Gallium nitride catalyzed the direct hydrogenation of carbon dioxide to dimethyl ether as primary product

The selective hydrogenation of CO2 to value-added chemicals is attractive but still challenged by the high-performance catalyst. In this work, we report that gallium nitride (GaN) catalyzes the direct hydrogenation of CO2 to dimethyl ether (DME) with a CO-free selectivity of about 80%. The activity of GaN for the hydrogenation of CO2 is much higher than that for the hydrogenation of CO although the product distribution is very similar. The steady-state and transient experimental results, spectroscopic studies, and density functional theory calculations rigorously reveal that DME is produced as the primary product via the methyl and formate intermediates, which are formed over different planes of GaN with similar activation energies. This essentially differs from the traditional DME synthesis via the methanol intermediate over a hybrid catalyst. The present work offers a different catalyst capable of the direct hydrogenation of CO2 to DME and thus enriches the chemistry for CO2 transformations.

following comments: 1. Abstract (line 20). The word "selectivity" is duplicated in the sentence: "… with a CO-free selectivity selectivity as high as …".
2. Lines 59-60. Please, revise the following statement: "Thus, taking into these properties … to DME account" (probably "Thus, into account these properties …" is the correct expression).
3. For the best GaN catalyst at optimum reaction conditions the authors report a STY of DME of 2.9 mmol/(g·h). In order to put this productivity value into context, it would be interesting to compare it with that obtained for some of the best-performing hybrid systems (e.g., Cu-based catalyst + zeolite) reported in the literature (such a comparison could be included in the Supporting Information). 4. Lines 101-102. "…, we synthesized bulk GaN with different crystal sizes by calcining the mixture of gallium nitrate and melamine." I would suggest, for the sake of clarity, to specify here that different crystal sizes were produced by changing the duration of the calcination treatment, as one may infer from the experimental section. 5. Line 115. "… (FID), DME is exclusively the main product." Since DME is not the only product detected by the FID, I suggest to remove the word "exclusively" to avoid confusion. 6. In line 123, it is mentioned the existence of an "induction period" of about 12 h before a stable performance is achieved. In Figure 4, however, one can see significant fluctuations in both CO2 conversion and product selectivities during the first 12 h of reaction. Can the authors discard analytical errors as the origin of this anomalous behavior? It seems questionable to me that this is a true "induction period" since, according to XPS, the nature of Ga and N species was the same in the fresh and spent samples (Fig. S1, Table S1). 7. Lines 159 and 274. "Absorption" should be replaced by "adsorption". 8. Table S3. The sum of product distributions does not always adds up to 100%. Please, check.
Reviewer #3: Remarks to the Author: The paper of Liu and co-workers describes a catalytic system for the CO2 reduction to DME. It is an extensive, multipart study, in which authors use different techniques to shed light on the efficiency and the mechanism of the process.
My main objection is the organization of the manuscript. The present version is very difficult to follow, and in my opinion it is because the authors present the full research paper in the form of a short communication. A signifincant portion of the results has been moved to the supplementary information, and there are too many references in the manuscipt to the supplement. In this case, the manuscript cannot stand on its own -the supporting information is often essential in the discussion. For instance -the manuscript does not even contain even the most basic information on the computational part (such as the functional used). My advice is to either rewrite the manuscript with the better suited form in mind, with the divisions into chapters and sections -as this will greatly improve the clarity of presentation; or focus the description on a strong point and supporting evidence.
Another issue is the analysis of the computational part. While the values of the adsorption energies bring useful information to the topic of study (although the observations are sometimes counterintuitive), the mechanistic study is very much incomplete. The main conclusion seems to be the competitive pathways via carboxyl and formate lead to the same -CH3 and -OH intermediates coadsorbed, regardless on the pathway. This does not explain the mechanism in any way, and in my view it only makes the computational part of the study look like it is an unnecessary addition. No other hydrogenation steps have been investigated, no information on the different modes of adsorption of the intemrediates has been provides, no analysis of the different character of adsorbed hydrogens has been carried out.
Contrary to that, the experimental part seems convincing and carried out with care. Sometimes it makes the impression of being too extensive -for instance omission of the paragraph devoted to carbonate promoters would not lead to the lesser scientific value of the research, and in my view it would only increase the clarity of presentation mentioned above.
