Math 344: Linear Algebra Homework 24, Solutions 8.A.1 Define T ∈ L(C2 ) by T (w, z) = (z, 0). Find all generalized eigenvectors of T . Solution: The matrix of T in the standard basis is given by 0 1 M (T ) = , 0 0 therefore the only eigenvalue of T is 0. Also note that M (T )2 = 0, so C2 = G(0, T ). This means that every vector of C2 is a generalized eigenvector of T . 8.A.3 Suppose that T ∈ L(V ) is invertible. Prove that G(λ, T ) = G( λ1 , T −1 ) for every λ ∈ F with λ 6= 0. Proof. Fix λ ∈ F with λ 6= 0. Suppose that v ∈ G(λ, T ). Then there is a positive integer n such that (T − λI)n v = 0. Multiplying this equation on the left by (−λ−1 T −1 )n we have 0 = (−λ−1 T −1 )n (T − λI)n v = (−λ−1 T −1 (T − λI))n v = (T −1 − λ−1 I)n v. This shows that v ∈ G(λ−1 , T −1 ). We conclude that G(λ, T ) ⊆ G(λ−1 , T −1 ). The same argument with the roles of λ and T being played by λ−1 and T −1 shows that G(λ−1 , T −1 ) ⊆ G(λ, T ). 8.A.5 Suppose that T ∈ L(V ), m is a positive integer, and v ∈ V is such that T m−1 v 6= 0 but T m v = 0. Prove that v, T v, T 2 v, . . . , T m−1 v is linearly independent. Proof. We start by noting that the hypotheses imply that v 6= 0. Suppose that a0 , . . . , am−1 are scalars in F such that a0 v + · · · + am−1 T m−1 v = 0. 1 2 Apply the operator T m−1 to this equation to see that a0 T m−1 v = 0; since T m−1 v 6= 0, we have a0 = 0. Therefore a1 T v + · · · + am−1 T m−1 v = 0. Next apply the operator T m−2 to this last equation to see that a1 T m−1 v = 0; then a1 = 0. Continuing in this manner we will conclude that all of the scalars a0 , . . . , am−1 are equal to 0. Hence v, T v, T 2 v, . . . , T m−1 v is linearly independent. 8.A.7 Suppose that N ∈ L(V ) is nilpotent. Prove that 0 is the only eigenvalue of N . Proof. Let λ ∈ F be an eigenvalue of N . Suppose that N n = 0 for some nonnegative integer n. Let v ∈ V be an eigenvector of N corresponding to λ. Then N v = λv, so 0 = N n v = λn v; and so λn = 0. We conclude that λ = 0. 8.A.8 Prove or give a counterexample: the set of nilpotent operators on V is a subspace of L(V ). This statement is false. Counterexample: Consider the operators S, T ∈ L(C2 ) defined by S(w, z) = (0, w); and T (w, z) = (z, 0). 2 2 Then S = T = 0, so S and T are nilpotent. Also, (S + T )(w, z) = (z, w), 2 so (S + T ) = I, which implies that S + T is not nilpotent. We have shown that the set of nilpotent operators on a vector space V need not be closed under addition. 8.B.1 Suppose that V is a finite-dimensional complex vector space, N ∈ L(V ) and that 0 is the only eigenvalue of N . Prove that N is nilpotent. Proof. By 8.21 (a) and 8.11 we have V = G(0, N ) = null (N dim V ). Therefore N dim V is the zero operator on V . We conclude that N is nilpotent. 3 8.B.2 Give an example of an operator T on a finite-dimensional real vector space such that 0 is the only eigenvalue of T but T is not nilpotent. Example: Let T ∈ L(R3 ) be defined by T (x, y, z) = (−y, x, 0). Note that T (0, 0, 1) = (0, 0, 0) so 0 is an eigenvalue of T . Now suppose that λ 6= 0 and T (x, y, z) = λ(x, y, z). It follows that z = 0, and λx = −y, λy = x. Therefore (λ2 +1)x = 0, so x = 0, which then implies that y = 0. We conclude that 0 is the only eigenvalue of T . Finally it is straightforward to compute T 4 (x, y, z) = (x, y, 0) for all (x, y, z) ∈ R3 , so we can see that T is not nilpotent. 