The Michelson–Morley Experiment

Sound needs air and ripples need water, so when Maxwell proved that light is a wave travelling at $c = 1/\sqrt{\mu_0\varepsilon_0}$ , it seemed obvious that light needed a medium too. That medium was the luminiferous ether: weightless, transparent, filling all of space, and defining the one frame in which light really moves at $c$ . If it existed, the Earth was ploughing through it at $30\ \mathrm{km/s}$ , and anyone riding the Earth should feel an ether wind — light sent upwind and back returning late, light sent across the wind returning early. Measure the difference and you have measured the Earth's absolute velocity through space. Albert Michelson set out to do exactly that, and the obstacle was brutal. The effect is not of order $v/c \approx 10^{-4}$ , as a first guess suggests, but of order $(v/c)^2 \approx 10^{-8}$ — the leading terms cancel, twice over — and a hundred-millionth of a round-trip time lies far beyond any clock. So Michelson built an instrument that needs no clock: split a beam of light in two, send the halves down perpendicular arms, bring them back, and let the light interfere with itself. Interference fringes are a ruler graduated in wavelengths, and they register a path difference of a hundred nanometres. In 1887, in a basement in Cleveland, Michelson and Edward Morley floated a sandstone slab on a bath of mercury, folded the light path to an effective eleven metres, and turned the whole apparatus slowly through the compass. The prediction was a shift of about four-tenths of a fringe. The fringes did not move — not that day, not at any hour, not at any season. Nature, asked to name a preferred direction for light, refused to answer, and classical physics never recovered.

Key takeaways

If light moves at $c$ relative to a stationary ether, an apparatus drifting through it at $v$ should find its round trip along the motion slower than the round trip across it, by $\Delta t \approx L\beta^2/c$ , where $\beta \equiv v/c$ .
The effect is second order in $\beta$ : the zeroth-order terms cancel exactly, and there is no first-order term at all. With $\beta \approx 10^{-4}$ for the Earth's orbit, the target is $\beta^2 \approx 10^{-8}$ — which is why the hunt demanded an interferometer, not a clock.
Rotating the table by $90^\circ$ swaps the arms and doubles the observable: $\Delta N = 2L\beta^2/\lambda$ . For the 1887 apparatus ( $L = 11\ \mathrm{m}$ , sodium $\lambda = 589\ \mathrm{nm}$ ) that is $0.373 \approx 0.4$ fringes — the figure Michelson and Morley themselves quoted.
The observed shift was under $0.01$ fringes: in their words, "certainly less than the twentieth part" of the expected value, "and probably less than the fortieth part". A roughly forty-fold null.
Key idea: a moving ether would make the two-way speed of light depend on direction at order $\beta^2$ , and the interferometer is a machine built to see precisely that. It saw nothing. Either matter conspires to hide the ether, or there is no ether and $c$ is simply the same in every direction, in every frame.

The core insight

Galilean velocity addition, applied to light, insists that the two-way speed of light must depend on direction whenever you move through the ether — not at first order in $\beta = v/c$ , where the upwind and downwind errors cancel, but at second order, $\Delta c/c = \beta^2/2$ . The Michelson interferometer converts that $10^{-8}$ asymmetry into a fringe shift of $0.4$ , a motion you can watch with your eye. The fringes never moved — and that stubborn fact is the crack through which special relativity entered physics.

Where this lives today

The experiment never ended; it was only rebuilt. Michelson's stone slab has become a pair of perpendicular optical cavities cut from ultra-stable silicon or sapphire, cooled to a few kelvin and rotated on an air bearing while their resonant frequencies are compared. A directional dependence of $c$ would show up as a beat note locked to the turntable and to the Earth's rotation. None has ever appeared, and the bound on $\Delta c/c$ now stands near $10^{-18}$ — about eight orders of magnitude beyond 1887, and the sharpest constraint we have on Lorentz invariance. The instrument itself has had a second career far larger than its first. A Michelson interferometer with $4\ \mathrm{km}$ arms, its light recycled thousands of times and its mirrors hung in vacuum, is what LIGO uses to detect gravitational waves, sensing arm-length differences of order $10^{-18}\ \mathrm{m}$ — a thousandth the width of a proton. The logic is unchanged from 1887: split a beam, send the halves down perpendicular arms, watch the fringes. Michelson built a machine to catch the ether wind and caught nothing; a century later the same machine caught two black holes colliding a billion light-years away.

