betaFeedback

Reading v1View latest version

WorksMathematics

Linear bounds and sharp exponential growth for commuting matrix algebra

Abstract

For any mm pairwise commuting matrices \{A_i\}_{i=1}^{m}\subset \M_n(\F), we prove \dim_\F \F[A_1,\cdots, A_m]\leq (2^m-1)(n-1)+1, answering Bergman's question on the existence of a linear bound for a fixed number mm.

Introduction

Let F\mathbb{F} be a field. For pairwise commuting matrices A1,…,Am∈Mn(F)A_1,\ldots,A_m \in\mathrm{M}_n(\mathbb{F}), this paper is concerned with linear upper bounds for the dimension of F[A1,…,Am]\mathbb{F}[A_1,\ldots,A_m].

Gerstenhaber problem

Gerstenhaber problem. For a single matrix A1A_1, the Cayley–Hamilton theorem says that the dimension of F[A1]\mathbb{F}[A_1] is at most nn. For two commuting matrices A1,A2A_1,A_2, the dimension of F[A1,A2]\mathbb{F}[A_1,A_2] is again at most nn, which was proved by Motzkin and Taussky [23] and, independently, by Gerstenhaber [11], both arguments resting on the irreducibility of the variety of commuting pairs. Linear-algebraic and commutative-algebraic proofs were found later [2, 20, 32, 3]. For four or more matrices, the corresponding bound fails [11]. Correspondingly, the variety of commuting mm-tuples of n×nn \times n matrices is reducible when m,n≥4m,n \geq4 [11, 12].

The case of three matrices lies between these two regimes and remains open.

Gerstenhaber Problem. Is dim⁡FF[A,B,C]≤n\dim_{\mathbb{F}}\mathbb{F}[A,B,C] \leq n for any pairwise commuting A,B,C∈Mn(F)A,B,C \in\mathrm{M}_n(\mathbb{F})?

We refer to [29, 15] for surveys. For n≤10n \leq10, the answer is affirmative: in this range, the variety of commuting triples of n×nn \times n matrices over an algebraically closed field is irreducible [13, 17, 25, 14, 30, 31], and Gerstenhaber’s argument for pairs then applies. For n≥29n \geq29, the variety of commuting triples is reducible [17, 24] (see also [19] for its components), so the geometric approach cannot settle this problem for all nn. Affirmative answers are known for various special classes of triples, for example, when one of the matrices has square zero [17] or small nullity [13], and for several combinatorially defined classes [26, 27, 9, 8]. Holbrook and O’Meara [16] have proposed a computational strategy for resolving the problem.

2020 Mathematics Subject Classification. Primary 15A27; Secondary 16G99, 13P10, 14L30, 52A39.

Key words and phrases. commuting matrices; algebras of matrices; Gerstenhaber problem; Bergman question.

Bergman question

It is well-known that every commutative subalgebra of Mn(F)M_n(\mathbb{F}) has dimension at most ⌊n24⌋+1\left\lfloor\frac{n^2}{4}\right\rfloor+1, by theorems of Schur [28] and Jacobson [18] (see also [22]), and this bound is sharp; the extremal algebras, however, require on the order of n2n^2 generators. For a fixed number of generators, subquadratic bounds arose in the representation theory of pp-adic groups, in connection with Bernstein’s uniform admissibility theorem [4]. It states that mm pairwise-commuting complex n×nn \times n matrices generate an algebra of dimension at most n2−21−mn^{2-2^{1-m}} (see [5 §4.10][6 Proposition 20]).

In the other direction, Bergman [3 §6] considered the supremum rmr_m of the ratios dim⁡FA/dim⁡FV\dim_{\mathbb{F}} A/\dim_{\mathbb{F}} V over all commutative subalgebras A⊂End⁡F(V)A \subset\operatorname{End}_{\mathbb{F}}(V) generated by mm elements. Explicitly, define

cm(F)=sup⁡n≥1, A1,…,Am∈Mn(F)AiAj=AjAidim⁡FF[A1,…,Am]n,rm=sup⁡Fcm(F).c_m(\mathbb{F})=\sup_{\substack{n\geq1,\ A_1,\ldots,A_m\in M_n(\mathbb{F})\\ A_iA_j=A_jA_i}}\frac{\dim_{\mathbb{F}}\mathbb{F}[A_1,\ldots,A_m]}{n},\qquad r_m=\sup_{\mathbb{F}}c_m(\mathbb{F}).

He asked the following.

Bergman Question. Let m≥3m \geq3. Is this rmr_m finite?

All the bounds above grow faster than linearly in nn, so they do not address Bergman’s question.

Linear Bound

Our main result answers the Bergman question affirmatively.

Theorem 1.1. Let F\mathbb{F} be any field and let m,n≥1m,n\geq1. For any A1,…,Am∈Mn(F)A_1,\ldots,A_m\in M_n(\mathbb{F}) commute pairwise, then

(i) dim⁡FF[A1,…,Am]≤(2m−1)(n−1)+1\dim_{\mathbb{F}}\mathbb{F}[A_1,\ldots,A_m]\leq(2^m-1)(n-1)+1.

(ii) lim⁡m→∞cm(F)1/m=lim⁡m→∞rm1/m=2\lim_{m\to\infty}c_m(\mathbb{F})^{1/m}=\lim_{m\to\infty}r_m^{1/m}=2.

(iii) if char⁡(F)=0\operatorname{char}(\mathbb{F})=0 (resp. a prime pp), then cm(F)=cm(Q)c_m(\mathbb{F})=c_m(\mathbb{Q}) (resp. cm(Fp)c_m(\mathbb{F}_p)).

Sketch of Proof of (i) and (ii)

(i) For the upper bound, the first ingredient is the maximal-rank argument. Put V=FnV=\mathbb{F}^n and R=F[A1,…,Am]R=\mathbb{F}[A_1,\ldots,A_m], and choose vv so that W=RvW=Rv has maximal dimension, we have Ann⁡R(v)\operatorname{Ann}_R(v) sends VV into WW and annihilates WW. Write P=F[x1,…,xm]P=\mathbb{F}[x_1,\ldots,x_m], and let QQ, HH, JJ be the annihilators in PP of VV, WW, V/WV/W, respectively. Then

H⊆J,HJ⊆Q,dim⁡F(P/H)=dim⁡FW.H\subseteq J,\qquad HJ\subseteq Q,\qquad\dim_{\mathbb{F}}(P/H)=\dim_{\mathbb{F}}W.

The second ingredient is the colength inequality

(dim⁡F(P/KL))1/m≤(dim⁡F(P/K))1/m+(dim⁡F(P/L))1/m.\left(\dim_{\mathbb{F}}(P/KL)\right)^{1/m}\leq\left(\dim_{\mathbb{F}}(P/K)\right)^{1/m}+\left(\dim_{\mathbb{F}}(P/L)\right)^{1/m}.

If r=dim⁡FWr=\dim_{\mathbb{F}}W and s=dim⁡F(P/J)s=\dim_{\mathbb{F}}(P/J), these ingredients yield

dim⁡FR≤(r1/m+s1/m)m≤(2m−1)r+s,\dim_{\mathbb{F}}R\leq(r^{1/m}+s^{1/m})^m\leq(2^m-1)r+s,

because s≤rs\leq r. Induction applies to the algebra induced on V/WV/W.

(ii) The exponential growth rate follows from constructing a lower bound. Write m=p+qm=p+q with p,q≥2p,q\geq2. From the coefficients of the product of two polynomials of degrees p−1p-1 and q−1q-1, together with one additional bilinear monomial, we obtain a graded algebra AA with mm degree-one generators. Its ideal of relations is generated by a single homogeneous polynomial of total degree

e=(m−2p−1).e = \binom{m - 2}{p - 1}.

Formal shift operators then give a faithful action of A/A>dA/A_{>d} on a space of dimension (d+m−1m−1)\binom{d+m-1}{m-1}. The ratio of the dimensions of the algebra to this space tends to ee as d→∞d \to\infty. Balancing p−1p-1 and q−1q-1 shows that cm(F)≥(m−2⌊m−22⌋)c_m(\mathbb{F}) \ge\binom{m-2}{\lfloor\frac{m-2}{2} \rfloor}, and the central binomial coefficient has exponential rate 2.

Proof of Theorem 1.1(i)

Colength inequality ingredient

Let P=F[x1,…,xm]P = \mathbb{F}[x_1,\ldots,x_m]. For an ideal II in PP, we define

ℓ(I):=dim⁡F(P/I).\ell(I) := \dim_{\mathbb{F}}(P/I).

We say II has finite colength if ℓ(I)\ell(I) is finite. If II is a monomial ideal, i.e., II is generated by some monomials xα=x1α1⋯xmαmx^\alpha= x_1^{\alpha_1}\cdots x_m^{\alpha_m} with an mm-tuple α=(α1,…,αm)∈Nm\alpha= (\alpha_1,\ldots,\alpha_m) \in\mathbb{N}^m, we define

Γ(I)=⋃xα∈I(α+R≥0m).\Gamma(I) = \bigcup_{x^\alpha\in I}(\alpha+\mathbb{R}_{\geq0}^m).

Lemma 2.1. An ideal II of PP has finite colength if and only if there exist ai∈N+a_i \in\mathbb{N}_+ with xiai∈Ix_i^{a_i} \in I for each i=1,…,mi=1,\ldots,m.

Proof. If xiai∉Ix_i^{a_i} \notin I for all ai≥1a_i \geq1. Then the images of 1,xi,xi2,…1,x_i,x_i^2,\ldots in P/IP/I are linearly independent, so ℓ(I)=∞\ell(I)=\infty. Conversely, assume xiai∈Ix_i^{a_i} \in I for each ii, and let xα∉Ix^\alpha\notin I. Then αi≤ai−1\alpha_i \leq a_i-1 for each ii since xiai∣xαx_i^{a_i}\mid x^\alpha implies xα∈Ix^\alpha\in I. So the monomials outside II are contained in {x1α1⋯xmαm:0≤αi≤ai−1}\{x_1^{\alpha_1}\cdots x_m^{\alpha_m}:0\leq\alpha_i\leq a_i-1\}, a set of at most ∏iai\prod_i a_i elements. Therefore, ℓ(I)≤∏iai<∞\ell(I)\leq\prod_i a_i<\infty. □\square

Lemma 2.2. Let K,LK,L be two monomial ideals of PP that have finite colength. Then

(i) Γ(K)\Gamma(K) is closed.

(ii) ℓ(K)=Vol⁡(R≥0m∖Γ(K))\ell(K)=\operatorname{Vol}(\mathbb{R}_{\geq0}^m\setminus\Gamma(K)).

(iii) KLKL has finite colength.

(iv) Γ(KL)=Γ(K)+Γ(L)\Gamma(KL)=\Gamma(K)+\Gamma(L).

