Taiwanese Journal of Mathematics Paper on Elliptic Integrals

TAIWANESE J OURNAL OF MATHEMATICS
Vol. 10, No. 6, pp. 1633-1660, December 2006

This paper is availableonlineat http://www.math.nthu.edu.tw/tjm/
EVALUATION OF COMPLETE ELLIPTIC INTEGRALS
OF THE FIRST KIND AT SINGULAR MODULI
HabibMuzaffar andKennethS. Williams
Abstract. Thecompleteellipticintegral of thefirst kindK(k) is definedfor
0 < k < 1 by
K(k) :=
_
2
0
d
_
1 k
2
sin
2
.
Thereal number k is called themodulus of theellipticintegral. Thecomple-
mentary modulus isk
= (1 k
2
)
1
2
(0 < k
< 1). Let beapositiveinteger.

Theequation
K(k
) =
K(k)
definesauniquereal number k() (0 < k() < 1) calledthesingular modulus
of K(k). Let H(D) denotetheformclass groupof discriminant D. Let d be
thediscriminant 4. Using somerecent results of theauthors on values of
the Dedekind eta function at quadratic irrationalities, a formula is given for
the singular modulus k() in terms of quantities depending upon H(4d) if
0 (mod 2); H(d) and H(4d) if 1 (mod 4); H(d/4) and H(4d) if
3 (mod4). Similarly aformulais givenfor thecompleteellipticintegral
K[
] := K(k()) intermsof quantitiesdependinguponH(d) andH(4d) if

0 (mod2); H(d) if 1 (mod4); H(d/4) andH(d) if 3 (mod4).
As an examplethecompleteellipticintegral K[
17] is determinedexplicitly
interms of gammavalues.
1. INTRODUCTION
Let k R besuch that
(1.1) 0 < k < 1.
Received J uly 19, 2004, acceptedSeptember 16, 2005.
Communicatedby Wen-Ching, WinnieLi.
2000Mathematics Subject Classification: Primary 11F20. Secondary 11E16, 11E25.
Key words and phrases: Completeelliptic integral of thefirst kind; Singular modulus; Dedekindeta
function; Webers functions.
Researchof thesecondauthor was supportedby Natural Sciences andEngineeringResearchCouncil
of Canadagrant A-7233.
1633
1634 HabibMuzaffar andKennethS. Williams
Thecompleteelliptic integral K(k) of thefirst kindis definedby
(1.2) K(k) :=
_
2
0
d
_
1 k
2
sin
2
=
_
1
0
dt
_
(1 t
2
)(1 k
2
t
2
)
.
Clearly
lim
k0
+
K(k) =

2
, lim
k1
K(k) = +.
Thequantityk iscalledthemodulusof theellipticintegral K(k). Thecomplemen-
tary modulusk
is definedby
(1.3) k
:=
_
1 k
2
.
From(1.1) and(1.3) weseethat
(1.4) 0 < k
< 1.
Thecompleteelliptic integral K(k
) of modulus k
is denotedby K
(k) so that
(1.5) K
(k) = K(k
) = K(
_
1 k
2
)
and
(1.6) lim
k0
+
K
(k) = +, lim
k1
(k) =

2
.
Let N. As k increases from0 to 1, the function K
(k)/K(k) decreases
from+ to 0. Hencethereis a uniquemodulus k = k() with 0 < k < 1 such
that
(1.7)
K
(k)
K(k)
=
.
Thereal number k() iscalledthesingular moduluscorrespondingto. Thevalue
of the complete elliptic integral K(k) of the first kind at the singular modulus
k = k() is denotedby
(1.8) K[
] := K(k()).
Thefirst fivesingular moduli are
k(1) =
1
2
,
k(2) =

2 1,
k(3) =
3 1
8
,
Evalation of CompleteElliptic Integrals 1635
k(4) = 3 2
2,
k(5) =
_
5 1
_
3
5
2
,
seefor example[1, p. 139]. Thevalues of K[
] for = 1, 2, . . . , 16 are given

in[1, Table9.1, p. 298]. Other values can befoundscatteredintheliterature. For
examplein [2, p. 277] thevalues
(1.9) k(22) = 99 70
2 + 30
11 + 21
22
and
(1.10) K[
22] =2
5/2
11
1/2
(7+5
2+3
22)
1/2
1/2
_
88
m=1
_
m
88
_
(
88
m
)
_
1/4
are given, where(x) is thegammafunction and
_
d
n
_
is theKronecker symbol.
Thevalues of k(25) andK[
25] aregivenin [5, p. 259].

Let H(D) denotetheformclassgroupof discriminantD. Let d bethediscrim-
inant 4. Usingsomerecent resultsof theauthors onvalues of theDedekindeta
functionat quadraticirrationalities, aformulaisgivenfor thesingular modulusk()
intermsof quantitiesdependinguponH(4d) if 0 (mod2); H(d) andH(4d) if
1 (mod4); H(d/4) andH(4d) if 3 (mod4), seeTheorem1inSection4.
Similarly a formula is given for thecompleteelliptic integral K[
] := K(k())
in terms of quantities depending upon H(d) and H(4d) if 0 (mod 2); H(d)
if 1 (mod 4); H(d/4) and H(d) if 3 (mod 4), seeTheorem1 in Section
4. Zucker [5, p. 258] has determined but not publishedthevalues of K[
] for
= 17, 18, 19 and 20, so as an example we determine explicitly the complete
elliptic integral K[
17] in terms of gamma values, see Theorem2 in Section 5.

