Analyticity of the generalized susceptibility in the linear response theory

Question

In linear response theory, the generalized susceptibility $\chi(\omega)$ is defined as $$\chi(\omega)=\int\limits_{0}^{\infty}\phi(t) e^{i\omega t} dt, ~~t\geq 0\tag{1}$$ where $\phi(t)$ is the response function$^1$. If it is assumed that $\chi(\omega)$ exists for all real, non-negative $\omega$, then its integral representation as given in (1) suggests that $\chi(\omega)$ also exists when ${\rm Im}\omega\geq 0$ i.e., in the complex upper half plane of $\omega$. This is because an additional damping factor increases the convergence of the integral (1). From (1), how can one argue that $\chi(\omega)$ is also analytic in the complex upper half plane of $\omega$? It's crucial in deriving the so-called Kramer's-Kronig relations in physics.

---

$^1$ On physical grounds, $\phi(t)$ is smooth (which @AFT enquired in his comment) and is also bounded as $t\to\infty$.

Answer 1

I believe that this follows from causality, namely that $\phi(t)=0$ for $t<0$. Physically, the response of the system cannot precede the perturbation that causes it.

In your integral defining $\chi(\omega)$, the lower limit of integration may therefore be extended to $-\infty$, making $\phi(t)$ and $\chi(\omega)$ a Fourier transform pair. Then the Titchmarsh theorem applies: $\chi(\omega)$ is analytic in the upper half complex plane of $\omega$.