这一节的逻辑构成是这样的:先定义了在countable集合上的级数,并且在绝对收敛下有Fubini定理成立,而后在uncountable集合上也可以定义级数,只要满足任何有限集合的sup存在。如果f是一个定义在uncountable集合上的绝对收敛级数,那么f非退化的点是至多可数的(证明需要用选择公理)。绝对收敛级数有很好理解的series laws,条件收敛级数则有Riemann的结论:可以收敛到任何实数L。

这一节的习题主要是完善正文中的定理证明,然而也不易。

Exercise 8.2.1

Exercise 3.6.3 says if there’s a function whose domain is on a finite subset of N\mathbf NN, then the range of this function is also finite. I’ll use this result later.
Since XXX is at most countable, XXX may be finite or countable. If XXX is finite, then the statement is obviously true. Now we consider the case when XXX is countable.
First suppose the series ∑x∈Xf(x)∑_{x∈X}f(x)∑x∈X​f(x) is absolutely convergent, then there’s a bijection g:N→Xg:\mathbf N→Xg:N→X such that ∑n=0∞f(g(n))∑_{n=0}^∞f(g(n))∑n=0∞​f(g(n)) is absolutely convergent. Given any element EEE in the set
S={∑x∈A∣f(x)∣∶A⊆X,Afinite}S=\left\{∑_{x∈A}|f(x)|∶A⊆X,A\text{ finite}\right\}S={x∈A∑​∣f(x)∣∶A⊆X,A finite}
We know there’s a finite set AAA such that E=∑x∈A∣f(x)∣E=∑_{x∈A}|f(x)|E=∑x∈A​∣f(x)∣ . So we can have a finite subset S′S'S′ of NNN such that g(S′)=Ag(S')=Ag(S′)=A, from Exercise 3.6.3 we can have S′S'S′ is bounded above by some natural number kkk, so
E=∑x∈A∣f(x)∣=∑n∈S′∣f(g(n))∣≤∑n=0k∣f(g(n))∣≤∑n=0∞∣f(g(n))∣E=∑_{x∈A}|f(x)| =∑_{n∈S'}|f(g(n))| ≤∑_{n=0}^k|f(g(n))| ≤∑_{n=0}^∞|f(g(n))| E=x∈A∑​∣f(x)∣=n∈S′∑​∣f(g(n))∣≤n=0∑k​∣f(g(n))∣≤n=0∑∞​∣f(g(n))∣
Since this is true for all EEE, we know that ∑n=0∞∣f(g(n))∣∑_{n=0}^∞|f(g(n))|∑n=0∞​∣f(g(n))∣ is an upper bound of SSS, so
sup⁡⁡S≤∑n=0∞∣f(g(n))∣<+∞\sup⁡S≤∑_{n=0}^∞|f(g(n))| <+∞sup⁡S≤n=0∑∞​∣f(g(n))∣<+∞
On the contrary, if M=sup⁡{∑x∈A∣f(x)∣∶A⊆X,Afinite}<+∞M=\sup\{∑_{x∈A}|f(x)|∶A⊆X,A\text{ finite}\}<+∞M=sup{∑x∈A​∣f(x)∣∶A⊆X,A finite}<+∞, we assume the series ∑x∈Xf(x)∑_{x∈X}f(x)∑x∈X​f(x) is not absolutely convergent, then given a bijection g:N→Xg:\mathbf N→Xg:N→X, we have SN=∑n=0N∣f(g(n))∣S_N=∑_{n=0}^N|f(g(n))|SN​=∑n=0N​∣f(g(n))∣ is not bounded above, so there exists a k∈Nk∈\mathbf Nk∈N such that ∑n=0k∣f(g(n))∣>M∑_{n=0}^k|f(g(n))| >M∑n=0k​∣f(g(n))∣>M, we let A=g({n:0≤n≤k})A=g(\{n:0≤n≤k\})A=g({n:0≤n≤k}), then AAA is finite, and
∑n=0k∣f(g(n))∣=∑x∈A∣f(x)∣>M∑_{n=0}^k|f(g(n))| =∑_{x∈A}|f(x)| >Mn=0∑k​∣f(g(n))∣=x∈A∑​∣f(x)∣>M
This leads to a contradiction.

Exercise 8.2.2

Since ∑x∈Xf(x)∑_{x∈X}f(x)∑x∈X​f(x) is absolutely convergent, the quantity MMM is well defined. For ∀n∈N∀n∈\mathbf N∀n∈N, consider the set {x∈X:∣f(x)∣>1/n}\{x∈X:|f(x)|>1/n\}{x∈X:∣f(x)∣>1/n}, this set is finite with cardinality at most MnMnMn, thus use Exercise 8.1.9, we know that
{x∈X:f(x)≠0}=⋃n=1∞{x∈X:∣f(x)∣>1/n}\{x∈X:f(x)≠0\}=⋃_{n=1}^∞\{x∈X:|f(x)|>1/n\}{x∈X:f(x)​=0}=n=1⋃∞​{x∈X:∣f(x)∣>1/n}
is at most countable.

