Академический Документы
Профессиональный Документы
Культура Документы
CHAPTER 1
a1 x1 + a2 x2 + . . . + an xn = b.
x + 2y = 5
2x + 3y = 4
x+y+ z=1
2x y + z = 3
3x
+ 2z = 4
has, for example, the solutions (x, y, z) = (2, 0, 1) and (x, y, z) =
(4, 1, 4).
(c) The system
(
x+y=2
x+y=3
has no solutions, for if there were a solution (x, y), we would have
2 = x + y = 3, which is false.
3
x + 2y = 5
2x + 3y = 4
(
(1)
(2)
(3)
(
x + 2y = 5
7y = 14
(
x + 2y = 5
y=2
The second equation in the last system says that y = 2. We can then
eliminate y from the first equation by substituting this value for y, so
that (3) is equivalent to
(4)
x+22=5
y=2
(5)
(
x=1
y=2
It follows that the system has the unique solution (x, y) = (1, 2).
The elimination we performed to obtain (2) from (*) can be recognized as the following operation: the first equation in (*) is multiplied
by 2, and is then added to the second equation. The coefficient of x
in the resulting sum of equations is then zero, i.e. the variable x is
eliminated there:
(
0 x + (2 2 + 3)y = 2 5 + 4
(
x + 2y = 5
7y = 14
The elimination of y from the first equation has a similar interpretation: the second equation in (3) is multiplied by 2 and added to the
first,
(
x + 2y = 5
y = 2 (2), add to first eq.
(
x+0y=522
y=2
The indicated operation is the main ingredient in the Gauss-Jordan
elimination method. We shall execute this operation repeatedly, in a
systematic way.
x + 4y 2z = 8
3x 2y 4z = 6
Solution. We first use the frist equation to eliminate x from the other
equations. From the second equation we subtract 2(first eq.) and
from the third we subtract 3(first eq.). The result is
x + 4y 2z = 8
y + 5z = 9
14y + 2z = 18
To simplify, we here divide the third equation by 2,
x + 4y 2z = 8
y + 5z = 9
7y + z = 9
x + 4y 2z = 8
y + 5z = 9
36z = 72
x + 4y 2z = 8
y + 5z = 9
z = 2
Using the third equation, we eliminate z from the first two ones:
x + 4y
= 4
y
= 1
z = 2
Finally, y is eliminated from the first equation using the second one:
x
= 0
y
= 1
z = 2
The system has the unique solution (x, y, z) = (0, 1, 2).
Here is the main principle. The first equation is used to eliminate the first unknown from the other equations. Then the (new)
second equation is used to eliminate the second unknown from the
subsequent equations, etc. The last equation will only contain the
last unknown, which has thus been calculated. After that we make a
backwards substitution to eliminate the variables "upwards, one at a
time, until the system is completely solved.
The simple rule above can not always be carried out. We illustrate
with a couple of simple examples.
Example 3. In the system
y 2z = 3
x + 2y z = 2
2x + 3y + z = 1
we cant as earlier use the first equation to eliminate x from the other
ones. On the other hand, the second equation can be used. To obtain
a system of the same form as earlier, we switch places of the first two
equations:
x + 2y z = 2
y 2z = 3
2x + 3y + z = 1
After this we can proceed as before. Do this! (The system has the
unique solution (x, y, z) = (2, 1, 2).)
(
3x + 2y = 4
.
Example 4. Solve the system
6x + 4y = 1
Solution. We eliminate x from the second equation by subtracting
2(first eq.) and get
(
3x + 2y = 4
0 = 7
The second equation is a false statement for all values of the variables
x and y. This means that the system has no solutions.
(
3x + 2y = 4
Example 5. Solve the system
.
6x + 4y = 8
Solution. Eliminating x from the second equation, we now get
(
3x + 2y = 4
0=0
The second equation is always satisfied, so the system is equivalent
to the single equation 3x + 2y = 4. This means that we can let y have
an arbitrary value, say y = t, and then x is uniquely determined by
y. Hence the system has infinitely many solutions
(
x = 34 32 t
,
t R.
y=t
Exercises.
1. Solve the following linear systems of equations:
(
(
(
x+y=2
2x 3y = 1
4x + 3y = 2
a)
b)
c)
x + y = 4
6x + y = 7
3x 5y = 6
2. Solve
the following systems:
x y+ z=4
2x y + 3z = 9
a)
b)
3x + 5y z = 0
3x + 2y 2z = 1
2x y 3z = 2
4x + 5y 4z = 2
x+y
=8
y + z = 2
d)
c)
x
+z=6 .
x 2y + z = 2
2x 5y + 3z = 2
y+z=4
3. Solve the following systems:
(
(
(
2x 31 y = 2
4x 3y = 4
6x + 9y = 5
a)
b)
c)
9
6x y = 4
3x + 4 y = 3
9x + 3y = 15
2
x + 2y + 3z = 10
1 2
3 10
()
is represented by 1 1 1 4 .
x+ y z= 4
2 1 1 5
2x y + z = 5
Example 6. We solve the system (*) in two ways: by writing down
complete equations as before, and, in parallel, by just manipulating
with the augmented matrix.
Eliminate x from eq.s 2 and 3:
x + 2y + 3z = 10
y 4z = 6
5y 5z = 15
3
10
1 2
0 1 4 6
0 5 5 15
1
1
Mult. eq. 2 by 1, eq. 3 by : | Mult. row 2 by 1, row 3 by :
5
5
x + 2y + 3z = 10
1 2 3 10
0 1 4 6
y + 4z = 6
0 1 1 3
y+ z= 3
Eliminate y from eq. 3 :
x + 2y + 3z = 10
y + 4z = 6
3z = 3
Row
1
0
3 row 2 :
2 3 10
1 4
6
0 3 3
Divide eq. 3 by 3 :
x + 2y + 3z = 10
y + 4z = 6
z= 1
Divide row 3 by 3 :
1 2 3 10
0 1 4 6
0 0 1 1
x + 2y
=7
y
=6
z=1
1 2 0 7
0 1 0 2
0 0 1 1
Eliminate y in eq. 1 :
x
=3
y
=6
z=1
Row 1 2 row 2 :
1 0 0 3
0 1 0 2
0 0 1 1
The system thus has the unique solution (x, y, z) = (3, 2, 1).
The admissible operations on the augmented matrix, i.e. those
giving rise to an equivalent augmented matrix, are the following:
(i)
(ii)
(iii)
(iv)
The operation (iv) is not really needed one can always circumvent it by other means. Here is an example of this.
Example 7. Solve the system
x + 2y 3z + w = 2
3x + 6y + 3z w = 2
2x + 4y + 3z + w = 9
2x + 4y 3z w = 13
Solution. The augmented matrix is
10
1
3
1
0
1
0
2 3 1 2
6 3 1 2
4 3
1
9
4 3 1 13
2 3 1 2
0 12 4 4
0 9 1 13
0 3 3 9
2 3 1 2
0 3 1 1
0 9 1 13
0 1 1 3
row 2 3row 1
row 3 2row 1
row 4 2row 1
divide row 2 by 4
divide row 4 by 3
Here we realize that we can not eliminate the variable in the second
column. We therefore skip that column and continue with the third
one. Interchange rows 2 and 4:
1 2 3 1 2
0 0 1 1 3
row 3 9row 2
0 0 9 1 13
0 0 3 1 1
row 4 3row 2
1 2 3 1 2
0 0 1 1 3
8 40
divide row 3 by 8
0 0 0
0 0 0
2 10
divide row 4 by 2
row 1- row 3
1 2 3 1 2
0 0 1 1 3
row 2 + row 3
1
5
0 0 0
0 0 0
1
5
row 4 is unnecessary - strike it!
row 1 + 3row 2
1 2 3 0 7
0 0 1 0 2
0 0 0 1 5
x + 2y = 1
1 2 0 0 1
0 0 1 0 2
.
z=2
i.e.
