• Aucun résultat trouvé

The results of this paper mainly concern the tangentially weakly defective join of subspace varieties

N/A
N/A
Protected

Academic year: 2022

Partager "The results of this paper mainly concern the tangentially weakly defective join of subspace varieties"

Copied!
6
0
0

Texte intégral

(1)

VARIETIES

MING YANG AND WEN LIU

Abstract. Using Terracini’s Lemma, we study the tangential weak defectivity of join of subspace varieties.

1. Introduction

In [2], L. Chiantini and G. Ottaviani introduced the tangential weak defectivity of join of projective varieties, which was related to uniqueness conditions concerning tensor decompo- sitions (see [1, 3, 4, 10]). In [12], the authors studied the non-tangentially weakly defective join of subspace varieties and used it to obtain some uniqueness results of block term tensor decompositions. The results of this paper mainly concern the tangentially weakly defective join of subspace varieties.

Throughout this paper, for basic definitions, notation and results, we follow [8].

Firstly we recall some basic concepts in algebraic geometry.

1.1. Notations. As in [8], for a finite dimensional complex vector space V,PV denotes the projective space associated to V, π denotes the projection of V\{0} onto PV; for a variety X ⊂PV, ˆX ⊂V denotes its inverse image under the projectionπ, which is the (affine) cone over X in V, and for x∈X, [x] denotes π(x).

Let S be a subset of PV, then the span hSi is by definition the range of π on the usual vector span of ˆS in V. The Zariski closure of S in PV will be denoted by ¯S.

When we need to specify the elements of S and its linear span, we use the notation {s1, s2,· · · } and hs1, s2,· · · i, respectively.

Forx∈X, ˆˆ T[x]X := ˆTxXˆ is the affine tangent space to X at [x].

1.2. Join of subspace varieties. LetAj, 1≤j ≤n, be finite dimensional complex vector spaces.

Definition 1.1. (See Definition 1 in [9]) Let kj ≤aj := dimAj, 1≤ j ≤ n be nonnegative integers. Subspace varieties, denotedSubk1,...,kn(A1⊗. . .⊗An)∈P(A1⊗. . .⊗An) are defined as

Subk1,...,kn(A1⊗. . .⊗An)

:={[T]∈P(A1⊗. . .⊗An)| ∀j ∃A0j ⊂Aj, dim A0j =kj, T ∈A01⊗. . .⊗A0n}.

Definition 1.2. IfXi, i= 1, . . . , k,k ≤nare projective algebraic varieties ofPn =PV, V = Cn+1, then the join of X1, . . . , Xk is

J(X1, . . . , Xk) :=[

{h[P1], . . . ,[Pk]i|Pi ∈Xˆi, 1≤i≤k},

1991Mathematics Subject Classification. classical algebraic geometry.

Key words and phrases. subspace variety, joins of varieties, tangential weak defectivity.

1

(2)

where Pi, i = 1, . . . , k, are linearly independent vectors in V. If X1 = · · · =Xk = X, then we write J(X1, ..., Xk) = σk(X) and we call this the k-th secant variety to X.

The following Terracini’s Lemma appears as Proposition 12.11 in [6] and we rephrased it in terms of join of projective varieties.

Theorem 1.3. Let Pi ∈ Xˆi be a general point of Xˆi for each i = 1, . . . , k, then for [P] :=

[P1+. . .+Pk],

[P]J(X1, . . . , Xk) = ˆT[P1]X1+· · ·+ ˆT[Pk]Xk. (1.1)

Definition 1.4. (see Definition 2.6 in [2]) Let Pi ∈ Xˆi be a general point of ˆXi, 1 ≤ i ≤ k. When for any j ∈ {1, . . . , k} and Qj ∈ Xˆj, ˆT[P1]X1 +· · · + ˆT[Pk]Xk contains ˆT[Qj]Xj only if [Qj] ∈ {[P1], . . . ,[Pk]}, we say J(X1, . . . , Xk) is not tangentially weakly defective.

Otherwise, we say thatJ(X1, . . . , Xk) istangentially weakly defective. J(X1, . . . , Xk) isweakly defective if the general hyperplane which is tangent to X1, . . . , Xk at some k general points [P1], . . . ,[Pk], is also tangent at some other point [Qj] 6= [P1], . . . ,[Pk], Qj ∈ Xˆj for some j ∈ {1, . . . , k}. Here general means in an open subset of the set of hyperplanes which are tangent toX1, . . . , Xk at k general points [P1],· · · ,[Pk] (see [5]).

