Ladder Determinantal Modules

• Ladder determinantal modules are direct sums of maximal-minor ideals defined by staircase-shaped ladder matrices, establishing a bridge between determinantal geometry and module theory.
• Their special fiber rings often degenerate to Hibi rings, enabling explicit computation of invariants like dimension, regularity, and multiplicity while also providing sharp Gorenstein criteria.
• Recent research employs blowup algebras, reduced Gröbner bases, and semidualizing module theory to reveal deep homological properties and combinatorial structures in ladder determinantal rings.

Ladder determinantal modules are algebraic objects attached to ladder-shaped patterns of variables in a matrix. In current commutative-algebra usage, the term most often denotes modules of the form $M=\bigoplus_{i=1}^r L$, where $L=I_n(X_S)$ is the ideal of maximal minors of a generic ladder matrix $X_S$ (Costantini et al., 29 Jul 2025). Closely related literature also uses the phrase for standard graded quotient modules $\Bbbk[L]/I_{L,\mathbf r}$ attached to one-sided mixed ladder determinantal ideals (Rajchgot et al., 2022), and studies semidualizing modules over ladder determinantal rings $A_t(Y)=A[Y]/I_t(Y)$ (Sather-Wagstaff et al., 2020). Across these settings, the subject links determinantal geometry, distributive lattices, Hibi rings, Gröbner and SAGBI degenerations, tableau combinatorics, and the homological theory of blowup algebras.

1. Ladder matrices, determinantal ideals, and module conventions

A ladder matrix is specified by intervals

$S_i=[u_i,v_i]\subseteq [m]\qquad (i=1,\dots,n),$

with

$$1=u_1\le u_2\le\cdots\le u_n\le m,\qquad 1\le v_1\le\cdots\le v_n=m,$$

and, after simplification,

$$u_{i-1}<u_i,\quad v_{i-1}<v_i,\quad u_i\le v_{i-1}+1\quad\text{for }i\ge 2.$$

The associated $n\times m$ ladder matrix $X_S$ has entry $L=I_n(X_S)$0 when $L=I_n(X_S)$1 and $L=I_n(X_S)$2 otherwise, so the nonzero positions form a staircase-shaped region. Writing

$L=I_n(X_S)$3

the maximal-minor ideal is

$L=I_n(X_S)$4

generated minimally by the nonzero $L=I_n(X_S)$5 minors indexed by

$L=I_n(X_S)$6

Recent work fixes $L=I_n(X_S)$7 and studies the ladder determinantal module

$L=I_n(X_S)$8

as a multigraded module built from $L=I_n(X_S)$9 copies of the same determinantal ideal (Costantini et al., 29 Jul 2025).

The same phrase occurs in adjacent senses. For one-sided mixed ladders, a standard graded module of interest is

$X_S$0

where $X_S$1 is generated by minors of prescribed sizes in nested ladder submatrices (Rajchgot et al., 2022). In another direction, the theory of semidualizing modules studies rank-one reflexive modules over ladder determinantal rings

$X_S$2

where $X_S$3 is a ladder in a generic matrix and $X_S$4 is generated by $X_S$5 minors lying entirely in $X_S$6 (Sather-Wagstaff et al., 2020).

The combinatorial parameters

$X_S$7

and, in the 2026 Gorenstein analysis,

$X_S$8

encode the widths of ladder rows and the step sizes along the left and right boundary (Costantini et al., 29 Jul 2025, Fouli et al., 20 Jan 2026). These parameters govern dimension, regularity, canonical data, multiplicity, and Gorensteinness.

2. Special fiber rings and Hibi-ring degeneration

For equigenerated ideals $X_S$9, the multi-Rees algebra and special fiber ring are

$\Bbbk[L]/I_{L,\mathbf r}$0

In the ladder determinantal case $\Bbbk[L]/I_{L,\mathbf r}$1, one presents $\Bbbk[L]/I_{L,\mathbf r}$2 as

$\Bbbk[L]/I_{L,\mathbf r}$3

with $\Bbbk[L]/I_{L,\mathbf r}$4 and defining ideal $\Bbbk[L]/I_{L,\mathbf r}$5 (Costantini et al., 29 Jul 2025).