Overall, the paper seems to be an example of those that carry too much information instead of sending one clear message. The manuscript has a potentnial, and the description of the reactivity of the GaN system is a valuable contribution to the field, but more work is needed to make it meet the standard.
Comment: Some statements are also to be revised, like at line 49 the claim according to which "metallic Cu catalyzes the hydrogenation of CO 2 to methanol" (there is no consensus about the active phase of this step!!!).
Response: Thank you and we agree with you. Indeed, the metallic Cu (Cu 0 ) is not unambiguously accepted as the only active phase for the CO 2 hydrogenation over the Cu-based catalysts although the Cu 0 phase has been revealed to be one of the most active sites for the CO 2 hydrogenation as reviewed in the reference (Wang, W., Wang, S., Ma, X. & Gong, J. L. Recent advances in catalytic hydrogenation of carbon dioxide. Chem. Soc. Rev. 40, 3703-3727 (2011)). For accuracy, the sentence is modified as "the metal based catalysts such as Cu/ZnO/Al 2 O 3 catalyze the hydrogenation of CO 2 to methanol..." (Lines 48-49 in the revised manuscript with yellow highlights). Moreover, the expression is applied in the cases throughout the manuscript, which is also highlighted in yellow color.

Comment:
Besides, if a direct hydrogenation process is carried out from CO 2 to DME, it is unanimously demonstrated that the activity-selectivity pattern of the catalyst cannot be considered as a mere linear combination of its performance in each of the two steps involved in the process (methanol synthesis from CO 2 and methanol dehydration into DME). From this point of view, the runs performed by co-feeding methanol and H 2 O at atmospheric pressure are to be reconsidered.

Response:
Thank you, and we accept your recommendation, which new experiments by using methanol or methanol + H 2 O as the reactants are performed, respectively. Indeed, the two steps involved in the reaction of CO 2 -to-DME over the typical hybrid catalysts should not be linearly combined over the GaN catalysts. Therefore, we investigate the reaction pathway of CO 2 hydrogenation over the GaN catalysts through catalytic evaluations under different W/F ( Figure 6a) and TPSR (Figure 6b). These results demonstrate that DME is the primary product and methanol is the secondary product, of which a different reaction pathway is proposed based on the DFT calculations.
To directly confirm the source of methanol as a secondary product, the experiment using DME and H 2 O as reactants is performed to realize the hydrolysis of DME to methanol (CH 3 OCH 3 + H 2 O = 2CH 3 OH), and the results are given in Figure R1a as follows and Figure S8a in the revised SI. At a reaction temperature of 320 ºC, methanol is the only product, indicating that the hydrolysis of DME occurs exclusively over the GaN catalyst. However, when the reaction is performed at a high temperature of 360 ºC, a comparable selectivity of methanol and CO 2 is obtained together with the production of H 2 (Not show), indicating that both the hydrolysis of DME and the steam reforming of DME (SRD, CH 3 OCH 3 + 3H 2 O = 2CO 2 + 6H 2 ) occur in a comparable extent.
To unambiguously confirm any other possibilities by combining your recommendation and our  Figure R1b), irrespective of the reaction temperatures, DME from the dehydration of methanol (2CH 3 OH = CH 3 OCH 3 + H 2 O) is always the main product. Moreover, a higher reaction temperature increases the selectivity of both CO and CO 2 . Thus, the steam reforming of methanol (SRM, CH 3 OH + H 2 O = CO 2 + 3H 2 ) and the decomposition of methanol (CH 3 OH = CO + 2H 2 ) occurs significantly at a higher reaction temperature, which is confirmed from the simultaneous release of H 2 . On the contrary, when methanol and H 2 O are co-fed as the reactants ( Figure   R1c), the very low selectivity of DME of 0.20 and 0.16% is obtained at the reaction temperatures of 320 and 360 ºC, respectively. Moreover, CO and CO 2 are the main products irrespective of the reaction temperatures, and the selectivity of CO is significantly increased with increasing the reaction temperature.
In fact, supported Cu is a good catalyst for SRM and the hybrid of supported Cu and a solid acid is also a good catalyst for the SRD reaction, the extent of which is strongly dependent on the specific catalyst and

Comment:
The presentation of many catalytic results is poor, not being well clear the need for removing CO among the formed products. It is obvious that its removal from the product distribution dramatically increases the selectivity of the other compounds.
