Solution: (a) S 1 = span. (b) S 2 = R n, x 1. x 1 + x 2 + x 3 + x 4 = 0. x 4 Solution: S 5 = x 2. x 3. (b) The standard basis vectors

Similar documents
Definition Suppose S R n, V R m are subspaces. A map U : S V is linear if

Advanced Linear Algebra Math 4377 / 6308 (Spring 2015) March 5, 2015

Solutions: We leave the conversione between relation form and span form for the reader to verify. x 1 + 2x 2 + 3x 3 = 0

x 1 + 2x 2 + 3x 3 = 0 x 1 + 2x 2 + 3x 3 = 0, x 2 + x 3 = 0 x 3 3 x 3 1

We see that this is a linear system with 3 equations in 3 unknowns. equation is A x = b, where

8 General Linear Transformations

x 2 For example, Theorem If S 1, S 2 are subspaces of R n, then S 1 S 2 is a subspace of R n. Proof. Problem 3.

Linear Algebra Practice Problems

can only hit 3 points in the codomain. Hence, f is not surjective. For another example, if n = 4

Practice Final Exam Solutions

2. (10 pts) How many vectors are in the null space of the matrix A = 0 1 1? (i). Zero. (iv). Three. (ii). One. (v).

TEST 1: Answers. You must support your answers with necessary work. My favorite number is three. Unsupported answers will receive zero credit.

S09 MTH 371 Linear Algebra NEW PRACTICE QUIZ 4, SOLUTIONS Prof. G.Todorov February 15, 2009 Please, justify your answers.

Math 110: Worksheet 3

1 Last time: inverses

The definition of a vector space (V, +, )

Solutions to Math 51 First Exam April 21, 2011

x y + z = 3 2y z = 1 4x + y = 0

MTH Midterm 2 Review Questions - Solutions

Fall TMA4145 Linear Methods. Exercise set Given the matrix 1 2

MATH 323 Linear Algebra Lecture 12: Basis of a vector space (continued). Rank and nullity of a matrix.

Rank and Nullity. MATH 322, Linear Algebra I. J. Robert Buchanan. Spring Department of Mathematics

Elementary Linear Algebra Review for Exam 2 Exam is Monday, November 16th.

Matrix invertibility. Rank-Nullity Theorem: For any n-column matrix A, nullity A +ranka = n

Algorithms to Compute Bases and the Rank of a Matrix

Notice that the set complement of A in U satisfies

Math 217 Midterm 1. Winter Solutions. Question Points Score Total: 100

Chapter 2. General Vector Spaces. 2.1 Real Vector Spaces

of A in U satisfies S 1 S 2 = { 0}, S 1 + S 2 = R n. Examples 1: (a.) S 1 = span . 1 (c.) S 1 = span, S , S 2 = span 0 (d.

Comps Study Guide for Linear Algebra

MATH SOLUTIONS TO PRACTICE PROBLEMS - MIDTERM I. 1. We carry out row reduction. We begin with the row operations

Math 308 Practice Test for Final Exam Winter 2015

Chapter 2 Subspaces of R n and Their Dimensions

LINEAR ALGEBRA REVIEW

Math 323 Exam 2 Sample Problems Solution Guide October 31, 2013

Math 265 Midterm 2 Review

Definitions for Quizzes

MTH 35, SPRING 2017 NIKOS APOSTOLAKIS

Midterm 1 Solutions Math Section 55 - Spring 2018 Instructor: Daren Cheng

is injective because f(a) = f(b) only when a = b. On the other hand, 1

MATH 20F: LINEAR ALGEBRA LECTURE B00 (T. KEMP)

Math 353, Practice Midterm 1

MATH PRACTICE EXAM 1 SOLUTIONS

Math 415 Exam I. Name: Student ID: Calculators, books and notes are not allowed!

1 Invariant subspaces

4.9 The Rank-Nullity Theorem

Linear Algebra II Lecture 8

Family Feud Review. Linear Algebra. October 22, 2013

MATH 2331 Linear Algebra. Section 2.1 Matrix Operations. Definition: A : m n, B : n p. Example: Compute AB, if possible.

