The inclusion of transverse effects in designing characteristic boundary conditions for the Euler and NavierStokes equations was discussed by different authors and has proved to give some improvement in the perspective of reducing numerical perturbations generated at open boundaries. Based on the most general characteristic formulation using nonorthogonal and rotated reference frames, an analysis of the different terms involved in such an approach is carried out in the present work. To achieve the best performance for the numerical behavior of the boundary condition, it is then shown that different transverse terms need to be treated differently when included in the definition of the incoming wave amplitude variations. The analysis is supported by a series of numerical tests involving an inviscid vortex convected at an angle or a purely acoustic wave propagating radially. It is concluded that the optimal behavior in terms of acoustic reflection by outgoing vorticity can be achieved by minimizing the contribution of the transverse terms related to the material derivatives along the bicharacteristics, either by properly relaxing them or by selecting a characteristic direction which is not necessarily orthogonal to the boundary.
Nomenclature= specific total energy (thermal kinetic), m 2 =s 2 E glb ; t = normalized global error for quantity F i = i-th component of the conservative flux vector-tensor J = Jacobian of the transformation between global and local generalized components j = 1 p , imaginary constant K = pressure relaxation coefficient, m 2 =kg L = reference length scale, m L = vector of the characteristic wave amplitude variations L t = generalized transverse propagation terms l, k = wavenumbers, 1=m M, M 1 = Mach number, reference Mach number n, t, k = basis of the local nonorthogonal reference frame p, p 1 = pressure and reference outlet pressure, respectively, Pa P = Jacobian of the transformation between conservative and primitive variables q = u i u i , m 2 =s 2 R v = vortex radius, m R = matrix of reflection coefficients S 1 , S = matrix of the right eigenvectors of A 1 and A , respectively s = entropy mode connected characteristic variable T = vector of transverse terms T = generalized transverse coupling terms t = time, s U, U = vectors of conservative and primitive variables, respectively u i , u 1 = i-th component of the velocity vector and reference velocity, respectively, m=s W = characteristic variables vector x i = i-th component of the position vector, m x 0 , y 0 = initial location of the vortex center, m ,= a dimensional pressure relaxation coefficients l , t = transverse relaxation coefficients = specific heat capacities ratio ij = Kronecker delta function " ijk = Levi-Civita symbol = base flow angle , = diagonal matrix of the eigenvalues of A 1 and A , respectively , , = local nonorthogonal reciprocal basis = eigenvalue of the nonconservative Jacobian, m=s, or reciprocal phase velocity, s=m i = i = j j 1=2 , , = coordinates in the local nonorthogonal reference frame (m) , 1 = density and ...