Compton: PhD physicist review (Step 2)

Reviewer role: PhD physicist, mainstream view. Document reviewed: Compton_1923.pdf (Compton, Phys. Rev. 21(5), 483-502, 1923).

Source gap (repeated from Step 1). The PDF is an image-only scan. In this review I rendered all 20 pages and read pp. 486 (scan p.4), 491, 492, 493, 494, 495, 496, 497, 498, 499 and 502 as images [read: page image]. Statements about later work (Compton-Simon, Bothe-Geiger, Klein-Nishina, Duane, Bohr-Kramers-Slater) come from [web] or [memory] and are tagged. Numerical checks marked [calc] were run by me with numpy.

Corrections to Step 1

1. facts.json summary says the total energy removed is "less than Thomson's by the ratio 1/(1+2 alpha)^2". Wrong. Eq. 28 (p.493) gives sigma = sigma_0/(1+2 alpha), so the ratio is 1/(1+2 alpha) [read: page image p.493]. The factor (1+2 alpha)^2 belongs to the true scattering coefficient, Eq. 29: sigma_s = sigma_0 (1+alpha)/(1+2 alpha)^2. The ratio of classical to true scattering, (1+2 alpha)^2/(1+alpha), in claims.md item 4 is correct.
2. claims.md item 3 and the key_claims entry present the gamma-ray result as a measured wavelength shift "0.022 to 0.068 A". The 0.022 A primary value is not independently measured. Table I (p.496) derives all wavelengths indirectly from absorption in lead, using Hull-Rice coefficients and an assumed lambda^3 extrapolation of fluorescent absorption. The "primary" 0.022 A is itself an input to the prediction. The 0.068 A (135 deg) is an inferred value; the theory column gives 0.063, about 8% off [read: page image p.496]. Only the Mo K-alpha spectrum (Fig. 4) is a direct spectroscopic measurement.
3. claims.md item 5 says Barkla-Ayers is "1.04 vs 1.10 predicted". Page 499 says "about 1.10" at roughly 0.7 A. Evaluating Eq. 27 at 0.7 A (alpha = 0.0347) gives a 40/140 degree ratio of 1.136 [calc], not 1.10. This is minor but worth noting: Compton's quoted figure is approximate. The Klein-Nishina value is 1.106 [calc], so the later correct theory is closer to Compton's rounded number.
4. facts.json says Compton "calls the paper's results unpublished". More exactly, footnote 2 on p.494 says it is hoped to publish soon the description of the experiments behind Fig. 4. This is accurate as a footnote; the earlier statement is a loose paraphrase.
5. Step 1 lists "Eq. 27 ... because the recoiling electrons move along the primary beam". This is right, and the paper itself flags the key step as unproven (p.491), as Step 1 says [read: page image p.491].

Step 1 otherwise checks out against the pages I read.

Background in four lines

Eq. 1 (momentum) and Eq. 2 (relativistic energy) are two equations for beta and nu_theta. Solving gives nu_theta = nu_0/(1 + 2 alpha sin^2(theta/2)), alpha = h/(m c lambda_0) [read: page image p.486]. I re-derived this with the standard four-momentum argument and it is correct. In wavelength form the shift is (h/mc)(1 - cos theta) = (2h/mc) sin^2(theta/2). With modern constants h/mc = 0.02426 A [calc], so 2h/mc = 0.0485 A; Compton's 0.0484 reflects 1922 constants.

Claim 1: A scattered X-ray quantum gives all its energy and momentum (h nu/c) to one electron, which recoils

• a. Mathematics. Consistent. Eqs. 1-3 are exact special-relativistic conservation laws, treating the electron as free and initially at rest. The relativistic electron energy and momentum are used correctly. One caveat: the paper treats the quantum's momentum as h nu/c by postulate, not derivation, which is the point of the claim.
• b. Experiment. Consistent with every modern experiment on photon-electron scattering: the recoil electron energy and angle agree with the formula from kinematics alone. At the time, the support was the beta-ray velocities in light elements (p.496-497: beta_max = 0.82 for RaC, average beta about 0.52 versus "somewhat greater than half" of c observed). That check is semi-quantitative only.
• c. Tests. Compton's own data (Mo K-alpha, graphite). Decisive later tests: Compton-Simon cloud-chamber and Bothe-Geiger coincidence experiments (1925) showed the scattered photon and recoil electron appear together in individual events, which killed the Bohr-Kramers-Slater statistical-conservation proposal [web: https://arxiv.org/pdf/1106.1365 and https://www.nobelprize.org/prizes/physics/1927/compton/facts/ for the 1927 Nobel award]. Replicated in countless later experiments [memory: high].
• d. If true. Already incorporated: light carries momentum h/lambda, photons are real localized quanta in interaction, energy and momentum are conserved event by event. It forced acceptance of the photon (Einstein 1905, [memory: high]) and led toward QED.
• Verdict: Established.

