Elliptic Curves and Group Law
Elliptic Curves and Group Law
Elliptic Curves and Group Law
Discriminant
Bezouts theorem
Definition of the group law
Associativity of the group law
Computing with the group law
Singular curves
In this chapter, we introduce the notion of elliptic curves. We will define the group law using the
chord and tangent process, which dates back the ancient Greeks.
2.1
(2.1)
where a1 , a2 , a3 , a4 , a6 K.
The homogenization of the curve E in (2.1) is given by
Y 2 Z + a1 XY Z + a3Y Z 2 = X 3 + a2 X 2 Z + a4 XZ 2 + a6 Z 3 .
(2.2)
The only point at infinity on this curve is [0 : 1 : 0], we denote this point by from now on. We
will see that this point is the neutral element in the group structure on E.
Example 2.1.0.3
18
In practice, it is often desirable to simplify Equation (2.1). This is possible provided that
char(K) 6= 2, 3. Indeed, when char(K) 6= 2, we can complete the square in (2.1). This amounts
to making the change of coordinates
(
x0 = x
y0 = y + 21 (a1 x + a3 ).
to obtain a curve with a medium Weierstrass equation
y2 = x3 + a02 x2 + a04 x + a06 .
(2.3)
(2.4)
We now give a criterion for a short Weierstrass cubic curve E to be an elliptic curve.
2.2
Discriminant
Definition 2.2 Let f (x) = a0 + a1 x + . . . + am xm , g(x) = b0 + b1 x + . . . + bn xn K[x] be
a0 a1 am1 am
0
0
0
0 a0 am2 am1 am
0
0
..
..
.
a
0
m
b0 b1
bn1 bn
0 0
0 b0
bn2 bn1 bn . . . 0
..
..
.
.
0
b0
b1
bn
1 0
1
0 1
0
1
R( f , g) = 0 0
1 1 0
0 1 1
0
1
0
1
0
0
0
1 = 5.
0
1
2.2 Discriminant
19
Lemma 2.2.1 Let f , g K[x] be two polynomials of degree m and n respectively. Then f and
g have a common factor which is non-constant if and only if there exist non zero polynomials
, K[x] such that deg < m, deg < n and f = g.
Proof. () If f and g have a common factor h which is non constant, then f = h and g = h;
so f = g.
() Suppose that f = g; then every irreducible factor of g divides either f or . However,
since deg < n, one of those irreducible factors must divide f .
Theorem 2.2.2 Let f , g K[x] be two polynomials of degree m and n respectively, given by
m
f (x) = ai xi , g(x) =
i=0
b jx j.
j=0
G(X,Y, Z) = B0 Z n + B1 Z n1 + + Bn ,
where Ai , B j K[X,Y ] are homogeneous of degree i and j respectively. Similarly to Definition 2.2, we define the resultant of F and G with respect to Z. This is either a polynomial
RF,G (X,Y ) K[X,Y ] or 0 by definition. Moreover, if it is different from 0 then this polynomial is
homogenous of degree at most mn: this follows because F and G are homogeneous polynomials.
Hence, an analagous of Theorem 2.2.2 holds, namely:
Theorem 2.2.3 Let K be an infinite field, and F, G K[X,Y, Z] be two homogeneous polyno-
mials of degree m and n respectively. If F(0, 0, 1) G(0, 0, 1) 6= 0, then the resultant RF,G (X,Y )
of F and G with respect to Z is either 0 or a homogeneous polynomial of degree mn (similarly
for RF,G (X, Z) and RF,G (Y, Z)). The polynomials F and G have a common factor if and only if
at least one between RF,G (X,Y ), RF,G (X, Z) and RF,G (Y, Z) is equal to 0.
The condition in the theorem ensures the right power of the variables are occurring. In general,
the resultant RF,G (X,Y ) of F and G, two homogeneous polynomials of degree m and n, with
respect to Z is either 0 or a homogeneous polynomial of degree at most mn: the degree will be
mn st where s,t are the smallest indexes for which Ai and B j , as defined above, are non-zero.
Example 2.2.3.1 The resultant of the polynomials F(X,Y, Z) = X 2 +Y 2 Z 2 and G(X,Y, Z) =
Y 2 Z X 3 + XZ 2 with respect to Z is
2
X +Y 2
0
1
0
2
2
0
X +Y
0 1
6
R(X,Y ) =
3
2
= Y .
