Math 317: Linear Algebra Homework 7 Solutions The following problems are for additional practice and are not to be turned in: (All problems come from Linear Algebra: A Geometric Approach, 2nd Edition by ShifrinAdams.) Exercises: Section 3.3: 1–6, 8, 18, 19 Section 3.4: 1,3,19 Turn in the following problems. 1. Section 3.3, Problem 9 Proof: Suppose that u, v, w ∈ Rn form a linearly independent set. We prove that u + v, v + 2w and −u + v + w form a linearly independent set. To this extent, let c1 (u + v) + c2 (v + 2w) + c3 (−u + v + w) = 0. We show that c1 = c2 = c3 = 0. By rearranging (collecting like terms) the linear combination given above, we have that c1 (u+v)+c2 (v+2w)+c3 (−u+v+w) = 0 =⇒ (c1 − c3 )u + (c1 + c2 + c3 )v + (2c2 + c3 )w = 0. Since {u, v, w} form a linear independent set, we know that c1 − c3 = 0, c1 + c2 + c3 = 0 and 2c2 + c3 = 0. From these equations we find that c1 = c3 , c2 = −c3 /2 and c1 + c2 + c3 = c3 + −c3 /2 + c3 = 0 =⇒ c3 = 0 =⇒ c1 = c2 = 0. Thus u + v, v + 2w and −u + v + w form a linearly independent set. 2. Section 3.3, Problem 10 Proof: Suppose that v1 , v2 , . . . , vk are nonzero vectors with the property that vi · vj = 0 whenever i 6= j. We prove that {v1 , . . . , vk } forms a linearly independent set. Suppose that c1 v1 + c2 v2 + . . . + ck vk = 0. We show that c1 = . . . = ck = 0. We take the dot product of this equation with vi to obtain c1 (v1 · vi ) + c2 (v2 · vi ) + . . . + ci (vi · vi ) + . . . ck (vk · vi ) = 0 · vi = 0. Using the fact that vi · vj = 0 whenever i 6= j, the above equation gives ci (vi · vi ) = 0 =⇒ c1 = 0 since vi 6= 0 and v1 · v1 = kvi k2 . Since this true for all i = 1, 2, . . . , k, we have that c1 = c2 = . . . = ck = 0 and hence {v1 , . . . , vk } forms a linearly independent set. 3. Section 3.3, Problem 11 (a) Proof : Suppose that v1 , v2 , . . . , vn are nonzero mutually orthogonal vectors in Rn . In the previous problem, we proved that V = {v1 , . . . , vn } forms a linearly independent set. Thus dim V = n, and V ⊂ Rn . Since dim Rn = n, it must be the case that V = Rn . Thus v1 , v2 , . . . , vn forms a basis for Rn . (b) Given any x ∈ Rn , we give an explicit formula for the coordinates of x 1 Math 317: Linear Algebra Homework 7 Solutions with respect to the basis {v1 , . . . , vn }. Since v1 , v2 , . . . , vn forms a basis for Rn , we can write x = c1 v 1 + c2 v 2 + . . . + cn v n , for some c1 , c2 , . . . , cn ∈ R. To find these c0i s, we take the dot product of vi on both sides to the equation above to obtain: x · vi = c1 (v1 · vi ) + c2 (v2 · vi ) + . . . + ci (vi · vi ) + . . . + cn (vn · vi ). Once again, we use the fact that vi · vj = 0 whenever i 6= j to obtain x · vi = ci (vi · vi ) =⇒ ci = (c) Recalling that projvi x = x·vi v, kvi k2 i x · vi x · vi = v i · vi kvi k2 from part (b), we see that x = c1 v 1 + c2 v 2 + . . . + cn v n x · v2 x · vn x · v1 v1 + v2 + . . . + vn = 2 2 kv1 k kv2 k kvn k2 n X = projvi x i=1 4. Section 3.3, Problem 15 Proof: Suppose that k > n. We prove that any k vectors in Rn must form a linearly dependent set. We recall that the columns of A form a linearly dependent set if N (A) 6= {0}. This is equivalent to showing that Ax = 0 has a nontrivial solution which is the same as showing that A is not a full rank matrix. Let A be the n × k matrix that contains any k vectors in Rn . Since rank(A) ≤ n < k (because the maximum number of nonzero rows in any echelon form of A is n), we have that A is not a full rank matrix and hence there are free variables that will ensure that Ax = 0 has infinitely many solutions which in turns imply that N (A) 6= {0}. 