Solving a linear PDE, I got the general solution $$f(x,t)=\frac{e^{-t/\tau}}x\sum^\infty_{n=1}a_n\sin(\gamma_n x)$$ where $\gamma_n$ is the nth positive root of $\tan x=x$.
To satisfy the initial condition we require $$\phi(x)=\frac1x\sum^\infty_{n=1}a_n\sin(\gamma_n x)$$
From a physical point of view, $\phi(x)=f(x,0)$ represents the initial temperature profile of the object, which can be arbitrarily defined. Hence, I assumed such a expansion form is possible for most sufficiently well-behaving $\phi(x)$, but I cannot prove it.
What are the necessary and sufficient conditions for expanding a function $\phi(x)$ into $\frac1x\sum^\infty_{n=1}a_n\sin(\gamma_n x)$?
Noteworthily, $\gamma_n\sim\frac{\pi}{2}+n\pi$ thus this series is ‘almost’ a Fourier sine series.
When this expansion turns out to be mathematically possible, how could one extract the $a_n$s? One possible way is performing fourier transform on both sides, so that $$\int^\infty_{-\infty}x\phi(x)e^{-i\omega x}dx = \sum^\infty_{n=1}a_n\delta(\omega-\gamma_n)$$ assuming $\omega>0$, but often does one have knowledge of $\phi(x)$ only on a physically meaningful region $[0,a]$.