WHEN ARE MULTIDEGREES POSITIVE?

FEDERICO CASTILLO, YAIRON CID-RUIZ, BINGLIN LI, JONATHAN MONTAÑO∗ ,

AND NAIZHEN ZHANG

m
A BSTRACT. Let k be an arbitrary field, P = Pm 1
k ×k · · · ×k Pk
p
be a multiprojective space over
k, and X ⊆ P be a closed subscheme of P. We provide necessary and sufficient conditions for
the positivity of the multidegrees of X. As a consequence of our methods, we show that when
X is irreducible, the support of multidegrees forms a discrete algebraic polymatroid. In alge-
braic terms, we characterize the positivity of the mixed multiplicities of a standard multigraded
algebra over an Artinian local ring, and we apply this to the positivity of mixed multiplicities of
ideals. Furthermore, we use our results to recover several results in the literature in the context
of combinatorial algebraic geometry.

1. I NTRODUCTION
mp
Let k be an arbitrary field, P = Pm
k ×k · · · ×k Pk
1
be a multiprojective space over k, and X ⊆
P be a closed subscheme of P. The multidegrees of X are fundamental invariants that describe
algebraic and geometric properties of X. For each n = (n1 , . . . , np ) ∈ Np with n1 + · · · + np =
dim(X) one can define the multidegree of X of type n with respect to P, denoted by degnP (X), in
different ways (see Definition 2.7, Remark 2.8 and Remark 2.9). In classical geometrical terms,
when k is algebraically closed, degnP (X) equals the number of points (counting multiplicity) in
the intersection of X with the product L1 ×k · · · ×k Lp ⊂ P, where Li ⊂ Pm i
k is a general linear
subspace of dimension mi − ni for each 1 6 i 6 p.
The study of multidegrees goes back to pioneering work by van der Waerden [60]. From a
more algebraic point of view, multidegrees receive the name of mixed multiplicities (see Defini-
tion 2.7). More recent papers where the notion of multidegree (or mixed multiplicity) is studied
are, e.g., [1, 9, 11, 23, 33, 36, 41, 42, 57].
The main goal of this paper is to answer the following fundamental question considered by
Trung [57] and by Huh [25] in the case p = 2.
• For n ∈ Np with n1 + · · · + np = dim(X), when do we have that degnP (X) > 0?
Our main result says that the positivity of degnP (X) is determined by the dimensions of the
images of the natural projections from P restricted to the irreducible components of X. First,
we set a basic notation: for each J = {j1 , . . . , jk } ⊆ {1, . . . , p}, let ΠJ be the natural projection
m mj1 mjk
ΠJ : P = Pm
k ×k · · · ×k Pk
1 p
→ Pk ×k · · · ×k Pk .
The following is the main theorem of this article. Here, we give necessary and sufficient condi-
tions for the positivity of multidegrees.
Date: September 24, 2020.
2010 Mathematics Subject Classification. Primary 14C17, 13H15; Secondary 52B40, 13A30.
Key words and phrases. positivity, multidegrees, mixed multiplicities, multiprojective scheme, projections, poly-
matroids, Hilbert polynomial.
∗ The fourth author is supported by NSF Grant DMS #2001645.

1
2 FEDERICO CASTILLO, YAIRON CID-RUIZ, BINGLIN LI, JONATHAN MONTAÑO, AND NAIZHEN ZHANG

m
Theorem A (Theorem 3.12, Corollary 3.13). Let k be an arbitrary field, P = Pm k ×k · · ·×k Pk
1 p

be a multiprojective space over k, and X ⊆ P be a closed subscheme of P. Let n = (n1 , . . . , np ) ∈

Np be such that n1 + · · · + np = dim(X). Then, degnP (X) > 0 if and only if there is an irreducible
component Y ⊆ X of X that satisfies the following two conditions:
(a) dim(Y) = dim(X).
(b) For each J = {j1 , . . . , jk } ⊆ {1, . . . , p} the inequality


nj1 + · · · + njk 6 dim ΠJ (Y)
holds.
When k is the field of complex numbers Theorem A is essentially covered by the geometric
results in [34, Theorems 2.14, 2.19],1 however their methods do not extend to arbitrary fields.
Here we follow an algebraic approach that allows us to prove the result for all fields, and hence
a general version for algebras over Artinian local rings (see Theorem B). The main idea in the
proof of Theorem A is the study of the dimensions of the images of the natural projections after
cutting by a general hyperplane (see Theorem 3.7).
We note that if p = 2 and X is arithmetically Cohen-Macaulay, the conclusion of Theorem A
in the irreducible case also holds for X (see [57, Corollary 2.8]). In Example 5.2 we show that
this is not necessarily true for p > 2.
If X is irreducible, then the function r : 2{1,...,p} → Z defined by r(J) := dim ΠJ (Y) is a


submodular function, i.e., r(J1 ∩ J2 ) + r(J1 ∪ J2 ) 6 r(J1 ) + r(J2 ) for any two subsets J1 , J2 ⊆
{1, . . . , p}, as proved in Proposition 5.1 (see also Definition 2.16). By the Submodular Theorem
(see, e.g., [7, Theorem 3.11] or [44, Appendix B]) and the inequalities of Theorem A, the points
n ∈ Np for which degnP (X) > 0 are the lattice points of a generalized permutohedron. Defined
by A. Postnikov in [49] generalized permutohedra are polytopes obtained by deforming usual
permutohedra. In recent years this family of polytopes has been studied in relation to other
fields such as probability, combinatorics, and representation theory (see [44, 45, 48]).
In a more algebraic flavor, we state the translation of Theorem A to the mixed multiplicities
of a standard multigraded algebra over an Artinian local ring (see Definition 2.13).
Theorem B (Corollary 3.14). Let A be an Artinian local ring and R be a finitely generated
standard Np -graded A-algebra. For each 1 6 j 6 p, let mj ⊂ R be the ideal generated by the
elements of degree ej , where ej ∈ Np denotes the j-th elementary vector. Let N = m1 ∩· · ·∩mp ⊂
R. Let n = (n1 , . . . , np ) ∈ Np be such that n1 + · · · + np = dim (R/ (0 :R N∞ )) − p. Then,
e(n; R) > 0 if and only if there is a minimal prime ideal P ∈ Min (0 :R N∞ ) of (0 :R N∞ ) that
satisfies the following two conditions:
(a) dim (R/P) = dim (R/ (0 :R N∞ )).
(b) For each J = {j1 , . . . , jk } ⊆ {1, . . . , p} the inequality
!
R
nj1 + · · · + njk 6 dim P −k
P + j6∈J mj
holds.
1In Remark 3.15 we briefly discuss how (over the complex numbers) Theorem A can be obtained by using the
results in [34, §2.2].
 WHEN ARE MULTIDEGREES POSITIVE? 3

For a given finite set of ideals in a Noetherian local ring, such that one of them is zero-
dimensional, we can define their mixed multiplicities by considering a certain associated stan-
dard multigraded algebra (see [58] for more information). These multiplicities have a long
history of interconnecting problems from commutative algebra, algebraic geometry, and com-
binatorics, with applications to the topics of Milnor numbers, mixed volumes, and integral
dependence (see, e.g., [25, 27, 55, 58]). As a direct consequence of Theorem B we are able to
give a characterization for the positivity of mixed multiplicities of ideals (see Corollary 4.3).
In another related result, we focus on homogeneous ideals generated in one degree; this case is
of particular importance due to its relation with rational maps between projective varieties. In
this setting, we provide more explicit conditions for positivity in terms of the analytic spread of
products of these ideals (see Theorem 4.4).
Going back to the setting of Theorem A, we switch our attention to the following discrete set

MSuppP (X) = n ∈ Np | degnP (X) > 0 ,
which we call the support of X with respect to P. When X is irreducible, we show that
MSuppP (X) is a (discrete) polymatroid (see §2.3, Proposition 5.1). The latter result was in-
cluded in an earlier version of this paper when k is algebraically closed, and an alternative proof
is given by Brändén and Huh in [3, Corollary 4.7] using the theory of Lorentzian polynomials.
An advantage of our approach is that we can describe the corresponding rank submodular func-
tions of the polymatroids, a fact that we exploit in the applications of Section 6. Additionally,
our results are valid when X is just irreducible and not necessarily geometrically irreducible
over k (i.e., we do not need to assume that X ×k k is irreducible for an algebraic closure k of k);
it should be noticed that this generality is not covered by the statements in [3] and [34].
Discrete polymatroids [24] have also been studied under the name of M-convex sets [46].
Polymatroids can also be described as the integer points in a generalized permutohedron [49],
so they are closely related to submodular functions, which are well studied in optimization, see
[38] and [52, Part IV] for comprehensive surveys on submodular functions, their applications,
and their history. There are two distinguishable types of polymatroids, linear and algebraic
polymatroids, whose main properties are inherited by their representation in terms of other
algebraic structures. Theorem A allows us to define another type of polymatroids, that we call
Chow polymatroids, and which interestingly lies in between the other two. In the following
theorem we summarize our main results in this direction.
Theorem C (Theorem 5.5). Over an arbitrary field k, we have the following inclusions of
families of polymatroids
     
Linear polymatroids ⊆ Chow polymatroids ⊆ Algebraic polymatroids .
Moreover, when k is a field of characteristic zero, the three families coincide.
If k has positive characteristic, then these types of polymatroids do not agree. In fact, there
exist examples of polymatroids which are algebraic over any field of positive characteristic but
never linear (see Remark 5.7).
Theorem A can be applied to particular examples of varieties coming from combinatorial
algebraic geometry. In §6.1 we do so to matrix Schubert varieties; in this case the multidegrees
are the coefficients of Schubert polynomials, thus our results allow us to give an alternative proof
to a recent conjecture regarding the support of these polynomials (see Theorem 6.3). In §6.2 and
§6.3 we study certain embeddings of flag varieties and of the moduli space M0,p+3 , respectively
4 FEDERICO CASTILLO, YAIRON CID-RUIZ, BINGLIN LI, JONATHAN MONTAÑO, AND NAIZHEN ZHANG

(see Proposition 6.7 and Proposition 6.8). In §6.4 we recover a well-known characterization for
the positivity of mixed volumes of convex bodies (see Theorem 6.9).
We now outline the contents of the article. In Section 2 we set up the notation used throughout
the document. We also include key preliminary definitions and results, paying special attention
to the connection between mixed multiplicities of standard multigraded graded algebras and
multidegrees of their corresponding schemes. Section 3 is devoted to the proof of Theorem A
and Theorem B. Our results for mixed multiplicities of ideals are included in Section 4. In
Section 5 we relate our results to the theory of polymatroids. In particular, we show the proof
of Theorem C. We finish the paper with Section 6 where the applications to combinatorial
algebraic geometry are presented.
We conclude the Introduction with an illustrative example. The following example is con-
structed following the same ideas in Proposition 5.4.

Example 1.1. Consider the polynomial ring S = k[v1 , v2 , v3 ][w1 , w2 , w3 ] with the N3 -grading
deg(vi ) = (0, 0, 0), deg(wi ) = ei for 1 6 i 6 3. Let T be the N3 -graded polynomial ring T =
k [x0 , . . . , x3 ] [y0 , . . . , y3 ] [z0 , . . . , z3 ] where deg(xi ) = e1 , deg(yi ) = e2 and deg(zi ) = e3 . Consider
the N3 -graded k-algebra homomorphism

x0 7→ w1 , x1 7→ v1 w1 , x2 7→ v1 w1 , x3 7→ v1 w1 ,
ϕ = T → S, y0 7→ w2 , y1 →7 v1 w2 , y2 →7 v2 w2 , y3 →7 (v1 + v2 )w2 ,
z0 7→ w3 , z1 7→ v1 w3 , z2 7→ v2 w3 , z3 7→ v3 w3 .

Note that P = Ker(ϕ) ⊂ T is an N3 -graded prime ideal. Let Y ⊂ P = P3k ×k P3k ×k P3k be the
closed subscheme corresponding to P. In this case, one can easily compute the dimension of
the projections ΠJ (Y) for each J ⊆ {1, 2, 3}, and so Theorem A implies that MSuppP (Y) is given
by all n = (n1 , . . . , n3 ) ∈ N3 satisfying the following conditions:

n1 + n2 + n3 = 3 = dim(Y),




n1 + n2 6 2 = dim Π{1,2} (Y) , n1 + n3 6 3 = dim Π{1,3} (Y) , n2 + n3 6 3 = dim Π{2,3} (Y) ,




n1 6 1 = dim Π{1} (Y) , n2 6 2 = dim Π{2} (Y) , n3 6 3 = dim Π{3} (Y) .