MATH 235. Final ANSWERS May 5, 2015

Systems of Linear Equations

Chapter SSM: Linear Algebra. 5. Find all x such that A x = , so that x 1 = x 2 = 0.

Lecture 03. Math 22 Summer 2017 Section 2 June 26, 2017

Row Space and Column Space of a Matrix

Chapter 3. Directions: For questions 1-11 mark each statement True or False. Justify each answer.

Math 54 HW 4 solutions

Math 113 Winter 2013 Prof. Church Midterm Solutions

Math 110 Answers for Homework 6

National University of Singapore. Semester II ( ) Time allowed: 2 hours

1 Last time: multiplying vectors matrices

Solutions to Math 51 Midterm 1 July 6, 2016

What is on this week. 1 Vector spaces (continued) 1.1 Null space and Column Space of a matrix

LINEAR ALGEBRA BOOT CAMP WEEK 1: THE BASICS

MATH2210 Notebook 3 Spring 2018

MATH 2360 REVIEW PROBLEMS

And, even if it is square, we may not be able to use EROs to get to the identity matrix. Consider

1. Let V, W be two vector spaces over F and let T : V W be a set theoretic map. Prove that the following are equivalent: T (cu + v) = ct (u) + T (v)

JORDAN NORMAL FORM. Contents Introduction 1 Jordan Normal Form 1 Conclusion 5 References 5

SUMMARY OF MATH 1600

MATH 152 Exam 1-Solutions 135 pts. Write your answers on separate paper. You do not need to copy the questions. Show your work!!!

Math 205, Summer I, Week 3a (continued): Chapter 4, Sections 5 and 6. Week 3b. Chapter 4, [Sections 7], 8 and 9

Math 369 Exam #2 Practice Problem Solutions

Chapters 5 & 6: Theory Review: Solutions Math 308 F Spring 2015

Math 4377/6308 Advanced Linear Algebra

Dimension. Eigenvalue and eigenvector

Test 1 Review Problems Spring 2015

2: LINEAR TRANSFORMATIONS AND MATRICES

Basis, Dimension, Kernel, Image

Chapter 2: Matrix Algebra

1. Give the dimension of each of the following vector spaces over R.

Definitions. Main Results

Definition 3 A Hamel basis (often just called a basis) of a vector space X is a linearly independent set of vectors in X that spans X.

T ((x 1, x 2,..., x n )) = + x x 3. , x 1. x 3. Each of the four coordinates in the range is a linear combination of the three variables x 1

Linear Transformations: Kernel, Range, 1-1, Onto

General Vector Space (3A) Young Won Lim 11/19/12

Basis, Dimension, Kernel, Image

Answers in blue. If you have questions or spot an error, let me know. 1. Find all matrices that commute with A =. 4 3

Linear Algebra Lecture Notes-I

Lecture 3q Bases for Row(A), Col(A), and Null(A) (pages )

Math 242 fall 2008 notes on problem session for week of This is a short overview of problems that we covered.

Math 24 Winter 2010 Sample Solutions to the Midterm

Math 235: Linear Algebra

Linear transformations

Review Notes for Linear Algebra True or False Last Updated: February 22, 2010

Math 2174: Practice Midterm 1

2. Every linear system with the same number of equations as unknowns has a unique solution.

Math 123, Week 5: Linear Independence, Basis, and Matrix Spaces. Section 1: Linear Independence

Linear Algebra: Homework 7

Math 313 Chapter 5 Review

Math 33A Discussion Notes

Transcription:

