Aplicaciones Ecuaciones Diferenciales PDF

DIFFERENTIAL
EQUATIONS
In the simplest model of population growth, the growth rate at any

time is proportional to the population size at that time. H o w would
we express such a situation mathematically? L e t N(t) denote the
size of a population at time t, t > 0. T h e n
' dN . '
— = rN(t), t>0 (8.1)
(We introduced this equation in Section 4.6.) E q u a t i o n (8.1) con-

tains the derivative of a function and such equations are there-
fore called differential equations. Modeling biological situations
frequently leads to differential equations.
Differential equations can contain any order derivative; for
example,
^1 + ^1
1 xy
dx dx
is a differential equation that contains the first and second deriva-

tive of the function y = y(x). I f a differential equation, such as
(8.1), contains only the first derivative, it is called a firsí-order
:
differential equation. ' ;" ' ;,V •
Throughout this chapter, we will restrict ourselves to/first
order differential equations of the form «•
-\ d¿ (8.2)
' dx /OOíOO
The right-hand side of (8.2) is the product of two functions; one
depends only on x, the other only on y. Such equations are called :
separable ditíerential equations (the reason for this ñ a m e will

become clear shortly). T h i s type of differential equation inciudes"
two special cases, namely, ', •-, ii
: )
' = / W . • 0 : (8.3)
dx
and
dy
?G0 (8.4)
dx
We discussed differential equations of the form (8.3) Ln Section 5.8.'

Differential equadous .K the form (8.4) include (8.1) and aré.:
v ;i
frequently used in bioíoîcaj modcls.''V.U% -::-'í -' *. r<-¡3gffi $
474 ChapterB • Differential Equations
8.1 SOLVING DIFFERENTIAL EQUATIONS
L e t ' s return to the growth model i n (8.1), and, to be concrete, let's choose
r = 2. T h i s results in the equation
dN
á b — =2N(t), !>0 (8.5)
i ró J
We are interested in finding a function N(t) that satisfies (8.5). Such a function
is called a solution of the differential equation. W e claim that with NQ = A'(0),
Z-t
2
N(l) = N e ',
Q í>0
¿JE. t ^ ~ ^ í is a solution of (8.5). T o check this, we differentiate N(t):

IOS - l
. —
2N(t)
A t any point (í, N(t)) oí the graph of N{t), the slope is equal to 2N(t) (see
Figure 8.1). _. (¡lf i ,
A first-ofrjer differential equation tells us what the derivative of. a
function is.. I n order to find a solution, we must therefore i n t é g r a t e . (Since it
is not always possible to i n t é g r a t e a function, it is not always possible to write
the solution of a differential equation in explicit form.)
W e begin w i t h a general method for solving differential equations of the
form (8.2)
dy
(8.6)
dx
We divide both sides of (8.6) by g ( y ) [assuming g(y) j¿ 0]:
> *< \
1 d
(8.7)
= / w
g(y) dx
Now, i f y = u(x) is a solution of (8.7), then ii(x') satisfies
1
u'(x) = f(x)
[«(*)]
and,therefore,
u(x)dx = / f{x)dx
Mi)
Slope 2N{t)
NU) = N e* 0
A R-JV.N? 8.1
Thr. ;r>kiion A/(í) 2
A'o* *: Tlse slope at (r, N ( 0 ) is 2N(t)
8.1 B Solving Differential Equations 475
B u t , since g[u(x)] = g(y) and u'(x) dx - dy, this is the same as
dy= [ f(x)dx
J iiy)
T h i s suggests the following procedure of solving separable differential
equations. Namely, we s e p á r a t e the variables x and y so that one side of the
equation depends only on y and the other side only on x. T o do so, we treat
dy and dx as if they were regular numbers: Writing (8.6) formally as
dy
-y- = f(x)dx
ând then integrating both sides
* ' C dy f
i / - 7 T = / fWdx
f¡ J g(y) J
yields equation (8.8). T h e method of separating the variables x and y works
since the right-hand side of (8.6) is of the special form f(x)g(y), which gave
this type of differential equation its ñ a m e . A note of caution: W e simply
divided (8.6) by g(y). W e must be careful, since g ( y ) might be 0 for some
v a l ú e s of y, in which case we cannot divide. W e will address this problem
in Subsection 8.1.2. I n the next two subsections, we discuss how to solve
differential equations of the form (8.3) and (8.4); we give an example of the
general type (8.2) in Subsection 8.1.3.
8.1.1 Puré Time Differential Equations

I n many appíications, the independent variable represents t i m e . T f the rate of
change of a function depends only on time, we cali the resulting differential
equation a p u r é time differential equation. Such a differential equation is of
the form
^ = f(x), x 6 / (8.9)
ax
where / is an interval and x represents time. W e discussed such equations in
Section 5.8, where we found that the solution is of the form
y{x) = f f(u)du + C (8.10)

JXQ
T h e constant C comes from finding the general antiderivative of / ( . r ) ; the
number XQ is in the interval / . T o determine C , we must phrase the problem
as an initial valué problem (see Section 5.8); if we assurne 3'.(A'O) = yo, then
plugging xn into (8.10) yields y(xo) =t C and tnus C = y . The. solution can 0
then be written as
y{ ) x = yo + / / ( « ) du
T o solve (8.9) formally, we s e p á r a t e variables; we write the differential

equation (8.9) in the form
dy = f(x)dx
and then i n t é g r a t e both sides:
f
dy= Lf(x)dx
476 Chapter 8 • Differential Equations
• Figure 8.2
The solution V (r) = 4 - eos r in Example 1
or
y(x) = J f(x)dx
w h i c h is the same as (8.10).
EESSEEUÍ Suppose that the volume V (/) of a cell at time t changes according to
dV_
= sinf withV(0)=3
dt
.C!m = Find V(t).
M = Solution
4 Since
V ( r ) = V ( 0 ) + / s i n n du
Jo
we find
V(0 = 3 + [-cosu] 0
= 3 - f ( - eos t + eos 0)
= 4 — eos t
since eos 0 = 1 (see F i g u r e 8.2). . • *^
I f w e changed the initial condition in E x a m p l e 1 to some oth'er v a l u é ,

the graph of the new solution would be obtained from the o í d solution by
shifting the solution vertically so that it would satisfy the new initial condition
(see Figure 8.3). ( T h i s was discussed in Section 5.8.)
8.1.2 Autonomous Differential Equations

Many of the differential equations that model biological situations, such as
(8.1), are of the form
dy
(8.11)
dx
where the right-hand side of (8.11) does not explicitiy depend on x. These
equations are called autonomous differential equations. |
8.1 ¡S Solving Differential Equations 477
V(l)
8
7
6
5
ÍA i'i
3
2
1
0
TT 27T í '
X Figure 8.3
The curve V(t) = 7 — cosí is obtained from V(t) = 4 — cosr by a vertical shift. The
function V(f) = 7 - eos t solves the differential equation in Example 1 with V(0) = 6
T o interpret the biological meaning of autonomous, let's return to the

growth model
dN
— = 2N(t) (8.12)
at
W e will see below that the general solution of (8.12) is
2
N(t) = Ce ' (8.13)
where C is a constant that can be determined if the population size at
one time is known. A s s u m e we conduct an experiment, where we follow a
population over time, w h i c h satisfies (8.12) with N(0) = 20. U s i n g (8.13), we
find N(Q) = C = 20. T h e n the size of the population at time t is given by
2
N{t) = 20e '
I f we repeat the experiment at time t = 10 (say) with the exact same initial
population size, then, everything else- being equal, the population evolves in
exactly the same w a y as the one starting at t = 0. U s i n g (8.13) now with
2 ü 2
A/(10) = 20, we find A/(10) = C e = 20 or C = 20e- °. T h e - s i z e of.the
population is then given by
20 2! 2 W)
N{t)=20e- e =20e «-
T h e graph of this solution can be obtained from the previous graph, where
N(0) = 20, by shifting the oíd graph 10 units to the right to the new starting
point (10, 20) (see F i g u r e 8.4).
T h i s means that a population starting with N = 20 follows the same
trajectory, regardless of w h e n we start the experiment. T h i s "makes sense
biologically: I f the growth conditions do not depend explicitiy on time, the
experiment should yield the same outeome regardless of when the experiment
is performed. I f the growth conditions for the population do change over time,
we would not be able to use an autonomous differential equation to describe
the growth of the population; we would need to explicitiy include the time
:
dependence in the dynamics. '
Formally, we can solve (8.11) by separa tion of variables; namely, dividing
•both sides of (8.11) by 'g(y) and multiplying both sides of (8.11) by dx, we
obtain
JÜ-=d*
g(y) '
478 Chapter 8 B Differential Equations
Shifted
(0, 20) (10, 20)

2
NO) = SO,? '
N(t) = 20e " " >
2 10
10 12 r
A. Figure 8.4
2
T h e graph of the solution N(t) = 20e ' is shifted to the n e w starting point (10, 20)
Integrating both sides gives
dy
- dx
s(y)
W e will discuss two cases: g(y) = k(y - a ) , and g(y) = k(y - a)(y - b). T h e
growth m o d e l (8.1) is an example of the first case; the logistic equation, wñich
w e saw in Section 3.3 for the first time, is an example of the second case.
Case 1 : g(y) = k(y — a) W e w i s h to solve
£ = kíy - a) (8.14)
where ik and a are constants. W e assume k y¿ 0. Separating variables yields

dy
= k dx (8.15)
y - a
where w e need to assume y ^ a in order to be able to divide by y - a. It
is somewhat arbitrary whether we leave k on the right-hand side or m o v e it
to the left-hand side. L e a v i n g it on the right-hand side is more convenient.
Integrating both sides of (8.15) results i n
dy
= kdx
y - a
or
ln|> V a | kx + C)
E x p o n e n t i a t i n g both sides yields
' | y - a | ES e
or
gCigkx
|y-fl|
R e m o v i n g the absolute v a l u é signs, we find
y -a c
-.e ê kx
C l
R e n a m i n g the constant by setting C = ± e , we can w r i t e the solution as
kx
y = Ce +a (S.!c)
( R e n a m i n g the constant only serves the purpose to get the equation in a more-
readable form.) A s in the case of a p u r é time differential equation, integration.
introduces a constant. I f we k n o w a point (XQ, yo) of the solution, then C can
be determined. We will refer to such a point as an initial condition.
T o obtain (8.15), we divided by y — a. W e are only allowed to do this as
long as y a. I f y = a, then dy/dx = 0 and the constant function y = a is a
solution of (8.14). We lost this solution when we divided (8.14) by y - a. Note
C l c
that the constant C in (8.16) is different from 0 since C = ± e and ¿? ' -?£ 0
for any real number C\. B u t we can combine the constant solution y = a and
the solution in (8.16) by allowing C to be equal to 0 in (8.16).
Before we turn to biological applications, we give an example where we
see how to solve such a differential equation and how to determine the v a l u é
oiC.
Example 2 Solve
dy
— = 2— 3y where VQ = 1 when XQ = 1
dx _ _ z/.'\ .j'r.f''9 tt»'
M '
Solution
Instead of trying to identify the constants C,k, and a in equation (8.16),
it is easier to solve the equation directly. W e s e p á r a t e variables and then
i n t é g r a t e , which results in
dy
= / dx
2-3y
Since an antiderivative of j r j ^ is - 5 ln |2 - 3yj, we find
— | ln 12 - 3y| = x + C x
Solving for y yields

ln|2-3y| = -3x-3Ci
-3x-3Ci
\2-3y\ = e
3c 3Á
2 - 3y = ±e' 'e'
3Cí
Setting C - ±e' , we find
2x
2-3y = Ce~ ''f.. .* ¿:£
T o determine C , we use the initial condition yo = 1 w h e n XQ = 1. T h a t is,

