[Paper Review] New linear independence measures for values of q-hypergeometric series

Igor Rochev|arXiv (Cornell University)|Jun 28, 2010

Advanced Mathematical Identities2 references3 citations

TL;DR

This paper establishes new linear independence measures for values of q-hypergeometric series by constructing auxiliary linear forms using difference operators and shift operators. Under specific conditions on q, P(z), and parameters αj, the method proves that certain combinations of f(σ)(αj qk) are linearly independent over ℚ, with an explicit lower bound of order H^{−μ−C₀/√log H} for non-zero linear forms, where μ = (M−1)/(1−Mγ) and M depends on deg P and the exponents s_j.

ABSTRACT

We prove linear independence results for values of (a certain class of) q-hypergeometric series in a quantitative form.

Motivation & Objective

• To establish effective linear independence measures for values of q-hypergeometric series f(z) = ∑ z^n / ∏_{k=1}^n P(q^k).
• To extend previous results by providing sharper quantitative bounds under weaker assumptions on the parameter q.
• To prove linear independence of the numbers 1, f^{(σ)}(α_j q^k) over ℚ under conditions ensuring no multiplicative dependence among α_j and no vanishing at poles.
• To refine the exponent in the lower bound estimate for linear forms in these values, improving upon earlier O((log H)^{3/2}) bounds.

Proposed method

• Construct auxiliary linear forms u_n(𝐱) and v_n(𝐱) using sequences of derivatives and shift operators, with v_n satisfying a recurrence v_n = P(q^n)v_{n-1} + u_n.
• Introduce the difference operator D_a = I - aB, where B is the backward shift, and use its properties to annihilate polynomial-exponential sequences.
• Define higher-order forms v_{l,n}(𝐱) = ∏_{k=1}^l ∏_{j=1}^m D_{α_j q^{-k}}^{s_j} (v_n(𝐱)) to control growth and ensure non-vanishing.
• Use the non-vanishing lemma (Lemma 5) to guarantee that for any non-zero coefficient vector, some v_{l,n} does not vanish.
• Scale the forms w_{l,n}(𝐱) = D^n q_1^{Sl(l+1)/2} q_2^{dn(n+1)/2} v_{l,n}(𝐱) to ensure integrality and control coefficient size.
• Apply the saddle-point method with l ≈ (L/a)^{1/2}, L = log H / log |q_1|, to derive the final lower bound on linear forms.

Experimental results

Research questions

• RQ1 Under what conditions on q, P(z), and α_j are the values f^{(σ)}(α_j q^k) linearly independent over ℚ?
• RQ2 Can the exponent in the lower bound for linear forms in these values be improved beyond O((log H)^{3/2})?
• RQ3 How does the structure of P(z), especially its degree and leading coefficient, affect the linear independence measure?
• RQ4 What role does the Diophantine approximation constant γ = log|q_2|/log|q_1| play in determining the sharpness of the bound?
• RQ5 Can the method be adapted to handle cases where P(0) = 0 or when all roots of P are rational?

Key findings

• The numbers 1, f^{(σ)}(α_j q^k) for 1 ≤ j ≤ m, 0 ≤ k < d, 0 ≤ σ < s_j are linearly independent over ℚ if γ < 1/M, where M is defined via deg P and s_j.
• A sharp lower bound is established: |A_0 + ∑ A_{j,k,σ} f^{(σ)}(α_j q^k)| ≥ H^{−μ−C₀/√log H}, with μ = (M−1)/(1−Mγ).
• The bound improves upon previous results that gave only exp(−C(log H)^{3/2}) estimates, now achieving a power-logarithmic decay rate.
• The constant C₀ depends only on q, P, m, α_j, s_j, and is absolute for fixed parameters.
• The method works uniformly for both cases P(z) = p_d z^d and general P(z), with M adjusted accordingly.
• The non-vanishing of the scaled forms w_{l,n}(𝐱) is guaranteed for some n in a short interval, enabling the use of the maximum modulus principle in the final estimate.