Exercise 8.2.3

Since both ∑x∈Xf(x)∑_{x∈X}f(x)∑x∈X​f(x) and ∑x∈Xg(x)∑_{x∈X}g(x)∑x∈X​g(x) are absolutely convergent, the sets S={x∈X:f(x)≠0}S=\{x∈X:f(x)≠0\}S={x∈X:f(x)​=0} and B={x∈X:g(x)≠0}B=\{x∈X:g(x)≠0\}B={x∈X:g(x)​=0} are at most countable. We denote ∑x∈Xf(x)=L∑_{x∈X}f(x)=L∑x∈X​f(x)=L and ∑x∈Xg(x)=M∑_{x∈X}g(x)=M∑x∈X​g(x)=M.
( a ) Let A⊆XA⊆XA⊆X be a finite set, then by proposition 7.1.11 we have
∑x∈A∣f(x)+g(x)∣≤∑x∈A∣f(x)∣+∑x∈A∣g(x)∣≤sup⁡⁡{∑x∈A∣f(x)∣∶A⊆X,Afinite}+sup⁡{∑x∈A∣g(x)∣∶A⊆X,Afinite}<∞\begin{aligned}∑_{x∈A} |f(x)+g(x)| &≤∑_{x∈A}|f(x)| +∑_{x∈A}|g(x)| \\&≤\sup⁡\left\{∑_{x∈A}|f(x)|∶A⊆X,A\text{ finite}\right\}+\sup \left\{∑_{x∈A}|g(x)|∶A⊆X,A\text{ finite}\right\}\\&<∞\end{aligned}x∈A∑​∣f(x)+g(x)∣​≤x∈A∑​∣f(x)∣+x∈A∑​∣g(x)∣≤sup⁡{x∈A∑​∣f(x)∣∶A⊆X,A finite}+sup{x∈A∑​∣g(x)∣∶A⊆X,A finite}<∞​
which means
sup⁡⁡{∑x∈A∣f(x)+g(x)∣∶A⊆X,Afinite}≤sup⁡⁡{∑x∈A∣f(x)∣∶A⊆X,Afinite}+sup⁡{∑x∈A∣g(x)∣∶A⊆X,Afinite}\sup⁡\left\{∑_{x∈A}|f(x)+g(x)|∶A⊆X,A\text{ finite}\right\}\\\leq\sup⁡\left\{∑_{x∈A}|f(x)|∶A⊆X,A\text{ finite}\right\}+\sup \left\{∑_{x∈A}|g(x)|∶A⊆X,A\text{ finite}\right\}sup⁡{x∈A∑​∣f(x)+g(x)∣∶A⊆X,A finite}≤sup⁡{x∈A∑​∣f(x)∣∶A⊆X,A finite}+sup{x∈A∑​∣g(x)∣∶A⊆X,A finite}
Use Definition 8.2.4 we know that ∑x∈X∣f(x)+g(x)∣∑_{x∈X}|f(x)+g(x)|∑x∈X​∣f(x)+g(x)∣ is absolutely convergent. Thus
∑x∈X∣f(x)+g(x)∣=∑x∈X,f(x)+g(x)≠0(f(x)+g(x))=∑x∈S∪B(f(x)+g(x))∑_{x∈X}|f(x)+g(x)| =∑_{x∈X,f(x)+g(x)≠0} (f(x)+g(x)) =∑_{x∈S∪B} (f(x)+g(x))x∈X∑​∣f(x)+g(x)∣=x∈X,f(x)+g(x)​=0∑​(f(x)+g(x))=x∈S∪B∑​(f(x)+g(x))
S∪BS∪BS∪B is at most countable, if S∪BS∪BS∪B is finite we can use Proposition 7.1.11(f) to get
∑x∈S∪B(f(x)+g(x))=∑x∈S∪Bf(x)+∑x∈S∪Bg(x)=∑x∈Sf(x)+∑x∈Bg(x)∑_{x∈S∪B} (f(x)+g(x)) =∑_{x∈S∪B} f(x)+∑_{x∈S∪B} g(x) =∑_{x∈S} f(x) +∑_{x∈B} g(x)x∈S∪B∑​(f(x)+g(x))=x∈S∪B∑​f(x)+x∈S∪B∑​g(x)=x∈S∑​f(x)+x∈B∑​g(x)
If S∪BS∪BS∪B is countable, we can have a bijection h:N→S∪Bh:\mathbf N→S∪Bh:N→S∪B, s.t.
∑x∈S∪B(f(x)+g(x))=∑n=0+∞(f(h(n))+g(h(n)))∑_{x∈S∪B} (f(x)+g(x)) =∑_{n=0}^{+∞}\bigg(f\Big(h(n)\Big)+g\Big(h(n)\Big)\bigg) x∈S∪B∑​(f(x)+g(x))=n=0∑+∞​(f(h(n))+g(h(n)))
Let SN=∑n=0N(f(h(n))+g(h(n)))=∑n=0Nf(h(n))+∑n=0Ng(h(n))S_N=∑_{n=0}^N\bigg(f\Big(h(n)\Big)+g\Big(h(n)\Big)\bigg) =∑_{n=0}^Nf(h(n)) +∑_{n=0}^Ng(h(n))SN​=∑n=0N​(f(h(n))+g(h(n)))=∑n=0N​f(h(n))+∑n=0N​g(h(n)) , we know that when N→∞N→∞N→∞, ∑n=0Nf(h(n))→L∑_{n=0}^Nf(h(n)) →L∑n=0N​f(h(n))→L since A⊆A∪BA⊆A∪BA⊆A∪B, by the same logic, ∑n=0Ng(h(n))→M∑_{n=0}^Ng(h(n)) →M∑n=0N​g(h(n))→M, thus the conclusion is valid.
( b ) If c=0c=0c=0 the conclusion is obviously true. Now suppose c≠0c≠0c​=0, Let A⊆XA⊆XA⊆X be a finite set, then by proposition 7.1.11 we have
∑x∈A∣cf(x)∣=∣c∣∑x∈A∣f(x)∣≤∣c∣sup⁡⁡{∑x∈A∣f(x)∣∶A⊆X,Afinite}<∞∑_{x∈A}|cf(x)| =|c| ∑_{x∈A}|f(x)| ≤|c| \sup⁡\left\{∑_{x∈A}|f(x)|∶A⊆X,A\text{ finite}\right\}<∞x∈A∑​∣cf(x)∣=∣c∣x∈A∑​∣f(x)∣≤∣c∣sup⁡{x∈A∑​∣f(x)∣∶A⊆X,A finite}<∞
Thus
sup⁡{∑x∈A∣cf(x)∣∶A⊆X,Afinite}≤∣c∣sup⁡⁡{∑x∈A∣f(x)∣∶A⊆X,Afinite}<∞\sup\left\{∑_{x∈A}|cf(x)|∶A⊆X,A\text{ finite}\right\}≤|c| \sup⁡\left\{∑_{x∈A}|f(x)|∶A⊆X,A\text{ finite}\right\}<∞sup{x∈A∑​∣cf(x)∣∶A⊆X,A finite}≤∣c∣sup⁡{x∈A∑​∣f(x)∣∶A⊆X,A finite}<∞
Use Definition 8.2.4 we know that ∑x∈Xcf(x)∑_{x∈X} cf(x)∑x∈X​cf(x) is absolutely convergent. Since we also have (f(x)≠0)⇔(cf(x)≠0)(f(x)≠0)⇔(cf(x)≠0)(f(x)​=0)⇔(cf(x)​=0), we know that
