Where’s the Moon’s Silhouette?

This is the first entry in an ongoing look at claims about the Sun and Moon — the “luminaries,” in the older sense of the word — drawn from a working review of Levi Miller's Globe Deconstruction. Question 9 of that review draft (p. 52) reads in full:

“Why do we never observe the entire circular silhouette of the Moon when approaching a solar eclipse? Photo manipulation always fails to detect evidence of the Moon.”
— Levi Miller, Globe Deconstruction, Question 9 (review draft, p. 52)

The short answer is that a new moon — which is exactly what causes a solar eclipse — is the one lunar phase that reflects essentially no sunlight toward Earth at all, day or night, eclipse or no eclipse; there's no missing photographic evidence, because there's nothing there to photograph. The longer answer, below, is worth walking through, because the same orbital mechanics that explains the invisibility also nails down the exact minute and sky position of totality months in advance, for a single city, and did happen, on schedule, after this page was first published.

The same objection also shows up, in slightly different words, on shapedebate.com's own Claim #3, which names four celestial “red flags”. The uniform brightness of a full moon and Jupiter's moon-shadow alignment now have pages of their own — Full-Moon Lighting and Jupiter’s Shadows — the moon-tilt illusion is written up as a test on the Self-Test Protocol page, and the fourth, “failed trajectory alignment during a solar eclipse,” is the same claim Question 9 makes, so the answer below covers both.

1 · Why a new moon has nothing to show you

The Moon emits no visible light of its own. Every bit of it you've ever seen was sunlight bouncing off its surface toward Earth. How much of that reflected light reaches you depends entirely on the Moon's position relative to the Sun and Earth — its phase.

A solar eclipse can only happen at exactly one phase: new moon, when the Moon sits between Earth and the Sun. That's not a coincidence or a special alignment for the eclipse specifically — it's the definition of new moon, and it happens roughly every 29.5 days regardless of whether it lines up precisely enough with Earth's orbital plane to cause an eclipse. At new moon, the hemisphere of the Moon facing Earth is the hemisphere facing away from the Sun. It is, from where you're standing, the dark side — not dark as in dim, but dark as in unlit, the same way the back of your own head is unlit at noon. There is no sunlit disk to see drifting into position, because through the days around new moon the face turned toward you is the face turned away from the Sun. Nor is there a new moon hanging in the sky the night before: at that phase the Moon isn't in the night sky at all — it rises and sets within about an hour of the Sun, so it is above the horizon almost exactly when the Sun is.

The claim assumes a dark, opaque circle should be visible against blue sky before it reaches the Sun. The reason it isn't has almost nothing to do with the Moon being dim, and everything to do with the blue being in the wrong place.

The blue is in front of the Moon, not behind it

This is the part worth getting right, because the obvious defence of the standard model — “the daytime sky is much brighter than an unlit Moon” — predicts precisely what the claim says we should see. Something much darker than its background is a hole. A black disc drifting across a bright blue sky would be about as conspicuous as an object gets. If that were the situation, the claim would be a good one.

It isn't the situation. The blue of a daytime sky is sunlight scattered by the air column between your eye and the Moon — the whole hundred kilometres of it sits in the foreground. The Moon is a quarter of a million miles further out, and can only block what lies behind it. Behind it is space, which contributes essentially nothing to the brightness of that patch of sky.

So the arithmetic of the patch where the new moon sits runs: all of the foreground scattering, which is unchanged, plus whatever faint light the Moon itself sends you. That second term is not quite zero — the Moon's night side is lit by earthshine, the full Earth hanging in its sky — but it is small, and it points the wrong way for the claim:

• — Sunlight at the Moon is about 130,000 lux; a full Earth delivers roughly 10–15 lux. Earthshine is around one ten-thousandth of sunlight.
• — Sunlit lunar rock has a surface brightness of a few thousand candela per square metre. Earthlit lunar rock is therefore about 0.3 cd/m².
• — Clear daytime sky runs 3,000–8,000 cd/m² well away from the Sun, and ten to a hundred times that in the glare within a few degrees of it — which is exactly where the new moon is.

The new moon is not darker than the sky around it. It is about one part in ten thousand brighter — and the eye needs roughly one part in a hundred to notice anything at all.

There is a check on this you may have made without noticing. A daytime quarter moon is easy to see, pale against a blue afternoon sky, and this is why: sunlit rock at a few thousand cd/m² is comparable to the sky's own brightness, so it registers. The earthlit face is ten thousand times fainter than that. Same object, same sky, one crosses the contrast threshold and the other misses it by two orders of magnitude.

