THE F -PURE THRESHOLD OF A SCHUBERT CYCLE

JUSTIN FONG AND MITSUHIRO MIYAZAKI

arXiv:2502.09559v1 [[Link]] 13 Feb 2025

Abstract. The F -pure threshold is the characteristic p counter part of the log canon-
ical threshold in characteristic zero. It is a numerical invariant associated to the singu-
larities of a variety, hence computing its value is important. We give a closed formula
for the F -pure threshold of the irrelevant maximal ideal of Schubert cycles, which
are the homogeneous coordinate rings of Schubert subvarieties of a Grassmannian.
The main point of the computation is to give an explicit formula for the a-invariant
of a Schubert cycle. The derivation of both formulas is made possible through the
combinatorics of the underlying poset of these rings.

1. Introduction
The F -pure threshold is a numerical invariant associated to the F -singularity class
of F -purity. It was first introduced in [TW04] as the positive characteristic counterpart
of the log canonical threshold of an ideal in characteristic zero, which is a measurement
of a singularity on the variety defined by the ideal. Although their definitions differ,
many important connections have been established between these two invariants, one
of which is that the log canonical threshold is approximated by the F -pure threshold
through reduction modulo p [MTW05].
The F -pure threshold is defined as follows: Let N = {0, 1, 2 . . . }, and S = k[x1 , . . . , xn ]
be a polynomial ring over an F -finite field k of characteristic p > 0 with irrelevant max-
imal ideal mS = (x1 , . . . , xn ), and I be a homogeneous ideal such that R = S/I is an
e e
F -pure domain. For any ideal J define its pe -th bracket power J [p ] = (j p : j ∈ J).
Given a homogeneous ideal a of R with preimage J in S, for each e ∈ N, define the
integer n o
e [pe ]
νe (a) = max r ∈ N : J r (I [p ] :S I) * mS .
Definition 1.1. The F -pure threshold of a ⊂ R is defined to be the limit
νe (a)
fpt(a) = lim .
e→∞ pe

When a = m is the irrelevant maximal ideal of R, we simply call fpt(m) the F -pure
threshold of R. The values of fpt(m) are relatively unknown, especially when R is
non-Gorenstein, since otherwise fpt(m) = −a(R) (when R is standard graded and a(R)
is the a-invariant of R), although successful computations of fpt(m) have been done
in [TW04], [Hir09], [MOY10], [CM15], [STV17], [CMSV18]. In this paper, we give a
formula for the F -pure threshold of the homogeneous coordinate ring of a Schubert
variety defined by γ inside a Grassmannian, shortened to Schubert cycle, and denoted
by Gγ where γ is a certain tuple of integers. We also give a closed formula for the
a-invariant of Gγ . The main results of the paper are as follows.
1
 THE F -PURE THRESHOLD OF A SCHUBERT CYCLE 2

Theorem A (Theorem 4.2, Corollary 3.6). Let γ = [a1 , . . . , ad ] be a tuple of integers

with 1 ≤ a1 < · · · < ad ≤ n, [a1 , . . . , ad ] 6= [n − d + 1, . . . , n], and let γ = [β0 , . . . , βs ] be
its decomposition into blocks of consecutive integers, where each block βi is followed by
a gap χi . For the homogeneous maximal ideal m of Gγ , one has
( i t
)
X X
fpt(m) = min |βj | + |χj | : 0 ≤ i ≤ t ,
j=0 j=i
( i t
)
X X
−a(Gγ ) = max |βj | + |χj | : 0 ≤ i ≤ t ,
j=0 j=i

where t = s if ad < n, and t = s − 1 if ad = n, |βj |, |χj−1| are the sizes of the blocks
and gaps.

Acknowledgments
The first author was partially supported by NSF Grant DMS-2100288. He would like
to thank Aldo Conca, Matteo Varbaro, Linquan Ma, Vaibhav Pandey, and Uli Walther
for helpful discussions.

