Final 3
Final 3
1 (10 points) Apply the third-order Taylor method with h = 1/2 to approximate the initial
value problem:
x0 + 2x2 = t2 − 1, x(0) = 0
Find the approximation value of x(1).
18 9 2 6
wn+1 − wn + wn−1 − wn−2 = hf (tn+1 , wn+1 ).
11 11 11 11
What is the order of the method? Is the method stable? Please give the detail of your
answers.
3 Consider the following differential equation with initial conditions,
(1 + t)x00 + t = 1 + y 0 ,
x0 y = tx2 − y 00 ,
x(0) = 1, x0 (0) = 0, y(0) = 0, y 0 (0) = 1.
i (5 points) Convert the above equations to a system of first-order differential equation.
ii (5 points) Advance the solution of the equation obtained in i through N = 2 time
steps using Euler’s method with time step h = 1/2. What are the numerical solutions
to approximate x0 (1) and y(1)?
4 Assume f (t, y) is continuous and satisfies a Lipschitz condition in y on the set R2 with
Lipschitz constant L. Consider to solve the following ODE
y 0 (t) = f (t, y).
by the one-step method:
wi+1 = wi + hφ(f, ti , wi , h),
where φ is continuous with respect to h.
i (5 points) Show that the method is consistent if φ(f, ti , yi , 0) = f (ti , yi ).
ii (10 points) Show that if φ satisfies the Lipschitz condition in y:
|φ(f, t, u, 0) − φ(f, t, v, 0)| ≤ Lφ |u − v|,
then the method is stable. That is, show that
|ui − vi | ≤ eihLφ |u0 − v0 |,
where ui and vi are the numerical solutions obtained by the one-step scheme with
initial data u0 and v0 respectively.
iii (10 points) Show that the Heun method:
h
wi+1 = wi + [f (ti , wi ) + f (ti + h, wi + hf (ti , wi )]
2
is consistent and stable.
1
5 (10 points) Suppose h > 0 and t0 = 0, t1 = 2h, t2 = 3h. Mimic Adams-Moulton methods
to derive a two-step method for y 0 = f (t, y). That is, find constants b0 , b1 , b2 , and local
truncation error τ2 such that
i (10 points) Prove that the degree of precision of In (f ) is less than 2n + 2 for any
choice of quadrature weights wi and points xi .
ii (10 points) Prove that if the degree of precision of In (f ) is 2n + 1, then x0 , x1 , . . . , xn
are the roots of φn+1 (x), the orthogonal polynomial of degree n + 1.
i (5 points) Show that the sequence xn converges to 1 and the order of convergence is
linear. You must show this directly without using any theorem.
ii (5 points) Modify the Newton’s method to regain quadratic convergence. Show your
modification indeed is quadratic directly.
i (5 points)
n−1
Y
P0,1,2,...,n (x) − P0,1,2,...,n−1 (x) = an (x − xi )
i=0
10 (10 points) Consider the linear system Ax = b, where A is an n × n matrix and x, b are
n×m matrices. Write a pseudo code for Gaussian Elimination without row interchanging.
Compute the operation counts for reducing the augmented matrix to an upper triangular
matrix.