arXiv reader

Positroids, Dressian and stable polynomials

Ayush Kumar Tewari

arXiv
2309.17091v1 · PDF · 29 September 2023
Categories
math.AG, math.CO
Licence
CC BY 4.0 The site shows the text.

Abstract

Our work is motivated by the connection established between Lorentzian polynomials and the Dressian in the seminal work of Brändén and Huh on Lorentzian polynomials. We analyze this relation for the class of positroids, and are able to show that in this case, we can relate a multiaffine homogenous stable polynomial to it. Additionally, we also highlight that a conjecture for matroids posed by Brändén and Huh is true when considered over the class of Rayleigh matroids which strictly contain the class of positroids. We collect these findings along with other results for further exploration.

Items 37

Theorem

Theorem 1

NE3W

Let V represent a nonegative point in Gr(k,n) and let \mathcal{M} be a positroid of rank k on [n] elements with bases \mathcal{B} that corresponds to V. Let p_{B} represent the nonzero Plücker coordinates of V, which also satisfy the quadratic Plücker relations. Then for \mu \in \text{Trop}\>Gr(\mathcal{M}) \subseteq Dr(\mathcal{M}), the multi affine homogenous stable polynomial f that represents V,

f = \sum_{B \in \mathcal{B}} p_{B} \> x^{B}

in \mathbb{K}[x_{1} , \hdots , x_{n}] is the one whose tropicalization is \mu.

Corollary

Corollary 2

CC2K

The following conditions are equivalent for any non-empty J \subseteq \{0, 1\}^{n} :

  1. J is the set of bases of a Rayleigh matroid on [n].

  2. The generating function f_{J} is a homogeneous 1-Rayleigh polynomial.

Essentially, Conjecture 3.12 [branden2020lorentzian] is true for the class of Rayleigh matroids.

Definition

Definition 3

BCVI

Set L^{0}_{n} = P^{0}_{n}, L^{1}_{n} = P^{1}_{n} , and

L^{2}_{n} = \{ f \in P^{2}_{n} \> \> | \> \> \mathcal{H}_{f} \> \> \text{is nonsingular and has exactly one positive eigenvalue} \}

For d\geq 2, L^{d}_{n} is defined recursively as follows

L^{d}_{n} = \{ f \in P^{d}_{n} \> \> | \> \> \partial_{i}f \in L^{d-1}_{n} \> \> \text{for all} \> \> i \in [n] \}

The polynomials in L^{d}_{n} are called strictly Lorentzian, and the limits of strictly Lorentzian polynomials are called Lorentzian.

Used by Theorem 15, Theorem 24, Proof
Definition

Definition 4

HV5H

A subset J \subset \mathbb{N}^{n} to be M-convex if it satisfies any one of the following equivalent conditions:

  • For any \alpha, \beta \in J and any index i satisfying \alpha_{i} > \beta_{i}, there is an index j satisfying

    \alpha_{j} < \beta_{j} \quad \text{and} \quad \alpha - e_{i} + e_{j} \in J

  • For any \alpha, \beta \in J and any index i satisfying \alpha_{i} > \beta_{i}, there is an index j satisfying

    \alpha_{j} < \beta_{j} \quad \text{and} \quad \alpha - e_{i} + e_{j} \in J \quad \text{and} \quad \beta - e_{i} + e_{j} \in J
Definition

Definition 5

TL9O

A function \mu from \mathbb{N}^{n} to \mathbb{R} \cup \{\infty\} is said to be M-convex if it satisfies the symmetric exchange property:

  • For any \alpha, \beta \in \text{dom}(\mu) and any i satisfying \alpha_{i} > \beta_{i} , there is j satisfying

    \alpha_{j} < \beta_{j} \quad \text{and} \quad \mu(\alpha) + \mu(\beta) \geq \mu(\alpha - e_{i} + e_{j} ) + \mu(\beta - e_{j} + e_{i} ).
Definition

Definition 6

WXK9

Set L^{0}_{n}(\mathbb{K}) = M^{0}_{n}(\mathbb{K}), L^{1}_{n}(\mathbb{K}) = M^{1}_{n}(\mathbb{K}), and

