Experimental and Computational Investigations of the Reactions between α,β‐Unsaturated Lactones and 1,3‐Dienes by Cooperative Lewis Acid/Brønsted Acid Catalysis

Abstract The reactions of α,β‐unsaturated δ‐lactones with activated dienes such as 1,3‐dimethoxy‐1‐[(trimethylsilyl)oxy]‐1,3‐butadiene (Brassard's diene) are barely known in literature and show high potential for the synthesis of isocoumarin moieties. An in‐depth investigation of this reaction proved a stepwise mechanism via the vinylogous Michael‐products. Subsequent cyclisation and oxidation by LHMDS and DDQ, respectively, provided six mellein derivatives (30–84 %) and four angelicoin derivatives (40–78 %) over three steps. DFT‐calculations provide insights into the reaction mechanism and support the theory of a stepwise reaction.

All these examples show that there is a need for a short and efficient synthesis of isocoumarins 1-6 and their derivatives as building blocks for natural product synthesis. For the increasing necessity of new drugs, caused by the growing population, resistances, and new diseases, we need to understand the mode of action of drugs, but also the chemical reactions. Therefore, it is a major challenge to develop general and predictable methods to achieve building block syntheses such as isocoumarins in an applicable way.
Summary Synthetic Investigation. After extensive catalyst screening, a cooperative catalyst system of AlMe 3 and Tf 2 CH 2 was found as the best working system for the investigated reaction. Getting the vinylogous Michael-product as the major product and determining its (E)-configuration by nOe measurements, the idea of direct cyclisation of this (E)configured Michael-product towards the desired isocoumarin was established. The crude product could readily be oxidised by DDQ to the desired isocoumarins. Overall, six different mellein derivatives and four different angelicoin derivatives could be synthesized in moderate to good yields over three steps (30-84 %).
Computational investigation. In order to receive more detailed information on the reaction mechanism and to understand as well as explain the unexpected high reactivity of the catalytic system Tf 2 CH 2 /AlMe 3 in the reactions between a,b-unsaturated lactones and 1,3-dienes, next we carefully analysed the reaction between lactone 7 and Brassards diene (8) computationally [M06-2X-D3/def2-QZVP/IEFPCM(toluene)//M06-L-D3/6-31 + G(d,p)/ IEFPCM(toluene)]. [44] In the absence of any catalyst, the reaction between 7 and 8 is thermodynamically favourable to yield endo-and exo-I1 (DG = À18.8 and À18.3 kcal mol À1 ). The cycloaddition proceeds through transition states TS1 endo and TS1 exo of very similar energies (DG°= 26.3 and 25.9 kcal mol À1 ). The forming C À C bond lengths significantly differ (2.07 and 2.00 vs. 3.05 and 3.08 , Scheme 6) within TS1 but no zwitterionic intermediates could be identified when following the intrinsic reaction coordinate (IRC) path. In line with this, most zwitterionic structures collapsed to the cycloadducts in separate calculations and stable structures were found to be significantly higher in energy (> 33 kcal mol À1 ). Therefore, it can be concluded that a putative background reaction should proceed through a concerted, yet asynchronous reaction. The high barrier of ca. 26 kcal mol À1 is qualitatively in line with the experimental finding that no cycloaddition product could be detected even upon heating to 100 8C. Instead, a decomposition of Brassards diene (8) was observed at elevated temperatures.
Scheme 5. Synthesis of alternariol derivatives from coumarins. Scheme 6. Calculated Gibbs free energies (in kcal mol À1 ) for the uncatalyzed cycloaddition between 7 and 8 (above) and structure as well as selected bond lengths (in ) for the transition states TS1 (below). Different mechanistic proposals can be suggested for the Tf 2 CH 2 /AlMe 3 -catalyzed reaction: Brønsted acid catalysis by Tf 2 CH 2 or a more acidic Tf 2 CH 2 -AlMe 3 adduct, Lewis acid catalysis by AlMe 3 , or alternative catalytic species formed in the reaction must be considered. We first investigated a Brønsted acid catalysis pathway as summarized in Scheme 7, with selected structures shown. This reaction starts with the protonation of the lactone by Tf 2 CH 2 (values for other Brønsted acids like TfOH or HCl are shown in the Supporting Information. According to our calculations, this process is highly endergonic (+ 45 kcal mol À1 ). The value is probably overestimated due to the unfavourable charge separation in the calculations and additional specific solvent-solute interactions not taken into account in continuum models. All attempts to locate concerted pathways failed in these cases and all transition-state guesses resulted in stepwise mechanisms with the formation of a zwitterionic intermediate. The most M06-2X-D3/def2-QZVP/IEFPCM-(toluene)//M06-L-transition state for the Brønsted acid catalysis TS2 a requires an activation free energy of 58.5 kcal mol À1 and results in the unstable zwitterion (E)-13-TMS + . In this transition state, the length of the forming C À C bond is 2.33 , while the second set of carbon atoms is still well separated (4.61 ). Interestingly, transition states leading to a Z-configured double bond within 13-TMS + are significantly lower in energy (DG°=+ 46.1 kcal mol À1 , not shown in Scheme 7). However, these structures are unproductive as they cannot react further to yield the cycloadduct, as observed and described experimentally. The zwitterion (E)-13-TMS + then collapses via a small barrier (TS2 b, DG°=+ 48.0 kcal mol À1 ) to give the protonated cycloadduct exo-I1-H + . In line with experimental results, the calculations (even though the barriers might be overestimated) clearly demonstrate that a simple Brønsted acid catalysis is not likely to be the origin of the high activities.
