This paper introduces supersymmetric elements in divided powers algebras, characterizes them through linear equations, and determines generators for specific cases, advancing the understanding of invariants in Lie superalgebras.
Contribution
It defines supersymmetric elements in divided powers algebras and provides explicit generators for particular cases, extending prior work on invariants of Lie superalgebras.
Findings
01
Characterization of supersymmetric elements via linear equations
02
Generators determined for cases n=0, n=1, m≤2, n=2
03
Extension of invariants description in divided powers algebras
Abstract
Description of adjoint invariants of general Linear Lie superalgebras gl(m∣n) by Kantor and Trishin is given in terms of supersymmetric polynomials. Later, generators of invariants of the adjoint action of the general linear supergroup GL(m∣n) and generators of supersymmetric polynomials were determined over fields of positive characteristic. In this paper, we introduce the concept of supersymmetric elements in the divided powers algebra Div[x1,…,xm,y1,…,yn], and give a characterization of supersymmetric elements via a system of linear equations. Then we determine generators of supersymmetric elements for divided powers algebras in the cases when n=0, n=1, and m≤2,n=2.
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Full text
Supersymmetric elements in divided powers algebras
Description of adjoint invariants of general Linear Lie superalgebras gl(m∣n) by Kantor and Trishin is given in terms of supersymmetric polynomials.
Later, generators of invariants of the adjoint action of the general linear supergroup GL(m∣n) and generators of supersymmetric polynomials were determined over fields of positive characteristic.
In this paper, we introduce the concept of supersymmetric elements in the divided powers algebra Div[x1,…,xm,y1,…,yn], and give a characterization of supersymmetric elements via a system of linear equations. Then we determine generators of supersymmetric elements for divided powers algebras in the cases when n=0, n=1, and
m≤2,n=2.
Introduction and notation
We start by recalling a description of invariants of conjugacy classes of matrices which is a classical problem in the invariant theory - see Chapter 1 of [4].
Let K be an infinite field of characteristic zero, V be a K-space of dimension m, E be the space of all K-linear maps of V
(a choice of a basis of V identifies E with m×m-matrices). The general linear group GL(V)
acts on E via conjugation g.a=gag−1 for g∈GL(V) and a∈E (corresponding to a change of basis of V).
Consider the space S(E) of all K-valued polynomial functions on E together with the action of GL(V) given as g.f(a)=f(g−1a) for g∈GL(V), f∈S(E) and a∈E.
By Chevalley’s restriction theorem (see Theorem 1.5.7 of [4]), the ring of invariants S(E)GL(V) is isomorphic to the ring of symmetric polynomials K[x1,…,xm].
Here the variables x1,…,xm are related to the coefficients σi(M) of the characteristic polynomial of m×m-matrices M.
The vertices of the Dynkin diagram Δ=Am corresponding to GL(m) are given by simple positive roots that can be labeled by 1,…,m. The action of the Weyl group W permutes the elements of the set {1,…,m}. If we denote by P the ring of polynomials in variables x1,…,xm, then the ring of symmetric polynomials
PW consists of invariants of P under the action induced by the Weyl group W.
The above classical correspondence was extended to the case of general linear superalgebras GL(m∣n) in characteristic zero by Kantor and Trishin in [2]. They have described the
polynomial invariants of general linear Lie supergroup gl(m∣n) and their connection to the algebra As of supersymmetric polynomials.
The algebra As consists of polynomials f(x∣y)=f(x1,…,xm,y1,…,yn) that are symmetric in variables
x1,…,xm and y1,…,yn separately, such that dTdf(x∣y)∣x1=y1=T=0.
Before the paper of Kantor and Trishin, motivated by a work of Kac and Schneuert, Stembridge proved a conjecture of Schneuert and described generators of As in [5].
An analogous problem for supergroups in the case of positive characteristic has been investigated first by La Scala and Zubkov in [3].
A complete description of generators of polynomial invariants of the adjoint action of the general linear supergroup G=GL(m∣n) and generators of As was obtained by
Grishkov, Marko and Zubkov in [1].
Since the universal enveloping algebra of a Lie superalgebra gl(m∣n) is isomorphic to (noncommuting) polynomials, while the universal enveloping algebra of its Cartan subsuperalgebra h is isomorphic to the ring K[x1,…,xm∣y1,…,yn], the invariants of this ring are the correct objects to consider
when the characteristic of K is zero.
If the characteristic p of the field K is positive, then instead of the universal enveloping algebra of a Lie superalgebra gl(m∣n) one considers the distribution algebra Dist(G) of G=GL(m∣n) and its Frobenius kernels. The basis of the distribution algebra of a maximal torus T of G
consists of products of binomial elements of the type (at)=∏i=1m(aiti), where (ax)=a!x(x−1)…(x−a+1), while
the basis of the distribution algebra of the unipotent subsupergroups U+ and U−, corresponding to unipotent upper and lower triangular matrices of G,
respectively, consists of products of
divided powers e(b)=∏i=1sei(bi), where x(a)=a!xa.
From now on, the characteristic p of the ground field K is positive.
Definition 1**.**
Denote by Div[x,y]=Div[x1,…,xm∣y1,…,yn] the algebra of divided power elements generated by two groups of commuting variables
x1,…,xm and y1,…,yn. The algebra structure is given via z(a)z(b)=(aa+b)z(a+b) for z=x1,…,xm;y1,…,yn.
Thus Div[x,y] is spanned by the products
x1(a1)…xm(am)y1(b1)…yn(bn), where the exponents ai and bj are non-negative integers.
Define the degree d of the element x1(a1)…xm(am)y1(b1)…yn(bn) to be the sum ∑i=1mai+∑j=1nbj.
Denote by Divk[x,y] the space of all elements of Div[x,y] of degree k.
It is clear that Div[x,y] is a graded algebra, where the grading is given by the degrees d.
We repeatedly use the property of divided powers that z(a)z(b)=0 if there is a p-adic carry when a is added to b; z(a)z(b) is a nonzero scalar multiple of z(a+b) otherwise.