The physical picture

Swim a hundred metres upstream and back in a river, then a hundred metres across and back. Both trips cover the same distance, but the up-and-down one always takes longer: going downstream you gain and going upstream you lose, yet you spend far more time losing than gaining — crawling upstream for a long while, flying downstream only briefly. What survives is a small excess, quadratic in the current. Light in a stationary ether should behave the same way. The parallel arm is the swimmer fighting the current; the transverse arm is the swimmer angling upstream to be carried straight across. Both arms are $L$ long, both journeys cross $2L$ of ether, and yet the times differ — but two cancellations conspire to make the difference tiny. First, the sign of the wind does not matter: reverse the current and both times are unchanged, because each depends on $v^2$ and not on $v$ , so no first-order effect can exist at all. Second, the transverse arm is slowed too, by just enough to swallow half the parallel arm's excess. What is left is $\Delta t \approx L\beta^2/c$ : one part in $10^8$ of the trip. Only an interferometer can see such a thing, because it never measures a time. It measures a phase, in units of the wavelength of the light itself.

Interactive Simulation

Predicted vs. Observed Fringe Shift — the aether hypothesis predicts a fringe shift as the interferometer rotates; the experiment saw essentially none (the null result that led to SR).

Figure

The 1887 interferometer — a beam split, sent down two ⟂ arms of effective length L = 11 m, and recombined, with the ether wind along one arm. Predicted Δt ≈ Lβ²/c ≈ 3.7×10⁻¹⁶ s: a path difference of just 110 nm, but that is 0.19 λ of sodium light, and the fringes can see it

Everything hangs on one number. The predicted path difference is $c\,\Delta t = L\beta^2 = 110\ \mathrm{nm}$ — a tenth of a micron out of a twenty-two-metre journey. Against a metre stick that is nothing; against the $589\ \mathrm{nm}$ wavelength of sodium light it is nineteen percent of a fringe, and rotating the table trades the arms and doubles it to $0.37$ . Michelson and Morley could read fringe positions to about a hundredth of a fringe: they had a factor of forty in hand, and they used it.

Figure

PUNCHLINE — the ether wind's entire predicted signature against what the instrument actually saw. Predicted ΔN = 0.373 ≈ 0.4 fringes; observed ΔN below 0.01. The gap is a factor ≳ 37, and it did not close at any hour or any season

The blue bar is the whole of nineteenth-century physics: Maxwell's waves, a medium to carry them, an Earth that moves. The red sliver is the measurement. The experiment did not merely fail to confirm the ether wind — it excluded it, at the speed the Earth is independently known to travel.

The core equations

Everything follows from one assumption applied twice: light moves at $c$ with respect to the ether, and velocities add the Galilean way, so an observer streaming through it at $v$ measures light going at $c \pm v$ along his motion. Take the parallel arm — the beam runs out against the wind at $c - v$ and returns with it at $c + v$ ; add the legs over a common denominator.

t_{\parallel} = \frac{L}{c-v} + \frac{L}{c+v} = \frac{2Lc}{c^{2}-v^{2}} = \frac{2L}{c}\cdot\frac{1}{1-\beta^{2}} \qquad (1)

where: $L$ — one-way arm length; $v$ — the apparatus's speed through the ether; $c$ — the speed of light in the ether; $\beta \equiv v/c$ . Physical reading: the round trip is always longer than $2L/c$ , whichever way the wind blows, because the result depends on $v^2$ — the slow upstream leg dominates the brief downstream one. The transverse arm needs a different picture. In the ether frame the far mirror slides sideways while the light is in flight, so a beam fired straight across would miss it: it must be aimed upstream. Its speed $c$ is then the hypotenuse of a triangle whose base is $v$ , and Pythagoras leaves only $\sqrt{c^2 - v^2}$ to carry it across. The return leg mirrors the outward one.

t_{\perp} = \frac{2L}{\sqrt{c^{2}-v^{2}}} = \frac{2L}{c}\cdot\frac{1}{\sqrt{1-\beta^{2}}} \qquad (2)

where: $\sqrt{c^2-v^2}$ — the component of the light's velocity that actually crosses the arm. Physical reading: the transverse arm is delayed too, but only by a square root's worth. This is the correction Michelson missed in 1881, and including it halves the entire effect. Subtracting and expanding for $\beta \ll 1$ is where the physics lives. Both arms give $2L/c$ at rest, so the leading terms cancel; both expansions are even in $\beta$ , so there is no first-order term for the subtraction to expose. What survives begins at $\beta^2$ .

