Manuscript IV@1 · IV.tex · proposition · 351cb2d7
Proposition: Evaluated spectral remainders
Model 2: spectral removal at fixed number cutoff › Pair and number estimates · lines 1775–1873 · proof included or linked
IV@1#prop:rates, IV@1#eq:rates, IV@1#eq:Pmajor, IV@1#eq:Schur2, IV@1#eq:mixedasymptotic
The pair kernel is Hilbert–Schmidt, and defines a bounded compact operator . There are finite constants , independent of the choices of masks satisfying (eq:masks), such that
For , is not Hilbert–Schmidt; more precisely,
for all sufficiently large , with .
Proof. For pair signs the denominator in (eq:master2) contains . Put , and use . With arbitrarily large ,
The radial integral of the last factor is finite:
Its integral for is , and for is .
For the scalar UV tail split and . On the latter, (eq:Pmajor) and give a bound whose integral over is . On the former, spatial transfer decay gives , with arbitrarily rapid integrated decay. Since
the pair estimate follows.
For the mixed signs, discard the bounded inverse frequency factor, but retain transfer decay. For , Peetre's inequality gives both bounds
For ,
The low-momentum part is Hilbert–Schmidt and the high part is bounded by the Schur test. The low gravitational tail has squared integral . For or , the appropriate bound in (eq:Schur2) has both Schur integrals ; the remaining low-, high- part decays faster. This proves the operator-norm rate. Each is Hilbert–Schmidt, so its norm limit is compact.
For the last assertion, the squared mixed estimate gives
Integration over proves the upper bound .
For a lower bound write , with and in a fixed compact set. Uniformly there,
The function is the Fourier transform of the compactly supported integrable function ; hence it is real analytic. It is nonzero if , because . It therefore cannot vanish on the open set . Choose a nonzero value there and an with . The TT norm is strictly positive. Continuity gives open sets of positive measure in on which the limit in (eq:mixedasymptotic) is bounded below. For , both masks equal one when is large. The Jacobian is , so integrating the lower bound gives a quantity bounded below by . The same region without radial truncation proves that is not Hilbert–Schmidt.
Displayed formulas in this passage
IV@1#eq:rates · line 1781
IV@1#eq:Pmajor · line 1797
IV@1#eq:Schur2 · line 1825
unlabeled · line 1787
unlabeled · line 1803
unlabeled · line 1816
unlabeled · line 1832
unlabeled · line 1844
IV@1#eq:mixedasymptotic · line 1852