p-adic Analysis Compared to Real

Lecture 1
Felix Hensel, Waltraud Lederle, Simone Montemezzani
October 12, 2011

1 Normed Fields & non-Archimedean Norms

Definition 1.1. A metric on a non-empty set X is a function

d : X × X −→ R≥0

satisfying the following properties:

(1) d(x, y) = 0 ⇐⇒ x = y

(2) d(x, y) = d(y, x) ∀x, y ∈ X

(3) d(x, y) ≤ d(x, z) + d(z, y) ∀x, y, z ∈ X

(X, d) is called a metric space.

Definition 1.2. A sequence (xn ) in a metric space (X, d) is called Cauchy sequence
if ∀ε > 0 ∃N ∈ N such that d(xn , xm ) < ε ∀m, n > N .
(X, d) is a complete metric space if any Cauchy sequence in X has a limit in X.

Definition 1.3. Let F be a field. A norm on F is a map k · k : F → R≥0 that

satisfies:

(1) kxk = 0 ⇐⇒ x = 0

(2) kxyk = kxkkyk ∀x, y ∈ F

(3) kx + yk ≤ kxk + kyk ∀x, y ∈ F (triangle inequality)

(F, k · k) is called a normed field.

A norm is trivial if k0k = 0 and kxk = 1 ∀x ∈ F \{0}.

1
Remark 1.1. A norm k · k on a field F induces a metric on F by:

F × F −→ R≥0 : (x, y) 7−→ kx − yk

This allows us to regard a normed field (F, k · k) as a metric space.

Proposition 1.1. For any x, y ∈ F we have:
(a) k1k = k − 1k = 1

(b) kxk = k − xk

(c) kx ± yk ≥ | kxk − kyk |

(d) kx − yk ≤ kxk + kyk

(e) kx/yk = kxk/kyk

(f) knk ≤ n ∀n ∈ N (on the left hand side: n := n · 1F ∈ F )

Proof. (a) k1k = k(±1) · (±1)k = k ± 1k2 =⇒ k ± 1k = 1

(b) k − xk = k − 1kkxk = kxk

(c) kxk = kx ± y ∓ yk ≤ kx ± yk + kyk =⇒ kxk − kyk ≤ kx ± yk

kyk = ky ± x ∓ xk ≤ ky ± xk + kxk =⇒ kyk − kxk ≤ ky ± xk
Thus kx ± yk ≥ | kxk − kyk |.

(d) Follows from (b) and the triangle inequality.

(e) kykkx/yk = kxk

(f) Follows by induction from (a) and the triangle inequality.

Definition 1.4. A norm is called non-Archimedean if it satisfies the strong triangle

inequality:
kx + yk ≤ max{kxk, kyk}.
If a norm does not satisfy the strong triangle inequality it is said to be Archimedean.
Remark 1.2. The strong triangle inequality clearly implies the triangle inequality.
We call a metric that is induced by a non-Archimedean norm an ultra-metric
and the corresponding metric space an ultra-metric space.
An ultra-metric satisfies the strong triangle inequality:

d(x, y) ≤ max{d(x, z), d(z, y)}.

2
Proposition 1.2. k · k is non-Archimedean ⇐⇒ knk ≤ 1 ∀n ∈ Z

Proof. ” =⇒ ”: Induction: k1k = 1 ≤ 1. Suppose that kkk ≤ 1 ∀k ∈ {1, . . . , n−1}.

knk = kn − 1 + 1k ≤ max{kn − 1k, k1k} = 1. Hence knk ≤ 1, ∀n ∈ N. Since we
have that knk = k − nk, the result follows.
”⇐=”:
n  
n n
X n k n−k
kx + yk = k(x + y) k = x y
k=0
k
n  
X n
≤ kxkk kykn−k
k=0
k
n
X
≤ kxkk kykn−k
k=0
≤ (n + 1)[max{kxk, kyk}]n

Hence kx + yk ≤ (n + 1)1/n max{kxk, kyk}.

