Final Homework

2019010531

2 , show that if m ≡ n ≡ k ≡ 1
mn
1. Let m, n be two distinct square-free integers, neither equals 1, let k = (m,n)
√ √ √ √
(mod 4), then {1, 4 (1 + m)(1 + k), 2 (1 + m), 2 (1 + n)} is an integral basis for OK , and d(K) = mnk.
1 1 1

√ √ √
Proof. Let K = Q( m, n), N = Q( n), l = (m, n), m = lm1 , n = ln1 . We first show the following lemma
Lemma 1. For α ∈ K, we have that α is integral over Z iff T rK/N (α), NK/N (α) are integers.

Proof. Suppose that α ∈ OK , then α is certainly integral over ON , and thus its minimal polynomial mα (t) has
coefficients in ON . Since T rK/N (α) = (−1)× (coefficient of t) and NK/N (α) is the constant term in mα (t), these
are both in ON and thus are algebraic integers. Conversely, if α is integral over OM , then α is integral over Z,
since ON is integral over Z.
√ √ √
We see that {1, m, n, k} is a basis for K over Q, thus every integer α of K is of the form
√ √ √
α = a0 + a1 m + a2 n + a3 k.

And the conjugates of α are

√ √ √
α1 = a0 − a1 m + a2 n − a3 k,
√ √ √
α2 = a0 + a1 m − a2 n − a3 k,
√ √ √
α3 = a0 − a1 m − a2 n + a3 k.
√ √ √ √ √ √
We have α + α1 = 2a0 + 2a2 n ∈ Q( n), α + α2 = 2a √0
+ 2a1 m ∈ Q( m), √
α + α3 = 2a0 + 2a
√3
k ∈ Q( k).
These are all integers, and we have OQ(√m) = Z[ 1+2 m ], OQ(√n) = Z[ 1+2 n ], OQ(√k) = Z[ 1+2 k ], thus 4a0 ≡
4a1 ≡ 4a2 ≡ 4a3 (mod 2).
(a) The case 2a0 , 2a1 , 2a2 , 2a3 are all rational integers.
√ √ √ √ √ √
We write α = 21 (a + b m + c n + d k), then the conjugate of α w.r.t. K/N is α′ = 21 (a − b m + c n − d k).
Thus
√ 1 √ √
T rK/N (α) = a + c n, NK/N (α) = (a2 + c2 n + 2ac n − b2 m − d2 k − 2bdm1 n).
4
√
By the lemma, we have T rK/N (α), NK/N (α) ∈ Z[ 1+2 n ], therefore (a2 + c2 n − b2 m − d2 k)/2 is a rational integer.
We have a + c − b − d ≡ a2 + c2 − b2 − d2 ≡ a2 + c2 n − b2 m − d2 k ≡ 0 (mod 2), thus the equation
√ √ √ √
1+ m 1+ n (1 + m)(1 + k)
x+y +z +w =α
2 2 4
has integer solution
x = (a − b − c + dm1 )/2, y = b − d, z = c − dm1 , w = 2d,
since m1 ≡ 1 (mod 2).
(b) The case 2a0 , 2a1 , 2a2 , 2a3 are all halves of odd integers.
√ √ √
We write α = 14 (a + b m + c n + d k) with a ≡ b ≡ c ≡ d ≡ 1 (mod 2), then the conjugate of α w.r.t. K/N
√ √ √
is α′ = 41 (a − b m + c n − d k). Thus
1 √ 1 2 √ √
T rK/N (α) = (a + c n), NK/N (α) = (a + c2 n + 2ac n − b2 m − d2 k − 2bdm1 n).
2 16
√
By the lemma, we have T rK/N (α), NK/N (α) ∈ Z[ 1+2 n
], therefore

a2 + c2 n − b2 m − d2 k ≡ 0 (mod 8), ac − bdm1 ≡ 0 (mod 4).

1
 If m1 ≡ 1 (mod 4), then ac − bd ≡ 0 (mod 4), therefore the possible cases are

(a, b, c, d) ≡(1, 1, 1, 1), (1, 3, 1, 3), (1, 3, 3, 1), (1, 1, 3, 3),

(3, 3, 1, 1), (3, 1, 1, 3), (3, 3, 3, 3), (3, 1, 3, 1) (mod 4).

In each case, we have a − b − c + d ≡ 0 (mod 4).

If m1 ≡ 3 (mod 4), then ac + bd ≡ 0 (mod 4), therefore the possible cases are

(a, b, c, d) ≡(1, 3, 1, 1), (1, 1, 1, 3), (1, 1, 3, 1), (1, 3, 3, 3),

(3, 1, 1, 1), (3, 3, 1, 3), (3, 1, 3, 3), (3, 3, 3, 1) (mod 4).