Proof. (i) By Lemma 2.1, there exist some ai∈N+a_i \in\mathbb{N}_+ such that xiai∈Kx_i^{a_i}\in K. Let ej=(δ1j,…,δmj)e_j=(\delta_{1j},\ldots,\delta_{mj}) and δij\delta_{ij} is the Kronecker delta which is 1 if i=ji=j, 0 otherwise. Then

Γ(K)=(⋃j=1m(ajej+R≥0m))∪(⋃xα∈Kαi≤ai−1 for all i(α+R≥0m))\Gamma(K)=\left(\bigcup_{j=1}^{m}(a_je_j+\mathbb{R}_{\geq0}^m)\right)\cup\left(\bigcup_{\substack{x^\alpha\in K\\\alpha_i\leq a_i-1\text{ for all }i}}(\alpha+\mathbb{R}_{\geq0}^m)\right)

since if αj≥aj\alpha_j\geq a_j, then α≥ajej\alpha\geq a_je_j coordinate-wise, and hence α+R≥0m⊂ajej+R≥0m\alpha+\mathbb{R}_{\geq0}^m\subset a_je_j+\mathbb{R}_{\geq0}^m, otherwise α∈∏i{0,1,…,ai−1}\alpha\in\prod_i\{0,1,\ldots,a_i-1\}. Then Γ(K)\Gamma(K) is a finite union of closed subsets. Therefore, Γ(K)\Gamma(K) is closed.

(ii) For α∈R≥0m\alpha\in\mathbb{R}_{\geq0}^m, we claim α∈Γ(K)\alpha\in\Gamma(K) if and only if x⌊α⌋∈Kx^{\lfloor\alpha\rfloor}\in K where ⌊α⌋=(⌊α1⌋,…,⌊αm⌋)\lfloor\alpha\rfloor=(\lfloor\alpha_1\rfloor,\ldots,\lfloor\alpha_m\rfloor). If α∈Γ(K)\alpha\in\Gamma(K), then α=β+u\alpha=\beta+u with xβ∈Kx^\beta\in K and u∈R≥0mu \in\mathbb{R}_{\geq0}^{m}, so βi≤αi\beta_i \leq\alpha_i for each ii. Since βi\beta_i are integers, βi≤⌊αi⌋\beta_i \leq\lfloor\alpha_i \rfloor, and x⌊α⌋=xβx⌊α⌋−β∈Kx^{\lfloor\alpha\rfloor}=x^\beta x^{\lfloor\alpha\rfloor-\beta}\in K. Conversely, if x⌊α⌋∈Kx^{\lfloor\alpha\rfloor}\in K, then α∈⌊α⌋+R≥0m⊂Γ(K)\alpha\in\lfloor\alpha\rfloor+\mathbb{R}_{\geq0}^{m}\subset\Gamma(K).

It follows that

R≥0m∖Γ(K)=⋃xα∉K(α+[0,1)m),\mathbb{R}_{\geq0}^{m}\setminus\Gamma(K)=\bigcup_{x^\alpha\notin K}\left(\alpha+[0,1)^m\right),

since for any α∈R≥0m\alpha\in\mathbb{R}_{\geq0}^{m}, α∈⌊α⌋+[0,1)m\alpha\in\lfloor\alpha\rfloor+[0,1)^m, this mm-cube lies in the complement if and only if x⌊α⌋∉Kx^{\lfloor\alpha\rfloor}\notin K, and distinct ⌊α⌋+[0,1)m\lfloor\alpha\rfloor+[0,1)^m are disjoint. By Lemma 2.1, there exist finitely many such mm-cubes, each of volume 1, so the right-hand side is measurable and

Vol⁡(R≥0m∖Γ(K))=#({xα:xα∉K})=ℓ(K).\operatorname{Vol}(\mathbb{R}_{\geq0}^{m}\setminus\Gamma(K))=\#\left(\{x^\alpha:x^\alpha\notin K\}\right)=\ell(K).

(iii) By Lemma 2.1, there exist ai,bi∈N+a_i,b_i\in\mathbb{N}_+ such that xiai∈Kx_i^{a_i}\in K and xibi∈Lx_i^{b_i}\in L. Then xiai+bi=xiaixibi∈KLx_i^{a_i+b_i}=x_i^{a_i}x_i^{b_i}\in KL for each ii. By Lemma 2.1, KLKL has finite colength.

(iv) If xα∈Kx^\alpha\in K and xβ∈Lx^\beta\in L, then xα+β=xαxβ∈KLx^{\alpha+\beta}=x^\alpha x^\beta\in KL, so

(α+R≥0m)+(β+R≥0m)=(α+β)+R≥0m⊂Γ(KL).(\alpha+\mathbb{R}_{\geq0}^{m})+(\beta+\mathbb{R}_{\geq0}^{m})=(\alpha+\beta)+\mathbb{R}_{\geq0}^{m}\subset\Gamma(KL).

It follows that Γ(K)+Γ(L)⊂Γ(KL)\Gamma(K)+\Gamma(L)\subset\Gamma(KL). Conversely, let γ∈Γ(KL)\gamma\in\Gamma(KL). Then γ=α+u\gamma=\alpha+u for xα∈KLx^\alpha\in KL and u∈R≥0mu\in\mathbb{R}_{\geq0}^{m}. The ideal KLKL is spanned by products xα′xβ′x^{\alpha'}x^{\beta'} with xα′x^{\alpha'} a monomial of KK and xβ′x^{\beta'} a monomial of LL as an FF-vector space. Since the spanning set consists of monomials and monomials are linearly independent, the monomial xαx^\alpha is equal to one of them. Let xα=xα′xβ′x^\alpha=x^{\alpha'}x^{\beta'} with xα′∈K,xβ′∈Lx^{\alpha'}\in K,x^{\beta'}\in L. Therefore, γ=α′+(β′+u)∈Γ(K)+Γ(L)\gamma=\alpha'+(\beta'+u)\in\Gamma(K)+\Gamma(L), and hence Γ(KL)⊂Γ(K)+Γ(L)\Gamma(KL)\subset\Gamma(K)+\Gamma(L). So (iv) holds. □\square

A monomial order is a well order ≤\leq on the set of monomials in PP that satisfies mm1≤mm2mm_1\leq mm_2 if m1≤m2m_1\leq m_2 for any monomials m,m1,m2∈Pm,m_1,m_2\in P. Fix a monomial ordering ≤\leq on PP. For a nonzero polynomial f∈Pf\in P, we denote by LM⁡(f)\operatorname{LM}(f) its greatest term with respect to ≤\leq, with its coefficient removed.

Lemma 2.3. Let f,g∈Pf,g\in P be nonzero. Then LM⁡(fg)=LM⁡(f)LM⁡(g)\operatorname{LM}(fg)=\operatorname{LM}(f)\operatorname{LM}(g).

Proof. Every monomial m1m2m_1m_2 occurring in fgfg satisfies m1m2≤LM⁡(f)m2≤LM⁡(f)LM⁡(g)m_1m_2\leq\operatorname{LM}(f)m_2\leq\operatorname{LM}(f)\operatorname{LM}(g) by the definition of monomial order. If there exists a monomial n1n2n_1n_2 occurring in fgfg such that n1n2=LM⁡(f)LM⁡(g)n_1n_2=\operatorname{LM}(f)\operatorname{LM}(g) with n1<LM⁡(f)n_1<\operatorname{LM}(f), then n1n2<LM⁡(f)n2≤LM⁡(f)LM⁡(g)=n1n2n_1n_2<\operatorname{LM}(f)n_2\leq\operatorname{LM}(f)\operatorname{LM}(g)=n_1n_2, which is a contradiction. □\square

Proposition 2.4. Let K,LK,L be two ideals of PP that have finite colength. Then

(1)

ℓ(KL)1m≤ℓ(K)1m+ℓ(L)1m.\ell(KL)^{\frac{1}{m}}\leq\ell(K)^{\frac{1}{m}}+\ell(L)^{\frac{1}{m}}.

Proof. If one of the ideals K,L,KLK,L,KL is equal to PP, the inequality is immediate. If one of them is 0, its colength is infinite, which contradicts the assumption that K,L,KLK,L,KL have finite colength. Assume that K,L,KLK,L,KL are nonzero and proper.

First assume that KK and LL are monomial ideals. Let Θ=Sm−1∩R≥0m\Theta=S^{m-1}\cap\mathbb{R}_{\geq0}^{m} where Sm−1S^{m-1} is the unit sphere in Rm\mathbb{R}^{m}, and let dθd\theta be the surface measure on Θ\Theta. For θ∈Θ\theta\in\Theta, set

ρK(θ)=min⁡{t≥0:tθ∈Γ(K)}.\rho_K(\theta)=\min\{t\geq0:t\theta\in\Gamma(K)\}.

By Lemma 2.2 (i), Γ(K)\Gamma(K) is closed. By Lemma 2.1, KK contains a power of every variable; then each ray in R≥0m\mathbb{R}_{\ge0}^{m} meets Γ(K)\Gamma(K). Therefore, ρK(θ)\rho_K(\theta) is well-defined. The function ρK\rho_K is bounded by mmax⁡{ai:1≤i≤m}\sqrt{m}\max\{a_i:1\le i\le m\}, where xiai∈Kx_i^{a_i}\in K are as in Lemma 2.1, and it is measurable since {θ:ρK(θ)≤t}={θ:tθ∈Γ(K)}\{\theta:\rho_K(\theta)\le t\}=\{\theta:t\theta\in\Gamma(K)\} is closed for t≥0t\ge0. In polar coordinates, by Lemma 2.2 (ii),

ℓ(K)=Vol⁡(R≥0m∖Γ(K))=∫Θ∫0ρK(θ)tm−1 dt dθ=1m∫ΘρK(θ)m dθ.\ell(K)=\operatorname{Vol}(\mathbb{R}_{\ge0}^{m}\setminus\Gamma(K))=\int_{\Theta}\int_{0}^{\rho_K(\theta)}t^{m-1}\,\mathrm{d}t\,\mathrm{d}\theta=\frac{1}{m}\int_{\Theta}\rho_K(\theta)^m\,\mathrm{d}\theta.

Since ρK(θ)θ∈Γ(K)\rho_K(\theta)\theta\in\Gamma(K) and ρL(θ)θ∈Γ(L)\rho_L(\theta)\theta\in\Gamma(L), by Lemma 2.2 (iv), their sum is in Γ(KL)\Gamma(KL). By Lemma 2.2 (iii), KLKL has finite colength. Therefore, ρKL(θ)≤ρK(θ)+ρL(θ)\rho_{KL}(\theta)\le\rho_K(\theta)+\rho_L(\theta) for θ∈Θ\theta\in\Theta. Apply Minkowski’s inequality to Lm(Θ,dθm)L^m(\Theta,\frac{\mathrm{d}\theta}{m}), the inequality (1) holds.

Now let KK and LL be arbitrary ideals. Let in⁡(K)\operatorname{in}(K) be the initial ideal of KK, the monomial ideal generated by LM⁡(f)\operatorname{LM}(f) for any f∈Kf\in K.

