Write a comment

Summary

The authors synthesize graphene-oxide aerogels (density 0.018–0.035 g/cm³) doped with natural uranyl or thorium nitrate to about 7–8% U/Th by mass, using a hydrothermal and freeze-drying route. They characterize the fuel two ways: (i) alpha activity via CR-39 solid-state nuclear track detectors with an AI (R-CNN/ResNet-101) track-counting pipeline and MCNP-derived solid-angle corrections, giving a measured 16.7 ± 1.5 pCi/mg and a thinning-extrapolated ~49 pCi/mg; and (ii) 14 MeV (D-T) neutron-induced fission, detecting fission-fragment (FF) tracks that escape the low-density matrix (5,590 ± 475 recorded, extrapolated to ~45,000 created). On the strength of the low-density-enables-FF-escape concept, they propose applications in fission-fragment rocket engines (FFRE), direct energy conversion (DEC), modular reactors, and radiotherapeutics.

The central experimental result, that FFs demonstrably escape a doped ultra-low-density aerogel and are detected as distinct tracks, is the first such demonstration I am aware of, and it is what the evidence supports. One scope note worth making up front: a CR-39 track records that a fragment arrived above the track-formation threshold, so the experiment establishes escape as a fact but not the escape-energy spectrum. The low-loss part of the premise (escape "without depositing all their energy as heat") is inferred from the range argument rather than measured directly. My comments concentrate on tightening the distance between the solid escape demonstration and the four downstream applications named in the title and abstract.

Strengths

1. A clean, first-of-its-kind measurement, honestly bounded. The core claim, CR-39 detection of FF escape from a doped low-density aerogel under 14 MeV neutron irradiation with a matched undoped control showing no true FF tracks, is a direct demonstration of the paper's founding premise. The control (104 AI-detected FF tracks in the undoped sample, all found to be false positives on human inspection, consistent with the stated 94% precision) is exactly the right check, and it is reported transparently.

2. The detector-method comparison cuts the right way. The authors show that gamma spectrometry is near-useless here (154 vs 132 background counts; loading 11 ± 6%) and use that to argue for CR-39 as the more reliable low-signal probe. They motivate the strong measurement with the weak one rather than hiding it.

3. Three independent loading estimates agree within error. CR-39 alpha (7 ± 0.7%), neutron-fission back-of-envelope (8 ± 1.2%), and gamma (11 ± 6%) all overlap around 7–8%. Cross-method consistency on the load fraction is more convincing than any single number, and the AI pipeline is documented with quantitative precision/recall/F1 (94% / 89% / 92%) against human grading.

4. Extrapolations are, in the body, labeled as such. The 49 pCi/mg and ~45,000-FF figures are presented as model-based projections (thinning below the 1.8 mm alpha range; correcting for the ~40% FF-active outer shell), and the conclusion repeats that the 49 pCi/mg enhancement is "beyond the scope of the presented experiment." (The abstract is the exception; see M4.)

Major points

M1. A closed-form escape law would upgrade §4 from two point estimates to a design tool, and it reproduces your own numbers

In §4 you state that an aerogel FFRE sheet of ~20 µm loses ~1% of FF energy on average, and a ~200 µm sheet ~10%. These are the two load-bearing numbers behind the FFRE case, but they appear as isolated estimates with no stated functional form.

For a sheet of thickness t, fragments born uniformly and isotropically, and a straight-line fragment range R, two related but distinct quantities have simple closed forms:

• Mean fractional energy loss of the fragments that escape a thin sheet (t ≤ R): f_loss ≈ t/(2R). This is the quantity your FFRE energy-loss estimates refer to.

• Fraction of fragments that escape at all from a thick source (t > R): f_esc ≈ R/(2t), i.e. only a surface layer of order R contributes. This is a count, not an energy loss, and it is the same physics as your statement that FFs come "from only the outer ~40% of the sample volume."

These are two different observables and should not be drawn as a single continuous curve: the thin-branch number is an energy loss, the thick-branch number is an escaped count. (This is also the distinction behind cheap test 2 below.)

