Observational Cosmology

High Redshift Research at Oxford

The observational cosmology programme at Oxford is primarily concerned with detection and measurement of dark energy and dark matter through low redshift (z < 3) cosmological surveys, accounting for astrophysical and measurement systematic effects and errors. Independent surveys and techniques have verified the presence of dark energy, but a physical understanding remains distant: there is no compelling explanation for dark energy’s existence or its magnitude (Peacock & Schneider, 2006; Frieman et al., 2008). The simplest possibility is a non-zero unchanging “cosmological constant”, Λ, described by an equation of state whose pressure and density are related via p = wρ, where w = −1. Dark energy might also arise dynamically from a slowly rolling scalar ﬁeld, where w ̸= −1 and possibly evolves with time, or general relativity may break down on cosmological scales.

Using a combination of techniques, the latest results show that ⟨w⟩ is consistent with −1 with a sub-5% statistical precision (Fig. 3). This represents major progress since the discovery of dark energy a decade ago: similar advances over the coming decade require substantial investment now to assemble the large astrophysical datasets required to make the next generation measurement.

• As standardised candles, Type Ia Supernovae (SNe Ia) are a mature probe of dark energy. Our work over the last 5 years has used SNe Ia to make the most precise measurement yet of ⟨w⟩. Systematics are now becoming dominant; most are caused by the inadequate low-redshift SN sample. Oxford has joined the new, and fully operational, Palomar Transient Factory (PTF) survey to construct the new low-redshift SN Ia dataset required to fully exploit existing and future SN Ia data, and to measure the variation of dark energy with redshift,w(z).

• Weak lensing has long been promoted as a direct way of measuring dark matter structure through its gravitational effect. 3D lensing in particular offers the possibility of measuring the growth in cosmic structure to test cosmological models, to map out the large-scale topology of structure and, in combination with other measures, to measure w(z). Oxford is taking a leading role in eliminating systematics in a new generation of lensing measurements.

•Baryon acoustic oscillations (BAO) in the matter power spectrum measured by galaxy redshift surveys promise a direct way of measuring the geometry of the universe. The challenge is to make large enough redshift surveys at the optimum redshifts for BAO detection and measurement. Oxford is a joint leader of the fastsound consortium proposing a large BAO survey at 1<z<2 using FMOS on Subaru.

Type Ia Supernovae and dark energy

Type Ia supernovae (SNe Ia) are exceptionally luminous and make excellent “standardised candles” for measuring the redshift-distance relation. The uniformity of their nuclear fuel results in a small dispersion in their peak brightnesses, further improved empirically: brighter SNe Ia have wider and bluer light-curves than their fainter counterparts. These provide distances precise to ∼7% (Conley et al., 2008), used to constrain cosmological models and measure w. The 5-year CFHT Supernova Legacy Survey (SNLS; Astier et al., 2006) discovered and conﬁrmed ≃450 SNe Ia to z ∼ 1. This survey has been highly productive, with 15 refereed publications since 2006. Oxford has played a key role, including leading the Gemini spectroscopy follow-up programme (PI: Hook) and the third-year cosmological analysis (Sullivan). Using these SN Ia data together with BAO and CMB probes (Eisenstein et al., 2005; Dunkley et al., 2009), we have made the most precise measurement of w (Fig. 3 - above; Sullivan et al. 2009, in prep.). The new result shows that dark energy acts like a vacuum energy (the cosmological constant) at a sub-5% statistical precision, or 7% including systematic effects.

Though in principle the SN Ia technique is limited only by the calibrations of the SN luminosities, in reality systematics complicate the analysis (Fig. 3 - above). One systematic arises from varying matter distributions along the line of sight to high-z SNe: magnification and de-magnification generates scatter in peak brightnesses (Jönsson et al., 2008). Our current PDRA, J. Jönsson, has shown that although the bias arising from gravitational lensing of individual SNe Ia is small for SNLS, correction for lensing can reduce the scatter in peak magnitudes by 6%, shrinking the error ellipses by 4 − 8% (Jönsson et al., 2009). The lensing effect on distant SNe can even be used to constrain the properties of dark matter haloes along the line of sight (Jönsson et al. in prep). JJ is now applying lensing corrections directly to SNLS data, the first such correction in cosmological SN analyses.

