Stolz–Cesàro theorem

From testwiki
Revision as of 10:56, 29 January 2025 by 2001:67c:10ec:578b:8000::2f1 (talk) (External links: Updated the link to point to the website that actually contains the content now.)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Template:Short description In mathematics, the Stolz–Cesàro theorem is a criterion for proving the convergence of a sequence. It is named after mathematicians Otto Stolz and Ernesto Cesàro, who stated and proved it for the first time.

The Stolz–Cesàro theorem can be viewed as a generalization of the Cesàro mean, but also as a l'Hôpital's rule for sequences.

Statement of the theorem for the Template:Math case

Let (an)n1 and (bn)n1 be two sequences of real numbers. Assume that (bn)n1 is a strictly monotone and divergent sequence (i.e. strictly increasing and approaching +, or strictly decreasing and approaching ) and the following limit exists:

limnan+1anbn+1bn=l. 

Then, the limit

limnanbn=l. 

Statement of the theorem for the Template:Math case

Let (an)n1 and (bn)n1 be two sequences of real numbers. Assume now that (an)0 and (bn)0 while (bn)n1 is strictly decreasing. If

limnan+1anbn+1bn=l, 

then

limnanbn=l. [1]

Proofs

Proof of the theorem for the Template:Math case

Case 1: suppose (bn) strictly increasing and divergent to +, and <l<. By hypothesis, we have that for all ϵ/2>0 there exists ν>0 such that n>ν

|an+1anbn+1bnl|<ϵ2,

which is to say

lϵ/2<an+1anbn+1bn<l+ϵ/2,n>ν.

Since (bn) is strictly increasing, bn+1bn>0, and the following holds

(lϵ/2)(bn+1bn)<an+1an<(l+ϵ/2)(bn+1bn),n>ν.

Next we notice that

an=[(anan1)++(aν+2aν+1)]+aν+1

thus, by applying the above inequality to each of the terms in the square brackets, we obtain

(lϵ/2)(bnbν+1)+aν+1=(lϵ/2)[(bnbn1)++(bν+2bν+1)]+aν+1<anan<(l+ϵ/2)[(bnbn1)++(bν+2bν+1)]+aν+1=(l+ϵ/2)(bnbν+1)+aν+1.

Now, since bn+ as n, there is an n0>0 such that bn>0 for all n>n0, and we can divide the two inequalities by bn for all n>max{ν,n0}

(lϵ/2)+aν+1bν+1(lϵ/2)bn<anbn<(l+ϵ/2)+aν+1bν+1(l+ϵ/2)bn.

The two sequences (which are only defined for n>n0 as there could be an Nn0 such that bN=0)

cn±:=aν+1bν+1(l±ϵ/2)bn

are infinitesimal since bn+ and the numerator is a constant number, hence for all ϵ/2>0 there exists n±>n0>0, such that

|cn+|<ϵ/2,n>n+,|cn|<ϵ/2,n>n,

therefore

lϵ<lϵ/2+cn<anbn<l+ϵ/2+cn+<l+ϵ,n>max{ν,n±}=:N>0,

which concludes the proof. The case with (bn) strictly decreasing and divergent to , and l< is similar.

Case 2: we assume (bn) strictly increasing and divergent to +, and l=+. Proceeding as before, for all 2M>0 there exists ν>0 such that for all n>ν

an+1anbn+1bn>2M.

Again, by applying the above inequality to each of the terms inside the square brackets we obtain

an>2M(bnbν+1)+aν+1,n>ν,

and

anbn>2M+aν+12Mbν+1bn,n>max{ν,n0}.

The sequence (cn)n>n0 defined by

cn:=aν+12Mbν+1bn

is infinitesimal, thus

M>0n¯>n0>0 such that M<cn<M,n>n¯,

combining this inequality with the previous one we conclude

anbn>2M+cn>M,n>max{ν,n¯}=:N.

The proofs of the other cases with (bn) strictly increasing or decreasing and approaching + or respectively and l=± all proceed in this same way.

Proof of the theorem for the Template:Math case

Case 1: we first consider the case with l< and (bn) strictly decreasing. This time, for each ν>0, we can write

an=(anan+1)++(an+ν1an+ν)+an+ν,

and for any ϵ/2>0, n0 such that for all n>n0 we have

(lϵ/2)(bnbn+ν)+an+ν=(lϵ/2)[(bnbn+1)++(bn+ν1bn+ν)]+an+ν<anan<(l+ϵ/2)[(bnbn+1)++(bn+ν1bn+ν)]+an+ν=(l+ϵ/2)(bnbn+ν)+an+ν.

The two sequences

cν±:=an+νbn+ν(l±ϵ/2)bn

are infinitesimal since by hypothesis an+ν,bn+ν0 as ν, thus for all ϵ/2>0 there are ν±>0 such that

|cν+|<ϵ/2,ν>ν+,|cν|<ϵ/2,ν>ν,

thus, choosing ν appropriately (which is to say, taking the limit with respect to ν) we obtain

lϵ<lϵ/2+cν<anbn<l+ϵ/2+cν+<l+ϵ,n>n0

which concludes the proof.

