We describe and analyse a quasi-Trefftz discontinuous Galerkin method for solving boundary value problems for the homogeneous diffusion-advection-reaction equation with piecewise-smooth coefficients. The discontinuous Galerkin (DG) weak formulation is derived including an interior penalty parameter and using the classical upwind numerical fluxes. We compare three different formulations which arise from different choices of the symmetrization parameter, including the Symmetric Interior Penalty Galerkin (SIPG) one. DG methods may require a higher number of degrees of freedom compared to continuous ones, especially in the case of high-order schemes; combining them with quasi-Trefftz functions allows for a reduction in the number of degrees of freedom. Trefftz schemes are high-order Galerkin methods whose discrete functions satisfy exactly the underlying partial differential equation (PDE) in each mesh element (for example, harmonic polynomials for the Laplace equation). Since a family of local exact solutions is needed, Trefftz basis functions can be easily computed for many PDEs that are linear, homogeneous and with piecewise-constant coefficients. However, if the equation has varying coefficients, in general, exact solutions are unavailable, hence the construction of discrete Trefftz spaces is not possible. The quasi-Trefftz methods have been introduced to overcome this limitation. They rely on discrete spaces of functions that are not exact solutions but elementwise ``approximate solutions'' of the PDE in the sense of Talyor polynomials: the Taylor polynomial of a given order of their image under the partial differential operator vanishes at the barycentre of a mesh element. A space-time quasi-Trefftz DG method has recently been studied, extending to the case of smoothly varying coefficients the space-time Trefftz DG scheme for the acoustic wave equation. Since it has shown excellent results, we propose a related method that can be applied to second-order diffusion-advection-reaction elliptic equation. We choose polynomial quasi-Trefftz basis functions, whose coefficients can be computed with a simple algorithm, which is initialized assigning a sort of Cauchy conditions and is based on the Taylor expansion of the coefficients of the PDE. The main advantage of Trefftz and quasi-Trefftz schemes over more classical ones is the higher accuracy for comparable numbers of degrees of freedom. We prove that the dimension of the quasi-Trefftz space is smaller than the dimension of the polynomial space of the same degree and that yields the same optimal convergence rates as the full polynomial space. The quasi-Trefftz DG method is well-posed, consistent and stable and we prove its high-order convergence. We present some numerical experiments in two dimensions that show excellent properties in terms of approximation and convergence rate.
In questo elaborato descriviamo e analizziamo un metodo quasi-Trefftz discontinuous Galerkin per risolvere problemi al contorno per l'equazione di diffusione-trasporto-reazione omogenea con coefficienti lisci a tratti. La formulazione debole discontinuous Galerkin (DG) include i classici flussi numerici upwind, un parametro di interior penalty e un parametro di simmetrizzazione che porta a tre diverse formulazioni, tra cui quella del metodo Symmetric Interior Penalty Galerkin (SIPG). I metodi DG spesso richiedono un maggior numero di gradi di libertà rispetto a quelli continui, soprattutto nel caso di metodi di alto ordine; combinandoli con funzioni quasi-Trefftz otteniamo una riduzione del numero di gradi di libertà. I metodi Trefftz sono metodi di alto ordine di tipo Galerkin le cui funzioni test e trial sono soluzioni esatte dell'equazione alle derivate parziali (EDP) in ogni elemento della mesh (ad esempio, polinomi armonici per l'equazione di Laplace). Poiché sono necessarie soluzioni locali esatte, le funzioni Treffz di base possono essere facilmente calcolate per molte EDP che sono lineari, omogenee e con coefficienti costanti a tratti. Tuttavia, se l'equazione ha coefficienti variabili, in generale, le soluzioni esatte non sono disponibili, quindi la costruzione di spazi discreti Trefftz non è possibile. I metodi quasi-Trefftz sono stati introdotti per superare questa limitazione. Si basano su spazi discreti di funzioni che non sono soluzioni esatte ma ``soluzioni approssimate'' dell'EDP in ogni elemento. Più precisamente, il polinomio di Taylor di un dato ordine della loro immagine tramite l'operatore differenziale parziale si annulla nel baricentro di un elemento della mesh. Un metodo