1 First-Order Equations

In this chapter we focus on problems that are given by equations involving some unknown
function and its derivative. For historical reasons they are called differential equations even
though they are usually formulated in terms of some rate of change that describes or models
important physical processes such as growth and decay.

Model: Natural Growth

The equation xN = ax describes the process of natural growth. In this model the quantity
x(t) grows at a rate proportional to itself. The constant a is often called the relative growth rate.
We know that the exponential function is the only possible solution since

We may also find the integral or antiderivative, where

Notice that we have redefined the arbitrary constant C, which indicates that we have a one-
parameter family of solutions or curves (shown).

1.1 Equations and Solutions

One way to classify a differential equation is by the order, indicated by the highest
derivative present. The natural growth model yN = ky is an example of a first-order equation. It
is also a linear equation since the variable y occurs only as itself, not as a power or any other
function.

Lesson 1: The solution (function) of any equation must satisfy that equation.

Example 1 Consider each of these functions.

(A)
The derivative of this function

We have recovered the original equation, hence y is a solution.

(B) Given yN = -t/y, we can verify the solution

since we have the identity

1.1 Equations and Solutions

# For practice, verify

An initial value problem (IVP) consists of a differential equation yN = ƒ(x,y) along with a
given (initial) value of the unknown function y(a) = b. The purpose of the initial value is to
specify the arbitrary parameter. The general solution y = F(x) + C is a family of solution curves,
while the particular solution is the curve passing through the point (a,b). Recall from basic
calculus that yN = 3x² has many solutions given by y = x³ + C. The IVP yN = 3x², y(0) = 1 has the
particular solution y = x³ + 1.

Example 2 We can show the general solution of the equation yN = y² is

since the derivative

The condition y(1) = 2 states that x = 1 and y = 2, so that

or c - 1 = 1/2 and c = 3/2. The particular solution then becomes

Notice that the interval of existence (domain) is (-4,3/2) in order to

satisfy the initial condition, the point (1,2), which is the left branch of
the graph. The right branch passes through the point (2,-2) and
becomes the particular solution of the problem yN = y², y(2) = -2.
# For practice, verify (-½,4)

A simple equation can display some interesting features, for example the IVP

clearly has no solution since y = ln*x* + C (below left) can’t satisfy the initial condition if x = 0. A
problem may have more than one solution for some initial condition, for example

as shown in the graph below (right).

1.1 Equations and Solutions

An important feature of the uniqueness property is that if two functions ƒ(x) and g(x) are
both solutions of the same equation and they share the same initial condition ƒ(a) = g(a) = y0, then
uniqueness guarantees that ƒ(x) = g(x). From a graphical point of view, we require that solution
curves should not intersect (see graphs above).

Application: Torricelli’s Law

Suppose that a tank has a hole in the bottom and the contents begins to leak out at a rate
(liter/min) determined by Torricelli’s Law, where y(t) is the depth and the volume is given by

For a cylindrical tank, we assume V = Ay and that

hence we have the problem

where c = k/A. We can verify that this IVP has a solution

Notice that if y0 = 0, we have two possible solutions

The first solution y1 implies that the tank was always empty while the second solution y2 suggests
that the tank was drained when t < 0 and then remains empty.

PS1.1
 1.2 Solution Curves

When the solution of an equation can be specified by some formula, we have found the
analytical solution. We may also consider the slope function yN = ƒ(x,y) that generates a series of
tangents to the solution curves at each point, forming the direction or slope field. This method
provides a qualitative solution for many first-order equations that may be difficult or even
impossible to solve by analytical methods. The direction field suggests the shape of each solution
curve. Consider the natural growth model yN = ky for the values k = 2 (left) and k = -1 (center).

The special equation yN = ƒ(y) is called autonomous or self-governing. The absence of

the independent variable (time) has important consequences that are best illustrated by comparing
the slope fields of the equations yN = ky and yN = y + t² (right). If we compare any two near-by
solution curves, corresponding to different initial conditions, then the autonomous equation will
always produce “parallel” graphs with the same slope for a given value of y(t), since the rate of
change depends only on the amount y and not on the time t.

The constant values of the slope function yN = k describe a family of curves called
isoclines which indicate where the slopes of the tangents to the curve are the same (iso).