Using linearity, it is easy to verify that
[TABLE]
Therefore we can replace the condition dTdf(x∣y)∣x1=y1=T=0 from the definition of the supersymmetric polynomial by the equivalent condition
[TABLE]
Now we are ready to define supersymmetric elements in Div[x,y].
Definition 2**.**
Define the derivations dxid and dyjd of Div[x,y] via
[TABLE]
and
[TABLE]
for every g∈Div[x,y], 1≤i≤m and 1≤j≤n.
If f∈Divk[x,y] is symmetric in variables x1,…,xm
and y1,…,yn separately, and
there is f′∈Divk−2[x,y] such that
[TABLE]
then f is called supersymmetric.
The supersymmetric elements form a graded subalgebra of Div[x,y], which is denoted by S.
Its homogeneous component of degree k is denoted by Sk.
The structure of the paper is as follows.
In Section 1, we determine the symmetric elements in the even divided power algebra Div[x1,…,xm].
In Section 2, we determine the supersymmetric elements in the divided power algebra Div[x1,y1] which serve as a bridge to the general supersymmetric case.
In Section 3, we characterize supersymmetric elements in Div[x,y] using the concepts of marked and unmarked monomials.
In Section 4, we describe generators of supersymmetric elements in Div[x1,…,xm,y1].
In Sections 5 and 6 we describe generators of supersymmetric elements in Div[x1,y1,y2] and Div[x1,x2,y1,y2], respectively.
1. Symmetric elements in the divided power algebra Div[x1,…,xm]
Let Σm be the symmetric group on m elements. It acts on the purely even algebra Div[x]=Div[x1,…,xm] of divided powers by permuting variables x1,…,xm.
In this section, we describe invariants Div[x]Σm, which we call symmetric elements of Div[x].
For an m-tuple μ=(μ1,…,μm) of non-negative integers denote its degree ∣μ∣=∑i=1mμi. Then every homogeneous element f∈Divk[x]
can be expressed as f=∑μ,∣μ∣=kfμx(μ).
Order monomials x(μ) of degree k legicographically.
If f∈Divk[x]Σm, then the largest monomial x(λ) for which fλ=0 satisfies
[TABLE]
The corresponding monomial x(λ) is called the leading term of f.
For a1≥…≥am≥0 denote by fa1,…,am the unique symmetric function
that has the leading term x1(a1)…xm(am) and all other terms of the form
xσ(1)(a1)…xσ(m)(am) for some σ∈Σm.
For example,
if a1>…>am≥0, then
[TABLE]
Proposition 1.1**.**
The algebra Div[x]Σm is generated by the set D of all elements fq,a2,…,am, where q=ps for an arbitrary non-negative integer s and arbitrary integers
q≥a2≥…≥am≥0.
The set D is minimal in the sense that none of the generators can be left out.
Proof.
Denote by B the algebra generated by all elements of D. Clearly, B⊆Div[x]Σm. The equality B=Div[x]Σm follows once we show that for every monomial
x1(a1)…xm(am), where a1≥…≥am≥0,
there is an element of B that has the leading term x1(a1)…xm(am). Indeed, take an arbitrary element g∈Div[x]Σm and assume that its leading term is
x1(a1)…xm(am). There is an element f of B that has the same leading term as g.
Then g−f has a smaller leading term, and by induction on the lexicographic order, we can assume that g−f belongs to B. Therefore g∈B.
If m=1, then the set D consists of elements x1(q), where q=ps and s≥0. It is easy to see that D is a minimal generating set of Div[x1].
Therefore, assume m>1.
If a1<p, then, up to a nonzero constant, x1(a1)…xm(am) equals
[TABLE]
which is the leading term of f1,…,1amf1,…,1,0am−1−am…f1,0,…,0a1−a2∈B.
Here fj1,…,1,0,…0 is the j-th elementary symmetric function
σj=∑1≤i1<…<ij≤mxi1…xij.
Assume now a1≥p and write the p-adic expansions of ai as
[TABLE]
for each i=1,…,m and sufficiently large s.
If a1=…=am, then, up to a nonzero constant,
x1(a1)…xm(am) is the leading term of
[TABLE]
Otherwise, there is an index 0≤t≤s such that
b1s=…=bms, b1,s−1=…=bm,s−1, …and br,t>br+1,t=br+2,t=…=bm,t for some r. In this case,
the product
[TABLE]
where ai′=∑j=0tbijpj for i=1,…,m are such that a1′≥…≥am′≥0,
is a multiple of x1(a1)…xm(am) by a constant c1.
Since there are no p-adic carries when adding the exponents in the above product, the constant c1 is not zero.
We can write
[TABLE]
for some nonzero constant c2 because there are no p-adic carries when adding the exponents on the right-hand side.
Since the exponents of the middle term satisfy
[TABLE]
there is an element fpt,…,pt,ar+1′−br+1,tpt,…,am′−bm,tpt of D
with the leading term x1(pt)…xr(pt)xr+1(ar+1′−br+1,tpt)…xm(am′−bm,tpt).
The element
[TABLE]
belongs to B and has the required leading term x1(a1)…xm(am).
It is clear that none of the generators corresponding to q=p0=1 can be omitted from the generating set D.
The reason why no element corresponding to q=ps>1 can be omitted from the generating set D is because no product of generating elements from D corresponding to powers pt, where t<s can have the leading element starting with x1(q) because we always encounter a p-adic carry when adding exponents at x1.
∎
2. Supersymmetric elements of Div[x1,y1]
From now on, as usual within the content of supergroups, we assume that the characteristic p of the ground field K satisfies p>2.
In this section, we write x=x1 and y=y1, for simplicity.
First, we determine a K-basis of supersymmetric elements of Div[x,y].
Proposition 2.1**.**
The following elements form a K-basis of supersymmetric elements in Divk[x,y] of degree k=pl+s, where 0≤s<p:
•
x(t)y(k−t), where t≡s+1,…,p−1(modp);
•
(s−j)!x(k−pr−j)y(pr+j)−s(s−1)…(j+1)x(k−pr−s)y(pr+s), where j=0,…,s−1
and 0≤r≤l;
•
∑r=0l(−1)rx(pr)y(k−pr)* if s=0.*
Proof.
