Linear Algebra Notes

David A. SANTOS [email protected]

January 2, 2010

REVISION

ii

Contents
Preface To the Student 1 Preliminaries 1.1 Sets and Notation . . . . . . . . . . 1.2 Partitions and Equivalence Relations 1.3 Binary Operations . . . . . . . . . . 1.4 Zn . . . . . . . . . . . . . . . . . . 1.5 Fields . . . . . . . . . . . . . . . . 1.6 Functions . . . . . . . . . . . . . . 2 Matrices and Matrix Operations 2.1 The Algebra of Matrices . . . 2.2 Matrix Multiplication . . . . 2.3 Trace and Transpose . . . . 2.4 Special Matrices . . . . . . . 2.5 Matrix Inversion . . . . . . . 2.6 Block Matrices . . . . . . . 2.7 Rank of a Matrix . . . . . . 2.8 Rank and Invertibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv v 1 1 4 6 9 13 15 18 18 22 28 31 36 44 45 55 65 65 70 71 76 76 79 81 84 87 91 6 Determinants 6.1 Permutations . . . . . . 6.2 Cycle Notation . . . . . . 6.3 Determinants . . . . . . 6.4 Laplace Expansion . . . 6.5 Determinants and Linear . . . . . . . . . . . . . . . . . . . . Systems . . . . . . . . . . . . . . . . . . . . . . . . . 111 111 114 119 129 136 138 138 139 143 147 149 149 152 158 164 169 174 178

7 Eigenvalues and Eigenvectors 7.1 Similar Matrices . . . . . . . . . . . . . 7.2 Eigenvalues and Eigenvectors . . . . . 7.3 Diagonalisability . . . . . . . . . . . . 7.4 Theorem of Cayley and Hamilton . . . 8 Linear Algebra and Geometry 8.1 Points and Bi-points in R2 8.2 Vectors in R2 . . . . . . . 8.3 Dot Product in R2 . . . . . 8.4 Lines on the Plane . . . . 8.5 Vectors in R3 . . . . . . . 8.6 Planes and Lines in R3 . . 8.7 Rn . . . . . . . . . . . . .

. . . .

. . . .

3 Linear Equations 3.1 Denitions . . . . . . . . . . . . . . . . . 3.2 Existence of Solutions . . . . . . . . . . . 3.3 Examples of Linear Systems . . . . . . . . 4 Vector Spaces 4.1 Vector Spaces . . . . 4.2 Vector Subspaces . . 4.3 Linear Independence 4.4 Spanning Sets . . . . 4.5 Bases . . . . . . . . 4.6 Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

A Answers and Hints 182 Answers and Hints . . . . . . . . . . . . . . . . 182

GNU Free Documentation License 262 1. APPLICABILITY AND DEFINITIONS . . . . . . 262 2. VERBATIM COPYING . . . . . . . . . . . . . 262 3. COPYING IN QUANTITY . . . . . . . . . . . . 262 4. MODIFICATIONS . . . . . . . . . . . . . . . 262 5. COMBINING DOCUMENTS . . . . . . . . . . 263 6. COLLECTIONS OF DOCUMENTS . . . . . . . 263 7. AGGREGATION WITH INDEPENDENT WORKS 263 5 Linear Transformations 97 8. TRANSLATION . . . . . . . . . . . . . . . . 263 5.1 Linear Transformations . . . . . . . . . . . 97 9. TERMINATION . . . . . . . . . . . . . . . . 263 5.2 Kernel and Image of a Linear Transformation 99 5.3 Matrix Representation . . . . . . . . . . . 104 10. FUTURE REVISIONS OF THIS LICENSE . . 263

iii Copyright c 2007 David Anthony SANTOS. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation; with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license is included in the section entitled GNU Free Documentation License.

Preface

These notes started during the Spring of 2002, when John MAJEWICZ and I each taught a section of Linear Algebra. I would like to thank him for numerous suggestions on the written notes. The students of my class were: Craig BARIBAULT, Chun CAO, Jacky CHAN, Pho DO, Keith HARMON, Nicholas SELVAGGI, Sanda SHWE, and Huong VU. I must also thank my former student William CARROLL for some comments and for supplying the proofs of a few results. Johns students were David HERNNDEZ, Adel JAILILI, Andrew KIM, Jong KIM, Abdelmounaim LAAYOUNI, Aju MATHEW, Nikita MORIN, Thomas NEGRN, Latoya ROBINSON, and Saem SOEURN. Linear Algebra is often a students rst introduction to abstract mathematics. Linear Algebra is well suited for this, as it has a number of beautiful but elementary and easy to prove theorems. My purpose with these notes is to introduce students to the concept of proof in a gentle manner. David A. Santos [email protected]

iv

To the Student
These notes are provided for your benet as an attempt to organise the salient points of the course. They are a very terse account of the main ideas of the course, and are to be used mostly to refer to central denitions and theorems. The number of examples is minimal, and here you will nd few exercises. The motivation or informal ideas of looking at a certain topic, the ideas linking a topic with another, the worked-out examples, etc., are given in class. Hence these notes are not a substitute to lectures: you must always attend to lectures. The order of the notes may not necessarily be the order followed in the class. There is a certain algebraic uency that is necessary for a course at this level. These algebraic prerequisites would be difcult to codify here, as they vary depending on class response and the topic lectured. If at any stage you stumble in Algebra, seek help! I am here to help you! Tutoring can sometimes help, but bear in mind that whoever tutors you may not be familiar with my conventions. Again, I am here to help! On the same vein, other books may help, but the approach presented here is at times unorthodox and nding alternative sources might be difcult. Here are more recommendations: Read a section before class discussion, in particular, read the denitions. Class provides the informal discussion, and you will prot from the comments of your classmates, as well as gain condence by providing your insights and interpretations of a topic. Dont be absent! Once the lecture of a particular topic has been given, take a fresh look at the notes of the lecture topic. Try to understand a single example well, rather than ill-digest multiple examples. Start working on the distributed homework ahead of time. Ask questions during the lecture. There are two main types of questions that you are likely to ask. 1. Questions of Correction: Is that a minus sign there? If you think that, for example, I have missed out a minus sign or wrote P where it should have been Q,1 then by all means, ask. No one likes to carry an error till line XLV because the audience failed to point out an error on line I. Dont wait till the end of the class to point out an error. Do it when there is still time to correct it! 2. Questions of Understanding: I dont get it! Admitting that you do not understand something is an act requiring utmost courage. But if you dont, it is likely that many others in the audience also dont. On the same vein, if you feel you can explain a point to an inquiring classmate, I will allow you time in the lecture to do so. The best way to ask a question is something like: How did you get from the second step to the third step? or What does it mean to complete the square? Asseverations like I dont understand do not help me answer your queries. If I consider that you are asking the same questions too many times, it may be that you need extra help, in which case we will settle what to do outside the lecture. Dont fall behind! The sequence of topics is closely interrelated, with one topic leading to another. The use of calculators is allowed, especially in the occasional lengthy calculations. However, when graphing, you will need to provide algebraic/analytic/geometric support of your arguments. The questions on assignments and exams will be posed in such a way that it will be of no advantage to have a graphing calculator. Presentation is critical. Clearly outline your ideas. When writing solutions, outline major steps and write in complete sentences. As a guide, you may try to emulate the style presented in the scant examples furnished in these notes.
1 My

doctoral adviser used to say I said A, I wrote B, I meant C and it should have been D!

Chapter

Preliminaries
1.1 Sets and Notation
1 Denition We will mean by a set a collection of well dened members or elements. 2 Denition The following sets have special symbols. N = {0, 1, 2, 3, . . .} Z = {. . . , 3, 2, 1, 0, 1, 2, 3, . . .} Q R C denotes the set of natural numbers. denotes the set of integers. denotes the set of rational numbers. denotes the set of real numbers. denotes the set of complex numbers. denotes the empty set.

4 Example Prove that between any two rational numbers there is always a rational number. Solution: Let (a, c) Z2 , (b, d) (N \ {0})2 ,
a b

3 Denition (Implications) The symbol = is read implies, and the symbol is read if and only if. <
c . d

Then da < bc. Now a b < a+c b+d ,

ab + ad < ab + bc = a(b + d) < b(a + c) = da + dc < cb + cd = d(a + c) < c(b + d) = whence the rational number a+c b+d lies between
a b

a+c c < , b+d d

and

c . d

so if r1 < r2 are rational numbers, then

 We can also argue that the averager of+two distinct numbers lies between the numbers and r
1 2

lies between them.

5 Denition Let A be a set. If a belongs to the set A, then we write a A, read a is an element of A. If a does not belong to the set A, we write a A, read a is not an element of A. 1

Chapter 1

6 Denition (Conjunction, Disjunction, and Negation) The symbol is read or (disjunction), the symbol is read and (conjunction), and the symbol is read not. 7 Denition (Quantiers) The symbol is read for all (the universal quantier), and the symbol is read there exists (the existential quantier). We have (x A, P(x)) ( A, P(x)) (1.1) (1.2) ( A, P(x)) (x A, P(x))

8 Denition (Subset) If a A we have a B, then we write A B, which we read A is a subset of B. In particular, notice that for any set A, A and A A. Also

A=B

N Z Q R C. (A B) (B A).

9 Denition The union of two sets A and B, is the set A B = {x : (x A) (x B)}. This is read A union B. See gure 1.1. 10 Denition The intersection of two sets A and B, is A B = {x : (x A) (x B)}. This is read A intersection B. See gure 1.2. 11 Denition The difference of two sets A and B, is A \ B = {x : (x A) (x B)}. This is read A set minus B. See gure 1.3.

A B

A B

A B

Figure 1.1: A B

Figure 1.2: A B

Figure 1.3: A \ B

12 Example Prove by means of set inclusion that (A B) C = (A C) (B C).

Sets and Notation Solution: We have, x (A B) C x (A B) x C (x A x B) x C (x A x C) (x B x C) (x A C) (x B C) x (A C) (B C),

which establishes the equality.

that is, the set of all ordered n-tuples whose elements belong to the given sets.

13 Denition Let A1 , A2 , . . . , An , be sets. The Cartesian Product of these n sets is dened and denoted by A1 A2 An = {(a1 , a2 , . . . , an ) : ak Ak },

 In the particular case when all the A

are equal to a set A, we write

A1 A2 An = An . If a A and b A we write (a, b) A2 . 14 Denition Let x R. The absolute value of xdenoted by |x|is dened by x x if x < 0, if x 0.

|x| =

It follows from the denition that for x R, |x| x |x|. t 0 = |x| t t x t. a R = a2 = |a|. 15 Theorem (Triangle Inequality) Let (a, b) R2 . Then |a + b| |a| + |b|. Proof: to |b| b |b| we obtain whence the theorem follows by 1.4. u (|a| + |b|) a + b (|a| + |b|), From 1.3, by addition, |a| a |a| (1.6) (1.3) (1.4) (1.5)

Chapter 1
Problem 1.1.5 Prove that (A B) \ (A B) = (A \ B) (B \ A). Problem 1.1.6 Write the union A B C as a disjoint union of sets. Problem 1.1.7 Prove that a set with n 0 elements has 2n subsets. Problem 1.1.8 Let (a, b) R2 . Prove that ||a| |b|| |a b|.

Homework
Problem 1.1.1 Prove that between any two rational numbers there is an irrational number. Problem 1.1.2 Prove that X \ (X \ A) = X A. Problem 1.1.3 Prove that X \ (A B) = (X \ A) (X \ B). Problem 1.1.4 Prove that X \ (A B) = (X \ A) (X \ B).

1.2 Partitions and Equivalence Relations

16 Denition Let S = be a set. A partition of S is a collection of non-empty, pairwise disjoint subsets of S whose union is S. 17 Example Let 2Z = {. . . , 6, 4, 2, 0, 2, 4, 6, . . .} = 0 be the set of even integers and let 2Z + 1 = {. . . , 5, 3, 1, 1, 3, 5, . . .} = 1 be the set of odd integers. Then (2Z) (2Z + 1) = Z, (2Z) (2Z + 1) = , and so {2Z, 2Z + 1} is a partition of Z. 18 Example Let 3Z = {. . . 9, , 6, 3, 0, 3, 6, 9, . . .} = 0 be the integral multiples of 3, let 3Z + 1 = {. . . , 8, 5, 2, 1, 4, 7, . . .} = 1 be the integers leaving remainder 1 upon division by 3, and let 3Z + 2 = {. . . , 7, 4, 1, 2, 5, 8, . . .} = 2 be integers leaving remainder 2 upon division by 3. Then (3Z) (3Z + 1) (3Z + 2) = Z, and so {3Z, 3Z + 1, 3Z + 2} is a partition of Z. (3Z) (3Z + 1) = , (3Z) (3Z + 2) = , (3Z + 1) (3Z + 2) = ,

Notice that 0 and 1 do not mean the same in examples 17 and 18. Whenever we make use of this notation, the integral divisor must be made explicit. 19 Example Observe which means that the real numbers can be partitioned into the rational and irrational numbers. R = (Q) (R \ Q), = (Q) (R \ Q),

Partitions and Equivalence Relations

20 Denition Let A, B be sets. A relation R is a subset of the Cartesian product A B. We write the fact that (x, y) R as x y. 21 Denition Let A be a set and R be a relation on A A. Then R is said to be reexive if (x A), x x, symmetric if ((x, y) A2 ), x y = y x, anti-symmetric if ((x, y) A2 ), (x y) (y x) = x = y, transitive if ((x, y, z) A3 ), (x y) (y z) = (x z). A relation R which is reexive, symmetric and transitive is called an equivalence relation on A. A relation R which is reexive, anti-symmetric and transitive is called a partial order on A. 22 Example Let S ={All Human Beings}, and dene on S as a b if and only if a and b have the same mother. Then a a since any human a has the same mother as himself. Similarly, a b = b a and (a b) (b c) = (a c). Therefore is an equivalence relation. 23 Example Let L be the set of all lines on the plane and write l1 l2 if l1 ||l2 (the line l1 is parallel to the line l2 ). Then is an equivalence relation on L.
a x 24 Example In Q dene the relation b y ay = bx, where we will always assume that the denominators are non-zero. Then is an equivalence relation. For a a since ab = ab. Clearly b b

a
s t

b
x x Finally, if a y and y b ayxt = bxsy. This gives

= ay = bx = xb = ya =

then we have ay = bx and xt = sy. Multiplying these two equalities ayxt bxsy = 0 = xy(at bs) = 0.

Now if x = 0, we will have a = s = 0, in which case trivially at = bs. Otherwise we must have at bs = 0 and so a s . b t 25 Example Let X be a collection of sets. Write A B if A B. Then is a partial order on X. 26 Example For (a, b) R2 dene a b a2 + b2 > 2.

Determine, with proof, whether is reexive, symmetric, and/or transitive. Is an equivalence relation? Solution: Since 02 + 02 2, we have 0 0 and so is not reexive. Now, ab a2 + b2 b a,

b2 + a2 so is symmetric. Also 0 3 since 02 + 32 > 2 and 3 1 since 32 + 12 > 2. But 0 1 since 02 + 12 2. Thus the relation is not transitive. The relation, therefore, is not an equivalence relation.

Chapter 1

27 Denition Let be an equivalence relation on a set S. Then the equivalence class of a is dened and denoted by [a] = {x S : x a}. 28 Lemma Let be an equivalence relation on a set S. Then two equivalence classes are either identical or disjoint. Proof: We prove that if (a, b) S2 , and [a] [b] = then [a] = [b]. Suppose that x [a] [b]. Now x [a] = x a = a x, by symmetry. Similarly, x [b] = x b. By transitivity (a x) (x b) = a b. Now, if y [b] then b y. Again by transitivity, a y. This means that y [a]. We have shewn that y [b] = y [a] and so [b] [a]. In a similar fashion, we may prove that [a] [b]. This establishes the result. u 29 Theorem Let S = be a set. Any equivalence relation on S induces a partition of S. Conversely, given a partition of S into disjoint, non-empty subsets, we can dene an equivalence relation on S whose equivalence classes are precisely these subsets. Proof: By Lemma 28, if is an equivalence relation on S then S=
aS

[a],

and [a] [b] = if a b. This proves the rst half of the theorem. Conversely, let S=

S , S S = if = ,

be a partition of S. We dene the relation on S by letting a b if and only if they belong to the same S . Since the S are mutually disjoint, it is clear that is an equivalence relation on S and that for a S , we have [a] = S . u

Homework
Problem 1.2.1 For (a, b) (Q \ {0})2 dene the relation as follows: a b a Z. Determine whether this b relation is reexive, symmetric, and/or transitive. Problem 1.2.2 Give an example of a relation on Z \ {0} which is reexive, but is neither symmetric nor transitive. Problem 1.2.3 Dene the relation in R by x y xey = yex . Prove that is an equivalence relation. Problem 1.2.4 Dene the relation in Q by x y 3y + h h Z such that x = . [A] Prove that is an 3 equivalence relation. [B] Determine [x], the equivalence 4 2 of x Q. [C] Is 3 5 ?

1.3 Binary Operations

30 Denition Let S, T be sets. A binary operation is a function SS T .

(a, b) (a, b)

Binary Operations

We usually use the inx notation a b rather than the prex notation (a, b). If S = T then we say that the binary operation is internal or closed and if S = T then we say that it is external. If ab =ba then we say that the operation is commutative and if a (b c) = (a b) c, we say that it is associative. If is associative, then we can write a (b c) = (a b) c = a b c, without ambiguity.

We usually omit the sign and use juxtaposition to indicate the operation . Thus we write
ab instead of a b. 31 Example The operation + (ordinary addition) on the set ZZ is a commutative and associative closed binary operation. 32 Example The operation (ordinary subtraction) on the set N N is a non-commutative, nonassociative non-closed binary operation. 33 Example The operation dened by a b = 1 + ab on the set Z Z is a commutative but nonassociative internal binary operation. For a b = 1 + ab = 1 + ba = ba, proving commutativity. Also, 1 (2 3) = 1 (7) = 8 and (1 2) 3 = (3) 3 = 10, evincing non-associativity. 34 Denition Let S be a set and : S S S be a closed binary operation. The couple S, is called an algebra. When we desire to drop the sign and indicate the binary operation by juxtaposition, we simply speak of the algebra S. 35 Example Both Z, + and Q, are algebras. Here + is the standard addition of real numbers and is the standard multiplication. 36 Example Z, is a non-commutative, non-associative algebra. Here is the standard subtraction operation on the real numbers 37 Example (Putnam Exam, 1972) Let S be a set and let be a binary operation of S satisfying the laws (x, y) S2 x (x y) = y, (1.7) Shew that is commutative, but not necessarily associative. Solution: By (1.8) x y = ((x y) x) x. By (1.8) again ((x y) x) x = ((x y) ((x y) y)) x. (y x) x = y. (1.8)

8 By (1.7) ((x y) ((x y) y)) x = (y) x = y x, which is what we wanted to prove.

Chapter 1

To shew that the operation is not necessarily associative, specialise S = Z and x y = x y (the opposite of x minus y). Then clearly in this case is commutative, and satises (1.7) and (1.8) but 0 (0 1) = 0 (0 1) = 0 (1) = 0 (1) = 1, and (0 0) 1 = (0 0) 1 = (0) 1 = 0 1 = 1, evincing that the operation is not associative. 38 Denition Let S be an algebra. Then l S is called a left identity if s S we have ls = s. Similarly r S is called a right identity if s S we have sr = s. 39 Theorem If an algebra S possesses a left identity l and a right identity r then l = r. Proof: Since l is a left identity r = lr. Since r is a right identity l = lr. Combining these two, we gather r = lr = l, whence the theorem follows. u 40 Example In Z, + the element 0 Z acts as an identity, and in Q, the element 1 Q acts as an identity. 41 Denition Let S be an algebra. An element a S is said to be left-cancellable or left-regular if (x, y) S2 ax = ay = x = y. Similarly, element b S is said to be right-cancellable or right-regular if (x, y) S2 xb = yb = x = y. Finally, we say an element c S is cancellable or regular if it is both left and right cancellable. 42 Denition Let S, and S, be algebras. We say that is left-distributive with respect to if (x, y, z) S3 , x(y z) = (xy) (xz). Similarly, we say that is right-distributive with respect to if (x, y, z) S3 , (y z)x = (yx) (zx). We say that is distributive with respect to if it is both left and right distributive with respect to .

Homework

Zn
Problem 1.3.1 Let S = {x Z : (a, b) Z , x = a + b + c 3abc}. Prove that S is closed under multiplication, that is, if x S and y S then xy S. Problem 1.3.2 Let S, be an associative algebra, let a S be a xed element and dene the closed binary operation by xy = x a y.
2 3 3 3

9
Problem 1.3.4 On R \ {1} dene the a binary operation a b = a + b ab,

where juxtaposition means ordinary multiplication and + is the ordinary addition of real numbers. Clearly is a closed binary operation. Prove that Prove that is both commutative and associative. Find an element e R \ {1} such that (a R \ {1}) (e a = a). Given e as above and an arbitrary element a R \ {1}, solve the equation a b = e for b.

Prove that is also associative over S S.

Problem 1.3.3 On Q] 1; 1[ dene the a binary operation a+b ab= , 1 + ab where juxtaposition means ordinary multiplication and + is the ordinary addition of real numbers. Prove that Prove that is a closed binary operation on Q] 1; 1[. Prove that is both commutative and associative. Find an element e R such that (a Q] 1; 1[) (e a = a). Given e as above and an arbitrary element a Q] 1; 1[, solve the equation a b = e for b.

Problem 1.3.5 (Putnam Exam, 1971) Let S be a set and let be a binary operation on S satisfying the two laws (x S)(x x = x), and ((x, y, z) S3 )((x y) z = (y z) x). Shew that is commutative. Problem 1.3.6 Dene the symmetric difference of the sets A, B as AB = (A \ B) (B \ A). Prove that is commutative and associative.

1.4 Zn
43 Theorem (Division Algorithm) Let n > 0 be an integer. Then for any integer a there exist unique integers q (called the quotient) and r (called the remainder) such that a = qn + r and 0 r < q. Proof: In the proof of this theorem, we use the following property of the integers, called the well-ordering principle: any non-empty set of non-negative integers has a smallest element.

Consider the set S = {a bn : b Z a bn}. Then S is a collection of nonnegative integers and S = as a 0 n S and this is nonnegative for one choice of sign. By the Well-Ordering Principle, S has a least element, say r. Now, there must be some q Z such that r = a qn since r S. By construction, r 0. Let us prove that r < n. For assume that r n. Then r > r n = a qn n = a (q + 1)n 0, since r n 0. But then a (q + 1)n S and a (q + 1)n < r which contradicts the fact that r is the smallest member of S. Thus we must have 0 r < n. To prove that r and q are unique, assume that q1 n + r1 = a = q2 n + r2 , 0 r1 < n, 0 r2 < n. Then r2 r1 = n(q1 q2 ), that is, n divides (r2 r1 ). But |r2 r1 | < n, whence r2 = r1 . From this it also follows that q1 = q2 . This completes the proof. u 44 Example If n = 5 the Division Algorithm says that we can arrange all the integers in ve columns as

10 follows:

Chapter 1

. . .

. . .

. . . 8 3 2 7 . . .

. . . 7 2 3 8 . . .

. . . 6 1 4 9 . . .

10 9 5 0 5 . . . 4 1 6 . . .

The arrangement above shews that any integer comes in one of 5 avours: those leaving remainder 0 upon division by 5, those leaving remainder 1 upon division by 5, etc. We let 5Z = {. . . , 15, 10, 5, 0, 5, 10, 15, . . .} = 0, 5Z + 1 = {. . . , 14, 9, 4, 1, 6, 11, 16, . . .} = 1, 5Z + 2 = {. . . , 13, 8, 3, 2, 7, 12, 17, . . .} = 2, 5Z + 3 = {. . . , 12, 7, 2, 3, 8, 13, 18, . . .} = 3, 5Z + 4 = {. . . , 11, 6, 1, 4, 9, 14, 19, . . .} = 4, and Z5 = {0, 1, 2, 3, 4}. Let n be a xed positive integer. Dene the relation by x y if and only if they leave the same remainder upon division by n. Then clearly is an equivalence relation. As such it partitions the set of integers Z into disjoint equivalence classes by Theorem 29. This motivates the following denition. 45 Denition Let n be a positive integer. The n residue classes upon division by n are 0 = nZ, 1 = nZ + 1, 2 = nZ + 2, . . . , n 1 = nZ + n 1. The set of residue classes modulo n is Zn = {0, 1, . . . , n 1}.

Our interest is now to dene some sort of addition and some sort of multiplication in Zn . 46 Theorem (Addition and Multiplication Modulo n) Let n be a positive integer. For (a, b) (Zn )2 dene a + b = r, where r is the remainder of a + b upon division by n. and a b = t, where t is the remainder of ab upon division by n. Then these operations are well dened. Proof: We need to prove that given arbitrary representatives of the residue classes, we always obtain the same result from our operations. That is, if a = a and b = b then we have a + b = a + b and a b = a b . Now a = a = (q, q ) Z2 , r N, a = qn + r, a = q n + r, 0 r < n, b=b
= (q1 , q1 ) Z2 , r1 N, b = q1 n + r1 , b = q1 n + r1 , 0 r1 < n.

Zn Hence a + b = (q + q1 )n + r + r1 ,
a + b = (q + q1 )n + r + r1 ,

11

meaning that both a + b and a + b leave the same remainder upon division by n, and therefore a + b = a + b = a + b = a + b . Similarly ab = (qq1 n + qr1 + rq1 )n + rr1 ,
a b = (q q1 n + q r1 + rq1 )n + rr1 ,

and so both ab and a b leave the same remainder upon division by n, and therefore a b = ab = a b = a b . This proves the theorem. u 47 Example Let Z6 = {0, 1, 2, 3, 4, 5} be the residue classes modulo 6. Construct the natural addition + table for Z6 . Also, construct the natural multiplication table for Z6 . Solution: The required tables are given in tables 1.1 and 1.2.

+ 0 1 2 3 4 5

0 0 1 2 3 4 5

1 2 1 2 2 3 3 4 4 5 5 0 0 1

3 3 4 5 0 1 2

4 5 4 5 5 0 0 1 1 2 2 3 3 4

0 1 2 3 4 5

0 0 0 0 0 0 0

1 0 1 2 3 4 5

2 0 2 4 0 2 4

3 4 0 0 3 4 0 2 3 0 0 4 3 2

5 0 5 4 3 2 1

Table 1.1: Addition table for Z6 .

Table 1.2: Multiplication table for Z6 .

We notice that even though 2 = 0 and 3 = 0 we have 2 3 = 0 in Z6 . This prompts the following denition. 48 Denition (Zero Divisor) An element a = 0 of Zn is called a zero divisor if ab = 0 for some b Zn . We will extend the concept of zero divisor later on to various algebras. 49 Example Let Z7 = {0, 1, 2, 3, 4, 5, 6} be the residue classes modulo 7. Construct the natural addition + table for Z7 . Also, construct the natural multiplication table for Z7

12

Chapter 1

+ 0 1 2 3 4 5 6

0 1 0 1 1 2 2 3 3 4 4 5 5 6 6 0

2 2 3 4 5 6 0 1

3 4 3 4 4 5 5 6 6 0 0 1 1 2 2 3

5 5 6 0 1 2 3 4

6 6 0 1 2 3 4 5

0 1 2 3 4 5 6

0 0 0 0 0 0 0 0

1 2 0 0 1 2 2 4 3 6 4 1 5 3 6 5

3 0 3 6 2 5 1 4

4 0 4 1 5 2 6 3

5 6 0 0 5 6 3 5 1 4 6 3 4 2 2 1

Table 1.3: Addition table for Z7 .

Table 1.4: Multiplication table for Z7 .

Solution: The required tables are given in tables 1.3 and 1.4. 50 Example Solve the equation 5x = 3 in Z11 . Solution: Multiplying by 9 on both sides 45x = 27, that is, x = 5. We will use the following result in the next section. 51 Denition Let a, b be integers with one of them different from 0. The greatest common divisor d of a, b, denoted by d = gcd(a, b) is the largest positive integer that divides both a and b. 52 Theorem (Bachet-Bezout Theorem) The greatest common divisor of any two integers a, b can be written as a linear combination of a and b, i.e., there are integers x, y with gcd(a, b) = ax + by.

Proof: Let A = {ax + by : ax + by > 0, x, y Z}. Clearly one of a, b is in A, as one of a, b is not zero. By the Well Ordering Principle, A has a smallest element, say d. Therefore, there are x0 , y0 such that d = ax0 + by0 . We prove that d = gcd(a, b). To do this we prove that d divides a and b and that if t divides a and b, then t must also divide then d. We rst prove that d divides a. By the Division Algorithm, we can nd integers q, r, 0 r < d such that a = dq + r. Then r = a dq = a(1 qx0 ) by0 .

Fields If r > 0, then r A is smaller than the smaller element of A, namely d, a contradiction. Thus r = 0. This entails dq = a, i.e. d divides a. We can similarly prove that d divides b. Assume that t divides a and b. Then a = tm, b = tn for integers m, n. Hence d = ax0 + bx0 = t(mx0 + ny0 ), that is, t divides d. The theorem is thus proved. u

13

Homework
Problem 1.4.1 Write the addition and multiplication tables of Z11 under natural addition and multiplication modulo 11. Problem 1.4.2 Solve the equation 3x2 5x + 1 = 0 in Z11 . Problem 1.4.3 Solve the equation 5x2 = 3 in Z11 .

Problem 1.4.4 Prove that if n > 0 is a composite integer, Zn has zero divisors.

Problem 1.4.5 How many solutions does the equation x4 + x3 + x2 + x + 1 = 0 have in Z11 ?

1.5 Fields
53 Denition Let F be a set having at least two elements 0F and 1F (0F = 1F ) together with two operations (multiplication, which we usually represent via juxtaposition) and + (addition). A eld F, , + is a triplet satisfying the following axioms (a, b, c) F3 : F1 Addition and multiplication are associative: (a + b) + c = a + (b + c), F2 Addition and multiplication are commutative: a + b = b + a, ab = ba (1.10) (ab)c = a(bc) (1.9)

F3 The multiplicative operation distributes over addition: a(b + c) = ab + ac F4 0F is the additive identity: 0F + a = a + 0F = a F5 1F is the multiplicative identity: 1F a = a1F = a F6 Every element has an additive inverse: a F, a + (a) = (a) + a = 0F F7 Every non-zero element has a multiplicative inverse: if a = 0F a1 F, The elements of a eld are called scalars. An important property of elds is the following. 54 Theorem A eld does not have zero divisors. aa1 = a1 a = 1F (1.15) (1.14) (1.11)

(1.12) (1.13)

14 Proof:

Chapter 1 Assume that ab = 0F . If a = 0F then it has a multiplicative inverse a1 . We deduce a1 ab = a1 0F = b = 0F . This means that the only way of obtaining a zero product is if one of the factors is 0F . u

55 Example Q, , + , R, , + , and C, , + are all elds. The multiplicative identity in each case is 1 and the additive identity is 0. 56 Example Let

Q( 2) = {a + 2b : (a, b) Q2 } 2b) + (c + 2d) = (a + c) + 2(b + d), 2d) = (ac + 2bd) + 2(ad + bc).

and dene addition on this set as (a + and multiplication as (a +

2b)(c +

Then Q + 2Q, , + is a eld. Observe 0F = 0, 1F = 1, that the additive inverse of a + 2b is a 2b, and the multiplicative inverse of a + 2b, (a, b) = (0, 0) is (a + 2b)1 = Here a2 2b2 = 0 since 1 = a 2b = a a2 2b2 2b a2 2b2 .

a+

2b

a2 2b2

2 is irrational.

57 Theorem If p is a prime, Zp , , + is a eld under multiplication modulo p and + addition modulo p. Proof: Clearly the additive identity is 0 and the multiplicative identity is 1. The additive inverse of a is p a. We must prove that every a Zp \ {0} has a multiplicative inverse. Such an a satises gcd(a, p) = 1 and by the Bachet-Bezout Theorem 52, there exist integers x, y with px + ay = 1. In such case we have 1 = px + ay = ay = a y, whence (a)1 = y. u 58 Denition A eld is said to be of characteristic p = 0 if for some positive integer p we have a F, pa = 0F , and no positive integer smaller than p enjoys this property. If the eld does not have characteristic p = 0 then we say that it is of characteristic 0. Clearly Q, R and C are of characteristic 0, while Zp for prime p, is of characteristic p. 59 Theorem The characteristic of a eld is either 0 or a prime. Proof: If the characteristic of the eld is 0, there is nothing to prove. Let p be the least positive integer for which a F, pa = 0F . Let us prove that p must be a prime. Assume that instead we had p = st with integers s > 1, t > 1. Take a = 1F . Then we must have (st)1F = 0F , which entails (s1F )(t1F ) = 0F . But in a eld there are no zero-divisors by Theorem 54, hence either s1F = 0F or t1F = 0F . But either of these equalities contradicts the minimality of p. Hence p is a prime. u

Functions

15

Homework
Problem 1.5.1 Consider the set of numbers Q( 2, 3, 6) = {a + b 2 + c 3 + d 6 : (a, b, c, d) Q4 }. Assume that Q( 2, 3, 6) is a eld under ordinary addition and multiplication. What is the multiplicative inverse of the element 2 + 2 3 + 3 6? Problem 1.5.2 Let F be a eld and a, b two non-zero elements of F. Prove that (ab1 ) = (a)b1 = a(b1 ). Problem 1.5.3 Let F be a eld and a = 0F . Prove that (a)1 = (a1 ). Problem 1.5.4 Let F be a eld and a, b two non-zero elements of F. Prove that ab1 = (a)(b1 ).

1.6 Functions
60 Denition By a function or a mapping from one set to another, we mean a rule or mechanism that assigns to every input element of the rst set a unique output element of the second set. We shall call the set of inputs the domain of the function, the set of possible outputs the target set of the function, and the set of actual outputs the image of the function. We will generally refer to a function with the following notation: D x T .

f:

Here f is the name of the function, D is its domain, T is its target set, x is the name of a typical input and f(x) is the output or image of x under f. We call the assignment x f(x) the assignment rule of the function. Sometimes x is also called the independent variable. The set f(D) = {f(a)|a D} is called the image of f. Observe that f(D) T.

f(x)

1 2 3

2 8 4

1 2 3

4 2

Figure 1.4: An injection.

Figure 1.5: Not an injection

61 Denition A function f :

X x

f(x)

is said to be injective or one-to-one if (a, b) X2 , we have a = b = f(a) = f(b).

16 This is equivalent to saying that f(a) = f(b) = a = b.

Chapter 1

62 Example The function in the diagram 1.4 is an injective function. The function represented by the diagram 1.5, however, is not injective, (3) = (1) = 4, but 3 = 1. 63 Example Prove that R \ {1} t: x is an injection. Solution: Assume t(a) = t(b). Then t(a) = t(b) a+1 = = = = = a1 (a + 1)(b 1) = = b+1 b1 (b + 1)(a 1) ab b + a 1 2b b R \ {1} x+1 x1

ab a + b 1 = 2a a = =

We have proved that t(a) = t(b) = a = b, which shews that t is injective.

1 2 3

4 2

1 2

4 2 8

Figure 1.6: A surjection

Figure 1.7: Not a surjection

64 Denition A function f : A B is said to be surjective or onto if (b B) (a A) : f(a) = b. That is, each element of B has a pre-image in A. A function is surjective if its image coincides with its target set. It is easy to see that a graphical criterion for a function to be surjective is that every horizontal line passing through a point of the target set (a subset of the y-axis) of the function must also meet the curve. 65 Example The function represented by diagram 1.6 is surjective. The function represented by diagram 1.7 is not surjective as 8 does not have a preimage.

66 Example Prove that t :

R x

R
3

is a surjection.

Functions Solution: Since the graph of t is that of a cubic polynomial with only one zero, every horizontal line passing through a point in R will eventually meet the graph of g, whence t is surjective. To prove this analytically, proceed as follows. We must prove that ( b R) (a) such that t(a) = b. We choose a so that a = b1/3 . Then t(a) = t(b1/3 ) = (b1/3 )3 = b. Our choice of a works and hence the function is surjective. 67 Denition A function is bijective if it is both injective and surjective.

17

Homework
Problem 1.6.1 Prove that h: is an injection. Problem 1.6.2 Shew that R\ f: x is a bijection. 3 2 R \ {3} 6x 2x 3 R x R x3

Chapter

Matrices and Matrix Operations

2.1 The Algebra of Matrices
68 Denition Let F, , + be a eld. An m n (m by n) matrix A with m rows and n columns with entries over F is a rectangular array of the form a11 a 21 A= . . . am1 a12 a22 . . . am2 a1n a2n , . . . amn

where (i, j) {1, 2, . . . , m} {1, 2, . . . , n}, aij F.

ij ] to denote the matrix A with entries aij . Notice that when we refer to the matrix we put parenthesesas in [aij ], and when we refer to a specic entry we do not use the surrounding parenthesesas in aij .

 As a shortcut, we often use the notation A = [a

69 Example 0 A= 1 1 2 1 3

is a 2 3 matrix and

is a 3 2 matrix.

2 1 B = 1 2 0 3

70 Example Write out explicitly the 4 4 matrix A = [aij ] where aij = i2 j2 . 18

The Algebra of Matrices Solution: This is 12 42 0 2 2 2 4 3 = 32 42 8 42 42 15 15 12 . 7 0

19

2 1 22 A= 2 3 42

11 12 12 12

12 22 22 22 32 22 42 22

12 32 22 32 32 32 42 32

3 0 5 12

8 5 0 7

71 Denition Let F, , + be a eld. We denote by Mmn (F) the set of all m n matrices with entries over F. Mnn (F) is, in particular, the set of all square matrices of size n with entries over F. 72 Denition The m n zero matrix 0mn Mmn (F) is the matrix with 0F s everywhere, 0F 0F = 0F . . . 0F 0F 0F 0F . . . 0F 0F 0F 0F . . . 0F 0F 0F 0F . . . . 0F

0mn

When m = n we write 0n as a shortcut for 0nn .

73 Denition The n n identity matrix In Mnn (F) is the matrix with 1F s on the main diagonal and 0F s everywhere else, 1F 0F In = 0F . . . 0F 0F 1F 0F . . . 0F 0F 0F 1F . . . 0F 0F 0F 0F . . . . 1F

74 Denition (Matrix Addition and Multiplication of a Matrix by a Scalar) Let A = [aij ] Mmn (F), B = [bij ] Mmn (F) and F. The matrix A + B is the matrix C Mmn (F) with entries C = [cij ] where cij = aij + bij .

20

Chapter 2

1 75 Example For A = 1 0

1 1 and B = 1 2 2 0

1 1 we have 1 3 3. 0.

1 A + 2B = 3 0

76 Theorem Let (A, B, C) (Mmn (F))3 and (, ) F2 . Then M1 Mmn (F) is close under matrix addition and scalar multiplication A + B Mmn (F), M2 Addition of matrices is commutative A+B=B+A M3 Addition of matrices is associative A + (B + C) = (A + B) + C M4 There is a matrix 0mn such that A + 0mn M5 There is a matrix A such that A + (A) = (A) + A = 0mn M6 Distributive law (A + B) = A + B M7 Distributive law ( + )A = A + B M8 1F A = A M9 (A) = ()A Proof: The theorem follows at once by reducing each statement to an entry-wise and appealing to the eld axioms. u (2.9) (2.8) (2.7) (2.6) (2.5) (2.4) (2.3) (2.2) A Mmn (F) (2.1)

Homework
Problem 2.1.1 Write out explicitly the 3 3 matrix A = [aij ] where aij = ij . Problem 2.1.2 Write out explicitly the 3 3 matrix A = [aij ] where aij = ij.

The Algebra of Matrices

Problem 2.1.3 Let a M= 0 a+b 2a a 0 c b , 1 1 a ab 2a ba 0 c b 1

21

N=

be square matrices with entries over R. Find M + N and 2M.

Problem 2.1.4 Determine x and y such that

3 1

x 2

1 2 +2 5 0

1 x

3 7 = 11 4

3 y

7 . 8

Problem 2.1.5 Determine 2 2 matrices A and B such that

1 2A 5B = 0

2 , 1

4 2A + 6B = 6

2 . 0

Problem 2.1.6 Let A = [aij ] Mnn (R). Prove that min max aij max min aij .
j i i j

Problem 2.1.7 A person goes along the rows of a movie theater and asks the tallest person of each row to stand up. Then he selects the shortest of these people, who we will call the shortest giant. Another person goes along the rows and asks the shortest person to stand up and from these he selects the tallest, which we will call the tallest midget. Who is taller, the tallest midget or the shortest giant?

Problem 2.1.8 (Putnam Exam, 1959) Choose ve elements from the matrix

11 24 12 23 6

17 10 5 4 20

25 13 14 1 7

19 15 2 8 21

16 3 18 , 22 9

no two coming from the same row or column, so that the minimum of these ve elements is as large as possible.

22

Chapter 2

2.2 Matrix Multiplication

77 Denition Let A = [aij ] Mmn (F) and B = [bij ] Mnp (F). Then the matrix product AB is dened n as the matrix C = [cij ] Mmp (F) with entries cij = l=1 ail blj : a11 a21 . . . ai1 . . . am1 a12 a22 . . . ai2 . . . am2 a1n a2n b11 . . b . 21 . ain . . . . b . n1 amn b1j b2j . . . bnj c11 b1p c21 . b2p . . = . . . . . bnp . cm1 cij c1p c2p . . . . . . . cmp

Observe that we use juxtaposition rather than a special symbol to denote matrix multiplication. This will simplify notation.In order to obtain the ij-th entry of the matrix AB we multiply elementwise the i-th row of A by the j-th column of B. Observe that AB is a m p matrix. 1 78 Example Let M = 3 2 5 6 and N = be matrices over R. Then 4 7 8 2 5 7 4 6 1 8 3 6 1 5 + 2 7 = 35+47 8 2 5 1 + 6 3 = 4 71+83

and

1 MN = 3 5 NM = 7

1 6 + 2 8 19 22 , = 43 50 36+48 5 2 + 6 4 23 34 = . 72+84 31 46

Hence, in particular, matrix multiplication is not necessarily commutative. 79 Example We have 1 1 1 1 1 1 1 2 1 1 1 1 1 2 1 1 0 1 = 0 2 0 0 0 0 0 0 , 0

over R. Observe then that the product of two non-zero matrices may be the zero matrix. 80 Example Consider the matrix 2 1 A = 0 1 4 4 3 1 0

Matrix Multiplication with entries over Z5 . Then 2 1 = 0 1 4 4 1 0 = 4 0 3 3 3 2 1 0 0 4 2 1 . 1 1 1 4 3 1 0

23

A2

 Even though matrix multiplication is not necessarily commutative, it is associative.

81 Theorem If (A, B, C) Mmn (F) Mnr (F) Mrs (F) we have (AB)C = A(BC), i.e., matrix multiplication is associative. Proof: To shew this we only need to consider the ij-th entry of each side, appeal to the associativity of the underlying eld F and verify that both sides are indeed equal to
n r

aik bkk ck j .
k=1 k =1

 By virtue of associativity, a square matrix commutes with its powers, that is, if A
Mnn (F), and (r, s) N2 , then (Ar )(As ) = (As )(Ar ) = Ar+s . 82 Example Let A M33 (R) be given by 1 A = 1 1 1 1 1 1 . 1 1

Demonstrate, using induction, that An = 3n1 A for n N, n 1. Solution: The assertion is trivial for n = 1. Assume its truth for n 1, that is, assume An1 = 3n2 A. Observe that 1 2 A = 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 3 1 = 3 3 1 3 3 3 3 = 3A. 3

Now

An = AAn1 = A(3n2 A) = 3n2 A2 = 3n2 3A = 3n1 A, and so the assertion is proved by induction.

24

Chapter 2

83 Theorem Let A Mnn (F). Then there is a unique identity matrix. That is, if E Mnn (F) is such that AE = EA = A, then E = In . Proof: It is clear that for any A Mnn (F), AIn = In A = A. Now because E is an identity, EIn = In . Because In is an identity, EIn = E. Whence In = EIn = E, demonstrating uniqueness. u 84 Example Let A = [aij ] Mnn (R) be such that aij = 0 for i > j and aij = 1 if i j. Find A2 . Solution: Let A2 = B = [bij ]. Then
n

bij =
k=1

aik akj .

Observe that the i-th row of A has i 1 0s followed by n i + 1 1s, and the j-th column of A has j 1s followed by n j 0s. Therefore if i 1 > j, then bij = 0. If i j + 1, then
j

bij =
k=i

aik akj = j i + 1.

This means that 1 0 0 A2 = . . . 0 0 2 3 1 2 0 1 . . . . . . 4 3 2 . . . 0 0 n1 n2 n3 . . . 1 0 n n 1 n 2 . . . . 2 1

0 0 0 0

Homework
Problem 2.2.1 Determine the product . 1 1 1 2 0 1 1 1 1 1 1 2

1 Problem 2.2.2 Let A = 1 1

0 1 1

0 a , B = 0 c b 1

b a c

c b. Find AB and BA. a

Matrix Multiplication

25

1 Problem 2.2.3 Find a + b + c if 2 3 0 0 Problem 2.2.4 Let N = 0 0 Problem 2.2.5 Let 2 0 0 0 3 2 0 0

2 3 1

3 1 1 2 3 2

1 2 3

1 a 2 = b c 3

a b c

a b. c

4 3 . Find N2008 . 2 0

be matrices in M44 (Z5 ) . Find the products AB and BA. Problem 2.2.6 Let x be a real number, and put 1 m(x) = x 0

2 1 A= 4 3

3 2 1 4

4 3 2 1

1 4 , 3 2

1 1 B= 1 1

1 1 1 1

1 1 1 1

1 1 1 1

0 1 0

If a, b are real numbers, prove that 1. m(a)m(b) = m(a + b). 2. m(a)m(a) = I3 , the 3 3 identity matrix.

x x2 . 2 1

Problem 2.2.7 A square matrix X is called idempotent if X2 = X. Prove that if AB = A and BA = B then A and B are idempotent.

Problem 2.2.8 Let 0 A = 1 2 0 1 2 0 0 0 0 . 1 2

Calculate the value of the innite series

I 3 + A + A2 + A3 + . Problem 2.2.9 Solve the equation 4 x x 1 = 0 4 2 0 1

26
over R. Problem 2.2.10 Prove or disprove! If (A, B) (Mnn (F))2 are such that AB = 0n , then also BA = 0n . Problem 2.2.11 Prove or disprove! For all matrices (A, B) (Mnn (F))2 , (A + B)(A B) = A2 B2 .

Chapter 2

1 Problem 2.2.12 Consider the matrix A = 3

0 non-zero 2 2 matrices B such that AB = 0

2 , where x is a real number. Find the value of x such that there are x 0 . 0

1 Problem 2.2.13 Prove, using mathematical induction, that 0 1 Problem 2.2.14 Let M = 1 0 Problem 2.2.15 Let A = 2 1 . Find M6 . 1

1 1 = 0 1

n . 1

3 . Find, with proof, A2003 . 0

Problem 2.2.16 Let (A, B, C) Mlm (F) Mmn (F) Mmn (F) and F. Prove that A(B + C) = AB + AC, (A + B)C = AC + BC, (AB) = (A)B = A(B). Problem 2.2.17 Let A M22 (R) be given by cos A= sin Demonstrate, using induction, that for n N, n 1. cos n An = sin n sin n . cos n sin . cos

Problem 2.2.18 A matrix A = [aij ] Mnn (R) is said to be checkered if aij = 0 when (j i) is odd. Prove that the sum and the product of two checkered matrices is checkered.

Matrix Multiplication
Problem 2.2.19 Let A M33 (R),

27

Prove that

1 A = 0 0 1 n A = 0 0 n 1 0

1 1 0

1 1 . 1
n(n+1) 2

n 1

Problem 2.2.20 Let (A, B) (Mnn (F))2 and k be a positive integer such that Ak = 0n . If AB = B prove that B = 0n .

a Problem 2.2.21 Let A = c .

b . Demonstrate that d A2 (a + d)A + (ad bc)I2 = 02

Problem 2.2.22 Let A M2 (F) and let k Z, k > 2. Prove that Ak = 02 if and only if A2 = 02 . Problem 2.2.23 Find all matrices A M22 (R) such that A2 = 02 Problem 2.2.24 Find all matrices A M22 (R) such that A2 = I2 Problem 2.2.25 Find a solution X M22 (R) for 1 X2 2X = 6 0 . 3

Problem 2.2.26 Find, with proof, a 4 4 non-zero matrix A such that 1 1 A 1 1 1 1 1 1 1 1 1 1 1 1 1 1 = 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 1 A = 0 1 1 0

0 0 0 0

0 0 0 0

0 0 . 0 0

Problem 2.2.27 Let X be a 2 2 matrices with real number entries. Solve the equation 1 X2 + X = 1 1 . 1

28
Problem 2.2.28 Prove, by means of induction that for the following n n we have 1 0 0 . . . 0 Problem 2.2.29 Let 1 A = 1 1 1 1 1 1 1 . 1 1 1 0 . . . 0 1 1 1 . . . 0 1 1 0 1 1 = 0 . . . . . . 0 1 3 3 1 0 . . . 0 6 3 1 0 (n1)n 2 (n2)(n1) . 2 . . . 1
n(n+1) 2

Chapter 2

Conjecture a formula for An and prove it using induction.

