Exploratory research · nitrogen-15
✨ Fuel assembly simulation
Staggered pellets must arrive with the right position and velocity, then merge into a hot, inward-moving fluid. The nitrogen-15 study contains a spherical burn surrogate and two-dimensional tests of the prepared fluid and its burn. These assembly results use a 1.131 × 10¹¹ kg nitrogen-15 test pulse. Their applicability to the selected 1.131 × 10¹³ kg pulse remains to be assessed.
Correction. The original launcher-error tests sorted displaced fuel into spherical shells and assigned a radial velocity profile. This erased the asymmetric arrival pattern and its transverse momentum before the burn. The resulting yields are radial-profile sensitivity tests. The revised launcher targets have a separate joint prepared-fluid test.
Pellet flight · Plasma handoff · Engineering tolerances · Correlated errors · Mesh audit · Axisymmetric handoff · Reactive dipole · Launcher acceptance · Reproduce the runs
Pellet release and first compression
The nitrogen-15 stage, B1, has the least gain of the four published pulses (7.96) and compresses one hundredth of the mass of the other stages into a similarly sized hot inflow. In its prior spherical contact calculation, only 82.3% of the mass reached the proposed contact state by reversing the hot flow. The outer 17.6% needs a different density history, and the central 0.08% is reserved for a seed. Those facts make B1 the hardest case on the available assembly evidence. Spherical contact study.
The flight calculation starts at release, 1.1 million km from the center. It uses the reference architecture of 1,000 organizational complexes and illustrates 100,000 spatially distributed outlets. The incoming mass is sampled in 128 shells and propagated ballistically to 100 µs before plasma start. Each sample represents a parcel of a packet. The middle 90% of the mass coasts for about 325–1,034 s. Inside the last 100 µs, the nominal flight sensitivity calculations use the archived spherical pressure trajectory as their guide. They apply radial displacements from timing and speed errors, and transverse displacements from pointing and residual aperture. Flight samples and assumptions.
The unperturbed pressure trajectory first reaches 1 g/cm³ in one shell about 71 µs before plasma start. The mean inside its outer radius reaches 1 g/cm³ about 26.4 µs before plasma start, at a 300 m radius. By then, about 83% of the mass is in radial bins above the threshold. The ideal reference ends with about 95% above it. Continuing the cold ballistic tangent instead would reach 1 g/cm³ locally at 83 µs and would put the central density at 187 g/cm³ by its nominal bulk crossing. Centimeter grains cannot reach that density by filling space. A physical contact model must take over before the bulk crossing. This packing test allows stream collisions to begin earlier.
| Flight case | Median radius | Mass ≥ 1 g/cm³ | Mass inside ideal 99% radius |
|---|---|---|---|
| Ideal geometry | 85.5 m | 95.0% | 99.0% |
| 1 µs / 3 nrad / 3 ppb | 85.6 m | 94.0% | 99.0% |
| 10 µs timing alone | 83.4 m | 91.5% | 97.4% |
| 30 nrad pointing alone | 98.1 m | 93.5% | 98.2% |
| 30 ppb speed alone | 87.1 m | 91.5% | 97.0% |
| 100 m parallel exit pupil | 112.7 m | 93.5% | 97.2% |
The radial density estimate uses equal-mass bins and excludes the most distant 0.5% of the samples. A local density increase can coexist with a worse overall profile. The 100 m aperture case illustrates what happens if a filled exit pupil stays parallel. The reference design therefore needs convergent optics or an equivalent correction to meet a much smaller terminal spot. The cases in the next section assume that correction is in place.
From 1 g/cm³ to the plasma solver
The radial runs switch to a single fluid at the nominal bulk crossing and evolve pressure, shock viscosity and grey radiation for 26.4 µs. Three cases use the nominal radii. One takes the temperatures from the prior reversible FEOS path and two deposit an assumed 31.0 or 123.9 GJ/kg above the FastEOS 100,000 K energy reference in each shell. These last two cases vary the effective heat per unit mass at a single time, the bulk crossing. Each case exchanges initial kinetic and internal energy at a common total energy. The arrival-error cases instead remap the shell radii by enclosed mass, retaining the reference temperature and velocity ordering and rescaling inward speeds to the common energy. This retains only the spherical distribution of radii. The mapping erases the relative location of the seed and fuel centroid, angular density and phase structure, transverse momentum, and parcel-level changes in velocity.
