RELATIVE RAMIFICATION AND INERTIA DEGREES. IAN KIMING Let K be an algebraic number eld with ring of integers O, and let p be a prime ideal of O and Kp the completion of K w.r.t. p . Let us choose a prime element 2 p, i.e., an element with p () = 1. The prime ideal p divides some prime number p. Let e and f denote the ramication index and the inertia degree of p, respectively. I.e., we have p (p) = e and #(O=p) = pf : So, e 1 p is an extension of the exponential valuation p on Q attached to the prime number p, and as we discussed earlier, we can in an obvious fashion view Qp as sitting inside Kp . It is clear that the eld extension Kp =Qp is a nite extension (if is a primitive element for the extension K=Q then also for Kp =Qp ). We will prove the following statement which is a `local' version of equation (3.5.3) of [1] (for number elds). Proposition 1. In the above situation we have: ef = [Kp : Qp ] : Before the proof we need a number of simple remarks. First notice that for m 2 N the ideal (pOp )m = pm Op is a principal ideal generated by m : For clearly m 2 pm Op since 2 p. On the other hand, if 2 pm Op we see that p () m (use that p (x1 + : : : + xs ) minfp (xi )g). Then p ( m ) 0 so that m 2 Op , hence 2 m Op . So pm Op = m Op , and we also see that pm Op consists precisely of the elements of Kp that have valuation m. In a similar fashion we prove the following: Op . Then a = pm Op where m := minfp () j 2 ag : In particular, a is a principal ideal, and Op is a principal ideal domain. Proof. Notice that m < 1 since a 6= 0. If 2 a then p () m so by the above we have 2 pm Op . Suppose that 2 pm Op . Then p ( ) m. Choose 2 a with p () = m. Then 6= 0 and p ( 1 ) 0 so that 1 2 Op . We deduce 2 Op a. Proposition 2. Let a be a non-zero ideal of Corollary 1. We have pOp = pe Op : Proof. Immediate from the previous proposition as p (p) = e. 1 2 IAN KIMING Applying Proposition 2 to the case K = Q we have that Zp is a principal ideal domain. Clearly Zp Op . In particular we can view Op as a Zp -module. As such Op is torsion-free as a Zp -module: This follows because Kp is a eld. We then claim that Op is nitely generated as Zp -module: If it were not there would be no bound on sizes of nite sets of Zp -linearly independent elements of Op ; but any nite number of Zp -linearly independent elements of Op are also Qp -linearly independent (Qp is the quotient eld of Zp ). Since Kp has nite dimension as vector space over Qp we reach a contradiction, and the claim follows. So Op is a torsion-free, nitely generated module over the principal ideal domain Zp . By the structure theory of torsion-free, nitely generated modules over a principal ideal domain (accept this, or consult for instance Appendix A.4 of [1]) we have () Op = Zp 1 : : : Zp r with certain elements 1 ; : : : ; r () 2 Op . We claim that r = [Kp : Qp ] : For rst of all we clearly have r [Kp : Qp ] by the above independence argument. On the other hand, Kp has a Qp basis 1 ; : : : ; s where s := [Kp : Qp ]. Now, if t is suciently large we have p (pt i ) 0 for all i, i.e., pt i 2 Op for all i. But then pt 1 ; : : : ; pt s are Zp -linearly independent elements of Op . By the theory of torsion-free, nitely generated modules over a principal ideal domain this implies s r, and we are done. Proof of Proposition 1. Look at the quotient Op =pOp : By () above we have Op =pOp = (Zp =pZp )r = (Fp )r where r = [Kp : Qp ] by () above. In particular, (y) #(Op =pOp ) = p[Kp :Qp ] : On the other hand, by Corollary 1 we have pOp = pe Op so that Op =pOp = Op =pe Op : Now notice that we have pm 1 Op =pm Op = Op =pOp (as abelian groups) for any m 2 N. This can be seen because pt Op = t Op : Then map x 7! 