Fourier Transforms and Their Applications

The Fourier transform answers a simple question with enormous consequences: which frequencies make up a signal? It converts a function of time into a function of frequency, revealing structure that is invisible in the original. From mobile phones and MRI scanners to solving differential equations, it is one of the most widely used tools in applied mathematics.

From Fourier Series to Fourier Transforms

A periodic function can be written as a Fourier series, a sum of sines and cosines at frequencies that are whole-number multiples of a fundamental. But most real signals, such as a single spoken word or a brief pulse, are not periodic. If we let the period grow without limit, the separate frequencies merge into a continuous spectrum, and the series becomes an integral. That integral is the Fourier transform.

The Definition

We use the common convention with angular frequency \( \omega \):

$$ F(\omega) = \int_{-\infty}^{\infty} f(t)\, e^{-i\omega t}\, dt $$

and the inverse transform recovers the original function:

$$ f(t) = \frac{1}{2\pi}\int_{-\infty}^{\infty} F(\omega)\, e^{i\omega t}\, d\omega $$

Since \( e^{-i\omega t} = \cos\omega t - i\sin\omega t \), the transform measures how strongly \( f \) correlates with oscillations at each frequency \( \omega \). The magnitude \( |F(\omega)| \) is the amplitude spectrum. Other textbooks place the factor \( \frac{1}{2\pi} \) differently or use ordinary frequency \( f \) in hertz, so always check the convention before comparing formulas.

Worked Example 1: The Rectangular Pulse

Find the Fourier transform of \( f(t) = 1 \) for \( |t| < a \) and \( f(t) = 0 \) otherwise.

Step 1. The function is zero outside \( [-a, a] \), so

$$ F(\omega) = \int_{-a}^{a} e^{-i\omega t}\, dt $$

Step 2. Integrate (for \( \omega \neq 0 \)):

$$ F(\omega) = \left[ \frac{e^{-i\omega t}}{-i\omega} \right]_{-a}^{a} = \frac{e^{-i\omega a} - e^{i\omega a}}{-i\omega} $$

Step 3. Use \( e^{i\theta} - e^{-i\theta} = 2i\sin\theta \):

$$ F(\omega) = \frac{-2i\sin(\omega a)}{-i\omega} = \frac{2\sin(a\omega)}{\omega} $$

Step 4. At \( \omega = 0 \), the integral is simply the area \( 2a \). This matches the limit of \( \frac{2\sin(a\omega)}{\omega} \), which is \( 2a \).

Interpretation: the first zero of the spectrum is at \( \omega = \frac{\pi}{a} \). A narrow pulse (small \( a \)) has a wide spectrum, while a long pulse has a narrow one. Short events need many frequencies to describe them.

Worked Example 2: The Two-Sided Exponential

Find the transform of \( f(t) = e^{-a|t|} \) with \( a > 0 \).

Step 1. Split the integral at zero:

$$ F(\omega) = \int_{-\infty}^{0} e^{at}e^{-i\omega t}\, dt + \int_{0}^{\infty} e^{-at}e^{-i\omega t}\, dt $$

Step 2. Each integral is a simple exponential:

$$ \int_{-\infty}^{0} e^{(a - i\omega)t}\, dt = \frac{1}{a - i\omega}, \qquad \int_{0}^{\infty} e^{-(a + i\omega)t}\, dt = \frac{1}{a + i\omega} $$

Step 3. Add the fractions:

$$ F(\omega) = \frac{(a + i\omega) + (a - i\omega)}{a^2 + \omega^2} = \frac{2a}{a^2 + \omega^2} $$

Check: \( F(0) = \frac{2}{a} \), which equals the total area \( \int e^{-a|t|}\, dt = 2 \times \frac{1}{a} \). The result is real and even, as it must be for a real, even function. This curve, called a Lorentzian, appears in spectroscopy as the shape of spectral lines.

A one-sided version, \( f(t) = e^{-at} \) for \( t \ge 0 \) and zero before, has transform \( \frac{1}{a + i\omega} \). This is exactly the frequency response of a simple RC low-pass filter, with \( a = \frac{1}{RC} \).

