Math 110: Linear Algebra Homework #4: David Zywina
Math 110: Linear Algebra Homework #4: David Zywina
DAVID ZYWINA
2.2: The Matrix Representation of a Linear Transformation Problem 1. (a) True. This is a consequence of L(V, W ) being a vector space. (b) True. Suppose = {v1 , , vn } and = {w1 , , wm }. Let A = [T ] = [U] . Then for each vj ,
m
T (vj ) =
i=1
Ai,j wi = U(vj ).
Since T and U agree on they must be equal. False. By denition it will be a n m matrix. True. This is Theorem 2.8(a) True. This is proved in Theorem 2.7. False. If V = W then we cant have equality (since the functions in both sets have dierent domains, and hence are dierent).
Problem 3. We have a linear map T : R2 R3 dened by T (a1 , a2 ) = (a1 a2 , a1 , 2a1 + a2 ). We have bases = {(1, 0), (0, 1)} and = {(1, 2), (2, 3)} of R2 , and basis = {(1, 1, 0), (0, 1, 1), (2, 2, 3)} of R3 . After computing T (1, 0) and T (0, 1) in terms of the basis we nd 2 1 (1, 1, 0) + 0(0, 1, 1) + (2, 2, 3) T (1, 0) = (1, 1, 2) = 3 3 T (0, 1) = (1, 0, 1) = (1, 1, 0) + (0, 1, 1) + 0(2, 2, 3). Therefore 1 [T ] = 0 1 . 2 0 3
1 3
We now compute T (1, 2) and T (2, 3) in terms of the basis (using our earlier calculations). T (1, 2) = T (1, 0) + 2T (0, 1) 1 2 = (1, 1, 0) + 0(0, 1, 1) + (2, 2, 3) + 2 ((1, 1, 0) + (0, 1, 1) + 0(2, 2, 3)) 3 3 2 7 (1, 1, 0) + 2(0, 1, 1) + (2, 2, 3) = 3 3
Date: September 29th.
DAVID ZYWINA
T (2, 3) = 2T (1, 0) + 3T (0, 1) 1 2 = 2 (1, 1, 0) + 0(0, 1, 1) + (2, 2, 3) + 3 ((1, 1, 0) + (0, 1, 1) + 0(2, 2, 3)) 3 3 4 11 (1, 1, 0) + 3(0, 1, 1) + (2, 2, 3) = 3 3 Therefore
[T ] = 2
2 3
7 3
11 3 4 3
3 .
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
= 1 + x2 = 0 = 2x
1 1 0 0 [T ] = 0 0 0 2 , 0 1 0 0
where = {1, x, x2 } is the standard basis of P2 (R). Problem 6. Given vector spaces V and W over a eld F , let L(V, W ) be the set of linear transformations from V into W . We dene addition and scalar mulitplication in L(V, W ) as described on page 82. We must proof Theorem 2.7(b), which says that under these operations L(V, W ) is itself a vector space. The set L(V, W ) contains the zero transformation T0 . In Theorem 2.7(a) it is show that L(V, W ) is closed under addition and scalar multiplication. What is left to check is the vector space axioms. All of the axioms are easy to check, and they come down to the equivalent axiom of the vector space V which we will implicitly use. Take arbitrary S, T, U L(V, W ) and scalars a, b. To check that the following functions are equal it suces to show that they agree on an arbitrary element v V . (see page 7 for the axioms) (VS 1) : (S + T )(v) = S(v) + T (v) = T (v) + S(v) = (T + S)(v)
(VS 2) : ((S + T ) + U)(v) = = = = = (VS 3) : (T + T0 )(v) = T (v) + T0 (v) = T (v) + 0 = T (v) (VS 4) : The map B := T is in L(V, W ), since we know L(V, W ) is closed under scalar multiplication. (T + B)(v) = T (v) + B(v) = T (v) T (v) = 0 = T0 (v) (VS 5) : (1 T )(v) = 1 T (v) = T (v) (VS 6) : ((ab)T )(v) = (ab)T (v) = a(bT (v)) = a(bT )(v) = (a(bT ))(v) (VS 7) : (a(S+T ))(v) = a(S+T )(v) = a(S(v)+T (v)) = aS(v)+aT (v) = (aS)(v)+(aT )(v) = (aS+aT )(v) (VS 8) : ((a + b)T )(v) = (a + b)T (v) = aT (v) + bT (v) = (aT )(v) + (bT )(v) = (aT + bT )(v) Problem 9. In this exercise V = C, a real vector space. Dene T : V V by T (z) = z. Given any complex numbers z1 = x1 + iy1 , z2 = x2 + iy2 V (where x1 , x2 , y1 , y2 R) and scalar c R, we have T (cz1 +z2 ) = T ((cx1 +x2 )+i(cy1 +y2 )) = (cx1 +x2 )i(cy1 +y2 ) = c(x1 iy1 )+(x2 iy2 ) = cT (z1 )+T (z2 ). Since z1 , z2 and c were arbitrary, we have that T is a linear map (of real vector spaces). Since T (1) = 1 and T (i) = i, we have [T ] = 1 0 0 1 , where = {1, i}. (S + T )(v) + U(v) (S(v) + T (v)) + U(v) S(v) + (T (v) + U(v)) S(v) + (T + U)(v) (S + (T + U))(v)
