3
$\begingroup$

If $\psi\left(x\right)$ is the Chebyshev psi function we know that $$\psi_{0}\left(x\right)=x-\sum_{\rho}\frac{x^{\rho}}{\rho}-\frac{\zeta'}{\zeta}\left(0\right)-\frac{1}{2}\log\left(1-x^{-2}\right)$$ where $$\psi_{0}\left(x\right)=\begin{cases} \psi\left(x\right)-\frac{1}{2}\Lambda\left(x\right), & x\textrm{ is a prime power}\\ \psi\left(x\right), & \textrm{otherwise} \end{cases}$$ and $x>1$. Now if $0<x<1$ the identity does not hold due to the presence of log function. But the other terms are defined.

Question: Is it known an explicit formula for $\psi\left(x\right)$ if $0<x<1$? I mean $$\psi\left(x\right)=x-\sum_{\rho}\frac{x^{\rho}}{\rho}+\textrm{something}.$$

I ask this question since it would be very useful for my work. I have only found this identity $$\sum_{n\leq1/x}\frac{\Lambda\left(n\right)}{n}=\log\left(\frac{1}{x}\right)-\gamma+\sum_{\rho}\frac{x^{\rho}}{\rho}-x+\frac{1}{2}\log\left(\frac{1+x}{1-x}\right),\,0<x<1$$ but I don't see how to link that to the psi function.

Meaning of this question I have an integral of this type $$\int_{0}^{\infty}\psi\left(t\right)f\left(t,z\right)dt$$ where $z$ is a complex number with positive real part. I would to use the explicit formula for $\psi\left(t\right)$ and integrate termwise but I can only use it for $t>1$. And using another technique for evaluate the integral I get some terms that match with the integration termwise for $t>0$. Unfortunately the other technique is not useful for a complete evaluation. So I'm asking if it possible to extend in some way the formula.

$\endgroup$
  • $\begingroup$ For $x>0$ there's always a formula for $\psi_0^{\phantom .}(x)$ as a contour integral with a factor $x^s \, ds$. When $x>1$ you make this smaller by moving the contour to the left, encountering poles that yield the contributions $x - \sum_\rho x^\rho/\rho$. When $x<1$, you make $x^s$ smaller by moving the contour to the right and encounter no singularities, so as expected $\psi_0^{\phantom .}(x) = 0$ in that case. $\endgroup$ – Noam D. Elkies Feb 24 '17 at 3:02
  • $\begingroup$ @NoamD.Elkies Of course, but I'm interested in a formula involving the terms $x-\sum_{\rho}\frac{x^{\rho}}{\rho}$. In other words, a funny way to write $0$ using that terms. I repeat, studying my integral in two different ways I saw that the the first terms match with a calculation of the explicit formula for $t>0$, so I'm asking if it is possible to extend it in some way, keeping $x-\sum_{\rho}\frac{x^{\rho}}{\rho}$ (who have perfectly sense fot $0<x<1$). $\endgroup$ – User Feb 24 '17 at 7:38
5
$\begingroup$

The Chebyshev function $\psi(x)$ is constant zero for $0<x<1$. It does not get more explicit than that.

$\endgroup$
  • $\begingroup$ Yes, but it is $0$ also for $1<x<2$. And in that case we have also the explicit formula. $\endgroup$ – User Feb 22 '17 at 15:54
  • 3
    $\begingroup$ @User: Yes, but the explicit formula $\psi(x)=0$ is simpler. In general, the explicit formula you are talking about is useful for large $x$, not for small $x$. I don't see any reason why you would like to bring in the zeta zeros for expressing the constant zero function. $\endgroup$ – GH from MO Feb 22 '17 at 15:57
  • $\begingroup$ I have an integral of this type $$\int_{0}^{\infty}\psi\left(t\right)f\left(t,z\right)dt$$ where $z$ is a complex number with positive real part. I would to use the explicit formula for $\psi\left(t\right)$ and integrate termwise but I can only use it for $t>1$. And using another technique for evaluate the integral I get some terms that match with the integration termwise for $t>0.$ Unfortunately the other technique is not useful for a complete evaluation. So I'm asking if it possible to extend in some way the formula. $\endgroup$ – User Feb 22 '17 at 16:06
  • $\begingroup$ @User: I see. I think it would more useful if you asked a more specific question about your integral. Whatever formula we generate for $\psi(t)$ valid for $0<t<1$, it will not harmonize with the original formula valid for $t>1$, exactly because $\log(1-t^{-2})$ has a serious singularity at $t=1$ (and so does the sum over the zeros). $\endgroup$ – GH from MO Feb 22 '17 at 16:14
  • $\begingroup$ Thank you. I proved that the integral converges if we use the complex log instead of the real log. If all terms converges can I use some type of analytic continuation? I essentially split the integral from $0$ to $1$ and $1$ to $\infty$ and all terms converges due to the particular form of $f(t,z).$ $\endgroup$ – User Feb 22 '17 at 16:17
0
$\begingroup$

Indeed the explicit formula is about the Laplace transform pairs :

$$\begin{array}{cc}\frac{1}{s}&\qquad & \frac{-\zeta'(s)}{s\zeta(s)} = \frac{1}{s} \left(\frac{1}{s-1}-\sum_\rho \frac{1}{s-\rho}- \sum_{n=1}^\infty (\frac{1}{s+2n}-\frac{1}{2n})+C\right) \\ \color{red}\uparrow & &\color{red}\uparrow \\ 1_{u > 0}& &\psi(e^u) = 1_{u > 0}(e^u-\sum_\rho \frac{e^{\rho u}}{\rho}- \sum_{n=1}^\infty \frac{e^{-2n u}}{-2n}- \ln 2\pi)\end{array}$$

$\endgroup$

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.