Professional Documents
Culture Documents
Calculus Instructor S Manual Components
Calculus Instructor S Manual Components
Starting from f (0) = 0 at constant velocity v, the distance function is f (t) = vt. When f (t) = 55t the
+
velocity is v = 55. When f ( t ) = 55t 1000 the velocity is still 65 and the starting value is f (0) = 1000. In
each case v is the slope of the graph of f . When v(t)is negative, the graph of f(t)goes downward. In that case
area in the v-graph counts as negative.
Forward motion from f (0) = 0 to f (2) = 10 has v = 6. Then backward motion to f (4) = 0 has v = -5. The
<
distance function is f (t)= 5t for 0 t 5 2 and then f ( t ) equals S(4 - t ) (not -5t). The slopes srs 5 and -6.
The distance f (3)= 5. The area under the v-graph up to time 1.5 is 7.5. The domain of f is the time interval
<
0 t 5 4 , and the range is the distance interval 0 5 f <
10. The range of v(t) is only 6 and -5.
+
The value of f ( t ) = 3t 1 at t = 2 is f (2) = 7 . The value 19 equals f (6).The difference f (4) - f(1) = 9.
That is the change in distance, when 4 - 1 is the change in time. The ratio of those changes equals S, which is
+ + +
the slope of the graph. The formula for f ( t ) 2 is 3t 3 whereas f ( t 2) equals it+7. Those functions have
+ +
the same slope as f : the graph of f ( t ) 2 is shifted u p and f (t 2) is shifted to the left. The formula for
f (5t)is 15t+l. The formula for 5f ( t )is 1st +5. The slope has jumped from 3 to 15.
The set of inputs to a function is its domain. The set of outputs is its range. The functions f ( t ) = 7+3(t-2)
+
and f ( t ) = vt C are linear. Their graphs are straight lines with slopes equal to S and v. They are the same
function, if v = 3 and C = 1.
0
)
for 0 < t < T
for T < t < 2T
for 2T < t < 3T
2Ofor t C . 2 20t for t 5 .2
525;22;t+lO 76;-30 9v(t)={
0 for t > .2 f(0 = { 4 for t 2 . 2 11 10%;12+%
13 f ( t ) = 0,30(t - I), 30; f ( t ) = -Sot, -60,30(t - 6 ) 15 Average 8, 20 17 40t - 80 for 1 5 t 2.5 <
2 1 O < t < 3 , - 4 0 5 f 52O;O<t<3T,O< f <6OT 233-7t 256t-2 273t+7
8 for O c t c T 8t for 0 5 t 5 T
29 Slope -2; 1 f < < 9
{
3 1 v ( t )=
-2 for T < t < 5 T = 1 {
2 -
for T 5 t 5 5T
1 6 f(t) = { o
30(t - 1) 15 t
30 t 2 2
05t52
2St54
t>4
+ +
1 8 v(t) = 8 then 1(after t = 2); f (t) = 6 8 t then 20 t.
+
20 1200 30%= 40s when 1200 = lox or x = 120 yearbooks. The slope is SO. If it goes above 40 you
can't break even.
22 Range = {0,20,40); the velocity is not defined at the jump.
+
24 f (t) = 4 t 1 (linear up) or -4t + 9 (linear down).
+
26 The function increases by 2 in one time unit so the slope (velocity) is 2; f (t) = 2 t C with constant C = f (0).
28 f (2t) = 2vt must equal 4vt so v = 0 and f=0. But ia(2t)2 does equal 4(iat2). To go four times as far in
twice the time, you must accelerate.
SO f (t) = 0 then 8 - 2 t (change at t = 4); slopes 0 and -2; range -2 5 f (t) 5 0.
+
32 f(t) = 3t = 12 at t = 4; then v = 6 gives f (t) = 12 6(t - 4) = 30 at t = 7. The extra distance was 18 in 3
time units; thus u(t) = 3 then 6.
%
34 C(F) = (F - 82) his slope g. 6
+
36 At t = 0 the reading was .061 10(.015) = .all. A drop of .061- .04 = .021 would take .021/.015 hours.
This was the Exxon Valder accident.
38 Domain 1 < t 5 5; range 5 f (t) < oo.
40 Domain 0 5 t < 4 and 4 < t 5 5 (omit t = 4); range $ 5 f (t) < oo
< <
42 Domain 0 t 5; range 2-' (or &) f (t) 5 1.
44 Jump from 0 to 1 at t = 0;jump from 2 to S at t = 0; jump from 0 to 1 at t = -2; jump from 0 to S at
t = 0; jump from 0 to 1at t = 0.
462f(3t)=2(3t-1)=6t-2;f(l-t)=(l-t)-l=-t;f(t-l)=(t-l)-l=t-2.
+ +
48 fi(t) = S t 3; f2(t) = S t 18.
6 0 aA function assigns an output to each input -"
+ + + + +
52 3(vt C) 1 has slope Sv; v(3t 1) C also has slope h;2(4vt C) has slope 8v; -ut + C has slope
+ + +
-v; vt C - C has slope v; v(vt C) C has slope v2.
64 A function cannot have two values (the upper and lower branches of X) at the same point. Apparently only
1.2 Jumps in Velocity (page 14)
When the velocity jumps from ul to y, the function u(t) is piecewise constant. The distance function
+
f (t) is piecewise linear. In the first time interval, f (t) = f (0) v l t. After the jump at t = 1, the formula is
+
f (t) = f (1) v j ( t - 1). In case fo = 6 all distances are increased by 6 and all velocities are t h e same.
With distances 1, 5, 25 at unit times, the velocities are 4 and 20. These are the slopes of the f -graph.
The slope of the tax graph is the tax rate. If f (t) is the postage cost for t ounces or t grams, the slope is the
cost per ounce (or per gram). For distances 0, 1, 4 , 9 the velocities are 1, 3, 5. The sum of the first j odd
numbers is f j = j2.Then flo is 100 and the velocity ulo is 19.
The piecewise linear sine has slopes 1, 0, -1, -1, 0, 1. Those form a piecewise constant cosine. Both
+
functions have period equal to 6, which means that f (t 6) = f (t) for every t. The velocities u = 1,2,4,8, .. .
have uj = d-l. In that case fo = 1and f j = 3. The sum of 1,2,4,8, 16 is 31. The difference 2 j - 2j-I equals
d-l. After a burst of speed V to time T, the distance is VT. If f (T) = 1and V increases, the burst lasts only
to T = 1/V. When V approaches infinity, f (t) approaches a s t e p function. The velocities approach a delta
function, which is concentrated at t = 0 but has area 1under its graph. The slope of a step function is eero o r
infinity.
2 f (6), f (7) are 66, 77 and -11, -13 and 4, 9. Then f (7) - f (6) is 11,-2,5.
4 The increases f (4) - f (1) are 12 - 3 = 9 and 14 - 5 = 9 and 18 - 9 = 9.
6 h(t) = .St + 3; the slopes of f , g, h are 3, 2.5 and 3 - 2.5 = .5.
8 f(t)=l+lOtforO<t< $,f(t)=2fort> &
1 0 f (3) = 12; g(f (3)) = g(12) = 25; g( f (t)) = g(4t) = 8t + 1.Distance increases four times as fast and velocity
is multiplied by 4.
+
12 10,160.50 is f(44,900) = 2782.50 .28(44,900 - 18,550).
<
1 4 F(x) = 2f ( i x ) = . E x for x 37,100; then F(x) = 5565 + .28(x - 37,100) up to x = 89,800;
+
then F(x) = 20,321 .33 (x - 89,800) up to x = 186,260; then F(x) = .28x beyond 186,260.
T h e 1991 r a t e s o n t h e front cover have only three brackets.
1 6 f (t) = 3 + 2 t for t 2 1 is continuous; f (t) = 4 + 2 t is discontinuous (because f (1) = 5). f (x) = .15x
1.3 The Velw'ty at an Instant (page 21)
+ .
then SO00 .28 (x - 18,550) has a jump at $18,550.
fl = 1,f2 = 3 , f3 = 7, f j = 3 - 1 ; fl = - I , fa = O , f3 = -1, f j = {-1 for odd j,Oforeven j)
= &(-l)j - 1).
The big triangle has area = $ (base)(height) = ij2and the j small triangles have area ij. Together they
+ + +
give rectangles of total area 1 2 .. . j. Note: Another drawing could move the diagonal line up by i.
The big triangle still has area ij2 ;
and the strip across the bottom has area j.
False when the vj are ( i ) j ; false when the uj are - ( i ) j ; true when all fj+, = f j
(p is the period) because then V j + p = fj+p - fj+,-1 equals f j - fj-1 = ~ j false ; when all uj = 1.
+
Assume fo = 0. First f j = j2, second f j = j, by addition third f j = j2 j, by division last f j = f (j2+j)
which is 1 2+ + + j.
f (99) = 9900 and f (101) = 10302;A f /At = 40212 = 201.
+
Take v = C, 2C, 3C, .. Then f = C, 3C, 6C,. ... The example f3 - 2f2 fl gives 6C - 2(3C) C = C. +
The answer is always C (by Problem 30).
+
fj+1 - 2fj fj-1 equals (fj+1 - fj) - ( f j - fj-1) = vj+l- vj. If v is velocity then is acceleration.
The period of v + w is SO, the smallest multiple of both 6 and 10. (Then v completes five cycles and
w completes three.) An example for functions is v = sin +
and w = sin y ( v w has a nice graph).
f(l2) = ( 1 + 2 + 1 + 0 ) + ( 1 + 2 + 1 + 0 ) + ( 1 + 2 + 1 + 0 ) = 1 2 . Then f(14) = 1 2 + 1 + 2 = 15
+ + + +
and f (16) = 15 1 0 = 16. f doesn't have period 4 since X I + 22 23 xd is not aero.
2j is 2 times 2j-l. Subtracting 2j-' leaves 2j-l. Similarly 3 j is 3 times 3'-I and subtraction leaves
2 times d-1.
fl - fo equals vl = 2fo = 2 so fl is 3; fa - fl equals = 2f1 = 6 so f2 is 9; then f3 is 27 and f4 is 81.
.
Problem 36 shows that f j = 3 j fits the requirement vj = 2fj- 1.
+
vj = f j - fj-1 equals d -d-'. Adding the V'S gives (fl - fo) (fa - fl) + (f3 - fa) + (fj - fj-1).
.
Cancelling leaves only f j - fo = d - 1.
- a s +
U(t) - U(t - 1) is aero except between t = 0 and t = 1 (where it equals 1). If this is the velocity, then the
distance is f(t) = t up to t = 1; then f(t) = 1: a 'short burst of speedn. If the square wave is distance,
then v(t) is a delta function at t = 0 minus a delta function at t = 1.
The sum jumps up by 1 at t = 0, 1, 2. Its slope is a sum of three delta functions.
I Forj= 1,Ndo
V j = f j - fj-1
Examples 2j and j2 and 2 j give vj = 2 and vj = 2j - 1 and vj = 2j-'.
FINDV (FINDF (vl , . VN)) brings back vl , . . ,VN. But FINDF (FINDV (fo, fi , . .,fN))
produces O s f l - f 0 , f 2 - f0, " ' , f N - foe
The average age increases with slope 1 except at a birth or death (when it is discontinuous).
Between the distances f (2) = 100 and f (6) = 200, the average velocity is 25. If f (t) = f t2 then f (6) = 9
and f (8) = 16. The average velocity in between in 3.5. The instantaneous velocities at t = 6 and t = 8 are S
1.4 Circular Motion (page 28)
and 4.
+ +
The average velocity is computed from f (t) and f (t h) by uave = h ( f ( t h) - f (t)). If f (t) = t2 then
+
uaw = 2 t h. &om t = 1 to t = 1.1 the average is 2.1. The instantaneous velocity is the limit of Uave. If the
distance is f (t) = *at2 then the velocity is u(t) = at and the acceleration is a.
On the graph of f (t), the average velocity between A and B is the slope of t h e secant line. The velocity
at A is found by letting B approach A. The velocity at B is found by letting A approach B. When the
velocity is positive, the distance is increasing. When the velocity is increasing, the car is accelerating.
3a(t'+2th+ha)- #at3
(a) = 6 (limit is 6); (b) 6(t+h)+2-("+21
h = 6 (limit also 6); (c) h +
= at $ah (limit
is at); (d) t+h-(t+h)'-(t-t)=
h = 1 - 2t - h (limit is 1- 2t); (e) = 0 (limit is 0 ) ; (f) the limit is u(t) = 2t
(and f(t) = t2 gives +
= 2t h).
U=a-O=
At 1 2; = 3-
2'
!M h Q= 1 h.
& + 6 lim +
= lim(1 h) = 1= slope of the parabola at t = 0.
u(t) = 3 - 2t gives a line through (0,3) and (1,l); f (t) = 3t - t2 gives a parabola through (0,O) and (3,O) with
maximum at (g, 5).
10 Slope of f (t) = 6t2 is u(t) = 12t; slope of u(t) = 12t is a = 12 = acceleration.
+
1 2 A f = i a ( t h)' - $a(t - h)2 = 2 ath; then $$= 9 = at = velocity at time t. The region under the
lie u = at is a trapezoid. Its area is the base 2h times the average height at.
1 4 ' h e (the slope is ); false (the curve is partly stdper and partly flatter than the secant line which gives
the average slope); true (because A f = AF); false (V could be larger than u in between).
1 6 The functions are t2 and t2 - 2 and 4t2. The velocities are 2t and 2t and 8t.
1 8 The graph is a parabola f (t) = i t 2 out to f = 2 at t = 2. After that the slope of f stays constant at 2.
3;
20 Area to t = 1 is *; to t = 2 is i ; to t = 3 is 2; to t = 4 is to t = 5 is *; area from t = 0 to t = 6 is zero.
The graph of f (t) through these points is parabola-line-parabola (symmetric)-line-parabola to zero.
22 f (t) is a parabola t - i t 2 through (0,0), (1, *), and (2,O); f (t) is the same parabola until (1, *), but the
second half goes up to (2,l); f (t) is the parabola 2t - t2 until (1,l) and then a horizontal line since u = 0.
+ + +
24 The slope of f is v(t) = at b; the slope of u is the constant a; f (t) = i t 2 t 1equals 41 when t = 8.
+
(The quadratic formula for i t 2 t - 40 = 0 gives t = -1 f d- = -1 f 9.)
26 f (t) = t - t2 has v(t) = 1- 2t and f (3t) = 3t - 9t2. The slope of f (3t) is 3 - 18t. This is 8v(St).
28 To find f (t) multiply the time t by the average velocity. This is because u,,(t) = f M td U -- t o
A ball at angle t on the unit circle has coordinates x = cos t and y = sin t. It completes a full circle at
t = 2 x . Its speed is 1. Its velocity points in the direction of the tangent, which is perpendicular to the radius
1.4 Circular Motion (page 28)
coming out from the center. The upward velocity is cos t and the horizontal velocity is - sin t.
A mass going up and down level with the ball has height f (t) = sin t. This is called simple harmonic motion.
The velocity is v(t) = cos t. When t = */2 the height is f = 1and the velocity is v = 0. If a speeded-up mass
reaches f = sin 2t at time t, its velocity is v = 2 COB 2t. A shadow traveling under the ball has f = cost and
v = -sin t. When f is distance = area = integral, v is velocity = slope = derivative.
110,(0,-1,-10 3(4cost,4sint);4and4t;4costand-4sint
5 3t; (cos 3t, sin 3t); -3 sin 3t and 3 cos 3t 7 x = cost; ,/-2/2; - 4 / 2 9 2n/3; 1;2a
11Clockwise starting at (1,O) 3
13 Speed 1 5 Area 2 1 7 Area 0
19 4 from speed, 4 from angle 2 1 f from radius times 4 from angle gives 1in velocity
i;
23 Slope average (1 - $)/(2/6) = = .256 25 Clockwise with radius 1from (1,0), speed 3
27 Clockwise with radius 5 from (0,5), speed 10 29 Counterclockwise with radius 1from (cos 1, sin I), speed 1
31Left andrightfrom (1,O) to (-1,0), v = -sint 33Upanddownbetween2 and-2;start 2sin8,v = 2cos(t+B)
35Upanddownfrom(O,-2)to(0,2);v=sinit 3 7 z = c o s E , y = s i360,
n ~ speed &,
vu, = cos %
39 I think there i s a stop between backward and forward motion.
2 The cosine of ,
is x = -A -, the sine is y =$; t
the tangent is = -&; the ball has a distance & to go
(draw triangle from (0,O) to (x, y) and back down at right angle); the speed is 1so the added time is 4 and
the total time is 9 + 4. Not easy.
4 x = R cos t and y = R sin t; velocity -R sin t and R cos t; distance and velocity triangles both grow by R.
6 The angle is 5+ 3t; the position is x = cos(: + 3t) = - sin 3t and y = sin(: +
3t) = cos 3t; the vertical
velocity is -3 sin 3t (= horizontal velocity of original ball).
The new mass at x = cost, y = 0 never meets the old mass at x = 0, y = sint. The distance between them
+
is always \/eosz t sinZt = 1.
+ +
f = sin(t a ) equals - sin t; the velocity is cos(t r ) which equals - cos t. The ball is a half-circle ahead
of the original ball.
+ + + +
f (t) = sin t cos t has f a = sinz t 2 sin t cos t cos2 t which is the same as 1 2 sin t cos t
+
(or 1 sin 2t). The maximum is at t = 45' = 2 when f a = 2. Then f, = fi. Its graph is a sine curve
with this maximum point: f (t) equals fisin(t +
).
The ball goes halfway around the circle in time n. For the mass to fall a distance 2 in time a we need 2 = ? a 2
so a = 4 1 9 .
1
+
36 Along the line z y = 1between (1,O) and (0,l). Starting at (1,O) the z and y velocities are -2 sint cost
+ i
and 2 sin t cos t. (Maybe introduce cos2 t = $ $ cos 2t and sin2 t = $ - cos 2t to find velocities
- sin 2t and sin 2t : Discuss.)
S8 Choose k = 2r. The speed is 21r and the upward velocity is 2% cos Zrt.
Starting with a right triangle, the six basic functions are the ratios of the sides. Two ratios (the cosine x/r
and the sine y/r) are below 1. Two ratios (the secant r/x and the cosecant r/y) are above 1. Two ratios (the
tangent and the cotangent) can take any value. The six functions are defined for all angles 8, by changing
from a triangle to a circle.
The angle 8 is measured in radiana. A full circle is 8 = 2r, when the distance around is 2rr. The distance
to angle 8 is BY. All six functions have period 2%. Going clockwise changes the sign of 8 and sin 8 and tan 8.
Since cos(-8) = cos 8, the cosine is unchanged (or even).
+ +
Coming fmm x2 9 = 9 are the three identities sin2 8 cos2 8 = 1 and tan28 1= see2 8 and +
+
1 cot2# = csc28. (Divide by r2 and x2 and y2.) The distance from (2,5) to (3,4) is d = a.The d b
tance from (1,O) to (cos(s - t), sin(s - t)) leads to the addition formula cos(s - t)= cos s cos t +
sin s sin t.
+
Changing the sign of t gives cos(s t) = cos s cos t - sin s sin t. Choosing s = t gives cos 2t = cos2 t - sin2t
+
or 2 cos2t - 1.Therefore $ ( l cos 2t) = cos2 t, a formula needed in calculus.
1Connect corner to midpoint of opposite side, producing 30' angle Sr 7 $ + area ?r28
9 d = 1, distance around hexagon < distance around circle, 11T; T; F; F
+
I S cos(2t t) = cos 2t cos t - sin 2t sin t = 4 cos3 t - 3 cos t
15$cos(s-t)+~cos(s+t);~cos(s-t)-?cos(s+t) 17cos8=sec8=flat8=nr
+
I9 Use cos(5 - s - t) = cos(5 - 8) COB t sin(; - 8) sin t 2 3 8 = $+ multiple of 2 r
25 8 = +: multiple of r 27 No 8 29 4 = q S 1 lOPl= a, 1OQ1= b
2 r,3r, -: radians equal 180°, 540°, -45'. Also 60°, 90°, 270' equal i, $ radians.
f, The alias of 480'
is 120' and the alias of -lo is 559'.
+ +
4 cos 2(8 r) is the same as cos(28 2u) which is cos 28. Since cos2 8 = ? + ? cos 26, this also has period r.
6 Notice the patterns in this table.
8 Straight distance a;quarter-circle distance 5;
semicircle distance also 5.
10 8 = (0 - + (1 - 9)' !+
= +
1- fi f = 2 - fi.Then 12d = 6.21. This is the distance around a
twelvesided figure that fits into the circle (curved distance is 2r.)
1 2 *om the inside front cover or the addition formulas: sin(* - 8) = sin 8, cos(r - 8) = -COB 8, sin(; + 6) =
+
cos 8, COS(; e) = -sin e.
+ +
1 4 sin 3t = sin(2t t) = sin 2t cos t cos 2t sin t. This equals (2 sin t cos t) cos t + (cos2 t - sin2 t) sin t or
3 sin t - 4 sin3 t.
1.5 Review of fiigonometry (page 33)
+ +
(cost + i sin t)2 = cos2t - sin2t 2 i sin t cos t. Then the double-angle formulas give cos 2t i sin 2t.
A complete solution is not expected! Finding a point like s = ~ 1 2t ,= 3x12 is not bad.
+ +
Formula (9) is sin(s t) = sin s cos t cos s sin t. Replacing t by -t gives formula (8) for sin(s - t).
+
tan(. t) =
+
-*
(Ask why this replacement is allowed. It is not easy for a student to explain.)
= :f
tan(s t) = l'_"PA:t::jt.
z:?;:::. TOsimplify, divide top and bottom by cos s and cost :
sec 8 = -2 when cos 8 = -+, which happens first at 0 = 120' = 2x13. Also at 8 = 240' = 4x13. Then
+
at all angles 2x13 27rn and 4x15 2 m . +
sin 8 = 8 at 8 = 0 and never again. Reason: The right side has slope 1 and the left side has slope cos 8 < 1.
(Draw graphs of sin 8 and 8. A solution with negative 8 would give a solution for positive 8 by reversing
sign.)
1tan 0 = 0 when 8 is a multiple of r. The ratio y/z is r u o when y = 0, so the point on the circle in Figure
1.20 has to be on the z axis.
+ +
A sin(%+ 4) equals A sin z cos 4 A e m z sin 4. Matching with a sin z b cos x gives a = A cos 4 and
b = A sin#. hen a 2 + b 2 = A ~ C O S ~ # + A ~ S ~T ~~ U SA
~ ~ =d =
A ~m . a n d t a n d = AA csin4 -g
o e d - b'
+
The distance squared from (0,O) to R is (a b cos 8)2 (b + which simplifies to
+ +
a2 + 2 a b cos 8 b2. Notice the parallelogram law: (diagonal)2 (other diagonal)2 = 2a2 + 2b2
+ + +
which is (side)2 (next side)2 (third side)2 (fourth side)2.
The amplitude and period of 2 sin x z are both 2.
+
By Problem 30, sin x cos z equals 6
sin(x + 2). The graph should show a sine function with maximum
near \/Z at x = i.
The graph of t sin t oscillates between f 45' lines. The graph of sin 4t sin t oscillates inside the g r a p h
of sin t. See the nraph on pane 294, at the end of Section 7.2.
2.1 The Derivative of a hnction (page 49)
CHAPTER 2 DERIVATIVES
+
The derivative is the limit of A f /At as At approaches sero. Here A f equals f ( t A t ) - f (t). The step
At can be positive or negative. The derivative is written v or d f ' d t or fl(t).If f (x) = 2x+ 3 and Ax = 4 then
A f = 8. If Ax = -1 then A f = -2. If Ax = 0 then Af = 0. The slope is not 0/0 but df/dx = 2.
The derivative does not exist where f (t) has a corner and v(t) has a jump. For f (t) = l / t the derivative
is - 1 t 2 The slope of y = 41%is dy/dz = -4/x2. A decreasing function has a negative derivative. The
independent variable is t or z and the dependent variable is f or y. The slope of # (is not) ( d y / d ~ ) ~The
.
+
slope of ( ~ ( z )is) ~2u(x) d u / d x by the square rule. The slope of (22 3)2 is 2 ( h + S ) 2 = 8x 12. +
1(b) and (c) +
S 12 3h; 13 3h; 3; 3 +
5 f (x) 1 7 -6 +9 2 z + A s + 1;2z+ 1
1 1 4 - A - -4
t+at t - ?r + lS7;9;corner 15A=1, B=-1 17F;F;T;F
1 9 b = B; m and M; m or undefined 2 1 Average z2 xl -,2x1 +
25~;nolimit(one-sidedlimits1,-1);1;1ift#0,-1ift=O 27 ft(3);f(4)-f(3)
29 2x4(4z3) = ax7 31 2
= 2u = 2+ 3s = - I 2,
* 2
f'(2) doesn't exist 8 6 2f = 4 u 3 2 5
2 (a) = becomes 2x at h = 0 (b) ~ z + 6 ~ ~ - ' a= lohXb+lrha
= 2% 5h becomes h at h = 0 +
(c) --
-
i&-h)a-(=-h)a 4zh
ii; = h always (d) = + oo as h -r 0
2h
+ +
4 x2 1,x2 10,d- 100
6 The line and parabola have slopes 1 and 22. So the touching point must have z = $. There y = for a
+
the line, y = (a)2 c for the parabola so c = a.
1
$f 2* -W) =~=-,;~~=y=-2;w~=B~=- ~ar-
a00 -
10 Az
= 1 + 2 ( x + ~ t ) + 3 ( ~ + ~ ~ ) ~ - 1 - 2 ~=
Az
- 3 2+62+3Ax.
~' Then $ = 2+6z.
.rfi
12 Af = t + ~ t - i) =- - ?t (t+At) = owd divide b y A t and set A t = 0 :
answer t -2t-0
, = -- 2
t3'
1 4 y = 3z2 has 2
= 3 times 22 and then 9
= 3 times 2 = 6.
l 8 A t z=2wewant y = 4 a n d 2
= ~ + 2 2 = 0 .S O A + ~ B + ~ and= B~ + ~ ( ; ) = o . ~ h e n ~ = - l
and A = 2.
1 8 1 - A = -.
~ + h 2-h = kr!;,
Divide by 2h because the centered difference went from z - h to
+
z
h (an average over distance 2h). Division by 2h leaves = 3; at h = 0 this is = 2 3.
d?+&)-sr(?) - 3-4, - 12 (forward difference); d A ) - u P - L ) =
20 The ratios are
h -x- ,yy= -24 (backward
difference); y v = = -18 (centered difference is closest).
22 The graph of f (t) has slope -2 until it reaches t = 2 where f (2) equals -1; after that it has slope zero.
So f' jumps from -2 to 0 (undefined at the jump).
24 is always zero, as At gets smaller. The limit of zero (unchanging number) is zero.
26 If has any limit then f (0) must be sero. (In this section functions an assumed to be civilized.) Then
ft(0) is the limit of w, which is %
and approaches 7. Example: f (z) = 7x x2. +
28 By the square rule S ( z ) ' = 2 z ( g ) = 21
30 If u = 1the square rule gives & ( I ) ~= 2 ( 1 ) 2 or 2
= 2 times 2.
This is possible because is zero and
2.2 Powers and Polynomials (page 56)
2 t i m e s xero i s xero.
+ +
32 In the figure, f (t At) is the height of the curve above t At; the time step A t is the distance from t
+
across to t At; the change A f is the height of one red "bullet" above the other. The secant line between
bullets has slope s. The tangent line at the lower bullet has slope f ' ( t ) .
34 For x = 0 and A x = 1 the function f ( x ) = x2 - x has A f = f (1)- f (0) = 0. But the slope f' at x = 0 is -1.
This problem will be worded more carefully in the future.
36 (a) False First draw a curve that stays below y = x but comes upward steeply for negative x. Then create
a formula like y = -x2 - 10. (b) False f ( x ) could be any constant, for example f ( x ) = 10. Note what is
true: If 5 1 and f ( x ) 5 x at some point then f ( x ) 5 x everywhere beyond t h a t point.
+ +
38 For f ( x ) = i x the graph of f ( x h ) = i ( x h ) is above it by the vertical distance i h . Then A f = $ h
is a horbontal line (down near the axis!) and i
= is also horbontal.
The derivative off = x4 is f = 4x3. That comes from expanding (x+h)' into the five terms x4 + 4x3h+ 6x2h2+
4xhS + h4. Subtracting x4 and dividing by h leaves the four terms, 4x3 + 6 x 2 h + 4xh2+h3.This is Af /h,
and its limit is 4x3.
+
The derivative of f = xn is f' = nxn-l. Now ( x h)" comes from the binomial theorem. The terms to
' +
look for are xn- h, containing only one h. There are n of those terms, so ( x h)" = xn + nxn- ' h . . . After +
subtracting xn and dividing by h, the limit of A f / h is d-l. The coefficient of xn-jhj, not needed here, is
' n choose j" = n!/j! (n-j)!,where n! means n(n-1) - (1).
The derivative of x - ~is -2x-'. The derivative of %'I2 is ix-'l2. The derivative of 3x+ ( l / x ) is 3 - l / x 2 ,
+
which uses the following rules: the derivative of 3f ( x ) is 3 f 1 ( x )and the derivative of f ( x ) g(x) is f'(x)+ g' ( x ).
Integral calculus recovers y from dyldx. If dy/dx = x4 then y(x) = x 5 / 5 .
2f(x)=+x7(or:x7+C) 4f'(x)=7(3)+5(3)=-1-U
x 2 xs '
+ +
6 f ( x ) = x4 2x2 1 so + +
= 4x3 2(2x) = 4x3 4x. Or use the square rule: = 2(x2 + I ) & (x2+ 1) =
+
2(x2 1)(2x)= 4x3 42. +
xn-l
8 = $(nxn-l) = - - 1 1
(n-I)!. Note the step n(n-;)...(l) - o...(l) = (n-l)!
2.2 Powers and Polynorniak (page 56)
First difficulty: The number of terms should be a whole number, so x is restricted to integers. Real difficulty:
Increasing x not only increases each of the terms in x x + + + . z, it also increases t h e number of
terme. If x increases by 1, then x x + + + + + +
x not only increases by 1 1 . . 1, but also by another
+
x (or maybe x 1).
+ 3 +
The slope of x $ is 1- which is sero at x = 1. At that point the graph of x $ levels off. (The function
reaches its minimum, which is 2. For any other positive x, the combination x +; is larger than 2.)
The function f (x) = $ has a negative derivative but f (x) never becomes negative. (To define f (x) for all x,
take f(x) = 2 - x up to x = 1.)
The units of f'are feet p e r second; the units of f" an, ft/sec2. The second 16 is 16 ft/sec2.
At a point where $ = 0, t h e tangemt to t h e g r a p h is horisontal. This may be a minimum point or
a maximum point; for y = x3 the origin is a .pause point".
~ ~ =J5L t h e n ~ y = * - ~ ~= ~ ~ = ( m u l t i p l y t o p a n d b o t t o m b y ~ + ~ z + h ) =
.*- x+h&(\E+ x+h)
Cancel x - x in the numerator and divide by h : 9 = m-(++ml.
-1
If yl' = x then y' = i x 2 (plus any constant C) . Then y = (i x3) plus Cx plus any constant D :
y = g1x S + C x + D .
2
The derivative of ( U ( X ) )is~2u(x) by the square rule. If u = xn then the derivative of zan is
*(nxn-l) = 2nxan-l which follows the power mle.
If g = ~ ( x then
) + +
(a) 4f(x) has slope 4v(x) (b) f(x) x has slope v(x) 1 (c) f(x+l) has slope v(x + 1)
+ +
(d) f(x) v(x) has slope v(x) J (x)
32 y = has 2 = = ,fi and second derivative ,fi and eventually the nth derivative is 1. Check
e we are led to $ = 1.
,*+=
n = 3 : y = $,yt = $,y" = x , ~ " ' = 1. ~ o t how
If g = - x - ~then f (x) = -12-I - i z - 2 36 $ = 2 f i is solved b y y = x2 (provided x > 0).
If y = yo + ex then E(z) = d &
= which approaches 1 as z m.
(a) High price elasticity means that the price curve steepens: as you buy more stock and get close to having
a corner on the market. (b) Low price elasticity mans that the curve flattens: switch to unlimited
service for making local phone calls.
y = x n has E = d e = n . ~ h e r e v e n u e x y = x n + lhas E = n + l .
a.
Marginal propensity to save is dS Elasticity is not needed because S and I have the same units. Applied
to the whole economy this is macroeconomics.
Relative growth of y and x is 3. A child is born with relatively large head s i ~ ey. Then growth of
the body catches up (n < 1).
In general 9 + + +
= a2 ab b2. We can directly verify f x + h ~ - z s = (x h ) l + (x + h)x + z2.
As h -r 0 this gives 2 + + +
= 3x2. Similarly $$ = a3 0 2 b ob2 b3 and directly
+ + + + +
= (z h)3 (x h)2x (x h)x2 x3. +
+ +
Two graphs touch when the difference y3 = y l - y 2 = x4 x3 - 72 5 is zero. At x = 1 we find y3 = 0
+
(graphs touch) and also y&= 4x3 3x2 - 7 = 0 (graphs are tangent). The curves don't cross.
The expected payoff can be greater than the cost of buying a ticket for every combination. This happens
when most other players have chosen from a small set of favorite 'lucky" numbers. The Massachusetts
lottery does have unequal popularity of different numbers, but not enough to advise buying every
combination. Better to choose the unpopular numbers.
2.3 The Slope and the Tangent Line (page 63)
A straight line is determined by 2 points, or one point and the slope. The slope of the tangent line equals
the slope of the curve. The point-slope form of the tangent equation is y - f (a) = f' (a)(x - a).
+
The tangent line to y = x3 x at x = 1 has slope 4. Its equation is y - 2 = 4(x - 1).It crosses the y axis
at y = -2 and the x axis at x = $. The normal line at this point (1,2) has slope - f . Its equation is y - 2 = -
.
(x-1) The secant line from (1,2) to (2, 10) has slope 8. Its equation is y - 2 = 8(x-1).
1* ; y - 6 = 3 ( ~ - 2 ) ; y - 6 = $ ( ~ - 2 ) ; ~ - 6 = - ~ ( ~ - 2 ) 3 y+l=3(x- l);y=3x-4
5y=x;(3,3) 7 y - a = (c+a)(x-a);y-a=2a(x-a) 9y= ~ ~ ~ + 2 ; ~ - 7 = - ~ ( ~ - 5 )
11y = 1 ; x = lSy-$=-$(~-a);~=~,x=2a 1 5 c = 4 , tangent at x = 2
1s
17(-3,19)and(5,2r) 19~=4,y=3-xtangentatx=l
+ + +
2 1 (1 h)'; 3h 3h2 h3; 3 3h ha; 3 + + 23 Tangents parallel, same normal
+
25 y = 2ax - a2, Q = (0, -a2); distance a2 !; angle of incidence = angle of reflection
27x=2p;focushasy=~=p 2 9 y - I = x + ~ - x = - 1 =
a fi' fi -Jz
31 y - a 2 = - & ( x - a ) ; y = a 2 + I - 2 , a = A 33 (-$)(1000)=10at x = 10 hours 35 a = 2
+
37 1.01004512; 1 10(.001) = 1.01 +
39 (2 Axl3 - (8 6Ax) =+ AX)^ + AX)^ 4 1 XI = :; x2 = fi
4o
4 3 T = 8 see; f (T) = 96 meters 45 a = f meters/sec2
2 y = x 2 + x h a s ~ = 2 ~ + 1 = 3 a t z = 1 , y = 2 . T h e t a n g e n t l i n e i s y - 2 = 3 ( x - 1 ) o ry=3x-1.The
normal line is y - 2 = -S(x 1 - 1)or y = -:+ g.
The secant line is y - 2 = m(x - 1)with
m= l w - = ~ + h .
+
4 y = x3 62 has 2 +
= 3x2 6 = 6 at x = 0, y = 0. The tangent liie is y = 6x. (Note how x3 disappears.)
+
The only crossing where x3 6 s = 62 is at x = 0.
6 x = y2 is y = fi with = Lx-ll2
2
2 4
= ?(?) = at x = 4. The tangent line is y - 2 = z(x 1 - 4).
+
8 (x - 1)(x - 2) is zero at x = 1 and x = 2. If this is the slope (it is x2 - 32 2) then the function can be
+ +
4x3 - ex2 2x. We can add any Cx D to this answer, and the slopes at x = 1 and 2 are still equal.
y = x4 - 2x2 has 2
= 4x3 - 42. At x = 1and x = -1 the slopes are zero and the y's are equal.
The tangent line (horizontal) is the same.
10 The slope from (a, l/a) to (c, l/c) is
A 1
= = -- .
1 So the secant line has equation y - = - (x - a).
,a A
As c approaches a this becomes y - = - 2(x - a), the equation of the tangent line. Note the slope
1 1
-5 for the function y = $.
1 2 If a -r b and c --+ b then approaches ft(b), the slope at b. Test on y = x2 and y = $.
+
1 4 If g(x) = f (x) 7, the tangent lines at x = 4 are parallel. But the perpendicular(!) distance between them
is less than 7, unless they are horizontal. (The vertical distance is 7.)
+ +
16 The problem requires 52 - 7 = x2 cx and (slopes) 5 = 22 c, at the same x. Then x = F.
Substitute into
the first equation: 5(?) - 7 = ( 9 I 2 c(?). +
Move all terms to the left side and simplify:
E4 f - 4L o o r c = *,hi.
2.4 The Derivative of the Sine and Cosine (page 70)
18 Tangency requires 42 = cz2 and also (slopes) 4 = 2cz at the same z. The second equation gives z = $ and
then the k t is = which has no solution.
+ + -
20 The parabolas pass through z = 1,y = 0 if 1 b c = 0 and d 1= 0. They are tangent (same slope)
if2+b=d-2.Thend=1andb=-~andc=2.Theparabol~~y=z~-3z+2and~=l-z~.
-
22 The tangent line at z = 1has equation y f (1) = f'(1) (z - 1). For the secant line change f'(1) to w.
For f (z) = z + + +
bz c (a parabola) we require f'(1) = 2 b to equal ~9+3b+c)-(1+b+c~
3-1 =4 b +
+ +
(Impossible!). So try a cubic like f (z) = x3 bz2. Then f '(1) = 3 26 equals 27+9b -(l+b)
,I1 = 13 + 4b
if b = -5, which gives one possible answer f (z) = z3 5z2. -
24Fory=~+1atx=aandy=z-z2atz=cwerequireequalsloperk=1-2c.
+
The normal line y - (a2 1) = g ( z - a) must go through the closest point y = c - c2 at z = c.
- +
(Compare Problem 23.) Then (c - c2) (a2 1 ) = d ( c - a). (Find solution not required:
(i
c - c2 - - c ) -~ 1= &(c - i+ C)yields a cubic equation for c. Calculuu will minimbe ( d i s t a n ~ e ) ~
which involves z4. Then derivative = 0 gives the same cubic.)
26 If a vertical ray is reflected horisontally, the tangent must go down at a 45O angle (slope -1). For y = at
z = a this means 2 3
= = -1 and a = in the figure.
28 (a) If y = 22 is the tangent line at (1,2), then y - 2 = - i ( x - 1)is the normal line. (b) As c approaches
k t M approaches ft(a). (c) The line through (2,s) with slope 4 is y - 3 = ((2 - 2).
a, the secant slope f k c-a
3OThe tangent lineis y- f(a) = f'(a)(z-a). Thisgoes through y =g(b) at z = bifg(b) - f ( a ) =f'(a)(b-a).
The slopes are the same if g' (b) = f'(a).
S2 When the circle touches the parabola y = f at z = a, the normal line has equation y - $ -!
= (z - a).
+
That line touches z = 0 when y = $ 1. The distance to (a, a2) equals the radius 1when
+ +
(a)2 ($ 1- a2)' = 12.This gives a = 0. The circle rests at the b o t t o m of this flatter parabola.
34 The secant lines all have Islopel 1 1so their limit the tangent line has I$ 1 I1. In other words
lg(a)l = h+ 1.
- 1
11.
36 If # U' X
= 7 then u(z) = 7v(z) and ul(z) = 7 J ( z ) and *v = 7. But (#)I = &(7) = 0.
38 The tangent line to y = $ at r = 1is y - 1= -l(z - + -
1). At 2 = 1 A z this gives y = 1 Az. The curve is
at height y = &. The difference is & - (1 - Az) = I-(I-A=)(I+AZ~
1+Az
+ X'
= LA$
40 The distance between curve and tangent line is of order AX)^. The tangent line ignores the second
derivative.
44 With acceleration changed from 3 to 2m/sec2, Example 4 has equal speeds when 2(T - 4) = V or T = ?V 4. +
+ +
The distance VT must equal 72 $(2)(T - 4)' when the cars meet. Then 72 f V2 = V(?V 4) +
+
gives 0 = f v2 4V - 72 and V = -8 + m. Check: V is less than 12 because the other car is slower.
+ -
46 To just pass the baton, the runners reach the same point at the same time (ut = -8 6t i t 2 ) and with the
+ -
same speed (v = 6 - t). Then (6 - t)t = -8 6t - i t h d i t 2 8 = 0. Then t = 4 and v = 2.
The derivative of y = sinz is y' = cos x. The second derivative (the derivative of the derivative) is
y" = -sin x. The fourth derivative is yl)"
= sin x. Thus y = sin z satisfies the differential equations yl) = -y
and = y. So does y = cos x, whose second derivative is -coa x.
{I'
All these derivatives come from one basic limit: (sin h ) / h approaches 1. The sine of .Ol radians is very close
2.4 The Derivative of the Sine and Cosine (page 70)
to .01. So is the tangent of .01. The cosine of .Ol is not .99, because 1 - cos h is much smaller than h. The
-
1. Therefore cos h is close to 1 ;ha and cos .0l ar .99995. We can replace h
ratio (1 - cos h)/h2 approaches 2
by x.
It
The differential equation { = -y leads to oscillation. When y is positive, y is negative. Therefore y'
is decreasing. Eventually y g a s below Bero and y" becomes positive. Then y' is increasing. Examples of
oscillation in real life are springs and heartbeats.
1 M=-W+-l.
-
sln h 8in2h-2 h ain2h-f h h
4 t a n h = 1.01hat h = O a n d h = f . l 7 ; t a n h = h a t h = 0 .
8 .995004 versus .995; .8776 versus .875; .866 versus .863; .9986295 versus .9986292.
10(a) - l+ca
1-cos2 h
h ha = 1 s i ; h 2 , + ( b ) wha- 4 -h2 +)l ( , ) sin2 h -
1-cos2h-
2
(d) 1-cos2h - -2 1 c d 2 h + 2(0) =o.
h
12 (,) &O) = tanh-tan0 = sinh
h h coa h
+ t
= 1 (b) $(0) = sin(-h)-sin(-ol - --
h
sin h , -1
h
1 4 The slopes of cos x and 1- $x2 are - sin x and -x (close for small x). The slopes of sin x and x - gx
1 3
(close for small x) are cos x and 1- ix2.
16 sin(z+h)-sin(z-hl -
2h
- (sin z cos h+cos zsin h)-(ain lco.h-c08 z sin h l -
2h
- 2 COB X Sin h + COB X.
1 8 (a) y - sin0 = (cos O)(X - 0) or y = x (tangent
- -is 45' line) (b) y - sin r = (cos r ) ( x - r) or y = -x +r
(c) y - sin f = (cos ): (x - ): or y - = 9 g(x-2).
+ +
20 (a) sin(% 1) = sin x cos 1 cos x sin 1. The derivative is cos x cos 1- sin zsin 1 which is cos(x 1). +
(b) 4 = ain(z+l+Az)-sin(z+l) - - sin(X+Az)-8inX -+ cos X = cos(x 1). +
Az Az Az
22 sin 2(z+h)-sin 22 - sin 2z(col, 2h- l)+cos22 sin 2h
. Then c O s ~ +
- ~0 and
h
- .
h = 2 w + 2. So the limit
is 2 =0+2coszX.
24 The maximum of y = sin x
that point is cosx - a s i n x =
+ &cos x is at x = & (or 30') where y = + fie = 2. The slope at
- $ = 0. Note that y is the same as 2 cos x shifted to the right by :.
26 (a) False (use the square rule) (b) n u e (because cos(-x) = cos x) (c) False for y = x2 (happens to be
true for y = sin x) (d) True (y" = slope of y' = positive when y' increases)
28 y = sin 52 has y" = -25 sin 5%so y satisfies the equation y" = -25y. (In general y = sin kx satisfies
y" = -k2 y.)
30 z ( n ) = limit of y("+Ax'-~(") sin(r+.Ol) = - sin .Ol
AZ .Fory=sinxandAx=.01theratiois .ol = -.99998.
2.5 The Product and Quotient and Power R u b (page 77)
2.5 The Product and Quotient and Power Rules (page 77)
The derivatives of sin z cos z and 1/cos z and sin a/ car z and tan3 z come from the product rule, reciprocal
mle, quotient m e , and power mle. The product of sin z times cos a ha. (uu)' = u J u'v = cos2x - sin2x. +
The derivative of l/u is -v'/v2, so the slope of sec x is sin x/coa2x. The derivative of u/v is (vu'- uv')/v2
+
so the slope of tan z is (cos2x sin2x)/cos2x = sec2x. The derivative of tan3 z is S tan2x sec2x. The
slope of zn is d-'and the slope of (u(z))" is d - l d u / d x . With n = -1 the derivative of (cos z)-I is
-l(coa sin x), which agrees with the rule for sec z.
+ .
Even simpler is the role of linearity, which applies to au(z) bu(z) The derivative is au' (x) tnr' (x). +
+
The slope of 3 sin z 4 cos z is 5 cos x - 4 sin x. The derivative of (3sin z 4 cos is 2(S sin x 4 cos x ) + +
(3 cos x - 4 sin x). The derivative of sin4x is 4 sin3 x cos z.
122 -
+
7 -z2 sin x 42 cos z 2 sin z +
* 5(z-2)(~-3)+(z-l)(z-3)+(z-l)(z-2)
9 22 - 1- 112 f i sin z cos z T 3z-1/2 sin2 z+ (sin COB z3
1S4Pcosz-z4sinz+cos4z-4zcos3zsinz 1 5 ~ z 2 c o s z 1 7 0 19-~(z-5)-s~3+~(5-z)-5/3(=O?)
2 1 3 ( ~ i n z c o s z ) ~ ( czo- +
s ~sin2 z) 2cos2z +
23 u'uwz duwz w'uur duuw + +25 -csc2 z - sec2 x
27 v - t coat V I = eo.t--t.int--t'.int
l+t 8
29 lot for t < 10, &
(l+qa
for t > 10
A = 2(ifT+tcost+ e)
~ ' = 2 ( c o s t - t s i n t + m - f$)
81 3$$$$;
+ +
SS u''u + 2utu' utf' ;ul"u 3u1'u' 3utu" u"' + + ;
S5 sin2 t; tan2 t; [(I t)'l2 - I] i +
59 T; F; F; T; F 4 1 degree 2n- 1/ degree 2n 43 u(t) =cost-tsint(t 5 g);u(t) = -;(t 2 5)
4 5 y = ~ + ~ h a . ~ = ~ a t z = ~ ( n o c r a . h ) a n d a t x = - L ( n o d i v e ) . T h e6 V 2
Lnh ~'
L= -+ (E ) a n d
fi
&'
=6~'h
-LT( L + 1)-
+
12 z312(3sin2 z cos z) $z112 sin3 z + (sin z) 'I2 cos z
14 G(fi +
1)?z-'l2 + +
z/jE(fi 2) i z - 1/2 (fi + +
1)(fi + + +
2) $x-'12 = (3% 6 6 2) ;z-'l2 (or other form).
1 6 1O(z - 6)' +10sinQz cos x.
1 8 csc2 z - cot2z = &, -
e.& = = 1SO the derivative is zero.
20 lain x+ca z)(coa .+sin j - ( a i n z-coa x)(cosz-sin z) -2
(sin x+coa x)'
23 a m x ha. derivative ain Z( -z d n X+COS ' - -z+sin o cos
2)-x C O Z~ ~ C O S1
ain2 z sina z (or other form).
2)+ 2)
24 [ U ( X ) ] ~ ( ~ U ( X )[ u ( z ) ] ~ ( ~ u ( z )
+
26 z cos z sin z - sin z = x cos x (we now have a function with derivative z cos z).
