Trichel: Physicist review (Step 2)

Reviewer: PhD physicist role. Source folder: one 15-page conference paper, Sattari, Castle and Adamiak (SCA), "Numerical Simulation of Trichel Pulses in a Negative Corona Discharge in Air", ESA Annual Meeting on Electrostatics 2010, paper K4 (University of Western Ontario). Nothing written by G. W. Trichel is in the folder. All statements about his 1938 work rest on the SCA paper's description, on the Investigator's [web] finding of his abstract, or on [memory]. No source gap for the SCA paper itself (text layer fully readable).

Tags: [read] from the document or Step 1 files; [web] with URL; [memory] with confidence; [calc] my own arithmetic, reproducible from numbers in the documents.

Plain-language background (for the reader)

A sharp metal needle held at a few kilovolts negative, facing a flat plate 1 cm away in air, produces a faint glow (negative corona) at the tip. The current is not steady. It arrives as a train of very short spikes (Trichel pulses) about every 3 to 20 microseconds. The standard picture [memory, high]:

1. The field at the tip is very high (of order 10^7 V/m, i.e. 100 kV/cm). A free electron there gains energy between collisions, ionizes air molecules and makes an avalanche (each ionization frees one more electron).
2. The avalanche leaves behind a cloud of slow positive ions near the tip and a cloud of electrons. The positive ions are pulled into the needle in nanoseconds. Electrons drift outward into weaker field, where oxygen captures them, forming slow negative ions. This is "attachment".
3. While the negative-ion cloud sits in front of the tip, its charge lowers the field at the tip below the level needed for avalanches, so the current stops.
4. The ions drift outward (mobility mu about 2e-4 m^2/(V s), so velocity = mu x E). When the cloud is far enough away the tip field recovers and the next avalanche starts.

Worked order-of-magnitude check [calc]. For a hyperbolic needle (tip radius R = 100 um, gap d = 1 cm, voltage V = 7 kV) the vacuum (Laplace) field at the tip is about E_tip = 2V / (R ln(4d/R)) = 2 x 7000 / (1e-4 x 6.0) = 2.3e7 V/m. A negative ion starting 10 um from the tip drifts by dz/dt = mu x E(z). Integrating gives R z + z^2 = 2 mu V t / ln(4d/R); at t = 10 us this gives z = about 2.1 mm. Crossing the whole gap in a uniform field would take d^2/(mu V) = about 71 us. So a 10 to 20 us period is the time for the cloud to move a millimetre or two: the right order of magnitude for a "space-charge clearing" picture. This is my own estimate and ignores space charge, so it is only a plausibility check.

Per-claim review

Claims are numbered as in claims.md.

Claim 1: negative point-plane corona in air gives a regular train of Trichel pulses, only in electronegative gases and a limited voltage range

a. Mathematics: not a mathematical claim. Not applicable.

b. Established experiment: the pulse train itself is among the best-documented phenomena in gas discharge physics. Trichel (Phys. Rev. 54, 1078, 15 Dec 1938) reported it from oscillograms; his abstract says current consists of discrete pulses whose size and frequency depend on current, point size and pressure, with frequency "independent of gap length" and "not affected by the electrical constants of the external circuit" [web] https://journals.aps.org/pr/abstract/10.1103/PhysRev.54.1078 (via Investigator, research/2026-10-04-aps.md; I did not re-fetch). Loeb's 1939 book reproduces his oscillograms (Fig. 243) and a sketch of the glow (Fig. 247) [web] https://archive.org/stream/in.ernet.dli.2015.74804/2015.74804.Fundamental-Processes-Of-Electrical-Discharge-In-Gases_djvu.txt (via Investigator). Later quantitative measurements: Loeb, Kip and Hudson 1941, Lama and Gallo 1974 (cited in the SCA paper, refs [6], [7] [read: p.14-15]), Dordizadeh, Adamiak and Castle 2017 [web] https://www.sciencedirect.com/science/article/abs/pii/S0304388616301735 (abstract only, via Investigator).

