Dr. Z.

’s Calc4 Lecture 14 Handout: General Theory of nth Order Linear Differential Equations

By Doron Zeilberger

Version of Dec. 11, 2016 (Thanks to Alex Eisenschmied)

The general form of a linear nth order linear diff.eq. is (after you divide by the coefficient of
dn y (n)
dtn (aka as y (t))

dn y dn−1 y dy
n
+ p1 (t) n−1
+ . . . + pn−1 (t) + pn (t)y(t) = g(t) ,
dt dt dt

where, in general, the coefficients, p1 (t), p2 (t), ..., pn (t) and g(t) are functions of t (not just
constants).

In shorthand notation it is written

y (n) (t) + p1 (t)y (n−1) (t) + . . . + pn−1 (t)y 0 (t) + pn (t)y(t) = g(t) .

Existence and Uniqueness Theorem for nth Order Linear Diff.eqs.

If the coefficient-functions p1 (t), p2 (t), . . . pn (t) and the right side, g(t), are continuous (i.e. “well-
behaved”, they have no breaks and do not blow up) in an open interval I containing the number
t0 , then we are guaranteed that there exists a unique solution y(t) to the initial value diff.eq.

y 00 (t) + p(t)y 0 (t) + q(t)y(t) = g(t) , y(t0 ) = z0 , y 0 (t0 ) = z1 , y (n−1) (t0 ) = zn−1 ,

(for any numbers z0 , z1 , . . . , zn−1 ).

In other words, if the functions p1 (t), p2 (t), . . . , pn (t),and g(t) do not blow up, and have no surprises
in I, and you want a function y(t) that satisfies the diff.eq. satisfying the initial conditions then
you are promised that there is such a function. Not only that, there is only one such function!

Problem 14.1: For each of the following diff.eq. initial value problems, and intervals, decide
whether the theorem promises you that there is a unique solution.

a. (t2 + 1) y 000 (t) + sin t y 00 (t) + (t3 + 1) y(t) = et , y(0) = 1 , y 0 (0) = 3 y 00 (0) =
5 ; −10 < t < 10

b. t2 y 0000 (t)+sin t y 00 (t)+(t3 +1) y(t) = et , y(0) = 1 , y 0 (0) = 3 , y 00 (0) = 5 , y 000 (0) =
−3 ; −10 < t < 10

Solution to 14.1a: We first divide by the coefficient of y 000 (t), t2 + 1, getting the equivalent
diff.eq. (initial value problem)

t3 +1 et
a.’ y 000 (t) + sin t 00
t2 +1 y (t) + t2 +1 y(t) = t2 +1

1
 3 t
So p1 (t) = tsin t t +1 e
2 +1 , p2 (t) = 0, p3 (t) = t2 +1 , and g(t) = t2 +1 . The numerators are all nice (continuous)

functions, and the denominator in these, t2 +1 never vanishes in the interval (−10, 10), so everything
is nice, and the theorem promises a unique solution.

Solution to 14.1b: We first divide by the coefficient of y 000 (t), t2 , getting the equivalent diff.eq.
(initial value problem)

t3 +1 et
b.’ y 0000 (t) + sin t 00
t2 y (t) + t2 y(t) = t2
3 t
So p1 (t) = 0 p2 (t) = sin t t +1 e
t2 , p3 (t) = 0 p4 (t) = t2 , and g(t) = t2 . The numerators are all nice
(continuous) functions, but the denominator in these, t2 vanishes at t = 0, and t = 0 happens to
lie in the given interval, so (at least one of, but in this case all of) the coefficients blow up and
hence are not continuous, and there is no guarantee.

Problem 14.2: Find the largest open interval for which the solution to

(t − 1)(t + 3) y (4) (t) + (cos t) y 00 (t) + sin t y 0 (t) + cos t y(t) = t ,

y(0) = 0, y 0 (0) = 1, y 00 (0) = −1, y 000 (0) = 4,

has a unique solution.

Solution to 14.2: After dividing by the coefficient of y (4) (t), (t − 1)(t + 3), we get the initial value
problem

cos t sin t cos t t

y (4) (t) + y 00 (t) + y 0 (t) y(t) = .
(t − 1)(t + 3) (t − 1)(t + 3) (t − 1)(t + 3) (t − 1)(t + 3)

The coefficients blow-up at t = −3 and t = 1. Our interval should contain t0 = 0, so the largest
interval is −3 < t < 1.

Ans. to 14.2: The largest interval is −3 < t < 1.

The amazing Principle of Superposition (only valid for homogeneous (linear) diff.eqs.!) that
we already know is applicable for first and second-order linear homog. diff.eqs. is of course valid
for any order!

If y(x) is some solution, and c is a constant, then we know automatically that cy(x) is also a
solution.

If y1 (x) and y2 (x) are two solutions, then we know automatically that y1 (x) + y2 (x) is also a
solution.

Repeating this process, we know that if y1 (x), . . . , yn (x) are some solutions of a homog. linear
diff.eq. then any linear combination

y(x) = c1 y1 (x) + . . . + cn yn (x)

2
is also a solution. But, given a candidate set of n specific solutions y1 (x), . . . , yn (x), can any
solution y(x) be expressed in terms of these proposed “atomic” functions?

There is beautiful criterion, called the Wronskian.

