MATH 219

Fall 2024
Lecture 1
Lecture notes by Özgür Kişisel

Content: Introduction, Direction Fields, Separable Equations (parts of sections

1.1 and 1.3, section 2.2 (homogenous equations as described in problem 30 is also
covered)).
Suggested Problems: (Boyce, DiPrima, 10th edition)
§1.1: 4, 10, 15-20
§1.3: 3, 4, 11, 18, 26
§2.2: 4, 6, 17, 23, 32, 36

1 Introduction
A differential equation is a functional equation that contains derivatives and al-
gebraic operations. Such an equation always has one or more dependent variables
which are functions of one or more independent variables. Alternatively, several
equations containing several variables interrelated with each other may be present.
If this is the case then we say that we have a system of differential equations. It
is important to note that the solutions of a differential equation are functions, and
not numbers.

Example 1.1 Say x is a function of t. Consider the differential equation

dx
+ 5x = et .
dt
The equation says that the derivative of x with respect to t plus 5 times x is equal to
et . The function x(t) is the unknown here and the aim is to find all functions x(t)
that satisfy the equation. We will not solve the equation for now. Is it clear to you
whether this equation has a solution or not? Can you guess how many solutions this
equation has?

1
Example 1.2 Say x and y are both functions of t. Consider the system of differ-
ential equations
dx
+y = t
dt
dy
+ xy = sin(t).
dt
In this case, a solution is an ordered pair of functions (x(t), y(t)) such that both
of these equations simultaneously hold.

Example 1.3 Suppose that u is a function of both x and y, but assume that x and
y are two independent variables. Since u is a multivariable function, we need to
consider partial derivatives of u: Expressions such as du/dx don’t make much sense
but ∂u/∂x or ∂u/∂y do. For example, let us consider the equation
∂u ∂u
+ = x + y.
∂x ∂y
This is a differential equation that relates the partial derivatives of u with respect to
x and y. Solutions will be functions u(x, y) that satisfy this equality. Can you find
any solutions?

Definition 1.1 A differential equation in which there is only one independent vari-
able is called an ordinary differential equation (abbreviated to ODE). A dif-
ferential equation in which there is more than one independent variable is called a
partial differential equation (abbreviated to PDE).

It is easy to visually distinguish a PDE from an ODE since a PDE will contain
partial derivatives.

Example 1.4 Let x be a function of t. Consider the differential equation

d2 x dx
2
+ x + x2 = 1.
dt dt
In this equation not only the first derivative, but also the second derivative of x
appears. The equation constrains the function x(t) by relating its second and first
derivatives and the function itself.

2
Definition 1.2 The largest number n such that the nth derivative appears in a
differential equation is called the order of that equation.

Example 1.5 The equation

d2 x dx
2
+ x + x2 = 1
dt dt
has order 2. Example 1.1 above is a first order ODE, whereas Example 1.2 is a
system of first order ODE’s. Example 1.3 is a first order PDE. The equation
 5
dx d3 x
+ 3 +x=0
dt dx

has order 3 and not 5: The 5th power appearing in the equation is an algebraic
operation, so it does not affect the order of the equation.

2 Why do we study differential equations?

It is a trivial matter to write down as many differential equations as one likes.
However, this is not the main focus of the subject; the topic is interesting mainly
because of its applications to real life problems. The dependent and independent
variables often represent certain physical quantities such as volume, price, velocity,
length, time etc. Applications are very diverse, spanning essentially all branches of
science and technology.
But why are differential equations useful in such areas? Why do we look at rates of
change, or rates of change of rates of change? One possible answer to this question
is related to complexity of expressions. Often, understanding or testing how a
function changes with respect to an independent variable is easier than writing a
direct formula for the function in terms of the variable. Actually, the process is often
completely opposite to creating a formula for your function by some wizardry. Here
is (arguably) the main idea of differential equations: Suppose that we would like to
find a formula for a certain function y(t) which comes up in some application that
we would like to consider. If we know y(t0 ) for a certain value of t0 of t, and if we also
know how fast y(t) changes when we change t, then maybe we can reconstruct y(t)
from this information. The jargon for “how fast” y(t) changes is the rate of change
of y(t). If we know the rate of change of y(t) in terms of t only, then calculus tells us

3
what to do (integrate rate of change to find the change in y(t)). If we know the rate
of change in terms of both t and y, then we cannot do an immediate integration.
Rather, we need to find out a way to deduce the change from rate of change. The
process is called “solving the differential equation”, and this is the main issue of this
course.
Can we solve all differential equations by some general algorithms? The answer to
this question is a resounding “certainly not”. Even fairly innocent looking differen-
tial equations are extremely difficult or impossible to solve by an algorithm. There
are many families of equations which are simple enough so that we can solve them.
We will see several such families in this course. We tend to model reality with the
simplest things available first, so many of these simpler families occur repeatedly
in applications. However, going further in depth in many applications, one runs
into difficult puzzles about differential equations. For example in many practical
subjects of interest such as turbulence or finance, there are certain problems about
differential equations whose solutions have been resisting the attempts of scientists
and they have remained unknown for many decades.
One approach to understanding differential equations is to give up the task of find-
ing explicit formulas for solutions and rather focus on obtaining qualitative and/or
partial information about the solutions. Along with qualitative methods (exam-
ple: I cannot find the formula for the solution, but can I prove that the solution will
be bounded by a certain number?), the subject of numerical methods is of great
interest. These two branches are mainstream in the current research in differential
equations.

