2 Numerical Methods for Initial Value Problems

Size: px
Start display at page:

Download "2 Numerical Methods for Initial Value Problems"

Transcription

1 Numerical Analysis of Differential Equations 44 2 Numerical Methods for Initial Value Problems Contents 2.1 Some Simple Methods 2.2 One-Step Methods Definition and Properties 2.3 Runge-Kutta-Methods 2.4 Linear Multi-Step Methods 2.5 Convergence of One-Step Methods 2.6 Zero-Stability of Linear Multistep Methods 2.7 Stepsize Control 2 Numerical Methods for Initial Value Problems TU Bergakademie Freiberg, SS 2012

2 Numerical Analysis of Differential Equations Some Simple Methods Consider an IVP y = f (t, y), y(t 0 ) = y 0. (IVP) Under the assumptions of Theorem 1.1, (IVP) possesses a unique solution y = y(t), say, on the interval I. We will approximate y(t) for t [t 0, t end ] I using the Euler method a (sometimes called the explicit Euler method), which can be regarded as a prototype for all numerical schemes for solving IVPs. As all methods we shall encounter in this course, Euler s method is based on discretization, i.e., approximating the solution y on only a discrete subset {t n : n = 0, 1,..., N} of the interval [t 0, t end ]. a LEONHARD EULER ( ) 2.1 Some Simple Methods TU Bergakademie Freiberg, SS 2012

3 Numerical Analysis of Differential Equations 46 We fix N N, set h := (t end t 0 )/N and define t n := t 0 +nh, n = 0, 1,..., N, i.e. t 0 < t 1 < t N 1 < t N = t end. The number h > 0 is called the stepsize, which we choose to be constant for convenience only. One step of Euler s method requires an approximation y n to y(t n ), n = 0, 1,..., N 1. Noting that, under the assumption y C 2 [t 0, t end ], y(t n+1 ) = y(t n + h) = y(t n ) + hy (t n ) h2 y (ξ) we obtain the y(t n ) + hy (t n ) = y(t n ) + hf (t n, y(t n )) y n + hf (t n, y n ), (Explicit) Euler scheme: y 0 given, y n+1 := y n + h f (t n, y n ) n = 0, 1,..., N Some Simple Methods TU Bergakademie Freiberg, SS 2012

4 Numerical Analysis of Differential Equations 47 A common way to visualize a first order ODE y = f(t, y) is provided by its associated direction field, obtained by drawing an arrow with slope y = f(t, y) at every point in the (t, y)-plane. The graph of the solution to the IVP y = f(t, y), y(t 0 ) = y 0, must pass through the point (t 0, y 0 ) and its tangents at every point (t, y(t)) must be parallel to the direction field. The following figure illustrates the Euler method approximating the solution of the logistic equation y = y(1 y) with IC y(0) = 1/10 using the step size h = 1. Rather than following its exact trajectory (which is, of course, impossible), the Euler scheme may be viewed at producing a piecewise linear approximation. At the starting point t 0 the Euler approximation uses the correct slope f(t 0, y 0 ) = 9/100. However, from the next point t 1 = h = 1 onwards the slope is wrong, and the approximation may deviate further and further from the exact solution. 2.1 Some Simple Methods TU Bergakademie Freiberg, SS 2012

5 Numerical Analysis of Differential Equations Euler Verfahren y =y(1 y), y(0)=1/ Some Simple Methods TU Bergakademie Freiberg, SS 2012

6 Numerical Analysis of Differential Equations 49 4 Richtungsfeld von y =x(y 2) und Loesungen mit Anfangswerten y(0) = 1:0.5:1 2.1 Some Simple Methods TU Bergakademie Freiberg, SS 2012

7 Numerical Analysis of Differential Equations 50 Does Euler s method converge, i.e., do the approximations converge to the exact solution y(t) as h 0? Note: each value of the step size h is associated with a (finite) collection of approximations y n = y n (h), n = 0, 1,..., N(h) := (t end t 0 )/h. The method is said to converge (in [t 0, t end ]) if lim h 0+ max y n(h) y(t n ) = 0. 0 n N(h) (Conv) Theorem 2.1 (Convergence of Euler s method) Under the assumptions of Theorem 1.1 the Euler method is convergent. More precisely, max y n(h) y(t n ) = O(h) as h 0. 0 n N(h) 2.1 Some Simple Methods TU Bergakademie Freiberg, SS 2012

8 Numerical Analysis of Differential Equations 51 exakte Loesung Schrittweite h 0 Schrittweite h 1 Schrittweite h 2 t_0 t 2.1 Some Simple Methods TU Bergakademie Freiberg, SS 2012

9 Numerical Analysis of Differential Equations 52 Modified Euler method: y 0 given, y n+1 := y n + h f ( t n h, y n hf (t n, y n ) ) n = 0, 1,..., N 1. Improved Euler method: y 0 gegeben, y n+1 := y n h[f (t n, y n ) + f (t n + h, y n + hf (t n, y n ))] n = 0, 1,..., N 1. The construction of these variants of Euler s method is also easily interpreted using the direction field. Both are convergent schemes in fact (by a trivial modification of the proof of Theorem 2.1) one can show that for both methods max y n(h) y(t n ) = O(h 2 ), as h 0. 0 n N(h) 2.1 Some Simple Methods TU Bergakademie Freiberg, SS 2012

10 Numerical Analysis of Differential Equations exakte Loesung Euler Verfahren modifiziertes Euler Verf. verbessertes Euler Verf Some Simple Methods TU Bergakademie Freiberg, SS 2012

11 Numerical Analysis of Differential Equations 54 Implicit Euler method: y 0 given, y n+1 := y n + h f (t n + h, y n+1 ) n = 0, 1,..., N 1. In contrast with the methods introduced so far, this is an implicit method, which means that determining y n+1 from y n requires the solution of a system of algebraic equations, which, unless f is linear in y, is generally nonlinear. As in Theorem 2.1 one can show that for the implicit Euler method max y n(h) y(t n ) = O(h) as h 0. 0 n N(h) 2.1 Some Simple Methods TU Bergakademie Freiberg, SS 2012

12 Numerical Analysis of Differential Equations 55 The schemes we have considered up to now involved two approximations of the solution at consecutive times. Such schemes are called one-step methods. Multistep methods are based on difference formulas involving more than two solution values. Examples are y(t + h) y(t h) 2h 3y(t) 4y(t h) + y(t 2h) 2h = y (t) + h2 6 y (t) + O(h 4 ), = y (t) h2 3 y (t) + O(h 3 ). When these formulas are evaluated at t = t n+1, we obtain Some Simple Methods TU Bergakademie Freiberg, SS 2012

13 Numerical Analysis of Differential Equations the explicit Midpoint-rule: (leapfrog method) y 0, y 1 given, y n+1 = y n 1 + 2h f (t n, y n ), n = 1, 2,..., N 1. (2.1) as well as an implicit method belonging to the so-called BDF-family (Backward Differentiation Methods) A BDF method: y 0, y 1 given, 3y n+1 4y n + y n 1 = 2hf (t n + h, y n+1 ), n = 1, 2,..., N 1. We observe that these methods can only proceed once, in addition to the initial vaue y 0, an approximation y 1 is available. All multistep methods require such a startup calculation. 2.1 Some Simple Methods TU Bergakademie Freiberg, SS 2012

14 Numerical Analysis of Differential Equations One-Step Methods Definition and Properties A one-step method (OSM) has the general form y n+1 = y n + h Φ f (y n+1, y n, t n ; h). pos- We will consider exclusively methods whose increment function Φ f sesses the following properties: (OSM) Φ f 0 (y n+1, y n, t n ; h) 0 (V 1 ) and Φ f (y n+1, y n, t n ; h) Φ f (y n+1, y n, t n ; h) M 1 y n+j yn+j. (V 2 ) j=0 For reasonable OSM property (V 2 ) follows from the Lipschitz continuity of f (cf. Theorem 1.1), which we always assume to hold. 2.2 One-Step Methods Definition and Properties TU Bergakademie Freiberg, SS 2012

15 Numerical Analysis of Differential Equations 58 Examples of more complicated OSMs: y n+1 y n = 1 4 h(k 1+3k 3 ), where k 1 = f (t n, y n ), k 2 = f (t n h, y n hk 1), k 3 = f (t n h, y n hk 2), (Example 1) an explicit OSM belonging to the class of Runge-Kutta methods a b (cf. 2.3) and the implicit Runge-Kutta method y n+1 y n = 1 2 h(k 1+k 2 ), where k 1 = f (t n, y n ), k 2 = f (t n + h, y n hk hk 2). (Example 2) a CARLE DAVID TOLMÉ RUNGE ( ) b MARTIN WILHELM KUTTA ( ) 2.2 One-Step Methods Definition and Properties TU Bergakademie Freiberg, SS 2012

16 Numerical Analysis of Differential Equations 59 We say the method (OSM) is convergent if lim h 0 t=t 0 +nh y n = lim h 0 t=t 0 +nh y n (h) = y(t) holds and does so for all IVPs which satisfy the assumptions of Theorem 1.1 where y(t) denotes the solution of such an IVP, uniformly for all t [t 0, t end ], for all solutions {y n (h)} = {y n } of (OSM) with initial values y 0 (h) satisfying lim h 0 y 0 (h) = y 0. Equivalent: The global discretisation error converges to 0 uniformly (as h 0): lim h 0 e n = e n (h) := y(t n ) y n (h) max e n(h) = lim 0 n N h 0 max y(t n) y n (h) = 0. 0 n N 2.2 One-Step Methods Definition and Properties TU Bergakademie Freiberg, SS 2012

