In silico design of dihydroazulene/vinylheptafulvene photoswitches for solar-energy storage guided by an all-around performance descriptor

The possibility to use the reversible cycling of molecular photoswitches between isomeric forms as a means to store and release solar energy has stimulated the development of candidate systems based on several different core structures, such as the dihydroazulene/vinylheptafulvene (DHA/VHF) couple. However, a major challenge in these efforts is to simultaneously realize many of the performance criteria required of the switches for such applications. Here, we take on this challenge by first introducing an all-around performance descriptor that combines three key criteria (related to energy density, storage time and light-absorption characteristics), and by then using density functional theory (DFT) methods to calculate its values for 52 newly designed DHA/VHF switches. Through this approach, we are able to identify several switches with excellent overall properties that contain a structural motif absent in all DHA/VHF compounds considered for solar-energy storage in the existing literature. For some of these switches, we also provide retrosynthetic analyses for their preparation and perform DFT calculations to demonstrate that they form the energy-storing VHF isomer through a facile DHA ® VHF photoisomerization reaction. All in all, we conclude that these switches show great promise for further development towards applications in solar-energy storage.


Introduction
The ever-increasing demand for sustainable energy solutions continues to trigger research into new materials for exploiting the energy that our sun inexhaustibly provides.Among molecular-based materials, various types of molecular photoswitches are of particular interest in light of well-known challenges to account for the variation in both solar influx and energy demand with the time of day, the weather and the season.2][3] When the absorbed solar energy subsequently is needed, these isomeric forms are made (catalytically) to return to their parent forms through a thermal back-reaction, whereby the excess of chemical energy is released as heat.Synthetic photoswitches tailor-made for this particular function are often referred to as molecular solar thermal energy (MOST) systems. 4,5r the design of an efficient MOST system, two thermochemical properties are of special interest.One is the (gravimetric) energy-storage density (ΔGstorage), which measures the free-energy difference between the photoproduct isomer and the parent isomer per unit weight, and thus indicates how much solar energy can be stored as chemical energy by the MOST system in question.Another is the activation free energy for the thermal back-reaction (ΔG ‡ ), which reflects how quickly the photoproduct isomer will revert to the (more stable) parent isomer in the absence of a catalyst, or, differently put, how long the solar energy can be stored.Generally, for an efficient MOST system, it is commonly accepted that the values of ΔGstorage and ΔG ‡ should exceed 0.30 MJ kg -1 and 100 kJ mol -1 , 2,6 respectively.
Besides the aforementioned thermochemical properties, certain photophysical and photochemical characteristics are also deemed desirable for a MOST system, such as a high quantum yield for the forward photoisomerization and an inertness towards photochemical side-reactions (e.g., photodegradation) that compete with the switching between the two isomers.Furthermore, the photoactive isomer should have a UV-vis absorption spectrum that is centered in the violet-blue regime, in which both the solar irradiance at the Earth's surface and the photon energy are sufficiently high to facilitate efficient photoconversion. 2 Moreover, this spectrum should mismatch as much as possible with that of the photoproduct isomer, so that both species absorb maximally at a unique wavelength. 2,3,5This is to ensure that the photoactive isomer does not have to compete with the photoproduct for light absorption, and to prevent the latter species from hampering the forward photoisomerization.However, despite its importance, the degree of spectral mismatch is rarely quantified or even considered in the design of MOST systems.
To date, several different core structures have been utilized for the development of MOST systems.These structures are typically low-weight organic molecules that, upon light absorption, undergo either cycloaddition, double-bond isomerization or electrocyclic reactions.][9][10][11][12][13] Other systems include azobenzene, [14][15][16][17][18] dithienylethene, [19][20][21][22] and fulvalene ruthenium [23][24][25][26] compounds.More recently, 1,2-azaborines have been brought into consideration for MOST applications, owing to their ability to photoisomerize into strained, Dewar-like structures. 27other commonly explored core structure in this field of research is the dihydroazulene/vinylheptafulvene (DHA/VHF) system shown in Fig. 1.Upon irradiation, the C1-C8a bond of DHA is cleaved to form VHF through an electrocyclic ring opening. 