Discontinuities of Monotone Functions

A monotone function has left and right limits everywhere, so every discontinuity is a jump. Only finitely many jumps can exceed 1/n1/n, so the discontinuities are at most countable — no matter how many there are.

f(x−)=sup⁡t<xf(t),f(x+)=inf⁡t>xf(t)f(x-) = \sup_{t<x} f(t), \qquad f(x+) = \inf_{t>x} f(t)
#{x:f(x+)−f(x−)>1/n}≤n (f(b)−f(a))\#\{x : f(x+) - f(x-) > 1/n\} \le n\,(f(b) - f(a))

Examples & Exercises

Let ff be increasing on [a,b][a, b]. Show that the left limit f(x−)=sup⁡{f(t):t<x}f(x-) = \sup\{f(t) : t < x\} exists at every x∈(a,b]x \in (a, b], and the right limit f(x+)=inf⁡{f(t):t>x}f(x+) = \inf\{f(t) : t > x\} exists at every x∈[a,b)x \in [a, b).

f(x−)=sup⁡{f(t):t<x},f(x+)=inf⁡{f(t):t>x}f(x-) = \sup\{f(t) : t < x\}, \qquad f(x+) = \inf\{f(t) : t > x\}
f(x−)≤f(x)≤f(x+)(left limit is left-continuous)f(x-) \le f(x) \le f(x+) \quad \text{(left limit is left-continuous)}
At x=q1=12x = q_1 = \frac{1}{2}: f(12−)=f(12)≈0.8373474478721619f(\frac{1}{2}-) = f(\frac{1}{2}) \approx 0.8373474478721619, and the jump f(12+)−f(12)=2−1=12f(\frac{1}{2}+) - f(\frac{1}{2}) = 2^{-1} = \frac{1}{2} — the left limit equals the value; the rise is only from the right.
Show proof
  1. For every t<xt < x, monotonicity gives f(t)≤f(x)f(t) \le f(x). So {f(t):t<x}\{f(t) : t < x\} is a nonempty bounded set with upper bound f(x)f(x), and by completeness its supremum f(x−):=sup⁡{f(t):t<x}f(x-) := \sup\{f(t) : t < x\} exists as a real number.
  2. This supremum is indeed the left limit. Given ε>0\varepsilon > 0, the definition of the supremum yields t0<xt_0 < x with f(t0)>f(x−)−εf(t_0) > f(x-) - \varepsilon. By monotonicity, t0<t<xt_0 < t < x implies f(t0)≤f(t)≤f(x−)<f(x−)+εf(t_0) \le f(t) \le f(x-) < f(x-) + \varepsilon, so ∣f(t)−f(x−)∣<ε|f(t) - f(x-)| < \varepsilon — i.e. lim⁡t↑xf(t)=f(x−)\lim_{t \uparrow x} f(t) = f(x-).
  3. The right side is symmetric: {f(t):t>x}\{f(t) : t > x\} is bounded with lower bound f(x)f(x), so its infimum f(x+):=inf⁡{f(t):t>x}f(x+) := \inf\{f(t) : t > x\} exists, and the same ε\varepsilon argument gives lim⁡t↓xf(t)=f(x+)\lim_{t \downarrow x} f(t) = f(x+). Hence a monotone function has both one-sided limits at every point.
  4. Moreover f(x−)≤f(x)≤f(x+)f(x-) \le f(x) \le f(x+) always holds: taking the supremum of f(t)≤f(x)f(t) \le f(x) over t<xt < x gives f(x−)≤f(x)f(x-) \le f(x), and similarly on the right. In the example, at x=q1=12x = q_1 = \frac{1}{2} the left equality holds (f(12−)=f(12)≈0.8373474478721619f(\frac{1}{2}-) = f(\frac{1}{2}) \approx 0.8373474478721619) while the right jumps by 12\frac{1}{2}.

Show that a monotone function ff on [a,b][a, b] has at most countably many discontinuities. Hint: show the jumps j(x)=f(x+)−f(x−)≥0j(x) = f(x+) - f(x-) \ge 0 sum to at most f(b)−f(a)f(b) - f(a), and that {x:j(x)>1/n}\{x : j(x) > 1/n\} is finite with at most n(f(b)−f(a))n(f(b) - f(a)) points.

