How antisolvent miscibility affects perovskite film wrinkling and photovoltaic properties

Charge carriers’ density, their lifetime, mobility, and the existence of trap states are strongly affected by the microscopic morphologies of perovskite films, and have a direct influence on the photovoltaic performance. Here, we report on micro-wrinkled perovskite layers to enhance photocarrier transport performances. By utilizing temperature-dependent miscibility of dimethyl sulfoxide with diethyl ether, the geometry of the microscopic wrinkles of the perovskite films are controlled. Wrinkling is pronounced as temperature of diethyl ether (TDE) decreases due to the compressive stress relaxation of the thin rigid film-capped viscoelastic layer. Time-correlated single-photon counting reveals longer carrier lifetime at the hill sites than at the valley sites. The wrinkled morphology formed at TDE = 5 °C shows higher power conversion efficiency (PCE) and better stability than the flat one formed at TDE = 30 °C. Interfacial and additive engineering improve further PCE to 23.02%. This study provides important insight into correlation between lattice strain and carrier properties in perovskite photovoltaics.

Many research groups have seen variations in the thickness of their perovskite films and a couple of groups have written specifically on the subject. The topic is important because smooth films make it easier to deposit contact layers that have uniform thickness. Although most people seem to want smooth films, the authors of this manuscript obtained higher efficiency is solar cells with the wrinkling.
I really like the new data in this manuscript. They have made films with many different compositions using several different processing conditions. They observed interesting trends that I doubt many people could have predicted. I definitely think this data should be published, but at this time I find the explanations for the observations not to be convincing. If the authors address my comments, then the manuscript might be important enough for Nature Communications. I find the statement "no systematic studies were carried out to understand the formation mechanism of a pseudo-epitaxial wrinkle morphology depending on composition and/or preparation condition" to be misleading. Reference 20 did far more than simply report that wrinkling exists. It provided xray diffraction and wafer curvature stress measurements at several stages in the film formation process to show how compressive stress causes the wrinkling. It also showed how changes in the solvent composition could mitigate the wrinkling. I think that the authors are building on the explanation in Reference 20 by adding that the wrinkling can more easily occur in a perovskite layer that rests on top of a viscous layer.
I think the claims that the wrinkling improves light management are very misleading. It is well known that texturing can help trap light in solar cells. In this case, however, there are approximately 60 nm variations in height with periodicities of around 15 microns. The schematics are not to scale and greatly exaggerate the extent of the surface height variation. I wouldn't expect such a small change in the surface angle to help much. The EQE plots and the change in Jsc support my position. At best, the light trapping helps just a tiny bit around 700 nm. Reducing the temperature of the diethyl ether deposition clearly had its impact by increasing Voc, which is unlikely to be a result of light trapping.
On page 16, the authors describe the FDTD simulations of how light will propagate in the solar cells. Their analysis of the data is simply "Light absorption is more enhanced in the wrinkled structure than in the flat one." If they properly analyze their FDTD data in a quantitative way and find out the percentage change in light absorption, I think it will support what I have said in the previous paragraph.
At the end of a paper that very thoroughly examines the affect of varying the temperature of the anti-solvent and the primary solvent composition on wrinkling and optoelectronic properties, the authors suddenly say that they raised the efficiency considerably by adding some KI, which has nothing to do with the rest of the paper. They don't actually say how much KI they added. I find that very unsatisfying. The authors must say how much additive they used. Instead of telling readers in the abstract that they reached 23 % efficiency (with KI), they should say what efficiency they obtained as a function of DE deposition temperature and explain that they get higher efficiency in the wrinkled films. That is the real take home message for this manuscript.
I do not understand why the authors say "Since the wrinkle texture certainly has the potential for epitaxial growth due to an underlying buckling process associated with a compressive stress." I don't see what epitaxial growth has to do with buckling or compressive stress. The authors say this work is motivated by obtaining epitaxial growth. I only see one line on page 16 that says that an XRD plot in the supplemental section shows better orientation for the low temperature antisolvent film.
The statement on page 14 that "This indicates lower defect concentrations at the bottom side of the films, which is attributed to the crystal growth direction from the top surface to the bottom." is not clear or well supported by data. The authors reference Stranks et al's manuscript (ref 30) on how stress affects trap density and photoluminescence, but do not say anything about it. I recommend summarizing the main points of that manuscript and using them to try to explain the observations reported in this manuscript. Why do the authors think stress affects the trap density?
