“…For any function h of the system state x, we write δ xy h = h(x) − h(y). Thus δ xy s = s(x) − s(y), whereas (δS) xy is the entropy variation of the external systems: it is supposed to depend only on x and y, as discussed below, but in general it is not the variation from y to x of any function of the system state alone [1,2]. If, under special conditions, it happens that there is a function S tot (x) such that (δS tot ) xy = S tot (x) − S tot (y) for all (x, y), the only stationary distribution p 0 (x) of s is p 0 (x) ∝ S tot (x), and detailed balance [2,3,4,5,6] holds R xy p 0 (x) = R yx p 0 (y) ;…”