I would like to know the value of the following integral:

$$\displaystyle\int_2^\infty ({\theta(y)-y})\frac{\mathrm d}{\mathrm dy}\frac{\ln(y-1)}{\ln(y)} \mathrm dy$$

(Where $\theta(y)$ is Chebyshev's First Function)

It appears on (4.16) of J. Barkley and L. Schoenfeld "Approximated Formulas for some functions of Prime Numbers".

I do not know where to start from, and neither Mathematica nor Mathlab seem to help.

How can I find at least an approximated value for it?

  • $\begingroup$ By parts maybe? $\endgroup$
    – Laplacian
    Jan 8, 2017 at 23:25
  • $\begingroup$ Where does the $\sum\limits_{p\leqslant y} \frac{1}{p}$ come from? $\endgroup$ Jan 9, 2017 at 19:01
  • $\begingroup$ @Daniel Integrating by parts, it is the derivative of $\theta(y)$ $\endgroup$ Jan 9, 2017 at 19:07
  • 3
    $\begingroup$ That's wrong. $\theta$ is constant between consecutive primes, and has a jump of height $\ln p$ at a prime $p$. Its distributional derivative is hence $\sum (\ln p)\delta_p$, where $\delta_a$ is the Dirac distribution at $a$. The classical derivative doesn't exist at primes, and is $0$ everywhere else, so that isn't useful here. If you integrate by parts, one term you get is $\int_2^k \frac{\ln (y-1)}{\ln y}\,d\theta(y)$, which is a Riemann-Stietjes integral. Due to the nature of $\theta$, you can also write that as $\sum\limits_{p\leqslant k}\frac{\ln (p-1)}{\ln p}\cdot \ln p$, which can $\endgroup$ Jan 9, 2017 at 19:16
  • 3
    $\begingroup$ be simplified to $\sum\limits_{p\leqslant k} \ln (p-1)$. Unfortunately, that doesn't look too helpful yet. $\endgroup$ Jan 9, 2017 at 19:16

1 Answer 1


Might be helpful. You have an integral of the form

$$\int f g'\ \text{d} y$$

Use by parts integration to get

$$\underbrace{\left(\theta(y) - y)\right)\frac{\ln(y-1)}{\ln(y)}\bigg|_2^{+\infty}}_{0} - \int_2^{+\infty} \left(\theta(y) - y\right)'\ \frac{\ln(y-1)}{\ln(y)}\ \text{d}y$$


$$\frac{\text{d}}{\text{d}y} \theta(y) = \sum_{p\leq y} \frac{1}{p}$$

$$-\left(\sum_{p\leq y} \frac{1}{p} - 1\right)\int_2^{+\infty} \frac{\ln(y-1)}{\ln(y)}\ \text{d}y$$

Your problem arises when you try to calculate the last integral, which does diverge.

We can approach it with a consideration like at infinity the integral approaches to the value $1$.

I think now it's your turn to make some approximation to that integral.

  • $\begingroup$ Thank you! In your last step, why do you use $$-\left(\sum_{p\leq y} \frac{1}{p} - 1\right)\int_2^{+\infty} \frac{\ln(y-1)}{\ln(y)}\ \text{d}y$$ instead of $$-\left(\sum^\infty \frac{1}{p} - 1\right)\int_2^{+\infty} \frac{\ln(y-1)}{\ln(y)}\ \text{d}y$$ ? $\endgroup$ Jan 9, 2017 at 6:42
  • $\begingroup$ Also, another problem I have is that, writting $$-\left(\sum_{p\leq y} \frac{1}{p} - 1\right)\int_2^{+\infty} \frac{\ln(y-1)}{\ln(y)}\ \text{d}y$$ as $$\left(\sum_{p\leq y} \frac{p-1}{p} \right)\int_2^{+\infty} \frac{\ln(y-1)}{\ln(y)}\ \text{d}y$$ , both terms diverge, and so its product, isn't it? $\endgroup$ Jan 9, 2017 at 7:08
  • $\begingroup$ @user3141592 For the first question, I wrote the sum in that way because the definition I remember about the $\theta(y)$ function is that it's the sum of the logarithm of all the prime numbers less or equal than $y$. Hence if $y$ is a finite number the sum must be finite! For the second question, the $-1$ term IS NOT inside the sum. There would be less confusion if I had written it as $$\left(\sum_{p\leq y} \frac{1}{p}\right) - 1$$ $\endgroup$
    – Laplacian
    Jan 9, 2017 at 12:09
  • $\begingroup$ But since $$\sum_{p\le k} \frac{1}{p}$$ is arout $$ln(ln(k))$$ and the integral $$\int_2^{k} \frac{\ln(y-1)}{\ln(y)}\ \text{d}y is around $k$, it does not converge, doest it? $\endgroup$ Jan 9, 2017 at 12:48
  • $\begingroup$ And is it sure that $$\theta(\infty)-\infty = 0$$?? $\endgroup$ Jan 9, 2017 at 14:03

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.