∑x∈Xcf(x)=∑x∈X,cf(x)≠0cf(x)=∑x∈Scf(x)∑_{x∈X} cf(x) = ∑_{x∈X,cf(x)≠0} cf(x) =∑_{x∈S} cf(x)x∈X∑​cf(x)=x∈X,cf(x)​=0∑​cf(x)=x∈S∑​cf(x)
If SSS is finite, use Proposition 7.1.11(g) we have ∑x∈Scf(x)=c∑x∈Sf(x)=c∑x∈Xf(x)∑_{x∈S} cf(x) =c∑_{x∈S} f(x) =c∑_{x∈X} f(x)∑x∈S​cf(x)=c∑x∈S​f(x)=c∑x∈X​f(x) . If SSS is countable, we can have a bijection h:N→Sh:\mathbf N→Sh:N→S, s.t.
∑x∈Scf(x)=∑n=0+∞cf(h(n))=c∑n=0+∞f(h(n))∑_{x∈S} cf(x) =∑_{n=0}^{+∞}cf(h(n)) =c∑_{n=0}^{+∞}f(h(n)) x∈S∑​cf(x)=n=0∑+∞​cf(h(n))=cn=0∑+∞​f(h(n))
The last equality comes from Proposition 7.2.14(b)
( c ) In the first case, let A⊆X1A⊆X_1A⊆X1​ be any finite set, then A⊆XA⊆XA⊆X, thus
∑x∈A∣f(x)∣≤sup⁡{∑x∈A∣f(x)∣∶A⊆X,Afinite}∑_{x∈A}|f(x)| ≤\sup\left\{∑_{x∈A}|f(x)|∶A⊆X,A\text{ finite}\right\}x∈A∑​∣f(x)∣≤sup{x∈A∑​∣f(x)∣∶A⊆X,A finite}
Which gives
sup⁡{∑x∈A∣f(x)∣∶A⊆X1,Afinite}≤sup⁡{∑x∈A∣f(x)∣∶A⊆X,Afinite}\sup \left\{∑_{x∈A}|f(x)|∶A⊆X_1,A\text{ finite} \right \} ≤\sup\left\{∑_{x∈A}|f(x)|∶A⊆X,A\text{ finite}\right\}sup{x∈A∑​∣f(x)∣∶A⊆X1​,A finite}≤sup{x∈A∑​∣f(x)∣∶A⊆X,A finite}
So ∑x∈X1f(x)∑_{x∈X_1}f(x)∑x∈X1​​f(x) is absolutely convergent.
By the same logic ∑x∈X2f(x)∑_{x∈X_2}f(x)∑x∈X2​​f(x) is absolutely convergent.
Conversely, if ∑x∈X1h(x)∑_{x∈X_1} h(x)∑x∈X1​​h(x) and ∑x∈X2h(x)∑_{x∈X_2} h(x)∑x∈X2​​h(x) are absolutely convergent. Then for any finite set A⊆XA⊆XA⊆X, we can have A=(A∩X1)∪(A∩X2)A=(A∩X_1 )∪(A∩X_2)A=(A∩X1​)∪(A∩X2​), as X1X_1X1​ and X2X_2X2​ are disjoint, we have
∑x∈A∣f(x)∣=∑x∈A∩X1∣f(x)∣+∑x∈A∩X2∣f(x)∣≤sup⁡{∑x∈A∣f(x)∣∶A⊆X1,Afinite}+sup⁡{∑x∈A∣f(x)∣∶A⊆X2,Afinite}\begin{aligned} ∑_{x∈A}|f(x)| &=∑_{x∈A∩X_1}|f(x)| +∑_{x∈A∩X_2}|f(x)| \\&≤\sup\left\{∑_{x∈A}|f(x)|∶A⊆X_1,A\text{ finite}\right\}+\sup\left\{∑_{x∈A}|f(x)|∶A⊆X_2,A\text{ finite}\right\}\end{aligned}x∈A∑​∣f(x)∣​=x∈A∩X1​∑​∣f(x)∣+x∈A∩X2​∑​∣f(x)∣≤sup{x∈A∑​∣f(x)∣∶A⊆X1​,A finite}+sup{x∈A∑​∣f(x)∣∶A⊆X2​,A finite}​
Then we have
sup⁡{∑x∈A∣f(x)∣∶A⊆(X1∪X2),Afinite}≤sup⁡{∑x∈A∣f(x)∣∶A⊆X1,Afinite}+sup⁡{∑x∈A∣f(x)∣∶A⊆X2,Afinite}\sup\left\{∑_{x∈A}|f(x)|∶A⊆(X_1∪X_2),A\text{ finite}\right\} \\≤\sup\left\{∑_{x∈A}|f(x)|∶A⊆X_1,A\text{ finite}\right\}+\sup\left\{∑_{x∈A}|f(x)|∶A⊆X_2,A\text{ finite}\right\}sup{x∈A∑​∣f(x)∣∶A⊆(X1​∪X2​),A finite}≤sup{x∈A∑​∣f(x)∣∶A⊆X1​,A finite}+sup{x∈A∑​∣f(x)∣∶A⊆X2​,A finite}
To prove
∑x∈X1∪X2f(x)=∑x∈X1f(x)+∑x∈X2f(x)∑_{x∈X_1∪X_2}f(x) =∑_{x∈X_1} f(x) +∑_{x∈X_2} f(x)x∈X1​∪X2​∑​f(x)=x∈X1​∑​f(x)+x∈X2​∑​f(x)
We only need to consider points where fff doesn’t vanish on, thus redefine
X={x∈X:f(x)≠0},X1={x∈X1:f(x)≠0},X2={x∈X2:f(x)≠0}X=\{x∈X:f(x)≠0\},\\X_1=\{x∈X_1:f(x)≠0\},\\X_2=\{x∈X_2:f(x)≠0\}X={x∈X:f(x)​=0},X1​={x∈X1​:f(x)​=0},X2​={x∈X2​:f(x)​=0}
we first see the case is clear when X=X1∪X2X=X_1∪X_2X=X1​∪X2​ is finite. Now if one is countable and the other is finite, we may assume X2X_2X2​ is finite, then there’s bijections g1:N→X1g_1:\mathbf N→X_1g1​:N→X1​ and g2:{i∈N:0≤i≤m}→X2g_2:\{i∈\mathbf N:0≤i≤m\}→X_2g2​:{i∈N:0≤i≤m}→X2​, such that
∑x∈X1f(x)=∑n=0∞f(g1(n)),∑x∈X2f(x)=∑n=0mf(g2(n))∑_{x∈X_1} f(x) =∑_{n=0}^∞f\Big(g_1 (n)\Big) ,\quad ∑_{x∈X_2} f(x) =∑_{n=0}^mf\Big(g_2 (n)\Big)x∈X1​∑​f(x)=n=0∑∞​f(g1​(n)),x∈X2​∑​f(x)=n=0∑m​f(g2​(n))
We define a bijection g:N→X1∪X2g:\mathbf N→X_1∪X_2g:N→X1​∪X2​ by
g(n)={g2(n),n≤mg1(n−m−1),n>mg(n)=\begin{cases}g_2 (n),&n≤m\\g_1 (n-m-1),&n>m\end{cases}g(n)={g2​(n),g1​(n−m−1),​n≤mn>m​

