id
stringlengths 24
24
| idx
int64 0
402
| paragraph
stringlengths 106
17.2k
|
|---|---|---|
623179fd8ab37333ac6662a7 | 22 | Fig. also shows parts of the dissociation curves of the three truncated basis set levels for SCS-CC2. We can clearly rule out the DZ level from further consideration. The TZ and QZ curves are relatively close to one another with the difference between both their minima being only 0.11 kcal/mol, suggesting the result is close to convergence (Table † ). The SCS-CC2/def2-TZVP minimum lies 0.34 kcal/mol above the CCSDR(3)/CBS interaction energy at r = 3.00 Å. In contrast, the SCS-CC2/def2-QZVP mimimim is only 0.23 kcal/mol higher. Considering that the CCSDR(3)/CBS interaction energy at r = 3.00 Å is -13.52 kcal/mol, the percentage errors for the SCS-CC2/TZ, QZ and CBS (3,4) minima are only 2.5, 1.7 and 0.96%, respectively; see Table for the SCS-CC2/CBS (3,4) minimum. |
623179fd8ab37333ac6662a7 | 23 | When choosing a reliable benchmark level of theory for the subsequent TD-DFT study, the accuracy of a method must be considered alongside the time and resources needed to treat the system sizes involved in this study as well as the large number of points needed to generate smooth dissociation curves. As SCS-CC2/def2-QZVP calculations are feasible across all excimer systems, obtaining SCS-CC2/CBS(3,4) curves offers the potential for a reference at a standard comparable with ground-state studies. Ideally, to establish the accuracy of using this level of theory, we would want to obtain CCSDR(3)/def2-TZVP data for the larger excimer systems to offer a comparison. However, at this stage, our computational resources prevent this proposed extension. Given the low percentage errors for the benzene test case for truncated basis sets, discussed above, we obtain SCS-CC2 values for all excimers with TZ, QZ and CBS (3,4). Table lists the values of D e and r e from minima across these basis set treatments, while the corresponding dissociation curves are given as Figs . † Without higher-level data for all systems we cannot make a definitive statement as to the accuracy of these results. That being said, prior to extrapolation r e appears already close to convergence across all systems and D e differs within 0.11 to 1.52 kcal/mol. |
623179fd8ab37333ac6662a7 | 24 | Our main incentive for the TD-DFT benchmark is to assess the performance of modern TD-DFT methods, with and without applied dispersion corrections, based on their resulting dissociation energy curves. For this purpose, we choose SCS-CC2/CBS (3,4) as the reference level for the subsequent study. We would like to point out that the curves discussed herein are an improvement on what has been acceptable in the field as binding energies with basis sets of QZ quality or higher being rare for aromatic exciplexes. Nevertheless, we recommend a future study dedicated solely to (single-reference) WFT methods that addresses CBS extrapolations and BSSE corrections more in detail. As such studies are beyond the scope of this work, we continue with the discussion of the TD-DFT results. |
623179fd8ab37333ac6662a7 | 25 | Having established SCS-CC2/CBS (3,4) as a reference method for the four excimer models, it is now possible to move on to benchmarking various DFAs. Of the four excimers (Fig. ), benzene is the smallest and has therefore received the most attention in previous computational studies. While the other excimers have received some attention, they are typically studied individually rather than comparatively. The binding of an excimer involves a change in the geometry upon excitation with the eclipsed dimer, also called the "perfect sandwich" structure, widely accepted as the most stable conformation of an excimer. The excitation responsible for excimer formation causes a displaced ground-state dimer to move into this eclipsed form, which is also associated with a reduction in distance between the monomers. The most energetically stable intermolecular separation as determined experimentally for aromatic excimers is reported in the range of 3.0-3.6 Å. Our SCS-CC2/CBS(3,4) r e values, reported in Table , fall within 0.03 Å of this range which is reasonable given highaccuracy theoretical calculations tend to predict shorter r e . The close and parallel stacking of the eclipsed formation facilitates excimer binding interactions: electrostatics, Pauli-repulsion, CT, exciton coupling and London dispersion. For example, parallel stacking and short intermolecular distance increases orbital overlap, promoting exciton delocalisation. The interplay of these various interactions may present a challenge for DFT methods to properly describe the dissociation energies and intermolecular distances. The main focus of this section is to investigate DFAs with this in mind. Before we discuss each model system individually, we study the impact of the amount of FE on the TD-DFT dissociation energies. Each model system is then first discussed without the addition of dispersion corrections before the impact of said corrections is studied separately. This section ends with an overarching discussion across all four dimers by means of statistical analysis. |
623179fd8ab37333ac6662a7 | 26 | In this section, we analyse the influence of FE on the description of excimer binding by global hybrids. This analysis was performed by varying percentages of FE, between 0 and 75%, for two underlying exchange-correlation (XC) approximations, PBE and BLYP. The resulting functionals of varied FE are detailed in Table . PBE and BLYP were chosen as they are the XC approximations behind many popular (double-)hybrids, and as using two different approximations may help us to better indicate the indi- vidual influence of FE or the underlying XC expression. The functionals are analysed through their ability to recover the minimum of the dissociation curve of each excimer which can be separated into the equilibrium distance (r e ) between the two monomers and the stability of the dimers represented by the dissociation energy (D e ). The ability of each method to describe these parameters is assessed by quantitative comparison to those calculated with our SCS-CC2/CBS (3,4) references by use of signed percentage error: |
623179fd8ab37333ac6662a7 | 27 | where i represents either D e or r e . In this definition of percentage error, positive values represent overestimation while negative ones represent underestimation of the quantity. Fig. details the resulting signed percentage errors for D e (top panel) and r e (bottom panel). The numerical values and corresponding dissociation energy curves are given in Section SI.4. † As Fig. shows, the percentage errors in D e and r e are considerably large across all systems regardless of FE percentage. Almost all functionals underestimate D e , with a magnitude that tends to increase with amount of FE and system size, averaging across structures these errors range from about -83 to -38%. The size-dependence of the error can be partially explained with the fact that dispersion effects are expected to increase with system size and that dispersion corrections have not been applied at this stage. While percentage errors in r e are predominantly positive, negative errors seem to be associated more with higher amounts of FE, with average magnitudes ranging from 4 to 17%. Recovery of each minimum characteristic does not meet conveniently at the same functionals, i.e. a good description of r e is not accompanied by small errors in D e across all the functionals. Despite the overall poor performance from global hybrids regardless of FE percentage we can still gain an insight into the impact FE has for the complicated effects of excimer binding. |
623179fd8ab37333ac6662a7 | 28 | Functionals with large amounts of FE tend to worsen the description of excimer binding across the four tested systems. Functionals with 20% FE yield the smallest errors in D e for each excimer, but the errors do increase with system size, yielding errors that range from -70.8 to 3.4%. Functionals containing 75% FE (-97.5 to -41.7% error range in D e ) are comparable to those with 0% (-97.4 to -22.7% error range in D e ) which are wellknown to produce large errors for excited states and offer no exception here. GGAs are therefore not explored in the benchmarking study of the following section. The poor performance associated with 37.5-75% FE, however, is somewhat surprising given that global hybrids with large amounts of FE generally describe CT and other long-range excitation effects better than smaller amounts of FE. High percentages of FE combined with PBE exchange and correlation appear to improve the description of r e in larger excimers. However this improved description of r e cannot overcome the poor description of D e by these functionals. |
623179fd8ab37333ac6662a7 | 29 | The PBE-based functionals generally yield larger dissociation energies than BLYP-based functionals and are closer to the reference despite still underestimating the excimer stability; for example, PBE20 yields errors in D e ranging between -62.1 and 3.4% while B3LYP errors range between -70.8 and -21.7%. A notable difference is seen for BLYP and BLYP75; they are unable to bind the anthracene and pyrene excimers. Interaction energy curves of CCS, which can be thought of as the HF equivalent for electronic excited states, were plotted to offer a comparison to a method with no electron-correlation but 100 % FE (see Fig. ). The CCS interaction energy curves possess only very shallow minima for the anthracene (D e = 0.31 kcal/mol) and pyrene (D e = 0.70 kcal/mol) excimers indicating that electron correlation plays a significant role in the stabilisation of those systems. PBEbased functionals can be thought to better describe the electron correlation contributions to the excimer states than BLYP based functionals. The difference between the two underlying XC functionals for the excimer state therefore parallels that of ground states, where PBE is more attractive than BLYP in the treatment of NCIs. No simple trend in performance with increasing FE is observed for the description of both D e and r e . Despite the generally improved performance of PBE based functionals, DFAs based on both XC expressions greatly underestimate the dissociation energy and overestimate r e with error trends that increase with FE and system size indicating the requirement of more sophisticated functionals for an accurate description of excimer binding. The ability of a global hybrid TD-DFA to describe each quantity seems to require a trade-off in accuracy to the other. As both D e and r e must be correctly described in order to predict excimer binding, global-hybrid TD-DFAs do not offer reliable results. Methods from the higher rungs of Jacob's Ladder and additive dispersion cor- rections have been effective in addressing some of short-comings associated with global hybrids for ground states, so their performance will be addressed in the following sections. |
623179fd8ab37333ac6662a7 | 30 | In this section, the interaction energy curves for each excimer across a range of TD-DFAs are analysed relative to the reference of SCS-CC2/CBS (3,4). The chosen range of TD-DFAs is based on either popularity, established accuracy for singlemolecule excitations, or novelty; they are: B3LYP, PBE38, BH-LYP (global hybrids), CAM-B3LYP, ωB97X (RS hybrids), B2PLYP, B2GP-PLYP (global DHDFAs), ωB2PLYP and ωB2GP-PLYP (RS-DHDFAs). These functionals occupy the top two rungs of Jacob's Ladder with varying exchange-correlation components and FE percentages, with and without RS, as detailed in Table . For all the dissociation curves discussed, the values associated with the minima, i.e. D e and r e , are listed in Table . Dissociation curves are shown in Figs. . ωB97X interaction energy curves are not shown here, but instead in Fig. † due to observed problems with getting smooth curves that could not be fixed with any of the usual convergence techniques. ωB97X will however be discussed as part of the overall statistical analysis (Section 4.4) with the values based on the actually observed minima under the assumption that they are good approximations to the true minima. Hereafter, deviations from the reference and associated percentage errors will be given as absolute values. Indication to underor overestimation will be discussed qualitatively. |
623179fd8ab37333ac6662a7 | 31 | For the benzene excimer, excited-state interaction energies calculated across varied inter-monomer distances at the SCS-CC2/CBS (3,4) level of theory (Fig. ) yield a dissociation energy of 13.65 kcal/mol at 2.97 Å (r e ). The ability to describe the shape and depth of the potential energy well differs from DFA to DFA. The assessed global hybrids (left panel in Fig. ) consistently underbind the benzene excimer, underestimating D e by 2.97-7.96 kcal/mol (22-58% error) and overestimating r e by 0.16-0.20 Å (5-7% error), corresponding to the largest errors for this system. |
