Heavy atom effects in the Paternò–Büchi reaction of pyrimidine derivatives with 4,4’-disubstituted benzophenones

The regioselectivity and the photochemical efficiency were investigated in the Paternò–Büchi reaction of 1,3-dimethylthymine (DMT) and 1,3-dimethyluracil (DMU) with benzophenone (1b) and some 4,4’-disubstituted derivatives (dimethoxy (1a), difluoro (1c), dichloro (1d), dibromo (1e) and dicyano benzophenone (1f)) that gives rise to two regioisomeric oxetanes, 2 and 3. The regioselectivity (the ratio of 2/3) decreased gradually for both DMT/DMU photochemical systems from 1a to 1f. That is, a halogen atom as an electron-withdrawing group (EWG) has a pronounced effect on the regioselectivity. However, the photochemical efficiency of the 1e systems did not show the expected increase, but decreased relative to systems with 1b. Temperature effects on the regioselectivity of 1b–e systems showed some interesting features for systems with heavy atoms (including the 1d and 1e systems), such as higher inversion temperatures, and an entropy-controlled regioselectively whereas the regioselectivity for two other systems (1b and 1c) is enthalpy–entropy controlled. A heavy atom effect is suggested to be responsible for these unusual phenomena based on the triplet-diradical mechanism of the Paternò–Büchi reaction.

The "heavy atom effect" is a term which has been used to describe the influence of "heavy atom" substitution on a spinforbidden transition such as various intersystem crossings. If heavy atoms are present in a Paternò-Büchi reaction, spin-transition processes would be affected, and this may lead to interesting results.
Abe et al. [14] investigated the effect of a sulfur atom on the stereoselective formation of oxetanes in Paternò-Büchi reaction of aromatic aldehydes with silyl O,S-ketene acetals to give trans-3-siloxyoxetanes and found that this was ca. 70/30 to 90/10. The trans-selectivity is explained by the sulfur atom effect in the silyl O,S-ketene acetal which controls the approach direction of the electrophilic oxygen of triplet n,π* aldehyde to the nucleophilic alkene. A fast ISC process of the triplet 1-alkylthio-1-siloxy-2-oxatetramethylene 1,4-diradical in competition with the bond rotation was proposed [14]. Griesbeck et al. [15] observed substantial 2 H-magnetic isotope effects on the diastereoselectivity of triplet photocycloaddition reactions. Weaker isotope effects on the diastereoselectivity were detected for the reaction of α-deuterated propionaldehyde [15].
In this work, we have investigated the Paternò-Büchi reaction of 1,3-dimethylthymine (DMT) and 1,3-dimethyluracil (DMU) with benzophenone (1b) and its 4,4'-disubstituted derivatives 1a and 1c-1f with the formation of the regioisomeric oxetanes 2 and 3 (Scheme 1). By changing the halogen at para positions in the benzophenones, the photochemical efficiency and the regioselectivity were significantly affected, and the effects cannot be considered as a pure electronic effect (of the electron-withdrawing groups, EWGs), by comparing the observations with those of systems of 1a (with electron-donating groups, EDGs), and 1b and 1f (also with EWGs). However, as a heavy atom effect, observations above can be rationalized based on the triplet mechanism of the Paternò-Büchi reaction.

