Scribd
Upload
68 views2 pages

MAT260 - Numerical Analysis - Cheat Sheet: Preliminaries

This cheat sheet provides a concise summary of key concepts in numerical analysis: - Explicit multistep methods are not A-stable, while implicit methods can be A-stable. The highest order A-stable multistep method is 2. - Polynomial fitting involves minimizing the residual between a polynomial and a function over an interval using an orthogonal basis. - Multistep methods satisfy order conditions to achieve a given order of accuracy. Stability depends on the roots of the characteristic equation. - Implicit Runge-Kutta methods solve for stage values implicitly. Stability depends on the region where the absolute value of the stability function is less than 1. - Nonlinear root finding includes fixed-point

Uploaded by

Wail Rimouche
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
68 views2 pages

MAT260 - Numerical Analysis - Cheat Sheet: Preliminaries

This cheat sheet provides a concise summary of key concepts in numerical analysis: - Explicit multistep methods are not A-stable, while implicit methods can be A-stable. The highest order A-stable multistep method is 2. - Polynomial fitting involves minimizing the residual between a polynomial and a function over an interval using an orthogonal basis. - Multistep methods satisfy order conditions to achieve a given order of accuracy. Stability depends on the roots of the characteristic equation. - Implicit Runge-Kutta methods solve for stage values implicitly. Stability depends on the region where the absolute value of the stability function is less than 1. - Nonlinear root finding includes fixed-point

Uploaded by

Wail Rimouche
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

MAT260 - Numerical Analysis - Cheat SheetTommy Odland, June 6, 2016

Preliminaries A-stability The highest order of an A-stable multistep method is 2.


No explicit multistep method is A-stable.
Trig sin(u ± v) = sin u cos v ± cos u sin v
Norms kABk ≤ kAk · kBk Runge-Kutta methods
kA + Bk ≤ kAk + kBk
Polynomial
Pn fit If we wish to approximate f (x) by the polynomial Pn (x) =
k=0 ak ϕk (x), where ϕi (x) and ϕj (x) are orthogonal with respect to the
Euler’s method and beyond weight function w(x), such that:
Z b
FWE: yn+1 = yn + hfn , explicit, not A-stable, first order
hϕi (x), ϕj (x)i = ϕi (x)w(x)ϕj (x) dx = 0
FWE: yn+1 = yn + hfn+1 , implicit, A-stable, first order a
Trapezoidal: yn+1 = yn + h2 [fn + fn+1 ], implicit, A-stable, second order then the minimum is obtained by setting:
h
2. order AB: yn+2 = yn+1 + (−fn + 3fn+1 ) explicit, not A-stable, second order
2
hϕk (x), f (x)i
ak =
hϕk (x), ϕk (x)i
Multistep methods IRK An implicit Runge-Kutta scheme is given by:
ν
s s X
ξj = yn + h aj,i f (tn + ci h, ξi )
X X
am yn+m = h bm f (tn+m , yn+m ) (1)
i=1
m=0 m=0
ν
X
Stability Stable if the root criterion is satisfied for LHS of (1): yn+1 = yn + h bj f (tn + cj h, ξj )
(i) All zeros must reside in the closed unit disc, (ii) and all zeros of unit j=1
modulus must be simple(not repeated). RK tableux The RK-tableux is:
c A
Order To obtain order p it is necessary and sufficient that, for k ≥ 1:
bT
s
X
am = 0
Stiff equations
m=0

s
X s
X Stability domain Assuming yn+1 = r(z)yn , the linear stability domain is:
k k−1
m am = k m bm D = {z ∈ C : |r(z)| < 1} (2)
m=0 m=0
s
X s
X Stability domain A method is A-stable if it preserves the decay of y0 = λy when
p+1 p Re λ < 0:
m am 6= (p + 1) m bm
m=0 m=0 C− := {z ∈ C : Re z < 0} ⊆ D
where 00 := 1. where D is defined in equation (2).
Error control A bound
γn
kg(vn ) − g(un )k < v0 − u0

PECE The predictor-corrector scheme is: 1−γ
(i) use explicit method to predict yn+1 , then (ii) use implicit method to
Newtons method Newtons method is:
correct the initial prediction.

Error control If we integrate from 0 < t < T with step size h, and we wish that x1 = x0 − J −1 (x0 )F (x0 )
the total error is below ε, then the single step error must be below εh/T .
Here it is assumed that errors do not accumulate. Finite difference schemes
Gershgorin Every eigenvalue of A lies within at least one ofPthe Gershgorin discs
Nonlinear algebraic equations D(aii , Ri ), where aii is a diagonal element and Ri = j6=i |aij |.
Fix-point theorem The iteration uk = g(uk−1 ) is a contraction mapping if: SOCD Second order central difference approximation:

g(uk ) − g(uk−1 ) < γ uk − uk−1


1
u00 (x) ≈ [u(x + h) − 2u(x) + u(x − h)] + O(h2 )
h2
for some real γ < 1. For more information, please see:
k+1 k 0 k (1) Iserles, Arieh. Numerical Analysis of Differential Equations. 2nd edition. Cambridge ; New
Convergence The fix point iteration u = g(u ) converges if |g (u )| < 1 for all York: Cambridge University Press, 2008.
relevant u. (2) Richard L. Burden. Numerical Analysis. 3rd ed. Prindle, Weber & Schmidt, 1985.

You might also like