17 Numerical Analysis of Differential Equations 60 We define the local discretion error (local truncation error) T n = T n (h) of a method at step n to be the difference of the left and right-hand sides of the expression defining the method (scaled in such a way that the differential equation is approximated as h 0) when the exact solution is substituted for the approximation. A method is said to be consistent of order p if T n (h) = O(h p ) as h 0. Example: Midpoint rule T n = y(t n+1) y(t n 1 ) 2h f(t n, y(t n )) = ] [y (t n ) + h2 6 y (t n ) + O(h 4 ) y (t n ) = h2 6 y (t n ) + O(h 4 ), i.e., we have consistency of order p = One-Step Methods Definition and Properties TU Bergakademie Freiberg, SS 2012

18 Numerical Analysis of Differential Equations 61 Warning: In the literature one often finds a slightly different definition of the local discretization error based on the method s defining equation in the form (OSM). This increases the order of consistency by one power of h. Motivation: Description of the error incurred in a single step independent of that accumulated from previous steps using the localization assumption y n = y(t n ). In this case one step of the scheme (OSM) compared with the exact solution compares as ŷ n+1 (h) = y(t n ) + hφ f (ŷ n+1, y(t n ), t n ; h), y(t n+1 ) = y(t n ) + hφ f (y(t n+1 ), y(t n ), t n ; h) + ht n+1 (h). Denoting the difference S n+1 (h) := y(t n+1 ) ŷ n+1 as step error, we obtain using (V 2 ), S n+1 h Φ f (y(t n+1 ), y(t n ), t n ; h) Φ f (ŷ n+1, y(t n ), t n ; h) + h T n+1 hm S n+1 + h T n+1, implying (1 hm) S n+1 ht n+1 so that S n behaves like ht n as h One-Step Methods Definition and Properties TU Bergakademie Freiberg, SS 2012

19 Numerical Analysis of Differential Equations 62 Relevance of S n : The integration of (IVP) from t 0 to t end requires (t end t 0 )/h steps. If step error S n is committed at step n, this results, under the simplifying assumption that individual error have no effect on each other, in an accumulated (global) error of t end t 0 h S(h) T (h). As we shall see shortly, the crucial property for the validity of this simplifying assumption is the stability of the method. 2.2 One-Step Methods Definition and Properties TU Bergakademie Freiberg, SS 2012

20 Numerical Analysis of Differential Equations Runge-Kutta Methods Derivation via Quadrature Fundamental theorem of calculus: y(t + h) = y(t) + [y(t + h) y(t)] = y(t) + Change of variables: s = t + τh, 0 τ 1 t+h t y (s) ds = y(t) + h 1 0 y (t + τh) dτ. Approximate integral with quadrature formula: 1 0 g(τ) dτ m β j g(γ j ). j=1 ( ) To integrate at least g 1 exactly, we require m j=1 β j = Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

21 Numerical Analysis of Differential Equations 64 We obtain y(t + h) y(t) + h = y(t) + h m β j y (t + γ j h) j=1 m β j f (t + γ j h, y(t + γ j h)) j=1 (RK-1) Problem now: don t know y(t + γ j h) = y(t) + h γ j 0 y (t + τh) dτ. Again approximate by quadrature, but using same nodes {γ j } m j=1 to avoid introducing new unknowns y(t + node h): as in ( ) γj 0 g(τ) dτ m α j,l g(γ l ), j = 1,..., m. ( ) l=1 To integrate at least g 1 exactly, require m l=1 α j,l = γ j, j = 1,..., m. 2.3 Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

22 Numerical Analysis of Differential Equations 65 This yields y(t + γ j h) y(t) + h = y(t) + h m α j,l y (t + γ l h) l=1 m α j,l f (t + γ l h, y(t + γ l h)) l=1 (RK-2) Introduce kj := f ( t + γ j h, y(t + γ j h) ), j = 1,..., m. (RK-2): (RK-1): kj f ( t + γ j h, y(t) + h y(t + h) y(t) + h m α j,l kl ), j = 1,..., m. l=1 m β j kj. j=1 2.3 Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

23 Numerical Analysis of Differential Equations 66 m-stage Runge-Kutta method (RKM): m y n+1 = y n + h β j k j ( j=1 k j = f t n + γ j h, y n + h where ) m α j,l k l, j = 1,..., m. l=1 (RKM) Shorthand notation for RKMs: Butcher Tableau a γ 1 α 1,1 α 1,m... γ m α m,1 α m,m a JOHN CHARLES BUTCHER ( 1933) β 1 β m 2.3 Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

24 Numerical Analysis of Differential Equations 67 Examples. The Butcher Tableau /2 1/2 represents a two-stage explicit RKM, the improved Euler method: [ y n+1 = y n + h 2 f(tn, y n ) + f(t n + h, y n + hf(t n, y n )) ] = y n + h 2 (k 1 + k 2 ), k 1 = f(t n, y n ), k 2 = f(t n + h, y n + hk 1 ) Note: a RKM is explicit when α j,l = 0 j l, i.e., if the coefficients α j,l form a strictly lower triangular matrix. 2.3 Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

25 Numerical Analysis of Differential Equations /4 1/4 2/3 1/4 5/12 1/4 3/4 represents a two-stage implicit RKM: k 1 = f ( t n, y n + h( 1 4 k k 2) ), k 2 = f ( t n h, y n + h( 1 4 k k 2) ), ( two generally nonlinear equations for k 1 and k 2 ) y n+1 = y n + h 4 (k 1 + 3k 2 ). (Example 2 in 2.2 is a further example of an implicit two-stage RKM.) 2.3 Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

26 Numerical Analysis of Differential Equations /2 1/ /6 4/6 1/6 represents an explicit three-stage RKM Kutta s third-order method: k 1 = f (t n, y n ), k 2 = f ( t n h, y n hk 1), k 3 = f ( t n + h, y n + h( k 1 + 2k 2 ) ), y n+1 = y n h(k 1 + 4k 2 + k 3 ). 2.3 Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

27 Numerical Analysis of Differential Equations /3 1/ /3 0 2/3 0 1/4 0 3/4 represents another explicit three-stage RKM Heun s third-order method: k 1 = f (t n, y n ), k 2 = f ( t n h, y n hk 1), k 3 = f ( t n h, y n hk 2), y n+1 = y n + h 4 (k 1 + 3k 3 ) (cf. Example 1 in 2.2). 2.3 Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

28 Numerical Analysis of Differential Equations /2 1/ /2 0 1/ /6 2/6 2/6 1/6 represents an explicit four-stage RKM Classical Runge-Kutta method: k 1 = f (t n, y n ), k 2 = f (t n h, y n hk 1), k 3 = f (t n h, y n hk 2), k 4 = f (t n + h, y n + hk 3 ), y n+1 = y n h(k 1 + 2k 2 + 2k 3 + k 4 ). 2.3 Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

29 Numerical Analysis of Differential Equations 72 An (equivalent) alternative way of representing (RKM) is m y n+1 = y n + h β j f (t n + γ j h, ỹ j ) j=1 m where ỹ j = y n + h α j,l f (t n + γ l h, ỹ l ), l=1 j = 1,..., m (RKM ) (simply set k j = f (t n + γ j h, ỹ j )). Interpretation: ỹ j : approximate solution y at time t n + γ j h k j : approximate slopes y at time t n + γ j h 2.3 Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

30 Numerical Analysis of Differential Equations Consistency of Runge-Kutta Methods All RKM can be written y n+1 = y n +h Φ f (y n+1, y n, t n ; h) where Φ f (y n+1, y n, t n ; h) = m β j k j. j=1 A RKM is consistent (and therefore convergent) if, and only if, m β j = 1. j=1 Determining the order of consistency of RKMs (or constructing m-stage RKMs with maximal order of consistency) often leads to rather involved calculations. 2.3 Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

31 Numerical Analysis of Differential Equations 74 As a tractable example, we shall analyze all explicit three-stage RKMs. Any such method has a Butcher Tableau of the form γ 2 γ γ 3 γ 3 α 3,2 α 3,2 0. We expand the local truncation error β 1 β 2 β 3 T n+1 (h) = y(t n+1) y(t n ) h = y(t n+1) y(t n ) h Φ f (y(t n ), t n ; h) 3 β j k j j=1 in powers of h (assuming y and f sufficiently differentiable). 2.3 Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

32 Numerical Analysis of Differential Equations 75 We consider scalar IVPs and, introducing the abbreviations F := f t + f y f and G := f tt + 2f ty f + f yy f 2, all derivatives of f evaluated at (t n, y(t n )) employ the chain rule to obtain y(t n+1 ) y(t n ) h On the other hand, we also have k 1 = f(t n, y(t n )) = f, = f F h (G + f yf )h 2 + O(h 3 ). k 2 = f(t n + hγ 2, y(t n ) + hγ 2 k 1 ) = f + hγ 2 F h2 γ 2 2G + O(h 3 ), k 3 = f(t n + hγ 3, y(t n ) + h(γ 3 α 3,2 )k 1 + hα 3,2 k 2 ) = f + hγ 3 F + h 2 (γ 2 α 3,2 F f y γ2 3G) + O(h 3 ). 2.3 Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

33 Numerical Analysis of Differential Equations 76 This means: T n+1 (h) = Consequences: [ 1 ] 3 j=1 β j f + [ 1 2 β ] 2γ 2 β 3 γ 3 F h + [ ( 1 3 β 2γ 2 2 β 3 γ 2 3) 1 2 G + ( 1 6 β 3γ 2 α 3,2 )F f y ] h 2 + O(h 3 ) (1) The Euler method is the only explicit one-stage RKM of order one (β 1 = 1). (2) All explicit two-stage RKM of order 2 are characterized by β 1 + β 2 = 1 and β 2 γ 2 = 1 2. Examples are the modified (β 1 = 0, β 2 = 1, γ 2 = 1 2 ) and the improved (β 1 = β 2 = 1 2, γ 2 = 1) Euler methods. There are no explicit two-stage RKMs of order three or higher. 2.3 Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

