NUMERICAL ANALYSIS PRELIMINARY EXAM August 2006 1. Consider the ODE boundary value problem d2 u = f (x), 0 < x < 1, dx2 du u(0) + (0) = 0, u(1) = 0. dx − Recall that the variational, or weak, formulation for this problem is to find u ∈ V for which a(u, v) = L(v), for all v ∈ V. a. Specify the appropriate vector space V and derive the functionals a(u, v) and L(v) in the variational form. b. Verify that if u is a weak solution, and u ∈ C 2 [0, 1], then u solves the boundary value problem. c. Explain how to implement the Ritz-Galerkin method for this problem. 2. Consider the parameterized family of time-marching methods un+1 − unj = j ∆t n+1 [α(unj−1 − 2unj + unj+1 ) + (1 − α)(un+1 + un+1 j−1 − 2uj j+1 )] h2 for the PDE ut = uxx , 0 < x < 1, t > 0, with homogeneous Dirichlet boundary conditions, where unj ≈ u(jh, n∆t), and α is real-valued. Determine the ranges of parameters α for which the time-marching method is unconditionally stable and for which it is always unstable. For remaining values of α (where the method is conditionally stable) find conditions on h, ∆t, and α which guarantee stability. 3. Consider the difference method n uni−1,j + uni+1,j + uni,j−1 + uni,j+1 − 4uni,j un+1 i,j − ui,j = ∆t h2 for the 2-d heat equation, ut = uxx + uyy . a. Perform a Von Neumann stability analysis to determine the range of values of ∆t and h for which this method is stable. b. Provide a general definition of local truncation error for evolution equations, and then compute the local truncation error for this method. 1 4. Derive the Gauss-Legendre quadrature formula, Z 1 −1 f (x) dx ≈ n X f (xk ) wk , k=1 which is exact for all polynomials of degree ≤ 2n − 1, i.e., explain how to obtain the xk ’s and wk ’s, and prove that your formula does indeed have the desired exactness. You may use the Hermite polynomial interpolation formula: If f ∈ C 2n [a, b] and {xk }nk=1 are distinct points in [a, b], then f (x) = n X k=1 f (xk )hk (x) + n X f 0 (xk )h̄k (x) + k=1 n f (2n) (ξ) Y [ (x − xk )]2 , (2n − 1)! k=1 where for k = 1, . . . , n, hk (x) = [1 − 2(x − xk )`0k (xk )][`k (x)]2 , h̄k (x) = (x − xk )[`k (x)]2 , n Y (x − xi ) . `k (x) = i=1,i6=k (xk − xi ) 5. The leapfrog method for the 1-d linear avection equation ut + aux = 0 is un − uni−1 − uin−1 un+1 i + a i+1 = 0. 2∆t 2h Analyze the stability and accuracy of this method. 2