2. Background
In this section we introduce the necessary background and notations that will be
present throughout this paper. Let R be an N-graded Noetherian normal domain and
S denote the set of nonzero homogeneous elements of R. Then S −1 R is a Z-graded
ring. Let I be a graded fractionary ideal in S −1 R. Then I is said to be divisorial if
I = R :S −1 R (R :S −1 R I), i.e., I is reflexive as an R-module. The set of divisorial ideals
of R, denoted Div(R), form a group by the operation I · J = R :S −1 R (R :S −1 R IJ). For
n ∈ Z we denote I (n) as the n-th power of I in Div(R). Note that if I ( R and n > 0,
then I (n) is the n-th symbolic power of I. If moreover R is Cohen-Macaulay and R0
is a field, then the graded canonical module ω can be realized as a graded fractionary
ideal of R. Since ω is reflexive, ω is divisorial. We call ω (−1) the anticanonical ideal.
The main theorem that we will use to compute F -pure thresholds in this paper is
the following:
Theorem 2.1. [STV17, Theorem 4.1] Let S be an n-dimensional standard graded poly-
nomial ring over an F -finite field k of characteristic p > 0. Let I be a homogeneous ideal
such that R = S/I is an F -pure normal domain, and let ω be the graded canonical mod-
e (e) (e)
ule of R. For each e ∈ N, suppose ω (1−p ) is generated in degrees −d1 < · · · < −dre .
(e) (e)
Then νe (m) ∈ {d1 , . . . , dre }.
2.1. Algebras with Straightening Laws. For the definition and basic facts of alge-
bras with straightening laws (ASL for short), we consult [DCEP82] and [BV88]. How-
ever, we use the term algebra with straightening law as the ordinal Hodge algebra in
[DCEP82]. By [DCEP82, Proposition 1.1], we see the following.
Lemma 2.2. Let A be a graded ASL on Π over B. If x ∈ A is represented as a
polynomial of elements in Π with coefficients in B, x=f (ξ1 , . . . , ξn ), then the standard
 THE F -PURE THRESHOLD OF A SCHUBERT CYCLE 3

representation of x is obtained by repeated application of the straightening relations of

A.
Let Π be a poset and Ω a subset of Π. Ω is called a poset ideal of Π if whenever
α ∈ Ω, β ∈ Π, β < α, one has β ∈ Ω. We see by [DCEP82, Proposition 1.2] the
following.
Lemma 2.3. Let A be a graded ASL on Π over B and Ω a poset ideal of Π. Set
I = AΩ. Then A/I naturally has a graded ASL structure on Π \ Ω over B.
Let A be a graded ASL on a poset Π over a commutative ring B. Suppose Ω is
a poset ideal in Π, and let I = AΩ. We say that I (or Ω) is straightening-closed if
every standard monomial appearing in the staightening relation of δτ for any δ, τ ∈ Ω
incomparable in Π have at least two factors in Ω.
Lemma 2.4. Suppose A is a graded ASL on a poset Π over a commutative ring B, Ω
a straightening-closed poset ideal of Π, and I = AΩ. Let u ∈ N. If z is an arbitrary
element of A, then z ∈ I u if and only if every standard monomial appearing in the
standard representation of z contains at least u factors in Ω.
P
Proof. Suppose α, β ∈ Π are incomparable, and let αβ = µ cµ µ be the straightening
relation. If one of α or β belongs to Ω, then every standard monomial µ appearing
in the straightening relation of αβ contains a factor from Ω, since each µ contains a
factor ζ ∈ Π with ζ ≤ α and ζ ≤ β, and Ω is a poset ideal. If both α and β belong
to Ω, then by the straightening-closed property of Ω, each standard monomial µ in the
straightening relation of αβ contains at least two factors from Ω.
By Lemma 2.2, the standard representation of an element in A can be obtained by
repeated use of the straightening relations. Therefore, given a monomial in Π with at
least u factors in Ω, every standard monomial appearing in its standard representation
has at least u factors in Ω. 
Remark 2.5. The statement of Lemma 2.4 is equivalent to the fact that I u is generated
as a B-module by the standard monomials containing at least u factors in Ω. In
particular, as an ideal of A, I u is generated by the standard monomials of length u in
Ω.
Proposition 2.6. Suppose A is a graded ASL on a poset Π over a commutative ring
B. Let Ω1 , . . . , Ωn be straightening-closed poset Tn ideals of Π, and set Ii = AΩi for
ui
i = 1, . . . , n. For any u1 , . . . , un ∈ N, the ideal i=1 Ii of A is generated by
{ξ1 ξ2 · · · ξu : ξ1 ≤ ξ2 ≤ · · · ≤ ξu , ξ1 , ξ2 , . . . , ξui ∈ Ωi for 1 ≤ i ≤ n} (2.1)
Tn ui
where u = max{u1 , . . . , un }. In particular, if A is a homogeneous ASL, then i=1 Ii
is generated in degree u.
Proof. We will show that I1u1 ∩ · · · ∩ Inun is generated by the set of standard monomials
(2.1) as an ideal in A. It is clear that the set (2.1) is contained in I1u1 ∩ · · · ∩ Inun .
Let x ∈ A be an arbitrary non-zero element and
X
x= bj ξj1 ξj2 · · · ξjrj (ξj1 ≤ ξj2 ≤ · · · ≤ ξjrj )
j
 THE F -PURE THRESHOLD OF A SCHUBERT CYCLE 4

