Analytic Continuation
Analytic Continuation
Oftentimes, we are presented with a definition of a function that is only valid in some
restricted domain. A basic example is the sum
∞
∑
zn (1)
n=0
which only makes sense (i.e. is finite) for |z| < 1. We want to somehow extend this function
outside this domain, preferably over the entire complex plane. The first question one should
ask is whether this is a unique procedure. The answer is essentially yes: there is a unique
function f (z) that is analytic (up to possible branch cuts and isolated singularities) that
agrees with (1) in the domain where (1) is defined. So if we can find an example of such
an f (z), we can say that we have “analytically continued” (1) to the full complex plane.
In the case of (1) the situation is very simple. We recognize the sum as giving 1/(1−z),
and therefore f (z) = 1/(1 − z) is the analytic continuation. Indeed it is analytic everwhere
except for a simple pole at z = 1.
Another famous example is the Gamma function, originally defined as
∫ ∞
Γ(z) = e−t tz−1 dt (2)
0
which only makes sense for Re(z) > 0. To make sense of Γ(z) for Re(z) < 0 we need to
analytically continue the expression (2). In this case the easiest way to proceed is to use
(2) to show that Γ(z) obeys a functional relation, and then apply this relation to determine
Γ(z) in the larger domain. The analytic relation here is
Although we use (2) to derive this, and (2) is only valid for Re(z) > 0, the key point is
that we should take (3) to be valid throughout the complex plane. A relation like (3) that
is valid in one region cannot suddenly cease to be true in another region, assuming that
the function is smoothly extended. Given (3) we can now compute in the larger domain.
E.g. we have Γ(−1/2) = −2Γ(1/2). More generally, using (3) as many times as necessary
we can always relate a given z in the left half-plane to one in the right half plane, and so
evaluate Γ(z) for all z. That is to say, we have succeeded in defining Γ(z) on the full plane.
Via an analytic continuation interpretation, one can make sense of certain otherwise
bizarre looking formulas. For example, “everyone knows” that
1
1 + 1 + 1 + 1 + ··· = − (4)
2
The reason this is “true” is as follows. The Zeta-function is defined as
∑∞
1
ζ(s) = (5)
n=1
ns
1
The sum converges for Re(s) > 1. Also, the sum in (4) is seen to be ζ(0). So if we can
analytically continue (5) to Re(s) < 1, then we can make sense of (4). One can derive the
following functional relation
( πs )
ζ(s) = 2s π s−1 sin Γ(1 − s)ζ(1 − s) (6)
2
Using this one finds
( πs ) ( ) ( )
1 1 πs π 3 s3 1 1
ζ(0) = lim sin ζ(1 − s) = lim − + ... − + ... = − (7)
π s→0 2 π s→0 2 48 s 2