2 The First Form : XaYb + Y cZd
Let K be a field of characteristic p > 0 and
S = K [X1, . . . , Xr, Y1, . . . , Ys, Z1, . . . , Zt] .
In this section we shall determine the Hilbert-Kunz function of the hypersurface of the following form :
f := XaYb+ YcZd
where Xa = X1a1. . . Xrar, Yb = Y1b1. . . Ysbs, Yc = Y1c1. . . Yscs, Zd = Z1d1. . . Ztdt, and r ≥ 1. Let q = pn, J = j | bj > cj , and set R = S/ < f >. Then 0 ≤ |J| := m ≤ s, and w.l.o.g., we assume that b1 > c1, . . . , bm > cm, bm+1 ≤ cm+1, . . . , bs≤ cs. We shall determine the assignment
HKR(q) := dimK
S < X1q, . . . , Xrq, Y1q, . . . , Ysq, Z1q, . . . , Ztq, f >
.
Let f + X[q] be the ideal of S generated by all Xiq’s, Yjq’s, Zkq’s, and f . Fix a term order on S, and denote by in f + X[q] the initial ideal of f + X[q]. Then by Lemma 1.1, we get that HKR(q) is equal to dimK
S/in f + X[q]
. By making use of Gr¨obner basis, we understand which monomials one has to add to fill the gap between in f +X[q]
and the ideal in(f )+X[q]. Throughout this section, it is not restrictive to assume that a1 ≥ a2 ≥ · · · ≥ ar > 0 , b1− c1 ≥ b2− c2 ≥ · · · ≥ bm − cm > 0 , cm+1− bm+1 ≥ cm+2 − bm+2 ≥ · · · ≥ cs− bs ≥ 0, and d1 ≥ d2 ≥ · · · ≥ dt> 0. Let u be the maximum of the integers a1 , b1− c1 , cm+1 − bm+1, and d1; that is, u is the greatest integer among all ai’s, (bj − cj)’s, (ch − bh)’s, and dk’s. We also denote by [y] the greatest integer less than or equal to y, and Sin(x) the elementary symmetric polynomial of degree i in n indeterminates x = (x1, . . . , xn). Let Iq be the ideal f + X[q] and define (v)+= max {0, v}.
In order to make it more easy to determine the Hilbert-Kunz function of R, we shall prove the following lemma.
Lemma 2.1. Let S = K [X1, . . . , Xr, Y1, . . . , Ys], r ≥ 1, s ≥ 1, and the ai, bj, and cj are all positive integers with b1 − c1 ≥ b2 − c2 ≥ · · · ≥ bm − cm > 0, cm+1− bm+1 ≥ cm+2 − bm+2 ≥
· · · ≥ cs− bs≥ 0. We denote by G the ideal generated by
X1q, . . . , Xrq, Y1q, . . . , Ysq, X1[q−αa1]+Ye+α(c−b)+, . . . , Xr[q−αar]+Ye+α(c−b)+, Y1[q−α(b1−c1)−c1]+Ye+α(c−b)+, . . . , Ym[q−α(bm−cm)−cm]+Ye+α(c−b)+, and XaYb,
where α is a positive integer, e = (e1, . . . , es), e1 = c1, . . . , em = cm, em+1 = bm+1, . . . , es = bs, (c − b)+= (0, . . . , 0, cm+1− bm+1, . . . , cs− bs). Then the dimension of S/G is equal to
qr+s−
r
Y
i=1
(q − ai)
s
Y
j=1
(q − bj) − qr
m
Y
j=1
(q − cj)
s
Y
h=m+1
[q − α(ch− bh) − bh]+
+
r
Y
i=1
(q − ai)
m
Y
j=1
(q − bj)
s
Y
h=m+1
[q − α(ch− bh) − bh]+
+
r
Y
i=1
(q − αai)+
m
Y
j=1
[q − α(bj− cj) − cj ]+
s
Y
h=m+1
[q − α(ch− bh) − bh]+
−
r
Y
i=1
[q − (α + 1)ai ]+
m
Y
j=1
[q − (α + 1)(bj− cj) − cj ]+
s
Y
h=m+1
[q − α(ch− bh) − bh]+.
