Compton: Science Contrarian review

Reviewer role: Science Contrarian (judges by experiment; separates "tested and failed", "never properly tested", "dismissed without testing"; holds the researcher to the same standard as the critics).

Source gap (repeated from Step 1, narrowed). The only document is Compton_1923.pdf, an image-only scan with no text layer. In this review I rendered and read pages 1, 3, 5, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20 as images [read: page image p.N]. Pages 2, 4, 6, 7, 8 were not re-read (Step 1 read 2, 4, 6). Everything about later events (Bothe-Geiger, Compton-Simon, Duane, BKS, Klein-Nishina, the unmodified line) is [web] where a URL is given, otherwise [memory] with confidence marked. Calculations below are mine, run in the project venv with numpy (constants h, m, c from CODATA values; Thomson cross-section 8pir_e^2/3).

Corrections to Step 1

1. facts.json summary, scattering coefficient ratio is wrong. It says the total energy removed is less than Thomson's "by the ratio 1/(1+2 alpha)^2". The abstract and Eq. 28 [read: page image p.1, p.11] give sigma = sigma_0/(1+2 alpha), i.e. ratio 1/(1+2 alpha). The square appears only in the split: sigma_s = sigma_0 (1+alpha)/(1+2 alpha)^2 and sigma_a = sigma_0 alpha/(1+2 alpha)^2 (Eqs. 29-30, [read: page image p.12]); these sum to sigma_0/(1+2 alpha). (Checked numerically: the integral of Eq. 27 over the sphere reproduces Eq. 29 exactly.)
2. claims.md claim 4 conflates total and scattered. "Total scattering is too high by the factor (1+2 alpha)^2/(1+alpha)" is true only of the re-radiated (true scattering) part. The total energy removed is overestimated by Thomson by (1+2 alpha).
3. Gamma-ray "wavelengths" are not measured spectroscopically. Step 1 says "wavelength measured via absorption in lead", which is right, but claim 3 presents 0.022 to 0.068 A as an observation. They are inferred from lead mass-absorption coefficients via an extrapolation using tau/rho proportional to lambda^3 and Hull-Rice data at 0.122 A, and the primary value 0.022 A is itself fixed by the same method [read: page image p.15]. Table I (observed vs calculated, A): 45 deg 0.030 vs 0.029; 90 deg 0.043 vs 0.047; 135 deg 0.068 vs 0.063 [read: page image p.15]. The 135 deg point differs from theory by about 8 percent, and the 90 deg point by 9 percent in the other direction.
4. Claim 5, the 1.10 forward/back ratio. Evaluating Eq. 27 myself at 0.708 A (alpha = 0.0343) gives I(40 deg)/I(140 deg) = 1.134, not "about 1.10". Compton's 1.10 is stated for roughly 0.7 A and not derived on the page; Barkla-Ayers measured 1.04 [read: page image p.17]. Klein-Nishina photon-number ratio at the same alpha is 1.105. So the paper's "about 1.10" is closer to what the later correct theory gives than to its own Eq. 27. Minor.
5. Omitted from Step 1: (a) Compton's own earlier empirical value was a shift of "roughly 0.03 A" at 90 deg, quoted as support for wavelength-independence [read: page image p.12], against the predicted 0.024 A; (b) the gamma-ray angular data of Fig. 7 sit systematically below the Eq. 27 curve at large angles (details under claim 5); (c) footnote on p.3 shows Compton reversed his own 1921 published position (softening = fluorescence admixture) [read: page image p.3].
6. Step 1 listed pages 5, 10-12, 14-16, 18, 19 as unread; those are now read, so the source-gap statement can be narrowed as above. Eqs. 6, 8, 21-30 were checked against my own computation and are algebraically consistent.