Using your own SRIM FF range in aerogel (you quote 1–1.2 mm; take R ≈ 1.1 mm), the thin-sheet energy-loss branch gives 20/(2·1100) = 0.91% at t = 20 µm (you report ~1%) and 200/(2·1100) = 9.1% at t = 200 µm (you report ~10%), so it reproduces both of your stated points to within rounding.

The ½ coefficient is not fitted, though it is not a one-line result either: it comes from averaging the in-material path fraction over birth depth and emission angle, with the finite range R cutting off the grazing-angle rays (the naive ∫dθ/cosθ would otherwise diverge). I confirmed the resulting curve against a Monte-Carlo sampling of birth depth and emission angle to four significant figures, and the derivation is short. Three concrete upgrades then follow, at essentially zero experimental cost:

1. State the energy-loss curve f_loss(t) = t/(2R) explicitly, so a reader can pick any sheet thickness for a target loss budget instead of interpolating between two points.

2. State the escape-count law R/(2t) as the quantitative version of your "outer ~40%" statement. One caveat: R/(2t) is the 1-D slab result, whereas your ~40% is a 3-D cuboid-shell estimate, so the two agree in spirit but not in exact geometry.

3. FF energy is bimodal (light ~95 amu and heavy ~139 amu groups) with different ranges. An energy-weighted two-group version of f_loss would make the FFRE energy number defensible rather than illustrative.

The FFRE section is where the paper's forward-looking claim lives, so a closed-form scaling that reproduces your data is probably the cheapest single way to strengthen it.

M2. Please separate "escape enhancement" (which you demonstrate) from "conversion efficiency" (which the aerogel does not address, and may make harder)

The abstract and §4 move from FF escape to "direct conversion … into electricity," citing ~50 mW/g existing power densities and the DARPA "Rads to Watts" push toward W/g. This blends two independent quantities:

• Escape fraction: the fraction of FFs that leave the fuel with useful energy. This is what your low-density matrix genuinely improves, and what you measured.

• Conversion efficiency η: the fraction of escaped-FF kinetic energy that becomes electricity in a DEC device. This is set by the collector, not by the fuel.

For a magnetic-sorting DEC (the mechanism a high-η design needs in order to separate fragments by E/q before decelerating them), the aerogel form does not help and arguably hurts. FFs are born 4π-isotropically throughout a 3-D volume, so the source has a large transverse emittance (roughly r_s·sinθ, of order 10⁻² m·rad for a centimetre-scale piece), while such an analyzer has an acceptance of order 10⁻³ m·rad. By Liouville's theorem the source phase-space volume cannot be compressed to fit, so a large fraction is lost before collection, and a bulk 3-D source is less favourable here than the thin planar films you cite as the current approach (a thin film emits into a narrow forward cone because the large-angle fragments self-absorb, giving a smaller emittance). For a retarding-field collector without E/q sorting, the limit is instead the intrinsic FF spread (more than ten charge states across two mass groups), which historically held fission-electric-cell efficiencies to the low tens of percent. In that non-magnetic case the aerogel's larger escape fraction can partly compensate, so the "worse" statement above is specific to the collimation-limited, magnetic-sorting route.

The scope worth stating is therefore narrow: the aerogel improves escape and escape energy, while conversion efficiency is a separate, collector-limited problem the fuel form does not solve. The ~50 mW/g and W/g targets belong to specific devices and read best if attributed as such rather than implied to be enabled by the fuel. One citation to double-check: ref. 32 (Spencer & Alam, Appl. Phys. Rev. 6, 031305, 2019) is a review of alpha/beta radioisotope batteries (specific power 1–50 mW/g; TiT₂ betavoltaics and Am-241 alphavoltaics), not a fission-fragment-DEC efficiency result, so it supports the ~50 mW/g figure but not high-η FF conversion.

