> Full-text translation from Chinese. The original remains authoritative for its research claims; this translation is not independent mathematical verification. # X4-DMSD: X–Four-Edge Mortise-and-Tenon Constraint Domains and Decimal Multiscale Search Methodology **X4-DMSD: Exact X–Four-Edge Joint Constraint Domains with Decimal Multiscale Search** Version: v0.1 Nature: Foundational methodology / propositions / conjectures / search architecture Target problem: Exact structural search for the Perfect Cuboid Core techniques: X-coupling, Four-Edge Joint, Exact Angle Fingerprint, DMSD, Lazy Refinement, Sound Pruning Canonical source of the original: UTF-8 Markdown --- ## Abstract This research proposes a new search methodology for highly sparse arithmetic-geometric problems such as the Perfect Cuboid. Rather than taking edge-length tuples, near-square gaps, floating-point residuals or enormous integer enumerations as the first-layer search objects, it first builds an exact mortise-and-tenon constraint domain jointly defined by **X structure, four-edge angles, shared-edge coupling and the global diagonal direction**. The central idea is: > A candidate enters the next layer only when every X–edge–angle interface can "click into place exactly": neither a little too much nor a little too little is acceptable. Here, "angles" are not judged by floating-point trigonometric functions. Pythagorean ratios, direction fingerprints and integer cross-products instead establish an exact predicate, so: $$ \operatorname{LOCK}\in\{0,1\}. $$ There is no approximate pass of the form: $$ 0.99999999 $$ that numerical near-fit. Next, this research specializes DMSD (Decimal Multiscale Serializable Domain), so that it no longer serializes every element of a huge integer domain, but instead serializes: $$ \boxed{ \text{X–joint compatible constraint blocks} } $$ That is, each decimal-scale block stores only PerfectClick-relevant summary states, shared-edge keys, angle keys, remaining possibilities and infeasibility certificates. If exact necessary conditions can determine for an entire block that: $$ \operatorname{PossibleClick}(B)=0, $$ the entire block is sound-pruned without expanding its elements. The purpose of X4-DMSD is therefore not to expand brute force from: $$ 10^{12} \rightarrow 10^{100} $$ to ever larger scales, but: $$ \boxed{ \text{element search} \rightarrow \text{constraint-domain search} \rightarrow \text{multiscale block elimination}. } $$ This paper establishes X4-DMSD's basic objects, Exact X Fingerprint, Four-Edge Joint, Shared-Edge Coupling, PerfectClick predicate, DMSD block semantics, sound-pruning conditions, lazy refinement, search state machine, certificate format and subsequent experimental program. This paper does not claim that a Perfect Cuboid has been found or that the method necessarily outperforms every existing search. It proposes a specialized search framework that is implementable, falsifiable, expandable scale by scale and does not rely on floating-point approximation as a theorem gate. --- # 1. Reframing the Problem The standard length representation of a Perfect Cuboid can be written as: $$ W = (a,b,c,d_{ab},d_{ac},d_{bc},g), $$ satisfying: $$ a^2+b^2=d_{ab}^2, $$ $$ a^2+c^2=d_{ac}^2, $$ $$ b^2+c^2=d_{bc}^2, $$ $$ a^2+b^2+c^2=g^2. $$ Traditional searches often take: $$ a,b,c $$ or Pythagorean tuples as the main enumeration variables. X4-DMSD changes the first-layer search unit. The first-layer search object is no longer: $$ (a,b,c), $$ but: $$ \boxed{ \text{Local X–four-edge world units admitting exact mortise-and-tenon coupling} } $$ and whether these local units can form global shared-edge closure. --- # 2. "A Click" Is Not a Numerical Metaphor Define: $$ \operatorname{LOCK}(J_1,J_2) \in \{0,1\}. $$ If two joints are exactly compatible at every theorem-relevant interface: $$ \operatorname{LOCK}=1. $$ Otherwise: $$ \operatorname{LOCK}=0. $$ It is forbidden to define: $$ 0< \operatorname{LOCK} < 1. $$ Thus: $$ \boxed{ \text{near fit} \neq \text{fit}. } $$ --- # 3. Do Not Compute Angles Directly If a rectangular face has edges: $$ a,b>0, $$ a face diagonal: $$ d, $$ and: $$ a^2+b^2=d^2, $$ the angle $\theta$ can be written as: $$ \cos\theta=\frac ad, $$ $$ \sin\theta=\frac bd. $$ There is therefore no need for floating-point computation of: $$ \theta=\arctan\left(\frac ba\right) $$ that angle. Define the exact edge-angle fingerprint: $$ \boxed{ \chi(a,b;d) = \left( \frac ad,\frac bd \right). } $$ --- # 4. Direction Fingerprint If comparing only directions, the diagonal is not even needed: $$ \theta_1=\theta_2 $$ if and only if: $$ \frac{b_1}{a_1} = \frac{b_2}{a_2}. $$ For integers or rational quantities: $$ \boxed{ b_1a_2=b_2a_1. } $$ Exact angle equality can therefore be decided entirely with integer multiplication. --- # 5. X Fingerprint A rectangle's two diagonals can be written as direction vectors: $$ v_+=(a,b), $$ $$ v_-=(a,-b). $$ Their central angle $\phi$ satisfies: $$ \cos\phi = \frac{a^2-b^2}{a^2+b^2}, $$ $$ \sin\phi = \frac{2ab}{a^2+b^2}. $$ Define: $$ \boxed{ \Xi(a,b) = \left( \frac{a^2-b^2}{a^2+b^2}, \frac{2ab}{a^2+b^2} \right). } $$ This is called the **X Fingerprint**. --- # 6. Why $\Xi$ Is More Suitable Than Floating-Point Angles If: $$ a,b\in\mathbb Q, $$ then: $$ \Xi(a,b)\in\mathbb Q^2. $$ Thus: - No trigonometric approximation is needed; - No angle tolerance is needed; - No degrees / radians conversion is needed; - Direct exact comparison is possible; - Modularization is possible; - Hashing is possible; - It can serve as a block-summary key. --- # 7. Four-Edge Joint A face consists of more than two edge lengths. In rectangle topology, four edges form two pairs of equal opposite edges. Define: $$ J_{ab} = ( E_a^{(1)}, E_b^{(1)}, E_a^{(2)}, E_b^{(2)}, X_{ab}, \chi_{ab}, \Xi_{ab} ). $$ where: $$ |E_a^{(1)}| = |E_a^{(2)}| = a, $$ $$ |E_b^{(1)}| = |E_b^{(2)}| = b. $$ --- # 8. X4 Joint This paper treats: $$ X+\text{four edges} $$ as a single indivisible search unit: $$ \boxed{ J^{X4}_{ab}. } $$ An X4 joint contains at least: 1. Four-edge topology; 2. Opposite-edge equality; 3. Diagonal X; 4. Exact angle fingerprint; 5. X central-angle fingerprint; 6. Orientation; 7. Scale-free ratio class; 8. Arithmetic provenance. --- # 9. Joint Equivalence If: $$ (a,b,d) \mapsto (ka,kb,kd) $$ where: $$ k>0, $$ then: $$ \chi $$ and: $$ \Xi $$ remain unchanged. X4 shape equivalence can therefore be defined by: $$ J_1 \sim_X J_2 $$ when the normalized fingerprints agree. --- # 10. Pythagorean Seed A primitive Pythagorean face can be generated from: $$ m>n>0 $$ as: $$ a=m^2-n^2, $$ $$ b=2mn, $$ $$ d=m^2+n^2. $$ where: $$ \gcd(m,n)=1 $$ and $m,n$ have opposite parity to give a primitive triple. An X4 joint can therefore use: $$ (m,n) $$ directly as its seed, rather than first being represented by enormous: $$ (a,b,d) $$ lengths. --- # 11. Scale Quotient If: $$ (a,b,d) = k ( m^2-n^2, 2mn, m^2+n^2 ), $$ all angle / X fingerprints are independent of: $$ k $$ that scale. Thus: $$ \boxed{ \text{first search shape seed, then scale}. } $$ This complements, rather than conflicts with, DMSD's Scale First principle: - The X4 quotient removes meaningless common scale; - DMSD manages the arithmetic-height scale of seeds / families / couplings. --- # 12. Three Face-X Joints A Perfect Cuboid has three face joints: $$ J_{ab}, $$ $$ J_{ac}, $$ $$ J_{bc}. $$ They are not three independent Pythagorean triples. Shared relationships: $$ J_{ab} \cap J_{ac} = a, $$ $$ J_{ab} \cap J_{bc} = b, $$ $$ J_{ac} \cap J_{bc} = c. $$ --- # 13. Shared-Edge Coupling Define the shared-edge operator: $$ J_{ab} \bowtie_a J_{ac}. $$ It holds if and only if the edges marked $a$ in the two joints are exactly compatible under the allowed global-scale convention. Similarly: $$ J_{ab} \bowtie_b J_{bc}, $$ $$ J_{ac} \bowtie_c J_{bc}. $$ --- # 14. Raw Equality and Ratio Equality When comparing in an integerized world: $$ a_{ab} = a_{ac}, $$ use raw equality. When still in scale-free joint space, one can compare: $$ \frac{a_{ab}}{d_{ab}} $$ and other normalized edge signatures, then solve for a common scale in subsequent global-scale solving. Coupling therefore has two modes: ### Integer Coupling $$ a_1=a_2. $$ ### Projective Coupling $$ a_1:a_2 $$ determined by a common scale equation. --- # 15. Global Scale Variables For three local joints: $$ J_{ab}^{(0)}, J_{ac}^{(0)}, J_{bc}^{(0)}, $$ assign respective scales: $$ k_{ab}, k_{ac}, k_{bc}. $$ Require: $$ k_{ab}a_{ab}^{(0)} = k_{ac}a_{ac}^{(0)}, $$ $$ k_{ab}b_{ab}^{(0)} = k_{bc}b_{bc}^{(0)}, $$ $$ k_{ac}c_{ac}^{(0)} = k_{bc}c_{bc}^{(0)}. $$ This is an exact rational scale-coupling system. --- # 16. Shared-Edge Scale Consistency The equations above are equivalent to: $$ \frac{k_{ac}}{k_{ab}} = \frac{a_{ab}^{(0)}}{a_{ac}^{(0)}}, $$ $$ \frac{k_{bc}}{k_{ab}} = \frac{b_{ab}^{(0)}}{b_{bc}^{(0)}}. $$ The third equation must be compatible with the first two: $$ \boxed{ \frac{c_{ac}^{(0)}}{c_{bc}^{(0)}} = \frac{k_{bc}}{k_{ac}}. } $$ If this fails, the three joints can never form a cuboid. This is a powerful exact pre-filter. --- # 17. Proposition X4-1: Triple-Joint Coupling Condition Given three local scale-free Pythagorean joints. There exist positive rational scales: $$ k_{ab},k_{ac},k_{bc} $$ that merge their