3 3
2 - 3 = Ce' or C = -e
Henee.
3 3x
2-3y = -e e~
or
2 1
3-3^
' 3 3
(see Figure 8.5). ^
We will now turn to two biological applications that are covered by
Case 1.
( E x p o n e n ü a l Population G r o w t h ) T h i s model is given by (8.1) and was
introduced in Section 4.6. W e denote by-A/(f) the population size at time t > 0
r
A Figure 8.5
The solution to Example 2
and assume N(0) = NQ > 0. T h e change in population size is desoribed by the

initial v a l u é problem
dN
[¿jo J r á t
Io 1
— = rN w h e r e N(0) - NQ (8.17)
dt
T h e parameter r is called the intrinsic rate of growth and is the per capita rate
of growth since
11
1 dN
r
~ Ñ~d7 ~
W h e n r > 0, this represents a growing population. W h e n r < 0, the size of the
/-^ \. i i, population decreases. (Note that the per capita growth rate r is independen!
of the population size.)
?
— W e can either solve (8.17) directly as in E x a m p l e 2, or use (8.14). L e t ' s
use (8.14). I f w e compare (8.17) with (8.14), we find k = r and a — 0. Henee,
using (8.16), we find the solution
; n
A (f) = Ce
Since N(0) — NQ = C , we write the solution as

r
. N(t) = N e ' Q '•• • (8.18)
'Y E q u a t i o n (8.18) shows that the population grows exponentially when

r > 0. W h e n r < 0, the population size decreases exponentially fast, W h e n
r = 0, the population size stays constant.
rl
W e show solution curves of N(t) = Noe for r > 0, r = 0, and r < 0 in
F i g u r e 8.6. E x p o n e n t i a l growth (or decay) is one of the most important grewth
equations in biOlogy. Y o u should therefore memorize both the differential
equation (8.17) and its solution (8.18), together with the graphs i n F i g u r e 8.6,
and k n o w that (8.17)' describes a situation where the per capita growth rate
(or intrinsic rate of growth) is a constant.
W h e n r > 0, the population size grows without bound (li
N{t) = oo). T h i s can be found when i n d i v i d u á i s are not limited by food
availability or by competition. I f we start a bacterial colony on a nutrient-rich
substrate by inoculating the-substrate w i t h a few bacteria, then the bacteria,
initially, can grow and divide unrestricted. Subsequently, when the substrate
becomes more crowded ahd the food source depleted, the growth will be
restricted, and a different'differential equation w i l l be required to describe
8.1 • Solving Differential Equations 481
N(t) r> 0
r=0
r< 0
Á. Figure 8.6
Solution curves for dN ¡dt = rN
this situation. ( W e w i l l discuss this in Case 2.) E x p o n e n t i a l decay in a popu-

lation can be seen w h e n the death rate exceeds the birth rate (for instance, in
cases of starvatiqn). Wy-*- -i
T h e type of growth in E x a m p l e 3 is referred to as Malthusian growth,
n a m e d a f t e r T h o m a s M a l t h u s ( 1 7 6 6 - 1 8 3 4 ) , a Britishclergyman and economist.
Malthus wrote about the consequences of unrestricted growth on the welfare
of humans. H e observed that exponential growth of the h u m a n population
would lead to starvation, because food supply would remain constant.
R e c a l l from Section 4.6 that when r < 0, (8.17) has the same form as
the differential equation that describes radioactive decay. N(t) would then
denote the amount of radioactive material left at time t. "We will revisit this
application in Problem 20.
> I -"""""SM ( R e s t r i c t e d Growth: v o n B e r t u l a n f T y E q u a t i o n ) Thi.s example describes the

simplest form of restricted growth, and can be used to describe the growth of
fish. W e denote by L(t) the length of the fish at age t and assume L ( 0 ) = L Q .
Then
^=k{A-L) (8.19)
/dt
where A is a positive constant that we w i l l interpret in the following. W e
assume LQ < A and explain this restriction subsequently. T h e constant k is
also positive; the equation says that the growth rate dL/dt is proportional to
, i t • A — L , so k should be interpreted as a constant of proportionality. W e see
u e a , j V
• " ;r "" <>' that the growth rate dL/dt is positive and decreases linearly with length as
long as L < A, and that the growth stops (that is, dL/dt = 0) when L = A.
T o solve (8.19), we s e p á r a t e variables and i n t é g r a t e . T h i s yieids
- fl-
dL r
= kdt
i —L J
Henee,
-In|A-¿| ^Jk/fCf "
After multiplying this equation by - 1 and exponentiating,

c kl
|A -L\ =e- >e-
or
:!
A - L = Ce-'
x
/ i
} * .'
A. Figure 8.7
The graph of the von Bertalanffy equation
_ C l
with C = ± e . Since L ( 0 ) = L o , we find
' A - LQ = C
T h e solution is then given by

k
A - L ( r ) = (A - L )e- '
Q
or
Lo
L(t) = A 1 - (8.20)
A
(see F i g u r e 8.7).
T h i s is the von Bertalanffy equation that we encountered previously.
Since» ,
l i m LÜ) = A
/-i-OO
we see that the parameter A denotes the asymptotic length of the fish.
Mathematically, there are no r e s t r i c í i o n s on L o ; biologically, however, we
require that 0 < Lo < A . Otherwise, the growth rate w o u l d be negative,
meaning that the fish shrinks in size. N o t e that A is an asymptotic length that
is never reached, since there is no finite age T w i t h L(T) — A ii L(0) < A .
R e t u r n i n g to (8.19), we can now interpret this differential equation:
T h e growth rate is proportional to the difference in the current length and
the asymptotic length, with k representing the constant of proportionality.
Since the difference between the current length and the asymptotic length
is decreasing over time, this alsq shows that the growth rate decreases over
time, implying that juveniles grow at a faster rate than adults. M o r e o v e r , the
growth rate is always positive. T h i s means that fish grow throughout_their
lives, which is indeed the case. 4
Case 2: g(y) = k{y — a)(y — b) W e now turn to differential equations of

the form
d
-T =Hy-a)(y-b) (8.21)
dx
where k, a, and b are constant. W e assume k =¿ 0. Separating variables and
integrating both sides yields
dy
= kdx (8.22)
(y-a)(y-b)
8.1 H Solving Differential Equations 483
provided y a and y sí b. W h e n a = b, w e must ñ n d an antiderivative of

1
. , which is — — . I n this case, we ñ n d
K l
• • 1
= + c
y - a
or
y = a -
kx + C
T h e constant C can then be determined from the initial condition. W h e n

y = a, then dy/dx = 0 and consequently y = constant (namely, y = a).
1
T o find the solution when a b, w e must use the partial fraction method lÜ
that we introduced in Section 7.3. L e t ' s first do an example.
[ • E B a s olve
dy
— = 2(y - l ) ( y + 2) with yo = 2 when ¿o = 0
dx
Solution
Separation of variables yields
dy
= / 2dx
(y-l)(y+2)
W e use the partial fraction method to i n t é g r a t e the left-hand side.
1 _ A B
+
( y - l ) ( y + 2) " y - 1 J+2
A(y+2) + fl(y-l)
( y - l ) ( y + 2)
(A + fl)y+2A - 5
=
(y-l)(y+2)
Comparing the last term to the i ñ t e g r a n d , we find
A + S = 0 and 2A - B = 1
which yields
A = —B and 2A - B = 3A = 1
and thus
A = - and fl =
3 3
U s i n g the partial fraction decomposition, we must i n t é g r a t e
,
dy= i 2dx
3 J \y - 1 y +2
w h i c h yields
- [ l n |y - 1 | - In |y~H 2|] = 2JC + Ci

3
484 Chapter 8 H Differential Equations
Simplifying this results in
ln = 6x + 3 C i
y + 2
y - l 3c 6x
e ê
y+ 2
y - l
= ±e"3Ci „6x
y+ 2
i
y - l
= Cebx
• y + 2
U s i ñ g the initial condition yo = 2 w h e n XQ = 0. we find
1
= C
T h e solution is therefore
y - 1
y+ 2
I f we want the solution in the form y = - f(x), we must solve for y. W e find
y - l •
6
ê * + 1
y
.i 1 - \e<"
6x
2e + 4
y 6 1
4 - e-'
(see F i g u r e 8.8).
W e r e t u r n now to (8.22) when a =¡¿ b and y ^ c? or b,W e use the

partial fraction method to simplify the integral on the left-hand side of (8.22).
W e decompose the integrand on the left-hand side of (8.22) into a sum of
100
.v
80
60
40
20
0
:
A Figure--8.8 •
T h e solution for E x a m p l e 5
simpler rational funcfions that we k n o w how to i n t é g r a t e ; namely, we write

the integrand in the f o r m
1 A B
( 8 2 3 )
(y-a)(y-b) y - a ' y - b '
where A and B are constants that we must find. W e do the following algebraic
manipulation on the right-hand side of (8.23):
A B A(y-b) + B(y-a)
+
y - a y - b (y - a)(y - b)
y(A + B) - (Ab + Ba)
(8.24)
íy-a)(y-b)
Comparing the last expression in (8.24) to the left-hand side of (8.23), we
conclude that the constants A and B must satisfy
A + = 0 and Ab + Ba - -1
Substituting B = - A from the first equation into the second, we find

1
Ab - Aa =-1 or A(b — a) =—1 or A =
a-b
and, t h e r e f o r e ¡
B = - -
a-b
T h a t is,
1 1 1
(8.25)
(y — a) (y — b) a — b |_y — a y — b
E q u a t i o n (8.25) allows us to e v a l ú a t e
dy
[y-a){y-b)
which is the left-hand side of (8.22). Namely,
dy 1 1
dy
{y - a){y — b) 'a — b J \y — a y — bj
1
[ l n | y - a | - l n | y - o | ] + Ci
a-b
Integrating the.right-hand side of (8.22) as well and combining the constants
of integration into a new constant, C 2 , we find
-[\n\y - a\-\n\y-b\] = kx + C 2
a-b'
or
y - a
ln = k{a-b)x + C {a - b)
2
y - b
Exponentiating this yields
y - a eC (a-b)
2 gk{a-b)x
y - b
or
y - g C (a-b) k(a-b)x
±e 2 e
y - b
L

the integrand in the. f o r m
1 A B
( 8 2 3 )
(y-a)(y-b) y - a ' y - b '
A B A(y-b) + B(y-a)
+
y - a y - b {y - a)(y - b)
y(A+ B) - {Ab + Ba)
(8.24)
íy-a)(y-b)
A + B = 0 and Ab + Ba = - 1
Substituting B = - A from the first equation into the second, we find
1
Ab — Aa = — 1 or A(b — a) =—1 or A =
a-b
and, therefore*
B - - -
a-b
T h a t is,
1 1 1
(8.25)
[y — a)(y — b) a — b \_y — a y — b_
dy
{y-a) (y - b)
dy 1 1
dy
(y — a)(y—b) a —b J \y — a y — bj
1
[ I n | y - a | - l n | y - o | ] + C]
a-b
of integration into a new constant, C 2 , we find
-[ln\y - a\ - l n \y - b\] = kx + C 2
a-b'
or
y - a
ln = k{a-b)x + C {a - b)
2
y - b
y - a eC {a-b)
2 £k{a-b)x
y - b
or
±e 2 e
y - b
i

the integrand in the. f o r m
1 B
( 8 2 3 )
{y-a){y-b) y - a ' y - b '
A B A(y-b) + B{y-a)
y - a +y - b {y - a)(y - b)
y{A + B) - {Ab + Ba)
(8.24)
íy-a){y-b)
A + B = 0 and Ab + Ba - -1
Substituting 5 = - A from the first equation into the second, we find

1
Ab - Aa -1 or A(b — a) = — 1 or A =
a-b
and, thereforej •
B = - -
a-b
T h a t is,
1 1 1
(8.25)
{y — a){y — b) a — b \_y — a y —b
dy
{y-a) {y - b)
dy 1 1
dy
{y-a){y-b) a-b y— a y —b
1
[ l n | y - a | - l n | y - o | ] + Ci
a-b
of integration into a new constant, C , we find 2
-[ln\y - a\ - l n \y - b\] = kx + C 2
a-b'
or
y - a
ln = k{a-b)x + C {a - b) 2
y - b
y - a eC {a-b)2 £k{a-b)x
y - b
or
±e 2 e
y - b
C2Ía
Defining C = ±e "*\ we find as the solution of (8.21)
k(a b)x
i r l = Ce -
y - b
W e can solve this for y, namely,
kLa b)x
y = a ^ (y -b)Ce -
k{a b]x k{ a b)x
y{l-Ce - )=a-bCe - -
a _ b C e H o - b ) x ;
y = ( 8 2 6 )
i - c e ^ - ^ ;
T h e constant C can then be determined from the initial condition. When
y = a or b, then dy/dx = 0 and, consequently, y = constant ( n a m e l y , y = a
or b).
T h e following example of density-dependent growth is one of the most
important applications of this case. I n E x a m p l e 3, we considered unrestricted
(or density-independent) growth. I t has the unrealistic feature. that if the
intrinsic rate of growth is positive, the size of a population grows without
bound. F r e q u e n t l y , the per capita growth rate decreases as the population
size ulereases. T h e growth rate thus depends on the population density. T h e
simplest such model for which the growth rate is density dependent is the
^ i logistic equation. I n the logistic e q u a t i o n ^ J h j y ^ e j í f ^ j ^ ^
^ " 1 J ) grow_without bound.
| T h e logistic equation was originally developed around 1835 by Pierre-
Francois V e r h u l s t , who used the term logistic for the equation. H i s w o r k was
completely forgotten until 1920, w h e n R a y m o n d P e a r l and L o w e l l J . Reed
published a series of papers on population growth. T h e y used the same
equation as V e r h u l s t . After discovering V e r h u l s t ' s work, P e a r l and R e e d
adopted the ñ a m e logistic for the equation. P e a r l and R e e d (1920) used the
logistic equation to predict fu ture growth of the U . S . population based on
census data from 1790 to 1920. T h e i r equation would have predicted about
185 million people in the U n i t e d State for the year 2000, w h i c h is a gross
•' '• underestimate of the actual population size (over 260 million people). Though
the equation does not seem to fit actual populations very well, it is a useful
model for analyzing growth under limiting resources.
j \ ( T h e Logistic E q u a t i o n ) T h e logistic equation describes the change in size of

