Chapter V · Limits of functions of a continuous variable. Continuous and discontinuous functions
The Heine–Borel theorem and uniform continuity
From infinitely many local facts to one global fact; covering a closed interval by finitely many intervals, and why the oscillation of a continuous function can be made small uniformly.
Why this matters
Continuity is a local property: near each point, separately, the function varies little. Many theorems need a global statement: one step size that works everywhere at once. The bridge from local to global is the Heine–Borel theorem of §106: if every point of a closed interval has an interval of its own around it, then finitely many of those intervals already cover the whole. In §107 Hardy uses it to prove that a continuous function on has small oscillation on every sufficiently short piece. That fact, today called uniform continuity, is exactly what the definite integral of Chapter VII needs.
The idea
Suppose each point of comes with an open interval . Infinitely many intervals certainly cover , since each point is in its own. Can we throw all but finitely many away? If the radii were all at least some fixed , yes: the intervals round would do. The danger is that shrinks towards somewhere, forcing ever more intervals.
Hardy’s second proof measures this danger with a function. For each let be the largest such that the window fits inside one interval of the cover. He shows that is continuous; on a closed interval it then has a positive least value (last lesson). Any window shorter than , placed anywhere, fits inside one interval of the cover, and finitely many such windows reach across .
Try this. With the first cover, drag the purple window along the bottom and change its length . Find the largest for which every position fits. Then choose the wavy cover and do the same. Finally choose on .
- ρ(x)
- δ(x)
- minimum m of ρ
- window of length l (drag)
- interval of the cover containing it; below: the finite sub-cover
What to notice
A window fits everywhere exactly when , where is the minimum of the blue curve ; then the green rows below show a finite sub-cover, one interval per window. For the wavy cover is small, so many intervals are needed, but still finitely many. For , sinks to at the left end: no positive works near , and no finite sub-cover exists. The interval is not closed, and the point has no interval of its own.
Precise statements
Hardy states the theorem for an arbitrary set of intervals inside , finite or infinite, overlapping or not.
Let be a set of intervals, each contained in , such that
- every point of other than and lies inside (not at an end of) at least one interval of , and
- is the left-hand end of at least one interval of , and the right-hand end of at least one.
Then a finite number of the intervals of already have properties 1 and 2.
The heart of the proof is a lemma about a single number .
Under the hypotheses of the theorem there is a positive number such that every sub-interval of of length is contained in at least one interval of .
The lemma gives the theorem at once: cover by finitely many overlapping sub-intervals of length (starting at , each shifted by , the last ending at ), and for each choose one interval of containing it.
Now fix and a function continuous on . By continuity, each point has an interval round it in which the oscillation of is less than (Hardy, whose tolerance is rather than , calls these the -intervals of ). Heine–Borel selects finitely many of them.
If is continuous on and , then can be divided into a finite number of sub-intervals in each of which the oscillation of is less than .
If is continuous on and , there is a such that, if is divided in any manner into sub-intervals of length less than , the oscillation of in each of them is less than .
Theorem II says: whenever are in and , with one serving every point. In ordinary continuity may depend on the point.
Try this. With the continuous function, increase the number of pieces until every piece is green, for ; then halve . Repeat with , which is continuous on . Then choose , and also .
- y = φ(x)
- piece with oscillation < ε
- piece with oscillation ≥ ε
What to notice
For both continuous functions a large enough makes every piece green, for any : that is Theorem II (and needs more pieces near , where it wiggles, but only finitely many). For the pieces touching always have oscillation , and for the pieces ending at an integer always have oscillation : at a discontinuity no partition helps.
Proofs
Hardy gives two proofs of the lemma. Here is the first, which uses Weierstrass’s theorem on points of accumulation (§19).
Argue by contradiction. If no works, there are bad windows of every length . Their left ends have a point of accumulation . But lies inside an interval of , which must then contain the small bad windows near .
Suppose no such exists
Then for each there is a sub-interval of that is contained in no interval of .
Take a point of accumulation of the
The numbers lie in , so by Weierstrass’s theorem they have a point of accumulation in (if only finitely many distinct values occur, one of them occurs for infinitely many ; take that one as ).
lies well inside some interval of
If , by hypothesis 1 there is an interval of with . If , hypothesis 2 gives an interval of with (and the are ); similarly if . Let be the distance from to the nearer end that matters.
A bad window falls inside it
Choose with and (possible for infinitely many , since is a point of accumulation). Then lies inside that interval of , contrary to its choice. So exists.
The second proof is the one the covering game draws: is continuous because shifting by shrinks the best window by at most , so ; its minimum on a closed sub-interval is positive; and any works there. (Hardy says “less than ”, not “”, because the largest window round a point need not itself fit.) A third proof is by bisection: if could not be covered by finitely many intervals of , neither could one of its halves, nor one of that half’s halves, and so on; the nested halves close on a point , which lies inside one interval of , and that one interval covers all the small halves near .
Try this. Press “Halve” repeatedly. Each orange piece is not yet inside a single interval of the cover and is halved again; green pieces are done.