The monomials outside in⁡(K)\operatorname{in}(K) form a basis of P/KP/K under the canonical projection. For independence, let rr be a combination of such monomials with r∈Kr\in K. If r≠0r\ne0, then LM⁡(r)\operatorname{LM}(r) is one of them, hence outside in⁡(K)\operatorname{in}(K). But r∈Kr\in K implies LM⁡(r)∈in⁡(K)\operatorname{LM}(r)\in\operatorname{in}(K), which is a contradiction. Therefore, r=0r=0, and the monomials outside in⁡(K)\operatorname{in}(K) are linearly independent. For spanning, let f∈Pf\in P with LM⁡(f)∈in⁡(K)\operatorname{LM}(f)\in\operatorname{in}(K). Then LM⁡(f)=xαLM⁡(h)\operatorname{LM}(f)=x^\alpha\operatorname{LM}(h) for some nonzero h∈Kh\in K, by Lemma 2.3, LM⁡(f)=LM⁡(xαh)\operatorname{LM}(f)=\operatorname{LM}(x^\alpha h). Let g=xαh∈Kg=x^\alpha h\in K, the element f1=f−LC⁡(f)gf_1=f-\operatorname{LC}(f)g, where LC⁡(f)\operatorname{LC}(f) is the coefficient of LM⁡(f)\operatorname{LM}(f) in ff, satisfies f1=0f_1=0 or LM⁡(f1)<LM⁡(f)\operatorname{LM}(f_1)<\operatorname{LM}(f). Iterating, the leading monomial strictly decreases at each step. This terminates since only finitely many monomials are smaller than a given one. The final remainder rr is an FF-linear combination of monomials outside in⁡(K)\operatorname{in}(K), and f−r∈Kf-r\in K. Therefore, f=r+Kf=r+K in P/KP/K. Therefore, ℓ(K)=ℓ(in⁡(K))\ell(K)=\ell(\operatorname{in}(K)).

By Lemma 2.3, in⁡(K)in⁡(L)⊂in⁡(KL)\operatorname{in}(K)\operatorname{in}(L)\subset\operatorname{in}(KL). Therefore,

ℓ(KL)1m=ℓ(in⁡(KL))1m≤ℓ(in⁡(K)in⁡(L))1m≤ℓ(in⁡(K))1m+ℓ(in⁡(L))1m=ℓ(K)1m+ℓ(L)1m.\begin{aligned} \ell(KL)^{\frac{1}{m}}=\ell(\operatorname{in}(KL))^{\frac{1}{m}}\le\ell(\operatorname{in}(K)\operatorname{in}(L))^{\frac{1}{m}} \\ &\le\ell(\operatorname{in}(K))^{\frac{1}{m}}+\ell(\operatorname{in}(L))^{\frac{1}{m}} \\ &=\ell(K)^{\frac{1}{m}}+\ell(L)^{\frac{1}{m}}. \end{aligned}

Maximal-rank argument ingredient

Lemma 2.5 (Flanders–Atkinson lemma, See [10 Lemma 2.2]). Let FF be an infinite field, and let VV be a finite-dimensional vector space over FF. For a commutative unital algebra R⊂End⁡F(V)R\subset\operatorname{End}_F(V), choose v∈Vv\in V such that r=dim⁡F(Rv)r=\dim_F(Rv) is maximal, and let I=Ann⁡R(v)I=\operatorname{Ann}_R(v). Then

(i) IV⊂RvIV\subset Rv,

(ii) I(Rv)=0I(Rv)=0,

(iii) I2=0I^2=0.

Proof. Let b1v,…,brvb_1v,\ldots,b_rv be a basis of RvRv, where b1,…,br∈Rb_1,\ldots,b_r\in R. Fix a∈Ia\in I, w∈Vw\in V, and for any t∈Ft\in F, the vectors

b1(v+tw),…,br(v+tw),a(v+tw)b_1(v+tw),\ldots,b_r(v+tw),a(v+tw)

are contained in R(v+tw)R(v+tw) of dimension at most rr by the choice of vv. Therefore, their exterior product vanishes. Since av=0av=0, this gives the polynomial identity

(b1v+tb1w)∧⋯∧(brv+tbrw)∧taw=0.(b_{1}v+tb_{1}w)\wedge\cdots\wedge(b_{r}v+tb_{r}w)\wedge taw=0.

The field is infinite, so the coefficient of tt is zero, i.e.,

b1v∧⋯∧brv∧aw=0.b_{1}v\wedge\cdots\wedge b_{r}v\wedge aw=0.

Since b1v,…,brvb_{1}v,\ldots,b_{r}v are independent, b1v∧⋯∧brv≠0b_{1}v\wedge\cdots\wedge b_{r}v\ne0, and hence

aw∈Span⁡{b1v,…,brv}=Rv.aw\in\operatorname{Span}\{b_{1}v,\ldots,b_{r}v\}=Rv.

Since the choice of a,wa,w is arbitrary, IV⊂RvIV\subset Rv, so (i) holds. Condition (ii) holds since commutativity gives a(bv)=b(av)=0a(bv)=b(av)=0 for b∈Rb\in R. By (i), we have I2V=0I^{2}V=0. Since RR is an algebra of F\mathbb{F}-linear maps, I2=0I^{2}=0, and hence (iii) holds.

Proof of Theorem 1.1 (i)

Proof of Theorem 1.1(i). Let V=FnV=\mathbb{F}^{n}, and let R=F[A1,…,Am]R=\mathbb{F}[A_{1},\ldots,A_{m}]. It is easy to see that

dim⁡FF[A1,…,Am]=dim⁡F‾F‾[A1,…,Am].\dim_{\mathbb{F}}\mathbb{F}[A_{1},\ldots,A_{m}] =\dim_{\overline{\mathbb{F}}}\overline{\mathbb{F}}[A_{1},\ldots,A_{m}].

Then we can replace F\mathbb{F} by its algebraic closure. Without loss of generality, we assume F\mathbb{F} is algebraically closed. So F\mathbb{F} is infinite, and hence Lemma 2.5 is available.

Fix m∈N+m\in\mathbb{N}_{+}. We prove this theorem by induction on n=dim⁡FVn=\dim_{\mathbb{F}}V. For n=1n=1, the matrices A1,…,AmA_{1},\ldots,A_{m} can be regarded as the elements in F\mathbb{F}. Then F[A1,…,Am]=F\mathbb{F}[A_{1},\ldots,A_{m}]=\mathbb{F} and

dim⁡FF[A1,…,Am]=1=(2m−1)(1−1)+1.\dim_{\mathbb{F}}\mathbb{F}[A_{1},\ldots,A_{m}] =1=(2^{m}-1)(1-1)+1.

Therefore, this theorem holds for n=1n=1. Assuming that this theorem holds for the positive integers that are strictly less than nn, we show that this theorem holds for nn. Choose the vector v∈Vv\in V such that r=dim⁡F(Rv)r=\dim_{\mathbb{F}}(Rv) is maximal, and let W=RvW=Rv. Since RR has an identity, r≥1r\ge1.

If W=VW=V, then the F\mathbb{F}-linear map

φ:R→V\varphi:R\to V
a↦av.a\mapsto av.

is surjective. If b∈Ker⁡(φ)b\in\operatorname{Ker}(\varphi), then bv=0bv=0. Commutativity gives b(cv)=c(bv)=0b(cv)=c(bv)=0 for any c∈Rc\in R, i.e., bV=0bV=0. Since RR is an algebra of F\mathbb{F}-linear maps, b=0b=0. Then Ker⁡(φ)={0}\operatorname{Ker}(\varphi)=\{0\}. It follows that φ\varphi is also injective. Therefore, φ\varphi is an isomorphism between RR and VV. So we have

dim⁡FR=dim⁡FV=n≤(2m−1)(n−1)+1,\dim_{\mathbb{F}}R=\dim_{\mathbb{F}}V=n\le(2^{m}-1)(n-1)+1,

the last inequality holds since 2m−1≥12^{m}-1\ge1 for m∈N+m\in\mathbb{N}_{+}.

Assume dim⁡FV/W>0\dim_{\mathbb{F}}V/W>0. Regard VV as a P=F[x1,…,xm]P=\mathbb{F}[x_{1},\ldots,x_{m}]-module via the action

f(x1,…,xm)⋅v=f(A1,…,Am)v.f(x_{1},\ldots,x_{m})\cdot v=f(A_{1},\ldots,A_{m})v.

Then

R≅P/Ann⁡P(V).(1)R\cong P/\operatorname{Ann}_{P}(V). \tag*{(1)}

Commutativity and W=PvW=Pv give Ann⁡P(W)=Ann⁡P(v)\operatorname{Ann}_{P}(W)=\operatorname{Ann}_{P}(v), so f↦fvf\mapsto fv induces P/Ann⁡P(W)≅WP/\operatorname{Ann}_{P}(W)\cong W, and hence

ℓ(Ann⁡P(W))=dim⁡F(W)=r.(2)\ell(\operatorname{Ann}_{P}(W))=\dim_{F}(W)=r. \tag*{(2)}

The image of Ann⁡P(W)=Ann⁡P(v)\operatorname{Ann}_{P}(W)=\operatorname{Ann}_{P}(v) in R=P/Ann⁡P(V)R=P/\operatorname{Ann}_{P}(V) is Ann⁡R(v)\operatorname{Ann}_{R}(v), by Lemma 2.5(i), Ann⁡P(W)V⊂W\operatorname{Ann}_{P}(W)V\subset W, and hence Ann⁡P(W)⊂Ann⁡P(V/W)\operatorname{Ann}_{P}(W)\subset\operatorname{Ann}_{P}(V/W). It follows that

ℓ(Ann⁡P(V/W))≤ℓ(Ann⁡P(W))=r.(3)\ell(\operatorname{Ann}_{P}(V/W))\leq\ell(\operatorname{Ann}_{P}(W))=r. \tag*{(3)}

Let RV/WR_{V/W} be the image of PP in End⁡F(V/W)\operatorname{End}_{F}(V/W), i.e., the algebra induced by RR on V/WV/W, generated by the identity and by the quotient map induced by A1,…,AmA_{1},\ldots,A_{m}. The action map P→End⁡F(V/W)P\to\operatorname{End}_{F}(V/W) has kernel Ann⁡P(V/W)\operatorname{Ann}_{P}(V/W) and image RV/WR_{V/W}, by the rank-nullity theorem and the inductive hypothesis,

ℓ(Ann⁡P(V/W))=dim⁡F(RV/W)≤(2m−1)(dim⁡F(V/W)−1)+1.(4)\ell(\operatorname{Ann}_{P}(V/W))=\dim_{F}(R_{V/W})\leq(2^{m}-1)(\dim_{F}(V/W)-1)+1. \tag*{(4)}

