Rankings Primer Arbiter Physicist Mathematician Contrarian QE Claims Sources

Blondlot: review by a PhD physicist (mainstream view)

Source gap (repeated from Step 1). The folder holds one document: a Google Books scan of the 1905 English translation of Blondlot's "N" Rays (161 PDF pages). It has no text layer, so I read it from rendered page images and Step 1's scratch OCR (kept in /tmp/bl), checking the key table against the image [read: page image p.83, book p.58]. Not in the folder: the French Comptes Rendus originals, Wood's Nature report, replication papers. Everything about Wood and other replications rests on [web] https://en.wikipedia.org/wiki/N_ray, not on the folder. Where I say so below, a conclusion rests on [memory] (physics background) plus my own arithmetic; I mark those.

Tags: [read] from the book; [web] URL; [memory] with confidence. My own arithmetic is marked [calc].

Corrections to Step 1

Polarization conflict stated too strongly. facts.json says X rays from a tube are "essentially unpolarized in this manner". That is not safe. Barkla showed X-ray polarization by scattering, first note in Nature March 1904, full account 1905 [web] https://en.wikipedia.org/wiki/Charles_Glover_Barkla. Tube (bremsstrahlung, "braking radiation") emission is also partially polarized [memory, medium]. The correct objection is that Blondlot's detector (the eye judging a faint spark) is not a polarimeter, not that X rays cannot be polarized.
Claim 1 as written in claims.md was disowned by Blondlot himself. His Preliminary Notice and Complementary Note 1 say the "polarization of X rays" effects belong to a different radiation, N rays, and that "the experiments are correct" only the identification changed [read: p.3-6 and Complementary Note (1)]. claims.md should say the claim was reassigned from X rays to N rays, not that it stands as a polarization proof.
The wavelength unit is not just an inference. Blondlot's Newton's-ring check counts how many N-ray rings fit in one dark-to-dark yellow ring, giving 0.0085 for index 1.04 [read: book p.59]. Yellow light is about 0.589 micrometres (um), so 0.589/0.0085 is about 69 rings, which fits only if the unit is um [calc]. So 8 to 18 nm is read from the book's own method, not guessed. Also, the text says the wavelengths are "much smaller than those of light" [read: book p.60].
Omitted history that matters. Blondlot first believed N rays had LONG wavelengths: his quartz indices (2.93, 2.62, 2.436, 2.29) were matched to Rubens' "residual rays" from rock salt (quartz index 2.18), i.e. far infrared [read: book p.13, Complementary Notes (7)]. In January 1904 he reversed this to wavelengths shorter than light. The same radiation family is therefore claimed to be both far-infrared-like (quartz index near 2.9) and extreme ultraviolet-like (aluminium index 1.04 to 1.85). Step 1 mentioned both numbers but not the reversal or the conflict.
Claim 2 detail missed. Water stops N rays but salt water and the ox eye pass them; Blondlot explains this by "a very small quantity of sodium chloride" [read: book p.45]. Rock salt was first listed as opaque, then reclassified as transparent when polished [read: note 4]. These are after-the-fact accommodations.
The "4 per cent" error claim fails against his own table (see Claim 4). Step 1 repeated it without checking.
Minor: facts.json says the quartz index ~2.9 "contradicts X-ray and optical physics". It is fine for far infrared, and wrong only for the claimed X-ray-like nature. Not an error in Step 1's facts, but the reasoning there is incomplete.

Claim 1: A small spark gap near an X-ray tube changes brightness with orientation, read as proof of polarization

What the book says [read: book p.3-6]. An induction coil makes a small, faint spark at a gap near an X-ray tube behind 40 cm aluminium foil. With the gap parallel to the cathode-ray direction the spark brightens, and a lead or glass sheet dims it. Turned perpendicular, nothing. Rotating the gap about the ray direction gives a "maximum to minimum" like a Nicol prism (a calcite polarizing prism) with the spark as "analyzer". He concludes "polarized in the complete sense". "The spark must be very small and faint", meaning it is judged by eye at the edge of visibility.

a. Mathematics consistent? There is almost no mathematics; the logic is an analogy. The step "brightness varies with gap angle, therefore the beam is polarized" does not follow, because a rotating gap also changes how it couples to stray fields from the coil, wires and tube, and the aluminium foil only blocks some of that. Aside from that, the report is internally coherent. Blondlot also supplies the geometry of the plane of action (the plane containing the cathode ray and the emitted ray) [read]. Interestingly this is the same plane that theory gives for the electric vector of bremsstrahlung [memory, medium], a coincidence I cannot credit to this experiment.