3 Solutions of first order ODE’s

Let us now concentrate on some simple families of differential equations. Let us
consider a single first order ODE. We have one dependent variable (say y) and one
independent variable (say t) related by a differential equation. A function y(t) that
satisfies the equation at all points t in an open interval (a, b) will be called a solution
of this ODE. The number a can be any finite number or −∞, similarly the number
b can be any finite number greater than a or +∞. The graph of y(t) is called a
solution curve for the ODE. We graph each solution curve over a connected open
interval; two disconnected graphs over two disjoint open intervals are considered to
be different solution curves, even if they can be expressed by the same formula.

4
Example 3.1 An ODE often has many solutions; typically infinitely many. For
example, consider the equation
dy
= sin(t).
dt
This is a very simple differential equation to solve, since on the left hand side we
have a first derivative and on the right hand side we have a function of t only - there
are no y’s that could make our job more difficult. This is actually a calculus question
in disguise. In order to find y(t), employ the fundamental theorem of calculus and
integrate the right hand side.
Z
y(t) = sin(t)dt = − cos(t) + c.

We obtained an infinite family of solutions, depending on the parameter c. When

we graph some of these solutions, we see that any two of these graphs differ by a
constant shift in the vertical direction. The graph below shows three solution curves
for three different values of c (and t belonging to the interval (0, 2π). Remember that
2π is slightly greater than 6).

Example 3.2 Let us next consider the differential equation

dy
= y.
dt
5
We cannot solve this example directly with the method used in the last example: If
we attempt to integrate the right hand side with respect to t, we need to integrate the
function y(t) that we don’t know yet. So we get stuck at this step. We will soon see
how to find all solutions of this equation systematically. For now just notice that the
equation asks for a function which is equal to its own derivative. We know such a
function from calculus: y(t) = et . This is one solution of the equation, but it is not
the only possibility: If we multiply this function by a constant, it is still a solution.
So y(t) = cet is a solution for any value of the constant c ∈ R. We again have a
family of infinitely many solutions depending on c. However, now the constant c
appears multiplicatively (recall that in the previous example it appeared additively).
Let us graph a few solutions again. The ratio of any two solutions is a constant (but
their difference is not constant this time).

Both of these examples above have infinitely many solutions, depending on a single
parameter c. If an additional condition on y(t) were given, for instance y(t0 ) = y0 ,
then the solution would be unique. In the first example this condition says
y(t0 ) = − cos(t0 ) + c = y0 ,
therefore c must be equal to y0 + cos(t0 ). So there is a unique solution y(t) =
− cos(t) + (y0 + cos(t0 )). In terms of the graph, this means that we are picking out

6
the solution curve passing through the point (t0 , y0 ).

Definition 3.1 Suppose that we have a first order ODE with dependent variable y
and independent variable t. A condition of the form y(t0 ) = y0 is called an initial
condition.

4 Direction Fields
As mentioned before, it is often rather difficult to find explicit solutions of an arbi-
trarily given differential equation. This is true even in the case of first order ODE’s.
For instance, one can safely bet that nobody on earth will ever be able to find
formulas for explicit solutions of a crazy looking equation like
dy 1
= y 3 + y cos(y) + sin(t3 ) + √ .
dt ty − y
Our aim in this section will not be to solve equations, but rather to say something
qualitative about the solutions without actually finding them. Suppose that we
isolate the derivative term in a first order ODE and write the equation in the form
dy
= f (t, y)
dt
where f is some two-variable function. Solving the equation is equivalent to finding
the solution curves. We expect to have infinitely many solution curves, depending
on a constant c in some way or another. Intuitively, the reason behind is that we
need to do “one integration” in order to solve a first order ODE, hence one constant
c will pop up.
We do not yet know what the solution curves are, but we have the following geometric
information about them: At each point (t, y) on the plane, the value of f (t, y)
equals the slope of the tangent line to the solution curve passing through this
point, simply because it is equal to the value of the derivative dy/dt. If we had the
solution curves, we could easily draw these tangent lines. But the situation is just
the opposite: We know what the tangent lines look like and we want to reconstruct
the solution curves. So we can do the following:

• First, plot line segments having slope f (t, y) at (t, y) for as many points (t, y)
as possible. Of course, since there are infinitely many points on the plane we

7
 cannot plot them all, but by choosing a fine mesh we can draw a large number
of them. It is important to notice that we do not need to solve the ODE for
this purpose but we just need to tabulate the values of f (t, y) at these points.