28 The three slabs have volume uuAw and uw Av and vw Au.
SO (a) Volume = %Ph = (l+talr)a(l+tj rrt4 ha. rate of change ( l+ta/')' ( l + t ) 4 r r t S - nlt 4+( lt+~t 5)/(l+t)'
' ) ' - n t 4 (1+t)2(1+t3/')at1/'
The limit of a, = (sinn)/n is zero. The limit of a, = n 4 / P is zero. The limit of a, = (-1)" is not
defined. The meaning of an 4 0 is: Only finitely many of the numbers (a,! can be greater than 6 (an
arbitrary positive number). The meaning of a, + L is: For every 6 there is an N such that lan - LI < E
if n > N. The sequence 1,1+ 1 4, + 4 +
&, . is not convergent because eventually those sums go past any
number L.
The limit of f (x) = sin x as x + a is sin a. The limit of f (x) = x/(xl as z -r -2 is -1, but the limit as
x -,0 does not exist. This function only has one-sided limits. The meaning of lim,,, f (x) = L is: For every
c there is a 6 such that 1f (z) - L( < c whenever 0 < (x- a1 < 6.
A + B means that A is a sufficient condition for B. Then B is true if A is true. A # B means that A is a
necessary and s d c i e n t condition for B. Then B is true if and only if A is true.
4 (a) [a, -r l]*[-a, +- 11 (b) [a, + O]=+[an- a,,-1 -+ 0] (c) [a, I n]*[an = n]
(d) [an -+ sin an -r O] (e) [an + o]=+[$ fails to converge] (f) neither implication.
6 Given any c > 0, there are & and such that 1f (z) - L/< a if 0 < 1z - a1 < and Ig(z) - MI < a if
O < l z - a l < & . W e b = smallerof & a n d 6 2 a n d a d d : l f ( z ) + g ( z ) - L - M I < a + e i f ~ < lz-al<6.
8 No limit 10 Limits equals f'(1) if t h e derivative exists. 1 2 rint = A coro + 1 = 0
1 4 1x1 = -z when z is negative; the limit of 2 is 1. 16 &kffd + f'(a) if the derivative exists.
$ =%=4
c-a
18 = z + 5 approaches 1 0 a s z + 5 2 0 , e approaches asz--r2
22 - tan - 1-rin COB x
x = I-gin r l+gin x = I-aim' x
7GZ-( l+ain x ) con t(l+ain x) l+rin t
- con t aPPM).ches = 0 at ==. 2
24 *(&) 4
approaches 1 = 3 1 as z 4 1 26 Statement (2) is the definition of a limit.
28 Given any a > O there is an Xsuch that lf(z)l < a i f z < X.
50 If(%)- 21 < c means I& - 21 < .O1 or1 - 1 < .O1 or 2 < .0111+zl. This is true for r > 199.
52 The limit is e = 2.718
54 (a and b)* 6 + oo (no limit) (c and d) fw + 6 as z + oo or z -r -00.
56 The range of z is 0 < lz - a1 < 6. If 6 is reduced the range becomes smaller. So it remains true that
If (z) - LI < a for all allowed z.
58 There is an N such that la, - LI < c for n > N. Also larn- LI < E f o r m > N (and thus IL- arnl < 0.
Now add: I ( a n - L ) + ( L - a m ) J < c + c o r ) a , - a r n l < 2 c
40 (a) .a93999 approaches L = .494. (b) With a simple pattern the professor will find L. With random
choice there is no hope. Maybe try .49301101-. with 1's in the 2nd, 3rd, 5th, and all prime number
positions. The limit requires C(.1)prime = unknown?
42 The average L has ''in in each decimal position: L = .; ?? b.
= $(.111..-) = Second method: The first
digit could be 0 or 1 (average h). After that is another random sequence with average &L, since
+ +
it is shifted by one decimal. So the average & $L is the same as L and & &L = L yields L = h.
44 For every 6 the number c = 2 h a the required (and silly) property: 1 cos a ) < 2 if 1x1 < 6.
Continuity requires the limit of f (x) to exist as x + a and to agree with f(a). The reason that z/lzl is not
continuous at z = 0 is : i t j u m p s fkom -1 t o 1. This function does have one-sided limits. The reason that
1/ cos z is discontinuous at x = r / 2 is that it approaches inhity. The reason that cos(l/z) is discontinuous
at x = 0 is Infinite oscillation. The function f (x) = has a simple pole at z = 3, where f has a double
pole.
The power zn is continuous at all z provided n is positive. It has no derivative at z = 0 when n is between
0 and 1. f(x) = sin(-z)/x approaches -1 as z 0, so this is a continuous function provided we define
-+
f (0) = -1.A "continuous function' must be continuous at all points in its domain. A 'continuable function"
can be extended to every point x so that it is continuous.
If f has a derivative at z = a then f is necessarily continuous at z = a. The derivative controls the speed at
which f (z) approaches f(a). On a closed interval [a, b], a continuous f has the extreme value property and the
intermediate value property. It reaches its maximum M and its minimum m, and it takes on every value
2.7 Continuous hnctions (page 89)
in between.
2 c = cosS r = -1. Then the function is (A) continuous and (B) differentiable.
4 c = 0 gives f (z) = 0 : both properties (A) and (B)
6 c = -2 gives f (z) = z3 : both properties (A) and (B)
8 c > 0 gives f (z) = zC: For 0 < c < 1this is not differentiable at z = 0 but is continuous for (z 2 0).
For c 2 1 this is continuous and differentiable when it is defined (z 2 0 for noninteger c).
+
10 Need z c = 1 at z = c which gives 2c = 1or c = .I1 Then z +i matches 1 at z = (continuous but
not differentiable).
1 2 c = 1gives continuity at z = 0. However sec z is not defined for all z 2 0, which spoils (A) and (B).
1 4 e = 1gives f (z) = +
= z 1which a w e s with 2c = 2 at z = 1 (continuous but not differentiable).
1 6 At z = c continuity requires c2 = 2c. Then c = 0 or 2. At z = c the derivative jumps from 22 to 2.
+
1 8 Iz CI is continuous, but not differentiable, at z = -c (slope jumps from -1 to 1).
22 cos $ 24 ,*
20 lz2 + c21 is continuous and differentiable at z = -c (slope jumps from -1 to 1).
26 f (z) = 1% - 11-'12
28 (a) Choose c = 1 (or any a less than 4). There is no 6 such that 132 - 71 < 1when z is within 6 of 1.
i.
(b) 132 - 31 < $ if lz - 11 < So take 6 = 8 1.
50 (a) One-sided limits: -r -1 as z = 0- and
7 ; + as z + o+. (b) sin 1x1 has a two-sided limit at z = 0.
(c) Iz2 - 11 has a sharp corner at z = 1 and -1. It equals 1- x2 between z = -1 and x = 1.
The slope changes from 2s to -22 and back to 22. One-sided limits at z = 1 and -1.
a? sin 1-0
5 2 U s e ~ s i n ~ ~ < l . T h e n ( a ) z 2 s i n ~ + 0 a s z + 0 ( b-) +Oasz+O.
3 )+
(c) f '(z) = z2(cos $ ) (- k) +
(sin (22) = - cos $ 22 sin $ has no l i t as z -r 0. (Part (c) needs the
chain rule or careful limits. Main point: f'(z) has no limit as z -+ 0 even though f'(0) = 0)
1 for s z o
is discontinuous but f '(z) = 1.
\
36 cos z is greater than 2s at s = 0; cos z is less than 2%at z = 1. The continuous function cos z - 22 changes
from positive to negative. By the intermediate value theorem there is a point where cos z - 2x = 0.
5
38 zsin $ approaches zero as x -r O (so it is continuous) because 1 sin 1 < 1. There is no derivative because
= % sin it = sin has no limit (infinite oscillation).
40 A continuous function is continuous at each point x in its domain (where f (z) is defined). A continuable
function can be defined at all other points z in such a way that it is continuous there too. f (z) = $
is continuous away from z = 0 but not continuable.
42 f(z) = z if z is a fraction, f (z) = 0 otherwise
44 Suppose L is the limit of f (z) as z -r a. To prove continuity we have to show that f (a) = L.
For any a we can obtain 1f (z) - LI < e, and this applies at x = a (since that point is not excluded any more).
Since a is arbitrarily small we reach f (a) = L : the function has the right value at z = a.
3.1 Linear Approximation (page 95)
On the graph, a linear approximation is given by the tangent line. At z = a, the equation for that line is
+ +
Y = f (a) ft(a)(x- a). Near x = a = 10, the linear approximation to y = z3 is Y = 1000 300(x - 10). At
z = 11the exact value is (11)3= 1331. The approximation is Y = 1300. In this case Ay = 531 and dy = 300.
+
If we know sin z, then to estimate sin(% Az) we add (cos x) Ax.
+ +
In terms of z and Az, linear approximation is f (z Az) m f (z) f t(x)Ax. The error is of order (Az)P or
(z - a)p with p = 2. The differential dy equals dy/dx times the differential dx. Those movements are along
the tangent line, where Ay is along the curve.
+ +
2 f ( z ) = $ and a = 2 : Y = f ( a ) f t ( a ) ( z - a ) = - i ( z - 2 ) . Tangent line is Y = 1 - az.
4 f (z) = sin z and a = $-:Y = sin 5+ 5)
cos q ( z - = 1. Level tangent line.
+
6 f (z) = sin2 z and a = 0 : Y = sin2 0 2 sin 0 cos O(z - 0) = 0. Tangent line is x axis.
+
8 f (z) = sin z, a = 0, Y = sin 0 cos 0(.02 - 0) = .02. Compare with sin .02 = .019999
+
1 0 f (z) = dl4,Y = l6'/' + f 16-3/4(15.99- 16) = 2 ti(-.01) = 1.9996875. Compare 15.99'/" = 1.9996874.
12 f (x) = sin z, a = a, Y = sin x + cos ~ ( 3 . 1 4- x) = -1(.00159). Compare sin 3.14 = -.00159.
1 4 Actual error: cos(.Ol) - 1= -4.99996 predicted error for f = cos z, f' = - cos z near x = 0 :
+ ( A Z ) ~ ~=~+(.oooi)(-1)
O) = -5 10-5
1 6 Actual error: (1.01)-~ - (1- .03) = .00059; predicted error for f = 5, 5f" = 9
: !j(.01)~ = .00060.
18 Actual error: a - (3 + ;(-.01)) = -4.6322 -'
predicted error for f = fi,f " = qza/l near
z = 9 : $(.01)~(&) = -4.6296 lo-'.
20 (1- u)-'I2 has derivative -$(I - u)-~/'(-I). At u = 0 these equal 1 and +.Then rr 1 + iu or
1
4 s m 1 + +z2.
22 df = - sin z dz and df = l x - ' ) ~ ~ ' ~ ! ~')(lldz
. = +dx
(x-1)
+
(by the quotient rule) and df = 2(x2 1)(2x)dx =
4x(z2 + 1)dz by the power rule. Notice dz in the formula for df.
24 A = 6z2 so dA = 122 dz. 26 V = xr2h so dV = w2dh = ~ ( 2 ) ~ ( 0 .= 5 )4 . 2 ) .
28 With 5 as z in Problem 20, the correction Am is mo times i z 2 = $ (:)'. Then Am times c2 is i m o v 2 =
energy equivalent to change in mass.
SO JG + !+
has derivatives $ (1 x)-'I2 and - (1 z ) - ~ / At i.
~ . z = 0 the second derivative is - The difference
3.2 Maximum and Minimum Problems (page 103)
y - Y between curve and tangent line is about ?yt'(~)x2= - f x2. This is negative so Y is higher than y.
If df /dx > 0 in an interval then f (x) is increasing. If a maximum or minimum occurs at x then f' (x) = 0.
Points where f' (x) = 0 are called stationary points. The function f (x) = 3x2 - x has a (minimum) at x = 8.
A stationary point that is not a maximum or minimum occurs for f (x) = xS .
Extreme values can also occur when ft(x) is not defined or at the endpoints of the domain. The minima of
1x1 and 5x for -2 5 x 5 2 are at x = 0 and x = -2, even though df /dx is not zero. x* is an absolute maximum
when f (x*) 2 f (x) for d x. A relative minimum occurs when f (x*) If (x) for all x near x*.
R(x) - C(x); v; z;
V = ~ ( -6 $)(12 - 22); x M 1.6
- profit ss x = &;zero 35 x = 2
+
39 A = r r 2 x2, x = i ( 4 - 27rr); rmin = 2
rnax area 2500 vs 7 = 3185 4 3 x = 2, y = 3 4 5 P(x) = 12 - x; thin rectangle up y axis
h=f,r =R 3
V = = $ of cone volume
r = ;&* best cylinder has no height, area 29rR2 from top and bottom (?)
r = 2, h = 4 5 5 25 and 0 66 8 and -oo
,/+W+ Jq2 ( 8 - x)2; $ = Z e 7 - e= 0 when sin a = sin c
y=x2=i ( 1 , - 1 , - ) 63m=lgivesnearestline 65m=i 67equal;z=i
$x2 7 1 T h e (use sign change of ft')
Radius R, swim 2R cos 8, run 2Re, time + ;rnax when sin 8 = &, rnin all run
2 ft(x) = 3x2 - 12 = 0 gives stationary points x = 2 and x = -2. The graph comes up to a local maximum at
x = -2 and falls to a local minimum at x = 2. ( You could ask for o sketch of the graph.)
4 ft(x) = 22 - 5
= 0 at x = 1. That stationary point (and endpoint) has f (1) = 3. Then f (x) increases to
16.5 at the endpoint x = 4.
( I22) = 0 at x = $. This stationary point is a minimum; f (x) -+ +oo at the endpoints.
6 ft(x) = (x - X ~ ) - ~ -
8 ft(x) = 22 - 4 for 0 5 x < 1,f '(x) = 22 for 1 < x I2. No stationary points. The endpoints give f (0) = 0
and f (2) = 0. These are maxima. The graph has a rough point (corner) at x = 1, where f (1) = -3 and
3.2 Maximum and Minimum Roblems (page 103)
+
32 We receive R(x) = ax + bx2 when the price per pizra is p(x) = a ba. In reverse: When the price is p we
sell z = pizzas. We expect b < 0 because additional pizzas are cheaper.
+
34 The profit crosses rero when 32 - z2 = 1+x%r 2x2 - 32 1= 0 or (22 - l)(x - 1) = 0. The profit is positive
between the roots z = $ and x = 1. The largest profit is when 3 - 22 = 21 or z = t.
+ + +
For cost 2 z2, the equation 32 - z2 = 2 x2 or 2x2 - 32 2 = 0 has no real roots (b2 - 4ac is - 7
in the quadratic formula). So the profit never crosses zero.
+
Volume of popcorn box = x(6 - z) (12 - x) = 72%- 18z2 z3. Then +
= 72 - 362 3x2. Dividing by 3 gives
+
x2 - 122 24 = 0 or z = 6 f d n = 6 f f i at stationary points. Maximum volume is at
x =6- a. (V has a minimum at z = 6 + 6, when the box has negative width.)
Classic: The side lengths are 12 - 22 so the volume is V (z) = x(12 - 2 ~ ) By ~ .the product rule
+
= (12 - 2 ~ 4 2 - 2x)(-4). Factor out 12 - 22, which is zero when x = 6 (no volume).
) 1~
Then 5 = 0 for 12 - 22 - 42 = 0 or z = 2. The maximum volume is V = 2(8)2 = 128 cubic inches.
Let x be the length of the sides perpendicular to the wall. The side parallel to the wall uses the remaining
200 - 22 feet. The area is A = 4200 - 22) and = 200 - 42 = 0 at x = 50. Then A = 6000 square
?
feet. Alternative: x is the length parallel to the wall and the other sides have length (200 - z). The
maximum of A = ix(200 - z) is still 5000 square feet.
Let the sides perpendicular to the existing fence have length x. This leaves 300 - 22 meters of fence so the
other sides are 150 - z meters. (Note: x > 50 is not allowed or this length would be below the existing
100 meters.) The area A = z(l5O - x) has = 150 - 22 = 0 when z = 75 meters which is not allowed.
3.2 Maximum and Minimum Problems (page 103)
In the American Math Monthly of May, 1990, Mary Embry-Wardrop finds the maximum area for each
orientation of the rectangle. Parallel to the hypotenuse, it can have corners at (2t, 0) and (0,3t) so one side
has length f i t . Draw the figure to see similar triangles. The other side has length '2-6". SOthe area is
m
t(12 - 6t) with maximum at t = 1as in boldface above.
+
46 The cylinder has radius r and height h. Going out r and up i h brings us to the sphere: r2 ( i h ) 2 = 1. The
volume of the cylinder is V = nr2h = x [ l - ($h)2]h. Then +
= n [ l - ( i h ) 2 ] n(-ih)h = O gives
1 = i h 2 . The best h is 3, so V = *[I- $13= A.Note: r2 + 5 = 1 gives r = fi.
e ~gives h = H ( l -
48 The equation in ~ r o b l 47 g). ~ h i the
n side area is A = 2nrH(1- 5 ) and
% = 27rH - 4 n H 5 = 0 for r = Z.
R
In that case A = 2 r f ~ ( $ )= $~RH.
+
50 The triangle with height y and base 1 z is similar to a triangle with height 8 and base z (hypotenuse along
: :+
the ladder). Then & = gives y = (1 z). The ladder length L = (1 x ) ~ y2 = (1 x ) ~ ( + + I + + 3)
.. + 3)
has $$ = ( 1 + ~ ) ~ ( - $ $ ?2)( 1 + z ) ( l + =(l+z)[-y - +2+ 9
Thus 51.
= 0 when z3 = 64 2
--
and x = 4 and L = 5" = 125.
rhe upper triangle has area z d n (twice a right triangle with side z and hypotenuse 1). Similarly the
lower triangle has area z d a . The derivatives are 4- - = and
,/--L--
4-2x2 These add to rero when 4-221 = or ( 1 - ~ ~ ) ( 4 - 2 z =
~ )( ~4 - ~ ~ ) ( 1 - 2 z ~ ) ~
,/D - =/,* , / JG3
+
or 16 - 32z2 20z4 - 4z8 = 4 - 17z2 + 20z4 - 4z8 or 12 = 15z2. When x2 = r; the areas are
4
66 To find where the graph of y(z) has greatest slope, solve d!Y
dx2
= 0 maximize 2).
For y = & and
A! = -2s and &
d = ( ~ + x ~ ) ~ ( - ~ ) + ~ x ( ~ x ) ( I + x ~ )
dz (l+z2)' (I++ = 0, the slope is greatest when
+
(1 z2)(-2) +
2z(4x) = 0 or 6z2 - 2 = 0 or x = -
3.3 Second Derivatives: Bending and Acceleration (page 110)
+ +
68 Suppose y is fixed. The minimum of z2 zy - y2 is where 2z y = 0 so m(y) = (-:)" iY
- y2 = - 5~ 2y.
The maximum of m(y) is sero. +
Now z is fixed. For the maximum of z2 zy - y2 take the y
derivative to get z - 2y = 0 or y = fz. The maximum is M(y) = z2 + z(?z) - = . zx
The minimum of M(y) is sero. So max of min equals min of max.
+ +
70 When $ is 2r, 4r, 6r, .. the slope 1- 2 cos($) 4zsin(!) equals 1- 2 0 = -1. So the wavy function
+
z 2 2 sin(:) is decreasing when z = &, &,
. .,nearer and nearer to z = 0 (but it is increasing at z = 0).
72 Y(X)= - 1 ~ 1 for 1x1 1has ymin = -1 and ym, = 0.
The direction of bending is given by the sign of fl'(x). If the second derivative is positive in an interval,
the function is concave up (or convex). The graph bends upward. The tangent lines are below the graph. If
fl'(z) < 0 then the graph is concave down, and the slope is decreasing.
At a point where f' (z) = 0 and f " (z) > 0, the function has a minixnun. At a point where f' (x) = 0 and
f" (x) < 0, the function has a maximum. A point where fl' (z) = 0 is an inflection point, provided f" changes
sign. The tangent line crosses the graph.
The centered approximation to f'(z) is [f (x + Ax) - f (x - Ax)]/2Az. The 3-point approximation to f'' (z)
is
+
[f(x + Ax) - 2f (x) f (x - ~ x ) ] / ( A z ) ~
The +
. second-order approximation to f (z Ax) is f (z) f '(z) Ax + +
f f " ( x ) ( ~ x ) ~Without
. that extra t- this is just the h e a r (or tangent) approximation. With that term
the error is AX)').
y = - l - z 2 ; no ... 6 False 7 'Rue 9 h e (f' has 8 rieros, f" has 7)
11z = 3 i s m i n : f N ( 3 ) = 2 l S z = O n o t m u o r m i n ; z = ismin: f " ( i ) = 8 1
16 z = F
is m u : f " ( y ) = - A x = is min: fl'(Zp) =
?
1 7 Concave down for z < -1 and z > (inflection points)
19 z = 3 i s m a x : f U ( 3 ) = -4;z=2,4 areminbut ff'=O 21 f ( A z ) = f(-Ax) 25 1 + z - $
251-$ 271-iz-!z2 29Ekror?f"(z)~z SIErrorOAz+$f"'(z)(~z)~
37 & = 1.01016i; Ifr = .go966 '9 Inflection 4 1 18 vs. 17 43 Concave up; below
1 2 fl(x) = 3z2 - 120 = 0 at z = 0 and z = 4. The second derivative 6%- 12 is -12 then +12. So
x = 0 is a maximum point, x = 4 is a minimum point.
14 ft(z) = 1 1- 60z9~ ~ =0 ~at z = 0 and x = g.
The second derivative ll0x9 - 540z8 has f"(0) = 0 and
,
' I 60
and f ( ) = 110(:)~ - 540(:)' > 0. Then z = 0 is not a minimum or maximum (check function)
and x = bo ia a minimum point.
+
16 fl(z) = 1 2cos2z = 0 at cos2z = -?
or 22 = 120' or 240°(?f or 5).
At those two stationary points
fM(z)= -4 sin 22 is negative (f is a maximum) and then positive (f is a minimum).
+ +
1 8 fl(x) = cos x sec2 x gives f "(z) = - sin x 2 sec2 x tan z = sin x(-1 -). 2+ COB
The inflection
X
points are 0, r,2 r , . . where sin x = 0. (Note c0s3 z = 2 is impossible.) Then f" > 0 for 0 < x < i and
-
r <x < (concave up). In the other intervals f is concave down. Watch blow-up at 2 and ?f when
COB 2 = 0.
+ +
~ x gives f "(z) = - sin x - 3(sin 2)' sin z 6 sin z (cos x) = 5 sin x - 9 sin3x.
20 f' (z) = cos z 3(sin x ) cos
%
Inflection points where sin z = 0 (at 0, r, . .) and also where (sin 2)' = (an angle x in each quadrant).
Concavity is updown-up from 0 to r. Then down-updawn from r to 2r.
An odd function has f (0) = 0 and f (-Ax) = -f (Ax). The second difference is zero (so is f" at x = 0).
+
f (z) = tan x has fl(x) = sec2 z and f "(x) = 2 see2 x tan z. The quadratic approximation is 0 l ( x - 0)+
;O(X - o ) or~ just x.
f (x) = l+z+x2 has fl(x) = 1+2z and fl'(x) = 2. Substituting x = 0 gives 1 + 1 ( x - 0 ) + ~ 2 ( ~ - 0 )=~ l+x+x2
+ +
(of course). Substituting z = 1 gives 3 3(z - 1) ?2(x - I ) ~which , + +
is again 1 x x2.
y = (1 - x)-a has y' = -2(1- x ) - ~and y" = 6(1- x)-'. Substituting x = 0 gives 1,-2,6. The quadratic is
1-2x+i6x2=1-2x+sx2.
+
f (-Ax) w f (0) fl(0)(-AX) +~"(O)(-AX)~
The cubic term is f f '"(0) ( A Z ) ~When
. f = z3 and f' = 3z2 and f" = 6%and f "' = 6, the third degree
+ + +
approximation is f (Ax) = 0 0 0 AX AX)^ = (Ax)', exactly right.
%
At x = 0.1 the difference - (1.11)= .00111 . . comes from the omitted terms z3 z4 x5 + + + .
At x = 2 the difference is & + +
- (1 2 4) = -8. This is large because x = 2 is far from the basepoint
x = 0.
+
For cos z the approximat ion around z = 0 is cos 0 (sin 0)x +4 (- cos 0)x2 = 1- $ x2. At lo= & radians
this is 1- (&)2 = .999847691 : correct to 8 decimals. At 1radian this is 1 - j! = .5. The correct
cos 1= .54 is not so close because x = 1is far from the basepoint x = 0.
If f (z) is even then f '(0) = 0 and f (z) rr f (0) + $ f "(0)x2. The tangent line y = f (0) is horizontal
(linear term not present).
+ +
f(1) = 3, f (2) = 2 + 4 + 8 = 14, f (3) = 3 9 27 = 39. The second difference is 3g-$+3 = 14. The true
+
f" = 2 62 is also 14. The e m r involves f"" which in this example is zero.
y(x) is concave up if the line segment between any two points on its graph stays above the graph.
The position, slope, and bending of y = f (x) are decided by f (x),fl(x), and f''(x). If 1f (z)1 -+ oo as z -r a,
the line z = a is a vertical asymptote. If f (x) -, b for large x, then x = b is a horizontal asymptote. If
3.4 Graphs (page 119)
+
f (z) - mz -r b for large z, then y = mx b is a sloping asymptote. The asymptotes of y = z2/(x2 - 4) are
x = 2, x = -2, y = 1.This function is even because y(-z) = y(x). The function sin kz has period 2rlk.
Near a point where dy/dx = 0, the graph is extremely flat. For the model y = Cz2, z = .I gives y = .01C.
A box around the graph looks long and thin. We soom in to that box for another digit of z*. But solving
dy/dx = 0 is more accurate, because its graph crosses the r axb. The slope of dy/dz is dfy/dxl. Each
derivative is like an inftnite soom.
To move (a,b) to (0,0), shift the variables to X = x - a and Y = y - b. This centering transform changes
y = f (x) to Y = -b + +
f (X a). The original slope at (a, b) equals the new slope at (0,O). To stretch the axes
by c and d, set x = cX and y = dY. The mom transform changes Y = F ( X ) to y = dP(x/c). Slopes are
multiplied by d/c. Second derivatives are multiplied by d/e2.
2 (a) 30 dark limes in 6 seconds (b) 8 beats in 6 seconds is 80 beatslminute; (c) Rule: Heart rate = cycles per
&second interval times 10.
4 zn is odd or even according as n is odd or even. Horisontal asymptote y = O if n = 0, a sloping asymptote
y = x if n = 1, and a vertical asymptote x = 0 if n is negative.
6$f-, is odd: vertical asymptotes x = 2 and x = -2, sloping asymptote y = -x for large z (ignore 4).
8 % is n o t o d d o r even. Vertical asymptote x = -1 and a sloping asymptote y = x (ignore 3 and 1).
1 0 Periodic a n d even: Maximum f (0) = 3, two local maxima and minima before f (a) = -3.
12 & is even. Vertical asymptotes at all multiples x = na, except at z = 0 where f (0) = 1.
1 4 1- 22 is n o t o d d o r even. Vertical asymptote x = 1, sloping asymptote y = -22.
=. '
1 6 Z:'::': is periodic, not odd or even, vertical asymptotes when sin z = cos x at z = f n r . +
1 8 ~ - ~ i s d e f i n e d f o r z ~ O a n d h a s v e r t i c a ~ a s y m p t o t e a t z = O 2. O f ( ~ ) = ( = 1- ~ ) ( ~ - ~ )
22 f (z) = $ 22 + +3 24 f (z) = & +
26 For a sloping asymptote, degree of P = 1 (degree of Q).
28 &
.
goes to
.
+a, on both sides of the asymptote z = 1, but &
goes to -00 on the left side (x < 1).
30 (a) False: has no asymptotes (b) True: the second difference on page 108 is even (c) False: f (z) = 1 1 +
is not even but f" = 0 is even (d) False: *. never touches sero.
82 p(z) = x - a or (x - a)3 or . .. 54 d l z is near z u o for small z, equals 1 at z = 1 and \/i at x = 2.
JBTheasymptotesarez=-2andy=-1. S8Thisistheseconddifference~ar(sinx)"=-sinz.
40 (a) The asymptotes are y = 0 and z = -3.48. (b) The asymptotes are y = 1 and z = 1 (double root).
42 The exact solution is z* = 4 = 1.73205. The room should find those digits.
46 The slope is sero near z = 1.5. This is the first step in Newton's method. The second step uses y' and y" at
z = 1.5 to come very near x* = 4.
48 The exact x = fi solves ,,dh = &j or ( 2 ~=) 15 +
~ x2 or z2 = 5.
+
50 The odd functions f, (x) approximate a square wave of height 1- $ 5 - . . = f (Substitute x = .) . 5
3.5 Parabolas, Ellipses, and Hyperbolas (page 128)
They overshoot the jump from - 2 to at x = 0. The overshoot moves closer and closer to x = 0 but does
not become smaller (Gibbs phenomenon).
5 2 d3- is defined only for x 2 -$ (local maximum); minimum near x = .95; no inflection point.
5 4 Maxima at x = .295,1.72,4.09; minima a t x = 1.39,2.84,5.34; inflection points at .78,1.6,2.3,3.5,4.7,5.9.
i i.
5 6 x sin $ has derivative sin - $ cos Minima and maxima alternate at x = .22, .13, .09, .07, - . . (approaching
zero); inflection points .l6, .11,.08, - . . .
58 Inflection points and second derivatives are h a r d e r to compute than maximum points and first derivatives.
(In examples, derivatives seem easier than integrals. For numerical computation i t i s t h e o t h e r way:
derivatives are very sensitive, integrals are smooth.)
+ +
The graph of y = x2 22 5 is a p a r a b o l a . Its lowest point (the vertex) is (x, y) = (-1,4). Centering by
+
X = x 1 and Y = y - 4 moves the vertex to (0,O). The equation becomes Y = x2.
The focus of this centered
f
parabola is (0, ). All rays coming straight down are reflected to the focus.
+ +
The graph of x2 4y2 = 16 is an ellipse. Dividing by 16 leaves x2/a2 y2/b2 = 1 with a = 4 and
b = 2. The graph lies in the rectangle whose sides are x = f 4 , y = f 2. The area is 7rab = 87r. The foci are at
x = f c = fa. The sum of distances from the foci to a point on the ellipse is always 8. If we rescale to
X = x/4 and Y = y/2 the equation becomes x2+
y2 = 1and the graph becomes a circle.
All these curves are conic sections - the intersection of a p l a n e and a cone. A steep cutting angle yields a hy-
perbola. At the borderline angle we get a p a r a b o l a . The general equation is Ax2+Bxy + + +
cy2 Dx E y F = 0. +
+ +
If D = E = 0 the center of the graph is at (0,O). The equation Ax2 Bxy Cy2 = 1 gives an ellipse when
4 A C > B ~The + +
. graph of 4x2 5xy 6y2 = 1 is an ellipse.
1 9 ~ = a ~ ~ w i t h a > *0 ; ;y ~
= -=a x 2 + a x w i t h a > 0
21 $ + y 2 = 1 ; + + ( y - 1 ) 2 = 1 2 3 $ + $ = 1 ; ~ + ~ = 1 ; x 2 + y 2 = 2 5
2 5 Circle, hyperbola, ellipse, parabola 27 2
= -:; y = -3-x 4 +5 29 5 0 =, aL ( 45 - 5
4 9.4 9 4)
3 1 C i r c l e ; ( 3 , 1 ) ; 2 ; X = ~ , ~ = ~3 3 3 x t 2 + y t 2 = 2 35y2-~x2=1;~-~=1;y2-x2=5
'1
57 25 - E-
%
39 = 1 + + +
39 # - 4y 4, 2x2 122 18; -14, (-3,2), right-left
41 ~ = ( f + , ~ ) ; y = f i ! j 43 ( ~ + y + 1 ) ~ = 0
+ + + + + + +
46 (a2 - 1)x2 2abxy (b2 - l)y2 2acx 2bcy c2 = 0;4(a2 b2 - 1);if a2 b2 < 1 then B2 - 4AC < 0
2 y = 3x2 - 122 has 2 = 62 - 12 and Xmin = 2. At this minimum, 3x2 is half as large as 122. Then
3.5 Parabolas, Ellipses, and Hyperbolas (page 128)
+
the shift X = x - 2 and Y = y 12 centers the equation to Y = 3X2.
4 y = (x - has vertex at (1,O) and focus above it at (1, f ). Note a = 1.
6 42 = 9 has vertex at (0,O) and opens to the right. The focus is at (A,
0). (Note the coefficient of
y2 in x = fy2.)
+
8 x2 9 9 = 9 is the ellipse $ + $ = 1with vertices (f3,0) on the major axis and foci ( 4 4 , 0 ) .
10 $ - = 1 is a hyperbola centered at (0,l). It opens right and left with vertices at (&2,1) and foci
at (&&,I).
12 (y - I ) -~&
, = 1 is a hyperbola centered at (0,l): a = 1 and b = i. It opens up and down with vertices
at (0,2) and (0,O). The foci are (0,1 f dq).
14 xy = 0 gives the two lines x = 0 and y = 0, a degenerate hyperbola with vertices and foci all at (0,O).
16 y = x2 - x has vertex at (i, +
-f). To move the vertex to (0,O) set X = x - !j and Y = y f . Then Y = P.
18 The parabola y = 9 - x2 opens down with vertex at (0,9).
20 The path x = t, y = t - t2 starts with = 2 2
= 1 at t = 0 (45' angle). Then y, = f at t = The i.
path is the parabola y = x - x2.
22 When x2 = a2 - b2 the equation gives
f
+ $ = 1 or $-= 5 or y = b2
xi-.
This is the height above
the focus.
+ + + + +
24 Squaring d ( x - c ) ~ y2 = 20 - d ( x c ) ~ y2 yields z2 - 2cz c2 3 = 4a2 x2 2cx c2 y2 + + + +
-4ad(x + +
c)2 y2. This is -4cx - 4a2 = - 4 a d ( x + +
c ) ~ 9. Divide by -4 and square again:
+ + + + +
c2x2 2a2cz a' = a2(x2 2cx c2 y2). Set c2 = a2 - b2 and cancel a2x2 2a2cx a' to leave + +
+
-b2x2 = -a2 b2 a2& Divide by a2b2 to find $ +$
= 1 (the ellipse!).
+
26 At z = 0 the equation becomes (x - 2y)2 (y - 2 ~ = )1 or~ 6x2 - 8xy 5y2 = 1. Then+
B2 = 64 < 4AC = 100 (ellipse).
+
28 (a) The line touches the circle at (xo, yo) because xg yi= r2. It is tangent because the slope = -32Yo
is perpendicular to the slope of the radius (product of slopes is 1). (b) The derivative of the ellipse
equation is $ + 3 $ = 0, so $ = -$ = slope of the line. (This is implicit differentiation.)
30 PFl = P R and QFl = QR (because Fl and R are mirror images) so step 2 is the same as step 1. Since step
2 holds for every Q on the tangent line, P must be on the straight line from F2 to R. The intersection
angles of these lines are a = /3.
52 The square has side s if the point (5,
5) is on the ellipse. This requires &(:)2 + = 1 or
s2 = 4($ + & ) - I = area of square.
54 The Earth has a = 149,597,870 kilometers (Problem 19 on page 469 says 1.5 10' km).
The eccentricity e = is 0.167 (or .02 on page 356). Then c = 2.5. lo6 and b = d=.
This b is very near a; our orbit is nearly a circle. Use Ma - &
M a - 2 . 10' km.
x,+~ = x i describes an iteration. After one step xl = xo. 3 After two steps z2 = F ( z l ) = xl3 = xo. 9 If
it happens that input = output, or z* = F(x*),then z* is a fixed point. F = x3 has t h r e e fixed points, at
z* = 0,1 and - 1.Starting near a fixed point, the x, will converge to it if IFt(x*)I< 1. That is because z,+l -
z* = F(z,) - F(z*) = F1(x*)(xn - x*).The point is called a t t r a c t i n g . The z, are repelled if IFt(x*)l> 1.For
F = z3 the fixed points have F' = 0 o r 3. The cobweb goes from (zo,zo) to (zo,xl ) to (xl ,x l ) and converges
to (z*, x*) = (0,O). This is an intersection of y = x3 and y = x, and it is super-attracting because F' = 0.
f (z) = 0 can be solved iteratively by z,+l = x, - cf (z,), in which case Ft(x*) = 1- cfl(x*). Subtracting
z* = z* - cf (z*), the error equation is x,+ 1 - x* FJ m(xn - x*). The multiplier is m = 1- cf' (x*).The
errors approach zero if -1 < m < 1. The choice c, = l/f'(x*) produces Newton's method. The choice c = 1 is
"successive s u b s t i t u t i o n " and c = l / f t (xo) is modified Newton. Convergence to z* is not certain.
We have three ways to study iterations x,+ 1 = F (x,) : (1) compute 21, x2, - - . from different so
(2) find the fixed points x* and test IdFldxl < 1 (3) draw cobwebs.
For a 5 3,2,+1 = asn - az: converges to a single fixed point. After a = 3 the limit is a Zcycle, which
means that the 8's alternate between two values. Later the limit is a Cantor set, which is a one-dimensional
example of a fractal. The Cantor set is self-similar.
1 zn+l = 2, - *
32; = 9+ & 5 XI = Z O ; 2 1 is not defined
zk -7
(00) 7 X* = 2; blows up; z* = 2 if xo < 3
xo < to X* = 0 ; zo > t o z* = 1 2 1 z,+l = x, - + 25 x4 = cot s = oo; x3 = cot = cot
s+l - 2-1 J X , + ~~ ) ~~ - 1
f ( z ) = 2+1 has f t ( x ) = +
- (2:,), SO Newton's formula is xn+l = x , -
2 z,+l -
=* '-
z, - ",
2' - 1
. The fixed points of this F satisfy z* = x* - which gives z* = 1 and z* = -1.
The derivatives F' = 1 - z* are 0 and -2. So the sequence approaches x* = 1, the correct zero of f ( z ) .
f ( z ) = z1l3 has f' ( z ) = &SO Newton's formula is z,+l = z , - 3 ~ 2 ' ~ ( x ! , ' ~
=) -22,. The graph of z1I3 is
vertical a t x = 0 ; the tangent line a t any z hits the axis a t -2s.
f ( z ) = x3 - 3 z - 1 = 0 : roots near 1.9, -.5, -1.6
For any zo, the new z1 is on the right side of the root. Then z2, xs, . . . approach steadily from the right.
Newton's method for f ( x ) = z 4 - 100 approaches z* = fiif zo > 0 and z* = -a
if zo < 0.
+
In this case the error a t step n 1 equals &
times (error a t step n ) 2 . In Problem 9 the multiplier is &
and convergence is quicker. Note to instructors: The multiplier is (this is + F " ( X * ) : see
Problem 31 of Section 3.8).
z3 - z = 0 gives X* = 1,0, and - 1. Newton's method has z l = xo - = -&.. & This equals -xo
U U
(producing a cycle) if zo = f 6
Between these limits we have lxll < lxol and Newton converges to
z* = 0. Between lzol = and lxol = the convergence to 1 or -1 looks complicated. For zo > fl
there is convergence t o x* = 1.
Between -5 5
and the graph of - tan z decreases from oo to -00. It crosses the 45' line once (at x = 0).
In each successive interval of length K , the same is true: one solution to z = - tan z in each interval.
Roots X* = 0 and z* = 2.03.
Roots a t x* = -1.3 and z* = .526.
(a) Fkom 1 - 2 ~ , + =~ (1- 2 ~= 1~- 42,) 4x:, +
~ cancel the 1's and divide by -2. Then x,+~ = 22, - 2x2.
. if I1 - 2x01 > 1, repeated squaring blows up.
(b) Every step squares 1 - 22, t o find the next 1 - 2 ~ , + ~So
If 11- 2x01 < 1, then repeated squaring gives 1- 22, -+ 0. Here I1- 2x01 < 1 puts 2x0 between 0 and 2.
Multiply zn+l = 22, - ax: by a and subtract from 1 t o find 1 - ax,+l = 1 - 2axn a2zZ, = ( 1 - +
At each step 1 - ax, is squared to find the next 1- az,+l. Then 1 - a x , -t O (or x , -+ if 11 - azol < 1. i)
This puts 1 - ax0 between -1 and 1: then 0 < so < 2/a.
Roots a t x* = -2.11485 and x* = .25410 and z* = 1.86081.
3.8 The Mean Value Theonm and l'H6pital's Rule (page 152)
3.8 The Mean Value Theorem and I'H6pital's Rule (page 152)
The Mean Value Theorem equates the average slope A f /Ax over an interval [a,b] to the slope df /dz at an
unknown point. The statement is A f / A x = fl(x)for some point a < c < b. It requires f (z)to be continuous
on the closed interval [a,b], with a derivative on the open interval (a, b). Roue's theorem is the special case
when f (a) = f (b) = 0, and the point c satisfies f'(c) = 0. The proof chooses c as the point where f reaches its
maximum o r minimum.
Consequences of the Mean Value Theorem include: If f'(z) = 0 everywhere in an interval then f(z) =
+
constant. The prediction f (z) = f (a) f'(c)(z - a) is exact for some c between a and z. The quadratic
+ + +
prediction f(z) = f(a) ft(z)(z - a) ~ f t t ( c ) (-z a)2 is exact for another c. The error in f(a) f'(a)(z - a)
is less than $M(z - where M is the maximum of 1fl1I.
A chief consequence is 1'Hbpital's Rule, which applies when f (z) and g(z) -+ 0 as z -+ a. In that case the
limit of f (z)/g(x) equals the limit of ft(x)/g' (x), provided this limit exists. Normally this limit is f' (a)/g' (a).
If this is also 010, go on to the limit of f" (x)/g" (x).
1 4 limx,s i.=
*
at every point c. By the MVT f (z) must have the same value at every pair of points a and b. By
trigonometry csc2 z - cos2 z = sin x - sin z = sin z = 1 at all points.
(This is not a case for 1'H6pital9s Rule! It is just limit of f (z) divided by limit of g(z).)
f (I-cos 2)-"'sin z
16 L = lim,,o = (by l'H6pital's Rule) lim,,o 1 or L = lim 2,/",i-":,s
.. This is again
i. But we can multiply by -& -+ 1to reach limx,o 2J& which is &. ~ h u Ls = & and L = -1
JZ'
(Note: The knowledge that 1- cos z M $
also gives lirn = lirn s = 1
1 8 limxdl 3sln x = sln 1 = 0 (not an application of 1'H8pitalys Rule).
l+~)~-l-nx n(l+z)"-'-n - n(n- ~ ) ( l + z ) ~=
-'n ( y ) .
2 0 limx,o I 22 = lim,,o 22 - (1'HGpital again) l i m x ~ o 2 -
lim,,~ m-
x
J1-2
= lim,,o -+-
1
-
1
'Ji--r - 1
-2+$=1.
The steps when f -+ co and g -+ co are L = lirn $ = lim 3
= (now comes l t H 6 p i t a l f o r o)
O
lirn
9-
-
1+-A
= which approaches 2 = - 1.
28 w - 1 i n z = 1
- s i n z c o a x c o s x ' The limit as x -+ O is f = 1.ok to use l'HBpitalys Rule for
s : L = limx,o =
cotz
lim,,o - C S C x cot x = lim & = 1 L. This gives L2 = 1 but does not eliminate L = -1; add the fact that
- CSC' z ,,,,
csc z and cot z have the same sign near x = 0.
SOMeanValue Theorem: f ( x ) - f ( y ) = f1(c)(z-y). Therefore I f ( z ) - f ( y ) [ = Ift(c)IIx- yl 5 15- yl since we
are given that I f' l 5 1 at all points. Geometric interpretation: If the tangent slope stays between -1 and
1, so does the slope of any secant line.
32 No: The converse of Rolle's theorem is false. The function f (x) = x3 has f ' = 0 at x = O (horizontal
tangent). But there are no two points where f (a) = f (6) (no horizontal secant line).
x 2 COB $
54li1n,,~$$ =lim,,o = limx,o (x cos ): = 0 because always lz cos f 1 5 1x1. However
= sin $+22 cos $
has no limit because sin $ oscillates as x -+ 0 (its graph is in Section 2.7).
gt(z) 1
3 6 If you travel 3000 miles in 100 hours then at some moment your speed is 3 0 miles p e r hour.
3 8 Mean Value Theorem: f (b) - f (a) = ft(c)( b - a ) is positive if ft(c) > 0 and b > a. Therefore f (b) - f (a) > 0
and f (b) > f (a). A function with positive slope is increasing (as stated without proof in Section 3.2).
CHAPTER 4 DERIVATIVES BY THE CHAIN RULE
z = f (g(z)) comes from z = f (y) and y = g(x). At z = 2 the chain (z2 - 1)3equals 3' = 27. Its inside
function is y = x2 - 1, its outside function is z = ys. Then dzldx equals Sy2dy/dx. The first factor is
evaluated at y = x2 - 1(not at y = z) . For z = sin(z4 - 1)the derivative is 4uS cos(x4 - 1).The triple chain
z = cos(x + 1)2has a shift and a s q u a r e and a cosine. Then dz/dz = 2 cos(x I)(- sin(x 1)). + +
The proof of the chain rule begins with Az/Az = (Az/Ay)(Ay/Ax) and ends with dz/dx= (dz/dy) (dy/dx).
Changing letters, y = cosu(z) has dy/dz = -sin u ( x ) e . The power rule for y = [u(z)ln is the chain rule
dy/dz = nun-' du. The slope of 5g(z) is 5gt(x) and the slope of g(5z) is 5gt(5x). When f = cosine and g =
F
sine and z = 0, t e numbers f (g(z)) and g(f (2)) and f (z)g(z) are 1 and sin 1and 0.
1z=y3,y=z2-3,z'=6z(z2-3)2 ~z=co~y,~=z~,z'=-3z~sinz~
5 z = f i , y = sinz, z' = c o s z / 2 d Z E 7 z = tany+ (l/tanz), y = l/z,zt = (9)sec2($) - ( t a n ~ ) - ~ s e c ' z
9z=cosy,y=z2+z+1,z'=-(2z+1)sin(z2+z+1) 1117cos17z lJsin(cosz)sinz
15z2cosz+2zsinz 1 7 ( c o s d ~ ) ~ ( ~ + 1 ) - ' 1 1~9 ~ ( 1 + s i n ~ ) - ~ ~ ~ ( c o s2z1)c 0 s ( & ) ( ; ~ 7 ~ )
2 3 8 ~ ' = 2 ( ~ ~ ) ~ ( 2 ~ ~ ) (225 z2 () z + 1 ) + ~ 0 s ( z + 7 r ) = 2 z + 2 - ~ o s z
27 (z2+ 1)2+ 1; sin U from 0 to sin 1; U(sin z) is 1and 0 with period 27r; R from 0 to z; R(sin z) is half-waves.
29 g(z) = z + 2, h(z) = z2 + 2; k(z) = 3 31 f t ( f (2))ft(z); no; ( - ~ / ( l / z ) ~ ) ( - l / z ~=) 1and f (f =z (2))
Js!j(iz+8)+8;$~+14;& 35f(g(z))=z,g(f(y))=y
37 f (g(z)) = & , d f (4)= 1- f ' f (f (z)) = Z = 9(9(z)), d f ( g ( 4 ) ) = = f (9(f (4))
39 f (y) = y - 1,g(z) = 1 4 3 2 cos(z2 + 1)- 4z2 sin(z2 + 1); -(z2 - I)-~/'; -(cos + (sin &)/4z3I2
40 False (The chain rule produces -1 : so derivatives of even functions are odd functions)
False (The derivative of f (z) = z is f' (z) = 1)False (The derivative of f ( l l x ) is f ' ( l l z ) times - 1/z2)
T N e (The factor from the chain rule is 1) False (see equation (8)).
42 Fkom z = go up to y = sin :. Then go across to the parabola 2 = y2. Read off z = d.;lf on the
horizontal z axis.
44 This is the chain rule applied to (a function of y). Its z derivative is its y derivative ($) times 2.
If z = y2 and y = z3 then 2
= 2y and $2 = 2(3x2). Check another way: = 2z3 and d(&)
d x dy
= 6 x2 .
46 2= (3u2)(3%') = 9z8 48 = 2 &$ so g = 22zf'(zZ) + 2f(z)%
52 2 = -nu(t)-n-l% 54 $ = - 5: 56 cos(sin z) cos z
58 (a) 53 (sum rule for derivatives) (b) 60 (chain rule)
60 Note that G' = cos(sin z) cos z and G" = - cos(sin z) sin z - sin(sin z)cos2 z. We were told that
H(z) = cos(cos z) should be included too.
