A work in progress: I've run this twice so far (the second trial sloppier than the first). The method is sound — I'd like 5+ trials and a real frequency measurement before I'd lean on the number.

You can measure one of the fundamental constants of the universe with a microwave oven and a bar of chocolate. The trick: a microwave makes a standing wave, so it has fixed hot spots. Measure the distance between them and you've measured half a wavelength. Multiply wavelength by frequency and you get a speed — the speed of light.

Hypothesis

If the oven's stated frequency is correct and the hot spots really are half a wavelength apart, then c = f × λ should land near the defined value, 2.998 × 10⁸ m/s, to within my measurement error.

Materials

• Microwave oven (frequency printed on the back panel — mine says 2450 MHz)
• A wide, flat chocolate bar (or a plate of grated cheese — anything that melts visibly)
• A ruler and a microwave-safe plate
• The key step: remove the rotating turntable, or flip it so the food can't rotate. Rotation is the whole reason microwaves don't normally cook in stripes — defeat it.

Method

1. Lay the chocolate flat on the stationary plate.
2. Heat in short bursts (~15–20 s) and stop the instant the first two or three spots start to melt.
3. The melted patches sit at the wave's antinodes. Measure the distance between the centres of adjacent melted spots — that distance is half a wavelength.
4. Wavelength λ = 2 × (spot spacing). Then c = f × λ.

Data

Worked example, Trial 1: λ = 2 × 0.061 m = 0.122 m; c = 2.450 × 10⁹ Hz × 0.122 m = 2.99 × 10⁸ m/s.

Result: c ≈ (2.9 ± 0.1) × 10⁸ m/s — about 3% below the true value of 2.998 × 10⁸ m/s.

Where the error comes from

Honest accounting. A 3% result from a chocolate bar is a good day, but I don't want to oversell it.

• Few spots, big leverage. With only 2–3 melt patches, a 2 mm ruler error on a ~60 mm spacing is already a ~3% swing. This is the dominant error, and it's why Trial 2 drifted.
• Frequency tolerance. "2450 MHz" is a nominal label. Domestic magnetrons can wander by tens of MHz and shift as they warm up. I treated f as exact; it isn't — and this is the biggest unquantified error here.
• Fuzzy spots. Melting spreads, so "centre of the spot" is a judgement call worth another millimetre or two.

What I'd trust, and what I wouldn't

• The method is sound — standing waves, half-wavelength spacing, c = f × λ. That part I'd defend.
• My number I wouldn't. Two trials isn't a measurement, it's an anecdote with error bars, and I haven't independently checked the oven's true frequency.

Next on the bench

1. Run 5+ trials and report the mean ± standard error.
2. Borrow an RF meter (or use a known-frequency source) to measure f instead of trusting the label.
3. Photograph the melt pattern and measure spot centres from the image rather than by eye.

Changelog

• 2026-05-30 — Added Trial 2 (which made the result worse, and more honest); wrote the error budget.
• 2026-02-02 — Started with a single trial.