You are on page 1of 9

0.

1 Practical Guide - Integrals with Parameters


"Functions of several variables" ! )(r, j, .) , q(n, , n, t, c, ,) .... they look so far away from the real life !
Actually, we have physical quantities, scalar elds or vector elds depending on parameters, more parameters
than we can handle.
Now about integrals with parameters
b(x;y;z)
_
a(x;y;z)
)(r, j, ., t)dt
all we need to do is nd an antiderivative G(r, j, ., t) , such that
0G(r, j, ., t)
0t
= )(r, j, ., t)
and computing the integral is easy
b(x;y;z)
_
a(x;y;z)
)(r, j, ., t)dt = G(r, j, ., t)j
t=b(x;y;z)
t=a(x;y;z)
= G(r, j, ., /(r, j, .)) G(r, j, ., a(r, j, .))
. .
A
When computing double or triple integrals, we reduce to successive simple integrals, which actually have "pa-
rameters" , but not so hard to compute. For instance
__
D
rdrdj =

2
_

2
_
_
4 cos t
_
0
r cos t rdr
_
_
dt =

2
_

2
_
_
4 cos t
_
0
cos t r
2
dr
_
_
dt = ...
Anyone can handle this.
Because it is a very simple case !
Well, "nd an antiderivative" this is precisely the problem. It may not be possible in any case.
We dont actually need that antiderivative "G", only the result of the dierence , not both terms ! Any
method should be ne.
Start with very simple examples.
Examples.
i) ) : [a, /] !R a continuous function and c 2 [a, /]
1(r) =
x
_
c
)(t)dt
Then 1 is derivable and 1
0
(r) = )(r) . This is well known from highschool analysis.
ii) Next, a bit more complicated / : [a, /] !R a derivable function.
G(r) =
b(x)
_
c
)(t)dt
We easily see that G(r) = 1(/(r)) , G is derivable and G
0
(r) = [1(/(r)]
0
= 1
0
(/(r)) /
0
(r) = )(r)/
0
(r)
iii) And next for a, / : [a, /] !R derivable functions we get
H(r) =
b(x)
_
a(x)
)(t)dt
1
In this case we may write
H(r) =
b(x)
_
a(x)
)(t)dt =
c
_
a(x)
)(t)dt +
b(x)
_
c
)(t)dt =
b(x)
_
c
)(t)dt
a(x)
_
c
)(t)dt = 1((/(r)) 1(a(r))
and we get H
0
(r) = [1((/(r)) 1(a(r))]
0
= 1
0
(/(r)) /
0
(r) 1
0
(a(r)) a
0
(r) = )(r)/
0
(r) )(r)a
0
(r)
iv) Finally for a function ) : 1 R
2
!R of class C
1
consider the integral
1(r) =
d
_
c
)(r, t)dt
Under "nice" conditions we may take the derivative
1
0
(r) =
0
0r
_
_
d
_
c
)(r, t)dt
_
_
=
d
_
c
0
0r
[)(r, t)] dt
Comment.
It seems to be all "wrong" ! We were supposed to integrate, nd antiderivative. All we did by now was taking
derivatives !
Well, the idea is : if we cannot nd an antiderivative as
0G(r, t)
0t
= )(r, t)
and directly compute 1(r) , maybe we may nd an antiderivative with respect to "t"
0
0r
[)(r, t)] =
0
0t
(?)
and then another antiderivative for 1
0
(r) so that
1(r) =
_
1
0
(r)dr + C
Its a very "strong" maybe. But it works in many cases, at least for ... school problems !
Anyway, better a hard task, instead of an impossible one.
v) When the previous cases are combined
H(r) =
b(x)
_
a(x)
)(r, t)dt
we also "combine" the results
H
0
(r) =
0
0r
_
_
_
b(x)
_
a(x)
)(r, t)dt
_
_
_ =
b(x)
_
a(x)
0
0r
[)(r, t)] dt + )(r, /(r))/
0
(r) )(r, a(r))a
0
(r)
In the following we will consider only the case
1(r) =
d
_
c
)(r, t)dt
where the integral could be denite or improper.
2
"Basic" algorithm"
Step I. take the partial derivative
@
@x
[)(r, t)]
Step II. integrate with respect to "t"
d
_
c
0
0r
[)(r, t)] dt = 1
0
(r)
Step III. integrate with respect to "r"
1(r) =
_
1
0
(r)dr + C
Step IV. To nd the value of C we need to nd some particular r = a for which we may compute the integral
1(a) =
d
_
c
)(a, t)dt
Examples.
1) Compute 1(r) dened by
1(r) =
+1
_
0
arctg(rt)
t(1 + t
2
)
dt , r 2 (1, 1)
Solution.
Step I
1
0
(r) =
0
0r
_
_
+1
_
0
arctg(rt)
t(1 + t
2
)
dt
_
_
=
+1
_
0
0
0r
_
arctg(rt)
t(1 + t
2
)
_
dt =
=
+1
_
0
1
t(1 + t
2
)
0
0r
(arctg(rt)) dt =
+1
_
0
1
t(1 + t
2
)
1
1 + (rt)
2
0
0r
(rt)dt =
=
+1
_
0
1
t(1 + t
2
)
1
1 + r
2
t
2
tdt =
+1
_
0
1
(1 + t
2
)(1 + r
2
t
2
)
dt =
Step II decompose into "simple" fractions
=
+1
_
0
1
1 r
2
_
1
1 + t
2

