment is integrated in the glove box and there is no oxygen introduction between different batches, is there a problem with the repeatability of the device?Similarly, what would happen if each batch was cooled to room temperature before oxygen was introduced.Can this solve the problem of oxygen-induced coalescence?
Our evaporation tool facilitates sample exchange within a vacuum-sealed load lock, as detailed in the Methods section.This load lock is also interconnected with a glove box.During the processing phase, the deposition chamber maintains an ultra-low vacuum level of approximately 1E-7 Torr, whereas the load lock can sustain a vacuum level of 1E-4 Torr.Therefore, we would like to highlight that achieving an entirely oxygen-free environment poses significant challenges, whether within a glove box or under a 1E-7 Torr vacuum.
In line with our experimental approach aimed at mimicking industrial processes (which is not supposed to involve glove boxes), we stored samples in the deposition queue within a load lock operating under a mild vacuum range of 1E-3 to 1E-4 Torr.
Before initiating the load lock transfer, it is important to note that the samples had never been exposed to ambient air.Instead, they were transferred through a glove box with an oxygen level of less than 10 ppm.Once located in the load lock, the system undergoes a pumping process to attain the vacuum level of 1E-4.Subsequently, a gate valve facilitates the transfer of samples into the deposition chamber.During this exchange, the temperature of the C60 crucible is around 240°C (which we assume is a similar condition to the industrial process).
In our system, the practicality of cooling down the source to room temperature without opening the gate valve between the deposition chamber and load lock is compromised, as the system automatically removes the sample from the chamber immediately after deposition, without allowing time for cooling.Nevertheless, we anticipate observing a performance drop in the subsequent deposition cycle, attributable to the coalescence effect explained in the manuscript.To answer the reviewer's question, we performed a new experiment to understand how heated/cooled powders affect the performance after the first cycle (following the reviewer's guide on this).Figure R4 shows how the cooling down of the crucible decreases the Voc and FF of the devices.The results confirm our previous observations.We believe there might be a misunderstanding here.The perovskite layers showed in Figures 1b and Figure S11 (now has been updated as S13) for the single-junction devices were fabricated using a hybrid method involving evaporation and spin coating.This approach differs from the one-step spin coating method that serves as our baseline for tandem cells, which is also used for high PCE cells (we note here the focus of this work is not perovskite itself).
In the hybrid method, the inorganic scaffold PbI2/CsBr layer is co-evaporated onto the substrate and subsequently transformed into the perovskite layer by introducing an ethanolic solution containing MACl, FAI, and FABr.This detail was explicitly highlighted in the caption of Figure 1, which reads: "Here, the perovskite layers of single-junction PSCs were fabricated using a hybrid method that consists of a two-step process.Initially, an inorganic template was evaporated, and then a solution conversion step was employed to accomplish conversion into the perovskite phase."We intentionally employed this hybrid method due to its ability to yield more homogeneous and reproducible devices on glass substrates (despite its lower performance than the onestep solution method), resulting in narrower device statistics.This choice played a critical role in our ability to conclude the repeatability issue observed during multiple C60 depositions.
6.In figure 1d, the HTL is NiO/MeO-2PACz.However, in Perovskite/silicon tandem solar cells fabrication, the HTL is 2PACz.Is there any difference for these two hole-selective materials?
The perovskite layer of single-junction devices in Figure 1d (main text) is deposited by evaporation/spin coating (hybrid method).During our optimizations, we found the suitable HTL combination for this method is NiOx/MeO-2PACz among many other options (we have not included this data here as it is out of scope).
However, the optimal HTL for one-step solution-processed perovskite/silicon tandem solar cells is 2PACz.The perovskite layer in perovskite layer in perovskite/ silicon tandem solar cell is deposited by the spin coating method.We note that our recently published article which delivers 32.5% tandem cells is based on a one-step solution-processed perovskite is also utilizes 2PACz as an HTL (doi.org/10.1038/s41586-023-06667-4). We believe this is well demonstrated in our previous works.

