Continuity, Differentiability and Bounded Variation
CSIR NET real analysis: continuity, uniform continuity, types of discontinuity, monotone functions, derivatives, Darboux, MVT, Taylor and bounded variation.
- 20 min
On this page · 14 sections
Prerequisites
- Sequences, Series, limsup and liminf: continuity is tested with sequences throughout.
- Real Numbers, Sets and Countability: the “one rational per jump” argument proves monotone functions have countably many discontinuities.
- Compactness of \( [a, b] \) is used here and proved in general in Metric Spaces and Several Variables.
The Hook: One Formula, Four Behaviours
Continuity
\( f : D \to \mathbb{R} \) is continuous at \( c \in D \) if for every \( \varepsilon > 0 \) there is \( \delta > 0 \) with \( \lvert f(x) - f(c) \rvert < \varepsilon \) whenever \( x \in D \), \( \lvert x - c \rvert < \delta \). Equivalently (sequential criterion), \( x_n \to c \) in \( D \) implies \( f(x_n) \to f(c) \). To prove discontinuity, find two sequences tending to \( c \) whose images have different limits.
Theorems on intervals
- Boundedness and extreme values. A continuous \( f : [a, b] \to \mathbb{R} \) is bounded and attains its sup and inf.
- Intermediate value theorem. A continuous function on an interval takes every value between any two of its values. So the image of an interval is an interval, and the image of \( [a, b] \) is a closed bounded interval.
- Fixed points. A continuous \( f : [a, b] \to [a, b] \) has a fixed point (apply the IVT to \( f(x) - x \)).
- Injective means monotone. A continuous injective function on an interval is strictly monotone, and its inverse is continuous.
The pathological zoo
| Function | Continuous at |
|---|---|
| Dirichlet: \( 1_{\mathbb{Q}}(x) \) | nowhere |
| \( x \cdot 1_{\mathbb{Q}}(x) \) | only \( x = 0 \) |
| \( x^2 \cdot 1_{\mathbb{Q}}(x) \) | only 0, and differentiable only at 0 |
| Thomae: \( f(\frac{p}{q}) = \frac{1}{q} \) in lowest terms (\( q \ge 1 \)), \( f = 0 \) on irrationals | exactly the irrationals |
| \( (x - a)(x - b) \cdot 1_{\mathbb{Q}}(x) \) | exactly \( a \) and \( b \) |
The set of points where any function \( f : \mathbb{R} \to \mathbb{R} \) is continuous is a \( G_\delta \) set (a countable intersection of open sets). By the Baire category theorem \( \mathbb{Q} \) is not \( G_\delta \), so no function is continuous exactly on the rationals, though Thomae’s function is continuous exactly on the irrationals.
Types of discontinuity
At an interior point \( c \), with \( f(c^-) \), \( f(c^+) \) the one-sided limits:
| Type | Condition | Example at 0 |
|---|---|---|
| Removable | \( f(c^-) = f(c^+) \ne f(c) \) | \( f(x) = 1 \) for \( x \ne 0 \), \( f(0) = 0 \) |
| Jump (first kind) | both exist, unequal | \( \operatorname{sgn} x \) |
| Second kind | at least one one-sided limit fails to exist | \( \sin\frac{1}{x} \), \( \frac{1}{x} \) |
Monotone functions
If \( f \) is increasing on \( (a, b) \), then at every point both one-sided limits exist and \( f(c^-) \le f(c) \le f(c^+) \). So:
- A monotone function has only jump discontinuities (no removable, no second kind, in the interior).
- The set of discontinuities is countable: the open intervals \( (f(c^-), f(c^+)) \) are disjoint, and each contains a rational.
- Conversely, any countable set \( \{x_n\} \subseteq (a, b) \) is the discontinuity set of an increasing function: \( f(x) = \sum_{x_n < x} 2^{-n} \).
- (Lebesgue) A monotone function is differentiable almost everywhere.
Uniform Continuity
\( f \) is uniformly continuous on \( D \) if for every \( \varepsilon > 0 \) there is one \( \delta > 0 \) that works at every point: \( \lvert x - y \rvert < \delta \Rightarrow \lvert f(x) - f(y) \rvert < \varepsilon \) for all \( x, y \in D \).