34 Numerical Analysis of Differential Equations 77 (3) Explicit three-stage RKM of order 3 are characterized by the four equations β 1 + β 2 + β 3 = 1, β 2 γ β 3 γ 2 3 = 1 3, β 2 γ 2 + β 3 γ 3 = 1 2, β 3γ 2 α 3,2 = 1 6. (One can show that none of these methods have order 4.) Examples include Heun s method (β 1 = 1 4, β 2 = 0, β 3 = 3 4, γ 2 = 1 3, γ 3 = α 3,2 = 2 3 ) and Kutta s method (β 1 = 1 6, β 2 = 2 3, β 3 = 1 6, γ 2 = 1 2, γ 3 = 1, α 3,2 = 2). (4) Similar (complicated) calculations reveal the existence of a two-parameter family of explicit four-stage RKMs, none of which have order 5. An example is the classical RKM; further examples are Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

35 Numerical Analysis of Differential Equations /3 1/ /3 1/ /8 3/8 3/8 1/8 (3/8-rule) /5 2/ /5 3/20 3/ /44 15/44 40/ / / /360 55/360 (Kuntzmann s method). 2.3 Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

36 Numerical Analysis of Differential Equations 79 This way of analyzing consistency of explicit RKMs becomes increasingly tedious for higher orders: order 3: 4 nonlinear equations for coefficients (see above) order 8: 200 nonlinear equations for coefficients. Butcher Theory a uses graph theory (trees) for systematic book-keeping of partial derivatives of f for computing the order of a given RKM. It does not, however, provide a technique for constructing RKMs of a desired order. In the following, we derive a sequence of necessary order conditions, obtained from the specific family of IVPs y = y + t l 1, y(0) = 0, (l N). a J. C. Butcher, The Numerical Analysis of Ordinary Differential Equations. Runge-Kutta and General Linear Methods. John Wiley & Sons, Chichester Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

37 Numerical Analysis of Differential Equations 80 Theorem 2.2 (Necessary order conditions for RKM) In order for the RKM defined by the Butcher tableau c A b to possess consistency order p there must hold b A k C l 1 e = (l 1)! (l + k)! = 1 l(l + 1)... (l + k) for l = 1, 2,..., p and k = 0, 1,..., p l. Notation: b := [β 1, β 2,..., β m ], A := [α j,ν ] 1 j,ν m, C := diag(γ 1, γ 2,..., γ m ) and e := [1, 1,..., 1] R m. 2.3 Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

38 Numerical Analysis of Differential Equations 81 Special cases of the necessary conditions of Theorem 2.2 are (for k = 0) b C l 1 e = m j=1 β j γ l 1 j = 1 l for l = 1, 2,..., p as well as (for l = 1 with k k + 1) b A k 1 e = 1 k! for k = 1, 2,..., p. Remark. An explicit m-stage RKM can have at most order m, since in this case A m = O (A ist a strictly lower trianglular matrix). In fact, for m 5 the maximal order p(m) attainable by an explicit m-stage RKM is bounded by p(m) m 1. The exact maximal orders for 1 m 12 are m p(m) Runge-Kutta Methods TU Bergakademie Freiberg, SS 2012

39 Numerical Analysis of Differential Equations Linear Multistep Methods Multistep methods are characterized by using solution approximations earlier than y n in the update formula for advancing the numerical approximation from y n to y n+1. This is in contrast with one-step methods, which use only y n, besides possibly additional intermediate (e.g. stage) quantities. Of these, the most important class of linear multistep methods (LMM) possesses the general form r α j y n+j = h j=0 r β j f (t n+j, y n+j ), n = 0, 1, 2,.... (LMM) j=0 defined by the parameters {α j } r j=0 and {β j} r j=0. Since these are unique only up to a (nonzero) scaling factor, they are typically normalized by the condition α r = 1. The method (LMM) is explicit if β r = 0, otherwise implicit. We first study various subfamilies of LMM. 2.4 Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

40 Numerical Analysis of Differential Equations Adams-Methods Adams-Methods have the specific form y n+r y n+r 1 = h r β j f (t n+j, y n+j ), (2.2) j=0 i.e., they are LMMs characterized by α r = 1, α r 1 = 1, in addition to α 0 = α 1 = = α r 2 = 0. The coefficients β j result from maximizing the order of consistency for a given value of r, which is r in the explicit case and r + 1 in the implicit case. The former class of methods are known as Adams-Bashforth methods, the latter Adams-Moulton methods. 2.4 Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

41 Numerical Analysis of Differential Equations 84 Alternative derivation of Adams methods: From the differential equation, we have y(t n+r ) = y(t n+r 1 ) + tn+r tn+r y (τ) dτ = y(t n+r 1 ) + f (τ, y(τ)) dτ. t n+r 1 t n+r 1 Choosing {β j } r j=0 in such a way that tn+r t n+r 1 g(τ) dτ h r β j g(t n+j ) j=0 is an interpolatory quadrature formula a yields the coefficiants of the corresponding (explicit or implicit) Adams method. a i.e., replace g by the unique interpolation polynomial passing through the points {(t n, g(t n )),..., (t n+r 1, g(t n+r 1 ))} of degree r 1 (explicit case) or {(t n, g(t n )),..., (t n+r, g(t n+r ))} of degree r (implicit case) 2.4 Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

42 Numerical Analysis of Differential Equations 85 For r = 1 to 4 the explicit Adams-Bashforth methods are as follows: r = 1 : y n+1 = y n + hf (t n, y n ), r = 2 : y n+2 = y n+1 + h 2 ( f (t n, y n ) + 3f (t n+1, y n+1 )), r = 3 : y n+3 = y n+2 + h 12 (5f (t n, y n ) 16f (t n+1, y n+1 ) +23f (t n+2, y n+2 )), r = 4 : y n+4 = y n+3 + h 24 ( 9f (t n, y n ) + 37f (t n+1, y n+1 ) 59f (t n+2, y n+2 ) + 55f (t n+3, y n+3 )). Forr r = 1 we recover the explicit Euler method. 2.4 Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

43 Numerical Analysis of Differential Equations 86 The implicit Adams-Moulton methods for r = 1 to 4 read: r = 1 : y n+1 = y n + h 2 (f (t n, y n ) + f (t n+1, y n+1 )), r = 2 : y n+2 = y n+1 + h 12 ( f (t n, y n ) + 8f (t n+1, y n+1 ) +5f (t n+2, y n+2 )), r = 3 : y n+3 = y n+2 + h 24 (f (t n, y n ) 5f (t n+1, y n+1 ) +19f (t n+2, y n+2 + 9f (t n+3, y n+3 ))), r = 4 : y n+4 = y n+3 + h ( 19f (tn, y n ) + 106f (t n+1, y n+1 ) f (t n+2, y n+2 ) + 646f (t n+3, y n+3 ) + 251f (t n+4, y n+4 ) ). For r = 1 we recover the trapezoidal rule. 2.4 Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

44 Numerical Analysis of Differential Equations Nyström Methods Nyström methods have the form y n+r y n+r 2 = h r β j f (t n+j, y n+j ). (2.3) j=0 The explicit Nyström method of order r = 2 is the midpoint rule (2.1). The implicit Nyström method of order r = 2 is Simpson s rule y n+2 y n = h 3 ( ) f (t n y n ) + 4f (t n+1, y n+1 ) + f (t n+2, y n+2 ). 2.4 Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

45 Numerical Analysis of Differential Equations BDF Methods In contrast with Runge-Kutta and Adams methods, which were derived using quadrature, the BDF methods (backward differentiation formulas) are based on numerical differentiation. Construction of linear r-step method: (1) Assume approximations y n, y n 1,..., y n r+1 available. (2) To compute new approximation y n+1, denote by p P r the unique interpolating polynomial through the r + 1 points {(t j, y j )} n+1 j=n r+1. (3) The missing interpolation condition p(t n+1 ) = y n+1 (we have not yet computed y n+1 ) is replaced by the requirement p (t n+1 ) = f(t n+1, y n+1 ), (2.4) i.e., that p satisfy the differential equation at t n Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

46 Numerical Analysis of Differential Equations 89 The Newton form of the resulting interpolation polynomial is given by p(t) = y n+1 + y n+1,n (t t n+1 ) + y n+1,n,n 1 (t t n+1 )(t t n ) + + y n+1,n,...,n r+2,n r+1 (t t n+1 ) (t t n r+2 ) in terms of the r + 1 divided differences y n+1,..., y n+1,n,...,n+1 r. Under the simplifying assumption of constant stepsize h, its derivative at t = t n+1 is r p (t n+1 ) = y n+1,n,...,n+1 j h j 1 (j 1)!. (2.5) j=1 Introducing the backward differences 0 y n+1 := y n+1 y n+1 := y n+1 y n, j y n+1 := ( j 1 y n+1 ) = j 1 y n+1 j 1 y n, j = 2, 3,..., 2.4 Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

47 Numerical Analysis of Differential Equations 90 p (t n+1 ) may be expressed as p (t n+1 ) = 1 h r j=1 j y n+1 j, which, combined with condition (2.4), yields the BDF method of order r as r j=1 1 j j y n+1 = hf(t n+1, y n+1 ), r N, (2.6) denoted BDFr for short. The index translation n+1 n+r and dividing both sides by the coefficient of y n+1 transforms the BDF formulas (2.6) to our standard notation (LMM) for linear multistep methods. 2.4 Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