Web checks (budget used: 3)

• BKS theory (1924) allowed only statistical energy conservation; Bothe-Geiger (Berlin) and Compton-Simon (Chicago), both 1925, tested it by coincidence/track methods and showed conservation holds event by event. [web] https://en.wikipedia.org/wiki/Bothe%E2%80%93Geiger_coincidence_experiment ; search summary also citing https://www.ias.ac.in/article/fulltext/reso/023/09/0943-0948 . The same sources note that Debye independently derived the shift, and that Duane challenged Compton on experimental grounds.
• NIST XCOM tables, aluminium: mass energy-absorption coefficient 0.02868 cm^2/g at 0.5 MeV and 0.02851 at 0.6 MeV; mass attenuation 0.0845 and 0.0780. [web] https://physics.nist.gov/PhysRefData/XrayMassCoef/ElemTab/z13.html . Used below. (0.022 A corresponds to about 0.56 MeV, using hc = 12.398 keV A [memory: high].)
• Nobel Prize 1927 (Compton, shared with C. T. R. Wilson): check failed (nobelprize.org returned HTTP 403). Stays [memory: high].
• Not checked: that Compton's later Phys. Rev. 22, 409 (1923) papers reported an unmodified line alongside the shifted line [memory: medium]; Duane's Harvard group initially failing to reproduce the shift and later conceding [memory: medium].

Claim 1. A scattered quantum gives all its energy and momentum (h nu / c) to a single electron, which recoils.

• a. Mathematics: Consistent. Eqs. 1-2 (momentum and energy conservation) with relativistic electron kinetic energy give Eq. 3, nu_theta = nu_0/(1 + 2 alpha sin^2(theta/2)), and the recoil speed Eq. 6. I checked beta_max = 2 alpha (1+alpha)/(1+2 alpha+2 alpha^2) = 0.823 for alpha = 1.103, matching the paper's 0.82 [read: page image p.16-17] and mean beta alpha/(1+alpha) = 0.524 (paper: 0.52). The paper's own treatment is quantum on one side (photon) and classical-relativistic on the other (electron), a hybrid, but an unambiguous one.
• b. Experiments: In 1923 the supporting evidence was indirect: secondary beta-ray speeds from gamma-ray scattering "somewhat greater than half that of light" on average [read: page image p.17], and the wavelength shift. The paper admits recoil electrons from ordinary X-rays "have not been observed" [read: page image p.15]. Direct event-by-event evidence (electron and scattered quantum correlated) came in 1925 (Compton-Simon, Bothe-Geiger) [web, see above].
• c. Testing: Strongly tested afterwards and replicated by independent labs on different continents (Chicago, Berlin) [web]. Duane contested it experimentally at first [web: challenge; his later position memory: medium].
• d. If true: Light must carry momentum in individual interactions and energy conservation holds per event; classical wave scattering (Thomson) is wrong for hard radiation; Bohr-Kramers-Slater is untenable. All of this is now mainstream, so nothing "must" change today.
• Verdict: Established. The 1923 paper alone made it Plausible and well-motivated; the 1925 coincidence/cloud-chamber experiments made it Established. This was not dismissed without testing: it was contested, then tested.

Claim 2. The shift is (2h/mc) sin^2(theta/2) = 0.0484 sin^2(theta/2) A, independent of incident wavelength.

• a. Mathematics: Consistent. 2h/mc computed = 0.04853 A, h/mc = 0.024263 A, matching the paper's 0.0484 and 0.0242 (small rounding of constants). Wavelength independence follows immediately from Eq. 5.
• b. Experiments: Paper's own test is a single X-ray line: Mo K-alpha (0.708 A) off graphite at 90 deg, observed 0.730 A, shift 0.022 vs 0.024 predicted [read: page image p.13]. Reading a calcite-crystal spectrum with broad peaks (Fig. 4), a 0.002 A discrepancy is about 9 percent of the shift and 0.3 percent of the wavelength. That is consistent, but it is a one-angle, one-wavelength test, with results declared "unpublished" and promised for later [read: page image p.12 footnote]. Wavelength independence rests on Compton's earlier report of "roughly 0.03 A" over 0.7 to 0.025 A, which differs by 25 percent from 0.024 [read: page image p.12].
• c. Tested: Yes, widely afterwards (angle series and different wavelengths; memory: medium for specific later papers, e.g. Compton and collaborators 1923-25). I did not verify individual replications by web, so I do not cite them. The 1925 coincidence experiments [web] used the same kinematics.
• d. If true: Photon momentum; sets the Compton wavelength as a fundamental constant. Mainstream already.
• Verdict: Established. Caveat for the learner: the 1923 paper by itself supports it at roughly 10 percent precision, at one angle for X-rays.

