I am broadly interested in using large-scale structure (LSS) observations to constrain cosmology. Over the course of my previous research, I had the opportunity to explore a large range of topics and methods in LSS research, from analytic calculations of systematic effects required for the interpretation of data from upcoming LSS surveys, analysis of numerical simulations, to data processing and clustering measurements.
For the full range of my previous research, please have a look at my publications .
After finishing my PhD last year I joined the Dark Energy Survey, one of the largest ongoing LSS surveys. Within DES I mainly work in the Combined Probes Working Group, coordinating the effort on combining weak lensing and clustering measurements. I also coordinate the higher-order statistics subgroup and the theory code comparison task force.
One of the main projects in my personal research is to develop CosmoLike, a likelihood analysis framework for combining different LSS probes consistently.
CosmoLike: Combining Cosmological Probes Consistently
Current and upcoming galaxy redshift surveys aim to determine the composition and initial conditions of the Universe and the nature of gravity and dark energy at unprecedented accuracy by measuring the positions and shapes of billions of galaxies. The different LSS observables, such as galaxy clustering, weak gravitational lensing, and the abundance of galaxy clusters, probe different aspects of
cosmic structure formation, all being sensitive to the late time/dark energy dominated phase of the Universe.
The tightest constraints on cosmological parameters will come from a joint analysis of these probes. If, however, measured from the same survey (or overlapping surveys), these LSS observables trace the same underlying density field. Hence, in contrast to combined constraints from BAO+CMB+supernovae, a multiplication of the individual posterior probability distributions is insufficient for combining LSS constraints. Instead, consistent forward modeling including all cross-correlations is necessary to correctly extract the cosmological information.
My current research focusses on 1) extracting the encoded cosmological information from these LSS probes jointly and consistently, and 2) combining this information with cosmological constraints from the CMB and supernovae. The first task is particularly challenging: due to the increase in survey volume, the interpretation of these measurements will be no longer be limited by statistical uncertainty, but by systematic errors that are correlated amongst probes (e.g., photometric redshifts, galaxy bias). The interpretation of such data requires a new generation of analysis methods, which need to incorporate cross-correlations amongst probes as well as detailed modeling of systematics.
For this purpose, we have developed CosmoLike, which is one of the most evolved software package for combined probes likelihood analyses. Currently, the code is only available to the DES collaboration; a public release is planned for early 2014. Features of the current version include
- Predictions for the observed cluster abundance and various tomographic two-point statistics of cosmic shear, galaxy-galaxy lensing, galaxy clustering, cluster clustering, cluster lensing, galaxy-CMB cross-correlations. These are modeled using either fitting functions for the non-linear matter power spectrum, or a state-of-the-art halo model + HOD code.
- Non-Gaussian covariances of these observables, including all cross-covariances
- Models for systematic uncertainties associated with shear calibration, photometric redshift estimates, intrinsic alignments, impact of baryons on the matter power spectrum, cluster finding and mass-observable relation (accounting consistently for correlated errors among different probes)
- Sub-second run-time per model evaluation (achieved through parallelized computation of fine-tuned look-up tables), which enables sampling of high dimensional parameter spaces
- Parallel MCMC sampling
- Extensions to include supernova data sets and the Planck cosmology likelihood are currently under development.
Weak Lensing with the Tully-Fisher Relation
We describe a new technique for reducing the shape noise in weak lensing measurements by an order of magnitude. Our method relies on spectroscopic measurements of disk galaxy rotation and makes use of the Tully-Fisher relation in order to control for the intrinsic orientations of galaxy disks. Using our new method, the shape noise ceases to be an important source of statistical error.
We produce CosmoLike forecasts for two spectroscopic weak lensing survey concepts; the first is a near-future design, intended to be within the capabilities of the next generation of large spectroscopic survey instruments, while the second represents a more ambitious medium-term survey. For comparison, we also forecast the performance two Large Synoptic Survey Telescope (LSST) cosmic shear analyses considering optimistic and conservative assumptions about LSST systematic errors. We find that Tully-Fisher weak lensing experiments are highly competitive with constraints from LSST alone, while evading the most important sources of theoretical and observational systematic error inherent in traditional weak lensing techniques.
The contours on the right show results of simulated likelihood analyses with CosmoLike. We show the 95% confidence regions for the TF-lensing based Stage III survey (black, solid; assuming an imaging survey + optical spectroscopy follow-up with ngal = 1.1/square arcminute over 5,000 square degree), a TF-Stage IV survey (red, dashed; assum-ing an imaging survey + optical and infrared spectroscopy follow-up with ngal = 1.1/square arcminute over 15,000 square degree) in comparison with the conventional cosmic shear analysis from LSST using an optimistic (green, dotted) and pessimistic scenario (blue, dotted-dashed) for the accuracy of LSST photometric redshifts and shear calibrations. We marginalize over shear calibration and (for LSST only) photometric redshift systematic errors.
See arXiv:1311.1489 for details.
Clustering Analysis of Green Valley Galaxies
Green valley galaxies are characterized by their intermediate colors between blue and red galaxies. They are strong candidates for a population of transitional galaxies which are currently shutting off star formation.
We study the clustering properties of near ultraviolet (NUV) - optical color selected luminosity bin samples of green valley galaxies. These galaxy samples are constructed by matching the Sloan Digital Sky Survey Data Release 7 with the latest Galaxy Evolution Explorer source catalog which provides NUV photometry. We present cross-correlation function measurements and determine the halo occupation distribution of these transitional galaxies using a new multiple tracer analysis technique.
Our analysis extends the halo-occupation formalism to model the cross-correlation function between a galaxy sample of interest and multiple tracer populations simultaneously. This method can be applied to commonly used luminosity threshold samples as well as to color and luminosity bin selected galaxy samples, and improves the accuracy of clustering analyses for sparse galaxy populations.
We confirm the previously observed trend that red galaxies reside in more massive halos and are more likely to be satellite galaxies than average galaxies of similar luminosity. While the change in central galaxy host mass as a function of color is only weakly constrained, the satellite fraction and characteristic halo masses of green satellite galaxies are found to be intermediate between those of blue and red satellite galaxies.
Details of the analysis can be found here.
The Gravitational Lensing Signal of Stacked Voids As the dynamics of underdense regions is dominated by dark energy at earlier times than the rest of the universe, the potential of voids to probe the nature of dark energy has been noted recently. We demonstrate that surveys like LSST and Euclid will be able to measure the gravitational lensing signal from stacking abundant medium-sized voids, thus providing direct constraints on the matter density profile of voids independent of assumptions on galaxy bias.
Shear signal of stacked voids in the redshift range 0.4 < zV < 0.6 for a Stage III as detailed in the paper.
Tidal Alignments as a Contaminant of the Galaxy Bispectrum We consider the effect of tidal alignments in combination with an orientation-dependent galaxy selection on the galaxy bispectrum using two toy-models of intrinsic galaxy alignments. We show that in the linear alignment model, intrinsic alignments result in an error in the galaxy bias parameters, but do not affect the inferred value of σ8 . In contrast, the quadratic alignment model results in a systematic error in both the bias parameters and σ8. However, the quadratic alignment effect has a unique configuration dependence that should enable it to be removed in upcoming surveys.
Weak Lensing Power Spectra for Precision Cosmology The ellipticity power spectrum is the central quantity for constraining cosmology with second order cosmic shear statistics. It is usually assumed that it can be expressed as an integral over the underlying matter power spectrum. This is true at second order in the gravitational potential. We extend the standard calculation, constructing all corrections to fourth order, accounting for multiple deflections along the light path, reduced shear, and magnification bias.