HILBERT FUNCTION AND HILBERT SERIES. PURE RESOLUTION ∗ This exposition uses essentially the approach to this subject given in the book of W. Bruns and J. Herzog “Cohen-Macaulay Rings”, Cambridge University Press 1993, 4.1. • Let M be the Z−module of all polynomials P ∈ Q [X] such that P (z) ∈ Z for all z ∈ Z. M is a free Z−module with basis (X + i) (X + i − 1) . . . (X − 1) X = i! If P ∈ M is a polynomial of degree d, then P = d X µ d−i (−1) ed−i i=0 µ ¶ X +i , i ∈ N. i ¶ X +i . i (1) Definition 1. We say that a numerical function f : Z → Z is of polynomial type (of degree d ) if there exists a polynomial P ∈ Q [X] (of degree d ) such that f (n) = P (n) for all n À 0. The zero polynomial has degree −1. Proposition 2. Let f : Z → Z be a function and d > 0 on integer. The following conditions are equivalent: a) f is of polynomial type of degree d. b) ∆ f (n) = f (n + 1) − f (n) is of polynomial type of degree d − 1. Proof. a) ⇒ b) This is easy. b) ⇒ a) We show this by induction on d. For d = 0 the implication is obvious. Let d > 0 and assume that ∆f (n) = f (n + 1) − f (n) is of polynomial type of degree d−1. Then there exist an integer no and a polynomial P ∈ Q[X] of degree d − 1 such that f (n + 1) − f (n) = P (n) for all n ≥ no. Then f (n + 1) = F (no ) + n X P (i) , for n ≥ no , i=no ∗ This lecture was held by Nicolae Radu: University of Bucharest, Faculty of Mathematics, Romania e-mail: [email protected] 1 2 Hilbert Function and Hilbert Series. Pure Resolution and this last sum is a polynomial function in n of degree d. Corollary 3. A function f : Z → Z is of polynomial type of degree d ≥ 0 if and only if ∆d f = ∆(∆d−1 f ) is a constant 6= 0 for n À 0. ∞ £ ¤ P Lemma 4. Let F = ai X i ∈ Z [[X]] X −1 and f : Z → Z , f (i) = ai ,for i=s all i ∈ Z. Then (i) n (1 − X) F = ∞ P ∆n f (i − n)X i . i=s (ii) For d > 0, the following conditions are equivalent: a) There exists a polynomial P ∈ Q[X] of degree d − 1 such that P (n) = f (i), n À 0. ¤ £ Q b) F = (1−X) , where Q ∈ Z X, X −1 and Q(1) = 0. d Proof. i) This follows easily by induction on n ii) a) ⇒ b) According to i), d (1 − X) F = ∞ X £ ¤ ∆d f (i − d)X i = Q ∈ Z X, X −1 , i=s because ∆d f (m) = 0, for m À 0. Suppose that Q(1) = 0. Then 0 = Q(1) = = d−1 ∞ X ∆d f (i − d) = i=s ∆ ∞ X (∆d−1 f (i + 1 − d) − ∆d−1 f (i − d)) = i=s (m) , for m À 0 and according to Corollary 1.3, ∆d−1 (m) is a constant 6= 0,a contradiction. b) ⇒ a) The relation d (1 − X) F = ∞ X ∆d f (i − d)X i = Q (4.1) i=s shows that ∆d f (m) = 0 for m À 0, thus ∆d−1 f (m + i) = ∆d−1 f (m) , for m À 0. The equality (1 − X) d−1 F = ∞ X i=s f (i − d + 1)X i = Q 1−X shows that ∆d−1 (m) 6= 0 for m À 0, thus f is of polynomial type of degree d − 1. 3 Hilbert Function and Hilbert Series. Pure Resolution Lemma 5. Under the assumption of Lemma 1.4, ii), if ¶ µ d−1 X X +i d−1−i , P = (−1) ed−1−i i i=0 then ei = Q(i) (1) i! (5.1) for i = 0, ..., d − 1. Proof. Let D =Q− d−1 i X (−1) i=0 Then H= £ ¤ i Q(i) (1) (1 − X) ∈ Z X, X −1 . i! d−1 i X (−1) d−i i! i=0 Q(i) (1) (1 − X) + D d (1 − X) . Since D(i) (1) = 0 for i = 0, 1, ..., d − 1, it follows that D is divisible by £ ¤ d D −1 (1 − X) . So (1−X) and for large n d ∈ Z X, X d−1 i X (−1) i=0 i! Qi (1) (1 − X) d−i = X P (n) X n , (5.2) n≥0 hence the equality holds for all n ∈ Z , because the coefficients of both series are polynomial functions in n. Since ¶ ∞ µ X 1 j−d−i−1 = Xj d−i d − i − 1 (1 − X) j=0 we have d−1 i X (−1) i=0 i! Q(i) (1) (1 − X) d−i ! Ãd−1 ∞ X X (−1) Q(i) (1) µj − d − i − 1¶ = Xj i! d − i − 1 j=0 i=0 Ãd−1 µ ¶! ∞ X X (−1)d−1−i j+i (d−1−i) = Q (1) Xj. i! i j=0 i=0 Now (5.2) and (5.3) imply (5.1). (5.3) 4 Hilbert Function and Hilbert Series. Pure Resolution Let R = ∞ L i=s Ri be a graded ring such that R0 is artinian local ring and R is finitely generated over R0 by elements of degree 1. ∞ L Let M = Mn by a finitely generated graded R-module. Then all Mn n=t have finite length, l(Mn ), the numerical function: H(M, ·) : Z → Z with H(M, n)P= l(Mn ) for all n ∈ Z is called the Hilbert function of M and HM = H(M, n)X n is called the Hilbert series of M. n∈Z Theorem 6. (Hilbert) Let M be a finitely generated graded R-module of dimension d. Then H(M, n) is of polynomial type of degree d − 1. Proof. Induction on the dimension d of M . If d = 0 M is artinian, thus H(M, n) = 0 for n large. For d > 0, let P ∈ Ass(M ), then P is a graded Rmodule. There exists a graded submodule N ⊂ M, such that N ≈ (R/P )(a). Since M is Noetherian there exist a chain 0 = N0 ⊂ N1 ⊂ . . . ⊂ Nn = M of graded submodules of M such that for each i we have Ni+1 /Ni ≈ (R/Pi )(ai ), where Pi is a graded prime ideal of R. Since the length of modules is additive on short exact sequences, it follows that H(M, n) = n X H((R/Pi (ai )), n). (6.1) i=1 Assuming that the assertion of theorem is proved for M = R/P, where P is a graded ideal with R of dim R/P = d, then from (6.1) the assertion follows for M, since the polynomials H((R/Pi (ai )), n) are zero or have positive leading coefficients, the degree of the sum is the maximum of dim(R/Pi ) − 1 and the maximum of dim(R/Pi ) is equal with d = dim M. Let be M = R/P , where P is a graded prime ideal of R and dim(R/P ) > 0. Then we may choose a homogeneous element x ∈ R/P , x 6= 0 of degree 1. The exact sequence 0 → (R/P )(−1) → R/P → R/(x, P ) → 0 gives the relation ∆H(R/P, n) = H(R/P, n + 1) − H(R/P, n) = H(R/(x, P ), n + 1). As dim R/(x, P ) = d−1, the induction hypothesis implies that ∆H(R/P, n) is of polynomial type of degree d-2, then Proposition 2 implies that H(R/P, n) is a polynomial type of degree d-1. 5 Hilbert Function and Hilbert Series. Pure Resolution Corollary 7. Let M 6= 0 be a finitely generated graded R- module of dimension d. Then QM HM (X) = , (1 − X)d where QM ∈ Z[X, X −1 ] with QM (1) 6= 0 and the coefficient ei of PM in representation (1) equals Q(i) (1) /i!. Definition 8. Let M be a graded R-module of dimension d. The unique polynomial PM such that H(M, n) = PM (n) for n À 0 is called the Hilbert polynomial of M and the multiplicity of M is defined to be e0 if d > 0 e(M ) = l(M ) if d = 0 , where e0 is the leading coefficient of PM in representation (1). Remark 9. The higher iterated Hilbert functions Hi (M, n), i ∈ N of a finitely generated graded R-module M are defined recursively as follows: X H0 (M, n) = H(M, n) and Hi (M, .n) = Hi−1 (M, j) for i > 0. j≤n Since ∆Hi (M, n) = Hi (M, n + 1) − Hi (M, n) = Hi−1 (M, n + 1), for i > 0, it follows from Proposition 1.3 