Claim 3. Experiment agrees: Mo K-alpha/graphite 0.022 vs 0.024 A; RaC gamma 0.022 to 0.068 A at 135 deg.

• a. Mathematics/arithmetic: Checked. 0.022 + 0.0484 sin^2(67.5 deg) = 0.0634 (paper's Table I: 0.063); at 90 deg 0.0463 (paper 0.047); at 45 deg 0.0291 (paper 0.029).
• b. Experiments, strongest honest case: The X-ray result is a clean spectroscopic observation, a displaced Ka line, and the shift has the right size and (from p.12) the right wavelength-independence. The gamma-ray trend (0.022 rising monotonically with angle) is qualitatively what the theory says.
• Critical points: (i) The gamma "wavelengths" are inferred from lead absorption via a calibration whose scale depends on an extrapolated tau/rho ~ lambda^3 law; the unscattered primary was assigned 0.022 A because that is where the extrapolation lands, then corroborated by an indirect value 0.025 A and Ellis's beta-ray line list (0.045, 0.025, 0.021, 0.020 A) [read: page image p.15]. A fitted scale and the prediction share an assumption. (ii) The abstract states that the scattered X-rays show a longer wavelength "with little or none" at the primary wavelength [read: page image p.3 and p.1]; the later literature found both a shifted and an unshifted line (see claim 6), so this is too strong as a statement about the data (memory: medium). (iii) The Mo/graphite figure is not tabulated; no uncertainty is given. These are weaknesses of the 1923 paper, not of the later-confirmed effect.
• c. Replication: The effect itself replicated (claim 1); the specific gamma-ray numbers were never the decisive test. Not checked by web.
• d. If true: as claim 2.
• Verdict: Split. Established for the existence and size of the shift. Weakly supported / indirect for the gamma-ray quantitative agreement as presented in this paper (the evidence was indirect and ~8 percent off at 135 deg). No scale item fits "weakly supported" exactly; the nearest is Established-by-later-work, with the 1923 gamma data not independently probative.

Claim 4. Thomson scattering fails for hard X-rays and gamma-rays; the quantum coefficients match Ishino and Hewlett.

