Page 1
MATHEMATICS OF COMPUTATIONVOLUME 48, NUMBER 177JANUARY 19K7. PAGES 351-370
Computation of the Néron-Tate Height
on Elliptic Curves
By Heinz M. Tschöpe and Horst G. Zimmer
For Daniel Shanks on the occasion of his 10 th birthday
Abstract. Using Néron's reduction theory and a method of Täte, we develop a procedure for
calculating the local and global Néron-Tate height on an elliptic curve over the rationals. The
procedure is illustrated by means of two examples of Silverman and is then applied to
calculate the global Néron-Tate height of a series of rank-one curves of Bremner-Cassels and
of a series of rank-two curves of Selmer. In the latter case, the regulator is also computed, and
a conjecture of S. Lang is investigated numerically.
In dealing with the arithmetic of elliptic curves E over a global field K, the task arises of
computing the Néron-Tate height on the group E(K) of rational points of E over K. Solving
this task in an efficient manner is important, for instance, in view of calculations concerning
the Birch and Swinnerton-Dyer conjecture (see [2]) or of the conjectures of Serge Lang [6],
The purpose of this note is to suggest a procedure for performing the necessary calculations.
1. Multiplication Formulas. Let the elliptic curve E over any field K be defined by
a generalized Weierstrass equation
(E) y2 A axxy A a3y = x3 + a2x2 A a4x A a6 (a, e K).
As usual, we introduce the quantities (see [10], [11])
A2 = a2 A 4a2, b4 = axa3 A 2a4, b6 = aj A 4a6,
Ag = axa6 - axa3a4 A 4a2a6 A a2a\ - a\,
c4 = b\ - 24b4, c6 = -A3 + 36A2A4 - 216A6,
and the discriminant
A = -A22A8 - 8A43 - 27A2 + 9A2A4A6 * 0,
as well as the absolute invariant
j = cl/ts,
belonging to E over K.
The fact that E is nonsingular implies the nonvanishing of the partial derivatives
of the polynomial
F(x, y) — y2 A axxy A a3y - jc3 - a2x2 - a4x - a6
Received May 19, 1986; revised June 30, 1986.
1980 Mathematics Subject Classification. Primary 14-04, 14H45, 14K07, 14K.15; Secondary 10B10,
14G20,14G25,14H25.
©1987 American Mathematical Society
0025-5718/87 $1.00 + $.25 per page
351
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 2
352 HEINZ M.TSCHÖPE AND HORST G. ZIMMER
at every rational point P e E(K):
(Íf(,),fo>))*<o.o).The addition law in the additive Abelian group E(K) of rational points on E over K
is given by the following formulas:
For P = (xP, yP), Q = (xQ, yQ) e E(K), denote the sum by P A Q =
(xP+Q,yP+Q).Then,
i L. \Ayr~yQ\2, lyp-yç]xP+Q= -(xP + xQ)+ ——-- \ +ax — -a2,
y xP xqj yxP Xqj
yP - yQyp+Q = v - v (xp ~ xp + q) - aixp+Q - a3 - yP if P * Q
Xp Xq
and
x2P = -2xP A t2P A axtp - a2, y2P = tP(xP - x2P) - axx2P - a3- yP
(2) im 3x\A2a2xPAa<-axyP ,f
2yP + axxP A a3
Generalizing classical formulas (see [3], [4], [15]), we obtain
Proposition 1. For a rational point P e E(K) and an r e N, the r-fold rational
point has coordinates
where <I>r, ^r, and 2Qr are polynomials in x and y with coefficients in
Z[ax, a2, a3, a4, a6] given by the following recursion formulas:
•$! = x, 02 = xA - b4x2 - 2b6x - A8,
Sl1=y, % = 0, % = 1,% = 2yAa1xAa3,
% = 3x4 + b2x3 + 3b4x2 A 3b6x A A8,
*4 = *2[2;c6 + A2x5 + 5A4*4 + lOA«,*3 + 10A8x2 +(A2A8 - A4A6)x + A4A8 - A62]
and for r > 2,
<Pr = xyvr — Tr_1M/r+1,
2%ttr = *?_x%+2 - %^2+i - %%[ax<i>r + a,*,2],
\J> = vl/3* — \Ir \J>3*2r+l *r Tr+2 V-1V+1'
%%r=%[*r2-l%+2-%-2*r2+l\-
Moreover, í>r, as a polynomial in x, has degree r2 and leading coefficient 1, whereas
^ (resp. <í,21<fír), as a polynomial in x, has degree (r2 — l)/2 (resp. (r2 — 4)/2)
and leading coefficient r (resp. r/2) provided that r is odd (resp. even). If we assign
the weight 2, 3 or i to x, y or a¡, then each term of $r has weight 2r2 and each term of
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 3
THE NÉRON-TATE HEIGHT ON ELLIPTIC CURVES 353
^fr (resp. <^2lyi,r) has weight r2 - 1 (resp. r2 - 4). The coefficients of <L>r, tyr, as
polynomials in x, y, belong already to Z[A2, A4, A6, A8].
From Proposition 1, one derives the following
Corollary. For r g N, we put ^_r = -%. Then, for r, n g N, we have
(4) *?n(P) = %2r\P)^2(nP)
and, more generally,
(4') *2»(p) = n*2m2("~')(™>iP).v=l
Furthermore,
(5) X -X %+n(P)%-ÁP)() ,P "" ViPWiP) '
r+1 %(P A Q)%(P - Q),
*?(P)*?(Q)
f,s , ,V+1 ^r\r T A¿)^Ar ~ S>) , ,r(6) Xrp-XrQ=(-l) -.T,2iPW,2/^-(Xp-tQ)
and finally, for r G N0,
(7) 4»2(2T) = $2,+ l(P)*2-,8(P), ^22(2rP) = ^2 + i(P)^2rs(P).
