Today it is my (Rico) turn to write the blog. It has been great fun joining the observing run this past week, and I am impressed with the team. It feels a bit like being in the cockpit of a starship: everyone instantly knows what the 100+ buttons do on the six monitors controlling the instrument. It’s an organized chaos I’m quickly growing to love. However, since I have arrived, the seeing deities have not been in the spirit of giving. To appease the seeing gods special agent Joseph finished his all important mission: The Enstickening. With the sticker in its proper place, we optimistically set up the instrument for this nights observations.
Special agent Joseph finishing The Enstickening.
Parker trying to get comfortable while studying for his exam.
The MagAO-X PSF getting ready for Christmas.
The night started with Elena and Katie operating the instrument for Miles. Miles is focused on further validating the polarimetric mode of the instrument. He is hoping to generate some beautiful Polarimetric Differential Imaging images of protoplanetary disks.
The first shift getting ready for the night.
After that we switched to Josh and Eden who spend the rest of the night star hopping for Leah, also in search of protoplanetary disks. It turns out Leah is remarkably good at finding previously unknown binaries for the hopping.
With only two observing nights remaining, a part of the group (among which the PI) called it an early night to start adjusting to a daytime schedule, and like the instrument says: “Als de kat van huis is, dansen de muizen op tafel”. We took full advantage, turning on DJ Haffert’s playlist and letting the music energize the final hours.
When the seeing just hits too hard and your DM’s can’t keep up.
The Burro I had a close encounter with during the day. I don’t think it likes me very much.
Just as things were settling into a rhythm, we got hit by a small earthquake. Some people quickly ran outside. However, I was too busy wrestling with a different kind of instability. I was trying to get the Neural Network (NN) to run its AO correction loop. When the ground is moving, and the control loop is fighting the atmosphere, you quickly learn what your priorities are. It’s fair to say my own internal NN has not been performing well on little sleep, as I almost walked straight into a Burro earlier this afternoon. The Burro was clearly not happy with this sudden close encounter, and I was also pretty startled. The instrument’s NN, however, survived the tremor, and helped us collect better data when the seeing was not ideal.
Fun ‘fact’ of the day: Donkeys Kill More People Than Plane Crashes
There are an estimated 44 million donkeys on the planet. A donkey kick can be anywhere between 0-2000psi, depending on the size and breed. As of 2013 there were 20000 planes on the planet. Due to these numbers it is safe to presume that this is true. In 2016 there were 325 aircraft related fatalities. In the same year there were 5478 admissions to emergency rooms across the world, 1478 of these resulted in the death of the donkeys victims. (the very reliable source: https://imgur.com/gallery/donkeys-kill-more-people-than-plane-crashes-QAhE6)
Song of the day
It is definitely better when not everything is in C.
“Eden, you have to stop making people blog” — anonymous
Taking that note, we will no longer have any people blogging. Eden asked me to fill in.
Today, I went out human-watching at sunset.
While we were out there, Miles kindly took a portrait of me to use on my dating profile.
I understand that these nerds have finally enstickened their instrument, albeit only halfway. Still, I told my guy at the weather agency that they’re cool, and he’s hooking them up with half-arcsecond seeing for tonight.
Fun Fact of the Day
A viscacha who goes into business is technically called a bizcacha.
Today marked the arrival of both Guanacos and our new team member Leah Albrow from MIT!
*Disclaimer: The Guanacos may have already arrived but I have yet to see one. For the time being, I must rely on third-party sightings of these mythical beings.*
Parker is attracting the attention of all of the animals on the mountain–just call him the Pied Piper of guanacos at this point.
Tonight also marked an important milestone in MagAO-X’s history. We conducted a whole night of observations with an unmodulated pyramid wavefront sensor using Rico’s new neural network!
The pyramid wavefront sensor exhibits a non-linear response, limiting its ability to accurately measure wavefronts. To mitigate this issue, AO groups will often modulate their pyramid wavefront sensors by spinning a little pyramid prism round and round (like a dreidel or a ballerina or me in my middle school computer lab with the little rolly chairs). But you can only spin a pyramid (or a Josh) so fast. The modulated pyramid is also limited in its sensitivity (the Josh, not so much). But with Rico’s non-linear reconstructor with a convolutional neural net (or NNRCNN for a fun acronym that rolls off the tongue), we don’t need to modulate our pyramid. We can run our AO loop fast and furiously, sensing wavefronts with the greatest of ease.
To celebrate, I have provided a list of things that are modulated and unmodulated.
