Linear and Matrices

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

2 Linear Transformations and Matrices

2.1 Linear Transformations Denition Let V1 and V2 be vector spaces. A linear transformation is a function T : V1 V2 with the following properties: 1. For any v, w V1 we have T (v + w) = T (v) + T (w). 2. For any v V1 , r R we have T (rv) = rT (v). In other words, a linear transformation is a function between vector spaces which is compatible with addition and scalar multiplication. In many cases, the parentheses in the notation will be dropped and we simply write T v for T (v). The following proposition is easy but very useful: Proposition Let T : V1 V2 be a linear transformation and let 01 and 02 be the zero vectors in V1 and V2 . Then we have T (01 ) = 02 . proof Multiplying the scalar 0 by any vector yields the zero vector. Consequently, we have T (01 ) = T (001 ) = 0T (01 ) = 02 where the second equality follows from the second part of the denition of linearity. Denition A bijective linear transformation T : V1 V2 is called a linear isomorphism. (We shall usually suppress the adjective linear.) Proposition Let T : V1 V2 be a linear isomorphism and let T 1 : V1 V2 be its inverse. Then T 1 is also a linear transformation. proof By the denition of the inverse T T 1 = IdV2 and T 1 T = IdV1 . Let v, w V2 ; we may write v = T (T 1 (v)) and w = T (T 1 (w)). We have T 1 (v + w) = T 1 (T (T 1 (v)) + T (T 1 (w))) = (T 1 T )[T 1 (v) + T 1 (w)] = T 1 (v) + T 1 (w) where the second equality follows from the fact that T is linear. If we let r R be a scalar, then T 1 (rv) = T 1 (rT (T 1 (v))) = (T 1 T )[rT 1 (v)] = rT 1 (v). Again, the second equality follows from the fact that T is linear. 2.2 Examples of Linear Transformations We begin by giving an example of a function which is not a linear transformation. Let f : R2 R be the function given by the rule f (x, y) = x2 +y +1. Since f (0, 0) = 1 = 0 we conclude by the proposition above that f is not linear.

We now give some examples of linear transformations. Let T : R R be given by the rule T (x) = ax where a R. To verify that T is a linear transformation, we must rst check that for any x, y R T (x + y) = a(x + y) = ax + ay = T (x) + T (y). Furthermore, we must also check that T (rx) = a(rx) = r(ax) = rT (x) for any scalar r. Let T : R2 R2 be given by the rule T (x, y) = (ax + by, cx + dy) where a, b, c, d R. Then T is also a linear transformation (the proof is left as an exercise). Now we give a more exotic example. Let S be the vector space of sequences described above, and dene T : S S by the rule T ((s1 .s2 , s3 , . . . )) = (s2 , s1 , s3 , . . . ). In other words, we exchange the rst two terms of the sequence. We verify that T is a linear transformation. Given sequences s = (sn ), t = (tn ) S, we have T (s + t) = T ((s1 + t1 , s2 + t2 , . . . )) = (s2 + t2 , s1 + t1 , . . . ) = (s2 , s1 , . . . ) + (t2 , t1 , . . . ) = T (s) + T (t). For any scalar r we also have T (rs) = (rs2 , rs1 , rs3 , . . . ) = r(s2 , s1 , . . . ) = rT (s). Thus T is a linear transformation. 2.3 Basic Properties of Matrices Denition A matrix A is a rectangular array of real numbers a11 a21 A= . . .

a12 a22

... ...

a1n a2n . . . . amn

am1 am2 . . .

The real numbers in the matrix are called its entries. The rows of the matrix are the horizontal sequences of entries and the columns are the vertical sequences of entries. We will use aij to denote the entry of the matrix A in the ith row and the jth column. (Notice the convention that the row number

precedes the column number). A matrix with m rows and n columns is called an m n matrix. Consider the matrix 3 2 1 A= . 0 2 1 We have a11 = 3 a12 = 2 a13 = 1 a21 = 0 a22 = 2 a23 = 1 The matrix has two rows 3 2 1 and three columns 3 1 2 2 1 . 1 and 0 2 1

Thus A is a 2 3 matrix. An m 1 matrix is called a column vector and a 1 n matrix is called a row vector. An n n matrix is called a square matrix. We now describe some operations on matrices. Let M (m, n) denote the set of all m n matrices, and let A, B M (m, n) so that a11 . . A= .

