Consider the function $f$ defined over $\mathbb{R}$ as $$f(x) = \coth x - \frac{1}{x}$$ if $x \neq 0$ and $f(0)=0$.
Since the function $\coth$ can be developed in series as $\coth x = \frac{1}{x} + \frac{x}{3} - \frac{x^3}{45} + \cdots$ around the origin, the function $f$ can easily be shown to be infinitely smooth over $\mathbb{R}$, strictly increasing, with limits $\pm 1$ in $\pm \infty$. The function $f$ is therefore a smooth bijection from $\mathbb{R} \rightarrow (-1,1)$.
My question is the following: Is there an analytic expression for the inverse $f^{-1} : (-1,1) \rightarrow \mathbb{R}$ of $f$ such that $f^{-1}(y) = x$ if and only if $f(x) = y$?
Context: The function $f$ naturally appears in an optimization problem I am considering. It would be useful for me to understand and possibly compute what the inverse is.
The “Langevin function” is invertible via Lagrange reversion:
$$L(x)=\coth(x)-\frac1x=a\mathop\iff^{x=\frac{y-1}{a+1}} y=e^\frac{2(y-1)}{a+1} \left(\frac{a+1}{a-1}(y-1)+1\right)\implies x=\frac1{a+1}\left(-1+\sum_{n=1}^\infty\frac{e^{-\frac{2n}{a+1}}}{n!}\left.\frac{d^{n-1}}{dt^{n-1}}e^\frac{2nt}{a+1}\left(\frac{a-1}{a+1}(t-1)+1\right)^n\right|_0\right)$$
Using general Leibniz rule:
$$\left.\frac{d^{n-1}}{dt^{n-1}}e^\frac{2nt}{a+1}\left(\frac{a-1}{a+1}(t-1)+1\right)^n\right|_0=\sum_{k=0}^{n-1}\binom{n-1}k\left.\frac{d^{n-1}}{dt^{n-1}}e^\frac{2n t}{a+1}\right|_0 \left.\frac{d^{n-1}}{dt^{n-1}} \left(\frac{a-1}{a+1}(t-1)+1\right)^n\right|_0$$
and summing over $k$ uses hypergeometric $\operatorname U(a,b,x)$. Therefore:
$$\bbox[2px, border: 2px solid silver]{L^{-1}(x)=-\frac1{x+1}\left(1+2\sum_{n=1}^\infty\frac{(-1)^n(x-1)^{n-1}}{e^\frac{2n}{x+1}(x+1)^nn!}\operatorname U\left(1-n,2,-\frac{4n}{x^2-1}\right)\right)}$$
shown here. The series converges more quickly as $x\to-1$ and symmetry extends the convergence radius to $x>0$