Preface

---

This is a tutorial made solely for the purpose of education and it was designed for students taking Applied Math 0330. It is primarily for students who have very little experience or have never used Mathematica before and would like to learn more of the basics for this computer algebra system. As a friendly reminder, don't forget to clear variables in use and/or the kernel.

Finally, the commands in this tutorial are all written in bold black font, while Mathematica output is in regular fonts. This means that you can copy and paste all comamnds into Mathematica, change the parameters and run them. You, as the user, are free to use the scripts to your needs for learning how to use the Mathematica program, and have the right to distribute this tutorial and refer to this tutorial as long as this tutorial is accredited appropriately.

Multistep Methods

---

The methods of Euler, Heun, Runge--Kutta, and Taylor are called single-step methods because they use only the information from one previous point to compute the successive point; that is, only the initial point (x₀, y₀) is used to compute (x₁, y₁) and, in general, y_k is needed to compute y_k+1. After several points have been found, it is feasible to use several prior points in the calculation. methods that use information at more than the last mesh point are referred to as miltistep methods. In this section, we discuss two types of multistep methods: Adams methods and backward differentiation methods. For simplicity, we assume throughout our exposition that the step length h is constant.

All multistep methods are not self-starting because some initial points are required to be determined. A desirable feature of a multistep method is that the local truncation error can eb determined and a correction term can be included, which improved the accuracy of the answer at each step. Also, it is possible to determine whether the step size is small enough to obtain an accurate value for y_n+1, yet large enough so that unnecessary and time-consuming calculations are eliminated. Using the combination of a predictor and corrector requires only two slope function evaluations compared to six evaluations required by RKF45. The predictor-corrector forms are among the most efficient known integration methods in terms of speed and accuracy. As a class of integration methods, the multistep sets are among the best, but individually as a predictor-corrector set, the choice for the best method varies depending on the application.

 

I. Adams Method

---

To integrate the initial value problem \( y' = f(x,y) , \quad y(x_0 ) = y_0 , \) on the mesh interval \( \left[ x_n , x_{n+1} \right] , \) we rewrite the problem in integral form: The basic idea of an Adams method is to approximate f(t,y(t)) by a polynomial P_k(t) of degree k and to use the polynomial to evaluate the integral on the right side of the above integral equation. John Couch Adams (1819--1892), an English mathematician and astronomer, is most famous as codiscoverer, with Joseph Leverrier, of the planet Neptume in 1846. He was associated with Cambridge University for most of his life, as student (1839--1843), fellow, Lowdean Professor, and diector of the Observatory. Adams was extremely skilled at computation; his procedure for numerical integration of differential equations appeared in 1883 in a book with Francis Bashforth on capilalry action.

Francis Bashforth (1819--1912), English mathematician and Anglican priest, was a classmate of J.C. Adams at Cambridge. He was particularly interested in ballistic and invented the Bashforth chronograph for measuring the velocity of artillery projectiles.

The polynomial P_k(t) of degree k contains k+1 coefficients that are determined from previously calculated data points. For example, suppose that we wish to use a first degree polynomial \( P_1 (t) =at+b . \) Then we need only the two data points \( (x_n , y_n ) \quad\mbox{and} \quad (x_{n-1} , y_{n-1}) . \) For P₁ to be an approximation to the slope function, we require that

Solving for a and b, we obtain Replacing the slope function by the polynomial \( P_1 (t) =at+b , \) and evaluating integral, we obtain The above explicit formula for \( y_{n+1} \) in terms of y_n and y_n-1 is called Adams--Bashforth. It has a local truncation error proportional to h³. However, this value p_n+1 is used as a predicted value of y_n+1 that approximates the actual solution \( y = \phi (x) \) at the mesh point x_n+1. This value p_n+1 is utilized in Adams--Moulton correction formula, which is our next topic. Forest Ray Moulton (1872--1952), an American astronomer and administrator of science, was for many years professor of astronomy at the University of Chicago. During World War I, he was in charge of the Ballistics Branch of the U.S. Army at Aberdeen (MD) Proving Ground. In the course of calculating ballistics trajectories he devised substantial improvements in the Adams formula.

Actually, the predictor formulas are based on Newton's backward difference interpolation formula:

where \( \nabla y_n = y_n - y_{n-1} \) is the backward finite difference and \( m^{\underline{n}} = m (m-1) \cdots (m-n+1) \) is the n-th falling factorial of m. Since the derivative of y is known, we obtain We can derive Nystrom midpoint predictor formula: followed by the corrector step: It is quite interesting to see that although the Euler and Nystrom methods have the widest range of real negative stability, their results are not quite as accurae as the results of the Hamming and the fourth order Adams, even the third order Adams methods.

 

II. Backward Differentiation Formula

---

Another type of multistep method uses a polynomial P_k(t) of degree k to approximate the actual solution \( y = \phi (x) \) of the considered initial value problem rather than its derivative \( y' = \phi' (x) , \) as in the Adams method. We then differentiate P_k(t) and set \( P'_k (x_{n+1}) \) equal to \( f \left( x_{n+1} , y_{n+1} \right) \) to obtain an implicit for mual for y_n+1. these are called backward differentiation formulas. These methods became widely used in the 1970s because of the work of C.William Gear (born in 1935, London, UK) on so called stiff differential equations, whose solutions are very difficult to approximate by discussed so far methods. A British-American mathematician C.W. Gear is well-known for his contributions in numerical analysis and computer science.

The simplest case uses a first degree polynomial \( P_1 (t) =at+b . \) The coefficients are chosen to match the computed values of the solution y_n and p_n+1

, from predictor stage. Hence, a and b must satisfy Since \( P'_1 (t) =a , \) the requirement that is just \( a = f \left( x_{n+1} , p_{n+1} \right) . \) Another expression for a comes from subtructing the first equation from the second, which gives Using these values of a, we get the first order backward differentiating formula which is just the backward Euler formula. This yields the Nystrom predictor-corrector formulas:

By using higher order polynomials and corresponding more data points, we can obtain backward differentiation formulas of any order. The second order Adams-Moulton formula is