Silylium ion mediated 2+2 cycloaddition leads to 4+2 Diels-Alder reaction products

The mechanism of silver(I) and copper(I) catalyzed cycloaddition between 1,2-diazines and siloxy alkynes remains controversial. Here we explore the mechanism of this reaction with density functional theory. Our calculations show that the reaction takes place through a metal (Ag+, Cu+) catalyzed [2+2] cycloaddition pathway and the migration of a silylium ion [triisopropylsilyl ion (TIPS+)] further controls the reconstruction of four-member ring to give the final product. The lower barrier of this silylium ion mediated [2+2] cycloaddition mechanism (SMC) indicates that well-controlled [2+2] cycloaddition can obtain some poorly-accessible IEDDA (inverse-electron demand Diels-Alder reaction) products. Strong interaction of d10 metals (Ag+, Cu+) and alkenes activates the high acidity silylium ion (TIPS+) in situ. This п-acid (Ag+, Cu+) and hard acid (TIPS+) exchange scheme will be instructive in silylium ion chemistry. Our calculations not only provide a scheme to design IEDDA catalysts but also imply a concise way to synthesise 1,2-dinitrogen substituted cyclooctatetraenes (1,2-NCOTs).

In this manuscript, Wang and Fan describe their computational studies on the mechanism of the formal inverse electron-demand Diels-Alder (IEDDA) reactions between phthalazines and siloxy alkynes catalyzed by Ag(I) and Cu(I) complexes. The experimental results for these transformations were reported by Rawal, Kozmin and coworkers in 2012 and 2014 (references 17 and 18 in the current manuscript). A related computational study on the same silver-catalyzed transformation was reported this year by Catak, Aviyente, Dedeoglu and co-workers (ChemCatChem, 2020, 12, 366;ref. 24).
In this latter work, the researchers concluded that the calculations supported the stepwise pathway for the [4+2] cycloaddition over the concerted mechanism. However, in the current manuscript, the authors propose a substantially different mechanism, which involves an initial [2+2] cycloaddition, followed by a silylium ion transfer. This mechanism is definitely intriguing. While such IEDDA reactions of azines are highly valuable as they lead to the synthesis of a variety of (hetero)cycles, catalytic variants are highly limited. In this context, understanding the mechanisms of these catalytic IEDDA reactions is crucial for the development of new and better catalytic reactions. Therefore, an in-depth computational analysis of such catalytic IEDDA reactions will be of interest to organic chemists working in the areas of catalysis and heterocyclic chemistry.
However, as listed below, several aspects of the current study arise questions and require more detailed analysis.
-The reactions of certain substituted 1,2-diazines with ynamines were shown to proceed via [2+2] cycloaddition, the ring opening of which gave diazacyclooctatetraene derivatives. The authors are recommended to cite the following articles on this topic: 1) Chem. Pharm. Bull. 1991, 39, 1713-1718. 2) Heterocycles, 1996 -There are some major differences between the results obtained in this work and in the ChemCatChem article mentioned above (ref 24): For instance, in Table 1, the uncatalyzed reaction was calculated to have a reaction barrier of 40.0 kcal/mol, whereas the barriers for the Ag(I)-catalyzed concerted and stepwise mechanisms were calculated to be 34.0 and 33.3 kcal/mol, respectively. However, in ref. 24, the Ag(I)-catalyzed concerted and stepwise pathways were computed to have barriers of Reviewer 1 Review Attachment 32.2 and 25.0 kcal/mol, respectively. In addition, the results for the complexation of Ag(I) to the ligand (2,2'-bipyridine), phthalazine and siloxy alkyne are highly different in the two studies. It is recommended that the authors provide an explanation for these differences.
-Scheme 2a and Figure 2a: The first step of the stepwise [2+2] cycloaddition is shown as the attack of phthalazine nitrogen to C1 of siloxy alkyne activated by the Ag(I) complex (conversion of Ag-int1 to Ag-int2). When the electron-rich nature of siloxy alkynes and electron-deficient nature of phthalazine are considered, an alternative initial step can be proposed to occur via the attack of the siloxy alkyne from C2 to the C1 of phthalazine. However, the authors did not discuss this possibility. It will be better if the authors compare the Gibbs energies for these two possible pathways.
-Scheme 2a and Figure 2a: The conversion of Ag-int2 to Ag-int3 seems problematic from an organic chemistry perspective. In this transformation, the silyl enol ether moiety was proposed to attack the iminium carbon intramolecularly, rather than the C-Ag bond. Given the highly electron-rich nature of such carbon-metal bonds, the C-Ag bond seems a better nucleophile candidate to attack the iminium carbon.
Moreover, the proposed product of this step, Ag-int3, continues to have a C-Ag bond. Such a C-Ag bond is present in most the intermediates, which look counterintuitive. Similar arguments can be made for the Cu(I)-catalyzed reaction.
-Some references (such as references 5, 19, 20 and 24) do not have page numbers and/or volume numbers.
-There are many grammatical errors and typos throughout the manuscript, which make the text difficult to follow and understand.