Sequential test for failure. \( f \) is not uniformly continuous on \( D \) iff there are \( x_n, y_n \in D \) with \( x_n - y_n \to 0 \) but \( \lvert f(x_n) - f(y_n) \rvert \ge \varepsilon_0 > 0 \).
| Criterion | Conclusion |
|---|---|
| \( f \) continuous on a compact set (Heine–Cantor) | uniformly continuous |
| \( f \) Lipschitz: \( \lvert f(x) - f(y) \rvert \le K \lvert x - y \rvert \) | uniformly continuous |
| \( f \) differentiable on an interval with \( f' \) bounded | Lipschitz (by MVT), hence UC |
| \( f \) continuous on \( (a, b) \), bounded interval | UC \( \iff \) \( f(a^+) \) and \( f(b^-) \) exist (finite) |
| \( f \) continuous on \( [a, \infty) \) and \( \lim_{x\to\infty} f(x) \) finite | UC |
| \( f \) continuous and periodic on \( \mathbb{R} \) | UC |
| \( f \) UC on a bounded set | \( f \) bounded |
| \( f \) UC, \( (x_n) \) Cauchy | \( (f(x_n)) \) Cauchy |
| Function | UC? | Reason |
|---|---|---|
| \( x^2 \) on \( \mathbb{R} \) | no | \( x_n = n + \frac{1}{n} \), \( y_n = n \): difference of squares \( \to 2 \) |
| \( x^2 \) on \( [0, 5] \) | yes | compact |
| \( \frac{1}{x} \) on \( (0, 1) \) | no | no limit at \( 0^+ \) |
| \( \frac{1}{x} \) on \( [1, \infty) \) | yes | \( \lvert f' \rvert \le 1 \) |
| \( \sin\frac{1}{x} \) on \( (0, 1) \) | no | no limit at \( 0^+ \) |
| \( x\sin\frac{1}{x} \) on \( (0, 1) \) | yes | extends continuously to \( [0, 1] \) |
| \( \sqrt{x} \) on \( [0, \infty) \) | yes | but not Lipschitz (slope blows up at 0) |
| \( \sin(x^2) \) on \( \mathbb{R} \) | no | \( x_n = \sqrt{n\pi + \frac{\pi}{2}} \), \( y_n = \sqrt{n\pi} \) |
| \( x \sin x \) on \( \mathbb{R} \) | no | unbounded oscillation |
| \( \frac{\sin x}{x} \) on \( (0, \infty) \) | yes | limits 1 at \( 0^+ \), 0 at \( \infty \) |
| \( e^x \) on \( \mathbb{R} \) | no | \( e^x \) on \( (-\infty, 0] \) is UC, on \( [0, \infty) \) not |
Sums and compositions of uniformly continuous functions are uniformly continuous; products need not be (\( x \cdot x \) on \( \mathbb{R} \)), though products of bounded UC functions are.
Differentiability
\( f'(c) = \lim_{h \to 0} \frac{f(c + h) - f(c)}{h} \). Differentiable implies continuous; \( \lvert x \rvert \) shows the converse fails. A continuous nowhere-differentiable function exists (Weierstrass; see Sequences and Series of Functions).
The mean value theorems
| Theorem | Hypotheses | Conclusion |
|---|---|---|
| Rolle | \( f \) continuous on \( [a,b] \), differentiable on \( (a,b) \), \( f(a) = f(b) \) | \( f'(c) = 0 \) for some \( c \in (a,b) \) |
| Lagrange MVT | same, without \( f(a) = f(b) \) | \( f(b) - f(a) = f'(c)(b - a) \) |
| Cauchy MVT | \( f, g \) as above | \( (f(b) - f(a))g'(c) = (g(b) - g(a))f'(c) \) |
| Taylor (Lagrange remainder) | \( f^{(n)} \) continuous on \( [a,b] \), \( f^{(n+1)} \) exists on \( (a,b) \) | \( f(b) = \sum_{k=0}^{n} \frac{f^{(k)}(a)}{k!}(b-a)^k + \frac{f^{(n+1)}(c)}{(n+1)!}(b-a)^{n+1} \) |
Consequences used in Part B:
- \( f' = 0 \) on an interval \( \Rightarrow \) \( f \) constant. \( f' = g' \Rightarrow f - g \) constant.
- \( f' > 0 \) on an interval \( \Rightarrow \) \( f \) strictly increasing. The converse fails: \( x^3 \) is strictly increasing with \( f'(0) = 0 \).
- \( \lvert f' \rvert \le K \Rightarrow \lvert f(x) - f(y) \rvert \le K \lvert x - y \rvert \).
- Between two roots of \( f \) lies a root of \( f' \). So if \( f' \) has \( k \) real roots, \( f \) has at most \( k + 1 \).
