I have a function defined as:


Are there any $z ∈ C$ (with nonzero imaginary part) such that $f(z)∈R$?

I tried substituting in $z=a+bi$ with $b≠0$ into the product and integral definitions of the Gamma function, but with little effect.




Introduce the variable $R$, denoting a real number. Then let $R=f(a+bi)$:


By definition $\Im(R)=0$. Re-arranging we get:


Because $R$ can be any real number, the 'real' part of the equation may always be true for some R. Therefore, equating the $\Im$ parts of each side of the equation, we get:


Because $a,b$ are strictly real numbers, $a^2+b^2$ must be real and the same on both sides.





It follows that:


Adding 1 to both sides:



EDIT II It turns out I may have proceeded incorrectly in the last step. You see, the last equation re-arranges to:

$\frac{z}{\Re(z)} = \frac{\Gamma(z)+1}{1+\Re(\Gamma(z))}$

And consequently:

$\frac{1+\Re(\Gamma(a+bi))}{\Re(z)} = \frac{\Gamma(z)+1}{z}$, which is the normal, real number solution. (i.e. $b=0$).

If we go back a few steps and instead eliminate the real part, we get the following:



Now, adding 1, we get:



Therefore, if $z=a+bi$ with $b≠0$ then the only time when $f(z)$ can be real is when:



The derivative of $f$ vanishes at many points on the real axis (at least once between every consecutive negative integers, just by looking at the limits there)

On some small enough complex neghbourhoods of those points, $f$ is then at least $2$-to-$1$, which means that there are complex inputs to $f$ that give real outputs.

  • $\begingroup$ Okay that makes sense. Thanks. I am hoping for a more general solution, if it exists at all. I'll edit the question with more of the work that I have done since. $\endgroup$ – KR136 Feb 7 '16 at 18:39
  • $\begingroup$ @mercio: I've never seen this argument before, and I really like it :D $\endgroup$ – Eric Stucky Feb 7 '16 at 18:57

I will provide a heuristic method to conclude that yes, such values exist, using a domain coloring visual argument. A domain coloring applies a color to each point in the complex plane based solely on its polar angle --- i.e. write $a+bi=Re^{i\theta}$ and we assign the colors based on $\theta$ (the $R$ decides brightness, typically).

If I apply the (Python, mpmath) domain coloring algorithm to the identity function $f(z)=z$ I get the following image:

domain coloring of the identity function.

The important thing about this image is that you notice the positive $x$-axis is colored red --- clockwise from red we move towards yellow and counterclockwise from red we get magenta. If we were to continuously go from magenta towards yellow in a clockwise direction, we must pass through the reddish area (the center of which colors the positive real numbers).

Now I will apply a domain coloring to your function $f(z)=\dfrac{\Gamma(z)+1}{z}$:

enter image description here

Notice in this image there are many "fingerlike" structures on the right. As Joanpemo noted, many values where $z \in \mathbb{R}$ yield real values (and hence are colored red in the image). Where else could $f$ be real?

Since $f$ is the sum and quotient of functions continuous on $\mathbb{C} \setminus \{0,-1,-2,\ldots\}$, we know that the function $f$ is continuous near the fingerlike structures and our image shows magenta --> red --> yellow color transitions (i.e. argument transitions from negative to positive) of $f(z)$ near the fingerlike structures.

It also appears that there is a sequence of strange areas going left along the $x$-axis near each nonpositive integer. These structures come from the fact that $\Gamma$ has poles at such points. Here's some zoomed in images of those places:

zoomed in near -1 zoomed in near -2 zoomed in near -4 zoomed in near -5 zoomed in near -6

To make these images just change XMIN, XMAX, et al to desired values:

import numpy as np
import matplotlib.pyplot as plt
from mpmath import *

cplot(lambda x: (gamma(x)+1)/x,[XMIN,XMAX],[YMIN,YMAX],points=50000,verbose=True)
  • $\begingroup$ Okay, that makes sense, logically, and with @mercio's reasoning makes a good case for the existence of such points. The next question, naturally, is what the pattern for the real outputs is, and if there is a regularity to it that which can be modeled algebraically. (Beyond just the negative integers, there also appear to be reddish areas for Re(z)>0, the "fingerlike" structures in your coloring chart). $\endgroup$ – KR136 Feb 7 '16 at 19:39

$$\forall\,n\in\Bbb N\;,\;\;\Gamma(n)=(n-1)!\implies f(n)=\frac{(n-1)!+1}n\in\Bbb Q\subset\Bbb R$$

  • $\begingroup$ Right. I will amend my question. My original intention was that $z=a+bi$ with $b≠0$. $\endgroup$ – KR136 Feb 7 '16 at 17:57
  • $\begingroup$ Well, then that seems to be a very tough problem, since $\;\frac{a+bi}{c+di}\in\Bbb R\iff ad-bc=0\;$ . $\endgroup$ – DonAntonio Feb 7 '16 at 18:03

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.