P), or oxygen atom transfer directly from DBTO. The absence of spectroscopic techniques for condensed-phase detection is an obstacle in studying very small and relatively short-lived intermediates such as O( 3 P). Kautsky and de Brujin faced the same problem in their efforts to uncover singlet oxygen. 3,4 Their solution was a "three-phase test" involving the photosensitizer dye, trypaflavine, and an oxygen acceptor dissolved separately on SiO 2 gel beads, which allowed for a millimeter of air separating the two molecules. By physically separating the site of oxidant generation from the site of oxidation, the experiment elegantly demonstrated that the oxidant produced upon irradiation was capable of diffusing through air. A challenge for highly reactive oxidants like O( 3 P) is the need for very short distances between reactants, which cannot be achieved with Kautsky's three-phase test. Porous shells of polymer nanocapsules offer a barrier in solution, capable of physically separating relatively large molecules while allowing for the diffusion of small molecules through very small pores (diameter, <1 nm) in the nanocapsule shell. 5 Small-angle neutron scattering revealed that the thickness of the shells in these vesicle-templated capsules is 1.0 ± 0.1 nm. 6 Long-term stability studies of nanocapsules showed no measurable efflux of molecules larger than the pore size over five years. 7,8 For example, nanocapsules loaded with pH-sensitive indicator dyes showed unhindered transport of protons while being impermeable to molecules larger than the pore size.

We designed an experiment involving the irradiation of DBTOloaded nanocapsules in the presence of an O( Therefore, we prepared a water-soluble DBTO derivative (1a). 13 The functionalization was shown to have no significant ameliorating effect on photodeoxygenation properties (Table S1). The optimal O( 3 P)-trap, 2a, was synthesized using a known procedure. Please do not adjust margins Please do not adjust margins through the oxidation of 2a by traditional means (i.e. using mCPBA). As shown in Figure 1, irradiation of 1a-loaded nanocapsules would only be expected to yield 2b if a freely diffusing oxidant was generated. The synthesis of polymer nanocapsules was accomplished using an aqueous suspension of self-assembled vesicles as templates with bilayers loaded with hydrophobic monomers and cross-linkers as scaffolds. Nanocapsules were constructed from vesicle templates which formed spontaneously in water with appropriate concentrations of surfactants, For loading the nanocapsules, a solution of water containing ~1-10 mM of a molecule to be encapsulated was used as the solvent (Figure 2A). Prior to polymerization, the vesicle size was monitored using dynamic light scattering (DLS) (Fig. 2B). A small size distribution centered at 100-200 nm was achieved either by ~2 h equilibration times or via extrusion. Following thermal polymerization (65°C for 8-12 h), nanocapsules were precipitated using methanol and then separated from the reaction solution by centrifugation and the decanting of the supernatant. The nanocapsules were washed extensively (>20 total washes) with methanol-water solutions, followed by water and acetonitrile. SEM Analysis of freeze-dried 6,19 nanocapsules confirmed the presence of spherical nanocapsules within the desired size range (Fig. 2C). Nanocapsule pore sizes were estimated by encapsulation of three dyes (Fig. S1) that were used as size probes: Procion Red (1.1 nm), Nile Blue A (1.0 nm), and 4-(phenylazo)benzoic acid (0.6 nm) in a similar fashion to a previously published size probe retention assay. 5,20, Following washing, Procion Red and Nile Blue A were shown to be retained by the nanocapsules, while 4-(phenylazo)benzoic acid was not suggesting an average nanocapsule pore size >0.6 nm and <1.0 nm. Using M06-2X/6-31G(d,p) geometry optimization and frequency calculations, the calculated diameters of the smallest cross-section of 1a and 2a were 1.01 and 1.40 nm, respectively, both of which were larger than the estimated nanocapsule pore size.

