Second order ordinary differential equations

Luis L. Bonilla

Universidad Carlos III de Madrid, 2.1.D17, bonilla@ing.uc3m.es

October 8, 2019

Bonilla (UC3M) Lecture 3 October 8, 2019 1 / 26

 Outline

Outline

Linear second order ODEs.

Variation of parameters.
Method of undetermined coefficients.
Reduction of order.
Driven systems, linear oscillators and resonance.

Bonilla (UC3M) Lecture 3 October 8, 2019 2 / 26

 Linear second order ODEs

General properties

Existence and uniqueness: a(t), b(t), f (t) continuous on interval I .

Then for each t0 in I and each set of values of y0 , v0 , the IVP:

y 00 + a(t)y 0 + b(t)y = f (t), y (t0 ) = y0 , y 0 (t0 ) = v0 ,

has a unique solution y (t) for t ∈ I .

Superposition principle for undriven ODE: y1 (t), y2 (t) are solutions of
the homogeneous ODE with f (t) = 0. Then y (t) = c1 y1 (t) + c2 y2 (t)
is also a solution (ci are arbitrary constants).
For independent solutions yi (t), with Wronskian determinant
W (y1 , y2 ) = y1 y20 − y2 y10 6= 0, y (t) = c1 y1 (t) + c2 y2 (t) is the general
d
solution. Abel formula: dt W (y1 , y2 ) = −a(t) W (y1 , y2 ), f (t) = 0.
Particular solution yp (t) of the driven ODE gives general solution:
y (t) = yp (t) + c1 y1 (t) + c2 y2 (t) if y1 (t), y2 (t) are independent
solutions of the homogeneous ODE.
Bonilla (UC3M) Lecture 3 October 8, 2019 3 / 26
 Linear second order ODEs

Superposition principle

Let y 00 + a(t)y 0 + b(t)y = 0 be a linear homogeneous ODE and y1 (t), y2 (t)

two of its solutions. Then y (t) = c1 y1 (t) + c2 y2 (t) is also a solution.

We insert this y (t) directly in the ODE and use linearity of the derivatives:

(c1 y1 + c2 y2 )00 + a(t)(c1 y1 + c2 y2 )0 + b(t)(c1 y1 + c2 y2 )

= (c1 y100 + c2 y200 ) + a(t)(c1 y10 + c2 y20 ) + b(t)(c1 y1 + c2 y2 )
= c1 [y100 + a(t)y10 + b(t)y1 ] + c2 [y200 + a(t)y20 + b(t)y2 ] = 0.

Here y100 + a(t)y10 + b(t)y1 = 0 and y200 + a(t)y20 + b(t)y2 = 0 because y1

and y2 are both solutions of the ODE.

Example: y1 = 1 + t and y2 = e t are solutions of ty 00 − (1 + t)y 0 + y = 0,

therefore y = c1 (1 + t) + c2 e t is also a solution.

Bonilla (UC3M) Lecture 3 October 8, 2019 4 / 26

 Linear second order ODEs

Independence of solutions
Let y 00 + a(t)y 0 + b(t)y = 0 be a linear homogeneous ODE and y1 (t),
y2 (t) two of its solutions. y1 (t), y2 (t) are linearly independent iff their
Wronskian is not zero.
0 0
 
 = 0 ⇐⇒ y1 y20 − y10 y2 = 0 ⇐⇒ y2 − y1 = 0
 y1 y2 
W [y1 , y2 ] ≡  0

0
y y 
1 2 y2 y1
d d d d y2
⇐⇒ ln y2 − ln y1 = 0 ⇐⇒ (ln y2 − ln y1 ) = ln =0
dt dt dt dt y1
y2
⇐⇒ = c ⇐⇒ y2 = c y1 .
y1
We have proven that the Wronskian is zero if and only if two functions are
linearly dependent: y2 (t) = c y1 (t). This is equivalent to the statement.
 1 + t et 
 
t  = (1 + t)e t − e t = te t 6= 0,
Example: W [1 + t, e ] =  
1 et 
so y1 = 1 + t and y2 = e t are linearly independent solutions of
ty 00 − (1 + t)y 0 + y = 0. The general solution is y (t) = c1 (1 + t) + c2 e t .
Bonilla (UC3M) Lecture 3 October 8, 2019 5 / 26
 Linear second order ODEs

Abel’s formula
Let y 00 + a(t)y 0 + b(t)y = 0 be a linear homogeneous ODE and y1 (t),
d
y2 (t) two of its solutions. Then dt W [y1 , y2 ] = −a(t) W [y1 , y2 ].
d d d d
W [y1 , y2 ] = (y1 y20 − y1 y20 ) = (y1 y20 ) − (y10 y2 )
dt dt dt dt
= (y1 y200 + y10 y20 ) − (y10 y20 + y100 y2 ) = y1 y200 − y2 y100
= y1 [−a(t)y 0 2 − b(t)y2 ] − y2 [−a(t)y10 − b(t)y1 ] = a(t)y2 y10 − a(t)y1 y20
= −a(t)(y1 y20 − y1 y20 ) = −a(t) W [y1 , y2 ].

