Cosmic Microwave Background Radiation Data Analysis

Principal Investigator 

George Smoot, 
Lawrence Berkeley National Laboratory 


Julian Borrill, 
Lawrence Berkeley National Laboratory 

Andrew Jaffe, 
Center for Particle Astrophysics, UC Berkeley

A Map Of The Imprint Of The Universe's  
Primordial Perturbations In The Cosmic  
Microwave Background Radiation 

Research Objectives

To develop the novel computational techniques necessary to extract fundamental cosmological parameters from forthcoming CMB datasets.

Computational Approach

Central to our task is being able to locate and describe the maximum of the likelihood function of the cosmological parameters given the data - here an N-pixel map generated by any one of the forthcoming CMB observations. At present N is at most a few thousand, but this will increase to tens and hundreds of thousands with the MAXIMA and BOOMERANG ballon flights, and to about a million with the MAP and PLANCK satellite missions. Our present approach scales as N-squared in size and N-cubed in time so a supercomputer like the T3E is essential to our work. However, even this will be unable to handle the largest datasets; we must therefore develop alternative algorithms either with better scaling properties, or that repeatedly analyse only a subset of the overall dataset at any time.


Our project received funding in September 1997. In the two months since then we have completed the development of two generations of the quadratic estimator formalism for performing a rapid search of the parameter space for the maximum of the likelihood function, and have implemented them on the T3E both in serial and in parallel (using the LAPACK and ScaLAPACK libraries respectively). We are currently working to demonstrate the ability of the parallel code to analyse at least the first generation MAXIMA/BOOMERANG data. To this end we are now developing simulated datasets, comprising the signal from a known underlying theory and a model of the instrument noise and sky foregrounds associated with each experiment.


The Cosmic Microwave Background provides a picture of the universe as it was a mere 100,000 years after the Big Bang. As the earliest possible photon image available to us it is our most powerful discriminant between different cosmological models. The unprecedently detailed CMB datasets obtained in the next 10 years will allow us to determine the fundamental cosmological parameters - in many cases currently known to no better than a factor of 2 - to the 1% level.


  1. `The Cosmic Microwave Background Anisotropy Experiments', G.Smoot, (astro-ph/9705135)
  2. `Estimating The Power Spectrum Of The Cosmic Microwave Background', J.R.Bond, A.H.Jaffe and L.Knox, (astro-ph/9708203)
  3. `Power Spectrum Estimators For Large CMB Datasets', Julian Borrill, (astro-ph/9712121)

Project-Related Web Sites

  1. Professor Smoot's CMB Research Group
  2. Professor Smoot's Home Page
  3. The COMBAT Project

Home | What's New | Events | About NERSC | Hardware | Software
Visualization | Research | Accounts | Training | Getting Started | Help | Search