623179fd8ab37333ac6662a7 | 32 | As discussed in the FE study, global hybrids are not capable of capturing the binding interactions of the benzene excimer. Poor performance of global hybrids is consistent with their inability to accurately describe the CT, exciton coupling and dispersion interactions inherent to excimer binding. The long-range correction in CAM-B3LYP reduces the overestimation in r e compared to its uncorrected counterpart B3LYP with an improvement from a deviation of 0.18 to 0.12 Å (central panel in Fig. ); however it also worsens the description of D e (2.97 kcal/mol underestimation for B3LYP vs. 5.08 kcal/mol for CAM-B3LYP). Double hybrids yield an improved description of interaction energies along the dissociation curve, with B2GP-PLYP being closer to the reference than B2PLYP (central panel in Fig. ), both of which underestimate D e by 1.66 and 2.29 kcal/mol, respectively (12 and 17% error). Double hybrids provide a more balanced description of the excimer binding than global and RS hybrids. This improvement is consistent with benchmarking trends of double-hybrid robustness for excitation energies and absorption spectra due to their perturbative correction. Improved performance by DHDFAs additionally parallels the finding that only double hybrids were able to properly describe an exciton-coupled ECD spectrum, for which WFT methods had to be applied earlier. However, as recently re-emphasised, global double hybrids still fail to correctly describe CT excitations. The combination of range-separation and perturbative correction improves the description of the benzene excimer dissociation energy curve, giving the curves closest to the reference at all intermolecular distances (right panel in Fig. ). The assessed RS-DHDFAs, ωB2PLYP and ωB2GP-PLYP, improve upon their uncorrected counterparts, with the former underestimating D e by 0.58 kcal/mol (4% error) and the latter slightly overestimating it by 0.25 kcal/mol (2% error). Close comparison of RS-DHDFAs with the reference interaction energy curves of the benzene excimer display potential to corroborate claims of their robustness for local-valence and long-range excitations. . In summary, it is observed that, for the benzene excimer, the description of binding is improved by climbing Jacob's Ladder with rung-five functionals showing considerable improvement over those belonging to rung four (see Table for the Jacob's Ladder classification of each assessed functional). |
623179fd8ab37333ac6662a7 | 33 | The naphthalene excimer is more than twice as stable as the benzene excimer, yielding a SCS-CC2/CBS (3,4) dissociation energy of 28.32 kcal/mol at 3.04 Å (Fig. ). When assessing DFA dissociation energies, one has to consider the well-documented issue of conventional TD-DFT methods often struggling to correctly order the first two excited states of the naphthalene monomer. For instance, TD-B3LYP and TD-BHLYP incorrectly predict the 1 L a state of the monomer to be lower than the 1 L b state. In the fully-stacked dimer, the first excimer state has, in fact, 1 L a character, but at the dissociation limit this turns out to be the second excited state. Indeed, we observe this for our reference method and most other methods, but want to point out that the aforementioned problem for the naphthalene monomer is also observed here, which can lead to the incorrect calculation of dissociation energies for B3LYP and BHLYP if this problem is not spotted. This is particularly a problem when a calculation is carried out without any symmetry, but less so if the programs used distinguish between excited-state symmetries, as the symmetries of the first two excited states differ. |
623179fd8ab37333ac6662a7 | 34 | The chosen exchange-correlation functionals uniformly predict a weaker binding with absolute errors of D e ranging between 3.65 and 16.53 kcal/mol (13-58% error). r e is largely overestimated with absolute deviations ranging from 0.01 to 0.33 Å (0.3-11% error), the exception being PBE38, which underestimates the distance by only 0.01 Å. The fourth-rung functionals behave differently between naphthalene and benzene excimers whereas fifth-rung functionals exhibit similar performance trends. For the naphthalene excimer PBE38 offers an r e closer to the reference than the BLYP based functionals, however underestimation of D e is on par with other DFAs belonging to this rung (46% error), consistent with results from our previous FE study in Section 4.1. B3LYP binds the excimer with a D e comparable to those of other global hybrids (48% error) while its long-range corrected counterpart, CAM-B3LYP, again improves the description of r e (reduced deviation from 0.33 to 0.20 Å) but yields D e akin to the global hybrids (53% error). While the relative trends between rung-four functionals differ between benzene and naphthalene excimers, the finding remains that global hybrids provide the worst results. |
623179fd8ab37333ac6662a7 | 35 | Double hybrids improve the description of excimer binding considerably with absolute errors in the order of 21-27% and 3-5% for D e and r e , respectively. Further improvement on conventional DHDFAs results from the inclusion of range-separation with absolute errors in the order of 13-20% and 1-2% for D e and r e , respectively. Despite the closer resemblance of RS-DHDFAs curves with the reference curve (right panel in Fig. ) the binding strength is insufficiently described. The best method, ωB2GP-PLYP underestimates the dissociation energy by 3.65 kcal/mol, which is a larger error than for the benzene dimer in Section 4.2.1. Most likely, this error can be attributed to missing dispersion, as its importance increases with the number of electrons. This will be further discussed in Section 4.4. |
623179fd8ab37333ac6662a7 | 36 | The anthracene excimer is slightly more stable than the naphthalene excimer, with an SCS-CC2/CBS (3,4) dissociation energy of 28.40 kcal/mol at a larger equilibrium monomer separation of 3.18 Å (Fig. ). The herein tested TD-DFAs uniformly underbind the anthracene excimer with absolute errors in D e between 5.71 and 21.52 kcal/mol and overestimate r e by between 0.05 and 0.43 Å. Trends in functional performance are similar to those of the naphthalene excimer, although with larger errors in D e . Each functional predicts a smaller excimer dissociation energy for the anthracene excimer than it does for the naphthalene excimer, so comparison with the larger reference value for the anthracene excimer results in larger errors for D e (20-76%). |
623179fd8ab37333ac6662a7 | 37 | It was previously noted that the D e of the anthracene excimer being larger than that of the naphthalene excimer may be due to a larger dispersion contribution to the excimer binding. Given that the TD-DFT methods discussed in this sections are unable to describe dispersion, this further supports the study of dispersioncorrected functionals for excimer binding in Section 4.3. |
623179fd8ab37333ac6662a7 | 38 | The pyrene excimer (Fig. ) is more stable than the anthracene excimer, with an SCS-CC2/CBS(3,4) dissociation energy of 36.33 kcal/mol at 3.19 Å. Similarly to the naphthalene dimer, we also observed the wrong order of states at the dissociation limit for some functionals, something that can be avoided when using symmetry in the calculation. |
623179fd8ab37333ac6662a7 | 39 | Each functional predicts similar dissociation energies as for the anthracene case (within 1.20-3.71 kcal/mol), which relative to the larger reference D e of the pyrene excimer yields larger errors (28-77%). The trends observed for the naphthalene and anthracene excimers hold true for the pyrene excimer, with an improved description of excimer stabilisation moving up Jacob's Ladder. |
623179fd8ab37333ac6662a7 | 40 | To summarise Section 4.2, the relative performance of the tested functionals follows a general trend: global hybrids and CAM-B3LYP perform the worst in terms of D e and r e , with a slight improvement in r e through range-separation; double hybrids improve upon the description of both well characteristics while range-separated double hybrids show the best description of both characteristics across the tested functionals. We saw an increase in the error of D e with system size which may be due to the lack of properly treating dispersion interactions. In the following section we explore the application of ground-state optimised dispersion corrections to assess their potential in accounting for the missing dispersion in excimer binding. |
623179fd8ab37333ac6662a7 | 41 | So far we have only discussed dispersion-uncorrected results. As dispersion-uncorrected TD-DFAs are able to partially predict excimer binding it is clear that interactions beyond dispersion are important, which aligns with the findings from an energy decomposition analysis reported in 2018. In our introduction we have pointed out how some studies make use of dispersion-uncorrected methods for the treatment of excimers, but our results clearly show that dispersion effects should not be neglected. Herein, we demonstrate the impact of ground-state dispersion corrections for the two extreme cases in our study, namely the benzene dimer as the smallest and the pyrene dimer as the largest system. |
623179fd8ab37333ac6662a7 | 42 | Ground-state dispersion corrections are often applied to excited-state studies justified by seemingly good outcomes from a few studies. However, it is reasonable to expect dispersion interactions to change upon electronic excitation 113 rendering any benefit application-dependent. Herein, we test these justifications for two schemes of additive DFT-D type dispersion corrections as applied to the excimer state without any statespecific adjustment. The DFT-D3(BJ) and DFT-D4-hereafter dubbed "D3(BJ)" and "D4"-dispersion energies were calculated for all points along the dissociation curves and added to the respective total excited-state energies. The analysis of these dispersion-corected dissociation energy curves focuses on four functionals B3LYP, CAM-B3LYP and B2PLYP and ωB2PLYP, i.e. two global functionals and their range-separated counterparts (minima given in Table ). A more generalised picture across all eight dispersion-corrected functionals will be provided in the overall statistical analysis in Section 4.4. |
623179fd8ab37333ac6662a7 | 43 | For the benzene excimer, both dispersion corrections overcorrect the binding predicted by each TD-DFA except ωB2PLYP (Fig. ). Whether D4 or D3(BJ) is closer to the reference depends on whether the dispersion-corrected method under-or overbinds the excimer; if the former is the case, the D4 version is closer to the reference, but in the latter case the D3(BJ) variant is closer. Across the four functionals shown here, absolute deviations in D e range from 0.46 to 7.82 kcal/mol for D4 and from 0.46 to 7.58 kcal/mol for D3(BJ). The corrections offer an improved but overestimated description of the optimal inter-monomer separation for B3LYP, CAM-B3LYP, and B2PLYP with errors between 0.03 and 0.07 Å for D3(BJ) and between 0.04 and 0.06 Å for D4. However, the improvement to r e is drastically over-shadowed by the greatly overstabilised dissociation energy. Dispersion-corrected and -uncorrected ωB2PLYP give the same overestimation of r e by 0.02 Å. As the smallest of the model systems, benzene excimer binding contains a smaller dispersion contribution than the larger models. With less dispersion to account for, more complex functionals are able to reasonably account for the other contributions to the dissociation energy without dispersion correction. The ground-state optimised corrections, thus, have a tendency to overestimate the dispersion contribution to the benzene excimer dissociation energy. |
623179fd8ab37333ac6662a7 | 44 | The pyrene excimer is larger and more strongly bound than the benzene excimer, which uncorrected TD-DFAs fail to predict with errors ranging from 13.21 to 26.62 kcal/mol (Fig. ). For this larger system, dispersion-corrected TD-DFAs offer some improve- Both dispersion corrections considerably improve the B3LYP dissociation energy to within 2.91 kcal/mol accompanied by an improved r e with an overestimation of 0.12 Å. Dispersioncorrected CAM-B3LYP does not show the same improvement to the dissociation energy as its global counterpart, yielding an underestimation of up to 8.93 kcal/mol for D e and an overestimation of r e by 0.14 and 0.13 Å for D3(BJ) and D4, respectively. B2PLYP-D3(BJ) and -D4 give dissociation energies with errors comparable to dispersion-corrected B3LYP (underbound by 2.05 and 1.73 kcal/mol), but with a smaller geometry error (overestimation of 0.07 and 0.08 Å). Dispersion corrected ωB2PLYP offers only a small improvement to the uncorrected functional such that ωB2PLYP-D3(BJ) and ωB2PLYP-D4 underbind the pyrene excimer by 12.70 and 12.67 kcal/mol, respectively, and overestimate r e by 0.08 Å. A negligible impact of D3(BJ) and D4 on both ωB2PLYP and ωB2GP-PLYP has also been noticed in ground-state benchmarking on the complete GMTKN55 database and its NCI category. For the larger system size of the pyrene excimer, the groundstate dispersion corrections improve TD-DFA dissociation energies and equilibrium distances. This is likely why others have recommended the use of ground-state optimised dispersion corrections for the calculation of this excimer's dissociation energy. While DFT-D type dispersion corrections offer some improvement to the description of excimer binding, their performance is far from a black-box solution for excited states. Inconsistent improvement by ground-state dispersion corrections for excited states reinforces the need for state-specific dispersion corrections for reliable and predictive TD-DFT methods. |