The sub-claim "only in electronegative gases" is NOT settled. Zhang, Xia, Jiang and Ouyang, Sci. Rep. 7:10135 (2017) report, from experiments with needle cathodes of 50-250 um radius, gaps of 1-20 mm, 1 MOhm ballast, that Trichel pulses occur "in various gases rather than only in electronegative gases", that positive ions "play a predominant role" and negative ions "are not necessary" [web] https://pmc.ncbi.nlm.nih.gov/articles/PMC5579005/ (abstract as relayed by the Investigator; my own re-fetch of that page returned a reCAPTCHA wall and the nature.com copy a login redirect, so the check failed and I cannot say which gases were tested). This is a peer-reviewed experimental result from one group, reported as their finding. It conflicts with the SCA introduction's statement. In my reading it does not contradict the existence of pulses in air, which is the part SCA actually uses.

Mixed corner worth noting: Trichel's "independent of gap length" versus Lama and Gallo's empirical law f proportional to V(V - V0)/(r d^2), quoted in Sattari's thesis, Ch. 8.2 [web] research/files/Sattari-2011-PhD-thesis-UWO.pdf, p.126-127 (I read the text). The two statements differ about gap dependence; Trichel's work was near onset and short on parameter range [memory, medium], so they may not be the same regime. Open; not resolved by any document here.

c. Tested: yes, repeatedly, by independent labs from 1938 on. Replicated: yes (Trichel 1938, Loeb-Kip-Hudson 1941, Lama and Gallo 1974, Soria et al. 2007 in oxygen, Zhang 2017, Dordizadeh 2017).

d. If untrue: nothing; it is established. If the "electronegative only" form were wrong (Zhang 2017), the standard negative-ion-clearing picture would need qualifying, not mainstream physics at large.

Verdict: Established (pulses in negative needle-plane air corona). The sub-claim "only in electronegative gases" is split off: Never properly tested as a universal statement; contested by one 2017 experimental paper, and I could not verify its gas list.

Claim 2: pulsing arises from space-charge dynamics; positive ions absorbed in nanoseconds switch ionization off; slow negative ions clearing sets the period

a. Mathematics: the mechanism is a qualitative picture, but the SCA model equations (below, claim 3) are the mathematical form. See claim 3.

b. Experiment: the mechanism dates from Trichel 1938 and Loeb [web] same sources as above, Trichel's abstract: "space-charge formation and clearing". It is the textbook explanation [memory, high]. Supporting experiments for ion clearing: dependence of frequency on voltage and gap, consistent with a drift time scaling roughly as d^2/(mu V). My order-of-magnitude check above is consistent. But challenges exist: - Zhang et al. 2017 (above) argue the positive-ion cloud rebuilding sets the interval and negative ions are not necessary. - Trichel's own abstract says pulses are unaffected by external circuit constants. In Sattari's thesis simulations the period depends on R_ext (100 kOhm) and on ion mobility [web] Sattari 2011 thesis, Ch. 7 conclusions, p.124 (via Investigator, research/2026-10-04-uwo-scholaris.md; thesis PDF in research/files). The two statements sit uneasily; the thesis does not resolve it (the value R_ext = 100 kOhm and gap capacitance 0.055 pF give an R C time constant of 5.5 ns, far below the microsecond period, so how R_ext lengthens or shortens the period in the model is not obvious to me [calc from thesis values]). - C. F. Gallo's unpublished microscope observations, relayed in Sattari's thesis Ch. 8.3: an "extremely narrow bright ionization channel" (0.12 to 2 um diameter) at the negative tip, not seen in positive corona; wobble of the channel on very sharp tips, offered as the explanation of the "mode shifts" Loeb and Trichel described; onset at -2.3 kV roughly independent of tip radius, against -4.7 kV calculated. The authors say the channel "is not described by present theory or simulations and may possibly be confined magnetically by the high current density" [web] thesis, Ch. 8.3. I checked the thesis text for the -2.3 kV statements (p.126, 8072). This is a single-source observation (Gallo's own, relayed by collaborators, no independent publication found). It deserves stating plainly: if the active region is a channel of order 1 um, then a 2D model with a 5 um mesh around a 100 um tip is not resolving the physics at the tip. Magnetic confinement: the claim is untested. A rough check [calc]: a 1 um channel carrying 2 mA has a current density of 2.5e9 A/m^2 and B at the edge = mu0 I/(2 pi r) = 4e-7 x 2e-3/(2 x 0.5e-6 ... ) = about 0.8 T with I = 2 mA over r = 0.5 um. Such a field is large, but the electrons' mean free path at 1 atm is about 0.5 um and the ions dominate current, so whether this "pinch" matters is not determined; I flag it as unquantified, not as supported.

c. Tested: the mechanism has been tested indirectly by many parametric studies and by numerical modelling (Morrow 1985, Napartovich 1997, Soria 2007 in oxygen, SCA 2010) [read: p.14-15, refs]. A direct neutral test of "negative ions clear -> period" (for instance changing ion mobility by changing gas composition or measuring ion cloud position) is not in the documents. Replicated in modelling by several groups; the modelling groups share the same basic assumptions, so this is agreement among models, not an independent confirmation.

d. If wrong: the picture of negative corona as ion drift-limited would need replacing with a cathode-sheath/Townsend-to-glow transition (Zhang 2017). Mainstream discharge theory would change in detail, not in kind.