Letting n tend to ∞ we get:

kx + yk ≤ max{kxk, kyk}.

Proposition 1.3. The following are equivalent:

(i) k · k is Archimedean

(ii) k · k has the Archimedean property:

given x, y ∈ F, x 6= 0 ∃n ∈ N such that knxk > kyk

(iii) sup{knk : n ∈ Z} = +∞

Proof. ”(i) =⇒ (iii)”: By Proposition 1.2 ∃n ∈ Z such that knk > 1

k→∞
=⇒ knk k = knkk −→ ∞.
”(iii) =⇒ (ii)”: Given x, y ∈ F, x 6= 0 we can choose n ∈ N such that:

kyk
knk > =⇒ knxk > kyk.
kxk

”(ii) =⇒ (i)”: Take x, y ∈ F, x 6= 0 such that kxk ≤ kyk. By (ii) ∃n ∈ N such

kyk
that knk > kxk ≥ 1, hence the result follows from Proposition 1.2.

3
Proposition 1.4. If k · k is non-Archimedean we have:

kx − ak < kak =⇒ kak = kxk

Proof.
kxk = kx − a + ak ≤ max{kx − ak, kak} = kak
and
kak = ka − x + xk ≤ max{ka − xk, kxk}.
Suppose that ka−xk > kxk, then kx−ak ≥ kak which contradicts the assumption.
Therefore we get that kak ≤ kxk and hence kak = kxk.
Proposition 1.5. Any triangle in an ultra-metric space (X, d) is isosceles and the
length of its base does not exceed the length of the other two sides.
Proof. Let x, y, z ∈ F . W.l.o.g. assume that d(x, y) < d(x, z). Then:

d(y, z) ≤ max{d(x, y), d(x, z)} = d(x, z)

d(x, z) ≤ max{d(x, y), d(y, z)} = d(y, z)

Therefore: d(x, z) = d(y, z).
Proposition 1.6. Let k · k be non-Archimedean. Any point of an open ball
B(a, r) := {x ∈ F : kx − ak < r} is its center. The same is true for closed
balls.
Proof. Fix any b ∈ B(a, r).
Choose x ∈ B(a, r), then

kx − bk = kx − a + a − bk ≤ max{kx − ak, ka − bk} < r.

Therefore we get that B(a, r) ⊂ B(b, r). Similarly, we get that B(b, r) ⊂ B(a, r)
and thus B(a, r) = B(b, r). This argument can easily be adapted to the case of
closed balls.

2 The Completion of a Normed Field

It is well-known how the real numbers can be constructed from the rationals as
equivalence classes of Cauchy sequences. In this section we will generalize this
construction to arbitrary normed fields.
Let (F, k · k) be a normed field. Let CF denote the set of all Cauchy sequences
in F . Componentwise addition and multiplication turns CF into a commutative
ring which obviously contains lots of zero divisors and hence is in no way a field.

4
However, we will get a field from CF by identifying sequences which should have
the same limit.
First we embed F into CF via the map

F → CF : a 7→ â := (a, a, a, ...).

Note that 0̂ and 1̂ are the neutral elements of addition and multiplication in CF .
Let
N := {(an ) ∈ F N | lim kan k = 0}
n→∞

be the set of all null sequences in F . Note that N ⊂ CF since every converging
sequence is a Cauchy sequence.

Proposition 2.1. This N is a maximal ideal in CF .

Proof. Claim 1: (N ,+) is a group.

