A Radio Measurement of the Star Formation History of the Universe

Most of the stars in the universe were formed by disk galaxies like our own Milky Way during an era poetically called “cosmic noon.” This era, occurring approximately 3 billion years after the Big Bang, marked the peak of not only star formation, but also black hole growth and dust attenuation. The dust that permeates all galaxies absorbs and scatters the ultraviolet and optical light primarily generated by massive stars, whose lives are so short that they provide an effectively instantaneous measurement of the star formation rate. Understanding the formation and evolution of galaxies at cosmic noon is essential to understanding how the universe appears and acts today---but the uncertainties imposed by dust are worst during this influential period. Radio emission is entirely unaffected by dust and is generated by supernova remnants of the same massive stars emitting primarily in the UV and optical. Radio observations of star-forming galaxies are therefore a powerful tracer of star formation rate, but it has not been until the past decade that radio observations have been sensitive enough to detect Milky Way-like galaxies at cosmic noon. In this dissertation, I calculate and combine the source counts from the deepest radio continuum image to date with the local luminosity function of radio sources to model the star formation history of the universe.

I determined the local luminosity functions for both star-forming galaxies (SFGs) and active galactic nuclei (AGNs) using N ~ 10,000 radio sources from the NRAO VLA Sky Survey (NVSS) cross-identified with galaxies in the 2MASS Extended (2MASX) catalog (Chapter 2). The AGNs and SFGs were separated using only radio and infrared data rather than optical emission-line diagnostics, which are not good quantitative measures of AGN-powered radio emission. Our sample of radio sources account for >99% of the total 1.4 GHz energy density in the nearby universe. The local radio-derived star formation rate density (SFRD) value of 0.0128 solar masses per year per cubic megaparsec is consistent with previous models for the SFRD derived using ultraviolet and infrared data.

In Chapter 3, I present radio source counts across eight decades of flux density spanning 0.25 microJy < S < 25 Jy determined from the deepest 1.4 GHz radio continuum image taken by the MeerKAT radio interferometer in South Africa and the archival NVSS component catalog. With an rms noise of 0.56 microJy/beam and a resolution of 7.6’’, the MeerKAT DEEP2 image is confusion-limited, so below S=10 microJy I calculated the source counts statistically from the confusion brightness distribution P(D). Above S=10 microJy the source counts were measured directly from the DEEP2 image and the NVSS component catalog.

The evolving energy-density function is the comoving energy density of radiation produced by sources at redshift z having spectral luminosity L_\nu at frequency \nu. Simple equations relate the brightness-weighted differential source counts S^2n(S) with the evolving energy-density function integrated over all redshifts (Appendix C). Using a combination of luminosity and density evolution, I developed evolutionary models (Chapter 4) for SFGs and AGNs that accurately predict the observed source counts given the local luminosity functions. Through the FIR/radio correlation, the product of luminosity and density evolution of radio sources is directly related to the total SFRD evolution, describing how many stars (by mass) were formed per year per comoving cubic megaparsec. The radio-derived model for SFRD evolution is similar to previous models based on UV/IR data, but predicts stronger star-formation evolution. Chapter 5 reviews the main conclusions of this dissertation and discusses future work.