Lights All Askew: Systematics in Galaxy Images from Megaparsecs to Microns

The stars and galaxies are not where they seem. In the process of imaging and measurement,
the light from distant objects is distorted, blurred, and skewed by several physical effects on scales
from megaparsecs to microns. Charge-coupled devices (CCDs) provide sensitive detection of this
light, but introduce their own problems in the form of systematic biases. Images of these stars
and galaxies are formed in CCDs when incoming light generates photoelectrons which are then
collected in a pixel’s potential well and measured as signal. However, these signal electrons can be
diverted from purely parallel paths toward the pixel wells by transverse fields sourced by structural
elements of the CCD, accidental imperfections in fabrication, or dynamic electric fields induced by
other collected charges. These charge transport anomalies lead to measurable systematic errors in
the images which bias cosmological inferences based on them. The physics of imaging therefore
deserves thorough investigation, which is performed in the laboratory using a unique optical beam
simulator and in computer simulations of charge transport.
On top of detector systematics, there are often biases in the mathematical analysis of pixelized
images; in particular, the location, shape, and orientation of stars and galaxies. Using elliptical
Gaussians as a toy model for galaxies, it is demonstrated how small biases in the computed image
moments lead to observable orientation patterns in modern survey data. Also presented are exam-
ples of the reduction of data and fitting of optical aberrations of images in the lab and on the sky
which are modeled by physically or mathematically-motivated methods.
Finally, end-to-end analysis of the weak gravitational lensing signal is presented using deep sky
data as well as in N-body simulations. It is demonstrated how measured weak lens shear can be
transformed by signal matched filters which aid in the detection of mass overdensities and separate
signal from noise. A commonly-used decomposition of shear into two components, E- and B-modes,
is thoroughly tested and both modes are shown to be useful in the detection of large scale structure.
We find several astrophysical sources of B-mode and explain their apparent origin. The methods
presented therefore offer an optimal way to filter weak gravitational shear into maps of large scale
structure through the process of cosmic mass cartography.