Benchmark performance of low-cost Sb2Se3 photocathodes for unassisted solar overall water splitting

Determining cost-effective semiconductors exhibiting desirable properties for commercial photoelectrochemical water splitting remains a challenge. Herein, we report a Sb2Se3 semiconductor that satisfies most requirements for an ideal high-performance photoelectrode, including a small band gap and favourable cost, optoelectronic properties, processability, and photocorrosion stability. Strong anisotropy, a major issue for Sb2Se3, is resolved by suppressing growth kinetics via close space sublimation to obtain high-quality compact thin films with favourable crystallographic orientation. The Sb2Se3 photocathode exhibits a high photocurrent density of almost 30 mA cm−2 at 0 V against the reversible hydrogen electrode, the highest value so far. We demonstrate unassisted solar overall water splitting by combining the optimised Sb2Se3 photocathode with a BiVO4 photoanode, achieving a solar-to-hydrogen efficiency of 1.5% with stability over 10 h under simulated 1 sun conditions employing a broad range of solar fluxes. Low-cost Sb2Se3 can thus be an attractive breakthrough material for commercial solar fuel production.

source." How was the intensity calibrated?
There are no major concerns regarding the data or how the authors have conducted the experiments (although there are a couple of minor questions outlined below). The work is well done. However, there is little new science brought forth. Sb2Se3 modified with CdS and TiO2 with at Pt catalyst has already been reported as a photocathode (see ACS Nano 2017, 11, 12753). The 1.1% sustained STH efficiency over 10 h is neither first nor best, and remains too low for practical water splitting. Admittedly the anode is the bigger problem, but it's not clear why this paper should be published in Nature Communications when many of the other references in Supplementary Table S1 are in a different tier of journals. This paper provides a few insights into materials engineering and device considerations, but even then, the ideas aren't new. Consequently, this paper is not recommended for publication in Nature Communications.
Minor questions about the data: 1. In the XRD shown in Figure 1, what would be the relative intensities of the (221) and (301) Bragg reflections in a truly randomly oriented sample? Is that what is shown in JCPDS 15-0861? The data for the fast and slow cooling look similar enough that I'm not convinced regarding the statement that a slight change in the XRD data implies that ribbons move during cooling.
cooling Sb 2 Se 3 films are slightly selenium poor (Se/Sb ~ 1.35) as similar with the previously reported CSS-Sb 2 Se 3 thin films (Solar Energy Mater. Solar Cells, 2018, 188, 177). We also have measured the grain size distribution depending on the cooling rate as shown in Figure   R2. The fast cooling sample revealed a slightly larger average value of grain size (~1038 nm) compared with the one of the slow cooling (~850 nm), while both samples had similar standard deviation (~370 nm). In our revised manuscript, we have added the results with relevant descriptions. We thank the reviewer again for the helpful comment. Figure R1. Cooling rate comparison during close space sublimation process between the fast and slow cooling Sb 2 Se 3 . Figure R2. Grain size distribution of Sb 2 Se 3 thin films depending on the cooling rate.

Revision made (colored blue):
(Line 11, Page 7) We denoted the sample prepared with N 2 -assisted cooling as the 'fast-cooling' sample, while the naturally cooled sample was denoted as the 'slow-cooling' sample. The fast cooling sample revealed a slightly larger average value of grain size (~1038 nm) compared with the one of the slow cooling (~850 nm), while both samples had similar standard deviation (~370 nm, Supplementary Fig. 3). The cooling rates for the fast cooling and the slow cooling until it has reached 200 °C are approximately 15.7 °C/min and 11.3 °C/min, respectively. In addition, the energy-dispersive X-ray spectroscopy (EDX) analysis showed that both fast and slow cooling Sb 2 Se 3 films are slightly selenium poor (Se/Sb ~ 1.35) as similar with the previously reported CSS-Sb 2 Se 3 thin films 31 .

Referee #1's Comment 2:
It is surprising that the IPCE improved from 45% to 95% in the short wavelength after fast cooling (Fig. 2d). Is there any optical transmitance for the substrate, Au coated FTO substrate.
How the Au deposited on the FTO and how thick the Au layer? As mentioned by the authors, the optical properties for the fast and slow cooling samples are nearly identical. As the authors proposed, TiO2 layer may contribute to the improved performance, however, the CSS can directly grow dense Sb2Se3 film (may cite Sol. RRL, 2: 1800128.2018 andNano Energy, 49, 346, 2018). It is still unclear why the TiO2 could boost the device performance significantly. In addition, in the KPFM images, the Fig.3c and 3f shown the surface profile.
The KPFM tip may not record the 1 um valley (till the FTO substrate?)

