Scribd
Upload
115 views

AbelSummation PDF

The document discusses Abel summation and Tauber's theorem. Abel summation provides a technique to calculate the sum of an infinite series by instead computing the limit as x approaches 1 of the series where each term is multiplied by x^k. Tauber's theorem shows that if a series is Abel summable and its terms approach 0 as 1/n, then the series is also convergent in the usual sense. The document proves Tauber's theorem by comparing the partial sums of the series to the Abel summation of the series.

Uploaded by

Hariharan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
115 views

AbelSummation PDF

The document discusses Abel summation and Tauber's theorem. Abel summation provides a technique to calculate the sum of an infinite series by instead computing the limit as x approaches 1 of the series where each term is multiplied by x^k. Tauber's theorem shows that if a series is Abel summable and its terms approach 0 as 1/n, then the series is also convergent in the usual sense. The document proves Tauber's theorem by comparing the partial sums of the series to the Abel summation of the series.

Uploaded by

Hariharan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Abel Summation

written by Oleg Ivrii

We begin our discussion with a problem to illustrate this technique:

Problem. Compute the infinite sum S = 1 − 1/3 + 1/5 − 1/7 + . . .

Solution. Let f (x) = x − x3 /3 + x5 /5 − x7 /7 + . . . , so S = f (1). Differentiating,


we compute f 0 (x) = 1 − x2 + x4 − x6 + · · · = 1+x
1
2 . Integrating from 0 to x, we find

that f (x) = arctan x, so S = f (1) = π/4. ( Recall that arctan x is the function from
R x dx
R → (−π/2, π/2) inverse of tan and equals to 0 1+x 2. )

The above computation is not entirely rigorous, for we have actually shown that f (x) =
arctan x for |x| < 1 (the radius of convergence of f (x) is 1, so we can only differentiate
term-by-term when |x| < 1).

The idea of Abel summation is the following: to find the sum ∞


P
k=0 ak , we instead
P∞ k
compute k=0 ak x and take limit x → 1. Of course to justify this technique, we should
show that if ∞
P P∞ k
k=0 ak tends to L, then limx→1 k=0 ak x also tends to L. This is known
as Abel’s theorem and will be proved next section.

So, in the above example, as S converges in the usual sense, Abel’s technique gives you
the answer you expect.

Abel summation is more powerful than usual summation as it can sum divergent series
1
like 1 − 1 + 1 − 1 + . . . , e.g 1 − x + x2 − x3 + · · · = 1+x and taking limit x → 1, we see that
this is 1/2.

Abel’s Theorem
Suppose that ∞
P P∞ k
k=0 ak converges to L. We show that f (x) = k=0 ak x converges at x = 1
and to L. It suffices to see that f (x) converges uniformly on [0, 1].

Let Sn = nk=0 ak . Thus we must show that ∞ k


P P
k=n ak x is small when n is sufficiently
large. Indeed, k=n ak xk = k=n (Sk+1 − Sk )xk = −Sn xn−1 + ∞
P∞ P ∞ P k−1 − xk ). What
k=n Sk (x
is important here is that the Sk are bounded and the xk decrease to 0. In literature, this is
known as Abel’s test for uniform convergence.
Tauber’s Theorem
An interesting question is the following: suppose a series ∞
P
k=0 ak is Abel summable. Under
P∞
what conditions on ak , does it follow that k=0 ak is summable in the usual way? Tauber
showed the condition an = o(1/n) is sufficient. We prove this.

Let Sn = nk=0 ak . We need to show that Sn − L → 0 as n → ∞. For this purpose, we


P

compare Sn and L to f (x) = ∞ k


P
k=0 ak x where x is close to 1:

|Sn − L| ≤ |Sn − f (x)| + |f (x) − L|


Xn X∞
k
≤ ak − ak x + |f (x) − L|

k=0 k=0
Xn X ∞
k
≤ ak (1 − x ) + ak xk + |f (x) − L|

k=0 k=n+1

The third term will be small (less than ) provided x will be close to 1.

The second term can be bounded by


∞
1 X 
|kak |xk ≤
n n(1 − x)
k=n+1

For the first term, we use the estimate (1 − xk ) = (1 − x)(1 + x + · · · + xk−1 ) ≤ k(1 − x).
Thus we get the bound (1 − x) · nk=0 |kak |. Tauber’s condition gives us that n1 nk=0 |kak |
P P

goes to 0, so for n sufficiently large, n1 nk=0 |kak | can be made less than . Thus we have
P

bounded the first term by n(1 − x).

We arrive at the estimate



|Sn − L| = n(1 − x) + + .
n(1 − x)

To balance out the first two terms, it is fair to take x = 1 − 1/n. Thus for n sufficiently
large, we have the estimate |Sn − L| ≤ 3 and the proof is complete.

You might also like