+
For z3 y3 = 2 the derivative dy/dz comes from implicit differentiation. We don't have to solve for y.
+ 2
Term by term the derivative is 3z2 sy2 = 0. Solving for dyldz gives -x2 /y2. At z = y = 1 this slope is
-1. The equation of the tangent line is y - 1 = -l(x - 1).
In related rates, we are given dg/dt and we want df /dt. We need a relation between f and g. If f = g2, then
+
(dfldt) = 2g(dg/dt). If f 2 92 = 1, then dfldt = - f % . If the sides of a cube grow by dsldt = 2, then its
volume grows by dV/dt = 3s2(2) = 6s2. To find a number (8 is wrong), you also need to know s.
.- --
+ 2 y g = 0 gives 2 = 1at (1,l); 22 + 2(y - 2 ) g = 0 also gives 2 = 1.
1 2 2(x - 2)
14 2 + 2 y 9 + ~ ( 2 ~ 9= - j-
= 0)yields
4.3 Inverse f i n ctions and Their Derivatives (page 170)
Distance to you is d m , rate of change is 2 with 2 = 560. (a) Distance = 16 and z = 8&
and rate is e ( 5 6 0 ) = 2 8 0 6 ; (b) z = 8 and rate is L(560)
,/iGiP = 280fi; (c) z = 0 and rate is zero.
10c(t - 3) = 8t divided by c(t - 3) = 4 gives 10 = 2t. So t = 5 and c = 2. The z and y distances between
ball and receiver are 2t - 10 and 12t - 60. The derivative of d ( 2 t - 10)2 + ( 1 2 - 60)2
= , m i l t - 51 is -,hz.
Volume = $xr3 has = 47rr2$. If this equals twice the surface area 4ar2 (with minus for evaporation)
than 2 = -2.
$ = 4 a radianslsecond; 0 = 22% - 6 cos 9 2 + 6 2 sin 9 g; at 9 = :, z = 3& and 6fig + 18&%
gives 2 = - 1 2 ~ ;at B = R, z = 0 and 2 = 0.
The functions g (x) = z - 4 and f (y) = y + 4 are inverse functions, because f (g(2)) = x. Also g (f (y)) = y.
The notation is f = g-' and g = f - l . The composition o f f a n d f - l is the identity function. By definition
z = gml(y) if and only if y = g(x). When y is in the range of g, it is in the domain of g-'. Similarly z is in
the domain of g when it is in the range of g-'. If g has an inverse then g(zl)#g(zz) at any two points. The
function g must be steadily increasing or steadily decreasing.
The chain rule applied to f (g(z)) = z gives (df/dy)(dg/dx) = 1. The slope of g-' times the slope of g
equals 1. More directly dxldy = l / ( d y / d x ) . For y = 22 1 and z = + i(y
- I), the slopes are dyldx = 2 and
1
dzldy = 3 . For y = z2 and z = a,
the slopes are dyldz = 2 x and dz/dy = 1/2&. Substituting x2 for y
gives d z l d y = 1/2x. Then (dz/dy)(dy/dx) = 1.
The graph of y = g(x) is also the graph of z = g-l(y), but with x across and y up. For an ordinary graph
of g-', take the reflection in the line y = z. If (3,8) is on the graph of g, then its mirror image (8,J) is on the
graph of g-'. Those particular points satisfy 8 = Z3 and 3 = log2 8.
The inverse of the chain z = h(g(z)) is the chain z = g-l(h-l(z)). If g(z) = 32 and h(y) = y3 then
z = (Sx)'= 27x3. Its inverse is z = ;z1/', which is the composition of g-l(y) = and h-'(z) = z1l3. iy
1s = Y+6
3
3x= ( z unrestricted -+ no inverse) 5 z = - s+ I 7 z = (1 + y) ' I 3
9 ( z unrestricted + no inverse) 11y = 1
2-0 132<f-'(z)<3 15fgoesupanddown
1 7 f(z)g(z) and &J 19 m # O ; m > O ; Iml >1 21 2
= 5 z 4 , z = ky-'/'
23!&=3z2.dl=$(l+y)-2/3
dz ' d!/ 25$=
(=-')a
-1 d l = -1
ds (u-1)) 2 7 y; y2 C +
2 9 f (g(z)) = -l/3z3; g-'(y) = $;g(ge'(z)) =z 39 2/& 4 1 I / ~ C O9S
4.4 In verses of Digonometric h n c tions (page 175)
z = SEE
A 4 z = &(f-' matches f ) 6 no inverse
2
{ Y
= -&;
YLO
2 = -i(z
10 z = Y5
- 1)2 (or -*).( Y 1)
f-' does not exist because f (3) is the same as f (5).
No two x's give the same y. 18 y = 5 and y = 2 - z (functions of z + y and zy lead t o suitable f )
The inverse of a piecewise linear function is piecewise l i n e a r (if the inverse exists).
! & = - ~ . d " = - l-
(-1)" d~ y2
- + - 1 ) 24 2
= - A . & - 1 -413
,r 9 dy - - %Y
26 = a*. 2
d~ - a d - b c
(cx+d)' ' d~ - ((ey-a)2 '
!&= y. 30 jumps a t 0, yl ,y2 to heights z l ,z2, z3; a piecewise const ant function has no inverse.
Hyperbola centered a t (-1,O) : shift the standard hyperbola zy = 1.
y = -3s for z < 0;y = -z for z 2 0. 36 The graph is the first quarter of the unit circle.
The graph starts a t (0,l) and increases with vertical asymptote a t z = 1.
1 = see2 2% so = cos2 z = 1
dY
42 2
= 1- cos z = 0 so $ = CQ. (The derivative does not exist.)
F i r s t p r o o f Suppose y = f (z). We are given that y > x. This is the same as y > f
S e c o n d p r o o f The graph of f (z) is above the 45' line, because f (z) > x. The mirror image is below
the 45' line so f-'(y) < y.
g(z) = 2 - 4, f ( y ) = 5y, g - q Y ) = y + 4, f - l ( z ) = ;,x = g
l.+4.
g(z) = z + 6 , f ( ~ =
) y3, g-'(y) = y - 6, f-'(2) = G;x
= G-6
g(z) = $ z + 4, f (y) = g(y), g-'(y) = 2y - 8, f-'(2) = 9 - ' ( 2 ) ; ~ = 2(2z - 8) - 8 = 4 z - 24.
z* = f -'(o)
f - ' ( ~ ) EJ f-'(y) + (%)(0 - y) is the same as z * rr z + &(o - f (z)), which gives Newton's method.
dY
= f -l(y) + ys -' s.
- F'( f (y))
t o f ( f q l ( y ) ) = y. This leaves the first term
The second term cancels the third because F' (f is equal
= f-'(y). G i s t h e a n t i d e r i v a t i v e o f f-' if F' = f .
To maximize yz - F ( z ) set the z derivative to zero: y = = f (z) or z = f-'(y). Substitute this
z into zy - F ( x ) : the maximum value is exactly G(y) from Problem 56. Now maximize zy - G(y).
The y derivative gives z = -'
or by Problem 56 z = f (Y). Substitute y = f (z) into z y - G(Y)
to find that the maximum value is z f ( z ) - G ( f ( z ) ) = z f ( z ) - [ f ( z ) z - F ( f - ' ( f ( z ) ) ] = F ( z ) .
N o t e : This is the L e g e n d r e t r a n s f o r m between F ( z ) and G(y) - important but not well known.
Since is increasing (then f -'exists), the function F ( x ) is convex (concave up). So is G(y).
The relation z = sin-' y means that y is the sine of x. Thus z is the angle whose sine is y. The number y
lies between -1 and 1. The angle z lies between -7r/2 and n/2. (If we want the inverse to exist, there cannot
be two angles with the same sine.) The cosine of the angle sin-' y is 43. The derivative of z = sin-' y is
4.4 Inverses of 'Rigonometric finctions (page 175)
The relation z = cos-' y means that y equals cos x. Again the number y lies between -1 and 1. This time
the angle z lies between 0 and a (so that each y comes from only one angle z). The sum sin-' +cos-' y = r/2.
(The angles are called complementary, and they add to a right angle.) Therefore the derivative of x = cos- y
is dzldy = -1143, the same as for sin-' y except for a minus sign.
The relation x = tan-' y means that y = t a n x. The number y lies between -oo and oo. The angle z lies
be tween -a/2 and a/2. The derivative is dzldy = 1/(1+ y2). Since tan-' y cot-' y = a/2, the derivative of +
cot-' y is the same except for a minus sign.
The relation z = sec-' y means that y = sec x. The number y never lies between -1 and 1. The angle x lies
between O and a, but never at x = r/2. The derivative of x = sec-' y is dxldy = 1/ I y I JyG.
1O , t , O
gsinx=Jl-yal-ya;Jwandl
lSy=O:l,-1,1;y=
21 2
- 2sin-I z
S ;,0, :
1:0,0,?
5 a is outside
sin-'
l l
15F;F;T;T; F;F
25 d~ =
1-5, $1
1
~dy c
17 &
o
dz
= -- 1
7 y = -&/2
s ~
,/c? l 9 2 = 3
= ~
dSJ
and &/2
+
-
~
cos 2
-
-
J
,* - -
The problem of summation is to add ul + . . + un. It is solved if we find f's such that uj = f j - fj-l. Then
UI + . ..+ un equals fn - fo. The cancellation in (fl - fo) + (fa - fl) + . . .+ (fn - fn- 1) leaves only fn and -foe
Taking sums is the reverse (or inverse) of taking differences.
The differences between 0, 1, 4, 9 are ul, uz,u3 = l , S , S . For f j = j2 the difference between flo and f9 is
ulo = 19. From this pattern 1 + + + +
3 5 .. 19 equals 100.
For functions, finding the integral is the reverse of fhding the derivative. If the derivative of f (x) is u(x),
then the integral of u(x) is f (x). If u(x) = 10%then f (x) = 5x2. This is the a r e a of a triangle with base x and
height 102.
Integrals begin with sums. The triangle under v = 103: out to x = 4 has area 80. It is approximated by
four rectangles of heights 10, 20, 30, 40 and area 100. It is better approximated by eight rectangles of heights
5'10, . . . ' 4 0 and area 90. For n rectangles covering the triangle the area is the sum of ($ & + + +
. . 40) =
80 + g. As n -+ oo this sum should approach the number 80. That is the integral of u = 10%from 0 to /.
Reason: Area under inverse function equals area above original function (provided f (0) = 0).
2 6 A f i ~5.3313556
5.2 Antiderivatives
Integration yields the a r e a under a curve y = v(z). It starts from rectangles with the base A x and heights
v(x) and areas v ( x ) A x . As A x -+ 0 the area vlAx + +
- - vnAz becomes the i n t e g r a l of v(x). The symbol for
the indefinite integral of v(z) is $ v ( x ) d x .
The problem of integration is solved if we find f (z) such that = v ( x ) . Then f is the a n t i d e r i v a t i v e of
v , and :$
v(x)dz equals f(6) minus f(2). The limits of integration are 2 and 6. This is a definite integral,
which is a number and not a function f (z) .
The example v(z) = z has f (z) = i x 2 . It also has f (x) = 1 2 ax + 1.The area under v(z) from 2 to 6 is 16.
1 9.
The constant is canceled in computing the difference f(6) minus f(2). If v(z) = z8 then f (z) = gx
+
The sum q + - - - vn = fn - fo leads t o the Fundamental Theorem $a v(z)dz = f ( b ) - f (a). The indefinite
b
integral is f (x) and the d e f i n i t e integral is f (b) - f (a). Finding the area under the v-graph is the opposite of
finding the s l o p e of the f -graph.
The Greek letter C indicates summation. In C; u j the dummy variable is j. The limits are j = 1 and j
1
= n, so the first term is v l and the last term is vn. When u j = j this sum equals %n(n +
1).For n = 100 the
1 1
leading term is z1002 = 6000. The correction term is % n= 60. The leading term equals the integral of u = z
from 0 to 100, which is written $ioox dx. The sum is the total area of 100 rectangles. The correction term is
the area between the sloping line and the rectangles.
slowly.
17g+~;fp-f8-fl+fo 19fl=l;n~+(2n+1)=(n+l)~
+ + + +
21 a b c = 1,2a 4b 8c = 5,3a 9b 2 7 ~ + + = 14; sum of squares 2S S400 = 80200; Elm = ,0025 = ;f
25 Sloo,lls H 350, Eloo,lls H .00587; Sloo,s = 25502500, E1oO,s= .0201 27 ul and y have the same sign
10 The first sum is close to e-l = .36788; the second is close to e = 2.71828; the product is extremely near 1.
12 Choose all a's and b's equal to 1. Then n2 # n.
I 4 fn - fo and fls - f3 (by telescoping: the other terms cancel).
I 6 C;=l ui = ~ zeoi2
~ ; z i and ~ + =~ (i -
5.4 Indefinite Integrals a d Substitutions (page 200)
Finding integrals by substitution is the reverse of the chain rule. The derivative of (sin x ) is~ 3(sin ~ ) ~ cx.o s
Therefore the antiderivative of 3(sin ~ ) ~ cxoiss (sin x)'. To compute 1(1+
sin x ) cos
~ x dx, substitute u =
+
1 sin x. Then du/dz = cos x so substitute du = COB x dx. In terms of u the integral is u2 du = 3u 1 3.
Returning to z gives the final answer.
+
The best substitutions for / tan(x+3) sec2(z+3)dz and /(x2+ l)lOzdx are u = t a n ( x 3) and u = x2 + 1.
+
Then du = sec2 (x 3)dx and 2x dx. The answers are &tan2(x 3) and + h(x2+
l ) l l . The antiderivative
of v du/dx is P2. $ 2%dx/(l + z2) leads to $9,
which we don't yet know. The integral d x / ( l + z2) is
known immediately as tan- lx.
1 $ ( 2 + ~ ) ~ / ~ + 3 (~x+l)"+'/(n+1)+C(n#-1) 5 & ( ~ ~ + 1 ) ~ + C7 - ! c o s 4 z + ~
9 -f c o s 4 2 x + ~ 11s i n - ' t + C 13 $ ( ~ + t ~ ) (~1 /+ t~2 -) l 1 2 + C 15 2&+x+C
+
1 7 sec x C 19 - cos x C+ +
2 1 $z3 $x312 2 3 - $(I - 2x)3/2 25 y = 6
27ix2 29asinz+bcosx 33F;F;F;F 35f(x-1);2f(:)
37 x - tan-' x 39 j i d u + +
4 1 4.9t2 Clt C2 + +
4 3 f (t 3); f (t) 3t; 3f (t); $ f (3t)
--
1 4 Write u = 1-t2 and du = -2t dt to give j(l-u)Ji;$ = - $ u ~ / ~~+ u ~ / ~= +-$(l-t2)3/2+
c i(l-t2)5/2+~
1 6 The integral of x112 + x2 is tx3/2 + ?x3 + C.
1 8 Set u = tan x and du = see2 z dx. The integral of u2du is $ tan3 x C. +
20 Write sin3 x as (1- cos2 x) sin z. The integrals of - cos2 z sin x and sin x give $ cos3 x - cos x C. +
22 Substitute y = cxn to find ncxn-' = ( C X ~ Match) ~ . exponents: n - 1 = 2n or n = -1. Match
coefficients: nc = c2 or c = n = -1. Answer y = - Itx.
24 y = + C 26 dy/dx = x/y gives y dy = x dx or y2 = x2 C or y = d m . +
5.5 The Definite Integral (page 205)
2s y = I +
#c1z4 c2z3 c3t2 c4+ cs + + + +
SO y = gi20 ~comes
s from y-'lady = ~ ' / ~ dorz2y112= 3%312(+~) 82 = z114 gives y = tz5I4 C 2 +
S4 (a) False: The derivative of j2(z) is f (z)$ (b) !the: The chain rule gives f (u(z))= %(u(z)) 2
times $ (c) False: These are h e operutions not inverse functions and (d) is The.
SBif(2z-1)+C;if(z2)+C S8$(z4+2z2+1)dz=~P+~z3+z+C.
4 0 ~ s e u = l + z ~ a n d d u = 2 z d z a n d z ~ = ~ - l . ~ h e n $ ~ - $ $ i s - ~ + ~ & - ~ ~ = ~ + ~ +
42 y = CIP +C2z2+C3z C4. +
If $'u(z)dz = f (z) + C, the constant C equals -f(a). Then at z = a the integral is zero. At z = b the
integral becomes f (b) - f (a). The notation f (z)]: means f (b) - f (a). Thus cos z]: equals -2. Also [cosz 31: +
equals -2, which shows why the antiderivative includes an arbitrary constant. Substituting u = 22 - 1 changes
JQ dz into Jl5 i 6 du (with limits on u).
The integral $ ' u(z)dz can be defined for any continuous function u(z), even if we can't find a simple
antiderivative. First the meshpoints 21, q ,. divide [a,b] into subinternab of length Axr, = xk - xk-1. The
upper rectangle with base Axk has height MI = maximum of v(x) in interval k.
The upper sum S is equal to AxlM1 + +
Ax2M2 .. . The lower sum 3 is Axlml Ax2m2 . . . The + +
area is between s and S. As more meshpoints are added, S decreases and 3 increases. If S and s approach
the same limit, that defines the integral. The intermediate sums S*,named after Riemann, use rectangles of
height u(z;). Here z; is any point between xk- 1and xk, and S*= Axkv(xi) approaches the area.
( 1 - ~ ) ~ u which
~ d uis lo
the same as t h e original (no progress). Compute by writing out z3(1- 32 3x2 - z3) and integrating +
each term: [:z4- % z 6 +gz6 - ;z7]i
=& 4 - % +3
g - 7.
1
20 sin 27rz has maximum MI = 1 and minimum m l = 0 in the interval to z = $; then Ma = 0 and m2 = -1 in
the interval t o z = 1. Thus S = 1 (1)and s = 1 (-1).
22 Maximum of z in the four intervals is: Mk = - $ ,0, $, 1. Minimum is rnr = -1, - $ ,0, $ . Then
S = $ ( - $ + 0 + $ + 1 ) = %1 a n d s = $ ( - I - - 2
1
+ o + $ ) = - 1
2'
24 The exact area is :$ z3dz = $1; = 4. Then S - 4 is less than S - s = Z3Ax. So S < 4.001 if z3Az < .001
or A x < &(.001)= .000125.
26 All midpoints of the intervals with A x = are fractions. So V (x*) = 1 a t these midpoints x*.
The upper Riemann sum S* is the sum of Ax's times 1 = length of interval of integration. This stays
the same as n -, oo but other choices of z* give S* = 0 : not Riemann integrable.
28 (Correction: Change v to M.)The graph of M ( z ) is above horizontal rectangles of total area
(+)(+) + (+)(a) +
---=
A
= 4. 5
With A x = the M's are 0, $, 1 with S = 1) = 5(0 + + i.
The m's are 0,0,4 with s = $(o 0 + + $) = i.
SO Check f(1) = $
: v ( z ) d z = 0. Check & v(x)dz = & (- $: v ( z ) d x )= -v(x). Then f ( z ) is correct.
z is chosen from 1,3,5,7 with equal probabilities f , its expected value (average) is 4. The expected value
If
of z2 is 21. I f z is chosen from 1,2, . -., 8 with probabilities k,
its expected value is 4.5. If x is chosen from
1 5 z 5 9, the chance of hitting an integer is zero. The chance of falling between x and z+ dz is p(x)dx = Qdx.
9
The expected value E ( z ) is the integral dx. It equals 5.
1
1~=~~~lx4dz=~equalsc4atc=fS
( ~~ ) =1 ~ 4$ ~ ~ ~ ~ ~ z d x = ~ e q u a l s c o s ~ c a t c = ~
5 =$
: 3= equals 3 at c = fi 7 $: v ( z )dz 9 False, take v ( x ) < 0
11 True; $ v(z)dz $ - $ + 4
: v ( z ) d z = :$ v ( z ) d x 13 False; when v ( z ) = z2 the function z2 -
b
5 is even
15False;takev(z)=1;factor~ismissing 17~,=&$~v(z)dz 19Oandf
5.6 Properties of the Integral and Average V h e (page 212)
uave = ? I-,
1
P d z = 0 which equals cS at c = 0.
uave = lo( 6 dz = 8 4 ~ which 1 ~ equals 4 at c = &4' = w. 16
uave = & $:,(sin z)Odz = 0 (odd function over symmetric interval -r to r). This equals (sin c)' at c = -r
and 0 and A.
5
8 2 $' x dx = z2]t = 24. Remember to reverse sign in the integral from 5 to 1.
1 0 False. The interval keeps length 3 but if u(z) = z the integral changes.
1 2 False: This is the average value of g.
1 4 False: -1 5 sin z 5 1 but the derivatives do not satisfy 0 5 cos z 0.<
+
16 (a) False: strictly speaking the antiderivatives of z2 are i z 3 C; this is odd only when C = 0
(b) False: ( z ) ~is even.
1 8 The average of g is = -1.
20 Property 6 proves both (a) and (b) because u(x) 5 lu(z)l and also -u(z) S Iu(x)l. So their integrals
maintain these inequalities.
22 If u is increasing then u(t) 5 u(z) when t 5 z. Apply Property 6: 1 ; u(t)dt 5 u(z) dt. N o t e v(x) is
constant in t h e last integral, which is t u(z)]$ = zu(z).
24 Suppose u, < c for n larger than N. ( N is now Fed.) Then the average vl+,+vs is lass than +'-+v~+(n-N'L.
n
As n -r oo this approaches = c. So the average goes below any c and must approach rero.
;
Yave = f 1 4-dz 3
= (area ) = 4%. A uniform distribution of Q along the base is different from a
uniform distribution of P along the semicircle.
This needle falls across a crack when y < zcos B (change the 1's in the Buffon needle figure to x's).
Following Problem 29, the shaded region lies under y = z cos t9 and under y = 1.
Keeping z < 1 (ahorter needles only) the area is COB 8 dB = z sin B];l2 = z.
This fraction $ = $ of the total area is the probability of falling across a crack.
S2 The square has area 1. The area under y = fi is $1,' fi dz = g.
+
S4 When z is replaced by -2, the function ?(u(z) u(-2)) is unchanged (even). The function
i(u(z) - u(- z)) becomes ;(u(-z) - u(z)) so signs are reversed (odd function).
36 fJ(-Z) = t(-x+h'-f
h
( -xl = (when f is even) lim i(x-hi-'lxl= -f'(x). Thus f' is odd.
58 Average sire is g.
The chance of an individual belonging to group 1is
X
2.
The expected siae is sum
of sire times probability: E(z) = C $. This exceeds f by the Schwars inequality:
(1x1 + 12,)~ 5 (12
* a 0 + + +
12)(z: -..+z:) is the same as G2 5 n z z ; .
40 This formula for f (z) jumps from 9 to -9. The correct formula (with continuous f ) is z2 then 18 - z2.
Then f (4) - f (0) = 2, which is I
,' u(z)dz.
42 The integral of u(z) - uaw is rero (equal positive and negative areas):
b
lo
uaVedz= (b - a)vave= u(z)dz.
5.7 The Fundamental Theorem and Its Applications (page 219)
The area f (z) = v(t) dt is a function of x. By Part 1of the Fundamental Theorem, its derivative is v(x).
In the proof, a small change A z produces the area of a thin rectangle. This area A f is approximately A x
times v(x). So the derivative of :$ t2 dt is x2.
b
The integral t2 dt has derivative -x2. The minus sign is because x is t h e lower limit. When both
limits a(z) and 6(z) depend on z, the formula for df /dz becomes v(b(x))db minus v(a(x)) g.
In the example
$23f t dt, the derivative is 9x.
The square 0 5 z 5 s, 0 5 y 5 s has area A = s2. If s is increased by As, the extra area has the shape of a n
L. That area AA is approximately 2s As. So dA/ds = 2s.
z = y2; 1
: y2dy = $1; = g; vertical strips have length 2 - &G
To integrate y(z), divide [a,b] into n pieces of length Ax = (b - a)/n. R, and Ln place a rectangle over each
piece, using the height at the right or left endpoint: R, = Az(yl - + +
yn) and L, = Ax(yo + +
. . yn- l).
These are Brst-order methods, because they are incorrect for y = x. The total error on [O,1] is approximately
y ( y ( l ) - y(0)). For y = cos s z this leading term is -Ax. For y = cos 2x2 the error is very small because
[O, 11 is a complete period.
Simpson's method is Sn = $ M, +i ~ nIt. is fourth-order, because the powers 1,x, x2,x3 are integrated
correctly. The coefficients of yo, ~ 1 1 2 ,y1 are 8, 8,8
times Az. Over three intervals the weights are Ax16 times
1- 4 - 2 - 4 - 2 - 4 - 1.Gauss uses t w o points in each interval, separated by AZ/&. For a method of order
p the error is nearly proportional to (Ax)P.
intervals), $ w 2.225 (four intervals is worse??), and 7 digits for T5. Curious that Mn = Tn for odd n.
10 The midpoint rule is exact for 1 and z. For y = z2 the integral from 0 to A z is and the rule gives
This error f - ~ ( A Z=) -~~ ( A Z does
) ~ equal - ~ ( y l ( h z -
) yl(0)).
12 The first and third integrals give accurate answers more easily.
14 Correct answer $. Tl = .5, Tlo M .66051, Tloow .66646. Ml w .707, Mlo M .66838, Mloo = .66673.
What is the rate of decrease of the error?
16 J!l
dz
= is approximated by T2 = I(? - 5 + 5 + ? 5) = $ and
s 2 =6r(r3 + 4 . kfi
2+c086nz
+2.$+4.$++ 1 1
and G1 = 2+cos(-en/&) + 2+cos(6n/&) = .776 (large error)
1 1
and G2 = w 1.5.
2 + c a ( 6 n ~ ) 2+cos(6nw)
+
In lo4 = 10,000, the exponent 4 is the logarithm of 10,000. The base is b = 10. The logarithm of lom times
+
10" is m n. The logarithm of 10m/lOn is m - n. The logarithm of 10, OWx is 4x.If y = bz then x = logby.
Here x is any number, and y is always positive.
A base change gives b = alOgab and bx = ax logab. Then 85 is 2''. In other words log2 y is log2 8 times
log8 y. When y = 2 it follows that log2 8 times log8 2 equals 1.
+
On ordinary paper the graph of y = m x b is a straight line. Its slope is m. On semilog paper the graph
of y = AW is a straight line. Its slope is log b. On log-log paper the graph of y = Axk is a straight line. Its
slope is k.
The slope of y = bx is dyldz = cbx, where c depends on b. The number c is the limit as h -4 0 of bh-l
ha
Since x = logby is the inverse, (dx/dy)(dy/dx) = 1.Knowing dy/dx = cbx yields dx/dy = l/cbx. Substituting
bx for y, the slope of logb y is l/cy. With a change of letters, the slope of logbx is l/cx.
2 4 h = 1 gives c = 9; h = .1 gives c = 2.6; h = .O1 gives c = 2.339; h = .001 gives c = 2.305; the limit is
c = In 10 = 2.3026.
bl/' -
2 6 (The right base is b = e.) With h = we pick the base so that ,174 = 1or b114 = (1 + i)
or b = (1 + i)'
= g. +
Generally b = (1 h)'lh which approaches e as h -+ 0.
10"-1 100"- 1
28 c = limh,o 7= limh,o = limhdo 7 =2
The number e is approximately 2.78. It is the limit of (1 + h) to the power l/h. This gives l.O1loO when
h = .01. An equivalent form is e = lirn (1 &)n. +
When the base is b = e, the constant c in Section 6.1 is 1. Therefore the derivative of y = ex is dyldz = ex.
The derivative of x = log, y is dxldy = l / y . The slopes at x = 0 and y = 1 are both 1. The notation for log, y
is In y, which is the natural logarithm of y.
The constant e in the slope of bx is e = In b. The function bz can be rewritten as ex ln b. Its derivative
is (In b)ex ln = (In b)bx. The derivative of e"(x) is e U ( X ) g . The derivative of esinz is esinx c o s x. The
derivative of eCxbrings down a factor c.
+
The integral of ex is ex C. The integral of eCx is i e c x + C. The integral of e u ( ~duldz
) is eU(X)+ C. In
general the integral of eU(")by itself is impossible to find.
3 5 2exl2 + J
! 2
3 7 e-z drops faster at z = 0 (slope -1); meet at z = 1; e-"/e-~ < e-9/e-3 < &j for z > 3
3 9 y - ea = ea(x - a); need -ea = -aea or a = 1
4 1 3/ = x x ( l n x + 1) = 0 at xmin = $;y" = x z [ ( l n x + 1)2+ $1 > O
4 3 $(e-" y) = e-"*d z - eqXy = 0 so e-%y = Constant or y = Cex
4 5 S 2L2 ] + &
2
4 7 z 1
ln2]-1
=1L:=_k
Ln2 21n2 49-e-x]r=1 10
51e1+"=e2-e 5 3 2sin~];= 0
55 $q d z = -e-U + C; +
J ( e u ) 2 g d z = +eZU C 5 7 yy' = 1 gives iy2= z + C or y = Jm
59 = (n - x)zn-'/ex < 0 for z > n; F(2x) < --+ 0 61 --!il
12rr
F* 117; (
:
)
6 F* 116; 7 digits
2 49e-7x 4 8e8' 6 (ln 3)exl n 3 = (1n 3)3% 8 4(ln 4)4" 10 A( 1e+l lq( l + x ) 1 2 (-$+l)el/x 14z2eZ
16 x2 + x2 has derivative 43,- 1 8 x-'Ix = e-(lnx)/x has derivative (-5+ kx2) e - ( l n z ) / x = ( v ) x - l / x
20 (1+ + +
+ e2 M 7.7 and (1 $)fi-+ 1.Note that (1 $)fiis squeezed between 1 and e l / f i which
approaches 1.
=~2.717
22 ( 1 . 0 0 1 ) ~ ~ ~ and ( 1 . 0 0 0 1 ) ' ~=~ 2.7181
~~ have 3 and 4 correct decimals. ( 1 . 0 0 0 0 1 ) ' ~=~ 2.71827 ~~~
+
has one more correct decimal. The difference between (1 $)" and e is proportional to i.
6.3 Growth and Decay in Science and Economics (page 250)
24 y = e-X solves +
= -y. The difference equation Y (x f ) = Y (x) - f Y (x) with Y (0) = 1gives Y ( f ) = 1
3 4 .
and Y(l) = (z) (Compare e-' = .37 with ( Q ) 4 = .32. See the end of Section 6.6.)
26 \/E; is the same as s l 2 . Its graph at x = -2,0,2 has the same heights f , 1,e as the graph of ex at
x = -1,o, 1.
28 (e3')(e7") = elox which is the derivative of & e l k
30 2-x = ,-x 1n2 which has antiderivative d e - ln~ =
in2
32 e-" +
x - ~has antiderivative -e-= + 34 -ecos x + eein x 36 xex - ex
+
38 ex meets xx at x = e. Their slopes are ex and xx(l In x) by Example 6. At x = e those slopes are ee and
2ee. The ratio $ = ( % )approaches
I infinity.
+
40 At x = 0 equality holds: e0 = 1 0 and e-O = 1- 0. (a) Beyond x = 0 the slope of ex exceeds the slope of
+ +
1 x (this means ex > 1).So ex increases faster than 1 x. (b) Beyond x = 0 the slope of e-l is
larger than the slope of 1- x (this means -e-' > -1). Since they start together, e-' is larger than 1- x.
42 xl/' = e(lnx)/xhas slope e(lnX)lz($ - 9)
=xl~('-lnx .
7) This slope is zero at x = e, when lnx = 1.
The second derivative is negative so the maximum of xl/" is elle. Check: $e(ln x ) / x ( y )=
~ )L 2 . L ] =
e ( l n ~ ) / ~ [ ( l - l t+ at x = e.
x
44 xe = ex at x = e. This is the only point where xee-" = 1 because the derivative is xe(-e-") + exe-le-' =
( i- l)xee-%. This derivative is positive for x < e and negative for x > e. So the function xee-%
increases to 1 at x = e and then decreases: it never equals 1again.
46 1;sin x eC08xdx= [-eCoSx]~ = -e-l e. +
48 f 1 2 - ~ d x= (by Problem 30) [&2-"]L1 = &(+ .- 3
- 2) =
2
50 Jr xe-"'dx = Jr ee-udu
2
= [-L
-u
m = 1
210 3.
11
52 Ji ] ~ 1(" 2 - e)
e l + ~ ' xdx = [ i e l + ~ ' =
z
54 $,' - e')lOe~dx = [-
1:- =-
11
56 y' (x) = 5 y(x) is solved by y = Ae5' (A is any constant). Choose A = 2 so that y(x) = 2 e 5 ~ h as y(0) = 2.
+
58 The asymptotes of (1 !)x = (?)% = (A)-'
are x = -1 (from the last formula) and y = e (from the
first formula).
6 0 The maximum of x6e-" occurs when its derivative (6%' - x6)e-" is zero. Then x = 6 (note that x = 0
is a minimum).
6 2 lim$ = limF6xS = limT30s' = limT120x3 = l i m F = l i m F = l i m F = 0.
ln2
If y' = cy then y(t) = yoect. If dyldt = 7y and yo = 4 then y(t) = 4e7t. This solution reaches 8 at t = 7.
l n 2 If y' = 3y and y(1) = 9 then yo was ge-'.
If the doubling time is T then c = T. When c is negative, the
solution approaches zero as t -+ oo.
The constant solution to dyldt = y + 6 is y = -6. The general solution is y = Aet - 6. If yo = 4 then A = 10.
+ +
The solution of dyldt = cy s starting from yo is y = Aect B = (yo :)ect - + e.
The output from the source
is e(ect - 1).An input at time T grows by the factor ec(t-T) at time t.
At c = lo%, the interest in time dt is dy = .O1 y dt. This equation yields y(t) = yOe-O1t. With a source
term instead of yo, a continuous deposit of s = 4000/year yields y = 40,00O(e - 1)after ten years. The deposit
6.3 Growth and Decay in Science and Economics (page 250)
required to produce 10,000 in 10 years is s = yc/(ect - 1)= 1000/(e - 1). An income of 4000/year forever (!)
comes from yo = 40,000. The deposit to give 4000/year for 20 years is yo = 40,000(1- e-2). The payment
rate s to clear a loan of 10,000 in 10 years is 1000e/(e - 1) per year.
47 (1.02)(1.03) + 5.06%; 5% by Problem 27 49 20,000 e(20-T)(-5)= 34,400 (it grows for 20 - T ears)
5 1 s = -cyoe"/(ed - 1) = - ( . 0 1 ) ( 1 0 0 0 ) e ~ ~ ~ / (-e -1)~ ~ 65 yo = -006 - e--005(48)1
Sbe4c=1.20soc=~ S724eS6s5=? WTo-oo;constant;to+oo
61 = 6OcY; = 60(-Y +
5); still Y , = 5
2 ~ = - t g i v e s d y = - t d t a n d y = - ~ t 2 + ~ . T h e n y = - 21t 2 + l a n d y = - i t 2 - l s t a r t f r o m l a n d - 1 .
4 2 = -y gives Y = -dt and In y = -t +
C and y = Ae-' (where A = eC). (Question: How does a negative
y appear, since eC is positive? Answer: $9
= In lyl leads to lyl = Ae-' and allows y < 0.)To start from
1 and -1 choose y = e-t and y = -e-t.
6 9 = 4t gives dy = 4t dt and y = 2t2 C = 2t2 + + 10. This equals 100 when 2t2 = 90 or t = m.
-
8$=e-4'hasy(t)=k+~= 7
4t
+ loa1 This only increases from 10 to 10; as t + oo. Before t = 0 we
- In 57
find y(t) = I when 4 4t
=9: ore-" = 37or t = .T.
10 The solutions of y' = y - 1 (which is also (y - 1)'= y - 1) are y - 1= Aex or y = Aex 1. Figure 6.7b is +
+
raised by 1unit. (The solution that was y = ex is lifted to y = ex 1. The solution that was y = 0 is lifted
to y = 1.)
1 2 To multiply again by 10 takes ten more hours, a total of 20 hours. If eloc = 10 (and e2OC= 100) then
10c = In 10 and c = In 10 ar .2S.
1 4 Following Example 2, the ratio e" would be 90% or .9. Then t = = ( 3 ) 5 5 6 8 = (In 1.8)5568 = 3273
1
24 Go from 4 mg back down to 1 mg in T hours. Then e--OIT = f and -.01T = In f and T = % = 139
hours (not so realistic).
26 The second-order equation is (& -c)($ -C)y = 9 - (c+C) 2 + c c y = 0. Check the solution y = ~ e " +~ e ~ '
+
by substituting into the equation: c2Aect C2BeCt - (c + C)(cAect + CBeCt) + cC(Aect + BeCt) does
equal sero.
+ +
28 Given mu = mu - vAm mAv - (Am)Av Am(v - 7); cancel terms to leave mAv - (Am)Av = 7Am;
divide by Am and approach the limit m & +
= 7. Then v = 7 in m C.At t = 0 this is 20 = 7 in 4 + C
so that v = 7 l n m + 2 0 - 7 l n 4 = 7 i n ~ + 2 0 .
30 Substitute y = Ae-t + B into y' = 8 - y to find -Ae-t = 8 - Ae-t - B. Then B = 8. At the start
+ +
yo = A B = A 8 so A = yo - 8. Then y = (yo - 8)e-t + 8 or y = + 8(1- e-').
52 Apply formula (8) to 2 = y - 1 with yo = 0. Then y(t) = (et - 1) = 1- et.
34 Formula (8) applied to 2 = -y - 1with yo = 0 gives y = *(e-' - 1)= e-t - 1.
56 (a) %
=3y+6givesy-+oo (b) $=-3y+6gives %
y - + 2 (c) =-3y-6gives y-+-2
(d) $ = 3y - 6 gives y -+-00.
+ 5:
38 Solve y' = y et by adding inputs at all times T times growth factors et-T : y(t) = e t - = e T d ~=
5,'etdT = tet. Substitute in the equation to check: (tet)' = tet et. +
40 Solve y' + +
y = 1 by multiplying to give ety' ety = et. The left side is the derivative of yet (by the
+
product rule). Integrate both sides: yet - yoeO= et - e0 or yet = yo et - 1 or y = yoe- + 1- e-t.
$1000 changes by ($1000) (-.04dt), a decrease of 40dt dollars in time dt. The printing rate should be s = 40.
First answer: With continuous interest at c = .09 the multiplier after a year is emo9 = 1.094 and the effective
rate is 9.4%. Second answer: The continuous rate c that gives an effective annual rate of 9% is eC= 1.09
or c = In 1.09 = .086 or 8.6%.
yo grows to yOe(*1)(20) = 50,000 so the grandparent gives yo = 50,000e-~a $6767. A continuous deposit s
grows to +(e(-l)(zO- 1) = 50,000 so the parent deposits s = j e = $785 p e r year.
+
Saving s = $1000/yr grows to ?(eelt - 1)= 50,000 when eelt = 1 - or .It = In6 or t = 17.9 years.
The deposit of 4dT grows with factor c from time T to time t, and reaches eC(t-T)4d~.With t = 2 add
4e~('-T) 1 - 4 e c - 4 e 2 ~
5;
deposits from T = 0 to T = 1 : e C ( 2 - T ) 4 d= ~ [- -c lo - -c
y(t) = (5000 - s ) e . 0 8 t + is sero when emo8' =
(Remember the deposit grows until it is withdrawn.)
,e = 5. Then .08t = ln 5 and t = a 20 years.
The natural logarithm of x is $ ' (or $). This definition leads to ln xy = In x + In y and in xn =
n In x. Then e is the number whose logarithm (area under l / x curve) is 1. Similarly ez is now defined as the
number whose natural logarithm is x. As x + oo, ln x approaches inRnity. But the ratio (In x)/& approaches
Eero. The domain and range of In x are 0 < x < oo, -00 < In x < oo.
+ x) is -=.
The derivative of in x is f1. The derivative of ln(1 1 The tangent approximation to ln(1 + x) at
z = 0 is x. The quadratic approximation is x - %x2.The quadratic approximation to ez is 1+ x + zx
1 2.
The derivative of ln u(x) by the chain rule is -&g.Thus (In cos x)' = - = -tan x. An antiderivative
+
of tan x is -In cos x. The product p = xeSx has lnp = 6 x In x. The derivative of this equation is p'/p = 6 + z.
1
Multiplying by p gives p' = xeSx(6 + 4) +
= 6xeSx eEx,which is L D or logarithmic differentiation.
The integral of u1(x)/u(z) is In u(x). The integral of 2x/(x2 + 4) is ln(x2 + 4). The integral of l/ex is 9.
+
The integral of l/(ct s) is w.
112,
write in 1x1 for the antiderivative of
The integral of 1/cos x, after a trick, is In ( ~ e xc + t a n x ) . We should
since this allows x < 0. Similarly $ du/u should be written lnlul.
-1
2 s* 61nx 7 - = cot x 9 11 ; l n t + c isin$
iln5 17-ln(ln2) 191n(sinx)+C 21-$ln(cos3x)+C 2S?(ln~)~+C
i +
Iny = ln(x2 1); $ = @ T i 29 2
= esin 2 cos x
4~ = e=ee'
dz ~ ~ l n ~ = e ~ l n x ; ~ = ~ e ~ ( Sl 6n1 xn y+ =~- )l s 0 y = ~ e9&
dz= O 57 0
-f 4 1 sec x 4 7 .l; .O95; .095310179 4 9 -.01; -.01005; -.010050335
l'H6pital: 1 65 663-21n2 67Rectangulararea?+.--+i<$;y=lnn
Maximumate 610 6Slogloeor& 651-x;l+zln2
Ractionisy=1whenln(T+2)-ln2=1orT=2e-2 6 9 y '(t+Z) = ~ + y = 1 - 1 ,+,
never equals 1
lnp = xln2;LD P l n 2 ; ~ D = e'1n2,p' = ln2 ezln2
2' = l a ; ylnx = xln y + = y; decreases after x = e, and the only integers before e are 1and 2.
2
2a*+f 4 = a
x2
6 Use (log, 10)(loglo x) = log, x. Then (loglo x) -
d 1 . I - - 1
loge10 z - x I n 10'
~ Y = ~ u s o ~ = * = *l n =
u
, . 1 O y = 7 l n 4 x = 7 l n 4 + 7 l n x s o ~ 7= ~ .
12in(l+x)from$+. 1 4 i l n ( 3 + 2 t ) ] ~ = i ( l n 5 - l n 3 ) =13 ~ 53 *
16 y = 4
z +I equals x - ,*.
Its integral is [;x2 - i ln(x2 + I)]: =2- i
ln 5.
18$$=-t=[-&l;=-l+=. 1
- 6.4 Logarithms (page 258)
~ 4- ~ J . - + o =
a dx z = J + = - l n u = - ~ n ( c o s z ) ] ~=
$ cos fi
J -dz = f In(sin ~ x )C. +
Set u = In ln z. By the chain rule 2
= & Our integral is $ i. % = In u = In (h(hx)) + C.
The graph starts at -oo when z = 0. It reaches xero when z = 5 and goes down again. At z = it
.~r stops.
+ +
In y = ln(z2 1) $ ln(z2 - 1).Then Ay d&z = X '-~ + I x'-1 = + &. Then 2= ~ 4 -
- - zxS
1 - JG'*
+ +
[In z]ir [ln 1~111: = (a - 0) (0 - ln I - 21) = 7r - In 2.
-& 1 Alternatively use ;;f&(z2) - $;IS(z) = $.
l n z = z.
This is $ +
with u = sec z tan z so the integral is in (see x +
tan x). See Problem 411
d
&(ln(z-a) - l n ( z + a ) ) = x a - & 1=- i x a (x+a) = -?e
x2-alo
1+-
Misprint! =+$F.
a+as = -
1 d P s + x -
J3c'+alx+Jz'+o'- . - -
1
JXm'
+ +
Linear: e.' rr 1 .1 = 1.1. Quadratic: e.' N 1 .1+-$(.I)~= 1.105. Calculator: em' = 1.105170918.
+ + +
Linear: e2 PJ 1 2 = 3. Quadratic: e2 N 1 2 $(22) = 5. Calculator: e2 = 7.389.
-
e*
Use l'H6pital's Rule: lim - = 1.
2-0 1
bZlnb
Use l'H6pital's Rule: lim -- In b. We have redone the derivative of bx at z = 0.
2-0 1
+ % + i+
-- -
'22
-t 0. This is
;
0 as y = -+ 0. (Emphasize: The factor y -t O
-+
line (no change by reversing z and y), roughly like the hyperbola (x - l)(y - 1) = (e - I ) ~ .
+
The equations dyldt = cy and dyldt = cy s and dyldt = u(y)u(t) are called separable because we can
I +
separate y from t. Integration of dy/y = c dt gives in y = c t constant. Integration of / d y / ( ~+ s/c) =
+ +
$ c dt gives ln(y f) = c t C. The equation dy/dz = -z/y leads to / y dy = - dx. Then y2 z2 = Ix +
constant and the solution stays on a circle.
The logistic equation is dy/dt = c y - by2.The new term -b# represents competition when cy represents
I
growth. Separation gives J dy/(cy - b#) = dt, and the y-integral is l/c times in . Substituting yo at
&
t = O and taking exponentials produces y/(c - by) = edyo/(c - b y o ) As t + oo,
y approaches g. That is the steady state where cy - by2 = 0. The graph of y looks like an 5,because it has an
c
16.
inflection point at 2
In biology and chemistry, concentrations y and z react at a rate proportional to y times z. This is the
Law of M a s s Action. In a model equation dyfdt = c(y)y, the rate c depends on y. The MM equation is
dyldt = -cy/(y + I
K). Separating variables yields y d y = / -c dt = -ct + C.
17et - 5 +
3 (qx2 l)lI3 6z 7 el-coe t ~ ( $ f + f1 1iy ) , =~
O ; t = Lb ~ o
1 5 ~ = l + e - ~ , y i s i n l S 17ct=ln3,ct=ln9
19 b = 10-',~ = 1 3 - 1 0 - ~ ; ~ = , 13.10'; at y = 6 (10)gives i n b = ct+ln* C- l o 6 b so t = 1900+ = 2091
2 1 d dips down and up (a valley) 2 3 sc = 1= sbr so s = ;,1 r = b
N
26 y = l + , - ~ t ( N - l ) i'2 = +0
+
27 Dividing cy by y K > 1 slows down y1
31 = *;
f ~ g = ~ * > o , = - - +u+K
c
multiply eulK
S~y'=(3-~)~;&=t+$;~=2att=i
= e-"IK euOIK (g)by K and take the Kth power to reach (19)
36Aet+D=Aet+B+Dt+t-,D=-l,B=-l;yo=A+BgivesA=l
3 7 y-, 1from yo > 0, y + --from yo < O;y+ 1 from yo > O,y+ -1 from yo < 0
$9 sc::,dy =$dt+ln(siny) = t + ~ = t + l n i .Thensiny= i e t stops at 1when t = l n 2
tydy=dtgives $ y 2 = t + C . T h e n C = $ a t t = ~ . ~ o y 2 = 2 t + 1 a n d y = ~ ~ .
4 ~=dzgivestan-ly=z+~.~hen~=~atx=~.Soy=tanx.
6 & +
= cos zdz gives ln(sin y) = sin z C.Then C = ln(sin 1) at x = 0. After taking exponentials
sin y = (sin l)esiP X. No solution after sin y reaches 1 (at the point where (sin 1)elin = 1).
+
8 eudy = etdt so eu = et C. Then C = ee - 1 at t = 0. After taking logarithms y = ln(et + ee - 1).
10 dM = +
= n. Therefore ln y = nln x C. Therefore y = (xn)(eC) = constant times xn.
6.6
y' = b$ gives y-%y = b dt and
(b) For
Correction: u =
Then y =
If a
[(a
= rYo and
is this slope
5
5,
4
infinite when b t = or t =
(a) Compare &<
Then
-
4s.
=
to obtain $ = 3 = cu. So u = me .
-: b t +
=
being eliminated.
+
= d then cy = dy dK and y =
+ +
+ +4
=
3.
+
C. Then C =
A.
+
6 $L&) *
Equation (14) is s = $(b T c - " ) . Turned upside down this is y =
-Y - b h
-1+2e-
n+(&)2
6.6 Powers Instead of Exponentids
with yo =
with d =
=* g .
F.
4.
+
factorialn and is equal to n(n - 1) . (1).At x = 1the series for e is 1+ 1 5
+
a.
