This thesis deals with developing Galerkin based solution strategies for several important classes of differential equations involving derivatives and integrals of various fractional orders. Fractional order calculus finds use in several areas of science and engineering. The use of fractional derivatives may arise purely from the mathematical viewpoint, as in controller design, or it may arise from the underlying physics of the material, as in the damping behavior of viscoelastic materials. The physical origins of the fractional damping motivated us to study viscoelastic behavior of disordered materials at three levels. At the first level, we review two first principles models of rubber viscoelasticity. This leads us to study, at the next two levels, two simple disordered systems. The study of these two simplified systems prompted us towards an infinite dimensional system which is mathematically equivalent to a fractional order derivative or integral. This infinite dimensional system forms the starting point for our Galerkin projection based approximation scheme. In a simplified study of disordered viscoelastic materials, we show that the networks of springs and dash-pots can lead to fractional power law relaxation if the damping coefficients of the dash-pots follow a certain type of random distribution. Similar results are obtained when we consider a more simplified model, which involves a random system coefficient matrix. Fractional order derivatives and integrals are infinite dimensional operators and non-local in time: the history of the state variable is needed to evaluate such operators. This non-local nature leads to expensive long-time computations (O(t2) computations for solution up to time t). A finite dimensional approximation of the fractional order derivative can alleviate this problem. We present one such approximation using a Galerkin projection. The original infinite dimensional system is replaced with an equivalent infinite dimensional system involving a partial differential equation (PDE). The Galerkin projection reduces the PDE to a finite system of ODEs. These ODEs can be solved cheaply (O(t) computations). The shape functions used for the Galerkin projection are important, and given attention. Calculations with both global shape functions as well as finite elements are presented. The discretization strategy is improved in a few steps until, finally, very good performance is obtained over a user-specifiable frequency range (not including zero). In particular, numerical examples are presented showing good performance for frequencies varying over more than 7 orders of magnitude. For any discretization held fixed, however, errors will be significant at sufficiently low or high frequencies. We discuss why such asymptotics may not significantly impact the engineering utility of the method. Following this, we identify eight important classes of fractional differential equations (FDEs) and fractional integrodifferential equations (FIEs), and develop separate Galerkin based solution strategies for each of them. Distinction between these classes arises from the fact that both Riemann-Liouville as well as Caputo type derivatives used in this work do not, in general, follow either the law of exponents or the commutative property. Criteria used to identify these classes include; the initial conditions used, order of the highest derivative, integer or fractional order highest derivative, single or multiterm fractional derivatives and integrals. A key feature of our approximation scheme is the development of differential algebraic equations (DAEs) when the highest order derivative is fractional or the equation involves fractional integrals only. To demonstrate the effectiveness of our approximation scheme, we compare the numerical results with analytical solutions, when available, or with suitably developed series solutions. Our approximation scheme matches analytical/series solutions very well for all classes considered.

The computation of Schrödinger equations in the semiclassical regime presents several enduring challenges due to the presence of the small semiclassical parameter. Standard approaches for solving these equations commence with spatial discretisation followed by exponentiation of the discretised Hamiltonian via exponential splittings. In this thesis we follow an alternative strategy${-}$we develop a new technique, called the symmetric Zassenhaus splitting procedure, which involves directly splitting the exponential of the undiscretised Hamiltonian. This technique allows us to design methods that are highly efficient in the semiclassical regime. Our analysis takes place in the Lie algebra generated by multiplicative operators and polynomials of the differential operator. This Lie algebra is completely characterised by Jordan polynomials in the differential operator, which constitute naturally symmetrised differential operators. Combined with the $\mathbb{Z}_2$-graded structure of this Lie algebra, the symmetry results in skew-Hermiticity of the exponents for Zassenhaus-style splittings, resulting in unitary evolution and numerical stability. The properties of commutator simplification and height reduction in these Lie algebras result in a highly effective form of $\textit{asymptotic splitting:} $exponential splittings where consecutive terms are scaled by increasing powers of the small semiclassical parameter. This leads to high accuracy methods whose costs grow quadratically with higher orders of accuracy. Time-dependent potentials are tackled by developing commutator-free Magnus expansions in our Lie algebra, which are subsequently split using the Zassenhaus algorithm. We present two approaches for developing arbitrarily high-order Magnus--Zassenhaus schemes${-}$one where the integrals are discretised using Gauss--Legendre quadrature at the outset and another where integrals are preserved throughout. These schemes feature high accuracy, allow large time steps, and the quadratic growth of their costs is found to be superior to traditional approaches such as Magnus--Lanczos methods and Yoshida splittings based on traditional Magnus expansions that feature nested commutators of matrices. An analysis of these operatorial splittings and expansions is carried out by characterising the highly oscillatory behaviour of the solution.

Submitted by Leonardo Cavalcante (leo.ocavalcante@gmail.com) on 2018-05-03T15:45:01Z No. of bitstreams: 1 Arquivototal.pdf: 9156008 bytes, checksum: 3a0383788f4458b1f695b8a3838f47bf (MD5) / Made available in DSpace on 2018-05-03T15:45:01Z (GMT). No. of bitstreams: 1 Arquivototal.pdf: 9156008 bytes, checksum: 3a0383788f4458b1f695b8a3838f47bf (MD5) Previous issue date: 2017-07-28 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / Abstract indisponível neste campo - O PDF foi entregue protegido para cópia / Resumo indisponível neste campo - O PDF foi entregue protegido para cópia