Verdict: Plausible, untested as the sole mechanism (strong form: negative-ion clearing is the controlling process); the weak form (space charge from ions suppresses the tip field, giving a pulse) is Established as a standard accepted explanation with competing details unresolved.

Claim 3: a 2D FEM-FCT three-species model reproduces a regular pulse train, a post-pulse, and a large delayed first pulse; first successful 2D FEM-FCT simulation at several voltages

a. Mathematics, internally consistent? Not as printed, on three points I verified from the deep read and extended.

1. Source terms in Eqs. (6)-(7). Printed: dNp/dt + div(Np Wp) = ki Np - beta Ne Np and dNn/dt + div(Nn Wn) = ka Nn - kd Nn. Ionization should create positive ions in proportion to the electron density (ki Ne) and attachment should create negative ions from electrons (ka Ne), as in Eq. (5), where electrons are lost at rate ka Ne [read: page image p.4, via deep read; I agree with the physics]. As printed, with kd = 0 the negative ions would grow exponentially from themselves and never take charge from electrons. Charge is not conserved between Eqs. (5)-(7). Almost certainly a typo in the paper; the code may be right. Status: printed equations inconsistent; actual code unknown.
2. Ionization law. Printed ki = 10^(11.93 - 4.493/EN), EN in Td. [calc] with N = 2.46e25 /m^3: ki = 6.6e11 /s at 1e6 V/m and 8.4e11 /s at 2e7 V/m, i.e. nearly constant and >> ka = 6e6 /s at every field above 1 Td. There would be no ionization threshold and no onset voltage, contradicting the paper's own onset of 4.7 kV. Sattari's thesis prints ki = 10^(11.93 - 449.3/EN) [web] thesis Ch. 6, p.91 (via Investigator; PDF in research/files). I recomputed [calc]: ki = 7.5 /s at 1e6 V/m (41 Td), 2.5e6 /s at 2e6 V/m (81 Td), 5.2e9 /s at 5e6 V/m, 6.7e10 /s at 1e7 V/m (407 Td), 2.4e11 /s at 2e7 V/m. Threshold ki = ka occurs at about 87 Td, i.e. 2.1e6 V/m. The corresponding Townsend coefficient alpha = ki/(mu_e E) = 1.3e5 /m at 1e7 V/m (using mu_e = 0.05), plausible for air at those fields [memory, medium]. So the exponent 4.493 in the ESA paper is a misprint; the deep read's inference is confirmed by the authors' own later text. I confirm this correction.
3. Units. If N in Eq. (1) is a number density, e is missing; if charge density (C/m^3, as in the figures), the recombination term beta Ne Np lacks a 1/e. As in the deep read. Minor typesetting-level inconsistency; not tested what the code does. Also: "N_O2 = 2.46e25 /m^3" [calc] equals the total air number density at 298 K and 1 atm (p/kT = 2.46e25), not the O2 density (about 5e24). The paper calls the gas "oxygen molecules" and takes rates from oxygen-based sources, so it is an oxygen-ion model applied to air; the thesis repeats this. This probably does not matter if rates were fitted in terms of E/N of air, but it is undocumented. Electron diffusion D = mu kB T/e at gas temperature (1.28e-3 m^2/s at 298 K [calc]) understates diffusion at E/N of hundreds of Td, where electron energies are of order eV [memory, medium].

