Introduction

Regular Perturbation Method

Nonlinear Oscillations

Math 537 - Ordinary Differential Equations

Lecture Notes – Perturbation Methods

Joseph M. Mahaffy,
hjmahaffy@sdsu.edui

Department of Mathematics and Statistics

Dynamical Systems Group
Computational Sciences Research Center
San Diego State University
San Diego, CA 92182-7720
http://jmahaffy.sdsu.edu

Fall 2019

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (1/47)
 Introduction
Regular Perturbation Method
Nonlinear Oscillations

Outline

1 Introduction
Power Series Review
Algebraic Examples
Error Function

2 Regular Perturbation Method

Motion in Resistive Medium
Kepler’s Laws
Precession of Perihelion

3 Nonlinear Oscillations
Duffing’s Equation
Poincaré-Lindstedt Method

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (2/47)
 Introduction Power Series Review
Regular Perturbation Method Algebraic Examples
Nonlinear Oscillations Error Function

Introduction

Mathematical Models rarely have an explicit form.

Require approximation and numerical methods.

Approximations often use perturbation methods:
The equations have a small term.
One physical process is significantly less important than
another dominant one.
Often rescale problem with a small parameter.
Use the small parameter to create a Taylor-like series
expansion.
These methods can be applied to ODEs, PDEs, algebraic
equations, and integral equations.

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (3/47)
 Introduction Power Series Review
Regular Perturbation Method Algebraic Examples
Nonlinear Oscillations Error Function

Regular Perturbation 1

Regular Perturbation: For conceptual purposes an ODE is used to

describe the approach:

F (t, y, y 0 , y 00 , ε) = 0, t ∈ I,

where t is the independent variable, defined in the interval I, and y is

the dependent variable.
The parameter, ε, explicitly appears in the equation and is generally
considered “small” with
ε  1.
It may also occur in the initial or boundary conditions.
Sometimes a parameter, λ, is “large,” then often introduce
1
ε= λ  1.

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (4/47)
 Introduction Power Series Review
Regular Perturbation Method Algebraic Examples
Nonlinear Oscillations Error Function

Regular Perturbation 2

Perturbation Series: Find an ε-power series of the solution to

the problem with

y(t) = y0 (t) + εy1 (t) + ε2 y2 (t) + . . .

The regular perturbation method assumes a solution to the ODE

in this form, where the functions y0 , y1 , y2 , . . . are found by
substituting into the ODE.
The first few terms of the ε-power series form an approximate
solution, called the perturbation solution or approximation.
Usually only a few terms are necessary.
The method is considered successful if the approximation is
uniform: i.e., the difference between the approximate and exact
solutions converges to zero at some defined rate as ε → 0, uniformly
on I (with ε < ε0 for some ε0 ).
Lecture Notes – Perturbation Methods —
Joseph M. Mahaffy, hjmahaffy@sdsu.edui (5/47)
 Introduction Power Series Review
Regular Perturbation Method Algebraic Examples
Nonlinear Oscillations Error Function

Regular Perturbation 3

Regular Perturbation: The dominant behavior comes from the

term, y0 (t), the leading order term, solving the unperturbed
problem:
F (t, y0 , y00 , y000 , 0) = 0, t ∈ I,
where this problem is chosen to be solvable.
The term εy1 (t), ε2 y2 (t), . . . are considered higher order correction
terms and are assumed to be small relative to the dominant behavior.
Singular Perturbation methods arise when the regular
perturbation methods fail.
The naive approach often fails for many reasons such as the problem
being ill-posed, the solution is invalid on all or parts of the domain,
like when there are multiple time or space scales.

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (6/47)
 Introduction Power Series Review
Regular Perturbation Method Algebraic Examples
Nonlinear Oscillations Error Function

Power Series Review

Power Series Review: Some important power series are listed.

f (x) = f (a) + f 0 (a)(x − a) + 1

2!
(x − a)2 + . . .

ex = 1+x+ 1 2
2!
x + 1 3
3!
x + ...
1 3 1 5 1 7
sin(x) = x− 3!
x + 5!
x − 7!
x + ...
1 2 1 4 1 6
cos(x) = 1− 2!
x + 4!
x − 6!
x + ...
p(p−1) 2 p(p−1)(p−2) 3
(1 + x)p = 1 + px + 2!
x + 3!
x + ..., |x| < 1

ln(1 + x) = x − 12 x2 + 13 x3 − 14 x4 + . . . , |x| < 1

p(p−1) p−2 2 p(p−1)(p−2) p−3 3
(a + x)p ap + pap−1 x +
x
= 2!
a x + 3!
a x + ...,  <1
a

These power series are commonly used for asymptotic expansions.