It also disposes of the related argument that image processing ought to be able to pull the disc out. There is no dark disc in the data to stretch. There is a patch of sky that differs from its surroundings by 0.01%, in the brightest, most turbulent part of the frame.

What you do see, once the Moon's near edge starts to overlap the Sun's disk, is exactly the silhouette the claim is asking about — a growing black bite taken out of the Sun, deepening to totality and then reversing. That's not the Moon suddenly becoming visible; it's the one moment a new moon can be seen at all, because for those few hours it isn't relying on reflected light to show up — it's blocking light instead, silhouetted directly against the one thing in the sky bright enough to outline it.

2 · A falsifiable prediction, made months ahead

If new-moon geometry is the real explanation, it should be possible to calculate — independently of any photograph, any footage, any claimed observation — exactly when the Moon will be in that position, from any given point on Earth, arbitrarily far in advance. That calculation is just applied orbital mechanics: known orbital elements for the Earth–Moon–Sun system, run forward in time.

The visualization below does exactly that for Reykjavík, Iceland, across the nine days surrounding the total solar eclipse of 12 August 2026. All nine daily Moon paths stay on screen at once, plotted by azimuth and altitude. Two sets of reference dots are laid over them, both at 17:50 UTC — the eclipse's own clock time. The Moon's nine march across the sky at about fourteen degrees a day. The Sun's nine sit almost on top of each other: at a fixed hour it returns to nearly the same place, sliding down by about a quarter of a degree a day as August wears on, which is the whole of its motion over the period. Pick a day and press play to send that day's Moon along its own path, phase included. Then step the selection from the 8th through to the 16th and watch what the Moon does at that fixed hour: it closes on the Sun day after day, lands on it on the 12th, and keeps right on going at the same pace. Nothing here is fitted after the fact — the underlying ephemeris (Astronomy Engine, cross-checked against JPL Horizons and NOVAS) computes raw positions the same way regardless of which day you're looking at.

The calculation places maximum eclipse at 17:48:38 UTC, Sun altitude 24.6° above Reykjavík's horizon, Moon illumination 0.006% — run independently of, and before, the event itself. A second, independently built eclipse calculator (sunpoint.org.uk, not affiliated with this page or with Astronomy Engine) puts local maximum at approximately 17:48:48 with the Sun at about 24.5° — Iceland doesn't observe daylight saving, so its local time is UTC year-round, making that a direct ten-second, 0.1° agreement between two independently written pieces of software predicting the same sixty-second window in the sky, months before it happened. Being precise about what that does and does not show: both descend from the same JPL ephemerides, so it tests the implementations, not the underlying physics. The independent test is the sky. And on 12 August 2026 it happened, on schedule, along a track running from the North Atlantic across Iceland and down to Spain, watched and photographed all the way — NASA's Astronomy Picture of the Day carried totality over Zaragoza the following morning. This page would rather cite a timestamped frame from Reykjavík itself than a summary; if you have one, send it and it goes here.

3 · Finding the Moon without using sunlight

§1 explains why there is nothing to see; §2 shows the position can be calculated in advance. There is a third answer, and it is the most direct: the Moon's position near new moon can be measured, by several independent methods that don't rely on reflected sunlight at all.

One thing to rule out first, because the page loses its footing if it overclaims. None of what follows produces a dark disc silhouetted against blue sky before the eclipse. That image isn't merely difficult, it's impossible, for the reason already given: an unlit object with nothing bright behind it has nothing to be silhouetted against. What follows is a different thing — ways to establish where the Moon is when you cannot see it, several of which work at new moon precisely because they never depended on sunlight in the first place.

The one you can photograph yourself

This is earthshine, and it answers the question almost literally. Sunlight strikes Earth, mostly bounces off cloud tops, travels to the Moon, illuminates the hemisphere the Sun isn't lighting, and returns to us. Earth is a far larger and more reflective object in the Moon's sky than the Moon is in ours, so the lunar “night” side is never truly black — only very faint next to the crescent beside it. Leonardo da Vinci worked out the explanation in the Codex Leicester around 1510. The old name for it is better than the technical one: the old Moon in the new Moon's arms.

No telescope is needed. A camera that allows a one-to-two second exposure at a wide aperture, resting on anything solid, records it a day or two either side of new moon: the dim disc comes out clearly while the crescent blows out to white.