...

am1 . . .

amn

a1n . . .

b11 . . . b1n . . . . . B= . . bm1 . . . . . . bmn

Then A + B M (m, n) is dened as the matrix with entries aij + bij , i.e. a11 + b11 . . . a1n + b1n . . . . . A+B = . . am1 + bm1 . . . amn + bmn If r R is a scalar, then we dene rA M (m, n) to be the matrix with entries raij ra11 . . . ra1n . . . rA = . . . . ram1 . . . ramn Taking these operations to be addition and scalar multiplication, we nd that M (m, n) has the structure of a vector space: Theorem The set M (m, n) of all m n matrices forms a vector space with respect to the operations dened above. The proof is left as an exercise. Remark Note that the space M (1, n) of row vectors is the same as Rn . Furthermore, the space M (m, 1) of column vectors is just Rm written vertically.

We will often take the liberty of writing elements of Rm as column vectors in M (m, 1). 2.4 Matrices and Linear Transformations The most important aspect of matrices is that they can be used to construct linear transformations. We describe how they may be used to transform vectors in Rn into vectors in Rm . We start with an illustrative example. Let A be the 3 2 matrix 2 1 A = 3 0 . 5 2 This matrix can be used to construct a function TA : R2 R3 , given by the rule 2x1 + x2 TA (x1 , x2 ) = 3x1 . 5x1 2x2 Note that we are writing the elements of R3 as column vectors. We encourage the reader to verify that TA is a linear transformation. Notice that TA (1, 0) equals the rst column of A and TA (0, 1) equals the second column. We generalize this now to arbitrary matrices. Let A be an m n matrix. We dene a function TA : Rn Rm by the rule
TA (x) = TA ((x1 , . . . , xn )) =

a11 x1 + a12 x2 + . . . + a1n xn a21 x1 + a22 x2 + . . . + a2n xn . . .

am1 x1 + am2 x2 + . . . + amn xn Again, we write the elements of Rm as column vectors. We prove that TA is linear. For any x, x Rn we have TA (x + x ) = TA ((x1 + x1 , . . . , xn + xn )) a11 (x1 + x1 ) + . . . + a1n (xn + xn ) . . = . am1 (x1 + x1 ) + . . . + amn (xn + xn )

a11 x1 + . . . + a1n xn a11 x1 + . . . + a1n xn . . + . . = . . am1 x1 + . . . + amn xn am1 x1 + . . . + amn xn = TA (x) + TA (x ).

Furthermore, for any scalar r we have TA (rx) = TA ((rx1 , . . . , rxn )) a11 rx1 + . . . + a1n rxn . . = . am1 rx1 + . . . + amn rxn a11 x1 + . . . + a1n xn . . = r . am1 x1 + . . . + amn xn = rTA (x). We have proven the following theorem: Theorem Let A be an m n matrix. Then the function TA : Rn Rm dened above is a linear transformation. The function TA is called the linear transformation associated to the matrix A. Our next goal is to classify linear transformations T : Rn Rm . We have already seen that m n matrices give many examples of such linear transformations. We shall prove that these are the only examples. First we introduce some notation. The vectors e1 = (1, 0, . . . , 0), e2 = (0, 1, 0 . . . , 0), . . . , en = (0, . . . , 0, 1) in Rn will be called the standard basis vectors. Generally, all the components of ei are zero except for the ith component which is one. Any vector x = (x1 , . . . , xn ) Rn may be written x = x1 e1 + x2 e2 . . . + xn en . If T : Rn Rm is a linear transformation then we have T (x) = T (x1 e1 + . . . + xn en ) = x1 T (e1 ) + . . . + xn T (en ). Consequently T is entirely determined by its values on the standard basis vectors, which we may write a11 a1n a21 a2n . . . . T (en ) = . . T (e1 ) = . . . . am1 amn Combining this with the previous formula, we obtain a11 a1n a21 a2n T (x) = x1 . + . . . + xn . . . . . am1
=

amn

a11 x1 + a12 x2 + . . . + a1n xn a21 x1 + a22 x2 + . . . + a2n xn . . .

am1 x1 + am2 x2 + . . . + amn xn

Consequently, T = TA for the matrix a11 . . A= .

