[Paper Review] Congruences for L-functions of additive exponential sums

Régis Blache|arXiv (Cornell University)|Jun 7, 2012

Coding theory and cryptography7 references3 citations

TL;DR

This paper establishes a non-trivial congruence modulo $ I_{ ho} $ for $ L $-functions associated with additive exponential sums over affine space in positive characteristic, extending Manin's congruence for Jacobians of curves. Using Dwork's theory and $ p $-adic analysis, it shows that the $ L $-function is congruent modulo $ I_{ ho} $ to a product of determinants involving matrices derived from the polynomial's coefficients and their Frobenius twists, with the congruence holding precisely when the $ p $-density of the support satisfies $ \delta_p(D_I) + n - \#I = \delta_p(D) $.

ABSTRACT

We give a congruence for L-functions coming from affine additive exponential sums over a finite field. Precisely, we give a congruence for certain operators coming from Dwork's theory. This congruence is very similar to the congruence of Manin for the characteristic polynomial of the action of Frobenius on the Jacobian of a curve defined over a finite field.

Motivation & Objective

• To extend Manin’s congruence for zeta functions of curves to $ L $-functions of additive exponential sums over affine space in positive characteristic.
• To establish a non-trivial congruence modulo $ I_{ ho} $, a $ \pi $-adic ideal defined by $ p $-density, rather than trivial modulo $ \pi $, for $ L $-functions arising from exponential sums.
• To characterize the $ L $-function’s structure in terms of matrices $ M(\Gamma_I) $ derived from the polynomial’s coefficients and their Frobenius actions.
• To provide a framework valid even when the Newton polygon has no horizontal slopes, overcoming limitations of prior congruences.
• To unify and generalize results from Artin-Schreier curves and projective hypersurfaces via a single $ p $-adic congruence.

Proposed method

• Define the $ L $-function $ L(\mathbb{A}^n, f; T) $ as the exponential generating series of additive exponential sums $ S_r(f) $ over finite field extensions.
• Use Dwork’s theory to lift coefficients $ c_{\mathbf{d}} $ to Teichmüller representatives $ \gamma_{\mathbf{d}} $ in $ \mathbb{Z}_p[\zeta_p] $, forming the matrix $ \Gamma $.
• Introduce the $ p $-density $ \delta_p(D) $ as a lower bound for $ q $-adic valuations of reciprocal roots and poles of the $ L $-function.
• Define the ideal $ I_{\delta} $ as the set of power series with coefficients having $ q $-adic valuation strictly greater than $ \delta i $, enabling non-trivial congruences.
• Construct matrices $ M(\Gamma_I) $ from the $ p $-minimal support of subsets $ D_I $, and define their Frobenius twists $ M(\Gamma_I)^{\tau^{m-1}} \cdots M(\Gamma_I) $.
• Establish the congruence $ L(\mathbb{A}^n, f; T) \equiv \prod \det(\mathbf{I}_{N_I} - \pi^{m(p-1)\delta} T M(\Gamma_I)^{\tau^{m-1}} \cdots M(\Gamma_I))^{(-1)^{\#I+1}} \mod I_{\delta} $, where the product runs over $ I $ with $ \delta_p(D_I) + n - \#I = \delta_p(D) $.

Experimental results

Research questions

• RQ1 Can a non-trivial $ p $-adic congruence be established for $ L $-functions of additive exponential sums over affine space, analogous to Manin’s congruence for Jacobians?
• RQ2 How does the $ p $-density $ \delta_p(D) $ of the support of a polynomial control the $ q $-adic valuation of the $ L $-function’s coefficients?
• RQ3 What is the precise algebraic structure of the $ L $-function modulo the ideal $ I_{\delta} $, and how is it encoded in matrices derived from the polynomial’s coefficients?
• RQ4 Under what conditions does the congruence reduce to a single determinant expression, and when is the right-hand side a polynomial?
• RQ5 How does this congruence recover or generalize known results for Artin-Schreier curves and projective hypersurfaces?

Key findings

• The $ L $-function $ L(\mathbb{A}^n, f; T) $ lies in the ring $ M_{\delta} $, consisting of power series with coefficients of $ q $-adic valuation at least $ \delta i $, where $ \delta = \delta_p(D) $.
• The congruence holds modulo the ideal $ I_{\delta} $, which captures series with coefficients of valuation strictly greater than $ \delta i $, ensuring non-triviality.
• The congruence expresses the $ L $-function as a product of determinants involving matrices $ M(\Gamma_I) $, raised to alternating powers based on the size of index sets $ I $, with $ \pi^{m(p-1)\delta} T $ scaling the variable.
• The product is restricted to subsets $ I \subset \{1,\dots,n\} $ such that $ \delta_p(D_I) + n - \#I = \delta_p(D) $, ensuring the correct valuation balance.
• In the case where $ \delta_p(D_I) + n - \#I > \delta_p(D) $ for all $ I $, the congruence simplifies to $ L(\mathbb{A}^n, f; T)^{(-1)^{n+1}} \equiv \det(\mathbf{I}_N - M(\Gamma)^{\tau^{m-1}} \cdots M(\Gamma)(\pi^{m(p-1)\delta} T)) \mod I_{\delta} $, yielding a single determinant.
• The result generalizes Miller’s matrix description of the Cartier operator on $ H^0(V, \Omega^{n-1}) $ for projective hypersurfaces, when $ f = yF $, by recovering his matrix as $ M(\Gamma_I)^{\tau^{m-1}} \cdots M(\Gamma_I) $.