Mathematician review: Popa-Simil, "Nano-Hetero Structure for direct energy conversion" (NSTI-Nanotech 2007)

Reviewer role: PhD mathematician (Step 2). Checks derivations, notation, units and hidden assumptions. Source tags: [read] from the document, [web] with URL, [memory] with confidence, [calc] my own arithmetic (Python, shown). Sources read by me: the paper (p.1-4, with p.2 rendered at 150 and 300 dpi, Eq.1 and Figs. 1-2 looked at directly), deep/ note, claims.md, facts.json, sites.md, the patent US20100061503A1 (paras 0115-0121, from the PDF in research/files/), research notes on Klassen (InTech 2013) and the MDPI 2026 review. I did not read the primary Anufrienko/Chernov or Yershov papers, nor the other patents in full.

No source gap: one 4-page paper with a full text layer. Everything below about the paper is [read]. The ~19-year record of later restatements by the same author (patents, 2011-2021 papers) was found by the Investigator; I read only the patent passages cited.

Before the claims: what the mathematics of this paper actually is

The paper has exactly one equation (Eq.1, a stopping-power formula) and one simulation (SRIM-2003 run, Figs. 1-2). It contains no derivation of the efficiency (>80%), of the module voltage, or of the 5000 kW/cm3 figure. Those are asserted [read: p.1, p.3, p.4]. So for claims 1-6, question (a) "is the mathematics internally consistent?" is largely "is there any mathematics, and do the asserted numbers follow from the paper's own inputs?" The only quantitative treatments I can check are Eq.1, the figures, the "10 keV/nm / 1200 electrons per nm3" paragraph, and (in the patent that the Investigator located) the energy/charge bookkeeping in paras 0116-0121.

Corrections to Step 1

I verified the deep read of p.2 against the page image. Most of it is right; four points need correcting or tempering.

1. Missing c in the log is probably a typesetting slip, not necessarily a conceptual error. I confirmed the printed log argument is "2mc β²/(1-β²)" with c to the first power [read: page image p.2, 300 dpi crop]. The prefactor does read "m c²". The standard argument is 2 m c² β²γ²/I [memory, high]. As printed it is dimensionally wrong, as Step 1 says. But the Word equation editor easily drops a superscript, and the same missing power affects nothing else. I would call it a probable typo that the reader should still see, not evidence the author misunderstands Bethe-Bloch. The Barkas sign and Bloch charge are more substantive.
2. Barkas sign and Bloch charge: confirmed on the image ("−Z1 L1(β) + Z2² L2(β)"). The text then calls the last term "Z2L2 Bloch effect" (no square), so text and equation disagree with each other as well as with the standard form (+Z1 L1 + Z1² L2) [memory, high; the form used by Ziegler 1999, the paper's own ref. 10]. Step 1 is right.
3. Prefactor: Step 1 is right (Z2 Z1², the target atom number times projectile charge squared). I mention this only because the crop makes the subscripts hard to see.
4. "Eq.1 overestimates SRIM by about 10x" needs a qualifier. I recomputed the U case: dE/dx = 0.307 (Z/A) z² L / β², with Z/A = 0.386, L = 0.62, β² = 0.00161, density 19.05 g/cm3, z = 55: about 263 keV/nm [calc], matching the deep read's 261. The 10x is real, but it measures "bare charge 55 instead of effective charge ~17". Since the paper never says SRIM used Eq.1 (SRIM uses Ziegler's scaled-proton method), the mismatch is not an error in Fig. 2; it shows Eq.1 is decorative. The effective charge that reproduces SRIM's ~26 keV/nm is z* ≈ 55 × sqrt(26/263) ≈ 17 [calc], in the middle of the deep read's 16-21 range. Good.
5. Step 1's claim "the Cs ion loses only ~9% of its energy" is plausible; I re-derived it roughly from the Fig. 2 plateaus: weighted mean ≈ 14 keV/nm over 600 nm gives ≈ 8-9 MeV [calc, rough]. Fine. A related point Step 1 did not draw out: since the ion exits with ~91 MeV, Fig. 2 is a rate (energy per length), not a "stopping" simulation, and no per-layer integrated deposit, electron yield, or charge-balance is computed anywhere in the paper.
6. Step 1 did not examine the claims' bookkeeping (efficiency vs voltage, power density, the "cold" claim). Those are my own analyses in the claims below.

