History of Mathematics Brief Version Victor Katz

Name:Stanislav Shur
Email:myidealist@gmail.com
IP:74.88.120.168
SOLUTIONS TO PROBLEMS
CHAPTER ONE
1. We write 275 as follows in Egyptian hieroglyphics (on the left) and Babylonian cuneiform
(on the right):
2.
1
0 10
50
(multiply by 10)
10
(double first line)
20
(double third line)
08
40
(double fourth line)
02
22
(halve first line)
0 10
(invert third line)
18 2 10
93
3.
1
7248
15 2 4
04
31 2
08
63
03
4 3 3 6 12
12 3
98 2 3 3 6 12
99 2 4
Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214
Name:Stanislav Shur
4.
2 11
11
IP:74.88.120.168
2 23
23
73
15 3
33
73
136
323
66
12
1246
276
12
12 276
6 66
5.
5 13 = (2 13) + (3 13) = 8 52 104 + 8 13 52 104 = 4 13 26 52
6 13 = 2(3 13) = 4 8 52 104 26 52 = 4 8 13 104
8 13 = 2(4 13) = 2 13 26
6. x + 17 x = 19. Choose x = 7; then 7 +
2 38 7 = 16 58 .
1
7
7 = 8. Since 19 8 = 2 38 , the correct answer is
7. (x + 23 x) 13 (x + 23 x) = 10. In this case, the obvious choice for x is x = 9. Then 9

added to 2/3 of itself is 15, while 1/3 of 15 is 5. When you subtract 5 from 15, you get
10. So in this case our guess is correct.
8. The equation here is (1 + 13 + 14 )x = 2. Therefore. we can find the solution by dividing
2 by 1 + 13 + 14 . We set up that problem:
1
124
1 18
2 36
4 72
12
8 144
The sum of the numbers in the right hand column beneath the initial line is 1 141
144 . So we
3
need to find multipliers giving us 144 = 144 72. But 1 3 4 times 144 is 228. It follows
that multiplying 1 3 4 by 228 gives 144 and multiplying by 114 gives 72. Thus, the
answer is 1 6 12 114 228.
2
Name:Stanislav Shur
IP:74.88.120.168
9. Since x must satisfy 100 : 10 = x : 45, we would get that x =

this up into a sum of two parts, 35100
and 10100
10
10 .
45100
10 ;
the scribe breaks
10. The ratio of the cross section area of a log of 5 handbreadths in diameter to one of 4
9
handbreadths diameter is 52 : 42 = 25 : 16 = 1 16
. Thus, 100 logs of 5 handbreadths
9
1
diameter are equivalent to 1 16 100 = 156 4 logs of 4 handbreadths diameter.
12.
8
7) 1 00
56
4
3
34 17 8
00 00 00 00
00
58
2 00
1 59
1 00
56
4
13. Since 3 18 = 54, which is 6 less than 60, it follows that the reciprocal of 18 is 3 13 , or,
putting this in sexagesimal notation, 3,20. Since 60 is (1 78 ) 32, and 78 can be expressed
as 52,30, the reciprocal of 32 is 1,52,30. Since 60 = 1 19 54, and 19 can be expressed as
1
1
6
40
15
10 + 90 = 60 + 3600 = 0; 06, 40, the reciprocal of 54 is 1, 06, 40. Also, because 60 = 16 64,
1
15
the reciprocal of 64 is 15
16 . Since 16 = 3, 45, we get that 16 = 56, 15. If the only prime
divisors of n are 2, 3, 5, then n is a regular sexagesimal.
14. 25 1, 04 = 1, 40 + 25, 00 = 26, 40. 18 1, 21 = 6, 18 + 18, 00 = 24, 18. 50 18 =
50 0; 3, 20 = 2; 30 + 0; 16, 40 = 2; 46, 40. 1, 21 32 = 1, 21 0; 01, 52, 30 = 1; 21 +
1; 10, 12 + 0; 00, 40, 30 = 2; 31, 52, 30.
15. Since the length of the circumference C is given by C = 4a, and because
C = 6r, it follows
17
that r = 23 a. The length T of the long transversal is then T = r 2 = ( 23 a)( 17
12 ) = 18 a.
7
The length t of the short transversal is t = 2(r 2t ) = 2a( 23 17
36 ) = 18 a. The area A of
the barge is twice the difference between the area of a quarter circle and the area of the
right triangle formed by the long transversal and two perpendicular radii drawn from
the two ends of that line. Thus
C 2 r2
A=2
48
2
a2 2a2
=2
3
9
2
= a2 .
9
16. Since the length of the circumference C is given by C = 3a, and because
C = 6r, it follows
that r = a2 . The length T of the long transversal is then T = r 3 = ( a2 )( 74 ) = 78 a. The

length t of the short transversal is twice the distance from the midpoint of the arc to
3
Name:Stanislav Shur
IP:74.88.120.168
the center of the long transversal. If we set up ourcircle so that it is centered on the
origin, the midpoint of the arc
has coordinates ( 2r , 23r ) while the midpoint of the long
transversal has coordinates ( 4r , 43r ). Thus the length of half of the short transversal is
r
a
2 and then t = r = 2 . The area A of the bulls eye is twice the difference between the
area of a third of a circle and the area of the triangle formed by the long transversal and
radii drawn from the two ends of that line. Thus
C2 1 r
T
A=2
36
22
9a2 1 a 7a
=2
36
24 8
= 2a2
7
1
9
= a2 .
4 64
32
17. The correct formula in the first case gives V = 56, while the Babylonian version gives
V = 12 (22 +42 )6 = 60 for a percentage error of 7%. In the second case, the correct formula
gives 488/3 = 162 23 , while the Babylonian formula gives V = 12 (82 + 102 )2 = 164, for an
error of 0.8%.
51
+ 10 = 1+0.4+0.0141666666+0.0000462962 = 1.414212963.
18. 1; 24, 51, 10 = 1+ 24
60 +
3600 216000
On the other hand, 2 = 1.414213562. Thus the Babylonian value differs from the true
value by approximately 0.00004%.
19. Because (1; 25)2 = 2; 00, 25, we have
2=
2; 00, 25 0; 00, 25 1; 25 (0; 30)(0; 00, 25)(1/1; 25).
An approximation to the reciprocal of 1;25 is 0;42,21,11. The

product of 0;30 by 0;00,25
by 0;42,21,11 is 0;00,08,49,25. The the approximation to 2 is 1; 25 0; 00, 08, 49, 25 =
1; 24, 51, 10, 35, which, with the last term truncated, is the Babylonian value.
20.
3 =
22 1 2
1
2
1
2
= 2 0; 15 = 1; 45. q
Since an approximate recipro
cal of 1;45 is 0;34,17.09, we get further that 3 = (1; 45)2 0; 03, 45 = 1; 45
(0; 30)(0; 03, 45)(0; 34, 17.09) = 1; 45 0; 01, 04, 17, 09 = 1; 43, 55, 42, 51, which we truncate to 1;43,55,42 because we know this value is a slight over-approximation.
129 2
2
21. 12 3 15 24 32 = 12 129
160 . (12 160 ) = (12.80625) = 164.0000391 . . .
22. v + u = 1; 48 = 1 45 and v u = 0; 33, 20 = 59 . So 2v = 2; 21, 20 and v = 1; 10, 40 = 106

90 .
56
Similarly, 2u = 1; 14, 40 and u = 0; 37, 20 = 90 . Multiplying by 90 gives x = 56, d = 106.
1
In the second part, v + u = 2; 05 = 2 12
and v u = 0; 28, 48 = 12
25 . So 2v = 2; 33, 48
769
and v = 1; 16, 54 = 600 . Similarly, 2u = 1; 36, 12 and u = 0; 48, 06 = 481
600 . Multiplying by
481
319
600 gives x = 481, d = 769. Next, if v = 360 and u = 360 , then v + u = 2 29 = 2; 13, 20.
161
7
Finally, if v = 289
240 and u = 240 , then v + u = 1 8 = 1; 52, 30.
4
Name:Stanislav Shur
IP:74.88.120.168
98569
23. The equations for u and v can be solved to give v = 1; 22, 08, 27 = 295707
216000 = 72000 and
67319
u = 0; 56, 05, 57 = 201957
216000 = 72000 . Thus the associated Pythagorean triple is 67319,
72000, 98569.
24. The two equations are x2 + y 2 = 1525; y = 23 x + 5. If we substitute the second equation
into the first and simplify, we get 13x2 + 60x = 13500. The solution is then x = 30,
y = 25.
25. If
we guess that the length of the rectangle is 60, then the width is 45 and the diagonal
is 602 + 452 = 75. Since this value is 1 78 times the given value of 40, the correct length
of the rectangle should be 60 1 78 = 32. Then the width is 24.
26. One way to solve this is to let x and x 600 be the areas of the two fields. Then the
equation is 23 x + 12 (x 600) = 1100. This reduces to 76 x = 1400, so x = 1200. The second
field then has area 600.
1
27. Let x be the weight of the stone. The equation to solve is then x 17 x 13
(x 17 x) = 60.
We do this using false position twice. First, set y = x 17 x. The equation in y is then
1
1
y 13
y = 60. We guess y = 13. Since 13 13
13 = 12, instead of 60, we multiply our
1
guess by 5 to get y = 65. We then solve x 7 x = 65. Here we guess x = 7 and calculate
1
the value of the left side as 6. To get 65, we need to multiply our guess by 65
6 = 10 6 . So
5
5
our answer is x = 7 65
6 = 75 6 gin, or 1 mina 15 6 gin.
1
(x 17 x).
28. We do this in three steps, each using false position. First, set z = x 17 x + 11
1
The equation for z is then z 13
z = 60. We guess 13 for z and calculate the value of
the left side to be 12, instead of 60. Thus we must multiply our original guess by 5 and
1
y = 65. If we now guess
put z = 65. Then set y = x 17 x. The equation for y is y + 11
y = 11, the result on the left side is 12, instead of 65. So we must multiply our guess
715
7
1
7
by 65
12 to get y = 12 = 59 12 . We now solve x 7 x = 59 12 . If we guess x = 7, the
7
left side becomes 6 instead of 59 12
. So to get the correct value, we must multiply 7 by
715
715
715
5005
37
37
12 /6 = 72 . Therefore, x = 7 72 = 72 = 69 72 gin = 1 mina 9 72 gin.
29.
Name:Stanislav Shur
IP:74.88.120.168
30. If we substitute the first equation into the second, the result is 30y (30 y)2 = 500 or
y 2 + 1400 = 90y. This equation has the two positive roots 20 and 70. If we subtract the
second equation from the square of the first equation, we get (x2 = 900)(xy(xy)2 =
500), or (x y)2 + x(x y) = 400, or finally (x y)2 + 30(x y) = 400. This latter
equation has x y = 10 as its only positive solution. Since we know that x = 30, we
also get that y = 20.
31. The equations can be rewritten in the form x + y = 5 56 ; x + y + xy = 14. By subtracting
the first equation from the second, we get the new equation xy = 8 16 . The standard
method then gives
x=
5 56
2
v
u
u
t
5 56
2
!2
11
1
8 =2 +
6
2
1
11
73
8 =2 =
8
144
6
12
11
7
1
49
=2 +
=3 .
144
12 12
2
Similarly, y = 2 13 .
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER TWO
1. Since AB = BC; since the two angles at B are equal; and since the angles at A and C are
both right angles, it follows by the angle-side-angle theorem that 4EBC is congruent
to 4SBA and therefore that SA = EC.
2. Because both angles at E are right angles; because AE is common to the two triangles;
and because the two angles CAE are equal to one another, it follows by the angle-sideangle theorem that 4AET is congruent to 4AES. Therefore SE = ET .
3. Tn = 1 + 2 + + n =
triangular number Tn .
4. n2 =
5.
(n1)n
2
8n(n+1)
2
n(n+1)
,
2
n(n+1)
.
2
Therefore the oblong number n(n + 1) is double the
and the summands are the triangular numbers Tn1 and Tn .
+ 1 = 4n2 + 4n + 1 = (2n + 1)2 .
6. Examples using the first formula are (3,4,5), (5,12,13), (7,24,25), (9,40,41), (11,60,61).
Examples using the second formula are (8,15,17), (12,35,37), (16,63,65), (20,99,101),
(24,143,145).
7. Consider the right triangle ABC where AB has unit length and the hypotenuse BC
has length 2. Then the square on the leg AC is three times the square on the leg AB.
Assume the legs AB and AC are commensurable, so that each is represented by the
number of times it is measured by their greatest common measure, and assume further
that these numbers are relatively prime, for otherwise there would be a larger common
measure. Thus the squares on AC and AB are represented by square numbers, where
the former is three times the latter. It follows that leg AC is divisible by three and
therefore its square is divisible by nine. Since the square on AB is one third that on AC,
it is divisible by three, and hence the side AB itself is divisible by three, contradicting
the assumption that the numbers measuring the two legs are relatively prime.
9. Let ABC be the given triangle. Extend BC to D and draw CE parallel to AB. By I29,
angles BAC and ACE are equal, as are angles ABC and ECD. Therefore angle ACD
equals the sum of the angles ABC and BAC. If we add angle ACB to each of these, we
get that the sum of the three interior angles of the triangle is equal to the straight angle
BCD. Because this latter angle equals two right angles, the theorem is proved.
10. Place the given rectangle BEF G so that BE is in a straight line with AB. Extend F G to
H so that AH is parallel to BG. Connect HB and extend it until it meets the extension
of F E at D. Through D draw DL parallel to F H and extend GB and HA so they meet
DL in M and L respectively. Then HD is the diagonal of the rectangle F DLH and
7
Name:Stanislav Shur
IP:74.88.120.168
so divides it into two equal triangles HF D and HLD. Because triangle BED is equal
to triangle BMD and also triangle BGH is equal to triangle BAH, it follows that the
remainders, namely rectangles BEF G and ABML are equal. Thus ABML has been
applied to AB and is equal to the given rectangle BEF G.
11. In this proof, we shall refer to certain propositions in Euclids Book I, all of which are
proved before Euclid first uses Postulate 5. (That occurs in proposition 29.) First,
assume Playfairs axiom. Suppose line t crosses lines m and l and that the sum of the
two interior angles (angles 1 and 2 in the diagram) is less than two right angles. We
know that the sum of angles 1 and 3 is equal to two right angles. Therefore 6 2 < 6 3.
Now on line BB 0 and point B 0 construct line B 0 C 0 such that 6 C 0 B 0 B = 6 3 (Proposition
23). Therefore, line B 0 C 0 is parallel to line l (Proposition 27). Therefore, by Playfairs
axiom, line m is not parallel to line l. It therefore meets l. We must show that the two
lines meet on the same side as C 0 . If the meeting point A is on the opposite side, then
6 2 is an exterior angle to triangle ABB 0 , yet it is smaller than 6 3, one of the interior
angles, contradicting proposition 16. We have therefore derived Euclids postulate 5.
Second, assume Euclids postulate 5. Let l be a given line and P a point outside the line.
Construct the line t perpendicular to l through P (Proposition 12). Next, construct the
line m perpendicular to line t at P (Proposition 11). Since the alternate interior angles
formed by line t crossing lines m and l are both right and therefore are equal, it follows
from Proposition 27 that m is parallel to l. Now suppose n is any other line through P .
We will show that n meets l and is therefore not parallel to l. Let 6 1 be the acute angle
that n makes with t. Then the sum of angle 1 and angle P QR is less than two right
angles. By postulate 5, the lines meet.
Note that in this proof, we have actually proved the equivalence of Euclids Postulate
5 to the statement that given a line l and a point P not on l, there is at most one line
through P which is parallel to l. The other part of Playfairs Axiom was proved (in the
8
Name:Stanislav Shur
IP:74.88.120.168
second part above) without use of postulate 5 and was not used at all in the first part.
12. One possibility: If the line has length a and is cut at a point with coordinate x, then
4ax + (a x)2 = (a + x)2 . This is a valid identity.
13. In the circle ABC, let the angle BEC be an angle at the center and the angle BAC be
an angle at the circumference which cuts off the same arc BC. Connect AE and continue
the line to F . Since EA = EB, 6 EAB = 6 EBA. Since 6 BEF equals the sum of those
two angles, 6 BEF is double 6 EAB. Similarly, 6 F EC is double 6 EAC. Therefore the
entire 6 BEC is double the entire 6 BAC. Note that this argument holds as long as line
EF is within 6 BEC. If it is not, an analogous argument by subtraction holds.
14. Let 6 BAC be an angle cutting off the diameter BC of the circle. Connect A to the
center E of the circle. Since EB = EA, it follows that 6 EBA = 6 EAB. Similarly,
6 ECA = 6 EAC. Therefore the sum of 6 EBA and 6 ECA is equal to 6 BAC. But the
sum of all three angles equals two right angles. Therefore, twice 6 BAC is equal to two
right angles, and angle BAC is itself a right angle.
15. In the circle, inscribe a side AC of an equilateral triangle and a side AB of an equilateral
pentagon. Then arc BC is the difference between one-third and one-fifth of the circum2
ference of the circle. That is, arc BC = 15
of the circumference. Thus, if we bisect that
arc at E, then lines BE and EC will each be a side of a regular 15-gon.
16. Let the triangle be ABC and draw DE parallel to BC cutting the side AB at D and the
side AC at E. Connect BE and CD. Then triangles BDE and CDE are equal in area,
having the same base and in the same parallels. Therefore, triangle BDE is to triangle
ADE and triangle CDE is to triangle ADE. But triangles withe the same altitude are
to one another as their bases. Thus triangle BDE is to triangle ADE as BD is to AD,
and triangle CDE is to triangle ADE and CE is to AE. It follows that BD is to AD
as CE is to AE, as desired.
17. Let ABC be the triangle, and let the angle at A be bisected by AD, where D lies on the
side BC. Now draw CE parallel to AD, meeting BA extended at E. Now angle CAD
is equal to angle BAD by hypothesis. But also angle CAD equals angle ACE and angle
BAD equals angle AEC, since in both cases we have a transversal falling across parallel
lines. It follows that angle AEC equals angle ACE, and therefore that AC = AE. By
proposition VI-2, we know that BD is to DC as BA is to AE. Therefore BD is to DC
as BA is to AC, as claimed.
18. Let a = s1 b + r1 , b = s2 r1 + r2 , . . ., rk1 = sk+1 rk . Then rk divides rk1 and therefore
also rk2 , . . . , b, a. If there were a greater common divisor of a and b, it would divide
r1 , r2 , . . ., rk . Since it is impossible for a greater number to divide a smaller, we have
shown that rk is in fact the greatest common divisor of a and b.
9
Name:Stanislav Shur
IP:74.88.120.168
19.
963 = 1 657 + 306
657 = 2 306 + 45
306 = 6 45 + 36
45 = 1 36 + 9
36 = 4 9 + 0
Therefore, the greatest common divisor of 963 and 657 is 9.
20. Since 1 x = x2 , we have
1 = 1 x + (1 x) = 1 x + x2
x = 1 x2 + (x x2 ) = 1 x2 + x(1 x) = 1 x2 + x3
x2 = 1 x3 + (x2 x3 ) = 1 x3 + x2 (1 x) = 1 x3 + x4
Thus 1 : x can be expressed in the form (1, 1, 1, . . .).

21.
46 = 7 6 + 4
6 = 14 + 2
4 = 22
23 = 7 3 + 2
3 = 12 + 1
2 = 21
Note that the multiples 7, 1, 2 in the first example equal the multiples 7, 1, 2 in the
second.
22. In Figure 2.16 (left), let AB = 7 and the area of the given figure be 10. q
The construction
7
7
described on p. 72 then determines x to be BS. This value is 2 49
4 10 = 2
q
9
7
4 =
q 2
7
49
2 +
4
3
2
= 2. The second solution is BE + ES = AE + ES = AS. This value is
10 =
7
2
9
4
7
2
3
2
= 5.
23. In Figure 2.16 (right), let AB = 10 and the area of the given figure be 39. The construction described on p. 72 then determines x to be BS. This value is 52 + 39 5 =
64 5 = 8 5 = 3.
24. Suppose m factors two different ways as a product of primes: m = pqr s = p0 q 0 r0 s0 .
Since p divides pqr s, it must also divide p0 q 0 r0 s0 . By VII30, p must divide one
of the prime factors, say p0 . But since both p and p0 are prime, we must have p = p0 .
After canceling these two factors from their respective products, we can then repeat the
argument to show that each prime factor on the left is equal to a prime factor on the
right and conversely.
10
Name:Stanislav Shur
IP:74.88.120.168
25. One standard modern proof is as follows. Assume there are only finitely many prime
numbers p1 , p2 , p3 , . . ., pn . Let N = p1 p2 p3 pn + 1. There are then two possibilities.
Either N is prime or N is divisible by a prime other than the given ones, since division
by any of those leaves remainder 1. Both cases contradict the original hypothesis, which
therefore cannot be true.
26. We begin with a square inscribed in a circle of radius 1. If we divide the square into four
isosceles triangles, each with vertex angle
a right angle, then the base of each triangle
has length b1 = 2 and height h1 = 22 . Then the area A1 of the square is equal to
4 12 b1 h1 = 2b1 h1 = 2. If we next construct an octagon by bisecting the vertex angles of
each of these triangles and connecting the points on the circumference, the octagon is
formed of eight isosceles triangles. The base of each triangle has length
b2 =
v
u
u
t
b1
2
!2
v
u
u
t
h2 =
v
u
u
t
+ (1 h1
)2
and height
b1
2
!2
+ h21
b2
1
2
!2
2h1 + 1 =
q
2 2h1 =
2+
2
Thus the octagon has area A2 = 8 12 b2 h2 = 4b2 h2 = 2 2 = 2.828427. If we continue

in this way by always bisecting the vertex angles of the triangles to construct a new
polygon, we get that the area An of the nth polygon is given by the formula An =
2n+1 12 bn hn = 2n bn hn , where
bn =
v
u
u
t
bn1
2
!2
+ (1 hn1
)2
v
u
u
t
and
hn =
bn1
2
v
u
u
t
!2
+ h2n1 2hn1 + 1 =
bn
1
2
!2
2 2hn1
The next two results using this formula are A3 = 3.061467 and A4 = 3.121445.
27. Since BC is the side of a decagon, triangle EBC is a 36-72-72 triangle. Thus 6 ECD =
108 . Since CD, the side of a hexagon, is equal to the radius CE, it follows that triangle
ECD is an isosceles triangle with base angles equal to 36 . Thus triangle EBD is a
36-72-72 triangle and is similar to triangle EBC. Therefore BD : EC = EC : BC or
BD : CD = CD : BC and the point C divides the line segment BD in extreme and
mean ratio.
28. Let ABCDE be the pentagon inscribed in the circle with center F . Connect AF and
extend it to meet the circle at G. Draw F H perpendicular to AB and extend it to
11
Name:Stanislav Shur
IP:74.88.120.168
meet the circle at K. Connect AK. Then AK is a side of the decagon inscribed in
the circle, while BF = AF is the side of the hexagon inscribed in the circle. Draw F L
perpendicular to AK; let N be its intersection with AB and M be its intersection with
the circle. Connect KN . Now triangles ANK and AKB are isosceles triangles with a
common base angle at A. Therefore, the triangles are similar. So BA : AK = AK : AN ,
or AK 2 = BA AN . Further, note that arc BKM has measure 54 , while arc BCG has
measure 108. It follows that 6 BF N = 6 BAF . Since triangles BF N and BAF also
have angle F BA in common, the triangles are similar. Therefore, BA : BF = BF : BN ,
or BF 2 = BA BN . We therefore have AK 2 + BF 2 = BA AN + BA BN =
BA (AN + BN ) = BA2 . That is, the sum of the squares on the side of the decagon
and the side of the hexagon is equal to the square on the side of the pentagon.
29. C =
360
7 15
5000 = 250, 000 stades. This value equals 129,175,000 feet or 24,465 miles.
The diameter then equals 7,787 miles.
12
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER THREE
1. Lemma 1: DA/DC = OA/OC by Elements VI3. Therefore DA/OA = DC/OC =
(DC +DA)/(OC+OA) = AC/(CO+OA). Also, DO 2 = OA2 +DA2 by the Pythagorean
Theorem.
Lemma 2: AD/DB = BD/DE = AC/CE = AB/BE = (AB + AC)/(CE + BE) =
(AB + AC)/BC. Therefore, AD2 /BD2 = (AB + AC)2 /BC 2 . But AD2 = AB 2 BD 2 .
So (AB 2 BD 2 )/BD2 = (AB + AC)2 /BC 2 and AB 2 /BD2 = 1 + (AB + AC)2 /BC 2 .
2. Set r = 1, ti and ui as in the text, and Pi the perimeter of the ith circumscribed polygon.
Then the first ten iterations of the algorithm give the following:
t1 = .577350269
t2 = .267949192
t3 = .131652497
t4 = .065543462
t5 = .03273661
t6 = .016363922
t7 = .0081814134
t8 = .004090638249
t9 = .002045310568
t10 = .001022654214
u1 = 1.154700538
u2 = 1.03527618
u3 = 1.008628961
u4 = 1.002145671
u5 = 1.0005357
u6 = 1.00013388
u7 = 1.000033467
u8 = 1.000008367
u9 = 1.000002092
u10 = 1.000000523
P1 = 3.464101615
P2 = 3.21539031
P3 = 3.159659943
P4 = 3.146086215
P5 = 3.1427146
P6 = 3.141873049
P7 = 3.141662746
P8 = 3.141610175
P9 = 3.141597032
P10 = 3.141593746
3. Let d be the diameter of the circle, ti the length of one side of the regular inscribed
polygon of 3 2i sides, and ui the length of the other leg of the right triangle formed from
the diameter and the side of the polygon. Then
ti+1 2
t2i
=
d2
t2i + (d + ui )2
or
ti+1 =
dti
ui+1 =
t2i + (d + ui )2
If Pi is the perimeter ofthe ith inscribed polygon, then

t1 =
d
2
= 0.5 and u1 =
t1
t2
t3
t4
t5
t6
t7
t8
=
=
=
=
=
=
=
=
3d
2
d2 ti+1 2 .
Pi
d
32i ti
d .
So let d = 1. Then
= 0.8660254. Then repeated use of the algorithm gives us:
0.500000000
0.258819045
0.130526194
0.06540313
0.032719083
0.016361731
0.008181140
0.004090604
u1
u2
u3
u4
u5
u6
u7
u8
=
=
=
=
=
=
=
=
0.866025403
0.965925826
0.991444861
0.997858923
0.999464587
0.999866137
0.999966533
0.999991633
P1
P2
P3
P4
P5
P6
P7
P8
=
=
=
=
=
=
=
=
3.000000000
3.105828542
3.132628656
3.13935025
3.141031999
3.141452521
3.141557658
3.141583943
13
Name:Stanislav Shur
t9 = 0.002045306
IP:74.88.120.168
u9 = 0.999997908
P9 = 3.141590016
4. We can prove the inequality

b < 2a + 1. To
side and noting that
simply by squaring each
find the approximands to 3, begin with 2 14 > 22 1 > 2 13 , or 74 > 3 > 53 . Since
2
3 = 13 52 + 2, we continue with 13 (5 + 15 ) > 13 52 + 2 > 13 (5 + 11
), or 26
> 3 > 57
15
33 .
1
1
1
1
1
1
2
2
Again, since 3 = 15 26 1, we get 15 (26 52 ) > 15 26 1 > 15 (26 51 ), or
1351
1325
265
780 > 3 > 765 = 153 .
5. Let the equation of the parabola be y = x2 +1. Then the tangent line at C = (1, 0) has
the equation y = 2x+2. Let the point O have coordinates (a, 0). Then MO = 2a+2,
OP = a2 + 1, CA = 2, AO = a + 1. So MO : OP = (2a + 2) : (1 a2 ) = 2 : (1 a) =
CA : AO.
6. a. Draw line AO. Then MS SQ = CA AS = AO2 = OS 2 + AS 2 = OS 2 + SQ2 .
b. Since HA = AC, we have HA : AS = MS : SQ = MS 2 : MS SQ = MS 2 :
(OS 2 + SQ2 ) = MN 2 : (OP 2 + QR2 ). Since circles are to one another as the squares
on their diameters, the latter ratio equals that of the circle with diameter MN to
the sum of the circle with diameter OP and that with diameter QR.
c. Since then HA : AS = (circle in cylinder):(circle in sphere + circle in cone), it
follows that the circle placed where it is is in equilibrium about A with the circle in
the sphere together with the circle in the cone if the latter circles have their centers
at H.
d. Since the above result is true whatever line MN is taken, and since the circles make
up the three solids involved, Archimedes can conclude that the cylinder placed where
it is is in equilibrium about A with the sphere and cone together, if both of them
are placed with their center of gravity at H. Since K is the center of gravity of the
cylinder, it follows that HA : AK = (cylinder):(sphere + cone).
e. Since HA = 2AK, it follows that the cylinder is twice the sphere plus the cone AEF .
But we know that the cylinder is three times the cone AEF . Therefore the cone
AEF is twice the sphere. But the cone AEF is eight times the cone ABD, because
each of the dimensions of the former are double that of the latter. Therefore, the
sphere is four times the cone ABD.
7. Since BOAP C is a parabola, we have DA : AS = BD 2 : OS 2, or HA : AS = MS 2 :
OS 2 . Thus HA : AS = (circle in cylinder):(circle in paraboloid). Thus the circle in
the cylinder, placed where it is, balances the circle in the paraboloid placed with its
center of gravity at H. Since the same is true whatever cross section line MN is taken,
Archimedes can conclude that the cylinder, placed where it is, balances the paraboloid,
placed with its center of gravity at H. If we let K be the midpoint of AD, then K
is the center of gravity of the cylinder. Thus HA : AK = cylinder:paraboloid. But
HA = 2AK. So the cylinder is double the paraboloid. But the cylinder is also triple the
volume of the cone ABC. Therefore, the volume of the paraboloid is 3/2 the volume of
the cone ABC which has the same base and same height.
14
Name:Stanislav Shur
IP:74.88.120.168
8. Let the parabola be given by y = a bx2 . Then the area A of the segment cut off by
the x axis is given by

a/b
Z a/b

1
A=2
(a bx2 ) dx = 2 ax bx3
3
0
0
r
r
r
4a a
a 2a a
=
.
= 2a
b
3 b
3 b
q
Since the area of the inscribed triangle is a
a
b,
the result is established.
9. Let the equation of the parabola be y = x2 , and let the straight line defining the segment
be the line through the points (a, a2 ) and (b, b2 ). Thus the equation of this line is
(a b)x + y = ab, and its normal vector is N = (a b, 1). Also, since the midpoint
b2 +a2
of that line segment is B = ( ba
,
2
2 ), the x-coordinate of the vertex of the segment
ba
2
is 2 . If S = (x, x ) is an arbitrary point on the parabola, then the vector M from
(a, a2 ) to S is given by (x + a, x2 a2 ). The perpendicular distance from S to the line
is then the dot product of M with N , divided by the length of N . Since the length of N
is a constant, to maximize the distance it is only necessary to maximize this dot product.
The dot product is (x+a, x2 a2 )(ab, 1) = axbx+a2 ab+x2 a2 = axbx+x2 ab.
The maximum of this function occurs when a b + 2x = 0, or when x = ba
2 . And, as
we have already noted, the point on the parabola with that x-coordinate is the vertex
of the segment. So the vertex is the point whose perpendicular distance to the base of
the segment is the greatest.
10. Let r be the radius of the sphere. Then we know from calculus that the volume of the
sphere is VS = 43 r3 and the surface area of the sphere is AS = 4r2 . The volume
of the cylinder whose base is a great circle in the sphere and whose height equals the
diameter has volume is VC = r2 (2r) = 2r3 , while the total surface area of the cylinder
is AC = (2r)(2r) + 2r2 = 6r2. Therefore, VC = 32 VS and AC = 32 AS , as desired.
11. Suppose the cylinder P has diameter d and height h, and suppose the cylinder Q is
constructed with the same volume but with its height and diameter both equal to f . It
follows that d2 : f 2 = f : h, or that f 3 = d2 h. It follows that one needs to construct the
cube root of the quantity d2 h, and this can be done by finding two mean proportionals
between 1 and d2 h, or, alternatively, two mean proportionals between d and h (where
the first one will be the desired diameter f ).
12. The two equations are x2 = 4ay and y(3a x) = ab. Pick easy values for a and b, say
a = 1, b = 1, and then the parabola and hyperbola may be sketched.
q
13. The focus of y 2 = px is at ( p4 , 0). The length of the latus rectum is 2 p p4 = p.

15
Name:Stanislav Shur
IP:74.88.120.168
p 2
2
14. The equation of the ellipse can be rewritten as 2a
x px+y 2 = 0 or as x2 2ax+ 2a
p y = 0,
or finally as
(x a)2
y2
= 1.
+
a2
pa/2
Therefore the center of the ellipse is at (a, 0) and b2 =

analogously.
pa
2 .
The hyperbola can be treated
15. Let the parabola be y 2 = px and the point C = (x0 , y0 ). Then the tangent line at C has
slope 2yp0 , and the equation of the tangent line is y = 2yp0 (x x0 ) + y0 . If we set y = 0,
we can solve this equation for x to get x = x0 .
16. a. Let the ellipse be given by the equation b2 x2 + a2 y 2 = a2 b2 . Let P have coordinates
2
(x0 , y0 ). Then the slope of the tangent line at P is ab 2xy0 . Thus the equation of line DK
0
is y = ab 2xy0 x. By solving this equation simultaneously with the equation of the ellipse,
0
we get the coordinates of the point D as ( ayb 0 , bxa0 ). It follows that the slope of the
tangent line at D is
b2 ay0 /b
a2 bx0 /a
y0
x0 ,
which is the slope of the diameter P G, as desired.
b. Given that the coordinates of P are (x0 , y0 ), it follows that tan =

Similarly, since the coordinates of D are
( ayb 0 , bxa0 ),
y0
x0
as before.
0 /a
it follows that tan = bx
ay0 /b =
ba2xy0 .
0
c. Take an arbitrary point S in the plane with rectangular coordinates (x, y) and oblique
coordinates (x0 , y 0 ). By drawing lines from S parallel to the two original axes and to
the two oblique axes, one can show that x = x0 cos y 0 cos(180 ) = x0 cos +
y 0 cos and that y = x0 sin +y 0 sin(180) = x0 sin +y 0 sin . If we replace x and y
in the equation of the ellipse by their values in terms of x0 and y 0 , we get the equation
specified in the problem, once we notice that b2 cos cos + a2 sin sin = 0, given
the values for tan and tan found in part b.
d. Let y = (tan )x be the equation of the diameter P G. If we solve this equation
simultaneously with the original equation for the ellipse, we find the coordinates of
the point P to be
x=
ab
b2 + a2 tan2
It follows that
0
a =
, y=
x2 + y 2 =
ab tan
b2 + a2 tan2
ab sec
b2
+ a2 tan2
Similarly,
b0 =
Then
2
a0 =
ab tan
b2 + a2 tan2
a2 b2
a2 b2 sec2
=
A
b2 + a2 tan2
or
A=
a2 b2
.
a0 2
16
Name:Stanislav Shur
IP:74.88.120.168
Similarly,
a2 b2
.
b0 2
If we substitute these values for A and C into the equation of the ellipse given in
C=
2
part (c), we get the equation
02
02 a
02
y =b
x0
a0 2
y0
b0 2
= 1, or
x0 b0
b0
0 0
0
0
0
0
=
(a
x
)(a
+
x
)
=
2
2
2 x1 x2
0
0
0
a
a
a
as desired.
e. Since P F = a0 sin( ) and CD = b0 , we have P F CD = a0 b0 sin( ) =
a2 b2
sin()
. But
AC
2
a sin sin )2 =
since b2 cos cos +a2 sin sin = 0, it follows that (b2 cos cos +
0 and therefore that
AC = a2 b2 (sin cos cos sin )2 = a2 b2 sin2 ( ).