A lex order $\Bbbk[L]/I_{L,\mathbf r}$6 on $\Bbbk[L]/I_{L,\mathbf r}$7 makes every maximal minor diagonal: $\Bbbk[L]/I_{L,\mathbf r}$8 Setting $\Bbbk[L]/I_{L,\mathbf r}$9, one obtains a second special fiber ring

$A_t(Y)=A[Y]/I_t(Y)$0

A key theorem identifies $A_t(Y)=A[Y]/I_t(Y)$1 with a Hibi ring over the distributive lattice $A_t(Y)=A[Y]/I_t(Y)$2: $A_t(Y)=A[Y]/I_t(Y)$3 Here the standard binomials form a Gröbner basis, and the ring is Cohen–Macaulay, normal, and Koszul (Costantini et al., 29 Jul 2025).

The SAGBI step is decisive. The generators $A_t(Y)=A[Y]/I_t(Y)$4 form a finite SAGBI basis for $A_t(Y)=A[Y]/I_t(Y)$5, with

$A_t(Y)=A[Y]/I_t(Y)$6

A further Gröbner degeneration gives

$A_t(Y)=A[Y]/I_t(Y)$7

where $A_t(Y)=A[Y]/I_t(Y)$8 is a squarefree quadratic monomial ideal. Consequently,

$A_t(Y)=A[Y]/I_t(Y)$9

coincide with the corresponding invariants of the Hibi ring and of the monomial degeneration $S_i=[u_i,v_i]\subseteq [m]\qquad (i=1,\dots,n),$0 (Costantini et al., 29 Jul 2025).

This Hibi-ring viewpoint is also central to the 2026 Gorenstein criterion. There the initial special fiber $S_i=[u_i,v_i]\subseteq [m]\qquad (i=1,\dots,n),$1 is again a Hibi ring on the distributive lattice $S_i=[u_i,v_i]\subseteq [m]\qquad (i=1,\dots,n),$2, and Gorensteinness is reduced to purity of the poset of join-irreducibles (Fouli et al., 20 Jan 2026).

3. Dimension, regularity, $S_i=[u_i,v_i]\subseteq [m]\qquad (i=1,\dots,n),$3-invariant, and multiplicity

For $S_i=[u_i,v_i]\subseteq [m]\qquad (i=1,\dots,n),$4, the analytic spread is

$S_i=[u_i,v_i]\subseteq [m]\qquad (i=1,\dots,n),$5

This is obtained by describing maximal cliques of a graph $S_i=[u_i,v_i]\subseteq [m]\qquad (i=1,\dots,n),$6 on $S_i=[u_i,v_i]\subseteq [m]\qquad (i=1,\dots,n),$7, where maximal cliques are precisely maximal chains in the distributive lattice between

$S_i=[u_i,v_i]\subseteq [m]\qquad (i=1,\dots,n),$8

The same clique combinatorics control regularity. If $S_i=[u_i,v_i]\subseteq [m]\qquad (i=1,\dots,n),$9 is the distinguished maximal clique of minimal sequence index $$1=u_1\le u_2\le\cdots\le u_n\le m,\qquad 1\le v_1\le\cdots\le v_n=m,$$0, then

$$1=u_1\le u_2\le\cdots\le u_n\le m,\qquad 1\le v_1\le\cdots\le v_n=m,$$1

Since $$1=u_1\le u_2\le\cdots\le u_n\le m,\qquad 1\le v_1\le\cdots\le v_n=m,$$2 is Cohen–Macaulay,

$$1=u_1\le u_2\le\cdots\le u_n\le m,\qquad 1\le v_1\le\cdots\le v_n=m,$$3

Special cases admit closed forms for $$1=u_1\le u_2\le\cdots\le u_n\le m,\qquad 1\le v_1\le\cdots\le v_n=m,$$4: if all offsets satisfy $$1=u_1\le u_2\le\cdots\le u_n\le m,\qquad 1\le v_1\le\cdots\le v_n=m,$$5, then $$1=u_1\le u_2\le\cdots\le u_n\le m,\qquad 1\le v_1\le\cdots\le v_n=m,$$6; if all $$1=u_1\le u_2\le\cdots\le u_n\le m,\qquad 1\le v_1\le\cdots\le v_n=m,$$7, then $$1=u_1\le u_2\le\cdots\le u_n\le m,\qquad 1\le v_1\le\cdots\le v_n=m,$$8 (Costantini et al., 29 Jul 2025).