Hence MSuppP (Y) = {(0, 0, 3), (0, 1, 2), (0, 2, 1), (1, 0, 2), (1, 1, 1)} ⊂ N3 . This set can also be
represented graphically as follows:
(3, 0, 0)

(0, 0, 3) (0, 3, 0)

Additionally, by using Macaulay2 [21] we can compute that its multidegree polynomial (see
Definition 2.10) is equal to:

degP (Y; t1 , t2 , t3 ) = t31 t32 + t31 t22 t3 + t31 t2 t23 + t21 t32 t3 + t21 t22 t23 .
 WHEN ARE MULTIDEGREES POSITIVE? 5

We note that here we are following the convention that MSuppP (Y) is given by the complemen-
tary degrees of the polynomial degP (Y; t1 , t2 , t3 ); for instance, the term t31 t32 corresponds to the
point (3, 3, 3) − (3, 3, 0) = (0, 0, 3) ∈ MSuppP (Y).

2. N OTATION AND P RELIMINARIES

In this section, we set up the notation that is used throughout the paper. We also present some
preliminary results needed in the proofs of our main theorems.
Let p > 1 be a positive integer. If n = (n1 , . . . , np ), m = (m1 , . . . , mp ) ∈ Zp are two multi-
indexes, we write n > m whenever ni > mi for all 1 6 i 6 p, and n > m whenever nj > mj for
all 1 6 j 6 p. For each 1 6 i 6 p, let ei ∈ Np be the i-th elementary vector ei = (0, . . . , 1, . . . , 0).
Let 0 ∈ Np and 1 ∈ Np be the vectors 0 = (0, . . . , 0) and 1 = (1, . . . , 1) of p copies of 0 and 1,
respectively. For any n = (n1 , . . . , np ) ∈ Zp , we define its weight as |n| := n1 + · · · + np . Let [p]
denote the set [p] := {1, . . . , p}.
For clarity of exposition we first introduce the main concepts in the theory of multidegrees
over an arbitrary field. Later, we also work over Artinian local rings; we highlight important
details in this more general setting in §2.2.
2.1. The case over a field. We begin by introducing a general setup for Theorem A and its
preparatory results.
Setup 2.1. Let k be an arbitrary field. Let R be a finitely generated standard Np -graded algebra
over k, that is, [R]0 = k and R is finitely generated over k by elements of degree ei with 1 6 i 6 p.
For each subset J = {j1 , . . . , jk } ⊆ [p] = {1, . . . , p} denote by R(J) the standard Nk -graded k-
algebra given by M
R(J) := [R](i1 ,...,ip ) ;
i1 >0,...,ip >0
ij =0 if j6∈J
L
for instance, for each 1 6 j 6 p, R(j) denotes the standard N-graded k-algebra R(j) := k>0 [R]k·ej .


For each 1 6 j 6 p, let mj ⊂ R be the ideal mj := [R]ej . Let N ⊂ R be the multigraded irrele-
vant ideal N := m1 ∩ · · · ∩ mp . For each J ⊆ [p], let NJ ⊂ R(J) be the corresponding multigraded
T 
irrelevant ideal NJ := j∈J mj ∩R(J) . Let X be the multiprojective scheme X := MultiProj(R)
(see Definition 2.2 below) and XJ be the multiprojective scheme XJ := MultiProj(R(J) ) for each
J ⊆ [p]. To avoid trivial situations, we always assume that X 6= ∅.

Definition 2.2. The multiprojective scheme MultiProj(R) is given by MultiProj(R) := P ∈
Spec(R) | P is Np -graded and P 6⊇ N , and its scheme structure is obtained by using multi-
homogeneous localizations (see, e.g., [28, §1]).
The inclusion R(J) ,→ R induces the natural projection
ΠJ : X → XJ , P ∈ X 7→ P ∩ R(J) ∈ XJ .
m
We embed X as a closed subscheme of a multiprojective space P := Pm p
k ×k · · · ×k Pk . Then,
1

for each J = {j1 , . . . , jk } ⊆ [p], ΠJ : X → XJ corresponds with the restriction to X and to XJ of

the natural projection
mj mj
ΠJ : P → Pk 1 ×k · · · ×k Pk k ,
mj1 mjk
and XJ becomes a closed subscheme of Pk ×k · · · ×k Pk .
6 FEDERICO CASTILLO, YAIRON CID-RUIZ, BINGLIN LI, JONATHAN MONTAÑO, AND NAIZHEN ZHANG

For any multi-homogeneous element x ∈ R, the closed subscheme MultiProj(R/xR) ⊆ X is

denoted by X ∩ V(x).
Notation 2.3. From now on, J = {j1 , . . . , jk } denotes a subset of [p]. Set r := dim(X) and r(J) :=
dim (ΠJ (X)) for each J ⊆ [p]. For a singleton set {i} ⊆ [p], r({i}) and Π{i} are simply denoted
by r(i) and Πi , respectively.
Note that the image of ΠJ : X → XJ can be described by the following isomorphism
!
R(J)
(1) ΠJ (X) =∼ MultiProj .
R(J) ∩ (0 :R N∞ )
P
∼ P

R
Remark 2.4. Since R = R(J) ⊕ j6∈J j , we obtain a natural isomorphism R(J) =
m of
j6∈J mj
Nk -graded k-algebras where k = |J|.
We now provide some preparatory results.
Lemma 2.5. Under Setup 2.1, the following statements hold:
(i) r = dim(X) = dim (R/ (0 :R N∞ )) − p.
(ii) There is an isomorphism
 
X 
∼ MultiProj R/ (0 :R N∞ ) +

(2) ΠJ (X) = mj  .
j6∈J
∼ MultiProj(R ) = XJ .
(iii) If (0 :R N∞ ) = 0, then ΠJ (X) =
  (J)
(iv) If (0 :R N∞ ) = 0, then 0 :R(J) N∞J = 0.

Proof. (i) This formula follows from [28, Lemma 1.2] (also, see [10, Corollary 3.5]).
(ii) From the natural maps R(J) ,→ R  R/ (0 :R N∞ ), we obtain a natural isomorphism

=∼
R(J) / R(J) ∩ (0 :R N∞ ) −→ R/ (0 :R N∞ ) (J) .

∼ R/ (0 :R N∞ ) + P mj . There-
 
By using Remark 2.4 it follows that R/ (0 :R N∞ ) (J) =


j6∈J
fore, the claimed isomorphism is obtained from (1).
(iii) It follows directly from part (ii) and Remark 2.4.
(iv) This part is clear. 

Let PR (t) = PR (t1 , . . . , tp ) ∈ Q[t] = Q[t1 , . . . , tp ] be the Hilbert polynomial of R (see, e.g.,
[23, Theorem 4.1], [9, Theorem 3.4]). Then, the degree of PR is equal to r and
PR (ν) = dimk ([R]ν )
for all ν ∈ Np such that ν  0. Furthermore, if we write
X 
t1 + n1
 
tp + np

(3) PR (t) = e(n1 , . . . , np ) ··· ,
n1 np
n1 ,...,np >0

then 0 6 e(n1 , . . . , nr ) ∈ Z for all n1 + · · · + np = r.

Remark 2.6. The following are basic properties of Hilbert polynomials.
 WHEN ARE MULTIDEGREES POSITIVE? 7

(i) Since Nk (0 :R N∞ ) = 0 for k  0 we have dimk ([R]ν ) = dimk ([R/(0 :R N∞ )]ν ) for ν  0.
Thus,
PR (t) = PR/(0:R N∞ ) (t).
(ii) Let L be a field extension of k. Then, R ⊗k L is a finitely generated standard Np -graded
L-algebra and dimL ([R ⊗k L]ν ) = dimk ([R]ν ) for all ν ∈ Np . Thus,

PR⊗k L (t) = PR (t).

In particular, one can always assume k is an infinite field (for instance, we can substitute
k by a purely transcendental field extension k(ξ)).

Under the notation of (3) we define the following invariants.

Definition 2.7. Let n = (n1 , . . . , np ) ∈ Np with |n| = r. Then:

(i) e(n, R) := e(n1 , . . . , np ) is the mixed multiplicity of R of type n.
(ii) degnP (X) := e(n1 , . . . , np ) is the multidegree of X = MultiProj(R) of type n with respect to
P.

As stated in the Introduction, in classical geometrical terms, when k is algebraically closed,

degnP (X) is also equal to the number of points (counting multiplicity) in the intersection of X
with the product L1 ×k · · · ×k Lp ⊂ P, where Li ⊆ Pm i
k is a general linear subspace of dimension
mi − ni for each 1 6 i 6 p (see [60], [9, Theorem 4.7]).
The multidegrees of X can be defined easily in terms of Chow rings and in terms of Hilbert
series.
m
Remark 2.8. The Chow ring of P = Pm p
k ×k · · · ×k Pk is given by
1

Z[H1 , . . . , Hp ]
A∗ (P) =  mp +1

Hm
1
1 +1
, . . . , Hp

where Hi represents the class of the inverse image of a hyperplane of Pm i

k under the natural
projection Πi : P → Pm i
k . Then, the class of the cycle associated to X coincides with
X m −n
[X] = degnP (X) Hm1
1 −n1
· · · Hp p p ∈ A∗ (P).
06ni 6mi
|n|=r
P ν
Remark 2.9. By considering the Hilbert series HilbR (t1 , . . . , tp ) := ν∈Np dimk ([R]ν ) tν1 1 · · · tpp
of R, one can analogously define the notions of mixed multiplicities and multidegrees (see
[43, §8.5], [9, Theorem A]). Here we quickly derive this analogous definition because we shall
use it in §6.1. Let S = k[x1,0 , x1,1 , . . . , x1,m1 ] ⊗k · · · ⊗k k[xp,0 , xp,1 , . . . , xp,mp ] be the multigraded
mp
polynomial ring corresponding with P = Pm k ×k · · · ×k Pk , that is P = MultiProj(S). By
1

considering an S-free resolution of R, we can write

Y
p
HilbR (t) = K(R; t)/(1 − t) m+1
= K(R; t1 , . . . , tp )/ (1 − ti )mi +1 ,
i=1
8 FEDERICO CASTILLO, YAIRON CID-RUIZ, BINGLIN LI, JONATHAN MONTAÑO, AND NAIZHEN ZHANG

where K(R; t) is called the K-polynomial of R (see [43, Definition 8.21]). Let C(R; t) ∈ Z[t1 , . . . , tp ]
be the sum of all the terms in K(R; 1 − t) of total degree equal to dim(S) − dim(R) (see [43, Def-
inition 8.45]). Then, if (0 :R N∞ ) = 0, we obtain the equality
X m −n
C(R; t) = degnP (X) tm
1
1 −n1
· · · tp p p .
06ni 6mi
|n|=r
P
Proof. From [9, Theorem A(I)] we have HilbR (t) = |k|=dim(R) Qk (t)/(1 − t)k where Qk (t) ∈
Z[t]. The assumption (0 :R N∞ ) = 0 gives that dim(R) = r + p (see Lemma 2.5(i)). Hence, by
using [9, Theorem A(II,III)] we obtain that degnP (X) = e(n; R) = Qn+1 (1) for all |n| = r. Also,
the assumption (0 :R N∞ ) = 0 and [9, Theorem 2.8(ii)] imply that Qk (1) = 0 when |k| = dim(R)
and ki = 0 for some 1 6 i 6P p.
After writing HilbR (t) = |k|=dim(R) Qk (t)(1 − t)m+1−k /(1 − t)m+1 , we obtain the equality
X
K(R; t) = Qk (t)(1 − t)m+1−k .
|k|=dim(R)

Making the substitution ti 7→ (1 − ti ) and

P P choosing the terms of total − dim(R) =
Pdegree dim(S)m−n
p
i=1 mi −r, it follows that C(R; t) =
m+1−k
|k|=dim(R) Qk (1)t = |n|=r Qn+1 (1)t . So, the
result is clear. 

