Measuring Exoplanet Atmospheres with HPF

Introduction: Transmission Spectroscopy

As we have covered in previous posts, HPF was designed and built for the goal of detecting exoplanets through the Doppler motion of their host stars.  Of course, there are other astrophysics experiments that can take advantage of HPF!  There are plenty of applications for a stable, high-resolution near-infrared spectrometer.  In today’s post, we will discuss one such application that allows us to learn more about nearby exoplanets: transmission spectroscopy.

Normally, it is extremely difficult to learn anything about the atmosphere of an exoplanet.  The planet itself is lost in the glare of its host star, and for a planet like Earth, the atmosphere is a tiny component of its overall radius–if Earth were the size of an apple, the atmosphere would only be as thick as a normal apple’s skin.  One special situation in which we can catch a glimpse of an exoplanet’s atmosphere is when the planet transits (or eclipses) its star.  When a transit occurs, a tiny (but measurable!) fraction of the star’s light will filter through the planet’s atmosphere, where atoms and molecules in that atmosphere will absorb light at very specific wavelengths.  Thus, by comparing the normal spectrum of the star with the spectrum as observed during transit, we can determine which chemical species are absorbing light in the planet’s atmosphere.  With the right contextual information, we may also be able to probe properties like the atmospheric temperature, pressure, and more.

Transmission spectroscopy examines starlight filtered through an exoplanet’s atmosphere to determine the composition and other properties of the planet’s atmosphere (Image credit: Christine Daniloff/MIT, Julien de Wit).

Planets orbiting very close to their host stars get so hot that their atmospheres begin to escape into space, trailing behind the planet like a giant comet tail.  Recent simulations of these escaping atmospheres suggest that absorption of near-infrared light by helium atoms should create a large signal for transmission spectroscopy.  The wavelength of this helium absorption is right in the middle of HPF’s waveband, so as an early science program for HPF, we attempted to detect this effect for an exoplanet transiting an M dwarf star.

The target: GJ 3470b

We have mentioned the M dwarf star GJ 3470 on this blog before–it was actually the star chosen for HPF’s first light image.  The star hosts a planet about the size of Neptune orbiting very close, which is the best type of planet for atmospheric characterization.  Other studies have already suggested that the planet should have a blue sky!  Based on what was already known about the planet’s size, density, and stellar radiation exposure, we predicted it should exhibit significant absorption from helium.  So we set out to determine whether HPF could measure it.

The experiment

When you need a planet to be in front of its star in order to make a measurement, the timing is crucial.  Careful planning is needed to observe the star at a time when a transit occurs at night from your observatory, and the weather has to be good.  In the case of GJ 3470b, we had to observe three such events in order to develop enough signal from the helium absorption.  In the meantime, we also observed the star when the planet wasn’t transiting, in order to compare the in- and out-of-transit cases.  All said, it took us about 5 months to get all of the observations required to do this experiment.

So did we see helium absorption?

To date, GJ 3470b is the smallest planet for which atmospheric helium absorption has been measured.  As a result, the helium signal is relatively small, and we had to be careful with how we merged all the observations into a single spectrum.  We also had to convince ourselves that the signal was not created by changes in the HPF instrument, or in Earth’s atmosphere.

An additional complication comes from the fact that the atmosphere of GJ 3470b is so hot that the temperature causes helium atoms to fly out of its atmosphere into space at high velocity, in a process known as a Parker wind.  This motion of the helium gas distorts the absorption feature we would otherwise expect to see, so we had to carefully model the effects of the Parker wind in order to make sure our data were consistent with the all of the physics involved in the interaction of the planet’s atmosphere with the intense stellar radiation.

After accounting for all of these effects, we determined that the HPF spectra show evidence for helium absorption caused by the atmosphere of the planet.  In the image below, we show the comparison between the spectrum of the star with and without the planet passing in front.  The spectrum shows a little more absorption at the wavelength of the helium feature during transit, in approximately the shape we expect from the wind-shifted motions of the atmosphere.  The signal is weak, but our careful analysis of the data in combination with the wind model suggests it is real.  Our detection is also supported by observations of the same planet by another team using the CARMENES spectrometer.