On s'intéresse à la théorie spectrale d'opérateurs semi-classiques non auto-adjoints en dimension un et plus précisément aux développements asymptotiques des valeurs propres. Ces derniers font intervenir des objets géométriques issus de la mécanique classique dans l'espace des phases complexifié et correspondent à une généralisation des conditions de quantification de Bohr-Sommerfeld au cadre non auto-adjoint. Plus précisément, dans un premier temps, on étudie le spectre de perturbations non auto-adjointes d'opérateurs pseudo-différentiels auto-adjoints en dimension un à l'aide de techniques d'analyse microlocale analytique et en corollaire, on établit que pour des perturbations PT-symétriques d'opérateurs auto-adjoints, le spectre est réel. Ensuite, on présente des conditions de quantification de Bohr-Sommerfeld pour des perturbations non auto-adjointes d'opérateurs de Berezin-Toeplitz du plan complexe auto-adjoints. Dans un second temps, on s'intéresse aux différentes quantifications du tore et plus précisément à la quantification de Berezin-Toeplitz du tore, à la quantification de Weyl classique du tore et à la quantification de Weyl complexe du tore. On établit des liens entre ces différentes quantifications notamment grâce à la transformée de Bargmann, puis à l'aide de simulations numériques, on met en évidence une conjecture sur des conditions de quantification de Bohr-Sommerfeld pour des perturbations non auto-adjointes d'opérateurs de Berezin-Toeplitz du tore auto-adjoints. / We interest ourselves in the spectral theory of non self-adjoint semi-classical operators in dimension one and in asymptotic expansions of eigenvalues. These expansions are written in terms of geometrical objects in a complex phase space coming from classical mechanics and correspond to a generalization of Bohr-Sommerfeld quantization conditions in the non self-adjoint case. First, we study non self-adjoint perturbations of self-adjoint pseudo-differential operators in dimension one by using techniques of analytic microlocal analysis. As a corollary, we establish for PT-symmetric perturbations of self-adjoint operators, that the spectrum is real. Then we show Bohr-Sommerfeld quantization conditions for non self-adjoint perturbations of self-adjoint Berezin-Toeplitz operators of the complex plane. In the second part, we look into quantizations of the torus, namely the Berezin-Toeplitz, the classical Weyl and the complex Weyl quantizations of the torus. We establish links between these different quantizations using Bargmann transform. We propose a conjecture, supported by numerical simulations, on Bohr-Sommerfeld quantization conditions for non self-adjoint perturbations of self-adjoint Berezin-Toeplitz operators of the torus.

This thesis is divided into three parts. In the first part we study mixed norm estimates for Riesz transforms associated with various differential operators. First we prove the mixed norm estimates for the Riesz transforms associated with Dunkl harmonic oscillator by means of vector valued inequalities for sequences of operators defined in terms of Laguerre function expansions. In certain cases, the result can be deduced from the corresponding result for Hermite Riesz transforms, for which we give a simple and an independent proof. The mixed norm estimates for Riesz transforms associated with other operators, namely the sub-Laplacian on Heisenberg group, special Hermite operator on C^d and Laplace-Beltrami operator on the group SU(2) are obtained using their L^pestimates and by making use of a lemma of Herz and Riviere along with an idea of Rubio de Francia. Applying these results to functions expanded in terms of spherical harmonics, we deduce certain vector valued inequalities for sequences of operators defined in terms of radial parts of the corresponding operators. In the second part, we study the chaotic behavior of the heat semigroup generated by the Dunkl-Laplacian ∆_κ on weighted L^P-spaces. In the general case, for the chaotic behavior of the Dunkl-heat semigroup on weighted L^p-spaces, we only have partial results, but in the case of the heat semigroup generated by the standard Laplacian, a complete picture of the chaotic behavior is obtained on the spaces L^p ( R^d,〖 (φ_iρ (x ))〗^2 dx) where φ_iρ the Euclidean spherical function is. The behavior is very similar to the case of the Laplace-Beltrami operator on non-compact Riemannian symmetric spaces studied by Pramanik and Sarkar. In the last part, we study mixed norm estimates for the Cesáro means associated with Dunkl-Hermite expansions on〖 R〗^d. These expansions arise when one considers the Dunkl-Hermite operator (or Dunkl harmonic oscillator)〖 H〗_κ:=-Δ_κ+|x|^2. It is shown that the desired mixed norm estimates are equivalent to vector-valued inequalities for a sequence of Cesáro means for Laguerre expansions with shifted parameter. In order to obtain the latter, we develop an argument to extend these operators for complex values of the parameters involved and apply a version of Three Lines Lemma.

The dissertation thesis investigates the modeling of the neural network activity with a focus on a multilayer forward neural network (MLP – Multi Layer Perceptron). In this often used structure of neural networks, time-varying neurons are used, along with an analogy in modeling hyperstructures of linear differential operators. Using a finite lemma and defined hyperoperation, a hyperstructure composed of neurons is defined for a given transient function. There are examined their properties with an emphasis on structures with a layout.

In the thesis we discuss binary hyperstructures of linear differential operators of the second order both in general and (inspired by models of specific time processes) in a special case of the Jacobi form. We also study binary hyperstructures constructed from distributive lattices and suggest transfer of this construction to n-ary hyperstructures. We use these hyperstructures to construct multiautomata and quasi-multiautomata. The input sets of all these automata structures are constructed so that the transfer of information for certain specific modeling time functions is facilitated. For this reason we use smooth positive functions or vectors components of which are real numbers or smooth positive functions. The above hyperstructures are state-sets of these automata structures. Finally, we investigate various types of compositions of the above multiautomata and quasi-multiautomata. In order to this we have to generalize the classical definitions of Dörfler. While some of the concepts can be transferred to the hyperstructure context rather easily, in the case of Cartesian composition the attempt to generalize it leads to some interesting results.