A characterization of length-factorial Krull monoids (Q2075855)

Let \(H\) be a commutative semigroup with identity. Denote by \(H^{\times}\) the group of invertible elements of \(H\). Call \(H\) reduced if \(H^{\times} =\{1\}\) and denote by \(H_{\mathrm{red}} = \{aH\times :a\in H\}\) the associated reduced semigroup. An element \(u\in H\) is said to be cancellative if \(au=bu\) implies that \(a=b\) for all \(a,b,u\in H\). Call \(H\) cancellative if all elements of \(H\) are cancellative and call \(H\) unit cancellative if whenever \(a=au\) we must have \(u\in H^{\times}\), for all \(a,u\in H\). In a welcome break from the past here's the enunciation found on page 1351 of the paper under review: ``Throughout this paper, a monoid means a commutative and unit-cancellative semigroup with identity''. Indeed, as Example 2.2 of the paper under review shows there are non-cancellative monoids that are unit cancellative. Here's another enunciation at the beginning of the paper saying, ``By an atomic monoid, we mean a commutative unit-cancellative semigroup with identity in which every non-invertible element is a finite product of irreducible elements. The monoids we have in mind stem from ring and module theory.'' (Here \(a\in H\) is ``irreducible'' if \(a=xy\), for \(x,y\in H\), implies \(x\in H^{\times}\) or \(y\in H^{\times})\). An atomic monoid \(H\) may be ``half factorial'' if every pair of (atomic) factorizations have the same length. The authors call \(H\) ``length factorial'' if for each \(a\in H\) every pair of distinct (atomic) factorizations have distinct lengths. ``Thus, an atomic monoid is factorial if and only if it is half-factorial and length factorial.'' A Krull monoid may be defined in a manner similar to Krull domains as: The monoid \(H\) is a Krull monoid if there is a family \(\{v_{i\in I}\}\) of discrete valuations of \(\langle H\rangle\) such that \(S\) is the intersection of valuation semigroups of the \(v_{i}\), and for every \(x\in H\), \(|\{i\in I|v_{i}(x)>0\}|<\infty\). By \(\mathcal{F}(P)\), for \(P\) a set, we usually mean the free monoid generated by elements of \(P\). The elements of \(\mathcal{F}(P)\) are of the form \(a=\prod p^{v_{g}(a)}\) where \(v_{g}(a)\in N\cup \{0\}\). Also for \(P\subseteq \langle H\rangle\), \(B(P)=\{B\in \mathcal{F}(G)| \sum_{g\in P}v_{g}(B)g=0\in \langle H\rangle\}\). Then comes the main theorem: Let \(H\) be a Krull monoid. Then \(H=H^{\times}\times \mathcal{F} (P_{0})\times H^{\ast}\), where \(P_{0}\) is a set of representatives of prime elements of \(H\), \(\mathcal{F}(P_{0})\times H\ast \approxeq H_{\mathrm{red}}\), and \( H^{\ast}\) is a reduced Krull monoid without primes. The class groups \(C(H)\) of \(H\) and \(C(H^{\ast})\) of \(H^{\ast }\) are isomorphic, and \(H\) is length-factorial if and only if \(H^{\ast}\) is length-factorial. Let \(G_{P^{\ast}}\subset C(H^{\ast})\) denote the set of classes containing prime divisors. Then \(H\) is length-factorial but not factorial if and only if every class of \(G_{P^{\ast}}\) contains precisely one prime divisor, \(H^{\ast} \approxeq B(G_{P^{\ast}})\), \(G_{P^{\ast}} = \{e_{1,1},\dots ,e_{1,t},e_{2,1},\dots ,e_{2,t},\dots ,e_{k,1},\dots ,e_{k,t},g_{1},\dots ,g_{k},e_{0,1},\dots ,e_{0,t},g_{0}\}\), and \[ C(H^{\ast}) = \langle e_{1,1},\dots ,e_{1,t},g_{1}\rangle \bigoplus \dots\bigoplus \langle e_{k,1},\dots ,e_{k,t},g_{k}\rangle \approxeq (\mathbb{Z}^{t}\bigoplus \mathbb{Z}/n\mathbb{Z})^{k}, \] where \(t\in N_{0}\), \(k,s_{0},s_{1},\dots,s_{t}\in \mathbb{N}\) with \(k+1\neq s_{0}+s_{1}+\dots+s_{t}\geq 2\), independent elements \(e_{1,1},\dots,e_{1,t}, e_{2,1},\dots,e_{2,t},\dots,e_{k,1},\dots,e_{k,t} \in C(H^{\ast})\) of infinite order and independent elements \(g_{1},\dots ,g_{k}\in C(H^{\ast})\), which are of infinite order in case \(t>0\) and of finite order for \(t=0\); \(s_{0}\) is the smallest integer such that \(s_{0}g_{i}\in \langle e_{i,1},\dots ,e_{i,t}\rangle\) and \(-s_{0}g_{i} = s_{1}e_{i,1}+\dots +s_{t}e_{i,t}\) for every \(i\in \lbrack 1,k]\); \(e_{0,j} = - \sum_{i=1}^{k}e_{i,~j}\) for all \(j\in \lbrack 1,t]\), \(g_{0} =- \sum_{i=1}^{k}g_{i}\) and \(n = \gcd (s_{0},\dot,s_{t})\). Moreover, \(C(H^{\ast})\) is a torsion group if and only if \(t=0\) and in that case we have \(C(H^{\ast})\approxeq (\mathbb{Z}/n\mathbb{Z})^{k}\), where \(n>2\), \(k\in N\) with \(k+1\neq n\), and \(\mathrm{ord}(g_{i}) = n\) for all \(i\in \lbrack 1,k]\).