{"id":"2334568f-1a28-435a-b7b6-26a816b0b5aa","arxiv_id":"2608.04954","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A 3D simulation of a convective silicon-burning shell in a 14 solar mass star burns out about 800 s earlier than a 1D MESA model, suggesting weaker convective boundary mixing and a convective-reactive energy profile.","lead":"Using 3D hydrodynamic simulations, this paper follows the silicon-burning shell of a 14 solar mass star from ignition to fuel exhaustion, the first time a complete silicon shell has been simulated in 3D. The 3D shell burns out earlier than a one-dimensional reference model because turbulent convection entrains less fresh fuel than standard 1D mixing prescriptions assume.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 3D-vs-1D lifetime comparison is confounded by different nuclear networks; the claim that lower CBM shortens the 3D shell's life by ~800s is not established unless fuel-consumption rates are shown to be network-independent.","rationale":"The reader's weakest assumption correctly identifies the reduced-network dependence of the central comparison. I agree with that assessment: the paper's own text (Sections 3, 6, and 7) provides explicit evidence that the 25-isotope network changes the dominant reaction pathways relative to MESA's 22-isotope network, and the paper even warns that mapping between substantially different networks can shift burning fronts. Because the central claim is a quantitative difference in shell lifetime attributed to CBM, the uncontrolled network difference is the most load-bearing concern. The proposed 1D MESA run with the same 25-isotope network is a decisive and practical test: it removes the network variable while preserving the 1D CBM prescription, thereby separating fuel-supply physics from nuclear-kinetics effects. The paper remains a valuable first 3D simulation of a complete convective silicon shell, with useful diagnostics of stiff boundaries and a plausible qualitative mechanism, so a conditional verdict is appropriate. No verdict change is needed relative to the reader's assessment, but the stated limitation should be made explicit in any acceptance recommendation.","tokens_in":22923,"tokens_out":7024,"duration_ms":83138,"concrete_test":"Run a 1D MESA model with the same 25-isotope network as the PROMPI simulation, using the same initial profile, CBM prescription, and weak-interaction treatment as the original 14SH21 model, and record the silicon-shell exhaustion time. If this 1D model also exhausts near ~2000 s (matching the 3D result) rather than ~3000 s, the lifetime gap is a network effect and the CBM-based explanation in the abstract needs revision. If the 1D model with the 25-isotope network still exhausts near ~3000 s, then the 3D model's earlier burnout is attributable to suppressed entrainment, supporting the paper's claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central causal claim, that the 3D silicon shell burns out ~800 s earlier because of weaker convective boundary mixing (CBM), rests on a comparison in which the two models differ not only in mixing treatment but also in nuclear network. Section 3 states that the MESA model uses the 22-isotope 'approx21_plus_Co56' network while PROMPI uses a 25-isotope network, and Section 7 explicitly warns that mapping a 1D model to 3D with a substantially different network can shift burning fronts or cause loss of hydrostatic equilibrium. Section 6 amplifies the issue: in the 3D model the dominant reaction flows involve (α,p) reactions and their inverses, e.g. 28Si(α,p)31P, whereas the MESA approx21 network burns primarily through (α,γ) channels. A network that processes fuel through faster (α,p) paths will consume silicon more rapidly even with identical entrainment. The paper therefore does not isolate CBM as the cause of the ~1000 s shorter burn time quoted in the text (the abstract's 'roughly 800 s' is itself inconsistent with that ~1000 s difference). The boundary-stiffness diagnostics (large Ri_b and declining ve/vrms in Figure 9) show that entrainment is weak, but they do not show that this weak entrainment, rather than the difference in nuclear kinetics, is what truncates the 3D shell's life. The 1D model's longer life is attributed to its CBM merging with an overlying flat-entropy layer at ~800 s, but no comparison of fuel-consumption rates before that merger is presented. Without controlling the network, the headline quantity is a joint consequence of network choice and CBM, and the causal attribution is not secure.