These formulas will be needed in the sequel.
2. Reduction Theory. Now let the elliptic curve E be defined by (E) over a
complete field K with respect to a discrete normalized additive valuation v, and
suppose that the corresponding residue field K of K is perfect. We assume the
equation defining E over K to be minimal with respect to the valuation v (see [11]).
Reducing E modulo v yields a cubic curve
(E) y2 + äxxy + ä3y = x3 + ä2x2 + ä4x + ä6 (à, G K)
over K with discriminant À. If Ä # 0, i.e., u(A) = 0, then Ë is an elliptic curve over
K, and E has good reduction at v. If, however, À = 0, i.e., u(A) > 0, then £ is a
rational curve over K, and E has bad reduction at v. In the latter case, E is said to
have multiplicative reduction or additive reduction modulo v, according as v(c4) = 0
or u(c4) > 0, respectively.
Denote by E0(K) the set of points in E(K) whose image under the reduction
map modulo v,
P: E(K)-*É(K),
is a nonsingular point on É over K. Then, EQ(K) is a subgroup of finite index in
E(K). Further, the set
EX(K) = [P = (*„ v>) g E(K)\v(xP) 4 -2, t;( v>) < -3}
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 4
354 HEINZ M.TSCHÖPE AND HORST G. ZIMMER
is a subgroup of E0(K), sind the restriction p0 to E0(K) of the reduction map p
induces an injective homomorphism of the factor group
~Po: E0(K)/EX(K) ^ Ë0(K)
to the nonsingular part Ë0(K) of Ë(K).
We shall use the following result (see [11]).
Proposition 2. The above groups satisfy
E(K) = E0(K) if E has good reduction at v,
#(E(K)/E0(K)) divides v(j) if E has multiplicative reduction at v,
and
#(E(K)/E0(K)) 4 4 if E has additive reduction at v.
3. Definition of Height Functions. Now let K be a global field, that is, an
algebraic number field or a function field of finite transcendence degree over its field
of constants k. Then K possesses a complete set MK of nonequivalent additive
valuations v satisfying the sum formula
(S) Ya \vv(c) = 0 forO^ce K
with some positive multiplicities Xv g R (cf. [7], [13]).
For an elliptic curve E over K, given by the Weierstrass equation (E), we
introduce the quantities
(8) ix, = min{t;(A2),^(A4),it;(A6),if;(A8)}.
Let P = (xP, yP) g E(K) be any rational point and 6 = (oo,oo) designate the
point at infinity. Then we define the local Weil height on E(K) with respect to v by
setting
... . . /-imii.((.„ip(i,)) ili*»,
(9) *-(P>=(-K ur-0.
Then the global Weil height on E(K) is simply the sum, with multiplicities, over the
local Weil heights
d(p)= £ Kd„(P)veMK
(see [13]).
In order to define the global Néron-Tate height on E(K), we proceed in the same
way as with the global Weil height. However, before introducing the local Néron-Tate
height on E(K), we need some estimates.
Proposition 3. The local Weil height onE(K) satisfies the following estimates:
{-(6u(, - D(A)) + 5a, < d,(P AQ) + d„(P - Q)
(10) -2dv(P)-2dv(Q)-v(xp-xQ)
4-2av if P,Q,P ± Q + e,
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 5
THE NERON-TATE HEIGHT ON ELLIPTIC CURVES 355
and
k{6K - v(A)) A 4av 4 dv(2P) - 4dv(P) - M*2(P))
4 -|a(, if2P*&,
where the constant av can be chosen to be 0 or -log 2 according as the valuation v of K
is discrete or archimedean, respectively (see [13]).
These estimates are obtained as generalizations of those given in [13], [14]. At the
same time, they sharpen those cited.
Remark 1. It is interesting to note that the authors of [2] suggested that a
sharpening of the estimates in [13], [14] should be possible. Proposition 3 appears to
be a step in this direction.
Employing (10), the inequalities (11) can be further generalized.
Corollary. For any m G N, there are (recursively computable) nonnegative
constants c, m, c2 m G R depending on E, K, and v such that, given an arbitrary point
P G E(K) with mP # 0, we have
(IV) cUm 4 dv(mP) - m2dv(P) - Jd(*¿(P)) *S c2,m.
We are now in a position to define the local Néron-Tate height on E(K) with
respect to v. Let m, n g N and m > 2. Then, for a rational point P G E(K) such
that m"P =é <S for each n G N, we define the local Néron-Tate height of P with
respect to v by the limit formula
<*> v.<,)-B./4í^i-líí3aáía\+±.(4).n^oo \ m l m I í¿
Proposition 4. For an elliptic curve E defined by a Weierstrass equation (E) over a
global field K and any valuation v of K, the function 8V m, defined by (12) on the
rational point group E(K), exists, is independent of the choice of m g N, so that
8V m = Sv, and fulfills the relations
(13) 8V(P AQ)A 8V(P - Q) - 28„(P) - 28V(Q) - v(xP - xQ) A MA) = 0
for any two points P = (xP, yP), Q = (xQ, yQ) G E(K) such that P,Q,P ±Q* 0,
and
(14) 8„(rP) - r\(P) - ±v{*?(P)) A ̂ -«(A) = 0
for any P = (xP, yP) G E(K) and r G N such that rP * 0.
Proof. The proof is an adaptation of the corresponding proof of the existence
theorem in [14]. Indeed, one exploits (10), (11) from Proposition 3 and (11') from the
corollary to establish the existence of 8V m. Then formulas (6) and (4) from the
corollary to Proposition 1 are utilized to prove that 8l: m fulfills the asserted relations
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 6
356 HEINZ M.TSCHÖPE AND HORST G. ZIMMER
(13) and (14). Finally, the independence of 8V m on m is a consequence of the
following
Corollary 1. The function 8V m on E(K) is related to the local Weil height on
E(K) through the estimate
(15)
where
dv(P)A±v(A) <cm,
m
1 /I I I il— ■mnx{\clm\,\c2m\).
