Russian Mathematical Surveys
RUS  ENG    JOURNALS   PEOPLE   ORGANISATIONS   CONFERENCES   SEMINARS   VIDEO LIBRARY   PACKAGE AMSBIB  
General information
Latest issue
Archive
Impact factor
Submit a manuscript

Search papers
Search references

RSS
Latest issue
Current issues
Archive issues
What is RSS



Uspekhi Mat. Nauk:
Year:
Volume:
Issue:
Page:
Find






Personal entry:
Login:
Password:
Save password
Enter
Forgotten password?
Register


Russian Mathematical Surveys, 2025, Volume 80, Issue 4, Pages 729–731
DOI: https://doi.org/10.4213/rm10257e
(Mi rm10257)
 

Brief communications

Tight lower bounds for Shannon entropy from ‘quantum pyramids’

A. S. Holevo

Steklov Mathematical Institute of Russian Academy of Sciences, Moscow
References:
Funding agency Grant number
Russian Science Foundation 24-11-00145
This work was supported by Russian Scientific Foundation under grant no. 24-11-00145, https://rscf.ru/en/project/24-11-00145/.
Received: 20.06.2025
Presented: V. M. Buchstaber
Accepted: 20.06.2025
Published: 24.10.2025
Document Type: Article
MSC: Primary 81P45; Secondary 81R15
Language: English
Original paper language: Russian

Computing accessible information for an ensemble of quantum states is a basic problem in quantum information theory, going back to its origin and important, in particular, for applications in quantum cryptography. The problem is solved in some special cases but, in general, remains open. In this note we show that a recently obtained optimality criterion [1], when applied to specific ensembles of states, leads to non-trivial entropy inequalities that are discrete relatives of the log-Sobolev inequality [2]. In this light, the hypothesis of an information-optimal measurement for an ensemble of equiangular equiprobable states — the quantum pyramid [3] — is reconsidered and the corresponding tight entropy inequalities are proposed.

Let $\mathfrak{S}$ be the convex set of quantum states (density operators) on $d$-dimensional Hilbert space $\mathcal{H}$. An ensemble of quantum states is a collection $\mathcal{E}=\{\pi_{j},\rho_{j}\}_{j=1,\dots,m}$, where $\rho_{j}\in \mathfrak{S}$ and $\pi=\{\pi_{j}\}$ is a probability distribution. The average state is $\overline{\rho}=\sum_{j}\pi_{j}\rho_{j}$. An observable is a set of Hermitian operators $\mathcal{M}=\{M_{k}\}_{k=1,\dots,n}$, where $M_{k}\geqslant 0$ and $\sum_{k=1}^{n}M_{k}=I$. The joint probability distribution of an ‘input’ $j$ and an ‘output’ $k$ is $p_{jk}=\pi_{j}\operatorname{Tr}\rho_{j}M_{k}$. The accessible information of the ensemble $\mathcal{E}$ is equal to

$$ \begin{equation} A(\mathcal{E})=\sup_{\mathcal{M}}I(\mathcal{E},\mathcal{M}), \end{equation} \tag{1} $$
where $I(\mathcal{E},\mathcal{M})=\sum_{j,k}p_{jk}\log(p_{jk}/(\pi_{j}p_{\cdot k}))$ is the Shannon mutual information between $j$ and $k$. It has been shown [4] that the supremum is attained at an observable $ \mathcal{M}$ with linearly independent components $M_{k}=|\varphi_{k}\rangle \langle \varphi_{k}|$, $k=1,\dots,n\leqslant d^{2}$, of rank 1.

Theorem 1. An observable $\mathcal{M}= \{|\varphi_{k}\rangle \langle \varphi_{k}|\}$ is optimal for problem (1) if and only if there exists a Hermitian operator $\Lambda$ such that the entropy inequality

$$ \begin{equation} -\sum_{j}\langle\psi|M_{j}'|\psi\rangle\log\langle\psi|M_{j}'|\psi\rangle \geqslant\langle\psi|\Lambda|\psi\rangle,\qquad M_{j}'=\overline{\rho}^{-1/2}\rho_{j}\overline{\rho}^{-1/2}, \end{equation} \tag{2} $$
holds for all unit vectors $\psi \in \mathcal{H}$; and the unit vectors $|\phi_{k}\rangle=|\varphi_{k}\rangle/\|\varphi_{k}\|$ turn (2) into equality:
$$ \begin{equation*} -\sum_{j}\langle \phi_{k}|M_{j}'|\phi_{k}\rangle \log \langle \phi_{k}|M_{j}'|\phi_{k}\rangle= \langle\phi_{k}|\Lambda |\phi_{k}\rangle,\qquad k=1,\dots,n. \end{equation*} \notag $$

The value of the accessible information is $A(\mathcal{E})=H(\pi)-\operatorname{Tr}\overline{\rho}\Lambda$.

Consider an ensemble $\mathcal{E}$ of $m$ equiprobable equiangular pure states on $m$-dimensional Hilbert space — a quantum pyramid [3]. Denote by $\xi $ the inner product between the state vectors and introduce the parameter $a=m^{-1}[\sqrt{1+(m-1)\xi}+(m-1)\sqrt{1-\xi}\,]$. Then inequality (2) reduces to tight lower bounds for the Shannon entropy of a discrete probability distribution. Namely, in the case of a moderately acute pyramid we have the following result.

