#kbosearch

Picture from the Colorado DMV.

I'm super excited to report that the two KBOs we are deciding between are officially called 2014 MU69 (PT1) and 2014 PN70 (PT3).  Those are a mouthful of random numbers and digits, huh?  That's why we call provisional designations like these "license plates".   My first thought was that I should get someone on our team to have this as their actual license plate.  I'd do it myself, but I don't own a car.  Sadly, CO license plates have a max of seven numbers.

They are also the first official names for our KBO targets, which means you can go to JPL's Horizons, type in those names, and learn where to look for the objects in the sky from any location in the Solar System.  You know, if they weren't among the dimmest objects discovered in the Kuiper Belt and could be easily spotted in ground-based telescopes.  

Once one submits positions of an object from more than one night to the Minor Planet Center (MPC), names are immediately assigned.  So, why did this take so long for our KBOs to get designations from the MPC?  Well, this one's our fault really.  We wanted to send the MPC our very best orbit positions, not the first ones that came out of our computers.  Or the second ones. So we waited until we had results that we thought were super reliable.

Since we didn't have submitted positions done by the announcement of the discovery of the two objects, we had to come up with alternate names.  Internally, PT1 had been given the designation 1110113Y, a name that has zero meaning outside of the KBO search team.  The name makes more sense if you think of it as 11-1-011-3Y . We called it that because it was discovered on HST visit pair 11-12, chip 1 (of 2) and it was the brightest on shift stack 011, though we could see it on many stacks (I explained our shift-stacking method here.  At one point last summer, I got really excited because I thought I discovered a KBO, but nope, it was object 1110113Y again, popping up in a different stack.  Nope, hopes dashed in 30 seconds flat.).  The last two numbers are just alphanumerics because it was one of hundreds of potential sources that Marc Buie's program picked up per shift stack that he had to vet to determine if it was a real object.  Using A-Z in addition to 0-9 gives us 1296 options for a 2 digit number, versus 100.  

Among, us, perhaps Simon Porter was the only one who had the whole string memorized.  We all called it object 11.  It was really weird reading posts online and seeing people talk about 1110113Y, when none of us called it that in our own discussions.  There was also object 7 (0720090F, found in the pilot), object 45 (4510067S), e3 (e31007AI - we switched to hexadecimal for the tens digit after visit pair 99-a0), and finally g1 (g12000JZ).

We really didn't want to be calling anything "object 11" in the news.   We never had 11 objects and the "visit pair 11 and 12" concept just took too long to explain.  So we needed alternate names, and Xena, Santa and Easter bunny were not our style.  We knew we'd be getting the license plates eventually; so we needed something ... forgettable.  Saying "remember that Xena thing that was in the news?" to my family and friends so they'd know what I was talking about when I was trying to discuss Eris was bad enough. We didn't want to unofficial names to have a life of their own.

By the fall, we had 11, e3 and g1 left as targetable in the fall, so how about PT1, PT2 and PT3-- Potential Targets 1, 2, 3.  And those names stuck with us so much that we stopped using 11, e3 and g1-- I've been asked if PT3 was e3 or g1 more than once.

So here are our names: 0720090F = 2014_MT69 1110113Y = 2014_MU69 (PT1) e31007AI = 2014_OS393 (PT2) g12000JZ = 2014_PN70 (PT3)

The minor planet center gives us these names based on when the object was discovered (they explain how here).  

Here's how it works.  I did not invent this system, so do not blame me for it.  Objects 11 and 7 were discovered in the 12th half-month of 2014, designated "M" (because they skip I and Z).  So the first 25 new objects (because they just skip “I” now) found from June 16-30 are 2014 MA, 2014 MB, ... 2014 MZ.  The next 25 are 2014 MA1, 2014 MB1, ... 2014 MZ1.  So by the time we sent our positions in, they'd been through the truncated alphabet 69 times already. Thus, 2014 MT69 and 2014 MU69, are consecutive designations.  If we sent our designations in earlier or later, we might have gotten nicer numbers (I've also made all the jokes about the designations already, so you don't have to).  Boy, July 16-31, 2014 (half month "O") was a busy, busy month, I wonder what was going on then.

Why is object 45 not on the list?  Well, you only get a name if you see an object on more than two nights.  We never followed up on it because we knew immediately that New Horizons couldn't reach it, even based on what little we know of its orbit.    It would have been dishonest to request Hubble time for an object what we knowingly couldn't reach.   Hubble time is very limited, and wasting the orbits that we had been given for following up an object for the sake of chasing it, would have jeopardized our ability to get more time to characterize the good objects. Object 7 was a different story.  Our first calculations had object 7 as a more reachable target than object 11, but we weren't able to rule it out without follow-up observations.

The SwRI contingent of the New Horizons post-Pluto encounter target search team during one of our daily phone meetings in Marc Buie’s office.

L to R: Marc Buie, Alex Parker (Berkeley), John Spencer, Cathy Olkin, Anne Verbiscer (UVA), and Simon Porter. Our first Hubble Space Telescope images were taken on June 16.  The last set of image were taken on June 26.  If we didn’t find 2 objects by June 30, the folks at Space Telescope Science Institute (who manage HST) wouldn’t be able to plan our the rest of our program in time for its July 7 start date.  We basically dropped everything that we were doing to work on the search until we got something.  Alex flew out from Berkeley on a one-way ticket.  Marc, who had trips to Hawaii and Finland planned during that time, had to cancel both of them.  Read that last sentence again.

Meanwhile, the New Horizons spacecraft was scheduled to come out of hibernation.  Members of the science team participated in a dry run for encounter time:  they practiced studying mock images and preparing a press release about what they found under a clock.  We didn’t do any of that because we had to focus all of our efforts on the encounter search.