In fact, 8V m = 8V is uniquely determined by the properties (14) and (15) and hence is
independent of the choice of m.
We can now define the global Néron-Tate height on E(K) as the sum, with
multiplicities, over the local Néron-Tate heights as follows (see [14]):
, , Í I KMP) ap + o,(16) 8(P) = vgmk
lo ifP = 0.
By the sum formula (S) we then obtain on the basis of (13) and (14):
Corollary 2. The global Néron-Tate height onE(K) fulfills the relations
(13') 8(P A Q) A 8(P - Q) - 28(P) - 20(g) = 0
and, for r G N,
(14') 8(rP)-r28(P) = 0.
Remark 2. Corollary 2 shows that the global Néron-Tate height 8 is a quadratic
form on E(K), whereas Proposition 4 implies that the local Néron-Tate height 8V is
"almost" a quadratic form on E(K).
4. Computation of the Néron-Tate Height. Again, let the elliptic curve E be given
by (E) over a global field K. Fix a nonarchimedean (discrete) valuation v of K.
Suppose that P = (xP, yP) G E(K) is a rational point satisfying v(xP) < nir
By Proposition 1, on choosing an m g N such that m > 2 and v(m) = 0, we have
Xm"P=*2»(p)-
Now v(xp) < ju.t, together with v(a¡) ^ fiv entails
«(<MP)) - rn2"v(xP), D(**.(P)) = (m2" - l)v(xP).
Hence
f (xm»P) = v(xp).
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 7
THE NÉRON-TATE HEIGHT ON ELLIPTIC CURVES 357
Thus we obtain from the limit formula (12) and the definition (9) of dv the asserted
relation
«„(P) = - HxP) A -¿i;(A) = d.(P) + ¿p(A).
Proposition 5. Suppose that a rational point P = (xP, yP) g E(K) satisfies the
inequality v(xP) < u„ for a nonarchimedean (discrete) valuation v of the global field
K. Then the local Néron-Tate height of P essentially coincides with the local Weil height
of P with respect to v; more precisely,
8v(P) = du(P) + ~v(A).
From Proposition 5 we get the following theorem, which is crucial for the
calculation of the Néron-Tate height on E(K).
Theorem 1. Let E be an elliptic curve defined by a Weierstrass equation (E) over an
algebraic number field K. Choose a discrete normalized additive valuation v of K and
suppose that the equation (E) is minimal with respect to v.* Then, for each nontorsion
point P G E0(K), the local Néron-Tate height of P is essentially equal to the local Weil
height of P with respect to v; more precisely,
8v(P) = dv(P)A±v(à).
Proof. The theorem can be found in [9]. For the convenience of the reader,
however, we give a proof.
By Proposition 5, we may confine ourselves to the case in which v(xP)> ¡iv. The
subcase in which v(xP)> u„ > 0 would lead to a contradiction to the choice of
P G EQ(K). Hence it remains to consider the subcase in which
v(xP)>nv = 0.
The reduction map of Section 2,
p0: E0(K)/EX(K) -» É0(K),
is an injective homomorphism. Since K is a number field, the residue field K of K
with respect to v is finite and hence so is the group E0(K). Therefore, for any
P g E0(K), there exists a number r g N such that rP g Ex(K). Choose r g N
minimal with this property. Then we have
v{xrP) <ßu = 0-
From this, since v(xP)> u„ = 0 and v(a¡) > u„ = 0, we conclude that
v($r(P)) > 0 and v{%(P)) > 0.
We claim
(17) v(xrP) = -v{^(P)).
*The required minimal model of E is found by Tate's algorithm [11].
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 8
358 HEINZ M.TSCHÖPE AND HORST G ZIMMER
By Proposition 5, Formula (14) of Proposition 4, and the definition (9) of dv, this
claim yields the asserted identity
«.(*) = i\{8^rp) - ri^ip)) + ̂ »(a)
= dv(P)A±v(A)
since v(xP) > ¡iv = 0.
To prove (17) it suffices to show that
(18) u(*,(P)) = 0.
This is accomplished by verifying (18), first for the lower r g N and then for general
r G N.
Let r = 2.
If v(3xp A 2a2xP A a4 - axyP) > 0 we would get a contradiction to the assump-
tion that P g E0(K). Hence it is enough to consider v(3xp A 2a2xP A a4 - axyP)
= 0. But then the asserted relation (18) follows directly from the formula (2) for x2P
and Proposition 1.
Let r = 3.
By the minimal choice of r, we have
v(%(P)) = 0 sind v(%(P))>0.
Now the decomposition formula (which can be verified without trouble)
%(P) = %(P)[%(P)(6x2 + b2xP A b4) - %\P)}
yields v(ty4(P)) = 0, and hence the relation from Proposition 1,
<t>3(P) = xP%2(P)-%(P)%(P),
leads to the identity v(<t>3(P)) = 0, as asserted in (18).
Finally, let r > 4.
Again, by the choice of r, we have
v(%(P)) = v(%(P)) = ••• = v(%_1(P)) = 0 and
i>(*,(P))>0.
Then Formula (5) from the corollary to Proposition 1 yields
v(x2P-xP) = 0 and v(x(r_X)P - xP) > 0,
so that another consequence of Formula (5), viz.,
^r+l(P) = ~[(X(r-l)P ~ Xp) +(xP — x2P)] —; r— ,
leads to i;(^r+1(/>)) = 0. Now the identity from Proposition 1,
*r(P) = xP*?(P)-%_x(P)%+x(P),
reveals that v(<±>r(P)) = 0, as asserted in (18). This proves Theorem 1.
