1
$\begingroup$

The infinite sums involving mobius function and a multiplicative function has got quite interest in past. In particular, sums of the form $$\sum_{d=1}^{\infty}\frac{\mu(d)}{f(d)}$$ for mobius function $\mu$ and multiplicative function $f$ have been investigated for various $f.$ I am interested in knowing about any arguments that could prove/disprove the non-negativity of the following sum $$\sum_{d=1}^{\infty}\frac{\mu(d)}{\mathrm{lcm}(d,\varphi(d))}$$ where $\varphi$ is the euler totient function. The function $f(d)=\mathrm{lcm}(d,\varphi(d))$ is not multiplicative and hence any standard techniques of treating multiplicative $f$ won't work here.

I would like to remark that the sum is absolutely convergent. To see this, one can consider the Lucas sequence $u_n=2^n-1$ and let $\mathrm{ord}_n(2)$ denote the multiplicative order of $2$ modulo $n.$ It is well known that $\mathrm{ord}_n(2)\mid \varphi(n).$ This gives that $$\mathrm{lcm}(n,\varphi(n))\ge \mathrm{lcm}(n,\mathrm{ord}_n(2)).$$ Thus, we have that $$\sum_{d=1}^{\infty}\frac{1}{\mathrm{lcm}(d,\varphi(d))}\le \sum_{d=1}^{\infty}\frac{1}{\mathrm{lcm}(d,\mathrm{ord}_d(2))}$$ and the convergence of right sum follows by proposition 1.4 in this published paper.

Thanks in advance for any help.

$\endgroup$

1 Answer 1

5
$\begingroup$

In general, it is better to approach such a question numerically, since your sum is absolutely convergent. However, in your particular case, it is possible to compute this explicitly without any numerical calculations. Notice that non-zero summands that appear in your sum correspond to squarefree $d$ (otherwise $\mu(d)=0$). Next, take a large $X$ and consider all squarefree $d$ with $2<d\leq X$. If such a $d$ is even, then $d=2d_1$ and $d_1>1$ is squarefree and odd. Since $\varphi(d)=\varphi(d_1)$ is even, we have $[d,\varphi(d)]=[d_1,\varphi(d_1)]$ (here $[a,b]=\mathrm{lcm}(a,b)$). On the other hand, $\mu(d)=\mu(2d_1)=-\mu(d_1)$, therefore $$ \frac{\mu(d)}{[d,\varphi(d)]}+\frac{\mu(d_1)}{[d_1,\varphi(d_1)]}=0. $$ Every even squarefree $d$ in $(2,X]$ is a member of one such pair, and same is true for odd squarefree $d$ with $1<d\leq X/2$, so $$ \sum_{d\leq X}\frac{\mu(d)}{[d,\varphi(d)])}=1-\frac12+\sum_{\text{odd }d \text{ in }(X/2,X]}\frac{\mu(d)}{[d,\varphi(d)]}. $$ This last summand can be estimated as follows $$ \left|\sum_{\text{odd }d \text{ in }(X/2,X]}\frac{\mu(d)}{[d,\varphi(d)]}\right|\leq \sum_{\text{odd }d>X/2}\frac{1}{[d,\mathrm{ord}_2(d)]}\ll \exp(-1/3(\ln X\ln\ln X)^{1/2})=o(1), $$ by the paper you linked. Therefore, your sum is equal to $1/2$.

$\endgroup$

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.