Therefore,
P F CD =
a2 b2 sin( )
= ab
ab sin( )
as claimed.
17. By Conics II8, if we pass a secant line through the hyperbola xy = 1 which goes
through points M and N on that curve and points T and U on the y-axis and x-axis
respectively (the asymptotes), then the segments T M and T N are equal. Thus, if we
let M approach N , then the secant line approaches the tangent line at N and therefore
the two line segments T N, NU between N and the asymptotes are equal. Therefore,
the triangles T SN and NRU are congruent. If the coordinates of N are (x0 , x10 ), then
T S = NR = x10 , and NS = x0 . So the slope of the tangent line T N U is
1/x0
1
TS
=
= 2.
SN
x0
x0
18. Let the parabola have the equation y 2 = px, with the focus at ( p4 , 0). Since the slope of
p
, it follows that the direction vector T of
the tangent line at the point P = (x, y) is 2y
the tangent line can be written in the form (2y, p). Similarly, the direction vector L of
the line parallel to the axis can be written as (1, 0) and the direction vector V of the line
from P to the focus can be written as (x p4 , y). Then the cosine of the angle between
T and L is 2y2 2 . The cosine of the angle between T and V is given by
4y +p
2y x
q
4y 2 + p2
r
p
4
+ py
p
4
2
py
r 2
2xy +
2y x +
p
4
2y
=q
.

= q

2 = q 2
2 x+ p
2 + p2
p
4y
+
p
4y
2
2
2
+y
4y + p
x+ 4
4
17
Name:Stanislav Shur
IP:74.88.120.168
Since these two cosines are equal, so are the angles.

19. If the two parallel lines are x = 0 and x = k and the perpendicular line is the x-axis, then
the equation of the curve satisfying the problem is y 2 = px(k x) or y 2 = kpx px2 .
This is the equation of a conic section.
20. Since the square of the distance between a point and a line is a quadratic function of the
coordinates x, y of the point, and since the same is true for the product of the distances
to two separate lines, the equation defining the locus in the three-line problem will be
a quadratic equation in x and y. Thus the locus will be a conic section, possibly a
degenerate one.
21. crd 120 =
4R2 R2 =
3R = 103; 55, 23; crd 30 =
q
q
R(2R crd 120 ) = 31; 03, 30;
crd 150 = 4R2 crd2 30 = 115; 54, 40; crd 15 = R(2R crd 150 ) = 15; 39, 47.
Similarly, crd 165 = 118; 58, 25 and crd 7 12 = 7; 50, 54.

22. Use a quadrilateral ABCD with AB = crd , BC = crd (180 ( + )), CD = crd ,
AD = 120 (the diameter of the circle). The diagonals are then AC = crd (180 ) and
BD = crd (180 ). Then apply Ptolemys theorem.
23. 120 crd (72 60) = crd (72)crd (120) crd (60)crd (108). So 120 crd(12) = 70; 32, 3
103; 55, 23 q60 97; 4, 56 = 1505; 11, 34. It follows that crd (12)
= 12; 32, 36. Then
q
2
crd (168) = 4 602 crd (12) = 119; 20, 33. Then crd (6) = 60(2 60 crd (168))
= 6; 16, 49. Similarly, crd (3) = 3; 8, 29; crd (1 12 ) = 1; 34, 15; and crd ( 34 ) = 0; 47, 7.
24. When = 90 , then = 23 510 and = 90 . When = 45 , we have sin =
sin(23 510 ) sin(45 ) and = 16 370 . Also tan = cos(23 510 ) tan(45 ), so = 42 270 .
By symmetry, the values for the declination at 270 and 315 are the negatives of the
values at 90 and 45 , respectively.
25. To calculate (60 , 45 ), we note that if = 60 , then = 20 300 and = 57 440 . Since
sin = tan tan 45 , we have = 21 570 and = = 35 470 . If = 90 , then
= 23 510 and = 90 . So = 26 140 and = = 63 460 .
27. Note that is the latitude where the sun is directly overhead at noon on the summer
solstice. The angular distance between the noon altitudes of the sun at the summer and
winter solstice is, given the assumption that at any given time the suns rays to every
point on the earth are parallel to each other, equal to the angle between the sun at noon
on the summer solstice and the sun at noon on the winter solstice, as viewed from the
center of the earth. And this angle, by Figure 3.34, is twice .
18
Name:Stanislav Shur
IP:74.88.120.168
28. L(, ) = 180 + 2(, ). When = 60 and = 36 , we calculate that sin =

tan tan = tan(20 300 ) tan(36 ); so = 15 460 and L = 211; 32, which corresponds to
14 hours, 6 minutes. Therefore, sunrise is 7 hours, 3 minutes before noon, or 4:57 a.m.
and sunrise is at 7:03 p.m.
29. If the length of day is 15 hours when = 90 , then 180 + 2(90 , ) = 225 . Therefore
sin(22 300 )
0
= 22 300 and, since sin = tan tan , we have tan = tan(23
510 ) , so = 40 53 .
30. The expression tan tan will be greater than 1 for = 23 12 when .4348 tan > 1, or
when tan > 2.2998, or when > 66 12 . When that occurs, the formula for L no longer
makes sense. Since when tan tan = 1, we know that L = 360 or 24 hours, it follows
that the sun does not set at all on the summer solstice when the latitude is greater than
66 12 .
31. If = 45 , the = 16 370 ; so SZ = = 45 16 370 = 28 230 . Similarly, if = 90 ,
then = 23 510 and SZ = 21 90 .
32. The sun is directly overhead at noon at latitude 20 when = 20 . Since sin = sinsin
23.5 ,
we find that = 59 . This value for the longitude of the sun occurs at approximately 60
days after the spring equinox and 60 days before the fall equinox, or at approximately
May 20 and July 21.
33. The maximal northerly sunrise point occurs when = 90 and therefore when = 23 510 .
0
When = 36 , we calculate that sin = sinsin233651
and = 29 590 north of east.
sin
34. When = 75 , we need to find so that sin

15 = 1. Clearly, = 15 , and since
0
sin 15 = sin 23 51 sin , it follows that = 39 48 . This value occurs approximately 40
days after the vernal equinox, or about April 30.
19
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER FOUR
1. Let x = Diophantus age at death. Then x = 16 x +
that 9x = 756 and x = 84.
1
12 x
+ 17 x + 5 + 12 x + 4. It follows
2. To solve x + y = 20; xy = 96, we set x = 10 + z, y = 10 z. Then 100 z 2 = 96; z 2 = 4;

and z = 2. Thus x = 12 and y = 8.
3. To find two squares whose difference is 60, set x2 = smaller square and x2 + 60 = larger
square. Then x2 + 60 = (x + 3)2 , where 3 is arbitrarily chosen. This equation reduces
289
1
1
to 6x = 51 and therefore x = 17
2 . The two squares are therefore 4 = 72 4 and 132 4 . In
the general case, the two squares are x2 and x2 + b = (x + a)2 , where a2 < b. It follows
2
that x = ba
2a and the two squares are found to be
b a2
2a
!2
and
b + a2
2a
!2
4. Let x2 be the least square and (x + m)2 = x2 + 2mx + m2 be the middle square. The
difference is 2mx + m2 . Therefore the largest square is x2 + 2mx + m2 + n(2mx +
m2 ) = x2 + (2m + 2mn)x + m2 + nm2 = (x + b)2 = x2 + 2bx + b2 . Provided that
m2 (1 + n) < b2 < m2 (1 + n)2 , the solution is
x=
b2 m2 nm2
.
2m + 2mn 2b
5. To solve x y = 10, x3 y 3 = 2170, set x = z + 5 and y = z 5. It follows that

(z + 5)3 (z 5)3 = 2170. This equation reduces to 30z 2 = 1920 or z 2 = 64 or z = 8.
Thus x = 13 and y = 3. In the general case, if x y = a and x3 y 3 = b, we set
x = z + a2 and y = z a2 If we substitute for x and y in the second equation, we get as
equation in z which reduces to z 2 =
a square.
4ba3
12a .
It follows that this latter expression must be
6. To solve x + y = 20, x3 + y 3 = 140(x y)2 , set x = 10 + z, y = 10 z. Then

(10 + z)3 + (10 z)3 = 140(2z)2 . This equation reduces to 2000 + 60z 2 = 560z 2 or
500z 2 = 2000 or z 2 = 4 or z = 2. Thus the solution is x = 12, y = 8. In general, if
x + y = a, x3 + y 3 = b(x y)2, we set x = a2 + z, y = a2 z. On substituting into the
second equation, we get 2( a2 )3 + 6 a2 z 2 = b(2z)2 , which reduces to
a3
4
a3
4
+ 3az 2 = 4bz 2 or
a
= (4b 3a)z 2 . Thus z 2 = 4(4b3a)
. This equation has a rational solution provided
that the right side is a square, and that condition is equivalent to Diophantus condition
that a3 (b 34 a) is a square.
20
Name:Stanislav Shur
IP:74.88.120.168
7. Simply divide the given square a2 into two squares. This is possible by II8.
8. We want to solve x + y = (x3 + y)3. We set x = 2z and y = 27z 3 2z (so that
x + y = (3z)3 ). Then (x3 + y)3 = (35z 3 2z)3 = (3z)3 . It follows that 35z 2 = 5. This
is impossible for rational z. But now note that 35 = 27 + 8 = 33 + 23 and 5 = 3 + 2.
In order that the equation in z be solvable in rationals, we need two numbers a and
3
+b3
is a square. So let a + b = 2 (where 2 is
b (to replace the 3 and 2) so that aa+b
3
3
arbitrary). Then b = 2 a and a + b must equal 2 times a square. This implies that
a3 + (2 a)3 = 8 12a + 6a2 = 2(square) or that 4 6a + 3a2 is a square. So set
16
4 6a + 3a2 = (2 4a)2 and solve for a. We get a = 10
13 and therefore b = 13 . Since it is
only the ratio of a and b which is important, we can choose a = 5, b = 8 and therefore
put x = 5z, y = 512z 3 5z and repeat the initial calculation. We then get 637z 3 = 13z
1
, so z = 17 . Then x = 57 ; y = 267
and z 2 = 49
343 is the desired solution.
9. Suppose the right triangle has legs a, b, hypotenuse c, and angle bisector d. Let r be the
length of that part of leg a from the right angle to the point where the bisector intersects
the leg. To make the right triangle with the angle bisector as hypotenuse a rational
triangle, we can set d = 5x and r = 3x. It follows that b = 4x. If we then let a = 3, we
have from Elements VI3 that c : (a r) = b : 4 or that c : (3 3x) = 4x : 3x. Thus
c = 4 4x and the reason why a was chosen to be 3 is evident. By the Pythagorean
theorem, we have (4 4x)2 = 32 + (4x)2 or 16 32x + 16x2 = 16x2 + 9. Thus 32x = 7
7
. To get integral answers, we can multiply through by 32. Thus the original
and x = 32
triangle is (96, 28, 100) and the bisector equals 35.
10. The diagram for Elements VI28 is Fig. 2.16. Let us assume that the proposed rectangle
has been constructed with base AS and area equal to c and that the defect is a square.
If we set AB = b, and BS = x, then AS = b x and x(b x) = c. Since the maximum
of the function f (x) = x(b x) occurs when x = 2b , and since this maximum is ( 2b )2 ,
it follows that c cannot exceed the value ( 2b )2 . This means that the area c of the given
rectilinear figure must not be greater than the area of the square on half the given line
of length b.
11. Assume that the theorem is true. Then AB 2 +BC 2 = 3AC 2 . But since AB = AC +BC,
we have (AC + BC)2 + BC 2 = 3AC 2 . This reduces to AC 2 + 2AC BC + 2BC 2 = 3AC 2
or AC BC + BC 2 = AC 2 . This in turn implies that BC(AC + BC) = AC 2 or that
AB BC = AC 2 . But this is precisely the statement that AB is cut in extreme and
mean ratio at C.
12. Suppose that three of the lines have equations x = a, x = b, x = c, that the other two
have equations y = d, y = e, and that the fixed line has length k. Then the equation
of the locus is (x a)(x b)(x c) = k(y d)(y e). Other arrangements of the lines
will give somewhat different equations, but in any case the locus is described by a cubic
equation in x and y.
21
Name:Stanislav Shur
IP:74.88.120.168
1
13. The equation is x 27 x 12
x 16 x 13 x 20 12 11 = 1. Multiplying by 84 and
simplifying gives 11x = 3696 or x = 336.
14. In 12 days the spouts will fill 12 + 6 + 4 + 3 = 25 tanks. Therefore, one tank will be filled
in 12
25 of a day.
15. This problem can be translated into two equations in two unknowns: x + 10 = 3(y 10);
y + 10 = 5(x 10). We can write these as x 3y = 40; 5x + y = 60. The solution
is then that A has x = 15 57 coins and B has y = 18 47 coins.
22
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER FIVE
1. We write, in order, the Chinese form of 56, 554, 63, and 3282:
2. The largest digit a so that (100a)2 < 142, 884 is a = 3. If we subtract 3002 = 90, 000 from
142,884, the remainder is 52,884. We then need to find b so that 2(100a)(10b) < 52, 884,
or 6000b < 52, 884. We take b = 7 and check that 6000b + (10b)2 < 52, 884. But this
inequality reduces to 42, 000 + 4900 < 52, 884. Since the left side equals 46,900, the
inequality is in fact true. Note that if we had taken b = 8, this second inequality would
not have been true. We now subtract 46,900 from 52,884 to get 5,984. We now need to
find c so that 2(370)c < 5984. We try c = 8 and check that 740c + c2 5984. But the
left side of this inequality is in fact equal to 5984, so the desired square root is 378.
3. 560 + 350 + 180 = 1090. Tx =
180
56
1090 100 = 16 109 .
560
1090
41
100 = 51 109
; Ty =
350
1090
12
100 = 32 109
; Tz =
4. In 15 days, the first channel fills the reservoir 45 times, the second channel 15 times, the
third, 6 times, the fourth, 5 times, and the fifth 3 times. It follows that in 15 days the
reservoir is filled 74 times. To fill it once then requires 15
74 of a day.
5. If x is the unknown amount, the conditions show that after the first tax the man had
2
4 2
6 4 2
3 x; after the second, he had 5 3 x; and after the third, he had 7 5 3 x. This amount
15
must equal 5. It follows that x = 175
16 = 10 16 pounds.
6. If x is the hypotenuse of a right triangle and 10 and d the legs, then x = d + 1 or
d = x 1. Then x2 = (x 1)2 + 100 and x = 50.5.
7. If we set x to be the length of a side of the city, draw a line through the center of the city
extending 20 pu north and 14 pu south, extend a line 1775 pu west from the bottom of
that line, and connect the end of that new line with the end of the line to the north, we
get a right triangle with legs x + 34 and 1775. Since we also have a similar triangle with
legs 20 and x2 , we get the proportion 20 : x2 = (x + 34) : 1775. The resulting equation is
x2 + 34x = 71000 and x = 250.
23
Name:Stanislav Shur
IP:74.88.120.168
50.4
x
8. If x is the depth of the well, the similarity relationship gives

54.6
0.4 = 57.5.
0.4
5 .
Thus x =
9. The simplest way is to set y = DC, x = CE. Then

93
3 120
10
y
y
=
and
=
.
x+5
5
x + 13 13
13 13
Simplifying these equations gives 40y = 30x + 400; 200y = 151x + 755. Solving these
simultaneously gives the solution x = 1245, y = 943 34 .
10. We begin with c6q= r = 10 and a6 =

300. Then c12 =
(10/2)2 + (10 8.6603)2 = 5.1764, and S24 = 12 1210c12 = 310.5859.
We then get a12 =

2.6105 and S48 =
c48 =
102 52 = 8.6603. Therefore, S12 = 12 61010 =
1
2
100 (c12 /2)2 = 9.6593. Next, c24 =
24 10 c24 = 313.2629. Next. a24 =

1
(c24 /2)2 + (10 a24 )2 = 1.3081. So S96 =

q
we get a48 = 100 (c48 /2)2 = 9.9786, so c96 =

then get S192 = 12 96 10 c96 = 314.1032.
q2
(c12 /2)2 + (10 a12 )2 =
100 (c24 /2)2 = 9.9144 and
48 10 c48 = 313.9350. Finally,
(c48 /2)2 + (10 a48 )2 = 0.6544. We
12. From Figure 5.8, we see that one-eighth of the volume of the double box-lid is
3
3
x2 ) dx = r3 r3 = 23 r3 . It follows that the entire volume is 16
3 r .
13. If x is the yield of good grain, y the yield of
grain, then the system of equations is
2x + y
3y +
x
+
In matrix form we get, in turn
1 0 2
0 0
0 3 1
1 3
4 1 0
8 1
1 1 1
1 1
It follows that 25z = 4 or z =
9
or x = 25
.
4
25 ;
Rr
0 (r
ordinary grain, and z the yield of worst
z
4z
2
1
0
1
3y + z = 1, 3y =
= 1
= 1
= 1
0
0
25
4
21
25 ,
0
3
1
1
or y =
2
1
0
1
7
25 ;
2x + y = 1, 2x =
18
25 ,
14. a. In this case, we see by trial that the solution is between 6 and 7. So we use 6 in the
synthetic division procedure:
6 | 16 192 1863.2
96
1728
6 | 16 288 |135.2
96
6 | 16 |384
|16
24
Name:Stanislav Shur
IP:74.88.120.168
For the next step, we will use decimals. The (positive) solution to 16x2 + 384x 135.2
is between 0 and 1. Again, we find by trial that the value is between 0.3 and 0.4. So our
next chart is as follows:
384
135.2
.3 | 16
4.8
116.64
.3 | 16 388.8 |18.56
4.8
.3 | 16 |393.6
|16
For a third step, we will try values between 0 and 0.1. Again, the closest value seems to
be 0.5, as in the following chart:
.05 | 16 393.6 18.56
.8
19.72
16 394.4 |0.96
Since the last value is relatively close to 0, we will leave the solution as 6.35. If we
wanted to go further, we could have used 0.4 in this last step and continued to find the
next decimal place.
b. In this case, we see by trial that the solution begins with a 2 in the 10s place, so
we try 20 in the chart:
20 |
0
20
15, 245
400
0
296, 900
6, 262, 506.25
5, 938, 000
20
20
20
14, 845
800
296, 900
280, 900
| 324, 506.25
20
40
20
14, 045 |577, 800

1, 200
20 |
60
20
|12, 845
20 |
1 |80
|1
The fourth degree polynomial we find for our next step has a solution between 0 and
1. We will therefore leave this solution as x = 20, but it is not difficult to continue
the solution further.
15. Qins method gives the following diagrams to produce 235 as the answer. We begin by
noting that the answer is a three-digit number beginning with 2.
200 |
0
200
55, 225
40, 000
200 |
200
200
|15, 225
200 |
1 |400
|1
25
Name:Stanislav Shur
IP:74.88.120.168
We next check that the second digit is a 3.

30 |
400
30
15, 225
12, 900
30 |
430
30
|2, 325
30 |
1 |460
|1
The final digit is a 5.

5 | 1 460 2325
5
2325
1 465
|0
16. Qins method gives the following diagrams to produce 234 as the answer. We begin by
noting that the answer is a three-digit number beginning with a 2.
200 |
0
200
0
40, 000
12, 812, 904

8, 000, 000
200 |
200
200
40, 000
80, 000
|4, 812, 904
200 |
400
200
|120, 000
1 |600
|1
The second digit is a 3.
30 |
600
30
120, 000
18, 900
4, 812, 904
4, 167, 000
30 |
630
30
138, 900
19, 800
|645, 904
30 |
660
30
|158, 700
1 |690
|1
The final digit is a 4.
4 | 1 690 158, 700 645, 904
4
2, 776
645, 904
1 694 161, 476
|0
26
Name:Stanislav Shur
IP:74.88.120.168
The third order coefficients occur, for example, in 600 = 3 200, 120, 000 = 3 2002 ,
and in 690 = 3 230 and 158, 700 = 3 2302 .
17. The diagram is as follows, where the solution is x = 23. We begin by noting that the
first digit is a 2.
20 | 1 0
0
0
279, 841
20
400
8, 000
160, 000
20 |
20
20
400
800
8, 000
24, 000
20 |
40
20
1, 200
1, 200
|32, 000
20 |
60
20
|2, 400
20 |
1 |80
|1
|119, 841
The second digit is a 3.

3 | 1 80 2, 400 32, 000 119, 841
3
249
7, 947
119, 841
1 83 2, 649 39, 947
|0
The fourth order coefficients show up in 80 = 4 20, 2400 = 6 202 , 32, 000 = 4 203 ,
and 160, 000 = 1 204 .
18. By the Pythagorean theorem, the altitude h of the lower triangle is given by
q
b2 ( 2c )2 .
c
c 2
2
2 b ( 2 ) as stated. Similarly, the area
c
c 2
4
4
2
2 a ( 2 ) . If x = A + B, then x = (A + B) =
B 4 = 2(A4 + 2A3 B + 2A2 B 2 + 2AB 3 + B 4 ) (A4
The area B of that triangle is then B =

q
A of the upper triangle is A =

A4 + 4A3 B + 6A2 B 2 + 4AB 3 +
2A2 B 2 + B 4 ) = 2(A2 + B 2 )(A2 + 2AB + B 2 ) (A2 B 2 )2 = 2(A2 + B 2 )x2 (A2 B 2 )2 .
The equation
for x followsimmediately. If a = 39, b = 25, and c =30, we have
A = 15 1521 225 = 15 1296 = 15 36 = 540. Similarly, B = 15 400 = 300.

Then 2(A2 + B 2 ) = 2(291, 600 + 90, 000) = 763, 200 and (A2 B 2 )2 = (201600)2 =
40, 642, 560, 000 as desired.
19. Using the notation from the description of Qin Jiushaos method, we first note that
M = 12. Then M1 = 12 3 = 4 and M2 = 12 4 = 3. Also P3 = 1, and P4 = 3.
Therefore, we need to solve two congruences: x1 1 (mod 3), and 3x2 1 (mod 4).
The solutions are x1 = 1 and x2 = 3. Therefore N = 0 4 1 + 1 3 3 = 9 9 (mod 12).
20. Using the notation from the description of Qin Jiushaos method, we first calculate
M = 11 5 9 8 7 = 27720. Then M1 = M 11 = 2520; M2 = M 5 = 5544;
27
Name:Stanislav Shur
IP:74.88.120.168
M3 = M 9 = 3080; M4 = M 8 = 3465; and M5 = M 7 = 3960. We then calculate

that M1 1 (mod 11); M2 4 (mod 5); M3 2 (mod 9); M4 1 (mod 8);
and M5 5 (mod 7). We next need to solve congruences: the solution to 1x1 1
(mod 11) is x1 = 1; to 4x2 1 (mod 5) is x2 = 4; to 2x3 1 (mod 9) is x3 = 5;
to 1x4 1 (mod 8) is x4 = 1; and to 5x5 1 (mod 7) is x5 = 3. Given that 3 of
the ri are equal to 0, we calculate N simply as n = 4 3080 5 + 6 3465 1 = 82, 390.
Subtracting off twice M, we get the solution as N = 82, 390 2 27, 720 = 26, 950, where
the answer is taken modulo 27,720.
28
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER SIX
1. We have KM = KL = BC = a and KA = b. It follows by the Pythagorean Theorem
that AM 2 = KM 2 AM 2 = b2 a2 . Thus the square on AM is the difference of the
two original squares, the one on AB and the one on P Q.
2. The rectangle ABCD is transformed into the gnomon AEGF KH of the same area by
the indicated construction. This gnomon is equal to the difference of the squares on AE
and F K. By the previous exercise, we can construct a square equal to that difference.
This square will therefore be equal to the rectangle, as desired.
3. Since AB = s, we have MN = r = 2s + 13 ( s 2 2 2s ). Thus rs = 12 + 62 16 = 2+6 2 .

Given that the area of the circle of radius MN = r equals the area
square of side
of the
r2
1
r
1
1
2+
2
2
2
AB = s, we have r = s or s2 = or s = . Thus = 6 , or = 2+62 , or
= 3.088311755 . . .
4. If we calculate the sum and difference of the given fractions, we get 0.878681752. If the
square of this side is equal to the area of a circle of diameter 1, then (0.878681752)2 = 4 ,
or = 4(0.878681752)2 = 3.088326491.
5. a. Let 6 ABC = . Then 6 ADC = . Set x = AC. In triangle ABC, we have
x2 = a2 + b2 2ab cos , while in triangle ADC, we have x2 = c2 + d2 2cd cos( ) =
c2 + d2 + 2cd cos . Setting the two expressions for x2 equal, we get a2 + b2 2ab cos =
c2 + d2 + 2cd cos and therefore
cos =
a2 + b2 c2 d2
.
2ab + 2cd
b. We get
x2 = a2 + b2
cd(a2 + b2 ) + ab(c2 + d2 )
2ab(a2 + b2 c2 d2 )
=
.
2ab + 2cd
ab + cd
c. cd(a2 + b2 ) + ab(c2 + d2 ) = a2 cd + b2 cd + c2 ab + d2 ab = (ac + bd)(ad + bc).

d. From parts b and c we get
ac + bd)(ad + bc)
x =
ab + cd
2
or
Similarly, we have
y = BD =
x = AC =
(ac + bd)(ad + bc)

.
ab + cd
(ac + bd)(ab + cd)

.
ad + bc
29
Name:Stanislav Shur
IP:74.88.120.168
6. a. From the law of cosines applied to triangle ABC, we get

b2 = a2 + x2 2ax cos 6 BAE = a2 + x2 2ax
b.
c.
d.
e.
f.
AE
= a2 + x2 2x AE.
a
Therefore b2 a2 = x(x 2AE).

Because x = 2AM, we have b2 a2 = x(2AM 2AE) = x(2EM) = 2x EM.
2
a2
.
Therefore EM = b 2x
By the same arguments as in parts a and b applied to triangle ADC, we get F M =
d2 c2
2x .
Let P be the area of quadrilateral ABCD. Then P is the sum of the areas of
triangles ABC and ADC. Therefore, P = 12 x BE + 12 x DF = 12 x(BE + DF ), and
P 2 = 14 x2 (BE + DF )2 .
Because BE + DF = BK, we get from part d and the Pythagorean Theorem that
P 2 = 14 x2 BK 2 = 14 x2 (BD 2 DK 2 ) = 14 x2 (y 2 EF 2 ).
From parts b and c, we have
EF = EM + F M =
(b2 + d2 ) (a2 + c2 )
b2 a2 d2 c2
+
=
.
2x
2x
2x
Substituting this value into the expression in part e, along with the values for x2 and
y 2 from exercise 5, we have
1 (ac + bd)(ad + bc) (ac + bd)(ab + cd) [(b2 + d2 ) (a2 + c2 )]2
P =
4
ab + cd
ad + bc
4x2
h
i
2
1
1
= (ac + bd)2
(b2 + d2 ) (a2 + c2 )
4
16
i
1 h
=
4(ac + bd)2 [(b2 + d2 ) (a2 + c2 )]2 .
16
2
g. We have s a = 12 (a + b + c + d) a = 12 (b + c + d a), with a similar result for the

three other cases.
h. First, we calculate (b + c + d a)(a + c + d b)(a + b + d c)(a + b + c d). By first
multiplying together the first two expressions and then the last two expressions and
then multiplying the two resulting expressions together, we find that this expression
becomes (2ab+2cd+c2 +d2 a2 b2 )(2ab+2cd+a2 +b2 c2 d2 ) = 8abcd+2(a2 b2 +
a2 c2 + a2 d2 + b2 c2 + b2 d2 + c2 d2 ) a4 b4 c4 d4 . On the other hand, if we multiply
out the numerator of the expression for P 2 from part f, we get 4(ac + bd)2 [(b2 +
d2 )(a2 +c2 )]2 = 4(a2 c2 +2abcd+b2 d2 [(b2 +d2 )2 2(b2 +d2 )(a2 +c2 )+(a2 +c2 )2 ] =
4(a2 c2 +2abcd+b2d2 )[b4 +2b2 d2 +d4 2a2 b2 2b2 c2 2a2 d2 2c2 d2 +a4 +2a2 c2 +c4 ] =
8abcd + 2(a2 c2 + b2 d2 + a2 b2 + b2 c2 + a2 d2 + c2 d2 ) a4 b4 c4 d4 , the same
expression
as in the first calculation. It follows that the area of the quadrilateral is
q
S = (s a)(s b)(s c)(s d), as asserted.
30
Name:Stanislav Shur
IP:74.88.120.168
7. We get the following:

1096x + 808 = 3y
3y 808 = 1096x
1096 = 365 3 + 1
3y 808 = (365 3 + 1)x 3(y 365x) 808 = x
t = y 365x
3t 808 = x
x + 808 = 3t
By inspection, we get x = 2, t = 270. Then y = t + 365x = 270 + 730 = 1000.. Since
2 is already the smallest possible solution for x, the result is x = 2, y = 1000, and
N = 808 + 1096 2 = 0 + 3 1000 = 3000.
8. We need to solve 137x + 23 = 60y. We perform Brahmaguptas algorithm:
137x + 23 = 60y
60y 23 = (2 60 + 17)x
60t 23 = 17x
17x + 23 = (3 17 + 9)t
17u + 23 = 9t
9t 23 = (1 9 + 8)u
9v 23 = 8u
8u + 23 = (1 8 + 1)v
8w + 23 = 1v
60y 23 = 137x
60(y 2x) 23 = 17x
17x + 23 = 60t
17(x 3t) + 23 = 9t
9t 23 = 17u
9(t 1u) 23 = 8u
8u + 23 = 9v
8(u 1v) + 23 = 1v
1v 23 = 8w
137 = 2 60 + 17
t = y 2x
60 = 3 17 + 9
u = x 3t
17 = 1 9 + 8
v = t 1u
9=18+1
w = u 1v
By inspection, we find that v = 31, w = 1. We then calculate u = 1v + w = 32,

t = 1u+v = 63, x = 3t+u = 221, and y = 2x+t = 505. So x = 221, y = 505 is a solution
to the equation. The general solution is then x = 221 + 60z, y = 505 + 137z. To get the
smallest value for x, choose z = 3. Then x = 41 and y = 94. Then N = 60y = 5640.
Since the solution is taken module 8220, we get N 5640 (mod 8220).
9. To solve 1096x + 1 = 3y, we apply Brahmaguptas algorithm:
1096x + 1 = 3y
3y 1 = 1096x
3y 1 = (365 3 + 1)x
3(y 365x) 1 = x
3t 1 = x
x + 1 = 3t
By inspection, we find that t = 1, x = 2. Then y
1096x + 10 = 3y, we simply multiply everything by 10:
1096 = 365 3 + 1
t = y 365x
= t + 365x = 731. To solve
x = 20, y = 7310.
10. We will show, via a generalizable example, that Brahmaguptas method does give a
solution to the equation rx + c = sy, assuming that the greatest common divisor of r
and s divides c. In fact, we will assume that the greatest common divisor is equal to
1, although one could generalize the procedure here to the case where it is greater than
1. We will apply the Euclidean algorithm to r and s and assume, for simplicity, that is
stops after four steps. We therefore have the following system:
rx + c = sy
sy c = (q1 s + r1 )x
st c = r1 x
sy c = rx
s(y q1 x) c = r1 x
r1 x + c = st
r = q1 s + r1
t = y q1 x
s = q2 r1 + r2
31
Name:Stanislav Shur
IP:74.88.120.168
r1 x + c = (q2 r1 + r2 )t
r1 (x q2 t) + c = r2 t
u = x q2 t
r2 t c = r1 u
r1 = q3 r2 + r3
r1 u + c = r2 t
r2 t c = (q3 r2 + r3 )u
r2 (t q3 u) c = r3 u
v = t q3 u
r3 u + c = r2 v
r2 = q4 r3 + 1
r2 v c = r3 u
r3 (u q4 v) + c = 1v
w = u q4 v
r3 u + c = (q4 r3 + 1)v
r3 w + c = 1v
1v c = r3 w
We now set w = 1, v = r3 + c, and solve for the other letters. We get u = w + q4 v =
r3 q4 + cq4 + 1, t = v + q3 u = r3 q3 q4 + cq3 q4 + q3 + r3 + c. Then x = u + q2 t = r3 q2 q3 q4 +
cq2 q3 q4 +q2 q3 +q2 r3 +q2 c+r3 q4 +cq4 +1 and y = t+q1 x = q1 x+r3 q3 q4 +cq3 q4 +q3 +r3 +c.
To prove that the method does in fact give us a solution to the original equation, we
need to substitute these values into that equation. In other words, we must show that
rx + c = sy. Substituting the value for r given in the first line of our process, we need
to show that
(q1 s + r1 )x + c = s(q1 x + r3 q3 q4 + cq3 q4 + q3 + r3 + c).
This is equivalent to showing that
r1 x + c = s(r3 q3 q4 + cq3 q4 + q3 + r3 + c).
We next substitute for s its value from the third line. At the same time, we substitute
for x the value we calculated above. We must therefore show that
r1 [q2 (r3 q3 q1 +cq3 q4 +q3 +r3 +c)+r3 q4 +cq4 +1]+c = (q2 r1 +r2 )(r3 q2 q4 +cq3 q4 +q3 +r3 +c).
This is in turn equivalent to showing that
r1 (r3 q4 + cq4 + 1) + c = r2 (r3 q3 q4 + cq3 q4 + q3 + r3 + c).
We next substitute for r1 from the fifth line. We thus must show that
(q3 r2 + r3 )(r3 q4 + cq4 + 1) + c = r2 (r3 q2 q4 + cq3 q4 + q3 + r3 + c).
To demonstrate this equality, it suffices to show that
r3 (r3 q4 + cq4 + 1) + c = r2 (r3 + c).
To do this, we finally substitute for r2 its value from the seventh line. We therefore must
demonstrate that
r3 (r3 q4 + cq4 + 1) + c = (q4 r3 + 1)(r3 + c).
That this final equation is true comes from multiplying it out. We have thus shown that
the values calculated by Brahmaguptas method in fact satisfy the original equation.
11. The Chinese method requires that the moduli be relatively prime. In this problem,
we note that if N 2 (mod 3) and N 3 (mod 4), then N 1 (mod 2) and
also N 5 (mod 6). Therefore, we may ignore the first congruence and solve the last
32
Name:Stanislav Shur
IP:74.88.120.168
three. Using the notation from chapter 5, we first calculate that M = 60. Then M1 = 12,
M2 = 15, M3 = 20, P1 = 2, P2 = 3, and P3 = 2. We must solve 2x1 1 (mod 5);
3x2 1 (mod 4) and 2x3 1 (mod 3). The solutions are x1 = 3, x2 = 3, x3 = 2.
Therefore, N = 4 12 3 + 3 15 3 + 2 20 2 359 (mod 60), and the smallest positive
N is 59. For the Indian method, we solve the first two congruences, then use that answer
along with the third, and the new answer along with the fourth. The solution of N 5
(mod 6) 4 (mod 5) requires solving the equation 6x + 1 = 5y. The solution by the
procedure of the previous problems is x = 4, y = 5 and then N = 4 6 + 5 = 29. We
next solve N 29 (mod 30) 3 (mod 4). We must solve 30x + 26 = 4y. We get
x = 1, y = 14 and N = 1 30 + 29 = 59. To solve N 59 (mod 60) 2 (mod 3),
we note that already 59 2 (mod 3); so N = 59 is the solution to the entire set of
congruences.
12. To solve this congruence in the Chinese fashion, we note that since the two moduli are
relatively prime, M1 = 60 and M2 = 137, while P1 = 60 and P2 = 17. We thus must
solve the two congruences 60x1 1 (mod 137) and 17x2 1 (mod 60). The second
congruence is not important, because its solution will ultimately be multiplied by 0. So
we simply apply the Euclidean algorithm to solve the first congruence. This amounts to
the following:
137 = 2 60 + 17
21+0=2
60 = 3 17 + 9
32+1=7
17 = 1 9 + 8
17+2=9
9= 18+1
1 9 + 7 = 16
so the solution is x1 = 16. Then N = 10 60 16 = 9600 1380 (mod 8220). Although
the Indian method detailed in the text begins with the same Euclidean algorithm, the
steps in the remainder of that process are different from the steps here.
13. This problem is equivalent to finding N to solve N 0 (mod 17) 1 (mod 75). In
the Chinese method, M1 = 75 and M2 = 17. Then P1 = 7 and P2 = 17. We then
solve 7x1 1 (mod 17) and 17x2 1 (mod 75). The solution to the first congruence
is unnecessary, because it will be multiplied by 0. We get the solution to the second
congruence by the Euclidean algorithm. We get
75 = 4 17 + 7
17 = 2 7 + 3
7= 23+1
By substitution, we get 1 = 53 17 12 75, so x2 = 53. Then N 1 17 53 901
(mod 1275). In terms of the original problem, we have 17 53 1 = 75m, so m = 12 and
n = 53.
In the Indian method, we use the Euclidean algorithm in Brahmaguptas procedure:
17n 1 = (4 17 + 7)m
17u 1 = 7m
17n 1 = 75m
17(n 4m) 1 = 7m
7m + 1 = 17u
75 = 4 17 + 7
u = n 4m
17 = 2 7 + 3
33
Name:Stanislav Shur
IP:74.88.120.168
7m + 1 = (2 7 + 3)u
7(m 2u) + 1 = 3u
v = m 2u
7v + 1 = 3u
3u 1 = 7v
7=23+1
3u 1 = (2 3 + 1)v
3(u 2v) 1 = 1v
w = u 2v
3w 1 = 1v
1v + 1 = 3w
We now choose v = 2, w = 1. Then u = w + 2v = 5, m = v + 2u = 12, and
n = u + 4m = 53. Thus, m = 12, n = 53 is the solution.
14. D(u0 v1 + u1 v0 )2 + c0 c1 = D(u0 v1 + u1 v0 )2 + (v02 Du20 )(v12 Du21 ) = 2Du0v1 u1 v0 +
D 2 u20 u21 + v02 v12 = (Du0 u1 + v0 v1 )2 .
15. To solve 83x2 + 1 = y 2, we begin by noting that (1, 9) is a solution for subtractive 2; that
is, 83 12 2 = 92 . If we compose this solution with itself, we get y1 = 83 12 + 92 = 164,
x1 = 9 + 9 = 18, b1 = 4. Therefore, (18, 164) is a solution for additive 4. But then we
can simply divide everything by 4 = 22 to get that (9, 82) is a solution for additive 1:
83 92 + 1 = 822 .
16. To show that (u1 , v1 ) is a solution, we calculate each side of the equation Du21 + 1 = v12
and show that they are equal. The left side is
L=D
2
1
uv(v 2 + 1)(v 2 + 3)
2
+1
1
= Du2 v 2 (v 2 + 1)2 (v 2 + 3)2 + 1
4
1
= (4 + v 2 )v 2 (v 2 + 1)2 (v 2 + 3)2 + 1
4
1
= (v 4 + 4v 2 )(v 2 + 1)2 (v 2 + 3)2 + 1.
4
The right side is