Key Properties

The transform is powerful because of a few simple rules:

• Linearity: the transform of \( \alpha f + \beta g \) is \( \alpha F + \beta G \).
• Time shift: \( f(t - t_0) \) transforms to \( e^{-i\omega t_0}F(\omega) \). Delaying a signal changes only its phase.
• Scaling: \( f(at) \) transforms to \( \frac{1}{|a|}F\!\left(\frac{\omega}{a}\right) \). Compressing in time stretches in frequency.
• Derivatives: \( f'(t) \) transforms to \( i\omega F(\omega) \). Differentiation becomes multiplication.
• Convolution: \( (f * g)(t) \) transforms to \( F(\omega)G(\omega) \). A difficult integral becomes a simple product.
• Parseval's theorem: \( \int |f(t)|^2\, dt = \frac{1}{2\pi}\int |F(\omega)|^2\, d\omega \). Energy is the same in both domains.

The derivative rule is why Fourier transforms solve linear differential equations: derivatives turn into powers of \( i\omega \), and the PDE becomes an algebraic equation.

The Discrete Fourier Transform

Computers work with samples, not continuous functions. For \( N \) samples \( x_0, \dots, x_{N-1} \), the discrete Fourier transform (DFT) is

$$ X_k = \sum_{n=0}^{N-1} x_n\, e^{-2\pi i kn/N}, \qquad k = 0, 1, \dots, N - 1 $$

Worked Example 3: A Four-Point DFT

Compute the DFT of \( x = (1, 2, 3, 4) \).

Step 1. With \( N = 4 \), the factor \( e^{-2\pi i/4} = -i \). So \( X_k = \sum x_n(-i)^{kn} \).

Step 2. \( k = 0 \): all factors equal 1, so \( X_0 = 1 + 2 + 3 + 4 = 10 \).

Step 3. \( k = 1 \): the factors are \( 1, -i, -1, i \):

$$ X_1 = 1 - 2i - 3 + 4i = -2 + 2i $$

Step 4. \( k = 2 \): the factors are \( 1, -1, 1, -1 \):

$$ X_2 = 1 - 2 + 3 - 4 = -2 $$

Step 5. \( k = 3 \): the factors are \( 1, i, -1, -i \):

$$ X_3 = 1 + 2i - 3 - 4i = -2 - 2i $$

Checks: \( X_0 \) is the sum of the data, \( X_3 \) is the complex conjugate of \( X_1 \) as expected for real data, and Parseval's rule holds: \( \sum |x_n|^2 = 30 \) and \( \frac{1}{4}(100 + 8 + 4 + 8) = 30 \).

Computing a DFT directly takes about \( N^2 \) operations. The Fast Fourier Transform (FFT), popularised by James Cooley and John Tukey in 1965, needs only about \( N\log_2 N \). For a million samples, that is the difference between roughly \( 10^{12} \) and \( 2 \times 10^{7} \) operations. In Python, numpy.fft.fft computes it in one line.

A Practical Calculation

If audio is sampled at 8000 Hz and we take an FFT of 1000 samples, the frequency resolution is \( \frac{8000}{1000} = 8 \) Hz, and the highest frequency we can detect is the Nyquist frequency, \( \frac{8000}{2} = 4000 \) Hz.

Applications

• Audio and music: equalisers, noise removal, pitch detection and compression formats all work in the frequency domain.
• Telecommunications: Wi-Fi and 4G/5G use OFDM, which sends data on many frequencies at once using the FFT.
• Medical imaging: MRI scanners collect data in the frequency domain, and an inverse Fourier transform builds the image.
• Image compression: JPEG uses the discrete cosine transform, a close relative, to discard fine detail the eye barely notices.
• Physics: in quantum mechanics, position and momentum wavefunctions are Fourier transform pairs, which leads to the uncertainty principle.
• Solving PDEs: transforming the heat equation in space turns it into an ODE in time.

Did You Know?

• Joseph Fourier developed his ideas while studying heat flow in the early 1800s, after serving as a scientific adviser on Napoleon's expedition to Egypt.
• Carl Friedrich Gauss described an FFT-like method around 1805 to compute asteroid orbits, but it was not widely noticed until much later.
• The uncertainty principle in quantum mechanics is the same mathematics as the rule that a narrow pulse needs a wide spectrum.

Key Takeaways

• The Fourier transform decomposes a signal into its frequency components.
• Narrow signals in time have wide spectra, and vice versa.
• Derivatives become multiplication by \( i\omega \) and convolution becomes multiplication, which simplifies many problems.
• The DFT handles sampled data, and the FFT computes it efficiently.
• Applications range from audio and wireless communication to MRI, image compression and quantum physics.

If Fourier analysis is part of your coursework or assignment, Amathify's subject experts and PhD scholars can provide clear step-by-step solutions. Message us on WhatsApp or use the Ask a Question form.