shared edges exactly if and only if the three pairwise scale ratios are consistent. This can be converted into a single cross-multiplication identity. Shared-edge compatibility can therefore reject exactly at the integer-seed layer. --- # 18. Global Body-X Signature A Perfect Cuboid also requires a body diagonal: $$ g. $$ Define: $$ \boxed{ \Gamma(a,b,c;g) = \left( \frac ag, \frac bg, \frac cg \right). } $$ If: $$ a^2+b^2+c^2=g^2, $$ then: $$ \Gamma $$ lies on the unit sphere. --- # 19. Body-X as Global Mortise One can interpret: $$ \Gamma $$ as a global socket / mortise. The three face joints combine to produce: $$ (a,b,c) $$ which must also fit exactly with: $$ g. $$ Thus: $$ \boxed{ \text{three local X4 joints} + \text{one global body-X joint}. } $$ --- # 20. PerfectClick Predicate Define: $$ \operatorname{PerfectClick} ( J_{ab}, J_{ac}, J_{bc}, \Gamma ). $$ Only when all of the following hold: $$ J_{ab} \bowtie_a J_{ac}, $$ $$ J_{ab} \bowtie_b J_{bc}, $$ $$ J_{ac} \bowtie_c J_{bc}, $$ and: $$ a^2+b^2+c^2=g^2 $$ and every theorem-required quantity lies in the target arithmetic domain, do we have: $$ \boxed{ \operatorname{PerfectClick}=1. } $$ --- # 21. Target Arithmetic Domain For the Perfect Cuboid: $$ \mathbb A = \mathbb Z_{>0}. $$ One can first use, in a normalized layer: $$ \mathbb Q_{>0}. $$ If the rational shape closes completely, then clear denominators: $$ \mathbb Q \rightarrow \mathbb Z. $$ --- # 22. PerfectClick Is Not a Score Forbidden: $$ \operatorname{PerfectClick} = 0.9999. $$ A separate search score may exist: $$ S_{\mathrm{heuristic}}, $$ but the theorem status remains: $$ \operatorname{PerfectClick} \in \{0,1\}. $$ --- # 23. Mortise-and-Tenon Constraint Domain Define: $$ \mathcal J_X $$ as all valid local X4 joints. Define the three-joint coupling domain: $$ \mathcal C_{X4} = \left\{ (J_{ab},J_{ac},J_{bc}) : \text{shared-edge compatible} \right\}. $$ Perfect target domain: $$ \boxed{ \mathcal P_{X4} = \left\{ W\in\mathcal C_{X4} : \operatorname{PerfectClick}(W)=1 \right\}. } $$ --- # 24. Rewriting the Search Problem Traditional search: $$ \text{Search }(a,b,c)\in\mathbb Z^3. $$ X4: $$ \boxed{ \text{Search }W\in\mathcal C_{X4}. } $$ This means that most tuples incapable of forming joints never enter the search domain. --- # 25. This Is Not Ordinary Constraint Programming Ordinary constraint programming also prunes. X4-DMSD additionally requires: 1. The basic search unit is a geometric joint, not a scalar variable; 2. Angle / X structure is canonicalized; 3. Scale quotienting is performed first; 4. Shared-edge coupling is a first-class index; 5. DMSD expands across orders of magnitude using constraint blocks; 6. Block summaries themselves admit merging and provable exclusion. --- # 26. Specializing DMSD The original DMSD's basic idea: $$ \text{Scale First} + \text{Summary} + \text{Lazy Refinement}. $$ X4-DMSD does not change this principle, but changes the property to: $$ \boxed{ \text{joint compatibility property}. } $$ --- # 27. X4-DMSD Block Define a multiscale block: $$ B_k^X. $$ It is not a materialized list of every ordinary integer within: $$ [10^k,10^{k+1}) $$ that interval. Instead, it is an X-joint seed / height / coupling region. --- # 28. Seed Height For example, for a Pythagorean seed: $$ (m,n), $$ one can define: $$ H_X(m,n) = \max(m,n) $$ or: $$ H_d(m,n) = m^2+n^2. $$ Decimal scale: $$ 10^k \le H_X < 10^{k+1}. $$ --- # 29. Block Summary X4-DMSD block summary: $$ \Sigma(B) = ( K_X, K_E, K_R, K_M, K_G, K_C ). $$ where: ### $K_X$ X fingerprint classes. ### $K_E$ Edge-ratio / shared-edge signature classes. ### $K_R$ Residue / modular masks. ### $K_M$ Possible mating classes. ### $K_G$ Body-X compatible summaries. ### $K_C$ Closure state: $$ \{ \mathrm{IMPOSSIBLE}, \mathrm{POSSIBLE}, \mathrm{UNKNOWN}, \mathrm{EXACT} \}. $$ --- # 30. The Meaning of IMPOSSIBLE Only when an exact necessary condition proves that the entire block contains no PerfectClick candidate: $$ K_C(B) = \mathrm{IMPOSSIBLE}. $$ may it be pruned. Thus: $$ \boxed{ \text{pruning must be sound}. } $$ --- # 31. Sound-Pruning Requirement Defining: $$ \operatorname{Prune}(B)=1. $$ must guarantee: $$ \boxed{ B\cap\mathcal P_{X4} = \varnothing. } $$ If it merely: > Looks unlikely to succeed, it cannot be pruned. --- # 32. No-False-Negative Principle For theorem-directed mode: $$ \boxed{ \operatorname{PerfectClick}(W)=1 \Rightarrow W \text{ must never be pruned by any ancestor block}. } $$ This is X4-DMSD's most important correctness property. --- # 33. Necessary-Condition Summaries A block can store, for example: - possible slope classes； - possible opposite-edge