Solvent-dependent complex reaction pathways of bromoform revealed by time-resolved X-ray solution scattering and X-ray transient absorption spectroscopy

The photochemical reaction pathways of CHBr3 in solution were unveiled using two complementary X-ray techniques, time-resolved X-ray solution scattering (TRXSS) and X-ray transient absorption spectroscopy, in a wide temporal range from 100 ps to tens of microseconds. By performing comparative measurements in protic (methanol) and aprotic (methylcyclohexane) solvents, we found that the reaction pathways depend significantly on the solvent properties. In methanol, the major photoproducts are CH3OCHBr2 and HBr generated by rapid solvolysis of iso-CHBr2-Br, an isomer of CHBr3. In contrast, in methylcyclohexane, iso-CHBr2-Br returns to CHBr3 without solvolysis. In both solvents, the formation of CHBr2 and Br is a competing reaction channel. From the structural analysis of TRXSS data, we determined the structures of key intermediate species, CH3OCHBr2 and iso-CHBr2-Br in methanol and methylcyclohexane, respectively, which are consistent with the structures from density functional theory calculations.


INTRODUCTION
An understanding of reaction mechanisms at atomic and electronic levels is one of the central themes in chemistry; many efforts have been devoted toward that goal with advances in chemical reaction dynamics. In particular, the development of time-resolved techniques has contributed to revealing mechanisms for various photochemical reactions in the gas phase and condensed media. Nevertheless, even for reactions of small molecules with simple chemical structures, it is still challenging to identify all of the major reaction intermediates along reaction pathways. One such example is the photochemistry of bromoform, CHBr 3 . Upon ultraviolet (UV) excitation, CHBr 3 dissociates to generate reactive bromine, which effectively depletes ozone in the troposphere and/or the stratosphere. 1,2 Due to both fundamental interest and environmental importance, photochemistry of CHBr 3 has been studied intensively using time-resolved optical spectroscopy. [3][4][5][6][7][8][9][10][11] However, the comprehensive photochemical reaction mechanisms of CHBr 3 have not been completely understood, mainly because this small molecule with an ostensibly simple structure exhibits unusually complex reaction pathways.
In the gas phase, upon excitation at UV wavelengths, the primary reaction channel of CHBr 3 is the cleavage of the C-Br bond to produce a CHBr 2 radical and a Br atom, CHBr 3 þ hv ! CHBr 2 þ Br: (R1) Subsequently, secondary dissociation of the CHBr 2 radical occurs to generate CBr þ HBr or CHBr þ Br (see Table S1). 9,10 Direct elimination of Br 2 is another possible reaction channel (CHBr 3 þ hv ! CHBr þ Br 2 ). 11 In the condensed phase, studies were focused on the dynamics of the iso-CHBr 2 -Br, which can be formed through roaming-mediated isomerization in the excited state 6 or solvation cage-induced geminate recombination of CHBr 2 and Br, [3][4][5][6] CHBr 3 þ hv ! iso À CHBr 2 À Br: (R2) Other short-lived intermediates or stable photoproducts, such as Br and Br 2 that react directly with O 3 (Br 2 þ hv ! 2Br, Br þ O 3 ! BrO þ O 2 ), have been rarely reported in solution; it might have been difficult to globally detect all the photoproducts, especially irreversible and stable species, due to limited time ranges (up to only a few nanoseconds) in most of the previous time-resolved spectroscopic studies. In addition, the photoproducts can further react with the surrounding solvent molecules in various ways, as exemplified by the Wohl-Ziegler reaction, in which the hydrogen in a solvent molecule is abstracted by the bromine atom. 7 An additional reaction can occur when the solvent is a nucleophile. For example, it was reported by Kwok et al. 3 that iso-CHBr 2 -Br formed by photolysis of CHBr 3 in aqueous solution undergoes hydrolysis, resulting in complex chemistry. Although time-resolved optical spectroscopy is highly sensitive to the formation of reaction intermediates, even for solution samples at low concentrations, it provides only limited structural insight in the reaction intermediates. In contrast, time-resolved X-ray solution scattering (TRXSS), which is also known as time-resolved X-ray liquidography (TRXL), is highly sensitive to the details of the molecular structure. TRXL can provide direct information on changes in the global molecular structure, which cannot be directly extracted from time-resolved optical spectroscopy, because X-rays are scattered off all atomic pairs present in sample molecules and the scattering pattern contains direct information on the distances of all atomic pairs.  