Multiplicity is governed by maximal cliques and tableaux. One has

$$1=u_1\le u_2\le\cdots\le u_n\le m,\qquad 1\le v_1\le\cdots\le v_n=m,$$9

and, separating the $$u_{i-1}<u_i,\quad v_{i-1}<v_i,\quad u_i\le v_{i-1}+1\quad\text{for }i\ge 2.$$0 case,

$$u_{i-1}<u_i,\quad v_{i-1}<v_i,\quad u_i\le v_{i-1}+1\quad\text{for }i\ge 2.$$1

For $$u_{i-1}<u_i,\quad v_{i-1}<v_i,\quad u_i\le v_{i-1}+1\quad\text{for }i\ge 2.$$2, the ladder determines a skew partition $$u_{i-1}<u_i,\quad v_{i-1}<v_i,\quad u_i\le v_{i-1}+1\quad\text{for }i\ge 2.$$3 via

$$u_{i-1}<u_i,\quad v_{i-1}<v_i,\quad u_i\le v_{i-1}+1\quad\text{for }i\ge 2.$$4

and

$$u_{i-1}<u_i,\quad v_{i-1}<v_i,\quad u_i\le v_{i-1}+1\quad\text{for }i\ge 2.$$5

Using the Morales–Pak–Panova hook-length-type formula,

$$u_{i-1}<u_i,\quad v_{i-1}<v_i,\quad u_i\le v_{i-1}+1\quad\text{for }i\ge 2.$$6

hence

$$u_{i-1}<u_i,\quad v_{i-1}<v_i,\quad u_i\le v_{i-1}+1\quad\text{for }i\ge 2.$$7

In the generic full-matrix case, the skew shape becomes ordinary rectangular, and the formula reduces to the classical hook-length formula for the degree of the Grassmannian. This gives a direct generalization of the Grassmannian degree formula from standard Young tableaux to standard skew Young tableaux (Costantini et al., 29 Jul 2025).

For a generic $$u_{i-1}<u_i,\quad v_{i-1}<v_i,\quad u_i\le v_{i-1}+1\quad\text{for }i\ge 2.$$8 matrix, with $$u_{i-1}<u_i,\quad v_{i-1}<v_i,\quad u_i\le v_{i-1}+1\quad\text{for }i\ge 2.$$9 and $n\times m$0,

$n\times m$1

$n\times m$2

For certain $n\times m$3 sparse matrices, the formulas recover earlier expressions for dimension, regularity, and $n\times m$4-invariant (Costantini et al., 29 Jul 2025).

4. Gorenstein special fiber rings

The 2026 analysis characterizes exactly when the special fiber ring of a ladder determinantal module is Gorenstein. For the module

$n\times m$5

attached to the generic ladder matrix $n\times m$6, one studies the join-irreducible poset $n\times m$7 of the distributive lattice $n\times m$8, and its enlargement $n\times m$9 for the Hibi ring of the initial special fiber (Fouli et al., 20 Jan 2026).

For $X_S$0, the following are equivalent: $X_S$1 Thus the problem splits into two parts: Gorensteinness of the single-copy special fiber $X_S$2, and a compatibility condition linking the number of summands $X_S$3 to the rank of the join-irreducible poset (Fouli et al., 20 Jan 2026).