0 0 0 1 5
w=5
Here y can have an arbitrary value (say t), and then the solution is
fixed. The solutions are thus
(x, y, z, w) = (1 2t, t, 2, 5),
t R.
11
x 2y = 1
4x 7y = 5
Solution. The augmented matrix is
1 2 1
2 3 4
row 2 - 2row 1
row 3 - 4row 1
4 7 5
1 2 1
2
0 1
0 0 1
The last line represents the equation 0 = 1, which is not satisfied for
any values of x and y. The given system thus lacks solutions.
(
x 2y + z = 3
Example 9. Solve the system
.
2x + 4y 2z = 6
Solution.
!
1 2 1
3
2 4 2 6
!
1 2 1 3
i.e.
0 0 0 0
row 2 + 2row 1
(
x 2y + z = 3
0=0
The last equation is always satisfied - it does not mean any condition
on the unknowns x, y, and z. The given system of equations is
therefore equivalent to the single equation x 2y + z = 3. Here we
can prescribe values for y and z arbitrarily; the variable x is uniquely
determined by these values. The general solution can be written
(x, y, z) = (3 + 2s t, s, t),
s, t R.
x + y az = 3
x ay z = 2
x 3y z = 2 a
12
Solution.
(1)
(2)
(3)
(4)
1
1
1 a
3
a 1
2
3 1 2 a
row 2 - row 1
row 3 - row 1
1
a
Divide row 3 by -4,
3
then interchange
a 1 a 1
1
1
a
3
1
1
1
(1 a) 4 (1 + a)
4
row 3 + (a + 1)row 2
a 1 a 1
1
1
a
3
1
1
1
(1
a)
(1
+
a)
.
4
4
0 14 (a 1)(3 a) 14 (a 1)(a + 3)
row 1 - row 2
5
2
1
2
(
xz=
y=
1 0 1
0 1 0
i.e.
5
2
1
2
One can e.g. prescribe z arbitrarily and then get x, y, z from the value
of z,
5
1
(x, y, z) =
+ t, , t ,
t R.
2
2
Case 2: a = 3. In this case, (*) says that 0 z = 2 6 0 = 12. This
is false, so the system lacks solutions for a = 3.
13
1
0
a
1
(1
a)
4
1
2
1 0 9+a
3a
1 0 3a
3+a
0 1 3a
1
1
0
0 0
1 0
0 1
3
1
(1 + a)
4
3+a
row 1 + arow 3
row 2 + 14 (a 1)row 3
3a
row 1 - row 2
a2 a+9
3a
3a
3+a
3a
x 9y 3z = 4
x y+ z=0
a)
b)
3x 2y + z = 2
3x + 5y z = 0
2x + 7y + 4z = 2
6x + 2y + 2x = 5
2x + 3y = 2
x + 2y = 3
c)
d)
x
+
3y
z
=
5
2x 3y = 8
10x y = 2
3x + y + z = 3
x 2y = 6
2x + y 3z = 4
e)
f)
3x
+
y
=
2
4x + 3y z = 2
5x 3y = 14
2x 3y + 4z = 1
x + y 4z = 7
3x + y z = 7
g)
h)
x + y + 2z = 1
x y + 5z = 1
4x 6y + 8z = 3
14
5.
6.
7.
8.
9.
2x 3y + 4z = 1
2x y + z = 3
3x + y z = 7
j)
i)
4x 2y + 5z = 0
x y + 5z = 1
2x y 2z = 9
4x 6y + 8z = 2
x + 3y + z w = 2
3x + 5y z w = 2
.
k)
5x y 5z + 3w = 0
2x + 3y 3z 2w = 2
2x 3y + z = 1
3x + y 4z = c
a) a = 2, b = 3, c = 0;
b) a = b = c = 0.
Hint: Several linear systems with the same coefficient
matrices but different right hand sides can be solved simultaneously with the same augmented matrix: one just writes the
different right hand sides next to each other.
Determine a and b so that the lines y 3x = 2 and 2y + ax = b
a) intersect at a point, b) are parallel and different, c) coincide.
2x+13
a
b
Determine a and b so that (x1)(x+2)
= x1
+ x+2
.
(This is an example of a decomposition in partial fractions
of a rational function. Such decompositions are frequently
used in the calculation of integrals, for example.)
Determine for all values of a the number of solutions to the
system
(
(
(
x + 2y = 2
x + 2y = 1
x + 2y = 1
a)
b)
c)
.
2x + ay = a
2x + ay = a
2x + a2 y = a
Determine all solutions for the following systems for all values of(the constant a:
(
(
x + 3y = 4
x + 3y = 3
x 3ay = 2
a)
b)
c)
2x + ay = a
2x + ay = a
ax 12y = a + 2
(
(
x + y + 3z = 1
x+ y=3
x 2ay = 3
d)
e)
f)
2x + y + z = 2
2x ay = 2
ax + 3y = 3a 1
3x + 3y + az = 0
x+ y+
z=1
x + y + az = 1
g)
h)
2x + ay
z=1
x + ay + z = a .
ax + 2y + (a + 3)z = 2
ax + y + z = a2
4. ANSWERS TO EXERCISES
15
10. Examine, for different values of a and b, the number of solutions to the following linear systems of equations:
(
(
ax + by = 2
ax + 2y = b
.
b)
a)
x+ y =1
3x + 2y = 5
11. Determine a, b, and c so that
3x2 + 6x 16 a
b
c
= +
+
.
3
x 4x
x x+2 x2
12. Determine the constants a, b, c so that the function f (x) =
ax2 + bx + c satisfies f (1) = 3, f (2) = 1 and f (1) = 7.
13. Determine the equation of a third degree polynomial whose
graph passes through the points (0, 1), (1, 1), (2, 1), and (1, 7).
4. Answers to Exercises
1. a) x = 1, y = 3
b) x = 1.1, y = 0.4 c) x = 28
, y = 18
.
29
29
2. (x, y, z) =
a) (2, 1, 2)
b) (1, 2, 3)
c) (6, 2, 0)
d) (5, 3, 1).
3. a) Has no solution. b) x = 1 + 43 y, y R. c) (x, y) = ( 65 , 0).
4. a) (x, y, z) = (1 + 32 t, t, 1 52 t), t R.
b) No solution.
c) (x, y, z) = (4, 2, 7).
d) No solution exists.
, 167 ) f) (x, y, z) = (4t + 5, 5t 6, t), t R
e) (x, y) = ( 10
7
g) (x, y, z) = (3 t, t, 1), t R
h) Has no solution.
i) (x, y, z) = (2, 1, 0)
j) (x, y, z) = (t, 2t5, 2),
,t R
k) (x, y, z, w) = (2, 1, 1, 2).