In Fig 4, the authors have both a grey scale and a color scale. I don't see how one can have two color scales for a 2D plot. We cannot see the grey scale. I don't think "Topographical tailoring" are good words to use in the title. I was partially expecting the paper to be about topological insulators, which it isn't at all. People won't know what the authors mean by "topographical tailoring." I suggest a title such as "How perovskite composition and antisolvent miscibility affect film wrinkling and optoelectronic properties." I think that the choice of solvents is so important that it should be stated clearly early in the main text, not just in the supplemental section. The way the manuscript is written, one might think that DMSO is the primary solvent. In fact, it is an additive and DMF is the main solvent. The authors might also want to clarify that the DMF probably evaporates first, which is probably why they write about the DMSO more.
On page 4, the authors refer to "blue frames" in Fig 1. It took me a while to notice the blue frames, which are very hard to see. It is slightly hard to follow the trends in how composition affects wrinkling by looking at microscope images, many of which are too small. I really like the plots of lambda vs composition in Fig S1 d and e. I recommend putting those plots in the main text. They helped me understand the trend instantly.
On page 10 (line 218), should "to form lower TDE" be "to form at lower TDE?" Why were the substrates at 15 degrees C during spin casting? That is unusual and inconvenient. There must be a good reason.
Reviewer #2 (Remarks to the Author): Park et al. reported a universal method to produce a pseudo-epitaxial perovskite layer via manipulating strain and compression relaxation in a perovskite film, which lead to a controllable wrinkled film structure. Deep study revealed the wrinkling mechanism of the compressing tensilestrained surface through a bilayer intermediate. Finally, the finding of hill sites showed longer carrier lifetimes and higher optical absorption than the valley sites, providing the basis for over 23% PCE and long-term operational stability of PSCs. This is an interesting and deep work on growth kinetics control of perovskite crystals, and the connection between topological arrangement and photocarriers behavior provides useful guidance for more efficient and stable PSCs design. However, there are still some concerns, which need more clarification: (1) The wrinkling structure only occurs under some specific components (i.e. (FAPbI3)0.875(CsPbBr3)0.125) at given conditions (i.e. TDE is as low as 5 ℃), thus it is hard to popularize this method and the cooling of diethyl ether will increase the production costs. Actually, there have been some components which may hardly form wrinkling structure (Advanced Materials, 2020, 32, 1907757;Science, 2019, 366, 749-753;Nature Photonics, 2019, 13, 460-466, etc.) but exhibit higher PCE and similar stability compared to this work, and all of them are fabricated under simpler process. The authors should comment on this fact and the potential advantages of this method.
(2) The bilayer intermediate is the key for wrinkling structure formation, however, whether such an intermediate will bring into the component inhomogeneity along with the vertical direction of the perovskite films? Moreover, the existence of wrinkling structure may also lead to the component inhomogeneity between hills and valleys. The authors should provide more information on these details.
(3) The amplitude (A) is about 100 nm for optimized perovskite film, although this value may have no obvious influence on Spiro-OMeTAD deposition (which obtains a thickness over 200 nm), however, some more stable HTLs such as PTAA, P3HT is hard to form continuous film when depositing on the wrinkle perovskite, resulting to the existence of lots of shunting paths. This is another concern that hinders the development of this method.
(4) In-situ absorption data showed an perovskite phase with thickness of 250 nm for TDE = 5 ℃, 362 nm for TDE = 15 ℃ and 450 nm for TDE = 30 ℃. Since anti-solvent was dropped on the top surface, it was reasonable to think that the pre-formed perovskite phase had a trend to distribute near the surface. But this evidence is not conclusive for a bilayer structure because partial preformed perovskite phase might also locate in bulk which could not been fully excluded so far. So, are there further characterizations or explanations on this especially a direct evidence of phase boundary?
(5) How to understand the decreased defect densities in the wrinkle structure compared with the flat morphology? Moreover, why the hill sites show longer carrier lifetimes than valley sites? The authors should give more description on the deep mechanism.
(6) A minor question is that some calculation details can be put into SI, and the description of the connection between perovskite components and wrinkling structures can be more concise so that the readers can easily get the important information.