2
1
R = (v + 2) (v 2 + 1)(v 2 + 3) 1
2

1
= (v 2 + 2)2 (v 2 + 1)2 (v 2 + 3)2 (v 2 + 1)(v 2 + 3) + 1
4
1 4
= (v + 4v 2 + 4)(v 2 + 1)2 (v 2 + 3)2 (v 2 + 2)2 (v 2 + 1)(v 2 + 3) + (v 2 + 2)2
4
1
= (v 4 + 4v 2 )(v 2 + 1)2 (v 2 + 3)2 + (v 2 + 1)2 (v 2 + 3)2 (v 2 + 2)2 (v 2 + 1)(v 2 + 3)
4
+ (v 2 + 2)2
1
= (v 4 + 4v 2 )(v 2 + 1)2 (v 2 + 3)2 + (v 2 + 1)(v 2 + 3)[(v 2 + 1)(v 2 + 3) (v 2 + 2)2 ]
4
+ (v 2 + 2)2
1
= (v 4 + 4v 2 )(v 2 + 1)2 (v 2 + 3)2 + (v 2 + 1)(v 2 + 3)(1) + (v 2 + 2)2
4
1
= (v 4 + 4v 2 )(v 2 + 1)2 (v 2 + 3)2 + 1.
4
2
34
Name:Stanislav Shur
IP:74.88.120.168
Thus the two sides are equal as asserted. Next, note that if u or v is even, then 12 uv is
an integer, so u1 is an integer. If both u and v are odd, then v 2 + 1 is even, so 12 (v 2 + 1)
is an integer and u1 is an integer. If v is even, then v 2 + 2 is even and (v 2 + 2) 12 (v 2 + 1)
is an integer. If v is odd, then v 2 + 1 is even, so 12 (v 2 + 1) is an integer. Thus in either
case, v1 is an integer.
17. To solve 13x2 + 1 = y 2 , we begin by noting that (1, 3) is a solution for subtractive
4: 13 12 4 = 32 . By the previous problem, set u1 = 12 1 3 10 12 = 180 and
v1 = 11[ 12 10 12 1] = 649. Then (180, 649) is the desired solution.
18. If Du2 + 2 = v 2 , then D(uv)2 + 1 = Du2 v 2 +1 = (v 2 2)v 2 +1 = v 4 2v 2 +1 = (v 2 1)2 ,
as desired. If Du2 2 = v 2 , then (u1 , v1 ) = (uv, v 2 + 1) solves Du21 + 1 = v12 . The proof
is virtually identical to the previous one.
19. To solve 61x2 + 1 = y 2, we begin by noting that 61 12 + 3 = 82 ; that is, that (1, 8) is a
solution for additive 3. We then need to solve 1m+8 = 3n. The general solution is easily
seen to be m = 1+3t, n = 3+t. We now choose t so that m2 is close to61: t = 2, m = 7,
= 5, b1 = 6149
= 4, and v1 = 61 25 4 = 39.
m2 = 49. Then take u1 = 17+8
3
3
2
2
We now have 61 5 4 = 39 ; that is, (5, 39) is a solution for subtractive 4. Now
use the result of problem 16. Set u1 = 12 5 39(392 + 1)(392 + 3) = 226, 153, 980 and
v1 = (392 + 2)[ 12 (392 + 1)(392 + 3) 1] = 1, 766, 319, 049. Then (u1 , v1 ) is a solution to
the original equation.
20. Since 13 x + 16 x = 12 x, the equation is

1 1
1 1
1 1
1 1
1 1
1 1
1
x x+
x +
x +
x +
x +
x +
x
2
2 2
2 4
2 8
2 16
2 32
2 64
This equation reduces to
1
128 x

= 1161.
= 1161, so x = 148, 608 is the solution.
21. In 1 day, the well is filled 2 + 3 + 4 + 5 = 14 times. Thus the well will be filled once in
1
1
2
14 of a day. In that time period, the first pipe will fill the well 14 2 = 14 full. Similarly,
3
4
5
the second pipe will fill 14 of the well, the third 14 , and the fourth 14 .
22. If t is the number of days until the second person overtakes the first, the equation is
5(t + 7) = 9t. The solution is t = 8 34 days.
23. If x is the amount held by the first traveler, y the amount held by the second, and p
the amount in the purse, then the problem results in two equations in three unknowns:
x+ 12 p = 2y; y + 23 p = 3x. The solution is a one-parameter family, expressible as y = 13
11 x;
30
p = 11 x. Since the solutions must be integers, x must be a multiple of 11. Thus x = 11,
y = 13, p = 30 is a solution, as is any positive integral multiple of that solution.
35
Name:Stanislav Shur
IP:74.88.120.168
24. If we calculate a table of sine values by Bhaskaras formula (not multiplying by 3438)
and compare them to actual sine values, we get the following:
Angle
Bhaskaras sine
Actual sine
0
0.00000
0.00000
10
0.17525
0.17365
20
0.34317
0.34202
30
0.50000
0.50000
40
0.64183
0.64279
50
0.76471
0.76604
60
0.86486
0.86603
70
0.93903
0.93969
80
0.98461
0.98481
90
1.00000
1.00000
An inspection of this table shows that the difference between the two values is greatest
at 10 degrees, where the actual difference is 0.00160, which corresponds to a percentage error
of less that 1%.
25. Brahmaguptas procedure gives us
1
12
(219
+
215)
(219 215)
2(3 34 )
2(3 34 )2
8
2
(4) = 948
= 890 + (434)
15
225
to the nearest integer. Bhaskaras procedure gives
sin(16) = sin(15 + 1) = sin(15) +
sin 16 = 3438
4 16 164
= 953
40, 500 16 164
to the nearest integer. The exact value is 948 to the nearest integer, so Bhaskaras answer
is in excess of the correct answer by approximately 0.5%.
26. We assume that yi is/n (is)3 /6n3 , given the approximation we have found for y.
We put this into the expression for x = cos s and use the formula for the sum of integral
cubes:
"
s3
2s (2s)3
(n 1)s ((n 1)s)3
s s
3+
+
x 1 lim
n n n
6n
n
6n3
n
6n3
s4 3
s2
+ lim
=1
[1 + 23 + + (n 1)3 ]
n 6n4
2
Pn1 3
s2 s4
s2 s4 1
i
=1
+ n
lim i=14
+
=1
2
6
n
2
6 4
s2 s4
+ .
=1
2
24
36
Name:Stanislav Shur
IP:74.88.120.168
We can similarly calculate a new value of y = sin s, using our knowledge of the double
sum formula as well. We have
y s
s3
6
2 "
s3
s3
s3
(2s)3
(2s)3
((n 1)s)3
+
+
+
+
+
+
+
+ lim
n
6n3
6n3
6n3
6n3
6n3
6n3
s5 3
s3
+ lim
=s
[1 + (13 + 23 ) + + (13 + 23 + + (n 1)3 )]
n 6n5
6
s3
s5
=s
+ n
lim
[n(13 + 23 + + (n 1)3 ) (14 + 24 + + (n 1)4 )]
5
6
6n "
Pn1 3
Pn1 4 #
s3 s5
i
i=1 i
=s
+
lim
i=15
4
6
6 n
n
n

s3 s5 1 1
+
=s
6
6 4 5
s5
s3
+
.
=s
6
120
s
n
!#
Putting the new value for y into the formulas for x and y will lead by a similar argument
to the next terms in both the sine and cosine series.
37
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER SEVEN
1. Al-Khwarizmis rule for solving bx + c = x2 translates to the formula
x=
v
u
u
t
!2
b
2
b
+c+ .
2
The main point of the geometric proof is that rectangle RBMN is equal to rectangle
N KT L. Then, because ( 2b )2 is equal to rectangle KHGT , we get that ( 2b )2 + c is
represented by rectangle MAGL, so that the square root in the formula is equal to the
side of that square, namely GA.
2. a. ( 13 x + 1)( 14 x + 1) = 20 transforms to x2 + 7x = 228. The formula then gives
x=
s
2
7
2
+ 228
7
= 12.
2
b. x2 +(10x)2 = 58 transforms to x2 +21 = 10x. The formula gives x = 5 25 21 =

7, 3.
3. a. Multiplying the equation by 2 gives x2 + 10x = 56. The solution is x =
b. Dividing the equation by 2 gives
11
5
2 2 = 3.
x2 +5x
= 24. The formula then gives x =
81 5 = 4.
q
121 5
4 2
4. The equation is
x
10 x
13
+
= .
10 x
x
6
If we multiply both sides by 6x(10 x) and simplify, we get x2 + 24 = 10x. The solutions
are then x = 6 and x = 4.
5. a. The equation is x2 = (10 x) 10. We can rewrite this as x2 + 10x = 10 10. The
formula then gives us that one part is
x=
v
u
u
t
10
2
!2
+ 10 10
The other part is

y = 10 x = 10 +
10
=
2
s
1
2
2
1
2 + 1000
2
1
2 .
2
1
2 + 1000.
2
b.
38
Name:Stanislav Shur
IP:74.88.120.168
6. a. If we set x = y 2 , the equation becomes

[y 2 (2y + 10)]2 = 8y 2 . Taking square
roots and q
rearranging gives us y 2 = (2 + 8)y + 10. The formula then gives us y =
1 + 2 + 13 + 8, and x = y 2 .
b. If we expand the left side, rearrange, and then square both sides, the equation
1
becomes 9x = x2 + 49
4 . An application of the formula gives x = 4 2 8. (We note
that if we use the plus sign in this result, we do not get a correct answer.) As an
then becomes (2y 2 + y)2 = 8y 2 or
alternative, we could set x = 2y 2 . The equation
1
2
2y 2 + y = 8y. The solution to this is y = 81
2 . Then x = 2y = 4 2 8 as
before.
7.
x2
1
x
1
3 13
1
x3
1
x2
6 23
20 30
6 0
12
30 40
40 60
60
10
80
80
120
120
1
x4
13 13
1
x5
1
x6
20
1
x7
26 23
40
320
320
480
160
160 240
240
If an represents the coefficient of x1n and bn represents the leftmost entry in the row which
begins two columns to the left of the column under x1n , then the method of calculation
2bn
shows that bn+2 = 2bn and that bn = 6an . But then an+2 = bn+2
6 = 6 = 2an .
8. The calculation is as follows:
10 1
4 10 0
20 2 58 75 125 96 94 140
2
0
5 5 10
2 8 25 25 96 94 140
8 20 20 86 94 140
20 0 66 54 140
16 4 40
4
0
8 2
50 90 20
50
50
50
50
10
90
90
90
90
10
20
20
20
20
20
9. Begin with the basic equation

(n + 1)
n
X
4
n
X
5
n
X
p
X
4
i=1
i=1
p=1
i=1
i =
i +
Given the result for the sum of fourth powers, we rewrite this in the form
n
X
i=1
i = (n + 1)
n
X
i=1
n
X
p=1
p
p5 p4 p3
+
+
.
5
2
3
30
39
Name:Stanislav Shur
IP:74.88.120.168
Therefore

n
n
n
n
6X
1 X
1X
1 X
5
4
3
i = n+
i
i +
i
5 i=1
2 i=1
3 i=1
30 i=1

n
n5 n4 n3
1
1
+
+
= n+
2
5
2
3
30
3
6
5
4
2
n
3n
n
n
=
+
+
5
5
2
10
n4 n3 n2
+
+
4
2
4
1
+
30
n2 n .
+
2
2
If we multiply through by 56 , we get the final result:

n
X
1
1
5
1
i 5 = n6 + n5 + n4 n2 .
6
2
12
12
i=1
10.
n1
X
2 2
(n 2n i + i ) = (n 1)n 2n
i=1
n
n5 n4 n3
n3 n2 n
+
3
2
6
5
2
3
30
n
7
n4
4
= (n 1)n (n 1)n +
15
30 30
1 4
1
8
4
= (n 1)n + n n
15
30
30
8
8 4
1 4
1
1
1
8 5
= n n + n n = n n4 n4 n.
15
15
30
30
15
2
30
4
11. We first note that the result is true for n = 1, for in that case the left side equals 2 1k
while the right side is 1k+1 + 1k . Now let us assume the result is true for n. For n + 1,
we get
(n + 2)
n+1
X
i=1
ik = ((n + 1) + 1)
= (n + 1)
n
X
k
X
i=1
ik + (n + 1)k
ik + (n + 1)k+1 +
i=1
n
X
i=1
n+1
X
i=1
ik+1 +
ik+1 +
n
X
p
X
ik + (n + 1)k
i=1
ik + (n + 1)k+1 +
p=1 i=1
p
n+1
X X
p=1
n
X
n+1
X
ik
i=1
ik .
i=1
Thus the result is true for n + 1 and is true for all n by mathematical induction.
x2
and substitute
c
x4
d
2
c x + c x = 0.
12. To solve x3 +d = cx, rewrite the equation of the given parabola as y =

into the equation y 2 x2 + dc x = 0 of the hyperbola. The result is
40
Name:Stanislav Shur
IP:74.88.120.168
If we multiply by c and divide by x, we get x3 cx + d = 0, an equation equivalent

to our original one. To sketch the two curves, note that the parabola has vertex at the
d
, 0), vertex of the right-hand branch (the
origin, while the hyperbola has center at ( 2c
d
d
only relevant one) at ( c , 0), and asymptote the line y = x 2c
. If one takes c = d = 2,
then the parabola does not intersect the asymptote, so cannot intersect the
hyperbola.
If one takes c = 3, d = 2, then the parabola and hyperbola intersect at (1, 33 ) and have
the same tangent line there. Therefore the curves are tangent there and that point is
the only intersection point. If one takes c = 4, d = 2, one can check by using a graphing
calculator that the curves intersect twice, once between x = 12 and x = 34 and once
between x = 1 and x = 2.
2
13. To solve x3 +d = bx2 , substitute y = xd into y 2 +dxdb = 0. The result is xd2 +dxdb = 0.
If one multiplies by x2 and divides by d, the result is d + x3 bx2 = 0, an equation
equivalent to the original one. The hyperbola and parabola intersect exactly once when
they have identical tangent lines at the intersection point (x0 , y0 ). The tangent line to
the hyperbola at that point is y = xy00 x + 2y0 , while the tangent line to the parabola
y0
d
2
0
there is y = 2yd0 x + y0 + dx
2y0 . If these lines are identical, then x0 = 2y0 or 2y0 = dx0 .
Substituting this value into the equation of the parabola and simplifying shows that
3d
x0 = 2b
3 . Then y0 = 2b . By substituting the value for x0 into the original equation,
we also get that 4b3 = 27d. For the curves to have no intersection, we must have
2
the hyperbola always above the parabola. Thus xd > db dx; xd2 > db dx; and
3
3
d > bx2 x3 for all x. But the maximum of bx2 x3 is 4b
27 . So 4b < 27d. The case
where the hyperbola and parabola intersect twice is then when 4b3 > 27d.
14. For three positive solutions to exist for a cubic equation written in modern terms as
x3 + qx2 + rx + t = 0, the left side of this equation must factor as (x m)(x n)(x p),
where m, n, p are all positive. Expanding this factored form, we get x3 (m + n +
p)x2 + (mn + mp + np)x mnp. Therefore, the coefficient of x is positive, while the
coefficient of x2 and the constant term must be negative. Writing this in al-Khayyamis
terms, we get the form x3 + cx = bx2 + d. Now to determine the conditions under
which this type of equation will in fact have three positive solutions, we rewrite it in
that
the form x3 bx2 + cx = d and call the left side of this equation f (x). We note
b
b2 3c
0
2
f (x) = 3x 2bx + c, and this derivative is 0 at the two critical values x = 3 3 .
For three positive solutions to exist, y = f (x) must cross the line y = d three times in
the first quadrant. A consideration of the graph of f (x) shows that it always crosses
that line at least once. For it to cross three times, the derivative must in fact equal 0
twice; thus b2 3c 0. In addition, the value of f (x) at the leftmost of the two critical
values, call it x1 , must be greater than d, that is, f (x1 ) > d.
15. If y = bx2 x3 , then y 0 = 2bx 3x2 and y 0 = 0 when x = 0 or when x = 2b
3 . The second
2b
derivative test shows that x0 = 3 makes y maximal. If we now consider the graph of
f (x) = x3 bx2 + d, we note that it has a maximum at 0 (and f (0) = d) and a minimum
41
Name:Stanislav Shur
IP:74.88.120.168
2b
4b
at 2b
3 (and f ( 3 ) = 27 + d). For this graph to cross the x-axis twice for x positive, this
minimum value must be negative. Thus there are two positive solutions to the cubic if
3
4b3
3
3
4b
27 + d < 0 or if 4b > 27d; there is one positive solution if 27 + d = 0 or if 4b = 27d;
3
3
and there are no positive solutions if 4b
27 + d > 0 or if 4b < 27d.
16. To solve x3 + d = cx, set f (x) = cx x3 . Then f 0 (x) = c 3x2 , which is 0 when x =
q
c
3.
c
(Of course, we only consider the positive value.) Then f ( 3c ) = 2c
3
3 is a maximum
value for f . The original
q equation then has two solutions if this value is greater than d,
one solution (at x = 3c ) if this value equals d, and no solutions if this value if less than
d. We can rewrite this condition as follows: there are two solutions if 4c3 > 27d2 ; there
is one solution if 4c3 = 27d2 ; and there are no solutions if 4c3 < 27d2 .
n
17. First, we note that Ckn+1 = Ck1
+ Ckn . For to count the number of ways to choose k
objects out of n + 1 objects, we can first count the ways to choose k objects out of n
objects, by neglecting the (n + 1)st object. Then, if we have that element as one of our
n
ways to complete the set of k objects. Now we can prove
set of k objects, there are Ck1
the result by induction on n. We know that the result is true for n = 1, because in
1
that case we must have k = 1, and then C11 = 21
1 C0 , because each side is equal to 1.
We then assume the result is true for n and show it is true for n + 1. By the induction
n+1 n
n
n + Cn
= nk+1
Ck1
hypothesis and the addition rule, Ckn+1 = Ckn + Ck1
k1 = k Ck1 .
k
n+1
k1
n+1
n + Cn
n
n
n
But also Ck1
= Ck1
k2 = Ck1 + nk+2 Ck1 = nk+2 Ck1 . We can rewrite this
n+1
n
= nk+2
last equation as Ck1
n+1 Ck1 . If we then substitute this result in the previous
equation, we get
Ckn+1 =
n+1 n
n + 1 n k + 2 n+1 n k + 2 n+1 n + 1 (k 1) n+1
Ck1 =
Ck+1 =
Ck+1 =
Ck1 .
k
k
n+1
k
k
Thus the inductive step is proved and the theorem is true by induction on n.
18. We need to solve the spherical triangle RNM with vertices at Rome, the North Pole, and
Mecca. We know from the given information that side m is 48 70 , side r is 68 150 , and angle N is 27 190 . The qibla is then angle R. From the formula of al-Battan we getcos n =
cos r cos m + sin r sin m cos N = cos(68 150 ) cos(48 70 ) + sin(68 150 ) sin(48 70 ) cos(27 190 )
= 0.8618. Therefore side n is 30 290 . We then calculate R from the law of sines:
sin R =
sin(27 190 ) sin(68 150 )

sin N sin r
=
= 0.8403.
sin n
sin(30 290 )
A quick glance at a globe should convince you that R is an obtuse angle. Therefore,
R = 122 500 . (Using al-Battans formula to find cos R will also give you this result.)
19. We need to solve the spherical triangle NP L with vertices at New York, the North Pole,
and London for the side p opposite the pole. The formula gives cos p = cos n cos ` +
42
Name:Stanislav Shur
IP:74.88.120.168
sin n sin ` cos P = cos 38 cos 49 + sin 38 sin 49 cos 74 = 0.6451. Therefore p = 49.83 . To
find the distance in miles, we divide this answer by 360 and multiply by 25,000. The
result is 3460 miles.
r
h cos
20. From Figure 7.16, we have r+h
= cos . Thus r = r cos + h cos and r = 1cos
.
To calculate r, substitute h = 652; 3, 18 and = 0 340 into the formula. The result is
13,331,731 cubits, which equals 19,997,597 feet, or 3,787 miles. The procedure is very
sensitive to a small change in the measured value of , and it is difficult to see how
can be measured with much precision.
21. Using the formula of Ptolemy, we find that in triangle CDB, tan CD/ tan B = sin BD
and tan BD/ tan C2 = sin CD. Similarly, in triangle ACD, we get tan CD/ tan A =
sin AD and tan AD/ tan C1 = sin CD. Equating the two expressions for tan CD derived from the first equations in each pair, we get tan A sin AD = tan B sin BD or
tan A/ tan B = sin BD/ sin AD. Equating the two expressions for sin CD gives us
tan BD/ tan C2 = tan AD/ tan C1 or tan C1 / tan C2 = tan AD/ tan BD.
22. We are given that AB = 60 , AC = 75 , and BC = 31 . Since AD and AE are
quadrants, we know that BD = 30 and CE = 15 . By the rule of four quantities,
sin CF : sin BF = sin CE : sin BD = sin 15 : sin 30 = .5176. Since CF = BF 31 ,
we have that .5176 sin BF = sin CF = sin(BF 31). Therefore, .5176 sin BF =
sin BF cos 31 sin 31 cos BF , or .5176 sin BF = .8571 sin BF .5150 cos BF. It follows that .5150 cos BF = .3395 sin BF , or that tan BF = 1.5169. Thus BF = 56 360
and CF = 25 360 . To find DF , we use equation 3.6. This implies that cos BF =
cos BD cos DF or that cos 56 360 = cos 30 cos DF . Then cos DF = .6356 and DF =
50 320 . Also we have cos CF = cos CE cos EF or cos 25 360 = cos 15 cos EF . Thus
cos EF = .9336 and EF = 21 . Since 6 A = arc DE, we have 6 A = 29 320 . To
find 6 C, we use the sine law: sin BC : sin A = sin AB : sin C. Thus sin C = .8280
and C = 55 590 . Similarly, from sin AC : sin B = sin BC : sin A, we calculate that
sin B = .9235 and therefore that B = 112 250 . (Note that because AC > AB, we must
have 6 B obtuse.)
23. We are given that AB = 60 , AC = 75 , and BC = 31 . Al=Battans formula gives
cos C =
cos 60 cos 31 cos 75

cos c cos a cos b
=
= 0.5591.
sin a sin b
sin 31 sin 75
Therefore C = 56 . The law of sines gives sin A = sin C sin a/ sin c = sin 56 sin 31/ sin 60
= 0.4930, so A = 29 320 . Similarly, sin B = sin C sin b/ sin c = sin 56 sin 75/ sin 60 =
0.9247. Since B must be obtuse, we find that B = 112 230 .
24. Given that A = 75 , B = 80 , and C = 85 , we find that MN = 180 A = 105 ,
LN = 180 B = 100 , and ML = 180 C = 95 . Although these sides are all greater
than quadrants, the method described in the text can be adapted to solve triangle
43
Name:Stanislav Shur
IP:74.88.120.168
LMN . The results are that L = 106.21, M = 101.77 , and N = 97.97 . Then
BC = 180 L = 73.79, AB = 180 N = 82.03 , and AC = 180 M = 78.23
25. For the first step, we set y1 =
q
p
= a. Since q = 47, 6; . . . and p = 45, 0, we get that a = 1;

q+y13
p , and, if we substitute
= qp+1
= 2,7;8,29,53,
= 0; 2, . . . Thus
p
45,0
3
q+(a+b)3
2
set y3 = a + b + c = q+y
=
.
p
p
that is, y1 = 1. For the second step, y2 = a + b. So y2 =

for y1 and y2 , we get the equation b =
qap+a3
p
b = 0; 2, and y2 = 1; 2. For the third step, we

Therefore,
c=
q (a + b)p + (a + b)3
47, 6; 8, 29, 53, 37, 3, 45 46, 30 + 1; 6, 12, 48
=
p
45, 0
37; 14, 42, 41, 37, 3, 45
= 0; 0, 49, . . .
=
45, 0
Thus c = 0; 0, 49 and y3 = 1; 2, 49. For the fourth step, we make a similar calculation
3
3
after setting y4 = a + b + c + d = q+y
p . The equation for d is
q p(a + b + c) + (a + b + c)3
p
47, 6; 8, 29, 53, 37, 3, 45 (45, 0)(1; 2, 49) + (1; 2, 49)3
.
=
45, 0
d=
It follows that d = 0; 0, 0, 43 and y4 = 1; 2, 49, 43.
44
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER EIGHT
1. Since there are 7200 sextarii in the cask, and since one-third of that amount is 2400, it
follows that 2400 + 6(200) = 3600 sextarii flow in through the first pipe. Then 2400
sextarii flow in through the second pipe and 1200 through the third.
2. One solution is to have the man and the goat go across first, then have the man return.
On his second trip, the man takes the cabbage and returns with the goat. On the third
trip, he takes the wolf across and returns alone. Finally, he brings the goat across again.
3. Since the ratio of 28 to 10 12 equals 8 to 3, a chord of length 6 in a circle of diameter 10 12
corresponds to a chord of length 16 in a circle of diameter 28. According to Abrahams
table, a chord of length 16 corresponds to an arc of 17 parts, 2 minutes, 16 seconds. To
convert this back to a circle of diameter 10 12 , we need to multiply this value by 38 . This
product is 6 parts, 23 minutes, 21 seconds.
5 1 6;23,21
4. The area of the sector is 4 2

= 16.772. To find the area of the corresponding
triangle, we can use the Pythagorean Theorem to find the altitude. This is 5.252 32 =
4.308. Thus the area of the triangle is 12 (4.308)(6) = 12.924 and the area of the segment
is 16.772 12.924 = 3.848.
5. An arc of length 5 12 in a circle of diameter 33 corresponds to an arc of length 4 23 = 4; 40
in a circle of diameter 28. If we use linear interpolation in Abrahams table, we find that
the ratio of 4; 40 4; 0, 55 to 5; 1, 44 4; 0, 55 is equal to 0.64. Therefore, we approximate
the corresponding chord by 4.64. Multiplying this by 33
28 gives 5.47 for the desired chord
length.
6. According to Abrahams formula, d =
s2
4h
+h=
82
8
+ 2 = 10.
7. If n is even, break this sum of sums into pairs. The first pair is 1 + (1 + 2), which, by the
previous proposition, equals 22 . Similarly, the second pair is (1 + 2 + 3) + (1 + 2 + 3 + 4),
which, by the previous proposition, equals 42 . Similarly, the next pair sums to 62 and so
on. If n is odd, then break this sum into pairs beginning with the pair (1+2)+(1+2+3).
This sum is equal to 32 . Since the first term of the sum is 12 , the entire sum is just the
sum of the squares of the odd numbers up to n.
8. In the sum (1 + 2 + 3 + + n) + (2 + 3 + + n) + (3 + + n) + + n, we have n
copies of n, n 1 copies of n 1, . . ., 2 copies of 2, and 1 copy of 1. Thus, this sum is
equal to 12 + 22 + + n2 as asserted.
45
Name:Stanislav Shur
IP:74.88.120.168
9. Each term in the left pair of square brackets sums with exactly one term in the right
pair of square brackets to the value 1 + 2 + + n. Since there are n such pairs of terms
(including one for which the corresponding term on the right is empty), the total sum is
n(1 + 2 + + n) as claimed.
10. We suppose that n is even. By exercises 7, 8, and 9, we have
n(1 + 2 + + n) = (12 + 22 + + n2 ) + (12 + 32 + + (n 1)2 ).
Also, by exercise 7, (1 + 2 + + n) = (22 + 42 + + n2 ) [1 + (1 + 2) + + (1 +
2 + + (n 1))] = (22 + 42 + + n2 ) (12 + 32 + + (n 1)2 ). By subtracting one
copy of 1 + 2 + + n from both sides, it follows that
(n 1)(1 + 2 + + n) = 3(12 + 32 + + (n 1)2 )
or
1
(n 1)(1 + 2 + + n) = 12 + 32 + + (n 1)2 .
3
By subtracting this equation from the first equation, we get that

1
n (n 1) (1 + 2 + + n) = 12 + 22 + + n2 ,
3
as desired. A similar argument proves the result if n is odd.

11. In 15 days, the first hole empties the barrel 3 times, the second hole empties it 5 times,
the third hole empties it 18 times, while the fourth hole empties it 30 times. Thus in 15
days the barrel is empties 56 times. It follows that the barrel can be empties one time
in 15/56 of a day, or approximately 6 hours, 26 minutes.
12. By taking the least common multiple, we find that 60 dinars will buy 30 litras of the first
drug, 20 litras of the second drug, 5 litras of the third drug, and 3 litras of the fourth
drug.
13. If we start with 1, we then get 1 1 + 1 3 = 4, then 1 1 + 2 3 + 1 9 = 16, then
1 1 + 3 3 + 3 9 + 1 27 = 64, and so on. In general, suppose we have a geometric series
1, a, a2 , a3 , . . .. If we put this series into the Pascal triangle, we get, first, 1 1 = 1, then
11+1a = 1+a, then 11+2a+1a2 = (1+a)2 , then 11+3a+3a2 +1a3 = (1+a)3 ,
and so on. In other words, the generalization of Jordanus formula is simply the binomial
theorem.
14. If we begin with the geometric series a4 , a3 , a2 , a, 1, we get, using Jordanus method, the
series a4 , a4 + a3 = a3 (a + 1), a4 + 2a3 + a2 = a2 (a + 1)2 , a4 + 3a3 + 3a2 + a = a(a + 1)3 ,
and finally a4 + 4a3 + 6a2 + 4a + 1 = (a + 1)4 . These terms form a geometric series with
common ratio (a + 1) : a.
46
Name:Stanislav Shur
IP:74.88.120.168
15. Every day the lion gains 1/7 1/9 = 2/63 of a foot. If we divide 50 by 2/63, we get 1575
days, which is the answer given by Leonardo. However, we can calculate that after 1571
2
= 49 55
days, the lion has climbed 1571 63
63 feet. Therefore, since on day 1572 he will
9
climb an additional 63 foot, he will have reached the top and be out of the hole before
evening.
16. We prove this result by induction. It is clearly true for n = 2. Suppose it is true for
2 +F
k = n1. We prove the result for k = n: Fn1 Fn+1 = Fn1 (Fn1 +Fn ) = Fn1
n1
n1
n1
= Fn2 (1)n .
Fn = (Fn Fn2 + (1) ) + Fn1 Fn = (Fn2 + Fn1 )Fn + (1)
For the second part, let
Fn
.
x = lim
n Fn1
Note that
Fn+1 Fn1
Fn+1 Fn1
=
= 1.
Fn
Fn
Fn
Taking limits in
this equation, we have x
finally, x =
1+ 5
2 ,
1
x
= 1, or x2 1 = x, or x2 x 1 = 0, or,
as desired.
17. We have x + y = 9, xy + x y = 21. A straightforward modern solution would start

by setting y = 9 x and substituting. The second equation then becomes, after some
algebraic manipulation, x2 11x + 30 = 0. This equation has two solutions, x = 6
and x = 5. Thus there are two answers to the problem: x = 6, y = 3; and x = 5,
y = 4. A solution more in keeping with Jordanus techniques would be to notice that
(x+y)2 4(xy+xy)+4 = x2 +2xy+y 2 4xy4(xy)+4 = x2 2xy+y 2 4(xy)+4 =
2
2
2
(x y)2 4(x
y) + 4 = [(x y) 2] . It follows that [(x y) 2] = a 4b + 4 or that
x y = 2 a2 4b + 4. With x y and x + y known, one can solve for
x and y as in
Jordanus Proposition I1. In this particular case, we have x y = 2 81 84 + 4 =
2 1. When x y = 3, we get x = 6, y = 3. When x y = 1, we get x = 5, y = 4.
18. Suppose x + y = 10, x3 + y2 = 4. If we multiply the second equation by 6, we have
2x + 3y = 24. Multiplying the first equation by 2 and subtracting from the new one, we
then get y = 4 and therefore x = 6. In the general case, multiply the second equation
by bc and the first one by c.
19. If x+y = 9 and x2 y 2 = 324, then xy = 18 and 4xy = 72. Since (x+y)2 = x2 +2xy +y 2 =
81, it then follows that (x y)2 = x2 2xy + y 2 = (x + y)2 4xy = 9. Therefore
x y = 3 and x = 6, y = 3. In general, if we have x + y =
a, x2 y 2 = b, we note that
q
(x y)2 = (x + y)2 4xy = a2 4 b. Therefore x y = a2 4 b, and we

can use
Proposition I1. In this particular case, the formula gives us xy = 81 72 = 9 = 3,

which, together with x + y = 9, gives us x = 6, y = 3.
47
Name:Stanislav Shur
IP:74.88.120.168
20. The distances covered in each of n equal subintervals are in the ratio 1 : 3 : 5 : 7 : :
(2n 1). One can see this by extending the diagram in Figure 8.8. The region over
the third subinterval will have area equal to 5 of the triangles AED, the region over the
fourth subinterval will have areal equal to 7 of these triangles, and so on. One can see this
another way by simply noting that the sum of the odd integers 1+3+5+ +(2n1) = n2
and the total distance traveled at the end of the given time is as the square of the time.
48
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER NINE
1. The traveler from Rome to Venice was traveling at the rate of 250
7 miles per day, while
250
the traveler from Venice to Rome was traveling at the rate of 9 miles per day. If t is
250
the number of days until they meet, then we have the equation 250
7 t + 9 t = 250, or
2250t + 1750t = 15750, or 4000t = 15750. It follows that t = 3 15
16 days. Thus the traveler
5
from Rome will have gone 140 8 miles while the traveler from Venice will have gone 109 38
miles.
2. If we let x be the investment of the third partner, then, since each partner should receive
a proportionate share of the profits, we get the equation
58
368 = 86.
58 + 87 + x
8
This equation reduces to 21344 = 86(145 + x) or 86x = 8874. The solution is x = 103 43
ducats. The second partner will then receive
87
4 368 = 129 ducats.
58 + 87 + 103 43
The third partner will receive the difference between 368 and the sum of 129 and 86.
Thus, he receives 368 215 = 153 ducats.
3. The second and third together can complete 6 jobs in 60 days; the first and third together
can complete 5 jobs in 60 days; the first and second together can complete 4 jobs in 60
days. By adding these results, we see that all three (doubled) can do 15 jobs in 60 days,
or that the first, second and third working together can do 7 12 jobs in 60 days. Since the
second and third can complete 6 jobs in this time, the first worker can do 1 12 jobs in 60
days, or 1 job in 40 days. Similarly, the second worker can do 2 12 jobs in 60 days, or 1
job in 24 days. And finally, the third worker can do 3 12 jobs in 60 days, or 1 job in 17 17
days.
4. Consider what happens in 60 hours. The first outlet fills the basin 30 times; the second
outlet fills it 20 times; and the third outlet fills it 15 times. Thus in 60 hours, the basin
is filled 65 times. Similarly, the first lower outlet empties the basin 20 times; the second
outlet empties it 15 times; and the third outlet empties it 12 times. Thus in 60 hours,
the basin is emptied 47 times. If all outlets are open, in 60 hours the basin is filled
1
65 47 = 18 times. Thus it is filled 1 time in 60
18 = 3 3 hours.
5. If we set x as the monthly interest rate in denarii per lire, then the problem leads to the
equation

x
100 1 +
20
4
= 160
or
4x3
x4
x 6x2
+
+
100 1 + +
5 400 8000 16000
= 160
49
Name:Stanislav Shur
IP:74.88.120.168
or, finally, to x4 + 80x3 + 2400x2 + 32000x

= 96000. If we reduce the original equation
4
x 4
x
to (1 + 20 ) = 1.6, we get 1 + 20 = 1.6 = 1.1248 or x = 2.494. On the other hand,
if we compare the fourth degree equation x4 + bx3 + cx2 + dx = e to (x + a)4 = e,
or x4 + 4ax3 + 6a2 x2 + 4a3 x + a4 = e, we note that this can be solved by completing
the fourth power provided that 4a = b,q6a2 = c, and 4a3 = d. If this is true, then
a3 = d4 , and a = 4b , so a2 = db and a = db . If we then add a4 = ( db )2 to both sides of
the original equation, our result is (x + a)4 = ( db )2 + e. Thus x + a =
x=
q
4
( db )2 + e
d
b
q
4
( db )2 + e and
as claimed.
2
6. If we let the two numbers be x and 10 x, the equation becomes 10xx
=
18 or
2x10
2
10x x = 18(2x
can be rewritten as a standard quadratic equation,
10). This
whose solution is x = 43 + 5 18. On the other hand, if we square both sides, we get
the equation 100x2 20x3 +x4 = 18(4x2 40x+100), which reduces to 720x+28x2 +x4 =
1800 + 20x3 . If we compare this to the general equation ax + bx2 + cx4 = d + ex3 and
use Pieros formula
v
!2
u
r
u
b
a
e
d
4
t
x=
,
+ +
4c
c 4c
2e
we get x = 4 72 + 1800 + 5 18 = 43 + 5 18, the correct solution.