GJ 3470b Transmission Spectrum

Spectra of GJ 3470 during (top) and outside of (bottom) the transit of its planet. The red curve shows our model of the expected absorption from helium in the planet’s atmosphere. Only during transits does the star’s spectrum show the expected absorption!

In short: the technique works!  This detection is an enticing preview of what’s possible in terms of atmospheric characterization of exoplanets using HPF.

Curious case of blue-shifted broad absorption in GJ3470b

While infrared helium absorption has been detected earlier in the atmospheres of larger planets, the most puzzling aspect of our new detection in the first warm Neptune was its unusually broad absorption profile.  This is especially true on the blue-wavelength side (leftwards in the image above), since the red side of the profile has more noise due to contamination from Earth’s atmosphere.

Line of Sight Velocoty of exosphere around GJ3470b

The line-of-sight velocity field of upper atmosphere around GJ 3470b. An Earth-based observer is towards the left side of the plot, and the host star GJ 3470 is towards the right side.  Bluer colors indicate more rapid motions towards Earth.

The reason turned out to be simple orbital dynamics. The orbit of GJ 3470b is not perfectly circular, and so its motion along the line connecting the star and Earth (the so-called “line of sight velocity”) is actually slowing down during the transit. Just like a slingshot, helium atoms escaping the gravitational pull of the planet continue to move towards us at the velocity they had the moment they escaped from the planet. Once you incorporate this velocity change into the modeled absorption profile from the helium atoms, it matches well with the observation!

What did we learn about the planet?

These measurements allowed us to estimate properties like the density, temperature, and velocity of helium atoms in the planet’s atmosphere.  In the image below, we have actually mapped out the distribution of helium surrounding the planet itself.

GJ 3470b model

Scale model of the star GJ 3470 (large black circle) with its planet (inner white circle) in transit. The red-yellow color scale indicates the density of helium atoms surrounding the planet, based on our transmission spectrum.

With powerful instruments and enough observing time, we can learn a lot more about a planet’s atmosphere.  Similar observations can reveal different chemical species, probe deeper into a planet’s atmosphere, and monitor changes over time.  Eventually, we will use this technique to look for changes in planetary atmospheres that could be caused by life!  So while this first step for HPF is exciting, it is only scratching the surface of an aspect of exoplanet study that will become increasingly important in the coming years.

Where can I find out more?

Details of this experiment were recently published in the AAS Journals in an article led by HPF team member Joe Ninan.  We encourage you to go there for more information!

Posted in Exoplanet Science, HPF Science | Leave a comment

G 9-40b: HPF’s first planet validation

HPF has validated its first planet called G 9-40b, a sub-Neptune sized planet (2 Earth radii) orbiting a nearby low mass M-dwarf star with an orbital period of 6 days. The paper has been published in the Astronomical Journal and is available freely on the arXiv preprint server here.

It takes a Village to Validate a Planet

What does it take to say that a planet candidate is a bone-fide planet? Even for promising planet candidates around nearby stars, it takes a number of follow-up observations to say that a planet candidate is really a planet, and not something else, such as an eclipsing binary star system or other astrophysical false positives that can imitate planet transits.

G 9-40 was originally presented to have a candidate transiting planet in a paper by Yu et al. 2018 using photometric data from the K2 mission. After a closer look at this candidate planet, we saw it looked interesting for many reasons.

First, at a distance of only ~90 light years, G 9-40b is among the 20 closest transiting planetary systems known (see Figure 1), and is currently the second closest transiting planets discovered by the K2 mission to date. Second, the host star is a nearby bright M-dwarf star—but being bright makes it easier for us both to get a good radial velocity precision on this target with HPF, and to observe additional transits of it from the ground at high precision. Further, the brightness of the host star along with the large transit depth of the planet (~0.35%, see Figure 2) makes G 9-40b one of the most favorable sub-Neptune-sized planets orbiting an M-dwarf for transmission spectroscopy with the James Webb Space Telescope (JWST) in the future—more on that below.

