Probing automorphic induction through zero correlations and secondary arithmetic terms

Abstract

Zero correlations offer a way to test the cuspidal structure of an automorphic L-function, but their leading self statistics retain only coarse degree and multiplicity data. We therefore adjoin a specified cuspidal L-function as a probe and study the resulting raw mixed correlations. Assuming a factorization into cuspidal L-functions over Q, we show that a mixed pair statistic detects the probe's occurrence and recovers its multiplicity. A suitable Gevrey-order-two test eliminates the smooth-density terms and gives a subexponential error. For arbitrary smooth plateau tests, the prime number theorem of Harcos and Thorner gives an error smaller than every fixed inverse power of log T, without a self-contragredience assumption. The stronger prime number theorem of Liu and Ye gives a subexponential remainder under its stated structural assumptions, with the possible norm-twist terms made explicit. To obtain arithmetic information beyond this overlap, we compute the correlation expansion at every fixed level and inverse-logarithmic order within the original strict small-support range. We treat the pair and higher correlations separately, using Jiang's rational-field fixed-data estimates to justify the required prime-power convergence. The pair calculation identifies the mixed arithmetic finite parts and gives detailed error estimates, including the contribution of norm twists. The higher-level calculation retains connected prime-power invariants, whose first cubic contribution appears at order (log T)^{-3}. Explicit third- and fourth-level formulas and examples explain the arithmetic content of these coefficients and the limitations of self-correlations. Varying powers of the probe separates their mixed components. These formulas, together with the distinct-index sieve, provide a collection of arithmetic probe responses toward the further problem of recovering automorphic induction structure.