X
Y
X
0
0
X 3
Y2 X
20
Definition 2.3 Let f be a polynomial of degree n in K[x], with leading coefficient an . The
discriminant of f is defined by
0
f = (1)n(n1)/2 a1
n R( f , f ),
Lemma 2.2.4 Let f be a polynomial of degree n in K[x], with leading coefficient an , and
write
f (x) = an (x ei ), with ei K.
i=1
Then, we have
f = a2n2
n
(ei e j )2 .
1i< jn
given by
0
1
0 = 4a3 c + a2 b2 + 18abc 4b3 27c2
0
3
21
g
0
x = f (x) = 0
E is singular at P
g = 2y = 0
y
f (x) = f 0 (x) = 0.
Let E : y2 = f (x) be a medium Weierstrass cubic, and e1 , e2 , e3 the roots of f (x). By Lemma 2.2.5
and Proposition 2.2.6, E is smooth if and only if E 6= 0, or equivalently, if and only if e1 , e2 , e3
are distinct. In other words, E is an elliptic curve if and only if f (x) has no repeated root.
Let us assume K R, then we can consider the R points of E. If E > 0 then f (x) has 3 real
roots and the graph of E(R) has two components. Meanwhile, if If E < 0 then f (x) has 1
real roots and the graph of E(R) has only one component. See Figures 2.1(b) and 2.1(a) for
illustrations of the set E(R).
e1
e2
e3
(a) > 0 e1 , e2 , e3 R.
e1
2.3
Bezouts theorem
Let C : f (x, y) = 0 be a plane curve, where f K[x, y] is a polynomial of degree m. Let
C0 : y = h(x) be another plane curve, where h K[x] is a polynomial of degree n (in one variable).
To find the intersection of C and C0 , we substitute h(x) for y, and solve the equation f (x, h(x)) = 0.
However, the curve C0 cannot always be given in this form. The notion of resultant, introduced
in the previous section, allows one to determine all intersection points even when the polynomial
defining the curves are not polynomials in one variable: we will see this procedure at the end of
this section.
Note that we cannot always expect to obtain all the intersection points unless K is an algebraically
closed field.
22
Theorem 2.3.1 Weak Bezout Theorem. Let K be an infinite field. Let F, G K[X,Y, Z] be
Then the set CF (K) CG (K) is finite, and contains at most mn points.
Proof. Let Pi = [xi : yi : zi ] CF (K) CG (K) for 1 i k be distinct points. Consider all the
lines through the pairs (Pi , Pj ), with 1 i < j k. Since K is infinite, we can find a point P0
which doesnt belong to any of these lines. Furthermore, by a change of coordinates, we can
assume that P0 = [0 : 0 : 1]. The fact that the points P0 , Pi , Pj are not co-linear for 1 i < j k,
implies that the points [xi : yi ] and [x j : y j ] are distinct in P1 (K).
Now consider the polynomials fi (Z) = F(xi , yi , Z) and gi = G(xi , yi , Z). Since Pi belongs to
CF (K) CG (K), we have fi (zi ) = gi (zi ) = 0. Therefore, by Theorem 2.2.2, fi (Z) and gi (Z) have
a common factor, which is non-constant. This means that RF,G (X,Y ) must vanish at [xi : yi ], i.e.
at Pi . Since, this is a homogeneous polynomial of degree at most mn, we must have k mn.
Let Pi = [xi : yi : zi ] CF (K) CG (K), for 1 i k, then RF,G (X,Y ) vanishes at all [xi : yi ]. If
K is algebraically closed, this means that
k
Note that there is a bijection between the points Pi and and the linear factors of RF,G (X,Y ).
We define the multiplicity of Pi to be mi . Analogously, the multiplicity I(P; CF (K), CG (K))
of P CF (K) CG (K) is that of the corresponding linear factor in RF,G (X,Y ). This gives
immediately the following theorem.
Theorem 2.3.2 Strong Bezout Theorem. Let K be an algebraically closed field. Let
I(P; CF , CG ) = mn.
PCF CG
The definition of the multiplicity I(P; CF , CG ) given above clearly depends on the choice
of coordinates. One can show that this is in fact not the case.
