陶哲轩实分析（上）11.6及习题-Analysis I 11.6

Exerise 11.6.1

If $I$ is of the form $[a,b]$ then by Proposition 11.6.1 $f$ can be Riemann integrable. So we suppose $I$ is of the form $(a,b),(a,b],[a,b)$.
Since $f$ is bounded, $\exists M>0,-M\le f(x)\le M,\forall x\in I$, choose $0<\epsilon <(b-a)/2$ to be a small number, then by Proposition 11.6.1, $f$ is Riemann integrable on $[a+\epsilon ,b-\epsilon ]$, in particular we can find a piecewise constant function $h$ on $[a+\epsilon ,b-\epsilon ]$ which majorizes $f$ and
$${\int }_{[a+\epsilon ,b-\epsilon ]}h<{\int }_{[a+\epsilon ,b-\epsilon ]}f+\epsilon$$
Define
$$H(x)=\{\begin{array}{ll}h(x),& x\in [a+\epsilon ,b-\epsilon ]\\ M,& x\in I\backslash [a+\epsilon ,b-\epsilon ]\end{array}$$
Then $H$ majorizes $f$ on $I$ and is piecewise constant, we have
$${\overline{\int }}_{I}f\le {\int }_{I}H={\int }_{[a+\epsilon ,b-\epsilon ]}h+2M\epsilon <{\int }_{[a+\epsilon ,b-\epsilon ]}f+(2M+1)\epsilon$$
By the same logic we have
$${\underline{\int }}_{I}f>{\int }_{[a+\epsilon ,b-\epsilon ]}f-(2M+1)\epsilon$$
Thus
$${\overline{\int }}_{I}f-{\underline{\int }}_{I}f<(4M+2)\epsilon$$
This means $f$ is Riemann integrable on $I$.

Exerise 11.6.2

Definition: Let $I$ be a bounded interval, and let $f:I\to \mathbf{R}$. We say $f$ is piecewise monotonic on $I$ iff there exists a partition $\mathbf{P}$ such that $f{|}_J$ is monotonic on $J$ for all $J\in \mathbf{P}$.
If $f$ is a bounded piecewise monotone function, then there exists a partition $\mathbf{P}$ such that $f{|}_J$ is bounded and monotonic on $J$ for all $J\in \mathbf{P}$. By Corollary 11.6.3 $f{|}_J$ is Riemann integrable on $J$ for all $J\in \mathbf{P}$.
We define
$$F_J(x)=\{\begin{array}{ll}f{|}_J(x),& x\in J\\ 0,& x\in I\backslash J\end{array}$$
By Theorem 11.4.1(g), $F_J$ is Riemann integrable on $I$, and we further have
$$f(x)=\sum _{J\in P}{F}_J(x)$$
So by Theorem 11.4.1(a), $f$ is Riemann integrable on $I$.

