Skip to content

Fourier Series, Fourier and Laplace Transforms

CSIR NET Physical Sciences: Fourier series, Dirichlet conditions, Parseval sums, Fourier transform pairs, convolution, Laplace transforms for ODEs, practice.

  • 17 min read
  • Updated
On this page · 13 sections
  1. Prerequisites
  2. The Hook: Why a Short Pulse Has a Broad Spectrum
  3. Fourier Series
  4. The Fourier Transform
  5. The Laplace Transform
  6. Fourier vs Laplace: Which One?
  7. Memory Tricks
  8. Don’t Confuse
  9. Common Mistakes
  10. Practice
  11. Quick Revision Table
  12. Key Takeaways
  13. What’s Next

Prerequisites

  • Linear ODEs and Special Functions: Laplace transforms are a method for exactly those initial-value problems, and Fourier series are another orthogonal expansion.
  • Uniform and pointwise convergence of series of functions, from Sequences and Series of Functions.
  • The Gaussian integral \( \int_{-\infty}^{\infty}e^{-ax^2}dx = \sqrt{\pi/a} \).

The Hook: Why a Short Pulse Has a Broad Spectrum

Fourier Series

Coefficients

A function with period \( 2L \) (or defined on \( (-L, L) \) and extended periodically) has the expansion

$$f(x) = \frac{a_0}{2} + \sum_{n=1}^{\infty}\left[a_n\cos\frac{n\pi x}{L} + b_n\sin\frac{n\pi x}{L}\right]$$

with

$$\boxed{a_n = \frac{1}{L}\int_{-L}^{L}f(x)\cos\frac{n\pi x}{L}\,dx, \qquad b_n = \frac{1}{L}\int_{-L}^{L}f(x)\sin\frac{n\pi x}{L}\,dx}$$

Writing the constant term as \( a_0/2 \) lets the same formula give \( a_0 \); then \( a_0/2 \) is the average of \( f \) over a period.

Symmetry halves the work. An even function has only cosines (\( b_n = 0 \)); an odd function has only sines (\( a_n = 0 \), including \( a_0 \)).

Half-range series. A function given only on \( (0, L) \) can be extended as even (cosine series) or odd (sine series). The odd extension is the one that vanishes at the ends, which is what a string fixed at both ends needs.

Complex form. \( f(x) = \sum_{n=-\infty}^{\infty}c_ne^{in\pi x/L} \) with \( c_n = \frac{1}{2L}\int_{-L}^{L}f(x)e^{-in\pi x/L}dx \). For real \( f \), \( c_{-n} = c_n^* \).

Convergence: the Dirichlet conditions

If \( f \) is periodic, has finitely many maxima, minima and finite jumps in a period, and is absolutely integrable over a period, then its Fourier series converges at every \( x \) to

$$\boxed{\frac{f(x^+) + f(x^-)}{2}}$$

which is \( f(x) \) where \( f \) is continuous and the midpoint of the jump where it is not. “Where it is not” includes the ends of the interval when the periodic extension jumps there: the series for \( f(x) = x \) on \( (-\pi, \pi) \) converges to 0 at \( x = \pm\pi \).

Gibbs phenomenon. Near a jump the partial sums overshoot by about 9% of the jump (0.0895 of the jump on each side), and the overshoot does not shrink as more terms are added; it only moves closer to the jump. The convergence is pointwise, not uniform.

Decay of coefficients. A function with a jump has coefficients falling as \( 1/n \). A continuous function with a kink (jump in \( f' \)) has coefficients falling as \( 1/n^2 \). Smoother functions give faster decay. This is a quick way to reject options.

Standard series on \( (-\pi, \pi) \)

\( f(x) \)SeriesSum it gives
\( x \)\( 2\sum_{n\ge1}\frac{(-1)^{n+1}}{n}\sin nx \)at \( x = \pi/2 \): \( 1 - \tfrac13 + \tfrac15 - \cdots = \tfrac{\pi}{4} \)
Square wave: \( -1 \) on \( (-\pi, 0) \), \( +1 \) on \( (0, \pi) \)\( \frac{4}{\pi}\sum_{n\ge1}\frac{\sin(2n-1)x}{2n-1} \)same Leibniz series
\( \lvert x\rvert \)\( \frac{\pi}{2} - \frac{4}{\pi}\sum_{n\ge1}\frac{\cos(2n-1)x}{(2n-1)^2} \)at \( x = 0 \): \( \sum\frac{1}{(2n-1)^2} = \frac{\pi^2}{8} \)
\( x^2 \)\( \frac{\pi^2}{3} + 4\sum_{n\ge1}\frac{(-1)^n}{n^2}\cos nx \)at \( x = \pi \): \( \sum\frac{1}{n^2} = \frac{\pi^2}{6} \); at 0: \( \sum\frac{(-1)^{n+1}}{n^2} = \frac{\pi^2}{12} \)

Note the decay rates: \( 1/n \) for the discontinuous \( x \) and square wave, \( 1/n^2 \) for the continuous \( \lvert x\rvert \) and \( x^2 \).