Case 2: we assume l=+ and (bn) strictly decreasing. For all 2M>0 there exists n0>0 such that for all n>n0,

an+1anbn+1bn>2Manan+1>2M(bnbn+1).

Therefore, for each ν>0,

anbn>2M+an+ν2Mbn+νbn,n>n0.

The sequence

cν:=an+ν2Mbn+νbn

converges to 0 (keeping n fixed). Hence

M>0ν¯>0 such that M<cν<M,ν>ν¯,

and, choosing ν conveniently, we conclude the proof

anbn>2M+cν>M,n>n0.

Applications and examples

The theorem concerning the Template:Math case has a few notable consequences which are useful in the computation of limits.

Arithmetic mean

Let (xn) be a sequence of real numbers which converges to l, define

an:=m=1nxm=x1++xn,bn:=n

then (bn) is strictly increasing and diverges to +. We compute

limnan+1anbn+1bn=limnxn+1=limnxn=l

therefore

limnx1++xnn=limnxn.

Given any sequence

(xn)n1

of real numbers, suppose that

limnxn

exists (finite or infinite), then

limnx1++xnn=limnxn.

Geometric mean

Let (xn) be a sequence of positive real numbers converging to l and define

an:=log(x1xn),bn:=n,

again we compute

limnan+1anbn+1bn=limnlog(x1xn+1x1xn)=limnlog(xn+1)=limnlog(xn)=log(l),

where we used the fact that the logarithm is continuous. Thus

limnlog(x1xn)n=limnlog((x1xn)1n)=log(l),

since the logarithm is both continuous and injective we can conclude that

limnx1xnn=limnxn.

Given any sequence

(xn)n1

of (strictly) positive real numbers, suppose that

limnxn

exists (finite or infinite), then

limnx1xnn=limnxn.

Suppose we are given a sequence (yn)n1 and we are asked to compute

limnynn,

defining y0=1 and xn=yn/yn1 we obtain

limnx1xnn=limny1yny0y1yn1n=limnynn,

if we apply the property above

limnynn=limnxn=limnynyn1.

This last form is usually the most useful to compute limits

Given any sequence

(yn)n1

of (strictly) positive real numbers, suppose that

limnyn+1yn

exists (finite or infinite), then

limnynn=limnyn+1yn.

Examples

Example 1

limnnn=limnn+1n=1.

Example 2

limnn!nn=limn(n+1)!(nn)n!(n+1)n+1=limnnn(n+1)n=limn1(1+1n)n=1e

where we used the representation of e as the limit of a sequence.

History

The ∞/∞ case is stated and proved on pages 173—175 of Stolz's 1885 book and also on page 54 of Cesàro's 1888 article.

It appears as Problem 70 in Pólya and Szegő (1925).

The general form

Statement

The general form of the Stolz–Cesàro theorem is the following:[2] If (an)n1 and (bn)n1 are two sequences such that (bn)n1 is monotone and unbounded, then:

lim infnan+1anbn+1bnlim infnanbnlim supnanbnlim supnan+1anbn+1bn.

Proof

Instead of proving the previous statement, we shall prove a slightly different one; first we introduce a notation: let (an)n1 be any sequence, its partial sum will be denoted by An:=m1nam. The equivalent statement we shall prove is:

Let

(an)n1,(bn)1

be any two sequences of real numbers such that

  • bn>0,n>0,
  • limnBn=+,

then

lim infnanbnlim infnAnBnlim supnAnBnlim supnanbn.

Proof of the equivalent statement

First we notice that:

  • lim infnAnBnlim supnAnBn holds by definition of limit superior and limit inferior;
  • lim infnanbnlim infnAnBn holds if and only if lim supnAnBnlim supnanbn because lim infnxn=lim supn(xn) for any sequence (xn)n1.

Therefore we need only to show that lim supnAnBnlim supnanbn. If L:=lim supnanbn=+ there is nothing to prove, hence we can assume L<+ (it can be either finite or ). By definition of lim sup, for all l>L there is a natural number ν>0 such that

anbn<l,n>ν.

We can use this inequality so as to write

An=Aν+aν+1++an<Aν+l(BnBν),n>ν,

Because bn>0, we also have Bn>0 and we can divide by Bn to get

AnBn<AνlBνBn+l,n>ν.

Since Bn+ as n+, the sequence

AνlBνBn0 as n+ (keeping ν fixed),

and we obtain

lim supnAnBnl,l>L,

By definition of least upper bound, this precisely means that

lim supnAnBnL=lim supnanbn,

and we are done.

Proof of the original statement

Now, take (an),(bn) as in the statement of the general form of the Stolz-Cesàro theorem and define

α1=a1,αk=akak1,k>1β1=b1,βk=bkbk1k>1

since (bn) is strictly monotone (we can assume strictly increasing for example), βn>0 for all n and since bn+ also Bn=b1+(b2b1)++(bnbn1)=bn+, thus we can apply the theorem we have just proved to (αn),(βn) (and their partial sums (An),(Bn))

lim supnanbn=lim supnAnBnlim supnαnβn=lim supnanan1bnbn1,

which is exactly what we wanted to prove.

References

Notes

Template:PlanetMath attribution