Search

Based on a previously introduced downscaling data assimilation algorithm, which employs a nudging term to synchronize the coarse mesh spatial scales, we construct a determining map for recovering the full trajectories from their corresponding coarse mesh spatial trajectories, and investigate its properties. This map is then used to develop a downscaling data assimilation scheme for statistical solutions of the two-dimensional Navier-Stokes equations, where the coarse mesh spatial statistics of the system is obtained from discrete spatial measurements. As a corollary, we deduce that statistical solutions for the Navier-Stokes equations are determined by their coarse mesh spatial distributions. Notably, we present our results in the context of the Navier-Stokes equations; however, the tools are general enough to be implemented for other dissipative evolution equations.

The determining modes for the two-dimensional incompressible Navier-Stokes equations (NSE) are shown to satisfy an ordinary differential equation of the form $dv/dt=F(v)$, in the Banach space, $X$, of all bounded continuous functions of the variable $s\in\mathbb{R}$ with values in certain finite-dimensional linear space. This new evolution ODE, named {\it determining form}, induces an infinite-dimensional dynamical system in the space $X$ which is noteworthy for two reasons. One is that $F$ is globally Lipschitz from $X$ into itself. The other is that the long-term dynamics of the determining form contains that of the NSE; the traveling wave solutions of the determining form, i.e., those of the form $v(t,s)=v_0(t+s)$, correspond exactly to initial data $v_0$ that are projections of solutions of the global attractor of the NSE onto the determining modes. The determining form is also shown to be dissipative; an estimate for the radius of an absorbing ball is derived in terms of the number of determining modes and the Grashof number (a dimensionless physical parameter). Finally, a unified approach is outlined for an ODE satisfied by a variety of other determining parameters such as nodal values, finite volumes, and finite elements.

The determining modes for the two-dimensional incompressible Navier-Stokes equations (NSE) are shown to satisfy an ordinary differential equation of the form $dv/dt=F(v)$, in the Banach space, $X$, of all bounded continuous functions of the variable $s\in\mathbb{R}$ with values in certain finite-dimensional linear space. This new evolution ODE, named {\it determining form}, induces an infinite-dimensional dynamical system in the space $X$ which is noteworthy for two reasons. One is that $F$ is globally Lipschitz from $X$ into itself. The other is that the long-term dynamics of the determining form contains that of the NSE; the traveling wave solutions of the determining form, i.e., those of the form $v(t,s)=v_0(t+s)$, correspond exactly to initial data $v_0$ that are projections of solutions of the global attractor of the NSE onto the determining modes. The determining form is also shown to be dissipative; an estimate for the radius of an absorbing ball is derived in terms of the number of determining modes and the Grashof number (a dimensionless physical parameter). Finally, a unified approach is outlined for an ODE satisfied by a variety of other determining parameters such as nodal values, finite volumes, and finite elements.

In this paper we present several new slightly nonlinear variants of the bipolar and the open mapping theorems in Banach spaces, which we abstracted from the recent developments in the theory of dual operator algebras. A new application of our techniques to the theory of operator algebras is also given.

The computation of turbulent flows necessitate a better understanding of turbulence and the development of algorithms and computational tools which are well adapted to the handling of large numbers of data. Turbulent flows are due to the superposition of a range of small and large eddies which interact and the study of their interaction is an important part of understanding turbulence. An inertial manifold is an exact (quasi-static) interaction law between small and large eddies. In relation with the concept of approximate inertial manifolds (AIM), Foias-Manley-Temam have shown the existence of a simple finite- dimensional manifold lying close to the attractor. By projecting the Navier- Stokes equations on this manifold we obtain a new numerical algorithm called the Nonlinear Galerkin Method. This algorithm is well-adapted to the large time solution of the Navier-Stokes equations and this has been broadly confirmed by the numerical tests which has been performed during this contract. After further tests and studies, this algorithm will soon be available for industrial implementations. (jhd)

Let $S$ and $T$ be (bounded) scalar operators on a Banach space $\scr X$ and let $C(T,S)$ be the map on $\scr B(\scr X)$, the bounded linear operators on $\scr X$, defined by $C(T,S)(X)=TX-XS$ for $X$ in $\scr B(\scr X)$. This paper was motivated by the question: to what extent does $C(T,S)$ behave like a normal operator on Hilbert space? It will be shown that $C(T,S)$ does share many of the special properties enjoyed by normal operators. For example, it is shown that the range of $C(T,S)$ meets its null space at a positive angle and that $C(T,S)$ is Hermitian if $T$ and $S$ are Hermitian. However, if $\scr X$ is a Hilbert space then $C(T,S)$ is a spectral operator if and only if the spectrum of $T$ and the spectrum of $S$ are both finite.

Let $S$ and $T$ be (bounded) scalar operators on a Banach space $\scr X$ and let $C(T,S)$ be the map on $\scr B(\scr X)$, the bounded linear operators on $\scr X$, defined by $C(T,S)(X)=TX-XS$ for $X$ in $\scr B(\scr X)$. This paper was motivated by the question: to what extent does $C(T,S)$ behave like a normal operator on Hilbert space? It will be shown that $C(T,S)$ does share many of the special properties enjoyed by normal operators. For example, it is shown that the range of $C(T,S)$ meets its null space at a positive angle and that $C(T,S)$ is Hermitian if $T$ and $S$ are Hermitian. However, if $\scr X$ is a Hilbert space then $C(T,S)$ is a spectral operator if and only if the spectrum of $T$ and the spectrum of $S$ are both finite.