\Delta t = t_{\parallel} - t_{\perp} = \frac{2L}{c}\left(\frac{1}{1-\beta^{2}} - \frac{1}{\sqrt{1-\beta^{2}}}\right) \approx \frac{L\beta^{2}}{c} \qquad (3)

where: $\Delta t$ — the round-trip time difference between the arms. Physical reading: a fractional effect of $\beta^2/2 \approx 5\times10^{-9}$ . For the 1887 arms $\Delta t = 3.7\times10^{-16}\ \mathrm{s}$ — no clock, then or now, could time it directly. The interferometer sidesteps the clock. A time difference is an optical path difference $c\,\Delta t$ , and path differences are read in wavelengths. Turning the table through $90^\circ$ exchanges the arms, so $\Delta t$ flips sign and the observable change is $2\Delta t$ — converting a static, invisible offset into a moving fringe pattern.

\Delta N = \frac{c\,(2\Delta t)}{\lambda} = \frac{2L\beta^{2}}{\lambda} \qquad (4)

where: $\Delta N$ — fringes the pattern shifts on a $90^\circ$ rotation; $\lambda$ — wavelength of the light. Physical reading: the signal grows linearly with $L$ , which is why the beam was bounced back and forth to reach an effective $11\ \mathrm{m}$ on a table barely a metre across.

Where the prediction comes from

Derivation AWhy the ether wind is a second-order effect, not a first-order one

The most important structural fact about this experiment is that the signal is $O(\beta^2)$ and not $O(\beta)$ . A first-order effect at $\beta \approx 10^{-4}$ would have been caught by cruder instruments decades before 1887; a second-order effect at $\beta^2 \approx 10^{-8}$ needed a mercury-floated stone slab and an eleven-metre optical path. The goal is to show that this smallness is not an accident of the arm lengths but a theorem: the zeroth-order terms cancel exactly, and there is no first-order term anywhere to cancel, because both round-trip times are even functions of $v$ .

Step 1. Write the parallel round trip. Against the wind the light closes on the mirror at $c-v$ ; with the wind it returns at $c+v$ . Adding, $t_\parallel = L/(c-v) + L/(c+v) = 2Lc/(c^2-v^2) = (2L/c)(1-\beta^2)^{-1}$ . Why: the legs are individually asymmetric in $v$ , but their sum is not — the odd parts cancel in the addition, the first hint that only $v^2$ can appear.

Step 2. Write the transverse round trip. The light must be launched at an angle so its sideways drift matches the mirror's; its speed along the arm is then $\sqrt{c^2-v^2}$ , and each leg takes $L/\sqrt{c^2-v^2}$ , giving $t_\perp = (2L/c)(1-\beta^2)^{-1/2}$ . Why: the transverse beam does not escape the wind — it pays for the drift with a slower crossing, and again only $v^2$ appears.

Step 3. Expand both binomially: $(1-\beta^2)^{-1} = 1 + \beta^2 + \beta^4 + \cdots$ and $(1-\beta^2)^{-1/2} = 1 + \tfrac12\beta^2 + \tfrac38\beta^4 + \cdots$ . Every power of $\beta$ is even: there is no $\beta^1$ term in either series — not a small one, none at all. Why: because both times depend on $v$ only through $v^2$ , no odd power can be generated. A first-order ether-wind effect in a two-way light path is forbidden by symmetry, not merely suppressed.

Step 4. Subtract. The zeroth-order pieces are both $2L/c$ and cancel exactly. The $\beta^2$ pieces do not: $\Delta t = (2L/c)\left[(1+\beta^2) - (1+\tfrac12\beta^2)\right] + O(\beta^4) = L\beta^2/c$ , the next correction being $5L\beta^4/(4c)$ . Why: this is the double cancellation — the transverse arm's own delay eats exactly half of the parallel arm's excess, and what remains is $10^{-8}$ of a $10^{-7}\ \mathrm{s}$ journey.

