Gold catalysts containing interstitial carbon atoms boost hydrogenation activity

Supported gold nanoparticles are emerging catalysts for heterogeneous catalytic reactions, including selective hydrogenation. The traditionally used supports such as silica do not favor the heterolytic dissociation of hydrogen on the surface of gold, thus limiting its hydrogenation activity. Here we use gold catalyst particles partially embedded in the pore walls of mesoporous carbon with carbon atoms occupying interstitial sites in the gold lattice. This catalyst allows improved electron transfer from carbon to gold and, when used for the chemoselective hydrogenation of 3-nitrostyrene, gives a three times higher turn-over frequency (TOF) than that for the well-established Au/TiO2 system. The d electron gain of Au is linearly related to the activation entropy and TOF. The catalyst is stable, and can be recycled ten times with negligible loss of both reaction rate and overall conversion. This strategy paves the way for optimizing noble metal catalysts to give an enhanced hydrogenation catalytic performance.

making the catalyst this way they make a material that has interstitial C in Au nanoparticles supported on carbon. They then carry out a standard model hydrogenation reaction (so there is no novelty in the catalysed reaction). The catalysis they observe with their catalyst is certainly interesting. But teh main claim is that their catalyst is far better than gold supported on carbon, silica or titania. But teh comparison is with three commercial catalysts and we have no idea how they are made and what they comprise. For a paper at this level this is unacceptable. They need to prepare there own Au comarison catalysts so that the particle sizes are similar and then do a comparison. The loss of gold can be a real problem and they need to show teh data for Au retained on teh catalyst after use. Typically quite a lot can be solubilised but it plates out on teh reactor and so is not observed in solution. Finally teh English needs attention. I am not commenting on the microscopy or the DFT calculations Reviewer #3 (Remarks to the Author): The authors claim to have supposedly made a catalyst material comprised of gold particles embedded in mesoporous carbon with carbon atoms occupying interstitial sites in the gold lattice. Furthermore, they claim his interstitial doped catalyst is superior to Au/TiO2 for the chemoselective hydrogenation of 3nitrostyrene.
Several things concern me about the microstructural characterization aspects of this paper.
Firstly, the Au Particle size distributions are not 'monodisperse' as repeatedly claimed in the text as proven by their own particle size distributions.
Secondly, it has not been proven that the Au particles are actually within the pores of the activated carbon spheres. TEM/STEM is a 2D projection technique unless applied in tomography mode. Indeed, several of the supplementary figures show gold particles in profile on the C surface. Furthermore, if the gold particles were really embedded in the interior of the activated C spheres how do the generated photoelectrons escape in the XPS experiments?
Finally, and even more concerning to me, is the lack of any convincing evidence that C exists as an interstitial species in the f.c.c. Au lattice. Measurement of Au (111) spacings by STEM HAADF imaging over six or so lattice planes to a 0.01A accuracy is simply not a realistic or feasible proposition. The authors should carefully evaluate the intrinsic errors associated with making such measurements.
Furthermore, even a (very high) 2at% C doping level in a 2nm Au particle would still only mean the incorporation 1-2 interstitial C atoms at most. How would this cause such a uniform expansion over (111) planes in a such 2nm Au nanoparticle? Response 4. We agree with the reviewer after re-checking references. The lattice contraction is highly related to a high proportion of outer shell undercoordinated atoms with respect to fully coordinated inner shell atoms. We have accordingly revised our manuscript.
Revised manuscript, Page 5: "It is well-known that lattice contraction occurs in small nanoparticles or clusters, and is induced by the large surface/volume ratio, ……" has been changed to "It is well-known that lattice contraction occurs in small nanoparticles or clusters, and is related to a higher proportion of outer shell undercoordinated atoms with respect to fully coordinated inner shell atoms, ……".

Comment 5.
Major: The binding energy (EB) is claimed to go linear with the coordination number, but should go rather on the 'surface/volume' ratio, express as the lowered-dimensionality reduced particle radius, r -1/3 , as many other properties do? See https://doi.org/10.1039/C3CS60421G.