Our methodis different fromthat of Zucker.
2. PRELIMINARY RESULTS
Let N and set
(2.1) q = e
so that 0 < q < 1. Wedefine

(2.2) Q
0
:=
n=1
(1 q
2n
),
(2.3) Q
1
:=
n=1
(1 + q
2n
),
(2.4) Q
2
:=
n=1
(1 +q
2n1
),
(2.5) Q
3
:=
n=1
(1 q
2n1
).
Since
Q
1
Q
2
=
n=1
(1 +q
n
), Q
0
Q
3
=
n=1
(1 q
n
),
wehave
Q
0
Q
1
Q
2
Q
3
=
n=1
(1 q
2n
) = Q
0
,
so that
(2.6) Q
1
Q
2
Q
3
= 1.
J acobi [3] [4, p. 147] has shownthat
(2.7) 16qQ
8
1
+ Q
8
3
= Q
8
2
.
Hehasalsoshownthat thesingular modulusk = k(), thecomplementarysingular
modulus k
(), andthecompleteelliptic integral K[
] = K(k()) aregivenby
(2.8) k() = 4
q
_
Q
1
Q
2
_
4
,
(2.9) k
() =
_
Q
3
Q
2
_
4
,
and
(2.10) K[
] =

2
_
Q
0
Q
2
Q
1
Q
3
_
2
,
see[3] [4, p. 146]. Next werecall that theDedekindetafunction(z) is defined
by
(2.11) (z) := e
iz/12
m=1
(1 e
2imz
), z C, Im(z) > 0,
and that Webers functions f(z), f
1
(z) and f
2
(z) are defined in terms of the
Dedekindetafunctionby
(2.12) f(z) : = e
i/24
_
1 +z
2
_
(z)
,
(2.13) f
1
(z) :=
_
z
2
_
(z)
,
(2.14) f
2
(z) :=
2
(2z)
(z)
,
see[9, p. 114]. It is convenient to set
f
0
(z) := f(z)
so that f
j
(z) isdefinedfor j = 0, 1, 2. From(2.1)-(2.5) and(2.11), wededucethat
(2.15) (
) = q
1/12
Q
0
,
(2.16) (2
) = q
1/6
Q
0
Q
1
,
(2.17) (
/2) = q
1/24
Q
0
Q
3
,
(2.18) ((1 +
)/2) = e
i/24
q
1/24
Q
0
Q
2
.
From(2.12)-(2.18) weobtain
(2.19) Q
0
= q
1/12
(
),
(2.20) Q
1
= 2
1/2
q
1/12
f
2
(
),
(2.21) Q
2
= q
1/24
f
0
(
),
(2.22) Q
3
= q
1/24
f
1
(
).
Then, from(2.6), (2.7), (2.20), (2.21) and (2.22), weobtain
(2.23) f
0
(
)f
1
(
)f
2
(
) =
2
and
(2.24) f
0
(
)
8
= f
1
(
)
8
+f
2
(
)
8
,
see[9, p, 114]. Then, from(2.8), (2.10) and (2.19) (2.23), weobtain k() and
K[
] interms of , namely,
(2.25) k() =
_
f
2
(
)
f
0
(
)
_
4
and
(2.26) K[
] =

2
(
)
2
f
0
(
)
4
.
Recent resultsof Muzaffar andWilliams[6] givethevaluesof (
), f
0
(
),
f
1
(
) andf
2
(
) for all N, seeSection3. Usingthesevalues in (2.25)

and (2.26), weobtainthesingular modulus k() and thecompleteelliptic integral
of thefirst kindK[
] inSection4.
3. EVALUATION OF (
), f
0
(
), f
1
(
) and f
2
(
)
Let d bean integer satisfying
(3.1) d < 0, d 0 or 1 (mod 4).
Let f bethelargest positiveinteger such that
(3.2) f
2
| d, d/f
2
0 or 1 (mod4).
Weset = d/f
2
Z so that
(3.3) d = f
2
, 0, 1 (mod 4) .
For aprimep, thenonnegativeinteger v
p
(f) is definedby p
vp(f)
| f, p
vp(f)+1
f.
Weset
(3.4)
p
(, f) =
_
p
vp(f)
1
_
_
1
_
p
__
p
vp(f)1
(p 1)
_
p
_
p
__,
where
_
p
_
is theLegendresymbol modulo p. Thequantity
p
(, f) is used in
Proposition1 below.
Thepositive-definite, primitive, integral, binary quadraticformax
2
+bxy +cy
2
isdenotedby (a, b, c). Itsdiscriminant isthequantityd = b
2
4ac, whichsatisfies
(3.1). Theclass of theform(a, b, c) is
(3.5) [a, b, c] = {(a(p, r), b(p, q, r, s), c(q, s)) | p, q, r, s Z, ps qr = 1} ,
where
a(p, r) = ap
2
+bpr+cr
2
, b(p, q, r, s) =2apq+bps+bqr+2crs, c(q, s) = aq
2
+bqs+cs
2
.
Thegroupof classesof positive-definite, primitive, integral, binary quadratic forms
of discriminant d under Gaussiancompositionisdenotedby H(d). H(d) isafinite
abelian group. We denoteits order by h(d). Theidentity I of thegroup H(d) is
theprincipal class
(3.6) I =
_
[1, 0, d/4] , if d 0 (mod 4),
[1, 1, (1 d)/4] , if d 1 (mod 4).
The inverse of the class K = [a, b, c] H(d) is the class K
1
= [a, b, c]
H(d). If p is a prime with
_
d
p
_
= 1, we let h
1
and h
2
be the solutions of
h
2
d (mod 4p), 0 h < 2p, withh
1
< h
2
. Theclass K
p
of H(d) is definedby
K
p
=
_
p, h
1
,
h
2
1
d
4p
_
.
Then
K
1
p
=
_
p, h
1
,
h
2
1
d
4p
_
=
_
p, h
2
,
h
2
2
d
4p
_
,
as h
1
+ h
2
= 2p. If p is a primewith
_
d
p
_
= 0, p f, theclass K
p
of H(d) is
definedby
K
p
=
_
_
[p, 0, d/4p], if p > 2, d 0 (mod 4),
[p, p, (p
2
d)/4p], if p > 2, d 1 (mod 4),
[2, 0, d/8], if p = 2, d 8 (mod 16),
[2, 2, (4 d)/8], if p = 2, d 12 (mod 16),
so that K
p
= K
1
p
. Wedo not defineK
p
for any other primes p.
As H(d) is a finiteabelian group, there exist positive integers h
1
, h
2
, . . . , h
and generators A
1
, A
2
, . . . , A
H(d) suchthat
h
1
h
2
h
= h(d), 1 < h
1
| h
2
| . . . | h
, ord(A
i
) = h
i
(i = 1, . . . , ),
and, for each K H(d), thereexist uniqueintegersk
1
, . . . , k
with
K = A
k
1
1
A
k
(0 k
j
< h
j
, j = 1, . . . , ).
We fix once and for all the generators A
1
, . . . , A
of the group H(d). For j =

