As pointed in Does the analytic continuation always exists? we know it doesn't always exist.

But: take the $\Gamma$ function: the first definition everyone meet is the integral one: $$ z\mapsto\int_{0}^{+\infty}t^{z-1}e^{-t}\,dt $$ which defines an holomorphic function on the half plane $\{\Re z>0\}$. Moreover we immediately get the functional equation: $$ \Gamma(z+1)=z\Gamma(z)\;,\;\;\;\forall\; \Re z>0. $$ This equation is used to extend the function on the whole complex plane (minus the negative integers)... but: WHY CAN WE DO THIS?!

We know that there is an holomorphic function $\Gamma$ which can be expressed as the integral on that half plane. Why are we allowed to write $$ \Gamma\left(\frac12\right)=-\frac12\Gamma\left(-\frac12\right) $$ for example? LHS is defined, RHS, NOT!!! But where's the problem? Simply let's define $\Gamma\left(-\frac12\right)$ in such a way... but why can we do this? How can I know that this function I named $\Gamma$ which is holomorphic on the above half plane admits an extension?

  • $\begingroup$ You are just using the same name for two differenti objects. $\endgroup$ – N74 Apr 13 '16 at 20:11

Once you have the functional equation for $\Gamma$, we can define a new function $\tilde \Gamma$ defined on the half-plane $\operatorname{Re} z > -1$ (except $z=0$) by $$ \tilde \Gamma(z) = \frac1z \Gamma(z+1). $$ It's clear that $\tilde \Gamma$ is holomorphic on $\{ \operatorname{Re} z > -1 \} \setminus \{ 0 \}$, and coincides with $\Gamma$ on $\operatorname{Re} z > 0$ (because of the functional equation). In other words, $\tilde \Gamma$ is an analytic continuation of $\Gamma$ to $\{ \operatorname{Re} z > -1 \} \setminus \{ 0 \}$. So we may as well call $\tilde \Gamma$ by $\Gamma$.

Repeating the above construction, we can define a "new $\Gamma$-function" on successively larger sets until we get something defined and holomorphic on the whole complex plane except the non-positive integers.

| cite | improve this answer | |
  • 1
    $\begingroup$ I don't agree. you are totally obfuscating the process of analytic continuation this way : what about the analytic continuation on a path of intersecting disks ? how do you know that the analytic continuation doesn't depend on the path (the fact that $\Gamma(s)$ is meromorphic ?) $\endgroup$ – reuns Apr 13 '16 at 21:47

The following are citations from the classic Applied and Computational Complex Analysis, Vol. I by P. Henrici. The chapter 3: Analytic Continuation provides a thorough treatise of the theme. Here we look at two aspects, which should help to clarify the situation.

At first we take a look when two functions $f(z)$ and $g(z)$ are analytic continuations from each other.

Theorem 3.2d: (Fundamental lemma on analytic continuation)

Let $Q$ be a set with point of accumulation $q$ and let $R$ and $S$ be two regions such that their intersection contains $Q$ and $q$ is connected. If $f$ is analytic on $R$, $g$ is analytic on $S$, and $f(z)=g(z)$ for $z\in Q$, then $f(z)=g(z)$ throughout $R\cap S$ and $f$ and $g$ are analytic continuations of each other.

We observe, we need at least a set $Q$ with an accumulation point where analytic functions $f$ and $g$ have to coincide. This set is part of the intersection of two regions $R$ and $S$ where $f$ and $g$ are defined. Finally we conclude that throughout $R\cap S$ the functions coincide.

The second aspect sheds some light at functional relationships in connection with analytic continuation. We can read in

Section 3.2.5: Analytic Continuation by Exploiting Functional Relationships

Occasionally an analytic continuation of a function $f$ can be obtained by making use of a special functional relationship satisfied by $f$. Naturally this method is restricted to those functions for which such relationships are known.

He continues with example 15 which seems that P. Henrici had precisely a user with OPs question in mind.

Example 15:

Let the function $g$ possess the following properties:

  • (a) $g$ is analytic in the right half-plane: $R:\Re (z)>0$

  • (b) For all $z\in R, zg(z)=g(z+1)$

We assert that $g$ can be continued analytically into the whole complex plane with the exception of the points $z=0,-1,-2,\ldots$.

We first continue $g$ into $S:\Re (z)>-1,z\neq 0$. For $z\in S$,let $f$ be defined by \begin{align*} f(z):=\frac{1}{z}g(z+1) \end{align*} For $z\in S, \Re(z+1)>0$. Hence by virtue of (a) $f$ is analytic on $S$. In view of (b) $f$ agrees with $g$ on the set of $R$. Since $S$ is a region, $f$ represents the analytic continuation of $g$ from $R$ to $S$. We note that $f$ satisfies the functional relation $f(z+1)=zf(z)$ on the whole set $S$.

Denoting the extended function again by $g$, we may use the same method to continue $g$ analytically into the set $\Re(z)>-2,z\neq 0,-1$, and thus step by step into the region $z\neq 0,-1,-2,\ldots$.

Of course this example addresses the Gamma Function $\Gamma(z)$ which is treated in detail in chapter 8, vol. 2. He then continues with further methods of analytic continuation, such as the principle of continuous continuation and the symmetry principle.

| cite | improve this answer | |

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.