Stabilizing ultrasmall Au clusters for enhanced photoredox catalysis

Recently, loading ligand-protected gold (Au) clusters as visible light photosensitizers onto various supports for photoredox catalysis has attracted considerable attention. However, the efficient control of long-term photostability of Au clusters on the metal-support interface remains challenging. Herein, we report a simple and efficient method for enhancing the photostability of glutathione-protected Au clusters (Au GSH clusters) loaded on the surface of SiO2 sphere by utilizing multifunctional branched poly-ethylenimine (BPEI) as a surface charge modifying, reducing and stabilizing agent. The sequential coating of thickness controlled TiO2 shells can further significantly improve the photocatalytic efficiency, while such structurally designed core-shell SiO2-Au GSH clusters-BPEI@TiO2 composites maintain high photostability during longtime light illumination conditions. This joint strategy via interfacial modification and composition engineering provides a facile guideline for stabilizing ultrasmall Au clusters and rational design of Au clusters-based composites with improved activity toward targeting applications in photoredox catalysis.

form reactive oxygen species) and high stability of the system. This two contradict propositions must be explained, i.e., gold should be oxidized by this mechanism, and thus either it can be dissolved or electron donor should be applied (considering gluthatione as electron donor, its regeneration should be proposed), 2) Dyes cannot be used for activity testing due to sensitization mechanisms (see reviews by B. Ohtani), other colorless compounds should be used for mechanism study (Application of chromate as testing molecule is also not recommended since it could sensitize titania (as has been published, e.g., Kuncewicz et al.)).
3) Authors concluded that titania played important role in the mechanism (this is obvious), but they did not perform any reference experiments for: (i) bare titania and (ii) titania with gold nanoclusters. 4) What about photocatalytic activity of ZnO-Au GSH clusters-BPEI (shown in Supplementary  Figure 9)? 5) Although, the existence of optimal thickness of titania is not surprising, especially in the case of amorphous titania with plenty of recombination sites (and many reports on similar behaviors have been already published; some of them should be cited), there is a question why crystallization of titania was not performed. 6) Authors presented plenty of supporting data, but some of them were not discussed, e.g., interesting data shown in Supplementary Figure 22 (impedance spectroscopy) should be discussed (at least in supplementary material file) and values on Y-axis should be shown. 7) Do authors use "Zeta potential" and "surface charge" interchangeable? 8) There are some strange phrases and mistakes, which should be corrected, e.g., "Recent years have seen..".

Reviewer #3 (Remarks to the Author):
This manuscript reports the use of a branched polyethylimine to act as interface and stabilizing agent of small Au cluster on the surface of silica and other metal oxides. Due to the strong tendency of Au clusters to decompose and agglomerate forming Au nanoparticles, the potential use of these Au clusters is very limited. The authors address this issue of stability, proposing that the main mechanism of Au decomposition is oxidation of ligands and these could be avoided due to the antioxidant activity of the polyimine. In addition strong Coulombic forces maintain the Au cluster attached to the branched polymer onto the metal oxide surface, avoiding metal leaching. The authors have shown the advantage of the proposed methodology by using the Au cluster on silica coated and coated it with TiO2, the resulting multicomponent nanoparticulated material acting as visible light photocatalyst for degradation of dyes and pollutants in water. Conoisdering the interest in the use of Au clusters in catalysis and photocatalysis, the proposed methodology could be of large interest. Publication in Nature Communications is recommended with some changes addressing the following comments: • The presence of the TiO2 shell increases considerably adsorption of the dye compared to silica. It is unclear how the authors have taken into account the increased adsorption when obtaining the corrected photocatalytic degradation data.
• It is known that amorphous TiO2 has much lower catalytic activity than crystalline TiO2 phases. The present procedure does not allow the typical calcination at 350 oC to crystallize amorphous titania. A comment on this issue (and possible a strategy to circumvent this issue) should be included in the manuscript. • To support the role of polyimine as stabilizer, the results of a control in which the SAP material is irradiated in the presence of an antioxidant will be important. • According to the author's proposal polyimine would undergo oxidation in the way to stabilize the Au cluster and would undergo presumably decomposition? Am I right? This indicates that the branched polyimine will be a sacrificial agent that eventually could loose activity. The authors should comment on this.
• Similarly, the role of silica on the photocatalytic activity is unclear. Since the authors have shown that the use of branched polyimine serves for different metal oxides, would it be possible to use this approach to deposit Au clusters on anatase TiO2 and text the photocatalytic activity? Other points: • Which is the Au+/Au(0) proportion of the Au GSH clusters determined by XPS? • It is unclear why the authors use the prefix bio to indicate that the polyethylimine is reducible. Please clarify. • It is indicated in the text that Au clusters could be used as visible light photosensitizer. A comment on the λmax absorption and other photophysical relevant properties of Au GSH as photosensitizer would be important.
Once these points have addressed, publication in Nature Communications should proceed. We greatly appreciate the reviewer's comment with regard to the loading amounts of Au GSH clusters in the SAB and SAP samples. We totally agree with the reviewer's opinion that the difference in the interaction between SiO 2 and Au GSH clusters will lead to the different loading amount of Au GSH clusters on SiO 2 surface. Accordingly, the loading amounts of Au GSH clusters in the SAB and SAP samples have been quantified by an inductively coupled plasma optical emission spectroscopy (ICP-OES). As shown in Supplementary Table 1, the addition amount of Au GSH clusters in SAB is demonstrated to be 0.86%, which is higher than that of SAP (0.28%). The different loading amounts of Au GSH clusters between SAB and SAP samples can be attributed to different interaction between SiO 2 and Au GSH clusters, as mentioned by the reviewer. The BPEI modified SiO 2 spheres for fabricating SAB sample exhibit relatively high positively charge surface with a zeta potential value of +36 mV ( Supplementary Fig. 4a), which facilitates the strong electrostatic interaction between SiO 2 spheres with negatively charged Au GSH clusters (-21 mV, Supplementary Fig. 4c) via coulombic forces. In contrast, the SiO 2 spheres display a weak positive charge with a zeta potential value of +5 mV ( Supplementary Fig. 4b) and the carboxylic groups in the glutathione ligand of Au GSH clusters are protonated under acidic conditions at pH 2 (J. Am. Chem. Soc. 2016, 138, 390-401), which subsequently leads to the inferior electrostatic attraction, resulting in the low addition amount of Au GSH clusters onto the surface of SiO 2 in SAP. For a fair comparison on the photoactivity of various SABT and SAPT samples, the photocatalytic performances of SAPT composites have been normalized with respect to the loading amount of Au GSH clusters in SAB, as revealed in Supplementary Fig. 22 in the revised supplementary information. The normalized RhB degradation efficiencies over different SAPT composites ( Supplementary Fig. 22) are worse than that of SABT samples (Fig. 3a) under identical reaction conditions, which suggests that the presence of BPEI layer is beneficial for the photoactivity enhancement since the photostability of Au GSH clusters in the SABT samples is significantly ameliorated.  suggest that the size of Au GSH clusters over SAB after 24 h visible light irradiation maintains unchanged, indicating that the BPEI layer as interfacial modification in the SAB system can provide a long time protection (24 h) with regard to the stabilization of Au GSH clusters under continuous visible light illumination (λ > 420 nm). To address the reviewer's concern, the irradiation time over SAB has been further prolonged to 36 h and 48 h and the size information of the Au GSH clusters in SAB has also been studied. As illustrated in Supplementary Fig. 9 in the revised supplementary information, after 36 h and 48 h visible light irradiation, the size of Au GSH clusters will increase to 2.0 nm and 2.1 nm, respectively. The slight increase of Au GSH clusters size may be attributed to the partial depletion of BPEI since the BPEI layer that serves as a reducing agent would undergo oxidation and/or decomposition in the way to stabilize the Au GSH clusters.