The one already printed in your tide table

Spring tides — the largest tidal range of the month — occur at new moon and full moon, when the Sun's and Moon's tidal axes line up. Neap tides, the smallest, fall at the quarters. Every port authority in the world publishes tide tables years ahead, and shipping depends on them being right.

Two honest limits on this one, because it is the weakest row in the table and the book (p. 166) disputes lunar tides anyway. Operational tide tables are harmonic fits to past gauge readings, so the amplitudes and phases are empirical rather than derived — but the frequencies being fitted are the Moon's orbital periods, taken from lunar theory. And syzygy only says the two tidal axes are aligned; on its own it does not decide whether the Moon is between Earth and Sun or opposite. What it does say is that if the lunar clock were wrong, spring tides would slide off new and full moon — everywhere, visibly, at every port.

The one an amateur can hear come back

Radio amateurs bounce signals off the Moon — EME, or moonbounce. In echo mode a single operator needs no partner: transmit a tone, and about two and a half seconds later your own signal returns, having travelled to the Moon and back. Lunar phase is irrelevant, because the signal reflects off rock, not off sunlight. Station software also predicts the Doppler shift from the ephemeris — up to roughly ±440 Hz at 144 MHz — and the echo arrives at the predicted offset. Distance and rate of approach, both confirmed, without a research budget — though not casually either: own-echo work on 2 m usually wants four or more long yagis and a kilowatt-class amplifier.

Caveat worth stating plainly: near new moon the Moon sits close to the Sun and solar radio noise degrades reception. Operators work around it with offset pointing, but this is harder at new moon than at full — it is not a clean eclipse-day demonstration.

Detecting the new Moon itself, at 10 GHz

The reason you can't see the Moon beside the Sun is a contrast problem specific to visible light — and contrast problems can be engineered around by changing wavelength. Optically the Sun outshines even a full Moon by a factor of order 10⁵, and outshines the earthlit face of a new moon by more like 10⁹. In the microwave band that ratio collapses.

That answers the claim on its own terms. The Moon can be detected right beside the Sun at new moon. It simply cannot be done with an instrument that responds only to visible light, which is what an eye is.

The silhouette, photographed — and predicted months ahead

NASA's Solar Dynamics Observatory orbits Earth and watches the Sun continuously. Two to five times a year the Moon passes between the spacecraft and the Sun, and SDO records it as a crisp black disc crossing the solar surface in extreme ultraviolet — the edge unusually sharp, because the Moon has no atmosphere to soften it. These transits necessarily happen at new moon; that is the only time the Moon is in the way.

The calendar is the part that matters. On 29 March 2025 SDO's project scientist published the year's upcoming transits in advance, including 25 July 2025 at 62% of the solar disc covered. The transit duly occurred, covering 62%, lasting about fifty minutes. A prediction made four months early, matching to the stated percentage — the same ephemeris as §2, doing the same job for a different instrument.

The Moon that has no phases at all

Two methods dispense with light entirely.

Cosmic rays striking the lunar surface produce gamma rays, and NASA's Fermi telescope images the Moon in them. Above about 31 MeV the Moon is brighter than the Sun. Because that emission has nothing to do with illumination, the gamma-ray Moon shows no phases whatsoever — in NASA's phrasing, at these energies it would always look full. At new moon it is essentially as bright as at full moon — the flux varies with the solar cycle's modulation of the cosmic-ray flux, but not with phase.

Better still, the Moon can be located by what fails to arrive. It blocks incoming cosmic rays, so detectors see a Moon-shaped hole in the sky. IceCube — buried in the ice beneath the South Pole, receiving no lunar light of any kind — has observed this shadow at better than 6σ and places its centre within 0.2° of the predicted position once the deflection of charged cosmic rays by the geomagnetic field is accounted for. Phase is irrelevant: the Moon blocks cosmic rays at new moon exactly as at full.

We haven't reproduced those sky maps here, because unlike the images above we couldn't establish clear reuse terms for them — but you can look at them directly. HAWC's own science page shows the Moon-shaped deficit in its cosmic-ray sky map, and IceCube's writeup on the companion Sun shadow explains why the Moon shadow is used as a routine check on where the detector is pointing — in their words, its reliability shows the angular resolution is well below one degree. The underlying IceCube measurement is in arXiv:1305.6811.