b. Consistent with established experiments? X rays do not refract or reflect by ordinary optics at his level (the book itself notes this and abandons the X-ray identification, [read: p.11]). X-ray polarization is real: Barkla 1904 to 1905 [web] https://en.wikipedia.org/wiki/Charles_Glover_Barkla. But Barkla used scattering, with ionization chambers as detectors, not an eyeball on a spark. Spark discharges do change with ionizing radiation (it lowers breakdown voltage) but that effect is symmetric, not dependent on gap azimuth in the way he reports [memory, medium].

c. Tested? The spark effect itself: no independent controlled test appears in the book. The N-ray program in general was tested by Wood (below). I found no source, in the folder or via the one Wikipedia page, specifically testing the spark-gap polarization experiment. By Blondlot's own later account the radiation involved was not X rays at all.

d. If true? An X-ray tube would emit a second, non-ionizing, transparent-to-metal radiation, and spark gaps would be analyzers for it; X-ray polarization would be detectable by eye with a spark. Little in textbook electromagnetism would have to change, since polarization of X rays is already accepted; what changes is the claim of an eye-readable detector.

Verdict (split). - X rays can be polarized: Established (Barkla), but not by this experiment. - The spark-gap-by-eye demonstrates it: Never properly tested in isolation (no blinded, instrumented repetition on record in the folder; not dismissed without testing, since Blondlot himself reassigned it to N rays). The reassigned form falls under Claim 5's verdict, Tested, not supported.

Claim 2: Many ordinary sources emit "N rays" that pass metals, wood, black paper but not water or wet paper

What the book says [read: book p.13-16, 45-52, Complementary Notes]. An Auer burner behind iron and aluminium focuses through a quartz lens onto the spark; transparent: aluminium, wood, black paper, glass, later copper 65 cm thick [read: note 5]; opaque: lead (later said to be due to oxide), platinum 4 mm, water, wet paper. Also sources: Nernst lamp, flames, sun, hardened steel; effects also from sound, speech, a bent cane, a clenched fist [read: book p.48-49, 79-80].

a. Mathematics. No equations to check. Internal consistency is poor: transparency is repaired post hoc (rock salt opaque then transparent; lead opaque then "metallic lead lets pass"; water opaque but salt water and ox eye transparent). A "fist" and "sound" produce the same effect as a Nernst lamp and a steel chisel, with no unifying physical quantity.

b. Consistent with experiment? No. Radiation that passes 65 cm of copper and 1.5 cm of aluminium, but not a film of water, has no counterpart. Known radiation that penetrates metal (hard X rays, gamma rays, neutrinos) is ionizing and photographically active; N rays are said to have "neither fluorescence nor photographic action" [read: p.11]. See Claim 4 for the numbers: at the wavelengths he later assigns, metals are opaque by orders of magnitude [calc below].

c. Tested? Yes, by others. Wood, visiting Blondlot's lab, covertly swapped the supposed source (a file) for wood and still got reports of N rays [web] https://en.wikipedia.org/wiki/N_ray. Kelvin, Crookes, Lummer and Rubens failed to reproduce it; Rubens attempted at the Kaiser's request and reported failure after about two weeks [web] same URL. Replication outside Nancy did not occur ("By 1905 no one outside Nancy accepted N-rays", same URL).

d. If true? A broad new radiation from every warm body that crosses metals but not water would overturn the understanding of black-body (thermal) emission, matter-radiation absorption, and detector physics; nobody would have missed it in radium or X-ray work.

Verdict. Tested, not supported (Wood, Rubens, Lummer, others [web]); the penetration part additionally Refuted by calculation when taken together with Claim 4's wavelengths (see calculation there).