= -r[A][B] = -r[A](bo - E(w - [A])). The changes w - [A] and bo - [ B ]an in the proportion rn
To change cy - by2 (with linear term) to a2 - x2 (no linear term), set x = dy - and a = 4
(We completed the square in cy - bd.)Now match integrals: The factor & is times 4
(from dx = 4 dy). The ratio =G
*,
A
,b -
is A.
(page 276)
2 6 .
The y line shows where y increases (by y' = f (y)) and where y decreases. Then the points where f (y) = 0
are either approached or left behind.
y' = cy(1- f ) agrees with y' = cy - b
+
1+ . +
n(4)
.
71
+ g)n.As n +,,
+
continuous
6.6 Powers Instead o f Exponentids (page 276)
compounding multiplies by ex. At z = 10% with continuous compounding, $1 grows to e - l n $1.105 in a year.
+ 1) = ay(t) yields y(t) = at times yo. The equation y(t + 1) = ay(t) + s is solved
The difference equation y(t
by y = at yo + s [ l + + - - + at-'].
deposits of s = 1 produce y = 7
The sum in brackets is or .e
?
aa.
When a = 1.08 and yo = 0, annud
1 . 0 8 -~1 after t years. If a = and yo = 0, annual deposits of s = 6 leave
1 after t years, approaching y, = 12. The steady equation y, = ay,
1 2 ( 1 - $) s gives y, = s / ( l - a).+
When i = interest rate per period, the value of yo = $1 after N periods is y(N) = (1 ilN. The deposit +
+
to produce y(N) = 1 is yo = (1 i)-N. The value of s = $1 deposited after each period grows to Y(N) =
+
6 ((1 i)N - 1).The deposit to reach y(N) = 1 is s = (1- (1 i)-N).
1
+
Euler's method replaces y' = cy by Ay = cyAt. Each step multiplies y by 1 c A t . Therefore y at t = 1 is+
+
(1 c ~ t ) ' / * ' ~ which
~, converges to yoec as At -+ 0. The error is proportional to A t , which is too large for
scientific computing.
35 ~m = 965 5 7 y(l.lao
- 1) = 57,275 3 9 y, = 1500 4 1 2; = 2.69; e
4 3 1.0142'~= 1.184 -+ Visa charges 18.4%
+ + +
2 y = 1 22 + i ( 2 ~ ) i~( 2 ~ ) .~. . Integrate each term and multiply by 2 to find the next term.
4 A larger series is 1 1 + + + +k + . - . = 3. This is greater than 1 1 - = e. + +5+i+
6 ln(1- i)"= n ln(1- n i) .(-a)
= -1. Take exponentials: (1- n e - l . Similarly i)n
+
ln(1 + ;)" = nln(1 ): w n(:) = 2. Take exponentials: (1 n e2. + i)n
8 The exact sum is e-' n .37 (Problem 6). After five terms 1 - 1 - = +i i+
= .375. &
By the quick method l n ( l + nn ( 5 ) -+ + 5)"-+
0. So (1 e0 = 1.Similarly l n ( l + :)"I n n 2 ( i ) -+ oo
SO (1 +;)na + 00.
Under the graph of f , the area from 1to 1 $ is ln(1 + + i).The rectangle inside this area has base $ and
height 4.Its a r e a is
I+,
so this is below l n ( l +
x+f
i).
y(0) = 0, y(1) = 1,y(2) = 3, y(3) = 7 (and y(n) = 2n - 1). 16 y(t) = ( f ) t .
~ ( t=) t (Notice that a = 1). 2 0 y(t) = st s[*]. +
22 ~ ( t= ) 5at s[%]. +
Ask for ky(0) - 6 = y(0). Then y(0) = -12. 26 Ask for - i y ( 0 ) 6 = ~ ( 0 )Then
. +
~ ( 0=) 4.
If -1 < a < 1then approaches E.1
The equation -dP(t +
1)+ b = cP(t) becomes -2P(t + + +
1) 8 = P ( t ) or P ( t 1) = - i ~ ( t ) 4. Starting +
from P(o)= O the solution is P ( t ) = 4[ ( - + l ] -1 2 ) I+:-
= - (-1 t
(1+ s)365 = 1.105156. - (Compare with em1 n 1+ . l + i ( . ~ =) 1.105.)
~ The effective rate is 5.156%.
Present value = $1,000 (1.1)-~'n $148.64.
Correction to formulas 5 and 6 on page 273: Change .05n to .05/n. In this problem n = 12 and
N = 6(12) = 72 months and .05 becomes .1 in the loan formula: s = $10,000 (.1)/12[1- ( I + w $185.
6.7 Hyperbolic finctions (page 280)
n=
n 1 . 1 r r 17
The interest is (.05)1000 = $50 in the first month. You pay $60. So your debt is now
$1000 - $10 = $990. Suppose you owe y(t) after month t, so y(0) = $1000. The next month's
+
interest is .05y(t). You pay $60. So y(t 1) = 1.05y(t) - 60. After 12 months
y(12) = (1.05)121000 - 60[=]. This is also +
(1000 - g)
(1.05) l2 M $841.
Compounding n times in a year at 100% per year gives (1 + i)".
Its logarithm is nln(1 + i)
rr n [ i - &]
= 1- &. Therefore (1 + i)"
w e(e-'I2") M e(1- &).
Use the loan formula with .09/n not .09n: payments s = 80,000- rr $643.70.
Then 360 payments equal $231,732.
Cosh x = 4 (ex + e-.I and sinh x = 4 (9- e-x) and cosh2 x - sinh2 x = 1.Their derivatives are s i n h x
and cosh x and zero. The point (x, y) = (cosh t, sinh t) travels on the hyperbola x2 - y2 = 1.A cable hangs in
the shape of a catenary y = a cosh:.
- ' '
The inverse functions sinh- x and tanh- x are equal to ln[x + d z ]and 4 in E.Their derivatives are
1 / d x 2 + 1and &. SOwe have two ways to write the antiderivative. The parallel to cosh x + sinh x = ex
is Euler's formula cos x + i sin x = eix. The formula cos x = $(cis + e-") involves i m a g i n a r y exponents. The
parallel formula for sin x is (ejx - e- jx) .
1e ~e-x
, , e'=-e-2z -
- isinh2x
7 sinh nx 9 3 sinh(3x 1) + 11 - sinh
cosh2 = - tanh x sech x
4coshx~inhz 15*(se~hJm)~ 176sinh5xcoshx
cosh(lnx)=~(x+~)=lat~=l21L 1 3' ,35 -1 = -5' '3 - 1P2 , 12 2Q0,0,1,00,00
i sinh(2x + 1) 27 5 cosh3 x 29 ln(1 + cosh x) 31 ex
S y dx = s s i n h t(sinh t dt);A = isinh tcosht - sydz;A1 = ;;A = 0 at t = 0 so A = i t .
e ~ = z + ~ ~ , y = l n [ x + d4 7~ f l] n I E I 4 9 sinh-' x (see 41) 5 1 -sech-'x
i In 3; oo 55 y(x) = f cosh cx; cosh cL - f
;
y" = y - 3 9 ; ( Y I ) ~= iy2
- y3 is satisfied by y = isech2:
g.
= g , c s c h x = 4.
3
In 4 -e-2 In 4
24 sinh(ln5) =
eln S-=- In 5
- 2 = y ; t a n h ( 2 l n 4 ) = e,1n4+ - 16-k - 255
a rn'
l6+& -
4
2 6 J' x ~ o s h ( x ~ ) d=x sinh(x2) C. +
2 8 f (tanh x ) ~C. +
30 J' coth x dx = J' -dx = ln(sinhx) C. + 3 2 sinh x + cosh x = ex and J enxdx = Be- + C.
3 4 y = tanh x is an odd function, with asymptote y = -1 as x -+ -oo and y = +1 as x + +oo. The
inflection point is (0,0).
36 y = sechx looks like a bell-shaped curve with ,y, = 1at x = 0. The x axis is the asymptote. But note
that y decays like 2e-' and not like e-".
38 To define y = cosh-' x we require x 2 1.Select the positive y (there are two y's so strictly there is no inverse).
For large values, cosh y is close to i e so
~ cosh-' x is close to ln 2%.
?
4 0 l n ( e ) approaches +oo as x + 1and -a, as x -+ -1. The function is odd (so is the tanh function).
The graph is an S curve rotated by 90'.
The quadratic equation for e' has solution eY = x f dm. Choose the plus sign so y -t oo as x --t oo.
Then y = ln(x d-) + is another form of y = cosh-' x.
The x derivative of x = sinh y is 1= cosh y $ . Then $ = =d --&
- 1 - t/l,a 1 -- of sinh-l x.
Integration by parts is the reverse of the product rule. It changes $ u dv into uv minus $ v du. In case u = z
and dv = e2'dz, it changes $ ~ e ~ to~ dxe& fz minus $ f
ehdx. The definite integral zeaXdzbecomes Se4
1
minus a.
In choosing u and dv, the derivative of u and the integral of dvldz should be as simple as possible.
Normally in z goes into u and ex goes into v. Prime candidates are u = z or z2 and v = sin z or cos x or ex.
When u = 2 we need two integrations by parts. For $ sin-' z dz, the choice dv = dz leads to x sin-% minus
$x d x / l / z P .
If U is the unit step function, dU/dz = 6 is the unit delta function. The integral from -A to A is U(A) -
U(-A)= 1. The integral of v(z)6(z) equals ~ ( 0 )The
. integral cos z 6(x)dz equals 1. In engineering, the
balance of forces -dv/dz = f is multiplisd by a displacement u(z) and integrated to give a balance of work.
1 9 i(z2+1)tan-'x- : + C 21 z3sinz+3z2cosz-6zsinz-6~0sx+C
23 e'(xS - 3 2 + 62 - 6) C + +
26 z tan x ln(cos x) C +
27 -1 29 -?e-' ! 31 -2 +
33 3 l n l 0 - 6 + 2 t a n - l 3 36 u = z n , v = e z S7u=zn,v=sinz S O U = (lnx)",v= z
41 u = z s i n x , v = e x - r $ e x s i n z d z i n 9 a n d - ~ z c o s z c x d z .Then u = -zcosx,v=ex +$excoszdx
in 10 and - I z s i n z e f d z (move to left side): f (zsinz- zcosz+cosx). Also try u = zex,v = -cosx.
43 / $u sin u du = $ (sin u - u cos u) = ? (sin z2 - z2 cos z2); odd 46 3- step function; 3ez. step function
49 Q x6(z)] - I6(z)dx = -1; u(z)6(z)j - $ v(z)6(z)dz 6 1 v(z) = $: f (z)dz
53 u(z) = $ #v(z)dz; i(: - G);f for z 5 ?,i ( 2 z - z2 - !) for z 2 i;f for z 5 ?,& for z 2 $.
55 u = z2, v = - cos z -z%os z - (22) sin z - $ 2 sin z dz
4 57 Compare 23
$0'
69 uw']; - u'uf - u'w]; + 1 : u'w' = [uw' - u'wl;
61 No mistake: e' cosh z - e2 sin hz = 1is part of the constant C
f - / f e 4 x d=
2 uv - $ v du = ~ (e4=) ~ e4=( f - &) c +
4 uv-$vdu=z($sin3z) - $ i s i n 3 z d z = :sin3z+ f c o s 3 z + ~
~ U ~ - $ ~ ~ ~ 2 =x ( ~ ~ )4 $ - J ~ ~ = $ ~ ~ - ~ + C
8 uv - $ v du = 2(!e4') - $(je4x)2z dz = (Problem 2) e4'(f - :+ j?i) + C
1 + I
1 0 ex cos z dz = ez sin x - ex sin z dz. Another integration by parts produces ez(sin z cos z)- ex cos x dz.
+
Move the last integral to the left side and divide by 2: answer ?ex(sin z cos z) + C.
1 2 Not by parts. Substitute u = x2, du = 22 dz : J' i e - ~ d u= -ie-" = -&e-~'
a + C.
7.2 Digonometric Integrals (page 293)
1 4 $ cos(1n z)dz = uv - $ vdu = cos(1n z)z + $ z sin(1nz) !dz = again by parts gives cos(1n z)z + sin(1n z)z
- $ z cos(1n z) $ dz. Move the last integral to the left and divide by 2: answer (cos(1nz) + sin(ln z)) + C.
16uv-$vdu=(lnz)$-$$$=(lnz)$-$+c.
18 uv - $ v du = cos-' (2z)z + $ z ,/- & = z C O S - ~ ( ~-X$)(1- 4z2)l12 + C.
20 $ z2 sin z dz = z2 (- cos z) + $ cos x(2z dz) = a g a i n by parts gives -z2 cos z + (sin z)22 - $ sin z(2 dz) =
answer: -z2 cos z + 25 sin z + 2 cos x + C.
22 uv - $ v du = z3(- cos z) + $(cos z)3z2dz = (use Problem 5) = -z3 cos z + 3z2sin z + 6zcos z - 6sin z + C.
24 uv- $ v d u = s e c - ' ~ ( 5 ) - sL1- IxId3 - $ sec-l z + + C.
26 u v - $ v d u = zcoshz-$coshzdz= z coshz-sinhz+C.
2 8 $ , ' e ~ d z = $ ~ ~ 0 e u ( 2 u d u ) = 2 e u ( u - 1 ) ~ ~ = 2J O
. ln(z2)=21nz;$~21nzdz=[2(zlnz-z)];=2.
52 r, z sin z dz = [sinz - zcos z]:, = 2s.
34 $,"I2 z2 sin z dz = (Problem 20) [-z2 cos z + 22 sin z + 2 cos x]:'~ = s - 2.
36 $ zneaxdz = z n 5 - - $ zn-'ea2dz. 38 $ zn sin z dz = -zn cos z + n $ zn-' cos x dz.
40 $ z(1n z)"dz = (ln z)"$ - $ $n(ln z)"-'$ = $(ln z)" - $ z(1n z)"-'dz.
42 Try u = tan-' z and dv = zex dz so v = (z - l)ex. Then $ v du = $ *ex dz. I believe this cannot be done
in closed form; that is true for $ $dz.
44 (a) e0 = 1;(b) v(0) (c) 0 (limits do not enclose zero).
1
46 6(2z)dz =
2
I,=-, 9
6(u) = 3 . Apparently 6(2z) equals i 6(2); both are zero for z # 0.
48 1,' 6(z - i ) d x = I-,,,
112
6(u)du = 1 ; $', ex6(z - i ) d z = I!/:/, eu+f 6(u)du = e l l 2 ; 6(z)6(z - 3) = 0.
50 $
,
! ~ ( zE)d z = (directly) [ i ( ~ ( z ) ) ' ) ] A= 4.
52 ,,
-2 = z gives v = -$ + c = -2, + I.-& dr i)
- U(z - gives a change in slope at z = : i
4
v = C for z 5 and v = C - (z - i) 1 to make v(1) = 0;
for z 2 ;; take C = 2
2 2
-- = 6(z - $) gives v = C for z < and v = C - 1 for z > take C = 1to make v(1) = 0. i;
54
(&)AX = 1.As Ax -
2 = & over the interval from z = -Ax to z = 0. Elsewhere AU = 0. The area under the graph is
0 the area is tall and thin. In the limit
56 (-1)" $ g ~ ( , - ~ ) d=z (-1)" %v(,) + (-1)"" $ -v(,]dz. n+ 1
$ 6(z)dz = 1.
To integrate sin4 z cos3 z, replace cos2 z by 1- sin2x. Then (sin4z - sin6 z) cos z dz is (u4 - u6)du. In
terms of u = sin z the integral is i u 5 - #u7. This idea works for sinm z cosn z if rn or n is odd.
If both m and n are even, one method is integration by parts. For $sin4z dz, split off dv = sinz dz.
7.2 lZigonometric Integrals (page 293)
Then - v du is $ S sin% cos2x. Replacing cos2 z by 1- sin% creates a new sin4 z dz that combines with
the original one. The result is a reduction to / sin2 z dz, which is known to equal 1( x - sin x cos x).
The second method usea the double-angle formula sin2 z = i ( 1 - cos 2x1. Then sin4 z involves cos2 2x.
Another doubling comes from cos2 21 = 4 ( 1 + c m 4x1.The integral contains the sine of &.
To integrate sin 6xcos 42, rewrite it as i s i n 10s + i s i n ix. The integral is -&cos l O x -f cos 2x. The
definite integral from 0 to 2r is sero. The product cos px cos qz is written as $ cos(p + q)z + 3
1cos(p - q)x.
Its integral is also sero, except if p = q when the answer is x.
With u = tan z, the integral of tanQz sec2 z is & tan1%. Similarly $ secg z(sec x tan x dz) = h s e c lox.
+
For the combination tanm z secn z we apply the identity tan2 z = 1 see2 x. After reduction we may need
$tan x d z = - l n c o s x a n d / s e c z d z = I n ( s e c x + t a n x ) .
5
cos x dx = 0 because the positive area from 0 to is balanced by the negative area from to r. This
is true for a n y odd power n = 1,3,5, (For even powers cosn x is always positive). The substitution
u =r - x and du = -dx 0
gives :$ eosn x dx = - I,cosn (r- u)du = $:(-l)n cosn u du.
So if n is odd, the integral equals minus the integral and must be sero.
i
(sin x) (sin x) = - cos(1+ 1)z + $ cos(1- 1)x is the double angle formula sin2x = l - " ~;(COB
2%)(cos x) =
i cos(2 + l ) x + i cos(2 - 1)x = "08 'X+CO~
2 X. TOderive equation (g), subtract cos(s + t) = cos s cos t -sin s
sin t from cos(s - t) = cos s cos t + sin t sin t. Divide by 2. Then set s = px and t = qx.
:/ sin 3x sin 52 d z = :/ - 'OS8= '08 d x = l -sin8x
a+ 16
+ sink
]o"=O. *
cos2 3%dx = 1+c~s6zdz= -sinIf +
I-, 6~ r = r .
$lm 1"
: $oar 2
e,]zr
CO"
sin z 22 sin 32 do = sbinx(- c0s4;+ca '%)dX= sin Zx( lX)dz=
Example 11 = Problem 62 for / see3 x dx and change / cos3 x dx to J(1- sin2 x) cos x dx.
Final answer
- s - q l n lsee x + tan x( 2 sin x 7 sin% + C. + +
A = 2 : 2cos(x+ E) = 2 c o s x c o s ~- 2 s i n x s i n t = cosx- f i s i n x . Therefore dx
(COsz-fisinx),
-
-
I 4 cos$+f) = itan
(X + 5)+ *'
Expand cos(x - a)= cos x cos o + sin x sin a, multiply by d m , and match with a cos x + b sin x.
Then cos a = -a and s i n o = is correct if tan o = (the right triangle has sides a and b).
When lengths are scaled by sec x, area is scaled by sec2x. The area from the equator to latitude z is then
proportional to sec2 x dx = tan x.
The graphs of sin2 x and cos2 x obviously give equal areas between 0 and 5 and between and r. The areas
add to :$ ldx = r so each area is 5.
7.3 Digonometric Substitutions (page 299)
1
62 Example 11ends with 2 sec3 z dx = sec x tan x + sec x dx. Divide by 2 to find I sec3 x dx =
4(sec x tan x + lnlsec x + tan X I )+ C.
The function 4 s suggests the substitution x = sin 8. The square root becomes COB 8 and dx changes to
<
cos 8 dB. The integral !(I- z2)312dx becomes cos48 dB. The interval $ x 5 1 changes to 853< 4.
For d m the substitution is x = a sin 8 with dz = a cos 8 dB. For z2 - a2 we use z = a see 8 with
I
dx = a sec 19 tan 8. (Insert: For x2 + a2 use x = a t a d ) . Then dx/(l + z2) becomes 5
dB, because
1+ tan2 8 = sec28. The answer is 8 = tan-' x. We already knew that 2
-
1
is the derivative of tan-' x.
l+x
+ +
The quadratic x2 2bz c contains a linear term 262. To remove it we complete the square. This gives
+ + + + + +
(z b)2 C with C = c - b2. The example x2 4% 9 becomes (x 2)2 5. Then u = x + 2. In case z2 enters
+ + +
with a minus sign, -x2 42 9 becomes -(x - 212 13. When the quadratic contains 4z2, start by factoring
out 4.
2 z = a s e c 0 , x 2 - a 2 = a 2 t a n 2d
0 ,m
$ A' = a s aet acne, t a n e d e = l n ~ ~ e ~ 0 + t a n 8 ~ = ~ n ~ ~ +
sec' Ode
4 z = ~tan0,1+9x2=sec20,$-&=I-= != % t a n - l ~ x + C .
6 z = sin0,/-- X' JGP = I s i ? ~ ~=~-cot0
s e = ---+C
8 z = a tan 0, z2 + a2 = a2 sec2 0, $ d m d z = $ a2 sec3 0d0 = use Problem 62 above:
g ( s e c ~ t a n 0+ lnlsec0 + tan01)= ;xdx-+ +hI$ + x2 2 + C GI
~ =e $ ~ = t a n -
1 4 ~ = s i n1 -0~ =,) 3$2 ~ - coss 0 + ~ =
14 = ,- q + m + c .
X
1 8 x = 2 t a n 8 , z 2 + 4 = 4sec20,$ $$ = $-2sec20 d0 = $2tan20d0 = $2(sec20- l)d0 = 2tan0-20 =
x - 2 tan-'% + C.
20 z = tan 0 , 1 + z2 = sec2 0,
--
-
I4
I -sec2 0 d0 = $ tan2 0 sec 0d0 = (use Problem 44 above)
i(secetan0-lnIsec0+tan01) = *(xd=- ~nld=+xl)+~.
22 z = sec0 : $ =J s:6:8"::fy + c = see-'x + C . For z = csc 0 the integral is $ -:::::ziy =
=0
-0 + C = -me-'x + C*. Both answers are right: sec-' z + csc-' x = sum of complementary angles in
Section 4.4 = 5 so the arbitrary constant C* is C - 5.
24 Set z2 = sec 0 and z4 - 1 = tan2 0 and 22 dx = sec 0 tan 0d0. Then J 2z:7& = J sec 8 tan8dO = 8 =
2secetanO
+ '
see- (z2).
2
26 z = sin 0 : ~ ! ~ -
(1z2)3/2dz= cos3 0(cos 0d0) = 2 $2
I: cos4 0d0 = (Problem 19 of Section 7.2)
1 .rr-3n
3
2(5)(5)(2) - g.
: $' A = J
28 x = s e c ~
1 42.T s e c ~ : ~ ~ d=B lnIsec0 + tanel = [lnlz+ J=ll: = ~n( 4 + a).
SO ,$! &
4 = [iln(z2 + l)]yl = 0 (odd function integrated from -1 to 1).
32 First use geometry: $i2
d s d z = half the area of the unit circle beyond z = $ which breaks into
-
dx COBz d2
341% = ~ s e c z d z = l n ~ s e c x t+a n x l + C ; $ & ( - ) = $ = - I sin22 =
-cotz +
s1n x
+
= 1-cosx
s mx +C;$ = $- = fiInlsec$ + tan$I+ C
1+-
36 z = tan 0 gives \ / '+I
A E-=/~=ln(secO+tan~)=~n(z+JZZ)=g. (b)Checkgt===
-&==.
'+I
(c) Thus sinh g = $(eg - e-g -A
) - 2 ( z + m - x + & i ) = 5 (
1 z 2 22
+
1
f+&&+ I = x .
X' 2' 1-1
r = cosh 19:jd+dz= 1
ssinh BdB = /tanha BdB = /(I -sech2B)dB = 8- tanh B = c a s h - l x - -+c
~ = / ~ /coth2 BdB= /(l+csch28)dB= 8-mthB =sinh-lx
z = s i n h B : / ~ ~ coshBdB= -w + C
Proof
by geometry: The z scale has factor f and the y scale has factor 4, so dA = dzdy is unchanged.
The idea of partial fractions is to express P(z)/Q(z) as a s u m of simpler terms, each,one easy to integrate.
To begin, the degree of P should be less than the degree of Q. Then Q is split into linear factors like z - 5
+ +
(possibly repeated) and quadratic factors like z2 z 1 (possibly repeated). The quadratic factors have two
complex roots, and do not alluw real linear factors.
A factor like z- 5 contributes a fraction A/(x - 5). Its integral is A In(x - 5). To compute A, cover up x - 5
in the denominator of P/Q. Then set z = 5, and the rest of P / Q becomes A. An equivalent method puts all
fractions wer a common denominator (which is Q).Then match the numerators. At the same point (z = 5)
this matching gives A.
A repeated linear factor (z - 5)' contributes not only A/(z - 5) but also B/(x - A quadratic factor like
+ + + + +
z2 z 1contributes a fraction (Cx D)/(z2 z 1) involving C and D. A repeated quadratic factor or a
+ + +
triple linear factor would bring in ( E z F)/(z2 z 1)2 or G/(z - 5)3. The conclusion is that any P/Q can
be split into partial &actions, which can always be integrated.
1
1A=-l,B=l,-Inz+ln(z-1)+C 3 2-3
1 - - 2-2 L
2 -L + =x+2
x+1
1
7:+? 93-&i 11-L a 2+2-1
1 1 1 ~ - hx ? +2-1
~ - . !2-2
& + A2-3
S
15 L L + L + W . A = - ' B = ' C=O,D=-'
x+1 r-1 xl+l 9 4 4 2
1 7 Coefficients of y :0 = -Ab + B; match constants 1= Ac; A = $, B =
1 9 ~ = l , t h e n~ = 2 a n d ~ = l ; ~ s + / w =
ln(z - I) + ln(z2 + z + 1) = ln(z - l)(z2 + z + 1)= ln(x3 - 1)
21u=e~;/~~~=/;t4:-J~=h(~)+c=h(~)+c
23 u = c o s ~ ;1-u
J ~ = - i J * 1-u
- i J * = l+u L
2 ln(l-u)-iln(l+u)= +In- + C. We can reach
i
In &
I,1 cose
= In -= ln(csc 9 - cot 9) or a different way In i ( I+COS
e) ',= ln a e e- In W =
i + ~ ~= sln 0
- ln(csc 9 + cot 9)
25 u = e z ; d u = e z d z = u d z ; J & d u = J1-u
~+J~= -2 ln(1- eZ) + Inez + C = -2 ln(1- e2) +x +C
2 7 z + 1 = u 2 , d z = 2 u d u ; JI+U
~=J[2-&]du=2u-21n(l+u)+~=
2J3;+1--2ln(l+ J2+1)+~
29 Note Q(a) = 0. Then = Q(.jIOg(a)
-+ A- by definition of derivative.
Q'(4 At a double root Q1(a) = 0.
1 -A
2 (z-l)(,+l, - + 5.
Cover up z - 1 and set z = 1to find A = $. Cover up z 1 and set z = -1 to find +
B=-i.~henJ&=iln(z-l)-~ In (z 1) = $In ' +
C. Method 1: (2-1)(2+1)
1
= A Z+I + B 2-1
+
and by matching numerators A + B = 0 and A - B = 1 so again A = $ and B = - i.
(z-3;z-2)
-
- 2-3
-3- - 2
2-2
1
z(z-l)(z+l) = -;
1
+ 2-1 z+1
112 *
8 -&
32 1
= 4 +& (first multiply by ( z - 1)2 and set z = 1 to find the coefficient 4).
- 112 &%+A
lo ( 2 - l ) ( + l - 2-1
1 * + * 2+2
;a+: l2 =2
%-I+&
14 z + l d z 2 + 0 z + 1 so &
z+1= z - l +2+1
l l a q q1 = - ; -1 g +1 = 1
+ y = e2'(1 - y) and
y = , l t + , -- z,;z-<
t -t
= tanh t. This is the S-curve.
22 Set u = ,/Z so u2 = x and 2u du = dz. Then /s d x = $ e 2 u du = (divide u + 1into -2u2 + 2u)
D
(u+l,2)du. Cover up (U- 1)2and set u = 1 to find B = 4; cover up (u + 1)2and set u = -1 to find
4 , match left and right to find A = C = 0. The integral is -+ 5 + f
. 7
D = -1. = = ---1 1
2e2t,l'
Check derivative: 5*T (2e2" = *.
2~'-1
Quicker integration: $ eg = - (u2 - I)-' - 1
- -z,zt,la
26 Set u3 = z - 8 so 3u2du = dx. Then J 2-8'/'dx -
- / &&
u3u2du
= (divide first) J(3 - &)du =
J 24 du
3u - (~+a)(u2-2~+4)3~
= - /(a 2
+ )du = 3u - 2 ln(u + 2) + J w
-2u
U2-2L4 d u = 3u - 2ln(u + 2)+
ln((u - 1)' + 3) - tan-'(?)
fi d-
+ C. Finally set u = ( z - 8)'13.
28 Set u4 = x so that 4u3du = dx. Then = / -= (divide first)
$ +";"F $(4u - 4 + &)du =
2u2-4~+41n(u+l)+C=2fi-443/;1+4ln($+l)+C.
SO Multiply = A + - . . by z - 1 and let x approach 1 to find A = lim f i = lim 1 =8
1.
An improper integral fab y(z)dx has lower limit a = -a, or upper limit b = a, or y becomes infinite in the
82
7.5 Improper Integrals (page 309)
5 x 5 b. The example 1
int e1-va1 a ; dx/x3 is improper because b = oo. We should study the limit of 1: dx/x3
1 For p > 1the improper integral I.1x-Pdx is finite.
ash+ oo. In practice we work directly with - ? X - ~ ] Y = 5.
For p < 1the improper integral
1
so
x-pdx is finite. For y = e-" the integral from 0 to oo is 1.
Suppose 0 <
u(x) 5 v(x) for all x. The convergence of I
v(x) dx implies the convergence of u(x)dx.
1 1
The divergence of u(x)dx implies the divergence of v(x) dx. Rom -oo to oo, the integral of l/(ex e-") +
.
converges by comparison with l/elXI Strictly speaking we split (-00, oo) into (-00, 0) and (0, oo) Changing .
to l/ (ex - e-') gives divergence, because ex = e-X at x = 0. Also dx/ sin x diverges by comparison with
$ dx/x. The regions left and right of rero don't cancel because oo - oo is not rero.
1 1-e
2 * = l 3 -2(1- x)lI2]b = 2 5 tan-' x]!,~~ = 7 i ( l n ~ ) ~=
] ;-oo
1-e 11 e-1
9xhx-XI:=-oo 1 1 l n ( l n ~ ) ] ~ ~ = o o1 3 ~ ( x + s i n x c o s x ) ] ~ = o o
15 k ] rdiverges for every p!
1-P
1 7 Less than $=
1 9 Less than 1: & + ,p. = tan-' x]: - -&I? = j+ 2
2 1 Less than 1; e-xdx = t, greater than -;
23 Less than 1,'e2dx + e 1 ; e-(x-l)ldx = e2 + e 1; ee-"du = e2 + fi
25 1,'gin1 :dx + JlW sin: dr less than 1+ 1 1*5 = 2 27 p! = p times (p - I)!; 1= 1times O!
-2
+ Jo* C-2d x = $6 so*1000e-.ltdt
29 u = x, dv = xe-"'dx
33 w= = 9=
:- x + ] ~
avo= 7- 31 = -10,000e-.~~]r= $10,000
e-~"2dx = -
e-r In 1
35 /,- .JOm
; 00-L
-In2 10 - 1.2
57 I;l2(sec x - tan x)dx = [In(secx + tan x) + ln(cos x)];12 = [ln(l + sin x)];12 = In 2.
The areas under sec x and tan x separately are infinite 39 Only p = 0
2 So ;
i= [k];
1 ds
diverges at x = 0 : infinite area 4
: & = [-
1 ln(1- x)]; diverges at x = 1: infinite area
A=[sin-lx]'l=~-(-q)=,
1-141-z'- 2
8 irw b
sin x dx is not defined because $a sin x dx = cos a - cos b does not approach a limit as b -+ oo
and a -,-00
;$ xe-"dx = [ - ~ e - ~ ] r
10
+ Jr e-"dx = 0 +1
12 Jrw is not defined because the area around x = -1 and x = 1 is infinite.
1 4 I:/2 tan x dx is not defined: it is 1,'
$ with u = cos x and the area is infinite.
16 & = (set u = ex - 1)$om 3 which is infinite: diverges at u = 0 if p 2 1,diverges at u = oo if p 5 1.
18 1,
1 dx
< 1,' = 1 : convergence 20
1: >1
: $$ = oo : divergence
22 x-"dx < Jyee-'dx = : Convergence
24 1,' d-dx < $:Ie(- 1nx)dx + JlIe ldx = [-xlnx + x];'~ + [x]ile = $ + 1: convergence (note xln x
1
--r 0
asX+O)
26!($
; - &)dx : the separate integrals would give oo - oo which is indeterminate, so combine ;- & =
< 5.The integral is less than IT .$= 1. Convergence.
28 $om x-1/2e-x dx (set x = u2) = 1; u-le-~' 2u du = 2 $;f e - ~ l d u= so this is (-i)!
6, Then (p + I)! =
7.5 Improper Integrals (page 309)
is no limiting value. If a = -b then the answer is sero - but we are not allowed to connect a and b.
S8 = (set z = u2) $r %
,
A 2u du
/r
= [2 tan-' u ] r = 2(5) = s; zeeZ cos z dz = (by parts)
[*(SinX - cosz) + f s i n z l r = 0.
-s
40 The red area in the right figure has an extra unit square (area 1) compared to the red area on the left.
8.1 Areas and Volumes by Sfices (page 318)
The area between y = 2 and y = z4 equals the integral of x' - x4. If the region ends where the curves
intersect, we find the limits on z by solving x' = x4. Then the area equals :$ (x' - x4)dx = - = 4 4 &.
When the area between y = 6and the y axis is sliced horisontally, the integral to compute is $ y2dy.
In three dimensions the volume of a slice is its thickness dz times its area. If the cross-sections are squares
of side 1- z, the volume comes from $ ( 1 - x )2 dx. Fkom z = 0 to z = 1,this gives the volume 3 1 of a square
pyramid. If the cross-sections are circles of radius 1- z, the volume comes from $ r ( 1 - x ) 'dx. This gives the
volume 8 of a circular cone.
For a solid of revolution, the cross-sections are circles. Rotating the graph of y = f (z) around the z axis gives
a solid volume $ r(f(x))'dx. Rotating around the y axis leads to $ r(f-1(y))2dy. Rotating the area between y =
f (z) and y = g(z) around the z axis, the slices look like washera. Their areas are r (f (x))' - r (g(x))2 = A(x)
so the volume is $ A(x)dx.
Another method is to cut the solid into thin cylindrical shells. Revolving the area under y = f (z) around
the y axis, a shell has height f (x) and thickness dz and volume 2- f (x)dx. The total volume is $ 2- f (x)dx.
[0 - (z2 - 2)ldz 4.
(a,0); area J-5 8 2
2 Intersect at (-fi,
0) and =
4 Intersect when y2 = y + 2 at (1,-1) and (4,2): area = + 2) - y2]dy= 8.
6 y = z1I5 and y = z4 intersect at (0,O) and (1,l):area = ~ 0 ' ( z - 1 ~ =6 -1
~ z4)dz 5 = l9 m'
8 y = $ meets y = 5 at (1,l); upper limit z = 3 : area = - 3 ) d z = [S+lj3 1 8 + & + + - 5 = =.
1 1 10
3x5 1 = -1
1 0 2 z = s i n 7 r z a t z = ? : a r e a = s o112( s i n r z - - 2 ~ ) d z = [ - ~ - z ~ :-
] ~ ~f.~ =
The region is a curved triangle between x = -1 (where e-2 = e) and z = 1 (where ex = e). Vertical strips
end at e-x for z < 0 and at ex for z > 0 : Area = s_O1(e- e-")dz +
(e - ex)dz = 2.
rhis region has y = 1as its base. The top point is at x = 9, y = 3, where 12 - z = &.Strips go
up to y = fi between z = 1 and z = 9. Strips go up to y = 12 - z between z = 9 and z = 11.
Area = J:(&- l ) d z + J c ( l 2 - 2 - 1)dx = g(27- 1) - 8 + 22- 20 = -6 = -34
3'
I'he triangle with base from z = -1 to z = 1and vertex at (0,l) fits inside the circle and parabola. Its area
?
is (2)(1) = 1. General method: If the vertex is at (t, d m ) on the circle or at (t, 1- t2) on the
parabola, the area is Jn or 1- t2. Maximum = 1at t = 0.
n
2
Volume = :J R sin zdz = [ 7 r ( 2 - s i n ~ )lo" = nz
Shells around the y axis have radius z and height 2sin z and volume (27rz)2 sin zdz. Integrate for the volume
of the galaxy: s:
4x2 sin zdz = [(*(sin z - z cos z)]: = 87r2 .
22 (a) Volume = J : ~ ( l + e ~ )=~d r(- +
z +2e $-) (b) Volume = 1' +
27rz(1+ ex)dz = [7rx2 27r(xex- ex)]; = 37r.
24 (a) Volume = s;l4 7r sin2 zdz+ J;/," r cos2 zdz = [ - - 74 10 + I y +r sin~ 2xI x"12/ 4= 2
n s i n 2x np4
8
- E + x4'
4 $' =
3 6 The tilted cylinder has circular slices of area 7rr2 (at all heights from 0 to h). So the volume is lo
7rr2dy = 7rr2h.
This equals the volume of an untilted cylinder (Cavalieri's principle: same slice areas give same volume).
38 (Work with region in figure.) The horisontal slice at height y is a square with side length d m .
The area is a2 - y2. So the volume is J:(a2 - y2)dy = !a3. Multiply by 8 to find the total volume g1 6a3 .
8.2 Length of a Plane C w e (page 324)
1 0 (a) The slices are rectangles. (b) The slice axes is time* y tan 6. (c) The volume is
1'2,/my tangdy = [-:(I -9)3/2 t a l k = ftan 6. (d) Multiply radius by r and volume by rs.
42 The area is the base length 24- time. the height
1
The length of a straight segment (Ax across, Ay up) H As = d.- Between two points
dd
on the graph of y z ,Ay is approximately dy/dx times Ax. The length of that piece is approximately
An infinitesimal piece of the curve has length ds =
the arc length integral is $ ds.
dx. Then
For y = 4 - z from z = 0 to x = 3 the arc length is 3 fidx = 3 6 . For y = 9 the arc length integral is
$,/Ga&.
The c w e z = cos t, y = sin t is the same as x2 + y2 = 1. The length of a curve given by z(t) ,y(t) is
8.2 Length of a Plane Curve (page 324)
$ d ( / ( d x ~ d+t )(dy/dt2)dt.
~ For example x = cost, y = sin t from t = r / 3 to t = s / 2 has length :/$
: dt. The
speed is dsldt = 1.For the special case x = t, y = f (t) the length formula goes back to J Jl+o)2dx.
27 Ellipse; two y's for the same x 29 Carpet length 2 # straight distance &
+
3 1 ( d ~=) (~ d ~ ) (~d ~ ) (~d ~ )ds~ =
; J(%)2+ + (g)2
+ (%)2dt;
ds = \/sin2 t + cos2 t + ldt = fidt; 27rfi; curve = helix, shadow = circle
SS L = $,' \ / W d x ; 4-dx = JW2du = 2L; stretch xy plane by 2 (y = x2 becomes !j = (:)2)
A surface of revolution comes from revolving a curve around a n axis (a line). This section computes
the s d a c e area. When the curve is a short straight piece (length As), the surface is a cone. Its area is
A S = 2xrAs. In that formula (Problem 13) r is the radius of t h e circle traveled by the middle point. The
line from (0,O) to (1,l) has length A s = 6,and revolving it produces area rfi.
When the curve y = f (x) revolves around the x axis, the area of the surface of revolution is the integral
/ 26(x) For y = x2 the integral to compute is /2 d d G Z % c . When y = x2 is
m around the y axis, the area is S = / f
revolved For the curve given by x = 2t, y = t2,
change ds to W d t .
5 2 r & G F m d x =
1
4rdx = 8 r 7 1,? 2axd1+(22)2dx = f (1 + 4x2)312j~= ~ [ 1 7-~11l ~
1
;
9 2rx\/zdx = 9 7 r a 11Figure shows radius s times angle 0 = arc 2aR
+
1 3 2mAs = r ( R R ) ( s -st) = r R s - rR'st because Rts - Rs' = 0
1 5 Radius a, center at (0,b); (%I2 +
= a2, surface area /,a'
2u(b asint)a dt = 4u2ab +
17 /
: 2rxd-dx =/
: +
= r2 2 r (write 2%- x2 = 1- (x - 1)2 and set z - 1= sine)
19 /Il22 r x J q d x (can be done)
2 1 Surface w a =;/ 2r:d-dz > I. = 2slnxjY = a, but volume = U ( $ ) ~=
~ aX
2s I,
2 r"
sin t d 2 sin2t + cos2 t dt = %sin t&CGZ dt = 2rJGFdu =
r u J S + 2 r sin-' *]il= 2 s + Z
y=,, b d n . Then the surface area is $:a 2": m m z (use table of integrals)
Discrete probability uses counting, continuous probability uses calculus. The function p(z) is the probability
b
density. The chance that a random variable falls between a and b is $' p(x)dx. The total probability is
p(z)dz = 1. In the discrete case C p , = 1. The mean (or expected value) is p = $ x p ( x ) d x in the
continuous case and p = C np, in the discrete case.
The Poisson distribution with mean X has p, = Xne-A/n!. The sum C p , = 1comes from the exponential
series. The exponential distribution has p(z) = e-= or 2eV2=or ae-=. The standard Gaussian (or n o r m a l )
distribution has G P ( z ) = e-z'12. Its graph is the well-known bell-shaped curve. The chance that the variable
falls below z is F ( z ) = +
p(x)dx. F is the cumulative density function. The difference F ( z dz) - F (z) is
about p(x)dx, which is the chance that X is between z and z dz. +
The variance, which measures the spread around p, is 02 = (x - p)2p(x)dx in the continuous case and
02 = C (n - Ic)2Pnin the discrete case. Its square root o is the standard deviation. The normal distribution
2 2
has p(z) = e-(X-p) 1 2 ~ 1 6 0 . If X is the average of N samples from any population with mean p and
variance 02, the Law of Averages says that X will approach t h e m e a n p. The Central Limit Theorem says that
8.4 Probability and Calculus (page 334)
the distribution for approaches a normal distribution. Its mean is p and its variance is 0 2 / ~ .
In a yes-no poll when the voters a m 50-50, the mean for one voter is p = 0(+)+ I(+)= ). The variance is
(0 - p)2po + (1- p)2pl = i.
For a poll with N = 100,F is 01- = There is a 95% chance that X (the
1
fraction saying yes) will be between p - 2F = ZJ - 1 and p + 2Z = + +.
m 1 =
Is 1,"a ~ e - ~ =d z[ - ~ e - ~ l ?+:J
~ - ~ d z f, =
=2 tan-lz*0J' e-f dz = 1- e-f; :J ae-ofdz = 1- e-a 17 &e-z/lodx = -e-'/10 00 1
lSj r 110 =,
19 Ezponcntial better than Poisson: 60 years -+ 1;
.01e-.~" dz = 1- e-.= = .45
2ly=~;thneareasnr~eachbccauscp-otop~thesameosptop+oand~easaddto1
+
28 -2pJ zp(z)dz 3 J p(z)dz = -2p p p2 . I = -p2 +
2 ~ ~ = 0 .1~ + 1 - ! + 2 - $ = 012; = (0 - 1)2: ++
(1- 1)2 ++
(2 - 1)2. = 5 - ?
27p=r1
29 Standard deviation (yes - no poll) <= = $ Poll showed = peaceful.
95% confidence interval is from -
81 95% confidence of unfair if more than =
602 + 3,
% &
or 93% to 100%peaceful.
= 2% away from 50% heads.
2% of 2500 = 50. So unfair if more than 1300 or less than 1200.
53 55 is 1.50 below the mean, and the area up to p - 1.50 is about 8% so 24 students fail.
below the mean and the area up to p - 1.30 is about 10%.
A grade of 57 is 1 . 3 ~
36 .999; .9991°00 = (1- &)lorn r~ $ because (1- !)" + $.
(4)
28 p = (pi + p 2 + s~ + . . . ) + ( ~ + ~ 3 + ~ 4 + . . . ) + ( ~ 3 + ~ 4 + - - . ) + - . . = ( I ) + +(:)+---=2.
30 p equals , h &, &, &, & in four tosses. It looks more bell-shaped with 16 tosses.
32 2000 f 20 gives 1700 t o 2300 as the 95% confidence interval.
34 The average has mean P = 30 and deviation a =
m- - 1. An actual average of = 31.25 is 1.25 a above
the mean. The probability of exceeding 1.25 a is about .1 from Figure 8.12b. With N = 256 we still
have 8000
256
= 31.25 but now a = 8 fl = $. To go 2.5 a above the mean has probability < .01.
36 (a) . 0 0 1 ( . 9 9 9 ) ~rr~.001(1-
~ &)lorn rr .001:. (b) Multiply the answer to (a) by 1000 (which gives 6)
since any of the 1000 players could have been the one to win. (c) The probability p, of exactly n
winners is "1000 choose n" times (.OOl)"(.Wg) looO-".This counts all combinations of n players times
the chance that the first n players are the winners. But "1000 choose n" = 1000(g99)..~(100~-n+1~
1(2).-.(n m n!
1
Multiplying by (.001)" gives p, rr 11 which is Poisson (= fish in fiench) with X = 1. With X times
1000 players, the chance of n winners is about sc-*.
8.5 Masses and Moments (page 340)
If masses m, are at distances z,, the total mass is M = x m n . The total moment around x = 0 is
M, = C m n x n . The center of mass is at 5 = My/M. In the continuous case, the mass distribution is given by
I I
the density p(z). The total mass is M = p(x)dx and the center of mass is at Z = xp(x)dx/M. With p = z,
I
the integrals from 0 to L give M = ~ ' / 2 and zp(z)dz = ~ ~and/ f =3 2L/3. The total moment is the same
as if the whole mass M is placed at X.
In a plane with masses mn at the points (z,, yn), the moment around the y axis is C m n x n . The center of
x
mass has f = mnxn/ E m n and g = C mnyn/ m n . For a plate with density p = 1, the mass M equals
the area. If the plate is divided into vertical strips of height y(z), then M = y(x)dz and M y = $ xy(x)dz.
For a square plate 0 5 z, y 5 L, the mass is M = L2 and the moment around the y axis is M, = L3/2. The
center of mass is at (f,ij) = (LIZ, LIZ). This point is the centroid, where the plate balances.
A mass rn at a distance z from the axis has moment of inertia I = mw2. A rod with p = 1 from z = a to
z = b has I, = b3/3 - a3/3. For a plate with p = 1 and strips of height y(z), this becomes I,, = I ~ ~ ~ ( x ) d x .
The torque T is force times distance.
-
+
8 M = 1 4 = 5; M, = l(1)+ 4(0) = 1,Mx = l(0) + 4(1) = 4;T = g
1 and 4
= 5.
1 a b ) ;MY =
1 0 M = 3(% 3xb(l- z)dz = 1% -7x 31 0= - and by symmetry M, = *;Z
b a = 9- a3
3ab 2
8.6 Force, Work, and Energy (page 346)
Work equals force times distance. For a spring the force F = k is proportional to the extension x (this is
.
1 3 This equals
Hooke's law). With this variable force, the work in stretching from 0 to z is W = kx dx = %hr
the increase in the potential energy V. Thus W is a definite integral and V is the corresponding indefinite
integral, which includes an arbitrary constant. The derivative dV/dx equals t h e force. The force of gravity is
8.6 Force, Work, and Energy (page 346)
In falling, V is converted to kinetic energy K = #mv2. The total energy K + V is constant (this is the
law of conservation of energy when there is no external force).
Pressure is force per unit area. Water of density w in a pool of depth h and area A exerts a downward force
F = w h A on the base. The pressure is p = wh. On the sides the pressure is still wh at depth h, so the total
force is I whl dh, where I is the side length at d e p t h h. In a cubic pool of side s, the force on the base is
F = wr3, the length around the sides is I = 4m, and the total force on the four sides is F = 2 m s 3 . The work
to pump the water out of the pool is W = $ whA dh = z1 w s 4.
.