Let (an ), (bn ) be null sequences. Let ε > 0 and N, M ∈ N such that ∀n > N :
kan k < ε/2 and ∀m > M : kbm k < ε/2. Then, by the triangle inequality, for all
k > max{N, M } : kak + bk k ≤ kak k + kbk k < ε. That the additive inverse of a null
sequence is again a null sequence is the direct consequence of Proposition 1.1 and
that 0̂ is a null sequence is obvious.
Claim 2: Let (an ) be a null sequence and (xn ) be any Cauchy sequence. Then,
(an xn ) is a null sequence.
First we prove that every Cauchy sequence is bounded. Let C := sup{kxn k|n ∈
N}. Since for big enough n we know that kxn k can only lie in some interval
[kxm k − ε, kxm k + ε] for an appropriate m, we get C < ∞.
Now we prove that the product of a bounded sequence with a nullsequence is
a nullsequence. Let ε > 0. Let now N ∈ N be such that ∀n > N : kan k < ε/C.
Then, for n > N : kan xn k < ε/C · C = ε.
Hence, N is an ideal.
Claim 3: N is maximal.
Let (an ) be a Cauchy sequence which is not a null sequence. We want to prove
that 1̂ ∈ (N, (an )). Since (an ) is not a null sequence, there exists a c > 0 and an
N > 0 such that for all n > N : kan k > c. Since for any ε > 0 and big enough
n, m we have
1 1 am − an ε
− = ≤ 2.
an am am an c
Define now (
1 if an = 0
bn :=
1/an else.

5
By the above, (bn ) is a Cauchy sequence. Define a null sequence (xn ) as follows
(
1 if an = 0
xn :=
0 else.

This is a null sequence since for big enough n, all xn are zero. Now,

1 = an b n + x n

for all n ∈ N and hence 1̂ ∈ (N, (an )).

Define
F̂ := CF/N.
Since N is maximal, F̂ is a field. We now extend the norm of F to a norm on F̂ .

Definition 2.1. Let a ∈ F̂ . Then, the norm of a is defined by

kak := lim kan k,

n→∞

where a = (an ) + N .

Proposition 2.2. This is a well-defined norm on F̂ .

Proof. If two Cauchy sequences (an ), (bn ) differ only by a null-sequence, then by
Proposition 1.1 we immediately get that

lim kan k = lim kbn k

n→∞ n→∞

and hence k · k is well-defined.

From the definition it directly follows that the elements with norm zero are
exactly the null sequences. Multiplicativity and the triangle inequality also follow
immediately.

Definition 2.2. The normed field (F̂ , k · k) is called the completion of F with
respect to the norm k · k.

This terminology is justified by the following theorem.

Theorem 2.3. The normed field (F̂ , k · k) is complete and F is dense in F̂ .

Proof. First we prove the second statement. Let (an ) be a Cauchy sequence in F .
Then, (ân ) is a sequence of (constant) Cauchy sequences and we have

lim kan − am k = 0.
n,m→∞

6
Hence, (ân ) converges to the class represented by (an ). Hence, F is dense in F̂ .
Now let (An ) be a Cauchy sequence of Cauchy sequences in F , hence a repre-
sentative of a Cauchy sequence in F̂ . Since F is dense and An is a Cauchy sequence
for every n, we know that there exists a Cauchy sequence (ân ) for all n such that
1
An − (ân ) < .
n
It follows that
(an ) − (An ) = ((an ) − (ân )) − (An − (ân ))
is a null sequence in F̂ . Therefore,
lim k(an ) − An k = 0
n→∞

.
The only thing left to check is that the field operations of F̂ come from F in a
continuous way.
Proposition 2.4. Let (an ) and (bn ) be Cauchy sequences in F ⊂ F̂ . Then,
lim (an + bn ) = ( lim an ) + ( lim bn )
n→∞ n→∞ n→∞
lim (an · bn ) = ( lim an ) · ( lim bn )
n→∞ n→∞ n→∞

Proof. From the proof of the above theorem follows

lim (an + bn ) = (an + bn ) = (an ) + (bn ) = ( lim an ) + ( lim bn )
n→∞ n→∞ n→∞

where we denote the class represented by (an ) also by (an ). Analog for multipli-
cation.