When the join-irreducible poset is connected, equivalently $X_S$4 for all $X_S$5, $X_S$6 is Gorenstein iff

$X_S$7

and

$X_S$8

An equivalent formulation is given purely in terms of the step sequences $X_S$9 and $L=I_n(X_S)$00: whenever $L=I_n(X_S)$01, the intervening $L=I_n(X_S)$02-values are all $L=I_n(X_S)$03 until the next jump, and similarly for the $L=I_n(X_S)$04-sequence (Fouli et al., 20 Jan 2026).

In general, the ladder decomposes into blocks determined by

$L=I_n(X_S)$05

with row intervals $L=I_n(X_S)$06. Then $L=I_n(X_S)$07 is Gorenstein iff each block satisfies the connected-case conditions and, in addition,

$L=I_n(X_S)$08

for all $L=I_n(X_S)$09, where

$L=I_n(X_S)$10

Finally, for the full module $L=I_n(X_S)$11, $L=I_n(X_S)$12 is Gorenstein iff every block satisfies those conditions and

$L=I_n(X_S)$13

for all $L=I_n(X_S)$14 (Fouli et al., 20 Jan 2026).

A particularly transparent consequence concerns the generic matrix. If $L=I_n(X_S)$15 for all $L=I_n(X_S)$16, so the ladder is the full generic $L=I_n(X_S)$17 matrix, then

$L=I_n(X_S)$18

When Gorensteinness holds,

$L=I_n(X_S)$19

Moreover, since $L=I_n(X_S)$20 is already $L=I_n(X_S)$21-rational in positive characteristic, Gorensteinness upgrades this to $L=I_n(X_S)$22-regularity (Fouli et al., 20 Jan 2026).

5. Blowup algebras, Gröbner bases, and reduced Gröbner structures

The 2024 blowup-algebra study treats modules

$L=I_n(X_S)$23

for a two-sided ladder determinantal ideal $L=I_n(X_S)$24, and

$L=I_n(X_S)$25

for a unit interval determinantal ideal $L=I_n(X_S)$26. The main objects are the multi-Rees algebra $L=I_n(X_S)$27 and the special fiber ring $L=I_n(X_S)$28 (Lin et al., 2024).

For the initial monomial modules $L=I_n(X_S)$29, the special fiber ideal has a Gröbner basis of Plücker-type quadratic binomials, and the Rees ideal has a Gröbner basis consisting of Eagon–Northcott-type linear relations together with those Plücker-type quadratic relations. A SAGBI lifting then transfers these defining equations to the actual multi-Rees algebra and special fiber of $L=I_n(X_S)$30. The resulting blowup algebras are of fiber type, and their defining ideals admit quadratic Gröbner bases (Lin et al., 2024).

The consequences are strong. For $L=I_n(X_S)$31,

• $L=I_n(X_S)$32 and $L=I_n(X_S)$33 are Koszul Cohen–Macaulay normal domains;
• in characteristic $L=I_n(X_S)$34 they have rational singularities;
• in positive characteristic they are $L=I_n(X_S)$35-rational;
• the powers of $L=I_n(X_S)$36 have linear resolutions.

For $L=I_n(X_S)$37, the paper establishes the analogous statements for the special fiber ring $L=I_n(X_S)$38, while noting that the full multi-Rees algebra is subtler because additional Koszul-type relations are expected beyond Eagon–Northcott and Plücker relations (Lin et al., 2024).

Reduced Gröbner-basis phenomena appear in a complementary 2023 study of blockwise determinantal ideals. There, one-sided ladder determinantal ideals are identified with Schubert determinantal ideals of vexillary permutations, and the elusive minors form the reduced Gröbner basis for any diagonal or anti-diagonal term order. For two-sided ladder determinantal ideals, Gorla’s generating set is shown to be the reduced Gröbner basis for lexicographic diagonal scanning orders. The characteristic pair is strong, and in the one-sided case the $L=I_n(X_S)$39-characteristic set is normal (Mou et al., 2023). This places ladder determinantal ideals within a general Gröbner-theoretic framework and clarifies the initial ideals used in degeneration arguments for ladder determinantal modules.