Proof : If [q − αai]+ = 0 for some i or [q − α(bj − cj) − cj ]+ = 0 for some j, then G is generated by
X1q, . . . , Xrq, Y1q, . . . , Ysq, Ye+α(c−b)+, and XaYb. Hence,
dimK S/G
= qr+s−
r
Y
i=1
(q − ai)
s
Y
j=1
(q − bj) − qr
m
Y
j=1
(q − cj)
s
Y
h=m+1
[q − α(ch− bh) − bh]+
+
r
Y
i=1
(q − ai)
m
Y
j=1
(q − bj)
s
Y
h=m+1
[q − α(ch− bh) − bh]+ .
From now on, we assume q − αai > 0 for each i and q − α(bj− cj) − cj > 0 for each j. Let lα be the minimum of
q − α(bj− cj) − 1 cj
,
q − 1
bh+ α(ch− bh)
j = 1, . . . , m, h = m + 1, . . . , s
.
Then we have q − α(bj0 − cj0) ≤ (lα+ 1)cj0 for some j0 with 1 ≤ j0 ≤ m or
q ≤ (lα+ 1)(bh0 + α(ch0 − bh0)) for some h0 with m + 1 ≤ h0 ≤ s, and q − α(bj− cj) − lαcj ≥ 1 for each j and q − lα(bh+ α(ch− bh)) ≥ 1 for each h.
We consider the ideals Gβ = G : Yβ[e+α(c−b)+], for β = 0, 1, 2, . . . , lα + 1. Since G0 = G, Glα+1 = S, and Gβ+1 = Gβ : Ye+α(c−b)+, we have the exact sequence of K-modules :
0 −→ S/Gβ+1 Ye+α(c−b)+−→ S/Gβ −→ S / < Gβ, Ye+α(c−b)+ > −→ 0.
It follows that
dimK S/G = dimK S/G0 =
lα
X
β=0
dimK
S / < Gβ, Ye+α(c−b)+ >
.
We determine dimK
S / < Gβ, Ye+α(c−b)+ >
as follows : For β = 0, the ideal < G0, Ye+α(c−b)+ > is generated by
X1q, . . . , Xrq, Y1q, . . . , Ysq, Ye+α(c−b)+, and XaYb. Hence,
dimK
S / < G0, Ye+α(c−b)+ >
= qr+s−
r
Y
i=1
(q − ai)
s
Y
j=1
(q − bj) − qr
m
Y
j=1
(q − cj)
s
Y
h=m+1
[q − α(ch− bh) − bh]+
+
r
Y
i=1
(q − ai)
m
Y
j=1
(q − bj)
s
Y
h=m+1
[q − α(ch− bh) − bh]+ .
For 1 ≤ β ≤ lα, the ideal Gβ = G : Yβ[e+α(c−b)+] is generated by
X1q−αa1, . . . , Xrq−αar, Y1q−α(b1−c1)−βc1, . . . , Ymq−α(bm−cm)−βcm, Ym+1q−β[bm+1+α(cm+1−bm+1)], . . . , Ysq−β[bs+α(cs−bs)], and XaY1(b1−βc1)+· · · Ym(bm−βcm)+Ym+1[bm+1−β(bm+1+α(cm+1−bm+1))]+· · · Ys[bs−β(bs+α(cs−bs))]+ .
Hence, the ideal < Gβ, Ye+α(c−b)+ > is generated by
X1q−αa1, . . . , Xrq−αar, Y1q−α(b1−c1)−βc1, . . . , Ymq−α(bm−cm)−βcm, Ym+1q−β[bm+1+α(cm+1−bm+1)], . . . , Ysq−β[bs+α(cs−bs)], XaY1(b1−βc1)+· · · Ym(bm−βcm)+Ym+1[bm+1−β(bm+1+α(cm+1−bm+1))]+· · ·
Ys[bs−β(bs+α(cs−bs))]+, and Ye+α(c−b)+ .
