Hemisystems in Hermitian Polar Spaces

• Hemisystems in Hermitian polar spaces are subsets of generators meeting each point in exactly half the available lines, offering a clear combinatorial configuration.
• They are constructed through advanced group-theoretic and algebraic methods, showcasing families like the Cossidente-Penttila and cyclotomic examples with strict arithmetic properties.
• These configurations yield strongly regular graphs and two-weight codes, impacting finite geometry, design theory, and coding theory.

A Hermitian polar space is a finite incidence geometry arising from the totally isotropic subspaces of a vector space equipped with a nondegenerate Hermitian form over a finite field. The archetype, the Hermitian surface $\mathrm{H}(3, q^2)$, is central to the theory of generalized quadrangles and has been a locus of deep structure in finite geometry and combinatorics. Hemisystems in such spaces—subsets of generators (lines or higher-dimensional analogs) meeting each point in precisely half the available generators—arise at the confluence of group actions, design theory, polar space combinatorics, and algebraic methods. The existence, classification, and construction of hemisystems remain focal problems, with key results revealing both infinite families and stringent nonexistence contradictions in higher ranks.

1. Structure and Fundamentals of Hermitian Polar Spaces

Let $V = \mathrm{GF}(q^2)^N$ with a Hermitian form $h$, i.e., $h(x, y) = \sum x_i y_i^q$ over $\mathrm{GF}(q^2)$. The Hermitian polar space $\mathrm{H}(n, q^2)$ comprises the totally isotropic subspaces: points correspond to $1$-spaces $〈x〉\subseteq V$ with $h(x,x)=0$, and the maximal totally isotropic subspaces (generators) are of projective dimension $d-1 = \left\lfloor\frac{n+1}{2}\right\rfloor-1$.

For the case $n=3$, $\mathrm{H}(3,q^2)$ is a generalized quadrangle of order $(q^2, q)$:

• Points are totally isotropic lines (projective $1$-spaces).
• Lines are totally isotropic planes ($2$-dimensional subspaces).
• Each line contains $q^2+1$ points, each point lies on $q+1$ lines.
• The space admits $|\mathcal{M}| = (q^2+1)(q^3+1)$ generators, and $|\mathrm{H}(3,q^2)| = (q^3+1)(q+1)$ points.

In higher dimensions, $\mathrm{H}(2d-1, q^2)$ is the unique classical, nondegenerate Hermitian polar space of rank $d$ (Smaldore, 2024).

2. Definition and Properties of Hemisystems

A hemisystem in $\mathrm{H}(3, q^2)$ is a set $\mathcal{H}$ of lines (generators) such that each point is incident with exactly $(q+1)/2$ lines from $\mathcal{H}$:

$$|\{\ell \in \mathcal{H} \mid P \in \ell\}| = \frac{q+1}{2},\quad \forall P \in \mathrm{H}(3, q^2)$$

and

$|\mathcal{H}| = \frac{(q^3+1)(q+1)}{2}$

(Bamberg et al., 2015, Korchmáros et al., 2017, Lavorante et al., 2021). Equivalently, in the dual elliptic quadric $Q^-(5,q)$, a hemisystem is a $(q+1)/2$-ovoid: a subset meeting every maximal totally singular subspace in $(q+1)/2$ points.

The hemisystem property imposes severe arithmetic restrictions and only exists for $q$ odd, inherited from the classical impossibility of partitioning the set of lines through a point into more than two equivalent classes (Smaldore, 2024). For $\mathrm{H}(3, q^2)$, hemisystems yield point-line configurations with uniform intersection numbers, producing strongly regular graphs and two-weight error-correcting codes (Lavorante et al., 2021).

3. Infinite Families and Construction Techniques

Major progress in the classification and construction of hemisystems in Hermitian polar spaces centers on explicit infinite families, unified by group-theoretic and algebraic constructions:

• Cossidente-Penttila Family: For all odd $q$, hemisystems invariant under subgroups isomorphic to $\mathrm{PSL}(2, q)\times C_{(q+1)/2}$ exist (Smaldore, 2024, Bamberg et al., 2015).
• Bamberg-Lee-Momihara-Xiang Cyclotomic Family: For all $q\equiv3 \pmod{4}$, hemisystems are constructed in the dual via index sets in the cyclotomic classes of $\mathbb{F}_{q^6}$, admitting the group $C_{(q^3+1)/4}:C_3$ as automorphisms (Bamberg et al., 2015). The construction uses the evaluation of Gauss sums and character theory on field traces.
• Maximal Curve Constructions (Korchmáros–Nagy–Speziali, Pallozzi Lavorante–Smaldore): For primes $p = 1 + 4a^2$ or more restrictively $p = 1 + 16 n^2$, explicit embeddings of maximal curves in $U_3$ yield hemisystems stabilized by $\mathrm{PSL}(2, p)\times C_{(p+1)/2}$ (Lavorante et al., 2021, Korchmáros et al., 2017).
• Singer-type Invariant Hemisystems: Computational evidence and explicit calculations for $q\leq29$ suggest the existence of hemisystems admitting a metacyclic group of shape $(q^2-q+1):6$ as automorphism group for all odd $q\not\equiv1\pmod{12}$, constructed from unions of field-theoretically defined orbits in $Q^-(5,q)$. Conjecturally, this subsumes a large new family (Bamberg et al., 2010).
• 2⁴●A₅-invariant Hemisystems: For $q\equiv3\pmod{4}$, hemisystems invariant under $2^4\cdot A_5$ subgroups are observed for various small $q$, with no currently known obstruction for arbitrarily large $q$ (Bamberg et al., 2010).