Reviewer #3
This paper aims to elucidate the effects of repeated C60 evaporation on single-junction and silicon/perovskite tandem solar cell devices, with a particular focus on the high efficiency (31%) achieved in the silicon/perovskite tandem configuration.The study appears to have meticulously conducted film analyses and calculations concerning the reuse of C60, as depicted in Fig. 2 and Fig. 3.If the conclusions of this study prove valid, they could offer invaluable insights to numerous research groups employing C60 and industrial institutions endeavoring to commercialize perovskite single-junction or tandem devices with a p-i-n configuration.However, a substantial set of questions and concerns arise concerning the primary aspect of this paper -the device data intended to substantiate the main claims of the authors.Addressing these concerns is imperative for mandatory revision, as outlined below, before I could recommend the publication of this manuscript in Nature Communications.
We would like to thank to the reviewer for the constructive comments and suggestions to further improve the quality of our manuscript.
Firstly, the pivotal data that supports the author's conclusions, as presented in Fig. 1, appears insufficient to robustly validate their claims.If I understand correctly, for singlejunction data, the authors fabricated the half-device (ITO/HTL/Perovskite) within the same batch.Subsequently, they evaporated C60 for the first set of substrates, stored them in a load lock (under low-vacuum conditions), evaporated C60 for the second set of substrates from the same batch as the first C60 devices set, similarly stored them in a load lock, and repeated this process up to the eighth C60 set before finally introducing BCP/Ag and conducting the experiment.This approach seems appropriate for assessing the influence of the number of C60 depositions.However, at this juncture, I'm curious that how many times the authors repeated this experiment, and the specifics of the experimental design.To ensure that the repeated use of C60 genuinely affects device performance, it is crucial for the authors to replicate this process in separate experiments at least 3-4 times (not for 3-4 substrates at one time) using the same batches of NiOx/SAM/Perovskite layers.Explicitly detailing the steps taken to ensure controlled conditions that exclusively account for the impact of C60 reuse, while excluding other potential performance variations, is essential to convince the readers.
We do agree with the reviewer that the repeatability of our observations is quite critical to draw a conclusion, which we were also quite careful before elaborating our interpretations.Before starting a more detailed interpretation, we repeated this experiment 3 times.First time, with 1.68 eV hybrid process single-junction perovskite solar cells as shown in Figure R5a.To verify the universality of the C60 behavior after multiple deposition cycles, we repeated the same experiment with 1.68 eV and 1.55 eV one step solution-process perovskite solar cells as shown in Figure R5b and c, respectively.Furthermore, we carried out the same experiment with solution-processed perovskite/tandem solar cells (Supplementary Information Figure S3).However, to build a further confidence on our work, following the reviewer's recommendation, we performed the experiment one more time as shown in Figure R5d.Figures R5 b, c and d have been added to the Supplementary Information as Figure S2.So, we believe that the readers will have full confidence about the repeatability/consistency/universality of this issue.We thank the reviewer once more for giving us an opportunity to bring a further confidence to our results.Additionally, measures to mitigate variables stemming from the time spent in the load lock need to be established (e.g., comparing substrates stored without using the load lock and substrates stored in the load lock for the same duration as the deposition times for the 1 st to 8 th C60 experiments, followed by experiments using fresh C60 to verify consistency).
We thank the reviewer for raising this interesting aspect.To answer this question, we designed a device fabrication batch by keeping the samples in; 1) Glove box with O2 and H2O levels <0.1 ppm 2) Load lock by keeping the samples in vacuum level 1E-4 Torr for 5 hours (assuming an equivalent time for one lot).Later, we deposited C60 layers by using as-received fresh powders in the same batch.Figure R6 summarizes the Voc and FF of those devices.We found that the Voc and FF of vacuum-stored substrates were unaffected by the vacuum as shown in Figure R6.The same results were obtained when we compared the performance of the fresh perovskite substrates and N2-stored perovskite substrates.(Figure R6 has been added to the Supplementary Information as Figure S16 a and b).
These findings confirm that the origin of Voc and FF losses is not due to the storage, it is related to the changes in C60 powder after multiple deposition cycles.Furthermore, a major factor contributing to the lack of persuasiveness in the device data is the behavior observed in the Fig. 1 graph.A drop is evident between the first and second data points, followed by almost consistent values between the 2nd and 8th data points, with a subtle decrease at the end.This pattern does not match well with the QFLS data in Figure 2. Efficiency remains nearly constant between the 2nd and 7th data points (excluding the 1st and 8th).It is imperative to elucidate whether the divergent efficiencies at the 1st and 8th data points are coincidental or genuinely attributed to the reuse of C60.The tandem data in Fig. 1 does not exhibit this discrepancy as prominently.The J-V curves in the tandem data are almost superimposed, making it challenging for readers to accept the efficiency drop posited by the authors.While a numerical efficiency drop of 0.6% presented, it's a small value; it might be more compelling to represent this reduction through the distribution illustrated in the supplementary materials.