.. Dimension In this section, we introduce the notion of dimension for a subspace. For a finite set, we can measure its size by counting its elements. We are interested in a measure of size on subspaces as well. Counting the number of elements will not suffice here since all nonzero subspaces have infinitely many elements. However, bases of subspaces are finite sets, of which we can measure the size. Definition... Suppose S R n is a subspace. The dimension of S, denoted dim(s), is the size of any basis of S. Example..2. What are the dimensions of the subspaces in Example.2.2? (a) S = span (b) S 2 = R n, (c) S = { }, ([ 2]), (d) S 4 = span( u,..., u k ), where ( u,..., u k ) is linearly independent, x (e) S 5 = x 2 x x + x 2 + x + x 4 =. x 4 ([ (a) is a basis for S 2]), so dim(s ) =. (b) The standard basis vectors.,.,...,. in Rn form a basis of R n, so dim(r n ) = n. You likely could have guessed this, given that R n is called n-dimensional Euclidean space. (c) ( ) is a basis for { }, so dim({ }) = the size of ( ) =. (d) ( u,..., u k ) is a basis for S 4, so dim(s 4 ) = k. (e) We can describe the solution to x + 2x 2 + x + 4x 4 = as x = 2x 2 x 4x 4 with x 2, x, x 4 free, so 2x 2 x 4x 4 2 4 S 5 = x 2 x = x 2 + x + x 4 x 4 2 4 = span,,. 2 4 Thus, dim(s 5 ) = because,, is linearly independent by Proposition.2.8.

2 Do you see any potential issues with Definition..? What if a subspace S had a basis of size 2 and a basis of size? Should dim(s) be 2 or? Could we compromise and say dim(s) = 2.5? What if S had a bases of size 2,, and 7? The point is that if a subspace S had two bases of different sizes, Definition.. would not make sense. Fortunately, this is never an issue. Theorem... Suppose S R n be a subspace. Any two bases of S have the same size. Proof. See Section.2, Problem 9b or Problem 5. Algorithms.2.4 and.2. give us a method for building up a basis for subspace S starting from the empty list and for removing vectors from a spanning list for S to get a basis for S. We now formalize the implications of these algorithms in Proposition..4 and Corollaries..6 and..5. In addition, we can say how many vectors we need to add or remove because we know each basis for S has size dim(s). Proposition..4. Suppose S R n is a subspace and u,..., u k S. () If ( u,..., u k ) is linearly independent, then dim(s) k vectors can be added to ( u,..., u k ) to make a basis for S. (2) If ( u,..., u k ) spans S, then k dim(s) vectors can be removed from ( u,..., u k ) to make a basis for S. Proof. These facts follows immediately from Proposition.2.6, Algorithm.2., and the fact that any basis of S must have size dim(s). Corollary..5. Any two of the following are sufficient for B S to be a basis for S: (i) span(b) = S. (ii) B is linearly independent. (iii) B = dim(s). Proof. Given (i) and (ii), B is a basis for S by definition. Given (i) and (iii), by Proposition..4, we can remove vectors from B to get a basis for S, so B is a basis for S. Given (ii) and (iii), by Proposition..4, we can add vectors to B to get a basis for S, so B is a basis for S. Corollary..6. Suppose S, S 2 are subspaces of R n with S S 2, then dim(s ) dim(s 2 ), with equality if and only if S = S 2. In particular, each subspace of R n except R n has dimension less than n. Proof. Let B be a basis for S, so dim(s ) = B. Since B is a linearly independent list of vectors in S S 2, we can add dim(s 2 ) B vectors to B to get a basis for S 2. By definition of dimension, dim(s ) dim(s 2 ). If dim(s ) = dim(s 2 ), then B is linearly independent and B = dim(s 2 ), so B is also a basis for S 2 by Corollary..5, forcing S 2 = span(b) = S. On the other hand, if S = S 2 then clearly, dim(s ) = dim(s 2 ). The final statement follows from the first with S 2 = R n. Now that we have the necessary definitions, we can state and prove one of the most important Theorems in linear algebra, the Rank-Nullity Theorem. For any linear map T : R n R m, intuitively, the more vectors that T sends to, the less vectors T hits. In other words, the larger ker(t ), the smaller range(t ) is. The Rank-Nullity Theorem makes this more precise using our measurement of size for a subspace, dimension. We begin by defining rank and nullity for maps and matrices and going through an example in order to motivate the result and clarify the proof.