2.3 Trace and Transpose

85 Denition Let A = [aij ] Mnn (F). Then the trace of A, denoted by tr (A) is the sum of the diagonal elements of A, that is
n

tr (A) =
k=1

akk .

86 Theorem Let A = [aij ] Mnn (F), B = [bij ] Mnn (F). Then tr (A + B) = tr (A) + tr (B) , tr (AB) = tr (BA) . (2.10) (2.11)

Proof: The rst assertion is trivial. To prove the second, observe that AB = ( n BA = ( k=1 bik akj ). Then
n n n n

n k=1

aik bkj ) and

tr (AB) =
i=1 k=1

aik bki =
k=1 i=1

bki aik = tr (BA) ,

whence the theorem follows. u 87 Example Let A Mnn (R). Shew that A can be written as the sum of two matrices whose trace is different from 0. Solution: Write A = (A In ) + In . Now, tr (A In ) = tr (A) n and tr (In ) = n. Thus it sufces to take = Since R has innitely many elements, we can nd such an . tr (A) n , = 0.

Trace and Transpose

29

88 Example Let A, B be square matrices of the same size and over the same eld of characteristic 0. Is it possible that AB BA = In ? Prove or disprove! Solution: This is impossible. For if, taking traces on both sides 0 = tr (AB) tr (BA) = tr (AB BA) = tr (In ) = n a contradiction, since n > 0. 89 Denition The transpose of a matrix of a matrix A = [aij ] Mmn (F) is the matrix AT = B = [bij ] Mnm (F), where bij = aji . 90 Example If a b c M = d e f , g h i a T M = b c 91 Theorem Let A = [aij ] Mmn (F), B = [bij ] Mmn (F), C = [cij ] Mnr (F), F, u N. Then ATT = A, (A + B) = A + B , (AC) = C A , (A ) = (A ) .
u T T u T T T T T T

with entries in R, then

d e f

g h . i

(2.12) (2.13) (2.14) (2.15)

Proof: The rst two assertions are obvious, and the fourth follows from the third by using induction. To prove the third put AT = (ij ), ij = aji , CT = (ij ), ij = cji , AC = (uij ) and CT AT = (vij ). Then
n n n

uij =
k=1

aik ckj =
k=1

ki jk =
k=1

jk ki = vji ,

whence the theorem follows. u 92 Denition A square matrix A Mnn (F) is symmetric if AT = A. A matrix B Mnn (F) is skewsymmetric if BT = B. 93 Example Let A, B be square matrices of the same size, with A symmetric and B skew-symmetric. Prove that the matrix A2 BA2 is skew-symmetric. Solution: We have (A2 BA2 )T = (A2 )T (B)T (A2 )T = A2 (B)A2 = A2 BA2 .

30

Chapter 2

94 Theorem Let F be a eld of characteristic different from 2. Then any square matrix A can be written as the sum of a symmetric and a skew-symmetric matrix. Proof: Observe that (A + AT )T = AT + ATT = AT + A, and so A + AT is symmetric. Also, (A AT )T = AT ATT = (A AT ), and so A AT is skew-symmetric. We only need to write A as A = (21 )(A + AT ) + (21 )(A AT ) to prove the assertion. u 95 Example Find, with proof, a square matrix A with entries in Z2 such A is not the sum of a symmetric and and anti-symmetric matrix. Solution: In Z2 every symmetric matrix is also anti-symmetric, since x = x. Thus it is enough 1 . 0

0 to take a non-symmetric matrix, for example, take A = 0

Homework
Problem 2.3.1 Write Problem 2.3.4 Let (A, B) (M22 (R))2 be symmetric matrices. Must their product AB be symmetric? Prove or disprove! Problem 2.3.5 Given square matrices (A, B) (M77 (R))2 such that tr A2 = tr B2 = 1, and (A B)2 = 3I7 , nd tr (BA).

as the sum of two 33 matrices E1 , E2 , with tr (E2 ) = 10. Problem 2.3.2 Give an example of two matrices A M22 (R) and B M22 (R) that simultaneously satisfy the following properties: 0 0 0 0 . and B = 1. A = 0 0 0 0 0 0 and BA = 0 0 0

1 A = 2 3

2 3 1

3 1 M33 (R) 2

a Problem 2.3.6 Consider the matrix A = c

M22 (R). Find necessary and sufcient conditions on a, b, c, d so that tr A2 = (tr (A))2 . Problem 2.3.7 Given a square matrix A M44 (R) such that tr A2 = 4, and (A I4 )2 = 3I4 , nd tr (A). Problem 2.3.8 Prove or disprove! If A, B are square matrices of the same size, then it is always true that tr (AB) = tr (A) tr (B). Problem 2.3.9 Prove or disprove! If (A, B, C) (M33 (F))3 then tr (ABC) = tr (BAC).

b d

3. tr (A) = tr (B) = 2.

0 2. AB = 0

0 .

4. A = AT and B = BT . Problem 2.3.3 Shew that there are (A, B, C, D) (Mnn (R))4 such that AC + DB = In , CA + BD = 0n . no matrices

Special Matrices
Problem 2.3.10 Let A be a square matrix. Prove that the matrix AAT is symmetric. Problem 2.3.11 Let A, B be square matrices of the same size, with A symmetric and B skew-symmetric. Prove that the matrix AB BA is symmetric. Problem 2.3.12 Let A Mnn (F), A = [aij ]. Prove that tr AAT =
n i=1 n j=1

31
a2 . ij

Problem 2.3.13 Let X Mnn (R). Prove that if XXT = 0n then X = 0n . Problem 2.3.14 Let m, n, p be positive integers and A Mmn (R), B Mnp (R), C Mpm (R). Prove that (BA)T A = (CA)T A = BA = CA.

2.4 Special Matrices

96 Denition The main diagonal of a square matrix A = [aij ] Mnn (F) is the set {aii : i n}. The counter diagonal of a square matrix A = [aij ] Mnn (F) is the set {a(ni+1)i : i n}. 97 Example The main diagonal of the matrix 0 1 A = 3 2 9 8 5 4 7

is the set {0, 2, 7}. The counter diagonal of A is the set {5, 2, 9}. 98 Denition A square matrix is a diagonal matrix if every entry off its main diagonal is 0F . 99 Example The matrix 1 0 A = 0 2 0 0 0 0 3

is a diagonal matrix.

100 Denition A square matrix is a scalar matrix if it is of the form In for some scalar . 101 Example The matrix 4 A = 0 0 0 0 4 0 = 4I3 0 4

is a scalar matrix.

102 Denition A Mmn (F) is said to be upper triangular if

((i, j) {1, 2, , n}2 ), (i > j, aij = 0F ), ((i, j) {1, 2, , n}2 ), (i < j, aij = 0F ),

that is, every element below the main diagonal is 0F . Similarly, A is lower triangular if that is, every element above the main diagonal is 0F .

32

Chapter 2

103 Example The matrix A M34 (R) shewn is upper triangular and B M44 (R) is lower triangular. 1 c 1 B= 0 0 1 1 1 F 0F 0 0 0 0 0 1

1 A = 0 0

a b 2 0 3 0

a 0 2 1 3 t

104 Denition The Kronecker delta ij is dened by if i = j if i = j

ij =

105 Denition The set of matrices Eij Mmn (F), Eij = (ers ) such that eij = 1F and ei j = 0F , (i , j ) = (i, j) is called the set of elementary matrices. Observe that in fact ers = ir sj . Elementary matrices have interesting effects when we pre-multiply and post-multiply a matrix by them.

106 Example Let 1 A = 5 9 Then 0 E23 A = 9 0 0 10 11 12 , 0 0 0 0 0 2 6 3 7 4 8 , 12 0 0 = 0 0 0 = 0 0 0 0 0 0 0 1 . 0 0 0 0 0 2 6 . 10

E23

10 11

AE23

107 Theorem (Multiplication by Elementary Matrices) Let Eij Mmn (F) be an elementary matrix, and let A Mnm (F). Then Eij A has as its i-th row the j-th row of A and 0F s everywhere else. Similarly, AEij has as its j-th column the i-th column of A and 0F s everywhere else. Proof: Put (uv ) = Eij A. To obtain Eij A we multiply the rows of Eij by the columns of A. Now
n n

uv =
k=1

euk akv =
k=1

ui kj akv = ui ajv .

Therefore, for u = i, uv = 0F , i.e., off of the i-th row the entries of Eij A are 0F , and iv = jv , that is, the i-th row of Eij A is the j-th row of A. The case for AEij is similarly argued.u The following corollary is immediate.

Special Matrices 108 Corollary Let (Eij , Ekl ) (Mnn (F))2 , be square elementary matrices. Then Eij Ekl = jk Eil .

33

109 Example Let M Mnn (F) be a matrix such that AM = MA for all matrices A Mnn (F). Demonstrate that M = aIn for some a F, i.e. M is a scalar matrix. Solution: Assume (s, t) {1, 2, . . . , n}2 . Let M = (mij ) and Est Mnn (F). Since M commutes with Est we have 0 . . . mt1 . . . 0 0 . . . ... ... ... ... ... 0 0 0 . . 0 0 . . . mtn = Est M = MEst = . . . . . . 0 0 . 0 0 0 ... . . . m1s m2s . . . m(n1)s mns . . . . . . . . . . . . 0 . . 0 . . . . . . . . . 0 . . . 0 .

mt2 . . . 0

For arbitrary s = t we have shown that mst = mts = 0, and that mss = mtt . Thus the entries off the main diagonal are zero and the diagonal entries are all equal to one another, whence M is a scalar matrix. 110 Denition Let F and Eij Mnn (F). A square matrix in Mnn (F) of the form In + Eij is called a transvection. 111 Example The matrix 1 = 0 0 1 7 3 9 13 6 7 , 2 3 1 6 2 0 1 0 4 0 1

T = I3 + 4E13

is a transvection. Observe that if

then 1 TA = 0 0 0 1 0

1 1 A = 5 6 1 2 4 1 0 5 1 1

that is, pre-multiplication by T adds 4 times the third row of A to the rst row of A. Similarly, 1 AT = 5 1 1 6 2 1 1 7 0 3 0 0 4 1 1 0 = 5 0 1 1 5 27 , 7

1 1 5 6 7 = 5 2 3 1

34 that is, post-multiplication by T adds 4 times the rst column of A to the third row of A. In general, we have the following theorem.

Chapter 2

112 Theorem (Multiplication by a Transvection Matrix) Let In + Eij Mnn (F) be a transvection and let A Mnm (F). Then (In + Eij )A adds the j-th row of A to its i-th row and leaves the other rows unchanged. Similarly, if B Mpn (F), B(In + Eij ) adds the i-th column of B to the j-th column and leaves the other columns unchanged. Proof: Simply observe that (In + Eij )A = A + Eij A and A(In + Eij ) = A + AEij and apply Theorem 107. u Observe that the particular transvection In + ( 1F )Eii Mnn (F) consists of a diagonal matrix with 1F s everywhere on the diagonal, except on the ii-th position, where it has a . 113 Denition If = 0F , we call the matrix In + ( 1F )Eii a dilatation matrix. 114 Example The matrix 4 = 0 0 1 7 3 1 4 4 7 = 5 6 3 1 2 4 7 , 3 0 0 1 0 0 1

S = I3 + (4 1)E11

is a dilatation matrix. Observe that if 1 1 A = 5 6 1 2 4 0 SA = 0 1 0 0 0 1 0 5 1 1 1 6 2

then

that is, pre-multiplication by S multiplies by 4 the rst row of A. Similarly, 1 1 AS = 5 6 1 2 1 4 7 0 0 3 0 1 0 0 4 1 0 = 20 6 1 4 2 1 7 , 3

that is, post-multiplication by S multiplies by 4 the rst column of A. 115 Theorem (Multiplication by a Dilatation Matrix) Pre-multiplication of the matrixA Mnm (F) by the dilatation matrix In + ( 1F )Eii Mnn (F) multiplies the i-th row of A by and leaves the other rows of A unchanged. Similarly, if B Mpn (F) post-multiplication of B by In + ( 1F )Eii multiplies the i-th column of B by and leaves the other columns of B unchanged.

Special Matrices Proof: This follows by direct application of Theorem 112. u

35

116 Denition We write Iij for the matrix which permutes the i-th row with the j-th row of the identity n matrix. We call Iij a transposition matrix. n 117 Example We have 1 0 = 0 0 2 0 0 1 0 0 1 0 0 0 0 . 0 1

I4

(23)

If

then

1 5 A= 9 13

4 6 7 8 , 10 11 12 14 15 16 3 2 3 4 12 , 8 16 4 8 . 12 16

and

1 9 (23) I4 A = 5 13 1 5 = 9 13

10 11 6 7

14 15

3 7

2 6

AI4

(23)

11 10 15 14

118 Theorem (Multiplication by a Transposition Matrix) If A Mnm (F), then Iij A is the matrix obtained n from A permuting the the i-th row with the j-th row of A. Similarly, if B Mpn (F), then BIij is the n matrix obtained from B by permuting the i-th column with the j-th column of B. Proof: We must prove that Iij A exchanges the i-th and j-th rows but leaves the other rows n unchanged. But this follows upon observing that Iij = In + Eij + Eji Eii Ejj n and appealing to Theorem 107. u

36

Chapter 2

119 Denition A square matrix which is either a transvection matrix, a dilatation matrix or a transposition matrix is called an elimination matrix. In a very loose way, we may associate pre-multiplication of a matrix A by another matrix with an operation on the rows of A, and post-multiplication of a matrix A by another with an operation on the columns of A.

Homework
Problem 2.4.1 Consider the matrices 1 0 A= 1 1 0 1 1 1 1 0 1 1 0 1 , 1 1 4 0 B= 1 1 2 1 1 1 4 0 1 1 2 1 . 1 1 is transformed into the matrix hg B= ed 2b 2a

g d 2a

by a series of row and column operations. Find explicit permutation matrices P, P , an explicit dilatation matrix D, and an explicit transvection matrix T such that B = DPAP T. Problem 2.4.3 Let A Mnn (F). Prove that if

i f 2c

Find a specic dilatation matrix D, a specic transposition matrix P, and a specic transvection matrix T such that B = TDAP.

Problem 2.4.2 The matrix a A = d g b e h c f i

(X Mnn (F)), (tr (AX) = tr (BX)), then A = B. Problem 2.4.4 Let A Mnn (R) be such that (X Mnn (R)), ((XA)2 = 0n ). Prove that A = 0n .

2.5 Matrix Inversion

120 Denition Let A Mmn (F). Then A is said to be left-invertible if L Mnm (F) such that LA = In . A is said to be right-invertible if R Mnm (F) such that AR = Im . A matrix is said to be invertible if it possesses a right and a left inverse. A matrix which is not invertible is said to be singular. 121 Example The matrix A M23 (R)

has innitely many right-inverses of the form

1 0 A= 0 1

0 0

R(x,y)

1 0 = 0 1 . x y

Matrix Inversion For 1 0 0 1 1 0 0 0 x 0 1 1 = 0 y 0 , 1

37

which will never give I3 regardless of the values of a, b, c, d, f, g.

regardless of the values of x and y. Observe, however, that A does not have a left inverse, for a b 0 a b 1 0 0 = c d c d 0 , 0 1 0 f g 0 f g

122 Example If = 0, then the scalar matrix In is invertible, for (In ) 1 In = In = 1 In (In ) . 123 Example The zero matrix 0n is singular. 124 Theorem Let A Mnn (F) a square matrix possessing a left inverse L and a right inverse R. Then L = R. Thus an invertible square matrix possesses a unique inverse. Proof: Observe that we have LA = In = AR. Then L = LIn = L(AR) = (LA)R = In R = R. u 125 Denition The subset of Mnn (F) of all invertible n n matrices is denoted by GLn (F), read the linear group of rank n over F. 126 Corollary Let (A, B) (GLn (F))2 . Then AB is also invertible and (AB)1 = B1 A1 . Proof: Since AB is a square matrix, it sufces to notice that B1 A1 (AB) = (AB)B1 A1 = In and that since the inverse of a square matrix is unique, we must have B1 A1 = (AB)1 . u 127 Corollary If a square matrix S Mnn (F) is invertible, then S1 is also invertible and (S1 )1 = S, in view of the uniqueness of the inverses of square matrices. 128 Corollary If a square matrix A Mnn (F) is invertible, then AT is also invertible and (AT )1 = (A1 )T . Proof: We claim that (AT )1 = (A1 )T . For AA1 = In = (AA1 )T = IT = (A1 )T AT = In , n where we have used Theorem 91. u

38 The next few theorems will prove that elimination matrices are invertible matrices.

Chapter 2

129 Theorem (Invertibility of Transvections) Let In + Eij Mnn (F) be a transvection, and let i = j. Then (In + Eij )1 = In Eij . Proof: Expanding the product (In + Eij )(In Eij ) = In + Eij Eij 2 Eij Eij = In 2 ij Eij = In , since i = j. u 130 Example By Theorem 129, we have 1 0 0 0 1 0 3 1 0 0 1 0 0 3 1 1 0 = 0 0 1 0 0 1 0 0 0 . 1

131 Theorem (Invertibility of Dilatations) Let = 0F . Then (In + ( 1F )Eii )1 = In + (1 1F )Eii . Proof: Expanding the product (In + ( 1F )Eii )(In + (1 1F )Eii ) = In + ( 1F )Eii +(1 1F )Eii +( 1F )(1 1F )Eii = In + ( 1F )Eii +(1 1F )Eii +( 1F )(1 1F ))Eii = In + ( 1F + 1 1F + 1F 1 1F ))Eii = In , proving the assertion. u

Matrix Inversion 132 Example By Theorem 131, we have 1 0 0

39

0 0 1 2 0 0 0 1 0

0
1 2

0 1 0 0 = 0 1 1 0 0

0 0 . 1

Repeated applications of Theorem 131 gives the following corollary. 133 Corollary If 1 2 3 n = 0F , then

is invertible and 1 0 0 . . . 0 0 2 0 . . . 0 0 0 3 . . . 0

1 0 0 . . . 0

0 2 0 . . . 0

0 0 3 . . . 0

0 0 0 . . .

0 0 0 . . . n

0 0 0 . . .

0 0 0 . . . n

1 1 0 0 . . . 0

0 1 2 0 . . . 0

0 0 1 3 . . . 0

0 0 0 . . . 0

1 n

0 0 0 . . .

134 Theorem (Invertibility of Permutation Matrices) Let Sn be a permutation. Then (Iij )1 = (Iij )T . n n Proof: By Theorem 118 pre-multiplication of Iij by Iij exchanges the i-th row with the j-th row, n n meaning that they return to the original position in In . Observe in particular that Iij = (Iij )T , and n n so Iij (Iij )T = In . u n n 135 Example By Theorem 134, we have 1 0 0

0 0 1

0 1 1 0 0 0

0 0 1

0 1 1 = 0 0 0

0 1 0

0 0 . 1

136 Corollary If a square matrix can be represented as the product of elimination matrices of the same size, then it is invertible.

40 Proof: This follows from Corollary 126, and Theorems 129, 131, and 134. u 137 Example Observe that 1 0 A = 0 3 0 0 0 4 1

Chapter 2

is the transvection I3 + 4E23 followed by the dilatation of the second column of this transvection by 3. Thus 1 0 0 and so 0 3 0 0 1 4 = 0 1 0 0 1 0 0 1 4 0 1 0 0 3 0 0 0 , 1

1 0 0

0 3 0

0 4 1

1 =

1 0 0

0 0 3 0 0 1 0
1 3

1 0 0 1 0 0

0 0
1 3

0 1 0 0 0 1 1 0 0 0 4 . 3 1

1 0 0

0 1 0

0 4 1

0 4 1

138 Example We have 1 0 0 1 1 0 1 1 1 = 0 1 0 1 1 0 0 1 0 0 1 0 0 1 0 0 1 , 1

Matrix Inversion hence 1 0 0 1 1 0 1 1 1 1 = 1 0 0 0 0 1 1 0 1 0 1 1 0 0 1 1 0 0 0 1 1

41

1 0 0 1 0 0

0 1 1 1 0 0 0 1 1 1 0 0 1 . 1

1 0 1 0 0 1

In the next section we will give a general method that will permit us to nd the inverse of a square matrix when it exists.

a b M22 (R). Then 139 Example Let T = c d a c Thus if ad bc = 0 we see that

d adbc c adbc b adbc a adbc

b d d c

b 1 = (ad bc) a 0

0 1

T 1 = 140 Example If

1 1 A= 1 1

1 1 1 1

1 1 1 1

1 1 , 1 1

42 then A is invertible, for an easy computation shews that 2

Chapter 2

1 1 1 1 A2 = 1 1 1 1 whence the inverse sought is 1 1 1 1 = A= 4 4 1 1 1 1 1 1 1 1 1 1

1 1 1 1

1 1 = 4I4 , 1 1

A1

1 1/4 1/4 1/4 1 1/4 1/4 1/4 = 1 1/4 1/4 1/4 1 1/4 1/4 1/4

1/4 1/4 . 1/4 1/4

141 Example A matrix A Mnn (R) is said to be nilpotent of index k if satises A = 0n , A2 = 0n , . . . , Ak1 = 0n and Ak = 0n for integer k 1. Prove that if A is nilpotent, then In A is invertible and nd its inverse. Solution: To motivate the solution, think that instead of a matrix, we had a real number x with |x| < 1. Then the inverse of 1 x is (1 x)1 = 1 1x = 1 + x + x2 + x3 + .

Notice now that since Ak = 0n , then Ap = 0n for p k. We conjecture thus that (In A)1 = In + A + A2 + + Ak1 . The conjecture is easily veried, as (In A)(In + A + A2 + + Ak1 ) = In + A + A2 + + Ak1 (A + A2 + A3 + + Ak ) = In and (In + A + A2 + + Ak1 )(In A) = In A + A A2 + A3 A4 + + Ak2 Ak1 + Ak1 Ak = In .

Matrix Inversion 142 Example The inverse of A M33 (Z5 ), 2 0 A = 0 3 0 0 3 0 = 0 2 0 0 0 3 0 0 4 0 0 0 4 0 0 , 4 0 1 0 0 0 1

43

is

A1

as 2 = 0 0 0 3 0

AA1

0 0 1 2 0 = 0 0 4 0

143 Example (Putnam Exam, 1991) Let A and B be different n n matrices with real entries. If A3 = B3 and A2 B = B2 A, prove that A2 + B2 is not invertible. Solution: Observe that (A2 + B2 )(A B) = A3 A2 B + B2 A B3 = 0n . If A2 + B2 were invertible, then we would have A B = (A2 + B2 )1 (A2 + B2 )(A B) = 0n , contradicting the fact that A and B are different matrices. 144 Lemma If A Mnn (F) has a row or a column consisting all of 0F s, then A is singular. Proof: If A were invertible, the (i, i)-th entry of the product of its inverse with A would be 1F . n But if the i-th row of A is all 0F s, then k=1 aik bki = 0F , so the (i, i) entry of any matrix product with A is 0F , and never 1F . u
1 Problem 2.5.1 The inverse of the matrix A = 1 1 1 1 a 1 is the matrix A = 1 b 1 . Determine a 1 1 0 1 1 2 1 2 3 Problem 2.5.2 A square matrix A satises A3 = 0n but A4 = 0n . Demonstrate that In + A is invertible and nd, with proof, its inverse.

Problem 2.5.3 Prove or disprove! If (A, B, A + B) (GLn (R))3 then (A + B)1 = A1 + B1 . Problem 2.5.4 Let S GLn (F), (A, B) (Mnn (F))2 , and k a positive integer. Prove that if B = SAS1 then Bk = SAk S1 . Problem 2.5.5 Let A Mnn (F) and let k be a positive

and b.

44
integer. Prove that A is invertible if and only if Ak is invertible.

Chapter 2

Problem 2.5.6 Let S GLn (C), A Mnn (C) with Ak = 0n for some positive integer k. Prove that both In SAS1 and In S1 AS are invertible and nd their inverses.

Problem 2.5.10 Let A

Problem 2.5.7 Let A and B be square matrices of the same size such that both A B and A + B are invertible. Put C = (A B)1 + (A + B)1 . Prove that

Mnn (F), n > 1, (a, b) F . Determine when A is invertible and nd this inverse when it exists. Problem 2.5.11 Let (A, B) (Mnn (F))2 be matrices such that A + B = AB. Demonstrate that A In is invertible and nd this inverse. Problem 2.5.12 Let S GLn (F) and A Mnn (F). Prove that tr (A) = tr SAS1 .

a b b . . . b

b a b . . . b

b b a . . . b

b b b . . . a

ACA ACB + BCA BCB = 2A.

Problem 2.5.8 Let A, B, C be non-zero square matrices of the same size over the same eld and such that ABC = 0n . Prove that at least two of these three matrices are not invertible.

Problem 2.5.13 Let A Mnn (R) be a skew-symmetric matrix. Prove that In + A is invertible. Furthermore, if B = (In A)(In + A)1 , prove that B1 = BT . Problem 2.5.14 A matrix A Mnn (F) is said to be a magic square if the sum of each individual row equals the sum of each individual column. Assume that A is a magic square and invertible. Prove that A1 is also a magic square.

Problem 2.5.9 Let (A, B) (Mnn (F))2 be such that A2 = B2 = (AB)2 = In . Prove that AB = BA.

2.6 Block Matrices

145 Denition Let A Mmn (F), B Mms (F), C Mrn (F), D Mrs (F). We use the notation A L= C for the block matrix L M(m+r)(n+s) (F). B D

and

 If (A, A ) (M

2 m (F)) ,

(B, B ) (Mmn (F))2 , (C, C ) (Mnm (F))2 , (D, D ) (Mm (F))2 , A S= C B , D A T = C

then it is easy to verify that

B , D

AA + BC ST = CA + DC

AB + BD . CB + DD

Rank of a Matrix 146 Lemma Let L M(m+r)(m+r) (F) be the square block matrix A L= 0rm C , B

45

with square matrices A Mm (F) and B Mrr (F), and a matrix C Mmr (F). Then L is invertible if and only if A and B are, in which case A L1 = 0rm
1

A1 CB1 B1

Proof:

Assume rst that A, and B are invertible. Direct calculation yields A 0rm C A B 0rm
1

CB

B1

AA = 0rm Im = 0rm = Im+r .

AA

CB

+ CB

BB1 0mr Ir

E H , with E Mm (F) and K Mrr (F), but that, Assume now that L is invertible, L1 = J K say, B is singular. Then Im 0rm 0mr = LL1 Ir

which gives BK = Ir , i.e., B is invertible, a contradiction. u

AE + CJ = BJ

A = 0rm

C E B J

H K

AH + BK , BK

2.7 Rank of a Matrix

147 Denition Let (A, B) (Mmn (F))2 . We say that A is row-equivalent to B if there exists a matrix R GLm (F) such that B = RA. Similarly, we say that A is column-equivalent to B if there exists a matrix C GLm (F) such that B = AC. We say that A and B are equivalent if (P, Q) GLm (F) GLn (F) such that B = PAQ.

46

Chapter 2

148 Theorem Row equivalence, column equivalence, and equivalence are equivalence relations. Proof: We prove the result for row equivalence. The result for column equivalence, and equivalence are analogously proved.

Since Im GLm (F) and A = Im A, row equivalence is a reexive relation. Assume (A, B) (Mmn (F))2 and that P GLm (F) such that B = PA. Then A = P1 B and since P1 GLm (F), we see that row equivalence is a symmetric relation. Finally assume (A, B, C) (Mmn (F))3 and that P GLm (F), P GLm (F) such that A = PB, B = P C. Then A = PP C. But PP GLm (F) in view of Corollary 126. This completes the proof. u 149 Theorem Let A Mmn (F). Then A can be reduced, by means of pre-multiplication and postmultiplication by elimination matrices, to a unique matrix of the form Ir 0(mr)r 0r(nr) 0(mr)(nr)

Dm,n,r =

(2.16)

called the Hermite normal form of A. Thus there exist P GLm (F), Q GLn (F) such that Dm,n,r = PAQ. The integer r 0 is called the rank of the matrix A which we denote by rank (A). Proof: If A is the m n zero matrix, then the theorem is obvious, taking r = 0. Assume hence that A is not the zero matrix. We proceed as follows using the Gau-Jordan Algorithm. GJ-1 Since A is a non-zero matrix, it has a non-zero column. By means of permutation matrices we move this column to the rst column. GJ-2 Since this column is a non-zero column, it must have an entry a = 0F . Again, by means of permutation matrices, we move the row on which this entry is to the rst row. GJ-3 By means of a dilatation matrix with scale factor a1 , we make this new (1, 1) entry into a 1F . GJ-4 By means of transvections (adding various multiples of row 1 to the other rows) we now annihilate every entry below the entry (1, 1). This process ends up in a matrix of the form 1F 0F P1 AQ1 = 0F 0F 0F b22 b32 . . . bm2 b23 b33 . . . bm3 b2n b3n . . . . bmn

(2.17)

Here the asterisks represent unknown entries. Observe that the bs form a (m 1) (n 1) matrix. GJ-5 Apply GJ-1 through GJ-4 to the matrix of the bs.

Rank of a Matrix Observe that this results in a matrix of the form 1F 0F P2 AQ2 = 0F 0F 0F 1F 0F . . . 0F c33 . . . cm3 c3n . . . . cmn

47

(2.18)

GJ-6 Add the appropriate multiple of column 1 to column 2, that is, apply a transvection, in order to make the entry in the (1, 2) position 0F . This now gives a matrix of the form 1F 0F P3 AQ3 = 0F 0F 0F 0F 1F 0F . . . 0F c33 . . . cm3 c3n . . . . cmn

(2.19)

The matrix of the cs has size (m 2) (n 2). GJ-7 Apply GJ-1 through GJ-6 to the matrix of the cs, etc. Observe that this process eventually stops, and in fact, it is clear that rank (A) min(m, n). Suppose now that A were equivalent to a matrix Dm,n,s with s > r. Since matrix equivalence is an equivalence relation, Dm,n,s and Dm,n,r would be equivalent, and so there would be R GLm (F), S GLn (F), such that RDm,n,r S = Dm,n,s , that is, RDm,n,r = Dm,n,s S1 . Partition R and S1 as follows S11 = S21 S31 S12 S22 S32 S13 S23 , S33

R11 R= R21

R12 , R22

S1

with (R11 , S11 )2 (Mrr (F))2 , S22 M(sr)(sr) (F). We have R11 RDm,n,r = R21 R12 Ir R22 0(mr)r 0(mr)r R11 = 0r(mr) R21 0(mr)r , 0r(mr)

48 and = Ir 0(sr)r 0(ms)r 0r(sr) Isr 0(ms)(sr) S12 S22 0(ms)(sr) S11 0(sr)(ns) S21 0(ms)(ns) S31 S13 S23 0(ms)(ns) . S13 S23 0(ms)(ns) 0r(ns) S12 S22 S32 S13 S23 S33

Chapter 2

Dm,n,s S1

Since we are assuming

S11 S21 0(ms)r

R11 R21

we must have S12 = 0r(sr) , S13 = 0r(ns) , S22 = 0(sr)(sr) , S23 = 0(sr)(ns) . Hence S1 S11 = S21 S31 0r(sr) 0(sr)(sr) S32 0r(ns) 0(sr)(ns) . S33

S11 0(mr)r = S21 0r(mr) 0(ms)r

S12 S22 0(ms)(sr)

The matrix

is non-invertible, by virtue of Lemma 144. This entails that S1 is non-invertible by virtue of Lemma 146. This is a contradiction, since S is assumed invertible, and hence S1 must also be invertible. u Albeit the rank of a matrix is unique, the matrices P and Q appearing in Theorem 149 are not necessarily unique. For example, the matrix 1 0 0 0 1 0

0(sr)(sr) S32

0(sr)(ns) S33

has rank 2, the matrix

1 0 0

0 1 0

x y 1

Rank of a Matrix is invertible, and an easy computation shews that 1 0 x 1 0 1 0 1 y 0 1 0 0 0 1 0 0

49

regardless of the values of x and y.

1 0 0 = 0 1 , 1 0 0

150 Corollary Let A Mmn (F). Then rank (A) = rank AT . Proof: Let P, Q, Dm,n,r as in Theorem 149. Observe that PT , QT are invertible. Then PAQ = Dm,n,r = QT AT PT = DT m,n,r = Dn,m,r , and since this last matrix has the same number of 1F s as Dm,n,r , the corollary is proven. u 151 Example Shew that 0 2 A= 0 1 3 0

has rank (A) = 2 and nd invertible matrices P GL2 (R) and Q GL3 (R) such that 1 0 PAQ = 0 1 0 . 0

We now subtract twice the second row from the rst, by effecting 1 2 3 2 0 1 0 1 0 3 = 0 0 0 0 . 1 0

Solution: We rst transpose the rst and 0 0 2 3 0 0 1 0 1

third columns by effecting 0 1 3 2 0 . 1 0 = 0 1 0 0 0

Finally, we divide the rst row by 3,

We conclude that

1/3 0 3 0 1 0

0 1

0 1 0 = 0 0 1 0 3 0 0 1

0 . 0

1/3 0

0 1 0 1

2 0 0 1

2 1

0 1 1 1 0 = 0 0 0

0 1

0 , 0

50 from where we may take

Chapter 2

1/3 0 1 P= 0 1 0 and

2 1/3 2/3 = 1 0 1

0 0 Q = 0 1 1 0

1 0 . 0

In practice it is easier to do away with the multiplication by elimination matrices and perform row and column operations on the augmented (m + n) (m + n) matrix In A 0nm . Im

152 Denition Denote the rows of a matrix A Mmn (F) by R1 , R2 , . . . , Rm , and its columns by C1 , C2 , . . . , Cn . The elimination operations will be denoted as follows. Exchanging the i-th row with the j-th row, which we denote by Ri Rj , and the s-th column by the t-th column by Cs Ct .

A transvection on the rows will be denoted by Ri +Rj Ri , and one on the columns by Cs +Ct Cs . 153 Example Find the Hermite normal form of 1 0 A= 1 1

A dilatation of the i-th row by a non-zero scalar F \ {0F }, we will denote by Ri Ri . Similarly, Cj Cj denotes the dilatation of the j-th column by the non-zero scalar .

0 0 . 1 2

Solution: First observe that rank (A) min(4, 2) = 2, so the rank can be either 1 or 2 (why

Rank of a Matrix not 0?). Form the augmented matrix 1 0 1 0 1 1 0 1 0 0 1 2 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 . 0 0 1 0 0 1

51

Perform R5 + R3 R5 and R6 + R3 R6 successively, obtaining 1 0 1 0 0 0 0 0 0 1 0 1 1 0 0 0 1 0 0 1 0 0 1 2 0 0 0 0 0 0 . 0 0 1 0 0 1 0 0 0 0 1 2

Perform R6 2R5 R6

Perform R4 R5

1 0 1 0 0 0 1 0 1 0 0 0

0 1 0 0 1 0

0 0 1 0 1 1

0 0 0 1 0 0

0 0 0 . 0 0 1 0 0 0 . 0 0 1

0 1 0 1 0 0

0 0 1 1 0 1

0 0 0 0 1 0

0 0 0 1 0 2

52 Finally, perform R3 R3

Chapter 2

We conclude that 1 1 0 1 0 0 1 0

1 0 1 0 0 0 0

0 1 0 1 0 0

0 0 1 1 0 1

0 0 0 0 1 0

0 0 0 0 0 . 1 0 0 0 2 1 0 0 1 1 0 2 1 0 0 = 1 0 0 0 1 . 0 0

0 1 1 0 0 0 0 1 1 2 1

154 Theorem Let A Mmn (F), B Mnp (F). Then rank (AB) min(rank (A) , rank (B)). Proof: We prove that rank (A) rank (AB). The proof that rank (B) rank (AB) is similar and left to the reader. Put r = rank (A) , s = rank (AB). There exist matrices P GLm (F), Q GLn (F), S GLm (F), T GLp (F) such that PAQ = Dm,n,r , Now Dm,p,s = SABT = SP1 Dm,n,r Q1 BT, from where it follows that PS1 Dm,p,s = Dm,n,r Q1 BT. Now the proof is analogous to the uniqueness proof of Theorem 149. Put U = PS1 GLm (R) and V = Q1 BT Mnp (F), and partition U and V as follows: U11 U= U21 U12 , U22 V11 V= V21 V12 , V22 SABT = Dm,p,s .

with U11 Ms (F), V11 Mrr (F). Then U11 UDm,p,s = U21

U12 Is U22 0(ms)s

0s(ps) 0(ms)(ps)

Mmp (F),

Rank of a Matrix and Ir Dm,p,s V = 0(mr)r U11 U21 V11 0(mr)(nr) V21 V11 0r(nr) V12 Mmp (F). V22 V12

53

From the equality of these two m p matrices, it follows that = 0(ms)(ps) 0(mr)r 0s(ps)

0(mr)(nr)

If s > r then (i) U11 would have at least one row of 0F s meaning that U11 is non-invertible by Lemma 144. (ii) U21 = 0(ms)s . Thus from (i) and (ii) and from Lemma 146, U is not invertible, which is a contradiction. u 155 Corollary Let A Mmn (F), B Mnp (F). If A is invertible then rank (AB) = rank (B). If B is invertible then rank (AB) = rank (A). Proof: Using Theorem 154, if A is invertible rank (AB) rank (B) = rank A1 AB rank (AB) , and so rank (B) = rank (AB). A similar argument works when B is invertible. u 156 Example Study the various possibilities for the rank of the matrix 1 A = b + c bc c + a a + b . ca ab 1 1

Solution: Performing R2 (b + c)R1 R2 and R3 bcR1 R3 , we nd 1 1 0 a b 0 0 . ac (b c)(a c) 1

Performing C2 C1 C2 and C3 C1 C3 , we nd 0 1 0 a b 0 0 0

We now examine the various ways of getting rows consisting only of 0s. If a = b = c, the last two rows are 0-rows and so rank (A) = 1. If exactly two of a, b, c are equal, the last row is a 0-row, but the middle one is not, and so rank (A) = 2 in this case. If none of a, b, c are equal, then the rank is clearly 3.

. ac (b c)(a c)

54

Chapter 2

Homework
Problem 2.7.1 On a symmetric matrix A Mnn (R) with n 3, R3 3R1 R3 C3 3C1 C3

successively followed by

are performed. Is the resulting matrix still symmetric? Problem 2.7.2 Find the rank of a + 1 a + 2 a + 3 a+4 a+2 a+3 a+4 a+5 a+3 a+4 a+5 a+6

a+4 a+5 a+6 a+7

a + 5 a + 6 M55 (R). a + 7 a+8

Problem 2.7.3 Let A, B be arbitrary n n matrices over R. Prove or disprove! rank (AB) = rank (BA) .

1 0 Problem 2.7.4 Determine the rank of the matrix 2 2 4 Problem 2.7.5 Suppose that the matrix x2

1 0 2 0

0 1 2 0

0 1 . 2 2

2 M22 (R) has rank 1. How many possible values can x assume? x

Problem 2.7.6 Demonstrate that a non-zero n n matrix A over a eld F has rank 1 if and only if A can be factored as A = XY, where X Mn1 (F) and Y M1n (F). Problem 2.7.7 Study the various possibilities for the rank of the matrix

1 a 1 b when (a, b) R2 .

a 1 b 1

1 b 1 a

b 1 a 1

Rank and Invertibility

55

1 m Problem 2.7.8 Find the rank of 1 1

1 1 m 1

0 1 1 m

1 1 as a function of m C. 0 2
2

a ab Problem 2.7.9 Determine the rank of the matrix ab b2 1 a Problem 2.7.10 Determine the rank of the matrix c ac

ab a2 b2 ab

ab b2 a2 ab

b ab . ab 2 a 1 b . d bd 1 x 2 2 1. Find x. 1

1 b c bc

1 a d ad

1 Problem 2.7.11 Let A M32 (R), B M22 (R), and C M23 (R) be such that ABC = 2 1

Problem 2.7.12 Let B be the matrix obtained by adjoining a row (or column) to a matrix A. Prove that either rank (B) = rank (A) or rank (B) = rank (A) + 1. Problem 2.7.13 Let A Mnn (R). Prove that rank (A) = rank AAT . Find a counterexample in the case A Mnn (C). Problem 2.7.14 Prove that the rank of a skew-symmetric matrix with real number entries is an even number.

2.8 Rank and Invertibility

157 Theorem A matrix A Mmn (F) is left-invertible if and only if rank (A) = n. A matrix A Mmn (F) is right-invertible if and only if rank (A) = m. Proof: Observe that we always have rank (A) n. If A is left invertible, then L Mnm (F) such that LA = In . By Theorem 154, n = rank (In ) = rank (LA) rank (A) , whence the two inequalities give rank (A) = n. Conversely, assume that rank (A) = n. Then rank AT = n by Corollary 150, and so by

56 Theorem 149 there exist P GLm (F), Q GLn (F), such that

Chapter 2

PAQ = This gives

In 0(mn)n

QT AT PT = In

0n(mn) .

QT AT PT PAQ = In

= =

AT PT PA = (QT )1 Q1 ((QT )1 Q1 )1 AT PT PA = In ,

and so ((QT )1 Q1 )1 AT PT P is a left inverse for A.

The right-invertibility case is argued similarly. u

By combining Theorem 157 and Theorem 124, the following corollary is thus immediate. 158 Corollary If A Mmn (F) possesses a left inverse L and a right inverse R then m = n and L = R. We use Gau-Jordan Reduction to nd the inverse of A GLn (F). We form the augmented matrix T = [A|In ] which is obtained by putting A side by side with the identity matrix In . We perform permissible row operations on T until instead of A we obtain In , which will appear if the matrix is invertible. The matrix on the right will be A1 . We nish with [In |A1 ].

If A Mnn (R) is non-invertible, then the left hand side in the procedure above will not reduce to In . 159 Example Find the inverse of the matrix B M33 (Z7 ),

6 0 B = 3 2 1 0

1 0 . 1

Rank and Invertibility Solution: We have

57

6 3 1

0 2 0

1 1 0 0 1 0

0 1 0

0 0 1

R1 R3

R3 6R1 R3 R2 3R1 R2

1 3 6 1 0 0

0 2 0 0 2 0 0 2 0 0 1 0

1 0 0 0 1 1 1 0 4 0 2 1 0 1 0 5 2 1 0 3 0 6 1 4

0 1 0 0 1 0 0 1 0 0 4 0

1 0 0 1 4 1 6 2 1

R2 2R3 R2 5R1 +R3 R1

3R1 R1 ; 4R3 R3 4R2 R2

5 0 0 1 0 0

4 1 . 4

We conclude that

6 3 1

0 2 0

1 0 1

3 = 6 4

0 4 0

4 1 . 4

0 160 Example Use Gau-Jordan reduction to nd the inverse of the matrix A = 4 3 nd A2001 .

1 3 3

1 4 . Also, 4

58 Solution: Operating on the augmented matrix 0 4 3 1 1 4 4 1 0 0 1 0 0 3 3 0 0 1

Chapter 2

R2 R3 R2

R3 3R2 R3

0 1 3 0 1 0

1 0 3 1 0 3

1 1 0 4 0 0

0 1 0 0 1 3 0 1 3

0 1 1

1 1 0 4 1 0 3 4 0 3 0 4 3 0 0

R3 +3R1 R3

R1 +R3 R1

R1 R2

0 1 0 1 0 0

0 1 0

1 1 0 0 0 1

1 0 0 0 0 1 0 0 1 0 0 1

3 1 3 1 3 3

4 1 4

0 1 4

0 1 4

Thus we deduce that 0 = 4 3 1 3 3

1 4 . 4

A1

From A1 = A we deduce A2 = In . Hence A2000 = (A2 )1000 = I1000 = In and A2001 = A(A2000 ) = n AIn = A. 161 Example Find the inverse of the triangular matrix A Mnn (R), 1 0 A = 0 . . . 0 1 1 0 . . . 0 1 1 1 . . . 0 1 1 1 . . . . 1

1 4 = A. 4

Rank and Invertibility Solution: Form the augmented matrix

59

1 0 0 . . . 0

1 1 0 . . . 0

1 1 1 . . . 0

1 1 1 . . . 1

1 0 0 . . . 0

0 0 1 0 0 1 . . . . . .

0 0

0 0 0 , . . . 1

and perform Rk Rk+1 Rk successively for k = 1, 2, . . . , n 1, obtaining 1 0 0 . . . 0 0 0 0 1 1 0 0 . . . 0 1 0 . . . 0 0 1 0 . . . 0 0 1 . . . 0 0 0 . . . 1 1 1 . . . 0 0 0 0 , . . . 1

whence 1 0 = 0 . . . 0 1 1 0 . . . 0 0 0 0 0 , . . . 1

1 1 . . . 0

A1

that is, the inverse of A has 1s on the diagonal and 1s on the superdiagonal. 162 Theorem Let A Mnn (F) be a triangular matrix such that a11 a22 ann = 0F . Then A is invertible. Proof: Since the entry akk = 0F we multiply the k-th row by a1 and then proceed to subtract kk the appropriate multiples of the preceding k 1 rows at each stage. u 163 Example (Putnam Exam, 1969) Let A and B be matrices of size 3 2 and 2 3 respectively. Suppose that their product AB is given by 8 AB = 2 2 2 2 5 4 . 4 5

60 Demonstrate that the product BA is given by 9 BA = 0 Solution: Observe that 0 . 9

Chapter 2

8 (AB)2 = 2 2

2 5 4

2 8 4 2 5 2

2 5 4

2 72 4 = 18 5 18

18 18 45 36 = 9AB. 36 45

Performing R3 + R2 R3 , R1 4R2 R2 , and 2R3 + R1 R3 in succession we see that 8 2 2 2 5 4 2 4 5 0 2 0 18 18 5 4 0 0 0 18 5 2 0 0 0 1 0 0 18 5 0 0 0 0 1 , 0

and so rank (AB) = 2. This entails that rank (AB)2 = 2. Now, since BA is a 2 2 matrix, rank (BA) 2. Also 2 = rank (AB)2 = rank (ABAB) rank (ABA) rank (BA) , and we must conclude that rank (BA) = 2. This means that BA is invertible and so (AB)2 = 9AB A(BA 9I2 )B = 03 BA(BA 9I2 )BA = B03 A BA(BA 9I2 )BA = 02 (BA)1 BA(BA 9I2 )BA(BA)1 = (BA)1 02 (BA)1 BA 9I2 = 02

= = = = =

Homework
Problem 2.8.1 Find the inverse of the matrix 3 1 M33 (Z7 ). 2

1 2 3

2 3 1

Rank and Invertibility

Problem 2.8.2 Let (A, B) M33 (R) be given by a A = 1 0 Find B1 and prove that AT = BAB1 . 1 Problem 2.8.3 Let A = 1 1 0 1 0 where x = 0 is a real number. Find A . x 0 0 1 1 0 1 0 is the matrix M = 1 1 1 1 1 1 0 a, nd (a, b). b b 0 1 c 0 , 0 0 B= 0 1 0 1 a 1 a . b

61

0 1 1

1 Problem 2.8.4 If the inverse of the matrix M = 1 0 1 Problem 2.8.5 Let A = 1 1 0 1 1

0 n 1 0 and let n > 0 be an integer. Find (A ) . 1

Problem 2.8.6 Give an example of a 2 2 invertible matrix A over R such that A + A1 is the zero matrix. Problem 2.8.7 Find all the values of the parameter a for which the matrix B given below is not invertible. 2 1 a

1 B= 0 2 Problem 2.8.8 Find the inverse of the triangular matrix

a+2 a 1

a 0 0 assuming that abc = 0.

2a b 0

3a 2b M33 (R) c

62
Problem 2.8.9 Under what conditions is the matrix b c 0

Chapter 2

a 0 c

invertible? Find the inverse under these conditions.

