Skip to content
Shadow Theory

Appendix A 4 October 2026

Explicit estimates for the periodic apparatus

Reading position 32 of 37

A Explicit estimates for the periodic apparatus

This appendix supplies the polynomial and Gaussian calculations used in Part I. All parameter bounds refer to that part's finite rectangle. They do not enlarge the law class of Part II.

A.1 Polynomial derivatives and clock normalization

Writing z=1−2az=1-2a, direct differentiation gives

B9′(a)=630a4(1−a)4,B9′′(a)=3158z(1−z2)3,B9′′′(a)=−3154(1−z2)2(1−7z2). B_9'(a)=630a^4(1-a)^4,\quad B_9''(a)=\frac{315}{8}z(1-z^2)^3,\quad B_9'''(a)=-\frac{315}{4}(1-z^2)^2(1-7z^2).

The first maximum is 315/128315/128; the second occurs at ∣z∣=1/7|z|=1/\sqrt7 and is below 1010. For the third, the only interior stationary squared arguments are 00 and 3/73/7, giving a bound 315/4315/4. Consequently the writer displacement has derivative bounds

∥b′∥∞<1478,∥b′′∥∞<150200,∥b′′′∥∞<2.958×107. \|b'\|_\infty<1478,\qquad \|b''\|_\infty<150200, \qquad \|b'''\|_\infty<2.958\times10^7.

For the four envelope derivatives the exact integrals Jk=∫01(B9(k)(a))2 daJ_k=\int_0^1(B_9^{(k)}(a))^2\dd a are

(J1,J2,J3,J4)=(44102431,5040143,272160143,181440011). (J_1,J_2,J_3,J_4) =\left(\frac{4410}{2431},\frac{5040}{143}, \frac{272160}{143},\frac{1814400}{11}\right).

Also ∫Ac2=176818/230945=Zc\int A_c^2=176818/230945=Z_c. The change of variable on both transition intervals gives ∥φ(k)∥2=2 52k−1Jk/Zc\|\varphi^{(k)}\|^2=2\,5^{2k-1}J_k/Z_c. Squaring the four claimed upper bounds 6,112,4000,1900006,112,4000,190000 verifies all four inequalities using rational arithmetic alone.

A.2 Driven Gaussian jets

Let d=vλd=v\lambda, r=y−qr=y-q, and let uu denote the normalized coherent Gaussian in G=φ(ξ)e−it/(2μ)uG=\varphi(\xi)e^{-it/(2\mu)}u. The start of every characteristic in the envelope support precedes the drive. Thus the characteristic solution depends on t+ξ/dt+\xi/d, and differentiation in ξ\xi is time differentiation divided by dd on q,q′,ϑq,q',\vartheta. The forcing F=Ω2b+b′′\mathcal F=\Omega^2b+b'' and the error e=q−be=q-b obey

∣e∣≤∣1−λ2∣ 2∥b′∥∞λΩ<.001,∣e′∣≤∣1−λ2∣ 2∥b′∥∞λ<.07. |e|\le\frac{|1-\lambda^2|\,2\|b'\|_\infty}{\lambda\Omega}<.001, \qquad |e'|\le\frac{|1-\lambda^2|\,2\|b'\|_\infty}{\lambda}<.07.

Indeed the total variation of the single positive hump b′b' is 2∥b′∥∞2\|b'\|_\infty; integrate the sine and cosine representations of e,e′e,e' against b′′(λt+ξ/v)b''(\lambda t+\xi/v). It follows that ∣q∣<25|q|<25, ∣q′∣<1630|q'|<1630, ∣q′′∣<170000|q''|<170000, and ∣q′′′∣<36000000|q'''|<36000000, using q′′=b′′−Ω2eq''=b''-\Omega^2e and q′′′=λb′′′−Ω2e′q'''=\lambda b'''-\Omega^2e'. Moreover ∣F∣<421000|\mathcal F|<421000 and ∣F˙∣<5.24×107|\dot{\mathcal F}|<5.24\times10^7.

Define the complex coefficient a1a_1 and real coefficient β\beta by

a1=q′+iμq′′d,β=μ2d(qq′′−q′2+Fe). a_1=\frac{q'+i\mu q''}{d},\qquad \beta=\frac{\mu}{2d}(qq''-q'^2+\mathcal F e).