Reviewer #3 (Remarks to the Author): This work by S.-G. Kim et al reports a systematic study on the wrinkle-like texture formation in a series of halide perovskite thin films. The structure and morphology of the wavy texture was controlled by varying the anti-solvent and substrate temperatures. The formation mechanism was investigated using a series of characterization tools such as in-situ PL and XRD. Using these textured perovskite thin film, solar cells with high efficiency up to 23% was fabricated. Although the formation of the textured surface has been reported by a number of groups previously, this work is more comprehensive and contains some interesting results. However, I feel that there are large amounts of inconsistency and many of the claims in this paper can not be supported by the experimental data. My specific points are listed below. Because of this, I do not recommend publication of this work, at least in the present form.
Specific points for the authors: 1. The authors mentioned epitaxial growth several times, however, there is no epitaxial growth in this work.
2. Regarding the growth mechanism: a. I can not understand where the lattice strain comes from as the growth from the TiO2 layer is not epitaxial. b. Then it is also hard to understand why the "wrinkle" surface can release the strain as the wrinkle is very much macroscopic with a wavelength on the order of micrometers. c. Furthermore, Fig 3a and 3b seems no difference to me. The peaks at low angels are so broad. It is unclear to me how the peak values were selected and how Fig 3c was generated. d. The in-situ PL does not provide constructive information regarding the formation mechanism. e. There is no experimental evidence for the bi-layer model.
3. Regarding Fig 4: It can be seen that in general the lower diethyl ether dripping temperature lead to longer lifetime for the whole film, and the lifetime for the bottom side of the film is better than top, but I do not see the "hill" is better than the "valley". 4. Line 320: the authors claim that the bottom of the film has lower defect density because the crystals growth from the top to bottom. I cannot understand why growth from top to bottom will lead to lower defect density at the bottom. More explanation and clarification are necessary. 5. Line 349: regarding the conducting AFM, the higher photoresponse may be from the larger thickness (more light absorption) at the "hill" sites over the "valley" sites. To me, this data is not convincing enough to prove the hill is better than the valley. 6. Based on the optical simulation, the light absorption is greatly enhanced in the textured film, but there is almost no difference in the device's photocurrent density.
7. Finally, the authors did not provide experimental details of the FA0.92Cs0.08PbBr0.15I2.85 solar cell fabrication in the method part. It is also confusing why the authors study the FA0.875Cs0.125 film in Figure 4 and part of Figure 5 and then switch to FA0.92Cs0.08 for solar cell fabrication.

Response to Reviewers Letters
Manuscript ID: NCOMMS-20-36179 -T Title: Topographical tailoring of photocarriers in perovskite solar cell Authors: Seul-Gi Kim 1 , Jeong-Hyeon Kim 1 , Philipp Ramming 2,3 , Yu Zhong 2,3 , Konstantin Schötz 3 , Seok Joon Kwon 4 Sven Huettner 2 , Fabian Panzer 3 , Nam-Gyu Park 1 * First of all, we thank the reviewers for their valuable comments on our manuscript of MS ID: NCOMMS-20-36179 -T entitled "Topographical tailoring of photocarriers in perovskite solar cell" (corresponding author: Nam-Gyu Park). Here we have addressed the queries from the reviewers and revised the manuscript according to the reviewers' comments. The revised parts were highlighted in green in the revised manuscript. In addition, Dr. Seok Joon Kwon was included as co-author because he did numerical simulation of wrinkled morphology.

Reviewer #1
Many research groups have seen variations in the thickness of their perovskite films and a couple of groups have written specifically on the subject. The topic is important because smooth films make it easier to deposit contact layers that have uniform thickness. Although most people seem to want smooth films, the authors of this manuscript obtained higher efficiency is solar cells with the wrinkling. I really like the new data in this manuscript. They have made films with many different compositions using several different processing conditions. They observed interesting trends that I doubt many people could have predicted. I definitely think this data should be published, but at this time I find the explanations for the observations not to be convincing. If the authors address my comments, then the manuscript might be important enough for Nature Communications.
(Answer) Many thanks for the encouraging comment on our work.
1. I find the statement "no systematic studies were carried out to understand the formation mechanism of a pseudo-epitaxial wrinkle morphology depending on composition and/or preparation condition" to be misleading. Reference 20 did far more than simply report that wrinkling exists. It provided xray diffraction and wafer curvature stress measurements at several stages in the film formation process to show how compressive stress causes the wrinkling. It also showed how changes in the solvent composition could mitigate the wrinkling.
I think that the authors are building on the explanation in Reference 20 by adding that the wrinkling can more easily occur in a perovskite layer that rests on top of a viscous layer.
