Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated

A finite-state actual-color framework for fixed translation covers


These are exact statements about all ternary relations, not selected lift templates. They apply to one fixed integer voltage assignment. No density-based lower bound or unbounded-color construction follows yet. No automaton enumeration has been run.

1. Symbol and finite residual states

Fix an integer matrix polynomial

A(z)=∑g=0GAgzg.A(z)=\sum_{g=0}^G A_gz^g.

For ordinary covers each nonzero entry of the binary base is one monomial, and the base columns are distinct and nonzero. Infinite column labels are V×ZV\times\mathbb Z; cyclic labels at degree mm are V×Z/mZV\times\mathbb Z/m\mathbb Z. An integer coefficient sequence ftf_t is a relation exactly when

∑g=0GAgft−g=0at every row time t.(1)\sum_{g=0}^G A_g f_{t-g}=0\quad\hbox{at every row time }t. \tag{1}

For ternary ftf_t, just before processing time tt store the pending row contributions r=(r0,…,rG−1)r=(r_0,\ldots,r_{G-1}). A transition labeled f∈{0,±1}Vf\in\{0,\pm1\}^V is allowed when

r0+A0f=0,rk′=rk+1+Ak+1f(0≤k<G),rG=0.(2)r_0+A_0f=0,\qquad r'_k=r_{k+1}+A_{k+1}f\quad(0\leq k<G),\quad r_G=0. \tag{2}

Use the finite state box

S=∏x∈X0≤k<G([−Bx,k,Bx,k]∩Z),Bx,k=∑g=k+1G∑v∣Ag(x,v)∣,N=∣S∣.(3)S=\prod_{\substack{x\in X\\0\leq k<G}} ([-B_{x,k},B_{x,k}]\cap\mathbb Z),\qquad B_{x,k}=\sum_{g=k+1}^G\sum_v|A_g(x,v)|, \quad N=|S|. \tag{3}

Every actual residual belongs to this box. If G=0G=0, use a single state, with transition condition A0f=0A_0f=0. Flag a transition by whether f≠0f\ne0.

Exact correspondences.

  • Finite-support ternary relations are precisely paths from state zero back to zero with at least one nonzero flag.
  • Cyclic degree-mm ternary relations are precisely such nonzero closed paths of length mm, starting at any state.

The first statement follows by accumulating pending row sums in (1). For the second, extend a closed path periodically. Repeated substitution in (2) forgets the initial state after GG steps, so the state is exactly the contribution of its previous GG coefficient vectors. There are no spurious periodic states, and (1) holds periodically.

2. Infinite proper colorings periodicize

Fix a palette of qq colors. A letter σ∈[q]V\sigma\in[q]^V specifies the colors at one sheet. For color cc, allow only transitions with fv=0f_v=0 if σv≠c\sigma_v\ne c.

Track Uc⊆SU_c\subseteq S, the endpoints of all nonzero allowed paths begun at state zero at any earlier time. Its update is

Uc′=Post⁡σ,c(Uc) ∪ {endpoints of nonzero allowed transitions from 0}.(4)U'_c=\operatorname{Post}_{\sigma,c}(U_c) \ \cup\ \{\text{endpoints of nonzero allowed transitions from }0\}. \tag{4}

Reject a word as soon as 0∈Uc′0\in U'_c for any cc. Otherwise the tracked sets avoid zero. Starting from all sets empty, this is a deterministic safety automaton with at most

K=2q(N−1)(5)K=2^{q(N-1)} \tag{5}

safe states. Its accepted finite words are exactly proper colorings of the corresponding finite column intervals. A finite relation contained in the interval must have zero residual immediately after its last supported column, since later zero coefficients cannot cancel a nonzero pending row sum.

Periodicization theorem. If the infinite configuration has a proper qq-coloring, it has one periodic in the sheet coordinate with period at most KK.

Indeed its positive half-line is an infinite safe path and contains a directed cycle of length at most KK. Repeat that cycle's word from the empty tuple. Each update (4) is monotone, and the empty tuple is contained in the cycle's initial tuple, so every repetition is safe. Its bi-infinite periodic extension is proper: a finite witness anywhere would translate into the safe positive half-line.

