Zulip Chat Archive

Stream: Is there code for X?

Topic: Gamma is strictly log convex


Snir Broshi (Nov 20 2025 at 22:46):

docs#Real.convexOn_log_Gamma and docs#Real.convexOn_Gamma are about convexity.
How can we prove StrictConvexOn rather than ConvexOn? Is there an easy way to get it from the ConvexOn theorem?

The current convexity proof uses Hölder's inequality (docs#MeasureTheory.integral_mul_le_Lp_mul_Lq_of_nonneg) which is a weak inequality, and while Wikipedia does mention some iff condition for equality such a theorem does not exist in Mathlib.

Wikipedia suggests showing ψ(1)(x)>0\psi^{(1)}(x) > 0 since the polygamma function has a series representation where all terms are positive, which then implies Γ(x)Γ(x)>Γ(x)2\Gamma''(x)\Gamma(x) > \Gamma'(x)^2, which implies strict-log-convexity.


Last updated: Dec 20 2025 at 21:32 UTC