be the standard representation of x. By Lemma 2.4, for each 1 ≤ i ≤ n, one has x ∈ Iiui
if and only if ξj1, ξj2 , . . . , ξjui ∈ Ωi for any j. Thus, I1u1 ∩ · · · ∩ Inun is contained in the
ideal generated by the set (2.1). 

2.2. Schubert Cycles. Suppose k is an algebraically closed field of arbitrary charac-

teristic. Let X = (xij ) be a generic d × n matrix with 1 ≤ d ≤ n. The homogeneous
coordinate ring of the Grassmannian Gr(d, V ) of d-dimensional subspaces of V , where
V is an n-dimensional vector space over k, with respect to the Plücker embedding, is
the k-algebra G(X) generated by the maximal minors of X. The maximal minors will
be identified with a tuple of their column indices [a1 , . . . , ad ], hence the set of maximal
minors of X is
Γ(X) = {[a1 , . . . , ad ] : 1 ≤ a1 < · · · < ad ≤ n}, (2.2)
which is a poset with respect to the partial order
[a1 , . . . , ad ] ≥ [b1 , . . . , bd ] ⇔ ai ≥ bi for all i.
It is easily verified that Γ(X) is a distributive lattice.
Let 0 = V0 ⊂ V1 ⊂ · · · ⊂ Vn = V be a flag of subspaces of V with dimk Vi = i.
For a fixed δ = [b1 , . . . , bd ] ∈ Γ(X), the corresponding Schubert variety inside the
Grassmannian Gr(d, V ) is the subvariety
Ω(δ) = {W ∈ Gr(d, V ) : dimk (W ∩ Vbi ) ≥ i, for i = 1, . . . , d}.
Lemma-Definition 2.7. Fix γ = [a1 , . . . , ad ] ∈ Γ(X) and set bi = n − ad−i+1 + 1 for
i = 1, . . . , d. Let γ̃ = [b1 , . . . , bd ]. The homogeneous coordinate ring of the Schubert
subvariety Ω(γ̃) of Gr(d, V ) (with respect to the Plücker embedding), called a Schubert
cycle, is
G(X; γ) = G(X)/J(X; γ), (2.3)
where J(X; γ) = {δ ∈ Γ(X) : δ  γ}G(X) is the ideal generated by the poset ideal
{δ ∈ Γ(X) : δ  γ} in G(X).
By Lemma 2.3, we see that G(X; γ) is a graded ASL on Γ(X) \ {δ ∈ Γ(X) : δ  γ}.
Note that
Γ(X) \ {δ ∈ Γ(X) : δ  γ} = {δ ∈ Γ(X) : δ ≥ γ}. (2.4)
We denote this set by Γ(X; γ). The Schubert cycle G(X; γ) is a k-algebra generated by
Γ(X; γ). Both G(X) and G(X; γ) are ASLs on the posets Γ(X) and Γ(X; γ) respectively
over k.
Throughout this article, we will denote G(X; γ) by Gγ . Over algebraically closed
fields of positive characteristic, it has been proven in [MR85, Theorem 2] that Schubert
varieties are F -split, hence are F -pure, so F -pure thresholds are defined for Gγ .
Given γ = [a1 , . . . , ad ] ∈ Γ(X), one may express
γ = [β0 , . . . , βs ], where βi = aki +1 , aki +1 + 1, . . . , aki+1 − 1, aki+1
for 0 ≤ i ≤ s, as a sequence of blocks of consecutive integers, where k0 = 0. We set
t = s if ad < n, and t = s − 1 if ad = n. After each block βi is a gap
χi = aki+1 + 1, . . . , aki+1 +1 − 1,
 THE F -PURE THRESHOLD OF A SCHUBERT CYCLE 5

which is a block of consecutive integers not appearing in γ, where ad+1 = n + 1. When

