Fun With Science / Globe Deconstruction / Earth Rotation
Question 11, the Compton generator, and the several thousand people who settle this for a living every day.
Question 11 is about a time-lapse at the South Pole, but the thing it is really asking is older and better than that, and worth answering on its own terms before touching the specifics. How do we know the Earth turns, rather than the heavens turning around a stationary Earth? From a deckchair those two are genuinely indistinguishable, and anybody who says otherwise is not taking the question seriously. They stop being indistinguishable the moment you go indoors.
Answered dailyPremise invertedThe short version
A turning sky and a turning Earth predict the same view of the sky and different physics in a windowless room. Ships, tunnel surveyors, drillers and airline crews use that difference as a working tool every day — none of them can see the sky while doing it. On the specifics: zero tangential speed at the axis is not zero angular rate, so the pole is where a Compton generator reads strongest, not weakest; a Foucault pendulum has been run there; and the ocean performs the same measurement continuously at every latitude, returning the length of the day to within half a per cent.
This chapter of the book is arguing geocentrism — a stationary Earth — and that matters here, because the geocentric alternative is not “nothing rotates.” It is that the relative rotation everyone agrees on belongs to the sky rather than to us. p. 180 quotes Einstein to exactly that effect — that the Sun at rest and the Earth moving, or the reverse, are “two different conventions concerning two different” coordinate systems.
Taken as a statement about steady motion in a straight line, that is correct, and it is the reason nobody can feel the Earth's orbit. But the book itself supplies the reason it does not extend to spin. p. 182 argues that relativity of motion “only applies to constant straight-line speed,” and that a body going in a circle “really is accelerating.” That is right, and it cuts cleanly:
Rotation is accelerated motion. Accelerated motion is detectable from the inside. So if the Earth turns, it must be possible to prove it in a sealed room — and if the sky turns instead, it must not.
That is the whole test, and it is a good one, because the two models genuinely diverge. A rotating sky exerts nothing on a pendulum, a gyroscope or a bucket of water in a basement. A rotating Earth carries everything on it into a rotating reference frame, where centrifugal and Coriolis effects appear whether or not anybody is looking — effects with a specific size, a specific direction, and a specific dependence on latitude that can be checked against a formula rather than an opinion.
So the honest question is not whether the sky appears to turn. It does. The question is whether the turning shows up somewhere the sky cannot reach.
Before going on, the strongest form of the geocentric reply deserves stating, because it is better than the one the chapter makes and it is not obviously wrong. “A rotating sky exerts nothing on a pendulum” is a Newtonian claim. In general relativity, rotating matter drags the local inertial frame with it — the Lense–Thirring effect, measured by Gravity Probe B and by satellite laser ranging. A sophisticated geocentrist (Sungenis and Bouw both go here) argues that a rotating shell of distant matter would drag the frame inside it, and that Foucault, Coriolis and the drifter loops are what that dragging feels like from the inside. On that reading, the sealed room is not decisive.
We think that is a real objection and it is worth granting. It is also, as an argument, an own goal three times over:
So the dichotomy above should be read as it is meant: everything from here on measures the relative rotation, and finds it real, quantitative and shaped like a sphere. Which of the two frames is “truly” turning is then a question we are happy to leave open — because either answer costs the chapter its conclusion. The first two bullets are what the rest of this page rests on, and neither depends on how that argument comes out.
It does, and the demonstration is not a museum exhibit. Several trades depend on Earth's rotation being real and quantitatively correct, and the striking thing about the list is what the jobs have in common: almost none of these people can see the sky while they work.
These are also not delicate experiments in need of interpretation. They are load-bearing. A tunnel that misses is a very expensive hole, and the Channel Tunnel bores met.
The most complete version of the test comes from somewhere nobody set it up on purpose. Free-floating ocean drifters trace small circular loops on top of whatever current carries them — inertial oscillations, caused by exactly the physics the instruments above exploit. The loops have a period that depends on latitude, and a direction that depends on hemisphere.
Fifteen NOAA drifting buoys, deployed for weather and climate work by people not arguing with anybody, give this:
A stationary Earth under a turning sky predicts no loops at all. A flat turntable predicts loops of identical period and identical direction everywhere, with nothing happening at the equator. A spinning sphere predicts what is observed. And the buoys are not instruments in any sense that an objection could attach to — a float has no sensitive axis, no calibration, and no bias term. It reports where it is; the rotation is in the shape of the path.