7. Pacioli has combined the terms 79x + 30 into the single term 109x. He then solved the
cubic equation 6x3 = 43x2 + 109x, or, dividing by 6, x3 = 7 16 x2 + 18 16 x. Given that we
can divide by x, this reduces to the quadratic equation x2 = 7 16 x + 18 16 . Pacioli solves
this equation by use of the standard method, but its solution is not a solution of the
original cubic.
. Therefore, we must
8. The sum of the integers from 1 to n may be expressed as n(n+1)
2
n(n+1)
find n which satisfies the equation
= 3240. This equation can be reduced to
2
2
2
n + n = 6480, or n + n 6480 = 0. The left side factors as (n 80)(n + 81). It follows
that the (positive) solution to the equation is n = 80, and it takes 80 days to pay off the
debt.
+
8
2 or x2 + 13 +
9. The equation is x(10 x) = 13 + 128. This reduces to 10x x2 = 13
q
8 2 = 10x. The quadratic formula in this case gives us x = 5 25 (13 + 8 2) =

q
q
5 12 8 2 = 5 2 3 2 2. These two numbers are the two desired parts of 10.

10. At the first stage, we add just the first odd number. We then skip 1 odd number (where
1 is the first triangular number). At the second stage, we add 4 = 22 odd numbers
and then skip 3 odd numbers (where 3 is the second triangular number). At the third
stage we add 9 = 32 odd numbers and then skip 6 odd numbers (where 6 is the third
triangular number). Therefore, we are always adding up numbers of the form 2n 1.
50
Name:Stanislav Shur
IP:74.88.120.168
We must determine at each stage, exactly which numbers we are adding. At the second
stage we add odd numbers 1 + 12 + 1(1+1)
= 3 to 1 + 12 + 1(1+1)
+ 22 1 = 6. At
2
2
+ 22 + 2(2+1)
= 10 to
the third stage, we add together odd numbers 1 + 12 + 1(1+1)
2
2
1(1+1)
2(2+1)
1 + 12 + 2 + 22 + 2 + 32 1 = 18. In general, at stage k 2, we begin with the
mth odd number, where
m=1+
k1
X
i=1
=1+
i +
k
X
i=2
(k 1)k(2k 1)
i
k+1
+
=1+
6
2
3
k3 k2 + 2
(k 1)k(2k 1) (k + 1)k(k 1)
+
=
,
6
6
2
k3 +k2
2 .
and conclude with the number m + k 2 1 =

given by
Sk =
k3 +k2
2
X
Therefore, the sum at stage k is
(2n 1),
3
2
n= k k2 +2
and we must prove that Sk = k 5 . We calculate:

k3 +k2
2
X
Sk = 2
n=
k2
n k2 = 2
2
k3 k2 +2
k3 k2 + 2 k3 + k2
+
2
2
k 2 = k 2 (k 3 + 1) k 2 = k 5
as claimed.
11. The fourth root of 10,556,001 must be a two-digit number beginning with 5, because
504 = 6, 250, 000 and 604 = 12, 960, 000. So, subtract 6,250,000 from the original number
to get remainder 4,306,001. We then guess that the next digit is 7. To check, we first
subtract 4 503 7 = 3, 500, 000 from this remainder to get 806,001. Next, we subtract
6 502 72 = 735, 000 to get 71,001. Next, we subtract 4 50 73 = 68, 600 to get 2401.
Finally, we note that since 74 = 2401, when we subtract this value we get 0. Therefore,
57 is the desired fourth root.
12. We need to sum the arithmetic progression 1 12 + (1 12 + 16 ) + (1 12 + 26 ) = + (1 12 + n1
6 )
1
n n1
and set this equal to 2955. The sum of the arithmetic progression is n 1 2 + 2 6 =
3n
2
2
+ n 12n = n +17n
12 . We therefore get the quadratic equation n + 17n = 35, 460. We
can solve this by the quadratic formula to get n = 180 days.
13. Let d be the number of dukes, e be the number of earls, and s be the number of soldiers.
Since each duke has under him twice as many earls as there are dukes, each duke has
under him 2d earls. Since there are d dukes, there are altogether e = 2d2 earls. Since
each earl has under him four times as many soldiers as there are dukes, each earl has
under him 4d soldiers. Since there are 2d2 earls, there are altogether s = 8d3 soldiers.
51
Name:Stanislav Shur
IP:74.88.120.168
s
But we also know that 200
= 9d, so that s = 1800d. Combining these last two equations
3
2
gives us 1800d = 8d or d = 225 or d = 15. It then follows that e = 2d2 = 450 and
s = 1800d = 27000.
14. We are given that r3 + d = cr and that s3 + d = cs. Also, since the sum of the three roots
must be 0 (the coefficient of the x2 term) and their product is d, it follows that the
third root is (r +s) and that rs(r +s) = d. Thus, t3 = (r +s)3 = r3 +3r2 s+3rs2 +s3 =
cr d + 3rs(r + s) + cs d = c(r + s) + 3d 2d = c(r + s) + d = ct + d, and t is a root
of x3 = cx + d.
15. We are given that t3 = ct + d. Then if r =
3
t
r +d=
2
3
2
t
+3
2
2
t
c3
2
2 !
t
+ c3
2
3t
= t3 + c +
2
2
t
c3
2
t
2
c 3( 2t )2 , we have
"
t
+3
2
c3
c3
2
t
2
2 #
t
2
+d
t
+ t3 ct = c +
2
2
t
c3
2
= cr.
3
A similar argument holds for s. To solve x3 + 3 = 8x, note that 3 is a root of
q x = 8x + 3.
Applying the formula, we have that two roots of x3 + 3 = 8x are 32 8 3( 32 )2 =
3
2
3 5
8 27
4 =
2 . Since the sum of all three roots is 0, and the sum of these two is
3, we know that the third root is 3.
16. The function y = x3 + cx d crosses the y-axis at y = d. Since y 0 = 3x2 + c, we know

that y 0 is never 0 and that the graph of y is always increasing. Thus, it only crosses the
x-axis once, and that must be when x is positive.
3x = 10. Here
17. We apply
formula to x3 +q
q c = 3 and d = 10. We get
q Cardanos q
3
3
3
3
x=
25 + 1 + 5
25 + 1 5 =
26 + 5
26 5.
18. q
We apply Cardanos
formula to x3 = 6x + 6. Here c = 6 and d = 6. We get x =
q
3
3 + 9 8 + 3 3 9 8 = 3 4 3 2.
that y 0 = 3x2 c. It follows that
19. Consider the graph of y = x3 cx d. We calculate
q
q
the graph has a local maximum when x = 3c and a local minimum when x = 3c .
Since the graph crosses the y axis when y = d, there must be a real positive solution of
x3 cx d = 0. There will be two negative solutions
provided that theqy-coordinate
of
q
q
c
c c
c
the local maximum is positive. But when x = 3 , we get that y = 3 3 + c 3 d =
52
Name:Stanislav Shur
2c
3
c
3
IP:74.88.120.168
d. This value is positive provided that
squaring both sides, if
( 3c )3
>
2c
3
c
3
> d, or, dividing by 2 and then
( d2 )2 .
20. To solve x4 + 4x + 8 = 10x2 , we rewrite this as x4 = 10x2 4x 8 and add 2bx2 + b2 to

both sides, where b is to be determined. The result is x4 2bx2 + b2 = (10 2b)x2 4x
8+ b2 . The left side is now the square of x2 b. For the right side to be a square, we need
(4)2 = 4(10 2b)(b2 8), or 16 = 8b3 + 40b2 + 64b 320, or 8b3 40b2 64b + 336 = 0.
We can simplify this equation to b3 5b2 8b+42 = 0. One solution to this cubic is b = 3.
To determine the others, we can divide
the polynomial by x 3 to get x2 2x 14. The
roots of this polynomial are 1 15. If b = 3, the fourth degree polynomial equation
, or
becomes x4 6x2 + 9 = 4x2 4x + 1. We simplify this to (x2 3)2 = (2x 1)2
x2 3 = (2x1). Using the plus sign gives us x2 2x2 = 0, whose roots are
x
=
1
3.
2
Using the minus sign gives us x + 2x 4 = 0, whose roots are x = 1 5. It is not
difficult to check that
these four quantities are all roots of the original equation. Suppose
we substitute
this value into
our fourth degree polynomial
we choose b = 1 + 15. If
+ 2 15)x2 + (1
+ 15)2 = (8 2 15)x2 4x + 8 + 2 15. This
equation, we get x4 (2
2
15))2
= (8 2 15)(x
simplifies to (x2 (1 +
(4 + 15)) . Taking square roots, we
2
3)(x
plus sign, we can rewrite
get x (1 + 15) = ( 5
(4 + 15)).
Using
the
2
this equation in the form x ( 5 3)x 1 + 3 + 5 15 = 0. The solution of
this equation is
q

1
( 5 3) 8 2 15 + 4 4 3 4 5 + 4 15
2
q

1 h
i
1
=
5 3 12 4 3 4 5 + 2 15 =
5 3 (2 5 3)
2
2
=1 3
or
1 + 5.
x=
Note that these roots are two of the roots found earlier.
If we use the minus sign, we get
the other two roots. Similarly, if we choose b = 1 15, we get the same four roots as
before. Thus there are precisely four solutions to our original quartic equation.
21. The Cardano formula for x3 + 3x = 36 gives us
x=
rq
3
182
+ 13
+ 18
rq
3
182
+ 13
18 =
q
3
325 + 18
q
3
325 18.
q
q
3
3
325 + 18 = b + a. Then
325 18 =
To show that this value is in fact 3, set
2
b a. If we multiply these two expressions together, we get 1 = b a . If we cube
the first expression and set equal the parts not having roots, we get 18 = 3ba + a3 . A
solution
to this system
is a = 32 , b = 13
4 . Thus the solution to our original equation is
q
q
x=(
13
4
+ 32 ) (
13
4
32 ) = 3, as desired.
22. If AE = B and A : E = S : R, then R : S = E : A = (B : A) : A = B : A2 and

S : R = A : E = (B : E) : E = B : E 2 . It follows that A2 = BS : R and E 2 = BR : S.
53
Name:Stanislav Shur
IP:74.88.120.168
In the case where B = 20, R = 1, S = 5, we get A2 = 20 5 : 1 = 100, so A = 10; and

E 2 = 20 1 : 5 = 4, So E = 2.
23. Let the two numbers be s and t. Then s t = B, s3 t3 = D, and s + t = E, where E
is unknown. We calculate that
4D B 3
4(s3 t3 ) (s t)3
(s t)[4(s2 + st + t2 ) (s2 2st + t2 )]
=
=
3B
3(s t)
3(s t)
2
2
3(s + 2st + t )
=
= (s + t)2 = E 2 .
3
Since E 2 is known, so is E. And since the sum E and difference B of the two numbers is
known, the two numbers can be easily calculated. In the case where B = 6 and D = 504,
= 100. Therefore, E = 10. Since the sum of the two numbers is
we have E 2 = 2016216
18
10 and the difference is 6, the two numbers are s = 8, t = 2.
25. Let us assume, as in Figure 9.1, that d = 3.5b. If one draws a pole of height p at E, then
one of height x1 at W1 (the intersection of line EV with the first line above the ground
line), then one of height x2 at W2 (the intersection with the second line), and so on, the
line connecting the tops of the poles also intersects V . Now draw the perpendicular from
V to the ground line, letting Y be its intersection with the ground line, T1 its intersection
with the first line above the ground line, T2 its intersection with the second line above the
ground line, and so on. Then by similarity, x1 : p = V W1 : V E = V T1 : V Y . Because
the first line above the ground line has the equation y = hb/(d + b) = h/4.5, we have
x1 : p = (3.5h/4.5) : h = 7 : 9. Therefore, x1 = 79 p. Similarly, x2 : p = V W2 : V E =
V T2 : V Y . We calculate that the second line above the ground line has the equation
y = 2bh/(d + 2b) = 2h/5.5. Therefore x2 : p = (3.5h/5.5) : h = 7 : 11. Therefore
7
7
7
x2 = 11
p. Similarly, x3 = 13
p, x4 = 15
p, and so on.
26. The construction shows that the difference between the distances of the intersection of
the two chords to the two vertices is a constant, namely, the difference between BC and
AC. The modern definition of a hyperbola is, in fact, the locus of points such that the
difference in their distances to two fixed points is a constant. Thus, Keplers construction
does give a hyperbola.
27. Given the calculations in the text, we need to solve the equation
(10 x)2 + 25
9
= .
2
(10 + x) + 25
25
= 0.
This equation reduces to 9x2 +180x+1125 = 25x2 500x+3125, or to 2x2 85x+250
85 5225
= 3.179.
The solution to this quadratic which makes sense in the problem is x =
4
5AB
Thus AB = 52 + 6.8212 = 8.46 and AG = 3 = 14.10.
54
Name:Stanislav Shur
IP:74.88.120.168
3B
28. Since A : B = 10 : 7, B : C = 7 : 3, and A + B + C = 180, we have A = 10B
7 , C = 7 ,
and 20B
7 = 180. The solution is B = 63 , so C = 27 and A = 90 . So if side b = 1, we
have side a = sec 27 = 1.12 and side c = tan 27 = 0.51.
29. We are given that + = and sin

sin = k. We want to find and . From the first
equation, we get sin = sin( ) = sin cos cos sin . Substituting this in the
second equation gives
sin cos cos sin

=k
sin
or
sin cot cos = k.
Therefore,
cot =
k + cos
sin
or
tan =
sin
.
k + cos
We can thus find and therefore . In the example, we are given that = 40 and
k = 74 . Therefore,
sin 40
tan =
= 0.2555,
1.75 + cos 40
and = 14 200 . It follows that = 25 400 .
30. Since this is a right triangle, we can use the formula cos B = sin A cos b to find b. We
get cos 70 = sin 50 cos b, so cos b = 0.4465 and b = 63.48 = 63 290 . By the law of sines,
sin a = sin A sin b/ sin B = sin 50 sin 63.48/ sin 70 = 0.7294, and a = 46.84 = 46 500 .
Again by the law of sines, sin c = sin a/ sin A = sin 46.84/ sin 50 = 0.9522 and c =
72.22 = 72 130 .
31. Recall that versin A = R R cos A, and note that in the versine formula, the sines are
in circles of radius R. It follows that we can rewrite the versine formula as
R R cos A
R2
=
(R R cos a) (R r cos(b c))
R sin bR sin c
or
1
1 cos A
=
.
cos a + cos b cos c + sin b sin c
sin b sin c
This equation is equivalent to sin b sin c sin b sin c cos A = cos a +cos b cos c+sin b sin c
or to cos a = cos b cos c + sin b sin c cos A.
32. We have cos( ) cos( + ) = cos cos + sin sin cos cos + sin sin =
2 sin sin .
33. Let us suppose that the radius of the circle in which the sines are calculated is 10,000,000.
Then we need to find such that sin = 2, 189, 109 in a circle of radius 10,000,000. This
55
Name:Stanislav Shur
IP:74.88.120.168
is equivalent, in modern terms, to determining such that sin = 0.2189109. By the

calculator, = 12 380 4200 . Then = 14 360 4000 and + = 39 540 400 . So
4, 378, 218 sin(27 150 2200 ) = cos( ) cos( + ) = cos(14 360 4000 ) cos(39 540 400 ) =
9, 676, 6037, 671, 527 = 2, 005, 076. If we do the actual multiplication, we get 2,005,087.
34. According to the text, Nlog x = r ln xr , where r is Napiers radius of 10,000,000. Therefore
r
Nlog xy = r ln xy
= r(ln r ln x ln y + ln r) r ln r = r ln xr + r ln yr r ln r = Nlog x +
r
= r(ln r ln x + ln y ln r) + r ln r =
Nlog y Nlog 1. Similarly, Nlog xy = r ln x/y
r
r
r ln x r ln y + r ln r = Nlog x Nlog y + Nlog 1.
35. By Keplers third law, the ratio of the period of Mars to the period of the earth equals
the 3/2 power of the ratio of their mean distances. Since the latter ratio is 1.524, we
calculate (1.524)3/2 = 1.88. So the period of Mars is 1.88 years = 687 days.
36. Since equal areas are swept out in equal times, the planet moves faster when it is closest
to the sun, that is, at its perigee.
37. Let us assume we have a moveable falling along a vertical line and also along an inclined
plane which makes an angle with the horizontal. Let a be the acceleration due to
gravity. Therefore, along theqvertical, the velocity v = at and the distance fallen d =
1 2
2d
2d
2
2 at . Thus t = a , or t =
a . Along the inclined plane, we know that the actual
distance traveled s, when the body is at vertical distance d from the initial point, is given
the plane is the same as the
by s = d csc . Also, by hypothesis, the velocity vp at s along
q
2d
velocity at d along the vertical. Thus vp (s) = v(d) = a a = 2da = 2as sin . But
since, in general, vp (t) = kt and s = 12 kt2 , we also have vp (s) = 2ks. It follows that
k = a sin and the velocity vp along the inclined plane is given by vp (t) = (a sin )t. Then
the distance s traveled along the plane is given by s = 12 (a sin )t2 . To determine the time
it takes to move a distance s, when the falling body has reached the pointqd = s sin , we
2d
2
just solve d csc = 12 (a sin )t2 for t. We get t2 = 2d
a csc , or t = csc a . Therefore
the ratio of the time to move a distance s along the inclined plane to the time to fall the
corresponding distance d is csc , which is exactly the ratio of the length of the plane to
the vertical distance. Since this is true for any two inclined planes, the times of descent
along two such planes of the same height are to one another as the lengths of the planes.
38. By exercise 24,
q the time of descent along the first plane, with height d1 and angle 1 ,
is t1 = csc 1 2da1 and the time of descent along the second plane, with height d2 and
q
angle 2 , is t2 = csc 2 2da2 . But the lengths of the planes are the same. That means
or that csc 1 : csc 2= d2 :
d1 . Theratio of the times of
that d1 csc 1 = d2 csc 2
descent is t1 : t2 = csc 1 d1 : csc 2 d2 = d2 d1 : d1 d2 = d2 : d1 . Thus this
ratio is inversely as the square root of the ratio of the heights of the planes.
56
Name:Stanislav Shur
IP:74.88.120.168
39. Suppose the initial velocity of the projectile is v0 . The horizontal distance x(t) is given
by x = (v0 cos )t, while the vertical distance y(t) is given by y = at2 + (v0 sin )t.
If we solve the first equation for the parameter t and substitute in the second, we get
x
2
y = a( v0 cos
) +(tan )x. Since y is a quadratic function of x, the graph of the function
is a parabola, as claimed.
40. From exercise 39, we have the equation expressing y as a function of x. The x coordinate
of the point where the projectile reaches the ground, or where y = 0, is found by solving
v02 cos2 tan
+
tan
=
0.
We
get
x
=
. If = 45 , and the distance
the equation v 2ax
2
a
cos
0
v0
traveled is 20,000, then 20, 000 = 2a
or va0 = 40, 000. Putting in this result into the
distance equation for = 30 , we get x = 40, 000 cos2 (30 ) tan(30 ) = 17, 321. We get
the same result when = 60 . We note that Galileos result is slightly in error.
41. From exercise 39, we have the equation expressing y as a function of x. This quadratic
1 2
function achieves its maximum when x = tan / v 2 2a
= 2a
v0 tan cos2 . The
cos2
0
1 2
y value at this point gives the maximum value, and that is y = 4a
v0 tan2 cos2 =
2
1 2
1 1 2
1 2
4a v0 sin . When = 45 , this expression is equal to 5000. So 4a 2 v0 = 5000 and 4a v0 =

10, 000. To find the maximum height when = 30 , we calculate y = 10, 000 sin2 (30 ) =
10, 000 14 = 2, 500. To find the maximum height when = 60 , we calculate y =
10, 000 sin2 (60 ) = 10, 000 34 = 7, 500. We note that Galileos result is slightly in error.
42. Given that d = at2 , we have that the distance fallen in 1 time interval is d = a 12 = a.
The distance fallen in the second interval is d = a 22 a 12 = 3a. The distance fallen
in the third interval is d = a 32 a 22 = 5a. In general, the distance fallen in the nth
interval is d = a n2 a (n 1)2 = (2n 1)a. Thus the ratio of the distances traveled
in equal intervals is 1 : 3 : 5 : : 2n 1.
57
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER TEN
1. We rewrite the equation x3 = 300x + 432 in the form x3 300x = 432. Thus, the first
faction is 0; the second faction is 300; and the third faction is 432. Since one solution
to the equation is x = 18, we know that the sum of the other two solutions is 18 and
their product is 24. These two
numbers satisfy the quadratic equation x2 + 18x + 24 = 0,
whose solutions are x = 9 57.
2. From the diagram, with x = F N , we have the three relations
x2 + 42 = AN 2 , (4
153
. If we substitute this last
x)2 + OC 2 = 153, and 4 : OC = AN : 153, or OC = 4 AN
equation into the second, and then use the first to evaluate AN 2 , we get the equation
(4 x)2 +
16 153
= 153.
x2 + 16
This reduces to (x2 + 16)(16 8x + x2 ) = 153x2 , or x4 8x3 121x2 128x + 256, or,
x4 = 8x3 + 121x2 + 128x 256. One finds that x = 1 and x = 16 are solutions. Since the
first faction is 8 and the last is 256, the two other solutions have their sum equal to 9
and their product equal to 16. Thus these
two solutions satisfy the quadratic equation
1
1
2
x + 16 = 9x. The solutions are 4 2 2 17.
3. We rewrite the equation xy + c = rx + sy in the form xy rx sy + c = 0, or
xy rx sy + rs rs + c = 0, or, finally, (x s)(y r) = rs c. If we let x0 = x s,
y 0 = y r, we can rewrite this equation in the form x0 y 0 = rs c, which is a hyperbola
with asymptotes the x0 and y 0 axes. In terms of the original equation, the axes of the
hyperbola are the lines y = r and x = s.
4. We add x2 to both sides of the equation b2 2x2 = 2xy +y 2 to get b2 x2 = x2 +2xy +y 2,
or b2 x2 = (x + y)2, or (x + y)2 + x2 = b2 . If we change coordinates from the original
orthogonal x and y coordinates to the new oblique coordinates x0 = x, y 0 = x + y, we
can rewrite this equation in the form x02 + y 02 = b2 , which is the equation of an ellipse.
(It is not, in general, a circle because the axes are not perpendicular to one another.)
5. By similarity, we have BE : BC = BD : AB = BD, since AB = 1. Therefore,
BE = BC BD. To construct the quotient of BE by BD, we draw those two lengths
at a convenient angle, connect DE, lay off AB = 1 along BD, and draw AC parallel
to DE. Then by similarity, BC : BE = BA : BD = 1 : BD, so BC = BE : BD as
desired.
6. First, draw a line through the center N of the circle parallel to LM , and let S be
the intersection of this line with the line MR. Connect the radius NR. Then SR2 =
N R2 NS 2 = LN 2 LM 2 = ( a2 )2 b2 . The solutions to the equation z 2 = az b2 are
58
Name:Stanislav Shur
IP:74.88.120.168
x = a2 ( a2 )2 b2 . These values are equal to LN SR = MS SR = MR or MQ,

as asserted.
7. Given the notation of p. 438 in the text, we let C = (x, y) be determined so that
CB CF = CH CD. This translates to y(ey + dek + dex) = (gy + f gl f gx)(cy + bcx)
or to ey 2 + deky + dexy = gcy 2 + gbcxy + f glcy + f glbcx f gcxy f gbcx2 . We can
rewrite this as (e gc)y 2 + (de gbc + f gc)xy + f gbcx2 + (dek + f glc)y f glbcx = 0,
which is the equation of a conic section.
8. We can rewrite the equation y 2 = cy (c/b)xy +ay ac in the form (a+c(c/b)xy)y =
ac. Therefore, if we set z = a + c (c/b)x y, the equation is of the form zy = ac,
namely, the equation of a hyperbola with asymptotes z = 0 and y = 0. In terms of
Figure 10.6, the line y = 0 is the line AK. The line z = 0 is the line whose equation is
y = a + c (c/b)x. Note that the slope of the line KN is (c/b), given that positive
y values are to the left of the line AK. Therefore, the asymptote is the line parallel to
KN which passes through the point D on AG extended at a distance NL = c from the
point G.
9. By similarity, GM : MC = CB : BL or BL = MC CB/GM = xy/2a y. Then
BK = KL BL = a xy/2a y. Note that x and y are the coordinates of the point
C on the curve CEG. But C is also on the parabola, whose equation is y 2 = ax. Since
BK is on the axis of that parabola, we can put the value of BK in for the x value. We
therefore get
xy
y =a a
2a y
2
2a2 ay xy
=a
2a y
2a3 a2 y axy
.
2a y
It follows that the equation of the curve CEG is 2ay 2 y 3 = 2a3 a2 y axy or
y 3 2ay 2 a2 y + 2a3 = axy.
10. In Figure 10.7, we see that the new axes are GC and GF . The latter one is the same
as the original y axis. The former
has been
q
rotated through an angle , whose sine is
BC
b
b 2
a2 b2
.
BG = a . Therefore, cos = 1 ( a ) =
a
11. If we set z = y + ab x + c, or y = z ab x c, and substitute this into the given equation,
we get (z ab x c)2 +
2bx
a (z
2
z =
ab x c) + 2c(z ab x c) =
!
f x2
a
+ ex + d. This reduces to
b2
2cb
f
+ e x + c2 + d,
+
x2 +
2
a
a
a
which is the equation

of a hyperbola. If we rewrite this as z 2 = rx2 + sx + t, then the
substitution x0 = rx turns this into an equation of the form z 2 = x02 + ux0 + v. Finally,
59
Name:Stanislav Shur
IP:74.88.120.168
we can substitute x00 = x0 + u2 and write this into the simpler form z 2 = x002 + m, which
is also the equation of a hyperbola.

12. The statement is true for n = 2, because 21 = 00 + 10 and 22 = 11 . Now suppose

the statement is true for n = m and let us prove it for n = m + 1. Suppose k m + 1.
Then k 1 m, and we have
!
m1
m
X
X
m+1
m
m
j
m
j
=
+
=
+
=
,
k
k
k1
k1
k1
k1
j=k1
j=k1
and the result is true for n = m + 1. Thus by induction, the result is true.

13. As Pascal noted, this result is true for n = 1, because 10 : 11 = 1 : 1. So let us

assume the result
for n = m and show it
is
true

forn = m + 1. By theinduction

m
m
m
m
m
m
hypothesis, k1 : k = k : (m k + 1). Then m+1
:
=
(
+
)
:
k
k
k
k1
k =

m
m+1
m
m+1
. Also, since m
k : k+1 = (k + 1) : (m k), we have k+1 : k =

mk+1

m
+ m
] : m = mk
+ 1 = m+1
. Therefore, by dividing the two previous equations
[ k+1
k k
k+1
k+1

m+1
m+1
m
m
m+1
m+1
k+1
k+1
we have m+1
:
=
[
:
][
:
k
k+1
k
k
k
k+1 ] = mk+1 m+1 = m+1k , and the
k
1 + mk+1
=
inductive step is proved. Thus, by mathematical induction, the result is proved for all
n.
14. This assertion is true for n = 1, because

1
0

0
0
= 1 : 1 = 1 : (1 0). So let us
assume that it is true for n = m 1 and prove it for n = m. We have

[

m1
k
m1
k1
]:
m1
k
= 1+[

m1
k1
m1
k

m
k
m1
k
]. By the result of the previous exercise, the

m1
k
k
k
m
last ratio is equal to m1(k1)
= mk
. It follows that m
= 1 + mk
= mk
,
k :
k
and the inductive step is proved. Thus, by mathematical induction, our result is true
for all n.
15. The probability of not throwing a six in a single throw is 56 . Therefore, the probability
625
of not throwing a six in any of the four throws is ( 56 )4 = 1296
. Thus, the probability of
625
671
throwing a least one six in four throws is 1 1296 = 1296 . That means that the odds in
favor of getting at least one six in four throws are 671 : 625.
16. The probability of not throwing a one in a single throw is 56 . Therefore, the probability
of not throwing a one in any of the three throws is ( 56 )3 = 125
216 . Thus, the probability of
91
getting at least one one in three throws is 1 125
=
.
Thus,
the odds against throwing
216
216
a one in three throws are 125 : 91.
17. Using Pascals theorem, with r = 3,
s = 4,and n = 6,

we see that the first player gets
6
6
6
that proportion of the stakes that 0 + 1 + 2 + 63 = 42 is to 26 = 64. Thus, the
60
Name:Stanislav Shur
IP:74.88.120.168
first player gets 42

64 of the stake and the second player gets
stake is split in the proportion 42 : 22, or 21 : 11.
22
64
of the stake; that is, the
18. At most three more games are necessary to decide the contest, so we may as well assume
that these three are played. Let us represent by A a win by the first player, by B a
win by the second player, and by C a win by the third player. There are 27 possible
outcomes for three games, where each one could result in a win for A, B, or C. These
are as follows: AAA, AAB, AAC, ABA, ABB, ABC, ACA, ACB, ACC, BAA, BAB,
BAC, BBA, BBB, BBC, BCA, BCB, BCC, CAA, CAB, CAC, CBA, CBB, CBC,
CCA, CCB, CCC. Since the first player wins whenever an A occurs before two Bs or
two Cs, and since this happens in 17 out of the 27 cases, we must give 17
27 of the stakes
to the first player. Since the portion of the stakes to be given to the other two players
5
of the stakes. That is, the stakes should be divided
must be equal, each of them gets 27
in the ratio 17 : 5 : 5.
19. We know that the probability of throwing a 7 with two dice is 16 , the probability of
5
throwing a 6 is 36
, and the probability of throwing something else if 25
36 . Thus the
1
1 25
37
expectation of the first player is 1 6 + 2 36 = 72 . The expectation of the second player
5
35
is 1 36
+ 12 25
36 = 72 .
5
20. The first player has a probability of 36
of winning on the first throw. If a 6 does not show
1
up then, which happens with probability 31
36 , then I have a probability of 6 of winning on
1
31
the second throw, for a probability of winning in the first pair of throws of 31
36 6 = 216 .
5
30
Thus in the first pair of throws, the first player has probability 36
= 216
of winning,
31
while I have probability 216 . If neither person wins on the first pair of throws, we can
simply discount that pair and start over. Thus the ratio of my probability to that of the
other player is 31 : 30.
21. The probability of A winning on the first draw is 13 . The probability of B being able
to draw is therefore 23 , so Bs probability of winning is 23 13 = 29 . The probability of C
4
being able to draw is 1 13 29 = 49 . Therefore Cs probability of winning is 49 13 = 27
.
1
2
4
Thus, the ratio of the probabilities of A to B to C is 3 to 9 to 27 , or 9 : 6 : 4.