Comments of reviewer 1:
In this manuscript, Wang and Fan describe their computational studies on the mechanism of the formal inverse electron-demand Diels-Alder (IEDDA) reactions between phthalazines and siloxy alkynes catalyzed by Ag(I) and Cu(I) complexes. The experimental results for these transformations were reported by Rawal, Kozmin and coworkers in 2012 and 2014 (references 17 and 18 in the current manuscript). A related computational study on the same silver-catalyzed transformation was reported this year by Catak, Aviyente, Dedeoglu and co-workers (ChemCatChem, 2020, 12, 366;ref. 24). In this latter work, the researchers concluded that the calculations supported the stepwise pathway for the [4+2] cycloaddition over the concerted mechanism. However, in the current manuscript, the authors propose a substantially different mechanism, which involves an initial [2+2] cycloaddition, followed by a silylium ion transfer. This mechanism is definitely intriguing. While such IEDDA reactions of azines are highly valuable as they lead to the synthesis of a variety of (hetero)cycles, catalytic variants are highly limited. In this context, understanding the mechanisms of these catalytic IEDDA reactions is crucial for the development of new and better catalytic reactions. Therefore, an in-depth computational analysis of such catalytic IEDDA reactions will be of interest to organic chemists working in the areas of catalysis and heterocyclic chemistry. However, as listed below, several aspects of the current study arise questions and require more detailed analysis.

Reply:
We are very grateful to reviewer 1's comment and valuable suggestions.

Question:
The reactions of certain substituted 1,2-diazines with ynamines were shown to proceed via [2+2] cycloaddition, the ring opening of which gave diazacyclooctatetraene derivatives. The authors are recommended to cite the following articles on this topic: 1) Chem. Pharm. Bull. 1991, 39, 1713-1718. 2) Heterocycles, 1996 Reply: The recommended reference is very valuable to our work. We have referred it in the introduction section (refer 12,13) with a brief discussion.

