Denominators of algebraic numbers in a number field (Q2254204)

For \( \gamma \) an algebraic number of degree \( n\geq 1 \) over \( \mathbb{Q}, \) let \( f(x)=a_nx^n+\dots+a_0\in\mathbb{ Z}[x]\) be the unique irreducible polynomial such that \(f(\gamma)=0\) and \(a_n>0. \) The leading coefficient \( a_n \) will be denoted by \( c(\gamma). \) The smallest positive integer \( d \) such that \( d\cdot \gamma \) is an algebraic integer is called the \textit{denominator} of \( \gamma \) and will be denoted by \( d(\gamma). \) It can be proved that \( d(\gamma)\mid c(\gamma) \mid d(\gamma)^n.\) Thus, if \( k=k(p)=v_p(d(\gamma))\) and \( t=t(p)=v_p(c(\gamma)), \) it follows that \[ k\leq t\leq nk, \] where, as usual, \(v_p(a) \) denotes the power of the prime \( p \) in the decomposition of the positive integer \(a\). Given a number field \(K,\) let \( \hat{K} \) denote the set of those elements of \( K \) which are primitive over \(\mathbb Q\) and let \( A \) denote the ring of integers of \( K. \) For a fixed prime number \( p \) and a fixed positive integer \( k\) let \[ T_p(k)=\{ t\geq 1:\text{there exists~}\gamma\in\hat{K}\text{~such that~}v_p(d(\gamma))=k\text{~and~}v_p(c(\gamma))=t \}. \] The authors study the relation between \(v_p(d(\gamma))\) and \( v_p(c(\gamma)) \) when \( \gamma\in \hat{K}, \) for \( K \) a fixed algebraic number field. More precisely, they completely determine the set \( T_p(k) \) in the case when \( K \) is an arbitrary number field and \( p \) is inert in \( K, \) and when \( K/\mathbb{Q} \) is a Galois number field and \( p \) splits completely or is the power of a prime ideal in \(K.\) The authors also prove that if \( K/\mathbb Q \) is a Galois number field and the prime \( p \) splits in \( A \) as a product of prime ideals in the form \( (p)=(\mathfrak p_1\cdot\dots\cdot \mathfrak p_r)^e, \) with \( r>1, \) then \[ |T_p(k)|\geq ek+1. \] These results indicate that \(T_p(k)\) depends neither on \(p\) itself nor on \(K,\) but rather on how \(p\) splits in \(K,\) the authors being led to claim the following. Conjecture. Let \(K\) be a Galois number field, \(A\) its ring of integers and \(p\) a prime number whose splitting as a product of prime ideals in \(A\) has the form: \[ (p)=(\mathfrak p_1\cdot\dots\cdot \mathfrak p_r)^e, \] where the common residual degree of \( \mathfrak p_i,~i=1,\dots,r, \) is equal to \( f. \) Then, for any positive integer \( k, \) we have \[ T_p(k)=\{nk-hf:h=0,\dots,(r-1)k+e-1\}. \] However, as it stands, the conjecture contradicts Theorem 3.1 from the paper, that is, it contradicts the lower bound on \(|T_p(k)|\) established above, as one might readily see by setting, for example, \( e=k=r=2 \) and \(f=1.\) It is likely that the authors meant \[ T_p(k)=\{nk-hf:h=0,\dots,(r-1)ek+e-1\}. \] The appendix written by Claus Fieker provides a complete characterization of the set \( T_p(k) \) in terms of the ramification and inertia degrees of the prime ideals \(\mathfrak p_i\) lying above \(p \) in \( A \) and concludes with a proof of the above conjecture. Nevertheless, the careful reader will notice yet another typo in the expression of the set \[ T_p(k)=\{nk-hf:nk-hf\geq f((e-1)k+1)\}, \] as it appears in the paper. Namely, the correct definition should be \[ T_p(k)=\{nk-hf:nk-hf\geq f(e(k-1)+1)\}. \] It is worth noting the fact that the results are proved using rather elementary, but very skillfully and cleverly applied techniques from algebraic number theory, and rely on a very brief list of references (which could, and should have been, in fact, slightly extended). As such, the paper is accessible to anyone possessing a good knowledge of basic algebraic number theory and Galois theory.