Remark 1.5. By semicontinuity (see Theorem III.12.8 of [7]), if for one particular set of gen- eral points{P1, . . . , Pk}, ˆT[P1]X1+· · ·+ ˆT[Pk]Xkcontains ˆT[Qj]Xj only if [Qj]∈ {[P1], . . . ,[Pk]}, then J(X1, . . . , Xk) is not tangentially weakly defective.

Remark 1.6. In [12], it is proved thatJ(Sub1,L1,L1(CI⊗CJ⊗CK), . . . ,Sub1,LR,LR(CI⊗CJ⊗ CK)) is not tangentially weakly defective, if

I ≥R, J, K ≥

R

X

r=1

Lr, Li+Lj > Lk ∀1≤i, j, k ≤R.

The main results in this paper are the following.

Theorem 1.7. AssumeI ≥R, J(Sub1,L1,L1(CI⊗CJ⊗CK), . . . ,Sub1,LR,LR(CI⊗CJ⊗CK))is tangentially weakly defective, if for some[Xj1]∈σLj1(PJ−1×PK−1), · · · , [Xjs]∈σLjs(PJ−1× PK−1), j1, . . . , js ∈ {1, . . . , R}, there exists [Xj0t] ∈ h[Xj1], · · · , [Xjs]i ∩σLjt(PJ−1×PK−1), but [Xj0t] is not in {[Xj1], · · · , [Xjs]}.

And we use the above theorem to obtain the following tangential weak defectivity results:

J(Sub1,L1,L1(CI ⊗CJ ⊗ CK),Sub1,L2,L2(CI ⊗CJ ⊗CK)) is tangentially weakly defective if I ≥ 2, J = K = 2Li2+Lj, ∀1 ≤ i, j ≤ 2 (2Li2+Lj is not independent from i, j) or I ≥ 2, min{J, K}= max{L1, L2}.

Theorem 1.8. J Sub1,L1,L1 CI⊗CJ⊗CK

, . . . , Sub1,LR,LR CI⊗CJ ⊗CK

is weakly de- fective if I ≥2, J, K ≥L1+· · ·+LR, L1, . . . , LR≥2.

Theorem 1.9. J Sub2,L1,L1 CI⊗CJ ⊗CK

, . . . , Sub2,LR,LR CI ⊗CJ ⊗CK

is tangen- tially weakly defective if I ≥2, J, K ≥L1+· · ·+LR, L1, . . . , LR≥2.

(3)

2. Proof of Theorem 1.7 And Its Corollaries

Proof. Without loss of generality, we assume that Xj0t =X12X2+· · ·+χRXR. Now a1⊗X1+· · ·+aR⊗XR

=a1 ⊗Xj0t −χ2a1⊗X2− · · · −χRa1⊗XR+a2⊗X2+· · ·+aR⊗XR

=a1 ⊗Xj0t + (a2−χ2a1)⊗X2+· · ·+ (aR−χRa1)⊗XR

=a1 ⊗Xj0t +a02⊗X2+· · ·+a0R⊗XR.

LetPi =ai⊗Xi, 1≤i≤R and Q=a1⊗Xj0t. Using Terracini’s Lemma 1.3, we have Tˆ[P1+···+PR]J(Sub1,L1,L1(CI⊗CJ ⊗CK), . . . ,Sub1,LR,LR(CI⊗CJ⊗CK))

= ˆT[P1]Sub1,L1,L1(CI⊗CJ⊗CK) +· · ·+ ˆT[PR]Sub1,LR,LR(CI⊗CJ⊗CK)

contains ˆT[Q]Sub1,Ljt,Ljt(CI ⊗CJ ⊗CK). By Remark 1.5, we have J(Sub1,L1,L1(CI ⊗CJ ⊗ CK), . . . ,Sub1,LR,LR(CI⊗CJ ⊗CK)) is tangentially weakly defective.

Corollary 2.1. J(Sub1,L1,L1(CI⊗CJ⊗CK),Sub1,L2,L2(CI⊗CJ⊗CK))is tangentially weakly defective if I ≥2, J =K = 2Li2+Lj, ∀1≤i, j ≤2.