Author's response:
The performance difference between the fast and slow cooling samples is due to the pin-hole formation and the resulting carrier recombination, as demonstrated by KPFM analysis. The enhancement by the fast cooling strategy is NOT related to the Au or TiO 2 layers as both of the fast and slow cooling samples have the same device configuration (FTO/Au/Sb 2 Se 3 /TiO 2 /RuO x , Fig. 2). We deposited 70 nm Au layer by a thermal evaporator.
The role of the Au layer on Sb 2 Se 3 photocathode was well investigated in our previous work (ACS Energy Lett., 2019, 4, 995). Briefly, the Au layer mainly acts as a hole selective contact, which facilitates the transfer of photo-generated holes. Of course, the 70 nm Au layer can reflect the incident light to some extent, but the more important role is selective hole transfer as proven by the fact that a transparent hole selective layer (Cu:NiO) is more efficient than reflective Au layer, indicating that the selective hole transport ability is more important than light reflection (ACS Energy Lett., 2019, 4, 995). The Au layer as a hole selective contact was also discussed in Sb 2 Se 3 solar cells (Solar Energy, 2019, 182, 96). As the reviewer pointed out, Sb 2 Se 3 can be directly grown on FTO substrate without Au or TiO 2 . We have taken some SEM images and measured the PEC performance without Au layer (please refer to the response to the comment #3 below). Regarding the role of TiO 2 , it is well known that TiO 2 can act as a protective layer as well as an n-type semiconductor on a p-type semiconductor layer to form a p-n junction (ACS Energy Lett., 2016, 1, 1127J. Mater. Chem. A, 2017, 5, 2180ACS Energy Lett. 2019, 4, 517). Additionally, the AFM and KPFM techniques are widely used to determine surface topography and surface potential distribution. The height recorded by AFM and KPFM ranged from a few nm to hundreds nm and even sometimes μm scale (Sol. Energy Mater. Sol. Cells, 2018, 183, 34;Nano Converg., 2014, 1, 27). In our data (Fig 3), it is also evident that a distinct potential peak was observed when there was a rapid drop of the height. In our revised manuscript, we have added some relevant references for explaining the capability of KPFM to determine a potential peak at a deep valley as follows.

Revision made (colored blue):
(Page 13, Line 7) In contrast, in the slow-cooling sample, the surface potential increased significantly with a rapid drop in the topography (Fig. 3f), indicating direct contact between the n-type TiO 2 layer and substrate due to pin-holes. It might be worth to note that the height recorded by AFM and KPFM ranged from a few nm to hundreds nm and even sometimes μm scale 38-39 . In such a case, the photo-excited electrons can be extracted laterally to the ribbons and they can recombine with the holes at the back contact as shown in Fig. 3h  Again, the role of the Au layer on the FTO on the PEC performance is unclear, it should compare the sample with and without Au layer. Here, the Sb2Se3 grown on the FTO.

Author's response:
As we mentioned in the above comments, the role of the Au layer as a hole selective contact was elucidated in our previous study (ACS Energy Lett., 2019, 4, 995), as well as in Sb 2 Se 3 solar cell research (Solar Energy, 2019, 182, 96). Here, as per the reviewer's comment, we have compared the PEC performance and microstructures of Sb 2 Se 3 photocathodes depending on the presence/absence of the Au layer. As shown in Figure R3, there was no noticeable microstructural difference between Sb 2 Se 3 thin films with and without the Au layer. On the other hand, PEC performance of Sb 2 Se 3 photocathodes significantly increased with the Au layer as shown in Figure R4. The results are in accordance with the previous studies, which showed much enhanced performance with the Au hole selective contact (ACS Energy Lett., 2019, 4, 995;ACS Nano, 2018, 12, 11088). In our revised manuscript, we have added the results of without Au layer sample with relevant descriptions and references. We thank the reviewer for the helpful comment improving the quality of our manuscript. Figure R3. SEM images of Sb 2 Se 3 (a-b) without and (c-d) with the Au bottom contact layer. Figure R4. PEC performance of (a) RuO x /TiO 2 /Sb 2 Se 3 /FTO and (b) RuO x /TiO 2 /Sb 2 Se 3 /Au/ FTO photocathodes.

Revision made (colored blue):
(Page 10, Line 2) Fig. 2a-b show the PEC performance of RuO x /TiO 2 /Sb 2 Se 3 /Au/FTO photocathodes using fast-and slow-cooling Sb 2 Se 3 films measured in pH 1 electrolytes. As we mentioned above, the Au layer acts as a hole selective contact, which facilitates the transfer of photo-generated holes while blocking the electrons backflow 32-33 . Without the Au layer, Sb 2 Se 3 photocathodes revealed relatively poor performance while nearly similar morphology of Sb 2 Se 3 was observed ( Supplementary Fig. 4), which verifies the role of the Au layer not affecting the growth of Sb 2 Se 3 , but assisting the transfer of photo-generated charges. The RuO x catalytic layer was deposited by the PEC method, while atomic layer deposition (ALD) was used for the TiO 2 layer, similar to a previous study 32 . In both samples, the onset potentials shifted towards a positive direction after the first scan due to activation of the RuO x catalyst 34 .

Supplementary Fig. 4 | Sb 2 Se 3 photocathodes without the Au bottom contact layer. a-b,
SEM images of fast-cooling Sb 2 Se 3 on FTO substrate. The morphology of Sb 2 Se 3 directly grown on the FTO substrate is nearly similar to that of Sb 2 Se 3 grown on Au/FTO. c, J-V curves of RuO x /TiO 2 /Sb 2 Se 3 /FTO photocathodes in pH 1 H 2 SO 4 electrolytes.

