Sequences and Series of Functions
CSIR NET real analysis: pointwise vs uniform convergence, M-test, Dini, interchange of limits, Weierstrass approximation, power series, radius of convergence.
- 18 min
On this page · 14 sections
Prerequisites
- Sequences, Series, limsup and liminf: Cauchy criterion, limsup and the root test, all reused here with a “sup over \( x \)” attached.
- Continuity, Differentiability and Bounded Variation: the properties we want limits to keep.
- Riemann integration is defined in Riemann, Improper and Lebesgue Integration; here we only use that continuous functions on \( [a, b] \) are integrable.
The Hook: Continuous Functions with a Broken Limit
Pointwise and Uniform Convergence
Let \( f_n, f : D \to \mathbb{R} \).
- Pointwise: \( f_n(x) \to f(x) \) for each \( x \in D \). The \( N \) may depend on \( x \) and \( \varepsilon \).
- Uniform: for every \( \varepsilon > 0 \) there is \( N \) with \( \lvert f_n(x) - f(x) \rvert < \varepsilon \) for all \( n \ge N \) and all \( x \in D \).
Cauchy criterion. \( (f_n) \) converges uniformly on \( D \) iff for every \( \varepsilon > 0 \) there is \( N \) with \( \sup_D \lvert f_n - f_m \rvert < \varepsilon \) for \( m, n \ge N \).
The routine
- Find the pointwise limit \( f \).
- Compute \( M_n = \sup_D \lvert f_n - f \rvert \), usually by calculus (set the derivative to zero) or by evaluating at a clever point.
- \( M_n \to 0 \) means uniform. To disprove uniformity, one point \( x_n \) with \( \lvert f_n(x_n) - f(x_n) \rvert \ge \varepsilon_0 \) suffices.
| \( f_n(x) \) | Domain | Pointwise limit | \( M_n \) | Uniform? |
|---|---|---|---|---|
| \( x^n \) | \( [0, 1] \) | 0 on \( [0,1) \), 1 at 1 | 1 | no |
| \( x^n \) | \( [0, a] \), \( a < 1 \) | 0 | \( a^n \) | yes |
| \( x^n \) | \( [0, 1) \) | 0 | 1 | no |
| \( \frac{nx}{1 + n^2x^2} \) | \( [0, 1] \) | 0 | \( \frac{1}{2} \) at \( x = \frac{1}{n} \) | no |
| \( \frac{x}{1 + nx^2} \) | \( \mathbb{R} \) | 0 | \( \frac{1}{2\sqrt{n}} \) at \( x = \frac{1}{\sqrt{n}} \) | yes |
| \( \frac{\sin nx}{\sqrt{n}} \) | \( \mathbb{R} \) | 0 | \( \frac{1}{\sqrt{n}} \) | yes |
| \( nxe^{-nx^2} \) | \( [0, 1] \) | 0 | \( \sqrt{\frac{n}{2e}} \) | no |
| \( \frac{x}{n} \) | \( \mathbb{R} \) | 0 | \( \infty \) | no (yes on bounded sets) |
| \( \left(1 + \frac{x}{n}\right)^n \) | \( [0, 1] \) | \( e^x \) | \( \to 0 \) | yes |
What Uniform Convergence Preserves
Uniform convergence of \( f_n \) alone does not give the third: \( \frac{\sin nx}{\sqrt{n}} \to 0 \) uniformly, but \( f_n'(x) = \sqrt{n}\cos nx \) diverges at \( x = 0 \). And \( \frac{x}{1 + nx^2} \to 0 \) uniformly on \( \mathbb{R} \), yet \( f_n'(0) = 1 \) for every \( n \), while the limit function has derivative 0.
Uniform convergence is sufficient, not necessary. \( x^n \to f \) non-uniformly on \( [0, 1] \), yet \( \int_0^1 x^n = \frac{1}{n+1} \to 0 = \int_0^1 f \). Pointwise convergence plus a uniform bound gives the integral interchange too (bounded convergence theorem; see Lebesgue integration).