28The first DHAs synthesized and tested for MOST applications had two cyano groups at C1 and a phenyl ring at C2, which was either unsubstituted or monosubstituted at the para position. 29,30However, despite exhibiting reasonable photoisomerization quantum yields (0.58 at best), the small ΔGstorage values (0.11 MJ kg -1 ) and short half-lives (3 h) of the corresponding VHF isomers meant that the room for improvement was ample.Subsequently, Hammett correlation studies showed that an electron-withdrawing group is needed at C1 for the thermal back-reaction (VHF ® DHA) to take place, due to the associated stabilization of a negative charge in the zwitterionic transition state. 31,32However, while a DHA compound bearing a single cyano group at C1 achieved an improved ΔGstorage value, its back-reaction did not proceed. 33Interestingly, this cyano group could be replaced by other substituents, such as amides, imidates or ketones, without impairing the photochemical reactivity of the DHA isomer. 34Furthermore, placing an additional cyano group at C7 allowed the backreaction to proceed, but ΔGstorage remained small (0.10 MJ kg -1 ). 35While benzannulation at different positions increased ΔGstorage by shifting the aromatic/antiaromatic balance between the two isomers, this approach also resulted in the formation of unstable VHFs and triggered photochemical side-reactions. 36,37e of the reasons for the small ΔGstorage values of the aforementioned DHA/VHF systems is the tendency of the s-cis VHF isomer produced by the ring opening to release steric strain through a rotation of the C2-C3 bond, which forms the s-trans VHF isomer.Therefore, several attempts have been made to introduce a "conformational lock" in the structure, so as to maintain s-cis VHF.Different saturated and/or unsaturated bridges have been tested, linking the C3 atom with an ortho phenyl group placed at C2. 38,39 While improved ΔGstorage values of around 0.4 MJ kg -1 were obtained, this strategy also brought about an undesirable lowering of the ΔG ‡ values. 38,391][42] In order to trigger this reaction, Nielsen and coworkers showed that Cu(I) is a potent catalyst applicable to both free and macrocyclic DHA/VHF systems. 43That same year, a catalytic flow reactor for this purpose was developed, alongside a photoconversion device for using the DHA/VHF couple to collect solar energy under outdoor conditions. 44rious aspects of the DHA/VHF couple that are central to its use as a MOST system have also been investigated by computational methods.For example, regarding the mechanism of the DHA ® VHF photoconversion, it has been shown that DHA becomes planar upon excitation to the lowest singlet excited state (S1) and that the formation of VHF is mediated by internal conversion to the ground state (S0) within ~600 fs. 45,46In other studies, Mikkelsen and coworkers have identified suitable density functional theory (DFT) methods for calculating optical and thermochemical properties of DHA/VHF systems, 47 and used such methods to predict how different electron donating or withdrawing groups influence the ΔGstorage values. 48espite these efforts, there seems to be no studies in the literature where computational methods are used in a "bottom-up" way to design DHA/VHF compounds that fulfill as many key performance criteria as possible within one and the same molecular framework.[37][39][40][41] This contrasts with the situation for other MOST systems like the norbornadiene/quadricyclane couple, for which bottom-up designs have indeed been reported. 49,50Accordingly, there is a need to better exploit well-known advantages of a computational approach vis-à-vis an experimental one in designing DHA/VHF MOST systems.The present work is an attempt to fill this gap.Specifically, we here introduce an easily calculable descriptor for assessing the overall potency of any MOST system by combining three separate performance criteria, and then use DFT methods to obtain the values of this descriptor for 52 newly designed DHA/VHF compounds.Besides its bottom-up nature and extensiveness, two other unique facets of this endeavor relative to previous research should be pointed out.First, while two of the criteria considered pertain to the values of ΔGstorage and ΔG ‡ , the third involves the spectral mismatch between the parent and photoproduct isomers, which, as noted above, is rarely ever taken into account when an optimal MOST system is constructed.Second, the introduction and use of an all-around descriptor to quantify the overall MOST potential facilitate the design of compounds that simultaneously fulfill several of the relevant key requirements, which is one of the major challenges in this field of research. 2,5,51For some of the DHA/VHF compounds that emerge as particularly promising from the screening, we also investigate whether the DHA ® VHF photoconversion is a facile reaction, as required for efficient MOST applications, and explore ways to improve their solubility in polar solvents like water, which is another major obstacle when devising MOST systems based on organic molecules.Furthermore, we propose different ways to synthesize the DHA/VHF compounds in question from commercially available reagents.