With that established, the specifics of Question 11 become tractable — and the fact at its centre turns out to be correct.
Question 11 is a summary; the chapter behind it runs pp. 178–194. A table on p. 190 gives air speeds of 1,040 mph at six feet rising to 1,050 mph at 37 miles, objecting that “the higher layers of air would have to travel mechanically faster to keep up.” p. 191 asks whether synchronised rotation happens “at all points of the ball — 0 mph air speed at axis.” p. 192 puts it most directly: “At the exact axial point, what makes a drone rotate in sync with the ground if it is not attached to it?”
What the chapter never does, anywhere in five hundred pages, is say what a Compton generator is. The term appears exactly once — in Question 11 — and is never defined. So that comes first.
A hollow ring of tubing, filled with water and sealed. Arthur Compton built the first in 1913, as a final-year undergraduate at the College of Wooster in Ohio, from one-inch brass tube bent into a circle eighteen inches across, with fine droplets of coal oil in the water and a microscope on a viewing window.
Lay it flat, let the water settle completely, then flip the whole thing through 180° about a horizontal diameter — over, and flat again. The water is now drifting around the tube, and you have a minute or two to time it before viscosity kills the motion. Lying flat, the ring's axis pointed up and the water was turning with the Earth at Ω sin φ; flipped, the tube turns at −Ω sin φ while the water's angular momentum has not changed. Relative to the tube, it circulates at twice the original rate. A Foucault pendulum with the bob replaced by a fluid:
ū = 2ΩR sin φ
Note the sin φ: zero at the equator, maximum at the poles. Compton's own nine-inch-radius ring gives 21.8 µm/s at Wooster, and 33.3 µm/s at the South Pole — 1.53× stronger. The 1913 Science paper is a demonstration; the precision came two years later, in a refined apparatus reported in Physical Review in 1915 under a title that says exactly what it is for — “A Determination of Latitude, Azimuth, and the Length of the Day Independent of Astronomical Observations.” Latitude to about 3%, and the length of the day to within roughly sixteen minutes — about one per cent — from a sealed ring of water on a bench. Those two accuracy figures come from standard secondary summaries of the 1915 paper, not from the paper itself, which we have not been able to read; treat them as approximate until someone checks the original.
The premise — “the rotational speed of the air at the axis point would be zero” — is true of one quantity and false of another.
| At the equator | At the pole | |
|---|---|---|
| Tangential speed across the surface | ~1,040 mph | 0 |
| Angular rate | 15.04°/h | 15.04°/h |
| Vertical component Ω sin φ — what local experiments feel | 0 | maximum |
A rigid body turning about an axis has the same angular velocity throughout; only distance from the axis, and so linear speed, varies. The pole does not stop rotating — it rotates in place. A spinning top is not stationary at its tip.
This is the part worth dwelling on, because it is a much harder thing to explain away than any single reading. Earth's rotation resolves into two local components, and they behave oppositely with latitude:
| Component | At the equator | At the pole | What depends on it |
|---|---|---|---|
| Vertical, Ω sin φ | zero | maximum | Foucault pendulum, Compton generator, inertial loops |
| Horizontal, Ω cos φ | maximum | zero | gyrocompasses and gyrotheodolites finding north |
So a spinning Earth predicts that at the pole, pendulums and water rings should read their strongest — and north-seeking gyros should stop working entirely. Both are observed. A gyrotheodolite cannot be used at either pole, and is unreliable above about 75° latitude, because the angle between Earth's rotation and local gravity becomes too small to resolve. That is not a physicist's caveat; it is an operating limit that mine surveyors work around.
Set out in full, the time-lapse argument runs: at the axis the air has no speed, so a camera up in it is not being dragged around by anything; if the ground below were rotating, the camera should hold its heading while the landscape turned beneath it; no such turning is seen; therefore it is not turning. Every step follows — provided the camera is genuinely decoupled from the Earth. That is the load-bearing assumption, and it fails differently depending on where the camera is.