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents the first 3D hydrodynamic simulations (PROMPI) of a complete convective silicon shell burning phase in a 14 solar-mass progenitor, evolved until fuel exhaustion. The authors compare the 3D run against the parent 1D MESA model, find that the 3D shell burns out earlier, and attribute this to weaker convective boundary mixing (CBM) caused by stiff boundaries. They also analyze the convective-reactive nature of the burning, including a double-peaked energy generation profile and the importance of (alpha,p) and reverse reactions. Resolution and opening-angle sensitivity are assessed with four simulations.","tokens_in":23140,"tokens_out":3320,"duration_ms":37548,"significance":"If the central claim is correct, the paper provides an important counterexample to the common assumption that 1D overshoot prescriptions overestimate or underestimate CBM in a uniform way; it suggests that CBM must depend on boundary stiffness (e.g., the bulk Richardson number). The study is also significant as the first 3D simulation to follow a silicon shell to exhaustion, and it includes a careful TKE budget, entrainment diagnostics, and Damköhler-number analysis. The authors are explicit about limitations of the reduced 25-isotope network, which is a strength. However, the headline quantitative claim is not yet supported by the evidence presented, mainly because the 3D and 1D models differ in their nuclear networks and reaction-flow treatment.","major_comments":[{"comment":"The abstract states that the silicon shell in the 3D model burns for 'roughly 800 s shorter', but Section 3 and Figures 2–3 state that the MESA model exhausts after around 3000 s while the 3D model exhausts 'just after 2000 s', a difference of approximately 1000 s. This quantitative inconsistency must be resolved, and the abstract should be corrected to match the actual difference or the text should be changed to support 800 s.","section":"Abstract and Section 3"},{"comment":"The central causal claim — that lower CBM in 3D is the reason for the shorter shell lifetime — is confounded by the use of different nuclear networks: MESA uses the 22-isotope 'approx21_plus_Co56' network while PROMPI uses a 25-isotope network. Section 6 shows that in the 3D model the dominant reaction flows involve (alpha,p) reactions and their inverses (e.g., 28Si(alpha,p)31P), whereas the 1D network burns primarily through (alpha,gamma) channels. A network that processes fuel through faster (alpha,p) paths will consume silicon more rapidly even with identical entrainment. The paper does not provide a control test, such as a 1D run with the 25-isotope network, a 3D run with the 22-isotope network, or at least a quantitative comparison of fuel-consumption rates before the 1D CBM merges with the overlying layer. Without such a control, the attribution of the lifetime difference to CBM alone is not established.","section":"Section 3 and Section 6"},{"comment":"The paper's own caveat in Section 7 states that mapping a 1D model to 3D with a substantially different nuclear network can lead to a loss of hydrostatic equilibrium or significant changes in the location of burning fronts. This is a direct threat to the validity of the 3D-vs-1D lifetime comparison. The authors should demonstrate that the burning-front location and the fuel-depletion history are sufficiently insensitive to the network difference for the claimed ~800–1000 s lifetime shift to be meaningful, or they should soften the causal claim in the abstract accordingly.","section":"Section 7"}],"minor_comments":[{"comment":"The abbreviation 'CMB' is used for 'convective boundary mixing' in the abstract; this should be 'CBM' for consistency with the rest of the paper.","section":"Abstract"},{"comment":"Caption of Figure 9: 'bulk Richardson, 𝑅𝑖𝑏, number' is awkward; consider 'bulk Richardson number, 𝑅𝑖𝑏'.","section":"Section 5"},{"comment":"'Damk\"ohlner numbers' appears with an incorrectly typeset umlaut; use 'Damköhler numbers'.","section":"Section 6"},{"comment":"The sentence 'We performed a one-zone 1 analysis around this radius' reads unclearly; consider 'We performed a one-zone analysis'.","section":"Section 6"},{"comment":"Missing space in 'at1.95×108cm'.","section":"Figure 11 caption"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of MNRAS and is of interest to the stellar-evolution and CCSN communities. The abstract overstates the confidence in the causal attribution of the lifetime difference; the body of the paper is more careful. The main concern is that the quantitative headline is not supported by the current comparison, and fixing it may require new calculations (e.g., a 1D run with the 25-isotope network or a simplified 3D run with a 22-isotope-like network). I therefore recommend major revision rather than rejection, as the qualitative finding — that 3D CBM is weaker than 1D overshoot for this stiff-boundary shell — is plausible and well-supported by the Richardson-number and entrainment diagnostics."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe thing to know: this is the first 3D simulation that follows a convective silicon-burning shell through to fuel exhaustion, and the central qualitative result—stiff boundaries, weak entrainment, no merger with the overlying flat-entropy layer—looks solid. The quantitative claim that the 3D shell burns ~800 s shorter because of weaker CBM is plausible but not fully isolated from the fact that the 3D and 1D models use different nuclear networks.