Skip to content
Shadow Theory

The coherent face and its budget

Reversible flow, resonant compression and the stability of declared coherence symbols.

Coherence keeps the modes of the carrier and changes how they evolve. The choice of symbol controls the phase rates, the resonances and the stability of the resulting budget.

6. The Coherent / Reversible Face

The coherent face uses a bounded real symbol hh of KK.

6.1 Admissible coherence symbols

Definition 6.1. Admissible coherence symbol.

An admissible coherence symbol is a Borel h:[0,∞)→Rh:[0,\infty)\to\mathbb R with: hh real-valued; hh bounded; h(0)=0h(0)=0; hh nondecreasing; if K≠0K\ne0, the coherent response is required to be nontrivial, h(K)≠0h(K)\ne0. The zero carrier is admitted as a degenerate case.

Example 6.2. Default symbol.

For λ0>0\lambda_0>0, hλ0(λ)=2πarctan⁡(λ/λ0)h_{\lambda_0}(\lambda) = \tfrac{2}{\pi}\arctan(\lambda/\lambda_0) is real, bounded by 11, satisfies h(0)=0h(0)=0, and is strictly increasing.

6.2 The coherence generator and unitary flow

Definition 6.3. Coherence generator.

C:=h(K)C := h(K). Bounded selfadjoint with ∥C∥≤∥h∥∞\|C\|\le\|h\|_\infty and C∈W∗(K)C \in W^*(K).

Definition 6.4. Reversible flow.

Ut:=e−itC=e−it h(K)U_t := e^{-itC} = e^{-it\,h(K)}, t∈Rt \in \mathbb R. A strongly continuous one-parameter unitary group.

Proposition 6.5. Mode preservation.

If Kej=λjejK e_j = \lambda_j e_j, then Cej=h(λj)ejC e_j = h(\lambda_j) e_j and Utej=e−ith(λj)ejU_t e_j = e^{-it h(\lambda_j)} e_j. Every spectral subspace of KK is UtU_t-invariant.

6.3 The reversible Fejér compression

Definition 6.6. Coherence Fejér operator.

ΓT(C):=(1−e−iTC)/(iTC)=1T∫0Te−itC dt\Gamma_T(C) := (1 - e^{-iTC})/(iTC) = \tfrac{1}{T}\int_0^T e^{-itC}\,dt, identity on ker⁡C\ker C.

Definition 6.7. Coherent compression.

CT:=ΓT∗ΓT=2(1−cos⁡(TC))T2C2\mathbf C_T := \Gamma_T^*\Gamma_T = \tfrac{2(1-\cos(TC))}{T^2 C^2}, with eigenvalues sinc2(Tμ/2)\mathop{\mathrm{sinc}}^2(T\mu/2) on (ker⁡C)⊥(\ker C)^\perp. Here sincx:=sin⁡x/x\mathop{\mathrm{sinc}}x := \sin x / x, with sinc0:=1\mathop{\mathrm{sinc}}0 := 1.

6.4 The coherence budget

Definition 6.8. Coherence compression budget: full convention.

Let

qT(μ)  :=  sinc2(Tμ/2),sincx:=sin⁡xx,sinc0:=1.q_T(\mu) \;:=\; \mathop{\mathrm{sinc}}^2(T\mu/2),\qquad \mathop{\mathrm{sinc}}x := \tfrac{\sin x}{x},\quad \mathop{\mathrm{sinc}}0 := 1.

For C=h(K)C = h(K), define the resonant zero-mode subspace

RT(C)  :=  EC ⁣({μ∈σ(C)∖{0}:Tμ2∈πZ∖{0}})H.\mathcal R_T(C) \;:=\; E_C\!\left(\left\{\mu\in\mathop{\mathrm{\sigma}}(C)\setminus\{0\}:\tfrac{T\mu}{2}\in\pi\mathbb Z\setminus\{0\}\right\}\right)\mathsf H.

Equivalently, RT(C)\mathcal R_T(C) is the part of (ker⁡C)⊥(\ker C)^\perp annihilated by CT=ΓT(C)∗ΓT(C)\mathbf C_T = \Gamma_T(C)^*\Gamma_T(C).