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

• a. Mathematics. Internally consistent (Eq. 3 to 5 to 31). Independence of lambda_0 follows because alpha lambda_0 = h/mc cancels. Shift at 90 degrees: 0.0242 A [calc].
• b. Experiment. Consistent with all Compton-scattering measurements (X-ray, gamma, and today's keV-GeV photon scattering) for scattering off free or loosely bound electrons [memory: high]. Paper's own cross-check: the earlier Bull. NRC statement that the shift at 90 degrees is "roughly 0.03" A over 0.7-0.025 A (p.494) is a rough agreement only.
• c. Tests. Compton's Fig. 4: Mo K-alpha 0.708 A, scattered line 0.730 A, shift 0.022 A observed versus 0.024 A predicted (p.495) [read: page image p.495]. The 0.002 A discrepancy is about 10% of the shift. Fig. 4 shows a broad, low-resolution calcite spectrum and a peak separation only a few steps wide; the paper gives no error bar. Compton's later, more careful measurements (Phys. Rev. 22, 409, 1923) are the better quantitative test [memory: medium; not checked on the web]. Independent replication: by Ross, Duane and others in the 1920s [memory: medium] and by every modern Compton spectrometer.
• d. If true. Nothing beyond Claim 1.
• Verdict: Established. (The paper's single measurement is a rough 10% test; the full statement is established by later work.)

Claim 3: Experiment agrees: Mo K-alpha at 90 deg shifts 0.708 to 0.730 A; RaC gamma from 0.022 to 0.068 A at 135 deg

• a. Mathematics. Predicted numbers recompute correctly: 0.708 + 0.0242 = 0.732 A. For gamma, 0.022 + 0.0484 sin^2(67.5 deg) = 0.0633 A [calc], and Table I column "lambda calc." (0.029, 0.047, 0.063) matches the formula. Self-consistent. Weak point: the gamma "observed" column is built from absorption coefficients with an assumed 1/lambda^3-type extrapolation and a "primary" value tuned to 0.022 A.
• b. Experiment. The Mo line fits within the stated resolution. The gamma data fit to roughly 5-8% (0.030 vs 0.029, 0.043 vs 0.047, 0.068 vs 0.063). Also, the gamma-ray source RaC emits a spectrum of lines (0.045, 0.025, 0.021, 0.020 A are quoted from Ellis, p.496), and Compton picks an "effective" wavelength. At such energies Compton scattering in lead competes with photoabsorption and, as understood later, pair production [memory: medium]. So Table I is a weak test, and its agreement should be read as consistent, not as precise confirmation.
• c. Tests. Compton himself. Better data came from later spectroscopy (Compton 1923 follow-up) and cloud-chamber tests [memory: medium]. I was unable to open those papers to check numbers.
• d. If true. Nothing new.
• Verdict: Split. Mo K-alpha shift: Established (with this paper's data as an early, rough but correct test). Gamma-ray Table I as evidence: Plausible, untested in this form; the absorption method is indirect, and its success is partly the result of tuning, though the formula itself is later established.

Claim 4: Classical Thomson scattering fails for hard X-rays and gamma rays; classical total scattering is too high by (1+2 alpha)^2/(1+alpha); the quantum coefficients match Ishino's and Hewlett's data

• a. Mathematics. Classical Thomson part: sound (fixed frequency, symmetric forward-back, sigma_0 = 8 pi N e^4/(3 m^2 c^4) per volume). I checked numerically that integrating Eq. 27 over a sphere reproduces Eq. 29 exactly: at 0.7 A, 0.9049 sigma_0 both ways; at 0.22 A, 0.7453; at 0.022 A, 0.2046 [calc]. So Eqs. 27, 28, 29 are mutually consistent. The ratio sigma_a = sigma - sigma_s = sigma_0 alpha/(1+2 alpha)^2 (Eq. 30) follows algebraically. But the underlying angular law (Eq. 27) rests on an admitted assumption (p.491): that two methods giving the same frequency shift must give the same intensity change. That is a heuristic, not a derivation (see Claim 5).
• b. Experiment. Qualitatively correct: Thomson cross-section works at low energy (Fig. 6 and p.499 note it) and falls at high energy. Quantitatively, Compton's formula is not the correct cross-section. The Klein-Nishina result (1929) gives, for the same ratio of total scattering to Thomson, 0.936 at 0.7 A, 0.829 at 0.22 A and 0.413 at 0.022 A [calc], versus Compton's 0.905, 0.745 and 0.205 for the "true scattering" and 0.935, 0.819 and 0.312 for his "energy removed" sigma. His sigma (Eq. 28) happens to agree with KN to first order in alpha but diverges at gamma energies. So the match with Ishino at 0.022 A (0.040 and 0.038 calc vs 0.045 and 0.042 observed) is not a confirmation of Eq. 29 as the true law: by Klein-Nishina the true value would be about twice Compton's number for Eq. 29's scattering part at alpha around 1.1 [calc]. Ishino's numbers are from 1917 and measured with heterogeneous RaC gamma rays; I cannot verify that on the web here [memory: low]. The fair summary is that Compton's agreement at gamma energies depended on poorly characterised data and a wavelength tuned to fit.
• c. Tests. Hewlett (carbon, soft X-rays), Ishino, Barkla-Ayers, as cited in the paper. The Klein-Nishina formula, derived from Dirac's electron theory, was confirmed by later precision measurements, and replaced Compton's formula [web: https://www.sciencedirect.com/science/article/abs/pii/S1355219802000059 for the Compton to Klein-Nishina development]; [memory: high] for the specifics.
• d. If true. Qualitative failure of classical scattering at short wavelength is already accepted. Compton's specific coefficients would not need mainstream change; they were corrected.
• Verdict: Split. Qualitative claim (Thomson fails at short wavelength): Established. Compton's quantitative formulas Eqs. 27-30: Historically superseded but instructive (replaced by Klein-Nishina; they agree to first order in alpha for the total-removal coefficient).

