|This article needs additional citations for verification. (December 2010)|
That is, L is viewed as a vector space over K, and mα is a linear transformation of this vector space into itself. The norm NL/K(α) is defined as the determinant of this linear transformation. Properties of the determinant imply that the norm belongs to K and
- NL/K(αβ) = NL/K(α)NL/K(β)
so that the norm, when considered on non-zero elements, is a group homomorphism from the multiplicative group of L to that of K.
of α, for g in the Galois group G of L/K.
- x + iy
- x2 + y2. This corresponds to the Galois group having two elements, the identity element which sends x + iy to itself, and complex conjugation, which sends x + iy to x − iy.
Notice that in this case the norm is the square of the "usual" norm, in the sense that we would normally consider the above number x+iy as having norm equal to the Euclidean distance from the origin in the complex plane, which is . In some sense therefore we need to speak of this field norm as an `arithmetic norm' as opposed to the usual Euclidean (distance) norm.
This is particularly important in algebraic number theory where we may have a number field K (a finite extension of the rationals which for simplicity we assume to be Galois), which is naturally itself a vector space of finite dimension d say over . The field norm from K down to of any element will be as defined above, namely the product of all of its d Galois conjugates. However we may also embed K in a standard way into the real or complex numbers or (according to whether K contains any imaginary numbers), and we then consider the Norm (mathematics) or absolute value of the element within the larger field in the usual way. This in general will be very different from the field norm. We give an example below.
Finally, if (as is common in algebraic number theory) we consider K as embedded inside a real vector space then both of these norms will in general be very different from the Euclidean norm inside (which is basis-dependent - see below).
As an example of this, consider the number field obtained from by adjoining , which is the splitting field for the polynomial . The Galois group of this extension is of order d=2, generated by the element which sends to , and so the norm of the number is:
We may also see this using the terminology of "formal definition 1" above if we fix a basis for the extension , say : then multiplication by the number sends 1 to while is mapped to . Hence the determinant of the mapping is that of the matrix which sends the vector (corresponding to the first basis element, ie 1) to and the vector (which represents the second basis element ) to , viz.:
The determinant of this matrix is again -1, showing that the two "formal" definitions of norm agree. However, if we view K as a subfield of the complex numbers in the normal way, then the norm (ie the absolute value) is of course .
Finally if we embed K inside then taking as a basis for K over we see that is represented by the vector and so its Euclidean norm inside is .
The norm of an algebraic integer is again an integer, because it is equal (up to sign) to the constant term of the characteristic polynomial.
In algebraic number theory one defines also norms for ideals. This is done in such a way that if I is an ideal of OK, the ring of integers of the number field K, N(I) is the number of residue classes in OK/I – i.e. the cardinality of this finite ring. Hence this norm of an ideal is always a positive integer. When I is a principal ideal αOK then N(I) is equal to the absolute value of the norm to Q of α, for α an algebraic integer.