1 − e−x
C07S0M.003: If f (x) = ln(x + e−x ), then f (x) = .
x + e−x
C07S0M.004: If f (x) = ex cos 2x, then f (x) = ex cos 2x − 2ex sin 2x.
C07S0M.005: If f (x) = e−2x sin 3x, then f (x) = 3e−2x cos 3x − 2e−2x sin 3x.
2
C07S0M.006: If g(t) = ln(tet ) = (ln t) + t2 ln e = (ln t) + t2 , then
1 1 + 2t2
g (t) = + 2t = .
t t
1
C07S0M.007: If g(t) = 3(e − ln t) , then g (t) = 15(e − ln t)
t 5 t 4
e −
t
.
t
2 + 3x
C07S0M.009: If f (x) = , then
e4x
3e4x − 4(2 + 3x)e4x 3 − 8 − 12x 12x + 5
f (x) = = =− .
(e4x )2 e4x e4x
dy dy dy
ey + xey = ; (1 − xey ) = ey ;
dx dx dx
dy ey dy ey
= ; = .
dx 1 − xey dx 1−y
In the last step we used the fact that xey = y to simplify the denominator.
C07S0M.012: Given sin (exy ) = x, we apply Dx to both sides and find that
dy dy
[cos (exy ) ] · exy · y + x = 1; xexy [cos (exy ) ] · = 1 − yexy cos (exy ) ;
dx dx
dy 1 − yexy cos (exy )
= .
dx xexy cos (exy )
1
dy dy dy
x
e +e y
=exy
y+x ; (ey − xexy ) = yexy − ex ;
dx dx dx
dy yexy − ex
= y .
dx e − xexy
dy dy dy 1 y y
1 = ey + yey ; = y = y = .
dx dx dx e + yey ye + y 2 ey x + xy
We used the fact that yey = x in the simplification in the last step.
Here is an alternative approach to finding dy/dx. Beginning with x = yey , we differentiate with respect
to y and find that
dx dy 1
= ey + yey , so that = y .
dy dx e + yey
C07S0M.015: Given ex−y = xy, we apply Dx to both sides and find that
dy dy dy
e x−y
1− =y+x ; x + ex−y = ex−y − y;
dx dx dx
dy ex−y − y dy xy − y (x − 1)y
= x−y ; = = .
dx e +x dx xy + x (y + 1)x
We used the fact that ex−y = xy to make the simplification in the last step.
x dy dy
ln y +· = 1+ ;
y dx dx
x dy
−1 · = 1 − ln y;
y dx
x − y dy
· = 1 − ln y;
y dx
dy y(1 − ln y)
= .
dx x−y
√
C07S0M.017: Given: y = (x2 − 4) 2x + 1 . Thus
1/2 1 2 1 1
ln y = ln (x − 4)(2x + 1)
2 1/2
= ln (x − 4)(2x + 1)1/2
= ln(x − 4) + ln(2x + 1) .
2
2 2 2
Therefore
1 dy x 1 5x2 + 2x − 4
· = 2 + = ,
y dx x − 4 2(2x + 1) 2(x2 − 4)(2x + 1)
2
and so
√
dy 5x + 2x − 4
2 (5x2 + 2x − 4) (x2 − 4) 2x + 1
= y(x) · = .
dx 2(x2 − 4)(2x + 1) 2(x2 − 4)(2x + 1)
1 1
ln y = ln(3 − x2 ) − ln(x4 + 1),
2 4
and therefore
1 dy x x3 x(3x2 + 1)
· =− − 4 = 2 .
y dx 3−x2 x +1 (x − 3)(x4 + 1)
Thus
1/3
(x + 1)(x + 2)
C07S0M.019: Given: y = . Then
(x2 + 1)(x2 + 2)
1
ln y = ln(x + 1) + ln(x + 2) − ln(x2 + 1) − ln(x2 + 2) ;
3
1 dy 1 1 1 2x 2x
· = + − 2 − 2 ;
y dx 3 x+1 x+2 x +1 x +2
1 1 1
ln y = ln(x + 1) + ln(x + 2) + ln(x + 3);
2 3 4
1 dy 1 1 1
· = + + ;
y dx 2(x + 1) 3(x + 2) 4(x + 3)
x
C07S0M.021: If y = x(e ) , then
3
1 dy ex
ln y = ex ln x; · = + ex ln x;
y dx x
dy (1 + x ln x)ex dy (1 + x ln x)ex
(ex )
= y(x) · ; = · x .