Claim 1: CIci structure converts fission-product kinetic energy to electricity at >80% efficiency

a. Internal consistency. The paper gives no calculation of efficiency [read: p.1 abstract: "may be higher than 80%"]. The Investigator found the number's origin in the patent US20100061503A1: para 0119 sets the Au "C" layer to stop 5 absorption units, the SiO2 insulators 0.5 each, the c layer (Al or LiH) 1 unit, "about 7 ionization energy absorption units in total"; para 0121 then says "The calculated efficiency of 84% is simply the efficiency of ionization energy stopped in the 'C' layer. To be a real accurate evaluation it has to consider all the electrons and excitations inside the material. That is hard to obtain by theoretic calculations" [read: US20100061503A1 paras 0119-0121]. So the author's own wording says the 84% is a fraction of stopping energy in one layer, an upper estimator, not an electrical efficiency. Two further checks: - The arithmetic does not reproduce: 5/(5+0.5+1+0.5) = 5/7 = 71%, not 84%. 5/(5+1) = 83.3% only if the insulators are ignored [calc]. The stated 84% matches neither exactly. - The decisive check is charge-and-voltage bookkeeping. Electrical energy out per collected electron is at most q·V, where V is the voltage at which the cell delivers current. Energy spent per electron-hole pair in ionization is W ≈ 25-30 eV [memory, medium: typical for solids; I did not web-verify]. A 100 MeV fragment therefore produces at most ≈ 3×10^6 free electrons [calc: 100 MeV / 30 eV]. If all are collected at V = 10 mV (the paper's own maximum, p.3, "less than 10 mV"), the electrical energy is 3.3×10^6 × 0.01 eV = 33 keV, a fraction 3×10^-4 of 100 MeV [calc]. To reach 80% by this route V would have to be about 0.8 × 30 V ≈ 24 V per collected electron [calc]; the paper's own breakdown limit is mV. Series stacking adds voltage across cells but does not change the energy per electron, because each electron is generated and collected within one cell. A converter cascade (claim 4) conserves power at best; it cannot create energy per electron. - The patent's own power example (para 0116): 167 MeV per fission = 26.7 pW per fission/s, "harvested at 10 mV and 2.67 nA" [read]. I verified the arithmetic: 167 MeV × 1.602×10^-19 J/eV = 2.675×10^-11 W; divided by 10 mV gives 2.675 nA [calc]. But 2.675 nA is 1.67×10^10 electrons per second per fission/s, i.e. 1.7×10^10 electrons per fission, while ionization can supply about 5.6×10^6 (167 MeV/30 eV) [calc]. The example assumes about 3000 times more charge than the energy can liberate. At 5.6×10^6 electrons and 10 mV the power is 9 fW, i.e. 0.033% of 26.7 pW [calc]. This is an internal inconsistency between "100% efficiency at 10 mV" and the energy cost of ionization. - Caveat on the objection: electron multiplication is not limited to W if the electrons gain energy from an applied field (it is a field-assisted device), but nothing in the paper describes that, and the energy then comes from the field source. The charge of the stopped fission ion itself (~+20 e) is negligible against 10^6 electrons.