Question:
There are some major differences between the results obtained in this work and in the ChemCatChem article mentioned above (ref 24): For instance, in Table 1, the uncatalyzed reaction was calculated to have a reaction barrier of 40.0 kcal/mol, whereas the barriers for the Ag(I)-catalyzed concerted and stepwise mechanisms were calculated to be 34.0 and 33.3 kcal/mol, respectively. However, in ref. 24, the Ag(I)-catalyzed concerted and stepwise pathways were computed to have barriers of 32.2 and 25.0 kcal/mol, respectively. In addition, the results for the complexation of Ag(I) to the ligand (2,2'-bipyridine), phthalazine and siloxy alkyne are highly different in the two studies. It is recommended that the authors provide an explanation for these differences.
Reply: There were two differences between our work and the ChemCatChem article  figure S1 in the SI and result section part III of the revised manuscript). Also, if they use the same reference as us, the barrier for the stepwise mechanism would be 31.3 kcal/mol which is quite close to our value.
The energy difference of concerted and stepwise mechanism is 7.2kcal/mol in ref 26, and only 0.7kcal/mol in our work. To make clear the cause of this difference, we have compared reaction profiles with the same model as ref26 (with TMS as protection group) under wb97xd/def2tzvp//M06-2x/def2tzvp,6-31G(d,p) level of theory with the same reference point as ref 26. The reaction barrier obtained from wb97xd/def2tzvp//M06-2x/def2tzvp,6-31G(d,p) level of theory is in accordance with ref26 (the uncatalyzed [4+2] barrier 46.6kcal/mol vs 45.2 in ref26, stepwise [4+2] mechanism 27.0 vs 25.0 in ref26, concerted mechanism 32.9kcal/mol vs 32.2kcal/mol in ref26). We further calculated reaction barrier with the same model under B2PLYPd3(BJ)/def2tzvp//M06-2x/def2tzvp,6-31G(d,p) level of theory and the barrier is in accordance with our work (uncatalyzed [4+2] mechanism 41.6kcal/mol vs 40.0kcal/mol in our work, stepwise [4+2] mechanism 25.4kcal/mol vs 26.1kcal/mol in our work, concerted [4+2] mechanism 28.1kcal/mol vs 26.8kcal/mol in our work). It turned out that energy differences of concerted and stepwise mechanism manly come from theoretical method to get the electronic energy (see details in table S2). For both uncatalyzed and catalyzed mechanism, the concerted barriers from wb97xd are higher than that from B2PLYPd3(BJ) by about 5kcal/mol, while the stepwise barriers are close. Since uncatalyzed mechanism is a pure organic system where double hybridized functional B2PLYPd3 is expected to work very well, we think although Wb97xd is often quite suitable for metal catalyzed system, in this reaction the barrier of concerted mechanism may be overestimated. To avoid such functional induced errors, we further compared rate determine barrier of our proposed SMC mechanism and [4+2] mechanism with different functional (wb97xd, pbe0, b3lyp-d3, pbe0-d3, m06-d3, b2plyp, mpw2plyp, b2plyp-d3, b2plypd3(BJ), B3lyp see details in SI table S3) and found the rate determine barrier of SMC mechanism lowed [4+2] mechanism 9.2-16.2kcal/mol (with TIPS group as used in this work. The value will be 9.8-11.2kcal/mol, as shown in table S1, if TMS group is used as in ref26) among all the tested functional. Our proposed SMC mechanism is more favored than [4+2] mechanism for every functional that has been tested. We have discussed the difference between two studies detailed in SI (part I) and in the revised manuscript (result section part III and introduction section). Figure 2a: The first step of the stepwise [2+2] cycloaddition is shown as the attack of phthalazine nitrogen to C1 of siloxy alkyne activated by the Ag(I) complex (conversion of Ag-int1 to Ag-int2). When the electron-rich nature of siloxy alkynes and electron-deficient nature of phthalazine are considered, an alternative initial step can be proposed to occur via the attack of the siloxy alkyne from C2 to the C1 of phthalazine. However, the authors did not discuss this possibility. It will be better if the authors compare the Gibbs energies for these two possible pathways.

Question: Scheme 2a and
Reply: Thanks for the comments. We have tested the C2 attack point and the barrier is 32.4kacl/mol, which is much high than C1 (19.5kcal/mol) case. We think the lower barrier of C1 attack mainly due to the positive charge on C induced by Ag(I) can be stabilized by O atom through conjugated effect. We have discussed this C2 attack pathway in the revised manuscript (part II in the results section). discussion at some points of the paper. It is crucial to make a thorough correction of the paper.
Reply: Thank you for comment. We have made a thorough correction of our manuscript.

Question:
The authors have chosen M06-2X for the structural optimization, although it is known that M06 functional describes better the features of transition metal complexes, as is the case for Cu and Ag catalysts. If they believe M062X is adequate, they should explain why, or otherwise reconsider to use M06, at least for the crucial schemes.