0 a b

Problem 2.8.10 Let A and B be n n matrices over a eld F such that AB is invertible. Prove that both A and B must be invertible. Problem 2.8.11 Find the inverse of the matrix

1 + a 1 1

1 1+b 1

1 1 1+c

Problem 2.8.12 Prove that for the n n (n > 1) matrix 0 1 1 . . . 1 1 0 1 . . . 1 1 1 0 . . . 1 ... ... ... ... ... 1 1 1 . . . 0 1 2 n 1 1 = n1 1 . . . 1 1 2n 1 . . . 1 1 1 2n . . . 1 ... ... ... ... ... 1 1 . . . 2n 1

Problem 2.8.13 Prove that the n n (n > 1) matrix 1 + a 1 1 . . . 1

1 1 1+a . . . 1

... ... ... ... ...

1 1 1 . . . 1+a

1 1 1 . . . 1na

1+a 1 . . . 1

has inverse

1 n a 1 1 1 a(n + a) . . . 1

1 1na 1 . . . 1

1 1 1na . . . 1

... ... ... ... ...

Rank and Invertibility

Problem 2.8.14 Prove that 1 (2n 1) (2n 3) . . . 3 2 n 2 1 2n3 2 . . . 2 + n2
2

63

3 1 (2n 1) . . . 5

5 3 1 . . . 7

7 5 3 . . . 9

. . .

has inverse

(2n 1) (2n 3) (2n 5) . . . 1 . . . 2 2 2 . . . 2 n2

2+n

2 2 + n2 2 n2 . . . 2

2 2 2 + n2 . . . 2

2 n2 2 . . . 2

Problem 2.8.15 Prove that the n n (n > 1) matrix 1 + a1 1 1 . . . 1 has inverse 1 a1 s a2 1 1 a2 a1 1 1 s a3 a1 . . . 1 an a1 1 a1 a2 1 a2 s a2 2 1 a3 a2 . . . 1 an a2 1 a1 a3 1 a2 a3 1 a3 s a2 3 . . . 1 an a3 ... ... ... ... ... , 1 an s a2 n 1 a1 an 1 a2 an 1 a3 a1 n . . . 1 1 + a2 1 . . . 1 1 1 1 + a3 . . . 1 ... ... ... ... ... 1 1 1 . . . 1 + an

where s = 1 +

1 a1

1 a2

+ +

1 . an

Problem 2.8.16 Let A M55 (R). Shew that if rank A2 < 5, then rank (A) < 5. Problem 2.8.17 Let p be an odd prime. How many invertible 2 2 matrices are there with entries all in Zp ? Problem 2.8.18 Let A, B be matrices of the same size. Prove that rank (A + B) rank (A) + rank (B).

64

Chapter 2

0 Problem 2.8.19 Let A M3,2 (R) and B M2,3 (R) be matrices such that AB = 1 1 BA = I2 .

1 0 1

1 1. Prove that 2

Chapter

3
a11 x1 + a12 x2 + a13 x3 + + a1n xn = y1 , a21 x1 + a22 x2 + a23 x3 + + a2n xn = y2 , . . .

Linear Equations
3.1 Denitions
We can write a system of m linear equations in n variables over a eld F

in matrix form as

am1 x1 + am2 x2 + am3 x3 + + amn xn = ym , a11 a 21 . . . am1 a12 a22 . . . am2 . . . a1n x1 y1 x y a2n 2 2 . = . . . . . . . . . ym xn amn

(3.1)

We write the above matrix relation in the abbreviated form AX = Y,

(3.2)

where A is the matrix of coefcients, X is the matrix of variables and Y is the matrix of constants. Most often we will dispense with the matrix of variables X and will simply write the augmented matrix of the system as a11 a12 a1n y1 a 21 a22 a2n y2 . [A|Y] = (3.3) . . . . . . . . . . . . . . . am1 am2 amn ym 164 Denition Let AX = Y be as in 3.1. If Y = 0m1 , then the system is called homogeneous, otherwise it is called inhomogeneous. The set {X Mn1 (F) : AX = 0m1 } 65

is called the kernel or nullspace of A and it is denoted by ker (A).

66

Chapter 3

 Observe that we always have 0

n1

ker (A) Mmn (F).

165 Denition A system of linear equations is consistent if it has a solution. If the system does not have a solution then we say that it is inconsistent. 166 Denition If a row of a matrix is non-zero, we call the rst non-zero entry of this row a pivot for this row. 167 Denition A matrix M Mmn (F) is a row-echelon matrix if All the zero rows of M, if any, are at the bottom of M. For any two consecutive rows Ri and Ri+1 , either Ri+1 is all 0F s or the pivot of Ri+1 is immediately to the right of the pivot of Ri . The variables accompanying these pivots are called the leading variables. Those variables which are not leading variables are the free parameters. 168 Example The matrices 1 0 0 0 0 0 0 0 1 2 0 0 1 2 , 3 0 1 0 0 0 0 0 0 0 1 0 0 0 1 1 , 0 0

are in row-echelon form, with the pivots circled, but the matrices 1 0 0 0 are not in row-echelon form. Observe that given a matrix A Mmn (F), by following Gau-Jordan reduction la Theorem 149, we can nd a matrix P GLm (F) such that PA = B is in row-echelon form. 169 Example Solve the system of linear equations 1 0 0 0 1 1 2 1 0 1 0 0 1 x 3 0 y 1 = . 1 z 4 w 2 6 0 1 0 1 0 1 0 0 1 2 , 1 0 1 0 0 0 0 0 0 0 1 1 0 0 , 0 1 0 0

Denitions Solution: Observe that the matrix of coefcients is already in row-echelon form. Clearly every variable is a leading variable, and by back substitution 2w = 6 = w = 6 2 = 3,

67

z w = 4 = z = 4 + w = 4 3 = 1, 2y + z = 1 = y = 1 1 z = 1, 2 2

x + y + z + w = 3 = x = 3 y z w = 0. The (unique) solution is thus x 0 y 1 = . z 1 w 3

170 Example Solve the system of linear equations x 1 3 y 0 = 1 . z 1 4 w

1 0 0

1 1 2 1 0 1

Solution: The system is already in row-echelon form, and we see that x, y, z are leading variables while w is a free parameter. We put w = t. Using back substitution, and operating from the bottom up, we nd z w = 4 = z = 4 + w = 4 + t, 2y + z = 1 = y = 1 2 1 2 z= 1 2 2 5 2 1 2 t= 1 2 5 2 1 2 t, 9 2 3 2 t.

x + y + z + w = 3 = x = 3 y z w = 3 + The solution is thus

t4tt=

3 9 x 2 2 t y 5 1 t 2 2 , t R. = z 4 + t t w

68 171 Example Solve the system of linear equations x 1 y 3 = . 0 z 1 w

Chapter 3

1 0

1 1 2 1

Solution: We see that x, y are leading variables, while z, w are free parameters. We put z = s, w = t. Operating from the bottom up, we nd 2y + z = 1 = y = 1 2 1 2 z= 1 2 1 2 s, 3 2 s t.

x + y + z + w = 3 = x = 3 y z w = The solution is thus x 5 3 s t 2 2 y 1 1s 2 2 , (s, t) R2 . = s z t w

5 2

172 Example Find all the solutions of the system x + 2y + 2z = 0, y + 2z = 1, working in Z3 . Solution: The augmented matrix of the system is 1 0 2 1 2 2

The system is already in row-echelon form and x, y are leading variables while z is a free parameter. We nd y = 1 2z = 1 + 1z, and x = 2y 2z = 1 + 2z. Thus x 1 + 2z = y 1 + 1z , z Z3 . z z

0 . 1

Denitions Letting z = 0, 1, 2 successively, we nd the three solutions x 1 = , y 1 z 0 x 0 = , y 2 1 z x 2 = . y 0 z 2

69

and

Homework
Problem 3.1.1 Find all the solutions in Z3 of the system x + y + z + w = 0, 2y + w = 2. Problem 3.1.2 In Z7 , given that 1 2 3 2 3 1 1 3 4 = 1 2 2 0 2 0 4 0 4 , 2 Problem 3.1.5 This problem introduces Hill block ciphers, which are a way of encoding information with an encoding matrix A Mnn (Z26 ), where n is a strictly positive integer. Split a plaintext into blocks of n letters, creating a series of n 1 matrices Pk , and consider the numerical equivalent (A = 0, B = 1, C = 2, . . . , Z = 25) of each letter. The encoded message is the translation to letters of the n 1 matrices Ck = APk mod 26. For example, suppose you want to encode the message COMMUNISTS EAT OFFAL with the encoding matrix 0 A = 3 0 1 0 0 0 0 , 2

nd all solutions of the system

1x + 2y + 3z = 5; 2x + 3y + 1z = 6; 3x + 1y + 2z = 0. Problem 3.1.3 Solve in Z13 : x 2y + z = 5, 2x + 2y = 7, 5x 3y + 4z = 1.

a 3 3 matrix. First, split the plaintext into groups of three letters: COM MUN IST SEA TOF FAL. Form 31 matrices with each set of letters and nd their numerical equivalent, for example, C 2 P1 = O = 14 . 12 M

Problem 3.1.4 Find, with proof, a polynomial p(x) with real number coefcients and degree 3 such that p(1) = 10, p(0) = 1, p(1) = 2, p(2) = 23.

70
Find the product AP1 modulo 26, and translate into letters: 0 AP1 = 3 0 1 0 0 0 2 14 O 0 14 = 6 = G , Y 24 12 2 system x0 + x1 x0 + x2 x0 + x3 . . . x0 + x100 x0 + x1 + x2 + + x100 Problem 3.1.6 Find all solutions in Z103 , if any, to the Hints: 0 + 1 + 2 + + 99 = 4950, = = = . . . = = 0, 1, 2, . . . 99,

Chapter 3

hence COM is encoded into OGY. Your task is to complete the encoding of the message.

4949.

99 77 103 74 = 1.

3.2 Existence of Solutions

We now answer the question of deciding when a system of linear equations is solvable. 173 Lemma Let A Mmn (F) be in row-echelon form, and let X Mn1 (F) be a matrix of variables. The homogeneous system AX = 0m1 of m linear equations in n variables has (i) a unique solution if m = n, (ii) multiple solutions if m < n. Proof: If m = n then A is a square triangular matrix whose diagonal elements are different from 0F . As such, it is invertible by virtue of Theorem 162. Thus AX = 0n1 = X = A1 0n1 = 0n1 so there is only the unique solution X = 0n1 , called the trivial solution.

If m < n then there are n m free variables. Letting these variables run through the elements of the eld, we obtain multiple solutions. Thus if the eld has innitely many elements, we obtain innitely many solutions, and if the eld has k elements, we obtain knm solutions. Observe that in this case there is always a non-trivial solution. u 174 Theorem Let A Mmn (F), and let X Mn1 (F) be a matrix of variables. The homogeneous system AX = 0m1 of m linear equations in n variables always has a non-trivial solution if m < n. Proof: We can nd a matrix P GLm (F) such that B = PA is in row-echelon form. Now AX = 0m1 PAX = 0m1 BX = 0m1 .

That is, the systems AX = 0m1 and BX = 0m1 have the same set of solutions. But by Lemma 173 there is a non-trivial solution. u 175 Theorem (Kronecker-Capelli) Let A Mmn (F), Y Mm1 (F) be constant matrices and X Mn1 (F) be a matrix of variables. The matrix equation AX = Y is solvable if and only if rank (A) = rank ([A|Y]) .

Examples of Linear Systems Proof: Assume rst that AX = Y, x1 x 2 X = . . . . xn

71

Now assume that r = rank (A) = rank ([A|Y]). This means that adding an extra column to A does not change the rank, and hence, by a sequence column operations [A|Y] is equivalent to [A|0n1 ]. Observe that none of these operations is a permutation of the columns, since the rst n columns of [A|Y] and [A|0n1 ] are the same. This means that Y can be obtained from the columns Ci , 1 i n of A by means of transvections and dilatations. But then
n

By performing Cn+1 n xj Cj Cn+1 on [A|Y] = [A|AX] we obtain [A|0n1 ]. Thus rank ([A|Y]) = j=1 rank ([A|0n1 ]) = rank (A).

Let the columns of [A|X] be denoted by Ci , 1 i n. Observe that that [A|X] Mm(n+1) (F) and that the (n + 1)-th column of [A|X] is x1 a11 + x2 a12 + + xn a1n x a + x a + + x a n 2 22 n 2n 1 21 = Cn+1 = AX = xi C i . . i=1 . . x1 an1 + x2 an2 + + xn ann

Y =
i=1

xi C i .

The solutions sought is thus x1 x 2 X = . . . . xn

rank (C) P Mp (q) such that B = AP.

Problem 3.2.1 Let A Mnp (F), B Mnq (F) and put C = [A B] Mn(p+q) (F) Prove that rank (A) =

3.3 Examples of Linear Systems

176 Example Use row reduction to solve the system x x + + 2y + 3z + 4w = 2y + 4z + 7w = 4y + 6z + 8w = 8 12 16

2x +

72 Solution: Form the expanded matrix of coefcients and apply row operations to obtain 1 1 2 2 3 4 2 4 4 6 7 8 8 12 16
R3 2R1 R3 R2 R1 R2

Chapter 3

The matrix is now in row-echelon form. The variables x and z are the pivots, so w and y are free. Setting w = s, y = t we have z = 4 3s, x = 8 4w 3z 2y = 8 4s 3(4 3s) 2t = 4 + 5s 2t. Hence the solution is given by x 4 + 5s 2t y t = . 4 3s z w s

1 0 0

3 4

0 0

1 3 0 0

8 4 . 0

177 Example Find R such that the system x + y z = 1, 2x + 3y + z = 3, x + y + 3z = 2, posses (i) no solution, (ii) innitely many solutions, (iii) a unique solution. Solution: The augmented matrix of the system is 1 1 2 3 1

By performing R2 2R1 R2 and R3 R1 R3 we obtain 1 0 0 1 1 1 1

1 1 3 . 3 2

By performing R3 ( 1)R2 R3 on this last matrix we obtain 1 0 0 1 1 1 1 1 0 +2 .

1 + 2 1 . 1 4

( 2)( + 3) 2

Examples of Linear Systems If = 3, we obtain no solution. If = 2, there is an innity of solutions x 5t = y 1 4t , z t If = 2 and = 3, there is a unique solution x 1 1 = . y + 3 1 z +3 178 Example Solve the system 6 3 1 0 1 x 1 2 0 y = 0 , z 0 1 2

73

t R.

for (x, y, z) (Z7 )3 .

Solution: Performing operations on the augmented matrix we have 6 3 1 0 2 0 1 1 0 0 1 2 1 3 6 1 0 0 0 2 0 0 2 0 1 2 0 0 1 1 1 2 4 1 2 3

R1 R3

R3 6R1 R3 R2 3R1 R2

This gives 2z = 3 = z = 5, 2y = 1 4z = 2 = y = 1, x = 2 z = 4. The solution is thus (x, y, z) = (4, 1, 5).

74

Chapter 3

Homework
Problem 3.3.1 Find the general solution to the system 1 1 2 4 1 1 0 1 2 0 1 1 2 4 0 1 0 1 2 0 1 a 1 1 b 1 2 c = 0 4 d 0 0 f 1

or shew that there is no solution.

Problem 3.3.2 Find all solutions of the system 1 1 1 1 1 1 1 1 1 2 1 1 1 4 3 1 1 3 4 4 1 a 3 2 b 4 3 c = 7 , 4 d 6 f 5 9

if any. Problem 3.3.3 Study the system

x + 2my + z = 4m; 2mx + y + z = 2; x + y + 2mz = 2m2 , with real parameter m. You must determine, with proof, for which m this system has (i) no solution, (ii) exactly one solution, and (iii) innitely many solutions. Problem 3.3.4 Study the following system of linear equations with parameter a. (2a 1)x + ay (a + 1)z = 1, ax + y 2z = 1, 2x + (3 a)y + (2a 6)z = 1. You must determine for which a there is: (i) no solution, (ii) a unique solution, (iii) innitely many solutions. Problem 3.3.5 Determine the values of the parameter m for which the system x (1 + m)x 2x is solvable. + y y my + + + (1 m)z 2z 3z = = = m+2 0 m+2

Examples of Linear Systems

Problem 3.3.6 Determine the values of the parameter m for which the system x x x x is solvable. Problem 3.3.7 It is known that the system ay + bx = c; cx + az = b; bz + cy = a possesses a unique solution. What conditions must (a, b, c) R3 fulll in this case? Find this unique solution. + y y y y + + + z z z z + + + t t t t = = = = 4a 4b 4c 4d

75

Problem 3.3.8 For which values of the real parameter a does the following system have (i) no solutions, (ii) exactly one solution, (iii) innitely many solutions? (1 a)x ax 2x + + + (2a + 1)y ay (a + 1)y + (a 1)z + (2a + 2)z = = = a, 2a + 2, a2 2a + 9.

Problem 3.3.9 Find strictly positive real numbers x, y, z such that x3 y2 z6 x4 y5 z12 x2 y2 z5 = = = 1 2 3.

Problem 3.3.10 (Leningrad Mathematical Olympiad, 1987, Grade 5) The numbers 1, 2, . . . , 16 are arranged in a 4 4 matrix A as shewn below. We may add 1 to all the numbers of any row or subtract 1 from all numbers of any column. Using only the allowed operations, how can we obtain AT ? 1 5 A= 9 13 2 6 10 14 3 7 11 15 4 8 12 16

Problem 3.3.11 (International Mathematics Olympiad, 1963) Find all solutions x1 , x2 , x3 , x4 , x5 of the system x5 + x2 = yx1 ; x1 x2 x3 x4 where y is a parameter. + x3 + x4 + x5 + x1 = = = = yx2 ; yx3 ; yx4 ; yx5 ,

Chapter

Vector Spaces
4.1 Vector Spaces
179 Denition A vector space V, +, , F over a eld F, +, is a non-empty set V whose elements are called vectors, possessing two operations + (vector addition), and (scalar multiplication) which satisfy the following axioms. (, b, ) V 3 , (, ) F2 , a c VS1 Closure under vector addition : + V, a b V, a + = + a b b a ( + b ) + = + ( b + ) a c a c (4.1)

VS2 Closure under scalar multiplication

(4.2)

VS3 Commutativity

(4.3)

VS4 Associativity

(4.4)

VS5 Existence of an additive identity

VS6 Existence of additive inverses

0 V :+ 0 =+ 0 = a a a

(4.5)

VS7 Distributive Law

V : + () = () + = 0 a a a a a ( + b) = + b a a ( + ) = + a a a a a 1F = () = () a a 76

(4.6)

(4.7)

VS8 Distributive Law

(4.8)

VS9

(4.9)

VS10

(4.10)

Vector Spaces 180 Example If n is a positive integer, then Fn , +, , F is a vector space by dening (a1 , a2 , . . . , an ) + (b1 , b2 , . . . , bn ) = (a1 + b1 , a2 + b2 , . . . , an + bn ), (a1 , a2 , . . . , an ) = (a1 , a2 , . . . , an ).

77

In particular, Z2 , +, , Z2 is a vector space with only four elements and we have seen the two-dimensional 2 and tridimensional spaces R2 , +, , R and R3 , +, , R . 181 Example Mmn (F), +, , F is a vector space under matrix addition and scalar multiplication of matrices. 182 Example If F[x] = {a0 + a1 x + a2 x + + an xn : ai F, n N} denotes the set of polynomials with coefcients in a eld F, +, then F[x], +, , F is a vector space, under polynomial addition and scalar multiplication of a polynomial. 183 Example If Fn [x] = {a0 + a1 x + a2 x + + ak xk : ai F, n N, k n} denotes the set of polynomials with coefcients in a eld F, +, and degree at most n, then Fn [x], +, , F is a vector space, under polynomial addition and scalar multiplication of a polynomial. 184 Example Let k N and let Ck (R[a;b] ) denote the set of k-fold continuously differentiable real-valued functions dened on the interval [a; b]. Then Ck (R[a;b] ) is a vector space under addition of functions and multiplication of a function by a scalar. 185 Example Let p ]1; +[. Consider the set of sequences {an } , an C, n=0 lp = {an } : n=0
n=0

|an |p < + .

Then lp is a vector space by dening addition as termwise addition of sequences and scalar multiplication as termwise multiplication: {an } + {bn } = {(an + bn )} , n=0 n=0 n=0 {an } = {an } , C. n=0 n=0 All the axioms of a vector space follow trivially from the fact that we are adding complex numbers, except that we must prove that in lp there is closure under addition and scalar multiplication. Since p p < + = < + closure under scalar multiplication follows easily. To prove n=0 |an | n=0 |an | closure under addition, observe that if z C then |z| R+ and so by the Minkowski Inequality Theorem 405 we have
N n=0

|an + bn |p

1/p

N n=0 n=0

|an |p |an |

1/p

+ +

N n=0 n=0

|bn |p

1/p

(4.11) .

p 1/p

|bn |

p 1/p

This in turn implies that the series on the left in (4.11) converges, and so we may take the limit as N + obtaining
1/p

|an + bn |p

n=0

1/p

|an |p

1/p

|bn |p

(4.12)

n=0

n=0

Now (4.12) implies that the sum of two sequences in lp is also in lp , which demonstrates closure under addition.

78 186 Example The set

Chapter 4

with addition dened as (a + b 2 + c 3) + (a + b 2 + c 3) = (a + a ) + (b + b ) 2 + (c + c ) 3, and scalar multiplication dened as (a + b 2 + c 3) = (a) + (b) 2 + (c) 3, constitutes a vector space over Q. 187 Theorem In any vector space V, +, , F , F, Proof: Hence or proving the theorem. u 188 Theorem In any vector space V, +, , F , V, v Proof: We have We have 0 = 0.

V = {a + b 2 + c 3 : (a, b, c) Q3 }

0 = ( 0 + 0 ) = 0 + 0 . 0 0 = 0, 0 = 0,

0F = 0 . v

Therefore or proving the theorem. u

v v v v 0F = (0F + 0F ) = 0F + 0F. v v v 0F 0F = 0F, v 0 = 0F,

189 Theorem In any vector space V, +, , F , F, V, v = 0 v = = 0F

190 Theorem In any vector space V, +, , F , F,

Proof: Assume that = 0F . Then possesses a multiplicative inverse 1 such that 1 = 1F . Thus = 0 = 1 = 1 0 . v v By Theorem 188, 1 0 = 0 . Hence 1 = 0 . v v v v v Since by Axiom 4.9, we have 1 = 1 = , and so we conclude that = 0 . u
F

= . v 0

V, v

() = () = (). v v v

Vector Subspaces Proof: whence that is Similarly, whence that is proving the theorem. u We have

79

0F = ( + ()) = + (), v v v v () + 0F = (), v v v () = (). v v

0 = ( ) = + (), v v v v () + 0 = (), v v () = (), v v

Homework
Problem 4.1.1 Is R2 with vector addition and scalar multiplication dened as x1 y1 x1 + y1 , + = x2 + y2 y2 x2 x1 x1 = 0 x2 V is a vector space over F if vector addition is dened as a b = ab, (a, b) (R+ )2 and scalar multiplication is dened as a = a , (, a) (R, R+ ). Problem 4.1.4 Let C denote the complex numbers and R denote the real numbers. Is C a vector space over R under ordinary addition and multiplication? Is R a vector space over C? Problem 4.1.5 Construct a vector space with exactly 8 elements. Problem 4.1.6 Construct a vector space with exactly 9 elements.

a vector space?

Problem 4.1.2 Demonstrate that the commutativity axiom 4.3 is redundant. Problem 4.1.3 Let V = R+ =]0; +[, the positive real numbers and F = R, the real numbers. Demonstrate that

4.2 Vector Subspaces

191 Denition Let V, +, , F be a vector space. A non-empty subset U V which is also a vector space under the inherited operations of V is called a vector subspace of V. 192 Example Trivially, X1 = { 0 } and X2 = V are vector subspaces of V.

193 Theorem Let V, +, , F be a vector space. Then U V, F and (, b) U2 it is veried that a + U. a b

U = is a subspace of V if and only if

Proof: Observe that U inherits commutativity, associativity and the distributive laws from V. Thus a non-empty U V is a vector subspace of V if (i) U is closed under scalar multiplication, that is, if F and U, then U; (ii) U is closed under vector addition, that is, if v v , ) U2 , then + U. Observe that (i) gives the existence of inverses in U, for take (u v u v = 1F and so U = U. This coupled with (ii) gives the existence of the zero-vector, v v for 0 = v v U. Thus we need to prove that if a non-empty subset of V satises the property

80

Chapter 4 stated in the Theorem then it is closed under scalar multiplication and vector addition, and viceversa, if a non-empty subset of V is closed under scalar multiplication and vector addition, then it satises the property stated in the Theorem. But this is trivial. u

194 Example Shew that X = {A Mnn (F) : tr (A) = 0F } is a subspace of Mnn (F). Solution: Take A, B X, R. Then tr (A + B) = tr (A) + tr (B) = 0F + (0F ) = 0F . Hence A + B X, meaning that X is a subspace of Mnn (F). 195 Example Let U Mnn (F) be an arbitrary but xed. Shew that CU = {A Mnn (F) : AU = UA} is a subspace of Mnn (F). Solution: Take (A, B) (CU )2 . Then AU = UA and BU = UB. Now (A + B)U = AU + BU = UA + UB = U(A + B), meaning that A + B CU . Hence CU is a subspace of Mnn (F). CU is called the commutator of U. 196 Theorem Let X V, Y V be vector subspaces of a vector space V, +, , F . Then their intersection X Y is also a vector subspace of V. Proof: Let F and (, b) (X Y)2 . Then clearly (, b) X and (, b) Y. Since X is a a a a vector subspace, + b X and since Y is a vector subspace, + b Y. Thus a a and so X Y is a vector subspace of V by virtue of Theorem 193. u R2 , +, , R are the set containing the zero-vector, any line through the origin, and R itself. The only vector subspaces of R3 , +, , R are the set containing the zero-vector, any line through the origin, any plane containing the origin and R3 itself.
2

+ X Y a b

 We we will soon see that the only vector subspaces of

Homework
Problem 4.2.1 Prove that a b 4 X = R : a b 3d = 0 c d is a vector subspace of R4 . Problem 4.2.2 Prove that a 2a 3b X = 5b : a, b R a + 2b a is a vector subspace of R5 .

Linear Independence
Problem 4.2.3 Let A Mmn (F) be a xed matrix. Demonstrate that S = {X Mn1 (F) : AX = 0m1 } is a subspace of Mn1 (F). Problem 4.2.4 Prove that the set X Mnn (F) of upper triangular matrices is a subspace of Mnn (F). Problem 4.2.5 Prove that the set X Mnn (F) of symmetric matrices is a subspace of Mnn (F). Problem 4.2.6 Prove that the set X Mnn (F) of skewsymmetric matrices is a subspace of Mnn (F). Problem 4.2.7 Prove that the following subsets are not subspaces of the given vector space. Here you must say which of the axioms for a vector space fail. a 2 2 b : a, b R, a + b = 1 R3 0 a 2 : a, b R , ab = 0 R3 b 0 a b : (a, b) R2 , a + b2 = 0 0

81

0 M22 (R)

Problem 4.2.8 Let V, +, , F be a vector space, and let U1 V and U2 V be vector subspaces. Prove that if U1 U2 is a vector subspace of V, then either U1 U2 or U2 U1 . Problem 4.2.9 Let V a vector space over a eld F. If F is innite, show that V is not the set-theoretic union of a nite number of proper subspaces. Problem 4.2.10 Give an example of a nite vector space V over a nite eld F such that V = V1 V2 V3 , where the Vk are proper subspaces.

4.3 Linear Independence

197 Denition Let (1 , 2 , , n ) Fn . Then the vectorial sum
n

j=1

is said to be a linear combination of the vectors i V, 1 i n. a a b M22 (R) can be written as a linear combination of the matrices 198 Example Any matrix c d 1 0 0 1 0 0 0 0 , , , , 0 0 0 0 1 0 0 1 a c b 1 0 0 = a +b d 0 0 0 1 0 + c 0 1 0 0 + d 0 0 0 . 1

jj a

for

199 Example Any polynomial of degree at most 2, say a + bx + cx2 R2 [x] can be written as a linear combination of 1, x 1, and x2 x + 2, for a + bx + cx2 = (a c)(1) + (b + c)(x 1) + c(x2 x + 2).

82 Generalising the notion of two parallel vectors, we have 200 Denition The vectors i V, 1 i n, are linearly dependent or tied if a
n

Chapter 4

(1 , 2 , , n ) Fn \ {0} such that

j=1

that is, if there is a non-trivial linear combination of them adding to the zero vector. 201 Denition The vectors i V, 1 i n, are linearly independent or free if they are not linearly a dependent. That is, if (1 , 2 , , n ) Fn then
n j=1

jj = 0 , a

jj = 0 = 1 = 2 = = n = 0F . a

 A family of vectors is linearly independent if and only if the only linear combination of them
giving the zero-vector is the trivial linear combination. 202 Example 1 4 7 , , 2 5 8 3 6 9

is a tied family of vectors in R3 , since

1 4 7 0 (1) 2 + (2) 5 + (1) 8 = 0 . 3 6 9 0

203 Example Let , be linearly independent vectors in some vector space over a eld F with characu v teristic different from 2. Shew that the two new vectors = and = + are also linearly x u v y u v independent. Solution: Assume that a( ) + b( + ) = 0 . Then u v u v

204 Theorem Let A Mmn (F). Then the columns of A are linearly independent if and only the only solution to the system AX = 0m is the trivial solution. Proof: Let A1 , . . . , An be the columns of A. Since x1 A1 + x2 A2 + + xn An = AX, the result follows. u

Since , are linearly independent, the above coefcients must be 0, that is, a + b = 0F and u v a b = 0F . But this gives 2a = 2b = 0F , which implies a = b = 0F , if the characteristic of the eld is not 2. This proves the linear independence of and + . u v u v

(a + b) + (a b) = 0 . u v

Linear Independence 205 Theorem Any family

83

containing the zero-vector is linearly dependent. Proof:

{ 0 , 1 , 2 , . . . , k } u u u

This follows at once by observing that u u u 1F 0 + 0F1 + 0F2 + . . . , +0Fk = 0

is a non-trivial linear combination of these vectors equalling the zero-vector. u

Homework
Problem 4.3.1 Shew that 1 1 1 , , 0 1 1 0 1 0 Problem 4.3.5 Let {1 , 2 , 3 , 4 } be a linearly indev v v v pendent family of vectors. Prove that the family { + , + , + , + } v v v v v v v v
1 2 2 3 3 4 4 1

is not linearly independent.

forms a free family of vectors in R3 .

is a linearly independent set of vectors in R and shew

Problem 4.3.2 Prove that the set 1 1 1 1 1 1 1 1 , , , 1 1 1 0 1 1 1 1

Problem 4.3.6 Let {1 , 2 , 3 } be linearly independent v v v 5 vectors in R . Are the vectors b 1 = 31 + 22 + 43 , v v v + 4 + 2 , b2 = v 1 v2 v3 + 4 + 3 , b3 = 9 v 1 v2 v3 b = + 2 + 5 , v v v

4 1 2 3

linearly independent? Prove or disprove!

Problem 4.3.7 Is the family {1, over Q? Problem 4.3.8 Is the family {1, over R?

2} linearly independent 2} linearly independent

1 2 that X = can be written as a linear combination of 1 1 these vectors. u v u v Problem 4.3.3 Let (, ) (Rn )2 . Prove that || = if and only if and are linearly dependent. v u v u Problem 4.3.4 Prove that 1 0 1 0 0 , , 0 1 1 0 1 1 0

Problem 4.3.9 Consider the vector space V = {a + b 2 + c 3 : (a, b, c) Q3 }. 1. Shew that {1, 2, 3} are linearly independent over Q. 2. Express 1 2 + 1 2 12 2 as a linear combination of {1, 2, 3}.

1 0 , 1 0

Problem 4.3.10 Let f, g, h belong to C (RR ) (the space of innitely continuously differentiable real-valued functions dened on the real line) and be given by f(x) = ex , g(x) = e2x , h(x) = e3x . 1 as 1 Shew that f, g, h are linearly independent over R. Problem 4.3.11 Let f, g, h belong to C (RR ) be given by f(x) = cos2 x, g(x) = sin2 x, h(x) = cos 2x. Shew that f, g, h are linearly dependent over R.

1 is a linearly independent family over R. Write 1 a linear combination of these matrices.

84

Chapter 4

4.4 Spanning Sets

206 Denition A family {1 , 2 , . . . , k , . . . , } V is said to span or generate V if every V can be u u u v s. written as a linear combination of the u j { 1 , 2 , . . . , k , . . . , } V w, u u u

207 Theorem If {1 , 2 , . . . , k , . . . , } V spans V, then any superset u u u also spans V. Proof:

This follows at once from

l

i=1

w u ii = 0F +

i=1

u ii .

u 208 Example The family of vectors 1 0 0 i = 0 , j = 1 , k = 0 0 0 1

a spans R3 since given b R3 we may write c a b = a + b + c. i j k c

209 Example Prove that the family of vectors 1 1 1 t 1 = 0 , t 2 = 1 , t 3 = 1 1 0 0

spans R3 .

Spanning Sets Solution: This follows from the identity

85

a 1 1 1 = (a b) + (b c) + c = (a b)1 + (b c)2 + c3 . t t t b 0 1 1 c 0 0 1

210 Example Since a b 1 = a c d 0 1 , 0 0 1 0 0 1 0 + b +c 0 0 0 1 0 0 0 +d 0 0 1

1 0 0 , the set of matrices 0 0 0 211 Example The set

0 , 0

0 0 is a spanning set for M22 (R). 0 1

{1, x, x2 , x3 , . . . , xn , . . .} spans R[x], the set of polynomials with real coefcients and indeterminate x. 212 Denition The span of a family of vectors {1 , 2 , . . . , k , . . . , } is the set of all nite linear combinau u u , , . . . , , . . . , } by tions obtained from the ui s. We denote the span of { u 1 u 2 uk span , , . . . , , . . . , . u u u
1 2 k

is a vector subspace of V. Proof:

213 Theorem Let V, +, , F be a vector space. Then u u u span 1 , 2 , . . . , k , . . . , V Let F and let
l l

u u u be in span 1 , 2 , . . . , k , . . . , (some of the coefcients might be 0F ). Then u u u and so span 1 , 2 , . . . , k , . . . , is a subspace of V. u + = x y

l k=1

= x

k=1

akk , = u y

k=1

bkk , u

u u u (ak + bk )k span 1 , 2 , . . . , k , . . . , , u V is the smallest vector subspace of V (in the sense of set

u u u 214 Corollary span 1 , 2 , . . . , k , . . . , inclusion) containing the set Proof:

then it contains every nite linear combination of them, and hence, it contains span 1 , 2 , . . . , k , . . . , . u u u u

If W V is a vector subspace of V containing the set {1 , 2 , . . . , k , . . . , } u u u

{1 , 2 , . . . , k , . . . , }. u u u

86

Chapter 4

1 0 0 , 215 Example If A M22 (R), A span 0 0 0 1 a 0

0 0 , 1 1

i.e., this family spans the set of all symmetric 2 2 matrices over R.

0 0 0 0 +c + b 1 0 1 0

1 a = c 0

1 then A has the form 0 c , b

216 Theorem Let V be a vector space over a eld F and let (, V 2 , F \ {0F }. Then v w) , = span , . span v w v w Proof: The equality a + b = a + (b1 )( v w v w),

proves the statement. u

217 Theorem Let V be a vector space over a eld F and let (, V 2 , F. Then v w) , = span + . span v w w, v w Proof: This follows from the equality a + b = a( + + (b a) v w v w) w.

Homework
Problem 4.4.1 Let R3 [x] denote the set of polynomials with degree at most 3 and real coefcients. Prove that the set {1, 1 + x, (1 + x)2 , (1 + x)3 } spans R3 [x].

1 0 1 Problem 4.4.2 Shew that 1 span 0 , 1 . 1 1 1 1 Problem 4.4.3 What is span 0 Problem 4.4.4 Prove that 1 span 0 0 1 , 0 1 0 0 , 1 1 1 0 , 1 0 1 = M22 (R). 0 0 0 , 0 0 0 0 , 1 1 1 ? 0

Bases
Problem 4.4.5 For the vectors in R3 , 1 1 1 3 = , = , = , = , a 2 b 3 c 1 d 8 5 0 2 1 span , b = span , d . a c

87

prove that

4.5 Bases
218 Denition A family {1 , 2 , . . . , k , . . .} V is said to be a basis of V if (i) they are linearly indepenu u u dent, (ii) they span V. 219 Example The family 0F . . . 0F = , e i 1F 0F . . . 0F

where there is a 1F on the i-th slot and 0F s on the other n 1 positions, is a basis for Fn . 220 Theorem Let V, +, , F be a vector space and let U = {1 , 2 , . . . , k , . . .} V u u u

be a family of linearly independent vectors in V which is maximal in the sense that if U is any other family of vectors of V properly containing U then U is a dependent family. Then U forms a basis for V. Proof: Since U is a linearly independent family, we need only to prove that it spans V. Take V. If U then there is nothing to prove, so assume that V \ U. Consider the set v v v }. This set properly contains U, and so, by assumption, it forms a dependent family. U = U{v There exists scalars 0 , 1 , . . . , n such that 0 + 11 + + nn = 0 . v u u Now, = 0 , otherwise the would be linearly dependent. Hence 1 exists and we have u
0 F i 0

and so the i span V. u u

= 1 ( + + ), v 1 u1 n un 0

From Theorem 220 it follows that to shew that a vector space has a basis it is enough to shew that it has a maximal linearly independent set of vectors. Such a proof requires something called Zrns Lemma, and it is beyond our scope. We dodge the whole business by taking as an axiom that every vector space possesses a basis.

88

Chapter 4

221 Theorem (Steinitz Replacement Theorem) Let V, +, , F be a vector space and let U = {1 , 2 , . . .} u u , , . . . , } be an independent family of vectors in span (U). Then there exist k of the V. Let W = {w1 w2 wk s, say { , , . . . , } which may be replaced by the s in such a way that ui u1 u2 uk wi span 1 , 2 , . . . , k , k+1 , . . . = span (U) . w w w u

Proof:

We prove this by induction on k. If k = 1, then = + + + w1 1 u1 2 u2 n un

for some n and scalars i . There is an i = 0F , since otherwise 1 = 0 contrary to the w are linearly independent. After reordering, we may assume that = 0 . assumption that the wi 1 F Hence = 1 ( ( + + )), u1 w1 n un 2 u2 1 and so span , , . . . , and u w u
1 1 2

u u w u span 1 , 2 , . . . , = span 1 , 2 , . . . , .

Assume now that the theorem is true for any set of fewer than k independent vectors. We may thus assume that that {1 , . . .} has more than k 1 vectors and that u Since k U we have w span 1 , 2 , . . . , k1 , k , , . . . = span (U) . w w w u

If all the i = 0F , then the {1 , 2 , . . . , k } would be linearly dependent, contrary to assumption. w w w Thus there is a i = 0F , and after reordering, we may assume that k = 0F . We have therefore But this means that = 1 ( ( + + + uk wk k1 wk1 + k+1 u k+1 + m u m )). 2 w2 1 w1 k span 1 , 2 , . . . , k , k+1 , . . . = span (U) . w w w u

= + + + wk 1 w1 2 w2 k1 wk1 + k u k + k+1 u k+1 + m u m .

This nishes the proof. u

222 Corollary Let {1 , 2 , . . . , n } be an independent family of vectors with V = span 1 , 2 , . . . , n . w w w w w w If we also have V = span 1 , 2 , . . . , , then u u u 1. n , 2. n = if and only if the {1 , 2 , . . . , } are a linearly independent family. y y y Proof: 1. In the Steinitz Replacement Theorem 221 replace the rst n i s by the i s and n u w follows. 2. If {1 , 2 , . . . , } are a linearly independent family, then we may interchange the rle of u u u and obtaining n. Conversely, if = n and if the are dependent, we could the wi ui ui as a linear combination of the remaining 1 vectors, and thus we would express some u
i

3. Any basis for V has exactly n elements.

3. This follows from the denition of what a basis is and from (2) of this corollary.

have shewn that some 1 vectors span V. From (1) in this corollary we would conclude that n 1, contradicting n = .

Bases u

89

223 Denition The dimension of a vector space V, +, , F is the number of elements of any of its bases, and we denote it by dim V. 224 Theorem Any linearly independent family of vectors {1 , 2 , . . . , k } x x x so that this latter family become a basis for V. Proof: u

in a vector space V can be completed into a family {1 , 2 , . . . , k , k+1 , k+2 , . . .} x x x y y Take any basis {1 , . . . , k , k+1 , . . . , } and use Steinitz Replacement Theorem 221. u u u

225 Corollary If U V is a vector subspace of a nite dimensional vector space V then dim U dim V. Proof: Since any basis of U can be extended to a basis of V, it follows that the number of elements of the basis of U is at most as large as that for V. u 226 Example Find a basis and the dimension of the space generated by the set of symmetric matrices in Mnn (R). Solution: Let Eij Mnn (R) be the n n matrix with a 1 on the ij-th position and 0s n(n 1) everywhere else. For 1 i < j n, consider the n = matrices Aij = Eij + Eji . The 2 2 Aij have a 1 on the ij-th and ji-th position and 0s everywhere else. They, together with the n matrices Eii , 1 i n constitute a basis for the space of symmetric matrices. The dimension of this space is thus n(n 1) n(n + 1) +n= . 2 2 227 Theorem Let {1 , . . . , n } be vectors in Rn . Then the s form a basis if and only if the n n matrix u u u s as the columns of A is invertible. A formed by taking the u Since we have the right number of vectors, it is enough to prove that the s are linearly u x1 x 2 independent. But if X = , then . . . xn Proof: x11 + + xnn = AX. u u

If A is invertible, then AX = 0n = X = A1 0n = 0n , meaning that x1 = x2 = xn = 0, so the s are linearly independent. u

Conversely, assume that the s are linearly independent. Then the equation AX = 0n has u a unique solution. Let r = rank (A) and let (P, Q) (GLn (R))2 be matrices such that A = P1 Dn,n,r Q1 , where Dn,n,r is the Hermite normal form of A. Thus

90

Chapter 4

y1 y 2 1 Put Q X = . Then . . . yn

AX = 0n = P1 Dn,n,r Q1 X = 0n = Dn,n,r Q1 X = 0n .

where j is the n-dimensional column vector with a 1 on the j-th slot and 0s everywhere else. e If r < n then yr+1 , . . . , yn can be taken arbitrarily and so there would not be a unique solution, a contradiction. Hence r = n and A is invertible. u

Dn,n,r Q1 X = 0n = y11 + + yrr = 0n , e e

Homework
Problem 4.5.1 In problem 4.2.2 we saw that a 2a 3b X = 5b : a, b R a + 2b a

is a vector subspace of R5 . Find a basis for this subspace.

Problem 4.5.2 Let {1 , 2 , 3 , 4 , 5 } be a basis for a vector space V over a eld F. Prove that v v v v v is also a basis for V. {1 + 2 , 2 + 3 , 3 + 4 , 4 + 5 , 5 + 1 } v v v v v v v v v v

Problem 4.5.3 Find a basis for the solution space of the system of n + 1 linear equations of 2n unknowns x1 + x2 + + xn = 0, x2 + x3 + + xn+1 = 0, ... ... ... xn+1 + xn+2 + + x2n = 0. Problem 4.5.4 Prove that the set V of skew-symmetric n n matrices is a subspace of Mnn (R) and nd its dimension. Exhibit a basis for V. Problem 4.5.5 Prove that the set is a vector subspace of R4 . Find its dimension and a basis for X. X = {(a, b, c, d)|b + 2c = 0} R4

nd a basis for this space.

Problem 4.5.6 Prove that the dimension of the vector subspace of lower triangular n n matrices is

n(n + 1) and 2

Coordinates
Problem 4.5.7 Find a basis and the dimension of

91

1 1 = X = span v1 , 1 1

1 1 = , v2 1 0

2 2 = . v3 2 1

Problem 4.5.8 Find a basis and the dimension of

1 1 X = span = , v1 1 1

1 1 = , v2 1 0

2 2 = . v3 2 2

Problem 4.5.9 Find a basis and the dimension of 1 X = span = v1 0 0 , 1

1 = v2 2

0 0

0 = v3 2

1 1 , = v4 0 0

1 . 0

Problem 4.5.10 Prove that the set a V = 0 0

b d 0

is a vector space of M33 (R) and nd a basis for it and its dimension.

c f M33 (R) : a + b + c = 0, g

a+d+g=0

4.6 Coordinates
228 Theorem Let {1 , 2 , . . . , n } be a basis for a vector space V. Then any V has a unique v v v v representation = a + a + + a . v nvn 2 v2 1 v1 Proof: Let = b + b + + b v 1v1 2 v2 n vn

be another representation of . Then v 0 = (a1 b1 )1 + (a2 b2 )2 + + (an bn )n . v v v , , . . . , } forms a basis for V, they are a linearly independent family. Thus we Since { v 1 v 2 vn must have a1 b1 = a2 b2 = = an bn = 0F , that is a1 = b1 ; a2 = b2 ; ; an = bn ,

proving uniqueness. u

92

Chapter 4

229 Denition An ordered basis {1 , 2 , . . . , n } of a vector space V is a basis where the order of the k v v v v , , . . . , } of a vector space V, Theorem 228 ensures that has been xed. Given an ordered basis { v 1 v 2 vn there are unique (a1 , a2 , . . . , an ) Fn such that The ak s are called the coordinates of the vector . v = a + a + + a . v nvn 2 v2 1 v1

1 3 230 Example The standard ordered basis for R is S = { i , j , k }. The vector 2 R3 for example, has 3 coordinates (1, 2, 3)S . If the order of the basis were changed to the ordered basis S1 = { i , k , j }, then 1 R3 would have coordinates (1, 3, 2)S . 1 2 3 we are using the standard basis.

 Usually, when we give a coordinate representation for a vector R , we assume that v

n

1 231 Example Consider the vector 2 R3 (given in standard representation). Since 3 1 1 1 1 = 1 1 + 3 , 2 0 1 1 3 0 0 1 1 has coordinates (1, 1, 3)B . We write 1 2 3 .

1 1 1 under the ordered basis B1 = 0 , 1 , 1 , 0 0 1

1 1 2 = 1 3 3

B1

Coordinates 232 Example The vectors of

93

are non-parallel, and so form a basis for R2 . So do the vectors 2 1 B2 = , . 1 1 3 Find the coordinates of 4 in the base B2 .

1 1 B1 = , 1 2

B1

Solution: We are seeking x, y such that

1 1 2 1 1 3 + 4 = x + y = 1 2 1 1 1 x y
B2

1 3 2 = 2 4 1 1 1 1 3 4 2

1 x 1 y

B2

Thus 2 = 1 =
1 3 1 3

1 1
1 3

Let us check by expressing both vectors in the standard basis of R2 : 3 4

B1

6 = 5

2 1 3 3 = 1 4 3 1

1 1 3 2 1 2 3 4

B2

1 1 7 = 3 + 4 = , 1 2 11

6 5

B2

2 1 7 = 6 5 = . 1 1 11

94

Chapter 4

In general let us consider bases B1 , B2 for the same vector space V. We want to convert XB1 to YB2 . We let A be the matrix formed with the column vectors of B1 in the given order an B be the matrix formed with the column vectors of B2 in the given order. Both A and B are invertible matrices since the Bs are bases, in view of Theorem 227. Then we must have AXB1 = BYB2 = YB2 = B1 AXB1 . Also, XB1 = A1 BYB2 . This prompts the following denition. 233 Denition Let B1 = {1 , 2 , . . . , n } and B2 = {1 , 2 , . . . , n } be two ordered bases for a vector u u u v v v s as its columns and let B M space V. Let A Mnn (F) be the matrix having the u nn (F) be the matrix having the s as its columns. The matrix P = B1 A is called the transition matrix from B1 to B2 v and the matrix P1 = A1 B is called the transition matrix from B2 to B1 . 234 Example Consider the bases of R3 1 1 1 B1 = 1 , 1 , 0 , 1 0 0 1 1 2 , , . B2 = 1 1 0 1 0 0

1 coordinates of 2 3

Find the transition matrix from B1 to B2 and also the transition matrix from B2 to B1 . Also nd the in terms of B2 .