5. Section 3.3, Problem 21 Proof : To show that {Av1 , Av2 , . . . , Avk } is linearly independent, we start with c1 (Av1 ) + c2 (Av2 ) + . . . + ck (Avk ) = 0, and show that c1 = c2 = . . . = ck = 0. By linearity of A we have that c1 (Av1 ) + c2 (Av2 ) + . . . + ck (Avk ) = A(c1 v1 + . . . + ck vk ) = 0. Since rank(A) = n for an m × n matrix, we know that Ax = 0 has only the trivial solution, i.e. x = 0. Thus, A(c1 v1 + . . . + ck vk ) = 0 =⇒ c1 v1 + . . . + ck vk = 0. Since {v1 , v2 , . . . , vk } is linearly 2 Math 317: Linear Algebra Homework 7 Solutions independent, then c1 = c2 = . . . = ck = 0. Thus, {Av1 , Av2 , . . . , Avk } is linearly independent. 6. Section 3.4, Problem 3d 1 0 1 1 . Then a row echelon form of A augmented by 2 0 0 −1 b2 − b 1 − b4 1 0 2 0 2 0 1 1 0 1 b2 − b1 . b = (b1 , b2 , b3 , b4 ) is given by [U |b̂] = 0 0 0 1 −1 b1 + b4 0 0 0 0 0 −2b2 + b3 − 2b4 A basis for C(A) is given by the pivot columns of A. Consulting the row echelon form of A, we find the columns 1, 2, and 4 all correspond to pivot columns. Thus, the 1st, 2nd,and 4thcolumns of A will form a basis for C(A). 1 −1 1 1 0 1 That is, C(A) = span , , and dim(C(A)) = 3. 0 2 2 −1 1 0 1 −1 1 1 0 2 Let A = 0 2 2 −1 1 −1 A basis rows in any echelon form of A. Thus, R(A) = forR(A) are thenonzero 1 0 0 1 0 0 span 2 , 1 , 0 and dim(R(A)) = 3. 0 0 1 2 1 −1 To find a basis for N (A), we solve Ax = 0. From the row echelon form of A, we see that x3 and x5 are free variables. We also obtain the following reduced system of equations to solve: x1 + 2x3 + 2x5 = 0 =⇒ x1 = −2x3 − 2x5 x2 + x3 + x5 = 0 =⇒ x2 = −x3 − x5 x4 − x5 = 0 =⇒ x4 = x5 . x1 −2x3 − 2x5 x2 −x3 − x5 = = x x In standard form we may write the solution as 3 3 x4 x5 x5 x5 3 Math 317: Linear Algebra Homework 7 Solutions −2 −2 −2 −2 −1 −1 −1 −1 x3 1 +x5 0 . Thus a basis for N (A) is given by N (A) = span 1 , 0 0 1 0 1 0 1 0 1 and dim(N (A)) = 2. A basis for N (AT ) is made up of the coefficients in the constraint equations for b so that Ax = b. Since we only have one zero row in our echelon form, we have one equation and thus a basis for N (AT ) is given by N (AT ) = constraint 0 −2 T span 1 , and dim(N (A )) = 1. −2 To check the orthogonality, take the dot product of any vector in N (A) with any vector in R(A) to see that it is 0. Similarly, take any vector in C(A) with any vector in R(AT ) to see that it is 0. 7. Section 3.4, Problem 22 (a) Proof : In this problem, we refer back to the properties proved in exercise 3.2.10. To prove that rank(AB) ≤ rank(A), we recall from exercise 3.2.10, that C(AB) ⊂ C(A). Thus, dim C(AB) ≤ dimC(A). However, rank(A) = dim C(A) and thus rank(AB) ≤ rank(A). (b) Proof: If n = p and B is nonsingular, then from exercise 3.2.10, we have that C(AB) = C(A) =⇒ dim C(AB) = dim C(A) =⇒ rank(AB) = rank(A). (c) Proof : From exercise 3.2.10, we know that N (B) ⊂ N (AB) and hence dim N (B) ≤ dim N (AB). Now dim N (B) = p−rank(B) and dim N (AB) = p − rank(AB) and thus p − rank(B) ≤ p − rank(AB) =⇒ rank(AB) ≤ rank(B). (d) Proof: If m = n and A is nonsingular, then from exercise 3.2.10, we have that N (B) = N (AB) which implies dim N (B) = dim N (AB) =⇒ p − rank(B) = p − rank(AB) =⇒ rank(B) = rank(AB). (e) Proof : Suppose that rank(AB) = n. We show that rank(A) = rank(B) = n. From the previous problems we have that n = rank(AB) ≤ rank(B) ≤ n =⇒ rank(B) = n. Similarly, we have that n = rank(AB) ≤ rank(A) ≤ n =⇒ rank(A) = n. 4