22. We note first that if p is prime, then kp is a multiple of p for 1 k p 1. After all,
this number is an integer expressed as a fraction with p a factor of the numerator and no
factor of the denominator can divide p. Thus every term of the expansion of (1 + 1)p is
congruent to 0 modulo p except the first term and the last. These terms are both equal
to 1, so 2p = (1 + 1)p 2 (mod p). Now assume that ap a (mod p). We want to
prove that (a + 1)p a + 1 (mod p). But again because all binomial coefficients except
the first and last are divisible by p, we have (a + 1)p ap + 1 a + 1 (mod p), and
the inductive step is proved.
61
Name:Stanislav Shur
IP:74.88.120.168
23. We consider the remainders of 1, a, a2 , . . . on division by p. Since there are only p 1

possible remainders, these remainders must ultimately repeat. So for some n and r we
have an+r ar (mod p), or ar (an 1) 0 (mod p). Given that if p divides a product,
it must divide one of the factors, and given that p and a are relatively prime, it follows
that an 1 0 (mod p), or that an 1 (mod p). Let n be the smallest positive
integer satisfying this congruence. Apply the division algorithm to n and p 1. We have
p 1 = kn + s, where 0 s < n. Then 1 ap1 akn+s (an )k as as (mod p).
Thus as 1 (mod p), where 0 s < n. But n is the smallest positive integer having
this property. So s = 0 and n divides p 1.
62
Name:Stanislav Shur
IP:74.88.120.168
SOLUTIONS TO PROBLEMS
CHAPTER ELEVEN
1. Fermats initial method for determining maxima and minima for a polynomial p(x) is to
set p(x1 ) = p(x2 ), or p(x1 ) p(x2 ) = 0, divide by x1 x2 , and then set x1 = x2 . That
is, the maximum or minimum can be found by setting x1 = x2 in the quotient
p(x1 ) p(x2 )
x1 x2
and setting this result equal to 0. Given that for polynomials, x1 x2 always divides
p(x1 ) p(x2 ) without remainder, one can determine the limit as x1 approaches x2 in
the quotient by simply setting x1 equal to x2 . Thus Fermats process is equivalent to
determining
p(x1 ) p(x2 )
lim
x1x2
x1 x2
and then setting this limit equal to 0. And this is the same process as determining p0 (x)
and setting it equal to 0. The same argument holds for Fermats second method, which
is equivalent to determining p0 (x) by calculating
p(x + e) p(x)
e0
e
lim
and then setting that equal to 0, again because e will always divide the difference p(x +
e) p(x) without remainder.
2. We set bx1 x31 = bx2 x32 . This is equivalent to bx1 bx2 = x31 x32 . If we divide
2
2
x = x1 = x2 now gives us
both sides by x1
qx2 , we get b = x1 + x1 x2 + x2 . Setting
q
b
2b b
2
b = 3x , or x = 3 . Thus the maximum value is 3 3 . Since the negative solution to
the equation gives a negative value to the expression bx x3 , it is the positive solution
which gives the maximum.
3. If p0 (a) = 0, and M = p(a), then
lim
xa
p(x) p(a)
= p0 (a) = 0.
xa
p(x)p(a)
But the quotient xa is a polynomial q(x), and q(a) = 0. Therefore, q(x) = (x

a)r(x), and p(x) M = p(x) p(a) = (x a)q(x) = (x a)2 r(x).
(x+e)3
3
2
2
3
3
3
4. We equate x3 and t+e
t . After expanding, we get tx + 3tx e + 3txe + te = tx + ex .
After canceling common terms and dividing by e, we have 3tx2 + 3txe + te2 = x3 .
Elimination of all terms with e then gives us the relation 3x2 t = x3 , or, finally, t = x3 .
Name:Stanislav Shur
IP:74.88.120.168
5. If (x, y) is a point on the curve, and if (x + e, y) is a point on the tangent line, then
yy
y
t+e
e = t , or y = t y. Thus, to find the subtangent, and therefore the tangent, we must
3
3
equate f (x, y) and f (x + e, t+e
t y). In the case x + y = pxy, we first rewrite this as
t+e
3
x3 + y 3 pxy = 0, and then equate this expression with (x + e)3 + ( t+e
t y) p(x + e) t y.
By expanding, we get
pe2
e
e2
e3
pe
y.
x3 +y 3 pxy = x3 +3x2 e+3xe2 +e3 +y 3 +3 y 3 +3 2 y 3 + 3 y 3 pxy xy pey
t
t
t
t
t
If we cancel common terms, divide through by e, and then eliminate all remaining terms
3
pxy3y3
in e, we get 3x2 + 3yt pxy
t px = 0. Solving this for t gives us t = 3x2 py .
6. Fermats method for finding the subtangent t involves setting up the equality tf (x + e)
(t + e)f (x), then canceling common terms, dividing by e, and removing any remaining
terms involving e. We can rewrite this in the form t(f (x + e) f (x)) = ef(x), or
t=
ef(x)
=
f (x + e) f (x)
f (x)
f (x+e)f (x)
e
The process of canceling common terms, dividing by e, and removing any remaining
terms involving e amounts to, in modern terms, calculating the limit of this expression
as e approaches 0. Since the limit of the denominator is f 0 (x), we get that t = ff0(x)
(x) . In
e
the modified method of the previous problem, we have f (x, y) = f (x + e, y + t y). By
modern methods, the expression on the right is approximately equal to f (x, y) + f
x e +
f ey
y t . Comparing this result to f (x, y) and dividing through by e gives us the equation
f
x
f y
y t
= 0. If we solve this for t, we get t = y(f /y)/(f /x).
7. We adequate
x2
a2
y2
b2
with
(x+e)2
a2
(1+(e/t)2 )y2
.
b2
By canceling common terms, dividing
through by e, and then removing all remaining terms with e, we get

Solving for t gives us t =
2 2
ab2yx
2 2
2 2
x
a b bb
2x
2
2
a x
x .
2x
a2
(2/t)y2
b2
= 0.
8. We need to find a double root to the polynomial (x3/2 )2 + v 2 2vx + x2 n2 = x3 +

x2 2vx + v 2 n2 . Since this is a cubic polynomial, we equate it to (x x0 )2 (x + b) =
x3 + (b 2x0)x2 + (x20 2x0 b)x + bx20. Comparing coefficients of like powers of x, we must
solve the equations: b 2x0 = 1; x20 2x0 b = 2v; and bx20 = v 2 n2 . Solving the first
equation for b and putting this into the second equation gives us x20 2x0 (1+2x0 ) = 2v,
or 2v = 3x20 2x0 , or, finally, v = x0 + 32 x20 .
1
9. Rewriting y 2 = x as y = x 2 , we need to find a double root to the polynomial (x1/2 )2 +

v 2 2vx + x2 n2 = x2 + (1 2v)x + v 2 n2 . We therefore equate this to (x
x0 )2 = x2 2x0 x + x20 by comparing coefficients of like powers. We get the equation
2
Name:Stanislav Shur
IP:74.88.120.168
1 2v = 2x
, which reduces to v x0 = 12 . It follows that the slope of the normal line
0
x
y0
is vx
= 1/20 = 2 x0 . Therefore, the slope of the tangent line is 21x0 .
0
10. To apply Descartes rule, we need to find a double root of the polynomial x2n + x2
2vx + v 2 m2 . By Huddes rule, this root will also be a root of 2nx2n + 2x2 2vx
or a solution of the equation 2nx2n1 + 2x 2v = 0. Since this root is x0 , we have
0
. Since the slope of the tangent line is vx
v = x0 + nx2n1
0
xn , we calculate this to be
nx2n1
0
xn
0
nxn1
0
as desired.
11. We multiply the x3 terms by 3, the x term by 1, and set the result equal to 0. We
2
2
get (9a 3b)x3 2b3ca x = 0, or (9a 3b)x2 2b3ca = 0. Thus, the maximum occurs
r
when x =
expression.
2b2 a/3c
9a3b
and the minimum occurs when x is the positive square root of that
12. Let us divide up the sphere of radius r into infinitely many cones of altitude r, each with
their vertex at the center of the sphere. The volume of each of these cones is then 13 rA,
where A is the (infinitesimal) area of the base on the surface of the sphere. If we add
all these volumes together, we get 13 rS, where S is the total surface area of the sphere.
Since S = 4r2 , we have V = 43 r3 .
13. We rewrite the result of exercise 12 of chapter 10 as
!
n
X
n+1
j
=
.
k
k
1
j=k1
When k = 3, this formula becomes

!
n
X
n+1
j
=
.
3
j=2 2
Expanding the binomial coefficients gives

!
n
n
j(j 1) X
j2 j
(n + 1)n(n 1) X
=
=
.
6
2
2
2
j=2
j=1
(Note that we can start the sum on the right at j = 1 since the value of the right side
for j = 1 is 0.) We rewrite this equation as
n
n
X
X
(n + 1)n(n 1)
=
j 2 =
j.
3
j=1
j=1
Name:Stanislav Shur
Therefore,
n
X
j2 =
j=1
IP:74.88.120.168
n3 n2 n
n3 n n(n + 1)
+
=
+
+ .
3
2
3
2
6
When k = 4, the formula becomes

!
n
X
n+1
j
=
4
j=3 3
or
n
X
(n + 1)n(n 1)(n 2)
j(j 1)(j 2)
=
.
24
6
j=1
If we multiply by 6 and expand the right side, we get

n
n
n
X
X
(n + 1)n(n 1)(n 2) X
=
j3 3
j2 + 2
j.
4
j=1
j=1
j=1
Using the known formulas for the sums of the integers and the squares of the integers
gives us
n
X
(n + 1)n(n 1)(n 2)
n3 n2 n
n2 n
+3
+
+
+
j =
2
4
3
2
6
2
2
j=1
3
n4 n3 n2
+
+ .
4
2
4
14. The formula from exercise 12 of chapter 10 becomes

!
n
X
n+1
j
=
.
5
4
j=4
If we expand the binomial coefficients and note that we can start the sum on the right
from j = 1, we get
n
(n + 1)n(n 1)(n 2)(n 3) X
j(j 1)(j 2)(j 3)
=
.
120
24
j=1
Multiplying both sides by 24 and simplifying gives

n
n
n
n
X
X
X
X
(n + 1)n(n 1)(n 2)(n 3)
=
j4 6
j 3 + 11
j2 6
j.
5
j=1
j=1
j=1
j=1
Replacing the sums of integers, integer squares, and integer cubes by their values gives
n
X
n5
6n
n4 n3 n2
n4 + n3 + n2
+6
+
+
j =
5
5
4
2
4
j=1
4
n2 n
+
+6
2
2
n3 n2 n
+
+
11
3
2
6
n
n5 n4 n3
+
+
.
5
2
3
30
Name:Stanislav Shur
IP:74.88.120.168
15. Let the equation of the parabola be y 2 = x and the vertical line BD be x = b2 . Then
the volume of the solid formed by rotating the parabola around BD is given by
V =
Z b
0
2 2
(b y ) dy =
4
= b y
2b2 y 3
y5
5
Z b
0
!b

(b4 2b2 y 2 + y 4) dy
2b5 b5
+
= b
3
5
5
8 5
b .
15
The cone with the same base and vertex as the paraboloid has volume 13 bb4 = 13 b5 =
5
5
15 b , so the ratio of the volumes is 8 : 5 as noted by Fermat. Also, the volume of the
8
cylinder circumscribing the paraboloid and cone is b5 . Thus the paraboloid is 15
of the
cylinder, as noted by ibn al-Haytham.
16. If we divide the interval [0, x0 ] as stated, then the area of the right-most circumscribing
n
n
rectangle is R1 = (1 m
)x0 pxk0 = (1 m
)pxk+1
0 . The area of the next rectangle is
n
n 2
n
n
n k+1 k+1
n k+1
k
) R1 . Similarly, the area of the
( m ( m ) )x0 p( m x0 ) = (1 m )p( m ) x0 = ( m
n 2(k+1)
third rectangle from the right is ( m )
R1 , and so on. Thus, the area under all of the
circumscribed rectangles is given by
A = R1
"
n
1+
m
R1
1
1
n
m
k+1
n
+
m
2(k+1)
k+1
=
1
+ +

n k+1
m
=
1+
n
m
n
m
2
k+1
n
m
n
1
pxk+1
0
m
k px0
n
We get the area under the curve by allowing m
to approach 1. Each of the terms in the
1
denominator approaches 1 and the area is k+1 pxk+1
0 , as desired.
17. Recall that, according to Fermats method of calculating the subtangent, if f (x) is a
ef (x)
function, then the subtangent t is found by equating t to f (x+e)f
(x) , then simplifying.
(x)
. Thus, if we have a product of
It will be convenient to rewrite this as 1t = f (x+e)f
ef (x)
two functions, say zu, we can calculate the subtangent tzu as follows:
1
tzu
z(x + e)u(x + e) z(x)u(x)

ez(x)u(x)
z(x + e)u(x + e) z(x)u(x + e) + z(x)u(x + e) z(x)u(x)
=
ez(x)u(x)
[z(x + e) z(x)]u(x + e) + z(x)[u(x + e) u(x)]
=
ez(x)u(x)
1
1
z(x + e) z(x) u(x + e) u(x + e) u(x) z(x)
= + .
=
ez(x)
u(x)
eu(x)
z(x)
tz tu
=
Name:Stanislav Shur
IP:74.88.120.168
Therefore, given the ratio u : v = w : z, we have uz = vw and therefore tuz = tvw .

Therefore t1z + t1u = t1v + t1w , or
1
1
1
tu tw + tu tv tv tw
1
= +
=
tz
tv tw tu
tv tw tu
and
tz =
tu tv tw
tu tw + tu tv tv tw
as stated. To convert this relationship to formulas involving derivatives, recall that if

y = f (x), then y 0 = tyy . To determine the product rule, we set u = 1, so z = vw. Since
tw
tu is infinite, the rule already derived tells us that tz = tvtv+t
or that
w
then have
z
vw vw
=
+
= wv 0 + vw0 ,
z0 =
tz
tv
tw
1
tz
1
tv
1
tw .
We
the product rule. Similarly, to determine the quotient rule, we set w = 1, so z =

tv
Since tw is infinite, we get tz = tutut
or t1z = t1v t1u . We then have
v

z
v 1
1
=
z =
tz
u tv tu
0
v
u.
v 0 vu0
uv 0 vu0
2 =
,
u
u
u2
the quotient rule.

18. If y 4 = x5 , then y = x5/4 and y 0 = 54 x1/4 . The arclength L of this curve from 0 to b is
given by
s
Z bq
Z b
25
1 + y 02 dx =
1 + x1/2 dx.
L=
16
0
0
1/2 , or x1/2 = 16 (u 1), then du = 25 x1/2 dx or dx =
If we set u = 1 + 25
16 x
25
32
512
25
(u
1)
du.
If
we
set
a
=
1
+
b,
then
the
arclength
formula becomes
625
16
32 1/2
du
25 x
512 a
512 a 3/2
(u 1)u1/2 du =
(u u1/2 ) du
625 1
625 1"
!
#

2
512 2 5/2 2 3/2 a 1024 a2 a
u u
=
a
+
.
=
625 5
3
625
5
3
15
1
L=
19. If y = x2 , then y 0 = 2x. Thus, the arclength formula in this case gives us
L=
Z bq
1 + (2x)2 dx =
Z bq
1 + 4x2 dx.
Thus, to calculate this integral, one needs to find the area under the curve y =
or y 2 4x2 = 1, a hyperbola.
1 + 4x2 ,
Name:Stanislav Shur
IP:74.88.120.168
20. We replace y by y+a and x by x+e in the formula x3 +y 3 = c3 . We get (x+e)3 +(y+a)3 =
c3 , or x3 + 3x2 e + 3xe2 + e3 + y 3 + 3y 2 a + 3ya2 + a3 = c3 . We then remove all terms
containing a power of a or e at least two and also delete the terms forming the original
expression. We are left with 3x2 e + 3y 2 a = 0. We next substitute y for a and t for e and
solve for the ratio y : t. We get 3x2 t + 3y 3 = 0, and y : t = 3x2 : 3y 2 = x2 : y 2 is the
slope of the tangent line.
21. If we set BG = P M = y and MRq= HG = a, we can rewrite the proportion CL : LF =
CB : BH in the form (f + ge) : g 2 2fge = 1 : (y a). Squaring both sides gives us
f 2 + 2fge
1
= 2
,
2
g 2fge
y 2ay
after neglecting powers of e higher than the first. This equation becomes g 2 2fge =
y 2 f 2 + 2fgy 2 e 2ayf 2 . Since y = fg , or y 2f 2 = g 2 , we have 2fgy 2 e + 2fge = 2ayf 2 , or,
gy 2 e + ge = ayf . Now we substitute y for a and t for e. We get (gy 2 + g)t = y 2 f , so
f y2
g
t = g+gy
2 . Since f = y , we can rewrite this as
BG CB 2
BG CK 2
y
=
=
.
t=
1 + y2
CG2
CE 2
y
t
Since the slope of the tangent line to the curve ANMO is given by
x
2
sec2 x, we have d tan
dx = sec x.
= 1 + y 2 = CG2 =
22. The first term of the square root of 1 + x is 1. We subtract this from 1 + x to get x,
then double the 1 and divide into x. This gives us the second term of the square root,
2
namely x2 . We then multiply 2 + x2 by x2 to get x + x4 . We subtract this from x to get
2
x4 . We double the two terms we already have to get 2 + x and divide the 2 into x4 .
2
The result, x8 , is then the third term of the square root. We then multiply 2 + x
2
by x8 to get x4
x3
8
x4
64 .
three terms we already have to give us 2 + x

us
x3
16
x4
64 .
We double the
and divide the 2 into
This gives
Subtracting this from x4 gives

x2
4
x3
8
x2
8
x3
8 .
for the fourth term of the series. We continue in this way. We therefore have
1+x=1+
x x2 x3 5x4 7x5
+
+
2
8
16 128 256
1 6
23. The power series for (1 x2 )1/2 is 1 12 x2 18 x4 16
x . To square this, we need to
square each term and also take twice the product of every pair of terms. We therefore
1 8
1 8
1 10
1 12
x + + 64
x + 64
x + + 256
x + We
get 1 x2 14 x4 18 x6 + + 14 x4 + 18 x6 + 16
4
6
8
note that the x and x terms already disappear. To check the x term, we would need
5 8
to find the next term in the original power series. That term turns out to be 128
x ,
Name:Stanislav Shur
IP:74.88.120.168
5 8
which, when doubled, gives 64
x . This term, combined with the two other terms in
8
x already calculated, again give a 0 coefficient. Every term beyond the x2 term does
disappear, and the square of the power series is 1 x2 as desired.
2
.001
.0001
.00001
24. We have log(1 + x) = x x2 + x3 x4 + . So log(1.1) = 1 .01
2 + 3 4 + 5
.000001
+ .0000001
= 0.09531018. Similarly, log(0.9) = 0.10536051, log(1.2) =
6
7
0.18232156, log(0.8) = 0.22314355, log(1.01) = 0.00995033, log(0.99) = 0.01005034,
log(1.02) = 0.01980263, and log(0.98) = 0.02020271. We have log(1) = 0. Since
1.22
2 = 1.21.2
0.80.9 , we have log 2 = 2 log(1.2)log(0.8)log(0.9) = 0.69314718. Since 3 = 0.8 ,
we have log 3 = log(1.2) + log(2) log(0.8) = 1.09861229. Also, log 4 = log(22 ) =
4
) = log 4log(0.8) = 1.60943791, log 6 = log(23) =
2 log 2 = 1.38629436, log 5 = log( 0.8
log 2 + log 3 = 1.79175947, log 8 = log(23 ) = 3 log 2 = 2.07944154, log 9 = log(32 ) =
2 log 3 = 2.19722458, and
q log(10) = log(2 5) = log 2 + log 5 = 2.30258509. To get
log 7, we note that 7 = 12 100 0.98, so log 7 = 12 [ log 2 + 2 log(10) + log(0.98)] =
1.94591015.
25. If we follow Newtons method but use the progression 4, 3, 2, 1, instead of 3, 2, 1, 0, we

3
get 4x2 x 3axx + 2ay x yx x + x3 yy + 2axy 4y 2 y = 0. If we factor out the terms with
x and y respectively and divide, we get
x
x(4y 3 2axy + ax2 x3 )
=
.
y
y(4x3 3ax2 + 2axy y 3 )
Using the original equation, this simplifies to
x
x(4y 3 2axy + axy y 3)
x(3y 3 axy)
=
=
y
y(4x3 3ax2 + 2axy x3 + ax2 axy) y(3x3 2ax2 + axy)
3y 2 ax
= 2
,
3x 2ax + ay
the same value as calculated using the original progression. A similar simplification takes
place if one uses any other arithmetic progression.
26. We set z = x a2 x2 . We then have two equations to which to apply Newtons rules:
y 2 a2 z = 0 and z 2 a2 x2 + x4 = 0. The first equation becomes 2y y z = 0, so
2
3
y
1
2 + 4x3 x
= 0, so xz = 2a x4x
. Then
z = 2y . The second equation becomes 2z z 2a xx
2z
y
x
y z
z x
2a2x4x3
4xy a2 x2
2
a
2x2
.
2y a2 x2
27. If we replace x by x + 1, we get the fluxional equation

2
y
=
+ 2 x2 2x = 2 2x + 2x2 2x3 + 2x4 + 2 x2 2x
x
x+1
= 4 4x + x2 2x3 + 2x4
8
Name:Stanislav Shur
Therefore, y = 4x 2x2 +
x3
3
x4
2
28. When x becomes x + o, the value

1
x+o
1
x
x(x+o)
x(x+o)
IP:74.88.120.168
2x5
5
1
x
becomes
1
x+o .
Thus the augments are as o to
o
1
= x(x+o)
. Therefore, the ratio of the augments is as 1 to x(x+o)
,
and by letting o vanish, we find that the ratio of the fluxions is as 1 to x12 .
29. When x becomes x + o, the value
1
xn
becomes
1
(x+o)n .
Thus the augments are as o to
nxn1 o + n(n1)
xn2 o2 +
1
1
2
=
.
(x + o)n xn
xn (x + o)n
Therefore, the ratio of the augments is as 1 to (nxn1 +(n(n1)/2)xn2 o+ )/xn (x+
n1
o)n . By letting o vanish, we find that the ratio of the fluxions is as 1 to nxx2n , or as 1
n
.
to xn+1
30. From the diagram, we have CT 2 = AT 2 + AC 2 . Using the basic rules of fluxions and
= (CT : AT )CT.
, or AT
= AT AT
It z
noting that AC is fixed, we get CT CT
=
represents the arc, we have, by the result demonstrated in section 11.4.3, z : AT
= CS : CT or that z : CT
= (CT :
CS : CT . It follows that z : (CT : AT )CT
AT ) (CS : CT ) = CS : AT . In other words, the fluxion of an arc to the fluxion of
its secant is as cosine to tangent. If we invert the ratios, the result becomes, in modern
terms, that the derivative of the secant is the quotient of the tangent by the cosine, or
d(sec z)/dz = tan z/ cos z = tan z sec z.
=
31. From the geometric result AS 2 + CS 2 = AC 2 , we derive the fluxional result AS AS
Also, from the geometric result CS : AC = AC : CT , we get CS CT = AC 2,
CS CS.
CT = CT
CS, or CS
= (CT
CS : CT . We therefore have, using
and therefore CS
the result from exercise 30 (and writing everything in fractions rather than ratios):
CS
zCS
AS
zCS
= CS CS = CS CT CS = zAT
=
=
.
AS
AS
AS CT
AS CT
AS AC
AC
= AC : CS, or that the fluxion of the arc to the fluxion of the sine
It follows that z : AS
is as radius to cosine.
q dy
dx
= ypq1
. If we multiply this
xp1
q
dy
dy
py
1
equation by xy and then cancel the equal terms y q and xp , we get y = p xdx , or dx
= qx
.
dy
py
p
qp
To apply the transmutation theorem, we note that z = yx dx = yx qx = y q y = q y.
32. Given that y q = xp , we have qy q1 dy = pxp1 dx, so
Therefore,
Z x0
0
1
y dx =
x0 y0 +
2
Z x0
0
1
1
z dx = x0 y0 +
2
2
Z x0
0
qp
y dx.
q
Name:Stanislav Shur
Since 1
qp
2q
p+q
2q ,
IP:74.88.120.168
we have
p+q
2q
Z x0
0
1
y dx = x0 y0
2
or
Z x0
0
y dx =
qx0y0
.
p+q
33. We have
x
d
y
xy + y dx xy x dy
y dx x dy
x + dx x
=
=
.
y + dy
y
y(y + dy)
y2
We can ignore the y dy in the denominator because it is infinitesimally small with respect
to the y 2 term.
dy
34. Assume that y = a + bx + cx2 + dx3 + ex4 + . Then dx
= b + 2cx + 3dx2 + 4ex3 + .
1
Since x+1
= 1 x + x2 x3 + , we can compare these two power series term by term to
get, in turn, b = 1, 2c = 1 or c = 12 , 3d = 1 or d = 13 , 4e = 1 or e = 14 , and so on.
Since y = 0 when x = 0, we get the power series log(x + 1) = x 12 x2 + 13 x3 14 x4 + .
2
3
35. We assume that x = a + by + cy 2 + dy 3 + ey 4 + . Then dx
dy = b + 2cy + 3dy + 4ey + .
2
3
4
But dx
dy also is equal to x + 1 = a + 1 + by + cy + dy + ey + . Since the logarithm of
1 is 0, we know that a = 0. Then we can compare the two series term by term. We get
a + 1 = b, so b = 1. Then 2c = b, or 2c = 1, or c = 12 . Next, 3d = c, or 3d = 12 , so d = 16 .
1
Next 4e = d, or 4e = 16 , so e = 24
. Thus the power series for x + 1, the exponential
1 2
1 3
1 4
function, is x + 1 = 1 + y + 2! y + 3! y + 4!
y + .
10
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER TWELVE
1. If dy =
a dx ,
x2 a2
then
Z
a dx
.
x2 a2
To integrate this, we substitute x = a cosh u. Then dx = a sinh u du. So
y=
y=
a2 sinh u du
a2 cosh2 u a2
a2 sinh u du
=
a sinh u
a du = au.
Since u = cosh1 xa , we have y = a cosh1 xa , or ay = cosh1 xa , or cosh ya = xa , or, finally,

x = a cosh ya , the standard equation of the catenary.
2. The horizontal force at (x, y) is T (x) cos , while that at the low point of the cord is
T (0) in the opposite direction. Since these forces balance, |T (0)| = |T (x)| cos . The
downward vertical force at (x, y) is s, while the upward vertical force is |T (x) sin .
Since these balance as well, we have s = |T (x)| sin . Dividing the second equation by
the first, we get
s
sin
dy
=
= tan =
.
|T (0)|
cos
dx
It follows that dy/dx = s/a, where a = |T (0)|/.
R
n dx
3. We can rewrite dp
. Thus, if we multiply
p = n dx in the form ln p = n dx, or p = e
m dx + ny dx + dy R= 0 by p, we get pm dx + pny dx + p dy = pm dx + y dp + p dy. If we
integrate, we have pm dx + py = k, or
y=
1
k
p
pm dx = e
In this particular case, m = 3x and n =

R
1
k
2
2
x (k + 3x dx) = x + x .
n dx
1
x.
me
n dx
dx .
Therefore, p = eln x = x. So y =
4. Given that ur = i cos + j sin and u = i sin +j cos , we get dur /d = isin +jcos
=u and du /d = icos jsin = ur . Next, since r = rur , the velocity v is given
by v= dr/dt = r(dur /dt) + (dr/dt)ur = r(dur /d)(d/dt) + (dr/dt)ur = r(d/dt)u +
(dr/dt)ur . We next calculate the acceleration:
dr d
d2
d du d d2 r
dr dur d
dv
=
u + r 2 u + r
+ 2 ur +
a=
dt
dt dt
dt
dt d dt
dt
dt d dt
!2
2
2
dr d
d
d r
dr d
d
=
u + r 2 u r
u
ur + 2 ur +
dt dt
dt
dt
dt
dt dt
d2 r
d
= 2 r
dt
dt
!2
ur
d2
dr d
+ r 2 +2
u .
dt
dt dt
11
Name:Stanislav Shur
IP:74.88.120.168
It follows that the radial component ar and the transverse component a of the acceleration are given respectively by
d
d2 r
ar = 2 r
dt
dt
!2
and
a = r
d2
dr d
.
+2
2
dt
dt dt
Since the force is central, a = 0. We multiply the differential equation expressing that
fact by r to get
!
2
dr d
2d
2 d
r 2 + 2r
=0
or
d r
= 0.
dt
dt dt
dt
It follows that r2 d
dt = k, where k is a constant. But the area swept out by the radius
vector between the angles 0 and 1 is given by
A=
Z 1
0
1 2
r d.
2
Therefore,
dA
dA d
1 d
1
=
= r2
= k.
dt
d dt
2 dt
2
Therefore, A = 12 kt + C and the area swept out depends only on time. This is Keplers
law of areas.
5. Triangle SQP has vertices S = (0, 0), P = (x, y), and Q = (x + dx, y + dy). The area of
the triangle is therefore 12 ((x+dx)y(y+dy)x) = 12 (xy+y dxyxx dy) = 12 (y dxx dy).
6. The conversion to differentials of Newtons description of the central force via QR and the
square of the area of triangle SQP was made on p. 335. Thus, the force is proportional
to
q
b d2 x x2 + y 2
,
x(y dx x dy)2
where we have replaced a by b in the expression to avoid confusion with the acceleration.
We will show that this expression is equivalent to the expression for ar in exercise 4,
given that a = 0. We set x = r cos and y = r sin . Then dx = dr cos r sin d and
dy = dr sin + r cos d. So y dx x dy = r sin dr cos r2 sin2 d r cos dr sin
r2 cos2 d = r2 d. This value is proportional to the infinitesimal area of the triangle
SP Q in polar coordinates, and we know that that value is a constant multiple of dt; that
is, r2 d = k dt. It follows that (y dx x dy)2 = k 2 dt2 . We next calculate d2 x:
d2 x = d2 r cos dr sin d dr sin r cos (d)2 r sin d2
= d2 r cos r cos (d)2 (r sin d2 + 2 sin dr d).
12
Name:Stanislav Shur
IP:74.88.120.168
Thus, we can translate Newtons force F as follows:

b[cos (d2 r r(d)2 ) sin (r d2 + 2 dr d)]r
r cos k 2 dt2
!2
"
#
b d2 r
b sin
d2
dr d
d
= 2
r
+ 2
r
+2
k
dt2
dt
k cos dt2
dt dt
F =
b
b sin
a = ar ,
a
+
r
k2
k 2 cos
since a = 0.
7. Given the results of exercise 6, we see that Hermanns equation is equivalent to ar =
or to ar = rK2 , or to
!
d2 r
k
d
r
=
.
dt2
dt
r2
1
r2
8. If we rewrite the equation a cx/b = x2 + y 2 in polar coordinates, we get a

cr cos /b = r. This equation can be rewritten as
!
c cos
a= 1
r
b
or r =
a
1
c cos
b
or r =
(c/b)(ab/c)
.
1 c/b cos
If we set e = c/b, this latter equation is the standard polar equation of a conic section
with eccentricity e. We know in fact that if e = 1, or b = c, we get a parabola; if e < 1,
or c < b, we get an ellipse; and if e > 1, or b < c, we get a hyperbola.
9. We know that a conic section can be written in polar coordinates as r = de/(1 e cos ),
where e is the eccentricity. If we set x = r cos this equation canqbe rewritten as r =
de/(1ex/r) = rde/(rex). This is equivalent to rex = de or to x2 + y 2 = ex+de,
q
or to x2 + y 2 = x + . Note that since || = e, we know that if || = 1, the conic is a

parabola; if || > 1, the conic is a hyperbola; and if || < 1, the conic is an ellipse. The
value determines the location of the center of the conic section.
10. If y = ex/a , then dy = 1a ex/a dx, d2 y = a12 ex/a dx2 , and d3 y = a13 ex/a dx3 . Therefore
a3 d3 y = ex/a dx3 = y dx3 . Now we assume that ex/a (a3 d3 y y dx3 ) = d[ex/a (A d2 y +
x/a[A d2 y dx +
B dy dx + Cy dx2 )]. If we take the differential on the right, we get 1
a e
B dy dx2 +Cy dx3 ]+ex/a[A d3 y +B d2 y dx+C dy dx2 ] = ex/a [A d3 y +(B a1 A)d2 y dx+
(C 1a B)dy dx2 a1 Cy dx3 ]. If we equate this expression with the one on the left above,
we have A = a3 ; B a1 A = B a2 = 0, so B = a2 ; and C a1 B = C a = 0, so C = a. It
follows that the left side above is the differential of ex/a (a3 d2 y + a2 dy dx + ay dx2 ). But
if y is a solution to a3 d3 y y dx3 = 0, then d[ex/a (a3 d2 y +a2 dy dx+ay dx2 )] = 0. Thus
ex/a (a3 d2 y + a2 dy dx + ay dx2 ) = k, for some k, or a3 d2 y + a2 dy dx + ay dx2 = kex/a .
13
Name:Stanislav Shur
IP:74.88.120.168
But if y 6= cex/a , this is not possible, unless k = 0. Therefore, y satisfies the equation
a3 d2 y + a2 dy dx + ay dx2 = 0, or a2 d2 y + a dy dx + y dx2 = 0.
11. We know that y = ex is a solution of y 000 6y 00 + 11y 0 6y = 0. So assume that ex (y 000
d x
[e (ay 00 + by 0 + cy)] = ex(ay 00 + by 0 + cy) + ex(ay 000 + by 00 + cy 0).
6y 00 + 11y 0 6y) = dx
It follows by comparing coefficients that a = 1; a + b = 6, so b = 5; and b+ c = 11,
d x 00
[e (y 5y 0 + 6y)] = ex(y 000 6y 00 + 11y 0 6y). But if y is a
so c = 6. Therefore dx
d x 00
solution to the original differential equation, this implies that dx
[e (y 5y 0 + 6y)] = 0.
x
00
0
00
0
x
So e (y 5y + 6y) = k for some k, or y 5y + 6y = ke . But if y 6= cex , this is not
possible, unless k = 0. Therefore, y 00 5y 0 + 6y = 0, as claimed.
12. The characteristic polynomial in this case is p3 6p2 + 11p 6 = (p 1)(p 2)(p 3).
We could divide everything by 6 and get the factor of the polynomial in the form Euler
wrote it as (1 p)(1 12 p)(1 13 p). It then follows that independent solutions are Aex,
Be2x, and Ce3x, and any solution is simply a combination of these.
13. Suppose y = uex satisfies the differential equation a2 d2 y + a dy dx + y dx2 = 0. We
calculate that dy = ex du+uex dx and that d2 y = ex d2 u+ex dx du+ex du dx+
2 uex dx2 . Therefore a2 d2 y + ady dx + y dx2 = a2 ex d2 u + (2a2 ex + aex) du dx +
(a2 2 uex +auex +uex)dx2 . To eliminate the term in du dx, we must have 2a2 +a =
1
1
. Then the coefficient of the dx2 term is ex (a2 4a12 a 2a
+ 1)u or 34 uex.
0, or = 2a
Dividing through by ex, we find that u must be a solution to a2 d2 u + 34 u dx2 = 0.
14. To solve the equation, we first multiply by du to get a2 du d2 u = 34 u dudx2 . Since the
differential of du2 is 2 du d2 u, we can integrate this equation with respect to u to get
a2
3 2
2
2
2
2
2
2
2
2
2 du = (C 8 u )dx , or 4a du = (K 3u )dx . This equation simplifies to
dx =
By making the substitution u =
Z
2a
du.
3u2
K2
sin v, we can integrate this differential equation:

Z
K
cos v dv
K 2 K 2 sin2 v 3
Z
2a
2a
2a
3u
=
f.
dv = v f = arcsin
K
3
3
3
x=
It follows that x + f =
2a
du =
2
K 3u2
2a
arcsin
3u
K
or
2a
3
2a (x + f )
= arcsin
3u
K ,
or, finally, that
!
(x + f ) 3
u = C sin
.
2a
14
Name:Stanislav Shur
IP:74.88.120.168
Q
2
2
15. Since P = 2xy 3 +6x2 y 2 +8x and Q = 3x2 y 2 +4x3 y +3, we have P
y = 6xy +12x y = x .
R
Therefore, we define f (x, y) to be (2xy 3 +6x2 y 2 +8x)dx+r(y) = x2 y 3 +2x3 y 2 +4x2 +r(y).
Taking the derivative of this function with respect to y and comparing it to Q shows
that r0 (y) = 3, so r(y) = 3y and the solution to the differential equation is the equation
x2 y 3 + 2x3 y 2 + 4x2 + 3y = k.
16. If ax = y 2 and by = z 2 , then z = (by)1/2 = (b(ax)1/2)1/2 = b1/2 a1/4 x1/4 . Since y =

R
(ax)1/2 , we have dy = 12 a1/2 x1/2 dx. Therefore, z dy = 12 b1/2 a3/4 x1/4 dx, and z dy =
R 1 1/2 3/4 1/4
a x
dx = 23 b1/2 a3/4 x3/4 . We next integrate this last expression with respect
2b
to x between 0 and x0 to get
V =
Z x0
0
8
2 1/2 3/4 3/4
7/4
b a x dx = b1/2 a3/4 x0 .
3
21
The modern method would be to evaluate a triple integral:

Z x0 Z ax Z by
0
dz dy dx =
Z x0 Z ax q
0
by dy dx =
Z x0
0
8
2
7/4
b(ax)3/4 dx = b1/2 a3/4 x0 .
3
21
17. We suppose that y(t, x) = (t + x) (t x). Since y(0, x) = f (x), we have f (x) =
(x) (x). Since y(t, 0) = y(t, l) = 0, we have (t + l) (t l) = 0, or (t + l) =
(t l). We note that f (x) = (x) (x) = f (x), so f (x) is an odd function.
Also, f (xl) = (xl)(x+l) = (x+l)(xl) = f (x+l), so f (x) is periodic
of period 2l. Since g(x) = 0 (x) 0(x), we have g(x) = 0 (x) 0(x) = g(x), so
g(x) is also an odd function. Finally, since is periodic, so is its derivative. Therefore,
g(x + l) = 0 (x + l) 0 (x l) = 0 (x l) 0 (x + l) = g(x l), and g(x) is periodic
of period 2l.
2
y
y
y
0
00
0
18. If y = F (t)G(x), then y
t = F (t)G(x), t2 = F (t)G(x), x = F (t)G (x), and x2 =
F (t)G00 (x). The original partial differential equation then implies that F 00 (t)G(x) =
00
00
F (t)G00 (x), or that FF = GG . Since the left hand expression is a function of t and the
right hand one is a function of x, each of these quotients must be equal to a constant,
say C. So F satisfies the ordinary differential equation F 00 = CF and G also satisfies
G00 = CG. The solution of these equations is the sum of real exponential functions, if C
is positive, or the sum of sines and cosines, if C is negative. But since y(t, 0) = y(t, l) = 0,
we have F (t)G(0) = 0 and F (t)G(l) = 0. This implies that both G(0) and G(l) = 0,
and this could
exponential functions.
Thus C < 0,
not happenif G were a sum of real