Claim 3: N rays behave like light: refracted, reflected, diffused, dispersed (aluminium-prism indices 1.04 to 1.85), diffracted

What the book says [read: book p.7-12, 53-57]. Quartz prism and lens refraction; reflection by polished glass; diffusion by ground glass; aluminium prism of 27 deg 15 min with indices 1.04, 1.19, 1.29, 1.36, 1.40, 1.48, 1.68, 1.85, each located by a slit filled with phosphorescent calcium sulphide, with a re-check by a 60 deg prism and a lens (which gave 1.15 rather than 1.19 for the second index) [read: p.54-55]. Check: with mean 1.42, minimum deviation in a 60 deg prism is predicted 30 deg 38 min and observed 31 deg 10 min; for 1.67 predicted 57 deg 42 min versus 56 deg 30 min [read: p.56].

a. Mathematics. Snell's law use is fine. Worked check [calc]: minimum deviation D = 2 arcsin(n sin(A/2)) - A. For n = 1.42, A = 60 deg: sin(A/2)=0.5, n sin = 0.71, arcsin = 45.23 deg = 45 deg 14 min, D = 30.5 deg, consistent with his 30 deg 38 min (my 45 deg 14 min vs his 45 deg 19 min incidence is a small rounding difference). So the arithmetic is right; the doubt is about the data, not the algebra. Problem: a dispersion of 8 well-separated discrete indices (not a continuum) means 8 lines in one beam from a thermal Nernst lamp, which is a continuous source.

Further, a curve in his dispersion appendix plots index minus 1 against wavelength and shows it rising with wavelength [read: p.70, partly OCR; plot axes only]. In every ordinary medium at short wavelengths, n-1 is negative and goes as lambda squared (index below 1 above the absorption edges, shrinking toward 1 as lambda decreases) [memory, high]. Blondlot's n-1 = 0.04 at lambda 0.0082 um and 0.85 at 0.0176 um rises by a factor of 21 for a doubling of lambda (lambda squared would give 4.6) and has the wrong sign [calc from the table, read p.58].

b. Consistent with experiment? No. For extreme ultraviolet (EUV) light of 8 to 18 nm, ordinary refraction is very weak and n slightly below 1; EUV optics use mirrors (multilayers) because no lens or prism material has n differing usefully from 1 and none is transparent [memory, high]. Quartz is opaque there, so a quartz prism or lens cannot refract it (index 2.9 for quartz is fine for far infrared of order 50 to 100 um [memory, medium; web check failed because the search quota was exhausted]).

c. Tested? The prism experiment was directly tested by Wood: he removed the aluminium prism in the dark and the observer kept "reading" the dispersed positions [web] https://en.wikipedia.org/wiki/N_ray. That test attacks precisely the data behind indices 1.04 to 1.85.

d. If true? Aluminium would need to be a prism-grade dispersive medium for EUV with n greater than 1 (opposite to every measured optical constant), and index would have to rise with wavelength (anomalous dispersion over a decade, with no absorption lines).

Verdict. Tested, not supported (Wood prism test); the dispersion law additionally Refuted by calculation / accepted dispersion theory (wrong sign and size of n-1 at 8 to 18 nm) and Internally inconsistent across his own wavelength scale: the same radiation family was said to refract in quartz at index 2.9 (far-infrared-like) and in aluminium at 1.04 to 1.85 (EUV-like).

Claim 4: N-ray wavelengths of about 0.008 to 0.018 um (8 to 18 nm), measured by gratings and Newton's rings, error under 4 per cent

What the book says [read: page image p.83, book p.58-60]. Table (um): index 1.04: 0.00813 / 0.00795 / 0.00839 for 200 / 100 / 50 lines per mm, probable 0.00815; 1.19: 0.0093 / 0.0102 / 0.0106, probable 0.0099; 1.4: 0.0117; 1.68: 0.0146; 1.85: 0.0176 / 0.0171 / 0.0184, probable 0.0176. The Newton's-ring method gave 0.0085 (versus 0.0081) and 0.017. He says "I cannot with certainty indicate the degree of approximation... I believe, nevertheless, that the relative errors do not exceed 4 per cent."

a. Mathematics [calc]. - Grating: d = 1 mm / 200 = 5 um. First-order angle sin(theta) = lambda/d = 0.00815/5 = 1.63e-3 rad = 0.093 deg. Tenth order: 0.93 deg; he measured the angle between left and right tenth fringes (about 1.9 deg) with a mirror, telescope and scale. Measurable in principle. For 50 lines per mm (d = 20 um), lambda = 0.0084 um gives 0.024 deg spacing, about 0.4 mrad, hard to read by eye on a faint phosphor slit 0.07 mm wide. That precision on a human-eye brightness judgment is implausible. - Consistency of his error claim: the three gratings for index 1.19 give 0.0093, 0.0102, 0.0106; the mean 0.0100 and spread (max-min)/mean = 0.0013/0.0100 = 13 per cent, about 6 to 7 per cent either side of the "probable" 0.0099. Even for index 1.04 the spread is (0.00839-0.00795)/0.00816 = 5.4 per cent. The "under 4 per cent" statement contradicts his own table. - Newton's rings: yellow 0.589 um over 0.0085 um is about 69 N-ray rings per yellow ring, which he treats as counted by eye on a phosphor [read: p.59]. Even if one pretends this is possible, Newton's rings require reflection from two surfaces; EUV reflectance at near-normal incidence from ordinary glass is of order 1e-3 or lower [memory, medium], so no fringes. - Photon energy: E = 1240 eV nm / lambda = 1240/8.15 = 152 eV for the shortest, and 1240/17.6 = 70 eV for the longest.