Since Ann⁡P(V/W)\operatorname{Ann}_{P}(V/W) maps VV into WW and Ann⁡P(W)\operatorname{Ann}_{P}(W) annihilates WW, the product Ann⁡P(W)Ann⁡P(V/W)\operatorname{Ann}_{P}(W)\operatorname{Ann}_{P}(V/W) annihilates VV, and hence

Ann⁡P(W)Ann⁡P(V/W)⊂Ann⁡P(V).(5)\operatorname{Ann}_{P}(W)\operatorname{Ann}_{P}(V/W)\subset\operatorname{Ann}_{P}(V). \tag*{(5)}

Therefore, let s=ℓ(Ann⁡P(V/W))s=\ell(\operatorname{Ann}_{P}(V/W)), by Formulas (2), (3), (4), (5), (6), and Proposition 2.4,

dim⁡F(R)=ℓ(Ann⁡P(V))≤ℓ(Ann⁡P(W)Ann⁡P(V/W))≤(ℓ(Ann⁡P(W))1m+ℓ(Ann⁡P(V/W))1m)m=(r1m+s1m)m=r+s+∑j=1m−1(mj)rm−jmsjm≤r+(2m−1)(dim⁡F(V/W)−1)+1+(2m−2)r=(2m−1)(r+dim⁡F(V/W)−1)+1=(2m−1)(dim⁡F(W)+dim⁡F(V/W)−1)+1=(2m−1)(n−1)+1,\begin{aligned} \dim_{F}(R)&=\ell(\operatorname{Ann}_{P}(V))\\ &\leq\ell(\operatorname{Ann}_{P}(W)\operatorname{Ann}_{P}(V/W))\\ &\leq\left(\ell(\operatorname{Ann}_{P}(W))^{\frac{1}{m}}+\ell(\operatorname{Ann}_{P}(V/W))^{\frac{1}{m}}\right)^{m}\\ &=\left(r^{\frac{1}{m}}+s^{\frac{1}{m}}\right)^{m}\\ &=r+s+\sum_{j=1}^{m-1}\binom{m}{j}r^{\frac{m-j}{m}}s^{\frac{j}{m}}\\ &\leq r+(2^{m}-1)(\dim_{F}(V/W)-1)+1+(2^{m}-2)r\\ &=(2^{m}-1)(r+\dim_{F}(V/W)-1)+1\\ &=(2^{m}-1)(\dim_{F}(W)+\dim_{F}(V/W)-1)+1\\ &=(2^{m}-1)(n-1)+1, \end{aligned}

where the sum is empty when m=1m=1. This completes the induction. □

Proof of Theorem 1.1(ii)

Preliminaries

Fix integers p,q≥2p,q\geq2 and set m=p+qm=p+q. Let

B=⨁r≥0Br,Br=F[x0,…,xp−1,z0,…,zq−1](r,r),\mathcal{B}=\bigoplus_{r\geq0}B_{r},\qquad B_{r}=F[x_{0},\ldots,x_{p-1},z_{0},\ldots,z_{q-1}]_{(r,r)},

where the subscript (r,r)(r,r) denotes the bi-degree, with degree rr with respect to the xx-variables and degree rr with respect to the zz-variables. For d≥0d\geq0, let VdV_{d} be an FF-vector space with the basis

vα,β(α∈Np, β∈Nq, ∣α∣+∣β∣=d).v_{\alpha,\beta}\quad(\alpha\in\mathbb{N}^{p},\ \beta\in\mathbb{N}^{q},\ |\alpha|+|\beta|=d).

Define

Cijvα,β={vα+ei,β−fj,βj>0,0,βj=0,0≤i<p, 0≤j<q.C_{ij}v_{\alpha,\beta} = \begin{cases} v_{\alpha+e_i,\beta-f_j}, & \beta_j > 0,\\ 0, & \beta_j = 0, \end{cases} \qquad0 \le i < p,\ 0 \le j < q.

where ei,fje_i,f_j are the respective coordinate vectors. For ∣μ∣=∣ν∣|\mu|=|\nu|, define an F\mathbb{F}-algebra homomorphism ρd:B→End⁡F(Vd)\rho_d:\mathcal{B}\to\operatorname{End}_{\mathbb{F}}(V_d) given by

ρd(xμzν)vα,β={vα+μ,β−ν,β≥ν coordinatewise,0,otherwise.\rho_d(x^\mu z^\nu)v_{\alpha,\beta} = \begin{cases} v_{\alpha+\mu,\beta-\nu}, & \beta\ge\nu\text{ coordinatewise},\\ 0, & \text{otherwise}. \end{cases}

Lemma 3.1. (i) dim⁡FVd=(d+m−1m−1)\dim_{\mathbb{F}}V_d=\binom{d+m-1}{m-1}.

(ii) {Cij:0≤i<p,0≤j<q}\{C_{ij}:0\le i<p,0\le j<q\} commute pairwise.

(iii) For each d≥0d\ge0,

Ker⁡(ρd)=B>d:=⨁r>dBr.\operatorname{Ker}(\rho_d)=\mathcal{B}_{>d}:=\bigoplus_{r>d}\mathcal{B}_r.

(iv) for any subalgebra A=⨁r≥0Ar\mathcal{A}=\bigoplus_{r\ge0}\mathcal{A}_r of B\mathcal{B}, the map ρd\rho_d induces a faithful action of A/A>d\mathcal{A}/\mathcal{A}_{>d} on VdV_d.

Proof. (i) Counting the number of bases of vα,βv_{\alpha,\beta} is equivalent to counting the number of tuples (α0,⋯ ,αp−1,β1,⋯ ,βq−1)(\alpha_0,\cdots,\alpha_{p-1},\beta_1,\cdots,\beta_{q-1}) satisfying

α0+⋯+αp−1+β0+⋯+βq−1=d.\alpha_0+\cdots+\alpha_{p-1}+\beta_0+\cdots+\beta_{q-1}=d.

The latter can be counted by the stars-and-bars method; there are (d+m−1m−1)\binom{d+m-1}{m-1} in total. Therefore,

dim⁡FVd=(d+m−1m−1).\dim_{\mathbb{F}}V_d=\binom{d+m-1}{m-1}.

(ii) It is easy to see that CijCkl=CklCijC_{ij}C_{kl}=C_{kl}C_{ij}.

(iii) Let xμzν∈Brx^\mu z^\nu\in\mathcal{B}_r with r>dr>d, and let vα,βv_{\alpha,\beta} be a basis element of VdV_d. Then ∣α∣+∣β∣=d|\alpha|+|\beta|=d. Since ∣ν∣=r>d≥∣β∣|\nu|=r>d\ge|\beta|, there exists some 0≤i0≤q−10\le i_0\le q-1 such that βi0<νi0\beta_{i_0}<\nu_{i_0}, and hence xμzνx^\mu z^\nu annihilates vα,βv_{\alpha,\beta}. Therefore, Br\mathcal{B}_r kills VdV_d for r>dr>d, i.e., B>d⊂Ker⁡(ρd)\mathcal{B}_{>d}\subset\operatorname{Ker}(\rho_d).

For the reverse inclusion, let P=∑i=0sPi=∑i=0dPi+∑j=d+1sPj∈Ker⁡(ρd)P=\sum_{i=0}^{s}P_i=\sum_{i=0}^{d}P_i+\sum_{j=d+1}^{s}P_j\in\operatorname{Ker}(\rho_d) where Pi∈BiP_i\in\mathcal{B}_i. Then the part ∑j=d+1sPj\sum_{j=d+1}^{s}P_j kills VdV_d, which implies ∑j=d+1sPj∈Ker⁡(ρd)\sum_{j=d+1}^{s}P_j\in\operatorname{Ker}(\rho_d). For the remainder part P′=∑i=0dPiP'=\sum_{i=0}^{d}P_i, assume, by contradiction, that P′≠0P'\ne0, i.e., P′=∑i=0rPiP'=\sum_{i=0}^{r}P_i, where r≤dr\le d, Pr≠0P_r\ne0. Let

Pr′(x,z)=∑∣ν∣=rPν(x)zνP'_r(x,z)=\sum_{|\nu|=r}P_\nu(x)z^\nu

and choose ν0\nu_0 with Pν0≠0P_{\nu_0}\ne0. For any γ∈Np\gamma\in\mathbb{N}^p with ∣γ∣=d−r|\gamma|=d-r, and the basis element vγ,ν0v_{\gamma,\nu_0}. Then

ρd(P′)vγ,ν0=∑∣μ∣=rcμvγ+μ,0≠0\rho_d(P')v_{\gamma,\nu_0}=\sum_{|\mu|=r}c_\mu v_{\gamma+\mu,0}\ne0

since ∣ν∣=∣ν0∣=r|\nu|=|\nu_0|=r and ν≤ν0\nu\leq\nu_0 implies ν=ν0\nu=\nu_0. For i<ri<r,

ρd(Pi′)vγ,ν0=∑∣ν∣=icνvγ+μ,ν0−ν\rho_d(P'_i)v_{\gamma,\nu_0}=\sum_{|\nu|=i}c_\nu v_{\gamma+\mu,\nu_0-\nu}

where ν0−ν≠0\nu_0-\nu\ne0. The terms of index <r<r contain only basis elements whose β\beta-part is nonzero. However, the terms of index rr contain basis elements with β\beta-part 0. Then ρd(P)vγ,ν0=ρd(P′)vγ,ν0≠0\rho_d(P)v_{\gamma,\nu_0}=\rho_d(P')v_{\gamma,\nu_0}\ne0, which contradicts the fact that P∈Ker⁡(ρd)P\in\operatorname{Ker}(\rho_d). It follows that P′=0P'=0, i.e., P=∑j=d+1sPj∈B>dP=\sum_{j=d+1}^{s}P_j\in\mathcal{B}_{>d}, and hence Ker⁡(ρd)⊂B>d\operatorname{Ker}(\rho_d)\subset\mathcal{B}_{>d}.

(iv) By (iii), A∩Ker⁡(ρd)=A>d\mathcal{A}\cap\operatorname{Ker}(\rho_d)=\mathcal{A}_{>d}, so Ker⁡(ρd∣A)=A>d\operatorname{Ker}(\rho_d|_{\mathcal{A}})=\mathcal{A}_{>d}, and hence ρd\rho_d induces a faithful action of A/A>d\mathcal{A}/\mathcal{A}_{>d} on VdV_d.

A Subalgebra of B\mathcal{B}

Lemma 3.2. Let m=p+qm=p+q, s=p+q−2s=p+q-2, and let

hl=∑0≤i<p0≤j<qi+j=lxizj(0≤l≤s),u=xp−1zq−2.h_l=\sum_{\substack{0\leq i<p\\0\leq j<q\\i+j=l}}x_i z_j\quad(0\leq l\leq s),\qquad u=x_{p-1}z_{q-2}.

and

A=F[h0,…,hs,u].\mathcal{A}=\mathbb{F}[h_0,\ldots,h_s,u].

Then

(i) the elements h0,…,hsh_0,\ldots,h_s are algebraically independent.