dx x dx x
ln x
C07S0M.022: Given: y = (ln x) , x > 1. Then
ln y = (ln x) ln (ln x) ;
1 dy 1 ln x 1 + ln (ln x)
· = ln (ln x) + = ;
y dx x x ln x x
dy 1 + ln (ln x) ln x
= · (ln x) .
dx x
x−2 1 1
lim = lim = .
x→2 x − 4 x→2 2x
2 4
Without l’Hôpital’s rule,
x−2 x−2 1 1
lim = lim = lim = .
x→2 x2 − 4 x→2 (x + 2)(x − 2) x→2 x + 2 4
ln(ln x) 1
C07S0M.028: By l’Hôpital’s rule, lim = lim = 0.
x→∞ ln x x→∞ ln x
ln(1 + x) 1 1
lim (cot x) ln(1 + x) = lim = lim = = 1.
x→0 x→0 tan x x→0 (1 + x) sec2 x 1·1
4
C07S0M.030: By l’Hôpital’s rule and the product rule for limits,
2
e1/x − 1 e1/x sin x 1/x
lim (e1/x − 1) tan x = lim = lim+ 2 = lim e = 1 · lim e1/x
= +∞.
x→0 + + x→0 cot x x→0 x csc2 x x→0+ x x→0+
Recall that l’Hôpital’s rule is valid even if the resulting limit is +∞ or −∞, although we are stretching the
hypotheses of the product rule a little here. We need a lemma.
Lemma: If
Proof: Given M > 0, choose δ > 0 such that, if x − a < δ then f (x) > p/2 and g(x) > 2M/p. Then,
for such x, f (x) · g(x) > (p/2) · 2M/p = M . Because M may be arbitrarily large positive, this implies that
f (x) · g(x) takes on arbitrarily large values if x > a and x is near a. That is,
C07S0M.031: After combining the two fractions, we apply l’Hôpital’s rule once, then use a little algebra:
sin x
1 1 1 − cos x − x2 (sin x) − 2x −2
lim − = lim 2 = lim = lim x .
x→0 x2 1 − cos x x→0 x (1 − cos x) x→0 2x(1 − cos x) + x2 sin x x→0 2(1 − cos x) + x sin x
Now if x is close to (but not equal to) zero, (sin x)/x ≈ 1, so the numerator in the last limit is near −1.
Moreover, for such x, cos x < 1 and x and sin x have the same sign, so the denominator in the last limit is
close to zero and positive. Therefore the limit is −∞.
C07S0M.032: We don’t need l’Hôpital’s rule here, although it may be applied. Without it we obtain
3
x2 x3 3x2 − 2x3 −2 0−2
lim − 2 = lim 3 = lim x = = −2.
x→∞ x+2 x +3 x→∞ x + 2x2 + 3x + 6 x→∞ 2 3 6 1+0+0+0
1+ + 2 + 3
x x x
Using l’Hôpital’s rule (three times) we find that
x2 x3 3x2 − 2x3
lim − 2 = lim
x→∞ x+2 x +3 x→∞ x3 + 2x2 + 3x + 6
6x − 6x2 6 − 12x −12
= lim = lim = lim = −2.
x→∞ 3x2 + 4x + 3 x→∞ 6x + 4 x→∞ 6
√ x2 − x − 1 − x
lim x2 − x − 1 − x = lim √ √
x→∞ x→∞ x2 − x − 1 + x
1
x2 − 2x − 1 x−2−
= lim √ √ = lim x = +∞.
x→∞ x2 − x − 1 + x x→∞ 1 1 1
1− − 2 −
x x x
5
ln x 1
C07S0M.034: ln lim x1/x = lim ln x1/x = lim = lim = 0. Therefore lim x1/x = 1.