Figure 1: Planet radius as a function of the distance of the planet system from Earth in parsecs (1 parsec is about 3.2 light years). G 9-40b (highlighted in the bright red) is among the closest transiting systems known to date. Red points show planets around low mass stars (M-dwarfs) and blue points show planets orbiting more massive hotter stars. Figure 12 from the paper.

Given these reasons, we decided to conduct more observations of this planet candidate and characterize it better to decide if this was a planet or not. To do so we used Adaptive Optics (AO) imaging observations, further precision transit follow-up observations from the ground, and precision radial velocities with HPF. Lets dive into each of these observations below.

Figure 2: Transits observed in the light curve of G 9-40 from the K2 mission. The transits are deep, making it a compelling candidate target for transmission spectroscopy in the future with the James Webb Space Telescope. Figure 7c from the paper.

Adaptive Optics Imaging

To check if G 9-40 is blended with another star—which could confuse us on which star is the true source of the transit, or indicate that the system would be an eclipsing binary system rather than a true planet—we used high contrast adaptive optics imaging observations from the ShaneAO system at Lick Observatory. These adaptive optics observations are often conducted by shooting a high power laser into the air (Figure 3) to generate a so-called laser guide star. This laser guide star is then used to correct for the distortion in the atmosphere by rapidly changing the shape of an ‘adaptive’ mirror in the optical system, allowing us to obtain extremely high contrast images of the star. From these observations we see no evidence of blending or other stellar companions nearby the star G 9-40, further suggesting that G 9-40 is the true source of the transits we see from K2 and in our ground-based data (see next section).

Figure 3: Adaptive optics image (left inset) and resulting contrast curve as observed with the ShaneAO Adaptive Optics (AO) System on the 3m Shane Telescope at Lick Observatory (right). The contrast curve on the left shows the brightness of the inset image (in the near-infrared K-band) as a function of the distance from the center of the star. We see no evidence of a background star in the image that could dilute and/or be the source of the transits we see in the K2 and/or ground-based data. Left: Figure 1 from the paper. Right image credit: Laurie Hatch.

Ground-based Transit Photometry using Engineered Diffusers

To constrain the planet parameters and further update the orbital ephemeris of the system, we used high-precision diffuser-assisted photometry using the ARCTIC imager on the 3.5m Telescope at Apache Point Observatory. We observed the transit using a custom-built narrow-band filter made by Semrock / AVR Optics  that operates in a band with little-to-no water absorption. We often call this filter the ‘cloud-killer’. If you are interested, you can actually buy a very similar filter from Semrock off-the-shelf here!

The plot below shows our ground-based transit, which agrees well with the transits observed by K2 further suggesting that our planet candidate is indeed a planet. Further, these observations allow us to pinpoint the timing of the transit extremely well, allowing us to predict midpoints at the start of the JWST era (~2021-2023) with less than 2minutes of error. Such small errors are important to enable easy scheduling for in-transit spectroscopic follow-up in the future.

Figure 4: Transit observations of G 9-40b. Left: Ground-based transit measurements—showing the flux of the star as a function of time, showing a clear transit dip—obtained using the ARCTIC imager and the Engineered Diffuser on the 3.5m Telescope at Apache Point Observatory (right). These observations allow us to further characterize key parameters of the planet, including its period and its radius. Image credit: Figure 7e from the paper (left), Gudmundur Stefansson (right).

But what is the diffuser-assisted photometry? Excellent question. This means that we used recent technology called Engineered Diffusers—developed by RPC Photonics—to help conduct the observations. Engineered Diffusers are nano-fabricated devices that you can easily place in a telescope beam-path to mold the focal plane image of a star to a stabilized top-hat shape (Figure 5). In doing so allows you to keep the image of the star stable throughout the night, minimizing photometric errors due to the atmosphere. Further, in spreading out the light over many pixels opens up the possibility to observe bright targets on large telescopes—especially useful in observing G 9-40—with the 3.5m Telescope at Apache Point Observatory. If you are interested in reading more about Engineered Diffusers for astronomy you can read more about them in a past press release here or here, or in our previous paper here.

