Complete metric space

Sequentially compact space

A topological space 𝑋 is said to be sequentially compact iff every sequence (π‘₯𝑛)βˆžπ‘›=1 in 𝑋 has a convergent subsequence with a limit in 𝑋. #m/def/topology

In general, sequential compactness is neither weaker nor stronger than compactness. However, the Main theorem describes when these conditions are equivalent.

Main theorem

Let 𝑋 be a second-countable topological space, e.g. a Metric space. Then 𝑋 is compact iff it is sequentially compact. #m/thm/topology

Proof

Let 𝑋 be a first-countable Compact space, and (π‘₯𝑛)βˆžπ‘›=1 βˆˆπ‘‹ be a sequence with end pieces π‘€π‘š ={π‘šπ‘›}βˆžπ‘›=π‘š. The intersection of finite end pieces will always be inhabited, since π‘š β‰₯𝑛 ⟹ π‘€π‘š βŠ†π‘€π‘›, so by Complement characterisation β‹‚βˆžπ‘›=1𝑀𝑛 β‰ βˆ… and thus the intersection of closures β‹‚βˆžπ‘›=1𝑀𝑛 β‰ βˆ…. Since Limit points are points contained in the closure of every end piece, π‘₯𝑛 has at least one limit point, and since Limit points are limits of convergent subsequences in a first-countable space, π‘₯𝑛 has a convergent subsequence. Therefore 𝑋 is sequentially compact.

For the converse, let 𝑋 be a second-countable sequentially compact space. Second countable implies LindelΓΆf, so without loss of generality we can take a countable open cover (π‘ˆπ‘›)βˆžπ‘›=1 βŠ†π‘‹. Let 𝑉𝑛 =⋃𝑛𝑖=1π‘ˆπ‘– be partial unions of the open cover. Assume π‘ˆπ‘› has no finite subcover, so 𝑋 βˆ–π‘‰π‘› β‰ βˆ… for all 𝑛 βˆˆβ„•, so we can construct a sequence (π‘₯𝑛)βˆžπ‘›=1 βˆˆπ‘‹ with π‘₯𝑛 βˆ‰π‘‰π‘›. Since 𝑋 is sequentially compact, π‘₯𝑛 has a convergent subsequence, and the limit thereof is a limit point of π‘₯𝑛 because Limit points are limits of convergent subsequences in a first-countable space. Therefore let π‘Ž βˆˆπ‘‹ be a Limit point of π‘₯𝑛. Since π‘ˆπ‘› covers 𝑋, π‘Ž βˆˆπ‘ˆπ‘š for some π‘š βˆˆβ„•, so both π‘ˆπ‘š and π‘‰π‘š are open neighbourhoods of π‘Ž. Since π‘Ž is a limit point of π‘₯𝑛, its neighbourhood π‘‰π‘š contains infinite π‘₯𝑛, which contradicts our construction. Therefore π‘ˆπ‘› must have a finite subcover.

Note the forward statement only requires the First countability axiom, whereas the converse requires both first-countability and LindelΓΆf.

Properties


#state/tidy | #lang/en | #SemBr

Footnotes

  1. Since at least one element must be repeated infinitely many times in a sequence by the Pigeonhole principle, yielding an eventually constant subsequence. ↩

  2. Closedness follows from the fact that it must be sequentially closed (since subsequences of a convergent sequence converge to the same limit). Boundedness is trivial, since an unbounded set contains divergent sequences. ↩