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (7/47)
 Introduction Power Series Review
Regular Perturbation Method Algebraic Examples
Nonlinear Oscillations Error Function

Quadratic Equation 1
Quadratic Equation: Consider the equation:
x2 + 2εx − 3 = 0, (1)
and assume the expansion x = x0 + εx1 + ε2 x 2 + ...
The Eqn. (1) satisfies:
(x0 + εx1 + ε2 x2 + . . . )2 + 2ε(x0 + εx1 + ε2 x2 + . . . ) − 3 = 0,
which gives
x20 − 3 + ε 2x0 x1 + 2x0 + ε2 x21 + 2x0 x2 + 2x1 + O ε3 = 0.

Solving each power of ε gives:

√ 1
x0 = ± 3, x1 = −1, x2 = ± √ , ...
2 3

The approximate solutions are

√ 2
x = 3 − ε + ε√ + . . . ,
2 3
√ 2
x = − 3 − ε − ε√ + . . . .
2 3

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (8/47)
 Introduction Power Series Review
Regular Perturbation Method Algebraic Examples
Nonlinear Oscillations Error Function

Quadratic Equation 2
Quadratic Equation; For the equation
x2 + 2εx − 3 = 0,
the quadratic formula gives exact solution;
p
x = −ε ± 3 + ε2 ,
which can be expanded by the binomial expansion
√ 2
x = −ε ± 3 ± ε√ + . . . .
2 3

, agreeing with our asymptotic expansion.

If we take ε = 0.1, then the exact and approximate solutions, are:
√ √
x = −0.1 + 3.01 = 1.634935157, xa = 3 − 0.1 + 0.01
√ = 1.634937560,
2 3

and
√ √ 0.01
x = −0.1 − 3.01 = −1.834935157, xa = 3 − 0.1 + √ = −1.834937560.
2 3

With ε = 0.01, the answers are indistinguishable until the 9th decimal place with
√
x = −0.01 ± 3.0001 = 1.722079675 and − 1.742079675.
Lecture Notes – Perturbation Methods —
Joseph M. Mahaffy, hjmahaffy@sdsu.edui (9/47)
 Introduction Power Series Review
Regular Perturbation Method Algebraic Examples
Nonlinear Oscillations Error Function

Transcendental Equation 1

Transcendental Equation: Consider the equation given by

x3 + ε sin(x) + a = 0,

which clearly cannot be solved exactly for x.

Take x = x0 + εx1 + ε2 x2 + . . . and find the solution to order ε2 .
Note that

sin(x) = sin(x0 + εx1 + ε2 x2 + . . . )

= sin(x0 ) + cos(x0 ) εx1 + ε2 x2 + . . . ) + . . .
sin(x0 ) + x1 cos(x0 )ε + O ε2 ,


=

so the equation becomes:

3
x0 + εx1 + ε2 x2 + O ε3 + ε sin(x0 ) + x1 cos(x0 )ε + O ε2



+ a = 0.

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (10/47)
 Introduction Power Series Review
Regular Perturbation Method Algebraic Examples
Nonlinear Oscillations Error Function

Transcendental Equation 2
Transcendental Equation: From


3
x0 + εx1 + ε2 x2 + O ε3 + ε sin(x0 ) + x1 cos(x0 )ε + O ε2



+ a = 0,

we expand to:

x30 + a + 3x20 x1 + sin(x0 ) ε + 3x20 x2 + 3x0 x21 + x1 cos(x0 ) ε2 + O ε3 = 0.

Solving iteratively gives:

1 sin(x0 ) 3x0 x21 + x1 cos(x0 )

x0 = −a 3 , x1 = − , x2 = − .
3x20 3x20

For a = 5, we have x0 = −1.709975947, x1 = 0.112896048, and

x2 = 0.009239083853.
With ε = 0.1, then a 3 term expansion of x gives:

x = x0 + 0.1x1 + 0.01x2 = −1.698593951.

Maple gives a numerical solution to the original equation as x = −1.698593473.

Lecture Notes – Perturbation Methods —
Joseph M. Mahaffy, hjmahaffy@sdsu.edui (11/47)
 Introduction Power Series Review
Regular Perturbation Method Algebraic Examples
Nonlinear Oscillations Error Function

Error Function 1
Error Function: The error function satisfies:
Z x Z x
1 2 2 2
erf(x) = √ e−t dt = √ e−t dt.
π −x π 0
In statistics with x > 0 and a random variable Y that is normally distributed
with mean 0 and variance 0.5, the error function, erf(x), describes the
probability of Y falling in the range [x, x].