b. Consistent with experiment? Partly. - Qualitative: pulse train, regular after the first, current spike with a post-shoulder, positive-ion deposition within ns, field at the tip peaking at 2.2-2.45e7 V/m at the pulse and recovering to 1.1e7 V/m between pulses [read: p.11, via deep read]. The tip field value is consistent with my Laplace estimate (2.3e7 V/m at 7 kV) and below that is physically sensible after space-charge screening. - Quantitative: periods 11.33, 6.31, 3.39, 2.46 us at -7, -8, -9, -11 kV against stated "experimental" 21.85, 13.32, 9.09, 5.07 us [read: p.13, Table 1]. Ratios 0.52, 0.47, 0.37, 0.49 [calc]. The authors say "approximately 50%" low and give "approximations in the numerical technique" as the explanation without testing any [read: p.14]. - The paper gives no simulated rise time or width, nor any comparison with measured amplitude or charge per pulse. The literature values it cites (rise 1.3 ns, width tens of ns) are not compared [read: p.1]. Sattari's thesis later gives simulated pulse durations of 0.2-0.3 us (Table 8.3) [web] thesis, via Investigator, i.e. an order of magnitude wider than the "tens of ns" literature value. Dordizadeh et al. 2017 found measured rise/fall/width independent of radius, voltage and gap, with "some discrepancies" against simulation [web] https://www.sciencedirect.com/science/article/abs/pii/S0304388616301735 (abstract). So a standing mismatch on pulse shape is documented, partly by the same group. - Unverifiable inputs: R_ext is not stated in the ESA paper (thesis uses 100 kOhm); the experimental geometry is not stated; the key rate constants come from an unpublished 1997 contractor report for ABB (an interested commercial party) [read: p.4, p.15].

New finding on the experimental comparison [calc]; this corrects the framing in Step 1 and extends the Investigator's lead. Lama and Gallo's empirical law, as quoted in the thesis, is f = K1 V (V - V0)/(r d^2) with K1 about 27 kHz mm^3/kV^2 and V0 = -2.3 kV. With r = 0.1 mm and d = 10 mm: T = 11.26, 8.12, 6.14, 4.81, 3.87 us at 7, 8, 9, 10, 11 kV. The first four match the thesis "Gallo's T" column (11.3, 8.1, 6.1, 4.8 us) to rounding. Now take the ESA Table 1 "experimental" values. Fitting y = 1/(T V) against V gives a straight line with residuals under 0.01 kHz/kV at all four points, slope 2.85 kHz/kV^2 (compare K1/(r d^2) = 2.7) and zero crossing V0 = 4.71 kV. That is, the ESA "experimental" curve is, to within the printed precision, the Lama-Gallo functional form with V0 = 4.7 kV, which is exactly the simulation's own computed onset voltage (4.7 kV), not Lama and Gallo's measured -2.3 kV. The simulated points themselves do not fit that form well (residuals up to 5.7 kHz/kV; 9 kV is the outlier). Interpretation, stated as inference: the ESA paper's experimental column looks like a fitted/analytic curve evaluated with V0 replaced by the computed onset (Fig. 17 shows the experimental data as a curve, not points [read: page image p.13]), and the thesis table uses the same law with Lama and Gallo's own V0. This alone changes the ratio by 1.3 to 1.9 and flips the sign of the model error (model underestimates in ESA 2010, overestimates in the thesis [web] thesis Table 7.1 via Investigator). I cannot determine from the documents which V0 is right for the 1 cm / 100 um case; the original paper (J. Appl. Phys. 45, 103) was not accessed (AIP abstract not reached by the Investigator; my search returned only citations). Consequence: the claimed "50% shortfall" is not a stable result of the model; it depends on what is plotted as "experiment".

The thesis itself is inconsistent about the direction (Ch. 8.2 says "overestimate" and then "underestimate") [web] thesis, via Investigator; I saw the "considerably overestimate" sentence myself (p.126).

Large first pulse (25-37 mA vs about 2 mA, 12 to 20 times the steady value) [read: p.7, p.12]. The authors explain it by charge-free initial space. No experimental evidence is cited that real first pulses are this large. A start-up transient in a simulation starting from "small initial space charge density" (magnitude unstated) is a numerical alternative they do not discuss [read: p.5]. At 7 kV the first interval (11.5 us) is actually shorter than the second (13.38 us), contrary to the text [read: p.7 via deep read]. The post-pulse is explained as a secondary avalanche from positive ions near the tip; the paper itself says it "can be a source of instability in the calculations" and that the time step "needs to be chosen very carefully" [read: p.8]; no experimental observation of it is claimed, and no mesh or time-step convergence study is reported [via deep read]. These are numerical features of unknown reality.