1, . . . , weset
ind
A
j
(K) := k
j
,
and for K, L H(d), wedefine : H(d) H(d)
h
(group of h
th roots
of unity) by
(K, L) = e
2i
j=1
ind
A
j
(K) ind
A
j
(L)
h
j
.
Thefunction has theproperties
(K, L) = (L, K), for all K, L H(d),
(K, I) = 1, for all K H(d),
(KL, M) = (K, M)(L, M), for all K, L, M H(d),
(K
r
, L
s
) = (K, L)
rs
, for all K, L H(d) andall r, s Z,
see[6, Lemma6.2]. It is knownthat for K(= I) H(d) thelimit
(3.7) j(K, d) = lim
s1
+
p
(
d
p
)
=1
_
1
(K, K
p
)
p
s
__
1
(K
1
, K
p
)
p
s
_
existsandis anonzero real number suchthat j(K, d) = j(K
1
, d), see[6, Lemma
7.6]. For n N andL H(d) wedefine
H
L
(n) := card{h | 0 h < 2n, h
2
d (mod4n),
_
n, h,
h
2
d
4n
_
= L}.
The properties of H
L
(n) are developed in [6, Section 5]. Then, for n N and
K H(d), weset
Y
K
(n) :=

LH(d)
(K, L)H
L
(n).
Properties of Y
K
(n) aregiven in[6, Section 7]. Further, for aprimep and aclass
K(= I) H(d), weset
(3.8) A(K, d, p) =
j=0
Y
K
(p
j
)
p
j
.
Next, for K(= I) H(d), weset
(3.9) l(K, d) =
p|d
pf
_
1 +
(K, K
p
)
p
_
p|f
A(K, d, p),
where theproducts are over all primes p satisfying thestated conditions. Finally,
for K H(d), wedefine
(3.10) E(K, d) =

_
|d|w(d)
48h(d)
LH(d)
L=I
(L, K)
1
t
1
(d)
j(L, d)
l(L, d),
see[6, Section9], where
(3.11) w(d) = 6, 4 or 2 according as d = 3, d = 4 or d < 4,
and
(3.12) t
1
(d) :=

p
_
d
p
_
= 1
_
1
1
p
2
_
.
Thefollowingevaluationof (
) followsimmediatelyfrom[6, Theorem1],
as (
) is real and positivefor N.

Proposition 1. Let N. Let d = 4 = f
2
, where and f are defined
in (3.2) and (3.3). Let K = [1, 0, ] H(d). Then
(
) = 2
3/4
1/4
1/4
p|f
p
p(,f)/4
_
_
||
m=1
_
m
||
_
(
m
)
_
_
w()
8h()
e
E(K,d)
,
where
p
(, f) is defined in (3.4) and
_
m
_
is theKronecker symbol.
Thefollowingresult is Theorem3 of [6].
Proposition 2. Let N. Let d = 4. Let K = [1, 0, ] H(d).
(a) 0 (mod 4). Set
M
0
= [4, 4, + 1] H(4d),
M
1
=
_
1, 0,

4
_
H
_
d
4
_
,
M
2
= [1, 0, 4] H(4d).
Let = 4
, where is a positive integer and 1, 2 or 3 (mod 4).

(i) 1 or 2 (mod 4) (so that is even and v
2
(f) = ). We have
f
0
(
) = 2
1
2
+3
e
E(K,d)E(M
0
,4d)
,
f
1
(
) = 2
2
+1
1
2
+2
e
E(K,d)E(M
1
,d/4)
,
f
2
(
) = 2
1
2
+3
e
E(K,d)E(M
2
,4d)
.
(ii) 3 (mod4) (so that (mod8) and v
2
(f) = + 1). If
3 (mod 8), we have
f
0
(
) = 2
1
32
+2
e
E(K,d)E(M
0
,4d)
,
f
1
(
) = 2
32
1
3.2
+1
e
E(K,d)E(M
1
,d/4)
,
f
2
(
) = 2
1
32
+2
e
E(K,d)E(M
2
,4d)
.
If 7 (mod 8), we have
f
0
(
) = e
E(K,d)E(M
0
,4d)
,
f
1
(
) =

2e
E(K,d)E(M
1
,d/4)
,
f
2
(
) = e
E(K,d)E(M
2
,4d)
.
(b) 1 (mod 4) (so that is even and f is odd). Set
M
0
=
_
2, 2,
+ 1
2
_
H(d),
M
1
= [4, 0, ] H(4d),
M
2
= [1, 0, 4] H(4d).
Then
f
0
(
) = 2
1/4
e
E(K,d)E(M
0
,d)
,
f
1
(
) = 2
1/8
e
E(K,d)E(M
1
,4d)
,
f
2
(
) = 2
1/8
e
E(K,d)E(M
2
,4d)
.
(c) 2 (mod 4) (so that is even and f is odd). Set
M
0
= [4, 4, + 1] H(4d),
M
1
=
_
2, 0,

2
_
H(d),
M
2
= [1, 0, 4] H(4d).
Then
f
0
(
) = 2
1/8
e
E(K,d)E(M
0
,4d)
,
f
1
(
) = 2
1/4
e
E(K,d)E(M
1
,d)
,
f
2
(
) = 2
1/8
e
E(K,d)E(M
2
,4d)
.
(d) 3 (mod 4) (so that (mod 8) and f 2 (mod 4)). Set
M
0
=
_
1, 1,
+ 1
4
_
H
_
d
4
_
,
M
1
= [4, 0, ] H(4d),
M
2
= [1, 0, 4] H(4d).
Then, for 3 (mod 8), we have
f
0
(
) = 2
1/3
e
E(K,d)E(M
0
,d/4)
,
f
1
(
) = 2
1/12
e
E(K,d)E(M
1
,4d)
,
f
2
(
) = 2
1/12
e
E(K,d)E(M
2
,4d)
,
and, for 7 (mod 8), we have
f
0
(
) =