Although in the proofs of Theorem A and Theorem B we do not exploit the fact that mul-
tidegrees can be defined as in Remark 2.8, we do encode the multidegrees in a homogeneous
polynomial that mimics the cycle associated to X in the Chow ring A∗ (P). The following ob-
jects are the main focus of this paper.
m
Definition 2.10. Let X ⊆ P = Pm p
k ×k · · · ×k Pk be a closed subscheme with r = dim(X). We
1

denote the multidegree polynomial of X with respect to P as the homogeneous polynomial

X m −n
degP (X; t1 , . . . , tp ) := degnP (X) tm
1
1 −n1
· · · tp p p ∈ N[t1 , . . . , tp ]
06ni 6mi
|n|=r

of degree m1 + · · · + mp − r. We say that the support of X with respect to P is given by


MSuppP (X) := n ∈ Np | degnP (X) > 0 .
Remark 2.11. Note that under the assumption (0 :R N∞ ) = 0 we obtain the equality degP (X; t) =
C(R; t).

2.2. The case over an Artinian local ring. In this subsection, we show how the mixed multi-
plicities are defined for a standard multigraded algebra over an Artinian local ring.
Setup 2.12. Keep the notations and assumptions introduced in Setup 2.1 and now substitute the
field k by an Artinian local ring A.
In this setting, the notion of mixed multiplicities is defined essentially in the same way as in
Definition 2.7.
 WHEN ARE MULTIDEGREES POSITIVE? 9

Definition 2.13. Let PR (t) = PR (t1 , . . . , tp ) ∈ Q[t] = Q[t1 , . . . , tp ] be the Hilbert polynomial of
R (see, e.g., [23, Theorem 4.1], [9, Theorem 3.4]). Then, as before, the degree of PR is equal to
dim (R/ (0 :R N∞ )) − p and
PR (ν) = lengthA ([R]ν )
P
for all ν ∈ N such that ν  0. If we write PR (t) = n1 ,...,np >0 e(n1 , . . . , np ) t1n+n
p tp +np



1
1
· · · np ,
∞
then 0 6 e(n1 , . . . , nr ) ∈ Z for all n1 + · · · + np = dim (R/ (0 :R N )) − p. For each n =
(n1 , . . . , np ) ∈ Np with |n| = dim (R/ (0 :R N∞ )) − p, we set that e(n, R) := e(n1 , . . . , np ) is
the mixed multiplicity of R of type n.
2.3. Polymatroids. In this subsection we include some relevant information about polyma-
troids.
Definition 2.14. Let E be a finite set and r a function r : 2E → Z>0 satisfying the following two
properties: (i) it is non-decreasing, i.e., r(T1 ) 6 r(T2 ) if T1 ⊆ T2 ⊆ E, and (ii) it is submodular,
i.e., r(T1 ∩ T2 ) + r(T1 ∪ T2 ) 6 r(T1 ) + r(T2 ) if T1 , T2 ⊆ E. The function r is called a rank
function on E. We usually let E = [p].
A (discrete) polymatroid P on [p] with rank function r is a collection of points in Np of the
following form
 
 X X 
P = x = (x1 , . . . , xp ) ∈ Np | xj 6 r(J), ∀J ( [p], xi = r([p]) .
 
j∈J i∈[p]
By definition, a polymatroid consists of the integer points of a polytope (the convex hull of P),
we call that polytope a base polymatroid polytope. We note that a polymatroid is completely
determined by its rank function.
Remark 2.15. If the rank function of P satisfies r({i}) 6 1 for every i ∈ [p], then P is called
a matroid. In other words, matroids are discrete polymatroids where every integer point is an
element of {0, 1}p . A general reference for matroids is [47].
In the following definition we consider the standard notions of linear and algebraic matroids
(see [47, Chapter 6]) and adapt them to the polymatroid case.
Definition 2.16. Let P be a polymatroid.
• We say P is linear over a field k if there exists a k-vectorspace V and subspaces
P
Vi , i ∈ [p] such that for every J ⊆ [p] we have r(J) = dimk j∈J Vj [47, Proposi-
tion 1.1.1]. The vector space V together with the subspaces Vi for 1 6 i 6 p, are a
linear representation of P.
• We say P is algebraic over a field k if there exists a field extension k ,→ L and intermedi-
V 
ate field extensions Li , i ∈ [p] such that for every J ⊆ [p] we have r(J) = trdegk j∈J Lj ,
V
where j∈J Lj is the compositum of the subfields, i.e., the smallest subfield in L con-
taining all of them [47, Theorem 6.7.1]. The field L together with the subfields Li for
1 6 i 6 p, are an algebraic representation of P.
3. A CHARACTERIZATION FOR THE POSITIVITY OF MULTIDEGREES
In this section, we focus on characterizing the positivity of multidegrees and our main goal
is to prove Theorem A and Theorem B. Throughout this section we continue using the same
notations and assumptions of Section 2.
10 FEDERICO CASTILLO, YAIRON CID-RUIZ, BINGLIN LI, JONATHAN MONTAÑO, AND NAIZHEN ZHANG

We begin with the following result that relates the Hilbert polynomial PR (t) ∈ Q[t] of R with
the dimensions r(J) = dim (ΠJ (X)) of the schemes ΠJ (X). It extends [57, Theorem 1.7] to a
multigraded setting.
Proposition 3.1. Assume Setup 2.1. For each J = {j1 , . . . , jk } ⊆ [p], let deg(PR ; J) be the degree
of the Hilbert polynomial PR in the variables tj1 , . . . , tjk . Then, for every such J = {j1 , . . . , jk } we
have that
deg(PR ; J) = r(J).
Proof. We may assume that (0 :R N∞ ) = 0 and k is an infinite field by Remark 2.6. Fix J =
{j1 , . . . , jk } ⊆ [p] and let w ∈ N be such that dimk ([R]n ) = PR (n) for every n = (n1 , . . . , np ) > w1.
Let (d1 , . . . .dk ) be such that δ := deg(PR ; J) = d1 + · · · + dk and tdj11 · · · tdjkk divides a term of PR .
Let q be a polynomial in the variables {ti | i 6∈ J} such that PR − q · tdj11 · · · tdjkk has no term
divisible by tdj11 · · · tdjkk . Let s = (si | i 6∈ J) ∈ Np−|J| be a vector of integers such that s > w1 and
q(s) 6= 0. Thus, if one evaluates ti = si in PR for every i 6∈ J one obtains a polynomial Q on the
variables tj1 , · · · , tjk of degree δ. On the other hand, by [9, Theorem 3.4], for nj1 , · · · , njk > w
this polynomial Q coincides with the Hilbert polynomial of the R(J) -module generated by [R]s 0 ,
where si0 = si if i 6∈ J and si0 = 0 otherwise. Call this module M.
Since (0 :R N∞ ) = 0, for every 1 6 i 6 p we have grade(mi ) > 1, and then there exist ele-
ments yi ∈ [R]ei which are non-zero-divisors (see, e.g., [6, Lemma 1.5.12]). From the fact that
s0 s0
y11 · · · ypp ∈ M, it follows that AnnR(J) (M) = 0. Therefore, δ = dim Supp(M) ∩ X(J) = r(J),

by [9, Theorem 3.4], finishing the proof. 

In the following remark we gather some basic relations for the radicals of certain ideals.
√ √
Remark
√ √ 3.2. (i)
√ Let I, J, K
√ ⊂ R be ideals. If J ⊂ K, then I + J ⊂ I + K. In particular, if
J = K, then I + J = I + K.
∞ )mk ⊂ (x) for some k > 0, it follows that
p
(ii)
q For any element x ∈ m 1 , since (x :R m1 1 (x) =
(x :R m∞ k (x :R m∞
p
1 )m1 = 1 ) ∩ m1 .

If k is an infinite field, then for each 1 6 i 6 p we say that a property P is satisfied by a

general element in the k-vector space [R]ei , if there exists a dense open subset U of [R]ei with
the Zariski topology such that every element in U satisfies the property P.
The following three technical lemmas are important steps for the proof of Theorem 3.7.
Lemma 3.3. Assume Setup 2.1 with k being an infinite field. Suppose thatq
R is a domain. Let
∞ x :R m∞
p

x ∈ [R]e1 be a general element. Then, we have the equality (x :R N ) = 1 .

Proof. Since (0 :R N∞ ) = 0, we have that ht(mj ) > 1 for every 1 6 j 6 p. Consider the following
finite set of prime ideals

S = P ∈ Spec(R) | P ∈ Min(mj ) for some 2 6 j 6 p and P 6⊇ m1 .
By using the Prime Avoidance Lemma and the fact that
k is infinite, for a general element
x ∈ [R]e1 we have that x 6∈ P∈S P. If P ∈ Min x :R m∞
S
1 , then ht(P) 6 1 by Krull’s Principal
Ideal Theorem, and so we would have that P ∈ S whenever P 6⊇ m1 and P ⊇ mj for some 2 6

j 6 p. Therefore, for any P ∈ Spec(R) and a general element x ∈ [R]e1 , if P ∈ Min x :R m∞ 1
∞ ∞ ∞ ∞
p p
we get P ⊇ (x :R N ) = (x :R (m1 ∩ m2 ∩ · · · ∩ mp ) ); so, (x :R N ) = (x :R m1 ). 
 WHEN ARE MULTIDEGREES POSITIVE? 11

The lemma below is necessary for some reduction arguments in Theorem 3.7.
Lemma 3.4. Assume Setup 2.1 with k being an infinite field. Suppose that R is a domain.
Let x ∈ [R]e1 be a general element and set Z = X ∩ V(x) = MultiProj(R/xR). Then, for each
J = {1, j2 , . . . , jk } ⊆ [p], the following statements hold:


(i) dim (ΠJ (Z)) = dim (XJ ∩ V(x)), where XJ ∩ V(x) = MultiProj R(J) /xR(J) .
(ii) dim(ΠL (Z)) = dim ΠL0 (XJ ∩ V(x)) , where L = J \ {1} and ΠL0 denotes the natural pro-


mj2 mjk mj mj
jection ΠL0 : Pm k ×k Pk ×k · · · ×k Pk
1
→ Pk 2 ×k · · · ×k Pk k .
Proof. For notational purposes, let bJ := m1 ∩ R(J) .
∼ MultiProj R/ (x :R N∞ ) + P mj . Since we are as-



(i) From (2) we have that ΠJ (Z) = j6∈J
suming 1 ∈ J, from Remark 2.4 we obtain the natural isomorphism
  ∼  X 
∞ = ∞
R(J) / x :R(J) bJ − → R/ (x :R m1 ) + mj ;
j6∈J

indeed, for ` > 0 and y ∈ R(J) , one notices that b`J · y ∈ xR(J) if and only if m`1 · y ∈ xR.
q
∞
P q P
By Lemma 3.3 and Remark 3.2(i) we have (x :R m1 ) + j6∈J mj = (x :R N∞ ) + j6∈J mj ,
q q
and by applying Lemma 3.3 to the ring R(J) we obtain (x :R(J) b∞ J ) = (x :R(J) N∞ J ). It fol-
lows that
s X
∼
q ∞ (x :R N∞ ) +

(4) R(J) (x :R(J) NJ ) = R mj ,
j6∈J

which gives the result.

∼ MultiProj R/ (x :R N∞ ) + P mj + m1 and that



(ii) By using (2) we obtain that ΠL (Z) = j6 ∈ J
∼ MultiProj R / (x :R N∞ ) + bJ . Since the isomorphism in (4) can be
  
Π 0 (XJ ∩ V(x)) =
L (J) (J) J
extended to
 
. s X
∼
. q 
R(J) (x :R(J) N∞
J ) + bJ = R
 (x :R N∞ ) + mj + m1  ,
j6∈J

the result follows from Remark 3.2(i). 