We next focused our attention on the potential Lewis acid catalysis by AlMe 3 as catalyst (Scheme 8). Again, the first step is the activation of the lactone through coordination to the Lewis acid. Based on our calculations this is a favourable process (DG = À11.3 kcal mol À1 ) that should occur readily. Again, all identified transition state structures resulted in stepwise reactions as discussed for the Brønsted acid catalysis above. The formation of the first C À C bond proceeds with an activation free energy of 14.8 kcal mol À1 through TS3 a and results in the zwitterionic intermediate (E)-13-AlMe 3 . Similar to the Brønsted acid catalysis described above, the formation of a Z-configured intermediate is also possible for AlMe 3 and proceeds with a comparable barrier. The zwitterion collapses in the next step without significant barrier via TS3 b (DG°=+ 1.5 kcal mol À1 ) with only a very small barrier. Based on the computed activation free energy of 14.8 kcal mol À1 , a reaction should be observable in the presence of catalytic amounts of AlMe 3 , although no product formation was detected experimentally under the screening conditions. As other functionals (DSD-BLYP-D3BJ, wB97X-D, B3LYP-D3BJ) resulted in similar barriers around 15 kcal mol À1 , we can exclude a systematic error in the M06-2X calculations. Similarly, our Scheme 7. Calculated Gibbs free energies (in kcal mol À1 ) for the Brønsted acid catalyzed cycloaddition between 7 and 8 (above) and structure as well as selected bond lengths (in ) for the transition states TS2 (below). Scheme 8. Calculated Gibbs free energies (in kcal mol À1 ) for the AlMe 3catalyzed cycloaddition between 7 and 8 (above) and structure as well as selected bond lengths (in ) for the transition states TS3 (below).
calculations further indicate that the interaction between AlMe 3 and with the lactone (DG = À11.3 kcal mol À1 ) is slightly stronger than an interaction with any of the three oxygen atoms of Brassards diene (À7.5 < DG < À4.1 kcal mol À1 ). Therefore, we have to conclude that either the solvent, which is present in a large excess, or the product interacts with the Lewis acid and lowers the reactivity and in turn increase the activation free energy.
Finally, we addressed the full catalytic system consisting of Tf 2 CH 2 and AlMe 3 . In previous investigations, Taguchi and co-workers proposed that Tf 2 CH 2 reacts with AlMe 3 to form the aluminum methide I2 and methane. I2 could then either react as a stronger Brønsted or Lewis acid to catalyze the Diels-Alder reaction. [33] Based on our calculations (Scheme 9), a Lewis acid-base pair is formed first in a thermoneutral reaction. The proposed aluminum methide I2 could then be formed in an exergonic reaction (DG = À27.3 kcal mol À1 ), but no transition states could be identified for this reaction. Instead, all transition states indicate that the isomeric O-substituted species I3 is formed instead. I3 is not only considerably more stable than its isomer I2, but it is also formed through a small activation free energy of 14.4 kcal mol À1 (TS4). The dual coordination of the AlMe 2 -fragment to both sulfonyl groups significantly contributes to the higher thermodynamic stability of I3. The up-field shifts for the Tf 2 CH-carbon and Tf 2 CH-proton reported by Taguchi and coworkers are both in agreement with I2 and the isomeric structure I3.
Both I2 and I3 could now act either as Brønsted acids or Lewis acids to catalyze the subsequent [4+2] cycloaddition.