The resulting profiles feed the same 128-zone, 27-isotope burn solver for 82.28 µs with the median rate update and grey opacity of 0.001 m²/kg. The reference is the published nitrogen-15 run. The pass line is 2.289 × 10²⁴ J of deposited nuclear energy, defined as modeled nuclear release less the weak-energy debit. Preparation losses are carried through the actual handoff state.
| Bulk crossing case | Handoff radius | Deposited energy retained | Spherical calculation |
|---|---|---|---|
| Reference thermal profile | 197.1 m | 99.3% | Pass |
| 31.0 GJ/kg uniform heat | 196.2 m | 93.6% | Pass |
| 123.9 GJ/kg uniform heat | 198.6 m | 86.0% | Below threshold |
| 1 µs / 3 nrad / 3 ppb | 213.5 m | 99.2% | Radial surrogate |
| 3 µs / 9 nrad / 9 ppb | 249.4 m | 101.2% | Radial surrogate |
| 10 µs / 30 nrad / 30 ppb | 378.6 m | 2.4% | Radial surrogate |
The radial surrogate for the largest input perturbation produced 2.4% of the reference yield even after extending the run to 120 µs. About 90.9% of its fuel was above 1 g/cm³ at handoff. This comparison tests whether a spherical density threshold alone predicts burn quality.
In a parallel-sector approximation, the transverse energy available from geometric obliquity is about 56 MJ/kg at 100,000 independent outlets or 5.6 GJ/kg at 1,000. The former is about 0.045% of the 123.9 GJ/kg contact energy in the earlier B1 calculation. This small adjacent-stream contribution cannot explain the required contact heating by itself. For two equal cold streams to thermalize 123.9 GJ/kg in a perfectly inelastic encounter requires a relative normal speed of about 1,000 km/s. At the flight RMS speed of 2,370 km/s, an equal-speed collision would need a separation angle of roughly 24°. Those are scales for choosing a contact model. Early breakup, encounters between nonadjacent beams, and the expansion of the outer material still determine where momentum and entropy end up.
Engineering targets and margin
Before the joint yield test, my central feasibility estimates for an advanced launcher with dedicated metrology were 1 µs RMS release timing, 3 nrad RMS pointing per transverse axis, and 3 parts per billion RMS speed error after the final charged correction. The conditional 80% credible intervals were 0.1–10 µs, 0.5–30 nrad, and 0.3–30 parts per billion. These are engineering judgments. They assume calibrated optical baselines and omit shared drift and excursions in the final neutral coast. Over a 100 km metrology baseline, 3 nrad is a 0.3 mm differential displacement. At 2,370 km/s, 3 parts per billion is about 0.13 ns in time of flight across that baseline. Over the full coast the speed error shifts arrival by 3.3 m, comparable to the 2.4 m from release timing alone. Existing dust accelerators demonstrate pickup timing and charge selection at far smaller grain size, speed and throughput. University of Colorado dust accelerator.
For residual charge I would budget a 5 V grain potential, with a subjective 80% range of 1–30 V in magnitude for successfully neutralized 1 cm grains. The isolated conducting sphere model gives about 1.3 × 10−9 C/kg at 5 V. The median transverse drift is 0.35 mm at 5 V in a uniform 1 µT field, and the small-expansion estimate for an uncompensated same-sign cylindrical beam is below 1 µm over the nominal flight. At 30 V those estimates are about 2 mm and 0.8 µm. This gives considerable margin against a few meters of trajectory displacement in the stated field and beam models. Charge tails, coherent chamber fields, neutralization heating and coating damage remain unmeasured. Spacecraft charge measurements for comparison.
For 100,000 outlets with independent 1% RMS mass scatter, the vector dipole of the aggregate mass distribution is of order 0.003% RMS under uniform coverage. A coherent imbalance survives that averaging. The resulting flow requires angular coverage and pressure exchange across sectors to evaluate.