Step 5. Convert to fringes. The path difference is $c\,\Delta t = L\beta^2$ ; a $90^\circ$ rotation reverses the sign of $\Delta t$ , so the change is $2L\beta^2$ and the pattern shifts by $\Delta N = 2L\beta^2/\lambda$ . Why: an absolute path difference is lost against the unknown zero of the fringe pattern, but a change in it is not — the rotation is what makes the effect observable at all. *Verified symbolically in verify_aq-sr-michelson-morley-experiment.py: both series carry zero coefficient at every odd power of $\beta$ ; $\Delta t$ has zero constant term, zero linear term, coefficient $L/c$ at $\beta^2$ and $5L/(4c)$ at $\beta^4$ ; and $\Delta N = 2L\beta^2/\lambda$ .*

Go deeper · Derivation B · 2 minHow a contracting arm nulls the experiment exactly

Derivation BHow a contracting arm nulls the experiment exactly

Suppose the null result is telling the truth and there really is an ether — then something must be cancelling the $0.4$ -fringe shift. The most economical guess, made by FitzGerald in 1889 and independently by Lorentz in 1892, is that the apparatus is not what we think: a body moving through the ether is shortened along its motion. This derivation shows that if the parallel arm shrinks by exactly $\sqrt{1-\beta^2}$ , the two round-trip times become equal — not approximately, not to leading order, but identically, at every $\beta$ and every orientation. Why that factor and no other is the subject of Lorentz–FitzGerald Contraction Hypothesis.

Step 1. Let the arm along the motion have length $L_\parallel = L\sqrt{1-\beta^2}$ , while the transverse arm keeps its length $L$ . Why: the hypothesis is anisotropic by construction — invented precisely to spoil the symmetry between the arms that produced $\Delta t \neq 0$ .

Step 2. Substitute into equation (1). Nothing about the light propagation changes; only the distance does. $t_\parallel = \dfrac{2L\sqrt{1-\beta^2}}{c}\cdot\dfrac{1}{1-\beta^2} = \dfrac{2L}{c}\cdot\dfrac{1}{\sqrt{1-\beta^2}}$ . Why: the factor $\sqrt{1-\beta^2}$ in the numerator cancels one of the two powers hidden in $(1-\beta^2)^{-1}$ , turning $1/(1-\beta^2)$ into $1/\sqrt{1-\beta^2}$ .

Step 3. Compare with equation (2). The result is literally $t_\perp$ . Hence $\Delta t = 0$ exactly — no expansion, no approximation — and $\Delta N = 0$ at every orientation and every speed through the ether. Why: the contraction does not make the effect small, it makes it zero. That is why the hypothesis was taken seriously rather than laughed at.

Step 4. Count the cost in nanometres. The $11\ \mathrm{m}$ arm must lose $L(1-\sqrt{1-\beta^2}) \approx L\beta^2/2 = 55\ \mathrm{nm}$ , shortening its round trip by $110\ \mathrm{nm}$ — precisely the $110\ \mathrm{nm}$ of extra path the ether wind was supposed to add. Why: the two numbers agree to the last digit because they are the same number rearranged; the hypothesis is not a fudge factor fitted to data but an algebraic identity.

Step 5. Count the cost in credibility. The shrinkage must apply to sandstone, brass and glass alike, by exactly the same factor, with no mechanism to enforce it — and it can never be seen, since any ruler brought to measure the shortened arm has shortened too. Why: universal, unexplained, and unmeasurable except through the one experiment it was invented to explain — the textbook definition of an ad hoc rescue. *Verified symbolically in verify_aq-sr-michelson-morley-experiment.py: substituting $L \to L\sqrt{1-\beta^2}$ into $t_\parallel$ returns $t_\perp$ identically, and the $55\ \mathrm{nm}$ / $110\ \mathrm{nm}$ numbers reproduce.*

Limiting cases

Case	Result	Meaning
Galilean limit $c \to \infty$	$\beta \to 0$ , $\Delta N \to 0$	infinitely fast light has no wind to reveal; the experiment probes the finiteness of $c$
First order in $\beta$	coefficient exactly $0$	forbidden by the evenness of $t_\parallel, t_\perp$ in $v$ — a two-way path can never show a first-order wind
Second order in $\beta$	$\Delta t = L\beta^2/c$ , $\Delta N = 2L\beta^2/\lambda$	the entire signal, $10^{-8}$ in size — the reason interferometry was mandatory
Reverse the wind, $v \to -v$	$\Delta N$ unchanged	both times go as $v^2$ : the experiment reads wind versus calm, never upwind versus downwind
Arm contracts by $\sqrt{1-\beta^2}$	$\Delta t \equiv 0$ for all $\beta$	the FitzGerald–Lorentz rescue — see Lorentz–FitzGerald Contraction Hypothesis
$c$ the same in all directions	$\Delta t \equiv 0$	no ether, nothing to cancel — the road to The Constancy of the Speed of Light and Einstein's Two Postulates

