Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
Inclusion Matrices and Chains
corollary_3: Ghorbani, Khosrovshahi, Maysoori and Mohammad-Noori's chain-based proof of Wilson's theorem that for t <= k <= v - t the t-subset versus k-subset inclusion matrix W_{tk} has a diagonal form with entries C(k-i,t-i) of multiplicity C(v,i) - C(v,i-1), i = 0, ..., t.
corollary_4: Ghorbani, Khosrovshahi, Maysoori and Mohammad-Noori's chain-based proof of Wilson's theorem that for t <= k <= v - t the system W_{tk} x = b has an integral solution exactly when R_{it} b is divisible by C(k-i,t-i) for every i = 0, ..., t.
theorem_1: Ghorbani, Khosrovshahi, Maysoori and Mohammad-Noori's theorem that, after each t-subset row label of the inclusion matrix W_{tk}(v) is replaced by the full-rank bottom of its rank chain, the resulting matrix has Smith normal form (I | O) with I of order C(v,t) whenever t <= k <= v - t, with Corollaries 1 and 2 on its p-rank and row space.
theorem_2: Ghorbani, Khosrovshahi, Maysoori and Mohammad-Noori's theorem that, for t <= k <= v - t, the inclusion matrix obtained from W_{tk} by replacing each k-subset column label by the top of its chain in the complemented rank-chain decomposition has Smith form (I | O), hence full p-rank for every prime p, and that its equation W x = lambda 1 corresponds to the signed-design equation W_{tk} x = lambda 1.
The copy read for this card is arXiv:0709.3144v1 (20 September 2007), 15 pages. The arXiv record carries no license field, so arXiv's assumed license applies (arXiv:0709.3144), every other right reserved.
E. Ghorbani, G. B. Khosrovshahi, Ch. Maysoori, M. Mohammad-Noori, "Inclusion Matrices and Chains," arXiv:0709.3144 (2007). Published in J. Combin. Theory Ser. A 115 (2008), 878--887, DOI 10.1016/j.jcta.2007.09.002.
Overview
The paper studies the ordinary inclusion matrix , whose rows and columns are indexed by the - and -subsets of , with entry precisely for containment (Section 1). Its motivating problem is to replace the signed-design system (equation (1)) by an integrally simpler system. The authors construct canonical row and column modifications of using a rank-preserving symmetric-chain decomposition of the Boolean lattice and determine their Smith forms.
For a finite set of positive integers, Section 2 presents Frankl's rank through a modified two-row tableau. The top row contains the elements of ; the entry below is the largest unused positive integer smaller than , or a formal symbol if none exists (equation (2)). The resulting identities are equation (3). This is presented as an alternative realization of the preceding lattice-walk definition of rank; no separately numbered theorem is given for their equivalence.
Section 3 defines a successor by adjoining the least positive integer absent from the tableau and, when is not full-rank, a predecessor by deleting the entry above the rightmost formal symbol. Iteration partitions into skipless symmetric chains , where every member has rank and the minimal member is full-rank. Remark 1 gives direct descriptions of iterated successors and predecessors. Remark 2 counts, for , the full-rank subsets of of rank at most as , and hence those of rank exactly , equivalently the chains of rank , as . The construction is identified in the Introduction with the classical de Bruijn decomposition, but the paper supplies its own explicit rank-tableau description.
In Section 4 each row label of is replaced by the minimal member of its chain, producing . If is the inclusion matrix from full-rank -sets to all -sets, then equation (5) decomposes into the blocks . The incidence-counting identity is equation (6), and equation (8) packages these identities as , where has with multiplicity .
Theorem 1 proves, for , that the Smith normal form of is , with of order . Its proof identifies the columns indexed by sets of rank at most as a square unimodular submatrix and establishes this recursively using equations (9) and (10). Consequently has full rank over every prime field (Corollary 1). Equation (8) further gives equality of the row spaces of and over fields whose characteristic divides none of the numbers (Corollary 2); this characteristic restriction is part of the stated result.