Response: Sorry for this. In fact, CO is produced via the reverse water-gas-shift (RWGS) reaction. Moreover, the CO 2 hydrogenation and the RWGS are parallel reactions over the GaN catalysts as revealed from much low activity for the CO hydrogenation ( Figure 5). This is also supported from the TPSR results given in Figure 6b, i.e., the further conversion of CO hardly occurs (corresponding discussions please see in lines 213-216). Therefore, the selectivity of products involving CO can indicate the catalytic performance via the different pathways, i.e. the direct CO 2 hydrogenation or the RWGS reaction. However, the RWGS reaction does not significantly affect the distributions among the hydrogenated products, i.e. methane, DME, methanol, etc. Thus, to calculate the distributions in hydrogenated products, the RWGS reaction should be excluded. Based on these considerations, the selectivity of different products is expressed including CO while the product distribution is presented without CO. For clarity and conciseness, the word "CO-free" is added in the cases without considering CO in the products in the revised manuscript, which is highlighted in yellow color.
Comment: Nevertheless, a thermodynamic reference is missing in the Figures, considering that the range of temperature is higher than that typically operated (<280 °C). Yet, the differences of the H 2 /CO 2 ratio (2 instead of 3) obviously result in dramatic changes which require thermodynamic references. In these sense, it is not sure that the reported STY values are really valuable. Comparison with other catalytic systems tested under similar conditions is needed.
Response: Thank you. Indeed, the optimized reaction temperature for the typical Cu-based hybrid catalysts, is 240-280 °C, which is the balance of the thermodynamics limitation and the sintering of Cu at higher temperatures. Thermodynamically, as reported in the references, e.g., Chem. Eng. Technol. 37 (2014) 1765-1777, there exists a minimum CO 2 conversion with increasing the temperature, i.e., first decrease followed by the increase of the CO 2 conversion, the extent of which is clearly dependent on the H 2 /CO 2 ratio and the pressure as given in Figure R2. Moreover, our calculations by using Aspen plus ( Figure R3) are consistent with the reported results. Under the conditions of 2.0 MPa and 360 °C, the equilibrium CO 2 conversion is 31.1% and 26.6% with a H 2 /CO 2 molar ratio of 3 and 2, respectively (Referring Figure R2 and Figures R3b-c). For the reaction results over GaN catalysts as given in Figure 3 in the manuscript, the maximum CO 2 conversion under the reaction conditions of 2.0 MPa and 360 °C is 16.8% and 10.6% with a H 2 /CO 2 molar ratio of 3 and 2, respectively, which is sufficiently lower than that of the equilibrium CO 2 conversion. That is why there are still some reports on the CO 2 hydrogenation to DME over the typical Cu-based catalysts operated at the reaction temperatures of higher than 280 °C (Please see Table R1). In these cases, the increasing temperature leads to the significant decrease in the selectivity of DME and clear increase in the selectivity of hydrocarbons over the typical Cu-based catalysts. Based on your comments, we gave the references of catalytic evaluations over Cu-based catalysts at low and high temperatures in Table S4, and made the comparison between Cu-based and GaN catalysts under similar reaction conditions as given in the following Table R1. It is observable that at the temperature of higher than or equal to 300 °C, the pressure of 2-5 MPa, GHSV of lower than 3600 mLg -1 h -1 and H 2 /CO 2 = 3, the GaN catalyst exhibits a lower CO 2 conversion but a much higher selectivity of DME than the Cu-based hybrid catalysts. Moreover, the STY of DME over GaN catalysts at 360 °C is similar as those over Cu-based hybrid catalysts at 300-350 °C. Noteworthy, the CaCO 3 -GaN catalyst exhibits a much higher STY of DME than the Cu-based catalysts at higher reaction temperature, lower pressure and similar GHSV. Therefore, the GaN catalysts favors the selective hydrogenation of CO 2 to DME at higher reaction temperature, and the promotion of CaCO 3 can further enhance the catalytic performance of CO 2 -to-DME.   Response: Thank you, and we accept your suggestion by considering the experimental errors. Therefore, error bars are added in Figures 1, 3-6 in the revised manuscript. With the very reasonable experimental errors, the trends are still valid.