• •- •— a population for which per capita growth is density dependent. I f w e denote
7
by A (O the population size at time t, then the change in growth is given by
the initial v a l u é problem:
L a ^ ^ v Aí-s*c^ if--''^
l — = r ^ ( i - — ) w i t h A ' ( 0 ) = Ao (8.27)
(w ~ » ^ A dt \ K)
v
i? w e,o'- s¡ i<?v>v* v- - where r and K are positive constants. T h i s is the simplest w a y of incorporating
i,-,.,, ¡i-, density dependence in the per capita growth rate, namely, it decreases linearly
w t n
~¡ -.. j . - » f OOÍ -e. i ^ * ' population size (see F i g u r e 8.9):
U\^¿C ****~*-*» 1 dN / N
A« S /=..|>^
0
¿« 4 v » - f > ' N dt \ K
W e can interpret the parameter r as follows: T h e per capita growth rate

is equal to r when A' = 0. T h e r e f o r e , one w a y to measure r is to grow the
organism at a very low density, that is, w h e n N is much smaller than K, so that
8.1 • Solving Differential Equations 487.
the per capita growth rate is c i ó s e to r . ( I n Problem 42, we discuss a different

way to find r.)
T h e quantity K is called the carrying capacity. L o o k i n g at Figure 8.9,-
we see that the per capita growth rate is 0 when the population size is at
the carrying capacity. Since the per capita growth rate is positive below K
and negative above K, the size of the population will increase below K and
decrease above K. T h e number K thus determines the population size that
can be supported by the environment. T o show this mathematically, we need
to solve (8.27). W e will then see from the solution of (8.27) that starting from
any positive initial population size the size of the population will eventually
A Figure 8.9 reach K [that is, l i m _ A ' ( f ) = KM /V(0) > 0].
í 0 0
The per capita growth rate in L e t ' s solve (8.27). W e write the right-hand side of (8.27) as
(he logistic equation is a
linearly decreasing function of - K) (8.28)
population size K
Comparing (8.28) with the right-hand side of i 21), we find
ks*-— a = 0 b :K i (8.29)
K
T h e solution of (8.27) is given i n (8.26). U s i n g (8.29), the solution of (8.27) is
therefore
0 - KCe
N(t) =
1 _ Ce-O/KXO-*)'
or
rr
-CKe CK
r
- Ce < C-e-
K
r
(8.30)
- '/C
Since /V(0) = Aô, we find
CK
A =
0
C - 1
Solving this for C ,
A ( C - 1) = CK
0 or C ( / V - K) =
0 NQ
thus
NQ
C = (8.31)
NQ K
Substituting (8.31) into .30), we find
K
N(t) = No-K
1
No '
or
K
N(t) (8.32)
K_
1 + No
It follows from (8.32) that

lim N{t) = K
We see from Figure 8.10, where we plot (8.32) for NQ > K and An < K,
that the solution N{t) approaches K, the carrying capacity, as t -* oo. When
NQ > K, then A ( í ) approaches K from above; when 0 < A ' < K, then N(t) 0
r 488 Chapter 8 • Differential Equations
NO)
\N D >K
A Figure 8.10
Solution curves for different initial valúes Ao
approaches K from below. T o use separation of variables to solve (8.27), we

need to exelude/ N = 0 and N = K. W h e n A ( 0 ) = 0, then dN/dt = 0 and,
7
consequently, N(t) = constant = N(0) = 0 for a l l t > 0. W h e n - A ( 0 ) = K ,
7
then dN/dt — 0, and, consequently, A (r) = con-stant = for all r > 0. [The
constant solution N(t) = K is contained in (8.32) i f w e choose NQ = K . ]
T J j e ^ p j ^ a n t ^ s ^ u d o n s „/C.. and 0 are. c a l l e r J _ e q u i l i b r i a _ T h e constant
solution N (i) = 0 is not a very interesting one. W e calLit t h e t r i v i a l equilibrium.
I f N(0) = 0, then nothing happens, that is, N(t) sfays equal to 0 for all later
times. T h i s m a k e s sense; if there aren't any i n d i v i d u á i s to begin with, then
there won't be any later on.
W e can show that i f 0 < NQ < j , then the solution curve is S-shaped,
as seen in F i g u r e 8.10. T h i s is characteristic of populations that show this
type of density-dependent growth. A t low densities, the growth is almost
like u n r e s t r i c t é d growth. A t higher densities, the growth is restricted and the
curve bends around and eventual!)' levéis off at the carrying capacity. I f thfi
population size is initially greater than the carrying capacity, the population
size decreases and becomes asymptotically (that is, when t oo) equal lo
the carrying capacity K. 4
8.1.3 Allometric Growth f U s ? c v

° p ^( í
Solving differential equations of the form (8.2) gets c ó m p l i c a t e d quickly.

B e f o r e we show the role such differential equations play in biology, we will
give an example in which we solve a differential equation of this type.
S o l v e
BEHulSBBI
dy
with yo = 0 w h e n XQ = 1
dx
Solution
T o solve this differential equation, we s e p á r a t e variables first and then

i n t é g r a t e . W e find
f dy r dx
J y+l 1 x
C a r r y i n g out the integration on both sides, we obtain
ln|y + l | = l n | x | + Ci (8.33)
Solving for y, we find

c
.|y + l | = e ' | x |
c
y+ 1= ±e 'x
Ct
Setting C = ±e . we obtain
y = C.x - 1
Using the initial condition yo — 0 when .vn = 1 allows us to determine C :
0 = C - 1 or C = 1
T h e solution is therefore
> =» x -1 ^
W e will now turn to a biological application, namely allometric growth.

W e have seen a number of allometric relationships throughout the earlier
chapters of this book. Tii£S£-are-typically.relationships between sizesjiLp.arts
of an organism (for instance, sJgjlJJength_and_T3^yêngth, or leaf á r e a and
s t e m ^ T á m e T e r ) . W e denote by Ly{t) and ¿ 2 ( 0 the sizes of two organs of an
individual of age t. W e say that L ¡ and L are related through an allometric
2
.;,., law if their specific growth rates are proportional, that is, if
1 dL\ 1 dLi
!
/^t — — - = k - - i (8.34)
1
. I 14.
Éí =. V_ U l f the constant k in (8.34) is equal to 1, then the growth is called isometric;
otherwise it is.called allometric. Cancelling dt on both sides of (8.34) and
integrating, we find
J L \ J L 2
or
l n | L i | = * l n | L | + Ci 2 (8-35)
Solving for L \ , we obtain

L i = CL\ (8.36)
C l
where C = ± e . (Since L i and L 2 are typically positive, the constant C will
typically be positive.)
[•PjfSTnfflriil I n a study of 45 species of unicellular algae, the relationship between cell

volume ( V ) and cell biomass ( 5 ) was found to be
a 7 9 4
' B a V
F i n d a differential equation that relates the relative growth rates of cell

biomass and volume.
Solution
T h e relationship between cell biomass and volume can be expressed as
0J9A
,- B(V) = CV (8.37)
where C is the constant of proportionality. Comparing (8.37) with (8.36), we

;
T F.,'¡Óc... -. o see that k = 0.794. I t therefore follows from (8.34) that / t \ "< • - • "'•' •
1 dB 1. 1 dV
W e can also get (8.38) by differentiating (8.37) with respect to V , that is,
0 794 1
— = C(0.794)V - -
dV
E q u a t i o n (8.37) allows us to e l i m í n a t e C. Solving (8.37) for C , we find
JH
- C = BV-° and, therefore,
^ = B V -0,794 ( f j 7 9 4 ) í / 0.794-l
dV
1
= (0.794)5 V "
R e a r r a n g i n g terms yields
dB — " dV
— = (0.794) —
B V
D i v i d i n g both sides by dt, we get
1 dB „ _ 1 dV
= (0.794)
B dt V dt
which is the same as (8.38).
[ • ) £ETnTnBrH ( H o m e o s t a s i s ) T h e nutrient c o n t e n í of a consumer (for instance, the percent

nitrogen of the consumer's biomass) can range from refiecting the nutrient
c o n t e n í of its food to being constant. T h e former is referred to as absence of
homeostasis, the l a t í e r as strict homeostasis. A model for h o m e o s í a l i c regu-
l a í i o n is p r o v i d e d in Sterner and E l s e r (2002). T h e model relates consumer's
nutrient c o n t e n í (denoled by y) to its food's nutrient c o n t e n í (denoted by A )
as follows:
d y
-l = l- • (8.39)
T. íi ; F
l
8X .
where 8 > 1 is a constant. Solve the differential equation and relate 6 lo
absence of homeostasis and stricí homeostasis.
Solution
W e can solve (8.39) using separation of variables
t dy _ 1 f dx_
1 7 ~ BJ x
Integrating and simplifying yields
\n\y\ = -ln \x\ + Ci
e (l/ff)ln|x|+Ci
]/e C
\y\ = \x\ e i
c l e
y = ±e ^x l
Since x and y are positive (they denote nutrient contents), we find
9
y = Cx'l
where C is a positive constant.
Absence of homeostasis refers to when the consumer reflects the food's
nutrient c o n t e n í . T h i s occurs when y — Cx and íhus when 8 = 1. Siria
homeostasis means í h a l í h e nutrient c o n t e n í of í h e consumer is independen!
of í h e n u í r i e n í c o n í e n í of the food, thal is, y = C; this occurs in the limit
4
8 ->• oo.
8.1 • Solving Differential Equations 491
,8.1*4 Problems y» . "

(8.1.1) is on a linear scale; the vertical axis, representing the size of
¡n Problems 1-8, solve each puré lime differential equation. the population, is on a logarithmic scale), you find that your
dv data fit a straight line that intercepts the vertical axis at 1 (log
1, — = x — sin x, where VQ = 0 for .tn = 0 scale) and has slope - 0 . 4 3 . Find a differential equation that
dx relates the growthrate of the population at time t to the size
dv of the population at-time f.
dx = e~ , where yr¡ = 10 for JCQ = 0
x
2. —
dy 19. Suppose that a population, whose size at time t is denoted
—- = - , where yrj = 0 when XQ = 1 by N(t), grows according to
dx
1 dN
1 r 8 4
ctx 1 +
, where yo = 1 when XQ — 0 Ñ17= < - °)
where r is a constant.
1
, where x{0) = 2 (a) Solve (S.40).
1 -
(b) Transform your solution in (a) appropriately so that the
ce
: cos(/ — 3), where .r(3) = 1 resulting graph is a straight line. H o w can you determine the
constant r from your graph?
ds 7
= v 3/ + 1, where s{0) = 1 (c) Suppose now that you followed á population over time
7
- 77 that evolved according to (8.40). Describe how you would
í 8. = 5 - 1 6 r , where A (3) = - 1 1 determine r from your data.
77
9. Suppose that the volume V(t) of a ceil at time t changes 20. Assume that W{t) denotes the amount of radioactive
according to material at time t. Radioactive decay is then described by the
dV differential equation
: c o s í with V(0) = 5 dW
dt
Find V(t). • -—=-xw(t) withW(0) = W 0 (8.41)
10. Suppose that the amount of phosphorus in a lake at time where X is a positive constant, called the decay constant.
f, denoted by P(t), follows the equation (a) Solve (8.41)."
dP (b) Assume that W(0) = 123 gr and VV(5) = 20 gr and that
—-=3t + l withP(0) = 0
dt time is measured in minutes. Find the decay constant k and
Find the amount of phosphorus at time t = 10. determine the half-life of the radioactive substance.
(0.1.2) 2 1 . Suppose that a population, whose size at time t is N(t),
grows according to
ln Problems 11-16, solve the given autonomous differentiaf
dN 1 9
equations. with/V(0) = 10 (8.42)
dy dt
: 3)', where yQ = 2 for XQ = 0 (a) Solve (8.42).
dx
(b) G r á p h N(t) as a function of r for 0 < t < 10. What
dj_ 12. 2(1 - y ) , where VQ = 2 for XQ = 0 .^-happens as r —>• 10? Explain in words what this mearis.
dx '
^ 2 2 . ) Denote by L{t) the length of a.fish at time r and assume
13. dx ; - I r , where JC(1) = 5 "HTíat the fish grows according to the von Bertalanffy equation
It '
dL
",14J : : 1 — 3JC, where l) = -2 = k{24-L{t)) withL(0)=2
S¿? dt ~dl
15. ± : : 2/i + l , where h(0) = 4

