For a convergent series ∑an→S, the power series ∑anxn approaches S as x→1− — Abel's theorem. Even the divergent Grandi series points to 21 in this limit, but that is the Abel sum, never "the sum" of the series.
Prove by summation by parts that if the partial sums Sn=a0+⋯+an converge to S, then limx→1−∑n=0∞anxn=S.
n=0∑Nanxn=SNxN+1+(1−x)n=0∑NSnxn
n=0∑∞anxn−S=(1−x)n=0∑∞(Sn−S)xn
Convergent example at x=0.9: S40(0.9)≈0.6420, closed form ln(1.9)≈0.6419 — the Abel sum points the same way as the ordinary sum ln2.Show proof
With S−1:=0 we have an=Sn−Sn−1. Since ∑n=0N(Sn−Sn−1)xn=∑n=0NSnxn−x∑n=0N−1Snxn, rearranging gives ∑n=0Nanxn=SNxN+1+(1−x)∑n=0NSnxn — Abel's transformation (summation by parts).
For fixed ∣x∣<1, letting N→∞ kills the boundary term SNxN+1→0 (SN is convergent hence bounded, xN+1→0). Thus ∑n=0∞anxn=(1−x)∑n=0∞Snxn — the right side converges absolutely since Sn is bounded.
Since (1−x)∑n=0∞xn=1, we get ∑anxn−S=(1−x)∑n=0∞(Sn−S)xn. Given ε>0, pick N0 with ∣Sn−S∣<ε/2 for all n>N0, and split the sum at N0.
Head: (1−x)∑n=0N0(Sn−S)xn≤(1−x)∑n=0N0∣Sn−S∣=(1−x)M (using xn≤1). M is a fixed constant, so this tends to 0 as x→1−. For x close enough to 1 the total is below ε. Hence limx→1−∑anxn=S. ∎
Prove the Abel sum of Grandi's series ∑n=0∞(−1)n is 21. Be explicit that the series itself diverges.
n=0∑N(−x)n=1+x1−(−x)N+1N→∞1+x1(∣x∣<1)
x→1−lim1+x1=21
x=0.99, N=100: S100(0.99)≈0.6846, closed form 1/1.99≈0.5025 — raising N and sending x→1− drives it to 1/2.Show proof
First, honestly: ∑n=0∞(−1)n diverges. The partial sums Sk=1 (k even), 0 (k odd) bounce forever with no limit (see the Grandi scene). What we compute now is not "the sum" of the series, but its Abel sum.
For ∣x∣<1 the geometric formula gives ∑n=0N(−x)n=1+x1−(−x)N+1. E.g. at x=0.9, N=20: S20(0.9)=1.91−(−0.9)21≈0.5839, with exact error 1.9∣(−0.9)21∣≈0.0576.
For ∣x∣<1, (−x)N+1→0, so letting N→∞ gives ∑n=0∞(−1)nxn=1+x1. At x=0.9 this is 1/1.9≈0.5263 — the true target of the S20 above.
Now let x→1−: 1+x1→21. Hence the Abel sum limx→1−∑n=0∞(−1)nxn=21. It coincides with the Cesàro sum 21, but neither is "the sum" of a divergent series. ∎
Apply Abel's theorem to the alternating harmonic series ∑n=0∞n+1(−1)n=ln2 to prove limx→1−∑n=0∞n+1(−1)nxn+1=ln2.
x=0.99, N=100: S100(0.99)≈0.6899, closed form ln(1.99)≈0.6881 — tending to ln2≈0.6931 as x→1−.Show proof
Set an=n+1(−1)n. Then ∑n=0∞an converges by the Leibniz test (alternating, n+11↘0), and its value is ln2≈0.6931 (numerical check: at N=100000 the partial sum is ≈0.69315218, error ≈5.0×10−6).
Abel's theorem (Problem 1) gives limx→1−∑n=0∞anxn=ln2. But ∑n=0∞n+1(−1)nxn+1=x⋅∑n=0∞anxn, and x→1, so the desired limit is 1⋅ln2=ln2.
A direct check agrees: integrating ∑n=0N(−t)n=1+t1−(−t)N+1 (for ∣t∣<1) from 0 to x gives ∑n=0Nn+1(−1)nxn+1=ln(1+x)−RN(x) with ∣RN(x)∣≤∫0xtN+1dt=N+2xN+2→0 (N→∞). So the power series sums to exactly ln(1+x).
Since ln(1+x)→ln2 (x→1−), the limit is ln2≈0.6931 — matching Abel's theorem's prediction. This shows the Abel sum never contradicts the ordinary sum for a convergent series. ∎