On the ice, camera and landscape are bolted to the same rotating object and share its motion exactly. No relative movement, ever — equally true in Reykjavík or Nairobi. Nothing about it is polar.
Airborne, the decoupling still doesn't happen, because the air is not stationary either. Zero tangential speed at the axis is not zero angular rate: the polar air column turns in place at 15.04°/h and carries anything floating in it. A drone tries to hold its own heading, with a magnetometer that is nearly useless at the pole and a MEMS gyro whose yaw drifts by degrees an hour — sensors that cannot resolve 15°/h, let alone subtract it.
Genuinely decoupled — on a mount holding fixed orientation in space — and the prediction is correct: the landscape really would turn beneath it, at 15.04°/h, which is 0.25° per minute. Across a twenty-to-forty-minute drone flight that is five to ten degrees, which nobody notices by eye in a shaky aerial shot.
So the argument fails not because it is silly, but because the effect is real and slow and the eye is the wrong instrument. Put something sensitive enough on that decoupled mount and it registers immediately — which is what a ring-laser gyroscope is, and why the chapter must then spend pp. 183–187 arguing those instruments measure something else.
Not with a ring of water, but with the pendulum version of the same physics. In the winter of 2001, Mike Town (University of Washington), John Bird (York University) and R. Allan Baker (Sonoma State) hung a Foucault pendulum down the six-storey stairwell of the new Amundsen–Scott South Pole Station while it was under construction — a 33 m wire carrying a 25 kg bob, in a sealed shaft where no moving air could disturb it. They measured a precession period of 24 hours ± 50 minutes against a predicted 23 h 56 m, consistent with the Earth turning clockwise as seen looking down from above the South Pole — which puts the swing plane drifting counter-clockwise across the floor, the mirror image of the northern hemisphere. They got g = 9.85 ± 0.03 m/s² along the way. Their account is here, posted to Baker's university page in October 2001.
As for the Compton generator specifically: it is a 1913 teaching demonstration accurate to about 1%, and ring-laser gyroscopes now measure the same rotation continuously to parts in 10⁸. Nobody hauling a brass water torus to Antarctica is not a gap in the evidence — and the modern separating test is Compton's test: rotate the instrument 180°, and the Earth signal reverses sign while instrument bias stays put. That method is in daily industrial use. It is simply no longer made of brass and water.
There is a small awkwardness in the question as posed, and then a much more useful point behind it. The awkwardness: it asks why no scientist has run this in Antarctica — but a scientist running it in Antarctica would be an institutional result from a research station, arriving through precisely the channels the book's later chapters decline to take at face value. If the answer would not be believed on arrival, requesting it is not a route to resolution. That is not a reason to dismiss the question; it is a reason to notice that a polar measurement cannot actually do the job being asked of it.
The useful point is bigger: a single location was never where the evidence lived.
The discriminating factor in the drifter data is the trend line, not the final point.
No individual buoy proves anything, and none is asked to. Any one reading could be a coincidence, a poorly-resolved spectrum, or a float shifted off f by the vorticity of the flow it is riding in — and indeed two of the fifteen are well out for exactly that kind of reason. What cannot be a coincidence is fifteen of them lying along a one-parameter curve and reversing handedness at the equator. The evidence is in the relationship between the measurements, not in any measurement. That is why the fit recovers the length of the day rather than merely being consistent with it: the constant is over-determined fifteen times, so it has nowhere to hide.
Which is precisely why the polar reading is unnecessary. You do not need the endpoint of a curve you have already fixed from both directions. A rotating globe does not just predict a big number at the pole; it predicts a specific value at every latitude, with one constant and no freedom to fit, and the pole is simply where that curve runs out of hemisphere. Its value there is settled by measurements taken nowhere near it.
Which turns the original question around into a better one. Not “why has nobody measured this in Antarctica,” but: the measurement has been made all the way from 63°N to 59°S, it traces the predicted curve, and the extrapolation to the pole is arithmetic. If that curve is wrong, it is wrong in public, in data anyone can pull this afternoon — and finding it wrong would be far more damaging to the standard model than any single Antarctic result could ever be.