ad = n, the last gap χs is empty. We remark that the sizes of the blocks and gaps of γ
can be used to characterize the Gorenstein property of Gγ , that is, Gγ is Gorenstein if
and only if |χi−1 | = |βi | for all i = 1, . . . , t [BV88, Corollary 8.13].
If Π is an arbitrary poset, and x, y ∈ Π, then y is a cover of x, denoted x ⋖ y, if x < y
and there is no z ∈ Π such that x < z < y. An upper neighbor ζ of γ is an element of
Γ(X) which covers γ. Upper neighbors of γ are
ζi = [β0 , . . . , β̃i , . . . , βs ], β̃i = aki +1 , aki +1 + 1, . . . , aki+1 − 1, aki+1 + 1
for 0 ≤ i ≤ t. Note that β̃i is obtained by increasing the last entry of βi by 1.
Set the following
J(X; ζi ) = {δ ∈ Γ(X) : δ  ζi }G(X),
Ji = J(X; ζi )/J(X; γ),
Γ(X; ζi ) = {δ ∈ Γ(X) : δ ≥ ζi },
Ωi = Γ(X; γ) \ Γ(X; ζi )
for 0 ≤ i ≤ t. Then Ji = Ωi Gγ and are distinct height 1 prime ideals for 0 ≤ i ≤ t.
Further, it holds that
γGγ = J1 ∩ · · · ∩ Jt . (2.5)
See [BV88, Propositions 5.2, 5.10]. Further, by [BV88, Corollary 9.18] we see that
(m)
Ji = Jim for any positive integer m and 0 ≤ i ≤ t. Moreover, by [BV88, Lemma 9.1],
we see the following.
Lemma 2.8. Ωi is straightening closed for 0 ≤ i ≤ t.
2.3. Summary for computing F -pure threshold. We outline the computation of
the F -pure threshold of Gγ in the list below.
1. Set the following,
i
X t
X
κi = |βj | + |χj | for 0 ≤ i ≤ t, (2.6)
j=0 j=i

κ = max{κi : 0 ≤ i ≤ t}, (2.7)

′
κ = min{κi : 0 ≤ i ≤ t}, (2.8)
t
(κi )
\
I= Ji . (2.9)
i=0

It follows from [BV88] that I is isomorphic to the canonical module of Gγ if one

ignores grading.
2. Compute a(Gγ ) by [BH92] and show −a(Gγ ) = κ. As a consequence, one sees
that I is isomorphic to the canonical module of Gγ as graded modules.
((κ−κi )(pe −1))
3. For e > 0, it holds that γ κ(p −1) ω (1−p ) = ti=0 Ji
e e T
. In particular,
κ(pe −1) (1−pe ) e
γ ω is generated in degree (κ − κ )(p − 1), and hence ω (1−p ) is gen-
′ e

erated in degree κ′ (1 − pe ).
 THE F -PURE THRESHOLD OF A SCHUBERT CYCLE 6

4. It will follow that fpt(m) = κ′ .

In the remainder of the article, we go through the details of each point.

3. The a-invariant of Schubert Cycles

In this section we compute the a-invariant of Schubert cycles Gγ by using a formula of
Bruns and Herzog in [BH92] that computes the a-invariant of an ASL. The a-invariant
of a positively graded Cohen-Macaulay ring R over a field k with canonical module ωR
is defined to be
a(R) = − min{j : (ωR )j 6= 0}.
See [GW78].
The principal chain of a distributive lattice Π is a chain P(Π) = ξ1 , . . . , ξm formed
inductively as follows: ξ1 is the unique minimal element of Π. If ξ1 , ξ2 , . . . , ξi is defined
and ξi is not the unique maximal element of Π,
ξi+1 = ν1 ⊔ · · · ⊔ νv ,
where ν1 , . . . , νv are elements which cover ξi , and ⊔ denotes the join in Π. If ξi is the
maximal element of Π, then the inductive construction terminates and ξ1 , . . . , ξi is the
principal chain. By using the principal chain, the a-invariant of an ASL is given as
follows.
Theorem 3.1. [BH92, Theorem 1.1] Let R be a graded ASL on a distributive lattice Π
with principal chain P(Π) = ξ1 , . . . , ξm . Then its a-invariant is given by
m
X
a(R) = − deg ξi . (3.1)
i=1

Since we assume Gγ has the standard grading, each deg ξi = 1, so (3.1) becomes
a(Gγ ) = −m = − P(Γ(X; γ)) ,
where P(Γ(X; γ)) is the cardinality of the principal chain of Γ(X; γ). Thus, computing
the a-invariant becomes a combinatorial problem in which we count the number of
elements making up the chain P(Γ(X; γ)).

