single variable calculus and applications · 2019-01-03 · single variable calculus a function is...
TRANSCRIPT
Sin
gle
Var
iabl
eCal
culu
san
dApp
licat
ions
Dudle
yCook
e
Trinity
Colleg
eD
ublin
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
1/
47
EC20
40Top
ic1
-Sin
gle
Var
iabl
eCal
culu
san
dApp
licat
ions
Rea
din
g
1Chap
ters
7.1-
7.3,
9.1-
9.4
and
10of
CW
2Chap
ters
6,7,
and
9of
PR
Pla
n
1Rev
iew
ofsingl
e-va
riab
leca
lculu
s
2O
ptim
izat
ion
3Con
cavi
tyan
dCon
vexi
ty
4A
pplic
atio
ns
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
2/
47
Sin
gle
Var
iabl
eCal
culu
s
Afu
nct
ion
isuse
fully
thou
ght
ofas
a“r
ule
”w
hic
hco
nve
rts
anin
put
(den
oted
typic
ally
byx)
into
anou
tput
(den
oted
typic
ally
byy).
Not
atio
nal
ly,we
write
y=
f(x
).In
wor
ds,
“yis
afu
nct
ion
ofx.”
xis
also
calle
dth
ein
dep
enden
tva
riab
leor
the
exog
enou
sva
riab
le.
yis
also
calle
dth
edep
enden
tva
riab
leor
the
endog
enou
sva
riab
le.
Inte
rms
ofec
onom
ics,
itis
impor
tant
tokn
oww
hat
thes
efu
nct
ions
‘look
like’
.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
3/
47
Eco
nom
icFun
ctio
ns
Mac
ro-
Nat
ional
inco
me
acco
unting.
The
link
bet
wee
nag
greg
ate
inco
me,
consu
mption
,in
vest
men
t,an
dgo
vern
men
tsp
endin
g.
Y=
C( Y
d) +
I(r
)+G
Mic
ro-U
tilit
yfu
nct
ions.
We
thin
kth
atin
div
idual
sfe
licity
dep
ends
onco
nsu
mption
.W
=u
(c)
Mic
ro-
Pro
duct
ion
funct
ions.
Indiv
idual
firm
sou
tput
dep
ends
onla
bor
.y
=f
(L)
Som
etim
eswe
leav
eth
efu
nct
ions
gener
alan
dot
her
tim
eswe
assu
me
afu
nct
ional
form
.It
alldep
ends
onhow
much
stru
cture
we
wan
tto
put
onth
eunder
lyin
gec
onom
ics.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
4/
47
Slo
pe
ofLin
ear
Fun
ctio
ns
The
sim
ple
stfu
nct
ion
we
consider
isa
linea
rfu
nct
ion.
Alin
eis
are
lation
ship
ofth
efo
rmy
=ax
+b.
Sta
rtat
any
poi
nt
(x0,y
0)
onth
elin
ean
dm
ove
alon
gth
elin
eso
that
the
x-c
oor
din
ate
incr
ease
sby
one
unit.
The
corr
espon
din
gch
ange
inth
ey-c
oor
din
ate
isca
lled
the
slop
eof
the
line.
The
slop
ete
llsus
the
rate
ofch
ange
:how
much
ych
ange
sw
hen
xch
ange
sby
agi
ven
amou
nt.
The
defi
nin
gch
arac
terist
icof
alin
eis
that
this
rate
ofch
ange
isco
nst
ant.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
5/
47
Slo
pe
ofLin
ear
Fun
ctio
ns
Cla
im:
the
slop
eof
y=
ax+
bis
a.
Pro
of:
pic
kx
=x 0
;th
isim
plie
sth
aty 0
=ax
0+
b.
Incr
ease
x 0to
x 0+
∆x.
The
corr
espon
din
gnew
valu
eof
yis
y 1=
a(x 0
+∆
x)+
b.
The
chan
gein
yis
∆y
=y 1
−y 0
=a(
x 0+
∆x)+
b−
(ax 0
+b).
This
sim
plifi
esto
∆y
=a∆
x.
Ther
efor
e,∆
y
∆x
=a
Rat
eof
chan
ge(o
fa
line)
isco
nst
ant.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
6/
47
Slo
pe
ofa
Non
-Lin
ear
Fun
ctio
n
Lin
ear
funct
ions
are
use
fulbec
ause
ther
ear
esim
ple
.H
owev
er,not
man
yec
onom
icphen
omen
aar
elin
ear.
Con
sider
the
follo
win
gnon
-lin
ear
funct
ion
(aquad
ratic)
y=
x2.
Ifwe
star
tat
x=
1an
din
crea
sex
by1,
then
ych
ange
sby
3(i.e
.,4−
1).
Ifwe
star
tat
x=
2how
ever
,th
enin
crea
sing
xby
1ch
ange
sy
by5
(i.e
.,9−
4).
Thus
the
sam
ech
ange
inx
lead
sto
diff
eren
tch
ange
sin
y.
We
cannot
ther
efor
edefi
ne
agl
obal
not
ion
ofth
eslop
efo
rnon
-lin
ear
funct
ions.
