Magnitude 19.9 TESS TNO: (55565) 2002 AW197
I said I was going to look for Sedna in my last TNO investigation, but then I discovered Sedna's magnitude in TESS was 20.8, not the 20.2 that I had seen referenced in a paper. It seemed wise to find an intermediate object to detect that was dimmer than Orcus (mag 19.1) before I tried to go after super-dim Sedna. (55565) 2002 AW197 (lovely name) is a magnitude 19.9 TNO that orbits beyond Pluto and is observed by TESS in sector 8, camera 1, CCD 3. It fit the bill as an intermediate target and proved a solid challenge. It also allowed for testing incremental enhancements to my detection process that will almost certainly help me detect Sedna when I get to that point. I'm building this bridge and crossing it at the same time.
1. This is a full frame image for TESS sector 8, camera 1, CCD 3. The red square is the region of interest for this study.
2. These are the 9 processed TESS frames of the region of interest in image 1 that will be differenced in subsequent image sets. Each frame is a median integration of 10 TESS observations (4 hours of data) on the date of observation. The static sky template is shown if you click and hold the image set above. The template is built from these 9 frames plus 5 more.
3. These are the frames of image set 2 with the static sky template removed. See if you can find the small, slow object moving across the frame before clicking and holding the image set to see the streaks frame.
4. This JPL's expected position for 55565 over the course of the observations.
Image set technical details:
Image 1: This image just shows where we're searching for 55565 in the context of a sector 8, camera 1, CCD 3 full frame image.
Image set 2: I started with 17 frames built from median integrations of 4 hours of data on each of the observation dates (files here). Each frame has a simple constant background removed (which is new), is clipped on the high-end at 98% and is brightness matched. These are the 9 frames from 2/6/2019-2/14/2019 that were preserved from the 17 through the whole detection process (3 were rejected before building the template and 5 more were rejected after differencing). The static sky template is a trimmed max of 14 frames. You can see a small difference in the 9 frames I show here and the static sky template.
Image set 3: These are the difference images of the 9 non-rejected frames. Difference images are calculated as a pixel-wise subtraction of the static sky template from each of the frames in image set 2. A floor of zero is enforced in the differencing. Remember to check out the streaks view of the difference images to see the trajectory of 55565 over time.
Image set 4: These are the difference images from image set 3 with JPL Horizon's position for 55565 plotted for the dates of the frame.
Other process notes:
- What constitutes a detection: 55565 led me to ask the question: when have you actually detected an object? You can often "see" something in the noise if you know where to look, but that's not really how it works when you're finding new objects. Since you generally need 3 observations to determine an initial orbit, at least three 3σ thresholded detections should suffice for a detection. I'm also a fan of 1σ thresholding and then culling anything smaller than 2 contiguous pixels, so I might use 3 of those sometimes. This criteria helps determine a good place to stop trying to improve the detection quality and move on to a more challenging target. For this study I found 3 detections at 3σ and 4 with the 1σ and cull method.
- Frame rejection: Frame rejection has been pretty useful, but I've largely glossed over it in this writeup. I'll be more specific here. I started with 17 frames. After removing the background from each frame I rejected 3 frames that had mean values that were outliers in the set of frame means. I then clipped, background matched, templated and finally differenced the remaining 14 frames. Next I rejected 5 difference frames whose means were substantial outliers in the set of difference image means. That left the 9 frames I present here.
- Background model: New for this study I estimated a constant background level and removed it from the cropped frames before I processed them. The model is a simple masked sigma-clipped median. For a 200x200px crop I think this is a good start. For full frame images it will likely have to be a 2D model. Very interestingly (and not surprising in retrospect), removing the background relieved a lot of work brightness matching was doing. All my coefficients went to nearly 1 and 0. Lots of 0.998s and 0.02s and the like.
Ideas & Todos:
- Detection trend: There's a weird trend in this detection. 55565 seems to have the strongest (brightest) detection on the first and last frame and no detection whatsoever in the middle frame. There is a bright star closest to where 55565 would be in the middle frame, but this could also be an artifact of my process somehow. I will watch out for similar trends in future studies.
- Image registration: The static sky template and the retained frames you can see in image set 2 look slightly different. It's subtle, but it's more variation than previous studies. It may just be brightness differences in the some of the rejected observations, but it might also be some small jitter in the spacecraft. Would image registration help here? It might be worth a try.
- PSF modeling: I bit the bullet and reinstalled my badly behaving version of Anaconda so I could use the latest version of photutils. It includes some PSF modeling tools that I need to try out. I may not be able to go much deeper without deconvolving a PSF - or that's what I keep thinking anyway.
So close to 20! At magnitude 19.9, (55565) 2002 AW197 is about half as bright as Orcus (mag 19.1). Sedna will be about half as bright again. The question is whether I go straight to Sedna next or choose another intermediate object to find before that. The objects with legit names (Orcus, Sedna etc) are more appealing to me for no good reason. Although Sedna is also a more interesting class of solar system object in general being very eccentric and therefore very, very distant at aphelion. It's also one of the solar system bodies that's used as evidence for the hypothetical Planet 9. I'll get to it soon. If not in the next study than surely the one after that.