Let f=∑i+j=kaijx(i)y(j) be a homogeneous element of Divk[x,y].
Then
[TABLE]
On the other hand, if
f′=∑u+v=k−2buvx(u)y(v), then
[TABLE]
Comparing both expressions, we infer that f is supersymmetric if and only if the linear system, consisting of equations
[TABLE]
in variables a0,k,…,ak,0, is consistent.
Assume that s≥2. Then the above system splits into blocks of two different types.
Blocks of the first type correspond to equations labeled by (i) where i=k−pr−1,…,k−pr−s for fixed 0≤r≤l. The first equation of this block is
[TABLE]
and the last equation is
[TABLE]
Row-reducing the augmented matrix of this linear system, we obtain that it is consistent if and only if
[TABLE]
Blocks of the second type correspond to equations (i) where i=k−pr−s−1,…,k−(r+1)p for 0≤r<l. Since the coefficient matrix of this system
is triangular and all coefficients on the main diagonal are nonzero, a system of the second type is always consistent.
An important observation is that the variable bk−pr−s−1,pr appears only in a block of the second type and
variable bk−pr−1,pr appears just in a block of the second type. This implies that the original system is consistent if and only if
all systems corresponding to blocks of the first type are consistent.
Since at,k−t, where t≡s+1,…,p−1(modp), only appear in a block of the second type, we derive that corresponding element
x(t)y(k−t) is supersymmetric. We obtain additional generating supersymmetric elements by setting
[TABLE]
for j=0,…s−1
and each 0≤r≤l in the equation (\ref∗).
If s=1, then each block of the first type consists of a single equation ak−pr,pr+ak−pr−1,pr+1=0,
showing that −x(k−pr)y(pr)+x(k−pr−1)y(pr+1), where 0≤r≤l, are additional generating supersymmetric elements.
If s=0, then there are only blocks of the first type, but there is an overlap between conditions from different blocks.
Namely, we get equations
[TABLE]
for 0≤r<l, which imply that
∑r=0l(−1)rx(pr)y(k−pr)
is supersymmetric.
∎
The lexicographic order on monomials x(i)y(j) is given as x(i1)y(j1)>x(i2)y(j2) if i1>i2,
or i1=i2 and j1>j2.
Let f∈Div[x,y] be such that f=∑i,jfi,jx(i)y(j). We call x(i)y(j) the leading term of f if fi,j=0, and
fk,l=0 implies x(i)y(j)≥x(k)y(l).
Now we describe algebra generators of the algebra S of supersymmetric elements in Div[x,y].
Proposition 2.2**.**
The algebra S of supersymmetric elements of Div[x1,y1] is generated by elements ∑r=0ps−1(−1)rx(pr)y(ps−pr) for s>0,
and xy(k−1)−ky(k) for k>0.
Proof.
Denote by A the algebra generated by elements from the text of the lemma, and by Ak its homogeneous component of degree k.
It is clear that A is a subalgebra of S.
We consider f∈Sk and proceed by the induction on the degree k.
It is clear that S1 is generated by the element x−y∈A. Assume now that Su=Au for u≤k.
By Proposition 2.1, all leading terms of f∈Sk are of the form x(t)y(k−t), where 0≤t≤k is not divisible by p, together with
one more term x(k) if k is divisible by p.
Denote by f(t,k−t) an element of Sk that has x(t)y(k−t) as its leading term.
If t≡0,p−1(modp),
then (x−y)f(t,k−t)∈S1Sk has a leading term x(t+1)y(k−t). If t=0, then A contains a generator with the leading term xy(k).
If t≥p and t is divisible by p,
then there are elements f(1,k−t)∈Sk−t+1 and f(t,0)∈St with leading terms xy(k−t) and x(t), respectively.
Their product f(1,k−t)f(t,0)∈Sk−t+1St has the leading term x(t+1)y(k−t).
Finally, A contains a generator f(ps,0) with the leading term x(ps). It t+1≡0(modp) and t+1=avpv+…a1p is the p-adic expansion of t+1, then
f(pv,0)av…f(p,0)a1∈A has the leading term x(t+1).
It follows from the inductive assumption and from Lemma 2.1 that for every f∈Sk+1
there is an element a∈Ak+1 that has exactly the same leading term as f. Subtracting a suitable scalar multiple of a from f we obtain another element
g∈Sk+1 with the smaller leading term. Induction on the lexicographical order of leading terms of elements in Sk+1 concludes the proof.
∎
Remark 2.3**.**
It would be interesting to know if the set of generators in Proposition 2.1 is minimal; meaning that if any of its element is omitted,
then the resulting set does not generate S.
Analogously to the arguments in the proof of Proposition 1.1, we cannot eliminate any generator with the leading term x(ps) and still generate S.
Also, if k≤p, then no element of the form xy(k−1)−ky(k) can be generated from elements of the form xy(l−1)−ly(l), where l<k. Therefore no element
xy(k−1)−ky(k) for k≤p can be omitted either. This suggests that the set of generators in Proposition 2.1 might be minimal, but further investigation is required.
Remark 2.4**.**
Note that in the case of classical supersymmetric polynomials over the field of characteristic zero, the algebra generators are given by −Ck=xyk−1−yk for k>0.
Since (k−1)![xy(k−1)−ky(k)]=xyk−1−yk, we have an exact correspondence of algebra generators for degrees k≤p.
If k>p, then we replace −Ck with xy(k−1)−ky(k).
To facilitate the transition through the degrees k that are multiples of p, we need generators with leading terms x(ps) for s>1. It turns out that is all that is required for Div[x,y].
3. Characterization of supersymmetric elements in Div[x1,…,xm,y1,…,yn]
For fields K of characteristic zero, generators of supersymmetric polynomials (as presented in [2]) are given as
[TABLE]
where σi(x) is the i-th elementary symmetric polynomial in variables x1,…,xm and pj(y) is the complete j-th symmetric polynomial in variables
y1,…,yn. Following Remark 2.4, we first show how to transform the generators Ct to supersymmetric elements of Div[x,y] over a field of characteristic p.
