WILSON CHEUNG

Department of Mathematics
State University of New York
Geneseo, NY 14454
PATRICK X. RAULT
Department of Mathematics
State University of New York
Geneseo, NY 14454
1. Introduction
ON UNIFORM BOUNDS FOR RATIONAL POINTS ON
QUADRATIC RATIONAL CURVES AND THIN SETS
Abstract. We show that for any > 0 the number of rational points of height
less than B on the image of a quadratic map from P
1
to P
n
(n 1) is bounded
above by CB/|R|
1/(4n(n+1))
+C

|R|

, where the point is that the constant C is

independent of the choice of the map. R is the resultant of the map and is a
nonzero integer. In the special case of quadratic plane curves we prove a bound
of CB/|R|
1/72
+ 4 which improves on a result of Browning and Heath-Brown
by establishing an inverse dependence on the resultant. Heath-Brown proved
that for any > 0 the number of rational points of height less than B on a
degree d plane curve is O
,d
(B
2/d+
); it is known that Heath-Browns theorem is
sharp apart from the . Our results shed light on the open question of whether
Heath-Browns theorem is true with = 0. This generalizes a previous result
on indenite quadratic forms by the second author, to any quadratic forms with
the condition that the resultant is odd and coprime to the discriminant. In the
process, we generalize another result by the second author on lattice points in
plane regions bounded by ellipses and hyperbolas.
The primary aim of this paper is to prove the following upper bound for the
number of points of bounded height on a quadratic curve.
Theorem 1.1. Let V be a quadratic rational curve in P
n
(with n 2) or thin set
in P
1
(n = 1), dened over Q, with parametrization (f
0
: f
1
: ... : f
n
) : P
1
P
n
.
Let R =

0i<jn
R(f
i
, f
j
) and D =
n

i=0
D(f
i
). Assume further that gcd(R, D) = 1.
Then for any > 0 there exist positive constants C and C

for which the number

of points on V (Q) with height less than B is at most
CB
|R|
1/(4n(n+1))
+C

|R|

.
Furthermore, if n = 2 then there exists a positive constant K for which there are
at most
KB
|R|
1/72
+ 4 points Q V (Q) with H(Q) < B.
Note, as discussed in [6], that the number of points Q P
n
(Q) of height at most
B is on the order of B
n+1
. The bound of order B given here relies on the fact that
2000 Mathematics Subject Classication. Primary 11G50; Secondary 11P21.
Key words and phrases. lattice, height, uniform, rational point, rational curve.
1
2 UNIFORM BOUNDS FOR RATIONAL POINTS ON QUADRATIC RATIONAL CURVES
we are counting points on a curve (or thin set) in P
n
; this ts into a conjectural
bound of order B
2/d
for degree d curves.
For a xed quadratic curve V , we know that the set of rational points V (Q)
is either empty or innite. In fact, the number of points of height at most B
is known to be asymptotically cB for some factor c = c(V ) dependent on the
curve. The content of our theorem is twofold. First, without loss of generality
the resultant is a nonzero integer, so for plane curves we have a uniform upper
bound of KB + 4; for a xed curve our result is not best possible, so the point
here is that our uniform bound is independent of the particular curve in question.
Second, we give an inverse dependence on the coecients dening the curve in the
form of an explicit formula involving resultants. Furthermore, our theorem yields
a corresponding result for thin sets in P
1
.
Theorem 1.1 is a direct generalization of the main theorem appearing in [6]. It
improves, for certain curves, on a previous uniform bound of O

