RankingsPrimerArbiterPhysicistMathematicianContrarianQESites logClaimsSourcesResearch notes

Science Contrarian review: SchusterAndHemsalech

Reviewer: Science Contrarian (Step 2). Date: 2026-10-04. Source: one document, A. Schuster and G. Hemsalech, "The Constitution of the Electric Spark" (abstract), Proc. Royal Society 64, pp. 331-336 (read 2 Feb 1899). I read all six pages of the extracted text myself, and the Step 1 deep note. The full paper (Phil. Trans. A 193, 1900, 189-213) is NOT in the project folder and I did not read it; the Investigator did not download it either (research/files/ is empty). Conclusions about the full paper rest on secondary citations (Royds, Hemsalech's thesis, Schaeffer) that the Investigator read in part.

Tag key: [read] from the project document; [web] URL given; [memory] with confidence; [calc] my own arithmetic.

Where this project sits

This is not a fringe claim. It is a mainstream 1899 measurement by an FRS, refereed by Crookes and Abney, who both recommended publication of the full paper (Crookes, 13 Feb 1899: "The paper marks an important step in understanding the electric spark", archivist's summary) [web] https://makingscience.royalsociety.org/items/rr_14_230 . So the contrarian question here is not "was a true anomaly suppressed?" (nothing in the record says so; the work was continued, cited in the 1911 Britannica and in 2018-2022 plasma papers per the Investigator). The useful question is narrower and more interesting: what does a slanted line on a moving film actually measure? A contemporary tested that, and the question was left open.

Investigator's findings that I weigh (kept whatever my verdict)

  1. Hull 1906 Doppler test (Proc. R. Soc. A 78, 80-81). I read the original text myself in the archive.org djvu text. [web] https://archive.org/stream/proceedingsroya02britgoog/proceedingsroya02britgoog_djvu.txt . Hull (Dartmouth, work at the Cavendish): with Cd/Hg amalgam electrodes in air, "no variation of the wave-lengths of the strong Cd, Hg, lines greater than 0.002 tenth-metres" between viewing along the discharge and at right angles; "Approximately the same result held good when a small capacity was inserted in the circuit, but in this case the discrepancies in the readings were larger." Conclusion: "the luminous particles do not acquire a velocity in the direction of the discharge greater than 150 metres per second. Hence the curving of the image ... as in the Feddersen experiment, and as recently studied by Schuster and Hemsalech for individual spectral lines, appears to be due not so much to motion of luminous particles as to the propagation along those particles of a condition of luminosity." [web, same URL]. Check [calc]: 0.002 A at 4800 A is v = c x 0.002/4800 = about 125 m/s, so Hull's 150 m/s is consistent with his stated resolution; a coherent 500 m/s drift would shift a line by about 0.008 A, four times his limit.
  2. Royds 1908 (Phil. Trans. A 208, 333) from Schuster's own laboratory: Schuster and Hemsalech "did not discuss any possibilities of the propagation of luminosity through vapour which may be, as a whole, stationary." (Investigator, read from page image p.333.) Same alternative as Hull, raised by a colleague.
  3. Referee reports (Crookes, Abney) accepted the work; Crookes asked for more on metal purification. Archivist's summaries only; images not transcribed.
  4. Hemsalech's 1901 Paris thesis (Investigator read parts): extends the self-induction result to 14 metals in two groups; iron-group lines strengthened, Cd, Zn, Mg, Al, Sb, Sn, Bi, Pb, Cu, Ag weakened; argues spark-vapour temperature is not much above arc temperature.
  5. Electrode origin conflict: abstract says bands extend "from both poles"; Hemsalech 1906 reportedly "solely from the positive electrode" (Nature 1908 snippet); Schaeffer 1908 says vapour forms at the negative pole. Three statements, none checked against the originals by me.
  6. Schaeffer 1908 credits the discovery that self-induction makes air lines disappear to Schuster and Hemsalech (Google snippet via Investigator); Royds 1908 repeated the film method. This is replication of the film method and of the self-induction effect by other laboratories, but Royds and Hull are the only ones I know of who questioned the velocity interpretation.

