Worksheet 8
Problem 3
Let A∈Mn×n(F) be similar to an upper-triangular matrix, and suppose that A has distinct eigenvalues λ1,…,λk with corresponding algebraic multiplicities m1,…,mk.
- Prove that trA=∑i=1kmiλi.
- Prove that detA=∏i=1kλimi.
Solution.
This follows from two facts:
- If A and B are similar, then they have the same eigenvalues.
- The eigenvalues of an upper-triangular matrix are its diagonal entries.
Fact 1 follows from the fact that if B=QAQ−1, then
det(B−λI)=det(Q(A−λI)Q−1)=det(A−λI),
i.e., A and B have the same characteristic equation. For Fact 2, if A is upper-triangular with diagonal entries a1,…,an, then because A−λI is also upper-triangular with diagonal entries a1−λ,…,an−λ, the characteristic equation of A is
pA(λ)=det(A−λI)=i=1∏n(ai−λ).
Thus, the roots of pA are exactly the diagonal entries, so the eigenvalues of A are also just the diagonal entries.
-
Note that tr(AB)=tr(BA) from a simple calculation:
tr(AB)=i=1∑n(AB)ii=i=1∑nj=1∑nAijBji=j=1∑ni=1∑nBjiAij=j=1∑n(BA)jj=tr(BA).
Thus, if A=QUQ−1, where U is an upper-triangular matrix, then
trA=tr(QUQ−1)=tr(Q−1QU)=trU.
But the diagonal entries of U are the eigenvalues of A counted with multiplicity, so
trU=i=1∑kmiλi.
-
Again, write A=QUQ−1, where U is upper-triangular. Then the determinant of U is just the product of the diagonal entries, which are the eigenvalues of A counted with multiplicity. Thus,
detA=detU=i=1∏kλimi.
Problem 4
Let V be a finite dimensional vector space over F, let T∈L(V) be invertible.
- Prove that if λ is an eigenvalue of T, then λ−1 is an eigenvalue of T−1.
- Prove that the eigenspace of T corresponding to λ is the same as the eigenspace of T−1 corresponding to λ−1.
- Prove that if T is diagonalizable, then T−1 is diagonalizable.
Solution.
These all follow from the fact that v is an eigenvector for T if and only if it's an eigenvector for T−1. If v is an eigenvector for T with eigenvalue λ, then
T(v)=λv⟹v=λT−1(v)⟹T−1(v)=λ−1v.
Note that because T is invertible, λ=0, so λ−1 is well-defined.
Since v is an eigenvector of T, we know that v=0, so this precisely says that v is an eigenvector for T with eigenvalue λ−1. Applying the forward direction with T,λ replaced with T−1,λ−1 and noting that (T−1)−1=T and (λ−1)−1=λ gives the reverse direction. (Alternatively, you can just do the same computation, but in reverse.)
-
Our computation above shows that if λ is an eigenvalue for T, then λ−1 is an eigenvalue for T−1.
-
Recall that the eigenspace of T corresponding to λ is the set
Eλ(T)={v∈V∣T(v)=λv}.
We showed that T(v)=λv if and only if T−1(v)=λ−1v, and this implies Eλ(T)=Eλ−1(T−1).
-
Let β={v1,…,vn} be an eigenbasis of V for T. Then β is also a set of eigenvectors for T−1. Since β was a basis of V, this means that β is also an eigenbasis of V for T−1.