Perturbation Theory and Asymptotic Expansions
A graduate student course at the
Dept. of Information Technology, Division of Scientific Computing, given
by visiting professor
Bengt Fornberg, May-June 2004.
Course topics
- Some introductory examples from applications
- Expansion methods in the case of algebraic equations
- Techniques for convergence improvement of truncated expansions
- Approximate solutions of linear and nonlinear ODEs
- Asymptotic expansion of integrals
- Perturbation series for ODEs and PDEs
- Boundary layer theory - matched asymptotic expansions
- WKB theory
- Multiple scale analysis
Course Motivation
The governing equations for most phenomena in nature and in the sciences
can be formulated in terms of PDEs, ODEs, integral equations, or in
combinations of these. The main approaches to obtain insights from such
equations are
Analytic Solutions: This approach alone is virtually never successful. For
all but the most trivial cases, realistic governing equations simply do
not
admit exact solutions in terms of elementary functions. This fundamental
obstacle is not due to any limitations in our ability to perform analytic
manipulations - the use of symbolic algebra packages (like Mathematica)
help only very little.
Numerical Solutions: This general approach is immensely powerful -
large-scale computer simulations are now often considered as the third
fundamental investigative technique (besides the long established ones of
theory and experiment). No other approach can come even close to solving
systems of hundreds or thousands of coupled nonlinear differential
equations that arise in many applications. However, single-minded
number-crunching suffers from notable limitations and difficulties, e.g.
- Coding can be very complex,
- We are often interested in situations where some parameter is very
small
(or very large). In such limits, computer costs often become
prohibitive,
- Numerics is not well suited to get 'leading behaviors' in precise
analytic form, suitable for further analysis.
Perturbation / Asymptotic Analysis: By replacing an analytically
unsolvable
problem with a sequence of analytically solvable ones, one can often avoid
the fundamental barrier that is encountered when searching for exact
solutions (Now, a package like Mathematica becomes extremely useful in
carrying out the difficult and lengthy - but feasible - manipulations
required). Asymptotic methods are usually most powerful precisely when
numerical approaches encounter their most serious difficulties, such as in
cases of small parameters, phenomena on vastly different scales etc.
Perturbation / asymptotic analysis can then provide accurate information
in
analytic forms which are very well suited for both understanding and for
further analysis.
The three general approaches above all complement each other. In most
applications, all three are required.
Schedule
All room numbers refer to building 2, MIC,
Polacksbacken.
Mon | 2004-05-17 | 10:15-12:00 | Room 2145 |
Tue | 2004-05-18 | 10:15-12:00 | Room 2145 |
Wed | 2004-05-19 | 08:15-10:00 | Room 2144 |
Wed | 2004-05-19 | 10:15-12:00 | Room 2145 |
|
Thu | 2004-05-27 | 10:15-12:00 | Room 2145 |
Fri | 2004-05-28 | 10:15-12:00 | Room 2145 |
|
Wed | 2004-06-02 | 10:15-12:00 | Room 2145 |
Thu | 2004-06-03 | 10:15-12:00 | Room 2145 |
Fri | 2004-06-04 | 10:15-12:00 | Room 2145 |
Examination and course credits
The form of the examination is not yet determined. The course will give 3
graduate student credit points. For more information, contact Elisabeth Larsson, at the address below.
Mathematica notebooks
Regular perturbation; polynomial roots
Reg_Pert_Pol_Roots.nb
ODE expansion around regular point
ODE_Exp_Reg_Pt.nb
ODE expansion around irregular singular point
LINDIF01.nb
Padé convergence acceleration for a Stieltjes function
Pade_01.nb
Conversion from Taylor- to continued fraction expansion
Taylor_to_cf.nb
Fourier-Laplace method to get ODE solution in integral form
FourierLaplace.nb
Laplace integral - integration by parts
Laplace_int_part.nb
Laplace's method - higher order terms
Laplace_high_order.nb
Laplace integrals - Watson's lemma
Watson_02.nb
Gamma function expansion using steepest descent
Gamma_As.nb
Perturbation expansions for the projectile problem:
--- By plugging in an assumed series and equating coefficients
Proj_Pert_1.nb
--- By parametric differentiation and by iteration
Hw09_P3.nb
First example on boundary layers
B_Layer_01.nb
Second example on boundary layers
B_Layer_02.nb
Numerical solution of boundary value problems
Num_ODE.nb
Elisabeth Larsson
bette@it.uu.se
Last modified: Sat May 8 09:09:56 MEST 2004