In each case, we have a − b − c − d ≡ 0 (mod 4).

Thus the equation √
√ √ √
1+ m 1+ n (1 + m)(1 + k)
x+y +z +w =α
2 2 4
always has integer solution

x = (a − b − c + dm1 )/4, y = (b − d)/2, z = (c − dm1 )/2, w = d.

√ √ √ √
So we have {1, 14 (1 + m)(1 + k), 12 (1 + m), 12 (1 + n)} is an integral basis for OK . And
 √ √ √ √ 
(1+ m)(1+ k) 1+ m 1+ n
1
 √ 4 √ 2
√ 2√ 
1 (1− m)(1− k) 1− m 1+ n 

d(K) = det 
 √ 4 √
(1+ m)(1− k)
2
√
1+ m
2√  2
1− n 
= mnk.
1 √ 4 √ 2
√ 2√ 
(1− m)(1+ k) 1− m 1− n
1 4 2 2

2. Let α be the real root of f = x3 − x + 1, and β, β̄ the conjugate complex roots of it, K := Q[α]. Show:
(a) β + β̄ = −α, β β̄ = −α−1 , [(β − α)(β̄ − α)(β − β̄)]2 = −23;
√ √
(b) K( −23) = Q(α, β, β̄), and K( −23)/Q is a Galois extension of degree 6, calculate the Galois group;
√
(c) Determine all subfields of K( −23)/Q. Which ones are Galois over Q?
√
(d) The only rational prime ramified in K( −23) is 23, and its ramification index is 2;
√ √ √
(e) K(√−23)/Q( −23) is a cyclic extension of degree 3, and each prime ideal in Q( −23) is unramified in
K( −23).

Proof. (a) We have x3 − x + 1 = (x − α)(x − β)(x − β̄), therefore α + β + β̄ = 0, −αβ β̄ = 1, and

Y
(ri − rj )2 = Disc(f ) = −Res(f, f ′ ) = −23,
i<j

where {r1 , r2 , r3 } = {α, β, β̄}.

√
(b) Clearly, we have K( −23) ⊆ Q(α, β, β̄), and
√ √
[K( −23) : Q] = [K( −23) : K] · [K : Q] = 2 · 3 = 6.

Since Q(α, β, β̄) = K(β, β̄) = K(β) ∼

= K[X]/(g(X)), where f (X) = (X − α)g(X) in K[X], we have

K(β) : Q] = [K(β) : K] · [K : Q] = 6.
√
Therefore K( −23) = Q(α, β, β̄), which is the splitting field of f over Q, thus Galois over Q with Galois group
the symmetric group S3 , which acts by permuting α, β, β̄.
(c) Let α, β, β̄ correspond to 1, 2, 3 respectively in the presentation S3 = {1, (1 2), (1 3), (2 3), (1 2 3), (1 3 2)}.
The order 2 subgroups of S3 are h(1 2)i, h(1 3)i, h(2 3)i, whose√fixed fields are Q(β̄), Q(β), Q(α), respectively.
The order 3 subgroup of S3 is h(1 2 3)i, whose fixed field is Q( −23). These are all the nontrivial subfields of

2
 √ √
K( −23)/Q by Galois theory, and Q( −23) is the only one Galois over Q, since h(1 2 3)i is the only normal
subgroup of S3 .
√ √ √
(d) We have L := K( −23) = K(( 1+ 2−23 )), with h = x2 − x + 6 the minimal polynomial of 1+ 2−23 . Note that
√
h has discriminant −23, it is separable after mod any prime not above 23, thus K( −23)/K is unramified at
primes not above 23.
Since K = Q(α), and disc(1, α, α2 ) = −23 which is square free, we have d(K) = −23 and {1, α, α2 } is an integral
basis for OK . Thus theP only rational prime ramified in K is p = 23, therefore 23 is the only rational prime
ramified in L. We have p ep fp = 3, with p runs through primes over 23. These primes one-one correspond to
the irreducible factors of f = (x + 10)2 (x + 3) (mod 23), therefore 23OK = p21 p2 .
Since L/Q is Galois, we see that p1 is inert in L and p2 is totally ramified in L. That is p = 23 has ramification
index 2.
√ √ √
(e) K( −23)/Q( √ −23) is Galois and of degree 3, thus cyclic. Since d(Q( −23)) = −23,√ the only rational prime
ramified
√ in Q( −23) is 23, with ramification index 2, therefore each prime ideal in Q( −23) is unramified in
K( −23).
√
3. (1) Show that the class number of K = Q( −103) is 5.
(2) Show that the only rational integer solutions of x2 + 2 = y 3 are (x, y) = (±5, 3).
√
−103
√
Proof. (1) Since −103 ≡ 1 (mod 4), we have OK = Z[ 1+ 2 ], d(K) = −103 and M (K) = 2
π 103 ≈ 6.46. The
rational primes less than M (K) are 2, 3, 5. We have

