DETECT: A Pipeline to Quantify Detection Thresholds in Rubin for Nearby Targets Embedded in Bright Host Galaxies

Abstract

The final stages of stellar evolution can be constrained by studying pre-SN variability. The incredible amount of data coming from the upcoming Rubin Legacy Survey of Space and Time (LSST) will be fundamental to this type of work. However, robustly measuring pre-SN variability can be hard, as even state-of-the-art image subtraction pipelines struggle when the target is embedded in a bright nearby galaxy. We developed Detection Efficiency and Threshold Estimation for Characterization of Transients (DETECT) to tackle this problem. It performs a series of source injection, image subtraction, and forced photometry to obtain reliable detection thresholds tailored to a specific location within a given host galaxy. We first validate the pipeline using simulated data from Rubin DP0 and then apply it to a sample of 15 targets found in Rubin DP1. We demonstrate that DETECT is capable of identifying pre-SN variability while calculating reliable upper limits and suppressing false positives for targets embedded in bright host galaxies. Most of the false positives in this work occurred when the signal-to-noise ratio (SNR) was between 5 and 10, while no false positives were found when the SNR was greater than 10. Finally, even though DETECT was originally developed in the context of pre-SN variability, it is broadly applicable to any situation where detections are uncertain and robust upper limits are needed.

Figures (16)

• Figure 1: The $r$-band science image for SN 2025brs at MJD = 60647 (left), as well as the corresponding $r$-band template (middle left) and difference images (middle right), taken from Rubin . The rightmost panel shows the calculated by performing forced photometry on the difference image at every location. The red cross shows the reported location of SN 2025brs. It is clear that the image subtraction pipelines did not work as expected, presumably due to the bright galaxy that SN 2025brs is embedded in.
• Figure 2: The left panel shows the of the difference image after injecting a galaxy and star in the science image, and only a galaxy in the template image of Rubin . This was done repeatedly for different magnitudes of the star and galaxy. The right panel shows which areas of this parameter space would be counted as detections using the default $> 5$ threshold. This shows that very faint targets that could not have been detected are incorrectly classified as detections when their host galaxies are very bright ($\textrm{m}_{\textrm{gal}} \lesssim 13 \; \textrm{mag}$).
• Figure 3: The left panel shows possible injection locations for SN 2025brs with the default settings. Note that in this case the injection locations naturally fall on an elliptical isophotal contour; however DETECT is written to identify appropriate injection locations for arbitrarily complex galaxy morphologies. Since the host galaxy is large and nearby, was able to find locations that are sufficiently spread out so they can all be injection together (i.e. they are in the same injection iteration). The injection locations in the right panel were selected with more strict criteria, and are therefore more spread out and divided over two different injection iterations, indicated by their color. The red cross in both panels indicates the location of SN 2025brs.
• Figure 4: The left panel shows the science image of SN 2025brs with bright ($m_{r} = 18$ mag) sources injected in the locations of one of the injection iterations shown in the right panel of Figure \ref{['fig:injection_locs']}. The right panel shows the difference image after performing image subtraction. The red cross in both panels indicates the location of SN 2025brs.
• Figure 5: The blue region in the left panel shows the locations in the host galaxy where the flux in the template is within 5% of the flux in the template at the SN location. The red cross shows the location of SN 2025brs. The right panel shows the distribution of fluxes that are found by performing forced photometry on the difference image at all of these locations. The dashed vertical line indicates the median of this distribution, which is equal to $863\pm307$ nJy.