Example 1 Consider the direction field for

The direction field of this equation demonstrates that the isoclines are
the oblique lines t - y = k. The flow of the field is used to sketch some
solution curves, including the curve through (say) the point (2,1). A
close examination of this field suggests that there is a special linear
solution y = t - 1, which the other curves seem to approach as t ÷ 4,
known as the asymptotic solution, denoted by

If we examine the second derivative, where

we observe that yO = 0 along this line, yO < 0 (concave down) below

the line and yO > 0 (concave up) above the line.
# For practice, let
1.2 Solution Curves

The constant solution y(t) = c of an equation yN = ƒ(t,y) is called an equilibrium solution,

which is stable given

For example, the equation yN = -2ty + t has one (constant) equilibrium solution y(t) = ½.

Application: Population Growth

The natural growth model PN = aP can be modified to limit the growth (decay) of the
population P(t), where PN = aP - bP². If we choose a = 0.06 and b = 0.0004, then we have the
special autonomous equation

This equation has two equilibria P = 0 and P = 150, however

is the limiting value or stable equilibrium as shown in the graph.

Example 2 Consider

The solution curves (top) of this non-autonomous equation are not

parallel. The family of parabolas y² - t = c or y² - c = t along which
the slope dy/dt = yN is constant are the isoclines. Observe that there
are no constant solutions (equilibria) y = k when the slope yN = 0,
however, as t ÷ 4 there are asymptotic solutions that split along the
curve y² = t: some “funnel” along the upper half

while others drop off along the lower half of the curve:

# For practice, try

Approximation: Euler’s Method

Thanks to modern graphing software we don’t require the analytical solution of an IVP to
plot its direction field and solution curves. Most of these programs instead rely on some
numerical method that generates a set of discrete values

that will approximate the values of the function y(x) that satisfy the IVP

The simplest scheme is the linear (tangent) approximation of this function at some point (a,g(a))
1.2 Solution Curves

which provides the connection between one data point and the next point, where the jump from
one point to the next is indicated by the step size

so we obtain an iterative formula

called Euler’s method.

Example 3 Consider the approximate the solution of

Beginning with the initial values xo = 0 and yo = 1, we compute the next

few approximations:

The exact (analytical) solution of this equation is the function

If we compare the approximations with the actual function values, we

get a sense of the error in this method.

For greater accuracy, we could reduce the step size h, which would
increase the number of calculations over a given interval. In that case,
the algorithm may be carried out by a calculator or computer.

# For practice, let

Example 4 The challenging nonlinear problem

has an approximation

With a step size (say) h = 0.1 the first three iterates are
1.2 Solution Curves

Continuing in this manner, we find some additional values.

Rather than accept the results with h = 0.1, we repeat this process with
even smaller step sizes h = 0.02 and 0.005 to obtain some alternative
approximations. Notice the rapid divergence in these approximations,
implying that Euler’s method becomes unstable as x increases. There
seems to be a possible asymptote in the solution curve as x ÷ 1, which
is predicted by the smallest step size h = 0.005.
# For practice, let

Other features of a solution may not be accurately predicted by Euler’s method, regardless
of the step size h or number of steps n. Consider the IVP

This equation has a typical solution

however this method has difficulty matching the graph’s oscillations.

Application: Velocity Model Revisited

Suppose that a baseball is dropped from a helicopter so that its velocity after t seconds is
once again determined by the initial value problem

Using Euler’s method, the discrete form of this linear equation is

We then generate the following approximate velocities for 0 # t # 10 and compare with the actual
solution v(t), obtained earlier.

Notice that the terminal velocity (equilibrium) is 200 ft/sec and that the ball has attained about
80% of this value after 10 seconds.
PS1.2
 1.3 Separable Equations

We express the natural growth model xN = kx in its integral form

where we used the typical formula dx = (dx/dt)dt to simplify the left side. We have created a
separable or differential equation that can then be integrated on both sides, a process that we
extend to other similar equations by matching functions with their appropriate differentials.
Example 1 Solve:
Separation of variables yields the differential equation

which has the integral form

There are many solutions, each depending on the value of C. There is

also a special or singular solution y = 1 that was lost when the
variables were separated. No value of C corresponds to this solution,
nor does it correspond to any limiting value since

# For practice, solve

It is often difficult to find an explicit solution (function) y = ƒ(x) that solves an equation.
We therefore must be content with the implicit form F(x,y) = 0.

