• 検索結果がありません。

Remarks on osculating linear spaces to projective varieties

N/A
N/A
Protected

Academic year: 2022

シェア "Remarks on osculating linear spaces to projective varieties"

Copied!
4
0
0

読み込み中.... (全文を見る)

全文

(1)

Bull Braz Math Soc, New Series 35(1), 123-126

© 2004, Sociedade Brasileira de Matemática

Remarks on osculating linear spaces to projective varieties

E. Ballico*

Abstract. LetX PN be an integraln-dimensional variety andm(X, P , i)(resp.

m(X, i)), 1 i N n+1, the Hermite invariants ofXmeasuring the osculating behaviour ofXatP (resp. at its general point). Here we provem(X, x)+m(X, y) m(X, x+y)andm(X, P , x)+m(X, y) m(X, P , x+y)for all integersx,y such thatx+y Nn+1, the casen=1 being known (M. Homma, A. Garcia and E.

Esteves).

Keywords: osculating spaces, order of contact.

Mathematical subject classification: 14N05.

1 Introduction

We work over an algebraically closed fieldK. LetXPN be an integral non- degenerate variety. Setn := dim(X). For any QX set m(X, Q,0) = 0 andm(X, Q,1) = 1. Fix an integeri with 2 ≤ iNn+1; if there is a linear spaceV with dim(V ) = i−1, QV and such thatVX contains a curveC withQC, setm(X, Q, i)= +∞; if there is no such linear space, let m(X, Q, i)be the supremum of the length of the connected component supported byQof all schemesXV, whereV is a linear space with dim(V ) = i−1 andQV. For all integersi with 0 ≤ iNn+1, let m(X, i)be the infimum of allm(X, Q, i)withQX. Thusm(X,0)=0 andm(X,1)=1. If n=1 the integersm(X, i), 0iN, are called the Hermite invariants of the curveX(see [6] or [3]). If char(K)=0 and dim(X)=1 we havem(X, i)=i for everyi; for curves in positive characteristic, see [3], Remark 1.5. For the osculating behaviour of surfaces in characteristic zero, see [7]. We always make

Received 5 September 2001.

*Partially supported by MIUR and GNSAGA of INdAM (Italy).

(2)

124 E. BALLICO

the convention that+∞ + +∞ = +∞anda + +∞ = +∞for all integers a≥0.

Remark 1. LetXPNbe an integraln-dimensional non-degenerate variety andQX. We have m(X, x)m(X, y)andm(X, Q, x)m(X, Q, y)if xyNn+1. If m(X, x) = +∞ (resp. m(X, Q, x) = +∞) and x < y, then m(X, x) < m(X, y)(resp. m(X, Q, x) < m(X, Q, y)). Fix an integeri with 2 ≤ iNn+1 and assume that there is no linear spaceV with dim(V )=i−1 and such thatVXcontains an irreducible curveCwith QC. Since the GrassmannianG(i, N+1)of all(i−1)-dimensional linear subspaces ofPN is algebraic, the value ofm(X, i)is computed by a maximum, not just a supremum, and in particularm(X, i) <+∞.

Here are our results.

Theorem 1. LetXPNbe an integraln-dimensional variety andx,yintegers with x ≥ 0, y ≥ 0 and x +yNn+1. Then m(X, x)+m(X, y)m(X, x+y).

Theorem 2. LetXPNbe an integraln-dimensional variety,PX, andx,y integers withx ≥0,y≥0andx+yNn+1. Thenm(X, P , x)+m(X, y)m(X, P , x+y).

Theorem 3. LetXPN be an integral n-dimensional variety,n ≥ 2, andC an integral curve contained inX. FixPC and a generalQC. Letx,y be integers withx ≥ 0,y ≥ 0andx+yNn+1. Thenm(X, P , x)+ m(X, Q, y)m(X, P , x+y).

In the case of curves Theorem 1 was proved using combinatorial techniques by M. Homma ([3], Th. 1). His proof was simplified by A. Garcia still using combinatorial techniques ([2]). E. Esteves gave a geometric proof of a more general inequality. In the same year M. Homma gave a very short proof of his original inequality and another proof of Esteves ’ inequality ([4]).

Remark 2. LetXPg1be the canonical model of a smooth curve of genus g. Assume that the Hermite invariants ofXare not classical in the sense of [6].

Such curves do exists in positive characteristics ([8], [5] or [3]). Theorem 1 gives a relation for the Hermite sequence of any Weierstrass point ofX.

