7

I am self-learning Real Analysis from the text, Understanding Analysis by Stephen Abbott. I'd like someone to verify, if my below proof and deductions are rigorous and technically correct.

[Abbott 6.4.6] Let

\begin{equation*} f( x) =\frac{1}{x} -\frac{1}{x+1} +\frac{1}{x+2} -\frac{1}{x+3} +\frac{1}{x+4} -\dotsc \end{equation*}

Show that $\displaystyle f$ is defined for all $\displaystyle x >0$. Is $\displaystyle f$ continuous on $\displaystyle ( 0,\infty )$? How about differentiable?

Proof.

Well-definedness of $\displaystyle f$.

Define

\begin{equation*} f_{n}( x) =\frac{( -1)^{n}}{x+n} \end{equation*} Then,

\begin{equation*} f( x) =\sum _{n=0}^{\infty } f_{n}( x) \end{equation*} Let $\displaystyle x_{0}$ be an arbitrary point, such that $\displaystyle x_{0} >0$. Fix $\displaystyle x=x_{0}$.

We have:

\begin{equation*} \frac{1}{x_{0}} \geq \frac{1}{x_{0} +1} \geq \frac{1}{x_{0} +2} \geq \dotsc \geq 0 \end{equation*} Moreover,

\begin{equation*} \lim \frac{1}{x_{0} +n} =0 \end{equation*} By the Alternating Series Test for convergence, $\displaystyle \sum _{n=0}^{\infty } f_{n}( x)$ converges pointwise on $\displaystyle x >0$.

Continuity of $\displaystyle f$.

Let $\displaystyle [ a,\infty )$ be any interval such that $\displaystyle a >0$. Let us group each of pair of terms of $\displaystyle f$ and write:

\begin{equation*} \begin{array}{ c l } f( x) & =\left(\frac{1}{x} -\frac{1}{x+1}\right) +\left(\frac{1}{x+2} -\frac{1}{x+3}\right) +\left(\frac{1}{x+4} -\frac{1}{x+5}\right) +\dotsc \end{array} \end{equation*}

Define:

\begin{equation*} g_{n}( x) =\frac{1}{( x+2n)} -\frac{1}{( x+2n+1)} =\frac{1}{( x+2n)( x+2n+1)} \end{equation*}

Then,

\begin{equation*} f( x) =\sum _{n=0}^{\infty } g_{n}( x) \end{equation*} Now, $\displaystyle g_{0}( x) =\frac{1}{a( a+1)} =M_{0}$. Moreover,

\begin{equation*} 0\leq g_{n}( x) \leq \frac{1}{( 2n)( 2n+1)} \leq \frac{1}{4n^{2}} =M_{n} \end{equation*}

for all $\displaystyle n\geq 1$. Since $\displaystyle \sum _{n=0}^{\infty } M_{n}$ converges, by the Weierstrass $\displaystyle M$-Test, $\displaystyle \sum_{n=0}^{\infty } g_{n}( x)$ converges uniformly on $\displaystyle [ a,\infty )$ for any $\displaystyle a >0$.

Since each $\displaystyle g_{n}( x)$ is continuous for $\displaystyle x >0$, by the Term-by-term continuity theorem, $\displaystyle f( x)$ is continuous on $\displaystyle [ a,\infty )$, where $\displaystyle a >0$. Thus, $\displaystyle f$ is continuous on $\displaystyle ( 0,\infty )$.

Differentiability of $\displaystyle f$.

Since each $\displaystyle g_{n}( x)$ is differentiable for $\displaystyle x >0$, by the Term-by-Term differentiability theorem, $\displaystyle f$ is differentiable on $\displaystyle [ a,\infty )$ where $\displaystyle a >0$. Thus, $\displaystyle f$ is differentiable on $\displaystyle ( 0,\infty )$.

Quasar
  • 5,644
  • 1
    What's the statement of the "Term-by-Term differentiabilty theorem". Just because $g_n(x)$ is differentiable and $G(x) = \sum_{n=0}^\infty g_n(x)$ it is not always true that $G$ is differentiable with $G'(x) = \sum_{n=0}^\infty g_n'(x)$. You haven't proved anything yet regarding the series derivative. – RRL Nov 20 '22 at 21:52
  • 1
    Your proof of convergence and continuity is correct but see caveat below. – RRL Nov 20 '22 at 22:10
  • Just for you curiosity, $f(x)=\Phi (-1,1,x)$ where appears the Hurwitz[Dash]Lerch transcendent function. – Claude Leibovici Nov 21 '22 at 08:35

1 Answers1

5

It can be invalid to sum a series by arbitrarily rearranging terms when the series is not absolutely convergent. However, you can sum any convergent series in pairs as you have done even if the series is only conditionally convergent because $f_{2n}(x) \to 0$ as $n \to \infty$ and

$$\sum_{j= 1}^n\{|f_{2j-2}(x)| -|f_{2j-1}(x)|\}=\sum_{k=0}^{2n-1}f_k(x) = \sum_{k=0}^{2n}f_k(x)- f_{2n}(x),$$

so that partial sums with odd and even numbers of terms converge to the same limit.

Nevertheless, this pairing argument is unnecessary. Since $\frac{1}{n+x} \to 0$ as $n \to \infty$ both monotonically and uniformly and $\sum_{k=0}^n (-1)^k$ is uniformly bounded for all $n$ and (trivially) for all $x$, the Dirichlet test for series of functions implies that the series convergence is uniform.

By showing that $\sum f_n'(x)$ is uniformly convergent you can finish the part about differentiability.

RRL
  • 92,835
  • 7
  • 70
  • 142
  • Arbitrarily grouping terms should be fine i.e. infinite addition is associative, as long as we have conditional convergence right? For example, if $s_n(x_0)$, the sequence of partial sums evaluated at a point $x_0$ converges, every subsequence $(s_{n_k})$ also converges to the same limit. – Quasar Nov 21 '22 at 06:44
  • @Quasar: I think that is correct. It is the rearrangement that is problematic. Nevertheless your pairing of the terms works and is a nice trick to prove uniform convergence albeit unnecessary in this case. – RRL Nov 21 '22 at 11:29
  • 1
    I’m still not sure what theorem you are referring to for differentiability. I’m sure you know that the uniform convergence of the series of derivatives along with convergence of the original series at one point is sufficient for termwise differentiation. – RRL Nov 21 '22 at 11:33
  • You're right and that is the statement of the Term-by-term differentiability theorem. – Quasar Nov 22 '22 at 18:47