Definition..7. For any linear map T : R n R m, define rank(t ) = dim(range(t )), nullity(t ) = dim(ker(t )), rank([t ]) = rank(t ), nullity([t ]) = nullity(t ). Example..8. Suppose T : R 5 R 4 is linear with [T ] = A = [ ] a... a 5 with 2 5 RREF(A) = 4 (a) Find a basis for range(a), and then find rank(t ), rank(a). (b) Find a basis for ker(a), and then find nullity(t ), nullity(a). (c) What is rank(t ) + nullity(t )? (a) The first two columns of RREF(A) are the pivot columns, so by Algorithm.2., ( a, a 2 ) is a basis for range(a). Hence, rank(t ) = rank(a) = 2. (b) Columns, 4, 5 are free, which means a solution description to A x = has x, x 4, x 5 free, and then x = 2x x 4 5x 5, x 2 = x x 4 5x 5. Thus, 2x x 4 5x 5 2 5 2 5 x x 4 5x 5 ker(t ) = x = x x 4 + + 5 = span,, 5. x 5 2 5 Since,, 5 is linearly independent - in fact, this list must be linearly independent by Proposition.2.8, it forms a basis for ker(t ). Hence, nullity(t ) = nullity(a) =. (c) Notice rank(t )+nullity(t ) = 2+ = 5, which is the dimension of the domain. The Rank-Nullity Theorem says that this is no coincidence. Theorem..9 (Rank-Nullity). Suppose T : R n R m is linear. Then, dim(range(t )) + dim(ker(t )) = n. Proof. Let [T ] = [ a... a n ] and R = RREF(A), which are both m n matrices. We have range(t ) = span( a,..., a n ) and ker(t ) = ker([t ]) = ker(r). Then, by Algorithm.2., range(t ) has a basis whose size is the number of pivot columns of R, so dim(range(t )) = # of pivot columns of R In addition, by Proposition.2.8, ker(t ) = ker(r) has a basis whose size is the number of free columns of R, so dim(ker(t )) = # of free columns of R

4 Since each column of R is either pivot or free, dim(range(t )) + dim(ker(t )) = # of columns of R = n. Example... Find linear maps T : R 4 R with each of the possible ranks and nullities. rank(t ) nullity(t ) Example [T ] range(t ) ker(t ) 4 { } R 4 span span,, 2 2 span, span, R span 4 Does not exist There does not exist a linear map T : R 4 R with rank(t ) = 4 because there is no 4-dimensional subspace of R. The Rank-Nullity Theorem limits the possibilities for range(t ) and ker(t ) for linear maps T : R n R m. In fact, it is the only limit on what range(t ) and ker(t ) can be, provided that they are subspaces, Proposition... Proposition... Given any subspaces S R n, S 2 R m satisfying dim(s ) + dim(s 2 ) = n, there exists a linear map T : R n R m so that Proof. Problem 5. ker(t ) = S, range(t ) = S 2. Exercises:. Determine the ranks and nullities of the following matrices, given in RREF. 2 (a) 5 7 4 (b) 2 2. What are the possible ranks and nullities for linear maps T : R R 4? Give an example for each possibility.

. Given an positive integers m, n, what are minimum and maximum possible values of rank(a) and nullity(a) for all m n matrices A? Describe examples to show these minimum and maximum values are possible. 4. Determine if a linear map T : R n R m with ker(t ) = S and range(t ) = S 2, with appropriate m, n exists. (a) S = span ([ 2, S 2 = span. 2]) (b) S = span 2, S 2 = span,. (c) S = span,, (d) S = span, 2, S 2 = span 2. S 2 = span,. (e) S = span,, S 2 = span,. 5 (f) S = span,, S 2 = span 2, 6 7. 4 8 (g) S = span 2, S 2 = span,,. 4 5. For each of the parts in Exercise (4) for which such a map exists, find T : R n R m with ker(t ) = S and range(t ) = S 2. I recommend finding a general construction first, Problem 5. Problems:. (2) Suppose S R n is a subspace. Show that the minimum size of a list of vectors which spans S is dim(s). This requires showing size dim(s) is possible, but any smaller size is impossible. 2. (2) Suppose S R n is a subspace. Show that the maximum size of a linearly independent list of vectors in S is dim(s). This requires showing size dim(s) is possible, but any larger size is impossible.. () Suppose A is an m n matrix. Show that every solution description to A x = has the same number of free variables. 4. () Suppose A is an m n matrix, and b R m. Show that if A x = b has a solution, then every solution description to A x = b has the same number of free variables. 5