We continue with the next auxiliary lemma that allows us to simplify the proof of Theo-
rem 3.7.
Lemma 3.5. Assume Setup 2.1 with k being an infinite field. Suppose that R is a domain and
r(1) > 1. Let x ∈ [R]e1 be a general element and set Z = X ∩ V(x) = MultiProj(R/xR). Then,
the following statements hold:
(i) If 1 ∈ J ⊆ [p], then dim (ΠJ (Z)) = r(J) − 1; in particular, dim(Z) = r − 1.
(ii) If 1 6∈ J and r(K) > r(J), where K = {1} ∪ J ⊆ [p], then dim (ΠJ (Z)) = r(J).
Proof. (i) First, from Lemma 3.4(i) it suffices to compute dim (XJ ∩ V(x)), where XJ ∩ V(x) =
∼ Π 0 (XJ ), where Π 0 de-
MultiProj R(J) /xR(J) . For J = {1, j2 , . . . , jk } ⊆ [p], note that Π1 (X) =


1 1
mj mj
notes the natural projection Π10 : Pm
k ×k Pk ×k · · · ×k Pk
1 2 k
→ Pm 1
k . Therefore, neither the
assumption nor the conclusion changes if we substitute R by R(J) and X by XJ , and we do so.
12 FEDERICO CASTILLO, YAIRON CID-RUIZ, BINGLIN LI, JONATHAN MONTAÑO, AND NAIZHEN ZHANG

From the short exact sequence

x
0 → R(−e1 ) →
− R → R/xR → 0,
we obtain PR/xR (t) = PR (t) − PR (t − e1 ). By using Proposition 3.1, deg(PR ; t1 ) = r(1) > 1
and so PR is non-constant as a univariate polynomial in the variable t1 . Thus, PR/xR (t) 6= 0
which implies that (x :R N∞ ) is a proper ideal. So, Krull’s Principal Ideal Theorem yields that
ht(x :R N∞ ) = 1 and that
dim(Z) = dim(R/(x :R N∞ )) − p = dim(R) − 1 − p = (r + p) − 1 − p = r − 1.
(ii) By using Lemma 3.4(ii), we can substitute R by R(K) and X by XK , and we do so. So, we
may assume that K = [p] and J = {2, . . . , p}. From (2) we get the isomorphism
∼
 
∞
(5) ΠJ (Z) = MultiProj R/ ((x :R N ) + m1 ) .

The equality
q
(x :R N ) + m1 = (x :R m∞
∞
p
(6) 1 ) + m1

follows from Lemma 3.3 and Remark 3.2(i). The assumption yields that


ht(m1 ) = dim (R) − dim R(J) = (r + p) − (r(J) + p − 1) > 2,

then as a consequence Krull’s Principal Ideal Theorem it follows that (x) = (x :R m∞

p p
√ 1 );
∞
p
therefore, Remark 3.2(i) implies that
(x :R m1 ) + m1 = m1 . By summing up, we obtain the
equalities dim R/ ((x :R N∞ ) + m1 ) = dim(R/m1 ) = r(J) + p − 1, and so the result follows.

The next important theorem computes the dimension of the image of the projections ΠJ after
cutting with a general hyperplane under certain conditions. For the proof of this result, we
need the following version of Grothendieck’s Connectedness Theorem. For that, we recall the
definitions

c(R) := min dim(R/a) | a ⊂ R is an ideal and Spec(R) \ V(a) is disconnected ,

sdim(R) := min dim(R/P) | P ∈ Min(R) and
√
ara(a) := min{n | (a1 , . . . , an ) = a and ai ∈ R}
p

for any ideal a ⊂ R.

Lemma 3.6 ([5, Proposition 2.1], [57, Lemma 2.6]). For two proper homogeneous ideals a, b ⊂
R, if min{dim(R/a), dim(R/b)} > dim(R/(a + b)), then
dim(R/(a + b)) > min{c(R), sdim(R) − 1} − ara(a ∩ b).
We are now ready to present the following theorem.
Theorem 3.7. Assume Setup 2.1 with k being an infinite field. Suppose that R is a domain and
r(1) > 1. Let x ∈ [R]e1 be a general element and set Z = X ∩ V(x) = MultiProj(R/xR). Then,
for each J ⊆ [p] we have that

dim(ΠJ (Z)) = min r(J), r J ∪ {1} − 1 .

 WHEN ARE MULTIDEGREES POSITIVE? 13

Proof. For each J = {j1 , . . . , jk } ⊆ K = {h1 , . . . , h` } ⊆ [p] we have that ΠJ (Z) = Π0J (ΠK (Z))
mh mh mj mj
where Π0J denotes the natural projection Π0J : Pk 1 ×k · · · ×k Pk ` → Pk 1 ×k · · · ×k Pk k . So,
from Lemma 3.5(i) it follows that the inequality “6” holds in the desired equality.
Due to Lemma 3.5, in order to show the reversed inequality “>”, it suffices to show that
dim(ΠJ (Z)) > r(J) − 1 when 1 6∈ J and r(K) = r(J), where K = {1} ∪ J ⊆ [p]. By using
Lemma 3.4(ii), we assume may that K = [p] and J = {2, . . . , p}. From

(5) and (6), the proof
would be complete if we prove the inequality dim R/ (x :R m∞ 1 ) + m1 > (rJ −1)+(p−1) =
r + p − 2.
By using Lemma 3.3 and Lemma 3.5(i) we obtain that
dim (R/(x :R m∞ ∞
1 )) = dim (R/(x :R N )) = (r − 1) + p = r + p − 1,
and since r(J) = r, we have
dim(R/m1 ) = dim(R(J) ) = r(J) + (p − 1) = r + (p − 1) = r + p − 1.
Moreover, (6) and Lemma 3.5(i) yield that
dim R/ ((x :R m∞ ∞



1 ) + m1 ) = dim R/ ((x :R N ) + m1 ) 6 (r − 1) + (p − 1) = r + p − 2.

Since x ∈ m1 , Remark 3.2(ii) gives that ara (x :R m∞

1 ) ∩ m1 = ara ((x)) = 1. As R is a domain,
c(R) = sdim(R) = r + p. Therefore, from Lemma 3.6 we obtain that
dim R/ ((x :R m∞


1 ) + m1 ) > min{r + p, (r + p) − 1} − 1 = r + p − 2.
So, the proof is complete. 
Notation 3.8. Let {x0 , . . . , xs } be a basis of the k-vector space [R]e1 . Consider a purely tran-
scendental field extension L := k(z0 , . . . , zs ) of k, and set RL := R ⊗k L and XL := X ⊗k L =
mp
MultiProj (RL ) ⊆ P ⊗k L = Pm L ×L · · · ×L PL . We say that z := z0 x0 + · · · + zs xs ∈ [RL ]e1 is
1

the generic element of [RL ]e1 .

In the following remark we explain that field extensions as in Notation 3.8 preserve the do-
main assumption.
Remark 3.9. Suppose that R is a domain and consider a purely transcendental field extension
k(ξ). Then, R ⊗k k(ξ) is also a domain; indeed, one can see that R ⊗k k(ξ) is a subring of
the field of fractions Quot(R[ξ]) of the polynomial ring R[ξ]. So, when R is a domain one can
extend k to an infinite field without loosing the assumption of R being a domain.
The lemma below shows that the Hilbert function modulo a generic element coincides with
the one module a general element.
Lemma 3.10. Assume Notation 3.8 with k being an infinite field. Let x ∈ [R]e1 be a general
element, then



dimk [R/xR]ν = dimL [RL /zRL ]ν
for all ν ∈ Np .
Proof. Let T be the polynomial ring T = k[z0 , . . . , zs ] and consider the finitely generated T -
algebra given by S = (R ⊗k T ) /w (R ⊗k T ) where w = z0 x0 + · · · + zs xs ∈ R ⊗k T . From the
Grothendieck’s Generic Freeness Lemma (see, e.g., [39, Theorem 24.1], [13, Theorem 14.4])
there exists an element 0 6= a ∈ T such that Sa is a free Ta -module. Hence, for any p ∈ Spec(T )
14 FEDERICO CASTILLO, YAIRON CID-RUIZ, BINGLIN LI, JONATHAN MONTAÑO, AND NAIZHEN ZHANG

inside the dense open subset D(a) ⊂ Spec(T ), if k(p) denotes the residue field k(p) = Tp /pTp
of Tp , one has that




dimk(p) [Sa ⊗Ta k(p)]ν = dimQuot(T ) [Sa ⊗Ta Quot(T )]ν = dimL [RL /zRL ]ν
for all ν ∈ Np . Note that for any β = (β0 , . . . , βs ) ∈ ks+1 with pβ = (z0 −β0 , . . . , zs −βs ) ∈ D(a)
one has the isomorphisms
∼ R ⊗k T ∼ R/ (β0 x0 + · · · + βs xs ) R.
Sa ⊗Ta k(pβ ) = =
(z0 x0 + · · · + zs xs , z0 − β0 , . . . , zs − βs )
So, the result follows. 
We now obtain Theorem A when X is an irreducible scheme.
Remark 3.11. We first provide a couple of general words regarding the proof of Theorem 3.12
below and where the irreducibility assumption comes into play. The proof is achieved by itera-
tively cutting with generic hyperplanes (following Notation 3.8) to arrive to a zero-dimensional
situation, and the main constraint is to control the dimension of the image of all the possible pro-
jections after cutting with a general hyperplane (see (7)). Our main tool to control those dimen-
sions is Theorem 3.7, where it is needed to assume that R is a domain.
√ When X is irreducible, by
just taking the reduced scheme structure Xred = MultiProj(R/ 0) we can easily reduce to the
case where R is a domain. To maintain the irreducibility assumption during the inductive pro-
cess, we use a “generic” version of Bertini’s Theorem as presented in [16, Proposition 1.5.10].
It should be noted that the usual versions of Bertini’s Theorem for irreducibility require X to
be geometrically irreducible and that the dimension of the image of certain morphism is bigger
or equal than two (see [32, Theoreme 6.10, Corollaire 6.11]). Finally, Lemma 3.10 is used to
relate the process of cutting with a generic hyperplane with the one of cutting with a general
hyperplane.
Theorem 3.12. Assume Setup 2.1. Suppose that X is irreducible. Let n = (n1 , . . . , np ) ∈ Np
such that |n| = r. Then, degnP (X) > 0 if and only if for each J = {j1 , . . . , jk } ⊆ [p] the inequality
nj1 + · · · + njk 6 r(J) holds.
Proof. From Proposition 3.1 it is clear that the inequalities nj1 +· · ·+njk 6 r(J) are a necessary
condition for degnP (X) = e(n; R) > 0. Therefore, it suffices to show that they are also sufficient.
Assume that nj1 + · · · + njk 6 r(J) for every J = {j1 , . . . , jk } ⊆ [p]. We may also assume
∞
√ (0 :R N ) = 0 by Remark 2.6(i). Hence, the condition of X being irreducible implies that
that
0 ⊂ R is a prime ideal. Since the associativity formula for mixed multiplicities (see, e.g.,
[9, Lemma 2.7]) yields that
   √ 
e(n; R) = lengthR√ R√0 · e n; R/ 0 ,
0

we can assume that R is a domain, and we do so. In addition, by Remark 2.6(ii), Proposition 3.1,
and Remark 3.9 we may also assume that k is an infinite field.
We proceed by induction on r. If r = 0, then [9, Theorem 3.10] implies e(0; R) > 0.
Suppose now that r > 1. Without any loss of generality, perhaps after changing the grading,
we can assume that n1 > 1. Let L, RL , XL and z be defined as in Notation 3.8. Let x ∈
[R]e1 be a general element. Set S = R/xR, Z = X ∩ V(x) = MultiProj(S), T = RL /zRL , W =
XL ∩ V(z) = MultiProj(T ) and n 0 = n − e1 . Then, [9, Lemma 3.9] and Lemma 3.10 yield
that e(n; R) = e(n 0 ; S) = e(n 0 ; T ). From [16, Proposition 1.5.10] we obtain that W is also an
 WHEN ARE MULTIDEGREES POSITIVE? 15

irreducible scheme. By the assumed inequalities and because n1 > 1 we have that for each
J = {j1 , . . . , jk } ⊆ [p] the following inequality holds
nj01 + · · · + nj0k 6 min r(J), r J ∪ {1} − 1 ,


(7)
and the latter is equal to dim(ΠJ (Z)) by Theorem 3.7. Moreover, by Lemma 3.10 and Propo-
sition 3.1, we also have dim(ΠJ (W)) = dim(ΠJ (Z)); here, by an abuse of notation ΠJ (W)
mj mj
denotes the image of the natural projection ΠJ : P ⊗k L → PL 1 ×L · · · ×L PL k restricted to W.
Finally, by using the inductive hypothesis applied to the irreducible scheme W, we obtain
that e(n; R) = e(n 0 ; T ) > 0, and so the result follows. 
Now we are ready to show the general version of Theorem A.
Corollary 3.13. Assume Setup 2.1. Let n = (n1 , . . . , np ) ∈ Np such that |n| = dim(X). Then,
degnP (X) > 0 if and only if there is an irreducible component Y ⊆ X of X that satisfies the
following two conditions:
(a) dim(Y) = dim(X).
(b) For each J = {j1 , . . . , jk } ⊆ [p] the inequality nj1 + · · · + njk 6 dim ΠJ (Y) holds.