Novelty claim ("first time FEM-FCT has been successfully applied to 2D Trichel pulses for different voltages") [read: p.14]: a statement of priority by the model's own developers; I cannot confirm or refute it. It is a statement about the literature, not physics.

c. Tested: compared with one experimental data set, indirectly, at 4 voltages, one run each, no uncertainties. Replicated: by the same group only (thesis 2011; 2017 and 2023 follow-ups) [web] https://iopscience.iop.org/article/10.1088/1361-6463/acdd0f/meta (2023 abstract: three rate-coefficient formulations compared, "validated by comparing them with the experimental data", via Investigator; I did not read the full text). No independent group replicated this FEM-FCT code.

d. If true: nothing changes in mainstream physics. It would be an application of drift-diffusion-Poisson, which is standard.

Verdict: Never properly tested as stated. Split: (i) "A drift-diffusion-Poisson model of this kind can produce a pulse train" is Plausible, untested in the strong sense, because other groups (Morrow 1985; Napartovich 1997; Soria 2007) found the same qualitatively [read: p.14-15 refs]; (ii) "reproduces the measured pulses" is Tested, not supported quantitatively: periods off by 2x (or the other way depending on the reference), no shape, amplitude or charge comparison, printed equations inconsistent; (iii) the large first pulse and post-pulse are Never properly tested (no experimental comparison). Internal consistency of the printed equations: Internally inconsistent as printed (typos), not necessarily in the code.

Claim 4: pulse frequency rises with applied negative voltage; simulated period falls from 11.33 us at -7 kV to 2.46 us at -11 kV, following the trend of Lama and Gallo

a. Mathematics: the table arithmetic is consistent: the "average" excludes the first interval (mean of 13.38, 10.67, 10.85, 10.4 = 11.33; mean of 3.53, 3.30, 3.35 = 3.39 [calc, matches Table 1]). At 9 kV the three intervals used are few, and no scatter is given.

b. Experiment: the trend (frequency rises roughly as V(V - V0)) is a standard measured result, see Lama and Gallo's law above and Trichel 1938. The simulated 1/T over V is not linear in V as the empirical form would require [calc]: 1/(T V) = 12.6, 19.8, 32.8, 37.0 kHz/kV at 7, 8, 9, 11 kV; the 9 kV point bends upward, so "follows a similar trend" holds only loosely. With Dordizadeh 2017, measured onset -4 to -6 kV (19 to 55 um tips) differs from Gallo's -2.3 kV [web] the ScienceDirect abstract via Investigator. So the experimental record itself has two different thresholds.

c. Tested: yes, experimentally many times (trend). The specific quantitative simulation prediction has been compared once, with a reference set that changes between the paper and the thesis (see claim 3). Replicated: trend yes; quantitative agreement no.

d. If untrue: nothing; the trend is established.

Verdict: Established (trend: frequency increases with voltage above onset). The quantitative simulated-vs-measured comparison: Not testable as stated in this paper (experimental geometry, resistor and the origin of the experimental values are not given, and the reference values appear to depend on a chosen V0).

Claim 5: the simulated periods are about half the measured ones, attributed (untested) to "approximations within the numerical technique"

a. Mathematics: ratios 0.519, 0.474, 0.373, 0.485 [calc], so "approximately 50%" holds at three voltages but not at 9 kV (37%).

b. Experiment: see claim 3. The size and even the sign of the discrepancy depend on which experimental reference is used. The candidate causes in the deep read are all hypotheses, none tested in the document: constant ion mobility 2e-4 (the thesis uses 2.24e-4 and 2.16e-4), the incorrectly printed ki (if the code used the misprint the avalanches would restart too easily and shorten the period; unknown), constant gamma = 0.01, attachment fixed at 6e6 /s, an oxygen-only model for air, unstated R_ext, 5 um mesh vs a possible 1 um active channel. Gallo's channel observation adds an uncommon explanation: with an effective tip radius of 10 um instead of 100 um, the thesis model matched Gallo's period to 15% and current to 23% [web] thesis Ch. 8.4, via Investigator. A fitted radius is not a prediction; this is a tuned parameter, so the agreement is consistent with, not proof of, the channel picture.

c. Tested: the explanation "approximations in the numerical technique" has not been tested in this paper. The thesis tests some sensitivities in simulation (period depends on R_ext and mobility; secondary emission has little effect), which is partial. No independent replication.

d. If the discrepancy were physical: it would indicate a missing mechanism at the tip (channel, field emission) that current continuum models do not include; mainstream discharge physics would need a sub-micron cathode-layer model, not new laws.

Verdict: Never properly tested (the stated cause); the discrepancy itself is Not testable as stated, because its size depends on an unstated or changing experimental reference.