Example 2 Solve:

Given yN = dy/dt, this equation has the separable form

The solutions or integral curves then become

or after some simplification

These implicit curves for various k values (initial conditions) are

shown (the “duck” graph).
# For practice, solve
1.3 Separable Equations

Example 3 Solve

We separate this equation and obtain the differential form

The implicit general solution becomes

where if x = 1 and y = 3, then C = 9. Instead of trying to graph this

equation, we consider the direction field and observe that the particular
solution (curve) through the point (1,3) is only defined on the interval
(-1,5). The solution curve through (say) the point (1,0) appears to be
defined on the interval (0,4), while we seem to have a complete curve
through the point (1,-2). Notice that the curves are separated by the
asymptotes y = ±/(5/3) where the slope dy/dx becomes infinite.
# For practice, let

This method should remind the reader of implicit differentiation- we simply carry out the
reverse process, beginning with the derivative yN = ƒ(x,y) and finding an implicit form F(x,y) = C.

Example 4 Consider a familiar case given by the IVP

In differential form, we have

which can be integrated on both sides to produce the implicit solution

The initial condition x = 0 and y = 1 requires that c = ½ and we get the

upper half circle x² + y² = 1, y $ 0.
# For practice, solve

The growth (decay) of some quantity may be expressed by the general equation

The solution is found by the typical separation-integration scheme:

Another method for solving this equation is to make a change in the dependent variable, say u =
ax + b, then by the chain rule
1.3 Separable Equations

which allows to recover the natural growth model uN = au. By inspection, the general solution is

where u = ax + b. Solving for x, we recover the expected solution

Application: Terminal Velocity

An immediate application of this result is the problem of an object released from 3000
feet, where the velocity v(t) was determined by

With a = -0.16 and b = 32, we obtain the general solution (velocity)

where the initial condition produces the particular solution

Notice that as t ÷ 4 we have the terminal velocity (stable equilibrium) 200 ft/sec.

This technique can be applied to any equation yN = F(ax + by + c), where the substitution u
= ax + by + c should reduce the equation to a separable or integral form.

Example 5 Solve
Following this scheme, we make the “obvious” changes

The second form (equation) uN = 1 + u² is separable, as promised, so

We have the implicit solution

expressed in terms of the original variables as

# For practice, solve

Project: Brachistochrone

The inverted cycloid (shown) has a slope

A bead sliding down this curve would require the least amount of time.

PS1.3
 1.4 Integrating Factors

We know that the equation dy/dx = 2xy is separable

and has an implicit solution ln*y*= x² + c, which we may also write as

Notice what happens when we differentiate the second version:

which simplifies to the homogeneous form of the original equation yN- 2xy = 0. The exponential
is the integrating factor ì since it allows to write the left side as a derivative.

To solve the non-homogeneous equation yN+ p(x)y = q(x) we would reverse this process:

where by assumption the left side must also be the derivative

This condition simplifies when y 0 to the separable form µN = µp(x), or

which has the familiar solution ln ì = Ip(x)dx, so we have the integrating factor

then (ìy)N = ìq(x), so finally the general solution becomes ìy = Iìq(x)dx.

We had the general growth (decay) equation xN = ax + b or nonhomogeneous form

The integrating factor here is easily determined:

so, after multiplying through by µ, the equation becomes

The integral of each side gives the expected general solution:

Example 1 Solve
1.4 Integrating Factors

We recognize that

so there is a similar integrating factor:

We now multiply this equation by this factor, so that

which we recognize has the form

The integral of both sides gives a general solution

# For practice, solve

Example 2 Solve
We first divide through by the expression x² + 1 and obtain the
standard form

with P(x) = 3x/(x² + 1) and Q(x) = 6x/(x² + 1), so the integrating factor

Multiplying the equation by this factor will produce

which has an integral

Finally, we obtain a general solution

# For practice, solve

In these examples we should notice any general solution is always the sum of two
functions: the solution of the homogeneous equation plus another function. A more powerful
version of this method called the variation of parameters can also be applied to the equation
1.4 Integrating Factors

and its homogeneous or separable form yN + p(x)y = 0 which has a complementary solution

Replacing (varying) the parameter c with an arbitrary function, we try a particular solution

that satisfies the nonhomogeneous equation

The general solution is the sum of the complementary and particular functions

which is equivalent to the solution obtained with the integrating factor:

If we return to the model xN - ax = b, we note that the complementary solution

suggests that we try a particular solution

which produces the equation

Our mystery function is then the integral

The complete general solution takes the familiar form

Example 3 Solve
When we modify the homogeneous solution

this equation becomes

Upon simplifying, we have the mystery function

and a particular solution

Finally, the general solution y = yc + yp .

