sciencebriefs
13:00in productionCh. 1 · A wave-built answer to a heat problem/ 13:00 · ceiling 15 min
Mathematics

Fourier analysis

1822

Joseph Fourier claimed in 1807, studying how heat spreads through a solid, that almost any function could be rebuilt from sine and cosine waves alone — a claim reviewers thought too bold to be rigorous, and one still central to JPEG images and digital audio.

Joseph Fourier's 1807 memoir on heat conduction, fully published in 1822, argued that a function describing an arbitrary shape or signal could be represented as a sum of simple sine and cosine waves, an idea narrower precursors from Clairaut, Lagrange and Gauss had used for specific problems but never claimed so generally. A review committee that included Lagrange and Laplace judged his 1811 submission insufficiently rigorous, and it took later mathematicians, particularly Dirichlet and Riemann, to establish precisely when and how such a series actually converges to the function it claims to represent. That precision matters because the claim is not unconditionally true: most functions require infinitely many terms in their Fourier series, and the series does not always converge, a limitation the underlying technique has carried since Fourier first proposed it, even as it became one of the most widely used tools in science and engineering.

Chapters & takeaways6
  1. 0:08
    A wave-built answer to a heat problem

    Fourier proposed in 1807 that the way heat spreads through a solid could be described using sums of sine and cosine waves.

  2. 2:10
    A bolder claim than earlier work

    Clairaut, Lagrange and Gauss had used similar trigonometric techniques for specific problems, but Fourier claimed the method could represent almost any function at all.

  3. 4:20
    A committee unconvinced by the rigour

    Lagrange and Laplace, reviewing Fourier's 1811 submission, judged the mathematics insufficiently rigorous even as they recognised its power.

  4. 6:30
    Precision arrives decades later

    Dirichlet and Riemann later worked out the precise conditions under which a Fourier series actually converges to the function it represents.

  5. 8:40
    Not every function cooperates

    Most functions require infinitely many terms in their Fourier series, and the series does not always converge to the original function at all.

  6. 10:50
    From heat to JPEG

    A close relative of Fourier's original technique now compresses digital photographs, decades after the underlying mathematics was first proposed for a very different problem.

Worth your time?

Yes. Study the whole thing.

4.5/ 5
What works
  • the contrast between Fourier's original, loosely rigorous proposal and its later formalisation by Dirichlet and Riemann shows mathematics correcting itself over time rather than being handed down complete
  • the huge range of unrelated fields depending on the same technique, heat, acoustics, crystallography, image compression, demonstrates real explanatory reach rather than narrow academic interest
  • being upfront that most functions do not have a Fourier series that neatly converges keeps the claim honest rather than oversold
What does not
  • the technique cannot deliver perfect time and frequency information simultaneously, a genuine trade-off rather than a solvable engineering problem
  • for unevenly spaced or gapped data, standard Fourier analysis tends to exaggerate long-period noise, requiring alternative methods rather than a straightforward application of the original technique
Study it if
  • anyone who wants to know why so much digital technology, from JPEG images to audio equalisers, quietly depends on sine waves
  • readers interested in how a mathematical claim can be simultaneously enormously useful and not fully rigorous when first proposed
  • anyone curious about the actual limits of representing a signal as a sum of frequencies
Skip it if
  • readers wanting the full technical convergence conditions spelled out in mathematical detail rather than in general terms
  • anyone hoping for a single unqualified fact rather than a claim later shown to hold only under specific conditions
The written brief4 min read

A wave-built answer to a heat problem

Joseph Fourier’s claim, first put forward in a 1807 memoir and given its full treatment in 1822, was that the temperature at every point in a solid object changing over time, an apparently complicated physical process, could be described mathematically as a sum of simple, well-understood sine and cosine waves added together. This was a striking claim because heat conduction has no obvious wave-like appearance on its surface, and Fourier’s insight was that decomposing an arbitrary function this way, breaking it into a series of oscillating components, could make an otherwise difficult problem mathematically tractable. The technique of building complicated functions out of trigonometric building blocks was not entirely new, but Fourier’s specific contribution was insisting the method should work for essentially any function at all, not merely for the narrow set of cases earlier mathematicians had tried it on.