B1

Solution: Let 1 A = 1 1 1 1 0 1 1 , B = 0 1 0 1 1 1 0 2 0 . 0

Coordinates The transition matrix from B1 to B2 is P = B1 A 1 = 1 1 0 = 0

1 2

95

1 1 0 0 1
1 2

2 0 0

The transition matrix from B2 to B1 is P1 1 = 2 2

1 = 2 2

1 1 1 1 1 1 0 0 .
1 2

1 1 1 1 1 0

1 1 1 0 0 0 1 0 0

0 1 1

0 1 1

Now,

0 0
1 2

1 0 = 2 1 0 2 1 = 4
11 2

0 0 . 2

YB2

As a check, observe that in the standard basis for R 1 2 3

1 = 2 2

0 1 1

0 1 0 2 1 3 2

B1

B2

B1

1 1 1 6 = 1 1 + 2 1 + 3 0 = 3 , 1 0 0 1

1 4
11 2

B2

2 6 1 1 11 = . = 1 1 4 1 + 3 0 2 1 0 1 0

96

Chapter 4

Homework
Problem 4.6.1 1. Prove that the following vectors are linearly independent in R4 Problem 4.6.2 Consider the matrix a 0 A(a) = 1 1 1 1 1 0 1 0 a 1

1 1 1 1 1 1 1 1 = , = , = , = . a1 a2 a3 a4 1 1 1 1 1 1 1 1 1 2 2. Find the coordinates of under the ordered ba 1 1 , , , ). sis ( a a a a

1 2 3 4

Determine all a for which A(a) is not invertible. Find the inverse of A(a) when A(a) is invertible. Find the transition matrix from the basis 1 1 1 1 1 1 1 0 B1 = , , , 1 1 0 0 0 0 0 1

1 1 . 1 1

1 2 3. Find the coordinates of under the ordered ba 1 1 , , , ). sis ( a a a a

1 3 2 4

to the basis

a 1 1 1 0 1 0 1 B2 = , , , . 1 0 a 1 1 1 1 1

Chapter

5
V L: W ,

Linear Transformations
5.1 Linear Transformations
235 Denition Let V, +, , F and W, +, , F be vector spaces over the same eld F. A linear transformation or homomorphism

is a function which is Linear: L( + b) = L() + L( b), a a

L() a a

Homogeneous: L() = L(), for F. a a

 It is clear that the above two conditions can be summarised conveniently into
L( + b ) = L() + L( b). a a Mnn (R) A R

236 Example Let

L:

Then L is linear, for if (A, B) Mnn (R), then

tr (A)

L(A + B) = tr (A + B) = tr (A) + tr (B) = L(A) + L(B). 237 Example Let L: Mnn (R) A Then L is linear, for if (A, B) Mnn (R), then Mnn (R) A
T

L(A + B) = (A + B)T = AT + BT = L(A) + L(B).

97

98 238 Example For a point (x, y) R2 , its reexion about the y-axis is (x, y). Prove that R: R2 R2

Chapter 5

is linear.

(x, y) (x, y)

Solution: Let (x1 , y1 ) R2 , (x2 , y2 ) R2 , and R. Then R((x1 , y1 ) + (x2 , y2 )) = = = = whence R is linear. 239 Example Let L : R2 R4 be a linear transformation with 1 1 1 1 L = ; L = 1 1 2 3 5 Find L . 3 Solution: Since 1 1 5 = 4 , 1 1 3 1 2 6 1 0 4 5 1 1 = 4L L = 4 = . L 3 1 1 2 2 6 3 3 9 R((x1 + x2 , y1 + y2 )) ((x1 + x2 ), y1 + y2 ) (x1 , y1 ) + (x2 , y2 ) R((x1 , y1 )) + R((x2 , y2 )),

2 0 . 2 3

we have

240 Theorem Let V, +, , F and W, +, , F be vector spaces over the same eld F, and let L : V W be a linear transformation. Then

Kernel and Image of a Linear Transformation L( 0 V ) = 0 W . Proof: hence Since we obtain the rst result. Now

99

V, L() = L(). x x x We have L( 0 V ) = L( 0 V + 0 V ) = L( 0 V ) + L( 0 V ), L( 0 V ) L( 0 V ) = L( 0 V ). L( 0 V ) L( 0 V ) = 0 W ,

from where the second result follows. u

0 W = L( 0 V ) = L( + ()) = L() + L(), x x x x

Homework
Problem 5.1.1 Consider L : R3 R3 , x x y z L y = x + y + z . z z Prove that L is linear. that Mnn (R) L: H is a linear transformation. Mnn (R) A1 HA1

Problem 5.1.2 Let A GLn (R) be a xed matrix. Prove

Problem 5.1.3 Let V be a vector space and let S V. The set S is said to be convex if [0; 1], x, y S, (1 )x + y S, that is, for any two points in S, the straight line joining them also belongs to S. Let T : V W be a linear transformation from the vector space V to the vector space W. Prove that T maps convex sets into convex sets.

5.2 Kernel and Image of a Linear Transformation

241 Denition Let V, +, , F and W, +, , F be vector spaces over the same eld F, and V T : W T () v

be a linear transformation. The kernel of T is the set

The image of T is the set

ker (T ) = { V : T () = 0 W }. v v

 Since T ( 0

V)

= 0 W by Theorem 240, we have 0 V ker (T ) and 0 W Im (T ).

Im (T ) = { w : V such that T () = = T (V). w v v w}

100 242 Theorem Let V, +, , F and W, +, , F be vector spaces over the same eld F, and V T : W T () v

Chapter 5

be a linear transformation. Then ker (T ) is a vector subspace of V and Im (T ) is a vector subspace of W. Proof: Let (1 , 2 ) (ker (T ))2 and F. Then T (1 ) = T (2 ) = 0 V . We must prove that v v v v + ker (T ), that is, that T ( + ) = 0 . But v1 v2 v1 v2 W proving that ker (T ) is a subspace of V. T (1 + 2 ) = T (1 ) + T (2 ) = 0 V + 0 V = 0 V v v v v

Now, let (1 , 2 ) (Im (T ))2 and F. Then (1 , 2 ) V 2 such that T (1 ) = 1 and w w v v v w ) = . We must prove that + Im (T ), that is, that such that T () = T( v 2 w2 w1 w2 v v + . But w1 w2 + = T ( ) + T ( ) = T ( + ), w1 w2 v1 v2 v1 v2 = + . This proves that Im (T ) is a subspace of W. and so we may take v v v
1 2

243 Theorem Let V, +, , F and W, +, , F be vector spaces over the same eld F, and V T : W T () v

be a linear transformation. Then T is injective if and only if ker (T ) = 0 V . Proof: T () = 0 W . x

Assume that T is injective. Then there is a unique V mapping to 0 W : x

By Theorem 240, T ( 0 V ) = 0 W , i.e., a linear transformation takes the zero vector of one space to the zero vector of the target space, and so we must have = 0 V . x Conversely, assume that ker (T ) = { 0 V }, and that T () = T (). We must prove that = . But x y x y T () = T () x y = = = = = as we wanted to shew. u T ( ) = 0 W x y T () T () = 0 W x y

= , x y

= x y 0V

( ) ker (T ) x y

Kernel and Image of a Linear Transformation

101

244 Theorem (Dimension Theorem) Let V, +, , F and W, +, , F be vector spaces of nite dimension over the same eld F, and V T : be a linear transformation. Then W T () v

dim ker (T ) + dim Im (T ) = dim V.

Proof: Let {1 , 2 , . . . , k } be a basis for ker (T ). By virtue of Theorem 224, we may extend v v v , , . . . , , this to a basis A = { v 1 v 2 v k v k+1 , k+2 , . . . , n } of V. Here n = dim V. We will now v v )} is a basis for Im (T ). We prove that (i) B spans shew that B = {T ( v k+1 ), T ( v k+2 ), . . . , T ( v n Im (T ), and (ii) B is a linearly independent family. Let Im (T ). Then V such that T () = Now since A is a basis for V we can write w v v w. = v
n i=1

Hence = T () = w v

ii . v

i=1

To prove the linear independence of the B assume that 0W =

n

since T (i ) = 0 V for 1 i k. Thus B spans Im (T ). v i T (i ) = T v

i T (i ) = v

i=k+1

i T (i ), v

i=k+1

i=k+1

This means that 0F . u

n i=k+1

ii ker (T ), which is impossible unless k+1 = k+2 = = n = v

ii v

245 Corollary If dim V = dim W < +, then T is injective if and only if it is surjective. Proof: Simply observe that if T is injective then dim ker (T ) = 0, and if T is surjective Im (T ) = T (V) = W and Im (T ) = dim W. u 246 Example Let L: M22 (R) A M22 (R) A A
T

Observe that L is linear. Determine ker (L) and Im (L) .

a Solution: Put A = c

0 0 a = L(A) = 0 0 b

b and assume L(A) = 02 . Then d

c a b 0 = (c b) d c d 1

1 . 0

102 This means that c = b. Thus a b : (a, b, d) R3 , ker (L) = b d 0 Im (L) = k k :kR . 0

Chapter 5

This means that dim ker (L) = 3, and so dim Im (L) = 4 3 = 1. 247 Example Consider the linear transformation L : M22 (R) R3 [X] given by a b = (a + b)X2 + (a b)X3 . L c d

Determine ker (L) and Im (L). Solution: We have

a b = (a + b)X2 + (a b)X3 = a + b = 0, a b = 0, = a = b = 0. 0 = L c d 0 0 2 : (c, d) R ker (L) = . c d

Thus

Thus dim ker (L) = 2 and hence dim Im (L) = 2. Now

(a + b)X2 + (a b)X3 = a(X2 + X3 ) + b(X2 X3 ). Clearly X2 + X3 , and X2 X3 are linearly independent and span Im (L). Thus Im (L) = span X2 + X3 , X2 X3 .

Homework
Problem 5.2.1 In problem 5.1.1 we saw that L : R3 R3 , x x y z L y = x + y + z z z Problem 5.2.2 Consider the function L : R4 R2 given by x y x + y . L = z xy w

is linear. Determine ker (L) and Im (L).

Kernel and Image of a Linear Transformation

1. Prove that L is linear. 2. Determine ker (L). 3. Determine Im (L). Problem 5.2.3 Let R3 a R4 L() a Problem 5.2.6 It is easy to see that L : R3 R2 , x x y z L y = y 2z z is linear. Determine ker (L) and Im (L). Problem 5.2.7 Let 1 0 1 L 0 = . 1 1 0 M22 (R) L: A R tr (A)

103

L:

satisfy 1 1 0 L 0 = ; 1 0 0

Determine ker (L) and Im (L).

2 1 1 L 1 = ; 0 0 0

Determine ker (L) and Im (L). Problem 5.2.8

1. Demonstrate that M22 (R) L: A M22 (R) AT + A

is linear. Determine ker (L) and Im (L).

Problem 5.2.4 It is easy to see that L : R2 R3 , x + 2y x L = x + 2y y 0

is a linear transformation.

2. Find a basis for ker (L) and nd dim ker (L) 3. Find a basis for Im (L) and nd dim Im (L). Problem 5.2.9 Let V be an n-dimensional vector space, where the characteristic of the underlying eld is different from 2. A linear transformation T : V V is idempotent if T 2 = T . Prove that if T is idempotent, then I T is idempotent (I is the identity function). I + T is invertible. ker (T ) = Im (I T )

is linear. Determine ker (T ) and Im (T ).

Problem 5.2.5 It is easy to see that L : R2 R3 , x y x L = x + y y 0

104

Chapter 5

5.3 Matrix Representation

Let V, W be two vector spaces over the same eld F. Assume that dim V = m and {i }i[1;m] is an ordered v } basis for V, and that dim W = n and A = {wi i[1;n] an ordered basis for W. Then a11 a 21 . . . an1 A a12 a 22 . . . . an2
A

L(1 ) v

w w w a111 + a212 + + an1n

L(2 ) v . . .

w w w a121 + a222 + + an2n . . .

. . .

. . .

. . .

w w w L(m ) = a1m1 + a2m2 + + anmn v

a1n a 2n = . . . anm

248 Denition The n m matrix

formed by the column vectors above is called the matrix representation of the linear map L with respect to the bases {i }i[1;m] , {i }i[1;n] . v w 249 Example Consider L : R3 R3 , x x y z = L y x + y + z . z z

a11 a 21 ML = . . . an1

a12 a12 . . . an2

. . .

a1n a2n . . . anm

Clearly L is a linear transformation.

1. Find the matrix corresponding to L under the standard ordered basis.

Matrix Representation

105

1 1 1 2. Find the matrix corresponding to L under the ordered basis 0 , 1 , 0 , for both the domain 0 0 1 and the image of L. Solution: 1 1 0 1 0 1 1. The matrix will be a 3 3 matrix. We have L 0 = 1, L 1 = 1 , and L 0 = 1 , 0 0 0 0 1 1 whence the desired matrix is 1 1 0 1 1 0 1 1 . 1 ,

2. Call this basis A . We have

1 1 1 1 1 0 = = 0 + 1 + 0 = L 0 1 0 1 0 1 0 0 0 0 1 0

and

1 0 1 1 1 2 = = 2 + 2 + 0 = L 1 2 0 1 0 2 0 0 0 0 1 0 1 0 1 1 1 3 L 0 = 2 = 3 0 + 2 1 + 1 0 = 2 1 1 0 0 1 1 0 1 0 2 2 0 3 2 . 1

whence the desired matrix is

250 Example Let Rn [x] denote the set of polynomials with real coefcients with degree at most n.

106 1. Prove that R3 [x] p(x) R1 [x]

Chapter 5

L:

is a linear transformation. Here p (x) denotes the second derivative of p(x) with respect to x. 2. Find the matrix of L using the ordered bases {1, x, x2 , x3 } for R3 [x] and {1, x} for R2 [x]. 3. Find the matrix of L using the ordered bases {1, x, x2 , x3 } for R3 [x] and {1, x + 2} for R1 [x]. 4. Find a basis for ker (L) and nd dim ker (L). 5. Find a basis for Im (L) and nd dim Im (L). Solution: 1. Let (p(x), q(x)) (R3 [x])2 and R. Then L(p(x) + q(x)) = (p(x) + q(x)) = p (x) + q (x) = L(p(x)) + L(q(x)), whence L is linear. 2. We have d2 dx 0 = 0(1) + 0(x) = , 0 0 = 0(1) + 0(x) = , 0

p (x)

L(1)

1 2

L(x)

d2 x dx2

L(x2 ) =

d2 dx

x2 2

L(x3 ) =

d2 dx

x3 2

= 6x

2 = 2(1) + 0(x) = , 0 0 = 0(1) + 6(x) = , 6

whence the matrix representation of L under the standard basis is

0 0 0 0

2 0

0 . 6

Matrix Representation 3. We have L(1) = d

2

107

dx2

0(1) + 0(x + 2)

L(x)

d2 x dx2

0(1) + 0(x + 2)

L(x2 ) =

d2 dx2

x2

2(1) + 0(x + 2)

L(x3 ) =

d2 dx

x3 2

= 6x

whence the matrix representation of L under the standard basis is 0 0 0 2 0 0 12 . 6

12 , = 12(1) + 6(x + 2) = 6

2 , 0

0 , 0

0 , 0

4. Assume that p(x) = a + bx + cx2 + dx3 ker (L). Then 0 = L(p(x)) = 2c + 6dx, This means that c = d = 0. Thus a, b are free and ker (L) = {a + bx : (a, b) R2 }. Hence dim ker (L) = 2. 5. By the Dimension Theorem, dim Im (L) = 4 2 = 2. Put q(x) = + x + x2 + x3 . Then L(q(x)) = 2 + 6(x) = (2)(1) + (6)(x). Clearly {1, x} are linearly independent and span Im (L). Hence Im (L) = span (1, x) = R1 [x]. 251 Example 1. A linear transformation T : R3 R3 is such that 2 T ( i ) = 1 ; 1 x R.

It is known that

3 T( j ) = 0 . 1

Im (T ) = span T ( i ), T ( j )

108 and that 1 ker (T ) = span 2 . 1

Chapter 5

2. Find and , and hence, the matrix representing T under the standard ordered basis. Solution: 1. Since T ( k ) Im (T ) and Im (T ) is generated by T ( i ) and T ( k ) there must be (, ) R2 with 2 3 2 + 3 + = . T ( k ) = T ( i ) + T ( j ) = 1 0 1 1 2 = T( i ) T( j ) T(k) 1 1 3 0 1 2 + 3 .

Argue that there must be and such that T ( k ) = T ( i ) + T ( j ).

2. The matrix of T is

1 Since 2 ker (T ) we must have 1 2 1 1 3 0 1

Solving the resulting system of linear equations we obtain = 1, = 2. The required matrix is thus 2 1 1 3 0 1 8 1 . 1

2 + 3 1 0 = . 2 0 1 0

If the linear mapping L : V W, dim V = n, dim W = m has matrix representation A Mmn (F), then dim Im (L) = rank (A).

Matrix Representation

109

Homework
Problem 5.3.1 Let T : R4 R3 be a linear transformations such that 1 0 1 T = 0 , 1 1 1 1 1 0 T = 1 , 1 1 0

Find the matrix of T with respect to the canonical bases. Find the dimensions and describe ker (T ) and Im (T ). Problem 5.3.2 1. A linear transformation T : R3 R3 has as image the plane with equation x + y + z = 0 and as kernel the line x = y = z. If 1 a T 1 = 0 , 1 2 2 3 T 1 = b , 5 1 1 1 T 2 = 2 . c 1

1 0 1 T = 0 , 1 1 0

1 1 2 T = 1 . 0 1 0

Find a, b, c.

2. Find the matrix representation of T under the standard basis. Problem 5.3.3 1. Prove that T : R2 R3

is a linear transformation. 2. Find a basis for ker (T ) and nd dim ker (T ) 3. Find a basis for Im (T ) and nd dim Im (T ).

x+y x T = xy y 2x + 3y

1 1 0 1 1 2 4. Find the matrix of T under the ordered bases A = , of R and B = 1 , 0 , 1 of R3 . 2 3 1 0 1 Problem 5.3.4 Let L: where R3 a R2 L() a ,

Clearly L is linear. Find a matrix representation for L if

x x + 2y . L y = 3x z z

110
1. The bases for both R3 and R2 are both the standard ordered bases. 1 1 1 2. The ordered basis for R3 is 0 , 1 , 1 and R2 has the standard ordered basis . 1 0 0

Chapter 5

1 1 1 1 1 3 , , and the ordered basis for R2 is A = , . 3. The ordered basis for R is 0 1 1 0 1 1 0 0

1 2 Problem 5.3.5 A linear transformation T : R2 R2 satises ker (T ) = Im (T ), and T = . Find the matrix 3 1 representing T under the standard ordered basis. Problem 5.3.6 Find the matrix representation for the linear map M22 (R) L: A under the standard basis 1 A = 0 1 0 , 1 0 R tr (A) 0 1

for M22 (R).

0 0 , 0 0

0 0 , 0 0

Problem 5.3.7 Let A Mnp (R), B Mpq (R), and C Mqr (R), be such that rank (B) = rank (AB). Shew that rank (BC) = rank (ABC) .

Chapter

Determinants
6.1 Permutations
252 Denition Let S be a nite set with n 1 elements. A permutation is a bijective function : S S. It is easy to see that there are n! permutations from S onto itself. Since we are mostly concerned with the action that exerts on S rather than with the particular names of the elements of S, we will take S to be the set S = {1, 2, 3, . . . , n}. We indicate a permutation by means of the following convenient diagram 2 1 = (1) (2) n . (n)

253 Denition The notation Sn will denote the set of all permutations on {1, 2, 3, . . . , n}. Under this notation, the composition of two permutations (, ) S2 is n 2 1 = (1) (2) = 1 2 n 1 (1) (2) (n) ( )(n) n n (n) .

( )(1) ( )(2)

The k-fold composition of is

k compositions

= k .

We usually do away with the and write simply as . This product of permutations is thus simply function composition. Given a permutation : S S, since is bijective,

1 : S S 111

112 exists and is also a permutation. In fact if

Chapter 6

then

2 1 = (1) (2) (2) 2

n , (n) (n) . n

(1) 1 = 1

Figure 6.1: S3 are rotations and reexions.

254 Example The set S3 has 3! = 6 elements, which are given below. 1. Id : {1, 2, 3} {1, 2, 3} with

2. 1 : {1, 2, 3} {1, 2, 3} with

1 2 Id = 1 2 1 2 1 = 1 3 1 2 2 = 3 2 1 2 3 = 2 1 1 2 1 = 2 3

3 . 3 3 . 2 3 . 1 3 . 3 3 . 1

3. 2 : {1, 2, 3} {1, 2, 3} with

4. 3 : {1, 2, 3} {1, 2, 3} with

5. 1 : {1, 2, 3} {1, 2, 3} with

Permutations 6. 2 : {1, 2, 3} {1, 2, 3} with

113

1 2 2 = 3 1

3 . 2

255 Example The compositions 1 1 and 1 1 can be found as follows. 1 2 1 1 = 1 3 3 1 2 2 2 3 3 1 = 1 3

(We read from right to left 1 2 3 (1 goes to 2, 2 goes to 3, so 1 goes to 3), etc. Similarly Observe in particular that 1 1 = 1 1 . Finding all the other products we deduce the following multiplication table (where the multiplication operation is really composition of functions). Id 1 2 3 2 1 Id Id 1 2 3 2 1 1 1 Id 2 1 2 3 2 2 1 Id 2 3 1 3 3 2 1 Id 1 2 1 1 2 3 1 Id 2 2 2 3 1 2 1 Id 1 2 1 1 = 2 3 3 1 2 1 1 3 3 1 = 2 2 2 3 = 3 . 1 3

2 3 = 2 . 2 1

The permutations in example 254 can be conveniently interpreted as follows. Consider an equilateral triangle with vertices labelled 1, 2 and 3, as in gure 6.1. Each a is a reexion (ipping) about the line joining the vertex a with the midpoint of the side opposite a. For example 1 xes 1 and ips 2 and 3. Observe that two successive ips return the vertices to their original position and so (a {1, 2, 3})(2 = Id ). Similarly, 1 is a rotation of the vertices by an angle of 120 . Three successive a rotations restore the vertices to their original position and so 3 = Id . 1 256 Example To nd 1 take the representation of 1 and exchange the rows: 1 This is more naturally written as 1 1 = 1 1 3 2 2 . 3

1 1 = 1 1

2 3

3 . 2

114 Observe that 1 = 1 . 1 257 Example To nd 1 take the representation of 1 and exchange the rows: 1 2 1 = 1 1 3 2 1 . 3

Chapter 6

This is more naturally written as

Observe that 1 = 2 . 1

1 1 = 1 3

2 1

3 . 2

6.2 Cycle Notation

We now present a shorthand notation for permutations by introducing the idea of a cycle. Consider in S9 the permutation We start with 1. Since 1 goes to 2 and 2 goes back to 1, we write (12). Now we continue with 3. Since 3 goes to 3, we write (3). We continue with 4. As 4 goes 6, 6 goes to 7, 7 goes 8, and 8 goes back to 4, we write (4678). We consider now 5 which goes to 9 and 9 goes back to 5, so we write (59). We have written as a product of disjoint cycles = (12)(3)(4678)(59). This prompts the following denition. 258 Denition Let l 1 and let i1 , . . . , il {1, 2, . . . n} be distinct. We write (i1 i2 . . . il ) for the element Sn such that (ir ) = ir+1 , 1 r < l, (il ) = i1 and (i) = i for i {i1 , . . . , il }. We say that (i1 i2 . . . il ) is a cycle of length l. The order of a cycle is its length. Observe that if has order l then l = Id . 1 = 2 2 1 3 3 4 5 6 9 6 7 7 8 8 4 9 . 5

 Observe that (i
have

. . . il i1 ) = (i1 . . . il ) etc., and that (1) = (2) = = (n) = Id . In fact, we (i1 . . . il ) = (j1 . . . jm )

if and only if (1) l = m and if (2) l > 1: a such that k: ik = jk+a mod l . Two cycles (i1 , . . . , il ) and (j1 , . . . , jm ) are disjoint if {i1 , . . . , il } {j1 , . . . , jm } = . Disjoint cycles commute and if = 1 2 t is the product of disjoint cycles of length l1 , l2 , . . . , lt respectively, then has order lcm (l1 , l2 , . . . , lt ) . 259 Example A cycle decomposition for S9 , is The order of is lcm (3, 4) = 12. 1 2 = 1 8

3 7

4 5 6 2

6 3

7 4

8 9 5 9

(285)(3746).

Cycle Notation 260 Example The cycle decomposition = (123)(567) in S9 arises from the permutation 1 = 2 2 3 3 1 4 5 4 6 6 7 7 8 5 8 9 . 9

115

Its order is lcm (3, 3) = 3.

261 Example Find a shufe of a deck of 13 cards that requires 42 repeats to return the cards to their original order. Solution: Here is one (of many possible ones). Observe that 7 + 6 = 13 and 7 6 = 42. We take the permutation (1 2 3 4 5 6 7)(8 9 10 11 12 13) which has order 42. This corresponds to the following shufe: For i {1, 2, 3, 4, 5, 6, 8, 9, 10, 11, 12}, take the ith card to the (i + 1)th place, take the 7th card to the rst position and the 13th card to the 8th position. Query: Of all possible shufes of 13 cards, which one takes the longest to restitute the cards to their original position? 262 Example Let a shufe of a deck of 10 cards be made as follows: The top card is put at the bottom, the deck is cut in half, the bottom half is placed on top of the top half, and then the resulting bottom card is put on top. How many times must this shufe be repeated to get the cards in the initial order? Explain. Solution: Putting the top card at the bottom corresponds to 1 2 2 3 3 4 4 5 5 6 6 7 7 8 8 9 10 . 1

9 10

Cutting this new arrangement in half and putting the lower half on top corresponds to 1 2 7 8 3 9 4 5 6 2 7 8 3 4 9 5 10 . 6

10 1

Putting the bottom card of this new arrangement on top corresponds to 1 2 6 7 3 8 4 9 5 6 7 8 2 3 9 4 10 = (1 6)(2 7)(3 8)(4 9)(5 10). 5

10 1

The order of this permutation is lcm(2, 2, 2, 2, 2) = 2, so in 2 shufes the cards are restored to their original position. The above examples illustrate the general case, given in the following theorem. 263 Theorem Every permutation in Sn can be written as a product of disjoint cycles.

116

Chapter 6 Proof: Let Sn , a1 {1, 2, . . . , n}. Put k (a1 ) = ak+1 , k 0. Let a1 , a2 , . . . , as be the longest chain with no repeats. Then we have (as ) = a1 . If the {a1 , a2 , . . . , as } exhaust {1, 2, . . . , n} then we have = (a1 a2 . . . as ). If not, there exist b1 {1, 2, . . . , n} \ {a1 , a2 , . . . , as }. Again, we nd the longest chain of distinct b1 , b2 , . . . , bt such that (bk ) = bk+1 , k = 1, . . . , t 1 and (bt ) = b1 . If the {a1 , a2 , . . . , as , b1 , b2 , . . . , bt } exhaust all the {1, 2, . . . , n} we have = (a1 a2 . . . as )(b1 b2 . . . bt ). If not we continue the process and nd = (a1 a2 . . . as )(b1 b2 . . . bt )(c1 . . .) . . . . This process stops because we have only n elements. u

264 Denition A transposition is a cycle of length 2.1 265 Example The cycle (23468) can be written as a product of transpositions as follows (23468) = (28)(26)(24)(23). Notice that this decomposition as the product of transpositions is not unique. Another decomposition is (23468) = (23)(34)(46)(68). 266 Lemma Every permutation is the product of transpositions. Proof: It is enough to observe that (a1 a2 . . . as ) = (a1 as )(a1 as1 ) (a1 a2 ) and appeal to Theorem 263. u Let Sn and let (i, j) {1, 2, . . . , n}2 , i = j. Since is a permutation, (a, b) {1, 2, . . . , n}2 , a = b, such that (j) (i) = b a. This means that (i) (j)
1i<jn

ij

= 1.

267 Denition Let Sn . We dene the sign sgn() of as sgn() =

1i<jn

(i) (j) ij

= (1) .

If sgn() = 1, then we say that is an even permutation, and if sgn() = 1 we say that is an odd permutation.

 Notice that in fact

sgn() = (1)I() ,

where I() = #{(i, j) | 1 i < j n and (i) > (j)}, i.e., I() is the number of inversions that effects to the identity permutation Id . 268 Example The transposition (1 2) has one inversion. 269 Lemma For any transposition (k l) we have sgn((k l)) = 1.
1A

cycle of length 2 should more appropriately be called a bicycle.

Cycle Notation Proof: Let be transposition that exchanges k and l, and assume that k < l: 1 = 1 2 2 ... ... k1 k1 k k + 1 ... l k + 1 ... l1 l1 l k l+1 l+1 ... ... n n

117

Let us count the number of inversions of : The pairs (i, j) with i {1, 2, . . . , k 1} {l, l + 1, . . . , n} and i < j do not suffer an inversion; The pair (k, j) with k < j suffers an inversion if and only if j {k + 1, k + 2, . . . , l}, making l k inversions; If i {k + 1, k + 2, . . . , l 1} and i < j, (i, j) suffers an inversion if and only if j = l, giving l 1 k inversions. This gives a total of I() = (l k) + (l 1 k) = 2(l k 1) + 1 inversions when k < l. Since this number is odd, we have sgn() = (1)I() = 1. In general we see that the transposition (k l) has 2|k l| 1 inversions. u 270 Theorem Let (, ) S2 . Then n Proof: We have sgn() = =
1i<jn ()(i)()(j) ij ((i))((j)) (i)(j)

sgn() = sgn()sgn().

1i<jn

1i<jn

(i)(j) ij

The second factor on this last equality is clearly sgn(), we must shew that the rst factor is sgn(). Observe now that for 1 a < b n we have (a) (b) ab = (b) (a) ba .

Since and are permutations, b = a, (i) = a, (j) = b and so (i) = (a), (j) = b. Thus ((i)) ((j)) (i) (j) and so
1i<jn

(a) (b) ab (a) (b)

((i)) ((j)) (i) (j)

=
1a<bn

ab

= sgn().

u 271 Corollary The identity permutation is even. If Sn , then sgn() = sgn(1 ). Proof: Since there are no inversions in Id , we have sgn(Id ) = (1)0 = 1. Since 1 = Id , 1 we must have 1 = sgn(Id ) = sgn(1 ) = sgn()sgn(1 ) = (1) (1) by Theorem 270. Since the values on the righthand of this last equality are 1, we must have sgn() = sgn(1 ). u 272 Lemma We have sgn(1 2 . . . l)) = (1)l1 . Proof: Simply observe that the number of inversions of (1 2 . . . l) is l 1. u

118 273 Lemma Let (, (i1 . . . il ) S2 . Then n (i1 . . . il )1 = ((i1 ) . . . (il )), and if Sn is a cycle of length l then . sgn() = (1)l1

Chapter 6

Proof: For 1 k < l we have ((i1 . . . il )1 )((ik )) = ((i1 . . . il )(ik )) = (ik+1 ). On a ((i1 . . . il )1 )((il )) = ((i1 . . . il )(il )) = (i1 ). For i {(i1 ) . . . (il )} we have 1 (i) {i1 . . . il } whence (i1 . . . il )(1 (i)) = 1 (i) etc. Furthermore, write = (i1 . . . il ). Let Sn be such that (k) = ik for 1 k l. Then = (1 2 . . . l)1 and so we must have sgn() = sgn()sgn((1 2 . . . l))sgn(1 ), which equals sgn((1 2 . . . l)) by virtue of Theorem 270 and Corollary 271. The result now follows by appealing to Lemma 272 u 274 Corollary Let = 1 2 r be a product of disjoint cycles, each of length l1 , . . . , lr , respectively. Then r sgn() = (1) i=1 (li 1) . Hence, the product of two even permutations is even, the product of two odd permutations is even, and the product of an even permutation and an odd permutation is odd. Proof: This follows at once from Theorem 270 and Lemma 273. u

275 Example The cycle (4678) is an odd cycle; the cycle (1) is an even cycle; the cycle (12345) is an even cycle. 276 Corollary Every permutation can be decomposed as a product of transpositions. This decomposition is not necessarily unique, but its parity is unique. Proof: This follows from Theorem 263, Lemma 266, and Corollary 274. u

277 Example (The 15 puzzle) Consider a grid with 16 squares, as shewn in (6.1), where 15 squares are numbered 1 through 15 and the 16th slot is empty. 1 5 9 2 6 10 3 7 4 8 (6.1)

11 12 15

13 14

In this grid we may successively exchange the empty slot with any of its neighbours, as for example 1 5 9 2 6 10 3 7 4 8 . 11 12 15 (6.2)

13 14

Determinants We ask whether through a series of valid moves we may arrive at the following position. 1 5 9 2 6 10 3 7 4 8

119

(6.3)

11 12 14

13 15

Solution: Let us shew that this is impossible. Each time we move a square to the empty position, we make transpositions on the set {1, 2, . . . , 16}. Thus at each move, the permutation is multiplied by a transposition and hence it changes sign. Observe that the permutation corresponding to the square in (6.3) is (14 15) (the positions 14th and 15th are transposed) and hence it is an odd permutation. But we claim that the empty slot can only return to its original position after an even permutation. To see this paint the grid as a checkerboard: B R B R R B R B B R R B (6.4)

B R R B

We see that after each move, the empty square changes from black to red, and thus after an odd number of moves the empty slot is on a red square. Thus the empty slot cannot return to its original position in an odd number of moves. This completes the proof.

Homework
Problem 6.2.1 Decompose the permutation

as a product of disjoint cycles and nd its order.

1 2

2 3

3 4

4 1

5 5

6 8

7 6

8 7

9 9

6.3 Determinants
There are many ways of developing the theory of determinants. We will choose a way that will allow us to deduce the properties of determinants with ease, but has the drawback of being computationally cumbersome. In the next section we will shew that our way of dening determinants is equivalent to a more computationally friendly one. It may be pertinent here to quickly review some properties of permutations. Recall that if Sn is a cycle of length l, then its signum sgn() = 1 depending on the parity of l 1. For example, in S7 , = (1 3 4 7 6) has length 5, and the parity of 5 1 = 4 is even, and so we write sgn() = +1. On the other hand, = (1 3 4 7 6 5)

120 has length 6, and the parity of 6 1 = 5 is odd, and so we write sgn() = 1. Recall also that if (, ) S2 , then n sgn() = sgn()sgn(). Thus from the above two examples = (1 3 4 7 6)(1 3 4 7 6 5)

Chapter 6

has signum sgn()sgn() = (+1)(1) = 1. Observe in particular that for the identity permutation Id Sn we have sgn(Id ) = +1. 278 Denition Let A Mnn (F), A = [aij ] be a square matrix. The determinant of A is dened and denoted by the sum sgn()a1(1) a2(2) an(n) . det A =
Sn

 The determinantal sum has n! summands.

279 Example If n = 1, then S1 has only one member, Id , where Id (1) = 1. Since Id is an even permutation, sgn(Id ) = (+1) Thus if A = (a11 ), then det A = a11 . 280 Example If n = 2, then S2 has 2! = 2 members, Id and = (1 2). Observe that sgn() = 1. Thus if a11 A= a21 a12 a22

then det A = sgn(Id )a1Id

(1) a2Id (2)

+ sgn()a1(1) a2(2) = a11 a22 a12 a21 .

281 Example From the above formula for 2 2 matrices it follows that det A 1 2 = det 3 4

= (1)(4) (3)(2) = 2, 2 (1)(4) (3)(2) 4

det B =

= and

1 det 3 10,

Observe in particular that det(A + B) = det A + det B.

0 det(A + B) = det 6

4 = (0)(8) (6)(4) = 24. 8

Determinants 282 Example If n = 3, then S2 has 3! = 6 members: Id , 1 = (2 3), 2 = (1 3), 3 = (1 2), 1 = (1 2 3), 2 = (1 3 2). . Observe that Id , 1 , 2 are even, and 1 , 2 , 3 are odd. Thus if

121

a11 A = a21 a31 then det A = sgn(Id )a1Id

a12 a22 a32

a13 a23 a33

(1) a2Id (2) a3Id (3)

+ sgn(1 )a11 (1) a21 (2) a31 (3)

+sgn(2 )a12 (1) a22 (2) a32 (3) + sgn(3 )a13 (1) a23 (2) a33 (3) +sgn(1 )a11 (1) a21 (2) a31 (3) + sgn(2 )a12 (1) a22 (2) a32 (3) = a11 a22 a33 a11 a23 a32 a13 a22 a31 a13 a21 a33 + a12 a23 a31 + a13 a21 a32 .

283 Theorem (Row-Alternancy of Determinants) Let A Mnn (F), A = [aij ]. If B Mnn (F), B = [bij ] is the matrix obtained by interchanging the s-th row of A with its t-th row, then det B = det A. Proof: Let be the transposition s = (t) t . (s)

Then (a) = (a) for a {1, 2, . . . , n} \ {s, t}. Also, sgn() = sgn()sgn() = sgn(). As ranges through all permutations of Sn , so does , hence det B sgn()b1(1) b2(2) bs(s) bt(t) bn(n) sgn()a1(1) a2(2) ast ats an(n) sgn()a1(1) a2(2) as(s) at(t) an(n) sgn()a1(1) a2(2) an(n)

= = = =

Sn

Sn

Sn

Sn

= det A. u

122 284 Corollary If A(r:k) , 1 k n denote the rows of A and Sn , then A(r:(1)) A (r:(2)) = (sgn()) det A. det . . . A(r:(n)) An analogous result holds for columns. Proof:

Chapter 6

Apply the result of Theorem 283 multiple times. u

285 Theorem Let A Mnn (F), A = [aij ]. Then det AT = det A. Proof: Let C = AT . By denition det AT = det C = =
Sn

sgn()c1(1) c2(2) cn(n) sgn()a(1)1 a(2)2 a(n)n .

Sn

But the product a(1)1 a(2)2 a(n)n also appears in det A with the same signum sgn(), since the permutation is the inverse of the permutation (1) (2) 1 2 (n) n

2 1 (1) (2)

n . (n)

286 Corollary (Column-Alternancy of Determinants) Let A Mnn (F), A = [aij ]. If C Mnn (F), C = [cij ] is the matrix obtained by interchanging the s-th column of A with its t-th column, then det C = det A. Proof: This follows upon combining Theorem 283 and Theorem 285. u

287 Theorem (Row Homogeneity of Determinants) Let A Mnn (F), A = [aij ] and F. If B Mnn (F), B = [bij ] is the matrix obtained by multiplying the s-th row of A by , then det B = det A. Proof: Simply observe that

sgn()a1(1) a2(2) as(s) an(n) = sgn()a1(1) a2(2) as(s) an(n) . u

Determinants

123

288 Corollary (Column Homogeneity of Determinants) If C Mnn (F), C = (Cij ) is the matrix obtained by multiplying the s-th column of A by , then

det C = det A.

Proof:

This follows upon using Theorem 285 and Theorem 287. u

It follows from Theorem 287 and Corollary 288 that if a row (or column) of a matrix consists of 0F s only, then the determinant of this matrix is 0F .

289 Example

x det x2 x3

1 1 1

a 1 = x det b x c x2

1 1 1

a b . c

290 Corollary

det(A) = n det A.

Proof:

Since there are n columns, we are able to pull out one factor of from each one. u

291 Example Recall that a matrix A is skew-symmetric if A = AT . symmetric. Prove that det A = 0.

Let A M2001 (R) be skew-

Solution: We have

det A = det(AT ) = (1)2001 det AT = det A,

and so 2 det A = 0, from where det A = 0.

292 Lemma (Row-Linearity and Column-Linearity of Determinants) Let A Mnn (F), A = [aij ].

For a

124 xed row s, suppose that asj = bsj + csj for each j [1; n]. Then
a11 a12 a22 . . . a(s1)2 bs2 + cs2 a(s+1)2 . . . an2 a1n a2n . . . a(s1)n bsn + csn a(s+1)n . . . ann

Chapter 6

a21 . . . a(s1)1 det bs1 + cs1 a (s+1)1 . . .

an1

. . .

a21 .. . a = det (s1)1 bs1 a (s+1)1 . ..

an1

a11

a12 a22 . . . a(s1)2 bs2 a(s+1)2 . . . an2

a1n a2n . . . a(s1)n bsn a(s+1)n . . . ann

. . .

a21 .. . a + det (s1)1 cs1 a (s+1)1 . ..

an1

. . .

. . .

a11

a12 a22 . . . a(s1)2 cs2 a(s+1)2 . . . an2

a1n a2n . . . a(s1)n csn a(s+1)n . . . ann

. . .

. . . .

An analogous result holds for columns. Proof: Put

a11 a12 a22 . . . a(s1)2 bs2 + cs2 a(s+1)2 . . . an2 a1n a2n . . . a(s1)n bsn + csn a(s+1)n . . . ann

a21 . . . a(s1)1 S= bs1 + cs1 a (s+1)1 . . .

an1

. . .

, . . .

a21 .. . a T = (s1)1 bs1 a (s+1)1 . ..

an1

a11

a12 a22 . . . a(s1)2 bs2 a(s+1)2 . . . an2

a1n a2n . . . a(s1)n bsn a(s+1)n . . . ann

. . .
. . .

Determinants and a21 .. . a U = (s1)1 cs1 a (s+1)1 . ..

an1

125

a11

a12 a22 . . . a(s1)2 cs2 a(s+1)2 . . . an2

a1n a2n . . . a(s1)n csn a(s+1)n . . . ann

. . .

Then
det S =
Sn

. . . .

sgn()a1(1) a2(2) a(s1)(s1) (bs(s)

+cs(s) )a(s+1)(s+1) an(n) = + =

Sn

sgn()a1(1) a2(2) a(s1)(s1) bs(s) a(s+1)(s+1) an(n) sgn()a1(1) a2(2) a(s1)(s1) cs(s) a(s+1)(s+1) an(n)

Sn

det T + det U.

By applying the above argument to AT , we obtain the result for columns. u 293 Lemma If two rows or two columns of A Mnn (F), A = [aij ] are identical, then det A = 0F . Proof: Suppose asj = atj for s = t and for all j [1; n]. In particular, this means that for any Sn we have as(t) = at(t) and at(s) = as(s) . Let be the transposition s = (t) t . (s)

Then (a) = (a) for a {1, 2, . . . , n} \ {s, t}. Also, sgn() = sgn()sgn() = sgn(). As runs through all even permutations, runs through all odd permutations, and viceversa. Therefore detA = = sgn()a1(1) a2(2) as(s) at(t) an(n) sgn()a1(1) a2(2) as(s) at(t) an(n)

Sn

sgn()=1

Sn

+sgn()a1(1) a2(2) as(s) at(t) an(n) =

sgn()=1
Sn

sgn() a1(1) a2(2) as(s) at(t) an(n)

a1(1) a2(2) as(t) at(s) an(n) =

sgn()=1
Sn

sgn() a1(1) a2(2) as(s) at(t) an(n)

a1(1) a2(2) at(t) as(s) an(n) = 0F . Arguing on AT will yield the analogous result for the columns. u

126

Chapter 6

294 Corollary If two rows or two columns of A Mnn (F), A = [aij ] are proportional, then det A = 0F . Proof: Suppose asj = atj for s = t and for all j [1; n]. If B is the matrix obtained by pulling out the factor from the s-th row then det A = det B. But now the s-th and the t-th rows in B are identical, and so det B = 0F . Arguing on AT will yield the analogous result for the columns. u 295 Example 1 det 1 1 a b 1 1 = a det a c 1 1 a d 1 1 b c = 0, d

since on the last determinant the rst two columns are identical.

296 Theorem (Multilinearity of Determinants) Let A Mnn (F), A = [aij ] and F. If X Mnn (F), X = (xij ) is the matrix obtained by the row transvection Rs + Rt Rs then det X = det A. Similarly, if Y Mnn (F), Y = (yij ) is the matrix obtained by the column transvection Cs + Ct Cs then det Y = det A. Proof: For the row transvection it sufces to take bsj = asj , csj = atj for j [1; n] in Lemma 292. With the same notation as in the lemma, T = A, and so, det X = det T + det U = det A + det U. But U has its s-th and t-th rows proportional (s = t), and so by Corollary 294 det U = 0F . Hence det X = det A. To obtain the result for column transvections it sufces now to also apply Theorem 285. u 297 Example Demonstrate, without actually calculating the determinant that 2 det 4 7 9 6 4 9 8 1

is divisible by 13.

Solution: Observe that 299, 468 and 741 are all divisible by 13. Thus 2 9 det 4 6 7 4 9 8 1 2 det 4 7 9 6 4 299 2 9 = 13 det 4 6 468 741 7 4 23 36 , 57

C3 +10C2 +100C1 C3

which shews that the determinant is divisible by 13. 298 Theorem The determinant of a triangular matrix (upper or lower) is the product of its diagonal elements.

Determinants Proof: Let A Mnn (F), A = [aij ] be a triangular matrix. Observe that if = Id then ai(i) a(i)2 (i) = 0F occurs in the product a1(1) a2(2) an(n) . Thus det A = = u 299 Example The determinant of the n n identity matrix In over a eld F is det In = 1F . 300 Example Find 1 det 4 7 Solution: We have 1 2 det 4 5 7 8 3 6 9 1 0 det 4 3 7 6 0 6 12 2 5 8 3 6 . 9
Sn

127

sgn()a1(1) a2(2) an(n)

(n)

sgn(Id )a1Id (1) a2Id (2) anId

= a11 a22 ann .

C3 3C1 C3

C2 2C1 C2

1 0 (3)(6) det 4 1 7 2 0,

0 1 2

since in this last matrix the second and third columns are identical and so Lemma 293 applies. 301 Theorem Let (A, B) (Mnn (F))2 . Then det(AB) = (det A)(det B).

Proof: Put D = AB, D = (dij ), dij = k=1 aik bkj . If A(c:k) , D(c:k) , 1 k n denote the columns of A and D, respectively, observe that
n

D(c:k) =
l=1

blk A(c:l) , 1 k n.

128 Applying Corollary 288 and Lemma 292 multiple times, we obtain det D = det(D(c:1) , D(c:2) , . . . , D(c:n) ) =
n j1 =1 n j2 =1

Chapter 6

n jn =1

b1j1 b2j2 bnjn

det(A(c:j1 ) , A(c:j2 ) , . . . , A(c:jn ) ). By Lemma 293, if any two of the A(c:jl ) are identical, the determinant on the right vanishes. So each one of the jl is different in the non-vanishing terms and so the map {1, 2, . . . , n} : l {1, 2, . . . , n} jl

is a permutation. Here jl = (l). Therefore, for the non-vanishing det(A(c:j1 ) , A(c:j2 ) , . . . , A(c:jn ) ) we have in view of Corollary 284, det(A(c:j1 ) , A(c:j2 ) , . . . , A(c:jn ) ) = (sgn()) det(A(c:1) , A(c:2) , . . . , A(c:n) ) = (sgn()) det A. We deduce that det(AB) = = = = as we wanted to shew. u By applying the preceding theorem multiple times we obtain 302 Corollary If A Mnn (F) and if k is a positive integer then det Ak = (det A)k . det D
n jn =1

b1j1 b2j2 bnjn det(A(c:j1 ) , A(c:j2 ) , . . . , A(c:jn ) )

Sn (sgn())b1(1) b2(2)

(det A)

bn(n)

(det A)(det B),

303 Corollary If A GLn (F) and if k is a positive integer then det A = 0F and det Ak = (det A)k .

Proof: We have AA1 = In and so by Theorem 301 (det A)(det A1 ) = 1F , from where the result follows. u

Homework
Problem 6.3.1 Let bc = det a a2 ca b b2 ab c . 2 c

Laplace Expansion
Without expanding either determinant, prove that 1 = det a2 a3 Problem 6.3.2 Demonstrate that a b c = det 2b 2c 2a bca 2c 2a 2b cab 1 b2 b3 1 2 . c 3 c

129

= (a + b + c)3 .

Problem 6.3.3 After the indicated column operations on a 33 matrix A with det A = 540, matrices A1 , A2 , . . . , A5 are successively obtained: A
C1 +3C2 C1

Determine the numerical values of det A1 , det A2 , det A3 , det A4 and det A5 . Problem 6.3.4 Prove, without actually expanding the determinant, that 1 6 det 8 7 2 2 1 6 3 4 3 5 1 8 6 7 14 21 7 0 0 1 3 1 4

A1

C2 C3

A2

3C2 C1 C2

A3

C1 3C2 C1

A4

2C1 C1

A5

is divisible by 1722.

Problem 6.3.5 Let A, B, C be 3 3 matrices with det A = 3, det B3 = 8, det C = 2. Compute (i) det ABC, (ii) det 5AC, (iii) det A3 B3 C1 . Express your answers as fractions. Problem 6.3.6 Shew that A Mnn (R), (X, Y) (Mnn (R))2 , (det X)(det Y) = 0 such that A = X + Y. That is, any square matrix over R can be written as a sum of two matrices whose determinant is not zero. Problem 6.3.7 Prove or disprove! The set X = {A Mnn (F) : det A = 0F } is a vector subspace of Mnn (F).