b. Consistency with experiment. The only measured efficiencies in the Investigator's findings are far below 80%: Yoshida/Tanabe laminated device ~10^-5 % (reported by Klassen et al., a second-hand statement in one published source) [read: research/2026-10-04-issp-intech-klassen.md, citing p.271]; planar U-235 layered structures 25% (Anufrienko et al. 2006, 2008, as reported by Klassen et al. p.272-273; I have not read the primary papers and do not know what "efficiency" is defined as there, or whether it is electrical power out / fission power in). The 25% is documented in one published source (a 2013 book chapter by a group that has its own related experiments); no second source found yet. It deserves weight, and a deep read of the primary papers is the single most useful follow-up. The older fission-electric-cell lineage (Safonov 1957, Krieve 1966, Ursu 1978-88) uses vacuum gaps and kV-scale voltages, which is how it avoids the per-electron-energy limit above; Sandia's ~60% is a theoretical estimate, per the 2026 MDPI review [read: research/2026-10-04-mdpi.md]. The MDPI review's own 0.85 We/Wth is an ideal ceiling equal to the fragment kinetic energy share (≈168/200 [memory, high]), not a measurement.

c. Tested? No measurement of a CIci structure by the authors is in the paper (SRIM only). The Investigator found no later published measurement of a built CIci structure through 2021; the patent describes a beam-test set-up (FIG. 15) but no results were found. Related independent tests (Yershov/Klassen, Anufrienko/Chernov) are of similar, not identical, structures. Replication of the specific 80%+: none.

d. If true. Not a violation of energy conservation (the upper bound is 168/200 ≈ 84% of total fission energy, and the 84% quoted is suspiciously close to that number; that is my observation, not a documented derivation). It would require charge collection at volt-scale potentials across nanometre insulators, i.e. field strengths of 24 V / few nm ≈ 10^9 V/m, near or above dielectric breakdown (SiO2 ~ 10^9 V/m [memory, medium]). Nothing in mainstream solid-state or radiation physics would need to change; the engineering assumptions would.

Verdict: Internally inconsistent for the strong form (>80% electrical efficiency at mV potentials, by the paper's and patent's own numbers). Never properly tested for the weak form (some useful fraction, order 1-25%, in thin-insulator layered stacks): Russian reports exist, I have not read them, and nothing exists for the specific stack. Split verdict, with the second part open.

Claim 2: mechanism is a difference in electron-shower intensity between adjacent conductors, electrons cross a nm insulator and charge the low-yield collector

a. Consistency. Qualitatively sensible: secondary-electron yield per unit path depends on material (SRIM Fig. 2 shows ionization rates differing by 2-4x between layers, e.g. U ~26 keV/nm vs. S/Mg/CF2 ~7 keV/nm [read: page image p.2; deep read]). But the sign convention is never worked out: net charge flows only if a net electron current leaves the high-yield layer and arrives in the low-yield one, and in a stack with ionization in every layer, electrons born in the collector also leave it. Net current is the difference of cross-layer fluxes, which the paper never computes. "Ballistic" transport across the insulator: the paper says free path is "several nm" and the insulator thickness is comparable to it [read: p.2-3]; these are quoted without an inelastic mean free path calculation (typical IMFPs for 1-1000 eV electrons are 0.5-10 nm [memory, medium]), and high-energy delta electrons (up to 1.65 keV from this ion, β² = 0.0016; deep read, I re-derived T_max = 2mc²β²γ² = 2 × 511 keV × 0.0016 = 1.65 keV [calc]) travel much further and go through the insulator regardless. Also, the insulator itself is ionized (SiO and CF2 layers in the stack stop 7 keV/nm), generating electron-hole pairs that will conduct under a field or leak charge. The model treats the insulator as passive. Hidden assumption: the "different shower intensity" means different electron emission from the layers' surfaces into the next layer, not different ionization in the bulk. Fig. 2 shows bulk ionization. Emission into the gap depends on the exit surface and the work function; the patent adds that this needs "high mass density and low extraction work function" (para 0115). Nothing quantitative.

