A theoretical investigation has been carried out on the reaction mechanism and kinetics of the thermal decomposition of hexafluoropropylene oxide (HFPO), a compound of importance in fluorocarbon thin film preparation in the electronics industry. Decomposition occurs via two pathways, 1(a) to give CF2 and CF3C(O)F and 1(b) to give CF2O and CF3CF. Rate coefficients calculated for reaction 1(a) are reported over the temperature range 300-1500 K and compared with the small number of available rate coefficients at selected temperatures within this range. Rate coefficients are reported for the first time for reaction 1(b) over the temperature range 300-1500 K. The geometries of the stationary points on the potential energy surfaces have been obtained at the MP2/6-311++G(d,p) level, and the energies of selected points along the minimum energy path (MEP) have been improved at the RCCSD(T)/AVDZ and RCCSD(T)/AVTZ levels. These improved energies were extrapolated to the complete basis set limit to obtain RCCSD(T)/CBS//MP2/6- 311++G(d,p) energies at each selected point on the MEP. The energies were then used in a dual-level direct dynamics method to calculate rate coefficients of the two decomposition reactions. The variational effect on the rate coefficients obtained is found to be small over the whole temperature range and tunnelling plays a small but significant role only at the lower temperatures. Comparison has been made of the computed reaction enthalpies, forward activation energies and rate coefficients computed at the RCCSD(T)/CBS//MP2/6-311++G(d,p) level with those computed with a number of different functionals and with the MP2 method. Reaction 1(a) is found to be the dominant reaction throughout the temperature range considered. Calculated rate coefficients for reaction 1(a) at the highest level used (improved canonical variational theory (ICVT) with small curvature tunnelling (SCT) with the RCCSD(T)/CBS//MP2/6-311++G(d,p) MEP) show reasonably good agreement with two recent sets of experimental values, although agreement with an older set is poor. This comparison highlights the need for more experimental rate coefficients for this thermolysis reaction over the whole temperature range considered, but particularly in the ranges 550-800 K and 1200-1500 K not currently covered by experimental measurements. ; Department of Applied Biology and Chemical Technology