Ab initio G2M calculations have been performed to investigate the potential energy surface for the reaction of C6H5 with O2. The reaction is shown to start with an exothermic barrierless addition of O2 to the radical site of C6H5 to produce phenylperoxy (1) and, possibly, 1,2-dioxaspiro[2.5]octadienyl (dioxiranyl, 8) radicals. Next, 1 loses the terminal oxygen atom to yield the phenoxy + O products (3) or rearranges to 8. The dioxiranyl can further isomerize to a seven-member ring 2-oxepinyloxy radical (10), which can give rise to various products including C5H5 + CO2, pyranyl + CO, o-benzoquinone + H, and 2-oxo-2,3-dihydrofuran-4-yl + C2H2. Once 10 is produced, it is unlikely to go back to 8 and 1, because the barriers separating 10 from the products are much lower than the reverse barrier from 10 to 8. Thus, the branching ratio of C6H5O + O against the other products is mostly controlled by the critical transition states between 1 and 3, 1 and 8, and 8 and 10. According to the calculated barriers, the most favorable product channel for the decomposition of 10 is C5H5 + CO2, followed by pyranyl + CO and o-benzoquinone + H. Since C6H5O + O and C5H5 + CO2 are expected to be the major primary products of the C6H5 + O2 reaction and thermal decomposition of C6H5O leads to C5H5 + CO, cyclopentadienyl radicals are likely to be the major product of phenyl radical oxidation, and so it results in degradation of the six-member aromatic ring to the five-member cyclopentadienyl ring. Future multichannel RRKM calculations of reaction rate constants are required to support these conclusions and to quantify the product branching ratios at various combustion conditions.