How
ever
,it
ispos
sible
todefi
ne
anot
ion
ofth
eslop
ew
hic
his
valid
when
the
chan
gein
xis
“sm
all.”
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
7/
47
Tan
gent
Lin
e
Con
sider
the
curv
ey
=x
2.
Also
consider
the
line
y=
4x−
4.
This
line
just
touch
esth
ecu
rve
y=
x2
atth
epoi
nt
(x,y
)=
(2,4
).H
ow?
y=
22an
dy
=8−
4.
Such
alin
eis
calle
da
tange
nt
line.
The
tange
nt
line
has
the
prop
erty
that
it“l
ook
sth
esa
me
asth
efu
nct
ion
arou
nd
the
poi
nt
that
itju
stto
uch
es.”
The
tange
nt
line
show
sth
era
teof
chan
geat
apoi
nt
for
smal
lch
ange
s.
The
slop
eof
the
tange
nt
line
isca
lled
the
der
ivat
ive
atth
epoi
nt
x 0.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
8/
47
Dia
gram
ofth
eTan
gent
Lin
e
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
9/
47
The
Tan
gent
Lin
eas
aLim
itof
Sec
ant
Lin
es
An
alte
rnat
ive
way
ofth
inki
ng
abou
tth
eta
nge
nt
line
isth
atit
isa
limit
ofse
cant
lines
.
Ase
cant
line
isa
line
join
ing
any
two
poi
nts
onth
ecu
rve.
The
tange
nt
line
isth
elin
eto
whic
hth
ese
cant
lines
get
“clo
ser”
asth
epoi
nts
onth
elin
ege
tcl
oser
and
clos
er.
This
issh
own
inth
edia
gram
whic
hfo
llow
sfo
rth
ecu
rve
y=
x2
and
the
poi
nt
(x,y
)=
(2,4
).
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
10
/47
Dia
gram
with
Sec
ant
Lin
es
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
11
/47
Der
ivat
ive
The
der
ivat
ive
ofa
funct
ion
f(x
)at
the
poi
nt
x 0is
den
oted
asf′ (
x 0).
The
tota
ldiff
eren
tial
off(x
)at
x 0is
defi
ned
byth
efo
llow
ing.
dy
=f′ (
x 0)d
x
The
tota
ldiff
eren
tial
isa
way
ofunder
stan
din
gth
elo
calra
teof
chan
geof
the
funct
ion
f(x
)ar
ound
the
poi
nt
x 0.
That
is,it
isan
alge
brai
cway
ofden
otin
gth
eslop
eof
afu
nct
ion.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
12
/47
Eco
nom
icSig
nifica
nce
ofth
eD
eriv
ativ
e:Pro
duct
ion
Cos
ts
Imag
ine
that
yre
pres
ents
the
cost
sof
product
ion
in(in
$)an
dx
the
quan
tity
produce
dby
afirm
.
The
der
ivat
ive
ata
give
nx 0
tells
us
how
cost
sch
ange
inre
spon
seto
ach
ange
inquan
tity
,pr
ovid
edth
ech
ange
issm
all.
For
exam
ple
,if
we
know
that
the
der
ivat
ive
atx
=2
is4,
this
tells
us
that
ifth
equan
tity
produce
dch
ange
sby
asm
allam
ount
dx,th
enth
eim
pac
ton
cost
isgi
ven
appr
oxim
atel
yby
the
tota
ldiff
eren
tial
dy
=4d
x.
Eco
nom
ists
hav
ea
spec
ialfo
rth
eder
ivat
ive
ofth
eco
stfu
nct
ion:
itis
calle
dm
argi
nal
cost
.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
13
/47
Rem
arks
Sin
ceth
era
teof
chan
geal
ong
acu
rve
isch
angi
ng
const
antly,
the
der
ivat
ive
has
tobe
com
pute
dse
par
atel
yat
each
pos
sible
valu
eof
x.
The
der
ivat
ive
isth
us
alo
calphen
omen
on:
itte
llsus
som
ethin
gab
out
the
rate
ofch
ange
inth
enei
ghbor
hood
ofa
poi
nt,
but
itgi
ves
no
info
rmat
ion
abou
tth
era
teof
chan
gegl
obal
ly.
For
exam
ple
,th
ein
form
atio
nth
atth
eder
ivat
ive
ofy
=x
2(i.e
.,dy
=2x
dx)
atx
=2
is4
tells
us
that
the
rate
ofch
ange
is4
when
xis
“clo
se”
to2.
Itdoes
not
give
any
info
rmat
ion
abou
tth
era
teof
chan
geat
x=
10,
and
soon
.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
14
/47
Der
ivat
ive
asa
Fun
ctio
n
Not
ice
that
we
can
also
thin
kof
the
der
ivat
ive
itse
lfas
afu
nct
ion
ofx.
Giv
ena
funct
ion
y=
f(x
),th
eder
ivat
ive
funct
ion
sim
ply
asso
ciat
esto
ever
yx
the
slop
eof
the
tange
nt
line
atx.
Typ
ical
ly,w
hen
we
talk
abou
tth
eder
ivat
ive,
we
mea
nth
eder
ivat
ive
asa
funct
ion.