623179fd8ab37333ac6662a7 | 45 | Beyond the binding region, in the medium-to-long range inter-monomer separation, TD-DFA interaction energy curves falsely predict a repulsive region. For the benzene excimer, all functionals as well as SCS-CC2/def2-TZVP exhibit that positive region falling between 4.3 Å and the asymptote. An extrapolation of SCS-CC2 to the CBS limit [CBS (3,4)] mostly corrects the unphysical repulsion of SCS-CC2/def2-TZVP, reducing the repulsion to less than 0.01 kcal/mol, which is way within the expected numerical noise for that method and a negligible value (see Fig. ). Larger excimers exhibit a similar repulsive region for global hybrids and range-separated hybrids (see Figs. S14-S17 † for further details). Herein, we focus on the benzene excimer by observing B3LYP and B2PLYP with and without dispersion corrections in the range between 4.00 and 12.00 Å (Fig. ). The extent of the repulsion for dispersion-uncorrected B3LYP is over twice that of B2PLYP, with maxima of 0.35 and 0.12 kcal/mol, respectively. The D3(BJ) correction reduces the repulsion yielding maxima of up to 0.02 kcal/mol, which brings the two functionals into close comparison with each other (maxima within 0.001 kcal/mol of each other). The D4 curves are also comparable in this region, although they exhibit larger maxima than D3(BJ) curves: 0.07 and 0.06 kcal/mol for B3LYP-D4 and B2PLYP-D4, respectively. The positive behaviour that remains after the applied corrections would not be expected with corrections appropriately parametrised for excited states. Interestingly, in a recent study of exciplex interaction energy curves, a similar unphysical repulsive hump exhibited by ωB97 also saw correction by ωB97X-D3 196 with DFT-D3(0), something that has not been noticed by the authors. As excimer binding is comprised of more than just dispersion effects, 8 dispersion-uncorrected functionals are able to predict some attraction in the binding region, with more complex functionals performing reasonably in binding the benzene excimer. In the medium-range, however, the dispersion-uncorrected functionals predict an unphysical repulsive region. Medium-range inter-monomer separations appear to be governed by dispersion, and can therefore only be accounted for by inclusion of the dispersion energy missing from the description by dispersion-uncorrected TD-DFAs. |
623179fd8ab37333ac6662a7 | 46 | With this closer look at ground-state dispersion corrections for the binding of benzene and pyrene excimers, it is clear that accounting for dispersion is important for accurate calculation of dissociation curves in both the binding region and beyond. While ground-state optimised dispersion corrections offer some improvement to excimer binding, a state-specific reparametrisation for excited states would be necessary for robust and reliable TD-DFA results. An overarching discussion of the performance of dispersion corrections for all benchmarked DFAs is presented in the following section. |
623179fd8ab37333ac6662a7 | 47 | Following our previous analysis of the interaction energy curves, we continue with an overall discussion of each functional's performance averaged across all systems both with and without dispersion corrections. Functional performance is almost universally assessed by mean absolute deviations (MADs) as the metric for benchmarking of quantum-chemical methods. Here, the MAD for each characteristic averaged for each method is calculated by following the general form: |
623179fd8ab37333ac6662a7 | 48 | where MET HOD i and REFERENCE i are the values of the property for the ith excimer, either D e or r e , and N the number of systems. An analysis of MADs is simply presented as a convenient metric to summarise the performance of a method across our four model structures, and we acknowledge a sample size of four does not offer MADs with statistical significance comparable to more comprehensive excited-state benchmarks. MADs for the nine functionals, with and without D3(BJ) and D4 dispersion corrections, compared to the SCS-CC2/CBS (3,4) reference are presented in Fig. , with corresponding numerical values reported in Table . † For the description of D e and r e , dispersion-uncorrected functionals display a descending trend for the MADs while ascending Jacob's Ladder (Fig. ). It should be noted that, for dispersion-uncorrected functionals, these deviations overwhelmingly correspond to underestimations of D e and overestimations of r e , whereas dispersion-corrected functionals vary in this regard (see ESI †). The largest errors in D e correspond to global hybrids and CAM-B3LYP yielding MADs in the order of 15.38 to 17.93 kcal/mol. Similarly, the largest MADs in r e range between 0.23 and 0.36 Å and are attributed to CAM-B3LYP, BHLYP and B3LYP, i.e. the global and RS hybrids with Becke88 exchange and LYP correlation. PBE38 gives a much smaller MAD of 0.10 Å, comparable to higher-rung DHDFAs. |
623179fd8ab37333ac6662a7 | 49 | For global hybrids, an increase in FE is associated with an increase in MAD for D e , corroborating the results of the FE study (Section 4.1). This increase in deviation supports the tendency of global hybrids with large amounts of FE to cause issues for describing excitations in TD-DFT. For example, global hybrids with large components of FE tend to produce blue-shifted excitation energies in single molecules. Despite the established better description of CT excitations with range-separated hybrids, CAM-B3LYP performs worse than B3LYP. On the other hand, ωB97X, despite the aforementioned problems with obtaining smooth dissociation curves, proves to be the best tested dispersion-uncorrected hybrid in this study with MADs of 10.45 kcal/mol and 0.17 Å. |
623179fd8ab37333ac6662a7 | 50 | We have already mentioned problems in state order for the naphthalene and pyrene dimers in Sections 4.2.2 and 4.2.4, but still need to elaborate more in detail on the fact that the dimers consist of polycyclic aromatic hydrocarbons (PAHs). Singlemolecule studies of PAHs have established that the first two π-π * excitations in PAHs-called L a and L b according to Platt 201 -are poorly described by many TD-DFT methods, including global hybrids. RS hybrids can improve the description of L a but blueshift L b excitations. Double hybrids, particularly the latest range-separated ones, have to date yielded the most accurate and balanced description of both states in PAHs. This, in addition to the aforementioned better description of exciton coupling, Its various dispersion-corrected variants all depend on slightly different underlying XC expressions, which is why dispersion-corrected results are not provided. |
623179fd8ab37333ac6662a7 | 51 | The application of D3(BJ) and D4 dispersion corrections shows significantly reduced MADs across all tested functionals, excluding ωB2(GP-)PLYP, with larger reductions observed for D e than for r e . However, the improvement to excimer binding by the tested dispersion corrections shown through this reduction in MADs is not predictable. While dispersion-uncorrected functionals yield errors due to underestimation of D e and overestimation of r e , dispersion corrections under or overcorrect depending on the system and functional. D4 and D3(BJ) yield similar D e MADs in the order of 1.96-6.58 kcal/mol and 1.48-6.59 kcal/mol, respectively. For the same underlying functional, the two iterations of DFT-D yield comparable MADs where differences between D4 and D3(BJ) MADs do not exceed 0.03 Å or 1.16 kcal/mol, with the exception of B2GP-PLYP where D4 gives considerably higher D e deviations than D3(BJ). |
623179fd8ab37333ac6662a7 | 52 | Dispersion-corrected global double hybrids give the best description of excimer binding with greatly improved r e (ranging from 0.05 to 0.07 Å) and the smallest devions in D e despite their tendency to overbind the excimer systems: their MADs for D e range from 1.48 kcal/mol [B2GP-PLYP-D3(BJ)] to 3.77 kcal/mol [B2GP-PLYP-D4]. Overestimation of exciplex binding from dispersion-corrected double hybrids was also noted in the study by Krueger and Blanquart where B2PLYP-D3(BJ) significantly overbound the exciplexes analysed which the authors acknowledged was largely due to the inclusion of a dispersion correction. Their tendency to overestimate D e suggests that excited-state parametrised dispersion corrections would be necessary for a reliable description by TD-DFT methods. |
623179fd8ab37333ac6662a7 | 53 | Dispersion-corrected RS-DHDFAs only slightly improve on dispersion-uncorrected results, with MADs for D e that range from 4.90 [ωB2GP-PLYP-D3(BJ)/D4] to 6.59 kcal/mol [ωB2PLYP-D3(BJ)]. MADs in r e are unchanged from dispersion-uncorrected RS-DHFAs (0.04-0.05 Å) remaining the best performing across all functionals dispersion-corrected or otherwise. As the first application of ground-state parametrised dispersion corrections to RS-DHDFAs for excited-state problems, these functionals display potential for TD-DFT dissociation energies that approach SCS-CC2/CBS(3,4) quality. However, their tendency to underestimate D e reinforces the necessity of excited-state parametrised dispersion corrections for a reliable description by TD-DFT methods. |
623179fd8ab37333ac6662a7 | 54 | From the discussion of dispersion-uncorrected functionals, double hybrids and range-separated double hybrids performed well for the more complicated excited-state interactions comprising excimer binding. As with range-separated double hybrids, the ground-state parametrised dispersion corrections do not reliably account for excited-state dispersion interactions, but without accounting for this missing dispersion TD-DFT cannot provide reliable results for excited states. While with current dispersion corrections both types of DHDFAs seem to be adequate, it is safe to assume that the better description of CT with RS-DHDFAs means that they will prevail once paired with state-specific corrections. |
623179fd8ab37333ac6662a7 | 55 | Global hybrids and CAM-B3LYP show a significant error reduction upon dispersion correction yielding MADs for D e in the order of 3.20-5.12 kcal/mol, comparable to the MADs of higherrung dispersion-corrected functionals, alongside an improved but still overestimated r e with MADs ranging from 0.05 to 0.12 Å. The dispersion corrections offer considerable improvement to excimer binding for these less sophisticated functionals which gave the largest MADs when dispersion-uncorrected. However, even with dispersion-corrected, global hybrids and CAM-B3LYP are still largely outperformed by DHDFAs. PBE38 seems to be able to compete with the best methods, but we would like to reiterate that it does not describe CT nor do related PBE-based hybrids describe exciton coupling correctly, which indicates that its good performance is most likely influenced by fortuitous error cancellation. |
623179fd8ab37333ac6662a7 | 56 | The binding of four different aromatic excimer models was analysed by means of dissociation curves. To our knowledge, this is the first study to provide single-reference wave function curves at the complete basis set (CBS) limit for aromatic excimer systems. More specifically, linear SCS-CC2/CBS (3,4) was established as a reliable reference that allowed us to shed light onto various TD-DFT methods including a detailed analysis of excimers with double-hybrid density functionals and the first application of range-separated double hybrids to such systems. Our main goal was to address the impact of Fock exchange, range separation, the perturbative nonlocal correction used in double hybrids, and London dispersion corrections. We analysed two quantities that characterise the minima along the dissociation curves, namely dissociation energies D e and inter-monomer equilibrium distances r e . |
623179fd8ab37333ac6662a7 | 57 | Overall, it turned out to be challenging to obtain a good description of both D e and r e with hybrid functionals, whereas double hybrids provided a more robust picture. High admixtures of Fock exchange in global hybrids usually led to smaller dissociation energies or even repulsive curves. Dispersion-uncorrected global hybrids with 50 % or less Fock exchange, range-separated hybrids, as well as global and range-separated double hybrids, all gave curves with distinct minima, with the latter functional type giving the best curves, most likely due to a better description of charge-transfer and exciton coupling. That being said, all dispersion-uncorrected TD-DFT methods produced large errors that increased considerably with system size. For the smallest system, the benzene excimer, dispersion-uncorrected methods see the best results from range-separated double hybrids, approaching the accuracy of SCS-CC2/CBS (3,4) quality. However, for larger systems there was greater disparity between even the most accurate TD-DFT methods and the reference. D e values were usually underestimated and r e values overestimated, which points to missing dispersion interactions as the likely reason. |