Also preserved by uniform limits: boundedness (for each \( f_n \) bounded), uniform continuity.
Dini’s theorem
If \( K \) is compact, \( f_n : K \to \mathbb{R} \) are continuous, \( f_n \to f \) pointwise with \( f \) continuous, and \( f_n(x) \ge f_{n+1}(x) \) for all \( x, n \) (or all increasing), then \( f_n \to f \) uniformly.
Each hypothesis is needed. Drop compactness: \( x^n \) on \( [0, 1) \) decreases to the continuous 0, not uniformly. Drop continuity of the limit: \( x^n \) on \( [0, 1] \). Drop monotonicity: \( \frac{nx}{1 + n^2 x^2} \) on \( [0, 1] \).
Series of Functions
\( \sum f_n \) converges uniformly if its partial sums do.
Weierstrass M-test. If \( \lvert f_n(x) \rvert \le M_n \) for all \( x \in D \) and \( \sum M_n < \infty \), then \( \sum f_n \) converges absolutely and uniformly on \( D \).
Uniform Dirichlet and Abel tests. If the partial sums of \( \sum a_n(x) \) are uniformly bounded on \( D \) and \( b_n(x) \) decreases to 0 uniformly on \( D \), then \( \sum a_n b_n \) converges uniformly. (Abel: \( \sum a_n(x) \) uniformly convergent, \( b_n(x) \) monotone in \( n \) and uniformly bounded.)
Necessary condition. If \( \sum f_n \) converges uniformly on \( D \), then \( \sup_D \lvert f_n \rvert \to 0 \).
| Series | Behaviour |
|---|---|
| \( \sum \frac{\sin nx}{n^2} \) | uniform on \( \mathbb{R} \) (M-test, \( M_n = \frac{1}{n^2} \)); sum continuous |
| \( \sum \frac{\sin nx}{n} \) | converges for every \( x \); uniform on \( [\delta, 2\pi - \delta] \) (Dirichlet), not on \( \mathbb{R} \); sum is \( \frac{\pi - x}{2} \) on \( (0, 2\pi) \), 0 at 0 |
| \( \sum \frac{(-1)^n}{n + x^2} \) | uniform on \( \mathbb{R} \) (alternating remainder \( \le \frac{1}{n+1} \)) but absolutely convergent at no \( x \) |
| \( \sum_{n \ge 0} \frac{x^2}{(1 + x^2)^n} \) | sum is \( 1 + x^2 \) for \( x \ne 0 \), 0 at 0: not uniform near 0 |
| \( \sum \frac{1}{n^x} \) | uniform on \( [a, \infty) \) for \( a > 1 \); not on \( (1, \infty) \) |
Two big theorems
Weierstrass approximation theorem. Every continuous \( f : [a, b] \to \mathbb{R} \) is a uniform limit of polynomials. (On an unbounded interval this fails: a uniform limit of polynomials on \( \mathbb{R} \) is a polynomial, so \( \sin x \) is not one.)
A continuous nowhere-differentiable function. Let \( \varphi(x) = \lvert x \rvert \) on \( [-1, 1] \), extended to \( \mathbb{R} \) with period 2. Then \( f(x) = \sum_{n \ge 0} \left(\frac{3}{4}\right)^n \varphi(4^n x) \) converges uniformly (M-test), so \( f \) is continuous, and it is differentiable nowhere (Rudin, Chapter 7). Weierstrass’s original example is a cosine series of the same type.
Arzelà–Ascoli. A sequence in \( C[a, b] \) that is uniformly bounded and equicontinuous has a uniformly convergent subsequence. (A sequence with uniformly bounded derivatives is equicontinuous, by MVT.)