12.4 ft lb; 2.424 .. ft lb 3 24000 lb/ft; 8 3 i ft lb 5 lox ft lb; lox ft lb 7 25000 ft lb; 20000 ft lb
9 864,000 Nkm 1 1 5.6. lo7 Nkm 13 k = 10 lb/ft; W = 25 ft lb 15 / 60wh dh = 48000cu, 1 2 0 0 0 ~
1 7 ~ W A HiwAH2
~; 19 96OOw 21 1 - ) 2 3 (800) (9800) kg 25 f force
--
4 W = J;(~OX - xS)dx = [10z2 - $1; = 116;V(2) - V(0) = 36 so V(2) = 41; k = = 20 - 3x2 = 8 at x = 2.
6 (a) At height h the burnt fuel weighs 100(&) = 4h so mass of fuel left = 100 - 4 h kg
(b) Work = $ Fdz = ~ t ~ ( 1 - 0 4h)gdh
0 = (1250) (9.8) Newton-km = 12,250,000joules.
8 The side length at height h is 800(1- A) 8 2 . The work is
= 800 - f h so the area is A = (800 - gh)
W = I whAdh = lOOh(800 - :h)ldh = 1 0 0 [ ( 8 0 0 ) ~ (-~ 1) 6~ 0 0 ( f ) v 3 + 4 1=
1 0 1 0 [ q - 16(5)5~+ y] = R]h.
10 The change in V = is AV = G ~ M ( & - rn) 1
= GmM- 20 = 22B&A R1 R1-10" The first factor
is the distance (20 feet) times the force (30 pounds). The second factor is the correction (practically 1.)
1 2 If the rocket starts at R and reaches x, its potential energy increases by G ~ m ( -b $). This equals $mu2
1 - 1 = ,K
(gain in potential = loss in kinetic energy) so H vM2 and x = (g v2
1 - .rM)-l. If the rocket
reaches x = oo then = & or v = d-
y = 25,000 mph.
1 4 A horizontal slice with radius 1 foot, height h feet, and density p lbs/ft3 has potential energy ~ ( 1 ) ~ h ~ d h .
Integrate from h = 0 to h = 4 : 1,' xphdh = 8*p.
16 (a) Pressure = w h = 62 h lbs/ft2 for water. (b) = so l = f h (c) Total force F = / whldh =
$:062h(th)dh = v ( 3 0 ) ' = 1,488,000 ft-lbs.
1 8 (a) Work to empty a full tank: W = $ W A H ~ = +(62)(25~)(20)'= 1110,000~ft-lbs = 973,000 ft-lbs
(b) Work to empty a half-full tank: W = I;,
wAhdh = = 2112, 5 0 0 ~ft-lbs = 730,000 ft-lbs.
2
20 Work to empty a cone-shaped tank: W = / wAhdh =:/ w d s d h = w m 2 T . For a cylinder (Problem
17) W = iwAH2 = wrr2$. So the work for a cone is half of the work for a cylinder, even though the
volume is only one third. (The cone-shaped tank has more water concentrated near the bottom.)
22 The cross-section has length 10 meters and depth 2 meters at one end and 1meter at the other end. Its area
is 10 times 1; = 15 m2; multiply by the width 4m to find the total volume 60m3. This is of the box
volume (10)(2)(4) = 80, so 31of the volume is saved. The force is perpendicular to the bottom of the pool.
(Extra question: How much work to empty this trapezoidal pool?)
9.1 Polar Coordinates (page 350)
Polar coordinates r and 8 correspond to z = r cos 8 and y = r sin 8. The points with r > 0 and 8 = r are
<
located o n t h e negative x axis. The points with r = 1and 0 8 5 r are located on a semicircle. Reversing
the sign of 8 moves the point (z, y) to (x, -y) .
Given z and y, the polar distance is r = dx2 +y2. The tangent of 8 is y/x. The point (6,8) has r = 1 0
and 8 = tan-lt. Another point with the same 8 is (S,4). Another point with the same r is (10,O). Another
point with the same r and tan8 is (-6, -8).
The polar equation r = cos 8 produces a shifted circle. The top point is at 8 = ul4, which gives r = 6 1 2 .
When 8 goes from 0 to 2r, we go two times around the graph. Rewriting as r2 = r cos 8 leads to the zy equation
x2 + y2 = x. Substituting r = cos 8 into z = r cos 8 yields z = cos28 and similarly y = cos 8 sin 8. In this form
z and y are functions of the parameter 8.
170<r<oo,-~<8< ~;0<r<oo,r~8<2r;~<r<~,0<8<2r;0~r<oo,-~<8~~
IQy=ztan8,r=zsecO 2 1 8 = ~ , a l l r ; r = 8 i 1n e + c o B e ; ~ = ~ ~ ~ 8 + ~ i n 8
2Sz2+y2=y 25z=r sin8c0~8,~=rsin~8,z~+~~=~
+ +
27 z2 y2 = x y, (X- ?)2 + (y - = ($)2
core
co82::in
29 x = co8e+8in 1 y = + S l (z2 y2)3 = x4
+ +
26 From x = cos2 8 and y = sin 8 cos 8, square and add to find x2 y2 = cos2 8(cos28 sin2 8) = cos2 8 = x.
+ + +
28 Multiply r = a cos 8 b sin 8 by r to find x2 y2 = ax by. Complete squares in x2 - ax = (x - -
and similarly in # - by to find (x - +
(y - t ) 2 = + t)
( i ) 2 . This is a circle centered at (; , with
radius r = -4 = I-.
50 The point z = cos2 8, y = sin2 0 is generally not at the polar angle 8. For example let 8 = = 30' : then
?
x = and y = f . The polar angle for this point has tangent = $, but the tangent of is 1
fie
Conclusion: an angle named 8 is not automatically the polar angle. See Problem 9.5.40.
52 The second figure is not a closed curve as it stands. As the parameter t keeps going, the spaces around
the circle fill up and the curve eventually closes (but the figure becomes less beautiful).
The circle of radius 3 around the origin has polar equation r = 3. The 45' line has polar equation 8 = r/4.
+
Those graphs meet at an angle of 90'. Multiplying r = 4cos 8 by r yields the xy equation x2 y2 = 4x.Its
graph is a circle with center at (2,O). The graph of r = 41 cost9 is the line x = 4. The equation r2 = cos28
+
is not changed when 8 -r -8 (symmetric across the x axis) and when 8 -r r 8 (or r -.* -r). The graph of
r = 1+ cos 8 is a cardioid.
The graph of r = A/(1+ e cos 8) is a conic section with one focus at (0,O). It is an ellipse if e < 1and a
+
hyperbola if e > 1. The equation r = 1/(1+ cos 8) leads to r x = 1which gives a parabola. Then r = distance
from origin equals 1- x = distance from directrix y = 1. The equations r = 3(1- x) and r = $(I- x) represent
a hyperbola and an ellipse. Including a shift and rotation, conics are determined by five numbers.
1Line y = 1 +
S Circle x2 y2 = 22 +
6 Ellipse 3x2 4y2 = 1- 2% 7 x, y, r symmetries
9xsymmetryonly 11Nosymmetry lSx,y,rsymmetries!
+ + +
x2 y2 = 6y 8x -.* (x - 4)2 (y - 3)' = s2,center (4,3) 1 7 (2,0), (0,0)
r = l - & B=S"*
2 , , r = 1+ $,8 = 2 ; ( 0 , 0 ) 21 r = 2 , 8 = f $ , f E , f s , f %
(xy)=(l,l) 25r=cos58has5petals 27(x2+y2-x)2=x2+y2
s 8) =~ - ~ , s i n 8 = ~ - + y = ~ , x = - S~S z = $ , r = - t
( ~ ~ + y ~ ) ~ = ( x 5~ l -c yo ~ 56.967
must try both tests.) Note that the maximum of y = rain8 = is y = 1as 8 + 0 : the line y = 1
is a horiarontd arymptote! As negative 8 approach rero, the spiral goes left toward the same asymptote y = 1.
1 4 r = 1- 2 sin 38 has y axia symmetry: change 8 to r - 8, then sin 3 ( r - 8) = sin(%- 38) = sin 38.
1 6 This is another case where t h e parameter t is not t h e polar angle. (The Earth completed a circle at
t = 1.)
1 8 (1= sin 28 and (1= cos 28 are lemniscates (or aspectaclesa). They meet when sin 28 = cos 28 or 28 = f
or9 8
or 8 = or 9 ( r can be positive or negative). They also meet at the origin r = 0.
20 r = l+cosB and r = 1 - c w 8 are cardioids (reaching right to r = 2,B = O and left t o r = 2 , 8 = r ) .
8
They meet when cos 8 = 0 at r = 1,8 = and r = 1 , B = 9.
They also meet at the origin r = 0.
221fcos8=f a n d c w 8 = 1 - r t h e n f = l - r a n d ? + 4 r - 4 = 0 . ~ h i s ~ i v e s r = - 2 - f i a n d
r = -2 + The first r is negative and cannot equal 1- cos 8. The second gives cos 8 = 1- r = 3 - fi
and 8 B SO0 or 8 B -80'. The curves also meet at t h e origin r = 0 and at the point r = -2,8 = 0 which
is & o r = +2,8 = r.
+
24 he limacon r = 1 bcos8 has z = rcos8 = cos8 bcos2 8 and + 2 = -sin8 - 2bcosBsin8. Then
9 +
= -coal - 2bcos2 8 2bsin28 which equals 1- 2b at 8 = r. T h e dimple begins at b = 2 1.
At b = 1it becomes the cusp in the cardioid.
26 The other 101 petals in r = cos 1018 are duplicates of t h e fist 101. For example 8 = r gives
r = cos 101%= -1which is also 8 = 0, r = +l. (Note that cw l O O r = +1 gives a new point.)
28 (a) Yes, z and y symmetry imply r symmetry. Reflections across the z axis and then the y axb take
(z, y) to (z, -y) to (-2, -y) which is reflection through the origin. (b) The point r = -1,B = satisfies
the equation r = cos 28 and it is the same point as r = 1 , B = 5.
S 0 ( a ) ? = 8 ( b ) z + y = lorrcosB+rsinB= l o r r = c o s O ~ s i n (O~ ) e l l i ~ s e z ~=+120~d~( ~ 0 ~ ~ 8 + 2 s i=n~1e )
32 (a) 8 = gives r = 1;this is z = 0,y = 1(b) The graph crosses the x axis at 8 = 0 and r where z = &
and z = 2. The center of the graph is halfway between at z = i(& - &) = ,a. The second focus is
twice as far from the origin at -*.
1-e
(Check: e = 0 gives center of circle, e = 1 gives second focus of
parabola at infinity.)
S4 r = A
and r = C+dcosO are the same if C = 1 and D = 5. For the mirror image across the y
axis, 8 becomes r - 8 and cos 8 changes sign.
86 Maxi-e
and ym, = =
'
y = l+ecose where $ = l+ecos O)Al +cos @+(Asin O)e sin 0
9.3 Slope, Length, and Area for Polar Curves (page 359)
A circular wedge with angle A8 is a fraction A8/2r of a whole circle. If the radius is r, the wedge area
1 4
is 2r 3A8. Then the area inside r = F(8) is $ i r 2 d 8 = $ ( ~ ( 8 ) ) ~The d ~ .area inside r = B2 from 0 to r is
1- 1
r S / l O . That spiral meets the circle r = 1 at 8 = 1.The area inside the circle and outside the spiral is 2 m.
A chopped wedge of angle A8 between r l and r2 has area i r Z ~ -8 ir:~t9.
The curve r = F(8) has z = r cos 8 = F (8)cos 8 and y = F (#)sin 8. The slope dyldz is dy/d8 divided
by dx/d8. For length (ds)" = ( d ~ ) (dy)2
~ + +
= (dr)2 (rd8)=. The length of the spiral r = 8 to 8 = r is
9.3 Slope, Length, and Area for Polar Cwves (page 359)
/ d a d o . The surface area when r = 8 is revolved around the z axis is / 2 r y ds = / 2x8 sin 8 d g d 8 .
The volume of that solid is / ry2dz = / re2 sin2# (cot3 8 - t9 sin 8)dB.
- l+tanl 0 F F
- /l+tan' B/F1 = Fe
22 r = 1- cos 8 is the mirror image of Figure 9 . 4 ~across the y axis. By Problem 20, tan J, = 5=w.
This is ,mtan $. So $
1 sin' Q
= = (check at 8 = n where $ = f).
24 By Problem 18 $ = c0'e+tane~-8ine~
- cos 8 tan @-sin 8 =- .cos 6 -2 sm 0 = -a
sin28 6 at 8 = t .At that point
= -1 =r COB 8=
~ y = r sin8 = cos 3 sin: = ;($).
cos2 f = ( 9 ) and The tangent line is y - ds = - L ( ~ - 51
fi
26 r = sec 8 has = sec 8 tan 8 and +
= dsec2 8 sec2 8 tan2 8 = d a = see2 8. Then arc length
= /;I4 sec2 8 d8 = tan 2 = 1. Note: r = sec 8 is the line r cos 8 = 1 or z = 1 from y = 0 up to y = 1.
28 r = B2 has $ = 28 and 2
= d w . Then zwc length = :/ 8 d m d 8 = [f (02 + 4)312];
= f[($ + 4 ) W - 43/21.
SO ds = dcosa 0 + sin2 8d8 = dB and surface area = $ 2 r y ds = / 2n r sin 8 ds = I:2/ 2, cos 8 sin 8 d8 = R.
Sf r = l + c o s 8 has 2
= 4 ( 1 + 2 cos 8+cos28) +sin2# = d m .~ l s yo = r sin 8 = ( ~ + C O#)sine.
S
Surface area $ 2n y ds = 2 r J Z l + +
: (1 cos 8)S12sin 8 dB = [ % a ( - $) (1 cos 8)5/2]b =
2rt/Z($)25/2 = 921.s
9.4 Compkx Numben (page 364)
+
The complex number 3 4 i has real part 3 and imaginary part 4. Its absolute value is r = 5 and its complex
conjugate is S - 4i. Its position in the complex plane is at (S,4). Its polar form is r cos B+ir sin 8 = reig (or beig).
Its square is -7 - l4i. Its nth power is P e i n b .
+ +
The sum of 1 i and 1 - i is 2. The product of 1 i and 1 - i is 2. I . . polar form this is fiei"/' times
fie-iu14. The quotient ( 1+i)/(l -i) equals the imaginary number i. The number ( 1 equals 18. An eighth
+
root of 1 is tu = (1 i ) / f i . The other eighth roots are w 2 , u s , .,w7,w8 = 1.
To solve day/dt8 = y, look for a solution of the form y = ect . Substituting and canceling e" leads to the
equation e8 = 1. There are e i g h t choices for c, one of which is (- 1 + ;)/a.
With that choice I ed I= e-@.
The red solutions are Re ed = L - ~ / ~ ~ c o B - &and Im eb = e - t / f i d n
55
is ( 2 + i ) ( 2 -i) = 4 - i2= 5.
+ + +
16 (cos 0 + i sin 0)3 = (cos3 0 3i2 cos 0 sin2 0) (3i cos2 0 sin 0 i3sin3 0). Match with eSie to find real part
cos 30 = cos3 0 - 3 cos 0 sin2 0 (or 4 cos3 0 - 3 cos 0). The imaginary part is sin 38 = 3 cos2 0 sin 0
- sin3 0 (or 3 sin 0 - 4sin3 0).
18 The fourth roots of rei0 are r'14 times eieI4, ei(e+2u)14,ei(e+4r)14,ei(e+6r)/4.Multiply (r'14)l to get r.
+ +
Add angles to get (40 12x)/4 = 0 3x. The product of the 4 roots is = -rei0.
+ +
20 (e")ttt e" = 0 gives (c3 1)e" = 0. Then c3 = -1 = eiu and c = eiuI3, ein, ei6r/3. The root c = eiu = -1
gives y = e-t. The other roots give y = e('+fiiItl2 and y = e('-fii)'12. (Note: Real solutions are:
y = et12cos $t and y = et12sin qt.)
22 (ee')" +6(ee')'+5eC' = o gives c2+6c+5 = 0 or ( c + 5 ) ( c + l ) = 0. Then c = -5 yieldsy = and c = -1
yields y = e-t.
24 c2 - 2c + 2 = 0 gives c = 1k i. Then the real part of e('+'lt is y = etcos t and the imaginary part is etsin t.
+ + +
26 e(-l*)' = e-' cost ie-'sin t spirals in to ec = e-' cos 1 ie-'sin 1w .2 .3i at t = 1.
28 2 5
= iy leads to y = eit = cos t + i sin t. Matching real and imaginary parts of (cog t +i sin t) = i(cos t +i sin t)
yields $ cos t = - sin t and $ sin t = cos t .
+
SO i(eie + e-") = i(cos 0 + isin 0 cos(-0) + isin(-0)) = i(2cos 8) = cos 0. Similarly sin 0 = &(eie - e-ie).
+ + +
. rectangular form is rR cos(0 4) irR sin(8 4). This
32 rei0 times Rei& equals ( r ~ ) e ~ ( ' + * )The
+ + +
equals ( r cos 0 + i r sin 0) (R cos 4 i R sin 4) = rR(cos 0 cos 4 - sin 0 s i n 4) irR(cos 8 sin 4 sin 0 cos 4).
+
3 4 Problem 30 yields cos iz = i(ei(ix) e-i(ix)) = i(e-" + &
ex) = cosh z;similarly sin i x = (ei(ix) - e-i(ix)) =
&(e-= - ex) = i sinh z. With x = 1 the cosine of i equals i(e-' + el)= 3.086. The cosine of i is larger
than I!
CHAPTER 10
z = 1 is t a n m l l = 1- 3
g g , ggg 15
27 2(.1234.. .) is 2 -
3 1 - l n ( l - .l) = -1n.9
4 1+(1-z)+(l-z)2+-.=
two series.
+ + +
1+ 5
INFINITE SERIES
33 Ainu
2
+
.9
3 9 All products like alb2 are missed; (1 1)(1+1) # 1 1
43 In 3 seconds the ball goes 78 feet
*
+
35 (n I)!
3 7 y = 1-bz
+
= Iz *) integration gives in z = z -
The geometric series 1 z z2 . . . adds up to 1/(1-x). It converges provided 1x1 < 1.The sum of n terms
is (1- xn)/(l- x). The derivatives of the series match the derivatives of 1/(1- z) at the point x = 0, where
.
the nth derivative is n! The decimal 1.111.. is the geometric series at z = .1and equals the fraction 1019. The
.
decimal .666.. multiplies this by A. The decimal .999.. is the same as 1. .
+ +
1Subtraction leaves G - zG = 1 or G =
17 &3
25 ( I + Z + X ~ + Z ~ . . . ) ( ~ - Z-z3-)
& . (l-ib,,
1 - 2 81,
-
&
2
1-lnz
3 f,,,,
7 .I42857 repeats because the next step divides 7 into 1 again
1 3 87.123 * 19 '"
+ X ~ =i+z2+z4++-
0.1-.0123...i~l-L
21
,
. 4.100.34
1
,, ,,
+
The derivative of the geometric series is 1/(1-x ) =~1 ZX 3x2 . - . . This also comes from squaring the
The differential equation
dy/dz = y2 is solved by the geometric series, going term by term starting from y(0) = 1.
+ . +
The integral of the geometric series is -In(l - x) = x x 2 / 2 - - . At z = 1 this becomes the harmonic
+ + . +
1 we find ln2 = f (412/2 ( f ) ' / ~ . . . The change from x to -z produces
In the geometric series, changing to z2 or -z2 gives 1/(1- z2) = 1+x2 +x4 . . - and 1/(1 x2) =
l-x2+x4-... , Integrating the last one yields z - $z3 +
1 51 - ...).
5 2.1+3.2~+4-3z2+...=
3 1-+ -
10+2(6)&
S
=dofeet
+-.-I.
At z = 0 this is -cp = -00.
6 Multiplying (1- x + z2 - . .) times (1 z z2 . .) term by term, the odd powers disappear and one of
+ + +
each even power survives. The product is 1 z2 z4 - .. which is f 1+z -
1-2'
the product of the
1 = .076923076923. , - so that c = 76923 and N = 6 and n = 6 (repeat after 6 digits starting immediately).
8 13
1 0 The decimal .010010001~ is not repeating because the number of zeros increases. So it cannot be a fraction.
1 2 The number 1.065065. . equals 1
14 (l+$+&+...)(l+&+&+...)
16 1 - 2 ~ + ( 2 2 ) ~ - - . - 1
18 iZ - tz2 + iX3
'
= - 0= 1
+ &+
= I+,2
- . . . = 2 - (;)a + ( 3
(looo)
3
65
3 ,+
which is I + &(-)
- ... =
+
4
I+,
.
=
.
Z+X'
a
+
1
,
=I+
4 +..-Expressedinfractionsthisis
X
1064
= -.
-@
3
+
i z 5 - . = tan- lx. The angle whose tangent is
+
-80
loo (1- +# - 81
292'-
(page 373)
- . . . + C and at
10.2 Convergence Tests: Positive Series (page 380)
2 6 j ( 1 + z2 + % 4 + . . .)dz = z + 2 3
+ 15.+ - +=. # I n by equation (13). This is IA d z which is
also tanh- 'x ( + C).
2 8 ( 1 + z + z 2 ) ( l + z + z 2 ) ( l + z + z 2= ) (1+2z+3z2+.-.)(1+z+z2) =1+3x+6x2+-..
+
30 . l + .02 .003 - . = + &+ +
2 ( & ) l + 3(&)3 . - . = by equation (5) = 10
= 81. &&
52 & - ;(&)l +
$(&)3 - = by equation (lob) = In ( 1 &) = I n 1.1.
. - a +
3 4 1 - 1 1 1 ' 2 - . . . = fi[&
3(3) + ~ ( 5 ) - $(&)3 + i(5)'
- - . - ]= by equation (18) = f i t a n - ' -L- a-&g.
36 y'(0) = [y(0)12 = b2; y"(0) = 2y(0)y1(0) = 2bS; y"'(0) = (from second derivative of y' = #) =
+
2y(0)y"(O) 2y1(0)y' (0) = 6b4. Then y(z) = b b2z 2b3($) +
6b4($) +
... = + +
+
b(1+ bz b2z2 b3z3 + +
-) which is y ( x ) = b --.
+ + +
38 The mean value is p = f 2 ( i ) ( f ) 3 ( i ) 2 ( i ) . - = (by equation (5) with z = = f *. = 4. i)
Why should you have t o wait for the fourth deal t o get the best hand??
4 0 Note: The equations referred to should be (10) and (13). Choose z = $ in (10a): f i ( f ) 2 . . . = - In + + 5.
In equation (13) choose z = so that i= 3. Then 2 ( i $(4)' + $(i)3
+
. .) = ln3. This +
converges faster because of the factor ($)2 between successive terms, compared t o $ in the first series.
The series are e q u a l because - In $ = In 3.
4 2 Eguation (18) gives tan-' rr - & +& +
w .10000 - .00333 .00002 rr .O9669 (which is .0967 t o four
decimal places).
4 4 The series 1- 4+
$- +
. . adds to .693. The fact that the first digit is 6 is settled when the sum stays
below .7, a t about the term +A
(other answers are equally acceptable!). The 50th power of equals
the 100th power of Also i. = 5 &
when an = 2"' or n l n a = 1001n2 or n = 1 0 0 E .
4 6 Equation (20) is r = 4(tan-' tan-' 4+ 5)
rr 4 ( 3 - - $(i)3
+ +(i)' )(i)7
+ $(i)g
- A($)''
+ + + +
+$ - f ($)3 j!(f)' - ) ( f ) 7 ) = 2 - .I6667 .02500 - .OO446 .00087 - .00018 1.33333 - .04938 .00329 +
-.00026 = 3.1415+ = 3.142. Note: &($)I3 and L ( l ) g will increase the total toward 3.1416.
9 3
48 x$ = (take z = e' in equation (10)) = -h(l - el).
+ +
The convergence of a1 a2 . - is decided by the partial sums s, = a1 . - - an. If the s, approach s, + +
then x a, = s. For the g e o m e t r i c series 1 z + +
. the partial sums are s, = (1- xn)/(l- x). In that case
s, 1/(1- z) if and only if 1x1 < 1. In all cases the limit s, + s requires that a, -+ 0. But the harmonic series
-+
The comparison test says that if 0 5 a, 5 b , then xan c o n v e r g e s if xbn converges. In case a decreas-
ing ~ ( z agrees
) with a, a t z = n, we can apply the i n t e g r a l test. The sum C a, converges if and only if
y ( x ) d x converges. By this test the pseries l / n P converges if and only if p is g r e a t e r than 1. For the
+ +
harmonic series (p = I), s, = 1 - - - l / n is near the integral f (n) = In n.
The r a t i o test applies when ~ , + ~ / a ,--+ L. There is convergence if ILI < 1, divergence if ILI > 1, and no
decision if L = 1or - 1. The same is true for the root test, when (a,)'/" -+ L. For a geometric-pseries
10.2 Conwgence Tests: Positive s& (page 380)
+
/ u ~ x ( n l)P/d'. Its limit is L = x so there is convergence if
combination an = sn/np, the ratio u ~ + ~ equals
+
1x1 < I. For the exponentid ez = C xn/n! the limiting ratio an+i/an or z/(n 1) is L = 0. This series always
converges because n! grows faster than any snor np.
There is no sharp line between convergence and divergence. But if C bn converges and an/bn approaches
L, it follons from the -1 comparison test that C an sbo converges.
+ +
1$ f - * - is smaller than 1 $ + +
2n 1
S a n = s n - S n - l = ~ 1~ s S = l ; ~ n = I , S = ~ ; n+l ~ n = h ~ - h ~ = h & , d = h 2
;,+,,
10 (Just drop the 100.)
l2 diverges by comparison with the smaller series C which diverges. Check that
n2 10 is less than n 10.+
14 C converges because it is less than C = &
which converges.
5
1 6 C cos(!) converges because it is below C 5.
!
18 C s i n 2 ( i ) converges because sin $ < gives sin2(:) < and C5 converges.5
20 z
C
converges because $40, which means that eventually ne < ?en and the series compares with
which converges.
22 The limit of &/;;C = . .
= L = 1.SOthe ratio test (and root test) give no decision.
24 The terms are (%)' = (1 - i ) " -r c-' so the ratio approaches L =
e-1
e-l
= 1. (Divergent series because its
terms don't approach sero.)
26 The ratios = -$& approach L = 0.
28 The ratios '=/$ = (*)" = ( F ) - n approach L = e-l.
+
30 (a) Put (f - f ) ( f - )! + +
... = 1together with ( i - f ) (f - i)+.. = to obtain s = 1+ 21 = 2.
b s
(b) l n i + l n $ + - . . + l n & = l n ( ? ) ( ~ ) . . - ( & ) = l n & approacheslnO=-oo (nosums).
10.3 Convergence Tests: AH Series (pge 384)
converges!)
In lon = n ln 10 so the sum is & !
C = oo (harmonic series diverges)
n-'In approaches 1 so the series cannot converge
9
C diverges by comparison with C ;fi if p 5 1. For p > 1 the terms % are eventualy smaller than
with 1< P < p. So 9
converges if p > 1.
C fi converges if g > 0 by the ratio test: a/& = ( F ) p . h + L = 0. 68 NO,
z
The series a, is absolutely convergent if the series z
lan( is convergent. Then the original series a, is
also convergent. But the series z
a, can converge without converging absolutely. That is called conditional
convergence, and the series 1- 1 5 +
1 - . . . is an example.
decreasing and the partial sums s 2 , ~ , are increasing. The difference between s, and 3,-1 is an. Therefore
the two series converge to the same number s. An alternating series that converges absolutely [conditionally]
[c(-
(not at all) is ~ ( - l ) ~ + l / n ~~ ) ~ + l /( n
~ (] - l ) ~ + ' ) .With absolute [conditional] convergence a reordering
cannot [can] change the sum.
1-coe 1
(trig identity) = tan :( - i ) ; s =
= cot x in
= - log(1- e') by 10a in Section 10.1;
take imaginary part
C* : converges conditionally (passes alternating series test) but not absolutely: C fi diverges
C 5 converges (ratio test: = & -r 0) so there is absolute convergence.
C(- l)"+lsin2 n diverges (terms don't approach zero)
x(-l)"+'+ : no absolute convergence because sin2 n > ;half of the time and x & diverges.
The terms alternate in sign but do not decrease steadily; still I believe there is conditional convergence.
C(- '
I)"+ 2'1" diverges (terms don't approach zero)
nl/" decreases steadily to 1so the alternating test is passed: x(-l)"+'(l- nl/") converges conditionally.
But nl/" > el/" > (1 + i) so that 11- nl/"l exceeds $ and there is no absolute convergence.
Yes, the sum x(-5) converges absolutely.
The terms alternate in sign but do not decrease to zero. The positive terms t ,?, A, i
. . . approach and so
does the sequence f , %, ,. &
The term after ss = 1- $-+ +i g= is a6 = -+. Then later terms bring the sum upward. So the
sum s = In2 is between g i 5
- = and g.
The difference between s and sloe is less than 7 1 the, next term in the series (because after that term
I01
comes - & and the sums stay between sloe and slol).
The error 101' - 102' + in the alternating series is smaller than the error in the positive series.
The series a1 +a2-a3+ar +a5 -ae +. is sure to converge (conditionally) if 0 5 a3,+3 < < as,
for every n. Then it passes the alternating series test when each pair of positive terms is combined.
(The series could converge without passing this particular test.)
26 The series 1- i+5 -I 4 +i +
-16 . . . = In 2 is in Section 10.1. Take half of every term and also insert zeros:
o+;-0- f + 0 + i - . . . = i l n 2 . ~ d the d twoseries term byterm: 1 + 0 + $ - ++ $+o+..-=
In 2, as the problem requires. This is allowed because the partial sums sk and 8:: of the first two series
add to the partial sums s, of the third series. Notice something strange: The third series can also be
produced from the first series only, by rearranging (two positive terms between negative terms). With
10.4 The Taylor Seriea for e2, sin x and COB x (page 390)
$ - - & - . = 13 - 1-)
A
The smallest positive number must include +$; then choose all minus signs: -. =
$ - i= 8. (This is for the Cantor set centered at zero. Add $ to obtain the number $ in the usual
Cantor set between 0 and 1.) With alternating signs the sum - ? 1-
27
+ a
= += '.
4
1+ 3
If a, converges then its terms approach zero: in particular lanl 5 C for some number C. Then C anzn
converges absolutely by comparison with C Clxln = $&.
10.4 The Taylor Series for ex, sin x and cos x (page 390)
The Taylor series is chosen to match f (x) and all its derivatives at the basepoint. Around x = 0 the series
+ +
begins with f (0) f' (0)x I f ' ' (0)z2. The coefficient of zn is f (0)/n!. For f (x) = e2 this series is xn/n!.
+ . +
For f (z)= cos x the series is 1- x2/2! x4/4! - . . For f (z) = sin z the series is x - x3/3! .. .. If the signs
+
were positive in those. series, the functions would be cosh x and sinh x. Addition gives cosh x sinh x = ex.
In the Taylor series for f ( x ) around x = a, the coefficient of (z - a)" is b , = f@)(a)/n!. Then bn(z - a)"
has the same derivatives as f at the basepoint. In the example f (x) = x2, the Taylor coefficients are
+
bo = a2, b1 = 2a, b2 = 1. The series b0 bl(z - a) b2(x -+ agrees with the original x2. The series for e2
around x = a has bn = ea/n!. Then the Taylor series reproduces the identity e2 = (ea)(&x-a).
We define e2, sin z, cos x, and also eie by their series. The derivative d/dz(l + z + ?x2 + . . -) = 1+ x+
translates to dl&($) +
= $. The derivative of 1 - ?x2 . is -x x3/3! - + . Using i2 = -1 the series
1+ i9 + +
+ . . splits into eie = cos 9 + i sin 8. Its square gives eaie = cos 28 i sin 28. Its reciprocal is
e-a = cos 0 - i sin 0. Multiplying by r gives reie = r cos 6 + ir sin 6, which connects the polar and rectangular
10.4 The T a y h Series for e2, sin x and cos x age 390)
forms of a complex number. The logarithm of eie is iB.
11 + 2 x + (az)l
,I + .;derivatives T ; 1 2 $ + + + 3 Derivatives in; 1 i x . + +
6 Derivatives 2nn!; 1 22 4x2+ + +
. 7 Derivatives -(n - I)!; -x - f - - $
g y = 2 - e z = 1 - x - ~ -2!. . . l l y = ~ - g6+ - - . = s b x 1 s y = x e x = x + x 2 + ~ +..-
I S I + ~ X + X ~ ; ~ + ~ ( X - ~ ) + ( X - ~ ) ~I~-(x-I)~ i9i-(~-i)+(~-i)~-...
2 1 (x - 1) - k$ + k$E - . . . = h ( l + (x - 1)) 2 3 e-lel-z = e-l(l- - 1) + -
26x+2x2+2x3 27 ?-g+=!- 720a 2 g x - &18+ i "600 sl1+x2+g SSI+X-$
3500~10pe;l + ~ ( x - l ) 7
.ib - - i b
- + 39x+$+g 11l + x + $ 4s 1+0x-x~
4 5 case = -,sine = -A 2i 47 99th powers -1, -i, e3"14, -i
49 e-iu/3 and -1; sum sero, product -1 6 3 iz, i; +
2ri 56 2e2
2 s b 2 x = 2 x - ~ + M -61. . . s o t h a t ( s b 2 x ) m = - 2 3 + T25x=
-...= -8 at x = 0. This agrees with
the chain rule for (sin 22)'". Also sin 2 r = 2 r - +
51
- . = 0.
4 f = 1, f f = &., f" = $ *, f"' = *,. ... Set z = 0 : f = 1,f' = -1, f" = 21, f'" = -3!, . .
1+2 1+2 1+2 1+2
The Taybr series is & + +
= 1- x $x2 - $x3 . . . = x&o(-l)nxn.
.
6 f = cosh x , f' = sinh x, f" = cosh z, . Evaluate at x = 0 : f = 1,f' = 0, f" = 1,. .. The Taylor series is
coshx= 1+xq;- +
2 x4 +*.*.
&,f" = -+1+2)
8 f = l n ( l + x), f' = , (one step behind Problem 4). Evaluate at x = 0 : f = 0, f' = 1,
a! + % - . = x - % + J-T+"'.
f" = -1, f'" = 21, . . The Taylor series is l n ( l + x) = x - 2 x3 x 4
1 0 d = cy + s, y" = cg', y"' = cy", . With yo = 0 this gives yb = s, y:
= cs, y r = c2s,. .. The Taylor series
i s y ( x ) = s ~ + c s $ + c ~ s ~ + ~ ~ ~ = ~ [ c z + c ~ $ + c d~ e~ + ~ -1)-
~ ] = ~
1 2 y' = y yields y" = y' = y and y"' = y .
Then y and all its derivatives equal 1 at z = 3. The Taylor series
+ + +
is y(x) = 1 (x - 3) h ( x - 3)2 . . = ex-3.
1 4 At x = 0 the equation gives y" = y = 1, y"' = y' = 0 and y"" = y" = y = 1 (even derivatives equal 1,
odd derivatives equal 0). The Taylor series is y(x) = 1 $ $ + + + +
. = I(- a( e-X) = cosh x. -
16 x3 and its derivatives at x = a are as, 3a2, 6a,6,0, . . .. The Taylor series is a' + 3a2(x - a)+
p- ( x - a)' +
#(x - a)' which agrees with x3.
1 8 At x = 2 r the cosine and its derivatives are 1,0, -1,0,1, . . . The Taylor series is cos x = 1- +
fZ=kf
4! --
. At x = 0 the function cos(x - 2r) and its derivatives again equal 1,0, -1,0,1, .
Now the Taybr series is cos(x - 29r) = 1- x 2 + x 4 - .
20 & haa derivatives &, &, . .. At x = 1those equal 1,1,2,6,. .. and the series is
2-x= l + ( x - l ) + ( x - 1 ) 2 + ( x - 1 ) 3 + ....
22 At x = 1the function x4 and its derivatives equal 1,4,12,24,O,O, . .. The Taylor series has five nonzero
terms: x4 = i + r ( x - I) +q(x-
I) 2 g(x-
24 +
I) 3 = +
24( x - 1l4.
2 has derivatives 2e2', 4e2*, 8e2', .. Evaluating at x = 1 gives e2, 2e2, 4e2, 8e2, . . The
24 The function '
2
+
Taybr series is eax = e2 2e2(x - 1) 4e2 + 8e2 + + (which is c2 times e2(~-')).
2ecosfi=1-L&F+L$
a! -... =1-:+& -... . (Note that sin fi would not succeed; the terms
fi,(a3,.
. are not acceptable in a Taylor series. The function has no derivative at x = 0.)
.in=
-
28 y--
- 2- ++&,-... -- I - - +x 2- x 4 -...,
2 6 120
30 sin z = P - H I+ . .. = x2 - x6 xl0 - . .. .
6 120 =+mi5
32 bL = ellnb = l + x l n b + f (xlnb12 + a m - .
10.5 Power Series (page 395)
3 4 1x1 = {
-x f o r x < O
+x for x > 0
so that 9
=
-1 f o r x < O
+1 f o r x > O
. When x is negative and n is not a whole number,
xn is a complex number. But still 5
= nlxln-'einn (-1) = nlxln-' (ein)n-' = nxn-I.
+ +
36 2%= exin2= 1 x 1n2 $(xln 2)' +
h(~ln2)~ +
. . . (OK to compute derivatives).
&
38 Compute sin-' x and its derivatives at x = 0 : sin-' x = 0, 1 x = 1, x(1- x2)-3/2 = 0,
(1- x2)-3/2 + 3x2(1 - x2)-5/2 = 1,gx(l - x2)-5/2 - 1523(1 - x2)-7/2 = 0, g(1 - x2)-5/2 + . . . = 9.
+ + +
The Taylor series for sin-' x starts with 0 x 0 ax 1 3 0 m9 x5. + +
4 0 Compute ln(cos x) and its derivatives at x = 0 : In 1= 0, - = - tan x = 0, - sec2 x = -1, -2 sec2 z tan x
zx +
= 0, -2 sec4 x - 4 sec2 x tan2 x = -2. The Taylor series for ln(cos x) starts with - 1 2 0 - ~x 2 4.
42 Compute tanh-' x (or $ l n ( z ) : Section 6.7) and its derivatives at x = 0 : tanh-' 0 = 0, = 1,
+
2x(1- x2)-2 = O92(1- x2)-2 4x2(1 - x2)-3 = 2. The series for tanh-' x starts with x + 0 + ~x 2 3.
+
4 4 Compute sec2 x and its derivatives at x = 0 : sec2 0 = 1,2 sec2 x tan x = 0,2 sec4 x 2 sec2 x tan2 x = 2.
+ +
The Taylor series for sec2 x starts with 1 Ox $x2 = 1 x2. +
+
4 6 (ei8)2 = e 2 i ~equals cos 28 i sin 28, so neither of the proposed answers is correct.
+
4 8 (a) e2ni/3 = cos ?f i sin ?f - 1+
- -2
iq-
(b) (e2ni/3)3 = e6ni/3 = 2ni -
e -l(c)(-i+iq)(-$+iq)=
'
4
- - 4-g=-i-$.Multiplybyanother(-k+i$)toget
2 4
f -i2i=1.
+
50 (2ei"/3 4 = e2ni13 and also (1 &) (1 &) = 1 2& + + - 3 = -2 + 2 6 i ; (4ein14)2= 1 6 e ~ " / ~
+ +
and also ( 2 f i i2&)(2fi /Z i2&) = 8 16i - 8 = 16i.
+
52 Write (e'8) (e-") = ei(8-t) in rectangular form: (cos s i sin s) (cos t - i sin t) = cos(s - t) +i sin(s - t). Collect
+
real and imaginary parts: cos(s - t) = cos s cos t sin s sin t and sin($ - t) = s i n s cos t - cos s sin t.
5 4 If e = $ then the number N = p![$ - (1- 1 l!+ 2! . . - f f
1 P
)] is an integer, because all d e n o m i n a t o r s
g o evenly i n t o t h e p! t e r m . But in parentheses is an alternating and decreasing series approaching
e-' = $.The error is less than the last term -$ so IN1 < 1. The only possible integer N is N = 0 which
is not correct. The contradiction means that e = E9 was not true: e is not a fraction.
If 1x1 < 1x1 and anXn converges, then the series a,xn also converges. There is convergence in a
symmetric interval around the origin. For z ( 2 x ) " the convergence radius is r = %.For x
xn/n! the radius
is r = oo. For x ( x - 3)" there is convergence for lx - 31 < 1.Then x is between 2 and 4.
x
Starting with f (x), its Taylor series anxn has an = f (0)/n!. With basepoint a, the coefficient of ( x - u ) ~ is
f(n)(a)/n!. The error after the xn term is called the remainder Rn (x). It is equal to f(n+l) (e)(X - a)n+l/(n I)! +
+
where the unknown point c is between a and x. Thus the error is controlled by the (n 1)st derivative.
The circle of convergence reaches out to the first point where f(x) fails. For f = 4/(2 - x), that point is
x = 2. Around the basepoint a = 5, the convergence radius would be r = 3. For sin x and cos x the radius is
r=oo.
The series for fiis the binomial series with p = $. Its coefficients are an = (Z)(-Z)(-z)
1 1 3 - In!. Its
convergence radius is 1. Its square is the very short series 1 x. +
10.5 Power Series (page 395)
1 7 f (c) = cos c < 1; alternating terms might not decrease (as required)
'n+l .p+1
1 9 f = Ll-z,lRnlS = G;(1-c)4 = I- 2
2 1 f(n+l)(x) = ,*, lRnl 5 ,&(A) 4 0 when x = ? and 1- c >
23 R2 = f (x) - f (a) - ft(a)(x - a) - ?f"(a)(x - a)2 so & = & = = 0 at x = a, R r = f'";
Generalired Mean Value Theorem in 3.8 gives a < c < c2 < cl < x
2 6 1 + ~ ~ ~ + f ( x 2~7 ()- 1 ~) ~ ; ( - l ) ~ ( n + l )
29 (a) one friend k times, the other n - k times, 0 5 k 5 n; 21 88 (16 - 1)'14 B 1.968
+
5 (1 + 1 ) = ( . ) ( . l ) ( 1 B 1.1105 + +
37 1 $ $;r = 4 1 z x2 fx3 + + + ix4
4~x2-$x4+&z6 461+:+?+$ 47.2727 49-'-'=-'
6 3 6lr=l,r= t-1
+ + +
40 f (g(z)) w a0 a1(blx b2x2 . -) a2(blx b s 2 + + + +
. rn a0 q b l x (alb2 ajbl)x +2 2. +
Test on f = k w 1 - z + x 2 ( w h i c h h a s a o = l , a l =-1,a2= 1) a n d g = & w x + x 2 (whichhas
b1 = 1 = ba). The formula correctly gives f (g(x)) = 1- x + (0)x2.
42 By Problem 40 with a0 = 0 the series starts with f (g(x)) = alblx + (alb2 + a2b:)x2. This agrees with
f(g(z)) = z when bl = &
and b2 = -$ -2.
= The example f = ez - 1 = x + "211+ ... has a1 = 1 and
i t -1
'
a2 = so that b1 = and b2 = 3.These are the coefficients in f - (x) = ln(1 + x) = x - $ + . ..
44 Quick method: Multiply (1 - x)(l z3 x6 + + +
. = 1- x x3 - x4 x6 - x7 ..
a) + + +
Slow method: 3 = -= (geometric series for -x - x2) = 1- x - x2 (x x2)2 - (x + + + +
+
(X x2)4 - (X + 1- +
ox2 ~3 - x4 ox5.+ +
46 $,'e-z3dxw ~ , ' ( ~ - ~ $
2 +- E! + E! -d
=)d
x ! 1- 1 + 1
5.2 - 1
7,e - +&
= .747 to 3 places.
6 24
48 At x = -1 the alternating series 5= n
+i- +.
converges (to ln(1- x) = ln2). The
= -1
derivative C x n - l = 1+ x + x 2 + +
= 1- 1 1- . diverges. Both series have r = 1; one series
converges at an endpoint of the interval -1 < x < 1 and the other doesn't.
.
50 If akl" approaches L then (anxn)'In approaches f By the root test the series anxn converges when
I < 1 and diverges when If 1 > 1. So the radius of convergence is r = L.
11.1 Vectors and Dot Products (page 405)
A vector has length and direction. If v has components 6 and -8, its length is ivl = 10 and its direction
vector is u = .6i - 4.The product of lvl with u is v. This vector goes from (0,O) to the point x = 6, y = -8.
A combination of the coordinate vectors i = (1,O) and j = ( 0 , l ) produces v = x i + y j .
To add vectors we add their components. The sum of (6, -8) and (1,O) is (7, -8). To see v i +
geometrically, put the tail of i at the head of v. The vectors form a parallelogram w i t h diagonal v i. +
(The other diagonal is v - i). The vectors 2v and -v are (12, -16) and (-6,8). Their lengths are 20 and 10.
In a space without axes and coordinates, the tail of V can be placed anywhere. Two vectors with the same
components o r t h e same length and direction are the same. If a triangle starts with V and continues with
W, the third side is V + W. The vector connecting the midpoint of V to the midpoint of W is &(V W). +
That vector is half of the third side. In this coordinate-free form the dot product is V W = IVI 1 cos 6. 13
+ +
Using components, V W = Vl Wl V2 W2 V1 Ws and (1,2,1) . (2, -3,7) = S. The vectors are per-
pendicular if V .W = 0. The vectors are parallel if V ita a multiple of W. V .V is the same as 1 ~ 1 ' . The dot
+ +
product of U V with W equals U.W V . W. The angle between V and W has cos t9 = V W/JVIIWI.
+ +
When V . W is negative then t9 is greater t h a n 90'. The angle between i j and i k is r / S w i t h cosine 4.
The Cauchy-Schwan inequality is IV W I 5 IVI I W 1, and for V = i j and W = i + +
k it becomes 15 2.
1(0,0,0);(5,5,5);3;-3;cos6=-1 32i-j-k;-i-7j+7k;6;l;cosB=~
6(y,-ul);(y,-ul,0),(u3,0,-ul) 7(0,0);(0,0,0) 9Cosineof6;projectionofwonv
11F;T;F 1 0 Zero; sum = 10 o'clock vector; sum = 8 o'clock vector times
15 45' 1 7 Circle x2 + 3 = 4; ( x - 1)2+ 9 = 4; vertical line x = 2; half-line x 2 0
~ S C O S ~ ~ = ~ - C O S ~ ? = ~ - C 25A.(A+B)=l+A.B=l+B.A=B.(~+B);e~uilatera1,60~
O S ~ ~ = $
fi'
27 a = A I, b = A J 29 (cos t, sin t) and (- sin t, cos t); (cos 2t, sin 2t) and (-2 sin 2t, 2 cos 2t)
SlC=A+B,D=A-B;C-D=A.A+B.A-A-B-B.B=Y~-Y~=O
88 u+v-w=(2,5,8),u-~+w=(O,-1,-2),-U+v+W=
(4,3,6)
SS c and d m ;b/a and da2 + b2 + c2
~ ~ M ~ = ? A + c , M ~ = A + ? B , M ~ = B + ~ c ; M ~ +~M ( A~++B M
+ c~) ==O
89 8 5 3 3; 2@ 5 x y + 4 1 Cancel a2c2 and b2da; then b2c2+ a2da >_ 2obcd because (bc - 20
4SF;T;T;F 45all2fi;cos6=-$
2V+W=i+a-k;2V-3W=2i-j+3k;lV12=2;V-W=~;COS~=?
4 v + w = (2,3,4,5);2V-3W = (-1,-4,-7,-io);lv12 = 4 ; v . w = io;cose = -&
6 (O,0,1) and (1, -1,o)
+
8 Unit vectors *(I, 1,l); i ( i j); -&(i - 2j + k); (i,
4-
?,?, i).
11.1 Vectors and Dot Products f ~ a a 405)
e
10 (cog 8, sin 8) and (cos 8, - sin 8); (r cos 8, r sin 8) and (r cos 8, -r sin 8).
12WewantV~(W-cV)=0orV~W=cV~V.Thenc=~=2and~-c~=(-1,0,1).
14 (a) Try two possibilities: keep clock vectors 1 through 5 or 1through 6. The five add to 1 + 2 cos 30' +
2 cos 60' = 2& = 3.73 (in the direction of 3:OO). The six add to 2 cos 15' + 2 cos 45' + 2 cos 75' = 3.88
which is longer (in the direction of 3:3O). (b) The 12 o'clock vector (call it j because it is vertical)
is subtracted from all twelve clock vectors. So the sum changes from V = 0 to V* = -12j.
16 (a) The angle between these unit vectors is 8 - 4 (or 4 - 8), and the cosine is U;:yl = cos 8 cos 4 + sin 8 sin 4.