- Inequalities: \( \lvert \sin x - \sin y \rvert \le \lvert x - y \rvert \); \( \frac{x}{1 + x} < \log(1 + x) < x \) for \( x > 0 \); \( e^x \ge 1 + x \).
L’Hospital’s rule. If \( f, g \) are differentiable near \( a \), \( g' \ne 0 \) there, \( f, g \to 0 \) (or \( \lvert g \rvert \to \infty \)) and \( \frac{f'}{g'} \to L \), then \( \frac{f}{g} \to L \). The converse fails: \( \frac{x + \sin x}{x} \to 1 \) as \( x \to \infty \), but \( \frac{1 + \cos x}{1} \) has no limit.
Smooth but not analytic. \( f(x) = e^{-1/x^2} \) (\( x \ne 0 \)), \( f(0) = 0 \), is infinitely differentiable with \( f^{(n)}(0) = 0 \) for all \( n \). Its Taylor series at 0 is identically zero, yet \( f(x) > 0 \) for \( x \ne 0 \).
Functions of Bounded Variation
For \( f : [a, b] \to \mathbb{R} \) and a partition \( P : a = x_0 < x_1 < \dots < x_n = b \), the total variation is
$$V_a^b(f) = \sup_P \sum_{i=1}^{n} \lvert f(x_i) - f(x_{i-1}) \rvert.$$\( f \) is of bounded variation (BV) if \( V_a^b(f) < \infty \). Think of it as the total vertical distance travelled by the graph.
| Result | Statement |
|---|---|
| Monotone | \( V_a^b(f) = \lvert f(b) - f(a) \rvert \) |
| Lipschitz | BV, with \( V \le K(b - a) \) |
| \( f' \) bounded on \( [a,b] \) | Lipschitz, hence BV |
| \( f' \) continuous (or Riemann integrable) | \( V_a^b(f) = \int_a^b \lvert f' \rvert \) |
| Additivity | \( V_a^b = V_a^c + V_c^b \) for \( a < c < b \) |
| Jordan decomposition | \( f \) is BV \( \iff \) \( f = g - h \) with \( g, h \) increasing (e.g. \( g(x) = V_a^x(f) \)) |
| Closure | sums, differences and products of BV functions are BV |
| Regularity | BV \( \Rightarrow \) bounded, countably many discontinuities (all of the first kind: one-sided limits exist), Riemann integrable, differentiable a.e. |
Continuous but not BV. \( f(x) = x \sin\frac{1}{x} \), \( f(0) = 0 \) on \( [0, 1] \). At \( x_k = \frac{1}{(k + \frac{1}{2})\pi} \), \( \lvert f(x_k) \rvert = x_k \), and the graph swings through 0 between consecutive \( x_k \). The variation is at least \( \sum_k x_k = \infty \) (a harmonic-type series).
Differentiable but not BV. \( f(x) = x^2 \sin\frac{1}{x^2} \), \( f(0) = 0 \): differentiable everywhere on \( [0, 1] \), but the same swing argument with \( x_k^2 = \frac{1}{(k + \frac{1}{2})\pi} \) gives \( \sum x_k^2 = \infty \). Its derivative is unbounded near 0.