\n\nWhat's genuinely new: the complete evolution to exhaustion, the double-peaked energy generation profile, and the convective-reactive interpretation supported by Damköhler numbers and the RA-ILES budgets. The paper is also honest about its own limitations—it explicitly says the MESA comparison is not one-to-one, flags the network and weak-interaction effects on iron-group composition, and warns that mapping with a different network can shift burning fronts. That authorship honesty counts for something.\n\nThe soft spots are real but not fatal. The stress-test concern lands: MESA uses approx21_plus_Co56, PROMPI uses a 25-isotope network with explicit (α,p) channels, and the paper itself says the dominant reaction pathways differ. So the ~1000 s lifetime difference (the abstract's 'roughly 800 s' is a bit loose but not wildly off) could be partly a network effect rather than purely a CBM effect. The paper doesn't present a like-for-like comparison of fuel consumption rates before the merger. The boundary-stiffness diagnostics (Ri_b, v_e/v_rms) show entrainment is weak, but they don't by themselves prove that weak entrainment, rather than faster fuel consumption, truncates the shell. Second, the main result rests on a single 3D realization (45med) run to exhaustion; the resolution study shows broad consistency, but the authors themselves note it's unclear whether the upper-boundary growth differences are stochastic. Third, data availability is 'on reasonable request'—for a simulation paper, that's a reproducibility weakness.\n\nWho this is for: stellar evolution and CCSN people, especially those calibrating CBM prescriptions in 1D codes. A serious referee should be assigned. The authors should be asked to (1) run a controlled comparison—same network in 1D and 3D, or at least show fuel-consumption rates are network-independent—and (2) tighten the number in the abstract to match the text. Even if the quantitative claim softens, the qualitative finding about stiff silicon-shell boundaries is worth publishing.\n\nMy take: send it to review.","headline":"First complete 3D silicon-shell burn to exhaustion; the qualitative stiff-boundary story holds, but the ~800 s lifetime difference is not cleanly separated from the nuclear-network difference.","tokens_in":23878,"tokens_out":4232,"would_cite":true,"duration_ms":43088,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Full 3D hydrodynamics makes a pre-supernova silicon shell burn out roughly 800 s earlier than a 1D model, because resolved, stiff convective boundaries entrain far less fresh fuel than the 1D diffusive overshoot prescription.","keywords":["silicon shell burning","convective boundary mixing","3D stellar hydrodynamics","convective-reactive burning","bulk Richardson number","quasi-statistical equilibrium","supernova progenitor","turbulent entrainment"],"falsifier":"Repeat the 3D simulation from the same mapped progenitor using the same 22-isotope network as the MESA run (or a substantially larger network), and check two things: whether the silicon shell still exhausts roughly 800 s earlier than the 1D model, and whether the upper-boundary entrainment velocity still falls to zero near 1500 s. If matching the networks removes most of the lifetime difference, or if the overlying silicon-rich layer is entrained once the reaction flow changes, the claim that stiff boundaries rather than network choice shorten the phase would be overturned.","tokens_in":22620,"feed_emoji":"🔥","tokens_out":14065,"duration_ms":137814,"temperature":0.7,"pith_summary":"This paper reports the first 3D hydrodynamic simulation of a complete silicon shell burning phase, run until fuel exhaustion for a non-rotating 14 solar-mass supernova progenitor. Its central claim is that the silicon shell burns out roughly 800 s earlier in 3D than in the corresponding 1D stellar evolution model, because resolved turbulent convection entrains far less fresh fuel than the 1D diffusive overshoot prescription. The mechanism is boundary stiffness: both convective boundaries have large bulk Richardson numbers (the stabilizing buoyancy jump divided by the turbulent kinetic energy), so the entrainment velocity drops to zero at the upper boundary within about 1500 s, whereas the 1D model's imposed mixing reaches a previously convective, silicon-rich layer above the shell and extends its life. The paper also establishes that the phase is convective-reactive, with nuclear burning timescales shorter than mixing timescales, and that the angular-averaged energy generation is double-peaked, with regions of near-zero or negative net heating produced by quasi-statistical-equilibrium readjustment and near-balanced forward and reverse $(\\alpha,p)$ reaction rates. A sympathetic reader would care because this implies 1D codes should