We have
[TABLE]
and
[TABLE]
Denote by d(j1,…,jn) the highest power of p dividing the product (j1!…jn!) and by dk=min{d(j1,…,jn)∣j1+…+jn=k}.
It is obvious that the sequence {dk}k=0∞ is nondecreasing. Denote by ℓk the highest term (in the lexicographic order)
y1(j1)…yn(jn)∈Div[y1,…,yn]
such that (j1,…,jnk) is not divisible by pdk+1.
If t≤m, then Ct represents a nonzero supersymmetric element Et of Div[x,y] and its highest term is x1…xt.
If t>m, then the expression pdt−mCt represents a nonzero supersymmetric element Et of
Div[x,y] and its highest term with respect to the lexicographic order is x1…xmℓt−m.
Next, we characterize the supersymmetric elements in Div[x,y].
To simplify the notation, we write x(i1,…,im)y(j1,…,jn) in place of the monomial x1(i1)…xm(im)y1(j1)…yn(jn).
Proposition 3.1**.**
A homogeneous element
[TABLE]
of Divk[x1,…,xm,y1,…,yn]
is supersymmetric if and only if for each t≥0 such that i1+j1=t=pl+s, where 0≤s<p and arbitrary
i2,…,im,j2,…,jm such that i2+…+im+j2+…+jn=k−t the set of equations
[TABLE]
[TABLE]
for 0≤r<l,
and
[TABLE]
for any permutations σ∈Σm and τ∈Σn,
is consistent.
Proof.
We compute
[TABLE]
[TABLE]
On the other hand, if
[TABLE]
then
[TABLE]
[TABLE]
Therefore f satisfies (dxd+dyd)(f)=(x1−y1)f′ if and only if the linear system, consisting of equations
[TABLE]
where i1+…+im+j1+…+jn=k−1 is consistent.
The above equation (i1j1) differs from the equation (i) only by inserting ”frozen” indices i2,…,im,j2,…,jm.
Repeating the arguments from the proof of Lemma 2.1, we conclude that the condition (dxd+dyd)(f)=(x1−y1)f′ is equivalent to
the consistency of the system given by the set of equations (3) and (4) involving fixed t≥0 such that i1+j1=t=pl+s, where 0≤s<p, 0≤r<l, and arbitrary ”frozen” indices
i2,…,im,j2,…,jm satisfying i2+…+im+j2+…+jn=k−t.
The claim now follows if we recall that a supersymmetric element is symmetric with respect to indices x1,…,xm and y1,…,yn separately.
∎
We call any equation of type (3), (4) or (5) a defining equation, and a linear system consisting of all equations
(3), (4) and (5) the defining linear system.
We would like to extend further the idea presented in the proof of Proposition 2.2 and describe supersymmetric elements in Div[x,y]
using the leading terms.
Let us call a monomial x(i1,…,im)y(j1,…,jn) and the corresponding variable ai1,…,im,y1,…,ynsymmetrized if i1≥i2≥…≥im and j1≥j1≥…≥jn.
Symmetrized variables are ordered with respect to the lexicographic order.
Using the action of the symmetric group Σm on the variables x1,…,xm and the action of the symmetric group Σn on the variables y1,…,yn, every monomial x(i1,…im)y(j1,…,jn) is conjugated to a unique symmetrized monomial x(u1,…,um)y(v1,…,vn), which we call its symmetrization. In this case we also call au1,…,um,v1,…,vn the symmetrization of ai1,…,im,y1,…,yn.
If the symmetrization of x(i1,…,im)y(j1,…,jn) is a leading term of a supersymmetric element in Div[x,y],
then we call x(i1,…,im)y(j1,…,jn) and the corresponding variable ai1,…,im,j1,…,jnmarked in Div[x,y]. If this is not the case, then we call them unmarked in Div[x,y].
The next lemma shows a connection between unmarked monomials from different Div[x,y].
Lemma 3.2**.**
*Assume that a monomial x~(i1′,…,im′′)y~(j1′,…,jn′′) in Div[x~1,…,x~m′,y~1,…,y~n′] is obtained from a monomial
x(i1,…,im)y(j1,…,jn) in Div[x1,…,xm,y1,…,yn] by deleting some of the variables xi and yj.
If the monomial x~(i1′,…,im′′)y~(j1′,…,jn′′) is unmarked in Div[x~1,…,x~m′,y~1,…,y~n′], then
the monomial x(i1,…,im)y(j1,…,jn) is unmarked in Div[x1,…,xm,y1,…,yn].
*
Proof.
Let
[TABLE]
be a supersymmetric element in Div[x1,…,xm,y1,…,yn].
Fix a set of indices {k1,…,ks} from {1,…,m} and a set of indices {l1,…,lt} from {1,…,n}.
Write
[TABLE]
where
[TABLE]
By Proposition 3.1, each f(ik1,…iks,jl1,…jlt) is a supersymmetric element in
[TABLE]
obtained by “freezing” variables xki for i=1,…,s, ylj for j=1,…,t and relabeling the remaining variables as x~1,…,x~m−s
and y~1, …, y~n−t.
If the monomial
[TABLE]
is unmarked, then for every supersymmetric polynomial f′ in Div[x~,y~] such that
a~i1′,…,im−s′,j1′,…,jn−t′=0 there is a higher a~u1′,…,um−s′,v1′,…,vn−t′=0.
If we apply this to f(ik1,…,iks,jl1,…,jlt), we obtain that
ai1,…,ik1,…,iks,…,im,j1,…,jl1,…,jlt,…,jn=0 implies that for some higher coefficient
au1′,…,ik1,…,iks,…,um−s′,v1′,…,jl1,…,jlt,…,vn−t′=0.
Therefore x(i1,…,im)y(j1,…,jn) is unmarked in Div[x1,…,xm,y1,…,yn].