x→∞ x→∞ x→∞ x x→∞ x x→∞
2x 2
lim 2xe−2x = lim = lim = 0.
x→∞ x→∞ e2x x→∞ 2e2x
Then
ln(e2x − 2x)
ln lim (e2x − 2x)1/x = lim ln(e2x − 2x)1/x = lim
x→∞ x→∞ x→∞ x
2e2x − 2 2 − 2e−2x 2−0
= lim = lim −2x
= = 2.
x→∞ e − 2x
2x x→∞ 1 − 2xe 1−0
Therefore lim (e2x − 2x)1/x = e2 .
x→∞
1/x2
C07S0M.036: Given: lim 1 − exp(−x2 ) .
x→∞
1/x2 2
2 1/x ln 1 − exp(−x2 )
ln lim 1 − exp(−x )2
= lim ln 1 − exp(−x ) = lim
x→∞ x→∞ x→∞ x2
2x exp(−x2 ) exp(−x2 ) 0
= lim = lim = = 0.
x→∞ 2x [1 − exp(−x2 )] x→∞ 1 − exp(−x2 ) 1−0
1/x2
Therefore lim 1 − exp(−x2 ) = e0 = 1.
x→∞
C07S0M.037: This is one of the most challenging problems in the book. We deeply regret publication of
this solution. First let u = 1/x. Then
x
1 (1 + u)1/u − e
L = lim x · 1+ − e = lim .
x→∞ x u→0+ u
C07S0M.038: First replace b with x to remind us what the variable in this problem is. Thus
√
x a x + x2 − a2
A(x) = 2πax +√ ln .
a x2 − a2 a
6
Then
x x + (x2 − a2 )1/2
= 2πa2 + 2πa2 lim ln .
x→a+ (x2 − a2 )1/2 a
Then apply l’Hôpital’s rule to the limit that remains:
e−1/x
2
1
dx = e−1/x + C. (Optional substitution: u = −1/x2 .)
2
C07S0M.044:
x3 2
7
√
x
C07S0M.045: Let u = 10 . Then
√
√ 1 10 x ln 10
du = (10 x
ln 10) · x−1/2 dx = √ · dx.
2 x 2
Therefore
√ √
10 x
2 2u 2 · 10 x
√ dx = du = +C = + C.
x ln 10 ln 10 ln 10
1
C07S0M.046: Let u = ln x. Then du = dx, so that
x
1 1 1 1
dx = du = − + C = − + C.
x(ln x)2 u2 u ln x
1
C07S0M.048: Let u = 1 + ln x. Then du = dx, and therefore
x
1 2 2
(1 + ln x)1/2 dx = u1/2 du = u3/2 + C = (1 + ln x)3/2 + C.
x 3 3
6x
C07S0M.049: 2x · 3x dx = 6x dx = + C.
ln 6
2 −1/3
C07S0M.050: Let u = 1 + x2/3 . Then du = x dx. Hence
3
dx 3 1 3 3
= du = (ln u) + C = ln(1 + x2/3 ) + C.
x1/3 (1 + x2/3 ) 2 u 2 2
12 − 24 ln 2
t= ≈ −6.68765951.
ln 2
But this value of t is negative, and in addition f (t) < 0 for all larger values of t. Thus the revenue is a
decreasing function of t for all t 0. Therefore the grain should be sold immediately.
Now
8
√ 1
100 2 exp 12 t − t1/2 exp 10 t
f (t) = 1/2
,
t
We used Newton’s method to solve f (t) = 0 and found the solution to be approximately 11.7519277504.
To make sure that this value of t maximizes the profit, we graphed f (t) for 0 t 21 (the graph is shown
following this solution). The profit upon cutting and selling after about 11.75 years will be approximately
$1202.37.