Supplementary
Particularly, we would like to emphasize that, despite various studies have observed the transformation of ultrasmall Au GSH clusters to larger Au NPs Our present work employees the BPEI as interfacial modification of SiO 2 -Au GSH clusters composites, which can maintain the size and structure of Au GSH clusters over 24 h under continuous visible light irradiation (λ > 420 nm) on the surface of SiO 2 supports. This is the first research work regarding how to stabilize the ultrasmall Au GSH clusters for designing efficient and stable Au GSH clusters-based composite photocatalysts. Additionally, the surface of SAB composites has been subsequently coated by a thickness tunable TiO 2 shell for constructing core-shell SiO 2 -Au GSH clusters-BPEI@TiO 2 (SABT) structures, which not only further contributes to stabilizing the ultrasmall Au GSH clusters but also improves the photocatalytic activities of Au GSH clusters during catalytic reactions under visible light illumination. Thus, our joint strategy via interfacial modification and sequential coating of semiconductor shell provides a simple and effective approach for stabilizing Au clusters with improved photocatalytic performance. Figure 8. TEM image (a) and HRTEM image (b) of SiO 2 -Au GSH clusters-BPEI composites (SAB) after visible light irradiation (λ > 420 nm) for 24 h; size distribution histogram (c) of Au GSH clusters over SAB after visible light irradiation (λ > 420 nm) for 24 h.
In this work, the photoelectrochemical (PEC) performances in Fig. 3d and f suggest that the separation efficiency of photoinduced charge carriers over the SABT-0.05 composite with thin layer of TiO 2 is the best among these samples because a thicker TiO 2 shell will block the transport of photogenerated electrons from Au GSH clusters to the back contact, thus reducing the overall PEC performances. Notably, with the increased thickness of TiO 2 shell, the adsorption capacity of SABT composites gradually increases, as shown in Fig. 3g, and the SABT-0.2 exhibits the strongest adsorption toward RhB among these samples. Therefore, the dye adsorption and charge carrier separation should have a synergistic effect on the catalytic performance, which rationalizes the best photoactivity for dye degradation achieved neither over SABT-0.2 with the highest RhB adsorption capacity nor SABT-0.05 with the most efficient charge carrier separation efficiency. It is the SABT-0.15, which could balance the combined influence of the RhB adsorption and charge carrier separation, that acquires the best photocatalytic performance. Based on the above discussion, the primary role of coating TiO 2 semiconductor layer onto SAB for ameliorating the photocatalytic efficiency of SABT is two-fold. One is to improve the separation and migration efficiency of photoinduced charge carriers from excited Au GSH clusters. The other role of TiO 2 shell is to enhance the adsorption capacity of the catalysts toward the reactants. By adjusting the thickness of TiO 2 shells, the photoactivity of SABT can be tuned due to cooperative adsorption and PEC properties. Therefore, the optimal SABT-0.15 composite that exhibits the highest activity toward photocatalytic degradation of RhB among these SABT samples is reasonable and understandable. Photocatalytic degradation of (a) RhB over blank SiO 2 spheres, TiO 2 , SAB, TAB and SABT composites with different TiO 2 shell thickness under visible light irradiation (λ > 420 nm) for 0.5 h; photocatalytic reduction of (b) p-methoxy nitrobenzene to p-methoxy aniline and over blank SiO 2 spheres, TiO 2 , SAB, TAB and SABT composites with different TiO 2 shell thickness under visible light irradiation for 5 h; (c) recycling photocatalytic degradation of RhB over optimal SABT-0.15 composite under visible light irradiation (λ > 420 nm); (d) transient photocurrent densities of SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness under visible light irradiation (λ > 420 nm); (e) scheme illustrating the photogenerated electron transport pathways between Au GSH clusters and TiO 2 shell; (f) cyclic voltammograms of SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness; (g) bar plots showing the remaining RhB in reaction solutions after being kept in dark for 3 h to achieve the adsorption-desorption equilibrium over SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness; (h) nitrogen adsorption−desorption isotherms of SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness; (i) surface area of SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness.
Corresponding revisions highlighted in red in the revised manuscript:  Comments 4: This work looks systematically performed and the analyses are quite thorough. The idea of using BPEI is interesting, but I do not think the significance of this work meets the requirement for the publication in this journal. Scientific Reports or other relevant journals would be a better place for this work if all the concerns are properly addressed.

Author reply:
At first, we appreciate the reviewer's evaluation that our work is interesting and the analyses are quite thorough. However, we do not think the assessment that the significance of this work cannot meet the requirement of the journal of Nature Communications is objective, in view of the following background in this significant research field.
Noble metal nanoparticles (MNPs) are uniquely suited for heterogeneous catalysis because of their relatively high specific surface areas and abundant active centers. The size of MNPs has been shown to be one of the most important factors that dictates the performance of a catalyst. Recently, considerable interest has been shown in controlling ultrasmall gold (Au) nanoparticles with atomic precision, which are often called Au clusters, due to their distinctive properties. These ligands protected Au clusters, typically glutathione-protected Au clusters (Au GSH clusters), can serve as both photosensitizer and catalytic center for multifunctional use in photoredox catalysis, such as water splitting, selective organic synthesis and pollutants degradation (for example, see . However, such ultrasmall Au GSH clusters suffer from serious instability under light irradiation due to its extremely high surface energy and large surface. The loading of Au GSH clusters onto different supports is often inefficient to avoid coalescence and agglomeration of these Au GSH clusters. Actually, up to now, the efficient control of long-term photostability of ultrasmall Au GSH clusters on the metal-support interface has never been achieved. In this paper, we report a facile and general strategy for enhancing the photostability of Au GSH clusters loaded on the surface of SiO 2 sphere by utilizing multifunctional branched poly-ethylenimine (BPEI) as a surface charge modifying, reducing and stabilizing agent. In addition, while simultaneously maintaining the photostability of Au GSH clusters, the sequential coating of TiO 2 shells for constructing core-shell SiO 2 -Au GSH clusters-BPEI@TiO 2 (SABT) nanostructures has been demonstrated to significantly ameliorate the photoactivity by regulating the photoelectrochemical and adsorption properties of SABT composites synergistically. This work is expected to stimulate broad scientific and technological interests in this special type of metal nanomaterial and provide a guideline in structural design of Au clusters-based composites with improved catalytic performance and long-term photostability.
We believe that this manuscript is of high interest to the general readership of Nature . However, the effective control of these ultrasmall Au GSH clusters with long-term stability on the substrates under in situ photo-irradiation conditions still remains a challenge, which becomes the main bottleneck for the development of Au clusters-based catalysts systems for the long-term use. In this work, we report a facile and general strategy of using multifunctional branched poly-ethylenimine as surface charge modifying, reducing and stabilizing agents to enhance the photostability of Au GSH clusters loaded on the surface of SiO 2 sphere supports.
Secondly, although some excellent works of Au clusters-semiconductor composites have been reported for photocatalytic solar energy conversions by randomly decorating Au clusters onto the surface of semiconductor supports, the rational structure design for achieving high efficient Au clusters-semiconductor composites is barely explored in previous reports. In this work, while simultaneously maintaining the photostability of Au GSH clusters, core-shell SiO 2 -Au GSH clusters-BPEI@TiO 2 (SABT) composites have been constructed by coating of the thickness-controlled TiO 2 shells to significantly ameliorate the photoactivity. Besides, the adsorption and photoelectrochemical properties of SABT composites can be well regulated by simply adjusting the thickness of TiO 2 shell, thereby synergistically ameliorating the photoactivity of SABT composites.
It is anticipated that this work could raise broad scientific and technological interests in this special type of metal nanomaterial and offer an avenue for rational designing of Au clusters-based composites photocatalyst with high catalytic efficiency and long-term stability for photoredox catalysis.