Proof. It is sufficient to prove the case L1, L2 < J =K <P2

r=1Lr. Let A, B and C denote complex vector spaces of dimensions I, J, K respectively. Split B = B1 ⊕ B0 ⊕B2 and C =C1⊕C0 ⊕C2, where B1, B0, B2,C1, C0, and C2 are of dimensions L1 −lb, lb,L2 −lb, L1−lc, lc,L2−lc, respectively, and lb =lc= 12Lj.

Recall that there is a normal form for a general point [p] of σL(PB×PC) (L is smaller than dimB and dimC), which is of the form

p=b1⊗c1+· · ·+bL⊗cL.

We may assume that all the bi,1 ≤ i ≤ L are linearly independent in B as well as all the ci,1 ≤ i ≤ L (Otherwise one would have [p] ∈ σL−1(PB ×PC)). Then a general element [ϕ]∈Sub1,L,L(PA⊗PB×PC) is of the form

ϕ=a⊗(b1⊗c1+· · ·+bL⊗cL), where a is a nonzero vector in A.

So we my consider

X1 =b1,1⊗c1,1+· · ·+b1,L1−lb⊗c1,L1−lb +b0,1 ⊗c0,1+· · ·+b0,lb ⊗c0,lc

∈(B1⊕B0)⊗(C1⊕C0)∼=CL1 ⊗CL1 and

X2 =b2,1⊗c2,1+· · ·+b2,L2−lc⊗c2,L2−lc +b0,1⊗c0,1+· · ·+b0,lb ⊗c0,lc

∈(B2⊕B0)⊗(C2⊕C0)∼=CL2 ⊗CL2. LetXj0 be a general point of σLj(PJ−1×PK−1) and set

Xj0 =X1−X2,

(4)

then we have

Xj0 =b1,1⊗c1,1+· · ·+b1,L1−lb ⊗c1,L1−lb−b2,1⊗c2,1− · · · −b2,L2−lc ⊗c2,L2−lc

has rank equal to Lj, which implies that [Xj0] is a point in σLj(PJ−1 ×PK−1). But [Xj0] is not in{[X1],[X2]}. From Theorem 1.7, we knowJ(Sub1,L1,L1(CI⊗CJ⊗CK),Sub1,L2,L2(CI

CJ ⊗CK)) is tangentially weakly defective.

Corollary 2.2. J(Sub1,L1,L1(CI⊗CJ⊗CK),Sub1,L2,L2(CI⊗CJ⊗CK))is tangentially weakly- defective if I ≥2, min{J, K}= max{L1, L2}.

Proof. It is sufficient to prove the case L1 ≤L2 =K. Let B and C denote vector spaces of dimensions J, K respectively. Split B =B1+B2 and C =C1 ⊕C2, where B1, B2, C1, and C2 are of dimensions L1,L2,L1, L2−L1, respectively.

Consider

X1 =b1,1⊗c1,1+· · ·+b1,L1 ⊗c1,L1 ∈B1⊗C1 ∼=CL1 ⊗CL1 and

X2 =b2,1 ⊗c1,1+· · ·+b2,L1 ⊗c1,L1 +b2,L1+1⊗c1,L1+1+· · ·+b2,L2 ⊗c2,L2

∈B2⊗(C1⊕C2)∼=CL2 ⊗CL2,

where {b1,1, . . . , b1,L1}, {b2,1, . . . , b2,L2},{c1,1, . . . , c1,L1}, and {c2,L1+1, . . . , c2,L2} are bases for B1, B2,C1 and C2, respectively.

Let [Xj0] be a general point of σLj(PJ−1×PK−1) and set Xj0 =X1+X2. Then we have

Xj0 = (b1,1+b2,1)⊗c1,1+· · ·+ (b1,L1 +b2,L1)⊗c1,L1 +b2,L1+1⊗c2,L1+1+· · ·+b2,L2 ⊗c2,L2. Xj0 has rank equal to L2, which implies that [Xj0] is a point in σL2(PJ−1×PK−1). But [Xj0] is not in{[X1],[X2]}. From Theorem 1.7, we knowJ(Sub1,L1,L1(CI⊗CJ⊗CK),Sub1,L2,L2(CI

CJ ⊗CK)) is tangentially weakly defective.