23
U 3 + 3U 2W 2UW 2 +V 2W 2VW 2 +W 3 = 0
2.4
24
Example 2.4.0.2 Let K = Q, and E : y2 = x3 2. The set of Q-rational points E(Q) contains
To define the group structure, we need to work with the K-rational points, i.e., the set
2
Recall that, if L be a line in P2 (K), then Bezouts Theorem implies that L E has three points
P, Q and R counted with multiplicity. For P, Q E(K), we denote the third point of intersection
of the line through P and Q with E by P Q.
We are now ready to define the group structure on E(K).
P Q := (P Q) .
P Q
P Q
Theorem 2.4.1 Let E be an elliptic curve defined over a field K. Then, E(K) is an abelian
group under the operation , with identity element (= [0 : 1 : 0]). In other words, we have
(i) P Q = Q P P, Q E(K) (commutativity);
(ii) P = P P E(K) (i.e., is the identity element);
(iii) Let P0 = P . Then P P0 = (i.e., the opposite of P is P = P );
(iv) P (Q R) = (P Q) R, P, Q, R E(K) (associativity).
Proof. The first statement is an immediate consequence of the definition. Indeed, the third point
of intersection of the line through P and Q is R = P Q = Q P, and the line through R and
intersects E at P Q = Q P.
To prove (ii) and (iii), let P E(K). If P = then, the (tangent) line at infinity Z = 0 intersects
E at three times by Bezouts Theorem; so
= = .
Otherwise, the line through P and is a vertical line, whose third point of intersection with
E(K) is P0 = P . Then, by definition, we have P P0 = ; so
P P0 = (P P0 ) = = , and
P = (P ) = P0 = P.
The last statement (iv) is harder and we will come back to it later, after some preliminaries.
25
through them. The conic C is unique if no 4 of these points are on the same line.
Proof. Let V be the set of all homogeneous polynomials in K[X,Y, Z] of degree 2. Then every
element of V is of the form
F(X,Y, Z) = v0 X 2 + v1 XY + v2 XZ + v3Y 2 + v4Y Z + v5 Z 2 ,
where v0 , . . . , v5 K. This is a vector space since
K, F V F V ,
F1 , F2 V F1 + F2 V .
The dimension of V is 6.
Let C be a conic in P2 (K). Then, by definition, C is given by an element F V . Note that
C is also the zero locus of F for all K . Let W be the subset of V consisting of the
polynomials corresponding to all conics passing through P1 , . . . , P5 . The conic C passes through
the point P = [x : y : z] if and only if
F(x, y, z) = v0 x2 + v1 xy + v2 xz + v3 y2 + v4 yz + v5 z2 = 0.
This is linear equation in (v0 , . . . , v5 ). Therefore, the elements in W are the solutions to a
homogeneous linear system of 5 equations in 6 variables. Hence, it is a vector subspace of
dimension at least 1. This means that, there is at least one conic passing through P1 , . . . , P5 .
We are now going to show that, if no 4 of the points P1 , . . . , P5 are on the same line, then
dim W = 1. Assume that dim W > 1. Then, there are two polynomials F1 , F2 , which are linear
independent, such that the conics
Ci (K) = Ci := [x : y : z] P2 (K) : Fi (x, y, z) = 0 , i = 1, 2,
go through P1 , . . . , P5 . So #C1 C2 5 > 4. So by Theorem 2.3.2, F1 and F2 have a common
factor which is non-constant. Since they are linearly independent (of degree 2 each), this common
factor must be a linear factor. In other words, C1 C2 contains a line, which contradicts our
assumption on the points P1 , . . . , P5 . So dim W = 1 as required.
Lemma 2.4.3 Let P1 , . . . , P8 P2 (K) be distinct. Suppose that no 4 of them are colinear; and
no 7 of them lie on the same conic. Then, the subspace of homogeneous cubic polynomials
which vanish at P1 , . . . , P8 has dimension 2.
Proof. Let V be the space of all homogeneous polynomials of degree 3 in K[X,Y, Z]. Then,
every element F V is of the form
F = v0 X 3 + v1 X 2Y + v2 XY 2 + v3Y 3 + v4 X 2 Z + v5 XZ 2 + v6 Z 3 + v7Y 2 Z + v8Y Z 2 + v9 XY Z,
10
where (v0 , . . . , v9 ) K . As in the proof of Lemma 2.4.2, we see that dim V = 10.