Common trapThe ether-wind signal is second order in v/c, not first order — and the naive estimate is not merely rough, it is qualitatively wrong

The commonest error, and one Michelson himself half-made in 1881, runs: "light goes at $c \pm v$ , so the fractional error is $v/c \approx 10^{-4}$ , and I need a part-in-ten-thousand measurement." That is wrong twice over. First, on a round trip the $+v$ and $-v$ legs cancel their first-order errors exactly: $t_\parallel = (2L/c)(1-\beta^2)^{-1}$ contains no $\beta$ term at all. A one-way measurement would show first order, but nobody in 1887 could synchronise clocks at the two ends of an arm. Second, the transverse arm is also delayed, by $(1-\beta^2)^{-1/2}$ , which halves what is left — Michelson's Potsdam analysis missed this and predicted a signal twice too large. The real target is a part in $10^8$ , not a part in $10^4$ .

Worked examples

1 · The fringe shift the 1887 apparatus should have seen

The interferometer had an effective arm length $L = 11\ \mathrm{m}$ and used sodium light, $\lambda = 589\ \mathrm{nm}$ . The Earth orbits at $v = 29.8\ \mathrm{km/s}$ . How many fringes should the pattern shift on a $90^\circ$ rotation?

Step 1. Get $\beta$ . With $c = 2.998\times10^8\ \mathrm{m/s}$ , $\beta = v/c = 2.98\times10^4 / 2.998\times10^8 = 9.94\times10^{-5}$ . Round to the working anchor $\beta = 10^{-4}$ , so $\beta^2 = 10^{-8}$ .

Step 2. Assemble the geometric factor: $\dfrac{2L}{\lambda} = \dfrac{22\ \mathrm{m}}{5.89\times10^{-7}\ \mathrm{m}} = 3.735\times10^{7}$ — the number of wavelengths of sodium light in one folded round trip, and the leverage the multiple reflections buy.

Step 3. Multiply by $\beta^2$ : $\Delta N = \dfrac{2L\beta^2}{\lambda} = 3.735\times10^{7} \times 10^{-8} = \mathbf{0.373\ fringes}$ . With the unrounded $\beta = 9.94\times10^{-5}$ it is $0.369$ — either way, about four-tenths of a fringe, and $0.4$ is exactly the number Michelson and Morley printed.

Step 4. Cross-check from the path difference: $c\,\Delta t = L\beta^2 = 110\ \mathrm{nm} = 0.187\lambda$ — a shift of $0.187$ fringes for one orientation, doubled to $0.373$ by the rotation. Against a fringe-reading precision of about $0.01$ , that is a signal-to-noise of nearly forty: not a marginal experiment, a decisive one.

2 · Turning "we saw nothing" into a number

Michelson and Morley reported the displacement as certainly under $1/20$ , and probably under $1/40$ , of the expected value — that is, $\Delta N < 0.01$ fringes. What upper limit does that place on the Earth's speed through the ether, and on the anisotropy of $c$ ?

Step 1. Use the scaling. $\Delta N = (2L/\lambda)\beta^2 \propto \beta^2$ , so a bound on $\Delta N$ becomes a bound on $\beta$ through a square root: $\beta_{\max} = \beta_{\text{pred}}\sqrt{\Delta N_{\text{obs}}/\Delta N_{\text{pred}}}$ .

Step 2. Substitute: $\beta_{\max} = 10^{-4}\sqrt{0.01/0.373} = 1.64\times10^{-5}$ , hence $v_{\max} = \beta_{\max}\,c = \mathbf{4.9\ km/s}$ . The Earth's motion through the ether is bounded below one-sixth of its known orbital speed of $29.8\ \mathrm{km/s}$ — and the Earth demonstrably does orbit at that speed, as stellar aberration had shown since 1728. An ether at rest with respect to the Sun is excluded outright.