Computational details
Using the Gaussian 16 software, 52 geometry optimizations, harmonic frequency calculations and intrinsic reaction coordinate (IRC) calculations 53 in the S0 state were carried out with the M06-2X density functional 54 in combination with the cc-pVDZ basis set. 55Potential-energy minima and transition structures (TSs) were characterized as stationary points having only real vibrational frequencies or one imaginary vibrational frequency along the relevant normal mode, respectively.Based on the calculated frequencies, Gibbs free energies at room temperature were derived within the standard ideal-gas and rigid-rotor approximations.Furthermore, based on the resulting geometries, more accurate electronic energies were obtained by performing single-point calculations at the M06-2X/cc-pVTZ level of theory.
For the S1 state, all calculations (geometries, frequencies and IRC) were done in the framework of timedependent DFT (TD-DFT) 56 at the same levels that were employed for the S0 state.Moreover, using the SHARC 2.0 package 57,58 interfaced with ORCA 4.2.1 59 and Gaussian 16, minimum-energy S1/S0 conical intersections (CIs) were located at the TD-M06-2X/cc-pVDZ level.Unless otherwise indicated, all calculations were performed in the gas phase, because the ideal solvents for the 52 neutral DHA/VHF compounds considered are non-polar, organic ones. 28,35,43,44In complementary calculations in which three additional compounds derivatized for use in polar solvents were investigated, the SMD approach 60 was employed to model a water solvent.
UV-vis absorption spectra were calculated from vertical excitation energies (at S0 molecular geometries) and oscillator strengths obtained at the TD-M06-2X/cc-pVTZ level.A sufficient number of excited states were included in the calculations to describe the full near-UV/visible range.The spectra were derived through convolution with Gaussian functions, each of them centered at a given excitation energy, with a height equal to the corresponding oscillator strength and a half-width of 0.3 eV.A motivation for this specific choice of half-width is given in Section 1.1 in the ESI.

