Synthesis and characterisation of novel composite sunscreens containing both avobenzone and octocrylene motifs

Avobenzone and octocrylene are popular sunscreen active ingredients. Experiments that probe the stability of avobenzone in binary mixtures with octocrylene are presented, together with the synthesis of a class of novel composite sunscreens that were designed by covalently linking avobenzone and octocrylene groups. Spectroscopy, both steady-state and time-resolved, of the fused molecules was performed to investigate the stability of the new molecules and their potential function as ultraviolet filters. Computational results are detailed for truncated versions of a subset of the molecules to reveal the energy states underlying the absorption processes of this new class of sunscreen. The results indicate that the combination of elements of the two sunscreen molecules into one molecule creates a derivative with good stability to UV light in ethanol and in which the main degradation pathway of the avobenzone component in acetonitrile is reduced. Derivatives containing p-chloro substituents are particularly stable to UV light.


1) Background information on Avobenzone and Octocrylene.
Avobenzone: The photochemistry of dibenzoylmethane UVA filters was studied by Schwack and co-workers. 1 While UVA photodegradation was low in polar solvents such as isopropanol and methanol, avobenzone 1 was sensitive to UVA light in the non-polar solvents cyclohexane and isooctane. This significant to sunscreens as high screening efficiency cannot be guaranteed if photochemical pathways can lead to loss of UV absorbing activity. Several photoproducts were identified by HPLC and GC-MS during an 8-hour solar-simulated irradiation in cyclohexane ( Figure S1). Figure S1. Photodegradation pathways of dibenzoylmethane. 1 It was suggested that keto/enol tautomerisation in avobenzone 1 precedes fragmentation in non-polar solvents. Evidence for this can be found in complementary NMR studies that demonstrated that while the tautomerisation equilibrium lies towards the enol form, around 3.5% was detected in the diketo form in cyclohexane-d12 (degradation is more rapid in cyclohexane) while none was observed in deuterated polar solvents. A study by Roscher and co-workers identified only p-tert-butylbenzoic acid and p-methoxybenzoic acid as products from an extended irradiation, and the study suggests that the benzoic acids were S3 the most stable photoproducts observed. 2 The carbon-centred benzoyl and phenacyl free radicals proposed by Schwack as the key intermediates in the fragmentation pathway, were observed experimentally as electron-spin resonance (ESR) spectroscopy signals that persist for several minutes after cessation of irradiation. 3 In an ethanolic solution of avobenzone 1, a UV absorption band around 356 nm with ππ* character is assigned to the convolution of two degenerate cis enols that are stabilised by an intramolecular hydrogen bond between the 1,3-dicarbonyl From Figure S2, it may be concluded that stabilisation of the ground-state chelated enol form to avoid formation of the excited triplet state keto will also stabilise the molecule against photodegradation. This could be by means of formulation in emollients of appropriate stabilising polarity, encapsulation, by including antioxidant additives such as glutathione or by modifying the structure of the molecule itself to shift the equilibrium towards the enol form; supported by a number of studies. [8][9][10][11][12] The diketo form of avobenzone that can be excited to form a triplet and/or the carbon-based radicals discussed above are potentially reactive towards biological substrates and the photosensitizing ability of avobenzone 1 has been illustrated in a number of studies where it is observed alongside increased cytotoxicity in human keratinocytes, 13 lipid peroxidation 14 and direct strand breaks in plasmid DNA. 15 Avobenzone is thought to have synergistic effects with other sunscreen additives and could potentially react with other components in a formulation. Photogenerated fragments from avobenzone led to DNA-damaging photoproducts when irradiated in the presence of cinnamate filters such as octinoxate or EHMC, with concurrent loss of UV activity. 16 The identified mechanism followed a [2+2] cycloaddition of the enolic form of the diketone to the electrophilic alkene, followed by ring-opening ( Figure S3). 17 S5 Figure S3. Reactive pathway identified for avobenzone fragmentation.
Octocrylene: Baker and co-workers identified a minor relaxation pathway that suggested the presence of a long-lived triplet state in the ultrafast transient electronic absorption spectra of octocrylene in both methanol and cyclohexane. 18 It is thought that this triplet state, which can be prepared through collision with avobenzone in the excited state, can thus improve avobenzone stability. 19 Figure S4). In a study by Kikuchi and co-workers, the excited states of avobenzone and a bis-alkylated analogues were reported. The single state energy (ES) was obtained from the intersection point of the UV absorption and fluorescence spectra and the triplet state energy (ET) was obtained from the first peak of phosphorescence in ethanol at 77 K. 25 Figure S4. Energies of the singlet and triplet states (of the two main forms) as summarised in the text. Energies* from Kikuchi et al. 25 were converted to kJ/mol from cm -1 by multiplying by hcNa. ND -Not determined.
The key process is the interaction of the singlet and triplet states. The S1 state of the keto form of avobenzone possesses mainly a 1 nπ* character, whereas that of the enol form possesses a 1 ππ* character. The intersystem crossing (ISC) between S7 [the singlet] 1 nπ* and [the triplet] 3 ππ* states should be much faster than that between 1 ππ* and 3 ππ* states, as suggested by El-Sayed's selection rule. 26 A further prerequisite for a transfer process to occur is a limit of intermolecular distance between a donor and acceptor molecule. For triplet-triplet interaction this distance is about 1 nm, compared with the longer distance of ~10 nm for singletsinglet processes. 27 To assess whether the mean free path (p) of the molecules in a given solution is on the order where such processes can occur, equation (1): can be used, where n is the number of molecules per volume (mols/L) and d is the molecular diameter (nm). 28 If an approximate diameter of a typical sunscreen molecule is 1 nm then a lower limit on the concentration for a triplet-triplet process could be approximated by equation (2): It is noted here that this concentration is much higher than the concentration used in our steady-state irradiation studies. In contrast, the concentrations used in real formulations exceed this lower limit by many orders of magnitude.  (3) and (4) below: (3) As a starting point for any discussion for the difference in stabilities between the variously substituted AVOCTO compounds in the following sections, reference can be made to the effect of para-substitution with electron donating groups (EDGs) on aromatic carbonyls ( Figure S5) as studied by phosphorescence, optically detected magnetic resonance, and other optical techniques by Kikuchi et al. 25 Figure S5. Molecules studied by Kikuchi et al. 25 There is near degeneracy between the T1 and T2 states in many studied aromatic carbonyls which may be explained by the large spin-orbit coupling constant of the carbonyl oxygen and its important role in the mixing between the T1 3 ππ*and T2 3 nπ* states and singlet states. The S1 state of the keto form of BM-DBM possesses mainly 1 nπ* character, whereas that of the enol form possesses mainly 1 ππ* character. The intersystem crossing (ISC) between [keto] 1 nπ* and 3 ππ* states should be much faster than that between [enol] 1 ππ* and 3 ππ* states, as suggested by El-Sayed's rule. The energy of the T1 state (mostly 3 ππ*) is reduced by the effect of para-substitution with an electron donating group while the energy of the T2 state (mostly 3 nπ*) is increased; as observed experimentally when benzaldehyde is compared to p-methylbenzaldehyde and p-methoxybenzaldehyde. 30 Kikuchi and co-workers 25 observed an increase in the T1 lifetime with avobenzone-like molecules substituted with tert-butyl and methoxy groups (i.e. BM-DBM vs. DBM) S9 and offered an explanation for the effect of para-substitution in lengthening the triplet state lifetimes: the purity of the T1 state is recovered by an electron-donating substituent because the 3 ππ* excitation energy decreases. A longer triplet lifetime means that the excited state molecule remains in a high energetic state for longer and thus has a greater probability of reacting further, a property that is unfavourable to its use as a sunscreen. In this work, the effect of an electronwithdrawing group (EWG, i.e. -Cl) is studied along with EDGs (-OMe,-tBu). S10 2) Materials and Methods (Synthesis) and NMR spectra.