Penetration check [calc, with [memory, medium] inputs; web check failed]. At 70 to 150 eV, the attenuation length (1/e depth) in aluminium is of order 0.05 to 0.1 um, and 100 um would be a very generous upper bound. Beer-Lambert: transmission T = exp(-x/L). With x = 1.5 cm = 15,000 um and L = 100 um, T = exp(-150) = 7e-66. With L = 0.1 um, T = exp(-150,000). For a 3 cm oak board, or copper 65 cm, even L = 1 mm gives exp(-650). Result: nothing at 8 to 18 nm can come through the screens Blondlot used (aluminium foil, 2 cm deal board, black paper, [read: book p.53]). The conclusion does not depend on the exact L, only that L is below a few mm.

b. Consistent with experiment? No: EUV is absorbed by air within centimetres to metres [memory, high] and cannot be handled in an open laboratory with wet cardboard slits. Real EUV work (13.5 nm lithography) requires vacuum.

c. Tested? No independent measurement of N-ray wavelength exists; the only test in the record is Wood's and the failed replications [web: https://en.wikipedia.org/wiki/N_ray]. Not dismissed without testing: tests were performed, though not on the grating data specifically.

d. If true? An 8 to 18 nm radiation traversing centimetres of metals and 65 cm of copper would break the known link between photon energy and absorption cross sections (photoabsorption in the 50 to 200 eV range is large for every element). It would require a massless carrier that does not interact with electrons in a standard way.

Verdict. Refuted by calculation (wavelengths he reports, combined with metal transparency, are incompatible with the attenuation law above) and Internally inconsistent (error bound versus own table; also the long-wavelength reversal noted in Corrections). Also Tested, not supported.

Claim 5: N rays are detected as a small brightening of a spark or dim phosphorescent surface (more seen normally, less tangentially) by the unaided eye under special viewing rules

What the book says [read: book p.72-76, 79-83]. Viewed normally, N rays make a dimly lit surface brighter; viewed nearly tangentially, dimmer; intermediate, "no appreciable effect" [read: p.72]. "Only the observer placed exactly in front of the sensitive screen perceives the effect... how illusory it would be to try to make an audience witness these experiments" [read: p.72-73]. Instructions: dark room, silence, no tobacco smoke, do not fix the eye on the source, look "like an impressionist painter", "some people, in fact, never succeed" [read: p.82-83]. The calcium-sulphide screen shows "some of the spots less luminous than the others... nebula", and speaking aloud or whistling "all the spots become distinct" [read: p.79-80]. Only the spark photographs offered (Figs. 2 and 3) are of the spark, not the rays [read: p.16].

a. Mathematics. No quantitative claim; a brightness change in a "normal" vs "tangential" direction is a modification of the angular distribution of emitted light, with no model. Not testable to a number.

b. Consistent with experiment? The physiological side is well established: in the dark, faint light is seen best with averted vision (not looking directly at it), perceived brightness of a barely visible patch fluctuates on a time scale of seconds, and expectation shifts it [memory, high]. Blondlot's viewing rules (look away, be "passive", no effort) are exactly the conditions that maximize those fluctuations. An effect that only the "sensitive" can see, that cannot be shown to an audience, and that is spoiled by "effort" or tobacco smoke is what subjective perception looks like.

c. Tested? Yes. Wood's blinded-in-practice test: with the prism removed or the source replaced by wood, observers still reported N rays [web] https://en.wikipedia.org/wiki/N_ray. Rubens and others tried and failed [web, same URL]. The Wikipedia text reports about 120 scientists and about 300 papers, but I did not check that against the primary list, and independent positive reports were largely French [web, same URL; counts not independently verified]. No instrument detection independent of the eye is in the book [read].

d. If true? Phosphor glow would have to depend on an unseen radiation that is reversible with viewing angle (more light in one direction, less in another, conserving total emission). Known luminescence physics (phosphor excitation, thermal quenching) would have to be extended by a new quantum state controlled by a non-photon agent that is also produced by speech.

