To gain insight into the mechanisms and kinetics of 2-azido-N,N-dimethylethanamine's (DMAZ's) thermal decomposition postulated reaction paths were simulated with ab initio and density functional theory quantum chemistry models. Four reaction types were modeled: (i) spin-allowed and spin-forbidden paths involving N-N-2 bond fission and nitrene formation, (ii) HN3 elimination with the formation of (dimethylamino)ethylene, (iii) N-N-2 bond fission with the formation of molecules with three- or four-membered heterocydic rings, and (iv) simple scission of C-H, C-N, and C-C bonds. The geometries of stationary points of the reactions were obtained with a MPWB1K/6-31+G(d,p) model. To locate and model the geometries of minimum energy intersystem crossing points for triplet nitrene formation and isomerization, unrestricted broken spin symmetry calculations were performed. Employed to model an analogous path for methyl azide's decomposition, this approach was found to yield results similar to those obtained with a CASSCF(10,8)/aug-cc-pVDZ model. Of the four reaction types studied, N-N-2 bond fissions with singlet or triplet nitrene formation were found to have the lowest barriers. Barriers for paths to cyclic products were found to be 2-4 kcal/mol higher. Kinetic rate expressions for individual paths were derived from the quantum chemistry results, and spin-allowed nitrene formation was found to be dominant at all temperatures and pressures examined. The expression 2.69 X 10(9) (s(-1))T-1.405 exp(-39.0 (kcal/mol)/RT), which was derived from QCISD(T)/6-31++G(3df,2p)//MPWB1K/6-31+G(d,p) results, was found to be representative of this reaction's gas-phase rate. Adjusted on the basis of results from self-consistent reaction field models to account for solvation by n-dodecane, the expression became 1.11 x 10(9) (s(-1))T-1.480 exp(-37.6 (kcal/mol)/RT). Utilizing this result and others derived in the study, a model of the decomposition of n-dodecane-solvated DMAZ was constructed, and it generated simulations that well-reproduce previously published measured data for the process.