uniolympiads Ελληνικά

Continuity

Bolzano and intermediate values, fixed points and horizontal chords, the extreme value theorem, uniform continuity (Heine–Cantor), discontinuities of monotone functions, sequential continuity, Cauchy's equation.

What to know by heart

Let f ⁣:[a,b]→Rf\colon [a, b] \to \R be continuous.

  • Bolzano. If f(a)f(b)<0f(a) f(b) < 0, there is ξ∈(a,b)\xi \in (a, b) with f(ξ)=0f(\xi) = 0.
  • Intermediate values. ff takes every value between f(a)f(a) and f(b)f(b). More generally: the image of an interval under a continuous function is an interval.
  • Extreme value theorem (Weierstrass). ff is bounded on [a,b][a, b] and attains a maximum and a minimum. Together with the previous item: f([a,b])=[m,M]f([a, b]) = [m, M].
  • Heine–Cantor. ff is uniformly continuous: for every ε>0\varepsilon > 0 there is δ>0\delta > 0 with ∣f(x)−f(y)∣<ε|f(x) - f(y)| < \varepsilon whenever ∣x−y∣<δ|x - y| < \delta, with the same δ\delta for all x,yx, y.
  • Sequential continuity. ff is continuous at x0x_0 if and only if f(xn)→f(x0)f(x_n) \to f(x_0) for every sequence xn→x0x_n \to x_0.
  • Monotone functions. A monotone ff on an interval has one-sided limits at every point. Its only discontinuities are jumps, and there are at most countably many. A monotone function whose image is an interval is continuous.
  • Continuous and injective on an interval ⇒\Rightarrow strictly monotone.

The two matching tools are intermediate values and monotonicity.

The basic lemmas, with short proofs

Bolzano via the supremum. Let f(a)<0<f(b)f(a) < 0 < f(b) and c=sup⁡{x∈[a,b]:f(x)<0}c = \sup\{x \in [a, b] : f(x) < 0\}. If f(c)<0f(c) < 0, by continuity f<0f < 0 also slightly to the right of cc (and c<bc < b), a contradiction. If f(c)>0f(c) > 0, then f>0f > 0 on some (c−δ,c](c - \delta, c], so c−δc - \delta is also an upper bound, a contradiction. Hence f(c)=0f(c) = 0.

Fixed point. If f ⁣:[a,b]→[a,b]f\colon [a, b] \to [a, b] is continuous, g(x)=f(x)−xg(x) = f(x) - x has g(a)≥0≥g(b)g(a) \ge 0 \ge g(b), so it vanishes: there is xx with f(x)=xf(x) = x.

Bolzano via the supremum, and a fixed point via g(x)=f(x)−xg(x) = f(x) - x

Horizontal chord. If f ⁣:[0,1]→Rf\colon [0, 1] \to \R is continuous with f(0)=f(1)f(0) = f(1), then for every positive integer nn there is x∈[0,1−1n]x \in \big[0, 1 - \frac1n\big] with f(x+1n)=f(x)f\big(x + \frac1n\big) = f(x). Proof: g(x)=f(x+1n)−f(x)g(x) = f\big(x + \frac1n\big) - f(x) satisfies

g(0)+g(1n)+⋯+g(n−1n)=f(1)−f(0)=0,g(0) + g\big(\tfrac1n\big) + \dots + g\big(\tfrac{n-1}{n}\big) = f(1) - f(0) = 0 ,

so either some term is 00 or two terms have opposite signs, and then you apply Bolzano between them. For a chord length not of the form 1n\frac1n the conclusion fails.

Extreme value theorem. If ff were unbounded, there would be xnx_n with ∣f(xn)∣>n|f(x_n)| > n. By Bolzano–Weierstrass a subsequence converges to some x∗∈[a,b]x^* \in [a, b], and continuity gives f(xnk)→f(x∗)f(x_{n_k}) \to f(x^*), a contradiction. With M=sup⁡fM = \sup f, take xnx_n with f(xn)>M−1nf(x_n) > M - \frac1n and again a convergent subsequence: its limit is a maximum point.

Heine–Cantor. If it failed, there would be ε>0\varepsilon > 0 and xn,ynx_n, y_n with ∣xn−yn∣<1n|x_n - y_n| < \frac1n but ∣f(xn)−f(yn)∣≥ε|f(x_n) - f(y_n)| \ge \varepsilon. A subsequence xnk→x∗x_{n_k} \to x^*, hence also ynk→x∗y_{n_k} \to x^*, and continuity at x∗x^* gives f(xnk)−f(ynk)→0f(x_{n_k}) - f(y_{n_k}) \to 0, a contradiction. In both proofs the closed bounded interval is essential: 1x\frac1x on (0,1](0, 1] and x2x^2 on R\R are not uniformly continuous.

