Computational Physics

The search for resonances and hadron spectroscopy is a computationaly intense endeavour. The data reduction of current data requires working with data sets that are 10's of terrabytes in size and development of algorithems to both maniipulat and compress these files are cruicial to the success of the process. In terms of analysis, the partial wave analyses that are carried out require computer resources to both generate large Monte Carlo data sets as well as analze the data to extract resonsnce signatures.

Over the past several years, we have developed a large suite of tools at CMU that facilitate this analysis. These include analysis tools for manipulating data from the CLAS experiment at Jefferson Lab--- the COBRA tools which faciliate easy access and compressed storage to CLAS data.

A tool for carrying out quantum field theory calculations has also been developed: qft++ and an article has been written describing this work: Numerical Object Oriented Quantum Field Theory Calculations. This tool set allows us to compute the needed physics amplitudes for our partial wave analysis using a covariant tensor formalism.

We have also developed a background subtraction method that is appropriat for inclusion in unbinned maximum likelihood fits. Separating Signals from Non-Interfering Backgrounds using Probabilistic Event Weightings The precedure is a generalization of sideband subtraction to high-dimension data sets by defining a set of nearest neighbors for each event. It allows each event to be assigned a quality factor that indicates the probability the event is signal or background.

Our work is carried out on a 286-core processor farm which we share with our lattice QCD colleagues at CMU. The farm has about 20 terrabytes of data storage and is built primarily with dual-quad-core AMD chips in 32 blade servers connected using a private channel-bonded gigabit network.

We are currently funded by an NSF Physics at the information Frontier (PIF) grant to develop grid-enabled partial wave analysis tools. This latter effort is collaborative bewteen Carnegie Mellon, University of Connecticut and Indiana University.

PWA Wiki
qft++ Code

© 2015 Curtis A. Meyer