Referee #1's Comment 4:
How the cells chemical composition post reliability test?
Author's response: We have measured Raman spectroscopy to investigate the chemical composition difference after the reliability test as per the reviewer's comment. Before the stability test, the Raman spectra of Sb 2 Se 3 shows one distinct peak at ≈ 190 cm −1 along with a shoulder peak at ≈ 208 cm −1 , both of which are attributed to the vibration modes in Sb 2 Se 3 phase ( Figure R5a). After the stability test, an additional peak located at ≈ 250 cm −1 appeared. The additional peak indicates the formation of by-products such as Sb 2 O 3 (≈ 254 cm −1 ) and/or several Se phases (e.g., monoclinic Se 8 rings at ≈ 253 cm −1 , rhombohedral Se 6 rings at ≈ 247 cm −1 , and amorphous Se at ≈ 250 cm −1 ), which result from the decomposition of Sb 2 Se 3 . In addition, there was also morphological destruction after the stability test ( Figure R5b-c). According to our previous study on the stability of Sb 2 Se 3 photocathodes (Adv. Energy Mater., 2019, 9, 1900179), the morphological destruction of Sb 2 Se 3 photocathode is caused by the photoreduction of TiO 2 accompanied by the degradation of Sb 2 Se 3 . In the revised manuscript, we have added the chemical composition results with relevant descriptions. We thank the reviewer for the helpful comment. Figure R5. Chemical composition and microstructures of Sb 2 Se 3 before and after stability test.
(a) Raman spectra of Sb 2 Se 3 photocathodes and SEM images of (b) before and (c) after stability test.

Revision made (colored blue):
(Page 15, Line 9) The RuO x /TiO 2 /Sb 2 Se 3 sample retained approximately 60% of initial photocurrent density after 35 hours in the neutral electrolytes, which is the best stability of Sb 2 Se 3 photocathodes reported so far ( Supplementary Fig. 9a). The photocurrent density of Pt/TiO 2 /Sb 2 Se 3 decreased more rapidly (42% photocurrent after 5 h), probably due to the larger bubbles at Pt surfaces, as evidenced by severe fluctuations in the enlarged photocurrent curves (Supplementary Fig. 9b-c). We measured Raman spectroscopy to investigate the chemical composition variation after the reliability test. Before the stability test, the Raman spectra of the RuO x /TiO 2 /Sb 2 Se 3 /Au/FTO photocathode showed one distinct peak at ≈ 190 cm −1 along with a shoulder peak at ≈ 208 cm −1 , both of which are attributed to the vibration modes in Sb 2 Se 3 phase ( Supplementary Fig. 8a). After the stability test, an additional peak located at ≈ 250 cm −1 appeared. The additional peak indicates the formation of by-products such as Sb 2 O 3 (≈ 254 cm −1 ) and/or several Se phases (e.g., Se 8 rings at ≈ 253 cm −1 , Se 6 rings at ≈ 247 cm −1 , and amorphous Se at ≈ 250 cm −1 ) as a result from the decomposition of Sb 2 Se 3 . In addition, there was also morphological destruction after the stability test ( Supplementary Fig. 8b-c).
According to our previous study on the stability of Sb 2 Se 3 photocathodes 25 , the morphological destruction of Sb 2 Se 3 photocathode is caused by the photo-reduction of TiO 2 accompanied by the degradation of Sb 2 Se 3 .

Supplementary Fig. 8 | Chemical composition and microstructures of Sb 2 Se 3 before and
after stability test. a, Raman spectra of Sb 2 Se 3 photocathodes and SEM images of b, before and c, after stability test.

Comments to the Author
The manuscript by Yang et al. describes a new synthesis method for Sb2Se3 photocathodes-a highly promising material for practical water splitting-and the development of a tandem cell with BiVO4. The performance of the Sb2Se3 is outstanding, the article is well written and appropriately cited, and the level of scientific discussion is high. I recommend to accept after addressing the following minor concerns:

Author's response:
We thank the reviewer for evaluating our work as a well-written and meaningful work. All of the comments made by the reviewer are helpful for improving the overall quality of our work.
Our detailed, point-by-point responses to the reviewer's comments can be found below.

Referee #2's Comment 1:
The title should be more descriptive of the contents of the article.

Author's response:
We have modified the title as per the reviewer's comment as follows.

Revision made (colored blue):
Benchmark performance of low-cost Sb 2 Se 3 photocathodes obtained by the fast-cooling strategy during close space sublimation for unassisted solar overall water splitting

Referee #2's Comment 2:
Page 15: "The photocurrent density of Pt/TiO2/Sb2Se3 decreased more rapidly (42% photocurrent after 5 h), probably due to the larger bubbles at Pt surfaces, as evidenced by severe fluctuations in the enlarged photocurrent curves ." what do the authors mean, that the releasing of large bubbles cause the Pt to detach?
Author's response: The detachment of Pt particles due to the releasing of large bubbles is one of the well-known degradation mechanisms in the Pt-decorated photocathodes for water splitting. It is widely reported in the literature (please refer to the section 4.3.2 in Chem. Soc. Rev., 2019, 48, 4979).
In the revised manuscript, we have added relevant descriptions and references on the detachment of Pt for better understanding.