r
2
1 + r
2
t
2
_
dt =
1
1 r
2
_
_
+1
_
0
1
1 + t
2
dt
+1
_
0
r
2
1 + r
2
t
2
dt
_
_
=
=
1
1 r
2
_
arctg tj
t!1
t=0

1
2
arctg(r
2
t)

t!1
t=0
_
=
1
1 r
2
_
lim
t!+1
arctg t arctg 0 lim
t!+1
1
2
arctg r
2
t arctg 0
_
=
=
1
1 r
2
_

2


4
_
=

4
1
1 r
2
Step III
1
0
(r) =

4
1
1 r
2
) 1(r) =
_
1
0
(r)dr =
_

4
1
1 r
2
dr =
=

4
_
1
2
_
1
1 r

1
1 + r
_
dr =

8
(lnj1 rj lnj1 + rj) + C
Step IV It seems the only possible choice r = 0 that allows computing the original intergal
1(0) =
+1
_
0
arctg(0 t)
t(1 + t
2
)
dt =
+1
_
0
0dt = 0
3
Consequently we have
0 = 1(0) =

8
(lnj1 0j lnj1 + 0j) + C = C ) C = 0
and we nally get
1(r) =
+1
_
0
arctg(rt)
t(1 + t
2
)
dt =

8
(lnj1 rj lnj1 + rj) =

8
ln(1 r
2
)

2) Compute the integral


=2
_
0
ln(1 + rcos t)
cos t
dt , r 2 (1, 1)
Solution.
This is an improper integral.
Step I Let
1(r) =
=2
_
0
ln(1 + rcos t)
cos t
dt
1
0
(r) =
0
0r
_
_
_
=2
_
0
ln(1 + rcos t)
cos t
dt
_
_
_ =
=2
_
0
0
0r
_
ln(1 + rcos t)
cos t
_
dt =
=2
_
0
1
cos t
0
0r
ln(1 + rcos t)dt =
=
=2
_
0
1
cos t
1
1 + rcos t
cos tdt =
=2
_
0
1
1 + rcos t
dt =
Step II Use the change of variable
tan
t
2
= j , t = 2 arctg j , dt = 2
1
1 + j
2
dj , t = 0 )j = 0 , t =

2
)j = 1 , cos t =
1 j
2
1 + j
2
=
1
_
0
1
1 + r
1y
2
1+y
2
2
1
1 + j
2
dj = 2
1
_
0
1
1 + j
2
+ r rj
2
dj = 2
1
_
0
1
1 + r + (1 r)j
2
dj =
=
2
1 + r
1
_
0
1
1 +
1x
1+x
j
2
dj =
2
1 + r
_
_
1 + r
1 r
arctg
_
1 r
1 + r
j
_

y=1
y=0
=
=
2
_
(1 + r)(1 r)
(arctg 0 arctg 1) =

2
_
(1 + r)(1 r)
=

2
p
1 r
2
Step III
1
0
(r) =

2
p
1 r
2
) 1(r) =
_

2
p
1 r
2
dr =

2
_
1
p
1 r
2
dr =

2
arcsinr + C
Step IV It seems the only possible choice r = 0 that allows computing the original intergal
1(0) =
=2
_
0
ln(1 + 0 cos t)
cos t
dt =
=2
_
0
0dt = 0
Consequently we have
0 = 1(0) =

2
arcsin0 + C = C ) C = 0
4
and nally we get
1(r) =
=2
_
0
ln(1 + rcos t)
cos t
dt =

2
arcsinr

2) Consider the integral (also improper)

_
=2
ln(1 + rcos t)
cos t
dt
make the change of variable n = t we get

_
=2
ln(1 + rcos t)
cos t
dt =
0
_
=2
ln(1 + rcos( n))
cos( n)
(dn) =
=2
_
0
ln(1 rcos n)
cos n
dn =
=
=2
_
0
ln(1 rcos n)
cos n
dn = 1(r)
Consequently it makes sense to consider the improper integral