L^{2}_{n}(\mathbb{K}) = \{ f_{t} \in M^{2}_{n}(\mathbb{K}) \>\> | \>\> \text{The Hessian of} \>\> f_{t} \>\> \text{has at most one eigenvalue in} \>\> \mathbb{K}_{\geq 0} \}

For d \geq 3, we define L^{d}_{n}(\mathbb{K}) as follows

L^{d}_{n}(\mathbb{K}) = \{ f_{t} \in M^{d}_{n}(\mathbb{K}) \>\> | \>\> \partial^{\alpha}f_{t} \in L^{2}_{n}(\mathbb{K}) \>\> \text{for all} \>\> \alpha \in \Delta^{d-2}_{n} \}

where \Delta^{d-2}_{n} represents the discrete (d-2)-th simplex.

Remark

Remark 8

7SSO

In [choe2004homogeneous], the half-plane property is also defined for a matroid. \mathcal{M} is said to satisfy the half plane property if the basis generating polynomial of \mathcal{M} satisfies the half-plane property. Additionally, \mathcal{M} is said to satisfy the weak half plane property if there exists a polynomial that satisfies the half-plane property and whose support is the basis of \mathcal{M} [choe2004homogeneous].

Uses Definition 7
Definition

Definition 9

V557

Let c be a fixed positive real number, and let f be a polynomial in \mathbb{R}[w_{1} , \cdots , w_{n}]. f is called c-Rayleigh if f has nonnegative coefficients and

\partial^{\alpha}f(x) \> \partial^{\alpha+e_{i}+e_{j}}f(x) \leq c \> \partial^{\alpha+e_{i}}f(x) \> \partial^{\alpha+e_{j}}f(x) \quad \text{for all} \>\> i,j \in [n], \alpha \in \mathbb{N}^{n}, x \in \mathbb{R}^{n}_{\geq 0}

When f is multi-affine, that is, when f has degree at most one in each variable, the c-Rayleigh condition for f is equivalent to

f(x) \partial^{i} \partial^{j}f(x) \leq c \>\> \partial^{i}f(x) \partial^{j}f(x) \>\> \text{for all distinct} \>\> i,j \in \mathbb{N}^{n}, \>\> \text{and} \>\> x \in \mathbb{R}^{n}_{\geq 0}
Used by Conjecture 17, Corollary 18, Proof, Problem 32
Remark

Remark 10

OGZU

We acknowledge that in some instances, all Plücker relations (not necessarily only the 3-term Plücker relations) are considered for the definition of the Dressian. However, by the work in [baker2019matroids] we know that these two notions coincide over the tropical hyperfield and hence our definition would remain consistent for our discussion.

Theorem

Theorem 11

DSOZ

Let \mu be a point in Dr(k,n). Then the following are equivalent

  1. \mu is a valuated matroid with the underlying matroid \mathcal{U}_{k,n}.

  2. \mu defines a M-convex function on the matroid \mathcal{U}_{k,n}.

  3. \mu satisfies the tropical three-term Plücker relations.

  4. \mu as a weight vector induces a regular matroidal subdivision of \Delta(k,n).

  5. \mu = \text{trop}(f_{t}), where f_{t} is a Lorentzial polynomial defined on the bases set \mathcal{B} of \mathcal{U}_{k,n}.

  6. \mu defines a metric for a generalised metric tree arrangement.

Theorem

Theorem 12

9OCX

Let \mu be a point in Dr(\mathcal{M}). Then the following are equivalent

  1. \mu is a valuated matroid with the underlying matroid \mathcal{M}.

  2. \mu defines a M-convex function on the matroid \mathcal{M}.

  3. \mu satisfies the tropical three-term Plücker relations.

  4. \mu as a weight vector induces a regular matroidal subdivision of \mathcal{P}_{M}.

  5. \mu = \text{trop}(f_{t}), where f_{t} is a Lorentzial polynomial defined on the bases set \mathcal{B} of \mathcal{M}.

Theorem

Theorem 13

BDWJ

Let f \in \mathbb{C}[x] be a multi-affine homogenous polynomial of degree k that represents a point V in Gr(k,n). Then, f is stable if and only if V is totally nonnegative.