Problem 10. We have a basis = {v1 , , vn } of V , and a linear map T : V V such that T (vj ) = vj + vj1 for j = 1, 2, , n (where we have v0 = 0 by convention). The n n matrix A = [T ] is dened to be the unique matrix such that
n
T (vj ) =
i=1
Ai,j vi for 1 j n.
DAVID ZYWINA
This is a matrix with 1s on the main diagonal, 1s on the diagonal above that, and 0s 1 1 1 1 .. . 1 everywhere else: .. . 1 1 Problem 11. Choose a basis {v1 , , vk } for the vector space W . We can extend this to a basis = {v1 , , vk , vk+1 , , vn } of V . When 1 j k, by construction vj W and hence T (vj ) W (this is where we use that W is T -invariant). Since T (vj ) W , we can write
k
T (vj ) =
i=1
Mi,j vi ,
However we know that the ([T ] )i,j are the unique scalars such that
n
T (vj ) =
i=1
([T ] )i,j vi .
So by comparison wtih (1) we see that ([T ] )i,j = 0, when 1 j k and k + 1 i n. This is the same as saying that [T ] is of the form A B O C ,
where A is a k k matrix and O is the (n k) k zero matrix. Problem 14. Suppose that there we have a linear combination a1 T1 + an Tn = 0, with a1 , , an R. Fix a j {1, , n}. Dene the polynomial g(x) = (a1 T1 + an Tn )(xj ), which has constant term j!aj . Thus g(0) = j!aj . However, by assumption a1 T1 + an Tn = 0 so g(x) = 0. This implies that j!aj = 0, which shows that aj = 0. Since j was arbitrary we have shown that a1 = = an = 0. Therefore T1 , , Tn are linearly independent. Problem 16. Let n = dim V = dim W . Let T : V W be a linear map. Choose a basis u1 , , uk of N(T ). We can then extend this to some basis = {v1 , , vnk , u1, , uk }
of V . Claim: T (v1 ), , T (vnk ) are linearly independent Proof. Suppose we have a linear combination a1 T (v1 ) + + ank T (vnk ) = 0. This implies that T (a1 v1 + +ank vnk ) = 0, since T is linear. Thus a1 v1 + +ank vnk N(T ), so we can write it as (2) a1 v1 + + ank vnk = b1 u1 + + bk uk
for some scalars bj . Since is a basis of V (and in particular linearly independent), we nd that all of coecients of (2) are zero. Therefore a1 = = ank = 0, which show the desired independence. Dene w1 = T (v1 ), , wnk = T (vnk ). These vectors are linearly independent by our claim, and we can extend to a basis of = {w1 , , wnk , wnk+1, , wn }. By construction, T (vi ) = wi and T (ui) = 0. Therefore we have 1 1 .. . 1 [T ] = , 0 .. . 0 where the rst n k diagonal elements are 1s and all the other entries are zeros. In particular, the matrix [T ] is diagonal. We now prove the dimension theorem in the case n = dim V = dim W . Let T : V W be a linear map. From above we know that we can choose a bases of V and of W such that the matrix A := [T ] is diagonal as above. After choosing these bases we see that the map T : V W can be viewed as the linear map LA : F n F n (this is the content of the next section). In particularly, it is apparent that nullity(T ) = nullity(LA ) = nullity(A) and rank(T ) = rank(LA ) = rank(A) . Thus it suces to prove that for an n n diagonal matrix A, nullity(A) + rank(A) = n. For a diagonal matrix A, let r be the number of zeros on the main diagonal of A. It is then easy to see that nullity(A) = r and rank(A) = n r. Thus nullity(A) + rank(A) = k + (n k) = n.