tie convective boundary mixing to the physical stiffness of each boundary, and because multidimensional burning can change the pre-supernova shell lifetime and structure that feed core collapse.","feed_headline":"Silicon shell burns out 800 s sooner in 3D","feed_subtitle":"Stiff convective boundaries entrain less fuel than 1D overshoot models, shortening the burning phase","key_machinery":"The load-bearing object is the bulk Richardson number, \\(\\mathrm{Ri}_B = \\ell \\$\\Delta$ b / v_{\\mathrm{rms}}^2\\), where \\(\\$\\Delta$ b = \\int $N^{2}$ \\, dr\\) is the buoyancy jump integrated across a convective boundary, \\(\\ell\\) a turbulent length scale, and \\(v_{\\mathrm{rms}}\\) the rms convective velocity; it measures how stiff a boundary is relative to the kinetic energy available to entrain material. It enters through the entrainment law \\(v_e / v_{\\mathrm{rms}} = A\\, \\mathrm{Ri}$_B^{{-n}}$\\), which the paper uses to show that both silicon-shell boundaries are on the high end of previously studied shells and that entrainment declines as the shell stiffens. The second mechanism is the Damköhler number, \\(\\mathrm{Da} = \\tau_{\\mathrm{mix}} / \\tau_{\\mathrm{nuc}}\\), the ratio of mixing to nuclear timescales, which marks the shell as convective-reactive and explains why composition is not homogenised. The third ingredient is the reduced 25-isotope network built around QSE (quasi-statistical-equilibrium) groups, whose near-balanced $(\\alpha,p)$ reactions and their inverses, notably \\(^{28}\\mathrm{Si}(\\$\\alpha$,p)^{31}\\mathrm{P}\\) and \\(^{31}\\mathrm{P}(p,\\$\\alpha$)^{28}\\mathrm{Si}\\), produce the alternating exoergic and endoergic patches and the double-peaked mean energy generation, with a \\(^{44}\\mathrm{Ti}\\) bottleneck (slow \\(^{44}\\mathrm{Ti}(\\$\\alpha$,\\gamma)^{48}\\mathrm{Cr}\\)) causing accumulation of \\(^{36}\\mathrm{Ar}\\), \\(^{40}\\mathrm{Ca}\\), and \\(^{44}\\mathrm{Ti}\\).","core_discovery":"The paper's central claim, stated on its own terms, is that the silicon shell in a 14 M$_\\odot$ progenitor burns for roughly 800 s less in 3D than in the 1D MESA model from which it was mapped, and that this difference is a hydrodynamic consequence of stiff convective boundaries rather than a network artefact. In the MESA model, the exponentially decaying diffusive overshoot eventually connects the active shell to a flat-entropy region left by an earlier convective episode, mixing in fresh \\(^{28}\\mathrm{Si}\\) and \\(^{32}\\mathrm{S}\\) and keeping the shell burning to about 3000 s. In the 3D simulation the entropy jump at the upper boundary stays large, the bulk Richardson numbers remain high, and the entrainment velocity falls to zero after about 1500 s, so the shell consumes its own fuel and dies just after 2000 s without incorporating the overlying layer. The authors further argue that the burning is convective-reactive: Damköhler numbers of order unity or larger mean nuclear processing outpaces turbulent homogenisation, so abundances develop strong radial gradients and the mean energy-generation profile becomes double-peaked, with a near-zero or negative region between the peaks. They attribute this structure to quasi-statistical-equilibrium groups whose reaction flow is carried by $(\\alpha,p)$ reactions and their inverses, with forward and reverse rates so similar that small temperature fluctuations make local regions exoergic or endoergic; the dominance of \\(^{56}\\mathrm{Ni}\\) over \\(^{54}\\mathrm{Fe}\\) in the 3D ash, unlike the 1D model, is explicitly interpreted as a reduced-network and weak-interaction effect rather than a result of the mixing.","pith_inferences":["A natural testable extension is to replace the 1D model's uniform CBM with a stiffness-dependent prescription calibrated by these Ri_B values; the resulting pre-supernova models would predict a smaller silicon-shell mass and a different density gradient at collapse, which can be checked against explosion simulations.","Because forward and reverse $(\\alpha,p)$ rates are so nearly balanced, nuclear rate uncertainties translate almost directly into uncertainty in the shell's lifetime and energy profile; more precise measurements of \\(^{28}\\mathrm{Si}(\\alpha,p)^{31}\\mathrm{P}\\) and the calcium–titanium flow rates would shrink the spread in predicted silicon-burning durations.","The double-peaked, partially endoergic energy-generation signature may be a generic fingerprint of convective-reactive burning in QSE-regulated material, worth searching for in other late-stage shells and shell-merger events, where it would be observable as alternating heating/cooling patches.","The 800 s figure comes from one progenitor, one network, and one mapping; the direction of the effect is more robust than its size, and mapping it across initial masses, metallicities, and rotation rates would show how much of the pre-collapse structure actually depends on multidimensional