Correlated launch errors
Here an error is correlated when whole sectors drift together. One hemisphere is early and the other late in the timing mode. In the speed mode one hemisphere has a positive fractional bias and the other a negative bias. A hemispheric split maximizes the dipole moment for a bounded scalar error on a uniform sphere, since each side takes the sign of its projection onto the dipole axis. For pointing, the polar dipole tips each ray toward the same pole at fixed magnitude. In the joint dipole, timing and speed add, pointing reinforces their displacement, and either the early or the late hemisphere receives the excess mass. A joint amplitude of 1 means ±1 µs timing, ±3 ppb speed, 3 nrad pointing, and ±1% hemispheric loading.
The flight calculation applies uΔt, RΔu/u, and RΔθ as radial or transverse displacements, then projects the displaced mass back into spherical shells at the nominal 1 g/cm³ bulk crossing. It tapers the errors toward zero in the central delivery region on a 20 m scale. The model retains the reference temperature and radial speed ordering and holds total prepared energy fixed. The projection gives a burn sensitivity to radial profile changes. It averages away angular shear and preserves the radial component only through the reference speed ordering. The polar cases report the centroid displacement about the fixed seed before this projection. At three times the joint test amplitude, it is 18.3 m with the late hemisphere heavier and 13.9 m with the early hemisphere heavier. The loading mode instead evolves the two hemispheres as separate spherical columns of different density and averages their nuclear outputs. It holds total mass fixed and omits pressure exchange across the equator. At ±20% loading, the light side’s yield drops sharply. The heavy side makes up part of the loss, leaving 66.8% of the published pulse energy overall.
The former pass and fail brackets are withdrawn as launcher tolerances. Spherical rank ordering erases the timing and speed dipoles along with the pointing dipole. For loading, separate hemispheric solves suppress pressure exchange across the equator. The archived numbers document the response of those surrogate models.
A common timing shift is an exact time translation if the seed and burn window shift with it. A common fractional speed error leaves some shell-to-shell phase error after retiming because flight times differ. The common-speed tests subtract the clock correction that minimizes the mass-weighted squared radial displacement. Separate timing corrections for each velocity group could remove more of that error. A change in aim can translate the pulse if all launchers and the seed aim at the same new point and their times are adjusted for the new path lengths.
For a fully correlated grain charge of the same sign, adding the reference uniform-field drift and uncompensated beam expansion gives an illustrative 3 m displacement at about 30 kV grain potential. A coherent electric field adds a displacement of approximately (q/m)Et²/2. That is a trajectory-budget scale in the specified 1 µT field and cylindrical beam model. The 5 V working target is set by the assumed neutralizer, and the estimates do not cover fast field variation, charge tails, or heating.
All six timing, pointing, and speed targets together
At full proposed scale, the independent error widths are 1 µs for timing, 3 ppb for speed, and 3 nrad for each pointing axis. The correlated errors are ±0.3 µs and ±1 ppb across hemispheres, plus a 1 nrad polar pointing dipole. The calculation superimposes all six before projecting the arriving fuel onto the radial grid. It also applies ±0.3% correlated loading and tests both the early-heavy and late-heavy orientations. Independent scatter uses a fixed Sobol sample with Gaussian tails clipped at three standard deviations. The independent release errors apply throughout the profile. The 20 m seed taper applies to the coherent trajectory errors.
The combined run feeds the scalar inputs into the same radial projection. Its yield cannot establish the simultaneous launcher budget. Axisymmetric flow can resolve an aligned dipole and its interaction with the seed. Independent outlet scatter and nonaligned perturbations also need a model for azimuthal structure or a validated statistical closure.
Correlated-error runner · Scan results · Profiles and solver records · Launcher accuracy specs.
Why the mesh changes the burn
Most of the 64-to-128 difference appears in the controls, before any launch errors enter the model. The coarse control starts preparation with 0.29% more energy. It radiates 0.83% of that energy before plasma start, compared with 0.33% for the 128-zone control. Its burn peaks at 513 MK instead of 540 MK. Both pulses are expanding throughout the final profile, with maximum temperatures below 40 MK at the end of the comparison window. Bypassing the sampling and remap entirely in the 64-zone zero-error control changes its yield by about 0.002 percentage points.