623179fd8ab37333ac6662a7 | 58 | The application of ground-state parametrised dispersion corrections generally reduced the errors of dispersion-uncorrected functionals but did not reach chemical accuracy most likely due to not having been designed to describe excited states. We have noticed overstabilisiation of some systems for some functionals, which occurred more often for the smaller systems. For some methods and systems, improvements of the minimum-energy regions were observed, but dissociation energies were still underestimated. To our knowledge, we were also the first to point out that most TD-DFT methods were unphysically repulsive in the mid range, something that could be reduced by applying dispersion corrections, which in turn indicated that dispersion was the most dominant contribution in that range. |
623179fd8ab37333ac6662a7 | 59 | Our study has shown that some of the latest and most modern TD-DFT methods, namely range-separated double hybrids, belong to the most robust and accurate when treating excited states, which parallels single-molecule studies. However, we have also shown that there is a need for the development of state-specific London dispersion corrections to achieve a reliable and robust TD-DFT description of excimers and related systems for all tested methods, including range-separated double hybrids. |
6305f16cf9e99c5db387eeb5 | 0 | crystallographic and spectroscopic data are shown to coalesce to the same picture of a predominant S=6 species containing the first one-electron oxidation product of two water molecules i.e. [O5O6] 3-. Progression of this form to the two-electron oxidised peroxo and three-electron oxidised superoxo forms, leading eventually to the evolution of triplet O 2 , is proposed to be the pathway Nature adopts to oxidise water. The study reveals the key electronic, magnetic and structural design features of Nature's catalyst which facilitates water oxidation to O 2 under ambient conditions. |
6305f16cf9e99c5db387eeb5 | 1 | Introduction Every oxygen molecule we breathe is produced from two water molecules in the photosystem 2 protein complex of higher plants, algae and cyanobacteria. This highly endothermic reaction is carried out during photosynthesis using visible light energy under ambient conditions. To perform this task a unique water oxidising catalytic complex, Mn 4 CaO 5/6 , evolved some three billion years ago. This complex oxidises two water molecules to molecular oxygen at a rate approaching 1000 s -1 at ambient temperatures and pressure. Besides being one of the most important reactions in biology, it is also of intense interest from a green energy perspective, where it is recognized to be the main barrier to the development of commercial solar devices for the generation of hydrogen from water. Water oxidation to dioxygen is a challenging due to the high endergonicity (E° = 0.82 V (vs NHE) at pH 7) for the reaction, 2H 2 O → O 2 + 4H + + 4e - and the associated need to remove four protons and four electrons with the formation of an oxygen-oxygen, O-O, bond. Two broad mechanistic proposals, either water nucleophilic attack of metal oxo or direct metal oxo radical coupling, have been proposed for artificial water oxidation catalysts. Somewhat similar proposals have been put forward for the WOC namely water nucleophilic attack or oxyl radical-oxo coupling These require the generation of a reactive oxo species in the final Kok-cycle S 4 state. Artificial catalysts generally use very high strength oxidising agents to generate reactive oxo species, either radical oxygen species or highly charged metal electrophilic species. The WOC on the other hand is limited to the approximately 1 V oxidising power of the nearby tyrosyl radical, Y Z OX . The current mechanisms for the WOC which propose generation of a reactive "hot" oxo species in the S 4 state need to explain how such a species can be generated when the oxidising capability available from the visible light energy available via the S 3 Y Z ox oxidant is around 1V. It is also unclear how triplet O 2 can be produced from peroxo with such a mechanism given that the last oxidising equivalent has been used. |
6305f16cf9e99c5db387eeb5 | 2 | Our starting point on the pathway to O5-O6 bond formation is an O5 oxo O6 hydroxo form of the WOC complex, Figure , formed after initial formation of the S 3 state. This corresponds to the S=3 form detected by EPR with four Mn (IV) ions. O6 corresponds to the new oxygen atom detected by XFEL after the second flash. For this oxo-hydroxo model seven broken symmetry, Ms, states are possible at the optimised geometry. We have shown that two of these, both M S =3, [Mn4(↓↓↓)Mn3(↑↑↑)Mn2(↑↑↑)Mn1(↑↑↑)] and [Mn4(↑↑↑)Mn3(↓↓↓)Mn2(↑↑↑)Mn1(↑↑↑)] are the lowest in energy and govern the spin density of the complex resulting in a spin distribution of close to 0.5, 0.5, 0,0 and 0.0 for Mn1-Mn4. This explains the set of two large (Mn1 and Mn2) and two small (Mn3 and Mn4) magnitude Mn hfcs observed using EDNMR spectroscopy. For the oxo-hydroxo form no changes are found in the IBOs in the region of 2.5 -2.0 Å and as mentioned above this model becomes unstable at bond distances less than 2.0 Å. By contrast significant changes are observed for both oxo-oxo forms. The IBOs which undergo significant changes are located by monitoring the root-mean-square deviation of every IBO from the initial partial charge distribution along the PES. Figures and identifies four main IBOs participating in bond making and breaking during the reaction. These are, the α and β spin orbitals of the Mn 4 O5 σ-bond, the β spin orbital of the Mn 1 O6 σ-bond and the α spin orbital of one of the π-bonding lone pair orbitals on O6. As the O5-O6 bond distance is decreased from the non-bonded oxo-oxo form, the α electron density of the Mn 4 O5 sigma bond evolves into a dz 2 orbital on Mn 4 , Figures and (pink), at an O5-O6 bond distance of around 2.2 Å for Ms=6 and 2.0 Å for Ms=3. Concurrently with this electron density rearrangement, the α density of the π-lone pair on O6, evolves to a σ bond between the O6 and O5 oxygens, Figures and (blue). A Mayer bond order analysis, Figure , also illustrates such a change with a decrease in the Mn 4 -O5 bond order from near 1.0 to near 0.5 and an increase in the O5-O6 bond order from 0.0 to near 0.4. In a similar fashion, Mulliken spin population analysis, Figure shows a change in spin population of Mn 4 from near 3.0 to 4.0 signalling a reduction from Mn 4 (IV) to Mn 4 (III). For the Ms=3 state, further progression along the PES shows the β-electron of the Mn 4 -O5 σ-bond evolves to become the β component of the O5-O6 σ-bond, Figure (yellow), and the Mn 1 -O6 σ-bond β electron density evolves into a dz 2 orbital on Mn 1 , Figure (green). For Ms=6, the β-electron of the Mn 4 -O5 sigma bond again evolves to become the β component of the O5-O6 sigma bond, Figure (yellow), while in this case a Mn 1 -O6 π-bond β electron density evolves into a dπ orbital on Mn 1 , Figure (green). Mulliken spin populations, Figure , correspondingly show an increase in spin population from 3 to 4 for Mn1 illustrating reduction of Mn1 to high spin Mn(III) for Ms=3 whereas for Ms=6 the electron transfer of a β electron to Mn1 results in occupation of a dπ orbital, Figure (green), resulting in a spin population of 2 and corresponding to a low spin form of Mn(III). The Mayer bond order values for both Ms states of Figure show an increase in the O5-O6 bond order to near 1.0 as the peroxo is formed. |
6305f16cf9e99c5db387eeb5 | 3 | Our key finding is that for the Ms=6 oxo-oxo form an electronic state corresponding to Mn4(↑↑↑↑)Mn3(↑↑↑)Mn2(↑↑↑)Mn1(↑↑↑)[O5O6](↓) is found as a shallow local minimum at an O5O6 distance of 2.0 Å. The IBOs, Figure , show that electron movement has occurred from O5 to Mn4 leading to a high spin Mn4 (III) and formation of a nascent two centre one electron O5-O6 bond. This species was identified previously by us as a shoulder on the Ms=3 state, see Figure . While a shoulder on the Ms=3 PES, it corresponds to a broad minimum energy structure on the Ms=6 surface due to the favourable antiferromagnetic coupling with all four Mn ions. We note that this species has been referred by us and others previously as an O5 oxo O6 oxyl form but is best and more appropriately described as [O5O6] 3-since negative spin density is present on both O5 and O6, clearly demonstrated by the spin density plot for this form in Figure . |
6305f16cf9e99c5db387eeb5 | 4 | Suga et al first reported an O5-O6 bond distance of 1.5 Å indicating peroxo formation. Later studies proposed an additional oxygen O x similar to O6 of the Suga et al structure but with an extended O5-O6/O x bond length of 2.1 Å. More recently Suga et al propose a best fit O5-O6 bond length of 1.9 Å. All structures of S 3 so far appear to rule out an oxohydroxo non-bonded form which requires an O5-O6 bond separation of at least 2.5 Å. The interpretation of the S=3 signal EPR from the 2-flash state based on BS-DFT analysis of calculated hfcs is highly indicative of an oxo-hydroxo form for the S 3 state. An oxohydroxo model is not however compatible with the structure obtained by XFEL. As described above the [O5O6] 3-model does agree with the XFEL structures. This corresponds to a broken symmetry Ms=6 spin state. This is not a true spin state, S. The true spin state energies can be obtained by diagonalization of the Heisenberg Dirac van Vleck Hamiltonian using J values obtained by analysing all possible BS states. Table shows the calculated J values and energies of the ground spin states using this procedure. From this an S=6 spin state is calculated to be the ground state spin. This therefore cannot be attributed to the species observed by EPR/EDNMR which has an S=3 ground state spin. The PES shows that the two species are related by the protonation state of O6. Intriguingly an S=6 species was proposed to be formed in the 2-flash S 3 state of spinach samples and was proposed to be the major component (80%) of native samples. The S=6 form was attributed to a so-called closed cubane form of the WOC cluster with a pentacoordinated Mn 4 (IV) ion formed before the second substrate binds. So far no structural experimental support for such a closed cubane structure of the WOC has been obtained for any S-state. It is therefore more likely (see below) that this species corresponds to the [O5O6] 3-form alluded to in this manuscript, also with S=6. Experimentally, no S=6 species has so far been reported in cyanobacteria samples where high resolution high field W-band EPR spectra obtained are attributed to an S=3 form Simulations of the W-band EPR spectra for the S=3 form are shown in Figure . Also shown are simulations for an S=6 form using the zero field splitting parameters reported for the spinach samples. From the simulations it is clear that the spectral intensity of the S= 6 |
6305f16cf9e99c5db387eeb5 | 5 | form is much less than that of the S=3 form. This suggests that the S=6 form would be difficult to detect in the W-band EPR experiment. More intriguingly as shown in Figure , even at a 70% contribution of the S=6, the S=3 form still dominates the spectral envelope with the S=6 form mainly contributing a distinctive shoulder at around 3500 -4000 mT to the overall spectral shape. It is clear from the spectra presented in Figure of Chrysina et al that a poorer fit between experimental and a simulated S=3 spectrum exists in this very region. Figure shows that inclusion of the S=6 form (70%) gives rise to a much improved fit to the experimental spectrum. Additional simulations varying the ratio of the two spin systems are presented in Figures S7 where we can estimate that an S=6 contribution of between 60-70% is optimal. |
6305f16cf9e99c5db387eeb5 | 6 | We therefore suggest that the seeming incompatibility between the XFEL and EPR data for the S 3 state, lies in the fact that the oxo-hydroxo and [O 2 ] 3-are in equilibrium. The [O 2 ] 3- form detected in the XFEL determined structure is not readily apparent in the EPR spectrum due to it S=6 nature and resultant low intensity compared with the S=3 form. Further simulations presented in the Figure suggest that the S=6 component is also likely a major component of the broadened W-band EPR spectrum caused by methanol and glycerol addition. It has been known for some time that the S=3 species does not correspond to all of the S 3 spin and that an EPR "undetectable" component observed only on near infrared (NIR) irradiation is also present in equilibrium with it. In our analysis, this undetectable component corresponds to the [O 2 ] 3-form. This is a different assignment to that previously made for the S=6 form detected in spinach samples where the S=6 form was attributed to a closed cubane form of the WOC cluster with a pentacoordinated Mn 4 (IV) ion, an intermediate formed prior to binding of the second substrate water. This however, in striking contrast to the model proposed here, is not supported by the XFEL structural data. In addition it has been shown that Mn(III) is required for NIR excitation and the large D value of 1.523 cm -1 for the S=6 form strongly suggests the presence of Mn(III) ion in the complex. It should be noted that it is possible that the peroxo form, Figure , is also present in a low concentration and its EPR spectrum is masked by the oxo-hydroxo form. The peroxo complex would have two Mn(III) ions present likely leading to a large D value similar to the [O 2 ] 3-form which would again lead to a low intensity EPR spectrum compared with the oxohydroxo form. |