Claim 5: Scattered intensity is forward-concentrated per Eq. 27, derived by treating the recoiling electron as a moving radiator with beta = alpha/(1+alpha); Barkla-Ayers fits imperfectly

• a. Mathematics. Internally consistent as algebra; every step I checked on pp. 491-493 follows from the premises (Doppler transformation of the dipole law (3/8)(1 + cos^2 theta') sin theta' d theta', Eq. 21, k = 3/8 by normalisation). The black-body check (Eq. 20, I/I' = (T/T')^4) is a legitimate consistency check of the Doppler intensity transformation (p.491). The premise, however, is not derived, and Compton says so ("I have not, however, succeeded in showing rigidly", p.491). It also fixes the forward intensity by classical theory (Eq. 26), another assumption. The nu_theta/nu_0 weighting in Eq. 23 ("energy scattered per second") is a correct bookkeeping of energy per quantum. A nuance: the real Klein-Nishina cross-section is a different function. At alpha = 1.1 forward/backward from Eq. 27 is 32.8, from KN 5.8 [calc]; they agree at small alpha. So the assumption is wrong at high energy, not just unproven.
• b. Experiment. The paper itself reports a disagreement: Barkla-Ayers 1.04 for the ratio versus about 1.10 predicted (p.499). Compton attributes this to wavelength-dependent ionization; this is a plausible but untested rescue. I note that Klein-Nishina predicts 1.106 at 0.7 A [calc], so the true theory also predicts about 1.1, which suggests the 1.04 figure was inaccurate or contaminated by excess scattering and ionization-measurement effects (the explanation itself is [memory: low] as to what was wrong).
• c. Tests. Only Barkla-Ayers and the assorted data in the paper; no replication in the paper. Later replaced by the Klein-Nishina formula and its tests (precision gamma-ray scattering) [memory: high; web: https://www.sciencedirect.com/science/article/abs/pii/S1355219802000059].
• d. If true. Nothing; as stated it is false in detail (shown by KN).
• Verdict: Historically superseded but instructive. Heuristic derivation from a non-rigorous premise, plausible at small alpha, wrong at large alpha. It is not "Refuted by calculation" in the paper's own regime; the refutation required the 1929 Dirac-based theory.

Claim 6: The theory holds only for loosely bound electrons in light elements; bound electrons in heavy elements need a different treatment

• a. Mathematics. Consistent. The paper's reasoning (p.502): the derivation assumes no constraining forces; recoil energy must exceed binding energy. The condition is physically right: if h nu is much greater than the binding energy, the electron acts as free.
• b. Experiment. Consistent with later observations: the Compton line broadens (Doppler broadening from the electron momentum distribution) and shifts for bound electrons, and an unshifted (coherent, Rayleigh) component appears for tightly bound electrons [memory: high]. The paper also speculates that a bound electron's quantum may "spread over a large number of electrons" and permit interference. That is qualitatively close to what is understood as coherent scattering, but described only as a guess in the paper; it is not tested here.
• c. Tests. Not tested in this paper (it is a stated limitation). Later work established the bound-electron treatment (impulse approximation, Compton profiles) [memory: high].
• d. If true. Nothing; it is the accepted domain of validity.
• Verdict: Established (as a stated limitation). Compton's speculation about spreading the quantum over many electrons: Not testable as stated in this paper (too vague).

Overall comments

• The paper's central physical content (photon momentum, the Compton formula) is correct and fundamental. Its quantitative intensity formulas were provisional, and its "tests" at gamma energies are the weakest part: indirect, tuned and heterogeneous-source. The pivotal confirmation was later: coincidence and cloud-chamber experiments in 1925 [web: https://arxiv.org/pdf/1106.1365].
• Duane's reservations and the Bohr-Kramers-Slater alternative: I did not read primary sources for the content of Duane's objection and cannot describe it with certainty [memory: low]. The web summary confirms that BKS was ruled out by the 1925 experiments.
• Web checks done (3): Bothe-Geiger/Compton-Simon and BKS outcome and Klein-Nishina success (arXiv 1106.1365 and ScienceDirect link above); Nobel 1927 (https://www.nobelprize.org/prizes/physics/1927/compton/facts/). The Ishino 1917 data quality, Compton's 1923 follow-up numbers and the Barkla-Ayers measurement quality remain [memory: low or medium] and unchecked.

Verdict summary