Revision made (colored blue):
(Page 15, Line 12) The photocurrent density of Pt/TiO 2 /Sb 2 Se 3 decreased more rapidly (42% photocurrent after 5 h), probably due to the larger bubbles at Pt surfaces, as evidenced by severe fluctuations in the enlarged photocurrent curves (Supplementary Fig. 7b-c). It should be noted that the detachment of Pt particles due to the releasing of large bubbles is one of the well-known degradation mechanisms in the Pt-decorated photocathodes for water splitting 42 .

(Reference) (Page 20, Line 22)
However, the photocurrent density of Sb 2 Se 3 photocathodes at 0.4 V RHE with and without BiVO 4 were 5.0 mA cm −2 and 2.2 mA cm −2 , respectively, which are higher than the photocurrents shown by the J-V curve.

Referee #2's Comment 5:
Fig 5 caption: "All analyses were conducted in 0.5 M phosphate buffer + 0.01 M V2O5 (pH 7.0)" -were faradaic efficiency measurements carried out to ensure that the 20 mM vanadium ions do not interfere with the hydrogen production? Since V5+ is much more easily reduced than proton.

Author's response:
We thank the reviewer's comment. As the reviewer pointed out, theoretically the V 5+ can be reduced prior to proton, possibly affecting the performance of our Sb 2 Se 3 photocathode-based tandem devices for water splitting. However, some previous studies showed that the presence of V 5+ in a potassium borate buffer solution doesn't produce any additional reduction or oxidation peaks and doesn't interfere with water reduction and water oxidation (Nature Energy, 2018, 3, 53). In order to verify whether V 5+ participates in the redox reactions or not, we have performed additional experiments as follows. As shown in Figure R5a, there are distinctive peaks in the LSV scans for a Pt electrode upon addition of V 5+ into strongly acidic electrolyte, indicative of a significant reduction of V 5+ . In contrast, there is no noticeable difference between with/without V 5+ electrolyte when measured in a neutral electrolyte (0.5 M KPi). These results imply that the reactivity of V 5+ , which is relatively stronger in an acidic electrolyte, significantly decreases in a neutral electrolyte. As we measured our Sb 2 Se 3 -based tandem device in a neutral electrolyte (0.5 M KPi), there is no significant change of both the Sb 2 Se 3 photocathode and the BiVO 4 photoanode upon adding V 5+ into our electrolyte as shown in Figure R6. It should be noted that the slight difference observed in the photocathode case ( Figure R6d), possibly due to parasitic light absorption by yellow V 5+ ions, does not affect the performance of our tandem device as the operation potential of the tandem device is around 0.4 V RHE . Accordingly, in any cases, it is reasonable to conclude that addition of V 5+ does not interfere with the hydrogen production by our Sb 2 Se 3 -based tandem device.
We have modified our manuscript with the results and descriptions on the reduction of V 5+ .
We appreciate the reviewer for improving the quality of our work.

Revision made (colored blue):
(Page 19, Line 6) However, owing to the low stability of BiVO 4 in phosphate, fast degradation of performance was observed for the tandem cell, and we addressed the stability issue by adding vanadium cation (V 5+ ) as done by Choi group 45 . It should be noted that theoretically the V 5+ can be reduced prior to proton, possibly affecting the performance of our Sb 2 Se 3 photocathode-based tandem devices for water splitting. As shown in Supplementary Fig. 16a, there are distinctive peaks in the LSV scans for a Pt electrode upon addition of V 5+ into strongly acidic electrolyte, indicative of a significant reduction of V 5+ . In contrast, there is no noticeable difference between with/without V 5+ electrolyte when measured in a neutral electrolyte (0.5 M KPi, Supplementary Fig. 16b). These results imply that the reactivity of V 5+ , which is relatively stronger in an acidic electrolyte, significantly decreases in a neutral electrolyte. As we measured our Sb 2 Se 3 -based tandem device in a neutral electrolyte (0.5 M KPi), there is no significant change of both the Sb 2 Se 3 photocathode and the BiVO 4 photoanode upon adding V 5+ into our electrolyte as shown in Supplementary Fig. 17. It is also noteworthy that the slight difference observed in the photocathode case ( Supplementary Fig. 17d), possibly due to parasitic light absorption by yellow V 5+ ions, does not affect the performance of our tandem device as the operation potential of the tandem device is around 0.4 V RHE . Accordingly, in any cases, it is reasonable to conclude that addition of V 5+ does not interfere with the hydrogen production by our Sb 2 Se 3 -based tandem device.

Referee #2's Comment 6:
Page 29: "and simulated solar light illumination (AM 1.5G, Newport Corporation) was used as the light source." How was the intensity calibrated?
Author's response: Calibration of the 1-sun level was performed using a standard Si reference cell certified by the Newport Corporation, consisting of a readout device and a 2 × 2 cm 2 calibrated solar cell made of monocrystalline silicon. During calibration, the Si reference cell was located at the same position of the sample for PEC measurement. We have added the calibration process in the revised manuscript.