Open-access LRMS and HRMS
Lower resolution mass spectra were recorded on a time of flight (TOF) mass spectrometer (Agilent 6130B single Quad) by the electrospray ionisation (ESI) method with a potential mass range of 50 -3,000 m/z. This is coupled with a coupled with an isocratic Agilent 1100 HPLC (without column) as an automatic Synthesis of precursors to avobenzone/ octocrylene (AVOCTO) compounds.
This compound was reported previously. 31

Preparation of ethyl 4-chlorobenzoate.
This is a known compound. 35
A procedure for an analogous compound was adopted. 38             adjusted to equate to 1,000 W/m 2 (100 mW/cm 2 ) sustained over 2 hours which is equivalent to the Sun's energy at Earth's surface (see Figure S6). Spectra were recorded on a UV-vis spectrometer (Cary 60 UV-vis, Agilent Technologies) at specified intervals up to 2 hours at a scan rate of 600 nm/min at 1 nm intervals with baseline correction.

Quantifying photostability
A key measurement of photostability is the percentage degradation of a molecule under the action of UV or solar light, and this can be quantified by equation (5) An AUC R > 0.80 has previously been used as a criterion for a photostable molecule. 39 The AUC R ratio has been converted into a percentage for the Tables in the main paper.

Transient Electronic Absorption Spectroscopy (TEAS).
The transient electronic absorption setup at the Warwick Centre for Ultrafast Spectroscopy has been described previously 40