Two checks I did that Step 1 did not

A. Identity of the "4925 double line". The authors call 4924.8 "the double line" of zinc. The NIST Atomic Spectra Database lists Zn II (singly ionised zinc) lines at 4923.889 (intensity 7) and 4924.013 (intensity 500), the 3d10 4d 2D to 4f 2F pair, and no strong neutral zinc line there. [web] https://physics.nist.gov/cgi-bin/ASD/lines1.pl (query Zn I, Zn II, 4920-4930 A; I fetched this directly). The paper's 4924.8 differs by 0.8 A from this, which is within what a prism-based wavelength scale of 1899 could allow, so I rate the identification medium-high, not certain. The blue triplet at 4810.7 is the neutral Zn I line (4810.5) [memory: high]. If correct, the "uniformly higher velocity" of 4925 versus 4811 (Table I) compares an ion line with a neutral line. Charged and neutral emitters would plausibly move differently in a field, and the authors' own explanation (molecules of different mass) and their alternative (doublet measurement artefact) both miss this. In 1899 "ion" in this sense was new (Thomson's electron 1897) [memory: high], so this is not a criticism of them; it is a lead the reader can check against the full paper's line list.

B. The speed-of-sound formula [calc]. The sound speed is v = sqrt(gamma R T / M). With M = 2 rho x 10^-3 kg/mol (rho "referred to hydrogen" meaning H2 = 2; 56 gives M = 112 g/mol, the Cd atomic mass, so cadmium vapour is treated as single atoms) this is v = k sqrt(T/rho) with k = 76.3 for gamma = 1.4 and 83.3 for gamma = 1.67. The authors' 80 lies between (gamma about 1.55). Their T = 2700 K is reproduced: 80 x sqrt(2700/56) = 555 m/s. Using the monatomic value 1.67, T = 2530 K; with 1.4, T = 3020 K. So the temperature is good to about +/-10 percent given the assumption, and the arithmetic is sound. But the assumption that the measured slope is a gas speed, rather than a luminosity front, is exactly what Hull challenged, so the 2700 K is conditional.

C. Timing consistency [calc]. Oscillation period 2 x 10^-6 s with C = 0.033 microfarad implies L = T^2/(4 pi^2 C) = about 3 microhenry, a reasonable loop inductance. Cadmium "reaches the centre" (5 mm from a pole) in 6 x 10^-6 s, i.e. about 830 m/s, higher than the quoted mean 560 m/s (the 560 is a different average). The 2 mm baseline of V1 at about 500 m/s takes about 4 x 10^-6 s, roughly two oscillation periods. The authors admit "the periodic time was rather too small ... The metallic molecule ... was possibly affected in its motion by the subsequent oscillation." So the velocities are averages over several current reversals, each of which would re-light vapour already present. That is how a luminosity front could mimic a drift. Because the oscillation period scales as the square root of C (my inference, not in the paper), the odd result that velocity falls as capacity rises at 0.5 cm (814, 556, 416 m/s for 2, 4, 6 jars) is what a timing-dependent luminosity pattern would produce better than a thermal-expansion picture. That is my speculation, flagged as such; the authors only call it "very curious".