3. Proof of Theorem 1.8

Lemma 3.1. Let A, B, C be complex vector spaces of dimensions I, J, K and ϕ = ai ⊗ (b1⊗c1+· · ·+bL⊗cL), where ai, bi, ci are basis for A, B, C, respectively. We have

TbϕSub1,L,L(A⊗B⊗C) =A⊗

* L X

i=1

bi⊗ci +

+B⊗ hai⊗cii+C⊗ hai⊗bii Proof. Pick a curveϕ(t) = ai(t)⊗

PL

i=1bi(t)⊗ci(t)

, whereϕ(0) = ϕ. Taking derivative with respect to t, we have

ϕ0(0) =a0i(0)⊗

L

X

i=1

bi⊗ci

!

+ai

L

X

i=1

b0i(0)⊗ci

!

+ai

L

X

i=1

bi⊗c0i(0)

! .

Since a0(0), b0(0), c0(0) are arbitrary vectors in A, B, C, we obtained Lemma 3.1.

(5)

Proof of Theorem 1.8. It is sufficient to prove the caseI = 2, J =K =PR r=1Lr.

Let A, B and C be complex vector spaces of dimensions I, J, K, respectively. Split B =L

1≤q≤RBq and C =L

1≤r≤RCr, where for 1≤q, r ≤R, Bq and Cr are of dimensions Lq, Lr, respectively.

Choose a general set {ϕp ∈ Subd1,Lp,Lp(CI ⊗CJ ⊗CK) : 1 ≤ p ≤ R}. Without loss of generality, we can assume

ϕp = (a1pa2)⊗(bp,1⊗cp,1+bp,2⊗cp,2+· · ·+bp,Lp⊗cp,Lp)∈Ap⊗Bp⊗Cp, for any 1≤p≤R, where {a1pa2},{bp,1, . . . , bp,Lp}and {cp,1, . . . , cp,Lp} are bases for Ap, Bp, Cp, respectively. Note that for a general set {ϕp ∈Ap⊗Bp⊗Cp : 1≤p≤R}.

Using Terracini’s Lemma 1.3 and 3.1, a general hyperplane tangent to J Sub1,L1,L1 CI⊗CJ ⊗CK

, . . . , Sub1,LR,LR CI⊗CJ⊗CK at [ϕ1+· · ·+ϕR] is of the form

H = X

1≤p≤R

pa1−a2)⊗

 X

1≤j6=k≤Lp

µp,j,kbp,j⊗cp,k0p,j,k bp,j⊗cp,j−bp,k⊗cp,k

It is straightforward to see that H is tangent to Sub1,Lp,Lp CI⊗CJ ⊗CK at ςp = (a1pa2)⊗X

j6=k

j −k

|j −k|µp,j,kbp,j ⊗cp,k0p,j,k(bp,j⊗cp,j+bp,k⊗cp,k), which is different from ϕ1,· · · , ϕR. So

J Sub1,L1,L1 CI⊗CJ ⊗CK

, . . . , Sub1,LR,LR CI⊗CJ⊗CK

is weakly defective.

4. Proof of Theorem 1.9 Proof. It is sufficient to prove the caseI = 2, J =K =PR

r=1Lr.

Let A, B and C be complex vector spaces of dimensions I, J, K, respectively. Split B =L

1≤q≤RBq and C =L

1≤r≤RCr, where for 1≤q, r ≤R, Bq and Cr are of dimensions Lq, Lr, respectively.

Choose a general set {ϕp ∈ Subd2,Lp,Lp(CI⊗CJ ⊗CK) : 1 ≤ p ≤ R}. Using Weierstrass normal forms (see Chapter 10 in [8] and Chapter IX in [11]) of tensors of multilinear rank (2, Lp, Lp), we can assume

ϕ1 =

L1

X

i=1

(a1ia2)⊗bi ⊗ci,

ϕ2 =

L1+L2

X

i=L1+1

(a1ia2)⊗bi⊗ci, ...

ϕR=

L1+···+LR

X

i=L1+···+LR−1+1

(a1ia2)⊗bi⊗ci,

(6)

where {a1, a2},{b1, . . . , bL1+···+LR},{c1, . . . , cL1+···+LR} are bases for A, B, C.