12
Response to the comments of reviewer 2:

Comments:
The manuscript on preparation of highly stable gold nanoclusters by B. Weng et al. is very interesting. Although, the overall quality of the manuscript is high, there are some important issues that should be explained before possible publications, as shown below:

Author reply:
We deeply appreciate the reviewer's evaluation that our work is very interesting and the overall quality of the manuscript is high. We have revised our manuscript with great attention according to your valuable comments.
Comments 1: The mechanism must be clarified. Authors proposed photo-generation of electrons (able to form reactive oxygen species) and high stability of the system. This two contradict propositions must be explained, i.e., gold should be oxidized by this mechanism, and thus either it can be dissolved or electron donor should be applied (considering gluthatione as electron donor, its regeneration should be proposed),

Author reply:
Thanks for your valuable comments. We agree with the reviewer's opinion that the presence of reactive oxygen species (ROSs) is unfavorable for the stability of Au GSH clusters and it has been reported that the onslaught of ROSs on organic ligands of Au GSH clusters can lead to the instability of Au GSH clusters and their aggregation into large Au nanoparticles (ACS Appl. Mater. Interfaces 2015, 7, 28105-28109; Sci. Rep. 2016, 6, 22742). Notably, in our reaction systems, the Au GSH clusters is found to be stable during photocatalytic reactions and the following factors are considered to be the decisive role for achieving high photostability of Au GSH clusters in SABT composites.
On one hand, BPEI with high reductive capability could serve as an electron donor and/or an effective reducing and stabilizing agent to protect the organic ligands of Au GSH clusters from being oxidized by ROSs during the photo-irradiation process. Specifically, due to the presence of BPEI layer, the R−SO 3 species belonging to the oxidation products of gluthatione have not been detected even after a long time irradiation (24 h) over the SAB composites, which indicates the critical role of BPEI layer in stabilizing Au GSH clusters against aggregation under continuous light illumination condition. Furthermore, the multiple polymer chains of primary, secondary and tertiary amine groups in BPEI could encapsulate the as-synthesized Au GSH clusters via crosslinking (Chem. Commun., 2014, 50, 88-90), which could fix these clusters on the surface of supports and hamper the migration of clusters for fusing, thus stabilizing the ultrasmall Au GSH clusters.
On the other hand, the intelligent design of core-shell SABT composites by coating TiO 2 shell can enhance the adsorption capacity of catalysts toward reactants (Fig. 3g), which would consume the ROSs timely and effectively to reduce the possibility of the reactions between ROSs with the organic ligands of Au GSH clusters, thus leading to the improvement of the photostability of Au GSH clusters in SABT composites.

13
Figure 3 | The enhancement of photocatalytic performance over SABT composites and the underlying mechanism. Photocatalytic degradation of (a) RhB over blank SiO 2 spheres, TiO 2 , SAB, TAB and SABT composites with different TiO 2 shell thickness under visible light irradiation (λ > 420 nm) for 0.5 h; photocatalytic reduction of (b) p-methoxy nitrobenzene to p-methoxy aniline and over blank SiO 2 spheres, TiO 2 , SAB, TAB and SABT composites with different TiO 2 shell thickness under visible light irradiation for 5 h; (c) recycling photocatalytic degradation of RhB over optimal SABT-0.15 composite under visible light irradiation (λ > 420 nm); (d) transient photocurrent densities of SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness under visible light irradiation (λ > 420 nm); (e) scheme illustrating the photogenerated electron transport pathways between Au GSH clusters and TiO 2 shell; (f) cyclic voltammograms of SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness; (g) bar plots showing the remaining RhB in reaction solutions after being kept in dark for 3 h to achieve the adsorption-desorption equilibrium over SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness; (h) nitrogen adsorption−desorption isotherms of SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness; (i) surface area of SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness.
Comments 2: Dyes cannot be used for activity testing due to sensitization mechanisms (see reviews by B. Ohtani), other colorless compounds should be used for mechanism study (Application of chromate as testing molecule is also not recommended since it could sensitize titania (as has been published, e.g., Kuncewicz et al.)).

Author reply:
Thanks for your helpful comments. We agree with the reviewer that the dye and chromate sensitized mechanisms could be present in the photocatalytic activity test. Actually, to eliminate possible dyes (i.e., RhB, and MB) and chromate (i.e., Cr(VI)) photosensitization effect on the activity, photocatalytic reduction of p-methoxy nitrobenzene to p-methoxy aniline over various SABT composites has also been performed in the original manuscript under visible light irradiation (λ > 420 nm), as shown in Fig. 3b in the revised manuscript (i.e., Fig. 16 in the original Supplementary Information). The higher photocatalytic activity of SABT composites than that of SiO 2 spheres and SAB sample verifies the positive role of TiO 2 shell in improving the photoactivity of SiO 2 -Au GSH clusters composites. And the SABT-0.15 shows the best photoactivity among these samples, elucidating that the thickness of TiO 2 shell plays a critical role in determining the photocatalytic performance of SABT toward p-methoxy nitrobenzene reduction. Moreover, due to the intelligently designed core-shell structure, all of the SABT composites show higher catalytic activity than the TiO 2 -Au GSH cluster-BPEI (TAB) composite. The photoactivity tendency toward

p-methoxy nitrobenzene reduction over SABT composites is similar with that of photocatalytic degradation of dyes and reduction of Cr(VI) and further corroborates our conclusions.
Additionally, it is well known that the sensitization efficiency of RhB during the photocatalytic reaction is generally low (Appl.  Photocatalytic degradation of (a) RhB over blank SiO 2 spheres, TiO 2 , SAB, TAB and SABT composites with different TiO 2 shell thickness under visible light irradiation (λ > 420 nm) for 0.5 h; photocatalytic reduction of (b) p-methoxy nitrobenzene to p-methoxy aniline and over blank SiO 2 spheres, TiO 2 , SAB, TAB and SABT composites with different TiO 2 shell thickness under visible light irradiation for 5 h; (c) recycling photocatalytic degradation of RhB over optimal SABT-0.15 composite under visible light irradiation (λ > 420 nm); (d) transient photocurrent densities of SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness under visible light irradiation (λ > 420 nm); (e) scheme illustrating the photogenerated electron transport pathways between Au GSH clusters and TiO 2 shell; (f) cyclic voltammograms of SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness; (g) bar plots showing the remaining RhB in reaction solutions after being kept in dark for 3 h to achieve the adsorption-desorption equilibrium over SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness; (h) nitrogen adsorption−desorption isotherms of SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness; (i) surface area of SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness.
Corresponding revisions highlighted in red in the revised manuscript: Figure 3 has been revised in the revised manuscript.

