Let $\epsilon >0$ and suppose that $\sum_{k=1}^{\infty}a_k\geq 1$. I need to show that there is $k_0$ and constant $C(\epsilon)$ such that $a_{k_0}\geq 2^{-k_0\epsilon}/C(\epsilon)$. I prove this by contradiction: assume that the inequality does not hold for any constant $C(\epsilon)$. Especially we can choose $C(\epsilon)=n$ so that $$ a_k<\frac{1}{n}2^{-k\epsilon} \quad \text{for all} \ n\in\mathbb{N} $$ But now using geometric series we have $$ \sum_{k=1}^{\infty}a_k<\frac{1}{n}\sum_{k=1}^{\infty}\bigg(\frac{1}{2^{\epsilon}}\bigg)^k<\frac{1}{n}\frac{1}{2^{\epsilon}-1}<1 $$ for $n$ large enough. This is a contradiction.

My question is: Is my proof correct and why the constant $C$ has to be dependent on $\epsilon$?

  • 2
    $\begingroup$ Congratulations on getting the lucky post number 3 million! (see the URL) $\endgroup$ – Wojowu Nov 15 '18 at 18:44
  • 1
    $\begingroup$ Didn't notice that! I hope I will get lucky this weekend! $\endgroup$ – peastick Nov 16 '18 at 7:04

Your original statement is $\forall \varepsilon>0$, $\exists C>0$ and $k_0\in{\mathbb N}$ such that $a_{k_0}\ge\frac{1}{C}2^{-k_0\varepsilon}$.

So if the negation should be something like:

$\exists \varepsilon_0>0$ such that for any $C>0$ and any $k\in{\mathbb N}$, it holds that $a_k<\frac{1}{C}2^{-k\varepsilon_0}$.

For THAT $\varepsilon_0>0$,then you can use directly the last inequality for $C=1$, Hence, $$1<\sum_{k=1}^\infty a_k \le \sum_{k=1}^\infty 2^{-k\varepsilon_0} =\frac{2^{-\varepsilon_0}}{1-2^{-\varepsilon_0}}< 1,$$ which is a contradiction.

So, your argument is almost correct, but you have to polish some steps.

EDIT Maybe some steps are not clear or wrong. Here is a better argument:

For that $\varepsilon_0$, take $C=\frac{1-2^{-\varepsilon_0}}{2^{1-\varepsilon_0}}>0$, so $$1\le \sum_{k=1}^\infty a_k \le C \sum_{k=1}^\infty 2^{-k\varepsilon_0} =C\frac{2^{-\varepsilon_0}}{1-2^{-\varepsilon_0}}=\frac{1}{2},$$ which is a contradiction.

  • $\begingroup$ Thanks for the answer, but could you explain why your last inequality is true? And your first inequality should be $\leq$. $\endgroup$ – peastick Nov 16 '18 at 7:03
  • $\begingroup$ @peastick Look the EDITED answer $\endgroup$ – Tito Eliatron Nov 16 '18 at 10:31

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.