Proof. We may assume that (0 :R N∞ ) = 0 by Remark 2.6(i). By the associativity formula for
mixed multiplicities (see, e.g., [9, Lemma 2.7]) we get the equation
X
degnP (X) = e(n; R) = lengthRP (RP ) · e(n; R/P).
P∈Min(R)
dim(R/P)=r+p

Thus, e(n; R) > 0 if and only if e(n; R/P) > 0 for some minimal prime P ∈ Min(R) of maximal
dimension. So, the result is clear from Theorem 3.12. 
Below we have a proof for Theorem B.
Corollary 3.14. Assume Setup 2.12. Let n = (n1 , . . . , np ) ∈ Np such that dim (R/ (0 :R N∞ )) −
p = |n|. Then, e(n; R) > 0 if and only if there is a minimal prime ideal P ∈ Min (0 :R N∞ ) of
(0 :R N∞ ) that satisfies the following two conditions:
(a) dim (R/P) = dim (R/ (0 :R N∞ )).  
(b) For each J = {j1 , . . . , jk } ⊆ [p] the inequality nj1 + · · · + njk 6 dim P+PR m − k holds.
j6∈J j

Proof. As in Corollary 3.13, after assuming that (0 :R N∞ ) = 0 and using the associativity
formula for mixed multiplicities, we obtain that e(n; R) > 0 if and only if e(n; R/P) > 0 for
some minimal prime P ∈ Min(R) of maximal dimension. Note that, for each P ∈ Min(R),
R/P is naturally a finitely generated standard Np -graded algebra over a field. So, the result
follows by using Theorem 3.12. 
Finally, for the sake of completeness, we provide a brief discussion on how Theorem 3.12
can be recovered (over the complex number) from the related results of [34, §2.2].
m
Remark 3.15. Assume k = C. For the closed subscheme X ⊂ P = Pm p
k ×k · · · ×k Pk , let Li be
1

the pullback of OPmi (1) to X. Take Ei to be |Li |. Following the notation in [34, §2.2], for each
k
∅ 6= J ⊆ [p], denote by  
ΦJ : X → P E∨
J
16 FEDERICO CASTILLO, YAIRON CID-RUIZ, BINGLIN LI, JONATHAN MONTAÑO, AND NAIZHEN ZHANG

the Kodaira map corresponding with the linear system EJ . Let τJ be the dimension of the closure
of the image of ΦJ ([34, Definition 2.12]). Consequently, it is easy to check that dim (ΠJ (X)) =
τJ . Thus, [34, Theorems 2.14, 2.19] translate into the following statement: dim (ΠJ (X)) > |J|
if and only if for general hyperplanes Hj ∈ |Πj OP i (1)| (j ∈ J), X ∩ ( j∈J Hj ) 6= ∅. The latter
∗
T
m
Q k
is equivalent to the condition [X] · j∈J [Hj ] 6= 0 on intersection of classes. Theorem 3.12 (over
the complex numbers) eventually follows from applying this statement finitely many times to
relevant index subsets J.

4. P OSITIVITY OF THE MIXED MULTIPLICITIES OF IDEALS

In this section, we characterize the positivity of the mixed multiplicities of ideals. The re-
sults obtained here are a consequence of applying Theorem B to a certain multigraded algebra.
For the particular case of ideals generated in one degree in graded domains we obtain a neat
characterization in Theorem 4.4.
Throughout this section we use the following setup.
Setup 4.1. Let R be a Noetherian local ring with maximal ideal m ⊂ R (or a finitely generated
standard graded algebra over a field k with graded irrelevant ideal m ⊂ R).
Let J0 ⊂ R be an m-primary ideal and J1 , . . . , Jp ⊂ R be arbitrary ideals (homogeneous in the
graded case). The multi-Rees algebra of the ideals J0 , J1 , . . . , Jp is given by
i i
Ji00 · · · Jpp ti00 · · · tpp ⊂ R[t0 , . . . , tp ],
M
R(J0 , . . . , Jp ) := R[J0 t0 , . . . , Jp tp ] =
i0 >0,...,ip >0

where t0 , . . . , tp are new variables. Note that R(J0 , . . . , Jp ) is naturally a standard Np+1 -graded
algebra and that, for 0 6 k 6 p, the ideal mk generated by elements of degree ek is given by
mk := Jk tk R(J0 , . . . , Jp ) ⊂ R(J0 , . . . , Jp ).
Let N := m0 ∩· · ·∩mp be the corresponding multigraded irrelevant ideal. Since J0 is m-primary,
we obtain that
i  i
Ji00 Ji11 · · · Jpp Ji00 +1 Ji11 · · · Jpp
M
T (J0 | J1 , . . . , Jp ) := R(J0 , . . . , Jp ) ⊗R R/J0 =
i0 >0,i1 >0...,ip >0

is a finitely generated standard Np+1 -graded algebra over the Artinian local ring R/J0 . For sim-
plicity of notation, throughout this section we fix R := R(J0 , . . . , Jp ) and T := T (J0 | J1 , . . . , Jp ).
Let r be the integer r := dim (MultiProj (T )) , which coincides with the degree of the Hilbert
polynomial PT (u1 , . . . , up+1 ) of the Np+1 -graded R/J0 -algebra T . From [58, Theorem 1.2(a)]
we have the equality r = dim (R/(0 :R (J1 · · · Jp )∞ )) − 1.
Definition 4.2. Under the above notations, for each n ∈ Np+1 with |n| = r, we say that
en (J0 | J1 , . . . , Jp ) := e (n; T )
is the mixed multiplicity of J0 , J1 , . . . , Jp of type n.
The main focus in this section is to characterize when en (J0 | J1 , . . . , Jp ) > 0. As a direct con-
sequence of Theorem B we get the following general criterion for the positivity en (J0 | J1 , . . . , Jp ).
Corollary 4.3. Assume Setup 4.1 and the notations above. Let n = (n0 , n1 , . . . , np ) ∈ Np+1
such that |n| = r. Then, en (J0 | J1 , . . . , Jp ) > 0 if and only if there is a minimal prime ideal
P ∈ Min (0 :T N∞ ) of (0 :T N∞ ) that satisfies the following two conditions:
 WHEN ARE MULTIDEGREES POSITIVE? 17

(a) dim (T/P) = dim (T/ (0 :T N∞ )).  

(b) For each J = {j1 , . . . , jk } ⊆ {0} ∪ [p] the inequality, nj1 + · · · + njk 6 dim P+P T
−k
j6∈J mj T
holds.
We now focus on the case where R is a graded k-domain and each ideal Ji is generated in one
degree. In this case, our characterization depends on the analyticspread of certain
 ideals; recall
that the analytic spread of an ideal I ⊂ R is given by `(I) := dim R(I)/mR(I) .
Theorem 4.4. Let R be a finitely generated standard graded domain over a field k with graded
irrelevant ideal m ⊂ R. Let J0 ⊂ R be an m-primary ideal and J1 , . . . , Jp ⊂ R be arbitrary ideals.
Suppose that, for each 0 6 i 6 p, Ji is generated by homogeneous elements of the same degree
di > 0. Let n = (n0 , n1 , . . . , np ) ∈ Np+1 such that |n| = dim(R) − 1. Then, en (J0 | J1 , . . . , Jp ) > 0
if and only if for each J = {j1 , . . . , jk } ⊆ [p] the inequality


nj1 + · · · + njk 6 ` Jj1 · · · Jjk − 1
holds.
Proof. First, note that r = dim(R) − 1.
Since J0 is m-primary, the kernel of the canonical map T  T 0 := R ⊗R R/m is nilpotent.
Therefore, the conditions (a),(b) in Corollary 4.3 are satisfied for T if and only if they are
satisfied for T 0 .
Consider the Np+1 -graded domain given by
M h ii h i h i
i i
F := J00 Ji11 · · · Jpp ti00 ti11 · · · tpp ⊂ R.
i0 d0 i1 d 1 ip d p
i0 >0,...,ip >0
i
Since [Ji00 ]i d [Ji11 ]i d · · · [Jpp ]ip dp = ∼ Ji0 Ji1 · · · Jipp ⊗R R/m, we have the isomorphism T 0 =
∼ F and
0 0 1 1 0 1
so T 0 is a domain. P
For any K = {h1 , . . . , hs } ⊆ {0} ∪ [p], since we have the natural isomorphism R ∼
h6∈K mh =
R(Jh1 , . . . , Jhs ) = R[Jh1 th1 , . . . , Jhs ths ], it follows that
X
dim T 0 mh T 0 = dim R(Jh1 , . . . , Jhs ) ⊗R R/m .

h6∈K

R(J ∼


After using the Segre embedding we get the isomorphism MultiProj h 1
, . . . , Jh s ) ⊗ R R/m =
Proj R(Jh1 · · · Jhs ) ⊗R R/m and, accordingly, from Lemma 2.5(i) we have

dim R(Jh1 , . . . , Jhs ) ⊗R R/m = dim R(Jh1 · · · Jhs ) ⊗R R/m + s − 1 = `(Jh1 · · · Jhs ) + s − 1,

(also, see [2, Corollary 3.10]).

So, en (J0 | J1 , . . . , Jp ) > 0P
if and only if for each K = {h1 , . . . , hs } ⊆ 0 ∪ [p] the inequality
nh1 + · · · + nhs 6 dim(T / h6∈K mh T 0 ) − s = `(Jh1 · · · Jhs ) − 1 holds.
0

For any K = {0, h2 , . . . , hs } ⊆ {0} ∪ [p], as J0 is m-primary, from [27, Theorem 5.1.4, Proposi-
tion 5.1.6] we obtain
X 

dim T 0 mh T 0 = dim R(J0 , Jh2 , . . . , Jhs ) ⊗R R/J0 = dim grJ R(J ,...,J ) R(Jh2 , . . . , Jhs )



0 h2 hs
h6∈K
= dim(R) + s − 1.
Therefore, we only need check the inequalities corresponding to the subsets J ⊆ [p], and so the
result follows. 
18 FEDERICO CASTILLO, YAIRON CID-RUIZ, BINGLIN LI, JONATHAN MONTAÑO, AND NAIZHEN ZHANG

Remark 4.5. Note that in Theorem 4.4 the conditions for the positivity of en (J0 | J1 , . . . , Jp ) do
not involve the m-primary ideal J0 (see [58, Corollary 1.8(a)]).
Remark 4.6. We note that if in Theorem 4.4 we have `(Ji ) = dim(R) for every
1 6 i 6 p,
then by [29, Lemma 4.7] for each J = {j1 , . . . , jk } ⊆ [p] we also have ` Jj1 · · · Jjk = dim(R).
Therefore, by Theorem 4.4 it follows that en (J0 | J1 , . . . , Jp ) > 0 for every n ∈ Np+1 such that
|n| = dim(R) − 1.

5. P OLYMATROIDS
We recall that Theorem A implies that MSuppP (X) (see Definition 2.10) is the set of integer
points in a polytope when X is irreducible. In this section we explore properties of these discrete
sets.
Following standard notations, we say that X is a variety over k if X is a reduced and irre-
ducible separated scheme of finite type over k (see, e.g., [56, Tag 020C]). In the following two
results we connect the theory of polymatroids (see §2.3) with MSuppP (X) when X is a variety.
m
Proposition 5.1. Let X ⊆ P = Pm p
k ×k · · · ×k Pk be a multiprojective variety over an arbitrary
1

field k. Then MSuppP (X) is a discrete algebraic polymatroid over k.