1200
1000
800
600
400
200
5 10 15 20
-200
C07S0M.053: If lots composed of x pooled samples are tested, there will be 1000/x lots, so there will be
1000/x tests. In addition, if a lot tests positive, there will be x additional tests. The probability of a lot
testing positive is 1 − (0.99)x , so the expected number of lots that require additional tests will be the product
of the number of lots and the probability 1 − (0.99)x that a lot tests positive. Hence the total number of
tests to be expected will be
if x 2. Next,
1000
f (x) = − − 1000(0.99)x ln(0.99);
x2
1
f (x) = 0 when = (0.99)x ln(100/99);
x2
(0.99)−x
x2 = ;
ln(100/99)
(0.99)−x/2
x= .
ln(100/99)
The form of the last equation is exactly what we need to implement the method of repeated substitution
(see Problems 23 through 25 of Section 3.8). We substitute our first “guess” x0 = 10 into the right-hand side
of the last equation, thus obtaining a “better” (we hope) value x1 and continue this process until the digits
in these successive approximations stabilize. Results: x1 = 10.488992, x2 = 10.514798, x3 = 10.516161,
x4 = 10.516233, and x5 = 10.516237 = x6 . The method is not as fast as Newton’s method but the formula
is simpler. (The graph of y = f (x) follows this solution to convince you that we have actually found the
minimum value of f .) We must use an integral number of samples, so we find that f (10) ≈ 195.618 and
f (11) = 195.571, so there should be 90 lots of 11 samples each and one lot of 10 for the most economical
results. Alternatively, it might be simpler to use 10 samples in every lot; the extra cost would be only about
9
24 cents. The total cost of the batch method will be about $978, significantly less than the $5000 cost of
testing each sample individually.
500
450
400
350
300
250
10 20 30 40
C07S0M.054: If f (x) = 12 x2 − 1
4 ln x, then
2 2
2 1 1 1 1
1 + [f (x) ] = 1 + x − 2
=x + + = x+ .
4x 2 16x2 4x
1 3
C07S0M.055: If f (x) = sin−1 3x, then f (x) = · Dx (3x) = √ .
1− (3x)2 1 − 9x2
7
C07S0M.056: If f (x) = tan−1 7x, then f (x) = .
1 + 49x2
1 2t 2
g (t) = · Dt t2 = √ = √ .
|t2 | (t2 )2 −1 t2 t −1
4 t t −1
4
et
C07S0M.058: If g(t) = tan−1 et , then g (t) = .
1 + e2t
1 sin x sin x
f (x) = · Dx (cos x) = − √ =− .
2 1 − cos2 x | sin x|
1 − (cos x)
2
C07S0M.060: If f (x) = sinh−1 2x, then f (x) = √ .
1 + 4x2
1 10
C07S0M.061: If g(t) = cosh−1 10t, then g (t) = · Dt (10t) = √ , t> 1
10 .
(10t)2 − 1 100t2 − 1
1
C07S0M.062: If h(u) = tanh−1 , then
u
10
1 1 u2 1 1 1
h (u) = 2 · Du =− 2 · 2 =− 2 =
1 u u −1 u u −1 1 − u2
1−
u
−1 1
C07S0M.064: If f (x) = tan , then
x
1 1 x2 −1 1
f (x) = · Dx = 2 · =− 2
1 x x + 1 x2 x +1
1+ 2
x
provided that x = 0.
√
C07S0M.065: If f (x) = arcsin x, then
1
1 1 1
f (x) = · Dx x1/2 = √ · x−1/2 = √ √ .
√ 2 1 − x 2 2 1 − x x
1 − ( x)
x 2
f (x) = · Dx x2 + sec−1 x2 = √ + sec−1 x2
|x2 | (x2 )2 −1 x −1
4
1 2x
f (x) = · Dx (x2 + 1) = 4 .
1 + (x2 + 1)2 x + 2x2 + 2
√
C07S0M.068: If f (x) = sin−1 1 − x2 , then
1 1 1 x
f (x) = · Dx (1 − x2 )1/2 = √ · (1 − x2 )−1/2 · (−2x) = − √ .