(Answer) We thank the reviewer for the comment. We agreed with the reviewer's opinion and revised manuscript (MS) to reflect that opinion. MS is revised as "Recently, an approach to control the perovskite morphology has been explored. For example, microscopic wrinkles have been observed for a certain composition of perovskite that suffers buckling of the perovskite thin film [16,17]. In particular, the buckling was explained as a result of local compressive stress relaxation [17]. However, detailed and comprehensive studies for effects of the microscopic wrinkles on the photovoltaic performances as well as wrinkling mechanism have not been reported yet. Here, we report a simple and yet effective experimental approach to control and optimize the microscopic geometry of the wrinkles of perovskite thin films to maximize the photovoltaic performances as well as long-time durability. We also suggest a theoretical model elucidating the wrinkling mechanism based on the detailed experimental data." in introduction part (p.3). And also, "Previous study on the wrinkling of perovskite thin films suggested that the compressive stress developed by the volume change during a fast perovskite formation led to wrinkling by using wafer curvature stress measurements [17]. This mechanism requires relatively long-time wrinkle formation dynamics up to several minutes to hours [19,20]. However, we observe that the wrinkles form within 10 s. Therefore, we have developed a more detailed model by which the overall morphology of wrinkles can be elucidated as well as the wrinkling mechanism based on previous reports. [17,18]. Using a model based on the thin film mechanics, the wrinkle geometry can be described as a function of the thickness and mechanical constants of the materials. We also derive relationships of λ and A of the wrinkles with the compositions and TDE (see detailed analysis in SI)." in p.7.
2. I think the claims that the wrinkling improves light management are very misleading. It is well known that texturing can help trap light in solar cells. In this case, however, there are approximately 60 nm variations in height with periodicities of around 15 microns. The schematics are not to scale and greatly exaggerate the extent of the surface height variation. I wouldn't expect such a small change in the surface angle to help much. The EQE plots and the change in Jsc support my position. At best, the light trapping helps just a tiny bit around 700 nm. Reducing the temperature of the diethyl ether deposition clearly had its impact by increasing Voc, which is unlikely to be a result of light trapping.
(Answer) Actually, as reviewer argued, the solar spectrum absorption does not seem to be significantly improved in the wrinkled perovskite layer as shown in EQE spectrum and the simulated net absorption by rigorous coupled-wave analysis (RCWA) as shown in Fig R1 (not included in the revised SI). There is tiny difference in the band from 750 nm. However, the wrinkle structure is on rear side of device (near back contact). Also, the wrinkle period is large (~13 μm) and the amplitude is also significantly smaller than λ/4. Therefore, it is difficult to say that the wrinkled structure is particularly helpful in light management. We remove FDTD in MS to avoid misleading. As reviewer commented, our experiments show impacts to mainly increasing Voc and FF by reducing TDE. Also, we observed prolonged carrier lifetime as TDE decreased as show in Fig. 4. Therefore, it is reasonable that the prolonged carrier lifetime is the main effect of the wrinkled morphology at lower TDE. In literature, response of EQE spectrum at longer wavelength is significantly related to carrier collection length (LC) which is proportional to carrier diffusion length (LD). [Nakane et al, J. Appl. Phys. 2016, 120, 064505.] We revised MS (on p.11) as "As listed in Table S1 and [23]. This implies that the enhanced photovoltaic performances of the wrinkled morphology is due mainly to the facilitated transport property of photo carriers." 4. At the end of a paper that very thoroughly examines the affect of varying the temperature of the anti-solvent and the primary solvent composition on wrinkling and optoelectronic properties, the authors suddenly say that they raised the efficiency considerably by adding some KI, which has nothing to do with the rest of the paper. They don't actually say how much KI they added. I find that very unsatisfying. The authors must say how much additive they used.
Instead of telling readers in the abstract that they reached 23 % efficiency (with KI), they should say what efficiency they obtained as a function of DE deposition temperature and explain that they get higher efficiency in the wrinkled films. That is the real take home message for this manuscript.