Response 5.
We agree with the reviewer. The shift of the binding energy is attributed to the particle size effect, in good agreement with the fact that EB shifts to a higher energy for clusters compared to that of the bulk metal, and the change is much related to the surface-to-volume ratio. Accordingly, we have revised our manuscript.
Revised manuscript, Page 8, several references have been added: "This shift is attributed to the particle size effect, in good agreement with the fact that EB shifts to a higher energy for clusters compared to that of the bulk metal and the change is a linear function of the average coordination number." has been changed to "This shift may be attributed to the particle size effect which is a final state effect according to the size-dependent electrostatic interaction between the cluster and an escaping photoelectron. It has also been reported that there is a negative core level shift for the surface atoms of macroscopic Au due to an initial state effect, which is 6s → 5d charge reorganization for the more undercoordinated surface atoms; but for nanoclusters, the initial state shift is mainly overcompensated by the electrostatic final state effect 41 . The present positive shift is in good agreement with the fact that there are shifts to a higher energy for clusters compared to that of the bulk metal. The change is determined by the surface-to-volume ratio, and the photoemission onset is influenced by an initial state effect involving charge transfer 40,42 .". and explain what would be the low-intensity grey signal in between such when decomposed. Have authors analysed the C 1s? A recent study on 3s levels on Au suggest that having interstitial C shifts the BE to higher values, see https://doi.org/10.1002/anie.201813037, which seems to go along the present XPS data.

Response 6.
(1) We agree that the EB shifts are small. For the XPS studies, a Al Kα source was used. A pass energy of 40 eV and a step size of 100 meV were used for a survey scan. An experimental resolution of 0.5 eV has been fitted from the Ag 3d 5/2 bulk peak. For a detailed analysis, the core-level lines obtained by XPS were numerically fitted by a convolution of a Gaussian and a Lorentzian profile with an additional parameter allowing asymmetry of the line, and the data was calibrated using the C 1s binding energy of 284.6 eV.
(2) Such a small change has also been observed by Huang and co-workers 1 . The Au 4f 7/2 binding energy shifts from 84.13 eV for 1.0 Å-thick Au film on a reduced TiO2(110) surface, to 84.06 eV for 1.5 Å Au and finally to 84.13 eV for 2.0 Å Au. The authors pointed out that this was the first experimental evidence for the charge transfer from the defective sites on the reduced TiO 2 (110) surface to Au clusters. However, as mentioned by the reviewer, the unequivocal conclusion on such shifts may be exaggerated. In ordered to not mislead the readers, we have revised our manuscript to avoid this.
(3) The low-intensity grey signals in the XPS spectra of the Au 4f level belong to Au + and Au 3+ which may originate from undercoordinated sites (Fig. 1a).
(4) We have analyzed the C 1s spectra as shown in Fig. 1b. But it should be mentioned that the carrier is a kind of activated carbon, which will have an effect on the C 1s spectra. Therefore, we did not discuss the results in the manuscript.
(5) We fully agree with this comment. To estimate the changes in the electronic structure of Au on the adsorption or absorption of C, the detection of a difference between the 3s core levels of metal centers around C and the corresponding centers in pristine gold before C insertion provides direct evidence. The given reference provides a calculation of the core levels of surface and subsurface atoms bound to C on Au(111) surfaces, and suggests that the 3s level of Au containing interstitial C shifts the BE to a higher value. However, to detect the 3s level of Au, high energy XPS with Cr Kβ (hν = 5946.7 eV) is required to measure electron attenuation lengths over a kinetic energy range of ~2500 to ~5900 eV. The availability of this instrument is very limited in China (maybe fewer than 3). We contacted the owners of the instruments and were told that the 3s level of a Au foil is undetectable with a commercial high energy XPS instrument. For a special measurement of the 3s level of Au, a substrate overlayer method is used in conjunction with synchrotron radiation (hν = 1800-4600 eV) excited XPS 2 . Currently this method and this instrument is not available for our samples. Therefore, we have discussed the 3s level of Au by DFT calculations in the revised manuscript, and leave it an open question for future study. We have revised our manuscript accordingly.
Revised manuscript, Page 8: (1) "This distinctive reversal of the binding energy shift in the Au 4f doublet as a function of Au particle size has been observed on Au/TiO2, and assigned to a combined contribution of the charge transfer from surface to clusters and the particle size effect 14 ." has been changed to "This distinctive reversal of the binding energy shift in the Au 4f doublet as a function of Au particle size has been observed in Au/TiO2 with similarly small changes, and may be assigned to a combined contribution of the charge transfer from surface to clusters, the initial state effect and the electrostatic final state effect 15 .".
Revised manuscript, Page 7, a sentence has been added: (2) " Fig. 2a shows measured and fitted Au 4f XPS spectra for Au nanocatalysts, and the major peaks for Au 0 in all can be fitted with minors of Au + and Au 3+ which may originate from undercoordinated sites 38 .".
Revised manuscript, Page 8: (4) "In the present case, C diffusion that leads to the charge transfer either from the carbon atom or from the s, p electron redistribution of Au would be reasonable, and this effect is dominant for the charge gain." has been changed to "In the present case, C diffusion that leads to a rearrangement of electron density may be dominant for the charge transfer. Viñes et al. found that the 3s core levels of Au around C shifts the binding energy to a higher value compared to a pristine Au(111) surface with no interstitial C, which is an evidence for the redistribution of orbitals 28 . The high energy XPS should be be used to give more distinct and direct evidence on interstitial C in future studies. Here a charge transfer either from the carbon atom or from the s, p electron redistribution of Au may be reasonable.".