• a. Mathematics: Algebra consistent (Eqs. 24-30; I integrated Eq. 27 numerically and obtained sigma_s/sigma_0 = (1+alpha)/(1+2 alpha)^2 to 12 digits, 0.9058 at alpha = 0.0343 and 0.2046 at alpha = 1.103). But this consistency is internal to the Eq. 27 assumption (see claim 5).
• b. Experiments / what the correct modern theory gives. Klein-Nishina (1929) [memory: high, not web-verified] total cross-section relative to Thomson, my computation: 0.937 at alpha = 0.0343 (Compton's total 1/(1+2 alpha) = 0.936, good) but 0.413 at alpha = 1.103 vs Compton's 0.312 (25 percent too low). Splitting energy: KN true-scattering fraction 0.265 vs Compton's 0.205; KN energy absorbed by recoil 0.148 vs Compton's 0.107. For aluminium at 0.56 MeV, KN absorption = 0.148 * 0.193 = 0.0286 cm^2/g, matching NIST's mass energy-absorption coefficient 0.0285-0.0287 [web: NIST XCOM, above]. Compton's own Eq. 30 gives 0.021 and Ishino's measurement was 0.021 [read: page image p.17], both about 27 percent below the NIST value. So the Compton-Ishino "agreement" is agreement between a low formula and a low measurement; the quoted agreement within "probable experimental error" would not have survived better data. Similarly true scattering: Ishino 0.045 (Al), Compton 0.040, KN 0.051 (my computation).
• Hewlett's carbon data at less than 0.5 A: Fig. 6 shows Hewlett's crosses scattered around the quantum curve, and Compton concedes "the agreement is perhaps within the experimental error" [read: page image p.16]. At 0.1-0.5 A the quantum curve (0.135 to 0.183 cm^2/g by my evaluation) differs from KN (0.141 to 0.184) by less than the scatter, so Hewlett's data cannot discriminate; they show only that classical 0.201 is too high at short wavelengths.
• The failure of Thomson at short wavelength (reduced scattering, forward concentration) was established experimentally before the paper and is not Compton's inference: [read: page image p.2-3 on prior results].
• c. Tested: Qualitative failure of Thomson for hard radiation: tested, replicated, solid. The specific coefficients, Eqs. 28-30: superseded by KN and by later measurement; no sign that they were suppressed, simply refined.
• d. If true: Classical electron scattering cannot hold for photon energies comparable with m c^2; needs QED (KN is the lowest-order result).
• Verdict: Split. Thomson fails at high photon energy: Established. Eqs. 28-30 as exact formulas: Historically superseded but instructive (right at alpha << 1, up to 25-45 percent wrong at alpha ~ 1).

Claim 5. Intensity is forward-concentrated per Eq. 27, derived by treating a recoiling electron as a moving radiator with beta = alpha/(1+alpha) (assumption admitted unproven; Barkla-Ayers 1.04 vs 1.10).

• a. Mathematics: The Doppler step is correct for frequency (Eq. 7 reduces to Eq. 3 when beta = alpha/(1+alpha)). The intensity step is explicitly an assumption: "I have not, however, succeeded in showing rigidly that if two methods of scattering result in the same relative wave-lengths ... they will also result in the same relative intensity" [read: page image p.9]. The motivating support is a black-body consistency check (Eq. 20, I/I' = (nu_m/nu_m')^4), which tests the Doppler transformation of intensity, not the claimed equivalence between recoil and a moving radiator. Also the forward normalisation I_0 is set equal to the classical Thomson value (Eq. 26) by fiat [read: page image p.11]. The derivation is not contradictory, so not "internally inconsistent", but it is a heuristic.
• b. Experiments: Fig. 7 compares Eq. 27 with ionisation-chamber measurements of RaC gamma rays scattered from iron [read: page image p.18, 19]. Issues: (i) the chamber was "so designed as to absorb the greater part of even the primary gamma-ray beam", yet Compton admits "it is not clear just how the ionization ... will vary with the wavelength", assuming "it appears probable that the variation will not be great" [read: page image p.19]; the scattered quanta have much lower energy at large angle, so the chamber response very likely did vary. (ii) The paper states "beautiful agreement ... not a single adjustable constant" [read: page image p.19], but data are normalised to the theoretical I_0. (iii) My evaluation of Eq. 27 at alpha = 1.103 gives I/I_0 = 0.234 (45 deg), 0.134 (60 deg), 0.069 (90 deg), 0.038 (135 deg). The text gives the measured value at 90 deg as 0.037 I_0 [read: page image p.19]: Eq. 27 overshoots it by a factor of about 1.9. In the Fig. 7 plot the data points lie visibly above the curve near 60 deg and below it from about 70 deg to 150 deg. "Beautiful agreement" is overstated; it is a good qualitative match on a compressed scale. (iv) The correct KN energy-weighted forward-to-90 deg ratio at this alpha is about 0.085 (my computation, KN number ratio 0.178 times scattered-energy factor 0.476), i.e. also above the plotted 90 deg datum; so the data do not clearly favour Eq. 27 over KN, and neither fits the 90 deg point if the chamber response is flat.
• Barkla-Ayers: 1.04 observed vs about 1.10 predicted; Compton calls the difference beyond probable experimental error [read: page image p.17] then argues ionisation response and excess scattering explain it. That explanation is plausible but not demonstrated; it removes the discrepancy by an unquantified correction. (My evaluation of Eq. 27 gives 1.13; see Correction 4.)
• c. Tested: Tested only by those two sets of 1910s-20s measurements; no neutral, controlled test of the intensity law by Compton. Klein-Nishina (1929) [memory: high] provided the correct derivation; later measurements of angular distributions are mainstream [memory: medium]; I did not web-verify them.
• d. If true: Nothing beyond claim 1; Eq. 27 is not what mainstream physics adopted.
• Verdict: Historically superseded but instructive. Qualitative forward concentration: Established. Eq. 27 as a quantitative law: wrong by up to a factor of about 2 in the ratio of large-angle to forward intensity at alpha ~ 1 (my KN comparison: Eq. 27 gives 0.069 vs KN energy 0.085 at 90 deg), derived by an assumption the author flagged himself.