Complementary to TRXL, X-ray transient absorption (XTA) spectroscopy can provide the dynamics of the electronic structure and local geometric structure around light-absorbing atoms by probing the excitation of core electrons to modified valence electron orbitals. [51][52][53][54] These two complementary techniques can be combined together to obtain more complete information on the reaction dynamics as was demonstrated in recent studies. 13,55 In this work, we applied TRXL and XTA to the photochemical reaction of CHBr 3 in methanol and methylcyclohexane (see Figs. 1 and 2). By simultaneously analyzing the datasets from TRXL and XTA, we characterized the dynamics of geometric and electronic structures to completely describe the reaction dynamics of CHBr 3 in solution. The results revealed the rich photochemistry of CHBr 3 including the formation of Br and Br 2 that directly relate to ozone depletion and solvent-dependent reaction pathways. Stable photoproducts, CH 2 Br 2 , BrCH 2 OH, C 2 H 2 Br 4 , cis-C 2 H 2 Br 2 , and trans-C 2 H 2 Br 2 were also observed and characterized using gas chromatography mass spectrometry (GCMS) and NMR spectroscopies.

RESULTS AND DISCUSSION TRXL measurement in methanol
To follow the photochemical reaction dynamics of CHBr 3 , we collected a series of TRXL data at various delay times as shown in Fig.  1(a). The experimental difference scattering curves, qDS(q), were obtained by subtracting the reference scattering signal measured at a negative time delay (-3 ns) from the scattering signals measured at positive delay times. It can be seen in Fig. 1(a) that the oscillatory features of the difference signal at positive delay times change considerably with time, reflecting the structural changes of the reacting molecules.
The total scattering signal S(q) from liquid can be described by the Fourier transform of the statistical correlation between scattering sites, g(r), 56 which can be expressed as the sum of contributions from the solute, solvent, and solute-solvent interaction (that is, the solvation cage). To explain the measured difference scattering signal, we fitted the experimental data at various delay times using theoretical scattering curves, which are a linear sum of these three contributions, as shown in Fig. S1 in the supplementary material. For the solute species, all possible reaction pathways were considered as shown in Table S2 in the supplementary material. Specifically, we considered the following solute species: CHBr 3 , CHBr 2 , CHBr 2 þ , Br, CBr 2 , CHBr, iso-CHBr 2 -Br, Br 2 , CH 3 OCHBr 2 , HBr, CH 2 Br 2 , (antiand gauche-) C 2 H 2 Br 4 , (transand cis-) C 2 H 2 Br 2 , BrCH 2 OH, and CH 3 Br. The geometries of these species were optimized using density functional theory (DFT) calculations, as shown in Fig. S2. The data analysis is described in Secs. 2 and 3 in the supplementary material and Fig. S1, while its detailed description can be found in previous reports. 12,15,19 Although the difference scattering curves, qDS(q,t), shown in Fig.  1(a) constitute a complete "fingerprint" of the structural rearrangements, the difference radial distribution functions (RDFs), rDR(r,t), in real space, shown in Fig. 1(b), provide a more intuitive picture of the changes in the molecular structure. The difference RDF represents the change in the atom-atom pair distribution function during the course of the reaction and thus is a measure of the change in radial electron density around an (average) excited atom as a function of interatomic distance r, weighted by the X-ray form factor. As shown in Fig. 1(b), several positive and negative peaks appear and evolve over time. Unequivocal peak assignments can be made by decomposing the signal into the contributions of the solute, solvent, and solvation cage as illustrated in Fig. S1. The negative peak at 3.25 Å appears at early delay times and can be attributed to the depletion of the BrÁÁÁBr interatomic pairs in CHBr 3 . Subsequently, strong positive (at $5.5 Å ) and negative (at $4 Å and $7 Å ) peaks become dominant after 30 ns, and can be Structural Dynamics ARTICLE scitation.org/journal/sdy ascribed to the changes in the intermolecular distances of solvent molecules due to thermal expansion of the solution. 