Liouville Lambda Function and Riemann Hypothesis

112 Views Asked by At

What is the exact statement involving the Liouville Lambda function, which is equivalent to Riemann Hypothesis, and true iff RH is true? Can anyone cite the sources for it and/or outline its proof in brief?

Just curious, because hardly any literature available regarding it, unlike other equivalent problems of RH. Thanks in advance!

1

There are 1 best solutions below

7
On BEST ANSWER

For ${\rm Re}(s) > 1$ we have $\sum_{n \geq 1} \lambda(n)/n^s = \zeta(2s)/\zeta(s)$, which is analogous to $\sum_{n \geq 1} \mu(n)/n^s = 1/\zeta(s)$ for the Mobius function. Set $L(x) = \sum_{n \leq x} \lambda(n)$, which is analogous to the function $M(x) = \sum_{n \leq x} \mu(n)$.

There is a link between the functions $\lambda$ and $\mu$: $\lambda(n) = \sum_{d^2 \mid n} \mu(n/d^2)$ for all $n \geq 1$, where the sum runs over squares dividing $n$. Show that implies $L(x) = \sum_{d \leq \sqrt{x}} M(x/d^2)$ for all $x \geq 1$, which is a link between $L$ and $M$ that lets us convert some properties of $M$ into similar properties of $L$.

Let's show RH is equivalent to the bound $L(x) = O(x^{1/2 + \varepsilon})$ for all $\varepsilon > 0$, where the $O$-constant is allowed to depend on $\varepsilon$. This is the $L$-analogue of RH being equivalent to the bound $M(x) = O(x^{1/2 + \varepsilon})$ for all $\varepsilon > 0$, which is more widely accessible in the literature.

To show RH implies $L(x) = O(x^{1/2 + \varepsilon})$ for all $\varepsilon > 0$, use the equation $L(x) = \sum_{d \leq \sqrt{x}} M(x/d^2)$ to show that if $M(x) = O(x^{1/2 + \varepsilon})$ for an $\varepsilon > 0$, then $L(x) = O(x^{1/2 + \varepsilon})$ for the same $\varepsilon$. It should be easier for you to find a proof in the literature that RH implies the estimate $M(x) = O(x^{1/2 + \varepsilon})$ for all $\varepsilon > 0$, so RH implies $L(x) = O(x^{1/2 + \varepsilon})$ for all $\varepsilon > 0$ too.

To show, conversely, that the bound $L(x) = O(x^{1/2 + \varepsilon})$ for all $\varepsilon > 0$ implies RH, start with the equation $$ \frac{\zeta(2s)}{\zeta(s)} = \sum_{n \geq 1} \frac{\lambda(n)}{n^s} = s\int_1^\infty \frac{L(x)}{x^{s+1}}\,dx $$ when ${\rm Re}(s) > 1$. We already know $\zeta(s)$ is meromorphic on $\mathbf C$, so $\zeta(2s)/\zeta(s)$ is meromorphic on $\mathbf C$. When $L(x) = O(x^{1/2 + \varepsilon})$ for an $\varepsilon > 0$, show the integral on the right side is absolutely convergent when ${\rm Re}(s) > 1/2 + \varepsilon$, which implies the right side is analytic for such $s$. Thus $\zeta(2s)/\zeta(s)$ is analytic when ${\rm Re}(s) > 1/2 + \varepsilon$. If there is an $s_0$ where $\zeta(s_0) = 0$ and ${\rm Re}(s_0) > 1/2 + \varepsilon$ then $\zeta(2s)/\zeta(s)$ has a pole at $s_0$ because $\zeta(2s)$ is analytic and nonvanishing when ${\rm Re}(s) > 1/2$. We have contradicted analyticity of $\zeta(2s)/\zeta(s)$ when ${\rm Re}(s) > 1/2 + \varepsilon$, so the bound $L(x) = O(x^{1/2 + \varepsilon})$ implies the zeta-function has no zeros with real part greater than $1/2 + \varepsilon$. So if $L(x) = O(x^{1/2 + \varepsilon})$ for all $\varepsilon > 0$, then $\zeta(s)$ has no zeros with ${\rm Re}(s) > 1/2$, which is RH.

If you have ever seen a proof that the bound $M(x) = O(x^{1/2 + \varepsilon})$ for all $\varepsilon > 0$ implies RH, you should recognize that the previous paragraph is essentially the same argument, with the role of the displayed integral identity above for ${\rm Re}(s) > 1$ replacing the identity $$ \frac{1}{\zeta(s)} = \sum_{n \geq 1} \frac{\mu(n)}{n^s} = s\int_1^\infty \frac{M(x)}{x^{s+1}}\,dx $$ for ${\rm Re}(s) > 1$.

The reason you probably don't see much written directly about relationships between RH and the Liouville function is that formulating and proving such results at a basic level is always just an exercise in adapting what is more widely written about $\mu(n)$ or $M(x)$, as I illustrated above.

Here are two things to try yourself.

  1. Locate a proof that the prime number theorem is equivalent to $M(x) = o(x)$ and adapt that method to prove the prime number theorem is equivalent to $L(x) = o(x)$.

  2. Locate a proof that RH is equivalent to convergence of $\sum_{n \geq 1} \mu(n)/n^s$ for ${\rm Re}(s) > 1/2$ and adapt that method to prove RH is equivalent to convergence of $\sum_{n \geq 1} \lambda(n)/n^s$ for ${\rm Re}(s) > 1/2$. This is a second equivalence between RH and the Liouville function. (The series $\sum_{n \geq 1} \lambda(n)/n^s$ and $\sum_{n \geq 1} \mu(n)/n^s$ turn out not to converge anywhere on ${\rm Re}(s) = 1/2$, which is analogous to $\sum_{n \geq 1} 1/n^s$ not converging anywhere on ${\rm Re}(s) > 1$ and is not as technical to prove as the RH equivalences.)