∎
The next lemma describes generators of the algebra S of all supersymmetric elements in Div[x,y].
Proposition 3.3**.**
For each marked symmetrized monomial x(i1,…,im)y(j1,…,jn) of Div[x,y] choose a supersymmetric element
S(x(i1,…,im)y(j1,…,jn)) in Div[x,y] such that x(i1,…,im)y(j1,…,jn) is its leading term.
Denote by B the subalgebra of S generated by all such elements S(x(i1,…,im)y(j1,…,jn)). Then B=S.
Proof.
For a nonzero element f∈S denote by ℓ(f)=x(i1,…,im)y(j1,…,jn) its leading term, and by
cℓ(f) the coefficient of f at ℓ(f).
We proceed by induction on the lexicographic order of the leading term ℓ(f). Clearly the minimal f=1 belongs to B.
Assume that all elements g∈S with ℓ(g)<ℓ(f) belong to B. Then f−cℓ(f)S(ℓ(f)) belongs to S and its leading term
is smaller than ℓ(f), which implies
f−cℓ(f)S(ℓ(f))∈B. This shows that f=S(ℓ(f))+(f−cℓ(f)S(ℓ(f))∈B.
∎
We use the above proposition to describe the algebra S once all marked and unmarked monomials of Div[x,y] and elements
S(x(i1,…,im)y(j1,…,jn)) are determined.
We need to accomplish two different tasks.
The first task is to determine all marked monomials by constructing supersymmetric elements that have these monomials as their leading term.
The second task is to show that every remaining monomial is unmarked. We assume that the coefficient ai1,…,im,j1,…,jn
appearing in the presentation of a supersymmetric element
[TABLE]
is not zero. Using equations characterizing supersymmetric elements, we need to derive that
ai1′,…,im′,j1′,…,jn′=0 for some coefficient higher than ai1,…,im,j1,…,jn.
In what follows we write
[TABLE]
where 0≤ru,sv<p for u=1,…,m and v=1,…,n.
The next lemma plays a vital role in what follows.
Lemma 3.4**.**
Assume that p divides some of i1,…,im and j1>0. Then the
monomial x1(i1,…,im)y1(j1,…,jn) is unmarked in
Div[x,y].
Proof.
Assume that p divides it. It follows from Proposition 2.1 that xt(it)y1(j1) in unmarked in Div[xt,y1].
We can also prove it directly by pointing out the equation that involves the variable ait,j1 such that all other appearing symmetrized
variables are higher than ait,j1.
If 0<s1<p, then this equation is
[TABLE]
If s1=0, then l1>0 and this equation is
[TABLE]
Adding variables x1,…,xr,…,xm,y2,…,yn concludes the claim.
∎
Let us note that as a special case, if j1>0 and im=0, then the monomial x1(i1,…,im)y1(j1,…,jn) is unmarked in Div[x1,…,xm,y1,…,yn].
To get an intuition about the structure of supersymmetric elements, apply Lemma 3.1, work with symmetrized variables corresponding to ai1…imy1…yn and look for the free variables given by an echelon form of the matrix of the total linear system given by equations of type (3) and (4).
Lemma 3.5**.**
For every k1>1, the monomial x1(pk1) is marked in Div[x,y].
Proof.
It is straightforward to verify that the monomial x1(pk1) is the leading term of the supersymmetric element
[TABLE]
in Div[x,y].
∎
4. Supersymmetric elements in Div[x1,…,xm,y1]
Throughout this section, we assume m>1. Let us start with the following lemma.
Lemma 4.1**.**
Let m>1.
Denote by C the subalgebra of Div[x1,…,xm,y1] generated by symmetrized monomials
•
ρi=x1…xi* for i=1,…m;*
•
ρm+j1=ρmy1(j1)=Em+j1* for j1>0 (corresponding to generators from characteristic zero case);*
•
x1(ps)* for s>1;*
•
τi1,…,im,j1=x(i1,…,im)y1(j1), where i1+…+im is divisible by p but none of i1,…,im are divisible by p,
and j1≥0.
Then C has a K-basis consisting of symmetrized monomials
•
x1(k)* where k is divisible by p;*
•
x(i1,…,im)y1(j1), where p does not divide i1,i2,…,im and j1≥0.
Proof.
The proof is left to the reader.
∎
By Lemma 3.5, the monomials x1(ps) for s>1 are marked in Div[x1,…,xm,y1].
The following lemma shows that the monomials ρi, ρm+j1, and τi1,…,im,j1 are also marked in Div[x1,…,xm,y1].
Hence all monomials appearing in Lemma 4.1 are marked in Div[x1,…,xm,y1].
Proposition 4.2**.**
Every symmetrized monomial x(i1,…,im)y1(j1), where j1≥0 and none of positive indices i1,…,im is divisible by p,
is marked in Div[x1,…,xm,y1].
Proof.
Recall the notation j1=pl1+s1 and iu=pku+ru for u=1,…,m, where 0≤ru,s1<p.
Denote by Mv the set of all ordered v-tuples of indices (m1,…,mv) such that
[TABLE]
and by M the union of all such Mv.
We show that the element
[TABLE]
that has
[TABLE]
as its leading term (corresponding to v=0), is supersymmetric in Div[x1,…,xm,y1].
Denote by Nv the set of all variables
[TABLE]
where (m1,…,mv)∈Mv, and by N the union of all such Nv.
Every defining equation, that involve a variable from N, contains
exactly two variables from N.
The variables
[TABLE]
and
[TABLE]
appear in the equation
[TABLE]
[TABLE]
corresponding to the values j=rm1+…+rmv+s1 and j=rm1+…+rmt+…+rmv+s1
where (m1,…,mt,…,mv)∈Mv−1 is obtained from (m1,…,mv)∈Mv by removing mt.
Assume (m1′,…,mt′,…,mv+1′)∈Mv+1 is obtained from (m1,…,mv)∈Mv by inserting mt′.