19 We performed a structural analysis of the difference scattering curve at 100 ps as shown in Fig. S3, and the analysis result shows that reaction (R1) fits the data much better than reaction (R2), which was postulated to be a major reaction channel in the condensed phase. [3][4][5][6] As can be seen in Fig. S3, the inclusion of iso-CHBr 2 -Br deteriorates the fitting quality significantly. When both reactions (R1) and (R2) FIG. 2. TRXL and XTA signals measured for CHBr 3 in methylcyclohexane after 267 nm laser excitation. (a) Difference scattering curves, qDS(q), measured at various delay times (black) and their theoretical fits (red) from a global fitting analysis described in the text. (b) Difference radial distribution functions, rDS(r), obtained by the sine-Fourier transform of qDS(q) shown in (a). The low q signals from 100 ps to 30 ns, and from 50 ns to 1 ls in (A), are divided by 3 and 6, respectively, for clarity. (c) XTA signals measured at various delay times. Experimental XTA spectra (black) were fit by theoretical XTA spectra (red) calculated from a global fitting analysis. Structural Dynamics ARTICLE scitation.org/journal/sdy were included in the analysis, the contribution of Reaction R2 converged to zero.

XTA measurement in methanol
With 267 nm excitation, various reaction channels are energetically allowed as shown in Table S1, and thus all of them were considered in the analysis of TRXL data. Because light atoms (C and H) have much lower scattering intensities than heavy atoms (Br), TRXL is less effective for distinguishing chemical species that have the same heavy atoms but different light atoms. For example, as shown in Fig. S4, TRXL cannot unambiguously distinguish Reaction R1 from another reaction, CHBr 3 ! CBr 2 þ HBr. To complement such limitation of TRXL and obtain electronic structural dynamics, we performed XTA measurement on the photochemical reaction of CHBr 3 because the method is sensitive to the change in electronic and local geometric structures around the X-ray absorbing atoms. Considering that the Br of CHBr 2 and free Br atoms have unrestricted open electron shells while the Br of CBr 2 and HBr are closed electron shell species, the two reaction channels should yield distinctly different XTA signals at the Br K-edge.
The Br K-edge XTA spectra in the X-ray absorption near edge structure (XANES) region for CHBr 3 in methanol were measured at various delay times from 263 nm excitation, as shown in Figs. 1(c) and 3(a).
A positive peak at 13 475 eV is observed at 120 ps, which corresponds to the absorption feature of Br (1s ! 4p transition), 57 providing a direct evidence for the formation of Br upon laser excitation of CHBr 3 . This feature disappears at later delay times, indicating the participation of Br in the formation of other species. It can be seen that the negative peak at 13 480 eV is shifted to higher energies over time while its intensity remains constant in the entire time range of our measurement (see Fig.  S5). The negative peak arises from the depletion of ground-state molecules induced by photoexcitation. The blue shift of the negative peak with a nearly constant intensity suggests that, at late delay times, reaction intermediates formed at the onset of the reaction are transformed into other intermediate species rather than returning to the ground state.
To extract the structural dynamics of transient intermediates, the experimental XTA spectra were fitted with the theoretical XANES spectra calculated from putative reaction intermediates [see Fig. 3 In particular, we identified the intermediates by fitting the XTA signal at 120 ps with reaction (R1) and CHBr 3 ! CBr 2 þ HBr, which exhibit distinctly different XTA spectra as shown in Fig. 3(a). In fact, in the fitting analysis of the 120 ps spectrum, the contribution of the latter channel (CHBr 3 ! CBr 2 þ HBr) converges to zero, confirming that CHBr 3 dissociates into the CHBr 2 radical and Br, rather than CBr 2 and HBr. The formation of CHBr 2 was confirmed by applying electron paramagnetic resonance (EPR) spectroscopy to a photolyzed solution of CHBr 3 in methanol frozen at 77 K (Sec. 5 in the supplementary material). A previous transient Raman spectroscopic study on the photochemical reaction of CHBr 3 in methanol also showed that the major product at 5 ns delay time is the CHBr 2 radical. 58 This result clarifies the uncertainty regarding the identities of intermediates that cannot be unequivocally resolved by the analysis of TRXL data.