Modulated:
My sleep schedule (some nights I go to bed early and wake up around noon, other nights, I dream about permanently residing in Florida and things don’t end well):
Florida.
Parker’s facial expressions:
Themanyfaces of Parker.
The number of group members on our observing run:
While we witnessed the departures of Matthijs, Tiffany, and Laird over the last few days, we also gained new LCO first-timers being Rico, and most recently, Leah!
Leah and co. enjoying the Dupont telescope.
Unmodulated:
Empanada volume on Sundays:
No explanation needed.Is this legal?
The wavefront sensor of the hour:
Unmodulated pyramid WFS (left) vs modulated pyramid WFS (right)The modulated pyramid image kinda reminds me of this 1988 classic.
Our excitement in getting to spend another 5 days at the best telescope in the world!
We’re all here to figure out puzzles like “Do rocky exoplanets have the same composition as Earth?”, “Do all protoplanets produce Halpha emission as they accrete?”, and “Can our AO system reach the contrast necessary to image low mass planets in reflected light?”
It turns out, we also have a lot of puzzles to solve that we didn’t realize we needed PhDs for, such as, “Why is the attempt to upgrade the AO computer cursed?” and “Why does Joseph no longer think computers are deterministic?”
The saga of the long-legged PI and the Cursed Computer (photo credit: Joseph)
Then there are the questions that make us think we should have gone into philosophy rather than science, such as, “Why do we find clouds so pretty even though we hate them?”
Sunset yesterday
The hardest questions are the deep mathematical ones like, “Why is the seeing when I’m on the telescope always worse than the median?”
Worse than median seeing
And then there are the idle puzzles one wonders about at 2 AM such as, “What is the collective noun for a bunch of graduate students? [my vote is for ‘pile’ as in the P in PhD]”, and “How many snacks does it take to satisfy said pile?”
Perhaps a better question is, “Can we get Trader Joe’s to sponsor MagAO-X?”
I’m optimistic that all of these questions will be answered in my lifetime.
But when the seeing gets bad or the sky clouds over, the puzzles we really like are the NYT crossword type (see 2024Ab Day 10). I’ve never made a crossword puzzle before. Tonight I experimented with two free grid-creation applications, and my main takeaway is that the folks edited by Will Schortz are quite impressive. Nevertheless, here for your enjoyment is the AO Puzzle:
Fact of the Day: A record 889 contestants participated in the 2025 American Crossword Puzzle tournament, which is held every year in Connecticut.
Today’s blog post is going to highlight my raison d’être for working with MagAO-X: its new polarimeter!
Polarimetric differential imaging
The MagAO-X polarimeter was initially installed in spring of 2025, consisting of a rotating half-wave plate (HWP) and polarizing beamsplitter cube (PBS). The combination of these two optics enables MagAO-X to measure the angle of linear polarization of light (i.e., the orientation the electric field is oscillating). Detecting polarized light is useful for imaging scattered light in astronomical scenes, in particular light reflecting off of circumstellar disks!
Schematic of polarized light generated by a linear polarizer. Design of the HWP and PBS for MagAO-X (credit: T. G. B. de Souza)
The major benefit of polarimetry is that light reflecting off disks, for example, is polarized, while the on-axis starlight is unpolarized. The PBS directs half the unpolarized starlight to each camera along with the small percentage of the fainter disk light which is polarized in orthogonal directions. By subtracting the two camera images the unpolarized starlight is almost completely cancelled out, leaving only the polarized circumstellar signal. This technique is powerful for attenuating the starlight, orders of magnitude better than using angular or spectral differential imaging, alone (although these methods can be combined with polarimetry for even better attenuation!).
Examples of polarimetric disk images from VLT/SPHERE (credit: C. Ginski and DESTINYS team)
Calibrating the MagAO-X polarimeter
The fundamental reason disks have polarimetric signal is due to the reflection of starlight towards the observer–reflections polarize light! One may wonder, what about all the reflections off the mirrors inside our instrument, won’t those also polarize the incoming light? Yes, indeed, every reflection will polarize the light, inducing instrumental polarization (IP), which cannot be distinguished from circumstellar signal. In addition, certain optical elements will rotate the angle of linear polarization of the incoming light, and can even turn linear polarization into circular polarization (crosstalk). Because the PBS only filters linear polarization states, any light which becomes circularly polarized effectively disappears, leading to a net loss in polarimetric throughput.