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 9
THE NERON-TATE HEIGHT ON ELLIPTIC CURVES 359
Remark 3. Theorem 1 makes it possible to calculate the local Néron-Tate height
8V(P) with respect to all discrete valuations v of the number field K for all
nontorsion points P G E(K).
This is true because Proposition 2 tells us that a suitable multiple rP of P belongs
to E0(K). Then we apply Theorem 1 to calculate 8v(rP) and use Formula (14) from
Proposition 4 to get the desired value of 8V(P) itself.**
Remark 4. Torsion points P G E(K) are of no interest in this connection since
their global Néron-Tate height is 8(P) = 0.
It remains to show how to compute the local Néron-Tate height S„ for archi-
medean valuations v of the number field K. From (4') in the corollary to Proposi-
tion 1, we get the formula
1 o(^.(P)) " 1 v{*Z,(m-lP))
which proves to be useful in the sequel.
Now, since we are interested here only in the case of K = Q, the field of rational
numbers, we confine ourselves to considering its completion Kx = QK = R with
respect to the ordinary absolute value v = vx = -log | |. Then Tate's method is best
suited for calculating 8V (see [12]).
Theorem 2. Let E be an elliptic curve defined by a Weierstrass equation (E) over
the field R of real numbers and denote by vx = -log| | the ordinary additive
archimedean valuation of R. Take an open subgroup T of P(R) such that all
P = (xP, yP) g r satisfy xP * 0.*"* For P g T such that 2"P * 6 for all n g N,
define the entities Tn, Wn, and Zn by putting
'T1 _ _n
' 1n+l - 7Xp z.n
where
Wn = 4T„ A b2T„2 A 2A4P„3 + A6P„4, Z„ = 1 - A4r„2 - 2A6P„3 - A8P„4.
Let
fi(P)- £ ^i^. \(P) = \l0g\Xp\ + \ß(P).n = 0 *•
Then the local Néron-Tate height of P with respect to vx is
«JP) = X(P)-^log|A|.
Proof See [12]. However, the assertion of Theorem 2 also follows from
Proposition 6. In the situation of Theorem 2 we have for n g N0,
1 *22(2"P) $2(2"P)i„ =-, W„ =-■-, Z„ =-:-.
Po = — ' Tn + i = ^T fom& N0,
\.2np i2np A2„p
** Added in proof. Joe Silverman, whom we wish to thank for some valuable hints, told us that he has
carried out similar height computations (unpublished) avoiding, however, the use of Proposition 2 by
employing Tate's local formulas (see [14]).
***Hence T is either £(R) or the identity component of £(H) according as £(R) is connected or
disconnected.
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 10
360 HEINZ M.TSCHÖPE AND HORST G. ZIMMER
Proof. The proof is carried out easily by means of the formulas (7) in the corollary
to Proposition 1.
Remark 5. The simplest way of finding a subgroup T of E(R) of the type desired
in Theorem 2 is by applying a birational transformation to (E) to obtain a model
(E') such that A¿ < 0. Then T = £"(R) itself will do.
In the special case of K = Q we are interested in, the set MQ consists in the
p-stdic valuations v corresponding to the primes p of Q and the additive valuation
vx = - log | | corresponding to the unique archimedean absolute value | | on Q. Of
course, the multiplicities in the sum formula (S) are all \L, = 1.
5. Examples. We are now in a position to calculate the Néron-Tate height 5 on the
group E(K) of rational points on an elliptic curve E over the rational number field
K = Q. To this end, we use the defining formula (16) for 8 with multiplicities
X v = 1 to reduce the computation of 8 to that of the local Néron-Tate heights 8V on
E(K). For discrete valuations v of Q, the height 8V is calculated by means of
Theorem 1 in accordance with Remark 3, and for the archimedean absolute value
vx = -log| |, the calculation of 8V is performed on the basis of Theorem 2.
(i) Examples of Silverman. We illustrate our procedure by verifying the height
calculations of Silverman [9].
(A) E: y2 A 21xy A 494y = x3 A 26x2,
P = (0,0)g£(Q),
A = -213 • 133 • 192.
Silverman obtains
5(P) = 0.010,492,....
We have
(a) 8„ (P) = 0.038,612,393,...,
(b) P <£ £0(Q) for p = 2,13 and 19; and
5„(P) = 0 for all primes p * 2, 13 or 19.
Now
13P e £0(Q) for p = 2,
3Pg£0(Q) for/> = 13,
2Pg£0(Q) for/; = 19.
One computes
%(P) = 2 • 13 • 19, %(P) = 23 • 133 • 192, *13(P) = -280- 1356 • 1942
and
x2P = -2-13, x3P = -2-19, xX3P= -24 ■ 5 ■ 13 • 19.
This leads to
802(13P) = -gln2, «Di3(3P) = ilnl3, 8VJ2P) = | In 19.
Hence, by (14) of Proposition 4,
S02(P) = ^ln2, 8VJP)= -^lnl3, 8jP) = - -^ In 19.
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 11
THE NÉRON-TATE HEIGHT ON ELLIPTIC CURVES 361
By (16) this adds up to
8(P) = 0.010,492,061,....
(B) E: y2 + Uxy A my = x3 4- Sx2,
P = (0,0)g£(Q),
A = -211 • 52 ■ 19.
Silverman gets
8(P) = 0.010,284,...
and we obtain similarly to (A)
8(P) = 0.010,284,005,....
(ii) The Bremner-Cassels Curves. Our procedure turns out to be particularly useful
for calculating the global Néron-Tate height on the elliptic curves
E : y2 = x3 + px
for primes p of Q such that p = 5 (mod 8), as they were considered by Bremner and
Cassels [1]. The authors exhibit points P g E(Q) of infinite order on 43 curves of
this type, where
r2 r • tP = (xP,yP) with Xp = —-, yP = —— forr,í,íGZ
s s3
such that
g.c.d.(r,s) = 1; r,/# 0 (mod/?); and
r = / == 1 (mod2), ,r = 0(mod2).