Revision made (colored blue):
(Methods) The Sb 2 Se 3 photocathodes were submerged in an acidic (H 2 SO 4 , pH ~ 1) or a neutral (phosphate buffer, pH ~ 6.25) electrolyte, and simulated solar light illumination. Calibration of the 1-sun level was performed using a standard Si reference cell certified by the Newport Corporation, consisting of a readout device and a 2 × 2 cm 2 calibrated solar cell made of monocrystalline silicon. During calibration, the Si reference cell was located at the same position of the sample for PEC measurement.

Referee #3
Comments to the Author  Supplementary Table S1 are in a different tier of journals. This paper provides a few insights into materials engineering and device considerations, but even then, the ideas aren't new. Consequently, this paper is not recommended for publication in Nature Communications.

Author's Response:
We thank the reviewer for stating that "There are no major concerns regarding the data or how the authors have conducted the experiments" and evaluating our manuscript as a welldone work. As the reviewer pointed out, there are many nice papers pertaining to photocathode materials in a different tier of journals (Supplementary Table S1). However, we would like to emphasize that our Sb 2 Se 3 -based PEC device has sufficient novelty to be published in Nature Communication in terms of not only the performance and but also the material's novelty. We are well aware that the development of cost-effective materials still remains a paramount challenge for the commercialization of PEC water splitting, despite the tremendous effort has devoted by researchers over decades. Sb 2 Se 3 is one of the attractive emerging materials for PEC water splitting in terms of cost, band gap, optoelectronic properties, photocorrosion stability, and processability (please refer to the recent highlight paper "Rapid advances in antimony triselenide photocathodes for solar hydrogen generation", J. Mater. Chem. A, 2019, 7, 20467). Thus, establishing a new benchmark performance of this emerging photocathode could be much more important than other high performance devices based on well-investigated materials. For example, we modified Table S1 to emphasize the novelty of our system compared with other photocathode materials (for review purpose only).
Additionally, we also report a novel strategy by using close-space sublimation, which is a relative scalable method. Our fast cooling strategy enabling smooth and pin-hole-free Sb 2 Se 3 thin films provides meaningful insight into how thermodynamically metastable morphology can be achieved. We believe that our finding contributes to other materials systems as well as the development of materials science. Of course, the performance and stability of our Sb 2 Se 3 based photocathodes should be further improved for practical water splitting. However, we believe that reporting the highest efficiency of the novel and emerging semiconductor can sufficiently provide the feasibility to be commercialized in the future as well as it is worth to receive attention from researchers worldwide. Therefore, we politely ask the reviewer to reconsider recommending our manuscript which has been revised based on the reviewer's comment below.

Referee #3's Comment 1:
In the XRD shown in Figure 1, what would be the relative intensities of the (221) and (301) Bragg reflections in a truly randomly oriented sample? Is that what is shown in JCPDS 15-0861? The data for the fast and slow cooling look similar enough that I'm not convinced regarding the statement that a slight change in the XRD data implies that ribbons move during cooling.

Author's response:
To quantify the relative intensities of each plane revealed in XRD data to a standard Sb 2 Se 3 powder (JCPDS 15-8601), we have calculated the texture coefficient T c , which is defined as

T c hkl =n I(hkl)/I o (hkl) ∑ I(hkl)/I o (hkl) n 1
where I(hkl) is the measured relative intensity of the peak corresponding to the hkl diffraction, I o (hkl) is the relative intensity from a standard powder sample (JCPDS 15-0861), and n is the total number of diffraction peaks used in the evaluation. A large T c value for a specific diffraction peak indicates preferred orientation along this direction. In the present case, we chose four diffraction peaks (n = 4) corresponding to 2θ values of 120, 211, 221, and 301. Figure R7 clearly shows that T c (120) of both fast and slow cooling samples is nearly zero while the other values are higher than 1, indicating both samples have (hk1) preferred orientation. Although both samples have a similar preferred orientation, it is also obvious that the fast cooling sample revealed higher Tc values of (211) and (301) planes and lower T c value of (221) plane, implying possible rearrangement of the ribbons. In our revised manuscript, we have added the quantitative analysis of the ribbon orientations based on the T c values with relevant descriptions. We thank the reviewer for improving the quality of our work. Figure R7. The texture coefficients of selected diffraction peaks in different Sb 2 Se 3 thin films.