Lines 18-30 of Page 11:
To investigate the effect of surface coating of TiO 2 shell on the photocatalytic performance and exclude the possible dye photosensitization effect of RhB on the activity 56 , the visible light photoactivities of SiO 2 spheres, TAB, SAB and SABT composites towards photocatalytic reduction of p-methoxy nitrobenzene to p-methoxy aniline have been evaluated, which is a typical six-electron reduction reaction (Supplementary Equation 1) 24, 27, 47, 57 . As presented in Fig. 3b, the higher photocatalytic activity of SABT composites than that of SiO 2 spheres and SAB sample verifies the positive role of TiO 2 shell in improving the photoactivity of SiO 2 -Au GSH clusters composites. And the SABT-0.15 shows the best photoactivity among these samples, elucidating that the thickness of TiO 2 shell plays a critical role in determining the photocatalytic performance of SABT toward p-methoxy nitrobenzene reduction. Moreover, due to the intelligently designed core-shell structure, all of the SABT composites show higher catalytic activity than the TAB composite.
Comments 3: Authors concluded that titania played important role in the mechanism (this is obvious), but they did not perform any reference experiments for: (i) bare titania and (ii) titania with gold nanoclusters.

Author reply:
We deeply thank the reviewer for this valuable comment. As the reviewer recommended, the photoactivities of bare TiO 2 and TiO 2 -Au GSH clusters-BPEI composite (TAB) toward photocatalytic RhB degradation and reduction of p-methoxy nitrobenzene have been supplemented in our revised manuscript. As illustrated in Fig. 3, it is clear that bare TiO 2 catalyst exhibits negligible photoactivity since the visible light (λ > 420 nm) cannot excite semiconductor TiO 2 with a band gap of 3.2 eV to generate electron-hole pairs for driving the photocatalytic reactions. Notably, the TAB sample with the decoration of Au GSH clusters shows obvious photoactivity toward degradation of RhBs and reduction of p-methoxy nitrobenzene, which confirms that the Au GSH clusters can act as photosensitizers to generate electron-hole pairs under visible light irradiation, thereby triggering various photocatalytic reactions. Furthermore, all of SABT composites exhibit higher photocatalytic efficiency than TAB sample, which indicates that the intelligently designed core-shell structure of SABT is favorable for enhancing the photocatalytic performance of Au GSH clusters-semiconductor composites.
Figure 3 | The enhancement of photocatalytic performance over SABT composites and the underlying mechanism. Photocatalytic degradation of (a) RhB over blank SiO 2 spheres, TiO 2 , SAB, TAB and SABT composites with different TiO 2 shell thickness under visible light irradiation (λ > 420 nm) for 0.5 h; photocatalytic reduction of (b) p-methoxy nitrobenzene to p-methoxy aniline and over blank SiO 2 spheres, TiO 2 , SAB, TAB and SABT composites with different TiO 2 shell thickness under visible light irradiation for 5 h; (c) recycling photocatalytic degradation of RhB over optimal SABT-0.15 composite under visible light irradiation (λ > 420 nm); (d) transient photocurrent densities of SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness under visible light irradiation (λ > 420 nm); (e) scheme illustrating the photogenerated electron transport pathways between Au GSH clusters and TiO 2 shell; (f) cyclic voltammograms of SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness; (g) bar plots showing the remaining RhB in reaction solutions after being kept in dark for 3 h to achieve the adsorption-desorption equilibrium over SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness; (h) nitrogen adsorption−desorption isotherms of SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness; (i) surface area of SiO 2 spheres, SAB and SABT composites with different TiO 2 shell thickness.

Corresponding revisions highlighted in red in the revised manuscript:
Lines 34-35 of Page 10 and Lines 1-6 of Page 11: Furthermore, the sample of TiO 2 -Au GSH cluster-BPEI (TAB) exhibits worse catalytic activity than the SABT composites, which indicates that the intelligently designed core-shell structure of SABT is favorable for enhancing the photocatalytic performance of Au GSH clusters-semiconductor composites. The pseudo-first order kinetic of the degradation of RhB based on the above data is shown in Supplementary Table 2. It is clear that all of the SABT samples show higher reaction rate than the SAB and TAB samples, among which SABT-0.15 exhibits the highest reaction rate of 0.119 min -1 among these SABT composites.

Page 26:
The photoactivities of bare TiO 2 and TiO 2 -Au GSH clusters-BPEI (TAB) have been supplemented in the revised Figure 3. Supplementary Figure 9)?

Author reply:
Thanks for your kind comments. To address the reviewer's concern, the photocatalytic activity of ZnO-Au GSH clusters-BPEI (ZAB) has been evaluated toward RhB degradation under visible light illumination (λ > 420 nm) for 9 h. As shown in Supplementary Figure 18a, nearly 96% of RhB can be degraded over ZAB. To investigate the dye sensitization effect of RhB on the activity, the photocatalytic performance of bare ZnO has also been tested. Notably, the sample of bare ZnO removes ca.79% of RhB with 9 h of visible light irradiation, which is governed by a dye self-sensitization process because the wide-band gap of ZnO semiconductor (3.2 eV) cannot be activated in the visible light region (J. Photochem. Photobiol. A 2004, 162, 317-322; J. Mater. Chem. A 2014, 2, 9380-9389). The ZAB with the decoration of Au GSH clusters exhibits moderate photoactivity enhancement than bare ZnO, which suggests that the random loading of ultrasmall Au GSH clusters onto ZnO supports without rational structure design is inefficient to achieve high efficient Au GSH clusters-semiconductor composites. For your kind reference, the photocatalytic activitives toward RhB degradation over ZAB composite and ZnO sample have been displayed as follows.  Corresponding revisions highlighted in red in the revised manuscript: Lines 11-23 of Page 9: Notably, it has been demonstrated that the CB edge potential of metal oxides (e.g., ZnO and rutile TiO 2 ) is more positive than the LUMO potential of Au GSH clusters, which enables the transformation of photoexcited electrons from Au GSH clusters to the metal oxide supports 15, 24, 53 . The photoactivities of BPEI modified metal oxides-Au GSH clusters composites (denoted as MAB) have been evaluated toward RhB degradation under visible light irradiation (λ > 420 nm). As shown in Supplementary Fig. 18, the samples of MAB exhibit moderate photoactivity enhancement than bare semiconductors, which suggests that the random loading of ultrasmall Au GSH clusters onto semiconductors without rational structure design is inefficient to achieve high efficient Au GSH clusters-semiconductor composites. Therefore, a thickness tunable TiO 2 shell has been coated onto the surface of SAB composites for designing core-shell SiO 2 -Au GSH clusters-BPEI@TiO 2 structures to construct high performance Au GSH clusters-semiconductor composites for solar energy conversion.

Corresponding revisions highlighted in red in the revised supplementary information:
Supplementary Figure 18 has been added.
Comments 5: Although, the existence of optimal thickness of titania is not surprising, especially in the case of amorphous titania with plenty of recombination sites (and many reports on similar behaviors have been already published; some of them should be cited), there is a question why crystallization of titania was not performed.

