I have recently read about zeta function regularization, a way of ascribing values to functions having simple poles in a point and to divergent series. The values obtained are the same as those obtained using Ramanujan's summation.

In short this says that a function having pole in $x_0$ should be ascribed the following value:

$$f(x_0)=\lim_{h\to0}\frac{f(x_0+h)+f(x_0-h)}2=\lim_{x\to x_0} \left(\frac{f (x)^2 f''\left( x \right)}{2 f'\left(x\right)^2}-f (x)\right)$$ (see here for derivation).

This way we can obtain:

$$1/0=0$$ $$\zeta(1)=\gamma$$ $$\Gamma(0)=-\gamma$$ $$\Gamma(-1)=\gamma-1$$ $$\Gamma(-2)=\frac{3}{4}-\frac{\gamma }{2}$$ $$\Gamma(-3)=\frac{\gamma }{6}-\frac{11}{36}$$ $$\Gamma(-4)=\frac{25}{288}-\frac{\gamma }{24}$$ etc.

But as we can see, these values of Gamma function do not satisfy the functional equation. It seems like every value were missing a non-real term.

We can fix the issue by introducing a value $\omega$ with the following properties:

$$0\, \omega = 1$$ $$\omega+a \ne \omega$$ $$a \omega \ne \omega$$ $$\Re(\omega)=\Im(\omega)=0$$ $$|\omega|=\infty$$

This way the values of the Gamma function at negative arguments will satisfy the functional equation:

$$\Gamma(0)=-\gamma+\omega$$ $$\Gamma(-1)=\gamma-1-\omega$$ $$\Gamma(-2)=\frac{3}{4}-\frac{\gamma }{2}+\frac\omega 2$$ $$\Gamma(-3)=\frac{\gamma }{6}-\frac{11}{36}-\frac\omega 6$$


etc. As you can see, these values perfectly satisfy the functional equation for Gamma.

Thus I wonder

  • What algebraic properties of real and complex numbers get lost with introduction of such $\omega$?
  • What is the general way of finding the coefficient at $\omega$ for an arbitrary function having a pole (I want an expression in derivatives, similar to the expression for the real part).
  • $\begingroup$ Why the downvotes? $\endgroup$
    – Anixx
    Dec 8 '14 at 4:37
  • $\begingroup$ Interesting question. I don't see the connection between Zeta regularization and your limit, could you provide a reference for that? $\endgroup$ Dec 8 '14 at 4:39
  • $\begingroup$ For the $\Gamma$ function, the values are the same as $~\displaystyle\lim_{\epsilon\to0}\frac{\Gamma(n+\epsilon)+\Gamma(n-\epsilon)}2$. $\endgroup$
    – Lucian
    Dec 8 '14 at 5:23
  • $\begingroup$ @Antonio Vargas there was a typo in the formula, sorry (actually I inserted the wrong formula). $\endgroup$
    – Anixx
    Dec 8 '14 at 5:44
  • $\begingroup$ Sorry, I wasn't clear. I wasn't commenting on the formula, I was wondering how it (or the new one) is connected to Zeta regularization. It's not obvious to me, so I was hoping you could provide a reference. Your other question doesn't say either. $\endgroup$ Dec 8 '14 at 5:46

Well at least for the second part of the question, each function having pole at $x_0$ can be regularized the following way:

$$f(x_0)=\lim_{h\to 0}\frac{f(x_0+h)+f(x_0-h)}2+\omega \lim_{h\to 0}h\frac{f(x_0+h)-f(x_0-h)}2 $$

Or at any point,

$$f(x)=\lim_{h\to 0}\frac{f(x+h)+f(x-h)}2+\lim_{h\to 0}\frac{f(x+h)-f(x-h)}2 $$


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.