ME673: Mathematical Methods for Engineers

Ordinary Di↵erential Equations

S. Gopalakrishnan
 Flows, heat transfer,
Electromagnetics etc Phenomena

Navier Stokes, heat equation,

Maxwell's equation etc Governing Equations

FDM, FVM, Galerkin, LBM etc Numerical method

Computation
 Flows, heat transfer,
Electromagnetics etc Phenomena

Error

Navier Stokes, heat equation,

Maxwell's equation etc Governing Equations

Error

FDM, FVM, Galerkin, LBM etc Numerical method

Error

Computation
Di↵erential equation: an equation relating an unknown function to
its various derivatives.
Solution: A function di↵erentiable sufficiently often so that when
the function and its derivatives are substituted in the di↵erential
equation an identity results.
ODE: DE with derivatives with respect to only one variable.
PDE: DE with derivatives with respect to more than one variable.
Order of DE: Order of the highest derivative in it.
Degree of DE: Power of the highest derivative in it.
Di↵erential equation: an equation relating an unknown function to
its various derivatives.
Solution: A function di↵erentiable sufficiently often so that when
the function and its derivatives are substituted in the di↵erential
equation an identity results.
ODE: DE with derivatives with respect to only one variable.
PDE: DE with derivatives with respect to more than one variable.
Order of DE: Order of the highest derivative in it.
Degree of DE: Power of the highest derivative in it.

" ✓ ◆2 # 32
du d 3u
1+ =k
dx dx 3
✓ ◆2 ✓ ◆2 ✓ ◆4 ✓ ◆6
2 d 3u du du du
=) k =1+3 +3 +
dx 3 dx dx dx
=) Order is 3, Degree is 2.
In general, an nth order ODE is written as

du 00 d 2 u
u (n) = f (x, u, u 0 , u 00 , . . . , u (n 1)
) where u 0 = ,u = , etc.
dx dx 2
A nth order linear ODE is written as
0
u (n) = q(x) Pn 1 (x)u Pn 2 (x)u ··· P0 (x)u (n 1)

or

u (n) + P0 (x)u (n 1)
+ P1 (x)u (n 2)
+ · · · + Pn 1 (x)u = q(x)

if q(x) = 0, then it is a homogenous ODE else it is

non–homogenous.
Nonlinear DEs contain powers of unknown function higher than
one or products of the unknown function and its derivatives or
powers of derivatives higher than one. For example,
✓ ◆2
d 2u du du 3 du
2
+u = 0, or = xu , or = xe x
dx dx dx dx
A general solution of an nth order linear ODE is a family of
solutions containing n arbitrary constants. Determining the values
of these n constants takes general solution to a specific solution.

u 0 = f (x, u) with u(x0 ) = u0 : general first order IVP

u 00 = f (x, u, u 0 ) with u(x0 ) = u0 , u 0 (x0 ) = u00 : general 2nd order IVP

u 00 = f (x, u, u 0 )withu(x0 ) = u0 , u(x1 ) = u1 : typical BVP
IVP: Initial Value Problem. conditions specify values of solution
and / or its derivatives at one point x = x0 .
BVP: Boundary Value Problem. conditions specify values of
solution and / or its derivatives at more than one point.
First Order ODEs:
du
= f (x, u) in general
dx
du
+ p(x)u = q(x) linear first order ODE
dx
If f (x, u) can be written as or is in the form of g (x) · v (u) where
v (u) 6= 0, then we have a seperable equation.

du 1
= g (x)v (u) =) du = g (x)dx
dx v (u)
Z Z
h(u)du = g (x)dx =) h(u)du = g (x)dx + c

where c is an arbitrary constant.