Consider

ψ1 =

L1−1

X

i=1

(a1ia2)⊗bi⊗ci+ (a1L1+1a2)⊗bL1+1⊗cL1+1

ψ2 =

L1+L2

X

i=L1+2

(a1ia2)⊗bi⊗ci+ (a1L1a2)⊗bL1 ⊗cL1.

It is obviously thatϕ1+· · ·+ϕR123· · ·+ϕR. So from Terracini’s Lemma 1.3, we knowJ Sub2,L1,L1 CI⊗CJ ⊗CK

, . . . , Sub2,LR,LR CI⊗CJ ⊗CK

is tangentially weakly

defective.

5. Acknowledgement

I am very grateful to Prof. Landsberg for many inspiring discussions, and unfailingly useful suggestions. I am also very grateful to Prof. Ottaviani to teach us the concept of tangential weak defectivity.

References

[1] E. Ballico and A. Bernardi, Symmetric tensor rank with a tangent vector: a generic uniqueness theorem, Proceedings of the American Mathematical Society, 140 (2012), pp. 3377–3384.

[2] L. Chiantini and G. Ottaviani,On generic identifiability of 3-tensors of small rank, SIAM Journal on Matrix Analysis and Applications, 33 (2012), pp. 1018–1037.

[3] L. Chiantini, G. Ottaviani, and N. Vannieuwenhoven, An algorithm for generic and low-rank specific identifiability of complex tensors, SIAM Journal on Matrix Analysis and Applications, 35 (2014), pp. 1265–1287.

[4] , On generic identifiability of symmetric tensors of subgeneric rank, arXiv preprint arX- iv:1504.00547, (2015).

[5] L. Chiantini and D. Sacchi, Segre functions in multiprojective spaces and tensor analysis. Trends Hist. Sci. to appear, 2016.

[6] D. Eisenbud and J. Harris,3264 & all that: Intersection theory in algebraic geometry, preparation, to appear, (2010).

[7] R. Hartshorne,Algebraic geometry, Graduate texts in mathematics, Springer, New York, 1977.

[8] J. M. Landsberg,Tensors: geometry and applications, American Mathematical Society, 2012.

[9] J. M. Landsberg and J. Weyman,On the ideals and singularities of secant varieties of segre varieties, Bulletin of the London Mathematical Society, (2007).

[10] G. OTTAVIANI, Identifiability and weak defectivity informal notes for lukecin school september 2-6, 2013, (2013).

[11] H. W. Turnbull and A. C. Aitken, An introduction to the theory of canonical matrices, Courier Corporation, 2004.

[12] M. Yang, On partial and generic uniqueness of block term tensor decompositions, ANNALI DEL- L’UNIVERSITA’DI FERRARA, 60 (2014), pp. 465–493.

Department of Computer Science/Mathematics, Southern Illinois University Carbon- dale, Carbondale, IL 62901, USA

E-mail address: yangmingmath@gmail.com

(Wen Liu)Department of Mathematics, Texas A&M University, College Station, TX 77843, USA

E-mail address: wenliu79@tamu.edu

Références

Documents relatifs

Summary : Hypersurfaces of a Euclidean space satisfying certain conditions on the projective.. curvature tensor are studied and local characterizations of certain types

In this informal survey article, we explain, mainly without proofs, what is known (or conjectured) in the following cases: cubic threefolds, Fano threefolds of degree 14 and index

1. Goxeter, Shephard and Artin groups. Local deformation theory of representations. The correspondence between projective reflections and their fixed points. Commuting and

Motivated by previous studies on the graph of algebraic structures, for any finite group G different from a cyclic group of prime power order, we define an undirected simple graph

We show that infinitesimal Torelli for the period of n-forms holds for abelian covers of algebraic varieties of dimension n &amp;#x3E; 2, under some explicit

V’Ve ask now in general the question that has been asked and answered for curves: what is the greatest possible geometric genus of an irreducible, nondegenerate variety

Then for any finite index normal subgroup H ⊂ π 1 (X y ) and character χ of the quotient, the identity component of the Zariski closure of the image of τ H,χ ◦ ρ is semisimple

Diante de tal exposição, podemos dizer que este livro contribui mais para rea- valiar posições acerca do autoconhecimento do que para apresentar uma nova tese que abarque todas