Parseval’s theorem

$$\boxed{\frac{1}{L}\int_{-L}^{L}\lvert f(x)\rvert^2dx = \frac{a_0^2}{2} + \sum_{n=1}^{\infty}\left(a_n^2 + b_n^2\right)}$$

The mean-square value of \( f \) is the sum of the powers in each harmonic. It turns any Fourier series into a numerical sum.

The Fourier Transform

Conventions differ. This page uses the symmetric one:

$$\boxed{\tilde f(k) = \frac{1}{\sqrt{2\pi}}\int_{-\infty}^{\infty}f(x)e^{-ikx}dx, \qquad f(x) = \frac{1}{\sqrt{2\pi}}\int_{-\infty}^{\infty}\tilde f(k)e^{ikx}dk}$$

Other books put the whole \( 1/2\pi \) on one side, or use \( e^{-2\pi i\nu x} \). Before matching an option, check which convention the question uses: answers differ by factors of \( \sqrt{2\pi} \).

Standard pairs

\( f(x) \)\( \tilde f(k) \)
\( e^{-ax^2} \)\( \frac{1}{\sqrt{2a}}e^{-k^2/4a} \)
\( e^{-x^2/2} \)\( e^{-k^2/2} \) (its own transform)
\( e^{-a\lvert x\rvert} \)\( \sqrt{\frac{2}{\pi}}\,\frac{a}{a^2 + k^2} \) (a Lorentzian)
\( 1 \) for \( \lvert x\rvert < a \), else 0\( \sqrt{\frac{2}{\pi}}\,\frac{\sin ka}{k} \) (a sinc)
\( \delta(x) \)\( \frac{1}{\sqrt{2\pi}} \)
\( 1 \)\( \sqrt{2\pi}\,\delta(k) \)
\( e^{ik_0x} \)\( \sqrt{2\pi}\,\delta(k - k_0) \)

The delta-function pair contains the key representation \( \delta(x) = \frac{1}{2\pi}\int_{-\infty}^{\infty}e^{ikx}dk \).

Properties

Operation on \( f(x) \)Effect on \( \tilde f(k) \)
Shift \( f(x - a) \)\( e^{-ika}\tilde f(k) \)
Modulation \( e^{ik_0x}f(x) \)\( \tilde f(k - k_0) \)
Scaling \( f(ax) \)\( \frac{1}{\lvert a\rvert}\tilde f(k/a) \)
Derivative \( f'(x) \)\( ik\,\tilde f(k) \)
\( xf(x) \)\( i\,\tilde f'(k) \)
Convolution \( (f*g)(x) = \int f(y)g(x - y)dy \)\( \sqrt{2\pi}\,\tilde f(k)\tilde g(k) \)
Real and even \( f \)real and even \( \tilde f \)

Plancherel (Parseval) theorem: \( \int\lvert f(x)\rvert^2dx = \int\lvert\tilde f(k)\rvert^2dk \) in this convention.

Uncertainty. For any \( f \), the widths of \( \lvert f\rvert^2 \) and \( \lvert\tilde f\rvert^2 \) satisfy \( \Delta x\,\Delta k \ge \tfrac{1}{2} \), with equality only for Gaussians. With \( p = \hbar k \) this is \( \Delta x\,\Delta p \ge \hbar/2 \).

The Laplace Transform

For \( f(t) \) defined for \( t \ge 0 \),

$$\boxed{F(s) = \mathcal L[f](s) = \int_0^\infty f(t)e^{-st}dt}$$

defined for \( \operatorname{Re}s \) large enough for the integral to converge.