Comment 7.
Major: The shifts on the EB can have multiple origins. Considering size effect, the smaller, the more atom-like are the NPs, and less diffuse electrons, which would make the EB to increase. However, the charge transfer to (from) Au could decrease (increase) it. However, when having larger surface/volume ratio, one has more undercoordinated sites, and such are known to reduce the BE, as they accumulate there charge density (see doi:10.1007/s11426-010-0086-z). The discussion should be reconsidered, as, being many factors, one cannot isolate them and rule most out so to explain solely based on one.

Response 7.
We agree with the comment that the shifts in binding energy can have multiple origins. In the case of Au, as mentioned by the reviewer, the negative core level shift for the surface atoms of macroscopic Au can be explained by an initial state effect due to 6s → 5d charge reorganization for more undercoordinated surface atoms which accumulate their charge density 1 . Such intraatomic charge transfer causes a smaller binding energy for less coordinated Au atoms than fully coordinated bulk atoms. For the nanoclusters, the systematic energy shift can be explained by a final state effect according to the size-dependent electrostatic interaction between the ionised cluster and escaping photoelectron. The initial state shift is mainly overcompensated by the electrostatic final state effect. As a result, the binding energy increases with the surface/volume value as highlighted in comment 5, and the photoemission onset is influenced by an initial state effect involving charge transfer 2 . The effect of the undercoordinated sites should be considered. Accordingly, we have revised our manuscript. Revised manuscript, Page 7, a sentence has been added: "There are many possible reasons for the shifts in binding energy.".

Revised manuscript, Page 7:
"This shift is attributed to the particle size effect, in good agreement with the fact that EB shifts to a higher energy for clusters compared to that of the bulk metal and the change is a linear function of the average coordination number" has been changed to "This shift may be attributed to the particle size effect which is a final state effect according to the size-dependent electrostatic interaction between the cluster and an escaping photoelectron. It has also been reported that there is a negative core level shift for the surface atoms of macroscopic Au due to an initial state effect, which is 6s → 5d charge reorganization for the more undercoordinated surface atoms; but for nanoclusters, the initial state shift is mainly overcompensated by the electrostatic final state effect 41 . The present positive shift is in good agreement with the fact that there are shifts to a higher energy for clusters compared to that of the bulk metal. The change is determined by the surface-to-volume ratio, and the photoemission onset is influenced by an initial state effect involving charge transfer 40,42 .".