6.4 Laplace Expansion

We now develop a more computationally convenient approach to determinants. Put Cij =
Sn

(sgn())a1(1) a2(2) an(n) .

(i)=j

130 Then det A = =

Sn (sgn())a1(1) a2(2) n i=1

Chapter 6

an(n) (6.5)

aij

Sn

(i)=j

(sgn())a1(1) a2(2)

a(i1)(i1) a(i+1)(i+1) an(n) =

n i=1

aij Cij ,

is the expansion of det A along the j-th column. Similarly, det A = =

Sn (sgn())a1(1) a2(2) n j=1

an(n)

aij

Sn

(i)=j

(sgn())a1(1) a2(2)

a(i1)(i1) a(i+1)(i+1) an(n) =

n j=1

aij Cij ,

is the expansion of det A along the i-th row. 304 Denition Let A Mnn (F), A = [aij ]. The ij-th minor Aij Mn1 (R) is the (n 1) (n 1) matrix obtained by deleting the i-th row and the j-th column from A. 305 Example If 1 2 A = 4 5 7 8 3 6 9

then, for example,

5 A11 = 8

6 , 9

4 A12 = 7

6 , 9

2 A21 = 8

3 , 9

1 A22 = 7

3 , 9

1 2 . A33 = 4 5

306 Theorem Let A Mnn (F). Then

n n

det A =
i=1

aij (1)i+j det Aij =

j=1

aij (1)i+j det Aij .

Proof:

It is enough to shew, in view of 6.5 that (1)i+j det Aij = Cij .

Now, Cnn = =
Sn

(n)=n

sgn()a1(1) a2(2) a(n1)(n1) sgn()a1(1) a2(2) a(n1)(n1)

Sn1

= det Ann ,

Laplace Expansion since the second sum shewn is the determinant of the submatrix obtained by deleting the last row and last column from A. To nd Cij for general ij we perform some row and column interchanges to A in order to bring aij to the nn-th position. We thus bring the i-th row to the n-th row by a series of transpositions, rst swapping the i-th and the (i + 1)-th row, then swapping the new (i + 1)-th row and the (i + 2)-th row, and so forth until the original i-th row makes it to the n-th row. We have made thereby n i interchanges. To this new matrix we perform analogous interchanges to the j-th column, thereby making n j interchanges. We have made a total of 2n i j interchanges. Observe that (1)2nij = (1)i+j . Call the analogous quantities in the resulting matrix A , Cnn , Ann . Then Cij = Cnn = det Ann = (1)i+j det Aij , by virtue of Corollary 284. u It is irrelevant which row or column we choose to expand a determinant of a square matrix. We always obtain the same result. The sign pattern is given by + + + + + . . . . . . . . . 307 Example Find 1 det 4 7 2 5 8 3 6 9 . . + . . . . . . . . . .

131

by expanding along the rst row. Solution: We have 5 = 1(1)1+1 det 8

det A

= 1(45 48) 2(36 42) + 3(32 35) = 0.

6 4 6 4 5 + 2(1)1+2 det + 3(1)1+3 det 9 7 9 7 8

308 Example Evaluate the Vandermonde determinant 1 det a a2 1 b b2 1 c . c2

132 Solution: 1 det a a2 1 b b2 1 c c2 1 = det a a2 0 ba b2 a2 0

Chapter 6

ba = det b2 c 2

= (b a)(c a) det

ca c2 a2

ca c2 a2

= (b a)(c a)(c b).

b+a c+a

309 Example Evaluate the determinant

det A = det

1 2 3 4

2 1 2 3

3 2 1 2

4 3 2 1

2000 1999 1998 1997

2000 1999 1998 . 1997 1

Solution: Applying Rn Rn+1 Rn for 1 n 1999, the determinant becomes 1 1 1 1 1 1 1 1 1 1 1 1 det 1 1 1 1 1 1 1 1 2000 1999 1998 1997 1 1 1 1 1 2 1 1 1 1 . 1 1

Laplace Expansion Applying now Cn + C2000 Cn for 1 n 1999, we obtain 2 2 2 0 0 2 2 0 0 0 0 2 det 0 0 0 0 0 0 0 0 2001 2000 1999 1998 2 2 2

133

2 2 2 2 0 3

This last determinant we expand along the rst column. We have 2 0 0 2001 det 0 0 2 0 0 0 2 2 0 0 2 2 2 0

1 1 1 1 . 1 1

1 1 1 = 2001(21998). 1 1

310 Denition Let A Mnn (F). The classical adjoint or adjugate of A is the n n matrix adj (A) whose entries are given by [adj (A)]ij = (1)i+j det Aji , where Aji is the ji-th minor of A. 311 Theorem Let A Mnn (F). Then (adj (A))A = A(adj (A)) = (det A)In . Proof: We have [A(adj (A))]ij = =
n k=1 n k=1

aik [adj (A)]kj aik (1)i+k det Ajk .

Now, this last sum is det A if i = j by virtue of Theorem 306. If i = j it is 0, since then the j-th row is identical to the i-th row and this determinant is 0F by virtue of Lemma 293. Thus on the diagonal entries we get det A and the off-diagonal entries are 0F . This proves the theorem. u The next corollary follows immediately. 312 Corollary Let A Mnn (F). Then A is invertible if and only det A = 0F and A1 = adj (A) det A .

134

Chapter 6

Homework
Problem 6.4.1 Find 1 det 4 7 2 5 8 3 6 9

by expanding along the second column.

a Problem 6.4.2 Prove that det c b

b a c

c 3 3 3 b = a + b + c 3abc. This type of matrix is called a circulant matrix. a

Problem 6.4.3 Compute the determinant 1 2 det 666 1 Problem 6.4.4 Prove that x + a det a a Problem 6.4.5 If b x+b b c 2 c = x (x + a + b + c). x+c 0 0 3 0 1 0 1 0 1 . 1000000 1 1

and xabc = 0, prove that

1 x det x x

1 a 0 0

1 0 b 0

1 0 = 0, 0 c

1 1 1 1 = + + . x a b c Problem 6.4.6 Consider the matrix

a b A= c d

b a d c

c d a b

d c . b a

Laplace Expansion
Compute AT A. Use the above to prove that det A = (a2 + b2 + c2 + d2 )2 . Problem 6.4.7 Prove that

135

0 a det 0 1 Problem 6.4.8 Demonstrate that

a 0 a 1

b b 0 1

0 0 = 2ab(a b). b 1

a 0 det c 0 Problem 6.4.9 Use induction to shew that

0 a 0 c

b 0 d 0

0 b 2 = (ad bc) . 0 d

1 1 0 det 0 . . . 0

1 0 1 0 . . . 0

1 0 0 1 0

. . .

1 0 0 0 . . . 1

. . .

1 0 0 n+1 . = (1) 0 0

Problem 6.4.10 Let 1 n n A= n . . . n n 2 n n . . . n n n 3 n . . . n n n n 4 n . . . n n n , n n

. . .

that is, A Mnn (R), A = [aij ] is a matrix such that akk = k and aij = n when i = j. Find det A. Problem 6.4.11 Let n N, n > 1 be an odd integer. Recall that the binomial coefcients and that for 1 k n, n1 n1 n . + = k k1 k
n k

satisfy

n n

n 0

= 1

136
Prove that 1 det n n1
n 1

Chapter 6

n 1

n 2 n 1

n n1 n n2 n n3

1 1
n 2

1
n 3

n n1 n n n = (1 + (1) ) . n2 1

Problem 6.4.12 Let A GLn (F), n > 1. Prove that det(adj (A)) = (det A)n1 . Problem 6.4.13 Let (A, B, S) (GLn (F))3 . Prove that adj (adj (A)) = (det A)n2 A. adj (AB) = adj (A) adj (B). adj SAS1 = S(adj (A))S1 . Problem 6.4.14 If A GL2 (F), , and let k be a positive integer. Prove that det(adj adj(A)) = det A.
k

Problem 6.4.15 Find the determinant 2 (b + c) det ab ac by hand, making explicit all your calculations. Problem 6.4.16 The matrix a d c b b a d c c b a d d c b a ab (a + c)2 bc ac bc (a + b)2

is known as a circulant matrix. Prove that its determinant is (a + b + c + d)(a b + c d)((a c)2 + (b d)2 ).

6.5 Determinants and Linear Systems

313 Theorem Let A Mnn (F). The following are all equivalent det A = 0F . A is invertible. There exists a unique solution X Mn1 (F) to the equation AX = Y. If AX = 0n1 then X = 0n1 .

Determinants and Linear Systems Proof: We prove the implications in sequence: = : follows from Corollary 312 = : If A is invertible and AX = Y then X = A1 Y is the unique solution of this equation. = : follows by putting Y = 0n1 = : Let R be the row echelon form of A. Since RX = 0n1 has only X = 0n1 as a solution, every entry on the diagonal of R must be non-zero, R must be triangular, and hence det R = 0F . Since A = PR where P is an invertible n n matrix, we deduce that det A = det P det R = 0F . u

137

The contrapositive form of the implications and will be used later. Here it is for future reference. 314 Corollary Let A Mnn (F). If there is X = 0n1 such that AX = 0n1 then det A = 0F .

Homework
1 Problem 6.5.1 For which a is the matrix 1 1 1 a 1 1 1 singular (non-invertible)? a

Chapter

Eigenvalues and Eigenvectors

7.1 Similar Matrices
315 Denition We say that A Mnn (F) is similar to B Mnn (F) if there exist a matrix P GLn (F) such that B = PAP1 .

316 Theorem Similarity is an equivalence relation. Proof: Let A Mnn (F). Then A = In AI1 , so similarity is reexive. If B = PAP1 (P n GLn (F) ) then A = P1 BP so similarity is symmetric. Finally, if B = PAP1 and C = QBQ1 (P GLn (F) , Q GLn (F)) then C = QPAP1 Q1 = QPA(QP)1 and so similarity is transitive. u Since similarity is an equivalence relation, it partitions the set of n n matrices into equivalence classes by Theorem 29. 317 Denition A matrix is said to be diagonalisable if it is similar to a diagonal matrix. Suppose that 1 0 A= . . . 0 K 1 0 K A = . . . 0 0 2 . . . 0 0 0 . . . 0 0 0 . . . . n 0 0 . . . . K n

Then if K is a positive integer

0 K 2 . . . 0

0 0 . . . 0

138

Eigenvalues and Eigenvectors In particular, if B is similar to A then

K 1

139

0 K 2 . . . 0

B = (PAP

)(PAP

) (PAP

) = PA P

K factors

so we have a simpler way of computing BK . Our task will now be to establish when a particular square matrix is diagonalisable.

= P

0 0 . . .

0 . . .

0 0 1 P , . . . K n

7.2 Eigenvalues and Eigenvectors

Let A Mnn (F) be a square diagonalisable matrix. Then there exist P GLn (F) and a diagonal matrix D Mnn (F) such that P1 AP = D, whence AP = DP. Put 0 1 0 0 0 0 2 0 , P = [P1 ; P2 ; ; Pn ], D= . . . . . . . . . . . . 0 0 0 n AP = DP = [AP1 ; AP2 ; ; APn ] = [1 P1 ; 2 P2 ; ; n Pn ], from where it follows that APk = k Pk . This motivates the following denition. 318 Denition Let V be a nite-dimensional vector space over a eld F and let T : V V be a linear transformation. A scalar F is called an eigenvalue of T if there is a = 0 (called an eigenvector) v such that T () = . v v

where the Pk are the columns of P. Then

319 Example Shew that if is an eigenvalue of T : V V, then k is an eigenvalue of T k : V V, for k N \ {0}. Solution: Assume that T () = . Then v v v T 2 () = TT () = T () = T () = () = 2. v v v v v ) = k, which is what we want. Continuing the iterations we obtain T k ( v v

320 Theorem Let A Mnn (F) be the matrix representation of T : V V. Then F is an eigenvalue of T if an only if det(In A) = 0F . Proof: is an eigenvalue of A there is = 0 such that A = A = 0 v v v v v In A = 0 det(In A) = 0F by Corollary 314.u v v det(In A) = 0F

321 Denition The equation

is called the characteristic equation of A or secular equation of A. The polynomial p() = det(In A) is the characteristic polynomial of A.

140 1 1 322 Example Let A = 0 0

Chapter 7

1 1 0 0

0 0 0 0 . Find 1 1 1 1

The characteristic polynomial of A. The eigenvalues of A. The corresponding eigenvectors. Solution: We have 1 1 det(I4 A) = det 0 0 1 0 0

1 = ( 1) det 0 0

1 0 0 0 1 1 0 1 1 0 0

= ( 1)(( 1)(( 1)2 1)) + ((( 1)2 1))

1 + det 1 1 0 1 1 0

1 1 1 1

= ( 1)(( 1)( 2)()) ( 2)() = ( 2)()(( 1)2 1) = ( 2)2 ()2 The eigenvalues are clearly = 0 and = 2. If = 0, then 1 1 = 0 0 1 1 0 0 0 0 1 1 0 0 . 1 1

0I4 A

Eigenvalues and Eigenvectors This matrix has row-echelon form 1 0 0 0 and if 1 0 0 0 0 a 0 1 b 0 = , 0 c 0 0 d 0 1 0 0 0 0 1 0 0 0 1 , 0 0

141

1 0 0 0

0 1 0 0

then c = d and a = b Thus the general solution of the system (0I4 A)X = 0n1 is a 1 0 b 1 0 = a + c . c 0 1 d 0 1 If = 2, then 1 1 = 0 0 1 1 0 0 0 0 1 1

2I4 A

This matrix has row-echelon form

0 0 . 1 1

1 0 0 0 and if

1 0 0 0

0 1 0 0

0 1 , 0 0

142

Chapter 7

then c = d and a = b Thus the general solution of the system (2I4 A)X = 0n1 is a 1 0 b 1 0 = a + c . c 0 1 d 0 1 Thus for = 0 we have the eigenvectors 1 0 1 0 , 0 1 0 1 and for = 2 we have the eigenvectors 1 0 1 0 , . 0 1 0 1 323 Theorem If = 0F is an eigenvalue of A, then A is non-invertible. Proof: Put p() = det(In A). Then p(0F ) = det(A) = (1)n det A is the constant term of the characteristic polynomial. If = 0F is an eigenvalue then p(0F ) = 0F = det A = 0F , and hence A is non-invertible by Theorem 313. u 324 Theorem Similar matrices have the same characteristic polynomial.

1 0 0 0

1 0 0 0

0 1 0 0

0 a 0 1 b 0 = , 0 c 0 0 d 0

Diagonalisability Proof: We have det(In SAS1 ) = det(SIn S1 SAS1 ) = det S(In A)S1 = (det S)(det(In A))(det S1 ) = (det S)(det(In A)) = det(In A), from where the result follows.u 1 det S

143

Homework
Problem 7.2.1 Find the eigenvalues and eigenvectors of 1 A= 1 1 1 0 Problem 7.2.5 Let A = 2 1 The eigenvalues of A. The corresponding eigenvectors. Problem 7.2.6 Describe all matrices A M22 (R) having eigenvalues 1 and 1. Problem 7.2.7 Let A Mnn (R). Demonstrate that A has the same characteristic polynomial as its transpose. 2 3 2 1 2. Find 0

Problem 7.2.2 Let A be a 2 2 matrix over some some eld F. Prove that the characteristic polynomial of A is 2 (tr (A)) + det A. Problem 7.2.3 A matrix A M22 (R) satises tr (A) = 1 and det A = 6. Find the value of det(I2 + A). Problem 7.2.4 A 2 2 matrix A with real entries has characteristic polynomial p() = 2 + 2 1. Find the value of det(2I2 + A).

The characteristic polynomial of A.

7.3 Diagonalisability
In this section we nd conditions for diagonalisability. 325 Theorem Let {1 , 2 , . . . , k } V be the eigenvectors corresponding to the different eigenvalues v v v {1 , 2 , . . . , k } (in that order). Then these eigenvectors are linearly independent. Proof: Let T : V V be the underlying linear transformation. We proceed by induction. For k = 1 the result is clear. Assume that every set of k 1 eigenvectors that correspond to k 1 distinct eigenvalues is linearly independent and let the eigenvalues 1 , 2 , . . . , k1 have corresponding eigenvectors 1 , 2 , . . . , k1 . Let be a eigenvalue different from the v v v 1 , 2 , . . . , k1 and let its corresponding eigenvector be . If were linearly dependent of the v v , , . . . , v1 v2 v k1 , we would have v (7.1) v v x + x11 + x22 + + xk1k1 = 0 . v Now by Theorem 240. This implies that v v v T (x + x11 + x22 + + xk1k1 ) = T ( 0 ) = 0 , v v v v x + x1 11 + x2 22 + + xk1 k1k1 = 0 . v

(7.2)

144 From 7.2 take away times 7.1, obtaining x1 (1 )1 + x2 (22 + + xk1 (k1 )k1 = 0 v v v

Chapter 7

(7.3)

Since i = 0F 7.3 is saying that the eigenvectors 1 , 2 , . . . , k1 are linearly dependent, v v v a contradiction. Thus the claim follows for k distinct eigenvalues and the result is proven by induction. u 326 Theorem A matrix A Mnn (F) is diagonalisable if and only if it possesses n linearly independent eigenvectors. Proof: Assume rst that A is diagonalisable, so there exists P GLn (F) and 1 0 D= . . . 0 such that 1 0 1 P AP = . . . 0 0 2 . . . 0 0 0 . . . 0 0 0 . . . . n 0 0 = [1 P1 ; 2 P2 ; ; n Pn ], . . . n 0 2 . . . 0 0 0 . . . 0 0 0 . . . n

Then

1 0 [AP1 ; AP2 ; ; APn ] = AP = P . . . 0

0 2 . . . 0

0 0 . . . 0

where the Pk are the columns of P. Since P is invertible, the Pk are linearly independent by virtue of Theorems 204 and 313. Conversely, suppose now that 1 , . . . , n are n linearly independent eigenvectors, with correv v sponding eigenvalues 1 , 2 , . . . , n . Put 1 0 D= . . . 0 0 2 . . . 0 0 0 . . . 0 0 0 . . . . n

P = [1 ; . . . ; n ], v v

Diagonalisability v Since Ai = ii we see that AP = PD. Again P is invertible by Theorems 204 and 313 since v are linearly independent. Left multiplying by P1 we deduce P1 AP = D, from where A the v k is diagonalisable. u 327 Example Shew that the following matrix is diagonalisable: 1 1 3 1 3 1 1 1 1

145

and nd a diagonal matrix D and an invertible matrix P such that A = PDP1 . Solution: Verify that the characteristic polynomial of A is 3 32 4 + 12 = ( 2)( + 2)( 3). The eigenvector for = 2 is 1 . 1 4 1 0 . 1 1 . 1 1 2 D= 0 0 0 2 0 0 1 , P = 0 1 3 1
1 5 1 5

The eigenvector for = 2 is

The eigenvector for = 3 is

We may take

1 0 1

We also nd

4 1 . 1

1 1 0

= 0

1 .
1 5

1 5

146

Chapter 7

Homework
Problem 7.3.1 Let A be a 22 matrix with eigenvalues 1 respectively. Determine A10 . 6. Find the eigenvalues of A. 7. Find the eigenvectors of A. 8. Find A10 . Problem 7.3.5 Consider the matrix 1 a 1 A = 0 1 b . 0 0 c 1 1 and 2 and corresponding eigenvectors and , 1 0

9 Problem 7.3.2 Consider the matrix A = 20

a 4. If A20 = c

1. Find the characteristic polynomial of A. 2. Find the eigenvalues of A. 3. Find the eigenvectors of A. b , nd a + d. d

4 . 9

Find the characteristic polynomial of A.

Explain whether A is diagonalisable when a = 0, c = 1. Explain whether A is diagonalisable when a = 0, c = 1. Explain whether A is diagonalisable when c = 1. Problem 7.3.6 Find a closed formula for An , if 7 A= 12 6 . 10

Problem 7.3.3 Let A M33 (R) have characteristic polynomial ( + 1)2 ( 3). 1 One of the eigenvalues has two eigenvectors 0 and 0 1 . The other eigenvalue has corresponding eigenvec1 0 1 tor 1. Determine A. 1 Problem 7.3.4 Let 0 0 A= 0 1 0 0 1 0 0 1 0 0 1 0 . 0 0

Problem 7.3.7 Let U Mnn (R) be a square matrix all whose entries are equal to 1. 1. Demonstrate that U2 = nU. 2. Find det U. 3. Prove that det(In U) = n1 ( n). 4. Shew that dim ker (U) = n 1. 5. Shew that n 0 U = P . . . 0 1 1 1 P= . . . 1 1 1 0 0 . . . 0 1 0 0 . . . 0 . . . 0 0 1 P , . . . 0 0 0 . . . . . . 0 1 0 0 . . . . . . . 1 1

where

0 1 0 . . . 0 1

.. . .

..

1. 2. 3. 4. 5.

Find Find Find Find Find

det A. A1 . rank (A I4 ). det(A I4 ). the characteristic polynomial of A.

Theorem of Cayley and Hamilton

147

7.4 Theorem of Cayley and Hamilton

328 Theorem (Cayley-Hamilton) A matrix A Mn (F) satises its characteristic polynomial. Proof: Put B = In A. We can write det B = det(In A) = n + b1 n1 + b2 n2 + + bn , as det(In A) is a polynomial of degree n. Since adj (B) is a matrix obtained by using (n 1) (n 1) determinants from B, we may write adj (B) = n1 Bn1 + n2 Bn2 + + B0 . Hence det(In A)In = (B)(adj (B)) = (In A)(adj (B)), from where n In + b1 In n1 + b2 In n2 + + bn In = (In A)(n1 Bn1 + n2 Bn2 + + B0 ). By equating coefcients we deduce In b1 In b2 In

= Bn1 = ABn1 + Bn2 = ABn2 + Bn3 . . .

bn1 In bn In

= AB1 + B0 = AB0 .

Multiply now the k-th row by Ank (the rst row appearing is really the 0-th row): An b1 An1 b2 An2 = = = . . . bn1 A bn In = = A2 B1 + AB0 AB0 . An Bn1 An Bn1 + An1 Bn2 An1 Bn2 + An2 Bn3

Add all the rows and through telescopic cancellation obtain An + b1 An1 + + bn1 A + bn In = 0n , from where the theorem follows. u

148 329 Example From example 327 the matrix 1 1 3 1 3 1 1 1 1

Chapter 7

has characteristic polynomial

( 3)( 2)( + 2) = 3 32 4 + 12, hence the inverse of this matrix can be obtained by observing that 1/6 1/3 = 1/6 1/3 5/6 1/6 1/6 1/6 . 1/3

A3 3A2 4A + 12I3 = 03 = A1 =

1 12

A2 3A 4I3

Homework
Problem 7.4.1 A 3 3 matrix A has characteristic polynomial ( 1)( + 2). What is the characteristic polynomial of A2 ?

Chapter

Linear Algebra and Geometry

8.1 Points and Bi-points in R2
a1 R2 is the set of all points A = with real number coordinates on the plane, as in gure 8.1. We use a2 0 the notation O = to denote the origin. 0 y

a1 A= a2 O x

Figure 8.1: Rectangular coordinates in R2 .

b1 a1 Given A = R2 and B = R2 we dene their addition as b2 a2 a1 b1 a1 + b1 . A+B = + = a2 + b2 b2 a2 149 (8.1)

150 Similarly, we dene the scalar multiplication of a point of R2 by the scalar R as a1 a1 . A = = a2 a2

Chapter 8

(8.2)

 Throughout this chapter, unless otherwise noted, we will use the convention that a point
A R2 will have its coordinates named after its letter, thus a1 A = . a2 330 Denition Consider the points A R2 , B R2 . By the bi-point starting at A and ending at B, denoted by [A, B], we mean the directed line segment from A to B. We dene 0 [A, A] = O = . 0 The bi-point [A, B] can be thus interpreted as an arrow starting at A and nishing, with the arrow tip, at B. We say that A is the tail of the bi-point [A, B] and that B is its head. Some authors use the terminology xed vector instead of bi-point. 331 Denition Let A = B be points on the plane and let L be the line passing through A and B. The direction of the bi-point [A, B] is the direction of the line L, that is, the angle ; that the line L 2 2 makes with the horizontal. See gure 8.2. 332 Denition Let A, B lie on line L, and let C, D lie on line L . If L||L then we say that [A, B] has the same direction as [C, D]. We say that the bi-points [A, B] and [C, D] have the same sense if they have the same direction and if both their heads lie on the same half-plane made by the line joining their tails. They have opposite sense if they have the same direction and if both their heads lie on alternative half-planes made by the line joining their tails. See gures 8.3 and 8.4 .

B A A

B D C D A

Figure 8.2: Direction of a bi-point

Figure 8.3: Bi-points with the same sense.

Figure 8.4: Bi-points with opposite sense.

Points and Bi-points in R2

151

 Bi-point [B, A] has the opposite sense of [A, B] and so we write

[B, A] = [A, B]. 333 Denition Let A = B. The Euclidean length or norm of bi-point [A, B] is simply the distance between A and B and it is denoted by ||[A, B]|| = We dene ||[A, A]|| = ||O|| = 0. A bi-point is said to have unit length if it has norm 1. A bi-point is completely determined by three things: (i) its norm, (ii) its direction, and (iii) its sense. 334 Denition (Chasles Rule) Two bi-points are said to be contiguous if one has as tail the head of the other. In such case we dene the sum of contiguous bi-points [A, B] and [B, C] by Chasles Rule [A, B] + [B, C] = [A, C]. See gure 8.5. 335 Denition (Scalar Multiplication of Bi-points) Let R \ {0} and A = B. We dene 0[A, B] = O and [A, A] = O. We dene [A, B] as follows. 1. [A, B] has the direction of [A, B]. 2. [A, B] has the sense of [A, B] if > 0 and sense opposite [A, B] if < 0. 3. [A, B] has norm ||||[A, B]|| which is a contraction of [A, B] if 0 < || < 1 or a dilatation of [A, B] if || > 1. See gure 8.6 for some examples. 2[A, B] (a1 b1 )2 + (a2 b2 )2 .

B A C 1 2

[A, B]

[A, B]

Figure 8.5: Chasles Rule.

Figure 8.6: points.

Scalar multiplication of bi-

152

Chapter 8

8.2 Vectors in R2
336 Denition (Midpoint) Let A, B be points in R2 . We dene the midpoint of the bi-point [A, B] as A+B 2
a +b 12 1 a2 +b2 2

337 Denition (Equipollence) Two bi-points [X, Y] and [A, B] are said to be equipollent written [X, Y] [A, B] if the midpoints of the bi-points [X, B] and [Y, A] coincide, that is, [X, Y] [A, B] X+B 2 = Y +A 2 .

See gure 8.7.

Geometrically, equipollence means that the quadrilateral XYBA is a parallelogram. Thus the bi-points [X, Y] and [A, B] have the same norm, sense, and direction. Y X || A Figure 8.7: Equipollent bi-points. || B

338 Lemma Two bi-points [X, Y] and [A, B] are equipollent if and only if y1 x1 b1 a1 . = b2 a2 y2 x2 Proof: This is immediate, since
a +y 12 1 a2 +y2 2

[X, Y] [A, B]

= b

b +x 12 1
2 +x2

y1 x1 b1 a1 , = b2 a2 y2 x2

as desired. u From Lemma 338, equipollent bi-points have the same norm, the same direction, and the same sense. 339 Theorem Equipollence is an equivalence relation.

Vectors in R2

153

Proof:

y1 x1 = Write [X, Y] [A, B] if [X, Y] if equipollent to [A, B]. Now [X, Y] [X, Y] since y2 x2

y1 x1 and so the relation is reexive. Also y2 x2 [X, Y] [A, B] y1 x1 = y2 x2 b1 a1 b2 a2

[A, B] [X, Y],

b1 a1 y1 x1 = y2 x2 b2 a2

and the relation is symmetric. Finally

[X, Y] [A, B] [A, B] [U, V]

y1 x1 b1 a1 = b2 a2 y2 x2

y1 x1 v1 u1 = y2 x2 v2 u2 [X, Y] [U, V],

b1 a1 v1 u1 = v2 au2 b2 a2

and the relation is transitive. u

340 Denition (Vectors on the Plane) The equivalence class in which the bi-point [X, Y] falls is called the vector (or free vector) from X to Y, and is denoted by XY. Thus we write y1 x1 . [X, Y] XY = y2 x2

If we desire to talk about a vector without mentioning a bi-point representative, we write, say, , thus v denoting vectors with boldface lowercase letters. If it is necessary to mention the coordinates of we v will write v = 1 . v v2

154

Chapter 8  For any point X on the plane, we have = , the zero vector. If [X, Y] then [Y, X] XX 0 v

v.

p1 . We call the position vector of P and we use boldface lowercase letters to denote the equality OP p2 OP = . p 1 3 342 Example The vector into which the bi-point with tail at A = and head at B = falls is 2 4 3 (1) 4 = . AB = 42 2 343 Example The bi-points [A, B] and [X, Y] with 3 1 A = ,B = , 4 2 3 7 X = ,Y = 7 9 3 (1) 4 7 3 = XY. = = AB = 97 2 42

p1 341 Denition (Position Vector) For any particular point P = R2 we may form the vector OP = p2

represent the same vector

1 + n 3 + n ,T = then the innite number of bi-points [S, T ] are representatives of In fact, if S = 2+m 4+m of the vectors AB = XY = ST .

Given two vectors , we dene their sum + as follows. Find a bi-point representative AB u v u v u . Then by Chasles Rule and a contiguous bi-point representative BC v + = + = AB BC AC. u v

Again, by virtue of Chasles Rule we then have

AB = AO + OB = OA + OB = b a

(8.3)

Vectors in R2

155

Similarly we dene scalar multiplication of a vector by scaling one of its bi-point representatives.We dene the norm of a vector R2 to be the norm of any of its bi-point representatives. v sum and scalar multiplication take the form v1 u1 Componentwise we may see that given vectors = , = , and a scalar R then their u v v2 u2 v u + = 1 + 1 , u v v2 u2 u 1 . = u u2

+ u v

B v

2 u u 1 u 2

Figure 8.8: Addition of Vectors.

Figure 8.9: Scalar multiplication of vectors.

344 Example Diagonals are drawn in a rectangle ABCD. If AB = and AC = , then BC = , x y y x , DA = , and BD = 2. CD = x x y y x

345 Denition (Parallel Vectors) Two vectors and are said to be parallel if there is a scalar such u v = . If is parallel to we write ||. We denote by R = { : R}, the set of all vectors that u v u v u v v v . parallel to v  is parallel to every vector. 0

u1 u 346 Denition If = , then we dene its norm as u u2 and is d , = . vectors u v u v u v v Solution: Observe that v v Hence the vector a v

u2 + u2 . The distance between two 1 2

v v 347 Example Let a R, a > 0 and let = 0 . Find a vector with norm a and parallel to . has norm 1 as v v v = v

= 1.

v has norm a and it is in the direction of . One may also take a . v v

156 348 Example If M is the midpoint of the bi-point [X, Y] then XM = MY from where XM = if T is any point, by Chasles Rule TX + TY = TM + MX + TM + MY = 2TM. = 2TM XM + MY 1 XY . 2

Chapter 8 Moreover,

349 Example Let ABC be a triangle on the plane. Prove that the line joining the midpoints of two sides of the triangle is parallel to the third side and measures half its length. Solution: Let the midpoints of [A, B] and [A, C] be MC and MB , respectively. We shew that BC = 2MC MB . We have 2AMC = AB and 2AMB = AC. Thus BC = BA + AC = AB + AC

= 2AMC + 2AMB = 2(MC A + AMB ) = 2MC A + 2AMB

as we wanted to shew.

= 2MC MB ,

350 Example In ABC, let MC be the midpoint of side AB. Shew that 1 CA + CB . CMC = 2 Solution: Since AMC = MC B, we have

CA + CB = CMC + MC A + CMC + MC B = 2CMC AMC + MC B = 2CMC ,

which yields the desired result.

351 Theorem (Section Formula) Let APB be a straight line and and be real numbers such that ||[A, P]|| = . ||[P, B]|| With = OA, b = OB, and = OP, then a p = b + a . p +

(8.4)

Vectors in R2 Proof: Using Chasles Rule for vectors, AB = AO + OB = + b , a Also, using Chasles Rule for bi-points, AP = AO + OP = + . a p

157

[A, P] = ([P, B]) = ([P, A] + [A, B]) = ([A, P] + [A, B]), whence [A, P] = [A, B] = AP = ( b ). a AB = = p a + + +

On combining these formul

from where the result follows. u a b

Figure 8.10: 8.2.6.

( + )( ) = ( b ) = ( + ) = b + , p a a p a a a

c
[A]. Problem

c
Figure 8.11: 8.2.6.

b
[B]. Problem

Figure 8.12: 8.2.6.

[C]. Problem

Figure 8.13: 8.2.6.

[D]. Problem

Figure 8.14: 8.2.6.

b f

c e

[E]. Problem

Figure 8.15: 8.2.6.

[F]. Problem

Homework
Problem 8.2.1 Let a be a real number. Find the dis 1 a 1 . tance between and 1 a 1 where = . v 1 Problem 8.2.3 Given a pentagon ABCDE, nd AB + BC + CD + DE + EA. Problem 8.2.2 Find all scalars for which v =
1 , 2

158
Problem 8.2.4 For which values of a will the vectors will be parallel? a+1 = , a 2 a 1 2a + 5 b = 2 a 4a + 3

Chapter 8
Problem 8.2.9 ABCD is a parallelogram. E is the midpoint of [B, C] and F is the midpoint of [D, C]. Prove that AC + BD = 2BC. Problem 8.2.10 Let A, B be two points on the plane. Construct two points I and J such that IA = 3IB, 1 JA = JB, 3

Problem 8.2.5 In ABC let the midpoints of [A, B] and [A, C] be MC and MB , respectively. Put MC B = , x , and CA = . Express [A] AB + BC + M , MB C = y z C MB [B] AMC + MC MB + MB C, [C] AC + MC A BMB in , , and . terms of x y z Problem 8.2.6 A circle is divided into three, four equal, or six equal parts (gures 8.10 through 8.15). Find the sum of the vectors. Assume that the divisions start or stop at the centre of the circle, as suggested in the gures. Problem 8.2.7 Diagonals are drawn in a square (gures ?? through ??). Find the vectorial sum + b + . a c Assume that the diagonals either start, stop, or pass through the centre of the square, as suggested by the gures.

and then demonstrate that for any arbitrary point M on the plane MA + 3MB = 4MI and 3MA + MB = 4MJ. Problem 8.2.11 You nd an ancient treasure map in your great-grandfathers sea-chest. The sketch indicates that from the gallows you should walk to the oak tree, turn right 90 and walk a like distance, putting and x at the point where you stop; then go back to the gallows, walk to the pine tree, turn left 90 , walk the same distance, mark point Y. Then you will nd the treasure at the midpoint of the segment XY. So you charter a sailing vessel and go to the remote south-seas island. On arrival, you readily locate the oak and pine trees, but unfortunately, the gallows was struck by lightning, burned to dust and dispersed to the winds. No trace of it remains. What do you do?

Problem 8.2.8 Prove that the mid-points of the sides of a skew quadrilateral form the vertices of a parallelogram.

8.3 Dot Product in R2

352 Denition Let (, b) (R2 )2 . The dot product b of and b is dened by a a a b a = 1 1 = a b + a b . a b 1 1 2 2 b2 a2

The following properties of the dot product are easy to deduce from the denition. DP1 Bilinearity DP2 Scalar Homogeneity DP3 Commutativity DP4 DP5 y z x z ( + ) = + , x y z ( + ) = + x z x y z x y (8.5) (8.6) (8.7) (8.8) (8.9)

() = () = ( x y ), R. x y x y = x y y x 0 x x

= 0 = x x x 0

Dot Product in R2 DP6 ||x|| = 353 Example If we put

159

x x

(8.10)

a 1 then we can write any vector = as a sum a a2 The vectors

1 0 i = , j = , 0 1

= a + a . a 1 i 2 j

satisfy i j = 0, and i = j = 1.

1 0 i = , j = , 1 0

355 Theorem

354 Denition Given vectors and b, we dene the angle between them, denoted by (, b), as the angle a a between any two contiguous bi-point representatives of and b . a a b = |||||| b || cos (, b). a a Using Al-Kashis Law of Cosines on the length of the vectors, we have || b ||2 = ||||2 + || b||2 2|||||| b|| cos (, b ) a a a a

Proof:

b b 2 b + = ||||2 + || b||2 2|||||| b|| cos (, b ) a a a a a a b = |||||| b || cos (, b ), a a a

( b )( b ) = ||||2 + || b||2 2|||||| b|| cos (, b ) a a a a a

|| b ||2 2 a b + || b ||2 = ||||2 + || b ||2 2|||||| b|| cos (, b) a a a

as we wanted to shew. u Putting (, b ) = a

2

in Theorem 355 we obtain the following corollary.

356 Corollary Two vectors in R2 are perpendicular if and only if their dot product is 0.

160

b a b

Chapter 8

Figure 8.19: Theorem 355.

358 Denition If b and = b = 1 we say that and b are orthonormal. a a a

357 Denition Two vectors are said to be orthogonal if they are perpendicular. If is orthogonal to b , we a write b. a  It follows that the vector is simultaneously parallel and perpendicular to any vector! 0 = { R2 : }. a x x a

359 Denition Let R2 be xed. Then the orthogonal space to is dened and denoted by a a Since | cos | 1 we also have

360 Corollary (Cauchy-Bunyakovsky-Schwarz Inequality) a b a b .

361 Corollary (Triangle Inequality)

Proof:

+ + . a b a b = = ( + b )( + b ) a a

|| + b ||2 a

||||2 + 2|||||| b|| + || b||2 a a = from where the desired result follows. u (|||| + || b ||)2 , a

+ 2 + a a a b b b

362 Corollary (Pythagorean Theorem) If b then a

|| + b ||2 = a a

Proof:

Since b = 0, we have a

|| + b||2 a

= + 2 b + b b a a a = = + 0 + b b a a ||||2 + || b ||2 , a

( + b )( + b) a a

Dot Product in R2 from where the desired result follows. u 363 Denition The projection of t onto (or the -component of t ) is the vector v v
t proj = (cos ( t , )) t v v

161

where (, t ) [0; ] is the convex angle between and t read in the positive sense. v v

1 v, v

common tail and join them together at their tails. The projection of t onto v is the shadow of t t in the direction of . To obtain proj we prolong if necessary and drop a perpendicular line v v v to it from the head of t . The projection is the portion between the common tails of the vectors and the point where this perpendicular meets t . See gure 8.20.

 Given two vectors and vector = , nd bi-point representatives of them having a t v 0

Figure 8.20: Vector Projections. 364 Corollary Let = 0 . Then a

x proj = cos (, ) x a x a

365 Theorem Let R2 \ { 0 }. Then any R2 can be decomposed as a x where R and . u a v a = + , x u v

1 = x a . a a 2 a a

x x Proof: We know that proj is parallel to , so we take = proj . This means that we must a u a a = proj . We must demonstrate that is indeed perpendicular to . But this x then take v x v a a is clear, as = proj x a a x a v

= a x = 0,

|| || x a = a x

a x a a a 2 a

completing the proof. u 366 Corollary Let be non-zero vectors in R2 . Then any vector R2 has a unique representation v w a , , as a linear combination of v w = s + t (s, t) R2 . a v w,

162 Proof: By Theorem 365, there exists a decomposition = s + s , v a v = s + t a v w.

Chapter 8

where is orthogonal to . But then || and hence there exists R with = Taking v v v w v w. t = s we achieve the decomposition To prove uniqueness, assume

367 Corollary Let , be non-zero, non-parallel vectors in R2 . Then any vector R2 has a unique p q a , , representation as a linear combination of p q = l + m, (l, m) R2 . a p q

Then (s p) = (q t) We must have s = p and q = t since otherwise would be parallel v w. v This completes the proof. u to w.

s + t = = p + q v w a v w.

Proof:

(s, t) R2 such that

q Consider = proj . Clearly and so by Corollary 366, there exists unique z q p z p

= a = =

q s tproj + t p q p

s + t p z st

p ||||2

q p

establishing the result upon choosing l = s t

p ||||2

q p

and m = t. u

+ t, p q

1 1 368 Example Let = , = . Write as the sum of two vectors, one parallel to and the other p q p q 1 2 . perpendicular to q
p Solution: We use Theorem 365. We know that proj is parallel to , and we nd q q p proj q 3 p q 3 5 = = . = 2q q 5 q 6 5

We also compute

Observe that

1 p proj = = . p q 6 1 1 5 5
3 5

3 5

2 5

= 6 6 = 0, 25 25 6 1 5 5

2 5

Dot Product in R2 and the desired decomposition is 1 = + . 6 1 1 5 5 C B H

3 5

163

2 5

A
Figure 8.21: Orthocentre.

369 Example Prove that the altitudes of a triangle ABC on the plane are concurrent. This point is called the orthocentre of the triangle. Solution: Put = OA, b = OB, = OC. First observe that for any , we have, upon a c x expanding, ( )( b ) + ( b)( ) + ( )( b ) = 0. x a c x c a x c a (8.11) Let H be the point of intersection of the altitude from A and the altitude from B. Then 0 = AHCB = (OH OA)(OB OC) = (OH )( b ), a c (8.12) (8.13)

and

Putting = OH in (8.11) and subtracting from it (8.12) and (8.13), we gather that x which gives the result. 0 = (OH )( b ) = CHAB, c a

0 = BHAC = (OH OB)(OC OA) = (OH b)( ). c a

Homework
a Problem 8.3.1 Determine the value of a so that 1a 1 be perpendicular to . 1 as the sum of two vectors, one parallel to and the p r other parallel to . s Problem 8.3.4 Prove that a
2

= ( i )2 + ( j )2 . a a

Problem 8.3.2 Demonstrate that a ( b + = 0 ) ( = b ) ( b )( ) = 0. c a a c 2 1 4 Problem 8.3.3 Let = , = , = . Write p r s 1 1 5

Problem 8.3.5 Let = 0 = b be vectors in R2 such that a = 0. Prove that a b + b = 0 = = = 0. a

3 Problem 8.3.6 Let (, ) (R2 )2 with |||| = 2 ||||. x y x y + 3 and 2 3 are perpendicular. Shew that 2 x y x y

164
Problem 8.3.7 Let , b be xed vectors in R2 . Prove a that if v R2 , = b , v v a then = b . a Problem 8.3.8 Let (, b ) (R2 )2 . Prove that a + a b
2

Chapter 8
Problem 8.3.10 Let , be non-zero vectors in R2 . x a Prove that proj
a proj x

a + b

a =2

+2 b

with 0 1.

= , a

Problem 8.3.9 Let , be vectors in R2 . Prove the pou v larisation identity: = 1 || + ||2 || ||2 . u v u v u v 4

Problem 8.3.11 Let (, ) R R2 be xed. Solve the a equation = a x R2 . for x

8.4 Lines on the Plane

370 Denition Three points A, B, and C are collinear if they lie on the same line. It is clear that the points A, B, and C are collinear if and only if AB is parallel to AC. Thus we have the following denition. 371 Denition The parametric equation with parameter t R of the straight line passing through the p1 v point P = in the direction of the vector = 0 is p2

x p1 = t. v y p2 x If = , then the equation of the line can be written in the form r y = t. r p v (8.14)

The Cartesian equation of a line is an equation of the form ax + by = c, where a2 + b2 = 0. We write (AB) for the line passing through the points A and B. 372 Theorem Let = 0 and let . An alternative form for the equation of the line = t is v n v r p v a Moreover, the vector is perpendicular to the line with Cartesian equation ax + by = c. b Proof: The rst part follows at once by observing that = 0 and taking dot products to both v n sides of 8.14. For the second part observe that at least one of a and b is = 0. First assume that ( ) = 0. r p n

Lines on the Plane

b c a = 0. Then we can put y = t and x = a t + a and the parametric equation of this line is

165

and we have

x y

c a

b = t a , 1

b a

c Similarly if b = 0 we can put x = t and y = a t + b and the parametric equation of this line is b

a = b a + b = 0. a b 1 1 = t , c a y b b x

and we have proving the theorem in this case. u

1 a = a a b = 0, b b a b

a2 +b2 has norm 1 and is orthogonal to the line ax + by = c. The vector

b a2 +b2

4 2 373 Example The equation of the line passing through A = and in the direction of = is v 3 5 4 x 2 = . 5 y3 1 2 374 Example Find the equation of the line passing through A = and B = . 1 3 Solution: The direction of this line is that of

The equation is thus

2 (1) 1 = . AB = 31 2 1 x + 1 = , R. 2 y1

166

Chapter 8

375 Example Suppose that (m, b) R2 . Write the Cartesian equation of the line y = mx+b in parametric form. Solution: Here is a way. Put x = t. Then y = mt + b and so the desired parametric form is x 1 = t . yb m

376 Example Let (m1 , m2 , b1 , b2 ) R4 , m1 m2 = 0. Consider the lines L1 : y = m1 x + b1 and L2 : y = m2 x + b2 . By translating this problem in the language of vectors in R2 , shew that L1 L2 if and only if m1 m2 = 1. Solution: The parametric equations of the lines are 1 x , = s L1 : m1 y b1 L2 :

1 . = t m2 y b2 x

yields

1 1 and = . Since the lines are perpendicular we must have = 0, which Put = v w v w m1 m2 0 = = 1(1) + m1 (m2 ) = m1 m2 = 1. v w

377 Theorem (Distance Between a Point and a Line) Let ( ) = 0 be a line passing through the r a n . If B is not a point on the line, then the distance from B to the point A and perpendicular to vector n line is n ( b ) a . n If the line has Cartesian equation ax + by = c, then this distance is |ab1 + bb2 c| . a2 + b2 Proof: Let R0 be the point on the line that is nearest to B. Then BR0 = b is orthogonal to r0 the line, and the distance we seek is ( b) r0 n |( b)| r0 n r0 b ||proj || = n = . n 2 n n n a n Since R0 is on the line, = , and so r0
r0 ||proj b || = n

as we wanted to shew.

n n a n n a n | b | | b | |( b )| r0 = = , | n n n

Lines on the Plane If the line has Cartesian equation ax + by = c, then at least one of a and b is = 0. Let us suppose a = 0, as the argument when a = 0 and b = 0 is similar. Then ax + by = c is equivalent to
c x a a = 0. y b 0

167

b1 a n We use the result obtained above with = , = , and B = . Then a n b 0 b2 a2 + b2 and

c b1 a | = a = |c ab1 bb2 | = |ab1 + bb2 c|, |( a b) n b b2

c a

giving the result. u

378 Example Recall that the medians of ABC are lines joining the vertices of ABC with the midpoints of the side opposite the vertex. Prove that the medians of a triangle are concurrent, that is, that they pass through a common point.

 This point of concurrency is called, alternatively, the isobarycentre, centroid, or centre of

gravity of the triangle. Solution: Let MA , MB , and MC denote the midpoints of the lines opposite A, B, and C, respectively. The equation of the line passing through A and in the direction of AMA is (with = + r . OA r AMA = + s . OB r BMA

x = ) r y

Similarly, the equation of the line passing through B and in the direction of BMB is These two lines must intersect at a point G inside the triangle. We will shew that GC is parallel to CMC , which means that the three points G, C, MC are collinear. Now, (r0 , s0 ) R2 such that OA + r0 AMA = OG = OB + s0 BMB , r0 AMA s0 BMB = OB OA,

that is or

Since MA and MB are the midpoints of [B, C] and [C, A] respectively, we have 2BMA = BC and 2AMB = AC = AB + BC. The relationship becomes 1 1 1 r0 (AB + BC) s0 (AB + AB + BC) = AB, 2 2 2

r0 (AB + BMA ) s0 (BA + AMB ) = AB.

168 (r0 + We must have s0 2 r0 s0 1)AB = ( + )BC. 2 2 r0 + r0 2 s0 2 + 1 = 0, s0 2 = 0,

Chapter 8

since otherwise the vectors AB and BC would be parallel, and the triangle would be degenerate. 2 Solving, we nd s0 = r0 = 2 . Thus we have OA + 2 AMA = OG, or AG = 3 AMA , and similarly, 3 3 BG = 2 BMB . 3

From AG = 2 AMA , we deduce AG = 2GMA . Since MA is the midpoint of [B, C], we have 3 GB + GC = 2GMA = AG, which is equivalent to GA + GB + GC = 0 . As MC is the midpoint of [A, B] we have GA + GB = 2GMC . Thus 0 = GA + GB + GC = 2GMC + GC.

This means that GC = 2GMC , that is, that they are parallel, and so the points G, C and MC all lie on the same line. This achieves the desired result.