Table

\( f(t) \)\( F(s) \)
1\( 1/s \)
\( t^n \)\( n!/s^{n+1} \)
\( t^{-1/2} \)\( \sqrt{\pi/s} \)
\( e^{at} \)\( 1/(s - a) \)
\( \sin\omega t \)\( \omega/(s^2 + \omega^2) \)
\( \cos\omega t \)\( s/(s^2 + \omega^2) \)
\( \sinh\omega t \)\( \omega/(s^2 - \omega^2) \)
\( \cosh\omega t \)\( s/(s^2 - \omega^2) \)
\( \delta(t - a) \), \( a \ge 0 \)\( e^{-as} \)
\( u(t - a) \) (unit step)\( e^{-as}/s \)

Rules

RuleStatement
First shift\( \mathcal L[e^{at}f(t)] = F(s - a) \)
Second shift\( \mathcal L[u(t - a)f(t - a)] = e^{-as}F(s) \)
Derivatives\( \mathcal L[f'] = sF - f(0) \); \( \mathcal L[f''] = s^2F - sf(0) - f'(0) \)
Integral\( \mathcal L\left[\int_0^tf(\tau)d\tau\right] = F(s)/s \)
Multiply by \( t \)\( \mathcal L[tf(t)] = -F'(s) \)
Divide by \( t \)\( \mathcal L[f(t)/t] = \int_s^\infty F(\sigma)d\sigma \)
Convolution\( \mathcal L\left[\int_0^tf(\tau)g(t - \tau)d\tau\right] = F(s)G(s) \)
Initial value\( f(0^+) = \lim_{s\to\infty}sF(s) \)
Final value\( f(\infty) = \lim_{s\to0}sF(s) \), only if all poles of \( sF \) lie in the left half-plane

The derivative rule is the whole point: it turns a linear ODE with initial conditions into an algebraic equation for \( F(s) \), with the initial conditions built in.

Inverse transforms are found by partial fractions and completing the square, then reading the table backwards. Example: \( \dfrac{1}{s^2 + 2s + 5} = \dfrac{1}{(s + 1)^2 + 4} \), which by the first shift is \( \tfrac{1}{2}e^{-t}\sin2t \). The general inversion integral (Bromwich) is a contour integral evaluated by residues; see Complex Analysis and Residues.

Fourier vs Laplace: Which One?

graph TD
    A[Problem] --> B{Domain}
    B -->|periodic, finite interval| C[Fourier series]
    B -->|whole real line, decaying| D[Fourier transform]
    B -->|t from 0, initial conditions given| E[Laplace transform]
    C --> F[Parseval: mean square equals sum of coefficient squares]
    D --> G[Derivative becomes ik; convolution becomes product]
    E --> H[Derivative becomes s F minus initial values]

Memory Tricks

Don’t Confuse

ThisvsThat
Fourier series value at a jump: midpointValue of \( f \) there (may be defined differently)
Pointwise convergence with Gibbs overshootUniform convergence (needs \( f \) continuous and periodic, with absolutely summable coefficients or similar smoothness)
\( \mathcal L[e^{at}f] = F(s - a) \) (first shift)\( \mathcal L[f(t - a)u(t - a)] = e^{-as}F \) (second shift)
\( \mathcal L[\sin\omega t] = \omega/(s^2 + \omega^2) \)\( \mathcal L[\sinh\omega t] = \omega/(s^2 - \omega^2) \)
Fourier convolution: \( \int_{-\infty}^{\infty} \)Laplace convolution: \( \int_0^t \)
Final value theorem for decaying \( f \)Not valid for \( \sin t \) or \( e^{t} \): poles on or right of the imaginary axis

Common Mistakes

Practice

All questions below are original practice questions written for this page. They are not from any CSIR NET paper.

Level 1: Foundation

  1. An even function on \( (-\pi, \pi) \) has which Fourier coefficients equal to zero?
  2. Write \( \mathcal L[t^2] \).
  3. To what value does the Fourier series of \( f(x) = x \) on \( (-\pi, \pi) \) converge at \( x = \pi \)?
Answers
  1. All \( b_n \). Only cosines (and the constant) survive.
  2. \( 2/s^3 \).
  3. 0. The periodic extension jumps from \( \pi \) to \( -\pi \); the midpoint is 0.

Level 2: Part B Style

Practice Q1. Using the Fourier series of \( \lvert x\rvert \) on \( (-\pi, \pi) \), the value of \( 1 + \frac{1}{3^2} + \frac{1}{5^2} + \cdots \) is:

(a) \( \pi^2/6 \) (b) \( \pi^2/8 \) (c) \( \pi^2/12 \) (d) \( \pi/4 \)

Answer: (b) π²/8

\( \lvert x\rvert = \frac{\pi}{2} - \frac{4}{\pi}\sum\frac{\cos(2n-1)x}{(2n-1)^2} \). At \( x = 0 \) (where \( \lvert x\rvert \) is continuous): \( 0 = \frac{\pi}{2} - \frac{4}{\pi}S \), so \( S = \pi^2/8 \). Option (a) is the sum over all \( n \); (c) the alternating sum; (d) the Leibniz series.