5. a) There is no solution.
b) x = y = z = t, t R.
6. a) a , 6
b) a = 6, b , 4
c) a = 6, b = 4.
d) The system of equations has respectively: exactly one
solution, no solutions, infinitely many solutions.
7. a = 5, b = 3.
8. a) If a = 4 there are infinitely many, otherwise unique.
b) If a = 4 there is no solution, otherwise unique.
c) If a = 2 there are infinitely many, if a = 2 none, otherwise a unique solution.
a
9. a) For a = 6 no solution. For a , 6, (x, y) = a6
, a8
.
a6
b) For a = 6: (x, y) = (3 3t, t), t R. For a , 6: (x, y) =
(0, 1).
c) For a = 2, no solution. For a = 2: (x, y) = (2 + 6t, t),
t R.
1
.
For a , 2: (x, y) = a+4
,
a+2
3(a+2)
4
d) For a = 2: insoluble. For a , 2: (x, y) = 3a+2
,
.
a+2 a+2
16
e) (x, y) =
10.
11.
12.
13.
6a2 2a+9
, 2a21+3
2a2 +3
.
, 15 , 3 .
f) a = 9: insoluble. a , 9: (x, y, z) = a15
a9 a9 a9
g) a , 1: insoluble. a = 1: (x, y, z) = (2t, 1 3t, t), t R.
h) a = 1: (x, y, z) = (1tu,
2 t, u), t, u R. a = 2: insoluble.
1
, a+2
, a+1
a < {1, 2}: (x, y, z) = a +2a+1
.
a+2
a+2
a) If a , 3 unique solution; if a = 3 and b = 5 infinitely many
solutions; if a = 3 and b , 5 no solutions.
b) If a = b = 2 infinitely many solutions; if a = b , 2 none;
if a , b a unique solution.
a = 4, b = 2, c = 1.
a = 3, b = 5, c = 1.
The polynomial x3 + 3x2 2x + 1.
CHAPTER 2
Vectors
1. Basic Definitions
In this situation we say that the line segment PQ represents the vector
u.
The direction and the magnitude of the vector u is defined as the
u + v = PR.
(This rule is sometimes called the parallelogram law of addition.)
17
18
2. VECTORS
If u = PQ is a vector, we denote by ~
u the vector with the same
1
OM = (u + v) .
2
PQ = OQ OP = v u.
1. BASIC DEFINITIONS
19
u = OP ; v = OQ ; w = OR.
By the center of mass of the triangle PQR we mean the point N defined
by
1
ON = (u + v + w) .
3
Prove that the three medians of the triangle PQR intersect at the point
N. (A median of a triangle is a line-segment connecting a vertex to
the mid-point of the opposite side.)
(2)
Solution. Let M be the mid-point of the segment PQ. Then (by (1))
1
1
RM = OM OR = (u + v) w = (u + v 2w) .
2
2
Moreover,
1
1
RN = ON OR = (u + v + w) w = (u + v 2w) .
3
3
We have shown that
2RM = 3RN,
so N lies on the median RM and divides it according to the ratio 2 : 1.
By symmetry of the expression (2), the point N is also on the other
medians.
Exercises.
a) OB + BA b) AB c) OM d) AM.
2. Simplify the following sums:
a) OB + BD + DC b) AC + CO + OB
c) AB + OA + BD d) BD BA + DL.
20
2. VECTORS
u = xe.
The vector e is called a basis for ` and x is the coordinate for u with
respect to the basis e.
Now let be a plane and let e1 , e2 be two non-parallel vectors in
. We claim that each vector u in can be written
(2)
u = x1 e1 + x2 e2
where x1 and x2 are real numbers. To show this, we first choose two
lines `1 and `2 in such that e1 is in `1 and e2 is in `2 . Let O be
the point of intersection between `1 and `2 . To help our geometric
intuition, we will in the following fix O as our "origin, and place all
21
vectors in so that they emanate from the point O. (So e1 = OP1 for
and u2 = x2 e2
u = x1 e1 + x2 e2 + x3 e3
u = u0 + u00 ,
u0 , u00 `.
22
2. VECTORS
(5)
(6)
The rule (5) corresponds to the rule for addition of vectors, for if
u = x1 e1 + x2 e2 + x3 e3
and v = y1 e1 + y2 e2 + y3 e3 ,
then
u + v = (x1 e1 + x2 e2 + x3 e3 ) + (y1 e1 + y2 e2 + y3 e3 )
= (x1 + y1 )e1 + (x2 + y2 )e2 + (x3 + y3 )e3 .
The proof of (6) is similar.
Projections. In the decomposition (4) of a vector u, the vector u0
is called the projection of u parallel to the line ` on the plane , and u00 is
called the projection of u parallel to on `.
If the line ` is normal to the plane , then the projection u0 is called
the orthogonal (or "right angled) projection of u on .
If we decompose both u and v in this way, we find
u + v = (u0 + v0 ) + (u00 + v00 ),
which means that
(u + v)0 = u0 + v0
and
Exercises.
7. Let OPQR be a tetrahedron and introduce a basis for 3-space
3. LINEAR DEPENDENCE
23
3. Linear Dependence
Let u1 , u2 , . . . , uk be a collection of vectors. We introduce some
basic terminology:
A vector u of the form
u = 1 u1 + . . . + k uk ,
where 1 , . . . , k are real numbers, is called a linear combination of the
vectors u1 , . . . , uk .
The collection u1 , u2 , . . . , uk is called linearly dependent if there are
real numbers 1 , . . . , k , not all equal to zero, such that
1 u1 + . . . + k uk = 0.
Otherwise, i.e., if
1 u1 + . . . + k uk = 0
1 = 2 = . . . = k = 0,
24
2. VECTORS
u2 = (2, 3, 1)
u3 = (1, 3, 2)
1 22 3 = 0
22 + 32 + 33 = 0
21 + 2 + 23 = 0
Solving this with the elimination method gives the only solution
1 = 2 = 3 = 0. Hence u1 , u2 , u3 is linearly independent and (by
the result of Example 2) is a basis for 3-space.
A collection of four or more vectors in 3-space is always linearly
dependent. To show this, it is enough to prove that four vectors
are linearly dependent, for then every larger collection is linearly
25
can then be described by the position vector OP. For a given basis e1 ,
e2 , e3 there are then unique numbers x1 , x2 , x3 such that
OP = x1 e1 + x2 e2 + x3 e3 .
The numbers x1 , x2 , x3 are the coordinates of the point P with respect
to the coordinate system Oe1 e2 e3 . If the coordinate system is clear from
the context, we can simply write
P = (x1 , x2 , x3 ).
Remark 5. In a similar way we can describe points in a plane .
Let O be a point in and e1 , e2 is a basis for . Then for each point P
in we can write
OP = x1 e1 + x2 e2
for unique x1 , x2 R. We say that x1 , x2 are the coordinates of P in
the coordinate system Oe1 e2 for . In short: P = (x1 , x2 ).
26
2. VECTORS
(1)
QP = tu
for some t R. The vector u is called a direction vector of `.
Denote by (a, b, c) the coordinates of u with respect to the basis e1 ,
e2 , e3 . Thus
u = ae1 + be2 + ce3 .