So
when
we
wan
tto
talk
abou
tth
eva
lue
ofth
eder
ivat
ive
ata
poi
nt
x,we
shal
lm
ention
itby
sayi
ng
“the
der
ivat
ive
atx
is..
.”N
otat
ional
ly,gi
ven
afu
nct
ion
y=
f(x
),we
shal
lden
ote
the
der
ivat
ive
funct
ion
as,
f′ (
x)
ordy
dx
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
15
/47
Rul
esfo
rCom
puting
Der
ivat
ives
Ther
ear
ea
num
ber
ofru
les
for
com
puting
der
ivat
ives
ofsp
ecifi
cfu
nct
ions:
1f(x
)=
a⇒
f′ (
x)
=0.
[con
stan
t]
2f(x
)=
mx
+b⇒
f′ (
x)
=m
.[lin
ear
funct
ion]
3f(x
)=
xn⇒
f′ (
x)
=nx
n−1
.[p
ower
funct
ion]
4g(x
)=
cf(x
)⇒
g′ (
x)
=cf
′ (x).
5h(x
)=
f(x
)+g(x
)⇒
h′ (
x)
=f′ (
x)+
g′ (
x).
6h(x
)=
f(x
)g(x
)⇒
h′ (
x)
=f′ (
x)g
(x)+
f(x
)g′ (
x).
[pro
duct
rule
]
7h(x
)=
f(x
)g(x
)⇒
h′ (
x)
=f′ (
x)g
(x)−
f(x
)g′ (
x)
[g(x
)]2
,pr
ovid
edg(x
)�=
0.
[quot
ient
rule
]
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
16
/47
ASim
ple
Exa
mpl
e
Con
sider
the
follo
win
g:y
=f(x
)=
αx−2
.W
eca
nal
sow
rite
this
asy
=f(x
)=
α x2.
Inth
efirs
tca
sewe
can
use
the
pow
erru
le.
Inth
ese
cond
case
,we
use
the
quot
ient
rule
.
Cas
e1:
y′ =
0·x
−2+
(−2)x
−2−1
·α=
−2α
/x
3<
0.
Cas
e2:
y′ =
0·x2
−α(2
x)
(x2)2
=−2
α/x
3<
0.
Inei
ther
case
we
get
the
sam
ere
sult.
As
xrise
s,y
lower
s.W
eal
sokn
ow,w
hen
x=
0,y
=0
and
when
x>
0,y
>0
and
when
x<
0,y
<0.
Inte
rms
ofw
hic
hru
leyo
uuse
,it
isbes
tto
use
the
sim
ple
ston
e(c
ase
1in
this
exam
ple
).
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
17
/47
Sim
ple
Eco
nom
icExa
mpl
es
Con
sider
the
product
ion
funct
ion,y
=f
(L).
Ass
um
eth
efo
llow
ing
funct
ional
form
:y
=12
L2−
L3.
By
(3),
(4)
and
(5),
dy dL
=24
L−
3L2
An
alte
rnat
ive
form
is,y
=6L
(3L
2+
4).
1W
eca
nuse
(6)
alon
gw
ith
(1),
(3),
(4)
and
(5).
dy dL
=6(3
L2+
4)+
6L(6
L)
=54
L2+
24
2A
lter
nat
ivel
y,not
eth
aty
=18
L3+
24L.
Now
apply
rule
s(3
),(4
)an
d(5
)to
obta
inth
esa
me
resu
ltas
abov
e.
Do
we
expec
tou
tput
togo
up
ordow
nas
we
use
mor
ela
bor
?Pro
bab
lyup.
This
assu
mption
rule
sou
tce
rtai
nfu
nct
ional
form
s.T
hat
is,we
thin
k,dy
dL
>0.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
18
/47
Ano
ther
Eco
nom
icExa
mpl
e
Con
sider
cost
and
reve
nue
funct
ions.
We
know
the
profi
tsof
afirm
are
give
nby
reve
nue
min
us
cost
s.
Π(y
)=
R(y
)−c(y
)
What
wou
ldwe
expec
tth
ere
venue
and
cost
funct
ions
tolo
oklik
e?Pro
bab
lyth
efo
llow
ing: R
′ (y)
>0
and
c′ (
y)
>0
As
we
produce
mor
e,we
mak
em
ore
reve
nue.
How
ever
,as
we
produce
mor
e,co
sts
rise
.
Cle
arly,th
ere
isa
trad
e-off
.At
som
epoi
nt
mar
ginal
reve
nue
(R′ (
y))
will
equal
mar
ginal
cost
(c′ (
y))
.T
hat
isth
epr
ofit
max
imiz
ing
conditio
n.
We
will
revi
sit
this
idea
repea
tedly.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
19
/47
Cha
inRul
e
Suppos
enow
that
yis
afu
nct
ion
ofu
and
uitse
lfis
afu
nct
ion
ofx.
Such
afu
nct
ion
isca
lled
aco
mpos
ite
funct
ion.
As
anex
ample
,co
nsider
y=
(x2+
1)2
.H
ere,
u=
x2+
1an
dy
=u
2.