Meanwhile, in accordance with the result of the TRXL analysis, XTA confirms that iso-CHBr 2 -Br is not observed at 120 ps, as shown in Fig. S6. While CHBr 2 and Br are identified as the intermediates at the earliest delay time, the analysis of the XTA spectrum at 120 ps shows that reaction (R1) alone cannot produce a perfect fit to the experimental data as can be seen in the upper plot in Fig. 4(a), indicating that other intermediates are also produced at the early stage of the photochemical reaction of CHBr 3 . The discrepancy between the experimental and theoretical spectra becomes more pronounced at late delay times [for example, at 3 ls, as can be seen in the upper plot in Fig. 4(b)]. By considering all the possible reaction pathways listed in Table S2, we found that the following solvolysis reaction is required to achieve good agreement between the experimental and simulated XTA spectra on time scales from picoseconds to microseconds, as can be seen in the lower plots in Figs. 4(a) and 4(b), Inspired by the result of this XTA data analysis, we reanalyzed the TRXL data by considering both Reaction R3 and Reaction R1. As a result, the fitting qualities of the TRXL data improve significantly at both early and late delay times, as can be seen in Figs. 4(c) and 4(d).
The solvolysis reaction of iso-CHBr 2 -Br [reaction (R3)] is highly exothermic (see Table S2), and a similar solvolysis reaction was observed in a transient resonance Raman spectroscopic study of CHBr 3 in water. 3 The reaction between CHBr 2 and one solvent molecule (CHBr 2 þ CH 3 OH ! CH 3 OCHBr 2 þ H) can also generate CH 3 OCHBr 2 , but it is an endothermic reaction with an energy barrier of 130.43 kJ/mol (Table S2) and thus is less favorable. In fact, the formation of HBr respectively. The combined global fitting approach, which is described in detail in Sec. 3 in the supplementary material, is a fully consistent and complete data analysis framework because it takes advantage of both the local electronic element-specific sensitivity of XTA and the global structural and thermodynamic sensitivity of TRXL. According to the kinetics determined from the TRXL data shown in Fig. 5(a), CHBr 2 , Br, CH 3 OCHBr 2 , and HBr are present at the onset of the    Table S5. The concentrations of CH 3 OCHBr 2 and HBr stay constant until they further increase at $40 ns, indicating that these species remain stable over the entire time interval of our measurement and the formation of these species occurs biphasically. We note that the increase in the concentrations of HBr and CH 3 OCHBr 2 at $40 ns is accompanied by the decrease in the concentration of CHBr 2 . We also noted that the amount of consumed Br is larger than the amount of formed Br 2 . These observations suggest that nongeminate association of CHBr 2 and Br at 40 ns generates iso-CHBr 2 -Br, which promptly reacts with methanol to form CH 3 OCHBr 2 and HBr. The association of CHBr 2 and Br to form iso-CHBr 2 -Br is energetically uphill, yet the solvolysis is strongly downhill, resulting in the overall exothermic reaction. According to our DFT calculations, the reaction barrier for the formation of iso-CHBr 2 -Br would be prohibitively high, and we suggest that the overall reaction is solvent-assisted.
