12
$\begingroup$

I am looking for the determinant

$$ \det(I_n + H_n) $$

where $I_n$ is the $n \times n$ identity matrix and $H_n$ is the $n \times n$ Hilbert matrix, whose entries are given by

$$ [H_n]_{ij} = \frac{1}{i+j-1}, \qquad\qquad 1 \le i,j \le n $$

Is anything known about this determinant for finite $n$ or about its asymptotic behaviour for $n \rightarrow \infty$?

More generally, are there results about the determinant of "identity plus Hankel" matrices or their asymptotic behaviour?

$\endgroup$
  • $\begingroup$ On the other hand, there is a nice formula for the determinant of a Cauchy matrix plus $J$, where $J$ is the $n \times n$-matrix whose all entries (not just the diagonal ones) are $1$. $\endgroup$ – darij grinberg Mar 8 '18 at 4:48
8
$\begingroup$

One thing you can say is that your determinant is the sum of determinants of the Cauchy matrices $C_S$ for subsets $S$ of $\{1,\ldots, n\}$, where $(C_S)_{i,j} = 1/(S_i + S_j - 1)$ for $1 \le i,j \le |S|$ (in the case $S = \emptyset$ we take the determinant to be $1$). This means $$ \det(I_n + H_n) = \sum_{S \subseteq \{1\ldots n\}} \frac{\prod_{1 \le i < j \le |S|} (S_i - S_j)^2}{\prod_{i,j=1}^{|S|} (S_i + S_j -1)}$$ For $n=1$ to $8$ your determinants are $$ 2,{\frac{29}{12}},{\frac{2927}{1080}},{\frac{659251}{224000}},{\frac{ 46508430817}{14817600000}},{\frac{616473989937916861}{ 186313420339200000}},{\frac{3577562384224548869428843}{ 1033954523962885324800000}},{\frac{1314142513507030576449489451528961} {365356847125734485878112256000000}} $$ I don't see an apparent pattern, nor does Maple's gfun package. The numerators and denominators don't seem to be in the OEIS.

EDIT: They are now. A295426 and A295427

$\endgroup$
  • $\begingroup$ The prime decompositions of denominators are kind of nice, although it is hard to see a pattern. The primes are very small compared to the denominators themselves. $\endgroup$ – Mark Sapir Feb 10 '18 at 1:26
  • 3
    $\begingroup$ Of course, all primes in the denominator for $\det(I_n + H_n)$ are at most $2n-1$. $\endgroup$ – Robert Israel Feb 11 '18 at 2:48
5
$\begingroup$

If you look at $H=\frac 1{(i+j-1)\pi}$ instead, then $\det(1 + H)\sim n^{3/8}$, as $n\to\infty$. This is basically proven in arXiv:1808.08009, arXiv:1905.03154

$\endgroup$

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.