Verdict (split). - The perceived brightness changes as human experience: Plausible, untested as a psychological effect and consistent with known vision; not tested here as such. - N rays as the cause: Tested, not supported (Wood; failed replications). - Detection by the eye under observer-dependent rules, as Blondlot states it, is also close to Not testable as stated by its own admission (cannot be shown to an audience).

Claim 6: Bricks, stones, sea water and ox eyes store N rays after sun exposure; compressed or strained bodies emit them indefinitely

What the book says [read: book p.44-52]. Sea water, bricks, stones and an ox eye "store up" N rays after sun exposure and release them later; sodium hyposulphite is "a powerful accumulator"; compressed wood, glass, rubber, bent canes, hardened steel, Rupert's drops (tempered-glass drops), hammer-hardened brass emit N rays; softened steel does not; a chisel active when hard, inactive when tempered out. A Gallo-Roman knife of the Merovingian era (a dated blade) "emits N rays just like a modern knife", so emission lasted "more than twelve centuries" [read: p.51]. Energy "is most likely borrowed from the potential energy corresponding to the strained state of tempered steel" [read: p.52].

a. Mathematics. Energy budget, my calculation [calc, inputs from [memory, medium]]. Strain energy density in highly stressed steel: u = sigma^2/(2E) with sigma about 1 GPa and E about 200 GPa gives u = (1e9)^2/(2 x 2e11) = 2.5e6 J/m^3 = 2.5 J/cm^3. If it were all released over 1200 years (3.8e10 s), the average power would be 2.5/3.8e10 = 6.6e-11 W per cm^3. At 100 eV per photon (1.6e-17 J) that is about 4e6 photons per second per cm^3, spread over all directions. So the claim is not refuted by energy conservation alone, which is a point in fairness to Blondlot. But the tempered state would have to relax to produce this, and tempered steel is not releasing stored elastic energy: the hardness comes from a metastable martensite structure, and that does not decay visibly over centuries at room temperature [memory, medium]. Nor has any heat or radiation from knives been measured. And at this rate the emitted EUV photons could not cross the aluminium plate, 1.5 cm thick, the book says they traverse (see Claim 4).

b. Consistent with experiment? Storage and slow release of light after sun exposure is real in phosphors (calcium sulphide, strontium aluminate); Blondlot himself calls it "phosphorescence" [read: p.46]. Salt water and an ox eye storing a radiation for hours are not known behaviors; water has no suitable long-lived trap states [memory, medium]. No known radiation is emitted indefinitely from hardened steel at detectable intensity.

c. Tested? Wood's swap of the supposed N-ray emitter (a steel file) for a piece of wood, with observers still reporting N rays, is a direct test of the steel-emission claim [web] https://en.wikipedia.org/wiki/N_ray. I found no source testing the storage claims (bricks, sea water, ox eye) separately; they rely on the same eye-and-phosphor detector, so Wood's result bears on them by extension only.

d. If true? Condensed matter physics would need a long-lived radiative channel in strained or sun-exposed solids and liquids, detectable at room temperature without heat, with a vacuum-ultraviolet-like quantum. It would also give a new use for archaeological metal (a dating tool), and nothing like it appears elsewhere.

Verdict (split). - Hardened-steel/strain emission: Tested, not supported (Wood's file-for-wood swap [web]); energy alone does not rule it out (my calculation shows about 6.6e-11 W/cm^3, small). - Storage by bricks, stones, sea water, ox eyes: Never properly tested on its own (not found in any source I checked; it was not dismissed without testing so much as buried when the whole program collapsed); inherits the Claim 5 detector problem.

Overall

Blondlot's reported measurements are the output of one detector, the human eye at the threshold of vision, and every quantitative result (indices, wavelengths) depends on it. The arithmetic of his optics is sound, the physics is not: wavelengths of 8 to 18 nm are incompatible with passing through aluminium, wood, copper and with refraction at index above 1. The reasoning I could not confirm on the web is flagged: aluminium attenuation length at 70 to 150 eV (conclusion robust to the exact value) and quartz far-infrared index (search quota exhausted). Roughly the whole case rests on the fact that Wood's covert prism removal and source swap left reports unchanged [web: https://en.wikipedia.org/wiki/N_ray].