Erratum: Chiral state conversion without encircling an exceptional point [Phys. Rev. A 96 , 052129 (2017)]

The numerical precision of the iterative algorithm (Runge-Kutta) was identified as insufficient when obtaining the results that were presented in Figs.1(a)-1(c) and the curves for γ = 0.008 in Figs.2(a)-2(c) of the original article.When we increased the precision to account for high levels of field amplification, no mode conversion was observed when the parameter loop was far away from the exceptional point (EP), which was mistakenly shown to be true as in Fig. 1(c) of the main text.In fact, for the parameter values listed there, the modal output should be |ψ 1 → |ψ 1 and |ψ 2 → |ψ 2 (corresponding to no mode conversion and no chirality).This numerical error leaves the main message of our paper intact, which states that chiral mode conversion can also take place even without enclosing or encircling an EP in a dynamic contour.The results based on the increased numerical precision show, however, that chiral mode conversion requires the contour to be at least in the direct vicinity of the non-Hermitian degeneracy.To illustrate this, we consider a new loop that now lies closer to the EP and we compare the state conversion for two cycling rates γ = 0.5 and γ = 0.1.Specifically, in the upper panel of Fig. 1 presented below, the loop is centered at (g 0 ,ρ) = (0.82,0.1) which excludes the EP.For the relatively nonadiabatic case, i.e., γ = 0.5, no mode conversion was observed.Conversely, upon making the encirclement adiabatic, i.e., γ = 0.1, both |ψ 1 and |ψ 2 (or any other linear combination of the two) end up in the same state |ψ 1 , thus corroborating the findings of our analytical results.For the purpose of comparison, in this corrected Fig. 1, we also show the EP-inclusive case that depicts mode conversion both for small and fast rates of encirclement.Figure 2 of the original text should also be considered in the light of the aforementioned changes.Those results hold true only in the cases where