Author reply:
Thanks for your valuable comments. We agree with the reviewer's opinion that the amorphous TiO 2 with defects could exhibit worse photoactivity than crystalline TiO 2 , which has also been demonstrated by previous works (J. Phys. Chem. B, 1997, 101 ., 2015, 8, 286-296). According to your precious suggestion, we have paid our endeavors to enhance the crystallization of TiO 2 , including calcination and hydrothermal treatment, since the high temperature is generally required to transform amorphous TiO 2 particle into an anatase one.
The sample of SABT-0.15 has been calcinated at 350 under Ar for 2 h and the calcinated sample is denoted as SABT-0.15-350. Even though the peak of anatase TiO 2 (A-TiO 2 ) can be observed in the XRD pattern ( Supplementary Fig. 23c), the color of SABT-0.15-350 sample changes to purple, as shown in Supplementary Fig. 23b, which corresponds to the color of Au nanoparticles (NPs), indicating that the Au GSH clusters aggregate into large Au NPs due to the high calcination temperature. Furthermore, the DRS spectrum of SABT-0.15-350 sample in Supplementary Fig. 23a exhibit a surface plasmon resonance (SPR) peak located at 550 nm belonging to the metallic Au NPs, confirming the fusion of Au GSH clusters into Au NPs with large size. More direct evidence comes from the TEM image of SABT-0.15-350 sample, as pictured in Supplementary Fig. 23d, and the size of Au NPs is demonstrated to be 7.8 nm ( Supplementary Fig. 23e), which suggests that the calcination is unsuitable for enhancing the crystallinity of TiO 2 since the high temperature can lead to the aggregation of Au GSH clusters.
We then treat the SABT-0.15 composite under hydrothermal condition at 180 for 12 h for crystallizing TiO 2 , and the obtained sample is labeled as SABT-0.15-180. The application of elevated temperatures and pressures in an aqueous solution could facilitate the conversion of amorphous TiO 2 into crystalline TiO 2 and cause an increase in its crystallinity. The XRD result in Supplementary Fig.  24c indicates that the TiO 2 in SABT-0.15-180 sample is anatase (A-TiO 2 ). However, the intelligently designed core-shell structure of SABT-0.15 is destroyed during the hydrothermal process, as revealed by TEM image in Supplementary Fig. 24d. Moreover, the coalescence of Au GSH clusters into larger metallic Au NPs has also been observed and confirmed by a series of techniques, as displayed in Supplementary Fig. 24a, b and d. The size of Au NPs in SABT-0.15-180 sample is calculated to be 3.0 nm ( Supplementary Fig. 24e). The fusion of Au GSH clusters may be ascribted to the high temperatures and pressures during the hydrothermal process, which is essential for crystallizing TiO 2 .
The photocatalytic performances of SABT-0.15-350 and SABT-0.15-180 composites have been evaluated toward photocatalytic RhB degradation and reduction of p-methoxy nitrobenzene under visible light illumination (λ > 420 nm), as shown in Supplementary Fig. 25. Both the SABT-0.15-350 and SABT-0.15-180 samples exhibit poor photoactivity as compared with SABT-0.15 composite, which could be arrtibuted to the aggregation of Au GSH clusters into large metallic Au NPs and the destroy of core-shell structure of SABT-0.15-180 composite. According to your precious suggestion, these results have been supplemented in our revised manuscript to provide more information for the readership with citation of relevant references in our revised manuscript.

Lines 30-35 of Page 11 and Lines 1-4 of Page 12:
In addition to the TiO 2 shell thickness, the crystallization of TiO 2 layer in composites is also a factor that affects the photocatalytic performance of SABT composites and it has been reported that the amorphous TiO 2 usually exhibits poor photocatalytic activity 58-62 . We have paid our endeavors to crystallize TiO 2 shell in SABT-0.15, including calcination and hydrothermal treatment, since the high temperature is generally required to transform amorphous TiO 2 particle into an anatase one. Even though the crystallization of TiO 2 can be improved by calcination and hydrothermal treatment, the Au GSH clusters in SABT-0.15 suffer from serious instability under high temperature and aggregate into metallic Au nanoparticles with large size, as displayed in Supplementary Fig. 23-24, thereby deteriorating the photocatalytic performance of SABT-0.15 samples toward various photocatalytic reactions under visible light illumination (λ > 420 nm) ( Supplementary Fig. 25). Kt

Page 23, References
where C 0 is the initial concentration of the reactant, C is the concentration of the reactant, t is the time, K is the adsorption coefficient of the reactant, kr is the reaction rate constant. Generally, due to the initial concentration of pollutant is low, the equation can be simplified as (Appl. where k′ is the apparent rate constant. The results of calculated reaction rate constants are 26 presented in Supplementary Table 2 in the revised supplementary information. It is clearly observed that the reaction rates of various SABT samples increase with increasing thickness of TiO 2 layer and the SABT-0.15 composite exhibits highest reaction rate of 0.119 min -1 due to the presence of optimum TiO 2 shell thickness, which is consistent with the photocatalytic degradation efficiency, as displayed in Fig. 3a. Additionally, we would like to emphasize that, the performance of photocatalysts is determined not only by the adsorption ability of catalyst but also the charge carrier separation efficiency. As shown in Fig. 3g, even though the SABT-0.2 composites have the strongest adsorption toward dyes, the photocatalytic efficiency of SABT-0.2 is lower than that of SABT-0.15 (Fig. 3a), which is attributed to the fact that the dye adsorption and charge carrier separation should have a synergistic effect on the catalytic performance.

Corresponding revisions highlighted in red in the revised manuscript:
Lines 2-5 of Page 11: The pseudo-first order kinetic of the degradation of RhB based on the above data is shown in Supplementary Table 2. It is clear that all of the SABT samples show higher reaction rate than the SAB and TAB samples, among which SABT-0.15 exhibits the highest reaction rate of 0.119 min -1 among these SABT composites.

Corresponding revisions highlighted in red in the revised supplementary information:
Supplementary