= c cos Cx + d sin Cx. To rewrite the
F (t) = a cos Ct + b sin Ct, and G(x)
first of these expressions in
the form A cos( Ct
), note that this latter expression
can be expanded as A cos Ct cos + A sin Ct sin . Therefore, we must have
A cos = a and A sin = b. This means that tan = ab and that A2 = a2 + b2 . Thus we
can find the appropriate
A and . A similar argument shows that we can rewrite G(x)
in the form B sin( Cx ).
15
Name:Stanislav Shur
IP:74.88.120.168
19. Let the sides of the isosceles triangle have length s and the base have length b. If
we draw radii from the center of the inscribed circle to the two sides and the base,
the radius to the base divides the base into two segments of length 2b , while the radii
to the sides divide each side into two segments of length 2b and s 2b . By drawing
lines from the three vertices to the center of the circle, we see that the area of the
= s + 2b . Also, by Herons formula, the area
triangle is given by A = 4 4b + 2 sb/2
2
q
of the triangle can be expressed as A = (s + 2b )( 2b )( 2b )(s 2b ) = 2b s2 b4 . If we

solve the first area expression for s = A 2b and substitute in the second, we get
q
2
A = 2b (A 2b )2 b4 = 2b A2 bA. If we square both sides of this equation and
simplify, we get 4A2 = b2 A2 b3 A, so (4 b2 )A2 + b3 A = 0, or A =
dA
db
b4 12b2
.
(b2 4)2
b3
.
b2 4
To minimize
A, we take the derivative with respect to b. We get

=
The derivative is 0
when b4 12b2 = 0, or when b2 = 12. So b = 2 3. Then the area A = 248 3 = 3 3, and
the side s = A 2b = 2 3.
20. The volume V of a cone is given by V = 13 r2 h, where r is the radius of the base and h
is the height. If we designate the slant
height by `, then `2 = h2 + r2 . The surface area
S is given by S = 12 `2r = `r = r h2 + r2 . If we solve the volume formula for h and

substitute in the surface area formula, we get
s
r q 2
9V 2
2 =
S = r
+
r
9V + 2 r6 =
2r4
r2
9V 2 + 2 r6
.
r
To minimize S, we take its derivative and set it equal to 0. We have

dS
=
dr
r
(6 2 r5 )
2 9V 2 +2 r6
r2
If the derivative is to be equal to 0, we must have
9V 2 + 2 r6
.
2 6
3 r
9V 2 +2r6
9V 2 + 2r6 = 0, or
3 2 r6 (9V 2 + 2 r6 ) = 0, or 2 2 r6 = 9V 2 . We can solve this for r to get r3 =

s
r=
We then have
3V
,
2
and
3 2V
.
2
3V r
3V r 2
3V
= 2r.
=
h= 2 =
r
r3
3V
21. As noted in the text, if y = cos z, we have zy2 = ay2 . Since y = a when z = 0, we have
= 1 . Now, taking further fluxions of the above expression, we get y2 = y2 . But
E
a
z
a
1/a
y = 0 when z = 0, so E = 0. Then 2 = 2 . Thus E = 2 = 3 . Similarly, we

z
16
Name:Stanislav Shur
IP:74.88.120.168
= 0 and E
= a15 . It follows that the first four non-zero terms of the
calculate that E
1 2
1 4
1 6
power series for y = cos z are y = a 2a
z + 4!a
3 z 6!a5 z .
2a3 4x3
a3

. If we evaluate this quantity
3
2 2a3 xx4
3( a2 x)2
3
3
a
4
3a
at x = a, we get 2a
2 = 3 a. Similarly, the derivative of the denominator is
2a2
3ax2
3a3
3
. Its value at x = 1 is 4a

4
3 = 4 . If we divide the value of the numerator by
3
3
4( ax )
the value of the denominator, we have the result 43 a 43 = 16
9 a.
22. The derivative of the numerator is
23. Let u = t/a. Then dt = a du. Therefore

Z x
t/a
dt = a
Z x/a
e du = a lim
Z x/a
v v
eu du = lim (aex/a aev ) = aex/a .

v
24. Let the rectangle be placed so its vertices are at (0, 0), (a, 0), (0, b), and (a, b), and let
the line be drawn through (0, 0), intersecting the extensions of the two opposite sides
at (u, b), and (a, t). Since ub = at , we have u = ab
t , so the two points at the end of
ab
the desired line segment are ( t , b) and (a, t). The square of the length of this line
2
2
segment is given by L = (a + ab
t ) + (t b) . To find the value of t which minimizes L,
ab
ab
we take the derivative: L0 = 2(a + t )( t2 ) + 2(t + b). The equation L0 = 0 can then
be written as t4 + bt3 a2 bt a2 b2 = 0. The polynomial on the left factors, so we get
give a solution to the problem, we must
(t3 a2 b)(t + b) = 0.Since t = b does not
3 2
3
3
2
have t = a b, or t = a b. Therefore, u = ab2 , and the desired
line segment
which
3
3
must have a minimum length goes through the point (a, a2 b) and ( ab2, b).
25. The curve y 2 = 4(2x)
is defined for 0 < x 2. For each x value in that interval, there
x
are two symmetrically placed y values, so the curve is symmetric about the x-axis. As
x approaches 0, y 2 approaches infinity; thus the curve is asymptotic to the y-axis. To
find the area bounded by the curve and the y-axis, it is easiest first to solve for x and
then integrate. Thus, we rewrite the equation of the curve in the form xy 2 = 8 4x and
solve this as x = y28+4 . Then the area is given by
A=
y
8
1
dy = 8 arctan
2
y +4
2
2
=4
4
= 4.
2
2
x
x2
x3
x4
1 2 + 3 4 + , while
2
3
4
= k1 ( x1 x2 + x3 x4 + ). Thus
= 1k ln a, or k = ln a.
26. If y = ln(1 + x), then y =

y = loga (1 + x) is given by y
If we take 1 + x = a, then 1
in general the series for

loga (1 + x) =
1
k
ln(1 + x).
17
Name:Stanislav Shur
IP:74.88.120.168
27. We have ln(1 + x) = j(1 + x)1/j j. But

1
1(j 1) 2 1(j 1)(2j 1) 3 1(j 1)(2j 1)(3j 1) 4
1
(1 + x) j = 1 + x
x +
x
x +
j
j 2j
j 2j 3j
j 2j 3j 4j
Since j is infinite,
j1
2j
= 12 ,
2j1
3j
= 23 ,
3j1
4j
j(1 + x) j = j +
= 34 , and so on. Therefore,
x x2 x3 x4
+
1
2
3
4
If we subtract j from this series, we get the standard power series for ln(1 + x).
28.
If y = arctan x, then by looking at the right triangle with legs 1 and x and hypotenuse
1 + x2 , we get sin y = sin(arctan x) = x 2 and cos y = cos(arctan x) = 1 2 . If
= sin y, then dp = (1+xdx2 )3/2
p =
arcsin p, we also know that dy = dp 2 .
1p
dx
dx
2
get dy = (1+x2)3/2 1 + x = 1+x2 .
1+x q
x
1+x2
and
1 p2
1+x
= cos y =
1
.
1+x2
Since y =
If we substitute in this expression for dp, we
29. If y = ax, then dy = ax+dx ax = ax(adx 1) = ax(ln a dx+ (ln a)2 dx + ) = ax ln a dx+
terms in higher powers of dx. Therefore, dy = (ln a)ax dx.
30. If y = tan x, then
tan x + tan dx
tan x
1 tan x tan dx
tan x + tan dx tan x + tan2 x tan dx
tan dx + tan2 x tan dx
=
=
1 tan x tan dx
1 tan x tan dx
2
2
(1 + tan x)dx
(1 + tan x) tan dx
=
= (1 + tan2 x)dx,
=
1 tan x tan dx
1 tan xdx
dy = tan(x + dx) tan x =
because tan dx = dx and tan x dx is infinitely small with relation to 1.

31. According to dAlembert, a magnitude y is the limit of a magnitude x, when x may
approach y within any given magnitude, though it may not exceed y. If we think of x as
being a function of t, then we could translate this into the statement that lim x(t) = y
if x(t) < y for all t and, given any > 0, there is a value t such that y x(t) < . This
is quite different from the modern idea of a limit, partly in that we usually define the
limit of x(t) as t approaches some particular quantity and also in that we do not have
any such restriction that x(t) is always less than y.
32. We have
i
1
1
P (x, i) = ( x + i x) =
.
i
i x+i+ x
18
Name:Stanislav Shur
Therefore p =
.
2 x
IP:74.88.120.168
Next,
"
#
"
#
1
1
1
x x+i
1
Q(x, i) =
=

i
i 2 x( x + i + x)
x+i+ x 2 x
"
#
1
1
i
=

2 =
.
i 2 x( x + i + x)
2 x( x + i + x)2
It follows that q =
1
.
8x x
Finally,
"
#
"
#
1
1
1 4x + ( x + i + x)2
1

+ =
R(x, i) =
i
i 8x x( x + i + x)2
2 x( x + i + x)2 8x x
q
1 2x + i + 2 x(x + i)
=
.
i 8x x( x + i + x)2
It follows that r =
1
8x x(2 x)2
1
.
32x2 x
33. Given that f (x + i) = f (x) + pi + qi2 + ri3 + , we can take the derivative with respect
to i. We get f 0 (x + i) = p + 2qi +3ri2 + . If we set i = 0, we have f 0 (x) = p. Next, take
the second derivative. We have f 00 (x + i) = 2q + 6ri + . Again setting i = 0, we get
f 00 (x) = 2q, or q = f 00 (x)/2. If we take the third derivative, we have f 000 (x + i) = 6r + .
If we again set i = 0, we get f 000 (x) = 6r, or r = f 000 (x)/6. We can continue this process
indefinitely.
19
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER THIRTEEN
1. There is only one way of getting three successes in three tries, namely by being successful
each time. Since the probability of one success in one try is a, the probability of three
consecutive successes is a3 . There are three ways of getting two successes in three tries,
namely, the failure could be on the first, the second, or the third try. Since the probability
of success in any one try is a and the probability of failure is b, the probability of any
one of these methods happening is a2 b. Therefore, the probability of two successes in
three tries is 3a2 b. The probability of one success in three tries is the same as that of
two failures in three tries, so is equal to 3ab2 . Finally, the probability of no successes is
the same as that of three failures, namely b3 .

n
ways of getting r successes in n tries, because each set of r successes
2. Since there are nr
corresponds to n r failures, and since the probability of any one of these ways of r
successes and n r failures is ar bnr , it follows that the total probability for r successes
n
in n tries is nr
ar bnr .

10 1 4 2 6
1 64
6 ( 3 ) ( 3 ) = 210 81 729
13,440
1 5 2 5
1
32
8064
. Also, P (S = 5) = 10
5 ( 3 ) ( 3 ) = 252 243 243 = 59,049 . Finally, P (S = 6)
59,049

10 1 6 2 4
13,440
1 16
3360
8064
3360
6 ( 3 ) ( 3 ) = 210 729 81 = 59,049 . Therefore, P (4 S 6) = 59,049 + 59,049 + 59,049
24,864
59,049 = 0.42.
3. If a = 13 , b = 23 , and n = 10, we calculate that P (S = 4) =

=
=
=
4. We calculate
log 19000
log c(s 1)
=
= 300.465
log(r + 1) log r
log 31 log 30
and
log 29000
log c(r 1)
=
= 210.597.
log(s + 1) log s
log 21 log 20
Thus, we take m = 301 and n = 211. Then
mt +
st(m 1)
20 50 300
= 301 50 +
= 24727.419
r+1
31
and
nt +
rt(n 1)
30 50 210
= 211 50 +
= 25550.
s+1
21
Since the latter number is the larger, we have N (1000) = 25, 550.
5. We calculate
log 190000
log c(s 1)
=
= 370.687
log(r + 1) log r
log 31 log 30
20
Name:Stanislav Shur
IP:74.88.120.168
and
log 290000
log c(r 1)
=
= 257.79.
log(s + 1) log s
log 21 log 20
Thus, we take m = 371 and n = 258. Then
mt +
st(m 1)
20 50 370
= 371 50 +
= 30485.5
r+1
31
and
nt +
30 50 257
rt(n 1)
= 258 50 +
= 31257.14.
s+1
21
Since the latter number is the larger, we have N (10, 000) = 31, 258.

X
N

p >
X
6. We know that P X
N p = 1 P N p > . Therefore, Bernoullis version is

equivalent to 1 P X
N p > > cP

or to P X
N p > <
1
c+1 .
or to 1 > (c + 1)P X
N p >
log 2
7. According to De Moivres procedure, the number of trials x is given by log 10log
9 = 6.6.
For the approximation procedure, we take q = 9, so x = 0.7 9 = 6.3. In either case,
this means that in 6 trials the odds are slightly less than even, while in 7 trials, they are
more than even.
1
. Therefore, to insure even odds,
8. The probability of winning the prize on one ticket is 40
log 2
one must buy x = log 40log 39 = 27.37, that is, 28 tickets.
9. Since bx + xabx1 represents the number of chances in which the event will succeed no
x
+xabx1
more than once, the probability that this will happen is given by b (a+b)
x . Since we
want this probability to be equal to 12 , we get the equation (a + b)x = 2bx + 2xabx1.
x
If we set a : b = 1 : q, we can rewrite this in the form ( qb + b)x = 2bx + 2xb
or
q
1 x
2x
1
x
(1 + q ) = 2 + q . Taking logarithms, we get x log(1 + q ) = log 2 + log(1 + q ). If we
assume that q is large, then the left side of this equation can be approximated by using
just the first term of the power series for the logarithm: x 1q . Thus, if we write z = xq , we
have z = log 2 + log(1 + z). The easiest way to solve this equation is by using a graphing
calculator. We find that z 1.678, so x 1.678q.
q
10. If u = t/ np(1 p), then du = dt/ np(1 p) and when t = n, u = n/ np(1 p) =
q
n/ p(1 p). Given that the integrand is symmetric about x = 0, we see that we can
21
Name:Stanislav Shur
IP:74.88.120.168
write
1
Z
2 np(1 p) Z
e[t /2np(1p)] dt = q
P = q
n
2np(1 p)
2np(1 p)
2 Z n/ p(1p) u2/2

=
e
du.
2 0
n/ p(1p)
eu
/2
du
To calculate this integral when p = .6, = .02, and n = 6498, we need to use a graphing
utility. The value of the integral in this case is 0.999. With the same values of p and ,
to find n so that P = .99 requires making several trials with the graphing utility. By
experimenting, we find that n = 3980 solves the problem.
11. According to the formula with P = 1, r = .04 and n = 50, the present value is
A=
(1.04)50 1
= 21.4822 pounds.
.04(1.04)50
12. Setting n = 86 k = 86 36 = 50 and A = 21.4822 from exercise 11, the formula for
present value with interest at 4% gives
Q=
1 (1.04)(21.4822)/50
= 13.8293 pounds.
0.04
13. According to Bayes theorem, expressed in its integral form, we have

P ((r < x < s)|X = n 1) =
R s n n1
(1 x) dx
r n1 x

R
1
n
0 n1
xn1 (1 x) dx
Rs
n1 x) dx
r n(x
.
R1
n1 xn ) dx
0 n(x
n+1
1
1
The integral in the denominator is (xn nx
n+1 )|0 = n+1 , while the integral in the
n
numerator is sn rn n+1
(sn+1 rn+1 ). Thus the desired probability is P = (n +
1)(sn rn ) n(sn+1 rn+1 ). In the particular case indicated, we have n = 11, r = 0.7,
and s = 1. Thus, P = 12(1 .711 ) 11(1 .712 ) = 11.763 10.848 = 0.915.
14. We again use Bayes theorem in its integral form. Here p = n, so the quotient of the two
integrals, as noted in the text, equals sn+1 rn+1 . If we take s = 1 and r = 12 , then the
n+1
1
= 2 2n+11 . It follows that
probability P that x is greater than 12 is given by P = 1 2n+1
the odds are 2n+1 1 to 1 for more than an even chance of the event happening again.
15. If E is the event that the third ball is white and F is the event that the first two balls were
white, then P (E|F ) = P (E F )/P (F ). We calculate the two probabilities on the right.
First, note that P (E F ) is the probability of drawing three white balls in three draws.
22
Name:Stanislav Shur
IP:74.88.120.168
Since we do not know the color composition of the balls in the urn, we can assume there
are three equally likely possibilities. Both balls could be white; both could be black;
or there could be one white and one black. In the first case, with probability 13 , it is
certain that three white balls will be drawn; in the second case, with probability 13 , it is
impossible that three white balls will be drawn; and in the third case, with probability 13 ,
the probability of drawing three white balls is 18 . Therefore. P (E F ) = 13 1 + 13 18 = 38 .
5
5
9
. Therefore, P (E|F ) = 38 12
= 10
.
Similarly, P (F ) = 13 1 + 13 14 = 12
16. The probability of A winning on the first flip is 12 . The probability of A winning on the
second flip is ( 12 )2 (since in this case the first flip must have been a head). Similarly, the
probability of A winning on the third flip is ( 12 )3 and the probability of A winning on the
nth flip is ( 12 )n . The expectation for A is the sum of the probabilities each multiplied
by the payoff for that outcome. Thus the expectation is 12 1 + ( 12 )2 2 + ( 12 )3 4 + +
P
1 i1
( 12 )n 2n1 + =
= 12 + 12 + 12 + . Since this sum is infinite, so is the
i=0 2i 2
expectation.
23
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER FOURTEEN
z
= ncyf
= phykz
. We then equate the
1. We solve each equation for x: x = mbycz
a
d
g
first and second expressions to get one equation in y and z and equate the second and
third to get a second equation: (bdae)y+(cdaf )z = mdan, (dhge)y+(dkgf )z =
)z
pd gn. We then solve the first equation for y = mdan(cdaf
and substitute in the
bdae
second equation. After simplifying, we get (dh ge)(md an) (dh ge)(cd af )z +
(bd ae)(dk gf )z = (bd ae)(pd gn). When we multiply out and solve for z, we get
z=
bgn + aep + hmd bdp ahd gem

.
hcd + bgf + aek haf gec bdk
To find x and y, it is simplest to replace the numerator in this expression analogously.
2. One example is y = x3 3x and x = y 2 2.
3. Note that equation 4 is the sum of equations 2 and 3, while equation 1 is the sum of
twice equation 2 and equation 3. Therefore, the system reduces to two equations in four
unknowns. In this case there is a two-parameter family of solutions and not a unique
solution.
4. We know that every solution of xn 1 = 0, except 1, is a power of the complex root

r
s
whose angle is 2
n . So suppose = and = . We need to show that is a power
of . So set rx s (mod n). Since n is prime, we know this equation has a solution
x = q. But then rq = s , or q = , as desired.
5. If v = x1 x2 , then v 2 = (x1 +x2 )2 4x1 x2 = t2 4c. Thus v satisfies a quadratic

equation
2
2
in t. It is simpler to rewrite this equation in the form v = b 4c, so v b2 4c. We
vb
b2 4cb
1 x2 )b
also have x1 = (x
=
.
It
follows
that
x
=
and the second solution
1
2
2
2
x2 is given by
b2 4cb
.
2
6. We set x = y+ y2 and substitute into the original equation. We get (y+ y2 )3 6(y+ y2 )9 =
0, or y 3 + y83 9 = 0. If we let r = y 3, then r + 8r 9 = 0, or r2 9r + 8 0, and r = 1
or r = 8. In the first case, we have y = 1, y = , and y = 2 . In the second case, we
have y = 2, y = 2, and y = 2 2 . It follows that the three roots of the original equation
are x1 = 1 + 2 = 3, x2 = + 2 2 , and x3 = 2 + 2. The six values of y can then be
24
Name:Stanislav Shur
IP:74.88.120.168
expressed as follows:
1
(3 + ( + 2 2 ) + 2 ( 2 + 2)) =
3
1
( + 2 2 + ( 2 + 2) + 2 3) =
3
1 2
( + 2 + 3 + 2 ( + 2 2 )) =
3
1
(3 + ( 2 + 2) + 2 ( + 2 2 )) =
3
1 2
( + 2 + ( + 2 2 ) + 2 3) =
3
1
( + 2 2 + 3 + 2 ( 2 + 2)) =
3
1
(7 + + 2 ) = 2
3
1
(7 2 + + 1) = 2 2
3
1
(7 + 2 + 1) = 2
3
1
(5 + 2 + 2 2 ) = 1
3
1
(5 2 + 2 + 2) = 2
3
1
(5 + 2 2 + 2) =
3
7. The three values are x1 x2 + x3x4 , x1 x3 + x2 x4 , and x1 x4 + x2 x3 . The value of the

variable which is multiplied by x1 determines the entire expression, and there are just
three possible values for that variable, as indicated.
8. The relationship between roots and coefficients shows that x1 + x2 + x3 + x4 = a,
x1 x2 + x1 x3 + x1 x4 + x2 x3 + x2 x4 + x3 x4 = b, x1 x2 x3 + x1 x2 x4 + x1 x3 x4 + x2 x3 x4 = c
and x1 x2 x3 x4 = d. It follows immediately that + + = b. For the second result,
note that + + = (x1 x2 + x3 x4 )(x1 x3 + x2 x4 ) + (x1 x2 + x3 x4 )(x1 x4 + x2 x3 ) +
(x1 x3 + x2 x4 )(x1 x4 + x2 x3 ) = x21 x2 x3 + x1 x22x4 + x1 x23 x4 + x2 x3 x24 + x21 x2 x4 + x1 x22 x3 +
x1 x2 x24 + x2 x23 x4 + x21 x3 x4 + x1 x2 x23 + x1 x2 x24 + x22 x3 x4 = (x1 + x2 + x3 + x4 )(x1 x2 x3 +
x1 x2 x4 + x1 x3 x4 + x2 x3 x4 ) 4x1 x2 x3 x4 = ac 4d. Finally, = x31 x2 x3 x4 + x21 x22 x23 +
x21 x22 x24 + x1 x32 x3 x4 + x21 x23x24 + x1 x2 x33 x4 + x1 x2 x3 x34 + x22 x23 x24 . Also, a2d + c2 4bd =
(x1 + x2 + x3 + x4 )2 x1 x2 x3 x4 + (x1 x2 x3 + x1x2 x4 + x1 x3 x4 + x2 x3 x4 )2 4(x1 x2 + x1 x3 +
x1 x4 + x2 x3 + x2 x4 + x3 x4 )(x1 x2 x3 x4 ). These two expressions are in fact equal. We
note, for example, that the terms in the first expression with cubes come from the
first term in the second expression, while the terms in the first expression which have
squares of three variables come from the second term in the second expression. The third
term in the second expression eliminates all the additional expressions in the first two
terms. Since we now know the sum of three values, the sum of the three values taken in
pairs, and the product of the three values, we know from the relationship between roots
and coefficients of a polynomial that the three values are the roots of the cubic equation
y 3 (+ +)y 2 +( + +)y = 0 or y 3 by 2 +(ac4d)y(a2 d+c2 4bd) = 0.
9. Using the notation of exercise 8, we have a = 0, b = 0, c = 12, and d = 3. Therefore,
the reduced cubic equation is y 3 12y 144 = 0. We can find one root by trial and error:
= 6. It follows that the other two roots have their sum equal to 6 and their product
2
Therefore,
equal to 24. Thus, they satisfy the quadratic equation
y + 6y + 24 = 0.
the two additional roots of the cubic are = 3 + 15 and = 3 15. In
particular, if the four roots of the quartic equation are x1, x2 , x3 , and x4 , we know that
x1 x2 + x3 x4 = 6. From the relationship between roots and coefficients, we also know
25
Name:Stanislav Shur
IP:74.88.120.168
that x1 x2 x3 x4 = 3. Therefore, x1 x2 and

of the quadratic equation
x3 x4 are the roots
2
z 6z + 3 = 0. Thus, x1 x2 = 3 + 6 and x3 x4 = 3 6. Also, we know that
(x1 + x2 ) + (x3 + x4 ) = 0, or x3 + x4 = (x1 + x2 ). Given further that x1 x2 x3 + x1 x2 x4 +
x1 x3 x4 + x2x
3 x4 = 12, we can rewrite
this equation as x2 x2 (x3 + x4 ) + x3x4 (x1 + x2 ) = 12
or as (3 + 6)(x1 + x2 ) + (3 6)(x1 + x2 ) = 12, or, finally, as 2 6(x1 + x2 ) = 12.
It follows that
x1 + x2 =
6. Therefore, x1 and x2 are the roots of the quadratic
2
equation w = 6w = 3 + 6 = 0. We therefore have
q
q
s
6 + 6 4(3 + 6)
6 + 6 4 6
1
1
x1 =
=
=
6+ 61 .
2
2
2
2
Similarly,
1
x2 =
6
2
1
61 .
2
We also find that x3 and x4 satisfy the quadratic equation w2

1
6+
x3 =
2
1
61
2
and
1
x4 =
6
2
6w + 3 6 = 0. So
1
61 .
2
10. Suppose that p and p2 + 3q 2 have a common factor. Then p2 and p2 + 3q 2 will have
the same factor, and so will 3q 2 . But since x and y are relatively prime, so are p and
q. Therefore, the only possible common factor of p2 and 3q 2 is 3. But we are assuming
that p is not divisible by 3, so this is also impossible.
11. Assume that p = 3r. Then 14 3r(9r2 + 3q 2 ) = 94 r(3r2 + q 2 ) is a cube. We first note that
the two factors here are relatively prime. Recall that p and q are relatively prime, with
p even. Thus r is even. Neither 2 nor 3 can divide 3r2 + q 2 because q is odd. If any
other number divided 3r2 + q 2 and r, it would also divide q 2 and therefore p. Since this
is impossible, the two factors are relatively
prime andtherefore both are cubes. As in
2
2
the text, we can factor q + 3r as q + r 3)(q r 3).
factors are
Since the two
relatively prime, both of them are also cubes. Thus q + r 3 = (t + u 3)3 . Then
q = t3 9tu2 = t(t2 9u2 ) and r = 3u(t2 u2 ), where t is odd and u is even. Also,
9
8 9
2
2
2
4 r is a cube, so 27 4 r = 3 r is a cube. That is, 2u)t u ) = 2u(t + u)(t u) is a cube.
Again, these factors are relatively prime because u is even. So each is a cube: t + u = f 3 ,
t u = g 3 , and 2u = h3 . Therefore, f 3 g 3 = h3 , or g 3 + h3 = f 3 . We have therefore
found a new sum of cubes equation with each term less than the corresponding numbers
in the original sum of cubes equation. The impossibility of a solution is thus proved in
this case too by infinite descent.
12. The residue of 1 is 1; of 5 is 5; of 52 is 12; and of 53 is 8. If we multiply each of these
residues by 2, we get the coset {2, 10, 11, 3}. If we multiply each of the residues by 4, we
get the coset {4, 7, 9, 6}.
26
Name:Stanislav Shur
IP:74.88.120.168
13. We have, modulo 13, 12 = 1, 22 = 4, 32 = 9, 42 = 3, 52 = 12, and 62 = 10. The squares

of the other integers less than 13 are the same as these. Therefore, the quadratic residues
modulo 13 are 1, 3, 4, 9, 10, and 12.
14. Let a be a primitive root modulo q, that is, a number such that no power less than the
(q 1)st is congruent to 1 modulo q. Such a primitive root always exists. Since aq1 1
q1
(mod q), we must have a 2 1 (mod q), and no other power of a less than q 1
q1
4
is
even,
then
b
=
a
has the property that
can be congruent to 1 modulo q. If q1
2
q1
2
b 1 (mod q) and thus 1 is a quadratic residue. But 2 is even if and only if
q 1 (mod 4).
27
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER FIFTEEN
1. Connect AD and BC. We show first that triangles ABC and ABD are congruent.
We know that they share a common side, AB. Also, AC = DB and 6 CAB = 6 ABD.
Thus, the triangles are congruent by side-angle-side. Therefore, AD = BC and 6 DAB =
6 CBA. We now show that triangles ACD and BCD are congruent. We have AD =
BC, AC = DB and 6 DAC = 6 CBD by subtracting equal angles from equal angles.
Therefore, the triangles are congruent by side-angle-side and so 6 C = 6 D.
2. Let (ABC denote the defect of triangle ABC. Then (ABD) + (BDC) = (180
6 A 6 ABD 6 ADB) + (180 6 C 6 CDB 6 CBD) = 360 6 A 6 C 180
6 ABC = 180 6 A 6 C 6 ABC = (ABC). Note that we have used the fact that
6 ABD + 6 CDB = 180.
3. Since DE = EF , we have AE = 12 DE = 12 EF = BE. Therefore, 4AEB is isosceles.
Similarly, triangles DAC and CBF are isosceles. Also, since AE = AD, BE = DC,
and 6 E = 6 D, triangles AEB and DAC are congruent. Similarly, both of these are
congruent to triangle CBF . It follows that AB = AC = BC, so 4ABC is equilateral.
Next, 3 + = (CBF ) + (DAC) + (AEB) + (ABC) = 180 6 F 6 BCF 6 F BC +
1806 D6 DCA6 CAD+1806 E6 EBA6 EAB+1806 BAC6 ABC6 BCA =
720 6 E 6 D 6 F (6 BAC + 6 CAD + 6 EAB) (6 ABC + 6 F BC + 6 EBA)
(6 BCA + 6 BCF + 6 DCA) = 180 6 E 6 D 6 F = (DEF ) = . Finally, if the
four smaller triangles were all congruent, then angles BCF , BCA, and ACD would all
be equal. Because their sum is 180, each angle would be 60. The same would be true
of all the other angles in the four triangles, so the defects would be 0, contrary to our
assumption.
4. Let (ABC) denote the excess of triangle ABC. Then (ABD) + (BDC) = (6 A +
6 ABD + 6 ADB 180) + (6 C + 6 CDB + 6 CBD 180) = 6 A + 6 C + 6 ABC 360 +
(6 ABD + 6 CDB) = 6 A + 6 C + 6 ABC 180 = (ABC).
5. We have cosh x = cosh(i(ix)) = cos(ix) = cosix. Also, sinh x = sinh(I9 ix)) =
i sin(ix) = i sin ix. Then cosh2 x sinh2 x = cos2 ix (i sin ix)2 = cos2 ix +sin2 ix =
1.
6. We have = 12 (1 + 2) 12 (1 2 ) cos 2 = 12 (1 + 2 ) 12 (1 2)(cos2 sin2 ) =
1
1
2
1
2
1
2
2
2
2
2 1 (1 cos ) + 2 1 sin + 2 2 (1 sin ) + 2 2 cos = 1 sin + 2 cos .
7. The normal line to the surface z = f (x, y) is a line in the direction of the gradient,
z z
namely ( x
, y , 1). The normal vector to the plane z = y x + is (, , 1). Thus
the plane will contain the normal line if these two vectors are perpendicular, that is, if
z
z
their dot product is zero. This amounts to the condition x
x
1 = 0, as stated.
28
Name:Stanislav Shur
IP:74.88.120.168
8. Since (A, B, C) is a normal vector to the plane, the normal line is given by the three
equations x = At + x0 , y = Bt + y0 , z = Ct + z0 .
z
9. Monges form of the equations of the normal line are x x0 + (z z 0 ) x
= 0 and
0
z
xx
z
0
0
0
y y + (z z ) y = 0. If we set t = z/x , then we have x = x t + x and, by
substituting in the first equation of Monge, t + z z 0 = 0, or z = t + z 0 . Then the
z
0
second equation of Monge becomes y y 0 t z
y = 0, or y = x t + y . These three
equations form the modern vector equation of the normal line.
10. We need to show that under either of these circumstances, if we add together the numbers
associated to each bridge, that is, if we count the number of letters in the path, the
total is 1 more than the number of bridges. We suppose that there are n regions. We
designate the number of bridges leading into region i by ki . We first suppose that
exactly two of the ki are odd,Psay k1 and k2 . Then the total numberPof letters is
k1 +1
k2 +1
k3
kn
1
ki + 1. Since the number of bridges is 12 ki , we have
2 + 2 + 2 ++ 2 = 2
shown that the number of letters in this case is 1 more than the number of bridges, so
the Euler path exists. If all of the ki are even, let us suppose that our path starts in
P
region 1. Then the total number of letters is k21 + 1 + k22 + k23 + + k2n = 12 k1 + 1.
Thus again the number of letters is 1 more than the number of bridges, and the Euler
path exists.
11. One possible Euler path in the first diagram is EADCBAEC. An Euler path in the
second diagram is CBDBADACAC. In each of these, every crossing between the same
two regions is on a different bridge.
29
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER SIXTEEN
1. We divide p 1 by m. We get p 1 = qm + r, where 0 r < m. Then 1 = ap1 =
aqm+r = (am)q ar = 1q ar = ar . But m is the smallest positive integer such that am = 1.
Therefore, r = 0, so m divides p 1.
2. We calculate 23 1, 36 1, 43 1, 56 1, and 62 1 (mod 7). All of the exponents,
namely 2, 3, and 6 divide p 1 = 6, as proved in exercise 1.
3. Note that a is a primitive root for p = 13 if and only if 6 is the smallest positive integer
such that a6 1 (mod 13). Note also that a is a primitive root if and only if a is.
We then check the integers in turn: Since 1 is not a primitive root, neither is 1 12.
Since 26 1 and no smaller power of 2 is congruent to 1, we know that 2 is a
primitive root for p = 13. It follows that 2 11 is also a primitive root. Next, we
have 33 1, so 3 is not a primitive root, nor is 3 10. Also, since 4 = 22 , we know
that 46 = 212 = 1, so 4 is not a primitive root, nor is 4 9. Also, 52 1, so 54 1
and 5 is not a primitive root, nor is 5 8. Finally, 66 1 and no smaller power is
congruent to 1. Therefore, 6 is a primitive root and so is 6 7. Thus the primitive
roots of 13 are 2, 6, 7, and 11.
4. Suppose p divides abc . Then N (p) divides N (abc ) = N (a)N (b)N (c) . But N (p)
is prime. Therefore N (p) must divide one of the factors, say N (a). But N (a) is also
prime. Therefore N (p) = N (a) and p = ua, where u is a unit.
5. N (3 + 5i) = 9 + 25 = 34. Since 34 = 2 17, we see that 3 + 5i factors into a product of
two primes, one with norm 2 and one with norm 17. We get 3 + 5i = (1 4i)(1 + i),
and since the norm of each of the factors is prime, the factors themselves must be prime.
6. If we cube the residues modulo 7, we get 13 = 1, 23 = 1, 33 = 6, 43 = 1, 53 = 6, and
63 = 6. Thus, the cubes modulo 7 are 1 and 6, and therefore 3 is not a cube. Also, these
cubes (or cubic residues) do not differ by 1. It therefore follows from Germains theorem
that there are no case 1 solutions to the Fermat equation for exponent 3.
7. Suppose p is prime and assume it factors into irreducible integers: p = q1 q2 qn . Then
since p divides the product, it must divide one of the factors, say q1 . But then q1 would
not be irreducible, contrary to our hypothesis. Thus p itself must be irreducible.
8. We have N (a + b 5) = a2 + 5b2. Since N (2) = 4, any factor of 2 must have norm equal
to 2, and no such number exists. Similarly, since N (3) = 9, any factor
of 3 must have
norm equal to 3, and again there is no such number. Also, N (2 5) =9. Again,
any factor of either of these numbers must have norm equal to 3. And N (1 5) = 6,
so a factor of either of these must have norm equal to 2 or 3, both of which are impossible.
30
Name:Stanislav Shur
IP:74.88.120.168
2 consists of sums of multiples of 2 2 = 4, 2(1 + 5) =