6305f16cf9e99c5db387eeb5 | 7 | Further experimental support for our S 3 state model comes from analysis of the X-ray emission spectroscopy (XES) data by Ibrahim et al. The 1F flash first moment XES shift can be confidently assigned to the Mn (III) to Mn (IV) oxidation of Mn 4 . The first moment shift for the 2F state is approximately 40% of the 1F shift based on the solution phase data and the most current time resolved crystal data, see Figure . In addition at least 10% of the oxidation change shift can be attributed to S 1 to S 2 oxidation based on the S state populations of the 1F state given in Abrahim et al. leaving around 30% Mn oxidation occurring in the S 2 to S 3 transition. This is what is predicted by our equilibrium model above. The 30% Mn oxidation can be attributed to formation of the oxo-hydroxo form where a Mn 1 (III) to Mn 1 (IV) oxidation occurs. The [O5O6] 3-form has however an identical overall Mn oxidation state to the S 2 state i.e. one Mn(III) and three Mn(IV) so will not give rise to a first moment shift. |
6305f16cf9e99c5db387eeb5 | 8 | The computational, structural and spectroscopic evidence above all points to an S 3 state involving an equilibrium between an O5 O6H oxo-hydroxo and an [O5O6] 3-species. The most recent XFEL structures for the S 3 state also reveal a very short O6 to Oε2Glu189 distance of 2.4 -2.5 Å suggesting a low barrier hydrogen bond between the two atoms. This strongly indicates that the S 3 state equilibrium is established by proton sharing between these two atoms as illustrated in Figure . Based on our combined computational, spectroscopic and structural analysis, we demonstrate that O-O bond formation has begun between the O5 and O6 atoms in the S 3 state with the generation of the [O5O6] 3-ion. This is the dominant species present in the S 3 state. Figure shows that this provides a low barrier pathway to subsequent formation of the peroxo form. |
6305f16cf9e99c5db387eeb5 | 9 | As indicated above this peroxo ion. This [O 2 ] 3-ion is stabilised by antiferromagnetic interaction with the Mn ions of the complex. Combined computational, crystallographic and spectroscopic data show that an equilibrium exists between an O5 oxo and O6 hydroxo form, S=3 spin state and a deprotonated O6 form containing a two-centre one electron bond in [O5O6] 3-which we identify as the form detected by XFEL crystallography. This form gives rise to an S=6 spin state which gives rise to a low intensity EPR spectrum compared with the accompanying S=3 state, making its detection via EPR difficult and overshadowed by the S=3 form. Simulations assuming a 70% contribution of the S=6 form give rise to a superior fit to the experimental EPR spectrum compared with an S=3 only form. The study reveals the key electronic, magnetic and structural design features of Nature's catalyst which allows water oxidation to O 2 to be uniquely performed under ambient conditions. |
672e12c27be152b1d071afbe | 0 | Methods for sp 2 C-H alkylation, particularly for electron deficient aromatics, are limited with traditional methods such as Friedel-Crafts alkylation reacting preferentially with electron rich organic aromatics. C-H alkylation is usually impractical and requires the use of superstoichiometric organolithium or toxic organomercury reagents (Fig. ). Instead of using a direct C-H functionalisation strategy, a halogen handle is often incorporated into the substrate to enable traditional palladium-based cross coupling reactions or recently developed photoredox nickel cross coupling (Fig. ). Minisci coupling does avoid the need for a functional handle and offers a reliable radical based route to C-H alkylate pyridines. |
672e12c27be152b1d071afbe | 1 | However, it also exhibits significant limitations such as providing solely C2 selectivity and requiring strong acidic conditions limiting its functional group tolerance (Fig. ). The generation of radical species through the photoactivation of electron donor-acceptor (EDA) complexes has been precedented since the 1950s, but it has only recently emerged as a useful tool in organic chemistry and remains underexploited in organic catalysis. EDA catalysis offers the advantage of producing radicals under mild conditions, with highly targeted photoreactivity achievable through substrate design. Initial applications of EDA catalysis were limited to highly specific donor-acceptor pairs, typically requiring a preinstalled leaving group. The introduction of redox auxiliaries (RAs) or 'redox tags' as components of EDA complexes has broadened its synthetic utility and in the presence of a donor species (D) can facilitate a charge transfer absorption band in the visible spectrum. |
672e12c27be152b1d071afbe | 2 | These auxiliaries can be attached to a specific functional group of the substrate molecule, enabling greater generality to EDA methodology across a much wider range of substrates. Upon photoexcitation, the EDA complex undergoes single-electron transfer (SET) to form a radical ion pair, which subsequently fragments to yield the desired neutral substrate radical (S • ) (Fig. ). Despite these advances, a significant limitation of current EDA methodologies lies in their use of the radicals generated upon fragmentation of the EDA complex. Most approaches rely on either highly pre-functionalised substrates for intramolecular reactions or use a narrow set of "radical traps" (e.g. silyl enol ethers, isocyanides, vinyl sulfones) as coupling partners. In this study, we overcome this limitation by demonstrating that electron-poor aromatic molecules can serve as versatile coupling partners for alkyl radicals generated via EDA fragmentation thereby establishing a more general sp 2 -sp 3 coupling method. |
672e12c27be152b1d071afbe | 3 | Our direct C-H functionalisation strategy enables selective modification at the most electrondeficient position on a broad range of electron-poor aromatics in an 'Anti-Friedel-Crafts' alkylation manner (Fig. ). Our photocatalyst-free method utilises inexpensive and commercially available catalysts, without requiring harsh acidic or pyrophoric reagents. The mild reaction conditions tolerate a wide range of functional groups, including halide handles to facilitate further down-stream functionalisation. Mechanistic investigations reveal that the EDA activation between DABCO (1,4-diazabicyclo[2.2.2]octane) and a redox active phthalimide ester (RAE) serves as the trigger for a chain-like, autocatalytic radical reaction. This high regioselectivity and functional group tolerance was exploited for the late-stage modification of pharmaceuticals and agrochemicals (Fig. ). The observed 'Anti-Friedel-Crafts' regioselectivity can be readily tuned with aromatic substituents, as predicted by density functional theory (DFT) calculations and machine learning models. |
672e12c27be152b1d071afbe | 4 | Phthalonitrile was selected as the initial substrate for C-H alkylation as cyano-aryl species are a useful synthon in various electrochemical and photochemical transformations. , Cyanoarenes can also effectively stabilise aryl radicals/anions and thus are also common motifs in photocatalytic organic dyes. The visible-light absorbing EDA complex was constructed from redox active phthalimide esters and the electron rich tertiary amine donor DABCO, which has been selected as our donor (D) due to its known but under-utilised activation of redox active esters (RAEs). The RAEs were synthesised from a range of carboxylic acids, with the phthalimide redox auxiliary (RA) serving as the electron deficient component in the EDA complex. Upon complexation and photo-induced fragmentation, the irreversible loss of CO2 drives the reaction forward entropically, generating a phthalimide anion and a neutral alkyl radical (S • ). This alkyl radial can then selectively alkylate the C4 site of the electron deficient aromatic acceptor. |
672e12c27be152b1d071afbe | 5 | Our goal was to develop a reaction using inexpensive, commercially available donors in conjunction with readily synthesised phthalimide RAEs, avoiding the complex donors/acceptors employed in previous EDA methodologies. This approach led to our standard conditions: DABCO (50 mol%), RAE (1, 1 equiv.), phthalonitrile (3 equiv.), and blue LED irradiation (λmax = 447 nm) in DMSO for 16 h at 25 °C under N2. At a 0.15 mmol scale, the desired product (2) was isolated in 84% yield (Entry 1, Table ). |
672e12c27be152b1d071afbe | 6 | Only catalytic amounts of amine are required, with no increase in yield observed with stoichiometric amounts of DABCO (Entry 5, Table ). Due to the low extinction coefficient of the EDA complex, 50 mol% of the inexpensive catalyst proved optimal, especially for less reactive substrates. The reaction favoured polar aprotic solvents, with DMF providing yields comparable to DMSO (Entry 6, Table ), while protic or less polar solvents significantly reduced the yield (Supplementary Section 3). |
672e12c27be152b1d071afbe | 7 | The mild conditions of this photocatalyst-free system proved advantageous compared to a reaction photocatalysed by a prototypical iridium polypyridyl complex, which resulted in lower yields (Entry 7, Table ). Using the aryl radical acceptor as the limiting reagent, while having the RAE in excess, also led to a reduced yield (Entry 8, Table ). Irradiation with 405 nm LEDs offered a comparable yield (Entry 9, Table ), in agreement with the broad absorption band of the EDA complex (Fig. ). The radical nature of the reaction was confirmed by the addition of a radical scavenger, TEMPO, which halted reactivity and product formation (Entry 10, Table ). |
672e12c27be152b1d071afbe | 8 | The addition of an auxiliary base to aid deprotonation resulted in yields comparable to those without additional base (Entries 11-12, Table ). However, there was a slight decrease in yield with potassium phthalimide, likely due to its nature as an electron-poor aromatic that competes as a radical acceptor. In both cases, the reaction rate increased significantly, suggesting that deprotonation is the rate-limiting step, warranting further mechanistic investigation (see below). |
672e12c27be152b1d071afbe | 9 | Crucially, the reaction tolerated a wide range of halides (15, 17-19; Fig. ). The tolerance of Cl/Br and other transition-metal-sensitive groups, such as methanesulfonyl and nitrile, demonstrates the robustness of this methodology enabling further downstream functionalisation. Notably, fluorinated and trifluorinated aromatics, common in pharmaceuticals and agrochemicals, often deactivate cross-coupling chemistry due to their electron-deficient nature. Yet, in our anti-Friedel-Crafts mechanism, the electron deficiency of these substrates aided homolytic aromatic substitution (5, 7, 9, 15, 16; Fig. ). |
672e12c27be152b1d071afbe | 10 | Testing extremely poor electron-poor systems, such as nitrobenzene (26), revealed no alkylated products, with the RAE failing to fragment. This was also the case for activated pyridine N-oxide (27). Both of these substrates are precedented to form EDA complexes and the lack of RAE fragmentation is likely due to the donor amine forming an alternative complex with these substrates and outcompeting the RAE. This demonstrates the substrate window, as the electron-poor systems must not form a more favourable EDA complex than the redox auxiliary. Conversely, to prevent alkylation of the phthalimide, substrates must be better radical acceptors than the phthalimide fragment released upon RAE fragmentation. This can be observed in the absence of an alternative aromatic acceptor or when progressively more electron rich (28) substrates are used (Supplementary Section 5.3). To mitigate this competing reaction, an excess of the aromatic coupling substrate (3-5 equiv.) was employed, a strategy common in EDA methodologies using silyl enol ethers, isocyanides, and other radical traps. This requirement for an excess radical-accepting reagent is not prohibitive, as it can be recovered from the reaction mixture, making the method suitable for expensive or late-stage aromatic acceptor substrates (Fig. ). |
672e12c27be152b1d071afbe | 11 | The radical attack giving anti-Friedel-Crafts regioselectivity was highly selective in the case of aryl aromatic acceptors, with loss of yield mainly arising from alkyl radical quenching and alkylation of the phthalimide fragment of the RAE. Meanwhile in the heteroaryl case, we observed some minor regioisomer formation notably at the C4 position on the pyridines. The overall anti-Friedel-Crafts selectivity however was retained, and total yield loss was minimised given the electron deficient nature of these heteroaryl species as improved nucleophilic alkyl radical acceptors (Supplementary Section 5.2). |
672e12c27be152b1d071afbe | 12 | The nucleophilic alkyl radicals generated from the RAE demonstrated exceptional scope, with tertiary, secondary, and primary radicals successfully coupling with 3-fluorocyanopyridine with a wide functional group tolerance (Fig. ). Only the methyl radical proving to be too high energy to fragment and hence unreactive. As expected, tertiary alkyl radicals proved the most successful due to their nucleophilicity and stability, providing a straightforward route to highly hindered quaternary carbon centres, common in many natural and biologically active products. The reaction showed tolerance to various functional groups, including ketones, aryl groups, alkenes, alcohols, and esters. Pharmaceutically relevant motifs, such as cyclic ethers and protected amines, were also retained making this protocol particularly suitable for latestage functionalisation of complex substrates. |
