Metrically Cauchy vs. Topologically Cauchy

179 Views Asked by At

Let $X$ be a topological vector space. A family $x_i \in X, i\in I,$ is called a topologically Cauchy family if for every NH $U$ of zero, there exists $i_0\in I$ such that $$ i,j\geqslant i_0 \Rightarrow x_i-x_j \in U $$ If the topology on $X$ is induced by a metric $d$, then the family is called metrically Cauchy if for every $\varepsilon >0$, there exists $i_0\in I$ such that $$ i,j\geqslant i_0 \Rightarrow d(x_i,x_j)<\varepsilon. $$

Allegedly, neither implies the other in general, which I'm having a hard time understanding. The notions are equivalent if the metric is shift invariant, e.g a norm.

Suppose $\tau _X$ is induced by a metric $d$. Let $x_i\in X, i\in I,$ be metrically Cauchy. Let $B := B(0,\varepsilon)$ be a NH of zero. We can find $i_0\in I$ such that the metric condition is satisfied. But if the implication "Metrically Cauchy $\Rightarrow$ Topologically Cauchy" fails in general, that must mean $x_i-x_j\in B$ is not necessarily true for some $i,j\geqslant i_0$.

Conversely, if $x_i$ is topologically Cauchy and $\varepsilon >0$, then $x_i-x_j\in B(0,\varepsilon)$, but then again $d(x_i,x_j)< \varepsilon$ has to fail sometimes.

  1. What are some examples of non-shift-invariant metrics and families (sequences, even?) for which either implication fails?
  2. Is shift invariance necessary for the notions to be equivalent?
  3. Are there such topological vector spaces (whose topology is induced by a metric) such that exactly one implication is true?
1

There are 1 best solutions below

1
On BEST ANSWER

Topologically Cauchy does not imply Metrically Cauchy.

Take an irrational number, say $e$. Define on $\mathbb Q$ $$ d(p,q) = \left\lvert \frac{1}{p-e} - \frac{1}{q-e} \right\rvert. $$ $d$ is not shift invariant. Note that $\tau _d$ is the subspace topology of the standard topology on $\mathbb R$. Put $$ x_n = \left ( 1+ \frac{1}{n} \right )^n, n\in\mathbb N $$ Then $x_n\to e$ in $\mathbb R$ so it is topologically Cauchy in $\mathbb Q$. On the other hand, given any $N\in\mathbb N$ one can find $m,n\geqslant N$ such that $d(x_m,x_n)$ is large.


For the converse, we can take in the same metric space $y_n = n,\ n\in\mathbb N$. Then $d(y_m,y_n)$ is eventually small, but clearly $y_n$ is not Topologically Cauchy.