2 ∤ d(K), −103 ≡ 1 (mod 8) ⇒ 2OK = p2 p¯2 , N (p2 ) = N (p¯2 ) = 2;

−103
2

3 ∤ d(K), = = −1 ⇒ 3OK is prime, N (p3 ) = 9;
3 3
−103
2

5 ∤ d(K), = = −1, ⇒ 5OK is prime, N (p5 ) = 25.
5 5
And the class group is generated by [p2 ].
√
Let α = 2 + 1+ 2−103 ∈ OK , we have N ((α)) = 25 , thus (α) = pi2 p¯2 j , with i + j = 5. Since 2 ∤ (α) in OK , we
have i = 0 or j = 0. Therefore [p2 ]5 = 1, and the assertion follows.
√ √
(2) We consider in K := Q( −2), which has class √ number √ h = 1 (therefore OK = Z[ −2] is principal). Suppose
that x, y ∈ Z satisfies x2 + 2 = y 3 , then (x + −2)(x − −2) = (y)3 in OK .
√ √
We first show that (x + −2) and√(x − −2) are relatively prime.√Suppose not,√let 0 6= p be a prime ideal √ of 2OK
dividing both of them,
√ then 2x, 2 −2 ∈ p, therefore p|(2)
√ or p|(
√ −2). Since ( −2) is prime, (2) = ( −2) , it
must be that p = ( −2). Therefore x ∈ p, thus x = −2(a + b −2) with a, b ∈ Z. Since x ∈ Z, we have a = 0
and therefore x ≡ 0 (mod 2). Then y ≡ 0 (mod 2) by the equation, but then y 3 ≡ 0 (mod 4) while x2 + 2 ≡ 2
(mod 4), a contradiction.
√ √ √
Since (x + −2) and (x − −2) are√relatively prime, we have (x + −2) = a3 for some ideal a of OK , by prime
decomposition. We have a = (c + d −2) for some c, d ∈ Z since OK is principal. So we have
√ √
x + −2 = c3 − 6cd2 + (3c2 d − 2d3 ) −2,
⇒ 3c2 d − 2d3 = 1, c3 − 6cd2 = x

If d = 1, then c = ±1, therefore x = ±5. If d = −1, the 3c2 = 1, which is impossible. Then the assertion
follows.

4. Let f (x) ∈ Z[x] be monic irreducible of degree ≥ 2, show that: there are infinitely many rational primes p such
that f (x) has no root in Fp .

K/Q

Proof. Let K be the splitting field of f over Q, with Galois group G. Let σp = p be the Artin symbol.
Let p be a rational prime unramified in K and p ∤ Disc(f ), that is, f mod p is separable. For such a p, f (x) ≡ 0
(mod p) has a solution implies that every element of sigmap fix some root of f since the decomposition type of
(f mod p) equals the cycle pattern of the Frobenius element σp ∈ G (view G as a subgroup of the symmetric
group on roots of f ).

3
 Let H = Gal(K/Q(α)) where α is a root of f . By considering the action of G on the set of left cosets of H in
G, let C = ∪g∈G gHg −1 (which is stable under conjugation), we see that

|C| ≤ [G : H](|H| − 1) + 1 = |G| − [G : H] + 1 < |G|.

By Chebotarev density theorem, we have that the set of primes

A = {p | p is unramified in K, p ∤ Disc(f ) and σp lies in C}

has Dirichlet density < 1. Since the number of primes ramified in K or deviding Disc(f ) is finite, the assertion
follows.

5. (1) Let p be an odd prime, then −1 is a square in Qp iff p ≡ 1 (mod 4).

(2) Calculate the square root of −1 in Q5 (accurate to the fourth digit of the 5-adic expansion).

Proof. (1) Let f (x) = x2 + 1, we have the fact that for an odd prime p, f (x) ≡ 0 (mod p) has a solution iff p ≡ 1
(mod 4).
If p ≡ 1 (mod 4), let 0 ≤ a0 ≤ p − 1 be such that f (a0 ) ≡ 0 (mod p). We construct a Cauchy sequence {ai }i≥0
in Zp whose limit a is the required root. Suppose that we are given an ∈ Zp with an ≡ a0 (mod p) and

f (an ) ≡ 0 (mod pn+1 ).

