Math 21b Differential Equations 1

Example. A particular species of bird population grows at a rate proportional to the bird population. Denote
x(t) the the bird population at time t. Write a differential equation which models x(t), and solve it.

dx
Solution. We need to express in terms of x and t. The rate of change being proportional to
dt
the population means
dx
= λx
dt
for some constant λ.
We can solve this differential equation by separating variables:
dx
= λdt
Z x Z
dx
= λdt
x
ln |x| = λt + c
x(t) = ±eλt+c
x(t) = beλt

where b = ±ec .

Example. Residents of two small islands regularly move between the two islands. Suppose that, each year,
3% of the residents of the first island move to the second island, and 5 % of the residents of island two move
to island one. Denote by

x1 (t) = population of island 1 at time t

x2 (t) = population of island 2 at time t

where t is measured in years. Write a continuous dynamical system modeling this situation.
dx1 dx2
Solution. Writing a continuous dynamical system means we need to express and in
dt dt
terms of x1 and x2 . The rate of change of island 1 is approximately −0.03x1 + 0.05x2 (the island
loses roughly 3% of its population per year, while gaining roughly 5% of the population of island
dx2
2). Similarly, = 0.03x1 − 0.05x2 . So we end up with
dt
dx1
= −0.03x1 + 0.05x2
dt
dx2
= 0.03x1 − 0.05x2
dt
which we can write in matrix form as
 
d~x −0.03 0.05
= ~x.
dt 0.03 −0.05
 d~x
Solving Dynamical Systems. Suppose = A~x is a continuous dynamical system and {~v1 , . . . , ~vn } is an
dt
eigenbasis for A.

1. How can we express ~x(t) in terms of the eigenbasis {~v1 , . . . , ~vn }?

Solution. For a fixed t, the state ~x(t) is simply a vector in Rn , so we can write it as a linear
combination of the basis vectors ~v1 , . . . , ~vn . We’ll get different coefficients for different t, so
the coefficients will themselves be functions of t:

~x(t) = a1 (t)~v1 + · · · + an (t)~vn .

d~x
2. Rewrite the differential equation = A~x(t) using your answer to the previous part. Use this to find
dt
a general solution to the continuous dynamical system.

Solution. Differentiating ~x(t) = a1 (t)~v1 + · · · + an (t)~vn gives

d~x
= a01 (t)~v1 + · · · + a0n (t)~vn .
dt
On the other hand, multiplying ~x(t) = a1 (t)~v1 + · · · + an (t)~vn by A gives

A~x(t) = a1 (t)A~v1 + · · · + an (t)A~vn

= a1 (t)λ1~v1 + · · · + an (t)λn~vn .

d~x
So, we can rewrite the differential equation = A~x as
dt
a1 (t)~v1 + · · · + an (t)~vn = a1 (t)λ1~v1 + · · · + an (t)λn~vn .

Since the vectors ~vi are a basis and therefore linearly independent, we these can only be
equal if
a01 (t) = λ1 a1 (t), a02 (t) = λ2 a2 (t), . . . a0n (t) = λn an (t).
This gives n differential equations a0i (t) = λi ai (t), which we recognize have solution ai (t) =
bi eλi t where bi is some constant that doesn’t depend on t. Plugging this back in gives

~x(t) = b1 eλ1 t~v1 + · · · + bn eλn t~vn

where b1 , . . . , bn are any real numbers.

d~x
3. What is the solution ~x(t) of = A~x satisfying the initial condition ~x(0) = c1~v1 + · · · + cn~vn ?
dt
Solution. We found in part (2) that the general solution is ~x(t) = b1 eλ1 t~v1 + · · · + bn eλn t~vn
where b1 , . . . , bn are any real numbers. Now, we just use the initial condition to solve for the
constants b1 , . . . , b)n.
On the one hand, plugging t = 0 into the general solution ~x(t) = b1 eλ1 t + · · · + bn eλn t gives
~x(0) = b1~v1 + · · · + bn~vn . On the other hand, we are also told that ~x(0) = c1~v1 + · · · + cn~vn .
Equating these two different ways of writing ~x(0) gives b1 = c1 , . . . , bn = cn .
Continuous Dynamical Systems. Suppose A has an eigenbasis ~v1 , . . . , ~vn with corresponding eigenvalues
d~x
λ1 , . . . , λn . The solution to the continuous dynamical system = A~x with initial condition ~x(0) = c1~v1 +
dt
· · · + cn~vn is
~x(t) = c1 eλ1 t~v1 + · · · + cn eλn t~vn .
Example. Consider the continuous dynamical system
 
d~x −2 0
= ~x.
dt −6 1
 
2
1. Find the solution satisfying the initial condition ~x(0) = .
1
 
−2 0
Solution. As long as the matrix A = is diagonalizable we can just use the
−6 1
above formula to find ~x(t); we just need to find the eigenvalues and eigenvectors. The
characteristic polynomial of A is fA (λ) = (−2 − λ)(1 − λ) so the eigenvalues are λ = −2, 1.
The corresponding eigenspaces are
   
0 0 1
E−2 = ker = span
−6 3 2
   
−3 0 0
E1 = ker = span .
−6 0 1
   
1 0
So the vectors ~v1 = and ~v2 = are an eigenbasis for A.
2 1
We need to express the initial condition ~x(0) as a linear combination of ~v1 and ~v2 . Using
your favorite method we find ~x(0) = 2~v1 − 3~v2 . Therefore we get
   
1 0
~x(t) = 2e−2t~v1 − 3et~v2 = 2e−2t − 3et .
2 1

2. Sketch the phase portrait of the continuous dynamical system. (The phase portrait is a drawing that
shows the different trajectories ~x(t) that can result from different initial conditions.)
Asymptotic Stability. A continuous dynamical system is asymptotically stable if limt→∞ ~x(t) = ~0 for
any initial condition ~x(0).

d~x
Example. Determine if the continuous dynamical system = A~x is asymptotically stable or not.
dt
 
−1 0
1. A =
0 −4
Solution. The eigenvalues are λ = −1, −4 so the general solution is

~x(t) = c1 e−t + c2 e−4t .

The limit limt→∞ ~x(t) = ~0 so the system is asymptotically stable.

Notice if we were looking at a discrete dynamical system ~x(t + 1) = A~x(t) this would not be
asymptotically stable, since | − 4| ≥ 1!
 
0 −1
2. A =
1 0
Solution. The eigenvalues are λ = ±i so the general solution is

~x(t) = c1 e−it + c2 eit

~x(t) = c1 (cos t − i sin t) + c2 (cos t + i sin t)
~x(t) = b1 cos t + b2 sin t

(where b1 = c1 + c2 and b2 = i(c1 − c2 )). The trajectories ~x(t) are circles, so the system is
not asymptotically stable.

d~x
Asymptotic Stability and Eigenvalues. A continuous dynamical system = A~x is asymptotically stable if
dt
and only if all of it’s complex eigenvalues have negative real part, i.e., for every complex eigenvalue λ = a + ib
we have a < 0.