(Page 7, Line 15)
As found in previous studies on Sb 2 Se 3 thin-film solar cells, (hk1) orientations, representing (Sb 4 Se 6 ) n nanoribbons oriented perpendicular or inclined relative to the substrate (Fig. 1h where I(hkl) is the measured relative intensity of the peak corresponding to the hkl diffraction, I o (hkl) is the relative intensity from a standard powder sample (JCPDS 15-0861), and n is the total number of diffraction peaks used in the evaluation. A large T c value for a specific diffraction peak indicates preferred orientation along this direction. In the present case, we chose four diffraction peaks (n = 4) corresponding to 2θ values of 120, 211, 221, and 301.
Supplementary Fig. 4 clearly shows that T c (120) of both fast and slow cooling samples is nearly zero while the other values are higher than 1, indicating both samples have (hk1) preferred orientation. Although both samples have a similar preferred orientation, it is also obvious that the fast cooling sample revealed higher T c values of (211) and (301)

Referee #3's Comment 2:
In Figure 2a-b, the origin of the shift to more positive potentials after the 1st sweep is described as RuOx activation? Is the surface the same after this activation? What do the SEM images look like after HER?
Author's response: As the reviewer mentioned, it is correct that the origin of the shift to more positive potentials after the 1st sweep is described as RuO x activation. As shown in Figure R8a-b, it seems that the particle size of RuO x slightly decreases upon activation, the degree of the change is not significant before and after activation. Additionally, there was some morphological destruction after the stability test ( Figure R8b-c). According to our previous study on the stability of Sb 2 Se 3 photocathodes (Adv. Energy Mater., 2019, 9, 1900179), the morphological destruction of Sb 2 Se 3 photocathode is caused by the photo-reduction of TiO 2 accompanied by the degradation of Sb 2 Se 3 . In the revised manuscript, we have added some descriptions on the changes after the stability test. Figure R8. SEM images of RuO x /TiO 2 /Sb 2 Se 3 /Au/FTO photocathodes (a) before and (b) after the activation of the RuO x layer.

Revision made (colored blue):
(Page 15, Line 9) In addition, there was also morphological destruction after the stability test ( Supplementary   Fig. 8b-c). According to our previous study on the stability of Sb 2 Se 3 photocathodes 25 , the morphological destruction of Sb 2 Se 3 photocathode is caused by the photo-reduction of TiO 2 accompanied by the degradation of Sb 2 Se 3 .
( Supplementary Information) a b Supplementary Fig. 5 | Chemical composition and microstructures of Sb 2 Se 3 before and after stability test. a. Raman spectra of Sb 2 Se 3 photocathodes and SEM images of b, before and c, after stability test.

Referee #3's Comment 3:
Can the authors avoid mixing thermodynamic and kinetic language? Current is cathodic (or anodic), but there is no such thing as cathodic potential. So, rather than scanning in the "cathodic direction," the authors should state that J-V curves were scanned from open circuit to more negative potential.

Author's response:
In the revised manuscript, we have modified the confusing descriptions as per the reviewer's comment. We thank the reviewer for clarifying our work.

Referee #3's Comment 4:
Can the authors provide data for multiple films? It is not clear if the data are representative of typical films or the best-performing films.

Author's response:
We thank the reviewer for the careful comment. Normally we presented the best-performing films. For example, for RuO x /TiO 2 /Sb 2 Se 3 /Au/FTO photocathodes, which we used as a benchmark photocurrent density, 30 mA cm −2 is the highest value, while normally 25 -30 mA cm −2 photocurrent is observed as shown in Figure R9. In addition, for the tandem devices, For tandem cell, photocurrent density near 0.4 V RHE of each photoanode and photocathode determines the overall STH efficiency. As shown in Figure R10, photocurrent density of BiVO 4 at 0.4 V RHE varies from 1.2 to 0.8 mA cm −2 , thus the overall STH efficiencies of the tandem cell range from 1.48 % to 0.98 %. We have modified the descriptions regarding benchmark performance in our revised manuscript.

Revision made (colored blue):
(Page 10, Line 6) The photocurrent density of the fast-cooling sample approached 30 mA cm −2 at 0 V RHE , which is not only the highest value obtained for a Sb 2 Se 3 photocathode but also among the best observed for all photoelectrodes used in PEC water splitting so far. Note that the data shown in Fig.2 were obtained from the best performing device, while normally 25 -30 cm −2 at 0 V RHE photocurrent density was observed in the fast-cooling Sb 2 Se 3 based photocathodes.

(Page 19, Line 9)
As shown in Fig. 5b, the operating point of the two photoelectrodes, as estimated by the intersection of two J-V curves, was approximately 1.2 mA cm −2 at 0.4 V RHE , which corresponded to a STH efficiency of 1.5%. It should be noted that photocurrent density of BiVO 4 at 0.4 V RHE varies from 1.2 to 0.8 mA cm −2 , thus the overall STH efficiencies of the tandem cell range from 1.48 % to 0.98 %.

Editorial comments
In an effort to ensure reproducibility of research data, we now also require that you provide a separate source data file. The source data file should, as a minimum, contain the raw data such as the antibody used. We also encourage you to include any other types of raw data that may be appropriate. An example source data file is available demonstrating the correct format: https://www.nature.com/documents/ncomms-example-source-data.xlsx The file should be labelled 'Source Data', with the title and a brief description included in your cover letter, and should be mentioned in all relevant figure legends using the template text below: "Source data are provided as a Source Data file."

Response:
As per the editor's comment, we have provided the source data in excel format and mentioned in all figure legends.