M3. The "novel reactor / modular reactor" framing would benefit from one neutronics sentence; the driven-subcritical source you actually demonstrate is clean and honest, but the title implies more

The title and conclusion invoke "modular reactors" and "novel reactor design," yet the manuscript contains no neutronics: no k, no enrichment, no moderator or reflector context. As written, the reactor framing outruns the data. There is also an intrinsic tension worth naming, in that the very low density that helps FFs escape simultaneously starves neutron multiplication (fewer fissile atoms per unit path, more leakage).

Quantitatively, natural U (0.72% ²³⁵U) at ~0.02 g/cm³ total density has a fissile atom density orders of magnitude below anything that can sustain a chain reaction. Because k_inf is a material property independent of size, an unmoderated natural-U aerogel stays below 1 regardless of assembly size, so no finite configuration can reach criticality without either (a) high enrichment plus a strong moderator and reflector (a thin-film plus Be-dominant lattice can approach criticality, but that is a different fuel form), or (b) an external neutron driver, which is in fact exactly your experimental setup (the D-T generator). Your 14 MeV demonstration is a driven, subcritical fission source, and that framing is scientifically clean.

A related scope point: §1 notes the method "could be applied to fissile isotopes … allowing for fission without fast neutrons," which quietly extends the claim from what is shown (natural U/Th under 14 MeV fast fission) to thermal fission of fissile isotopes. The escape physics carries over (similar fragment masses and energies), but the neutronics does not, so one sentence saying so would keep the scope matched to the data.

I would suggest either (i) dropping "reactor" from the title and abstract and reframing the terrestrial application as a driven subcritical FF source / DEC target, which is what the data support, or (ii) adding one paragraph on the enrichment/moderator/driver regime a reactor application would require, together with the escape-versus-criticality tension. Either keeps the title's scope matched to what is shown.

M4. The "created FF" and "49 pCi/mg" figures rest on geometry corrections that a direct measurement could close cheaply, as you note yourself

You extrapolate 5,590 detected to 18,032 escaped (solid angle) to ~45,000 created (dividing by the ~40% outer shell), and 16.7 to 49 pCi/mg (thinning below the 1.8 mm alpha range). Each step is reasonable and labeled, but each also multiplies the uncertainty by a model factor. Two of them are directly testable with samples you can already make (see cheap tests). Until then, I would recommend the abstract carry the measured numbers as primary (16.7 pCi/mg; 5,590 FF tracks) and the extrapolated ones as clearly secondary. The body largely does this, but the abstract leads with "~49 pCi/mg," which could be read as a measured value.

M5. The escape estimate assumes a homogeneously dispersed dopant; a clustered distribution would shorten the in-fuel range and cut escape

The SRIM FF range of 1–1.2 mm is computed for homogeneous C₂O at 0.02 g/cm³ (§3). But the fuel is 7–8% U by mass introduced by uranyl adsorption onto graphene-oxide sheets, and uranyl species tend to segregate rather than disperse atom-by-atom. Inside a dense uranium/oxide aggregate the fragment range is far shorter than in the 0.02 g/cm³ matrix, so fragments born in such aggregates lose more energy, or fail to escape, relative to the bulk-density SRIM run. Escape fraction and escape energy therefore depend on the dopant micro-distribution, not only on the sample geometry of M1. This is the assumption I would most want pinned down, because it sits directly under the low-loss-escape premise. Cheap test: an EDS or EELS map on the SEM you already use, together with a SRIM run at the measured local dopant density, would show whether the dispersion is fine enough for the bulk-density range to apply.

Minor points

• Gamma S/N (Fig 2): 154 vs 132 background counts (11 ± 6%) is barely above background, as you say. It works as a negative-result motivation for CR-39; you might also state the significance and the convention used, e.g. about 22 net counts, which is ~1.3σ referenced to √(S+B) ≈ 17 (or ~1.9σ referenced to √B ≈ 11.5), so the reader sees how weak it is and by which measure.

• Alpha exposure in air rather than vacuum (§2.1): good that you note the extra attenuation. Consider quantifying the ~2% air-path loss you mention alongside the 32.5 ± 2.6% solid-angle term in the systematic budget.

