Let $f_k$ be a real numbers such that $\sum_{k=1}^\infty f_k < \infty$. For each $R > 0$, define the convergent sum $$v(R) = \sum_{k=1}^\infty f_k(b_k(R)e^{-ky} - c_k(R)e^{ky})$$ where $0 \leq y \leq R$ and where $b_k$ and $c_k$ are functions of $k$ and $R$ and satisfy $$b_k(R) = 1+ \frac{1}{e^{2kR}-1}\quad \text{and}\quad c_k(R) = b_k(R)-1 = \frac{1}{e^{2kR}-1}$$

Apparently this sum uniformly converges (see this) in $n$ as function of $R$. How do I show this? I.e. if $v_n(R) = \sum_{k=1}^n f_k(b_k(R)e^{-ky} - c_k(R)e^{ky})$ how do I show that $v_n \to v$ uniformly as function of $R$?

I didn't understand how the explanation is given in that thread and I think my question is too basic to be asked there. Both the coefficients have look "like" $e^{-2kR}$ so it seems like it should be uniform but I don't know the argument.

  • $\begingroup$ Is $R>\delta >0$? Or can $R$ be any real number? $\endgroup$ – Mark Viola Apr 21 '15 at 14:23
  • $\begingroup$ @Dr.MV Sorry $R > 0$, and $0 \leq y \leq R$. $\endgroup$ – RealMax Apr 21 '15 at 14:24
  • $\begingroup$ Yes, but is it restricted to be greater than a pre-assigned number? $\endgroup$ – Mark Viola Apr 21 '15 at 14:25
  • $\begingroup$ No, just that it is a positive number. $\endgroup$ – RealMax Apr 21 '15 at 14:26
  • 1
    $\begingroup$ You're welcome. My pleasure. And note, that if $f_k$ is the $k$th term of a Generalized Fourier series, then we certainly have that it converges in $\mathscr{L}^2$ (i.e., $\sum_k |f_k|^2$ converges). $\endgroup$ – Mark Viola Apr 21 '15 at 16:17

I'm just fumbling here, but if you compute a bit you get $$ b_k(R) e^{-ky} - c_k(R) e^{ky} = \frac 1 {e^{2kR}-1} \left( e^{2kR} e^{-ky} - e^{ky} \right). $$ For $R$ large enough, you can easily bound this by some constant times $$ e^{-ky} - e^{k(y-2R)}. $$ Since $R \ge y$, the second term will be $\le e^{-ky}$. So for large $R$, you can bound your summand by $$ | f_k (b_k(R) e^{-ky} - c_k(R) e^{ky}) | \le C e^{-ky} |f_k| \le C |f_k|. $$ Since the $\sum f_k$ converges, so does this new sum (assuming $f_k\ge 0$), and since the bound doesn't depend on $R$, it converges uniformly in $R$ (again, for large $R$).

Please double check all of this, I'm not sure if I messed up along the way...

EDIT: Ouch, Dr. MV had the more elegant idea obviously.


Assume that $\sum_k f_k$ converges absolutely.

Note that

$$\begin{align} |b_k(R)e^{-ky}-c_k(R)e^{ky}|&=\left|\frac{e^{2kR}e^{-ky}-e^{ky}}{e^{2kR}-1}\right|\\\\ &=\left|\frac{\sinh k(R-y)}{\sinh kR}\right|\\\\ &<1 \end{align}$$


$$\begin{align} \left|v(R) \right|&=\left|\sum_{k=1}^{\infty}f_k\left(b_k(R)e^{-ky}-c_k(R)e^{ky}\right) \right|\\\\ &\le \sum_{k=1}^{\infty}|f_k|\,\left|\left(b_k(R)e^{-ky}-c_k(R)e^{ky}\right)\right| \\\\ &\le \sum_{k=1}^{\infty}|f_k| \\\\ \end{align}$$

which converges by hypothesis!

Assume that $\sum_k |f_k|^2$ converges.


$$\begin{align} \left|v(R) \right|^2&=\left|\sum_{k=1}^{\infty}f_k\left(b_k(R)e^{-ky}-c_k(R)e^{ky}\right) \right|^2\\\\ &\le \sum_{k=1}^{\infty}|f_k|^2\,\,\sum_{k=1}^{\infty}\left|\left(b_k(R)e^{-ky}-c_k(R)e^{ky}\right)\right|^2 \\\\ &=\sum_{k=1}^{\infty}|f_k|^2\,\,\sum_{k=1}^{\infty}\left|\frac{\sinh k(R-y)}{\sinh kR}\right|^2 \end{align}$$

The first term converges by hypothesis. For the second term, note that

$$\left|\frac{\sinh k(R-y)}{\sinh kR}\right|^2\le e^{-2ky}$$

and $\sum_{k=1}^{\infty} e^{-2ky}$ converges for any fixed $y>0$.


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.