Arithmetic Operations on Series: The Theorem of Convergence and Multiplication

This passage presents the proof of a theorem stating that if two infinite series converge, their respective terms can be multiplied to obtain a new convergent series, even without the condition of absolute convergence.

Some author2 min read

Arithmetic operation over series.

Series are operating additionally and scaling nicely. However, the case for multiplication can be tricky. Let's look at the following theorem.

Theorem\textbf{Theorem} If ∑n=0∞an\sum_{n=0}^{\infty} a_n converges absolutely, ∑n=0∞an=A\sum_{n=0}^{\infty} a_n=A, ∑n=0∞bn=B\sum_{n=0}^{\infty} b_n=B, cn=∑k=0nakbn−k(n=0,1,2,…)c_n=\sum_{k=0}^n a_k b_{n-k} \quad(n=0,1,2, \ldots).

Then ∑n=0∞cn=AB\sum_{n=0}^{\infty} c_n=A B

Proof. let An=∑i=0naiA_n = \sum_{i=0}^n a_i, Bn=∑i=0nbiB_n = \sum_{i=0}^n b_i, Cn=∑i=0ncnC_n = \sum_{i=0}^n c_n. Now we want to show that ∑n=0∞cn=AB\sum_{n=0}^{\infty} c_n=A B, which can be derived from lim⁡n→∞Cn=AB\lim_{n\to \infty}C_n = AB,

which can be derived from lim⁡n→∞Cn=lim⁡n→∞AnB\lim_{n\to \infty}C_n =\lim_{n \to \infty}A_n B, which can be derived from lim⁡n→∞Cn−AnB=0\lim_{n\to \infty}C_n - A_n B = 0

Since we have Cn−AnB=∑i=0n∑k=0iakbi−k−AnB=∑k=0nak∑i=0n−kbi−AnB=∑k=0nak∑i=0n−kbi−AnB=∑k=0nakBn−k−AnBC_n -A_n B = \sum_{i=0}^n \sum_{k=0}^i a_k b_{i-k} -A_n B = \sum_{k=0}^n a_k \sum_{i=0}^{n-k}b_i-A_nB = \sum_{k=0}^n a_k \sum_{i=0}^{n-k}b_i-A_nB = \sum_{k=0}^n a_k B_{n-k}-A_nB

=∑k=0nak(B−∑i=n−k+1∞bi)−AnB=\sum_{k=0}^n a_k (B - \sum_{i={n-k+1}}^\infty b_{i})-A_nB

Let's denote f(x)=−∑i=x+1∞bi,f(x) = -\sum_{i=x+1}^{\infty} b_i, then. we will have

Cn−AnB=∑k=0nak(B+f(n−k))−AnB=∑k=0nakf(n−k)C_n -A_n B = \sum_{k=0}^n a_k (B +f(n-k))-A_nB = \sum_{k=0}^n a_k f(n-k)

So all we want to show is as nn goes to infinity the right hand side ∑k=0nakf(n−k)\sum_{k=0}^n a_k f(n-k) goes to zero.

We can accomplish this by noting as xx goes to infinity f(x)f(x) goes to zero due to the fact that BnB_n goes to BB as nn goes to infinity.

Then we will have for every ϵ>0\epsilon >0, there exists an NN such that for every n′>Nn' >N, we will have ∣f(n′)∣<ϵ.|f(n')|<\epsilon. Then, for n>Nn >N,

∣∑k=0nakf(n−k)∣=∣∑k=0nan−kf(k)∣=∣∑k=0Nan−kf(k)+∑k=N+1nan−kf(k)∣≤∣∑k=0Nan−kf(k)∣+∣∑k=N+1nan−kf(k)∣<∣∑k=0Nan−kf(k)∣+A′ϵ|\sum_{k=0}^n a_k f(n-k)| = |\sum_{k=0}^n a_{n-k} f(k)| = |\sum_{k=0}^N a_{n-k} f(k)+ \sum_{k=N+1}^n a_{n-k} f(k)| \le |\sum_{k=0}^N a_{n-k} f(k)| +| \sum_{k=N+1}^n a_{n-k} f(k)| < |\sum_{k=0}^N a_{n-k} f(k)| + A' \epsilon,

where A′A' is the sum of ∣an∣|a_n| and exists due to ∑an\sum a_n converges absolutely.

Then if we sending nn to ∞\infty, we will have

lim sup⁡n→∞∣∑k=0nakf(n−k)∣≤A′ϵ\limsup_{n \to \infty} \left|\sum_{k=0}^n a_k f(n-k)\right| \le A' \epsilon

due to ϵ\epsilon being arbitrary, we have the limit to be zero. And therefore, we have the original.