Laboratory · Historical model comparison · Explanatory preview

Michelson–Morley: what does a null result decide?

Hold the apparatus fixed and compare the predicted rotation shift with and without longitudinal contraction.

Change one setting and compare the consequences

Static worked example
Model inputs · SI units
Predict first

Your next accepted calculation shows the before-and-after values beside this note. No score or answer gate is used. The note stays in this tab and is not uploaded or saved.

Supported range: -0.95 to 0.95. Scientific notation is accepted: type 10⁻⁹ as 1e-9.
Experiment settings equal one-way arm length, vacuum wavelength
Supported range: 0.001 to 100 m. Scientific notation is accepted: type 10⁻⁹ as 1e-9.
Supported range: 1 × 10⁻⁹ to 0.001 m. Scientific notation is accepted: type 10⁻⁹ as 1e-9.

Changes here apply with Apply settings.

90° rotation shift · one common linear scale
Ether without contraction0.4 fringes
Ether with longitudinal contraction0 fringes

Bar range: −0.4 to +0.4 fringes. The center line is zero. The scale is recalculated after Apply.

Accepted calculation 0. Every model row shares these inputs.

Inputs used for these results
Equal one-way arm length
11 m
Vacuum wavelength
5.5 × 10⁻⁷ m
Signed ether-wind speed divided by c
0.0001
Computed values · shared inputs · not experimental observations
ObservableEther without contractionEther with longitudinal contraction
90° rotation shift (fringes)0.40
Leading order in β² (fringes)0.40
Parallel round-trip time (s)7.338 × 10⁻⁸7.338 × 10⁻⁸
Perpendicular round-trip time (s)7.338 × 10⁻⁸7.338 × 10⁻⁸
Arm-time difference before rotation (s)3.669 × 10⁻¹⁶0

For these equal arms, the contraction hypothesis cancels the rotation signal. A null result therefore does not by itself distinguish this contracted-ether model from every other null-predicting account. The arm-time difference at one orientation is not the full 90° rotation fringe shift. The β² expression is a low-speed approximation, not the exact high-speed curve.

An instrument splits a beam of light along two arms at right angles, sends each out and back, and compares the returns. If the Earth moved through a fixed medium for light, turning the instrument should shift the pattern measurably; the shift found was far smaller, and this lab shows what that null result does and does not decide.

In an ether at rest, light out and back along an arm of length L parallel to the ether wind takes 2L/(c(1 − β2)), and across it 2L/(c√(1 − β2)), where β = v/c. Turning the apparatus through 90 degrees swaps the arms, and the fringe pattern should shift by about 2Lβ2/λ. With arms of 11 m, light of 550 nm and β = 10−4, about the Earth's orbital speed, the ether without contraction predicts 0.4 fringes. If the arm along the motion is shortened by √(1 − β2), the two times become equal and the prediction is 0. A null result therefore rules out the resting ether without contraction, but it cannot choose between contraction in an ether and Einstein's kinematics, which both predict zero. The relativity paper mentions only unsuccessful attempts to detect the Earth's motion relative to the light medium, and names no experiment.

These are host reference calculations with modern SI calibration and illustrative settings. They are not historical measurements, a reviewed historical dataset, or publication of the strict 1904 shelf. No FrankenSim WASM calculation is claimed.

What is being calculated?

The apparatus has equal one-way arm lengths before any hypothesized contraction. The wind is aligned with one arm, and the displayed fringe prediction is the change after a 90° rotation. Unequal arms, angular sweeps, source coherence and an empirical upper bound are not modeled here.

t∥=2Lc(1−β2),t⊥=2Lc1−β2t_{\parallel}=\frac{2L}{c(1-\beta^2)},\qquad t_{\perp}=\frac{2L}{c\sqrt{1-\beta^2}}
ΔN90∘=2cλ(t∥−t⊥)≃2Lλβ2\Delta N_{90^\circ}=\frac{2c}{\lambda}(t_{\parallel}-t_{\perp})\simeq\frac{2L}{\lambda}\beta^2

These formulas describe the uncontracted case. For the contracted case the longitudinal arm is shortened by √(1 − β²), which equalizes the two round-trip times for equal rest lengths. The numerical owner evaluates the small difference in a cancellation-resistant form.

Try doubling the arm length, then doubling the wavelength, then reversing the wind. Finally choose a much larger speed ratio and inspect where the leading β² approximation departs from the exact model expression. Large ratios are mathematical stress tests, not claims about terrestrial wind speeds.

Prediction is not a measurement

No observed points, digitized fringe shifts or experimental confidence bounds have been added to these plots. A reviewed source-specific dataset, with its geometry, wavelength, protocol and uncertainty, is still needed before making a numerical comparison with a historical experiment.

The reference implementation is src/physics/reference/shelfOptics.ts. Its functions own the arm times, fringe shifts, drag speeds and wave residuals; the interface only projects their results. Numerical precision in a table does not imply measurement accuracy.

Inspect the reference calculations

This comparison belongs to the special-relativity discovery route.

Embed this laboratory on another page. It starts from its worked defaults, not your current settings.