Corrections to Step 1

  1. facts.json key claim 1 says vapour "crosses a 1 cm gap in about 6e-6 s". The abstract says the vapour "reach[es] the centre of the spark (the gap being 1 cm long)" in about 6 x 10^-6 s for cadmium [read: p.6]. Reaching the centre means 5 mm. claims.md item 1 states it correctly; facts.json should be fixed. The summary paragraph in facts.json states it correctly too.
  2. claims.md item 5 gives only "no simple relation to atomic weight". The abstract also says "those having comparatively low atomic weights, viz., aluminium and magnesium, have higher molecular velocities" before concluding "There is obviously no simple law" [read: p.4]. Royds's one-line summary ("The lighter metals had a greater velocity") is therefore not in conflict with the abstract; it reports the first half. The Investigator flagged this as a discrepancy; it is a difference of emphasis. Both halves belong in claim 5, including bismuth's 1420 m/s on some lines.
  3. facts.json says the deep read and digest found "no deep read file exists"; one does exist (deep/TheConstoftheElectSpark-4e82a8.md). Minor.
  4. The "[memory] medium" note that Hemsalech later worked in Paris is now supported: his 1901 thesis was published in Paris (Hermann) [web] https://archive.org/details/recherchesexpri01hemsgoog (Investigator).
  5. facts.json "mainstream_conflict: None evident". This is too strong. Hull 1906 directly contradicts the moving-vapour reading of the slopes (above), and the Investigator logged it. Correct to "contested interpretation (Hull 1906)".
  6. Nothing else numerically wrong: Table I values, disc speed (about 104 m/s at the mean annulus radius [calc, deep note]) and the 80 sqrt(T/rho) reconstruction check out.

Claim-by-claim review

Claim 1. Metal vapour crosses the gap at measurable speed (about 6 x 10^-6 s to the centre for Cd; 400-1500 m/s)

a. Mathematics: Internally consistent as reported. Velocities are derived from slope on a film moving about 100 m/s, with the time axis fixed by the disc. The authors state large measurement uncertainty (displacements "subject to considerable variations", "a very slight error in measurement will also affect the results considerably", doubtful values marked "?"). The table is not monotonic (1.54 cm, 4811: 1210, 1526, 492?), so precision is poor, as they admit. Note that V1 is a mean over the first 2 mm and V2 an average of several points, two different quantities that the abstract does not always label. b. Experimental consistency: The observation (metal lines slanted, air lines straight) is a direct photographic fact, repeated by Royds 1908 in the same laboratory, and by Schenck 1901 and Milner (cited by Strutt 1913; not read by me). Hull 1906 found no Doppler shift above 150 m/s in a Cd/Hg spark in air, which conflicts with a drift of 500 m/s under the Doppler reading. His conditions (small capacity, not the six-jar bank; time-integrated; viewing along vs across) differ, and a symmetrical outflow from both poles could cancel a net shift while broadening the lines. I have not verified whether Hull's lines were broadened; his note does not say. So the conflict is real but not conclusive. c. Tested: Measured by the authors, repeated by Royds (Manchester, Schuster's laboratory, same method). Independently probed by Hull with a different method, result negative at his sensitivity. No test that distinguishes moving vapour from moving luminosity at the same conditions (for example Doppler measurements on a gated six-jar spark) is on record that I know of. Not dismissed; discussed in print. d. If true: very little changes; it extends gas dynamics to the spark gap. If instead Hull is right, then the slopes measure the speed at which excitation propagates, a different quantity with different physics (a conduction or ionisation front). Verdict: Split. (i) The observation of slanted lines on the moving film and the apparent propagation speeds: Established as a measurement (repeated by Royds). (ii) Interpretation of those speeds as the bulk speed of vapour: Never properly tested at the authors' conditions, with one contradicting preliminary test (Hull 1906, "tested, not supported" at his conditions only); Royds, from Schuster's own lab, flagged the alternative. I do not call it refuted.

Claim 2. Moving film: metal lines inclined, air lines straight; slope measures the motion

a. Mathematics: Geometry is sound: a source that is steady in space and brief in time gives a straight line; one whose position of emission changes with time gives a slope equal to (film speed) x (dx/dt). 100 m/s film speed against 500 m/s signal means the line is inclined at roughly 10 degrees to the time axis, hence measurable, and small errors matter (as they say). b. Experimental consistency: Consistent with Feddersen 1862 and with all later moving-film work. The confusion between particle motion and luminosity front (Hull; Royds) is the open point. The authors' own caution applies: "the path of the metallic particles is not always straight, and, if straight, its image does not necessarily coincide with the slit". c. Tested: the method was adopted and repeated by others (Schenck 1901, Royds, Milner); the interpretation of the slope was challenged in 1906 and 1908. Nothing I found settles it. d. If true: nothing; it is a method. The modern line of time-resolved spectroscopy (Kerr cells, streak cameras) descends from it (Springer snippet per Investigator). Verdict: Established as an experimental method; the reading "slope equals particle velocity" is Never properly tested (see claim 1).