Assessing stabilisation effects in binary mixtures in ethanol
Binary mixtures of avobenzone and octocrylene (Merck) were prepared in ethanol and acetonitrile to observe any synergistic effects between the compounds.
Octocrylene is often found in formulations with avobenzone. The absorbance change in the system was measured both at the absorbance peak and across the entire UVA range (area under the curve, AUC). In the binary mixtures in ethanol there was a reduced loss of activity at the UVA peak absorbance when avobenzone was combined with octocrylene (-3%). This is also the case across the entire UVA range. A stoichiometric ratio of 1:3 (avobenzone: octocrylene) was used as this is typical of a sunscreen product and results in an approximately equal absorbance at both peak maxima. From these data alone, it is difficult to distinguish the effect of spectral overlap and photon shielding from that of a true stabilising effect, especially when the overall loss of activity is low. When modelled with linear rates of photodegradation (y = a + b*x) the rate constants (b) are within error of each other and therefore these data do not appear to show any significant stabilisation at these concentrations in ethanol. Experimental data are recorded in Table 2 (main paper) and displayed in Figure S8. Octocrylene was irradiated alone separately and showed no degradation.
Assessing stabilisation effects in binary mixtures in acetonitrile.
When assessed for stability under the same conditions used in the ethanol experiments, the results for acetonitrile (data in Table 4, main paper) show that the avobenzone UVA activity loss is significantly greater in the polar, non-protic solvent than in polar protic, with an increase to ~70% loss of activity during an hour of irradiation. In acetonitrile, the data show significant difference between avobenzone in the two-component mixture versus avobenzone alone. As a control, one cuvette containing only avobenzone was not irradiated and the loss of activity was considerably less (~7% vs. ~70%). This indicates that the major factor in driving the loss of the UVA activity is a light-driven process and there may be a minor effect of tautomerisation that occurs without irradiation and is due to the solvent environment alone. In any case in the two-component mixture the decrease in the UVA absorbance activity is reduced in line with the average degradation at peak absorbance but whether there is a direct stabilisation mechanism, or this is simply a S52 result of spectral overlap is unclear. Details of each experimental result are presented in Figure S9. . Plots of the UV-vis activity of AVOCTO compounds 10a-10e in acetonitrile (main paper, Table 3) are shown in Figure S10:  10a-10e, avobenzone and chloroavobenzone in acetonitrile (main paper, Table 3).  Figure S12. TAS for avobenzone presented as a false colour heat map in acetonitrile photoexcited at the pump wavelength 355 nm. S58 Figure S13. Transient at 363 nm for AVOCTO1 10a (left), AVOCTO4 10d (middle) and avobenzone (right) in acetonitrile after photoexcitation at 355 nm (350 nm for AVOCTO1 10a). An approximate ground state bleach recovery for AVOCTO1 10a and AVOCTO4 10d was ~20% and for avobenzone it was ~30%.
NMR study of AVOCTO4 10d photostability during TEAS analysis.
A complementary study that was carried out to confirm the stability of AVOCTO4 10d involved taking an NMR spectrum before and then immediately after TEAS analysis, in deuterated acetonitrile. The two spectra produced are displayed below ( Figure S14). There was no significant change in the ratio of the peaks that correspond to the relative amount of the enol and keto forms (marked with dashed blue lines on the spectra). This study suggests that the same molecular entity is present before and after irradiation; but does not preclude that other molecules are formed that were not detected. Combined with the ultrafast studies this seems to confirm that the processes that produce a loss of activity in the UV are ultrafast and then, at longer times, the parent molecule is recovered. S59 Figure S14. NMR spectra of AVOCTO4 10d before and after 60 minutes irradiation (same conditions as before in the longer pump wavelength scheme) in acetonitrile. S60

6) Computational AVOCTO calculations.
All calculations were performed using the NWChem software package. 44 Due to the large number of atoms and single bonds in the full avobenzone-octocrylene molecules, it was not possible to achieve convergence at the lower level of theory, hence only the truncated forms were studied. Density functional theory (DFT) geometry optimisations for the chelated enol and diketo structures of the avobenzone section of each molecule were studied only. These geometry optimisations were carried out to determine the most stable, lowest energy conformations in the ground state. These calculations were conducted in implicitly modelled acetonitrile, using the conductor-like screening model (COSMO, with SMD) built into NWChem. 45,46 The relaxation of the initial enol and diketo structures of each truncated molecule was initially carried out using DFT at the PBE0/6-31g* level of theory. This initial structure was then further optimized by increasing the basis set to 6-311++g**, before arriving at the final structure, which was calculated at the PBE0/6-311++g** level of theory.
Once the six optimised structures were attained, time-dependent DFT (TD-DFT) was carried out to attain the vertical excitation energies of the singlet (S1-5) states of each species in acetonitrile using the same COSMO model, using TD-DFT at the PBE0/6-311++g** level of theory. The state characters were also calculated during these TD-DFT calculations and assigned manually. S61   Table S1. Predicted singlet excited state vertical excitation energies for AVOCTO1 10a in its diketo and enol form calculated at the TD-DFT/PBE0/6-311++g** level of theory in implicitly modelled acetonitrile. The evolution-associated difference spectra are included here to supplement Figure   7 and Table 4 of the main paper. These traces are effectively fits to the experimental data and are summarised by the time constants in Table 4. NB: Second trace τ2 has amplitude in the SE region (500 nm) while the fourth trace τ4 is purely a GSB containing no contribution from either the ESA (400 nm) or SE regions. Figure S15. EADS of AVOCTO1 10a and AVOCTO4 10d presented as traces with associated time constants τ1 to τ4.