# For practice, solve
1.4 Integrating Factors

Application: Mixture Problem

We have a tank or body of water with a given volume V that contains a certain amount of
a solute (pollutant) x(t) measured with appropriate units (grams, pounds, etc). The concentration
is the function c(t) = x(t)/V(t). If the rates of flow (volume/time) in and out of the tank are given
as ri and ro, the change in the amount of solute can be expressed as

Dividing by ªt, taking the limit as ªt ÷ 0, we get the linear equation

In summary, the rate of change dx/dt = RATE IN - RATE OUT.

Suppose a large tank contains 1000 L of pure water and a brine solution begins to flow
into the tank at a rate of 6 L/min. The resulting solution is uniformly mixed and flows out of the
tank at the same rate. If the concentration of salt entering the tank is 1 kg/L, we seek the amount
of the salt in the tank x(t). Applying this model with consistent units, we get

We can once again take advantage of the form xN = ax + b, with a = -3/500 and b = 6, so that

The initial condition x(0) = 0 gives C = -1000, so finally we have

Application: Changing Volume

A 120 gal tank initially contains 90 lb of salt dissolved in 90 gal of water. Brine
containing 2 lb/gal of salt flows into the tank at a rate of 4 gal/min, and the mixture flows out of
the tank at the rate of 3 gal/min. So, the volume is no longer constant, but given by the function
V(t) = 90 + t. Applying the above model, we get an equation for the rate of change (lb/min):

In the typical linear format xN + Px = Q, we would obtain the integrating factor

that produces a general solution determined by

Finally, the initial condition x(0) = 90 gives C = -904.

PS1.4
 1.5 Population Models

The key model for the study of first order equations has been the natural growth equation
dy/dt = ky, which has a simple, but unrealistic increasing solution for t > 0. A more accurate
description of the growth of a population takes into account the birth rate â and death rate ä:
â(t) is the number of births per unit of population per unit of time,
ä(t) is the number of deaths per unit of population per unit of time.
The change in population p over an interval of time ªp = births - deaths, or

so taking the usual limit as ªt ÷ 0, we get the general population model

Observe that the natural growth model pN = âp has a constant birth rate â and death rate ä = 0.

Example 1 Suppose that an alligator population p(t) initially has 100 members and
that the death rate ä = 0, while the birth rate â = 0.0005p. This
information produces the nonlinear initial value problem

This equation is easily solved by separating variables, so that

Using the initial condition p(0) = 100 gives C = -1/100, so finally the
particular solution

Like the natural growth model, as time increases, in this case t ÷ 20,
the gator population “blows up” as shown. The nonlinear nature is
illustrated by the different “gaps” in the curves at each time t.
# For practice, solve pN = kp², p(0) = 10, p(10) = 24.

An important example of limited growth is the logistic model. In order to correct the
unbounded natural growth problem, we assume that the birth rate will decrease as the population
increases, so â = á - bp, for some positive constants á and b. If the death rate ä remains constant,
the general population equation takes the nonlinear form

or the simpler version pN = ap - bp². This equation is often expressed in the more useful form
1.5 Population Models

We may now interpret this equation as a correction of the natural growth model dP/dt = rP, which
assumes a constant relative growth rate r = PN/P. The logistic model assumes a rate of change
that varies jointly with the population P and the capacity for growth M - P, where M is limiting
value or stable equilibrium of the system. Other textbooks often consider the “fractional”
population and express the logistic model in the alternative form

where they have set L to be the carrying capacity.

The logistic model, like any autonomous equation xN = ƒ(x) is separable and may be
integrated directly or with the use of partial fractions.