Construction of the Principal Chain. Let γ = [a1 , . . . , ad ] ∈ Γ(X), and γ =

[β0 , . . . , βs ] be its expression as a sequence of consecutive blocks in Γ(X; γ). The join
ζ ⊔ η of two elements ζ, η ∈ Γ(X) is by taking the maximum of their corresponding
entries. We describe the algorithm that constructs the principal chain of P(Γ(X; γ)) =
ξ1 , . . . , ξm .
• The only minimal element of Γ(X; γ) is γ = [a1 , . . . , ad ], hence ξ1 = γ will be
the first element of P(Γ(X; γ)).
• Suppose that ξ1 , ξ2 , . . . , ξi is defined and ξi 6= [n − d + 1, . . . , n]. Let ξi =
(i) (i) (i)
[ν0 , ν1 , . . . , νsi ] be the block decomposition of ξi and set ti = si if the last
(i)
entry of νsi is not n and ti = si − 1 otherwise. Increase the last entry of each
(i)
νj by 1 for 0 ≤ j ≤ ti .
 THE F -PURE THRESHOLD OF A SCHUBERT CYCLE 7

• The process stops after a finite number of steps when one reaches

[n − d + 1, . . . , n − 1, n],

which is the only maximal element of Γ(X; γ). Thus, ξm = [n−d+1, . . . , n−1, n]
will always be the last element of P(Γ(X; γ)).

Example 3.2. Suppose X is a 6 × 11 generic matrix (so d = 6, n = 11), and let

γ = [2, 3, 4, 6, 8, 9] ∈ Γ(X). The principal chain P(Γ(X; γ)) will be:

γ = ξ1 = [2, 3, 4, 6, 8, 9],
ξ2 = [2, 3, 5, 7, 8, 10],
ξ3 = [2, 4, 6, 7, 9, 11],
ξ4 = [3, 5, 6, 8, 10, 11],
ξ5 = [4, 5, 7, 9, 10, 11],
ξ6 = [4, 6, 8, 9, 10, 11],
ξ7 = [5, 7, 8, 9, 10, 11],
ξ8 = [6, 7, 8, 9, 10, 11]

(boldfaced numbers are the previous entries that increased by 1).

We define a twist operation on Γ(X). For a = [a1 , . . . , ad ] ∈ Γ(X), set bi = n −

ad−i+1 + 1 for 1 ≤ i ≤ d, and define ã = [b1 , . . . , bd ]. It is clear that ã ∈ Γ(X) and
˜ = a. Further, if c = [n − d + 1, . . . , n − 1, n], then c̃ = [1, 2, . . . , d]. By mapping
ã
this operation, the construction process of the principal chain becomes the following
algorithm.

Algorithm 2. Let [a1 , . . . , ad ] be an arbitrary tuple of positive integers where 1 ≤ a1 <

· · · < ad . We define the algorithm for building a chain of d-tuples from [a1 , a2 , . . . , ad ]
to [1, 2, . . . , d] as follows:
(1) (1) (1) (1) (i) (i)
Set [a1 , . . . , ad ] = [a1 , . . . , ad ]. If [a1 , . . . , ad ], . . . , [a1 , . . . , ad ] are defined and
(i) (i) (i+1) (i+1)
[a1 , . . . , ad ] 6= [1, . . . , d], then define [a1 , . . . , ad ] by
(
(i) (i) (i) (i)
(i+1) aj − 1, if j = 1 and a1 > 1, or j > 1 and aj > aj−1 + 1
aj = (i)
aj , otherwise.

If one starts this algorithm from γ̃, then one gets the twisted principal chain ξ˜1 , ξ˜2 , . . . , ξ̃m.
In particular, the length of the principal chain and that of the chain obtained by Algo-
rithm 2 are the same.
 THE F -PURE THRESHOLD OF A SCHUBERT CYCLE 8