and Theorem 6 that Hi (M, n) is of polynomial type of degree d + i − 1, where d = dim M. In the representation of Hi (M, n) as in (1), for n large, a0 = e(M ). Let A be a local noetherian ring with the maximal ideal N and let I be an ∞ L N -primary ideal. The graded ring R = I i /I I+1 is noetherian, R0 = R/I i=0 is a local artinian ring and R is generated over R0 by the elements of degree ∞ L 1. If E is a finitely generated A- module, then grI (E) = I i E/I I+1 E is a i=0 finitely generated graded R-module, H1 (E, n) = l(E/I n E) and the associated polynomial is called the Hilbert-Samuel polynomial of E with respect to I. Proposition 10. Let M be a finitely generated graded R-module of finite projective dimension and let M X 0→ R(−j)βpj → ... → R(−j)β0j → M → 0 j j 6 Hilbert Function and Hilbert Series. Pure Resolution be a graded free resolution of M. Then HM = SM HR , where SM = p X X βαj X j . α=0 j=1 In particular, for R = k[X1 , ..., Xn ], the polynomial ring over the field k , we have SM HM = . (1 − X)n Proof. The Hilbert function is additive on short exact sequences, so HM = p X i=0 (−1)i X βij HR(−j) j and HR(−j) = X j HR . If R = k[X1 , ..., Xn ] then HR = 1 (1−X)n . Corollary 11. Let R = k[X1 , ..., Xn ] , where k is a field and M be a finitely generated graded R-module of dimension d. Then: (i) SM = (1 − X)n−d QM . (i) (ii) n − d = inf{i | SM (1) 6= 0}. (n−d−i) (iii) ei = (−1)n−d SM (1)/(n − d + i)! . Proof. It follows from Proposition 10 and Corollary 7. Definition 12. Let R = k[X1 , ..., Xn ] , where k is a field and I ⊂ R a graded ideal. We say that R/I has a pure resolution of type (d1 , ..., dp ) if its minimal free graded resolution has the form 0 → R(−dp )βp → ... → R(−d1 )β1 → R → R/I → 0 . Then d1 < d2 < ... < dp . Theorem 13. Let R = k[X1 , ..., Xn ] be a polynomial ring over a field k and I be an ideal of R such that R/I has a pure resolution of type (d1 , ..., dp ) where p = dim R − dim(R/I) (for example if R/I is Cohen-Macaulay). Then: Q (i) βi = (−1)i+1 dj /(dj − di ). j6=i (ii) e(R/I) = (1/p!) p Q i=1 di . 7 Hilbert Function and Hilbert Series. Pure Resolution Proof. From Proposition 10 and Corollary 11, ii) we obtain the system of linear equations p P (−1)i βi = −1 i=1 (j) SR/I (1) = 0, j = 0, 1, ..., p − 1 , because d0 = 0, β0 = 1. Applying elementary row operations which do not affect the solution of this system, we obtain the system: p P (−1)i βi = −1 i=1 p P i=1 p P i=1 (−1)i βi di = 0 (−1)i βi d2i = 0 ................................... p P i=1 (−1)i βi dp−1 = 0, i whose solution is stated in (i). (ii) By Corollary 1.11,iii, (p) e(R/I) = (−1)p SR/I (1)/p! = = 1 p! p Q i=1 di ( P p−1 Q ¡ ¢ P (−1)p+1 βi dpi = (di −dj ) j=1 Q (di −dj ) ). j6=i It remains to show that the sum in this expression equals 1. For this, take the rational complex function p−1 Q f (z) = (z − j) j=1 p Q (z − dj ) j=1 which has simple poles at worst in the points d1 , ..., dp and the sum of all residues of this function at all points including ∞ is zero. Then p P i=1 p−1 Q (di −j) ( Q (di −dj ) ) = −Res∞ f (z) = Res0 f (1/z) = z2 j=1 j6=i = Res0 Ã 1 z p−1 Q j=1 (1 − jz) p Q j=1 ! −1 (1 − dj z) = 1. 8 Hilbert Function and Hilbert Series. Pure Resolution
© Copyright 2026 Paperzz