Let an+1 = an + tpn+1 , t ∈ Zp . We want

f (an+1 ) = f (an ) + 2an · tpn+1 + · · · ≡ 0 (mod pn+2 ).

It suffices to take
−f (an )
t≡ (mod p).
2an pn+1
P∞
Conversely, if −1 is a square in Qp , that is, there is some r=0 ar pr ∈ Zp such that
∞
X
( ar pr )2 + 1 = 0.
r=0

So we have a20 + 1 ≡ 0 (mod p), therefore p ≡ 1 (mod 4).

P∞
(2) We want to find r=0 ar 5r ∈ Z5 such that
∞
X
( ar 5r ) + 1 = 0.
r=0

We first take a0 = 2, then by the discussion in (1), and

−f (a0 )
a1 ≡ ≡ 1 (mod 5),
2a0 · 5
we have a1 = 2 + 1 · 5. Then
−f (a1 )
≡ 2 (mod 5),
2a1 · 52
so a2 = 2 + 1 · 5 + 2 · 52 . Then
−f (a2 )
≡ 1 (mod 5),
2a2 · 53
so a3 = 2 + 1 · 5 + 2 · 52 + 1 · 53 . Then
−f (a3 )
≡ 3 (mod 5),
2a3 · 54
so a4 = 2 + 1 · 5 + 2 · 52 + 1 · 53 + 3 · 54 .

6. Let E be a finite extension of Qp , vp be the p-adic valuation, let v be the unique extension to E of vp . Show
that:

4
 (a) For x ∈ E, if v(x) > 1
p−1 , then the series

x2 xn
exp(x) = 1 + x + + ··· + + ···
2! n!
n
converges, and if n ≥ 2, then v( xn! ) > v(x). Thus

v(x) = v(exp(x) − 1).

(b) If x, y ∈ E, v(x) > 1

p−1 , v(y) > 1
p−1 , then

exp(x + y) = exp(x) exp(y).

(c) If v(x) > 0, x ∈ E, then the siries

x2 x3 xn
log(1 + x) = x − + − · · · + (−1)n+1 + ···
2 3 n
n
converges. If v(x) > 1
p−1 , then when n ≥ 2, we have v( xn ) > v(x). Thus

v(log(1 + x)) = v(x).

(d) If v(x − 1) > 0, v(y − 1) > 0, x, y ∈ E, then

log(xy) = log(x) + log(y).

(e) If v(x) > 1

p−1 , x ∈ E, then
exp(log(1 + x)) = 1 + x, log(exp(x)) = x.
In particular, when r ≥ e
[ p−1 ] + 1, the maps

log : Ur → prE , exp : prE → Ur = 1 + prE

form an inverse pair of group isomorphisms between the multiplicative group Ur and the additive group
prE , where pE is the prime ideal of E, e = e(pE /p).
(f) The group homomorphism f : U1 = 1 + pE → E, x 7→ log(x) has kernel
n
ker(f ) = {α ∈ E | ∃ n ≥ 0, s.t. αp = 1}.

Proof. (a) Write n = a0 + a1 p + · · · + ar pr , 0 ≤ ai ≤ p − 1. Then

n n n
v(n!) = b c + b 2 c + · · · + b r c,
p p p
and
n
b c = a1 + a2 p + a3 p2 · · · + ar pr−1
p
n
b 2 c = a2 + a3 p + · · · + ar pr−2
p
···
n
b c = ar
p
Therefore P
p0 − 1 p1 − 1 p2 − 1 pr − 1 n − ai
v(n!) = a0 + a1 + a2 + · · · + ar = .
p−1 p−1 p−1 p−1 p−1
Thus P
xn 1 ai
v( ) = n(v(x) − )+ .
n! p−1 p−1
P
So we have v(xn /n!) → +∞ as n → +∞, if v(x) > 1
p−1 , and therefore exp(x) converges. If n ≥ 2, then ai ≥ 1,
∑
n− ai
therefore (n − 1)v(x) > n−1
p−1 ≥ p−1 , then v(x /n!) > v(x). Thus v(exp(x) − 1) = v(x).
n

5
 1 1
(b) If v(x) > p−1 , v(y) > p−1 , then exp(x), exp(y), exp(x + y) all converge. Then we have exp(x + y) =
exp(x) exp(y) since this is an identity of formal series:
X (x + y)n XX n  
n r n−r
= x y /n!
n! r=0
r
n≥0 n≥0
XX
n
xr y n−r
= ·
r! (n − r)!
n≥0 r=0
X n x X ym
=( )( ).
n! m!
n≥0 m≥0