Comment: Not even all the conclusions of the work are fully convincing, since many results are obtained at low space velocity, far enough from a full kinetic control suitable to assess a superior behaviour or preferential paths under the adopted conditions.
Response: Thank you, but we cannot fully agree with you. From the previous publications shown in Table  R2, the CO 2 hydrogenation to DME over the typical hybrid catalysts are performed with a GHSV of about 1500 and 10000 mLg -1 h -1 . However, increasing GHSV from 4333 to 34666.67 mLg -1 h -1 decreases the CO 2 conversion from 21.5 to 8.2% and STY of DME from 5.9 to 2.3 mmolg -1 h -1 (Chem. Eng. J. 348, 713-722 (2018)). Therefore, we employed a medium GHSV of 3000 mLg -1 h -1 to keep a reasonable CO 2 conversion, and comparable results are obtained (Table S4).
Thermodynamically, as reported in the references, e.g., Chem. Eng. Technol. 37 (2014) 1765-1777, there exists a minimum CO 2 conversion with increasing the temperature, i.e., first decrease followed by the increase of the CO 2 conversion, the extent of which is clearly dependent on the H 2 /CO 2 ratio and the pressure as given in Figure R2. Moreover, our calculations by using Aspen plus ( Figure R3) are consistent with the reported results. Under the conditions of 2.0 MPa and 360 °C, the equilibrium CO 2 conversion is 31.1% and 26.6% with a H 2 /CO 2 molar ratio of 3 and 2, respectively (Referring Figure R2 and Figures   R3b-c). For the reaction results over GaN catalysts as given in Figure 3 in the manuscript, the maximum CO 2 conversion under the reaction conditions of 2.0 MPa and 360 °C is 16.8% and 10.6% with a H 2 /CO 2 molar ratio of 3 and 2, respectively, which are only about half of the equilibrium CO 2 conversions. Thus, the results in our cases are under kinetically controlled rather than thermodynamically limited.
Considering your concern, the different mechanism, and the new-type of catalyst, the kinetics of the GaN catalyzed CO 2 -to-DME is worthy to be studied in our future work. Comment: For these reasons, I'm not sure that the MS meets the requirements for submission in this journal, so suggesting a deep revision and a more appropriate resubmission in another journal.
Response: Thank you very much for your patience in reviewing our manuscript. We have modified the manuscript based on your and the other reviewers' comments. With the new experimental and DFT results, as you may see, both the clarity and the quality of the revised manuscript is significantly improved.
Response: Thank you. To discard the experimental errors, we repeated the catalytic evaluations and modified the Fig. 4. Now the new catalytic results with the error bars are present. It is observable that the changes of CO 2 conversion and product selectivities are significant during the first 12 h of reaction. Thus, the induction period does exist in the catalytic evaluation. In the case of XPS results, the binding energies for Ga and N are similar before and after the catalytic evaluations ( Figure S1), however, we cannot be sure that the surface propriety of GaN catalyst is unchangeable during the catalytic evaluations. Furthermore, the induction period can be explained by the results of NH 3 -TPD over the fresh and spent catalysts ( Figure   6a and Table S3). You can also find the results in Figure R4. The results indicate that a gradual loss of the acidity with increasing the time on stream, and is consistent with the decreased CO 2 conversion during the initial induction period. Moreover, the secondary reactions of DME and/or methanol to HCs may be inhibited due to the loss of the acidity, leading to the increased DME selectivity and the decreased selectivity of HCs during the induction period ( Figure 4). Response: Thank you. We agree with you. The word has been modified. For the first place, the corresponding sentence was deleted based on the comments of another reviewer. For the other place, the sentence was modified as "…the slight difference between the adsorption energy of H 2 and…"and was shown in the line 273.
Comment: 8. Table S3. The sum of product distributions does not always add up to 100%. Please, check.
Response: Thank you. Table S3 used to represent the catalytic results over GaN promoted by different promoters. In fact, the carbon balances for all the catalytic evaluations in this work are always better than 95%, indicating that the results of product selectivities are convincing. However, based on the comments of another reviewer, the catalytic results over promoted GaN are removed.

The point-by-point response to Reviewer 3
Comment: The paper of Liu and co-workers describes a catalytic system for the CO 2 reduction to DME. It is an extensive, multipart study, in which authors use different techniques to shed light on the efficiency and the mechanism of the process.