Figure 5: An Engineered Diffuser is capable of molding the image of a star into a broad and stable top-hat shape. b) We used an Engineered Diffuser in the telescope beam path of the 3.5m Telescope at Apache Point Observatory to stabilize the image of the star and allow us to obtain extremely high photometric precisions allowing us to precisely characterize the parameters of G 9-40b. Image credit: RPC Photonics (left), Gudmundur Stefansson (right).

Precision Radial Velocities from HPF

Last but not least, this paper includes some of the first precision near-infrared (NIR) radial velocities (RVs) from the Habitable-zone Planet Finder. In a previous paper led by AJ Metcalf, we showed that HPF is capable of 1.5m/s RV precision in the NIR (see blog post here), enabled by the precise near-infrared HPF Laser Frequency Comb (LFC) also discussed in detail in our previous blog post here. Using precision NIR RVs from HPF enabled by the HPF LFC, we placed an upper limit on the mass of G 9-40b of 12 MEarth. We hope to obtain a robust mass measurement with future RV measurements with HPF in the future. In doing so will allow us to constrain the composition of the planet, and to inform future atmospheric follow-up efforts with JWST.

Figure 6: Precise radial velocities from HPF (left) on the 10m Hobby-Eberly Telescope (right) allowed us to place an upper limit on the mass of the planet of 12 Earth masses. We hope to get a further precise mass constraint by continuing to observe G 9-40 in the future. Image credit: Figure 11a from the paper (left), Gudmundur Stefansson (right).

Posted in Uncategorized | 9 Comments

HPF Commissioning: A New Standard in the Near Infrared

We’ve been quiet…

Our last post, made about a year ago, shared our excitement at having delivered HPF to its permanent home on HET, and recording the instrument’s first data.  Since then, we have been busy doing the hard work of “commissioning” the instrument.  For an astronomical instrument, commissioning refers to a period of performance verification and fine-tuning that occurs once the instrument arrives at the telescope.  Certain tests and adjustments can only be done in conjunction with the telescope, so some patience is required as the instrument team makes improvements before real science can begin.

We are happy to report that HPF has now been released to the HET community for full science operations!  We have also collected enough data that we can start to evaluate the measurement precision of the spectrograph, and determine whether our on-sky performance matches our original project goals.  Here, we will show you some of that data, and explain why we are confident that HPF will break new ground in the study of the nearest exoplanet systems.  Technical details of this work can be found in an upcoming edition of Optica.

Barnard’s Star: the M Dwarf Standard

While the ultimate goal of any Doppler spectrograph is to find lots of exoplanets, boring is better during the commissioning phase.  The only way to test the stability and precision of your end-to-end measurement system–from the telescope, through the fiber optics, and ultimately the optics and detector of the spectrograph–is to make repeated measurements of a star with little or no variability.  That way, any variability seen in the measurements must be caused by the instrument, rather than the star itself.  In other words, the less variability we measure in observations of our stable “standard star,” the better the instrument is performing.

For the M dwarf stars HPF is designed to target, the best standard star is also one of the most famous stars in the sky.  Barnard’s star–named for the American astronomer E. E. Barnard–is the second-closest star system to the Sun (after the alpha Centauri triple system).  It is a little more than 14 percent as massive as the Sun, which is typical of the stars HPF will survey for exoplanets.

Barnard’s star can be seen moving relative to background stars in this time-lapse image. Barnard’s star has the highest on-sky velocity of any star. (Image courtesy Steve Quirk).

Barnard’s star is extremely old–possibly as old as the Milky Way Galaxy itself.  As such, it is rotating extremely slowly, and exhibits little of the stellar activity that causes radial velocity variability in other stars.  That low variability makes it an excellent commissioning standard for HPF.  Furthermore, its location near the celestial equator means it has been observed by telescopes in both the Northern and Southern Hemispheres, so we can compare our results to those of other spectrographs.