_
0
ln(1 + rcos t)
cos t
dt
because we may split as

_
0
ln(1 + rcos t)
cos t
dt =
=2
_
0
ln(1 + rcos t)
cos t
dt +

_
=2
ln(1 + rcos t)
cos t
dt =
= 1(r) 1(r) =

2
arcsinr

2
arcsin(r)r = arcsinr
and we nally get

_
0
ln(1 + rcos t)
cos t
dt = arcsinr

3) Compute the improper integral


=2
_
0
arctg(rtg t)
tg t
dt , r 2 (1, 1)
Solution.
Step I Let
1(r) =
=2
_
0
arctg(rtg t)
tg t
dt
1
0
(r) =
0
0r
_
_
_
=2
_
0
arctg(rtg t)
tg t
dt
_
_
_ =
=2
_
0
0
0r
_
arctg(rtg t)
tg t
_
dt =
=2
_
0
1
tg t
0
0r
(arctg(rtg t)) dt =
5
=
=2
_
0
1
tg t
1
1 + (rtg t)
2
tantdt =
=2
_
0
1
1 + r
2
tg
2
t
dt =
Step II Make the change of variable tg t = n , t = arctg n , dt =
1
1+u
2
dn , t = 0 )n = 0 , t !

2
)n !+1
=
+1
_
0
1
1 + r
2
n
2
1
1 + n
2
dn =
+1
_
0
1
1 r
2
_
r
2
1 + r
2
n
2
+
1
1 + n
2
_
dn =
=
1
1 r
2
_
_
+1
_
0
r
2
1 + r
2
n
2
dn
+1
_
0
1
1 + n
2
dn
_
_
=
1
1 r
2
_
1
2
arctg(r
2
n
2

u!+1
u=0
arctg nj
u!+1
u=0
_
=
=
1
1 r
2
_
1
2
lim
u!+1
arctg(r
2
n
2
) arctg 0
_
lim
u!+1
arctg n arctg 0
__
=
=
1
1 r
2
_

4


2
_
=
3
4(1 r
2
)
Stept III
1
0
(r) =
3
4(1 r
2
)
) 1(r) =
_
3
4(1 r
2
)
dr =
3
4
_
1
1 r
2
dr =
=
3
4
_
1
2
_
1
1 r

1
1 + r
_
dr =
3
8
(ln(1 r) ln(1 + r)) + C
Step IV It seems the only possible choice r = 0 that allows computing the original intergal
1(0) =
=2
_
0
arctg(0 tg t)
tg t
dt =
=2
_
0
0dt = 0
Consequently we have
0 = 1(0) =
3
8
(ln(1 0) ln(1 + 0)) + C = C ) C = 0
and we get
1(r) =
=2
_
0
arctg(rtg t)
tg t
dt =
3
8
ln(1 r
2
)

4) Find the derivative of the function


1(r) =
cos x
_
sin x
ln(2 + sinrt)dt
Solution
1
0
(r) =
_
_
cos x
_
sin x
ln(2 + sinrt)dt
_
_
0
=
0
0r
cos x
_
sin x
ln(2+sinrt)dt +ln(2+sin(rcos r))(sinr) ln(2+sin(rsinr)) cos r =
=
cos x
_
sin x
0
0r
ln(2 + sinrt)dt ln(2 + sin(rcos r)) sinr ln(2 + sin(rsinr)) cos r =
=
cos x
_
sin x
cos(rt) t
2 + sinrt
dt ln(2 + sin(rcos r)) sinr ln(2 + sin(rsinr)) cos r
6

5) Prove that
1
_
0
sin,r
r
dr =
1
_
0
sinr
r
dr for all , 0 and
1
_
0
sinr
r
dr =

2
Solution.
make the change of variable ,r = t , dr =
1

dt and we get
1
_
0
sin,r
r
dr =
1
_
0
sint
t

1
,
dt =
1
_
0
sint
t
dt =
1
_
0
sinr
r
dr
We do not prove the value is actually ,2.
6) Here is a dierent "trick". Compute the integral
1
_
0
cos ar cos /r
r
2
dr , a 6= / , a, / 6= 0
Solution
Remark that
cos ar cos /r =
a
_
b
rsin(tr)dt
Also remember the integral is convergent and
1
_
0
sinr
r
dr
Next we use a theorem about changing the integration order
1
_
0
cos ar cos /r
r
2
dr =
1
_
0
1
r
2
_
_
a
_
b
rsin(tr)dt
_
_
dr =
a
_
b
_
_
1
_
0
1
r
2
rsin(tr)dr
_
_
dt =
=
a
_
b
_
_
1
_
0
sin(tr)
r
dr
_
_
dt =
Now we must distinguish several cases:
a) a, / 0 in this case for any t 2 [a, /] ) t 0 and
1
_
0
sin(tr)
r
dr =