One easily checks that P g £0(Q) for all primes p of Q and all points P G P(Q)
displayed in [1]. (Notice that 2yP and 3x2P + p are relatively prime.) This leads to
Proposition 7. For the points P g ^(Q) of infinite order on the Bremner-Cassels
curves in [1], the Néron-Tate height is
8(P) = 8JP)A ¿In|A|+lnM.
(iii) Modular Elliptic Curves. In [16, pp. 75-113], N. M. Stephens and J. Daven-
port list 68 modular elliptic curves E of rank 1 with a rational point P g E(Q) of
infinite order. We computed the Néron-Tate heights of these points P.* Comparison
of the Néron-Tate height of the generator of the 63rd curve in their table with the
Néron-Tate height of the point in Silverman's second example (see (i) (B) above)
shows that the two values agree. It turns out, as one easily checks, that the
corresponding two curves are birationally isomorphic (see Table 1).
+ We have compared the height values in our Table 1 with those in a corresponding (unpublished) table
of Silverman containing up to six digits behind the period. They agree (except for the sixth digits of the
curves 58A, 61A, 135A, 153A, 189C and for the fifth and sixth digit of the curve 185D).
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 12
362 HEINZ M.TSCHÖPE AND HORST G. ZIMMER
Table 1
1 . ) al = 0 a2 = 0 a3 = 1 a4 =-1 aó
P = ( 0 ; 0 )global height: .0255557041237A
2 . ) al = 0 a2 = 1 a3 = 1 a4 = 0 a6P ■ ( 0 S 0 )global height: .03140825354443A
3 .) al = 1 a2 =-1 a3 = 1 a4 = 0 aóP = ( 0 ; 0 )
global height: .04649074231953A
4 .) al = 0 a2 =-1 a3 = 1 a4 =-2 aó
57E P = ( 2 ; 1 )global height: .018787296303
5 .) al = 1 a2 =-1 a3 = 0 a4 =-1 aó
P » ( 0 J 1 )global height: .0212101539258A
6 .) al = 1 a2 = 0 a3 = 0 a4 =-2 aóP = ( 1 ; 0 )global height: .039593865681
61A
7 . ) al = 1 a2 = 0 a3 = 0 a4 =-1 a6P = (-1 ¡ 1 )global height: .187757
65A
8 . ) al = 0 a2 = 0 a3 = 1 a4 = 2 aó
77FP = ( 2 ; 3 )global height: .049013989032
9 .) al = 1 a2 = 1 a3 = 1 a4 = -2 aóP = ( 0 ; 0 )
79A global height: .048832105054
10 .) al = 1 a2 = 0 a3 = 1 a4 =-2 aó
82A p = < ° ! ° »°¿il global height: .112353462459
11 . ) al = 1 a2 = 1 a3 = 1 a4 = 1 aó
83A P = ( 0 ; 0 )global height: . 033Ó4Ó1.47057
12 .) al = 0 a2 = 0 a3 = 0 a4 =-4 aó
88A P = í 2 i 2 )global height: .020132182168
13 .) al = 1 a2 = 1 a3 = 1 a4 =-1 aó
8qc P = ( 0 ! 0 )oy^ global height: .056052440615
14 . ) al = 0 a2 = 0 a3 = 1 a4 = 1 aó
91A P = < 0 ! 0 )^ global height: .071190075334
15 .) al = 0 a2 = 1 a3 = 1 a4 =-7 aó
q1R P = (-1 ; 3 )3 global height: .529622543205
16 .) al = 0 a2 = 0 a3 = 0 a4 =-1 aó
P ■ < 1 5 1 )9¿C" global height: .024904198649
99A
17 .) al = 1 a2 =-1 a3 = 1 a4 =-2 aóP = ( 0 ; 0 )global height: .151285092281
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 13
THE NÉRON-TATE HEIGHT ON ELLIPTIC CURVES
Table 1 (continued)
18 .) al = 0 a2 = 1 a3 = 1 a4 =-1 aó =-1
101A P " <_1 ' ° '' global height: .032351726475
19 .) al = 1 a2 = 1 a3 = 0 a4 =-2 aó = 0
102E P ■ ("I i 2 'u"* global height: .07162694647
20 .) al = 1 a2 = 1 a3 = 0 a4 =-7 aó = 5
106A P = < 2 ¡ 1 )global height: .034456340202
21 . ) al = 0 a2 = 1 a3 = 0 a4 = 0 a6 = 4P = ( 0 i 2 )
1 ltIV global height: .119959949363
22 .) al = 1 a2 =-1 a3 = 1 a4 = 4 aó = 6P = ( 0 i 2 )
117A global height: .56516781309
23 .) al = 1 a2 = 1 a3 = 0 a4 = 1 aó = 1
118A p = < ° ' 1 '1 IOA global height: .043953097838
24 .) al = 0 a2 =-1 a3 = 1 a4 =-7 aó = 1
121D P - ' « > 3 >*■ 'u global height: .04439257808
25 . ) al = 1 a2 = 0 a3 = 1 a4 = 2 aó = 0122A P = ' l ' 1 '
global height: .060421607704
26 .) al = 0 a2 = 1 a3 = 1 a4 =-10 a6 =123A P = ' 1 ' l 'u:->A global height: .420260708766
27 .) al = 0 a2 = 1 a3 = 0 a4 =-2 a6 = 1
124B P - < 1 I 1 )global height: .2602Ó5346941
28 . ) al = 0 a2 = 1 a3 = 0 a4 = 1 aó
128C P = « 0 ¡ 1 )global height: .216105582237
29 .) al = 0 a2 =-1 a3 = 1 a4 =-19 aó =
12QE P = ( 1 ¡ 4 )l¿yfi' global height: .04997957634
30 . ) al = 1 a2 = 0 a3 = 1 a4 =-33 aó =P = ( 2 i 2 )
"OH global height: .585232076797