48 Numerical Analysis of Differential Equations 91 The BDF methods for r = 1 to 6 are: r = 1 : y n+1 y n = hf (t n+1, y n+1 ), r = 2 : 3 2 y n+2 2y n y n = hf (t n+2, y n+2 ) r = 3 : 11 6 y n+3 3y n y n y n = hf (t n+3, y n+3 ), r = 4 : y n+4 4y n+3 + 3y n y n y n = hf (t n+4, y n+4 ), r = 5 : y n+5 5y n+4 + 5y n y n y n u n = hf (t n+5, y n+5 ), r = 6 : y n+6 6y n y n y n y n y n u n = hf (t n+6, y n+6 ). For r = 1 we recover the implicit Euler method. 2.4 Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

49 Numerical Analysis of Differential Equations The Local Discretization Error of LMM Inserting the exact solution into (LMM), using the fact that y(t) satisfies y (t n ) = f (t n, y(t n )), we obtain after dividing by h T n+r (h) = 1 r r α j y(t n+j ) h β j y (t n+j ). h j=0 j=0 Expanding all evaluations of y and y at t = t n y(t n+j ) = y(t n ) + jhy (t n ) + (jh)2 2 y (t n+j ) = y (t n ) + jhy (t n ) + (jh)2 2 y (t n ) + (jh)3 6 y (t n ) + (jh)3 6 y (t n ) +... and collecting all terms with the same power of h, we obtain... y (t n ) +..., 2.4 Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

50 Numerical Analysis of Differential Equations 93 T n+r (h) = 1 r r α j y(t n ) + (jα j β j ) y (t n ) h j=0 j=0 r ( ) + h j 2 2 α j jβ j y (t n ) j=0 r ( ) + + h q j q+1 (q + 1)! α j jq q! β j y (q+1) (t n ) + O(h q+1 ). j=0 The LMM is consistent, i.e., T n (h) 0 as h 0, if and only if r α j = 0 as well as r jα j = r β j. (2.7) j=0 j=0 j=0 Its order of consistency is p N whenever the first p + 1 terms in brackets vanish. 2.4 Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

51 Numerical Analysis of Differential Equations Characteristic Polynomials The polynomials formed with the coefficients {α j } r j=0 and {β j} r j=0 of a LMM, known as its characteristic polynomials ρ(ζ) := r α j ζ j and σ(ζ) := j=0 r β j ζ j, (2.8) j=0 play an important role in its analysis. For an r-step method we have ρ P r and, for an implicit r-step method, also σ P r. When an r step method is explicit the degree of σ ist strictly less than r. The two consistency conditions (2.7) can be formulated in terms of the two characteristic polynomials as ρ(1) = 0 und ρ (1) = σ(1). (2.9) 2.4 Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

52 Numerical Analysis of Differential Equations 95 Example: For the Adams-Moulton method with r = 2 y n+2 = y n+1 + h ( ) f (t n, y n ) + 8f (t n+1, y n+1 ) + 5f (t n+2, y n+2 ), 12 we have α 0 = 0, α 1 = 1, α 2 = 1, and the characteristic polynomials are ρ(ζ) = ζ 2 ζ, σ(ζ) = 1 12 β 0 = 1 12, β 1 = 8 12, β 2 = 5 12 ( 5ζ 2 + 8ζ 1 ). 2.4 Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

53 Numerical Analysis of Differential Equations 96 Because of α 0 + α 1 + α 2 = 0, 0 α α α 2 (β 0 + β 1 + β 2 ) = 0, 0 α α α 2 (0 β β β 2 ) = 0, 1 6 (0 α α 1 + 8α 2 ) 1 2 (0 β β β 2 ) = 0, 1 24 (0 α α α 2 ) 1 6 (0 β β β 2 ) = this method is consistent of order p = Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

54 Numerical Analysis of Differential Equations Starting Values When an r-step method (r > 1) is started, it is necessary to calculate, besides y 0, the additional solution approximations y 1, y 2,..., y r 1. To maintain the method s order p, it is sufficient to calculate these using any other method of order at least p 1. Explanation: Since only a fixed number of steps (r 1) is performed with this startup method, independently of h, no power of h is lost in the accumulation of step errors as h 0; for this reason the order of accuracy depends only on that of the method used from step r onwards. Example: To generate y 1 to start the midpoint rule (2.1), which is consistent of order p = 2, one can use the explicit Euler method without loss of order. 2.4 Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

55 Numerical Analysis of Differential Equations Predictor-Corrector Methods A popular technique for avoiding the expensive solution of algebraic equation systems required by implicit LMM consists of (a) first computing an initial approximation ŷ n+r of y(t n+r ) using an explicit method with the same step size (predictor) and (b) then inserting ŷ n+r in place of y n+r in an implicit method (corrector). Example: Combination of r = 1 Adams-Bashforth method (explicit Euler) with Adams-Moulton method with same step number (trapezoidal rule): ŷ n+1 = y n + hf (t n, y n ), y n+1 = y n + h 2 (f (t n, y n ) + f (t n+1, ŷ n+1 )), (predictor step), (corrector step). It can be shown: this predictor/corrector combination has consistency order p = Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

56 Numerical Analysis of Differential Equations One-Step vs. Multistep Methods Advantages of one-step methods: One-step methods are self-starting. Changing the stepsize h possible witout extra effort. Integration across discontinuities of the solution possible without loss of order if these are grid points. Advantage of multistep methods: one evaluation of right hand side per time step (important if f expensive to evaluate). 2.4 Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

57 Numerical Analysis of Differential Equations Convergence of One-Step Methods Theorem 2.3 (Relation between local and global discretization error) Under the given assumptions (cf. Theorem 1.1 as well as (V 2 )) there holds ( ) e n ( e 0 + (t n t 0 ) max T j exp(m(t n t 0 )). 1 j n In particular: a one-step method is convergent if, and only if, it is consistent. 2.5 Convergence of One-Step Methods TU Bergakademie Freiberg, SS 2012

58 Numerical Analysis of Differential Equations 101 Theorem 2.4 (Total error of explicit OSM) Let the explicit one-step method (OSM) be realized in floating point arithmetic with machine precision ε as ỹ n+1 = ỹ n + hφ f (ỹ n, t n ; h) + ε n+1, ỹ 0 = y 0 + ε 0. If ε n ε and T n T for all n = 0, 1,..., then y(t n ) ỹ n ( ε 0 + (t n t 0 )(T + ε h )) exp(m(t n t 0 )). Dilemma: For h large, in general (local error) T large; for h small T becomes small, however ε h grows, i.e., total error dominated by rounding error. This fact favors high order methods, which achieve small T for moderate values of the step size h. 2.5 Convergence of One-Step Methods TU Bergakademie Freiberg, SS 2012

59 Numerical Analysis of Differential Equations Zero-Stability of Linear Multistep Methods Convergence of one-step methods: Contribution of n-th step error S n bounded by e M(t end t 0 ) S n, which still goes to zero as h 0. This property of the error is called zero-stability, and it always holds for OSM. This is not the case for LMM. Example: For the LMM y n+2 3y n+1 + 2y n discretization error is = hf (t n, y n ) the local T n+2 (h) = h 2 y (t n ) + O(h 2 ). We apply it to IVP y (t) = 0, y(0) = 0 with exact solution y(t) 0. (2.10) For starting values y 0 = y 1 = 0 it indeed yields the exact solution. 2.6 Zero-Stability of Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

60 Numerical Analysis of Differential Equations 103 However, setting y 1 = h to simulate an error of order h when approximating y(t 1 ), one obtains for the approximation of the solution at t = 1 the following values (h = 1/N): N y N Since y 1 0 as h 0 these initial data are admissible. Nonetheless, y N fails to converge to y(t N ) = y(1) = 0. The approximations satisfy the difference equation with solution y n+2 3y n+1 + 2y n = 0, n = 0, 1, 2,..., (2.11) y n = (2 n 1)y 1 (2 n 2)y 0 = 2y 0 y n (y 1 y 0 ), n Zero-Stability of Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

61 Numerical Analysis of Differential Equations Solution of Linear Difference Equations Solutions of a homogeneous linear difference equation of order r α r y n+r + α r 1 y n+r α 0 y n = r α j y n+j = 0 (2.12) j=0 can be obtained by hypothesizing solutions of the form y n = ζ n ( Ansatz ). Inserting in (2.12) and dividing by ζ n leads to the condition r α j ζ j = 0, j=0 i.e., y n = ζ n solves (2.12) if ζ is a root of the first characteristic polynomial ρ(ζ) (cf. (2.8)). 2.6 Zero-Stability of Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

62 Numerical Analysis of Differential Equations 105 If the characteristic polynomial ρ has the factored form ρ(ζ) = α r (ζ ζ 1 )(ζ ζ 2 ) (ζ ζ r ) with r distinct roots {ζ j } r j=1, then {ζn j }r j=1 constitute r linearly independent solutions of (2.12). Due to the linearity and homogeneity of (2.12), any linear combination y n = c 1 ζ n 1 + c 2 ζ n c r ζ n r, c 1,..., c r arbitrary, (2.13) of these is also a solution. Coefficients {c j } r j=1 determined by fixing first r values of y n: c 1 + c c r = y 0, c 1 ζ 1 + c 2 ζ c r ζ r = y 1, c 1 ζ r c 2 ζ r c r ζ r 1 r = y r Zero-Stability of Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