2e
E(K,d)E(M
0
,d/4)
,
f
1
(
) = e
E(K,d)E(M
1
,4d)
,
f
2
(
) = e
E(K,d)E(M
2
,4d)
.
4. FORMULAE FOR k() AND K[
]
From(2.25), (2.26), Proposition1 andProposition2, weobtainthemainresult
of thispaper, namely, theformulaefor thesingular modulusk() andthecomplete
elliptic integral of the first kind K[
] at the singular modulus valid for every

N.
Theorem 1. Let N. Let d = 4. Let K = [1, 0, ] H(d).
(a) 0 (mod 4). Set
M
0
= [4, 4, + 1] H(4d), M
2
= [1, 0, 4] H(4d).
Then
k() = e
4(E(M
0
,4d)E(M
2
,4d))
.
Let = 4
, where is a positive integer and 1, 2 or 3 (mod 4). Then

K[
] = 2
1/2
1/2
p|f
p
p(,f)/2
_
_
||
m=1
_
m
||
_
(
m
)
_
_
w()
4h()
e
2E(K,d)4E(M
0
,4d)
,
where
=
_
_
1
2
+1

5
2
, if 1 or 2 (mod 4),
1
3 2

5
2
, if 3 (mod 8),
5
2
, if 7 (mod 8),
(b) 1 (mod 4). Set
M
0
=
_
2, 2,
+ 1
2
_
H(d), M
2
= [1, 0, 4] H(4d).
Then
k() = 2
1/2
e
4(E(M
0
,d)E(M
2
,4d))
and
K[
] = 2
3/2
1/2
1/2
p|f
p
p(,f)/2
_
_
||
m=1
_
m
||
_
(
m
)
_
_
w()
4h()
e
2E(K,d)4E(M
0
,d)
.
(c) 2 (mod 4). Set
M
0
= [4, 4, + 1] H(4d), M
2
= [1, 0, 4] H(4d).
Then
k() = e
4(E(M
0
,4d)E(M
2
,4d))
and
K[
] = 2
2
1/2
1/2
p|f
p
p(,f)/2
_
_
||
m=1
_
m
||
_
(
m
)
_
_
w()
4h()
e
2E(K,d)4E(M
0
,4d)
.
(d) 3 (mod 4). Set
M
0
=
_
1, 1,
+ 1
4
_
H
_
d
4
_
, M
2
= [1, 0, 4] H(4d).
Then, for 3 (mod 8), we have
k() = 2
1
e
4(E(M
0
,d/4)E(M
2
,4d))
and
K[
] = 2
7/6
1/2
1/2
p|f
p
p(,f)/2
_
_
||
m=1
_
m
||
_
(
m
)
_
_
w()
4h()
e
2E(K,d)4E(M
0
,d/4)
,
and, for 7 (mod 8), we have
k() = 2
2
e
4(E(M
0
,d/4)E(M
2
,4d))
and
K[
] = 2
1/2
1/2
1/2
p|f
p
p(,f)/2
_
_
||
m=1
_
m
||
_
(
m
)
_
_
w()
4h()
e
2E(K,d)4E(M
0
,d/4)
.
5. EVALUATION OF K[
17]
In this section we use Theorem1 to evaluatethe completeelliptic integral of
the first kind K[
17]. Thus = 17, d = 4 = 68, = 68 and f = 1 in

thenotationof Sections3and4. ThegroupH(68) of classesof positive-definite,
primitive, integral binary quadratic formsof discriminant 68 under compositionis
H(68) = {I, A, A
2
, A
3
}, A
4
= I,
where
I = [1, 0, 17], A = [3, 2, 6], A
2
= [2, 2, 9], A
3
= [3, 2, 6].
In order to determine K[
17] explicitly using Theorem 1, we must determine