Response 8.
Thank you for this comment. C diffusion in the Au lattice may lead to the rearrangement of electron density that may be dominant for the charge transfer. As mentioned by the reviewer, charge transfer from C to gold or gold to C is possible since other factors for example the s, p electron redistribution of Au, are combined together. All these factors should be discussed. We have revised our manuscript accordingly.
Revised manuscript, Page 8, a reference has been added: "In the present case, C diffusion that leads to the charge transfer either from the carbon atom or from the s, p electron redistribution of Au would be reasonable, and this effect is dominant for the charge gain." has been changed to "In the present case, C diffusion that leads to a rearrangement of electron density may be dominant for the charge transfer. Viñes et al. found that the 3s core levels of Au around C shifts the binding energy to a higher value compared to a pristine Au(111) surface with no interstitial C, which is an evidence for the redistribution of orbitals 28 . The high energy XPS should be be used to give more distinct and direct evidence on interstitial C in future studies. Here a charge transfer either from the carbon atom or from the s, p electron redistribution of Au may be reasonable.". possibly to some C-free areas of the surface 1 . It could open another channel to move C to all possible sites in the lattice. The H adsorption would be further improved due to the increasing concentration of C on surface. In addition, the dynamics of the C movement to the surface or its insertion back to the Au matrix may also inhibit the accumulation of the organic substances, and enhance the reusability. In addition, a continuous exchange of C atoms between surface and subsurface translates into a dynamic equilibrium for Au. As a result, the dynamics of C movement to the surface or its re-insertion to the Au matrix may be responsible for a uniform expansion over the (111) plane in a such a 2 nm Au nanoparticle. We have revised our manuscript accordingly.
Revised manuscript, Page 6: Several sentences have been added: "It should be mentioned that the observed lattice expansion is small possibly due to the relatively low C solubility and the intrinsic lattice contraction in small size particles.
However, a continuous exchange of C atoms in Au between surface and subsurface has been reported so that a dynamic equilibrium is established. The mobility of carbon in the lattice may cause residual vacancies and in turn produce a relatively uniform lattice expansion 38 .".  Table 4). The locations of carbon atoms on the surface and subsurface were comparatively investigated for clean and C-containing Au(111) and Au(211) surfaces and subsurfaces (Supplementary Tables 5   and 6). The isolated molecules and atoms were optimized in a box of 20 Å×20 Å×20 Å with the same criterion as those for geometric optimization of the Au surfaces. The adsorption energies were calculated by Eads = Eadsorbate/surface -Esurface -Eadsorbate, where Eadsorbate/surface, Esurface and Eadsorbate are respectively the total energy of a surface covered with an adsorbate, the clean surface slab, and an isolated adsorbate. The transition states of hydrogen dissociation on the surfaces were located by the dimer method 12 where the criteria for the convergence of electronic, and ionic relaxations were set to be 10 -6 eV and 0.025 eV Å -1 , respectively. The geometrical structures for the transition states were confirmed by subsequent vibrational calculations carried out using the numerical finite difference method, which generated only one imaginary frequency for a specific transition state.  Step octahedral -3.50 Low tetrahedral -4.26 High tetrahedral -4.40 Step tetrahedral -3.94

Reviewer #2
Comment. This is a potentially interesting paper but the catalysis needs a lot of attention. The claim is that by making the catalyst this way they make a material that has interstitial C in Au nanoparticles supported on carbon.
They then carry out a standard model hydrogenation reaction (so there is no novelty in the catalysed reaction). I am not commenting on the microscopy or the DFT calculations.
Response. We thank you for reviewing our work on a new catalytic material that has interstitial C in Au nanoparticles supported on carbon. The novelty is the formation of interstitial C in the Au lattice, and the charge transfer between the C and Au which has not been reported to the best of our knowledge. A standard model hydrogenation reaction was chosen because a very weak catalytic hydrogenation activity was observed over gold loaded on a less-active support such as carbon and silica in the literature. Therefore, it is acceptable that the high activity and selectivity for the present interstitial catalyst originates from the charge transfer between interstitial C and Au. We are now using our novel catalysts in some challenging reactions, and shall report the results in the future.         Actually, we also synthesized Au/TiO2, Au/SiO2 and Au/C reference samples with our mesoporous carbon carriers using the deposition-precipitation or deposition-reduction method. TEM images of the synthesized catalysts are shown in Fig. 7. The catalytic performance in the selective hydrogenation of 3-vinylaniline for these catalysts was similar to the customized ones supported on the same carrier, and also compared to the well-documented literature. For example, the TOF value for our mesoporous Au/TiO2 was estimated to be 226 h -1 , and both Au/C and Au/SiO2 were almost inactive to this reaction. However, the synthesis of these samples was complicated. It has been reported that the nanoparticle sizes and crystal phases of transitional oxide carriers will for certain affect the catalytic performance of gold nanoparticles. Since commercial products are easily obtained for comparison, we decided to purchase customized samples for our research.

Comment 2.
The loss of gold can be a real problem and they need to show the data for Au retained on the catalyst after use. Typically quite a lot can be solubilised but it plates out on the reactor and so is not observed in solution.

