JOVANA
Library Glossary Getting Started Three Levels Fields How it works Mission
Join the mission
All guides

Nested Intervals and Why Limits Exist

The nested interval theorem captures completeness geometrically and pins down a single point. We then see how completeness — through Cauchy sequences — is what guarantees limits exist at all.

Closing in on a point

The nested interval theorem says: if I_1 ⊇ I_2 ⊇ I_3 ⊇ ... is a shrinking chain of closed bounded intervals I_n = [a_n, b_n], each containing the next, then their intersection is nonempty — there is at least one point lying in every interval. If in addition the lengths b_n - a_n shrink to 0, that point is unique. This is completeness wearing a geometric costume.

Proof that the intersection is nonempty.
  Nesting gives, for all m, n:   a_m <= b_n.
  (any left endpoint sits left of any right endpoint)
  So the set A = { a_n : n in N } of left endpoints
  is bounded above (every b_n is an upper bound).
  By COMPLETENESS, c = sup A exists in R.
  - c is an upper bound of A, so a_n <= c for every n.
  - each b_n is an upper bound of A, and c is the LEAST one, so c <= b_n.
  Hence a_n <= c <= b_n, i.e. c is in [a_n, b_n] for EVERY n.
  The intersection contains c, so it is nonempty.   QED

Uniqueness when lengths -> 0:
  if c and c' both lie in every I_n then |c - c'| <= b_n - a_n -> 0,
  forcing c = c'.   (closedness matters: open (0,1/n) intersect to empty)
Nested intervals reduced to a single supremum.

Completeness is what makes limits exist

Now the payoff for the whole track. A Cauchy sequence is one whose terms eventually cluster: for every epsilon > 0 there is an N so that |a_m - a_n| < epsilon whenever m, n ≥ N. This is a condition you can check using only the sequence itself, with no candidate limit in hand. The deep theorem is that in R, every Cauchy sequence converges — its limit exists. This property is precisely equivalent to completeness.

This is the whole point of the real numbers. Over Q the limit can fail to exist: the decimal truncations 1, 1.4, 1.41, 1.414, ... form a Cauchy sequence of rationals that clusters tighter and tighter, yet there is no rational number for it to converge to — it is aiming at the square root of 2, the very hole completeness fills. Pass to R and the hole is gone; the limit is there waiting. Completeness is the one fact that guarantees the limits analysis is built on actually exist.

Sketch: in R, every Cauchy sequence (a_n) converges.
  1. A Cauchy sequence is BOUNDED
     (beyond some N all terms lie within 1 of a_N; finitely many before).
  2. By Bolzano-Weierstrass (itself a child of completeness),
     a bounded sequence has a convergent SUBSEQUENCE a_{n_k} -> L.
  3. Cauchy + a subsequence reaching L forces the WHOLE sequence to L:
     given epsilon, pick N from the Cauchy condition (using epsilon/2),
     pick a subsequence index n_k >= N with |a_{n_k} - L| < epsilon/2.
     Then for n >= N:
       |a_n - L| <= |a_n - a_{n_k}| + |a_{n_k} - L| < epsilon/2 + epsilon/2 = epsilon.
  So a_n -> L.   QED
Every step above leans on completeness.
Cauchy implies convergent in R — completeness behind every line.