ratios； - possible shared-edge gcd classes； - possible modular square classes； - possible body-diagonal residues； - possible scale-ratio compatibility； - known impossible angle pairings. --- # 34. Block-Level Join For two blocks: $$ B_i, B_j, $$ define: $$ B_i \bowtie B_j. $$ This does not generate the entire Cartesian product. Compare summaries first: $$ \Sigma(B_i), \Sigma(B_j). $$ If summary-level necessary conditions are incompatible: $$ \boxed{ B_i\bowtie B_j=\varnothing. } $$ the entire block pair is excluded. --- # 35. Summary Before Elements X4-DMSD's central computational rule: $$ \boxed{ \text{summary join first} \rightarrow \text{refine only survivors}. } $$ Not: $$ \text{materialize} \rightarrow \text{compare every pair}. $$ --- # 36. Decimal Multiscale Tree Root: $$ B_0. $$ Children: $$ B_{0,0}, \ldots, B_{0,9}. $$ Or, by seed height, split: $$ [10^k,10^{k+1}) $$ into: $$ 10 $$ logarithmic / arithmetic sub-blocks. --- # 37. Lazy Refinement If: $$ K_C(B) = \mathrm{IMPOSSIBLE}, $$ stop. If: $$ K_C(B) = \mathrm{EXACT}, $$ output a certificate directly. Only if: $$ K_C(B) = \mathrm{POSSIBLE} $$ or: $$ \mathrm{UNKNOWN}, $$ then: $$ B \rightarrow \{B_1,\ldots,B_m\}. $$ --- # 38. Arbitrary Scale Representation Thus, even if: $$ H_X \sim 10^{100}, $$ the system does not require materializing: $$ 10^{100} $$ seeds. It needs only to serialize: $$ B_{100}^X. $$ Whether to refine afterward is decided by the compatibility summary. --- # 39. Scale Jump If a property operator can act directly on: $$ B_k $$ without expanding lower layers, one can make the transition: $$ k=10 \rightarrow 100 \rightarrow 1000. $$ This is the "order-of-magnitude expansion" for which DMSD is actually responsible in this problem. --- # 40. This Is Not Skipping Proof A scale jump does not mean skipping proof. It requires the block operator: $$ F_B $$ to hold over the entire block. For example: $$ \operatorname{ResidueMask}(B). $$ If the mask is already disjoint from the residues necessary for PerfectClick, reject the entire block exactly. --- # 41. Residue Projection For edge / diagonal square conditions, choose a modulus: $$ m. $$ Store: $$ R_m(B) \subseteq (\mathbb Z/m\mathbb Z)^n. $$ If: $$ R_m(B) \cap R_m^{\mathrm{Perfect}} = \varnothing, $$ then: $$ B $$ can be sound-pruned. --- # 42. Multi-Modulus Intersection Multiple moduli: $$ m_1,\ldots,m_t $$ can be combined: $$ R(B) = \bigcap_i R_{m_i}(B). $$ This is essentially a block-level modular sieve. --- # 43. Angle-Key Index Because: $$ \frac ba $$ defines the direction, one can use a reduced ratio: $$ \operatorname{AngleKey}(a,b) = \frac{b/\gcd(a,b)}{a/\gcd(a,b)}. $$ For a Pythagorean seed, a canonical key can also be generated directly from: $$ m,n $$ those seed values. --- # 44. X-Key Index $$ \Xi(a,b) $$ can be reduced to a rational pair: $$ (\xi_1,\xi_2). $$ Define: $$ \operatorname{XKey} = \operatorname{Reduce} ( a^2-b^2, 2ab, a^2+b^2 ). $$ This is actually a Pythagorean-triple-like key. --- # 45. Shared-Edge Key If a local joint edge: $$ e $$ has primitive representation: $$ e_0, $$ the global-scale coupling can be recorded as: $$ [e_0:k]. $$ A block summary can store: - prime support； - valuation ranges； - gcd class； - possible scale ratio. --- # 46. Prime-Valuation Summary If shared-edge compatibility requires: $$ k_1e_1=k_2e_2, $$ then, for any prime: $$ p, $$ we have: $$ v_p(k_1)+v_p(e_1) = v_p(k_2)+v_p(e_2). $$ A block can store valuation intervals first. If the intervals are disjoint, it can be rejected directly. --- # 47. Body-Diagonal Residue Gate Even after the three face joints have been merged, one still needs: $$ a^2+b^2+c^2 $$ to be a square. The block level can therefore store: $$ a^2+b^2+c^2 \bmod m. $$ If none of the possible residues is a square residue: $$ \boxed{ \text{global body-X impossible}. } $$ --- # 48. Search State Machine X4-DMSD candidate states: $$ \mathrm{BLOCK}, $$ $$ \mathrm{LOCAL\_JOINT}, $$ $$ \mathrm{PAIR\_COUPLED}, $$ $$ \mathrm{TRIPLE\_COUPLED}, $$ $$ \mathrm{BODY\_COMPATIBLE}, $$ $$ \mathrm{PERFECT\_CLICK}, $$ $$ \mathrm{REJECTED}. $$ --- # 49. Early Rejection Priority Suggested gate order: 1. local Pythagorean/X validity; 2. angle/X key canonicalization; 3. shared-edge ratio feasibility; 4. valuation / residue compatibility; 5. triple scale consistency; 6. body-square residue; 7. exact body diagonal; 8. final PerfectClick certificate. Run cheaper gates first. --- # 50. Candidate Graph Each surviving local X4 joint is a node: $$ v_i. $$ If two joints may share an edge: $$ e_{ij}\in E. $$ This therefore forms: $$ G_X=(V,E). $$ A Perfect Cuboid candidate becomes: $$ \boxed{ \text{Triangle / motif search with consistency of three shared-edge types} } $$ rather than arbitrary triple enumeration. --- # 51. Typed Edges A graph edge is not a single relation. Types: $$ E_a, $$ $$ E_b, $$ $$ E_c. $$ The target motif must therefore simultaneously contain: $$ a,b,c $$ these three types of compatibility. --- # 52. Hypergraph Form More naturally, a hyperedge can be used directly: $$ h = (J_{ab},J_{ac},J_{bc}). $$ If the three-joint scale system is consistent, then: $$ h $$ becomes a triple-coupled world seed. --- # 53. DMSD Hyperblock Further: $$ \mathbb B = (B_{ab},B_{ac},B_{bc}) $$ is a hyperblock of three X-joint blocks. First determine: $$ \operatorname{PossibleTripleLock}(\mathbb B). $$ If false, do not expand the entire hyperblock. --- # 54. Transfer of Complexity Ordinary search complexity lies mainly in: $$ |\mathcal A| \cdot |\mathcal B| \cdot |\mathcal C|. $$ X4-DMSD aims to shift the cost to: $$ \boxed{ \text{number of surviving compatible blocks} + \text{refinement cost}. } $$ Whether it is genuinely smaller must be measured experimentally. --- # 55. Methodological Conjecture X4-C1 **Constraint-Domain Compression Conjecture** For Perfect Cuboid search, building an exact local X4-joint domain before shared-edge coupling gives a significantly higher effective candidate density than ambient integer-tuple search. This is an experimental conjecture, not a current theorem. --- # 56. Methodological Conjecture X4-C2 **Multiscale Block Sparsity Conjecture** As arithmetic height increases, most DMSD X4 blocks are excluded before element materialization by at least one exact necessary condition among: - scale-ratio incompatibility； - modular residue； - valuation； - body-square； - angle-key; these conditions. --- # 57. Methodological Conjecture X4-C3 **Exact-Click Isolation Conjecture** If a Perfect Cuboid solution exists, its corresponding X4-joint triple will, under sufficiently fine DMSD refinement, fall into an: $$ \mathrm{EXACT} $$ block, rather than being recognizable only as an asymptotic near-hit. --- # 58. Methodological Conjecture X4-C4 **Scale-Independent Structural Signature Conjecture** If two huge-scale Perfect candidates differ only by common scale, their local fingerprints and coupling skeletons in X4-DMSD canonicalize to the same shape signature. Scale duplicates can therefore be eliminated before materialization. --- # 59. Soundness Theorem Goal A future proof should establish: If: $$ \operatorname{Prune}(B)=1, $$ then: $$ B\cap\mathcal P_{X4} = \varnothing. $$ Prove this separately for each type of pruning operator. --- # 60. Completeness Goal If the search tree refines without bound: $$ B \rightarrow B_i \rightarrow B_{ij} \rightarrow\cdots $$ and every nonempty candidate cell is eventually refined, then every finite-height PerfectClick candidate should be reached. This requires a formal coverage proof. --- # 61. Search Certificate If a PerfectClick is found: $$ \mathcal C_{\mathrm{click}} $$ output at least: 1. Three local X4 seeds; 2. Local Pythagorean certificates; 3. Shared-edge scale solution; 4. Merged $a,b,c$; 5. Three face diagonals; 6. Body diagonal; 7. All exact equations; 8. Primitive reduction; 9. Originating DMSD block path. --- # 62. Rejection Certificate A block rejection outputs at least: $$ \mathcal C_{\mathrm{reject}} = ( B, \text{gate}, \text{witness}, \text{scope} ). $$ For example: ```text gate = body_residue modulus = 16 possible_body_residues = {3,5,6} square_residues = {0,1,4,9} intersection = empty ``` --- # 63. UNKNOWN Is a Valid State If the block summary is insufficient to decide: $$ K_C(B) = \mathrm{UNKNOWN}. $$ One cannot prune directly because: - It is too large; - It has not yet been computed; - An operator is missing; those are not grounds for pruning. --- # 64. POSSIBLE Is Not a Candidate Witness Either $$ \mathrm{POSSIBLE} $$ means only that necessary gates have not yet excluded it. One cannot claim: $$ \operatorname{PerfectClick}=1. $$ --- # 65. Three Layers of Truth X4-DMSD adopts: ### Block Truth $$ \mathrm{IMPOSSIBLE}, \mathrm{POSSIBLE}, \mathrm{UNKNOWN}. $$ ### World Truth Whether exact constraints hold. ### Theorem Truth Whether there is a complete PerfectClick. The three layers must not be conflated. --- # 66. Specialized DMSD Property Algebra Original DMSD properties are usually counts / histograms / residue vectors. The X4-DMSD property becomes: $$ \boxed{ \text{compatibility algebra}. } $$ The merge operator: $$ \oplus $$ does not merely sum, but may be: - set union； - mask intersection； - relation composition； - valuation convolution； - edge-key join. --- # 67. Example: Residue Mask Merge