• Then, sketch approximate solution curves of the differential equation. The key
point is that they must be tangent to the line segments constructed above at
each of their points. Again, this cannot be accomplished fully, but we can do
it approximately in a satisfactory way.

This will give a rough idea about what the solution curves of the differential equation
look like. Notice that at the end of this process we will not have formulas for the
solution curves. Nevertheless, we will obtain their approximate graphs. The initial
picture that we get after drawing the line segments is called a direction field (or
a slope field).
There is a practical method which sometimes makes the plotting of the direction
field easier: If the curves of the form f (t, y) = c are easy to draw, then we can place
them on the plot as a preliminary step (in a way to be erased later on). Along such
a curve, the slope of the tangent line will be equal to c throughout, so all of the slope
lines will be parallel to each other along this curve. This observation significantly
reduces the effort necessary to plot the direction field.

Example 4.1 Consider the differential equation

dy
= y(y − 1).
dt
Here, f (t, y) = y(y − 1) and it is special in the sense that it depends only on y but
not on t. Let us find f (t, y) for a few values of y, listed in the table below:
y f(t,y)
-0.5 0.75
0 0
0.5 -0.25
1 0
1.5 0.75
This means that in order to draw the direction field, for any point with y-coordinate
−0.5, we should place a line segment of slope 0.75 etc. Expand the table by making

8
this calculation for more values of y, maybe by using a computer. Plotting the line
segments with the calculated slopes results in the direction field below:

This plot was actually obtained using the subroutine “dfield8” on Matlab, but it can
easily be drawn by hand, given enough time. Now let us plot a few solution curves on
the same graph; the rule is: “follow the direction field”. The solution curves should
be sketched in a way that they are tangent to the line segments at each point.

9
Using this sketch, we can say several things about the solution curves even though
we don’t have a single formula for any solution curve at this point:

• Solutions that have an initial value y(0) between 0 and 1 remain between 0 and
1 and tend to 0 as t → +∞. These solutions are monotone decreasing.

• Solutions that have an initial value y(0) larger than 1 tend to infinity (actually
even before t has a chance to march off to infinity, but this is more difficult to
observe here. Let us discuss this later).These solutions are monotone increas-
ing.

10
 • Solutions that have an initial value y(0) less than 0 tend to 0 when t → ∞.
These solutions are also monotone increasing.

Of course, these are observations rather than proofs and care must be taken before
they are seriously used. For instance, after this observation, one can go back and
carefully prove that the solutions are monotone increasing for y > 1 by looking at
the sign of y(y − 1).

5 Separable Equations
Now, let us explicitly solve some special types of ODE’s.
As a first attempt for solving a first order ODE, it is enticing to try integrating both
sides of the equation. But this method dramatically fails for a general equation of
the form dy/dt = f (t, y): If f (t, y) is going to be integrated with respect to t, we
need to know what y is in terms of t. But this requires the knowledge of the solution
of the problem itself, so there is a circular reasoning here and it doesn’t work.
Separable equations form a special family of ODE’s where we can somehow alge-
braically separate the two variables.

Definition 5.1 Suppose that f (t, y) can be written in the form M (t)/N (y) for some
functions M and N . Then the differential equation
dy M (t)
=
dt N (y)
is called a separable differential equation.

We can find all solutions of a separable ODE as follows:

dy M (t)
=
dt N (y)
dy
N (y) = M (t)
Z dt Z
dy
N (y) dt = M (t)dt
dt
Z Z
N (y)dy = M (t)dt

11
(We can pass to the fourth equation from the third one by using the chain rule.) In
the last line, the left hand integral is purely with respect to y and the right hand
integral is purely with respect to t. Therefore, in principle both can be evaluated
and we obtain a relation between y and t. This is an implicit relation. In some cases
we can explicitly solve for y(t), but in general an implicit relation is all that we can
find.
It is a fact that in many examples the integrals obtained in this process might be
difficult to find, and even impossible sometimes. This we regard as some other type
of difficulty irrelevant to our discussion. In the worst case scenario, one can evaluate
the integrals by symbolic or numerical integrators.

