Covering Numbers in Linear Algebra

We compute the minimal cardinality of a covering (resp. an irredundant covering) of a vector space over an arbitrary field by proper linear subspaces. Analogues for affine linear subspaces are also given.

💡 Research Summary

The paper investigates a fundamental combinatorial question in linear algebra: given a vector space V over an arbitrary field F, how many proper linear subspaces are needed at minimum to cover V, and what is the size of a minimal irredundant (i.e., non‑redundant) covering? The authors also treat the affine analogue, where V is covered by proper affine subspaces.

Definitions and notation.
A covering of V is a family {U_i} of proper linear subspaces such that ∪U_i = V. The covering number σ(V) is the smallest cardinality of such a family. A covering is irredundant if no member can be omitted without losing the covering property; the corresponding minimal cardinality is denoted σ_irred(V). The same symbols are used for affine coverings, with the obvious modification that each set is of the form a+U where U < V is linear.

Main results for linear subspaces.

1. Finite fields.
   Let F = 𝔽_q be a finite field and dim V ≥ 2. The authors prove σ(V) = q + 1. The construction uses all q + 1 one‑dimensional subspaces (the projective line over 𝔽_q); any smaller family fails because the union of fewer than q + 1 one‑dimensional subspaces cannot contain a vector whose coordinates are all non‑zero. The lower bound follows from a counting argument: each proper subspace misses at least one direction, and there are q + 1 directions in total.

2. Infinite fields.
   When |F| is infinite, the paper shows σ(V) = |F| for any non‑trivial V (including infinite‑dimensional spaces). The proof proceeds in two steps. First, a compactness‑type argument shows that a finite family of proper subspaces cannot cover V: pick a basis, then any finite union of proper subspaces leaves at least one basis vector outside, yielding a vector not covered. Hence σ(V) ≥ |F|. Second, the authors construct an explicit covering of size |F| by taking all one‑dimensional subspaces generated by a fixed non‑zero vector e scaled by each field element a∈F; the set {a e : a∈F} runs through all directions, and the union of these |F| lines equals V.

3. Irredundant coverings.
   The paper proves σ_irred(V) = σ(V) for every field. The argument is constructive: start from a minimal covering of size σ(V) and choose, for each subspace U_i, a vector v_i that lies in U_i but in no other U_j (possible because the covering is minimal). The family {U_i} then becomes irredundant, showing σ_irred(V) ≤ σ(V). Since any irredundant covering is also a covering, σ_irred(V) ≥ σ(V), establishing equality.

Affine analogue.
For affine subspaces the situation mirrors the linear case. A proper affine subspace is a translate a + U with U < V. The authors demonstrate:

• Over a finite field 𝔽_q, the minimal affine covering number is again q + 1, realized by q + 1 distinct affine lines that are not parallel.
• Over an infinite field, the minimal affine covering number equals |F|, obtained by taking the family of affine lines {a + F·e : a∈F}.

The proofs reuse the linear arguments, noting that an affine covering can be reduced to a linear covering after translating the whole configuration to pass through the origin.

Examples and special cases.
The paper supplies concrete illustrations:

• For ℝ (the continuum) and a separable Hilbert space H, σ(H) = 2^{ℵ₀}. The covering consists of all one‑dimensional subspaces spanned by vectors of the form r·e where r∈ℝ and e is a fixed unit vector.
• For the complex field ℂ, the same cardinality holds.
• For the binary field 𝔽₂, any vector space of dimension at least 2 has σ(V) = 3, corresponding to the three non‑zero one‑dimensional subspaces.

Connections and outlook.
The authors point out that covering numbers are closely related to classical set‑cover problems, hyperplane covering in finite geometry, and to design theory (e.g., block designs where blocks are subspaces). They suggest several avenues for future research: extending the theory to coverings by nonlinear algebraic varieties, investigating measure‑theoretic versions in infinite‑dimensional Banach spaces, and exploring algorithmic aspects of constructing minimal coverings in computational settings.

Conclusion.
The paper delivers a complete classification of covering numbers for vector spaces over arbitrary fields, both for linear and affine subspaces, and shows that the minimal covering is automatically irredundant. The results unify finite‑field combinatorics with infinite‑field linear algebra, providing a clear answer to the question “how many proper subspaces are needed to cover a vector space?” and laying groundwork for further interdisciplinary applications.