Example:

W [1 + t, e t ] = te t =⇒ ln W [1 + t, e t ] = t + ln t
d 1 1+t
=⇒ ln W [1 + t, e t ] = 1 + = ,
dt t t
which is Abel’s formula as ty 00 − (1 + t)y 0 + y = 0 divided by t yields
y 00 − (1 + t)y 0 /t + y /t = 0 and therefore a(t) = − 1+t
t .
Bonilla (UC3M) Lecture 3 October 8, 2019 6 / 26
 Linear second order ODEs

Solving linear ODEs with constant coefficients

Let y 00 + ay 0 + by = 0 be a linear homogeneous ODE with constant
coefficients. We try exponential solutions: y = e rt .

0 = (e rt )00 + a(e rt )0 + b(e rt ) = r 2 e rt + are rt + be rt = (r 2 + ar + b)e rt

√
2 −a ± a2 − 4b
=⇒ r + ar + b = 0 =⇒ r = characteristic roots.
2
Cases:
(a) a2 > 4b: √ Two different real√roots r+ and r− give solution
2 2
y = (c1 e t a −4b/2 + c2 e −t a −4b/2 )e −at/2 .
(b) a2 = 4b: one double real root r0 = − 2a give solution
y = (c1 + c2 t)e −at/2 .
q
a2
(c) a2 < 4b: two complex conjugate roots − 2a ± iω, ω = b− 4, give
solution y = [c1 cos(ωt) + c2 sin(ωt)] e −at/2 .
Bonilla (UC3M) Lecture 3 October 8, 2019 7 / 26
 Linear second order ODEs

Visualizing solutions: damped spring (spiral point/focus)

Bonilla (UC3M) Lecture 3 October 8, 2019 8 / 26

 Linear second order ODEs

Visualizing solutions: vector fields

For an autonomous second order ODE:

y 00 + f (y , y 0 ) = 0,

1 Write it as two first order ODEs:

y0 = v,

0
v = −f (y , v ).

2 On a grid of the phase plane (y , v ) draw at each point a velocity

vector (v , −f (y , v )) with the same length.
• For a linear ODE with constant coefficients, f (y , y 0 ) = ay 0 + by .
• A first order ODE y 0 = f (t, y ) can be written as the system t 0 = 1,
y 0 = f (t, y ), whose vector field (1, f (t, y )) is the direction field studied in
the previous lectures.
Bonilla (UC3M) Lecture 3 October 8, 2019 9 / 26
 Linear second order ODEs

Visualizing solutions: vector field damped spring

y 00 + 0.4y 0 + 65y = 0, equivalent to y 0 = v , v 0 = −65y − 0.4v .

Bonilla (UC3M) Lecture 3 October 8, 2019 10 / 26

 Linear second order ODEs

Visualizing solutions: vector field undamped spring (center)

y 00 + 4y = 0, equivalent to y 0 = v , v 0 = −4y .

Bonilla (UC3M) Lecture 3 October 8, 2019 11 / 26

 Linear second order ODEs

Visualizing solutions: saddle point

y 00 + y 0 − 2y = 0, equivalent to y 0 = v , v 0 = 2y − v .

Bonilla (UC3M) Lecture 3 October 8, 2019 12 / 26

 Variation of parameters

Variation of parameters
Inhomogeneous ODE: ay 00 + by 0 + cy = F (t).
y1 (t), y2 (t) are independent solutions of the homogeneous ODE.
Particular solution of the inhomogeneous ODE:
Z t Z t
y2 (s)F (s)ds y1 (s)F (s)ds
yp (t) = −y1 (t) + y2 (t) ,
t0 a(s)W (y1 , y2 )(s) t0 a(s)W (y1 , y2 )(s)

where W (y1 , y2 ) = y1 y20 − y2 y10 is the Wronskian determinant.

General solution is the sum of a particular solution of the
inhomogeneous ODE plus a general solution of the homogeneous
(F = 0) ODE:
y (t) = yp (t) + c1 y1 (t) + c2 y2 (t).
Proof of formula: Write ODE as y 00 + ba y 0 + ca y = Fa , and for
y = u1 (t)y1 (t) + u2 (t)y2 (t) find y1 u10 + y2 u20 = 0, y10 u10 + y20 u20 = F
a, and
solve.
Bonilla (UC3M) Lecture 3 October 8, 2019 13 / 26
 Variation of parameters

Example: ty 00 − (1 + t)y 0 + y = t 2 e 2t

Solutions of homogeneous ODE: y1 = e t , y2 = 1 + t.