3 The field of p-adic numbers Qp

Definition 3.1. Let p ∈ N prime. For 0 6= x ∈ Q we define the p-adic order of x

max{n ∈ N|pn divides x}, if x ∈ Z
ordp (x) =
ordp (a) − ordp (b), if x = a/b, b 6= 0, a, b ∈ Z
Remark 3.1. For x = a/b, y = c/d ∈ Q we have
ac
ordp ( ) = ordp (ac) − ordp (bd) (1)
bd
= ordp (a) + ordp (c) − ordp (b) − ordp (d)
a c
= ordp ( ) + ordp ( )
b d

7
and if x 6= −y
a c ad + cb
ordp ( + ) = ordp ( ) = ordp (ad + cb) − ordp (bd) (2)
b d bd
≥ min{ordp (ad), ordp (cb)} − ordp (b) − ordp (d)
= min{ordp (a) + ordp (d), ordp (c) + ordp (b)} − ordp (b) − ordp (d)
= min{ordp (a) − ordp (b), ordp (c) − ordp (d)}
a c
= min{ordp ( ), ordp ( )}
b d
Definition 3.2. On Q we define the p-adic norm
 −ord (x)
p p
, if x 6= 0
|x|p =
0, if x = 0

Proposition 3.1. |.|p is a non-Archimedean norm on Q.

Proof. |x|p = 0 ⇐⇒ x = 0 follows from the definition of the |.|p norm.

|xy|p = |x|p |y|p follows from (1).
For the strong triangle inequality we have

|x + y|p = p−ordp (x+y) ≤ max{p−ordp (x) , p−ordp (y) } = max{|x|p , |y|p }

from (2).
Remark 3.2. Unlike the euclidean norm on Q, given two numbers a, b ∈ Q with
|a|p < |b|p we can’t always find a third number c ∈ Q so that |a|p < |c|p < |b|p . In
particular, |.|p only takes values in {pk |k ∈ Z} ∪ {0}.

Definition 3.3. Let p ∈ N be prime. The field of p-adic numbers Qp is defined as

the completion of Q with respect to |.|p , and its elements are equivalence classes
of Cauchy sequences.
For an element a ∈ Qp and a Cauchy sequence (an ) representing a, we defined the
norm of a as
|a|p = lim |an |p
n→∞

Remark 3.3. By remark 3.2, the norm of a ∈ Qp only takes values in

{pk |k ∈ Z}∪{0}, just like |.|p . Also, if |a|p = pk 6= 0, then for any Cauchy sequence
representing a there is an N so that |an |p = pk for n > N .
How does an element of Qp look like? The following proposition will give a
way to construct one of them.

8
Proposition 3.2. Let d−m 6= 0 and 0 ≤ di < p integers. Then the partial sums of
the series

a = d−m p−m + d−m+1 p−m+1 + · · · + d−1 p−1 + d0 + d1 p + d2 p2 + . . . (3)

form a Cauchy sequence and therefore a is an element of Qp .

Proof. Let  > 0. Then we can find N ∈ N so that p−N < , and for n, k > N ,
WLOG k > n, we have
k
X n
X k
X
i
di p − di p i
= di pi ≤ max{|dn+1 pn+1 |p , . . . , |dk pk |p } ≤ p−N < 
i=−m i=−m p i=n+1 p

We might now wonder if every element of Qp looks like a series (3).

The following propositions will help us prove that every a ∈ Qp actually has a
unique Cauchy sequence representing it that looks like (3).

Proposition 3.3. Let x ∈ Q with |x|p ≤ 1. Then for any i there is a unique
integer α ∈ {0, 1, . . . , pi − 1} so that |x − α|p ≤ p−i .