0.5

-0.5

-1

-3 -2 -1 0 1 2 3
Lecture Notes – Perturbation Methods —
Joseph M. Mahaffy, hjmahaffy@sdsu.edui (12/47)
 Introduction Power Series Review
Regular Perturbation Method Algebraic Examples
Nonlinear Oscillations Error Function

Error Function 2

Since
lim erf(x) = 1,
x→∞

the complementary error function is given by:

2
Z ∞ 2
erfc(x) ≡ 1 − erf(x) = √ e−t dt.
π x

Let s = t − x, then

2
Z ∞ 2 2 2
Z ∞ 2
erfc(x) = √ e−(s+x) ds = √ e−x e−(s +2sx)
ds.
π 0 π 0

As a first order approximation, we see for x > 0

2 2
Z ∞ 2 2 2
Z ∞ 2 2
erfc(x) = √ e−x e−(s +2sx)
ds ≤ e−x √ e−s ds = e−x .
π 0 π 0

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (13/47)
 Introduction Power Series Review
Regular Perturbation Method Algebraic Examples
Nonlinear Oscillations Error Function

Error Function 3

Complementary Error Function is better approximated using integration by

parts:

2
Z ∞ 2 2
Z ∞ 2
erfc(x) = √2 e−x e−(s +2sx)
ds = √2 e−x e−s e−2sx ds
π π
0 0
 ∞ Z ∞ 
2 1 2 1 2
√2 e−x − e−s e−2sx  + (−2s)e−s e−2sx ds

= π 2x 0 2x 0
 Z ∞ 
2 1 1 2
= √2 e−x − se−s e−2sx ds
π 2x x 0
" 2
∞ Z ∞ !#
2 1 1 se−s e−2sx  1 2
= √2 e−x − −  + (1 − 2s2 )e−s e−2sx ds
π 2x x 2x  2x 0
0

2

1 1
Z ∞ 2

= √2 e−x − (1 − 2s2 )e−s e−2sx ds
π 2x 2x2 0

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (14/47)
 Introduction Power Series Review
Regular Perturbation Method Algebraic Examples
Nonlinear Oscillations Error Function

Error Function 4

Complementary Error Function is approximated for large x by the following:

2 2
!
e−x e−x
erfc(x) = √1 +O as x → ∞.
π x x2

1 2 3 4 5
erf(x) 0.8427 0.9953 0.9999779 0.9999999846 1.0
Approx 0.7924 0.9948 0.9999768 0.9999999841 1.0

A useful asymptotic expansion given by Wikipedia is

2 ∞
" #
e−x X 1 · 3 · 5 · · · (2n − 1)
erfc(x) = √ 1+ (−1)n ,
x π n=1
(2x2 )n

which agrees with our computation above.

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (15/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Motion in Resistive Medium 1

Resistive ODE: A body mass m with initial velocity V0 moves in a straight line,
where the resistive force has magnitude av − bv 2 with v(τ ) being the velocity of
the object.
It is assumed that b  a are constants, then Newton’s Law gives:

dv
m dτ = −av + bv 2 , v(0) = V0 .

With the maximum velocity of V0 and a time scaling based on the decay rate of
the linear equation, a/m, we let

v τ
y= V0
and t= m/a
.

This change of variables creates the dimensionless problem with t > 0:

dy
dt
= −y + εy 2 , y(0) = 1,

where
bV0
ε≡ a
 1.
This last assumption is that the quadratic resistive force is small compared to the
linear force.
Lecture Notes – Perturbation Methods —
Joseph M. Mahaffy, hjmahaffy@sdsu.edui (16/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Motion in Resistive Medium 2

Resistive ODE: The scaled model is a Bernoulli’s equation and is readily solved
exactly.
One makes the substitution w = y −1 , so dw
dt
= −y −2 dy
dt
, so transformed the scaled
model becomes:
dw
dt
− w = −ε, w(0) = 1.
This is a linear ODE with the solution:
 Z t 
w(t) = et 1−ε e−s ds ,
0

or
w(t) = et 1 + ε(e−t − 1) ,

or
e−t
y(t) = ,
1 + ε(e−t − 1)
which is just a slightly altered form of the linearized scaled model.

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (17/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Motion in Resistive Medium 3

Perturbation Method: For this solution we assume a solution of the form:

y(t) = y0 (t) + εy1 (t) + ε2 y2 (t) + . . . ,

and substitute this into the scaled model giving:

y00 + εy10 + ε2 y20 + · · · = −(y0 + εy1 + ε2 y2 + . . . ) + ε(y0 + εy1 + ε2 y2 + . . . )2 .

Collecting powers of ε yields a series of linear ODEs:

y00 = −y0 ,
y10 = −y1 + y02 ,
y20 = −y2 + 2y0 y1 , ...

The initial condition satisfies:

y0 (0) + εy1 (0) + ε2 y2 (0) + · · · = 1,

which gives the sequence of initial conditions:

y0 (0) = 1, y1 (0) = y2 (0) = · · · = 0.

