I'm looking for alternate methods of proving the following result:
$$\int_{0}^{\infty} \psi^{(2)} (1+x) \ln (x) \, dx = \frac{\pi^2}{6} \left( \gamma + \ln (2\pi)-12 \ln A +1\right) = \zeta'(2) + \zeta(2)$$ Where $\psi$ is the polygamma function, $A$ is the Glaisher-Kinkelin constant and $\gamma$ is the Euler-Mascheroni constant.
Here is how I originally proved it:
Begin with the known result $$\int_{0}^{\infty} \frac{x \ln x}{e^{2 \pi x}-1} \, dx = \frac{1}{2} \zeta'(-1)=\frac{1}{24}-\frac{1}{2} \ln A$$
then use the following well-known property of the Laplace transform $$\int_{0}^{\infty} f(x) g(x) \, dx = \int_{0}^{\infty} \left( \mathcal{L} f\right) (y) \left( \mathcal{L}^{-1} g \right) (y) \, dy$$
Let $f(x) = \frac{x^2}{e^{2 \pi x} - 1}$ and $g(x) = \frac{\ln x}{x}$ then: $$\left(\mathcal{L} f\right) (y) = -\frac{1}{8 \pi^3} \psi^{(2)} \left(1+\frac{y}{2\pi}\right)$$ $$\left(\mathcal{L}^{-1} g\right)(y) = -\gamma - \ln (y)$$
$$\implies \frac{1}{24} - \frac{1}{2} \ln A = \frac{1}{8\pi^3}\int_{0}^{\infty} \gamma \, \psi^{(2)} \left(1+\frac{y}{2\pi}\right) \, dy + \frac{1}{8\pi^3} \int_{0}^{\infty} \psi^{(2)} \left(1+\frac{y}{2\pi}\right) \ln (y) \, dy$$
Proceed with the substitution $y \mapsto 2\pi x \implies dy = 2 \pi\,dx$ $$\implies \frac{1+\gamma}{24} - \frac{1}{2} \ln A = \frac{1}{4\pi^2} \int_{0}^{\infty} \psi^{(2)} \left(1+x\right) \ln (2 \pi x) \, dx$$ Now separate $\ln (2\pi x) = \ln (2 \pi) + \ln (x)$ and the result quickly follows.
I was wondering if there are any alternative methods that don't rely on the initial $\frac{1}{2} \zeta'(-1)$ result.
Motivation:
I've been experimenting with the following identity: $$\int_{0}^{\infty} \frac{x^{2n-1}}{e^{2\pi x}-1}\, dx = \frac{(-1)^{n-1} B_{2n}}{4n}\,\text{ for } n \in \mathbb{N}$$
I noticed that if one takes the partial derivative with respect to $n$ and then proceeds to take the limit as $n \to 1$, then we arrive at our $-\frac{1}{2} \zeta'(-1)$ result. This means it is possible for one to define the 'derivative' of the Bernoulli numbers at $n=1$.
More precisely: $$B_{2}^{\prime} = \frac{1}{4} - 2 \ln A$$ is this result already known, and if so, are any other Bernoulli derivatives definable in a similar way using the Glaisher-Kinkelin constant? (I know it is possible to do it with Glaisher-Kinkelin-like constants such as the Bendersky-Adamchik constants due to their similar connection to hyperfactorials, $\zeta'(-2)$ and $\zeta(3)$ for example). I'm not sure what meaning one can give to the 'derivative' of the Bernoulli numbers, but figure 2 of this paper seems to roughly agree with my result.