spazio-tempo quasi-Trefftz DG è stato recentemente studiato, estendendo al caso di coefficienti variabili il metodo spazio-tempo Trefftz DG per l'equazione delle onde acustiche. Poiché ha mostrato eccellenti risultati, proponiamo un metodo correlato per l'equazione ellittica del secondo-ordine di diffusione-trasporto-reazione. Scegliamo delle funzioni di base polinomiali quasi-Trefftz, i cui coefficienti possono essere calcolati con un semplice algoritmo, che viene inizializzato assegnando una sorta di condizioni di Cauchy e si basa sull'espansione di Taylor dei coefficienti dell'EDP. Il principale vantaggio dei metodi Trefftz e quasi-Trefftz rispetto a quelli più classici è la maggiore accuratezza per un numero comparabile di gradi di libertà. Dimostriamo che la dimensione dello spazio quasi-Trefftz è inferiore alla dimensione dello spazio polinomiale dello stesso grado e che si ottengono gli stessi ordini di convergenza ottimali dello spazio polinomiale completo. Il metodo quasi-Trefftz DG è ben posto, consistente e stabile e dimostriamo la convergenza di ordine alto. Presentiamo alcuni esperimenti numerici in due dimensioni che mostrano ottime proprietà in termini di approssimazione e ordini di convergenza.
A quasi-Trefftz discontinuous Galerkin method for the homogeneous diffusion-advection-reaction equation with piecewise-smooth coefficients
PERINATI, CHIARA
2022/2023
Abstract
We describe and analyse a quasi-Trefftz discontinuous Galerkin method for solving boundary value problems for the homogeneous diffusion-advection-reaction equation with piecewise-smooth coefficients. The discontinuous Galerkin (DG) weak formulation is derived including an interior penalty parameter and using the classical upwind numerical fluxes. We compare three different formulations which arise from different choices of the symmetrization parameter, including the Symmetric Interior Penalty Galerkin (SIPG) one. DG methods may require a higher number of degrees of freedom compared to continuous ones, especially in the case of high-order schemes; combining them with quasi-Trefftz functions allows for a reduction in the number of degrees of freedom. Trefftz schemes are high-order Galerkin methods whose discrete functions satisfy exactly the underlying partial differential equation (PDE) in each mesh element (for example, harmonic polynomials for the Laplace equation). Since a family of local exact solutions is needed, Trefftz basis functions can be easily computed for many PDEs that are linear, homogeneous and with piecewise-constant coefficients. However, if the equation has varying coefficients, in general, exact solutions are unavailable, hence the construction of discrete Trefftz spaces is not possible. The quasi-Trefftz methods have been introduced to overcome this limitation. They rely on discrete spaces of functions that are not exact solutions but elementwise ``approximate solutions'' of the PDE in the sense of Talyor polynomials: the Taylor polynomial of a given order of their image under the partial differential operator vanishes at the barycentre of a mesh element. A space-time quasi-Trefftz DG method has recently been studied, extending to the case of smoothly varying coefficients the space-time Trefftz DG scheme for the acoustic wave equation. Since it has shown excellent results, we propose a related method that can be applied to second-order diffusion-advection-reaction elliptic equation. We choose polynomial quasi-Trefftz basis functions, whose coefficients can be computed with a simple algorithm, which is initialized assigning a sort of Cauchy conditions and is based on the Taylor expansion of the coefficients of the PDE. The main advantage of Trefftz and quasi-Trefftz schemes over more classical ones is the higher accuracy for comparable numbers of degrees of freedom. We prove that the dimension of the quasi-Trefftz space is smaller than the dimension of the polynomial space of the same degree and that yields the same optimal convergence rates as the full polynomial space. The quasi-Trefftz DG method is well-posed, consistent and stable and we prove its high-order convergence. We present some numerical experiments in two dimensions that show excellent properties in terms of approximation and convergence rate.È consentito all'utente scaricare e condividere i documenti disponibili a testo pieno in UNITESI UNIPV nel rispetto della licenza Creative Commons del tipo CC BY NC ND.
Per maggiori informazioni e per verifiche sull'eventuale disponibilità del file scrivere a: unitesi@unipv.it.
https://hdl.handle.net/20.500.14239/16361