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (18/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Motion in Resistive Medium 4

Perturbation Method: The sequence of linear ODEs are easily solved to give:

y0 (t) = e−t ,
y1 (t) = e−t − e−2t ,
y2 (t) = e−t − e−2t + e−3t , ...

It follows that the approximate solution to O ε2 satisfies:

ya (t) = e−t + ε e−t − e−2t + ε2 e−t − e−2t + e−3t .

Recall that the exact solution is

e−t
y(t) = ,
1 + ε(e−t − 1)

which has a Taylor series expansion in ε of

y(t) = e−t + ε e−t − e−2t + ε2 e−t − e−2t + e−3t + O ε3 ,

agreeing with the solution from the perturbation method.

Note: It is rare that one can obtain the exact solution for comparison.
Lecture Notes – Perturbation Methods —
Joseph M. Mahaffy, hjmahaffy@sdsu.edui (19/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Kepler’s Laws 1

Two Body Problem: Consider the motion of a planet in the solar

system.
Planet has mass m, and Sun has mass M .
The position is r(t).
Newton’s Law gives r 00 = F.
Newton postulated that the gravitational force is proportional
to the product of the masses and inversely proportional to the
square of the distance between the masses.
In vector form the Gravitational force satisfies
GM m
F=− ir ,
r2
where ir is a unit vector in the radial direction and
G = 6.67 × 10−8 cm/g · s2 .
Lecture Notes – Perturbation Methods —
Joseph M. Mahaffy, hjmahaffy@sdsu.edui (20/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Kepler’s Laws 2
Two Body Problem: From Newton’s Law, we have the vector ODE
GM m
mr 00 = − ir .
r2

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (21/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Kepler’s Laws 3
Two Body Problem: The vector ODE satisfies:

GM
r 00 = − ir ,
r2

where
r(t) = r(t) cos(θ(t))ix + r(t) sin(θ(t))iy = x(t)ix + y(t)iy .
Thus,

dx dr dθ
= cos(θ) − r sin(θ) ,
dt dt dt
dy dr dθ
= sin(θ) + r cos(θ) .
dt dt dt

It follows from the vector ODE that

d2 x d2 r d2 θ
2
dr dθ dθ GM

= cos(θ) − 2 sin(θ) − r sin(θ) − r cos(θ) = − cos(θ),
dt2 dt2 dt dt dt2 dt r2
2 2 2 2
d y d r dr dθ d θ dθ GM

= sin(θ) + 2 cos(θ) + r cos(θ) − r sin(θ) = − sin(θ).
dt2 dt2 dt dt dt2 dt r2

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (22/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Kepler’s Laws 4

Two Body Problem: Multiplying the x 00 equation by cos(θ) and the y 00

equation by sin(θ) and adding the results gives the nonlinear ODE:
2
d2 r

dθ GM
−r =− .
dt2 dt r2

Similarly, multiplying the x 00 equation by sin(θ) and the y 00 equation by cos(θ)

and subtracting the results gives the nonlinear ODE:

d2 θ dr dθ
r +2 = 0.
dt2 dt dt

Thus, we have two nonlinear coupled 2nd order ODEs, which with the initial
conditions (4 of them):

r(0) = r0 = r0 ir0 and r 0 (0) = v0 = vr0 ir0 + vθ0 iθ0 ,

provide a unique solution for r(t) and θ(t), describing the motion of the
planet for t > 0.
Lecture Notes – Perturbation Methods —
Joseph M. Mahaffy, hjmahaffy@sdsu.edui (23/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Kepler’s 2nd Law 1

Kepler’s 2nd Law: Take the second nonlinear ODE:

d2 θ dr dθ
r +2 = 0.
dt2 dt dt

and multiply by r, which gives:

d2 θ dr dθ d(r2 θ 0 )
r2 2
+ 2r = = 0.
dt dt dt dt

This is integrated to give

dθ p0
r2 = ,
dt m
where p0 is a constant depending on the initial position and velocity.
The quantity mr2 θ 0 is the angular momentum, and this result shows that
angular momentum of the planet is conserved (same for all time).

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (24/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Kepler’s 2nd Law 2

Kepler’s 2nd Law: Suppose at t1 the planet is at P1 = (r1 , θ1 ) and at t2 the
planet is at P2 = (r2 , θ2 ).
With ∆t = t2 − t1 , the position vector sweeps a sector with area:
" # Z θ2 Z r(θ) Z θ2 Z t2
∆A = Area = r dr dθ = 1
2
r2 dθ = 1
2
r2 θ 0 dt,
θ1 0 θ1 t1

where r(θ) is the plane curve traced by the planet’s orbit.

From the conservation of angular momentum (integrand above being
constant), it follows that
p0
∆A = ∆t.
2m

Property (Kepler’s 2nd Law)

The position vector of the planet’s orbit sweeps out sectors of equal area in equal
time intervals.