Uses Definition 7 · Used by Proof
Corollary

Corollary 14

5XVA

Let f \in \mathbb{R}[x] be a multi-affine homogenous polynomial of degree k that represents a point V in Gr(k,n). Then, f is stable if and only if V is totally nonnegative. Moreover, if \mathbb{K} is a real closed field and f \in \mathbb{K}[x] be a multi-affine homogenous polynomial of degree k that represents a point V in Gr(k,n) then, f is stable if and only if V is totally nonnegative.

Uses Definition 7 · Used by Proof
Proof of Corollary 14

With the well-known phase theorem [choe2004homogeneous, Theorem 6.1] we know that every homogeneous stable polynomial has non-negative real coefficients up to scalar multiples. Hence, the stability of the polynomial in Theorem 13 over \mathbb{C} is equivalent to stability of over \mathbb{R} [purbhooemail].

Since, the field of real numbers also allows quantifier elimination, therefore by invoking the Tarski principle as done in [branden2020lorentzian] , we can extend Corollary 14 over real closed fields as well.

Uses Theorem 13, Definition 7
Theorem

Theorem 15

S2M9

The following conditions are equivalent for any function \mu : \Delta(k,n) \rightarrow \mathbb{Q} \cup \{ \infty \}

  1. the function \mu is M-convex;

  2. there is a Lorentzian polynomial in \mathbb{K}[w_{1} , \hdots , w_{n}] whose tropicalization is \mu.

Uses Definition 3
Theorem

Theorem 16

5LWK

Let V represent a nonnegative point in Gr(k,n) and let \mathcal{M} be a positroid of rank k on [n] elements with bases \mathcal{B} that corresponds to V. Let p_{B} represent the nonzero Plücker coordinates of V, which also satisfy the quadratic Plücker relations. Then for \mu \in \text{Trop}\>Gr(\mathcal{M}) \subseteq Dr(\mathcal{M}), the multi affine homogenous stable polynomial f that represents V,

f = \sum_{B \in \mathcal{B}} p_{B} \> x^{B}

in \mathbb{K}[x_{1} , \hdots , x_{n}] is the one whose tropicalization is \mu.

Uses Definition 7
Proof of Theorem 16

We consider the homogenous multi-affine polynomial,

f = \sum_{B \in \mathcal{B}} p_{B} \> x^{B}

where x^{B} = \prod_{b \in B} x_{b}, which represents V. By Corollary 14, this polynomial is stable over \mathbb{K} with coefficients satisfying the quadratic Plücker relations. We also realize that the tropicalization \text{trop}(f) is defined by the tropicalization of the coefficients which are just Plücker coordinates. However, these tropical Plücker relations are the relations that define a point in the tropical Grassmannian \text{TropGr}(k,n). Therefore, we realize that \text{trop}(f) has coefficients that satisfy the tropical Plücker relations, which defines a point \mu \in \text{TropGr}(k,n) \subseteq Dr(\mathcal{M}).

Uses Corollary 14, Definition 7
Conjecture

Conjecture 17

IZQX

The following conditions are equivalent for any non-empty J \subseteq \{0, 1\}^{n} :

  1. J is the set of bases of a matroid on [n].

  2. The generating function f_{J} is a homogeneous \frac{8}{7} -Rayleigh polynomial.

Uses Definition 9
Corollary

Corollary 18

VPI9

The following conditions are equivalent for any non-empty J \subseteq \{0, 1\}^{n} :

  1. J is the set of bases of a Rayleigh matroid on [n].

  2. The generating function f_{J} is a homogeneous 1-Rayleigh polynomial.

Essentially, Conjecture 17 is true for the class of Rayleigh matroids.

Uses Definition 9
Proof of Corollary 18

For the implication (i) \implies (ii) by the definition of Rayleigh matroids in [choe2006rayleigh], the basis generating function of a Rayleigh matroid is 1-Rayleigh. Also, if f is a 1-Rayleigh polynomial, then f is c-Rayleigh polynomial for all c \geq 1. Additionally, for the opposite implication i.e., (ii) \implies (i) is straightforward in this case by [branden2020lorentzian, Theorem 3.10], which implies that J is the basis of a matroid, which in turn is Rayleigh by our assumption in this implication. Therefore, this helps us to see that Conjecture 17 is true for the class of Rayleigh matroids.