Under (D3), the full coherence budget convention used in this paper is

BT(C)  :=  {−log⁡det⁡∗ ⁣(CT∣(ker⁡C)⊥),RT(C)={0},+∞,RT(C)≠{0}.\mathfrak B_T(C) \;:=\; \begin{cases} -\log\det\nolimits_*\!\big(\mathbf C_T\big|_{(\ker C)^\perp}\big), & \mathcal R_T(C) = \{0\},\\[0.3em] +\infty, & \mathcal R_T(C)\ne\{0\}. \end{cases}

Thus any nonzero coherence mode annihilated by CT\mathbf C_T produces infinite full compression loss. If (ker⁡C)⊥={0}(\ker C)^\perp = \{0\}, we set BT(C)=0\mathfrak B_T(C) = 0.

In finite rank, with nonzero coherence eigenvalues μj\mu_j, this reads

BT(C)  =  {−∑jlog⁡sinc2(Tμj/2),sinc(Tμj/2)≠0 for every j,+∞,sinc(Tμj/2)=0 for at least one j.\mathfrak B_T(C) \;=\; \begin{cases} -\displaystyle\sum_j \log\mathop{\mathrm{sinc}}^2(T\mu_j/2), & \mathop{\mathrm{sinc}}(T\mu_j/2)\ne 0\text{ for every }j,\\[0.6em] +\infty, & \mathop{\mathrm{sinc}}(T\mu_j/2) = 0\text{ for at least one }j. \end{cases}

When no confusion is possible we write BT(K)\mathfrak B_T(K) for BT(h(K))\mathfrak B_T(h(K)).

Remark 6.9. Full versus reduced resonant convention.

The coherent Fejér compression differs from the dissipative Fejér transmission in one essential respect. The dissipative eigenvalues

fT(λ)  =  1−e−TλTλf_T(\lambda) \;=\; \frac{1-e^{-T\lambda}}{T\lambda}

are strictly positive for λ>0\lambda>0, whereas the coherent compression eigenvalues

qT(μ)  =  sinc2(Tμ/2)q_T(\mu) \;=\; \mathop{\mathrm{sinc}}^2(T\mu/2)

vanish at the resonant values Tμ/2∈πZ∖{0}T\mu/2 \in \pi\mathbb Z\setminus\{0\}. There are therefore two distinct conventions.

  1. Full determinant convention. The convention used in the paper unless explicitly stated otherwise. The kernel of CC is ignored because ΓT(C)\Gamma_T(C) is the identity there and contributes no loss. Any resonant zero mode inside (ker⁡C)⊥(\ker C)^\perp is not ignored: it makes the full budget equal to +∞+\infty.

  2. Reduced resonant pseudodeterminant convention. One may instead remove both ker⁡C\ker C and the resonant zero modes:

    HT,red  :=  (ker⁡C⊕RT(C))⊥,\mathsf H_{T,\mathrm{red}} \;:=\; \big(\ker C \oplus \mathcal R_T(C)\big)^\perp,

    and define

    BT,red(C)  :=  −log⁡det⁡∗ ⁣(CT∣HT,red),\mathfrak B_{T,\mathrm{red}}(C) \;:=\; -\log\det\nolimits_*\!\big(\mathbf C_T\big|_{\mathsf H_{T,\mathrm{red}}}\big),

    when the determinant on the reduced subspace is defined. In finite rank this is the product over the remaining strictly positive compression eigenvalues only:

    BT,red(C)  =  −∑μj≠0sinc(Tμj/2)≠0log⁡sinc2(Tμj/2).\mathfrak B_{T,\mathrm{red}}(C) \;=\; -\sum_{\substack{\mu_j\ne 0\\ \mathop{\mathrm{sinc}}(T\mu_j/2)\ne 0}} \log\mathop{\mathrm{sinc}}^2(T\mu_j/2).

On nonresonant scale windows, RT(C)={0}\mathcal R_T(C) = \{0\}, so the full and reduced conventions agree up to the harmless omission of ker⁡C\ker C, where the compression is the identity. At resonant scales they do not agree: the reduced budget is a pseudodeterminant after deleting total-loss modes, and is therefore not the full compression-loss budget.

Proposition 6.10. Properties of BT\mathfrak B_T.

For T>0T>0 the full budget satisfies BT(K)≥0\mathfrak B_T(K)\ge0, with equality if and only if C=0C=0 on the retained space. The compression CT\mathbf C_T is a Borel function of KK, and BT(UKU∗)=BT(K)\mathfrak B_T(UKU^*)=\mathfrak B_T(K) whenever the determinant convention is transported by UU. Hence the budget is basis-independent.

