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

Let $(X,+)$ be an abelian group and $d$ a metric on $X$. Suppose $\{a_n\}$ and $\{b_n\}$ are Cauchy sequences. What conditions on the relation between the group operation and the metric are sufficient to show that $\{a_n+b_n\}$ is Cauchy? The proof in $\mathbb{R}$ simply relies on the fact that the Euclidean metric is induced by a norm: $$d(a_m+b_m,a_n+b_n)=\|(a_m+b_m)-(a_n+b_n)\|\leq\|a_m-a_n\|+\|b_m-b_n\|$$ I believe this proof works so long as the triangle inequality relates $+$ and the norm which induces $d$.

I think a weaker assumption which still works is if the metric is translation invariant. Then,

$$\begin{align*} d(a_m+b_m,a_n+b_n) &\leq d(a_m+b_m,a_n+b_m) + d(a_n+b_m,a_n+b_n) \\ &= d(a_m,a_n) + d(b_m,b_n) \end{align*}$$

Are there weaker assumptions under which the sum is Cauchy? What is an example in this case? Is there an example of a space and Cauchy sequences whose sum is not Cauchy?

This is probably not a standard way of phrasing the question. I'm just trying to strip down assumptions to clarify my understanding. I thought about this for a while but was having trouble coming up with examples.

share|cite|improve this question
The natural requirement is simply that $+$ be continuous with respect to $d$. – Brian M. Scott Nov 23 '11 at 6:14
No, continuity of $+$ is not enough. For example, consider $\mathbb R$ with the metric $d(x,y) = |\arctan(x) - \arctan y|$. Then $a_n = n$ and $b_n = (-1)^n - n$ are Cauchy but $a_n + b_n = (-1)^n$ is not Cauchy. – Robert Israel Nov 23 '11 at 7:05
Ok, so that's why I was having trouble proving that comment! Artan is even Lipschitz continuous, huh? Nice example! And that metric doesn't seem to be translation invariant, so it's not induced by a norm. This should be promoted to full answer. – dls Nov 23 '11 at 7:11
What about a space which isn't translation invariant, but preserves the Cauchy property under addition? – dls Nov 23 '11 at 7:26
This question is really a precursor to the question "when is the completion of a topological group also a topological group"?… – dls Nov 27 '11 at 18:03
up vote 6 down vote accepted

A sufficient condition weaker than translation invariance is that translations are uniformly equicontinuous, i.e that for every $\epsilon > 0$ there is $\delta > 0$ such that for all $x, y, z$, $d(x,y) < \delta$ implies $d(x+z, y+z) < \epsilon$. For then if $\{a_n\}$ and $\{b_n\}$ are Cauchy sequences, take $N$ so for $n, m > N$, $d(a_n, a_m) < \delta$ and $d(b_n, b_m) < \delta$, and note that $d(a_n + b_n, a_m + b_m) \le d(a_n + b_n, a_m + b_n) + d(a_m + b_n, a_m + b_m) < 2 \epsilon$.

share|cite|improve this answer
I ran into this while trying to prove the comment that regular continuity was sufficient; just define away the issue! I'm having trouble thinking of an example here which isn't translation invariant. But this brings up a question, related to the arctan example. If $d$ is induced by a norm, and $f : \mathbb{R} \rightarrow \mathbb{R}$, can we impose a growth-condition on $f$ so that $f \circ d$ is a metric? Can we refine this condition to say $f \circ d$ preserves Cauchy sequences under $+$? Maybe should make this own question unless answer is "no". – dls Nov 23 '11 at 16:25
Looking at the Wikipedia page for uniform space, they give the example $d(x,y)=|e^x-e^y|$ and state that this has a different uniform structure than the usual metric. Though I don't understand "uniform structure" yet, maybe I'm thinking, can you put a condition on the uniform structure which guarantees preservation of Cauchy sequences under addition? – dls Nov 23 '11 at 18:12
Instead of $f \circ d$, it's better to work with $d \circ (f \otimes f)$, i.e. $(x,y) \to d(f(x),f(y))$. For any one-to-one function from $X$ to a metric space $Y$ with metric $d$, $d \circ (f \otimes f)$ is a metric on $X$. This will typically not be translation-invariant. In the case $X=Y=\mathbb R$ with $d(x,y)=|x−y|$, it will satisfy my condition if there are constants $a$, $b$ with $0<a<f′(x)<b$ everywhere. – Robert Israel Nov 23 '11 at 19:58
Right, I see why $f \circ d$ is incorrect. So in the case $X=Y=\mathbb{R}$, we can think of level sets of the graph of $d$ over $\mathbb{R}^2$. If $d$ is translation invariant, all level sets are parallel to the line $y=x$. In the $\arctan$ example, this is clearly not the case... the level sets diverge away from this line as $x\rightarrow\pm\infty$. I think your condition says that the level sets are bounded between two lines parallel to the line $y=x$. Haven't actually tried computing this, but I think I've satisfied my curiosity for today! – dls Nov 23 '11 at 21:18

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.