DAVID ZYWINA
2.3: Composition of linear transformations and matrix multiplication Problem 1. (a) False. This is an incorrect statement of Theorem 2.11. If dim W = dim Z then the expression [T ] [U] doesnt make sense [you cannot multiply matrices unless their dimensions match up]. Note that even if dim W = dim Z then the statement is still false, since in general matrices dont commute. (b) True. This is Theorem 2.14. (c) False. We have bases and of V and W respectively. Since U is a map from W to Z, the symbol [U] will not always make sense (since (resp. ) need not be a basis of W (resp. Z) ). (d) True. If = {v1 , , vn } then, Iv (vi ) = vi . This shows that [IV ] = I. (e) False. If V = W then T 2 is not dened. However, even if V = W the statement may fail. Consider the easiest counter example: dene T : R R by T (x) = x. We have bases = {2} and = {1} fo R. Then [T 2 ] = [T ] = [2], but ([T ] )2 = [2]2 = [4]. 1 0 2 (f) False. Let A = . A =I 0 1 (g) False. If T is a map of the form F n F m then Theorem 2.15 shows that T = LA for some matrix A. Otherwise, it is unclear what LA means (for example, what does LA : P2 (R) P2 (R) mean?). 0 1 . A2 = 0 (h) False. Let A = 0 0 (i) True. This is Theorem 2.15(c). (assuming A and B have the same dimensions!) (j) True. This is the denition of I (see page 89). Problem 3. We have linear maps T : P2 (R) P2 (R) and U : P2 (R) R3 dened by T (f (x)) = f (x)(3 + x) + 2f (x) and U(a + bx + cx2 ) = (a + b, c, a b). We also have bases = {1, x, x2 } and = {(1, 0, 0), (0, 1, 0), (0, 0, 1)} of P2 (R) and R3 respectively. (a) U(1) = (1, 0, 1) = (1, 0, 0) + 0 (0, 1, 0) + (0, 0, 1) U(x) = (1, 0, 1) = (1, 0, 0) + 0 (0, 1, 0) (0, 0, 1) U(x2 ) = (0, 1, 0) = 0 (1, 0, 0) + (0, 1, 0) + 0 (0, 0, 1) Thus 1 1 0 [U] = 0 0 1 . 1 1 0 T (1) = 2 = 2 1 + 0x + 0x2 T (x) = 3 + 3x = 3 1 + 3x + 0x2 T (x2 ) = 6x + 4x2 = 0 1 + 6x + 4x2
Thus
2 3 0 [T ] = 0 3 6 . 0 0 4 UT (1) = U(2) = (2, 0, 2) = 2(1, 0, 0) + 0(0, 1, 0) + 2(0, 0, 1) UT (x) = U(3 + 3x) = (6, 0, 0) = 6(1, 0, 0) + 0(0, 1, 0) + 0(0, 0, 1) U(x2 ) = U(6x + 4x2 ) = (6, 4, 6) = 6(1, 0, 0) + 4(0, 1, 0) 6(0, 0, 1)
Thus
2 6 6 [UT ] = 0 0 4 . 2 0 6 We now verify Theorem 2.11, which says that [UT ] = [U] [T ] . 1 1 0 2 3 0 2 6 6 [U] [T ] = 0 0 1 0 3 6 = 0 0 4 = [UT ] . 1 1 0 0 0 4 2 0 6 3 (a) Since h = 3 2x + x2 , we have [h(x)] = 2 . Since U(h(x)) = (4, 2, 2) we 1 1 1 . have [U(h(x))] = 5 We now verify Theorem 2.14, which says that [U(h(x))] = [U] [h(x)] . 1 1 0 3 1 0 0 1 2 = 1 = [U(h(x))] . [U] [h(x)] = 1 1 0 1 5 Problem 8. We now prove Theorem 2.10 (see page 87 for the statements). To show that two functions are the same it suces to show that they agree on their domains. (a) Take any v V . Then (T (U1 + U2 ))(v) = T ((U1 + U2 )(v)) = T (U1 (v) + U2 (v)) = T (U1 (v)) + T (U2 (v)) = (T U1 )(v) + (T U2 )(v) = (T U1 + T U2 )(v). Since this holds for all v V , we have T (U1 + U2 ) = T U1 + T U2 . Take any v V . Then ((U1 +U2 )T )(v) = (U1 +U2 )(T (v)) = U1 (T (v))+U2 (T (v)) = (U1 T )(v)+(U2T )(v) = (U1 T +U2 T )(v). Since this holds for all v V , we have (U1 + U2 )T = U1 T + U2 T . (b) This part does not use linear algebra. It is a consequence of the general fact that the composition of functions is associative. (c) For any v V , (T I)(v) = T (I(v)) = T (v).