6 5. () Suppose T : R n R m is linear map and S is a subspace of R n. Recall that T (S) = {T ( x) x S}. Show that dim(t (S)) dim(s). 6. () Suppose T : R n R m is an injective linear map and S is a subspace of R n. Recall that T (S) = {T ( x) x S}. Show that dim(t (S)) = dim(s). 7. () Suppose T : R n R m is a linear map satisfying dim(t (S)) = dim(s) for all subspaces S of R n. Show that T is injective. 8. Suppose S R m is a subspace. Show that there exists a linear map T : R n R m with range(t ) = S if and only if n dim(s). 9. () Suppose S R m is a subspace with dim(s) = n. Show that there exists an injective linear map T : R n R m with range(t ) = S.. () If S R n and S 2 R m are subspaces with dim(s ) = dim(s 2 ), show that there exists a linear map T : R n R m so that T (S ) = S 2.. The purpose of this problem is to give another proof of the Rank-Nullity Theorem. Suppose T : R n R m is linear. Let ( u,..., u k ) be a basis for ker(t ), and extend it to a basis ( u,..., u k, v,..., v r ) of R n. (a.) (2) Show that r + k = n. (b.) (2) Show that range(t ) = span(t ( v ),..., T ( v r )). (c.) () Show that (T ( v ),..., T ( v r )) is linearly independent. (d.) (2) Conclude that dim(ker(t )) + dim(range(t )) = n. 2. (2+) Suppose T : R n R m is linear. Let ( u,..., u k ) be a basis for ker(t ), and extend it to a basis ( u,..., u k, v,..., v r ) of R n. Use the Rank-Nullity Theorem to show that (T ( v ),..., T ( v r )) is linearly independent.. (2) Suppose A, C are m n matrices with A C. Show that rank(a) = rank(c). 4. (2+) For what values of n does there exists a map T : R n R n with ker(t ) = range(t )? Justify your answer. 5. Prove Proposition... 6. (2) Prove or give a counterexample: For any m n matrix A, rank(a) is the number of nonzero rows in RREF(A). 7. (2) Prove or give a counterexample: For any m n matrix A, nullity(a) is the number of rows of all zeros in RREF(A). 8. (+) Suppose S R n is a subspace. Show that the minimum number of relations for a relation form of S is n dim(s). This requires showing that n dim(s) relations is possible, but any less is impossible.

9. (4) If T, T 2 : R n R m are linear maps with range(t ) range(t 2 ), show that there exists a linear map U : R n R n so that T = T 2 U. 2. () If T, T 2 : R n R m are linear maps with range(t ) = range(t 2 ), show that there exists an invertible linear map U : R n R n so that T 2 = T U. 2. (4) If T, T 2 : R n R m are linear maps with ker(t ) ker(t 2 ), show that there exists a linear map U : R m R m so that T 2 = U T. 22. () If T, T 2 : R n R m are linear maps with ker(t ) = ker(t 2 ), show that there exists an invertible linear map U : R m R m so that T 2 = U T. 2. () If T, T 2 : R n R m are linear maps with range(t ) = range(t 2 ) and ker(t ) = ker(t 2 ), show that there exists an invertible linear map U : range(t ) range(t ) so that T 2 = U T. 24. () If T, T 2 : R n R m are linear maps with rank(t ) = rank(t 2 ), show that there exists an invertible linear map U : R m R m so that range(t 2 ) = range(u T ). 25. () If T, T 2 : R n R m are linear maps with rank(t ) = rank(t 2 ), show that there exists an invertible linear map W : R m R m so that ker(t 2 ) = ker(t W ). 26. () If T, T 2 : R n R m are linear maps with rank(t ) = rank(t 2 ), show that there exists invertible linear maps U : R n R n and W : R m R m so that T 2 = W T U. 27. (5) Suppose T : R n R n is a nilpotent linear map. Show that T n = O. 7