Direct differentiation, including the scalar compensation phase, gives

uξ=(a1r+iβ)u,uξξ=[a12r2+((a1)ξ+2iβa1)r+iβξ−a1qξ−β2]u,(a1)ξ=(q′′+iμq′′′)/d2,βξ=μ2d2(qq′′′−q′q′′+F˙e+Fe′).\begin{aligned}u_\xi&=(a_1r+i\beta)u,\\ u_{\xi\xi}&=\left[a_1^2r^2+ ((a_1)_\xi+2i\beta a_1)r+ i\beta_\xi-a_1q_\xi-\beta^2\right]u,\\ (a_1)_\xi&=(q''+i\mu q''')/d^2,\\ \beta_\xi&=\frac{\mu}{2d^2} (qq'''-q'q''+\dot{\mathcal F}e+\mathcal F e'). \end{aligned}

Since d≥126.99873d\ge126.99873, substitution of the preceding bounds yields

∣a1∣<19,∣(a1)ξ∣<25,∣β∣<273,∣βξ∣<367,∣qξ∣<13,μ∣q′∣<17. |a_1|<19,\quad |(a_1)_\xi|<25,\quad |\beta|<273, \quad |\beta_\xi|<367,\quad |q_\xi|<13, \quad \mu|q'|<17.

For example the numerator bounding β\beta is μ+(25⋅170000+16302+421000⋅.001)\mu_+(25\cdot170000+1630^2+421000\cdot.001); for βξ\beta_\xi it is μ+(25⋅36000000+1630⋅170000+5.24 107⋅.001+421000⋅.07)\mu_+(25\cdot36000000+1630\cdot170000 +5.24\,10^7\cdot.001+421000\cdot.07). The denominators are respectively 2d−2d_- and 2d−22d_-^2.

For the Gaussian measure ∣u∣2 dr|u|^2\dd r, ∥rku∥2=(2k−1)!!/2k\|r^k u\|^2=(2k-1)!!/2^k. The quadratic polynomial in uξξu_{\xi\xi} therefore has coefficient moduli bounded by 361,10399,75143361,10399,75143. Applying the triangle inequality to these three monomials and to their derivatives proves

uξuξξ∥⋅∥28783000∥r ⋅∥21163000∥py ⋅∥51101485000 \begin{array}{c|rr} &u_\xi&u_{\xi\xi}\\\hline \|\cdot\|&287&83000\\ \|r\,\cdot\|&211&63000\\ \|p_y\,\cdot\|&5110&1485000 \end{array}

as strict upper bounds. Here py(P(r)u)=[−iP′(r)+(ir+μq′)P(r)]up_y(P(r)u)=[-iP'(r)+(ir+\mu q')P(r)]u, which explains the last row without omitting the phase momentum. Leibniz' rule now gives

∥Gξξ∥<112+12(287)+83000=86556<110000,∥pyGξξ∥<112(18)+12(5110)+1485000<1.55 106,∥yGξξ∥<25(86556)+112/2+12(211)+63000<2.23 106.\begin{aligned}\|G_{\xi\xi}\|&<112+12(287)+83000=86556<110000,\\ \|p_yG_{\xi\xi}\|&<112(18)+12(5110)+1485000<1.55\,10^6,\\ \|yG_{\xi\xi}\|&<25(86556)+112/\sqrt2+12(211)+63000 <2.23\,10^6. \end{aligned}

These explicit calculations prove both weighted bounds used in Lemma 8.1.

A.3 Rotor enclosures and the quiet-prefix current

For completeness, an outward square-root operation on a nonnegative rational xx can be defined using integers alone. Set n=1032n=10^{32}, k=⌊⌊n2x⌋⌋k=\lfloor\sqrt{\lfloor n^2x\rfloor}\rfloor, and return k/nk/n if k2≥n2xk^2\ge n^2x, otherwise (k+1)/n(k+1)/n. This is an upper enclosure whose error is at most 10−3210^{-32}. Applying it to the four rational squares in Lemma 12.1 gives

d0<.006517177225775,ϵR<.000715486047289,d<.007232663273064,q<.003982026605165. \begin{aligned} d_0&<.006517177225775,&\quad \epsilon_R&<.000715486047289,\\ d&<.007232663273064,&q&<.003982026605165. \end{aligned}