(ii) there is a monic polynomial f∈F[H0,…,Hs,U]f\in\mathbb{F}[H_0,\ldots,H_s,U] of degree e=(m−2p−1)e=\binom{m-2}{p-1} in UU and homogeneous of total degree ee such that

A≅F[H0,…,Hs,U]/(f).\mathcal{A}\cong\mathbb{F}[H_0,\ldots,H_s,U]/(f).

under Hi↦hiH_i\mapsto h_i, U↦uU\mapsto u. Consequently,

A=⨁i=0e−1F[h0,…,hs]ui.\mathcal{A}=\bigoplus_{i=0}^{e-1}\mathbb{F}[h_0,\ldots,h_s]u^i.

(iii) Let Rd=ρd(A)=F[ρd(h0),…,ρd(hs),ρd(u)]R_d=\rho_d(\mathcal{A})=\mathbb{F}[\rho_d(h_0),\ldots,\rho_d(h_s),\rho_d(u)]. Then

Rd≅A/A>d,R_d\cong\mathcal{A}/\mathcal{A}_{>d},

and

dim⁡FRd=∑j=0e−1(d−j+m−1m−1)=(d+mm)−(d+m−em).\dim_{\mathbb{F}}R_d=\sum_{j=0}^{e-1}\binom{d-j+m-1}{m-1}=\binom{d+m}{m}-\binom{d+m-e}{m}.

Proof. (i) Let a,γ1,…,γsa,\gamma_1,\ldots,\gamma_s be algebraically independent indeterminates over F\mathbb{F} with a=zq−1xp−1a=z_{q-1}x_{p-1}, and put

f(t)=a∏i=1p−1(t−γi)=∑i=0p−1zq−1xiti,g(t)=∏j=ps(t−γj)=tq−1+∑j=0q−2zjtj.f(t)=a\prod_{i=1}^{p-1}(t-\gamma_i)=\sum_{i=0}^{p-1}z_{q-1}x_i t^i,\qquad g(t)=\prod_{j=p}^{s}(t-\gamma_j)=t^{q-1}+\sum_{j=0}^{q-2}z_jt^j.

Since

f(t)g(t)=a∏i=1s(t−γi),(6)f(t)g(t)=a\prod_{i=1}^{s}(t-\gamma_i), \tag*{(6)}

where deg⁡f(t)=p−1\deg f(t)=p-1 and deg⁡g(t)=q−1\deg g(t)=q-1, comparing the coefficient of tlt^l gives

∑i+j=lxizj=(−1)s−laεs−l(γ1,…,γs)\sum_{i+j=l} x_i z_j = (-1)^{s-l}a\varepsilon_{s-l}(\gamma_1,\ldots,\gamma_s)

for 0≤l≤s0\le l\le s, where εs−l\varepsilon_{s-l} is the (s−l)(s-l)-th elementary symmetric polynomial. For simplicity, denote εi(γ1,…,γs)\varepsilon_i(\gamma_1,\ldots,\gamma_s) as εi\varepsilon_i. By Artin’s Theorem (see [21 Corollary 3.11]), the field extension

F(γ1,…,γs)/F(γ1,…,γs)Sn\mathbb{F}(\gamma_1,\ldots,\gamma_s)/\mathbb{F}(\gamma_1,\ldots,\gamma_s)^{S_n}

is finite, so they have the same transcendence degree. Since the transcendence degree of F(γ1,…,γs)\mathbb{F}(\gamma_1,\ldots,\gamma_s) is ss, by the fundamental theorem of symmetric polynomials, the transcendence degree of

F(γ1,…,γs)Sn≅F(ε1,…,εs)\mathbb{F}(\gamma_1,\ldots,\gamma_s)^{S_n}\cong\mathbb{F}(\varepsilon_1,\ldots,\varepsilon_s)

is also ss. Therefore, ε1,…,εs\varepsilon_1,\ldots,\varepsilon_s are algebraically independent. If a,ε1,…,εsa,\varepsilon_1,\ldots,\varepsilon_s are algebraically dependent over F\mathbb{F}, there exists a nonzero polynomial P∈F[y0,…,ys]P\in\mathbb{F}[y_0,\ldots,y_s] such that

P(a,ε1,…,εs)=∑i=0kQi(ε1,…,εs)ai=0P(a,\varepsilon_1,\ldots,\varepsilon_s)=\sum_{i=0}^{k}Q_i(\varepsilon_1,\ldots,\varepsilon_s)a^i=0

where Qi∈F[y1,…,ys]Q_i\in\mathbb{F}[y_1,\ldots,y_s]. Since aa is an indeterminate over F\mathbb{F}, this equality implies

Q1(ε1,…,εs)=⋯=Qs(ε1,…,εs)=0,Q_1(\varepsilon_1,\ldots,\varepsilon_s)=\cdots=Q_s(\varepsilon_1,\ldots,\varepsilon_s)=0,

Since ε1,…,εs\varepsilon_1,\ldots,\varepsilon_s are algebraically independent, these polynomials Q1,…,QsQ_1,\ldots,Q_s are zero polynomial in F[y1,…,ys]\mathbb{F}[y_1,\ldots,y_s]. It follows that P=0P=0 in F[y0,…,ys]\mathbb{F}[y_0,\ldots,y_s], which contradicts that PP is nonzero. Therefore, a,ε1,…,εsa,\varepsilon_1,\ldots,\varepsilon_s are algebraically independent. By the definition of h0,…,hsh_0,\ldots,h_s, we have

hs=a,hs−j=(−1)jaεj(γ1,…,γs).h_s=a,\qquad h_{s-j}=(-1)^j a\varepsilon_j(\gamma_1,\ldots,\gamma_s).

Therefore, h0,…,hsh_0,\ldots,h_s are algebraically independent.

(ii) Consider the polynomial

F∗(U)=∏I⊂{1,…,s}∣I∣=q−1(U+a∑i∈Iγi)=Ue+∑i=1e−1aiPi(γ1,…,γs)Ue−i.(7)F_*(U)=\prod_{\substack{I\subset\{1,\ldots,s\}\\ |I|=q-1}}\left(U+a\sum_{i\in I}\gamma_i\right)=U^e+\sum_{i=1}^{e-1}a^iP_i(\gamma_1,\ldots,\gamma_s)U^{e-i}. \tag*{(7)}

where

Pi(γ1,…,γs)=∑I1≠⋯≠Ii⊂{1,…,s}∏t=1i(∑r∈Itγr).P_i(\gamma_1,\ldots,\gamma_s)=\sum_{\substack{I_1\ne\cdots\ne I_i\subset\{1,\ldots,s\}}}\prod_{t=1}^{i}\left(\sum_{r\in I_t}\gamma_r\right).

Since PiP_i is symmetric and homogeneous of degree ii, by the fundamental theorem of symmetric polynomials, Pi(γ1,…,γs)P_i(\gamma_1,\ldots,\gamma_s) is a linear combination of some ∏j=1sεjbj\prod_{j=1}^{s}\varepsilon_j^{b_j} for some bj≥0b_j\ge0 satisfying i=∑j=1sjbji=\sum_{j=1}^{s}jb_j. By Formula (3.2), we have

ai∏j=1sεjbj=ai−∑j=1sbj∏j=1s(aεj)bj=ai−∑j=1sbj∏j=1s((−1)jhs−j)bj=ai−∑j=1sbj(−1)∑j=1sjbj∏j=1shs−jbj=(−1)ihsi−∑j=1sbj∏j=1shs−jbj.\begin{aligned} a^i\prod_{j=1}^{s}\varepsilon_j^{b_j} = a^{i-\sum_{j=1}^{s}b_j}\prod_{j=1}^{s}(a\varepsilon_j)^{b_j} \\ &= a^{i-\sum_{j=1}^{s}b_j}\prod_{j=1}^{s}\left((-1)^j h_{s-j}\right)^{b_j} \\ &= a^{i-\sum_{j=1}^{s}b_j}(-1)^{\sum_{j=1}^{s}jb_j}\prod_{j=1}^{s}h_{s-j}^{b_j} \\ &= (-1)^i h_s^{i-\sum_{j=1}^{s}b_j}\prod_{j=1}^{s}h_{s-j}^{b_j}. \end{aligned}

The exponent of hsh_s is nonnegative since i=∑j=1sjbj≥∑j=1sbji=\sum_{j=1}^{s}jb_j\geq\sum_{j=1}^{s}b_j. Therefore, F∗(U)F_\ast(U) is the image under the map Hi↦hiH_i\mapsto h_i of a polynomial

F(H0,…,Hs,U)=Ue+∑i=1e−1P~i(H0,…,Hs)Ue−i∈F[H0,…,Hs,U].F(H_0,\ldots,H_s,U)=U^e+\sum_{i=1}^{e-1}\widetilde{P}_i(H_0,\ldots,H_s)U^{e-i}\in\mathbb{F}[H_0,\ldots,H_s,U].

with P~i\widetilde{P}_i homogeneous of degree ii. Consider the factor U+a∑i∈{p,…,s}γi=U−uU+a\sum_{i\in\{p,\ldots,s\}}\gamma_i=U-u in Formula (7), it follows that F∗(u)=0F_\ast(u)=0. Therefore, F(h0,…,hs,u)=0F(h_0,\ldots,h_s,u)=0.

Regard F(h0,…,hs)\mathbb{F}(h_0,\ldots,h_s) as a subfield of F(a,γ1,…,γs)\mathbb{F}(a,\gamma_1,\ldots,\gamma_s). By Formula (3.2), the field F(h0,…,hs)\mathbb{F}(h_0,\ldots,h_s) is generated over F\mathbb{F} by aa and the aεia\varepsilon_i, and hence also by aa and the εi\varepsilon_i since εi=(−1)ihs−ihs\varepsilon_i=(-1)^i\frac{h_{s-i}}{h_s}. Every permutation of γ1,…,γs\gamma_1,\ldots,\gamma_s fixes aa and each εi\varepsilon_i, so it fixes KK, and it sends uu through the ee distinct elements

−a∑i∈Iγi-a\sum_{i\in I}\gamma_i

for some ∣I∣=q−1|I|=q-1. Therefore, if G(U)∈F(h0,…,hs)[U]G(U)\in\mathbb{F}(h_0,\ldots,h_s)[U] is nonzero and G(u)=0G(u)=0, applying these permutations shows that GG has at least ee distinct roots, so deg⁡UG≥e\deg_U G\geq e. Since F∗F_\ast has degree ee and is monic, it is the minimal polynomial of uu over F(h0,…,hs)\mathbb{F}(h_0,\ldots,h_s).

Let G(H0,…,Hs,U)∈F[H0,…,Hs,U]G(H_0,\ldots,H_s,U)\in\mathbb{F}[H_0,\ldots,H_s,U] satisfy G(h0,…,hs,u)=0G(h_0,\ldots,h_s,u)=0. Dividing by the monic polynomial FF in the variable UU, we have

G=QF+R,G=QF+R,

where Q,R∈F[H0,…,Hs][U]Q,R\in\mathbb{F}[H_0,\ldots,H_s][U] and deg⁡UR<e\deg_U R<e. Then R(h0,…,hs,u)=0R(h_0,\ldots,h_s,u)=0.