Example 3.3. Let γ̃ = [3, 4, 6, 8, 9, 10] be the twist of γ from Example 3.2. The
construction of the chain of tuples by Algorithm 2 is:
γ̃ = ξ˜1 = [3, 4, 6, 8, 9, 10],
ξ˜2 = [2, 4, 5, 7, 9, 10],
ξ˜3 = [1, 3, 5, 6, 8, 10],
ξ˜4 = [1, 2, 4, 6, 7, 9],
ξ˜5 = [1, 2, 3, 5, 7, 8],
ξ˜6 = [1, 2, 3, 4, 6, 8],
ξ˜7 = [1, 2, 3, 4, 5, 7],
ξ˜8 = [1, 2, 3, 4, 5, 6]
(boldfaced numbers are the previous entries that decreased by 1).
For a = [a1 , . . . , ad ] ∈ Γ(X) with a 6= [1, 2, . . . , d], we define m(a, j) = aj − 2j for j
with aj > j, and m(a) = max{m(a, j) : aj > j}. Further, we denote the length of the
sequence of elements in Γ(X) obtained by Algorithm 2 starting from a by τ (a). The
following fact is easily seen.
Lemma 3.4. If j < aj = aj+1 − 1, then m(a, j + 1) = m(a, j) − 1. In particular, if one
decomposes a into blocks as a = [α0 , α1 , . . . , αu ] where α0 is a special block defined by
{ar : ar = r}, which may be empty, then m(a) is obtained by one of the first entries of
αi , 1 ≤ i ≤ u, and for the other entries of a with aj > j, m(a, j) < m(a).
Next, we show the following.
Theorem 3.5. For a = [a1 , . . . , ad ] ∈ Γ(X) with a 6= [1, 2, . . . , d], it holds that
τ (a) = m(a) + d + 1.
Proof. We prove by induction on τ (a). Since a 6= [1, 2, . . . , d], there are at least
2 elements a and [1, 2, . . . , d] in the sequence obtained by Algorithm 2 starting from
a. Therefore, we begin to consider the case where τ (a) = 2. If τ (a) = 2, then a =
[1, 2, . . . , d − 1, d + 1]. Thus m(a) = 1 − d, and m(a) + d + 1 = 2 = τ (a).
Next, suppose that τ (a) > 2 and let b = [b1 , . . . , bd ] be the element obtained from
Algorithm 2 from a. Then, b 6= [1, 2, . . . , d], τ (b) = τ (a) − 1 and τ (b) = m(b) + d + 1 by
induction hypothesis. Let a = [α0 , α1 , . . . , αu ] be the block decomposition of Lemma
3.4, αi = ari , . . . , asi for 1 ≤ i ≤ u. Then by Lemma 3.4,
m(a) = max{m(a, ri ) : 1 ≤ i ≤ u},
and m(a, j) < m(a) for j with r1 ≤ j ≤ d and j 6= ri for 1 ≤ i ≤ u. We consider by
separating into three cases.

Case 1: ar1 > r1 + 1.

In this case it holds that br1 > r1 . Further, m(b, j) = m(a, j) < m(a) for j with j 6= ri
for any i by the first part of Lemma 3.4. On the other hand, m(b, ri ) = m(a, ri ) − 1 ≤
 THE F -PURE THRESHOLD OF A SCHUBERT CYCLE 9

m(a) − 1 for any i with 1 ≤ i ≤ u. Thus, m(b) ≤ m(a) − 1. On the other hand,
m(a) = max{m(a, ri ) : 1 ≤ i ≤ u}
= max{m(b, ri ) : 1 ≤ i ≤ u} + 1
≤ m(b) + 1.
Therefore, we see that m(a) = m(b) + 1 and
m(a) + d + 1 = (m(b) + d + 1) + 1
= τ (b) + 1
= τ (a).

Case 2: ar1 = r1 + 1 and ar1 +1 = r1 + 2.

In this case, br1 = r1 , br1 +1 = r1 + 2. Therefore, by the same argument as in Case 1,
we see that


m(b) = max {m(b, ri ) : 2 ≤ i ≤ u} ∪ {m(b, r1 + 1)} = m(a) − 1,
since m(b, r1 + 1) = m(a, r1 ) − 1. The remaining arguments are the same as that of
Case 1.

Case 3: ar1 = r1 + 1 and ar1 +1 > r1 + 2.

In this case r2 = r1 + 1 and m(a, r2 ) = m(a, r1 + 1) ≥ m(a, r1 ). Therefore, m(a) =
max{m(a, ri ) : 2 ≤ i ≤ u}. Further, by the same argument as in Case 1, we see that
m(b) = max{m(b, ri ) : 2 ≤ i ≤ u}. The remaining arguments are the same as that of
Case 1. 
By Theorem 3.5, Lemma 3.4, and Theorem 3.1, we see the following.
Corollary 3.6. Let γ = [a1 , . . . , ad ] ∈ Γ(X) and γ = [β0 , . . . , βs ] be its decomposition
into blocks of consecutive integers. Let t = s if ad < n, and t = s − 1 if ad = n. The
a-invariant of the associated Schubert cycle Gγ satisfies
(
1, if γ = [n − d + 1, . . . , n − 1, n],
−a(Gγ ) =
κ, otherwise.