Parameter for spectral mismatch and all-around performance descriptor
In order to include the spectral mismatch between the parent and photoproduct isomers among the MOST performance criteria, one needs to define a parameter for measuring this mismatch in a relevant and quantitative way.For example, the difference in wavelengths of maximum UV-vis absorption of the two isomers is not a very useful parameter, as it only provides a relevant measure when the corresponding spectra consist of single and relatively sharp peaks with identical or near-identical widths.Another tentative parameter is the root mean square deviation between the spectra; however, as it has no upper bound, the values of this parameter are not transferable between MOST systems of different chemical types.With these reservations, the dimensionless parameter that we are proposing is the product of the normalized absorption spectra of the two isomers (denoted 1 and 2), integrated over the full near-UV/visible regime (300-800 nm).Specifically, denoting this parameter as the spectral overlap integral S, where n (with unit length -1 ) is the product of the normalization constants for A1 and A2, determined in such a way that S11 = S22 = 1.With this definition, the goal to maximize the spectral mismatch is equivalent to minimize the value of S. Since this parameter can only take on values between 0 (maximal mismatch) and 1 (no mismatch), it can be used to compare any two MOST systems in a meaningful way.Furthermore, as illustrated in Fig. 2, it can also be applied when the spectra in question have different numbers of peaks and the peaks have different widths.Noting that ΔGstorage and ΔG ‡ values of 0.30 MJ kg -1 and 100 kJ mol -1 , respectively, have been put forth as suitable thresholds towards which the design of MOST systems should be aimed, 2,6 it is natural to define an analogous threshold value for S. Here, following an analysis presented in Section 1.2 in the ESI, we suggest that this threshold is set to 0.40.In a situation when the spectra of the two isomers consist of single peaks in the near-UV/visible range, this value corresponds to a difference in the wavelengths of maximum absorption of about 75 nm.
Given that an ideal MOST system maximizes ΔGstorage and ΔG ‡ and minimizes S, it is then natural to use these three parameters to define an all-around descriptor of the overall MOST potential as where MAD is an acronym for MOST All-around Descriptor.Here, the 4/3 factor is introduced in order for a system that precisely meets each of the three separate thresholds (ΔGstorage ≥ 0.30 MJ kg -1 , ΔG ‡ ≥ 100 kJ mol - 1 , S ≤ 0.40) to have a MAD value of exactly 100 in the corresponding units (GJ Fig. 3 presents the comparison of the previously reported ΔGstorage and ΔG ‡ values for compounds E1-E3 38,39,48 with the values calculated in this work.Pleasingly, the agreement between the two sets of data is generally very good, which lends credence to our computational methodology.Additionally, we also calculated the S and MAD values for E1-E3, whereby E3 was found to achieve the largest MAD value by an appreciable margin (159 vs. 50 for E2).This finding corroborates that E3 is among the best DHA/VHF MOST systems available today. 39However, its large MAD value is partly a result of overcompensation.Indeed, while its ΔG ‡ value of 84 kJ mol -1 lies firmly below the 100 kJ mol -1 threshold, this deficiency is well compensated for by its excellent ΔGstorage and S values of 0.47 MJ kg -1 and 0.33, respectively.The corresponding values reported in the original studies 38,39,48 are given in parentheses.
Starting from the DHA/VHF core structure, the 52 compounds subjected to the screening were generated by varying the C1 substituents and considering two different aryl groups at C2, both in the absence of a secondary connection between C3 and the aryl, and in the presence of three different such connections.Regarding the C1 substituents, it has been highlighted elsewhere that an electron-withdrawing group is needed for the thermal back-reaction to take place. 31,32Therefore, in addition to already proposed monocyano and dicyano substitution patterns, 29,30,33 we also explored monosubstitution with a carboxylic acid, a 2-pyridyl sulfonyl (SO2Py) group, a methyl ester, and a trifluoromethyl group, as well as disubstitution with SO2Py and a methyl group.All of these substitutions were introduced with a specific goal in mind.The carboxylic acid should provide higher water solubility.The SO2Py group, commonly used as a directing group in transition-metal catalysis, [63][64][65][66][67] might facilitate the Cu(1)-catalyzed back-reaction 43,44 by bringing the Cu + center closer to the C1-C2 bond.However, its strong electron-withdrawing capacity could also lower the ΔG ‡ value, which is why disubstitution with SO2Py alongside an electron-donating methyl group was investigated as well.For a similar reason, it was also deemed natural to consider the comparatively weaker methyl ester and trifluoromethyl electron-withdrawing groups (the former by a mesomeric effect and the latter by an inductive effect).
For the C2 aryl group, in turn, either a phenyl or a 2-pyridyl was used for their ability to act as a directing group in the Cu(1)-catalyzed back-reaction.Finally, regarding the secondary connection between C3 and the aryl, we tested ethylenediyl (-CH2CH2-), methylene (-CH2-) and -NH-(which provides some π-conjugation), as well as having no connector at all (to allow for C2-C3 rotations in the VHF isomer).Altogether, the combination of these four possibilities with the seven C1 substituents and the two C2 aryl groups yields 56 different DHA/VHF compounds.Out of these, only four have been previously described. 30,33,38The remaining 52 ones subjected to the screening are shown in Fig. 4. The key results (calculated ΔGstorage, ΔG ‡ , S and MAD values) underlying the screening are summarized in Table 1, which also includes a quantity -the energy storage efficiency (ESE) -obtained by dividing ΔGstorage with the energy of maximum absorption of the DHA isomer.This quantity corresponds to the percentage of solar energy that is transformed to and stored as chemical energy in the VHF isomer.A number of observations can be made from Table 1.For example, provided that 40 compounds (77%) have a ΔG ‡ that exceeds 100 kJ mol - 1 , it appears straightforward to find a DHA/VHF MOST system that can store the absorbed solar energy for a long time.Achieving a large ΔGstorage, on the other hand, is more difficult, as only 21 compounds (40%) reach the 0.30 MJ kg -1 threshold.Notably, among those compounds, the average ESE value is 28%, which is quite comparable to that of 34% achieved by photosynthesis. 68Notwithstanding, from the statistics it seems that an even greater challenge facing the design of a DHA/VHF MOST system is realizing sufficient spectral mismatch between DHA and VHF.In fact, only five compounds (10%) show an S value below the desired 0.40 limit, whereas as many as 22 compounds (42%) have an S value of 0.90 more.This challenge is likely to reflect that the size of the π-system and the conjugation length are relatively similar in the two isomers, which contrasts with the situation for other MOST systems (like norbornadienes and 1,2-azaborines) whose πsystems are more distinctly altered upon photoisomerization.The success of the NH connector is a result of an intricate interplay between several different factors, and can be better understood if one compares the corresponding molecules with their methylenic counterparts (for instance, comparing 23 with 16), which have nearly identical molecular masses and geometries.In 2016, Mikkelsen and coworkers showed that the incorporation of a primary amino group at C3 helps increase ΔGstorage. 48In the present case, the NH connector functions as a secondary amine and although its electrondonating capacity is weaker due to the conjugation of its lone pair with the C2 aryl group, it is sufficiently strong to provide a boost in ΔGstorage (from 0.24 MJ kg -1 in 16 to 0.33 MJ kg -1 in 23).However, this comes at the cost of a decrease in ΔG ‡ (from 117 kJ mol -1 in 16 to 100 kJ mol -1 in 23).Still, this decrease is not severe and is not incompatible with long energy-storage times.Despite these effects, the main advantage of the NH connector is its ability to drastically increase the spectral mismatch, with the S value of 0.84 for 16 being reduced to a markedly better 0.30 for 23.Notably, as further detailed in Section 2.2 in the ESI, while the additional conjugation in 23 shifts the absorption of the DHA isomer only marginally (from 354 nm in 16 to 346 nm in 23), it does cause a pronounced red-shift in the absorption of the VHF isomer (from 376 nm in 16 to 440 nm in 23), which explains the improved S value.All in all, these results provide strong support for the conclusion that a NH connector between C3 and an aryl group at C2 is a key structural element for constructing potent MOST systems based on the DHA/VHF core.
Having completed the screening, the subsequent parts of this work will explore the extent to which some of the compounds with the highest MAD scores exhibit other desirable MOST features.For this, we will primarily focus on 23 and, to a lesser extent, on 20 (which is structurally very similar to 23 and whose ΔGstorage, ΔG ‡ , S scores are almost identical to those of 23).The first feature of interest is ease of synthesis.