The Hubble Space Telescope images are quite different from ground-based images, and we took an entirely different approach from what I described in my last KBO search post.

Explaining what made this so demanding is a little on the technical side, but I’m going to give it a shot.  This blog has an “ask” feature if you want to know more.

From HST’s perspective outside Earth’s atmosphere, the stars stand still.  From image to image, you can find them in the same place with the same shape and brightness, so it’s easy to remove them cleanly.  This is fantastic news: without the atmosphere, starlight isn’t so smeared and they appear smaller, so it’s less likely that starlight will blend into a KBO, making it impossible to spot.  WFC3 (the HST camera we used) has higher resolution than most ground-based cameras to accommodate HST’s sharpness.  

Yet these advantages come at a price.  In pixel-space, our KBOs can move a rather larger distance, making it hard for the eye to spot it, the way the KBOs jump out in ground-based data when you shuffle through images in rapid succession.  Our atmosphere also does a reasonable job at shielding us from cosmic rays.  You still seem them in images taken at sea level, but wow-oh-wow do HST images have cosmic rays.  Cosmic rays show up in images as bright white dots and streaks, and I’m not exaggerating when I say that they are as many of them as there are stars in some images.  Did I mention we are looking in the star-choked galactic plane?  That’s a lot of cosmic rays.

With these problems, taking two images and blinking them simply wasn’t feasible.  I wasn’t the only person who, used to the old method, required a lot of convincing.  Thus our team came up with the novel idea of stacking shifted images. 

Let’s see how well I can write this out:

In my last post, I talked about hunting for KBOs in the ground-based Magellan data.  I found 94 main-belt asteroids and 6 Kuiper belt objects, only two of which even have the possibility of being targetable (the better of those two had been found by our team in previous years and was known not to be reachable.)  Well, objects in our solar system appear to move relative to background stars in pictures, and the closer an object is, the further it will have moved between two images.  We also know that an object with high inclination (i.e. has a tilted orbit like Pluto does), will be out of our reach.  Thus, the only objects that will work are objects called “cold classicals.”  

We found that for a typical search field set (a set of 5 images, and then another set of 5 taken a few hours later), that 19 different orbit rates were sufficient to describe all the cold classical orbits that would work for New Horizons.  For example, let’s say there was a KBO with an orbit that most closely resembled Possible Orbit #7.  Between images 1 and 3 in the first set, we could expect our object to travel 17 pixels up and 13 pixels to the left, no matter where it was on the image.  If it matched Possible Orbit #8, it would have moved 20 pixels up and 15 pixels to the left.  However, if it matched Possible Orbit #11, it would have travelled 30 pixels up and 10 pixels to the left. 

So we took image 1, and made a copy with the entire picture shifted 17 pixels up and 13 pixels to the left.  We also made copies that were shifted 20 up and 15 left, as well as 30 up and 10 left.  Basically, we 19 different versions of image 1 for every Possible Orbit.  And then, 19 times over, we calculated what we needed to make images 2, 4 and 5 match image 3 too.  So if there was a KBO that matched Possible Orbit #7, it would now show up *in the exact same place* in each of the five images in the shift set for Possible Orbit #7.  It would move a little, but wouldn’t travel too far between images in the shift set for Possible Orbit #8  (the edges of the KBO might overlap between images).  The object certainly would not line up from image to image in the shift set for Possible Orbit #11!

We took the five images for each of the 19 Possible Orbit’s shift sets and run a modified median filter on them, to make a single combined image.  Only bright spots that registered on most images would make it through.  Not all things that made it through were KBOs.  We could have bad luck and there would be a cosmic ray or a bit of a core of a star that showed up in the same place multiple times.  Even so, we could take this image and create a list of possibilities and vet these possibilities as potential objects.  

A real object would be of a certain brightness (absolute mag of about +27, for you astronomer folks who know what that means).  A real object would also show up in the *other* set of five images at the predicted location (also stacked and filtered in the same way).  We would not expect that the brightness of the potential object would be especially different from the first set to the second set.  We would also see a fainter, more spread-out version of it in other stack images we made  (i.e. an object that had an orbit that matched Possible Orbit #7 would appear in Possible Orbit #8’s stack image too).  When we pulled up the original images, we should see it on all 10 images, discounting bad luck with stars and cosmic rays.  Success!

Of course this is all slightly more complicated than I made it out to be here.  The number of Possible Orbits for each set of fields depended on whether HST managed to take the entire set of 5 images before it got blocked by the Earth (this happens every 90 minutes).  If that happened, the set of five images would be taken over a much longer period of time, and we’d have to make 78 orbit groups, instead of 19.  

To make matters worse, the group of Possible Orbits were slightly different for each of the 20 visit pairs in pilot program as well (because the time between images didn’t match up perfectly).  Thus Possible Orbit #7 in Visits 3 and 4 did not stand for the same KBO motion as the Possible Orbit #7 for Visits 5 and 6.  Luckily, careful computer programming allows you to get around all this.

While the shift-stacking technique worked out really well for us, you would not want to use it to search for other kinds of unknown objects in the Solar System.  For one, there would be too many stacks to make.  We only cared about KBOs with very particular orbits that New Horizons could reach, but many KBOs have highly inclined orbits, so we’d have to create more possible orbits that point at different angles.  We also have to include many different speeds for objects closer to the sun, such as main belt asteroids were moving at different rates.   Even with just 19 Possible Orbits, 10 images became 228 images (19 rates * (5 images per visit + 1 stack) * 2 visits ).  I hope you have some good computers.  Our uncompressed images are 168 MB each, so that’s 38GB.  We had 20 of these visit sets. You can kiss your entire hard disk goodbye pretty quick with this project.