Solutions Homework 2 6.7 Prove that if V is a complex inner-product space, then hu, vi = ku + vk2 − ku − vk2 + ku + ivk2 i − ku − ivk2 i 4 for all u, v ∈ V . Proof. Expanding each term using linearity in the first term and conjugate linearity in the second yields ku + vk2 = hu + v, u + vi, = kuk2 + hu, vi + hv, ui + kvk2 , = kuk2 + hu, vi + hu, vi + kvk2 , ku − vk2 = hu − v, u − vi, = kuk2 − hu, vi − hu, vi + kvk2 , ku + ivk2 = hu + iv, u + ivi, = kuk2 + hu, ivi + hiv, ui + kivk2 , = kuk2 − ihu, vi + ihv, ui + |i|2 kvk2 , = kuk2 − ihu, vi + ihu, vi + kvk2 , ku − ivk2 = hu − iv, u − ivi, = kuk2 − hu, ivi − hiv, ui + kivk2 , = kuk2 + ihu, vi − ihv, ui + |i|2 kvk2 , = kuk2 + ihu, vi − ihu, vi + kvk2 . Now ku + vk2 − ku − vk2 = 2(hu, vi + hu, vi), and i(ku + ivk2 − ku − ivk2 ) = 2i(−ihu, vi + ihu, vi) = 2(hu, vi − hu, vi). Together these show that ku + vk2 − ku − vk2 + ku + ivk2 i − ku − ivk2 i = 4hu, vi. 6.8 A norm on a vector space U is a function k k : U → [0, ∞) such that kuk = 0 if and only if u = 0, kαuk = |α|kuk for all α ∈ F and all u ∈ U , and ku + vk ≤ kuk + kvk for all u, v ∈ U . Prove that a norm satisfying the parallelogram equality comes from an inner product (in other words, show that if k k is a norm on U satisfying the parallelogram equality, then there is an inner product h , i on U such that kuk = hu, ui1/2 for all u ∈ U ). Proof. Let hu, vi = ku + vk2 − ku − vk2 + ku + ivk2 i − ku − ivk2 i . 4 1 2 This is positive definite as k2uk2 + k(1 + i)uk2 i − k(1 − i)uk2 i , 4 4kuk2 + |1 + i|2 kuk2 i − |1 − i|2 kuk2 i = , 4 = kuk2 . hu, ui = This is positive and only zero when u = 0 by the properties of the norm k k. The inner product is conjugate symmetric as kv + uk2 − kv − uk2 + kv + iuk2 i − kv − iuk2 i , 4 ku + vk2 − | − 1|ku − vk2 + |i|k − iv + uk2 i − | − i|kiv + uk2 i = , 4 ku + vk2 − ku − vk2 + ku − ivk2 i − ku + ivk2 i = , 4 = hu, vi. hv, ui = 6.9 Suppose n is a positive integer. Prove that 1 sin(x) sin(2x) sin(nx) cos(x) cos(2x) cos(nx) √ , √ , √ ,..., √ , √ , √ ,..., √ π π π π π π 2π is an orthonormal list of vectors in C[π, π], the vector space of continuous real-valued functions on [π, π] with inner product Z π hf, gi = f (x)g(x)dx. −π Proof. The first vector is orthogonal to the others as for all k = 1, . . . , n, π Z π 1 sin(kx) 1 −1 √ , √ =√ sin(kx)dx = √ cos(kx) = 0, π 2π 2π −π k 2π −π similarly for all k = 1, . . . , n, π Z π 1 cos(kx) 1 1 √ , √ =√ cos(kx)dx = √ sin(kx) = 0. π 2π 2π −π k 2π −π Now for all k = 1, . . . , n and l = 1, . . . , n Z sin(kx) cos(lx) 1 π √ , √ = sin(kx) cos(lx)dx, π −π π π Z 0 Z π 1 = sin(kx) cos(lx)dx + sin(kx) cos(lx)dx , π −π 0 3 and substituting u = −x in the first integral shows Z 0 Z π sin(kx) cos(lx) 1 √ − sin(−ku) cos(−lu)du + sin(kx) cos(lx)dx , = , √ π π π π 0 Z π Z π 1 = − sin(ku) cos(lu)du + sin(kx) cos(lx)dx , π 0 0 = 0. Finally for all k, l = 1, . . . , n with k 6= l, integrating by parts shows Z sin(kx) sin(lx) 1 π √ = , √ sin(kx) sin(lx)dx, π −π π π π Z 1 sin(kx) cos(lx) k π = − cos(kx) cos(lx)dx , + l π l −π −π k cos(kx) cos(lx) √ = , √ , l π π and another integration by parts shows Z cos(kx) cos(lx) 1 π √ , √ cos(kx) cos(lx)dx, = π −π π π π Z 1 cos(kx) sin(lx) k π = sin(kx) sin(lx)dx , + l π l −π −π k sin(kx) sin(lx) √ = , √ . l π π Therefore, sin(kx) sin(lx) k 2 sin(kx) sin(lx) √ √ , √ , √ = 2 . l π π π π E D E D sin(lx) cos(kx) cos(lx) √ √ √ √ , = 0, which implies that , = 0. Since k = 6 l this implies sin(kx) π π π π They are orthonormal as Z π 1 2 1 √ = dx = 1, 2π −π 2π and by the half angle formulas, for all k = 1, . . . , n Z π Z sin(kx) 2 1 π 1 − cos(2kx) 2 √ = 1 sin (kx)dx = dx = 1, π −π π −π 2 π and Z π Z π cos(kx) 2 1 1 1 + cos(2kx) 2 √ = cos (kx)dx = dx = 1. π −π π −π 2 π 4 6.10 On P2 (R), consider the inner product given by Z 1 hp, qi = p(x)q(x)dx. 0 Apply the Gram-Schmidt procedure to the basis (1, x, x2 ) to produce an orthonormal basis of P2 (R). Solution: The first polynomial, 1, is a unit vector as Z 1 2 1dx = 1, k1k = 0 thus e1 (x) = 1. The second polynomial will be R 1 x − 0 xdx 1 √ x − 21 1 e2 (x) = =2 3 x− . = qR x − 1 1 2 1 2 2 (x − 2 ) dx 0 The third polynomial is given by R R 1 1 x2 − 0 x2 dx) 1 − 12 0 x2 x − 12 dx) x − 12 e3 (x) = , x2 − 1 − x − 1 3 2 = qR 1 x2 − x + x2 1 6 1 2 6 , dx −x+ √ 1 = 6 5 x2 − x + . 6 6.11 What happens if the Gram-Schmidt procedure is applied to a list of vectors that is not linearly independent? Solution: The Gram-Schmidt procedure can be describe inductively using orthogonal projections as follows. For a list of vectors (v1 , . . . , vk ) v1 e1 = U1 = span(e1 ), kv1 k vj+1 − PUj vj+1 ej+1 = Uj+1 = span(e1 , . . . , ej+1 ). kvj+1 − PUj vj+1 k 0 When the list is linearly dependent, let l be the smallest integer such that (v1 , . . . , vl ) is linearly independent and vl+1 ∈ span(v1 , . . . , vl ). Then the procedure produces an orthonormal basis (e1 , . . . , el ) for Ul =span(e1 , . . . , el ) =span(v1 , . . . , vl ). Since vl+1 ∈ Ul , at the next step of the procedure vl+1 − PUl vl+1 0 el+1 = el+1 = = . kvl+1 − PUl vl+1 k k0k This is not defined, however the procedure can be continued if we simply throw out vl+1 . Proceeding in this manner will produce an orthonormal basis for span(v1 , . . . , vk ), but there will be fewer than k vectors in this basis. 6.13 Suppose (e1 , . . . , em ) is an orthonormal list of vectors in V . Let v ∈ V . Prove that kvk2 = |hv, e1 i|2 + · · · + |hv, em i|2 if and only if v ∈ span(e1 , . . . , em ). 