1 − (1 − x2 ) x2 2 |x| 1 − x2
sinh x
C07S0M.070: If f (x) = ln cosh x, then f (x) = = tanh x.
cosh x
C07S0M.071: If f (x) = tanh2 3x + sech2 3x, then f (x) ≡ 1 by Eq. (5) in Section 7.6. Therefore f (x) ≡ 0.
Alternatively, f (x) = 6 tanh 3x sech2 3x − 6 sech 3x sech 3x tanh 3x ≡ 0.
11
√
C07S0M.072: If f (x) = sinh−1 x2 − 1, then
1 1 1 x
f (x) = √ 2 · Dx (x − 1)
2 1/2
=√ · (x2 − 1)−1/2 · 2x = √ .
1+ x2 − 1 1+x −1
2 2 |x| x2 − 1
√
C07S0M.073: If f (x) = cosh−1 x2 + 1, then
1 1 1 2 x
f (x) = √ 2 · Dx (x2 + 1)1/2 = √ · (x + 1)−1/2 · 2x = √ .
x2 + 1 −1 x2 2 |x| x2 + 1
1 −2x 2
f (x) = · (−2x) = 2 = 3 .
1 − (1 − x2 )2 2x − x4 x − 2x
√ √ √
C07S0M.077: Let x = 2u. Then dx = 2 du and 4 − x2 = 4 − 4u2 = 2 1 − u2 . Therefore
x
1 2
√ dx = √ du = arcsin u + C = arcsin + C.
4 − x2 2 1 − u2 2
√ √
C07S0M.079: Let u = ex . Then du = ex dx and 1 − e2x = 1 − u2 . Hence
ex 1
√ dx = √ du = arcsin u + C = arcsin ex + C.
1 − e2x 1 − u2
12
demonstrating that your intelligence, guided by experience, is still required to interpret what the computer
tells you.
√ √ √
C07S0M.081: Let u = 23 x, so that x = 32 u. Then dx = 3
2 du and 9 − 4x2 = 9 − 9u2 = 3 1 − u2 . So
1 3 1 1 1 1 2x
√ dx = · √ du = arcsin u + C = arcsin + C.
9 − 4x2 2 3 1 − u2 2 2 3
C07S0M.083: The integrand resembles the derivative of the inverse tangent of something, so we let u = x3 .
Then du = 3x2 dx, so that x2 dx = 13 du. Then
x2 1 1 1 1
6
dx = 2
du = arctan u + C = arctan x3 + C.
1+x 3 1+u 3 3
1 1 2x 1 1 1 1
√ dx = √ dx = √ du = arcsec |u| + C = arcsec x2 + C.
x x −1
4 2 x x −1
2 4 2 u u2 −1 2 2
√
C07S0M.089: Let u = x if necessary, but by inspection an antiderivative is f (x) = k cosh x1/2 for some
constant k. Because
k sinh x1/2
f (x) = k sinh x1/2 · Dx x1/2 = ,
2x1/2
it follows that k = 2 and therefore that
√
sinh x √
√ dx = 2 cosh x + C.
x
√ √
C07S0M.092: Let u = 2x, so that 4x2 − 1 = u2 − 1 and du = 2 dx. Then
1 1 1 1 1
√ dx = √ du = cosh−1 u + C = cosh−1 2x + C.
4x2 − 1 2 u2 − 1 2 2
√ √ √
C07S0M.093: Let u = 23 x, so that x = 32 u. Then dx = 32 du and 4x2 + 9 = 9u2 + 9 = 3 u2 + 1.
Therefore
1 3 1 1 1 1 2x
√ dx = · √ du = sinh−1 u + C = sinh−1 + C.
4x2 + 9 2 3 u2 + 1 2 2 3
14
√
C07S0M.094: We expect to see an integrand containing u2 + 1, so let u = x2 . Then du = 2x dx, and
thus
x 1 1 1 1
√ dx = √ du = sinh−1 u + C = sinh−1 x2 + C.
x4 + 1 2 u2 + 1 2 2
See the Index in the textbook for references to further information on the Riemann zeta function ζ(z).