 The centroid of ABC satises thus

and divides the medians on the ratio 2 : 1, reckoning from a vertex.

GA + GB + GC = 0 ,

Homework
Problem 8.4.1 Find the angle between the lines 2x y = 1 and x 3y = 1. Problem 8.4.2 Find the equation of the line passing 2 1 through and in a direction perpendicular to . 1 1 Prove that E, I, J are collinear. Problem 8.4.5 Let ABCD be a parallelogram. Let E and F be points such that 1 AE = AC 4 3 and AF = AC. 4

Demonstrate that the lines (BE) and (DF) are parallel. Let I be the midpoint of [A, D] and J be the midpoint of [B, C]. Demonstrate that the lines (AB) and (IJ) are parallel. What type of quadrilateral is IEJF? Problem 8.4.6 ABCD is a parallelogram; point I is the midpoint of [A, B]. Point E is dened by the relation 1 IE = 3 ID. Prove that 1 AB + AD AE = 3

Problem 8.4.3 ABC has centroid G, and A B C satises AA + BB + CC = 0 . Prove that G is also the centroid of A B C . Problem 8.4.4 Let ABCD be a trapezoid, with bases [A, B] and [C, D]. The lines (AC) and (BD) meet at E and the lines (AD) and (BC) meet at F. Prove that the line (EF) passes through the midpoints of [A, B] and [C, D] by proving the following steps. Let I be the midpoint of [A, B] and let J be the point of intersection of the lines (FI) and (DC). Prove that J is the midpoint of [C, D]. Deduce that F, I, J are collinear.

and prove that the points A, C, E are collinear. Problem 8.4.7 Put OA = , OB = b , OC = . Prove a c that A, B, C are collinear if and only if there exist real

Vectors in R3
numbers , , , not all zero, such that + b + = 0 , + + = 0. a c Problem 8.4.8 Prove Desargues Theorem: If ABC

169
and A B C (not necessarily in the same plane) are so positioned that (AA ), (BB ), (CC ) all pass through the same point V and if (BC) and (B C ) meet at L, (CA) and (C A ) meet at M, and (AB) and (A B ) meet at N, then L, M, N are collinear.

8.5 Vectors in R3
We now extend the notions studied for R2 to R3 . The rectangular coordinate form of a vector in R3 is a1 = . a a2 a3 In particular, if 0 0 1 i = 0 , j = 1 , k = 0 1 0 0

a1 = as a sum then we can write any vector a a2 a3

b1 a1 = and = , their dot product is b b2 Given a a2 b3 a3

= a + a + a . a 1 i 2 j 3k

and We also have and

= a b + a b + a b , a b 1 1 2 2 3 3 = a a2 + a2 + a2 . 1 2 3

i j = j k = k i = 0,

379 Denition A system of unit vectors i , j , k is right-handed if the shortest-route rotation which brings i to coincide with j is performed in a counter-clockwise manner. It is left-handed if the rotation is done in a clockwise manner.

i = j = k = 1.

170

Chapter 8

To study points in space we must rst agree on the orientation that we will give our coordinate system. We will use, unless otherwise noted, a right-handed orientation, as in gure 8.22. k k j i j i

Figure 8.22: Right-handed system.

Figure 8.23: Left-handed system.

The analogues of the Cauchy-Bunyakovsky-Schwarz and the Triangle Inequality also hold in R3 .

We now dene the (standard) cross (wedge) product in R3 as a product satisfying the following properties. 380 Denition Let (, , , ) R3 R3 R3 R. The wedge product : R3 R3 R3 is a closed x y z binary operation satisfying CP1 Anti-commutativity: = () x y y x ( + ) = + , x z y x y z y ( + ) = + x z y x z x y (8.15)

CP2 Bilinearity:

(8.16)

CP3 Scalar homogeneity:

CP4

() = () = () x y x y x y = x x 0 i j = k, j k = i , k i = j

(8.17)

(8.18)

CP5 Right-hand Rule:

(8.19)

y1 x1 = and = be vectors in R3 . Then y y2 381 Theorem Let x x2 y3 x3 = (x y x y ) + (x y x y ) + (x y x y ). x y 2 3 3 2 i 3 1 1 3 j 1 2 2 1 k

Vectors in R3 Proof: Since i i = j j = k k = 0 we have

171

(x1 i + x2 j + x3 k )(y1 i + y2 j + y3 k ) = x1 y2 i j + x1 y3 i k

= x1 y2 k x1 y3 j x2 y1 k from where the theorem follows. u 382 Example Find 1 0 . 0 1 3 2 Solution: We have ( i 3 k )( j + 2 k ) = i j + 2 i k 3 k j 6 k k = = 3 i 2 j + k . 1 0 3 = . 0 1 2 3 2 1 k 2 j 3 i + 60

+x3 y1 k i + x3 y2 k j

+x2 y1 j i + x2 y3 j k

+x2 y3 i + x3 y1 j x3 y2 i ,

Hence

383 Theorem The cross product vector is simultaneously perpendicular to and . x y x y Proof: We will only check the rst assertion, the second verication is analogous. () = x x y = = completing the proof. u (x1 i + x2 j + x3 k )((x2 y3 x3 y2 ) i

x1 x2 y3 x1 x3 y2 + x2 x3 y1 x2 x1 y3 + x3 x1 y2 x3 x2 y1 0,

+(x3 y1 x1 y3 ) j + (x1 y2 x2 y1 ) k )

172 c c 384 Theorem ( b ) = ( a ) b ( a b ). a c Proof: ( ) = (a + a + a )((b c b c )+ a b c 1 i 2 j 3k 2 3 3 2 i = a1 (b3 c1 b1 c3 ) k a1 (b1 c2 b2 c1 ) j

Chapter 8

+(b3 c1 b1 c3 ) j + (b1 c2 b2 c1 ) k ) a2 (b2 c3 b3 c2 ) k + a2 (b1 c2 b2 c1 ) i

= (a1 c1 + a2 c2 + a3 c3 )(b1 i + b2 j + b3 i ) = ( a ) b ( a b ), c c

+a3 (b2 c3 b3 c2 ) j a3 (b3 c1 b1 c3 ) i

+(a1 b1 a2 b2 a3 b3 )(c1 i + c2 j + c3 i )

completing the proof. u 385 Theorem (Jacobis Identity)

Proof: From Theorem 384 we have ( ) = () (), a b c a c b a b c b( c a ) = ( b ) ( b ), a c c a () = () (), c a b c b a c a b and adding yields the result. u

( ) + () + () = . a b c b c a c a b 0

x y x y 386 Theorem Let (, ) [0; [ be the convex angle between two vectors and . Then |||| = |||||||| sin (, ). x y x y x y

Proof: We have ||||2 x y = (x2 y3 x3 y2 )2 + (x3 y1 x1 y3 )2 + (x1 y2 x2 y1 )2 = x2 y2 2x2 y3 x3 y2 + x2 y2 + x2 y2 2x3 y1 x1 y3 + 2 3 3 2 3 1 +x2 y2 + x2 y2 2x1 y2 x2 y1 + x2 y2 1 3 1 2 2 1 = (x2 + x2 + x2 )(y2 + y2 + y2 ) (x1 y1 + x2 y2 + x3 y3 )2 1 2 3 1 2 3 = ||||2 || y ||2 ( x y )2 x

= ||||2 || y ||2 sin2 (, ), x x y

x y = ||||2 || y ||2 ||||2 || y ||2 cos2 (, ) x x

Vectors in R3 whence the theorem follows. The Theorem is illustrated in Figure 8.24. Geometrically it means that the area of the parallelogram generated by joining and at their heads is . u x y x y
x y

173

Figure 8.24: Theorem 386.

The following corollaries are now obvious. 387 Corollary Two non-zero vectors , satisfy = 0 if and only if they are parallel. x y x y 388 Corollary (Lagranges Identity) 2 2 || y ||2 = ||x|| ||y|| ( x y )2 . x

389 Example Let R3 , ||x|| = 1. Find x

Solution: By Lagranges Identity,

|| i ||2 + || j ||2 + || k ||2 . x x x

2

|| i ||2 = x x

|| k ||2 = x x

Problem 8.5.1 Consider a tetrahedron ABCS. [A] Find AB + BC + CS. [B] Find AC + CS + SA + AB. Problem 8.5.2 Find a vector simultaneously perpendic 1 1 ular to 1 and 1 and having norm 3. 0 1

and since ( i )2 + ( j )2 + ( k )2 = x x x x

|| j ||2 = x x

( i )2 = 1 ( x i )2 , x

( j )2 = 1 ( x j )2 , x

( k )2 = 1 ( k )2 , x x

= 1, the desired sum equals 3 1 = 2.

Problem 8.5.4 Prove or disprove! The cross product is associative. Problem 8.5.5 Prove that = 0 follows from the x x anti-commutativity of the cross product. Problem 8.5.6 Expand the product ( b )( + b ). a a a Problem 8.5.7 The vectors , b are constant vectors. () = (). Solve the equation a x b b x a Problem 8.5.8 The vectors , b , are constant veca c tors. Solve the system of equations 2 + = b , x y a 3 + = , y x a c

Problem 8.5.3 Find the area of the triangle whose ver 1 0 0 tices are at P = 0, Q = 1, and R = 0. 0 0 1

174
Problem 8.5.9 Prove that there do not exist three unit vectors in R3 such that the angle between any two of 2 . them be > 3 Problem 8.5.10 Let R3 be a xed vector. Demona strate that X = { R3 : = 0 } x a x is a subspace of R3 . Problem 8.5.11 Let (, b ) (R3 )2 and assume that a a b = 0 and that and b are linearly independent. a Prove that , b , b are linearly independent. a a

Chapter 8
Problem 8.5.12 Let (, b ) R3 R3 be xed. Solve a the equation = , a x b for . x Problem 8.5.13 Let h , k be xed vectors in R3 . Prove that L: R3 R3 (, ) x y R3

is a linear transformation.

k + h x y

8.6 Planes and Lines in R3

390 Denition If bi-point representatives of a family of vectors in R3 lie on the same plane, we will say that the vectors are coplanar or parallel to the plane. 391 Lemma Let , in R3 be non-parallel vectors. Then every vector of the form v w u ((a, b) R2 arbitrary) is coplanar with both and Conversely, any vector t coplanar with both v w. v can be uniquely expressed in the form and w t = p + q v w. = a + b u v w,

Proof: This follows at once from Corollary 367, since the operations occur on a plane, which can be identied with R2 . u

x points on the same plane and that R = y is another arbitrary point on this plane. Since A, B, and C z are non-collinear, AB and AC, which are coplanar, are non-parallel. Since AR also lies on the plane, we have by Lemma 391, that there exist real numbers p, q with AR = pAB + qAC. OR = OA + p(OB OA) + q(OC OA),

A plane is determined by three non-collinear points. Suppose that A, B, and C are non-collinear

By Chasles Rule,

is the equation of a plane containing the three non-collinear points A, B, and C. By letting = OR, r = OA, etc., we deduce that a = p( ) + q( ). r a b a c a Thus we have the following denition.

Planes and Lines in R3

175

Componentwise this takes the form

392 Denition The parametric equation of a plane containing the point A, and parallel to the vectors u is given by and v = p + q. r a u v x a1 = pu1 + qv1 , y a2 = pu2 + qv2 , z a3 = pu3 + qv3 .

The Cartesian equation of a plane is an equation of the form ax + by + cz = d with (a, b, c, d) R4 and a2 + b2 + c2 = 0. 393 Example Find both the parametric equation and the Cartesian equation of the plane parallel to the

1 1 0 and and passing through the point . vectors 1 1 1 1 0 2 Solution: The desired parametric equation is

x 1 1 = s + t . y + 1 1 1 z2 1 0 This gives s = z 2, t = y + 1 s = y + 1 z + 2 = y z + 3 and x = s + t = z 2 + y z + 3 = y + 1. Hence the Cartesian equation is x y = 1. 394 Theorem Let and be non-parallel vectors and let = p + q be the equation of the plane u v r a u v and . If is simultaneously perpendicular to and then containing A an parallel to the vectors u v n u v ( ) = 0. r a n

a Moreover, the vector b is normal to the plane with Cartesian equation ax + by + cz = d. c Proof: The rst part is clear, as = 0 = . For the second part, recall that at least one u n v n of a, b, c is non-zero. Let us assume a = 0. The argument is similar if one of the other letters is non-zero and a = 0. In this case we can see that x= b c d y z. a a a

176 Put y = s and z = t. Then x y z

d a

Chapter 8

is a parametric equation for the plane. u

= s +t 1 0 0 1

b a

c a

395 Example Find once again, by appealing to Theorem 394, the Cartesian equation of the plane parallel 1 1 0 to the vectors 1 and 1 and passing through the point 1. 1 0 2

1 1 1 Solution: The vector 1 1 = 1 is normal to the plane. The plane has thus equation 1 0 0 x 1 y + 1 1 = 0 = x + y + 1 = 0 = x y = 1, z2 0

as obtained before.

396 Theorem (Distance Between a Point and a Plane) Let ( ) = 0 be a plane passing through the r a n . If B is not a point on the plane, then the distance from B to the point A and perpendicular to vector n plane is n ( b ) a . n Proof: Let R0 be the point on the plane that is nearest to B. Then BR0 = b is orthogonal r0 to the plane, and the distance we seek is ( b) r0 n |( b)| r0 n r0 b ||proj || = n = . n 2 n n n a n Since R0 is on the plane, = , and so r0 n a n n a n | b | | b | |( b )| r0 n r0 b ||proj = = , || = | n n n n

as we wanted to shew. u

Given three planes in space, they may (i) be parallel (which allows for some of them to coincide), (ii) two may be parallel and the third intersect each of the other two at a line, (iii) intersect at a line, (iv) intersect at a point.

Planes and Lines in R3 397 Denition The equation of a line passing through A R3 in the direction of = 0 is given by v 398 Theorem Put OA = , OB = b , and OC = . Points (A, B, C) (R3 )3 are collinear if and only if a c + + = . a b b c c a 0 = t, t R. r a v

177

Rearranging, gives

Proof: If the points A, B, C are collinear, then AB is parallel to AC and by Corollary 387, we must have ( )( b ) = 0 . c a a = . c b c a a b 0

Further rearranging completes the proof. u

399 Theorem (Distance Between a Point and a Line) Let L : = + , = 0 , be a line and let B be a r a v v point not on L. Then the distance from B to L is given by ||( b)|| a v . v

Proof: If R0 with position vector is the point on L that is at shortest distance from B then r0 BR0 is perpendicular to the line, and so The distance we must compute is BR0 = || b||, which is then given by r0 || = ||BR0 v || = ||(r0 b ) v || . ||r0 b v v ( b) = ( b ), r0 v a v = ||( a b ) v || , ||r0 b|| v ||BR0 || = ||BR0 || sin = ||BR0 || . v v v 2

a Now, since R0 is on the line t0 R such that = + t0. Hence r0 v giving

proving the theorem. u

 Given two lines in space, one of the following three situations might arise:
Homework
Problem 8.6.1 Find the equation of the plane passing through the points (a, 0, a), (a, 1, 0), and (0, 1, 2a) in R3 .

(i) the lines intersect at a point, (ii) the lines are parallel, (iii) the lines are skew (one over the other, without intersecting).

Problem 8.6.2 Find the equation of plane containing the point (1, 1, 1) and perpendicular to the line x = 1 + t, y = 2t, z = 1 t.

178
Problem 8.6.3 Find the equation of plane containing the point (1, 1, 1) and containing the line x = 2y = 3z. Problem 8.6.4 Find the equation of the plane perpendicular to the line ax = by = cz, abc = 0 and passing through the point (1, 1, 1) in R3 . Problem 8.6.5 Find the equation of the line perpendicular to the plane ax + a2 y + a3 z = 0, a = 0 and passing through the point (0, 0, 1). Problem 8.6.6 The two planes x y z = 1, x z = 1, x 1 1 = + t .. 0 y 2 1 0 z

Chapter 8

Problem 8.6.8 Points a, b, c in R3 are collinear and it is known that = i 2 j and b = 2 k 3 i . Find a c a b c . Problem 8.6.9 Find the equation of the plane which is 1 3 and . equidistant of the points 2 1 1 1

intersect at a line. Write the equation of this line in the form x = + t, t R. v y a z

Problem 8.6.10 (Putnam Exam, 1980) Let S be the solid in three-dimensional space consisting of all points (x, y, z) satisfying the following system of six conditions: x 0, y 0, z 0,

Problem 8.6.7 Find the equation of the plane passing 1 through the points 0 , 1 2 and parallel to the line 1 1

2x + 4y + 3z 36, Determine the number of vertices and the number of edges of S. 2x + 3z 24.

x + y + z 11,

8.7 Rn
As a generalisation of R2 and R3 we dene Rn as the set of n-tuples x 1 x 2 : xi R . . . . x n

The dot product of two vectors in Rn is dened as x1 y1 x y = 2 2 = x y + x y + + x y . x y 1 1 2 2 n n . . . . . . yn xn The norm of a vector in Rn is given by = x . x x

Rn As in the case of R2 and R3 we have

179

400 Theorem (Cauchy-Bunyakovsky-Schwarz Inequality) Given (, ) (Rn )2 the following inequality x y holds || . x y x y
n n n

Proof:

Put a =
k=1

x2 , b = k
k=1 n

xk yk , and c =
k=1 n

y2 . Consider k
n n

f(t) =

(txk yk )2 = t2
k=1 k=1

x2 2t k
k=1

xk yk +
k=1

y2 = at2 + bt + c. k

This is a quadratic polynomial which is non-negative for all real t, so it must have complex roots. Its discriminant b2 4ac must be non-positive, from where we gather
n 2 n n

4
k=1

xk yk

x2 k
k=1 k=1

y2 k

This gives from where we deduce the result. u

x ||2 x y

401 Example Assume that ak , bk , ck , k = 1, . . . , n, are positive real numbers. Shew that
n 4 n n n 2

ak bk ck
k=1

a4 k
k=1 k=1

b4 k
k=1

c2 k

Solution: Using CBS on

n

n k=1 (ak bk )ck

once we obtain
n 1/2 n 1/2

k=1

ak bk ck
1/2

a2 b2 k k
k=1 k=1

c2 k

Using CBS again on

n k=1

n k=1

a2 b2 k k

we obtain
n k=1 n k=1

ak bk ck

a2 b2 k k a4 k

1/2

n k=1 n k=1

c2 k

1/2

1/4

b4 k

1/4

n k=1

c2 k

1/2

which gives the required inequality. 402 Theorem (Triangle Inequality) Given (, ) (Rn )2 the following inequality holds x y Proof: We have + + . x y x y = =

|| + b ||2 a

||||2 + 2|||||| b|| + || b||2 a a = (|||| + || b ||)2 , a

+ 2 + a a a b b b

( + b )( + b ) a a

180 from where the desired result follows. u We now consider a generalisation of the Euclidean norm. Given p > 1 and Rn we put x x
n 1/p p

Chapter 8

=
k=1

|xk |

(8.20)

Clearly

We now prove analogues of the Cauchy-Bunyakovsky-Schwarz and the Triangle Inequality for ||||p . For this we need the following lemma. 403 Lemma (Youngs Inequality) Let p > 1 and put 1 p + 1 q = 1. Then for (a, b) ([0; +[)2 we have

(8.21) (8.22) (8.23)

= || x

= 0 = 0 x
p

, R

ab Proof:

ap bq + . p q

Let 0 < k < 1, and consider the function [0; +[ x R

k

f:

Then 0 = f (x) = kxk1 k x = 1. Since f (x) = k(k 1)xk2 < 0 for 0 < k < 1, x 0, x = 1 1 and is a maximum point. Hence f(x) f(1) for x 0, that is xk 1 + k(x 1). Letting k = p ap x = q we deduce b 1 ap a 1+ 1 . bq/p p bq Rearranging gives ab b1+p/q + from where we obtain the inequality. u The promised generalisation of the Cauchy-Bunyakovsky-Schwarz Inequality is given in the following theorem. 404 Theorem (Hlder Inequality) Given (, ) (Rn )2 the following inequality holds x y x y Proof: If p = 0 or Young Inequality we have = 0 there is nothing to prove, so assume otherwise. From the || x y x
p

x k(x 1)

ap b1+p/qp b1+p/q p p

|xk |p |yk |q |xk | |yk | p + q . x p y q x p p y q q

Rn Adding, we deduce
n k=1

181

|xk | |yk | x p q y

= = =

1 pp x p p x 1 1 + p q 1. x
p p p p

n k=1

+ y

|xk |p + y q y
q q q q

1
q q

n k=1

|yk |q

This gives
n k=1

The result follows by observing that

n

|xk yk | x
n

k=1

xk yk

k=1

|xk yk | x

As a generalisation of the Triangle Inequality we have 405 Theorem (Minkowski Inequality) Let p ]1; +[. Given (, ) (Rn )2 the following inequality holds x y + x y p p + p. x y Proof: From the triangle inequality for real numbers 1.6 |xk + yk |p = |xk + yk ||xk + yk |p1 (|xk | + |yk |) |xk + yk |p1 . Adding
n n n k=1

|xk + yk |p

|xk ||xk + yk |p1 +

k=1 k=1

|yk ||xk + yk |p1 .

(8.24)

By the Hlder Inequality

n k=1

|xk ||xk + yk |p1

= =

n k=1 n k=1

|xk |p |xk |p

1/p

n k=1 n k=1

|xk + yk |(p1)q |xk + yk |p

1/q

1/q

1/p

(8.25)

In the same manner we deduce

n

+ x y
p

p/q p

k=1

Hence (8.24) gives + x y

p p n

|yk ||xk + yk |p1 y

+ x y
p/q p

p/q . p

(8.26)

=
k=1

from where we deduce the result. u

|xk + yk |p x

+ x y

+ y

+ x y

p/q , p

Homework

182
Problem 8.7.1 Prove Lagranges identity: aj bj
2

Chapter 8
Problem 8.7.4 Let Rn be a xed vector. Demona strate that X = { Rn : = 0} x a x is a subspace of Rn . Problem 8.7.5 Let i Rn , 1 i k (k n) be k a non-zero vectors such that i j = 0 for i = j. Prove that a a these k vectors are linearly independent. Problem 8.7.6 Let ak 0, 1 k n be arbitrary. Prove that
n 2

1jn

1jn

a2 j

1jn

b2 j

1k<jn (ak bj

aj bk )2

and then deduce the CBS Inequality in Rn . Problem 8.7.2 Let i Rn for 1 i n be unit vectors a n = 0 . Prove that n with i=1 ai 1i<jn ai aj = . 2 Problem 8.7.3 Let ak > 0. Use the CBS Inequality to shew that n n 1 a2 n2 . k a2 k k=1 k=1

ak
k=1

n(n + 1)(2n + 1) 6

k=1

a2 k . k2

Appendix

A
x X \ (X \ A) X \ (A B) x X x (X \ A) xXxA x X A.

Answers and Hints

1.1.2

1.1.3

x X (x (A B)) x X (x A x B) (x X x A) (x X x B) x (X \ A) x (X \ B) x (X \ A) (X \ B).

1.1.6 One possible solution is 1.1.8 We have giving Similarly, gives

A B C = A (B \ A) (C \ (A B)). |a| = |a b + b| |a b| + |b|, |a| |b| |a b|. |b| = |b a + a| |b a| + |a| = |a b| + |a|, |b| |a| |a b|.

The stated inequality follows from this.

1.2.1 a a since a = 1 Z, and so the relation is reexive. The relation is not symmetric. For 2 1 since 2 Z a 1 but 1 2 since 1 Z. The relation is transitive. For assume a b and b c. Then there exist (m, n) Z2 such 2 a that b = m, b = n. This gives c a a b = = mn Z, c b c and so a c. 1.2.2 Here is one possible example: put a b = 5 12.
a2 +a a 22 +2 5 6 5 a2 +a b

= a + 1 Z. On the other hand, the relation is not symmetric, since 5 2 as Z. It is not transitive either, since
5 +5 3
2

Z. Then clearly if a Z \ {0} we have a a since

3 +3 12
2

52 +5 2

Z = 5 3 and

Z = 3 12 but

= 15 Z but 2 5, as
52 +5 12

Z and so

183

184
1.2.4 [B] [x] = x + 1 Z. [C] No. 3
3 . 2

Appendix A

1.3.1 Let = 1 + i 2

Then 2 + + 1 = 0 and 3 = 1. Then

x = a3 + b3 + c3 3abc = (a + b + c)(a + b + 2 c)(a + 2 b + c), y = u3 + v3 + w3 3uvw = (u + v + w)(u + v + 2 w)(u + 2 v + w). Then (a + b + c)(u + v + w) = au + av + aw + bu + bv + bw + cu + cv + cw, (a + b + 2 c)(u + v + 2 w) = au + bw + cv +(av + bu + cw) +2 (aw + bv + cu), and (a + 2 b + c)(u + 2 v + w) = au + bw + cv +(aw + bv + cu) +2 (av + bu + cw). This proves that xy = (au + bw + cv)3 + (aw + bv + cu)3 + (av + bu + cw)3 3(au + bw + cv)(aw + bv + cu)(av + bu + cw), which proves that S is closed under multiplication. 1.3.2 We have where we may drop the parentheses since is associative. Similarly x(yz) = x(y a z) = (x) (a) (y a z) = x a y a z, (xy)z = (x a y)z = (x a y) (a) (z) = x a y a z. By virtue of having proved x(yz) = (xy)z, associativity is established. 1.3.3 We proceed in order. Clearly, if a, b are rational numbers, |a| < 1, |b| < 1 = |ab| < 1 = 1 < ab < 1 = 1 + ab > 0, whence the denominator never vanishes and since sums, multiplications and divisions of rational numbers a+b a+b are rational, is also rational. We must prove now that 1 < < 1 for (a, b) ] 1; 1[2 . We have 1 + ab 1 + ab 1 < a+b <1 1 + ab 1 ab < a + b < 1 + ab 1 ab a b < 0 < 1 + ab a b (a + 1)(b + 1) < 0 < (a 1)(b 1).

Since (a, b) ] 1; 1[2 , (a + 1)(b + 1) > 0 and so (a + 1)(b + 1) < 0 giving the sinistral inequality. Similarly a 1 < 0 and b 1 < 0 give (a 1)(b 1) > 0, the dextral inequality. Since the steps are reversible, we have a+b established that indeed 1 < < 1. 1 + ab

Answers and Hints

Since a b = a+b b+a = = b a, commutativity follows trivially. Now 1 + ab 1 + ba b+c 1 + bc

185

a (b c)

a a+

b+c 1 + bc b+c 1+a 1 + bc a(1 + bc) + b + c a + b + c + abc = . 1 + bc + a(b + c) 1 + ab + bc + ca

= One the other hand,

(a b) c

a+b 1 + ab

a+b +c 1 + ab a+b 1+ c 1 + ab (a + b) + c(1 + ab) 1 + ab + (a + b)c a + b + c + abc , 1 + ab + bc + ca

= = whence is associative. If a e = a then If a b = 0, then

a+e = a, which gives a + e = a + ea2 or e(a2 1) = 0. Since a = 1, we must have e = 0. 1 + ae a+b = 0, which means that b = a. 1 + ab

1.3.4 We proceed in order. Since a b = a + b ab = b + a ba = b a, commutativity follows trivially. Now a (b c) = = = One the other hand, (a b) c = = = whence is associative. If a e = a then a + e ae = a, which gives e(1 a) = 0. Since a = 1, we must have e = 0. If a b = 0, then a + b ab = 0, which means that b(1 a) = a. Since a = 1 we nd b = a . 1a (a + b ab) c a + b ab + c (a + b ab)c a + b + c ab bc ca + abc, a (b + c bc) a + b + c bc a(b + c bc) a + b + c ab bc ca + abc.

186

Appendix A

+ 0 1 2 3 4 5 6 7 8 9 10

0 0 1 2 3 4 5 6 7 8 9 10

1 1 2 3 4 5 6 7 0 9 10 0

2 2 3 4 5 6 7 8 9 10 0 1

3 3 4 5 6 7 8 9 10 0 1 2

4 4 5 6 7 8 9 10 0 1 2 3

5 5 6 7 8 9 10 0 1 2 3 4

6 6 7 8 9 10 0 1 2 3 4 5

7 7 8 9 10 0 1 2 3 4 5 6

8 8 9 10 0 1 2 3 4 5 6 7

9 9 10 0 1 2 3 4 5 6 7 8

10 10 0 1 2 3 4 5 6 7 8 9

0 1 2 3 4 5 6 7 8 9 10

0 0 0 0 0 0 0 0 0 0 0 0

1 0 1 2 3 4 5 6 7 8 9 10

2 0 2 4 6 8 10 1 3 5 7 9

3 0 3 6 9 1 4 7 10 2 5 8

4 0 4 8 1 5 9 2 6 10 3 7

5 0 5 10 4 9 3 8 2 7 1 6

6 0 6 1 7 2 8 3 9 4 10 5

7 0 7 3 10 6 2 9 5 1 8 4

8 0 8 5 2 10 7 4 1 9 6 3

9 0 9 7 5 3 1 10 8 6 4 2

10 0 10 9 8 7 6 5 4 3 2 1

Table A.1: Addition table for Z11 .

Table A.2: Multiplication table Z11 .

1.3.5 We have xy = = = = = = = proving commutativity. 1.4.1 The tables appear in tables A.1 and A.2. 1.4.2 Observe that 3x2 5x + 1 = 0 = 4(3x2 5x + 1) = 40 = x2 + 2x + 1 + 3 = 0 = (x + 1)2 = 8. We need to know whether 8 is a perfect square modulo 11. Observe that (11 a)2 = a, so we just need to check half the elements and see that 2 2 2 2 2 1 = 1; 2 = 4; 3 = 9; 4 = 5; 5 = 3, whence 8 is not a perfect square modulo 11 and so there are no solutions. 1.4.3 From example 50 x2 = 5. Now, the squares modulo 11 are 0 = 0, 1 = 1, 2 = 4, 3 = 9, 4 = 5, 5 = 3. Also, (11 4)2 = 7 = 5. Hence the solutions are x = 4 or x = 7. 1.4.5 Put f(x) = x4 + x3 + x2 + x + 1. Then f(0) = 1 1 f(3) = 121 0 f(6) = 1555 4 f(9) = 7381 0 mod 11 mod 11 mod 11 mod 11 f(1) = 5 5 f(4) = 341 0 f(7) = 2801 7 f(10) = 11111 1 mod 11 mod 11 mod 11 mod 11 f(2) = 31 9 f(5) = 781 0 f(8) = 4681 6 mod 11 mod 11 mod 11
2 2 2 2 2 2 2

(x y) (x y) [y (x y)] x [(x y) x] y [(y x) x] y [(x x) y] y (y y) (x x) y x,

Answers and Hints

1.5.1 We have 1 2+2 3+3 6 = = = = = = 1.5.2 Since

187

2+2 33 6 ( 2 + 2 3)2 (3 6)2 2+2 33 6 2 + 12 + 4 6 54 2+2 33 6 40 + 4 6 ( 2 + 2 3 3 6)(40 4 6) 402 (4 6)2 ( 2 + 2 3 3 6)(40 4 6) 1504 16 2 + 22 3 30 6 18 376

(a)b1 + ab1 = (a + a)b1 = 0F b1 = 0F , (a)b1 = (ab1 ).

we obtain by adding (ab1 ) to both sides that

Similarly, from a(b1 ) + ab1 = a(b1 + b1 ) = a0F = 0F , we obtain by adding (ab1 ) to both sides that a(b1 ) = (ab1 ). 1.6.1 Assume h(b) = h(a). Then h(a) = h(b) = = = Now, a3 = b3 a3 b3 = 0 (a b)(a2 + ab + b2 ) = 0

a 2 3a2 + . 2 4 This shews that b2 + ab + a2 is positive unless both a and b are zero. Hence a b = 0 in all cases. We have shewn that h(b) = h(a) = a = b, and the function is thus injective. b2 + ab + a2 = b + 6a 6b = 2a 3 2b 3 6a(2b 3) = 6b(2a 3) 12ab 18a = 12ab 18b 18a = 18b a = b,

1.6.2 We have f(a) = f(b)

proving that f is injective. Now, if then that is 6x = y(2x 3). Solving for x we nd

f(x) = y, y = 3, 6x = y, 2x 3 x=

3y . 2y 6 Since 2y 6 = 0, x is a real number, and so f is surjective. On combining the results we deduce that f is bijective.

188
1 2.1.1 A = 2 3 1 2.1.2 A = 2 3 1 8 . 27 3 6 . 9 0 b 2a 0 2c 0 , 2 2M = 0 2a + 2b 2a 4a 2a 0 2c 2b . 2

Appendix A

1 4 9

2 4 6

a + 1 2.1.3 M + N = a 2a 2.1.4 x = 1 and y = 4. 13 2.1.5 A = 15

2.2.2

2 2.2.1 0

2.1.8 The set of border elements is the union of two rows and two columns. Thus we may choose at most four elements from the border, and at least one from the central 3 3 matrix. The largest element of this 3 3 matrix is 15, so any allowable choice of does not exceed 15. The choice 25, 15, 18,l 23, 20 shews that the largest minimum is indeed 15. 2 2

1 5 , B = 6 3

0 1

a AB = c + a a+b+c 1 2.2.3 2 3 2 3 1 3 1 1 2 3 2 1 2 3 1 14 2 = 11 11 3 14 11 11

b a+b a+b+c

a + b + c , BA = b+c a + b + c a+b+c a+b+c c

b+c a+b c+a

c b a

14 11, whence a + b + c = 36. 11 0 0 0 0 4 0 0 0 12 0 0 4 3 , N = 0 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0 4 and N = 0 0 0 0 0 0 0 0 0 0 0 0 0 0 . Hence 0 0

0 0 2 2.2.4 An easy computation leads to N = 0 0

any powerfrom the fourth onis the zero matrix.

Answers and Hints

189

0 0 2.2.5 AB = 04 and BA = 0 0

2 2 2 2

0 0 0 0

2.2.6 For the rst part, observe that

3 3 . 3 3

m(a)m(b)

0 a 1 0 b 1 a2 b2 a 1 b 1 2 2 0 0 1 0 0 1 1 0 a+b 2 2 a b + ab a b 1 2 2 0 0 1

(a + b) 0 m(a + b)

0 1 0

(a + b) 2 1
2

a+b

For the second part, observe that using the preceding part of the problem, 1 m(a)m(a) = m(a a) = m(0) = 0 0 0 1 0

giving the result. 2.2.7 Observe that

0 02 = I3 , 2 1

A2 = (AB)(AB) = A(BA)B = A(B)B = (AB)B = AB = A. Similarly, B2 = (BA)(BA) = B(AB)A = B(A)A = (BA)A = BA = B. 2.2.8 For this problem you need to recall that if |r| < 1, then a + ar + ar2 + ar3 + + = This gives 1+
1 2 1 4

a . 1r = 2 , 3 = 2. 4 , 3

+
1 23

1 42

+
1 25

1 43

+ =

1 1
1 2 1 4

1 4

+ =
1 23

and 1+
1 2

1 22

+ =

1 1

1 2

190
By looking at a few small cases, it is easy to establish by induction that for n 1 0 = 1 2n1 2 0 1 22n1 0 0 0 0 1 22n1 1 22n = 0 0 0 1 22n 0 0 0 . 1 22n

Appendix A

A2n1

A2n

This gives 1 1 1 1 + 4 + 42 + 43 + I 3 + A + A2 + A3 + = 1 + 1 + 1 + 2 23 25 0 2.2.9 Observe that 4 x x 4 = x 4 2 x 4 x 4 x 16 x = 0 4

2 1 2

+
1 4

1 23

+
1 42

1 25

+
1 43

0 0 1+
1 2

1+

+ 0

+ +
1 22

1 23

= 2 3 0 +

4 3

2 3 4 3

0 0 . 2

0 16 x2

and so we must have 16 x2 = 1 or x = 17. 1 2.2.10 Disprove! Take A = 0

0 0 and B = 0 0

1 . Then AB = B, but BA = 02 . 0 0 . Then 0 0 = (A + B)(A B). 1

0 2.2.11 Disprove! Take for example A = 1

0 1 and B = 1 1 0 1 = 2 1

1 A2 B2 = 0 2.2.12 x = 6. 32 2.2.14 32 32 . 32 2
1001 1002

0 2.2.15 A2003 = 21002 31001

2.2.17 The assertion is clearly true for n = 1. Assume that is it true for n, that is, assume cos(n) An = sin(n) sin(n) . cos(n)

Answers and Hints

Then An+1 = AAn

191

cos sin

sin cos(n) sin(n) cos

sin(n) cos(n)

cos cos(n) sin sin(n) sin cos(n) + cos sin(n) sin(n + 1) , cos(n + 1)

cos sin(n) sin cos(n) sin sin(n) + cos cos(n)

and the result follows by induction.

cos(n + 1) sin(n + 1)

2.2.18 Let A = [aij ], B = [bij ] be checkered n n matrices. Then A + B = (aij + bij ). If j i is odd, then aij + bij = 0 + 0 = 0, which shows that A + B is checkered. Furthermore, let AB = [cij ] with cij = n aik bkj . If i is k=1 even and j odd, then aik = 0 for odd k and bkj = 0 for even k. Thus cij = 0 for i even and j odd. Similarly, cij = 0 for odd i and even j. This proves that AB is checkered. 2.2.19 Put 0 J = 0 0 0 J2 = 0 0 0 0 0 1 0 0 1 1 . 0

We rst notice that

This means that the sum in the binomial expansion

1 3 0 , J = 03 . 0
n

An = (I3 + J)n =
k=0

n nk k I J k

is a sum of zero matrices for k 3. We thus have An =

n1 In + nI3 J + 3 n 2 n2 I3 J2

1 0 0 1 0 0

0 1 0 n 1 0
n 2

0 0 0 + 0 0 1

n 0 0

n(n+1) 2

n 0 n + 0 0 0

0 0 0

0 0

n 2

n 1

giving the result, since

n 2

n(n1) 2

and n +

n(n+1) . 2

192
2.2.20 Argue inductively, A2 B = A(AB) = AB = B A3 B = A(A2 B) = A(AB) = AB = B . . . Am B = AB = B. Hence B = Am B = 0n B = 0n .

Appendix A

a 2.2.22 Put A = c

b . Using 2.2.21, deduce by iteration that d Ak = (a + d)k1 A.

a 2.2.23 c

b , bc = a2 a b , a2 = 1 bc a b = X I. Then d 0 0 +I= 6 3 0 . 4

a 2.2.24 I2 , c

a 2.2.25 We complete squares by putting Y = c a + bc c(a + d)

2

b(a + d) 1 = Y 2 = X2 2X + I = (X I)2 = 6 bc + d2

This entails a = 0, b = 0, cd = 6, d2 = 4. Using X = Y + I, we nd that there are two solutions,

1 3 2.2.26 The matrix

0 1 , 3 3

0 . 1

clearly satises the conditions.

1 1 A= 1 1

1 1 1 1

1 1 1 1

1 1 1 1

Answers and Hints

193

a 2.2.27 Put X = c

b . Then d

1 X2 + X = 1

The last equation holds

1 ca + dc + c = 1 1 cb + d2 + d = 1

2 a + bc + a = 1 ab + bd + b = 1

2 a + bc + a = 1 b(a + d + 1) = 1

c(a + d + 1) = 1 (d a)(a + d + 1) = 0

d=a= 1 2 1 c = b = 2a + 1 1 2 a + (2a + 1)2 + a = 1

Thus the set of solutions is

1 3 4a4 + 8a3 + a2 3a = 0 a { , 1, 0, }. 2 2 1 { 1 1 0 , 1 1 1 1/2 , 1/2 0 1 J = 1 1 1/2 3/2 , 1/2 1/2 1 1 . 1 1/2 } 3/2

2.2.29 Observe that A = 2I3 J, where I3 is the 3 3 identity matrix and 1 1 1

Observe that Jk = 3k1 J for integer k 1. Using the binomial theorem we have An = = = = (2I3 J)n
n k=0 n k

(2I3 )nk (1)k Jk

2n I3 +

1 n nk J n (1)k 3k k=1 k 2 3 1 2n I3 + J((1)n 2n ) 3 n n+1 (1)n 2n (1) + 2 1 n n (1)n + 2n+1 3 (1) 2 (1)n 2n (1)n 2n

(1)n 2n n n . (1) 2 n n+1 (1) + 2

2.3.1 There are innitely many solutions. Here is one: 1 A = 2 3 2 3 1 3 9 1 = 2 3 2 2 3 1 3 10 1 + 0 0 2 0 0 0 0 0 . 0 1 . Another set is 1

1 2.3.2 There are innitely many examples. One could take A = 1

1 1 and B = 1 1

194

Appendix A

2.3.3 If such matrices existed, then by the rst equation tr (AC) + tr (DB) = n. By the second equation and by Theorem 86, 0 = tr (CA) + tr (BD) = tr (AC) + tr (DB) = n, a contradiction, since n 1. 1 2.3.4 Disprove! This is not generally true. Take A = 1 1 3 and B = 0 2 1 2 3 . 2 0 . Clearly AT = A and BT = B. We have 1

2 A= 0

0 0 and B = 0 0

0 . 2

but

3 AB = 3 3 (AB)T = 1

2.3.6 We have a = tr c b a c d Thus tr A2 is the condition sought. 2.3.7 tr (A I4 )2

tr A2

and

b a + bc = tr ca + cd d 2

ab + bd = a2 + d2 + 2bc d2 + cb

a tr c

b = (a + d)2 . d

= (tr (A))2 a2 + d2 + 2bc = (a + d)2 bc = ad,

= = = =

tr A2 2A + I4 tr A2 2tr (A) + tr (I4 ) 4 2tr (A) + 4 2tr (A) ,

and tr (3I4 ) = 12. Hence 2tr (A) = 12 or tr (A) = 6. 2.3.8 Disprove! Take A = B = In and n > 1. Then tr (AB) = n < n2 = tr (A) tr (B). 1 2.3.9 Disprove! Take A = 0 0 0 , B = 0 0 1 0 , C = 1 0 0 . Then tr (ABC) = 1 but tr (BAC) = 0. 0

Answers and Hints

2.3.10 We have (AAT )T = (AT )T AT = AAT .

195

2.3.11 We have

(AB BA)T = (AB)T (BA)T = BT AT AT BT = BA A(B) = AB BA.

2.3.13 Let X = [xij ] and put XXT = [cij ]. Then

0 = cii =
k=1

x2 = xik = 0. ik

2.4.1 Here is one possible approach. If we perform C1 C3 on A we obtain 1 0 A1 = 1 1 Now perform 2R1 R1 on A1 to obtain 2 0 A2 = 1 1 0 1 1 1 2 0 1 1 0 1 1 1 so take 2 0 D= 0 0 0 1 0 0 0 0 1 0 0 0 . 0 1 0 1 1 1 1 0 1 1 0 1 1 1 so take 0 0 P= 1 0 0 1 0 0 1 0 0 0 0 0 . 0 1

Finally, perform R1 + 2R4 R1 on A2 to obtain 4 0 B= 1 1 2 1 1 1 4 0 1 1 2 1 1 1 so take 1 0 T = 0 0 0 1 0 0 0 0 1 0 2 0 . 0 1

196
2.4.2 Here is one possible approach. a d g c f i

Appendix A

b e h

P:3 1

P :C1 C2

g d a h e b

h e b g d a

i f c g d a

i f c

T :C1 C2 C1

D:23 3

h g e d ba

i f c g d 2a

hg ed 2b 2a

i f 2c

Thus we take 0 P = 0 1 1 T = 1 0 2.4.3 Let Eij Mnn (F). Then 0 0 . = . . 0 0 0 0 . . . 0 0 ... . . . . . . . . . . . . a1i a2i . . . an1i ani ... . . . . . . . . . . . . 0 0 . . , . 0 0 0 1 0 1 0 , P = 0 1 0 0 0 0 , 1 1 D = 0 0 1 0 0 0 0 , 1 0 0 . 2

0 1 0

0 1 0

AEij

where the entries appear on the j-column. Then we see that tr (AEij ) = aji and similarly, by considering BEij , we see that tr (BEij ) = bji . Therefore i, j, aji = bji , which implies that A = B.

Answers and Hints

2.4.4 Let Est Mnn (R). Then

197

where the entries appear on the i-th row. Thus

0 . . . Eij A = aj1 . . . 0

0 . . . aj2 . . . 0

... ... ... ... ...

0 . . . ajn , . . . 0

which means that i, j, aji ajk = 0. In particular, a2 = 0, which means that i, j, aji = 0, i.e., A = 0n . ji 2.5.1 a = 1, b = 2. 2.5.2 Claim: A1 = In A + A2 A3 . For observe that (In + A)(In A + A2 A3 ) = In A + A2 A3 A + A2 A3 + A4 = In , proving the claim. 2.5.3 Disprove! It is enough to take A = B = 2In . Then (A + B)1 = (4In )1 =
1 I 4 n

. . . 2 (Eij A) = aji aj1 . . . 0

0 . . . aji aj2 . . . 0

... ... ... ... ...

aji ajn , . . . 0 . . .

but A1 + B1 =

1 I 2 n

1 + 2 In = In .

2.5.8 We argue by contradiction. If exactly one of the matrices is not invertible, the identities A = AIn = (ABC)(BC)1 = 0n , B = In BIn = (A)1 (ABC)C1 = 0n , C = In C = (AB)1 (ABC) = 0n , shew a contradiction depending on which of the matrices are invertible. If all the matrices are invertible then 0n = 0n C1 B1 A1 = (ABC)C1 B1 A1 = In , also gives a contradiction. 2.5.9 Observe that A, B, AB are invertible. Hence A2 B2 = In = (AB)2 = = AABB = ABAB AB = BA,

by cancelling A on the left and B on the right. One can also argue that A = A1 , B = B1 , and so AB = (AB)1 = B1 A1 = BA. 2.5.10 Observe that A = (a b)In + bU, where U is the n n matrix with 1F s everywhere. Prove that A2 = (2(a b) + nb)A ((a b)2 + nb(a b))In . 2.5.11 Compute (A In )(B In ).

198
2.5.12 By Theorem 86 we have tr SAS1 = tr S1 SA = tr (A). 2.7.2 The rank is 2.

Appendix A

2.7.3 If B is invertible, then rank (AB) = rank (A) = rank (BA). Similarly, if A is invertible rank (AB) = rank (B) = and so rank (BA) = 0. 2.7.4 Observe that 1 0 2 2 1 0 2 0 0 1 2 0 0 1 2 2
R3 2(R1 +R2 )R3 R4 2R1 R4

1 rank (BA). Now, take A = 0

0 0 and B = 0 0

1 0 . Then AB = B, and so rank (AB) = 1. But BA = 0 0

0 , 0

whence the matrix has three pivots and so rank 3.

1 0 0 0

1 0 0 2

0 1 0 0

0 1 , 0 2

2.7.5 The maximum rank of this matrix could be 2. Hence, for the rank to be 1, the rows must be proportional, which entails x2 x = = x2 2x = 0 = x {0, 2}. 4 2 2.7.6 Assume rst that the non-zero n n matrix A over a eld F has rank 1. By permuting the rows of the matrix we may assume that every other row is a scalar multiple of the rst row, which is non-zero since the rank is 1. Hence A must be of the form a1 1 a1 A= . . . n1 a1 a2 1 a2 . . . n1 a2 a1 1 an a2 = 1 . . . . . . an n1 an an 1 n1 := XY,

which means that the claimed factorisation indeed exists. Conversely, assume that A can be factored as A = XY, where X Mn1 (F) and Y M1n (F). Since A is non-zero, we must have rank (A) 1. Similarly, neither X nor Y could be all zeroes, because otherwise A would be zero. This means that rank (X) = 1 = rank (Y). Now, since rank (A) min(rank (X) , rank (Y)) = 1, we deduce that rank (A) 1, proving that rank (A) = 1. 2.7.7 Effecting R3 R1 R3 ; aR4 bR2 R4 successively, we obtain 1 a 1 b a 1 b 1 1 b 1 a b 1 a 1 1 a 0 0 a 1 ba ab 1 b 0 a2 b2 b

1 . a b ab

Answers and Hints

Performing R2 aR1 R2 ; R4 + R3 R4 we have 1 0 0 0 a 1 a2 ba 0

199

1 ba 0 a2 b2

Performing (1 a2 )R3 (b a)R2 R3 we have 1 0 0 0 a 1 a2 0 0

1 ab . ab 2(a b)

1 ba

a2 + 2ab b2 a2 b2

Performing R3 R4 R3 we have

1 ab . 3 2 2a 2b a + ab 2(a b)

Performing 2aR4 + (a + b)R3 R4 we have 1 0 0 0 a 1 a2 0 0

1 0 0 0

a 1 a2 0 0

1 ba 2a(a b) a2 b2

1 ab . 2 2 a(a b ) 2(a b)

b 1 ab a(a2 b2 ) 4a2 4ab a4 + a2 b2 ba3 + ab3

ba 2a(a b) 0

Factorising, this is

Thus the rank is 4 if (a + 2 + b)(a b)(a 2 + b) = 0. The rank is 3 if a + b = 2 and (a, b) = (1, 1) or if a + b = 2 and (a, b) = (1, 1). The rank is 2 if a = b = 1 and a = 1. The rank is 1 if a = b = 1. 2.7.8 rank (A) = 4 if m3 + m2 + 2 = 0, and rank (A) = 3 otherwise. 2.7.9 The rank is 4 if a = b. The rank is 1 is a = b = 0. The rank is 0 if a = b = 0. 2.7.10 The rank is 4 if (a b)(c d) = 0. The rank is 2 is a = b, c = d or if a = b, c = d. The rank is 1 if a = b and c = d.

