To prove that tr(A∗)=−1 if and only if the matrix In+A∗ is invertible, we will consider different cases based on the rank of matrix A.
1. **Case 1: n=2**
Let A=[acbd]. The adjugate matrix A∗ is given by:
A∗=[d−c−ba]
The determinant of I+A∗ is:
det(I+A∗)=det[1+d−c−b1+a]
By direct computation, we have:
det(I+A∗)=(1+d)(1+a)−(−b)(−c)=1+a+d+ad−bc
Since A is not invertible, det(A)=ad−bc=0. Therefore:
det(I+A∗)=1+a+d
This determinant is nonzero if and only if a+d=−1. Since tr(A∗)=a+d, we conclude that tr(A∗)=−1 if and only if I+A∗ is invertible.
2. **Case 2: n≥3 and rank(A)≤n−2**
If rank(A)≤n−2, then any n−1 rows of A are linearly dependent. This implies that any (n−1)×(n−1) submatrix of A has dependent rows, and hence A∗=0. In this case:
tr(A∗)=0=−1
and:
I+A∗=I
which is clearly invertible. Thus, tr(A∗)=−1 if and only if I+A∗ is invertible.
3. **Case 3: n≥3 and rank(A)=n−1**
Since rank(A)=n−1, the null space of A is one-dimensional. Every column of A∗ lies within the null space of A, implying rank(A∗)=1. Let B=A∗. For rank 1 matrices, we have:
B2=(tr(B))B
This means the algebra generated by B consists of span{I,B}. If I+B is invertible, it must lie within this algebra. Suppose the inverse of I+B is of the form I+αB. Then:
(I+B)(I+αB)=I+(1+α)B+α(tr(B))B=I+(1+α+αtr(B))B
For this to equal I, we must have:
1+α+αtr(B)=0⟹α(1+tr(B))=−1
If tr(B)=−1, there is no solution for α, and I+B is not invertible. If tr(B)=−1, we can solve for:
α=1+tr(B)−1
and verify that I+αB is indeed the inverse of I+B. Thus, tr(A∗)=−1 if and only if I+A∗ is invertible.
■