As far as I understand, DIA-Sources that occur within 1" of each other will be associated to the same DIA-Object.
When we have strongly lensed transient, we may have multiple images that will be resolved and detected but have a separation below 1". Let’s for simplicity’s sake assume that the strong lensing system is a double, so we only have two images. Let’s denote them Img1 and Img2. I assume that no DIAObject is associated with the position of the strongly lensed images before hand. In that case then I have the following questions:
Given an initial detection of a transient at Epoch 0, is it correctly understood that the centroid, with which DIAObject membership of later detections is decided, is fixed to the centroid determined at Epoch 0? Or is the centroid of the DIAObject allowed to change over time as more DIASources are associated with the object?
Assume that Img1 is detected at Epoch 1. Then, at a later visit Epoch 2, both Img1 and Img2 are detected and resolved. As far as I understand, then the DIASource of Img2 will be associated to the DIAObject of Img1. Will the DIAObject then contain two different DIASources at Epoch 2, or will the two DIASources be combined into a singular detection?
Assume that both Img1 and Img2 are detected at Epoch 1 and constitute two different DIASources. At Epoch 2, both Img1 and Img2 are detected again. In this case of simultaneous detection of resolved DIASources, how is the creation of and association with a DIAObject handled?
I saw mention of an end-of-night process to tidy up duplicate DIAObjects / other association issues in this thread. How would edge cases like the ones I’ve described be handled in such a process? Are they being considered?
The broad answer to your questions is that we do not currently deblend detection on the image difference, so it is not possible to create multiple DiaSources from the same visit that are closer than about 4.5". That radius is approximate, and is based on the (varying) footprint size of the DiaSources. The association radius is 1" as you state, and any DiaSource detected within that radius of a DiaObject will be associated with that DiaObject. We do have plans to implement deblending for image difference detections in the future, which would allow DiaSources to be created that are closer together.
Given an initial detection of a transient at Epoch 0, is it correctly understood that the centroid, with which DIAObject membership of later detections is decided, is fixed to the centroid determined at Epoch 0? Or is the centroid of the DIAObject allowed to change over time as more DIASources are associated with the object?
That is currently correct. The DiaObject location is fixed at the epoch of initial detection. We plan to update the DiaObject location in the future as more DiaSources are associated, but that is not yet implemented.
Thanks for the answer, Ian. I’m specifically interested in the case where the two sources are not blended, but have low separation. Would two such sources within 4.5" of each other simply be assumed to be blended together regardless? Also, where does this 4.5" number come from? As far as I remember, the expected PSF is on the order of ~1", no?
Sources are considered blended if their pixel footprints overlap, which is at minimum a 21x21 pixel box centered on the sources. My 4.5" is a rough estimate of the typical footprint size of faint sources (bright sources may have substantially larger footprints).
Thanks a lot @isullivan for your answers. I am still a bit confused and would appreciate clarification for the following scenario (considering the current version of the pipeline):
Epoch 0: Two transients, t1 and t2 (e.g., two images of a lensed transient), appear for the first time at positions x1 and x2, separated by 3″. Will this produce one DiaSource and create one DiaObject?
Epoch 1: The same two transients appear again at x1 and x2. Again 1 DiaSource will be produced and associated to the one DiaObject created at Epoch 0?
Epoch 2: t1 vanishes, only t2 remains at x2. Will a DiaSource be created at x2? If x2 is outside the 1″ association radius of the DiaObject from Epoch 0, will a new DiaObject be created?
I’ll add another wrinkle to this question from the Solar System side. Outer solar system small bodies (typically known as Trans-Neptunian Objects, TNOs, Kuiper Belt Objects, or KBOs) are faint moving objects that should show up as DIASources and DIAObjects. Occasionally (~5% of the time for LSST-like observations?), these will be resolvable similar brightness sources, e.g., two r=24 objects with a separation of 1". As a solar system object, they will both move… in a single night their relative position and brightness will probably be fixed to within detection limits. Similar to the strong lensing transients, the binary nature itself will evolve over time: each object can have its own lightcurve (can be up to ~1 mag), the relative position will change detectably over ~months, and there can be significant amounts of time in the mutual orbit where the two objects will be inseparately close together.
These are a bit of an edge case, even within the solar system, but they are scientifically quite valuable and interesting, so understanding how these will be tracked internally is valuable. The solar system aspects (tracking and linking detections together into orbit fits) have also been raised occasionally to those working on solar system analyses. But I thought I would chime in here to provide another angle to the discussion.
Yes, I expect that this would create one DiaSource and one DiaObject with the current code. We have recently resolved one of the major blockers for re-implementing difference image deblending, so I anticipate we will be able to work on that algorithm in the coming months.
If they are within one footprint’s width of each other they will again produce a single DiaSource. If their relative brightnesses have changed, they might either create a new DiaObject or be associated with the previous DiaObject.
Yes, if t1 vanishes and the centroid of the DiaSource is more than 1" away from the previous location a new DiaObject will be created.