Practice Q2. The inverse Laplace transform of \( \dfrac{1}{s^2 + 2s + 5} \) is:

(a) \( e^{-t}\sin2t \) (b) \( \tfrac{1}{2}e^{-t}\sin2t \) (c) \( \tfrac{1}{2}e^{t}\sin2t \) (d) \( e^{-t}\cos2t \)

Answer: (b)

\( s^2 + 2s + 5 = (s + 1)^2 + 4 \). Since \( \mathcal L[\sin2t] = 2/(s^2 + 4) \), \( \mathcal L^{-1}[1/(s^2 + 4)] = \tfrac{1}{2}\sin2t \), and the shift \( s \to s + 1 \) multiplies by \( e^{-t} \). Option (a) loses the \( \tfrac{1}{2} \); (c) has the shift sign wrong.

Practice Q3. With \( \tilde f(k) = \frac{1}{\sqrt{2\pi}}\int f(x)e^{-ikx}dx \), the transform of \( e^{-\lvert x\rvert} \) is:

(a) \( \sqrt{\frac{2}{\pi}}\frac{1}{1 + k^2} \) (b) \( \frac{1}{1 + k^2} \) (c) \( e^{-k^2/2} \) (d) \( \sqrt{\frac{2}{\pi}}\frac{k}{1 + k^2} \)

Answer: (a)

\( \int_{-\infty}^{\infty}e^{-\lvert x\rvert}e^{-ikx}dx = \frac{1}{1 + ik} + \frac{1}{1 - ik} = \frac{2}{1 + k^2} \); divide by \( \sqrt{2\pi} \). Option (b) has a convention-dependent factor missing; (d) would be the transform of an odd function, but \( e^{-\lvert x\rvert} \) is even, so its transform is real and even.

Practice Q4. The square wave equal to \( -1 \) on \( (-\pi, 0) \) and \( +1 \) on \( (0, \pi) \) is expanded in a Fourier series. Which statement is correct?

(a) The series has only cosine terms (b) The coefficients decay as \( 1/n^2 \) (c) The partial sums overshoot near \( x = 0 \) by about 9% of the jump, however many terms are taken (d) The series converges uniformly on \( (-\pi, \pi) \)

Answer: (c)

The function is odd, so only sines appear (not (a)); it has jumps, so \( b_n = \frac{4}{\pi n} \) for odd \( n \) decays as \( 1/n \) (not (b)); a uniform limit of continuous partial sums would be continuous, so (d) is false. The Gibbs overshoot is about 0.0895 of the jump of 2, i.e. the partial sums peak near 1.18.

Level 3: Part C Style

Practice Q5. The Fourier series of \( f(x) = x^2 \) on \( (-\pi, \pi) \) is \( \frac{\pi^2}{3} + 4\sum_{n\ge1}\frac{(-1)^n}{n^2}\cos nx \). Which of the following follow from it?

  1. \( \sum_{n\ge1}\frac{1}{n^2} = \frac{\pi^2}{6} \).
  2. \( \sum_{n\ge1}\frac{(-1)^{n+1}}{n^2} = \frac{\pi^2}{12} \).
  3. \( \sum_{n\ge1}\frac{1}{n^4} = \frac{\pi^4}{90} \).
  4. The series converges uniformly on \( [-\pi, \pi] \).

(a) 1 and 2 only (b) 1, 2 and 3 only (c) 3 and 4 only (d) all four

Answer: (d) all four

Statement 1: at \( x = \pi \), \( \cos n\pi = (-1)^n \): \( \pi^2 = \frac{\pi^2}{3} + 4\sum\frac{1}{n^2} \). True.

Statement 2: at \( x = 0 \): \( 0 = \frac{\pi^2}{3} + 4\sum\frac{(-1)^n}{n^2} \), so \( \sum\frac{(-1)^{n+1}}{n^2} = \frac{\pi^2}{12} \). True.

Statement 3: Parseval. \( \frac{1}{\pi}\int_{-\pi}^{\pi}x^4dx = \frac{2\pi^4}{5} \); \( a_0 = \frac{2\pi^2}{3} \) gives \( \frac{a_0^2}{2} = \frac{2\pi^4}{9} \); \( a_n^2 = \frac{16}{n^4} \). So \( 16\sum\frac{1}{n^4} = \frac{2\pi^4}{5} - \frac{2\pi^4}{9} = \frac{8\pi^4}{45} \), giving \( \frac{\pi^4}{90} \). True.