Proof. The case where γ = [n − d + 1, . . . , n − 1, n] is a direct consequence of Theorem

3.1. (In this case, Gγ is isomorphic to the polynomial ring with one variable over k.)
We assume that γ 6= [n−d+1, . . . , n−1, n]. By Theorem 3.5, Lemma 3.4, and Theorem
3.1, we see that
−a(Gγ ) = d + 1 + max{n − aki+1 + 1 − 2(d − ki+1 + 1) : 0 ≤ i ≤ t}
= max{ki+1 + ((n − aki+1 ) − (d − ki+1 )) : 0 ≤ i ≤ t}.
By the definition of block decomposition [β0 , . . . , βs ] of γ, we see that
i
X
ki+1 = |βi | and
j=0
 THE F -PURE THRESHOLD OF A SCHUBERT CYCLE 10

s t
! s t
X X X X
(n − aki+1 ) − (d − ki+1 ) = |βj | + |χj | − |βj | = |χj |
j=i+1 j=i j=i+1 j=i
for any i. Therefore, it follows that
( i t
)
X X
−a(Gγ ) = max |βj | + |χj | : 0 ≤ i ≤ t = κ.
j=0 j=i


Since κi−1 − κi = |χi−1 | − |βi | for 1 ≤ i ≤ t, we see by [BV88, Theorem 8.12]
(κ ) (κ )
that J0 0 ∩ · · · ∩ Jt t is isomorphic to the canonical module of Gγ if we ignore the
(κ )
grading. Further by [BV88, Corollary 9.18], we see that Ji i = Jiκi for 0 ≤ i ≤ t.
Moreover, by Proposition 2.6, the intersection J0κ0 ∩ · · · ∩ Jtκt is generated in degree
max{κi : 0 ≤ i ≤ t} = κ. Therefore, we see the following.
Proposition 3.7. The graded canonical module ω of Gγ is
(κ0 ) (κt )
J0κ0 ∩ · · · ∩ Jtκt = J0 ∩ · · · ∩ Jt .

4. The F -pure Threshold

In this section we consider the degree of the symbolic powers of anticanonical modules
and F -pure threshold of Gγ .
(κ ) (κ )
Let m be a positive integer. Then since ω = J0 0 ∩ · · · ∩ Jt t , we see that ω (−m) =
(−mκ0 ) (−mκt )
J0 ∩ · · · ∩ Jt . On the other hand, since γGγ = J0 ∩ · · · ∩ Jt by (2.5), we see
(m(κ−κ0 )) (m(κ−κt ))
γ mκ ω (−m) = J0 ∩ · · · ∩ Jt .
Since each m(κ − κi ) is a non-negative integer, it follows from Proposition 2.6 and
[BV88, Corollary 9.18] that γ mκ ω (−m) is generated in degree
max{m(κ − κi ) : 0 ≤ i ≤ t} = m(κ − κ′ ).
Since γ is an element of Gγ of degree 1, we have the following.
Proposition 4.1. Let ω be the graded canonical module of Gγ and m a positive integer.
Then ω (−m) is generated in degree −mκ′ .
By Theorem 2.1, we have the following.
Theorem 4.2. Suppose that the underlying field of Gγ is algebraically closed of char-
acteristic p > 0. Let m be the irrelevant maximal ideal of Gγ . Then it holds that
fpt(m) = κ′ .
e
Proof. Let e be a positive integer. By Proposition 4.1, we see that ω (1−p ) is generated
in degree κ′ (1 − pe ). Thus, by Theorem 2.1 we see that νe (m) = κ′ (pe − 1). Therefore,
νe (m)
fpt(m) = lim = κ′ .
e→∞ pe

