Skip to content

input-format.md

Latest commit

 

History

History
76 lines (53 loc) · 5.07 KB

input-format.md

File metadata and controls

76 lines (53 loc) · 5.07 KB
Raw

Writing an input file

The --input to lewron run is either a path to a Markdown file or an inline prompt string. What you write determines the mode and the build depth, so it is worth knowing the shape of a good input.

Discovery mode: a model description

A discovery-mode input is a short Markdown file (one to three paragraphs) describing a new model. Include:

  1. What the model is — one or two sentences of physics context (the field content, the symmetry being extended).
  2. The tree-level potential, in LaTeX. This is the heart of the input.
  3. Field content and symmetries — name the fields and any relevant symmetries or charges.
  4. The zero-temperature vacuum — which fields get a VEV at $T=0$, and any restriction on the field ranges you care about.
  5. The analysis request — what observable you want and under which criterion.

Here is the shipped ALP example (examples/ALP.md), annotated:

The model extends the SM by a light pseudo-Nambu-Goldstone boson (ALP) $S$,    ← context
arising from the spontaneous breaking of a global U(1) symmetry at a scale f ≫ v
The tree-level potential is:
$$
V(H,S) = -(μ_H² - A f cos δ)|H|² + λ|H|⁴ + μ_S² f² (1 - cos(S/f))            ← the potential
         - A f (|H|² - v²) cos(S/f - δ),
$$
where H is the SM Higgs doublet. ... The Higgs parameters are fixed by         ← field content,
m_h = 125.20 GeV and v ≃ 174 GeV.                                                  fixed inputs

At T = 0, the Higgs acquires its standard VEV and ⟨S⟩ = 0.                      ← T=0 vacuum
We are interested in the case where 0 < δ < π/2.
The field value of S is considered within one period ...                       ← field-range scope

We are interested in a strongly first-order electroweak phase transition.      ← the request

The effective potential is computed at one-loop order with Parwani             ← optional convention
resummation. Bubble nucleation should be computed, with T_n defined by            hints
S_3(T_n)/T_n ≃ 140.

You do not have to pin every convention. LeWRON proposes background directions, parametrization, and resummation at checkpoints, and you confirm or adjust there. The conventions you do state (here, Parwani resummation and the nucleation criterion) are taken as given.

A note on the Higgs VEV. Be explicit about your normalization — whether $v \simeq 246$ GeV or $v \simeq 174$ GeV — as the ALP example does. The symbol $v$ is overloaded across the literature, and stating it avoids a class of avoidable sign/factor confusion in the derivation.

Reproduction mode: a paper plus a target

A reproduction-mode input is short — name the paper and the target figure or table, and note anything to skip. The shipped example (examples/2009.14295.md) is a single line:

Reproduce Table 1 of 2009.14295. Ignore the bubble wall velocity stuff.

From this, LeWRON classifies the run as reproduction mode, fetches the paper, and infers the paper's conventions (background directions, parametrization, renormalization, resummation) into a reconstruction plan you approve at bootstrap. You can pass this kind of request inline rather than as a file:

lewron run --input "Reproduce Table 1 of 2009.14295, neglecting the wall velocity."

How your wording sets the build depth

Bootstrap reads the request and picks a build preset. The cues:

  • Mentioning the nucleation temperature ($T_n$), a strongly first-order transition, or gravitational waves → the full build (one-loop + thermal). This is the common case.
  • "Vacuum stability at $T=0$" → T0_loop (renormalized $T=0$ potential, no thermal).
  • A thermal request that explicitly opts out of one-loop corrections → tree_thermal (you will be asked to confirm the skip first).
  • A purely symbolic request → tree (tree-level derivation only).

So the ALP example, by asking for a strongly first-order EWPT and nucleation, selects the full build and ends in the Explorer.

Tips

  • Write the potential in LaTeX exactly as you mean it; it is read literally.
  • State the symmetries and the $T=0$ vacuum explicitly — they drive the symmetry analysis and background-direction proposals.
  • If you have a specific resummation scheme or nucleation criterion in mind, say so; otherwise sensible defaults apply (Parwani resummation, $S_3/T \simeq 140$).
  • To skip the scalar parametrization — keep every scalar Lagrangian parameter as a free input instead of trading couplings for physical masses and mixings — just say so in the input (e.g. "skip the parametrization; use the Lagrangian parameters directly"). Bootstrap reads the request and the tree-level stage leaves the parametrization empty, deferring the choice of physical inputs to the Explorer. You can also request this later at a tree-level checkpoint.
  • Keep it concise. The input is the prompt that produced the run, and shorter, precise inputs reproduce more reliably.

See also: the shipped inputs under examples/ and the examples guide.