DAVID ZYWINA
Since T I and T agree on their domains, they must be equal. Similarily we can prove that IT = T . (d) Take any v V . (a(U1 U2 ))(v) = a((U1 U2 )(v)) = a(U1 (U2 (v))) ((aU1 )U2 )(v) = (aU1 )(U2 (v)) = a(U1 (U2 (v))) (U1 (aU2 ))(v) = U1 ((aU2 )(v)) = U1 (aU2 (v)) = a(U1 (U2 (v))) This shows that the functions a(U1 U2 ), (aU1 )U2 and U1 (aU2 ) agree on all elements of V , therefore these functions are equal. Theorem. (Generalization of Theorem 2.10) Let V, W, Z, X be vector spaces. (a) Let U1 , U2 L(V, W ) and T L(W, Z), T (U1 + U2 ) = T U1 + T U2 . Let U1 , U2 L(W, Z) and T L(V, W ), (U1 + U2 )T = U1 T + U2 T. (b) For U2 L(V, W ), U1 L(W, Z) and T L(Z, X), T (U1 U2 ) = (T U1 )U2 . T IV = IW T = T. (Where IV and IW are the identity maps on V and W respectively.) (d) For U2 L(V, W ), U1 L(W, Z) and a scalar a, a(U1 U2 ) = (aU1 )U2 = U1 (aU2 ). The proofs given before carry over verbatim in this more general situation. Problem 12. We are give linear maps T : V W and U : W Z. (a) Suppose that UT is one-to-one, we will now show that T is one to one. Take any v N(T ). Then (UT )(v) = U(T (v)) = U(0) = 0, and since UT is one-to-one this implies that v = 0. Therefore we must have N(T ) = 0, and hence T is one-to-one. Note that U need not be one-to-one. Consider the following example: Dene T : R R2 and U : R2 R by T (x) = (x, 0) and U(x, y) = x. The composition UT is the identity function on R and is in particular one-to-one. However, the function U is not one-to-one since U(0, 1) = 0. (b) Suppose that UT is onto, we will now show that U is onto. Take any z Z. Since UT is onto we know there is a v V such that (UT )(v) = z. This shows that z = U(T (v)) is in the range of U. Since z Z was arbitary we nd that R(U) = Z; therefore U is onto. Note that T need not be onto. Consider the example from part (a): Dene T : R R2 and U : R2 R by T (x) = (x, 0) and U(x, y) = x. The composition UT is the identity function on R and is in particular onto. However, the function T is not onto since (0, 1) R(T ). (c) Suppose that U and T are one-to-one. We show that UT is one-to-one also. Take any v N(UT ). Since U(T (v)) = (UT )(v) = 0, we nd that T (v) = 0 [here we used that U is one-to-one]. But this then implies that v = 0 since T is also one-to-one. Therefore (c) For T L(V, W ),
N(UT ) = 0, and hence UT is one-to-one. Now suppose that U and T are onto. We show that UT is onto also. Take any z Z. Since U is onto, there is a w W such that U(w) = z. Since T is onto, there is a v V such that T (v) = w. Combining everything we nd that (UT )(v) = U(T (v)) = U(w) = z. Since z Z was arbitrary we nd that R(UT ) = Z. Therefore UT is onto as desired. Problem 13. Let A and B be n n matrices.
n
tr(AB) =
i=1 n
(AB)i,i
n
=
i=1 k=1 n n
= =
k=1
(BA)k,k
= tr(BA) Recall that At is the n n matrix such that (At )i,j = Aj,i. tr(A ) =
i=1 t n
(A )i,i =
i=1
Ai,i = tr(A)