E(I, 68) and E(A
2
, 68) (see Lemma 14). This requires finding j(A
m
, 68)
(m = 1, 2, 3) (seeLemma13). To computej(A
m
, 68) (m = 1, 2, 3) from(3.7)
wemust determinethoseprimes p satisfying
_
1
p
_
=
_
p
17
_
= 1 for which K
p
= I
andthosefor whichK
p
= A
2
. Thisdependsuponwhetherp isof theformx
2
+17y
2
for integersx andy or of theform2x
2
+2xy +9y
2
for integersx andy. By class
fieldtheory theformer occursif andonly if thequarticpolynomial x
4
+x
2
+2x+1
istheproduct of four linear factors(modp). Thisleadsustoconsider thearithmetic
of thefieldK = Q(), where is aroot of x
4
+x
2
+ 2x + 1.
Let f(x) betheirreduciblequartic polynomial givenby
(5.1) f(x) = x
4
+x
2
+ 2x + 1 Z[x].
Thediscriminant of f(x) is 272 = 2
4
17 andits Galois groupis D
8
(thedihedral
group of order 8) [8, p. 441]. Thefour roots of f(x) are
1
2
(i + (1 + 4i)
1
2
),
1
2
(i (1 + 4i)
1
2
),
1
2
(i + (1 4i)
1
2
),
1
2
(i (1 4i)
1
2
),
wherez
1
2
denotes theprincipal valueof thesquareroot of thecomplex number z.
Let
=
1
2
(i + (1 + 4i)
1
2
)
and set
(5.2) K = Q()
so that K is the totally complex quartic field Q((1 + 4i)
1
2
). Thus the number
of real embeddings of K is r
1
= 0 and the number of imaginary embeddings is
2r
2
= 4. Thering of integersof K is
(5.3) O
K
= Z + Z +Z
2
+Z
3
,
see[8, p. 441]. As K is monogenic, its discriminant d(K) = disc(f(x)) = 272.
It is known that O
K
has classnumber h
K
= 1 [8, p. 435] so that it is a principal
ideal domain. Asr
1
+r
2
1 = 0 +2 1 = 1 weknowby Dirichletsunit theorem
that O
K
has asinglefundamental unit. This unit can betakento be [8, p. 441].
Theregulator
R(K) =2 log|| =log
i + (1 + 4i)
1
2
2
2
=log
_
1 +
_
2 + 2
17+
17
4
_
0.732,
see[8, p. 441]. Thequartic fieldK contains auniquesubfield(= Q, K), namely,
Q(i). Theonly roots of unity in O
K
are1 and i. Thus thenumber of roots of
unity in O
K
is w(K) = 4.
Wenowgivethefactorization of f(x) modulo aprimep. Weusethenotation
(m) todenoteamonicirreduciblepolynomial of degreem withinteger coefficients.
Thus g(x) (2)(2) (mod p) means that g(x) is theproduct of two distinct monic
irreduciblequadratic polynomials modulo p and h(x) (2)
2
means that h(x) is
thesquareof amonic irreduciblequadratic polynomial modulo p. Fromclass field
theory or indeed by elementary arguments one can show that the factorization of
f(x) (mod p), wherep is aprime= 2, 17, is givenas follows:
If
_
1
p
_
=
_
p
17
_
= 1 and p = u
2
+ 17v
2
for someintegersu and v
then
f(x) (1)(1)(1)(1) (mod p).
If
_
1
p
_
=
_
p
17
_
= 1 and p = 2u
2
+ 2uv + 9v
2
for someintegersu and v
then
f(x) (2)(2) (mod p).
If
_
1
p
_
= 1,
_
p
17
_
= 1
then
f(x) (2)(2) (mod p).
If
_
1
p
_
= 1,
_
p
17
_
= 1
then
f(x) (1)(1)(2) (mod p).
If
_
1
p
_
=
_
p
17
_
= 1
then
f(x) (4) (mod p).
For p = 2
f(x) (2)
2
(mod 2)
and for p = 17
f(x) (1)(1)(1)
2
(mod 17).
Using theseresults, a standard algebraic number theoretic argument gives thefac-
torizationof theprincipal ideal pO
K
into primeidealsin O
K
, wherep is aprime.
Lemma 1. Let p be a prime = 2, 17.
(i) If
_
1
p
_
=
_
17
p
_
= 1 and p = x
2
+ 17y
2
for some integers x and y
then
pO
K
= PQRS, N(P) = N(Q) = N(R) = N(S) = p,
where P, Q, R, S are distinct prime ideals of O
K
.
(ii) If
_
1
p
_
=
_
17
p
_
= 1 and p = 2x
2
+ 2xy + 9y
2
for some integers x and y
then
pO
K
= PQ, N(P) = N(Q) = p
2
,
where P and Q are distinct prime ideals of O
K
.
(iii) If
_
1
p
_
= 1,
_
17
p
_
= 1
then
pO
K
= PQ, N(P) = N(Q) = p
2
,
where P and Q are distinct prime ideals of O
K
.
(iv) If
_
1
p
_
= 1,
_
17
p
_
= 1
then
pO
K
= PQR, N(P) = N(Q) = p, N(R) = p
2
,
where P, Q and R are distinct prime ideals of O
K
.
(v) If
_
1
p
_
=
_
17
p
_
= 1
then
pO
K
= P, N(P) = p
4
,
where P is a prime ideal.
(vi) 2O
K
= P
2
, N(P) = 2
2
, where P is a prime ideal.
(vii) 17O
K
= PQR
2
, N(P) = N(Q) = N(R) = 17, where P, Q and R are
distinct prime ideals.
The next lemmadetermines theclass K
p
of H(68) when p is a primesuch
that
_
68
p
_
= 1.
Lemma 2. Let p be a prime such that
_
68
p
_
= 1. Then
K
p
= I p = x
2
+ 17y
2
for some integers x and y,
K
p
= A
2
p = 2x
2
+ 2xy + 9y
2
for some integers x and y,
K
p
= A or A
3
p = 3x
2
2xy + 6y
2
for some integers x and y.
Proof. As
_
68
p
_
= 1 thereexist integersx and y such that
p = x
2
+ 17y
2
or 2x
2
+ 2xy + 9y
2
, if
_
1
p
_
=
_
17
p
_
= 1,
and such that
p = 3x
2
2xy + 6y
2
, if
_
1
p
_
=
_
17
p
_
= 1.
We recall that as p is a prime the only classes representing p are K
p
and K
p
1
.
Hence
p =x
2
+17y
2
=[1, 0, 17] representsp =I =K
p
or K
1
p
=K
p
=I,
p =2x
2
+2xy+9y
2
=[2, 2, 9] representsp =A
2
=K
p
or K
1
p
=K
p
=A
2
,
p =3x
2
2xy+6y
2
=[3, 2, 6] representsp =A
3
=K
p
or K
1
p
=K
p
=AorA
3
.
This completestheproof of Lemma2.
Definition 1. For s > 1 and , {1, +1} wedefine
A
,
(s) : =

p = 2, 17
_
1
p
_
= ,
_
17
p
_
=
_
1 +
1
p
s
_
1
and
B
,
(s) : =

p = 2, 17
_
1
p
_
= ,
_
17
p
_
=
_
1
1
p
s
_
1
.
For brevity we just writeA
+1,+1
(s), A
+1,1
(s), ... as A
++
(s), A
+
(s), ... re-
spectively. In viewof Lemmas 1 and 2 wecan split each of A
++
(s) and B
++
(s)
into two products as
A
++
(s) = A
++
(s)A
++
(s), B
++
= B
++
(s)B
++
(s),
where
A
++
(s) : =