(b) u3 = (- sin 4, cos 4) is perpendicular to u2. Its angle with ul is + 4 - 8, whose cosine is - sin(8 - 4).
5
The cosine is also = - cos 8 sin 4 + sin 8 cos 4. To get the formula sin(8 + 4) = sin 8 cos 4 + cos 8 sin 4,
take the further step of changing 8 to -8.
18 (a) The points tB form a line from the origin in the direction of B. (b) A + tB forms a
+
line f'rom A in the direction of B. [c) sA tB forms a plane containing A and B.
(d) v . A = v B means 3 = fixed number 9I I
where dl and 82 are the angles from v to A and B.
Then v is on the plane through the origin that gives this fixed number. (If IAl = IB I the plane bisects the
angle between those vectors.)
20 The choice Q = ( i , i) makes PQR a right angle because Q- P = ( i , i ) is perpendicular to R-Q = (-$, i).
The other choices for Q lie on a circle whose diameter is PR. (Fkom geometry: the diameter subtends a
;
right angle from any point on the circle.) This circle has radius and center i ; in Section 9.1 it was
the circle r = sin 8.
22 If a boat has velocity V with respect to the water and the water has velocity W with respect to the land,
then the boat has velocity V + W with respect to the land. The speed is not IVI + IWI
+
but IV WI.
24 For any triangle PQR the side PR is twice as long as the line A B connecting midpoints in Figure 11.4.
(The triangle PQR is twice as big as the triangle AQB.)Similarly lPRl = 21WI based on the triangle PSR.
Since V and W have equal length and are both parallel to PR, they are equal.
26 (a) I = (cos 8, sin 8) and J = (- sin 8, cos 8). (b) One answer is I = (cos 8, sin 8,O), J = (- sin 8, cos B,0) and
K = k. A more general answer is I = sin ~ ( C OB, Ssin 8,O),J = sin 4(- sin 8, cos 8,O) and K = cos 4(0,0,1).
2 8 1 . ~ = ~ . ~-- 2- = 0 .~ d d i + j = & ~ t o i - j = @ t o f i n d i = $ ( ~ + ~ ) . ~ o b s t i t u t e b a c k
4
t0find~=~I-J).ThcnA=2i+3~=~i(I+J)+~(I-J)=a1+Uwitha=fi+~
andb=fi-9.
SO IA. i12 ~ . for A = ( x , y, z) : x2 + y2 + z2 = lA12.
+ IAmj12+ )A kJ2= 1 ~ 1 Check
32 T ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ P R = A + B ~ ~ ~ Q S = B - A . T ~ ~ ~ I P R ~ ~ + ~ Q S I
(B . B - B A - A . B + A A) which equals 2A .A + 2B B = sum of squares of the four side lengths.
34 Thediagonalsare A + B and B - A. Suppose I A + B ~ = ~A . A + A - B + B - A + B . Bequals I B - A J ~
= B B - A . B - B A + B B. After cancelling this is 4A B = 0 (note that A B is the same as B A).
The region is a rectangle.
38 IA+B12 = A . A + A . B + B . A + B . B . I f t h i s e q u a l s A - A + B . B (and always A - B = B - A ) ,then
2A B = 0. So A is perpendicular to B.
a
38 In Figure 11.4, the point P is of the way along all medians. For the vectors, this statement means
A + i M s = $Ma = -C + $MI. To prove this, substitute -A - $ C for M1 and A $B for Maand +
C + ?A for MI.Then the statement becomes $A = LC 3 =! A + $B = -$C + $A.This is true because
11.2 P h e s and Projections (page 414)
B=-A-C.
40 Choow W = (1,1,1). Then V W = Vl + V2 + V3. The Schwan inequality IV .W12 5 IV121W12
+ + + +
is (V, v2 v3y 5 3(Vf v,2 V?).
+ +
42 IA BI 5 1A1+ IBI or ICI 5 IAI IBI says that any side length is less than t h e sum of t h e o t h e r t w o
+ + +
side lengths. Pmof: IA B12 5 (using Schwan for A B)IA12 21AllBI IB12 = ((IAl+ IBI)'.
+
44 ( V W ( = IVI + I W ( only if V and W are in the same direction: W is a multiple cV with c 2 0. Given
V = i + 2k this leads to W = c (i + 2 k) (for example W = 2i + 4k).
A plane in space is determined by a point Po= (xo, yo, zo) and a normal vector N with components (a, b, c).
The point P = (x, y, z) is on the plane if the dot product of N with P - Po is rero. (That answer was not PI)The
+ +
equation of this plane is a(x - w) b(y - yo) c(s - so) = 0. The equation is also written as ax+ by+ cz = d,
where d equals a+ + +
by() CEO or N - P O A parallel plane has the same N and a different d. A plane through
the origin haa d = 0.
+ +
The equation of the plane through Po= (2,1,0) perpendicular to N = (3,4,5) is hr 4y 5s = 10. A second
point in the plane is P = (0,0,2). The vector from Po to P is (-2, -1,2), and it is
perpendicular to N. (Check by dot product). The plane through Po = (2,1,O) perpendicular to the a axis has
N = (0,0,1) and equation s = 0.
The component of B in the direction of A is IBI cos 6, where 6 is the angle between the vectors. This
is A . B divided by ( A1. The projection vector P is (B1 cos 6 times a unit vector in the direction of A. Then
P = (IBlcosO)(A/IAl) simplifies to (A B)A/IA12. When B is doubled, P is doubled. When A is doubled,
P is n o t changed. If B reverses direction, then P reverses direction. If A reverses direction, then P stays
t h e same.
When B is a velocity vector, P represents the velocity in t h e A direction. When B is a force vector,
P is t h e force component along A. The component of B perpendicular to A equals B - P. The shortest
+
to the plane a z + by cz = d is along the normal vector. The distance from the origin is
and the point on the plane closest to the origin + b2 + c2). The
distance from Q = (xl, yl, zl) to the plane is Id - ax1 - byl -
1(0,0,0)and(2,-l,O);N=(l,2,3) S(0,5,6)and(0,6,7);N=(l,O,O)
S (1,1,1) and (1,2,2); N = (1,1,-1) +
7 x y =3 9 x + 2y + z = 2
11 Parallel if N V = 0; perpendicular if V = multiple of N
13 i +j + k (vector between points) is not perpendicular to N ; V . N is not zero; plane through first three
is z + y + z = 1; z + y - z = 3 succeeds; right side must be zero
1 5 a z + b y + c s = O;a(z- zo) + b(y- yo) + c ( z -20) = 0 1 7 cosd = $,$,$
19 & A has length $ 2 1 P = +A has length iIAl 2 3 P = -A has length IAl 25 P = 0
27 Projection on A = (1,2,2) has length 5; force down is 4; mass moves in the direction of F
2 9 IPlmi,= = distance from plane to origin 31 Distances 2 and 2 both reached at
6 6 (5, $, -5)
33 i + j + k ; t = -$;(+,-+,-$);&
35 Same N = (2, -2,l); for example Q = (0,0,1); then Q + :N = (i,-:, y) is on second plane; $ 1 ~ 1= 2
37 3i + 4;
(3t,4t) is on the line if 3(3t) + 4(4t) = 10 or t = g;P = (SO 25 ) IPI = 2
259 g 9
392z+2(?- ~ z ) ( - ~ ) = ~ s25o z ~= ; ~3 =~ + 4 ~ = 1 0 ~ i vf e s ~ =
4 1 Use equations (8) and (9) with N = (a,b) and Q = (zl, yl) m;
4 3 t = A.B B onto A
4 5 aVL = +LI - iLIII; a V F = $LII + iLIII
47V.LI=2-l;V.LIr =-3-1,V.LII1=-3-2;thus~.2i=1,~.(i-fij)=-4, andV= $ i + q j
2 P = (6,0,0) and Po= (0,0,2) are on the plane, and N = (1,2,3) is normal. Check N - (P - Po)=
(l,2,3).(6,O, -2) = 0.
4 P = (1,1,2) and Po = (0,0,0) give P - Po p e r p e n d i c u l a r t o N = i + j - k . (The plane i s ~ + ~ - z = 0
and P lies on this plane.)
6 The plane y - z = 0 contains the given points (0,0,0) and (1,0,0) and (O,1,1). The normal vector is N = j - k.
(Certainly P = (0,1,1) and Po = (0,0,0) give N - (P - Po)= 0.)
+
8 P = (x,y,z) lies on the plane if N - (P - Po)= l ( z - 1) 2(y - 2) - 1(z + 1) = 0 or x + 2y - z = 4.
1 0 z + y + z = zo + yo + zo or ( z - 20) + (y - yo) + (Z - ZO) = 0.
1 2 (a) No: the line where the planes (or walls) meet is not perpendicular to itself. (b) A third plane
perpendicular to the first plane could make any angle with the second plane.
1 4 The normal vector to 32 + 4y + 72 - t = 0 is N = (3,4,7, -1). The points P = (1,0,0,3) and Q = (0,1,0,4)
are on the hyperplane. Check (P - Q) . N = (1,-1,0,-1) . (3,4,7,-1) = 0.
16 A curve in 3D is the intersection of two surfaces. A line in 4D is the intersection of three hyperplanes.
5
18 If the vector V makes an angle 4 with a plane, it makes an angle - 4 with the normal N. Therefore
V.N =COS(: -4) =sin@. The normalto the zyplane is N = k , so sin4 =
~Tvl;imi
,A
= and 4 = $ 2.
20 The projection P = A B is :A = (1,-1,O). Its length is IPI =
I I
a.
Here the projection onto A equals A!
22 If B makes a 60' angle with A then the length of P is IBI cos 60' = 2 $ = 1. Since P is in the direct ion of
A it must be
24 The projection is P =
IAl'
M A= $ (i + j). Its length is IPI = q.
N B
26 A is along N = (1,-1,l) so the projection of B = (1,1,5) is P = -N = $(I,-1,l).
NI~ ,I 8
.A
28 P = *AB and the perpendicular projection is B - P. The dot product P . (B - P) or P . B - P . P is
zero: A B -1 % ~ A = 0.
30 We need the angle between the jet's direction and the wind direction. If this angle is 4, the speed over land
is 500 + 50 cos 4.
32 The points at distance 1 from the plane z + 2y + 22 = 3 fill two parallel planes x + 2y + 22 = 6 and
11.3 Cross Products and Determinants (page 423)
x + 2y + 2%= 0. Check: The point (0,0,0) on the last plane is a distance & = = 1from the
plane x + 2 y + 2 2 = 3.
The plane through (1, 1, 1) perpendicular to i + 2j + 2k is z + 2y + 22 = 5. Its distance from (0, 0, 0)
is#-=#.
I I
The distance is zero because t h e t w o planes meet. They are not parallel; their normal vectors (1, 1, 5)
and (3, 2, 1) are in different directions.
The point P = Q + t N = (3 + t , 3 + 2t) lies on the line z + 2y = 4 if (3 + t) + 2(3 + 2t) = 4 or 9 + 5t = 4 or
t = -1. Then P = (2,l).
The drug runner takes ? second to go the 4 meters. You have 5 meters to travel in the same second.
Your speed must be 10 m e t e r s p e r second. The projection of your velocity (a vector) onto the drug
runner's velocity equals t h e d r u g runner's velocity.
+
rhe equation a z + by+ cz = d is equivalent to a x ab y + 2 s = 1.So the three numbers e = 2, f = a,b g = 2
determine the plane. (Note: We say that three points determine a plane. But that makes 9 coordinates!
We only need the 3 numbers e, f , g determined by those 9 coordinates.)
l'wo planes ax + by + cz = d and ex + f y + gz = h are (a) parallel if the normal vector (a, b, c) is a multiple
of (e, f , g) (b) perpendicular if the normal vectors are perpendicular (c) at a 45' angle if the normal
vectors are at a 45' angle: HI'
= $.
+
The aVR lead is in the direction of A = -i j. The projection of V = 2i - j in this direction is
p = A-VA= T .
2 ) . TSh e l e n g t h o f P i s
-3( - i + j ) = ( 2 , - 3 @
142
If V is perpendicular to L, the reading on that lead is zero. If $ V (t)dt is perpendicular to L then
$ V(t) . Ldt = 0. This is the area under V(t) L (which is proportional to the reading on lead L).
The cross product A x B is a vector whose length is 1 A1 lBl sin 8. Its direction is perpendicular to A
and B. That length is the area of a parallelogram, whose base is IAl and whose height is IBI s i n 19. When
+ +
A = ali a j and B = bli b j , the area is lalb2 - a2bll. This equals a 2 by 2 determinant. In general
IA . ~l~ +
IA x BI' = ~ A I ~ I B I ?
+ + + +
The vectors ali a j ask and bli b j b3k have cross product (a2b3 - agb2)i (a3bl - albg)j + +
+ + + +
(alb2 - a2bl)k. The vectors A = i J k and B = i j have A x B = -i j. (This is also the 3 by 3
l i j k l
determinant 1 1 1 .) Perpendicular to the plane containing (0,0,0), (1,1,1),(1,1,0) is the normal vector N
I1 l o 1
= -i + j. The area of the triangle with those three vertices is &a,
which is half the area of the parallelogram
with fourth vertex at (2,2,1).
Vectors A, B, C from the origin determine a box. Its volume IA. (B x C) I comes from a 3 by 3 determinant.
There are six terms, t h r e e with a plus sign and t h r e e with minus. In every term each row and column is
represented once. The rows (1,0,0), (0,0,1), and (0,1,0) have determinant =-I. That box is a cube, but its sides
form a left-handed triple in the order given.
+ +
If A, B, C lie in the same plane then A . (B x C) is zero. For A = xi yj zk the first row contains the
letters x,y,z. SO the plane containing B and C has the equation A . B x C = 0. When B = i + j and C = k
that equation is x - y = 0. B x C is i - j.
A 3 by 3 determinant splits into t h r e e 2 by 2 determinants. They come from rows 2 and 3, and are multiplied
by the entries in row 1. With i, j, k in row 1, this determinant equals the cross product. Its j component is
-(alb3 - a3bl), including the minus sign which is easy to forget.
bl b3 bl b2
41 area2 = (iab)' + (iac)' + (ibc)' = (, IA x Bl)' when A = ai - bj,B = ai - ck
43 A = i ( 2 . 1- (-1)l) = z; fourth corner can be (3,3)
+ bj;lalb2 - a2blI;A x B = . . . + (alb2 - a2bl)k
45 ali + a j and bli
47 A x B; from Eg. (6), (A x B) x i = -(a3b1 - a1b3)k + (alb2 - a2bl)j;(A . i)B - (B . i)A =
al(bli + b j+ b3k) - bl(ali + aj+ ask)
49N=(Q-P)x(R-P)=i+j+k;area +&;z+y+z=2
i j k
8 cos6 sin6 0 =Oi+Oj+k(-cos2B-sin28)=-k.
sin6 - cos 19 0
116
vectors A x B, C , D can be put in any cyclic order (see 'useful factsB about volume of a box, after
Theorem 11G). We choose (A x B ) ( C x D).
50 For a parallelogram choose S so that S - R = Q - P. Then S = (2,3,3). The area is the length of the
i j k
cross-product (Q- P ) x (R- P ) = 0 1 2 = - 2 i + a -k. Its length is t/22+22+12 = 3.
1 2 2
Oneway toproduceaboxistochaweT= P + S a n d U = Q + S a n d V = R+S. (Then S T W
comes from shifting In that case the three edges from the origin are O P and 0 Q
1 8
The equations 3x+y = 8 and z+y = 6 combine into the vector equationx[
1] + y[ ]=[ ] = d. The left
Z
side is Au with coefficient matrix A = [ ] and unknown vector u = [ 1. The determinant of A is 2, so this
1 1 Y
problem is not singular. The row picture shows two intersecting lines. he column picture shows xa y b = d , +
3 1 1 1 -1
where a = [ ] and b = [ 1. The inverse matrix is A-' = TJ[
3
1. The solution is u = A-'d = [ 1 1.
A matrix-vector multiplication produces a vector of dot products from the rows, and also a combination of
the columns:
d -b
If the entries are a, b, c, d, the determinant is D = ad - be. A-' is 1 -C a
] divided by D. Cramer's Rule
shows components of u = A-'d as ratios of determinants: x = (b2dl - bld2)/D and y = (ald2 - a2dl)/D.
A matrix-matrix multiplication MV yields a matrix of dot products, from the rows of M and the columns
0fV:
The last line contains the identity matrix, denoted by I. It has the property that I A = A I = A for every
matrix A, and Iu = u for every vector u. The inverse matrix satisfies A-'A = I. Then Au = d is solved by
multiplying both sides by A-l, to give u = ~ - ' d . There is no inverse matrix when d e t A = 0.
a = (1,1,1), b = (1,2,3), and d = (0,8,4), the equations for x and y are 3x 6 y = 1 2 and 6x + 1 4 y = 28. +
Solving them also gives the closest line to the data points (1,0), (2,8), and (3,4). The solution is x = 0, y = 2,
+
which means the best line is horixontal. The projection is Oa 2 b = (2,4,6). The three error components are
(-2,4, -2). Check perpendicularity: (1,1,1)(-2,4, -2) = 0 and (1,2,3) . (-2,4, -2) = 0. Applying calculus
+
to this problem, x and y minirni~ethe sum of squares E = (-x - y)2 (8 - x - 2y) 2 (4 - x - s ~ ) ~ . +
1z = 5 , y = 2 , D = - 2 , [ : ] = 5 [ : ] + 2 [ - : ] 3r=3,y=l9[:] =3[:]+[1i],D=-8
gs=&I: 1; = 9 = 3 , y = - ; [ I3 0
8 ] 3
- 8 -1 llg
1 5 A-' = -:f 1 if ad - bc = 1 1 7 Are parallel; multiple; the same; infinite
19 ~ u l t i ~ loft seach tther; in the same direction as the columns; infinite
2 1 dl = .34,d2 = 4.91 2 3 .96z + .02y = .58, .04s + .98y = 4.92; D = .94,z = .5, y = 5
25 a = 1 gives any x = -y; a = -1 gives any s = y
5
arD=-2,A-1=-$[3 - : ] ; D = - ~ , ( ~ A ) - 2~ = L A - ~-2 ; 'D(A-1)-'
=I = original A;
D = -2 (not +2), (-A)-' = -A-'; D = 1,I-' = I
+ + +
3 1 AB =
ae+bg af +bh
+
ce dg cf dh + I '
aecf aedh bgc f bgdh
-af ce - af dg - bhce - bhdg
= (ad - bc)(eh - f g)
~SA-'=[: -;],B-l= [a 1
-:] 35 Identity; B-'A-' 3 7 Perpendicular; u = v x w
8 The solution is z = s, y= (ok to use Cramer's Rule) (solution breaks down if ad = bc);
;
'
&!It[;]+fi[:]= [:];I =."".
1 0 In Problem 4, s = IGI
2 4
= and y = 1
s
t2 4
= 0 (no solution)
16 If ad - bc = 1 then
0
18z = = - In Cramer's Rule this - signals
0' 0
6 2 6 2 6 2
that a solution might (or might not) exist.
20 z - y = dl
and 92 - 9y = d2 can be solved if d2 = 9dl.
22
+
.96z .02y = dl
The sums down the columns of A are .96 +.04 = 1 and .02 .98 = 1. +
21 is .04z + .98y = d2.
+
Reason: Everybody has to be accounted for. Nobody is lost or gained. Then z y (tota1 population before
move) equals dl + d2 (total population after move).
24 Determinant of A = 1 .96 .02
.04 .98 I =.94;~-'= -
eg4
1
-.O4
1
-*02 (columns still add to 1); A- 'A = I.
.96
I
32 I O!I+ 1 :-
3 0
1 doesnot equal Example of equality:
3 3 0 0 3 3
34 AB = I
2 6
0 2
has determinant 4 so (AB)-' = i I 2 -6
] . Check that this is also
36 CIB-' A-l ABC equals the identity matrix (because it collapses to C-'B-'BC which
is C - l C which is I). Then the inverse of ABC is c - ~ B - ~ A - ' .
38 (a) Find z and y from the normal equations. First compute a a = 3 and a - b = b a = 0 and b b = 2 and
a . d = 12 and b d = 2. The normal equations
33: Oy = 12 +
give z = 4 , y = 1.
ox 2y = 2 +
+ +
(b) The projection p = z a yb equals 4(1,1,1) 1(-1,0,1) = (3,4,5). Error d - p = (-1,2, -1).
Check perpendicularity of error: (-1,2, -1) . (1,1,1)= 0 and (-1,2, -1) . (-1,0,1) = 0.
40 Computes-a= 3 and a . b = b - a = 2 and b . b = 6 a n d a - d = 5 a n d b . d = 6 . The normal
3z+2y=5 =
equation (14) is with solution z = and y = 8 = 4 71.The nearest combination
2z+6y=6
+
z a y b is p = (5, y , y).
The vector of three errors is d - p = ($, -$, t).
It is perpendicular
to a and b. The best straight line is f = z + yt = . + $t.
.
42M=
0 1
[ , ,]
and.= [ i :].
44 Suppose u # 0 but Au = 0. Then A-' can't exist. It would multiply 0 (the zero vector) and produce u.
Three equations in three unknowns can be written as Au = d. The vector u has components x, y, z and A is
a 3 by 3 matrix. The row picture has a plane for each equation. The first two planes intersect in a line, and
all three planes intersect in a point, which is u. The column picture starts with vectors a,b,c from the columns
+ +
of A and combines them to produce xa y b zc. The vector equation is z a y b zc = d. + +
The determinant of A is the triple product a . b x c. This is the volume of a box, whose edges from the origin
are a, b, c. If det A = 0 then the system is singular. Otherwise there is an inverse matrix such that A-'A = I
(the identity matrix). In this case the solution to Au = d is u = A-'d.
The rows of A-' are the cross products b x c , c x a,a x b, divided by D. The entries of A-' are 2 by 2
determinants, divided by D. The upper left entry equals (b2c J - b3c2)/D. The 2 by 2 determinants needed
for a row of A-l do not use the corresponding column of A.
The solution is u = A- 'd. Its first component x is a ratio of determinants, Id b c I divided by la b c 1. Cramer's
Rule breaks down when det A = 0. Then the columns a,b,c lie in the same plane. There is no solution to
+ +
z a y b zc = d, if d is not on that plane. In a singular row picture, the intersection of planes 1 and 2 is
parallel to the third plane.
1 2 2
3 25 1
[ 20 2 [I] [i] [P -: [I] [HI
= 3 =
1 0 0
det A = 0, add 3 equations -,0 = 1 7 5a + l b + Oc = d, A-' =
[: :] 3
[i -;I [: a]
b x c; a . b x c = 0; determinant is zero 116, 2, 0; product of diagonal entries
-2 4 0 2 -1
A-l= ,.-I= - 15 Zero; same plane; D is zero
:1 2;
18 12 0
] , det AB = 12 = (det A) times (det B)
A+C = [ -:-: :]
2 3 -3
, det(A+ C ) is not det A + det C
2) (3)- (0)(61 -(4)(3)+(0)(0)
P = ~ 6 = l,q = 6 =-2 25(A-')-'isalwaysA
New second equation 32 = 0 doesn't contain y; exchange with third equation; there is a solution
Pivots 1,2,4, D = 8; pivots 1,-1,2, D = -2 37 a12 = 1,a21 = 0, C aijbjk = row a, column k in AB
. [ ::][]=[HI :][]=[XI
1 1 0 1 0 0
A [ : :
6 By inspiration (x, y, z ) = (1, -1,l). By Cramer's Rule: det A = -1 and then
0 1 0 1 0 0 1 1 0
x = - -1 1 1 1 = l , y = z 1 1 1 = - I + = ' -1 1 1 1 = 1
0 1 1 0 0 1 0 1 0
8 x + 2 y + 2 z = O -+ x + 2 y + 2 z = O -+ x + 2 y + 2 z = O z=-8 -+
x=o
when y = 0 .
z=1
18 Choose d = I 0 I as right side. The same steps as in Problem 16 end with y - 32 = 0 and -y + 3z = 1.
111
Addition leaves 0 = 1.No solution. Note: The left sides of the three equations add to zero.
20 BC =
0 -1
and CB = [ -20 -13
-:]
There is a solution only if the right sides (components of d) also add to zero.
22 2 4 = [: ]
0 0 6
1: has determinant 48 which is 8 times det A. If an n by n matrix is multiplied by 2,
26 The inverse of A B is B-~A-'. The inverses come in reverse order (last in - first out: shoes first!)
1 0 0 0 1 0 0 0 1 0 1 0 0 0 1
The position vector R ( t ) along the curve changes with the parameter t. The velocity is dR/dt. The acceler-
+ + +
ation is d2a/dt2. If the position is i t j t2k, then v = j 2 t k and a = 2k. In that example the speed is
Ivl= d xThis equals ds/dt, where s measures the distance along t h e curve. Then s = (ds/dt)dt.
The tangent vector is in the same direction as the velocity, but T is a unit vector. In general T = v/lvl and
+
in the example T = (j 2 t k)/d-.
Steady motion along a line has a = rero. If the line is x = y = z, the unit tangent vector is T = (i +j + k)/fi.
If the speed is lvl = &,the velocity vector is v = i +j + k. If the initial position is (1,0,0), the position vector is
+ +
R(t) = ( 1 + t ) i + t j + t k. The generalequation of aline is x = xo+tul, y = yo t v 2 , z = 10 tv3. Invector
notation this is R(t) = Q + t v. Eliminating t leaves the equations (x - xo)/ul = (y - yo)/@ = (I - I ~ ) / V ~ .
A line in space needs t w o equations where a plane needs one. A line has one parameter where a plane has two.
The line from & = (1,0,O) to (2,2,2) with lvl = 3 is R(t) = (1+ t ) i + 2 t j + 2 t k.
Steady motion around a circle (radius r, angular velocity w) has x = r cos wt, y = r sin wt, z = 0. The velocity
+
is v = -rw sin w t i r w coa w t j. The speed is Ivl = rw. The acceleration is a = -rw2 (cos w t i + sin wt j),
which has magnitude rw2 and direction toward (0,0). Combining upward motion R = tk with this circular
+
motion produces motion around a helix. Then v = -rw sin wt i r w cos w t j k and lvl= + d-.
4 3 u = (8,3,2); projection perpendicular to v = (1,2,2) is (6, -1, -2) which has length a
46 z = G(t), y = F(t); y = z2I3;t = 1and t = -1 give the same z so they would give the same y; y = G(F-'(x))
+
2 The path is the line z y = 2. The speed is d(dz/dt)2 (dy/dt)2 = fi. +
4 = 6 - 2t = 0 at t = 3, so the highest point is x = 18,y = 9. The curve is the parabola y = x -
and a = -2tJ.
6 (a) x2 = y so this is a parabola (b) 2 = y2 so y = xSI2 is a power curve (c) In z = t ln 4 so
y=m 4 x is a logarithmic curve.
8 The direction of the line is 4i + Sj. This is nonnal to the plane 4x + Sy + Oz = 0. (The right side could be
any number.) One line in this plane is ix + Sy = 0, a = 0. (A point that satisfies those two equations
also satisfies the plane equation.)
10 The line is (z, y, z) = (3,1, -2) t(-1, + -3,f). Then at t = 3 this gives (0, 0, 0). The speed is
tame
= = 9. For speed et c h m e (z, y,z) = (S,l, -2) + -&(-3, -1,2).
1 2 x = eos et , = sin et has velocity 2 = (- +
sin et)et , $ = (cos et )et and speed J ( d ~ / d t ) ~ (dy/dt)2 = et .
The circle is complete when et = 2x or t = In 2 ~ .
1 4 z2 + y2 = ( l + t ) 2 + (2-t)2 is a minimum when 2(1+t) -2(2 - t) = 0 or 4t = 2 or t = 5.
1 The path
+
crosses y = z when 1 t = 2 - t or t = (again) at x = y = 5. S The line never crosses a parallel
line l i i e x = 2 + t , y = 2 - t .
16 (b)(c) (d) give the same path. Change t to 2t, -t, and t3, respectively. Path (a) never goes through (1,l).
+
18 If z = 1 v ~ =t 0 and y = 2 + v2t = 0, the k t gives t = -;?; and then the second gives 2 - 2 = 0
or hrl - v 2 = 0. This line crosses the 45' line unless vl = v2 or v l - v 2 = 0. In that case z = y leads
to 1 = 2 and is impossible.
20 If 2% + vg= 0 along a path then &(x2 + y2) = 0 and z2 + y2 = constant.
22 If a is a constant vector the path must be a straight line (with uniform motion since z = xo + xlt and
y = yo + y t are the only functions with =0 = 3).
If the path is a straight line, a must be
in t h e same direction as t h e line (but not necessarily constant).
+ + i. +
24 z = 1 2cos $ and y = 3 2sin Check (z - 1)' (y - 3)' = 4 and speed = 1.
26 la1 = $ when the motion is along a straight line. On a curve there is a turning component - for example
x = c o e t , y = s i n t has $ = l a n d t h e n 9 = 0 b u t a = - c o s t i - s i n t j i s n o t rero.
dt - d ( d z ~ d t ) ~
2% dd + +
( d ~ / d t ) ~( d ~ l d t =
)~ = 7. The path leaves (1,2,0) when t = 0 and arrives at
(13,8,4) when t = 2, so the distance is 2 . 7 = 14. Also 1 2 ~ + 62 + 42 = 1 4 ~ .
+
SO If the parametric equations are x = cos 8, y = sin 0, a = 8, the speed is d ( d ~ / d t ) ~( d ~ l d t+) ~( d ~ l d t ) ~
= \/(sin2 8 + cos2 O)(d8/dt)2 + (d8/dt)2 = \/Zldd/dtl. (In Example 7 the speed was &.) So take 8 = t / f i
for speed 1.
32 Given only the path y = f ( x ) , it is impossible to find the velocity but still possible to find the
tangent vector (or the slope).
+
34 z = cos(1- e-'), y = sin(1- e-') goes amund the unit circle x2 y2 = 1 with speed e-'. The path starts
at (1,O) when t = 0; it ends at x = cos 1,y = sin 1when t = oo. Thus it covers only one radian
I
(because the distance is J(ds/dt)dt = e-' = 1). Note: The path x = cos e-*, y = sin e-' is also acceptable,
12.2 Plane Motion: Projectiles and Cycloids (page 457)
A projectile starts with speed uo and angle a. At time t its velocity is dx/dt = vo cos a,dy/dt = vo sin a - g t
(the downward acceleration is g) .Starting from (0,0), the position at time t is x = vo cos a t, y = v o sin a t - -gt
1 2.
The flight time back to y = 0 is T = 2vo(sin a)/g. At that time the horbontal range is R = ( v i sin 2a)/g.
The flight path is a parabola.
The three quantities uo, a, t determine the projectile's motion. Knowing uo and the position of the target,
we cannot solve for a. Knowing a and the position of the target, we can solve for uo.
A cycloid is traced out by a point on a rolling circle. If the radius is a and the turning angle is 0, the center
of the circle is at x = aB, y = a. The point is at x = a(B - sin B ) , y = a(1- cos B), starting from (0,O). It travels
a distance 3r2 in a full turn of the circle. The curve has a cusp at the end of every turn. An upside-down
cycloid gives the fastest slide between two points.
37 y = 2 sin 8 - sin 28 = 2 sin 8(1- cos 8); x2 + y2 = 4(1- cos 0)" r = 2(1- cos 8)
20
dttance a c m 8
= tan 8 and z = distance across = 3 = 2a cot 0. The length O B is 2asin 8 (from
the polar equation of a circle in Figure 9.2c, or from plane geometry). Then the height of
+
B is (OB)(sin 8) = 20 sin2 8. The identity 1 cot2 8 = csc2 8 gives 1 ( e l 2 = Y + a.
38 On the line z = g y the distance is ds = \ / ( d ~ ) ~ ( d ~=) d(lr/2)"+1
~ +
dy. The last step in equation (5)
=d
integrates csz~@G to give , 3 5 f [ 2 4 ~ m % = dIra+kfi.
a
40 I have read (gut don't believe) that the rolling circle jumps as the weight descends.
The curvature tells how fast the c w e turns. For a circle of radius a, the direction changes by 27 in a
distance f r a , so n = 1/a. For a plane e w e y = f (z)the formula is n = ly"l/(l+ (yt)2)S/2. The curvature of
. a point where y" = 0 (an inflection point) the curve is momentarily
y = sinz is lain x l / ( l + C O S ~ X ) ' / ~At
straight and n = zero. For a space c w e n = Iv x al/(v13.
The normal vector N is perpendicular to the curve (and t h e r d o r e t o v and T) It is a unit vector .
along the derivative of T, so N = Tt/lTtI. For motion around a circle N points inward. Up a helix N also
points inward. Moving at unit speed on any curve, the time t is the same as the distance 8 . Then lvl = 1and
@s/dt2 = 0 and a is in the direction of N.
+
Acceleration equals d2s/dt2 T n1vl2 N. At unit speed around a unit circle, those components are zero
and one. An astronaut who spins once a second in a radius of one meter has la1 = w2 = ( 2 d 2 meters/sec2,
which is abbut 4g.
1&
-p 3 $ 6 O (line) 7 * 5 9 (- sin t2, cos t2); (- cos t2, - sin t2)
11(cost,sint);(-sint,-cost) l ~ ~ ~ ~ s i n t , f c o s t , ~ ) ; ~ v ~ = S , r = ~t;a ~n @
l o=n! g e r ;
1
2 y=Inxhasn=
(l+v qa/== &, = (x2+l)t2
x
. Maximum of n when its derivative is ~ e r o :
(z2 + + 1
A central force points toward t h e origin. Then R x b R / d t Z = 0 because these vectors a r e parallel.
128
12.4 Pokr Coordinates and Planetary Motion (page 468)
In polar coordinates, the outward unit vector is ur = cos 8 i + sin 8 j. Rotated by 90' this becomes ue =
-sin 8 i + cos 8 j. The position vector R is the distance r times ur. The velocity v = dR/dt is (dr/dt)ur +
+
( r d8/dt)ue. For steady motion around the cirde r = 5 with 8 = 4t, v is -20 sin 4 t i 20 cos i t j and lvl is
20 and a is -80 cos 4 t i - 80 sin 4 t j.
For motion under a circular force, 9 times d8/dt is constant. Dividing by 2 gives Kepler's second law
f
dA/dt = r2d8/dt = constant. The first law says that the orbit is an ellipse with the sun at a focus. The
+
polar equation for a conic section is l/r = C - D cos 8. Using F = ma we found gee q = C.So the path is a
conic section; it must be an ellipse because planets come around again. The properties of an ellipse lead to
the period T = 2 x a S 1 2 / t / m , which is Kepler's third law.
2 The point (3,s) is at 8 = 2. So u r = &(i +j) and ue = Afi( - i +j). If v = i +j then v = &ur. This is
the velocity when $ = \/Z and g = 0. (Better question: If R = 3i + 3j then
R= ur. Answer r = fi.)
4 r = 1- cos8 has $ = sine% = 2sin8. Then v = 2 sin 8 ur + 2(1- cos 8)ue. The cardioid is covered as
8 g a s from 0 to 2n. With = 2 the time required is r .
6 The path r = 1,8 = sin t goes along the unit circle from 8 = 0 to 8 = 1 radian, then backward to
8 = -1radian, and oscillates on this arc. The velocity from equation (5) is v = r 2 u e = cos t ue;
the acceleration is a = -cos2t u r - sin t ue : part radial from turning, part tangential
from change of speed. v = 0 when cos t = 0 (top and bottom of arc: 8 = 1 or -1).
8 The distance r8 around the circle is the integral of the speed 8t : thus 46 = 4t2 and 8 = t2. The
circle is complete at t = 6. At that time v = r g u o = 4 ( 2 & ~ and a = -4(8r)i 4(2)j. +
g
1 0 The line x = 1is r cos 8 = 1or r = see 8. Integrating r2 = sec28% = 2 gives tan 8 = 2t. The point
(1,l)at B = f is reached when tan8 = 1= 2t; then t = 3. 1
1 2 Since ur has constant length, its derivatives are perpendicular to itself. In fact %= 0 and du
= ue.
1 4 R = reie has 9 = $eie + + +
2 s ( i e i e g ) i r g e i e i2r(g)2ee'. (Note repeated term gives factor 2.)
&
The coefficient of eie is +
- r($)'. The coefficient of ieie is 2 2 % r e . These are the ur
12.4 Polar Coordinates and Planetary Motion (page 468)
and ue components of a.
16 The period of a satellite above New York is 1 d a y = 86,400 seconds. Then 86,400 = 7= o3l2
gives a = 4.2 lo7 meters = 420,000 km.
18 The period of the moon reveals the mass of the earth: 28 days 436400z = a ( 3 8 0 , O O O ) ~ ~ ~
m
gives A4 = 5.54 lo2' kg. Remember to change 380,000 km to meters.
20 (a) False: The paths are conics but they could be hyperbolas and possibly parabolas.
(b) W e : A circle has r = constant and ? & = constant so 3 = constant.
(c) False: The central force might not be proportional to f .
22 T = &
, (~oOO]~P p~ .268 seconds.
+ + + +
24 1= C r - Dz is 1 D z = C r or 1 2Dx ~~x~= c 2 ( x 2 y2). Then (C2 - D2)z2 + Cay2- 2 0 2 = 1.
26 Substitute z = -c,y = $ and use c2 = a2 - b2. Then $ $ = $ + + = = 1.9
28 If the force is P = -m(r)u,., the left side of equation (11) becomes -a(r). Gravity has a(r) = GM
r2'
30 Multiply gee + q = $ by qe and integrate: + ;$
I FdB = A+c.
=
2s2
Substituting u = q2
and ue = 2qqe (or $ = + = 2)gives 8, + = 3 2U + C or ug = -4u2 + 8uC - 4. Integrate
u1 u1
The graph of z = f (z, y) is a snrfsce in three-dimensional space. The level curve f (z, y) = 7 lies down in
the base plane. Above this level curve are all points at height 7 in the surface. The plane z = 7 cuts through
the surface at those points. The level curves f (z, y) = c are drawn in the zy plane and labeled by c. The family
of labeled curves is a contour map.
For z = f (z, y) = z - y2, the equation for a level curve is x2 - y2 = c. This curve is a hyperbola. For
z = z - y the curves are straight b e a . h e 1 curves never cross because f(x,y) cannot equal two numbers
c and cf. They crowd together when the surface is steep. The curves tighten to a point when f reaches a
maximum or minimum. The steepest direction on a mountain is perpendicular to the level curve.
S z derivatives oo, -1, -2, -4e-' (flattest) 6 Straight lines 7 Logarithm curves
9 Parabolas 1 1 No: f = (z+ y)n or (az +by)" or any function of az by + 1S f (z, Y)= 1- z2 - y2
1 6 Saddle 1 7 Ellipses 4z2+ = c2 19 Ellipses 52' + +
y2 = c2 4cz z2 +
21 Straight lines not reaching (1,2) 28 Center (1,l); f = 2 y2 - 1 + 26 Four, three, planes, spheres
27 Less than 1, equal to 1, greater than 1 29 Parallel lines, hyperbolas, parabolas
3 1 $ : 482 - 3 2 = 0, z = 16 hours 8 3 Plane; planes; 4 left and 3 right (3 pairs)
34 The function f (z, y) is the height above the ground. The level curve f = 0 is the outline of the shoe.
The partial derivative a f /ay comes from fixing x and moving y. It is the limit off (x,y A y ) - f (x,y))/Ay. +
If f = eZ2sin y then a f / a z = 2 e 2 sin
~ y and af /ay = eax cos y. If f = (z2+#)'I2 then f, = x/(x2 y 2 ) 112 +
and f, = y/(x2+ y2)1/2. At (zo, yo) the partial derivative f, is the ordinary derivative of the partial function
f (z, yo). Similarly f, comes from f (xo,y). Those functions are cut out by vertical planes z = zo and y = y o ,
while the level curves are cut out by horisontal planes.
The four second derivatives are f,,, fw ,fyx, fyy . For f = zy they are 0,1,1,0. For f = cos 2%cos 3y
they are -4 coa 2x cos Sy, 6 sin 2 x sin Sy, -9 cos 2x cos Sy. In those examples the derivatives fw and fm
are the same. That is always true when the second derivatives are continuous. At the origin, cos 22 cos 3y is
curving down in the z and y directions, while zy goes up in the 45' direction and down in the -45' direction.
13
9&7;&7
+ 2zy2; -1 + 2yz2
ll*;&~
s 3z2y2 - 22; 2z3y - eY
l32,3,4
5 -a;
&
156(z+iy),6i(z+iy),-6(z+iy)
-2z 3 ;
(z'+u')
22'-
1 7 (f = !) f,, = +;
= f,, = f,, = 9; 2q, 19 -a2 cos azcos by, ab sin a z sin by, -b2 cos a z cos by
2 1 Omit line z = y; all positive numbers; f, = -2(z - Y ) - ~ , f, = 2(z -
23 Omit z = t; all numbers; A,
3, w,
25 z > 0, t > O and z = 0, t > 1 and z = -1, -2, . . ,t = e, e2,. - ;f, = (1nt)z1"'-', ft = (In z)tl""-'
a
82
=7 +
a=ra- t
~ 3j s ~
at--7'
24 Domain: halfplane where z + t > 0; range: all real numbers; f= = & = ft.
26 Domain: all (x, y) with lyl 5 1; range: all numbers with absolute value 1f (x, y) 1 <r
( s i n c e - 1 5 c a ~ 5 l a n d O 5 c o s - ~ y 5 r ) ; ~ = - s xcos-'y,%=
in -ma
28 2 = -u(z) and 5=~(y).
J O f(z, y) = 1'$ = lny - lnx = ln f; f.
=- 1
;and f, = 1 (confirming Problem 28).
32 g(y) = e-', or any multiple Ae-'g. S4 g(y) = ecau or any multiple A ~ C ' Y .
56 fz = 5(,*)e-sa14t. Then 1%. = ft = + 44te - x a 1 4 t .
+ 4 t ft = fzz + f, =
38 e-mat-n 'sin mz cos ny solves ft = f== f,. Also f = ~ e - ( z a + ~ a ) /has
2 + a -(~~+,~)/4'.
(-$ + +)e
:
40 f = sin(z - t) peaks when z - t = 5 and f = 1. If t increases by At = then z increases by Ax = i.
The wave velocity is 2
= 1. The other function sin(x t) has + 2
= -1 (velocity 1 to the left).
42 $$ = gives =3 & and also = 9.
Use fSt = ft. to find = 9 3.
44 (a) 2
= y and a = x when zy > 0 (two quadrants); = -y and2 = -x when xy < 0 (other two
quadrants); %
doesn't exist but = 0 up the y axis, = 0 but% 5
doesn't exist on the z axis.
(b) 2 = 22 and = 2y except f is not continuous when x = 0 and y # 0. OK at (0,O).
46 (a) ( ~ 2 9 ~=
2) (a, i);
(24, ~ 4 =
) (isi); ) ( ~,~
approaching ( 0 4 ) (b) ( ~ 2~, 2 = 4 4 =) (LO) approaching (1,o)
(c) (q, a)= (z4,y4) = (1,O) but n o limit (d) (22, b)= (2, O), (x4, y4) = (4,O) has no S i i t .
48 (a) The limit is d m ( c o n t i n u o u ~function) (b) The limit is % provided b # 0 (c) The limit is
provided a + b # 0 (d) The limit is ,* except n o limit at (0,0).
50 Along y = mz the function is x , ~ ~ ~-r, 0
a (the ratio is near & for small z). But on the parabola y = x2
the function is & = 4.SOthis function f (z, y) has n o limit: not continuous at (0,O).
62 (a) - - - = y(-*$) -r 0 because always 1,
e i. (b) .&-
5l equals 0 on the axes but i on 45' lines;
n o S i i t ; (c)
,m/a yn/3
= (zrnI2f12) ( 2m+,n ) -r 0 if m > 0, n > 0, because the second factor is 5 5
as in (a). For negative z and y, m and n should be positive integers. Arther problem by same method:
zayb
zrn+," - r O i f a > F a n d b > q .
The tangent line to y = f (z) is y - yo = ft(x0)(x - w). The tangent plane to w = f (z, y) is w - wo =
(af/&c)o (x - XO) +(af/@)O (y - yo). The normal vector is N = (fx,fy ,-1). For w = x3 y3 the tangent +
+
equation at (1,1,2) is w - 2 = S(x - 1) S(y - 1). The normal vector is N = (S,5, -1). For a sphere, the
direction of N is o u t fkom the origin.
The surface given implicitly by F(z, y, a) = c has tangent plane with equation ( a F / a ~(z ) ~- xo)+
(aF/@)(y - yo) + ( ~ F / ~ Z )-~zO) ( Z= 0. For zyz = 6 at (1,2,3) the tangent plane has the equation
+ - + - +
6(x - 1) S(y 2) 2(s 8) = 0. On that plane the differentials satisfy 6dz Jdy 2dz = 0. The differ- +
ential of z = f (z, y) is dz = fxdx + fydy. This holds exactly on the tangent plane, while Az rr fxAx fyAy +
+
holds approximately on the surface. The height z = 32 7y is more sensitive to a change in y than in z,
13.3 Tangent Planes and Linear Approximations (page 488)
+ +
The linear approximation to f (z, y) is f (20,yo) (af /r3x)o(x- xO) (af/*)O(y - yo). This is the same as
Af t +
a (af/ax)Az (X/ay)Ay. The error is of order (AX)' +
( A ~ ) ' . For f = sin zy the linear approximation
around (0,O) is fL = 0. We are moving along the tangent plane instead of the surface. When the equation is
given as F ( z , y, z) = c, the linear approximation is FxAz FyAy FIAz = 0. + +
Newton's method solves g(z, y) = 0 and h(z, y) = 0 by a linear approximation. Starting from z,, yn the
+ +
equations are replaced by gx A x gy A y = -g(xn, yn) and hx A x h y A y = -h(xn, yn) The steps Az and .
Ay go to the next point ( X ~ + ~ , Y ~ Each
+ ~ ) solution
. has a basin of attraction. Those basins are likely to be
frac tale.
lz-l=y-l;N=j-k 3 ~ - 2 = $ ( ~ - 6 ) - 32 Y( - 3 ) ; N = $ i - $ j - k
5 2(2-1)+4(y-2)+2(z-l)=O;N=2i+4i+2k 72-1=z-l;N=i-k
9 Tangent plane 2zo(z - zo) - 2z0(z - xo) - 2yo(y - yo) = 0; (0,0,O) satisfies this equation because
yg
z i - zg - = 0 on the surface; cos 6 = N.k - 3
m=,/-.-fi (surface is the 45' cone)
11 dz = 3dz - 2dy for both; dz = 0 for both; Az = 0 for 32 - 2y, Az = .00029 for z3/y2; tangent plane
1 3 z = zo+ Fzt; plane 6(z - 4) + 12(y - 2) + 8(z - 3) = 0; normal line z = 4 + 6t,y = 2 + 12t,z = 3 + 8t
1 5 Tangent plane 4(z - 2) + 2(y - 1)+ 4(z - 2) = 0; normal line z = 2 + 4t, y = 1+ 2t, z = 2 + 4t; (0,0,0)
= -Q8 = sQ8 + Q = Q8 + 40
27 = 1,t = 10 gives Q = 40 :
Pt = - Q ~= S Q ~ + ~ = Q ;~Q +~ =~- ~ O , Q ~ = - ~ , P ~ = ~ O , P ~ = ?
3
2N=i+j+k;(z-3)+(y-4)+(z-10) =O 4N=i+2j-k;x+2y=z-1
6 N = 2i +qi +4k;2(x - 1) + 4(y - 2) + 4(z - 1) = 0
8 N = 87ri + 4 d - k; 8 r ( r - 2) + 4s(h - 2) = V - 87r
i j k
10v= 1 4 -1 =-i-j- 5k (both planes go through (0,0,0) and so does the line!)
2 3 -1
13.3 Tangent P h e s and Linear Approximations (page 488)
i j k
12 N t = 21 + 41 - k and N2 = 2i + 6j - k give v = 2 +
4 -1 = 21 4k tangent to both surfaces
2 6 -1
+
1 4 The direction of N is 2xy2i 2x2yj -k = 8i + 4 - k. +
SOthe line through (1,2,4) has x = 1 8t,y = 2 + It,
z=4-t.
1 6 The normal line through (xo, YO, zo) has direction N = (F.i + +
FA Fsk)o. This is the radial line from the
origin if (Fz)o = cxo, (F,)o = cyo , + +
= czo. Then F is a function of x2 y2 z2 and the surface
is a sphere.