d (a) Solve (8.43).
ds (b) Use your solution in (a) to determine k under the
16. ÜL :5--/V,wbere7/(2) = 3 assumption that L ( 4 ) — 10. Sketch the graph of L ( / ) for
this valué of k.
/ Suppose that a population, whose size at time t is denoted
(c) F i n d the length of the fish when / = 10.
&y N (r), grows according to
(d) F i n d the asymptotic length of the fish, that is, find
— =0.3/V(7) with/V(0) = 20 linv-*oo ¿(O- - ;
dt
Solve this differential equation, and find the size of the 23. Denote by L(t) the length of a certain fish at time í and
population at time t = 5. assume that this fish grows according to the von Bertalanffy
/ 18. 'Suppose that you follow the size of a population over equation
lime. When you plot the size of the population versus time on dL
— = ¿(Loo - £•(')) with 1.(0) = 1 (8.44)
a semilog plot (that is, the horizontal axis, representing time, dt ,
492 Chapter 8 • Differential E q u a t i o n s
where k and Loo are positive constants. A study showed that 36. Find a solution of
the asymptotic length is equal to 123 in. and that it takes this dy 2
y +4
fish 27 months to reach halí its asymptotic length. dx
that passes through (0, 2).
(a) U s e this information to determine the constants k and
• 37.)Suppose that the size of a population at time r is denoted
í.oo¡ii(8.44). ';• ; 1
"By'A'(f) and that A^í) satisfies the differential equation
[Hint: Solve (8.44).]
(b) Determine the length of the fish after 10 months. — =0.34/VÍ 1 - — J with N(0) = 50
(c) H o w long will it take until the fish reaches 9 0 % of its Solve this differential equation, and determine the size of the'
asymptotic length? pojíulation in the long run; that is, find l i m _ K N(t). / X )
^ ' 2 4 ? ) L £ t A ' ( ; ) d e n o t e t h e s i z e o f a p o p u l a t i o n a t t i m e í . Assume/538^ Assume that the size of a population, denoted by NU)
iSát the population exhibits exponential growth. evolves according to the logistic equation. Find the intrinsic
(a) I f you plot lo'g/v(r) versus r, what do you get? rate of growth if the carrying capacity is 100, N(0) = 10, and
N(l) = 20.
(b) F i n d a differential equation that describes the growth of
this population and sketch possible solution curves. 39. Suppose that NU) denotes the size of a population at
timé t and rhat
25. U s e the partial fraction method to solve
dN
dy
~d~t~
dx
(a) Solve this differential equation when JY(0) = 10.
where yo = 2 for XQ = 0. ,
(b) Solve this differential equation when N(0) = 90.
26. U s e the partial fraction method to solve
dy (c) Graph your solutions in (a) and (b) in the same c o o r d í n a t e
= yíl-v) system.
dx
where yo = 2 for XQ - 0. (d) Find l i m / ^ o c N U ) for your solutions in (a) and ( b ) .
27. Use the partial fraction method to solve 40. Suppose that the size of a population, denoted by NU),
dy satisfies
= y(y - 5) d_N_ N'
dx
dt - 0.7*11-g (8.4:
where yo = 1 for XQ = 0.
(a) Determine all equilibria by solving dN/dt = 0.
."¡'28. U s e the partial fraction method to solve
(b) Solve (8.45) for ( i ) N(0) = 10, (ii) N(0) = 35, (iii) N(Q) =
dy
= (y - D O - 2) 50, and ( i v ) N(0) = 0. F i n d lim,_>oo N U ) for each of the four
dx
where yo = 0 for XQ - 0. initial conditions.
29. U s e the partial fraction method to solve (c) Compare your answer in (a) to the limiting valúes you
dy ¿ o u n d in (b).
-f-=2y(3 y) -4*41. I L e t N U ) denote the size of a population at time /. Assume
dx
where yo = 5 for XQ = 1. lÜSt the population evolves according to the logistic equation.
30. Use the partial fraction method to solve Furthermore, assume that the intrinsic growth rate is 5 and
that the carrying capacity is 30.
Ú- dy 1 - 2
— — - y — 2v (a) Find a differential equation that describes the growlh of

dt 2*
where yp = —3 for IQ = 0. this population.
In Problems 31-34, solve the given differential equations. (b) Without solving the differential equation in ( a ) , sketch
solution curves of N U ) as a function o f ; when ( i ) A'(O) = 10,
dy
31. = v(l+y) m N(Q) = 20, and (iii) N(0) = 40.
dx
¿'{42.} Logistic growth is described by the differential equation
dy
32. = (1 + y ) 2
dN
dx rN 1
It
dy
33. = (1 + ) ' ) 3
The solution of this differential equation with initial condition
dx
A>(0) = NQ is given by
dy K
34. = ( 3 - y ) ( 2 + y) (8,46)
dx NU) =
1 + ( 4 " 1 U-
35, («) Use the partial f
f du 1 (a) Show that
+ C K - NQ\ , 1 NU)
2^ r = iln + -ln (8.47)
-
(b) Use.your result in (a) to find a solution of • NQ I
dy by solving (8.46) for r. K NU)
y (b) The expression for r in (8.47) can be used to e s t í m a t e
dx
that passes through ( i ) (0, 0), ( i i ) (0, 2 ) , and (iii) (0, 4). r. Suppose we follow a population that grows according to
1
.A' i 8.2 H Equilibria and Their Stability 493
the logistic equation, and find that N(0) = 10, N(5) 22, dy
49 1+1 with y = 5 if ¿o = 2
WfiüO) 53 30, and #(200) = 30. Estimate r. , . t - ¿ x - 1
0
t £ y (Seleaion at a single locus) W e consider" one locus

sin t
with two alíeles, A i and A-¡, in a randomlv mating diploid ¡Í5CÚ — with un = 3 if IQ — 0
2
' a/ u +1
1
population. T h a t is, each individual in the population is

either of type A\A\, A\Ai, or A2A2. We denote by p{t) = re ' with rn = 1 if fn = 0
the frequency of the A i alíele and by q(t) the frequency of the dt
Ai alíele in the popujation a time t. Note that p(t) + a(t) = ^
We denote the fkness'of the A / A j type by w¡j and assume •' ¿y = - - with.ro = 2 if yo = 3
2 y
that i ^ u = 1, u>i2 = 1 - J / 2 , and 1022 = 1 — *, where j
is a nonnegative constant less than or equal to 1. That is, | 53' ;- (Adapted from Reiss, 1989) In a case study by Taylor et
íhe fitness of the heterozygote AÂi is halfway between the "~áf. (1980), where the maximal rate of oxygen consumption (in
í
niness of the two homozygotes, and the type A\A\ is the mi s~ ) for nine species of wild African mammals was plotted
littest. I f s is small, we can show that approximately agatnst body mass (in kg) on a log-log plot, it was found that
the data points fall on a straight line with slope approximately
dp 1 ,.
s p i l
p) with p(0) = po (8.48) equal to 0.8. Find a differential equation that relates maximal
* ' 2 oxygen consumption and body mass.
(a) Use separalion of variables and the partial fraction
method to find the solution of (8.48). . \ ^ S f í | Consider the following differential equation, which is of
(b) Suppose po se 0.1 and s = 0.01; how long'will take until "Tnirportance in population genetics:
p(0 = 0.5? aíx)g{.x) - ±j- [b(x)g(x)] 53 0
(c) F i n d l i m , . ><xs p{t). Explain in words what this limit means.
where b{x) > 0.
(ü.1.3)
ln Problems 44-52, solve each differential equation with the (a) Define y = b(x)g(x), and show that y satisfies
given initial condition. a(x) 1 rfv
V (8.49)
dy b{x)-• - 2 ^ = °
44. - í - = 2- with vo = 1 if XQ 1 (b) S e p á r a t e variables in (8.49), and show that if y > 0, then
dx x
d] 1 a(x)
with yo 2 if .t 53 0 y — C exp •dx
45. — = 0
b(x)
ax
!55. \ When phosphorus content in Daphnia was plotted against
dy_ with VQ 2ííx =0
46. 0 ptiosphorus conten t of its algal food on a log-log plot, a straight
dx x +1
line with slope 1/7.7 resulted (see Sterner and Elser, 2002;
dy x
data from D e M o t t et al., 1998). F i n d a differential equation
47. = (y + l)e~ with y = 2 if XQ = 0 0
dx that relates phosphorus content of Daphnia to the phosphorus

= x 2y2 with yo = 1 if ¿o = 1
dy_ content of its alsai food.
48.
dx
8.2 EQUILIBRIA A N D THEIR STABILITY ' ' % [ '
I n Subsection.8.1.2, we learned how to solve autonomous differential equations

and graphed t,heir solutions as functions of the independent variable for given
initial conditions. F o r instance, logistic growth
(8.50)
with initial condition A ( 0 ) = NQ has the solution given i n (8.32) and graphed
in F i g u r e 8.10 for different initial v a l ú e s .
"The solution of a differential equation can inform us about the long-term
behavior, as we saw in the case of logistic g r o w t h . N a m e l y , i f Ao > 0, then
/y( )f K , the carrying capacity, as t -* 0 0 , and i f Aô = 0, then N(t) = 0 for
all í > 0. A l s o , i f NQ = K , then N(t) = K for all / > Ü. W h a t is so special
about NQ = K 0 1 NQ = 0? W e see from E q u a t i o n (8.50) that i f N = K or
N = 0, then dN/dt == 0, implying that NQ) is constant.
Constant solutions form a v e r y special class of solutions of autonomous
differential equations. T h e y are called point equilibria or simply equilibria.
T h e constant solutions N = K and N - fl noint - -i-u •
ff ü a r e o m í
equation. ~ P equdjbna o f the íqgfefc.
i n t h i s secíjo/), we vw'JJ consider autonomous ¿JjJif£-j-¿*s?jJ^J
ene r o n n
Z = íW , J8.51)
wfiere we wiíí typicaííy tfiink o f x as time. W e / w í l l l e a r n how to find

point e q u i l i b r i a , and we will discuss what they can tell us about the long-term
behavior of the solution y = y ( . r ) ; that is, the behavior of y t » a s * - > o o . I f w e
can solve (8.51), we can study the solution directly to obtain information about
the long-term behavior. B u t what should we do i f we cannot solve (8.51)?
Candidates for describing the long-term behavior are the constant
solutions or equilibria y = y ( r e a d " y h a t " ) t h a t satisfy g(y) = 0. (Such
solutions, of course. need not exist.) A ? .i.,<
I f y satisfies
g(y) = o
then y is an equilibrium of
dy
••dx
L e t ' s look at equilibria in more detail beforewe discuss specific exampies.

C) T h e basic property of equilibria is that if, initially (say, at x = 0 ) , y(Q) = y
N> \ J and y is an equilibrium, then y{x) = y for all x > 0.
_[ \ A phy§ical analogue of equilibria is provided i n F i g u r e 8.11. O n the left
side, a ball rests on top of a hill; on the right side, a ball rests at the bottom of
A Figure 8.11 ' a valley. I n either case, the ball is in e q u i l i b r i u m because it does not move.
Stability illustrated with a ball O f great interest is the stability of equilibria. W h a t w e mean by this is
on a hill and in a valley best explained using our example of a ball on a hill versus a ball in the v a l l e y .
I f we perturb the ball by a small amount—that is, if we move it out of its
equilibrium slightly—the ball on the left side w i l l roll down the hill and not
return to the top, whereas the ball on the right side will return to the bottom
of the valley. W e cali the situation on the left side unstable and the situation
on the right side stable.
T h e analogue of stability for equilibria of differential equations is as
follows. Suppose that y is an equilibrium of ^ = g(y), that is, g(y) = 0. We
say y is locally stable i f the solution returns to y after a small perturbation;
this means that we look at what happens to the solution when w e start closc
to the equilibrium interpreting this as having moved the solution a w a y from
the e q u i l i b r i u m by a small amount—this is called a small perturbat-isn. K
the solution does not return after a small perturbation, we say y is unstable.
T h e s e concepts will be developed in the following subsection, in which we
will discuss a graphical and an analytical method for analyzing stability oí
equilibria. T h i s is then followed by a number of applications.
8.2.1 A First Look at Stability

G r a p h i c a l A p p r o a c h Suppose that g(y) is of the form given in Figure 8.12.
T o find the equilibria of (8.51), we set g(y) = 0. Graphically, this means that
if we graph g{y) (that is, the derivative oí y w i t h respect to x) as a function of
y, then the equilibria are the points of intersection oíg(y) w i t h the horizontal
T n e constant solutions N = K and N = 0 are point equilibria of the logistic

equation.
I n this section, we will consider autonomous differential equations of
the form
= s i y )
T x " ^
where we w i l l typically think of x as time. W e / w i l l l e a r n how to find
point equilibria, and we will discuss what they can tell us about the long-term
behavior of the solution y = y(x); that is, the behavior of y{x) as x - > co. I f we
can solve (8.51), we can study the solution directly to obtain information about
the long-term behavior. B u t what should we do i f we cannot solve (8.51)?
Candidates for describing the long-term behavior are the constant
solutions or equilibria y = y ( r e a d " y h a t " ) t h a t satisfy g(y) = 0. (Such
solutions, of course, need not exist.) .o? .),,•. ,-. , .,
.'• I f y satisfies' :•' ..' ' V
then y is an equilibrium of
dy •
s ( y )
T x =
L e t ' s look at equilibria in more detai'l before we discuss specific exampies.