(c) Write n = a0 + a1 p + · · · + ar pr , 0 ≤ ai ≤ p − 1, with ar > 0, then pr ≤ n, so v(n) ≤ r ≤ ln n

ln p . Therefore

ln n
v(xn /n) = nv(x) − v(n) ≥ nv(x) − ,
ln p
which tends to +∞ if v(x) > 0, so log(1 + x) converges in this case. For n ≥ 2, write n = pa n0 , with (a0 , p) = 1.
If a > 0, then
v(n) a a 1 a 1
= a ≤ a = · ≤ .
n−1 p n0 − 1 p −1 p − 1 1 + p + · · · + pa−1 p−1
If a = 0, then
v(n) 1
=0< .
n−1 p−1
So we have
xn
v( ) = nv(x) − v(n) > v(x)
n
1 1
if v(x) > p−1 . Thus v(log(1 + x)) = v(x) if v(x) > p−1 .
(d) If v(x − 1) > 0, v(y − 1) > 0, x, y ∈ E, then
xy − 1 = (x − 1)(y − 1) + (x − 1) + (y − 1) ⇒ v(xy − 1) > 0.
Therefore log(x), log(y), log(xy) all converge. Then we have the equality log(xy) = log(x) + log(y) since this is
an identity of formal series: Let D be the formal derivative operator, we have
D(t)
D log(1 + t) = ,
1+t
therefore

D(t) + D(s) + tD(s) + sD(t) D(t) D(s)

D log (1 + t)(1 + s) = = + = D log(1 + t) + log(1 + s) .
(1 + t)(1 + s) 1+t 1+s


So log (1 + t)(1 + s) = log(1 + t) + log(1 + s), for they have the same constant term.
1 1
(e) If v(x) > p−1 , then v(log(1+x)) = v(exp(x)−1) = v(x) > p−1 , so the series log(1+x), exp(x), exp(log(1+x)),
log(exp(x)) all converge. Then we have the equality since these are identities of formal power series:
e
For r > p−1 , exp maps prE into Ur since v(exp(x)−1) = v(x), and log maps Ur into prE since v(log(1+x)) = v(x).
Then the assertion for the particular case follows by the equality just shown.
(f) Let 1 + x be in the kernel of log : U1 → E. For n sufficiently large, we have |x|p < p−1/(p−1) . Then
n

p  n
X
n

pn p
(1 + x) −1= xj
j=1
j

pn


has absolute value < p−1/(p−1) (since for 0 < j < pn we have | j xj | ≤ |px| < |p| ≤ p−1/(p−1) ), and therefore
n n
|(1 + x)p − 1| = | log((1 + x)p )| = |pn log(1 + x)| = 0.
Thus, (1 + x) is a pn -th root of unity.
Conversely, every p-power root of unity in E is a torsion element of U1 . But E has characteristic zero, so (E, +)
is torsion-free. Therefore all p-power roots of unity are in the kernel of the homomorphism log : U1 → E.

6
7. For a number field K, define the Dedekind zeta function ζK (s), show that it admits a meromorphic extension
to some strip domain in Re(s) < 1, and give the details of calculating the residue at s = 1.
Definition 2. Let K be a number field, the Dedekind zeta function of K is defined by the series
X
ζK (s) = N (a)−s ,
a

where a ranges over the integral ideals of K, and N (a) denotes its absolute norm.
Q
We have the product form ζK (s) = p (1 − N (p)−s )−1 , where p ranges over nonzero prime ideals. The product
converges absolutely and uniformly in the domain Re(s) ≥ 1 + δ for any δ > 0: since #{p|p} ≤ n and N (p) =
[OK : p] ≥ p, thus X X
log(1 − N (p)−s ) ≤ n log(1 − p−s ) ,
p p

and we know that ζ(s) converges absolutely and uniformly in Re(s) ≥ 1 + δ.

We want to prove the following theorem
Theorem 3. (Analytic Class Number Formula) Let K be a number field of degree n. The Dedekind zeta function
ζK (s) extends to a meromorphic function on Re(s) > 1 − n1 that is holomorphic except for a simple pole at s = 1
with residue
2r1 (2π)r2 hK RK
lim (s − 1)ζK (s) = ,
s→1+ wK |DK |1/2
where r1 and r2 are the number of real and complex places of K, respectively, hK := #cl(OK ) is the class
number, RK is the regulator, wk := #µK is the number of roots of unity, and DK := disc(OK ) is the absolute
discriminant.