The effects of instrumental polarization are mitigated to first order by the HWP itself–by placing the HWP as early as possible in the instrumental beam path, we can cancel out any IP induced downstream of the HWP through another layer of differential subtraction by rotating the HWP and taking multiple images. This will not correct for any crosstalk though, which requires more sophisticated methods. Modern polarimeters overcome both IP and crosstalk by directly modeling the polarimetric response of the instrument and using the solution for correcting astronomical observations. This modeling requires calibration sequences injecting light with known polarization states into the instrument and measuring the response.
MagAO-X does not have the capability of injecting polarized light on the same path as the on-sky beam, so I designed a polarization generator which can be aligned on the exterior of the instrument precisely for measuring the calibrated polarimetric response. I got it designed and built in Tucson before shipping here to the telescope, where it will remain in the clean room for future polarimetric calibrations!
The polarization generator aligned in front of the HWP, ready to inject a flat-field of polarized light.
Initial results from the polarimetric calibration
After a week of Joseph and Jared working tirelessly on getting my calibration data in the right format with the right metadata, I was able to reduce it and begin the measurements of the polarimetric response of MagAO-X!
Normalized single difference flux at different HWP and IMR angles.
These images show the direct polarimetric response of the image on the two cameras at the three filters we can use for polarimetry (r, i, and z). Each data point is the fractional polarized flux measured between the two cameras: +1 means 100% of the input vertically polarized light was measured, -1 means 100% of the input vertically polarized light has now become horizontally polarized, and 0 means 0% of the input flux was measured. An ideal curve goes from +1 at HWP=0° to -1 at HWP=45°. Shifts in the locations of the curves indicate retardance in the instrument, and when the curves are squished less than ±1 it indicates a loss in polarimetric efficiency due to crosstalk.
From these curves we can see immediately by eye a lack of sensitivity at r-band, which can be attributed to the PBS–it has non-ideal performance below 650nm, which is half of the r bandpass. We can also see that the image rotator (IMR) has clear retardance and crosstalk effects as the different colored curves shift and squish at different IMR angles.
The overall polarimetric efficiency can be estimated from the maxima and minima of the normalized single-difference curves. A curve that goes from +1 to -1 has 100% efficiency, while a curve that goes from +0.4 to -1 only has 70% efficiency. The above plot shows the estimated polarimetric efficiency for each filter at different image rotator angles. The highlighted region indicates the typical image rotator angles used on-sky by MagAO-X.
First off, we can see that at i- and z-band we can reach an efficiency of almost 100% without any additional polarimetric components in the instrument, which is great! However, this is reached at image rotator angles that we never use on-sky due to the fixed orientation of the instrument pupil offset compared to the telescope pupil–this is difficult to change because this angle is optimized to minimize the effects of dead actuators on the deformable mirror (DM) and all pupil stops in the instrument are co-aligned with the DM. Thankfully, even with the current pupil offset we’re only losing 50% of the polarized flux, at most.
Future work
Now that these data have been measured, I am going to begin working on fitting polarimetric models for the instrument. In a few nights, I will also take measurements of polarized and unpolarized standard stars, which allows me to measure the polarimetric response of the telescope’s tertiary mirror (which we know has IP and crosstalk). Hopefully I will also get some time on a circumstellar disk, too!
Before next observing run I am going to work on the design of a polarimetric compensator to optimize the polarimetric efficiency by inverting the effects of the instrumental retardance and crosstalk. This will require new optical components and motorized stages, which means I’ll be preparing a new proposal asking for money! If all goes well, we should have an optimal polarimeter in time for the 2026 observing runs.
Other news
Following the arrival of Rico, we’ve had to say goodbye to Matthijs, keeping our Dutch equivalency number constant
Hallo, Rico; tot ziens, Matthijs!A trio of knit-wearing AO scientists
As we progress further and further in the run, our attempts for consistent sleep are getting more and more exasperated, to the point of insanity for some
A Faraday-cage construction around one of the dorms
Observations are continually mostly well, although our luck with the weather is becoming more tenuous. Everyone keep your fingers crossed and knock on wood we can finish out the remaining 5 nights successfully!
Song of the day
twenty one pilots – Polarize
Fact of the day
Hawai’i is the extinct and endangered species capital of the world. Most native Hawaiian plants and animals are endemic, meaning they are found nowhere else in the world but Hawaiʻi. Once they are lost to extinction, they are gone forever, largely due to habitat loss and impacts of invasive species. Current estimates show 95 of 142 endemic bird species, 50% of land snails, and 134 species of plants have been extinguished from the Hawaiian islands.
