Heights bound root multiplicities #
A root β of multiplicity m of a polynomial P over a number field K contributes the
factor (X - β) ^ m, whose m linear factors each have coefficient height H(β). Gelfond's
inequality bounds the product of the heights of the factors by the height of P, which gives
m * h(β) ≤ h(P) + [K : ℚ] * deg P * log 2
for relative logarithmic heights. Thus a polynomial of small height cannot vanish to high order
at a point of large height. In terms of indices: if P ≠ 0, r ≥ deg P and r * h(β) is at
least s⁻¹ * (h(P) + [K : ℚ] * r * log 2) for some s > 0, then the index of P at β with
weight r is at most s. This is Roth's lemma in one variable, the base case of the inductive
bound on indices of auxiliary polynomials in the proof of Roth's theorem.
Main results #
Polynomial.rootMultiplicity_mul_logHeight₁_le:m * h(β) ≤ h(P) + [K : ℚ] * deg P * log 2.MvPolynomial.weightedIndex_toMvPolynomial_le: the index form of Roth's lemma in one variable, for a univariate polynomial placed in one coordinate.
References #
- L. Pottmeyer, Diophantine Approximation (2022), §3.4, Lemma 3.4.3.
- The statement of the corresponding lemma in Ralf Stephan's
DiophantineApproximation/RothBaseCase.leaninrwst/Subspace-Theorems, snapshotc26d909dc48cf7fe(Apache 2.0), was consulted for reference; the proof here is independent.
Roth's lemma in one variable, multiplicity form: the multiplicity of a root β of a
polynomial P over a number field, times the relative logarithmic height of β, is at most
h(P) + [K : ℚ] * deg P * log 2. The zero polynomial has root multiplicity zero.
Roth's lemma in one variable, index form. Let P ≠ 0 be placed in the coordinate i,
with deg P ≤ r i. If r i * h(a i) ≥ s⁻¹ * (h(P) + [K : ℚ] * r i * log 2) for some s > 0,
then the index of P at a with weights r is at most s. Heights are relative. Replacing
log 2 by a larger constant, such as Pottmeyer's 4, only strengthens the hypothesis.