C07S0M.097: We use some results in Section 7.6. By Eqs. (36) and (37),
1
−1 1 1 1+ x 1 x+1
tanh = ln = ln = coth−1 x
x 2 1 2 x−1
1−
x
(provided that |x| > 1). By Eq. (35), if 0 < x 1 then
√ √
−1 1 1 1 1 1 − x2 1 + 1 − x2
cosh = ln + − 1 = ln + √ = ln = sech−1 x
x x x2 x x2 x
√
by Eq. (38). Note that x2 = x because x > 0.
15
If x(0) = 1 and x (0) = 0, then A = 1 and kB = 0, so A = 1 and B = 0. If x(0) = 0 and x (0) = 1, then
A = 0 and kB = 1, so A = 0 and B = 1/k.
C07S0M.099: The graphs of y = cos x and y = sech x are shown following this solution, graphed for
0 x 6. One wonders if there is a solution of cos x = sech x in the interval (0, π/2). We graphed
y = f (x) = sech x − cos x for 0 x 1 and it was clear that f (x) > 0 if x > 0.25. We graphed y = f (x)
for 0 x 0.25 and it was clear that f (x) > 0 if x > 0.06. We repeated this process several times and
could see that f (x) > 0 if x > 0.0002. Instability in the hardware or software made further progress along
these lines impossible. Methods of Section 11.8 can be used to show the desired inequality, but Ted Shifrin
provided the following elegant argument.
First, sech x 1 sec x if x is in I = [0, π/2). Moreover, sech x = sec x only for x = 0 in that interval;
otherwise, sech x < sec x for x in J = (0, π/2). So sech2 x < sec2 x if x is in J.
But tanh x = tan x if x = 0. Therefore tanh x < tan x for x in J because Dx tanh x < Dx tan x for
such x. That is,
sinh x sin x
<
cosh x cos x
if x is in J. All the expressions involved here are positive, so
if x is in J. That is,
for x in J. But cosh x cos x = 1 if x = 0, and we have now shown that Dx (cosh x cos x) < 0 if x is in J.
Therefore
for x in J. In other words, sech x cos x for x in I and sech x > cos x for x in J.
Because cos x 0 < sech x for π/2 < x < 3π/2 and because cos x is increasing, while sech x is
decreasing, for 3π/2 x < 2π, it follows that the least positive solution of f (x) = 0 is slightly larger than
3π/2, about 4.7, exactly as the figure suggests.
We then applied Newton’s method to the solution of f (x) = 0 with x0 = 4.7, with the following results:
x1 ≈ 4.7300338216, x2 ≈ 4.7300407449, x3 ≈ 4.7300407449. Thus x3 is a good approximation to the least
positive solution of cos x cosh x = 1.
0.5
1 2 3 4 5 6
-0.5
-1
16
C07S0M.100: If f (x) = sinh−1 (tan x), then
Therefore
sec x dx = sinh−1 (tan x) + C
if sec x > 0; that is, if k is an odd integer and kπ/2 < x < (k + 2)π/2. On the intervals where sec x < 0, we
find that
sec x dx = − sinh−1 (tan x) + C.
cosh x cosh x
g (x) = 2 = = sech x,
1 + sinh x cosh2 x
and therefore
sech x dx = tan−1 (sinh x) + C
for all x.
1 1 x1/2 − 2
F (x) = 1/2
− = ,
2x x 2x
so F (x) = 0 when x = 4. Clearly F is decreasing on (0, 4) and increasing on (0, +∞), so the global
minimum value of F (x) is
F (4) = 2 − ln 4 = 2 − 2 ln 2 > 2 − 2 · 1
x1/3 − ln x
x ←− x − .
1 −2/3 1
x −
3 x
Beginning with x0 = 100, we get x5 ≈ 93.354461.
For part (c), suppose that j(x) = x1/p is tangent to the graph of g(x) = ln x at the point (q, ln q). Then
q 1/p
= ln q and j (q) = g (q). Hence
1 (1/p)−1 1
q = ; q 1/p = p;
p q
p = ln q = ln pp = p ln p; ln p = 1, so p = e.
17