Then ∑x∈X1∪X2f(x)=∑n=0∞f(g(n))∑_{x∈X_1∪X_2} f(x) =∑_{n=0}^∞f\Big(g(n)\Big)∑x∈X1​∪X2​​f(x)=∑n=0∞​f(g(n)) , further we have for any N∈NN∈\mathbf NN∈N:
∑n=0N+mf(g(n))=∑n=0mf(g2(n))+∑n=0Nf(g1(n))∑_{n=0}^{N+m}f\Big(g(n)\Big) =∑_{n=0}^mf\Big(g_2 (n)\Big) +∑_{n=0}^Nf\Big(g_1 (n)\Big) n=0∑N+m​f(g(n))=n=0∑m​f(g2​(n))+n=0∑N​f(g1​(n))
Let N→+∞N→+∞N→+∞, we get the conclusion.

If both sets are countable, there’s bijections g1:N→X1g_1:\mathbf N→X_1g1​:N→X1​ and g2:N→X2g_2:\mathbf N→X_2g2​:N→X2​ such that
∑x∈X1f(x)=∑n=0∞f(g1(n)),∑x∈X2f(x)=∑n=0∞f(g2(n))∑_{x∈X_1} f(x) =∑_{n=0}^∞f(g_1 (n)) ,\quad ∑_{x∈X_2} f(x) =∑_{n=0}^∞f(g_2 (n)) x∈X1​∑​f(x)=n=0∑∞​f(g1​(n)),x∈X2​∑​f(x)=n=0∑∞​f(g2​(n))
We define a bijection g:N→X1∪X2g:\mathbf N→X_1∪X_2g:N→X1​∪X2​ by
g(n)={g1(k),n=2k,k∈Ng2(k),n=2k+1,k∈Ng(n)=\begin{cases}g_1 (k),& n=2k,k∈N \\g_2 (k),&n=2k+1,k∈N\end{cases}g(n)={g1​(k),g2​(k),​n=2k,k∈Nn=2k+1,k∈N​
Then ∑x∈X1∪X2f(x)=∑n=0∞f(g(n))∑_{x∈X_1∪X_2} f(x) =∑_{n=0}^∞f(g(n))∑x∈X1​∪X2​​f(x)=∑n=0∞​f(g(n)) , further we have for any N∈NN∈\mathbf NN∈N:
∑n=02N+1f(g(n))=∑n=0Nf(g2(n))+∑n=0Nf(g1(n))∑_{n=0}^{2N+1}f(g(n)) =∑_{n=0}^Nf(g_2 (n)) +∑_{n=0}^Nf(g_1 (n)) n=0∑2N+1​f(g(n))=n=0∑N​f(g2​(n))+n=0∑N​f(g1​(n))
Let N→+∞N→+∞N→+∞, we get the conclusion.
The statement of h(x)h(x)h(x) can then be derived by substituting fff to hhh.
( d ) First we prove ∑y∈Yf(ϕ(y))∑_{y∈Y}f(ϕ(y))∑y∈Y​f(ϕ(y)) is absolutely convergent. Choose any finite A⊆YA⊆YA⊆Y, as ϕϕϕ is a bijection, ϕ(A)⊆Xϕ(A)⊆Xϕ(A)⊆X is finite. Let

L=sup⁡⁡{∑x∈A∣f(x)∣∶A⊆X,Afinite}<∞L=\sup⁡\left\{∑_{x∈A}|f(x)|∶A⊆X,A\text{ finite}\right\}<∞L=sup⁡{x∈A∑​∣f(x)∣∶A⊆X,A finite}<∞