If Q0 = (x0 , y0 , z0 ) and P = (x, y, z), then the vector QP = OP OQ has
coordinates (x x0 , y y0 , z z0 ). Hence (1) can be cast in the form
x x0 = at
t R.
(2)
y y0 = bt ,
z z0 = ct
The relation (2) is known as the parametric representation of the line `.
Example 1. Consider the line ` which passes through the points
Q = (3, 6, 5) and R = (4, 3, 3). A direction vector of ` is
QR = OR OQ = (4, 3, 3) (3, 6, 5) = (1, 3, 2).
Since ` passes through Q, we conclude that
x3=t
t R,
y + 6 = 3t ,
z + 5 = 2t
is a parametric representation for `.
In a similar way, one can represent a plane in space. Namely, let
Q be a point in and u, v a basis for (i.e. two non-parallel vectors
which are parallel to ). Then a point P in space belongs to if and
only if
(3)
QP = su + tv
for some s, t R. The numbers s, t are the coordinates for P in the
coordinate system Quv of . If we denote
u = ae1 + be2 + ce3
v = a0 e1 + b0 e2 + c0 e3
Q = (x0 , y0 , z0 ),
x x0 = as + a0 t
(4)
y y0 = bs + b0 t
z z0 = cs + c0 t
27
s, t R.
x 1 = s + t
y 2 = s 3t
z = s 3t
s, t R
is a parametric representation of .
Example 3. Let us determine the point of intersection between the
line ` of Example 1 and the plane of Example 2. To this end, we
separate between the t-parameters in the representations for ` and
and write
x 3 = t1
x 1 = s + t2
`:
,
:
y
+
6
=
3t
y 2 = s 3t2 .
z + 5 = 2t1
z = s 3t2
At the point of intersection, the values of x, y, and z must match, i.e.,
3 + t1 = 1 s + t2
s + t1 t2 = 2
s + 3t1 + 3t2 = 8 .
6 + 3t1 = 2 s 3t2
5 + 2t1 =
s 3t2
s + 2t1 + 3t2 = 5
After Gaussian elimination, this gives s = 2, t1 = 1, and t2 = 3.
Inserting t1 = 1 into the equation for ` gives (x, y, z) = (2, 9, 7).
The equation of a plane. We claim that planes in space correspond precisely to equations of the form
(5)
Ax + By + Cz = D
28
2. VECTORS
y
=
s
z
=
t
for some s, t R. Thus (5) describes the plane passing through
the point (D/A, 0, 0), which is parallel to the vectors (B/A, 1, 0),
(C/A, 0, 1).
Conversely, one can, by eliminating s and t in the parametric representation (4), prove that any plane can be described by an equation
of the form (5). We omit the details, since we shall anyway find
an easier method to prove this later on. Instead, we turn to some
examples.
Example 4. Consider again the plane
x 1 = s + t
:
y 2 = s 3t
z = s 3t
s, t R
1
=
s
+
t
x
1 = s + t
y 2 = s 3t
x + y 1 =
4t
x+ z1=
z = s 3t
2t
x
1 = s + t
x + y
1=
4t .
3x y + 2z 1 =
0
Since we can always choose s and t so that the first two equations are
satisfied, we see that a point (x, y, z) belongs to if and only if the
third equation is satisfied, i.e., iff
3x y + 2z 1 = 0.
Lines in space can be described as the intersection between two
non-parallel planes. This is illustrated by the following example.
Example 5. The points (x, y, z) which belong to both of the planes
2x + y 3z = 5
and x + 2y z = 4
3y + z = 3
x + 2y z = 4
29
x = 2 + 5t
y=1 t .
z =
3t
The intersection line thus passes through the point (2, 1, 0) and has
direction vector (5, 1, 3).
Remark 6. To describe a line in space we need two equations of the
form Ax + By + Cz = D. On the other hand, if we only consider points
in a plane , and if x, y denote coordinates with respect to some
coordinate system in , then one equation Ax + By = C is sufficient to
describe a line. If for example A , 0, then (x, y) satisfies the equation
if and only if
(
x = (C/A) (B/A)t
y=
t
for some t R. This is a parametric representation of a line in
passing through the point (C/A, 0), with direction vector (B/A, 1).
Exercises.
12. Give a parametric representation for the line passing through
the points (1, 1, 4) and (2, 3, 5).
13. Consider the lines
x=2+ t
x = 1 + 2t
`1 :
, `2 :
y=1 t
y=
t .
z =
2t
z = 1 + t
Do they intersect? Are they parallel?
14. Determine whether the following lines intersect:
x = 1 + 15t
x = 6 65t
y = 4 21t
y = 11 + 91t .
z = 5 + 33t
z = 16 143t
30
2. VECTORS
x = 10 + 5t
x
=
1
+
t
and
y= 5+ t .
y=
t
z = 2 + 2t
z = 5 + t
16. Find a parametric representation of the plane which passes
through the points (2, 3, 0), (1, 5, 2), and (1, 4, 3).
17. Are the four points (1, 1, 0), (0, 4, 1), (1, 0, 1), (1, 3, 2)
co-planar?
18. Find, in the form Ax + By + Cz = D, the equation for the
plane which passes through the point (1, 1, 2) and which
contains the line
x=3 t
y = 2 + 2t .
z = 1 3t
19. Prove that a line with direction-vector (a, b, c) is parallel to the
plane Ax + By + Cz = D if and only if
Aa + Bb + Cc = 0.
20. Find a parametric representation for the line which passes
through the point (1, 2, 4) and which is parallel to the planes
2x + y z = 3
and
3x 3y + z = 0.
21. Determine the equation for the plane which contains the line
(
3x + 4y + z = 5
,
x y
= 6
and which passes through the mid-point on the segment between the points (1, 1, 2) and (3, 1, 4).
5. Answers to Exercises
1. a) a b) b a c) 21 (a + b) d) 21 (b a).
2. a) OC
b) AB
c) OD
d) AL.
5. ANSWERS TO EXERCISES
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
x= 1+ t
y = 1 + 4t .
z = 4 + t
The lines do not intersect and they are not parallel.
The lines
coincide.
x = 5 + 2t
`:
y = 4 + 2t .
z =
t
x = 2 s 3t
:
y = 3 + 2s + t .
z = 2s + 3t
Yes.
: 5x 3y 14z = 60.
: x y z = 0.
x = 1 + 2t
y = 2 + 5t .
z = 4 + 9t
13x + 36y + 7z = 83.
31
CHAPTER 3
34
1
where e = kvk
v is a unit vector. Hence the absolute value |(u|v)|
equals the length of v times the length of the orthogonal projection
of u on a line with direction vector v. The sign of (u|v) is positive if
the angle between u and v is acute, and negative if it is obtuse.
The scalar product satisfies the following rules.
(I) (u|v) = (v|u)
(II) (u + v|w) = (u|w) + (v|w)
(III) (tu, v) = t (u, v)
(IV) (u|u) p0 with equality u = 0.
(V) kuk = (u|u)
The rule (I) is called symmetry of the scalar product; (II) and (III)
together are called linearity in the first argument; (IV) is called positive
definiteness.