Substituent effects
To investigate substituent effects of benzophenones in the Paternò-Büchi reaction, photochemical reactions of DMT/ DMU with 1a-f in acetonitrile-d 3 were performed in Pyrex NMR tubes. The regioselectivity (the ratio of 2/3) and the yield were measured directly from the 1 H NMR spectra of crude product mixtures and are listed in Table 1. The substituent effect of benzophenones on the regioselectivity (2/3) is similar to our previous observations [10], a gradual decrease according to their electronic effect from 1a to 1e.  In our previous papers [10,13], the photochemical [2 + 2] cycloadditions of DMT and DMU with benzophenones generate two series of regioisomeric oxetanes, 2 and 3, via 1,4-diradical intermediates, and reveal notable substituent effects on the regioselectivity and the photochemical efficiency. The reactions initiated by triplet benzophenones with EDGs give a higher proportion of 2 and a lower photochemical efficiency, whilst benzophenones with EWGs yield a lower proportion of 2 and have a higher efficiency.
The data in Table 1 show that the regioselectivity (2/3) and the photochemical efficiency correlates clearly with electronic property of substituents. The benzophenones with EWGs give more efficient Paternò-Büchi reactions (except 1e systems) and lower ratios of 2/3, and the benzophenones with EDGs undergo less efficient Paternò-Büchi reactions and have higher ratios of 2/3, in accord with our previous observations [10,13]. However, the photochemical efficiency of the 1e systems decreased significantly. According to our understanding of these oxetanes [10], this low efficiency was considered to be due to poor stability of the photoproducts, in particular oxetanes 3.
To verify this speculation, the yield and the regioselectivity were tracked over an irradiation time of 15 min for the DMT-1e (5 mM/5 mM) system ( Figure 1). Figure 1 clearly shows that the yield increases with irradiation time and the ratio of 2/3 is slightly higher (37:63) during the initial irradiation period (1-3 min), and then becomes constant (33:67) on further irradiation (> 3 min). However, this change is within the experimental error of 5%. According to our understanding of the stability of oxetanes, 3 are less stable than 2. The constant ratio of 2/3 implies that no significant decomposition of the photo-product oxetanes occurs in the photochemical reaction. In other words, the stability of photoproducts in the systems is not responsible for the low yields. Therefore, the effect of halosubstituted benzophenones on the Paternò-Büchi reaction is not a "pure" substituent effect.

Temperature effects
In our previous papers [9,11], the photochemical [2 + 2] cycloadditions of DMT/DMU with benzophenones revealed notable temperature effects on the regioselectivity and the photochemical efficiency. We have demonstrated that the temperature-dependent regioselectivity is derived from the conformational properties of the intermediate triplet 1,4-diradicals.
The observations show that the reaction temperature influences the regioselectivity by changing the populations of two regioisomeric diradicals as a result of differences in the potential energies of two stable conformers, the productive conformation of the triplet diradical and the unproductive conformation of the triplet diradical, for each regioisomeric diradical [9,11].
To investigate further the temperature effects in four systems with 4,4'-dihalo-substituted benzophenones, we carried out the Paternò-Büchi reactions of DMT with 1b-1e over a temperature range of −30 to 70 °C: Notable temperature effects were observed. Both the photochemical efficiency and the regioselectivity (2/3) decreased with increasing temperature from the general trend (Table 2). Table 2 shows clearly that efficiencies of the 1b system are lower than those of the 1c and 1d systems except for the values at the initial three temperatures, i.e., the efficiency of 1b system is the most sensitive to reaction temperature, and the efficiency of 1e system is the lowest among the four systems studied. In addition, the data show that a higher reaction temperature is unfavorable for Paternò-Büchi reactions.
The regioselectivity (2/3) data in Table 2 have been plotted in Figure 2 against the inverse absolute temperature according to the Eyring formalism [16]. (1) Where k 2 and k 3 are the overall rate constants of the reactions leading to the two regioisomers 2 and 3, respectively.
The Eyring plots obtained are nonlinear across the whole temperature range. The nonlinear Eyring plot is indicative of a change of the selectivity-determining step during the change in the reaction temperature [1,[17][18][19].  Although the Eyring plots obtained were nonlinear over whole temperature range, strict linearities (correlation coefficients R > 0.99) on both sides of inversion points were found. The temperature at the point of inversion is called the inversion temperature, T inv , of the system. The temperatures increase gradually from H-to Br-substituted Paternò-Büchi systems ( Table 2). In our previous paper [11], the temperature effect of the Paternò-Büchi reaction DMU with three benzophenones, 1b, 1c and 1f, was investigated, and similar inversion temperatures could be obtained from the Eyring plots, 295 K for 1b, According to the Eyring theory, when this relationship is plotted, the slope corresponds to the difference in the overall activation enthalpies (ΔΔH ≠ ) and the intercept represents the difference in the overall activation entropies (ΔΔS ≠ ) (Figure 2). The inversion temperature reveals two sets of activation parameters (ΔΔH 1 ≠ and ΔΔS 1 ≠ (T > T inv ), ΔΔH 2 ≠ and ΔΔS 2 ≠ (T < T inv )), which were obtained from the slope and the intercept of the linear plot for each system.  [7]. Therefore, the regioselectivity in the Paternò-Büchi reaction is strongly temperature-dependent. Moreover, these activation parameters (ΔΔH ≠ and ΔΔS ≠ ) increase gradually from the F-to Br-benzophenones systems, with the exception of ΔΔH 1 ≠ for 1d. In addition, the values of ΔΔH ≠ are similar to T inv ΔΔS ≠ for the 1b and 1c systems since the ratio of 2/3 is ~50:50. However, the values of ΔΔH ≠ are less than T inv ΔΔS ≠ for two other systems (−2.1 J/mol for 1d, −3.2 kJ/mol for 1e). In other words, this is an entropy-determined selection for the regioselectivity over the whole temperature range investigated.

