Probabilistic solutions of the so-called Schrödinger boundary data problem provide for a unique Markovian interpolation between any two strictly positive probability densities designed to form the input–output statistics data for the process taking place in a finite-time interval. The key issue is to select the jointly continuous in all variables positive Feynman–Kac kernel, appropriate for the phenomenological (physical) situation. We extend the existing formulations of the problem to cases when the kernel is not a fundamental solution of a parabolic equation, and prove the existence of a continuous Markovian interpolation in this case. Next, we analyze the compatibility of this stochastic evolution with the original parabolic dynamics, which is assumed to be governed by the temporally adjoint pair of (parabolic) partial differential equations, and prove that the pertinent random motion is a diffusion process. In particular, in conjunction with Born’s statistical interpretation postulate in quantum theory, we consider stochastic processes which are compatible with the Schrödinger picture quantum evolution.
Abstract: Probablistic solutions of the so called Schrödinger boundary data problem provide for a unique Markovian interpolation between any two strictly positive probability densities designed to form the input-output statistics data for a certain dynamical process taking place in a finite-time interval. The key problem is to select the jointly continuous in all variables positive semigroup kernel, appropriate for the phenomenological (physical) situation. The issue of deriving a microscopic dynamics from the input-output statistics data (analyzed in terms of densities) was addressed, as the Schrödinger problem of a probabilistic interpolation, in a number of publications [1] – [6]. We shall consider Markovian propagation scenarios so remaining within the well established framework, where for any two Borel sets A, B ⊂ R on which the strictly positive boundary densities ρ(x, 0) and ρ(x, T) are defined, the transition probability m(A, B) from the set A to the set B in the time interval T> 0 has a bi-variate density given in a specific factorized form: m(x, y) = f(x)k(x, 0, y, T)g(y), with marginals: ∫
We invert the well-developed strategy of studying dynamics in terms of probability densities and investigate the problem of the most likely microscopic propagation scenario, which is consistent with the given priori (possibly phenomenological) input-output statistics data for the process taking place in a finite-time interval. A solution of this so-called Schr\"odinger problem is known to provide an adequate probabilistic framework for the measure preserving dynamics which is Markovian. We pay particular attention to the subclass of nonstationary solutions, determined by unitary-wave-packet evolution (Schr\"odinger wave mechanics). The existence of the pertinent Markovian diffusion is known on general grounds, but no explicit demonstration (through detailed computational arguments) until now was available even in the simplest cases. We give a definitive probabilistic description of the free quantum dynamics as a stochastic process solving Schr\"odinger's interpolation problem. The Markov diffusion arised as a particular case singled out by a suitable Feynman-Kac semigroup.
Relaxation to equilibrium of a drifted Brownian motion is quantified by a transition probability density function, whose main (multiplicative) entry is an inferred Feynman-Kac kernel of the Schrödinger semigroup operator. Although seemingly devoid of a natural probabilistic significance (except for its explicit path integral definition), the pertinent kernel relaxes to equilibrium as well. The implicit Feynman-Kac potential
V(x), continuous, confining and bounded from below, may take negative values. If positive,
V(x) can be interpreted as the killing rate of the decaying diffusion process. In case of relaxing F-K kernels the killing effects are tamed (often overcompensated). The taming inavoidably appears in conjunction with the existence of the negativity subdomains of
V(x) in
R. If locally
V(x)<0, its sign inversion
−V(x) can be interpreted as the branching (cloning, alternatively bifurcation) rate in the course of the other wise free random motion. The arising killed diffusion processes with branching, we interpret as the possible path-wise background of tamed (relaxing) Feynman-Kac diffusions. We present acomputer-assisted path-wise arguments, towards a consistency of the killing/branching taming scenario, for a number of nonlinear model systems in one space dimension. Special attention is paid to Feynman-Kac potential shapes in the double well form, where an analytic access to eigenvalues and eigenfunctions is scarce. Throughout the paper the dynamics refers to the positive real time. Since the Newton-type equations of motion for admissible classical trajectories have a Euclidean form (due to the sign inverted force term), we give a brief resume of a couple of their explicit solutions, without recourse to the Euclidean time intuitions, and the instanton lore of related quantum model systems.
In Nelson [10], it is shown that for each solution of a Schr\"odinger equation
ϕ(x,t), x∈Rd, t≧0 , with
∥ϕ∥=1 , there exists a diffusion process which has ... probability density at time
t given byMore precisely the following is shown in [10]: suppose that we are given a real valued function
V(x), x∈Rd , and a complex valued function
ϕ0(x), x∈Rd , with
∥ϕ0∥=1 , and consider the Schrodinger equation Then, under the assumption that
ϕ and
b~ are sufficiently regular, there exists a diffusion process
{Xt}, t≧0 , with initial density
∣ϕ0(x)∣2 and generator
(1/2)Δ+b~(x,t)⋅∇ such that the probability density of this diffusion at time
t≧0 is
∣ϕ(x,t)∣2 .In addition, in [10] it is also shown that the diffusion process
{Xt}, t≧0 , solves the second order stochastic differential equation
(1/2)(D∗D+DD∗)Xt=−∇V(Xt) , where
D and
D∗ are Nelson's forward and backward stochastic derivatives.Generally sPeaking, the functions
ϕ and
b~ may be singular, and the existence of the corresponding diffusion process must be studied carefully.Carlen
[1,2] , Meyer and Zheng [8] and Nagasawa [9] considered rigorously the construction of a diffusion process with a given initial density and a generator.
ln order to construct a Markovian propagator for the diffusion process, Carlen
[1,2] used partial differential equation methods.Meyer and Zheng [8] restricted themselves to the case when
ϕ(x,t)=ϕ0(x), ∀t≧0 and considered the construction of the diffusion process through the relationships between Markov processes and Dirichlet forms.We may consult Nagasawa [9] about recent developments of this area.In Carlen [1], it is assumed that
V(x), x∈Rd , is a Rellich class potential and
∥∇ϕ0∥2<∞ , and shown that there exists a diffusion process
{Xt}, t≧0 , which has the probability density
∣ϕ(x,t)∣2 at time
t≧0 and admits the representation