Example 5.1 Find all solutions of the ODE

dy 2 2
= tet −ln(y ) .
dt

Solution: The equation is scary looking at first sight, but with a small algebraic
manipulation we can rewrite it as
2
dy tet
= 2
dt y
therefore it is separable. Multiply both sides by y 2 and integrate with respect to t.
Z Z
2
y dy = tet dt
2

2
y 3 /3 = et /2 + c
 2 1/3
y = et /2 + c

There are a few remarks to be made: The integral on the right hand side can be
obtained by using the substitution u = t2 . Important: There are two indefinite
integrals, yet one c is enough since equality of the integrands implies that the an-
tiderivatives differ by a constant. Also note that the constant c appears in the final
result in an awkward place; it is neither additive nor multiplicative. We should just
leave it where it is.

Example 5.2 Find all solutions of the ODE

dy
= y(y − 2)t.
dt
12
Solution: We may rewrite the equation as
dy t
=
dt 1/y(y − 2)
but some care is needed. The two equations above are equivalent if and only if y
is different from 0 or 2. We will take care of these cases separately. Now, for a
moment assume that y 6= 0 and y 6= 2. Then, as before,
Z Z
dy
= tdt
y(y − 2)
Z Z Z
dy dy
− = tdt
2(y − 2) 2y
1 t2
(ln |y − 2| − ln |y|) = + c
2 2
y−2
ln = t2 + c
y
y−2 2
= cet
y
2
y=
1 − cet2
A few remarks are in order: To pass from the first line to the second line, write
1 A B
= +
y(y − 2) y−2 y
and solve for A, B. A second issue: The reader is probably annoyed by now by the
careless use of the constant c: Passing from line 3 to line 4, the constant c is multi-
plied by 2, but it still takes all possible real values, therefore the new constant can be
written as c instead of 2c. Passing from line 4 to line 5, everything is exponentiated,
so we should have written ec instead. Lifting the absolute values gives ±ec . This is
again an arbitrary constant, it can take all real values except for 0. So we should
note that in the last equation c is an arbitrary constant different from 0.
What about y = 2 and y = 0? These constant functions are both solutions because
d(2)/dt = 2(2 − 2)t and d(0)/dt = 0(0 − 2)t are correct identities. Actually, the so-
lution y = 2 corresponds to the missing value c = 0 in the formula above. However,
the solution y = 0 does not correspond to any special value of c (unless one prefers
to think about it as the degenerate case c = ∞).

13
There is a subtle point about the argument concerning y = 2 or y = 0. You may
rightfully ask “what if y(t) 6= 2 for some value t and y(t) = 2 for some other value
t?” After all, y is a function; it is not a number. This possibility seems to create
more cases which are not discussed in the argument above. The answer (which I
think would be unclear for the reader at this stage) is that everything is alright since
there is a fact saying that two different solution curves cannot intersect at a point
(under mild hypotheses). However this fact depends on a deep and important result,
the existence-uniqueness theorem. We will talk about this theorem later.

6 Homogenous Equations
Some first order ODE’s are not separable, yet they become separable after a simple
substitution. In this section we will discuss one family of such examples: homoge-
nous equations.

Definition 6.1 A first order ODE is said to be homogenous if it can be written

in the form
dy y 
=h
dt t
for some function h.

The given form of an equation may be deceptive and this can make it tricky to check
whether the equation is homogenous. A quick test for homogeneity is as follows: If
dy/dt = f (t, y) = h (y/t), then f (λt, λy) = h ((λy)/(λt)) = h (y/t) = f (t, y) for any
λ 6= 0. Therefore, if f (λt, λy) 6= f (t, y) even for one value of λ, then the ODE is not
homogenous.

Example 6.1 Let

dy
= t + y.
dt
Then f (λt, λy) = λ(t + y) 6= t + y for almost all values of λ. Therefore this ODE is
not homogenous.

Example 6.2 Let

dy t+y
= .
dt t−y

14
Then, we can rewrite this equation as

dy 1 + y/t
= .
dt 1 − y/t
1+v
If we set h(v) = then
1−v
dy y 
=h
dt t
therefore the equation is homogenous.

Suppose now that we have a homogenous equation dy/dt = h (y/t). Substitute

v = y/t. The right hand side will clearly be h(v). In order to compute the left hand
side, notice that y = vt. By the product rule for differentiation,

dy d(vt) dv
= =v+t
dt dt dt
Therefore the equation can be written as
dv
v+t = h(v)
dt
dv h(v) − v
=
dt t

This final equation is separable, hence we can solve it as before.

Example 6.3 Let us solve the equation

dy 1 + y/t
= .
dt 1 − y/t

15
 1+v
Setting h(v) = as above, we obtain
1−v
dv h(v) − v
=
Z dt Z t
dv dt
=
h(v) − v t
Z Z
dv dt
1+v =
1−v
−v t
(1 − v)dv
Z
= ln |t| + c
1 + v2
1
arctan(v) − ln |1 + v 2 | = ln |t| + c.
2

Finally, plugging in v = y/t gives us an implicit relation between y and t. It seems to

me that it should be extremely difficult to write y solely in terms of t in this problem,
so let us leave it like this.

16