Response Letter
Title: "Benchmark performance of low-cost Sb 2 Se 3 photocathodes obtained by the fastcooling strategy during close space sublimation for unassisted solar overall water splitting"

Reviewer #1
General comment: The authors replied well to the reviewers' comments. As shown in the comments of Reviewer #3 and the authors' response, the work seems promising and demonstrates a benchmark photocurrent performance for Sb2Se3 photocathodes deposited using the CSS system.

General reply:
We thank the reviewer for mentioning that our work seems promising and demonstrates a benchmark performance for Sb 2 Se 3 photocathodes, which is an emerging low-cost material.
While we understand the reviewer's concern in some points, however, we respectfully disagree with the reviewer for the other points. Detailed point-by-point response can be found below.

Comment #1:
The reviewer agrees with the authors that Sb2Se3 is an emerging candidate for the photoelectrochemical application, but the authors reported lots of previous similar work, such as ACS Energy Lett.,2019, 4, 995;ACS Nano, 2018, 12, 11088, with similar device structure.
Here, the CSS deposition of the Sb2Se3 absorber layer for photovoltaics application is not a novelty (Solar Energy Mater. Solar Cells, 2018, 188, 177, Sol. RRL, 2: 1800128.2018and Nano Energy, 49, 346, 2018 although the authors claim the cooling rate could significantly impact on the photoelectrochemical performance.

Response #1:
Regarding the similarity to the previous works, using a similar device structure (ex,  (Nat. Cat., 2018, 1, 412). Although all of the synthetic methods and device structures of both the photoanode and the photocathode had been already known, their tandem device result was worth to be published in the prestigious journal and has been cited more than 90 times up to the present.
Moreover, our current work achieved lots of scientific advances in Sb 2 Se 3 photocathodes in the following points of views; 1) The first demonstration of the cooling rate effect on the morphology of Sb 2 Se 3 2) Insight toward synthesis meta-stable morphology during CSS deposition 3) Demonstration of the importance of pinhole-free films for high performance 4) The highest level of photocurrent among not only Sb 2 Se 3 but also all photoelectrodes for PEC water splitting 5) The first unassisted water splitting by Sb 2 Se 3 -based photocathodes 6) A promising level of STH efficiency among photoanode-cathode tandem cells 7) The first demonstration of the role of V 5+ ions in enhancing the stability of the PEC tandem devices (so far only tested in a half cell reaction) 8) Demonstration of the potential of Sb 2 Se 3 , as a low-cost p-type semiconductor which is one of the most important but less investigated field We are not insisting that our result is the most impactful research in this decade. But we still believe that our result will significantly contribute to the PEC research field and is also worth to be published in Nature Communications.

Comment #2:
However, the fundamental mechanism for improved PEC performance is not clear, particularly, the chemical composition, microstructure did not show significant differences.

Response #2:
We should note that it is common sense in the PEC research field that direct contact between the top and bottom contact can cause significant degradation of the performance, even in the case of the chemical composition and optoelectronic properties have negligible differences.
For example, Luo et al. reported the effect of a thin blocking layer to prevent shunt pathways in Cu 2 O nanowire photocathodes (Nano Lett. 2016, 16, 1848. As shown in Fig. R1, the PEC performance of Cu 2 O photocathodes significantly increased upon deposition of a very thin blocking Cu 2 O layer. There were no significant differences in chemical composition, microstructure, and optoelectronic properties, thereby demonstrating the detrimental effect of shunt pathways. Although the reviewer pointed out the fundamental mechanism for improved PEC performance is not clear, we strongly believe that we have rather proven the performance degradation mechanism in the presence of pinholes by showing rapid potential drops near pinholes by KPFM analysis, which had not yet been experimentally demonstrated.
That is why the Reviewer #3 mentioned "Generating pinhole-free Sb 2 Se 3 is crucial for longterm overall solar water splitting at zero bias." In our revised manuscript, we have modified our manuscript to clarify the importance of preventing pin-holes formation in order to avoid any confusion of readers.

Revision made (colored blue):
(Line 18, Page 13) In such a case, the photo-excited electrons can be extracted laterally to the ribbons and they can recombine with the holes at the back contact as shown in Fig. 3h

Comment #3:
Particularly, the mechanism for the Au layer increase almost 3 times of the photocurrent ( Fig.R4) has no detailed analysis, but only cite the previous work.

Response #3:
Regarding to the use of the Au layer, there are some comprehensive studies on hole-selective materials for Sb 2 Se 3 -based photocathodes and solar cells (ACS Energy Lett., 2019, 4, 995;Solar Energy, 2019, 182, 96). In these papers, the authors have already reported the role of hole selective layers by comparing different types of hole selective layers as well as the performance enhancement mechanism while measuring the band positions and the resistivity, which had been cited in the present study. We believe that it is unnecessary to repeat all observations, which will make the present work deviated from the main point. Thus, it is absolutely valid to cite the previous well-known phenomena in our manuscript instead of repeating the details.

Comment #4:
And the low overpotential of the Sb2Se3, i.e., 0.4V still lower than other systems has not been investigated.