Step 3. Recast the bound as an anisotropy of $c$ . The arms measure two-way speeds $c_\parallel = 2L/t_\parallel = c(1-\beta^2)$ and $c_\perp = 2L/t_\perp = c\sqrt{1-\beta^2}$ , so the predicted fractional anisotropy is $\Delta c/c = \beta^2/2 = 5.0\times10^{-9}$ , and $\Delta N = 4L(\Delta c/c)/\lambda$ .

Step 4. Invert it: $\Delta c/c < \dfrac{\lambda\,\Delta N_{\text{obs}}}{4L} = \dfrac{5.89\times10^{-7} \times 0.01}{44} = \mathbf{1.3\times10^{-10}}$ . With a stone slab and fringes viewed by eye, the speed of light was shown to be the same in perpendicular directions to one part in ten billion. Modern rotating cryogenic optical resonators now push that to $\Delta c/c \lesssim 10^{-18}$ — eight further orders of magnitude, and still nothing.

Physics of the portrait

Michelson is the rare physicist whose fame rests on a result he never wanted. He was, first and last, an instrument builder, and the ether hunt was to him one more precision measurement — that the answer came out zero he regarded for years as a personal failure.

Origins · 2 minAlbert Abraham Michelson and how this was discovered

Albert Abraham Michelson

1852–1931 · Potsdam · Cleveland · Chicago

Born in Strelno in Prussian Poland and carried to the California gold country as a toddler, Michelson talked his way into the U.S. Naval Academy, graduated in 1873, and began measuring the speed of light with apparatus he built on the sea wall at Annapolis. In 1881, on study leave in Europe, he ran the first ether-drift interferometer at Potsdam and saw no fringe shift — though his treatment of the transverse arm was flawed, an error Alfred Potier pointed out to him and Lorentz corrected in 1886. Joining forces with the chemist Edward Morley at Cleveland, he rebuilt the instrument an order of magnitude better and, in 1887, obtained the most famous null result in physics. He never abandoned the ether emotionally, calling it "that beloved old monster" and running ever-larger versions of the hunt well into the 1920s. In 1907 he received the Nobel Prize in Physics "for his optical precision instruments and the spectroscopic and metrological investigations carried out with their aid" — the first American to win a Nobel Prize in any of the sciences. The citation does not mention the ether at all: Michelson was honoured for building the ruler, and history remembers him for what the ruler failed to find.

History

1881

— Michelson runs the first ether-drift interferometer in Potsdam, with arms of 1.2 m, and finds no fringe shift; his treatment of the transverse arm is incorrect.

1886

— Lorentz, taking up an objection Alfred Potier had raised, shows the transverse arm is also delayed, halving the predicted shift and putting it below the reach of the 1881 instrument.

1887

— Michelson and Morley, in Cleveland, float a sandstone slab on mercury and fold the light path to an effective 11 m: predicted shift about 0.4 fringes, observed shift under 0.01.

1889

— FitzGerald proposes in Science that bodies contract along their direction of motion, which would null the experiment exactly.

1907

— Michelson receives the Nobel Prize in Physics, the first American awarded one in the sciences.

1932

— Kennedy and Thorndike run an unequal-arm version, testing whether the speed of light depends on the laboratory's velocity as well as its orientation.

2015

— rotating cryogenic resonators bound any anisotropy of the speed of light at Δc/c ≲ 10⁻¹⁸.

Going further

Texts. Taylor & Wheeler, Spacetime Physics, 2nd ed., Ch. 3 · Rindler, Introduction to Special Relativity, 2nd ed., §1 · Griffiths, Introduction to Electrodynamics, 4th ed., §12.1 · Michelson and Morley, "On the Relative Motion of the Earth and the Luminiferous Ether", American Journal of Science 34 (1887) 333.

Provenance. The arm-time formulas and the second-order difference $\Delta t \approx L\beta^2/c$ are the standard treatment (Griffiths §12.1, Rindler §1, Jackson §11.1). Anchors, all recomputed in verify_aq-sr-michelson-morley-experiment.py: $c = 2.998\times10^8\ \mathrm{m/s}$ ; Earth's orbital speed $29.8\ \mathrm{km/s}$ ( $\beta = 9.94\times10^{-5}$ ); sodium $\lambda = 589\ \mathrm{nm}$ ; effective arm $L = 11\ \mathrm{m}$ . These give $\Delta N = 0.3735$ at $\beta = 10^{-4}$ and $0.369$ at the unrounded $\beta$ — the $\approx 0.4$ fringes Michelson and Morley quote is a rounding of this same calculation, not an independent number. A verifier should also reproduce $\Delta t = 3.67\times10^{-16}\ \mathrm{s}$ , the path difference $110\ \mathrm{nm} = 0.187\lambda$ , the bounds $v < 4.9\ \mathrm{km/s}$ and $\Delta c/c < 1.3\times10^{-10}$ , and the identity $t_\parallel(L\sqrt{1-\beta^2}) = t_\perp$ .