Retrosynthetic analysis of compound 23
Compounds 20 and 23 contain two structural elements that are not present in any of the DHA/VHF systems synthesized to date: the NH connector and a single carboxylate derivative substituent at C1.At first glance, the ester 23 could be obtained by esterification of 20, which, in turn, may be produced from a dicyano parent compound through reduction to the monocyano species with DIBAL-H 33 and subsequent hydrolysis.The NH connector, on the other hand, is a different prospect.In 2016, Nielsen and coworkers synthesized a DHA/VHF system bearing an unsaturated connector (-CHCH-), 39 but this moiety was not present on the starting material.Rather, it was generated at the last step of the synthesis from the saturated precursor.Therefore, it is necessary to define an alternative approach to the synthesis of 20 or 23 that incorporates the NH connector.Fig. 5 presents a possible retrosynthetic sequence, decomposing 23 into simple, commercially available reagents.In 2008, Maes and coworkers described the synthesis of indole derivatives via tandem palladium-catalyzed C-N/C-C cross coupling between 4-chloroquinolines and 2-chloroanilines. 70The reaction involves an initial Buchwald-Hartwig amination immediately followed by intramolecular cyclization, forming exclusively a 5membered ring.This process would allow the installation of both the phenyl group and the NH connector in a single step, from 2-chloroaniline and the ketone intermediate I2, activated as a vinyl triflate.The cyclopentanone of this intermediate could come from an intramolecular Heck reaction, forming the C3-C3a bond.This process requires activating the carboxylic acid, for instance as an acyl chloride; a similar process has been described using anhydrides. 71To ensure that the correct carboxylate is activated, the methyl ester has to be formed prior to this step to act as a protecting group.
Intermediate I1 in Fig. 5 is formally a succinic acid derivative, but a sequence of alkylation of the succinic anhydride enolate and transesterification with methanol would afford little control over the final position of the tropylium relative to the methyl ester, as the methyl ester can be formed at any of the positions.Better regiocontrol is realized if the carboxylic acid is incorporated to methyl acrylate in the form of a nucleophilic reagent, and the enolate intermediate is captured with a tropylium salt (i.e.tetrafluoroborate).In this way, deprotonation of 1,3-dithiane with BuLi/CuI would generate the lithium dialkylcuprate, which undergoes 1,4addition to methyl acrylate; the resulting enolate is trapped by the tropylium cation to afford α-alkylation.Removal of the thioacetal protecting group and oxidation of the resulting aldehyde would finally yield I1.
This seven-step process (conjugate addition ® enolate trapping ® aldehyde deprotection ® aldehyde oxidation ® acid activation ® intramolecular Heck ® triflate formation ® C-N/C-C cross coupling) for the preparation of 23, with only one additional hydrolysis step required for its conversion to 20, indicates that both compounds are synthetically feasible.Further support for this notion is provided in Section 3 in the ESI, which includes a more detailed synthetic sequence based on the analysis in Fig. 5 and also an alternative sequence starting instead from 3-oxindole.