(_) T h e basic property of equilibria is that if, initially (say, at x = 0 ) , y ( 0 ) = y
f \. \ J and y is an equilibrium, then y(x) = y for all x > 0.
J_ \ A physjcal analogue of equilibria is provided i n F i g u r e 8.11. O n the left
side, a ball rests on top of a hill; on the right side, a ball rests at the bottom of
a
A Figure 8.11 ' valley. I n either case, the ball is in equilibrium because it does not move.
Stability illustrated with a ball O f great interest is the stability of equilibria. W h a t we mean by this- is
on a hill and in a valley best explained using our example of a ball on a hill versus a ball in the valley.
I f we perturb the ball by a small a m o u n t — t h a t is, i f we move it out of its
equilibrium slightly—the ball on the left side w i l l roll down the hill and not
return to the top, whereas the ball on the right side will return to the bottom
of the valley. W e cali the situation on the left side unstable and the situation
on the right side stable.
T h e analogue of stability for equilibria of differential equations is as
follows. Suppose that y is an equilibrium of | £ = g(y), that is, g(y) = 0. We
say ) ' is Jocally stable i f the solution returns to y after a small perturbation;
this means that we look at what happens to the solution when we start closc
to the equilibrium interpreting this as having m o v e d the solution away from
the equilibrium by a small amount—this is called a small perturbation. If
the solution does not return after a small perturbation, we say y is unstable.
T h e s e concepts will be developed in the following subsection, in which we
will discuss a graphical and an analytical method for analyzing stability of
equilibria. T h i s is then followed by a number of applications.
8.2.1 A First Look at Stability

Graphical Approach Suppose that g(y) is of the form given in Figure 8.12.
T o find the equilibria of (8.51), we set g(y) = 0. G r a p h i c a l l y , this means that
if we graph g{y) (that is, the derivative of y w i t h respect to x) as a function of
y, then the equilibria are the points of intersection of g(y) w i t h the horizontal
A Figure 8.12 .
T h e function g(y). T h e arrows c i ó s e to the equilibria i n d í c a t e the type of stability
axis, which is the y-axis i n this case since y is the independent variable. ( L o o k
at the labels of the axes i n F i g u r e 8.12 to see what is graphed in F i g u r e 8.12.)
We see in Figure 8.12 that for our choice of g(y), the equiiibria are at y = 0,
y i , and y . 2
W h y does this w o r k ? R e m e m b e r , we are discussing autonomous differ-

ential equations. T h i s means that the derivative of y (dy/dx) is a function of
y; it does not depend explicitiy on x. T h i s allows us to graph the derivative
of y as a function of y. Since dy/dx-= g(y), this means graphing g(y) as a
function of y. W e can then use the graph of g{y) to say something about the
f á t é b f a solution based on its current valué. Namely, if the current v a l u é y
is such that g(y) > 0 (that is, dy/dx > 0 ) , then y w i l l increase as a function
o í * . I f y is such that g(y) < 0 (that is, dy/dx < 0), then y w i l l decrease as a
function of x. T h e points y where g{y) = 0 are the points where y w i l l not
change as a function of x [since g(y) = dy/dx = 0]. These are the (point)
equilibria. • . f
E q u i l i b r i a are characterized by the property that a system'in an equilib-

rium state stays there for all later times (unless some external f o r c é disturbs
the system). T k i s does not imply, that the system w i l l necessarily reach a
particular equilibrium w h e n starting from some initial v a l u é that is different
from the equilibrium, ñ o r does. it imply that it will return to the equilibrium
after a small perturbation.
Whether or not the system w i l l return to an equilibrium after a small
perturbation depends on the local stability of the equilibrium. B y this we
mean the fdllowing. Suppose that the system is i n equilibrium, which we
denote by y. W e apply a small perturbation to the system, so that after the
perturbation the new state of the system is
y = y +z •• (8.52)
where z is small and m a y be positive or negative. W e explain what can happen

using the function g(y) in Figure 8.12.
Suppose that y = 0 and we subject this equilibrium to a small perturba-
tion. I f the new v a l u é y = y + z = z'> 0 for z small, then dy/dx > 0, that is, y
will increase (see a r r o w ) . I f the new v a l u é y;= y + z = z < 0 íor z small, then
dy/dx < 0 and y w i l l decrease (see arrow). I n either case, the system will not
return to 0. W e say that y = 0 is unstable. D r a w i n g arrows on the horizontal
axis helps determine stability.
W e now turn to the equihbrium y = y\. W e perturb the equilibrium
to y = y + z for z small. L o o k i n g at Figure 8.12, i f z > 0, then dy/dx < 0
r
dN
di
60
40
. \ - -11-
20 - /
0
x¿ 100\ W
-20
-40 -
A F i g u r e 8.13
The graph oídN/di versus N in Example 1
and henee y decreases. I f z < 0, then dy/dx > 0 and y increases. T h e system
will therefore r e t u r n to the e q u i l i b r i u m yi after a small perturbation. W e
say that yi is lócally stable. T h e attribute locallv refers to the fact that the
system w i l l r e t u r n to yj if the perturbation is sufficiently¿malí, I t does not say
anything about w h a t happens w h e n the perturbation is large. F o r instance, if
the perturbation z is large and the n e w v a l u é is less than 0, then dy/dx < 0
and the system w i l l not return to y\.
A n a l y z i n g the equilibrium yn i n the same way-shows that it is an unstable
equilibrium, j u s t l i k e the equilibrium y — 0.
T h e preceding discus.sion illustrates that it is not necessarily the case
that the system w i U r e a c t í "an e q u i l i b r i u m v a l u é . T h e only e q u i l i b r i u m in
t h e n w i l 1
Figure 8.12 that can be reached is y^. I Í J ^ Í a l i y . . ^ 0 ) . . e X Q . ' l ' 2 ) ' -
approach y\. I f y ( 0 ) < 0, y(x) -> - c e , and if y > yi, then y(x) -+ ce.
T h e ""preceding discussion also illustrates that we can only subject an
equilibrium to a small perturbation i f we want to learn something about its
stability: I f we perturb too much so that the valué after the perturbation
is either less than 0 or greater than yi, the solution will not return to the
equilibrium .valué y\.
1
fjjJJSnHBpB L e t N{i) denote the size of a population at time t that evolves according to
the logistic equation
for N > 0
dt V 100,
F i n d the equilibria and analyze their stability.
Solution
T o find the equilibria, we set ^ = 0. T h a t is,
N
2N 1 = 0
100,
We see that either
Ni = 0 or ^ 2 = 100
T o analyze the stability, we draw the graph of dN/dt versus N in Figure 8.13.
Note that N > 0, since N represents the size of a population. T o perturb the
trivial e q u i l i b r i u m Ni = 0, we therefore choose only v a l ú e s that are slightly
bigger than 0. W e see from the graph that such small perturbations result in .
8.2 ii E q u i l i b r i a a n d T h e i r Stability 497
dN/dt > O, and henee N increases (see arrow). T h i s implies that Ni = 0 is an

unstable equilibrium.
T o perturb the nontrivial equilibrium A 2 — 100, we can either increase
r
or decrease the population size a bit. I f we decrease it, then dN/dt > 0 and
the population size will increase. I f we increase it, then dN/dt < 0 and the
population size will decrease. T h a t is, if the system is subjected to a small
perturbation about the nontrivial equilibrium Nj = 100, the population size
will return to the equilibrium valué of Nj = 100, as indicated by the arrows.
T h i s implies that A'2 = 100 is a locally stable equilibrium. -4
Analytical Approach W e assume that y is an equilibrium of

dy
dx
[thus y satisfies g{y) = 0]. W e consider a small perturbation about the
equilibrium y, which we express as
y = y +z
where z is small and may be either positive or negative. T h e n
dy ^ ^ - _j_ -j dz
dx dx dx
since dy/dx = 0 (y is a constant). W e find
dt ' .
— = giy 4- z)
dx
I f z is sufficiently small, we can approximate g(y + z) by its linear approxima-
tion. T h e linear approximation oí g(y) about y is given by
L(y) = g(y) + (y-y)g'(y)

Since g{y) = 0,
L{y) = (y-y)g'(y)
Therefore, the linear approximation of g(y + z) is given by
LXy + z) = íy + z - y)g'(y) = zg'iy)
I f we set
A. = g'íy)
then
É'-Xz
is the ñ r s t - o r d e r approximation of the perturbation. I t has the. solution

= Z (0)e^ (8.53)
which has the property

lim z(x) = 0 if X < 0
Since y ( . r ) = y + z ( x ) , it follows that if A' < 0, the system returns to the

equilibrium y after a small perturbation z { 0 ) . T h i s means that y is locally
stable if A < 0. O n the other hand, i f A > 0, then z(x) does not go to 0 as
x -+ 0 0 , implying that the system w i l l not return to the equilibrium y after a
small perturbation. and y is called unstable. T h e v a l u é k = g'(y) is called an
eigenvalue; it is the slope of the tangent fine of g ( y ) at y.
STABILITY CRITERION ' Consider the differential equation
w h e r e ¿ ( y ) is a differentiable function. A s s u m e that y is an equilibrium,

that is, g(y) = 0. T h e n
J
y is locally stable ifg (y) <.0 : . . •
:
y is unstable if gf {y) > Q ".'* ' '
W h e n the eigenvalue k == 0. the first-order a p p r o x i m a t i o n ^~ = kz

T
does not a ü o w us i o d r a w any conclusions about the behavior of'z(x) since
higher-order t'errns then become important.
W e w i s h to tie this analytical approach i n w i t h the m o r e informal
graphical analysis i n the beginning of this section. L o o k i n g at F i g u r e 8.12, we
find that the slope of the tangent line at y = 0 is positive, that is, g'(0) > 0;
similarly, w e find g ' ( y i ) < 0 and g'iyi) > 0. Henee, the equilibria 0 and y?
are unstable and the equilibrium y\ is locally stable, as found in the graphical
analysis.
L e t ' s try the analytical approach out on E x a m p l e 1. T h e r e
N
g{N)=2N í l -
íoo
T o differentiate g(N), we multiply out,
2
N
g(N) = 2N-
50
Differentiating g(N) yields
- A 'I I f A' = 0, then g'(0) = 2 > 0; thus N -- 0 is an unstable equilibrium. If

N = 100, then g'(100) = 2 - 200/50 = - 2 < 0; thus N = 100 is a locally
stable equilibrium.
T h e analytical approach is more powerful than the graphical approach:
I n addition to determining whether an equilibrium is locally stable or unstable,
it allows us to say something about how quickly a solution returns to an
equilibrium after a small perturbation. T h i s follows from (8.53). T h e r e , we
found that the perturbation has the approximate solution
kx
Z(x) = z(Ú)e
I f X > 0, then the larger k, the faster the solution moves away from the
equilibrium. I f k < 0, then the more negative k, the faster the solution will
return to the equilibrium after a small perturbation (see F i g u r e 8.14).
T h e preceding derivation for the analytical stability criterion was based
on linearizing g[y) about the equilibrium y. Since the linearization is only
cióse for v a l ú e s cióse to y [unless g(y) is l i n e a r ] , the perturbations about the
equilibrium must be small, and henee the stability analysis is always local, that
is, w i t h i n c i ó s e vicinity of the equilibrium. W h e n g(y) is linear, the'analysis is
exact; in particular, we can compute exactly how quickly a solution returns
to a locally stable equilibrium (or moves away from an unstable equilibrium)
after a perturbation.
8.2 H Equilibria and Their Stability 499
A. Figure 8.14
x
T h e graph of i(x) = z(Q)e''- for X > 0 and ;. < 0
Example 2 g ( y ) is linear Show that the differential equation

dy
= l - y (8.54)
dx
has a locally stable equilibrium at y = 1, and determine how quickly a solution
starting at y(0) = yo 9= 1 will reach y = 1.
Solution
Since g ( y ) = 0 for y = 1, y = 1 is an equilibrium. Differentiating g(y) = 1 — y,
we ñ n d g'(y) = — 1 , which is negative regardless of y. Therefore, y = 1 is a
locally stable equilibrium.
W e can solve (8.54) exactly using separation of variables:
— ln 11 — y | =t x + C j
Cl x
1 - y = ±e~ e-
x
/ \-y = Ce-
Cl
where we set ±e~ = C. Solving for y yields
x
y{x) = 1- Ce~
If we set y(0) = yo, then yo = 1 - C , and the solution is

x
y(x) = 1 - (1 - yo)e-
W e see that for any initial v a l u é yo,

l i m y(x) - 1
and it takes an infinite amount of time to reach the equilibrium y — 1 if

yo ¿ !•
Instead of asking how long it takes .until the equilibrium is reached
(which yielded the uninformative answer: an infinite amount of time), we
compute the time it takes to reduce the initial deviation from the equilibrium,
ím
yo - 1, to a fraction e~ , that is, we wish to determine the number XR such that
i
y{x )-l
R = - (yo-D
e
dcvÍBiion írom X—Xfí inilisl devmlior

Chapter 8 • Differential Equations •
Since. y(x ) R 1 = •(1 — yo)e X R

, we find
•(1 - y )e'-XR -i - i' {yo - 1)

Q
, - X R l
XR = 1
(see F i g u r e 8.15). I t takes one unit of time to reduce the-initial deviation to a

l
fraction e~ \ this does not depend on how large the initial deviation is. "4
> 1 fffTnTTJ¡J^ff Using Both Approaches Suppose that

dy
= >'(4 - y)
dx
F i n d the equilibria of this differential equation and discuss their stability,
using both the graphical and the analytical approach.
Solution
T o find the equilibria, we set dy/dx = 0. T h a t is,
, '' y(4-y) = 0
which yields
yi = 0 and yi = 4
I f we set g{y) = y ( 4 - y ) and graph giy) (see Figure 8.16), we see that )>j = 0
is unstable, since if we perturb yi = 0 to y = z > 0 where z is small, then
giy) = dy/dx > 0, and if we perturb yi — 0 to y = z < 0 where z is small,
J--f
then g{y) = dy/dx < 0. T h a t is, in either case, y will not return to 0. On
the other hand. yi is a locally stable equilibrium. A small perturbation to the
right of y = yi results in dy/dx < 0, whereas a small perturbation to the left
results i n dy/dx > 0. I n either case, the solution y will return to y = 4. 2
I f w e use the analytical approach, we need to compute the eigenvalues.