Response: Thank you very much for your kind evaluation and positive comments on our manuscript. We have revised the manuscript based on your and the other reviewers' valuable comments, and the point-by-point reply to each of your specific comment is given as follows. provides, no analysis of the different character of adsorbed hydrogens has been carried out.
Response: Thank you very much for the valuable comments. Based on your comments, the DFT calculations have been greatly improved. The adsorption of different intermediates during the CO 2 hydrogenation is systematically studied and the results are listed in Figure S16. In addition, the reaction energies and activation energies for various elementary steps are calculated. The whole pathway of CO 2 hydrogenation to DME is shown in Figure 7 ( Figure R5 below), Figures S13-S21 and Tables S7-S8. In particular, we find that over the (110) plane the activated CO 2 is preferable to be hydrogenated to the formate species while over the (100) plane CO 2 is preferable to be hydrogenated to the carboxyl species, due to the different activation energies ( Figure S17). Subsequently, over the (100) plane, the formed carboxyl species are hydrogenated and dehydrated, and finally converted to the methyl species. The DFT results also indicate that DME is formed via the coupling of the formate and the methyl over the (100)/(110) interface. More importantly, during the whole pathway, the methoxyl (CH 3 O * ) species are absent, which is consistent with the results of DRIFTS (Table S6). This indicates methanol can hardly be formed from the hydrogenation of CH 3 O * over GaN. Even though the CH 2 OH * species are present, the hydrogenation of CH 2 OH * to methanol is difficult due to a high activation energy (1.28 eV). Instead, the CH 2 OH * is preferable to be dissociated into CH 2 * and OH * with a much lower activation energy of 0.26 eV, which provides the precursor for the formation of CH 3 * . Therefore, the improved DFT calculation results not only provide a reasonable reaction mechanism for the formation of DME but also explain well why DME rather than methanol is formed as the primary product on GaN. The theoretical and experimental results together contribute to the comprehensive understanding of the CO 2 -to-DME on GaN catalysts. Comment: Contrary to that, the experimental part seems convincing and carried out with care. Sometimes it makes the impression of being too extensive -for instance omission of the paragraph devoted to carbonate promoters would not lead to the lesser scientific value of the research, and in my view it would only increase the clarity of presentation mentioned above.
Response: Thank you. We accept your suggestion, and the contents devoted to the carbonate promoters are greatly simplified by keeping the necessary results over the CaCO 3 -GaN catalyst. The modified sentences are highlighted in Lines 158-162 of the revised manuscript with yellow highlights.
Comment: Overall, the paper seems to be an example of those that carry too much information instead of sending one clear message. The manuscript has a potential, and the description of the reactivity of the GaN system is a valuable contribution to the field, but more work is needed to make it meet the standard.

Reviewers' Comments:
Reviewer #1: Remarks to the Author: After revision, the quality of MS appears to be significantly improved, being most of the points arisen convincingly addressed. However, what is not fully convincing yet is the significance of the values obtained (i.e., STY(DME) values on GaN incredibly low -more than one order of magnitude -if compared with literature data obtained at lower temperature on copper-based catalysts!!!) as well as the presentation of many results (CO is the main product in the adopted conditions and every attempt to show something of different is misleading), comments about thermodynamics (it is confirmed that, in the range of temperature chosen, DME formation is hindered), conditions selected for experiments (i.e., range of temperature). For these reasons, I cannot express a positive evaluation for this MS, definitely suggesting its rejection.
Reviewer #2: Remarks to the Author: The authors have provided suitable answers to my (minor) comments and made the required changes in the manuscript. Also, in my view, the authors have reasonably addressed the main issues raised by the other reviewers and, in consequence, have significantly improved the overall quality and clarity of the manuscript. Therefore, I recommend its acceptance for publication. I have only a few additional minor comments on formal/grammar aspects that could probably be amended at the editing stage if finally accepted: -Some figures are placed before they are mentioned in the text.
-In line 75, it should be "a much lower GHSV" instead of "a much low GHSV". -Lines 296-297: "is significantly more favorable than".
-Lines 341-342: "HCs and oxygenates including methanol are produced" Reviewer #3: Remarks to the Author: The remarks made with respect to the original submission have been addressed by the authors. The computational part has been significantly improved, and in my view it provides sufficient support for the experimental part. I am pleased to recommend the revised version of the manuscript for publication.