63 Numerical Analysis of Differential Equations 106 Example: The characteristic equation of the difference equation (2.11) is ρ(ζ) = ζ 2 3ζ + 2 = (ζ 1)(ζ 2). Conclusion: If ρ(ζ) has roots of modulus > 1, the associated LMM cannot be convergent. If two (or more) roots coincide, i.e., ζ i = ζ j, then ζi n and ζj n are linearly dependent. A further, linearly independent solution is y n = nζ n 1 i. For a root ζ i of multiplicity three, y n = n 2 ζ n 2 i Example: The (consistent) LMM applied to y (t) = 0. is linearly independent etc. y n+2 2y n+1 + y n = h 2 [ f (tn+2, y n+2 ) f (t n, y n ) ]. Conclusion: If ρ(ζ) has multiple roots of modulus 1, the associated LMM is also not convergent. 2.6 Zero-Stability of Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

64 Numerical Analysis of Differential Equations 107 Example: The consistent LMM applied to (2.10). y n+3 2y n y n y n = h 4 f(t n, y n ) Conclusion: For multiple roots of modulus > 1, the associated LMM is convergent (linear growth is outweighed by exponential decay). A LMM is called zero-stable, if the roots {ζ j } r j=1 polynomial ρ(ζ) satisfy (a) ζ j 1 for all roots. (b) ζ j < 1 for all multiple roots. of its first characteristic Examples: All Adams-methods are zero-stable. The BDF-methods are zero-stable for 1 r Zero-Stability of Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

65 Numerical Analysis of Differential Equations 108 These examples indicate that zero-stability is a necessary condition for convergence of LMMs. In fact, this condition is also sufficient. Theorem 2.5 (Dahlquist, 1956) For LMM applied to (IVP), consistency + zero-stability convergence. Remarks 2.6 (a) The first characteristic polynomial of a consistent LMM always possesses the root ζ = 1 (cf. (2.9)). (b) We will see that, besides zero-stability, there are further important stability properties for numerical methods for IVPs. (c) All single-step methods are zero-stable. 2.6 Zero-Stability of Linear Multistep Methods TU Bergakademie Freiberg, SS 2012

66 Numerical Analysis of Differential Equations Stepsize Control In practice, numerical methods for solving IVPs for ODEs almost never employ a constant stepsize h. Rather, it is much more efficient to adapt h to the local behavior of the solution y in the sense that rapid change in y necessitates small h slow variation of y allows large h. Here we will introduce one (of many) stepsize control schemes aimed at maintaining the (easily estimated) local discretization error T n near a prescribed tolerance tol T n tol, n = 1, 2,.... For systems of ODEs (particularly when solution components vary strongly in magnitude) absolute error tolerances for each solution component as well as a global error tolerance are typically supplied. 2.7 Stepsize Control TU Bergakademie Freiberg, SS 2012

67 Numerical Analysis of Differential Equations 110 Theorem 2.3 states that bounding the local discretization error also results in bounding the global discretization error (the actual quantity of interest). The local discretization error may be estimated by using two methods of different consistency orders, say, p and q where p < q, for computing y n from y n 1 : y n = y n 1 + hφ f (y n 1, t n 1 ; h) and ŷ n = y n 1 + h Φ f (y n 1, t n 1 ; h) with associated local discretization errors This implies T n = y(t n) y(t n 1 ) h T n = y(t n) y(t n 1 ) h Φ f (y(t n 1 ), t n 1 ; h) = O(h p ), Φ f (y(t n 1 ), t n 1 ; h) = O(h q ). T n T n = Φ f (y n 1, t n 1 ; h) Φ f (y n 1, t n 1 ; h) = 1 h (ŷ n y n ). 2.7 Stepsize Control TU Bergakademie Freiberg, SS 2012

68 Numerical Analysis of Differential Equations 111 Because T n T n = T n (1 + O(h q p )) T n, the quantity provides a (rough) estimate for T n. 1 h y n ŷ n T n Whenever 1 h y n ŷ n > tol, we discard the stepsize h, replacing it with h determined by ) p ( h = α h tol h y n ŷ n. ( ) This choice of h is motivated as follows: discarded stepsize h: 1 h y n ŷ n T n = ch p + O(h p+1 ) ch p, desired stepsize h: tol = T n = c h p + O( h p+1 ) c h p. (α is added as a safety factor, e.g. α = 0.9.) 2.7 Stepsize Control TU Bergakademie Freiberg, SS 2012

69 Numerical Analysis of Differential Equations 112 Starting again from y n 1, approximations y n and ŷ n now of y(t n 1 + h) 1 are then generated and this process repeated until h y n ŷ n tol holds. When this is achieved, ( ) is used to propose a (trial) stepsize for the next step (n n + 1). An elegant way to limit the computing effort is to compute y n and ŷ n using two RKM (of different orders) whose Butcher tableaus differ only in the vector b (same A and c). As a result, the slopes k j of their stages are also the same. Such pairs of RKMs are called embedded RKMs and denoted c A b b e.g /2 1/2. Here the explicit Euler method (p = 1) is embedded in the improved Euler method (p = 2). 2.7 Stepsize Control TU Bergakademie Freiberg, SS 2012

70 Numerical Analysis of Differential Equations 113 Another popular embedded pair is the Fehlberg 4(5)-formula: consisting of two six-stage RKMs of orders 4 and Stepsize Control TU Bergakademie Freiberg, SS 2012

71 Numerical Analysis of Differential Equations 114 A further example is the Dormand-Prince 4(5)-formula: Here a six-stage RKM of order 4 is embedded in a seven-stage method of order Stepsize Control TU Bergakademie Freiberg, SS 2012

Numerical Methods for Differential Equations

Numerical Methods for Differential Equations Numerical Methods for Differential Equations Chapter 2: Runge Kutta and Multistep Methods Gustaf Söderlind Numerical Analysis, Lund University Contents V4.16 1. Runge Kutta methods 2. Embedded RK methods

More information

Scientific Computing: An Introductory Survey

Scientific Computing: An Introductory Survey Scientific Computing: An Introductory Survey Chapter 9 Initial Value Problems for Ordinary Differential Equations Prof. Michael T. Heath Department of Computer Science University of Illinois at Urbana-Champaign

More information

Numerical Methods for Differential Equations

Numerical Methods for Differential Equations Numerical Methods for Differential Equations Chapter 2: Runge Kutta and Linear Multistep methods Gustaf Söderlind and Carmen Arévalo Numerical Analysis, Lund University Textbooks: A First Course in the

More information

CS 450 Numerical Analysis. Chapter 9: Initial Value Problems for Ordinary Differential Equations

CS 450 Numerical Analysis. Chapter 9: Initial Value Problems for Ordinary Differential Equations Lecture slides based on the textbook Scientific Computing: An Introductory Survey by Michael T. Heath, copyright c 2018 by the Society for Industrial and Applied Mathematics. http://www.siam.org/books/cl80

More information

CHAPTER 5: Linear Multistep Methods

CHAPTER 5: Linear Multistep Methods CHAPTER 5: Linear Multistep Methods Multistep: use information from many steps Higher order possible with fewer function evaluations than with RK. Convenient error estimates. Changing stepsize or order

More information

MTH 452/552 Homework 3

MTH 452/552 Homework 3 MTH 452/552 Homework 3 Do either 1 or 2. 1. (40 points) [Use of ode113 and ode45] This problem can be solved by a modifying the m-files odesample.m and odesampletest.m available from the author s webpage.

More information

Lecture Notes to Accompany. Scientific Computing An Introductory Survey. by Michael T. Heath. Chapter 9

Lecture Notes to Accompany. Scientific Computing An Introductory Survey. by Michael T. Heath. Chapter 9 Lecture Notes to Accompany Scientific Computing An Introductory Survey Second Edition by Michael T. Heath Chapter 9 Initial Value Problems for Ordinary Differential Equations Copyright c 2001. Reproduction

More information

Solving Ordinary Differential equations

Solving Ordinary Differential equations Solving Ordinary Differential equations Taylor methods can be used to build explicit methods with higher order of convergence than Euler s method. The main difficult of these methods is the computation

More information

Numerical Methods - Initial Value Problems for ODEs

Numerical Methods - Initial Value Problems for ODEs Numerical Methods - Initial Value Problems for ODEs Y. K. Goh Universiti Tunku Abdul Rahman 2013 Y. K. Goh (UTAR) Numerical Methods - Initial Value Problems for ODEs 2013 1 / 43 Outline 1 Initial Value

More information

Numerical solution of ODEs

Numerical solution of ODEs Numerical solution of ODEs Arne Morten Kvarving Department of Mathematical Sciences Norwegian University of Science and Technology November 5 2007 Problem and solution strategy We want to find an approximation

More information

Chapter 5 Exercises. (a) Determine the best possible Lipschitz constant for this function over 2 u <. u (t) = log(u(t)), u(0) = 2.