T h e eigenvalue associated with y i = 0 is
AJ =g'{0) = 4 - 2v|v=o = 4 > 0
which implies that yj is unstable. T h e eigenvalue associated with V2 = 4 is
A 2 = g'(4) = 4 - 2 y | v = 4 = -4 < 0
which implies that y = 4 is locally stable.

2 4
Vf, ~ 1
1 + e - ' ( . v - 1)
0
A. Figure 8.15
A n illustration of the time XR in Example 2
8.2 fl Equilibria and Their Stability 501
giy) = y (4 - )
y
A Figure 8.1 6
T h e graph of giy) in E x a m p i e 3
8.2.2 Single Compartment or Pool

T h i s example is adapted from D e A n g e l i s (1992). Compartment models are
frequently used to model the flow of matter (nutrients, energy, and so forth).
T h e simplest such model consists of one compartment, for instance, a fixed
Rate q volume V of water (such as a tank or l a k e ) containing a solute (such as
phosphorus). A s s u m e that water enters the compartment at a constant rate q
and leaves the compartment at the same rate (see F i g u r e 8.17). ( H a v i n g the
same input and output rate k e e p s ' t h é volume of the pool constant.) W e will
A. Figure 8.17
investígate the effects of different input concentrations of the solution on the
lnput and output rates are the
concentration of the solution in the p o o l .
same: T h e water in the tank
remains at the same level W e denote by Cit) the concentration of the solution in the compartment
at time t. T h e n the total mass of the solute is C ( f ) V, where V is the volume of
the compartment. F o r instance, i f the concentration of the solution is 2 grams
per liter and the volume of the compartment is 10 liters, then the total mass
1
of the solute in the compartment is.2 g l i t e r s ^ times 10 liters, w h i c h is equal
to20g. ,
I f C ; is the concentration'of the incoming solution and q is the rate
at.which water enters, then qC¡, the input loading, is the rate at which
mass enters. F o r instance, i f the concentration of the incoming solution is
- 1 _1
5 g l i t e r s and the rate at w h i c h the solution enters is 0.1 liter s , " t h e n the
- 1
input loading—that is, the rate at which mass enters—is 5 g l i t e r s times
- 1 -1
0.1 liter s , which is equal to 0 . 5 . g s .
I f we assume that the solution i n the compartment is w e l l mixed, so
that the outflowing solution has the same concentration as the solution i n the
compartment—namely C ( r ) at time t—then qC{t) is the rate at which mass
leaves the compartment at time t.
C(t)\
These different processes Can be schematically illustrated in a flow
diagram, as shown in Figure 8.18. Because mass is conserved i n the system,
we can use the conservation of mass Iaw to derive an equation that describes
the flow of matter in this system:
A. F i g u r e 8.18
Flow diagram for the single
compartment model rate of change
rate at which rate at w h i c h
of mass of
mass enters mass leaves
solute in pool
, C < C,
0
,, Q > Q
1 ii V
s
/
/
/
A. Figure 8.19
The solution curves for the single compartment model for-different valúes of CQ
or, w r i t i n g C for C ( r ) (and being careful not to confuse C with a constant).

', d
—(CV) = qC¡-qC (8.55)
Since V is constant, w e can write (8.55) as

dC
(8.56)
~dt
T h i s is a h n e a r differential equation of the type discussed in Subsection 8.1.2,
Case 1. I t can be solved using separation of variables. W e skip the steps, to
c o n c é n t r a t e on the discussion of the system. I f C ( 0 ) = CQ, then the solution
,-c.V - £ i- -t « i of the differential equation is
co- ,-(g/V)t
ca) = C¡ 1 - (8.57)
. i- L C,
Solution curves for different valúes of Co are shown in F i g u r e 8.19. We

conclude f r o m (8.56) that C¡ is the only e q u i l i b r i u m . L o o k i n g at the solution
C ( / j i n (8.57), w e see that "
l i m C ( 0 = C]
í-i-00
regardless of the initial concentration CQ in the compartment. T h i s shows that
C¡ is globally stable. Global,stability. implies that no matter^how much we
perturb" the equilibrium,,the system .will, r e t u r n to the equilibrium.
W e c a n also obtain this result using the eigenvalue method. W e write
C(r) = C / . + z(f)
T h e n , as before,
d C i l ) d D Z
- l r j . ¿M ^
V
that is,
dz Q
(8.5S)
since C¡ - C ( 0 = -z(t). Note that (8.58) is exact, that is, we need not use
a linear a p p r o x i m a t i o n for the right-hand side of (8.56), since it is already
linear. W e see from (8.58). that —q/V is the eigenvalue associated with the
equilibrium Cj. Since q and V are both positive, the eigenvalue is negative,
8.2 9 Equilibria and Their Stability 503
and it follows that C¡ is locally stable. W e can obtain more information here
since (8.58) is exact with solution
{ v
z(t) = z(0)e- ^ >' (8.59)
which shows that for any perturbation z(0), z(t) -*• 0 as í -> oo. T h a t is, the
system returns to the e q u i l i b r i u m C¡ [that is, C ( r ) -> C¡ as t -y oo] regardless
of how much we perturb the system. T h e reason that the eigenvalue method
allows us to show global stability lies in the fact that the differential equation
(8.55) is linear. I n j o t h e r cases, the eigenvalue method only allows us to
obtain local stabifity. because we must first linearize and, consequently, the
linearized differential equation is only an approximation.
E q u a t i o n (8.59) shows that the system returns to the equilibrium C / _
exponentially. fast and that the eigenvalue - (q/V) determines the time scale,
that is, how quickly the system reaches the equilibrium. T h e larger q/V,
the faster the system recovers from a perturbation. I n this context, we can
define a return time to equilibrium (just as in E x a m p l e 2 ) , denoted by TR. B y
convention, the return time to equilibrium, T , is defined as the amount of R
l
time it takes to reduce the initial difference CQ — C¡ to a fraction e~ , that is,
L
- C ( T R ) - C ¡ = e ~ ( C Q - C ¡ ) (8.60)
U s i n g (8.57), we see that
V T
C(7» = C/ 1 - ( l - - £ ^ J E - W > *
which we can rewrite as

( / K ) 7
C(JR)-C¡ = (C -C/) - ?
0 e * (8.61)
E q u a t i n g (8.60) and (8.61), we find

1 7
e - (C -C ) = 0 / (Co-C/)*--^ "*
which vields
1 = ^ T R or TR -
V q
T h a t is, the return time increases w i t h the volume V and decreases with the
flow rate q. This should be intuitively clear; the larger the volume and the
smaller the input rate, the longer it takes to return to equilibrium. W e can
show that f can also be interpreted as the mean residence time of a molecule
R
of the solute', that is, T is the average time a molecule of the solute spends in
R
the compartment before leaving, w h e n the system is i n equilibrium.
8.2.3 The Levins Model

T h e ecological importance of spatial structure to the maintenance of popula-
tions was pointed out by A n d r e w a r t h a and B i r c h (1954) based on studies of
insect populations. T h e y observed that although local populations become fre-
quently extinct, their patches become subsequently recolonized by migrants
from other patches occupied by this population, thus allowing the population
to persist globally. Fifteen years later, i n 1969, R i c h a r d L e v i n s introduced
the concept of metapopulations ( L e v i n s , 1969). T h i s was a major theoretical
advance; this concept provided a theoretical framework for studying spatially
structured populations.
A metapopuiation is a collection of subpopulations. E a c h subpopulation
occupies a patch, and different patches'are linked via migration of individuáis
between patches (see F i g u r e 8.20). I n t h i s setting, we only keep track of what

proportion of patches are occupied by subpopulations. Subpopulations go
O extinct at a constant rate, denoted by m {in stands for m o r t a l i t y ) . Vacant
patches can be colonized at a rate that is proportional to the fraction of
occupied patches; the constant of proportionality is denoted by c (c stands
-o for colonization rate). I f w e denote by__p{t) the fraction of patches that are
o occupied at time t, then writing"p = p\t),
o dp i'
— = cp(l - p) -mp
' ' - f..,v.». -'>....,
(8.62)
dt > \ . • ., f.
A. Figure 8.20
T h e first term on the right hand-side describes the colonization process. Note
A schematic flescription of a
that an increase in the fraction of occupied patches occurs only i f a vacant
metapopulation model. The
patch becomes occupied; henee the product p(l — p) in the first term on the
shaded patches are occupied;
arrows indícate migration right-hand side. T h e minus sign in front of m shows that an extinction event
events ; decreases the fraction of occupied patches.
W e w i l l not solve (8.62), but í o c u s on its equilibria. W e set
cp(l — p) — mp = 0
A f t e r factoring'cp, we obtain
m
11 p = 0
c
-» I ^* " -VA
-> r which has the two solutions:
m
Pi = 0 and pi = 1
We cali the solution p\ = 0 a trivial solution because it corresponds to the

situation in w h i c h all patches are vacant. Since i n d i v i d u á i s are not-created
spontaneously, a vacant patch can be recolonized only through migration
from other occupied patches. T h e r e f o r e , once a metapopulation is extinct,
it stays extinct. T h e other equilibrium pi = 1 - m/c is only relevant when
P2 e ( 0 , 1 ] , because p represents a fraction that is a number between 0 and
1. Since m and c are both positive, it follows immediately that p < 1 for all 2
choices of m and c. T o see when p > 0, we check

2
1-™ >0
c
which holds w h e n
m < c
T h a t is, the n o n t r i v i a l equilibrium p "=_1 — m/c is in ( 0 , 1 ] i f the_extinction
2
rate m is less than the colonization rate c. I f m > c, then there is only one
equilibrium in [0, 1], namely p\ = 0. W e ¡Ilústrate this in F i g u r e s 8.21 and
8.22; looking at the graphs, we can analyze the stability of the equilibria..
Case 1: m > c. T h e r e is only the trivial equilibrium p\ = 0. F o r any

p e (0, 1], we see that dp/dt < 0; henee the fraction of occupied patches
declines. T h e equilibrium is locally and globally stable.
Case 2: m < c. T h e r e are two equilibria, namely 0 and 1 — m/c. The

trivial e q u i l i b r i u m p\ = 0 is now unstable, since i f we perturb p\ « 0
to some v a l u é in (0, 1 — m/c), then dp/dt > 0, which implies that p{¡)
increases. T h e system will therefore not return to 0.
T h e other equilibrium, p = 1 — m/c, is locally stable. After a small
2
perturbation of this equilibrium, to the right of p , dp/dt < 0, and to the

2
left of p2, dp/dt > 0; therefore, the system will return to p . 2

A Figure 8.21
T h e case m > c
cp(\ - p) - .mp •
A Figure 8.22
T h e case m < c
W e can also use the eigenvalue approach to analyze the stability of the
equilibria. I n addition, this w i l l allow us to obtain information on how quickly
the system returns to the stable equilibrium. W e set
m ¿
g(p) = cpiX - P) ~ P - C í> -<= f - f
T o linearize this function about the equilibrium valúes, we must find
g'(p) =.c-2cp - m
N o w , if pi = 0, then
> o
whereas, if pz = 1 — m / c , then
'\ m
2c 1 - — n% — c — 2c + 2m — m = m
• ,i - -
F r o m this we see that c - m is the eigenvalue corresponding to p\ 0 and
m - c is the eigenvalue corresponding to p¿ = 1 - m/c. W e find that
ifc-m<0 then p¡ = 0 is locally stable
if m - c < 0 then pi = 0 is unstable and p 2 = 1 — is locally stable
T o summarize our results, if m > c, then p\ = 0 is the only equilibrium
in [ 0 , 1 ] and p\ = 0 is locally stable. ( I n fací, the graphical analysis showed
that pi = 0 is globally stable.) I f m < c, then there are two equilibria in [ 0 , 1 ] .
T h e equilibrium p\ = 0 is now unstable and the equilibrium pi = 1 - m/c is
locally stable..
A Figure 8.23
The graph of g(N) illustrating the Allee effect
8.2.4 The Allee Effect \JO>fif?m £.. A l ' e . c . (< T Í : ¡ 4 ' " ' !
I n a sexually r é p j o d u c i n g species, i n d i v i d u á i s may experience a dispropor-
tionately l o w r é c r m í m é n t rate when the population density falls below a
(
.,. certain level,-due to Iack~of suitable mates. T h i s is called an A l l e e effect

( A l l e e , 1931). A simple e x t e n s i ó n of the logistic equation incorporates this
effect. W e denote by N(t) the size of a population at time r; then, writing
N = N(l),
dN ( N
— = r N ( N - a ) i l - - $.63)
dt \> K
where r, a, and K are positive constants. W e assume that 0 < a < K. We
will see that, as in the logistic equation, K denotes the c a r r y i n g capacity.
T h e constant a is a threshold population size, below which the recruitment
rate is negative, meaning that the population will shrink and ultimately go to
rt
extinction. -«ct
T h e e q u i l i b r i a of (8.63) are given by Ñ = 0, a, and K. W e set
; 2
g(N) = rN(N - « ) ( ! - A
K K K
A graph of g(N) is shown in Figure 8.23. Differentiating g(N) yields