To demonstrate that 1a and 2a were unable to pass through the nanocapsules barrier, fluorescent derivatives were encapsulated and completely retained by the nanocapsules. Like DBTO, 1a was not fluorescent, but the deoxygenation product of 1a, i.e. the sulfide 1b, was fluorescent. Nanocapsules loaded with 2a could not be prepared since 2a is not water-soluble; however, a water-soluble derivative of 2a, i.e. 3, was prepared and found to fluoresce. Fluorescence spectroscopy of 1b-loaded nanocapsules and 3-loaded nanocapsules was performed at an excitation wavelength (λ ex ) of 270 nm, with emission peaks at 368 and 344 nm, respectively. As shown in Fig. 3A-B, when compared to the spectra of 1b and 3 alone, the results indicated that both 1b and 3 were successfully encapsulated and retained after extensive washing. Since 3 had limited solubility in water, lower concentrations of 3 (≤1 mM) were used to prepare 3loaded nanocapsules, which was proposed as the cause of the weak emission observed for the 3-loaded nanocapsules.

Nonetheless, the successful encapsulation of 1b and 3 suggested that 1a and 2a are too large to pass through pores of the nanocapsule barrier, and thus, sufficient separation of 1a and 2a can be achieved. Please do not adjust margins Please do not adjust margins 1a-loaded nanocapsules were irradiated with broadly emitting UV light (fwhm, 325-375 nm), and fluorescence spectroscopy was performed at 0, 3, and 5 h (Fig. 3C). A fluorescence spectrum consistent with 1b, increased over time, which demonstrated that encapsulation did not prevent photodeoxygenation of 1a. After the last fluorescence spectra was taken, the nanocapsules were filtered off, and the HPLC analysis of the supernatant did not reveal any trace of 1a or 1b, indicating that no leakage occurred during photolysis. Together these results confirmed that 1a is encapsulated and retained in the nanocapsules throughout photolysis.

To determine if the photodeoxygenation of 1a generates a small diffusing oxidant, 1a-loaded nanocapsules with 2a present in the exterior solution were irradiated as shown in Figure 1. Nanocapsules loaded with 1a were used for experimental trials, while 1b-loaded nanocapsules and "empty" nanocapsules were used as two different types of photocontrol trials. Experimental and photocontrol trial solutions contained nanocapsules and 20 ± 2 mM of 2a dissolved in acetonitrile. Two different degassing methods were examined: argon-sparging and freeze-pump-thaw. Degassing via argon-sparging is known to leave behind residual O 2 , while a freeze-pump-thaw method can reduce O 2 to insignificant concentrations.

1 Degassed solutions were irradiated using broadly emitting fluorescent bulbs (fwhm 325-375 nm) for 5 h. By a procedure described in SI, the maximum concentration of 1a (in nanocapsules) in the experimental solutions was estimated to be 8.5 mM or lower. A total of twelve experimental and ten photocontrol trials were performed. Overall, the photolysis of 1a-loaded nanocapsules in the presence of 2a resulted in the formation of 8-11 µM of 2b. Depending on the method of degassing, little (<3 µM) or no 2b was observed in photocontrol experiments containing 2a and 1b-loaded nanocapsules or "empty" nanocapsules. The results of ≥6 trials for the photolysis of argon-sparged solutions (Fig. 4A) indicate that the average change in concentration of 2b in experimental solutions was 10.4 µM, amounting to a 4x increase in 2b formation in the experimental trials relative to the photocontrols. On average, the increase in 2b observed in photocontrol experiments with "empty" nanocapsules was 2.7 µM. In photocontrol solutions containing 1b-loaded nanocapsules, there was no increase in 2b concentration. Three trials were performed under freeze-pump-thaw conditions for photocontrols and experimental trials (Fig. 4B). The initial concentration of 2b was found to be zero in each trial. The photolysis of 1a-loaded nanocapsules in the presence of 2a resulted in the formation of 8.4 µM of 2b, on average. In the controls degassed by freeze-pump-thaw, where O 2 is presumably insignificant, there was no change in the concentration of 2b after irradiation. Therefore, we attributed the formation of 2b observed in the argon-sparged control experiments (Figure 2A, photocontrol) to the presence of residual O 2 in solution. In the absence of O 2 , the oxidation of 2a was only observed upon photolysis of 1a in free solution or 1a-loaded nanocapsules, confirming that the oxidant resulted from 1a photodeoxygenation. As a control, we examined if 2b could be the result of a thermal reaction or direct photoproduct of 2a. Minor photochemical degradation of 2a was observed; however, 2b was not observed in the absence of O 2 . Using GC-MS, products of degradation were identified as thiophenol, diphenyl disulfide, 2,2-bis((phenylthio)methyl)-propane (4a), and 2,2bis((phenylthio)methyl)-cyclopropane. In the dark under ambient air, 2a was found to oxidize to 2b; although, the process occurred very slowly over a period of a month. In an additional control, a solution containing 80 µM 1a, "empty" nanocapsules, and 2a (degassed by freeze-pumpthaw) was photolyzed for 5 h, resulting in complete conversion of 1a to 1b. Following photolysis, the observed concentration of 2b was 7.4 µM, or slightly less than the average observed in solutions containing 1a-loaded nanocapsules. If 1a leaked from 1a-loaded nanocapsules to cause the oxidation of 2a observed This journal is © The Royal Society of Chemistry 20xx