Question 10 asks how it was ever established that air layers “stay in perfect rotational sync across all heights of the atmosphere,” and p. 190 puts numbers to it: 1,040 mph at six feet, rising to 1,050 mph at 37 miles. The objection is that the higher layers would have to travel mechanically faster to keep up with the lower ones, and that this “goes against demonstrable engineering principles.”
The arithmetic is correct. Speed is angular rate times radius, so air further out does move faster, and checking the figures gives 1,039.2, 1,041.9, 1,044.8 and 1,049.0 mph — his table is right to the mile per hour.
Two things follow that the chapter doesn't consider.
The whole difference is 9.7 mph spread over 37 miles of altitude — about a quarter of a mile per hour per mile of climb.
| Wind shear | Change per mile of altitude |
|---|---|
| Required by co-rotation | 0.26 mph |
| An ordinary jet stream | ~20 mph |
Everyday weather sustains vertical shear roughly 75× larger than anything co-rotation demands, in every frontal system, on every continent, without difficulty. Whatever engineering principle is supposed to forbid the smaller number has to explain why the larger one happens daily.
The word doing the damage is “keep up.” It pictures a ground that has started spinning and an atmosphere scrambling to catch it. Nothing is catching anything. Uniform rotation requires no force to maintain — the air has been turning with the planet since both condensed, and angular momentum is conserved. There is no principle stating that a fluid cannot rotate with the body it rests on. There is one stating it will carry on doing so unless something stops it.
The stronger answer is to reject the premise outright. The atmosphere is not in perfect rotational sync, nothing in the standard model says it is, and the ways it departs are not a difficulty for the theory — they are the subject matter.
Air that lags the surface is an easterly wind. Air that runs ahead of it is a westerly. Every weather map is a map of the atmosphere failing to co-rotate.
The trade winds are air lagging the surface in the tropics. The mid-latitude westerlies are air running ahead of it. Jet streams are departures of 100 to 250 mph, ten to twenty times the entire co-rotation gradient in his table. Higher still, the thermosphere genuinely does super-rotate, measurably faster than the ground beneath it. A model requiring perfect synchronisation would be refuted by any weather forecast; the actual model predicts the departures and their pattern.
And that pattern is the same signature the drifters gave in §3, in a different fluid. Through the lowest kilometre or so of atmosphere — of order a kilometre, deeper in strong winds and shallower in calm — friction couples the air to the ground, and the wind direction turns with height: clockwise going up in the northern hemisphere, anticlockwise in the southern. This is the Ekman spiral, it is in the opening chapters of every meteorology textbook, and weather balloons measure it twice a day from roughly 800 stations worldwide.
The chapter rests its case on two further arguments. Both deserve answering directly rather than by assertion.
A fibre-optic gyroscope reportedly read differently at different altitudes on the same line of latitude — “more than a full degree per hour” between Pike's Peak at 14,115 ft and the Eastern Plains at 4,000 ft — taken as showing the instrument is not measuring Earth's rotation at all. The globe model does predict no altitude dependence, so that part is stated correctly. The question is whether the measurement is.
Credit where it is due, because it is unusual: the book says plainly that the data is unsourced. p. 187 reads, “I could not find the formal document with their exact measurements. We need multiple teams to perform this to see if the results are true.” That is more candour than this genre usually manages — and it means the central table is, by its own author's assessment, the weakest evidence in the chapter.
Two ordinary things would produce the reported variance. Mounting tilt: a gyroscope with its sensitive axis vertical reads Ω sin(φ + tilt), and at 38.84°N the true signal is 9.433°/h with a sensitivity of 0.204°/h per degree of tilt. Under five degrees of tilt reproduces the entire claimed effect — and levelling an instrument to better than a degree on a mountain summit, in cold and wind, does not happen by default. A 10,000-foot altitude change also brings a large temperature swing, which is the other thing fibre-optic gyro bias tracks. And the control is not clean: the comparison is presented as same-latitude, but the table lists Guanella Pass at 39.59°N against 38.84°N for the other two sites, which alone predicts about 0.15°/h of variance.