p = 2, 17
Kp = I
_
1 +
1
p
s
_
1
, A
++
(s) : =

p = 2, 17
Kp = A
2
_
1 +
1
p
s
_
1
and
B
++
(s) : =

p = 2, 17
Kp = I
_
1
1
p
s
_
1
, B
++
(s) : =

p = 2, 17
Kp = A
2
_
1
1
p
s
_
1
.
Lemma 3. For s > 1 we have
A
,
(s) =
B
,
(2s)
B
,
(s)
, where , {1, +1},
and
A
++
(s) =
B
++
(2s)
B
++
(s)
, A
++
(s) =
B
++
(2s)
B
++
(s)
.
Proof. Wejust provethefirst equalityastheother twoequalitiescanbeproved
similarly. Wehave
A
,
(s)B
,
(s) =
p = 2, 17
_
1
p
_
= ,
_
17
p
_
=
_
1 +
1
p
s
_
1
p = 2, 17
_
1
p
_
= ,
_
17
p
_
=
_
1
1
p
s
_
1
=
p = 2, 17
_
1
p
_
= ,
_
17
p
_
=
_
1
1
p
2s
_
1
= B
,
(2s),
fromwhich theassertedresult follows.
For s > 1 theRiemannzetafunctionis givenby
(s) =
p
_
1
1
p
s
_
1
,
wheretheproduct is taken over all primes p. If D is an integer with D 0 or 1
(mod 4) theDirichlet L-series L(s, D) (s > 1) is given by
L(s, D) =
p
_
_
1
_
D
p
_
p
s
_
_
1
.
Weprove
(i) (s) =
_
1
1
2
s
_
1
_
1
1
17
s
_
1
B
(s)B
+
(s)B
+
(s)B
++
(s),
(ii) L(s, 4) =
_
1
1
17
s
_
1
B
(2s)
B
(s)
B
+
(2s)
B
+
(s)
B
+
(s)B
++
(s),
(iii) L(s, 17) =
_
1
1
2
s
_
1
B
(2s)
B
(s)
B
+
(s)
B
+
(2s)
B
+
(s)
B
++
(s),
(iv) L(s, 68) = B
(s)
B
+
(2s)
B
+
(s)
B
+
(2s)
B
+
(s)
B
++
(s).
Proof. We just givetheproofs of (i) and (ii). Equations (iii) and (iv) can be
proved similarly. Let
X = {(1, 1), (1, +1), (+1, 1), (+1, +1)}.
First weprove(i). Wehave
_
1
1
2
s
__
1
1
17
s
_
(s) =

p=2, 17
_
1
1
p
s
_
1
=
(,)X
p = 2, 17
_
1
p
_
=,
_
17
p
_
=
_
1
1
p
s
_
1
,
fromwhich(i) nowfollows by Definition1.
Next weprove(ii). Wehave
L(s, 4) =

p
_
_
1
_
4
p
_
p
s
_
_
1
=
_
1
1
17
s
_
1
p=2, 17
_
_
1
_
4
p
_
p
s
_
_
1
=
_
1
1
17
s
_
1
(,)X
p = 2, 17
_
1
p
_
= ,
_
17
p
_
=
_
1

p
s
_
1
=
_
1
1
17
s
_
1
A
(s)A
+
(s)B
+
(s)B
++
(s),
and (ii) follows usingLemma3.
B
(s)
4
= L(s, 4)
1
L(s, 17)
1
L(s, 68)B
(2s)
2
(s),
B
+
(s)
4
=
_
1
1
2
s
_
2
L(s, 4)
1
L(s, 17)L(s, 68)
1
B
+
(2s)
2
(s),
B
+
(s)
4
=
_
1
1
17
s
_
2
L(s, 4)L(s, 17)
1
L(s, 68)
1
B
+
(2s)
2
(s),
B
++
(s)
4
=
_
1
1
2
s
_
2
_
1
1
17
s
_
2
L(s, 4)L(s, 17)L(s, 68)
B
(2s)
2
B
+
(2s)
2
B
+
(2s)
2
(s),
Proof. We obtain the asserted equalities by solving the equations (i)-(iv) in
Lemma4 for B
(s), B
+
(s), B
+
(s) and B
++
(s).
TheDedekindzetafunctionfor thefieldK is givenby
K
(s) =
P
_
1
1
N(P)
s
_
1
,
wheretheproduct is takenover all primeidealsof O
K
.
K
(s) =
_
1
1
2
2s
_
1
_
1
1
17
s
_
3
B
(4s)B
+
(2s)
2
B
+
(2s)B
++
(2s)
2
B
+
(s)
2
B
++
(s)
4
.
Proof. We split
K
(s) into seven products and make use of Lemma 1 to
recognizeeach of theseproducts in terms of theB
,
. Wehave
K
(s) =
1
7
,
where
1
: =

P|2O
K
_
1
1
N(P)
s
_
1
=
_
1
1
4
s
_
1
=
_
1
1
2
2s
_
1
,
2
: =

P|17O
K
_
1
1
N(P)
s
_
1
=
_
1
1
17
s
_
3
,
3
: =

p = 2, 17
_
1
p
_
=
_
17
p
_
= 1
P|pO
K
_
1
1
N(P)
s
_
1
=

p = 2, 17
_
1
p
_
=
_
17
p
_
= 1
_
1
1
p
4s
_
1
= B
(4s),
4
: =

p = 2, 17
_
1
p
_
= 1,
_
17
p
_
= 1
P|pO
K
_
1
1
N(P)
s
_
1
=

p = 2, 17
_
1
p
_
= 1,
_
17
p
_
= 1
_
1
1
p
2s
_
2
= B
+
(2s)
2
,
5
: =

p = 2, 17
_
1
p
_
= 1,
_
17
p
_
= 1
P|pO
K
_
1
1
N(P)
s
_
1
=

p = 2, 17
_
1
p
_
= 1,
_
17
p
_
= 1
_
1
1
p
s
_
2
_
1
1
p
2s
_
1
= B
+
(s)
2
B
+
(2s),
6
: =

p = 2, 17
_
1
p
_
=
_
17
p
_
=1
p=x
2
+17y
2
P|pO
K
_
1
1
N(P)
s
_
1
=
p = 2, 17
_
1
p
_
=
_
17
p
_
=1
Kp=I
_
1
1
p
s
_
4
= B
++
(s)
4
,
7
: =
p = 2, 17
_
1
p
_
=
_
17
p
_
=1
p=2x
2
+2xy+9y
2
P|pO
K
_
1
1
N(P)
s
_
1
=