31 .) al = 0 a2 =-1 a3 = 1 a4 = 1 aó = 0,,,. P = < o ; o )1-51A global height: .108047599334
136A
2 . ) al = 0 a2 = 1 a3 = 0 a4 =-4 aó
P = (-2 ! 2 )global height: .115753990413
33 . ) al = 1 a2 = 1 a3 = 0 a4 =-1 aó =
1W P = ( 0 ! 1 ),-5ÖI!, global height: .08368409567
34 .) al = 0 a2 = 1 a3 = 1 a4 =-12 aó
-wir. P = (-3 ! 4 )1411i global height: .017243387509
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 14
364 HEINZ M.TSCHÖPE AND HORST G. ZIMMER
Table 1 (continued)
35 . ) a
1411 P
30 . ) a
142E I
37 . ) a
142F r,
38 . ) a
145A I
39 . ) a
148AÜ
40 . )
152A
41 . ) a
153AP
42 . )
153C
43 . ) a
155C I
44 . )
156E
45 . ) a
158D Pg
46 . ) a
158E pg
47 . )
162K
43 . )
163A
49 . )
'■Ak
50 . )
171A
51 . )
172A
= 0 a2 =■= ( 0 ; 0 )
obal height:
= 1 a2 == (-1 ¡ 1 )
obal height:
= 1 a2 =-= ( 1 ; 1 )
obal height:
= 1 a2 =-= ( 0 ¡ 1 )
obal height:
= 0 a2 =■= (-1 ¡ 2 )
obal height:
= 0 a2 =
= (-1 ! 2 )obal height:
= 0 a2 == ( 5 ; 13 )
obal height:
= 0 a2 == ( 0 ; 1 )
obal height:
= 0 a2 == ( 1 ; 0 )
obal height:
= 0 a2 =■= ( 1 ; 1 )
obal height:
= 1 a2 == ( 0 i 1 )
obal height:
= 1 a2 == (-1 ¡ 4 )
obal height:
= 1 a2 == ( 2 ; 0 )
obal height:
= 0 a2 =
= ( 1 ! 0 )obal height:
= 1 a2 =
= ( 0 ! 2 )obal height:
= 0 a2 == ( 2 ¡ 4 )
obal hei ght:
= 0 a2 == ( 2 Î 1 )
obal he i ght:
1 a3 = 1
.099247618232
1 a3 = 0
.090456855492
1 a3 = 1
.010894571319
1 a3 = 1
.292228814932
1 a3 = 0
.043120589701
1 a3 = 0
.032707434794
0 a3 = 1
.056444869251
0 a3 = 1
.034740140542
1 a3 = 1
.092071901309
1 a3 = 0
.073707206024
1 a3 = 0
.03958438143
1 a3 = 1
.019495140155
1 a3 = 0
.152967441934
0 a3 = 1
.094954616249
1 a3 = 0
.044978395458
0 a3 = 1
.112983434413
1 a3 = 0
.380069831503
a4 =-1
a4 =-1
a4 =-12
a4 =-3
a4 =-5
a4 =-1
a4 = 6
a4 =-3
a4 =-1
a4 =-5
a4 =-3
a4 =-9
a4 =-6
a4 =-2
a4 =-6
a4 = ó
a4 =-13
aó = 0
a6 =-1
aó = 15
aó = 2
aó = 1
aó = 3
aó = 27
aó = 2
aó = 1
aó = 6
aó = 1
aó = 9
aó = 8
aó = 1
aó = 4
aó = 0
aó = 15
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 15
THE NÉRON-TATE HEIGHT ON ELLIPTIC CURVES
Table 1 (continued)
52 . )
175A
53 . )
175C
54 . )
176A
55 . )
184B
50 . )
184C
57 . )
185A
58 . )
185B
59 . )
185D
60 . )
189A
61 . )
189C
62 . )
190C
63 . )
190D
04 . )
192Q
65 . )
196A
66 . )
197A
67 . )
1981
68 . )
200C
= 0 a2 == ( 7 ¡ 2 )
obal height:
= 0 a2 == (-3 ; 12 )
obal height:
= 0 a2 == ( 1 ¡ 2 )
obal height:
= 0 a2 =
= ( 2 ! 1 )obal height:
= 0 a2 == ( 0 ; 1 )
obal height:
= 0 a2= ( 0 ; 2 )
obal height:
= 1 a2 == ( 3 ; 2 )
obal height:
= 0 a2 == ( 4 ; 12 )
obal height:
= 0 a2 == (-1 ¡ 1 )
obal height:
= 0 a2 == (-3 ; 9 )
obal height:
= 1 a2 == ( 1 ; 2 )
obal height:
= 1 a2 == ( 13 ; 33
obal height:
= 0 a2 ■= ( 3 ; 2 )
obal height
= 0 a2 == ( 0 ; 1 )
obal height:
= 0 a2 == ( 1 ¡ 0 )
obal height:
= 1 a2 == (-1 ¡ 5 )
obal height:
= 0 a2 == (-1 i 1 )
obal height:
-1 a3 = 1
.332314998542
-1 a3 = 1
.040236666901
-1 a3 = 0
.087531915126
-1 a3 = 0
.051533618406
1 a3 = 0
.061565455601
-1 a3 = 1
.055139483611
0 a3 = 1
.712032645336
1 a3 = 1
.057028352204
0 a3 = 1
.031006094417
0 a3 = 1
.931621776106
1 a3 = 0
.065910740941
-1 a3 = 1)
.010284005728
-1 a3 = 0
.675801867206
-1 a3 = 0
.043017725483
0 a3 = 1
.069433995882
1 a3 = 0
.097521495699
1 a3 = 0
.146605513301
a4 =-143
a4 =-33
a4 = 3
a4 =-4
a4 = 0
a4 =-5
a4 =-4
a4 =-156
a4 =-3
a4 =-24
a4 = 2
a4 =-48
a4 =-4
a4 =-2
a4 =-5
a4 =-13
a4 =-3
a6 = 748
aó = 93
aó = 1
aó = 5
aó = 1
aó = ó
aó =-3
aó = 700
aó = 0
aó = 45
aó = 2
aó = 147
aó =-2
aó = 1
aó = 4
aó = 4
aó =-2
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 16
366 HEINZ M.TSCHÖPE AND HORST G. ZIMMER
6. Lang's Conjectures. Silverman [9] used his above-cited examples of rank-one
elliptic curves E over Q to estimate the constants c,, c2 in S. Lang's Conjecture 2
(see [6]) about lower bounds for the Néron-Tate height 8 on nontorsion points in
£(Q). We wish to carry through a similar estimation with respect to Lang's
Conjecture 3 (see [6]) for Selmer's [8] rank-two elliptic curves E over Q.