mixing."],"forward_implications":["1D stellar evolution codes should stop applying a single convective boundary mixing value to all boundaries; this paper shows the length-scale overshoot prescription in the MESA model mixes too much, while mixing tied to boundary stiffness, for instance through the bulk Richardson number, would not.","The silicon shell burning phase is a convective-reactive event: within the shell, nuclear processing and turbulent transport run on comparable timescales, so composition gradients and energy generation cannot be inferred from the temperature profile alone.","The direction of the 1D-versus-3D boundary mixing difference is not universal: earlier oxygen, carbon, and neon shell studies found 3D entrainment stronger than 1D mixing, while the stiff silicon shell shows the opposite, so each convective boundary must be assessed on its own physical conditions.","If 3D is the better description, the pre-supernova structure changes at the level of silicon-shell lifetime and final shell extent, altering the mass, entropy, and composition profile that the collapsing core and the supernova shock will encounter."],"supporting_citations":[{"why":"Supplies the PROMPI hydrodynamics code and the bulk-Richardson-number entrainment framework used to measure boundary stiffness in the silicon shell.","marker":"Meakin & Arnett (2007)"},{"why":"Provides the non-rotating 14 M$_\\odot$, Z=0.001 progenitor model (14SH21) whose silicon shell is mapped to the 3D domain.","marker":"Whitehead et al. (2026)"},{"why":"The MESA stellar evolution code that produced the 1D comparison model and its time-dependent mixing and burning solution.","marker":"Paxton et al. (2011, 2019)"},{"why":"The exponential overshoot prescription whose diffusive mixing beyond convective boundaries is the 1D behaviour the 3D result is contrasted with.","marker":"Freytag et al. (1996)"},{"why":"The exponentially decaying overshoot length-scale formulation adopted by the MESA model to mix beyond the convective boundary.","marker":"Herwig (2000)"},{"why":"Defines the mass-dependent SH21 convective boundary mixing used in the progenitor model, the prescription whose 3D counterpart is weaker.","marker":"Scott et al. (2021)"},{"why":"Supplies the QSE-group description of silicon burning and the electron-fraction sensitivity used to interpret reaction flows and iron-group products.","marker":"Hix & Thielemann (1996)"},{"why":"Provides earlier shell-burning bulk Richardson numbers and entrainment-law constants that bracket the high stiffness of the silicon-shell boundaries.","marker":"Rizzuti et al. (2022)"},{"why":"Earlier neon-shell entrainment results used as the baseline for the silicon shell's Ri_B evolution and entrainment decline.","marker":"Rizzuti et al. (2023)"},{"why":"The 2D late-stage burning study that previously showed alternating exoergic and endoergic energy generation, supporting the double-peaked profile interpretation.","marker":"Arnett & Meakin (2011)"}],"fun_headline_variants":["3D silicon shell burns 800s shorter than 1D model","Stiff boundaries limit entrainment, cutting 3D silicon shell burn","Convective-reactive silicon shell: 3D burn ends sooner","3D simulation: less fuel mixing shortens silicon shell burning"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The reduced 25-isotope network used in the 3D run reproduces the energy generation and fuel-consumption rates well enough that the roughly 800 s lifetime difference is a mixing effect and not a network effect, even though the 1D comparison model burns with a different 22-isotope network.","fun_headline_variants_meta":{"raw":{"variants":["3D silicon shell burns 800s shorter than 1D model","Stiff boundaries limit entrainment, cutting 3D silicon shell burn","Convective-reactive silicon shell: 3D burn ends sooner","3D simulation: less fuel mixing shortens silicon shell burning"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000426,"raw_usage":{"total_tokens":2306,"prompt_tokens":1190,"completion_tokens":1116,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":806,"completion_tokens_details":{"reasoning_tokens":1039}},"tokens_in":806,"tokens_out":1116,"duration_ms":12208,"temperature":1.0,"reasoning_tokens":1039,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T12:58:48.266051+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the 3D simulation from the same mapped progenitor using the same 22-isotope network as the MESA run (or a substantially larger network), and check two things: whether the silicon shell still exhausts roughly 800 s earlier than the 1D model, and whether the upper-boundary entrainment velocity still falls to zero near 1500 s. If matching the networks removes most of the lifetime difference, or if the overlying silicon-rich layer is entrained once the reaction flow changes, the claim that stiff boundaries rather than network choice shorten the phase would be overturned.","supporting_citations":[],"review_version":1}