Interception of heavy atom effects
Based on the triplet mechanism of the Paternò-Büchi reaction, it is possible to have a more detailed discussion on a heavy atom effect on the Paternò-Büchi reaction based on the phenomena noted above. The formation of two regioisomers in the Paternò-Büchi reaction is detailed in Scheme 2.
Among these processes, there are four factors that determine the regioselectivity [7]. (i) Initial O-attacked site selection at the double bond by 3 benzophenones*, k 2a + k 2a' versus k 3a + k 3a' , (ii) equilibrium constants, : This factor only operates when the conformational change is faster than the ISC process, (iii) the relative rate constants of the ISC processes in the triplet 1,4diradicals, k 2I /k 2I' versus k 3I /k 3I' and (iv) the relative rate constants of the bond-forming and bond-breaking step from the singlet 1,4-diradical C5-1 DR1 and C6-1 DR1, k 2b /k 2c versus k 3b / k 3c .
According to the Curtin-Hammett principle [20], the ratios of the productive conformers of singlet diradicals C5-1 DR1 and C6-1 DR1 are determined not only by the populations of C5-3 DR1 and C6-3 DR1 but also by the relative rate constants of ISC processes, the k 2I /k 2I' and k 3I /k 3I' . The former is determined by the equilibrium constants, K 2 and K 3 , whilst the latter processes (k 2I , k 2I' , k 3I and k 3I' ) would be accelerated by heavy atoms. Thus, the equilibrium between the productive conformers and the unproductive conformers, of the triplet 1,4diradicals, would be achieved at a higher temperature for the system with heavy atoms than that without heavy atoms. Due to the energy barriers between the two stable conformers [9], the equilibrium is more favorable for the formation of oxetanes 3 rather than oxetanes 2 at a higher temperature. This would lead to a higher inversion temperature and a higher ratio of 2/3. In addition, the ISC process from singlet excited state to triplet excited state is very fast, 10 11 s −1 for benzophenones, and not affected by heavy atoms, but the ISC process of triplet benzophenones to singlet ground state would be accelerated, reducing lifetime of triplet benzophenones. Finally, triplet benzophenones with a short lifetime would give rise to a less efficient Paternò-Büchi reaction. The Paternò-Büchi reaction of DMT/DMU with benzophenones generates two regioisomers 2 and 3. The oxetanes were numbered by using subscript "1" for DMT and "2" for DMU, e.g., oxetanes from DMT-1a system are denoted as 2a 1 and 3a 1 . For oxetanes mentioned in this work, most were reported in our previous papers [9,10] except for the following oxetanes. 3a 2 has not been detected by 1 H NMR. 3d 1 , 3d 2 , 3e 1 , 2e 2 and 3e 2 could be detected by 1 H NMR, but could not be isolated because of their poor stability towards acid and silica gel. 2e 1 was isolated and the characterization data of 2e 1 was as follows:

Photoproduct assay
The Paternò-Büchi reactions of DMT/DMU with benzophenones were performed in acetonitrile-d 3 . A solution of the reactants was placed in a Pyrex NMR tube (transmitted light > 290 nm), purged with high purity N 2 for 10 min and then irradiated with a 125 W high-pressure Hg lamp at 10 °C. The sample tubes were placed on a merry-go-round equipment moving around the Hg lamp. Photoproducts 2 and 3 have no significant absorption for light at above 290 nm. Hence, a secondary photolysis of the oxetane products (2 or 3) should not occur unless there is prolonged irradiation. Compositions in photoreaction mixture were quantified by 1 H NMR spectroscopy (300 MHz) directly on the crude product mixture, using the sum of 5-methyl (5-H) and 6-H signals as internal standards. The yields and the ratios of the two regioisomeric oxetanes were obtained from the peak areas of 5-methyl and 6-H for DMT system and those of 5-H and 6-H for DMU system in the 1 H NMR spectra. The experimental error was within 5%.