(Answer) We thank the reviewer for the comment. We modified abstract "A power conversion efficiency (PCE) of 21.00% is observed for the sinusoidal winkled morphology formed at TDE = 5 o C, which is higher than that of 19.46% for the flat one at TDE = 30 o C due to the improved voltage and fill factor." We added information about how much KI added in "Device Fabrication" section of the revised supporting information (SI) as "For K-doped  (Answer) We modified MS on p.15. "Moreover, it is also notable that k1 is lower at the bottom than at the top (i.e., k1 = 3.6 × 10 6 s -1 in spot T-a at the top vs. k1 = 2.6 × 10 6 s -1 in spot B-f at the bottom). This indicates the bottom side of the films has the lower defect concentration. This would confirm again that that the crystal grows from the top surface (initially crystallized part with more defects) to the bottom (retarded crystallization in relatively DMSO-rich environment) which can further allow lower defects." Also, we improved the description about the correlation between local strain, defect density and transient PL properties in the manuscript, summarizing the main points of Ref 30 now on p.14-15, reading "The difference in the recombination rate at the hill and valley sites can be attributed to the difference of the local defect densities at the hill and the valley sites. It was known that both tensile as well as compressive strain in halide perovskite thin films led to an increase in the defect density [29]. Areas with higher local strain was reported to result in faster PL decay [30]. Atomistic calculations based on the first-principle models, the defect density was indeed proportional to the degree of the local strain. Indeed, we have observed that k1 decreases with higher amplitude wrinkle morphology formed at lower TDE, which indicates that defects densities at the hill sites decreases with higher amplitude. This can be attributed to the reduced structural defects such as grain boundary defects at the hill sites because local strain is additionally alleviated at structural defects [31]." 7. In Fig 4,   8. I don't think "Topographical tailoring" are good words to use in the title. I was partially expecting the paper to be about topological insulators, which it isn't at all. People won't know what the authors mean by "topographical tailoring." I suggest a title such as "How perovskite composition and antisolvent miscibility affect film wrinkling and optoelectronic properties." (Answer) We changed the title as follows "How antisolvent miscibility affects perovskite film wrinkling and photovoltaic properties".
9. I think that the choice of solvents is so important that it should be stated clearly early in the main text, not just in the supplemental section. The way the manuscript is written, one might think that DMSO is the primary solvent. In fact, it is an additive and DMF is the main solvent.
The authors might also want to clarify that the DMF probably evaporates first, which is probably why they write about the DMSO more.
(Answer) To clearly show that DMF is the main solvent, we modified MS on p.8 and added (Answer) According to the reviewer's comment, we modified Fig. 1 and sets of optical images were moved to the revised SI to clearly show trends of wrinkling depending on TDE and compositions as follows.  11. On page 10 (line 218), should "to form lower TDE" be "to form at lower TDE?" (Answer) We modified the sentence.
12. Why were the substrates at 15 degrees C during spin casting? That is unusual and inconvenient. There must be a good reason.
(Answer) Since temperature near 15 o C was found to be critical in miscibility of DMSO and DE. At temperature below 15 o C, miscibility starts to decrease. Thus, to control the miscibility of DMSO/DE mixture, the critical temperature of 15 o C was applied to the substrate (see Fig.   S4 in the revised SI). In addition, temperature of substrate was varied because the substrate temperature was found to affect viscosity (η) of viscous precipitate (Fig. R2) and evaporation rate of solvents. providing the basis for over 23% PCE and long-term operational stability of PSCs. This is an interesting and deep work on growth kinetics control of perovskite crystals, and the connection between topological arrangement and photocarriers behavior provides useful guidance for more efficient and stable PSCs design. However, there are still some concerns, which need more clarification: (Answer) We appreciate the encouraging comments on our work.
(1) The wrinkling structure only occurs under some specific components (i.e. (FAPbI3)0.875(CsPbBr3)0.125) at given conditions (i.e. TDE is as low as 5 ℃), thus it is hard to popularize this method and the cooling of diethyl ether will increase the production costs.
Actually, there have been some components which may hardly form wrinkling structure (Advanced Materials, 2020, 32, 1907757 (Answer) It is important to tune the various parameters to maximize the PSCs performances while maintaining the number of the factors as small as possible for the practical applications.
