I would like to know whether the set of $n\times n$ Hermitian matrices whose all ${{n}\choose{k}}$ principal $k\times k$ sub-matrices---the matrices obtained by removing $n-k$ columns as well as the corresponding rows---are positive semidefinite is a well-studied set, and if so under which name. This is for a given $1\leq k\leq n$. If $k=1$, this is the set of $n\times n$ matrices whose diagonal entries are non-negative, and, if $k=n$, it is the set of $n\times n$ positive-semidefinite matrices. I am interested in what is known for the general case $1\leq k\leq n$, in particular for the non-trivial case $1<k<n$, for $n\geq 3$.


RC Thompson had a series of problems exploring questions regarding principal minors of Hermitian matrices in full:

Principal submatrices of normal and Hermitian matrices

Principal submatrices II: The upper and lower quadratic inequalities

Principal submatrices III: Linear inequalities

Principal submatrices IV. On the independence of the eigenvalues of different principal submatrices

Principal submatrices V: Some results concerning principal submatrices of arbitrary matrices

Principal submatrices VI. Cases of equality in certain linear inequalities

Principal submatrices VII: Further results concerning matrices with equal principal minors

Principal submatrices. VIII. Principal sections of a pair of forms

Principal submatrices IX: Interlacing inequalities for singular values of submatrices

  • $\begingroup$ It is unclear to me if this answers the question (I am not saying it does not). Did Thompson consider matrices with positive principal k-minors? If so, in which of the 9 papers you listed? If so, what did he prove about them? $\endgroup$ – Moishe Kohan Dec 3 '16 at 16:06
  • $\begingroup$ For example in the second paper he discusses lower and upper quadratic bounds on the eigenvalues of the matrix based on the eigenvalues of the $(n-1) \times (n-1)$ submatrices. Positivity can be derived from these inequalities. $\endgroup$ – Nick R Dec 4 '16 at 9:51
  • $\begingroup$ Dear @NickR , many thanks for your contribution. There are some results in those papers that might turn out to be useful, but I am not sure your answer is what I was hoping for, exactly because of the generality of some of the results there, and of the specialization of some other ones. It seems more like a starting point to attack the specific problem I am interested in, rather than a direct analysis of it. $\endgroup$ – Marco Dec 8 '16 at 0:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.