Author reply:
Thanks for your kind comments. We agree with the reviewer's opinion that the amorphous TiO 2 with defects could exhibit worse photoactivity than crystalline TiO 2 , which has also been demonstrated by previous works (J. Phys. Chem. B 1997, 101, 3746- 2015, 8, 286-296). According to your precious suggestion, we have paid our endeavors to enhance the crystallization of TiO 2 , including calcination and hydrothermal treatment, since the high temperature is generally required to transform amorphous TiO 2 particle into an anatase one.
The sample of SABT-0.15 has been calcinated at 350 under Ar for 2 h and the calcinated sample is denoted as SABT-0.15-350. Even though the peak of anatase TiO 2 (A-TiO 2 ) can be observed in the XRD pattern ( Supplementary Fig. 23c), the color of SABT-0.15-350 sample changes to purple, as shown in Supplementary Fig. 23b, which corresponds to the color of Au nanoparticles (NPs), indicating that the Au GSH clusters aggregate into large Au NPs due to the high calcination temperature. Furthermore, the DRS spectrum of SABT-0.15-350 sample in Supplementary Fig. 23a exhibit a surface plasmon resonance (SPR) peak located at 550 nm belonging to the metallic Au NPs, confirming the fusion of Au GSH clusters into Au NPs with large size. More direct evidence comes from the TEM image of SABT-0.15-350 sample, as pictured in Supplementary Fig. 23d, and the size of Au NPs is demonstrated to be 7.8 nm (Supplementary Fig. 23e), which suggests that the calcination is unsuitable for enhancing the crystallinity of TiO 2 since the high temperature can lead to the aggregation of Au GSH clusters.
We then treat the SABT-0.15 composite under hydrothermal condition at 180 for 12 h for crystallizing TiO 2 , and the obtained sample is labeled as SABT-0.15-180. The application of elevated temperatures and pressures in an aqueous solution could facilitate the conversion of amorphous TiO 2 into crystalline TiO 2 and cause an increase in its crystallinity. The XRD result in Supplementary Fig.  24c indicates that the TiO 2 in SABT-0.15-180 sample is anatase (A-TiO 2 ). However, the intelligently designed core-shell structure of SABT-0.15 is destroyed during the hydrothermal process, as revealed by TEM image in Supplementary Fig. 24d. Moreover, the coalescence of Au GSH clusters into larger metallic Au NPs has also been observed and confirmed by a series of techniques, as displayed in Supplementary Fig. 24a, b and d. The size of Au NPs in SABT-0.15-180 sample is calculated to be 3.0 nm ( Supplementary Fig. 24e). The fusion of Au GSH clusters may be ascribted to the high temperatures and pressures during the hydrothermal process, which is essential for crystallizing TiO 2 .
The photocatalytic performances of SABT-0.15-350 and SABT-0.15-180 composites have been evaluated toward photocatalytic RhB degradation and reduction of p-methoxy nitrobenzene under visible light illumination (λ > 420 nm), as shown in Supplementary Fig. 25. Both the SABT-0.15-350 and SABT-0.15-180 samples exhibit poor photoactivity as compared with SABT-0.15 composite, which could be arrtibuted to the aggregation of Au GSH clusters into large metallic Au NPs and the destroy of core-shell structure of SABT-0.15-180 composite. According to your precious suggestion, these results have been supplemented in our revised manuscript to provide more information for the readership with citation of relevant references in our revised manuscript.

Lines 30-35 of Page 11 and Lines 1-4 of Page 12:
In addition to the TiO 2 shell thickness, the crystallization of TiO 2 layer in composites is also a factor that affects the photocatalytic performance of SABT composites and it has been reported that the amorphous TiO 2 usually exhibits poor photocatalytic activity 58-62 . We have paid our endeavors to crystallize TiO 2 shell in SABT-0.15, including calcination and hydrothermal treatment, since the high temperature is generally required to transform amorphous TiO 2 particle into an anatase one. Even though the crystallization of TiO 2 can be improved by calcination and hydrothermal treatment, the Au GSH clusters in SABT-0.15 suffer from serious instability under high temperature and aggregate into metallic Au nanoparticles with large size, as displayed in Supplementary Fig. 23-24, thereby deteriorating the photocatalytic performance of SABT-0.15 samples toward various photocatalytic reactions under visible light illumination (λ > 420 nm) ( Supplementary Fig. 25).

Comments 3:
To support the role of polyimine as stabilizer, the results of a control in which the SAP material is irradiated in the presence of an antioxidant will be important.

Author reply:
Thanks for your valuable comments. To address the reviewer's concern, the sample of SAP has been modified by BPEI (BPEI-SAP) and the as-obtained BPEI-SAP composite is irradiated under visible light (λ > 420 nm) for 10 h. As displayed in Supplementary Fig. 10a, the HRTEM image of BPEI-SAP after light illumination suggests that the aggregation of Au GSH clusters could be inhibited to some extent. The size of Au GSH clusters in BPEI-SAP after 10 h light irradiation is calculated to be 2.0 nm ( Supplementary Fig. 10b), which is smaller than that of SAP after light irradiation (6 nm, Fig. 1f), confirming the critical role of BPEI layer in stabilizing the ultrasmall Au GSH clusters.  (a) Schematic illustration of synthesis procedure for SiO 2 -Au GSH clusters-BPEI composites (SAB) and photostability testing of as-prepared SAB; transmission electron microscopy (TEM) images of SAB (b) before and (c) after visible light irradiation (λ > 420 nm) for 10 h and SiO 2 -Au GSH clusters-pH (SAP) (e) before and (f) after visible light irradiation for 10 h; the corresponding models of (d) SAB and (g) SAP after visible light irradiation for 10 h; high-resolution X-ray photoelectron spectroscopy (XPS) spectra of (h) S 2p and (i) Au 4f for SAB before/after visible light irradiation; (j) UV-vis diffuse reflectance spectrum (DRS) spectrum of SAP after 10 h visible light irradiation. Insets in the b and c are the corresponding models of SAB and SAP before visible light irradiation. Note: The histograms in c and f correspond to the particle size distributions of Au GSH clusters in the SAB and SAP after visible light irradiation for 10 h, respectively. The yellow spheres in Fig.1d represent Au GSH clusters and the purple spheres in Fig. 1g represent Au nanoparticles.

Corresponding revisions highlighted in red in the revised manuscript:
Lines 31-35 of Page 6 and Lines 1-2 of Page 7: To further confirm the role of BPEI in inhibiting the growth of Au GSH clusters, the SAP sample is subsequently modified by BPEI layer (BPEI-SAP) and the obtained BPEI-SAP composite is irradiated under visible light (λ > 420 nm) for 10 h. The HRTEM image of BPEI-SAP after light illumination in Supplementary Fig. 10a suggests that the aggregation Au GSH clusters could be prevented to some extent and the size of Au GSH clusters is calculated to be 2.0 nm ( Supplementary  Fig. 10b), confirming the critical role of BPEI layer in stabilizing the ultrasmall Au GSH clusters.

Corresponding revisions highlighted in red in the revised supplementary information:
Supplementary Figures 10 has been added.

Comments 4:
According to the author's proposal polyimine would undergo oxidation in the way to stabilize the Au cluster and would undergo presumably decomposition? Am I right? This indicates that the branched polyimine will be a sacrificial agent that eventually could loose activity. The authors should comment on this.

Author reply:
Thank you very much for your valuable comments. We certainly agree with the reviewer that BPEI may be consumed ultimately and the Au GSH clusters over SAB may lead to aggregation after BPEI layer is depleted if we irradiate the SAB sample for a very long time. In our original manuscript, the SAB sample has been irradiated under visible light irradiation (λ > 420 nm) for 24 h to investigate the efficiency of BPEI on inhibiting the fusion of Au GSH clusters. The results in Supplementary Fig. 8 suggest that the size of Au GSH clusters over SAB after 24 h visible light irradiation maintains unchanged, indicating that the BPEI layer as interfacial modification in the SAB system can provide a long time protection (24 h) with regard to the stabilization of Au GSH clusters under continuous visible light illumination (λ > 420 nm). To address the reviewer's concern, the irradiation time over SAB has been further prolonged to 36 h and 48 h and the size information of the Au GSH clusters in SAB has also been studied. As illustrated in Supplementary Fig. 9 in the revised supplementary information, after 36 h and 48 h visible light irradiation, the size of Au GSH clusters will increase to 2.0 nm and 2.1 nm, respectively. The slight increase of Au GSH clusters size may be attributed to the partial depletion of BPEI layer since the BPEI that serves as a reducing agent would undergo oxidation and/or decomposition in the way to stabilize the Au GSH clusters.
Particularly, we would like to emphasize that, despite various studies have observed the transformation of ultrasmall Au GSH clusters to larger Au NPs irradiation (λ > 420 nm) on the surface of SiO 2 supports. This is the first research work regarding how to stabilize the ultrasmall Au GSH clusters for designing efficient and stable Au GSH clusters-based composite photocatalysts. Additionally, the surface of SAB composites has been subsequently coated by a thickness tunable TiO 2 shell for constructing core-shell SiO 2 -Au GSH clusters-BPEI@TiO 2 (SABT) structures, which not only further contributes to stabilizing the ultrasmall Au GSH clusters but also improves the photocatalytic activities of Au GSH clusters during catalytic reactions under visible light illumination. Thus, our joint strategy via interfacial modification and sequential coating of semiconductor shell provides a simple and effective approach for stabilizing Au clusters with improved photocatalytic performance. Note: The EDX spectrum in Supplementary Fig. 9e evidences the presence of Au, O and Si elements over SAB sample and the detected element Cu can be attributed to the use of Cu grid, which serves as the support for TEM analysis.

