Algebras whose multiplication algebra is semiprime. A decomposition theorem (Q2475068)

In 1937, \textit{N. Jacobson} [ Duke Math. J. 3, 544--548 (1937; Zbl 0018.05005, JFM 63.0088.03)], proved the following structure theorem for non-necessarily associative algebras: For an algebra \(A\) the following assertions are equivalent: (1) \(A\) is finite-dimensional and \(M(A)\) is semiprime. (2) \(A =\bigoplus_{i=0}^n B_i\) is a direct sum of ideals, one of them, say \(B_0\), is a finite-dimensional null algebra and the others are finite-dimensional simple algebras. Moreover, \(B_0 = \text{Ann}(A)\) (the annihilator of \(A\)) and \(M(B_i)\) is isomorphic to a full matrix algebra over a finite-dimensional division algebra \(\Delta_i\), for each \(1\leq i\leq n\), where for an algebra \(A\), \(M(A)\) means the subalgebra of \(\text{End}(A)\) generated by the identity and all left and right multiplication operators. Note that for associative algebras with zero annihilator, Jacobson's Theorem is the Classical Wedderburn Decomposition Theorem. In [Proc. Lond. Math. Soc. (3) 24, 307--323 (1972; Zbl 0232.46041)], \textit{B. Yood} gave a topological extension of the Wedderburn Theorem and in [J. Algebra 282, No. 1, 386--421 (2004; Zbl 1124.17002)], the authors of the paper under review, motivated by the lattice version of Yood's Theorem given in [\textit{A. Fernández López} and \textit{M. I. Tocón Barroso}, J. Algebra 242, No. 1, 60--91 (2001; Zbl 0985.06004)], obtained a Yood's Theorem for algebras with zero annihilator: For an algebra \(A\) with zero annihilator, the following conditions are equivalent: (i) \(A\) is \(\varepsilon\)-decomposable. (ii) \(A\) is an \(\varepsilon\)-atomic multiplicatively semiprime algebra. (iii) \(\varepsilon\)-\(\text{Rad}(A) = 0\). Moreover, in this case, every minimal \(\varepsilon\)-closed ideal of \(A\) is a multiplicatively prime algebra (recall that an algebra \(A\) is said to be \(\varepsilon\)-decomposable if \(A =(\sum_{B\in\mathbf m} B)^\wedge\) with \(\mathbf m\) the set of all minimal \(\varepsilon\)-closed ideals of \(A\), an algebra \(A\) is said to be \(\varepsilon\)-atomic if each nonzero \(\varepsilon\)-closed ideal of \(A\) contains a minimal \(\varepsilon\)-closed ideal, the \(\varepsilon\)-radical of \(A\) is defined by \(\varepsilon\)-\(\text{Rad}(A) =\bigcap_{U\in \mathbf M} U\) with \(\mathbf M\) the set of all maximal \(\varepsilon\)-closed ideal of \(A\), and \((-)^\wedge\) denotes the closure operator associated to the Galois connection between the lattice of ideals of an algebra \(A\) and the lattice of ideals of its algebra of multiplication given by the pair of maps \(I\to I^{\text{ann}}\) and \({\mathcal P}\to {\mathcal P}_{\text{ann}}\) where \[ I^{\text{ann}}:=\{F\in M(A)\;|\;F(I)=0\}\quad\text{ and }\quad{\mathcal P}_{\text{ann}}:=\{a\in A\;|\;{\mathcal P}(a)=0\} \] i.e., the closure of an ideal \(I\) of \(A\) is \(I^\wedge:=(I^{\text{ann}})_{\text{ann}}\) and the closure of an ideal \({\mathcal P}\) of \(M(A)\) is \({\mathcal P}^\vee =({\mathcal P}_{\text{ann}})^{\text{ann}}\) and therefore an ideal \(I\) of \(A\) is \(\varepsilon\)-closed if \(I=I^\wedge\) and an ideal \(\mathcal P\) of \(M(A)\) is \(\varepsilon'\)-closed if \({\mathcal P}={\mathcal P}^\vee\). Finally, An algebra \(A\) is called multiplicatively semiprime (prime) if both \(A\) and \(M(A)\) are semiprime (prime) algebras). In the paper under review the authors avoid the assumption of zero annihilator imposed in this previous result, and thus obtain an infinite-dimensional extension of Jacobson's Theorem. Let us denote by \(\alpha(A)\) the multiplicative semiprime radical of \(A\). The authors prove that if \(M(A)\) is semiprime, \(\alpha(A)=\text{Ann}(A)\) (this result follows, in the finite dimensional case, as a consequence of Jacobson's Theorem). Moreover, they prove that for every algebra \(A\), \(\text{Ann}(A)\) is zero or a minimal \(\varepsilon\)-closed ideal of \(A\) and that, if \(\text{Ann}(A)\) is nonzero and \(\varepsilon\)-quasicomplemented in \(A\) (an \(\varepsilon\)-closed ideal \(U\) of an algebra \(A\) will be said to be \(\varepsilon\)-quasicomplemented in \(A\) if there exists an \(\varepsilon\)-closed ideal \(V\) of \(A\) such that \(A = (U \oplus V)^\wedge\)) then \(A=\text{Ann}(A)\oplus A^2\) with \(A^2\) the unique \(\varepsilon\)-quasicomplement of \(\text{Ann}(A)\). The main result of the paper states: For an algebra \(A\), the following conditions are equivalent: (i) \(A\) is \(\varepsilon\)-decomposable. (ii) \(A\) is strongly \(\varepsilon\)-atomic and \(M(A)\) is semiprime. (iii) \([\varepsilon\text{-}\text{Rad}(A),A] = 0\) and \(\mathbf M^*=\{\text{Ann}(B):\;B\in \mathbf m^*\}\). Moreover, in this case, every \(B\in\mathbf m^*\) is a multiplicatively prime algebra, where \(\mathbf m^*\) denotes the set of all minimal \(\varepsilon\)-closed ideals of \(A\) that are different from \(\text{Ann}(A)\), \(\mathbf M^*\) denotes the set of all maximal \(\varepsilon\)-closed ideals of \(A\) that are different from \(A^2\), and an algebra \(A\) is strongly \(\varepsilon\)-atomic if for each nonzero \(\varepsilon\)-closed ideal \(U\) of \(A\) such that \(U \neq \text{Ann}(A)\) there exists \(B\in \mathbf m^*\) such that \(B\subset U\).