b. Experiment. Klassen et al.: thinning the insulator from 200 µm to 1 µm increased conversion efficiency by 3 orders (second-hand, one source, quantitative papers not read) [read: p.272 of InTech chapter, per research note]. This supports a thin-insulator trend. Fission-electric cells show the collection principle works with large gaps. Anufrienko et al. state "converters of nuclear energy to electricity by secondary electrons" (title of their 2006 paper) [read, as a title in a reference list]. Alpha/beta-voltaic secondary-emission batteries are established at low power [memory, high]. These support the mechanism in outline.

c. Tested? By others in related structures, yes (Russian groups; Nagoya). By the authors, no. Replicated: the trend (thin insulator helps) appears in two independent groups' reports as relayed by one chapter; no replication of the authors' stack.

d. If true. Nothing in mainstream physics needs to change; this is applied secondary-electron emission physics.

Verdict: Plausible, untested as a mechanism in outline, with partial support from related experiments by others. The specific quantitative picture (ballistic crossing, net current) is Not testable as stated until a net-current model is written down.

Claim 3: a reactor harvesting fission energy this way would "run cold"

a. Consistency. Energy budget: of ≈200 MeV per fission, about 168 MeV is fragment kinetic energy, ~5 MeV prompt neutrons, ~7 MeV prompt gammas, ~8 MeV beta, ~7 MeV gamma from decay, and ~12 MeV antineutrinos that escape [memory, high; typical textbook values]. Even a perfect fragment converter leaves roughly 30 MeV (15%) that appears as heat in structure and coolant, plus decay heat after shutdown, which does not come from the fragments in the structure at all. At the paper's own illustrative stack (9% of energy lost by one ion in 600 nm) and 80% conversion of only the portion deposited in the structure, the heat fraction is higher still. "Cold" is relative; the sentence "less energy is left as heat" is true if conversion works at all, but "cold" is not quantified anywhere in the paper [read: p.1]. This is arithmetic, no inconsistency in the paper itself.

b/c. No experiment on heat reduction exists. The interplay with claim 6 matters (below): at 5000 kW/cm3 even a 15% heat fraction is 750 W/cm3, above conventional light-water core power density (~100 W/cm3 [memory, medium]).

d. Nothing would change.

Verdict: Plausible, untested (if claim 1's weak form holds, heat is reduced by the converted fraction; "cold" as written is not testable).

Claim 4: mV-level modules stacked and boosted to grid voltage by DC/AC and AC/AC cascade, possibly with Josephson and Andreev effects

a. Consistency. Voltage boosting is possible in principle, but not free. Power through a converter is conserved at best: boosting 10 mV to 10 kV needs a ratio of 10^6, with current dropping by 10^6. The paper does not give a loss model. Three problems: - Switching devices at 10 mV: a silicon diode or MOSFET gate drive needs hundreds of mV [memory, high]; a converter whose threshold is 50x the input voltage is mostly dissipation unless it is superconducting. The paper gestures at that ("possibly Josephson and Andreev effects") [read: p.3]. - Josephson effect: a junction biased at V oscillates at f = 2eV/h = 483.6 MHz/µV; at 10 mV that is about 4.8 THz [calc]. That is a way to convert DC to AC, but at THz, and extracting that as power at 50 Hz is not described. Andreev reflection is a charge-transfer process at a normal-superconductor boundary; it does not by itself raise voltage. Neither is quantified. - Series stacking to raise voltage (also mentioned) adds ~10 mV per cell: 10^6 cells in series for 10 kV, each with insulator breakdown margin. Possible geometrically in nano-layers (hence the idea) but then the leakage current through the insulators in the 10 kV chain, and the radiation-induced conductivity of every insulator, set the loss. Not discussed.

b. No data. c. Never tested. The patent (para 0116) gives a thermal-protection switch and a "protection resistor" but no converter performance.

d. Nothing would change; it is a circuit-engineering question.