1 m x then gives the desired isomorphism. Consequently, we easily deduce by induction on v u that we have #(pu Op =pv Op ) = #(Op =pOp )v u for v u. In particular, (z) #(Op =pOp ) = #(Op =pe Op ) = #(Op =pOp )e = #(O=p)e = pef where we used the isomorphism Op =pOp = O=p discussed earlier. Comparing (y) and (z) the proposition follows. RELATIVE RAMIFICATION AND INERTIA DEGREES. 3 Now let us look at a relative situation: Suppose that L is a nite extension of K , OL the ring of integers of L, and P a prime ideal of OL that divides the given prime p of K : Q K L [ [ [ OK OL [ [ [ p p P Z Considering the completions we have then a tower of eld extensions Qp Kp LP : One denes the relative ramication index of LP over Kp (or sometimes one says just of P over p), denoted by e(LP =Kp ) (or just as e(P=p)), as the exponent of P in the prime factorization (in OL ) of the ideal p OL . Thus, the number e that we had attached to K in the beginning is in fact the relative ramication index of Kp over Qp and could also be denoted e(Kp =Qp ). Similarly, one has the notion of relative inertia degree of LP over Kp (or of P over p): There is a natural homomorphism OK ! OL =P; the kernel of this is some ideal a of OK ; then OK =a is a subring of the nite eld OL =P; so a is a prime ideal of OK and hence also a maximal ideal; we have that a is contained in P; but p is also a maximal ideal of OK contained in P; if we had a 6= p then a + p = OK , and we would deduce the contradiction 1 2 P; so, a = p, and we have an injection of nite elds OK =p ,! OL =P : The above relative inertia degree, denoted by f (LP =Kp ) (or just as f (P=p)), is dened as the degree of this extension of nite elds, i.e., as the dimension of OL =P as vector space over OK =p. We say that LP =Kp is unramied (or that P is unramied over p) if e(LP =Kp ) = 1 : We say that LP =Kp is totally ramied (or that P is totally ramied over p) if f (LP =Kp ) = 1 : By the next proposition we will then have: LP =Kp unramied , e(LP =Kp ) = 1 , f (LP =Kp ) = [LP : Kp ] ; and LP =Kp totally ramied , f (LP =Kp ) = 1 , e(LP =Kp ) = [LP : Kp ] : We have now the following easy consequence of the denitions and results above. 4 IAN KIMING Proposition 3. We have e(LP =Qp ) = e(LP =Kp ) e(Kp =Qp ) ; f (LP =Qp ) = f (LP =Kp ) f (Kp =Qp ) ; and [LP : Kp ] = e(LP =Kp ) f (LP =Kp ) : Proof. By denition the exponent of p in the prime factorization of pOK is e(Kp =Qp ). Similarly, the exponent of P in the prime factorization of pOL is e(LP =Kp ). So, the exponent of P in the prime factorization of pOL is e(Kp =Qp ) e(LP =Kp ); but by denition this exponent is e(LP =Qp ). The second claim follows from denitions thus: pf (LP =Qp ) = #(Z=p)f (LP =Qp ) = #(OL =P) = #(OK =p)f (LP =Kp ) = #((Z=p)f (Kp =Qp ) )f (LP =Qp ) = pf (LP =Kp )f (Kp =Qp ) : Now, the third claim follows from the 2 rst and Proposition 1: [L : Q ] e(LP =Qp ) f (LP =Qp ) [LP : Kp ] = P p = = e(LP =Kp ) f (LP =Kp ) : [Kp : Qp ] e(Kp =Qp ) f (Kp =Qp ) Finally, we can easily derive from the last proposition the analogous statement in a truly relative situation: Suppose further that M is a nite extension of L, and e is a prime of M dividing P. Then we have: that P e(MPe =Kp ) = e(MPe =LP ) e(LP =Kp ) ; f (MPe =Kp ) = f (MPe =LP ) f (LP =Kp ) ; and of course the following statement that is already in the proposition: [MP e =Kp ) : e =Kp ) f (MP e : Kp ] = e(MP The proofs of the rst 2 statements are easy consequences of Proposition 3 and are left to the reader. References [1] H. Koch: `Number Theory. Algebraic Numbers and Functions'. Graduate Studies in Mathematics 24, AMS 2000. Department of Mathematics, University of Copenhagen, Universitetsparken 5, DK2100 Copenhagen , Denmark. E-mail address : [email protected]
© Copyright 2026 Paperzz