Chapter 5 Exercises. (a) Determine the best possible Lipschitz constant for this function over 2 u <. u (t) = log(u(t)), u(0) = 2. Chapter 5 Exercises From: Finite Difference Methods for Ordinary and Partial Differential Equations by R. J. LeVeque, SIAM, 2007. http://www.amath.washington.edu/ rjl/fdmbook Exercise 5. (Uniqueness for

More information

Applied Math for Engineers

Applied Math for Engineers Applied Math for Engineers Ming Zhong Lecture 15 March 28, 2018 Ming Zhong (JHU) AMS Spring 2018 1 / 28 Recap Table of Contents 1 Recap 2 Numerical ODEs: Single Step Methods 3 Multistep Methods 4 Method

More information

Initial value problems for ordinary differential equations

Initial value problems for ordinary differential equations Initial value problems for ordinary differential equations Xiaojing Ye, Math & Stat, Georgia State University Spring 2019 Numerical Analysis II Xiaojing Ye, Math & Stat, Georgia State University 1 IVP

More information

Module 4: Numerical Methods for ODE. Michael Bader. Winter 2007/2008

Module 4: Numerical Methods for ODE. Michael Bader. Winter 2007/2008 Outlines Module 4: for ODE Part I: Basic Part II: Advanced Lehrstuhl Informatik V Winter 2007/2008 Part I: Basic 1 Direction Fields 2 Euler s Method Outlines Part I: Basic Part II: Advanced 3 Discretized

More information

Jim Lambers MAT 772 Fall Semester Lecture 21 Notes

Jim Lambers MAT 772 Fall Semester Lecture 21 Notes Jim Lambers MAT 772 Fall Semester 21-11 Lecture 21 Notes These notes correspond to Sections 12.6, 12.7 and 12.8 in the text. Multistep Methods All of the numerical methods that we have developed for solving

More information

Lecture 4: Numerical solution of ordinary differential equations

Lecture 4: Numerical solution of ordinary differential equations Lecture 4: Numerical solution of ordinary differential equations Department of Mathematics, ETH Zürich General explicit one-step method: Consistency; Stability; Convergence. High-order methods: Taylor

More information

Ordinary Differential Equations

Ordinary Differential Equations CHAPTER 8 Ordinary Differential Equations 8.1. Introduction My section 8.1 will cover the material in sections 8.1 and 8.2 in the book. Read the book sections on your own. I don t like the order of things

More information

Fourth Order RK-Method

Fourth Order RK-Method Fourth Order RK-Method The most commonly used method is Runge-Kutta fourth order method. The fourth order RK-method is y i+1 = y i + 1 6 (k 1 + 2k 2 + 2k 3 + k 4 ), Ordinary Differential Equations (ODE)

More information

Consistency and Convergence

Consistency and Convergence Jim Lambers MAT 77 Fall Semester 010-11 Lecture 0 Notes These notes correspond to Sections 1.3, 1.4 and 1.5 in the text. Consistency and Convergence We have learned that the numerical solution obtained

More information

multistep methods Last modified: November 28, 2017 Recall that we are interested in the numerical solution of the initial value problem (IVP):

multistep methods Last modified: November 28, 2017 Recall that we are interested in the numerical solution of the initial value problem (IVP): MATH 351 Fall 217 multistep methods http://www.phys.uconn.edu/ rozman/courses/m351_17f/ Last modified: November 28, 217 Recall that we are interested in the numerical solution of the initial value problem

More information

NUMERICAL SOLUTION OF ODE IVPs. Overview

NUMERICAL SOLUTION OF ODE IVPs. Overview NUMERICAL SOLUTION OF ODE IVPs 1 Quick review of direction fields Overview 2 A reminder about and 3 Important test: Is the ODE initial value problem? 4 Fundamental concepts: Euler s Method 5 Fundamental

More information

Linear Multistep Methods I: Adams and BDF Methods

Linear Multistep Methods I: Adams and BDF Methods Linear Multistep Methods I: Adams and BDF Methods Varun Shankar January 1, 016 1 Introduction In our review of 5610 material, we have discussed polynomial interpolation and its application to generating

More information

Numerical Differential Equations: IVP

Numerical Differential Equations: IVP Chapter 11 Numerical Differential Equations: IVP **** 4/16/13 EC (Incomplete) 11.1 Initial Value Problem for Ordinary Differential Equations We consider the problem of numerically solving a differential

More information

8.1 Introduction. Consider the initial value problem (IVP):

8.1 Introduction. Consider the initial value problem (IVP): 8.1 Introduction Consider the initial value problem (IVP): y dy dt = f(t, y), y(t 0)=y 0, t 0 t T. Geometrically: solutions are a one parameter family of curves y = y(t) in(t, y)-plane. Assume solution

More information

HIGHER ORDER METHODS. There are two principal means to derive higher order methods. b j f(x n j,y n j )

HIGHER ORDER METHODS. There are two principal means to derive higher order methods. b j f(x n j,y n j ) HIGHER ORDER METHODS There are two principal means to derive higher order methods y n+1 = p j=0 a j y n j + h p j= 1 b j f(x n j,y n j ) (a) Method of Undetermined Coefficients (b) Numerical Integration

More information

1 Ordinary Differential Equations

1 Ordinary Differential Equations Ordinary Differential Equations.0 Mathematical Background.0. Smoothness Definition. A function f defined on [a, b] is continuous at ξ [a, b] if lim x ξ f(x) = f(ξ). Remark Note that this implies existence

More information

Bindel, Fall 2011 Intro to Scientific Computing (CS 3220) Week 12: Monday, Apr 18. HW 7 is posted, and will be due in class on 4/25.

Bindel, Fall 2011 Intro to Scientific Computing (CS 3220) Week 12: Monday, Apr 18. HW 7 is posted, and will be due in class on 4/25. Logistics Week 12: Monday, Apr 18 HW 6 is due at 11:59 tonight. HW 7 is posted, and will be due in class on 4/25. The prelim is graded. An analysis and rubric are on CMS. Problem du jour For implicit methods

More information

Initial value problems for ordinary differential equations

Initial value problems for ordinary differential equations AMSC/CMSC 660 Scientific Computing I Fall 2008 UNIT 5: Numerical Solution of Ordinary Differential Equations Part 1 Dianne P. O Leary c 2008 The Plan Initial value problems (ivps) for ordinary differential

More information

Ordinary Differential Equations II

Ordinary Differential Equations II Ordinary Differential Equations II CS 205A: Mathematical Methods for Robotics, Vision, and Graphics Justin Solomon CS 205A: Mathematical Methods Ordinary Differential Equations II 1 / 29 Almost Done! No

More information

Multistage Methods I: Runge-Kutta Methods

Multistage Methods I: Runge-Kutta Methods Multistage Methods I: Runge-Kutta Methods Varun Shankar January, 0 Introduction Previously, we saw that explicit multistep methods (AB methods) have shrinking stability regions as their orders are increased.

More information

COSC 3361 Numerical Analysis I Ordinary Differential Equations (II) - Multistep methods

COSC 3361 Numerical Analysis I Ordinary Differential Equations (II) - Multistep methods COSC 336 Numerical Analysis I Ordinary Differential Equations (II) - Multistep methods Fall 2005 Repetition from the last lecture (I) Initial value problems: dy = f ( t, y) dt y ( a) = y 0 a t b Goal:

More information

Scientific Computing with Case Studies SIAM Press, Lecture Notes for Unit V Solution of

Scientific Computing with Case Studies SIAM Press, Lecture Notes for Unit V Solution of Scientific Computing with Case Studies SIAM Press, 2009 http://www.cs.umd.edu/users/oleary/sccswebpage Lecture Notes for Unit V Solution of Differential Equations Part 1 Dianne P. O Leary c 2008 1 The

More information

Chapter 6 - Ordinary Differential Equations

Chapter 6 - Ordinary Differential Equations Chapter 6 - Ordinary Differential Equations 7.1 Solving Initial-Value Problems In this chapter, we will be interested in the solution of ordinary differential equations. Ordinary differential equations

More information

Introduction to the Numerical Solution of IVP for ODE

Introduction to the Numerical Solution of IVP for ODE Introduction to the Numerical Solution of IVP for ODE 45 Introduction to the Numerical Solution of IVP for ODE Consider the IVP: DE x = f(t, x), IC x(a) = x a. For simplicity, we will assume here that

More information

CS520: numerical ODEs (Ch.2)

CS520: numerical ODEs (Ch.2) .. CS520: numerical ODEs (Ch.2) Uri Ascher Department of Computer Science University of British Columbia ascher@cs.ubc.ca people.cs.ubc.ca/ ascher/520.html Uri Ascher (UBC) CPSC 520: ODEs (Ch. 2) Fall

More information

AN OVERVIEW. Numerical Methods for ODE Initial Value Problems. 1. One-step methods (Taylor series, Runge-Kutta)

AN OVERVIEW. Numerical Methods for ODE Initial Value Problems. 1. One-step methods (Taylor series, Runge-Kutta) AN OVERVIEW Numerical Methods for ODE Initial Value Problems 1. One-step methods (Taylor series, Runge-Kutta) 2. Multistep methods (Predictor-Corrector, Adams methods) Both of these types of methods are

More information

Part IB Numerical Analysis

Part IB Numerical Analysis Part IB Numerical Analysis Definitions Based on lectures by G. Moore Notes taken by Dexter Chua Lent 206 These notes are not endorsed by the lecturers, and I have modified them (often significantly) after

More information

Ordinary differential equations - Initial value problems

Ordinary differential equations - Initial value problems Education has produced a vast population able to read but unable to distinguish what is worth reading. G.M. TREVELYAN Chapter 6 Ordinary differential equations - Initial value problems In this chapter

More information

Review Higher Order methods Multistep methods Summary HIGHER ORDER METHODS. P.V. Johnson. School of Mathematics. Semester

Review Higher Order methods Multistep methods Summary HIGHER ORDER METHODS. P.V. Johnson. School of Mathematics. Semester HIGHER ORDER METHODS School of Mathematics Semester 1 2008 OUTLINE 1 REVIEW 2 HIGHER ORDER METHODS 3 MULTISTEP METHODS 4 SUMMARY OUTLINE 1 REVIEW 2 HIGHER ORDER METHODS 3 MULTISTEP METHODS 4 SUMMARY OUTLINE

More information

Ordinary Differential Equations II

Ordinary Differential Equations II Ordinary Differential Equations II CS 205A: Mathematical Methods for Robotics, Vision, and Graphics Justin Solomon CS 205A: Mathematical Methods Ordinary Differential Equations II 1 / 33 Almost Done! Last