That means that ml′=ml for l<t and ml+1′=ml for l≥t. Then the variables
[TABLE]
and
[TABLE]
appear in the equation
[TABLE]
[TABLE]
corresponding to the values j=rm1+…+rmv+s1 and j=rm1′+…+…+rmv+1′+s1.
Since there are no other equations involving variables from N, we can set the value of every variable not in N equal to zero. Then the linear system turns into
a smaller one, consisting of equations
[TABLE]
[TABLE]
for all (m1,…,mv)∈Mv and (m1′,…,mv+1′)∈Mv+1 as above.
Since
[TABLE]
and
[TABLE]
by setting
[TABLE]
we obtain a solution of the defining linear system and the corresponding supersymmetric element.
∎
In the next proposition we consider monomials in Div[x1,x2,y1].
The result then extends to monomials in Div[x1,…,xm,y1,…,yn].
Proposition 4.3**.**
Every symmetrized monomial
x(i1,i2)y1(j1) such that i2>0 and p divides i1 or i2
is unmarked in Div[x1,x2,y1].
Proof.
If j1>0, then the statement follows from Lemma 3.4. Therefore assume j1=0.
If r2=0, then k2>0 and ai1,i2,0=(−1)k2ai1,0,i2=ai1+i2,0,0 which is higher than ai1,i2,0, showing that ai1,i2,0 is unmarked.
If r2>0, then the equation
[TABLE]
shows that ai1,i2,0=0 implies
ai1,pk2+r2−j,j=apk2+r2−j,i1,j=0 for some j>0.
The equation
[TABLE]
shows that in this case apk1+j−i,pk2+r2−j,i=0 for some i<j.
Since apk1+j−i,pk2+r2−j,i is higher than ai1,i2,0, the claim follows.
∎
Corollary 4.4**.**
Every symmetrized monomial
x1(i1,…,im)y1(j1,…,jn)
such that p divides ir>0 for some r>1, or p divides i1 and i2>0,
is unmarked in Div[x,y].
The algebra A of all supersymmetric elements in Div[x1,…,xm,y1] is generated by elements S(M), where M is one of the symmetrized monomials
•
ρi=x1…xi* for i=1,…m;*
•
ρm+j1=ρmy1(j1)=Em+j1* for j1>0 (corresponding to generators from characteristic zero case);*
•
x1(ps)* for s>1;*
•
τi1,…,im,j1=x1(i1,…,im)y1(j1), where i1+…+im is divisible by p but none of i1,…,im are divisible by p,
and j1≥0.
Proof.
Lemma 3.5 shows that every monomial x1(k), where k is divisible by p is marked. Proposition 4.2 shows that every symmetrized
monomial x(i1,…,im)y1(j1), where p does not divide any positive i1,i2,…,im and j1≥0, is also marked.
On the other hand, Corollary 4.4 shows that all remaining monomials are unmarked.
Since all monomials M from the text of the theorem are marked,
Lemma 4.1 shows that each marked monomial is the leading term of some element of the algebra A.
Proposition 3.3 concludes the proof.
∎
The following lemma is needed later.
Lemma 4.6**.**
Assume that f=∑(i1,i2,j1)a(i1,i2,j1)x(i1,i2)y(j1) is a supersymmetric element in Div[x1,x2,y1].
If p divides j1 and i2, and i1≥p, then
ai1,i2,j1=ai1−p,i2+p,j1=…=as1,i1+i2−s1,j1.
Proof.
It is enough to show ai1,0,i2+j1=as1,i1−s1,i2+j1.
We consider the following system of equations:
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
When we expand the sum ∑i=1s1−1(is1)ai1−i,0,i2+j1+i using equations (es1−1′) through (e1′)
we observe that the term as1−l−i,i1−s1+l,i2+j1+i, where 1≤l≤s1−1 and 0≤i<s1−l appears with the coefficient −(is1−l)(s1−ls1).
When we expand the sum ∑i=1s1−1(is1)as1−i,i1−s1,i2+j1+i using equations (es1−1) through (e1)
we observe that the term ai1−s1+l,s1−l−i,i2+j1+i, where 1≤l≤s1−1 and 0≤i<s1−l appears with the coefficient −(il+i)(l+is1).
Since (is1−l)(s1−ls1)=i!(s1−l−i)!l!s1!=(il+i)(l+is1), we obtain
[TABLE]
This together with
[TABLE]
implies ai1,0,i2+j1=as1,i1−s1,i2+j1.
∎
Remark 4.7**.**
It is not true that for each supersymmetric element f=∑(i1,i2,j1)a(i1,i2,j1)x(i1,i2)y(j1) the conditions i1≡u1(modp), i2≡u2(modp) imply ai1,i2,0=au1,u2,0.
As a counterexample, there is a supersymmetric element f for which a7,1,0=a4,4,0 and p=3.
5. Supersymmetric elements in Div[x1,y1,y2]
Throughout this section, we assume that n>1. Let us introduce the concept of the height of n-tuple (y1,…,yn) in a form suitable to a more general setting
of Div[x,y].
Definition 3**.**
Let j1≥…≥jn≥0. If
•
j1≤p−1* and for every 1≤t≤n such that st=p−1 we have jt+1=0 and su=p−1 for 1≤u<t, or*
•
j1>p−1* and s1=…=sn−1=p−1,*
then we say that (j1,…,jn) has the height h=1; otherwise we say that it has the height h>1.
For Div[x1,y1,y2] we have n=2 and a simpler description - (j1,j2) is of the height one if and only if either 0<j1<p−1 and j2=0, or s1=p−1.
Lemma 5.1**.**
Denote by C the subalgebra of Div[x1,y1,y2] generated by symmetrized monomials
•
x1(ps)* for s>0;*
•
x1y(j1,j2)* for (j1,j2) of height h=1;*
•
x1(2)y(j1,j2)* for (j1,j2) of height h>1.*
Then C has a K-basis consisting of symmetrized monomials
•
x1(i1), where r1=0;
•
x1(i1)y(j1,j2), where r1≥1 and (j1,j2) has height h=1;
•
x1(i1)y(j1,j2), where r1≥2 and (j1,j2) has height h>1.