Uses Definition 9
Theorem

Theorem 19

B32X

The positive tropical Grassmannian \text{Trop}^{+} Gr(k,n) equals the positive Dressian Dr^{+}(k,n).

Theorem

Theorem 20

VQ25

Let \mathcal{M} be a positroid. Then the positive Dressian Dr^{+}(\mathcal{M}) equals the positive tropical positroidal cell \text{Trop}^{+} \Pi_{\mathcal{M}}.

Theorem

Theorem 21

VB6V

Let \mu be a point in Dr^{+}(d,n). Then the following are equivalent

  1. \mu is a realizable valuated matroid with the underlying positroid \mathcal{U}_{d,n}.

  2. \mu defines a realizable M-convex function.

  3. \mu satisfies the positive tropical three-term Plücker relations.

  4. \mu as a weight vector induces a regular positroidal subdivision of \Delta(d,n).

  5. \mu = \text{trop}(f_{t}), where f_{t} is a Lorentzial polynomial defined on the bases set \mathcal{B} of the positroid \mathcal{U}_{d,n}.

Theorem

Theorem 22

NW7Z

Let \mathcal{M} be a positroid and \mu be a point in Dr^{+}(\mathcal{M}). Then the following are equivalent

  1. \mu is a realizable valuated matroid with the underlying positroid \mathcal{M}.

  2. \mu defines a M-convex function on the positroid \mathcal{M}.

  3. \mu satisfies the positive tropical three-term Plücker relations.

  4. \mu as a weight vector induces a regular positroidal subdivision of \mathcal{P}_{\mathcal{M}}.

  5. \mu = \text{trop}(f_{t}), where f_{t} is a Lorentzial polynomial defined on the bases set \mathcal{B} of the positroid \mathcal{M}.

Proof of Theorem 22

The equivalence between (i) \iff (ii) is by definition. The equivalences for (ii) \iff (iii) are established in [arkani2021positive, Proposition 8.2]. We concentrate on proving the equivalence of (i). We first concentrate on proving the equivalence (i) \iff (iii). We already know that a point \mu \in Dr^{+}(\mathcal{M}) \implies \mu \in Dr(\mathcal{M}), hence the implication that \mu defines a valuated matroid follows from the inclusion in the Dressian Dr(\mathcal{M}), the only nontrivial part being the realizability of \mu. The stratification of the tropical positive Grasmmanian for a positroid is discussed in [arkani2021positive] and it is defined in terms of positroidal cells as

\text{Trop}^{+} \Pi_{\mathcal{M}} = \text{val}(\Tilde{\Pi}_{\mathcal{M}}(\mathcal{R}_{\geq 0})) = Dr^{+}(\mathcal{M})

where \mathcal{R}_{\geq 0} represents the positive part of the generalized Puiseux series and val() represents the valuaton map. The equality in ? implies that all positive Plücker vectors are realizable therefore by [arkani2021positive, Theorem 9.2], [brandt2021tropical, Remark 3.1.7] implies that the valuated matroid \mu on the positroid \mathcal{M} is realizable. So (i) \iff (iii). The implication in (v) is straightforward with the inclusion of \mu \in Dr(\mathcal{M}) restricted to a positroid.

Theorem

Theorem 23

7OMY

Let \mu = (\mu_{1} , \hdots, \mu_{k}) be a sequence of valuated matroids such that its sequence of underlying matroids M = (M_{1}, \hdots, M_{k}) is a flag matroid. Then the following are equivalent:

  1. \mu is a point on FlDr(M), i.e. it satisfies tropical incidence-Plücker relations and Grassmann-Plücker relations,

  2. \mu is a valuated flag matroid with underlying flag matroid M.