¿
2N
2N + — N
K K. K •(2NK +2aN -2N2
- aK)
7
We can compute the eigenvalue g'{Ñ) associated with the e q u i l i b r i u m A .
;
ifÑ = 0 then g'(0) = (-aK) <0 i ^ - ••..-.
ifÑ = a then g'(a) = j?a{K - a)'> 0 y..--.-
ifÑ = K then g'{K) = j?K(a - K)'< 0 t\ i-m
A s we c o n t i n u é , you should compare the results from the eigenvalue

method with the graph of g(N).
Since g'(Q) < 0, it follows that A' = 0 is locally stable. L i k e w i s e , since
g'(K) < 0, it follows that Ñ = K is locally stable. T h e equilibrium Ñ = a
is unstable since g'(a) > 0. This, is also evident from F i g u r e 8.23. T h i s is an
example where both stable equilibria are locally but not globally stable. \ln v'<
8.2 m Equilibria and Their Stability 507
W e see from F i g u r e 8.23 that if 0 < N(0) < a, then N (t) -+ 0 as t oo.
I f a < N(0) < K oí N(0) > K, then N(t) -> K as f -> oo. T o interpret
our results, we see that if the initial population 7V(0) is too small [that is,
N(Q) < a ] , then the population goes extinct. I f the initial p o p u í a t i o n is large
enough [that is, A/(0) > a], the population persists. T h a t is, the parameter
a is a threshold level. T h e recruitment rate is only large enough when the
population size exceeds this level.
8.2.5 Problems
(8^2.1) , r'é.) (A simple model of predation) Suppose that NU) denotes
Suppose that rhre size of a population at timé t. T h e population evolves
dy according to the logistic equation but, in addition, predation
v (2 - y ) reduces the size of the population so that the rate of change
dx
/is given by . .•
(a) Find the equilibria of this differential equation.
(b) Graph dy/dx as a function of y, and use your graph to = N 11 - — 1 - (8.64)
dt \~ 50 J 5+ N
disct*SS the stability of the equilibria.
T h e first term on the right-hand side describes the logistic
(c) Compute the eigenvalues associated with each equilib-
growth; the secorid term describes the effect of predation.
rium and discuss the stability of the equilibria.
(a) Set
f 2i\ Suppose that
N\ 9N
g(N) = N { \ - -
/ = (2-y)(3-y) 5+ N
dx
and graph g{N).
(a) F i n d the equilibria of this differential equation.
(b) Graph dy/dx as a function of y, and use your graph to (b) F i n d all equilibria of (8.64).
discuss the stability of the equilibria. (c) Use your graph in (a) to determine the stability of the
(c) Compute the eigenvalues asseeiated with each equilib- equilibria found in (b).
rium and discuss the stability of the equilibria. (d) Use the method of eigenvalues to determine the stability
3, Suoposethat of the equilibria found in (b).
d\ 7. (Logistic equation) Assume that the size of a population
dx evolves according to the logistic equation with intrinsic rate
(a) Find the equilibria of this differential equation. of growth r = 2. Assume that /V(0) = 10.
(b) Graph dy/dx as a function of y, and use your graph to (a) Determine the carrying capacity K if the population
discuss the stability of the equilibria. grows fastest when the population size is 1000. (Hint: Show
(c) Compute the eigenvalues associated with each equilib- that the graph of dN/dt as a function of N has a
rium and discuss the stability of the equilibria. m á x i m u m at K/2.)
4. Suppose that (b) I f N{0) = 10, how long will it take the population size to
- dy reach 1000?
= y(l-y)(y-2)
dx
(c) F i n d lim,-,.:» N(t).
(a) F i n d the equilibria of this differential equation.
8. (Logistic equation) T h e logistic curve N(t) is an í - s h a p e d
(b) Graph dy/dx as a function of y, and use your graph to
curve that satisfies
discuss the stability of the equilibria.
dN f N\
(c) Compute the eigenvalues associated with each equilib-
rium and discuss the stability of the equilibria. —- = r N \ \ - - j with N(0) = A' 0 (8.65)
5.! (Logistic equation) Assume that the size of a population
evolves according to the logistic equation with intrinsic rate of when NQ < K.
growth r = 1.5. Assume that the carrying capacity K — 100. (a) Use the differential equation (8.65) to show that the
(a) F i n d the differential equation that describes the rate of inflection point of the logistic curve ís at exactly half the
saturation valué of the curve. (Hint: do not solve (8.65);
growth of this population.
instead differentiate the right-hand side of (8.65) with
(b) F i n d all equilibria, and discuss the stability Qf the
respect to t.)
equilibria using the graphical approach.
(b) T h e solution N(t) of (8.65) can be defined for a l l . ' 6 R .
(c) F i n d the eigenvalues associated with the equilibria, and Show that N(t) is symmetric about the inflection point and
use the eigenvalues to determine stability of the equilibria. that N ( 0 ) = A'o- That is, first use the solution of (8.65) that is
Compare your answers with your results in (b). ' '.. given in (8.32) and find the time ÍQ so that N (ÍQ) = K/2, that
R
is, the inflection point is at t = tr¡. Compute N(tQ + h) and (b) . Solve the differential equation in (a) when C ( 0 ) = 0 and
7 —
A ('0 h) for h > 0, and show that find lim/ôo C ( r ) .
(c) Find all equilibria of the differential equation and discuss
•' •' N(tQ + k) - ~jm y - N(io - h)
their stability.
Use a sketch of the graph of N(t) to explain why this shows
, 14. Suppose that a tank holds 1000 liters of water, and 2 kg
that A'(f) is symmetric about the inflection point (ÍQ, A (/Q)).
of salt is poured into the tank.
¡ 9.J Suppose that a fish population evolves according to the
- 1
ftígistic equation, and that a fixed number of fish per unit time (a) Compute the concentration of salt in g l i t e r .
are removed. T h a t is, (b) A s s u m e now that you want to reduce the salt concentra-
dN N tion. One method would be to r e m o v e a certain amount of
rN H the salt water from the tank and then replace it by p u r é water.
~dt " ~ ~K
How much salt water do you have to replace by p u r é water
Assume in the following that r =-2 and K = 1000. - 1
to obtain a salt concentration of 1 g l i t e r ? '
(a) F i n d possible equilibria, and discuss their stability when
(c) Another method to reduce the salt concentration would
H = 100.
be to hook up an overfiow pipe and pump p u r é water into
(b) What is the maximal harvesting rate that maintains a
the tank. T h i s way, the salt concentration would be gradually
positive population size? . •<<•;.•»/. ír' , n
reduced. Assume that you have two pumps, one that pumps
í o ) Suppose that a fish population evolves according to - 1
water at a rate of 1 liter s , the other at a rate of 2 liters s . - 1
a'logistic equation and that fish are Tiarvested at a rate F o r each pump, find out how long it would fake to reduce
proportional to the population size. Ií-N(t) denotes the the salt concentration from the original concentration to
population size at time r, then - 1
1 g l i t e r , and how much p u r é water is needed in each case.
dN (Note that the rate at which water enters the tank is equal to
= rN 1 hN
- dr the rate at which waterleaves thetank.) Compare the amount
Assume that r = 2 and K = 1000. of water needed using the pumps with the. amount of water
(a) Find possible equilibria and discuss their stability when needed in part ( b ) .
h = 0.1, using the graphical approach, and find the maximal
•fl5j) Assume the single compartment model we introduced in
harvesting rate that maintains a positive population size.
Subsection 8.2.2. Denote by C ( r ) the concentration at time i
(b) Show that " i f h < r = 2, then there is a nontrivial and assume
equilibrium. Find the equilibrium.
— = 0.37(254 - C(f)) f o r r > 0
(c) Analyze the stability of the equilibrium found in (b) using dt
(i) the eigenvalue approach and (ii) the graphical approach. (a) Find the equilibrium concentration. • -tU*<
(8.2.2) (b) Assume that the concentration is suddenly increased
from the equilibrium concentration to 400. F i n d the return
11. Assume the single compartment model defined in
time to equilibrium, denoted by TR, which is the amount of
Subsection 8.2.2: I f C(r) is the concentration of the solute - 1
time until the initial difference is reduced to a fraction e .
at time r, then dC/dr is given by (8.56), that is,
dC q „ (c) Repeat (b) for the case when the concentration is sud-
— = -(C, - C) denly increased from the equilibrium concentration to 800.
dt V '
where q, V, and C¡ are defined as in Subsection 8.2.2. U s e the (d) A r e the valúes for TR computed in (b) and (c) different?
graphical approach to discuss the stability of the equilibrium 16. Assume the compartment model "as in Subsection 8.2.2.
c = c,. Suppose that the equilibrium concentration is C¡, and the ini-
Í 2 i Assume the single-compartment model defined in Sub- tial concentration is CQ. Express the time it takes until the ini-
section 8.2.2; that is, denote by C(t) the concentration of the tial deviation CQ — C¡ is reduced to a fraction p in terms of TR.
solute at time / and assume that
17, Assume the compartment model as in Subsection 8.2.2.
dC
— = 3 ( 2 0 - C ( r ) j for ; > 0 (8.66) Suppose that the equilibrium concentration is C¡. T h e time
dt TR has an integral representation which can be generalized to
(a) Solve (8.66) when C(0) = 5 . systems with more than one compartment. Show that
(b) Find lim,_»8oC(/). •°° C(t)-C,
(c) U s e your answer in (a) to determine t so that C ( r ) = 10.
^li. Assume the single-compartment model defined in Sub- [Hint: Use (8.57) to show that
section 8.2.2; that is, denote by C(t) the concentration of the
solution at time t and assume that the concentration of the C(0) - c,
- 1
incoming solution is 3 g l i t e r s and the rate at which mass and intégrate both sides with respect to / from 0 to oo.]
- 1
enters is 0.2 liter s . Furthermore, assume that the volume 18. Use the compartment model defined in Subsection 8.2.2
of the compartment V = 400 liters. to investígate how the size of a lake inñuences nutrient
(a) Find the differential equation for the rate of change of dynamics in the lake after a perturbation. Mary L a k e and
the concentration at time t. Elizabeth L a k e are two fictitious lakes in the North Woods
N~25o
x
ZM~ ZN - ü, 1N
19.-ZH 4úOO
1ÓOQ
iooo
8.2 • Equilibria and Their Stability 509
that are used as experimental lakes tostudy nutrient dynamics. (a) Set g{p) = 0.5p(l -p) - 1.5p. G r a p h g{p) for p e [0, 1].
1 5
Mary Lake has'avolume of 6.8 x 10- m- , Elizabeth Lake has
twice this volume, namely, 13.6 x 1 0 m . Both lakes have (b) F i n d all equilibria of (8.68) that are in [0, 1], and
3 3
the same ¡nflow/outflow rate q = 170 liters s . Because determine their stability using your graph in (a).
both lakes share the same drainage área, the concentration
C¡ of the incoming solute is the same for both lakes, (c) Use the eigenvalue approach to analyze the stability of
namely, C¡ = 0 . 7 m g l i t e r . Assume that in the beginning the equilibria that you found in (b).
-1
; : :.-.s Í K ~ Í - - Z - \ lakes are ir. equilibrium. that is. the