Proof.
The proof is left to the reader.
∎
The next statement is formulated for a more general case Div[x1,y1,…,yn] instead of Div[x1,y1,y2].
Lemma 5.2**.**
If (j1,…,jn) has the height h=1, then every symmetrized monomial
x1y(j1,…,jn) is marked in Div[x1,y1,…,yn].
Proof.
If j1≤p−1 and for every 1≤t≤n such that st=p−1 we have jt+1=0 and su=p−1 for 1≤u<t, then
x1y1(j1,…,jn) is marked since it is the leading term of E1+j1+…+jn.
Therefore we can assume s1=…=sn−1=p−1. If sn=p−1, then a1,j1,…,jn does not appear in any defining equation and is therefore marked.
If sn=p−1, then we have a series of p−1−sn pairs of equations for each choice of t=1,…,n−1:
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
These are the only defining equations involving variables a1,jn,j1,…,jn−1, a0,jn+1,j1,…,jn−1,
[TABLE]
and
[TABLE]
for all t=1,…,n−1.
When we set all variables that are not listed above to zero and a1,j1,j2,…,jn=1, the values of the remaining variables are
[TABLE]
[TABLE]
for j=1,…,p−1−sn and
[TABLE]
for j=1,…,p−2−sn.
Denote by gj1,…,jn the symmetric function analogous to fi1,…,im, defined on y1,…,yn instead of x1,…,xm.
Then
[TABLE]
is a supersymmetric element that has the leading term x1y(j1,…,jn), showing that a1,j1,…,jn is marked.
∎
For the next three lemmas, we work inside Div[x1,y1,y2].
Lemma 5.3**.**
Every symmetrized monomial
x1(2)y(j1,j2) is marked in Div[x1,y1,y2].
Proof.
The statement is true for (j1,j2) of height h=1 by Lemma 5.2.
First, assume that j1>j2.
Then the element
[TABLE]
is a supersymmetric element with the leading term x1(2)y1(j1)y2(j2), showing that a2j1j2 is marked.
This is easy to verify since the variables corresponding to summands in this expression appear just in the following
defining equations
[TABLE]
for t=0,1
and
[TABLE]
for t=0,1.
If j1=j2=j, then
[TABLE]
is a supersymmetric element with the leading term x1(2)y(j,j) showing that a2jj is marked.
This is easy to verify this since the variables corresponding to summands in this expression appear only in the
defining equations (A0)=(B0) and (A1)=(B1).
∎
Lemma 5.4**.**
Assume that f=∑(i1,j1,j2)a(i1,j1,j2)x(i1)y(j1,j2) is a supersymmetric element in Div[x1,y1,y2].
If r1=1, s1<p−1, l1>0 and s2=0, then ai1,j1,j2=0 implies that some au1,v1,v2=0, where u1>i1.
Proof.
First assume l2>0. Then the claim follows by considering the system of equations
[TABLE]
If l2=0, then j2=0. Considering the series of s1+1 pairs of equations
[TABLE]
implies the claim.
∎
Lemma 5.5**.**
If r1=1 and (j1,j2) has the height h>1, then the symmetrized monomial x1(i1)y(j1,j2) is unmarked in Div[x1,y1,y2].
Proof.
Since (j1,j2) has height h>1, we have 0≤s1<p−1.
If s2>0, then the equations
[TABLE]
[TABLE]
imply that ai1,j1,j2 is unmarked since ai1−1,j2,j1+1 is a linear combination of terms higher than ai1,j1,j2.
If s2=0 and l2>0, then ai1,j1,j2 is unmarked by Lemma 5.4.
Finally, if j2=0, then j1>p−1 and ai1,j1,0 is unmarked by Lemma 5.4.
∎
Corollary 5.6**.**
If r1=1 and (j1,…,jn) has the height h>1, then every symmetrized monomial x1(i1)y(j1,…,jn) is unmarked in Div[x1,y1,…,yn].
Proof.
If (j1,…,jn) has the height h>1, then there is a pair (ju,jv), obtained by removing entries from (j1,…,jn), that has the height h>1.
Use Lemmas 5.5 and 3.2 to complete the proof.
∎
Theorem 5.7**.**
The algebra A of all supersymmetric elements in Div[x1,y1,y2] is generated by elements S(M), where M is one of the symmetrized monomials
•
x1(ps)* for s>0;*
•
x1y(j1,j2)* for (j1,j2) of height h=1;*
•
x1(2)y(j1,j2)* for (j1,j2) of height h>1.*
Proof.
Lemmas 5.1, 3.5, 5.2 and 5.3 describe all marked monomials.
Lemma 3.4 and Corollary 5.6 describe all unmarked monomials. Lemma 5.1 and Proposition 3.3 conclude the proof.
∎
6. Supersymmetric elements in Div[x1,x2,y1,y2]
Lemma 6.1**.**
Denote by C the subalgebra of Div[x1,x2,y1,y2] generated by following symmetrized monomials:
•
x1;
•
x1(ps)* for some s>0;*
•
x1x2y(j1,j2), where (j1,j2) has the height h=1;
•
x(i1,i2)y(j1,j2), where r1=r2=1, k2>0 and s1=p−1;
•
x(i1,i2)y(j1,j2), where r1=r2=2.
Then C has a K-basis consisting of the following monomials
•
x1(i1);
•
x1(i1)x2y(j1,j2), where r1>0 and (j1,j2) has the height h=1;
•
x(i1,i2)y(j1,j2), where r1>0, r2=1, k2>0 and s1=p−1;
•
x(i1,i2)y(j1,j2), where r1,r2≥2.
Proof.
The proof is left for the reader.
∎
Lemma 6.2**.**
Every symmetrized monomial x1x2y(j1,j2), where (j1,j2) has the height h=1, is marked in
Div[x1,x2,y1,y2].
Proof.
The monomial x1x2y(j1,j2) is the leading term of the supersymmetric element E2+j1.
∎
If j1=p−1, then the claim of Lemma 6.2 is a particular case of Lemma 6.3.
