Dynamical-mean-field cluster-correlation expansion of NV$^-$ spin-echo decoherence in a realistic substitutional-nitrogen bath

Abstract

The Hahn-echo time $T_2$ of nitrogen-vacancy (NV$^-$) ensembles in nitrogen-rich diamond follows the empirical relation $T_2\cdot$[N] = 160 $μ$s ppm, independent of field between 2 and 30 mT. Cluster-correlation-expansion (CCE) calculations that treat the Overhauser field of the surrounding P1 centres as a frozen, sampled mean field overestimate it at high field by up to a factor of two, an excess recently attributed to additional dark defects. We show that the excess is largely an artefact of freezing the field. The P1 centres outside each cluster flip-flop in pairs that conserve their total $S_z$, generated by kinetic Monte Carlo at golden-rule rates from the dipolar Hamiltonian without adjustable parameters, and each cluster is propagated in the resulting time-dependent field. This dynamical mean field interpolates between the static-mean-field and mean-field-free limits. For a $^{14}$N P1 bath it reduces $T_2\cdot$[N] at 30 mT from 264 $\pm$ 16 to 188 $\pm$ 12 $μ$s ppm and gives 167 $\pm$ 11 $μ$s ppm at 10 mT, 1.7 and 0.4 standard deviations from experiment, with no dark spins. Letting each centre jump independently gives 147 $\pm$ 9 $μ$s ppm at 30 mT; the correlation of the flip-flop partners lengthens $T_2$ by 20-30 percent at high field. The Overhauser field is neither frozen nor motionally narrowed: 68-83 percent of its autocorrelation survives 200 $μ$s. At 2.5 mT, where the P1 electron and nucleus hybridise, $T_2$ remains below experiment for any choice of rates, so part of the field dependence remains ($T_2$ rises by a factor 1.69 $\pm$ 0.07 instead of 2.15 $\pm$ 0.08 from 2.5 to 30 mT).