Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am across two definitions for unconditional convergence for which it is not immediately obvious to me that they are equivalent. Here are the definitions. Throughout, $\frak{X}$ will denote a Banach space.

Definition 1. Given a series $\sum x_n$ in $\frak{X}$, we say that this series converges unconditionally to $x$ if for every $\varepsilon>0$, there is a finite subset $J\subseteq \mathbb{N}$ such that for every finite subset $I$ such that $J\subseteq I\subseteq\mathbb{N}$ one has $\|x - \sum_{i\in I}x_i\|<\varepsilon$.

Definition 2. A series $\sum x_n$ in $\frak{X}$ is said to converge unconditionally to $x$ if for any permutation $\sigma:\mathbb{N}\to \mathbb{N}$, the series $\sum x_{\sigma(n)}$ converges to $x$.

I was wondering how one might go about proving that these two definitions are equivalent.

share|cite|improve this question
up vote 4 down vote accepted

This is proved in detail in:

T.H. Hildebrandt, On unconditional convergence in normed vector spaces, Bull. Amer. Math. Soc. 46 (12) (1940), 959–962, MR0003448.

In fact, Hildebrandt proves the equivalence of five properties A–E. Your definition 1 is his condition E and your definition 2 is his condition A.

Since Hildebrandt proves the equivalence of A and E (both directions) on page 960 and the paper is freely accessible it makes little sense to reproduce the clean and clear argument here.

Let me mention that condition A goes back to

W. Orlicz, Über unbedingte Konvergenz in Funktionenräumen (I), Studia Math. 4 (1933), 33–37.

while (according to Hildebrandt) condition E was studied by Moore in

E.H. Moore, General analysis, Memoirs of the American Philosophical Society, vol. 1, part 2, 1939, p. 63.

share|cite|improve this answer
Hey Theo, thanks so much for the reference, it was very helpful! – Deon Sep 5 '11 at 23:56

1 implies 2

Let consider that $\sum x_n$ converge unconditionally to $x$ according to definition 1. Let $\sigma$ be an arbitrary permutation of $\mathbb{N}$, now by definition 1 given $\varepsilon >0$, there is a finite subset $J_\varepsilon\subset\mathbb{N}$ such that for every finite subset $I$ such that $J_\varepsilon$ is contained in it $\|x - \sum_{i\in I}x_i\|<\varepsilon$.

Now, given $\varepsilon > 0$ take $N_\varepsilon\in\mathbb{N}$ such that $$J_\varepsilon\subset \{\sigma(n)\,|\,n\leq N_\varepsilon\}$$ which can be easily calculated by $N_\varepsilon=\max \sigma^{-1}J$. Now, we have that for every $m\geq N_\varepsilon$, $$\{\sigma(n)\,|\,n\leq m\}$$ contains $J_\varepsilon$ and it is finite. So $$\|x - \sum_{i\in \{\sigma(n)\,|\,n\leq m\}}x_i\|<\varepsilon$$ i.e. $$\|x-\sum_{n\leq m}x_{\sigma(n)}\|<\varepsilon$$ Hence, $$\sum x_{\sigma(n)}$$ converge to $x$ by definition, because for every $\varepsilon>0$ there is a $N_\varepsilon$ such that for every $m\geq N_\varepsilon$, $\|x-\sum_{n\leq m}x_{\sigma(n)}\|<\varepsilon$.

2 implies 1

For the other sense, we will use counterimplication. Suppose that $\sum x_n$ does not converges inconditionally to $x$ acording definition 1, then we have that there is a $\varepsilon>0$ such that for every finite set $J\subset\mathbb{N}$ there exist a bigger finite subset $I_J$ containing it such that $$\|x - \sum_{i\in I}x_i\|\geq\varepsilon$$ Let, thanks to the axiom of choice $g$ be the function that for each finite sets $J$ selects the finite set $J_I$ with the spetial property. And define a seguence of sets as follows by the recursion theorem $$A_0=\{0\}$$ $$A_{2n+1}=g(A_{2n})$$ $$A_{2n+2}=\{n\in\mathbb{N}\,|\,n\leq (\max A_{2n+1}+1)\}$$ We can see that each $A_i$ is finite, and that in the end they cover all $\mathbb{N}$, we only have to apply induction.

Now, we can define a permutation $\sigma:\mathbb{N}\rightarrow\mathbb{N}$ such that covers the $A_i$ one by one in order, which makes that $\sigma$ has as images from some inital segments of $\mathbb{N}$ sets of the form $g(A_{2n})$. $\sum x_{\sigma(n)}$ does not converge to $x$. Since for the existenting $\varepsilon$ we have that given any $N\in\mathbb{N}$ there will be an $M_N\in\mathbb{N}$ such that $$\sigma(\{n\leq M_n\})=g(A_{2i})$$ for some $i$. And so, $$\|\sum_{n\leq M_n}x_{\sigma(n)}-x\|\geq \varepsilon$$ as desired.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.