Proof. Let x = a/b with a and b relatively prime. Since |x|p = p−ordp (a)+ordp (b) ≤ 1
we get ordp (b) = 0, that is, b and pi are relatively prime for any i. We can then
find integers m and n so that npi + mb = 1. For α = am we get
a a
|α − x|p = am − = |mb − 1|p ≤ |mb − 1|p = |npi |p = |n|p p−i ≤ p−i
b p b p

There is exactly a multiple cpi of pi so that cpi + α ∈ {0, 1, . . . , pi − 1}, and we

have
|cpi + α − x|p ≤ max{|α − x|p , |cpi |} ≤ max{p−i , p−i } = p−i

Theorem 3.4. Let a ∈ Qp with |a|p ≤ 1. Then there is exactly one Cauchy
sequence (an ) representing a so that for any i
i) 0 ≤ ai < pi
ii) ai ≡ ai+1 (mod pi )

Proof. Let (cn ) be a Cauchy sequence representing a. Since |cn |p → |a|p ≤ 1, there
is an N so that |cn |p ≤ 1 for any n > N . (If |a|p = 1 this still holds because of
remark 3.3)
By replacing the first N elements we can find an equivalent Cauchy sequence so

9
that |bn |p ≤ 1 for any n. Now, for every j = 1, 2, . . . let N (j) be so that N (j) ≥ j
and
|bi − bi0 |p ≤ p−j ∀i, i0 ≥ N (j)
From the previous proposition we know that for any j we can find integers 0 ≤
aj < pj (condition i)) so that

|aj − bN (j) |p ≤ p−j

These aj also satisfy condition ii):

|aj+1 − aj |p = |aj+1 − bN (j+1) + bN (j+1) − bN (j) + bN (j) − aj |p

≤ max{|aj+1 − bN (j+1) |p , |bN (j+1) − bN (j) |p , |bN (j) − aj |p }
≤ max{p−j−1 , p−j , p−j } = p−j

This sequence is equivalent to (bn ): for any j take i ≥ N (j)

|ai − bi |p = |ai − aj + aj − bN (j) + bN (j) − bi |p

≤ max{|ai − aj |p , |aj − bN (j) |p , |bN (j) − bi |p }
≤ max{p−j , p−j , p−j } = p−j

so |ai − bi | → 0.
Now, to show uniqueness, let (dn ) be another Cauchy sequence satisfying condi-
tions i) and ii) and let (an ) 6= (dn ), that is, for some i0 , ai0 6= di0 . Since ai0 and
di0 are between 0 and pi0 , ai0 6= di0 (mod pi0 ). From condition ii) we have that for
i > i0 , ai = ai0 6= di0 = di (mod pi0 ), that is, ai 6= di (mod pi0 ) and therefore

|ai − di |p > p−i0

doesn’t converge to 0 and (an ) and (dn ) aren’t equivalent.

Remark 3.4. For a ∈ Qp with |a|p ≤ 1 we can write the Cauchy sequence (an )
representing a from the previous proposition as

ai = d0 + d1 p + · · · + di−1 pi−1

for di ∈ {0, 1, . . . , p − 1} and a is represented by the convergent series

∞
X
a= di pi
i=0

which we can think of as a number written in base p which keeps extending to the
left
a = . . . dn . . . d1 d0

10
Remark 3.5. If a ∈ Qp with |a|p = pm > 1 then a0 = apm satisfies |a0 | = pm p−m = 1
and we can then write
∞
X ∞
X
a = a0 p−m = p−m ci p i = di pi
i=0 i=−m

with d−m = c0 6= 0 and a becomes a fraction in base p with finitely many digits
after the point and which extends infinitely to the left

a = . . . dn . . . d1 d0 .d−1 d−2 . . . d−m

Definition 3.4. This way of writing a ∈ Qp as a number written in base p which

keeps extending to the left is called the p-adic expansion of a. This will either look
like
a = . . . dn . . . d1 d0
for di ∈ {0, 1, . . . , p − 1}, if |a|p ≤ 1, or like

a = . . . dn . . . d1 d0 .d−1 d−2 . . . d−m

for di ∈ {0, 1, . . . , p − 1} and d−m 6= 0, if |a|p > 1.

11