Response #4:
Here we respectfully disagree the reviewer's comment for the following reasons. First, our Sb 2 Se 3 photocathode revealed the highest level of photocurrent and photovoltage among the previously reported Sb 2 Se 3 photocathodes as shown in Fig. R2a. It should be also noted that the maximum photovoltage obtained by a semiconductor is in relation to its band gap. The Sb 2 Se 3 photocathode in the present study is located at the black dashed line (the green star) which represents SQ−0.4 V. It is obvious that our Sb 2 Se 3 photocathode achieved a higher figure of merit than most of the previous photoanodes and photocathodes, and actually revealed close to single crystalline photocathodes such as Si and InP. Thus, we strongly believe that our Sb 2 Se 3 photocathodes revealed meaningfully high onset potential not only among previous Sb 2 Se 3 photocathodes but also all other photoelectrode materials considering its small band gap. Figure R2. (

Comment #5:
By considering the Pt or RuOx cocatalyst and the 70 nm Au hole collection layer are all precious materials in the RuOx/TiO2/Sb2Se3/Au/FTO and Pt/TiO2/CdS/Sb2Se/Au/FTO structure will significantly limit the scale application and may weaken the claimed "low cost" Sb2Se3 but an expensive device.

Response #5:
This comment is partially correct, but it is widely well-known that the most important building block in a PEC water splitting device in terms of cost-effectiveness is the light-absorbing semiconductors. For example, the cost portion of electrocatalysts in the c-Si based PV-EC system is just approximately 1 % (whether the electrocatalyst is Ir-Ru or NiFe-NiMo) for overall cost of hydrogen production (Energy Environ. Sci., 2014, 7, 3828). Additionally, in the high STH systems based on III-V semiconductors, the cost of InGaP/GaAs ($175) per unit solar collection area is much greater than other parts such as catalysts (Pt and IrO x , $8) and membranes (127 mm-thick Nafion, $5) (Energy Environ. Sci., 2016, 9, 2354. Moreover, the community is recognizing that noble metals are easy to recycle, as well as price of hydrogen expected from non-noble metal electrocatalysts is not competitive to the one with noble metal ones (Nat. Energy, 2019, 4, 430). It is also noteworthy that there is an alternative low-cost materials (Cu:NiO) for hole selective contact of Sb 2 Se 3 photocathodes (ACS Energy Lett., 2019, 4, 995). That is also an important merit of Sb 2 Se 3 photocathode for further reducing the overall cost, considering the fact that single crystalline semiconductors (such as GaAs and InP) cannot be grown on cost-effective substrates. Aforementioned descriptions are generally accepted in the PEC water splitting field, which is well supported by John Turner (who had been working in this field over 30 years) in his commentary article (Science, 2014, 344, 469) in which the importance of the development of cost-effective semiconductors having good optoelectronic properties is highly emphasized. In the revised manuscript, we have added some descriptions to emphasize the importance of semiconducting materials in terms of cost-effectiveness of PEC water splitting devices.

Revision made (colored in blue):
(Line 11, Page 11) Given the low-cost and relatively short history of Sb 2 Se 3 as well as the simple preparation and low material usage due to the high α, the high photocurrent density of ~30 mA cm −2 at 0 V RHE clearly demonstrates the strong potential of Sb 2 Se 3 as a promising photocathode material. It is worth emphasizing that the most important building block in a PEC water splitting device in terms of cost-effectiveness is the light-absorbing semiconductors. For example, the cost portion of electrocatalysts in the c-Si based PV-EC system is just approximately 1 % (whether the electrocatalyst is Ir-Ru or NiFe-NiMo) for overall cost of hydrogen production 37 . Additionally, in the high STH systems based on III-V semiconductors, the cost of InGaP/GaAs ($175) per unit solar collection area is much greater than other parts such as catalysts (Pt and IrO x , $8) and membranes (127 mm-thick Nafion, $5) 38 . Moreover, the community is recognizing that noble metals are easy to recycle, as well as price of hydrogen expected from non-noble metal electrocatalysts is not competitive to the one with noble metal ones 39 . Thus, despite the use of relatively expensive catalysts and a hole selective contact layer, the high performance of our device demonstrates the feasibility of cost-effective Sb 2 Se 3 based photoelectrodes for PEC water splitting.

Comment #6:
By forming a tandem PEC cells with BiVO4, the efficiency is not promising for 10 h stability.
Overall, it does not show clearly innovative evidence and understanding to publish in Nature Communications when more similar results are published in other journals. Thus, I recommend the manuscript to be rejected and may submit to elsewhere.

Response #6:
As we presented in Fig 6. in the manuscript, 10 h stability is the 2 nd longest record for PEC tandem devices (the first one is BiVO 4 -Cu 2 O tandem cells revealing 12 h stability). Our Sb 2 Se 3 photocathode, operating stably over 30 h, is also the most stable Sb 2 Se 3 photocathodes.
Additionally, as we mentioned above, our work is the first demonstration of the role of V 5+ ions in enhancing the stability of the PEC tandem devices, which is another novelty regarding the stability of our device. We are not insisting that our work represents a matured technology ready to commercialize, but we still believe that our work has sufficient novelty and significance to be published in Nature Communications.

Reviewer #2
Remarks to the Author: The authors have carefully considered and thoroughly and satisfactorily responded to all of