The Setting
Let be an order parameter of the zero-temperature Sherrington-Kirkpatrick model and the value function of the stochastic control problem behind its Parisi formula. The gradient is the feedback of the diffusion driven by a Brownian motion and started from , and the second moment of is the function from which the contact potential is built. The questions are whether this diffusion exists and is unique, and how starts and changes in .
The Control Problem
A Brownian system on a measurable space is a probability measure on , its law, together with a standard Brownian motion under , each a random variable on .✦ Expectations are taken under .
An order parameter is a function on the time interval that is nonnegative, nondecreasing and right-continuous, and whose extension by zero, the function on equal to on and to elsewhere, is Lebesgue integrable.✦ Below, also denotes this extension.
For a time , the increment filtration of after is the natural filtration of the increments of after : is the -algebra generated by , .✦ A bounded control from time is a real process on , indexed by the time elapsed since , that is progressively measurable for and satisfies for all and .✦
The value function of is the function of a time and a position given by the supremum, over bounded controls from time , of the expectation of the control payoff, the random variable in brackets: (1)✦ a negative time being read as wherever or is evaluated. The gradient is the derivative of at ✦, and the curvature is the derivative of at .✦
The Diffusion
A controlled diffusion of driven by is a real process on , progressively measurable for the natural filtration of , such that for -almost every the path is continuous on , starts at and satisfies (2)✦ The set of controlled diffusions of driven by is written .
A bounded Lipschitz drift is a jointly measurable function of a time and a position , with a bound and a constant , such that and for all , and .✦ The Picard iterates of a bounded Lipschitz drift along a real process on are the processes with and (3)✦ The drift correction of along is the process , the limit taken at each time and sample point✦, and the drift equation solution of driven by is the process , the Picard solution of .✦
For a horizon , the clamp of a time to the strip is ✦, and the clamped drift is .✦ For it is the drift of (2). The strip drift at is the drift , which is a bounded Lipschitz drift.✦ The horizons are for natural numbers , increasing from toward .✦
The progressive version of is the process , the limit of at the last points not after of the grids of mesh .✦ The driving Brownian motion of is .✦
The localized diffusion is the drift equation solution driven by for the strip drift at , the Picard solution of (4)✦ The raw diffusion limit is , at each time and sample point✦, and the raw feedback is for and for . The diffusion process of driven by is (5)✦
The second moment of the gradient along a real process is the function ✦; along the diffusion process it is (6) and the second moment of the curvature along is .✦
The Claim
Theorem 1. Let be a Brownian system with law and Brownian motion , let be an order parameter, let be the diffusion process of driven by , and let be the second moment of the gradient along . Then (7)✦ and for every and -almost every , (8)✦ Moreover (9)✦ and is differentiable on , from the right at , with (10)✦
In the paper, the result reads: the diffusion has a unique strong solution on ; on every strip with , ; in particular is on with and .
What Is Checked
The theorem is stated over a Brownian system and an order parameter : the value function of (1) with its gradient and curvature, the controlled diffusions of (2), the diffusion process (5) built from the strip drifts at the horizons by Picard iteration and a limit in , and the second moments of the gradient and of the curvature along it.
The library also shows that every horizon lies in .✦