The integral for ∆A gives more by relating the area with the initial angular
momentum per unit mass of the planet.
Lecture Notes – Perturbation Methods —
Joseph M. Mahaffy, hjmahaffy@sdsu.edui (25/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Kepler’s 1st and 3rd Laws 1

Kepler’s 1st and 3rd Laws: From the conservation of angular momentum
with
dθ p0 dθ p0
r2 = or = ,
dt m dt mr2
this is substituted into the other nonlinear ODE giving:

d2 r p2 1 GM
2
− 02 3 = − 2 ,
dt m r r

which is an autonomous nonlinear ODE that can be solved for r(t), but results
in a horrendous solution.
The shape of the planet’s orbit is obtained by examining r(θ).
Differentiating and using the conservation of angular momentum gives
 
dr dr dθ p0 dr p0 d 1
= = =− ,
dt dθ dt mr2 dθ m dθ r
so
d2 r p0 d2 p2 d2
   
1 dθ 1
2
=− = − 20 2 2 .
dt m dθ2 r dt m r dθ r
Lecture Notes – Perturbation Methods —
Joseph M. Mahaffy, hjmahaffy@sdsu.edui (26/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Kepler’s 1st and 3rd Laws 3

Kepler’s 1st Law: It follows that
d2 GM m2
 
1 1
+ = .
dθ 2 r r p20

This 2nd order linear ODE has the solution:

1 GM m2
= A cos(θ) + B sin(θ) + .
r p20
dr
With the initial conditions r(θ0 ) = r0 and dθ (θ0 ) = r0 vr0 /vθ0 , then for vr0 = 0
the solution satisfies:
2
a(1 − e )
r(θ) = ,
1 + e cos(θ − θ0 )
where
p20 p20
= a(1 − e2 ) and = 1 + e.
GM m2 GM m2 r0
For 0 < e < 1 (eccentricity of the ellipse) the expression for r(θ) gives:

Property (Kepler’s 1st Law)

The orbit of the planet around the sun is an ellipse.
Lecture Notes – Perturbation Methods —
Joseph M. Mahaffy, hjmahaffy@sdsu.edui (27/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Kepler’s 1st and 3rd Laws 3

Kepler’s 3rd Law: Let T be the time for the planet to complete one revolution
of its orbit, so
√
p0 T p 2mπa2 1 − e2
= area of the ellipse = πa2 1 − e2 or T = .
2m p0

However, p p0
1 − e2 = √ ,
m GM a
so
4π 2 a3
T2 = .
GM
This gives

Property (Kepler’s 3rd Law)

The square of the orbital period is proportional to the third power of the length of
the semimajor axis of the elliptical orbit.

Our derivation using Newton’s law gives a precise prediction of the period for a
planet orbiting the sun.
Lecture Notes – Perturbation Methods —
Joseph M. Mahaffy, hjmahaffy@sdsu.edui (28/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Precession of Perihelion 1

Precession of Perihelion: Kepler’s 1st Law states that the orbit

of a planet is elliptical.

The point of the orbit closest to the sun is called perihelion,

while the furthest point is called aphelion.
Astronomical data show that the perihelion of Mercury
advances about 574 seconds of arc/century ( π2 radians
= 3.24 × 105 seconds of arc).
Newton’s law predicts no advance.
The multi-body problem using Newton’s law can account for
about 531 seconds of arc/century.
This leaves about 43 seconds of arc/century, which required
theory of general relativity because of the speed of the planet
Mercury and the accumulation of small effects over time.

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (29/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Precession of Perihelion 2

Relativistic Effects: Dynamics of planetary motion are complex.

The sun moves, which has a complex effect on the observed

motion of the planets.
This is from complex effects of the relative motion of two moving
frames of reference.
Relativistic effects are small except for:
When motion is comparable to the speed of light.
Time allows the accumulation of many small effects.
We examine relativistic effects by studying the simplified
two-body problem.
The initial approximation uses the Newtonian ODE derived
for studying Kepler’s Laws to which relativistic effects are
added.
Lecture Notes – Perturbation Methods —
Joseph M. Mahaffy, hjmahaffy@sdsu.edui (30/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Precession of Perihelion 3

Relativistic Effects: Let r(t) be the radial distance between the sun and the
planet.
Define the dimensionless variable:

a(1 − e2 ) p20
ω0 = , with a(1 − e2 ) = ,
r GM m2

then our Newtonian equation from before

d2 GM m2
 
1 1
+ = .
dθ2 r r p20

becomes
d2 ω0
+ ω0 = 1,
dθ2
which has the solution:
ω0 (θ) = 1 + e cos(θ − θ0 ).