_
B
2/d+
_
, due to
Heath-Brown [4], for arbitrary degree d plane curves. In [2] and [3] this uniform
bound is decreased to cB
2/d
+k (for some positive uniform constants c and k) for all
curves except possibly higher degree rational curves. In [6] we give an example of a
curve of each degree d which asymptotically have cB
2/d
points (for some constant
c), to demonstrate that Heath-Browns theorem is sharp except possibly for the
. However, none of these bounds include the extra inverse dependence on the
(nonzero integer) resultant found in theorem 1.1, which gives a bound of the form
cB
2/d
+ k as a corollary. It is an open question whether a uniform bound of the
form cB
2/d
+ k holds for arbitrary plane curves. for a further description of the
theorem, such as an explicit discussion of the constants appearing in the theorem,
see the introductory section of [6].
As described in detail in [6], the methods of [2] and [3] are not applicable to
higher degree rational curves. Our key stepping stone, proposition 3.1, is entirely
insensitive to the degree. Indeed, it is written so that a strengthening of the
following theorem 1.2 to all higher degrees would immediately generalize theorem
1.1 and resolve the aforementioned open question.
Theorem 1.2. Let B, C be positive real numbers, and let f, g be integral binary
quadratic forms. If R(f, g)D(f)D(g) = 0, then the area of the region of all points
(u, v) for which |f(u, v)| < B and |g(u, v)| < C is at most 16
(BC)
1/2
|R(f,g)|
1/4
.
Fix r Z, and let L denote a sublattice of Z
2
for which |Z
2
: L| r. Then the
number of primitive points (u, v) in L such that |f(u, v)| < Br and |g(u, v)| < Cr
is at most 32
(BC)
1/2
|R(f,g)|
1/4
+ 8.
For example, in the case when L = Z
2
and r = 1, the content of this theorem
mirrors that of theorem 1.1 above. First, because R(f, g) is a nonzero integer we
give a uniform upper bound of 32(BC)
1/2
+8, independent of the particular forms
f and g; indeed, for xed curves, one could establish a sharp asymptotic formula.
UNIFORM BOUNDS FOR RATIONAL POINTS ON QUADRATIC RATIONAL CURVES 3
Second, we give an inverse dependence on the coecients of f and g by way of an
explicit formula involving the resultant.
Note that the theorem does not provide a formula for the area of the region;
indeed we do not claim a lattice point bound of the form 2(Area + 4), except in
the case of convex regions. Exact formulas for these areas are computed in [1];
each formula is a sum of terms of the form
2
_
|D(f)|
arccos
_
D(f) P(f, g)
_
4|R(f, g)|
_
+
2
_
|D(g)|
arccos
_
D(g) P(f, g)
_
4|R(f, g)|
_
and
2
_
D(f)
arccosh

D(f) P(f, g)
_
4|R(f, g)|

+
2
_
D(g)
arccosh

D(g) P(f, g)
_
4|R(f, g)|

where D denotes the discriminant, R the resultant, and P the polarization of the
discriminant.
Note that this does not directly give a bound in terms of the resultant because
the resultant and discriminant can indeed change independently. In the following
chart we demonstrate this with two families of examples depending on a parameter
. In the rst example we have an innite family of polynomials with bounded
discriminant and arbitrarily large resultant; the opposite holds for the second
example.
Table 1. Examples of independence of resultant and discriminant.
f g R(f, g) D(f) D(g)
xy y
2
x
2
xy
2
(1 +
2
) 1 1
x
2
y
2
( + 1)x
2
y
2
1 4 4( + 1)
The above are also examples of innite families of curves which satisfy our
condition that gcd(R(f, g), D(f)D(g)) = 1.
2. Lattice points in bounded plane regions
We will begin by proving the key ingredient of all of our results, theorem 1.2,
while introducing the necessary denitions. Let f and g be rational binary qua-
dratic forms. Let B and C be positive real numbers. Consider the region
R
(f,g)
(B, C) :=
_
(u, v) R
2
: |f(u, v)| < B, |g(u, v)| < C
_
.
We give an example of the region R in the following graph, where the lightly-
shaded region denotes the region within either |f(x, y)| < B or |g(x, y)| < C,
whereas the darkly-shaded region denotes R
(f,g)
(B, C).
4 UNIFORM BOUNDS FOR RATIONAL POINTS ON QUADRATIC RATIONAL CURVES
20 10 0 10 20
20
10
0
10
20
x
y
Figure 1. Demonstrating the darkly shaded region R
(f,g)
(B, C).
We call a nonzero lattice point (u, v) Z
2
is primitive if the coordinate entries
u and v are relatively prime. Let A(R) denote the area of region R and let
N
(f,g)
(B, C) denote the number of primitive integral lattice points (u, v) in that
same region.
In bounding the number of lattice points, and later the number of rational points,
the following invariants of binary forms will be instrumental in our calculations.
These are dened in more detail in [5] and [6].
Denition 2.1.
I. Let h(x, y) and k(x, y) be binary linear forms. We dene their resultant
R(h, k) to be the determinant of their coecient matrix.
II. Let f be a binary degree d form which factors linearly over Q as f =

d
i=1
f
i
. We dene its discriminant D(f) to be the following:
D(f) =
d

i=1
d

j=1j=i
R(f
i
, f
j
) =
d1

i=1
d

j=i+1
R(f
i
, f
j
)
2
.
III. Let f and g (resp.) be binary forms of degrees d and e (resp.) that factor
linearly over Q as f =