A bolder claim than earlier work

Elements of the same technique had appeared before Fourier in narrower forms: Alexis Clairaut used a related method for orbital calculations in 1754, Joseph-Louis Lagrange applied similar ideas to vibrating strings in 1759, and Carl Friedrich Gauss used a comparable cosine-and-sine method around 1805 to interpolate the orbit of an asteroid, developing along the way an early technique for computing the result efficiently that anticipated the modern fast Fourier transform. The astronomer Friedrich Wilhelm Bessel independently applied Fourier series to a completely different problem, Kepler’s equation, publishing his own results in 1819, three years before Fourier’s own comprehensive treatment finally appeared in print in 1822. What set Fourier apart from these narrower precedents, according to later historical assessment, was chiefly the boldness of claiming the method could represent an arbitrary function in general, not any particular new mathematical machinery.

A committee unconvinced by the rigour

What has held up, decisively, is the practical reach of the underlying idea, even though its rigorous justification took decades longer than Fourier’s own confident claims suggested it would. A review committee including Lagrange and Laplace found Fourier’s 1811 submission insufficiently rigorous, and it fell to later mathematicians, particularly Peter Gustav Lejeune Dirichlet and Bernhard Riemann, to establish the precise conditions under which a function’s Fourier series actually converges to that function, turning an informal, powerful idea into a properly justified mathematical theory. Once that foundation was secured, the technique proved applicable far beyond its origin in heat conduction, extending to vibration and acoustics, optics, quantum mechanics, and eventually to numerical methods including the fast Fourier transform, which makes computing these decompositions practical at the speed modern applications require.

Precision arrives decades later

What does not hold, at least not unconditionally, is the claim in its boldest original form. Most functions actually require infinitely many terms in their Fourier series to be represented at all, and the series does not always converge to the function it is meant to represent, meaning the technique’s applicability depends on conditions that took later mathematicians to work out precisely rather than being automatic for any function whatsoever. A separate and more fundamental limitation concerns time and frequency together: a representation of a signal over time carries no information about its frequency content, while a Fourier transform of that same signal carries perfect frequency information but loses all sense of when, in time, each frequency actually occurred, a trade-off that cannot be engineered away and that has motivated newer techniques such as the short-time Fourier transform and wavelet transforms to manage.

Not every function cooperates

The practical stakes of getting this technique right run through an enormous range of modern technology. JPEG image compression relies on a close mathematical relative of the Fourier transform, the discrete cosine transform, applied to small blocks of a digital photograph, discarding the weakest frequency components to shrink file size while preserving most of the visible detail. Related transform techniques underpin nuclear magnetic resonance and mass spectrometry instruments used in chemistry and medicine, and X-ray crystallography reconstructs the atomic structure of crystals directly from diffraction patterns using the same underlying mathematics. Audio equalisers, digital radio receivers, and sound analysis tools used in forensics and acoustics all depend on decomposing a signal into its frequency components in essentially the way Fourier first proposed for an entirely unrelated problem involving heat.

From heat to JPEG

Yes, and the value here is in seeing how far an idea can travel from its original, narrow purpose once its mathematical foundations are properly secured. Fourier set out to solve a specific physics problem, how heat spreads through a solid object, and ended up proposing a method general enough that later generations found it indispensable for image compression, medical imaging and digital audio, none of which existed in any form when he first wrote it down. The honesty about where the technique’s limits actually lie, in convergence and in the trade-off between time and frequency information, makes this a genuinely instructive case of a powerful idea whose true scope took much longer to pin down than its initial usefulness might have suggested.

Same field · Mathematics4 of 15
Up next in Science

FOXP2

· 13:00

FOXP2 was dubbed the language gene after a mutation was linked to a family's speech disorder, but the same gene shapes birdsong and bat echolocation, and a 2018 study found no evidence it was recently and specially selected in humans.

13:00