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (31/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Precession of Perihelion 4
For the theory with the relativistic effects, we again define the dimensionless
a(1−e2 )
variable, ω = r
.
The relativistic equation requires a working knowledge of tensor calculus, so we
omit derivation of the equations of motion.
It can be shown that the angular momentum is still conserved and ω(θ) satisfies
the equation:

d2 ω GM m 2
 
2
+ ω = 1 + εω 2 , where ε = 3 ,
dθ p0 c
with c being the speed of light.
p2
From before we have r ≈ a(1 − e2 ) = GM0m2 , so the angular velocity is
approximated by:
dθ p0 GM m
r = ≈ .
dt mr p0
√
From the definition of ε, we find ε is a ratio of the planet’s speed to the speed of
light, which for Mercury satisfies:
ε = O 10−9 .

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (32/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Precession of Perihelion 5
Regular Perturbation: Let

ω(θ) = ω0 (θ) + εω1 (θ) + O ε2 ,

and consider the ODE in ω with our perturbation method:

d2 ω
+ ω = 1 + εω 2 .
dθ2
Thus,


2
ω000 + εω100 + ω0 + εω1 + O ε2 = 1 + ε ω0 + εω1 + O ε2


.

The zeroth order terms give:

ω000 + ω0 = 1,

which from the initial value problem before has the solution:

ω0 (θ) = 1 + e cos(θ − θ0 ).

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (33/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Precession of Perihelion 6

The ε-order terms give the ODE:

2
e2 e2

ω100 + ω1 = ω02 = 1 + e cos(θ − θ0 ) = 1+ + cos(2(θ − θ0 )) + 2e cos(θ − θ0 ).
2 2

This is solved using the method of undetermined coefficients yielding:

e2 e2
ω1 (θ) = A cos(θ) + B sin(θ) + 1 + − cos(2(θ − θ0 )) + eθ sin(θ − θ0 ).
2 6

The homogeneous part and first 3 terms of the particular solution are bounded
solutions, which remain very small when multiplied by ε.
However the last term comes from resonance in this ODE and results in an
unbounded solution (from the θ factor), so results in the dominant behavior over
time due to relativistic effects.

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (34/47)
 Introduction Motion in Resistive Medium
Regular Perturbation Method Kepler’s Laws
Nonlinear Oscillations Precession of Perihelion

Precession of Perihelion 7
Asymptotic Behavior: The resonance term is the only significant one in the
ω1 (θ) solution, so the asymptotic solution is
 
ω(θ) ≈ 1 + e cos(θ − θ0 ) + ε(θ − θ0 ) sin(θ − θ0 )
q
= 1 + e 1 + (εθ̄)2 cos(θ̄ − φ), with θ̄ = θ − θ0 ,

where φ = arctan(εθ̄) ≈ εθ̄ for |εθ̄|  1.

The phase angle φ varies with θ, (and thus, time).
From this the perihelion of the orbit advances by an amount approximately equal
to 2πε = 4.9 × 10−7 radian for each revolution of Mercury around the sun, or the
perihelion precesses.
With an 88-day revolution, Mercury goes through 415 revolutions/century.

Property
By the relativistic theory, the perihelion of Mercury’s orbit precesses by 43
seconds of arc for each century.

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (35/47)
 Introduction
Duffing’s Equation
Regular Perturbation Method
Poincaré-Lindstedt Method
Nonlinear Oscillations

Spring-Mass Oscillators
Spring-Mass System: Consider a mass, m, connected to a nonlinear spring
with restoring force ky + ay 3 with y being the displacement from equilibrium.
Newton’s second law gives:

d2 y
m = −ky − ay 3 , τ > 0,
dτ 2
and initial conditions:
dy
y(0) = A, (0) = 0.
dτ
This problem cannot be solved exactly, but for a  k,it suggests a perturbation
method.
From the initial conditions, we scale y by the amplitude A.
For the scaling of time, we ignore the cubic term and examine the
pODE,
my 00 + ky = 0, which has periodic solutions with a frequency of k/m or period
p
T = 2π m/k.
This suggests a scaling of
τ y
t= p and u= .
m/k A

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (36/47)
 Introduction
Duffing’s Equation
Regular Perturbation Method
Poincaré-Lindstedt Method
Nonlinear Oscillations

Duffing’s Equation 1

Duffing’s Equation: With the previous scaling, the nonlinear mass-spring

equation becomes:

ü + u + εu3 = 0, u(0) = 1, u̇(0) = 0, t > 0,

with the “small” dimensionless parameter (assuming aA2  k):

aA2
ε≡  1.
k

Perturbation method suggests a solution of the form:

u(t) = u0 (t) + εu1 (t) + ε2 u2 (t) + . . .

This is inserted into the equation above to give:

ü0 + εü1 + ε2 ü2 + · · · + u0 + εu1 + ε2 u2 (t) + · · · + ε(u0 + εu1 + ε2 u2 (t) + . . . )3 = 0.

