SUSY scans and dark matter Pat Scott Oskar Klein Centre for Cosmoparticle Physics & Department of Physics, Stockholm University May 27 2010 Mostly based on: JCAP 1001:031 (2010; arxiv:0909.3300) JHEP 1004:057 (2010; arxiv:0910.3950)
Supersymmetry (SUSY) Extension of the standard model to include superpartners of all known particles Superpartners differ only in mass and spin from SM particles With conservation of R-parity, lightest superparticle (LSP) is stable => if neutral, LSP is a WIMP Pat Scott, May 27 2010 Nordic Astrophysics 2010, Visby 2
Supersymmetry (SUSY) Extension of the standard model to include superpartners of all known particles Superpartners differ only in mass and spin from SM particles With conservation of R-parity, lightest superparticle (LSP) is stable => if neutral, LSP is a WIMP A (relatively) simple implementation is the Constrained Minimal Supersymmetric SM (CMSSM) 4 parameters (m 0, m 1/2, tanβ, A 0 ) + 1 sign (sgnµ) Pat Scott, May 27 2010 Nordic Astrophysics 2010, Visby 3
SUSY Scanning Goal: given a particular version of supersymmetry, determine which parameter combinations fit all experiments, and how well Issue 1: Combining fits to different experiments Easy composite likelihood (allows marginalisation of errors/nuisances and full statistical inference) L 1 x L 2 = χ 1 2 + χ 2 2 Issue 2: Finding the points with the best likelihoods Tough grid scans, MCMCs, nested sampling or genetic algorithms Public codes: SuperBayeS, SFitter, Fittino Pat Scott, May 27 2010 Nordic Astrophysics 2010, Visby 4
SUSY Scanning Constraints Issue 1: Combining fits to different experiments Precision electroweak tests at LEP (particle) LEP limits on sparticle masses (particle) B-factory data (rare decays, b --> sg) (particle) Muon anomalous magnetic moment (particle) Dark matter relic density from WMAP (astro/cosmo) Want to include more astro/cosmo observables Various indirect detection targets, observations (Fermi, HESS, etc) Direct detection limits Neutrino telescopes Pat Scott, May 27 2010 Nordic Astrophysics 2010, Visby 5
Indirect detection & Segue 1 Nearby ultra-faint dwarf spheroidal galaxy d = 23kpc, M/L ~ 1300 Provides strongest constraint on DM annihilation cross-section Pat Scott, May 27 2010 Nordic Astrophysics 2010, Visby 6
SUSY Scanning - Algorithms Issue 2: Finding the points with the best likelihoods SUSY parameters map very non-linearly to observables => Likelihood function is a messy one Grid and random scans do not cut it 'Bayesian' techniques: MCMCs, nested sampling 'Frequentist' techniques (optimised for profile likelihood): genetic algorithms, others?? Most previous work (e.g. Gondolo & Baltz, Allanach et al, Ellis et al, Trotta-Ruiz-et al) used Bayesian scanning to do Bayesian inference Interest has been growing in quasi-frequentist analyses but most people still use Bayesian scanning techniques to perform them Pat Scott, May 27 2010 Nordic Astrophysics 2010, Visby 7
Genetic Algorithms (GAs) Evolutionary algorithms based on natural selection Individuals (points in the parameter space) are selected and cross-bred to create offspring (new points) Selection and breeding occur according to ranking by a fitness function (the likelihood in our case) Do not use the likelihood gradient => good for messy parameters spaces, with e.g. holes, spikes, etc. Scale better than MCMCs/nested sampling with dimensionality Highly optimised for finding maxima (frequentist) rather than mapping the full likelihood surface/integral (Bayesian) Pat Scott, May 27 2010 Nordic Astrophysics 2010, Visby 8
Indirect detection with GAs GAs find much better fits than traditional (Bayesian) methods like nested sampling (χ 2 = 9.35 vs. χ 2 = 13.51). Previously-missed region found at m χ ~ 400 GeV Pat Scott, May 27 2010 Nordic Astrophysics 2010, Visby 9 (Akrami, PS, Edsjö, Conrad, Bergström, JHEP 2010)
Direct detection with GAs Best-fit point is actually ruled out by direct detection (under standard halo assumptions). Secondary maximum is still OK Pat Scott, May 27 2010 Nordic Astrophysics 2010, Visby 10
Summary SUSY scans allow one to Consistently test which SUSY parameters are allowed or disallowed, and at what confidence level Make complete and probabilistic predictions for the properties of dark matter in a given SUSY setup First applications to indirect detection give weak constraints, but technique is definitely promising Genetic algorithms can help in tackling some of the more thorny technical issues Pat Scott, May 27 2010 Nordic Astrophysics 2010, Visby 11