2
and
1
_
0
cos ar cos /r
r
2
dr = .. =
a
_
b
_
_
1
_
0
sin(tr)
r
dr
_
_
dt =
a
_
b
_

2
_
dt = (a /)

2
b) a, / < 0 in this case for any t 2 [a, /] ) t < 0 and
1
_
0
sin(tr)
r
dr =

2
7
and
1
_
0
cos ar cos /r
r
2
dr = .. =
a
_
b
_
_
1
_
0
sin(tr)
r
dr
_
_
dt =
a
_
b
_

2
_
dt = (/ a)

2
c) / < 0 < a in this case
1
_
0
sin(tr)
r
dr =
_

2
for t < 0

2
for t 0
and consequently
1
_
0
cos ar cos /r
r
2
dr = .. =
a
_
b
_
_
1
_
0
sin(tr)
r
dr
_
_
dt =
0
_
b
(

2
)dt +
a
_
0

2
dt =

2
(a + /)

Comment.
This problem is actually more complicated than it looks. We also need to prove the integral is convergent.
Otherwise the "trick" may not work.
Consider a similar integral
1
_
0
sinar sin/r
r
2
dr
The same trick does not apply because this improper integral is not convergent.
It is not convergent since
sinar sin/r
r
2
=
2 cos
axbx
2
cos
ax+bx
2
r
2
lim
x&0
2 cos
axbx
2
cos
ax+bx
2
x
2
1
x
2
= 2
Consequently the integral may be compared (by limit comparison test) to
1
_
0
1
r
2
dr which is divergent.
thus the integral
1
_
0
sinar sin/r
r
2
dr
is also divergent, which implies the integral
1
_
0
sinar sin/r
r
2
dr =
1
_
0
sinar sin/r
r
2
dr
. .
DIV
+
1
_
1
sinar sin/r
r
2
dr
. .
CONV
= divergent
is divergent.
7) Consider Lobachevskys integral
)(r) =
x
_
0
ln(cos t)dt , r 2 [0,

2
]
Prove it veries the following functional equation
)(r) = 2)(

4
+
r
2
) 2)(

4

r
2
) rln2
8
Then compute )(

2
)
Solution.
The equation is equivalent to
2)(

4
+
r
2
) 2)(

4

r
2
) rln2 )(r)
. .
A(x)
= 0
Which actually means (r) is a constant function.
All we need to do is
- prove its derivative is zero and
- nd some value r = a for which we may compute and nd that (0) = 0
Take the derivative
)
0
(r) =
_
_

x
_
0
ln(cos t)dt
_
_
0
= ln(cos r))

0
(r) =
_
2)(

4
+
r
2
) 2)(

4

r
2
) rln2 )(r)
_
0
=
= 2)
0
(

4
+
r
2
)(

4
+
r
2
)
0
2)
0
(

4

r
2
)(

4

r
2
)
0
ln2 )
0
(r) =
= 2 ln(cos(

4
+
r
2
)
1
2
(1) ln(cos(

4

r
2
)
1
2
ln2 (1) ln(cos r) =
= ln
_
cos(

4
+
r
2
) cos(

4

r
2
)2
1
cos r
_
=
= ln
_
2
(cos

4
cos
x
2
sin

4
sin
x
2
)(cos

4
cos
x
2
+ sin

4
sin
x
2
)
cos r
_
=
= ln
_
2
(
1
p
2
cos
x
2

1
p
2
sin
x
2
)(
1
p
2
cos
x
2
+
1
p
2
sin
x
2
)
cos r
_
=
= ln
_
2
1
2
(cos
x
2
sin
x
2
)(cos
x
2
+ sin
x
2
)
cos r
_
= ln
_
(cos
2 x
2
sin
2 x
2
)
cos r
_
= ln(1) = 0
Next for r = 0 we clearly have
)(0) =
0
_
0
ln(cos t)dt = 0
Consequently
(0) = 2)(

4
+ 0) 2)(

4
0) 0 ln 2 )(0) = 0
Finally for r =

2
Lobachevskys integral becames
)(

2
) = 2)(

4
+

2
2
) 2)(

4


2
2
)

2
ln2 ,
, )(

2
) = 2)(

2
) 2)(0)

2
ln2 , )(

2
) =

2
ln2

You might also like