Derivatization of compounds 20 and 23 for better solubility in polar solvents
Another desirable feature of MOST systems is that they are soluble in polar solvents like water.However, for the majority of MOST systems hitherto developed based on organic molecules, this is a major hurdle and nonpolar, organic solvents are required. 28,35,43,44In this light, the reason for including a carboxylic acid among the C1 substituents considered in the screening is precisely its naturally high solubility in water.Accordingly, we expect compound 20 to be at least partially soluble in water, especially in its deprotonated state (20A) at neutral pH.Compound 23, on the other hand, does not contain any functional group that enhances water solubility.However, whereas it is well known that the addition of almost any functional group to the DHA/VHF core can lead to drastic changes in the chemical properties, [34][35][36][37]48 one key advantage of 23 is that it can be derivatized with minimal effect on the DHA/VHF core by replacing the methyl group of its methyl ester by another substituent. In his way, it might be possible to improve the water solubility without altering the ΔGstorage, ΔG ‡ and S values too much.
However, in order to maintain ΔGstorage above the 0.30 MJ kg -1 threshold, there is clearly a limit on how large this substituent can be.Noting that the ΔGstorage of 0.33 MJ kg -1 attained by 23 comes from a DHA/VHF freeenergy difference of ~91 kJ mol -1 and a molecular mass of ~277 g mol -1 , it might be possible to increase the latter value to ~304 g mol -1 without going below the 0.30 MJ kg -1 threshold (assuming that the free-energy difference is relatively independent on the choice of substituent).Although this margin is quite small, it is sufficient for replacing the methyl group of 23 with ethanolamine (yielding compound 23A in a protonated state at neutral pH) or ethylene glycol (yielding compound 23B), whose NH3 + /OH groups should increase water solubility.With these changes introduced, it is of interest to compare how well 20A, 23A and 23B perform compared to the parent compounds.Therefore, their ΔGstorage, ΔG ‡ and S and MAD values were calculated in the same way as before, but this time accounting for the presence of a water solvent with the use of the SMD approach. 60The results are given in Fig. 6.Encouragingly, the results for 20A are very similar to the previous results for 20 (see Table 1).23A and 23B, on the other hand, perform much worse than 23 (see Table 1).This is due primarily to a decrease in ΔG ‡ , likely because of a stabilization of the TS in the water solvent, and an increase in the S value.As shown in Section 2.3 in the ESI, this increase arises from the absorption of the DHA and VHF isomers in water being brought into the same region through a pronounced red-shift in the absorption of the latter species.
Taken together, the results in Fig. 6 suggest that 20 (in the form of 20A) could well be a suitable MOST system for use in a polar solvent and, in the case of 23, that using a non-polar, organic solvent is preferable over derivatization towards a polar solvent.Importantly, however, the worse performance of 23A and 23B relative to 23 is not an intrinsic consequence of their ethanolamine (23A) and ethylene glycol (23B) substituents, but is rather due to the polar water environment.Indeed, this is clear from the observation in Fig. 6 that the ΔGstorage, ΔG ‡ and S and MAD values for 23B calculated in the gas phase agree very well with the corresponding results for 23 (see Table 1).Furthermore, from complementary calculations summarized in Section 4 in the ESI that compare the gas-phase values for 23B (MAD = 133) with those calculated in different solvents, it can be inferred that while this compound largely maintains its excellent MOST properties in a non-polar toluene solvent (MAD = 120), the properties deteriorate with increasing solvent polarity (MAD = 105/77 in acetonitrile/water), again primarily because of a decrease in ΔG ‡ and an increase in the S value.