Continuous and injective ⇒ strictly monotone. If not, there are x1<x2x_1 < x_2 with f(x1)<f(x2)f(x_1) < f(x_2) and y1<y2y_1 < y_2 with f(y1)>f(y2)f(y_1) > f(y_2). For t∈[0,1]t \in [0, 1] put p(t)=(1−t)x1+ty1p(t) = (1 - t)x_1 + t y_1 and q(t)=(1−t)x2+ty2q(t) = (1 - t)x_2 + t y_2. Then q(t)>p(t)q(t) > p(t) for every tt, while h(t)=f(q(t))−f(p(t))h(t) = f(q(t)) - f(p(t)) is continuous with h(0)>0>h(1)h(0) > 0 > h(1). A t0t_0 with h(t0)=0h(t_0) = 0 gives two different points with the same value, a contradiction.

Discontinuities of a monotone function. Let ff be increasing. At an interior discontinuity xx we have f(x−)<f(x+)f(x^-) < f(x^+). The open intervals (f(x−),f(x+))\big(f(x^-), f(x^+)\big) for different xx are disjoint (for x<yx < y, f(x+)≤f(y−)f(x^+) \le f(y^-)), and each contains a rational. So there are at most countably many. Moreover, the image of ff meets (f(x−),f(x+))\big(f(x^-), f(x^+)\big) at most at f(x)f(x), so a jump leaves a gap in the image. That is why a monotone function with an interval as image is continuous.

Cauchy's equation under continuity, briefly. If f(x+y)=f(x)+f(y)f(x + y) = f(x) + f(y) for all x,yx, y, then f(q)=qf(1)f(q) = q f(1) for every rational qq (induction, together with f(0)=0f(0) = 0, f(−x)=−f(x)f(-x) = -f(x)). For arbitrary xx take rationals qn→xq_n \to x: by continuity f(x)=lim⁡f(qn)=xf(1)f(x) = \lim f(q_n) = x f(1). More in the functional equations sheet.

Typical tricks

  • The right difference. For "there is xx with f(x)=g(x)f(x) = g(x)" work with f−gf - g. For a fixed point with f(x)−xf(x) - x, for a chord with f(x+h)−f(x)f(x + h) - f(x).
  • Telescoping sums for signs. If some values of a continuous gg add up to 00, they cannot all be strictly positive or all strictly negative.
  • Take the maximum point. On a compact interval the maximum exists. Write the condition of the problem there and get a contradiction or an equality.
  • Polynomials. A polynomial of odd degree takes every real value. A polynomial of even degree with positive leading coefficient has a minimum and image [m,∞)[m, \infty).
  • Dense sets. A continuous function is determined by its values on a dense set (rationals, dyadics). Prove the identity there and pass to the limit.
  • Injectivity from the relation. If f(f(x))f(f(x)) is injective, so is ff. Together with continuity you get monotonicity for free.

How it appears in the IMC

  • IMC 2025, Day 1, Problem 1: a geometric question about tangent lines reduces to whether a polynomial takes every real value. Odd degree: intermediate values. Even degree: bounded on one side.
  • IMC 2026, Day 2, Problem 7: every horizontal section of a set is the image of a continuous function on [0,1][0, 1], hence an interval, and this gives a lower bound for its length.
  • SEEMOUS 2026, Problem 4: a strictly increasing function onto R\R is continuous, because a jump would leave a gap in the image. From there the solution continues with derivatives.
  • IMC 2000, Day 1, Problem 1: fixed points for monotone (not necessarily continuous) f ⁣:[0,1]→[0,1]f\colon [0, 1] \to [0, 1]. A good exercise to see which of the above survives without continuity.

The last exercise of the ladder is a bridge to IMC 2025, Day 1, Problem 1.

Pitfalls

  • The domain must be an interval. f(x)=1xf(x) = \frac1x on [−1,0)∪(0,1][-1, 0) \cup (0, 1] changes sign without a root.
  • Closed and bounded. xx on (0,1)(0, 1) has no maximum. 1x\frac1x on (0,1](0, 1] is unbounded.
  • The intermediate value property does not imply continuity. Derivatives have the Darboux property without necessarily being continuous (see the mean value theorems). For monotone functions, however, the intermediate value property does give continuity.
  • Sequential continuity needs all sequences. sin⁡1x\sin\frac1x has f(xn)→0f(x_n) \to 0 for xn=1nπx_n = \frac{1}{n\pi}, but no limit at 00.

Further reading

  • Kaczor & Nowak, Problems in Mathematical Analysis I, §2.1 (Monotonic Sequences)
  • Gelca & Andreescu, Putnam and Beyond, §3.2.4 (Derivatives and Their Applications)

Contest problems using these tools

No problems with these tools yet.