The
chai
nru
leis
use
dto
dea
lw
ith
such
case
s.Suppos
ewe
hav
ey
=f(u
)w
her
eu
=g(x
),so
that
y=
f(g
(x))
.T
he
chai
nru
leis
give
nby
,dy
dx
=f′ (
u)g
′ (x)
This
says
that
the
der
ivat
ive
ofy
with
resp
ect
tox
iseq
ual
toth
eder
ivat
ive
ofy
with
resp
ect
tou
tim
esth
eder
ivat
ive
ofy
with
resp
ect
tox.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
20
/47
Pro
ofof
the
Cha
inRul
e
Let
y=
f(g
(x))
.N
owle
tu
=g(x
)so
that
y=
f(u
).U
sing
the
tota
ldiff
eren
tial
,we
hav
e
dy
=f′ (
u)d
u(1
)
Sin
ceu
=g(x
),we
can
use
the
tota
ldiff
eren
tial
her
eto
get
du
=g′ (
x)d
x(2
)
Use
(2)
tosu
bst
itute
for
du
in(1
):
dy
=f′ (
u)d
u=
f′ (
u)[ g
′ (x)d
x] =
f′ (
u)g
′ (x)d
x(3
)
This
implie
sth
atdy
/dx
=f′ (
u)g
′ (x).
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
21
/47
Exa
mpl
esus
ing
the
Cha
inRul
e
Inth
eab
ove
case
,we
hav
eu
=g(x
)=
x2+
1an
df(u
)=
u2.
Then
,f′ (
u)
=2u
and
g′ (
x)
=2x
,an
dby
the
chai
nru
le,we
hav
e,
dy
dx
=2u
×2x
=4x
(x2+
1)
Anot
her
exam
ple
:su
ppos
ey
=(x
1/2+
3x2)5
.T
hen
(show
for
yours
elf)
,
dy
dx
=5(x
1/2+
3x2)4
×[ (1
/2)x
−1/2+
6x]
Not
ice
the
diff
eren
ce.
Inth
efirs
tca
se,we
can
multip
lyou
ty
=(x
2+
1)2
=x
4+
2x2+
1an
duse
the
usu
alru
les.
Inth
ese
cond
case
,we
can
do
that
,but
itis
not
ago
od
idea
asit
bec
omes
very
mes
sy.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
22
/47
Eco
nom
icExa
mpl
e
Aga
in,co
nsider
the
tota
lre
venue
ofa
firm
,R
.It
also
dep
ends
onth
ete
chnol
ogy
(i.e
.,th
esh
ape
ofth
epr
oduct
ion
funct
ion)
and
labor
input.
That
is,
R=
k(y
);y
=f
(L)
⇒R
=k
(f(L
))
The
der
ivat
ive
ofth
isfu
nct
ion
is:
dR dl
=dR dy
df
dL
=k′ (
y)f
′ (L)
Eco
nom
icm
eanin
g:dR dy
ism
argi
nal
reve
nue
(as
abov
e).
df
dL
isth
em
argi
nal
product
ofla
bor
.
We
concl
ude
that
the
MRPL
iseq
ual
toth
eM
Rtim
esth
eM
PL.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
23
/47
Der
ivat
ive
ofth
eIn
vers
eFun
ctio
n
Con
sider
the
funct
ions
y=
f(x
)=
x2
and
x=
g(y
)=
√ y.
We
can
thin
kof
thes
etw
ofu
nct
ions
asin
vers
es.
Ifwe
take
xas
the
input,
apply
fto
itan
dth
enpas
sth
isou
tput
thro
ugh
the
funct
ion
g,we
get
bac
kx.
The
der
ivat
ives
ofin
vers
efu
nct
ions
are
rela
ted
toea
chot
her
.
Ify
=f(x
)an
dx
=g(y
)ar
ein
vers
efu
nct
ions,
we
hav
e,
g′ (
y)
=1 dy
dx
=1
f′ (
x)
Inth
eab
ove
case
,f′ (
x)
=2x
.T
her
efor
e,
g′ (
y)
=1
f′ (
x)
=1 2x
=1
2√ y
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
24
/47
Exp
onen
tial
and
Log
arithm
icFun
ctio
nsI
Con
sider
the
follo
win
gfu
nct
ion: y
=b
x
yis
the
dep
enden
tva
riab
le,x
isin
dep
enden
t,an
db
isth
e(fi
xed)
bas
eof
the
expon
ent.
We
can
chan
geth
ebas
e.A
com
mon
bas
eis
e=
2.71
828.
...
This
isca
lled
the
nat
ura
lbas
e.
Why
do
we
care
abou
tth
is?
Bec
ause
ex
isit’s
own
der
ivat
ive
itis
easy
towor
kw
ith.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
25
/47
Exp
onen
tial
and
Log
arithm
icFun
ctio
nsII
Now
consider
the
rela
tion
ship
bet
wee
n8
and
64.
Cle
arly,82
=64
.
The
expon
ent
2is
the
loga
rith
mof
64to
the
bas
e8.
That
is,
log8(6
4)
=2.
