Systematic derivation of the conjugation relation between saddle-point parameters
Derive, in a rigorous and self-contained manner, the conjugation relation \(\tilde u = -\bar u\) for the saddle-point parameters in the large-\(k\) analysis of sparse non-Hermitian random matrices, starting from the general operator identity \(\tilde g = \operatorname{sign}(\hat C)\,\bar g\) where \(\hat C\) is defined by the kernel \(C(X,X') = k\,[1-\hat p\big(i\tilde X^{\dagger}X - iX^{\dagger}\tilde X\big)]\). Establish the conditions under which this relation holds and justify its use in obtaining the circular law and spectral-edge shapes for the ensembles considered.
Sponsor
References
We leave systematic derivation of this formula to future work, and we shall see below that it does not only allow us to rederive the circular law, but also much more sophisticated shapes of spectral edges for more complicated random matrix ensembles in the next section.