Thus, dimK
S / < Gβ, Ye+α(c−b)+ >
=
r
Y
i=1
(q − αai)
m
Y
j=1
[q − α(bj − cj) − βcj ]+
s
Y
h=m+1
[q − β(bh+ α(ch− bh))]+−
r
Y
i=1
[q − (α + 1)ai]+
m
Y
j=1
q − α(bj − cj) − βcj − (bj − βcj)+
+ s
Y
h=m+1
[q − β(bh+ α(ch− bh))]+
−
r
Y
i=1
(q − αai)
m
Y
j=1
[q − α(bj− cj) − (β + 1)cj ]+
s
Y
h=m+1
[q − (β + 1)(bh+ α(ch− bh))]+
+
r
Y
i=1
[q − (α + 1)ai]+
m
Y
j=1
[q − α(bj − cj) − βcj − ujβ]+
s
Y
h=m+1
[q − (β + 1)(bh+ α(ch− bh))]+
=
r
Y
i=1
(q − αai) × ( m
Y
j=1
[q − α(bj − cj) − βcj ]+
s
Y
h=m+1
[q − β (bh+ α (ch − bh))]+
−
m
Y
j=1
[q − α(bj− cj) − (β + 1)cj ]+
s
Y
h=m+1
[q − (β + 1)(bh+ α(ch− bh))]+ )
−
r
Y
i=1
[q − (α + 1)ai]+× ( m
Y
j=1
q − α(bj− cj) − βcj− (bj − βcj)+
+ s
Y
h=m+1
[q − β(bh + α(ch−
bh))]+−
m
Y
j=1
[q − α(bj− cj) − βcj− ujβ]+
s
Y
h=m+1
[q − (β + 1)(bh+ α(ch− bh))]+ )
,
where ujβ = max cj , [bj− βcj ]+ . Now, we have
dimK S/G
= qr+s−
r
Y
i=1
(q − ai)
s
Y
j=1
(q − bj) − qr
m
Y
j=1
(q − cj)
s
Y
h=m+1
[q − α(ch− bh) − bh]+
+
r
Y
i=1
(q − ai)
m
Y
j=1
(q − bj)
s
Y
h=m+1
[q − α(ch− bh) − bh]+
+
r
Y
i=1
(q − αai) × ( l
α
X
β=1
" m Y
j=1
[q − α(bj − cj) − βcj ]+
s
Y
h=m+1
[q − β(bh+ α(ch− bh))]+
−
m
Y
j=1
[q − α(bj− cj) − (β + 1)cj ]+
s
Y
h=m+1
[q − (β + 1)(bh+ α(ch− bh))]+
#)
−
r
Y
i=1
[q − (α + 1)ai]+ × ( lα
X
β=1
" m Y
j=1
[q − α(bj − cj) − βcj− (bj− βcj)+]+
s
Y
h=m+1
[q − β(bh+
α(ch− bh))]+−
m
Y
j=1
[q − α(bj − cj) − βcj − ujβ]+
s
Y
h=m+1
[q − (β + 1)(bh+ α(ch− bh))]+
#) .
Let (∗) be the term
lα
X
β=1
" m Y
j=1
[q − α(bj− cj) − βcj ]+
s
Y
h=m+1
[q − β(bh+ α(ch− bh))]+
−
m
Y
j=1
[q − α(bj − cj) − (β + 1)cj ]+
s
Y
h=m+1
[q − (β + 1)(bh+ α(ch− bh))]+
# . Since q − α(bj0 − cj0) ≤ (lα+ 1)cj0 for some j0 with 1 ≤ j0 ≤ m or
q ≤ (lα+ 1) bh0+ α(ch0 − bh0)
for some h0 with m + 1 ≤ h0 ≤ s, (∗) is equal to
m
Y
j=1
[q − α(bj− cj) − cj ]+
s
Y
h=m+1
[q − α(ch− bh) − bh]+.
Let (∗∗) be the term
lα
X
β=1
" m Y
j=1
q − α(bj− cj) − βcj− (bj − βcj)+
+ s
Y
h=m+1
[q − β(bh+ α(ch− bh))]+
−
m
Y
j=1
[q − α(bj − cj) − βcj − ujβ ]+
s
Y
h=m+1
[q − (β + 1)(bh+ α(ch− bh))]+
# . Since
m
Y
j=1
[q − α(bj− cj) − βcj− ujβ ]+
s
Y
h=m+1
[q − (β + 1)(bh+ α(ch− bh))]+
=
m
Y
j=1
q − α(bj − cj) − (β + 1)cj − (bj− (β + 1)cj)+
+ s
Y
h=m+1
[q − (β + 1)(bh+ α(ch− bh))]+,
where β = 1, 2, . . . , lα− 1, the term (∗∗) is equal to
m
Y
j=1
[q − (α + 1)(bj − cj) − cj ]+
s
Y
h=m+1
[q − α(ch − bh) − bh]+
−
m
Y
j=1
[q − α(bj − cj) − lαcj − ujlα ]+
s
Y
h=m+1
[q − (lα+ 1)(bh+ α(ch− bh))]+.