  3. \mu induces a subdivision of the base polytope of M into base polytopes of flag matroids,

  4. the projective tropical linear spaces \overline{trop(\mu_{i})} form a flag \overline{trop(\mu_{1})} \subseteq \hdots \subseteq \overline{trop(\mu_{k})}

Used by Proof
Theorem

Theorem 24

713R

Let \mu = (\mu_{1}, \hdots, \mu_{k}) be a sequence of valuated matroids such that its sequence of underlying matroids M = (M_{1}, \hdots, M_{k}) is a flag matroid. Then the following are equivalent

  1. \mu is a point on FlDr(M),

  2. there exist a sequence of Lorentzian polynomials f_{1}, \hdots f_{k} such that \mu = (\text{trop}(f_{1}) = \mu_{1}, \hdots, \text{trop}(f_{k})= \mu_{k}), where (\mu_{1}, \hdots \mu_{k}) forms a valuated flag matroid and f_{i} is a Lorentzial polynomial defined on the bases set B_{i} of M_{i}.

Uses Definition 3
Proof of Theorem 24

We first consider the implication (i) \implies (ii). Each valuated matroid \mu_{i} corresponds to a M-convex function on M_{i} by definition, and by [branden2020lorentzian, Theorem 3.20] this further corresponds to a Lorentzian polynomial f_{i} such that \text{trop}(f_{i}) = \mu_{i}. Therefore, for \mu \in FlDr(M), there exists a sequence of Lorentzian polynomials f_{1}, \hdots f_{k}, such that \mu = (\text{trop}(f_{1}) = \mu_{1}, \hdots, \text{trop}(f_{k}) = \mu_{k}), where f_{i} is a Lorentzial polynomial defined on the bases set B_{i} of M_{i}. We now try to prove the implication (ii) \implies (i). Given a sequence of Lorentzian polynomials such that \mu = (\text{trop}(f_{1}) = \mu_{1}, \hdots, \text{trop}(f_{k})= \mu_{k}) is a valuated flag matroid on the underlying matroid M = (M_{1}, \hdots , M_{k}), which is equivalent to saying that \mu \in FlDr(M) by 23.

Uses Theorem 23, Definition 3
Definition

Definition 25

KD7T

A homogeneous polynomial h(x) \in \mathbb{R}[x_{1} , \cdots , x_{n}] is hyperbolic with respect to a vector e \in \mathbb{R}^{n} if h(e) \neq 0, and if for all x \in \mathbb{R}^{n} the univariate polynomial t \rightarrow h(te - x) has only real zeros.

Lemma

Lemma 26

4GX4

Let P \in \mathbb{R}[x_{1} , \cdots , x_{n}] be a homogeneous polynomial. Then P is stable if and only if P is hyperbolic with \mathbb{R}^{n}_{+} \subseteq \Lambda_{+}.

Uses Definition 7
Definition

Definition 27

UZEQ

Let X \subset \mathbb{C}^{n} be a variety which is equidimensional of codimension c \leq n-1. X is called positively hyperbolic if for every linear subspace L in the positive Grassmannian Gr^{+}(c, n) and every x \in X, the imaginary part Im(x) does not belong to L \setminus \{0\}. A projective variety in \mathbb{P}^{n-1} is positively hyperbolic if its affine cone in \mathbb{C}^{n} is.

Definition

Definition 28

PB9I

Let \mathcal{M} be a matroid on the ground set E with basis \mathcal{B}. Let \text{Pr}(e) denote the probability of an element e being present in basis element B, which is chosen uniformly at random. The matroid \mathcal{M}(E, \mathcal{B}) is said to satisfy the negatively correlated property if

\text{Pr}(ef) \leq \text{Pr}(e)\text{Pr}(f)

for all pair of distinct e,f \in E.

Definition

Definition 29

P7WP

A matroid \mathcal{M} is said to be balanced if all its minors including itself satisfy the negative correlation property.

Problem

Problem 30

SSH8

Verify Conjecture 17 for the class of transversal matroids.

Problem

Problem 31

NNWM

Find the largest class of transversal matroids for which it satisfies the half-plane property.

Source and licence

The text on this page is from arXiv:2309.17091v1 by Ayush Kumar Tewari, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/2309/2309.17091.jsonl.