Proof. We consider the associated Np -graded k-domain R. Let ξ be
the generic point of X and
set L := OX,ξ . For each J ⊆ [p], let XJ = ΠJ (X) = MultiProj R(J) and ξJ be the generic point
of XJ , and notice that

OXJ ,ξJ = f/g | f, g ∈ R(J) , g 6= 0, deg(f) = deg(g) ⊆ L.
For 1 6 i 6 p, let Li := OXi ,ξi . Then, for each J ⊆ [p], we have that OXJ ,ξJ = j∈J Lj and that
V

dim (XJ ) = trdegk OXJ ,ξJ (see, e.g., [20, Theorem 5.22], [22, Exercise II.3.20]). Finally, the


result follows from Theorem A. 
In [57, Corollary 2.8] it is shown that the conclusion of Theorem 3.12 holds if p = 2 and X is
arithmetically Cohen-Macaulay. The following example shows that this result does not always
hold for p > 2.
Example 5.2. Let S = k[x1 , . . . , x12 , y1 , . . . , y12 ] be a polynomial rings with an N12 -grading in-
duced by deg(xi ) = deg(yi ) = ei for 1 6 i 6 12. Let ∆ be the simplicial complex given by
the boundary of the icosahedron. We note that ∆ is a Cohen-Macaulay complex (because it
is a triangulation of the sphere S2 [54, Corollary II.4.4]), but it is not a (poly)matroid (see
Remark 2.15) since not every restriction is pure [54, Proposition III.3.1].
Let J∆ = {(xi1 + yi1 ) · · · (xik + yik ) | {i1 , . . . , ik } ∈
/ ∆} and X∆ = MultiProj(S/J∆ ) ⊂ P = P1k ×
· · · × P1k . The definition of J∆ is a modification on the definition of I∆ , the Stanley-Reisner
ideal of ∆ with monomials in the variables {x1 , . . . , x12 } [43, Chapter 1], [54, Chapter II]. It
can be easily verified that I∆ is the initial ideal of J∆ with respect to any elimination order
with {x1 , . . . , x12 } > {y1 , . . . , y12 }. Since the ideal J∆ is obtained from I∆ by a linear change of
variables, we have a similar primary decomposition as [43, Theorem 1.7], so no component
is supported on any coordinate subspace and thus J∆ is saturated with respect to the irrelevant
ideal of S. By [24, Corollary 3.3.5], X∆ is arithmetically Cohen-Macaulay. Moreover, since
Hilbert functions are preserved by Gröbner degenerations, the multidegree of X∆ coincides
with C(S/I∆ ; t) (see Remark 2.9). Thus, MSuppP (X∆ ) consists of all the incidence vectors of
the facets of ∆ [43, Theorem 1.7] and then it is not a polymatroid.
 WHEN ARE MULTIDEGREES POSITIVE? 19

With Proposition 5.1 in hand, we can introduce the following class of polymatroids.
Definition 5.3. A polymatroid P is Chow over a field k if there exists a variety X ⊆ P = Pm
k ×k
1
m
· · · ×k Pk p such that MSuppP (X) = P.
The following statement follows as an easy corollary of the main result in [37], when k is an
infinite field. Here we give a simple direct argument for an arbitrary field k.
Proposition 5.4. A linear polymatroid over an arbitrary field k is Chow over the same field.
Proof. Let V be a k-vector space and V1 , . . . , Vp be arbitrary subspaces. Let S be the polynomial
ring S = Sym(V) = k[x1 , . . . , xq ], where q = dimk (V). By using the isomorphism [S]1 = ∼ V, we
identify each Vi with a k-subspace Ui of [S]1 . For each 1 6 i 6 p, let {xi,1 , xi,2 , . . . , xi,ri } ⊂ [S]1
be a basis of the k-vector space Ui . Let T be the Np -graded polynomial ring
T := k yi,j | 1 6 i 6 p, 0 6 j 6 ri , deg(yi,j ) = ei .
 

Induce an Np -grading on S[t1 , . . . , tp ] given by deg(ti ) = ei and deg(xj ) = 0. Consider the

Np -graded k-algebra homomorphism
yi,0 7→ ti for 1 6 i 6 p
ϕ = T → S[t1 , . . . , tp ],
yi,j 7→ xi,j ti for 1 6 i 6 p, 1 6 j 6 ri .
Note that P := Ker(ϕ) is an Np -graded prime ideal. Set R := T/P and X := MultiProj(R). By
construction, for each J ⊆ [p], we obtain the isomorphism
∼ k xi,j ti | i ∈ J, 1 6 j 6 ri [ti | i ∈ J] ⊂ S[t1 , . . . , tp ];
 
R = (J)
thus, it is clear that
dim R(J) = trdegk k xi,j ti | i ∈ J, 1 6 j 6 ri [ti | i ∈ J]

 

= trdegk k xi,j | i ∈ J, 1 6 j 6 ri [ti | i ∈ J]

 

   
X X
= dimk  Ui  + |J| = dimk  Vi  + |J|.
i∈J i∈J
P

Therefore, Lemma 2.5 yields that dim (ΠJ (X)) = dimk i∈J Vi , and so the result follows
from Theorem A. 
The following is the main theorem of this section. Here we summarize the results presented
above to show that the class of Chow polymatroids lies in between the ones introduced in
Definition 2.16.
Theorem 5.5. Over an arbitrary field k, we have the following inclusions of families of poly-
matroids
     
Linear polymatroids ⊆ Chow polymatroids ⊆ Algebraic polymatroids .
Moreover, when k is a field of characteristic zero, the three families coincide.
Proof. The first inclusion follows from Proposition 5.4; the second from Proposition 5.1. In the
characteristic zero case linear and algebraic polymatroids coincide by [30, Corollary, Page 166]
(also, see Remark 5.6). 
Remark 5.6. The result mentioned above from [30] is stated for matroids but the arguments go
unchanged for polymatroids.
20 FEDERICO CASTILLO, YAIRON CID-RUIZ, BINGLIN LI, JONATHAN MONTAÑO, AND NAIZHEN ZHANG

Remark 5.7. Over finite fields there are algebraic matroids that are not linear. An example is
the Non-Pappus matroid described in [47, Page 517], it is algebraic over any field of positive
characteristic but not linear over any field.
Classifying linear polymatroid rank functions is a difficult problem. For linear matroids over
a field of characteristic zero, the poetically titled “The missing axiom of matroid theory is lost
forever” [59] together with a recent addition [40] shows that there is no finite list of axioms
that characterize which rank functions are linear. For fields of positive characteristic, Rota
conjectured in 1971 that for each field there is a list of finite restrictions. A proof of Rota’s
conjecture has been announced by Geelen, Gerards, and Whittle, but expected to be several
hundred of pages long. Little is known about the algebraic case. In [31] there is an example of
a matroid that is not algebraic over any field: the Vamos matroid V8 [47, Page 511]. For these
reasons we do not expect a further characterizations of Chow polymatroids.
We finish this section with the following question.
Question 5.8. Are all algebraic polymatroids Chow?

6. A PPLICATIONS
In this section we relate our results to several objects from combinatorial algebraic geometry.

6.1. Schubert polynomials. Let Sp be the symmetric group on the set [p]. For every i ∈
[p − 1] we have the transposition si := (i, i + 1) ∈ Sp . Recall that the set S = hsi , 1 6 i < pi
generates Sp . The length l(π) of a permutation π is the least amount of elements in S needed
to obtain π. Alternatively, the length is equal to the number of inversions, i.e., l(π) = {(i, j) ∈
[p] × [p] : i < j, π(i) > π(j)}. The permutation π0 = (p, p − 1, · · · , 2, 1) (in one line notation) is
p(p−1)
the longest permutation, it has length 2 .
Definition 6.1. The Schubert polynomials Sπ ∈ Z[t1 , . . . , tp ] are defined recursively in the fol-
Q
lowing way. First we define Sπ0 := i tp−i i , and for any permutation π and transposition si
with l(si π) < l(π) we let
Sπ − si Sπ
Ssi π = ,
ti − ti+1
where Sp acts on Z[t1 , . . . , tp ] by permutation of variables. For more information see [18,
Chapter 10].
Next we define matrix Schubert varieties following [43, Chapter 15]. Let k be an algebraic
closed field and Mp (k) be the k-vector space of p × p matrices with entries in k. As an affine
variety we define its coordinate ring as Rp := k[xij : (i, j) ∈ [p] × [p]]. Furthermore we consider
an Np -grading on Rp by letting deg(xij ) = ei .
Definition 6.2. Let π be a permutation matrix. The matrix Schubert variety Xπ ⊂ Mp (k) is the
subvariety
Xπ = {Z ∈ Mp (k) | rank(Zm×n ) 6 rank(πm×n ) for all m, n},
where Zm×n is the restriction

to the first m rows and n columns. This is an irreducible variety
and the prime ideal I Xπ is multihomogeneous [43, Theorem 15.31]. By [43, Theorem 15.40],
the Schubert polynomial

Sπ equals the multidegree polynomial of the variety corresponding to
the ideal I Xπ (see Definition 2.10).
 WHEN ARE MULTIDEGREES POSITIVE? 21
P
Following [45] we say a polynomial f = n cn tn ∈ Z[t1 , · · · , tp ] have the Saturated Newton
Polytope property (SNP for short) if supp(f) := {n ∈ Np | cn > 0} = ConvexHull{n ∈ Np | cn >
0} ∩ Np , in other words, if the support of f consist of the integer points of a polytope. In
[45, Conjecture 5.5] it was conjectured that the Schubert polynomials have SNP property and
they even conjectured a set of defining inequalities for the Newton polytope in [45, Conjecture
5.13]. A. Fink, K. Mézáros, and A. St. Dizier confirmed the full conjecture in [15]. As noted
by the authors of [26] the combination of Proposition 5.1 (they use the equivalent [3, Corollary
10.2]) and [43, Theorem 15.40] (which is also included in [26, Theorem 6]) is enough to give
an alternative proof to [45, Conjecture 5.5].
Theorem 6.3. For any permutation π, the Schubert polynomial Sπ has SNP and its Newton
polytope is a polymatroid polytope.
The Newton polytope of a polynomial f is by definition the convex hull of the exponents
in the support of f, however in by our convention in Definition 2.10 MSupp consists of the
complementary exponents. This does not change the conclusion that the resulting polytope is a
polymatroid polytope.
Codimensions of projections. We now use Theorem A to give a combinatorial interpretation
for the codimensions of the natural projections of matrix Schubert varieties. First we need some
terminology.
A diagram D is a subset of a p × p grid whose boxes are indexed by the set [p] × [p]. The
authors of [45] define a function θD : 2[p] 7→ Z as follows: for a subset J ⊆ [p] and c ∈ [p] , we
construct a word WD c (J) by reading the column c of [p] × [p] from top to bottom and recording

• ( if (r, c) ∈
/ D and r ∈ J,
• ) if (r, c) ∈ D and r ∈/ J,
• ? if (r, c) ∈ D and r ∈ J;
c c (J) + # ? in W c (J), and finally θ (J) =
Pp i
let θD (J) = # paired “()” in WD D D i=1 θD (J).

Example 6.4. For examplae, let D be the diagram depicted in Figure 1 and J = {2, 3}, then
θD (J) = 3.

F IGURE 1. Example of a diagram in [5] × [5]

For any π ∈ Sp we can define its Rothe diagram as

Dπ := {(i, j) | 1 6 i, j 6 n, π(i) > j and π−1 (j) > i} ⊂ [p] × [p].
For example when π = 42531 then Dπ is the diagram of Figure 1.


Theorem 6.5. Let π ∈ Sp , then for any J ⊆ [p] the projection ΠJ Xπ onto the rows indexed
by J has codimension θDπ ([p]) − θDπ (J 0 ), where J 0 = [p]\J is the complement of J.
22 FEDERICO CASTILLO, YAIRON CID-RUIZ, BINGLIN LI, JONATHAN MONTAÑO, AND NAIZHEN ZHANG

Proof. In [15, Theorem 10] the authors givea proof of [45, Conjecture 5.13], which in our setup
(recall the indexing in Definition 2.10) states that MSupp(Xπ ) is equal to
 
 X X 
n ∈ Np |



(p − 1) − nj 6 θDπ (J), ∀J ( [p], (p − 1) − nj = θDπ ([p]) .
 
j∈J j∈[p]
P
The first inequalities can be rewritten as (p − 1)|J| − θDπ (J) 6 j∈J nj , and combining them
P
with (p − 1)p − θDπ ([p]) = j∈[p] ni we obtain
X  
0
nj 6 (p − 1)|J | − θDπ ([p]) − θDπ (J) .
j∈J 0

= θDπ ([p]) − θDπ (J 0 ) for every J ⊆ [p], as

By Theorem A we must have that codim ΠJ Xπ
we wanted to show. 
Remark 6.6. Notice that θDπ ([p]) counts the total number of boxes in Dπ , which is equal to the
length of π (see [43, Definition 15.13]). So the case J = [p] of Theorem 6.5 above is equivalent
to the well-known fact that the codimesion of a matrix Schubert variety is equal to the length of
the permutation (see [43, Theorem 15.31]).
6.2. Flag varieties. We now focus on a multiprojective embedding of flag varieties. We first
review some terminology. For more information the reader is referred to [18] or [4].
In this subsection we work over an algebraically closed field k. Consider the complete flag
variety Fl(V) of a k-vector space V of dimension p + 1. This variety parametrizes complete
flags, i.e., sequences V• := (V0 , · · · , Vp+1 ) such that {0} = V0 ⊂ V1 ⊂ V2 ⊂ · · · ⊂ Vp ⊂ Vp+1 = V,
and each Vi is a linear subspace of V of dimension i. One can embed this variety in a product
of Grassmannians Fl(V) ,→ Gr(1, V) × Gr(2, V) × · · · × Gr(p, V) as the subvariety cut out by
incidence relations.
Furthermore, each Grassmannian can be embedded in a projective space via the Plücker
embedding ιi : Gr(i, V) → Pm i
k for 1 6 i 6 p. By considering the product of these maps, we
mp
obtain a multiprojective embedding of ι : Fl(V) ,→ Pm k ×k · · · ×k Pk . For convenience we also
1

call ι the Plücker embedding. The proposition below computes the corresponding multidegree
support.
Proposition 6.7. Let V be a k-vector space of dimension p + 1 and let X be the image of the
mp
Plücker embedding ι : Fl(V) ,→ P = Pm
k ×k · · · ×k Pk , then
1