| Radial zones | Zero-error control | Early-heavy radial surrogate | Late-heavy radial surrogate |
|---|---|---|---|
| 64 | 91.17% | 90.81% | 90.81% |
| 128 | 99.26% | 99.21% | 99.21% |
| 256 | 102.68% | 102.89% | 102.89% |
All percentages in this table use the same published nitrogen-15 pulse as denominator. The 256-zone runs subdivide the existing 128-zone preparation profile in enclosed mass. They retain each original cell’s volume and mass-weighted temperature and interpolate the boundary velocities. This tests the forward solver on a nested mesh with the same source information. The lower joint result changes yield by -0.390%, -0.055%, and +0.202% relative to its own control at 64, 128, and 256 zones, respectively. The 256-zone control peaks at 610 MK, compared with 540 MK at 128 zones. Absolute yield and peak temperature still shift with mesh, and the yield shifts are larger than the measured effect of the target errors.
To isolate the burn, I also took the exact prepared 128-zone control, rebinned it to 64 zones, matched its initial energy, and skipped the preceding preparation. It produced 93.48% of the 128-zone control’s yield. Halving the timestep cap changed that ratio by less than 0.00001 percentage points. Halving the artificial-viscosity length scale, which halves the linear coefficient and quarters the quadratic coefficient, raised the ratio to 99.34%. This shows sensitivity to the numerical treatment of compression and shocks. The weaker viscosity also raised the peak temperature to 584 MK, so matching the integrated yield alone gives a poor test of physical accuracy.
Removing the seed taper and restoring the Gaussian tails together gave 99.21% in the 128-zone late-heavy case. The standard version gave 99.21%. A kinematic estimate gives about 0.10% of the nominal kinetic energy in transverse motion at the bulk crossing for the combined target case. If each parcel carried that angular momentum to its nominal plasma-start radius, the estimate would be 0.31%. For the isolated 30 nrad polar dipole, the plasma-start estimate rises to about 15%. The spherical projection discards that motion. This estimate follows the nominal radii and assumes no torque between bulk crossing and plasma start. The omitted motion is one part of the error. The projection also erases the seed-relative displacement, angular shear and asymmetric merger shocks. The mesh comparison measures the convergence of the radial surrogate.
The flight screen also assumes that the final optics focus the beam’s own width. A small pointing error in the beam axis can coexist with a broad beam at the target. The filled, parallel 100 m exit pupil above illustrates that separate failure mode.
Audit results · Prepared-profile audit runner · Solver records and inputs.
Axisymmetric handoff test
A two-dimensional cylindrical calculation now evolves the prepared fuel from the nominal 300 m bulk crossing to plasma start, about 26.4 µs later. The grid carries mass, both momentum components, total energy and a passive seed tracer. It resolves pressure exchange across the equator. The seed starts centered and moves with the fluid. A smooth polar pointing dipole vanishes on the axis and reaches its specified amplitude at the equator. The joint cases apply the proposed coherent timing, speed, pointing and loading inputs, with the pointing bias aligned to reinforce the phase displacement. Cylindrical conservation equations.
This is a first-order finite-volume calculation with an ideal fully ionized gas and equilibrium radiation pressure. The imposed fluid handoff deposits overlapping parcels into a common cell while conserving their total energy and axial momentum. The setup assumes the bulk-crossing temperature profile and uses the same fluid for the seed and fuel. Nuclear reactions, radiative transport, kinetic contact before bulk crossing and independent azimuthal errors remain outside this test.