In order to prove the analytic class number formula we need an asymptotic estimate for the number of integral
ideals a with absolutePnorm N (a) bounded by a parameter t ∈ R>0 , this is necessary for us to understand the
behavior of ζK (s) = a N (a)−s as s → Q 1+ . Our strategy is to count points in Log(OK ∩ K × ) that lie inside
a suitably chosen region S of R r1 +r2
' [ τ R]+ that we will then scale by t. In order to bound this count as
a function of t we need a condition on S that ensures that the count grows smoothly with t; this requires S to
have a “reasonable” shape. We first introduce the following results in real analysis that we will use.
Definition 4. (a) Let X and Y be two metric spaces, a function f : X → Y is Lipschitz continuous if there
exists c ∈ R>0 such that for all x1 , x2 ∈ X we have

d(f (x1 ), f (x2 )) < c · d(x1 , x2 ).

(b) A set B in a metric space X is d-Lipschitz parametrizable if it is the union of the images of a finite number
of Lipschitz continuous functions fi : [0, 1]d → X.
Lemma 5. Let S ⊆ Rn be a measurable set whose boundary ∂S := S̄ \ S 0 is (n − 1)-Lipschitz parametrizable.
Then
#(tS ∩ Zn ) = µ(S)tn + O(tn−1 ), as t → +∞,
where µ is the standard Lebesgue measure on Rn .

With this lemma, we have the corollary

Corollary 6. Let Λ be a lattice in an R-vector space V ' Rn and let S ⊆ V be a measurable set whose boundary
is (n − 1)-Lipschitz parametrizable, then

µ(S) n
#(tS ∩ Λ) = t + O(tn−1 ).
covol(Λ)

Recall that covol(Λ) = µ(F ) for any fundamental region F for Λ, so the ratio µ(S)/covol(Λ) = µ(S)/µ(F )
in Corollary 6 does not depend on the normalization of the Haar measure µ. However, we plan to apply the
corollary to Λ = OK and want to replace covol(OK ) with |DK |1/2 , which requires us to use the normalized Haar
measure on KR :
µ(S) := 2r2 µRn (S).

7
We want to estimate the quantity #{a : N (a) ≤ t}. As a first step, we restrict our attention to principal ideals
(α) ⊆ OK , so we want to estimate the quantity #{(α) : N (α) ≤ t}, which is equivalent to the cardinality of
×
{α ∈ K × ∩ OK : N (α) ≤ t}/OK .

We define
×
KR,≤t := {x ∈ KR× : N (x) ≤ t} ⊆ KR× ⊆ KR ,
× × ×
then we want to estimate the cardinality of (KR,≤t ∩ OK )/OK , here we view OK , OK as subsets of KR through
× ×
the natural map j : K → KR . By Dirichlet unit theorem, we have OK ' µK × U , where U ⊆ OK is free of rank
r1 + r2 − 1. Thus we have the wK -to-1 map
× × ×
(KR,≤t ∩ OK )/U → (KR,≤t ∩ OK )/OK ,
×
and we will estimate the cardinality of (KR,≤t ∩ OK )/U .
We now define a surjective map
ν : KR× → KR,1
×
, x 7→ xN (x)−1/n .
× Q +
The image of KR,1 under Log is precisely the trace zero hyperplane Rr01 +r2 of Rr1 +r2 ' [ τ R] , in which
×
Log(U ) = Log(OK ) = ΛK is a lattice. Let us fix a fundamental domain F for the lattice ΛK in R01 +r2 , then
r

S := ν −1 (Log−1 (F ))

is a set of unique coset representatives for the quotient KR× /U . We define

S≤t := {x ∈ S : N (x) ≤ t} ⊆ KR ,

then we want to estimate the cardinality of S≤t ∩ OK . The set OK is a lattice in the R-vector space KR of
dimension n. We have tS≤1 = S≤tn , so we can estimate the cardinality of S≤t = t1/n S≤1 via Corollary 6
with S = S≤1 and Λ = OK by replacing t with t1/n , provided that the boundary of S≤1 is (n − 1)-Lipschitz
parametrizable, which we now argue.
The kernel of the Log map is {±1}r1 × U (1)r2 , where U (1) = {z ∈ C : z z̄ = 1} is the unit circle in C. We thus
have a continuous isomorphism of locally compact groups
∼
KR× = (R)r1 × (C)r2 −→ Rr1 +r2 × {±1}r1 × [0, 2π)r2 (1)
x = (x1 , · · · , xr1 , z1 , · · · , zr2 ) 7−→ (Log(x)) × (sign x1 , · · · , sign xr1 ) × (arg z1 , · · · , arg zr2 ).

