Almost fully decomposable infinite rank lattices over orders. (Q995597)

Let \(R\) be a Dedekind domain and let \(K\) be its field of fractions. Let \(A\) be a finite dimensional \(K\)-algebra. An `\(R\)-order in \(A\)' is an \(R\)-submodule \(\Lambda\) of \(A\) (a submodule of \(A\) as an \(R\)-module) which satisfies (i) \(\Lambda\) is finitely generated as an \(R\)-module, (ii) \(\Lambda\) is closed under the multiplication of \(A\); (iii) \(K\Lambda=A\). For example, let \(p\) be a rational prime, let \(\zeta\) denote a primitive \(p\)-th root of unity, and put \(K=\mathbb{Q}(\zeta)\) so that \(R=\mathbb{Z}[\zeta]\) is the ring of integers in \(K\). Let \(C_p\) denote the cyclic group of order \(p\). Then \(\Lambda=RC_p\) is an \(R\)-order in \(A=KC_p\). Let \(\Lambda\) be an \(R\)-order in \(A\). A `\(\Lambda\)-lattice' is a (left) \(\Lambda\)-module \(L\) for which \(L\) is finitely generated and projective as an \(R\)-module. A `generalized \(\Lambda\)-lattice' is a (left) \(\Lambda\)-module \(L\) for which \(L\) is projective as an \(R\)-module. In the case \(\Lambda=\mathbb{Z}[\zeta]C_p\) given above, the linear dual \(\Lambda^*=\Hom_R(\Lambda,R)\) is a generalized \(\Lambda\)-lattice in a natural way: The algebra map \(m\colon\Lambda\otimes\Lambda\to\Lambda\) dualizes to yield the map \(\Delta\colon\Lambda^*\to\Lambda^*\otimes\Lambda^*\), from which one can define a left action of \(\Lambda\) on \(\Lambda^*\): for \(\lambda,x\in\Lambda\), \(f\in\Lambda^*\), \((\lambda\cdot f)(x)=\sum_{(f)}f_{(2)}(\lambda)f_{(1)}(x)\), where \(\Delta(f)=\sum_{(f)}f_{(1)}\otimes f_{(2)}\). A generalized \(\Lambda\)-lattice \(L\) is `fully decomposable' if \(L\cong\coprod_{i\in I}E_i\) where the \(E_i\) are \(\Lambda\)-lattices. \(L\) is `almost fully decomposable' if the \(A\)-module \(KL=K\otimes_RL\) decomposes as \(M_0\oplus M_1\) where \(M_0\) is finitely generated and \(L\cap M_1\) is a fully decomposable generalized \(\Lambda\)-lattice. Note that \((\mathbb{Z}[\zeta]C_p)^*\) is fully decomposable: One has \((\mathbb{Z}[\zeta]C_p)^*\cong\underbrace{\mathbb{Z}[\zeta]\oplus\mathbb{Z}[\zeta]\oplus\cdots\oplus\mathbb{Z}[\zeta]}_p\) where \(\mathbb{Z}[\zeta]\) is a \(\mathbb{Z}[\zeta]C_p\)-lattice with left action defined as \((\sum_{g\in C_p}a_gg)\cdot x=\varepsilon(\sum_{g\in C_p}a_gg)x\), with \(\varepsilon\colon\mathbb{Z}[\zeta]C_p\to\mathbb{Z}[\zeta]\), \(g\mapsto 1\), \(\forall g\in C_p\). Certainly, \((\mathbb{Z}[\zeta]C_p)^*\) is almost fully decomposable. The paper under review concerns almost fully decomposable generalized \(\Lambda\)-lattices which are not fully decomposable. Let \(\Lambda\)-Lat denote the additive category of generalized \(\Lambda\)-lattices. The author's main result is the following (Theorem 2). Let \(R\) be a Dedekind domain with quotient field \(K\), and let \(\Lambda\) be an \(R\)-order in a finite dimensional \(K\)-algebra \(A\) such that \(A/\text{Rad}(A)\) is separable over \(K\). Assume that \(A_\nu\) is not semisimple for some \(\nu\in\text{Spec\,}R\). Then there exists an almost fully decomposable, but not fully decomposable \(L\in\Lambda\)-Lat with \(\dim(KL)=\aleph_0\). A question comes to mind: Let \(R=\mathbb{Z}[\zeta]\), \(K=\mathbb{Q}(\zeta)\), and let \(\Lambda\) be an \(R\)-order in \(KC_p\). Does there exist an almost fully decomposable \(L\in\Lambda\)-Lat which is not fully decomposable?