p = 2, 17
_
1
p
_
=
_
17
p
_
=1
Kp=A
2
_
1
1
p
2s
_
2
= B
++
(2s)
2
.
Multipying
1
,
2
, ... ,
7
together, weobtaintheassertedequality.
B
++
(s)
8
=
_
1
1
2
s
_
2
_
1 +
1
2
s
_
2
_
1
1
17
s
_
4
L(s, 4)
1
L(s, 17)L(s, 68)
B
(4s)
2
B
+
(2s)
4
B
+
(2s)
4
B
++
(2s)
4
K
(s)
2
(s)
1
,
B
++
(s)
8
=
_
1
1
2
s
_
2
_
1 +
1
2
s
_
2
L(s, 4)
3
L(s, 17)L(s, 68)
B
(4s)
2
B
(2s)
4
B
++
(2s)
4
K
(s)
2
(s)
3
.
Proof. The first equality follows by replacing B
+
(s)
4
in the square of the
equality in Lemma 6 by its value given in Lemma 5. The second equality then
follows fromB
++
(s)
8
B
++
(s)
8
= B
++
(s)
8
and the value of B
++
(s)
8
given by
Lemma5.
Lemma 8.
(i) B
(2)B
+
(2)B
+
(2)B
++
(2) =
36
2
289
,
(ii) t
1
(68) =
289
36
2
B
+
(2)B
+
(2).
Proof. By Lemma4(i) wehave(as (2) =
2
/6)
B
(2)B
+
(2)B
+
(2)B
++
(2) =
_
1
1
2
2
__
1
1
17
2
_
(2) =
36
289
2
,
whichis (i). Then
t
1
(68) =
1
B
(2)B
++
(2)
=
289
36
2
B
+
(2)B
+
(2).
by (3.12), Definition1and (i).
Lemma 9.
lim
s1
+
_
K
(s)
(s)
_
=

2
4
17
log
_
1 +
_
2 + 2
17 +
17
4
_
.
Proof. By [7, Theorem7.1, p. 326] wehave
lim
s1
+
(s 1)
K
(s) =
2
r
1
+r
2
r
2
R(K)h(K)
w(K)|d(K)|
1/2
=

2
4
17
log
_
1+
_
2+2
17+
17
4
_
.
As
lim
s1
+
(s 1)(s) = 1
theassertedresult follows.
Lemma 10.
L(1, 4) =

4
, L(1, 17) =
2
17
log(4 +
17), L(1, 68) =

2
17
.
Proof. Dirichlets class number formula [7, Theorem7.1, p. 326] for the
quadratic fieldQ(
d) of discriminant d asserts that

L(1, d) =
2h(d) log(d)
d
, if d > 0,
and
L(1, d) =
2h(d)
w(d)
_
|d|
, if d < 0,
where h(d) is the class number of Q(
d), (d) is the fundamental unit > 1 of

Q(
d) when d > 0, and w(d) = 2, 4 or 6 according as d < 4, d = 4 or

d = 3 whend < 0. As
h(4) = 1, h(17) = 1, h(68) = 4, (17) = 4 +
17
theassertedresult follows.
Lemma 11.
lim
s1
+
_
B
(s)
B
++
(s)
_
2
=
17
17B
(2)
2
B
+
(2)B
+
(2)
4 log(4 +
17)
.
Proof. By Lemma5 wehave
_
B
(s)
B
++
(s)
_
2
=
_
1
1
2
s
_
1
_
1
1
17
s
_
1
L(s, 4)
1
L(s, 17)
1
B
(2s)
2
B
+
(2s)B
+
(2s).
Lettings 1
+
and appealingto Lemma10, weobtaintheassertedlimit.
Lemma 12.
lim
s1
+
_
B
++
(s)
B
++
(s)
_
2
=
24
17
17
log
_
1 +
_
2 + 2
17 +
17
4
_
B
(4)
1
B
(2)B
+
(2)
1
B
+
(2)
1
B
++
(2)
2
.
Proof. By Lemma7 wehave
_
B
++
(s)
B
++
(s)
_
2
=
_
1+
1
2
s
__
1
1
17
s
_
L(s, 4)
1
B
(4s)
1
B
(2s)B
+
(2s)
1
B
+
(2s)
1
B
++
(2s)
2
_
K
(s)
(s)
_
.
Lettings 1+ andappealingtoLemmas9and10, weobtaintheassertedlimit.
Wenote(inthenotationof Section3) that
K
2
= [2, 2, 9] = A
2
,
K
17
= [17, 0, 1] = [1, 0, 17] = I,
(A
j
, A
k
) = i
jk
,
l(A
j
, 68) =
_
1 +
(A
j
, A
2
)
2
__
1 +
(A
j
, I)
17
_
=
_
1 +
(1)
j
2
__
1 +
1
17
_
=
_
_
9
17
, if j = 1, 3,
27
17
, if j = 2.
Lemma 13.
j(A, 68) = j(A
3
, 68) =
17
17B
+
(2)B
+
(2)
24 log
_
1 +
_
2 + 2
17 +
17
4
_,
j(A
2
, 68) =
17
17B
+
(2)B
+
(2)
4 log(4 +
17)
.
Proof. For r = 1, 2, 3 wehaveby (3.7)
j(A
r
, 68) = lim
s1
+
p = 2, 17
_
1
p
_
=
_
17
p
_
_
1
(A
r
, K
p
)
p
s
__
1
(A
r
, K
p
)
p
s
_
.
Thus, by Lemmas 1and 2, wehave
j(A
r
, 68) = lim
s1
+
_
1
p
_
=
_
17
p
_
= 1
Kp = I
_
1
1
p
s
_
2
_
1
p
_
=
_
17
p
_
= 1
Kp = A
2
_
1
(1)
r
p
s
_
2
_
1
p
_
=
_
17
p
_
=1
_
1
i
r
p
s
__
1
i
r
p
s
_
.
Hence
j(A
2
, 68) = lim
s1
+
B
++
(s)
2
A
(s)
2
= lim
s1
+
1
B
(2s)
2
_
B
(s)
B
++
(s)
_
2
(by Lemma3)
=
1
B
(2)
2
lim
s1
+
_
B
(s)
B
++
(s)
_
2
.
Thedeterminationof j(A
2
, 68) nowfollows by Lemma11.
Finally
j(A, 68) = j(A
3
, 68) = lim
s1
+
B
++
(s)
2
A
++
(s)
2
A
(2s)
1
= lim
s1
+
B
(2s)
B
++
(2s)
2
B
(4s)
_
B
++
(s)
B
++
(s)
_
2
(by Lemma3)
=
B
(2)
B
++
(2)
2
B
(4)
lim
s1
+
_
B
++
(s)
B
++
(s)
_
2
.
Thedeterminationof j(A, 68) nowfollows by Lemma12.
Lemma 14.
E(I, 68) =
1
4
log
_
1 +
_
2 + 2
17 +
17
4
_
+
1
16
log(4 +
17),
E(A
2
, 68) =
1
4
log
_
1 +
_
2 + 2
17 +
17
4
_
+
1
16
log(4 +
17).
Proof. From(3.10) wehavefor r = 0, 1, 2, 3
E(A
r
, 68) =