System of equations for f (t) = t e 2t :
 t  0   
e 1+t u1 0
= .
et 1 u20 t e 2t

Then, subtracting equations, −t u20 = t e 2t =⇒ u20 = −e 2t , and

u2 = − 21 e 2t + c2 . Also, u10 e t = −(1 + t)u20 = (1 + t)e 2t , which yields
u10 = (1 + t)e t . Then u1 = t e t + c1 .
General solution: y (t) = u1 (t)y1 (t) + u2 (t)y2 (t):

t − 1 2t
y (t) = e + c1 e t + c2 (1 + t).
2
Try particular solution yp = (At + B)e 2t .

Bonilla (UC3M) Lecture 3 October 8, 2019 14 / 26

 Undetermined coefficients

Method of undetermined coefficients

F (t) yp (t)

Pn (t) = a0 t n + a1 t n−1 + . . . + an t s (A0 t n + A1 t n−1 + . . . + An )

Pn(t)e αt t s (A0 t n + A1 t n−1 + . . . + An )e αt
cos βt
Pn (t) t s [(A0 t n + A1 t n−1 + . . . + An ) cos βt
sin βt
+(B0 t n + B1 t n−1 + . . . + Bn ) sin βt]

Particular solutions of the ODE ay 00 + by 0 + cy = F (t) depending on the

form of the source term F (t). Here s is the smallest nonnegative integer
(s = 0, 1, or 2) that will ensure that no term in yp (t) is a solution of the
corresponding homogeneous equation. Equivalently, for the three cases, s
is the number of times 0 is a root of the characteristic equation, α is a
root of the characteristic equation, and iβ is a root of the characteristic
equation, respectively.
Bonilla (UC3M) Lecture 3 October 8, 2019 15 / 26
 Undetermined coefficients

Example: y 00 − 2y 0 − 3y = t
Variation of parameters. Solution of homogeneous ODE:
y = u1 e −t + u2 e 3t , because r 2 − 2r − 3 = 0 has the solutions r = −1 and
r = 3. Now u1 (t) and u2 (t) selected so that
u10 e −t + u20 e 3t = 0.
Then
−u10 e −t + 3u20 e 3t = t.
This yields (adding the two equations): 4u20 e 3t = t, hence u20 = 4t e −3t .
Integrating by parts, u2 = −( 12 t
+ 361
)e −3t + c2 . From the first equation,
u10 e −t = −u20 e 3t =⇒ u10 = − 4t e t . Integrating by parts, u1 = 1−t t
4 e + c1 .
2−3t −t
We obtain y (t) = 9 + c1 e + c2 e . 3t

Undetermined coefficients. Particular solution: yp = at + b. Inserting in

ODE: −2a − 3at − 3b = t =⇒ 2a + 3b + (1 + 3a)t = 0. Then 1 + 3a = 0
and 2a + 3b = 0, which produce a = − 13 , b = 29 . Thus yp = − 3t + 29 .
Adding general solution of the homogeneous ODE, same result as before.
Bonilla (UC3M) Lecture 3 October 8, 2019 16 / 26
 Undetermined coefficients

b2
Example: y 00 + by 0 + cy = cos t, with c − 4 = ω2 > 0
Undetermined coefficients. Particular solution: yp = A cos t + B sin t.
Inserting in ODE:
(c − 1)A + Bb = 1, (c − 1)B − Ab = 0 =⇒ A = (c − 1)γ, B = bγ with
γ −1 = (c − 1)2 + b 2 . Then

(c − 1) cos t + b sin t
y (t) = + e −bt/2 (d1 cos ωt + d2 sin ωt),
(c − 1)2 + b 2

where we have used r 2 + br + c = 0 to solve the homogeneous ODE.

Resonance: b = 0, ω = 1, y (0) = 0, y 0 (0) = 0.

resonance
t sin t
2

40
20
t
20 40 60 80 100
-20

cos t−cos ωt t
y (t) = ω 2 −1
→ 2 sin t, as ω → 1. -40

Bonilla (UC3M) Lecture 3 October 8, 2019 17 / 26

 Reduction of order

Reduction of order

Let y1 (t) be a solution of a(t)y 00 + b(t)y 0 + c(t)y = 0. The substitution

y (t) = y1 (t)v (t) transform the ODE a(t)y 00 + b(t)y 0 + c(t)y = F (t) in a
first-order ODE for v 0 :
u10 0
 
00 F (t)
a(t)v + b(t) + 2 v = .
u1 y1 (t)
y
Example. y 00 − 1+t 0
t y + t = 0 is solved by y1 = 1 + t. y = (1 + t)v gives:

(1 + t)2 0 v 00 1 2
(1 + t)v 00 + 2v 0 − v = 0 =⇒ 0 = 1 + −
t v t 1+t
te t
R te t dt
or ln v 0 = t + ln (1+t)
t
2 =⇒ v 0 = (1+t) 2 =⇒ v = (1+t)2
=
te t et
− 1+t + e dt = 1+t . Thus y = e t is the other independent solution.
t
R

Bonilla (UC3M) Lecture 3 October 8, 2019 18 / 26

 Driven systems, linear oscillators and resonance

Nonlinear autonomous ODE

d y 02
y 00 + V 0 (y ) = 0 =⇒ 0 = y 0 [y 00 + V 0 (y )] = [ + V (y )].
dt 2
y 02
Then 2 + V (y ) = C and
Z
dy
± p = t − t0 .
2[C − V (y )]
Example: Pendulum: ml θ̈ = −mg sin θ =⇒ 12 θ̇2 + gl (1 − cos θ) = C .

Bonilla (UC3M) Lecture 3 October 8, 2019 19 / 26

 Driven systems, linear oscillators and resonance

Linear oscillations and resonance

Damped pendulum with force −γ θ̇ − mgl sin θ:

γ g
θ̈ +
θ̇ + sin θ = 0.
ml l
q
Use sin θ ≈ θ and t = ω0 t̃, with ω0 = gl ,

γ
θ̈ + 2β θ̇ + θ = 0, √ , (damped spring).
β=
2m gl
p
λ2 + 2βλ + 1 = 0 gives λ = −β ± β 2 − 1.
√ √
2 2
β > 1, overdamped pendulum: θ = e −βt (ae β −1t + be β −1t ).
0 ≤ βp< 1, underdamped pendulum: θ(t) = ce −βt cos(Ωt + ϕ),
Ω = 1 − β 2 . Also θ = e −βt (a cos Ωt + b sin Ωt), a + ib = ce iϕ .

Bonilla (UC3M) Lecture 3 October 8, 2019 20 / 26

 Driven systems, linear oscillators and resonance

Visualizing solutions: driven damped spring

Bonilla (UC3M) Lecture 3 October 8, 2019 21 / 26

 Driven systems, linear oscillators and resonance

Visualizing solutions: driven undamped spring

Bonilla (UC3M) Lecture 3 October 8, 2019 22 / 26

 Driven systems, linear oscillators and resonance

Particular solution: transfer function

Add periodic forcing term (coefficient can be set equal to 1):

θ̈ + 2β θ̇ + θ = cos ωt.

For a force e iωt

1
θ(t) = Re H(iω)e iωt = Re e iωt ,
1− ω2 + 2iβω

where H(r ) = 1/(r 2 + 2βr + 1) is the transfer function.

Amplitude and phase shift of transfer function:

1 2βω
M(ω) = p , ϕ(ω) = − arctan .
(1 − ω )2 + 4β 2 ω 2
2 1 − ω2

If ω < 1, −π < ϕ(ω) < 0. Steady solution is θ(t) = M(ω) cos[ωt + ϕ(ω)].

Bonilla (UC3M) Lecture 3 October 8, 2019 23 / 26

 Driven systems, linear oscillators and resonance

Gain and phase shift (Bodé plots)

Gain M(ω) (ratio of amplitude of response to force amplitude):

Resonance:
1 p
max M(ω) = p , ωmax = 1 − 2β 2 .
2β 1 − β 2

As β → 0+, M(ω) → +∞.

Bonilla (UC3M) Lecture 3 October 8, 2019 24 / 26
 Driven systems, linear oscillators and resonance

Resonance
Spring driven by a periodic forcing term:

θ̈ + 2β θ̇ + θ = cos ωt.

General solution:
p
θ(t) = M(ω) cos[ωt + ϕ(ω)] + ce −βt cos[Ωt + ϕ0 ], Ω= 1 − β2

1 2βω
M(ω) = p , ϕ(ω) = − arctan .
(1 − ω 2 )2 + 4β 2 ω 2 1 − ω2

Undamped oscillator with θ(0) = θ̇(0) = 0:

cos ωt − cos t 2 sin (1−ω)t

2 (1 + ω)t t
θ(t) = 2
= 2
sin =⇒ θ ≈ sin t (ω → 1).
1−ω 1−ω 2 2

Bonilla (UC3M) Lecture 3 October 8, 2019 25 / 26

 Soft and hard springs

Soft and hard springs

my 00 = −ky ∓ αy 3 − cy 0 ; +: soft spring; -: hard spring.

k 2 α 4
V (y ) = y ± y .
2m 4m
02
For c = 0, y2 + V (y ) = E and V (ytp ) = E , ytp > 0, give the period
Z y Z ytp
du du
± p = t − t0 =⇒ P = 4 p .
2[E − V (u)] 0 2[E − V (u)]

Bonilla (UC3M) Lecture 3 October 8, 2019 26 / 26