Sub-Linear Root Detection, and New Hardness Results, for Sparse Polynomials Over Finite Fields

We present a deterministic 2^O(t)q^{(t-2)(t-1)+o(1)} algorithm to decide whether a univariate polynomial f, with exactly t monomial terms and degree <q, has a root in F_q. A corollary of our method — the first with complexity sub-linear in q when t is fixed — is that the nonzero roots in F_q can be partitioned into at most 2 \sqrt{t-1} (q-1)^{(t-2)(t-1)} cosets of two subgroups S_1,S_2 of F^*_q, with S_1 in S_2. Another corollary is the first deterministic sub-linear algorithm for detecting common degree one factors of k-tuples of t-nomials in F_q[x] when k and t are fixed. When t is not fixed we show that each of the following problems is NP-hard with respect to BPP-reductions, even when p is prime: (1) detecting roots in F_p for f, (2) deciding whether the square of a degree one polynomial in F_p[x] divides f, (3) deciding whether the discriminant of f vanishes, (4) deciding whether the gcd of two t-nomials in F_p[x] has positive degree. Finally, we prove that if the complexity of root detection is sub-linear (in a refined sense), relative to the straight-line program encoding, then NEXP is not in P/Poly.

💡 Research Summary

The paper investigates the computational problem of detecting roots of sparse univariate polynomials over finite fields. A sparse polynomial (often called a t‑nomial) is a polynomial that contains exactly t non‑zero monomial terms, and the authors focus on the case where the degree of the polynomial is strictly less than the size q of the field F_q.

Main algorithmic contribution
The authors present a deterministic algorithm whose running time is
  2^{O(t)}·q^{(t‑2)(t‑1)+o(1)}.
When t is treated as a constant, the exponent of q becomes (t‑2)(t‑1), which is strictly smaller than q for any fixed t≥3, and for t=2 the algorithm runs in essentially constant time. The algorithm proceeds by exploiting the multiplicative group structure of F_q^. It selects two nested cyclic subgroups S₁⊂S₂ of F_q^ and shows that any non‑zero root of the input polynomial must lie in a coset of either S₁ or S₂. The number of such cosets is bounded by 2·√(t‑1)·(q‑1)^{(t‑2)(t‑1)}. Consequently, the search space is reduced from q‑1 elements to a dramatically smaller set, and exhaustive checking of these cosets yields the desired decision.

Corollary on root structure
From the same analysis the authors deduce a structural description of the set of non‑zero roots: they can be partitioned into at most 2√(t‑1)·(q‑1)^{(t‑2)(t‑1)} cosets of the two subgroups S₁ and S₂, with S₁ being a subgroup of S₂. This gives a clear algebraic picture of how sparsity forces roots to concentrate in highly regular patterns.

Common linear factor detection
The technique extends to the problem of detecting a common linear factor among k t‑nomials when both k and t are fixed. By mapping each polynomial’s potential linear factors to the same family of cosets, the algorithm checks for intersections among the k sets. The resulting procedure also runs in sub‑linear time in q, providing the first deterministic algorithm with this property for the multi‑polynomial setting.

Hardness results for unrestricted t
When t is not bounded, the paper establishes NP‑hardness (under BPP‑reductions) for four natural decision problems over prime fields F_p:

1. Root existence – deciding whether a given t‑nomial has a root in F_p.
2. Square‑factor detection – deciding whether the square of a linear polynomial divides the t‑nomial.
3. Discriminant zero test – deciding whether the discriminant of the t‑nomial vanishes (equivalently, whether the polynomial has a repeated root).
4. GCD degree test – deciding whether the greatest common divisor of two t‑nomials has positive degree.

These reductions show that, despite the apparent simplicity of sparse representations, the associated algebraic decision problems remain computationally intractable when the sparsity parameter t is part of the input.

Implications for straight‑line program (SLP) encodings
The authors also consider a more compressed representation of polynomials via straight‑line programs. They prove that if root detection for SLP‑encoded polynomials could be performed in sub‑linear time (in a refined sense that accounts for the size of the program rather than the expanded degree), then the complexity class NEXP would not be contained in P/Poly. This conditional lower bound links the efficiency of root‑finding algorithms to a major separation in non‑uniform complexity theory, indicating that a dramatically faster algorithm would have far‑reaching consequences beyond the immediate problem.

Overall significance
The paper delivers a rare combination of algorithmic breakthroughs and hardness proofs. On the algorithmic side, it shows that for fixed sparsity the root‑finding problem over finite fields can be solved in time sub‑linear in the field size, a result that had not been achieved before. The structural insight about roots lying in a bounded number of cosets of nested subgroups is both elegant and powerful, enabling extensions to common‑factor detection among multiple polynomials. On the hardness side, the authors demonstrate that once sparsity is allowed to grow with the input, the natural algebraic decision problems become NP‑hard, even under randomized reductions, and that any further improvement in the SLP model would imply a major complexity‑theoretic separation. Together, these contributions deepen our understanding of how sparsity interacts with algebraic computation and provide a clear roadmap for both practical algorithm design and theoretical limits in the area of finite‑field polynomial arithmetic.