The properties (I),(III), (IV) and (V) are immediate; (II) can be
realized in the following way: If w = e is a unit vector, then, by the
geometrical interpretation of the scalar product, we can see that (II)
means precisely that
2. ORTHONORMAL BASES
35
and v = y1 e1 + y2 e2
are two vectors in , then by the rules (I)(III) for the scalar product,
(u|v) = x1 y1 (e1 |e1 ) + (x1 y2 + x2 y1 )(e1 |e2 ) + x2 y2 (e2 |e2 ).
This expression becomes particularly simple if the basis vectors e1 , e2
are unit vectors making a right angle with each other. Namely, then
(e1 |e1 ) = (e2 |e2 ) = 1 and (e1 |e2 ) = 0, and thus
(3)
(u|v) = x1 y1 + x2 y2 .
when
j , k.
Three vectors e1 , e2 , e3 are said to form an orthonormal basis for threedimensional space if (i) they have unit length, and (ii) they are pairwise orthogonal. We can summarize the definition of orthonormal
basis by the equation
(
0 if j , k
(e j |ek ) =
.
1 if j = k
36
and v = y1 e1 + y2 e2 + y3 e3 ,
then
(4)
(u|v) = x1 y1 + x2 y2 + x3 y3 .
and v = (2, 2, 1)
kuk = 16 + 1 + 1 = 3 2
kvk = 4 + 4 + 1 = 3.
Since (u|v) = kukkvk cos , this gives
cos =
(u|v)
9
1
=
= .
kukvk 3 2 3
2
To this end, let e1 , e2 be an orthonormal basis for the plane and put
u = (cos )e1 + (sin )e2
Since u and v then have length 1, while the angle between them is
, the definition of scalar product shows that
(u|v) = cos( ).
(We have here used the fact that cos is even: cos( ) = cos( ).)
37
First note that ` has direction vector QS = (2, 2, 1), so it has the
parametric representation
x = 2 + 2t
`:
y = 1 2t .
z = 1 + t
We now seek the closest point R ` to P. To this end we put
38
This gives t = 4/9 and PR = 19 (19, 17, 4). The distance from P to `
is thus
1
74
PR
=
.
192 + 172 + 42 =
9
3
Distance from a point to a plane. Let be a plane in space, Q
a point in , and n a unit normal vector to . (This means that n has
unit length and points in a direction orthogonal to .)
The distance d between a point P and is again defined as the
smallest distance between P to some point in . We assert that
(7)
d = n|QP .
To prove this, it suffices to note that the orthogonal projection of the
d = Ax + By + Cz D .
Ax + By + Cz = D.
39
2
dividing with A + B2 + C2 . The distance formula then becomes
Ax + By + Cz D
(10)
d=
.
A2 + B2 + C2
Example 2. We shall determine an equation for the plane which
passes through the point (1, 1, 2) and has the normal
x = 1 + 3t
`:
y = 3 + 2t .
z = 2 t
The direction vector of `, i.e. the vector (3, 2, 1) must then be a normal vector of . Since the point (1, 1, 2) belongs to M, the equation
of the plane becomes
3(x 1) + 2(y (1)) + (1)(z 2) = 0,
i.e.,
3x + 2y z = 1.
The distance from the point P = (5, 6, 7) to is thus (see (10))
21
|3 5 + 2 6 1 7 + 1|
= .
14
32 + 22 + 12
Distance between two lines. Given two lines which do not intersect, we define their distance to be the smallest possible distance
between one point on the first line and one on the second one. The
following example comprises a method to compute this kind of distance.
Example 3. Consider the lines `1 and `2 having parametric representations
x
=
3
+
2t
x=3+ t
`1 :
; `2 :
y=
t
y = 4 + 3t .
z = 1 t
z = 2 + 2t
We shall compute the (shortest) distance between `1 and `2 . Direction
vectors for the two lines are (2, 1, 1) and (1, 3, 2). Let be the plane
parallel to these directions, which passes through the point (3, 0, 1)
on `1 . Then `1 and also, is parallel to `2 . Hence the distance d
40
x = 3 + 2t + s
:
y=
t + 3s .
z = 1 t + 2s
Elimination of s and t gives the equation
x y + z = 2
for the plane . The distance from the point (3, 4, 2) on `2 to is,
according to (10)
3
|3 4 + 2 + 2|
d=
= = 3.
3
1+1+1
2 : x + 4y + z = 5.
kn1 kkn2 k
2
12 + 22 + 22 12 + 42 + 12
This gives = 3/4. This is the obtuse angle between the planes.
There is also another possibility, namely if we substitute n1 for n1
above. This leads to the acute angle = /4.
cos =
Angle between a line and a plane. To determine the angle between a line ` and a plane , one first computes the acute angle
between ` and a normal vector to . The angle between ` and is
defined by + = /2.
Example 5. Suppose that
x=2+ t
`:
y=3+ t
z = 1 + 4t
: 4x 11y 5z = 2.
41
Let denote the angle between the direction vector (1, 1, 4) of ` and
the normal vector (4, 11, 5) of . Then
1 4 + 1 (11) + 4 (5)
1
= .
2
12 + 12 + 42 42 + 112 + 52
This gives = 2/3, which is obtuse. Hence the acute angle between
` and the normal to is = = /3. The angle between ` and
is thus = /2 = /6.
cos =
Exercises.
5. Compute the distance between the point (1, 2, 3) and the line
x= 1 t
y = 4 + 2t .
z = 3 t
6. The line ` is the intersection between the planes x+2y2z = 5
and 2x y + z = 0. Determine the point on ` which is closest
to the origin.
7. The line ` passes through the point (1, 2, 3) and is perpendicular to the plane 2x 3y + z = 3. Find the distance between
` and the point (4, 5, 6).
8. Determine, in the form Ax + By + Cz = D, the equation of the
plane which consists of all points which have equal distance
to the points (1, 2, 0) and (1, 0, 2).
9. Find the distance from the plane 3x 4y + 12z = 13 to the
points (0, 0, 0) and (2, 1, 3). Are these points on the same or
on opposite sides of the plane?
10. a) Determine, in the form Ax + By + Cz = D, an equation for
the plane M which passes through the points (2, 3, 0) and
(2, 2, 2), and is parallel to the line x = 2 t, y = 1 + t, z = 2 t.
b) Find the distance between the point (3, 1, 0) and M.
11. Find the point in the plane through the points (1, 3, 1), (1, 1, 0),
(1, 3, 2) which is closest to the point (2, 2, 1).
12. a) Prove that the lines
x
=
1
+
t
x=3+ t
`1 :
and
`2 :
y=2 t
y=2+ t
z = 3 + 2t
z = 2 3t
intersect at a point.
b) Find the distance between the point (3, 4, 5) and the
plane spanned by `1 and `2 .
42
13. A ray of light is emitted from the point (3, 2, 1) and reflected
off the plane x 2y 2z = 0. The reflected ray passes the point
(4, 1, 6). At which point does the ray hit the plane?
14. Determine the distance between the lines
a) (x, y, z) = t(3, 3, 1) and (x, y, z) = (1, 0, 0) + t(1, 1, 1).
b) (x, y, z) = (1, 2, 3) + t(0, 1, 1) and (x, y, z) = (1, 1, 1) +
t(2, 3, 1).