Lik
ewise,
ifwe
hav
ey
=b
xth
enx
=lo
gb(y
).W
eal
sonot
eth
atif
y=
ex
then
x=
log e
(y)≡
ln(y
).A
gain
,th
ere
isa
use
fulre
sult,as
dln
(x)/
dx
=1
/x.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
26
/47
Rul
esfo
rExp
onen
tial
and
Log
arithm
icFun
ctio
ns
Log
arithm
ican
dex
pon
ential
funct
ions
are
very
com
mon
inec
onom
ics.
The
bas
icru
les
for
diff
eren
tiat
ing
expon
ential
and
loga
rith
mic
funct
ions
are
the
follo
win
g.
1f(x
)=
lnx⇒
f′ (
x)
=1
/x
2f(x
)=
lnu(x
)⇒
f′ (
x)
=u′ (
x)/
u(x
)3
f(x
)=
ex⇒
f′ (
x)
=ex
4f(x
)=
eu(x
)⇒
f′ (
x)
=eu(x
) u′ (
x)
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
27
/47
Exa
mpl
esof
Fun
ctio
ns
Som
est
raig
htf
orwar
dex
ample
sof
expon
ential
and
loga
rith
mic
funct
ions
are:
f(x
)=
ln(x
α)⇒
f′ (
x)
=αx
(1/x
α)
=α
/x
f(x
)=
ln(x
2+
3x+
1)⇒
f′ (
x)
=(2
x+
3)/
(x2+
3x+
1)
f(x
)=
e5x⇒
f′ (
x)
=5e
5x
f(x
)=
5ex
2⇒
f′ (
x)
=10
xex
2
You
shou
ldbe
com
fort
able
with
allof
thes
e.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
28
/47
Log
arithm
icD
iffer
entiat
ion
Suppos
ewe
wan
tto
diff
eren
tiat
ey
=( x
2) x2
.W
epr
oce
edas
follo
ws:
1Tak
elo
gson
bot
hsides
: lny
=x
2ln
x2
=2x
2ln
x
2D
iffer
entiat
ebot
hsides
with
resp
ect
tox.
On
the
left
-hand
side,
byth
ech
ain
rule
,we
get
(1/y)(
dy
/dx).
On
the
right-
hand
side,
byth
epr
oduct
rule
,we
hav
e4x
lnx
+2x
.A
fter
re-a
rran
ging,
we
obta
in
dy
dx
=y(4
xln
x+
2x)
=(x
2)x
2[4
xln
x+
2x]
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
29
/47
Eco
nom
icExa
mpl
e:Ela
stic
ity
ofD
eman
d
Suppos
ewe
hav
ea
dem
and
funct
ion
q=
D(p
).W
efe
elhap
pyas
sum
ing
D′ (
p)
<0;
that
is,th
edem
and
curv
eslop
esdow
nwar
ds.
How
ever
,how
does
quan
tity
chan
geas
pric
ech
ange
sal
ong
the
curv
e?
The
elas
tici
tyof
dem
and
isdefi
ned
as,
ε pq≡
dq q dp p
=p q
dq
dp
Itis
inte
rest
ing
tonot
eth
atwe
can
repr
esen
tth
eel
astici
tyof
dem
and
ina
mor
eco
nci
sem
anner
as ε pq
=d(ln[q
])d(ln[p
])
Thes
eex
pres
sion
sar
eeq
uiv
alen
t.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
30
/47
Eco
nom
icExa
mpl
e:Ela
stic
ity
ofD
eman
d
To
see
this,not
eth
atby
using
the
chai
nru
le,we
hav
e
dln
[q]
dln
[p]
=d
ln[q
]q
×dq
dln
[p]
=d
ln[q
]dq
×dq
dp×
dp
dln
[p]
Sin
ced
ln[q
]/dq
=1
/q
and
dp
/d
ln[p
]=p
(by
the
inve
rse
rule
),th
ere
sult
follo
ws.
The
alte
rnat
ive
form
ula
for
the
elas
tici
tyca
nbe
use
fulin
som
eci
rcum
stan
ces.
Suppos
ewe
hav
eth
edem
and
funct
ion
q=
pa
wher
ea
<0
isa
const
ant.
(Why
must
we
hav
ea
<0?
)
We
can
com
pute
the
elas
tici
tyve
ryea
sily
asfo
llow
s.Sin
ceq
=p
a,
we
hav
eln
q=
aln
p.
Thus
dln
[q]/
dln
[p]=
ash
owin
gth
atth
eel
astici
tyof
dem
and
isa
const
ant
ever
ywher
e.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
31
/47
Hig
her
Ord
erD
eriv
ativ
es
We
call
f′ (
x)
the
firs
tder
ivat
ive
ofth
efu
nct
ion
f.
Sin
ceth
eder
ivat
ive
itse
lfis
afu
nct
ion,we
can
take
it’s
der
ivat
ive
–th
isis
calle
dth
ese
cond
der
ivat
ive
and
den
oted
d2y
dx
2or
f′′ (
x).
For
mal
ly,
d2y
dx
2=
d(d
ydx)
dx
Suppos
ef(x
)=
x5.
Then
,f′ (
x)
=5x
4an
df′′ (
x)
=20
x3.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
32
/47
Hig
her
Ord
erD
eriv
ativ
es
Sin
ceth
ese
cond
der
ivat
ive
isal
soa
funct
ion,we
can
also
take
its
der
ivat
ive.