The set S≤1 consists of 2r1 connected components, one for each element of {±1}r1 . We can parametrize each of
these component using n real parameters as follows:
• r1 + r2 − 1 parameters in [0, 1) that encode a point in F as an R-linear combination of Log(ε1 ), · · · ,
Log(εr1 +r2 −1 ), where {ε1 , · · · , εr1 +r2 −1 } is a basis for U ;
• r2 parameters in [0, 1) that encode an element of U (1)r2 ;
• a parameter in (0, 1] that encodes the nth-root of the norm.
These parameterizations define a continuously differentiable bijection from the set

C = [0, 1)n−1 × (0, 1] ⊆ [0, 1]n

to each of the 2r1 disjoint components of S≤1 ; it can be written out explicitly in terms of exponentials and
the identity function. The boundary ∂C is the boundary of the unit n-cube, which is clearly (n − 1)-Lipschitz
parametrizable; thus each component of S≤1 , and therefore S≤1 itself, has a boundary that is (n − 1)-Lipschitz
parametrizable.
We now apply Corollary 6 to the lattice OK and the set S≤1 in the n-dimensional R-vector space KR with t
replaced by t1/n , since S≤t = t1/n S≤1 . This yields

µ(S≤1 ) µ(S≤1 )
#(S≤t ∩ OK ) = (t1/n )n + O((t1/n )n−1 ) = t + O(t1−1/n ) (2)
covol(OK ) |DK |1/2

Our next task is compute µ(S≤1 ); as noted in the statement after corollary 6, we must use the normalized Haar
measure µ on KR . We will use the isomorphism in (1) to make a change of coordinates, we just need to understand

8
 Q
how this affects the Haar measure µ on KR = v|∞ Kv ' RP r1 ×Cr2 . In terms of the standard Lebesgue measures
dx and dA on R and C, we have µ = (dx)r1 (2dA)r2 . For each factor of KR× ' (R× )r1 × (C× )r2 ⊆ Rr1 × Cr2 , we
define the maps

R× → R × {±1}, x 7→ (log |x|, sgn x), ±el ←[ (l, ±1), dx 7→ el (dl)µ{±1}

C× → R × [0, 2π), z 7→ (2 log |z|, arg z), el/2+iθ ←[ (l, θ), 2dA 7→ 2el/2 d(el/2 )dθ = el dldθ,
where dl is the Lebesgue measure on R, µ{±} is the counting measure on {±1}, and dθ is the Lebesgue measure
on [0, 2π). We thus have
∼
KR× −→ Rr1 +r2 × {±1}r1 × [0, 2π)r2 , µ 7→ eT(·) µRr1 +r2 µr{±1}
1
µr[0,2π)
2
,

where the trace function T(·) sums the coordinates of a vector in Rr1 +r2 .
We now make one further change of coordinates:

Rr1 +r2 −→ Rr1 +r2 −1 × R

x = (x1 , · · · , xr1 +r2 ) 7−→ (x1 , · · · , xr1 +r2 −1 , y := T (x))
eT(x) µRr1 +r2 7−→ ey µRr1 +r2 −1 dy

If we let π0 : Rr1 +r2 → Rr1 +r2 −1 denote the coordinate projection, then the Lebesgue measure of π0 (F ) in
Rr1 +r2 −1 is, by definition, the regulator RK .
The Log map gives us a bijection
∼ 1 1 2 2

S≤1 −→ F + (−∞, 0]
,··· , , ,··· ,
n n n n
1 1 2 2

x = N (x) ν(x) 7−→ Log ν(x) + log N (x) , · · · , , , · · · , .
1/n
n n n n
The coordinate y ∈ (−∞, 0] is given by y = T(Log x) = log N (x), so we can view S≤1 as an infinite union of
cosets of Log−1 (F ) parameterized by ey = N (x) ∈ (0, 1]. Under our change of coordinates we thus have
∼
KR× −→ Rr1 +r2 −1 × R × {±1}r1 × [0, 2π)r2
S≤1 −→ π0 (F ) × (−∞, 0] × {±1}r1 × [0, 2π)r2 .

Since RK = µRr1 +r2 −1 (π0 (F )), we have

Z 0
µ(S≤1 ) = ey RK 2r1 (2π)r2 dy = 2r1 (2π)r2 RK .
−∞

So we have  2r1 (2π)r2 R 

K
#(S≤t ∩ OK ) = t + O(t1−1/n ). (3)
|DK |1/2
We are now ready to prove the analytic class number formula. Our main tool is the following theorem
Theorem 7. Let K be a number field of degree n as t → ∞, the number of nonzero integral ideals a of absolute
norm N (a) ≤ t is
 2r1 (2π)r2 h R 
K K
t + O(t1−1/n ), as t → +∞,
wK |DK |1/2
where r1 and r2 are the real and complex places of K, respectively, hK := #cl(OK ) is the class number, RK is
the regulator, wk := #µK is the number of roots of unity, and DK := disc(OK ) is the absolute discriminant.