If child blocks: $$ B_i $$ have possible residues: $$ R(B_i), $$ the parent has: $$ R(B) = \bigcup_iR(B_i). $$ If: $$ R(B) $$ is already entirely incompatible, the parent can be pruned directly. --- # 68. Example: Compatibility Intersection If block $A$ has possible edge keys: $$ K_A, $$ and block $B$ has possible edge keys: $$ K_B, $$ then: $$ K_{A\bowtie B} = K_A\cap K_B. $$ If: $$ K_A\cap K_B=\varnothing, $$ then: $$ A\bowtie B=\varnothing. $$ --- # 69. Example: Scale-Ratio Set If a local block has possible primitive shared-edge sets: $$ E_A, E_B, $$ reduced ratio classes can be stored: $$ R_{AB} = \left\{ \frac{e_A}{e_B} \right\}. $$ If a three-joint ratio cycle: $$ R_{ab,ac} R_{ac,bc} R_{bc,ab} $$ cannot contain: $$ 1, $$ the triple is impossible. --- # 70. Triple Ratio Cycle For an exact triple: $$ \frac{k_{ac}}{k_{ab}} \cdot \frac{k_{bc}}{k_{ac}} \cdot \frac{k_{ab}}{k_{bc}} = 1. $$ Thus, if any block-level ratio intervals / sets satisfy: $$ 1 \notin R_1R_2R_3, $$ they can be rejected directly. --- # 71. Geometric View of Angle Coupling Visually, each local joint is a puzzle piece. But the program actually stores not a bitmap shape, but: $$ \chi, \Xi, E, R, V_p. $$ that is, an exact symbolic mortise/tenon signature. --- # 72. Formalizing "Neither Too Much Nor Too Little" Human visual language: > A little too much is unacceptable; a little too little is unacceptable. X4 formal language: $$ \boxed{ \text{cross-products equal} } $$ $$ \boxed{ \text{shared-edge scale equations solvable} } $$ $$ \boxed{ \text{all exact quadratic constraints zero} } $$ $$ \boxed{ \text{arithmetic domains satisfied}. } $$ --- # 73. The Role of Floating Point Floating point can be used only for: - UI； - animation； - rough ordering； - visual inspection. It cannot decide: $$ \operatorname{LOCK}. $$ --- # 74. No Angular Tolerance It is forbidden to use: $$ |\theta_1-\theta_2| < 10^{-12} $$ as a lock. Formally use: $$ b_1a_2=b_2a_1 $$ or equality of normalized rational fingerprints. --- # 75. No Scale Tolerance Forbidden: $$ |k_1e_1-k_2e_2|<\epsilon. $$ Formally use rational equality. --- # 76. No Body Tolerance Forbidden: $$ |a^2+b^2+c^2-g^2|<\epsilon. $$ Formally use: $$ a^2+b^2+c^2=g^2. $$ --- # 77. X4-DMSD MVP The first version need not search to enormous scales. The MVP should first: ### M1 Generate primitive Pythagorean X4 joints. ### M2 Canonical X / angle keys. ### M3 Shared-edge pair join. ### M4 Triple scale-consistency join. ### M5 Body-diagonal exact gate. ### M6 Build DMSD height blocks. ### M7 Block residue summaries. ### M8 Lazy refinement. ### M9 Bounded baseline comparison. ### M10 Scale-jump experiment. --- # 78. Control A：Known Pythagorean Reconstruction Verify that the local X4 generator fully reconstructs: $$ (3,4,5), $$ $$ (5,12,13), $$ and other known joints. --- # 79. Control B：Known Euler Brick Euler brick: $$ (a,b,c) = (44,117,240) $$ The three face diagonals close, but the body does not. X4-DMSD should successfully form a triple-joint world, but: $$ \operatorname{PerfectClick}=0. $$ This is a very important negative control. --- # 80. Control C：Near-Hit Trap Input many near-square and near-angle candidates. No approximate match may pass: $$ \operatorname{LOCK}. $$ --- # 81. Control D：Scale Duplicate For: $$ W $$ and: $$ kW, $$ the X4 shape signatures should be identical. They must not be counted again as two distinct structural solutions. --- # 82. Control E：Huge DMSD Empty Block Artificially construct a block on the scale of: $$ 10^{100} $$ that magnitude, whose residue / valuation summary proves that no solution exists. Require the system to reject it without materializing elements. This validates a core DMSD capability. --- # 83. Baseline 1：Ambient Edge Search Test: $$ a,b,c\le N. $$ Record: - tuples tested； - face-square tests； - body-square tests； - runtime. --- # 84. Baseline 2：Pythagorean Triple Join Traditionally, generate triples, then join them by shared edges. This is a fairer baseline that X4-DMSD must outperform or at least offer distinct structural value against. --- # 85. Baseline 3：X4 Without DMSD Use exact X4 joints / coupling, but no multiscale block pruning. This separates: $$ \text{X4 gain} $$ from: $$ \text{DMSD gain}. $$ --- # 86. Baseline 4：X4-DMSD Complete version: $$ \boxed{ \text{X4} + \text{block summaries} + \text{sound prune} + \text{lazy refinement}. } $$ --- # 87. Metrics Measure at least: ### Element Avoidance Ratio $$ R_E = 1- \frac{ N_{\mathrm{materialized}} }{ N_{\mathrm{ambient}} }. $$ ### Block Prune Rate $$ R_B = \frac{ N_{\mathrm{pruned\ blocks}} }{ N_{\mathrm{visited\ blocks}} }. $$ ### Structural Yield $$ Y_S = \frac{ N_{\mathrm{triple\ coupled}} }{ N_{\mathrm{generated\ local\ joints}} }. $$ ### Exact Candidate Yield $$ Y_C = \frac{ N_{\mathrm{body\ compatible}} }{ N_{\mathrm{triple\ coupled}} }. $$ --- # 88. Correctness Metrics Most importantly: $$ \boxed{ \mathrm{FalseNegative}=0. } $$ across all known controls. If X4-DMSD misses a known exact solution through pruning, the method fails outright. --- # 89. Cross-Scale Consistency If first a summary is produced at: $$ B_k $$ that block, and it is then refined, the parent summary must agree with the merged children: $$ \boxed{ \Sigma(B_k) = \bigoplus_i \Sigma(B_{k,i}) } $$ for summary fields declared exactly composable. --- # 90. DMSD Serialization A block can serialize: ```text block_id scale height_interval seed_family x_keys edge_keys residue_masks valuation_masks possible_mates body_masks closure_state children certificate ``` --- # 91. Canonical Block Hash Each block summary can have a canonical hash: $$ H(B) = \operatorname{SHA256} ( \operatorname{CanonicalSerialize}(\Sigma(B)) ). $$ Used for: - cache； - distributed search； - AI agent handoff； - deduplication. --- # 92. Distributed / Agent Search Because blocks can be independently refined, one can distribute: $$ B_1,\ldots,B_n $$ to different workers / agents. Each needs to return only: - summary； - children； - reject certificate； - an exact candidate. --- # 93. World Search Ledger Each time: $$ B \rightarrow B_i $$ write to: $$ \mathcal L. $$ If a solution is eventually found, one can replay the full path: $$ \text{root block} \rightarrow \cdots \rightarrow \text{PerfectClick}. $$ --- # 94. Search Without Materialization Ideally: $$ B_{100} $$ is rejected directly from its summary. The system then proves: > A particular entire $10^{100}$ arithmetic-scale region cannot contain a compatible X4 world. This is X4-DMSD's genuine attack on "orders of magnitude". --- # 95. This Does Not Claim That an Entire $10^{100}$ Range Has No Solution Block semantics must be extremely precise. Only the following can be excluded: $$ \boxed{ \text{The seed / constraint region represented by that block}. } $$ If the block definition does not cover all numbers, one cannot claim that an entire numerical interval has no solution. --- # 96. Coverage Map DMSD must therefore preserve: $$ \operatorname{Coverage}(B). $$ and prove: $$ \bigcup_i \operatorname{Coverage}(B_i) = \operatorname{Coverage}(B). $$ Otherwise, pruning statistics have no theorem-level meaning. --- # 97. Representation Artifact Audit Decimal blocks may produce base-related patterns. X4-DMSD must therefore distinguish: $$ \text{mathematical structure} $$ from: $$ \text{decimal partition artifact}. $$ Perform a cross-base audit when necessary: $$ 10, 2, 3, 16, \ldots $$ --- # 98. The Role of Decimal Decimal is not a theorem primitive. It is merely: - A human-readable scale; - DMSD serialization grammar; - Logarithmic partition. All PerfectClick equations are independent of base. --- # 99. Cross-Base Invariance Goal If a pruning rule is genuinely arithmetic structure, it should give consistent theorem outcomes for the same candidate coverage across different block bases. --- # 100. X4-DMSD's Actual Research Proposition Not: > Decimal can prove the Perfect Cuboid. But: $$ \boxed{ \text{Exact joint semantics can recompile the Perfect Cuboid search space into a multiscale serializable constraint domain.} } $$ Then experimentally test: $$ \boxed{ \text{Whether this representation can push the effective search scale far beyond ordinary element search.} } $$ --- # 101. Relationship to GERES The GERES main line studies: $$ \text{world reconstruction} \rightarrow \text{low-dimensional arithmetic geometry}. $$ The X4-DMSD main line studies: $$ \text{exact local joints} \rightarrow \text{multiscale constraint search}. $$ Neither line replaces the other. --- # 102. Intersection of the Two Lines If X4-DMSD finds: $$ \mathrm{PerfectClick}=1, $$ it directly produces a GERES-compatible exact world certificate. If it finds only a: $$ (0,1,1,1) $$ near-world, this can be sent back to GERES for analysis of its arithmetic geometry. --- # 103. Relationship to the X-Diagonal Geometry Branch X4-DMSD can be viewed as a discrete exact-search specialization of the original X-Diagonal / center-axis coupling. Originally: $$ \text{visual geometry} $$ now becomes: $$ \boxed{ \text{canonical X keys} + \text{joint compatibility algebra}. } $$ --- # 104. First-Phase Series Roadmap ### X4-P0 This paper: methodology. ### X4-P1 Exact X4 Joint Algebra. ### X4-P2 Shared-Edge Scale Coupling Engine. ### X4-P3 PerfectClick Runtime. ### X4-P4 DMSD Block