9. If A =
(2, 1 + 5),then A
2 + 2 5, and (1 + 5)(1 + 5) = 4 + 2 5. Each of these numbers is a multiple
2
also 2 = (1)(4 + (4 + 2 5) +
of 2 (by some
integer in the 2domain), so A (2). But
(1)(2 + 2 5)), so 2 A , and therefore (2) A2 . Therefore (2) = A2 .
10. Since 3 is a primitive root modulo 7, we set h = 33 6 (mod 7). Then if r is a fixed
root of the equation, we set i = ri +rih , for i = 1, 2, 3. We get 1 = r+r6 , 2 = r2 +r5 ,
and 3 = r3 + r4 . We now need to find the cubic equation satisfied by the i . We note
that 31 + 21 21 = r3 + r2 + r + 2 + r6 + r5 + r4 = 1. Therefore, 1 (and also 2 and
3 ) satisfy the equation y 3 + y 2 2y 1 = 0. We now need to solve this cubic equation.
We simplify it by setting y = z 13 . We then get (z 13 )3 + (z 13 )2 2(z 13 ) 1 = 0,
7
7
or z 3 73 z 27
= 0. If we rewrite this in the form z 3 = 73 z + 27
, we can apply Cardanos
formula. We have z = + , where
v
u
u
3
t
7
+
54
s

7
54
2
3
and =
v
u
u
3
t
54
s

7
54
2
3
7
9
With a bit of algebraic manipulation, we get

1
=
3
s
3
7
(1 + 3 3)
2
and
1
=
3
s
3
7
(1 3 3).
2
Since y = z 13 , it follows that the three roots of the equation in y are 1 = 13 (1++),
2 = 13 (1 + 2 + ), 3 = 13 (1 + + 2 ), where is a complex cube root of 1.
r2 + 1 = 1 r, or
To solve the original equation, we note that since 1 = r + r6 , we haveq
r2 1 r + 1 = 0. There are two solutions to this equation: r = 12 (1 21 4). These
two solutions are, in fact, the original r and r6 , since 21 4 = (r r6 )2 . We find that
the other powers of r can be written similarly in terms of 2 and 3 .
P
i
2
11. We first note that 1 + 2 + 3 = 18
i=1 r = 1. Thus the coefficient of x in the cubic
equation for the i must be +1. For the remainder of the problem, it is convenient to use
both negative and positive powers of r in the representation of i . Thus, for example,
1 = (r + r1 ) + (r7 + r7 ) + (r8 + r8 ). We now multiply the i in pairs. We find
1 2 = (r + r1 ) + 2(r2 + r2 ) + 2(r3 + r3 ) + 3(r4 + r4 ) + 2(r5 + r5 ) + 3(r6 + r6 ) +
(r7 + r7 ) + (r8 + r8 ) + 3(r9 + r9 ). Also, 1 3 = 2(r + r1 ) +3(r2 + r2 ) +3(r3 + r3 ) +
(r4 + r4 ) + 3(r5 + r5 ) + (r6 + r6 ) + 2(r7 + r7 ) + 2(r8 + r8 ) + (r9 + r9 ). Finally,
2 3 = 3(r + r1 ) + (r2 + r2 ) + (r3 + r3 ) + 2(r4 + r4 ) + (r5 + r5 ) + 2(r6 + r6 ) +
P
i
3(r7 + r7 ) + 3(r8 + r8 ) + 2(r9 + r9 ). Therefore, 1 2 + 1 3 + 2 3 = 6 18
i=1 r = 6.
It follows that the coefficient of the x term in the cubic equation must be 6. Finally,
we calculate 1 2 3 by multiplying 1 2 (already calculated) by 3 . We find that each
positive and negative power of r occurs 11 times, while r0 = 1 occurs 18 times. Thus the
P
i
product is equal to 11 18
i=1 r + 18 = 11 + 18 = 7. Thus the constant term in the cubic
equation must be 7, and the equation itself must be x3 + x2 6x 7 = 0 as asserted.
31
Name:Stanislav Shur
IP:74.88.120.168
12. To show that 1 = r + r18 , 8 = r8 + r11 , and 7 = r7 + r12 are the roots of x3
1 x2 + (1 + 4 )x (2 + 2 ) = 0, we check the symmetric functions of these three
values. We have 1 + 8 + 7 = r + r18 + r8 + r11 + r7 + r12 = 1 , so the sum of
the i is the negative of the coefficient of the x2 term. Also, 1 8 + 1 7 + 8 7 =
(r9 + r12 + r7 + r10 ) + (r8 + r13 + r6 + r11 ) + (r15 + r + r18 + r4 ) = 1 + 4 . Therefore,
the sum of products of the i taken two at a time is equal to the coefficient of the x
term. Finally, the product 1 8 7 = r16 + r14 + r17 + r2 + r5 + r3 + 2 = 2 + 2, and the
product of the i is the negative of the constant term. From the basic properties of the
symmetric functions, we have that the i are the roots of the given polynomial.
q
13. From the equation x2 1 x + 1 = 0, where 1 = r + r18 , we have x = 12 (1 12 4) =

q
1
18 r2 2 + r17 ) = 1 (r + r18 (r r18 )2 ) = 1 ((r + r18 ) (r r18 )). Thus
(r
+
r
2
2
2
the two roots of this equation are r and r18 .
14. Constructing the double of a cube of side 1 requires constructing the cube root of 2.
But this number satisfies the irreducible polynomial equation x3 = 2, whose degree is
not a power of 2. Therefore, one cannot construct the double of a cube of side 1 using
straightedge and compass. Also, if one could construct one-third of an arbitrary angle,
one could, in particular, construct the cosine of one-third of a sixty degree angle, that
is, cos 20 . The triple angle formula for cosine is cos 3 = 4 cos3 3 cos . If we let
= 20 and set x = cos 20 , this equation becomes 12 = 4x3 3x, or 8x3 6x 1 = 0.
This equation is irreducible and its degree is not a power of 2. Therefore, its solution is
not constructible using straightedge and compass.
15. A graphical analysis of y = x3 + 6x 6 shows that the equation x3 + 6x = 6 has one real
root and therefore two complex ones. Thus adjoining the real root gives an extension F of
Q of degree three, which does not contain the complex roots. They satisfy an equation of
degree two over this field, so adjoining one of those roots gives us an additional extension
of degree two. Therefore, the splitting field of the cubic equation had degree 6 over Q.
Therefore the Galois group of the equation has order 6. There are only two groups of
order 6, S3 and the cyclic group. Every subgroup of the cyclic group is normal. In this
case, however, field F is not a normal extension of Q because it does not contain the
conjugates of the real root, so the subgroup of G corresponding to F is not normal. It
follows that G = S3 . This group does have a normal subgroup H of degree 3, namely
the cyclic group generated by a cyclic permutation of three elements, say (1, 2, 3). The
index of H in G is 2, so both the order of H and its index in G are prime, as required.
16. One of the groups of order six is the cyclic group. The other is generated by two elements
and , with 3 = 1, 2 = 1, and = 2 . The group tables can easily be written out
from this description. To show that there cannot be any additional groups, note that a
group of order 6 must contain a cyclic, normal subgroup of order 3, generated by , say.
It must also contain an element of order 2. Since the elements i j , i = 0, 1, 2, j = 0, 1
32
Name:Stanislav Shur
IP:74.88.120.168
are all the elements of the group, the group will be determined once we know . There
are only two possibilities. If = , then the group is Abelian and it is easy to see
that it is cyclic. If = 2 , then we have the non-Abelian group mentioned above.
17. There are three Abelian groups of order 8. The first is the cyclic group of order 8. The
second has two generators, and , with 4 = 1 = 2 and = . The third has three
generators , , and , with 2 = 2 = 2 = 1, and with all the generators commuting
with each other. There are also two non-Abelian groups of order 8. The first has two
generators, and , with 4 = 1 = 2 and with = 3 . The second one has two
generators, and , with 4 = 1, 2 = 2 , and = 3 .

a b
is in SL(2, p) if ad bc = 1. We must count the number of solutions
c d
of this equation modulo p. Let us rewrite the equation as a = (1 + bc)/d. It is evident
that if d 6= 0, then any choices of b and c determine a value for a. There are p 1 nonzero choices for d and p2 choices for b and c together, thus giving us p2 (p 1) choices
altogether. On the other hand, if d = 0, then our equation becomes bc = 1, and any
non-zero choice for b determines a value for c. In this case, in addition, the choice of a
is arbitrary. Thus, with d = 0, there are (p 1)p solutions to the equation. Therefore,
the total number of solutions is p2 (p 1) + p(p 1) = p(p + 1)(p 1) = p(p2 1), and
that is the order of the group SL(2, p). The group P SL(2, p) is the quotient group of
SL(2, p) by its subgroup of multiples of the identity matrix. The only multiples of the
identity matrix which have determinant 1 are I and (p 1)I (because 1 and p 1 are
the only solutions to x2 1 (mod p)). Therefore, the subgroup has only two elements
and the order of the quotient group is half the order of SL(2, p), namely 12 p(p2 1).
18. A matrix
19. First, note that the linear fractional transformations do act on P1 (p), because if (x1 , y1)
(x2 , y2 ), then the transforms of both z1 = x1 /y1 and z2 = x2 /y2 are the same. Second,
note that as mentioned in the solution to exercise 18, the subgroup of SL(2, p) consisting
of multiples of the identity only consists of two elements, I and (p 1)I. Now let
LF (p) denote the group of linear fractional transformations acting
onP1 (p), and consider

a
b , then (M)(z) =
the function : P SL(2, p) LF (p) such that if M =
c d

a b
e f
=
(az + b)/(cz + d). We note that is a group homomorphism:
c d
g h

ae + bg af + bh
, while the composite of the two transformations takes z to
ce + dg cf + dh

ez+f
gz+h + b

ez+f
c gz+h
+d
aez + af + bgz + bh
(ae + bg)z + (af + bh)
=
.
cez + cf + dgz + dh
(ce + dg)z + (cf + dh)
It is clear that is an onto homomorphism. We will show that it is 1-1. Assume (M)
is the identity transformation. That is, assume that (az + b)/(cz + d) = z. Therefore
cz 2 + (d a)z + b = 0 for every z in P1 (p). Thus, c = b = 0 and a = d. But since
ad = a2 = 1, we also know that a = 1 or a = p 1. Therefore, the only elements
33
Name:Stanislav Shur
IP:74.88.120.168
of SL(2, p) taken to the identity are the elements in the subgroup of multiples of the
identity. It follows that the kernel of is the identity of P SL(2, p), and that is a group
isomorphism, as claimed.
20. Assume that the group G is not cyclic. Then no element of the group has index p2 .
Therefore, every element other than the identity must have index p. Choose one such
element , and let H be the subgroup of G generated by . According to one of the
Sylow theorems, H must be normal in a group of order p2 , therefore in G itself. Now
let be another element of order p not in H, and let K be the subgroup of G generated
by . K is also normal in G, and by counting we see that every element of G can
be written as i j , 0 i < p, 0 j < p. Furthermore, H K = {1}, because no
element of H could generate K and vice-versa. It remains to determine the product
. We know that 1 1 = (1 ) 1 K, because K is normal. Similarly
1 1 = (1 1 H, because H is normal. Therefore 1 1 = 1, or
= . In other words, the group G is Abelian. Therefore, it is simply the product of
two cyclic groups of order p. Thus, there are exactly two groups of order p2 , the cyclic
group of order p2 and the product of two cyclic groups of order p. Both are Abelian.
21. Since S 1 T S = T r , we have S 2 T S 2 = S 1 S 1 T SS = S 1 T r S = (S 1 T S)r = (T r )r =
2
j
T r . Similarly, for any j, we have S j T S j = T r . In particular, setting j = q, we have
q
q
T = S q T S q = T r . It follows that T r 1 = 1 and therefore that rq 1 is divisible by
p; that is, rq 1 (mod p).
22. Let be the real fifth root of 2 and a primitive fifth root of unity. Then the zeros of
x5 2 are , , 2 , 3 , and 4 . An element of the Galois group is determined by its
action on and on . The group element must take to one of the roots of the equation
and to a power of from 1 to 4. So define the group element ab (1 a 4, 0 b 4)
by ab () = a and ab () = b . The mapping which takes the substitution x0 ax + b
(mod 5) to the group element ab is then an isomorphism of groups. To demonstrate
this, we show that the mapping preserves the group operation. The composition of the
substitutions x0 ax + b and x0 cx + d is x0 a(cx + d) + b acx + (ad + b). On the
other hand, ab cd () = ab ( c ) = ac and ab cd() = ab ( d ) = (ab ())d ab () =
ad b = ad+b . This group element corresponds under the mapping to the substitution
x0 acx + (ad + b) as asserted, where all the operations are taken modulo 5. It follows
that the Galois group and the group of substitutions are isomorphic and therefore that
the Galois group has 20 elements.
23. We can take x3 + x + 1 0 as the irreducible congruence modulo 5. This is irreducible,
because none of the residues modulo 5 satisfy this congruence. If we then assume that
is a solution of this congruence, the set F = {a0 + a1 + a2 2 |0 a0, a1 , a2 4, 3 =
1} is a field under the ordinary operations and has order 53 = 125.
34
Name:Stanislav Shur
IP:74.88.120.168
24. It is easiest to do this in matrix form. If we let

X=
x
,
y
X = (x, y),
and B =
a b
,
b c
then the quadratic form F can be written in the form F = X t BX, where, in general,
the exponent t indicates the transpose of a matrix. If we let
A=
then the substitution can be written in either the form X = AX 0 or X t = X 0t At. Thus,
we have F 0 = X 0t At BAX 0 . Let C = At BA, so F 0 = X 0t CX 0 . To get back to F , we
make the substitution, X 0 = A1 X, or X 0t = X t (A1 )t . But (A1)t = (At )1 . Since
this substitution turns F 0 = X 0t CX 0 into X t (At )1 CA1X = X t (At )1 AtBAA1X =
X t BX = F , the substitution X 0 = A1 X is the desired inverse substitution. Given that
the determinant of A is 1, we know that
A1 =
We can write this substitution in the form x0 = x y, y 0 = x + y.

25. If AB = 0, then 0 = det(AB) = det(A) det(B). Therefore, either det(A) = 0 or
det(B) = 0.
26. The characteristic equation of A may be written as 2 (a + d) + ad bc = 0. Thus,
we need to show that A2 (a + d)A + (ad bc)I = 0. We calculate:

a b
c d
a2 + bc ab + bd
ac + cd bc + d2
2
a b
(a + d) c d

a2 + ad ab + bd
ac + cd ad + d2
ad bc
0
0
ad bc
ad bc
0
0
ad bc
0 0
.
0 0
27. It is easiest to do this problem in matrix form. Let 1 , 2 , 3 be the three roots
of the characteristic equation of A = (aik ), that is, the eigenvalues of A. These are
not necessarily distinct nor real in general. We further suppose that to each i there
corresponds an eigenvector vi = (vi1 , vi2 , vi3 ), where the three eigenvectors are linearly
independent. Let W be the matrix formed by the vi. Finally, we let Y be the column
vector of the yi and, if ui = vi1 y1 + vi2 y2 + vi3 y3 , we let U be the column vector formed
by the ui . We can write the differential equation formed by the sum of the original
equations multiplied by the vij in the matrix form W Y 00 + (W A)Y = 0. We then
make the substitution U = W Y or, equivalently, Y = W 1 U . We then have the new
matrix differential equation U 00 + (W AW 1 )U = 0. However, W AW 1 = D, under the
conditions mentioned, is the diagonal matrix whose entries along the diagonal are 1 ,
2 , 3 . Thus, the differential equation is of the simple form U 00 + DU = 0. This is
35
Name:Stanislav Shur
IP:74.88.120.168
equivalent to the three equations u00i + i ui= 0 for i =1, 2, 3. If is real and negative,
then the solution to this equation is u = Ae t + Be t . This solution is, in general,
not bounded
If is real and positive, then the solution to the equation is
as t .
u = A sin t + B cos t, which is bounded as t . If were complex, then the
solution would be a product of an exponential and a trigonometric term, which again
would not be bounded. Finally, a deeper study of non-distinct eigenvalues shows that
if two or more of the i are equal, then we may not be able to diagonalize the matrix
and in that case the solutions would again involve exponential factors which would be
unbounded. It follows, therefore, that for the solution to make physical sense, each of
the i must be real and positive, and they all must be distinct. Once we have the three
solutions corresponding to the matrix equation U 00 + DU = 0, we recover the solutions
Y to the original equation by making the substitution Y = W 1 U .
28. The characteristic equation belonging to the quadratic form 2x2 +4xy +5y 2 is (2)(5
) 4 = 0, or 2 7 + 6 = 0. The roots of this equation are 1 = 1 and 2 = 6. To
determine the solution (x1 , y1 ) corresponding to 1 = 1, we solve (2 1)x + 2y = 0, or
x + 2y = 0. Thus x1 = 2y1 . But since we want x21 + y12 = 1, we have 4y12 + y12 = 1, or
y12 = 15 . We pick y1 = 1 and therefore x1 = 2 . To determine the solution (x2 , y2 )
5
5
corresponding to 2 = 6, we solve (2 6)x + 2y = 0, or 4x + 2y = 0. Thus y2 = 2x2 .
Since also x22 + y22 = 1, we have x22 +4x22 = 1, or x22 = 15 . We pick x2 = 15 and therefore
y2 =
2 .
5
Thus the orthogonal substitution which converts the original quadratic form
into the form u2 + 6v 2 is x =
2 u + 1 v,
5
5
y = 15 u +
2 v.
5
29. (3+4i+7j+k)(23i+jk) = (6+127+1)+(9+871)i+(3+143+4)j+(3+2+

4+21)k = 12 9i + 18j + 24k. To divide 3 +4i +7j + k by 2 3i + j k, we multiply both
dividend and divisor by 2 + 3i j + k. The denominator is then 4 + 9 + 1 + 1 = 15, while
the numerator is (6 12 + 7 1) + (9+ 8+ 7+ 1)i +(3 +14 +3 4)j +(3 +2 4 21)k =
25i + 10j 20k. It follows that the quotient is 53 i + 23 j 43 k.
30. In general, (a + bi + cj + dk)(e + fi + gj + hk) = (ae bf cg dh) + (af + be + dg
ch)i +(ag + ce+ df bh)j + (ah + de+ bg cf)k. To demonstrate that the modulus of the
product is the product of the moduli, we must show that (ae bf cg dh)2 +(af + be+
dg ch)2 +(ag +ce+df bh)2 +(ah+de+bg cf)2 = (a2 +b2 +c2 +d2 )(e2 +f 2 +g 2 +h2 ).
To do this, note that the 16 squares of the terms on the left side are precisely the 16
products on the right. Also, every other term in the expansion on the left side occurs
twice, once positive and once negative, so all of these terms cancel out.
31. By calculating all 3 3 determinants, we see that each of them has value 0. On the
other hand

2 1

5 3 = 1,
so the order of the maximal non-vanishing determinant is 2. To solve the system, we
solve the third equation for u: u = v + 8x y + 12z. We then substitute in the first
36
Name:Stanislav Shur
IP:74.88.120.168
and second equations. The first equation reduces to v + 18x y + 27z = 0, while the
second reduces to 2v + 36z 2y + 54z = 0. Since these equations are equivalent, we
can solve either one for v. We get v = 18x y + 27z. Substituting this value back into
the equation for u gives us u = 10x 15z. In these results, we have that x, y, and z
are arbitrary.
32. Since the largest non-vanishing determinant of the matrix of coefficients has order 2, the
rank of the matrix is 2. Given the solution to the system worked out in exercise 31,
we see that the solution depends on three arbitrary constants, x, y, and z. Thus if we
choose x = 1, y = 0, z = 0, we get the solution (10, 18, 1, 0, 0). If we choose x = 0,
y = 1, z = 0, we get the solution (0, 1, 0, 1, 0). If we choose x = 0, y = 0, z = 1, we get
the solution (15, 27, 0, 0, 1). These three vectors form a basis for the set of solutions.
33. The system associated to the chosen basis for the set of solutions to the original equation
is simply the system whose coefficients are those basis elements:
10u + 18v
v
15u + 27v
+ x
+ y
+ z
= 0
= 0
= 0
Since there is a 3 3 non-vanishing determinant, the rank of this system is 3. We solve

this system by solving the first equation for x, the second for y, and the third for z.
Thus, x = 10u 18v, y = v, and z = 15u 27v. A basis for the set of solutions to
this system is found by first setting u = 1, v = 0, and then setting u = 0, v = 1. Thus
the basis consists of the two vectors (1, 0, 10, 0, 15) and (0, 1, 18, 1, 27). The system
associated to this basis is
u
v
+ 10x
18x + y
+ 15z
27z
= 0
= 0.
The solution to this system, namely u = 10x 15z, v = 18x y + 27z is the same as
the solution to the original system, as worked out in exercise 31.
34. If G satisfies the usual axioms of a group, namely a set with an associative binary
operation which has an identity and in which every element has an inverse, then we see
that G satisfies Webers criteria. First, Weber states the associative law. Second, the
equation AX = B will have the unique solution X = A1 B. And if r = s , then
on multiplying both sides on the right by 1 , we get r = s. Now let us assume that
G satisfies Webers criteria. We know that the operation is associative. To find the
identity, find the unique solution to XA = A for any A. To see that X is the identity,
note that since BXA = BA, we can cancel the A to get BX = B, where B is arbitrary.
Similarly, BXB = BB, so by canceling the B on the left, we have XB = B. Thus X is
the identity. Therefore, the unique solution Y to the equation AY = X is the inverse of
A. Thus, a set satisfying Webers criteria also satisfies the modern group conditions.
35. The standard modern definition of a field is a set F with two binary operations, + and
, under the first of which the set is an Abelian group, and under the second of which
37
Name:Stanislav Shur
IP:74.88.120.168
F {0} is also an Abelian group. The only further requirement is that the distributive law
holds: a(b + c) = ab + ac. Weber has three other axioms in his definition: a(b) = ab,
(a)(b) = ab, and a 0 = 0. These three axioms can, however, be proved from the
other axioms in the field definition.
38
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER SEVENTEEN
1. Given that limx f (x + 1) f (x) = , we know that given any positive number
M, there exists a positive number N such that if x N , then f (x + 1) f (x) > M .
Therefore, for i = 1, 2, . . . n, we have f (N + i) f (N + i 1) > M . The arithmetic
mean of these n expressions also satisfies the same inequality. Thus
f (N + n) f (N )
> M,
n
or
f (N + n) f (N )
= M + ,
n
where > 0. Let x = N + n. The equation then becomes

f (x) f (N )
= M + or f (x) = f (N ) + (x N )(M + ).
xN
Therefore,
f (x)
f (N )
N
=
+ 1
(M + ).
x
x
x
Since N is fixed, we see that as x , f (x)
x approaches M + . Thus eventually,
f (x)
x > M and therefore
f (x)
= .
lim
x x
2. We assume that a > 1, Then limx ax+1 ax = limx ax(a 1) = . It follows
from the theorem in exercise 1 that
ax
= .
x x
lim
1
Also, we have limx log(x + 1) log x = limx log( x+1
x ) = limx log(1 + x ) = 0.
By the theorem in the text, we have
lim
log x
= 0.
x
3. Let > 0. We need to find > 0 such that | sin(x + ) sin x| < , whenever || < .
We choose = . Then for || < , we have
| sin(x + ) sin x| =

2 sin

1
1
1
cos x + 2 || 1 = || < = .
2
2
2
Therefore, sin x is continuous at x.
39
Name:Stanislav Shur
IP:74.88.120.168
4. Given any > 0, we know there exists an h such that if x h, then

k<
f (x + 1)
< k + .
f (x)
f (h+i)
Since this inequality is true for f (h+i1)
for i = 1, 2, . . . , n, it is true for the geometric
mean of all these quotients. Therefore,
k<
v
u
u f (h + 1) f (h + 2)
n
t
f (h)
f (h + 1)
f (h+n) 1/n
= k + , where
f (h) )
f (x) 1/xh
f (x)
= k + , or f (h) = (k + )xh ,
( f (h) )
each side to the power x1 . We get f (x)1/x
expression f (h)1/x 1 and the expression
It follows that (
f (h + n)
< k + .
f (h + n 1)
< < . We set x = h + n. Then

or f (x) = f (h)(k + )xh . We now raise
= f (h)1/x (k + )1h/x . As x , the

(k + )1h/x (k + ). It follows that
f (x)1/x approaches k + , where < < . Because is arbitrary, we see that f (x)1/x
approaches k, as desired.
5. We know that limx

limx x1/x = 1.
x+1
x
= 1. Therefore, by the theorem of exercise 4, we have
6. Given > 0, we can find n such that
1
2n2
< . But
1
n!
<
1
.
2n1
So
1
1
1
1
1
1
1
+
+
+ < n1 + n + n+1 + < n2 < .
n! (n + 1)! (n + 2)!
2
2
2
2
Therefore, any finite sum of the reciprocals of the factorials where n! > 2n1 is also less
than , and by the Cauchy criterion, the series converges.
7. Each of the ui (x) is a continuous function in a neighborhood of x = 1, and sk (x) =
Pk
P
k
k
i=1 ui (x) = x . We let s(x) =
i=1 ui(x) = limk x . We see that for any x < 1,
s(x) = 0; for x = 1, s(x) = 1; while for x > 1, s(x) is not defined. Therefore, s(x) is not
a continuous function of x in any neighborhood of x = 1. The problem with Cauchys
proof in this instance is that although given any positive , for each k there exists k
such that |sk (1 + a) sk (1)| = |(1 + a)k 1| < whenever |a| < k , there is no that
will work for all k. In fact, since the curve y = xk gets steeper near 1 with increasing k,
for given , we have lim k = 0.
8. If y = ax, then
ax+i ax
ax(ai 1)
ai 1
x
y = lim
= lim
= a lim
.
i0
i0
i0
i
i
i
0
40
Name:Stanislav Shur
IP:74.88.120.168
To calculate the last limit, we set ai = 1 + , or i = loga (1 + ), where approaches
0 with i. We need to determine now lim0 log (1+)

, but it is easier to work with the
a
= loga (1 + )1/ . As approaches 0, the argument of this

reciprocal. We have loga (1+)
logarithm function approaches e. Therefore,

lim
i0
ai 1
1
=
,
i
loga e
and
y0 =
ax
.
loga e
9. Let us assume that f (xj ) is the smallest value of the f (xi ) and that f (xk ) is the largest
of these values. Let us designate by S the sum f (x0 )(x1 x0 ) + f (x1 )(x2 x1 ) +
+ f (xn1 )(xn xn1 ). We then have f (xj )(b a) S f (xk )(b a), or f (xj )
S
ba f (xk ). By the intermediate value theorem, there is a value x [a, b] such that
S
. Therefore, S = f (x)(b a). But since x0 x xn , we can write x as
f (x) = ba
x = x0 + (b a), where 0 1. Thus we have S = (b a)f (x0 + (b a)) as desired.
10. First, we show that the quantity U has the property that all smaller x have property M.
For if there were an x < U such that x did not have property M, then U x > 2Dq for
some q, or x < U 2Dq . This means that there is an element ui in Bolzanos sequence
such that x < ui . But this contradicts the definition of the ui , that M is valid for all
x less than ui . Second, we show that there is no V > U such that all smaller x have
property M. If such a V existed, then V U > 2Dr for some r. Thus U + 2Dr < V and
therefore all x smaller than U + 2Dr satisfy property M. But this means that there is a ui
in Bolzanos sequence such that ui > U , contradicting the definition of U as the limit of
the (strictly increasing) sequence. Therefore, U is the greatest number of those of which
it can be asserted that all smaller x have property M. In other words, U is the least
upper bound of all the x which have property M.
11. Since the number 0.66 . . . 6 is in A no matter how many 6s are in the expansion, the
least upper bound of A must be greater or equal to 0.66 . . . = 23 . But if it were greater
than 23 , there would be an element in A greater than 23 , which is impossible. So the least
upper bound is equal to 23 .
12. Since (1 + 1)3 > 3, while 13 < 3, we begin Bolzanos process by trying 1 + 12 . We check
that (1 + 12 )3 > 3, so we next try 1 + 14 . In this case, (1 + 14 )3 < 3, so the first two
terms of the desired sequence are 1, 1 + 14 . We next note that (1 + 14 + 18 )3 < 3, and also
1 3
1
(1 + 14 + 18 + 16
) < 3. So 1 + 14 + 18 and 1 + 14 + 18 + 16
are the next two in the sequence.
1
1
1
After that, we try adding, in turn, 32 , 64 , and 128 , each of which gives a value too large.
1
1 3
1
1
But (1 + 14 + 18 + 16
+ 256
) < 3, so 1+ 14 + 18 + 16
+ 256
is the next number in the sequence.
1
1
1
1
1 3
+ 256
+ 2048
) < 3.
Again, if we add 512 or 1024 , the value is too large. But (1 + 14 + 18 + 16
1
1
1
1
1
In decimals, 1 + 4 + 8 + 16 + 256 + 2048 is equal to 1.4418945, which to three decimal
places is equal to 3 3.
41
Name:Stanislav Shur
IP:74.88.120.168
13. We calculate 0 (x) = memx and 00 (x) = m2 emx =
mx
.
Ae
Therefore,
(x)
emx
=
mx = A.
00 (x)
Ae
cos ny. Therefore,

Similarly, 0 (y) = n sin ny and 00 (y) = n2 cos ny = A
(y)
cos ny
=
= A.
00
(y)
A cos ny
We then have v = aenx cos ny = (x)(y). Therefore

(y)
(x)
2v 2v
(y) (x)
= 0,
+ 2 = 00 (x)(y) + (x) 00 (y) =
2
x
y
A
A
so v is a solution to the given partial differential equation.
14. We write the sine series for 12 x in the form
1 x
= a1 sin x + a2 sin 2x + a3 sin 3x + .
2
Then we calculate
ak =
Z
0
1
x
sin kx dx =
x sin kx dx.
We integrate by parts by setting u = x, dv = sin kx dx, du = dx, and v = k1 cos kx.

Then
1
Z
0
1
x
x sin kx dx =
cos kx
k
1
= cos k.
k

0
1
+
k
Z
0

1
1
cos kx dx = cos k + 2 sin kx
k
k
0
Note that if k is even, this value is k1 and if k is odd, the value if k1 . Therefore, the
sine series is given by
1
1
1
1
x = sin x sin 2x + sin 3x sin 4x + .
2
2
3
4
1
1
5
If we replace x by 2 , we get 4 = sin 2 12 sin + 13 sin 3
2 4 sin 2 + 5 sin 2 ,
or 4 = 1 13 + 15 . This is the same result we get by setting x = 1 in the power
series for arctan x and is also the result which Leibniz found by using his transmutation
theorem applied to a circle.
42
Name:Stanislav Shur
IP:74.88.120.168
k
15. We let s(x) =
n=0 un (x), sk (x) =
n=0 un (x), and rk (x) = s(x) sk (x). We want
to show that s(x) is continuous in I. So let x I and choose > 0. Then, since the
series converges uniformly in I, we can find N such that |rn (x)| < 3 for every n > N
and for every x I. Choose some n > N. Since sn (x) is continuous, there is a > 0
such that |sn (x + a) sn (x)| < 3 whenever |a| < and x + a I. Then, if |a| <
and x + a I, we have |s(x + a) s(x)| = |sn (x + a) + rn (x + a) sn (x) rn (x)|
|sn (x + a) sn (x)| + |rn (x + a)| + |rn (x)| 3 + 3 + 3 = . Therefore, s(x) is continuous
in I as claimed.
16. First, since the denominator in the definition of vk does not vanish as h approaches 0,
we have
2h
= uk .
lim vk (h) = uk + lim
h0
h0 ((k 1)h + 1)(kh + 1)
k
Second, we claim that u1 + u2 + + uk = k+1
. This is clearly true for k = 1. To prove
this by induction, we assume the truth of the statement and calculate the sum up to
k + 1:
1
k(k + 2) + 1
k
+
=
k + 1 (k + 1)(k + 2)
(k + 1)(k + 2)
2
k + 2k + 1
(k + 1)2
k+1
=
=
=
.
(k + 1)(k + 2)
(k + 1)(k + 2)
k+2
u1 + u2 + + uk + uk+1 =
k
Thus the result is true for all k. We then have uk = limk k+1
= 1. Third, we claim
k
2kh
that v1 (h) + v2(h) + + vk (h) = k+1 + kh+1 . Again, we note that this is true for k = 1.
We assume the truth of this statement for k and show that it is also true for k + 1:
k+1
X
vi (h) =
i=1
k
X
vi (h) + vk+1(h)
i=1
=
=
=
=
=
k
k+1
k+1
k+2
k+1
k+2
k+1
k+2
k+1
k+2
+
+
+
+
+
2kh
1
2h
+
+
kh + 1 (k + 1)(k + 2) (kh + 1)((k + 1)h + 1)
2kh((k + 1)h + 1) + 2h
(kh + 1)((k + 1)h + 1)
2k(k + 1)h2 + 2kh + 2h
(kh + 1)((k + 1)h + 1)
2h(k + 1)(kh + 1)
(kh + 1)((k + 1)h + 1)
2(k + 1)h
.
(k + 1)h + 1
P
k
2kh
The result is therefore true by induction. Then vk (h) = limk k+1
+ kh+1
= 1+2 = 3.
P
P
P
Finally, limh0 vk (h) = limh0 3 = 3, while uk = 1. Therefore, limh0 vk (h) 6=
P
uk , as claimed.
43
Name:Stanislav Shur
IP:74.88.120.168
17. If = (A1 , A2 ) and = (B1 , B2 ), we define < to mean that A1 is strictly included
in B1 ; that is, there are rational numbers in B1 which are not in A1 . Now, this ordering
satisfies the transitive law. For suppose < and < = (C1 , C2 ). Then A1 is strictly
included in B1 and B1 is strictly included in C1 . Therefore A1 is strictly included in C1 ,
so < . We also show the trichotomy law: for any two real numbers , , either <
or = or < . So let us suppose that 6= . We compare the sets A1 and B1 . If
one is strictly included in the other, then we are done, according to our definition. So
suppose neither is strictly included in the other. Thus, there exists a rational number
/ B1 and a rational number b with b B1 and b
/ A1 . By the
a with a A1 and a
definition of a cut, we must have a B2 and b A2 . Since a A1 and b A2 , we have
a < b. But since b B1 and a B2 , we also have b < a. This is a contradiction, and the
trichotomy law holds.
18. Let = (A1 , A2) and = (B1 , B2 ). To define + , we need to define a particular set
C1 of rationals which is the lower set of a pair defining a cut. We let C1 be the set
of all rational numbers formed by adding any element in A1 to any element in B1 . Let
C2 be all the remaining rational numbers. To show that = (C1 , C2 ) is a cut, we must
show that every number in C1 is less than every number in C2 . So suppose c C1 and
d C2 . Then c = a + b, where a A1 and b B1 . Since d is not the sum of any pair
/ B1 . Therefore, d a B2 . So d a is larger
of numbers from A1 and B1 , then d a
than any number in B1 . In particular, d a > b and c = a + b < d. Thus = (C1 , C2 )
is a cut. To show that + = + , we just note that the rational numbers formed by
adding any element in A1 to any element in B1 are the same as those formed by adding
any element in B1 to any element in A1 .
19. Suppose {i } form a bounded increasing sequence of cuts. There is then one cut , and
therefore infinitely many, such that i < for all i. So let A2 be the set of all such and
A1 be all cuts not in A2 . Then any element 1 A1 has the property that i > 1 for
some i. So 1 is less than every cut 2 A2 . We can show that there is a cut which is
either the greatest in A1 or the least in A2 . The first case is impossible, since the sequence
is increasing. Therefore A2 and is the desired limit number. Now take an increasing
sequence of real numbers defined by fundamental sequences (a1i ), (a2i ), (a3i ), . . ., which
is bounded above by the rational number B. Note that b can be thought of as the
fundamental sequence (b, b, b, . . .). Since the sequence is an increasing sequence, we can
replace each real number by an equivalent sequence for which all of the components
of (a1i ) are less than all of the components of (a2i ) which are in turn less than all of
the components of (a3i ), and so on. Now consider the sequence a of rational numbers
(a11 , a22 , a33 , . . .). This sequence is an increasing sequence of rational numbers, bounded
above by b. This sequence is a fundamental sequence, for if not, then for some , there
would be infinitely many pairs of numbers in the sequence whose distance from each
other was greater than , and thus the sequence would not be bounded. Therefore a is
a fundamental sequence of rational numbers which therefore represents a real number,
and this number is the limit of the original sequence of real numbers.
44
Name:Stanislav Shur
IP:74.88.120.168
20. Since {bi } is a fundamental sequence not defining the limit 0, there are positive numbers
r and L such that r < |bi|, |bj | < L for every i, j. In particular, for every i, j, we have
|bi bj | > r2 or |bi1bj | < r12 . Also, since {ai} is a fundamental sequence, there is a number
K such that |ai | < K for every i. So, given > 0, choose N so that |bi bj | <
|ai aj | <

a
i

bi
r2
2L
r2
2K
and
whenever i, j > N . Then

aj bj ai bi aj |bj ai bi aj |
|bj ai bj aj + bj aj bi aj |
<
=
=
2

bj
bi bj
r
r2
<

|bj (ai aj )| |aj (bi bj )|
L r2 K r2
+
=
+ = .
+
<
r2
r2
r2 2L r2 2K
2 2
Therefore, { abii } is a fundamental sequence.