Corresponding revisions highlighted in red in the revised manuscript:
Lines 27-31 of Page 6: When the irradiation time over SAB is further prolonged to 36 h and 48 h, the size of Au GSH clusters will increase to 2.0 nm and 2.1 nm, respectively, as illustrated in Supplementary Fig. 9. The slight increase of Au GSH clusters size is reasonable, which may be attributed to the partial depletion of BPEI since the BPEI layer that serves as a reducing agent would undergo oxidation and/or decomposition in the way to stabilize the Au GSH clusters.

Corresponding revisions highlighted in red in the revised supplementary information:
Supplementary Figure 9 has been added in the revised supplementary information.
Comments 5: Similarly, the role of silica on the photocatalytic activity is unclear. Since the authors have shown that the use of branched polyimine serves for different metal oxides, would it be possible to use this approach to deposit Au clusters on anatase TiO2 and text the photocatalytic activity?

Author reply:
Thanks for your comments. The insulation SiO 2 spheres has been chosen as the inert supports to predominantly focus on investigating the photosensitizer role of Au GSH clusters in photocatalytic applications due to its rather low optical absorption ( Supplementary Fig. 3). Additionally, the SiO 2 spheres with well-defined structure is favorable for the construction of core-shell SiO 2 -Au GSH clusters-BPEI@TiO 2 composites with high efficiency since the reported Au GSH clusters-semiconductor composites in literatures are generally fabricated by randomly loading the Au To address the reviewer's concern, the Au GSH clusters have been deposited on the surface of anatase TiO 2 (Fig. R1g, as shown below) via a pH value adjusted process and interfacial modification process. The stability of obtained anatase TiO 2 -Au GSH clusters-pH composites (ATAP) and anatase TiO 2 -Au GSH clusters-BPEI composites (ATAB) have been investigated under visible light illumination (λ > 420 nm). The HRTEM image in Fig. R1a reveals that the Au GSH clusters in ATAP suffer from fusion even after 0.5 h light irradiation, and the size of Au nanoparticles (NPs) is demonstrated to be 1.9 nm (Fig. R1d). As for the sample of ATAB, after 3 h visible light irradiation, the size and structure of Au GSH clusters remain unchanged (Fig. R1b and e), indicating the important role of BPEI in protecting the Au GSH clusters from being oxidized and restraining the growth of Au GSH clusters. The EDX spectrum in Fig. R1h evidences the presence of Au, O and Ti elements over ATAB sample after light illuminated for 3 h. Unfortunately, when the irradiation time of ATAB is further extended to 5 h, the Au NPs with size of 1.8 nm are detectable, which confirms the slight aggregation of Au GSH clusters, as shown in Fig. R1c and f. The fusion of Au GSH clusters on the surface of anatase TiO 2 may be ascribed to the presence of abundance surface hydroxyl group (Environ. Sci. Technol. 2013, 47, 2777-2783; J. Am. Chem. Soc. 2017, 139, 10020-10028), which could facilitate formation of . OH that decompose BPEI layer, thus resulting in the formation of Au NPs with large size. The above inference has been evidenced by the synthesis of rutile TiO 2 -Au GSH clusters-BPEI composites (RTAB), among which the rutile TiO 2 (Fig. R1i) is obtained by calcinating anatase TiO 2 at 850 for 5 h. The sample of RTAB has been exposed to continuous visible light irradiation (λ > 420 nm) for 10 h under ambient conditions and the size information of Au GSH clusters is given by the TEM analysis. As revealed in Fig. R2a and c, the Au GSH clusters maintain the size of 1.4 nm on the surface of rutile TiO 2 for RTAB composites after photo-irradiation. In constrast, the sample of rutile TiO 2 -Au GSH clusters-pH (RTAP) has also been fabricated and irradiated under visible light for 10 h, as illustrated in Fig. R2b and d. The Au GSH clusters have aggregated into Au NPs with size of 3 nm, which indicates the effect of BPEI modification on enhancing the stability of Au GSH clusters under visible light illumination.
The photocatalytic performances of rutile TiO 2 nanoparticles and RTAB composites have been studied by degradation of RhB under visible light irradiation (λ > 420 nm). As shown in Supplementary Fig. 18, the RhB degradation efficiency over RTAB composites is nearly 96%, which is higher than that of rutile TiO 2 (79%). The observed photoactivity enhancement is attributed to the addition of Au GSH clusters, which could generate electron-hole pairs under visible light irradiation to drive the photocatalytic degradation of RhB. Notably, due to the random loading of ultrasmall Au GSH clusters onto TiO 2 supports, the RTAB exhibits moderate photoactivity enhancement than bare rutile TiO 2 , which suggests that the random loading of ultrasmall Au GSH clusters onto semiconductors without rational structure design is inefficient to achieve high efficient Au GSH clusters-semiconductor composites. Figure R1. HRTEM image of (a) anatase TiO 2 -Au GSH clusters-pH composites (ATAP) after visible light irradiation (λ > 420 nm) for 0.5 h; HRTEM images of anatase TiO 2 -Au GSH clusters-BPEI composites (ATAB) after visible light irradiation (λ > 420 nm) for (b) 3 h and (c) 5 h; size distribution histogram of Au GSH clusters over (d) ATAP after visible light irradiation (λ > 420 nm) for 0.5 h; size distribution histograms of Au GSH clusters over ATAB after visible light irradiation (λ > 420 nm) for (e) 3 h and (f) 5 h; XRD pattern of (g) anatase TiO 2 ; (h) EDX spectrum of ATAB after visible light irradiation (λ > 420 nm) for 3 h; XRD pattern of (i) rutile TiO 2 .

Corresponding revisions highlighted in red in the revised manuscript:
Lines 5-14 of Page 7: Additionally, the surfaces of different metal oxide supports, including rutile TiO 2 , ZnO and ZrO 2 , have been positively charged by the BPEI modification ( Supplementary Fig. 11), which can subsequently interact with the negatively charged Au GSH clusters ( Supplementary Fig. 4c) via an electrostatic self-assembly method to produce metal oxide-Au GSH clusters-BPEI composites (MABs) and the photostability of MABs has been investigated under the same conditions as that for SAB and SAP. As illustrated in Supplementary Fig. 12, the mean diameter of Au GSH clusters on the surfaces of various metal oxide supports after continuous visible light (λ > 420 nm) for 10 h is determined to be 1.4 nm, indicating the critical role of BPEI on inhibiting the aggregation of Au GSH clusters and excluding the supports effect on the photostability enhancement of Au GSH clusters.

