# Data dictionary

All numerical records are **synthetic model outputs**, exported from the unmodified `results/verification.json`. Values are not fitted material parameters, experimental observations, measured energies, physical feature sizes, or manufacturing yields. Each Hugging Face configuration has a single `data` split. No train/test partition or benchmark performance claim is implied. Floating values preserve Python's deterministic round-trip representation.

## `dose_frontier` — 36 rows

Source array: `dose_table`. Four targets `q` and nine comb orders `m` are included. The original array order is preserved.

| Column | Meaning |
|---|---|
| `m` | Integer number of equally spaced phase gates and ideal pitch-multiplication order, at least 2. |
| `q` | Required dimensionless committed fraction at each prescribed interstitial midpoint, strictly between 0 and 1. |
| `action` | Balanced dimensionless action per marking gate, `u_m(q)`, obtained by inverting `G_m(u)=q`. |
| `total_action` | Minimum total dimensionless marking action `U_m(q)=m*action` within the ideal exact-node class. |
| `ratio` | Dimensionless ratio `total_action/m^3`, approaching `c_q` as `m` increases. |

The conversion from total marking action to incident marking fluence requires a calibrated marking coefficient: `E_mark=total_action/sigma`. Reset/commit fluence, heat and duration are additional resources.

## `cubic_constants` — 4 rows

Source mapping: `cubic_constants`; rows sorted by numerical `q`.

| Column | Meaning |
|---|---|
| `q` | Fixed required midpoint fraction. |
| `c_q` | Numerically calculated positive coefficient solving the manuscript's infinite-product equation; limit of `U_m(q)/m^3`. |

These are floating/high-precision calculations, not interval-certified decimal enclosures.

## `diffusion` — 16 rows

Source array: `diffusion_examples`. Four comb orders and four hypothetical angular diffusivities are included. The normalized first-order marking rate is 1, the target ideal midpoint fraction is 0.5, and the modeled gate ordering is that of the original implementation.

| Column | Meaning |
|---|---|
| `m` | Comb order and number of patterned marking gates. |
| `diffusion` | Hypothetical normalized angular diffusivity `D_theta=D*k^2`; not an experimentally calibrated material value. |
| `action` | Per-gate dimensionless marking action chosen for ideal no-diffusion midpoint fraction 0.5. |
| `time` | Total normalized marking duration, equal to `m*action` at marking rate 1; excludes reset and commit times. |
| `max_dark` | Maximum calculated retained fraction among prescribed nominal dark nodes. |
| `min_bright` | Minimum calculated retained fraction among prescribed interstitial midpoints. |
| `max_dark_bound` | Maximum corresponding analytic protecting-gate upper bound used by the original discrete-model check. |
| `grid_error` | Largest profile difference between grids with `n=64*m` and `2*n=128*m`; numerical refinement evidence, not a rigorous continuum error bound. |
| `n` | Number of spatial points on the coarser periodic finite-difference grid. |

Fractions and actions are dimensionless; angular position is periodic. Diffusion may destroy exact dark zeros and m-fold symmetry. The data do not demonstrate fabricated-feature resolution.

## `hazard_compiler` — 4 rows

Source array: `hazard_examples`. All rows concern the same specified three-product synthetic 3D hazard target evaluated on a `24^3` grid. Each product has three fringe factors. Repeated passes are evaluated through their exact aggregate formula in software; reported counts charge all physical passes and gates.

| Column | Meaning |
|---|---|
| `error` | Maximum observed conversion-fraction error over the supplied `24^3` grid. |
| `hazard_error` | Maximum observed hazard-field error over that grid. |
| `bound` | Conservative whole-domain conversion-error envelope from the manuscript's hazard bound using the analytic target-hazard envelope `H_max<=2.56`; it may exceed the trivial fraction-error bound 1. |
| `grid_bound` | Error-bound expression using grid maxima of the target hazard and per-pass survival; not a whole-domain certificate. |
| `passes` | Number of repeated serial passes required by the constructive program. |
| `factor_gates` | Total patterned marking-gate count, three per pass. |
| `primitive_ideal_maps` | Count of ideal marking/filter/commit maps, seven per pass in this example. |
| `mark_action` | Total dimensionless marking action, `action*factor_gates`. |
| `action` | Common dimensionless weak marking action per fringe factor. |

The existence theorem is separate from this finite example. Smaller error can have rapidly growing resource cost; the comb's cubic law is not a generic 3D efficiency bound.

## `claims` — 12 records

Format: UTF-8 JSON Lines, one object per line. Every object has the same seven string fields. These are editorial statements of scope grounded in the manuscript, not additional measured data, benchmark labels or machine-verified proofs.

| Field | Meaning |
|---|---|
| `claim_id` | Stable packaging identifier `C01` through `C12`. |
| `title` | Short descriptive title. |
| `statement` | Claim or release-status statement, with its intended domain. |
| `status` | Evidence category, such as conditional analytic result or synthetic simulation. |
| `assumptions` | Principal conditions necessary for interpreting the statement. |
| `evidence` | Locations of derivations, finite checks or provenance supporting that scope statement. |
| `limitation` | Explicit boundary preventing a stronger interpretation. |

A readable rendering appears in `CLAIM_LEDGER.md`. The source of truth for mathematical proofs remains the manuscript. Original full ledgers and provenance remain in `results/` and `source/`.