Additional systematics might rise from our poor knowledge of SN Ia progenitor physics: dependencies on metallicity or age could manifest as redshift evolution (Sullivan et al., 2006; Howell et al., 2007; Sullivan et al., 2009). This “population drift” is empirically calibratable to the accuracy required by current analyses. However, from a theoretical stand-point metallicity is an important parameter that affects the SN Ia light curve shape (Podsiadlowski et al., 2006, 2008). Understanding this at the detail required for the next level of precision will improve the utility of SNe Ia as standard candles and is a key driver for new low-z surveys, where exquisite data on many events can be assembled. One possibility is the use of physical information encoded in the SN spectra. Our former graduate student, Justin Bronder, demonstrated a correlation between Si II strength and SN luminosity from two years of SNLS/Gemini spectra: empirical corrections based on Si perform as well as traditional light-curve shape corrections (Bronder et al., 2008). This work has been expanded to the full SNLS sample in the thesis of current student Emma Walker.

Weak lensing surveys, measurement and analysis

In contrast to SN Ia cosmology, where astrophysical systematics are becoming the dominant uncertainty, weak lensing studies are still dominated by measurement systematics. Traditional shape measurement methods, such as KSB (Kaiser et al., 1995), work well for Gaussian-shaped galaxies and PSFs, but in reality PSFs are exceedingly non-Gaussian, varying both temporally and spatially, and galaxies are poorly described by Gaussians. To remedy this, in 2000 Miller started working with graduate student Catherine Heymans on a Bayesian model-fitting approach. The resulting algorithm, lensﬁt, (Miller et al., 2007; Kitching et al., 2008) uses a full Bayesian methodology that recovers the correct galaxy shape in the presence of any PSF, and crucially enables the sensitivity of the measurement to be quantified: faint galaxies have increasingly diminishing sensitivity to lensing shear as their signal-to-noise decreases, an effect calibrated entirely within lensﬁt. The technique is currently the leading method of those that can be applied to real data (including KSB and “shapelets”) in the “Great08” challenge. We are currently applying lensﬁt to the CFHT Legacy Survey (Wide: CFHTLSW) as part of the CFHTLS Archive Research Survey (CARS) collaboration (Erben et al., 2009). The “real data” version of lensﬁt copes with chip-to-chip and temporal PSF variations, optimum combination of likelihoods from dithered exposures, astrometric shifts between dithered exposures, and non- cosmological sources of shear (camera distortion and atmospheric refraction). The PSF is measured from stars in the fields being analysed, and the requisite prior distributions in size, magnitude and ellipticity are obtained from the CFHTLS Deep survey data. We will produce the first catalogue containing accurate galaxy shapes and broad-band photometric redshifts (photo-z) in June 2009, based on measurements of 10 galaxies in 170 deg to iAB < 24.5. This will be the largest and most accurate weak-lensing survey to date.

The key to obtaining unbiased shape measurement and to quantifying the sensitivity is the Bayesian nature of the measurement. We have also demonstrated that a full implementation of Bayesian measurement corrects for bias in photo-z estimates that would otherwise contaminate 3D lensing signals (Edmondson et al., 2006). CARS uses the Benítez (2000, BPZ) pseudo-bayesian approach which still outputs a single best-guess redshift value rather than a Bayesian probability distribution Our aim now is to combine Bayesian shape measurement with fully Bayesian photo-z measurement in collaboration with Heavens & Kitching (Edinburgh), to obtain an unbiased measure of 3D lensing signals. The techniques can be applied immediately to CARS and GEMS data and in the future to DES and Pan-STARRS data.

Baryon Acoustic Oscillation Surveys

The Subaru telescope’s FMOS facility (Dalton et al., 2008; Iwamuro et al., 2008) provides an excellent opportunity to measure baryon acoustic oscillations (BAO) from a sample at 1 < z < 2 (Glazebrook & Blake, 2005; Dalton, 2009). The Subaru Telescope SAC are currently considering a large-scale FMOS survey (fastsound, Totani, Dalton, Matsubara & Glazebrook) to measure redshifts for ∼ 6 × 10 galaxies over 300 deg in this redshift range, commencing early 2011 and observing for 3–5 years. These observations will build heavily on the FMOS GTO observations.

Observational Cosmology