Corollary 3 recovers Wilson's diagonal form for when : its diagonal entries are with multiplicity for . This is a diagonal form, not asserted there to be the invariant-factor-ordered Smith form. Corollary 4 gives, for , the exact integral-solvability criterion for : for every , the vector must be integral; sufficiency follows from the unimodular minor in Theorem 1 and equation (13). Remark 3 specializes the transformed system to signed designs, while Remark 4 identifies the rows of as a concrete basis for the row space obtained by stacking . The paper explicitly attributes the diagonal-form result to Wilson and the integral-solvability theorem to Wilson and, in the signed-design case, Graver–Jurkat; it also says the Smith form of is implicit in the work of Bier and credits Bier with the concrete basis of Remark 4. Its contribution is the chain-based derivation.
Section 5 applies the same construction after complementation. Complementing every set of every rank chain gives a second chain partition, and replacing each column label by the largest member of its chain in that partition gives , decomposed by column size in equation (14). Equation (15) is the corresponding weighted intertwining identity. For , Theorem 2(i) proves that also has Smith form and full rank in every characteristic. Theorem 2(ii) states that is equivalent, via , to equation (1). The scope is finite Boolean-lattice incidence, integral linear algebra, and signed designs; the paper contains no result about additive relations in sets of integers.
Relation to E774
For E774, take a finite subset of the candidate infinite set and identify subsets of with subsets of the index set . Under this identification, records only which -element supports are contained in which -element supports. It is completely independent of the integer values .
A failure of dissociation, by contrast, is value-dependent: it is represented by a nonzero signed coefficient vector on giving an additive relation, equivalently by disjoint index sets with equal corresponding sums. Neither this equality nor its coefficients occur in . Proportionate dissociation asks for uniformly large relation-free subcollections of every finite , while the finite-union conclusion asks for a bounded coloring with no monochromatic signed relation. The paper proves neither an extraction theorem nor a coloring/partition theorem of this kind.
The potentially usable ingredient is organizational rather than decisive. Section 3 gives a canonical symmetric-chain stratification of all supports by cardinality, and Theorem 1 supplies unimodular bases for uniform containment constraints. Thus, if an E774 argument reduces an auxiliary counting or averaging step to a system exactly of the form , Corollary 4 can decide integral solvability through the divisibility conditions ; Corollary 3 can likewise diagonalize the associated incidence operator. The full-rank conclusions in Corollary 1 and Theorem 2(i) may prevent modular degeneracy in such an auxiliary system.
These tools do not control the hypergraph of signed additive relations, whose edges depend on the values in and may have unbounded size. In particular, the rank in Sections 2–3 is a combinatorial rank of an index subset, not additive rank or dissociativity. Remark 6 also warns that the full-rank property is special to the authors' chain decomposition and does not hold for arbitrary symmetric-chain decompositions. Accordingly, the relation to E774 is weak and hypothetical: the paper is relevant as a source of incidence-matrix and chain machinery, but it supplies no torsion-free additive input and does not resolve, or directly advance, the stated finite-union problem.
Read status: claims checked for the definitions of Sections 1 to 5, Theorems 1 and 2 and Corollaries 1 to 4, read clause by clause on the page images of arXiv:0709.3144v1; the proofs of Theorem 1 and Corollaries 3 and 4 followed, the argument for Theorem 2 read for structure. The journal version was not compared. Nothing here is independently reviewed. Result pages: theorem_1, corollary_3, corollary_4 and theorem_2.
Bears on. E0774: the paper proves nothing about dissociated sets or additive relations, and decides nothing about the problem.
Results.
- Theorem 1 (p. 8), with Corollaries 1 and 2 (p. 10): has Smith normal form for .
- Corollary 3 (p. 10): Wilson's diagonal form of for .
- Corollary 4 (p. 11): Wilson's criterion for an integral solution of for .
- Theorem 2 (p. 14): has Smith form , and corresponds to the signed-design equation (1), for .
No file of this source is held: no license on record permits its redistribution, and the card cites the edition it names above.