Regarding Figure 1c, it is noteworthy that the mean average open-circuit voltage values resulted in a slight drop from the 2nd to the 4th deposition.Subsequently, from the 5th to the 8th deposition, a significant decline was observed, which aligns with the results of the QFLS (Quasi Fermi Level Splitting) measurements.In our QFLS analysis, we specifically compared the results obtained from the 1st, 6th, 7th, and 8th depositions to reduce the experimental load.
Furthermore, the PCEs exhibited a slight decrease due to the reuse of C60, primarily affecting the open-circuit voltage, as it did not significantly impact the short-circuit current densities.To validate our findings, we conducted the experiments four times.
1. We reported the repeated deposition results in Figure R5. 2. To further verify our results independent from the perovskite absorber layers, we repeated these experiments with solution-processed perovskite absorbers with energy bandgaps of 1.55 and 1.68 eV, and again we observed the same trend (see results in S2) 3. Later, we considered adding a CaF2 layer between the perovskite and C60 layers thinking this contact displacement can solve the performance loss issues (which we understood later the C60 itself has degenerated after repeated processes).
Even in this case, we saw the same trend as shown in Figure S13.4. Following the reviewer's suggestions, we repeated this experiment one more time with 1.68 eV hybrid perovskite absorber and gave the results, Figure S2.
Concerning the tandem devices, the J-V curves represent the best-performing device with fresh and thermally aged as-received C60.
Notably, we observed the same behavior in tandem devices as in the single-junction devices, as illustrated in Figure S3.So, we believe the achieved results show a strong consistency.
I acknowledge that repeating the entire experiment for all conditions is a considerable undertaking.Therefore, I recommend focusing on replicating 2-3 types of devices: 1) using fresh C60, and 2) applying 8 cycles of heating and cooling for C60 (w/o fabricating device), followed by fabricating the device repeatedly using the same batch of perovskite layers, while verifying the absence of any load lock storage influence.
Additionally, it could be beneficial to introduce one more condition in the middle, involving 4 cycles of heating and cooling, if feasible for the authors.
We thank the reviewer for helping us to further prove the repeatability of our observations.We performed 8 cycles of the process without fabricating devices and used this powder to fabricate a new deposition.For this, we used identical perovskite half-device stacks fabricated in the same batch.It seems that even in this case, the Voc and FF of fabricated perovskite solar cells follow the same trend as shown in Figure R7.Our experiments performed for the other queries of the reviewer (e.g., storing in vacuum and N2, repeating, etc) are complementary to this answer.We express our gratitude to the reviewer for their valuable recommendation.After thorough deliberation with several authors, we have opted to include this information in the Supplementary Information.This choice was made to prevent excessive space consumption within the main text and to maintain the readers' focus on the in-depth interpretations.We trust that the reviewer will appreciate our decision.
The revised manuscript addresses the three reviewers' comments, which reinforce the significance of the research findings and contribute to a comprehensive understanding of the impact of C60 aggregafion on perovskite solar cell performance.The inclusion of the updated data and figures significantly enhances the manuscript's quality and provides a comprehensive understanding of the influence of C60 aggregafion on perovskite solar cells.The authors addressed my queries safisfactorily and recommended for publicafion.
Reviewer #2 (Remarks to the Author): During this revision,the authors have well adressed all my concerns.I have no further comment and recomend accepfing this paper.
Reviewer #3 (Remarks to the Author): I applaud the authors for their thorough and conscienfious responses to the reviewers' comments.It is evident that the authors have effecfively addressed my previous remarks, primarily through repeated experiments, which have solidified their claims as factual.Therefore, I strongly recommend that the manuscript be published in Nature Communicafions at its current stage.The addifional comments provided below are my personal suggesfions aimed at further enhancing the clarity of the authors' arguments.No further response is necessary.
With regard to Figure 1e, my intenfion was to point out the potenfial confusion for readers in interprefing the JV curve in this figure.Given that the JV curves appear almost idenfical, a casual observer might erroneously conclude that there is a difference in the single-juncfion but not in the tandem configurafion when C60 is reused.To mifigate this potenfial misconcepfion, it might be advisable to exclude the JV curve and present the stafisfical data in Figure S3.Alternafively, even if the tandem data is removed from Figure 1, it may help streamline the core argument.
Regarding the last comment, the suggesfion is not to introduce the retested data (Fig. R5, R7) as a separate figures.Instead, I propose integrafing these data points into Figure 1c, as it could yield a more stafisfically robust outcome.I believe that this approach would befter emphasize the degradafion in efficiency resulfing from the reuse of C60.