 THE F -PURE THRESHOLD OF A SCHUBERT CYCLE 11

Remark 4.3. Let S be a polynomial ring where Gγ ⊂ S. With respect to the diagonal
term order on S, the initial algebra in(Gγ ) is the Hibi ring Rk (Γ(X; γ)) on the poset
Γ(X; γ) (see [Miy07]). We observe that Gγ and Rk (Γ(X; γ)) both share the same
a-invariant and F -pure threshold.
By the structure theorem of Birkhoff, Γ(X; γ) = J(P ) is the lattice of poset ideals
of P , where P is the poset of join irreducibles of Γ(X; γ). Let −∞ and ∞ be two new
elements outside of P such that −∞ < x < ∞ for all x ∈ P . Define [−∞, ∞] = {x ∈
P : −∞ ≤ x ≤ ∞}. For an arbitrary poset Π, if x, y ∈ Π such that x ≤ y, a saturated
chain from x to y is a sequence of elements z0 , z1 , . . . , zt of Π such that
x = z0 ⋖ z1 ⋖ · · · ⋖ zt = y
(recall that zi ⋖ zi+1 means zi+1 is a cover of zi ). The length of this chain is t. Next,
define
dist(−∞, ∞) = min{t : there is a saturated chain in P from −∞ to ∞ of length t}.
Recalling the fact that in(Gγ ) is a level and anticanonical level Hibi ring on the poset
(n)
Γ(X; γ), it follows from [Miy20, Theorem 3.11] that for any n ∈ Z, the module ωin(Gγ ) is
generated in degree n · rank[−∞, ∞] if n > 0, and n · dist(−∞, ∞) if n < 0. Therefore,
the a-invariant and F -pure threshold of in(Gγ ) are respectively
−a(in(Gγ )) = rank[−∞, ∞] and fpt(m) = dist(−∞, ∞).
In fact, both quantities follow from the general results of [CM15]. These are equal to
the a-invariant and F -pure threshold of Gγ as well.

References
[BH92] Winfried Bruns and Jürgen Herzog. On the computation of a-invariants. Manuscripta
Math., 77(2-3):201–213, 1992.
[BV88] Winfried Bruns and Udo Vetter. Determinantal rings, volume 45 of Monografı́as de
Matemática [Mathematical Monographs]. Instituto de Matemática Pura e Aplicada (IMPA),
Rio de Janeiro, 1988.
[CM15] Takahiro Chiba and Kazunori Matsuda. Diagonal F-thresholds and F-pure thresholds of
Hibi rings. Comm. Algebra, 43(7):2830–2851, 2015.
[CMSV18] Aldo Conca, Maral Mostafazadehfard, Anurag K. Singh, and Matteo Varbaro. Hankel
determinantal rings have rational singularities. Adv. Math., 335:111–129, 2018.
[DCEP82] Corrado De Concini, David Eisenbud, and Claudio Procesi. Hodge algebras, volume 91 of
Astérisque. Société Mathématique de France, Paris, 1982. With a French summary.
[GW78] Shiro Goto and Keiichi Watanabe. On graded rings. I. J. Math. Soc. Japan, 30(2):179–213,
1978.
[Hir09] Daisuke Hirose. Formulas of F-thresholds and F-jumping coefficients on toric rings. Kodai
Math. J., 32(2):238–255, 2009.
[Miy07] Mitsuhiro Miyazaki. A sufficient condition for a Hibi ring to be level and levelness of
Schubert cycles. Comm. Algebra, 35(9):2894–2900, 2007.
[Miy20] Mitsuhiro Miyazaki. Fiber cones, analytic spreads of the canonical and anticanonical ideals
and limit Frobenius complexity of Hibi rings. J. Math. Soc. Japan, 72(3):991–1023, 2020.
 THE F -PURE THRESHOLD OF A SCHUBERT CYCLE 12

[MOY10] Kazunori Matsuda, Masahiro Ohtani, and Ken-ichi Yoshida. Diagonal F -thresholds on
binomial hypersurfaces. Comm. Algebra, 38(8):2992–3013, 2010.
[MR85] V. B. Mehta and A. Ramanathan. Frobenius splitting and cohomology vanishing for Schu-
bert varieties. Ann. of Math. (2), 122(1):27–40, 1985.
[MTW05] Mircea Mustaţǎ, Shunsuke Takagi, and Kei-ichi Watanabe. F-thresholds and Bernstein-
Sato polynomials. In European Congress of Mathematics, pages 341–364. Eur. Math. Soc.,
Zürich, 2005.
[STV17] Anurag K. Singh, Shunsuke Takagi, and Matteo Varbaro. A Gorenstein criterion for
strongly F -regular and log terminal singularities. Int. Math. Res. Not. IMRN, (21):6484–
6522, 2017.
[TW04] Shunsuke Takagi and Kei-ichi Watanabe. On F-pure thresholds. J. Algebra, 282(1):278–297,
2004.

Department of Mathematics, Purdue University, West Lafayette, IN 47907, USA

Email address: fong11@[Link]

Osaka Central Advanced Mathematical Institute, Osaka Metropolitan University,

3-3-138 Sugimoto, Sumiyoshi-ku Osaka 558-8585, Japan
Email address: mmyzk7@[Link]