8.B.3 Suppose that T ∈ L(V ). Suppose that S ∈ L(V ) is invertible. Prove that T and S −1 T S have the same eigenvalues with the same multiplicity. Proof. Let k be a nonnegative integer. Then (S −1 T S)k = (S −1 T S)(S −1 T S) · · · (S −1 T S) = S −1 T k S; also for any operators T1 and T2 , S −1 (T1 + T2 )S = S −1 T1 S + S −1 T2 S. It follows for any polynomial p ∈ P(F), that p(S −1 T S) = S −1 p(T )S. Let n = dim V , and fix some λ ∈ F. Applying the last statement to the polynomial p(z) = (z − λ)n , we have (S −1 T S − λI)n = S −1 (T − λI)n S. We now claim that S −1 affords a bijection G(λ, T ) → G(λ, S −1 T S). Note that the truth of this claim implies that T and S −1 T S have the same eigenvalues with the same multiplicity. Let v ∈ G(λ, T ). Then [S −1 (T − λI)n S](S −1 v) = S −1 (T − λI)n v = 0, which proves that S −1 maps G(λ, T ) to G(λ, S −1 T S). Similarly, for w ∈ G(λ, S −1 T S), (S −1 T S − λI)n w = 0 =⇒ S −1 (T − λI)n Sw = 0 =⇒ (T − λI)n Sw = 0, 4 which shows that Sw ∈ G(λ, T ). Then since w = S −1 (Sw) we conclude that the map S −1 : G(λ, T ) → G(λ, S −1 T S) is bijective. 8.B.5 Suppose that V is a finite-dimensional complex vector space and T ∈ L(V ). Prove that V has a basis consisting of eigenvectors of T if and only if every generalized eigenvector of T is an eigenvector of T . Proof. Suppose that V has a basis consisting of eigenvectors of T . Let λ1 , . . . , λm be the distinct eigenvalues of T . Then V = E(λ1 , T ) ⊕ · · · ⊕ E(λm , T ). Moreover, in any case, V = G(λ1 , T ) ⊕ · · · ⊕ G(λm , T ). Since E(λi , T ) ⊆ G(λi , T ) for i ∈ {1, . . . , m}, we conclude that G(λi , T ) = E(λi , T ); this means that every generalized eigenvector of T is an eigenvector of T . Now assume that every generalized eigenvector of T is an eigenvector of T . Then G(λi , T ) = E(λi , T ), for all i ∈ {1, . . . , m}. This then implies, using 8.23, that V = E(λ1 , T ) ⊕ · · · ⊕ E(λm , T ). Hence V has a basis consisting of eigenvectors of T . 8.B.10 Suppose that V is a finite-dimensional complex vector space and that T ∈ L(V ). Prove that there exist D, N ∈ L(V ) such that T = D + N , the operator D is diagonalizable, N is nilpotent, and DN = N D. Proof. We begin by writing V = G(λ1 , T ) ⊕ · · · ⊕ G(λm , T ), where λ1 , . . . , λm are the distinct eigenvalues of T . We define an operator D ∈ L(V ) by setting D|G(λi ,T ) = λi IG(λi ,T ) , 5 where IG(λi ,T ) denotes the identity operator on G(λi , T ). Clearly the matrix of D with respect to any basis coming from the given decomposition of V is a diagonal matrix, so D is diagonalizable. Now set N = T − D. Note that N |G(λi ,T ) = (T − λi I)|G(λi ,T ) , and since the last operator is nilpotent, we conclude that N is nilpotent. Since we have T = D +N , the proof will be complete once we prove that DN = N D. To see this note that DN = D(T − D) = DT − D2 , and N D = (T − D)D = T D − D2 . Then since D is a scalar operator on G(λi , T ), we see that DT = T D when restricted to G(λi , T ), but then the fact that V decomposes as a direct sum of the G(λi , T ) implies that DT = T D. We conclude that DN = N D.