Important Theorem If the functions p1 (t), . . . , pn (t) are continuous (i.e. don’t blow up and don’t
have breaks) in an open interval I, and if the n functions y1 (t), . . . , yn (t) are all solutions of the
homog. linear diff.eq.

y (n) (t) + p1 (t)y (n−1) (t) + . . . + pn−1 (t)y 0 (t) + pn (t)y(t) = 0 .

and the so-called Wronskian (see below), W (y1 , y2 , . . . , yn )(t), is not always zero, then Every
solution of of that diff.eq., y(x), can be written as a linear combination

y(t) = c1 y1 (t) + . . . + cn yn (t) ,

for some numbers c1 , . . . , cn .

In that case we say that y1 (t), y2 (t), . . . , yn (t) form a fundamental set of solutions.

Important Definition

The Wronskian determinant of a list of functions y1 (t), . . . , yn (t) is:

y1 y2 ... yn
y10 y20 ... yn0
... ... ... ...
... ... ... ...
(n−1) (n−1) (n−1)
y1 y2 ... yn

Reminder about determinants

The determinant of a 2 × 2 matrix is defined by:

a1 b1
= a 1 b2 − a 2 b1 .
a2 b2

The determinant of a 3 × 3 matrix is (if you expand with respect to the top row)

a1 b1 c1
a2 b2 c2 =
a3 b3 c3

equals
b2 c2 a c2 b c2
a1 − b1 2 + c1 2
b3 c3 a3 c3 b3 c3

3
Higher order determinants are computed analogously, but it is much more efficient to use row
operations and column operations for larger determinants.

Problem 14.3: Decide whether the following functions are linearly independent or linearly depen-
dent. In the latter case find a linear relation among them.

a. y1 (t) = 1 + t, y2 (t) = 2 + t, y3 (t) = 3 + t

b. y1 (t) = 1 + t, y2 (t) = 1 + t2 , y3 (t) = 1 + t3

Solution to 14.3a: We are looking for numbers k1 , k2 , k3 such that

k1 (1 + t) + k2 (2 + t) + k3 (3 + t) = 0 ,

for all t. Simplifying, we get

(k1 + k2 + k3 )t + (k1 + 2k2 + 3k3 ) = 0 .

Setting the coefficients of t and 1 to 0 we get

k1 + k2 + k3 = 0 , k1 + 2k2 + 3k3 = 0 .

From the first, we get k3 = −k1 − k2 . Plugging into the second, we get

k1 + 2k2 + 3(−k1 − k2 ) = 0 ,

So
k1 + 2k2 − 3k1 − 3k2 = 0 ,

−2k1 − k2 = 0 ,

so k2 = −2k1 and going back to k3 : k3 = −k1 − (−2k1 ) = −k1 + 2k1 = k1 . Taking, for example,
k1 = 1, we get
k1 = 1 , k2 = −2 , k3 = 1 .

So a linear relation is:

1 · (1 + t) − 2 · (2 + t) + 1 · (3 + t) = 0 .

Ans. to 14.3a: The three functions are linearly dependent and a linear relation is: 1 · (1 + t) −
2 · (2 + t) + 1 · (3 + t) = 0.

Solution to 14.3b: We are looking for numbers k1 , k2 , k3 such that

k1 (1 + t) + k2 (1 + t2 ) + k3 (1 + t3 ) = 0 ,

for all t. Simplifying, we get

(k1 + k2 + k3 ) + k1 t + k2 t2 + k3 t3 = 0 .

4
Setting the coefficients of t, t2 and t3 to zero, we get
k1 = 0 , k2 = 0 , k3 = 0 , .
So the only solution is the trivial solution, and the three functions are linearly independent.

Ans. to 14.3b: The three functions are linearly independent.

Problem 4.4: Verify that the given functions are solutions of the diff.eq. and determine their
Wronskian. a.
xy 000 (x) − y 00 (x) = 0 ; 1 , x , x3 .

b.
y 000 (x) + y 0 (x) = 0 ; 1 , sin x , cos x

Solution to 4.4a: For y(x) = 1 we have

xy 000 (x) − y 00 (x) = x1000 − 100 = 0
For y(x) = x we have
xy 000 (x) − y 00 (x) = xx000 − x00 = 0
For y(x) = x3 we have
xy 000 (x) − y 00 (x) = x(x3 )000 − (x3 )00 = x(6) − 6x = 0 ,
so there three functions are indeed solutions. Now the Wronskian is The determinant of a 3 × 3
matrix is (if you expand with respect to the top row)
1 x x3
0 1 3x2 = 6x .
0 0 6x

Answer to 4.4a: The three functions are indeed solutions and their Wronskian is 6x.

Solution to 4.4b: For y(x) = 1 we have

y 000 (x) + y 0 (x) = 1000 + 10 = 0
For y(x) = sin x we have (y 0 (x) = cos x, y 00 (x) = − sin x, y 000 (x) = − cos x)
y 000 (x) + y 0 (x) = − cos x + cos x = 0
For y(x) = cos x we have (y 0 (x) = − sin x, y 00 (x) = − cos x, y 000 (x) = sin x)
y 000 (x) + y 0 (x) = sin x − sin x = 0
so there three functions are indeed solutions. Now the Wronskian is The determinant of a 3 × 3
matrix is (if you expand with respect to the top row)
1 sin x cos x
0 cos x − sin x = − cos2 x − sin2 x = −1 .
0 − sin x − cos x

Answer to 4.4b: The three functions are indeed solutions and their Wronskian is −1.