The point-by-point response to Reviewer #1
Comment: After revision, the quality of MS appears to be significantly improved, being most of the points arisen convincingly addressed.
Response: Thank you very much for your kind evaluation and positive comments on our manuscript, and the point-by-point reply to each of your specific comments is given as follows.

Comment:
However, what is not fully convincing yet is the significance of the values obtained (i.e., STY(DME) values on GaN incredibly low -more than one order of magnitude -if compared with literature data obtained at lower temperature on copper-based catalysts!!!) as well as the presentation of many results (CO is the main product in the adopted conditions and every attempt to show something of different is misleading), comments about thermodynamics (it is confirmed that, in the range of temperature chosen, DME formation is hindered), conditions selected for experiments (i.e., range of temperature). For these reasons, I cannot express a positive evaluation for this MS, definitely suggesting its rejection.
Response: Thank you so much for your kind comments on our work. However, we cannot fully agree with you.
As stated clearly in the manuscript, the importance of our work lies in (1)  Regarding your comment "i.e., STY(DME) values on GaN incredibly low -more than one order of magnitude -if compared with literature data obtained at lower temperature on copper-based catalysts!!!", it is not the case if the typically reported and our results are examined as show in Table S4.
For your convenience, Table S4 is appended as follows. This work DME over GaN catalysts is 0.56 ~ 2.9 mmolg -1 h -1 , which is very comparable with those over the Cu-based hybrid catalysts, i.e., 0.51 ~ 7.2 mmolg -1 h -1 . If the highest space-time yield of DME over the Cu-based hybrid catalyst (7.2 mmolg -1 h -1 ,) is concerned, it is achieved at a favorably higher pressure of 5.0 MPa, which is still less than three times higher than that over the GaN catalyst at an unfavorably lower pressure of 2.0 MPa (2.9 mmolg -1 h -1 ). More importantly, as a new catalyst, there is still a large room for the further optimization of GaN, which is worthy to be done in the near future.
In the case of your comment of "the presentation of many results (CO is the main product in the adopted conditions and every attempt to show something of different is misleading), comments about thermodynamics (it is confirmed that, in the range of temperature chosen, DME formation is hindered), conditions selected for experiments (i.e., range of temperature)", the most important finding of our work is that GaN is an active, stable, and selective catalyst for the direct hydrogenation of CO 2 to DME as a primary product and the reaction proceeds via the coupling of CH 3 * and HCOO * , which is completely different from the traditional route of the DME synthesis over a hybrid catalyst via the methanol intermediate. Definitely, the further optimization of the GaN catalyst and the reaction conditions are very necessary, and improved catalytic performance for CO 2 -to-DME is reasonably expected.
Indeed, we agree with you that the formation of CO is thermodynamically favored at higher temperatures while the formation of DME is thermodynamically favored at lower temperatures. Thus, the improvement of the lower-temperature performance of the GaN catalyst is unquestionably one of the most important work for studies in the near future. Fortunately, a reasonable mechanism is proposed in our work, which can provide valuable guidelines for the catalyst optimization.
As a common practice in the references, the CO-free selectivity of DME over Cu-based hybrid catalysts is reported for better understanding the catalytic process. Indeed, as clearly stated in the manuscript, we do not deny the high selectivity of CO over the GaN catalyst at a higher reaction temperature of 360°C (~60%). However, it is still very reasonable in comparison with those over the Cu-based hybrid catalyst reported in the references, e.g., the selectivity of CO = 56.1% at 240 °C (Appl. Catal. B. 162, 57-65 (2015)); the selectivity of CO = 69.8% at 260 °C (Chem. Eng. J. 348, 713-722 (2018)); the selectivity of CO = 79.5% at 300 °C (React. Kinet. Mech. Catal. 130, 179-194 (2020)).
More importantly, even at a higher reaction temperature of 360°C, the highest space-time yield of DME of 2.9 mmolg -1 h -1 is achieved over the GaN catalyst, which is clearly higher than some of those over the Cu-based hybrid catalysts, e.g., the space-time yields of DME of 1.7 mmolg -1 h -1 at 260°C and 0.51 mmolg -1 h -1 at 350°C as given in Table S4. Thus, GaN is not only a new but also an efficient catalyst for the direct hydrogenation of CO 2 to DME.