1 0 = 0 0

a 1 a2 0 0

1 ba 2a(a b) 0

b 1 ab a(a b)(a + b) a(a + 2 + b)(a b)(a 2 + b)

200
2.7.11 Observe that rank (ABC) rank (B) 2. Now, 1 2 1 1 x 2 2 1 1 1 0 0 1 x+2 3 2 5 , 1

Appendix A

R2 +2R1 R2 R3 R1 R3

has rank at least 2, since the rst and third rows are not proportional. This means that it must have rank exactly two, and the last two rows must be proportional. Hence x+2 5 = = x = 13. 3 1 1 2.7.13 For the counterexample consider A = i 2.8.1 We form the augmented matrix 1 2 3 2 3 1 3 1 2 1 0 0 0 1 0 0 0 1 i . 1

From R2 2R1 R2 and R3 3R1 R3 we obtain 1 0 0 From R2 R3 we obtain 1 0 0 2 6 2 3 2 0 1 5 4 0 1 0

0 0 . 1 0 1 . 0

2 2 6

3 0 2

1 4 5

0 0 1

Now, from R1 R2 R1 and R3 3R2 R3 , we obtain 1 0 0 From 4R2 R2 and 4R3 R3 , we obtain 0 2 0 3 0 2 4 4 0 0 0 1

6 1 . 4

1 0 0

0 1 0

3 0 1

4 2 0

0 0 4

6 4 . 2

Answers and Hints

Finally, from R1 3R3 R3 we obtain

201

We deduce that 1 2 3

1 0 0 2 3 1

0 1 0

0 0 1

4 2 0

2 0 4

0 4 . 2 2 0 4 0 4 . 2

3 1 2

4 = 2 0

2.8.2 To nd the inverse of B we consider the augmented matrix

0 0 1 Performing R1 R3 , R3 R3 , in succession, 1 0 0

0 1 a

1 a b

1 0 0

0 1 0

0 0 . 1

a 1 0

b a 1

0 0 1

0 1 0

1 0 . 0

Performing R1 + aR2 R1 and R2 aR3 R2 in succession, 1 0 0 0 1 0 b + a2 0 1 0 a 1 a 1 0

1 0 . 0

Performing R1 (b + a2 )R3 R3 , R1 R1 and R2 R2 in succession, we nd 1 0 0 0 1 0 0 0 1 b a2 a 1 a 1 0 1 0 , 0

whence 2 b a = a 1 a 1 0 1 0 . 0

B1

202
Now, 0 0 1 0 1 0 a b c 0 1 a 1 a 0 1 0 0 0 1 0 1 a b c b a2 a 1 a 1 0 0 a 1 0 0 1 0 b 0 b a2 a 1 1 0 0 a 1 0 0 c 1 0 0

Appendix A

BAB1

= which is what we wanted to prove.

AT ,

2.8.3 First, form the augmented matrix: 1 1 1 1 0 0 0 1 1 0 0 x 1 0 0 0 1 0 0 0 . 1 0 0 . 1 0 0 . 1

Perform R2 R1 R2 and R3 R1 R3 :

0 1 1

0 0 x

1 1 1

0 1 0

Performing R3 R2 R3 :

Finally, performing

1 R3 R3 : x

1 0 0

0 1 0

0 0 x

1 1 0

0 1 1

1 0 0

0 1 0

0 0 1

1 1 0

0 1 1 x

0 0 . 1 x

2.8.4 Since MM1 = I3 , multiplying the rst row of M times the third column of M1 , and again, the third row of M times the third column of M1 , we gather that 1 0 + 0 a + 1 b = 0, 0 0 + 1 a + 1 b = 1 = b = 0, a = 1.

Answers and Hints

1 0 n 1 0. Row-reducing, (A ) = n (n 1)n 1 2

203
0 0. 1

n 2.8.5 It is easy to prove by induction that A = 0 2.8.6 Take, for example, A = 1 1 = A1 . 0

1 n n(n + 1) 2

0 1 n

0 1 n

2.8.7 Operating formally, and using elementary row operations, we nd a2 1 a2 +2a2 a2 a2 5+2a a2 5+2a a2 5+2a a+4 1 B1 = 2 2 a2 5+2a a 5+2a a2 5+2a 2a a a22a+5 a2 5+2a 5+2a a2 5+2a Thus B is invertible whenever a = 1 6.

2.8.8 Form the augmented matrix a 0 0 2a b 0 3a 2b c 1 0 0 0 1 0 0 0 . 1

Perform

Now perform R1 2R2 R1 and R2 2R3 R2 in succession, to obtain 1 0 1 1/a 2/a 0 0 0 1/b 2/c . 0 1 0 0 1/c 0 0 1 Finally, perform R1 + R3 R1 to obtain

1 1 1 R1 R1 , R2 R2 , R3 R3 , in succession, obtaining a b a 0 0 1 2 3 1/a 0 1/b 0 . 0 1 2 0 0 1 0 0 1/c

Whence a 0 0

1 0 0 2a b 0

0 1 0

0 0 1

1/a 0 0

2/b 1/b 0

1/c 2/c . 1/c 1/c 2/c . 1/c

3a 2b c

1/a = 0 0

2/b 1/b 0

204
2.8.9 To compute the inverse matrix we proceed formally as follows. The augmented matrix is b c 0 Performing bR2 cR1 R2 we nd a 0 c 0 a b 1 0 0 0 1 0 0 0 . 1

Appendix A

Performing aR3 + R2 R3 we obtain

b 0 0 b 0 0

a ca c

0 ab b

1 c 0

0 b 0

0 0 . 1 0 0 . a

a ca 0

0 ab 2ab

1 c c

0 b b

Performing 2R2 R3 R2 we obtain

b 0 0 Performing 2cR1 + R2 R1 we obtain

a 2ca 0

0 0 2ab

1 c c

0 b b

0 a . a

2bc 0 0 From here we easily conclude that

0 2ca 0

0 0 2ab

c c c

b b b

a a . a

b c 0 as long as abc = 0.

a 0 c

0 a b

1 2b 1 2a c 2ba

1 2c b 2ac 1 2a

a 2bc 1 2c 1 2b

2.8.10 Since AB is invertible, rank (AB) = n. Thus n = rank (AB) rank (A) n = rank (A) = n, n = rank (AB) rank (B) n = rank (B) = n, whence A and B are invertible.

Answers and Hints

2.8.11 Form the expanded matrix 1 + a 1 1 1 1+b 1 1 1 1+c 1 0 0 0 1 0 0 0 . 1

205

Perform bcR1 R1 , abR3 R3 , caR2 R2 . The matrix turns into bc + abc ca ab bc ca + abc ab bc ca ab + abc bc 0 0

0 ca 0

0 0 . ab

Perform R1 + R2 + R3 R1 the matrix turns into ab + bc + ca + abc ca ab ab + bc + ca + abc ca + abc ab ab + bc + ca + abc ca ab + abc bc 0 0 ca ca 0

ab 0 . ab

Perform

1 R ab+bc+ca+abc 1

R1 . The matrix turns into 1 ca + abc ab 1 ca ab + abc

bc ab+bc+ca+abc ca ab+bc+ca+abc

1 ca ab

0 0

ca 0

ab ab+bc+ca+abc

ab

Perform R2 caR1 R2 and R3 abR3 R3 . We get 1 0 0 1 abc 0 1 0 abc

1 R ABC 3 bc ab+bc+ca+abc abc ab+bc+ca+abc ab c ab+bc+ca+abc
2 2

ca ab+bc+ca+abc

ab ab+bc+ca+abc a bc ab+bc+ca+abc
2

ca

c a ab+bc+ca+abc
2

a bc ab+bc+ca+abc

ab

a2 b2 ab+bc+ca+abc

Perform

1 R ABC 2

R2 and 1 0 0 1 1 0

R3 . We obtain
ca ab+bc+ca+abc 1 b ab ab+bc+ca+abc a ab+bc+ca+abc 1 c

1 0 1

bc ab+bc+ca+abc c ab+bc+ca+abc b ab+bc+ca+abc

ca b(ab+bc+ca+abc)

a ab+bc+ca+abc

ab c(ab+bc+ca+abc)

Finally we perform R1 R2 R3 R1 , getting 1 0 0 0 1 0 0 0 1

a+b+bc ab+bc+ca+abc c ab+bc+ca+abc b ab+bc+ca+abc 1 b c ab+bc+ca+abc b ab+bc+ca+abc a ab+bc+ca+abc 1 c

ca b(ab+bc+ca+abc)

a ab+bc+ca+abc

ab c(ab+bc+ca+abc)

206
We conclude that the inverse is

Appendix A

b+c+bc

ab+bc+ca+abc c ab+bc+ca+abc b ab+bc+ca+abc

c ab+bc+ca+abc c+a+ca ab+bc+ca+abc a ab+bc+ca+abc

b ab+bc+ca+abc a ab+bc+ca+abc
a+b+ab ab+bc+ca+abc

2.8.16 Since rank A2 < 5, A2 is not invertible. But then A is not invertible and hence rank (A) < 5. 2.8.17 Each entry can be chosen in p ways, which means that there are p2 ways of choosing the two entries of an arbitrary row. The rst row cannot be the zero row, hence there are p2 1 ways of choosing it. The second row cannot be one of the p multiples of the rst row, hence there are p2 p ways of choosing it. In total, this gives (p2 1)(p2 p) invertible matrices in Zp . 2.8.18 Assume that both A and B are m n matrices. Let C = [A B] be the m (2n) obtained by juxtaposing A to B. rank (C) is the number of linearly independent columns of C, which is composed of the columns of A and B. By column-reducing the rst n columns, we nd rank (A) linearly independent columns. By column-reducing columns n + 1 to 2n, we nd rank (B) linearly independent columns. These rank (A) + rank (B) columns are distinct, and are a subset of the columns of C. Since C has at most rank (C) linearly independent columns, it follows that rank (C) rank (A) + rank (B). Furthermore, by adding the n + k-column (1 k n) of C to the k-th column, we see that C is column-equivalent to [A + B B]. But clearly rank (A + B) rank ([A + B B]) = rank (C) , since [A + B B] is obtained by adding columns to A + B. We deduce rank (A + B) rank ([A + B B]) = rank (C) rank (A) + rank (B) , as was to be shewn. 2.8.19 Since the rst two columns of AB are not proportional, and since the last column is the sum of the rst two, rank (AB) = 2. Now, 2 0 (AB)2 = 1 1 1 0 1 1 0 = 1 1 1 2 1 0 1 1 1 = AB. 2

Since BA is a 2 2 matrix, rank (BA) 2. Also,

2 = rank (AB) = rank (AB)2

= rank (A(BA)B) rank (BA) ,

whence rank (BA) = 2, which means BA is invertible. Finally, (AB)2 AB = 03 = A(BA I2 )B = 03 = BA(BA I2 )BA = B03 A = BA I2 = 02 , since BA is invertible and we may cancel it. 3.1.1 The free variables are z and w. We have 2y + w = 2 = 2y = 2 w = y = 1 + w, and x + y + z + w = 0 = x = y z w = 2y + 2z + 2w. Hence 0 0 x 0 0 0 y 1 = + z + w . 1 0 z 0 0 0 1 w

This gives the 9 solutions.

Answers and Hints

3.1.2 We have 1 2 3 x 1 = y 2 3 z 2 3 1 2 3 1 3 x 5 1 y = 6 , z 2 0 2 0 4 0 5 4 4 6 = 3 . 2 0 3 5 3 2 5 3 7

207

Hence

3.1.3 The augmented matrix of the 1 2 5

system is 2 2 3 1 0 4 5 7 1

1 3 5 4 1 6 = 2 2 0 0

R3 5R1 R3 R2 2R1 R2

6R3 7R2 R3

1 0 0 1 0 0

2 6 7 2 6 0

1 2 1 1 2 8

Backward substitution yields 6y = 2z 3 = 2 9 3 = 15 = 2 = 11 6y = 11 2 = y = 22 = 9, Conclusion : Check: 1 2 9 + 9 = 8 = 5, 2 1 + 2 9 = 20 = 7, 5 1 3 9 + 4 9 = 14 = 1. 3.1.4 We need to solve the system a b + c d = p(1) = 10, a = p(0) = 1, a + b + c + d = p(1) = 2, a + 2b + 4c + 8d = p(2) = 23. Using row reduction or otherwise, we nd a = 1, b = 2, c = 3, d = 4, and so the polynomial is p(x) = 4x3 3x2 + 2x 1. 3.1.5 Using the encoding chart 0 A 13 N 1 B 14 O 2 C 15 P 3 D 16 Q 4 E 17 R 5 F 18 S 6 G 19 T 7 H 20 U 8 I 21 V 9 J 22 W 10 K 23 X 11 L 24 Y 12 M 25 Z x = 2y 1z + 5 = 2 9 1 9 + 5 = 14 = 1. x = 1, y = 9, z = 9. 8z = 7 = 5 8z = 5 7 = z = 35 = 9,

208
we nd M 12 P2 = U = 20 , 13 N I 8 P3 = S = 18 , 19 T S 18 P4 = E = 4 , 0 A T 19 P5 = O = 14 , 5 F

Appendix A

Thus

F 5 P6 = A = 0 . 11 L

Finally, the message is encoded into

0 A 14 O 4 E 18 S 20 U AP2 = 10 = K , AP3 = 24 = Y , AP4 = 2 = C , AP5 = 5 = F , AP6 = 15 = P . W 22 K 10 A 0 M 12 A 0 OGY UKA SYM ECA OFK APW.

3.1.6 Observe that since 103 is prime, Z103 is a eld. Adding the rst hundred equations, 100x0 + x1 + x2 + + x100 = 4950 = 99x0 = 4950 4949 = 1 = x0 = 77 Now, for 1 k 100, This gives x1 = 26, x2 = 27, . . . , x77 = 102, x78 = 0, x79 = 1, x80 = 2, . . . , x100 = 22. 3.3.1 Observe that the third row is the sum of the rst two rows and the fourth row is twice the third. So we have 1 1 2 4 1 1 1 1 1 0 1 2 0 1 2 4 0 0 1 2 0 1 2 4 1 1 1 0 0 0
R3 R1 R2 R3 R4 2R1 2R2 R4

mod 103.

xk = k 1 x0 = k 78 = k + 25.

1 1 0 0 1 0 0 0 0 1

0 0 0 0 1 0 0 0 0

1 0 0 0 1 1 0 0 0

0 0 0 0 1 0 0 0 0

1 0 0 1 0 0 0 0 1

1 1 0 0 0 1 1 0 0 0

R2 R5 R2 R1 R5 R1

Rearranging the rows we obtain 1 0 0 0 0 0 1 0 0 0 0 1 1 0 0 0 1 0 0 0 1 0 0 0 0 0 1 1 . 0 0

Answers and Hints

Hence d and f are free variables. We obtain c = 1, b = 1 c d = 2 d, a = f. The solution is 1 0 a 0 0 1 b 2 c = 1 + d 0 + f 0 . 0 1 d 0 1 0 0 f 1 1 3.3.2 The unique solution is 1. 1 1 3.3.3 The augmented matrix of the system is 2m 1 1 Performing R1 R2 . 1 2m 1 1 1 2m 2 4m . 2 2m

209

Performing R2 R3 .

1 2m 1 1 1 2m

2m 1 1

1 1 2m

4m 2 . 2 2m 4m 2 . 2m 2

2m 1 1

1 2m 1

Performing R2 R1 R1 and R3 2mR1 R3 we obtain 1 0 0 2m 1 2m 1 4m2 1 2m 1 1 2m

2 2m 4m . 2 2 8m

4m

210
If m =
1 2

Appendix A
the matrix becomes 1 0 0 1 0 0 1 0 0 2 3 2 0

and hence it does not have a solution. If m = 1 0 0 Performing R3 (1 + 2m)R2 R3 we obtain 1 0 0

1 , 2

by performing 2m 1 1 1 1

1 R 12m 2

R2 and

1 R 12m 3

R3 , the matrix becomes

1 + 2m

2m(m2) . 12m 2(1 + 2m)

4m

2m 1 0

1 1 2 + 2m

4m
2m(m2) 12m 2(1+2m)(1m2 ) 12m

If m = 1 then the matrix reduces to

The solution in this case is

1 0 0

2 1 0

1 1 0

4 2 . 0

1 If m = 1, m = 2 we have the solutions

x z = y 2 + z . z z x . = 13m y 12m (1+2m)(1m) z 12m

m1 12m

3.3.4 By performing the elementary row operations, we obtain the following triangular form: ax + y 2z = 1, (a 1)2 y + (1 a)(a 2)z = 1 a, (a 2)z = 0. If a = 2, there is an innity of solutions: x 1 + t = y 1 t R. t z

Answers and Hints

Assume a = 2. Then z = 0 and the system becomes ax + y = 1, (a 1)2 y = 1 a, 2x + (3 a)y = 1. We see that if a = 1, the system becomes x + y = 1, 2x + 2y = 1, and so there is no solution. If (a 1)(a 2) = 0, we obtain the unique solution 1 x a1 = 1 . y a1 z 0 1 x m2 3.3.5 The system is solvable if m = 0, m = 2. If m = 2 there is the solution y = m+3 . m2 m+2 z m2 x a + d + b c y c d b + a 3.3.6 There is the unique solution = . z d + c b + a cd+b+a t b c 0 = a 0 c 0 x c a y = b . a z b a 0 c
1 2b

211

3.3.7 The system can be written as

The system will have the unique solution x y z

b c 0

0 a b

1 c b a a 2bc c , 1 b 2c 1 a 2b

1 2a c 2ba

1 2c b 2ac 1 2a

b2 + c2 a2 2bc a2 + c2 b2 2ac 2 a + b2 c2 2ab

212
as long as the inverse matrix exists, which is as long as abc = 0 3.3.8 We rst form the augmented matrix, 1 a a 2 2a + 1 a a+1 2a + 2 0 a1 a 1 a 2 3a + 1 a a+1 3a + 1 3a2 5a 1 2a + 2 0 a1 2a + 2 2a2 2a 3a 5 3a + 2

Appendix A

2a + 2 2 a 2a + 9

R1 +R2 R1

R2 aR1 R2 R3 2R1 R3

. After (5a 1)R2 + 3a2 R3 R2 , this las matrix becomes 1 0 0 3a + 1 2a + 2 a3 3a2 + 2a 3a 5

1 0 0

2a + 2 2 a 2a + 9 3a + 2

2 3a + 2 a2 8a + 5

3a + 2

0 5a 1

Exchanging the last two rows and factoring, 1 0 0 3a + 1 5a 1 0 2a + 2 3a 5 a(a 1)(a 2)

4 3 2 3a 9a + 18a 10a 2 . 2 a 8a + 5

3a + 2 a2 8a + 5 (a 1)(3a3 6a2 + 12a + 2)

Thus we must examine a {1, 2, 3} and a {0, 1, 2}. Clearly, if a(a 1)(a 2) = 0, then there is the unique solution z= 2 + 12a 6a2 + 3a3 , a (a 2) y= 2a3 3 a2 + 6a + 10 , a (a 2) 0 = 2, x=

2 a3 a2 + 4a + 6 a (a 2)

If a = 0, the system becomes x + y + 2z = 0, which is inconsistent (no solutions). If a = 1, the system becomes 3y + 4z = 1, which has innitely many solutions, y= 4 1 z, 3 3 x= 2 4 + z, 3 3 z=z . x + y = 1, 2x + 2y = 8, 2x + y z = 0,

If a = 2, the system becomes x + 5y + 6z = 2, 2x + 2y = 6, 2x + 3y + z = 9,

which is also inconsistent, as can be seen by observing that (x + 5y + 6z) 6(2x + 3y + z) = 2 18 = 13x 13y = 18, which contradicts the equation 2x + 2y = 6.

Answers and Hints

3.3.9 x y z = = = 22 36 23 312 22 37 .

213

3.3.10 Denote the addition operations applied to the rows by a1 , a2 , a3 , a4 and the subtraction operations to the columns by b1 , b2 , b3 , b4 . Comparing A and AT we obtain 7 equations in 8 unknowns. By inspecting the diagonal entries, and the entries of the rst row of A and AT , we deduce the following equations a1 = b1 , a2 = b2 , a3 = b3 , a4 = b4 , a1 b2 = 3, a1 b3 = 6, a1 b4 = 9. This is a system of 7 equations in 8 unknowns. We may let a4 = 0 and thus obtain a1 = b1 = 9, a2 = b2 = 6, a3 = b3 = 3, a4 = b4 = 0. 3.3.11 The augmented matrix of this system is y 1 0 0 1

1 y 1 0 0

0 1 y 1 0

0 0 1 y 1

1 0 0 1 y

Permute the rows to obtain

0 0 0 . 0 0

Performing R5 + yR1 R5 and R4 R1 R4 we get 1 0 0 0 0 0 1 0 y 1

1 0 0 1 y

0 1 0 y 1

0 y 1 1 0

1 1 y 0 0

y 0 1 0 1

0 0 0 . 0 0

1 1 y 1 y

y 0 1 y 1 y2

y 1 1 0

0 0 0 . 0 0

214
Performing R5 R2 R5 and R4 + yR2 R4 we get 1 0 0 0 0 0 0 1 0 0 0 y 1 1 y2 y

Appendix A

1 1 y y1 y1

y 0 1 y 1 y2

Performing R5 yR3 R5 and R4 + (y2 1)R3 R4 we get 1 0 0 0 0 0 0 1 1 0 0 0 y 1 0 0 1 y y3 + 2y 1 y2 + y 1

0 0 0 . 0 0

y 0 1 y2 + y 1 1 y y2

Performing R5 + R4 R5 we get

0 0 0 . 0 0

Upon factoring, the matrix is equivalent to 1 0 0 0 0 0 0 1 0 0 0 y 1 0 0

1 0 0 0 0

0 1 0 0 0

0 y 1 0 0

1 1 y y3 + 2y 1 y3 + y2 + 3y 2

y 0 1 y2 + y 1 0

0 0 0 . 0 0

1 1 y (y 1)(y2 + y 1) (y 2)(y2 + y 1)

y 0 1 y2 + y 1 0

Thus (y 2)(y2 + y 1)x4 = 0. If y = 2 then the system reduces to 1 0 0 0 0 0 0 1 2 0 1 5 0 1 0 0 0 2 1 0 0 1 2 5 0

0 0 0 . 0 0

0 0 0 . 0 0

Answers and Hints

In this case x5 is free and by backwards substitution we obtain x1 t x2 t x3 = t , x4 t t x5

215

t R.

If y2 + y 1 = 0 then the system reduces to

In this case x4 , x5 are free, and

1 0 0 0 0

0 1 0 0 0

0 y 1 0 0

1 1 y 0 0

y 0 1 0 0

0 0 0 . 0 0

Since y2 s s = (y2 + y 1)s ys, this last solution can be also written as x1 yt s x2 ys yt = ys t , x3 x4 s t x5

yt s x1 2 x2 y s yt s x3 = , ys t x4 s x5 t

(s, t) R2 .

(s, t) R2 .

Finally, if (y 2)(y2 + y 1) = 0, then x4 = 0, and we obtain

x1 0 x2 0 x3 = 0 . x4 0 0 x5

216
v v 4.1.1 No, since 1F = is not fullled. For example

Appendix A

4.1.2 We expand (1F + 1F )( + b ) in two ways, rst using 4.7 rst and then 4.8, obtaining a and then using 4.8 rst and then 4.7, obtaining (1F + 1F )( + b ) = (1F + 1F ) + (1F + 1F ) b = + + b + b , a a a a (1F + 1F )( + b ) = 1F ( + b ) + 1F ( + b ) = + b + + b . a a a a a + + + = + + + . a a b b a b a b + = + , a b b a

1 1 1 1 = . 1 = 1 0 1

We thus have the equality

Cancelling from the left and b from the right, we obtain a which is what we wanted to shew.

4.1.3 We must prove that each of the axioms of a vector space are satised. Clearly if (x, y, ) R+ R+ R then x y = xy > 0 and x = x > 0, so V is closed under vector addition and scalar multiplication. Commutativity and associativity of vector addition are obvious. Let A be additive identity. Then we need x A = x = xA = x = A = 1. Thus the additive identity is 1. Suppose I is the additive inverse of x. Then x I = 1 = xI = 1 = I = Hence the additive inverse of x is Now and (x y) = (xy) = x y = x y = ( x) ( y), ( + ) x = x+ = x x = (x ) (x ) = ( x) ( x), 1 . x 1 . x

whence the distributive laws hold. Finally, and and the last two axioms also hold.

1 x = x1 = x, ( x) = ( x) = (x ) = x = () x,

4.1.4 C is a vector space over R, the proof is trivial. But R is not a vector space over C, since, for example taking i as a scalar (from C) and 1 as a vector (from R) the scalar multiple i 1 = i R and so there is no closure under scalar multiplication. 4.1.5 One example is 0 0 0 0 1 1 1 1 3 , , , , , , , . (Z2 ) = 0 0 1 1 0 0 1 1 0 1 0 1 0 1 0 1

Addition is the natural element-wise addition and scalar multiplication is ordinary element-wise scalar multiplication.

Answers and Hints

4.1.6 One example is

217

Addition is the natural element-wise addition and scalar multiplication is ordinary element-wise scalar multiplication. 4.2.1 Take R and a b = X, a b 3d = 0, x c d a b = X, a b 3d = 0. y c d

0 0 0 1 1 1 2 2 2 2 (Z3 ) = , , , , , , , , . 0 1 2 0 1 2 0 1 2

Then

Observe that

a a + a a b b + b b + = + = x y . c c + c c d d d + d (a + a ) (b + b ) 3(d + d ) = (a b 3d) + (a b 3d ) = 0 + 0 = 0,

4.2.2 Take

meaning that + X, and so X is a vector subspace of R4 . x y a1 a2

Put s = a1 + a2 , t = b1 + b2 . Then

2a2 3b2 2a1 3b1 = = u , R. , v 5b2 5b1 a2 + 2b2 a1 + 2b1 a2 a1

a1 + a2 s 2(a1 + a2 ) 3(b1 + b2 ) 2s 3t + = u v = 5t X, 5(b1 + b2 ) (a1 + a2 ) + 2(b1 + b2 ) s + 2t s a1 + a2 since this last matrix has the basic shape of matrices in X. This shews that X is a vector subspace of R5 . 4.2.7 We shew that some of the properties in the denition of vector subspace fail to hold in these sets.

218

Appendix A

multiplication. 1 1 0 = , = . Then W, W but + = W as 1 1 = 1 = 0. Hence W is not closed x y x y 1 Take x 1 y 0 0 0 0 1 Take = x 0 under vector addition. 1 1 . Then Z but = x x 0 0 1 1 = 0 0 1 Z as 1 + (1)2 = 2 = 0. So Z is not closed 0

0 0 = , = 2. Then V but 2 = V as 02 + 22 = 4 = 1. So V is not closed under scalar x x Take x 2 1 0 0

under scalar multiplication.

4.2.8 Assume U1 U2 and U2 U1 . Take U2 \ U1 (which is possible because U2 v U1 ) and U1 \ U2 u + U , thenas is also in U the sum of two vectors in U must (which is possible because U1 U2 ). If u v u 1 1 1 also be in U1 giving + = U , u v u v 1 + U , thenas also in U the sum of two vectors in U must also be in U a contradiction. Similarly if u v v
2 2 2 1

giving

another contradiction. Hence either U1 U2 or U2 U1 (or possibly both). 4.2.9 Assume contrariwise that V = U1 U2 Uk is the shortest such list. Since the Uj are proper subspaces, y y x k > 1. Choose U1 , U2 Uk and choose U1 . Put L = { + | F}. Claim: L U1 = . For if x x L U then a F with = + a and so = a U , a contradiction. So L and U are disjoint. y u u u y 1 1 0x 1 0 0x We now shew that L has at most one vector in common with Uj , 2 j k. For, if there were two elements of F, a = b with + a, + b Uj , j 2 then y x y x contrary to the choice of . x (a b) = ( + a) ( + b) Uj , x y x y x

+ = U , u v u u 2

Conclusion: since F is innite, L is innite. But we have shewn that L can have at most one element in common with the Uj . This means that there are not enough Uj to go around to cover the whole of L. So V cannot be a nite union of proper subspaces. 4.2.10 Take F = Z2 , V = F F. Then V has the four elements 0 0 1 1 , , , , 0 1 0 1 with the following subspaces 0 0 0 0 1 1 V1 = , , V2 = , , V3 = , . 0 0 0 1 0 1

It is easy to verify that these subspaces satisfy the conditions of the problem.

Answers and Hints

4.3.1 If 1 1 1 a 0 + b 1 + c 1 = 0 , 1 0 0 a + b + c 0 b + c = 0 . 0 c

219

then

This clearly entails that c = b = a = 0, and so the family is free. 4.3.2 Assume

Then

1 0 1 1 1 1 0 1 1 1 a + b + c + d = . 0 0 1 1 1 1 1 1 1 0 a + b + c + d = 0, a + b c + d = 0, a b + c = 0, a b c + d = 0.

Subtracting the second equation from the rst, we deduce 2c = 0, that is, c = 0. Subtracting the third equation from the fourth, we deduce 2c + d = 0 or d = 0. From the rst and third equations, we then deduce a + b = 0 and a b = 0, which entails a = b = 0. In conclusion, a = b = c = d = 0. Now, put 1 1 1 1 1 1 2 1 1 1 x + y + z + w = . 0 1 1 1 1 1 1 1 1 1 x + y + z + w = 1, x + y z + w = 2, x y + z = 1, x y z + w = 1. Solving as before, we nd 1 1 1 1 1 1 1 1 1 1 1 2 2 + = . 2 2 1 1 1 0 1 1 1 1 1 1

Then

220
4.3.5 We have

Appendix A

a non-trivial linear combination of these vectors equalling the zero-vector. 4.3.7 Yes. Suppose that a + b 2 = 0 is a non-trivial linear combination of 1 and 2 with rational numbers a and b. If one of a, b is different from 0 then so is the other. Hence b a + b 2 = 0 = 2= . a b The sinistral side of the equality 2 = is irrational whereas the dextral side is rational, a contradiction. a 4.3.8 No. The representation 2 1 + ( 2) 2 = 0 is a non-trivial linear combination of 1 and 2. 4.3.9 1. Assume that If ac = 0, then a + b 2 + c 3 = 0, a, b, c, Q, a2 + b2 + c2 = 0.

(1 + 2 ) (2 + 3 ) + (3 + 4 ) (4 + 1 ) = 0 , v v v v v v v v

2b2 a2 3c2 = 3. b 2 = a c 3 2b2 = a2 + 2ac 3 + 3c2 2ac The dextral side of the last implication is irrational, whereas the sinistral side is rational. Thus it must be the case that ac = 0. If a = 0, c = 0 then 3 b , b 2+c 3=0 = c 2 and again the dextral side is irrational and the sinistral side is rational. Thus if a = 0 then also c = 0. We can similarly prove that c = 0 entails a = 0. Thus we have b 2 = 0, which means that b = 0. Therefore 2, 3} are linearly independent over Q. a + b 2 + c 3 = 0, a, b, c, Q, a = b = c = 0. 1+ 2 2 12 + 4 + 12 12 4 1 1 1 2 + 3+ 2 2 1 1 2+ 3. 2 2

This proves that {1,

2. Rationalising denominators, 1 2 + 1 2 12 2 = = = 4.3.10 Assume that aex + be2x + ce3x = 0. Then c = ae2x bex . Letting x +, we obtain c = 0. Thus and so aex + be2x = 0, b = aex . Again, letting x +, we obtain b = 0. This yields aex = 0.

Since the exponential function never vanishes, we deduce that a = 0. Thus a = b = c = 0 and the family is linearly independent over R. 4.3.11 This follows at once from the identity cos 2x = cos2 x sin2 x, which implies cos 2x cos2 x + sin2 x = 0.

Answers and Hints

4.4.1 Given an arbitrary polynomial p(x) = a + bx + cx2 + dx3 , we must shew that there are real numbers s, t, u, v such that p(x) = s + t(1 + x) + u(1 + x)2 + v(1 + x)3 . In order to do this we nd the Taylor expansion of p around x = 1. Letting x = 1 in this last equality, s = p(1) = a b + c d R. Now, Letting x = 1 we nd Again, Letting x = 1 we nd Finally, so we let v = d R. In other words, we have p(x) = a + bx + cx2 + dx3 = (a b + c d) + (b 2c + 3d)(1 + x) + (c 3d)(1 + x)2 + d(1 + x)3 . 4.4.2 Assume contrariwise that 0 1 1 = a + b . 1 0 1 1 1 1 a = 1, b = 1, a b = 1, p (x) = b + 2cx + 3dx2 = t + 2u(1 + x) + 3v(1 + x)2 . t = p (1) = b 2c + 3d R. p (x) = 2c + 6dx = 2u + 6v(1 + x). u = p (1) = c 3d R. p (x) = 6d = 6v,

221

Then we must have

1 0 1 0 1 which is impossible. Thus 1 is not a linear combination of 0 , 1 and hence is not in span 0 , 1 . 1 1 1 1 1 4.4.3 It is 1 a 0 0 0 + b 0 0 0 0 + c 1 1 1 a = c 0 c , b

i.e., this family spans the set of all skew-symmetric 2 2 matrices over R. 4.5.1 We have

a 0 1 3 2 2a 3b 5b = a 0 + b 5 , 2 1 a + 2b 0 1 a

222
so clearly the family

Appendix A

spans the subspace. To shew that this is a linearly independent family, assume that 0 0 1 3 0 2 a 0 + b 5 = 0 . 2 0 1 0 0 1

1 2 , 0 1 1

0 3 5 2 0

Then it follows clearly that a = b = 0, and so this is a linearly independent family. Conclusion: 1 0 2 3 0 , 5 1 2 1 0 is a basis for the subspace. 4.5.2 Suppose 0 = = a(1 + 2 ) + b(2 + 3 ) + c(3 + 4 ) + d(4 + 5 ) + f(5 + 1 ) v v v v v v v v v v

Since {1 , 2 , . . . , 5 } are linearly independent, we have v v v

(a + f)1 + (a + b)2 + (b + c)3 + (c + d)4 + (d + f)5 . v v v v v a + f = 0, a+b=0 b+c=0 c+d=0 d + f = 0.

Solving we nd a = b = c = d = f = 0, which means that the {1 + 2 , 2 + 3 , 3 + 4 , 4 + 5 , 5 + 1 } v v v v v v v v v v

are linearly independent. Since the dimension of V is 5, and we have 5 linearly independent vectors, they must also be a basis for V.

Answers and Hints

223

4.5.3 The matrix of coefcients is already in echelon form. The dimension of the solution space is n 1 and the following vectors in R2n form a basis for the solution space

1 1 0 . . . 0 a1 = , 1 1 0 . . . 0

1 0 1 . . . 0 a2 = , . . . , 1 0 1 . . . 0

an1

1 0 . . . 1 = 1 . 0 . . . 0 1

(The second 1 occurs on the n-th position. The 1s migrate from the 2nd and n + 1-th position on a1 to the n 1-th and 2n-th position on an1 .) 4.5.4 Let AT = A and BT = B be skew symmetric n n matrices. Then if R is a scalar, then (A + B)T = (A + B), so A + B is also skew-symmetric, proving that V is a subspace. Now consider the set of 1 + 2 + + (n 1) = (n 1)n 2

matrices Ak , which are 0 everywhere except in the ij-th and ji-spot, where 1 i < j n, aij = 1 = aji and i + j = k, 3 k 2n 1. (In the case n = 3, they are 0 1 0 1 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 , 0 0 0 0 1 0 1 0

, (n 1)n . 2

for example.) It is clear that these matrices form a basis for V and hence V has dimension 4.5.5 Take (, ) X2 and R. Then u v a b = , b + 2c = 0, u c d

a b = , b + 2c = 0. v c d

224
We have

Appendix A

and to demonstrate that + X we need to shew that (b + b ) + 2(c + c ) = 0. But this is easy, as u v (b + b ) + 2(c + c ) = (b + 2c) + (b + 2c ) = 0 + 0 = 0. Now

a + a b + b + = u v , c + c d + d

It is clear that

0 0 1 a a 0 2 0 b 2c = a + c + d = 0 1 0 c c 1 0 0 d d 1 0 0 0 2 0 , , 0 1 0 0 0 1

are linearly independent and span X. They thus constitute a basis for X. n(n + 1) matrices Eij Mn (F) for 1 i j n. 2 4.5.7 dim X = 2, as basis one may take { , }. v1 v2 4.5.8 dim X = 3, as basis one may take { , , }. v1 v2 v3 4.5.9 dim X = 3, as basis one may take { , , }. v1 v2 v3 4.5.6 As a basis we may take the c a f + 0 g 0 b

4.5.10 Let R. Observe that a b 0 d 0 0 and if a + b + c = 0,

d 0

a + d + g = 0,

a + b + c = 0,

c a + a 0 f = 0 g

b + b

d + d 0

a + d + g = 0, then

c + c f + f g + g

a + a + b + b + c + c = (a + b + c) + (a + b + c ) = 0 + 0 = 0, a + a + d + d + g + g = (a + d + g) + (a + d + g ) = 0 + 0 = 0, proving that V is a subspace. Now, a 0 0 a + b + c = 0 = a + d + g = a = b c, g = c b c b c b 1 1 = b = 0 d f d f 0 0 0 0 0 0 b+cd 0 g b + c d. Thus 0 1 0 +c 0 0 0 0 0 1 and

1 0 + d 0 0 0 1

0 1 0

0 0 + f 0 0 0 1

0 0 0

0 1 . 0

Answers and Hints

It is clear that these four matrices span V and are linearly independent. Hence, dim V = 4. 4.6.1 1. It is enough to prove that the matrix 1 1 A= 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2

225

is invertible. But an easy computation shews that 1 1 2 A = 1 1 1 1 1 1

1 1

1 1 1 1

1 1

whence the inverse sought is 1 1 1 1 = A= 4 4 1 1

1 1 = 4I4 , 1 1 1/4 1/4 1/4 1/4 1/4 1/4 1/4 1/4 1/4 1/4 . 1/4 1/4

1 1 1 1

A1

2. Since the k are four linearly independent vectors in R4 and dim R4 = 4, they form a basis for R4 . Now, we a want to solve x 1 y 1 A = z 1 1 w 1 1 1 1 1 1 1 1 1 x 1 1 y 2 = 1 z 1 1 w 1

1 1/4 1 1/4 = 1 1/4 1/4 1

and so

It follows that

1 1/4 x 2 1/4 y 1 =A = 1 1/4 z 1/4 1 w

1/4 1/4 1/4 1/4

1/4 1/4 1/4 1/4

1/4 1 5/4 1/4 2 1/4 . = 1/4 1 1/4 1/4 1 1/4

1 1 1 1 1 2 5 1 1 1 1 1 1 1 = + . 4 4 4 4 1 1 1 1 1 1 1 1 1 1

226
The coordinates sought are 5 1 1 1 , , , 4 4 4 4 3. Since we have 1 1 1 1 1 2 1 1 1 1 1 1 1 5 = + , 4 4 4 4 1 1 1 1 1 1 1 1 1 1 5 1 1 1 , , , 4 4 4 4 1 a 1 a+1 [3] a a1 a 1 1 a1 a a a1 1 a1 1 1 a 1 1 . .

Appendix A

the coordinates sought are

4.6.2 [1] a = 1, [2] (A(a))1

1 a1 1 = 1 a1 1 0 0 0 1

0 1a 1 a

0 1 0 1

1 a1 a 1 a a1

2+a

a+1

5.1.1 Let R. Then

x + a L y + b z + c

proving that L is a linear transformation.

a x L y + L b , c z

a b c x y z + a + b + c x + y + z c z

(x + a) (y + b) (z + c) (x + a) + (y + b) + (z + c) z + c

Answers and Hints

5.1.2 L(H + H ) = = = proving that L is linear. A1 (H + H )A1 A1 HA1 + (A1 H A1 ) L(H) + L(H ),

227

5.1.3 Let S be convex and let , b T (S). We must prove that [0; 1], (1 ) + b T (S). But since , b a a a belong to T (S), S, S with T () = , T () = b . Since S is convex, (1 ) + S. Thus x y x a y x y + ) T (S), T ((1 ) x y which means that that is, as we wished to show. x 5.2.1 Assume y ker (L). Then z (1 )T () + T () T (S), x y (1 ) + b T (S), a

that is

x 0 L y = 0 , 0 z x y z = 0, x + y + z = 0, z = 0.

and L is injective.

This implies that x y = 0 and x + y = 0, and so x = y = z = 0. This means that 0 ker (L) = 0 , 0 By the Dimension Theorem 244, dim Im (L) = dim V dim ker (L) = 3 0 = 3, which means that Im (L) = R3 and L is surjective.

5.2.2 1. If a is any scalar,

x x x + ax x x y y y + ay (x + ax ) + (y + ay ) x + y y y x + y = L + aL , +a = L + a = L = z z z + az z z xy x y (x + ax ) (y + ay ) w w w w w + aw

228
whence L is linear. 2. We have,

Appendix A

Thus dim ker (L) = 2. In particular, the transformation is not injective.

0 x 0 y x + y 0 = = x = y, x = y = x = y = 0 = ker (L) = : z R, w R . L = z z 0 xy w w

3. From the previous part, dim Im (L) = 4 2 = 2. Since Im (L) R2 and dim Im (L) = 2, we must have Im (L) = R2 . In particular, the transformation is surjective. a 5.2.3 Assume that b ker (T ), c

Then

1 1 a 0 = (a b) + b + c . 0 1 0 b 1 0 0 c 0 0 0 0 = a T b c 0 1 1 + bT + cT (a b)T 0 0 1 1 0 0

1 2 1 1 1 0 (a b) + b + c 1 0 1 0 0 0 a+b+c b c . a + b + c 0

Answers and Hints

It follows that a = 0 and b = c. Thus

229

and so dim ker (T ) = 1. By the Dimension Theorem 244,

0 :cR , ker (T ) = c 1 1

dim Im (T ) = dim V dim ker (T ) = 3 1 = 2. We readily see that 2 1 1 1 0 1 = + , 0 1 1 0 0 0 1 1 0 1 Im (T ) = span , . 1 1 0 0 x + 2y 0 x = L x + 2y = 0 . y 0 0 2 x = y . 1 y

and so

5.2.4 Assume that

Then x = 2y and so

This means that dim ker (L) = 1 and ker (L) is the line through the origin and (2, 1). Observe that L is not injective.

Then (x, y) R2 such that

a By the Dimension Theorem 244, dim Im (L) = dim V dim ker (L) = 2 1 = 1. Assume that b Im (L). c x + 2y a x = L x + 2y = b . y c 0

230
This means that a x + 2y 1 = (x + 2y) . = 1 b x + 2y 0 c 0 x y 0 x L = x + y = 0 . y 0 0 0 ker (L) = , 0

Appendix A

Observe that L is not surjective. 5.2.5 Assume that

Then x + y = 0 = x y, that is, x = y = 0, meaning that

and so L is injective.

Then (x, y) R2 such that

a By the Dimension Theorem 244, dim Im (L) = dim V dim ker (L) = 2 0 = 2. Assume that b Im (L). c x y a x L = x + y = b . y c 0 1 1 a x y = x + y . = 1 1 b x + y 0 0 0 c 1 1 , 1 1 0 0

This means that

Since

are linearly independent, they span a subspace of dimension 2 in R3 , that is, a plane containing the origin. Observe that L is not surjective. 5.2.6 Assume that x L y = z x y z 0 = . 0 y 2z

Answers and Hints

231

injective.

3 : z R . Hence dim ker (L) = 1, and so L is not Then y = 2z; x = y + z = 3z. This means that ker (L) = z 2 1 Now, if x x y z = L y = y 2z z a . b

Then

Now,

1 1 1 a x y z = x + y + z . = 2 1 0 y 2z b 1 1 1 3 2 = 2 1 0 1 1 , 1 0

and

are linearly independent. Since dim Im (L) = 2, we have Im (L) = R2 , and so L is surjective. 5.2.7 Assume that a 0 = tr c a c b d = c d b 1 = d 0 d b = a + d. d 0 0 + b 0 1 0 . 0 1 0 +c 1 0 0 , 0

Then a = d and so,

and so dim ker (L) = 3. Thus L is not injective. L is surjective, however. For if R, then = tr 0

5.2.8

1. Let (A, B)2 M22 (R), R. Then L(A + B) = = = = proving that L is linear. (A + B)T + (A + B) AT + BT + A + B AT + A + BT + B L(A) + L(B),

232

Appendix A

a 2. Assume that A = c

b ker (L). Then d 0 0 0 a = L(A) = c 0 b a + b d c 2a = b+c d b + c , 2d

whence a = d = 0 and b = c. Hence

and so dim ker (L) = 1.

0 ker (L) = span 1

1 , 0

3. By the Dimension Theorem, dim Im (L) = 4 1 = 3. As above, L(A) = b + c 2a b+c 2d 0

from where

2 a

0 0 + (b + c) 1 0 0 0 , 1 0

1 0 + d 0 0 1 0 , 0 0

0 , 2

5.2.9

Observe that proving the result.

2 Im (L) = span 0

0 . 2

(I T )2 = I 2T + T 2 = I 2T + T = I T, The inverse is I 1 T , for 2 (I + T )(I proving the claim. We have ker (T ) x T () ker (T ) x x 1 1 1 1 1 T ) = I + T T T 2 = I + T T T = I, 2 2 2 2 2

I() T () ker (T ) x x (I T )() ker (T ) x

5.3.1 Observe that

Im (I T ) . x

1 1 1 1 a 2 1 0 1 b = d + (2a c b) + (d 2a + 2c + b) + (a + c) . 0 1 1 1 c 0 0 0 1 d

Answers and Hints

Hence

233

a b T c d

1 0 1 0 d 0 + (2a c b) 1 + (d 2a + 2c + b) 0 + (a + c) 1 1 1 1 1 ab a b . a + 2d

1 1 dT + (2a c b)T 1 1

1 0 + (d 2a + 2c + b)T 1 0

1 1 1 2 + (a + c)T 1 0 0 0

This gives

The required matrix is therefore

1 1 0 T = 1 , 0 1 0

0 1 1 T = 1 , 0 0 0 1 1 1 1 1 0

0 0 0 T = 0 , 1 0 0 0 0 0

0 0 0 T = 0 . 0 2 1

This matrix has rank 2, and so dim Im (T ) = 2. We can use

2d a ab 0 2d b dimension theorem dim ker (T ) = 2. If 0 = T = a b , Hence the vectors in ker (T ) have the form c c a + 2d 0 d d 2 0 2 0 and hence we may take , as a basis for ker (T ). 0 1 1 0

0 0 . 2 1 1 , as a basis for Im (T ). Thus by the 1 1 1 0

234
5.3.2 1. Since the image of T is the plane x + y + z = 0, we must have

Appendix A

a + 0 + 1 = 0 = a = 1,

3 + b 5 = 0 = b = 2,

1 + 2 + c = 0 = c = 1.