yíj} (A metapopulation model with density-dependent
1
extinc-
::-.:e.-.:-a: :-. ::' :r.t soiu'.ion in both lakes is 0.7 mg l i t e r " .
tion) Denote by p = p(t) the fraction of occupied patches in
exrerirrsent co.-.s:s:s of increasing the concentration a metapopulation model, and assume that
of the solution by 10% in each lake at time 0 and then
1
watching how the concentration of the solution in each lake Y = cpQ - p) - p forr>0 (8.69)
changes with time. A s s u m e the single compartment model to 2
where c > 0. T h e term p describes density-dependent
make predictions on how the concentration of the solution
extinction of patches, namely the per patch extinction rate is
wil! evolve. (Note that 1 rcr of water corresponds to 1000
p and a fraction p of patches are occupied, resulting in an
liters of water.) 2
extinction rate of p . T h e colonization of vacant patches is
(a) F¡nd the initial concentration CQ of the solution in each the same as in Levins model.
. •.; : Z that is, hnmediately after the 10% increase in
•• : - :cr.centration). (a) Set g(p) = cp(\ - p) - p 2
and sketch the graph of g(p).
(b) Use Equation (8.57) to determine how the concentration
of the solution changes over time in each lake. Graph your (b) F i n d all equilibria of (8.69) in [0,1], and determine their
stability.
results.
(c) Which lake returns to equilibrium faster? Compute the (c) l s there a nontrivial equilibrium when c > 0? Contrast
return time to equilibrium, TR, for each lake, and explain your findings with the corresponding results in Levins model.
-how this is related to the eigenvalues corresponding to the
equilibrium concentration C¡ for each lake. 23. (Habitat destruction) In Subsection 8.2.3, we introduced
19. Use the compartment model defined in Subsection 8.2.2 L e v i n s model. T o study the effects of habitat destruction on
to investígate the effect of an increase in the input concen- a single species, we modify Equation (8.62) in the following
tration C¡ on the nutrient concentration in a lake. Suppose way. W e assume that a fraction D of patches is permanently
destroyed. Consequently, only patches that are vacant and
a iake in a pristine environment has an equilibrium phos-
- 1 undestroyed can be successfully colonized. These patches
phorus concentration of 0.3 m g . T h e volume V of the lake
6 3
have frequency 1 - p(t) — D i f p(t) denotes the fraction-of
is 12.3 x 1 0 t n and the infiow/outflow rate q is equal to
_ 1
occupied patches at time t. T h e n
220 liters s . Conversión of land in the drainage área of the
lake to agricultural use has increased the input concentra- ^ = cp{\ - p - D)-mp (8.70)
- 1 - 1
tion from 0.3 mg l i t e r to 1.1 mg l i t e r . Assume that this
(a) E x p l a i n in words the meaning of the different terms in
increase happened instantaneously. Compute the return time
(8.70).
to the new equilibrium, denoted by TR, in days, and find
the nutrient concentration in the lake TR units of time after (b) Show that there are two possible equilibria, the trivial
3
the change in input concentration. (Note that 1 nv of water equilibrium p\ = 0 and the nontrivial equilibrium pi =
corresponds to about 1000 liters of water.) 1 — D — -jr. Sketch the graph of p as a function of D.
2
(8.2.3)
/
(c) Assume that m < c, such that the nontrivial equilibrium
2 0 . ) (Levins. model) Denote by p = p(t) the fraction of
is stable when D = 0. F i n d a condition for D such that the
occupied patches in a metapopulation model, and assume
nontrivial equilibrium is between 0 and 1, and investígate
that -r :
the stability of both the nontrivial equilibrium and the trivial
= 2 p ( l -p)-p for i > 0 (8.67) equilibrium under this condition.
(a) Set {p) = 2/7(1 -p)-p-
g Graph g{p) for p e [0, i ] .
(d) Assume the condition that you derived in ( c ) , that is, the
(b) Find all equilibria of (8.67) that are in [0, 1], and nontrivial equilibrium is between 0 and 1. Show that when
\ determine their stability using your graph in (a). the system is in equilibrium, the fraction of patches that are
(c) Use the eigenvalue approach to analyze the stability of vacant and undestroyed, that is, the sites that are available
the equilibria that you found in (b). for colonization, is independen'! of D. Show that theeffecíive
21. (Levins model) Denote by p = p(t) the fraction of colonization rate in equilibrium, that is, c times the fraction of
occupied patches in a metapopulation model, and assume available patches", is equal to the extinction rate. This shows
that that the effective birth rate of new colonies balances their
extinction rate at equilibrium.
— = 0 . 5 p ( l - p) - 1.5p forí>0 (8.68)
(8.2.4) 25. (Allee effect) Denote by N(t) the size of a population at

í 24.) (Allee effect) Denote by N(t) the size of a population at time /, and assume that
time t, and assume that dN ( N \ - •*
í o r ; 8 7 2
— = 0.3A'(/V-17) f 1 — j - ° ^ ' )
dN í N
— = 2N(N - 10) ^ 1 - — ; for r > 0 71)
(a) Find all equilibria of (8.72).
(a) Find ail equilibria of (8.71).
(b) U s e the eigenvalue approach to determine the stability
(b) U s e the eigenvalue approach to determine the stability of the equilibria found in (a).
of the equilibria found in (a).
(c) Set (c) Set
N
(.V)=2A'(A>-10) (1-755) g(N) = 03N(N - 17) 1
5
200
for N > 0 and graph g(N). Identify the equilibria of (8.72)
for N > 0 and graph g(N). Identify the equilibria of (8.71)
on your graph, and use your graph to determine the stability
on your graph, and use your graph to deteraiine the stability
of the equilibria. Compare your results with your findings in
of the equilibria. Compare your results with your findings in
(b). Use youT graph to give a graphical interpretation of the
(b). U s e your graph to give a graphical interpretation of the
eigenvalues associated with the equilibria.
eigenvalues associated with the equilibria.
8.3 S Y S T E M S O F A U T O N O M O U S E Q U A T I O N S ( O P T I O N A L ) •
I n the preceding two sections, we discussed models that could be described
by a single differential equation. I f we wish to describe models in w h i c h
several quantities interact, such as a competition model in w h i c h various
species interact, more than one differential equation is needed. W e cali this
then a system of differential equations. W e will restrict ourselves again to
autonomous systems, that is, systems whose dynamics do not depend explicitiy
on the independent variable (which is typically time),
T h i s section is a preview of Chapter 1 1 , where we will discuss systems
of differential equations in detail. A thorough analysis of systems of differ-
ential equations requires a fair amount of theory, which we w i l l develop in
Chapters 9 and 10. Since we are not yet equipped with the right tools to
analyze systems of differential equations, this section will h e rather informal.
A s with m o v i e previews, you will not k n o w the full sto.ry a f t e r y o u finished
this sections, but reading it w i l l (hopefully) convince you that systems of
differential equations provide a rich tool for modeling biological systems.
8.3.1 A Simple Epidemic Model

W e begin our discussion of systems of autonomous differential equations with
a classical m o d e l of an infectious. disease, the K e r m a c k - M c K e n d r i c k model
(1927; 1932; 1933). W e consider a population of fixed size N that, at time /,
can be divided into three classes: the susceptibles, S(t), who can get infected:
the infectives, / ( / ) , who are infected and can transmit the disease; arid the
removed class, R(t), who are immune to the disease. T h e flow among these
classes can be described by
S —y l —• R
W e assume that the infection spreads according to the mass action law that
we encountered' in the discussion of chemical reactions. E a c h susceptible
becomes infected at a rate that is proportional to the number of'infectives
/ . E a c h infected individual recovers at a constant rate. A gain in the class ol
infectives is a simultaneous loss in the class of susceptibles. L i k e w i s e , a'.gain
in the class of recovered individuáis is a loss in the class of infectives. W e can
8.5 • Review Problems 519
10. Assume the hierarchical competidor, model with two interaction is modeled as
species, introduced in Subsection 8.3.3. Specifically, assume dV
°Va - ~)-bVN
dp\ dt K
— =2 P l ( l - Pi) - Pi dN
= cVN -dN
dp~> ~dt
(a) Suppose the herbivore number is equal to 0. What
-jf- = 6 ^ ( 1 - pi - P2) - P2 - 2pipi
differential equation describes the dynamics of the plant
(a) Use the zero-isocline approach to ñ n d all equilibria
biomass? Can you explain the resulting equation? Determine
graphically.
the plant biomass equilibrium ¡n the absence of herbivores.
(b) Determine the numerical valúes of all equilibria.
11. Assume the hierarchical competition model with two (b) Now assume that herbivores are present. Describe the
- '.roduced in Subsection 8.3.3. Specifically, assume effect of herbivores on plant biomass, that is, explain the term
dpi -bVN in the first equation. Describe the dynamics of the
= 3pi(l - P i ) - P i herbivores, that is, how their population size increases and
It
dp2 p ,i . T
what contributes to decreases in population size.
— = _ V d ~ Pl ~ PÍ) ~ P2 ~ ¿P\Pl •
2
(c) Determine the equilibria by solving
at
i L se :.-;£ zero-isocline approach to find all equilibria dV , dN
— = 0 and =0
graphically.
dt dt
(b) Determine the numerical valúes of all equilibria.
A l s o determine the equilibria also graphically. E x p l a i n why
12. (Adapted from Crawley, 1997) Denote by V plant this model implies that "plant abundance is determined solely
biomass, and by N herbivore number. T h e plant-herbivore by attributes of the herbivore," as stated i n Crawley (1997).
,8.4 KeyTerms
Chapter 8 Review: Topics Discuss the following definitions 9. A J l o m é t r i c growth
and concepts. 10. E q u i l i b r i u m
1. Differential equation 11. Stability
2. Separable differential equation 12. Eigenvalue
3. Solution of a differential equation 13. Single-compartment model
4. P u r é time differential equation 14. Levins model
5. Autonomous differential equation 15. A l l e e effect
6. Exponential growth 16. K e r m a c k - M c K e n d r i c k model
7. V o n Bertalanffy equation 17. Z e r o isocline
8. Logistic equation 18. Hierarchical competition model
18.5 Review Problems

1. {Newton's law of cooling) Suppose that an object has and with T = 2 1 ° C . Use 7"(20) = 2 8 ° C to determine the
a
temperatura T and is brought into a r ó o m that is kept at a constant it.]

constant temperature T . Newton's law of cooling states that
a
2. (Adapted from Caín et ai, 1995) I n this problem we discuss
the rate of temperature change of the object is proportional a model for clonal growth in the white clover Trifolium repens.
to the difference between the temperature of the object and T. repens is a widespread perennial clonal'piant species. It
the surrounding m é d i u m . spreads through stolon growth (a stolon is a horizontal stem).
(a) Denote by 7"(/) the temperature at time r, and explain B y mapping the shape of a clone over time, Cain et al.
why estimated stolon elongation and die-back rates as follows.
dT Denote by S(t) the stolon length of the clone at time r.
— =k(T -T ) a
dt Cain et al. observed that the change in stolon length was
is the differential equation that expresses Newton's law of proportional to stolon length, that is,
cooling. dS
— cxS
(b) Suppose that it takes the object 20 min to cool from 3 0 ° C dt i
to 2 8 ° C in a room whose temperature is 21°C. How long Introducing the proportionality constant r, called the net
will it take the object to cool to 25°C, if it is at 30°C when growth rate, we find
it is brought into the room? [Hint: Solve the differential d
-l = rS (8.84)
equation in (a) with the initial condition 7/(0) = 3 0 ° C dt

Aplicaciones Ecuaciones Diferenciales PDF

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Aplicaciones Ecuaciones Diferenciales PDF

Transféré par

Droits d'auteur :

Formats disponibles

DIFFERENTIAL

In the simplest model of population growth, the growth rate at any

(We introduced this equation in Section 4.6.) E q u a t i o n (8.1) con-

is a differential equation that contains the first and second deriva-

separable ditíerential equations (the reason for this ñ a m e will

We discussed differential equations of the form (8.3) Ln Section 5.8.'

8.1 SOLVING DIFFERENTIAL EQUATIONS

¿JE. t ^ ~ ^ í is a solution of (8.5). T o check this, we differentiate N(t):

B u t , since g[u(x)] = g(y) and u'(x) dx - dy, this is the same as

^and then integrating both sides

8.1.1 Puré Time Differential Equations

y{x) = f f(u)du + C (8.10)

T o solve (8.9) formally, we s e p á r a t e variables; we write the differential

and then i n t é g r a t e both sides:

w h i c h is the same as (8.10).

since eos 0 = 1 (see F i g u r e 8.2). . • *^

I f w e changed the initial condition in E x a m p l e 1 to some oth'er v a l u é ,

8.1.2 Autonomous Differential Equations

T o interpret the biological meaning of autonomous, let's return to the

(0, 20) (10, 20)

Integrating both sides gives

Case 1 : g(y) = k(y — a) W e w i s h to solve

where ik and a are constants. W e assume k y¿ 0. Separating variables yields

E x p o n e n t i a t i n g both sides yields

Since an antiderivative of j r j ^ is - 5 ln |2 - 3yj, we find

Solving for y yields

T o determine C , we use the initial condition yo = 1 w h e n XQ = 1. T h a t is,

480 Chapter 8 • Differential Equations

and assume N(0) = NQ > 0. T h e change in population size is desoribed by the

Since N(0) — NQ = C , we write the solution as

'Y E q u a t i o n (8.18) shows that the population grows exponentially when

this situation. ( W e w i l l discuss this in Case 2.) E x p o n e n t i a l decay in a popu-

> I -"""""SM ( R e s t r i c t e d Growth: v o n B e r t u l a n f T y E q u a t i o n ) Thi.s example describes the

-In|A-¿| ^Jk/fCf "

After multiplying this equation by - 1 and exponentiating,

T h e solution is then given by

Case 2: g(y) = k{y — a)(y — b) W e now turn to differential equations of

provided y a and y sí b. W h e n a = b, w e must ñ n d an antiderivative of

T h e constant C can then be determined from the initial condition. W h e n

Separation of variables yields

W e use the partial fraction method to i n t é g r a t e the left-hand side.

Comparing the last term to the i ñ t e g r a n d , we find

- [ l n |y - 1 | - In |y~H 2|] = 2JC + Ci

Simplifying this results in

W e r e t u r n now to (8.22) when a =¡¿ b and y ^ c? or b,W e use the

simpler rational funcfions that we k n o w how to i n t é g r a t e ; namely, we write

Substituting B = - A from the first equation into the second, we find

8.1 H Solving Differential Equations 485

simpler rational funcfions that we k n o w how to i n t é g r a t e ; namely, we write

8.1 H Solving Differential Equations 485

simpler rational funcfions that we k n o w how to i n t é g r a t e ; namely, we write

Substituting 5 = - A from the first equation into the second, we find

j \ ( T h e Logistic E q u a t i o n ) T h e logistic equation describes the change in size of

W e can interpret the parameter r as follows: T h e per capita growth rate

the per capita growth rate is c i ó s e to r . ( I n Problem 42, we discuss a different

It follows from (8.32) that

approaches K from below. T o use separation of variables to solve (8.27), we

8.1.3 Allometric Growth f U s ? c v

Solving differential equations of the form (8.2) gets c ó m p l i c a t e d quickly.

T o solve this differential equation, we s e p á r a t e variables first and then

Solving for y, we find

Using the initial condition yo — 0 when .vn = 1 allows us to determine C :

W e will now turn to a biological application, namely allometric growth.