2
$\begingroup$

Let $$\sum_{n=1}^{\infty} g_n(x)$$ converge uniformly to a function $g(x)$ on $A \subseteq \mathbb{R}$. Can we say that the series defined by $$\sum_{n=1}^{\infty} f_n(x) = \sum_{n=1}^{\infty} \frac{1}{n}g_n(x)$$ converges uniformly to a function $f(x)$ on $A \subseteq \mathbb{R}$? If so, prove it. If not, give a counter example.

I have not been able to find a counterexample to this proposition, and my intuition says that this statement is true. When trying to prove it, I attempt to make use of the Cauchy Criterion for the Uniform Convergence of a Series which states that for any $\epsilon > 0,$ there exists an $N \in \mathbb{N}$ such that $|g_{m+1}(x)+g_{m+2}(x)+...+g_n(x)| < \epsilon$ whenever $n>m \geq N$ and $x \in A.$

Now, I consider the analogous sum for the second series; namely, $|\frac{1}{m+1}g_{m+1}(x)+\frac{1}{m+2}g_{m+2}(x)+...+\frac{1}{n}g_n(x)|.$ If we add the restriction that $g_n(x) \geq 0$ for all $n \in \mathbb{N}$, the result falls out immediately as this sum will be less than or equal to the sum in the preceding paragraph. However, I do not know how to do it for general $g_n(x).$

Any help would be appreciated, but please try to only give hints.

Thanks

$\endgroup$
1
$\begingroup$

If $g$ is bounded, the answer is yes.

For any $N \ge 1$, $\sum_{n=1}^N \frac{g_n(x)}{n} = \frac{\sum_{n=1}^N g_n(x)}{N}+\int_1^N \frac{\sum_{n \le t} g_n(x)}{t^2}dt$, so

$\left|\sum_{n=1}^N \frac{g_n(x)}{n}-\int_1^\infty \frac{\sum_{n \le t} g_n(x)}{t^2}dt\right| \le \frac{\left|\sum_{n=1}^N g_n(x)\right|}{N}+\int_N^\infty \frac{\left|\sum_{n \le t} g_n(x)\right|}{t^2}dt$. Take $\epsilon > 0$ and then $N_0$ so that $|\sum_{n=1}^N g_n(x) - g(x)| \le \epsilon$ for all $x \in A$ and $N \ge N_0$. Then for any $x \in A$ and any $N \ge N_0$, $\left|\sum_{n=1}^N \frac{g_n(x)}{n}-\int_1^\infty \frac{\sum_{n \le t} g_n(x)}{t^2}dt\right| \le \frac{g(x)+\epsilon}{N}+\int_N^\infty \frac{g(x)+\epsilon}{t^2}dt \le 2\frac{g(x)+\epsilon}{N} \le 2\frac{C+\epsilon}{N_0}$ is arbitrarily small.

$\endgroup$
2
  • 1
    $\begingroup$ Is there any way to approach this problem without the use of integrals? I am in analysis and we have not covered that theory yet. $\endgroup$
    – L. Tim
    Apr 14 '20 at 0:56
  • $\begingroup$ @L.Tim You're welcome for the answer! I'm not sure. $\endgroup$ Apr 14 '20 at 1:15
0
+50
$\begingroup$

Yes, we can say that. We can even prove that using the following idea. Put $h_0(x)=0$ for each $x\in A$ and $h_k=\sum_{n=1}^k g_n$ for each natural $k$. Then $g_n=h_n-h_{n-1}$ for each natural $k<m$ and so for each $x\in A$ we have $$\sum_{n=k}^{m} f_n(x) = \sum_{n=k}^{m} \frac{1}{n}g_n(x)=\sum_{n=k}^{m}\frac{1}{n}(h_n(x)-h_{n-1}(x))=$$ $$-\frac 1{k}h_{k-1}(x)+\sum_{n=k}^{m-1}\frac{1}{n(n+1)} h_n(x)+\frac 1{m}h_m(x).$$

$\endgroup$

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.