1 2. Observe that 1 ker (T ) and so 1 1 0 T 1 = 0 . 0 1 Thus 2 1 T 0 = T 1 T 1 0 1 3 = , 1 2 5 1

1 0 T 1 = T 2 T 1 0

1 1 = , 1 2 1 1

1 0 = T T T 0 1 2 1 The required matrix is therefore 3 2 5 1 2 1

1 1 = . 1 0 1 1

1 0 . 1

Answers and Hints

5.3.3 1. Let R. We have u x T + v y = x + u T y + v x + u + y + v x + u y v 2(x + u) + 3(y + v) u+v x+y + uv xy 2u + 3v 2x + 3y

235

proving that T is linear. 2. We have

u x T + T , v y

dim ker (T ) = 0, and whence T is injective. 3. By the Dimension Theorem, dim Im (T ) = 2 0 = 2. Now, since 1 x+y 1 x = x + y , = T x y 1 1 y 3 2 2x + 3y whence 1 1 Im (T ) = span 1 , 1 . 3 2

x + y 0 0 x T = 0 x y = 0 x = y = 0, y 0 2x + 3y 0

4. We have

and

3 1 1 0 11/2 1 11 5 13 = T = 1 = 2 1 2 0 2 1 5/2 2 8 13/2 1 1 0 4 1 1 0 15/2 1 15 19 7 = T = 2 = 2 1 2 0 2 1 7/2 3 11 1 19/2 1 0

236
The required matrix is 11/2 5/2 13/2 15/2 7/2 19/2 .

Appendix A

5.3.4 The matrix will be a 2 3 matrix. In each case, we nd the action of L on the basis elements of R3 and express the result in the given basis for R3 . 1. We have 0 0 1 0 2 1 L 0 = , L 1 = , L 0 = . 1 0 3 1 0 0 2. We have 1 3 2 0 0 . 1

The required matrix is

The required matrix is

1 1 1 1 3 3 L 0 = , L 1 = , L 1 = . 3 3 2 0 0 1 1 3 3 3 3 . 2

3. We have

1 1 1 1 2 L 0 = = 2 + 3 = , 3 1 0 3 A 0 1 3 1 0 1 L 1 = = 0 + 3 = , 3 1 0 3 A 0 1 1 3 1 1 L 1 = = 1 + 2 = . 2 1 0 2 A 1 2 3 0 3 1 . 2

The required matrix is

Answers and Hints

237

2 5.3.5 Observe that Im (T ) = ker (T ) and so 3

Now

2 0 T = . 0 3 2 6 1 1 2 1 T = T 3 = 3T T = , 9 3 1 3 1 0 1 4 2 1 2 0 T = T 2 = T 2T = . 6 1 3 1 3 1 6 9 4 . 6

and

The required matrix is thus

5.3.6 The matrix will be a 1 4 matrix. We have

1 tr 0 0 tr 0 tr 0 tr 0 1 0

0 = 1, 0 1 = 0, 0

Thus

0 = 1. 1

0 = 0, 0

ML = (1 0 0 1). 5.3.7 First observe that ker (B) ker (AB) since X Mq1 (R), BX = 0 = (AB)X = A(BX) = 0. Now dim ker (B) = = = = = q dim Im (B) q rank (B) q rank (AB) q dim Im (AB) dim ker (AB) .

238
Thus ker (B) = ker (AB) . Similarly, we can demonstrate that ker (ABC) = ker (BC) . Thus rank (ABC) = = = = = dim Im (ABC) r dim ker (ABC) r dim ker (BC) dim Im (BC) rank (BC) .

Appendix A

6.2.1 This is clearly (1 2 3 4)(6 8 7) of order lcm(4, 3) = 12. 6.3.1 Multiplying the rst column of the given matrix by a, its second column by b, and its third column by c, we obtain abc abc = a2 a3 abc b2 b3 abc 2 . c 3 c

We may factor out abc from the rst row of this last matrix thereby obtaining 1 abc = abc det a2 a3 1 = det a2 a3 1 b2 b3 1 2 . c 3 c

Upon dividing by abc,

1 b2 b3

6.3.2 Performing R1 + R2 + R3 R1 we have

1 2 . c 3 c

a b c = det 2b 2c

2a bca 2c a+b+c bca 2c

2a 2b cab

Factorising (a + b + c) from the rst row of this last determinant, we have 1 1 1 . = (a + b + c) det 2b b c a 2b 2c 2c cab

a + b + c = det 2b 2c

a + b + c . 2b cab

Answers and Hints

Performing C2 C1 C2 and C3 C1 C3 ,

239

This last matrix is triangular, hence

1 = (a + b + c) det 2b 2c

0 b c a 0

. 0 c a b

= (a + b + c)(b c a)(c a b) = (a + b + c)3 , as wanted. 6.3.3 det A1 = det A = 540 by multilinearity. det A2 = det A1 = 540 by alternancy. det A3 = 3 det A2 = 1620 by both multilinearity and homogeneity from one column. det A4 = det A3 = 1620 by multilinearity, and det A5 = 2 det A4 = 3240 by homogeneity from one column. 6.3.5 From the given data, det B = 2. Hence det ABC = (det A)(det B)(det C) = 12, det 5AC = 53 det AC = (125)(det A)(det C) = 750, (det A3 B3 C1 ) = 6.3.6 Pick R \ {0, a11 , a22 , . . . , ann }. Put a11 a21 X = a31 . . . an1 0 a22 a32 . . . an2 0 0 a33 . . . an3 . . . . . . . . . . . . . . . . . . 0 0 0 . . . ann (det A)3 27 = . (det B)3 (det C) 16

and

Clearly A = X + Y, det X = (a11 )(a22 ) (ann ) = 0, and det Y = n = 0. This completes the proof. 6.3.7 No. 6.4.1 We have 4 2(1)1+2 det 7 6 1 + 5(1)2+2 det 7 9 3 1 + 8(1)2+3 det 4 9 3 6

0 Y = 0 . . . 0

a12 0 . . . 0

a13 a23 . . . 0

a1n a2n a3n . . .

det A

2(36 42) + 5(9 21) 8(6 12) = 0.

240
6.4.2 Simply expand along the rst row

Appendix A

a a det c

b c b det b a

b c + c det b a

a = a(a2 bc) b(ca b2 ) + c(c2 ab) = a3 + b3 + c3 3abc. c

6.4.3 Since the second column has three 0s, it is advantageous to expand along it, and thus we are reduced to calculate 1 3(1)3+2 det 2 1 1 0 0 1 1 1

Expanding this last determinant along the second column, the original determinant is thus

2 3(1)3+2 (1)(1)1+2 det 1

1 = 3(1)(1)(1)(1) = 3. 1

6.4.5 Expanding along the rst column,

0 1 1 1 x det b 0 0 b 0 0 0 c 0 c 1 1 1 1 1 1 +x det a 0 0 x det a 0 0 0 b 0 0 0 c 1 1 1 1 x det xabc xbc + x det a 0 0 a 0 0 0 c a det 0 0 0

1 x det x x

1 a 0 0

1 0 b 0

1 0 0 c

1 0 b

1 0 . 0

Answers and Hints

Expanding these last two determinants along the third row, 1 abc xbc + x det a 0 1 1 1 0 0 x det a 0 0 c 1 1 + xb det a 0 1 a b 1 0 0

241

1 abc xbc + xc det

1 0

= It follows that

abc xbc xca xab.

abc = x(bc + ab + ca), whence 1 bc + ab + ca 1 1 1 = = + + , x abc a b c as wanted. 6.4.7 Expanding along the rst row the determinant equals a a det 0 1 b 0 1 0 a + b det b 0 1 1 0 a 1 0 b 1

a ab det 1 2ab(a b),

b a + ab det 1 1

b 1

= as wanted. 6.4.8 Expanding along the rst row, the determinant equals a a det 0 c 0 d 0

b 0 + b det 0 c 0 d

a 0 c

b 0 . d

Expanding the resulting two determinants along the second row, we obtain a ad det c as wanted. 6.4.9 For n = 1 we have det(1) = 1 = (1)1+1 . For n = 2 we have 1 det 1 1 = 1 = (1)2+1 . 0 b a + b(c) det c d b = ad(ad bc) bc(ad bc) = (ad bc)2 , d

242
Assume that the result is true for n 1. Expanding the determinant along the rst column

Appendix A

1 1 0 det 0 . . . 0

1 0 1 0 . . . 0

1 0 0 1 0

. . .

1 0 0 0 . . . 1

. . .

1 0 0 0 0

0 1 1 det 0 . . . 0

0 0 1 0

. . . . . .

0 0 0 . . . 1

1 0 1 0 . . . 0

0 0 0 0 1 . . . . . . 1 0 0 0 0

= =

1(0) (1)(1)n (1)n+1 ,

1 1 0 1 det 0 . . . 0

giving the result.

6.4.10 Perform Ck C1 Ck for k [2; n]. Observe that these operations do not affect the value of the determinant. Then

1 n n det A = det n . . . n

n1 2n 0 0 . . . 0

n1 0 3n 0 . . . 0

n1 0 0 4n 0

. . .

. . .

n 1 0 0 . 0 0

Answers and Hints

Expand this last determinant along the n-th row, obtaining,

243

det A

(1)1+n n(n 1)(2 n)(3 n) 1 1 0 (2)(1) det 0 . . . 0 1 0 1 0 . . . 0 1 0 0 1 0 1 0 1 0 . . . 0 1 0 0 0 . . . 1

n 1 2 n 0 1+n (1) n det 0 . . . 0

n1 0 3n 0 . . . 0

n1 0 0 4n 0

. . .

n1 0 0 0 . . . 1

. . .

n 1 0 0 0 0

1 0 0 1

. . .

1 0 0 0 . . . 1

. . .

1 1 0 (n!) det 0 . . . 0 (n!)(1)n (1)n+1 n!,

. . .

. . .

1 0 0 0 0

1 0 0 0 0

= =

upon using the result of problem 6.4.9. 6.4.11 Recall that

n k

n nk

,
n

k=0

n k

= 2n

and
n

(1)k
k=0

n k

= 0,

if n > 0.

Assume that n is odd. Observe that then there are n + 1 (an even number) of columns and that on the same row, n n is on a column of opposite parity to that of nk . By performing C1 C2 + C3 C4 + + Cn Cn+1 C1 , k the rst column becomes all 0s, whence the determinant if 0 if n is odd.

244
6.4.15 I will prove that (b + c) det ab ac =
2

Appendix A

ab (a + c)2 bc

ac bc (a + b)2

Using permissible row and column operations, (b + c) det ab ac

2

= 2abc(a + b + c)3 .

ab (a + c)2 bc

ac bc (a + b)2

C1 +C2 +C3 C1

Pulling out a factor, the above equals b + c (a + b + c) det a + c a+b

ab ac b + 2bc + c 2 2 det ab a + 2ca + c bc 2 2 ac bc a + 2ab + b b2 + 2bc + c2 + ab + ac ab ac det ab + a2 + 2ca + c2 + bc a2 + 2ca + c2 bc ac + bc + a2 + 2ab + b2 bc a2 + 2ab + b2 (b + c)(a + b + c) ab ac det (a + c)(a + b + c) a2 + 2ca + c2 bc 2 2 (a + b)(a + b + c) bc a + 2ab + b

ab a2 + 2ca + c2 bc

ac bc a2 + 2ab + b2

and performing R1 + R2 + R3 R1 , this is

Factoring this is

2a + 2b + 2c (a + b + c) det a+c a+b

ab + a + 2ca + c + bc a2 + 2ca + c2 bc

ac + bc + a + 2ab + b bc 2 2 a + 2ab + b (a + b)(a + b + c) , bc a2 + 2ab + b2 a+b bc

which in turn is

2(a + b + c) (a + b + c) det a+c a+b 2

(a + c)(a + b + c) a2 + 2ca + c2 bc

2 (a + b + c) det a + c a+b

a+c a2 + 2ca + c2 bc

a2 + 2ab + b2

Answers and Hints

Performing C2 (a + c)C1 C2 and C3 (a + b)C1 C3 we obtain

245

(a + b + c) det a + c a+b
2

a c 0 a2 ab ac

2 a ab ac 0

a b

This last matrix we will expand by the second column, obtaining that the original determinant is thus

a + c (a + b + c)2 (a + c) det

a+b

a ab ac 2 + (a2 + ab + ac) det a+c 0

2 a ab ac

a b

This simplies to

(a + b + c)2 (a + c)(a + b)(a2 + ab + ac ) +(a2 + ab + ac)(a2 ab ac + bc) = = a(a + b + c)3 ((a + c)(a + b) a2 ab ac + bc) 2abc(a + b + c)3 ,

as claimed.

6.4.16 We have

246

Appendix A

a d det c b

b a d c

c b a d

d c b a

R1 +R2 +R3 +R4 R1 =

a + b + c + d d det c b

a+b+c+d a d c 1 a d c 1 b a d 1 b a d 1 c b a 0

a+b+c+d b a d

a + b + c + d c b a

1 d (a + b + c + d) det c b 1 d (a + b + c + d) c b 1 a d c

C4 C3 +C2 C1 C4 =

1 d (a + b + c + d)(a b + c d) c b

c b + a d b a + d c ad+cb 1 1 a d c b a d

0 1 1 1 1 b+a a a+d 1

R2 +R3 R2 , R4 +R3 R4 =

1 d + c (a + b + c + d)(a b + c d) c b+c 1

1 a+d d c+d 1 a+d c+d

0 0 1 0

(a + b + c + d)(a b + c d) d + c b+c

b + a a+d 0 db ca d b ca 1

C1 C3 C1 , C2 C3 C2 =

= = =

(a + b + c + d)(a b + c d)(d + c b a)(c a) (d b)(b + c a d) (a + b + c + d)(a b + c d) ((c a)(c a) + (c a)(d b) (d b)(c a) (d b)(b d)) (a + b + c + d)(a b + c d)((a c)2 + (b d)2 ).

d + c b a (a + b + c + d)(a b + c d) b+cad

0 (a + b + c + d)(a b + c d) d + c b a b+cad

b + a a+d

Answers and Hints

247

 Since

(a c)2 + (b d)2 = (a c + i(b d))(a c i(b d)),

the above determinant is then (a + b + c + d)(a b + c d)(a + ib c id)(a ib c + id). Generalisations of this determinant are possible using roots of unity. 7.2.1 We have 1 det(I2 A) = det 1 = ( 1)2 1 = ( 2), 1 1

whence the eigenvalues are 0 and 2. For = 0 we have

This has row-echelon form

1 0I2 A = 1 1 0 1 0 1 . 0

1 . 1

If

then a = b. Thus

1 a 0 = 0 b 0 1 a = a 1 b

1 and we can take as the eigenvector corresponding to = 0. Similarly, for = 2, 1 1 2I2 A = 1 If then a = 3b. Thus 1 0 1 0 3 . 0 3 , 3

which has row-echelon form

3 a 0 = 0 b 0

1 a = a 3 b

248

Appendix A

7.2.5

1 and we can take as the eigenvector corresponding to = 2. 3 We have det 2 1 2 3 2 1 2

det(I3 A)

= = = = The eigenvalues are 1, 1, 5. If = 1, 1 (I3 A) = 2 1

(2 3 4) + 2(2 2) + ( 1) ( 4)( + 1) 5( + 1) (2 4 5)( + 1) ( + 1)2 ( 5)

3 det 2

2 2 + 2 det 1

2 2 + det 1

3 2

2 4 2

1 a 0 2 b = 0 c 1 0

a = 2b + c

2 1 We may take as eigenvectors 1 , 0, which are clearly linearly independent. 0 1 If = 5, 5 (5I3 A) = 2 1 2 2 2 1 a 0 2 b = 0 0 c 5

2 a 1 = b + c . 0 1 b 0 c 1

a = c, b = 2c,

a 1 = c . 2 b 1 c

Answers and Hints

1 We may take as eigenvector 2 . 1

249

7.2.6 The characteristic polynomial of A must be 2 1, which means that tr (A) = 0 and det A = 1. Hence A a must be of the form b c , with a2 bc = 1, that is, a2 + bc = 1. a

7.2.7 We must shew that det(In A) = det(In A)T . Now, recall that the determinant of a square matrix is the same as the determinant of its transpose. Hence det(In A) = det((In A)T ) = det(IT AT ) = det(In AT ), n as we needed to shew. 7.3.1 Put 1 D= 0 0 1 , ,P = 1 2 1 P1 = 0 1 A10 = PD10 P1 = 1 0 1 0 1 4 1 . 1 1 1 = 0 1 1023 . 1024 0 . 1

We nd

Since A = PDP1

7.3.2

0 1 0 1024

3. For = 1 we have

2. ( 1)( + 1) = 0 = {1, 1}. 9 20

9 1. A has characteristic polynomial det 20

= ( 9)( + 9) + 80 = 2 1 = ( 1)( + 1). +9

2 so we can take as an eigenvector. 5 For = 1 we have 9 20 4 a a = 1 = 8a = 4b = a = b , 2 b b 9

4 a a = 1 = 10a = 4b = a = 2b , 5 b b 9

1 so we can take as an eigenvector. 2

250
4. We can do this problem in at least three ways. The quickest is perhaps the following.

Appendix A

Recall that a 2 2 matrix has characteristic polynomial 2 (tr (A)) + det A. Since A has eigenvalues 1 and 1, A20 has eigenvalues 120 = 1 and (1)20 = 1, i.e., the sole of A20 is 1 and so A20 has characteristic polynomial ( 1)2 = 2 2 + 1. This means that tr A20 = 2 and so tr A20 = 2.

The direct way would be to argue that 20 1

A20

2 5 2 5

1 0

1 1 0 2 0 , 1

1 1 0 0 1 2

0 2 5 1

2 1 5 2 1 1 2

and so a + d = 2. One may also use the fact that tr (XY) = tr (YX) and hence tr A20 = tr PD20 P1 = tr PP1 D20 = tr D20

= 2.

7.3.3 Put 1 D= 0 0 0 1 0 0 0 , 3 1 X = 0 0 1 1 0 1 1 . 1

Then we know that A = XDX1 and so we need to nd X1 . But this is readily obtained by performing R1 R2 R1 and R2 R3 R3 in the augmented matrix 1 0 0 getting 1 = 0 0 1 1 0 0 1 . 1 1 1 0 1 1 1 1 0 0 0 1 0 0 0 , 1

X1

Answers and Hints

Thus A = XDX1 1 1 1 1 0 1 1 0 0 0 0 1 1 0 0 0 1 0 4 4 . 3

251

0 1 0

0 1 0 0 0 3

1 1 0

0 1 1

7.3.4 The determinant is 1, A = A1 , and the characteristic polynomial is (2 1)2 . 7.3.6 We nd + 7 det(I2 A) = det 12 = 2 3 + 2 = ( 1)( 2). 10 6

2 A short calculation shews that the eigenvalue = 2 has eigenvector and that the eigenvalue = 1 has 3 3 eigenvector . Thus we may form 4

This gives

2 D= 0

0 , 1

2 P= 3

3 , 4

P1

4 = 3

3 . 1

2 A = PDP1 = An = PDn P1 = 3

3 2 0 4

0 4 3 1

3 8 2 + 9 = 12 2n 12 1

6 2 + 6 . n 92 8

7.4.1 The eigenvalues of A are 0, 1, and 2. Those of A2 are 0, 1, and 4. Hence, the characteristic polynomial of A2 is ( 1)( 4). 8.2.1 8.2.2 2a2 2a + 1 =
1 2

()2 + ()2 =

1 2

= 22 =

1 4

1 = = . 8

8.2.4 a = 1 or a = 8. 1 8.2.5 [A] 2( + ) 2, [B] + 1, [C] ( + + ) x y z x y z x y z 2 8.2.7 [F]. 0 , [G]. b , [H]. 2 0 , [I]. 0 .

8.2.3 0

8.2.6 [A]. 0 , [B]. 0 , [C]. 0 , [D]. 0 , [E]. 2(= 2 d ) c

252

Appendix A

8.2.8 Let the skew quadrilateral be ABCD and let P, Q, R, S be the midpoints of [A, B], [B, C], [C, D], [D, A], respec tively. Put = OX, where X {A, B, C, D, P, Q, R, S}. Using the Section Formula 8.4 we have x This gives = a + b, p 2 = b+ c, q 2 = c + d, r 2 = d + a. s 2

AC + BD = AD + BC = 2BC. 8.2.10 We have IA = 3IB IA = 3(IA + AB) = 3IA 3AB. Thus we deduce IA + 3IA = 3AB 4AI = 3AB AI = 3 AB. 4 4IA = 3AB

8.2.9 We have 2BC = BE + EC. By Chasles Rule AC = AE + EC, and AC + BD = AE + EC + BE + ED = But since ABCD is a parallelogram, AD = BC. Hence

This means that QP = RS and so PQRS is a parallelogram since one pair of sides are equal and parallel. BD = BE + ED. We deduce that AD + BC.

= a c , p q 2

= a c . s r 2

Similarly 1 JA = 3 JB

. Thus we take I such that AI = Now

4JA = AB AJ = 1 AB 4

3JA = JA AB

3JA = JB

3 AB and J such that AJ = 4 MA + 3MB = = =

1 AB. 4 MI + IA + 3IB

and 3MA + MB = = =

4MI,

4MI + IA + 3IB

8.2.11 Let G, O and P denote vectors from an arbitrary origin to the gallows, oak, and pine, respectively. The conditions of the problem dene X and Y , thought of similarly as vectors from the origin, by X = O + R(O G), Y = P R( P G), where R is the 90 rotation to the right, a linear transformation on vectors in the plane; the fact that R is 90 leftward rotation has been used in writing Y. Anyway, then X+Y R(O P ) O+ P = + 2 2 2

4MJ.

4MJ + 3JA + JB

3MJ + 3JA + MJ + JB

Answers and Hints

253

is independent of the position of the gallows. This gives a simple algorithm for treasure-nding: take P as the O + R(O) (hitherto) arbitrary origin, then the treasure is at . 2 8.3.1 a = 8.3.3 2 1 4 = = 2 + 3 = 2 + 3. r s p 1 1 5 = () + () a a i i a j j
1 2

a 8.3.4 Since a1 = i , a2 = j , we may write a from where the assertion follows. 8.3.5

+ b = 0 a

= = =

( + ) = a a b a 0 a () = 0 a a
2

= 0.

a Since = 0 , we must have a

= 0 and thus = 0. But if = 0 then + b = 0 a = = b = 0 = 0,

8.3.6 We must shew that But

since b = 0 .

9 (2 + 3)(2 3) = 4||||2 9||||2 = 4( ||||2 ) 9||||2 = 0. x y x y x y y y 4 8.3.7 We have R2 , ( b ) = 0. In particular, choosing = b , we gather v v a v a But the norm of a vector is 0 if and only if the vector is the 0 vector. Therefore b = 0 , i.e., = b . a a a b
2

(2 + 3)(2 3) = 0. x y x y

( b )( b ) = || b ||2 = 0. a a a

8.3.8 We have

= whence the result follows. 8.3.9 We have || + ||2 || ||2 u v u v =

a = 2 b + b b a a a
2

( b )( b ) a a

2 b + b a

v v u v u u v v u v = + 2 + ( 2 + ) u u = 4, u v

( + )( + ) ( )( ) u v u v u v u v

giving the result.

254
8.3.10 By denition
a proj x a a proj x 2 a a

Appendix A

proj

a = = Since 0 ()2 a x 2 x a

1 by the CBS Inequality, the result follows.

()2 a x , a 2 2 ||x|| a

a x x a x ||||2 2 a a

8.3.11 Clearly, if = 0 and = 0 then there are no solutions. If both = 0 and = 0, then the solution set is the a a x whole space R2 . So assume that = 0 . By Theorem 365, we may write = + with proj = || and . a u a v a x u v a Thus there are innitely many solutions, each of the form = + = x a + = + , x u v a v a v ||a||2 ||a||2 where . v a

2 1 8.4.1 Since = is normal to 2x y = 1 and b = is normal to x 3y = 1, the desired angle can be a 3 1 obtained by nding the angle between the normal vectors: (, b ) = arccos a 8.4.2 2(x 1) + (y + 1) = 0 or 2x + y = 1. 8.4.3 By Chasles Rule AA = AG + GA , BB = BG + GB , and CC = CG + GC . Thus 0 = = = = whence the result. 8.4.4 We have: The points F, A, D are collinear, and so FA is parallel to FD, meaning that there is k R \ {0} such that FA = kFD. Since the lines (AB) and (DC) are parallel, we obtain through Thales Theorem that FI = kFJ and FB = kFC. This gives FA FI = k(FD FJ) = IA = kJD. Similarly FB FI = k(FC FJ) = IB = kJC. Since I is the midpoint of [A, B], IA + IB = 0 , and thus k(JC + JD) = 0 . Since k = 0, we have JC + JD = 0 , meaning that J is the midpoint of [C, D]. Therefore the midpoints of [A, B] and [C, D] are aligned with F. AA + BB + CC a b b a = arccos 5 1 = arccos = . 4 10 2

GA + GB + GC ,

(GA + GB + GC) + (GA + GB + GC )

AG + GA + BG + GB + CG + GC

Answers and Hints

Let J be the intersection of the lines (EI) and (DC). Let us prove that J = J.

255

Since the points E, A, C are collinear, there is l = 0 such that EA = lEC. Since the lines (ab) and (DC) are parallel, we obtain via Thales Theorem that EI = lEJ and EB = lED. These equalities give EA EI = l(EC EJ ) = IA = lJ C, EB EI = l(ED EJ ) = IB = lJ D. Since I is the midpoint of [A, B], IA + IB = 0 , and thus l(J C + J D) = 0 . Since l = 0, we deduce J C + J D = 0 , that is, J is the midpoint of [C, D], and so J = J. 8.4.5 We have: By Chasles Rule AE = AF = 1 AC 4 3 AC 4 AB + BE = AD + DF = 1 AC 4 3 AC 4 ,

and

Adding, and observing that since ABCD is a parallelogram, AB = CD, AB + BE + AD + DF = AC

The last equality shews that the lines (BE) and (DF) are parallel. 1 Observe that BJ = 1 BC = 2 AD = AI = IA . Hence 2 proving that the lines (AB) and (IJ) are parallel. Observe that

BE = DF.

BE + DF = AD + DC AB AD .

BE + DF = AC AB AD

IJ = IA + AB + BJ = AB,

1 1 1 IE = IA + AE = DA + AC = CB + FC = CJ + FC = FC + CJ = FJ, 2 4 2 whence IEJF is a parallelogram. 1 ID 3

8.4.6 Since IE =

and [I, D] is a median of ABD, E is the centre of gravity of ABD. Let M be the midpoint of [B, D], and observe that M is the centre of the parallelogram, and so 2AM = AB + AD. Thus 2 1 1 AE = AM = (2AM) = (AB + AD). 3 3 3 To shew that A, C, E are collinear it is enough to notice that AE = 1 AC. 3

8.4.7 Suppose A, B, C are collinear and that

whence ( + ) b + = 0 and clearly ( + ) + = 0. Thus we may take = , = + , and = . a c Conversely, suppose that + b + = 0 , a c ++=0

||[A, B]|| = . Then by the Section Formula 8.4, ||[B, C]|| + c a b = , +

for some real numbers , , , not all zero. Assume without loss of generality that = 0. Otherwise we simply change the roles of , and and Then = ( + ) = 0. Hence a + b , + b = ( + ) = = a c c + and thus [O, C] divides [A, B] into the ratio , and therefore, A, B, C are collinear.

256
8.4.8 Put OX = for X {A, A , B, B , C, C , L, M, N, V}. Using problem 8.4.7 we deduce x + + = , v a a 0 + + = , v a a 0 + + = , v a a 0 1 + + = 0, 1 + + = 0, 1 + + = 0.

Appendix A

(A.1) (A.2) (A.3)

From A.2, A.3, and the Section Formula 8.4 we nd

b c b c = = l, whence ( ) l = b . In a similar fashion, we deduce c ( ) = , m c a

This gives

( ) l + ( ) + ( ) = 0 , m n ( ) + ( ) + ( ) = 0,

( ) = b . n a

and appealing to problem 8.4.7 once again, we deduce that L, M, N are collinear. 8.5.1 [A] AS, [B] AB. 1 1 1 = = ( + + )( + ) = = . i j k i j j i a 1 1 1 0 0 1
3 2 3 2

8.5.2 Put

Then either

or

3 a 3 a = = a 2

will satisfy the requirements. 8.5.3 The desired area is

3 a 3 = 2 a 0

3 2

PQPR

0 1 1 0 1 1

1 1 1

3.

8.5.4 It is not associative, since i ( i j ) = i k = j but ( i i ) j = 0 j = 0 . 8.5.6 2 b a

x x x x 8.5.5 We have = by letting = in 8.15. Thus 2 = 0 and hence = 0 . x x x x y x

Answers and Hints

8.5.7 () = () () () = () () = = 0. b x a x b x a b a x a x b b x a a b x a x b : R}. The answer is thus { x x a b = (ab)a + 6b + 2a c x 2 12 + 2 a ) + 6 + 3 c a b = (a c a y 2 18 + 3 a

257

8.5.8

8.5.9 Assume contrariwise that , b , are three unit vectors in R3 such that the angle between any two of them a c 2 < 1 , < 1 , and < 1 . Thus c . Then a b b c a is > 3 2 2 2 + + a b c
2

< =

1+1+1111 0,

c c a +2 b + 2 b + 2 a

c +

which is impossible, since a norm of vectors is always 0. 8.5.10 Take (, ) X2 and R. Then u v

proving that X is a vector subspace of Rn . 8.5.11 Since , b are linearly independent, none of them is 0 . Assume that there are (, , ) R3 such that a + b + b = 0 . a a (A.4) () = 0, taking the dot product of A.4 with yields 2 = 0, which means that = 0, since = 0. Since a a b a a a obtaining respectively, = 0 and = 0. This establishes linear Similarly, we take the dot product with b and a b independence. 8.5.12 Since = b , there are no solutions if b = 0. Neither are there solutions if = 0 and b = 0 . If a a x a a = = , then the solution set is the whole of R3 . Assume thus that = 0 and that and are linearly a b both a b 0 a b , , are linearly independent, and so they constitute a basis for R3 . Any R3 can be independent. Then a b a b x written in the form = + b + b . x a a We then have b = a x = a a a b + (( b ) () b ). a a 2 b + ( b ) a a a

( + ) = + = + = , a u v a u a v 0 0 0

= = = from where

2 b + ( 1) b = 0 , a a 1 which means that = 0 and = 2 , since , b , b are linearly independent. Thus a a a |||| 1 = x a a
2

a b a

b ( b ) a a a b,

in this last case.

a b

258
8.5.13 Let , , , be vectors in R3 and let R be a scalar. Then x y x y L((, ) + ( , )) x y x y = = L( + , + ) x x y y ( + ) k + h ( + ) x x y y

Appendix A

= 8.6.1 The vectors

= k + k + h + h x x y y L(, ) + L( , ) x y x y

and

a (a) 2a 0 1 = 1 a a0 0 (a) a 11 =0 2a 2a 0

are coplanar. A vector normal to the plane is

The equation of the plane is thus given by

2a a 2a = 2 . 1 0 3a a 2a a 2a x a 3a2 y 0 = 0, a za 2ax + 3a2 y az = a2 .

that is,

8.6.2 The vectorial form of the equation of the line is 1 1 = + t . r 0 2 1 1 1 1 Since the line follows the direction of 2, this means that 2 is normal to the plane, and thus the equation of 1 1 (x 1) 2(y 1) (z 1) = 0.

the desired plane is

Answers and Hints

8.6.3 Observe that (0, 0, 0) (as 0 = 2(0) = 3(0)) is on the line, and hence on the plane. Thus the vector 10 1 1 0 = 1 1 1 0

259

lies on the plane. Now, if x = 2y = 3z = t, then x = t, y = t/2, z = t/3. Hence, the vectorial form of the equation of the line is 1 1 0 = +t = t r 0 1/2 . 1/2 1/3 1/3 0 1 This means that 1/2 also lies on the plane, and thus 1/3 1 1 1/6 = 1 1/2 4/3 3/2 1/3 1 1 4 3 x y + z = 0. 6 3 2

is normal to the plane. The desired equation is thus

8.6.4 Put ax = by = cz = t, so x = t/a; y = t/b; z = t/c. The parametric equation of the line is 1/a x = t 1/b , t R. y 1/c z 1/a Thus the vector 1/b is perpendicular to the plane. Therefore, the equation of the plane is 1/c 1/a x 1 0 1/b y 1 = 0 , 0 1/c z1 x y z 1 1 1 + + = + + . a b c a b c bcx + cay + abz = ab + bc + ca.

or We may also write this as

260

Appendix A

a 8.6.5 A vector normal to the plane is a2 . The line sought has the same direction as this vector, thus the equation a2 of the line is x 0 a = + t 2 , y 0 a z a2 1

t R.

8.6.6 We have Hence if z = t,

x z y = 1 = 1 y = 1 = y = 2. x t 1 1 1 = + t . = 0 y 2 2 1 0 t z 2 1 1 1 0 = 1 2 1 (1) 1 1 1 = 0 1 1 1 2 1

8.6.7 The vector

lies on the plane. The vector

is normal to the plane. Hence the equation of the plane is

8.6.8 We have = i + 2 j and b = 2 k 3 i . By Theorem 398, we have c a a b c = b = 2 k + 3 i + i 2 j = 4 i 2 j 2 k . a c a 8.6.9 4x + 6y = 1 8.6.10 There are 7 vertices (V0 = (0, 0, 0), V1 = (11, 0, 0), V2 = (0, 9, 0), V3 = (0, 0, 8), V4 = (0, 3, 8), V5 = (9, 0, 2), V6 = (4, 7, 0)) and 11 edges (V0 V1 , V0 V2 , V0 V3 , V1 V5 , V1 V6 , V2 V4 , V3 V4 , V3 V5 , V4 V5 , and V4 V6 ). 8.7.2 Expand
n i=1

1 x 1 1 y = 0 = x + y z = 2. z+1 1

8.7.3 Observe that

ai

= 0.

n k=1

1 = n. Then we have
n 2 n

n2 =
k=1

=
k=1

(ak )

1 ak

a2 k
k=1 k=1

1 a2 k

giving the result.

Answers and Hints

8.7.4 Take (, ) X2 and R. Then u v

261

proving that X is a vector subspace of Rn . 8.7.5 Assume that

( + ) = + = 0 + 0 = 0, a v a u v a u

a a 1 1 + + k k = 0 . Taking the dot product with j and using the fact that i j = 0 for i = j we obtain a a a Since j = 0 = ||aj ||2 = 0, we must have j = 0. Thus the only linear combination giving the zero vector is the a trivial linear combination, which proves that the vectors are linearly independent. a 0 = 0 j = j j j = j ||aj ||2 . a a

8.7.6 This follows at once from the CBS Inequality by putting a1 1

and noticing that

1 a2 2 2 = , = v u . . . . . . an n n k2 = n(n + 1)(2n + 1) . 6

k=1

Bibliography

[Cul] CULLEN, C., Matrices and Linear Transformations, 2nd ed., New York: Dover Publications, 1990. [Del] DELODE, C., Gomtrie Afne et Euclidienne, Paris: Dunod, 2000. [Fad] FADDEV, D., SOMINSKY, I., Recueil dExercises dAlgbre Suprieure, 7th ed., Paris: Ellipses, 1994. [Hau] HAUSNER, M., A Vector Space Approach to Geometry, New York: Dover, 1998. [Lan] LANG, S., Introduction to Linear Algebra, 2nd ed., New York: Springer-Verlag, 1986. [Pro] PROSKURYAKOV, I. V., Problems in Linear Algebra, Moscow: Mir Publishers, 1978. [Riv] RIVAUD, J., Ejercicios de lgebra, Madrid: Aguilar, 1968. [Tru] TRUFFAULT, B., Gomtrie lmentaire: Cours et exercises, Paris: Ellipses, 2001.

262

GNU Free Documentation License

Version 1.2, November 2002 Copyright c 2000,2001,2002 Free Software Foundation, Inc. 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed.

Preamble
The purpose of this License is to make a manual, textbook, or other functional and useful document free in the sense of freedom: to assure everyone the effective freedom to copy and redistribute it, with or without modifying it, either commercially or noncommercially. Secondarily, this License preserves for the author and publisher a way to get credit for their work, while not being considered responsible for modications made by others. This License is a kind of copyleft, which means that derivative works of the document must themselves be free in the same sense. It complements the GNU General Public License, which is a copyleft license designed for free software. We have designed this License in order to use it for manuals for free software, because free software needs free documentation: a free program should come with manuals providing the same freedoms that the software does. But this License is not limited to software manuals; it can be used for any textual work, regardless of subject matter or whether it is published as a printed book. We recommend this License principally for works whose purpose is instruction or reference.

1. APPLICABILITY AND DEFINITIONS

This License applies to any manual or other work, in any medium, that contains a notice placed by the copyright holder saying it can be distributed under the terms of this License. Such a notice grants a world-wide, royalty-free license, unlimited in duration, to use that work under the conditions stated herein. The Document, below, refers to any such manual or work. Any member of the public is a licensee, and is addressed as you. You accept the license if you copy, modify or distribute the work in a way requiring permission under copyright law. A Modied Version of the Document means any work containing the Document or a portion of it, either copied verbatim, or with modications and/or translated into another language. A Secondary Section is a named appendix or a front-matter section of the Document that deals exclusively with the relationship of the publishers or authors of the Document to the Documents overall subject (or to related matters) and contains nothing that could fall directly within that overall subject. (Thus, if the Document is in part a textbook of mathematics, a Secondary Section may not explain any mathematics.) The relationship could be a matter of historical connection with the subject or with related matters, or of legal, commercial, philosophical, ethical or political position regarding them. The Invariant Sections are certain Secondary Sections whose titles are designated, as being those of Invariant Sections, in the notice that says that the Document is released under this License. If a section does not t the above denition of Secondary then it is not allowed to be designated as Invariant. The Document may contain zero Invariant Sections. If the Document does not identify any Invariant Sections then there are none. The Cover Texts are certain short passages of text that are listed, as Front-Cover Texts or Back-Cover Texts, in the notice that says that the Document is released under this License. A Front-Cover Text may be at most 5 words, and a Back-Cover Text may be at most 25 words. A Transparent copy of the Document means a machine-readable copy, represented in a format whose specication is available to the general public, that is suitable for revising the document straightforwardly with generic text editors or (for images composed of pixels) generic paint programs or (for drawings) some widely available drawing editor, and that is suitable for input to text formatters or for automatic translation to a variety of formats suitable for input to text formatters. A copy made in an otherwise Transparent le format whose markup, or absence of markup, has been arranged to thwart or discourage subsequent modication by readers is not Transparent. An image format is not Transparent if used for any substantial amount of text. A copy that is not Transparent is called Opaque. Examples of suitable formats for Transparent copies include plain ASCII without markup, Texinfo input format, LaTeX input format, SGML or XML using a publicly available DTD, and standardconforming simple HTML, PostScript or PDF designed for human modication. Examples of transparent image formats include PNG, XCF and JPG. Opaque formats include proprietary formats that can be read and edited only by proprietary word processors, SGML or XML for which the DTD and/or processing tools are not generally available, and the machine-generated HTML, PostScript or PDF produced by some word processors for output purposes only. The Title Page means, for a printed book, the title page itself, plus such following pages as are needed to hold, legibly, the material this License requires to appear in the title page. For works in formats which do not have any title page as such, Title Page means the text near the most prominent appearance of the works title, preceding the beginning of the body of the text. A section Entitled XYZ means a named subunit of the Document whose title either is precisely XYZ or contains XYZ in parentheses following text that translates XYZ in another language. (Here XYZ stands for a specic section name mentioned below, such as Acknowledgements, Dedications, Endorsements, or History.) To Preserve the Title of such a section when you modify the Document means that it remains a section Entitled XYZ according to this denition. The Document may include Warranty Disclaimers next to the notice which states that this License applies to the Document. These Warranty Disclaimers are considered to be included by reference in this License, but only as regards disclaiming warranties: any other implication that these Warranty Disclaimers may have is void and has no effect on the meaning of this License.

2. VERBATIM COPYING
You may copy and distribute the Document in any medium, either commercially or noncommercially, provided that this License, the copyright notices, and the license notice saying this License applies to the Document are reproduced in all copies, and that you add no other conditions whatsoever to those of this License. You may not use technical measures to obstruct or control the reading or further copying of the copies you make or distribute. However, you may accept compensation in exchange for copies. If you distribute a large enough number of copies you must also follow the conditions in section 3. You may also lend copies, under the same conditions stated above, and you may publicly display copies.

3. COPYING IN QUANTITY
If you publish printed copies (or copies in media that commonly have printed covers) of the Document, numbering more than 100, and the Documents license notice requires Cover Texts, you must enclose the copies in covers that carry, clearly and legibly, all these Cover Texts: Front-Cover Texts on the front cover, and Back-Cover Texts on the back cover. Both covers must also clearly and legibly identify you as the publisher of these copies. The front cover must present the full title with all words of the title equally prominent and visible. You may add other material on the covers in addition. Copying with changes limited to the covers, as long as they preserve the title of the Document and satisfy these conditions, can be treated as verbatim copying in other respects. If the required texts for either cover are too voluminous to t legibly, you should put the rst ones listed (as many as t reasonably) on the actual cover, and continue the rest onto adjacent pages. If you publish or distribute Opaque copies of the Document numbering more than 100, you must either include a machine-readable Transparent copy along with each Opaque copy, or state in or with each Opaque copy a computer-network location from which the general network-using public has access to download using public-standard network protocols a complete Transparent copy of the Document, free of added material. If you use the latter option, you must take reasonably prudent steps, when you begin distribution of Opaque copies in quantity, to ensure that this Transparent copy will remain thus accessible at the stated location until at least one year after the last time you distribute an Opaque copy (directly or through your agents or retailers) of that edition to the public. It is requested, but not required, that you contact the authors of the Document well before redistributing any large number of copies, to give them a chance to provide you with an updated version of the Document.

4. MODIFICATIONS
263

You may copy and distribute a Modied Version of the Document under the conditions of sections 2 and 3 above, provided that you release the Modied Version under precisely this License, with the Modied Version lling the role of the Document, thus licensing distribution and modication of the Modied Version to whoever possesses a copy of it. In addition, you must do these things in the Modied Version: A. Use in the Title Page (and on the covers, if any) a title distinct from that of the Document, and from those of previous versions (which should, if there were any, be listed in the History section of the Document). You may use the same title as a previous version if the original publisher of that version gives permission. List on the Title Page, as authors, one or more persons or entities responsible for authorship of the modications in the Modied Version, together with at least ve of the principal authors of the Document (all of its principal authors, if it has fewer than ve), unless they release you from this requirement. State on the Title page the name of the publisher of the Modied Version, as the publisher. Preserve all the copyright notices of the Document. Add an appropriate copyright notice for your modications adjacent to the other copyright notices. Include, immediately after the copyright notices, a license notice giving the public permission to use the Modied Version under the terms of this License, in the form shown in the Addendum below. Preserve in that license notice the full lists of Invariant Sections and required Cover Texts given in the Documents license notice. Include an unaltered copy of this License. Preserve the section Entitled History, Preserve its Title, and add to it an item stating at least the title, year, new authors, and publisher of the Modied Version as given on the Title Page. If there is no section Entitled History in the Document, create one stating the title, year, authors, and publisher of the Document as given on its Title Page, then add an item describing the Modied Version as stated in the previous sentence. Preserve the network location, if any, given in the Document for public access to a Transparent copy of the Document, and likewise the network locations given in the Document for previous versions it was based on. These may be placed in the History section. You may omit a network location for a work that was published at least four years before the Document itself, or if the original publisher of the version it refers to gives permission. For any section Entitled Acknowledgements or Dedications, Preserve the Title of the section, and preserve in the section all the substance and tone of each of the contributor acknowledgements and/or dedications given therein. Preserve all the Invariant Sections of the Document, unaltered in their text and in their titles. Section numbers or the equivalent are not considered part of the section titles. Delete any section Entitled Endorsements. Such a section may not be included in the Modied Version. Do not retitle any existing section to be Entitled Endorsements or to conict in title with any Invariant Section. Preserve any Warranty Disclaimers.

B.

C. D. E. F. G. H. I.

J.

K.

L. M. N. O.

If the Modied Version includes new front-matter sections or appendices that qualify as Secondary Sections and contain no material copied from the Document, you may at your option designate some or all of these sections as invariant. To do this, add their titles to the list of Invariant Sections in the Modied Versions license notice. These titles must be distinct from any other section titles. You may add a section Entitled Endorsements, provided it contains nothing but endorsements of your Modied Version by various partiesfor example, statements of peer review or that the text has been approved by an organization as the authoritative denition of a standard. You may add a passage of up to ve words as a Front-Cover Text, and a passage of up to 25 words as a Back-Cover Text, to the end of the list of Cover Texts in the Modied Version. Only one passage of Front-Cover Text and one of Back-Cover Text may be added by (or through arrangements made by) any one entity. If the Document already includes a cover text for the same cover, previously added by you or by arrangement made by the same entity you are acting on behalf of, you may not add another; but you may replace the old one, on explicit permission from the previous publisher that added the old one. The author(s) and publisher(s) of the Document do not by this License give permission to use their names for publicity for or to assert or imply endorsement of any Modied Version.

5. COMBINING DOCUMENTS
You may combine the Document with other documents released under this License, under the terms dened in section 4 above for modied versions, provided that you include in the combination all of the Invariant Sections of all of the original documents, unmodied, and list them all as Invariant Sections of your combined work in its license notice, and that you preserve all their Warranty Disclaimers. The combined work need only contain one copy of this License, and multiple identical Invariant Sections may be replaced with a single copy. If there are multiple Invariant Sections with the same name but different contents, make the title of each such section unique by adding at the end of it, in parentheses, the name of the original author or publisher of that section if known, or else a unique number. Make the same adjustment to the section titles in the list of Invariant Sections in the license notice of the combined work. In the combination, you must combine any sections Entitled History in the various original documents, forming one section Entitled History; likewise combine any sections Entitled Acknowledgements, and any sections Entitled Dedications. You must delete all sections Entitled Endorsements.

6. COLLECTIONS OF DOCUMENTS
You may make a collection consisting of the Document and other documents released under this License, and replace the individual copies of this License in the various documents with a single copy that is included in the collection, provided that you follow the rules of this License for verbatim copying of each of the documents in all other respects. You may extract a single document from such a collection, and distribute it individually under this License, provided you insert a copy of this License into the extracted document, and follow this License in all other respects regarding verbatim copying of that document.

7. AGGREGATION WITH INDEPENDENT WORKS

A compilation of the Document or its derivatives with other separate and independent documents or works, in or on a volume of a storage or distribution medium, is called an aggregate if the copyright resulting from the compilation is not used to limit the legal rights of the compilations users beyond what the individual works permit. When the Document is included in an aggregate, this License does not apply to the other works in the aggregate which are not themselves derivative works of the Document. If the Cover Text requirement of section 3 is applicable to these copies of the Document, then if the Document is less than one half of the entire aggregate, the Documents Cover Texts may be placed on covers that bracket the Document within the aggregate, or the electronic equivalent of covers if the Document is in electronic form. Otherwise they must appear on printed covers that bracket the whole aggregate.

8. TRANSLATION
Translation is considered a kind of modication, so you may distribute translations of the Document under the terms of section 4. Replacing Invariant Sections with translations requires special permission from their copyright holders, but you may include translations of some or all Invariant Sections in addition to the original versions of these Invariant Sections. You may include a translation of this License, and all the license notices in the Document, and any Warranty Disclaimers, provided that you also include the original English version of this License and the original versions of those notices and disclaimers. In case of a disagreement between the translation and the original version of this License or a notice or disclaimer, the original version will prevail. If a section in the Document is Entitled Acknowledgements, Dedications, or History, the requirement (section 4) to Preserve its Title (section 1) will typically require changing the actual title.

9. TERMINATION
You may not copy, modify, sublicense, or distribute the Document except as expressly provided for under this License. Any other attempt to copy, modify, sublicense or distribute the Document is void, and will automatically terminate your rights under this License. However, parties who have received copies, or rights, from you under this License will not have their licenses terminated so long as such parties remain in full compliance.

10. FUTURE REVISIONS OF THIS LICENSE

The Free Software Foundation may publish new, revised versions of the GNU Free Documentation License from time to time. Such new versions will be similar in spirit to the present version, but may differ in detail to address new problems or concerns. See http://www.gnu.org/copyleft/. Each version of the License is given a distinguishing version number. If the Document species that a particular numbered version of this License or any later version applies to it, you have the option of following the terms and conditions either of that specied version or of any later version that has been published (not as a draft) by the Free Software Foundation. If the Document does not specify a version number of this License, you may choose any version ever published (not as a draft) by the Free Software Foundation.

Que a quien robe este libro, o lo tome prestado y no lo devuelva, se le convierta en una serpiente en las manos y lo venza. Que sea golpeado por la parlisis y todos sus miembros arruinados. Que languidezca de dolor gritando por piedad, y que no haya coto a su agona hasta la ltima disolucin. Que las polillas roan sus entraas y, cuando llegue al nal de su castigo, que arda en las llamas del Inerno para siempre. -Maldicin annima contra los ladrones de libros en el monasterio de San Pedro, Barcelona.