Counter-examples Table
| False claim | Counter-example |
|---|---|
| Continuous \( \Rightarrow \) differentiable | \( \lvert x \rvert \); Weierstrass function (nowhere) |
| Differentiable \( \Rightarrow \) \( f' \) continuous | \( x^2 \sin\frac{1}{x} \), \( f(0) = 0 \) |
| Continuous \( \Rightarrow \) uniformly continuous | \( x^2 \) on \( \mathbb{R} \); \( \frac{1}{x} \) on \( (0,1) \) |
| Uniformly continuous \( \Rightarrow \) Lipschitz | \( \sqrt{x} \) on \( [0, 1] \) |
| Uniformly continuous \( \Rightarrow \) bounded | \( f(x) = x \) on \( \mathbb{R} \) |
| Product of UC functions is UC | \( x \cdot x \) on \( \mathbb{R} \) |
| Intermediate value property \( \Rightarrow \) continuous | \( \sin\frac{1}{x} \), value 0 at 0 |
| \( f \) strictly increasing \( \Rightarrow f' > 0 \) | \( x^3 \) at 0 |
| \( f'(c) > 0 \Rightarrow f \) increasing near \( c \) | \( x + 2x^2 \sin\frac{1}{x} \), \( f(0) = 0 \): \( f'(0) = 1 \), not monotone near 0 |
| \( C^\infty \Rightarrow \) equals its Taylor series | \( e^{-1/x^2} \) |
| Continuous \( \Rightarrow \) BV | \( x \sin\frac{1}{x} \) |
| Differentiable \( \Rightarrow \) BV | \( x^2 \sin\frac{1}{x^2} \) |
| A function continuous only on \( \mathbb{Q} \) exists | impossible (Baire); Thomae is the reverse |
| A derivative can have a jump | impossible (Darboux) |
| \( \frac{f}{g} \to L \Rightarrow \frac{f'}{g'} \to L \) | \( \frac{x + \sin x}{x} \) at \( \infty \) |
Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| uniform continuity (one \( \delta \) for all points) | continuity (a \( \delta \) for each point) | |
| Lipschitz (bounded difference quotients) | uniformly continuous (weaker: \( \sqrt{x} \)) | |
| differentiable at \( c \) | \( f' \) continuous at \( c \) | |
| jump discontinuity (both one-sided limits exist) | second kind (one fails) | |
| \( f' > 0 \) (sufficient for increasing) | \( f' \ge 0 \) (necessary for increasing) | |
| BV (finite total travel) | bounded (finite range); \( \sin\frac{1}{x} \) is bounded but not BV |
Common Mistakes
Practice
All questions below are original practice questions written for this page. They are not from any CSIR NET paper. In Level 3, as in Part C, one or more options may be correct.
Level 1: Foundation
- Is \( \sqrt{x} \) uniformly continuous on \( [0, \infty) \)? Is it Lipschitz there?
- For which \( a \) is \( f(x) = x^a \sin\frac{1}{x} \) (\( f(0) = 0 \)) differentiable at 0?
- State Darboux’s theorem.
Answers
- UC: yes. Lipschitz: no. On \( [1, \infty) \), \( \lvert f' \rvert \le \frac{1}{2} \); on \( [0, 1] \) it is UC by compactness; combine. Not Lipschitz since \( \frac{\sqrt{h} - 0}{h} = \frac{1}{\sqrt{h}} \to \infty \).
- \( a > 1 \). \( \frac{f(h)}{h} = h^{a-1}\sin\frac{1}{h} \to 0 \) iff \( a > 1 \).
- If \( f \) is differentiable on \( [a, b] \), then \( f' \) takes every value between \( f'(a) \) and \( f'(b) \).
Level 2: Part B Style
Practice Q1. Which of the following functions is uniformly continuous on \( (0, 1) \)?
(a) \( \frac{1}{x} \) (b) \( \sin\frac{1}{x} \) (c) \( x \sin\frac{1}{x} \) (d) \( \frac{1}{1 - x} \)
Answer: (c)
\( x\sin\frac{1}{x} \to 0 \) as \( x \to 0^+ \) and \( \to \sin 1 \) as \( x \to 1^- \), so it extends continuously to \( [0, 1] \) and is UC. (a) and (d) blow up at an end; (b) has no limit at \( 0^+ \).
Practice Q2. The number of real roots of \( x^5 + x^3 + x - 1 = 0 \) is:
(a) 0 (b) 1 (c) 3 (d) 5
Answer: (b)
\( f'(x) = 5x^4 + 3x^2 + 1 > 0 \), so \( f \) is strictly increasing and has at most one root. \( f(0) = -1 < 0 < 2 = f(1) \), so by the IVT it has exactly one.
Practice Q3. Let \( f : [0, 2] \to \mathbb{R} \) be differentiable with \( f(0) = 1 \) and \( f'(x) \le 3 \) for all \( x \). The largest possible value of \( f(2) \) is:
(a) 3 (b) 6 (c) 7 (d) no maximum exists
Answer: (c)
MVT: \( f(2) = f(0) + 2f'(c) \le 1 + 6 = 7 \), attained by \( f(x) = 1 + 3x \).
Practice Q4. With the value 0 at \( x = 0 \) in each case, which function is not of bounded variation on \( [0, 1] \)?
(a) \( x^2 \sin\frac{1}{x} \) (b) \( x \sin\frac{1}{x} \) (c) \( \sqrt{x} \) (d) \( \lfloor 4x \rfloor \)
Answer: (b)
(b) has infinite variation (the swings at \( x_k = \frac{1}{(k + 1/2)\pi} \) add up like a harmonic series). (a) has bounded derivative \( \lvert 2x\sin\frac{1}{x} - \cos\frac{1}{x} \rvert \le 3 \), so it is Lipschitz. (c) and (d) are monotone.