Reply:
We have reoptimized geometries of Ag catalyzed IEDDA with M06 functional and M06-d3 functional. The difference between M06 and M06-d3 functional is relatively small (the average difference between C-Ag bond 0.004 Å, C-N bond within 0.001 Å, bond angle within 0.31° among the tested geometries, see details in figure S2).
We further compared the parameters between M06-d3 and M06-2X optimized geometries of Ag catalyzed IEDDA. The difference between the two functional is relatively acceptable (the average difference of N-Ag bond 0.04Å, average difference of C-Ag bond 0.1Å, average differences of bond angle 1.25°. see details in figure S2). To evaluate the influence of geometry difference to the reaction barrier, we further calculated Gibbs energy profile of our proposed SMC mechanism under B2PLYPd3(BJ)/def2tzvp//M06-d3/def2tzvp,6-31G(d,p) level of theory and the result is very similar with our original work (the difference of rate determine barrier is only 0.2kcal/mol, and the averaged difference for transition states and intermediates is 0.46kcal/mol, see details in figure  S3).
For Cu catalyzed IEDDA, the difference between the two functional is relatively smaller than Ag case (the averaged difference between C-Cu bond 0.07 Å, C-N bond within 0.06 Å, bond angle within -0.65° among the tested geometries, see details in figure S2). To evaluate the impact posed by these geometries differences on reaction energy path way, we also calculated Gibbs energy profile of Coper(I) catalyzed SMC mechanism under B2PLYPd3(BJ)/def2tzvp//M06-d3/6-31G(d,p) level of theory and the result is also very similar with our previous work (the difference of rate determine barrier is 1.4kcal/mol, and the averaged difference is 0.25kcal/mol. see details in figure S4).
Therefore, although M062x and M06-d3 predict slightly different geometries, our final reaction energy profile based on M062x and M06-d3 geometries are quite similar. In the ref 26 (ChemCatChem. 12, 366-372 (2020)), the author have tested a series of functional for geometries and found that M062x is suitable for the current system. Recently, a detailed benchmark investigating on the transition metal reaction barrier heights in organometallic reactions including 3d-5d transition metals has also shown that M06-2X performs surprisingly well and even, M06-2X outperforms M06 (M. A. Iron, T. Janes, J. Phys. Chem. A. 2019, 123, 3761-3781). So, we consider the M06-2x functional is also appropriate for both Ag and Cu catalyzed IEDDA.

Question:
According to the experimental work by Rawal, Cu seems to be more efficient in promoting the reaction than Ag. However, the calculations point to the opposite situation, when comparing the activation barrier in Figures 2 and 6. Explain it or elaborate it further.
Reply: Thanks for the question. We have carefully evaluated Ag and Cu catalyzed IEDDA and found that Ag performs better than Cu in the same condition. In the Ag catalyzed IEDDA (J. Am. Chem. Soc. 2012, 134, 1, entry 14). We have chosen AgOTf and CuOTf as catalyst salt for Ag and Cu catalyzed IEDDA for convenient, and this could also introduce some differences compared to the experiments. Rawal also reported that Cu can catalyze IEDDA between pyridopyridazines and siloxy alkynes (Org. Lett. 2014, 16, 3236−3239. Figure1, entry 16) when refluxed in dichloroethane (DCE) while Ag was inefficient to this reaction. The temperature of this reaction as high as 83.5℃ and we think this reaction may not proceed through our proposed SMC mechanism.

Question:
There are many mentions during the discussion to the formation of the 1,2-NCOT product, which is regarded as a dynamically stable compound. According to the authors it could appear in certain conditions. But it is not clear to me if this compound has ever been isolated in this type of reactions, or whether it is just a theoretical prediction without any experimental support. Please clarify. Thus, the work is too preliminary for publication in this form Reply: At the beginning we did not notice clear experimental evidences for Nitrogen-substituted COT as unexpected byproduct of IEDDA, and it is just a theoretical prediction. Thanks for Reviewer 1, he recommended us two excellent example 1) Chem. Pharm. Bull. 1991, 39, 1713-1718. 2) Heterocycles, 1996 In the recommended reference, substituted 1,2-diazines with ynamines were shown to proceed both via [2+2] cycloaddition and [4+2] IEDDA cycloaddition at the same time, the [2+2] cycloaddition followed with ring expansion to give diazacyclooctatetraene derivatives. They obtained [4+2] IEDDA product and diazacyclooctatetraene derivatives with comparable yield. We also found some example of direct [2+2] between alkynes and aromatic rings, which then followed by ring expansion to obtain COT similarities. The yield of such reaction is relatively low. 1) Chem. Commun., 2007, 5119-5133. 2) Tetrahedron 57 (2001) 7575-7606. 3) Can. J. Chem. 81 37-44 (2003). 4) Journal of Organometallic Chemistry 693 (2008) 894-898. So, this prediction do have some experimental supports, and it indicates a very concise way to construct Nitrogen-substituted COT similarities.