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (37/47)
 Introduction
Duffing’s Equation
Regular Perturbation Method
Poincaré-Lindstedt Method
Nonlinear Oscillations

Duffing’s Equation 2
Duffing’s Equation: From IVP above, we obtain the following sequence of IVPs

ü0 + u0 = 0, u0 (0) = 1, u̇0 (0) = 0,

ü1 + u1 = −u30 , u1 (0) = 0, u̇1 (0) = 0, . . .

The first IVP is easily solved:

u0 (t) = cos(t).

To solve the second IVP, we use the trig identity cos(3t) = 4 cos3 (t) − 3 cos(t), so

ü1 + u1 = − 41 3 cos(t) + cos(3t) ,

with u1 (0) = 0, u̇1 (0) = 0.

Once again, the method of undetermined coefficients is employed to yield the

solution:
1
cos(3t) − cos(t) − 38 t sin(t).


u1 (t) = 32
The last term (resonance again) is clearly unbounded for large t.
This term is called a secular term and is inconsistent with the physical problem.
Higher order approximations will contain secular terms and not cancel out this
approximation.
Lecture Notes – Perturbation Methods —
Joseph M. Mahaffy, hjmahaffy@sdsu.edui (38/47)
 Introduction
Duffing’s Equation
Regular Perturbation Method
Poincaré-Lindstedt Method
Nonlinear Oscillations

Duffing’s Equation 3
Duffing’s Equation is given by:
3
ü + u + εu = 0, u(0) = 1, u̇(0) = 0, t > 0,
with approximate solutions:
u0 (t) = cos(t).
and  
u0 (t) + εu1 (t) = cos(t) + ε 1 cos(3t) − cos(t) − 3


t sin(t) .
32 8

Below is a Phase Portrait of v(t) vs u(t), where v = u̇.

Duffing’s Equation

1
ε = 0.2

0.5

0
v

-0.5

Exact, ue
Approx u0
-1 Approx u1

-1 -0.5 0 0.5
Lecture 1 – Perturbation
Notes 1.5 Methods —
Joseph M. Mahaffy, hjmahaffy@sdsu.edui u(39/47)
 Introduction
Duffing’s Equation
Regular Perturbation Method
Poincaré-Lindstedt Method
Nonlinear Oscillations

Duffing’s Equation 4

1 function du = duffDE(t,u)
2 % Duffing's ODE
3 ep = 0.2;
4 du1 = u(2);
5 du2 = -u(1) - ep*(u(1))ˆ3;
6 du = [du1;du2];
7 end

1 mytitle = 'Duffing''s Equation'; % Title

2 xlab = '$u$'; % X-label
3 ylab = '$v$'; % Y-label
4 tt = linspace(0,2*pi,500);
5 [te,ue] = ode23(@duffDE,tt,[1;0]);
6 ep = 0.2;
7 u0 = cos(tt);
8 v0 = -sin(tt);

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (40/47)
 Introduction
Duffing’s Equation
Regular Perturbation Method
Poincaré-Lindstedt Method
Nonlinear Oscillations

Duffing’s Equation 5

9 u1 = cos(tt) + ep*((1/32)*(cos(3*tt)-cos(tt))...
10 -(3/8)*tt.*sin(tt));
11 v1 = -sin(tt) + ep*((1/32)*(-3*sin(3*tt)+sin(tt))...
12 -(3/8)*(tt.*cos(tt)+sin(tt)));
13 plot(ue(:,1),ue(:,2),'b-','LineWidth',1.5);
14 hold on
15 plot(u0,v0,'r-','LineWidth',1.5);
16 plot(u1,v1,'m-','LineWidth',1.5);
17 grid % Adds Gridlines
18 h =legend('Exact, $u e$', 'Approx $u 0$',...
19 'Approx $u 1$', 'Location','southeast');
20 set(h,'Interpreter','latex')
21 text(-0.2,0.8,'$\varepsilon = 0.2$','FontSize',14,...
22 'interpreter','latex');
23 xlim([-1.2 1.5]);
24 ylim([-1.2 1.2]);

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (41/47)
 Introduction
Duffing’s Equation
Regular Perturbation Method
Poincaré-Lindstedt Method
Nonlinear Oscillations

Duffing’s Equation 5

25 fontlabs = 'Times New Roman'; % Font in labels

26 xlabel(xlab,'FontSize',14,'FontName',fontlabs,...
27 'interpreter','latex');
28 ylabel(ylab,'FontSize',14,'FontName',fontlabs,...
29 'interpreter','latex');
30 title(mytitle,'FontSize',16,'FontName',...
31 'Times New Roman','interpreter','latex');
32 set(gca,'FontSize',12); % Axis tick font size
33 print -depsc duff plot.eps % Create figure ...
as EPS file

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (42/47)
 Introduction
Duffing’s Equation
Regular Perturbation Method
Poincaré-Lindstedt Method
Nonlinear Oscillations

Duffing’s Equation Revisited

Duffing’s Equation Revisited: The regular perturbation method applied to
the equation:
ü + u + εu3 = 0, u(0) = 1, u̇(0) = 0, t > 0,
resulted in a secular term due to resonance in the solution, giving an unbounded
solution.
The regular perturbation method fails to correct for variations in the period,
which accumulate over time leading to solutions out-of-phase.

