Finite-grid Hom-module and Hom-cokernel dimensions
Let k be any field and S a finite bounded lattice. The interval-pair poset is
P = { (a,b) ∈ WithTop S × WithTop S | a ≤ b },
ordered coordinatewise, where WithTop S adjoins a genuinely new top element. Let A be its
actual incidence algebra. Let e be the sum of the vertex idempotents δ_(p,p) indexed by
diagonal pair vertices p = (a,a), where a ∈ WithTop S. Let I be the two-sided ideal
generated by e (equivalently, by these diagonal-pair vertex idempotents). Write C = (A/I)ᵐᵒᵖ.
The finite hook generator L = HookSum has opposite
endomorphism algebra T = (End L)ᵐᵒᵖ, which is identified with C by the existing hook-
endomorphism algebra equivalence. The objects of ModuleCat T and ModuleCat C are left
modules, encoding the actual right-module convention.
For p ∈ P, the minimal complement consists of the minimal elements of
{a ∈ P | ¬a ≤ p}. The number D(P) = maxComplementCard (S := P) is the maximum cardinality of
these minimal complements over all p ∈ P. The classification and grid counts are given in
minimal-complement cardinalities in finite grids.
For a persistence module M, let H(M) = homFunctor.obj M be the actual natural-transformation
module from L, and let K(M) be the actual categorical cokernel of H applied to the
two-injective differential of M. If M is pointwise finite-dimensional, the library proves
that the canonical restricted-field dimension of K(M) is finite and
HasProjectiveDimensionLE K(M) (D(P)). The bound uses the actual quotient-module action through
the algebra equivalence and the existing diagonal-quotient projective-dimension theorem. The
scalar restriction and its comparison with the natural Hom-space action are constructed with
explicit identity-on-vectors linear equivalences; neither compatible scalar actions nor a
projective-dimension estimate is an input.
For an actual coordinate grid G : CoordinateGrid n with nonempty factors and n ≥ 1, a
pointwise finite-dimensional M : PModule k G.Point satisfies
HasProjectiveDimensionLE K(M) (2*n) and
HasProjectiveDimensionLE (H(M)) (2*n-2). Every supplied certified minimal finite-rank-projective
resolution of M has zero termAt j for j > 2*n-2. This is a highest-nonzero-degree statement,
so adding terminal zero terms does not affect it; no bound on the stored length field is claimed.
The bounds hold over arbitrary fields and do not require a finite coefficient field or a
finite-dimensionality premise beyond pointwise finite-dimensionality of M.
These results are the ordinary ambient projective dimensions of the actual modules. Their
real-coordinate minimal-resolution transfer is proved separately in
the finite-presentation resolution result,
using the actual finite-grid Kan reconstruction and zero-padded uniqueness. Together they prove
the 2*n-2 upper degree bound for every certified minimal resolution of an originally finitely
presented real-coordinate module. These general upper-bound theorems do not prove a lower witness
or equality for 2*n-2, and they do not establish the final signed-stability result or a sharp
bound for arbitrary grids. The grid count used here is the existing finite-grid
complement-cardinality bound.
An explicit module on the finite Boolean square has a hook-Hom object of exact projective
dimension one. Set Square = Bool × Bool and let M be the principal-downset module supported
at its bottom. Over every field, M is not
rank projective, while its actual hook-generator Hom module has ordinary categorical projective
dimension exactly one. The obstruction is the nonsplit rank-exact quotient from the sum of the two
hooks ending at the incomparable atoms. The existing Hom comparison turns rank nonprojectivity of
M into ordinary nonprojectivity of its Hom module. The finite-lattice Hom bound and the
minimal-resolution dimension criterion then show that every certified minimal finite
rank-projective resolution of M has a nonzero degree-one term and zero terms above degree one.
Finite-poset resolution existence derives an actual certificate with these properties from the
field alone. This is a witness for this Boolean square, not a sharpness result for the general
coordinate-grid bound or an ordinary dimension claim for M. The fixture is
TDA.Persistence.FiniteRankProjectiveHomNonprojectiveExamples; use open scoped
TDA.Persistence.FiniteRankProjectiveHomNonprojectiveExamples to select the preorder category on
Bool × Bool. See also the API guide's actual Hom section.
The source context is the upper-bound argument of Theorem 5.16 and the upper-bound consequence of Proposition 5.17 and Theorem 5.18 in BOOS24-v3, together with Lemma 5.11's Hom and resolution comparison. The library reuses its actual two-injective rows and minimal-resolution comparison, proves finite-dimensionality and quotient transport for the actual Hom cokernel, and specializes the resulting bounds to finite coordinate grids. The real-coordinate transfer is stated and documented separately.