If y has left or right inverse x, then yxy = y, and x will suffice. If y is a unit then multiply by y inverse on the left and right, and x has to ve the inverse of y.
Conversely, let y generate H, and let the idempotent e generate H. For some x and z, xy = e, and ze = y. Write yxy = ye = zee = ze = y. Therefore, R is von neumann iff every principal left ideal is generated by an idempotent. By symmetry, the same holds for principal right ideals.
Set c = f-f*e. Note that c and e span f (and hence H), and c*e = 0.
Let b be the idempotent of R*c, so that b = xc, and b*e = xc*e = 0. Consider e+b-eb. Premultiply by e and b, and get e and b respectively. Thus e+b-eb generates H, and is contained in H, and H is principal.
Repeat this n times and every finitely generated left ideal is principal. By symmetry, the same holds for finitely generated right ideals.
Conversely, assume R is von neumann and left noetherian. Every left ideal is generated by an idempotent, making it a summand. This makes R a semisimple ring.
With R von neumann, any of the four chain conditions, left acc, right acc, left dcc, right dcc, makes R semisimple, whence all four chain conditions apply.
If S is the quotient of a von neumann ring R, pull y ∈ S back to y′ in R, find x′ in R, and map this forward to x in S, making S von neumann.
If R is the direct product of von neumann rings, select an xi for each yi, building an x satisfying yxy = y. Thus R is von neumann. The direct sum is also von neumann, though an infinite direct sum does not contain 1.
Let M be a module over a division ring K. In other words, M is a K vector space. If U is a subspace, assign it a basis, then extend the basis to all of M. The "rest" of the basis defines V, with U+V = M. Therefore M is semisimple.
Let M = Kn, and the endomorphisms of M, or the n×n matrices in K, forms a von neumann ring. Take finite direct products of such rings to build a semisimple ring. This is a round about proof that semisimple implies voneuman.
Take an infinite direct product of matrices over K, to find nonartinian voneuman rings, which are not semisimple.
Let the space be the interval [0,1] and suppose c(S) is von neumann. Let f(x) = x. Fix a nonzero x, and we're talking about the reals, whence g(x) has to be the inverse of f(x). Therefore g(x) = 1/x on (0,1], g is unbounded, and g cannot be continuous. The ring is not von neumann after all.