Proof. For real μ\mu, 0≤qT(μ)≤10\le q_T(\mu)\le1, with qT(μ)=1q_T(\mu)=1 exactly when μ=0\mu=0, since ∣sin⁡x∣<∣x∣|\sin x|<|x| for x≠0x\ne0. Each nonzero retained mode therefore contributes a strictly positive loss, or infinite loss at resonance. In the Fredholm case, the positive compact operator I−CTI-\mathbf C_T has a spectral product with the same conclusion: zero loss forces every factor to be one. Functional calculus and unitary invariance of the determinant prove the remaining assertions. ◻

6.5 Symbol-class stability

Proposition 6.11. Borel-symbol stability.

Let K≥0K\ge 0 be selfadjoint, and let h1,h2:[0,∞)→Rh_1, h_2 : [0,\infty)\to\mathbb R be bounded Borel symbols. Then

∥h1(K)−h2(K)∥  ≤  sup⁡λ∈σ(K)∣h1(λ)−h2(λ)∣.\|h_1(K) - h_2(K)\| \;\le\; \sup_{\lambda\in\mathop{\mathrm{\sigma}}(K)}|h_1(\lambda)-h_2(\lambda)|.

For a spectral window [0,R][0,R],

∥(h1(K)−h2(K)) EK([0,R])∥  ≤  ∥h1−h2∥∞,[0,R],\big\|(h_1(K) - h_2(K))\,E_K([0,R])\big\| \;\le\; \|h_1 - h_2\|_{\infty,[0,R]},

where the right-hand side denotes the supremum over [0,R]∩σ(K)[0,R]\cap\mathop{\mathrm{\sigma}}(K), or over [0,R][0,R] if the symbols are declared on the whole interval.

Proof. Let φ=h1−h2\varphi = h_1 - h_2. By the bounded Borel functional calculus, h1(K)−h2(K)=φ(K)h_1(K) - h_2(K) = \varphi(K) and ∥φ(K)∥≤∥φ∥L∞(σ(K))\|\varphi(K)\| \le \|\varphi\|_{L^\infty(\mathop{\mathrm{\sigma}}(K))}. Moreover, (h1(K)−h2(K)) EK([0,R])=(φ 1[0,R])(K)(h_1(K) - h_2(K))\,E_K([0,R]) = (\varphi\,\mathbf 1_{[0,R]})(K), so the same functional-calculus norm bound gives the windowed estimate. ◻

Lemma 6.12. Finite-rank nonresonant determinant-budget stability.

Let PR:=EK([0,R])P_R := E_K([0,R]) have finite rank r<∞r < \infty. Fix T>0T>0 and define qT(u):=sinc2(Tu/2)q_T(u) := \mathop{\mathrm{sinc}}^2(Tu/2). For a real bounded Borel symbol hh, define the finite-window coherence budget

BT,R(h)  :=  −∑λ∈σ(K∣PRH)m(λ)log⁡qT(h(λ)),\mathcal B_{T,R}(h) \;:=\; -\sum_{\lambda\in\mathop{\mathrm{\sigma}}(K|_{P_R\mathsf H})}^{m(\lambda)}\log q_T(h(\lambda)),

whenever all terms are positive. Here eigenvalues are counted with multiplicity, and modes with h(λ)=0h(\lambda) = 0 contribute −log⁡qT(0)=0-\log q_T(0) = 0.

Let h1,h2h_1, h_2 be real bounded Borel symbols. Set

M  :=  max⁡{∥h1∥∞,[0,R],∥h2∥∞,[0,R]},LT,M  :=  sup⁡∣u∣≤M∣qT′(u)∣.M \;:=\; \max\{\|h_1\|_{\infty,[0,R]}, \|h_2\|_{\infty,[0,R]}\},\qquad L_{T,M} \;:=\; \sup_{|u|\le M}|q_T'(u)|.

Assume there is δ>0\delta > 0 such that qT(hℓ(λ))∈[δ,1]q_T(h_\ell(\lambda)) \in [\delta,1] for every λ∈σ(K∣PRH)\lambda\in\mathop{\mathrm{\sigma}}(K|_{P_R\mathsf H}) and for ℓ=1,2\ell = 1,2. Then both finite-window budgets are full-convention nonresonant budgets on PRHP_R\mathsf H, and

∣BT,R(h1)−BT,R(h2)∣  ≤  r LT,Mδ ∥h1−h2∥∞,[0,R].\big|\mathcal B_{T,R}(h_1) - \mathcal B_{T,R}(h_2)\big| \;\le\; \frac{r\,L_{T,M}}{\delta}\,\|h_1 - h_2\|_{\infty,[0,R]}.