68w(68)
48h(68)
3
m=1
(A
m
, A
r
)
1
t
1
(68)
j(A
m
, 68)
l(A
m
, 68)
=
289
17
1728
B
+
(2)B
+
(2)
3
m=1
i
mr
l(A
m
, 68)
j(A
m
, 68)
(by Lemma8(ii))
=
17
17
192
B
+
(2)B
+
(2)
_
i
r
j(A, 68)
+ 3
i
2r
j(A
2
, 68)
+
i
3r
j(A
3
, 68)
_
.
Theassertedresultsnowfollowbytakingr = 0 andr = 2 andappealingtoLemma
13.
FromProposition2(b) and Lemma14weobtain
f
0
(
17) = 2
1/4
_
1 +
_
2 + 2
17 +
17
4
_
1/2
in agreement with[9, p. 721].
Theorem 2.
K[
17] = 2
9/2
17
1/2
1/2
(
17 4)
1/8
(1 +
_
2 + 2
17 +
17)
3/2
_
68
m=1

_
m
68
_
(
68
m
)
_
1/8
.
Proof. We apply Theorem1(b) with = 17 so that K = [1, 0, 17] = I and
M
0
= [2, 2, 9] = A
2
. Weobtain
K[
17] = 2
3/2
1/2
17
1/2
_
68
m=1
_
m
68
_
(
68
m
)
_
1/8
e
2E(I,68)4E(A
2
,68)
.
By Lemma14 wehave
2E(I, 68)4E(A
2
, 68) =
3
2
log
_
1 +
_
2 + 2
17 +
17
4
_
1
8
log(4+
17),
so that
e
2E(I,68)4E(A
2
,68)
=
_
1 +
_
2 + 2
17 +
17
4
_
3/2
(4 +
17)
1/8
= 2
3
(
17 4)
1/8
_
1 +
_
2 + 2
17 +
17
_
3/2
,
and Theorem2follows.
In a similar manner it can beshown that thesingular modulus k(17) is given
by
k(17) =
1
2
(
V ) = 0.006156 . . . ,
where
U = 21 + 5
17 8
_
2 + 2
17 6
_
2
17 2
and
V = 19 5
17 + 8
_
2 + 2
17 + 6
_
2
17 2.
REFERENCES
1. J . M. Borweinand P. B. Borwein, Pi and the AGM, Wiley, NewYork, 1987.
2. J . G. Huard, P. Kaplan, andK. S. Williams, TheChowla-Selbergformulafor genera,
Acta Arith., 73 (1995), 271-301.
3. C. G. J . J acobi, Fundamenta Nova Theoriae Functionum Ellipticarum, Regiomonti,
1829.
4. C. G. J . J acobi, Gesammelte Werke, Vol. I, Chelsea, NewYork, 1969, pp. 49-239.
5. G. S. J oyceandI. J . Zucker, Special valuesof thehypergeometricseries, Math. Proc.
Camb. Phil. Soc.. 109 (1991), 257-261.
6. H. Muzaffar and K. S. Williams, Evaluation of Webers functions at quadratic irra-
tionalities, JP J. Algebra Number Theory Appl., 4 (2004), 209-259.
7. W. Narkiewicz, Elementary and Analytic Theory of Algebraic Numbers, Springer-
Verlag, BerlinHeidelbergNewYork, 1990.
8. M. Pohst and H. Zassenhaus, Algorithmic Algebraic Nunber Theory, Cambridge
University Press, 1989.
9. H. Weber, Lehrbuch der Algebra, Vol III, Chelsea, NewYork, 1961.
Habib Muzaffar
Department of Mathematics,
University of Toledo,
Toledo, Ohio 43606, USA
E-mail: hmuzaff@utnet.utoledo.edu
KennethS. Williams
Centrefor Research inAlgebraandNumber Theory,
School of Mathematics and Statistics,
CarletonUniversity, Ottawa,
Ontario K1S 5B6, Canada
E-mail: kwilliam@connect.carleton.ca

Taiwanese Journal of Mathematics Paper on Elliptic Integrals

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Taiwanese Journal of Mathematics Paper on Elliptic Integrals

Încărcat de

Drepturi de autor:

Formate disponibile

TAIWANESE J OURNAL OF MATHEMATICS

Vol. 10, No. 6, pp. 1633-1660, December 2006

< 1). Let beapositiveinteger.

] := K(k()) intermsof quantitiesdependinguponH(d) andH(4d) if

] for = 1, 2, . . . , 16 are given

25] aregivenin [5, p. 259].

17] in terms of gamma values, see Theorem2 in Section 5.

so that 0 < q < 1. Wedefine

(), andthecompleteelliptic integral K[

) for all N, seeSection3. Usingthesevalues in (2.25)

of the group H(d). For j =

) is real and positivefor N.

, where is a positive integer and 1, 2 or 3 (mod 4).

] at the singular modulus valid for every

, where is a positive integer and 1, 2 or 3 (mod 4). Then

17]. Thus = 17, d = 4 = 68, = 68 and f = 1 in

17] explicitly using Theorem 1, we must determine

17), L(1, 68) =

d) of discriminant d asserts that

d), (d) is the fundamental unit > 1 of

d) when d > 0, and w(d) = 2, 4 or 6 according as d < 4, d = 4 or

S-ar putea să vă placă și