Positive Definiteness Determined from Symmetrized Products

  • POTW
  • Thread starter Euge
  • Start date
In summary, "positive definiteness determined from symmetrized products" is a mathematical concept used to determine if a symmetric matrix is positive definite. It takes into account the symmetry of the matrix and considers all possible combinations of elements to determine the positivity of the resulting values. This method differs from others by providing more accurate results, being applicable to matrices of any size, and having various real-world applications. However, it may have limitations in terms of computational cost and applicability to matrices with complex or imaginary elements. Overall, it is a useful tool in scientific research for analyzing positive definiteness and its applications in optimization, physics, economics, and data analysis.
  • #1
Euge
Gold Member
MHB
POTW Director
2,073
243
Suppose ##P## and ##Q## are self-adjoint linear operators on a finite dimensional, complex inner product space. Assume both ##P## and the symmetrized product ##PQ + QP## are positive definite. Show that ##Q## must also be positive definite.
 
  • Like
Likes Greg Bernhardt and topsquark
Physics news on Phys.org
  • #2
Note that as [itex]P[/itex] and [itex]Q[/itex] are self-adoint on a finite-dimensional complex inner product space [itex](V, \langle \cdot, \cdot \rangle)[/itex], their eigenvalues are real and they each have a basis of orthogonal eigenvectors.

By definition, a linear map [itex]L: V \to V[/itex] is positive definite if and only if [itex]\langle Lx ,x \rangle \geq 0[/itex] for every [itex]x \in V[/itex]. If in addition [itex]L[/itex] is self-adjoint, it is positive definite if all of its eigenvalues are non-negative, since we can take an orthonormal basis of eigenvectors [itex]x_i[/itex] with eigenvalues [itex]\lambda_i[/itex] so that for arbitrary [itex]v = \sum_i v_i x_i[/itex], [tex]
\langle Lv ,v \rangle = \sum_i \sum_j v_i v_j^{*} \lambda_i \langle x_i, x_j \rangle = \sum_i |v_i|^2 \lambda_i.[/tex]

Let [itex]x[/itex] be an eigenvector of [itex]Q[/itex] with eigenvalue [itex]\lambda[/itex]. Then by positive-definitenes of [itex]PQ + QP[/itex], [tex]
\begin{split}
0 &\leq \langle (PQ + QP)x, x \rangle \\
&= \langle PQ x, x \rangle + \langle QPx, x \rangle \\
&= \lambda \langle Px, x \rangle + \langle Px, Qx \rangle \qquad \mbox{(by self-adjointness of $Q$)} \\
&= \lambda \langle Px, x \rangle + \langle Px, \lambda x \rangle \\
&= \lambda \langle Px, x \rangle + \lambda \langle Px, x \rangle \qquad \mbox{(since $\lambda \in \mathbb{R}$)} \\
&= 2\lambda \langle Px, x \rangle. \end{split}[/tex] Now [itex]\langle Px ,x \rangle \geq 0[/itex] by positive definiteness of [itex]P[/itex], so we must have [itex]\lambda \geq 0[/itex] and [itex]Q[/itex] is positive-definite.
 
  • Like
Likes Greg Bernhardt and topsquark

Similar threads

Replies
1
Views
1K
Replies
2
Views
3K
Replies
31
Views
2K
Replies
3
Views
1K
Replies
0
Views
738
Back
Top