A product of two integrals
The gamma recurrence continues meromorphically from , but it does not on its own exclude zeros. A second identity relates values at and .
Theorem 1.14.1
(Euler's reflection formula)
As meromorphic functions on ,
In particular the two sides are finite and equal at every .
We will first establish it in the strip , then explain its meromorphic meaning. The proof uses two results developed separately: gamma's integral and recurrence, and the keyhole-contour evaluation
From gamma to a beta-type integral
Proof
(Real exponents in )
Fix real . Since the integrand is nonnegative, Tonelli's theorem permits multiplication of the two gamma integrals:
Make the change of variables and on the positive quadrant. Its inverse is , ; this is a diffeomorphism from onto the positive quadrant, with Jacobian . Since , the powers of cancel, so
The first integral is one. With , so that , and , the second becomes . Equation (2) proves (1) for real .
Proof X-ray
Where the cancellation comes from
Each hypothesis has one job. Reality of makes the double integrand nonnegative, which is what licenses Tonelli; complex exponents are reached later by the identity theorem, not by this computation. The range is needed three times: for , for , and for convergence of (2) at both ends.
The key move is the substitution , . The exponents of and add to , so the integrand carries exactly , and the Jacobian cancels it. The double integral then splits into times a one-variable integral in which all dependence on sits. The computation closes by quoting the contour evaluation (2); the gamma side contributes nothing beyond positivity and the product structure.
Extending the identity
Proof
(Continuation by the identity theorem)
On the connected strip , both gamma factors are holomorphic and has no zero. The two sides of (1) agree on the real interval , which has accumulation points inside the strip. The identity theorem therefore proves (1) throughout the strip.
The gamma recurrence gives a meromorphic continuation to , with simple poles only at nonpositive integers; the right side is meromorphic there. Composition with preserves meromorphicity. Since these meromorphic functions agree on the strip, the identity theorem on the connected complement of their discrete poles extends their equality throughout the plane as an equality of meromorphic functions.
At an integer, one reads that statement through Laurent expansions or limits; the displayed quotient is not a finite value there.
Common mistake
(Do not evaluate the identity at a pole)
At both and have poles, and substituting into either side as though both were finite numbers is meaningless. The identity asserts that the Laurent expansions agree, principal parts included. Here and , so the left side is , while : the principal parts are both .
Half-integers and the absence of zeros
Corollary 1.14.2
(Gamma has no zeros)
The meromorphic function has no zeros: at every that is not a pole, that is, at every .
Proof
(Two finite factors with a nonzero product)
Take a noninteger . Both and are finite, while the right side of (1) is nonzero. Neither factor can therefore be zero. At positive integers the recurrence gives ; nonpositive integers are poles rather than zeros.
Consequently extends to an entire function whose zeros are exactly the nonpositive integers, all simple.
Worked example 1.14.3
(Half-integer values)
For real , the integral defining is strictly positive. At , (1) gives , and positivity selects the positive root. The recurrence, read as for negative arguments, then gives
The signs alternate between consecutive poles, as nonvanishing and continuity on each interval require.
Worked example 1.14.4
(Residues of gamma at the poles)
Fix an integer . Near , the factor is holomorphic with value . Since ,
Dividing by gives
The edge case gives residue , matching with .
Worked example 1.14.5
(The modulus of gamma on the critical line)
For real , put . Then . For , conjugating the integral gives . Since , formula (1) becomes
At this recovers ; as the modulus decays like .
Practice with solutions
Quick check
(Compute from .)
Solution · Answer
The recurrence gives .
Quick check
(What does (1) give for ?)
Solution · Answer
It is .
Quick check
(Why can the identity theorem be used after proving (1) only for real ?)
Solution · Answer
Both sides are holomorphic on a connected strip and agree on a set with an accumulation point there.
Quick check
(Why does the nonvanishing argument treat positive integers separately?)
Solution · Answer
At a positive integer, has a pole, so the noninteger argument using two finite factors does not apply; the factorial formula gives a nonzero value directly.