Bayesian selection of sign μ within mSUGRA in global fits including WMAP5 results

Farhan Feroz, B. C. Allanach, M. P. Hobson, Shehu AbdusSalam, Roberto Trotta, Arne WeberJournal of High Energy Physics 2008 · DOI

TL;DR

This paper uses a Bayesian approach to determine the preferred sign of the Higgsino mass parameter ($\mu$) in the Minimal Supersymmetric Standard Model (mSUGRA), incorporating WMAP5 data. They find weak to moderate evidence for $\mu > 0$, with the anomalous magnetic moment of the muon ($g-2)_\mu$ being the dominant factor driving this preference.

What Problem Does This Paper Address?

The Minimal Supersymmetric Standard Model (MSSM) is a popular extension to the Standard Model of particle physics, aiming to solve issues like the hierarchy problem and provide a dark matter candidate. A simplified version, mSUGRA, has fewer parameters but still has a significant parameter space. One crucial parameter is $\mu$, which determines the sign of the Higgsino mass. This sign affects various physical observables, and distinguishing between the $\mu > 0$ and $\mu < 0$ scenarios is important for pinning down the correct model. Previous studies have faced challenges due to the high dimensionality and degeneracies in the parameter space, making robust model comparison difficult. This paper aims to address this by using a sophisticated Bayesian inference technique to perform a global fit to existing experimental data, including the latest WMAP5 cosmological data, and quantify the evidence for each sign of $\mu$.

Methods

The core of this paper lies in its application of Bayesian inference for model selection. Specifically, they employ the MULTINEST algorithm, a nested sampling technique designed to efficiently handle highly degenerate and multi-modal posterior distributions. This is crucial for exploring the complex parameter space of mSUGRA. MULTINEST allows for the calculation of Bayesian evidence, which is the key quantity for comparing different models (in this case, mSUGRA with $\mu > 0$ versus mSUGRA with $\mu < 0$). They also use this framework to derive posterior probability distributions for various parameters and observables, providing updated constraints. The paper also explores the impact of different prior choices on the results, a critical aspect of Bayesian analysis.

Datasets

The paper utilizes a comprehensive set of experimental data, including:

  • WMAP5 data: Specifically, the relic density of cold dark matter ($\Omega_{DM}h^2$).
  • B-physics observables: Including the branching ratio of $b \to s\gamma$ and the isospin asymmetry of $B \to K^*\gamma$.
  • Anomalous magnetic moment of the muon ($g-2)_\mu$: A key observable that shows a tension with the Standard Model prediction.
  • Electroweak precision measurements: Such as the W boson pole mass ($M_W$) and the effective leptonic mixing angle ($\sin^2\theta_W$).
  • LEP Higgs search likelihood: Used to constrain the Higgs mass.

The paper does not introduce new datasets but rather combines and updates existing ones with the latest WMAP5 results.

Key Insights and Ideas

  • The $(g-2)_\mu$ constraint is king: The anomalous magnetic moment of the muon is the most significant factor driving the preference for $\mu > 0$. Without it, the evidence for either sign is roughly equal.
  • Bayesian evidence is powerful for model selection: The paper demonstrates how Bayesian evidence, calculated via MULTINEST, provides a robust way to compare competing models, even in complex, high-dimensional parameter spaces.
  • Prior choice matters, but the data is starting to speak: While the choice of prior range and measure can influence the strength of the evidence, the results show a consistent weak to moderate preference for $\mu > 0$, suggesting the data is beginning to constrain the model.
  • Consistency checks are crucial: The paper highlights the importance of checking the consistency between different experimental constraints, showing how Bayesian methods can quantify tensions between datasets.

Key Results

  • Weak to moderate evidence for $\mu > 0$: The ratio of posterior probabilities $P(\mu > 0) / P(\mu < 0)$ ranges from 6 to 61, depending on the prior choice. This indicates a preference for the $\mu > 0$ scenario.
  • $(g-2)_\mu$ dominates model selection: Removing the $(g-2)_\mu$ constraint makes the evidence for $\mu > 0$ and $\mu < 0$ nearly equal, underscoring its importance.
  • Updated parameter constraints: The paper provides updated posterior probability distributions for mSUGRA parameters, showing qualitative agreement with previous studies but with improved precision due to the new data and methods.
  • Consistency between key observables: The analysis shows no statistically significant inconsistency between the $(g-2)_\mu$ and $BR(b \to s\gamma)$ constraints when considering the $\mu > 0$ branch, especially when other observables are included.

Limitations and Future Work

  • Prior dependence: The authors acknowledge that the evidence ratio is still somewhat dependent on the choice of prior measure and range. This indicates that the current data, while informative, is not yet powerful enough to unambiguously distinguish between the models.
  • Statistical uncertainty: While MULTINEST provides robust evidence calculations, the statistical uncertainty in the evidence ratio is still relatively small compared to the prior dependence.
  • Future work: The authors suggest that future high-quality data from the Large Hadron Collider (LHC) will be crucial for casting more light on this issue and providing a more definitive answer. The developed Bayesian framework and MULTINEST technique are well-suited for analyzing such future data.

Summary updated

BibTeX
@article{feroz2008bayesian,
  author = {Farhan Feroz and B. C. Allanach and M. P. Hobson and Shehu AbdusSalam and Roberto Trotta and Arne Weber},
  title = {{Bayesian selection of sign μ within mSUGRA in global fits including WMAP5 results}},
  journal = {Journal of High Energy Physics},
  year = {2008},
  doi = {10.1088/1126-6708/2008/10/064},
}
Figure