For an estimate of the change in Brønsted acidity, we calculated the reaction free energies for the isodesmic proton-transfer reactions as shown in Scheme 10. As these reactions are either almost thermoneutral (I2) or highly unfavourable (I3), one has to conclude that I2 and I3 cannot be considered as significantly stronger Brønsted acids compared to the free Tf 2 CH 2 . Consequently, a Brønsted acid catalysis is rather unlikely as the origin of the catalytic activity.
Therefore, we focused on the Lewis acid catalysis pathway and wondered how I2 and I3 activate lactone 7. The interaction of I2 with the lactone 7 leading to 7-AlC is again an exergonic reaction and comparable yet slightly weaker than that of AlMe 3 (Scheme 11, left). Interestingly, the Zconfiguration is more stable compared to the E-configuration, which might be attributed to the stronger C À H···O hydrogen bond (Scheme 11). However, the subsequent C À C bond formation occurs through TS4 a with an activation free energy of only 11.1 kcal mol À1 . In contrast to the previous systems, the zwitterionic intermediate (E)-13-AlC next collapses through a comparable barrier of 10.5 kcal mol À1 to give the endo cycloadduct endo-I1-AlC. The higher stability of the zwitterionic intermediate with respect to the second bond formation also explains, why the Michael adduct 13 is observed as the main product in the absence of oxidants or fluoride salts (Scheme 1 d).
Alternatively, I3 could be the active catalyst of the AlMe 3 -Tf 2 CH 2 mixture and the calculated Gibbs free energies are summarized in Scheme 11 (right). Based on our computations, the interaction of I3 with the lactone 7 is considerably weaker (DG = À2.1 kcal mol À1 ) than that of I2. In contrast, the direct comparison of the isomeric adducts 7-AlC and 7-AlO (Scheme 11, below) reveals, that the latter is thermodynamically preferred over the former by 12.4 kcal mol À1 (not shown in Scheme 11). This can be attributed to the high intrinsic stability of the free Lewis acid I3. Interestingly, no additional hydrogen bonds (e.g. between the Tf 2 C-H and the O-atom of the ester) stabilize these complexes. The subsequent CÀC bond formation proceeds via TS5 a with an activation barrier of 16.8 kcal mol À1 , followed by the cyclization via TS5 b with an activation free energy of 9.8 kcal mol À1 . Again, the second C À C bond formation occurs much faster than the first one, but as the barriers of both steps are closer in energy than e.g, in Scheme 8, the life time of the intermediate (E)-13-AlO should also be larger. Scheme 9. Formation of the catalytically active species between Tf 2 CH 2 and AlMe 3 (free energies in kcal mol À1 ) (above) and structures of the transition state TS4 and the potential catalyst I3 and selected bond lengths (in , below). Scheme 10. Calculated free energies for the isodesmic proton-transfer reactions between I2 and I3 and the Tf 2 CH anion (in kcal mol À1 ).
When comparing the different mechanistic pathways of Schemes 7, 8, 9, and 11, a Lewis acid catalysis by the AlMe 3 -Tf 2 CH 2 -mixture results in the lowest activation free energy. Among the different isomers of the catalyst, the computational data indicate that I2 is most likely the catalytically active species. The calculated activation free energies for the Lewis acid catalyzed reactions are probably underestimated as both Lewis acids are likely to interact with the solvent molecules in the system. Given the similar interaction energies with carbonyl groups, it can be expected that the solvent-solute interactions are also comparable in both cases. These findings are also in line with previous 13 C NMR investigations by Taguchi and colleagues, as a stronger change in chemical shifts was observed for AlMe 3 -Tf 2 CH 2 than for AlMe 3 alone. [26] This indicates that coordination to the former results in a larger lowering of the LUMO of the Michael acceptor.

Conclusion
Experimental and computational investigations show that the reaction of 5,6-dihydro-2H-pyran-2-one (7) and (Z)-((1,3dimethoxybuta-1,3-dien-1-yl)oxy)trimethylsilane (8) (Brassards diene) catalyzed by AlMe 3 and Tf 2 CH 2 undergoes a stepwise mechanism and no concerted Diels-Alder like reaction, making a competing vinylogous Michael addition possible. The experiments show that the major Michaelproduct is (E)-configured (E)-13, which could be verified by nOe-spectra and the convenient conversion into the cyclized product 12 by LHMDS. Oxidation of the intermediates 12 with DDQ gives the aromatic isocoumarins 14. Overall, six mellein derivatives and four angelicoin derivatives could be synthesized in moderate to good yields over three steps (30-84 %). The computational results underline the experimental results, showing the vinylogous Michael addition and the AlMe 3 /Tf 2 CH 2 -system as catalyst to be energetically favoured, in comparison to the direct formation of the Diels-Alder product and single AlMe 3 as catalyst.