||The lowest singlet and triplet potential energy surfaces of N2O and their intersection are studied using various ab initio methods including MP2, QCISD(T), CCSD(T), CASSCF and multireference configuration interaction method (MRCI). On the singlet surface, two local minima are found, linear NNO (A) and cyclic structure D. The latter lies ∼64 Full-size image (<1 K) higher in energy and is separated from the former by a barrier of about 15 Full-size image (<1 K) at the QCISD(T)/6-311+G(3df)//MP2/6-311+G(2d) and CCSD(T)/6-311+G(3df)//B3LYP/6-311G(d) levels. Both A and D can be produced from N2+O(1D) without barrier with exothermicity of 88.3 and 23.9 Full-size image (<1 K), respectively, at QCISD(T)/6-311+G(3df)//MP2/6-311+G(2d). On the triplet surface, no stable bound N2O structure exists although some plateau on the surface is found in the vicinity of the bent structure B, 73–77 Full-size image (<1 K) above linear A at the QCISD(T) and CCSD(T) levels. Singlet–triplet intersections are located both at the bent geometry (B1) with ∠NNO=114∘ and at the linear structure C. The computed energy of C, 60.3 Full-size image (<1 K) at the MRCI(10,9)/6-311+G(3df) level, closely agree with the experimental activation energy for N2O decomposition. C is minimum on the seam of crossing and has higher spin–orbit coupling than those for bent intersection structures. Thus, the spin-forbidden fragmentation N2O(1Σ+)→N2(1Σg+)+O(3P) should occur via structure C as a “transition state”. The calculations demonstrated that the use of QCISD(T), CCSD(T), full-valence active space CASSCF, or MRCI theoretical levels is essential to compute accurate relative energies of B1 and C.