(8)  
 X k X
k−1 X
p 
p+1 

MSuppP (X) = n ∈ N | 1 6 nk 6
p
(p − j) − ni , ∀k ∈ [p], nj = ;
 2 
j=1 i=1 j=1

Proof. We need to compute the dimension of ΠJ (X)) for each J = {j1 , . . . , jk } ⊆ [p]. The key
observation is that ΠJ (X) is isomorphic to the partial flag variety FlJ (V): it parametrizes flags
W• := {0} = V0 ⊂ V2 ⊂ V2 ⊂ · · · ⊂ Vk ⊂ Vk+1 = V, where dim Vk = jk . Hence
X
dim (ΠJ (X)) = dim (FlJ (V)) = di dj = S(J);
16i<j6k+1
P
here, for each J = {j1 , . . . , jk } ⊆ [p], we set S(J) := 16i<j6k+1 di dj , where di := ji − ji−1 and
by convention j0 := 0, jk+1 := p + 1. For a proof of the second equality see [4, §1.2]. From
 WHEN ARE MULTIDEGREES POSITIVE? 23

Theorem A it follows that

 
 X X
p 
p+1

(9) MSuppP (X) = n ∈ Np | nj 6 S(J), ∀J ⊆ [p], nj = .
 2 
j∈J j=1

It can be checked that the description in (9) coincides with the one in (8). 
m
The pullbacks of the classes Hi from Pm p
k ×k · · · ×k Pk to Fl(V) are called the Schubert
1

divisors, so Proposition 6.7 amounts to a criterion for which powers of these classes intersect.
These intersections are called Grassmannian Schubert problems in [50]. In [50, Theorem 1.2]
K. Purbhoo and F. Sottile give a stronger statement by providing an explicit combinatorial
formula using filtered tableau to compute the exact intersection numbers.

6.3. A multiprojective embedding of M0,p+3 . The moduli space M0,p+3 parametrizes ra-
tional stable curves with p + 3 marked points. Here we apply our methods to an embedding
considered in [8]. The starting point is the closed embedding Ψp : M0,p+3 −→ M0,p+2 ×k Ppk
constructed by S. Keel and J. Tevelev in [35, Corollary 2.7]. By iterating this construction we
obtain an embedding M0,p+3 ,→ P1k ×k P2k ×k · · · ×k Ppk (see [8, Corollary 3.2]). In [8], R. Cav-
alieri, M. Gillespie, and L. Monin computed the corresponding multidegree which turns out to
be related to parking functions. As an easy consequence of our Theorem A, we can compute its
support.
Proposition 6.8. Let X be the image of M0,p+3 ,→ P = P1k ×k P2k ×k · · · ×k Ppk , then
 
X
k X p
(10) MSuppP (X) = n ∈ Np | ni 6 k, ∀1 6 k 6 p − 1, ni = p .
i=1 i=1



Proof. First, as explained in [8, §3] we have dim Π[p] (X) = dim Π{p} (X) = p. Also, by
∼ M0,p+2 , and thus dim Π


construction Π[p−1] (X) = [i−1] (X) = i − 1 for every 2 6 i 6 p. So, by


induction one gets dim Π{i} (X) = i for all 1 6 i 6 p.
To use Theorem A, we must compute dim (Π
J (X)) for all J ⊆
[p]. Let m := max{j | j ∈ J},
then as explained above we have dim Π[m] (X) = dim Π{m} (X) = m and so we must have
dim (ΠJ (X)) = m. By Theorem A we obtain that MSuppP (X) is equal to
X X
p
{n ∈ N |
p
ni 6 max{j | j ∈ J}, ∀J ⊆ [p], ni = p},
i∈J i=1

but it is straightforward to check that the inequalities in (10) are enough to describe the same
set. 
The cardinality of MSuppP (X) is equal to the Catalan number Cn (see [53, Exercise 86]).
For a comprehensible survey on Catalan numbers see [53].

6.4. Mixed Volumes. In this subsection we assume k is an algebraically closed field. We

begin by reviewing the definition of mixed volumes of convex bodies, as a general reference
see [14, Chapter IV]. Let K = (K1 , . . . , Kp ) be a p-tuple of convex bodies in Rd . The volume
polynomial v(K) ∈ Z[w1 , . . . , wp ] is defined as
v(K1 , . . . , Kp ) := Vold (w1 K1 + · · · + wp Kp ).
24 FEDERICO CASTILLO, YAIRON CID-RUIZ, BINGLIN LI, JONATHAN MONTAÑO, AND NAIZHEN ZHANG

This is a homogeneous polynomial of degree d. If the coefficients of v(K) are written as

d
n
n V(K; n)w , then the numbers V(K; n) are called the mixed volumes of K. A natural question
to ask is: when are mixed volumes positive? The relation between mixed volumes and toric
varieties (see (11) below) together with Theorem A allows us to give another proof of a classical
theorem formulated on the non-vanishing of mixed volumes [51, Theorem 5.1.8].

Pp Let K = (K1 , .P
Theorem 6.9. . . , Kp ) be a p-tuple
P of convex

bodies in Rd . Then, V(K; n) > 0 if
and only if i=1 ni = d and i∈J ni 6 dim i∈J Ki for every subset J ⊆ [p].

We first indicate how to reduce to the case of polytopes. The basic idea is that convex bodies
can be approximated by polytopes in the Hausdorff metric [51, Section 1.8]. However, the
condition for positivity as stated in Theorem 6.9 is a priori not stable under limits. To fix this
we invoke an equivalent condition more suitable for the limiting argument.
Lemma 6.10. It suffices to show Theorem 6.9 for polytopes.
Proof. This follows from two facts. The first is that mixed volumes V(K; n) are continuous
[51, Theorem 5.1.7] and monotonous [51, Equation
p 5.25] on each entry. The second fact is that
for a given sequence K = (K1 , . . . , Kp ) ⊂ Rd of convex bodies, by [51, Lemma 5.1.9] the
following conditions are equivalent:
P P P
(1) pi=1 ni = d and i∈J ni 6 dim


i∈J K i for every subset J ⊆ [p].
(2) There exist line segments Si,1 , Si,2 , . . . , Si,ni ⊆ Ki , for every i, such that {Si,j }16i6p,16j6ni
has segments in d linearly independent directions.
We now assume the statement of Theorem 6.9 is true when each Ki is a polytope and show that
it follows in the case where each Ki is an arbitrary convex body.
If V(K; n) > 0 then by continuity we can find polytopes P = (P1 , · · · , Pp ) with V(P; n) > 0
and Pi ⊆ Ki for each i ∈ [p]. By assumption, the sequence P satisfies condition (2) above and
hence so does the sequence K.
Conversely, suppose condition (2) holds for K, then by continuity we can find polytopes
P = (P1 , . . . , Pp ) with Pi ⊆ Ki arbitrarily close so that (2) holds for P too. Due to the assumption,
we have V(P; n) > 0 and thus V(K; n) > V(P; n) > 0 by monotonicity. 
To finish the proof of Theorem 6.9 we need some preliminary results about toric varieties
and lattice polytopes. As an initial step we recall some facts about basepoint free divisors;
a general reference is [12, Section 6]. Let Σ be a fan and let P be a lattice polytope whose
normal fan coarsens Σ. Then, P induces a basepoint free divisor DP in the toric variety YΣ
[12, Proposition 6.2.5]. Here, being basepoint free means that the complete linear series |DP |
induces a morphism φP : YΣ → Pm ∗ ∗
k for some mi ∈ N such that φ (H) = DP ∈ A (YΣ ), where
i
mi
H is the class of a hyperplane in the projective space Pk .
Lemma 6.11. Let K1 , . . . , Kp be lattice polytopes and let K := K1 + · · · + Kp be their Minkowski
sum . Let Y be the toric variety associated to Σ, the normal fan of K, then
Q mj P
for each J ⊆ [p] we
have a map φJ : Y → j∈J Pk such that dim (φJ (Y)) = dim i∈J Ki .

Proof. The fan Σ is the common refinement of the normal fans of K1 , . . . , Kp [61, Proposition
7.12], so each Ki induces a basepoint free divisor Di on Y = YΣ and thus also a map φi :
Y → Pm i
k .QBy the universal property of fiber products these maps induce a canonical map
m
φJ : Y → j∈J Pk j for each J ⊆ [p]. It remains to compute the dimensions of the images of
these maps.
 WHEN ARE MULTIDEGREES POSITIVE? 25
Q m
By composing with the Segre embedding ϕJ : j∈J Pk j → Pm k we obtain a map ϕ ◦ φJ :
∗ (Pm ), we have that ϕ∗ (H) =
P J
Y → Pm . Let H be the class of a hyperplane in A i∈J Hi ∈
Qk  k J
m
A∗ j∈J Pk
j
where the each Hi is the pullback of a hyperplane in the i-th factor (see, e.g.,
P
[22, Exercise 5.11]). Then (ϕJ ◦ φJ )∗ (H) = i∈J Di ∈ A∗ (Y). ThisP means that the morphism
ϕJ ◦ φJ corresponds to the complete linear series |DKJ | where KJ = j∈J Ki , hence the image
P
has dimension dim(KJ ) = dim( j∈J Ki ) [12, Theorem 6.1.22]. 
Lemma 6.12. In the setup of Lemma 6.11, if J = [p] then after scaling each polytope if neces-
sary, φ = φJ is an embedding.
P
Proof. By construction the normal fan of K[p] = pi=1 Ki is Σ, so the corresponding divisor is
ample [12, Theorem 6.1.14]. By replacing the list of polytopes by large enough scalings we
obtain a very ample divisor, hence an embedding. 
Proof of Theorem 6.9. By using Lemma 6.10, we can assume that each Ki is a polytope. Ad-
ditionally, we can reduce to the case where each Ki is a lattice polytope since any polytope
can be approximated by lattice polytopes (see [17, Page 120]). Let K = K1 + · · · + Kp and let
Y be the toric projective variety associated to the normal fan of K. Each lattice polytope Ki
induces a basepoint free divisor Di on Y. As explained in [17, Eq. (2), Page 116], the funda-
mental connection between mixed volumes and intersection products is given by the following
equation
n

(11) V(K; n) = Dn1 1 · · · Dp p /d!,
where the numerator is the intersection product of the divisors in Y. Notice that positivity of
mixed volumes is unchanged by scaling so whenever Q needed we can scale each polytope.
By Lemma 6.12 we have an embedding φ : Y → pi=1 Pm i
k such that the pullback of each
Q p m
Hi ∈ A∗ i


i=1 Pk is Di . By using the projection formula [19, Proposition 2.5(c)], we can
n

consider the product Hn1 1 · · · Hp p /d! instead of the one in (11). From the fact that Y is irre-
ducible we are now in the Setup 2.1. So, the result follows by Remark 2.8, Theorem A, and
Lemma 6.11. 

ACKNOWLEDGMENTS
We thank the reviewer for his/her suggestions for the improvement of this work. We would
like to thank Chris Eur, Maria Gillespie, June Huh, David Speyer, and Mauricio Velasco for
useful conversations. We are also grateful with Frank Sottile for useful comments on an earlier
version (in particular for the simplification of the statement of Proposition 6.7). Special thanks
to Brian Osserman for many insightful conversations and encouragements. The computer alge-
bra system Macaulay2 [21] was of great help to compute several examples in the preparation
of this paper.