6. Semidualizing modules and quotient-module perspectives

A semidualizing module $L=I_n(X_S)$40 over a commutative noetherian ring satisfies

$L=I_n(X_S)$41

For a normal domain, semidualizing modules are exactly rank-one reflexive modules with vanishing self-Ext, so they define classes in the divisor class group (Sather-Wagstaff et al., 2019). Ladder determinantal rings are normal domains, and this makes semidualizing modules a natural module-theoretic refinement of their divisor-class structure (Sather-Wagstaff et al., 2020).

A general tensor-product mechanism describes them. If $L=I_n(X_S)$42 is a PID, $L=I_n(X_S)$43 a normal domain containing $L=I_n(X_S)$44, and $L=I_n(X_S)$45 a ladder, then

$L=I_n(X_S)$46

If $L=I_n(X_S)$47 is semidualizing over $L=I_n(X_S)$48 and $L=I_n(X_S)$49 semidualizing over $L=I_n(X_S)$50, then

$L=I_n(X_S)$51

is semidualizing over $L=I_n(X_S)$52; when $L=I_n(X_S)$53 is a field, different pairs give distinct semidualizing classes (Sather-Wagstaff et al., 2019). For arbitrary ladders and all minor sizes, semidualizing modules over $L=I_n(X_S)$54 decompose as

$L=I_n(X_S)$55

where $L=I_n(X_S)$56 is the number of non-Gorenstein ladder components in a suitable decomposition of $L=I_n(X_S)$57 (Sather-Wagstaff et al., 2020).

Over a field, large classes of ladder determinantal rings have only trivial semidualizing modules. For one-sided ladders and arbitrary $L=I_n(X_S)$58,

$L=I_n(X_S)$59

and for two-sided ladders with $L=I_n(X_S)$60 and no coincidental inside corners,

$L=I_n(X_S)$61

In these cases the only semidualizing modules are the ring itself and, when it exists, the canonical module (Sather-Wagstaff et al., 2018).

The $L=I_n(X_S)$62 case with coincidental inside corners is different. If a 2-connected ladder decomposes as

$L=I_n(X_S)$63

into 2-connected pieces without coincidental inside corners, then

$L=I_n(X_S)$64

and therefore

$L=I_n(X_S)$65

where $L=I_n(X_S)$66 is the number of non-Gorenstein factors among the $L=I_n(X_S)$67. In particular, every power $L=I_n(X_S)$68 occurs for suitable $L=I_n(X_S)$69 ladder determinantal rings (Sather-Wagstaff et al., 2018).

A different module-theoretic perspective comes from one-sided mixed ladder determinantal ideals. There the standard graded quotient module

$L=I_n(X_S)$70

is Cohen–Macaulay, and its Castelnuovo–Mumford regularity is computed through a vexillary permutation $L=I_n(X_S)$71: $L=I_n(X_S)$72 where $L=I_n(X_S)$73 is defined from the rank filling of the Young diagram $L=I_n(X_S)$74, and $L=I_n(X_S)$75 denotes the size of the largest antidiagonal path (Rajchgot et al., 2022). This connects ladder quotient modules to Grassmannian Kazhdan–Lusztig ideals and vexillary Grothendieck polynomials, and shows that ladder determinantal modules also sit naturally inside Schubert-theoretic and $L=I_n(X_S)$76-theoretic combinatorics.

Taken together, these developments show that ladder determinantal modules form a broad family rather than a single construction. Direct sums of maximal-minor ideals lead to special fiber rings with Hibi degenerations, explicit invariants, and sharp Gorenstein criteria (Costantini et al., 29 Jul 2025, Fouli et al., 20 Jan 2026). Blowup-algebra methods reveal quadratic Gröbner presentations and strong homological properties (Lin et al., 2024). Semidualizing-module theory exposes a rigid but highly structured interaction with divisor class groups (Sather-Wagstaff et al., 2020, Sather-Wagstaff et al., 2018). Quotient-module formulations connect ladders to Schubert and Grothendieck combinatorics (Rajchgot et al., 2022). This suggests that the topic is best understood as a common interface between sparse determinantal geometry, toric degeneration, and the homological theory of modules over ladder determinantal rings.