Q2. What is the main source of inaccuracy in the conventional finite element method when solving?
Numerical dispersion is the main source of inaccuracy in the conventional finite element method when solving Helmholtz problems at high frequencies.
Q3. What is the general argument for the use of wave-based methods?
The general argument for the use of wave-based methods is that plane waves are canonical solutions of the underlying equations and are therefore better suited than polynomials to construct an approximation basis.
Q4. What is the way to improve the FEM and wave-based DGM solutions?
The use of local h-refinement is a valid option to improve both the FEM and the wave-based DGM solutions and will be used in the next section.
Q5. How many tests have been performed without significant change in the results?
2. A large number of tests have also been performed using quadrature orders up to 2p + 10, without inducing any significant change in the results, thereby indicating that the resulting integration error can be considered negligible.
Q6. What is the central argument for using a basis of plane waves?
The central argument for using a basis of plane waves is that it follows directly from the governing equations and thus it allows embedding key features of the underlying physics into the numerical method.
Q7. What is the first step in the study of the underlying system of equations?
The convergence with respect to p- and h-refinements is firstly considered, followed by the study of the conditioning properties of the underlying system of equations.
Q8. What is the continuity of the normal flux across the edge between the elements e and e?
The continuity of the normal flux across the edge between the elements e and e′ is directly enforced by writing:Feue = −Fe′ue′ = fe,e′ (ue,ue′ ) , (17)where fe,e′ is the so-called numerical flux which is discussed below.
Q9. What is the generalized discretization error for the L2-norm?
It is shown that for smooth solutions, the global discretization error measured in the L2-norm behaves asymptotically like h(Nw−1)/2−1.
Q10. How many different benchmark problems are used to assess the performance of high-order methods?
The performance of high-order methods are known to be problem dependent [12], therefore a total of four different benchmark problems are used to assess the ability of these methods to tackle a large variety of problems.
Q11. What is the numerical method for constant coefficient matrices?
The numerical method is described here for constant coefficient matrices but it could be extended to the case where the matrices are functions of x and y [18] by approximating them as piecewise constant functions.
Q12. What is the approximate solution uh and the associated test function vh?
The approximate solution uh and the associated test function vh are constructed using the classical H1-conforming hierarchical polynomial shape functions.