A classic from 1956. Note the limiting assumptions of linearity and time independence of the input-output connection.

Phys. Rev. 104, 1760–1770 (1956)
Causality and the Dispersion Relation: Logical Foundations
Abstract
References
John S. Toll
Princeton University, Princeton, New Jersey and University of Maryland, College Park, Maryland
Received 31 January 1956; revised 24 August 1956; published in the issue dated December 1956

"Strict causality" is the assumption that no signal whatsoever can be transmitted over a space-like interval in space-time, or that no signal can travel faster than the velocity of light in vacuo. In this paper a rigorous proof is given of the logical equivalence of strict causality ("no output before the input") and the validity of a dispersion relation, e.g., the relation expressing the real part of a generalized scattering amplitude as an integral involving the imaginary part. This proof applies to a general linear system with a time-independent connection between the output and a freely variable input and has the advantage over previous work that no tacit assumptions are made about the analytic behavior or single-valuedness of the amplitude, but, on the contrary, strict causality is shown to imply that the generalized scattering amplitude is analytic in the upper half of the complex frequency plane. The dispersion relations are given first as a relation between the real and imaginary parts of the generalized scattering amplitude and then in terms of the complex phase shift.