Chapter IV · Limits of functions of a positive integral variable
Oscillating sequences
A sequence that neither tends to a limit nor to plus or minus infinity oscillates, finitely if it is bounded and infinitely if not.
Why this matters
The last lesson defined three definite behaviours: tending to a limit , to , or to . Most sequences you can write down do none of these. §62 gives them a name, oscillation, and splits them into two kinds according to whether they stay within fixed bounds. The result is a complete classification: every sequence behaves in exactly one of five ways, and much of the rest of the chapter asks which of the five a given sequence falls into.
The idea
jumps between and for ever. It certainly does not tend to or , since it never exceeds in size. Does it tend to some limit ? Try to find one.
Try this. Keep . Move the candidate limit anywhere you like, and look for an after which no point is orange.
- band l ± δ
- l
- n₀ (drag)
- n ≥ n₀, good
- n ≥ n₀, bad
What to notice
With near the odd terms escape, with near the even ones, and with in between both do. A band of total width cannot contain two points at distance apart, whatever is.
So does none of the three things. Hardy’s definition of oscillation is simply the name for this: it is defined negatively, as the failure of all three. That makes oscillation a large and varied class. It includes sequences whose values repeat in a cycle, like or , sequences that never repeat but settle near two values, like , and sequences whose size grows without bound, like .
The last example suggests the split. Some oscillating sequences stay in a fixed range ; others do not. Classify the gallery yourself.
Try this. For each sequence, choose one of the five verdicts before reading the reason. Use “show the largest so far” and the number of terms shown. Look hardest at numbers 3, 6, 7 and 13.
- φ(n)
- ± largest |φ(n)| so far
What to notice
The dashed lines show the largest size so far. If they keep moving outwards as you show more terms, the sequence is unbounded (it tends to or oscillates infinitely); if they stop, it is bounded (it tends to a limit or oscillates finitely). Number 6 goes backwards at every odd step yet tends to ; number 13 changes sign at every step yet tends to .
Precise statements
A sequence oscillates (as ) if it does not tend to a limit, does not tend to and does not tend to .
An oscillating sequence oscillates finitely if there is a number such that for every , and oscillates infinitely if there is no such .
A sequence with for all is called bounded. So the five behaviours are:
| behaviour | bounded? | example |
|---|---|---|
| tends to a limit | yes | |
| oscillates finitely | yes | |
| tends to | no | |
| tends to | no | |
| oscillates infinitely | no |
The table claims that boundedness splits the five behaviours exactly as shown. Two facts are needed.
If tends to a limit , there is a number with for every .
If or , there is no with for every .
A bounded sequence either tends to a limit or oscillates finitely. An unbounded sequence either tends to , tends to , or oscillates infinitely. Every sequence does exactly one of the five.
Proofs
The tail is trapped in a band around of width ; the finitely many terms before it are bounded by their largest size. Take the larger of the two bounds.
Use the definition with
There is an with for all . For those , .
Deal with the first few terms
The terms are finitely many, so one of them, say of size , is largest in absolute value (take if ).
Combine
Put . Every is either below , when , or at least , when .
Any proposed bound is itself a value of that the sequence must eventually exceed.
Suppose for every
If , apply Definition II with : for large , , contradicting .
The case
If , then for large , , so : again a contradiction.
The corollary now follows by sorting: a bounded sequence cannot tend to , so if it does not converge it oscillates, and then finitely; an unbounded one cannot converge, so it tends to or oscillates, and then infinitely. The five cases exclude one another because a sequence cannot have two different limits (an exercise below), and cannot tend both to a limit and to (bounded versus unbounded).
Finally, the example we started with, done properly.
It is bounded, so by the classification we only need to show that it has no limit. If it had a limit , then from some point on both and would be within of , which is impossible.
Bounded
for every , so works.
Suppose ; take
There is an with for all . Among the there are even ones and odd ones.
Contradiction
An even gives and an odd one gives . Then , which is false.
Worked examples and exercises
Classify: (a) ; (b) ; (c) ; (d) ; (e) .
Solution
(a) Values : bounded, no limit (as for ), so oscillates finitely. (b) for all , and for large the values are near (even ) or (odd ), so there is no limit: oscillates finitely, despite the huge early values. (c) Unbounded, and the odd terms are eventually negative while the even ones are eventually huge, so it neither tends to nor to : oscillates infinitely. (d) Tends to : it is below once . (e) For the size is unbounded and the sign alternates: oscillates infinitely.
Show that tends to , although for every .
Hint
Solution
Since , we have . Given , if (for ; any works otherwise) then . So for large . On the other hand and , which is smaller. The sequence takes a step back at every odd , but “oscillates” is defined by what a sequence does not do, and this one does tend to .
Show that a sequence cannot tend to two different limits and .
Hint
Solution
Suppose and with , and let . For large , , and for large , ; so for large both hold at once. For such an , , a contradiction.
Is this true? “If oscillates, and for arbitrarily large we have both and , then oscillates.” Prove it or give a counterexample.
Solution
False. Take , which oscillates, and . For every odd , , and for every even , , so both happen for arbitrarily large . But . Crossing an oscillating sequence infinitely often says nothing about one’s own behaviour.
Find the least such that for all . Then do the same with replaced by .
Hint
Solution
Even : means . Odd : means . So fails (its term is exactly ), while (term ) and every later succeed. The least is .
In general, write . Even need , odd need , that is . Working through the parities as above gives if is odd and if is even, which is Hardy’s answer . Check: , even, .
Let be a real number that is not an integer. Show that oscillates finitely. (Use the rules for limits of sums and products from Limits of sums, products and quotients.)
Hint
Hint
Solution
Put . Since is not an integer, . The sequence is bounded by , so we must only rule out a limit. Suppose . The shifted sequences and also tend to (shifting does not change a limit), so
and dividing by the constant , . Now . The left side tends to (a shift of ) and the right side tends to . So , and . But then , while it equals for every . This contradiction shows that there is no limit: oscillates finitely. (For integral every term is .)
Check yourself
A sequence “oscillates” in Hardy’s sense when:
Oscillation is defined negatively. Going up and down is not enough: goes up and down and tends to . Changing sign is not enough: changes sign and tends to . Repeating in a cycle is one way to oscillate, but oscillates without ever repeating.
A bounded sequence either tends to a limit or oscillates finitely.
True. A bounded sequence cannot tend to or (those are unbounded), so if it has no limit it oscillates, and its oscillation is finite because it is bounded.
Which sequence oscillates finitely?
for every (the largest is at ), and for large the values are near , so it oscillates and is bounded. The first and third are unbounded (infinite oscillation), and the last tends to .
If the first million terms of a sequence include values larger than , the sequence cannot oscillate finitely.
False. Finite oscillation asks only for some bound , however large; finitely many huge terms just make huge. starts enormous and oscillates finitely.
Hardy’s language
Oscillation is defined in a negative manner: a function oscillates when it does not do certain other things.
Summary
- A sequence oscillates if it tends neither to a limit nor to nor to ; the definition is purely negative.
- It oscillates finitely if some bounds every , and infinitely otherwise.
- Convergent sequences are bounded; sequences tending to are not. So bounded sequences converge or oscillate finitely, unbounded ones tend to or oscillate infinitely.
- Every sequence does exactly one of the five things. Going backwards now and then, or changing sign, does not by itself make a sequence oscillate.
- To show there is no limit, suppose exists and choose small enough to produce a contradiction, as for with .