| Input, 2 m cells | Axial fuel offset from seed | Density sampled by seed tracer | Temperature sampled by seed tracer |
|---|---|---|---|
| Control | -0.00 m | 29.71 g/cm³ | 41.5 MK |
| 1 nrad polar pointing | +0.80 m | 29.61 g/cm³ | 41.4 MK |
| 9 nrad polar pointing | +7.00 m | 23.36 g/cm³ | 38.9 MK |
| 30 nrad polar pointing | +21.78 m | 6.38 g/cm³ | 25.3 MK |
| ±6 µs hemispheric timing | -4.90 m | 33.75 g/cm³ | 43.1 MK |
| ±12 µs hemispheric timing | -9.62 m | 50.30 g/cm³ | 49.2 MK |
| ±9 ppb hemispheric speed | +0.33 m | 38.27 g/cm³ | 41.7 MK |
| ±15 ppb hemispheric speed | -0.26 m | 55.49 g/cm³ | 43.8 MK |
| ±5% hemispheric loading | +2.37 m | 29.69 g/cm³ | 41.4 MK |
| ±20% hemispheric loading | +9.48 m | 29.44 g/cm³ | 41.3 MK |
| Coherent inputs, early side heavier | -0.81 m | 29.93 g/cm³ | 41.5 MK |
| Coherent inputs, late side heavier | -1.10 m | 29.91 g/cm³ | 41.5 MK |
| Cell width | 30 nrad axial fuel offset from seed | 30 nrad seed-tracer density relative to same-grid control |
|---|---|---|
| 4 m | +21.77 m | 28.7% |
| 2 m | +21.78 m | 21.5% |
| 1 m | +21.69 m | 19.2% |
The large pointing case displaces the fuel from the seed and lowers the seed-tracer density. The displacement is much less mesh-sensitive than the density sampled by the seed. The 1 nrad displacement is smaller than a cell on the finest grid, and the seed-density response at that amplitude remains unresolved. Reversing the dipole reverses the displacement and preserves the seed density in the reflection check. The handoff results establish an asymmetric response in this model. A yield tolerance requires a resolved reactive, radiating evolution through compression and burn, with the contact and seed treatment checked independently.
Axisymmetric runner · Handoff records and checks. Run python axisymmetric.py?v=3a96357d7cd1 B1-backward-feos128-refined.npz --cell-m 2 --pointing-nrad 30 --out handoff.json with NumPy installed.
Yield through the asymmetric burn
The prepared-fluid simulations put the isolated coherent pointing tolerance on the order of 10 nrad for this polar pattern. The 2 m grid brackets a crossing of 90% of published B1 between 9 and 12 nrad. On the completed 1 m sweep, 15 nrad passes and 30 nrad is the first sampled amplitude below 90%. The physical tolerance remains open because the handoff is imposed and the transition shifts with mesh.
The reactive calculation uses the cylindrical finite-volume grid throughout compression, ignition and expansion. The solver evolves both momentum components and all 27 isotope abundances, uses the existing B1 electron EOS table and median carbon-rate update, and feeds net nuclear energy back into the fluid. An implicit local reaction update handles the fastest channels. The grey radiation step exchanges energy between neighboring cells with a flux limiter and allows free escape at a low-density surface. This is a numerical approximation to equilibrium grey transport. The first-order chemistry and transport splitting, surface rule, and stretched outer grid all limit the precision of the result.
Yield here is nuclear energy deposited after the weak-energy debit. Escaping grey radiation has a separate ledger. The published B1 reference is 2.543 × 1024 J. The zero-error control uses the same mesh, transport and handoff assumptions as the corresponding dipole. These comparisons hold the seed at its nominal initial location unless a seed variant is identified.
| Inner cell width | Energy vs. published B1 | Peak temperature | Through |
|---|---|---|---|
| 4 m | 81.8% | 379 MK | 82.3 µs |
| 2 m | 94.5% | 417 MK | 82.3 µs |
| 1 m | 99.0% | 439 MK | 82.3 µs |
The EOS spot check differs from the spline reference by less than 0.01%. A fixed-density full-network burn approaches the stiff BDF reference as the substep shrinks. The hot-start replay still changes with resolution, so the pointing result is an order-of-magnitude estimate. The main runs carry the energy ledger through 82.3 µs. Shorter sensitivity runs show their recorded end times.