Claim 3. Discharge starts through air; later oscillations pass through metal vapour; self-induction removes air lines

a. Mathematics: No equations; qualitative argument is coherent. The self-induction reasoning (a coil lengthens the discharge and gives time for vapour to fill the gap) is dimensionally reasonable: with L larger by a factor of 100, the period grows tenfold. b. Experimental consistency: The self-induction result was reported by Hemsalech in Comptes rendus 129, 285 (1899), extended to 14 metals in his 1901 thesis, and credited to Schuster and Hemsalech by Schaeffer 1908 and by de Gramont (Investigator, snippets; the thesis read in parts) and in the 1911 Britannica. This became a standard tool of spark analytical spectroscopy [memory: medium]. That is a real replication and use. The thesis adds a complication the abstract does not: iron-group lines are strengthened by self-induction and the spark becomes arc-like, so air-line suppression is not universal in effect on the metal lines. c. Tested: by Hemsalech (the co-author), then by others (Schaeffer; Royds). The electrode-origin question (both poles, positive only, negative only) shows that the detail of where the vapour comes from remains disputed between the cited sources; I have not read Hemsalech 1906 or Schaeffer 1908 in full. d. If true: nothing in mainstream physics; Feddersen's view that torn-off particles do not take part in the discharge is revised, which the abstract itself says is only partly right ("some evidence that this is partly true, especially with magnesium"). Verdict: Plausible, and supported by experiment in its qualitative form (air lines fade with added self-induction; replicated). The stronger statement, that the later oscillations run entirely through metal vapour, is not directly measured in the abstract (no current-path measurement) and is Never properly tested in the sense of a neutral current-path probe; it is inferred from spectra.

Claim 4. Speed of the vapour approximates the speed of sound in it; V = 80 sqrt(T/rho); T about 2700 K for Cd; molecule not lighter than normal

a. Mathematics: Consistent. My check shows the coefficient 80 sits between gamma = 1.4 (76) and 1.67 (83); the authors' reproduction of 2700 K holds, and 2530-3020 K spans the plausible gamma range. The step to "mass not much smaller than that determined near the boiling point" is correct for what it says: for a fixed v and T, a lighter emitter would need a lower T. It is circular in the sense that the temperature is not independently measured; they say 2700 K "seems a possible value", which is the correct level of claim. b. Experimental consistency: Hemsalech 1901 argues that spark vapour temperature does not exceed the arc temperature by much (thesis, per Investigator). Modern spark plasmas have higher electron temperatures and ionisation [memory: medium; not checked on the web]; but late-time vapour near the centre can be cooler. The measured line emission is from ions in several cases (see my check A), so a vapour that is "molecules" of cadmium is already partly ionised; the paper's neutral-molecule picture is a 1899 simplification. c. Tested: no independent temperature measurement in the project documents. The Hull result bears on whether the premise (the slope is a gas speed) holds. d. If true: nothing changes. Verdict: Plausible, untested. The arithmetic is internally consistent; the physical premise is contested (claim 1).

