James V. answered 6d
Making Linear Algebra Concrete | Harvard & Yale Grad, 35+ Yrs Exp.
I'll prove that if A and B are commuting positive definite matrices, then AB is also positive definite.
Proof:
Since A and B are positive definite matrices that commute (AB = BA), they can be simultaneously diagonalized by a unitary matrix. This is a key fact from linear algebra: commuting Hermitian matrices share the same eigenvectors.
Step 1: Simultaneous diagonalization
Since A and B are both positive definite, they are Hermitian. Because they commute, there exists a unitary matrix U such that:
- A = UDₐU*
- B = UDᵦU*
where Dₐ and Dᵦ are diagonal matrices with positive real entries (the eigenvalues of A and B respectively, which are positive since A and B are positive definite).
Step 2: Express AB
AB=(UDAU∗)(UDBU∗)=UDADBU∗AB=(UDAU∗)(UDBU∗)=UDADBU∗
Since diagonal matrices commute, we have DₐDᵦ = DᵦDₐ.
Step 3: Show AB is Hermitian
(AB)∗=(UDADBU∗)∗=UDBDAU∗=UDADBU∗=AB(AB)∗=(UDADBU∗)∗=UDBDAU∗=UDADBU∗=AB
So AB is Hermitian.
Step 4: Show AB is positive definite
For any nonzero vector x: x∗(AB)x=x∗UDADBU∗xx∗(AB)x=x∗UDADBU∗x
Let y = U*x (note that y ≠ 0 since U is unitary and x ≠ 0): x∗(AB)x=y∗DADByx∗(AB)x=y∗DADBy
Since Dₐ and Dᵦ are diagonal matrices with positive entries λ₁, λ₂, ..., λₙ and μ₁, μ₂, ..., μₙ respectively:
y∗DADBy=∑i=1nλiμi∣yi∣2y∗DADBy=∑i=1nλiμi∣yi∣2
Since each λᵢ > 0 and each μᵢ > 0, and y ≠ 0 (so at least one |yᵢ|² > 0), we have: x∗(AB)x=∑i=1nλiμi∣yi∣2>0x∗(AB)x=∑i=1nλiμi∣yi∣2>0
Therefore, AB is positive definite. ∎