Investigator findings: what I kept and how I weighed them

• Trichel identity and career (Gervais W. Trichel, 1898-1990, Berkeley Ph.D. 1938, then Army Ordnance, rockets, Chrysler) [web] https://archive.org (Redmond and Smith, Project Whirlwind; Loeb 1939 preface), https://prabook.com/web/gervais_william.trichel/90642. Physics reading: this explains why the corona papers stop after 1939 without suggesting suppression; I did not find any evidence of dismissal of his work. Source count: Redmond and Smith, Loeb preface, Prabook; Prabook is a compiled database. I did not re-open these pages.
• Citation error: SCA ref. [2] Phys. Rev. 55, 382 (1939) is Trichel's POSITIVE-corona paper, while the negative paper is Phys. Rev. 54, 1078 (1938) [web] https://link.aps.org/doi/10.1103/PhysRev.55.382 (via Investigator). Kept; the thesis repeats the error. Also noted by the deep read.
• Sattari thesis details (ki corrected, R_ext 100 kOhm, Cd 0.055 pF, mobilities, reference set, pulse duration 0.2-0.3 us, pressure dependence, charge per pulse about 115 pC at r = 100 um and 22 pC at r = 10 um): kept. Check [calc]: a 1.8 mA triangular pulse 0.3 us wide carries about 0.27 nC; the thesis's 115 pC is within a factor of 2.5, and 115 pC x 88 kHz = 10 uA vs 13.5-33 uA reported averages: same order. Sound.
• Gallo's narrow channel and -2.3 kV onset: kept in claim 2 and 5. Single source, relayed second-hand by collaborators (an interested party for the thesis's improved fit); no independent publication found. It is evidence that deserves an experimental test; a microscope at a negative tip is a feasible experiment.
• Zhang 2017 (pulses in various gases): kept as a contest to claims 1-2. Gas list unverified (my web check failed).
• Related work noted but unread: Kokovin 2021, Ferreira 2025, Friebe 2021, Eindhoven 1987 thesis.
• Not reached by the Investigator and not resolved here: the Lama-Gallo abstract (AIP), Zentner 1970 rise time, Napartovich 1997, Trichel's 1938 dissertation (WorldCat blocked), IEEE TIA 2011 journal version of the ESA paper.

Corrections to Step 1

1. claims.md caveat "printed ionization formula ... appear inconsistent": confirmed and resolved. The exponent is 449.3, not 4.493 (thesis, Investigator). With the corrected formula the threshold is about 87 Td (2.1e6 V/m) [calc].
2. facts/claims treat Table 1's experimental column as Lama and Gallo's measurements. My fit [calc] shows those four values follow f = 2.85 V (V - 4.7) kHz to within the printed precision, while Lama and Gallo's own V0 (-2.3 kV) yields about 11.3, 8.1, 6.1, 4.8 us (the thesis values). The factor of two is therefore tied to the V0 used, not necessarily to the model. Inference, not confirmed against the original paper.
3. Deep read claims the first interval is "usually longer"; at 7 kV the first interval (11.5 us) is shorter than the second (13.38 us). Deep read already noted this; claims.md does not.
4. Deep read (section 4) states the electron mobility 0.05 m^2/Vs gives a drift above typical measurement (about 2e5 m/s); I agree but rate it [memory, medium] and did not verify on the web. The conclusion above does not depend on it.
5. Deep read says "the 15 numbered references" but lists 19; harmless.
6. The deep read describes the lack of rise time and width as "cannot be confirmed from the paper". Add: the thesis gives 0.2-0.3 us durations, which are inconsistent with the "tens of ns" literature value the paper itself quotes [web] thesis Table 8.3 via Investigator.
7. Step 1 says no source gap: agreed for the SCA paper, but Trichel's own writing is absent; statements on his 1938 findings are [web] abstract only.

Web checks done (budget about 6)

1. https://pmc.ncbi.nlm.nih.gov/articles/PMC5579005/ : reCAPTCHA wall, failed.
2. WebSearch for the Lama-Gallo abstract: only citations found. The search summary attributes to the Lama-Gallo paper the same wording as Trichel's abstract; that is a search tool's paraphrase, not evidence, and I do not rely on it.
3. https://www.nature.com/articles/s41598-017-10118-2 : login redirect, failed. Facts that depend on these remain tagged to the Investigator's [web] notes.

Verdict summary