Although the concentration profiles determined by TRXL [ Fig.  5(a)] and XTA [ Fig. 5(b)] are not identical because of the different concentrations of CHBr 3 solutions used for these two measurements, they share a common set of rate constants. We note that iso-CHBr 2 -Br is absent in the kinetics extracted from either TRXL [ Fig. 5(a)] or XTA [ Fig. 5(b)]. The absence of iso-CHBr 2 -Br does not necessarily mean that it is not formed, but it may suggest that iso-CHBr 2 -Br has transformed into other species within our time resolution ($100 ps), probably via the solvolysis reaction with methanol to generate CH 3 OCHBr 2 and HBr. Indeed, a previous study on the photochemical reaction of CHBr 3 in water using transient resonance Raman spectroscopy 3 revealed that iso-CHBr 2 -Br reacts rapidly with H 2 O to generate CHBr 2 OH and HBr via hydrolysis and thus was not detected on the picosecond time scale. In this study, CH 3 OCHBr 2 and HBr rather than iso-CHBr 2 -Br, are observed at the earliest delay time as can be seen in Figs. 5(a) and 5(b), which suggests that HBr and CH 3 OCHBr 2 were generated via Reaction R3, the solvolysis of iso-CHBr 2 -Br with methanol, within 100 ps.
Our kinetic analysis shows that CH 3 OCHBr 2 is an important photoproduct that survives even until late delay times in methanol as shown in Figs. 5(a) and 5(b). In Fig. 4, we showed that the formation of CH 3 OCHBr 2 is required to obtain a good fit to the experimental XTA and TRXL data at both early and late delay times. To determine the structure of CH 3 OCHBr 2 , in Fig. 6(a), we show the solute-only (specifically, CH 3 OCHBr 2 -only) difference scattering curve at 10 ns obtained by subtracting the solvent and cage terms and the contributions of other reaction channels from the experimental difference scattering curve. For comparison, we also show the theoretical difference scattering curve calculated for CH 3 OCHBr 2 , which gives an excellent agreement with the experimental curve. Their corresponding radial intensities are shown in Fig. 6(b). The large negative dip at 3.25 Å corresponds to the depletion of the BrÁÁÁBr interatomic pair in CHBr 3 . The small negative dip at 1.8 Å and a small peak at 2.2 Å are caused by a superposition of the C-Br distance of 1.9 Å in the depleted CHBr 3 and the prolonged C-Br distance of 2.09 Å in the generated CH 3 OCHBr 2 , which is a unique signature of CH 3 OCHBr 2 . The C-Br distances in other intermediates are close to the C-Br distance in the parent CHBr 3 molecule and thus, in contrast to the feature of   Fig. 6(b). The XTA spectra further confirm that CH 3 OCHBr 2 is the major photoproduct in methanol at late delay times. In Fig. 3, the XTA spectrum at 20 ls is compared with the calculated XTA spectra of CH 3 OCHBr 2 and CHBr 2 , and it can be seen that CH 3 OCHBr 2 alone can reproduce most of the features in the experimental XTA spectrum.

TRXL and XTA measurements in methylcyclohexane
The presence of O-H bonds in the solvent is essential to drive a solvolysis reaction. 3 In aprotic solvents, O-H bonds are absent and the solvolysis is not expected to occur, thus presumably leading to the increase in the lifetime of iso-CHBr 2 -Br. Indeed, previous studies showed that the lifetime of iso-CHBr 2 -Br depends strongly on the solvent properties. In nonpolar solvents such as methylcyclohexane and cyclohexane, iso-CHBr 2 -Br survives up to several nanoseconds to microseconds, 4,5 but it decays rapidly within 13 ps in neat bromoform 4 and within 1 ns in polar solvents such as acetonitrile. 5 To determine the structure of iso-CHBr 2 -Br in solution, we also studied the photochemistry of CHBr 3 in methylcyclohexane using TRXL and XTA. Figure 2 shows the TRXL and XTA data of CHBr 3 in methylcyclohexane at various delay times together with the least squares fit. The same analysis protocol used for the data in methanol was applied to the data in methylcyclohexane. Figure S7 in the supplementary material and Fig. 7 show the analysis results of the TRXL and XTA data, respectively. The analyses of these data as shown in Figs. 7 and S8 indicate that the isomer formation channel is required to obtain a satisfactory fit to both experimental XTA and TRXL data in methylcyclohexane.