672e12c27be152b1d071afbe | 13 | The reaction is also scalable to gram scale, maintaining similar isolated yields in the alkylation of phthalonitrile from 0.15 mmol (84 %) to 6 mmol scale (82%, 1.233 g). This reaction therefore offers promising potential for industrial applications with a simple protocol; scalability in both batch and flow; and the use of cheap and commercially available catalysts. |
672e12c27be152b1d071afbe | 14 | Building on this versatility, we employed this general C-H anti-Friedel-Crafts strategy to regioselectively functionalise a range of pharmaceutical and agrochemical compounds. The late-stage alkylation with N-Boc 4-methylpyperidine was selected as piperidines are the most common nitrogen heterocycle found in drug molecules. This included: the antiretroviral Neviparine (46); the fungicide Boscalid (47); and the steroid biosynthesis inhibitor Metyrapone (48) in moderate yield (Fig. ). Notably, the excess of electron-poor aromatic radical acceptors was largely recovered, such that all pharmaceutical acceptors were purified in high yield based on recovered starting material (77-88 %), thus making the use of excess reagent non-prohibitive for late-stage or expensive substrates. Furthermore, we employed a redox-active ester of Gemfibrozil, a lipid-regulating fibrate, in the C-H alkylation of 3fluoropicolinonitrile in good yield (49, 66 %). This shows the ability of this methodology to not only alkylate aromatic molecules but furnish carboxylic acid drug molecules with aromatic groups. These results highlight the methodology's practical applicability in the late- To better understand the EDA complexation, mechanistic pathway and selectivity for this methodology, we performed DFT calculations to elucidate the mechanism of this anti-Friedel-Crafts C-H functionalisation. The model system (Entry 1, Table ) was selected based on its optimal reactivity, and the calculations were carried out at ωB97XD/6-31g(d,p) level of theory (see Supplementary Information for details). |
672e12c27be152b1d071afbe | 15 | Upon photoexcitation, an electron from the nitrogen lone pair of the DABCO is promoted to the redox active ester's valent π orbital. The resultant radical cation and anion pair generated can then fragment into a DABCO cationic radical, phthalimide anion, CO2 and the methyl cyclohexyl radical (R•). This fragmentation is slightly endergonic, with an overall Gibbs formation energy of +3.9 kcal mol -1 . While rapid backward electron transfer (BET) |
672e12c27be152b1d071afbe | 16 | immediately after photoexcited SET can serve as a deactivation pathway, the release of CO2 gas renders the fragmentation irreversible, driving the reaction forward. The alkyl radical R•, formed in situ, will then rapidly and selectively attack the electron poor aromatic phthalonitrile at the C4 site to form an aryl radical adduct (I1, ∆G = -2.7 kcal mol - ). |
672e12c27be152b1d071afbe | 17 | The aryl radical adduct (I1) can then proceed through either a hydrogen atom transfer (HAT) or deprotonation to yield the experimentally observed product, 4-(1methylcyclohexyl)phthalonitrile (Fig. ). In the HAT pathway, hydrogen abstraction by the DABCO radical cation leads to the final reaction product (P) in an overall exergonic process by -53.6 kcal mol -1 , followed by proton exchange between the phthalimide anion and DABCO(H), regenerating the DABCO catalyst (-61.4 kcal mol -1 ). However, this pathway is hindered by a high energy barrier (+29.5 kcal mol -1 ), suggesting a slow reaction at 25 °C. |
672e12c27be152b1d071afbe | 18 | In our optimised substrate scope conditions where we utilise 5 mol% Cs2CO3, the carbonate acts as the Brønsted base in the initial stages of the reaction in a highly thermodynamically favoured deprotonation (-45.8 kcal mol -1 , Fig. .2.1). This helps to accelerate the reaction into this propagative chain reaction and relying less on the slow initial EDA fragmentation (Fig. ). |
672e12c27be152b1d071afbe | 19 | Cyano-aryl radical anions, analogous to I2, are readily generated and well precedented via SET using electrochemistry, being used both as a reagent or as an electron shuttle to reduce a reagent in situ -as in our mechanism. Chemical generation of this radical anionic intermediate is also precedented via base assisted homolytic aromatic substitution with deprotonation of an aryl radical via addition of a superstoichiometric base. Although this strategy has been limited to a highly restricted synthetic scope, with some methods also relying on toxic organomercury reagents. The aryl radical anion (I2) must then be oxidised to form the observed product, either (a) by a DABCO radical cation, thereby quenching the radical reaction (Fig. ); or (b) by reduction of another redox active ester, thereby propagating the reaction (Fig. ). This induces another fragmentation event directly, whilst avoiding the slower EDA complex fragmentation pathway. |
672e12c27be152b1d071afbe | 20 | To gain further insight into the efficiency of these redox processes, Marcus theory was applied to calculate the kinetic barriers (Table .4.1). The terminatory SET with the DABCO cation, which regenerates DABCO for further EDA complexation, has an estimated barrier of +9.3 kcal mol -1 . In contrast, the propagative pathway, involving the direct reduction of the RAE, is much more favourable with an estimated barrier of only +0.4 kcal mol -1 , suggesting that this step is diffusion-controlled at room temperature. This implies that the radical anion exists in very low concentrations and undergoes rapid SET, leading to further fragmentation and chain propagation of the cyclohexyl radical R • . |
672e12c27be152b1d071afbe | 21 | The reaction is nevertheless not a perfect runaway chain reaction, as it does require continued irradiation to reach completion. In addition to cyano-aryl radical anion quenching, there is another deactivation pathway, whereby the ester radical anion reduces a DABCO cation to reform the starting EDA complex (Fig. ). This process is thermodynamically favourable (-41.3 kcal mol -1 ) but less likely to occur, as it involves two short-lived radical species that are present in only small concentrations. |
672e12c27be152b1d071afbe | 22 | The complexity of the deprotonation pathways, as well as their similar kinetic barriers, makes it challenging to directly determine their relative contributions to the overall reaction mechanism. To address this, we probed the DFT-supported mechanistic hypothesis involving radical anion propagation through experimental kinetic studies (Fig. ). These experimental results were then compared to microkinetic modelling (MKM) simulations using the DFTcomputed Gibbs energies and activation barriers. |
672e12c27be152b1d071afbe | 23 | We attribute the induction period to the slow and photon-inefficient EDA fragmentation, resulting in a slow generation of product (P). However, if photolytic EDA fragmentation was the rate-determining step through the reaction, one would expect the reaction rate to decline as the concentration of RAE decreases. Instead, an increasing rate was observed in the second phase, indicating a chain-propagative radical mechanism, which is consistent with previous EDA methodologies. The chain propagation in our proposed mechanism, as modelled by DFT, depends on a base to deprotonate the aryl radical adduct (I1 in Fig. ). This step would be rate-limiting in the initial phase of the reaction, as the phthalimide anion has not yet accumulated in solution from RAE fragmentation. |
672e12c27be152b1d071afbe | 24 | To confirm that the observed rate increase was specifically due to the availability of a base, we expanded the study to include a wider range of inorganic and organic bases. This revealed a clear correlation between increasing reaction rate and increasing base strength (pKaH), as shown in Fig. . Cs2CO3 enabled the fastest reaction with completion reached in 3 h (Fig. , black trace), rather than 8 h without base (Fig. , purple trace), all without decreasing the yield. Potassium phthalimide proved slightly slower (blue trace), while the less basic tetrachloro-analogue of potassium phthalimide (green trace) resulted in significantly slower reactions with respect to the addition of Cs2CO3. However, in all cases, the addition of a base accelerated the reaction, whereas the addition of protonated phthalimide retarded the reaction compared to the addition of no auxiliary base (Fig. , yellow trace). Taken together, this base variation supports a deprotonation step being key to reaction propagation; rather than a stabilising chemical interaction with phthalimide as seen in some photochemical methodologies. Microkinetic modelling simulations, based on DFT-computed energy barriers, further supports the mechanistic proposal by replicating the experimental observation of an initial induction phase followed by an acceleration of the reaction rate when no base is added (Fig. ). Inclusion of the phthalimide anion in the model reduces the induction period and leads to a ca. |
672e12c27be152b1d071afbe | 25 | Furthermore, the simulated concentration profiles from MKM confirm that most product formation arises via the chain-propagation pathway, as radical generation from EDA excitation alone is too slow and photon inefficient compared to the propagation pathway. The inclusion of base enables the propagation pathway to begin immediately, thereby reducing the reliance on the slow EDA fragmentation until sufficient concentration of base (phthalimide anion) accumulates. |
672e12c27be152b1d071afbe | 26 | Under the optimised conditions (50 mol% DABCO, 5 mol% Cs2CO3; Supplementary Section 5.4), we determined an internal quantum yield of Φ = 0.136. While a value of Φ > 1 would strongly indicate a chain-propagation mechanism, where more than one equivalent of product is generated per photon absorbed, a value of Φ < 1 does not rule out a chain-propagation process, as it accounts for the entire reaction, including nonproductive pathways such as the rapid BET within the initial EDA complex. Low quantum yields are characteristic of many EDA activation pathways due to these nonproductive processes, particularly those using amine donors to activate electron-deficient RAEs, such as tert-chloro-RAEs (Φ = 0.02). For comparison, we observed an initiation quantum yield of Φ = 0.008. Using established methods, the chain lengths of this autocatalytic reaction can be estimated by dividing the observed quantum yield by the initiation quantum yield, the estimated chain length of this autocatalytic radical chain reaction was calculated to be 17.0 (Φestimated = Φmeasured / Φinitiation; Supplementary Section 5.4) . This supports the presence of a highly efficient autocatalytic chain process, consistent with the experimental and computational findings. |
672e12c27be152b1d071afbe | 27 | The combination of the RAE and the amine donor, DABCO, results in EDA complex formation, producing a distinct charge transfer band absorption band. This absorption band tails into the visible range, allowing absorption at 450 nm and consequent photolysis of the RAE (Fig. ). In contrast, the individual reagents show only negligible absorbance in the visible region (Fig. and Supplementary Section 5.1). The measured spectra are consistent with the TD-DFT simulated spectra, showing the EDA complex's absorbance maxima at 368 nm and absorbance tailing well into the visible region (Supplementary Section 6.3). The 1:1 ratio of components within the EDA complex was confirmed via a Job plot, wherein the ratio of donor and acceptor is plotted against the maximum absorbance (Fig. , Supplementary Section 5.5). Another important mechanistic route considered was the role of the phthalimide anion, formed upon fragmentation of the RAE, as an electron-rich donor that could activate a new EDA complex. Additionally, the 5 mol% Cs2CO3 added to accelerate the reaction, could similarly act as a donor in a new EDA and not just as a general base. Notably, the addition of 5 mol% Cs2CO3 does not increase the absorption of the chromophore at 450 nm (Fig. ). |
672e12c27be152b1d071afbe | 28 | The aryl radical anion (I2) is expected to exist only at very low concentrations due to its predicted rapid reaction with RAE, as supported by DFT calculations. The viability of this step was probed by cyclic voltammetry where phthalonitrile and 4-(1methylcyclohexyl)phthalonitrile ( ) exhibit half wave potential reduction potentials (E1/2) of -2.26 V and -2.38V vs Fc/Fc + respectively (Fig ). Both species are sufficiently negative to reduce the RAE (1) as its reduction potential is -1.62 V vs Fc (Fig. ). As a result, the radical anion electron can feasibly shuttle from the anionic product (I2) to the phthalonitrile starting material, as evidenced by both the cyclic voltammetry and DFT (Fig. and Fig. |