Response 2.
We fully agree with the reviewer that the loss of gold can be a real problem for a liquid reaction.
Several tests have been performed to confirm the negligible leaching of gold for the present catalyst.
(1) A solid quenching test: A solid quenching test was performed using mercapto functional group-containing mesoporous silica SH-SBA-15 as the trapping agent. Once the metal species leach into solutions, mercapto functional group will trap them, and quench the catalysis by leaching metal species in liquid reactions. The unchanged hydrogenation activity and selectivity in the presence of SH-SBA-15 demonstrates there is no leaching of gold with the present C-Au/OMC catalysts regardless of Au particle size, and the solubilized gold in solution during reaction and redeposition on the carrier surface after reaction can be excluded (Fig. 1).
(2) Reusability: Both the initial reaction rate (r0) and turn over number (TON) in ten successive runs remained almost the same (Fig. 2), indicating the number of active centers does not significantly change.
(3) XPS spectra: The XPS spectrum was taken for the re-used catalyst, and shows indistinct changes compared with the fresh catalyst (Fig. 3).
(4) TEM image: the average Au particle size for the re-used catalyst also remained almost unchanged, as shown by TEM images (Fig. 4).
(5) Au concentration measurement: The Au concentration in the reused catalyst after 10 runs remained at 0.91 wt%, similar to that determined for the fresh catalyst.
In these cases, the leaching of gold nanoparticles can be excluded. A surface reaction on the Au nanoparticles instead of a reaction in solution catalyzed by leached Au is the best model for the present reaction.    The paragraph has been re-written: "The leaching of gold species from the solid catalyst which can catalyze the hydrogenation has been argued by some researchers. A solid quenching test was performed using mesoporous silica SH-SBA-15 containing mercapto functional groups as the trapping agent. Once the metal species leach into solution, mercapto functional group will trap them, and quench the catalysis by leaching metal species in liquid reactions 58 . The unchanged hydrogenation activity and selectivity in the presence of SH-SBA-15 demonstrates there is no leaching of gold with the present C-Au/OMC catalysts regardless of Au particle size, and solubilized gold in solution during the reaction and redeposition on the carrier surface after the reaction can be excluded ( Supplementary Fig. 16). The reusability of C-Au-2.4/OMC was also tested. Both the initial reaction rate (r0) and turn over number (TON) in ten successive runs remained almost the same ( Supplementary Fig. 17), indicating the number of active centers does not significantly change. The re-used catalysts were characterized.
The XPS spectrum shows no distinct changes compared with the fresh catalyst. The Au concentration in the reused catalyst after 10 runs remained at 0.91 wt%, also similar to that determined for the fresh catalyst. In addition, the average Au particle size for the re-used catalyst also remained almost unchanged, as shown by TEM images (Supplementary Fig. 18).".

Comment 3.
Finally the English needs attention.

Response 3.
Thanks for this comment. The manuscript has been revised by a professor who is a native English speaker, and some typos have been corrected.

Reviewer #3
Comment. The authors claim to have supposedly made a catalyst material comprised of gold particles embedded in mesoporous carbon with carbon atoms occupying interstitial sites in the gold lattice. Furthermore, they claim this interstitial doped catalyst is superior to Au/TiO2 for the chemoselective hydrogenation of 3-nitrostyrene.
Several things concern me about the microstructural characterization aspects of this paper.

Response.
We thank you for the comment.

Comment 1.
Firstly, the Au Particle size distributions are not 'monodisperse' as repeatedly claimed in the text as proven by their own particle size distributions.

Response 1.
We thank for this comment, and have removed the description of "monodisperse" in the whole manuscript.
Revised manuscript, Page 4: "Here, monodispersed gold nanoparticles with carbon atoms occupying interstitial sites in the lattice (C-Au)" has been changed to "Here, uniform gold nanoparticles with carbon atoms occupying interstitial sites in the lattice (C-Au)".
Revised manuscript, Page 5: "Monodispersed nanoparticles with uniform sizes can be clearly observed in large domains in the high-angle annular dark-field scanning transmission electron microscope (HAADF-STEM) images ( Supplementary Fig.   4)." has been changed to "Nanoparticles with uniform sizes can also be clearly observed in large areas in high-angle annular dark-field spherical aberration corrected-scanning transmission electron microscope (HAADF-ACSTEM) images (Figs. 1c,d and Supplementary Fig. 4).". In addition, we found that if the nanoparticles are partially exposed to the pore space, the generated photoelectrons can escape in the XPS experiments but the detected concentration is much lower than the ideal concentration 2 . Similar results have been found for the present catalysts. Therefore, the nanoparticles are partially embedded into the mesopore walls with partial exposure to the pore space.
Second, we fully agree that TEM/STEM is a 2D projection technique. We conclude that the gold nanoparticles are located inside the mesoporous carbon spheres instead of being stuck to the outer shell for the following reasons: Interestingly, partially exposed nanoparticles can be clearly observed. Similar results have also been found in the present Au nanoparticles in ordered mesoporous carbon (Fig. 3).
Last but not least, the outer surface area of the sphere is as low as 23 m 2 /g. On consideration of the coordination-assisted self-assembly approach and the high surface areas of mesoporous structure of the carbon sphere (594 m 2 /g), we conclude that the gold nanoparticles are located inside the structure during the assembly and carbonization. Gold nanoparticles are partially embedded in the carbon pore walls, and partially exposed to the pore space.