Level 3: Part C Style
Practice Q5. Let \( f : \mathbb{R} \to \mathbb{R} \). Which statements are true?
(a) If \( f \) is differentiable and \( f' \) is bounded, then \( f \) is uniformly continuous. (b) If \( f \) is uniformly continuous, then \( f \) is bounded. (c) If \( f \) is continuous and periodic, then \( f \) is uniformly continuous. (d) If \( f \) is differentiable everywhere, then \( f' \) has no jump discontinuity.
Answer: (a), (c) and (d)
- (a) True. MVT gives \( \lvert f(x) - f(y) \rvert \le \sup\lvert f' \rvert \cdot \lvert x - y \rvert \).
- (b) False. \( f(x) = x \) is uniformly continuous and unbounded.
- (c) True. With period \( T \), \( f \) is UC on the compact \( [-T, 2T] \); any two points within distance \( \delta < T \) can be translated together into that interval.
- (d) True. Darboux: \( f' \) has the intermediate value property, which a jump would violate.
Practice Q6. Let \( f(x) = x^2 \sin\frac{1}{x} \) for \( x \ne 0 \) and \( f(0) = 0 \), on \( [-1, 1] \). Which statements are true?
(a) \( f \) is differentiable at 0 and \( f'(0) = 0 \). (b) \( f' \) is continuous at 0. (c) \( f \) is of bounded variation on \( [-1, 1] \). (d) \( f' \) is Riemann integrable on \( [-1, 1] \).
Answer: (a), (c) and (d)
- (a) True. \( \frac{f(h)}{h} = h \sin\frac{1}{h} \to 0 \).
- (b) False. For \( x \ne 0 \), \( f'(x) = 2x\sin\frac{1}{x} - \cos\frac{1}{x} \); the second term has no limit at 0.
- (c) True. \( \lvert f' \rvert \le 3 \) everywhere, so \( f \) is Lipschitz on \( [-1, 1] \), hence BV.
- (d) True. \( f' \) is bounded and continuous except at the single point 0, so it is Riemann integrable (Lebesgue’s criterion).
Practice Q7. Let \( f : [0, 1] \to \mathbb{R} \) be monotone increasing. Which statements are true?
(a) The set of discontinuities of \( f \) is countable. (b) \( f \) is Riemann integrable on \( [0, 1] \). (c) \( f \) is continuous at some point of \( (0, 1) \). (d) Every discontinuity of \( f \) in \( (0, 1) \) is removable.
Answer: (a), (b) and (c)
- (a) True. Each jump interval \( (f(c^-), f(c^+)) \) contains its own rational, and these intervals are disjoint.
- (b) True. Monotone functions are Riemann integrable (\( U - L \le (f(1) - f(0)) \cdot \text{mesh} \)).
- (c) True. In fact at uncountably many points: \( (0, 1) \) is uncountable and the discontinuities are countable.
- (d) False. Interior discontinuities of a monotone function are jumps. \( f = 1_{[1/2, 1]} \) has a jump at \( \frac{1}{2} \), not a removable discontinuity.
Quick Revision Table
| Idea | Fact |
|---|---|
| Continuous on \( [a, b] \) | bounded, attains bounds, UC, IVT |
| Monotone function | only jumps, countably many; differentiable a.e. |
| No function | continuous exactly on \( \mathbb{Q} \) |
| UC on \( (a, b) \) | iff continuous extension to \( [a, b] \) |
| Bounded \( f' \) | Lipschitz \( \Rightarrow \) UC and BV |
| Darboux | derivatives have the IVP, no jumps |
| MVT | \( f(b) - f(a) = f'(c)(b - a) \) |
| Rolle count | \( f' \) has \( k \) roots \( \Rightarrow \) \( f \) has at most \( k + 1 \) |
| \( x^a \sin\frac{1}{x} \) | cont. \( a > 0 \); diff. and BV \( a > 1 \); \( C^1 \) \( a > 2 \) |
| BV | \( = \) difference of increasing functions |
| \( C^1 \) on \( [a, b] \) | \( V_a^b(f) = \int_a^b \lvert f' \rvert \) |
Key Takeaways
What’s Next
- Sequences and Series of Functions: when limits of continuous or differentiable functions keep those properties.
- Riemann, Improper and Lebesgue Integration: integrability of monotone and BV functions, and the fundamental theorem.
- Metric Spaces and Several Variables: continuity on compact and connected sets in general.
- Real Analysis: Quick Revision.