d
i=1
f
i
and g =

e
j=1
g
j
. We dene their resultant
R(f, g) to be the following symmetric polynomial of resultants of the linear
forms:
R(f, g) =
d

i=1
e

j=1
R(f
i
, g
j
).
Note that when f and g have integer coecients, the factors of R(f, g)
are permuted by the Galois group; hence R(f, g) is an integer. Note further
that for any scalars and we have R(f, g) =
e

d
R(f, g).
IV. Let = (f
0
, f
1
, ..., f
n
) be an (n + 1)-tuple of binary integral forms. We
dene the resultant R = R() and discriminant D = D() to be the
UNIFORM BOUNDS FOR RATIONAL POINTS ON QUADRATIC RATIONAL CURVES 5
following:
R =

0i<jn
R(f
i
, f
j
)
D =

0in
D(f
i
).
Note that this last denition concerns the rational points in the next section.
2.1. Indenite quadratic forms. We begin by noting that the case of intersec-
tions of hyperbolas was previously proven by the second author in [5]. We recall
the full theorem here:
Theorem 2.1. Let f and g be indenite binary quadratic forms such that R(f, g) =
0. Let R(f, g) denote the resultant of f and g and let B and C be positive real
numbers. Then:
(1) The area of the region bounded by the curves |f(x, y)| = B and |g(x, y)| = C
is at most 16
(BC)
1/2
|R(f,g)|
1/4
.
(2) The number of primitive integral lattice points (u, v) satisfying |f(u, v)| < B
and |g(u, v)| < C is at most 32
(BC)
1/2
|R(f,g)|
1/4
+ 8.
As is traditional in the study of lattice points, we make use of strong results for
convex regions, such as the following lemma which was also proven in [5].
Lemma 2.1. Let P be a closed convex plane region symmetric about the origin
with area A and containing L primitive integral lattice points. Then L 2A + 2.
Indeed, the essence of the proof of theorem 2.1 was to bound the region inside
4 parallelograms of combined area at most 16
(BC)
1/2
|R(f,g)|
1/4
, and to apply lemma 2.1.
2.2. Integer points in the unit circle. We will rst concern ourselves the case
when f(x, y) = x
2
+y
2
and g(x, y) = ax
2
+by
2
, for any nonzero real a and b, and
bound the region of all (u, v) for which f(u, v) 1 and |g(u, v)| 1. Specically,
in this section we will provide several convex sets which cover the set of lattice
points and bound their area above by a function on the order of 1/|R|
1/4
. As we
shall see, this seemingly trivial case is quite critical. The resultant can be easily
calculated to be R = (a b)
2
, and we assume throughout this paper that R = 0.
We will prove by cases.
Case 1: Assume that a and b both have absolute value at most 1. In this case
|g(x, y)| = 1 is the equation for either an ellipse or a symmetric pair of hyperbolas.
Then the x- and y-intercepts of the second curve are greater than or equal to 1,
so there are no intersection points and the region is bounded only by x
2
+y
2
= 1,
which has area . Since |a| and |b| are bounded by 1, we have 0 = R = (ab)
2
< 4,
so that the area is at most 2
1/2
/|R|
1/4
, as desired. Furthermore, lemma 2.1 tells
UNIFORM BOUNDS FOR RATIONAL POINTS ON QUADRATIC RATIONAL CURVES 13
3. Mapping to rational points of bounded height
As theorem 1.1 is a generalization of the main theorem of [6], we will give a
general argument which encompasses both the prior result and our new result.
Indeed, the following gives a range of generality which will be applicable for any
results about higher degree curves which may be done in the future; our main
theorem is a simple corollary.
Proposition 3.1. Fix a degree d and let T be a collection of degree d forms in
Z[x, y] with the following property:
There exist positive constants (dependent only on d) K, K

, and ,
such that for any positive B and C, for any f, g T satisfying
R(f, g)D(f)D(g) ,= 0, for any positive integer r and any lattice
L with [Z
2
: L[ r, then the number of primitive lattice points
(u, v) L satisfying [f(u, v)[ < Br and [g(u, v)[ < Cr is at most
K(BC)
1/d
/[R(f, g)[

+ K

.
Let V be the image of a map = (f
0
: f
1
: ... : f
n
) : P
1
P
n
, where each
component function of the map is in T . Let R =

0i<jn
R(f
i
, f
j
) and D =
n

i=0
D(f
i
). Assume further that gcd(R, D) = 1. Then for any > 0 there exist
positive constants (dependent on d) C, C

