The gamma function has a long and interesting history (check out (Davis 1963) excellent review), but one application does not seem to have shown up: minimal surfaces.

A minimal surface is one where the average curvature is always zero; it bends equally in two opposite directions. This is equivalent to having the (locally) minimal area given its boundary: such surfaces are commonly seen as soap films stretched from frames. There exists a rich theory for them, linking them to complex analysis through the Enneper-Weierstrass representation: if you have a meromorphic function g and an analytic function f such that  is holomorphic, then

produces a minimal surface .

When plugging in the hyperbolic tangent as g and using f=1 I got a new and rather nifty surface a few years back. What about plugging in the gamma function? Let .

We integrate from the regular point to different points in the complex plane. Let us start with the simple case of .

The surface is a billowing strip, and as we include z with larger and larger real parts the amplitude of the oscillations grow rapidly, making it self-intersect. The behaviour is somewhat similar to the Catalan minimal surface, except that we only get one period. If we go to larger imaginary parts the surface approaches a horizontal plane. OK, the surface is a plane with some wild waves, right?

Not so fast, we have not looked at the mess for Re(z)<0. First, let’s examine the area around the z=0 singularity. Since the values of the integrand blows up close to it, they produce a surface expanding towards infinity – very similar to a catenoid. Indeed, catenoid ends tend to show up where there are poles. But this one doesn’t close exactly: for re(z)<0 there is some overshoot producing a self-intersecting plane-like strip.

The problem is of course the singularity: when integrating in the complex plane we need to avoid them, and depending on the direction we go around them we can get a complex phase that gives us an entirely different value of the function. In this case the branch cut corresponds to the real line: integrating clockwise or counter-clockwise around z=0 to the same z gives different values. In fact, a clockwise turn adds [3.6268i, 3.6268, 6.2832i] (which looks like – a rather neat residue!) to the coordinates: a translation in the positive y-direction. If we extend the surface by going an extra turn clockwise or counterclockwise a number of times, we get copies that attach seamlessly.

 

OK, we have a surface with some planar strips that turn wobbly and self-intersecting in the x-direction, with elliptic catenoid ends repeating along the y-direction due to the z=0 singularity. Going down the negative x-direction things look plane between the catenoids… except of course for the catenoids due to all the other singularities for . They also introduce residues along the y-direction, but different ones from the z=0 – their extensions of the surface will be out of phase with each other, making the fully extended surface fantastically self-intersecting and confusing.

So, I think we have a simple answer to why the gamma function minimal surface is not well known: it is simply too messy and self-intersecting.

Of course, there may be related nifty surfaces. is nicely behaved and looks very much like the Enneper surface near zero, with “wings” that oscillate ever more wildly as we move towards the negative reals. No doubt there are other beautiful things to look for in the vicinity.