More information

Explicit One-Step Methods

Explicit One-Step Methods Chapter 1 Explicit One-Step Methods Remark 11 Contents This class presents methods for the numerical solution of explicit systems of initial value problems for ordinary differential equations of first

More information

Ordinary differential equation II

Ordinary differential equation II Ordinary Differential Equations ISC-5315 1 Ordinary differential equation II 1 Some Basic Methods 1.1 Backward Euler method (implicit method) The algorithm looks like this: y n = y n 1 + hf n (1) In contrast

More information

On interval predictor-corrector methods

On interval predictor-corrector methods DOI 10.1007/s11075-016-0220-x ORIGINAL PAPER On interval predictor-corrector methods Andrzej Marcinia 1,2 Malgorzata A. Janowsa 3 Tomasz Hoffmann 4 Received: 26 March 2016 / Accepted: 3 October 2016 The

More information

The Initial Value Problem for Ordinary Differential Equations

The Initial Value Problem for Ordinary Differential Equations Chapter 5 The Initial Value Problem for Ordinary Differential Equations In this chapter we begin a study of time-dependent differential equations, beginning with the initial value problem (IVP) for a time-dependent

More information

2tdt 1 y = t2 + C y = which implies C = 1 and the solution is y = 1

2tdt 1 y = t2 + C y = which implies C = 1 and the solution is y = 1 Lectures - Week 11 General First Order ODEs & Numerical Methods for IVPs In general, nonlinear problems are much more difficult to solve than linear ones. Unfortunately many phenomena exhibit nonlinear

More information

The collocation method for ODEs: an introduction

The collocation method for ODEs: an introduction 058065 - Collocation Methods for Volterra Integral Related Functional Differential The collocation method for ODEs: an introduction A collocation solution u h to a functional equation for example an ordinary

More information

Notes for Numerical Analysis Math 5466 by S. Adjerid Virginia Polytechnic Institute and State University (A Rough Draft) Contents Numerical Methods for ODEs 5. Introduction............................

More information

Solving PDEs with PGI CUDA Fortran Part 4: Initial value problems for ordinary differential equations

Solving PDEs with PGI CUDA Fortran Part 4: Initial value problems for ordinary differential equations Solving PDEs with PGI CUDA Fortran Part 4: Initial value problems for ordinary differential equations Outline ODEs and initial conditions. Explicit and implicit Euler methods. Runge-Kutta methods. Multistep

More information

Mathematics for chemical engineers. Numerical solution of ordinary differential equations

Mathematics for chemical engineers. Numerical solution of ordinary differential equations Mathematics for chemical engineers Drahoslava Janovská Numerical solution of ordinary differential equations Initial value problem Winter Semester 2015-2016 Outline 1 Introduction 2 One step methods Euler

More information

Lecture V: The game-engine loop & Time Integration

Lecture V: The game-engine loop & Time Integration Lecture V: The game-engine loop & Time Integration The Basic Game-Engine Loop Previous state: " #, %(#) ( #, )(#) Forces -(#) Integrate velocities and positions Resolve Interpenetrations Per-body change

More information

You may not use your books, notes; calculators are highly recommended.

You may not use your books, notes; calculators are highly recommended. Math 301 Winter 2013-14 Midterm 1 02/06/2014 Time Limit: 60 Minutes Name (Print): Instructor This exam contains 8 pages (including this cover page) and 6 problems. Check to see if any pages are missing.

More information

Math 660 Lecture 4: FDM for evolutionary equations: ODE solvers

Math 660 Lecture 4: FDM for evolutionary equations: ODE solvers Math 660 Lecture 4: FDM for evolutionary equations: ODE solvers Consider the ODE u (t) = f(t, u(t)), u(0) = u 0, where u could be a vector valued function. Any ODE can be reduced to a first order system,

More information

ECE257 Numerical Methods and Scientific Computing. Ordinary Differential Equations

ECE257 Numerical Methods and Scientific Computing. Ordinary Differential Equations ECE257 Numerical Methods and Scientific Computing Ordinary Differential Equations Today s s class: Stiffness Multistep Methods Stiff Equations Stiffness occurs in a problem where two or more independent

More information

Lecture IV: Time Discretization

Lecture IV: Time Discretization Lecture IV: Time Discretization Motivation Kinematics: continuous motion in continuous time Computer simulation: Discrete time steps t Discrete Space (mesh particles) Updating Position Force induces acceleration.

More information

Ordinary Differential Equations

Ordinary Differential Equations Chapter 13 Ordinary Differential Equations We motivated the problem of interpolation in Chapter 11 by transitioning from analzying to finding functions. That is, in problems like interpolation and regression,

More information

Euler s Method, cont d

Euler s Method, cont d Jim Lambers MAT 461/561 Spring Semester 009-10 Lecture 3 Notes These notes correspond to Sections 5. and 5.4 in the text. Euler s Method, cont d We conclude our discussion of Euler s method with an example

More information

Mini project ODE, TANA22

Mini project ODE, TANA22 Mini project ODE, TANA22 Filip Berglund (filbe882) Linh Nguyen (linng299) Amanda Åkesson (amaak531) October 2018 1 1 Introduction Carl David Tohmé Runge (1856 1927) was a German mathematician and a prominent

More information

Section 7.4 Runge-Kutta Methods

Section 7.4 Runge-Kutta Methods Section 7.4 Runge-Kutta Methods Key terms: Taylor methods Taylor series Runge-Kutta; methods linear combinations of function values at intermediate points Alternatives to second order Taylor methods Fourth

More information

5.6 Multistep Methods

5.6 Multistep Methods 5.6 Multistep Methods 1 Motivation: Consider IVP: yy = ff(tt, yy), aa tt bb, yy(aa) = αα. To compute solution at tt ii+1, approximate solutions at mesh points tt 0, tt 1, tt 2, tt ii are already obtained.

More information

CHAPTER 10: Numerical Methods for DAEs

CHAPTER 10: Numerical Methods for DAEs CHAPTER 10: Numerical Methods for DAEs Numerical approaches for the solution of DAEs divide roughly into two classes: 1. direct discretization 2. reformulation (index reduction) plus discretization Direct

More information

Finite Differences for Differential Equations 28 PART II. Finite Difference Methods for Differential Equations

Finite Differences for Differential Equations 28 PART II. Finite Difference Methods for Differential Equations Finite Differences for Differential Equations 28 PART II Finite Difference Methods for Differential Equations Finite Differences for Differential Equations 29 BOUNDARY VALUE PROBLEMS (I) Solving a TWO

More information

Introduction to standard and non-standard Numerical Methods

Introduction to standard and non-standard Numerical Methods Introduction to standard and non-standard Numerical Methods Dr. Mountaga LAM AMS : African Mathematic School 2018 May 23, 2018 One-step methods Runge-Kutta Methods Nonstandard Finite Difference Scheme

More information

4 Stability analysis of finite-difference methods for ODEs

4 Stability analysis of finite-difference methods for ODEs MATH 337, by T. Lakoba, University of Vermont 36 4 Stability analysis of finite-difference methods for ODEs 4.1 Consistency, stability, and convergence of a numerical method; Main Theorem In this Lecture

More information

Exact and Approximate Numbers:

Exact and Approximate Numbers: Eact and Approimate Numbers: The numbers that arise in technical applications are better described as eact numbers because there is not the sort of uncertainty in their values that was described above.

More information

Multistep Methods for IVPs. t 0 < t < T

Multistep Methods for IVPs. t 0 < t < T Multistep Methods for IVPs We are still considering the IVP dy dt = f(t,y) t 0 < t < T y(t 0 ) = y 0 So far we have looked at Euler s method, which was a first order method and Runge Kutta (RK) methods

More information

Ordinary Differential Equations

Ordinary Differential Equations Chapter 7 Ordinary Differential Equations Differential equations are an extremely important tool in various science and engineering disciplines. Laws of nature are most often expressed as different equations.

More information

Southern Methodist University.

Southern Methodist University. Title: Continuous extensions Name: Lawrence F. Shampine 1, Laurent O. Jay 2 Affil./Addr. 1: Department of Mathematics Southern Methodist University Dallas, TX 75275 USA Phone: +1 (972) 690-8439 E-mail:

More information

Chapter 11 ORDINARY DIFFERENTIAL EQUATIONS

Chapter 11 ORDINARY DIFFERENTIAL EQUATIONS Chapter 11 ORDINARY DIFFERENTIAL EQUATIONS The general form of a first order differential equations is = f(x, y) with initial condition y(a) = y a We seek the solution y = y(x) for x > a This is shown

More information

Ordinary Differential Equations

Ordinary Differential Equations Ordinary Differential Equations Professor Dr. E F Toro Laboratory of Applied Mathematics University of Trento, Italy eleuterio.toro@unitn.it http://www.ing.unitn.it/toro September 19, 2014 1 / 55 Motivation

More information

Astrodynamics (AERO0024)

Astrodynamics (AERO0024) Astrodynamics (AERO0024) 5. Numerical Methods Gaëtan Kerschen Space Structures & Systems Lab (S3L) Why Different Propagators? Analytic propagation: Better understanding of the perturbing forces. Useful

More information

FIRST-ORDER ORDINARY DIFFERENTIAL EQUATIONS III: Numerical and More Analytic Methods David Levermore Department of Mathematics University of Maryland

FIRST-ORDER ORDINARY DIFFERENTIAL EQUATIONS III: Numerical and More Analytic Methods David Levermore Department of Mathematics University of Maryland FIRST-ORDER ORDINARY DIFFERENTIAL EQUATIONS III: Numerical and More Analytic Methods David Levermore Department of Mathematics University of Maryland 30 September 0 Because the presentation of this material

More information

2.29 Numerical Fluid Mechanics Fall 2011 Lecture 20

2.29 Numerical Fluid Mechanics Fall 2011 Lecture 20 2.29 Numerical Fluid Mechanics Fall 2011 Lecture 20 REVIEW Lecture 19: Finite Volume Methods Review: Basic elements of a FV scheme and steps to step-up a FV scheme One Dimensional examples d x j x j 1/2

More information

The family of Runge Kutta methods with two intermediate evaluations is defined by

The family of Runge Kutta methods with two intermediate evaluations is defined by AM 205: lecture 13 Last time: Numerical solution of ordinary differential equations Today: Additional ODE methods, boundary value problems Thursday s lecture will be given by Thomas Fai Assignment 3 will

More information

Computational Techniques Prof. Dr. Niket Kaisare Department of Chemical Engineering Indian Institute of Technology, Madras

Computational Techniques Prof. Dr. Niket Kaisare Department of Chemical Engineering Indian Institute of Technology, Madras Computational Techniques Prof. Dr. Niket Kaisare Department of Chemical Engineering Indian Institute of Technology, Madras Module No. # 07 Lecture No. # 05 Ordinary Differential Equations (Refer Slide

More information

INTRODUCTION TO COMPUTER METHODS FOR O.D.E.