Then ∑y∈A∣f(ϕ(y))∣=∑ϕ(y)∈ϕ(A)∣f(ϕ(y))∣≤L∑_{y∈A}|f(ϕ(y))| =∑_{ϕ(y)∈ϕ(A)}|f(ϕ(y))| ≤L∑y∈A​∣f(ϕ(y))∣=∑ϕ(y)∈ϕ(A)​∣f(ϕ(y))∣≤L, so
sup⁡⁡{∑y∈A∣f(ϕ(y))∣∶A⊆Y,Afinite}≤L<∞\sup⁡\left\{∑_{y∈A}|f(ϕ(y))|∶A⊆Y,A\text{ finite}\right\}≤L<∞sup⁡⎩⎨⎧​y∈A∑​∣f(ϕ(y))∣∶A⊆Y,A finite⎭⎬⎫​≤L<∞

Next, to prove ∑y∈Yf(ϕ(y))=∑x∈Xf(x)∑_{y∈Y}f(ϕ(y)) =∑_{x∈X} f(x)∑y∈Y​f(ϕ(y))=∑x∈X​f(x) , we see that for any y∈Yy∈Yy∈Y such that f(ϕ(y))≠0f(ϕ(y))≠0f(ϕ(y))​=0, we can choose one x=ϕ(y)∈Xx=ϕ(y)∈Xx=ϕ(y)∈X such that f(x)≠0f(x)≠0f(x)​=0, vice versa.(In this step we use the Axiom of Choice if YYY and XXX are uncountable) Thus the sets {y∈Y:f(ϕ(y))≠0}\{y∈Y: f(ϕ(y))≠0\}{y∈Y:f(ϕ(y))​=0} and {x∈X:f(x)≠0}\{x∈X:f(x)≠0\}{x∈X:f(x)​=0} have the same cardinality, thus are both finite or countable. In the finite case the statement is obviously true by Proposition 7.1.11( c ). If both sets are countable, we let a bijection g:N→{y∈Y:f(ϕ(y))≠0}g:\mathbf N→\{y∈Y: f(ϕ(y))≠0\}g:N→{y∈Y:f(ϕ(y))​=0}, then ϕ∘g:N→{x∈X:f(x)≠0}ϕ∘g:\mathbf N→\{x∈X:f(x)≠0\}ϕ∘g:N→{x∈X:f(x)​=0} is also a bijection, and we have

∑y∈Yf(ϕ(y))=∑n=0∞f(ϕ(g(n)))∑x∈Xf(x)=∑n=0∞f(ϕ∘g(n))=∑n=0∞f(ϕ(g(n)))∑_{y∈Y}f(ϕ(y)) =∑_{n=0}^∞f(ϕ(g(n))) \\ ∑_{x∈X} f(x) =∑_{n=0}^∞f(ϕ∘g(n)) =∑_{n=0}^∞f(ϕ(g(n))) y∈Y∑​f(ϕ(y))=n=0∑∞​f(ϕ(g(n)))x∈X∑​f(x)=n=0∑∞​f(ϕ∘g(n))=n=0∑∞​f(ϕ(g(n)))
Then we have ∑y∈Yf(ϕ(y))=∑x∈Xf(x)∑_{y∈Y}f(ϕ(y)) =∑_{x∈X} f(x)∑y∈Y​f(ϕ(y))=∑x∈X​f(x) .

Exercise 8.2.4

Let ∑n=0∞an=L∑_{n=0}^∞a_n =L∑n=0∞​an​=L, we also know that ∑n=0∞∣an∣∑_{n=0}^∞|a_n |∑n=0∞​∣an​∣ doesn’t exist. Now assume the conclusion is wrong, then one of the three assertions below must be true:

  • i. ∑n∈A+an∑_{n∈A^+}a_n∑n∈A+​an​ converges, but ∑n∈A−an∑_{n∈A^-}a_n∑n∈A−​an​ is not conditionally convergent.
  • ii. ∑n∈A−an∑_{n∈A^-}a_n∑n∈A−​an​ converges, but ∑n∈A+an∑_{n∈A^+}a_n∑n∈A+​an​ is not conditionally convergent.
  • iii. both ∑n∈A+an∑_{n∈A^+}a_n∑n∈A+​an​ and ∑n∈A−an∑_{n∈A^-}a_n∑n∈A−​an​ converges.

In case i and ii, we know that ∑n=0∞an=∑n∈A+an+∑n∈A−an∑_{n=0}^∞a_n =∑_{n∈A^+}a_n +∑_{n∈A^-}a_n∑n=0∞​an​=∑n∈A+​an​+∑n∈A−​an​ , so we can have ∑n∈A−an=∑n=0∞an−∑n∈A+an∑_{n∈A^-}a_n =∑_{n=0}^∞a_n -∑_{n∈A^+}a_n∑n∈A−​an​=∑n=0∞​an​−∑n∈A+​an​ in case i and ∑n∈A+an=∑n=0∞an−∑n∈A−an∑_{n∈A^+}a_n =∑_{n=0}^∞a_n -∑_{n∈A^-}a_n∑n∈A+​an​=∑n=0∞​an​−∑n∈A−​an​ in case ii, both are contradiction since the left side is not convergent but the right side is convergent.
In case iii, we let ∑n∈A+an=L1∑_{n∈A^+}a_n =L_1∑n∈A+​an​=L1​ and ∑n∈A−an=L2∑_{n∈A^-}a_n =L_2∑n∈A−​an​=L2​, thus we have:
∑n=0∞∣an∣=∑n∈A+∣an∣+∑n∈A−∣an∣=∑n∈A+an−∑n∈A−an=L1−L2∑_{n=0}^∞|a_n |=∑_{n∈A^+}|a_n |+∑_{n∈A^-}|a_n |=∑_{n∈A^+}a_n -∑_{n∈A^-}a_n =L_1-L_2 n=0∑∞​∣an​∣=n∈A+∑​∣an​∣+n∈A−∑​∣an​∣=n∈A+∑​an​−n∈A−∑​an​=L1​−L2​
Which is a contradiction to the fact that ∑n=0∞∣an∣∑_{n=0}^∞|a_n |∑n=0∞​∣an​∣ doesn’t exist. Thus the original conclusion must be true.