• Range citations: you quote SRIM alpha range 3.5 cm in air, 1.8 mm penetration depth, and FF range 1–1.2 mm in aerogel / 2–3 cm in air. Please state the assumed aerogel composition and density used for the FF SRIM run in the caption (you give C₂O at 0.02 g/cm³ in §3; please confirm the alpha run uses the same, and see M5 on local vs bulk density).

• Isp claim (§4): the ">500,000" figure for FFRE is quoted without a reference for the specific configuration, and no thrust or mass-flow is measured here, so it is a property of the FFRE concept (refs 6–9) rather than of this fuel. Point to whichever ref gives it.

• Reference 31 (Stefanishin et al.) has duplicated author fields (Latin plus Cyrillic transliteration of the same names); worth cleaning up before the journal version if not already done.

• Terminology: "true alpha activity … is 49 ± 4.5 pCi/mg" (§3) reads as a measured fact; "projected" or "extrapolated true activity" would match your own §5 caveat.

Cheap tests that would close the biggest gaps

1. Directly measure thinned (<1.8 mm) samples. This is the single highest-value experiment: it tests both the 49 pCi/mg alpha extrapolation and the thin-branch energy-loss law of M1 in one shot, using fabrication you already have. You flag it as future work, and it is worth prioritizing, since two headline numbers depend on it.

2. Energy-resolved FF escape versus thickness. Measure residual FF energy (track diameter and depth on CR-39 is already your observable) as a function of sample thickness. This distinguishes "escaped with full energy" from "escaped after partial energy loss": it separates the escape count you report from the energy-loss fraction the FFRE case actually needs, and it is what would turn the escape demonstration into a low-loss-escape demonstration.

3. A dopant-distribution map (EDS/EELS) plus a SRIM run at the local dopant density (M5), and a two-group (light/heavy FF) SRIM range table in the SI, so the energy-weighted escape estimate in M1 is reproducible.

Overall assessment

Significance: valuable. The experimental core (the first CR-39 demonstration of fission-fragment escape from a doped ultra-low-density aerogel, with a clean control and a documented AI counting pipeline) is a genuine contribution to the FFRE/DEC fuel-form question. I would not yet call it important or landmark, because the demonstration is a driven subcritical source at laboratory scale rather than a device or a reactor, and because it establishes escape as a fact rather than measuring the escape-energy spectrum the applications need.

Strength of evidence: solid for the escape demonstration, incomplete for the named applications. The FF-escape and loading measurements are well controlled and cross-checked. The four downstream applications (FFRE efficiency, direct conversion, modular reactor, radiotherapy) are at present motivated rather than evidenced: the direct-conversion efficiency and the reactor-criticality claims are, as written, not supported by the data (M2, M3), the leading numbers (49 pCi/mg, 45,000 FF) are extrapolations (M4), and the escape estimate depends on a dopant-distribution assumption that is not yet verified (M5).

Recommendation. This is a paper worth having in the record, and the fixes are almost entirely textual and cheap: add the escape-fraction law (M1), which reproduces your own numbers and turns §4 into a design tool; separate escape enhancement from conversion efficiency (M2); add one neutronics sentence or reframe "reactor" as "driven subcritical source" (M3); keep measured numbers primary in the abstract (M4); and pin down the dopant distribution (M5). None of these requires new apparatus, and together they align the claims with what is a genuinely nice measurement.

I would gladly share the escape-formula derivation and the subcritical-multiplication estimates referenced above if they are useful to the authors.

Sources consulted for this review:

• Spencer, M. G. & Alam, T., "High power direct energy conversion by nuclear batteries," Appl. Phys. Rev. 6, 031305 (2019). DOI: 10.1063/1.5123163. A radioisotope (alpha/beta) battery review (specific power 1–50 mW/g; TiT₂ betavoltaics / Am-241 alphavoltaics), not a fission-fragment-DEC efficiency result.

• The emittance and thin-foil argument in M2 rests on Liouville's theorem and the fission-electric-cell / magnetic-collimator DEC design-study literature; it is qualitative.

Competing interests

The author declares that they have no competing interests.

Use of Artificial Intelligence (AI)

The author declares that they used generative AI to come up with new ideas for their review.