This
isca
lled
the
third
der
ivat
ive
and
den
oted
f′′′
(x)
toin
dic
ate
that
this
funct
ion
isfo
und
byth
ree
succ
essive
oper
atio
ns
ofdiff
eren
tiat
ion,
star
ting
with
the
funct
ion
f.
One
can
continue
this
proce
ssbut
we
will
typic
ally
not
gobey
ond
the
seco
nd
der
ivat
ive.
Exa
mple
continued
:Suppos
ef(x
)=
x5.
Then
,f′ (
x)
=5x
4,f
′′ (x)
=20
x3,f
′′′(x
)=
60x
2.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
33
/47
Con
cave
and
Con
vex
Fun
ctio
ns
Afu
nct
ion
f(x
)is
calle
dco
nca
veif
f′′ (
x)≤
0at
allpoi
nts
ofits
dom
ain.
An
econ
omic
exam
ple
ofa
conca
vefu
nct
ion
isth
epr
oduct
ion
funct
ion
(usu
ally
).
Afu
nct
ion
f(x
)is
calle
dco
nve
xif
f′′ (
x)≥
0at
allpoi
nts
ofits
dom
ain.
Ifth
ein
equal
itie
sar
est
rict
,th
enth
efu
nct
ion
isca
lled
strict
lyco
nca
veor
strict
lyco
nve
x.
Not
e:Con
cavi
tyan
dco
nve
xity
do
not
say
anyt
hin
gab
out
the
firs
tder
ivat
ive
(incr
easing
ordec
reas
ing
funct
ions)
.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
34
/47
Exa
mpl
esof
Con
cavi
tyan
dCon
vexi
ty
The
funct
ion
f(x
)=
x2
isco
nve
xon
the
dom
ain
x≥
0bec
ause
f′′ (
x)
=2
>0
for
allx≥
0.
The
funct
ion
g(x
)=
lnx
isco
nca
veon
the
dom
ain
x>
0bec
ause
g′′ (
x)
=−1
/x
2<
0fo
ral
lx
>0.
Afu
nct
ion
may
be
eith
erco
nca
venor
conve
xon
agi
ven
dom
ain.
For
exam
ple
,co
nsider
f(x
)=
−(2
/3)x
3+
10x
2+
5xon
the
dom
ain
x≥
0.
Inth
isca
se,f′′ (
x)
=−4
x+
20;so
,
1f′′ (
x)≥
0if
x≤
5(c
onve
x)2
f′′ (
x)
<0
ifx
>5
(con
cave
)
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
35
/47
Eco
nom
icExa
mpl
esof
Con
cavi
tyan
dCon
vexi
ty
Con
sider
the
follo
win
gpr
oduct
ion
funct
ion:
y=
lα.
We
know
,y′ =
αlα−1
>0.
But
y′′
=(α
−1)α
lα−2
.T
his
ison
lypos
itiv
ew
hen
α<
1.T
hat
isco
nsist
ent
with
dim
inishin
gm
argi
nal
retu
rns
tola
bor
.
Con
sider
the
reve
nue
and
cost
sfu
nct
ions
from
bef
ore.
Now
we
mig
ht
suppos
eth
efo
llow
ing.
R′ (
y)
>0;
R′′(y
)<
0;c′ (
y)
>0;
c(y
)′′>
0
As
we
produce
mor
e,we
mak
em
ore
reve
nue,
but
aspr
oduct
ion
rise
sev
erhig
her
,th
ead
ditio
nal
unit
gener
ate
less
reve
nue
than
the
last
.
How
ever
,as
we
produce
mor
e,co
sts
rise
,an
das
we
continue
topr
oduce
,ea
chex
tra
unit
ism
ore
cost
lyth
atth
ela
st.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
36
/47
Intr
odu
ctio
nto
Unc
onst
rain
edO
ptim
izat
ion
We
can
also
rela
tese
cond-o
rder
conditio
ns
tonec
essa
ryan
dsu
ffici
ent
conditio
ns
for
max
imiz
atio
npr
oble
ms.
That
is,to
find
the
max
imum
,it
isnot
enou
ghto
“set
the
firs
tder
ivat
ive
toze
ro”.
We
nee
dto
chec
kse
cond
order
conditio
ns
explic
itly.
Two
fam
iliar
exam
ple
s:
1Suppos
eth
ata
mon
opol
ist
face
sa
dem
and
curv
ep
=10
0−
qan
dhas
product
ion
cost
sgi
ven
byc(q
)=
q2.
Ifth
em
onop
olist
isin
tere
sted
inm
axim
izin
gpr
ofits
,how
much
shou
ldhe
produce
?
2Suppos
eth
ata
firm
has
fixe
dco
stof
100
and
variab
leco
sts
give
nby
v(q
)=
q2.
Ifth
efirm
wan
tsto
min
imiz
eav
erag
eco
sts
ofpr
oduct
ion,
how
much
shou
ldit
produce
?
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
37
/47
Nec
essa
ryCon
dition
sfo
rO
ptim
izat
ion
Ifa
diff
eren
tiab
lefu
nct
ion
f(x
)re
aches
its
max
imum
orm
inim
um
atx∗
then
f′ (
x∗ )
=0.
