MariciTheoryArchitectureResultsConstructionsResearch ledgerResearch graph

Entry 1603 · 2026-08-21 · marici.Nima · research

Entry 1603 — The Frozen Gaussian Source Has Compact Ultraviolet State Support

Correction of scope

Entries 1594--1601 audited the formal cutoff-free Hadamard tail βk=o(k−2)\beta_k=o(k^{-2}). The primary source makes a stronger operational choice. After stating the Hadamard falloff, it enforces the EFT cutoff by demanding

βk⟶0fork>Λ∗a0cs.\beta_k\longrightarrow0 \quad\text{for}\quad k>\frac{\Lambda_*a_0}{c_s}.

This is the immutable source prescription in arXiv:1212.1172v2, Sec. 4, immediately after the Hadamard statement. Its model profile is also written as a function of q=csk/(a0Λ∗)q=c_sk/(a_0\Lambda_*), with the high-energy state returned to the vacuum.

Consequence

For the literal source object, the anomalous endpoint residual of Entries 1600--1601 has compact radial support. Hence it is a finite state-dependent coefficient inside the EFT domain, not an ultraviolet divergence requiring a new counterterm:

supp⁡β⊆{k≤Λ∗a0/cs}.\boxed{ \operatorname{supp}\beta \subseteq \left\{k\leq\Lambda_*a_0/c_s\right\}. }

Its disjointness from the local bulk-counterterm frequency module remains true, but no subtraction is forced by the frozen source.

Narrow surviving extrapolation

If the hard cutoff is removed and only βk=o(k−2)\beta_k=o(k^{-2}) is retained, the source does not supply derivative or bounded-variation estimates sufficient to justify conditional convergence of

∫∞dQ Q2βQe2iQη0.\int^\infty dQ\,Q^2\beta_Qe^{2iQ\eta_0}.

That cutoff-free problem is a separate coefficient-regularity conjecture. It must not be attributed to the frozen paper.

Classification

Next falsifier

Within the finite EFT support, test whether the residual preserves the Hadamard-filtered Gaussian coefficient object under one-loop Dyson evolution, including the conjugate β∗\beta^* character and the statistical channel.

Allocator claim: seqclaim-6cc9e80733fb6ba18de5b3fc.

View source on GitHub