Lines 12-22 of Page 9:
Notably, it has been demonstrated that the CB edge potential of metal oxides (e.g., ZnO and rutile TiO 2 ) is more positive than the LUMO potential of Au GSH clusters, which enables the transformation of photoexcited electrons from Au GSH clusters to the metal oxide supports 15, 24, 53 . The photoactivities of BPEI modified metal oxides-Au GSH clusters composites (denoted as MAB) have been evaluated toward RhB degradation under visible light irradiation (λ > 420 nm). As shown in Supplementary Fig. 18, the samples of MAB exhibit moderate photoactivity enhancement than bare semiconductors, which suggests that the random loading of ultrasmall Au GSH clusters onto semiconductors without rational structure design is inefficient to achieve high efficient Au GSH clusters-semiconductor composites. Therefore, a thickness tunable TiO 2 shell has been coated onto the surface of SAB composites for designing core-shell SiO 2 -Au GSH clusters-BPEI@TiO 2 structures to construct high performance Au GSH clusters-semiconductor composites for solar energy conversion.

Corresponding revisions highlighted in red in the revised supplementary information:
Supplementary Figure 12 has been revised and Supplementary Figure 18 has been added. The discussion on the stability of Au GSH cluster on anatase TiO 2 has also been supplemented in the form of Appendix and Supplementary Figure A1 and A2 have also been added in the revised supplementary information.
Comments 6: Which is the Au+/Au(0) proportion of the Au GSH clusters determined by XPS?

Author reply:
Thanks for your valuable comments. As you kindly suggested, the Au + /Au 0 proportion of the Au GSH clusters has been calculated based on the XPS result in Fig. 1a and the Au 0 content is found to constitute ∼90% of all Au atoms.

Corresponding revisions highlighted in red in the revised manuscript:
Lines 30-31 of Page 7: The Au 0 content determined by XPS is found to constitute ∼90% of all Au atoms.
Comments 7: It is unclear why the authors use the prefix bio to indicate that the polyethylimine is reducible. Please clarify.

Author reply:
Thanks for your valuable comments. As you kindly suggested, the prefix "bio" has been deleted in our revised manuscript.

Corresponding revisions highlighted in red in the revised manuscript:
Lines 20-25 of Page 5: The SiO 2 spheres become positively charged (Supplementary Fig. 4a) after surface modification with branched poly-ethylenimine (BPEI) 36 , a conjugated reducible dendrimer, which leads to a substantial electrostatic attraction with negatively charged Au GSH clusters ( Supplementary Fig. 4c) by coulombic forces, thereby forming SiO 2 -Au GSH clusters-BPEI composites (SAB) with strong interfacial interaction between Au GSH clusters and SiO 2 supports Comments 8: It is indicated in the text that Au clusters could be used as visible light photosensitizer. A comment on the λmax absorption and other photophysical relevant properties of Au GSH as photosensitizer would be important.

Author reply:
Thanks for your valuable comments. As you kindly suggested, the photoluminescence (PL) excitation spectrum of Au GSH clusters has been performed ( Supplementary Fig. 1c), which shows a maximum at 400 nm and coincides well with the absorption shoulder (∼400 nm) observed in the absorption spectrum. Additionally, more discussions regarding the absorption and photophysical properties of Au GSH clusters have been supplemented in our revised manuscript, which faithfully confirms that the Au GSH clusters can act as a visible light photosensitizer.
UV-vis absorption spectrum of in Supplementary Fig. 1c suggests that the Au GSH clusters show an absorption onset at ∼520 nm with a distinct shoulder around 400 nm, which is attributed to the highest occupied molecular orbital-lowest unoccupied molecular orbital (HOMO-LUMO) transition originated from the ligand-to-metal charge transfer, indicating that Au GSH clusters could be used as visible light photosensitizers. The photoluminescence (PL) excitation spectrum exhibits a maximum at 400 nm, which coincides well with the absorption shoulder observed in the absorption spectrum. The PL emissive spectra of Au GSH clusters with different excitation wavelength in Supplementary Fig. 1f show a low energy emission band with the peak maximum at 605 nm, which is ascribed to the triplet metal-centered state, and the shape of the PL spectra is independent of the excitation wavelength. The large Stokes shift in the emission where the absorption band shoulder appears at around 400 nm and the emission maximum is seen at 605 nm, is consistent with the excited state being a ligand-metal charge transfer type.  Supplementary Fig. 1c suggests that the Au GSH clusters show an absorption onset at ∼520 nm with a distinct shoulder around 400 nm, which is attributed to the highest occupied molecular orbital-lowest unoccupied molecular orbital (HOMO-LUMO) transition originated from the ligand-to-metal charge transfer, 15 indicating that Au GSH clusters could be used as visible light photosensitizers. The photoluminescence (PL) excitation spectrum exhibits a maximum at 400 nm, which coincides well with the absorption shoulder observed in the absorption spectrum. The PL emissive spectra of Au GSH clusters with different excitation wavelength in Supplementary Fig. 1f show a low energy emission band with the peak maximum at 605 nm, which is ascribed to the triplet metal-centered state 32 , and the shape of the PL spectra is independent of the excitation wavelength 15 . The large Stokes shift in the emission where the absorption band shoulder appears at around 400 nm and the emission maximum is seen at 605 nm, is consistent with the excited state being a ligand-metal charge transfer type 15, 33 .

Page 20, References
Ref. 33 has been added. clusters to grow into large Au nanoparticles. As compared with the soluble antioxidant ascorbic acid, the BPEI with the multiple polymer chains of primary, secondary and tertiary amine groups can encapsulate the as-synthesized Au GSH clusters via crosslinking, which can fix these clusters on the surface of supports and hamper the migration of clusters, thereby stabilizing the ultrasmall Au GSH clusters.

Comment 2:
A last comment is that I do not agree that growing from 1.2 to 2.0 or 2.4 nm is a small particle size increase. What is important is the relative particle size increase and this change is substantial indicating a change from a cluster to nanoparticles. Author reply: Thanks for your valuable comments. We agree with the reviewer that the increase of particle size suggests the change from clusters to nanoparticles. As you kindly suggested, we have revised our manuscript with great attention.

Corresponding revisions highlighted in red in the revised manuscript:
Lines 27-35 of Page 6 and Lines 1-2 of Page 7: When the irradiation time over SAB is further prolonged to 36 h and 48 h, the size of Au GSH clusters will increase to 2.0 nm and 2.1 nm, respectively, as illustrated in Supplementary Fig. 9, which indicates the slight aggregation of Au GSH clusters. Such size increasement of Au GSH clusters may be attributed to the partial depletion of BPEI since the BPEI layer that serves as a reducing agent would undergo oxidation and/or decomposition in the way to stabilize the Au GSH clusters. To further investigate the role of BPEI in inhibiting the growth of Au GSH clusters, the SAP sample is subsequently modified by BPEI (BPEI-SAP) and the obtained BPEI-SAP composite is irradiated under visible light (λ > 420 nm) for 10 h. The HRTEM image of BPEI-SAP after light illumination in Supplementary Fig. 10a suggests that, as compared with SAP samples (Fig. 1b-d), the serious aggregation of Au GSH clusters could be prevented to some extent and the size of Au GSH clusters is calculated to be 2.0 nm (Supplementary Fig. 10b).