Time-dependent concentration changes of various photoproducts were extracted from the simultaneous GF analyses of the TRXL and XTA data as shown in Figs. 5(c) and 5(d), respectively. According to the kinetics determined from the TRXL data, at the onset of the reaction, the CHBr 2 radical, Br, and iso-CHBr 2 -Br coexist, which is in contrast to the reaction in methanol where iso-CHBr 2 -Br is absent. Then, iso-CHBr 2 -Br returns to CHBr 3 in 73 ns with a unimolecular rate constant of 7.7 6 1.6 Â 10 6 s À1 . The Br atoms recombine to generate Br 2 with a bimolecular rate constant of 5.04 6 0.7 Â 10 8 M À1 s À1 , which is smaller than that in methanol by an order of magnitude. CHBr 2 stays stable on the time scale of microseconds.
As was the case for the result in methanol, the XTA data also provide direct information on the reaction dynamics, as shown in Fig.  2(c) and S9 in the supplementary material. The positive peak at 13 475 eV at early delay times corresponds to the 1s ! 4p transition of Br and arises from the formation of Br, which disappears at late delay times when Br atoms react to form other species. In contrast to the reaction in methanol, the negative peak in the difference XTA spectra in methylcyclohexane stays at the same energy throughout the time interval while its intensity decreases by about a factor of two at late delay times, suggesting that $50% of the photoproducts return to CHBr 3 (Fig. S9). Indeed, in methylcyclohexane, iso-CHBr 2 -Br accounts for about half of the photoproducts and relaxes back to CHBr 3 in tens of nanoseconds [ Fig. 5(d)], while the recovery of CHBr 3 was not observed in methanol in the time range of our measurements.
To better visualize the formation of iso-CHBr 2 -Br in methylcyclohexane at the onset of the reaction, the contributions of the solvent and cage terms as well as the contribution of reaction (R1) were subtracted from the difference X-ray scattering data at 100 ps. The resulting solute-only (specifically, iso-CHBr 2 -Br-only) difference scattering curve and its difference radial intensity agree well with the theoretical scattering curve and its radial intensity calculated for the isomerization reaction as shown in Figs. 6(e) and 6(f), respectively, confirming the formation of iso-CHBr 2 -Br. In the difference radial intensities shown in Fig. 6(f), the large negative dip at 3.25 Å corresponds to the depletion of the BrÁÁÁBr interatomic distance in CHBr 3 while the positive peaks at 2.5 Å and 5.1 Å arise from the Br-Br 0 and Br 0 ÁÁÁBr 00 interatomic pairs, respectively, which are characteristic of iso-CHBr 2 -Br.

Structural refinement of CH 3 OCHBr 2 in methanol and iso-CHBr 2 -Br in methylcyclohexane
We refined the geometric structures of CH 3 OCHBr 2 in methanol and iso-CHBr 2 -Br in methylcyclohexane using the TRXL data, as shown in Fig. 6. The structural parameters were optimized together with the excited state fraction (c) to compensate for the reduction of the amplitude of the difference scattering curve as the structural parameters are varied. The confidence intervals and correlations of all parameters were obtained from the v 2 contour plots as explained in Sec. 4 in the supplementary material.  Fig. 6(a)]. To do so, we used the solute-only difference scattering curve at 10 ns, as shown in Fig. 6(a). Variation of either R or / changes the Br-Br 0 distance, indicating a strong negative correlation between R and / [ Fig. 6(c)]. On the other hand, H has a much weaker correlation with both R and / [see Figs. 6(d) and S10] because the methyl group consists of relatively light atoms compared with Br and therefore the change of its absolute position influences the simulated difference scattering signal only slightly. The optimal values for R, /, and H within the 68% confidence are R ¼ 2.  Table I.
For iso-CHBr 2 -Br, the Br-Br 0 bond length (R) and Br 0 -Br-C angle (/) were refined [see the inset in Fig. 6(e)]. To do so, we used the solute-only difference scattering curve at 100 ps, as shown in Fig. 6(e). As can be seen in Figs. 6(g) and 6(h), strong correlations are observed for both (R, /) and (R, c) pairs of parameters because the changes of R or / directly affect the Br 0 -Br 00 distance. The refined values of R and / with 68% confidence are R ¼ 2.68 6 0.02 Å and / ¼ 128 6 1 , which results in the Br 0 -Br 0 distance of 5.08 6 0.02 Å (see Table I).