Exercise 8.2.5

Why1: A+A^+A+ and A−A^-A− are infinite.
Assume not, then one of A+A^+A+ and A−A^-A− are finite, then either ∑n∈A+an∑_{n∈A^+}a_n∑n∈A+​an​ or ∑n∈A−an∑_{n∈A^-}a_n∑n∈A−​an​ is a finite series, thus must be convergent, a contradiction.
Why2: ∑m=0∞af+(m)∑_{m=0}^∞a_{f_+ (m)}∑m=0∞​af+​(m)​ and ∑m=0∞af−(m)∑_{m=0}^∞a_{f_- (m)}∑m=0∞​af−​(m)​ are not absolutely convergent.
Since af+(m)≥0,∀ma_{f_+ (m)}≥0,∀maf+​(m)​≥0,∀m and af−(m)<0,∀ma_{f_- (m)}<0,∀maf−​(m)​<0,∀m, if we assume ∑m=0∞af+(m)∑_{m=0}^∞a_{f_+ (m)}∑m=0∞​af+​(m)​ or ∑m=0∞af−(m)∑_{m=0}^∞a_{f_- (m)}∑m=0∞​af−​(m)​ is absolutely convergent, then we can know from Proposition 7.4.3 that ∑m=0∞af+(m)=∑n∈A+an∑_{m=0}^∞a_{f_+ (m)} =∑_{n∈A^+}a_n∑m=0∞​af+​(m)​=∑n∈A+​an​ or ∑m=0∞af−(m)=∑n∈A−an∑_{m=0}^∞a_{f_- (m)} =∑_{n∈A^-}a_n∑m=0∞​af−​(m)​=∑n∈A−​an​ , in particular either ∑n∈A+an∑_{n∈A^+}a_n∑n∈A+​an​ or ∑n∈A−an∑_{n∈A^-}a_n∑n∈A−​an​ is absolute convergent, which is a contradiction.
Why3: The map j→njj→n_jj→nj​ is injective.
Let k≠jk≠jk​=j, then either k<jk<jk<j or j<kj<kj<k, if k<jk<jk<j, the definition of njn_jnj​ guarantees that nj≠nkn_j≠n_knj​​=nk​. If k<jk<jk<j we can similarly prove nk≠njn_k≠n_jnk​​=nj​, so this map is injective.
Why4: Both Case I and II occur an infinite number of times.
Assume Case I only occurs finite number of times, this means for some M∈NM∈\mathbf NM∈N, we’ll have
∑0≤i<jani≥L,∀j>M∑_{0≤i<j}a_{n_i } ≥L,\quad ∀j>M0≤i<j∑​ani​​≥L,∀j>M
Thus we have ∑0≤i≤Mani+∑M<iani≥L∑_{0≤i≤M}a_{n_i} +∑_{M<i}a_{n_i} ≥L∑0≤i≤M​ani​​+∑M<i​ani​​≥L, or ∑M<iani≥L−∑0≤i≤Mani∑_{M<i}a_{n_i } ≥L-∑_{0≤i≤M}a_{n_i}∑M<i​ani​​≥L−∑0≤i≤M​ani​​ , notice that as iii increases, ∑M<iani∑_{M<i}a_{n_i}∑M<i​ani​​ decreases since all ani<0a_{n_i}<0ani​​<0, so ∑M<iani∑_{M<i}a_{n_i}∑M<i​ani​​ converges, which means ∑n∈A−an∑_{n∈A^-}a_n∑n∈A−​an​ converges, a contradiction.
When assume Case II occurs finite number of times, we can prove a contradiction similarly.

Why5: The map j→njj→n_jj→nj​ is surjective.
For ∀n∈N∀n∈\mathbf N∀n∈N, either n∈A+n∈A^+n∈A+ or n∈A−n∈A^-n∈A−. Assume no jjj can make nj=nn_j=nnj​=n, this means either Case I or Case II stops at nnn, so can only occur finite number of times, which is a contradiction.
Why6: lim⁡j→∞anj=0\lim_{j→∞} a_{n_j}=0limj→∞​anj​​=0
From Corollary 7.2.6 we have lim⁡n→∞an=0\lim_{n→∞} a_{n}=0limn→∞​an​=0, since both Case I and II occur an infinite number of times, when j→∞j→∞j→∞ we must have nj→∞n_j→∞nj​→∞.

Why7: lim⁡j→∞∑0≤i<jani=L\lim_{j→∞} ∑_{0≤i<j}a_{n_i} =Llimj→∞​∑0≤i<j​ani​​=L
From Why6 we know that for ∀ϵ>0,∃J∈N∀ϵ>0,∃J∈\mathbf N∀ϵ>0,∃J∈N,s.t.∣anj∣<ϵ,(j>J)|a_{n_j} |<ϵ,(j>J)∣anj​​∣<ϵ,(j>J). Start from nJn_JnJ​, we search for a number K>JK>JK>J such that ∑0≤i<Kani<L∑_{0≤i<K}a_{n_i} <L∑0≤i<K​ani​​<L and ∑0≤i<K+1ani≥L∑_{0≤i<K+1}a_{n_i} ≥L∑0≤i<K+1​ani​​≥L, this KKK must exist, otherwise we’ll get a contradiction with Why4.
We denote Sk=∑0≤i<kaniS_k=∑_{0≤i<k}a_{n_i}Sk​=∑0≤i<k​ani​​ , notice that:
If Sk<LS_k<LSk​<L, we shall add positive ania_{n_i}ani​​ to SkS_kSk​, the positive adding shall stop once Sk≥LS_k≥LSk​≥L, so Sk≤L+∣ani∣S_k≤L+|a_{n_i} |Sk​≤L+∣ani​​∣, if k>K>Jk>K>Jk>K>J, we can have Sk≤L+ϵS_k≤L+ϵSk​≤L+ϵ.
If Sk≥LS_k≥LSk​≥L, we shall add negative ania_{n_i}ani​​ to SkS_kSk​, the negative adding shall stop once Sk<LS_k<LSk​<L, so Sk>L−∣ani∣S_k>L-|a_{n_i} |Sk​>L−∣ani​​∣, if k>K>Jk>K>Jk>K>J, we can have Sk>L−ϵS_k>L-ϵSk​>L−ϵ.
Combined we get ∣Sk−L∣<ϵ,k>K|S_k-L|<ϵ,k>K∣Sk​−L∣<ϵ,k>K, this means lim⁡k→∞Sk=L\lim_{k→∞} S_k=Llimk→∞​Sk​=L.