Example 2 We earlier had the equation PN = 0.0004(150 - P)P, which is separable

so that using partial fractions on the left side we get

The integral of each side produces the implicit solution

which after some annoying algebra becomes the general solution

Notice that the limit of this function

matches the maximum value M in the equation. The graph shown

illustrates the particular solution for various initial values Po.
# For practice, solve
Rather than repeat this exhaustive process for each case, we could derive the particular
solution of the general problem PN = rP(M - P), P(0) = Po as

Application: Rumors

A rumor has started in some city with a population M = 100 thousand people. The rumor
1.5 Population Models

begins with 10 thousand people and after 1 week has spread to 20 thousand. The limited number
of people in this city suggests that we have a logistic function

The condition P(1) = 20 permits us to determine the factor r = ln(9/4)/100 . 0.008109.

Application: Threshold Model

A modification of the logistic model is the threshold equation

where the positive dominant quadratic term suggests that the population might expand out of
control. If we return to a familiar problem similar to the second example, where now

we should observe that r < 0 in the logistic solution. Thus, we immediately obtain the solution

The number M = 150 is the threshold value since there will be two different outcomes depending
on whether the initial value P0 < 150 or P0 > 150. For example, with P0 = 100, then

so as t ÷ 4, then P ÷ 0 or extinction. With, say, P0 = 200, then

will explode as t increases, an event often called doomsday.

If we only wish to know “what happens” in the logistic model or any similar problem we
don’t need to grind out the analytic solution. The limiting value or equilibrium solution

is already present in the equation xN = kx(M - x), where xN = 0 gives two equilibria x = 0 and x =
M. The stable value as determined by the sign of the derivative, which depends on the initial
condition: when x(0) > M and xN < 0, then x(t) will decrease and approach x(t) = M, while if x(0)
< M and xN > 0, then each solution curve x(t) will increase towards the (stable) equilibrium.
1.5 Population Models

A distinct feature of the logistic or S-curve (left) is the inflection point at x = M/2, where
xO = 0 (verify). The stability of each equilibrium is illustrated by the “flow” along the phase line
(right). The flow into a point (solution) implies stability, while the flow out instability.

Example 3 Describe the equilibria.

(A)
We can also think of the equation xN = ƒ(x) as a typical two-
dimensional plot with a horizontal axis or phase line x(t). In this case
ƒ(x) = x² - 1, so we have a parabola with two x-intercepts that define
two equilibrium solutions x = ±1. Since xN > 0 for x < -1, the flow is to
the right; when -1 < x < 1, then xN < 0 so the flow is to the left; and
once again xN > 0 for x > 1, so the flow continues to the right. Hence,
x = -1 is the stable equilibrium and x = 1 is the unstable value.

(B)
There are two distinct equilibria x = 0 and x = 1. Notice that the
derivative only changes sign at x = 0 from xN < 0 to xN > 0, so this value
is unstable. Near x = 1 the derivative xN > 0, so the flow is “through”
that value, making that equilibrium semi-stable. These features are
shown in the given graph by the direction field and solution curves.

# For practice, consider {0,2,unstable;1,stable}

Application: Harvesting Model

The logistic model can be modified by allowing for the harvesting of a population at
regular time intervals, given by the equation

Notice that this model predicts the usual limiting value x = m for no harvesting (h = 0). The right
side is quadratic suggesting that we may rewrite this model as

with a stable limiting solution x = N and an unstable threshold solution x = H. Suppose that we
set k = 1, m = 4 and choose the “tuning” parameter h = 3, then xN = x(4 - x) - 3 has equilibrium
solutions H = 1 and N = 3. If this model were counting (say) how many hundreds of fish x(t)
were in a lake, then the (minimum) threshold would be 100 fish with an equilibrium value of 300
fish, given a constant harvesting rate h = 300 fish/year. If the number of fish drops below 100,
then eventually all the fish would disappear from the lake.

If we let the parameter h = 4, the nature of the equilibria changes since x(4 - x) - 4 = 0 has
only one real solution x = 2. If h > 4, there are no equilibrium solutions. The equilibrium
condition x(4 - x) - h = 0 is the graph of a parabola
1.5 Population Models

which has a critical point (4,2) where the equilibrium x = c changes.

Project: Bifurcations

The behavior of an autonomous equation may often depend on the value of some
parameter k. The number and the nature of the equilibria will change according to this parameter
as in the harvesting model. This change in the equilibrium solutions is called a bifurcation, a
word that literally means to divide or split into parts. In the harvesting model above, the value h
= 4 is the bifurcation point since the nature of the solutions change as h ÷ 4. We may explore
various models and their bifurcation diagrams that predict other types of behavior.

PS1.5