Given a Laurent series $$f(z) = \sum_{k=-\infty}^\infty a_k(z-z_0)^k,$$ its derivative is obviously also a Laurent series. However, upon integration, the $\frac1{z-z_0}$ term introduces a $\ln(z-z_0)$ term and therewith an essential singularity turning the integral of a Laurent series into something that generally is not a Laurent series again. So one could define a new kind of series including that $\ln$ term, and upon integration obtain another new term due to $\int\ln z\,dz = z(\ln z-1)$, or more generally, $$\int z^k\ln z\,dz = \frac{z^{k+1}[(k+1)\ln z -1]}{(k+1)^2}$$ for $k\neq-1$.

So the series $$f(z) := \sum_{k=-\infty}^\infty [a_k+b_k\ln(z-z_0)](z-z_0)^k$$ with $b_{-1}=0$ is complete under both derivation and integration at the expense of introducing a branch cut and an essential singularity at $z_0$. But does such a series definition make sense? How would one obtain the coefficients $b_k$ now?

For $b_{-1}\neq0$, one obtains $\int\frac{\ln z}z\,dz = \frac12\ln^2 z$, the inclusion of which requires the inclusion of $\int\ln^2z\,dz = z(\ln^2z -2\ln z+2)$ and therewith of $z^m\ln^2z$ for $m\neq-1$. And should the extended series include a $\frac{\ln^2z}z$ term, its integration would require a $\ln^3z$ term and so on, so in an even more general sense we'd have $$f(z) := \sum_{k=-\infty}^\infty \sum_{l=0}^\infty c_{kl} \ln^l(z-z_0)\cdot (z-z_0)^k.$$ Now you could go on and include $l<0$ terms, but their integrals would require the inclusion of the logarithmic integral $\operatorname{li}(z-z_0)$, and integrating that would include $\operatorname{Ei}(z-z_0)$ and hypergeometric functions etc. - so, does this ever stop at a (up to powers) finite set of base functions?

  • 2
    $\begingroup$ For $k = -1$ the integral $\int \frac{\log z}z\,dz = \int z^k\,\log z\, dz$ is not elementary. Same problem as before ... $\endgroup$ – martini Sep 30 '13 at 8:50
  • 2
    $\begingroup$ @martini You're right, which leads to including $\ln^2 z$, which leads to including $\int\ln^2z/z\propto\ln^3z$ and so on, so all powers of $\ln$ would need to be included as well... So the $[a_k+b_k\ln(z-z_0)]$ actually becomes a power series of $\ln(z-z_0)$, i.e. $$f(z) := \sum_{k=-\infty}^\infty \sum_{l=0}^\infty c_{kl}\ln^l(z-z_0)\cdot(z-z_0)^k$$ $\endgroup$ – Tobias Kienzler Sep 30 '13 at 8:59

Perhaps consult the literature on "transseries" ...

(plug) G. Edgar, Transseries for beginners, Real Analysis Exchange 35 (2010) 253--310

The set of transseries includes (one-sided) Laurent series, and is closed under many operations, including integration.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.