In Section 3, Remark 2, we observed that the Néron-Tate height 5 is a quadratic
form on the rational point group E(Q). This property of 8 is tantamount to the fact
that the function
ß(P,Q) = HHPAQ)-8(P)-8(Q)}
Table 2
al = 0 a2 » 0 a3 - 0 a4 ■ 0 aó - -388800
PI » ( 76 / 1 $ 224 / 1 )
P2 - ( 124 / 1 I 1232 / 1)
The trans-formation with (risit(u) > ( 0 I 0 I 0 | 2 ) leads to
al - 0 a2 » 0 a3 » 0 a4 - 0 aó - -Ó075
Pl-< 19 / 1 | 28 / 1 )
p the local height decimal
2 (1/3 )*ln( 2 ) .2310490601803 < 13 / 12 )*ln< 3 ) 1.1901633127235 (1/3 )*ln( S ) .536479304144co -.220039705773
The global height is 1.73765197128
P2-( 31 / 1 i 134 / 1 )
p the local height decimal
2 (1/3 )*ln( 2 ) .2310490601863 (13/12 )*ln( 3 ) 1.1901633127235 ( 1 / 3 )*ln( 5 ) .536479304144<=° -.068619441325
The global height is 1.889072235727
Pl+P2=( 241 / 4 i-3689 / 8 )
p the local height decimal
2 ( 4 / 3 >*ln( 2 > .9241962407463 (13/12 )*ln< 3 ) 1.1901633127235 ( 1 / 3 )*ln( 3 ) .536479304144oo .18319182537
The global height is 2.834030682983
Regulator : 3.125459338543
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 17
THE NERON-TATE HEIGHT ON ELLIPTIC CURVES 367
Table 2 (continued)
al - 0 a2 = 0 a3 - 0 a4 = 0 aó » -26142912
PI » ( 26572 / 9 I 4329280 / 27 )
P2 - ( 61516 / 25 ( 15244064 / 125 )
The trans-formation with (rlsitju) - ( 0 I 0 I 0 ) 2 ) leads to
al » 0 a2 - 0 a3 - 0 a4 ■ 0 aó = -408483
Pl-( 6643 / 9 ; 541160 / 27 )
p the local height
2 (1/3 )#ln( 2 )3 (25/12 )*ln( 3 )41 (1/3 )*ln( 41 )ee
The global height is 4.431508332622
decimal
.2310490601862.2887756013911.237857355568.673826315477
P2»( 15379 / 25 ¡ 1905508 / 125 )
p the local height
2 (1/3 )*ln( 2 )3 (13/12 )*ln< 3 )5 (1/1 )*ln( 5 )41 (1/3 )*ln( 41 )OS
The global height is 4.857361833623
decimal
.2310490601861.1901633127231.6094379124341.2378573555Ó8.588854192712
Pl+P2=( 133393 / 784 ¡ 46655225 / 21952 )
p the local height decimal
2 (7/3 )*ln< 2 )3 (13/12 )*ln( 3 )7 (1/1 )*ln( 7 )41 (1/3 )*ln( 41 )
The global height is 6.030575031739
1.6173434213061.1901633127231.9459101490551.237857355568.039300793087
Regulator : 18.87131764437
for P, Q g £(Q) represents a symmetric bilinear form on P(Q). If E has rank two
over Q and P = Px, Q = P2 are two basis points of E(Q), the quantity
Ä=|det(j8(P,,P7)),._1> R
is called the regulator of the elliptic curve E over Q. In addition to the Néron-Tate
height of the basis points P,, P2 of the rank-two curves E in Selmer's tables [8], we
have also computed their regulator R. More detailed information about Selmer's
curves is to be found in [6]. To begin with, we list in detail two examples, namely the
curves with A = 30 and A = 246 in [8] (see Table 2).
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 18
368 HEINZ M.TSCHÖPE AND HORST G. ZIMMER
In analogy to Silverman [9], we now use these Selmer curves to estimate the
constants in Lang's Conjecture 3. Suppose E over Q is given in Weierstrass normal
form
(E) y2 = x3 + ax + b (a,6eZ).
Following Lang [6], we define the height of E over Q to be the number
H(E) = max{|a|3, |è|2},
so that approximately
h(E)= -log/7(£)-6ftl,oo,
where again vx = - log | | denotes the additive archimedean valuation of Q. Let N
stand for the conductor of E over Q (see [11]).