As the reviewer stated, the cooling process in manufacturing for mass production can increase the production costs. However, the method in this study does not require such a highperformance cooling process such as liquid nitrogen or huge cooling towers in petrochemical process because range of cooling temperature is moderate (0 ~ 30 o C). Therefore, influence on mass production cost may be marginal. Compared with previous studies mentioned by the reviewer, our method did not require new material and additional time for process. Our method using the miscibility difference by controlling TDE is very simple. Moreover, we proposed a mechanism for forming a wrinkle structure that can encompass the entire composition of perovskite, which is quite unique as compared to previous studies. In the present study, we would like to suggest a new dimension such as morphology control to tune the photovoltaic properties of perovskite materials, which can provide additional aspect of optimizing the PV performances to the community. Actually, we are planning to extend our work to incorporate multi-scale (or hierarchical) wrinkles to address broadband solar spectrum to control further the optical path length and the photo-carrier lifetime. Based on the fundamental study and detailed data shown in the present work, researchers can obtain substantial benefit to control the thin film morphologies to improve the perovskite photovoltaic performances.
( This is another concern that hinders the development of this method. (Answer) We clearly recognize the concern raised by the reviewer. Fortunately, the amplitude of the wrinkle is about 100 nm, and therefore a conformal polymeric HTL layer is expected to form onto the surface of the wrinkled perovskite film when the HTL thickness is comparable to or greater than the amplitude itself. Typically, the thickness of the polymeric HTL is around 100-200 nm, and the surface of the perovskite is favored by the polymeric HTL without concerns on the delamination or rupturing of the HTL. Therefore, the wrinkled surface of perovskite would be safe from the shunt. Also, if other deposition methods for HTL is used such as air knife, it will be freer from generation of shunting paths. [J. ding et al, Joule, 2019, 3, 402-416.] (4) In-situ absorption data showed a perovskite phase with thickness of 250 nm for TDE = 5 ℃, 362 nm for TDE = 15 ℃ and 450 nm for TDE = 30 ℃. Since anti-solvent was dropped on the top surface, it was reasonable to think that the pre-formed perovskite phase had a trend to distribute near the surface. But this evidence is not conclusive for a bilayer structure because partial pre-formed perovskite phase might also locate in bulk which could not been fully excluded so far. So, are there further characterizations or explanations on this especially a direct evidence of phase boundary?
(Answer) We appreciate the comments raised by the reviewer. We conducted additional experiment to confirm the wrinkling mechanism of a bilayer structure and modified MS and SI. We revised MS and added the modified Fig. 2. in p.9-10 as "To further confirm the bilayer model for the wrinkling mechanism, we have numerically simulated the morphological evolution of the thin film wrinkling based on temporal evolution of the wrinkle geometry (see details in SI) [22]. As shown in Figs. S7 and S8, we can find that the bilayer model provides qualitatively similar wrinkling morphologies accompanied by 2D fast Fourier transform (FFT) images to the experimentally observed images. We have also tested again the bilayer model by examining the optical diffraction patterns of the wrinkled thin films (Fig. 2g). As shown in Fig.   2h, the optical diffraction patterns would exhibit different patterns (i.e., concentric ring patterns for the wrinkled bilayer, while dot or single ring pattern for the wrinkled monolayer) with different configurations as denoted in Fig. 2g. Indeed, we observe concentric ring patterns at glass side (bottom) of film just after contacted with diethyl ether (10 s after spin started), and the patterns disappears with time, whereas the transmitted concentric ring patterns was sustained for long time as shown in Fig. 2i. This can be compared to the diffraction patterns of the wrinkled perovskite films obtained from reflected side and transmitted side which are commonly sustained over long time (see Figs. S9a and b). With the theoretical analysis supported by numerical calculations and experimental observations of the diffraction patterns, we can suggest that the wrinkling of the perovskite thin films can be elucidated by a bilayer model." , 1  is Poisson's ratio, and f E is elastic modulus of the elastic capping layer, respectively. The wrinkling results from the relaxation of the in-plane compressive stress denoted as 0  . The origin of the in-plane stress comes from the difference of mechanical responses of the elastic capping film and the underlying viscoelastic substrate.
For example, we can suggest that the difference is due mainly to the thermal expansion coefficient of the two layers [S13]. The absolute value of the compressive stress developed by discrepancy of the thermal expansion coefficients can be expressed as follows.