R EFERENCES
[1] P. B. Bhattacharya, The Hilbert function of two ideals, Proc. Cambridge Philos. Soc. 53 (1957), 568–575. ↑1
[2] C. Bivià-Ausina and J. Montaño, Analytic spread and integral closure of decomposable modules, arXiv
preprint arXiv:2001.08313 (2020). ↑17
[3] P. Brändén and J. Huh, Lorentzian polynomials (2020). arXiv:1902.03719. ↑3, 21
[4] M. Brion, Lectures on the geometry of flag varieties, Topics in cohomological studies of algebraic varieties,
Trends Math., Birkhäuser, Basel, 2005, pp. 33–85. ↑22
26 FEDERICO CASTILLO, YAIRON CID-RUIZ, BINGLIN LI, JONATHAN MONTAÑO, AND NAIZHEN ZHANG

[5] M. Brodmann and J. Rung, Local cohomology and the connectedness dimension in algebraic varieties, Com-
ment. Math. Helv. 61 (1986), no. 3, 481–490. ↑12
[6] W. Bruns and J. Herzog, Cohen-Macaulay rings, 2nd ed., Cambridge Studies in Advanced Mathematics,
Cambridge University Press, 1998. ↑10
[7] F. Castillo and F. Liu, Deformation Cones of Nested Braid Fans, International Mathematics Research Notices
(202006). rnaa090. ↑2
[8] R. Cavalieri, M. Gillespie, and L. Monin, Projective embeddings of M0,n and parking functions, arXiv
preprint arXiv:1912.12343 (2020). ↑23
[9] Y. Cid-Ruiz, Mixed multiplicities and projective degrees of rational maps, arXiv preprint arXiv:2001.00547
(2020). to appear in Journal of Algebra. ↑1, 6, 7, 8, 9, 10, 14, 15
[10] Y. Cid-Ruiz and A. Simis, Degree of rational maps and specialization, arXiv preprint arXiv:1901.06599
(2019). to appear in International Mathematics Research Notices. ↑6
[11] A. Conca, E. De Negri, and E. Gorla, Cartwright-Sturmfels ideals associated to graphs and linear spaces, J.
Comb. Algebra 2 (2018), no. 3, 231–257. ↑1
[12] D. A. Cox, J. B. Little, and H. K. Schenck, Toric varieties, Graduate Studies in Mathematics, vol. 124,
American Mathematical Society, Providence, RI, 2011. ↑24, 25
[13] D. Eisenbud, Commutative algebra with a view towards algebraic geometry, Graduate Texts in Mathematics,
150, Springer-Verlag, 1995. ↑13
[14] G. Ewald, Combinatorial convexity and algebraic geometry, Graduate Texts in Mathematics, vol. 168,
Springer-Verlag, New York, 1996. ↑23
[15] A. Fink, K. Mészáros, and A. St. Dizier, Schubert polynomials as integer point transforms of generalized
permutahedra, Adv. Math. 332 (2018), 465–475. ↑21, 22
[16] H. Flenner, L. O’Carroll, and W. Vogel, Joins and intersections, Springer Monographs in Mathematics,
Springer-Verlag, Berlin, 1999. ↑14
[17] W. Fulton, Introduction to toric varieties, Annals of Mathematics Studies, vol. 131, Princeton University
Press, Princeton, NJ, 1993. The William H. Roever Lectures in Geometry. ↑25
[18] , Young tableaux, London Mathematical Society Student Texts, vol. 35, Cambridge University Press,
Cambridge, 1997. With applications to representation theory and geometry. ↑20, 22
[19] , Intersection theory, Second, Ergebnisse der Mathematik und ihrer Grenzgebiete. 3. Folge. A Series
of Modern Surveys in Mathematics [Results in Mathematics and Related Areas. 3rd Series. A Series of
Modern Surveys in Mathematics], vol. 2, Springer-Verlag, Berlin, 1998. ↑25
[20] U. Görtz and T. Wedhorn, Algebraic geometry I, Advanced Lectures in Mathematics, Vieweg + Teubner,
Wiesbaden, 2010. Schemes with examples and exercises. ↑18
[21] D. R. Grayson and M. E. Stillman, Macaulay2, a software system for research in algebraic geometry. Avail-
able at http://www.math.uiuc.edu/Macaulay2/. ↑4, 25
[22] R. Hartshorne, Algebraic geometry, Springer-Verlag, New York-Heidelberg, 1977. Graduate Texts in Mathe-
matics, No. 52. ↑18, 25
[23] M. Herrmann, E. Hyry, J. Ribbe, and Z. Tang, Reduction numbers and multiplicities of multigraded structures,
J. Algebra 197 (1997), no. 2, 311–341. ↑1, 6, 9
[24] J. Herzog and T. Hibi, Monomial ideals, Graduate Texts in Mathematics, vol. 260, Springer-Verlag London,
Ltd., London, 2011. ↑3, 18
[25] J. Huh, Milnor numbers of projective hypersurfaces and the chromatic polynomial of graphs, Journal of the
American Mathematical Society 25 (2012), no. 3, 907–927. ↑1, 3
[26] J. Huh, J. P. Matherne, K. Mészáros, and A. St. Dizier, Logarithmic concavity of schur and related polyno-
mials, 2019. ↑21
[27] C. Huneke and I. Swanson, Integral closure of ideals, rings, and modules, Vol. 13, Cambridge University
Press, 2006. ↑3, 17
[28] E. Hyry, The diagonal subring and the Cohen-Macaulay property of a multigraded ring, Trans. Amer. Math.
Soc. 351 (1999), no. 6, 2213–2232. ↑5, 6
[29] , Cohen–macaulay multi–rees algebras, Compositio Math. 130 (2002), no. 3, 319–343. ↑18
[30] A. W. Ingleton, Representation of matroids, Combinatorial Mathematics and its Applications (Proc. Conf.,
Oxford, 1969), Academic Press, London, 1971, pp. 149–167. ↑19
[31] A. W. Ingleton and R. A. Main, Non-algebraic matroids exist, Bull. London Math. Soc. 7 (1975), 144–146.
↑20
 WHEN ARE MULTIDEGREES POSITIVE? 27

[32] J.-P. Jouanolou, Théorèmes de Bertini et applications, Université Louis Pasteur, Département de
Mathématique, Institut de Recherche Mathématique Avancée, Strasbourg, 1979. ↑14
[33] D. Katz, S. Mandal, and J. K. Verma, Hilbert functions of bigraded algebras, Commutative algebra (Trieste,
1992), World Sci. Publ., River Edge, NJ, 1994, pp. 291–302. ↑1
[34] K. Kaveh and A. G. Khovanskii, Complete intersections in spherical varieties, Selecta Math. (N.S.) 22 (2016),
no. 4, 2099–2141. ↑2, 3, 15, 16
[35] S. Keel and J. Tevelev, Equations for M0,n , Internat. J. Math. 20 (2009), no. 9, 1159–1184. ↑23
[36] A. Knutson and E. Miller, Gröbner geometry of Schubert polynomials, Ann. of Math. (2) 161 (2005), no. 3,
1245–1318. ↑1
[37] B. Li, Images of rational maps of projective spaces, Int. Math. Res. Not. IMRN 13 (2018), 4190–4228. ↑19
[38] L. Lovász, Submodular functions and convexity, Mathematical programming: the state of the art (Bonn,
1982), Springer, Berlin, 1983, pp. 235–257. ↑3
[39] H. Matsumura, Commutative ring theory, 1st ed., Cambridge Studies in Advanced Mathematics volume 8,
Cambridge University Press, 1989. ↑13
[40] D. Mayhew, M. Newman, and G. Whittle, Yes, the ‘missing axiom’ of matroid theory is lost forever, Trans.
Amer. Math. Soc. 370 (2018), no. 8, 5907–5929. ↑20
[41] M. Michałek, L. Monin, and J. Wiśniewski, Maximum likelihood degree and space of orbits of a C* action,
arXiv preprint arXiv:2004.07735 (2020). ↑1
[42] M. Michałek, B. Sturmfels, C. Uhler, and P. Zwiernik, Exponential varieties, Proc. Lond. Math. Soc. (3) 112
(2016), no. 1, 27–56. ↑1
[43] E. Miller and B. Sturmfels, Combinatorial commutative algebra, Graduate Texts in Mathematics, vol. 227,
Springer-Verlag, New York, 2005. ↑7, 8, 18, 20, 21, 22
[44] F. Mohammadi, C. Uhler, C. Wang, and J. Yu, Generalized permutohedra from probabilistic graphical mod-
els, SIAM J. Discrete Math. 32 (2018), no. 1, 64–93. ↑2
[45] C. Monical, N. Tokcan, and A. Yong, Newton polytopes in algebraic combinatorics, Selecta Math. (N.S.) 25
(2019), no. 5, Paper No. 66. ↑2, 21, 22
[46] K. Murota, Discrete convex analysis, SIAM Monographs on Discrete Mathematics and Applications, Society
for Industrial and Applied Mathematics (SIAM), Philadelphia, PA, 2003. ↑3
[47] J. G. Oxley, Matroid theory, Oxford Science Publications, The Clarendon Press, Oxford University Press,
New York, 1992. ↑9, 20
[48] A. Postnikov, V. Reiner, and L. Williams, Faces of generalized permutohedra, Doc. Math. 13 (2008), 207–
273. ↑2
[49] A. Postnikov, Permutohedra, associahedra, and beyond, Int. Math. Res. Not. IMRN 6 (2009), 1026–1106.
↑2, 3
[50] K. Purbhoo and F. Sottile, A Littlewood-Richardson rule for Grassmannian permutations, Proc. Amer. Math.
Soc. 137 (2009), no. 6, 1875–1882. ↑23
[51] R. Schneider, Convex bodies: the Brunn-Minkowski theory, expanded, Encyclopedia of Mathematics and its
Applications, vol. 151, Cambridge University Press, Cambridge, 2014. ↑24
[52] A. Schrijver, Combinatorial optimization. Polyhedra and efficiency. Vol. B, Algorithms and Combinatorics,
vol. 24, Springer-Verlag, Berlin, 2003. Matroids, trees, stable sets, Chapters 39–69. ↑3
[53] R. P. Stanley, Catalan numbers, Cambridge University Press, New York, 2015. ↑23
[54] R. P. Stanley, Combinatorics and commutative algebra, 2., Birkhäuser, Boston, 1996. ↑18
[55] B. Teissier, Cycles évanescents, sections planes et conditions de whitney, Astérisque 7 (1973), no. 8, 285–
362. ↑3
[56] The Stacks project authors, The stacks project, 2019. ↑18
[57] N. V. Trung, Positivity of mixed multiplicities, Math. Ann. 319 (2001), no. 1, 33–63. ↑1, 2, 10, 12, 18
[58] N. V. Trung and J. Verma, Mixed multiplicities of ideals versus mixed volumes of polytopes, Trans. Amer.
Math. Soc. 359 (2007), no. 10, 4711–4727. ↑3, 16, 18
[59] P. Vámos, The missing axiom of matroid theory is lost forever, J. London Math. Soc. (2) 18 (1978), no. 3,
403–408. ↑20
[60] B. L. Van der Waerden, On Hilbert’s function, series of composition of ideals and a generalization of the
theorem of Bezout, Proc. roy. acad. amsterdam, vol. 31, 1929, pp. 749–770. ↑1, 7
[61] G. M. Ziegler, Lectures on polytopes, Graduate Texts in Mathematics, vol. 152, Springer-Verlag, New York,
1995. ↑24
28 FEDERICO CASTILLO, YAIRON CID-RUIZ, BINGLIN LI, JONATHAN MONTAÑO, AND NAIZHEN ZHANG

(Castillo) M AX P LANCK I NSTITUTE FOR M ATHEMATICS IN THE S CIENCES , I NSELSTRASSE 22, 04103
L EIPZIG , G ERMANY
Email address: [email protected]

(Cid-Ruiz) M AX P LANCK I NSTITUTE FOR M ATHEMATICS IN THE S CIENCES , I NSELSTRASSE 22, 04103
L EIPZIG , G ERMANY
Email address: [email protected]
URL: https://ycid.github.io

(Li) D EPARTMENT OF S TATISTICS , 310 H ERTY D RIVE , U NIVERSITY OF G EORGIA , ATHENS , GA 30602,
USA
Email address: [email protected]

(Montaño) D EPARTMENT OF M ATHEMATICAL S CIENCES , N EW M EXICO S TATE U NIVERSITY , PO B OX

30001, L AS C RUCES , NM 88003-8001
Email address: [email protected]

(Zhang) I NSTITUT F ÜR D IFFERENTIALGEOMETRIE , L EIBNIZ U NIVERSIT ÄT H ANNOVER , W ELFENGARTEN

1, 30167 H ANNOVER , G ERMANY
Email address: [email protected]