Rea
son:
Con
sider
the
tota
ldiff
eren
tial
dy
=f′ (
x)d
x
Ifth
efu
nct
ion
reac
hes
am
axim
um
orm
inim
um
atx∗
then
itm
ust
be
impos
sible
toin
crea
seor
dec
reas
eth
eva
lue
ofth
efu
nct
ion
bysm
all
chan
ges
inx.
How
ever
,if
f′ (
x∗ )
�=0,
then
itis
alway
spos
sible
tom
ake
dy
larg
eror
smal
lerby
mak
ing
(sm
all)
appr
opriat
ech
ange
sin
x.
Ther
efor
e,we
must
hav
ef′ (
x∗ )
=0
ata
max
imum
orm
inim
um
.
This
isca
lled
anec
essa
ryco
nditio
nbec
ause
itca
nnot
guar
ante
eth
atx
isin
dee
da
max
imum
orm
inim
um
.It
isen
tire
lypos
sible
that
f′ (
x∗ )
=0
but
x∗
isnei
ther
am
axim
um
nor
am
inim
um
.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
38
/47
Glo
balan
dLoca
lM
axim
a
Apoi
nt
x∗
isca
lled
agl
obal
max
imum
ofth
efu
nct
ion
f(x
)if
f(x
∗ )≥
f(x
)fo
ral
lx
inth
edom
ain
off.
Apoi
nt
x∗
isca
lled
alo
calm
axim
um
ofth
efu
nct
ion
f(x
)if
ther
eis
a“s
mal
lin
terv
al”
cente
red
atx∗
such
that
f(x
∗ )≥
f(x
)fo
ral
lx
inth
esm
allin
terv
al.
Not
eth
enec
essa
ryco
nditio
non
lyid
entifies
alo
calm
axim
um
orm
inim
um
,but
not
agl
obal
max
imum
orm
inim
um
.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
39
/47
Sec
ond
Ord
eror
Suffi
cien
tCon
dition
sfo
rO
ptim
izat
ion
We
also
hav
eth
efo
llow
ing
conditio
ns:
1If
f′ (
x∗ )
=0
and
f′′ (
x∗ )
<0,
then
x∗
isa
loca
lm
axim
um
off(x
).2
Iff′ (
x∗ )
=0
and
f′ (
x∗ )
>0,
then
x∗
isa
loca
lm
inim
um
off(x
).
Not
eth
atit
ispos
sible
that
apoi
nt
may
be
am
axim
um
orm
inim
um
and
yet
not
satisf
yth
esu
ffici
ent
conditio
nfo
rm
axim
izat
ion
orm
inim
izat
ion.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
40
/47
Eco
nom
icExa
mpl
e
We
now
retu
rnto
the
econ
omic
exam
ple
sab
ove.
Suppos
eth
em
onop
olist’s
profi
tfu
nct
ion
isgi
ven
by,
Π(y
)=
py−
c(y
)=
(100
−y)y
−y
2
The
nec
essa
ryco
nditio
nim
plie
sth
atΠ
′ (y)
=10
0−
2y−
2y=
0or
y=
25.
The
firm
’sav
erag
eco
stis
give
nby
AC
(y)
=10
0/y
+y.
The
nec
essa
ryco
nditio
nAC
′ (y)
=0
give
s−1
00/y
2+
1=
0or
y2−
100
=0.
Ther
efor
e,y
=10
ory
=−1
0.Sin
ceq
=−1
0is
not
pos
sible
,we
must
hav
ey
=10
.
To
chec
kth
atth
ese
are
indee
dth
em
axim
um
and
min
imum
,we
can
eith
erplo
tth
etw
ofu
nct
ions,
orch
eck
the
seco
nd
order
conditio
ns
for
optim
izat
ion.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
41
/47
Eco
nom
icExa
mpl
eCon
tinu
ed...
Inth
em
onop
olist’s
prob
lem
,not
eth
atth
ese
cond
der
ivat
ive
Π′′ (
y)
=−4
for
ally,(in
par
ticu
lar,
fory
=25
)an
dhen
cey
=25
isa
loca
lm
axim
um
.In
the
firm
’spr
oble
m,AC
′′ (y)
=20
0/q
3;
subst
ituting
y=
10gi
ves
AC
′′ (10
)=
200
/10
00=
1/5
>0.
Hen
ce,
y=
10is
alo
calm
inim
um
.
We
do
not
know
whet
her
thes
ear
elo
calor
glob
alm
inim
a.T
he
impor
tance
ofco
nca
vean
dco
nve
xfu
nct
ions
stem
from
the
fact
that
the
loca
lm
axim
aof
conca
vefu
nct
ions
are
also
glob
alm
axim
a;th
elo
calm
inim
aof
conve
xfu
nct
ions
are
also
glob
alm
inim
a.