672e12c27be152b1d071afbe | 29 | Typical Friedel-Crafts alkylation proceeds by the attack of a highly electrophilic carbocation towards the most nucleophilic site of an aromatic ring. In contrast, our 'anti-Friedel-Crafts' selectivity relies upon a nucleophilic alkyl radical attacking the most electrophilic site of an aromatic ring. 1,2 Thus, regioselectivity is largely governed by the ability of each carbon site in the aromatic system to accommodate an additional radical electron, with steric effects also playing a role. |
672e12c27be152b1d071afbe | 30 | The experimental 1 H NMR-observed substitution products (Fig. ) confirm the validity of this approach and can be successfully predicted using Fukui indices -a natural bond orbital (NBO)based metric that describes localisation of extra electron density at each carbon centre in the aromatic ring, predicting the stability of the radical aryl intermediate. The higher the Fukui index at a carbon site, the better its ability to accommodate an extra electron, making the site more susceptible to be attacked by the alkyl radical. An exception is Boscalid (47, Fig. ), where significant steric hindrance shields the position alpha to the carbonyl group, limiting the predictive power of the solely electronic Fukui index. |
672e12c27be152b1d071afbe | 31 | To better account for steric effects, we expanded our dataset to include more aromatic substrates with electron-withdrawing groups, bringing the total to 30 molecules and 124 potentially active sites (Fig. ). That way steric factors for each molecule were integrated alongside electronic effects in our regioselectivity predictions. For this expanded dataset, we defined the ground truth alkylation site as the site with the highest Fukui index that is not sterically hindered (e.g. |
672e12c27be152b1d071afbe | 32 | In an effort to circumvent the use of quantum chemical calculations to predict regioselectivity using well-established electronic descriptors without the need for quantum chemical calculations, we applied Hammett parameters to experimental ortho (N-heteroarene-only), meta, and para positions in N-heteroarenes (Supplementary Section 6.7). Summing the Hammett parameters for each functional group at every active site resulted in an accuracy of 53% when compared to the ground truth (Supplementary Section 6.7). We attribute this relatively low accuracy to the possibility that the relationship between Hammett parameters and alkylation regioselectivity is not additive. Consequently, we turned to other methods to better understand and predict this complex relationship. |
672e12c27be152b1d071afbe | 33 | Given the limitations of using Hammet parameters alone, we next used these summed parameters as descriptors to train a machine learning model, specifically eXtreme Gradient Boosting (XGB). We performed leave-one-out cross-validation, where the classifier was trained on all but one molecule and then used to predict the active sites in the remaining molecule. The classifier was trained to identify whether each atomic site was active, and across all sites in each molecule, the one with the highest predicted probability was defined as the most active site. This machine learning approach significantly improved prediction accuracy, yielding a cross-validation accuracy of 93 % compared to the 53% obtained by summing the Hammett parameters (Supplementary Section 6.7). However, the model's dependence on experimental Hammett parameters for different functional groups still poses limitations, particularly for more complex molecules. |
672e12c27be152b1d071afbe | 34 | To address this limitation, we implemented a machine learning model on a purely structurebased descriptor, Smooth Overlap of Atomic Positions (SOAP). One key advantage of SOAP descriptors is that they are fast to generate, requiring only 3D structures derived from SMILES strings, such as those generated using RDKit. This approach also eliminates the need for computationally expensive DFT and NBO calculations to determine Fukui indices, as well as the manual effort of compiling Hammett values for each molecule (see Supplementary Sections 6.1 and 6.8). |
672e12c27be152b1d071afbe | 35 | By utilizing SOAP features, the XGB classifier correctly predicted the most active site in 28 out of 30 molecules, achieving an impressive accuracy of 93% (Fig. ). This high performance is particularly notable given the small dataset of 124 atomic sites and 30 molecules. One incorrectly predicted molecule was 2-(methylsulfonyl)pyrimidine, highlighting a limitation of our dataset -this sulfur-containing molecule was unique, leaving the classifier with no other sulfur-containing examples to learn from. In addition to XGB, other classification models, including random forest, logistic regression, neural networks, and Gaussian process, were tested, achieving comparable or lower performance (Table .8.1). |
672e12c27be152b1d071afbe | 36 | Notably, when comparing the accuracy of these models using SOAP features derived from UFF-optimised geometries with RDkit versus those generated from DFT-relaxed structures (Table .8.1), we found no significant difference in performance. This underscores the robustness and efficiency of our approach, as it does not rely on time-consuming DFT calculations while maintaining high predictive accuracy. Building on the robust performance and efficiency of our model using SOAP features, we next sought to demonstrate the predictive power of the XGB classifier in a real-world setting. For this, we applied the classifier to predict the most active site for four completely new molecules to the model (Supplementary Sections 6.1 and 6.8). These predictions, made without the need of DFT calculations, were completed in a matter of seconds on a standard computer using the XGB classifier trained on the 30 known molecules, with SOAP features derived from SMILES strings. Experimental validation using 1 H NMR confirmed the accuracy of the model, which correctly identified the most active site for all four new substrates (Fig. .8.1). It is important to note, however, that while the model accurately predicts the most active site, it does not provide information on whether other sites in the molecule may also be active. |
672e12c27be152b1d071afbe | 37 | We have developed a simple, direct and scalable method for anti-Friedel-Crafts alkylation of electron poor aromatics by exploiting the photolytic fragmentation of an electron donoracceptor (EDA) complex propagated through radical anion autocatalysis. This autocatalytic mechanism is strongly supported by DFT computation and kinetic evidence, which both demonstrate an increasing reaction rate and an accelerated reaction upon addition of an auxiliary base. Our strategy enables regioselective, photocatalyst-free C-H alkylation using only inexpensive, readily available and non-toxic catalysts. Notably, the catalytic fragmentation of the EDA complex eliminates the need for any exogenous oxidants or reductants. |
672e12c27be152b1d071afbe | 38 | substrates. We envisage further investigation in exploiting radical anion autocatalysis for reaction and the development of alternative propagative mechanisms to exploit electron donor acceptor fragmentation. The advancement of redox auxiliaries also offers exciting possibilities, both for creating more photon-efficient EDA complexes and for incorporating these complexes productively into catalytic processes, rather than using them merely as disposable redox activators. |
64be21ebb053dad33ac8fe86 | 0 | sulfonic acid (TfOH) in hexafluoroisopropanol (HFIP) promotes a Brønsted acid-assisted Brønsted acid catalysis strategy for arylating epoxides and a range of alcohols (Scheme 1a). Our lab recently reported conditions for setting quaternary carbon centers in site-selective Friedel-Crafts reactions using unactivated tertiary alcohols and catalytic combinations of FeX3/HX (Scheme 1b). Aside from a few isolated examples with cycloalkanols, the Cook group found that a mixture of FeCl3/AgSbF6 is effective at arylating unactivated secondary alcohols, albeit with little site-selectivity (Scheme 1c). less hindered ortho-site (Scheme 2a). We further demonstrate derivatization of the alkylated phenolic compound (3ah) with a Conditions: reactions performed on 0.1 mmol scale, phenol 1a (1 equiv), alcohol 2a (5 equiv), 18 h. b Determined by NMR analysis of the crude reaction mixture using 1,3,5-trimethoxybenzene as internal standard. c With 3 equiv 2a. d Isolated yield. |
64be21ebb053dad33ac8fe86 | 1 | We hypothesize that Lewis acid-enhanced Brønsted acidity can render unactivated secondary alcohols reactive in Friedel-Crafts chemistry. 3-tert-Butylphenol (1a) and cyclohexanol (2a) were selected to test for reactivity since they are readily available and conversion to product could be conveniently quantified by NMR analysis (Table ). Catalytic amounts of Fe(III) salts (2.5 mol%) were initially examined with stoichiometric quantities of HCl (2 equiv) at 140 °C. The desired alkylation product (3aa) was formed in 50-57% NMR yields (entries 1-2). Of all the Lewis acid catalysts tested, including Zn(OAc)2 and Fe(II) salts (see SI), ZnCl2 performed the best, providing the product in 63% yield (entry 3). Lowering the temperature from 140 °C to 120 °C was detrimental to yield (39%, entry 4), and increasing the ZnCl2 loading to 5 mol% enhanced product formation (71%, entry 5). However, further increasing the amount of Zn-catalyst to 30 mol% does not improve the reaction outcome (entry 6). CSA is found to be effective at supporting this transformation, albeit providing a lower yield even when cyclohexanol (2a) was employed as the solvent (39%, entry 7). Solid CSA is desirable because it addresses the concern of volatile HCl escaping from the reaction vessels at high temperature. Reducing the amount of alkylating agent in the reaction mixture from solvent quantities to 5 equivalents improved the yield to 68% (entry 8). The reactivity is maintained by reducing the amount of CSA from 2 equivalents to 0.75 equivalents, which resulted in an isolated 74% yield of 3aa when using 3 equivalents of alcohol 2a. |
64be21ebb053dad33ac8fe86 | 2 | The ortho-selective alkylation under Zn/CSA-catalysis conditions was similarly observed between unsubstituted phenol (1n) and 2-adamantanol (2h) where a mixture of alkylated products formed, with o-alkylation occurring as the major product (25%), followed by o-/o-dialkylation (14%), and palkylation (8%) (Scheme 3a). This contrasts from the paraselectivity observed in the analogous Fe/HCl-catalyzed Friedel-Crafts alkylation with tertiary alcohols and HFIPmediated alkylation with tertiary alkyl bromides. We postulate that the zinc catalyst plays a role in templating reactivity and that the phenolic group directs reactivity (through a zinc phenolate species). Upon subjecting anisole (9) to the catalysis conditions, only a small amount of the para-alkylated product (p-10) was observed by NMR analysis (Scheme 3b). This observation emphasizes the importance of the phenolic group in directing both reactivity and selectivity. We found that the catalytic Zn/CSA system favors ortho-selectivity (o-11, >16:1 o/p) even with the more hindered tert-butanol at 80 °C, albeit with lower reactivity compared to the Fe/HCl system (Scheme 3c). At 140 °C, the site-selectivity reverses to favor p-11 (40% We turned to kinetics studies to derive a rate law for this transformation using 3-tert-butylphenol (1a) and 2adamantanol (2h) as the model system (Figure ). Initial rates of the alkylation reaction were measured by varying the concentrations of ZnCl2, CSA, phenolic 1a, alcohol 2h, ZnCl2, and CSA (see SI). These experiments revealed the rate to be largely independent of the concentration of ZnCl2, suggestive of saturation kinetics and sequestration by substrate (Figure ). In the absence of ZnCl2, the initial rate deviates significantly (4.3-fold slower) from the trendline in Figure (See SI, Table ) and is indicative of background reactivity proceeding through a different ZnCl2-free mechanism. The initial rates follow a first-order dependence on the concentration of CSA, half-order dependence on the concentration of phenolic 1a, and first-order dependence on the concentration of 2adamantanol (2h), giving the rate law: rate |
64be21ebb053dad33ac8fe86 | 3 | therefore propose it to be the resting state of the catalytic cycle. For the reaction to proceed, one of the phenolate ligands must dissociate from zinc in exchange for CSA to coordinate, leading to complex B; hence the half-order dependence on [phenol] and first-order dependence on [CSA]. Complexation of the Brønsted acid to zinc effectively enhances its acidity, enabling it to activate alcohol 2 towards Friedel-Crafts alkylation, potentially via transition state D. This SN2-type pathway would be favored in the relatively non-polar PhCl solvent; however, we are unable to rule out the SN1-like pathway where ionization occurs with loss of water. Release of product in the presence of excess phenolic substrate turns over the zinc catalyst. |
64be21ebb053dad33ac8fe86 | 4 | In summary, the combination of zinc and CSA catalysts promotes the first direct ortho-selective Friedel-Crafts alkylation of phenolic derivatives with unactivated secondary alcohols. The free phenolic group is found to be important for reactivity and site-selectivity, which is rationalized through zincmediated templation that biases alkylation at the orthoposition over the generally more accessible para-position. |
