We prove that a ring R is an $n \times n$ matrix ring (that is, $R \cong \mathbb {M}_n(S)$ for some ring S) if and only if there exists a (von Neumann) regular element x in R such that $l_R(x) = R{x^{n-1}}$ . As applications, we prove some new results, strengthen some known results and provide easier proofs of other results. For instance, we prove that if a ring R has elements x and y such that $x^n = 0$ , $Rx+Ry = R$ and $Ry \cap l_{R}(x^{n-1}) = 0$ , then R is an $n \times n$ matrix ring. This improves a result of Fuchs [‘A characterisation result for matrix rings’, Bull. Aust. Math. Soc. 43 (1991), 265–267] where it is proved assuming further that the element y is nilpotent of index two and $x+y$ is a unit. For an ideal I of a ring R, we prove that the ring $(\begin {smallmatrix} R & I \\ R & R \end {smallmatrix})$ is a $2 \times 2$ matrix ring if and only if $R/I$ is so.

We use George Bergman's recent normal form for universally adjoining an inner inverse to show that, for general rings, a nilpotent regular element $x$ need not be unit-regular. This contrasts sharply with the situation for nilpotent regular elements in exchange rings (a large class of rings), and for general rings when all powers of the nilpotent element $x$ are regular.

Let $\mathfrak{g}=\mbox{Lie}(G)$ be the Lie algebra of a simple algebraic group $G$ over an algebraically closed field of characteristic $0$. Let $e$ be a nilpotent element of $\mathfrak{g}$ and let $\mathfrak{g}_e=\mbox{Lie}(G_e)$ where $G_e$ stands for the stabiliser of $e$ in $G$. For $\mathfrak{g}$ classical, we give an explicit combinatorial formula for the codimension of $[\mathfrak{g}_e,\mathfrak{g}_e]$ in $\mathfrak{g}_e$ and use it to determine those $e\in \mathfrak{g}$ for which the largest commutative quotient $U(\mathfrak{g},e)^{\mbox{ab}}$ of the finite $W$-algebra $U(\mathfrak{g},e)$ is isomorphic to a polynomial algebra. It turns out that this happens if and only if $e$ lies in a unique sheet of $\mathfrak{g}$. The nilpotent elements with this property are called non-singular in the paper. Confirming a recent conjecture of Izosimov, we prove that a nilpotent element $e\in \mathfrak{g}$ is non-singular if and only if the maximal dimension of the geometric quotients $\mathcal{S}/G$, where $\mathcal{S}$ is a sheet of $\mathfrak{g}$ containing $e$, coincides with the codimension of $[\mathfrak{g}_e,\mathfrak{g}_e]$ in $\mathfrak{g}_e$ and describe all non-singular nilpotent elements in terms of partitions. We also show that for any nilpotent element $e$ in a classical Lie algebra $\mathfrak{g}$ the closed subset of Specm  $U(\mathfrak{g},e)^{\mbox{ab}}$ consisting of all points fixed by the natural action of the component group of $G_e$ is isomorphic to an affine space. Analogues of these results for exceptional Lie algebras are also obtained and applications to the theory of primitive ideals are given.