Statement 4: the periodic extension of \( x^2 \) is continuous (\( f(-\pi) = f(\pi) \)), and \( \sum\lvert a_n\rvert = 4\sum\frac{1}{n^2} < \infty \), so the Weierstrass M-test gives uniform convergence. True.

Practice Q6. Solve \( y'' + y = 1 \) for \( t \ge 0 \) with \( y(0) = y'(0) = 0 \) by Laplace transform. Which of the following are correct?

  1. \( Y(s) = \dfrac{1}{s(s^2 + 1)} \).
  2. \( y(t) = 1 - \cos t \).
  3. The final value theorem gives \( \lim_{t\to\infty}y(t) = 1 \).
  4. The maximum value of \( y \) is 2.

(a) 1 and 2 only (b) 1, 2 and 4 only (c) 1, 2 and 3 only (d) all four

Answer: (b) 1, 2 and 4 only

Statement 1: \( s^2Y + Y = 1/s \). True.

Statement 2: \( \frac{1}{s(s^2 + 1)} = \frac{1}{s} - \frac{s}{s^2 + 1} \), so \( y = 1 - \cos t \). True.

Statement 3: \( sY = 1/(s^2 + 1) \to 1 \), but \( sY \) has poles at \( \pm i \) on the imaginary axis, so the theorem does not apply; \( y \) oscillates between 0 and 2 forever. False.

Statement 4: \( 1 - \cos t \) reaches 2 at \( t = \pi \). True.

Practice Q7. Let \( f(x) = e^{-x^2/2} \) and \( g(x) = f(x - 3) \), with the symmetric Fourier convention. Which of the following are correct?

  1. \( \tilde f(k) = e^{-k^2/2} \).
  2. \( \lvert\tilde g(k)\rvert = \lvert\tilde f(k)\rvert \).
  3. \( \int_{-\infty}^{\infty}\lvert\tilde g(k)\rvert^2dk = \sqrt\pi \).
  4. The convolution \( f*f \) is a Gaussian of twice the variance of \( f \).

(a) 1 and 2 only (b) 1, 2 and 3 only (c) 1, 2 and 4 only (d) all four

Answer: (d) all four

Statement 1: the standard self-reciprocal Gaussian. True.

Statement 2: a shift multiplies the transform by \( e^{-3ik} \), a pure phase. True.

Statement 3: Plancherel: \( \int\lvert g\rvert^2dx = \int e^{-(x-3)^2}dx = \sqrt\pi \). True.

Statement 4: the transform of \( f*f \) is \( \sqrt{2\pi}\,\tilde f^2 = \sqrt{2\pi}\,e^{-k^2} \), and inverting gives \( \sqrt{\pi}\,e^{-x^2/4} \): a Gaussian with variance 2, twice the variance 1 of \( f \) (read as a distribution shape). This is the rule that variances add under convolution, which reappears in Probability Distributions and the Central Limit Theorem. True.

Quick Revision Table

ItemResult
Coefficients on \( (-L, L) \)\( a_n, b_n = \frac{1}{L}\int f\cos, \frac{1}{L}\int f\sin \); constant \( a_0/2 \)
At a jump\( \frac{1}{2}[f(x^+) + f(x^-)] \)
Gibbsovershoot about 9% of jump, does not vanish
Parseval (series)\( \frac{1}{L}\int\lvert f\rvert^2 = \frac{a_0^2}{2} + \sum(a_n^2 + b_n^2) \)
Sums\( \frac{\pi^2}{6}, \frac{\pi^2}{8}, \frac{\pi^2}{12}, \frac{\pi^4}{90} \)
Gaussian\( e^{-ax^2} \to \frac{1}{\sqrt{2a}}e^{-k^2/4a} \)
Lorentzian\( e^{-a\lvert x\rvert} \to \sqrt{2/\pi}\,a/(a^2 + k^2) \)
FT rules\( f' \to ik\tilde f \); shift \( \to \) phase; convolution \( \to \) product
Laplace\( \mathcal L[f'] = sF - f(0) \); \( \mathcal L[e^{at}f] = F(s - a) \)
Useful integrals\( \int_0^\infty\frac{\sin t}{t}dt = \frac{\pi}{2} \); \( \int\frac{\sin^2k}{k^2}dk = \pi \)

Key Takeaways

What’s Next