Builds on. The Luminiferous Ether · Ether Drift and the Absolute Rest Frame · Galilean Transformations · Inertial Frames of Reference. Leads to. Lorentz–FitzGerald Contraction Hypothesis · The Constancy of the Speed of Light · Einstein's Two Postulates.

If the effect is so tiny, how could a nineteenth-century experiment possibly see it?

Because the interferometer never measures the tiny thing directly: it measures a path difference in units of the wavelength of light, and sodium light provides a very fine ruler at $589\ \mathrm{nm}$ . The predicted path difference is $110\ \mathrm{nm}$ — nineteen percent of that ruler — so on rotation the fringes should slide by $0.37$ of a fringe spacing, a visible motion. The time difference behind it, $3.7\times10^{-16}\ \mathrm{s}$ , is untouchable by any clock even today; but the light itself is a clock, ticking every $2.0\times10^{-15}\ \mathrm{s}$ . The design turns an impossible timing problem into an easy counting problem.

Could the Earth simply be dragging the ether along with it?

That was the first escape route, and it fails. If the ether near the ground moved with the Earth there would be no local wind and the null would be automatic — but a fully dragged ether contradicts stellar aberration, in which the apparent positions of stars shift over the year by about $20.5$ arcseconds, exactly the angle $v/c$ . That shift requires starlight to arrive in an ether frame that does not travel with the telescope. Ether drag also cannot accommodate Fizeau's moving-water result, which shows only partial drag. Each rescue that saves one experiment breaks another — the corner physics found itself in by 1890.

Michelson and Morley did not literally report "less than 0.01 fringes". What did they report?

Their paper says the observed displacement was "certainly less than the twentieth part" of the expected $0.4$ fringes, "and probably less than the fortieth part". One-twentieth of $0.4$ is $0.02$ ; one-fortieth is $0.01$ . The $\Delta N < 0.01$ used throughout this article is their more confident bound, and the one usually cited. It matters that they phrased it as an upper limit: a null result is never "we measured zero", it is "we measured something smaller than we could have seen" — and the size of what you could have seen is the whole content of the claim.

Did the Michelson–Morley result lead Einstein to special relativity?

Less directly than the textbooks suggest. Einstein told Robert Shankland in the 1950s that he first learned of the experiment through Lorentz's writings, was unsure when he had first heard of it, and had assumed its result true without consciously building on it; the experiments he named as decisive for his own path were stellar aberration and Fizeau's measurement of light in moving water. What Michelson–Morley did was arguably larger: it made the problem undeniable for the whole community, so that when the postulates arrived in 1905 there was already a wound in classical physics that only they could close. Being the crisis is not the same as being the clue.

Why does the transverse arm matter at all? Its light never travels along the wind.

Because in the ether frame the transverse mirror is running away sideways while the light is in flight. To hit it, the beam must be aimed upstream, and then only the component $\sqrt{c^2-v^2}$ of its velocity carries it across the arm — so that trip is delayed too, by $(1-\beta^2)^{-1/2}$ . This is no footnote: it removes exactly half of the parallel arm's excess. Were the transverse arm truly unaffected, $\Delta t$ would be $2L\beta^2/c$ rather than $L\beta^2/c$ — Michelson's failure to include it in 1881 is why his first prediction was twice too large.

Doesn't the interferometer's own arm mismatch swamp the effect?

It would, if the experiment tried to measure an absolute path difference: the arms could never be cut equal to within a hundred nanometres, and thermal drift alone moves them far more. This is exactly why the apparatus is rotated. Whatever constant offset the arms carry rides along unchanged; only the ether-wind term, which depends on the arms' orientation relative to the motion, reverses. The signal is therefore a periodic fringe shift, locked to the turntable's angle and to the Earth's daily and yearly motion — everything static is common-mode and drops out, the trick every lock-in measurement uses today.