Paper detail

Polynomially scaling spin dynamics simulation algorithm based on adaptive state space restriction

The conventional spin dynamics simulations are performed in direct products of state spaces of individual spins. In a general system of n spins, the total number of elements in the state basis is >4^n. A system propagation step requires an action by an operator on the state vector and thus requires >4^2n multiplications. It is obvious that with current computers there is no way beyond about ten spins, and the calculation complexity scales exponentially with the spin system size. We demonstrate that a polynomially scaling algorithm can be obtained if the state space is reduced by neglecting unimportant or unpopulated spin states. The class of such states is surprisingly wide. In particular, there are indications that very high multi-spin orders can be dropped completely, as can all the orders linking the spins that are remote on the interaction graph. The computational cost of the propagation step for a ktuples-restricted densely connected n-spin system with k<<n is O(n^2k). In cases of favourable interaction topologies (narrow graphs, e.g. in protein NMR) the asymptotic scaling is linear.