26
27
L1 E : P, Q, P Q
L2 E : Q R, Q R,
L3 E : P Q, R, (P Q) R
M1 E : P Q, , P Q
M2 E : Q, R, Q R
M3 E : Q R, P, P (Q R)
M3
M2
M1
L1
rP
rQ
rP Q
L2
rQ R
rQ R
L3
rP (Q R)
(P Q) R
PQ
We can equivalently define the group law by saying that P Q R = 0(= ) if and only if
P, Q, R are the three points of intersection of a line L with E (counted with multiplicities).
The extreme case is when L is a line of inflection. In that case L E is one point P with
multiplicity 3, which means 3P = 0.
28
2.4.2
Let E be an elliptic curve over K, and let L be a line defined over K. Let P1 , P2 and P3 be the
intersection points of E and L over K. If any two of the Pi for i = 1, 2, 3 is K-rational, so is
the third.
Proof. Let us assume for simplicity that E is given by a short Weierstrass equation (the general
case is analogous and it involves more calculations). Let E be given by y2 = f (x) = x3 + ax + b.
Let L be a vertical line L : x = c, with c K by hypothese. If no point (c, y) belongs to E, then
the intersection is given by the point with multiplicity
p 3 and is always apK-rational point.
Therefore, let us assume that L E consists of (c, f (c)) and , where f (c) K, since
by
poins are K-rational. Thep
statement then is equivalent to say that
p hypotheses at least two p
f (c) K if and only if f (c) K. Notice that if f (c) = 0 then L is the tangent line to E
at (c, 0).
Let L : y = mx + c with m, c K. The intersection L E is given by
(mx + c)2 = x3 + ax + b.
By moving all terms to the same side, expanding and then factorizing, we get
x3 m2 x2 + (a 2mc)x + (b c2 ) = (x x1 )(x x2 )(x x3 ) = 0,
in K
where x1 , x2 , x3 K are the roots of the cubic. Since two intersection poins are K-rational then
two between x1 , x2 and x3 are in K. By equating the terms of degree 2, we get x1 + x2 + x3 = m2 .
Hence, since the line L is defined over K, we have that x1 , x2 , x3 K.
We now give a more explicit description of the group law on E(K), but only for a curve E given
by a short Weierstrass equation.
Proposition 2.4.6 Let E : y2 = x3 + ax + b be an elliptic curve given by a short Weierstrass
y2
If x1 = x2 and y1 = y2 6= 0 then set m = 2y1 1 ; otherwise, set m = xy11 x
.
2
2
Let x3 = m x1 x2 and y3 = y1 + m(x3 x1 ), then P1 P2 = (x3 , y3 ).
Proof. We note that (1) and (2) are just a restatement of Theorem 2.4.1 (ii) and (iii). So we
only need to prove (3). In that case, let L : y = mx + c be the line through P1 and P2 . If P1 = P2 ,
3x2 +a
then L is the tangent line at P1 with m = 2y1 1 and c = y1 mx1 . Otherwise, L is the line with
1
slope m = yx22 y
x1 and x-intercept c = y1 mx1 = y2 mx2 . The intersection L E is then given by
(mx + c)2 = x3 + ax + b.
By moving all terms to the same side, expanding and then factorizing, we get
x3 m2 x2 + (a 2mc)x + (b c2 ) = (x x1 )(x x2 )(x x3 ) = 0,
29
where x1 , x2 , x3 K are the roots of the cubic, counted with multiplicity. By equating the terms
of degree 2, we get x1 + x2 + x3 = m2 , and the points (x1 , y1 ), (x2 , y2 ) and (x3 , y3 ). We note
that if xi K then yi = mxi + c K and the intersection point (xi , yi ) is defined over K. We
also note that, if two of the roots x1 , x2 , x3 are defined over K, then so is the third one since
x1 + x2 + x3 = m2 K.
Example 2.4.6.1 Let E : y2 = x3 + 73, and P = (2, 9), Q = (3, 10).
23
2
2
(b) The slope of the tangent line at P is m = 2yPP = 3(2)
2(9) = 3 ; so its equation is y = 3 x + 3 .
Let R = (xR , yR ) be the third point of intersection of this line with E. Then, we have
32
23
143
2
2xP + xR = m2 . So xR = ( 23 )2 2(2) = 32
9 , and yR = 3 ( 9 ) + 3 = 27 . Hence
143
32
2P = R = (xR , yR ) = (xR , yR ) = ( 9 , 27 ).
y y
(c) The slope of the line through P and Q is m = xQQ xPP = 109
32 = 1, and the equation of the line
is y = x + 7. Let R = (xR , yR ) be the third point of intersection of this line with E. Then,
we have xP + xQ + xR = m2 . So xR = (1)2 2 3 = 4, and yR = xR + 7 = 4 + 7 = 3.
