0:005:05

Recursos

Resumen

Take the determinant proof that Catalan's constant G is irrational (Theorem 1.1) and explain why it works. Assume G is rational, then show that two bounds on the same determinant cannot both hold.

Why the usual route failed

  • The criterion [0:15]: if G = a/b, every nonzero integer form AG − B is at least 1/b in size. So a proof needs nonzero integer forms that tend to zero.
  • Example and non-example: Apéry's forms for ζ(2) and ζ(3) keep their decay after the denominators are cleared. Zudilin's Catalan approximations converge fast, but their integer forms do not tend to zero.
  • Family results prove that some even beta value is irrational, but they never single out β(2) = G.

The determinant argument

  • Rationality [1:14]: Δ_N has size 48N, and its entries are moments that combine 1, G and ζ(2). Chebyshev rows with high Taylor contact cancel ζ(2). So if G ∈ ℚ, Δ_N ∈ ℚ.
  • Floor [1:50]: for a nonzero rational, log|Δ| = Σ v_p(Δ) log p, so bounding the denominators bounds |Δ| from below. Three things produce the bound:
    • an estimate at the prime 2;
    • two denominator layers (p², p) at the odd primes;
    • the prime number theorem, which adds up the losses.
    Nonvanishing comes from arithmetic, not positivity: for N = p prime, Frobenius reduces the matrix to three fixed matrices, and an exact certificate shows they are invertible. The result is liminf 𝓛_N > −2.29084.
  • Toy check [2:56]: suppose x is a nonzero rational whose denominator divides 72, so |x| ≥ 1/72. If an integral shows |x| < 1/100, there is no room left for x.
  • Ceiling [3:23]: with no hypothesis at all, Δ_N is an integral of Vandermonde products with logarithmic interactions. A Chebyshev expansion and certified rational trial coefficients give limsup 𝓛_N ≤ −2.290939875.
  • Collision [3:57]: along the same sequence of primes, the floor sits about 10⁻⁴ above the ceiling, so G is irrational. This gives no irrationality measure.

Reuse it [4:37]

When you read another determinant irrationality proof, look for four parts:

  1. rationality under the hypothesis
  2. a nonzero determinant
  3. a denominator floor
  4. an integral ceiling

All four live on the n² scale. Then check that the floor beats the ceiling.

Why Catalan's constant G is irrational

OpenAI Math Preprints

Clase 10 / 100

Ver clases