Then we enunciate, in the case of rank-two curves,
Lang's Conjecture 3. There is a basis {PVP2} of E(Q) modulo torsion such
that 8(PX) 4 8(P2) and
8(PX) 4 cxH(E)l/2A ■ N«^2 ■ logiV -(2/f3)l/2,
8(P2) 4 c2H(E)1/l2 ■ NciN) • logJV • c
for some positive real constants c, cx, c2, where
lim e(N) = 0.N -»oo
Now the constants cx sind c2 in Lang's Conjecture 3 satisfy the inequalities
/ H(E)l/24 ■ N*»"2 ■ \ogN-(2/]/3)l/2Y1
c^[-m-j '
(H(E)l/12-N«N)-logN-cYl
C2>\ t{P2) j •
On choosing c = 1 and putting, in analogy to the example on p. 166 of [6],
e(N) = (\ogN-\oglogN)~1/2,
we obtain for the constants cx and c2 the estimates^
cx > 0.021,784,..., c2> 0.002,709,....
Here we let E range over the rank-two curves in [8] and take the maximal values for
cx said c2, which are attained at the curves with A = 246 and A = 30, respectively.
For the sake of completeness, we include here the numerical estimates of the
constants cx and c2 for all values of A in Selmer's table [8] in order to show how cx
and c2 oscillate as A varies (see Table 3).
1t This estimation is based on the assumption that the points in Selmer's table [8] are of minimal height.
We wish to thank M. Reichert for verifying this on a Siemens PC MX-2 for Selmer's curves with A = 30,
37, 65, 91, 110, 124, 126, 163, 182, 217, 254, 342, 468 and 469. Only for A = 254, the point P, + P2 is to
be taken instead of P-, since it has a slightly smaller height value.
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 19
THE NERON-TATE HEIGHT ON ELLIPTIC CURVES 369
Table 3
19
3037658691
110
12k126127I32153163182183201
203209210
217
218219246254271273282
309335342345348370379390397399407420433435436446^53462468469477497498
O.OO513126O.OI927030.00483937O.OO345923O.OO629273O.OO36522O.OI2277O.OO3742650.004330630.00414683O.OI830990.01053350.004581410.004450210.005247470.005116710.00954740.007217880.01260480.00308153O.OO3271990.005001260.02178430.005313650.003706660.004721230.01828640.00457362O.OO274443O.OO352578O.O1422140.01757280.00282848O.OO384730.01118180.003491610.0042891O.OO2675020.0121 164O.OO5185120.01394960.004828680.00418040.004157060.0108269O.OO388973O.OO339982O.OO744997O.OO5312930.0171181
0.001103890.002709030.0009623820.001825010.002279750.0004383370.001538550.001749930.001142670.0009409480.002447980.001111880.0004830740.0005170170.001995030.001209150.001164550.0008122240.00121204
O.OOO3I9989O.OO2OIO74O.OOI7I349O.OOI939O5O.OOO936257O.OOO782067O.OOO663O38O.OOI83782O.OO227242O.OOI65495O.OOIO3604O.OOI4I568O.OOI9I220.001179460.0007112330.0008590680.0005969490.0006248690.001186860.001075260.0005229350.001475410.001600660.001450410.001685460.0008779930.000981010.0003044540.00111861O.OOO907821O.OO14825
Fachbereich 9 Mathematik
Universität des Saarlandes
D-6600 Saarbrücken. West Germany
1. A. Bremner&J. W. S. Cassels, "On the equation Y2 = X2(X + p)," Math. Comp., v. 42, 1984,
pp. 257-264.2. J. P. Buhler, B. H. Gross&D. B Zagier, "On the conjecture of Birch and Swinnerton-Dyer for
an ellliptic curve of rank 3," Math. Comp., v. 44,1985, pp. 473-481.
3. J. W. S. Cassels, "Diophantine equations with special reference to elliptic curves," J. London
Math. Soc, v. 41, 1966, pp. 193-291.
4. H. G. FOLZ. Ein Beschränktheitssatz fur die Torsion von 2-defizienten elliptischen Kurven über
algebraischen Zahlkörpern, Ph.D. Thesis, Saarbrücken, 1985.
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
Page 20
370 HEINZ M.TSCHÖPE AND HORST G. ZIMMER
5. S. Lang, Elliptic Curves: Diophantine Analysis, Springer-Verlag, Berlin and New York, 1978.
6. S. Lang, "Conjectured Diophantine estimates on elliptic curves," Progr. Math., v. 35, 1983, pp.
155-171.7. S. Lang, Fundamentals of Diophantine Geometry, Springer-Verlag, Berlin and New York, 1983.
8. E. Selmer, "The Diophantine equation ax2 + by3 + cz3," Acta Math., v. 85,1951, pp. 203-362.
9. J. H. Silverman, "Lower bound for the canonical height on elliptic curves," Duke Math. J., v. 48,
1981, pp. 633-648.
10. J. T. Täte, "The arithmetic of elliptic curves," Invent. Math., v. 23,1974, pp. 179-206.
11. J. T. Täte, "Algorithm for finding the type of a singular fibre in an elliptic pencil," in Modular
Functions of One Variable IV, Lecture Notes in Math., vol. 476, Springer-Verlag, Berlin and New York,
1975, pp. 33-52.12. J. T. Täte, Letter to J.-P. Serre, Oct. 1,1979.
13. H. G. Zimmer, "On the difference of the Weil height and the Néron-Tate height," Math. Z., v. 147,
1976, pp. 35-51.14. H. G. Zimmer, "Quasifunctions on elliptic curves over local fields," J. Reine Angew. Math., v.
307/308,1979, pp. 221-246.15. H. G. Zimmer, "Torsion points on elliptic curves over a global field," Manuscripta Math., v. 29,
1979, pp. 119-145.
16. Modular Functions of One Variable IV (B. J. Birch & W. Kuyk, eds). Lecture Notes in Math., vol.
476, Springer-Verlag, Berlin and New York, 1975.
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use