Skip to content
Shadow Theory

Sealed or Leaky Section 7

Conservation of complete information and operational sealing

Section 8 of 17

7 Conservation of complete information and operational sealing

For a class F\mathcal F of measurable tests f:X→[0,1]f:X\to[0,1], define

dF(P,Q)=sup⁡f∈F∣∫f d(P−Q)∣. d_{\mathcal F}(P,Q) =\sup_{f\in\mathcal F}\left|\int f\,d(P-Q)\right|.

Complete measurable access gives total variation. If the observer only reads π:X→Y\pi:X\to Y, the class of tests g∘πg\circ\pi, with all measurable 0≤g≤10\le g\le1, gives TV(π∗P,π∗Q)\TV(\pi_*P,\pi_*Q). For more restricted apparatus F\mathcal F may be smaller still.

Theorem 7.1 (Transport of accessible distinguishability)

Status: Proved.

For a measurable evolution Φ:X0→X1\Phi:X_0\to X_1 and an accessible test class F1\mathcal F_1,

dF1(Φ∗P,Φ∗Q)=dΦ∗F1(P,Q),Φ∗F1={f∘Φ:f∈F1}. d_{\mathcal F_1}(\Phi_*P,\Phi_*Q) =d_{\Phi^*\mathcal F_1}(P,Q),\qquad \Phi^*\mathcal F_1=\{f\circ\Phi:f\in\mathcal F_1\}.

Consequently Φ∗F1=F0\Phi^*\mathcal F_1=\mathcal F_0 ensures conservation of accessible distinguishability; inclusion in F0\mathcal F_0 ensures contraction, and reverse inclusion ensures nondecrease. For a bimeasurable bijection with complete measurable access,

TV(Φ∗P,Φ∗Q)=TV(P,Q). \TV(\Phi_*P,\Phi_*Q)=\TV(P,Q).

No corresponding conservation follows for a restricted observation class from invertibility alone.

Proof

For each test, change of variables gives ∫f d(Φ∗P−Φ∗Q)=∫(f∘Φ) d(P−Q)\int f\,d(\Phi_*P-\Phi_*Q)=\int(f\circ\Phi)\,d(P-Q). Taking the relevant suprema proves the identity and inclusions. A bimeasurable bijection permutes the class of all measurable bounded tests, proving the total-variation identity.

□

A finite reversible counterexample.

Let X={0,1}2X=\{0,1\}^2, observe π(x,y)=x\pi(x,y)=x, and let Φ(x,y)=(y,x)\Phi(x,y)=(y,x). Let UU be the fair-bit law and put Pb=δb⊗UP_b=\delta_b\otimes U, b=0,1b=0,1. Initially the accessible laws are δ0\delta_0 and δ1\delta_1, at total variation one. After the swap, Φ∗Pb=U⊗δb\Phi_*P_b=U\otimes\delta_b, and the accessible laws coincide exactly, although the complete laws remain at total variation one. Thus reversible dynamics can create exact snapshot sealing of a preparation family. Applying the swap again restores the distinction.

Persistent forward sealing under reversible dynamics.

The distinction is not confined to a one-time coincidence. Take X={0,1}ZX=\{0,1\}^{\mathbb Z} with its product sigma-algebra, fair product equilibrium EE, and the invertible shift (Φx)j=xj−1(\Phi x)_j=x_{j-1}. The observer reads coordinate zero and may choose future observation times but cannot reverse the shift or read other coordinates. For ∣a∣≤1|a|\le1, let

dPadE(x)=1+a(2x0−1). \frac{dP_a}{dE}(x)=1+a(2x_0-1).

Initially the observed bit differs from equilibrium by ∣a∣/2|a|/2 in total variation. For every integer n≥1n\ge1, the entire future output sequence is (x−n,x−n−1,…)(x_{-n},x_{-n-1},\ldots), whose law under PaP_a is the same fair product law as under EE. Every allowed future adaptive observation therefore has the same law. Nevertheless TV(Φ∗nPa,E)=∣a∣/2\TV(\Phi_*^nP_a,E)=|a|/2 at every time. The omitted distinction has moved into an inaccessible coordinate, and exact forward operational sealing coexists with reversible equilibrium-preserving source dynamics.

What retained records prevent.

If an observer retains the earlier record, the cumulative record at a later time has that record as a measurable marginal. Data processing then gives

TV(Pearlier record,Qearlier record)≤TV(Pcomplete retained record,Qcomplete retained record). \TV(P_{\text{earlier record}},Q_{\text{earlier record}}) \le \TV(P_{\text{complete retained record}},Q_{\text{complete retained record}}).

The examples concern future access after the original distinction has been discarded or never archived. They do not erase an actually retained record. Likewise, if implementing Φ−1\Phi^{-1} is an admissible operation, the original tests can be recovered and complete operational sealing of their distinction is impossible. Mathematical existence of an inverse is not physical access to that inverse.

For a Bohmian flow that is bimeasurably invertible on an equilibrium full-measure invariant domain, and a nonequilibrium initial law P0≪E0P_0\ll E_0, equivariance gives TV(Pt,Et)=TV(P0,E0)\TV(P_t,E_t)=\TV(P_0,E_0). This conserves fine-grained nonequilibrium; it neither forbids observable relaxation nor proves any record is able to reveal the surviving difference. Existence and equivariance alone must be supplemented by the flow's uniqueness and measurable inverse on the relevant domain to invoke the equality. The distinction between fine-grained conservation and coarse-grained relaxation is standard in pilot-wave relaxation analyses. In particular, invertible dynamics can create operational sealing; Theorem 7.1 states the exact access condition.