← 2017 Paper 2

UPSC 2017 Maths Optional Paper 2 Q4c — Step-by-Step Solution

20 marks · Section A

Rearrangement of Series; Riemann's Theorem · Real Analysis · Read the full method →

Question

Let n=1xn\displaystyle\sum_{n=1}^\infty x_n be a conditionally convergent series of real numbers. Show that there is a rearrangement n=1xπ(n)\displaystyle\sum_{n=1}^\infty x_{\pi(n)} of the series n=1xn\displaystyle\sum_{n=1}^\infty x_n that converges to 100100.

Technique

Riemann rearrangement — split into positive/negative subseries (both divergent, terms 0\to0, from conditional convergence), greedily overshoot/undershoot the target 100100, bound each crossing error by a single term 0\to0.

Solution

This is a special case (target S=100S=100) of the Riemann rearrangement theorem. The proof has three parts: (A) extract the positive and negative subsequences and establish their key properties; (B) describe a greedy rearrangement; (C) prove it converges to 100100.

Step 1 — Positive and negative parts; consequences of conditional convergence

Let pn=max(xn,0)0p_n=\max(x_n,0)\ge0 (positive parts) and qn=max(xn,0)0q_n=\max(-x_n,0)\ge0 (negative parts), so

xn=pnqn,xn=pn+qn.x_n=p_n-q_n,\qquad |x_n|=p_n+q_n.

Conditional convergence means xn\sum x_n converges but xn=\sum|x_n|=\infty.

Claim: pn=+\displaystyle\sum p_n=+\infty and qn=+\displaystyle\sum q_n=+\infty.

Suppose, for contradiction, that pn<\sum p_n<\infty. Since qn=pnxnq_n=p_n-x_n and both pn\sum p_n and xn\sum x_n converge, qn=pnxn\sum q_n=\sum p_n-\sum x_n would converge; then xn=(pn+qn)<\sum|x_n|=\sum(p_n+q_n)<\infty, contradicting conditional convergence. Symmetrically qn<\sum q_n<\infty also fails. Hence both pn=qn=+\sum p_n=\sum q_n=+\infty.

Also: xn0x_n\to0 (terms of a convergent series), so pn0p_n\to0 and qn0q_n\to0.

Now discard the zero terms and list the strictly positive terms of the series in their original order as a1,a2,a3,a_1,a_2,a_3,\ldots (with ak=+\sum a_k=+\infty, ak0a_k\to0) and the absolute values of the strictly negative terms as b1,b2,b3,b_1,b_2,b_3,\ldots (with bk=+\sum b_k=+\infty, bk0b_k\to0). Each original nonzero xnx_n appears exactly once among the aa‘s or the bb‘s; this is what lets us build a genuine rearrangement (a bijection π\pi). (Zero terms, if any, can be inserted anywhere and do not affect convergence; place them, say, after each negative block.)

Step 2 — The greedy construction

Build the rearranged series in blocks, alternating “add positives until you exceed 100100” and “subtract positives’ counterpart, i.e. add negatives until you drop below 100100”:

  1. Take the fewest leading positive terms a1,,am1a_1,\ldots,a_{m_1} so that the partial sum first exceeds 100100:
a1++am1>100,a1++am11100.a_1+\cdots+a_{m_1}>100,\quad a_1+\cdots+a_{m_1-1}\le100.

(Possible because ak=+\sum a_k=+\infty, so the positive sums are unbounded.) 2. Then append the fewest leading negative terms b1,,bn1-b_1,\ldots,-b_{n_1} so the running sum first drops below 100100:

(a1++am1)(b1++bn1)<100,\big(a_1+\cdots+a_{m_1}\big)-\big(b_1+\cdots+b_{n_1}\big)<100,

with the previous step (n11n_1-1 negatives) still 100\ge100. (Possible because bk=+\sum b_k=+\infty.) 3. Resume with the next unused positives am1+1,,am2a_{m_1+1},\ldots,a_{m_2} until the sum again first exceeds 100100; then the next unused negatives until it again first drops below 100100; and so on indefinitely.

Because both ak\sum a_k and bk\sum b_k diverge, each step can always be completed, and every aka_k and bkb_k is eventually used (we always continue from where we left off and each block uses at least one new term and ratchets through the lists). So this defines a bijection π\pi of N\mathbb N, i.e. a genuine rearrangement xπ(n)\sum x_{\pi(n)}.

Step 3 — Convergence to 100100

Let SNS_N denote the partial sums of the rearranged series. By construction the running sum crosses the level 100100 at the end of each block, alternately overshooting above and undershooting below. Consider any “turning point” partial sum:

0<Send of +block100amj.0<S_{\text{end of }+\text{block}}-100\le a_{m_j}. 0<100Send of blockbnj.0<100-S_{\text{end of }-\text{block}}\le b_{n_j}.

Between consecutive turning points the partial sums are monotone (adding only positives, or only negatives), so every partial sum SNS_N lies within distance max(amj,bnj)\max(a_{m_j},b_{n_j}) of 100100 for the current block. Precisely, for all NN in the jjth pair of blocks,

SN100max(amj,bnj).|S_N-100|\le\max\big(a_{m_j},\,b_{n_j}\big).

As jj\to\infty we use infinitely many terms, so the indices mj,njm_j,n_j\to\infty; since ak0a_k\to0 and bk0b_k\to0, we get amj0a_{m_j}\to0 and bnj0b_{n_j}\to0. Therefore

lim supNSN100limjmax(amj,bnj)=0,\limsup_{N\to\infty}|S_N-100|\le\lim_{j\to\infty}\max(a_{m_j},b_{n_j})=0,

i.e. SN100S_N\to100. Hence

Answer

  n=1xπ(n)=100.  \boxed{\;\sum_{n=1}^\infty x_{\pi(n)}=100.\;}
We post more of this — worked solutions, CSAT trap breakdowns, guide chapters — a few times a week on Telegram. Free, no sign-in. Join

This solution is part of the Maths Coverage Map — 13 years, mapped. Get the take-away PDF free.