Hence P Q = R = (4, 3).
Proof. By definition, the identity element E(K 0 ); also P = (x, y) E(K 0 ) implies that
P = (x, y a1 x a3 ) E(K 0 ). So we only need to show that
P, Q E(K 0 ) P Q E(K 0 ).
If P, Q E(K 0 ), then the slope of the line through P and Q belongs to K 0 , and (generalizations
of) the formulas in Proposition 2.4.6 show that the coordinates of P Q are in K 0 .
2.5
Singular curves
A Weierstrass cubic y2 = x3 +ax+b = f (x) is singular if its discriminant = (4a3 +27b2 ) = 0,
so, if and only if f (x) has at least a double root e. In that case, there is a unique singular point
P0 = (e, 0). Even though such curves are not elliptic curves, they are still useful. Let Ens (K) be
the set of all non-singular points, that is
Ens (K) = E(K) \ {P0 }.
We will show that Ens (K) is a group.
Claim: Ens (K) is an abelian group with the same group law as before. This works because
P, Q 6= P0 P Q 6= P0 .
30
There are two subcases.
Case 1:
The cubic f (x) has a triple root e K. By expanding f (x) = (x e)3 = x3 + ax + b, we see that
e = a = b = 0. So E : y2 = x3 , and the point (0, 0) is a cusp. This is called the additive case.
y 2 = x3
: Ens (K) K +
x
(x, y) 7
,
y
7 0,
e1
Proof. We need to check that is a group homomorphism, which is also a bijection. Let
(x, y) Ens (K) \ {}, then xy 6= 0 and y2 = x3 . Setting t = xy , we see that (x, y) = (t 2 ,t 3 ), and
that is indeed a bijection, whose inverse is the map
: K + Ens (K)
t 7 (t 2 ,t 3 ),t 6= 0,
0 7 .
In the proof above, we used the following fact. Let : G H be a map between two
groups. Then is a group homomorphism if and only if (g1 G g2 ) = (g1 ) H (g2 ). To
show this it is equivalent to check that, whenever g1 G g2 G g3 = eG (the identity element
in G) then (g1 ) H (g2 ) H (g3 ) = eH and also that (eG ) = eH .
Case 2:
The cubic f (x) has a double root e1 = e2 = e 6= 0 and a simple root e3 . Since the sum of the
roots must be zero, it follows that e3 = 2e. So we can write E : y2 = (x e)2 (x + 2e). In
that case, the singular point P0 = (e, 0) is a node. By making a translation, we can assume that
E : y2 = x2 (x + a), with a K , and P0 = (0, 0). This is called the multiplicative case. Let
be a root of the polynomial x2 a in K.
31
: Ens (K) K
(x, y) 7 u :=
7 1,
y + x
,
y x
2e
y/x + t +
=
y/x t
u+1
.
u1
x = t a =
u+1
u1
2
2 = 2
4 2 u
4au
(u + 1)2 (u 1)2
=
=
.
(u 1)2
(u 1)2 (u 1)2
u+1
4au(u + 1)
4au
=
.
2
u 1 (u 1)
(u 1)3
4au
4au(u + 1)
and y :=
(u 1)2
(u 1)3
+
= 1,
(u 1)2
(u 1)3
or equivalently
4ua(u 1) + 4ua(u + 1) = (u 1)3 .
This is a cubic in u whose constant term is u1 u2 u3 = 1, hence u1 u2 u3 = 1.
In case (1), since K, (x, y) Ens (K) u K . So induces an isomorphism Ens (K) ' K
in that case.
L = K() = {s + r a : s, r K}.
32
y + x
(y + x)2
y2 + 2xy + 2 x2 (y2 + ax2 ) + 2xy
=
=
=
= s + r,
y x (y x)(y + x)
y2 2 x2
y2 ax2
uu = (s + r)(s r) = s2 ar2 = 1.
So
Ens (K)
= u L : uu = 1 .
Example 2.5.2.1 Let K = R, L = C and E : y2 = x2 (x 1). Then E(C) = C and E(R) is