Exercise 8.2.6

We only have to prove lim inf⁡N→∞∑m=N∞af(m)=+∞\liminf_{N→∞}∑_{m=N}^∞a_f(m) =+∞N→∞liminf​∑m=N∞​af​(m)=+∞. With the same definition of A+A^+A+ and A−A^-A− in Lemma 8.2.7, we find increasing bijections f+:N→A+f_+:N→A^+f+​:N→A+ and f−:N→A−f_-:N→A^-f−​:N→A−. We define a sequence n0,n1,n2,…n_0,n_1,n_2,…n0​,n1​,n2​,… of natural numbers recursively as below:
Let n0=f+(0),n1=f−(0)n_0=f_+ (0),n_1=f_- (0)n0​=f+​(0),n1​=f−​(0). Suppose that j≥2j≥2j≥2 is a natural number, and nin_ini​ has already be defined for all i<ji<ji<j. Then we let
k=max⁡{n:f−(n)∈{ni:i<j}}k=\max \Big\{n:f_- (n)∈\{n_i:i<j\}\Big\}k=max{n:f−​(n)∈{ni​:i<j}}
For example, if j=2j=2j=2, then n0n_0n0​ and n1n_1n1​ has been defined, then k=1k=1k=1. Now
(I) If ∑0≤i<jani<k∑_{0≤i<j}a_{n_i} <k∑0≤i<j​ani​​<k, we set nj=min⁡{n∈A+:n≠nifor all i<j}n_j=\min\{n∈A^+:n≠n_i\text{ for all }i<j\}nj​=min{n∈A+:n​=ni​ for all i<j}
(II) If ∑0≤i<jani≥k∑_{0≤i<j}a_{n_i}≥k∑0≤i<j​ani​​≥k, we set nj=min⁡{n∈A−:n≠nifor all i<j}n_j=\min\{n∈A^-:n≠n_i\text{ for all }i<j\}nj​=min{n∈A−:n​=ni​ for all i<j}
Intuitively, we compare ∑0≤i<jani∑_{0≤i<j}a_{n_i}∑0≤i<j​ani​​ with an increasing KKK, which increases by 111 once we add an negative number to the sum. From the proof of Exercise 8.2.5 we can see that

  • A+A^+A+ and A−A^-A− are infinite.

  • ∑m=0∞af+(m)∑_{m=0}^∞a_{f_+ (m)}∑m=0∞​af+​(m)​ and ∑m=0∞af−(m)∑_{m=0}^∞a_{f_- (m)}∑m=0∞​af−​(m)​ are not absolutely convergent.

  • The map j→njj→n_jj→nj​ is injective.

And we can prove the following results:

  • Both Case I and II occur an infinite number of times.

Assume Case I occur only finite number of times, then from some JJJ and KKK we’ll get ∑0≤i<Jani≥K∑_{0≤i<J}a_{n_i} ≥K∑0≤i<J​ani​​≥K and Case II happens infinitely, but as we add negative numbers to ∑0≤i<Jani∑_{0≤i<J}a_{n_i}∑0≤i<J​ani​​ , we’ll have ∑0≤i<jani≤∑0≤i<Jani∑_{0≤i<j}a_{n_i} ≤∑_{0≤i<J}a_{n_i}∑0≤i<j​ani​​≤∑0≤i<J​ani​​ , and K→∞K→∞K→∞, this will lead to a contradiction since ∑0≤i<Jani∑_{0≤i<J}a_{n_i}∑0≤i<J​ani​​ is finite.
Assume Case II occur only finite number of times, then from some JJJ and KKK we’ll get ∑0≤i<Jani<K∑_{0≤i<J}a_{n_i} <K∑0≤i<J​ani​​<K, and ∑0≤i<jani<K,∀j>J∑_{0≤i<j}a_{n_i} <K,∀j>J∑0≤i<j​ani​​<K,∀j>J, this means ∑m=0∞af+(m)∑_{m=0}^∞a_{f_+ (m)}∑m=0∞​af+​(m)​ is convergent, a contradiction.

  • The map j→njj→n_jj→nj​ is surjective.

  • lim⁡j→∞anj=0\lim_{j→∞} a_{n_j}=0limj→∞​anj​​=0.