Reaction energetics and thermodynamics of the bulk solvent
As the photogenerated intermediate species recombine back to the ground state, excess energy is released as heat to the surrounding solvent, resulting in changes of temperature, pressure, and density of the solvent, and ultimately the variation of intermolecular radial distributions in the solvent. 19 Although for a single solvent molecule, this variation might be negligible, given the large relative amount of solvent molecules over solutes (with a concentration of 40 mM in methanol as an example, the ratio between the number of solute and solvent molecules is 1:618), the integrated solvent signal can be comparable to or even larger than the solute-only signal. In the difference X-ray scattering signals, the transient solvent response typically dominates at late delay times after the intermediate molecules relax back to the ground state, and it has a large amplitude at low to medium q values or in a high r-region, as shown in Figs. S1 and S7. In the GF analysis, the time dependence of the thermodynamic quantities such as the changes in density Dq(t) and temperature DT(t) of the bulk solvent can be monitored as the heat is released from the solutes and propagates through the solution. Figure S11 shows the temporal variations of Dq(t) and DT(t) in methanol and methylcyclohexane after laser excitation.

Stable photoproducts
We identified the stable photoproducts by analyzing the photolyzed solution of CHBr 3 in methanol (see Secs. 5 and 6 in the supplementary material for the details). First, we noticed that the transparent CHBr 3 /methanol solution changed to orange color during the experiment (Fig. S12), which can be ascribed to the formation of stable Br 2 as confirmed by the electronic absorption spectra measured at UV to visible wavelengths (Fig. S13). Also, from the analysis of the photolyzed solution with GCMS and NMR spectroscopy, we found that CH 2 Br 2 , BrCH 2 OH, C 2 H 2 Br 4 , cis-C 2 H 2 Br 2 , and trans-C 2 H 2 Br 2 are formed with comparable yields (see Table S3 and Figs. S15-S27 in the supplementary material) from various reactions following the photoexcitation of CHBr 3 . For example, CH 2 Br 2 was formed through the following abstraction reaction: as was confirmed by the formation of CHDBr 2 from the photolysis of CHBr 3 in CD 3 OD. It should be noted that CH 2 Br 2 , which is observed from the photolysis of CHBr 3 in methanol, is not observed in methylcyclohexane, confirming that CH 2 Br 2 is formed via the abstraction reaction [reaction (R4)]. The identification of these various photoproducts is discussed in detail in Secs. 5 and 6 in the supplementary material. Including these species in the GF analysis does not improve the fitting quality significantly, and the concentration of these species is below the threshold of the signal-to-noise ratio of our experimental data. Our data analysis illustrates that these stable photoproducts are probably formed on longer time scales, and the secondary photochemistry of CHBr 3 involves numerous inter-related reactions.

CONCLUSION
Complex photochemical reaction pathways of CHBr 3 in methanol and methylcyclohexane were elucidated using picosecond TRXL and XTA. The homolytic C-Br bond cleavage leading to the formation of the CHBr 2 radical, Br, and Br 2 is a common reaction channel observed in both solvents. In methanol, the rapid solvolysis of iso-CHBr 2 -Br [reaction (R3)] yields CH 3 OCHBr 2 and HBr at 100 ps, which survive for tens of microseconds. In contrast, in methylcyclohexane, iso-CHBr 2 -Br does not undergo solvolysis and instead isomerizes back to CHBr 3 with a time constant of $70 ns, which is specific to the concentration and excitation energy used in our experiment. Our study illustrates the amazing complexity of the photochemical reaction pathways of CHBr 3 , which vary with the solvent properties, as schematically shown in Fig. 8. Besides this complexity, the combination of two time-resolved X-ray techniques proved to be capable of providing a complete description for all major photochemical reaction pathways for this environmentally important molecule.  Government retains for itself, and others acting on its behalf, a paid-up nonexclusive, irrevocable worldwide license in the said article to reproduce, prepare derivative works, distribute copies to the public, and perform publicly and display publicly, by or on behalf of the Government.