In mathematics, a p-adically closed field is a field that enjoys a closure property that is a close analogue for p-adic fields to what real closure is to the real field. They were introduced by James Ax and Simon B. Kochen in 1965.
Let K be the field ℚ of rational numbers and v be its usual p-adic valuation (with
v
(
p
)
=
1
). If F is a (not necessarily algebraic) extension field of K, itself equipped with a valuation w, we say that
(
F
,
w
)
is formally p-adic when the following conditions are satisfied:
w extends v (that is,
w
(
x
)
=
v
(
x
)
for all x in K),
the residue field of w coincides with the residue field of v (the residue field being the quotient of the valuation ring
{
x
∈
F
:
w
(
x
)
≥
0
}
by its maximal ideal
{
x
∈
F
:
w
(
x
)
>
0
}
),
the smallest positive value of w coincides with the smallest positive value of v (namely 1, since v was assumed to be normalized): in other words, a uniformizer for K remains a uniformizer for F.
(Note that the value group of K may be larger than that of F since it may contain infinitely large elements over the latter.)
The formally p-adic fields can be viewed as an analogue of the formally real fields.
For example, the field ℚ(i) of Gaussian rationals, if equipped with the valuation w given by
w
(
2
+
i
)
=
1
(and
w
(
2
−
i
)
=
0
) is formally 5-adic (the place v=5 of the rationals splits in two places of the Gaussian rationals since
X
2
+
1
factors over the residue field with 5 elements, and w is one of these places). The field of 5-adic numbers (which contains both the rationals and the Gaussian rationals embedded as per the place w) is also formally 5-adic. On the other hand, the field of Gaussian rationals is not formally 3-adic for any valuation, because the only valuation w on it which extends the 3-adic valuation is given by
w
(
3
)
=
1
and its residue field has 9 elements.
When F is formally p-adic but that there does not exist any proper algebraic formally p-adic extension of F, then F is said to be p-adically closed. For example, the field of p-adic numbers is p-adically closed, and so is the algebraic closure of the rationals inside it (the field of p-adic algebraic numbers).
If F is p-adically closed, then:
there is a unique valuation w on F which makes F p-adically closed (so it is legitimate to say that F, rather than the pair
(
F
,
w
)
, is p-adically closed),
F is Henselian with respect to this place (that is, its valuation ring is so),
the valuation ring of F is exactly the image of the Kochen operator (see below),
the value group of F is an extension by ℤ (the value group of K) of a divisible group, with the lexicographical order.
The first statement is an analogue of the fact that the order of a real-closed field is uniquely determined by the algebraic structure.
The definitions given above can be copied to a more general context: if K is a field equipped with a valuation v such that
the residue field of K is finite (call q its cardinal and p its characteristic),
the value group of v admits a smallest positive element (call it 1, and say π is a uniformizer, i.e.
v
(
π
)
=
1
),
K has finite absolute ramification, i.e.,
v
(
p
)
is finite (that is, a finite multiple of
v
(
π
)
=
1
),
(these hypotheses are satisfied for the field of rationals, with q=π=p the prime number having valuation 1) then we can speak of formally v-adic fields (or
p
-adic if
p
is the ideal corresponding to v) and v-adically complete fields.
If K is a field equipped with a valuation v satisfying the hypothesis and with the notations introduced in the previous paragraph, define the Kochen operator by:
γ
(
z
)
=
1
π
z
q
−
z
(
z
q
−
z
)
2
−
1
(when
z
q
−
z
≠
±
1
). It is easy to check that
γ
(
z
)
always has non-negative valuation. The Kochen operator can be thought of as a p-adic (or v-adic) analogue of the square function in the real case.
An extension field F of K is formally v-adic if and only if
1
π
does not belong to the subring generated over the value ring of K by the image of the Kochen operator on F. This is an analogue of the statement (or definition) that a field is formally real when
−
1
is not a sum of squares.
The first-order theory of p-adically closed fields (here we are restricting ourselves to the p-adic case, i.e., K is the field of rationals and v is the p-adic valuation) is complete and model complete, and if we slightly enrich the language it admits quantifier elimination. Thus, one can define p-adically closed fields as those whose first-order theory is elementarily equivalent to that of
Q
p
.