DHA ® VHF photoconversion of compound 23
Despite the favorable steady-state properties of 20 and 23 documented herein, these compounds would not be particularly useful as MOST systems unless their DHA ® VHF photoconversion processes are facile.Therefore, we decided to investigate this issue by mapping the relevant parts of the S0 and S1 potential energy surfaces (PESs) of 23 through TD-M06-2X calculations.Given the chemical similarity of 20 and 23, the results for 23 are likely to be relevant also for 20.Moreover, seeking to address the DHA ® VHF photoconversion only, no other deactivation channels leading to the formation of undesirable side-products were considered.
The results are presented in Fig. 7 and some of the associated key structures are shown in Fig. 8.In the most stable DHA diastereoisomer of 23 (denoted 23-DHA-1 in Fig. 7), the H atom at C8a and the CO2Me group at C1 are syn oriented (see also Fig. 8).Starting from this species, the corresponding Franck-Condon (FC) point in the S1 state (with ππ* character) is populated by UV-A irradiation of 3.59 eV (345 nm).This point is connected to an S1 minimum at 2.94 eV, at which the seven-membered ring adopts a planar geometry.Such intermediate planarization of the DHA core has also been observed in non-adiabatic molecular dynamics simulations of other DHA/VHF compounds. 46From this minimum, which is further characterized in Section 5 in the ESI, accessing the photoreactive CI with the S0 state at 2.67 eV requires that an energy barrier of 0.31 eV is surmounted.Given that the associated TS lies 0.34 eV below the FC point, it seems likely that the photoexcited system is sufficiently energetic to overcome this barrier.
Regarding the photoreactive CI, it exhibits a number of features that point to the fact that it does funnel the system forward to the 23-VHF-1 photoproduct diastereoisomer, rather than back to the starting 23-DHA-1 diastereoisomer.First, with a C1-C8a distance of 2.97 Å (see Fig. 8), it occurs "beyond" the TS (at 2.50 Å) between 23-DHA-1 and 23-VHF-1 along the S0 PES, which means that while 23-VHF-1 can be reached in a barrierless fashion, 23-DHA-1 cannot.Second, as illustrated in Section 6 in the ESI, neither the S0 gradient nor the S1-S0 gradient difference vectors at the CI show any tendency of the system being funneled back to 23-DHA-1.These very features further ensure that accessing this CI through light absorption by 23-VHF-1, which is also possible (see Fig. 7), is not detrimental to the energy storage.Rather, this improves the MOST functionality of the system by increasing its photostability.All in all, then, the mechanism for the DHA ® VHF photoconversion that emerges from the calculations appears fully compatible with the use of 23 as a MOST system.(1.17 eV) means that a fully reversible photoswitching process is possible.This has a key implication for the design of a synthetic route towards 23 in that a diastereoselective sequence is not needed, as any given diastereosisomer will be transformed into the other as the photochemical-thermal reaction cycle is continuously operated.

Fig. 1
Fig. 1 Structure and atom numbering of the dihydroazulene/vinylheptafulvene MOST system.

Fig. 2
Fig. 2 Calculated S values for the spectral mismatch between the parent and photoproduct isomers for model spectra of different qualitative types.The overlap region is highlighted in yellow-black font.

Fig. 3
Fig.3Structures of three previously developed DHA/VHF MOST systems and their ΔGstorage (in MJ kg -1 ), ΔG ‡ (in kJ mol -1 ), S and MAD values calculated in this work.The corresponding values reported in the original studies38,39,48 are given in parentheses.

