User avatar
[ TSBT's Pirate ]
[ TSBT's Pirate ]
Posts: 10363
Joined: Thu Oct 04, 2012 1:22 pm
Location: roaming the planet


Post by Alez »


DBN Upper Bound is a Polymath project related to the theory of the Riemann zeta function, based at this Polymath wiki. For the background, theory and discussion on results, please head to the proposal and subsequent threads on Prof. Tao's blog. The computational part of the research is being hosted at this github project.

This Boinc project now plans to scale up the computations, and achieve tighter conditional bounds, using Internet-connected computers. You can contribute to the research by running a free program on your computer.

For a visual understanding with cool graphics and a summary of the results achieved, please check this visual guide created by Rudolph, who has also written the Arb scripts used for the scaled up computations

Types of Computations

There are multiple types of computations that are relevant to the project

1. Storedsums - One-time computation of a Taylor coefficients matrix at a certain height X. This can then be used repeatedly to evaluate approximations to Ht(z) near X using Taylor polynomials. For more details check this Arb or Pari/GP code

The currently available jobs are computing this matrix for X near 2x10^20, which should likely lead to Λ <= 0.11 in subsequent steps (conditional on RH being true upto this X).

2. Establishing a Barrier near height X - Here we actually evaluate Ht(z) approximations along a rectangle contour mesh [(X,X+1),(y0,1)] for small increments in t from [0,t0]. If the winding number turns out to be 0 for each such t, a barrier region is established. The storedsums matrix is used as a helper and significantly speeds up this step.

Till now this step has been quick enough to perform on a standalone machine, but in the future even this could be processed through the grid computing setup. For more details, check this Arb or Pari/GP code

3. Lemmabounds for an N range - Sometimes the analytically derived lower bound for Ht(z) is not positive at Na=sqrt(X/(4π)) and lemmabounds have to be calculated for each N between Na and Nb where the analytical bound becomes positive. We are currently working around this step by carefully choosing X, but for instance the unconditional bound of Λ <= 0.22 required this. For more details, check this Arb or Pari/GP code
The best form of help from above is a sniper on the rooftop....