Inth
em
onop
olist’s
prob
lem
,th
epr
ofit
funct
ion
isco
nca
ve;hen
cey
=25
isa
glob
alm
axim
um
.In
the
firm
’spr
oble
m,th
eAC
funct
ion
isco
nve
xon
the
dom
ain
qy
>0
and
hen
ceit
isa
glob
alm
inim
um
.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
42
/47
Ano
ther
Exa
mpl
e:Sup
ply,
Dem
and,
and
Tax
es
Suppos
ewe
hav
ea
mar
ket
with
dem
and
funct
ion
qd
=a 0
−a 1
pan
dsu
pply
funct
ion
qs=
b0+
b1p.
The
gove
rnm
ent
wan
tsto
impos
ean
exci
seta
xt
onth
ism
arke
t.W
hat
rate
shou
ldth
ego
vern
men
tch
oos
eif
itwan
tsto
max
imiz
eta
xre
venue?
Withou
tta
xation
,m
arke
teq
uili
briu
mis
det
erm
ined
byth
eco
nditio
nq
d=
qs.
Thus,
a 0−
a 1p
=b
0+
b1p,w
hic
hgi
ves,
p∗
=a 0
−b
0
a 1+
b1
and
q∗
=a 0
b1+
a 1b
0
a 1+
b1
When
ata
xt
isim
pos
ed,th
eeq
uili
briu
mco
nditio
ns
require
that
(i)
qd
=q
san
d(ii)
pb
=p
s+
t.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
43
/47
Sup
ply
and
Dem
and
Con
tinu
ed...
Equili
briu
mnow
requires
a 0−
a 1p
b=
b0+
b1(p
b−
t).
This
give
s
p∗ b
=a 0
−b
0+
b1t
a 1+
b1
,p∗ s
=a 0
−b
0−
a 1t
a 1+
b1
,q∗ t
=a 0
b1+
a 1b
0−
a 1b
1t
a 1+
b1
The
tax
reve
nue
asa
funct
ion
ofth
eta
xra
tet
is:
T(t
)=
tq∗ t
=t
[ a 0b
1+
a 1b
0−
a 1b
1t
a 1+
b1
]
We
wan
tto
find
tto
max
imiz
eT
(t).
First
,we
iden
tify
tsu
chth
atdT
/dt
=0.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
44
/47
Sup
ply
and
Dem
and
Con
tinu
ed...
Diff
eren
tiat
ing
T(t
)an
deq
uat
ing
toze
rogi
ves
dT dt
=[ a 0
b1+
a 1b
0−
2a1b
1t
a 1+
b1
] =0
This
give
sus
auniq
ue
valu
eof
t:
t∗=
a 0b
1+
a 1b
0
2a1b
1
We
now
hav
eto
chec
kw
het
her
t∗co
rres
pon
ds
toa
max
imum
ornot
.W
ehav
e
T′′ (
t)=
−2a
1b
1
a 1+
b1
Thus,
the
seco
nd
der
ivat
ive
isa
const
ant,
and
since
a 1>
0,b
1>
0it
isal
way
sneg
ativ
e.T
his
show
sth
atth
eta
xre
venue
funct
ion
T(t
)is
conca
ve,an
dhen
cet∗
isnot
only
alo
calbut
also
agl
obal
max
imum
.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
45
/47
Rem
ark
onth
eN
eces
sary
Con
dition
for
Opt
imiz
atio
n
The
conditio
nf′ (
x)
=0
ata
max
imum
orm
inim
um
isva
lidon
lyif
xis
inth
e“i
nte
rior
”of
the
dom
ain
ofth
efu
nct
ion.
Bas
ical
ly,if
we
look
atth
ear
gum
ent
give
nfo
rsh
owin
gth
atif
x∗
isa
max
imum
orm
inim
um
,th
enf′ (
x)
=0,
itsh
ould
be
not
iced
that
the
argu
men
tdep
ended
onth
eab
ility
tom
ake
smal
lch
ange
sin
x.
How
ever
,at
a“b
oundar
ypoi
nt”
we
cannot
mak
ece
rtai
nch
ange
s.For
inst
ance
,if
the
funct
ion
isdefi
ned
for
allx
inth
ein
terv
al[a
,b],
then
ata,
we
can
only
incr
ease
x;sim
ilarly
atb,we
can
only
dec
reas
ex.
Thus,
the
argu
men
tth
atf′ (
x)
=0
isnot
true
ifth
em
axim
um
occ
urs
ata
orb.
We
can
modify
the
nec
essa
ryco
nditio
nf′ (
x)
=0
toac
count
for
thes
e“b
oundar
ypoi
nts
”but
we
will
dea
lw
ith
such
situ
atio
ns
when
we
discu
ssco
nst
rain
edop
tim
izat
ion.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
46
/47
Rou
ndup
You
shou
ldnow
be
able
todo
the
follo
win
g:
1D
iffer
entiat
efu
nct
ions
(inc.
log
and
exp.)
.
2Chec
kfo
rco
nca
vity
/con
vexi
ty.
3A
pply
thes
eid
eas
tom
icro
econ
omic
prob
lem
s.
4U
nder
stan
dth
eim
por
tance
ofnec
essa
ryve
rsus
suffi
cien
tco
nditio
ns.
Dudle
yCooke
(Trinity
Colleg
eD
ublin)
Sin
gle
Var
iable
Calc
ulu
sand
Applica
tions
47
/47