[Paper Review] A Group-Theoretic Framework for the Construction of Packings in Grassmannian Spaces

A.R. Calderbank, R. H. Hardin+3 more|ArXiv.org|Aug 1, 2002

graph theory and CDMA systems8 references4 citations

TL;DR

This paper presents a group-theoretic framework using orthogonal spaces and the Clifford group to construct optimal packings of $2^k$-dimensional subspaces in $\mathbb{R}^{2^i}$, achieving maximal minimal distance via totally isotropic subspaces in $\Omega^+(2i,2)$. Key results include infinite families of optimal Grassmannian packings linked to Kerdock sets, Barnes-Wall lattices, and quantum codes, with explicit constructions meeting the simplex and orthoplex bounds.

ABSTRACT

By using totally isotropic subspaces in an orthogonal space Omega^{+}(2i,2), several infinite families of packings of 2^k-dimensional subspaces of real 2^i-dimensional space are constructed, some of which are shown to be optimal packings. A certain Clifford group underlies the construction and links this problem with Barnes-Wall lattices, Kerdock sets and quantum-error-correcting codes.

Motivation & Objective

• To develop an algebraic framework for constructing optimal packings of $n$-dimensional subspaces in $\mathbb{R}^m$ with maximal minimal distance.
• To unify and generalize known constructions of Grassmannian packings using group-theoretic structures, particularly the extraspecial 2-group and Clifford group.
• To establish connections between Grassmannian packings, Kerdock sets, Barnes-Wall lattices, and quantum-error-correcting codes through a common algebraic foundation.
• To provide explicit constructions of infinite families of optimal packings, including those meeting the simplex and orthoplex bounds.
• To tabulate and classify parameters of these packings up to dimension 128, identifying optimal configurations.

Proposed method

• The construction uses the orthogonal space $\Omega^+(2i,2)$ and its associated quadratic form on the quotient of an extraspecial 2-group $E$ by its center, yielding totally isotropic subspaces.
• The Clifford group $L$, generated by $X(a)$, $Y(b)$, permutation matrices, and the Hadamard matrix $H$, acts on the vector space $V = \mathbb{R}^{2^i}$, preserving the Grassmannian metric.
• Projection matrices $\Pi_P = A^tA$ represent $n$-planes $P \in G(m,n)$, and the distance is computed via $d^2(P,Q) = \frac{1}{2}\|\Pi_P - \Pi_Q\|^2$, linking geometry to Frobenius norm.
• The main constructions arise from orbits of totally isotropic subspaces under the Clifford group, with Theorem 1 yielding $N = 2^{i^2 + i + 1}(2^i - 1)\prod_{j=1}^{i-1}(4^j - 1)$ planes in $G(2^i, 2^{i-1})$.
• Theorem 3 constructs a distinct family using cyclic shifts of a base plane in odd dimensions $m = p$, yielding $N = p(p+1)/2$ planes in $G(p, (p-1)/2)$ with $d^2 = (p+1)^2/(4(p+2))$.
• Theoretical bounds (simplex and orthoplex) are used to verify optimality, with equality indicating optimal configurations.

Experimental results

Research questions

• RQ1 Can a unified algebraic framework be developed to construct optimal Grassmannian packings using group-theoretic structures?
• RQ2 How do constructions based on totally isotropic subspaces in $\Omega^+(2i,2)$ yield infinite families of optimal packings in $G(2^i, 2^{i-1})$?
• RQ3 What is the role of the Clifford group in linking Grassmannian packings to Kerdock sets, Barnes-Wall lattices, and quantum codes?
• RQ4 Do the packings constructed via cyclic shifts in odd dimensions $m = p$ achieve the simplex bound and thus are optimal?
• RQ5 How do the derived constructions compare to previously known packings in terms of distance and cardinality?

Key findings

• Theorem 1 constructs optimal packings in $G(2^i, 2^{i-1})$ with $N = 2^{i^2 + i + 1}(2^i - 1)\prod_{j=1}^{i-1}(4^j - 1)$ planes, achieving the simplex bound when $N \leq \binom{m+1}{2}$.
• Theorem 2a produces packings in $G(2^i, 2^{i-1})$ with $N = 2^{i^2 + i + 1}(2^i - 1)\prod_{j=1}^{i-1}(4^j - 1)$, matching known optimal values in dimensions 4, 8, 16, 32, 64, 128.
• Theorem 3 yields a distinct family of optimal packings in $G(p, (p-1)/2)$ for odd primes $p$, with $N = p(p+1)/2$ planes and $d^2 = (p+1)^2/(4(p+2))$, meeting the simplex bound.
• In dimension $m=8$, the construction yields $N=420$ planes in $G(8,4)$ with $d^2=1$, matching the simplex bound and confirming optimality.
• For $m=16$, the framework produces $N=16200$ planes in $G(16,8)$ with $d^2=1$, achieving the orthoplex bound and confirming optimality.
• The paper tabulates optimal packings up to dimension 128, identifying constructions (1), (1a), (2a)–(2e), and (3) as achieving or matching known optimal bounds.