Reviewer #1
The revised manuscript addresses the three reviewers' comments, which reinforce the significance of the research findings and contribute to a comprehensive understanding of the impact of C60 aggregation on perovskite solar cell performance.The inclusion of the updated data and figures significantly enhances the manuscript's quality and provides a comprehensive understanding of the influence of C60 aggregation on perovskite solar cells.The authors addressed my queries satisfactorily and recommended for publication.
We thank the reviewer for reviewing the manuscript and appreciate the recommendation for the publication.

Reviewer #2
During this revision, the authors have well addressed all my concerns.I have no further comment and recommend accepting this paper.
We are grateful for the reviewer's comment and recommendation for publication.

Reviewer #3
I applaud the authors for their thorough and conscientious responses to the reviewers' comments.It is evident that the authors have effectively addressed my previous remarks, primarily through repeated experiments, which have solidified their claims as factual.Therefore, I strongly recommend that the manuscript be published in Nature Communications at its current stage.The additional comments provided below are my personal suggestions aimed at further enhancing the clarity of the authors' arguments.No further response is necessary.
We thank the reviewer for the feedback which helps to further improve the manuscript quality.We appreciate it.
With regard to Figure 1e, my intention was to point out the potential confusion for readers in interpreting the JV curve in this figure.Given that the JV curves appear almost identical, a casual observer might erroneously conclude that there is a difference in the single-junction but not in the tandem configuration when C 60 is reused.To mitigate this potential misconception, it might be advisable to exclude the JV curve and present the statistical data in Figure S3.Alternatively, even if the tandem data is removed from Figure 1, it may help streamline the core argument.
Indeed, we provide Figure 1e with the table of photovoltaic parameters which clearly shows the 1 % drop in FF % from as-received fresh C 60 to thermally cycled as-received C 60 .In addition, the two J-V curves of fresh and thermally cycled as-received C60 don't overlap and aren't identical, especially for the open circuit voltage point and maximum power point.Therefore, we prefer to keep these J-V curves still in the main text.We hope that the reviewer will understand our reasoning for this.
Regarding the last comment, the suggestion is not to introduce the retested data (Fig. R5, R7) as a separate figures.Instead, I propose integrating these data points into Figure 1c, as it could yield a more statistically robust outcome.I believe that this approach would better emphasize the degradation in efficiency resulting from the reuse of C 60 .

Figure
Figure R1: MALDI-TOF analysis of fresh and thermally cycled as-received C60 thin films.

Figure R2 :
Figure R2: PL of the fresh and thermally cycled as-received and sublimed C60.

Figure R4 :
Figure R4: Statistical distributions of Voc and FF for single-junction perovskite solar cells using fresh and cooled thermally cycled as-received C60 powder.

Figure R6 :
Figure R6: Influence of the perovskite absorber storage on cells performance: Statistical distributions of Voc and FF for fresh substrates and vacuum-and N2 glove box-stored perovskite absorbers before C60 deposition.

Figure R7 :
Figure R7: Statistical distributions of Voc and FF of fabricated solar cell devices with fresh C60 powder and cooled C60 powder after 8 thermal cycles.