Theorem 2. For $m\geqslant 2$ and $p\equiv a^2>(m-1)/m$

$$ \begin{equation} -\sum_{j=1}^{m}t_{j}\log t_{j}\geqslant \mu_{0}(p) \biggl(\,\sum_{j=1}^{m}\sqrt{t_{j}}\,\biggr)^{2}-\mu_{1}(p); \end{equation} \tag{3} $$
here $t_{j}\geqslant 0$ and $\sum_{j=1}^{m}t_{j}=1$ (in fact, $t_{j}=|\langle\psi|e_j\rangle|^2$ for an orthonormal basis $\{|e_j\rangle\}$) and
$$ \begin{equation} \mu_{0}(p) =\frac{\sqrt{p(1-p)/(m-1)}\,[\log p-\log((1-p)/(m-1))]} {(\sqrt{p}+(m-1)\sqrt{(1-p)/(m-1)}\,)(\sqrt{p}-\sqrt{(1-p)/(m-1)}\,)}\,, \end{equation} \tag{4} $$
$$ \begin{equation} \mu_{1}(p) =\frac{\sqrt{p}\,\log p-\sqrt{(1-p)/(m-1)}\, \log((1-p)/(m-1))}{\sqrt{p}-\sqrt{(1-p)/(m-1)}}\,. \end{equation} \tag{5} $$
The graphs of the functions $t_{1},\dots,t_{m}$ in the left- and riht-hand sides of (3) are tangent at the points obtained from $t_{1}=p$, $t_{k}=(1-p)/(m-1)$, $k=2,\dots,m$, by permutations.

For $m\geqslant 3$ the value $p=(m-1)/m$ is allowed, giving the inequality

$$ \begin{equation} -\sum_{j=1}^{m}t_{j}\log t_{j}\geqslant \log m-\frac{m\log(m-1)}{m-2}[1-B(P;P_{U})^{2}], \end{equation} \tag{6} $$
where $B(P;P_{U})=\sum_{j=1}^{m}\sqrt{t_{j}/m}$ is the Bhattacharyya coefficient between the probability distribution $P=(t_{1},\dots ,t_{m}) $ and the uniform distribution $P_{U}$. It can be compared with the lower bound of Theorem 3.11 in [5] in terms of the total variation distance. For $P$ close to $P_{U}$ the bound (6) is considerably better, while close to degenerate probability distributions it can be negative. This bound also turns out to be the entropy inequality (2) for strongly acute pyramids with $1/m\leqslant p\leqslant (m-1)/m$.

In the case of the ensemble of flat pyramid ($\xi =-1)$ the corresponding tight entropy inequality is given by the following statement.

Theorem 3. For all complex $z_{j}$ satisfying $\sum_{j=1}^{m}|z_{j}|^{2}=1$ and $\sum_{j=1}^{m}z_{j}=0$,

$$ \begin{equation} -\sum_{j=1}^{m}|z_{j}|^{2}\log|z_{j}|^{2}\geqslant \begin{cases} 1, & m\leqslant 6; \\ \log m-\dfrac{m-2}{m}\log (m-1), & m\geqslant 7. \end{cases} \end{equation} \tag{7} $$
In the case $m\leqslant 6$ equality in (7) is attained for $z_{1}=-z_{2}=1/\sqrt{2}$ and $z_{j}=0$ for $j\geqslant 3$; and in the case $m\geqslant 7$ for $z_{1}=\sqrt{(m-1)/m}$ and $z_{j}=-\sqrt{1/((m-1)m)}$ for $j\geqslant 2$ (and for all permutations of such $z_{j})$.

Corollary. The conjectures concerning globally optimal observables for the accessible information of an ensemble of quantum pyramid [3] hold true in the cases of acute and flat pyramids.

The proofs are presented in the preprint [6], which also contains derivations of tight entropy inequalities in the case of obtuse ($-1<\xi<0$) quantum pyramids.


Bibliography

1. A. S. Holevo, Lobachevskii J. Math., 44:6 (2023), 2033–2043  mathnet  crossref  mathscinet  zmath
2. L. Gross, Amer. J. Math., 97:4 (1975), 1061–1083  crossref  mathscinet  zmath
3. B.-G. Englert and J. Řeháček, J. Modern Opt., 57:3 (2010), 218–226  crossref  mathscinet  zmath  adsnasa
4. E. Davies, IEEE Trans. Inform. Theory, 24:5 (1978), 596–599  crossref  mathscinet  zmath
5. S. M. Moser, Information theory, Lecture notes, 6th ed., ETH Zürich, Zürich; National Yang Ming Chiao Tung Univ. (NYCU), 2018, xvii+570 pp. https://moser-isi.ethz.ch/docs/it_script_v617.pdf
6. A. S. Holevo and A. V. Utkin, Quantum accessible information and classical entropy inequalities, 2025, 42 pp., arXiv: 2506.06700

Citation: A. S. Holevo, “Tight lower bounds for Shannon entropy from ‘quantum pyramids’”, Russian Math. Surveys, 80:4 (2025), 729–731
Citation in format AMSBIB
\Bibitem{Hol25}
\by A.~S.~Holevo
\paper Tight lower bounds for Shannon entropy from `quantum pyramids'
\jour Russian Math. Surveys
\yr 2025
\vol 80
\issue 4
\pages 729--731
\mathnet{http://mi.mathnet.ru/eng/rm10257}
\crossref{https://doi.org/10.4213/rm10257e}
Linking options:
  • https://www.mathnet.ru/eng/rm10257
  • https://doi.org/10.4213/rm10257e
  • https://www.mathnet.ru/eng/rm/v80/i4/p181
  • Citing articles in Google Scholar: Russian citations, English citations
    Related articles in Google Scholar: Russian articles, English articles
    Успехи математических наук Russian Mathematical Surveys
     
      Contact us:
     Terms of Use  Registration to the website  Logotypes © Steklov Mathematical Institute RAS, 2025