Proof. For each eigenvalue λ\lambda in the finite window, put aℓ(λ):=qT(hℓ(λ))a_\ell(\lambda) := q_T(h_\ell(\lambda)) for ℓ=1,2\ell = 1,2. By hypothesis aℓ(λ)∈[δ,1]a_\ell(\lambda)\in[\delta,1]. Since x↦log⁡xx\mapsto \log x has derivative bounded by δ−1\delta^{-1} on [δ,1][\delta,1],

∣log⁡a1(λ)−log⁡a2(λ)∣  ≤  δ−1∣a1(λ)−a2(λ)∣.|\log a_1(\lambda) - \log a_2(\lambda)| \;\le\; \delta^{-1}|a_1(\lambda) - a_2(\lambda)|.

The function qTq_T is smooth, hence Lipschitz on [−M,M][-M,M] with Lipschitz constant LT,ML_{T,M}. Therefore

∣a1(λ)−a2(λ)∣  ≤  LT,M∣h1(λ)−h2(λ)∣  ≤  LT,M∥h1−h2∥∞,[0,R].|a_1(\lambda) - a_2(\lambda)| \;\le\; L_{T,M}|h_1(\lambda) - h_2(\lambda)| \;\le\; L_{T,M}\|h_1 - h_2\|_{\infty,[0,R]}.

Summing over the rr eigenvalues in the finite window gives the stated estimate. ◻

Corollary 6.13. Small perturbations stay nonresonant.

In the setting of Lemma 6.12, suppose hh satisfies qT(h(λ))≥δq_T(h(\lambda)) \ge \delta for every λ∈σ(K∣PRH)\lambda\in\mathop{\mathrm{\sigma}}(K|_{P_R\mathsf H}), and let kk be another real bounded Borel symbol with ε:=∥h−k∥∞,[0,R]≤1\varepsilon := \|h - k\|_{\infty,[0,R]} \le 1. Put Mh:=∥h∥∞,[0,R]M_h := \|h\|_{\infty,[0,R]} and LT,Mh+1:=sup⁡∣u∣≤Mh+1∣qT′(u)∣L_{T,M_h+1} := \sup_{|u|\le M_h+1}|q_T'(u)|. If LT,Mh+1 ε≤δ/2L_{T,M_h+1}\,\varepsilon \le \delta/2, then

qT(k(λ))  ≥  δ2on σ(K∣PRH),q_T(k(\lambda)) \;\ge\; \tfrac{\delta}{2} \qquad \text{on }\mathop{\mathrm{\sigma}}(K|_{P_R\mathsf H}),

and

∣BT,R(k)−BT,R(h)∣  ≤  2r LT,Mh+1δ ε.\big|\mathcal B_{T,R}(k) - \mathcal B_{T,R}(h)\big| \;\le\; \frac{2r\,L_{T,M_h+1}}{\delta}\,\varepsilon.

Remark 6.14. Scope of the stability statement.

Proposition 6.11 is a general bounded-functional-calculus estimate. Lemma 6.12 and Corollary 6.13 are deliberately finite-rank, nonresonant determinant statements. They do not assert an infinite-dimensional determinant-continuity theorem without additional Fredholm, heat-kernel, zeta, or finite-volume hypotheses.

6.6 Properties shared by the two faces

Lemma 6.15.

Cdom(K)⊆dom(K)C\mathop{\mathrm{dom}}(K)\subseteq\mathop{\mathrm{dom}}(K) and KC=CKKC=CK on dom(K)\mathop{\mathrm{dom}}(K); likewise KUt=UtKKU_t=U_tK on dom(K)\mathop{\mathrm{dom}}(K), [ΛT(K),Us]=0[\Lambda_T(K), U_s]=0 for all s,t,Ts,t,T. Both {e−tK}t≥0\{e^{-tK}\}_{t\ge 0} and {Us}s∈R\{U_s\}_{s\in\mathbb R} lie in W∗(K)W^*(K) and commute with each other.

Proof. All bounded members are Borel functions of KK; their commutation with KK includes preservation of its operator domain. The algebra W∗(K)W^*(K) is abelian. ◻

The carrier is canonical given its closed form and Hilbert metric. The coherent face is uniquely determined once the specific symbol hh is fixed; the admissible symbol class permits multiple choices.

Published source: The Canonical Spectral-Curvature Field: A Unified Operator Origin for Coherence, Dissipation, and Spectral Budgets. Read the original PDF alongside the web treatment.