Claim 5. No simple relation of velocity to atomic weight; different lines may give different velocities; velocity falls with capacity at short gaps

a. Mathematics: The table is internally consistent with the text: 6-jar values are about 416-545 for gaps 0.51 and 1.03 cm and "practically identical" in the authors' words; the odd 1.54 cm entries are marked doubtful. I confirm that for 0.51 cm the 4925 line gives 814, 556, 416 and the 4811 line gives 1014, 668, 529 (two, four, six jars), i.e. a fall of about a factor of two, as stated. b. Experimental consistency: The abstract itself says low-atomic-weight Al and Mg are faster (Al over three times Zn, "numbers not laying any claim to accuracy"; Mg too scattered to measure) while Bi gives 1420 m/s on some lines and Zn equals Cd; hence "no simple law". This is a statement about scattered data. The 4925 vs 4811 difference may be explained by the doublet-measuring artefact the authors propose, or by the ion/neutral difference I noted (check A), or by a real velocity difference; the abstract defers to higher dispersion. The decline with capacity runs against the authors' expectation and they decline to call it established. c. Tested: Royds 1908 gives a velocity table by metal and line and reports differing inclinations "most marked in the bismuth spectrum" (the same pattern) and impurity calcium H and K lines (from silver electrodes) that light first at the centre of the gap and propagate to the electrodes. That observation is hard to read as vapour drifting from a pole (the Ca vapour must come from the electrodes), and fits a luminosity-propagation reading better. It is single-sourced (Royds, read by the Investigator from the page image), and from Schuster's own laboratory. I have not seen the plate. d. If true: for the line-dependent velocities, if real, they would show that different emitters move differently (ions vs neutrals), which affects how spark spectra are interpreted. If a measurement artefact, nothing. Verdict: Not testable as stated for the atomic-weight statement (a null statement about scattered data; "no simple law"); the line-dependent velocity and the capacity trend are Never properly tested (the authors themselves reserve judgement). The Royds calcium observation should be kept as pertinent evidence for the luminosity-propagation reading.

Was the work wrongly rejected? How it was treated

Nothing was rejected or suppressed on this record. It was published in the Proceedings and then in the Philosophical Transactions after favourable refereeing; it was repeated and extended, and the self-induction technique was adopted. Hull's 1906 critique is a one-page preliminary note with a different apparatus; I found no later rejoinder from Schuster, and I did not see Hull follow it up (not searched). So the open question is a neglected one: nobody I can find repeated Hull's Doppler test on a gated, six-jar spark. It lies in the category "never properly tested," not "dismissed without testing."

Rigour applied to the authors

The authors hedge honestly: they flag doubtful values with a query, say the Al numbers do not claim accuracy, withhold judgement on the doublet and the capacity trend, and admit the oscillation period was too short. The weaker points are their own: V1 and V2 are mixed in the text, the 6 x 10^-6 s centre time (830 m/s) differs from the quoted 560 m/s average without comment, and they do not discuss that vapour arriving from two poles meets in the middle. No unit errors found. No unfalsifiable claim.

What would settle the open points

  1. Read the full 1900 Phil. Trans. paper (tables for all metals, line list with wavelengths) and Royds 1908 pp. 334-347 (velocity table, calcium observation).
  2. Check the identity of the 4925 and 4811 lines (Zn II vs Zn I) in that line list.
  3. A gated Doppler measurement on a six-jar spark of 0.033 microfarad, viewed along and across the axis (Hull's test under the 1899 conditions). Equivalent today: streak-camera plus high-resolution spectrometer on a pulsed spark.
  4. Hull's note is preliminary; check whether lines were broadened.

Verdict summary

  1. Vapour crossing the gap at speed: observation Established; reading as bulk vapour velocity Never properly tested (one contradicting preliminary Doppler test, Hull 1906).
  2. Moving-film method: Established as a method; slope-equals-particle-velocity Never properly tested.
  3. Air first, then metal vapour; self-induction removes air lines: Plausible, replicated in qualitative form; strong form (all later oscillations through vapour) Never properly tested.
  4. Speed of sound, V = 80 sqrt(T/rho), T about 2700 K: Plausible, untested (arithmetic consistent).
  5. No simple atomic-weight law; line and capacity effects: Not testable as stated (null statement); line-dependent and capacity effects Never properly tested.