The linear homogenous equation u 0 + p(x)u = 0 is separable.
R
p(x)dx
u = ce

If the DE is of the type

du ⇣u ⌘ u
=f letting u = vx or v =
dx x x
du dv dv dv dx
=x + v =) x + v = f (v ) =) =
dx dx dx f (v ) v x
Radioactive decay of nuclei
dN dN
/ N =) = N
dt dt
Separable as
dN
= dt
N
with N|t=0 = N0 (IVP)
t
N = N0 e

Newton’s law of cooling

dT
/ (T Tsurr )
dt
d✓
= c✓ ✓ = T Tsurr
dt
Separable as
d✓
= cdt
✓
Exact equations: We have
du
= f (x, u)
dx
Let
M(x, u)
f (x, u) =
N(x, u)
Then the first equation is

M(x, u)dx + X (x, u)du = 0

if M = @@x and N = @@u then d = 0, which is the exact di↵erential

of some function (x, u) which implies = constant Now,

@M @2 @N @2
= & =
@u @x@u @x @u@x
. we have
@M @N
=
@u @x
use this condition to determine if Mdx + Ndu = 0 , is exact or not.
Example:
(2 + x 2 u)du + xu 2 dx = 0
M = xu 2 , N = 2 + x 2 u
@M @N @M @N
) = 2xu, = 2xu =) = =) exact equation
@u @x @u @x
To determins :

@ x 2u2
M= = xu 2 =) = + h(u)
@x 2
@
) = x 2 u + h0 (u) = N = 2 + x 2 u =) h0 (u) = 2
@u
x 2u2
) = + 2u = constant
2
is the solution implicit in u.
Integrating factor
In general, Mdx + Ndu = 0 is not exact. Integrating factor is a
function I (x, u) such that

I (Mdx + Ndu) = 0 becomes exact

For that , it is necessary that,

@ @
(IM) = (IN)
@u @x
The general first order linear equation
du
+ p(x)u = g (x)
dx
has an integrating factor F (x) as pure function of x. Writing the
equation as

du + [p(x)u g (x)]dx = 0, M = p(x)u g (x), N = 1

Then

F (x)du + F (x)[p(x)u g (x)]dx = 0 needs to be exact

@ @
) [F (x)(p(x)u g (x))] = (F (x) · 1)
@u @x
F 0 (x) R
F (x) · p(x) = F 0 (x), or = p(x) =) F (x) = e p(x)dx
F (x)
Now
d(F · u) = Fdu + udF
Since
@F
F 0 (x) = = F (x)p(x), dF = F (x) · p(x)dx
dx
) d(F · u) = F [du + u · p · dx]
| {z }
g ·dx
Z
) d(F · u) = F · g · dx F · u = F · g · dx + c
Z
1
u= F (x) · g (x)dx + c
F (x)
To make it convenient for solving an IVP
du
+ p(x) · u = g (x), u(x0 ) = u0
dx
Let Rx
F (s)·g (s)ds
F (x) = e x0
=) F (x0 ) = 1
Then Z x
1 c
u(x) = F (s) · g (s)ds +
F (x) x0 F (x)
Z x0
1 c
) u(x0 ) = F (s) · g (s)ds + =) u0 = c
F (x) x0 F (x0 )
Solution to IVP is
Z x
1 U0
u(x) = F (s) · g (s)ds +
F (x) x0 F (x)
Kircho↵’s Law for simple electrical circuit:
dI q dq
V (t) = IR + L + I = (current)
dt C dt

dV (t) dI d 2I I dV (t)
) =R +L 2 + : let = V1 (t)
dt dt dt C dt
d 2I dI I
L + R + = V1 (t) : Linear 2nd order ODE
dt 2 dt C
If there is no capacitor in the circuit
dI
L + RI = V (t) : Linear 1st order ODE
dt
Linear second order ODEs
The general second order linear ODE is

d 2u du
2
+ p0 (x) + p1 (x)u = g (x) a<x <b
dx dx
One can write it as L[u] = g (x), where

d2 d
L = di↵erential operator = 2
+ p0 (x) + p1 (x)
dx dx
Example
d2 d
L= +3 + 2, u = e x
dx 2 dx
) L[u] = e x + 3 · e x + 2e x = 6e x
L is a linear operator, for any constants A & B and functions u &
v to which L can be applied, we have

L[Au + Bv ] = A · L[u] + B · L[v ]