15. Consider the lines
x = 12 t
x
=
8
3t
.
and
`2 :
`1 :
y = 4 2t
y= 2 t
z = 1 + t
z = 3 + t
Determine, in the form Ax + By + Cz = D, an equation for the
plane which is parallel to `1 and `2 and has the same distance
to the two lines.
16. a) Determine, in the form Ax+By+Cz = D, an equation for the
plane M which passes through the points (2, 1, 3), (1, 2, 2),
and (1, 0, 2).
b) Determine the angle between M and the plane 2x + y
z = 1.
17. Determine the angle between the plane x + 2y z = 0 and the
line (x, y, z) = (3, 5, 1) + t(1, 1, 0).
18. A tetrahedron has corners A = (1, 2, 0), B = (1, 3, 1), C(1, 1, 0),
and D(1, 3, 2). Determine the angle between the plane containing the side BCD and the line containing the edge AB.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
4. Answers to Exercises
The side-lengths
are
3,
3,
and
2. The cosines of angles are
a) 3x 2y + z = 12.
b) 1/ 14.
(1, 1, 1).
a) 1/ 14.
b) 1/ 3.
4. ANSWERS TO EXERCISES
15.
16.
17.
18.
x + 2y + 5z = 1.
a) x + 2y + z = 3.
/3.
/6.
b) /3.
43
CHAPTER 4
46
x2 y2
= 1,
a2 b2
47
y2 = 4ax.
48
Now suppose that x and y are coordinates with respect to an ONsystem Oe1 e2 in the plane. We shall investigate the geometric meaning of the equation (1). In the preceding section, we saw that ellipses,
hyperbolas, and parabolas are all described by second-degree equations. We shall here show that, except for certain "pathological
cases, these three basic types of curves can be used to describe all
second-degree curves.
If A = B = C = 0, then (1) is a first-degree equation
Dx + Ey = F.
This is (unless D = E = 0) the equation for a line. In the sequel,
we can hence assume that at least one of the coefficients A, B, C are
non-zero.
The main idea for our solution of (1) involves changing coordinates to a new ON-system, where the equation has a simpler form.
We start by showing that, by a suitable rotation of the basis vectors,
we can get rid of the coefficient B for xy. Thus we introduce new
basis vectors e01 , e02 by
e01 = cos e1 + sin e2 ,
e02 = sin e1 + cos e2 .
It is easy to see that e01 , e02 is an orthonormal basis (since e1 , e2 is so).
Let (x0 , y0 ) be the coordinates for a point P relative to the system
Oe01 e02 . Then
49
where we have used the familiar "double angle formulas for cos and
sin.
If B = 0 no rotation is necessary, and we can take = 0. If B , 0
we choose such that
AC
.
cot 2 =
B
The relation (2) then shows that B0 = 0. In all cases, our new coordinate system turns our equation into the type
2
(3)
A0 (x0 )2 + C0 y0 + D0 x0 + E0 y0 = F,
for suitable constants A0 , C0 , D0 , E0 . Since we have assumed that at
least one of the numbers A, B, C are not zero, it is easy to see that at
least one of A0 , C0 are not zero.
Case 1. We first consider the case when both A0 and C0 are non-zero.
We can then complete squares in (3) to obtain
D0 2
E0 2 (D0 )2 (E0 )2
0
0
(4)
A0 x0 +
+
C
y
+
=
+
+ F.
2A0
2C0
4A0
4C0
We then make a new change of coordinates by
D0
E0
00
0
,
y
=
y
+
.
2A0
2C0
This means that the origin in the old system is moved to the point
which has x0 y0 -coordinates (D0 /2A0 , E0 /2C0 ). If
x00 = x0 +
F0 =
(D0 )2 (E0 )2
+
+ F,
4A0
4C0
A0 (x00 )2 + C0 (y00 )2 = F0 .
50
y00 = A0 x00 ,
which means a "degenerate hyperbola, or rather: two intersecting
straight lines.
Case 2. Now suppose that one of the numbers A0 , C0 in (3) are
zero. We can w.l.o.g. assume that A0 = 0 and C0 , 0. Then (3) says
that
C0 (y0 )2 + D0 x0 + E0 y0 = F.
(6)
,
y
=
y
+
4C0 D0 ,
D0 4C0 D0
2C0
we see that (6) transforms into the equation
x00 = x0
(7)
C0 (y00 )2 + D0 x00 = 0.
3. ANSWERS TO EXERCISES
3. Answers to exercises
1.
2.
3.
5.
6.
7.
8.
(3, 5).
a) (4, 0).
b) (12, 0).
2
2
x + 2y = 8.
a) (5, 0).
b) ( 40, 0).
5x2 4y2 = 20.
y2 = 12x2 ; focus (3, 0).
51
CHAPTER 5
where is the angle between u and v (in the interval [0, ]).
Definition 13. The cross-product (or vector-product) u v of u and
v is the unique vector w with the following properties:
(i) w is orthogonal to u and to v,
(ii) kwk = A(u, v),
(iii) If u and v are not parallel, then u, v, w is a positively oriented
triple.
53
54
Remark 14. Note that if the vectors u and v are parallel, then
A(u, v) = 0 and u v = 0.
Example 1. Let e1 , e2 , e3 be a positively oriented orthonormal basis.
Then
e1 e2 = e2 e1 = e3 ,
e3 e1 = e1 e3 = e2 ,
e2 e3 = e3 e2 = e1 .
These formulas are obvious geometrically.
The cross-product can (like the scalar product) be described using
an orthogonal projection. Suppose that v , 0 and that M is a plane
with normal vector v. Denote by u0 the orthogonal projection of u on
M. Then A(u, v) = A(u0 , v), because ku0 k is the height from v of the
parallelogram spanned by u and v. Let w = u v. So w is the vector
orthogonal to u and v, such that kwk = ku0 kkvk and that u, v, w is a
positively oriented triple. It is now seen that
(9)
u v = kvk T(u0 ),
(u1 + u2 ) v = u1 v + u2 v.
u v = v u.
u = x1 e1 + x2 e2 + x3 e3
v = y1 e1 + y2 e2 + y3 e3
2. CROSS-PRODUCT
55
and
(e1 e1 ) e2 = 0 e2 = 0.
56
Example 2. Consider three points P0 , P1 , P2 , which in a positively oriented orthonormal system have coordinates (2, 3, 2), (4, 1, 1), and
(2, 1, 1) respectively. We shall compute the area of the triangle
P0 P1 P2 . This area equals to half the area of the parallelogram spanned
v = y1 e1 + y2 e2
3. VOLUME-PRODUCT
57
Exercises.
1. Prove that if u + v + w = 0, then
2u v = v w + w u.
2. Find the area of the triangle which in a positively oriented
ON-system has its vertices at the points
a) (1, 2, 3), (3, 4, 1), (2, 0, 2).
b) (5, 1, 1), (2, 3, 2), (3, 2, 3).
c) (1, 0, 0), (0, 1, 0), (0, 0, 1).
3. In a positively oriented ON-system, the points (1, 1, 1) and
(0, 3, 3) are on the line `1 , and (2, 2, 4) and (4, 4, 4) are on `2 .
Find the distance between `1 and `2 .
4. Prove that
(u v) w = (u|w) v (v|w) u.
Hint: To simplify the computations, one can choose an
ON-basis e1 , e2 , e3 such that u = x1 e1 and v = y1 e1 + y2 e2 .