Exerise 11.6.3

We consider ${\int }_{[0,N]}f$, by Proposition 11.6.1, ${\int }_{[0,N]}f$ exists for all $N\in \mathbf{N}$. We let
P={[n,n+1):n=0,1,…,N−1}∪{N}\mathbf P=\{[n,n+1):n=0,1,\dots,N-1\}\cup \{N\}P={[n,n+1):n=0,1,…,N−1}∪{N}
be a partition of $[0,N]$, and define $\overline{f}(x)=f(n),x\in [n,n+1),\overline{f}(N)=f(N)$, then $\overline{f}$ majorizes $f$, thus
$${\int }_{[0,N]}f\le {\int }_{[0,N]}\overline{f}=\sum _{J\in P}\overline{f}(x)=\sum _{n=0}^{N-1}f(n)$$
Similarly if we define $\underline{f}(x)=f(n+1),x\in [n,n+1),\underline{f}(N)=f(N)$, then $\underline{f}$ minorizes $f$, so
$${\int }_{[0,N]}f\ge {\int }_{[0,N]}\underline{f}=\sum _{J\in P}\underline{f}(x)=\sum _{n=1}^{N}f(n)$$
Thus we have ${\sum }_{n=1}^{N}f(n)\le {\int }_{[0,N]}f\le {\sum }_{n=0}^{N-1}f(n)$. Thus if ${\sum }_{n=0}^{\infty }f(n)$ converges, we can see that ${\sum }_{n=0}^{\infty }f(n)$ is an upper bound for ${\int }_{[0,N]}f,\forall N>0$, so ${\sup }_{N>0}{\int }_{[0,N]}f\le {\sum }_{n=0}^{\infty }f(n)$. On the other hand, if ${\sup }_{N>0}{\int }_{[0,N]}f$ is finite, and assume ${\sum }_{n=0}^{\infty }f(n)$ diverges, since $f(n)\ge 0,\forall n$, we must have ${\sum }_{n=0}^{\infty }f(n)=+\infty$, thus there is $N^{\prime }$ such that ${\sum }_{n=0}^{N^{\prime }}f(n)>\sup (N>0){\int }_{[0,N]}f+f(0)$, this means ${\sum }_{n=1}^{N^{\prime }}f(n)>{\sup }_{N>0}{\int }_{[0,N]}f\ge {\int }_{[0,N^{\prime }]}f\ge {\sum }_{n=1}^{N^{\prime }}f(n)$, a contradiction.

Exerise 11.6.4

We can define
$$f(x)=\{\begin{array}{ll}1,& x\in [n+1/4,n+3/4],n\in \mathbf{N}\\ 0,& x\in [0,+\infty ),x\notin [n+1/4,n+3/4],n\in \mathbf{N}\end{array}$$
then ${\sum }_{n=0}^{\infty }f(n)=0$, but ${\int }_{[0,N]}f=N/2$, thus ${\sup }_{N>0}{\int }_{[0,N]}f$ is not finite.
Conversely we can define
$$f(x)=\{\begin{array}{ll}1,& x=n,n\in \mathbf{N}\\ 0,& x\in [0,+\infty ),x\notin n,n\in \mathbf{N}\end{array}$$
Then ${\sum }_{n=0}^{\infty }f(n)=+\infty$, but ${\int }_{[0,N]}f=0$, thus ${\sup }_{N>0}{\int }_{[0,N]}f$ is finite.

##Exerise 11.6.5
When $p\le 0$, we have $1/n^p\nrightarrow 0$, thus ${\sum }_{n=1}^{\infty }(1/n^p)$ diverges.
When $p>0$, the function $f(x)=1/x^p$ is monotonic decreasing on $[1,+\infty )$, and $f(x)\ge 0,\forall x\in [1,+\infty )$, thus the condition of Proposition 11.6.4 is satisfied.
Consider
$${\int }_{[1,N]}f={\int }_{[1,N]}\frac{1}{x^p}$$
Use Exercise 10.4.3 and some integral calculation we can show that
$${\int }_{[1,N]}f={\int }_{[1,N]}\frac{1}{x^p}=\{\begin{array}{ll}\frac{1}{1-p}(N^{1-p}-1),& p\ne 1\\ \ln N,& p=1\end{array}$$
Thus if $0<p<1$, we have $N^{1-p}\to \infty$ as $N\to \infty$, so ${\sup }_{N>0}{\int }_{[0,N]}f$ is not finite.
if $p=1$, then $\ln N\to \infty$ as $N\to \infty$, so ${\sup }_{N>0}{\int }_{[0,N]}f$ is not finite.
If $p>1$, then for any $N$ we have
$${\int }_{[1,N]}f=\frac{1}{1-p}\left(\frac{1}{N^{p-1}}-1\right)=\frac{1}{p-1}\left(1-\frac{1}{N^{p-1}}\right)<\frac{1}{p-1}$$
so ${\sup }_{N>0}{\int }_{[0,N]}f$ is finite.
Thus we can conclude ${\sum }_{n=1}^{\infty }(1/n^p)$ only converges when $p>1$.