Since q − α(bj0 − cj0) ≤ (lα+ 1)cj0 or q ≤ (lα+ 1) bh0+ α(ch0 − bh0), we have
m
Y
j=1
[q − α(bj − cj) − lαcj− ujlα ]+ = 0 or
s
Y
h=m+1
[q − (lα+ 1)(bh+ α(ch− bh))]+= 0 . Thus, (∗∗) is equal to
m
Y
j=1
[q − (α + 1)(bj− cj) − cj ]+
s
Y
h=m+1
[q − α(ch− bh) − bh]+.
So,
dimK S/G
= qr+s−
r
Y
i=1
(q − ai)
s
Y
j=1
(q − bj) − qr
m
Y
j=1
(q − cj)
s
Y
h=m+1
[q − α(ch− bh) − bh]+
+
r
Y
i=1
(q − ai)
m
Y
j=1
(q − bj)
s
Y
h=m+1
[q − α(ch − bh) − bh]+
+
r
Y
i=1
(q − αai)+
m
Y
j=1
[q − α(bj− cj) − cj ]+
s
Y
h=m+1
[q − α(ch− bh) − bh]+
−
r
Y
i=1
[q − (α + 1) ai ]+
m
Y
j=1
[q − (α + 1)(bj− cj) − cj ]+
s
Y
h=m+1
[q − α(ch− bh) − bh]+.
Since u is the maximum of the integers among all ai’s, (bj − cj)’s, (ch− bh)’s, and dk’s, we have
q−v
u := min (
hq−1 ai
i ,hq−c
j−1 bj−cj
i ,h
q−bh−1 ch−bh
i ,h
q−1 dk
i
1 ≤ i ≤ r, bj − cj > 0, ch− bh > 0 1 ≤ j ≤ m, m + 1 ≤ h ≤ s, 1 ≤ k ≤ t
)
for q 0, where v = 1 or 1 + cj for some j or 1 + bh for some h.
Let lu be the integerq−v
u , and be the remainder of q − v divided by u. Then lu = q−v−u and one has q − luai > 0, q − lu(bj− cj) − cj > 0, q − lu(ch− bh) − bh > 0, and q − ludk > 0 for all i, j, h, and k. On the other hand, by the definition of lu, at least one of [q − (lu+ 1)ai]+’s, [q − (lu+ 1)(bj − cj) − cj ]+’s, [q − (lu+ 1)(ch− bh) − bh]+’s, and [q − (lu+ 1)dk]+’s must be zero.
Proposition 2.2. Let f := XaYb+ YcZd. Then HKR(q) = qr+s+t− qt
r
Y
i=1
(q − ai)
s
Y
j=1
(q − bj)
− qr
m
Y
j=1
(q − cj)×
( l
u
X
α=1
" s Y
h=m+1
[q − α(ch− bh) − bh]
#
×
" t Y
k=1
(q − αdk) −
t
Y
k=1
[q − (α + 1)dk]+
#)
+
r
Y
i=1
(q − ai)
m
Y
j=1
(q − bj)×
( l
u
X
α=1
" s Y
h=m+1
[q − α(ch− bh) − bh]
#
×
" t Y
k=1
(q − αdk) −
t
Y
k=1
[q − (α + 1)dk]+
#)
+
lu
X
α=1
( r Y
i=1
(q − αai)
m
Y
j=1
[q − α(bj − cj) − cj ]
s
Y
h=m+1
[q − α(ch− bh) − bh]
−
r
Y
i=1
[q − (α + 1)ai]+
m
Y
j=1
[q − (α + 1)(bj− cj) − cj]+
s
Y
h=m+1
[q − α(ch− bh) − bh] )
× ( t
Y
k=1
(q − αdk) −
t
Y
k=1
[q − (α + 1)dk]+ )
, where lu is the integer q−v
u , and 0 ≤ m ≤ s .