Bull Braz Math Soc, Vol. 35, N. 1, 2004

(3)

OSCULATING SPACES 125

2 The proofs

Proof of Theorem 1. It is sufficient to prove the caseyx. The result is obvious ifx =0 orm(X, x)= +∞. The second part of Remark 1 gives that eitherm(X, y)= +∞orm(X, y) < m(X, y+1), proving the casex =1. Hence we may assumex≥2. Fix a general pair(P , Q)X×Xand linear subspaces V andW with dim(V )=x−1,PV,XV containing a zero-dimensional schemeA(P )of lengthm(X, x)withA(P )red = {P}, dim(W )=y−1,QW andXW containing a zero-dimensional schemeZ(Q)of lengthm(X, y)with Z(Q)red = {Q} (use the last part of Remark 1). The linear span VW ofVW has dimension at most x +y −1. We choose a linear space M with dim(M) = x+y −1 and VWM. There is a flat family of pairs {Qt, Wt}tT such thatT is an integral curve,oT, Qo = Q, Wo = W, for everytT, QtX,Wt is a linear subspace ofPN with dim(Wt) = y −1, WtXcontains a zero-dimensional schemeZ(Qt)such thatZ(Qt)red = {Qt} and length(Z(Qt))m(X, y), Z(Qo) = Z(Q), and there is aT with Qa = P. Indeed, since m(X, Q, y) = +∞, we may find such a flat family withm(X, Qt, y)=m(X, y)and length(Z(Qt))=m(X, y)for generaltT. By the properness of the GrassmannianG(x +y, N +1) of all(x +y−1)- dimensional linear subspaces ofPN, we may construct (taking if necessay a finite covering ofT) a flat family{Mt}tT of(x+y−1)-dimensional linear subspaces ofPN withMo =MandWtVMt for everyt. In particularPMa. By the properness of the Hilbert scheme Hilb(X)ofX, the schemeMaXcontains a zero-dimensional subscheme of lengthm(X, x)+m(X, y)withP as support;

here we useQt = Qa for generaltT and henceZ(Qt)A(P ) = ∅and length(Z(Qt)A(P )) = length(Z(Qt))+length(A(P )) for generaltT. Thus m(X, P , x +y)m(X, x)+m(X, y). Since P is general, we have m(X, x+y)=m(X, P , x+y), concluding the proof.

Proofs of Theorems 2 and 3. Just copy verbatim the proof of Theorem 1 with Pfixed and not general. For the proof of Theorem 3 take a flat family{Qt, Wt}tT

withQtCfor everyt.

References

[1] Esteves, E.: A geometric proof of an inequality of order sequences, Comm. Algebra 21(1) (1993), 231–238.

[2] Garcia, A.: Some arithmetic properties of order-sequences of algebraic curves, J.

Pure Applied Algebra85(1993), 259–269.

Bull Braz Math Soc, Vol. 35, N. 1, 2004

(4)

126 E. BALLICO

[3] Homma, M.: Linear systems on curves with no Weierstrass points, Bol. Soc. Brasil Mat. (N.S.)23(1992), 93–108.

[4] Homma, M.: On Esteves’ inequality of order sequences of curves, Comm. Algebra 21(1) (1993), 3685–3689.

[5] Laksov, D.: Weierstrass points on curves, in: Tableaux de Young et foncteurs de Schur en algebre et géometrie, Astérisque87/88(1981), 221–247.

[6] Laksov, D.: Wronskians and Plücker formulas for linear systems on curves, Ann.

Scient. Éc. Norm. Sup.17(1984), 565–579.

[7] Piene, R. and Tai, H.: A characterization of balanced rational normal scrolls in terms of their osculating spaces, in: Enumerative Geometry, Proc. Sitges 1987, pp. 215–224, Lect. Notes in Math.1436, Springer, 1990.

[8] Schimdt, F. K.: Die Wronskisch Determinante in belebigen differenzierbaren Funktionenkörper, Math. Z.45(1939), 62–74.

E. Ballico

Department of Mathematics University of Trento 38050 Povo (TN) ITALY

E-mail: [email protected]

Bull Braz Math Soc, Vol. 35, N. 1, 2004

参照

関連したドキュメント

(4) The author is informed by one of the referees that another proof of (1.2) is possi- ble by using the Helffer-Sjöstrand random walk representation; for the representation and

§ 6, the fibered power conjecture together with Lang’s conjectures should have very strong implications for counting rational points on varieties of general type, similar to the

This is a consequence of a more general result on interacting particle systems that shows that a stationary measure is ergodic if and only if the sigma algebra of sets invariant

GUE, eigenvalues of random matrices, Hermitian Brownian motion, non-colliding Brownian motions, Weyl chamber, queues in series, Burke’s theorem, reversibility, Pitman’s representa-

Abstract: By using the Euler-Maclaurin’s summation formula and the weight coefficient, a pair of new inequalities is given, which is a decomposition of Hilbert’s inequal- ity..

In this paper, we establish a one-parameter family of Harnack inequalities connecting the constrained trace Li-Yau differential Harnack inequality to the constrained trace

Makowski [3], who refers the reader to [2], where Blundon originally published this inequality, and where he actually proves more, namely that this is the best such inequality in

In this note, we give yet another proof and show that the G-A Mean inequality is merely a result of simple iteration of a well-known lemma.. The following