Comment 3.
Finally, and even more concerning to me, is the lack of any convincing evidence that C exists as an interstitial species in the f.c.c. Au lattice. Measurement of Au (111) spacings by STEM HAADF imaging over six or so lattice planes to a 0.01 Å accuracy is simply not a realistic or feasible proposition. The authors should carefully evaluate the intrinsic errors associated with making such measurements.
Response 3. We appreciate this comment. As mentioned in the literature 1 , the ex situ surface and subsurface C detection on Au regular surfaces is a challenging task. The in situ detection, for example ambient pressure X-ray photoemission spectroscopy (APXPS) may be useful for the detection of such C species in noble-metal systems.
However, for the nanoparticles, we have made several attempts, and the lattice expansion and electronic structure are the most repeatable attempts. The lattice expansion is reliable to some extent because: (1) It is well-known that lattice contraction occurs in small nanoparticles or clusters, and is related to a higher proportion of outer shell undercoordinated atoms with respect to fully coordinated inner shell atoms which is proportional to the reciprocal of the particle size (Table 1). For example, about 3.0% contraction is found for an Au nanoparticle with 1.6 nm in size (in comparison with the bulk Au with a d(111) of 2.350 Å) 2 . Therefore, the observed small lattice expansion may be evidence for the interstitial C. The smaller expansion of the C-Au catalyst compared to bulk Au is possibly due to the overall relatively low C solubility and the intrinsic lattice contraction in small nanoparticles. (2) A continuous exchange of C atoms in Au between surface and subsurface has been reported so that a dynamic equilibrium is established. The mobility of carbon in the lattice may cause residual vacancies and in turn a relatively uniform lattice expansion 3 . The final state for the dynamic C moieties would be to aggregate in graphite or amorphous carbon phases, similar to our experimental observations 1 .
We must keep in mind that direct evidence that C exists as an interstitial species in the f.c.c. Au lattice is very important and deserves further thorough investigation. For example, the detection of a difference between the 3s core levels of metal centers around C and the corresponding centers in pristine gold before C insertion gives direct evidence.
The solubility (S) of C in a particle with radius r is estimated to be = 0 (2 / ) where S0 is the C solubility in bulk Au, defined as the ratio of the amounts of solute and solvent, is the surface tension, V is the volume of a metal atom, k is the Boltzmann constant, and T is the melting temperature in bulk phase. Here, S0 = 0.08 at%, = 1.42 J m -2 , V = 1.69×10 -29 m 3 , k = 1.38×10 -23 J K -1 , T = 1337.58 K and r = 0.8, 0.95, 1.2, 1.95 and 4.5 nm. For σ and S0, the values in refs [2] and [3] were used.
The calculation used here is to show the significant increase of C solubility in the Au lattice with a decrease in particle diameter. The ideal solubility may not exactly follow the calculation. For example, if we use this formula to calculate the C solubility in Pd, it will be about 1.77 at% for a particle with a diameter of 2.5 nm.
The S0 value was taken from reference [4]. However, several reports have found about 10 at% C can be inserted into the Pd lattice 5 . Therefore, the C doping concentration in Au may not be as low as indicated by the calculations. To avoid this misunderstanding, we have removed the detailed data on C solubility.
(2) We fully understand your doubts about the insertion of C in the Au lattice. As mentioned by Reviewer #1, there is a rapid transition between the surface and subsurface C atoms 6 . A continuous exchange of C atoms between the surface and subsurface translates into a dynamic equilibrium for Au. The presence of subsurface C would facilitate the mobility of carbon in the lattice and possibly to some C-free areas of the surface. The final state for the dynamic C moieties would be to aggregate in graphite or amorphous carbon phases, similar to our experimental observations. Therefore, the dynamics of the C movement to the surface or its re-insertion in the Au matrix may be responsible for a uniform expansion over (111) planes in a such 2 nm Au nanoparticle.