Simple Hartree-type equations lead to dynamics of a subsystem that is not completely positive in the sense accepted in mathematical literature. In the linear case this would imply that negative probabilities have to appear for some system that contains the subsystem in question. In the nonlinear case this does not happen because the mathematical definition is physically unfitting as shown on a concrete example.PACS numbers: 02.30. Sa, 03.65.Bz, Db, 11.10.Lm
I. COMPLETE POSITIVITY AND NONLINEARITYLinear maps that are positive but not completely positive (CP) [1][2][3][4][5][6][7][8] have been shown to play an essential role in characterization of degree of entanglement of correlated quantum systems [9][10][11], an important issue for quantum computation and cryptography. A possibility of an experimental verification of CP of quantum evolutions was discussed in the context of the neutral kaon decay problem in [12]. Although the notion of CP may seem somewhat abstract and technical, it has a simple physical interpretation for linear maps. We begin with a system, labelled "1", whose dynamics is given by some positive map φ t 1 (a) = a(t), φ t 1 : A → A where A is a set of bounded operators acting in a Hilbert space H 1 . To avoid technicalities we assume H 1 is finite dimensional. In linear quantum mechanics a reversible dynamics is given by φ t 1 (a) = U t aU −1 t where U t is unitary. We require positivity of φ t 1 since if a is a density matrix we want the same to be true for a(t). Now consider a density matrix ρ 1+2 (0) of some bigger system "1+2" consisting of the original one plus a system whose dimension is m and which evolves trivially (its U t = 1). The initial density matrix of "1+2" is of the formIn the context of CP maps it is more standard to use the isomorphic form [13]It follows, the argument continues, that since the dynamics on A is given by φ t 1 each of the entries evolves by a kl → φ t 1 (a kl ) and the whole density matrix is mapped intowhich in linear quantum mechanics reduces toIf ρ 1+2 (t) = φ t 1+2 ρ 1+2 (0) is to be a density matrix it should not lead to negative probabilities. Moreover one should be able to do the construction for any m. If this is the case the map φ t 1 is said to be CP. The dynamics one typically thinks of in quantum mechanics is linear and therefore the notion of CP was initially defined only for linear maps [14]. However there are many situations in quantum physics where the dynamics is nonlinear. A nonlinear evolution of observables in Heisenberg picture is typical of quantum optics and field theory. Nonlinearly evolving states appear in mean field theories (Hartree-type equations [15]), soliton theory (nonlinear Schrödinger equations), and various attempts of nonlinear generalizations of quantum mechanics. Although the latter theories do not yet correspond to any concrete physical situation they have led to some formal developments especially due to the famous "Einstein-PodolskyRosen malignancy" discussed by Gisin and others [16][17][18][19] (see Appendix C).The argu...