Claim 6. The theory holds only for loosely bound electrons in light elements.

• a. Mathematics: Consistent, and honestly stated. The paper identifies the tacit assumption of no constraining forces, and the criterion that recoil kinetic energy must exceed the binding energy [read: page image p.20].
• b. Experiments: The paper itself points to "excess scattering" at small angles and long wavelengths (above about 0.5 A, p.16) and X-ray reflection/interference as unexplained [read: page image p.16, p.20]. Later experiments found an unshifted line along with the shifted one from graphite [memory: medium, not web-verified].
• Researcher's own speculation to hold to account: the suggestion that a tightly bound electron lets the "quantum spread itself over a large number of electrons" [read: page image p.20] was not what the mainstream adopted; the standard account is coherent scattering from the atom or lattice as a whole (negligible recoil) plus incoherent scattering from individual electrons, with binding effects broadening the shifted line [memory: medium]. This is a speculation labelled as such, not an error.
• c. Tested: The restriction is a boundary, not a testable positive claim. Consistent with later experiments [memory: medium].
• d. If true: Requires a treatment of bound-electron and coherent scattering (done later: atomic form factors, Bragg diffraction).
• Verdict: Established in its weak form (the free-electron treatment applies when photon energy greatly exceeds binding energy). The "spread over many electrons" mechanism: Plausible, untested as stated in 1923; later replaced by coherent scattering (Historically superseded but instructive).

Contrarian assessment: how the claim was treated

This is a case where the heterodox claim won, and it is useful to say why. It contradicted the 1922 consensus (wave theory of light, Thomson scattering), met resistance from Duane experimentally and from Bohr-Kramers-Slater theoretically [web], and was tested rather than dismissed: the decisive 1925 experiments were run by independent groups [web]. The contrarian lesson cuts both ways. (1) Anomalous data (the shifted line) beat theory, as the standards demand; the quantum-vs-wave consensus had to yield. (2) Compton's paper also shows the researcher-side failings the standards require me to call out: it declares results "unpublished" while citing them; it overstates "beautiful agreement" with data that are indirect and ionisation-based; it rests Eq. 27 on an acknowledged unproven assumption; and its extrapolated gamma wavelengths and the 25 percent error in the high-alpha total coefficients were masked by large experimental error bars on Ishino's numbers. Nothing in the paper is fabricated or a unit error; I found none, and the arithmetic reproduces.

Single decisive items for a reader: the 1925 coincidence experiments settle claim 1; the NIST energy-absorption value of 0.0286 cm^2/g for aluminium at 0.56 MeV versus Compton/Ishino's 0.021 settles that the 1923 absorption formula was approximate at alpha ~ 1.

Summary table