Power Series
\( \sum_{n \ge 0} a_n (x - c)^n \) has a radius of convergence \( R \in [0, \infty] \): it converges absolutely for \( \lvert x - c \rvert < R \) and diverges for \( \lvert x - c \rvert > R \).
$$\boxed{\frac{1}{R} = \limsup_{n\to\infty} \lvert a_n \rvert^{1/n}} \quad \text{(Cauchy–Hadamard)}, \qquad R = \lim_{n\to\infty} \left\lvert \frac{a_n}{a_{n+1}} \right\rvert \text{ if this limit exists.}$$| Series | \( R \) |
|---|---|
| \( \sum x^n \), \( \sum n^k x^n \), \( \sum \frac{x^n}{n^k} \) | 1 |
| \( \sum \frac{x^n}{n!} \) | \( \infty \) |
| \( \sum n! \, x^n \) | 0 |
| \( \sum \frac{n!}{n^n} x^n \) | \( e \) |
| \( \sum \left(1 + \frac{1}{n}\right)^{n^2} x^n \) | \( \frac{1}{e} \) |
| \( \sum 2^n x^{2n} \) | \( \frac{1}{\sqrt{2}} \) |
| \( \sum x^{n^2} \) | 1 |
| \( \sum \left(2 + (-1)^n\right)^n x^n \) | \( \frac{1}{3} \) |
Properties inside the radius
- A power series converges uniformly on every closed interval \( [c - r, c + r] \) with \( r < R \), but not necessarily on \( (c - R, c + R) \) (\( \sum x^n \) on \( (-1, 1) \)).
- Its sum is infinitely differentiable on \( (c - R, c + R) \), and may be differentiated and integrated term by term. The radius does not change, though behaviour at the endpoints can.
- \( a_n = \frac{f^{(n)}(c)}{n!} \): the series is the Taylor series of its sum. Two power series that agree on a set with a limit point inside the interval are identical.
At the endpoints
| Series | \( x = 1 \) | \( x = -1 \) |
|---|---|---|
| \( \sum x^n \) | diverges | diverges |
| \( \sum \frac{x^n}{n} \) | diverges | converges (conditionally) |
| \( \sum \frac{x^n}{n^2} \) | converges | converges |
| \( \sum \frac{(-1)^n x^n}{n} \) | converges | diverges |
Abel’s theorem. If \( \sum a_n \) converges, then \( \sum a_n x^n \) converges uniformly on \( [0, 1] \) and \( \lim_{x \to 1^-} \sum a_n x^n = \sum a_n \). This evaluates endpoint sums:
$$1 - \frac{1}{2} + \frac{1}{3} - \dots = \log 2, \qquad 1 - \frac{1}{3} + \frac{1}{5} - \dots = \frac{\pi}{4}.$$The converse fails: \( \lim_{x \to 1^-} \sum (-1)^n x^n = \frac{1}{2} \), but \( \sum (-1)^n \) diverges.
Counter-examples Table
| False claim | Counter-example |
|---|---|
| Pointwise limit of continuous functions is continuous | \( x^n \) on \( [0, 1] \) |
| Pointwise convergence allows \( \lim \int = \int \lim \) | \( nxe^{-nx^2} \) on \( [0, 1] \): integrals \( \to \frac{1}{2} \) |
| Uniform convergence of \( f_n \) gives \( f_n' \to f' \) | \( \frac{\sin nx}{\sqrt{n}} \); \( \frac{x}{1 + nx^2} \) at 0 |
| Uniform convergence is needed for \( \lim \int = \int \lim \) | \( x^n \) on \( [0, 1] \) (integrals still converge to 0) |
| Continuous limit \( \Rightarrow \) uniform convergence | \( \frac{nx}{1 + n^2x^2} \) on \( [0, 1] \) |
| Dini without compactness | \( x^n \) on \( [0, 1) \) |
| Uniform convergence \( \Rightarrow \) absolute convergence | \( \sum \frac{(-1)^n}{n + x^2} \) |
| A power series converges uniformly on its whole open interval | \( \sum x^n \) on \( (-1, 1) \) |
| Termwise differentiation keeps endpoint convergence | \( \sum \frac{x^n}{n^2} \) converges at 1; \( \sum \frac{x^{n-1}}{n} \) does not |
| \( \lim_{x \to 1^-} \sum a_n x^n \) exists \( \Rightarrow \sum a_n \) converges | \( a_n = (-1)^n \) |
| Polynomials approximate continuous functions uniformly on \( \mathbb{R} \) | \( \sin x \), \( e^x \) |
| Continuous \( \Rightarrow \) differentiable somewhere | \( \sum (\frac{3}{4})^n \varphi(4^n x) \) |
Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| uniform on \( [0, a] \) for every \( a < 1 \) | uniform on \( [0, 1) \) (often false) | |
| M-test (sufficient, gives absolute too) | uniform convergence (can hold without absolute) | |
| radius of \( f \), \( f' \), \( \int f \): equal | endpoint behaviour of \( f \), \( f' \), \( \int f \): can differ | |
| Abel’s theorem (series converges \( \Rightarrow \) radial limit) | its converse (false) | |
| Weierstrass approximation (on \( [a, b] \)) | approximation on \( \mathbb{R} \) (fails) |
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
- Does \( x^n \) converge uniformly on \( [0, 1] \)? On \( [0, \frac{1}{2}] \)?