And the alternative on offer predicts the wrong pattern. If these instruments are sensing a vortex in some medium rather than the rotation of the ground beneath them, the signal should vary with altitude and have no particular reason to track latitude. The world's large ring lasers show the reverse. They sit at wildly different elevations — Wettzell in Bavaria at around 610 m, the Cashmere cavern at Christchurch near sea level — and each reads Ω sin φ for its own latitude, continuously — the best of them, the Wettzell G ring, to parts in 10⁸. The claimed effect is on the axis where the data is flat; the axis where the data is not flat goes unmentioned.
The book lists reasons Foucault pendulums are hard to run: an elliptical swing precesses on its own through Airy precession, a suspension stiffer in one direction biases the path, temperature changes the wire, draughts push the bob, traffic vibration distorts it, and a misaligned electromagnetic drive can steer a museum pendulum into a false rate. Every one of those is true. These devices are genuinely finicky, and a badly set-up one produces nonsense.
What none of it explains is why the artifacts should trace a curve. Those error sources bear no relationship to latitude. The precession rate does — it follows sin φ, from 36.6 hours at Compton's Wooster to 26.6 at Reykjavík to 23.93 at the pole. A stiff suspension does not know what latitude it is at.
One note on the source. The book cites Heike Kamerlingh Onnes's 1879 doctoral thesis as documenting why these pendulums fail. That thesis is titled “Nieuwe Bewijzen voor de aswenteling der aarde” — “New Proofs for the Axial Rotation of the Earth.” Onnes catalogued the error sources precisely so the experiment could be done properly and the rotation established more rigorously. The analysis is real and worth conceding; the work it comes from reaches the opposite conclusion to the one it is cited for. (His Nobel came in 1913, thirty-four years later, for liquefying helium.)
The gyroscope chapter argues against one instrument. It is worth saying plainly how many there are, because the ring laser is the one the argument knows about and it is not the case.
Earth's rotation has been detected by at least five mutually unrelated physical mechanisms. Not five products — five different pieces of physics, with no shared calibration chain and no common scale factor.
The crudest quantitative result here is a few per cent and the best is parts per billion — something like six orders of magnitude apart, with no disagreement anywhere. A systematic error that faked rotation in a ring laser would have to fake a differently-scaled phase in a caesium interferometer, and again in superfluid helium at a scale set by an atomic constant, and again in a brass gyrocompass with no scale factor at all, all by coincidence and all in the same direction.
The ordinary version of this is stronger than the exotic one. Every ocean-going ship finds true north with a gyrocompass, and it works only because the Earth turns. The international performance standard, IMO Resolution A.424(XI), sets the permitted settle-point error at ±0.75° multiplied by the secant of the latitude. That formula is not a convention: north-seeking torque goes as Ω cos(latitude), so the error must scale as its reciprocal. Modern fibre-optic and resonator gyrocompasses are still specified the same way, at 0.1° and 0.08° secant latitude. A stationary Earth predicts nothing of the sort, and the formula has been in maritime law since 1979.
What we are not claiming. Nuclear-spin (NMR) gyroscopes are sensitive enough — bias instabilities of 0.01 to 0.04 °/hr against Earth's 15.04 °/hr — but we could find no published measurement of Earth rate by one, so they are left out of the count. Comagnetometer experiments do subtract Earth's rotation as a modelled term, and they take that value from VLBI and ring lasers, which makes them a consistency check rather than an independent one. Nor do most of these papers publish a standalone absolute value with an error bar; they report agreement with the accepted figure at some level. The honest statement is that all five agree with the IERS value of 15.04107 °/hr within their stated uncertainties, not that all five independently returned it. And under the most aggressive lumping — treating all interferometry as one idea and all rotating-frame mechanics as another — the count falls to three: Newtonian mechanics, wave interference, and macroscopic quantum coherence. Three is still three.
A turning sky and a turning Earth are indistinguishable if you only ever look up. They are not indistinguishable below decks, underground, or inside a steel drill casing — and that is where the answer has been settled, commercially and repeatedly, for a century. On Question 11's own ground: the fact is right and the inference is not. Surface speed at the axis is zero, angular rate is not, local experiments respond to angular rate, and the pole is consequently where a Compton generator reads hardest — while north-seeking gyros fail there completely, exactly as the same equations require. A pendulum has been run at the pole. And the ocean runs the whole experiment continuously, at every latitude, returning the length of the day to a tenth of a percent. None of it needs a trip to Antarctica.