On tracking mounts — and a safety warning

A computerised mount slewing to the Moon deserves precision, because the argument is weaker than it first looks and it's better to say so ourselves. A GoTo mount finding the Moon is not independent evidence of the Moon's position; it's evidence that the mount and the eclipse prediction use the same lunar theory. The real force is consistency: one body of solar-system dynamics underwrites eclipse canons, the Astronomical Almanac, planetarium software, occultation timings, tide tables and spacecraft navigation. For the Moon to be missing before an eclipse, that single theory would have to fail in a way that spoils eclipse predictions while leaving everything else it powers intact.

Summary

Every row answering “no” in the second column is a method that would work identically if the Moon emitted no light and reflected none. The claim that the Moon cannot be located before an eclipse turns out to be a claim about one narrow band of the electromagnetic spectrum, and about one instrument — the eye.

4 · What would actually count against this

This is a testable claim, not just a testable-sounding one, and it's worth being explicit about what would have broken it: if totality over Reykjavík had occurred at a meaningfully different time or sky position than 17:48 UTC / 24.6° altitude, or if the Moon's phase in the lead-up had not thinned on the predicted schedule — 21% on the 8th, 5.6% on the 10th, 1.4% on the 11th, and 0.006% at 17:48 on the 12th — that would have been a real problem for the model, not a rounding error. None of that happened. A single independently-run alternative calculator landing within ten seconds and a tenth of a degree of the same prediction is the kind of cross-check this page would rather have than not have, and it's included above for exactly that reason.

Two objections worth answering before they are made. The first: ancient astronomers predicted eclipses without a globe, so prediction proves nothing. The Saros cycle does predict eclipse dates, from the arithmetic of repeating intervals, and the Babylonians had it. What it cannot give you is which city, at what minute, with the Sun at what altitude, and for how many seconds — sixty-five of them at Reykjavík and none at all a few hundred miles inland. Local circumstances come out of three-dimensional geometry and the figure of the Earth, and there is no pattern-matching route to them.

The second: you have only shown this for an eclipse day. The opposite, in fact — the invisibility is a property of new moon, not of eclipses. The Moon's orbit is tilted about 5.1° to Earth's, so in a typical year eleven or twelve of the twelve or thirteen new moons pass above or below the Sun and produce no eclipse whatever. Every one of them is exactly as invisible as this one was. The eclipse is the rare month when the geometry lets you see the object that is always there.

Sources & further reading

• Levi Miller, Globe Deconstruction (review draft), Question 9, p. 52 — the claim answered on this page.
• Shape Debate, Claim #3 (shapedebate.com) — “failed trajectory alignment during a solar eclipse,” the same underlying objection as Question 9. The other three red flags named in Claim #3 are answered elsewhere on this site: full-moon brightness on Full-Moon Lighting, Jupiter's moon-shadow alignment on Jupiter’s Shadows, and the moon-tilt illusion as a test on the Self-Test Protocol page.
• Don Cross, Astronomy Engine (MIT licence) — the ephemeris library used to compute every Sun/Moon position on this page, tested against JPL Horizons and NOVAS.
• Fred Espenak, NASA GSFC, Eclipse Predictions.
• sunpoint.org.uk, Reykjavík Solar Eclipse 12 August 2026 — independent second calculation used as a cross-check in §2.
• Earthshine photograph: ESO/B. Tafreshi (eso1210a), CC BY 4.0. Leonardo da Vinci's account is in the Codex Leicester, c. 1510.
• A. Hafez et al., “A radio determination of the time of the New Moon,” Monthly Notices of the Royal Astronomical Society 439, 2271 (2014); arXiv:1312.5956 — the 10.8 GHz Sun/Moon contrast ratio and the 21 May 2012 eclipse cross-check in §3.
• NASA, Solar Dynamics Observatory — lunar transits; the April–July 2025 transit predictions were published on the mission blog on 29 March 2025.
• NASA, “Moon Glows Brighter Than Sun in Images From NASA's Fermi” — the phase-independent gamma-ray Moon.
• Radio image of the Moon: ALMA (ESO/NAOJ/NRAO), NRAO gallery, CC BY 4.0.
• IceCube Collaboration, observation of the cosmic-ray Moon shadow, arXiv:1305.6811 — the 0.2° pointing figure in §3.
• HAWC science pages and IceCube, “The Sun also casts a shadow on IceCube” (2018) — cosmic-ray shadow sky maps, linked rather than reproduced.
• NOAA, spring and neap tides — spring tides at new and full moon.