, and

for which the number of points

on V (Q) with height less than B is at most
CB
2/d
[R[

1
+ C

[R[

.
Furthermore, if n = 2 then there exist positive constants K and (depending only
on d) for which there are at most
KB
2/d
|R|

+ d
2
points Q V (Q) with H(Q) < B.
We will spend much of this section proving this lemma. In the following deni-
tion we give a name to the quantity bounded by theorem 1.1.
Denition 3.1. Let V be a variety in P
n
.
For any rational point Q = (x
0
: x
1
: ... : x
n
) V (Q) (represented by coprime
integer coordinates), we dene the height of Q to be H(Q) = max
0in
[x
i
[.
For any positive real number B, we write N
V
(B) for the height-counting function
of the number of rational points on V of height at most B:
N
V
(B) = [Q V (Q) : H(Q) < B[ .
Recall that we are considering a rational curve V = (P
1
). As detailed in [6],
the points of bounded height are all counted by functions of the form
14 UNIFORM BOUNDS FOR RATIONAL POINTS ON QUADRATIC RATIONAL CURVES
(1) N
r

(B) :=
1
2

_
gcd(u, v) = 1;
(u, v) Z
2
: max
0in
[f
i
(u, v)[ < rB;
gcd
_
f
0
(u, v), ..., f
n
(u, v)
_
= r
_

and the height-counting function is bounded via the following sum:

(2) N
V
(B)

r|R
N
r

(B).
We now make use of the hypothesis of proposition 3.1. Let T be a set of degree
d polynomials which satises this hypothesis, and assume that f
0
, f
1
, . . . , f
n
T
for which RD ,= 0. Let r be a positive integer. Then there exist positive constants
(dependent only on d) K, K

, and , such that for any positive B and for any

i ,= j the region bounded by [f
i
(x, y)[ = Br and [f
j
(x, y)[ = Br has its number
of primitive lattice points bounded above by KB
2/d
/[R(f, g)[

+K

. When r = 1,
this becomes
(3) (i, j) N
1
(f
i
,f
j
)
(B) < KB
2/d
/[R(f, g)[

+ K

.
Notice that equation 3 involves the counting function N
r

whereas equation 2
involves N
1

in the special case when n = 1. We will now improve on this in the

case when r > 1. The following lemma is the key tool in reconciling this dierence.
Lemma 3.1. Let f be an integral binary degree d form. Let p be a prime for
which p D(f) and let k be a positive integer. Then there exist at most d lattices
L
1
, L
2
, . . . , L
e
(with 0 e d) in Z
2
for which:
Every (u, v) Z
2
such that p
k
[ f(u, v) lies on one of the lattices.
For each j, we have [Z
2
: L
j
[ p
k
.
Proof. First we will prove that f(x, y) has at most d roots in P
2
(Z/p
k
Z). Since
Z/pZ is a eld, the polynomial must have at most d roots in it. Fix one such root,
without loss of generality satisfying f(x, 1) = 0. Specically, x in Z for which
f(, 1) 0 mod p. Let denote the reduction mod p map. Since D(f) ,= 0, we
know that is not a double root, so f

() ,= 0. By Hensels Lemma, lifts to a

unique value
1
in Z/p
2
Z for which f(
1
) = 0. Similarly, since p D(f) we can
always lift up to a next level. Hence in P
2
(Z/p
k
Z) the polynomial f(x, y) has at
most d roots.
Let (u
1
, v
1
), (u
2
, v
2
), . . . , (u
e
, v
e
) denote the corresponding elements of Z
2
with
nonnegative values less than p
k
, where e is the number of roots (so e d). For
each j, we dene L
j
to be the lattice of all points which reduce to (u
j
, v
j
): L
j
:=
(u
j
, v
j
), (p
k
, 0), (0, p
k
). Note that a generating set of two vectors can be found
for each of these. Furthermore, every (u, v) for which p
k
[ f(u, v) lies on one of L
j
.
UNIFORM BOUNDS FOR RATIONAL POINTS ON QUADRATIC RATIONAL CURVES 15
It remains to determine the index of these lattices. We will prove that [Z
2
: L
j
[
p
k
using induction. Let L
k
j
denote a lattice found above, for the prime power p
k
.
Since the elements of (u, v) L
k
j
satisfy p
k
[ f(u, v), they are roots of f(u, v) in
Z/p

Z, for 1 k; this naturally denes larger lattices L

j
, for 1 < k.
In the case that k = 1, we have Z
2
L
1
j
pZ
2
. The claim follows from the fact
that

Z
2
: L
1
j

must be a proper divisor of [Z

2
: pZ
2
[ = p
2
.
Now assume that for 1 < k we have [Z
2
: L

j
[ p

. We have Z
2
L
k1
j

L
k
j
p
k
Z
2
. By the induction hypothesis, we have

Z
2
: L
k1
j

p
k1
. Hence

Z
2
: L
k
j

is greater than p
k1
and is a divisor of p
2k
. The conclusion follows.