Finite-obstruction corollary. If the infinite family has no proper qq-coloring, the interval of KK sheets is already uncolorable. Otherwise its K+1K+1 visited states would contain a repeat and give the periodic coloring above. For ordinary covers, this obstruction projects to every cyclic degree m≥Km\geq K: column labels stay distinct, and identifying rows preserves the ternary relation.

3. An exact finite decision problem for cyclic degrees

For each letter σ\sigma and color cc, form an N×NN\times N transfer matrix. Its entries are subsets of {0,1}\{0,1\}, recording whether paths with zero-only or nonzero coefficient flag exist. Concatenation combines flags by logical OR and takes unions over intermediate states. There are at most

D=4qN2(6)D=4^{qN^2} \tag{6}

possible qq-tuples of these matrices. A cyclic word is proper precisely when its product has no diagonal entry containing flag one, in any color, by the closed-path correspondence.

Thus the proper cyclic coloring words form a regular language. To track only lengths, let RjR_j be the set of all transfer tuples realizable by length-jj words. The evolution Rj↦Rj+1R_j\mapsto R_{j+1} is deterministic on at most 2D2^D subsets. Hence the set of degrees admitting a proper qq-coloring is ultimately periodic, with preperiod plus period at most 2D2^D. If any degree is uncolorable, one is uncolorable by that bound.

This is finite and exact for the fixed symbol A(z)A(z). It is not a practical search bound, and no computation of these state spaces is claimed. It does not decide a supremum over unbounded voltage assignments.

4. Wrapped relations are real, not an automaton artifact

An infinite proper coloring prohibits nonzero zero-to-zero paths. It need not prohibit nonzero cycles at other residual states. For example,

A(z)=(100z110001100011),det⁡A(z)=1−z.A(z)= \begin{pmatrix} 1&0&0&z\\ 1&1&0&0\\ 0&1&1&0\\ 0&0&1&1 \end{pmatrix}, \qquad \det A(z)=1-z.

This is an ordinary four-cycle incidence symbol with one shifted incidence. Its infinite cover is an infinite path, so every finite subfamily is rationally independent. Every cyclic cover is an even cycle and has a nonzero alternating ternary relation. Thus its infinite coloring number is one, while every finite cyclic coloring number is two. Periodicization alone does not remove wrapped relations.

5. A sharp factor-two bridge

Let χ∞\chi_\infty be the dissociated coloring number of the infinite ordinary translation cover. It is finite, since coloring by base type works. For every positive m≥2Gm\geq2G,

χ1(A~m)≤2χ∞.(7)\chi_1(\widetilde A_m)\leq2\chi_\infty. \tag{7}

Split the cyclic sheets into two consecutive arcs of lengths ⌊m/2⌋,⌈m/2⌉\lfloor m/2\rfloor,\lceil m/2\rceil. Each length LL is at most m−Gm-G, so its columns and their supporting rows, spanning L+GL+G consecutive sheet positions, unroll injectively into the infinite cover. Use disjoint copies of a χ∞\chi_\infty-palette on the two arcs.

The finite-obstruction corollary gives the eventual opposite bound:

χ∞≤χ1(A~m)≤2χ∞for all sufficiently large m.(8)\boxed{\chi_\infty\leq\chi_1(\widetilde A_m) \leq2\chi_\infty \quad\hbox{for all sufficiently large }m.} \tag{8}

The preceding four-cycle attains the factor two. For a growing family of fixed symbols, unbounded infinite coloring would therefore give the required unbounded finite cover coloring. The missing theorem is still a lower bound for χ∞\chi_\infty under an extraction-compatible choice of symbols; rational rank does not supply it.

The finite-state and cyclic statements are author-recorded; they were derived by a separate contributor and checked by the author, including initial/final residuals, the flag semiring, periodic-state consistency, and interval projection. No experiments or external theorems are used.

A second separate reviewer checked the residual bounds, acceptance conditions, monotone periodicization, cyclic flags, ultimate periodicity, and unrolling bound, finding no gap. No search or computation was performed.