 The Observations

The Hobby-Eberly Telescope is queue scheduled, meaning resident astronomers record data for multiple users over the course of a single night.  This method of scheduling is ideal for repeated observations of a single star over time.  From February through July of 2018, HPF recorded the spectrum of Barnard’s star on every night possible.

Our HPF observations kicked into high gear in May, when the Laser Frequency Comb (LFC) from our team members at NIST and CU was fully operational at the HET.  The LFC is based on Nobel-prize-winning technology, and uses carefully modulated lasers and optics to create a perfectly uniform array of laser lines that can be used to provide an ultra-stable wavelength reference against which HPF measures Doppler shifts in a star’s spectrum.  We discuss the LFC in more detail in a companion post.

By the time we shut down HPF for some post-commissioning maintenance, we had made 21 high-quality measurements of Barnard’s star using HPF with the LFC.  These observations span 86 days, which is enough to quantify the baseline stability of the HPF system.  Let’s take a look at the results:

The Results

We would be remiss if we did not emphasize that working all of the kinks out of an ultra-precise Doppler spectrograph is a years-long process, and we are far from done making improvements to the instrument and our analysis techniques.  With that said, our early observations of Barnard’s star are extremely promising!

HPF radial velocities of Barnard’s star. Blue points represent individual spectra, while red points are the average of measurements made within a 20-minute series. The 1.5 meter per second precision is the best ever achieved in the near infrared!

Our current data series on Barnard’s star show a stability of around 1.5 meters per second, which is the best ever demonstrated on an infrared instrument.  For comparison, the most stable measurements of Barnard’s star ever reported come from the HARPS spectrograph, and they check in at around 1.2 meters per second.  Until now, HARPS has been the industry standard in ultra-precise Doppler spectroscopy, so for HPF–with its newer and more complicated infrared technology– to nearly match HARPS so soon after going on sky is a major accomplishment!

Long-time readers of this blog may recall that our goal measurement precision is 1 meter per second, while our current precision appears to be around 1.5 meters per second.  What noise sources might be remaining?  It is probably a combination of stellar activity, contamination from Earth’s atmosphere, small exoplanets, and some yet-to-be identified instrumental systematics.  We are continuing to check everything in an effort to squeeze every last drop of precision out of HPF.

A Planet Around Barnard’s Star?

Just after we finished our commissioning observations on Barnard’s star, the Red Dots team announced they had discovered a candidate exoplanet orbiting Barnard’s star.  The proposed planet, Barnard b, is around 3 times the mass of the Earth and takes 233 days to complete an orbit, so its Doppler signal is quite small.  Nevertheless, HPF’s precision is good enough to at least see some trace of this planet, so what gives?

As it turns out, cosmic coincidence prevents us from having much information on Barnard b at this point.  The orbit of the proposed planet is eccentric, which means the Doppler signal is more pronounced at some phases of its orbit than others.  Through nothing but luck, our HPF-LFC observations completely missed the most dynamic section of the Barnard b phase curve.

The orbital model of Barnard b (blue), with HPF measurements (gold) folded to the orbital phase. Our measurements have not yet covered the maximum of the eccentric orbit.

Thus, while our HPF measurements do not rule out the proposed planet, they cannot yet confirm it, either.  This is just one of many examples of how exoplanet detection is a data-intensive process!


HPF’s first year on sky is in the books, and things are looking quite promising!  We are already setting new records for Doppler precision in the near-infrared, and continue to push towards even better stability.

Now, the fun part begins.  HPF science operations officially started on December 1, 2018.  Whereas Barnard’s star is one of the most familiar, well-studied stars in the sky, we will now expand our operations to examine stars about which relatively little is known.  As always, we will share any exciting new results with you here.

We will continue to observe Barnard’s star, both to validate our long-term instrumental performance, and to learn more about any planets that may orbit this ancient star.  It is as quiet as stars come, but it may still be harboring some exciting secrets.

Posted in Exoplanet Science, HPF Hardware, HPF Science, Project Development | Leave a comment