Assume that R=∑i=0kQi(H0,…,Hs)UiR=\sum_{i=0}^{k}Q_i(H_0,\ldots,H_s)U^i with Qi∈F[H0,…,Hs]Q_i\in\mathbb{F}[H_0,\ldots,H_s]. If R≠0R\neq0, then there exists some Qi0≠0Q_{i_0}\neq0. Substitute Hj↦hjH_j\mapsto h_j, then R(h0,…,hs,U)∈F(h0,…,hs)[U]R(h_0,\ldots,h_s,U)\in\mathbb{F}(h_0,\ldots,h_s)[U] of degree <e<e. By the minimality of the degree of FF, we have each Qi(h0,…,hs)=0Q_i(h_0,\ldots,h_s)=0. It follows that Qi0(h0,…,hs)=0Q_{i_0}(h_0,\ldots,h_s)=0, which contradicts the fact h0,…,hsh_0,\ldots,h_s are algebraically independent. Therefore, G=QFG=QF for some Q∈F[H0,…,Hs,U]Q\in\mathbb{F}[H_0,\ldots,H_s,U].

(iii) By Lemma 3.1, we have Ker⁡(ρd∣A)=A>d\operatorname{Ker}(\rho_d|_A)=A_{>d}. Therefore,

Rd≅A/Ker⁡(ρd∣A)=A/A>d,(8)R_d\cong A/\operatorname{Ker}(\rho_d|_A)=A/A_{>d}, \tag*{(8)}

and Lemma 3.1 show that RdR_d acts faithfully on VdV_d.

Let F[h0,⋯ ,hs]≤d−j\mathbb{F}[h_0,\cdots,h_s]_{\le d-j} be the F\mathbb{F}-subspace of polynomials of degree ≤d−j\le d-j. A monomial of total degree at most d−jd-j in s+1=m−1s+1=m-1 variables is determined by its exponent vector α=(α0,⋯ ,αs)\alpha=(\alpha_0,\cdots,\alpha_s) such that

α0+⋯+αs≤d−j.(9)\alpha_0+\cdots+\alpha_s\le d-j. \tag*{(9)}

Add one variable αs+1\alpha_{s+1} such that

α0+⋯+αs+αs+1=d−j,(10)\alpha_0+\cdots+\alpha_s+\alpha_{s+1}=d-j, \tag*{(10)}

then counting the number of tuples (α0,⋯ ,αs+1)(\alpha_0,\cdots,\alpha_{s+1}) satisfying Formula (9) is equivalent to counting the number of tuples (α0,⋯ ,αs+1)(\alpha_0,\cdots,\alpha_{s+1}) satisfying Formula (10). The latter can be counted by the stars-and-bars method; there are (d−j+m−1m−1)\binom{d-j+m-1}{m-1} in total. Therefore,

dim⁡FF[h0,⋯ ,hs]≤d−j=(d−j+m−1m−1).(11)\dim_{\mathbb{F}}\mathbb{F}[h_0,\cdots,h_s]_{\le d-j}=\binom{d-j+m-1}{m-1}. \tag*{(11)}

By (ii) and u∈B1u\in\mathcal{B}_1

A/A>d=⨁j=0e−1F[h0,⋯ ,hs]≤d−juj,A/A_{>d}=\bigoplus_{j=0}^{e-1}\mathbb{F}[h_0,\cdots,h_s]_{\le d-j}u^j,

by Formulas (8), (11), we have

dim⁡FRd=dim⁡FA/A>d=∑j=0e−1dim⁡FF[h0,⋯ ,hs]≤d−j=∑j=0e−1(d−j+m−1m−1).(12)\begin{aligned} \dim_{\mathbb{F}}R_d&=\dim_{\mathbb{F}}A/A_{>d}=\sum_{j=0}^{e-1}\dim_{\mathbb{F}}\mathbb{F}[h_0,\cdots,h_s]_{\le d-j}\\ &=\sum_{j=0}^{e-1}\binom{d-j+m-1}{m-1}. \tag*{(12)} \end{aligned}

Consider the polynomial

∑j=0e−1(1+t)d−j+m−1=(1+t)d+m−e((1+t)0+⋯+(1+t)e−1)=(1+t)d+m−e(1+t)e−1t=1t(1+t)d+m−1t(1+t)d+m−e.\begin{aligned} \sum_{j=0}^{e-1}(1+t)^{d-j+m-1} &=(1+t)^{d+m-e}\bigl((1+t)^0+\cdots+(1+t)^{e-1}\bigr)\\ &=(1+t)^{d+m-e}\frac{(1+t)^e-1}{t}\\ &=\frac{1}{t}(1+t)^{d+m}-\frac{1}{t}(1+t)^{d+m-e}. \end{aligned}

By Formula (12), dim⁡FRd\dim_{\mathbb{F}}R_d is the coefficient of the term tm−1t^{m-1} on the left-hand side, and on the right-hand side, tm−1t^{m-1} has coefficient (d+mm)−(d+m−em)\binom{d+m}{m}-\binom{d+m-e}{m}. Therefore,

dim⁡FRd=(d+mm)−(d+m−em).\dim_{\mathbb{F}}R_d=\binom{d+m}{m}-\binom{d+m-e}{m}.

Proof of Theorem 1.1(ii)

Proof of Theorem 1.1(ii). Fix p,qp,q. By Lemma 3.2, there exists an algebra Rd=F[ρd(h0),…,ρd(hs),ρd(u)]R_d=\mathbb{F}[\rho_d(h_0),\ldots,\rho_d(h_s),\rho_d(u)] such that

dim⁡FRd=(d+mm)−(d+m−em),(13)\dim_{\mathbb{F}} R_d=\binom{d+m}{m}-\binom{d+m-e}{m}, \tag*{(13)}

where e=(m−2p−1)e=\binom{m-2}{p-1}. Then, by Lemma 3.1 and Formula (13),

cm(F)≥dim⁡FRddim⁡FVd=(d+mm)−(d+m−em)(d+m−1m−1).c_m(\mathbb{F})\geq\frac{\dim_{\mathbb{F}}R_d}{\dim_{\mathbb{F}}V_d}=\frac{\binom{d+m}{m}-\binom{d+m-e}{m}}{\binom{d+m-1}{m-1}}.

Letting d→∞d\to\infty, we have

cm(F)≥lim⁡d→∞(d+mm)−(d+m−em)(d+m−1m−1)=e.c_m(\mathbb{F})\geq\lim_{d\to\infty}\frac{\binom{d+m}{m}-\binom{d+m-e}{m}}{\binom{d+m-1}{m-1}}=e.

Let p=⌊m−22⌋+1p=\left\lfloor\frac{m-2}{2}\right\rfloor+1 and q=m−pq=m-p. Then

cm(F)≥e=(m−2p−1)=(m−2⌊m−22⌋).(14)c_m(\mathbb{F})\geq e=\binom{m-2}{p-1}=\binom{m-2}{\left\lfloor\frac{m-2}{2}\right\rfloor}. \tag*{(14)}

For each integer 0≤i≤m−20\leq i\leq m-2, we have

(m−2⌊m−22⌋)≥(m−2i).\binom{m-2}{\left\lfloor\frac{m-2}{2}\right\rfloor}\geq\binom{m-2}{i}.

Therefore,

(m−2⌊m−22⌋)≥1m−1∑i=0m−2(m−2i)=2m−2m−1.(15)\binom{m-2}{\left\lfloor\frac{m-2}{2}\right\rfloor}\geq\frac{1}{m-1}\sum_{i=0}^{m-2}\binom{m-2}{i}=\frac{2^{m-2}}{m-1}. \tag*{(15)}

By Theorem 1.1 (i), Formulas (14) and (15), we have

2m−2m−1m≤cm(F)m≤rm≤sup⁡n(2m−1)(n−1)+1nm≤2m−1m.\sqrt[m]{\frac{2^{m-2}}{m-1}}\leq\sqrt[m]{c_m(\mathbb{F})}\leq r_m\leq\sqrt[m]{\sup_n\frac{(2^m-1)(n-1)+1}{n}}\leq\sqrt[m]{2^m-1}.

Letting m→∞m\to\infty, we have

lim⁡m→∞2m−2m−1m=2andlim⁡m→∞2m−1m=2.\lim_{m\to\infty}\sqrt[m]{\frac{2^{m-2}}{m-1}}=2\quad\text{and}\quad\lim_{m\to\infty}\sqrt[m]{2^m-1}=2.

and hence

lim⁡m→∞cm(F)m=lim⁡m→∞rmm=2.\lim_{m\to\infty}\sqrt[m]{c_m(\mathbb{F})}=\lim_{m\to\infty}\sqrt[m]{r_m}=2.

Proof of Theorem 1.1(iii)

Preliminaries

Lemma 4.1. Let KK be a subfield of a field FF, and let A1,…,Am∈Mn(F)A_1,\ldots,A_m\in M_n(F) commute pairwise. Then there exists a finite extension L/KL/K and pairwise commuting matrices A~1,…,A~m∈Mn(L)\widetilde{A}_1,\ldots,\widetilde{A}_m\in M_n(L) such that

dim⁡LL[A~1,…,A~m]≥dim⁡FF[A1,…,Am].\dim_L L[\widetilde{A}_1,\ldots,\widetilde{A}_m]\geq\dim_F F[A_1,\ldots,A_m].

Proof. Let M1,…,MdM_1,\ldots,M_d be an F\mathbb{F}-basis of F[A1,…,Am]\mathbb{F}[A_1,\ldots,A_m]. Regarding each MiM_i as a column vector of its n2n^2 entries, we obtain an n2×dn^2 \times d matrix over F\mathbb{F} of rank dd, and hence there exists a nonzero d×dd \times d minor Δ\Delta. Let

B=K[all entries of A1,…,Am,{dij},Δ−1]⊂F.B=\mathbb{K}[\text{all entries of }A_1,\ldots,A_m,\{d_{ij}\},\Delta^{-1}]\subset\mathbb{F}.

where {dij}\{d_{ij}\} is the set of coefficients of one linear combination of M1,…,MdM_1,\ldots,M_d under the monomials generated A1,…,AmA_1,\ldots,A_m. Only finitely many entries occur, so BB is a finitely generated K\mathbb{K}-algebra. Let m\mathfrak{m} be a maximal ideal of BB, and let L=B/m\mathbb{L}=B/\mathfrak{m}. By the weak version of Hilbert’s Nullstellensatz (see [1 Corollary 7.10]), L\mathbb{L} is a finite extension of K\mathbb{K}.