Verdict: Never properly tested. Not refuted, but the paper omits the loss analysis that decides it. The Josephson/Andreev remark is Not testable as stated.

Claim 5: superconducting actinide compounds (PuCoGa5, Tc ~18.5 K; PuCoRh5, Tc ~8.7 K) can serve as harvesting structure, with damage balanced by self-repair over the reactor lifetime

a. Consistency. The Tc values are the discovery values (PuCoGa5 18.5 K, Sarrao et al. 2002 [memory, high]; the paper's 8.7 K for PuCoRh5 I could not confirm and treat as [read] only). The claimed "dynamic equilibrium of damage and repair" is stated in words with no rate equation. A simple one (generation rate G of displaced atoms vs. annealing rate proportional to defect density) would give a steady state n = G/k; the paper gives neither G nor k. In PuCoGa5 itself the Pu-239 alpha decay (5 MeV alpha plus 86 keV U-235 recoil) produces damage; the paper's own Frenkel-defect discussion starts there [read: p.3-4]. Because the structure is also bombarded by fission fragments at the claimed power density, the displacement rate is many orders of magnitude above self-irradiation. A cryogenic operating temperature (~18 K or below) strongly suppresses thermal annealing, so "dynamic repair" is least available exactly where superconductivity is required. This is my reasoning from standard defect physics [memory, medium], not a calculation.

b. Experiment. [web] Published work on PuCoGa5 reports that self-irradiation reduces Tc roughly linearly with sample age (Jutier et al.), by about 0.2 K per month initially, slowing as defects accumulate; damage is much more severe than predicted by standard radiation damage theory; defects act as pinning centres (https://www.osti.gov/biblio/917542 ; https://www.osti.gov/etdeweb/biblio/20760943 ; https://www.sciencedirect.com/science/article/abs/pii/S0304885306013345 ; summary from the search results, I did not read the full papers). That is direct evidence that an alpha-decay dose alone degrades Tc on a time scale of months, against a claimed self-repair over reactor life. It is also true that the rate slows as defects accumulate, which means saturation of some kind, which is consistent with the author's "dynamic equilibrium" idea in form; whether the saturated Tc is above the operating temperature is not established in the sources I saw. The benefit of Pu-based superconductors to the device idea (a superconducting collector with no resistive loss) is separate from the question.

c. Never tested at a fission-fragment dose; PuCoGa5 aging studies exist. d. Nothing changes. Note: using plutonium as structural conductor with 239Pu cross-sections (Table 1 in the paper) means the layer is itself a heat source and a neutron absorber; the paper lists absorption cross-sections but does not integrate them into the reactor.

Verdict: Never properly tested, with the self-repair part Tested, not supported only in the narrow sense that alpha self-irradiation measurements on PuCoGa5 show Tc declining with age (sources above); no test under fission-fragment exposure.

Claim 6: power density raised from ~0.2-0.5 kW/cm3 to a theoretical 5000 kW/cm3

a. Consistency. The ratio is 10^4. This is a statement of a "theoretical" maximum with no derivation [read: p.4]. I checked what it means: metallic uranium has about 4.8×10^22 atoms/cm3; fissioning every one at 200 MeV (3.2×10^-11 J) releases about 1.5×10^12 J/cm3 [calc]. At 5×10^6 W/cm3 that is burned out in 3.6 days [calc]. Required thermal neutron flux in pure U-235 (σ_f ≈ 585 b [memory, high]): φ = P/(N σ E) ≈ 5.6×10^15 n/cm²/s [calc], comparable to the best research reactors; for a fast spectrum with σ ≈ 1 b, φ ≈ 3×10^18 n/cm²/s [calc], far beyond any reactor. So 5000 kW/cm3 is not forbidden by energy conservation, but it is a flux and heat-removal claim, and the converter has nothing to do with flux. A nano-stack geometry cannot lift the neutron flux; at best it removes heat flux limits if conversion works. The TMS 2008 paper (Investigator lead) says "up to 1,000 times" rather than 10^4, so the author's own figures are inconsistent across papers (10^3 vs 10^4) [read: Investigator note, laro.lanl.gov abstract; I did not read the TMS paper].

b. Experiment. None. c. Never tested. d. Nothing in physics changes; engineering limits (neutron economy, burnup, heat) would.

