Pith. sign in
module module moderate

IndisputableMonolith.Geometry.ReggeActionNonlinearHessianProof

show as:
view Lean formalization →

Exact second directional variation of the full nonlinear Regge action along a line in vertex-potential space. Discrete-gravity and Regge-calculus workers cite it when the Hessian of the action, not just its linearization, is needed. The argument splits the action along a line into a canonical quadratic piece plus remainder, then discharges the Hessian identity from first-order tangency or near-zero linearization hypotheses.

claimAlong a line $t \mapsto q + t\,v$ in vertex-potential space, the nonlinear Regge action admits the exact second-directional-variation identity $S(q+tv)=S(q)+t\,DS_q(v)+\tfrac12 t^2 H_S(q)(v,v)+R(t)$, where the quadratic term is the directional Hessian and $R$ is the cubic-scale remainder. Equivalent formulations package the same fact as first-order tangency of the action derivative or as linearization of that derivative near zero.

background

Regge calculus replaces smooth spacetime by a simplicial complex whose geometry is carried by edge lengths (or dual vertex potentials). The nonlinear Regge action $S$ is built from deficit angles and volumes; both depend on the Cayley–Menger determinants and on $\arccos$ of dihedral data, so second derivatives are a long chain-rule expansion.

The parent module ReggeActionSecondVariation records the usable targets: a nonlinear second-variation identity and a cubic-remainder bound, with the heavy Cayley–Menger/$\arccos$ calculus left in named input structures until fully expanded. The present module supplies the exact directional-Hessian statement those targets presuppose.

Working objects are the restriction of $S$ to an affine line, its canonical quadratic part (the directional Hessian form), and the residual remainder along that line. Auxiliary calculus facts (eventual differentiability from $C^\infty$, second derivative of a constant shift) support the 1D reductions.

proof idea

The module is organized around a 1D split and two equivalent analytic interfaces.

First, actionAlongLine_canonical_split decomposes the restricted action into canonicalQuadraticAlongLine plus canonicalRemainderAlongLine. Differentiability lemmas (differentiableAt_eventually_of_contDiffAt_top, deriv_differentiableAt_of_contDiffAt_top, hasSecondDerivAt_const_add) justify passing derivatives through that split.

Two named targets then encode the same Hessian claim: ActionDerivativeFirstOrderTangencyTarget (the derivative of $S$ is tangent to its linear part at first order) and ActionDerivativeLinearizationNearZeroTarget (linearization of the action derivative near the base point). Bridge lemmas show each target implies the nonlinear directional Hessian, and that near-zero linearization yields first-order tangency. The headline NonlinearReggeDirectionalHessianTheorem packages the resulting exact second-variation identity.

why it matters in Recognition Science

Without an exact directional Hessian for the full nonlinear action, cubic remainder estimates cannot be stated cleanly. Downstream, ReggeActionCubicTaylorBound isolates the final local third-order Taylor bound in finite-dimensional vertex-potential space; its doc-comment says that bound is needed "after the nonlinear Hessian has been identified." This module is that identification step.

In the broader Recognition geometry stack, controlling the second variation of the discrete action is the bridge from kinematic simplex data to dynamical stability and to continuum limits. The module does not itself close the Cayley–Menger expansion; it freezes the Hessian claim so the cubic Taylor work can proceed against a fixed interface.

scope and limits

used by (1)

From the project-wide theorem graph. These declarations reference this one in their body.

depends on (1)

Lean names referenced from this declaration's body.

declarations in this module (142)

… and 62 more