Consider two distinct polynomials f and g from the set f

0
, f
1
, ..., f
n
. Let
r [ R(f, g), with prime factorization r =
(r)

i=1
p
k
i
i
; here (r) represents the number
of prime factors of r.
Let (u, v) Z
2
such that gcd(u, v) = 1 and r [ gcd(f(u, v), g(u, v)). Under
these conditions, the second author proved in lemma 2.1 of [6] that r [ R(f, g).
Since for each i we have p
k
i
i
[ gcd(f(u, v), g(u, v)), the same lemma gives us that
p
k
i
i
[ R(f, g). Hence, with our assumption that gcd (R(f, g), D(f)D(g)) = 1 in
proposition 3.1, we have p
i
D(f).
We use lemma 3.1 to dene, for each i, the lattices L
i
j
computed for the prime
power p
k
i
i
. Then, for each i, we have (u, v)
e

j=1
L
i
j
. Let o denote the set of (r)-
dimensional vectors whose entries are in 1, 2, 3, . . . , e; note that [o[ = e
(r)

d
(r)
. Then (u, v)

sS
(r)

i=1
L
i
s(i)
.
For any s o, the set L
s
:=
(r)

i=1
L
i
s(i)
is an intersection of lattices of coprime
indices [Z
2
: L
i
s(i)
[ p
k
i
i
. Hence the set L
s
is a lattice of index which is at least the
product of these indices: [Z
2
: L
s
[ r.
We conclude that the set of all primitive points (u, v) such that r [ gcd(f(u, v), g(u, v))
is contained in a union of at most d
(r)
lattices, each of index at least r.
Recall the hypotheses of proposition 3.1 above, which states that the number of
primitive points (u, v) in L such that [f(u, v)[ < Br and [g(u, v)[ < Cr is at most
KB
2/d
/[R(f, g)[

+ K

.
Since there may be up to d
(r)
such lattices L, we see that there are at most
d
(r)
_
KB
2/d
/[R(f, g)[

+ K

_
points (u, v) such that r [ gcd(f(u, v), g(u, v)).
Noting that there is a factor of 2 dierence between the number of primitive
lattice points and the number of projective equivalence classes, we arrive at the
following bound:
16 UNIFORM BOUNDS FOR RATIONAL POINTS ON QUADRATIC RATIONAL CURVES
N
r
(f,g)
(B) d
(r)

_
KB
2/d
2[R(f, g)[

+
K

2
_
.
Combining this with equation 3 yields N
(f:g)(P
1
)
(B)

r|R(f,g)
d
(r)

_
KB
2/d
2|R(f,g)|

+
K

2
_
.
Let
0
(r) denote the number of divisors of r, and recall our earlier denition that
(r) is the number of prime factors of r. Then 2
(r)

0
(r), with equality when r is
squarefree. Then we have (r) log
2

0
(r), and (r) log
2
d (log
2

0
(r))(log
2
d).
Hence d
(r)

0
(r)
log
2
d
.
Thus

r|R(f,g)
d
(r)

r|R(f,g)

0
(r)
log
2
d
by substitution, which is in turn less than

r|R(f,g)

0
(R(f, g))
log
2
d
. Since there are exactly
0
(R(f, g)) terms being added, the
sum is
0
(R(f, g))
1+log
2
d
. Regardless, this sum is at most
0
(R(f, g))
d
.
Hence:
(4) N
(f:g)(P
1
)
(B) (
0
(R(f, g)))
d

_
KB
2/d
2[R(f, g)[

+
K

2
_
.
As in [6], we again use the fact that for any > 0, there exists a constant h()
for which
0
(n) h()n

; for our case we will need H() = h

d
(/d). We apply this
bound separately to each part of equation 3 to yield, for any > 0 and > 0,
(i, j) N
(f
i
:f
j
)(P
1
)
(B)
KH()B
2/d
2[R(f
i
, f
j
)[