...

am1 . . .

amn

a1n . . . .

We have proved the following theorem: Theorem Let T : Rn R be a linear transformation. Then T = TA for some A M (m, n). The jth column of A is equal to T (ej ) for j = 1, . . . , n. Putting the previous two theorems together, we obtain the following Corollary There is a one-to-one correspondence between linear transformations T : Rn Rm and m n matrices. Remark There is a more sophisticated version of this result. In the exercises, we shall see that the set of linear transformations from Rn to Rm forms a vector space L(Rn , Rm ). The ideas above may be used to show that L(Rn , Rm ) is isomorphic to M (m, n). 2.5 Composition and Matrix Multiplication Theorem Let S : V1 V2 and T : V2 V3 be linear transformations. Then T S : V1 V3 is also a linear transformation. proof Let v, w V1 . We have (T S)(v + w) = T (S(v + w)) = T (S(v) + T (w)) = T (S(v)) + T (S(w)) = (T S)(v) + (T S)(w) (the second and third equalities rely on the linearity of S and T ). If r is any scalar then (T S)(rv) = T (S(rv)) = T (rS(v)) = rT (S(v)) = r(T S)(v) (again, the second and third equalities use the linearity of S and T ). Therefore, T S is linear. Let A be an n k matrix and B an m n matrix. Let TA : Rk Rn and TB : Rn Rm denote the linear transformations associated to A and B. The theorem above implies that TB TA is linear. Using the results from 2.4, we conclude that TB TA = TC for some m k matrix C. How do we compute C from A and B? We derive a formula for C in terms of A and B. To prevent possible confusion, we now use e1 , . . . , ek to denote the standard basis for Rk and f1 , . . . , fn to denote the standard basis for Rn . The key to the formula is the fact that the columns of C are equal to the vectors TC (e1 ), . . . , TC (ek ). Lets

compute TC (e1 ) to see the general pattern: TC (e1 ) = (TB TA )(e1 ) = TB (TA (e1 )) a11 a21 . = TB . . an1 = a11 TB (f1 ) + a21 TB (f2 ) + . . . an1 TB (fn ) b1n b12 b11 b2n b22 b21 . + a21 . + . . . an1 . = a11 . . . . . . bm1

bm2

bmn

b11 a11 + b12 a21 + . . . + b1n an1 = b21 a11 + b22 a21 + . . . + b2n an1 . bm1 a11 + bm2 a21 + . . . + bmn an1 The same computation shows that b11 a1j + b12 a2j + . . . + b1n anj TC (ej ) = b21 a1j + b22 a2j + . . . + b2n anj . bm1 a1j + bm2 a2j + . . . + bmn anj We summarize this computation in the following proposition: Proposition Let A be an n k matrix and B am m n matrix. Let C be the matrix for the composite linear transformation TB TA : Rk Rm . This is a m k matrix with entry ai1 b1j + ai2 b2j + . . . ain bnj in the ith row and jth column. Exercises Transposes 1)Let T : R2 R2 be the linear transformation dened by the rule T (x, y) = (x + y, x + 2y). a)Show that T is injective. b)Show that T is surjective. c)Compute a formula for the inverse of T . 2)Let V and W be xed vector spaces. Let L(V, W ) be the set of all linear transformations from V to W . Given linear transformations T1 and T2 in L(V, W ) then we can dene T1 + T2 to be the linear transformation given by (T1 + T2 )(v) = T1 (v) + T2 (v) for all v V . Also if r R, then we can dene rT to be the linear transformation given by (rT )(v) = rT (v) for all v V . Show that with the above denitions of addition and scalar multiplication, L(V, W ) is a real vector space. 3)Let V and W be xed vector spaces. Let T : V W be a linear transformation. Let K be the subset of V given by K = {v V : T (v) = 0}

Show that K is also a vector space. This vector space is called the kernel of T and is denoted ker(T ). Let L be the subset of W given by L = {w W : v V T (v) = w}

Show that L is a vector space. This vector space is often called the image of T and is denoted im(T ). Hint: For these two questions, you basically need to check that these subsets are closed under addition and scalar multiplication.

You might also like