Lemma 6.3**.**
Every symmetrized monomial x(i1,i2)y(j1,j2), where r1=r2=1, k2>0 and s1=p−1, is marked in
Div[x1,x2,y1,y2].
Proof.
We modify the proof of Lemma 5.2. To start, we recall the proof of Lemma 5.2 restated for Div[x1,y1,y2].
If s2=p−1, then ai1,j1,j2 does not appear in any defining equation and is therefore marked.
If s2=p−1, then we have a series of p−1−s2 pairs of equations
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
that are all the defining equations involving variables
[TABLE]
[TABLE]
and
[TABLE]
When we set all variables not listed above to zero and a1,j1,j2=1, the values of the remaining variables are
[TABLE]
[TABLE]
for j=1,…,p−1−s2 and
[TABLE]
for j=1,…,p−2−s2.
Then
[TABLE]
is a supersymmetric element that has the leading term x1y(j1,j2), showing that a1,j1,j2 is marked.
Now we make necessary adjustments for Div[x1,x2,y1,y2].
If s2=p−1, then ai1,i2,j1,j2 does not appear in any defining equation and is therefore marked.
If s2=p−1, then we have a series of p−1−s2 quadruples of equations
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
that are all the defining equations involving variables
[TABLE]
[TABLE]
[TABLE]
and
[TABLE]
When we set all variables not listed above to zero and ai1,i2,j1,j2=1, then values of the remaining variables are
[TABLE]
[TABLE]
for j=1,…,p−1−s2 and
[TABLE]
for j=1,…,p−2−s2.
Then
[TABLE]
is a supersymmetric element that has the leading term x1(i1)x2(i2)y1(j1)y2(j2), showing that a1,j1,j2 is marked.
∎
Lemma 6.4**.**
Every symmetrized monomial x(i1,i2)y(j1,j2), where r1=r2=2, is marked in
Div[x1,x2,y1,y2].
The statement is true if (j1,j2) has the height h=1 by Lemma 6.3.
Consider the following defining equations
[TABLE]
for t=0,1;
[TABLE]
for t=0,1;
[TABLE]
for t=0,1;
[TABLE]
for t=0,1;
If j1>j2, then
[TABLE]
is a supersymmetric element that has the leading term x(i1,i2)y(j1,j2), showing that ai1i2j1j2 is marked.
This is easy to verify since the variables corresponding to summands in this expression appear only in the
defining equations (A1t),(A2t),(B1t),(B2t) for t=0,1.
If j1=j2=j, then s1=s2=s and
[TABLE]
is a supersymmetric element that has the leading term x(i1,i2)y(j,j) showing that ai1i2jj is marked.
This is easy to verify this since the variables corresponding to summands in this expression appear only in the
defining equations (A0t)=(B0t) and (A1t)=(B1t).
∎
Remark 6.5**.**
The modification of the proofs of Lemmas 6.3 and 6.4 from the proofs of Lemmas 5.2 and 5.3 by ”doubling” the value of r to
r1=r2 shows how to proceed in general to build marked elements in Div[x1,…,xm,y1,y2] from those in Div[x1,y1,y2] in the case when r1=…=rm.
It is not clear if the assumption r1=…=rm is really needed, but we require the modification only in those cases.
Proposition 6.6**.**
The symmetrized monomials x(i1,i2)y(j1,j2) such that
•
p* divides i1 or i2 and either i2>0 or j1>0; or*
•
r1=1* or r2=1 and (j1,j2) has height h>1; or*
•
r1>0*, r2=1, k2>0, s1=p−1 and (j1,j2) has height h=1
*
are unmarked in Div[x1,x2,y1,y2].
Proof.
The first part of the statement follows from Lemma 3.4 and Corollary 4.4.
The second part follows from Lemmas 5.5 and 3.2.
For the last part, take (j1,j2) of the form (j1,0), where 0≤j1<p−1 and consider the defining equation
[TABLE]
By Lemma 4.6, we have ai1,i2−1,j1+1,0=ai1+i2−1,0,j1+1,0. Since ai1+i2−1,0,j1+1,0 and all variables
ai1,i2+s1−i,j1−s1+i,0 for i=0,…,s1−1 are higher than ai1,i2,j1,0 we conclude that ai1,i2,j1,0 is unmarked.
∎
Theorem 6.7**.**
The algebra S of all supersymmetric elements in Div[x1,x2,y1,y2] is generated by elements S(M), where M is one of the symmetrized monomials
•
x1;
•
x1(ps)* for some s>0;*
•
x1x2y(j1,j2), where (j1,j2) has height h=1;
•
x(i1,i2)y(j1,j2), where r1=r2=1, k2>0 and s1=p−1;
•
x(i1,i2)y(j1,j2), where r1=r2=2.
Proof.
Lemmas 6.1, 3.5, 6.2, 6.3 and 6.4 describe all marked monomials.
Proposition 6.6 describes all unmarked monomials. Lemma 6.1 and Proposition 3.3 conclude the proof.
∎
Bibliography5
The reference list from the paper itself. Each links out to its DOI / PubMed record.
1[1] Grishkov, A.N., Marko, F. and Zubkov, A.N., Generators of supersymmetric polynomials in positive characteristic , Journal of Algebra 349 (2012) 38-49.
2[2] Kantor, Issai and Trishin, Ivan, The algebra of polynomial invariants of the adjoint representation of the Lie superalgebra g l ( m | n ) 𝑔 𝑙 conditional 𝑚 𝑛 gl(m|n) , Commun. in Algebra, 25(7) (1997), 2039-2070.
3[3] La Scala, R. and Zubkov, A.N., Donkin-Koppinen filtration for general linear supergroup , Algebr. Represent. Theory 15 (2012), no. 5, 883-899.
4[4] Springer, T.A, Invariant theory , Lecture Notes in Mathematics 585, Springer-Verlag, Berlin, Heiderberg, New York, 1977.
5[5] Stembridge, J.R., A characterization of supersymmetric polynomials , J. Algebra 95(2) (1985), 439-444.