+
K

H()
2

R(f
i
, f
j
)

.
Forcing < /2 yields the more aesthetic bound for the height-counting function
of a thin set, with a uniform constant of KH(/2)/2 in the main term. This bound
is the special case of proposition 3.1 when n = 1:
(5) (i, j)N
(f
i
:f
j
)(P
1
)
(B)
KH(/2)B
2/d
2[R(f
i
, f
j
)[
/2
+
K

H()
2

R(f
i
, f
j
)

.
3.1. General case of rational curves in P
n
. We proceed to prove the general
case for any positive integer n. For any integers i and j with 0 i < j n,
the thin set (f
i
: f
j
)(P
1
) is a projection of V = (f
0
: f
1
: ... : f
n
)(P
1
); thus
N
V
(B) N
(f
i
:f
j
)(P
1
)
(B). Hence for any > 0 equation 4 gives us an upper bound
for the height-counting function for V :
(6) (i, j)N
V
(B)
KH(/2)B
2/d
2[R(f
i
, f
j
)[
/2
+
K

H()
2

R(f
i
, f
j
)

.
Recall from denition 2.1 that R =

0i<jn
R(f
i
, f
j
). Thus we have that [R[
max
0i<jn
[R(f
i
, f
j
)[
n(n+1)/2
. It follows that, for specic i and j,
1
|R(f
i
,f
j
)|

1
|R|
2/(n(n+1))
.
This inequality gives a uniform bound for the rst term in equation 5.
UNIFORM BOUNDS FOR RATIONAL POINTS ON QUADRATIC RATIONAL CURVES 17
The product denition for R also reveals that, for any i and j, [R(f
i
, f
j
)[ divides
R and is therefore at most [R[. This inequality gives a uniform bound for the
second term in equation 5.
Combining these bounds, we have
(7) N
V
(B)
KH(/2)B
2/d
2[R[
/(n(n+1))
+
K

H()
2
[R[

.
This concludes the proof of the main part of proposition 3.1.
3.2. Plane curves. In the special case of plane curves, where n = 2 in theorem
1.1, the following result of Heath-Brown (lemma 3 of [2]) allows us to improve the
bound signicantly.
Lemma 3.2. Let F Z[X
0
, X
1
, ..., X
n
] be a primitive form of degree d 2 and
let V : F = 0 denote the zero locus of F. Let [[F[[ denote the maximum modulus
of the coecients of F. Then either [[F[[ = O(B

), where = d
_
n + d
n
_
, or
else there exists a form G Z[X
0
, X
1
, ..., X
n
] of degree d, not proportional to F,
such that every x = (x
0
, x
1
, ..., x
n
) counted by N
V
(B) also satises the equation
G(x) = 0.
Combined with Bezouts theorem, the second case implies that N
V
(B) d
2
. In
our case where n = 2 we compute = d(d + 1)(d + 2)/2. Thus the lemma reveals
that either N
V
(B) d
2
or [[F[[ = O(B
d(d+1)(d+2)/2
).
Given a map = (f
0
: f
1
: f
2
) we nd the implicit representation for V = (P
1
)
using an alternate version of a three-polynomial resultant R

(X : Y : Z; f
0
: f
1
:
f
2
), involving the variables X, Y, Z, in a method generalized from [7] to degree d.
Specically, for each i, let f
i
(x, y) = a
0,d
x
d
+a
0,d1
x
d1
y +. . . a
0,0
y
d
. We will take
the determinant of a matrix using the following fundamental building block M,
which is a 3 (2d + 1) matrix:
M :=
_
_
X 0 0 a
0,d
a
0,d1
a
0,0
Y 0 0 a
1,d
a
1,d1
a
1,0
Z 0 0 a
2,d
a
2,d1
a
2,0
_
_
.
Furthermore, we let v
1
= [x, y]
(d1)
= [x
d1
, x
d2
y, , y
d1
]
T
, let v
2
=
[x, y]
(2d1)
, and let v be the concatenation of v
1
and v
2
; this has 3d terms.
For 0 d 1, we dene a matrix M

gotten by concatenating columns

of zeroes to the beginning of M and d 1 columns of zeroes to the end of M.
For any , we see that [X, Y, Z] = [f
0
(x, y), f
1
(x, y), f
2
(x, y)] if an only if

M

v = 0.
We thus dene R

(X : Y : Z; f
0
: f
1
: f
2
) to be the determinant of the 3d 3d
matrix formed by concatenating these matrices M