where   denotes the difference of thermal expansion coefficients of the two layers, S  , S E and H are Poisson's ratio, elastic modulus, and the thickness of the underlying viscoelastic substrate, respectively. The strain developed by the thermal expansion discrepancy is assuredly proportional to the temperature change T  . For most of the elastic-viscoelastic bilayer system, ≪ 1, and therefore, we can simplify eq (S2) as follows. 31 The critical compressive stress corresponding to can be calculated as follows [S12], In eq (S10), we assumed that the thickness of the initially formed elastic layer is not dependent Based on a typical phase diagram of spinodal decomposition, we can find that

Effect of the composition of perovskite materials on the wrinkling
We observed that the substitution of FA with Cs or MA and I with Br resulted in the decrease in λ and the increase in A at a certain substitution ratio. The smaller size of the substituents can increase σ0, which increase λ according to eq (S3) and (S4). Regarding the increased A, η is decreased with increasing the amount of Cs and Br or MA and Br (see Fig. 1b and c). According to ref S12, amplitude (A) is derived function of dimensionless growth rate (s), characteristic time scale (τ) and formation time (t) ( A = A 0 , s = α-μR/Ef and τ = η/Ef). The A is exponentially anti-proportional to η. Therefore, when η is decreased, A is enlarged. Except for the specific ratio, however, the compositions with z ≥ 0.25 or x ≥ 0.8 formed a solid bottom layer, which leads to a very large η (see Fig. S6) and thereby a significant increase of characteristic time scale (τ) to about 10 4~1 0 5 times, resulting in less formation of wrinkled texture.
4. Effect of the annealing condition on the wavelength of the wrinkles Given a condition of σ0 > σc, wrinkling starts with long wavelength (λ0) which will be eventually narrowed and saturated as the stress is being relaxed until σ0 = σc [S15]. However, in perovskite film formation process, λ0 cannot be saturated because the bottom layer is solidified before it is saturated, which may lead to a residual compressed stress after spincoating [S16]. The slight decrement of λ after annealing is evidence of the presence of residual stress because the relaxation of residual stress will further decrease λ as shown in Figs. S2d and S2e.

Null contribution of Ef and vf
Ef and νf can be also assumed to be constant due to a small difference in Ef between 10.2~11.8 GPa for FAPbI3 and 9.7~12.3 GPa for FAPbBr3 even upon replacing iodide with bromide [S17] and small νf of perovskite (0.28~0.33) [S18].

Effect of TSub
At fixed temperature such as TDE = 15 o C, λ increases, while A decreases with increasing TSub from 5℃ to 15℃ (see Fig. 1d). Upon increasing TSub, hf is expected to increase because the miscibility between DMSO and diethyl ether is enhanced by elevating TSub. This can lead to an increase in hf but decreases in A.

Evolution of the wrinkle pattern of the bilayer
To confirm the wrinkling mechanism of a bilayer structure observed in our experiments, we provide a computer simulation of the temporal morphological evolution of the surface wrinkles of the bilayer. For this simulation, we employed a typical finite-difference method for 2D simulation box (800800) with periodic boundary condition. According to the theoretical and numerical scheme suggested by Im and Huang [S12], we modeled the morphological evolution of the bilayer wrinkles as shown in Fig. S7. As shown in Fig. S8, one can find that the simulated wrinkle morphology is similar to the experimentally observed morphology. The similarity is confirmed again by comparing the 2D fast Fourier transform (2D FFT) signals obtained from the simulated and experimentally observed morphologies, in which isotropic wrinkles pattern with notable concentric ring patterns which correspond to the characteristic length scale (i.e., λC) of the wrinkles. The computer simulated bilayer wrinkle morphology strongly supports that the wrinkling mechanism of the perovskite thin film hinges on the relaxation of the in-plain compressive stress developed in the elastic-viscoelastic bilayer.   (5) How to understand the decreased defect densities in the wrinkle structure compared with the flat morphology? Moreover, why the hill sites show longer carrier lifetimes than valley sites? The authors should give more description on the deep mechanism.
(Answer) First of all, the wrinkle structure results from the stress relaxation across the thin film, and therefore, it is more probable to have lower local strain-concentrated geometry which can host local defects like fractures, grain boundary defects compared to the case of a flat thin film.