| Prepared-fluid case | Inner cell width | Energy vs. published B1 | Energy vs. same-mesh control | Through |
|---|---|---|---|---|
| Zero error | 2 m | 91.2% | 100.0% | 82.3 µs |
| 9 nrad | 2 m | 92.0% | 100.9% | 82.3 µs |
| 12 nrad | 2 m | 86.4% | 94.8% | 82.3 µs |
| 15 nrad | 2 m | 68.9% | 75.5% | 82.3 µs |
| 30 nrad | 2 m | 0.9% | 1.0% | 82.3 µs |
| Zero error | 1 m | 95.1% | 100.0% | 82.3 µs |
| 1 nrad | 1 m | 95.2% | 100.1% | 82.3 µs |
| 9 nrad | 1 m | 100.3% | 105.4% | 82.3 µs |
| 15 nrad | 1 m | 96.5% | 101.5% | 82.3 µs |
| 30 nrad | 1 m | 1.2% | 1.3% | 82.3 µs |
| 1 nrad with independent scatter | 2 m | 88.6% | 97.1% | 82.3 µs |
| 9 nrad with independent scatter | 2 m | 86.1% | 94.5% | 56.0 µs |
| 15 nrad with independent scatter | 2 m | 51.7% | 56.7% | 82.3 µs |
| 1 nrad with independent scatter | 1 m | 98.0% | 103.1% | 82.3 µs |
| Combined budget, early side heavier | 2 m | 88.5% | 97.1% | 56.0 µs |
| Combined budget, late side heavier | 2 m | 88.5% | 97.1% | 56.0 µs |
The short diagnostic runs record the time and axial centroid when cumulative deposition first reaches 10% of the published pulse. The hot-start replay reaches that point at 32.9 µs after nominal plasma start, compared with 32.47 µs in the published spherical calculation. The prepared control reaches it at 34.5 µs around z = +0.0 m, while 9 nrad reaches it at 35.6 µs around z = +9.3 m. This centroid measures where early deposition occurred.
Starting the imposed fluid evolution 40 µs before plasma start instead of 26.4 µs gives 86.0% in the control and 86.3% at 9 nrad. Lowering the handoff temperature by 10% at fixed initial energy gives 90.1% and 93.9%, respectively. At 9 nrad, shifting the central material with the rest of the pulse gives 91.5%. The 12 nrad result changes from 86.4% to 88.1% with half the Courant limit. These are handoff and numerical sensitivity cases. The earlier control falls below 90% before any pointing error is applied.
A statistical check in three dimensions
For uniform coverage and independent Gaussian errors of standard deviation σ per tangent axis, projecting the pointing field onto the smooth polar dipole gives three Cartesian components with standard deviation σ√(3/(2Neff)). Here Neff is the number of independent, equally weighted groups, or 1/Σwi2 for normalized group weights and approximately isotropic coverage. This is a model of the coherent low-order part of the random error. For 3 nrad scatter per axis, its random dipole has a 95th percentile of 0.32 nrad for 1,000 independent groups and 0.032 nrad for 100,000. A shared aiming bias adds as a vector and does not average away.
The scatter cases also sample independent azimuthal and meridional pointing plus radial timing and speed errors, then deposit mass, momentum and energy into cylindrical cells. The calculation assumes local mixing of the unresolved azimuthal streams. This tests the effect of a broadened and thermalized arrival distribution. Coherent higher angular modes and patch-to-patch shocks in three dimensions remain outside the closure. With 1 nrad coherent pointing, 3 nrad independent pointing per transverse axis, 1 µs independent timing, and 3 ppb independent speed, yield is 88.6% on the 2 m grid and 98.0% on the 1 m grid. A second 2 m scatter draw gives 89.8%. At 9 nrad, adding the same scatter changes the 2 m result from 92.0% to 86.1%. With the coherent timing and speed split and both loading orientations added, the lower 2 m case gives 88.5%. The wider budget falls below the yield threshold on the 2 m grid. The tighter budget is tested below.
Reactive runner · Microphysics checks · Dipole statistics · Yield results · Full histories and versioned runners.
Joint launcher acceptance test
The revised ten-entry budget passes the declared B1 simulation screen. The revised budget halves the timing, pointing, and speed allowances. It keeps the loading and charge values and defines how they are measured. The yield criterion is at least 90% of the published B1 deposited nuclear energy, or 2.289 × 1024 J. All ten entries.
The coarse screen applies all ten values together. It crosses both polar pointing signs and both hemispheric loading signs with timing and speed either adding or opposing. Additional cases reverse the coast fields or halve the timestep cap. Loading scatter is aligned with either timing tail. The finite-group cases sample 1,000 equal-area groups in three dimensions and project their arriving streams onto the fluid grid with the same local azimuthal mixing closure. From 2,048 candidate fields, the two selected cases are near the 95th percentile and at the maximum of a low-order displacement score. They are stress patterns selected by geometry. Two interior cases use half of every amplitude and the full independent limits with the coherent offsets removed.