j(x):=f(x+)−f(x−)≥0,∑xj(x)≤f(b)−f(a)j(x) := f(x+) - f(x-) \ge 0, \qquad \sum_{x} j(x) \le f(b) - f(a)
D=⋃n=1∞{x:j(x)>1n}countable union of finite setsD = \bigcup_{n=1}^{\infty} \left\{x : j(x) > \frac{1}{n}\right\} \quad \text{countable union of finite sets}
n=4n = 4 (threshold 14\frac{1}{4}): exactly one jump exceeds 14\frac{1}{4} — k=1k = 1 (size 12\frac{1}{2}) — within the bound 4(f(1)−f(0))≈7.9999997615814214(f(1) - f(0)) \approx 7.999999761581421 (2−2=142^{-2} = \frac{1}{4} is not >14> \frac{1}{4}).
Show proof
  1. First, discontinuities can only be jumps. By problem 1, f(x−)f(x-) and f(x+)f(x+) always exist, so failure of lim⁡t→xf(t)=f(x)\lim_{t \to x} f(t) = f(x) forces f(x−)<f(x+)f(x-) < f(x+) — neither removable nor oscillatory discontinuities are possible; only a jump j(x)=f(x+)−f(x−)>0j(x) = f(x+) - f(x-) > 0 remains.
  2. For finitely many points x1<⋯<xmx_1 < \cdots < x_m, the jump intervals [f(xi−),f(xi+)][f(x_i-), f(x_i+)] are disjoint (monotonicity orders them) and all lie inside [f(a),f(b)][f(a), f(b)]. Hence ∑i=1mj(xi)≤f(b)−f(a)\sum_{i=1}^{m} j(x_i) \le f(b) - f(a), and since mm is arbitrary, the total of all jumps satisfies ∑xj(x)≤f(b)−f(a)\sum_x j(x) \le f(b) - f(a).
  3. Fix nn: if j(x)>1nj(x) > \frac{1}{n} at mnm_n points, then ∑j(x)≥mn⋅1n\sum j(x) \ge m_n \cdot \frac{1}{n}, so mn≤n(f(b)−f(a))m_n \le n(f(b) - f(a)) — a finite set. But the discontinuity set D={x:j(x)>0}=⋃n=1∞{x:j(x)>1n}D = \{x : j(x) > 0\} = \bigcup_{n=1}^{\infty} \{x : j(x) > \frac{1}{n}\} is a countable union of finite sets, hence at most countable.
  4. Numerical check (terms=24terms = 24, n=4n = 4): the jumps are 2−k2^{-k} (k=1,…,24k = 1, \dots, 24), and the only one exceeding 14\frac{1}{4} is 2−1=122^{-1} = \frac{1}{2} (2−2=142^{-2} = \frac{1}{4} is not a strict exceedance). Against the bound 4(f(1)−f(0))≈7.9999997615814214(f(1) - f(0)) \approx 7.999999761581421, we get 1≤7.999…1 \le 7.999\ldots — exactly as the theorem predicts.

Let (qk)(q_k) enumerate the rationals of (0,1)(0, 1) and g(x)=∑k:qk<x2−kg(x) = \sum_{k : q_k < x} 2^{-k}. Show that gg is increasing on [0,1][0, 1], discontinuous exactly at the rationals, and continuous at every irrational.

g(x)=∑k:qk<x2−k,g(qm+)−g(qm)=2−mg(x) = \sum_{k : q_k < x} 2^{-k}, \qquad g(q_m+) - g(q_m) = 2^{-m}
if x∉Q and N large, ∑k>N2−k<ε and δ=dist⁡(x,{q1,…,qN})>0\text{if } x \notin \mathbb{Q} \text{ and } N \text{ large, } \sum_{k > N} 2^{-k} < \varepsilon \text{ and } \delta = \operatorname{dist}(x, \{q_1, \dots, q_N\}) > 0
At x=q3=23x = q_3 = \frac{2}{3}: g(23)≈0.841285169124603g(\frac{2}{3}) \approx 0.841285169124603, jumping up by 2−3=18=0.1252^{-3} = \frac{1}{8} = 0.125 — g(23+)≈0.966285169124603g(\frac{2}{3}+) \approx 0.966285169124603.
Show proof
  1. The series always converges: 0≤∑k:qk<x2−k≤∑k=1∞2−k=10 \le \sum_{k : q_k < x} 2^{-k} \le \sum_{k=1}^{\infty} 2^{-k} = 1, so g(x)g(x) is well defined with values in [0,1][0, 1]. If x<yx < y then {k:qk<x}⊆{k:qk<y}\{k : q_k < x\} \subseteq \{k : q_k < y\}, so the sum only gains terms and g(x)≤g(y)g(x) \le g(y) — gg is increasing.
  2. Let x=qmx = q_m. The sum defining g(qm)g(q_m) excludes k=mk = m (the condition qk<qmq_k < q_m leaves out qmq_m itself), but as t↓qmt \downarrow q_m, eventually qm<tq_m < t for all t>qmt > q_m, adding 2−m2^{-m}. Hence g(qm+)=g(qm)+2−mg(q_m+) = g(q_m) + 2^{-m} — gg jumps by 12m\frac{1}{2^m} at qmq_m, so it is discontinuous there.
  3. Now let xx be irrational. Given ε>0\varepsilon > 0, pick NN with ∑k>N2−k<ε\sum_{k > N} 2^{-k} < \varepsilon (possible since the tail tends to 0). Since xx is irrational, δ:=min⁡1≤k≤N∣x−qk∣>0\delta := \min_{1 \le k \le N} |x - q_k| > 0. If ∣t−x∣<δ|t - x| < \delta, none of q1,…,qNq_1, \dots, q_N lies strictly between xx and tt, so the difference g(t)−g(x)g(t) - g(x) comes only from the tail: ∣g(t)−g(x)∣<ε|g(t) - g(x)| < \varepsilon. Hence gg is continuous at xx.
  4. So gg is discontinuous at every rational and continuous at every irrational — a concrete example of a monotone function discontinuous on a dense countable set (and, by problem 2, on no larger a set than dense). Numerically, at q3=23q_3 = \frac{2}{3}: g(23)≈0.841285169124603g(\frac{2}{3}) \approx 0.841285169124603, g(23+)≈0.966285169124603g(\frac{2}{3}+) \approx 0.966285169124603 — the gap 0.125=2−30.125 = 2^{-3} matches the caption's jump.

3. Continuity