INTRODUCTION TO COMPUTER METHODS FOR O.D.E. INTRODUCTION TO COMPUTER METHODS FOR O.D.E. 0. Introduction. The goal of this handout is to introduce some of the ideas behind the basic computer algorithms to approximate solutions to differential equations.

More information

Introduction to Initial Value Problems

Introduction to Initial Value Problems Chapter 2 Introduction to Initial Value Problems The purpose of this chapter is to study the simplest numerical methods for approximating the solution to a first order initial value problem (IVP). Because

More information

Second Order ODEs. CSCC51H- Numerical Approx, Int and ODEs p.130/177

Second Order ODEs. CSCC51H- Numerical Approx, Int and ODEs p.130/177 Second Order ODEs Often physical or biological systems are best described by second or higher-order ODEs. That is, second or higher order derivatives appear in the mathematical model of the system. For

More information

Parallel Methods for ODEs

Parallel Methods for ODEs Parallel Methods for ODEs Levels of parallelism There are a number of levels of parallelism that are possible within a program to numerically solve ODEs. An obvious place to start is with manual code restructuring

More information

Integration of Ordinary Differential Equations

Integration of Ordinary Differential Equations Integration of Ordinary Differential Equations Com S 477/577 Nov 7, 00 1 Introduction The solution of differential equations is an important problem that arises in a host of areas. Many differential equations

More information

Ordinary Differential Equations

Ordinary Differential Equations Ordinary Differential Equations Introduction: first order ODE We are given a function f(t,y) which describes a direction field in the (t,y) plane an initial point (t 0,y 0 ) We want to find a function

More information

1 Error Analysis for Solving IVP

1 Error Analysis for Solving IVP cs412: introduction to numerical analysis 12/9/10 Lecture 25: Numerical Solution of Differential Equations Error Analysis Instructor: Professor Amos Ron Scribes: Yunpeng Li, Mark Cowlishaw, Nathanael Fillmore

More information

Introductory Numerical Analysis

Introductory Numerical Analysis Introductory Numerical Analysis Lecture Notes December 16, 017 Contents 1 Introduction to 1 11 Floating Point Numbers 1 1 Computational Errors 13 Algorithm 3 14 Calculus Review 3 Root Finding 5 1 Bisection

More information

Ordinary Differential Equations

Ordinary Differential Equations Ordinary Differential Equations We call Ordinary Differential Equation (ODE) of nth order in the variable x, a relation of the kind: where L is an operator. If it is a linear operator, we call the equation

More information

Lecture Notes on Numerical Differential Equations: IVP

Lecture Notes on Numerical Differential Equations: IVP Lecture Notes on Numerical Differential Equations: IVP Professor Biswa Nath Datta Department of Mathematical Sciences Northern Illinois University DeKalb, IL. 60115 USA E mail: dattab@math.niu.edu URL:

More information

Validated Explicit and Implicit Runge-Kutta Methods

Validated Explicit and Implicit Runge-Kutta Methods Validated Explicit and Implicit Runge-Kutta Methods Alexandre Chapoutot joint work with Julien Alexandre dit Sandretto and Olivier Mullier U2IS, ENSTA ParisTech 8th Small Workshop on Interval Methods,

More information

PowerPoints organized by Dr. Michael R. Gustafson II, Duke University

PowerPoints organized by Dr. Michael R. Gustafson II, Duke University Part 6 Chapter 20 Initial-Value Problems PowerPoints organized by Dr. Michael R. Gustafson II, Duke University All images copyright The McGraw-Hill Companies, Inc. Permission required for reproduction

More information

Exam in TMA4215 December 7th 2012

Exam in TMA4215 December 7th 2012 Norwegian University of Science and Technology Department of Mathematical Sciences Page of 9 Contact during the exam: Elena Celledoni, tlf. 7359354, cell phone 48238584 Exam in TMA425 December 7th 22 Allowed

More information

The Milne error estimator for stiff problems

The Milne error estimator for stiff problems 13 R. Tshelametse / SAJPAM. Volume 4 (2009) 13-28 The Milne error estimator for stiff problems Ronald Tshelametse Department of Mathematics University of Botswana Private Bag 0022 Gaborone, Botswana. E-mail

More information

Modeling & Simulation 2018 Lecture 12. Simulations

Modeling & Simulation 2018 Lecture 12. Simulations Modeling & Simulation 2018 Lecture 12. Simulations Claudio Altafini Automatic Control, ISY Linköping University, Sweden Summary of lecture 7-11 1 / 32 Models of complex systems physical interconnections,

More information

Numerical Methods for Differential Equations

Numerical Methods for Differential Equations CHAPTER 5 Numerical Methods for Differential Equations In this chapter we will discuss a few of the many numerical methods which can be used to solve initial value problems and one-dimensional boundary

More information

Numerical Analysis of Differential Equations Numerical Solution of Parabolic Equations

Numerical Analysis of Differential Equations Numerical Solution of Parabolic Equations Numerical Analysis of Differential Equations 215 6 Numerical Solution of Parabolic Equations 6 Numerical Solution of Parabolic Equations TU Bergakademie Freiberg, SS 2012 Numerical Analysis of Differential

More information

Runge-Kutta and Collocation Methods Florian Landis

Runge-Kutta and Collocation Methods Florian Landis Runge-Kutta and Collocation Methods Florian Landis Geometrical Numetric Integration p.1 Overview Define Runge-Kutta methods. Introduce collocation methods. Identify collocation methods as Runge-Kutta methods.

More information

The Plan. Initial value problems (ivps) for ordinary differential equations (odes) Review of basic methods You are here: Hamiltonian systems

The Plan. Initial value problems (ivps) for ordinary differential equations (odes) Review of basic methods You are here: Hamiltonian systems Scientific Computing with Case Studies SIAM Press, 2009 http://www.cs.umd.edu/users/oleary/sccswebpage Lecture Notes for Unit V Solution of Differential Equations Part 2 Dianne P. O Leary c 2008 The Plan

More information

9.6 Predictor-Corrector Methods

9.6 Predictor-Corrector Methods SEC. 9.6 PREDICTOR-CORRECTOR METHODS 505 Adams-Bashforth-Moulton Method 9.6 Predictor-Corrector Methods The methods of Euler, Heun, Taylor, and Runge-Kutta are called single-step methods because they use

More information

Numerical Methods for the Solution of Differential Equations

Numerical Methods for the Solution of Differential Equations Numerical Methods for the Solution of Differential Equations Markus Grasmair Vienna, winter term 2011 2012 Analytical Solutions of Ordinary Differential Equations 1. Find the general solution of the differential

More information

16.7 Multistep, Multivalue, and Predictor-Corrector Methods

16.7 Multistep, Multivalue, and Predictor-Corrector Methods 740 Chapter 16. Integration of Ordinary Differential Equations 16.7 Multistep, Multivalue, and Predictor-Corrector Methods The terms multistepand multivaluedescribe two different ways of implementing essentially

More information

ORBIT 14 The propagator. A. Milani et al. Dipmat, Pisa, Italy

ORBIT 14 The propagator. A. Milani et al. Dipmat, Pisa, Italy ORBIT 14 The propagator A. Milani et al. Dipmat, Pisa, Italy Orbit 14 Propagator manual 1 Contents 1 Propagator - Interpolator 2 1.1 Propagation and interpolation............................ 2 1.1.1 Introduction..................................

More information

Chapter 8. Numerical Solution of Ordinary Differential Equations. Module No. 1. Runge-Kutta Methods

Chapter 8. Numerical Solution of Ordinary Differential Equations. Module No. 1. Runge-Kutta Methods Numerical Analysis by Dr. Anita Pal Assistant Professor Department of Mathematics National Institute of Technology Durgapur Durgapur-71309 email: anita.buie@gmail.com 1 . Chapter 8 Numerical Solution of

More information

Research Article Diagonally Implicit Block Backward Differentiation Formulas for Solving Ordinary Differential Equations

Research Article Diagonally Implicit Block Backward Differentiation Formulas for Solving Ordinary Differential Equations International Mathematics and Mathematical Sciences Volume 212, Article ID 767328, 8 pages doi:1.1155/212/767328 Research Article Diagonally Implicit Block Backward Differentiation Formulas for Solving

More information

Simple ODE Solvers - Derivation

Simple ODE Solvers - Derivation Simple ODE Solvers - Derivation These notes provide derivations of some simple algorithms for generating, numerically, approximate solutions to the initial value problem y (t =f ( t, y(t y(t 0 =y 0 Here

More information