- Find the radius of convergence of \( \sum n! \, x^n \).
- State the Weierstrass M-test.
Answers
- No on \( [0, 1] \) (discontinuous limit). Yes on \( [0, \frac{1}{2}] \): \( M_n = 2^{-n} \to 0 \).
- \( R = 0 \). \( \frac{a_n}{a_{n+1}} = \frac{1}{n+1} \to 0 \).
- If \( \lvert f_n(x) \rvert \le M_n \) on \( D \) and \( \sum M_n < \infty \), then \( \sum f_n \) converges absolutely and uniformly on \( D \).
Level 2: Part B Style
Practice Q1. The radius of convergence of \( \sum_{n \ge 1} \left(1 + \frac{1}{n}\right)^{n^2} x^n \) is:
(a) 1 (b) \( e \) (c) \( \frac{1}{e} \) (d) 0
Answer: (c)
\( \lvert a_n \rvert^{1/n} = \left(1 + \frac{1}{n}\right)^n \to e \), so \( R = \frac{1}{e} \).
Practice Q2. Let \( f_n(x) = \frac{nx}{1 + n^2 x^2} \) on \( [0, 1] \). Then \( (f_n) \):
(a) converges uniformly to 0 (b) converges pointwise to 0, but not uniformly (c) does not converge pointwise (d) converges uniformly to a non-zero function
Answer: (b)
\( f_n(0) = 0 \); for \( x > 0 \), \( f_n(x) \le \frac{nx}{n^2x^2} = \frac{1}{nx} \to 0 \). But \( f_n(\frac{1}{n}) = \frac{1}{2} \) for all \( n \), so \( M_n \ge \frac{1}{2} \).
Practice Q3. The radius of convergence of \( \sum_{n \ge 0} 2^n x^{2n} \) is:
(a) 2 (b) \( \frac{1}{2} \) (c) \( \sqrt{2} \) (d) \( \frac{1}{\sqrt{2}} \)
Answer: (d)
With \( y = x^2 \), \( \sum (2y)^n \) converges iff \( \lvert y \rvert < \frac{1}{2} \), so \( \lvert x \rvert < \frac{1}{\sqrt{2}} \). Option (b) is the radius in \( y \), not in \( x \).
Practice Q4. \( \sum_{n=1}^{\infty} \frac{n}{2^n} \) equals:
(a) 1 (b) \( \frac{3}{2} \) (c) 2 (d) 4
Answer: (c)
\( \sum n x^n = \frac{x}{(1-x)^2} \) for \( \lvert x \rvert < 1 \); at \( x = \frac{1}{2} \): \( \frac{1/2}{1/4} = 2 \).
Level 3: Part C Style
Practice Q5. Let \( f_n(x) = x^n \). Which statements are true?
(a) \( (f_n) \) converges pointwise on \( [0, 1] \). (b) \( (f_n) \) converges uniformly on \( [0, 1) \). (c) \( (f_n) \) converges uniformly on \( [0, a] \) for every \( a \in (0, 1) \). (d) \( \lim_{n \to \infty} \int_0^1 f_n(x)\,dx = \int_0^1 \lim_{n\to\infty} f_n(x)\,dx \).
