I wrote this sum (out of the blue) and wondered if it has a closed form:

$$\sum_{k=1}^{\infty} L^{\frac{1}{k}} \cdot(-1)^{k+1}$$ where $L \in \mathbb{N}$

I thought of a sum that would use "$\text{k-root}$" but with alternating sign ($+$ to $-$ etc..)

I couldn't find a way to do so, the only thing I did is write a program that calculates it, so I post here for help. Thank you so much!! :-)

  • $\begingroup$ Does it converge? $\endgroup$ Aug 3 '20 at 3:01
  • $\begingroup$ @JackyChong Well in my program I checked values $k = 1,000,000+$ and it seems to converge but I am not sure.. $\endgroup$
    – MathAsker
    Aug 3 '20 at 3:02
  • 1
    $\begingroup$ Have you tried the divergence test? $\endgroup$ Aug 3 '20 at 3:02
  • 1
    $\begingroup$ Individual terms converge to $1$ not $0$. $\endgroup$ Aug 3 '20 at 3:05
  • $\begingroup$ I found that If this sum converged, your partial sums too.$\sum_{k=1}^{\infty} L^{\frac{1}{k}} -1$ $\endgroup$
    – Player0
    Aug 3 '20 at 3:31

It doesn’t converge unless $L=0$. For $n\ge 1$ let $s_n=\sum_{k=1}^nL^{1/k}(-1)^{k+1}$. If $L\ge 1$, then $L^{1/k}\ge 1$ for all $k\ge 1$, so $|s_{n+1}-s_n|\ge 1$ for all $n\ge 1$. Thus, the sequence of partial sums is not Cauchy and cannot converge.

  • $\begingroup$ Wow nice answer sir! But something is weird, because in my program, it seems to get fixed on a specific number (according to L of course) $\endgroup$
    – MathAsker
    Aug 3 '20 at 3:07
  • 1
    $\begingroup$ @StackOMeow: I would check the program; for $L=2$ the partial sums for a modest number of terms should appear to oscillate between values not too far from $0.7$ and $1.7$. $\endgroup$ Aug 3 '20 at 3:11
  • $\begingroup$ Ohh right! it is indeed, for some even values of $k$ the sum = $c$ and for odd $k$ equals $c+1$ . So will we be able to find a closed form of the partial sums for even / odd $k$'s ? thank you sir! $\endgroup$
    – MathAsker
    Aug 3 '20 at 3:16
  • $\begingroup$ @StackOMeow: You’re welcome! I don’t know for sure about the odd and even subsequences, but it wouldn’t surprise me. $\endgroup$ Aug 3 '20 at 6:53

I can prove its subsequences's convergence but can't find their exact values. $(L-L^{\frac{1}{2}})+(L^{\frac{1}{3}}-L^{\frac{1}{4}})+...+(L^{\frac{1}{n-1}}-L^{\frac{1}{n}}) = L+(-L^{\frac{1}{2}}+L^{\frac{1}{3}})+(-L^{\frac{1}{4}}+L^{\frac{1}{5}})+...-L^{\frac{1}{n}} <L-1$

$L+(-L^{\frac{1}{2}}+L^{\frac{1}{3}})+(-L^{\frac{1}{4}}+L^{\frac{1}{5}})+...+(-L^{\frac{1}{n-1}}+L^{\frac{1}{n}}) =(L-L^{\frac{1}{2}})+(L^{\frac{1}{3}}-L^{\frac{1}{4}})+...+L^{\frac{1}{n}} >1$


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.