Semantics. ### X4-P5 Sound Modular / Valuation Pruning. ### X4-P6 Bounded Baseline Benchmark. ### X4-P7 Decimal Scale Jump. ### X4-P8 Cross-Base Audit. ### X4-P9 Distributed Huge-Scale Search. ### X4-P10 Theorem Extraction / Formal Verification. --- # 105. The First Genuine MVP Do not immediately go to: $$ 10^{100}. $$ First use: $$ H_X\le10^4 $$ or smaller. Compare: 1. ambient integer search; 2. Pythagorean triple join; 3. X4 exact join; 4. X4-DMSD. Only when: $$ R_E, R_B, Y_S $$ start to show a clear advantage, then make a scale jump. --- # 106. Scale Jump Gate Allow the following transition only after these conditions hold: $$ 10^k \rightarrow 10^{k+\Delta}. $$ Require: 1. block coverage verified; 2. summary merge exact; 3. no-false-negative controls pass; 4. pruning certificates replayable; 5. cross-base artifact audit at least partially passed. --- # 107. The Most Important Counterexample If the DMSD block summary is too coarse, it is possible that: $$ B $$ contains just one rare PerfectClick, but the summary incorrectly judges: $$ \mathrm{IMPOSSIBLE}. $$ This is the most dangerous failure of the entire method. Thus: $$ \boxed{ \text{soundness first, compression second}. } $$ --- # 108. Exactness Priority It is preferable to have many: $$ \mathrm{UNKNOWN} $$ blocks than to prune incorrectly. Early X4-DMSD should be conservative. --- # 109. Formal Theorem Goal The long-term aim is to prove: **X4-DMSD Sound Coverage Theorem** If: $$ \mathcal T $$ is a complete X4-DMSD search tree, and every pruning operator is sound, then for every finite-height Perfect Cuboid witness: $$ W^\ast, $$ there must be an unpruned path: $$ B_0 \rightarrow B_1 \rightarrow \cdots \rightarrow W^\ast. $$ --- # 110. Formal Compression Goal Next prove: For some summary operators: $$ \Sigma(B) $$ has computational cost sublinear in block cardinality, or even: $$ O(1) $$ or: $$ O(\log |B|). $$ This is the theoretical core that genuinely supports "order-of-magnitude expansion". --- # 111. X4-DMSD's Philosophical Position The world itself is not a string of numbers. Numbers are only: $$ \text{world projection}. $$ What X4-DMSD attempts to search for is not: $$ \text{numbers that happen to work}, $$ but: $$ \boxed{ \text{structures that are allowed to exist together}. } $$ DMSD, in turn, is responsible for: $$ \boxed{ \text{Deciding scale by scale which structures merit existence without expanding the entire enormous world.} } $$ --- # 112. Claim Boundary As of this paper: Established: - Exact angle fingerprint； - X fingerprint； - X4 joint； - shared-edge coupling； - scale-coupling semantics； - body-X signature； - PerfectClick predicate； - X4 constraint domain； - DMSD block semantics； - summary / join / lazy refinement； - sound-pruning requirement； - search certificate； - baseline and acceptance roadmap. Not yet proved: - That X4-DMSD necessarily outperforms every existing Perfect Cuboid search; - The multiscale block sparsity conjecture; - That scale jumps can reach arbitrarily large heights while remaining practical; - That a Perfect Cuboid exists; - That a PerfectClick witness has been found; - That every pruning operator has been formally proved sound. --- # 113. Conclusion The core of X4-DMSD is not: $$ \text{Running brute force out to larger numbers}. $$ Instead: $$ \boxed{ \text{First change what merits being searched.} } $$ Ordinary element search asks: $$ \text{Which tuple }(a,b,c)\text{ works?} $$ X4 search first asks: $$ \boxed{ \text{Which X–four-edge joints are eligible for mortise-and-tenon coupling with each other?} } $$ DMSD then asks: $$ \boxed{ \text{Which enormous-scale constraint blocks merit expansion?} } $$ The whole method can therefore be written as: $$ \boxed{ \text{Pythagorean / X Seeds} \rightarrow \text{Exact X4 Joints} \rightarrow \text{Shared-Edge Coupling} \rightarrow \text{DMSD Constraint Blocks} \rightarrow \text{Sound Pruning} \rightarrow \text{Lazy Refinement} \rightarrow \text{PerfectClick}. } $$ where: $$ \boxed{ \operatorname{PerfectClick}\in\{0,1\}. } $$ No tolerance. No "just short". No "almost closed". Only: $$ \boxed{ \text{Click.} } $$ or: $$ \boxed{ \text{No click.} } $$ This is the first version of X4-DMSD's foundational methodology. --- ## Next Phase **X4-DMSD P1 — Exact X4 Joint Algebra** The next paper should formally complete: 1. Pythagorean seed canonicalization; 2. $\chi$ / $\Xi$ canonical form; 3. opposite-edge semantics; 4. shared-edge scale equations; 5. triple-ratio-cycle theorem; 6. exact joint hash; 7. local-to-global reconstruction; 8. Euler brick and known near-hit controls. After P1 is complete, work on the actual X4 runtime can begin.