The proofs are similar to that of Exercise 8.2.5.
Finally we shall prove lim⁡j→∞∑0≤i<jani>M\lim_{j→∞} ∑_{0≤i<j}a_{n_i} >Mlimj→∞​∑0≤i<j​ani​​>M for any M>0M>0M>0, thus complete the proof. For any M>0M>0M>0, we can find a natural number K>MK>MK>M, and as lim⁡j→∞anj=0\lim_{j→∞} a_{n_j}=0limj→∞​anj​​=0, there’s JJJ such that

∣anj∣<1,∀j≥J|a_{n_j} |<1,\quad ∀j≥J∣anj​​∣<1,∀j≥J
We’ll eventually find a j′>Jj'>Jj′>J such that ∑0≤i<j′ani≥K+1∑_{0≤i<j'}a_{n_i} ≥K+1∑0≤i<j′​ani​​≥K+1, and adding negative ania_{n_i}ani​​ to the sum would not let the new sum falls below KKK, once the sum is between KKK and K+1K+1K+1, we would continue to add positive ania_{n_i}ani​​ and push the sum to K+2K+2K+2 and even larger, thus we have ∑0≤i<jani≥K>M,∀j>j′∑_{0≤i<j}a_{n_i} ≥K>M,∀j>j'∑0≤i<j​ani​​≥K>M,∀j>j′, and the conclusion follows.

陶哲轩实分析(上)8.2及习题-Analysis I 8.2相关推荐

  1. 陶哲轩实分析命题10.1.7

    设$X$是$\mathbf{R}$的子集合,$x_0$是$X$的极限点,设$f:X\to\mathbf{R}$是函数,并设$L$是实数,则下述命题在逻辑上等价: (a):$f$在$x_0$处在$X$上 ...

  2. 陶哲轩实分析习题9.1.1

    设$X$是实直线的子集合,并设$Y$是集合,使得$X\subseteq Y\subseteq \overline{X}$,证明$\overline{Y}=\overline{X}$. 证明:因为$X\ ...

  3. 陶哲轩实分析引理10.4.1:反函数定理

    设$f:X\to Y$是可逆函数,反函数为$f^{-1}:Y\to X$.设$x_0\in X,y_0\in Y$,且$y_0=f(x_0)$(它蕴含$x_0=f^{-1}(y_0)$).如果$f$在 ...

  4. 陶哲轩实分析:微积分基本定理剖析

    本文致力于深入探讨微积分基本原理的本质 我们基于严格的实数理论.序列极限理论.级数理论.函数理论.导数理论建立起来了黎曼积分体系.我们苦苦追寻思考,最终发现了黎曼积分的最终控制函数.正是因为我们发现了 ...

  5. 陶哲轩实分析:数学家对于R和Q集可数性的探究

    本文意淫了康托尔是如何证明了R是不可数集的,还意淫了某位不知名数学家是如何证明Q是可数集的.本文较细致的分析了Q和R可数性证明过程. 于是我便在1873年完成了轰动世界的证明!没错,就是我证明了R是不 ...

  6. 陶哲轩实分析定理17.3.8(三)

    本文继承了这篇博文. 为了证明$f$在$x_0$处可微,我们只用证明,存在线性映射$T$,使得 \begin{equation} \lim_{x'\to x_0;x'\neq x_0}\frac{f( ...

  7. 陶哲轩实分析 定理 8.2.2 (无限和的富比尼定理) 证明

    设$f:\mathbb{N}\times\mathbb{N}\to\mathbb{R}$是函数.使得$$\sum_{(n,m)\in\mathbb{N}\times\mathbb{N}}f(n,m)$ ...

  8. 陶哲轩实分析习题17.1.2

    陶哲轩实分析习题17.1.2 转载于:https://www.cnblogs.com/yeluqing/archive/2012/09/10/3828300.html

  9. 《陶哲轩实分析》部分勘误

    我在读<陶哲轩实分析>,作者是陶哲轩,译者王昆扬.2008年11月第一版,第一次印刷.我在此添加一部分中译本印刷错误,若网友发现了另外的错误,请在评论里补充,由我代为添加.若有不当之处,敬 ...

  10. 陶哲轩实分析 命题 8.2.6 证明

    设$X$是任意的集合(可以是不可数的),并设$f:X\to \mathbb{R}$和$g:X\to\mathbb{R}$是函数.使得$\sum_{x\in X}f(x)$和$\sum_{x\in X} ...

最新文章

  1. Linux下快速分区格式化大于2T大容量存储
  2. python 帮助文档,撰写函数文档,并查看函数文档
  3. jeecms 2012 源码分析(2) 前台栏目页静态化分析
  4. codefirst数据库迁移
  5. python一个图画两条曲线_用python建立两个Y轴的XY曲线图方法
  6. java控制台计算数字_java从控制台接收一个数字
  7. Qt5.9绘制文字(drawText函数)用法
  8. fc oracle,使用 CLI 配置 FC
  9. 基于多域连接卷积神经网络的精神分裂症脑功能网络分类
  10. mysql 取出全部数据库_php取出mysql数据库中所有数据
  11. freemarker的测试结果框架_java必背综合知识点总结(框架篇)
  12. bat随机打开目录下的一个音乐文件_Tomcat bin目录详解
  13. 在Windows下安装和配置Node.js环境v8.11.3与遇到的问题
  14. 前端框架VUE的基础使用
  15. activiti启动流程实例
  16. 不要在锁的作用域之外通过指针或引用传递要保护的数据
  17. 大数据技术 - 学习之路(一)
  18. 深度置信网络(Deep belief network)matlab初解
  19. vue使用contenteditable 实现光标处插入自定义图片
  20. 不知道如何raw转换成jpg格式?可以试试这几个软件

热门文章

  1. Java之父詹姆斯•高斯林趣闻
  2. springboot整合阿里云oss上传文件(图片或视频)
  3. 怎样修改电脑时间同步北京时间
  4. 两台android相互ADB实现一台安卓手机给另一台安卓手机ADB
  5. 我们和蹦迪博主聊了聊,哪家大厂人在夜店最受欢迎?
  6. C语言正则表达式详解 regcomp() regexec() regfree()详解
  7. 设计模式之工厂模式(简单工厂、工厂方法、抽象工厂)
  8. 雷电9模拟器安装magisk和lsposed
  9. OLD:用于单变量和多变量异常值检测的SPSS宏
  10. linux+开机启动sshd_Linux sshd服务自动启动