Proof. In order to count the nonzero OK -ideals a of absolute norm N (a) ≤ t we group them by ideal class. For
the trivial class, we just need to count nonzero principal ideals (α), equivalently, the number of nonzero α ∈ OK
×
with N (α) ≤ t, modulo the unit group OK . Dividing (3) by wK to account for the wK -to-1 map
× ×
S≤t ∩ OK −→ (KR,≤t ∩ OK )/OK ,

we obtain  2r1 (2π)r2 R 

K
#{(α) ⊆ OK : N (α) ≤ t} = t + O(t1−1/n ) (4)
wK |DK |1/2

9
To complete the proof we now show that we get the same answer for every ideal class; the nonzero ideals a
of norm N (a) ≤ t are asymptotically equidistributed among ideal classes. Fix an ideal class [a], with a ⊆ OK
nonzero. Multiplication by a gives a bijection
×a ×
{b ∈ [a−1 ] : N (b) ≤ t} −→ {0 6= (α) ⊆ a : N (α) ≤ tN (a)} ' {0 6= α ∈ a : N (α) ≤ tN (a)}/OK .

Let S[a],≤t denote the set on the RHS. The estimate in (4) derived from Corollary 6 applies to any lattice in KR ,
not just OK . Replacing OK with a in (4) we obtain
 2r1 (2π)r2 R 
K
#S[a],≤t = tN (a) + O(t1−1/n )
wK · covol(a)
 2r1 (2π)r2 RK 
= tN (a) + O(t1−1/n )
wK · covol(OK )N (a)
 2r1 (2π)r2 R 
K
= t + O(t1−1/n ).
wK |DK |1/2

Note that the RHS does not depend on the ideal class [a]. Summing over ideal classes yields
X  2r1 (2π)r2 h R 
K K
#{0 6= b ⊆ OK : N (b) ≤ t} = #S[a],≤t = t + O(t1−1/n ).
wK |DK |1/2
[a]

We need the following two lemmas

Lemma 8. Let a1 , a2 , · · · be a sequence of complex numbers and let σ be a real number. Suppose that

a1 + · · · + at = O(tσ ), as t → ∞.
P
Then the Dirichlet series an n−s defines a holomorphic function on Re(s) > σ.
P
Proof. Let A(x) := 0<n≤x an . Writing the Dirichlet sum as a Stieltjes integral, for Re(s) > σ, we have
X Z ∞
an n−s = x−s dA(x)
1−
Z ∞
A(x) ∞
= − A(x) dx−s
x s 1− 1−
Z ∞
=0− A(x)(−sx−s−1 ) dx
1−
Z ∞
A(x)
=s dx
1 − xs+1

The integral on the RHS converges locally uniformly on Re(s) > σ and the lemma follows.
Lemma 9. Let a1 , a2 , · · · be a sequence of complex numbers that satisfies

a1 + · · · + at = ρt + O(tσ ), as t → ∞,
P
for some σ ∈ [0, 1), ρ ∈ C× . Then the Dirichlet series an n−s converges on Re(s) > 1 and has a meromorphic
continuation to Re(s) > σ that is holomorphic except for a simple pole at s = 1 with residue ρ.

Proof. Let bt := at − ρ, then b1 + · · · + bt = O(tσ ) and

X X
an n−s = ρζ(s) + bn n−s .
P
We know that bn n−s defines a holomorphic function on Re(s) > σ by lemma 7, and ζ(s) has a meromorphic
continuation to Re(s) > 1 with a simple residue at s = 1 with residue 1, then the assertion follows.

Now we can prove theorem 3.

10
 Proof. We have X X
ζK (s) = N (a)−s = at t−s ,
a t≥1

where a ranges over nonzero ideals of OK , at := #{a : N (a) = t}. Let

2r1 (2π)r2 hK RK
ρK := ,
wK |DK |1/2

then by theorem 7, we have

a1 + · · · + at = #{a : N (a) ≤ t} = ρK t + O(t1−1/n ), as t → ∞.

Thus theorem 3 follows by lemma 9.

8. Let L/K be a Galois extension of number fields, denote G := Gal(L/K). Give the definitions of saying a prime


ideal of K ramified or unramified in L. For an unramified prime ideal p and a prime P over it, define L/K
P .
State and prove the Chebotarev density theorem.

11