In this chapter $$p$$ is always a (fixed) prime integer.

The $$p$$-adic numbers $$Q_p$$ are the completion of the rational numbers with respect to the valuation $$\nu_p( p^v \cdot a / b) = v$$ if $$p$$ divides neither $$a$$ nor $$b$$. They form a field of characteristic 0 which nevertheless shows some behaviour of the finite field with $$p$$ elements.

A $$p$$-adic numbers can be represented by a "$$p$$-adic expansion" which is similar to the decimal expansion used for the reals (but written from left to right). So for example if $$p = 2$$, the numbers $$1$$, $$2$$, $$3$$, $$4$$, $$1/2$$, and $$4/5$$ are represented as $$1(2)$$, $$0.1(2)$$, $$1.1(2)$$, $$0.01(2)$$, $$10(2)$$, and the infinite periodic expansion $$0.010110011001100...(2)$$. $$p$$-adic numbers can be approximated by ignoring higher powers of $$p$$, so for example with only 2 digits accuracy $$4/5$$ would be approximated as $$0.01(2)$$. This is different from the decimal approximation of real numbers in that $$p$$-adic approximation is a ring homomorphism on the subrings of $$p$$-adic numbers whose valuation is bounded from below so that rounding errors do not increase with repeated calculations.

In GAP, $$p$$-adic numbers are always represented by such approximations. A family of approximated $$p$$-adic numbers consists of $$p$$-adic numbers with a fixed prime $$p$$ and a certain precision, and arithmetic with these numbers is done with this precision.

Pure $$p$$-adic numbers are the $$p$$-adic numbers described so far.

 ‣ PurePadicNumberFamily( p, precision ) ( function )

returns the family of pure $$p$$-adic numbers over the prime p with precision "digits". That is to say, the approximate value will differ from the correct value by a multiple of $$p^{digits}$$.

 ‣ PadicNumber( fam, rat ) ( operation )

returns the element of the $$p$$-adic number family fam that approximates the rational number rat.

$$p$$-adic numbers allow the usual operations for fields.

gap> fam:=PurePadicNumberFamily(2,20);;
0.010110011001100110011(2)
0.0101(2)
gap> 3*a;
0.0111(2)
gap> a/2;
0.101(2)
gap> a*10;
0.001(2)


See PadicNumber (68.2-2) for other methods for PadicNumber.

##### 68.1-3 Valuation
 ‣ Valuation( obj ) ( operation )

The valuation is the $$p$$-part of the $$p$$-adic number. See also PValuation (15.7-1).

 ‣ ShiftedPadicNumber( padic, int ) ( operation )

ShiftedPadicNumber takes a $$p$$-adic number padic and an integer shift and returns the $$p$$-adic number $$c$$, that is padic * $$p$$^shift.

 ‣ IsPurePadicNumber( obj ) ( category )

The category of pure $$p$$-adic numbers.

 ‣ IsPurePadicNumberFamily( fam ) ( category )

The family of pure $$p$$-adic numbers.

#### 68.2 Extensions of the p-adic Numbers

The usual Kronecker construction with an irreducible polynomial can be used to construct extensions of the $$p$$-adic numbers. Let $$L$$ be such an extension. Then there is a subfield $$K < L$$ such that $$K$$ is an unramified extension of the $$p$$-adic numbers and $$L/K$$ is purely ramified.

(For an explanation of "ramification" see for example [Neu92, Section II.7], or another book on algebraic number theory. Essentially, an extension $$L$$ of the $$p$$-adic numbers generated by a rational polynomial $$f$$ is unramified if $$f$$ remains squarefree modulo $$p$$ and is completely ramified if modulo $$p$$ the polynomial $$f$$ is a power of a linear factor while remaining irreducible over the $$p$$-adic numbers.)

The representation of extensions of $$p$$-adic numbers in GAP uses the subfield $$K$$.

 ‣ PadicExtensionNumberFamily( p, precision, unram, ram ) ( function )

An extended $$p$$-adic field $$L$$ is given by two polynomials $$h$$ and $$g$$ with coefficient lists unram (for the unramified part) and ram (for the ramified part). Then $$L$$ is isomorphic to $$Q_p[x,y]/(h(x),g(y))$$.

This function takes the prime number p and the two coefficient lists unram and ram for the two polynomials. The polynomial given by the coefficients in unram must be a cyclotomic polynomial and the polynomial given by ram must be either an Eisenstein polynomial or $$1+x$$. This is not checked by GAP.

Every number in $$L$$ is represented as a coefficient list w. r. t. the basis $$\{ 1, x, x^2, \ldots, y, xy, x^2 y, \ldots \}$$ of $$L$$. The integer precision is the number of "digits" that all the coefficients have.

A general comment:

The polynomials with which PadicExtensionNumberFamily is called define an extension of $$Q_p$$. It must be ensured that both polynomials are really irreducible over $$Q_p$$! For example $$x^2+x+1$$ is not irreducible over $$Q_p$$. Therefore the "extension" PadicExtensionNumberFamily(3, 4, [1,1,1], [1,1]) contains non-invertible "pseudo-p-adic numbers". Conversely, if an "extension" contains noninvertible elements then one of the defining polynomials was not irreducible.

 ‣ PadicNumber( fam, rat ) ( operation )
 ‣ PadicNumber( purefam, list ) ( operation )
 ‣ PadicNumber( extfam, list ) ( operation )

(see also PadicNumber (68.1-2)).

PadicNumber creates a $$p$$-adic number in the $$p$$-adic numbers family fam. The first form returns the $$p$$-adic number corresponding to the rational rat.

The second form takes a pure $$p$$-adic numbers family purefam and a list list of length two, and returns the number $$p$$^list * list. It must be guaranteed that no entry of list is divisible by the prime $$p$$. (Otherwise precision will get lost.)

The third form creates a number in the family extfam of a $$p$$-adic extension. The second argument must be a list list of length two such that list is the list of coefficients w.r.t. the basis $$\{ 1, \ldots, x^{{f-1}} \cdot y^{{e-1}} \}$$ of the extended $$p$$-adic field and list is a common $$p$$-part of all these coefficients.

$$p$$-adic numbers admit the usual field operations.

gap> efam:=PadicExtensionNumberFamily(3, 5, [1,1,1], [1,1]);;


A word of warning:

Depending on the actual representation of quotients, precision may seem to "vanish". For example in PadicExtensionNumberFamily(3, 5, [1,1,1], [1,1]) the number (1.2000, 0.1210)(3) can be represented as [ 0, [ 1.2000, 0.1210 ] ] or as [ -1, [ 12.000, 1.2100 ] ] (here the coefficients have to be multiplied by $$p^{{-1}}$$).

So there may be a number (1.2, 2.2)(3) which seems to have only two digits of precision instead of the declared 5. But internally the number is stored as [ -3, [ 0.0012, 0.0022 ] ] and so has in fact maximum precision.

 ‣ IsPadicExtensionNumber( obj ) ( category )

The category of elements of the extended $$p$$-adic field.

gap>  efam:=PadicExtensionNumberFamily(3, 5, [1,1,1], [1,1]);;
true


 ‣ IsPadicExtensionNumberFamily( fam ) ( category )
Family of elements of the extended $$p$$-adic field.
gap> efam:=PadicExtensionNumberFamily(3, 5, [1,1,1], [1,1]);;