Please do not adjust margins Please do not adjust margins in experimental solutions, then the concentration of leaked 1a would have to be at least 80 µM. The detection limit of 1a and 1b was found to 0.5 µM, and neither 1a or 1b were observed in the supernatant following irradiation. Therefore, the increase in 2b observed in the experimental trials cannot be explained by leakage of 1a. The polymer nanocapsules were measured to have a nanometer-thick shell, 6 which is about the size of the smallest cross-section of 1a and 2a. In the synthesis of rotaxane-like structures, a short linker threaded through a nanopore was not able to connect two molecules located on opposite sides of the shell. 21 This observation suggested that direct physical contact between 1a and 2a was extremely unlikely. Nevertheless, we considered a hypothetical scenario of physical contact between 1a and 2a through an event involving the partial insertion of 1a and 2a into the same small hole (or pore) on either side of the nanocapsule shell. Since 1a has a rod-like shape and 2a has a dendritic shape, the most likely collision event within the pore would be between the phenyl groups of 1a and 2a. This unlikely event of direct oxygen atom transfer would likely require arene oxide intermediates, but no phenolic products of 2a, which would be expected for arene oxide intermediates, were observed in any experiment. A collision event between the sulfoxide of the excited 1a and the sulfide of 2a would be required for direct oxygen atom transfer resulting in 2b. The probability of a productive collision event within a pore of the nanocapsules is very small and cannot explain the 8-11 µM increase in 2b. The diffusion distance of O( 3 P) in this system is predicted to be slightly less than 65 nm, 17 and thus, a freely diffusing O( 3 P) would be capable of traversing the nanocapsule intact. The results from the experiments described above demonstrate that photodeoxygenation of 1a inside of the nanocapsules generates a freely diffusing intermediate that oxidizes 2a to 2b. The oxidation of 2a to 2b upon irradiation in the presence of O 2 (Figure 2A, photocontrol) raises the possibility that the intermediate formed could be O 2 . While the current experiments cannot rule this out, the preponderance from previous studies have led to the conclusion that the direct irradiation of DBTO and its derivatives result in photodeoxygenation by a unimolecular mechanism. Additionally, the selective irradiation of DBTO in the presence of diphenyl sulfoxide produced no diphenyl sulfide, which would be expected if a bimolecular exciplex was involved in the photodeoxygenation mechanism. 1 The possibility of two O( Applying Occam's razor, the simplest explanation of the oxidation of 2a through an impermeable barrier upon irradiation of 1a is that the photodeoxygenation of 1a generates a small freely diffusing oxidant. Since 1a undergoes deoxygenation by unimolecular mechanism and has same chemoselectivity as DBTO, these results are consistent with the identity of this freely diffusing oxidant as being O( 3 P). The employed experimental scheme answers a key mechanistic question about the nature of the oxidant in these photooxygenations. The experimental scheme also offers a viable method for studying other short-lived reactive intermediates.