21. We first show that the product sequence {ai bi } is a fundamental sequence. We know that
there exists a number K such that |ai | < K for all i and also a number L such that |bi| < L

and also
for all i. So given > 0, choose N so that for i, j > N , we have |bi bj | < 2K

|ai aj | < 2L . Then |ai bi aj bj | = |ai bi aibj +aibj aj bj | |ai(bi bj )|+|bj (ai aj )| <

+ L 2L
= 2 + 2 = . Thus {aibi } is a fundamental sequence. To show that the
K 2K
product makes sense, we need to show that if {a0i } is any other sequence equivalent to A
and if {b0i } is any other sequence equivalent to B, then {a0i b0i } is equivalent to {ai bi }. So

and |bi b0i | < 2K
, where
given > 0, choose M so that for i > M , we have |ai a0i | < 2L
L and K are chosen as above to work for both the original and primed sequences. Then

+L 2L
= , and the
|ai bi a0ib0i | = |aibi ai b0i +ai b0i a0i b0i | |ai (bi b0i)|+|b0i (ai a0i )| < K 2K
two sequences are equivalent. Finally, note that if C = {aibi }, then B = {bi } = { aai bi i },
so B = C
A , where the division is defined as in exercise 20.
22. We have
dw
du + i dv
=
=
dz
dx + i dy

=

u
x
u
x
v
+ i x
v
+ i x

u
x
dx +
dx +
dy + i
v
x
dx +
dx + i dy
v
y
i u
y
i dy
dx + i dy
(dx + i dy)
dx + i dy
R
u
y
u
x
v
y
dy

v
+ i x
dx +
u
x
v
+ i x
i dy
dx + i dy
u
v
+i .
x
x
RR
23. If curl = 0, then C dr = A curl da = 0, where C is a closed curve bounding

the region A. Therefore, if is any curve connecting two points, and if designates
any other curve connecting the two points but in the opposite direction, then the curve
C consisting first of and then of is a closed curve, over which the integral is 0.
But this integral is the difference of the integrals over and over . It follows that
the integrals over and are equal, so the
integral depends
only on the endpoints of
RR
RRR
the curve. Similarly, if div = 0, then
S da =
V div dV = 0, where S is
45
Name:Stanislav Shur
IP:74.88.120.168
a closed surface bounding the solid region V . Therefore, if , are any two surfaces
with the same boundary curve, the sum of the two surfaces, with the second taken
in the opposite direction, is a closed surface. The integral over this closed surface is
zero, which implies that the integrals over the two surfaces and are equal. Thus, the
integral depends only on the boundary curve and not the particular surface of which the
curve is the boundary.
46
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER EIGHTEEN
1. We want to find the equation y = ax + b giving the best linear function representing
the data points (2.0, 2.5), (4.0, 4.5), (5.0, 7.0), (6.0, 8.5). Thus, we have four equations to
determine the two values a and b. These equations are 2.5 = 2.0a+b, or 2.5+b+2.0a =
0; 4.5 = 4.0a + b, or 4.5 + b + 4.0a = 0; 7.0 = 5.0a + b, or 7.0 + b + 5.0a = 0;
and 8.5 = 6.0a + b, or 8.5 + b + 6.0a = 0. To use the method of least squares,
we want to take the two partial derivatives of the sum of the squares function and
set them equal to 0. The sum of the squares is given by S = (2.5 + b + 2.0a)2 +
(4.5 + b + 4.0a)2 + (7.0 + b + 5.0a)2 + (8.5 + b + 6.0a)2 . We have S
b = 2(2.5 +
b + 2.0a) + 2(4.5 + b + 4.0a) + 2(7.0 + b + 5.0a) + 2(8.5 + b + 6.0a) and S
a =
2(2.5+b+2.0a)2.0+2(4.5+b+4.0a)4.0+2(7.0+ b+5.0a)5.0+2(8.5+ b+6.0a)6.0.
Setting the two equations equal to 0 and simplifying gives the following system of two
equations in two unknowns: 4b + 17a = 22.5 and 17b + 81a = 109. Solving this system
using Cramers rule gives
b=

22.5

109

4

17
17
30.5
81
= 0.87 and a =
=

17
35
81

4

17
22.5
53.5
109
=
= 1.53.
35
35
The desired straight line is then y = 1.53x 0.87.

2
e(x /c ) , and y 00 = 32
e(x /c ) + 4x
e(x /c ) .
2. If y = c1 e(x /c ) , then y 0 = c32x
c
c5
An inflection point occurs in this function occurs when y 00 = 0. Thus, we must solve
2
2
4x
2
= 3
. We get 2x2 = c2 , or x2 = c2 , or, finally, x = c2 .
c5
c
3. Using a table of the normal curve, we find that a percentile rank of 75 corresponds to
a z-score of 0.675. Therefore, one probable error from the mean, which corresponds to
a percentile rank of 75, is at distance approximately 0.675 from the mean, where is
the standard deviation. By
the result of exercise 40, a distance c of one modulus from
the mean corresponds to 2.
4. To construct a histogram of children of parent of height 68.5, we can use the following
table:
< 62.7
62.7 63.7
63.7 64.7
64.7 65.7
65.7 66.7 66.7 67.7
11
16
25
31
67.7 68.7
68.7 69.7
69.7 70.7
70.7 71.7
71.7 72.7
> 72.7
34
48
21
18
The median value calculated from the original table is 68.2 inches.
47
Name:Stanislav Shur
IP:74.88.120.168
Note that it is not possible from the data given to construct a histogram of heights
of parents of children of height 69.2. One can construct a histogram of the number of
children of height 69.2 inches whose parents are of various heights. This would be come
from the following table:
< 64
64 65 65 66
67 68
13
38
71 72
> 72
68 69 69 70 70 71
48
66 67
33
18
The median value calculated from the original table is 68.5.

5. The histogram for the heights of all adult children can be constructed from the following
table:
< 61.7
61.7 62.7 62.7 63.7 63.7 64.7 64.7 65.7
5
32
59
48
65.7 66.7 66.7 67.7 67.7 68.7 68.7 69.7

117
138
120
167
70.7 71.7 71.7 72.7 72.7 73.7

64
41
69.7 70.7
17
99
> 72.7
14
The median value here is 68.2. By calculation, the standard deviation is 2.54.
The histogram for the heights of all parents can be constructed from the following table:
< 64
64 65
65 66
66 67
67 68
12
20
33
70 71
71 72
22
11
68 69 69 70
49
41
72 73 > 73
6
The median value here is 68.5. By calculation, the standard deviation is 1.94.
6. It is actually not entirely clear what Playfair met. The price of wheat in 1821 was 54
shillings, but the graph for wages does not extend that far. If we assume that the wages
are, say, 32, then the ratio of the two prices is 1.69. However, in the period 178590, it
appears that the same ratio is 42:25=1.68, so wheat would have been relatively cheaper
48
Name:Stanislav Shur
IP:74.88.120.168
at that point. But in any other period, the ratio of the price of wheat to the wages of
a mechanic are higher than those two values, so the price of wheat is relatively higher
than either the assumed ratio in 1821 or the ratio in 178590.
7. There are four countries in which the line from population to revenue rises. These are
Spain, Britain, Portugal, and the United Provinces. But the line for Britain is clearly
much steeper than that for the other three countries. It follows that the revenues of
Britain in proportion to its population are greater than for any other country. In other
words, the people of Britain are excessively taxed.
8. It appears that the size of the African and European domains are each roughly onequarter of the total area, while the Asiatic domains from about one-half the total area.
9. Reading the histogram using the scale of one square equals 10 flowers, we see that
approximately 24 flowers had seven petals and 133 had five petals.
49
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER NINETEEN
1. Assuming that angle B is acute, we drop a perpendicular from angle B to the opposite
side b, intersecting that side at point D. Designate BD by h, AD by r, and DC by b r.
In right triangle BDC, we have cos a = cos h cos(b r). In right triangle BDA, we have
r
cos c
cos a
cos A = tan
tan c and cos c = cos h cos r. Therefore, cos h = cos r = cos(br) . So
cos c cos(b r)
cos c(cos b cos r + sin b sin r)
=
= cos c cos b + cos c sin b tan r
cos r
cos r
= cos c cos b + cos c sin b tan c cos A = cos c cos b + sin b sin c cos A.
cos a =
2. Note that in chapter 3, the triangle formulas assumed that the sides were measured in
degrees. Naturally, we can also measure the sides in radians and get the same formulas.
But if we measure the sides in linear measure, then we can only use the formulas as
written if the radius of the sphere is 1, for in that case the linear measure of an arc is
equal to its radian measure. If the sphere has radius K, then the linear measure x of an
x
arc whose radian measure is r is given by x = Kr. Thus, r = K
, and we can substitute
for the radian measure, in any of the formulas, including the one in exercise 6, the linear
measure divided by K. So the formula of exercise 6 becomes
cos
a
b
c
b
c
= cos cos
+ sin sin cos A.
K
K
K
K
K
3. The power series for y = cosh x is given by y = 1 +

series for y = sinh x is given by y = x +
x
3!
x
5!
x2
2!
x4
4!
+ , while the power
+ . If we substitute the terms from

a2
=
2K 2
2
2
2 2
b
A
+ Kc 2 + bKc4 2bcKcos
.
2
K2
2
still has denominator K
the power series through the second power into Taurinuss formula, we get 1 +
2
b
c
b c
a
(1 + 2K
2 )(1 + 2K 2 ) K K cos A. This simplifies to K 2 =
If we multiply through by K 2 and neglect the term which
because this term becomes much smaller than the other terms as K we get
a2 = b2 + c2 2bc cos A, exactly the law of cosines.
4. Since sin B =
1
cosh x ,
B
tan =
2
we have sin B =
s
1 cos B
=
1 + cos B
2
.
ex +ex
Then cos B =
v
u
u 1 ex ex
u
ex +ex
t
x
x
1+
e e
ex+ex
1 sin2 B =
ex ex
.
ex +ex
So
2ex 2x
= e
= ex
2ex
as desired. This argument works in reverse as well, so the two results are in fact equivalent.
50
Name:Stanislav Shur
IP:74.88.120.168
5. Let us write y = (x). Given that tan y2 = ex, we have tan2

e2x + 1. Therefore,
1
y
cos =
,
2
e2x + 1
so
y
sin =
2
y
2
= e2x, so sec2
y
2
ex
=
.
e2x + 1
e2x + 1
1
Then
1
2ex
y
ex
2
1
y
= 2x
= x
.
=
sin y = 2 sin cos = 2
x
2x
2x
2
2
e
+1
e +e
cosh x
e
+1 e
+1
Also.
cos y = cos2
1
e2x
1 e2x
ex ex
y
y
sin2 = 2x
2x
= 2x
= x
= tanh x.
2
2
e
+1 e
+1
e
+1
e + ex
To get the power series through the terms of degree 2, note that
1
1
sin (x) = (cosh x)1 = (1 + x2 + )1 = 1 x2
2
2
and
x + 16 x3 +
1
1
= (x + x3 + )(1 + x2 + )1
cos (x) = tanh x =
1 2
6
2
1 + 2x +
1
1
= (x + x3 + )(1 x2 + ) = x +
6
2
6. Rewrite the first formula as sin A cot (b) = sin B cot (a). Then we note that

x
1 2
cos (x)
x
=
x
1
cot (x) =
sin (x)
2
1 12 x2
1
1
x 1 + x2 x,
2
where the smaller the value of x, the better the approximation. If we use this approximation in the first formula, we get sin A b = sin B a, or sina A = sinb B , the law of
sines. Next, we rewrite the second formula in the form cos A cos (b) cos (c) sin (a) +
sin (b) sin (c) = sin (a). If we then approximate the quantities in the formula by the
terms of their power series up to degree 2, we get

1
1
cos A b c 1 a2 + 1 b2
2
2

1
1
1 c2 = 1 a2 .
2
2
If we multiply through by 2 and simplify, we get 2bc cos A a2bc cos A b2 c2 + 12 b2 c2 =

a2 . If we then neglect terms which have at least three linear factors because we are
assuming these are all small we get a2 = b2 + c2 2bc cos A, the law of cosines.
51
Name:Stanislav Shur
IP:74.88.120.168
7. From a sin(A + C) = b sin A and the law of sines, we have sinb B = sina A = sin(A+C)
.
b
Therefore, sin(A + C) = sin B, or, interchanging A and B, sin(B + C) = sin A. From
cos A + cos(B + C) = 0, we have cos(B + C) = cos A. The sine result implies that
either A = B + C or that A = (B + C). The cosine result shows that the second
equation is the correct one. Thus, A + B + C = .
8. Consider the plane z = a (a k) with coordinates u, v, sitting above the hemisphere
x2 + y 2 + z 2 = k 2 . If we take an arbitrary point on the plane, say (u, v, a) and connect
it by a straight line to the origin, then we can calculate the point (x, y, z), where the
line intersects the hemisphere. This line has direction vector (ut, vt, at). Its intersection
point (x, y, z) with the hemisphere is found by setting the length of that vector equal to
2
k 2 . We get u2 t2 + v 2 t2 + a2 t2 = k 2 , so t2 = u2+vk2 +a2 , and t = 2 k 2 2 . Thus we have
a +u +v
uk
,
x = ut =
2
a + u2 + v 2
y = vt =
a2
vk
,
+ u2 + v 2
and
ak
z = at =
.
2
a + u2 + v 2
9. To do this problem requires knowledge of Gausss formula for curvature. In particular,

we need to know that the curvature k(p) of a surface at p is the product zxx(p)zyy (p),
where the surface is given by the equation z = z(x, y) and the axes are chosen so that
p = (0, 0, 0) and zx (0, 0) = zy (0, 0) = zxy (0, 0) = 0. Since the sphere has constant
curvature, we can calculate it just at the origin. Thus,
q we write the equation of the
2
2
2
2
sphere in the form x + y + (z + k) = k , or z = k 2 x2 y 2 k. Then zx =
x
k2 x2y2
and zxx =
y2 k2
(k2x2 y2)3/2
. It follows that zxx(0, 0) = kk3 = k1 . Similarly,
zyy (0, 0) = k1 . Therefore, the curvature of the sphere is ( k1 )2 = k12 . If we replace k by

ik in this expression (taking a sphere of imaginary radius), then the curvature becomes
1
= k12 .
(ik)2
10. We have
1
a+r
1
a+u
1
a2 u2 + v
= k ln
,
, s = k ln
, and t = k ln
2
ar
2
au
2
a2 u2 v
where u = r cos and v = r sin . We calculate

a+r
tanh = tanh ln
k
ar
=
a+r
ar
ar 2 + a+r
a+r
ar
ar a+r

a+r 1/2
ar 1/2
ar
a+r
=
1/2

a+r
ar 1/2
+
ar
a+r
2
2
2
(a + r) 2(a r ) + (a r)2
1/2 !
(a + r)2 (a r)2
A similar calculation shows that tanh ks =

cosh x = 1
, we have
u
a
r cos
a
r
4r2
= .
4ar
a
and tanh kt =
v
.
a2 u2
Since
1tanh2 x
cosh
1
s
=q
k
1
u2
a2
a
=
,
a2 u2
52
Name:Stanislav Shur
IP:74.88.120.168
1
t
a2 u2
a2 u2
cosh = r
=
=
,
2 u2 v 2
2 r2
k
a
a
v2
1 a2 u2
and
cosh
=q
k
1
r2
a2
a
.
a2 r2
Therefore,
t
a
a
a2 u2
s
=
= cosh .
cosh cosh =
k
k
k
a2 u2 a2 r2
a2 r2
11. We first multiply the first two quantities: (2i + 3j 4k)(3i j + k) = (2 9)[ij] + (2 +
12)[ik] + (3 4)[jk]. We then multiply 11[ij] + 14[ik] [jk] by i + 2j k. Since any
repeated product of a unit is 0, we get 11[ijk] + 28[ikj] [jki]. But [ikj] = [ijk] and
[jki] = [ijk]. Thus this sum is equal to (11 28 1)[ijk] = 18[ijk].
P
12. When we multiply the expressions ji i together, the only non-zero terms are those
where we take distinct i from each of the n expressions. In each such case of a product
of distinct i , we must permute the factors to get the single unit [12 n ] of order
n. We will then get a positive sign if the permutation is even and a negative one if the
permutation is odd. Each coefficient of the product of distinct i will have one entry
from each row and each column of the matrix (ij ). Thus the total coefficient of the
nth order unit will be the sum of all possible products of n terms, each one having a
single factor from each row and column, where the sign of each term is positive if the
permutation needed to put that product into the natural order of the column numbers is
even and negative if it is odd. But that sum of products is exactly det(ij ) as claimed.
13. By trial, we see that the path W RST V JHGBCDF KLM N P OZXW passes through
each vertex exactly once and returns to the starting point at W .
15. Suppose we have a minimal map requiring 5 colors which contains a digon. Remove one
boundary line from the digon, thus merging it with one of its neighbors. Then we have a
map with one fewer country than the original one. Thus it can be colored with 4 colors.
Now replace the boundary line. The digon is now bounded by only two countries, each
colored differently. Thus there is color available (of the four colors) to color the digon.
This contradicts the original hypothesis.
53
Name:Stanislav Shur
IP:74.88.120.168
CHAPTER TWENTY
1. To show that the tables determine an associative algebra over the real numbers, we
must check associativity for the basis elements. For each algebra, that means checking
associativity for the triples iij, iji, ijj, jii, jij, and jji. For each of the algebras
defined here, associativity does hold for all of those triples. There are other associative
algebras of dimension 2 over the real numbers. For example, the complex numbers are
such an algebra. If we take the basis for the complex numbers to be i, j, then the four
multiplications are i i = i, i j = j i = j, and j j = i.
2. The matrix A =
3. The matrix A =

0 1
0 0
1 1
0 0

satisfies A2 = 0 and is therefore nilpotent.

satisfies A2 = A and is therefore idempotent.
4. If we multiply both the numerator and denominator of 1/(1+2i3j+4k) by the conjugate

quaternion 12i+3j 4k, the resulting denominator is the integer 1+4+9+16 = 30. It
1
1
1
1
2
(12i+3j 4k) = 30
15
i+ 10
j 15
k.
follows that the desired multiplicative inverse is 30
5. The description of the real number s requires a complete knowledge of the set E in
advance. In some sense, the element s represents a subset of E, and sets with subsets as
elements are inconsistent.
6. If a set A is well-ordered and we have two elements , with 6= , then consider the
subset B = {, }. By the well-ordering principle, B has a least element. If this element
is , then < . If this element is , then < . In any case, one of the three relations
= , < , or > must hold.
7. The separation axiom requires a definite propositional function to define a set. But
the function P defining the set of people whose hair is cut by the barber is not definite,
because the laws of logic do not determine whether P (x) holds for a particular x, namely
the barber.
8. Let z = x + iy, and assume y > 0. We want to show that the imaginary part of f (z) is
also positive. We calculate:
f (x + iy) =
a(x + iy) + b
ax + b + iay
acx2 + (ad + bc)x + bd + i(ady bcy)
=
=
.
c(x + iy) + d
cx + d + icy
(cx + d)2 + c2 y 2
Given that ad bc = 1, we see that the imaginary part of f (z) is

value is clearly positive.
y
.
(cx+d)2+c2y2
This
54
Name:Stanislav Shur
IP:74.88.120.168
9. To find the sum of P1 = (2, 5) and P2 = (4, 9), we first find the line connecting them.
Its equation is y = 2x + 1. To find where this his line intersects the curve again, we
need to solve the equation (2x + 1)2 = x3 + 17, or x3 4x2 4x + 16 = 0. Given
that x = 2 and x = 4 are solutions, we easily find that x = 2 is the third solution.
Thus the point P30 = (2, 3). Therefore the sum is P3 = (2, 3). To double (2, 3),
we need the tangent line to the curve at that point. We calculate that 2y dy = 3x2 dx,
so dy/dx = 3x2 /2y = 12/6 = 2. The tangent line is then given by y = 2x + 7. We
next find the intersection of this line with the curve by solving (2x + 7)2 = x3 + 17, or
x3 4x2 28x 32 = 0. The solution to this equation other than the double solution
at x = 2 is x = 8. Therefore, the point P30 = (8, 23) and the point P3 which is double
(2, 3) is the point (8, 23).
10. We first double (3, 8). The slope of the tangent line to the curve at (3, 8) is found by
calculating 2y dy = (3x2 43) dx, or dy/dx = 3x2 43/2y = 16/16 = 1. The tangent
line is then y = x + 11. To find the second intersection of this line with the curve,
we solve (x + 11)2 = x3 43x + 166 or x3 x2 21x + 45 = 0. Since 3 is a double
root, the other root is found to be x = 5, and the point of intersection is (5, 16).
Therefore, the point which is equal to 2 (3, 8) = (5, 16). To find 3 (3, 8), we add
(3, 8) and (5, 16). The line connecting these two points is y = 3x 1. To find the
third intersection of this line with the curve, we solve (3x 1)2 = x3 43x + 166, or
x3 9x2 37x+1650. Given that 3 and 5 are roots, we easily find the third root to be
x = 11. The intersection point is then (11, 32) and 3 (3, 8) = (11, 32). We next add
(3, 8) and (11, 32). Again, the line connecting these two points is y = 5x+23. To find
the third intersection of this line with the curve, we solve (5x + 23)2 = x3 43x + 166,
or x3 25x2 +187x 363 = 0. Since 3 and 11 are roots of this equation, we find the third
root to be x = 11 and therefore the third intersection point is (11, 32). It follows
that 4 (3, 8) = (11, 32). It is now clear that the sum of 3 (3, 8) and 4 (3, 8) is the
additive identity, namely, the point at infinity. Thus the order of (3, 8) is 7. It follows
that 5 (3, 8) = (5, 16) and 6 (3, 8) = (3, 8) (or, of course, one can calculate these
values directly).
11. The subgroup of multiples of the identity in SL(2, 8) consists only of the identity, since in
the field with 8 elements, the only non-zero solution to m2 = 1 is m = 1. Thus the order
of P SL(2,
8) isequal to the order of SL(2, 8). To calculate that order, note that in any
a b
matrix
in the group, ad bc = 1. It follows that, if a 6= 0, then d = (1 + bc)/a.
c d
Thus, there are 7 possible choices for a and 8 possible choices for b and c, for a total
of 7 8 8 = 448 choices. On the other hand, if a = 0, we have the equation bc = 1.
Thus, there are 8 arbitrary choices for d and 7 arbitrary choices for c (which of course
cannot equal 0). Thus in this case there are 8 7 = 56 choices. The total number of
possibilities is then 448 + 56 = 504, and that is the order of SL(2, 8) and also the order
of P SL(2, 8).
12. To calculate the order of P SL(3, 4), we begin by calculating the order of GL(3, 4). Note
that an element of GL(3, 4) can be thought of as a non-singular linear transformation T
55
Name:Stanislav Shur
IP:74.88.120.168
acting on the 3-dimensional vector space V over GF (4). Let {1 , 2 , 3 } be an ordered

basis of V . Any non-singular linear transformation T takes {1 , 2 , 3 } into an ordered
basis {T (1), T (2 ), T (3)}, and given an ordered basis {1, 2 , 3 }, there is a nonsingular linear transformation T such that T (1 ) = 1, T (2 ) = 2 , and T (3) = 3 . So
there are exactly as many elements of GL(3, 4) as there are ordered bases of V . To count
ordered bases of V , we note that a vector in V is a 3-tuple with elements in GF (4). So
there are 43 1 choices for the first vector 1 in the basis (because (0, 0, 0) cannot be a
basis element). Once 1 is chosen, we can take 2 to be any vector which is not in the
subspace spanned by 1 , i.e. any vector which is not a multiple of 1 . So since there
are 4 multiples, there are 43 4 choices for 2. Similarly, 3 can be any vector which is
not in the subspace spanned by 1 and 2 , i.e., not a linear combination of 1 and 2.
Since there are 42 such linear combinations, there are 43 42 choices for 3 . So there
are (43 1)(43 4)(43 42 ) = 63 60 48 = 181, 440 ordered bases for V and therefore
that many elements in GL(3, 4). Now SL(3, 4) is the subgroup of GL(3, 4) consisting of
those elements with determinant 1. Since the set of elements with determinant is a
coset of SL(3, 4), and there are three possible non-zero determinants, it follows that the
order of SL(3, 4) is 13 the order of GL(3, 4), or 63 60 16 = 60, 480. Finally, P SL(3, 4)
is the factor group of SL(3, 4) by the subgroup of multiples of the identity. Since this
subgroup consists of elements I, where 3 = 1, and since each of the three non-zero
elements of GF (4) has the property that its cube is 1, the subgroup of multiples of the
identity has order 3. It follows that the order of P SL(3, 4) is 13 the order of SL(3, 4),
namely 63 20 16 = 20, 160. The order of A8 is 12 8! = 20, 160, so these two groups have
the same order.
13. By the normal rules for composing cyclic permutations, we have
AB = (1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11)(5, 6, 4, 10)(11, 8, 3, 7) = (1, 2, 3, 8, 4, 11, 9, 10, 6, 5, 7),
BA = (5, 6, 4, 10)(11, 8, 3, 7)(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11) = (1, 2, 7, 3, 10, 8, 9, 5, 4, 6, 11),
AC = (1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11)(1, 12)(2, 11)(3, 6)(4, 8)(5, 9)(7, 10)
= (1, 12, 2)(3, 7, 11)(4, 9, 6)(5, 10, 8),
and
CA = (1, 12)(2, 11)(3, 6)(4, 8)(5, 9)(7, 10)(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11)
= (1, 11, 12)(2, 6, 10)(3, 8, 5)(4, 9, 7).
14. The graph will consist of an outer pentagon, an inner pentagon, and a dot in the center,
with each vertex of the outer pentagon connected to two adjoining vertices and to two
vertices of the inner pentagon, and with each vertex of the inner pentagon connected to
two adjoining vertices, two vertices of the outer pentagon and the dot in the center.

15. Because 12
= 924, if the woman has no discriminating ability, the probability of
6
picking
six cups correctly is 1/924, and the probability of picking five cups correctly is

6
6 5 /924 = 36/924. The total probability is then 37/924 = 0.04.
56
Name:Stanislav Shur
IP:74.88.120.168
16. Under the hypothesis that the lady has no discriminatory ability, that is, that p = 12 ,
there are 210 possible selections she could make of which cup in each pair had the
milk added first. In other words, the probability of her making any given
selection is
1
1 10
10
= ( 2 ) . But of the 2 possible selections she could make, there are 10
8 which have
210
8 selections correct (and 2 incorrect). Thus P (X = 8|p = 1/2) =

10
8
(1/2)10 .
17. Here we are assuming that the lady has probability p of choosing the correct cup of tea
in each pair. Thus, the probability of making k correct choices out of 10 and therefore
10 k incorrect choices is pk (1 p)10k . But of all the selections she could make, for a
given k there are 10choosek selections which have k correct choices and10 k incorrect
k
10k
.
choices. Therefore, the probability of getting exactly k correct choices is 10
k p (1[)
So if our critical region consists of 8, 9, or 10 correct choices, the probability of an
observation
in that region is the sum of the probabilities for the individual numbers,
10 9
8
2
10
namely 10
8 p (1 p) + 9 ; (1 p) + p . That is, this value is the probability of
an observation in the critical region, assuming the lady has probability p of choosing
correctly. In other words, this expression is the power P (R|H2 ).
18. We prove this by induction. We know that the first differences of first degree polynomial
functions are constant. So assume that the nth order differences of nth degree polynomial
functions are constant. Then suppose we have a polynomial of degree n + 1. If we look
at all terms of lower degree, the (n + 1)st order differences are all 0 by the induction
hypothesis. So we just need to consider the term xn+1 . The first differences of that
function are given by (x + 1)n+1 xn+1 = (n + 1)xn + terms of lower degree. Since
the polynomial on the right is one of degree n, its nth order differences are constant.
Therefore, the (n + 1)st order differences of the original polynomial xn+1 are constant,
and by induction, the general theorem is proved.

19. The pyramidal numbers are numbers of the form n3 , (n 3), and can therefore be
expressed in the form of a cubic polynomial: 16 n3 12 n2 + 13 n. These numbers are the
numbers 1, 4, 10, 20, 35, . . .. Their first differences are the triangular numbers 3, 6, 10,
15, . . .. Their second differences are the integers 3, 4, 5, . . ., and their third differences
are all constantly 1. Thus to calculate the pyramidal numbers, one starts with the third
differences, notes that the first second difference is 3, the first first difference is 3, and
the first pyramidal number is 1. We can then use the Difference Engine to calculate
by finding in turn the integers, the triangular numbers, and the pyramidal numbers by
repeated addition.
2
1
1
20. We first calculate for n = 1: 12 21
2+1 + B2 ( 2! ) = 0. Therefore 6 + B2 = 0, and B2 = 6 .
1
432
3
1
Next, we take n = 2: 12 41
4+1 + 6 2 + B4 ( 4! ) = 0. Therefore, 10 + 3 + B4 = 0,
65432
1
1 6
1 654
. Finally, take n = 3: 12 61
= 0. So
and B4 = 30
6+1 + 6 ( 2 ) 30 432 + B6
6!
5
1
5
1
14 + 2 30 + B6 = 0, and B6 = 42 .
57
Name:Stanislav Shur
IP:74.88.120.168
21. Since the machine begins in state q1 and initially reads a 1, according to instruction (a)
the machine then reprints that 1, moves one square to the right, and remains in state
q1 . Thus the tape still has the two ones it had initially. Next, the machine reads the
second 1, so it reprints that 1, moves one further square to the right, and still remains
in state q1 . The tape still has just the two ones. But since the next square to the right
is blank, we follow instruction (b). The machine thus prints a 1 in that blank square,
moves one further square to the right, and changes to state q2. There are now three
ones on the tape and no further instructions. So the machine stopes with a tape with
three ones. Note that if the tape initially had n ones, then it would remain in state q1
through n iterations of instruction (a). After each iteration, the tape will still have n
ones. But after these n iterations, the machine will be reading a blank square, so it will
follow instruction (b) by printing one further 1 and then stopping. Thus the tape will
end with n + 1 ones, and the machine represents the function f (n) = n + 1.
22. We are to begin with a tape which is all blank except for an initial 0, n 1s beginning
at the second square, and a 0 in the following square. We want to end with a tape in a
similar configuration except that there are 2n 1s. Therefore we must get the machine to
copy each of the given 1s over at the end of the string. One way to do this is to begin by
having the machine read the leftmost 1, change it to an x (so that it is marked), then
move along the row until it gets to a blank square. It should put a 1 in that blank square.
This is the copy of the first 1. The machine should then move back until it reaches the
x, change that back to a 1, move one square to the right, change that 1 to an x, and
then continue to the right until it reaches a blank. It should then put a 1 in that square
and move back to the left until it again reaches an x. This process should be repeated
until all the initial 1s have been changed into xs and back to 1s. The configuration will
then be a 0 followed by n 1s followed by a 0 followed by another n 1s. We then must
eliminate the middle 0 and put a 0 at the end. The actual states and instructions can
now be specified, where q1 is the initial state and where the machine begins by reading
the leftmost 1 on the tape.
a. If the machine is in state q1 and reads a 1, it prints an x, moves one square to the
right, and changes to state q2 .
b. If the machine is in state q2 and reads a 1, it moves one square to the right and
remains in state q2 . If the machine is in state q2 and reads a 0, it moves one square
to the right, and changes to state q3.
c. If the machine is in state q3 and reads a blank, it prints a 1, moves one square to the
left, and changes to state q4 .
d, If the machine is in state q4 and reads either a 1 or a 0, it moves one square to the
left and remains in state q4. If it is in state q4 and reads an x, it changes the x to a
1, moves one square to the right, and changes to state q5 .
e. If the machine is in state q5 and reads a 1, it prints an x, moves one square to the
right, and changes to state q6 . If it is in state q5 and reads a 0, it moves one square
to the right and changes to state q7 .
f. If the machine is in state q6 and reads a 0 or a 1, it moves one square to the right,
58
Name:Stanislav Shur
IP:74.88.120.168
and remains in state q6. If it is in q6 and reads a blank, it prints a 1, moves one
square to the left and changes to state q4 .
g. If the machine is in state q7 and reads a 1, it replaces it by a 0, moves one square to
the left and changes to state q8 .
h. If the machine is in state q8 and reads a 0, it replaces it by a 1, moves one square to
the right and changes to state q5 .
59

History of Mathematics Brief Version Victor Katz

Enviado por

Dados do documento

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

History of Mathematics Brief Version Victor Katz

Enviado por

Direitos autorais:

Formatos disponíveis

Name:Stanislav Shur

(double first line)

(double third line)

(double fourth line)

(halve first line)

(invert third line)

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

7 = 8. Since 19 8 = 2 38 , the correct answer is

7. (x + 23 x) 13 (x + 23 x) = 10. In this case, the obvious choice for x is x = 9. Then 9

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

9. Since x must satisfy 100 : 10 = x : 45, we would get that x =

the scribe breaks

that r = a2 . The length T of the long transversal is then T = r 3 = ( a2 )( 74 ) = 78 a. The

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

19. Because (1; 25)2 = 2; 00, 25, we have

2; 00, 25 0; 00, 25 1; 25 (0; 30)(0; 00, 25)(1/1; 25).

An approximation to the reciprocal of 1;25 is 0;42,21,11. The

22. v + u = 1; 48 = 1 45 and v u = 0; 33, 20 = 59 . So 2v = 2; 21, 20 and v = 1; 10, 40 = 106

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

Therefore the oblong number n(n + 1) is double the

and the summands are the triangular numbers Tn1 and Tn .

+ 1 = 4n2 + 4n + 1 = (2n + 1)2 .

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

Thus 1 : x can be expressed in the form (1, 1, 1, . . .).

= 2. The second solution is BE + ES = AE + ES = AS. This value is

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

Thus the octagon has area A2 = 8 12 b2 h2 = 4b2 h2 = 2 2 = 2.828427. If we continue

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

The diameter then equals 7,787 miles.

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

If Pi is the perimeter ofthe ith inscribed polygon, then

= 0.8660254. Then repeated use of the algorithm gives us:

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

4. We can prove the inequality

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

Since the area of the inscribed triangle is a

the result is established.

13. The focus of y 2 = px is at ( p4 , 0). The length of the latus rectum is 2 p p4 = p.

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

Therefore the center of the ellipse is at (a, 0) and b2 =

The hyperbola can be treated

which is the slope of the diameter P G, as desired.

b. Given that the coordinates of P are (x0 , y0 ), it follows that tan =

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

part (c), we get the equation

AC = a2 b2 (sin cos cos sin )2 = a2 b2 sin2 ( ).

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

Since these two cosines are equal, so are the angles.

3R = 103; 55, 23; crd 30 =

R(2R crd 120 ) = 31; 03, 30;

Similarly, crd 165 = 118; 58, 25 and crd 7 12 = 7; 50, 54.

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

28. L(, ) = 180 + 2(, ). When = 60 and = 36 , we calculate that sin =

34. When = 75 , we need to find so that sin

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214

2. To solve x + y = 20; xy = 96, we set x = 10 + z, y = 10 z. Then 100 z 2 = 96; z 2 = 4;

5. To solve x y = 10, x3 y 3 = 2170, set x = z + 5 and y = z 5. It follows that

It follows that this latter expression must be

6. To solve x + y = 20, x3 + y 3 = 140(x y)2 , set x = 10 + z, y = 10 z. Then

Address:8617 21st avenue,Brooklyn,NY,US Zip Code:11214