Proof : Let u be the maximum of the integers a1, b1− c1, cm+1− bm+1, and d1. Let < be the lexicographic order on S and define
ej = cj for j = 1, . . . , m and eh = bh for h = m + 1, . . . , s.
Then XaYb is bigger than YcZd and Ye = Y1e1. . . Yses = Y1c1. . . YmcmYm+1bm+1. . . Ysbs. We determine a Gr¨obner basis of the ideal
Iq =< X1q, . . . , Xrq, Y1q, . . . , Ysq, Z1q, . . . , Ztq, f >,
by means of Buchberger’s algorithm (Algorithm 1.9). By this algorithm, the elements X1q, . . . , Xrq, Y1q, . . . , Ysq, Z1q, . . . , Ztq, Xi(q−δai)+Ye+δ(c−b)+Zδd, i = 1, . . . , r, δ = 1, . . . , l, Yj[q−δ(bj−cj)−cj]+Ye+δ(c−b)+Zδd, j = 1, . . . , m, δ = 1, . . . , l, and XaYb+ YcZd,
form a Gr¨obner basis of the ideal Iq, where l =h
q−1 d1
i
. Thus, the ideal in(Iq) is generated by X1q, . . . , Xrq, Y1q, . . . , Ysq, Z1q, . . . , Ztq, Xi(q−δai)+Ye+δ(c−b)+Zδd, i = 1, . . . , r, δ = 1, . . . , l, Yj[q−δ(bj−cj)−cj]+Ye+δ(c−b)+Zδd, j = 1, . . . , m, δ = 1, . . . , l, and XaYb.
Now we have to compute the dimension of S/in(Iq). In order to do this, we consider the ideals Kα = in(Iq) : Zαd for α = 0, 1, . . . , l + 1, where Zαd = Z1αd1. . . Ztαdt. Since K0 = in(Iq), Kl+1 = S, and Kα+1 = Kα : Zd, we have the exact sequence of K-modules :
0 −→ S/Kα+1 −→ S/KZd α −→ S / < Kα, Zd> −→ 0 . It follows that
dimK S/in(Iq) = dimK S/K0 =
l
X
α=0
dimK
S / < Kα, Zd> .
We compute dimK
S / < Kα, Zd>
as follows : For α = 0, the ideal < K0, Zd> is generated by
X1q, . . . , Xrq, Y1q, . . . , Ysq, Z1q, . . . , Ztq, XaYb, and Zd. Let S1 = K [X1, . . . , Xr, Y1, . . . , Ys], and S2 = K [Z1, . . . , Zt]. Then
dimK
S / < K0, Zd>
= dimK
S1 / < X1q, . . . , Xrq, Y1q, . . . , Ysq, XaYb >
× dimK
S2 / < Z1q, . . . , Ztq, Zd>
=
"
qr+s−
r
Y
i=1
(q − ai)
s
Y
j=1
(q − bj)
#
×
"
qt−
t
Y
k=1
(q − dk)
#
= qr+s+t− qt
r
Y
i=1
(q − ai)
s
Y
j=1
(q − bj) − qr+s
t
Y
k=1
(q − dk) +
r
Y
i=1
(q − ai)
s
Y
j=1
(q − bj)
t
Y
k=1
(q − dk).
For 1 ≤ α ≤ l, the ideal Kα = in(Iq) : Zαd is generated by
X1q, . . . , Xrq, Y1q, . . . , Ysq, Z1q−αd1, . . . , Ztq−αdt, Xi(q−δai)+Ye+δ(c−b)+Z(δ−α)d, i = 1, . . . , r, δ = 1, . . . , l, Yj[q−δ(bj−cj)−cj]+Ye+δ(c−b)+Z(δ−α)d, j = 1, . . . , m, δ = 1, . . . , l, and XaYb.
Hence, the ideal < Kα, Zd> is generated by
X1q, . . . , Xrq, Y1q, . . . , Ysq, Z1q−αd1, . . . , Ztq−αdt, X1(q−αa1)+Ye+α(c−b)+, . . . , Xr(q−αar)+Ye+α(c−b)+, Y1[q−α(b1−c1)−c1]+Ye+α(c−b)+, . . . , Ym[q−α(bm−cm)−cm]+Ye+α(c−b)+, XaYb, and Zd.