Relative hemisystems generalize the notion to subsets intersecting only external points (that is, points outside a fixed embedded symplectic subquadrangle) in exactly $q/2$ lines (Bamberg et al., 2015). Sufficient criteria for the existence of relative hemisystems have been unified via semiregular group actions and orbit-stabilizer techniques.

4. Classification, Nonexistence, and Higher Dimensions

The classification and nonexistence results sharply demarcate the landscape:

• In $\mathrm{H}(3, q^2)$, only hemisystems (no other $m$-ovoids) can exist (Smaldore, 2024).
• All known infinite families and sporadic examples align with the group-theoretic constructions or their extensions described above (Bamberg et al., 2015, Bamberg et al., 2010, Lavorante et al., 2021).

For higher rank Hermitian polar spaces $\mathrm{H}(2d-1, q^2)$, $d>2$, the nonexistence of hemisystems (i.e., $m$-ovoids with $m=\frac12(q^d+1)$) is established for all odd $q>3$ via the connection to classical distance-regular graphs of negative type:

• Vanhove established that a hemisystem would yield a distance-regular induced subgraph with parameters corresponding to Weng's "last category" (Adriaensen et al., 19 Nov 2025);
• Tian et al. eliminated this category for diameter $3$, and by subgraph induction it is ruled out for all $d>2$, thus no hemisystems in $\mathrm{H}(2d-1,q^2)$, $d>2$, for odd $q\ne3$ (Adriaensen et al., 19 Nov 2025).

5. Automorphism Groups and Combinatorial Implications

Automorphism groups play a crucial role in both existence proofs and classification:

• Cossidente-Penttila hemisystems admit $\mathrm{PSL}(2, q)\times C_{(q+1)/2}$.
• Cyclotomic hemisystems are preserved by $C_{(q^3+1)/4}:C_3$ (Bamberg et al., 2015).
• Maximal-curve induced examples admit index-2 subgroups as stabilizers, intimately tied to the structure of the curves and their orbits (Korchmáros et al., 2017, Lavorante et al., 2021).

Hemisystems correspond (dually) to strongly regular graphs with parameters $(v, k, \lambda, \mu)$ directly computable from $q$:

$$v = \frac{(q^3 + 1)(q + 1)}{2},\quad k = \frac{(q^2 + 1)(q - 1)}{2}$$

and codes of two weights via the Klein correspondence (Lavorante et al., 2021).

Relative hemisystems exploit group actions with orbit structures mirroring those of the quadrangle automorphism group, with explicit semiregularity and orbit partitioning conditions ensuring existence (Bamberg et al., 2015).

6. Computational Classification and Open Problems

Systematic computational techniques—ranging from GAP/FinInG-based orbit enumeration to solving integer linear programs with Gurobi—have yielded complete classifications for small $q$ and revealed possible new infinite family patterns, notably the Singer-invariant and $2^4\cdot A_5$-invariant hemisystems (Bamberg et al., 2010). In $H(3,4^2)$, exhaustive search showed all 240 relative hemisystems are equivalent to the Penttila–Williford example (Bamberg et al., 2015).

Central open problems include:

• Full classification in $H(3, q^2)$, especially for $q\equiv1\pmod{4}$, where known cyclotomic techniques fail (Bamberg et al., 2015).
• Existence of hemisystems with trivial automorphism group.
• Generalizations to relative $m$-systems for alternative embedded subquadrangles and to higher rank Hermitian spaces (Bamberg et al., 2015).
• The Landau conjecture on infinitely many primes of the form $p=1+16n^2$, critical for the corresponding maximal-curve based examples (Korchmáros et al., 2017).
• The existence of new infinite families not covered by the current metacyclic or group-theoretic frameworks.

A summary table of currently known infinite families in $H(3,q^2)$:

Family/Type	Parameter restriction	Automorphism group
Cossidente-Penttila	$q$ odd	$\mathrm{PSL}(2, q)\times C_{(q+1)/2}$
Bamberg-Lee-Momihara-Xiang	$q\equiv3\pmod{4}$	$C_{(q^3+1)/4}:C_3$
Singer-invariant (conj.)	$q$ odd, $q\not\equiv1 \pmod{12}$	$(q^2-q+1):6$
$2^4\cdot A_5$-invariant (conj.)	$q\equiv3\pmod{4}$	$2^4\cdot A_5$
Maximal-curve families	$p=1+4a^2$ or $1+16 n^2$	$\mathrm{PSL}(2, p)\times C_{(p+1)/2}$

7. Connections, Impact, and Future Directions

The study of hemisystems in Hermitian polar spaces directly impacts the combinatorial theory of $m$-ovoids, the theory of regular graphs, association schemes, and finite geometries. Hemisystems are central to the existence of partial spreads, block designs, and geometries with prescribed automorphism groups. Through duality, they yield projective two-weight codes and strongly regular graphs, creating bridges to coding theory (Lavorante et al., 2021).

Progress in the field depends on new group-theoretic, algebro-combinatorial, and computational techniques. The extension of current existence proofs to $q$ even, to higher-dimensional Hermitian spaces, and the exploration of exotic automorphism groups remain open frontiers. Moreover, the link between the arithmetic of maximal curves, deep number-theoretic conjectures such as Landau’s, and combinatorial configurations in finite geometry continues to represent an area of active research (Korchmáros et al., 2017, Lavorante et al., 2021, Smaldore, 2024, Adriaensen et al., 19 Nov 2025).