In the revised MS, we now address this point by adding the following sentences on p. 14 "From the exponential fitting of the PL decay curve, we can deduce the decaying rate k1 such that ( ) ∝ (−2 ⋅ 1 ) (see the SI for details and Table S2). From the exponential fitting, we found that k1 is lower at the hill sites, while higher at the valley sites. Also, it decreases as TDE is lowered (i.e., 3.6 × 10 6 s -1 or 4.4 × 10 6 s -1 for the spots T-a or T-c (hill sites) vs. 5.7 × 10 6 s -1 or 7.4 × 10 6 s -1 for the spots T-b or T-d (valley sites)). The difference in the recombination rate at the hill and valley sites can be attributed by the difference of the local defect densities at the hill and the valley sites. It was reported that both tensile as well as compressive strain in halide perovskite thin films lead to an increase in the defect density [29]. Areas with higher local strain can result in faster PL decay [30]. Atomistic calculations based on the first-principle models, the defect density is indeed proportional to the degree of the local strain. Indeed, k1 is observed to decrease with higher amplitude wrinkle morphology (with lower TDE), which indicates that defects densities at the hill sites decrease with amplitude. This can be attributed to the reduced structural defects such as grain boundary defects at the hill sites because local strain is additionally alleviated at structural defects [30]. The reduced grain boundary defects should be accompanied by the enhanced uniformity of the grain sizes, which can be checked by the narrower distribution of the grain areas (see Figs. S16)." and we added Fig   (6) A minor question is that some calculation details can be put into SI, and the description of the connection between perovskite components and wrinkling structures can be more concise so that the readers can easily get the important information.
(Answer) We moved some calculation details to SI for the readers.

Reviewer #3 (Remarks to the Author):
This work by S.-G. Kim et al reports a systematic study on the wrinkle-like texture formation in a series of halide perovskite thin films. The structure and morphology of the wavy texture was controlled by varying the anti-solvent and substrate temperatures. The formation mechanism was investigated using a series of characterization tools such as in-situ PL and XRD.
Using these textured perovskite thin film, solar cells with high efficiency up to 23% was fabricated. Although the formation of the textured surface has been reported by a number of groups previously, this work is more comprehensive and contains some interesting results.
However, I feel that there are large amounts of inconsistency and many of the claims in this paper can not be supported by the experimental data. My specific points are listed below.
Because of this, I do not recommend publication of this work, at least in the present form.
(Answer) We thank the reviewer for the positive response to our work.
1. The authors mentioned epitaxial growth several times, however, there is no epitaxial growth in this work.
(Answer) We agree with the reviewer's opinion. We removed all expressions of epitaxial growth from MS.
2. Regarding the growth mechanism: a. I can not understand where the lattice strain comes from as the growth from the TiO2 layer is not epitaxial.
b. Then it is also hard to understand why the "wrinkle" surface can release the strain as the wrinkle is very much macroscopic with a wavelength on the order of micrometers.
(Answer) For both questions of (a) and (b), wrinkling (not epitaxial, we eliminated "epitaxial") is not related to the substrate ETL material morphology but related to a bilayer model upon antisolvent treatment in our work. The lattice strain is also related to and strongly dependent on temperature of diethyl ether. From the bilayer mechanism in perovskite film formation, the rapidly generated thick upper elastic layer (i.e. TDE=30 o C) hinders relaxation of compressive stress at surface, leading to a flat surface, while thin hf can release most of compressive stress at surface by forming wrinkled surface before crystallization. Therefore, flat perovskite shows irregular grain shape and many cracked or embedded grains in SEM image. We revised MS on p.14-15 as "From the exponential fitting of the PL decay curve, we can deduce the decaying rate k1 such that ( ) ∝ (−2 ⋅ 1 ) (see the S.I. for details and Table S2). From the exponential fitting, we found that k1 is lower at the hill sites, while higher at the valley sites.
Also, it decreased as TDE lowered (i.e., 3.6 × 10 6 s -1 or 4.4 × 10 6 s -1 for the spots T-a or T-c (hill sites) vs. 5.7 × 10 6 s -1 or 7.4 × 10 6 s -1 for the spots T-b or T-d (valley sites)). The difference in the recombination rate at the hill and valley sites can be attributed by the difference of the local defect densities at the hill and the valley sites. It is known that both tensile as well as compressive strain in halide perovskite thin films lead to an increase in the defect density. [29] Areas with higher local strain can result in faster PL decay [30]. Atomistic calculations based on the first-principle models, the defect density is indeed proportional to the degree of the local strain. Indeed, we observed that k1 decreases with higher amplitude wrinkle morphology (with lower TDE), which indicates that defects densities at the hill sites decreases with higher amplitude. This can be attributed by the reduced structural defects such as grain boundary defects at the hill sites because local strain is additionally alleviated at structural defects. [30] The reduced grain boundary defects should be accompanied by the enhanced uniformity of the grain sizes, which can be checked by the narrower distribution of the grain areas (see Figs. S16)." and we added Fig S16 in SI.