The nominal coarse runs span 90.41% to 91.53% of published B1. The highest-score finite-group case gives 96.48% on the 1 m grid. The two interior cases give 91.57% and 90.43%. The 0.75 m hot-start replay gives 100.26% of the published reference. The guarded yield is the lower of the two finer results minus their absolute difference and a 1.09 percentage-point timestep allowance measured by halving the coarse Courant limit. This is a numerical screening margin. The automated gate also checks the run settings, matched initial mass and energy, energy and composition ledgers, and yield change over the final 2 µs.
| Prepared handoff | Coarser fine-grid yield | Finer-grid yield | Guarded yield |
|---|---|---|---|
| Nominal bulk crossing | 96.42% (1 m) | 97.60% (0.75 m) | 94.14% |
| 10% lower temperature at fixed energy | 92.98% (1 m) | 93.64% (0.75 m) | 91.22% |
| Earlier start at 40 µs | 95.15% (1 m) | 96.72% (0.75 m) | 92.49% |
The combined earlier and cooler start gives 91.91% in its 1 m control and 93.13% with all ten inputs. This combined case is a separate stress test. These handoff states are discrete tests. The calculation assumes the prepared profile and instantaneous fluid mixing at each selected start.
In a profiled 2 m replay, the timestep fell from 18 ns to 6 ns between 42.1 and 43.2 µs, when the deposited energy went from 99.70% to 99.90% of that run’s final total. A dilute outflow cell set the timestep. The model uses nonrelativistic fluid dynamics and an equilibrium photon contribution to the equation of state even in optically thin outflow. The reported screening yield integrates the main burn and its terminal plateau. Resolving late outflow speed and photon transport needs a different treatment. Shrinking the 2 m fine region from 56 m to 40 m and 32 m reduced run time, but the yield fell from 90.41% to 86.60% and 79.72%. A faster mesh needs to preserve the compression and burn.
The coast check uses 1 cm radius grains at 1,000 kg/m³, a path up to 1.1 × 109 m, and a nonseed release speed of at least 280 km/s. With a 25 V absolute tail, an effective electric field up to 1 V/m, and a magnetic field up to 1 µT, the absolute drift bound is 6.6 cm. The modeled allowance is 0.1 m. The effective electric field includes any beam self-field in the coast envelope. The separately supplied seed has its own trajectory requirement.
For a metrology test, estimate the common clock shift and remove it only if the seed and shot window are retimed with it. Fit the declared hemisphere and polar modes on mass-weighted channels. Report the amplitudes of those modes and the independent residual RMS, per transverse axis for pointing. Evaluate signed charge about its common mean and record the absolute tail separately. Use upper uncertainty bounds for the measured magnitudes and lower bounds for effective group count and minimum speed. The simulation uses a common axis for the coherent modes. Physical qualification also needs measured launcher accuracy, nonaligned and higher angular modes, and resolved pellet breakup and merger.
Executable acceptance test · Declared inputs and case matrix · Test output · Run histories, source, and replay instructions.
Code, inputs and records
Download the assembly runner, the archived B1 pressure trajectory, the flight results and the handoff summary. The run archive contains the starting profiles and completed result records. Download the raw traces for the merger cases, arrival-error cases, or extended burn. SHA-256 hashes identify the trajectory, results and solver inputs. Doubling the draw count changes the mass above 1 g/cm³ by at most one percentage point across the cases at the bulk crossing and plasma handoff. 320,000-draw check.
Use Python 3.10 and extract the plasma reproduction package alongside the assembly runner and pressure trajectory. From that directory, with the package’s code/ and data/ directories available, run
python -m pip install -r code/requirements.txt
python assemble.py?v=243f5ea7c649 B1-backward-feos128-refined.npz --out assembly
python assemble.py?v=243f5ea7c649 B1-backward-feos128-refined.npz --out assembly \
--research-root code --baseline data/n15.json
python correlated.py?v=cca85f6a93c4 B1-backward-feos128-refined.npz \
--research-root code --baseline data/n15.json \
--mode timing-us --amplitude 3 --out timing-3.json
python correlated.py?v=cca85f6a93c4 B1-backward-feos128-refined.npz \
--research-root code --baseline data/n15.json --zones 256 \
--mode all-six-late-heavy --amplitude 1 --out combined-256.json