5. Prove that
(u v) w = u (v w)
if and only if u is parallel to w, or u and w are both orthogonal
to v.
Hint: Use the preceding exercise.
3. Volume-product
The cross-product and the scalar product can be combined to form
the volume-product V(u, v, w) of three vectors in space:
(15)
V(u, v, w) = (u v|w).
The motivation for the name is that the absolute value of V(u, v, w)
equals to the volume of the parallelepiped spanned by the vectors u, v, w,
if they are placed with their tails at one and the same point. To show this,
note that the vector
uv
e=
ku vk
is a unit normal to the plane spanned by u and v. Denote by P the
parallelepiped spanned by u, v, w. From elementary geometry we
know that the volume of P equals the area of the "base parallelogram
spanned by u, v, times the height h of P above the plane spanned by
u and v. Then h is the length of the orthogonal projection of w on the
58
normal of the plane, i.e., h = |(e|w)|. Since the base parallelogram has
area A(u, v) = ku vk, we infer that the volume of P equals
A(u, v)h = ku vk |(e|w)| = |(u v|w)| .
The asserted property of the volume product is proved.
When two vectors u and v are parallel we have u v = 0, so
V(u, v, w) = 0 in this case. More generally, the volume-product is
zero when the parallelepiped is degenerate, i.e., when the vectors u,
v, w are linearly dependent. On the other hand, if u, v, w are linearly
independent, the sign of (u v|w) depends on whether or not the
two vectors u v and w lie on the same side of the plane spanned by
u and v. The volume product V(u, v, w) is positive when the triple u,
v, w is positively oriented and negative otherwise. Since the volume
of the parallelepiped is the same regardless of how we choose to
permute the vectors u, v, w, only the sign can change under such a
permutation:
V(u, v, w) = V(w, u, v) = V(v, w, u) = V(u, w, v)
= V(v, u, w) = V(w, v, u).
Combining the computational rules for the cross- and scalar- products, we find that
(16)
for all real numbers s and t. Thus we have linearity in the first argument
for the volume product. We similarly have linearity in the second
and in the third argument. In short: the volume product is tri-linear,
i.e., linear in each of its three arguments.
Now let e1 , e2 , e3 be a basis for space, and take three vectors
u = (x1 , x2 , x3 ), v = (y1 , y2 , y3 ), w = (z1 , z2 , z3 ), where coordinates are
given according to the chosen basis. Then by linearity in the different
arguments,
V(u, v, w) = V(x1 e1 + x2 e2 + x3 e3 , v, w)
= x1 V(e1 , v, w) + x2 V(e2 , v, w) + x3 V(e3 , v, w)
X
= ... =
xi y j zk V(ei , e j , ek ).
Here the sum is over all possible choices of of i, j, k {1, 2, 3}, but
since V(ei , e j , ek ) = 0 if two of them coincide, only six terms can be
non-zero. Furthermore
V(e1 , e2 , e3 ) = V(e3 , e1 , e2 ) = V(e2 , e3 , e1 ) = V(e1 , e3 , e2 )
= V(e2 , e1 , e3 ) = V(e3 , e2 , e1 ).
3. VOLUME-PRODUCT
59
60
Exercises.
6. Motivate the identity
(u v|w) = (u|v w) .
7. The volume of a tetrahedron spanned by three vectors u, v,
w, rooted at the same point, equals to 1/6 of the volume of the
parallelepiped spanned by u, v, and w. Find the volume of
the tetrahedron with vertices at
a) (2, 1, 0), (3, 5, 2), (4, 1, 2), (6, 1, 5).
b) (2, 2, 3), (2, 1, 3), (1, 4, 2), (0, 5, 1).
8. A tetrahedron with volume 5 has three of its vertices at the
points (2, 1, 1), (3, 0, 1), (2, 1, 3). The fourth vertex is on the
positive y-axis. Determine its y-coordinate.
9. For which values of a and b are the three vectors
(a, b, b) ,
(b, a, b) ,
(b, b, a)
linearly dependent?
10. For which values of a are the four points
(0, 2, 1) ,
(a, 1, 0)
(3, 3, a) ,
(3, 3, 1 + a)
4. QUATERNIONS
61
ij = a + bi + cj.
x = x0 + x1 i + x2 j + x3 k.
i2 = j2 = k2 = 1
ij = ji = k , jk = k j = i ,
ki = ik = j,
then multiplication of 4-tuples will satisfy all other rules of calculation. Hamilton coined the term quaternions for the set of 4-tuples
with this multiplication. He also showed how to define division by
a non-zero quaternion.
In 1878, the German mathematician Frobenius proved that if we
want to define multiplication of n-tuples such that division by nonzero elements is always possible, and if we only are prepared to
abandon the commutative law for multiplication, then n must be
either 1, 2, or 4. Except for the real and complex fields, Hamiltons
quaternions are the only possibility.
Hamilton called the number x0 the scalar part of the quaternion (20)
and x1 i + x2 j + x3 k is the vector part. If one multiplies two quaternions
with scalar parts zero and uses the identities in (21), one finds after a
little calculation that
(x1 i + x2 j + x3 k)(y1 i + y2 j + y3 k) = (x1 y1 + x2 y2 + x3 y3 )+
+ (x2 y3 x3 y2 )i + (x3 y1 x1 y3 ) j + (x1 y2 x2 y1 )k.
62
The scalar part of the right hand sign equals the negative of the scalar
product of the vectors in the left hand side, and the vector part of
the right hand side equals the cross-product of the vectors in the left
hand side. The scalar product is older, even though the name was
invented by Hamilton, but the cross-product was discovered in this
way, as a by-product of multiplication of quaternions. Hamilton also
interpreted the quaternions geometrically and defined the scalar- and
cross-products in a basis-independent way, as we have done in this
chapter.
Exercises.
11. Prove that the formula (21) implies that
(x0 + x1 i + x2 j + x3 k)(x0 x1 i x2 j x3 k) = x20 + x21 + x22 + x23 ,
and that one therefore can define division by non-zero quaternions.
5. Answers to Exercises
a) 3 2. b) 26/2. c) 3/2.
3.
a) 4/3. b) 10.
y = 8.
a = b or a = 2b.
a = 1 and a = 9/4.
2.
3.
7.
8.
9.
10.
CHAPTER 6
Matrices
1. Basic properties
Definitions. By a p n-matrix we mean an array of numbers,
arranged in the form
21 a22 . . . a2n
A = ..
..
..
.
.
.
A = (a jk ) j,k=1 .
In the special case when p = n we say that A is a square matrix of order
n.
p,n
p,n
A + B = (a jk + b jk ) j,k=1 .
Example 1.
!
!
!
2 3 1
3 2 1
5 1 0
+
=
.
4 2 1
2 1 2
6 3 3
For a scalar t we define tA as the matrix with entries ta jk .
Example 2.
!
!
2 3 1
4 6 2
.
2
=
4 2 1
8 4 2
63
64
6. MATRICES
..
..
c x + c12 x2 + . . . + c1n xn = z1
11 1
..
c jk = b j1 a1k + b j2 a j2 + + b jp apk .
21 a22 . . . a2n
A = ..
..
..
.
.
.
21 b22 . . . b2n
and B = ..
..
.. .
.
.
.