Answer: (a), (c) and (d)
- (a) True. Limit 0 on \( [0, 1) \), 1 at 1.
- (b) False. \( \sup_{[0,1)} x^n = 1 \) for every \( n \).
- (c) True. \( \sup_{[0,a]} x^n = a^n \to 0 \).
- (d) True. \( \int_0^1 x^n = \frac{1}{n+1} \to 0 \), and the limit function is 0 except at one point, so its integral is 0. The interchange holds without uniform convergence.
Practice Q6. Consider \( F(x) = \sum_{n \ge 1} \frac{x^n}{n^2} \). Which statements are true?
(a) The radius of convergence is 1. (b) The series converges uniformly on \( [-1, 1] \). (c) \( F \) is differentiable on \( (-1, 1) \). (d) The term-by-term derivative series converges at \( x = 1 \).
Answer: (a), (b) and (c)
- (a) True. \( \left(\frac{1}{n^2}\right)^{1/n} \to 1 \).
- (b) True. M-test with \( M_n = \frac{1}{n^2} \) on \( [-1, 1] \).
- (c) True. Power series are differentiable inside the interval of convergence.
- (d) False. The derivative series is \( \sum \frac{x^{n-1}}{n} \), which at \( x = 1 \) is the harmonic series.
Practice Q7. Let \( f_n(x) = \frac{x}{1 + nx^2} \) on \( \mathbb{R} \). Which statements are true?
(a) \( f_n \to 0 \) uniformly on \( \mathbb{R} \). (b) \( (f_n') \) converges pointwise on \( \mathbb{R} \). (c) \( (f_n') \) converges uniformly on \( \mathbb{R} \). (d) \( \left(\lim_n f_n\right)'(0) = \lim_n f_n'(0) \).
Answer: (a) and (b)
- (a) True. \( \lvert f_n \rvert \) is maximised at \( x = \frac{1}{\sqrt{n}} \) with value \( \frac{1}{2\sqrt{n}} \to 0 \).
- (b) True. \( f_n'(x) = \frac{1 - nx^2}{(1 + nx^2)^2} \): equals 1 at \( x = 0 \) and tends to 0 for \( x \ne 0 \).
- (c) False. The pointwise limit of \( f_n' \) is discontinuous at 0, while each \( f_n' \) is continuous.
- (d) False. The limit function is 0, with derivative 0, but \( f_n'(0) = 1 \) for every \( n \).
Quick Revision Table
| Idea | Fact |
|---|---|
| Uniform convergence | \( \sup_D \lvert f_n - f \rvert \to 0 \) |
| Preserves | continuity, integrals, boundedness |
| Derivatives | need \( f_n' \) uniform + \( f_n(x_0) \) convergent |
| Dini | compact, continuous, monotone, continuous limit \( \Rightarrow \) uniform |
| M-test | \( \lvert f_n \rvert \le M_n \), \( \sum M_n < \infty \) |
| Weierstrass approximation | polynomials dense in \( C[a, b] \) |
| Radius | \( \frac{1}{R} = \limsup \lvert a_n \rvert^{1/n} \) |
| Inside radius | uniform on compact subsets; termwise \( d/dx \), \( \int \) with same \( R \) |
| Abel | \( \sum a_n \) converges \( \Rightarrow \) \( \lim_{x \to 1^-} \sum a_n x^n = \sum a_n \) |
| Useful sums | \( \sum nx^n = \frac{x}{(1-x)^2} \), \( \sum \frac{x^n}{n} = -\log(1 - x) \) |
Key Takeaways
What’s Next
- Riemann, Improper and Lebesgue Integration: the monotone and dominated convergence theorems, which replace uniform convergence for integrals.
- Metric Spaces and Several Variables: \( C[a, b] \) with the sup norm as a complete metric space.
- Continuity, Differentiability and Bounded Variation for the single-function versions.
- Real Analysis: Quick Revision.