18 df = y z d x + x z d y + x y d z .
20 Direct method: R = l; !
and ,
BR1 =A RI +Ra and BRa = A..
Rl+Ra If R1= 1 and R2 = 2 then is
four times larger ($ vs. i;more sensitive t o R1). By chain rule: - = - 4 and - & # = - -.
1
R R
1 2
22 (a) Common sense: 2 hits in 5 at bats (: = .I) raises an average that is below .4. Mathematics:
A = I has
dA = IT> 0 if y dx > x dy. This is again $ > :or .4> A. (b) The player has
x = 200 hits since = .5. We want to choose Ax = Ay (all hits) so AA reaches .005.
But AA nr * l d = ,&AX = .005 when Ax = 4 hits. Check: = .505 (to 3 decimals).
If averages are rounded down we need Ax = 5 hits.
24 (1) c is between xo and + Ax by the Mean Value Theorem (2) C is between yo and yo + Ay
(3) the limit exists if fx is continuous (4) the limit exists if fy is continuous.
26 P = and Q = so BE =
r+.2 r+.2 Bs mland ' 6t = A r+.2'
At 8 = .ql t = 10 this gives =
&=%=
andBP=a=L
at .6 3'
28 Take partial derivatives with respect to b: 22% b + %+x = 0 or = a.
Similarly 2 x g b +
1= 0 +
gives = e.
1 Then is larger (in magnitude) when x = 2.
30 (a) The third surface is 1=0. (b) Newton uses the tangent plane to the graph of g, t h e tangent plane t o
t h e graph of h, and I = 0.
32 :AX - Ay = iand -Ax + = igive Ax = Ay = -:. The new point is (-1, -I), an exact solution.
The point (5, i)is in the gray band (upper right in Figure 13.lla) or the blue band on the front
cover of the book.
34 3a2Ax - Ay = -a3 and -Ax + OAy = a give Ax = -a and Ay = -2a3. The new point is (0, -2a3) on
the y axis. Then OAx - Ay = -2a3 and -Ax + 3(4a6)Ay = 8a9 give Ay = 2a3 and Ax = 160'.
The new point (16a9,0) is the same as the start (a, 0) if 16a8 = 1 or a = f &. In these cases
Newton's method cycles. Question: Is this where the white basin ends along the x axis?
36 By Problem 34 Newton's method diverges if 16a8 > 1: for instance (so,
yo) = (1,O) as in Example 9
in the text.
38 A famous fractal shows the three basins of attraction - see almost any book displaying fractals. Remarkable
property of the boundary points between basins: they touch all three basins! lly to draw 3 regions
with this property.
40 Problem 39 has 2xAx - Ay = y - x2 and Ax - Ay = y - x. Subtraction gives (22 - 1)Ax = x - x2. Then
+
x Ax = x + & &.
= By the second equation this is also y + Ay. Now find the basin:
+ ;
If x < 0 then Ax > 0 but x Az still < 0 : moving toward 0. If 0 5 x < then z + Ax < 0. So the basin
for (0,O) has all x < 3.
The line x = gives blowup. If ? < x 2 1then Ax > 0. If x > 1then Ax < 0
5
+
but x Ax = & +
2 1 (because x2 - 2% 12 0). So the basin for (Ill) has all x > $.
13.4 Directional Derivatives and Gradients (page 495)
42J=[
1 1
] is singular; g and h have the same tangent planes. Newton's equations 2Ax + 2Ay = -2 and
Ax + Ay = -1 have infinitely many solutions.
41f gives the rate of change of f(x,y) in the direction u. It can be computed from the two deriva-
tives H/ibr and H / a y in the special directions (1,O) and (0,l). In terms of ul, u2 the formula is &f =
fxul +
fyu2. This is a dot product of u with the vector (fx, fy), which is called the gradient. For the linear
+
function f = ax by, the gradient is grad f = (a, b ) and the directional derivative is &f = (a,b ) u.
The gradient V f = (fz, f,) is not a vector in three dimensions, it is a vector in the base plane. It
is perpendicular to the level lines. It points in the direction of steepest climb. Its magnitude lgrad f 1 is
+
For f = x2 3 the gradient points o u t from t h e origin and the slope in that
steepest direction is I(&, 2y)1 = 2r.
The gradient of f (x, y, z ) is (fX,fy, fs). This is different from the gradient on the surface F(z, y, z) = 0, which
is -(Fz/Fz)i- (F,/F,)j. naveling with velocity v on a curved path, the rate of change of f is df /dt = (grad f ) v.
When the tangent direction is T, the slope of f is df Ids = (grad f ) . T. In a straight direction u, df /ds is the
same as t h e directional derivative &f.
a'A),&f = fie-'
1 5 grad f = (eZ-Y, -ez-Y) = (e-', -e-') at P ; u = (1
. .
17gradf=Oatmaximum;levelcurveisonepoint 19N=(-l,l,-l),U=(-1,1,2),L=(l,1,0)
2 1 Direction -U = (-2,0, -4)
25 f = (x + 2y) and (x + 2y)'; i +
23 -U = (a,
+
e,
e,)
+
2j; straight lines x 2y = constant (perpendicular to i 25)
= f(&, 3);grad g = f(216, - 3)+ C, g = f( 2 6 %+ f i y )
f = f(3
fi), +C
27 grad f
29 9 = constant along ray in direction u = v;. grad B = *
.i 4;
= -4i+3* us grad B = 0
3 1 U = (f., f,, f: +
f i ) = (-1, -2,5); -U = (-1, -2,5); tangent at the point (2,1,6)
+ +
33 grad f toward 2i j at P,j at Q, -21 j at R; (2, ?) and (2$,2); largest upper left, smallest lower right;
z,,, > 9;z goes from 2 to 8 and back to 6
$6 f
= id(/(.- ! i ! i
1 ) 2 + ( ~ - 1 2) ;(,,,,,)o.o= (&,a)
37 Figure C now shows level curves; (grad f 1 is varying; f could be xy
+
39 x2 xy; el-,; no function has 2
= y and = -x because then fz, # f,.
4 1 v = (1,2t); T = v/d-; = v ((2, 2t2) = 2t + 4t3; %
= (2t + 4 t 3 ) / d s 2
13.4 DIreetiond Derivatives and Gradients (page 495)
4 s V = (2,s); T = 5;
5= V (2x0 -k It, -2y0 - 6t) = 4xo - 6~ - lot; % = %
4 ~ v = ( e ~ , 2 e " , - e - ~ ) ; ~ = & ; g r a d f = ( &z' l)=(e-t,e-",et),5=1+2-1,g=
J' I
(V(
2
47 v = (-2sin2t,2cos2t),T = (-sin2t,cos2t); grad f = (y,z), % = -2sin22t + 2cos2 2t, = ;%;
sero slope because f = 1on this path
49~-1=2(~-4)+3(g-5);f =1+2(~-4)+3(y-5) Slgrad f . T = O ; T
2 grad +
f = fJ+ fJ = 3i 4;41f = 3(%)+ 4($) = 5 at every point P.
4 grad f = l0#j : &f = -10fl;&f(P) = 10. 6 grad = .zi~i-d-l'f,l
+w'+z 1 I
8 g r a d f =-,&@$+m-l&i&#&qm
2+11+ +A
A-(3ul
du1
+ d q ) = 3 - 1 u = O if ul = 3 d q . Then u: = 9(1- u:) or IOU: = 9 or u l =
+ d$ = \/iTj.
7
The chain rule applies to a function of a ihnetion. The x derivative off (g(z,y)) is a f /ax = (af/ag)(ag/ax).
+
The y derivative is a f l a y = (aflag)(agl*). The example f = (z+y)" has g = x y. Because agla x = ag/ay
we know that af/& = M/ay. This partial differential equation is satisfied by any function of z y. +
Along a path, the derivative of f (z(t),y(t)) is df ldt = (BfI&)(dxldt) +(af lay)(dyldt). The derivative of
+ +
f (x(t),y(t), z(t)) is fxxt fyyt fsnt. If f = zy then the chain rule gives df /dt = y dx/dt x dy/dt. That +
is the same as the product rule! When z = ult and y = uat the path is a straight line. The chain rule for
+
f (x,y) gives df ldt = fxul fyu2. That is the directional derivative 41f.
+
The chain rule for f (z(t,u),y (t, u)) is a f /at = (afI&)(&/at) (afIlly)(*/at) . We don't write df ldt
because f also depends on u. If z = r cos 8 and y = rsin 8, the variables t, u change to r and 8. In this case
+ +
af f a r = (M/ax) C 0 8 8 (aflay) sin 8 and a f /a8 = (af/&c)(-r sin 8) (df/dy)(r cos 8). That connects the
derivatives in rectangular and polar coordinates. The difference between ar/az = z/r and &=/ax = l/ cos 8
is because y is constant in the first and 8 ia constant in the second.
With a relation like zzyz = 1, the three variables are not independent. The derivatives (af/ax), and
(aflax). and (aflaz) mean that y is held constant, and r is constant, and both are constant. For
+
f = z2 y" +
z2 with zyz = 1, we compute (BflBz), from the chain rule af/ax (t3f/lly)(ay/&). In that
rule aslax = -I/& from the relation zyz = 1.
$ = ,/-
90
+ J901+901 (45) = 3 mph; $$= ,/-
60
(60) + ,/-
45
(45) rr 74 m ~ h
13.5 The Chain Rule (page 503)
2 s & = u 1 ~ + u 2 ~ + u 3 26
~ = 1 with x and y fixed; & =6
2 f z = lOa(ax+ by)' +
and f , = 10b(ax by)'; b f , = a f y 4 f, = - L
Z+ ~ Y and fv = &;7f= = fu*
6 = %$ + %$ dx d
is the product rule y a +x$. In terms of u and v this is $(uv) = v 2 + ug.
+
8 ftt = c2n(n- l ) ( z ct)n-2 which equals c2f , .
Choose C = -c : f = ( x - ct)" also has ftt = c2f,, .
Since sin(0 - t ) is decreasing (it is - sin t), you go down. At t = 4, your height is -sin4 and your
velocity is - cos(-4) = -COB 4.
(a) fr = 2reZiB,f rr = 2e", foe = ~ ~ ( 2 i ) ~
and
e ~frr" + I= + 9
= 0. Take real parts throughout to find
the same for ? cos 28 (and imaginary parts for r2 sin 28). (b) Any function f (reiB)has
fr = ei8f '(reiB)and frr = (eiB)2"(reiB)
f and fe = ireiBf '(reiB)and fee = i2reiBf ' + (ireiB)2f ".
+
Any f (reiB)or any f ( x i y ) will satisfy the polar or rectangular form of Laplace's equation.
!!=I'*(, + ,$y) t+2t5 1+2t3
= dl+tr = ,,l+tl0
i
since = we must find &
= 0. The chain rule gives f%- 5 = &(et) - &(2et) = 0.
g +
= (4t3)(1) (0)(1)= 4t3.
The rocket's position is x = 6t, y = t2. Its speed from (0,O)is $ 4 ( ~ t ) (t2)2
~ =36t+2t5
J(6t)a+(ta)a'
+
At t = 0 this
= 6. The rate of change of 8 = tan-' $ = tan-' $ is
A
speed is 1.
At t = 0 this is 8 .
1+(il2
4+
Driving south = (.O5)(70) = 3.5 degrees per hour. Southeast now gives = (.05)= (.01)=
4
ra 3.4 degrees per hour. is larger going south.
dta - fik+aa+attB
- 82 at a, at ax at = ( f r u l + f u u 2 + f z u 3 ) . u l + ( f z u l + f v u 2 + f ~ u 3 ) v u 2 + ( f z u l + f ~ 2 + f x ~ 3 ) z u 3= .
--
= % ( l ) + g ( u ) and
2 t and similarly +
= 2u. Check: f = (t u ) - +
~ 2tu = t2 u2 has ft = 2t and f, = 2u.
sinx+siny=O givescosx+cosyg = 0 a n d -sinx-siny(g)2+cosy$ =o. hen *=
dz and
C O S ~
fi = ainz+sin y+
&a cosy .
+ y%(cx, cy) = f (x, y). At c = 1this becomes
The c derivative of f (ex, cy) = cf (x, y) is x g ( c x , cy)
x%(xsy) y waf( x , y ) = f(x,y). Test on f =
+ d m :x . + y J- -x- F- J =- -d w .
Test on f = @ : = m.Other examples: f (x, y) = d u x 2 + bxy + cy2
orf=Ax+Byorf=x
The minimum of d2 = (x - + (y - yl)2 occurs at the rough point (xi,y l ) The graph of d is a cone and
grad d is a u n i t vector that points o u t f r o m (xi,y l ) The graph of f = lxy 1 touches bottom along the lines x
= 0 and y = 0. Those are "rough linesn because the derivative does n o t exist. The maximum of d and f
must occur on the b o u n d a r y of the allowed region because it doesn't occur inside.
When the boundary curve is x = x(t), y = y(t), the derivative of f (x, y) along the boundary is fxxt + fyyt
(chain rule). If f = x2 + 2 3 and the boundary is x = cost, y = sint, then df /dt = 2 s i n t cos t. It is zero at
the points t = 0, n l 2 , n, Ja/2. The maximum is at (0,f1)and the minimum is at ( & I ,0). Inside the circle f
has an absolute minimum at (0,0).
To separate maximum from minimum from saddle point, compute the second derivatives at a stationary
point. The tests for a minimum are f n > 0 and f n f w
> f&. The tests for a maximum are f n < 0 and f-fW 2
> fw In case ac < b2 or f,. fw < f&, we have a
saddle point. At all points these tests decide between concave up and concave d o w n and "indefiniten. For
f = 8%' - 6xy + y2, the origin is a saddle point. The signs of f at (1,O) and (1,3) are + and -.
The Taylor series for f (x, y) begins with the terms f (0,O) + xfx + yfy + zx
1 2f- + x y f V +ly2fyy.
2 The
coefficient of xn ym is p + m f / a x n a y m ( O , 0) divided by n!m! To find a stationary point numerically, use New-
13.6 Maxima, Minima, and Saddle Points ('page 512)
( z - l ) ( y- 1) $ ( y - l ) l ]
49 z = 1, y = -1 : f, = 2, f , = -2, f,, = 2, f,, = 0, f , = 2; series must recover z2 y2 +
+
51 Line z - 2y = constant; z y = constant
53 gf,. + zyf., + $f,,lo*o; f,, > 0 and f,,f,, f:;
> at (0,o);f, = f, = 0 55 A z = -1, AY = -1
57 f = z2(12 - 42) has fmM = 16 at (2,4);line has slope -4, y = 3 has slope = -4
59 If the fence were not perpendicular, a point to the left or right would be closer
A restriction g(x, y) = k is called a constraint. The minimking equations for f (z, Y) subject to g = k are
%/ax = Xag/ax, %/a
= X a g / a , and g = k. The number X is the Lagrange multiplier. Geometrically, grad
f is parallel to grad g at the minimum. That is because the level curve f = fmi, is tangent to the constraint
curve g = k. The number X turns out to be the derivative of fmin with respect to k. The Lagrange function is
13.7 Constraints and Lagrange Multipliers (page 519)
L = f (x,y ) - X(g(x,y ) - k) and the three equations for x, y, X are aL/Ebr = 0 and a L / a y = 0 and aL/t3X = 0 .
With two constraints g(x, y, z ) = kl and h ( x ,y, a) = k2 there are two multipliers Xl and At.
The five unknowns are x , y , s , X I , and X2. The five equations are fx = Xlgx + X2hx,fy= Xlgy + X2hx,
fz = Xlgs + X2hs,g = 0 , and h = 0 . The level surface f = fmin is tangent to the curve where g = kl and
h = k2. Then grad f is perpendicular to this curve, and so are grad g and grad h. With nine variables and
six constraints, there will be six multipliers and eventually 15 equations. If a constraint is an inequality g 5 k ,
then its multiplier must s a t e X 0 at a minimum.
&
1 f = x2 + ( k - 2 ~ ) ~ =; 22 - 4(k - 2%)= 0; (y, %),!$ S X = -4, xmin = 2, ymin = 2
5X = ,& '
: ( x ,y) = ((f2116,0)or ( 0 , f 2lI6),fmin = 2113; = 3 : ( 2 ,Y ) = (*I, * I ) , f m =~2
i,
7 X = ( x ,y) = (2,-3); tangent line is 22 - 3y = 13
9 ( I - c ) ~ + ( - ~ - c ) ~ + ( ~ - ~ - ~ - c ) ~ + ( z - isminimized
~-c)~ at a = -?,b= j , c =
11 (1,-1) and (-1,l); X = -+
13 f is not a minimum when C crosses to lower level curve; stationary point when C is tangent to level curve
15 Substituting = = = 0 and L = fmin leaves =X
17 x2 is never negative; (0,O); 1 = X(-3y2) but y = 0;g = 0 has a cusp at (0,O)
~ ~ ~ X = X ~ + X ~ , ~ ~ = X ~ , ~ Z = X ~ - X ~ , X + ~ + Z = ~ , X - Z = ~ ~1 at ~ V(+,0,-$)
~ S X ~ = O , X ~ = ~ ,
2 1 (l,0,0);(0,1,0);(A1, X 2 , 0 ) ;x = y = 0 2S g;
and X = 0
25 (1,0,0), (0,1,0),(0,0,1); at these points f = 4 and -2 (min) and ~ ( m a x )
27 By increasing k , more points are available so fmu goes up. Then A = % 20
29 (0,O);X = 0; fh stays at 0
+ +
3 1 5 = X I X 2 , 6 = X 1 As, X 2 2 0 , X3 5 0; subtraction 5 - 6 = X 2 - X3 or -1 2 0 (impossible);
+
x = 2004, y = -2000 gives 52 6y = -1980
+ + +
3 s 22 = 4X1 12,2y = 4X1 As, X 2 2 0, X3 1 0,4x 4y = 40; max area 100 at (10,0)(0,10);min 25 at (5,5)
2 x2 + y2 = 1 and 2xy = X(2x) and x2 = X(2y) yield 2X2 + X 2 = 1. Then X = -& gives x,, =f 9,
1+ -& = X(a + ih). Put Xb = 2 in the second equation and square: (2h)2 = $ + h2 or h2 = b2/12.
If R is the rectangle 0 5 x 5 4,4 5 y 5 6, the integral $$ x dA can be computed two ways. One is x dy dx,
when the inner integral is xy]: = 2x. The outer integral gives x2]t = 16. When the x integral comes first it
equals $ x dx = ix2]t = 8. Then the y integral equals 8y]z = 16. This is the volume between t h e base
rectangle and t h e p l a n e x = x.
The area R is $$ l d y dx. When R is the triangle between x = 0, y = 22, and y = 1, the inner limits on y
1. The area is .I;1
are hc and 1. This is the length of a t h i n vertical strip. The (outer) limits on x are 0 and 2
1 Now the strip is horizontal and the outer integral
In the opposite order, the (inner) limits on x are 0 and 2y.
is :$ iydy = 3.1 When the density is p(x, y), the total mass in the region R is $sp dx dy. The moments are
My = $$ px dx dy and Mx = $$ py dx dy. The centroid has Z = M,/M.
27 1,'Jt(22 - 3y + 1)dx dy = a 1: 1:
29 f (x)dz = $:(" Idy dx 31 5 0 , 0 0 0 ~
3 3 $ ~ $ ~ x 2 d x d y = ~3 5 2 J o' / a $Ow Idx dy = a4
*-L
3 7 +C;=lC;=l f (-,
n
a)
n
is exact for f = 1,x, y, xy 39 Volume 8.5 4 1 Volumes in 2,2 ln(1 + 4)
43 $
1 1
5,'
: $: xsdx dy = $: &dy = 1n2;J0 Jo xzYdydx = e d x = In2
45 With long rectangles yiAA = C AA = 1but s$y dA = i
2 $: 2xy dx = x 2 y ] =
~ (e2 - 1)y; J22e 2
(e - I ) dy ~ = (e2 - l)$]ie= (e2 - 1)(2e2- 2) = 2(e2 - 112;
$: = 1 " ~e =
711
.L.sZe
11' 2
&
y
= l n 2 e - l n 2 = I n % = 1.
4 $: yeZYdx = e z ~ ] =
? - eY; S (e2y - eY)dy = [ i e 2 y - e'l'o = ; e 2 - e + - , 21 . J 3 ,/A=
d
24T%Ty]:=
24- + +
- 2 4 m ; the x integral is [$(6 2x)3/2 - $(3 2x)3/2]!-l= $g3I2 - $53/2 - $43/2 + $.
Note! 3 + 22 + y is not zero in the region of integration.
6 The region is above y = x3 and below y = x (from 0 to 1). Area = $i(z - x3)dx = [$- $1: = z.
1
8 The region is below the parabola y = 1 - x2 and above its mirror image y = x2 - 1.
Area = $-1(1- x2 - x2 + 1)dx = [2x- $x3]?1 =
1
g.
10 The area is all below the axis y = 0, where horirontal strips crws from x = y to x = lyl (which is -y). Note
0
that the y integral stops at y =O. Area = l!lJ;Ydxdy= J-l -2ydy= [-y2]!!l = 1.
1 2 The strips in Problem 6 from y = x3 up to x are changed to strips from x = y across to z = y113.T he outer
integral on y is by chance also from 0 to 1. Area = J,'(y1/3 - y)dy = [gy4/3- ?y2]i = a-
14 Between the upper parabola y = 1- x2 in Problem 8 and the x axis, the strips now cross from the
left side x = -4- to the right side x = + d G . This half of the area is $--
1 ,p=i
sodx dy =
$,' 24I-y' dy = -$(I - y)3/2]A = $.The other half has strips from left side to right side of y = x2 - 1
or x = *JG. This area is J_OIJ- dx dy (also k).
16 The triangle in Problem 10 had sides x = y, x = -y, and y = -1. Now the strips are vertical. They go
from y = -1 up to y = x on the left side: area = J_O1$f d dx = J:~(X + +
1)dx = ?(x l)2]?1 = The i.
so
strips go from -1 up to y = -z on the right side: area = :/ dy dx = (-2
yl
1)dx =+ ?.
Check: ?+?
= 1.
18 The triangle has corners at (0,O) and (-1,0) and (-1, -1). Its area is /_Ol$;' dy dx = Jb$_;Y
dx dy(= 4).
20 The triangle has corners at (0,O) and (2,4) and (4,4). Horinontal strips go from x = to x = y : :
area = J,'Cl2
dx dy = 4. Vertical strips are of two kinds: from y = x up to y = 22 or to y = 4.
22 (Hard Problem) The boundary lines are y = $x from (-2, -1) to (0,0), and y = -22 from (0,O) to
(1, -2), and y = or x = -3y - 5 from (-2, -1) to (1, -2). (This is the hardest one: note first the
slope -3.) Vertical strips go from the third line up to the first or second: area = J:2fi-6,
dy dz+
5 4.
*.
1: 4::- dy dx = + = Horinontal strips cross from the k t or third Iines to the second:
5 ),3
area = $ _ ; ' J ' ~ ~ ~+ J_O~J;'/~
dx' dy dx dy = '
q +q = f*
a bx/a
24 The top of the triangle is (a, b). &om x = 0 to a the vertical strips lead to Jo Jblc dy dx =
- = $ - $.
[gb+(x-a)(d-b)/(c-a) Fkom x = o to c the strips go up to the third side:
c
Ja j'x 1c
The s i m is $+ + 2 - = 9 bw.
d y d x = [ b x + ~ ~ - ~ ] ~ = b ( c - a2 ) 2c
This is half of a parallelogram.
I:,/," g d x dy = ~ : [ f ( ~Y),- f(0, Y ) I ~ Y .
+ '~ - ~ + ~
The integral /:/: x2dx dy has the usual midpoint error -%for the integral of z2 (see Section 5.8).
The y integral lo 1
dy = 1 is done exactly. So the error is -&
(and the same for JJ dy). The
integral of zy is computed exactly. Errors decrease with exponent p = 2, the order of accuracy.
14.2 Change to Better Coordinates (page 534)
+
We change variables to improve the limits of integration. The disk z2 y2 5 9 becomes the rectangle
0 I. r I 8,O I.8 I. 2*. The inner limits of $ ' d y dz are y = f a .In polar coordinates this area integral
becomes /I r dr d8 = 9%
A polar rectangle has sides dr and r dB. Two sides are not straight but the angles are still 90'. The area
bet-n the circles r = 1 and r = 3 and the rays 8 = 0 and 8 = r / 4 is B(S2 - 12)= 1. The integral $$ z dy dx
changes to $$ r2cos 0 dr dB. This is the moment around the y axis. Then 5 is the ratio My/M. This is the z
coordinate of the centroid, and it is the average value of x.
In a rotation through a, the point that reaches (u,v) starts at x = u cos a - v sin a, y = u sin a v cos a. +
A rectangle in the uv plane comes from a in zy. The areas are equal so the stretching factor is
J = 1. This is the determinant of the matrix
$1(u eos a - v sin a)2du dv.
cosa -sina
sina cosa I . The moment of inertia I$ z2dz dy changes to
For single integrals dz changes to (dx/du)du. For double integrals dx d y changes to J du du with J =
a ( x , y ) / a ( u ; v ) .The stretching factor J is the determinant.of the 2 by 2 matrix ax/*
functions z(u,v) and y(u, v ) connect an zy region R to a uv region S , and dx dy =-fi
ax/& , ~h~
ay/au a y / h I
J d u d v = a& of I€.
For polar coordinates z = u cos v and y = u sin v (or r sin 8). For z = u,y = u+ 4v the 2 by 2 determinant is
J = 4. A square in the uv plane comes from a parallelogram in xy. In the opposite direction the change has
u = z and u = i(y 1 This J is constant because this change of variables is linear.
- z ) and a new J = 3.
I$,"'"$~rdrdO=~ SS=qnarter-~irclewithu~Oandv~0;$,'$~~dudv
6 R is symmetric across the y axis; I,' loQ u du dv = $ divided by area gives (ti, ir) = (4/3r,4 / 3 4
7 2~,'l~$;:~-dy dz;zy region R* becomes R in the x*y* plane; dz dy = dz'dy* when region moves
J=l az/ar*
ay/ar*
&lag*
ay/ae* I=/ cos 8*
sine*
-r* sin 8*
r*cos0* 1 3~14 1
= r* ;J I 4 r*dr*de*
1 I
23 J = 2 1 = 3, area Jo 1 so3du dv = 3; J =
1 2e2u+v e2u+v
eu+2v 2eu+2v - 3e3u+3u,Jo1 Jo1 3e3u+3vdudv =
(e3+3w- e3')dv = i(e6 - 2e3 + 1)
25 Comers (2, Y) = (o,o),(l,o), (1,f (111, (0, f (0)); (& 1) gives 2 = is
Y = f (+I; J = I 1
vf.(u)
0
f (u)
2 Area = ~3~~
h y dy dz splits into two equal parts left and right of z 1
= 0 : 2 1, J,
Gdy
- dz-
2 $'&I2(4- - z)dz = + sin-' z - z2]p12 = sin-' 9 = 2. The limits on
JJ dx dy are lP2J'Yy dx d y for the lower triangle plus JJ 212 J- -m
,/s
. -
dx d y for the circular top.
--
g=.*=.&. 2
For g divider$
(:/ a3( l - cos b ) by the area
sin B)r dr dB = T 9
to find g = E ( l - cos b).
8 The limits on r,B are extremely awkward for R*. Contrast with the simple limits 0 r* 5 1, f B* <
when the coordinates are recentered at (0,l). (A point on the lower boundary of the wedge has
sin
r= by the law of sines.)
+
10 The centroid (0,g) of R mwes up to the centroid (0,j7 1)of R*. The centroid of a circle is its
+
center (l,2). The centroid of the upper half is (l,2 )! because a half-circle has J;J;(~ sin B)r dr dB = 18
divided by its area ?f (which gives $).
9 3+/4 =
[==I 344 1
4 4 lo(r s i n B + l ) 2 r d r d B = ~ J s i n 2 8 d B + ~ J s i n B d B + ~ ~ d B = [ ~ - ~ - ~ c o s B + z ] , / 4
Problem 18 has J = I 1 1
I = 1. So the area of R is l x area of unit square = 1. Problem 20 has
v u
J = I -2u 2v
I= 2(u2 d), + and integration over the square gives area of R =
+ f.
$: :$ 2(u2 d ) d u dv = Check in z, y coordinates: area of R = 2 1 (1- z2)dz = $.So
1
ar/& ar/ay
a s / a z ae/ay = 1 -y/r2 z/r2
1 'Ir I
= = .; As in equation 12, this new J is A.
00
)I-,
~ = (u)(v) - $ vdu = ( z ) ( - e - ~ ' / ~
x2e - ~ ' /dz +$-,00 e-z1/2dx = 0 +6 by Example 5. Divide
by 6 to find cr2 = 1.
R is an infinite strip above the interval [OJ] on the z axis. Its boundary z = 1 is r cos 6 = 1or r = see 6.
The limits are 0 5 r 5 sec 6 and 0 5 6 5 5.
The integral is '/$: $let $ d B *
= $I:2 ( 0 0 )d6 = infinite.
For a finite example integrate (z2 y2)-1/2 = f a
Six important solid shapes are a box, prism, cone, cylinder, tetrahedron, a n d sphere. The integral
$$$ dx dy dz adds the volume dx d y d z of small boxes. For computation it becomes t h r e e single integrals.
I
The inner integral dz is the length of a line through the solid. The variables y and z are held constant. The
double integral $$ dz dy is the a r e a of a slice, with z held constant. Then the z integral adds up the volumes of
slices.
+ +
If the solid region V is bounded by the planes z = 0, y = 0, z = 0, and z 2y 32 = 1, the limits on the
inner z integral are 0 and 1- 2y - 3s. The limits on y are 0 and 1 z(1- 1 In
3s). The limits on z are 0 and 3.
the new variables u = x, v = 2y, w = 32, the equation of the outer boundary is u v + +
w = 1. The volume
1
of the tetrahedron in uuw space is 8. mom dx = du and dy = dv/2 and dz = dw/J, the volume of an xyz box
is dx dy dz = 81d u dv dw. So the volume of V is g1
g.
1$:
$:dz dy dz =
3 0 I y I z 5 z 5 1 and all other orders zzy, yzz,zzy,zyz; all six contain (O,O,O); to contain (1,0,1)
14.3 DipIe h tegrah (page 540)
23 Horizontal slices are circles of area lrr2 = r ( 4 - I); volume = lolr(4 - z)dz = 8 r ; centroid
4
has z = 0,g = 0 , r = ~ t Z l ( -
4 z)dz/8r =
25 I = $ gives zeros; = fl
f dy dz, = lo.l : f dx dz, BvB'I = f dx
', J!, $!1(y2 + z2)dx dy dz = y;JJJ x2dV = 3 JJJ(x - T ) 2 d V =
27 J! i;
29
1
: dx dy dz = 6 S 1 Tkapesoidal rule is second-order; correct for 1,x, y, z, xy, xz, yz, xyz
1 1 1
2 ~ h e a r e a o f 0 < x sy < z 5 1 i s $ o 1x Y dxdydx.Thefourfacesarex=O,y=x,z= y , z = l .
4
I,' $: x dx dy dz = J; $dy dz = I,' $dz = &. Divide by the volume a to find f = i;
1 S
and j? = 2 ; by symmetry E = ;l.
1 2
$,'~:~:ydzd~ lo
d z = lo d d y d z = $dz= I,'
6 Volume of half-cube = J', l!,
J', dz dy dz = 4.
$0' dx dy dz = /
: 2(z +
1)dz = [(z l)l]; = 3. +
8 /_II
10 $
,!
dz dy dz = I!,
Ji1(y 1)dy dz = + w d r= /!,
+ +
= (tetrahedron).[w]~,
#
12 The plane faces are x = 0, y = 0, z = 0, and 2x y z = 4 (which goes through 3 points). The volume
2 4-2s 4-2s-y 2 4-2s
is $0 /o 0
I dz dy dz = lo/, (4 - 22 - y)dy dz = v d x = [-w]i 12 = -ST-*
= A? l6
Check: Multiply standard volume by (4)(4)(2) =
(4-y-s)/2
y.
Check: Double the volume in Problem 11.
14 Put dz last and stop at z = 1: lolo lo dx dy dz = lo
1 4-z 1 4-2
Y d y dz =
$0'
qL'& = [ (4;;)=]; = 4=-3= - 37 12 m*
16 (Still tetrahedron of Problem 12: volume still ). Limits of integration: the top vertex
falh from (0,0,4) onto the y axis at (0, -4,O). The corner (2,0,0) stays on the x axis.
The corner (O,4,0) swings up to (O,O,4). The volume integral is 16
J!~J: dz dydz = 7.
18 The plane z = x cuts the circular base in half, leaving z 2 0. Volume = lo
41-za
/:
dz dy dx =1
/,'2~1/~dx=[-$(l-x 2 ) l o = 3.
312 1 2
20 Lying along the x axis the cylinder goes from x = 0 to x = 6. Its slices are circular disks y2 (z - 1)2= 1 +
resting on the x axis. Volume = lo1- A- ds
l+dl-ua dz dy dx = still 6 r .
22 Change variables to X = f, Y = f , Z = f; then dXdYdZ = Volume = "2".
abc dXdYdZ = l/l
1 Centroid (z,g,g) = ( a x , by, cZ) = (q,q, i
f ). (Recall volume and centroid ( i , f ,4) of standard
tetrahedron: this is Example 2.)
24 (a) Change variables to X = j,Y = t ,Z = F. + +
Then the solid is Xa Y2 Z2 = 1, a unit sphere of volume
4%
,. Therefore the original volume is (4)(2)( $) = 9 9.
( b) The hypervolume in 4 dimensions is 1 m,
following the pattern of 1 for interval, f for triangle, for tetrahedron. 6
26 Average of f = jVf (x, y, z)dV/
,/ dV = integral of f(x?y,a) divided by the volume.
28 Volume of unit cube = Ez1 A 1/Az
C:I?~(AX)~= 1.
SO In one variable, the midpoint rule is correct for the functions 1 and x. In three variables it is correct for
1, X? Y? a? xY?m ?YZ? W E *
i i
52 Simpson's Rule has coefficients ,$ , over a unit interval. In three dimensions the 8 corners of the cube will
have coefficients (a)3 = A.
The center will have ( i ) 3 = E.
The centers of the 12 edges will have
= A.
The centers of the 6 faces have = &. (Check: 8(1) 64 12(4) 6(16) = 216.) + + +
When N3 cubes an stacked together, with N small cubes each way, there are only 2N 1 meshpoints +
+
along each direction. This makes (2N 1)3points or about 8 per cube. (Visualire the 8 new points
of the cube as having x , y, z equal to rero or $.)
The t b r a spherical coordinates are p#8. The point at x = y = z = 1has p = fi,# = cos-ll/t/S, 8 = ~ 1 4 .
The angle # is measured from t h e r axis. 8 is measured from t h e x axis. p is the distance to t h e origin,
where r was the distance to t h e r axis. If p#8 are known then x = p sin 4 cos 8, y = p sin # sin 8, s = p cos 4.
The stretching factor J is a 3 by 3 determinant and volume is I$I
9 sin 4 dr d4 dB.
< <
The solid region 1 p 5 2,O 5 # r ,0 S 8 <_ 2 r is a hollow sphere. Its volume is 4 ~ ( -2 l3)/s. ~ &om
the 4 and t9 integrals the area of a spherical shell at radius p equals 47rp2. Newton discovered that the outside
gravitational attraction of a sphere is the same as for an equal mass located at t h e center.
(a
B)=(D n ,3-~ , ~ )4 ( r , 8 , ~ ) = ( 5 , ~ 0 ~ - ~ $ , 5 ) ; ( p , 4 , 8 ) = ( 5 J Z , ; , ~ 0 ~ - ~ 9 )
z ) = ( l ) 8z=ronthepositivexaxis(z~0,y=0(=8),~=0)
-
I =gs
in t. The unit sphere intersects the plane y = z.
+
12 The surface z = 1+ r2 = 1 z2 + y2 is a paraboloid (parabola rotated around the z axis). The region is
above the halfdisk 0 5 r 5 1 , O 5 8 5 r. The volume is qr.
< 1 4.
1 4 This is the volume of a half-cylinder (because of 0 B 5 r) : height 7r, radius r, volume ~r
16 The upper surface p = 2 is the top of a sphere. The lower surface p = see 4 is the plane z = p cos 4 = 1.
(The angle # = is the meeting of sphere and plane, where sec 4 = 2.) The volume is
2~"$
: (v) sin 4 d 4 = 2aI-t cos 4 - &]:I3 = 2x1- $ - -+ + - &
- 3'
i]
18 The region 1 5 p < 3 is a hollow sphere (spherical shell). The limits 0 < # 5 f keep the part that lies
above a 45' cone. The volume is y(1-9).
20 Fkom the unit ball p 5 1 keep the part above the cone # = 1 radian and inside the wedge 0 8 5 1 radian. <
!
Volume = $,'sin #d# = z1( l - coo I).
23 The curve p = 1- cos # is a cardioid in the zz plane (like r = 1- cos 9 in the z y plane). So we have a
cardioid of revolution. Its volume is 9
as in Problem 9.3.35.
24 Mass = Jo
2n
So
n R
J0 p sin #(p + 1)dp d# dB = JrR3 + 2 r R 2 . 1 1
26 Newton's ochievcment The cosine law (see hint) gives cos a = -P . Then integrate :
J$/(~% + &)dV. The second integral is
uses the same u = D2 - 2pD cos #
$ JJf 5=
+ p2 = q2 as in the text:
s.
so
%-=
29D
The first integral over q5
d u 2 D -
9
,, $+-
[*$I: = &(-1 -1 'D+P = D(Df-PI) The 9 integral gives 2x and then the p integral is
:$ a r e j w p 2 d p = w.
The two integrals give as Newton hoped and expected.
28 The small movement produces a right triangle with hypotenuse AD and almost the same angle a. So the
new small side Aq is A D cos a.
3 0 f $ q d A = 47rp2D g.
+ +
Divide by 4xp2 to find = D
2
for the shell. Then the integral over p gives
42 The ball comes to a stop at Australia and returns to its starting point. It continues to oscillate in harmonic
motion y = R cos(dc/m t).
15.1 Vector Fields (page 554)
A vector field assigns avector to each point (x, y) or (x, y, 2). In two dimensions F(x, y) = M(x, ~ ) i + N ( xyU.
,
+
An example is the position field R = x i y j (+x k). Its magnitude is lRl = r and its direction is o u t &om t h e
+
1 2 y2). The level curves are circles, and they are perpendicular
origin. It is the gradient field for f = i(x
to the vectors R.
+
Reversing this picture, the spin field is S = -y i x j. Its magnitude is IS1 = r and its direction is a r o u n d
the origin. It is not a gradient field, beeause no function has a f / a x = -y and a f / a y = x. S is the velocity
field for flow going a r o u n d t h e origin. The streamlines or field lines or integral curves are circles. The flow
field pV gives the rate at which mase is moved by the flow.
A gravity field from the origin is-proportional to F = a/rswhich has IF / = l/r2. This is Newton's inverse
square law. It is a gradient field, with potential f = l/r. The equipotential curves f (x, y) = c are circles. They
are perpendicular to the field lines which are rays. This illustrates that the gradient of a function f (x, y) is
perpendicular to its level curves.
+
The velocity field y i x j is the gradient of f = xy. Its streamlines are hyperbolas. The slope dy/dx of a
streamline equals the ratio N/M of velocity components. The field is tangent to the streamlines. Drop a leaf
onto the flow, and it goes along a streamline.
lf(x,y)=z+2y
11F=3i+j
27 F = y i + yj
~ l F = $ y i - ~ x j
35 = gE = g:;
5f(x,y)=sin(x+y)
7 F = zyi + $j, f ( z , y) =
lSF=i+2yj
19 z = - l ; y = - ~ + C
9
29 F = -:fG
= g ~ ;
9
1SF=2xi-2yj
2 1 2 = - ~I/'
+
sJ g = Sza = - L 2 1O
5f(x,y)=1n(x2+#)=21nr
mMG
+
= 0 so f cannot depend on x; streamlines are vertical (y = constant)
17F=ez-'i-eZ-gj
- ~ 2 + y 2 = C 2 3 $ = - x Y = z Y; x 2 + y 2 = C
(xi y j) - ((Z-l)a+va an((x- 1) i + ~ j )
d ~4 ,/=-
U = A 2I
f(r) = C gives circles
asparallel
+ +
57 T; F (no equi~otentials);T; F (not multiple of xi yj zk)
+
39 F and F i and 2F have the same streamlines (different velocities) and equipotentials (different potentials).
But if f is given, F must be grad f.
- #x2 = c.
10 If *
z f i = C or y = @/z2. The streamlines solve dy/dz = N / M = z/2y so 2y dy = x dz or y2
= -y then f = -yx+ a n y function C(y). In this c w a, = -x+ dC which can't give = x.
12 % = 1and % = -3; F = i - 3j has parallel l i e s z - 3y = c as equipotentials.
14 % = %c - 2 and % = 2y;F = (22 - 2)i + 2yj leads to circles (z - 1)2+ y2 = c around the center (1,O).
16 = iX cos y and 5 = -iX sin y ; F = ez (cos yi - sin yl) leads to curves e' cos y = c which stay inside
a strip like lyl < g. (They come in along the top, turn near the y axis, and leave along the bottom.)
18 %=
X
3 and = 4; +
F = - 3 1 f j has the rays f = c as equipotentials (omit the axis z = 0).
20 $ = x gives y = zx +
1 2 C (parabolas).
Y
+
22 $ = - gives y2 x2 = C (circles).
24 $ = 3 gives y = 2x + C (parallel lines).
1
Work is the integral of F dR. Here F is the force and R is the position. The d o t product finds the
component of F in the direction of mwement dR = dx i + dy j. The straight path (z, y) = (t, 2t) goes from (0,0)
+
at t = 0 to (1,2) at t = 1with dR = dt i 2 d t j.
Another form of dR is Tds, where T is the unit tangent vector to the path and the arc length has ds =
\/(dx/dt)t + (dy/dt)2. + a)/&
For the path (t, 2 t ) , the unit vector T is (i and ds = fidt. For F = 3i j, +
F . T ds is still bdt. This F is the gradient of f + +
= Sx y. The change in f = 3z y from (0,0) to (1,2)is 6.
+
When F = grad f, the dot product F dR is (af /ax)dx (af/*)dy = df. The work integral from P to Q
is df = f(Q) - f(P).In this case the work depends on the endpoints but not on the path. Around a closed
+ +
path the work is zero. The field is called conservative. F = (1 y) i x j is the gradient of f = x + xy. The
work from (0,O) to (1,2) is 3, the change in potential.
+
For the spin field S = -y i x j, the work does depend on the path. The path (x, y) = (3 cos t, 3 sin t) is a
+
circle with S-dR = -y dx x d y = 9 dt. The work is 1 8 r around the complete circle. Formally / g(z, y)ds is
the limit of the sum g(q,yi)Asi.
The four equivalent properties of a conservative field F = M i + N j are A: zero work a r o u n d closed paths,
15.2 Line Integrals (page 562)
B: work depends only on endpoints, C: gradient field, D: aM/ay = aN,/ax. Test D is passed by F =
(y + 1)i + 4.The work J F dB around the circle (cos t, sin t) is Bero. The work on the upper semicircle equals
the work on the lower ~ e m i c i r c l e(clockwise). This field is the gradient of f = x + xy, so the work to (-1,O)
is -1 starting f i o m (0,o).
a
2 1 x = a cos t, y = a sin t, ds = a dt, M = J'," a3 cos2 t dt = ra3, (Z, p) = (O,O)by symmetry
23 T = =&
F
)-;,
+
= 3 x i + 4 j =6t i+4j,ds = 2 ~ - d t , F - ~ d s
+
20t dt; F dR = (6ti 4). (2 dti 2t dt j) = 20t dt; work = 20t dt = 30
=(6ti+4~).(+)2dmdt
m =
25 1f 9 9 = + +
then M = ay b, N = ax c, constants a, b, c
(1 1) 2 (14 = 1
27 F = 4xj (work = 4 from (1,0) up to (1,l)) 29 f = [x - 2y](,:,) = -1 $1f = [zy I(,, )
33 Not conservative; (ti - tj) (dt i dt j) = + SO : (t2i - tj) (dt i
dt = 0; J + 2t dt j) = -t2dt = -$
56 = ax, +
= 2x b, so a = 2, b is arbitrary 57 a
a, = 2ye-' = E; f = -y2e-'
aM
39 - =
&r T = g ; f = r = d m = ~ x i + y j l
4 1 ~ = ( x - ~ ) i + ( x + ~h)aj s = - 1 , % = 1 , n o f 4821r;O;O
2 ~ o t e d s = \ / s i n ~ t + c o s ~ t d t = d then
. ~ x d s = ~ ~ l ~ c o s t and
dt~= zl y d s = ~ : ' ~ s i n t c o s t d t = f.
4 Around the square 0 5 x, y 5 3, JsOy dx = -9 along the top (backwards) and J : -x dy = -9 up the
right side. All other integrals are rero: answer -18. By Section 15.3 this integral is always -2 x area.
6 J $ d t = J d s = a r c length = 6.
8 Yes The field xi is the gradient off = $x2. Here M = x and N = 0 so we have Mdx+Ndy = f (Q)- f (P).
More directly: up and down movement has no effect on x dx. /
10 N o t much. Certainly the limit of C ( A S ) ~ is rero.
12 = 0 and = 1; n o t conservative, take straight path x = 1- t, y = t : J F d R = y dx dy = I +
J,' t(-dt) + dt = 4.
14 = and F is the gradient of f = xev. Then I F . dR = f (Q) - f ( P ) = -1.
16 # F;not conservative, choose straight path z = 1-t, y = t : -y2dz+x2dy = Jt2dt+(l-t)2dt = 8.
18 5 has M = = -xnY(x2 + y2)-(n/2)-1. This agrees with % so 7
,2-n
R is a
2 2
(c) W = :$ 2 +
(+ dt) s ( y dt) = f . But k+ # M (not consewative).
I,' +
(d) W = cd+"(z dt y dt) = eZ+Y- 1. Then = ec'+ and = ee'+ (conservative).
28 F = x2j on the circle x = c ~ s t , ~ = s i has
n t J F - ~ R J~'cos2t(cost
= dt) = 0.
: t2dt = but /,' t2(2t dt) = 1.
SO / x2dy = J 1
82 # + $01 4
(not conservative): x2Ydx x d d y = 2t3dt = but :$ t2(t2)dt t(t2)2(2tdt) = 17 + s.
34 The potential is f = )ln(x2 + +
y2 1). Then f (1,l) - f (o,o) = 1x1 S. f
86 +
-t2(-2t dt) (1 - t2)(2t dt) = 1(as before). On the quarter-circle ending at t = f :
J,"'~(- sin 2t) (-2 sin 2t dt) + (cos 2t) (2 cos 2t dt) = 2 = as before. 8
38 - = -2ycZ - 2ycS # 0. N o potential f (2, y).
4 0 ~ y'+za
= ~ h a s ~ # ~ .
dy
4 2 BatE = - if and only if b = c. Then f (x, y) = #ax2 bxy i d d . + +
44 Puc because / F dR = $ y dx. False because F = yi is not conservative. (The area underneath
depends on the curve.) !the because the area is 7r (and $ y dx = c"
sin t(sin t dt) = r.)
+
The work integral f M dx N dy equals the double integral J J ( N ~-My)& d y by Green's Theo-
rem. For F = 3i +
4 the work is sero. For F = 4 and -yi the work equals the area of R. When
M = a f /ax and N = a f l a y , the double integral is zero because fxy = fF. The line integral is zero because
+
f(Q) = f (P) when Q = P (closed curve). An example is F = y i x j. The direction on C is counterclock-
wise around the outside and clockwise around the boundary of a hole. If R is broken into very simple pieces
+
with crosscuts between them, the integrals of M dx N d y cancel along the crosscuts.
Test D for gradient fields is a M / a y = aN/&. A field that passes this test has IF
. dR = 0. There is a
+
solution to f. = M and f, = N. Then df = M dx N dy is an exact differential. The spin field 8/13 passes test
D except at r = 0. Its potential f = 9 increases by 2 r going around the origin. The integral I$(N, - M,)dx dy
is not rero but 2r.
