The now well-established model of hierarchical galaxy formation posits that galaxies grow over time by merging with other galaxies, through major mergers of comparable mass, minor mergers, or satellite accretion. However, the precise role that mergers and interactions play in driving galaxy evolution, and how that role has changed over the age of the universe, has been widely debated. Galaxy mergers can impact galaxies in a number of ways, including their morphology, their sizes, the motion of their stars and gas, their gas content, their star formation rates, and their nuclear activity. This range is due to many factors, including 1) the definition/selection of mergers, 2) the difficulty of selecting a consistent sample across all redshifts, and 3) what the particular science question is. Galaxy mergers are particularly difficult to identify at high redshift since their large distances result in them looking smaller and fainter, and low surface brightness features such as tidal tails become very difficult to identify. The most robust method for identifying galaxy mergers at high redshift is through visual classification, and our group has conducted many studies that look at samples of galaxy mergers selected this way (e.g., Kartaltepe et al. 2010, 2012, 2015; Magagnoli et al. in preparation). While this method is incredibly time-consuming, it is still the most reliable method for identifying galaxies with non-regular morphologies at high redshift since the human eye is adept at pattern recognition. While quantitative methods are less reliable for picking out galaxy mergers (see Kartaltepe et al. 2010, for example) new machine learning techniques show promise (e.g., Peth et al. 2016, Huertas-Company et al. 2009, 2015).
In order to quantify the impact that galaxy mergers have had over the age of the universe, it is important to first quantify the merger rate and how it evolves with redshift. While seemingly a straightforward measurement, this question has been one of much debate since the earliest studies in the 1980s (Zepf & Koo 1989). Today, there is a good handle on how the merger rate evolves out to z~1 (strongly, as (1+z)3, e.g., Kartaltepe et al. 2007) and how selection effects associated with different parent samples and merger identification techniques can impact the results (e.g., Lotz et al. 2011). However, at higher redshift (out to z~3 and beyond) studies have found a perplexing range of results.
Most observational studies find strong evolution with redshift out to z~1 and then a flattening out to z ∼ 3 (e.g., Mantha et al. 2018, Duncan et al. 2019), while modern simulations have found that the merger rate continues to increase out to z~3 (e.g., Rodriguez-Gomez et al. 2015). In order to resolve this discrepancy, one must invoke a change in the pair observability timescale (Snyder et al. 2017, Duncan et al. 2019). This is likely due to a combination of factors, including a change in the observability of pair galaxies with high specific star formation rates and gas fractions and a change in the time that passes between when a pair is observed and the final coalescence of the system, though more work is required to truly understand merger rates at high redshift.