Let π:B→L\pi:B\to\mathbb{L} be the canonical quotient map, and for each ii, let A~i\widetilde{A}_i be the matrix whose (s,t)(s,t)-entry is the image under π\pi of the (s,t)(s,t)-entry of AiA_i. Since AiA_i and AjA_j commute, we have AiAj−AjAi=0A_iA_j-A_jA_i=0. Each entry of the right-hand side is a polynomial in the entries of AiA_i and AjA_j, and π\pi is a ring homomorphism. Therefore, every entry of A~iA~j−A~jA~i\widetilde{A}_i\widetilde{A}_j-\widetilde{A}_j\widetilde{A}_i is zero, i.e., A~iA~j−A~jA~i=0\widetilde{A}_i\widetilde{A}_j-\widetilde{A}_j\widetilde{A}_i=0. It follows that A~1,…,A~m\widetilde{A}_1,\ldots,\widetilde{A}_m commute pairwise. Since Δ⋅Δ−1=1\Delta\cdot\Delta^{-1}=1 holds in BB, we have π(Δ)⋅π(Δ−1)=π(Δ⋅Δ−1)=π(1F)=1L\pi(\Delta)\cdot\pi(\Delta^{-1})=\pi(\Delta\cdot\Delta^{-1})=\pi(1_{\mathbb{F}})=1_{\mathbb{L}}. Then π(Δ)\pi(\Delta) is a unit of L\mathbb{L}.

For each ii, let M~i\widetilde{M}_i be the matrix whose (i,j)(i,j)-entry is the image under π\pi of the (i,j)(i,j)-entry of MiM_i. We arrange M~1,…,M~d\widetilde{M}_1,\ldots,\widetilde{M}_d in the same way as M1,…,MdM_1,\ldots,M_d. Then the minor Δ~\widetilde{\Delta} corresponding to the position Δ\Delta is the image of Δ\Delta under π\pi. Since Δ~=π(Δ)\widetilde{\Delta}=\pi(\Delta) is a unit of L\mathbb{L}, i.e., Δ~≠0\widetilde{\Delta}\ne0, and hence Δ~\widetilde{\Delta} is a nonzero d×dd\times d minor. It follows that M~1,…,M~d\widetilde{M}_1,\ldots,\widetilde{M}_d are linearly independent. Therefore, we have

dim⁡LL[A~1,…,A~m]≥d=dim⁡FF[A1,…,Am].\dim_{\mathbb{L}}\mathbb{L}[\widetilde{A}_1,\ldots,\widetilde{A}_m]\ge d=\dim_{\mathbb{F}}\mathbb{F}[A_1,\ldots,A_m].

Lemma 4.2. Let LL be a field, and let A~1,…,A~m∈Mn(L)\widetilde{A}_1,\ldots,\widetilde{A}_m\in M_n(\mathbb{L}) commute pairwise with characteristic polynomials p1,…,pmp_1,\ldots,p_m respectively. Consider the splitting field E\mathbb{E} over LL of p1,…,pmp_1,\ldots,p_m, and let V=EnV=\mathbb{E}^n. Then

(i) E/L\mathbb{E}/L is a finite extension.

(ii) the algebra E[A~1,…,A~m]\mathbb{E}[\widetilde{A}_1,\ldots,\widetilde{A}_m] can be decomposed as a direct sum of local E\mathbb{E}-algebras, i.e.,

E[A~1,…,A~m]=⨁iRi.\mathbb{E}[\widetilde{A}_1,\ldots,\widetilde{A}_m]=\bigoplus_i R_i.

(iii) VV can be decomposed into the corresponding simultaneous generalized eigenspaces

V=⨁jVj.V=\bigoplus_j V_j.

(iv) each RjR_j acts faithfully on VjV_j.

Proof. (i) Let P(x)=∏i=1mpi(x)∈L[x]P(x)=\prod_{i=1}^{m}p_i(x)\in L[x]. Then E\mathbb{E} is the splitting field of PP over LL given by adding finitely many roots of finitely many polynomials. Therefore, E/L\mathbb{E}/L is a finite extension. (ii) The E-algebra E[A~1,…,A~m]\mathrm{E}[\widetilde{A}_1,\ldots,\widetilde{A}_m] has finite dimension over E, hence an Artinian ring. By the structure theorem for Artinian rings, it can be decomposed as a direct sum of local E-algebras

E[A~1,…,A~m]=⨁i=1sRi.\mathrm{E}[\widetilde{A}_1,\ldots,\widetilde{A}_m]=\bigoplus_{i=1}^{s}R_i.

(iii) Let eie_i be the idempotent element of RiR_i. Then

∑i=1sei=1,eiej={ei,i=j,0,i≠j.(16)\sum_{i=1}^{s}e_i=1,\qquad e_i e_j= \begin{cases} e_i, & i=j,\\ 0, & i\ne j. \end{cases} \tag*{(16)}

and Ri=E[A~1,…,A~m]eiR_i=\mathrm{E}[\widetilde{A}_1,\ldots,\widetilde{A}_m]e_i. Let Vi=eiVV_i=e_iV, by Formula (16), we have

V=⨁i=1sVi.V=\bigoplus_{i=1}^{s}V_i.

Since RiR_i is integral over E, E⊂E[A~iej]⊂Rj\mathrm{E}\subset\mathrm{E}[\widetilde{A}_i e_j]\subset R_j, and hence RjR_j is integral over E[A~iej]\mathrm{E}[\widetilde{A}_i e_j]. Since RjR_j is local, by the going-up theorem (see [1 Theorem 5.11]), E[A~iej]\mathrm{E}[\widetilde{A}_i e_j] has a unique maximal ideal. Let mij(x)m_{ij}(x) be the minimal polynomial of A~iej\widetilde{A}_i e_j on VjV_j. Then

E[A~iej]≅E[x]/(mij(x)E[x]).\mathrm{E}[\widetilde{A}_i e_j]\cong\mathrm{E}[x]/(m_{ij}(x)\mathrm{E}[x]).

Since the left-hand side is local, there exists a unique maximal ideal of the right-hand side. This implies that mij(x)=p(x)tm_{ij}(x)=p(x)^t for some irreducible polynomial p(x)p(x). Let λ\lambda be a root of p(x)p(x). Then λ\lambda is a root of mij(x)m_{ij}(x), i.e, λ\lambda is the eigenvalue of A~iej\widetilde{A}_i e_j on VjV_j, and hence λ∈E\lambda\in\mathrm{E}. Therefore, p(x)=(x−λ)kp(x)=(x-\lambda)^k, and hence mij(x)=(x−λ)ktm_{ij}(x)=(x-\lambda)^{kt}. It follows that A~iej\widetilde{A}_i e_j has only one eigenvalue on VjV_j.

(iv) Assume that Aei∈RiAe_i\in R_i acts as zero in ViV_i for some A∈E[A~1,…,A~m]A\in\mathrm{E}[\widetilde{A}_1,\ldots,\widetilde{A}_m]. For any j=1,…,sj=1,\ldots,s, Aeiejv=A(0)=0Ae_i e_jv=A(0)=0 if i≠ji\ne j, Aeiejv=Aeiv=0Ae_i e_jv=Ae_i v=0 if i=ji=j. So AeiAe_i acts as zero on VV. Since E[A~1,…,A~m]\mathrm{E}[\widetilde{A}_1,\ldots,\widetilde{A}_m] acts faithfully on VV, and Aei∈E[A~1,…,A~m]Ae_i\in\mathrm{E}[\widetilde{A}_1,\ldots,\widetilde{A}_m], hence Aei=0Ae_i=0. Therefore, RiR_i acts faithfully on ViV_i. □

Proof of Theorem 1.1 (iii)

Proof of Theorem 1.1 (iii). Let K be the prime subfield of F. For any pairwise commuting matrices A1,…,Am∈Mn(K)A_1,\ldots,A_m\in M_n(\mathrm{K}), it is easy to see

dim⁡KK[A1,…,Am]=dim⁡FF[A1,…,Am]≤cm(F).\dim_{\mathrm{K}}\mathrm{K}[A_1,\ldots,A_m]=\dim_{\mathrm{F}}\mathrm{F}[A_1,\ldots,A_m]\le c_m(\mathrm{F}).

Since the choice of A1,…,AmA_1,\ldots,A_m is arbitrary, taking the supremum gives

cm(K)≤cm(F).c_m(\mathrm{K})\le c_m(\mathrm{F}).

For the reverse inequality, let A1,…,Am∈Mn(F)A_1,\ldots,A_m\in M_n(\mathrm{F}) commute pairwise. By Lemma 4.1, there exists a finite extension L/K\mathrm{L}/\mathrm{K} and pairwise commuting matrices A~1,…,A~m∈Mn(L)\widetilde{A}_1,\ldots,\widetilde{A}_m\in M_n(\mathrm{L}) such that

dim⁡LL[A~1,…,A~m]≥dim⁡FF[A1,…,Am].(17)\dim_{\mathrm{L}}\mathrm{L}[\widetilde{A}_1,\ldots,\widetilde{A}_m]\ge\dim_{\mathrm{F}}\mathrm{F}[A_1,\ldots,A_m]. \tag*{(17)}

Let E\mathbb{E} be the splitting field over L\mathbb{L} of the characteristic polynomials of A~1,…,A~m\widetilde{A}_1,\ldots,\widetilde{A}_m. Since L/K\mathbb{L}/\mathbb{K} is finite, so is E/K\mathbb{E}/\mathbb{K}. By Lemma 4.2, there exist decompositions

E[A~1,…,A~m]=⨁iRi,V=⨁jVj\mathbb{E}[\widetilde{A}_1,\ldots,\widetilde{A}_m]=\bigoplus_i R_i,\qquad V=\bigoplus_j V_j

such that RjR_j acts faithfully on VjV_j. It follows that

dim⁡EE[A~1,…,A~m]n=∑jdim⁡ERjn=∑jdim⁡EVjndim⁡ERjdim⁡EVj.(18)\frac{\dim_{\mathbb{E}}\mathbb{E}[\widetilde{A}_1,\ldots,\widetilde{A}_m]}{n}=\sum_j\frac{\dim_{\mathbb{E}}R_j}{n}=\sum_j\frac{\dim_{\mathbb{E}}V_j}{n}\frac{\dim_{\mathbb{E}}R_j}{\dim_{\mathbb{E}}V_j}. \tag*{(18)}

and RjR_j can be regarded as the subalgebra of End⁡E(Vj)\operatorname{End}_{\mathbb{E}}(V_j) for each jj. Since dim⁡EVjn≤1\frac{\dim_{\mathbb{E}}V_j}{n}\leq1, by Formula (18), there exists at least one component W=Vj0W=V_{j_0} such that

dim⁡ERj0dim⁡EW≥dim⁡EE[A~1,…,A~m]n,(19)\frac{\dim_{\mathbb{E}}R_{j_0}}{\dim_{\mathbb{E}}W}\geq\frac{\dim_{\mathbb{E}}\mathbb{E}[\widetilde{A}_1,\ldots,\widetilde{A}_m]}{n}, \tag*{(19)}

otherwise the right-hand side is strictly less than the left-hand side in Formula (18). Let λi\lambda_i be the eigenvalue of A~iej0∣W\left.\widetilde{A}_i e_{j_0}\right|_W in WW, and let Ni=A~iej0∣W−λiN_i=\left.\widetilde{A}_i e_{j_0}\right|_W-\lambda_i. Then N1,…,NmN_1,\ldots,N_m are pairwise commuting nilpotent operators on WW, and Rj0=E[N1,…,Nm]R_{j_0}=\mathbb{E}[N_1,\ldots,N_m].