The strict comparisons with D=.007233D=.007233 and Q=.003983Q=.003983 follow. The trigonometric estimates used there also have elementary rational checks: alternating Taylor sums through degrees 14,1614,16 bound cos⁡(1.3)\cos(1.3), and degrees 15,1715,17 bound sin⁡(1.3)\sin(1.3). They give cos⁡(1.3)>1/4\cos(1.3)>1/4, tan⁡(1.3)<4\tan(1.3)<4, and sec⁡(1.3)+tan⁡(1.3)<8\sec(1.3)+\tan(1.3)<8. Since the positive exponential Taylor sum through degree 1414 at 2.12.1 exceeds 88,

∫01.3sec⁡3t dt=12[sec⁡(1.3)tan⁡(1.3)+log⁡(sec⁡(1.3)+tan⁡(1.3))]<8. \int_0^{1.3}\sec^3 t\dd t =\tfrac12[\sec(1.3)\tan(1.3)+ \log(\sec(1.3)+\tan(1.3))]<8.

The same cosine upper sum at .998(1.22).998(1.22) gives cos⁡(.998(1.22))/2+.01<1/5\cos(.998(1.22))/2+.01<1/5, placing both compact waves strictly inside their respective soft-classifier plateaux.

The Sobolev constants in the quiet-prefix estimate require no private auxiliary result. For a normalized sector rotor rir_i with 0≤Vi≤giK/80\le V_i\le g_iK/8, energy and graph-norm conservation from the initial flat wave give

∥ri′∥2≤mgiK/4,∥ri′′∥≤mgiK/2. \|r_i'\|^2\le mg_iK/4,\qquad \|r_i''\|\le mg_iK/2.

Thus ∥ri′∥≤K/h\|r_i'\|\le K/h and ∥ri′′∥≤K2/h2\|r_i''\|\le K^2/h^2. On a circle of length two, ∥f∥∞2≤∥f∥2/2+2∥f∥∥f′∥\|f\|_\infty^2\le\|f\|^2/2+2\|f\|\|f'\|. Apply this first to rir_i and then to ri′r_i' to obtain ∥ρr∥∞≤3K/h\|\rho_r\|_\infty\le3K/h and ∥jr∥∞≤8K\|j_r\|_\infty\le8K. The factorized pretrigger wave errors then give

∥δjX∥1≤4hϵpre,∥δρ∥1≤3ϵpre, \|\delta j_X\|_1\le4h\epsilon_{\rm pre},\qquad \|\delta\rho\|_1\le3\epsilon_{\rm pre},

and hence the lifted-rank integrand is at most 36Kϵpre36K\epsilon_{\rm pre}. Integrating up to 1.221.22 is bounded by 120Kϵpre<6 10−18120K\epsilon_{\rm pre}<6\,10^{-18} as used in the main text.

A.4 Gaussian tails and the whole holding interval

After capture, every comparison centre is within .012.012 of its intended centre, including the offset z0z_0. For a>0a>0, integration by parts gives

∥1∣r∣≥aγ∥2≤e−a2aπ,∥1∣r∣≥arγ∥2≤e−a2π(a+12a). \|\one_{|r|\ge a}\gamma\|^2\le \frac{e^{-a^2}}{a\sqrt\pi},\qquad \|\one_{|r|\ge a}r\gamma\|^2\le \frac{e^{-a^2}}{\sqrt\pi}\left(a+\frac1{2a}\right).

Take a=7.98a=7.98. The positive Taylor sum of ea2e^{a^2} through degree 200200 exceeds 4 10274\,10^{27}. With μ∣q′∣<.000701\mu|q'|<.000701, these inequalities give comparison tail norms below 10−1410^{-14} and momentum norms below 10−1310^{-13} on all four holding bands. They also imply the weaker wrong-inner-region amplitude bound 10−1210^{-12} at capture.

For each inner/outer pair choose a smooth cutoff changing from zero to one across the intervening band of width two, with ∣χ′∣≤15/16|\chi'|\le15/16. Any path that starts in the inner region and leaves its outer region has cutoff variation at least one. Equivariance therefore bounds its wave probability by ∫tb3∫∣χ′(Z)jZ∣ dq dt\int_{t_b}^3\int |\chi'(Z)j_Z|\dd q\dd t. This is a band integral of absolute current; no pointwise boundary trace estimate is needed. Expanding the exact auxiliary wave as G+EG+E bounds its integrand by the comparison current plus (eA10−13+10−14pA+eApA)/μ−(e_A10^{-13}+10^{-14}p_A+e_Ap_A)/\mu_-. Summing the four cutoffs, multiplying by the actual-law cap 1616, and using 3−1.28=43/253-1.28=43/25 proves the displayed holding bound in Part I.