1. BASIC PROPERTIES
65
3
1 2 3 2 = 10 .
1
Example 3.
3 6 9
3
2 1 2 3
= 2 4 6 .
1 2 3
1
Example 4.
!
!
!
4 2 1 2
8 16
=
,
2 1 2 4
4 8
!
!
!
1 2 4 2
0 0
=
.
2 4 2 1
0 0
The last example shows that the order between the factors is
essential for matrix multiplication. In other words, matrix multiplication is non-commutative: it is possible (and very common) to have
AB , BA.
In Example 1, B is a 2 2 matrix and A is a 2 3 matrix. The
product BA is therefore a 2 3 matrix. The product AB is not defined
in this case. In order that both AB and BA be defined, it is necessary
and sufficient that A be a n p and B a p n matrix (with the same n
and p). Then BA is an n n matrix and AB is p p. This is illustrated
by examples 2 and 3.
66
6. MATRICES
1 0 . . . 0
0 1 . . . 0
E = En = .. ..
..
. .
.
0 0 ... 1
is called the identity matrix of order n.
Notice that E is the neutral element for matrix multiplication, i.e.
we have
EA = AE = A
for all n n matrices A.
While matrix multiplication fails to be commutative, it obeys the
other rules of calculation.
Theorem 19. Matrix multiplication obeys the associative law
C(BA) = (CB)A
(5)
and the distributive laws
(6)
B(A + A0 ) = BA + BA0
(B + B0 )A = BA + BA0 .
(We here assume that the dimensions of the matrices are such that the sums
and products make sense.)
Proof. The formulas can easily be verified by direct evaluation.
Nonetheless, we shall give an alternative argument for the associative
law (5).
Consider the relation (1) as a function F : Rn Rp , which to
each n-tuple (x1 , . . . , xn ) Rn associates a p-tuple (y1 , . . . , yp ) Rp .
Likewise (2) can be regarded as a function G from Rp to Rq , which
to (y1 , . . . , yp ) associates (z1 , . . . , zq ). The matrix product BA will then
correspond to the composite function G F, and (5) follows from the
associate law from composition of functions:
H (G F) = (H G) F.
We shall in this course only be concerned with matrices whose
entries are real numbers. Nonetheless, we want to mention that
matrices with complex entries can be handled in the same way, as in
the following example.
1. BASIC PROPERTIES
67
!
0 i
K=
,
i 0
JK = KJ = I
KI = IK = J.
If this is compared with the formulas for multiplication of quaternions in the preceding chapter, one realizes that Hamiltons quaternions x0 + x1 i + x2 j + x3 k can be identified with the set of complex 2 2
matrices of the form
!
x0 + ix2 x2 + ix3
x0 E + x1 I + x2 J + x3 K =
.
x2 + ix3 x0 ix1
The computational rules for quaternions can then be seen as special
cases of the rules for matrix multiplication.
Before we close this section, we define a new matrix operation
called transposition. If A is a p n matrix, then the transpose of A At is
defined as the n p matrix whose rows are the columns of A:
21 a22 . . . a2n
12 a22 . . . ap2
t
If A = ..
..
.. .
..
.. then A = ..
.
.
.
.
.
.
(AB)t = Bt At .
Notice that the last rule says that transposition reverses the order of a
matrix product.
Exercises.
1. Let
1 0 2
A = 0 3 1
2 2 1
0 1 1
B = 2 2 0
1 2 3
2 1
C = 1 1 .
1 2
68
6. MATRICES
Compute
a) AB b) BA c) At Bt d) (A + 3B)C e) CCt
2. Let
!
!
1 1
1 2
A=
and
B=
.
1 1
3 4
f) Ct C.
0 5 3
A = 0 0 3 .
0 0 0
5. Show that, in order to verify all statements in Example 5, it
suffices to prove that
I2 = J2 = K2 = IJK = E.
2. Matrix inverse
Let
A = ..
..
..
.
.
.
x1
x2
x = ..
.
xn
y1
y2
y = .. .
.
yp
The linear equation system (1) can then be written in the matrix form:
(7)
Ax = y.
We shall here discuss (7) in the important case when n = p, i.e., when
A is a square matrix of order n.
2. MATRIX INVERSE
69
BA = E.
and
!
,
!
3 2
B=
,
2 1
70
6. MATRICES
AC = D
k = 1, . . . , n.
AB = E.
1 1 1
A = 1 2 3
1 3 2
2. MATRIX INVERSE
71
x1 + x2 + x3 = y1
x1 + 2x2 + 3x3 = y2
x1 + 3x2 + 2x3 = y3
x1 + x2 + x3 = y1
x2 + 2x3 = y1 + y2
x2 x3 =
y2 + y3
...
3x1
= 5y1 y2 y3
3x2
= y1 y2 + 2y3
3x3 = y1 + 2y2 y3
We see that the system has a unique solution x for each right hand
side y, so the matrix A is invertible. The last system also shows that
5 1 1
A1 = 1 1 2 .
3 1 2 1
Computational rules for the inverse. If both of the matrices A
and B are invertible, then the product AB is also invertible, and
(AB)1 = B1 A1 .
The order between factors is thus reversed after inversion. This is
realized by observing that if A and B are invertible then the matrix
D = B1 A1 satisfies
D(AB) = B1 (A1 A)B = B1 EB = B1 B = E,
and similarly
(AB)D = E.
Thus AB is invertible with inverse D.
Finally, we leave it to the reader to check that if A is invertible,
then At is invertible and
(At )1 = (A1 )t.
72
6. MATRICES
Exercises.
6. Determine which matrices are invertible. Also determine the
inverse
a) 0 1 2 b) 2 1 1 c) 0 1 1 .
1 1 0
1 1 4
0 0 1
7. Let
1 0 a
A = 0 1 1 .
1 1 0
Calculate A1 for those values of a for which A is invertible.
8. Find the inverse matrices of A and of A2 where
1 2 3
A = 2 3 1 .
1 1 1
9. Find a matrix X which solves the matrix equation AXB = C
where
!
!
1 2 3
1 2 1
1 1
0 1 2
A=
, B =
.
, C =
1 2
2 1 2
0 0 1
10. Let A and B be two nn-matrices such that EAB is invertible.
Prove that E BA is invertible and that
(E BA)1 = E + B(E AB)1 A.
3. Answers to Exercises
7
2 5 0
2
2 5
1. a) 7 4 3 b) 0 6 2 c) 5
0
3 4 1
7 12 1
4 14
5 1 4
6
d) 16 5 e) 1 2 1 f)
3
10 29!
4
1 !5
5
12
4
9
2. a)
b)
.
17 10
20 9
!
9t 3t
3.
, t R.
3t t
1 50 705
4. b) (E + A)10 = 0 1 30 .
0 0
1
2 7
6 12
2 1
!
3
.
6
3. ANSWERS TO EXERCISES
1 2 1
1 1 1
1 .
c) 12 1 1
6. a) 0 1 2 . b) Not invertible.
0 0
1
1
1 1
1 a a
1
1 a 1 for a , 1.
7. A1 = 1a
1 1 1
2 1 7
10 7 2
2 5 and (A2 )1 = 19 5 8 8 .
8. A1 = 13 1
1 1 1
2 4 13
!
0 3 6
9. X = A1 CB1 =
.
1 3 4
73