The flow form of Green's Theorem is fC M d y - N dx = $ h ( M x + Ny)& dy. The normal vector in F-nds
points o u t across C and In1 = 1and n ds equals dy i - dx j. The divergence of Mi Nj is Mx Ny. For F =+ +
zi the double integral is Ill
dt = area. There is a source. For F = y i the divergence is zero. The divergence
of R/r2 is zero except at r = 0. This field has a point source.
A field with no source has properties E = aero flux through C, F = equal flux across all paths
from P t o Q , G = existence of s t r e a m function, H= zero divergence. The stream function g satisfies
+
the equations ag/* = M and a g / a x = -N. Then aM/ax aN/ay = 0 because aag/azay = aag/ay &.
The example F = yi has g = &y2. There is not a potential function. The example F = xi - yj has g =
xy and also f = &x2- iy2.
This f satisfies Laplace's equation f= +
fyy = 0, because the field F is both
conservative and source-free. The functions f and g are connected by the Cauchy-Riemann equations
15.3 Green 's Theorem (page 571)
1 $ln
( a cos t) a cos t dt = aa2; Nz - My = 1,$1dz dy = area aa2
3$ ~ z d z + $ ~ z d z = ~ , ~ , - ~ , = O , $ $ ~ d z d ~ = ~
5 $ z2y dz = $:"(a cos t)2(asint)(-a sin t dt) = -$ $:"(sin 2t)2dt = -$;
Nz - My = -z2,$$(-z2)dz dy = So So -r2cos2B r drd9 = -$
2n a
7$zdy-ydz=$~(cos2t+sin2t)dt=a;$$(1+l)dzdy=2 (area) = r ; $ z 2 d y - z y d z = $ + l ;
$l +
(22 z)dz dy =
'+Y
=
15.4 Surface Integrals (page 581)
so M2 + Nu = 0 : Double integral = 0. Along the bottom edge (where y = 0 and n = -j) the line integral
is .nds = 1,' : -
= -1. The right side (x = 1and n = i) yields J =-,/---=l-fi, &$
a
,*
2 +O l + y lo
B;rcL m o s s the top (y = 1,n = j, notice ds = -dx!) /: = fi - 1. Down the left side (notice
ds = -dy !) gives +l. Adding the four sides I n ds = 0.
20 F = grad r = (2, F) has F . n = 0 along the + axis where n = -j and y = 0. On the unit circle
n i. equal to F (unit vector pointing outward) so F .n = 1. Amund the semicircle
~ ~ ~ n d s = ~ ~ l d 8 = r . T h e d o u b l e i n t e g r aMl h2 a=s& ( f ) = $ - f g = $ - $ = f .Similarly
!.
N, = a6u( n+ ) = $ and M, + Nu = f = The double integral is I,'
i ( r dr d8) = T .
22 f F n ds is the same through a square and a circle because the difference is j J ( E + +)dz dy =
I$ div Fdx dy = 0 over the region in betwem.
26 div 3 = A(+f)
82 2 +u + #-(,*$)
This is infinite at x = y = 0.
= = 0. Integrating ,*
24 j(cos3 y dy - sin3 x dx) = $$(o- 0)dx dy = 0. A different example would be more revealing.
,m gives g = #ln (x2 +y2) = Inr.
38 g(Q) = J ' F nds starting from g(P) = 0. Any two paths give the same integral because forward on one
and back on the other gives f F .n ds = 0, provided the tests E - H for a stream function are passed.
+
4 0 With M2 N, = 0 we can solve ag/ay = M = 3x2 - 3y2 and ag/ax = - M = 6x9 to find
g = SYC2y -ys. Then f2 = g, = A4 and fu = -g, = N.
42 Mdy - Ndx is an exact differential if =- z.
(Then there is a stream function g.)
4 4 ~ I ~ . d R = f - ~ d x + x d ~x =area#O.
2
46 Simply connected: 2, 3, 6(?), 7. The other regions contain circles that can't shrink to points.
A small piece of the surface a = f ( x , y) is nearly flat. When we go across by dz, we go up by (az/ak)dx. That
+
movement is Adz, where the vector A is i ds/dx k. The other side of the piece is Bdy, where B = j (az/ay)k. +
+
The cross product A x B is N = - a s f a x i - as/* j k. The area of the piece is dS = INldx dy. For the
surface x = xy, the vectors are A = fl J-dx d y and N = -y i - x + k. The area integral is
IldS=i+yk.
With parameters u and u, a typical point on a 45' cone is x = u cos u, y = u sin v, z = u. A change
+
in u moves that point by A du = (cos v i + sin v j k)du. The change in u moves the point by Bdu =
(-u sinv i + u cos v j)dv. The normal vector is N = A x B = -u cos v i - u sin v j u k. The area is +
dS = 6 udu dv. In this example A B - 0 so the small piece is a a rectangle and d S = IAIIBldu du.
For flux we need ndS. The unit normal vector n is N = A x B divided by INI.For a surface z = f (z,Y),
15.4 Surface In te~rah (page 581)
the product n d S is the vector N dz dy (to memorize from table). The particular surface z = zy has n d S =
+ + +
(-yi - qj k)dz dy. For F = zi yj zk the flux through z = zy is F .n d S = -xy dz dy.
On a 30' cone the points are z = 2ucos v, y = 2usin v, z = u. The tangent vectors are A = 2 cos v i
+ 2 s i n v j + k and B = - 2 u a i n v i + 2 u c o s v j . This cone has n d S = A x B du dv = ( - 2 u c o s v i - 2
+ + +
u sin v j 4 u k ) d u dv. For F = x i y j z k , the flux element through the cone is F . n d S = zero. The
reason for this answer is that F is along t h e cone. The reason we don't compute flux through a Miibius strip
is that N c a n n o t b e d e 5 e d (the s t r i p is n o t orientable).
+
N = -221 - 2yj k; d S = dl + 4z2 + 4 9 dz dy; lo sod
2% 2
w r dr d0 = t(173/2 - 1)
N=-i+j+k;ds=4dxdy; area f i r
+
N = (y2 - z2)i - 2zyj k; d S = dl + +
(y2 - z2)2 4z2y2dz dy = dl +
(y2 + z ~ ) dy;
~ ~ z
1:"
6.-
:$ r dr d0 =
Jz +
+a +
N = 2i +
k; dS = 3dz dy; 3(area of triangle with 22 2y 5 1) =
r a J W 15 $ :
$-': zy(& dz dy) = $
'J: $,"I4 sin2 4 cos 4 sin 0 cos 0(sin 4 d4 d0) = 0 l 9 A = i+j+2k; B = j+k; N = -i-j+k; d S = f i du dv
+
A = -sinu(cosv i + s i n v j) +cosu k ; B = -(3+cosu)sinv i ( 3 + cosu)cosv j;
N = -(3+cosu)(cosucosvi+cosusinvj + s i n u k ) ; d S = (3+cosu)dudv
$$(-ME - N% + +
P)dz dy = IS(-2z2 - 2y2 s)dx dy = $$ -r2(r dr dB) = -8n
+ +
F . N = -z y z = 0 on plane
N=-i-j+k,F=(v+u)i-uj,/$~.Nds=$$-vdudv=~
I$dS = so2n So2n (3 + cos u)du dv = 12r2 3 1 Yes 3 3 No
A = i + flcosOj+ f 1 s i n 0 k ; B = - f s i n B j + f c o s B k ; N = f f ' i - f c o s 0 j - f s i n B k ; d S = lNldzdB=
f ( z ) J W dz d0
2 ~ = - 2 x i - 2 ~ j + k a n d d ~ = \ / 1 + 4 ~ ~ + 4 ~ ~ d x d ~2n.$,~,/hii
d e s nr ~d ~r dd ~0==~ ~
g ( J J J / ~- 173/2).
AN=-3i-qj+k a n d d s = m d x d y . Thenarea=$:$,'adzdy=&.
6N=-
4-
xi - L + k a n d d S =
,/- ,/&.d x d Then area = so2 I~l l1-=
,~
/-
[-2*JT-7'i]:,, = fir.
~ N r= r-+Ik - and d S = ~ V z fi dx dy. Then area = J02n fab a r d r dB = fir(b2 - a2).
d dy =
1 0 N = -i - j + k and dS = 4 dx dy. Then surface area = times base area = 2 h .
z = d= gives N = + k and dS =
$;$. Then area = 4$'$d0)-" ,/-•
adxdy
16 On the sphere d S = sin 4 d4 d0 and g = x2 + y2 = sin2 4. Then so2 n son/2 sin3 4 d4 d0 = 2745) =
1 8 x = 2 cos v, y = 2 sin v, and d S = 2 d u dv. Then $$ g d S = so
3 '
lo2 cos v(2du dv) = 0.
2lr 3
20~=vi+j+k,~=ui+j-k,~=~x~=-2i+(u+v)j+(v-u)k,d~=,/4+2~~+2v~dudv.
22 ~ = c o s v i + s i n u j =
, ~- u s i n v i + u c o s v j + k , =~ sinvi- cosvj + u k , d ~= f i d u d v .
24 $$ F . n d S = 1:" $2fi -r3dr d0 = -242. 26 $$ F n d S = s $ O d S = 0.
15.5 The Divergeace Themem (page 588)
+ + + + +
28 F .n d S = ((u u)i - uvj). (-21 (u u)j (u - u)k)du du = (2u 2v - u2v - v 2 u ) d u dv.
So
Integrate with u = r cos 0, u = r sin B :
2% 1
+
(2r cos B 2r sin B - r3 cos2 B sin B - ? sin2 B cos B)rdrdB = 0.
SO A = cosBi+sin#j - 2rk,B = -rsinBi+ r c o s # j , N = A x B = 2?cosBi+2r2sinBj+rk,
2u a
$ $ k . n d ~ = / $ k . ~ d u d u = / ~I, r d r d B = r a 2 a s i n ~ x a m p l e 1 2 .
52 I think a 'triple Mijbius stripn is orientable.
+
34 The plane z = ax by has roof area = I,/- times base area. So choose for example a = 1 and b = a.
In words, the basic balance law is flow in = flow out. The flux of F through a closed surface S is the
+ + +
double integral I/ F .ndS. The divergence of M i Nj Pk is Mx + Ny Pg, and it measures t h e source
at t h e point. The total source is the triple integral /fl
div F dV. That equals the flux by the Divergence
Theorem.
For F = 5zk the divergence is 5. If V is a cube of side a then the triple integral equals 5aS. The top surface
where z = a has n = k and F n = 5a. The bottom and sides have F . n = gero. The integral // F n d S equals
5a3.
The field F = has div F = 0 except a t t h e origin. // F ndS equals 4 r over any surface around the
+
origin. This illustrates Gauss's Law: flux = 4 r times source strength. The field F = x i y j - 22 k has div
F = 0 and / ' F n d S = 0. For this F, the flux out through a pyramid and in through its base an equal.
+ +
The symbol V stands for (a/&)i (a/ay)j (a/az)k. In this notation div F is V . F. The gradient of f is
Vf. The divergence of grad f is V Vf o r v 2 f . The equation div grad f = 0 is Laplace's equation.
+
The divergence of a product is div(uV) = u div V (grad u ) V. Integration by parts in 3D is
$11 u div Vdx dy dz = - 1's +
V . grad u dx d y d~ /$ u V n dS. In two dimensions this becomes
+ + +l
$lu(aM/& aN/t3y)dx d y = - / ( M au/& N au/t3y)dx d y u V .n ds. In one dimension it becomes
integration by parts. For steady fluid flow the continuity equation is div pV = +/at.
2 /$F.ndS=$$/0dV =O
1 1 1
4 $$F.P~s=$,$ , $o(2z+2y+2z)dzdydz=1+l+l=~.
6 $1 F .ndS = (directly) $1 , 2 s s lo a 2 2
dS = 4ra2. By the Divergence Theorem: $ ;p sin 4 dp d4 de = 4ra2
8$$~.nds=$, I,
2s a a
.
12n b
$, 3 p 4 ~ i n 4 d p d 4 d t 9 = - p
1 0 div F = O + zegsinz - ze*sinz = 0 so $$F * n d=~0.
+
1 2 An integral over a box with small side a is near ca3. Here div F = 22 1has integral a' + a3, which is
near a3 because a is small. Then c = 1,which equals div F on the plane x = 0.
1 4 R - n = ( z i + g j + z k ) . i = z = 1 onone faceofthe box. On the fiveotherfaces R-n=2,3,0,0,0.
~hei~t~~rslis$~$~ld~dz+$~$~2dzdz+$~$,'~ 3 lds zo d i~v=R1. =
8l . + l + l = 3 and
$: # 3dx dy dz = 18.
16 The normal vectors to the cube are n = fi, fj,fk. Then $1 I,' 1;
F ndS = +
z dx dy
1,'$.(-z)dx dy I, + +
$', 'z dz dz I,' $,'(-z)dxdz + + I,'
$$,',' Ody dz $', ldy dz = 1.
Also $f/div FdY = $, So 1 1 1
$, ldz dy dz = 1.
18 grad f .n is the d i m t i o n a l derivative in the normal direction n (also written 2).
The Divergence Theorem gives $11 div (grad f ) d V = $$ grad f . n d S = $$ g d s .
But we are given that div (grad f) = + + j,, f, is sero.
20 Suppose F is perpendicular to n on the surface; then $$ F ndS = 0. Example on the unit sphere:
F is any q( z, y, z) times the spin field -yi xj+ .
22 The spin field F = -yi + z j has div F = 0 and F n = 0 on the unit sphere.
24 The flux of F = RIP3 through an area A on a sphere of radius p is Alp2, because I F I= 113and F
is outward. The spherical box has A/$ = sin 4d4 dB on both faces (minus sign for face pointing in).
No flow through sides of box perpendicular to F. So net flow = rero.
26 When the density p is constant (incompressible Bow), the continuity equation becomes div V = 0. If the Bow
is irrotational then F = grad f and the continuity equation is div (p grad f ) = -dp/dt.
If also p = constant, then div grad f = 0: Laplace's equation for the 'potential."
28 Extend E-F-G-H in Section 15.3 to 3 dimensions: E The total flux $$ F ndS through every closed
surface is rero F. Through all surfaces with the same boundary $IF ndS is the same
G There is a stream field g for which F = curl g H. The divergence of F is sero (this is the quick test).
SO The boundary of a solid ball is a sphere. A sphere has no boundary. Similarly for a cube or a cylinder - the
boundary is a closed surface and that s d a c e ' a boundary is empty. This is a crucial fact in topology.
+ + + +
The curl of Mi Nj Pk is the vector (Py- Ng )i (MI - Px)j (Nx - My)k. It equals the 3 by 3
i j k
+
determinant a/& slay a/& The curl of x2i z2k is xero. For S = yi - (z z)j yk the curl is + +
M N P
2i - 2k. This S is a spin field a x R = $( curl F ) x R, with axis vector a = i - k. For any gradient field
+ +
f.i frJ fsk the curl is sero. That is the important identity curl grad f = rero. It is based on,.f = f,,and
15.6 Stokes' Theorem and the Curl of F (page 595)
The curl measures the spin (or turning) of a vector field. A paddlewheel in the field with its axis along
n has turning speed f n . curl F. The spin is greatest when n is in the direction of curl F. Then the angular
1
velocity is ZI curl FI.
Stokes' Theorem is fCF . d R = JJS(curl F) n dS. The curve C is the boundary of the surface S. This is
Green's Theorem extended to three dimensions. Both aides are sero when F is a gradient field because t h e
curl is sero.
+
curl curl - grad div
0 -divgrad
O
= V21!!
I
f,v)k. This is f curl F = f ( v j - v,k) added to (grad f ) x F =
f,v j - &vk. Again the identity extends to any F.
+ +
P x G = (Np - Pn)i (Pm - Mp)j (Mn - Nm)k. Its divergence is the sum of x, y, and o derivatives:
1N.p + Np. - P.n - Pn.] + + + +
[P,m Pm, - Mvp - Mp,] [M,n Mn, - Nzm - Nm,]. Note that m
multiplies P, - Nz,the first component of curl F. This starts G *curl F - F curl G , as we want.
46 False. Certainly G x F would be perpendicular to F but V x F is something different. For example P = i yk +
h a s V x F = i s o ( V x F ) - F = 1.
15.6 Stokes' Theorem and the Curl of F (page 595)
48 S = roof, its shadow = ground floor, C = edge of roof, shadow of C = boundary of ground floor. Similarly
+ + ?.
for spherical cap r2 y2 z2 = 1 above z = Note C is on the plane z = and its shadow is a circle
around the shadow of the cap, down on the plane z = 0.
60 curl V = curl (-xk) = j. A wheel in the zz plane has n = j so it spins at full speed. A wheel perpendicular
to j will not spin, if it is in the r y plane with n = k.
16.1 Linear Algebra (page 602)
+
If Ax = b and AX = B, then A times 22 3X equals Zb+SB. If Ax = 0 and AX = 0 then A times 22 3X +
equals 0. In this case x and X are in the nullspace of A, and so is the combination 2x + 3X.The nullspace
contains a l l solutions to Ax = 0. It is a subspace, which means that when x and X are in the nullspace, eo
+
are all combinations ax bX. If x = (1,1,1) is in the nullspace then the columns add to the zero column,
so they are dependent.
Another subspace is the column space of A, containing all combinations of the columns of A. The system
Ax = b can be solved when b is in the column space. Otherwise the best solution comes from AX = ATb.
Here is the transpose matrix, whose rows are the columns of A. The nullspace of AT contains all solutions
to = 0. The column space of A= (row space of A) is the fourth fundamental subspace. Each subspace
has a basis containing as many independent vectors as possible. The number of vectors in the basis is the
dimension of the subspace.
When Ax = Ax the number X is an eigenvalue and x is an eigewector. The equation dy/dt = Ay has the
exponential solution y = eA' yo. A 7 by 7 matrix has seven eigenvalues, whose product is the determinant D.
If D is nonsero the matrix A has an inverse. Then Ax = b is solved by x = A-lb. The formula for D contains
-
7! = 5040 terms, so x is better computed by elimination. On the other hand Ax = Xx means that A X I has
determinant zero. The eigenvalue is computed before the eigenvector.
19detF=3
2
Z ~ F - ~ = + [ _ ~ -:]
'I
1-X 1
1-X 1 = (1 - - 3(1- A) + 2 = X3 - 3X2 = 0 if X = 3 or X = 0 (repeated)
1 1-X
1
L 4
12 The row space of A = is the line of multiples of (So is the column space.)
- :.
1 0
] = haa no solution because the first equations give x l = 1 and z2 = 1, which
[i] [i1
solution zl = $ , x 2 = i.
The vectors and a m combinations of columns of C.
Ax is a combmation of the co umns. If Ax = 0 happens with a nonsero z, the columns are 'dependent."
I[ ] [ ] [ 1]
If Ax = 0 only happens when x = 0, the columns are 'independent."
, , are three independent vectors in ID a they form a bml. The problem asks for
[ ::]
t ree vectors with positive components not lying in a common plane.
= haa determinant 25 - 16 = 9 which is a2 = (det F ) ~ .
1 :1 1 -:1 .
of F. General reasoning: If Fx = Xx then Fax = XFx = X2x. So x is also an eigenvector of F a .
26 Exponential solutions to 2 = Gy are yl = eQt and y2 = e' The combination f yl +
starts from yo = [$I+[-$]=[:I* L - J
Ex = 3 x is solved by x = 1i 1 L1J
or any multiple 11 41 .
SO A sero eigenvalue means there is a solution to Ax = OZ = 0. In this case A is not invertible and its
16.2 Different i d Equations (page 610)
+
The solution to y' - 5y = 10 is y = Ae6' B. The homogeneous part Ae5' satisfies y' - 5y = 0. The particular
+
solution B equals -2. The initial condition yo is matched by A = yo 2. For y' - 5y = ekt the right form is
y = ACS' + ~ e %For y' - s ~ = cost the form is y = Ae5' +tacos t + b s i n t.
+ +
The equation y" 4y' 5y = 0 is second-order because i t begins w i t h y". The pure exponential solutions
come from the roots of X2+4X+5 = 0, which are r = -1 and s = -4. The general solution is y = ~ e ~- e ~ +
- l ~ .
Changing 4y' to rero yields pure oscillation. Changing to 2y' yields X = -1f 2i, when the solutions become
y = ~ e eos +
- 2 t~ ~ e sin- Pt.~This oscillation is underdamped. A spring with rn = 1, d = 2, k = 5 goes
directly to zero. An electrical network with L = 1,R = 2, C = also goes t o rero (underdamped) .
One particular solution of y'' + 1 If the right side is cos t, the form of yp is s1c o s t. If
4y = e' is e' times 5.
the right side is 1then yp = f . If the right side is cos 2 t we have resonance and yp contains an extra factor t.
+ + +
2 -2a sin 2t 26 cos 2t a cos 2t b sin 2t = 4 sin 2t; collect sin 2t and cos 2t terms to find -2a +
6 = 4 and
6,
2b + a = 0; then b = a = -8, 8 + 4
y = -5 cos 2 t 5 sin 2t.
+ +
4 aet cost - aet sin t bet sin t betcost = 2et cost; collect terms to find a b = 2, -a+ +
b = 0; then
+
a = 1,b = l y = et(cos t sin t). This integral of 2et cost could be found by parts; here the correct form
was assumed at the start.
8 8
6 Ae-' solves the homogeneous equation yt + y = 0; then with A = the combination - % cos 2t $ sin 2t +
+
solves y' y = 4 sin 2t with y(0) = 0.
8 y = constant solves the homogeneous equation y' = 0, so a constant can be added to any particular solution
of y' = 2et cos t (the constant A is the same as the integration constant C!).
+
10 Substitute homogeneous solution plus particular solution y = Aet BeM into the equation; then kB - B = 1
and also A + B = 0 from yo = 0; thus B = -A = and y =
tekt
-= n.
et -t ekt Apply l9H6pital's rule
(the variable is k!) to find limk,l = limkdl i= tet.
+ + +
1 2 Homogeneous solution plus particular solution = Ae-' Bet C t D. Substitute to find 2Bet + C + Ct + D =
+ i,
et t and B = C = 1,D = -1; then y = ~ e #et + +
- ~ t - 1.(Initial value yo determines A.)
+ + +
1 4 Particular solution is B Dt; substitute to find D = B Dt t; then B = -1, D = -1; general solution
y=~et-t-l.
16 Substitution gives (A2 + 9)eAt = 0. (a) X = 3i or X = -3i (b) Pure oscillation because the X's are
+
pure imaginary (c) General solution y = Ae3" Be-3it = a cos St b s i n St. +
+ +
18 Substitution gives (A2 + 6X + 9)eAt = 0. (a) X2 6X + 9 = (A 3)2 and the equation (A + 3)2 = 0 has
the repeated solution X = -3 (b) The general solution is y = ~ +
e ~ - t e ~- ~ ~ .
20 k = 0 gives overdamping; k = 1 gives critical damping; k = 2 gives underdamping.
22 Substitute (te-')I' + 3(te-')' +2(teet) = (t - 2)e-' + 3(1- t)e-' +
2te-' = e-' # rero. So te-' is not a
homogeneous solution. (It would be if the coefficients in the equation changed from 1,3,2 to 1,2,1; then
+ +
X2 2X 1has X = -1 as a double root.)
+ + +
24 Substitute y- t o find -w2a cos wt - w2b sin wt - wa sin wt wb cos wt a cos wt b sin wt = sin wt. Then
+
(1- w2)a wb = 0 and -wa + (1 - w2)b = 1, which gives a =
w +( -w ) iW and b = 1-w2
W2+(lTW2)2
26 The graph of cos 5t - cos 3t will look like the figure following the exercises in Section 7.2, which shows
sin lox sin x = (cos 11%- cos 9%).
28 Substitute to find (iwL R &)AeiWt = veiWt. Then A = i W ~ +v~ + ( i w C ) - ' .
+ +
+
30 Substitute ypatticular = A sin t + B cos t to find -A sin t - B cos t 3A sin t + 3B cos t = sin t. Then
4
2A = 1 and B = 0 give y = sin t. (Note: the homogeneous solution involves cos a t and sin a t . )
+ + +
32 Substitute ypar,icu~ar= Aet cos t + Bet sin t to find -2Aet sin t 2Bet cos t 2Aet cos t 2Bet sin t = et cos t.
Then - 2 A + 2 B = O a n d 2 8 + 2 A = l a n d A = B = 1
g.
,kt -ect
3 4 Eguation (2) used l'H6pital's Rule for the limit of y = as k approached c. This exercise uses the series
e" = 1+ kt + i k 2t2 + . . and ect = 1+ ct + +cat2+ . to find - - kt-ct+ f k'tl- +c2t'+--. -
-
k-c - k-c
+
t + $(k + c)t2 + . . . As k approaches c this is t c t 2 + +
- = t ( l + c t . .) = tect.
+
3 6 Substitute yparticular= Act to find Aet Aet = et. Then A = and y = 2I et .
1
38 yo = . I gives A + B = l;yh = O gives - A - 4 B = 0 ; the solution has A = !,B= -5, and y = $e-'- i e - 4 t .
At t = 2 r the value is #e-2u - ife- 8" r;r 3
40 In exponential solutions AcM all derivatives are proportional to the function. For / = 6 8 substitute y = xn
to find n(n - l)xn-" 6x2". Comparing exponents gives n - 2 = 2n or n = -2. Then n(n - 1) = 6
and a solution is y = x - ~ (If
. the 6 were changed, this solution would become Ax-?)
44 Runge-Kutta is stable with h = .02 but 100 steps with h = .03 lead to yloo ry 670,000,000,000.
A graph is a set V of nodes or vertices and a set E of edges. With 6 nodes, a complete graph has 16
edges. A spanning tree has only 6 edges. A tree is defined as a graph with no loops, and it is spanning if it
contains all nodes. It has one path between each pair of nodes.
To find a path from node i to node j , two search methods are depth first search and breadth first
search. As nodes are reached, DFS looks out from the latest node for a new one. BFS looks out from the
earliest node. DFS must be prepared to backtrack to earlier nodes. In case of fire, BFS locates all doors from
the room you are in before it chooses one.
1Two then two then last one; go around hexagon 3 Six (each deletes one edge)
6 Connected: there is a path between any two nodes; connecting each new node requires an edge
13 Edge lengths 1,2,4
15 No; 1,3,4 on left connect only to 2,3 on right; 1,3 on right connect only to 2 on left 17 4
l9Yes 21F(mayloop);T 2616
4 A spanning tree omits one edge from each square. There are 4 x 4 choices of edges to omit: 16 spanning trees.
6 A loop is a sequence of edges from xo to XI, xl to x2, * , xh to xo : the loop (or circuit) is a closed path
on the graph. A loop invobing k nodes uses k edges. That leaves only 8 - k edges to connect the loop
to the remaining 9 - k nodes, which is impossible.
8 Breadth first search finds all neighbors, then all neighbors of neighbors, and so on. The farthest state
is found last.
10 Please send answer!
12 A matching is a set of edges, no two of which touch the same node.
14 The shortest path tree in network A contains the edges of length 4, 5, 1.
16 The loop in network B contains the edges of length 1, 4, 2, 3. Minimum spanning tree by Method 1, starting
top left: 7, 2, 3, 1, 5, 6, 8. Minimum spanning tree by Method 2 (increasing lengths, avoiding loops):
1, 2, 3, (avoid 4), 5, 6,7, 8.
18 If one node has 4 edges going out, then every other node has one of these edges coming in: a node with 0
edges cannot exist.
20 Suppose a spanning tree contains edge (8) but not (6). Then it is shorter if we substitute edge (6)
for edge (8). Also it is still a spanning tree: It contains the correct number of edges and no loops. Edge (6)
cannot be in a loop since edge (8) was removed; and those are the only edges into their common node.
22 (a) Starting from node s, a tree that connects each node to the next is perfect for depth first search. There
are no backtracks. (b) For breadth first search the perfect tree connects node s directly to each other
node. On that tree, depth first search would have to backtrack after each new node.
24 The code intends to find shortest distances dij between every pair of nodes (like a road atlas).
Correction: The loop on k should be the outer loop. Then k = 1 allows direct edges i to j and paths
from i to 1to j. Next k = 2 also allows paths from i to 2 to j and from i to 2 to 1to j
and from i to 1 to 2 to j. Eventually all intermediate nodes and all paths are allowed.
26 Maximum spanning tree: Method 1adds the longest edge that goes out from the current tree (start at any
node). Method 2 adds edges in decreasing order, longest first, rejecting any edge that closes a loop.
Notes on the Text: Classroom Experience
Chapter 1makes a start on three topics: functions in general, the sine and cosine in particular, and
computing. Here are brief comments so you will know what is needed later.
1. finctions and graphs (essential): Section 1.1is a solid starting point for calculus. You may give
+ +
it more than a day, especially if you introduce f (t 2) and f (t) 2 and f (2t) and 2f ( t ) .
These notes offer ideas about other graphing activities. One purpose is to maintain the interest
of those who have already taken calculus, without giving them an enormous advantage.
My favorite is the forward-back function graphed on page 4. Section 1.2 goes on to other piecewise
linear models like income tax - and says explicitly that "I hope you like them but you don't have
to learn them." Any of these examples, especially the delta function mentioned briefly, can be
passed over. It is Section 1.3 that compares average to instantaneous for y = z2.
2 . Sines and cosines (these are optional in Chapter 1): My intention is to see trigonometry in use -
for points on a circle as well as sides of a triangle. Many classes will not spend substantial time
on the review, but it must be available. The figure on page 31 leads neatly to cos(s - t).
Section 2.4 computes derivatives of sin x and cos z in the normal way from 2.
The limits of
and 9 are fully developed there. But students may understand these functions better (and
also the motion described by x = cost, y = sin t) by following a point on a circle. That is the
outstanding example of a parameter.
3. Computing in calculus (optional): The computing section is placed in a way that allows you to
discuss it or not. This topic is especially dependent on the local situation. (M.I.T. does not do
much computing in the first year, and does nothing with graphing calculators.) But calculators
are so convenient that we will see them more and more. They have the advantage of requiring
less faculty time, as well as being personal and portable and not too expensive.
The valuable thing is to see graphs (better than numbers). The example of 3= versus zs is
quite good - those graphs are surprisingly close for 2.2 < z < 3.2. It is a challenge to find their
intersection. It is a real challenge to find the only value of b for which bz never goes below zbfor
positive z. The main point is to see how this happens - the graphs of ez and zeare tangent at
z = e. In Section 6.2 we know the derivatives and verify eZ 2 ze.
Computing needs to be separate from the stream of ideas that launch calculus in Chapter 2.
'1000 Points of Lightn is purely for entertainment. See the College Mathematics Journal of
November 1990, and a forthcoming American Mathematical Monthly paper by Richert.
Graphing Activities
Ithaca College has developed a course that begins with a study of graphs and includes larger
projects throughout the course. They find that ''pqsing questions about graphs results in abundant
class participation and student response." I can confirm this from the Boston Workshop for Mathematics
Faculty.
The Ithaca materials will be published in the future - here are three examples suggested by and
adapted from their graphing activities. They can all be done in class.
EXAMPLE 1 These graphs describe flights from Los Angeles toward Washington:
velocity in mph
6OoL 0
. 1 1 1 1 1 & n e t
in hours
2 4 2 4 2 4 6
(a) Explain each flight in complete sentences. Does the third flight end in Washington or California?
(b) Try to give a formula for each velocity (in terms of time).
(c) Sketch a graph of the distance traveled versus time - same horizontal axis, distances on the vertical
axis (with numbers).
(d) Which graph is most realistic? Can you make it more realistic?
(e) If the first plane climbs slowly over bad weather in Chicago, how does that d e c t the graphs of
velocity and distance?
EXAMPLE 2 These graphs describe the distance traveled from New York toward Boston.
I distana in kilometers
hours
4
(a) Describe the five trips. For each graph, when is the distance increasing and when is it decreasing?
(c) Sketch the five velocity graphs. Which (if any) are realistic?
*(e) Draw the distance graph for travel from your room to this class.
*(f) Draw the height graph for raising a flag hand over hand.
0 Here is a word problem that uses the connection between v = 8t and f = 4t2, and then finds
roots of a quadratic. It could be discussed after Section 1.3 or with tangent lines.
Problem: A car goes 240 feet in 8 seconds. Up to time tl it has u = 8t (constant acceleration);
after that time it has v = 8tl (constant velocity). Find tl, by computing the distance f (t) and solving
f (8) = 240. Draw the graphs of f (t) and u(t).
+
Solution: 4t: 8t1(8 - tl) = 240 yields t l = 6. The graph of u(t) is piecewise linear. The graph of
f (t) shows a parabola changing into its tangent line at t = 6.
EXAMPLE 3 Students start lining up for registration at 8:00 at the rate of 200 students per hour.
From 12 noon the rate decreases steadily until 2:00, when 1,100 students have arrived and the rate
stabilizes at 100 per hour. At 9:00,250 students per hour are allowed to enter. At 1:00 the gym closes
for a faculty lunch. After 2:00 the entering rate is 150 students per hour. At 4:00 the last student
arrives and enters.
(a) Draw the graph of the arrival rate. The horizontal axis gives the time
(between 8:00 a.m. and 4:00 p.m.).
(b) Draw the graph of the rate at which students enter the gym vs. time between 8:00 and 4:OO.
(c) Graph the total number of students who have arrived at the gym vs. time between 8:00 and 4:00.
(d) Graph the number of students who have entered the gym vs. time between 8:00 and 4:OO.
(e) Give a rule for the number of students who have entered as a function of elapsed time since 8:00.
I am convinced by the experience of others: It is good for students to work in small groups. There
are so many reports about the success of this idea that it has to be accepted as valuable. It will be
implemented in different ways, and the comments from Ithaca about group projects are representative:
"The approach changed the students' attitudes toward mathematics. The projects engage the
curiosity of the good students and challenge them, but this does not come at the expense of average
and weaker students. In fact, cooperative work with good, motivated students bolsters the others.'
"A common fear about groups is that one student may do nothing but still get the same grade
aa the members who did all the work. This has not been a great problem .. . (others say the same).
Students experience cooperative learning. They talk to each other about mathematical ideas and they
form friendships with other mathematics students."
I personally believe that we too often lose sight of the human part of learning mathematics.
Section 2.2: My class resists the binomial theorem but likes Pascal's triangle. This inconsistency
suggests that theorems are not as attractive as patterns! I made a transparency of Pascal's
+
triangle down to (1 I)'', and colored all the even coefficients. They fall into neat triangles
which the class can more or less explain.
Here is a !kuaFalse question that looks too good to be true. The derivative of zn at z = n ia nn.
These topics deserve more attention than in the past. The iteration zn+l = F(zn) is a simple
idea and it is the basis of scientific computing. Its simplicity means that a numerical experiment like
zn+l = cosz, is instantly available. On the TI-81 no program is necessary: use .7 ENTER and on a
new line cos Ans. Now each press of ENTER gives the next x,+l as the cosine of the previous Answer.
Here is the output, six steps at a time:
The limit of zn+l = cos zn is the underlying equation x = cos x. Linear convergence is apparent from
this triangle of numbers. At the solution = - sin x M -.68, so every six steps multiply the error by
about (-.68)6 (J &. This explains why one new decimal is correct after six steps. Compare with the
quadratic convergence of Newton's method for x - cos x = 0 :
Thus two Newton steps match thirty steps of simple iteration. But Newton requires the derivative -
in fact 100 partial derivatives for 10 equations in 10 unknowns. This leaves s place for iteration.
What to cover in a calculus course? Certainly we must teach Newton's method, which uses the
central idea of linear approximation and the tangent line. I believe that simple iteration should also be
included, because its convergence is a perfect calculus problem: z,+l - x = F(z,) - F(x). Approximate
the right side by times x, - x. The error is multiplied at every step by g. But where is this
derivative evaluated? The Mean Value Theorem says between x, and z. Linear approximation says at
the solution z.Convergence is guaranteed if I I < 1 everywhere.
Some new courses compare the iterations
Those discrete systems match the key differential equations in this book: 2
= cy and 2
= cy e and +
2 = ey - by2. The logistic model 3 describes a population in successive years, and no derivative is
involved. The solutions can approach a steady state or a cycle or chaos.
+
Note: The text introduces chaos through Newton's method for x2 1 = 0. The iteration is easy
1
to try: xn+l = I(xn - &). Most classes will touch chaos and fractals very lightly - that is absolutely
normal, since there is so much calculus to do. But the book must offer the possibility to read something
that Newton and Leibniz and even Riemann did not know. I will be grateful for any comments on your
experience, and also on the last exercises in Sections 3.6 and 3.7.
Problem 3.7.48 starts with y(x) = 3.4(x - x2) and asks for graphs of y(y(x)) and y(y(y(x))).
Those are polynomials of degree 4 and 8. The tenth iterate has degree 21° = 1024, but this polynomial
is easy to graph (by computer). So is the hundredth iterate. Maybe a polynomial of degree 21°0 h a
never been seen before! Its graph is very surprising at first:
At each x the graph shows xloo, the result of 100 iterations. What do we expect from one more iter-
ation? The graph above will flip over for the polynomial of degree 21°1, which sends .452 to 3 4 2 and
vice versa. The function cos(cos(cos would have a horizontal graph at height z =.7390851.
It is unusual to have a full chapter with this title. The chain rule is generally squeezed in earlier,
with a quick explanation of f (g(z)). But this is an idea of major importance. It is how new functions
are created. The four sections of this chapter (including inverse functions and implicit differentiation
and related rates) make a natural development from f (g(z))and the chain rule.
Two special cases we see earlier. One is the power rule un-'2. The other allows examples like
sin2z, when 2 = 2 is a constant. I would call attention to those cases when the general chain rule is
presented. And I really feel that2 =es is so clear and simple that a much longer argument could
only obscure the key idea.
Inverse functions also deserve serious discussion (and exercises). Draw graphs to see f and to
explain why f must be one-to-one. This section is preparing for y = ez and x = In y. Also for y = sin z
and z = sin-' y which follow immediately.
To me the correct notation is z = f -'(y). The inverse of y = z3 is z = y1I3. We should not switch
z and y! Don't say that 'the inverse of the function y = z3 is the function y = x'/~."
The inverse of f (g(z)) is a basic idea, to reverse to g-l(f -I(%)). The student can get this.
+ +
The strips are trapezoids. Their areas are f (zl - 22) (yl y2) and f (zl x2)(m - yl ) . The average
is f (zlM - z2y1). This formula for the area of a triangle appears in Chapter 11, as half of a vector
parallelogram or a cross product or a 2 by 2 determinant. So we have an algebraic proof, when
the graph is a straight line. We need a geometrical proof that still applies when the function is
not linear and its graph between (zl, yl) and (z2,y2) is curved.
Geometrical proof: Draw the extra line to (z2,yl) in the second figure. The shaded rectangle and
triangle have the same height yl and the same base. The triangular area is half the rectangular
area. The same argument applies above the extra line (rectangle at the top left, triangle down to
the origin). That leaves the remaining piece which is above yl and beyond x2. This piece belongs
to both of the strips and also to the triangle, whether the graph i s curved or straight. Then adding
the (curved) trapezoids gives twice the area of the (curved) triangle.
Any class discussion can wait until you have computed area in polar coordinates (Chapter 9).
There is also a ucalculus proof" that the triangle area is the average of the strip areas. For
infinitesimal pieces the three areas are z dy and -y dz and f (z dy - y dz). The minus sign
is because a line integral goes counterclockwise from (zl ,yl) to (z2,y2), so dx is negative. Then
the three line integrals, around any region, give the area. This is the first application of Green's
Theorem (page 564), but it is not a surprise - areas are always found from strips. What was a
surprise was the realization that f (z dy - y dz) is the same as f r2 dB. By substituting z = r cos 0
and y = r sin 8 this is confirmed.
Numerical Integration
This section emphasizes an important fact about numerical integration: the error is proportional to
(Az)P. The order of accuracy p comes from a simple test: apply the integration rule to the power func-
tions 1, z, z2, *. The first power to be wrong is p. The rule fails to compute :/ zpdz. The trapezoidal
rule has p = 2 and Simpson has p = 4. New rules are constructed so as to integrate powers correctly.
This error term (Az)P is what all numerical analysts use - not the exact expression which involves
+
the (p 1)st derivative of y(z). I do not discuss the exact error integrals at all - life is too short.
0 Question: What i s the error exponent p for Riemann sums? The answer depends on the
evaluation points zf. We know that p = 2 for midpoints and p = 1 for left endpoints. Also p = 2
for alternating left and right endpoints! To make the question sensible for any Riemann sum,
why not pick the points zf at random?
This is now a computing problem (and also a probability problem). The student can use the random
number generator - an excellent tool for numerical experiments. Tky l,'z dz with intervals A z = &
and &;
repeat 100 times with random z f . I believe the expected error is proportional to ( ~ z ) ' ? The
appearance of a fraction p = is unusual to say the least.
Differential Equations
A significant part of "calculus reform" is to bring out more clearly the purpose of calculus. That
leads to an earlier appearance of differential equations. In the simplest case (Section 2.2) we are just
reversing the derivative, as in $ = 32'. This is important (it motivates integration). It is also easy
to discuss. We cannot let students finish calculus, as many do, without hearing the words "differential
equation."
In the second semester we can do much more. Here is an outline for three specific first-order
equations. The applications are in the book along with second-order equations (those use complex
numbers). The key equations are
1 gives exponential growth or decay: y = meCt.Equation 2 ha8 a constant source term. The logistic
equation 3 has competition from the nonlinear mass action term - this is basic in the life aeiences.
With the benefit of hindsight, 2 and 3 can be rewritten to show pure exponential growth:
Notice how all these solutions approach steady states ym when c < 0. The exponential ed decays
to zero. The differential equation approaches $ = 0 (no movement):
It may be useful to collect together the methods of solution, especially for the equation $ = cy+s :
A. Assume y = ~ e " B + (undetermined coefficients)
B. Multiply the equation by (integrating factor)
C. Add up outputs from each input: meCt+ e ~ ( ~ - ~ ) s d r (Duhamel)
The logistic equation is also solved by separation of variables and integration of ,hl'.
This is a good
contribution from partial fractions (which are quite useful for quadratics - we probably overemphasize
!
the high degree cases). Note that z = converts the logistic equation to = -cz b 9 +
Notes on Chapter 7
7.1 Integration by Parts: Most likely it will take all your time to practice integration by parts
(this is a technique of integration that deserves to survive). If no time is left for the delta function
or the application to engineering, that is no problem.
The exercises include n integrations by parts, leading to Taylor's formula with integral remainder
term. Here is Problem 19with a clever choice v = : / x tan-' z dz = tan-' z - ) $1 dx.
0 7.2 Trigonometric Integrals: Example 3 is the integral of cos2z. Show why the substitution
u = cos x does not succeed. Then double angle formulas make it easy (alsocos4z and cos2x sin2z).
One integral that is needed later is / see3 x dx.
Maybe the substitution u = tan f z should not come Youtof the blue". Draw a unit circle with
a typical point (cosx, sin z) in the first quadrant. Then draw the chord to that point from the
left-hand point (- 1,O). By geometry, the chord makes an angle of with the horizontal. The
figure shows that tan f = -The chord crosses the vertical axis at (0, u). To be truthful, I
don't teach this substitution.
7.3 Trigonometric Substitutions: Usually the integral of t / D is pure manipulation. The
student must Bee this as area under a circle. Figure 7.5a shows why the answer has two parts
Z'zd1-22
and sin-'%. Notice the paradox for a complete circle in Problem 53, when the B
integral goes from zero to 2x.
7.4 Partial Fractions: The application in the text is to the integration of rational functions.
But partial fractions can also improve the derivative (Hendel, College Math Journal 11/91). The
fourth derivative of zf+sz,+2 looks unpleasant - no pattern is visible. Computer algebra can give
a formula for any denvatwe, but it won't be as neat as this:
Infinite Series
The geometric aeries is absolutely essential. It is easy to get involved with convergence tests and
lose that explicit formula. I think some numbers are good in Section 10.1. It might be interesting to
take a few steps of the Gauss algorithm on page 372 that computes r .
Apologies for Problem 68 on page 385, which asks for a series whose convergence you cannot decide.
Maybe the student gets a chance to be creative. I wouldn't go overboard with the remainder terms in
the Taylor series (page 392). But the Taylor series itself is the best connection between finite series
and the mainstream of calculus.
This chapter is an interesting challenge - to present just enough of linear algebra, even as far as
A-I and least squares, but not too much. I try to give explanations in R2 and RS that remain useful
in Rn. The row picture and column picture of a linear system are both important. Do not forget that
a matrix-vector multiplication Av is a combination of the columns of A.
Before that come dot products and cross products. Dot products are for everybody - and the
formula for c o d is generally well received. Cross products are less fun, but formula (4) for A-I in
Section 11.5 is neat. So is the area of a triangle - neater than I realized:
0 Area of a triangle T The 3 by 3 determinant is in Problem 11.3.44:
This is half of three crow products. Therefore three triangular areas. One triangle connects (0,O)
to (xl, yl) and (x2,M). Two more triangles go from (0,O) to the other two edgea of T. If (0,O) is
inside T, the three small triangles add to the complete triangle. But the sum gives the correct
area even when (0,O) is outside T - get the class to see why.
Notice that the sum does not depend on the position of the origin. Adding a constant to XI, x2, q
leaves the sum unchanged. So does adding a constant to yl, yr, ys. Drawing a picture is better,
with (0,O) outside and areas canceling to leave the area of T.
0 Area of a polygon P Barry Cipra actually met two carpet-layers in a Minnesota coffee shop,
who wanted a formula for the area of a carpet. Inspired by caffeine he proposed
It's right, and a good student can see why. Divide the polygon into triangles, by connecting each
edge to the origin. My experience in this direction was with a lawyer, who wanted the azea of a
property and not a carpet. 'If I count steps walking around the boundary, what is the area?" He
was very disappointed when I needed to know the shape. Now I would tell him the carpeblayer
formula.
I hope you don't mind saving the equation of a line for Section 12.1, which follows immediately.
If you don't want projectiles and curvature and Kepler's Laws, you can stop after 12.1. I like the
last law with the power.
Partial Derivatives
Students seem to get the knack of computing 2 and treating y as constant. The idea of the
gradient is not ao easy. I keep drawing surfaces to teach this chapter, with tangent planes and normal
vectors. In explaining the analysis, the summary on page 471 (multivariable vs. single variable) goes a
long way. The study of continuity for f (2, y) can get technical quickly - perhaps it is more important
to understand the geometry.
One high point of the chapter is to maximize by solving = 0 and = 0. Again, draw the
tangent plane. The Steiner problem gives an example with some substance, but it could be omitted.
There is a lot to do in Section 13.6. Then 13.7 with constraints and Lagrange multipliers is certainly
optional (but important - few calculus texts try to explain what X represents!)
Caspar Curjel has a book of multivariable exercises (McGraw-Hill) based on figures like those on
page 496. In one variable we easily connect f (z) and & to graphs and slopes. It is not so simple for
two variables (or three). Should the n-dimensional case be developed in this chapter?
Vector Fields and Green's Theorem
These topics generally come late in multivariable calculus. Many times they don't come at all -
for an engineering or physics major they are important, for a biologist or an economist less so. This
note is to call attention to the treatment of one specific question: Which pairs of functions M(z, y)
and N(z, y) are the partial derivatives of a single function, M = and N = g?
In this case the vector field Mi+ N j is the gradient of f (z, y). The field is conservative. The integral
+
I Mdz Ndy is $ df = change in f . The differential is exact. This is a special and highly important
situation (not only in physics), and the test is simple:
9'
Added in proof (College Mathematics Journal, May 1991): Find the sloping asymptote of
41 The example is better with -10 < z < 30 instead of -1 < z < 10
61 .Near z = 1, the distance up is about 9 times the distance across."
104 Problem 52 should maximize not minimize
206 - v to M in Problem 28
Change
258 +
Problem 46 is ln(z d m ) = ...
265 Change to c = bmK in line -3
267 Change to y/(e - by) in Problem 18
273 Change .05n to .05/n in 5 and 6
280 Remove f in Problem 5
310 GM = 4 10" in Problem 34 (otherwise it's a small world)
359 The last read-through question is for '
j ry2dz (not ds)
1 3
402 The figure shows w = [ ] not [ ]
411 Example 8 Find the nearest point to the origin on the plane
z+2y+2z=5
429 Equation (8) gives A-'d not A%
444 Change BC to CB in Problem 20
520 Problem 32: Explain why As > 0 and X4 > 0 and f- = 2
526 Change the second part of Problem 3 to ( +
dy dz/(z y)'
540 In Problem 13 find the volume below z = 5
In Problem 15 find the volume below the cone 4
2
- +z = 1.