674755f77be152b1d019291d | 0 | Dipyrrins (dipyrromethenes 1) are well-known for their ability to act as ligands in coordination chemistry. Dipyrrinato(metal) complexes feature different properties depending on their structural diversity, redox or catalytic properties, and ligand exchange propensity. Depending on the metal employed, they bear one or more ligands and are classified as homoleptic or heteroleptic. They have found applications in materials science, photonics, and photomedicine. In the latter context they have opened chemical space in the medicinal bioinorganic chemistry arena, primarily as anticancer and antimicrobial photosensitizers. The metal complexes have interesting post excitation characteristics through excitonic coupling (Davydov splitting), photoinduced electron transfer, charge transfer states leading in either radical pair (RP-ISC) or spin orbit coupling (SOC-ISC) intersystem crossing. There is a continuous competition between the 1 -* electronic transitions upon photoexcitation or the charge separated states formation which is modulated by the polarity of the environment polarity. |
674755f77be152b1d019291d | 1 | Focusing on the dipyrrinato complexes with Group 13 elements, the boron analogues (BODIPY, 2) are the most famous dipyrrin chelates due to their facile preparation and functionalization. Their advantageous emissive features direct the focus towards applications as fluorescence probes and sensors and in optoelectrical devices and bioimaging. Likewise, gallium and indium complexes can be coordinated with dipyrrin ligands. Frequently, these dipyrrin complexes, in comparison with the boron alternatives, are considered as poor emissive chelates; however, this is not a hard-and-fast rule. Recently, Wan et al. reported the synthesis of the stable monomeric gallium chelate 3 (the analogue of BODIPY), which resembled the highly fluorescence profile and the photophysical properties of BODIPY dyes along with a similar tetrahedral geometry. The tris(dipyrrinato)gallium(III) complex 4 showed poor fluorescence emission (Φf = 0.024 in hexane), whilst the monomeric complex 3 had highly luminescent properties (Φf = 0.82 in DCM, 0.91 in toluene). Factors that affect the absorption and emission properties include the metal center but also the number and nature of the coordinated ligands. Sazanovich et al. showed that by altering the 5-substitution and by increasing the size of the aryl group in zinc bis(dipyrrins), the non-radiative decay could be diminished, with a considerable increase in the fluorescence quantum yield. Similarly, the homoleptic tris(dipyrrinato)indium(III) complexes 5a and 5d were emitting less (Φf = 0.074 in hexane and Φf = 0.028 in toluene, respectively) than the heteroleptic analogues 5b,c, which exhibit stronger fluorescence features (Φf = 0.41 and Φf = 0.34 in toluene, respectively). Lastly, Gutsche et al. evaluated the phototoxicity of indium(III), iron(III), and gallium(III) tris(dipyrrinato) complexes bearing pentafluorophenyl moieties with alcohols and thiocarbohydrates. They tested the phototoxicity against various tumor cell lines and against the bacterium S. aureus. Promising phototoxic results were exhibited by the glycosylated analogues of tris(dipyrrinato)gallium(III) complexes. Only few reports have been published to date on aluminum dipyrrin complexes. Aluminum finds a wide range of uses in material, pharmaceutical sciences, and food industry; whereas organometallic complexes containing aluminum are used as catalysts. Notably, Al is the most abundant metallic element and the third most abundant of all elements in the earth's crust making it a crucial target in developing more sustainable metal containing materials and drugs. An example of aluminum drugs in photodynamic therapy (PDT) is the sulfonated aluminum phthalocyanine Photosens, which has been clinically approved for the treatment of some cancers. Giannopoulos et al. described the synthesis of monomeric dipyrrinato aluminum (ALDIPY) complexes bearing mesityl substituents at the α-pyrrolic positions (6,7) whilst the photophysical properties were beyond the aim of the work. The introduction of bulky aryl groups at the α-position of the pyrrole ring resulted in a steric hindrance which stabilizing the reactive Al(III) center towards the formation of the mono dipyrrin aluminum complexes. Additionally, Ikeda et al. reported N2O2type aluminum dipyrrins (8,9) which displayed an absorption maximum at ca. 600 nm and exhibited moderate to high fluorescence quantum yields. Lack of a α-pyrrolic substituents allows for the formation of 1:3 metal-ligand complexes of trivalent metals. 20 years ago a patent reported the structures of tris(dipyrrinato)aluminum(III) complexes, where aluminum adopted an octahedral coordination geometry. Apart from this patent, which was directed towards optoelectronic applications, we are unaware of other chemical or photophysical characterizations of such complexes. In this study, we sought to develop a library of novel homoleptic tris(dipyrrinato)aluminum(III) [AL(DIPY)3] complexes with the aim of investigating their structural and photophysical properties and possible use in photodynamic therapy. |
674755f77be152b1d019291d | 2 | Reaction yields for the DIPY were relatively good and in the range of 30-90%. N-H protons of the dipyrrin moiety show a broad weak resonance (sometimes not easy to detect) in the range of 11-13 ppm in their 1 H NMR spectra, as their arrangement results in a planar conformation with a fast tautomeric exchange of the N-H proton between the nitrogen atoms of the pyrrolic units. The aromatic properties of these compounds are demonstrated by the 1 H NMR shifts where β-protons appear in the range of 6.3-6.7 ppm and α-protons appear in the range of 7.5-7.7 ppm. Aluminum salt AlCl3 in mild basic conditions was utilized for the N-H deprotonation and complex formation. The absence of α-or βpyrrole substituents resulted in a less sterically hindered environment which, along with the presence of the aluminum trivalent metal ion formed the tris(dipyrrinato) complexes. Attempts for in situ complexation, of the respective dipyrrins without purification, as it is common in the case of BODIPY complexes, were unsuccessful similar to reports for other trivalent Group 13 metal ion complexes. In our case, TLC reaction monitoring showed predominately the dipyrrin spot and by-product formation with a less intense orange spot of the desired AL(DIPY)3 compound. Therefore, dipyrrin purification prior to complexation is essential in order to access the corresponding AL(DIPY)3 complexes in good yields. The majority of the aluminum complexes were obtained in moderate to high yields of 50-90% with the lowest yield being found for the mesityl derivative (13b, 18%). |
674755f77be152b1d019291d | 3 | Irrespective of increasing the number of AlCl3 equivalents unreacted dipyrrin remained after complexation but could be recovered via column chromatography. The majority of these complexes adopt a symmetrical octahedral geometry with clear signals in the proton and carbon NMR spectra testifying the expected symmetry. As mentioned, α-pyrrole protons of the dipyrrin precursors appear at a lower field ~7.5-7.6 ppm, whilst in the octahedral complex configuration they appear at higher field ~6.7-7.0 ppm. This stems from the increasing shielding effect (double) by the two adjacent dipyrrinato moieties. Moreover, mass spectrometry was employed and their exact mass was confirmed by APCI method. Additionally, the octahedral aluminum coordination center was confirmed via Al NMR with a signal/resonance in the range of 6.5-7.15 ppm which is characteristic of the octahedral coordination (between -46 and 40 ppm). The insufficient solubility of AL(DIPY)3 13g resulted in practical issues, thereby it was difficult to proceed with the complete photophysical characterization, but it was characterized by UV-Vis, NMR and MS. |
674755f77be152b1d019291d | 4 | With the aim to investigate the capability for post functionalization on the periphery of AL(DIPY)3 complexes and introduction of polar groups or palladium-catalyzed coupling reactions were tested (Scheme 2). First, ester hydrolysis of the 4methoxycarbonylphenyl derivative 13k was conducted straightforward yielding the corresponding carboxylic acid 16 in 95% yield. The presence of carboxylic acid groups can facilitate the biological applicability since they increase the solubility in more polar solvents. Moreover, the introduction of carboxylic acid group can act as an anchor side for further modification, such as bioconjugation. |
674755f77be152b1d019291d | 5 | Next, the pentafluorophenyl substituents of 13d enabled the nucleophilic aromatic substitution on the respective para-fluorine position under basic conditions. First attempt was to introduce the propargyl group as described by the reaction yielded the 4-hydroxy-2,3,5,6-tetrafluorophenyl derivative 17 in good yield (85%). Purification of the complex was tedious and different solvent systems with increasing polarity were required. Similarly, deprotection of 13e to compound 18 proceeded smoothly with tert-butyl ammonium fluoride (TBAF) in 92% yield. Finally, a palladium-catalyzed cross-coupling reaction was performed by using a BODIPY moiety proving that AL(DIPY)3 complexes can be stable and the formation of supramolecular complexes is feasible. Specifically, the Suzuki-Miyaura reaction of 13c and four eq. of the borylated BODIPY 19 yielded the desired product 20 in very good yield (72%). |
674755f77be152b1d019291d | 6 | Single crystal X-ray diffraction analysis was performed to determine the molecular geometry of four dipyrrins (Fig. ) and the majority of homoleptic tris(dipyrrinato)aluminum(III) complexes. Crystal structures of such aluminum complexes have not been reported yet. The dipyrrin core mainly adopted a planar conformation with the contribution of the intramolecular hydrogen bonds (N•••H) and an average distance ~2.06 Å. Compound 12h crystallized with two independent molecules in the asymmetric unit. One molecule presented an intramolecular hydrogen bond keeping it in a planar conformation whereas the other had a more twisted conformation and no intramolecular hydrogen bond (Fig. ). The dihedral angle between the dipyrrin plane and the 5substituent is locked between 55 -90 degrees in the order: 12f with 52.57(3)° < 12a with 67.52(4)° < 12g with 83.45(2)° < 12h with 93.20( )° (Fig. ). The trimeric octahedral configuration of tris(dipyrrinato)aluminum(III) complexes was confirmed by singlecrystal X-ray crystallography. The crystal structure obtained of the phenyl tris(dipyrrinato) derivative 13a (Fig. ) is isostructural to the iron(III) analogue. The dihedral angle between the dipyrrin and the substituent for the two moieties is between 65-76° and the structure is tightly packed without any solvent (Fig. ). Compound 13b (Fig. ) exhibited disorder of the mesityl group and showed a crystal lattice tightly packed with encapsulated solvent and a dihedral angle of 83.44° between the dipyrrin plane and the mesityl group (Fig. ). Indium and gallium analogues have been reported by Thoi et al. with equal asymmetric units, similar metal-nitrogen distances and degrees of dihedral angles of mesityl group and dipyrrin moieties. Complex 13c (Fig. ) crystallized with one AL(DIPY)3 and a partially occupied hexane in the asymmetric unit. The solvate molecules are located in a channel parallel to the a-axis (Fig. ). The bromophenyl vs. dipyrrin twist angle was 120.2°. A similar structure of 13d (Fig. ) contained half of the molecule in the asymmetric unit and a partially occupied hexane. The dihedral angle between the dipyrrins and C6F5 groups was 74-84°. The solvate molecules are incorporated in channels running parallel to the c-axis (Fig. ). Analogous structures of gallium and indium were previously reported by us. Figure . View of the molecular structure in the crystal of 13a, 13b, 13c,13d (disordered hexanes omitted), and 13f. Thermal ellipsoids show atomic displacement at 50% probability. |
674755f77be152b1d019291d | 7 | Complex 13f (Fig. ) crystallized with one molecule in the asymmetric unit and exhibited an intermolecular interaction between oxygen and hydrogen (C-H•••O) of 2.61 Å. The crystal structure shows tight packing in 'stacks' parallel to the a-axis with weak H-bonding and no solvent (Fig. ), and dihedral angles between the dipyrrin plane and the substituents of 32 -66°. Both 13h (Fig. ) and 13i (Fig. ) showed significant rotational disorder because of the formation of atropisomers (rotation about the dipyrrin-aryl bond. The structure of 13h contained a partially occupied disordered hexane and showed dihedral angles between the dipyrrin plane and the substituents of 64-99°. The structure of 13i showed tightly packed molecules without solvate molecules. The naphthyl group was twisted by 74-98° from the dipyrrin plane. Compound 13j (Fig. ) crystallized with one molecule of the complex and three solvent sites in the asymmetric unit (Fig. ). The dihedral angle between the dipyrrin plane and the anthracene unit is between 88-92°. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.