5 Proof. First, extend (e1 , . . . , em ) to an orthonormal basis, (e1 , . . . , en ). Then since v ∈V, n X v= ai ei . i=1 Note that, hv, ej i = * n X + ai ei , ej = i=1 n X ai hei , ej i = aj , i=1 and kvk2 = * n X ai e i , i=1 = = n X n X n X + aj e j , j=1 ai āj hei , ej i, i=1 j=1 n X |ai |2 , i=1 = n X |hv, ei i|2 . i=1 2 Now assume that kvk = |hv, e1 i|2 + · · · + |hv, em i|2 , then n X |hv, ei i|2 = 0, i=m+1 and since all of the terms in the sum are non-negative, it must be that each is zero. That is, for i = m + 1, . . . , n, |hv, ei i|2 = 0, which implies ai = hv, ei i = 0. Thus v = a1 e1 + · · · + am em and v ∈span(e1 , . . . , em ). For the converse, assume that v ∈span(e1 , . . . , em ). This implies that ai = 0 for i = m + 1, . . . , n, thus m X kvk2 = |hv, ei i|2 . i=1 6.16 Suppose U is a subspace of V . Prove that U ⊥ = {0} if and only if U = V . Proof. Assume that U ⊥ = {0}, and let v ∈ V . Since h0, vi = 0, and 0 is the only vector in U ⊥ , v ∈ (U ⊥ )⊥ . Since (U ⊥ )⊥ = U , this implies that v ∈ U , and V ⊂ U . Therefore, as U is assumed to be a subspace of V , U = V . Now assume, conversely, that U = V , and let u ∈ U ⊥ . Let (e1 , . . . , en ) be an orthonormal basis for V then u = hu, e1 ie1 + · · · + hu, en ien . Since U = V , ei ∈ U for all i = 1, . . . , n, this implies that hu, ei i = 0. Therefore u = 0, and U ⊥ = {0}. 6 2.1.1 For V = R3 , with inner product given by the dot product, hx, yi = x · y, v1 = (1, 0, 1), v2 = (0, 1, 1), v3 = (1, 3, 3), apply the Gram-Schmidt procedure to obtain an orthonormal basis {e1 , e2 , e3 }. Then write v = (1, 1, 2) in terms of {e1 , e2 , e3 }. Solution: First set (1, 0, 1) e1 = √ , 2 then (0, 1, 1) − (v2 · e1 )e1 , e2 = k(0, 1, 1) − (v2 · e1 )e1 k (0, 1, 1) − = (1,0,1) √1 √ 2 2 k(0, 1, 1) − (1/2, 0, 1/2)k (−1, 2, 1) √ = , 6 , and finally e3 = (1, 3, 3) − (v3 · e1 )e1 − (v3 · e2 )e2 , k(1, 3, 3) − (v3 · e1 )e1 − (v3 · e2 )e2 k (1, 3, 3) − = (1,0,1) √4 √ 2 2 − (−1,2,1) √8 √ 6 6 k(1, 3, 3) − (2, 0, 2) − (−4/3, 8/3, 4/3)k (1, 1, −1) √ = . 3 Now v · e1 = compute √3 , 2 v · e2 = √3 , 6 , and v · e3 = 0. Therefore v = √3 e1 2 + q 3 e, 2 2 to check 3 3 √ e1 + √ e2 = (3/2, 0, 3/2) + (−1/2, 1, 1/2), 2 6 = (1, 1, 2) = v. 2.1.2 For V = R2 , hx, yi = 2x1 y1 − x1 y2 − x2 y1 + 2x2 y2 , apply the Gram-Schmidt procedure to v1 = (1, 0), v2 = (0, 1) to obtain an orthonormal basis {e1 , e2 }. Solution: First, since kv1 k2 = 2, e1 = v1 (1, 0) = √ . kv1 k 2 Next, v2 − hv2 , e1 ie1 e2 = (0, 1) − 1 h(1, 0), (0, 1)i(1, 0) , 2 (0, 1) − (−1)(1, 0) = p , h(1, 1), (1, 1)i (1, 1) = √ . 2 2.2.1 Prove that U ⊥ is always a subspace of V . 7 Proof. Let v ∈ U ⊥ , then for all u ∈ U , hv, ui = 0. This implies that hcv, ui = chv, ui = 0 for all u ∈ U , and thus cv ∈ U ⊥ . Therefore U ⊥ is closed under scalar multiplication. Let v1 , v2 ∈ U ⊥ , then for all u ∈ U , hv1 , ui = hv2 , ui = 0. This implies that hv1 + v2 , ui = hv1 , ui + hv2 , ui = 0 for all u ∈ U , and thus v1 + v2 ∈ U ⊥ . Therefore U ⊥ is closed under vector addition. This implies that U ⊥ is a subspace of V . 2.2.2 For V = R3 , hx, yi = x1 y1 + x2 y2 − x1 y3 − x3 y1 + 4x3 y3 , extend the orthonormal list e1 = (1, 0, 0), e2 = (0, 1, 0) to an othonormal basis. Solution: The vector v3 = (0, 0, 1) completes the list to a basis for V . v3 − hv3 , e1 ie1 − hv3 , e2 ie2 e3 = , k(0, 0, 1) − (−1)(1, 0, 0) − (0)(0, 1, 0)k (1, 0, 1) = √ . 3