Fig. 4
Fig. 4 The 52 DHA/VHF compounds whose MOST potential was tested in this work.

Fig. 5
Fig. 5 Retrosynthetic analysis of compound 23 and the reactions involved in each step.Molecules drawn in green are commercially available.

Fig. 6
Fig. 6 Structures of derivatized compounds 20A, 23A and 23B and their calculated ΔGstorage (in MJ kg -1 ), ΔG ‡ (in kJ mol -1 ), S and MAD values.For 23B, values in parentheses are the corresponding ones calculated in the gas phase.

Fig. 7
Fig. 7 PESs describing the interconversion of DHA and VHF isomers of compound 23 in the S0 and S1 states.Electronic energies are given in eV relative to the S0 energy of the most stable isomer (denoted 23-DHA-1).CIs are indicated with hourglass symbols.
62g -1 mol -1 ).Henceforth, these units are omitted but implied whenever MAD values are discussed.With this definition, it is one of the goals of the present work to find DHA/VHF compounds with as large MAD values as possible and to understand chemically how such values can be attained.Before describing and presenting the results of the screening of different DHA/VHF compounds for their MOST potential, it is important to ascertain that the M06-2X-based calculations performed in this work provide reliable predictions of ΔGstorage, ΔG ‡ and S. To this end, and as further discussed below, such calculations were performed to obtain ΔGstorage and ΔG ‡ values for three previously characterized DHA/VHF MOST systems in the literature (here denoted E1, 48 E2 38 and E339).Moreover, as detailed in Section 2.1 in the ESI, for one DHA/VHF compound it was demonstrated that the S value computed with the M06-2X density functional is quantitatively very similar to the S values obtained with two alternative density functionals: CAM-B3LYP 61 and wB97X-D.62Inall calculations, the existence of different isomeric forms (of DHA, VHF and the TS for the thermal back-reaction) was accounted for by Boltzmann-averaging the values of ΔGstorage, ΔG ‡ and S over the relevant isomers (for DHA, two isomers are possible if the C1 position is heterosubstituted; for VHF, up to four isomers are possible upon C1 heterosubstitution and C2-C3 rotation; for the TS, the backreaction can occur through a conrotatory or a disrotatory mechanism.

Table 1
Calculated ΔGstorage (in MJ kg -1 ), ESE, ΔG ‡ (in kJ mol -1 ), S and MAD values for the 52 DHA/VHF compounds studied in this work (Pareto-optimal compounds are highlighted against a colored background; yellow if the MAD value is smaller than 100 and green if the MAD value is at least 100) regardless of their MAD values.The 18 retained compounds are highlighted in yellow or green font in Table1.In the second step, 19, 20, 23, 24 and 52 were identified as the subset of these compounds having a MAD value of at least 100, meaning that they either (23 and 24) meet all separate thresholds (ΔGstorage ≥ 0.30 MJ kg -1 , ΔG ‡ ≥ 100 kJ mol -1 , S ≤ 0.40), or (19, 20 and 52) compensate the nonfulfilment of a threshold by sufficiently exceeding another.Interestingly, these five compounds, among which 20, 23 and 24 are the stand-out performers with MAD values of 144-147, are structurally very similar.Specifically, the only variation among 19, 20, 23 and 24 is the C1 substituent, and 52 is the 2-pyridyl variant of 24.Moreover, the common structural denominator for the five compounds is their NH connector between C3 and the aryl group.Notably, this motif is present also in two other compounds, 21 and 51, that attain a MAD value of at least 100 but are not Pareto-optimal.
69 the first step of the screening of the 52 DHA/VHF systems, only those compounds whose ΔGstorage, ΔG ‡ and S values fulfill the conditions for Pareto optimality69were retained.Accordingly, any given compound with a specific set of ΔGstorage, ΔG ‡ and S values is Pareto-optimal only if there is no other compound that improves one (or more) of these values without worsening the other ones.Through this step, the number of compounds subsequently subjected to the MAD-based screening is reduced from 52 to 18 on simple numerical grounds,