Comments of reviewer 3:
In this manuscript, Fan and co-workers reported their investigations on the mechanism of silver(I)-and copper(I)-catalyzed [2+2] cycloaddition reactions between 1,2-diazines and siloxyalkynes by density function theory (DFT). The results showed the migration of silylium ion [triisopropylsilyl ion (TIPS+)] controlled the reconstruction of four-member ring to give the final product. This work not only disclosed that the SMC mechanism is more favorable for this silver(I) and copper(I) catalyzed IEDDA reaction, but also presented a powerful method for the synthesis of 1,2-dinitrogen substituted cyclooctatetraene (1,2-NCOTs). In addition, the paper was well organized and written. Overall, this work is innovative and interesting, and this reviewer would like to recommend the acceptance for Communications Chemistry after addressing the following concerns: Reply: We are very grateful to comment and good suggestion.

Question:
The reaction profiles of ketone-enol tautomerism pathway (Figure 4a) showed that the direct intramolecular isomerization via Ag-TS4b has an activation free energy of 18.6 kcal/mol (path b); however, in the copper(I) catalyzed IEDDA reaction (Figure 5a), path b is energetically unfavored due to the overall activation free energy of 33.8 kcal/mol (from Cu-int1 to Cu-TS4b). What causes this difference in energy barriers should be explained.
Reply: Thanks for the nice comment. There was some misleading due to our poor expression. We have rewritten the related paragraphs to avoid such misleading. The rate determine step of path b is ketone-enol tautomerism step, the barrier relative to zero point (Ag-int1 or Cu-int1) of this step is 23.9kcal/mol (Ag-TS4b) for silver(I) and 33.8kcal/mol (Cu-TS4b) for copper(I) catalyzed IEDDA. Path b is energetically unfavored relative to Path a for both silver(I) and coper(I) catalyzed IEDDA. For silver(I) and copper(I) catalyzed IEDDA, the difference of the rate determined barrier is 9.9 kcal/mol for path b, 6.5 kcal/mol for path a, and 6.0 kcal/mol for [4+2] path. In each case the barrier for copper(I) is higher which is in accordance with experiment. We think this is partly due to relative smaller radius and higher positive charge density of copper(I) atom, which lead to a stronger coordination of the substrate. According to our calculation, formation of L-Cu+ is 10kcal/mol more exothermic than formation L-Ag+ (-33.0kcal/mol vs -23.1kcal/mol). Formation of the п-complex Cu-int1 is much more exothermic than Ag-int1(-30.3kcal/mol vs -60.8kcal/mol, figure 1). п-complex Cu-int1 is more stable than Ag-int1 in solution, which lead to a relatively lower reference point of copper(I) catalyzed IEDDA. This is main cause of relatively higher barrier of Cu catalyzed IEDDA. There was also some experiment confirmed that copper(I) п-complex is more stable than silver(I) п-complex (ref 29). In addition, copper(I) atom is more prone to be oxidized, NBO charge distribution shows that there was evident electron transfer from copper(I) to siloxy alkyne which may also lead to stronger coordinate of copper(I) to siloxy alkynes (NBO charge distribution shows that the electron changed from -0.164 to -0.271 in the C2 and 0.293 to 0.248 in the C3 when siloxy alkyne 2a coordinated with copper(I) to form Cu-int1, electron charge changed into -0.218 in C2 and 0.312 in C3 when siloxy alkyne 2a coordinated with silver(I) to form Ag-int1). We have discussed this in detail in our revised manuscript in the discussion section and results section (part I).