- piece inside one interval of the cover
- piece still to be halved
What to notice
After a few rounds every piece is green: finitely many pieces, each inside one interval, so finitely many intervals cover . The process must stop, because if it did not, the orange pieces would nest down to a point , and the interval belonging to would swallow them.
Collect, for every point, the intervals round it on which the oscillation is less than . They satisfy the hypotheses of Heine–Borel, so finitely many suffice; their end points cut into the required pieces.
Each point has -intervals
If , continuity at gives with on . All values there lie within of , so the oscillation is at most . At and use one-sided intervals and .
Apply Heine–Borel
The set of all these intervals satisfies hypotheses 1 and 2, so finitely many of them, say , also cover in that sense.
Cut at all their end points
The end points of that lie in divide it into finitely many pieces. Each piece lies inside one (it lies between consecutive end points, so no starts or stops within it, and some covers its interior), and oscillation can only decrease on a sub-interval (§104, (2)). So each piece has oscillation less than .
Apply Theorem I with (strictly, with some ) and let be the length of the shortest piece. A short interval can straddle at most two neighbouring pieces, and the oscillation over two neighbouring pieces is at most the sum of theirs.
Pieces from Theorem I
Choose and divide into finitely many pieces , each of oscillation less than . Let be the length of the shortest; because there are finitely many.
A short interval meets at most two pieces
An interval of length less than cannot contain a whole piece in its interior, so it lies within the union of at most two successive pieces .
Add the oscillations
By property (3) of §104, the oscillation on is at most that on plus that on , which is less than ; on the short interval it is no larger.
Worked examples and exercises
(i) Divide into the closed intervals . Do they satisfy the hypotheses of Heine–Borel? (ii) Give each in the interval cut down to lie in , with . Find a finite subset with properties 1 and 2.
Hint
Solution
(i) No: the points lie at the ends of intervals and inside none, so hypothesis 1 fails. (The conclusion is trivially true, as the set is already finite; the example shows what “inside” means.)
(ii) Hardy’s choice: the intervals belonging to , namely , ending with one whose right end is (for the last, use the interval of the point , or of itself, which is ). Consecutive intervals overlap by , so every interior point is inside one of them; is the left end of the first and the right end of the last.
For on and a given , find a as in Theorem II.
Hint
Solution
On any sub-interval of of length less than , two points satisfy . So the oscillation is at most , and works (the oscillation is then , since the sup and inf over a closed sub-interval are values). The bound comes from the right-hand end; on alone, would do. One for the whole interval is possible because the slope is bounded there.
is continuous at every point of . Show that the conclusion of Theorem II fails on for . Which step of the proof breaks?
Hint
Solution
Given any , choose an integer with . The points and both lie in , so they are less than apart, and , . So a division of that has the interval (of length less than ) as one of its pieces has a piece with oscillation . As was arbitrary, no works.
The proof breaks at the very start: Heine–Borel needs every point of a closed interval to have an -interval, including the end point , but is not even defined at , and no value there would make it continuous. The -intervals of points near shrink to nothing.
Give each rational in (lowest terms, ) the interval , cut down to . Show that if some points of lie in none of these intervals. You may use the fact that finitely many open intervals covering have total length greater than .
Hint
Hint
Solution
Suppose every point of lies inside (or, for and , at the appropriate end of) one of the intervals. By Heine–Borel finitely many of them already do, and these have total length greater than .
But the total length of any finite collection is at most the sum over all denominators. For there are two intervals (at and ), each cut to length : total . For there are at most intervals, of total length at most . So any finite collection has total length less than
using and . This contradicts “greater than ”. Hence some points are covered by no interval. (Those points are irrational numbers that are badly approximable by fractions.) Hardy stresses that the shortcut “the total length is less than , so they cannot cover” is exactly what needs Heine–Borel when there are infinitely many intervals.
Check yourself
Which family of open intervals covers its set but has no finite sub-family that does?
Finitely many of the intervals have a largest , say , and miss . Heine–Borel does not apply because is not closed. In the other options the set is closed and every radius is at least , so finitely many intervals suffice, as the theorem promises.
For on the whole line, there is one such that whenever .
False. Take : then once . The line is not a closed bounded interval, and the slope of is unbounded on it. On any , Theorem II does give such a .
What is the difference between continuity on and the conclusion of Theorem II?
Continuity: for each and there is . Theorem II: for each there is working at all points at once. On a closed interval the first implies the second; that is the content of the theorem, and it needs Heine–Borel.
Hardy’s language
It is impossible, without the use of this or some similar theorem, to prove that a function continuous throughout an interval necessarily possesses an integral over that interval.
Summary
- Heine–Borel: if every point of lies inside some interval of a family (end points at an end), finitely many of the intervals do the same job. The interval must be closed.
- Key lemma: one window length fits inside some interval of the family wherever it is placed (Hardy proves it by Weierstrass’s theorem, or by minimising the continuous function ).
- For continuous on : finitely many pieces of oscillation (Theorem I), and even one length such that every piece shorter than has oscillation (Theorem II, uniform continuity).
- Both fail off closed intervals ( on , on the whole line) and at discontinuities.