Verdict: Refuted by calculation in the strong form (5000 kW/cm3 as a sustainable power density of fuel: needs the neutron fluxes above and burns the fuel in days). Weak form (higher local power density from better heat removal) is Plausible, untested. Note this verdict applies to the number's use; the paper gives no stated assumptions, so "theoretical" may mean a cooling-limited rather than a flux-limited bound; this ambiguity should be resolved by the author.

Summary table

Other mathematical notes kept for the record: the stack in the text is 700 nm and the figures show 600 nm (deep read, confirmed against the page image); thin layers are drawn ~10 nm thinner than stated; the "1200 electrons per nm3, <30 atoms" figure matches solid iodine (23 atoms, 1240 e) but not the dominant U and Au layers (~4400-4700 e/nm3) [read: deep read, densities from handbook values, memory high]; the paper cites Ziegler as "ZIGLER" [read: p.4].

Investigator findings I weighed (kept whatever the verdict)

• US20100061503A1 (84% as a stopping-fraction estimator; the 26.7 pW / 10 mV / 2.67 nA arithmetic): used above for claim 1. The author's own wording says the figure is an upper estimator and needs experimental build-up (para 0121). That candour should be recorded alongside the inconsistency.
• US20130121449A1 (extends CIci to fusion and transmutation, "nuclear entanglement and quantum non-localities"), US8067758B2 (radiation shielding), TMS 2008, Cleantech 2007, MRS 2008, Nanotech 2007 sibling (">90%" efficiencies), ILENRS 2012, NETS 2013, TechConnect 2015, Granthaalayah 2021: not read by me except as titles/abstracts in the Investigator's notes. The sibling Nanotech 2007 paper (research/files/Nanotech2007v4-3021.pdf) claims higher efficiency (>90%), so the author's efficiency claim rose between two 2007 papers without a new derivation [per Investigator]. A deep read is recommended.
• Klassen/Yershov (ISSP RAS) triple-layer experiments, Anufrienko/Chernov 25%, Yoshida/Tanabe 10^-5 %: these are the empirical anchors for claims 1-2. Single published source (one book chapter) for each number; primary papers (Russian) not yet read. They are not dismissed here; they are the main evidence for the weak form of claim 1.
• Gabbar et al. 2026 (MDPI): an independent review calls the technology "mainly conceptual ... patent-stage"; its 0.85 We/Wth is an ideal ceiling, which is the same kind of number as the 84%.
• Historical record: both applications abandoned (reason unknown, file wrappers not read); no measured CIci results found 2007-2026; competing explanations (no funding for fabrication and beam time; unpublished or negative results; unmeasured because the voltage/charge budget above makes the device uninteresting at mV) are the Investigator's and mine; the USPTO file wrappers and Alabama beam-time records would distinguish them. This is an absence in the record that I treat as an anomaly to be explained, not as evidence of failure.
• Forward citations (US10699820B2, WO2011011504A1, etc.) are not read; they show others cited the filings, which is not evidence it works.

Web checks made (budget 6, used 1)

• PuCoGa5 Tc 18.5 K and self-irradiation Tc decrease: confirmed [web] (OSTI/ScienceDirect links above; search snippets only). Not checked: W ≈ 25-30 eV per pair in solids; PuCoRh5 Tc; the 100 W/cm3 PWR figure; SiO2 breakdown field. These are tagged [memory] medium; the verdicts above do not turn on them (claim 1's inconsistency holds for any W between ~10 and ~100 eV: at W = 100 eV the 10 mV efficiency is still 10^-4).