Power Series and Summation by Parts

A power series converges if the absolute value of the variable is less than the radius of convergence, determined by the limit supremum of the coefficients; the sum of a product series can be computed using partial sums and their differences; if the partial sums of a series are bounded and the coefficients are non-increasing and converge to zero, the product series converges; an alternating, absolutely decreasing, and converging-to-zero sequence leads to a convergent series; and if the radius of convergence of a power series is 1, the coefficients are non-increasing and converge to zero, the series converges for all points on the unit circle except possibly at 1.

Some authorUpdated 3 min read

Theorem\textbf{Theorem} Given the power series Σcnzn\Sigma c_n z^n, put

α=lim sup⁡n→∞∣cn∣n,R=1α.\alpha=\limsup _{n \rightarrow \infty} \sqrt[n]{\left|c_n\right|}, \quad R=\frac{1}{\alpha} .

(If α=0,R=+∞\alpha=0, R=+\infty; if α=+∞,R=0\alpha=+\infty, R=0.) Then Σcnzn\Sigma c_n z^n converges if ∣z∣<R|z|<R, and diverges if ∣z∣>R|z|>R.

Proof:

Let L=lim sup⁡n→∞(∣cnzn∣)1/n=∣z∣lim sup⁡n→∞∣cn∣1/n=∣z∣αL= \limsup_{n \to \infty} \left(|c_nz^n| \right)^{1/n}= |z| \limsup_{n \to \infty} |c_n|^{1/n} = |z|\alpha

If L<1L<1, then ∣z∣<α|z| < \alpha, converges

If L>1L>1, then ∣z∣>α|z| > \alpha, diverges

and we can't conclude anything when L=1L =1.

Example\textbf{Example} we can show that series ∑znΓ(n+1)\sum \frac{z^n}{\Gamma(n+1)} has R=+∞R = +\infty.

First note that

lim sup⁡n→∞(∣zn/Γ(n+1)∣)1/n=∣z∣lim sup⁡n→∞Γ(n+1)−1/n\limsup_{n \to \infty } \left(|z^n/\Gamma(n+1)|\right)^{1/n} = |z|\limsup_{n \to \infty } \Gamma(n+1)^{-1/n} =∣z∣lim sup⁡n→∞((2πn)1/2(n/e)n)−1/n=∣z∣lim⁡n→∞(2πn)−12n(n/e)−1=0= |z| \limsup_{n \to \infty }((2\pi n)^{1/2} (n/e)^n)^{-1/n} = |z| \lim_{n \to \infty }(2\pi n )^{-\frac{1}{2n}}(n/e)^{-1} = 0

Theorem\textbf{Theorem} Given two sequences {an},{bn}\left\{a_n\right\},\left\{b_n\right\}, put

An=∑k=0nakA_n=\sum_{k=0}^n a_k

if n≥0n \geq 0; put A−1=0A_{-1}=0. Then, if 0≤p≤q0 \leq p \leq q, we have

∑n=pqanbn=∑n=pq−1An(bn−bn+1)+Aqbq−Ap−1bp\sum_{n=p}^q a_n b_n=\sum_{n=p}^{q-1} A_n\left(b_n-b_{n+1}\right)+A_q b_q-A_{p-1} b_p

.

Proof:

Consider the following:

∑n=pqanbn=∑n=pq(An−An−1)bn\sum_{n=p}^q a_n b_n = \sum_{n=p}^q (A_n - A_{n-1}) b_n =∑n=pqAnbn−∑n=p−1q−1Anbn+1= \sum_{n=p}^q A_n b_n - \sum_{n=p-1}^{q-1} A_n b_{n+1} =∑n=pq−1An(bn−bn+1)+Aqbq−Ap−1bp= \sum_{n=p}^{q-1} A_n (b_n - b_{n+1}) + A_{q}b_q -A_{p-1}b_{p}

Theorem\textbf{Theorem} Suppose

(a) the partial sums AnA_n of Σan\Sigma a_n form a bounded sequence;

(b) b0≥b1≥b2≥⋯b_0 \geq b_1 \geq b_2 \geq \cdots;

(c) lim⁡n→∞bn=0\lim _{n \rightarrow \infty} b_n=0.

Then Σanbn\Sigma a_n b_n converges.

Proof:

Since the partial sum AnA_n of series ∑an\sum a_n form. a bounded sequence, we will have ∣An∣<M|A_n| < M for all nn. Choose NN big enouch so that ∣bn∣<ϵ2M|b_n| < \frac{\epsilon}{2M} for all n>Nn > N.

Then we will have the following for the partial sum ∑i=pqanbn\sum_{i=p}^q a_nb_n for p,q>np,q > n.

∣∑i=pqanbn∣=∣∑i=pq−1An(bn−bn+1)+Aqbq−Ap−1bp∣≤M∣∑i=pq−1(bn−bn+1)+bq+bp∣=2M∣bq∣<ϵ\left|\sum_{i=p}^q a_nb_n\right| = \left|\sum_{i=p}^{q-1} A_n(b_{n}- b_{n+1}) +A_{q}b_{q}-A_{p-1}b_{p}\right| \le M \left|\sum_{i=p}^{q-1} (b_{n}- b_{n+1}) +b_{q}+b_{p}\right| =2 M \left|b_q \right|< \epsilon

So it follows from Cauchy criteria.

Comment\text{Comment}. Alternating, absolutely decreasing and limiting to zero sequence has convergent series.

Theorem.\text{Theorem.} Suppose the radius of convergence of Σcnzn\Sigma c_n z^n is 1 , and suppose c0≥c1≥c2≥⋯ ,lim⁡n→∞cn=0c_0 \geq c_1 \geq c_2 \geq \cdots, \lim _{n \rightarrow \infty} c_n=0. Then Σcnzn\Sigma c_n z^n converges at every point on the circle ∣z∣=1|z|=1, except possibly at z=1z=1.

Proof. take an=zna_n = z^n and bn=cnb_n = c_n as of the above theorem. Then we only need partial ∑zn\sum z^n to be bounded.

Indeed, we have ∣∑i=0kzn∣=∣1−zk+11−z∣≤1+∣z∣k+1∣1−z∣=2∣1−z∣|\sum_{i=0}^kz^n| = \left|\frac{1-z^{k+1}}{1-z}\right| \le \frac{1 + |z|^{k+1}}{\left| 1 -z \right|} = \frac{2}{\left| 1 -z \right|} since ∣z∣=1|z| =1 and unbounded when z=1z=1, which we can't conclude anything based on above theorem