Since E/L\mathbb{E}/\mathbb{L} and L/K\mathbb{L}/\mathbb{K} are finite extensions, the field extension E/K\mathbb{E}/\mathbb{K} is finite. The field K\mathbb{K} is the prime subfield of F\mathbb{F}, so K\mathbb{K} is perfect, and hence the finite extension L/K\mathbb{L}/\mathbb{K} is separable. By the primitive element theorem, there exists an element α∈E\alpha\in\mathbb{E} such that E=K(α)\mathbb{E}=\mathbb{K}(\alpha). Regard WW as an K\mathbb{K}-vector space, define

B1=αI+N1,Bi=Ni(2≤i≤m).B_1=\alpha I+N_1,\qquad B_i=N_i\quad(2\leq i\leq m).

These mm operators commute pairwise. The part αI\alpha I is semisimple over K\mathbb{K} since its minimal polynomial is separable, and it commutes with N1N_1. By Jordan-Chevalley decomposition, the uniqueness of the decomposition gives that the semisimple part of B1B_1 is αI\alpha I, and hence there exists a polynomial g(x)∈K[x]g(x)\in\mathbb{K}[x] such that αI=g(B1)\alpha I=g(B_1). Therefore,

αI∈K[B1]andK[B1,…,Bm]=E[N1,…,Nm]⊂End⁡K(W).\alpha I\in\mathbb{K}[B_1]\quad\text{and}\quad\mathbb{K}[B_1,\ldots,B_m]=\mathbb{E}[N_1,\ldots,N_m]\subset\operatorname{End}_{\mathbb{K}}(W).

The action is faithful since restriction of scalars does not identify distinct endomorphisms. Therefore, by Formula (17), we have

cm(K)≥dim⁡KK[B1,⋯ ,Bm]dim⁡KW=[E:K]dim⁡EE[N1,⋯ ,Nm][E:K]dim⁡EW=dim⁡EE[N1,⋯ ,Nm]dim⁡EW=dim⁡ERj0dim⁡EVj0≥dim⁡EE[A~1,⋯ ,A~m]n=dim⁡LL[A~1,⋯ ,A~m]n≥dim⁡FF[A1,⋯ ,Am]n.\begin{aligned} c_m(\mathbb{K}) &\ge\frac{\dim_{\mathbb{K}}\mathbb{K}[B_1,\cdots,B_m]}{\dim_{\mathbb{K}} W} \\ &= \frac{[E:\mathbb{K}]\dim_E E[N_1,\cdots,N_m]}{[E:\mathbb{K}]\dim_E W} \\ &= \frac{\dim_E E[N_1,\cdots,N_m]}{\dim_E W} \\ &= \frac{\dim_E R_{j_0}}{\dim_E V_{j_0}} \\ &\ge\frac{\dim_E E[\widetilde{A}_1,\cdots,\widetilde{A}_m]}{n} \\ &= \frac{\dim_L L[\widetilde{A}_1,\cdots,\widetilde{A}_m]}{n} \\ &\ge\frac{\dim_F F[A_1,\cdots,A_m]}{n}. \end{aligned}

Since the choice of A1,⋯ ,AmA_1,\cdots,A_m is arbitrary, taking the supremum gives cm(F)≤cm(K)c_m(\mathbb{F}) \le c_m(\mathbb{K}). □

Acknowledgement.

Acknowledgement. We would like to thank Prof. Kaiming Zhao for his encouragement and helpful suggestions, as well as for informing us of an interesting and beautiful recent work [7]. We would also like to thank the referee for his/her detailed comments and kindness.

AI disclosure.

AI disclosure. During the preparation of this manuscript, the authors used LLMs to assist in drafting and checking parts of the mathematical arguments, organize references, and improve the clarity and readability of the exposition. All mathematical statements and proofs in the final manuscript were written, reviewed, and verified by the authors, who take full responsibility for the content of this manuscript.

SCHOOL OF MATHEMATICAL SCIENCES, HEBEI NORMAL UNIVERSITY, SHIJIAZHUANG, HEBEI 050024, CHINA. Email address: [email protected]

SCHOOL OF SCIENCE AND ENGINEERING, THE CHINESE UNIVERSITY OF HONG KONG, SHENZHEN, GUANGDONG 518172, CHINA. Email address: [email protected]

SCHOOL OF MATHEMATICAL SCIENCES, HEBEI NORMAL UNIVERSITY, SHIJIAZHUANG, HEBEI 050024, CHINA. Email address: [email protected]

SCHOOL OF SCIENCE AND ENGINEERING, THE CHINESE UNIVERSITY OF HONG KONG, SHENZHEN, GUANGDONG 518172, CHINA. Email address: [email protected]

References

  1. [1]M. F. Atiyah and I. G. Macdonald. Introduction to commutative algebra. Addison-Wesley Publishing Co., Reading, Mass.-London-Don Mills, Ont., 1969.
  2. [2]José Barría and P. R. Halmos. Vector bases for two commuting matrices. Linear and Multilinear Algebra, 27(3):147–157, 1990.DOI
  3. [3]George M. Bergman. Commuting matrices, and modules over artinian local rings. arXiv preprint arXiv:1309.0053, 2013.arxiv.org/abs/1309.0053
  4. [4]I. N. Bernstein. All reductive p-adic groups are tame. Functional Analysis and Its Applications, 8(2):91–93, April 1974.
  5. [5]I. N. Bernstein and A. V. Zelevinsky. Representations of the group GL(n, F), where F is a local non-Archimedean field. Uspehi Mat. Nauk, 31(3(189)):5–70, 1976.
  6. [6]J. Bernstein. Representations of p-adic groups. Lectures at Harvard University, Fall 1992, written by K. E. Ruelumhart, unpublished notes, 1992.
  7. [7]Cunguang Cheng, Gengqiang Liu, Kaiming Zhao, and Yueqiang Zhao. Commutative subalgebras of matrix algebras. Manuscript in preparation, 2026.
  8. [8]Ron Cherny, Tam An Le Quang, and Matthew Satriano. On combinatorial algebras generated by three commuting matrices. arXiv preprint arXiv:2511.22092, 2025.arxiv.org/abs/2511.22092
  9. [9]Ron Cherny, Matthew Satriano, and Yohan Song. On the algebra generated by three commuting matrices: combinatorial cases. Electron. J. Combin., 31(4):Paper No. 4.43, 20, 2024.arxiv.org/abs/2402.16334
  10. [10]Clément de Seguins Pazzis. Local linear dependence seen through duality II. Linear Algebra Appl., 462:133–185, 2014.DOI
  11. [11]Murray Gerstenhaber. On dominance and varieties of commuting matrices. Ann. of Math. (2), 73:324–348, 1961.DOI
  12. [12]Robert M. Guralnick. A note on commuting pairs of matrices. Linear and Multilinear Algebra, 31(1–4):71–75, 1992.DOI
  13. [13]Robert M. Guralnick and B. A. Sethuraman. Commuting pairs and triples of matrices and related varieties. Linear Algebra Appl., 310(1–3):139–148, 2000.DOI
  14. [14]Yongho Han. Commuting triples of matrices. Electron. J. Linear Algebra, 13:274–343, 2005.DOI
  15. [15]J. Holbrook and K. C. O’Meara. Some thoughts on Gerstenhaber’s theorem. Linear Algebra Appl., 466:267–295, 2015.DOI
  16. [16]John Holbrook and Kevin C. O’Meara. A computing strategy and programs to resolve the gerstenhaber problem for commuting triples of matrices. arXiv preprint arXiv:2006.08588, 2020.arxiv.org/abs/2006.08588
  17. [17]John Holbrook and Matjaž Omladič. Approximating commuting operators. Linear Algebra Appl., 327(1–3):131–149, 2001.DOI
  18. [18]N. Jacobson. Schur’s theorems on commutative matrices. Bull. Amer. Math. Soc., 50:431–436, 1944.
  19. [19]Joachim Jelisieiew and Klemen Šivic. Components and singularities of Quot schemes and varieties of commuting matrices. J. Reine Angew. Math., 788:129–187, 2022.arxiv.org/abs/2106.13137
  20. [20]Thomas J. Laffey and Susan Lazarus. Two-generated commutative matrix subalgebras. Linear Algebra Appl., 147:249–273, 1991.DOI
  21. [21]J. S. Milne. Finite Fields and Galois theory. Kea Books, Ann Arbor, MI, [2022] ©2022.
  22. [22]M. Mirzakhani. A simple proof of a theorem of Schur. Amer. Math. Monthly, 105(3):260–262, 1998.DOI
  23. [23]T. S. Motzkin and Olga Taussky. Pairs of matrices with property L. II. Trans. Amer. Math. Soc., 80:387–401, 1955.DOI
  24. [24]Nham V. Ngo and Klemen Šivic. On varieties of commuting nilpotent matrices. Linear Algebra Appl., 452:237–262, 2014.arxiv.org/abs/1308.4438
  25. [25]Matjaž Omladič. A variety of commuting triples. Linear Algebra Appl., 383:233–245, 2004.DOI
  26. [26]Jenna Rajchgot and Matthew Satriano. New classes of examples satisfying the three matrix analog of Gerstenhaber’s theorem. J. Algebra, 516:245–270, 2018.arxiv.org/abs/1711.10109
  27. [27]Jenna Rajchgot, Matthew Satriano, and Wanchun Shen. Some combinatorial cases of the three matrix analog of Gerstenhaber’s theorem (research). In Advances in mathematical sciences, volume 21 of Assoc. Women Math. Ser., pages 181–201. Springer, Cham, [2020] ©2020.DOI
  28. [28]J. Schur. Zur Theorie der vertauschbaren Matrizen. J. Reine Angew. Math., 130:66–76, 1905.
  29. [29]B. A. Sethuraman. The algebra generated by three commuting matrices. Math. Newsl., 21(2):62–67, 2011.
  30. [30]Klemen Šivic. On varieties of commuting triples. Linear Algebra Appl., 428(8–9):2006–2029, 2008.DOI
  31. [31]Klemen Šivic. On varieties of commuting triples III. Linear Algebra Appl., 437(2):393–460, 2012.DOI
  32. [32]Adrian R. Wadsworth. The algebra generated by two commuting matrices. Linear and Multilinear Algebra, 27(3):159–162, 1990.DOI

Paper details

vlnc:W_ugy3fpr7
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC0-1.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

LLMs

View statement
Looking for related papers…
Contents