0.5

-0.5

-1

-1.5
0 5 10 15 20 25

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (43/47)
 Introduction
Duffing’s Equation
Regular Perturbation Method
Poincaré-Lindstedt Method
Nonlinear Oscillations

Poincaré-Lindstedt Method 1

Poincaré-Lindstedt Method: With Duffing’s equation a perturbation is

introduced into the time scale.
Specifically, let
u(τ ) = u0 (τ ) + εu1 (τ ) + ε2 u2 (τ ) + . . . ,
where
τ = ωt with ω = 1 + εω1 + ε2 ω2 + . . . ,
so ω0 = 1, matching the frequency of the unperturbed problem.
This scaling changes Duffing’s equation (with 0 differentiation with respect to τ )
to
ω 2 u00 + u + εu3 = 0, u(0) = 1, u0 (0) = 0, τ > 0,
Substitution gives:

(1 + εω1 + . . . )2 (u00 00 3
0 + εu1 + . . . ) + (u0 + εu1 + . . . ) + ε(u0 + εu1 + . . . ) = 0,

and
u0 (0) + εu1 (0) + · · · = 1, u00 (0) + εu01 (0) + · · · = 0.

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (44/47)
 Introduction
Duffing’s Equation
Regular Perturbation Method
Poincaré-Lindstedt Method
Nonlinear Oscillations

Poincaré-Lindstedt Method 2

Poincaré-Lindstedt Method: Collect coefficients of the powers of ε to give the

following differential equations:

u00
0 + u0 = 0, u0 (0) = 1, u00 (0) = 0,

and
u00 00 3
1 + u1 = −2ω1 u0 − u0 , u1 (0) = u01 (0) = 0, . . .
The solution to the first equation is:

u0 (τ ) = cos(τ ),

so the second DE equation becomes:

u00 3 3 1


1 + u1 = 2ω1 cos(τ ) − cos (τ ) = 2ω1 − 4
cos(τ ) − 4
cos(3τ ).

Since cos(τ ) is a solution to the homogeneous equation, this term on the right
side leads to a particular solution, leading to a secular term of the form τ cos(τ ).
3
This term is eliminated by taking ω1 = 8
.

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (45/47)
 Introduction
Duffing’s Equation
Regular Perturbation Method
Poincaré-Lindstedt Method
Nonlinear Oscillations

Poincaré-Lindstedt Method 3

3
Poincaré-Lindstedt Method: With ω1 = 8
, the second DE equation becomes:

u00 1
1 + u1 = − 4 cos(3τ ), u1 (0) = u01 (0) = 0.

The general solution is given by:

1
u1 (τ ) = c1 cos(τ ) + c2 sin(τ ) + 32
cos(3τ ),

which with the initial conditions gives:

1


u1 (τ ) = 32
cos(3τ ) − cos(τ ) .

Thus, a first-order, uniformly valid perturbation solution is

ε


u(τ ) = cos(τ ) + 32
cos(3τ ) − cos(τ ) + . . . ,

where
3ε
τ =t+ 8
t + ...

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (46/47)
 Introduction
Duffing’s Equation
Regular Perturbation Method
Poincaré-Lindstedt Method
Nonlinear Oscillations

Poincaré-Lindstedt Method 4
Duffing’s Equation is given by:
3
ü + u + εu = 0, u(0) = 1, u̇(0) = 0, t > 0,

with the Poincaré-Lindstedt approximate solution:

ε cos(3τ ) − cos(τ )
+ . . . ,
u(τ ) = cos(τ ) + 32

where
τ = t + 3ε
8
t + ...

Below is a Phase Portrait and Time Series of v(t) vs u(t), where v = u̇.

Duffing’s Equation Duffing’s Equation

1 1
ε = 0.2 ε = 0.2

0.5
0.5

0
0 u
v

-0.5

-0.5
-1
Exact, ue Exact, ue
Approx u0 Approx u0
-1 Approx u1 Approx u1
-1.5
-1 -0.5 0 0.5 1 1.5 0 5 10 15 20 25
u t

Lecture Notes – Perturbation Methods —

Joseph M. Mahaffy, hjmahaffy@sdsu.edui (47/47)