The Color of Sunspots: Studying Solar Activity

In this blog, we have highlighted how stellar activity can hinder and even masquerade as planetary signals (see Gliese 581 and Gliese 667C). The stars discussed in those posts were important, not only because they may harbor habitable zone planets, but also because they are M Dwarfs. HPF is explicitly being built to discover and confirm planets just like the ones discussed around these stars. There are definitely challenges to overcome, including being able to separate the stellar signals from the planetary signals, however HPF’s use of the near-infrared should help with these problems.

False color image of a solar prominence as seen by NASA’s SDO mission. They have many more images and a fantastic youtube page.

The NIR has a number of advantages when it comes to stellar noise. One of the biggest is that because sunspots are cooler, their emission peak (think of this as a “temperature color.”) is located in or close to the NIR. This reduces the contrast between the normal surface (the photosphere) and the spot areas. We ‘see’ less of a difference between the two areas. Additionally, the effects of high energy events such as solar flares and coronal mass ejections is lessened as they emit much if not most of their light in the visible, ultraviolet, and X-rays.

In addition to its primary mission of searching for habitable planets around M stars, the reduction of stellar noise in the NIR could potentially enable a second exciting exoplanet study with HPF. Today, astronomers are searching for Earth 2.0. We describe this ideal planet as an Earth-sized planet, orbiting a Sun-like star, within the habitable zone. The challenges in finding our twin are immense. Currently we are only sensitive to large planets and planets at shorter periods. To put some numbers to it, the best instruments in the world can currently measure a Doppler shift of ~1 m/s (walking speed). Just as a side, remember, we are doing this across the galaxy! But back to Earth 2.0. If you were an alien observing the Sun, the largest signal you would see is from our most massive planet, Jupiter. Jupiter tugs on the Sun, making it move at about 12 m/s (the top speed of Usain Bolt – the world’s fastest man). The Earth induces a mere 0.1 m/s on the Sun’s velocity (about the speed of a turtle walking on land). This is a TINY number and it is hidden in the signal of other planets and the stellar noise previously discussed!

Since we expect a reduction in the noise around M-Dwarfs when using HPF, the question arose: “could HPF also observe solar type stars and confirm low-mass planets due to a reduction in stellar noise?” Or in other words, could it help discover Earth 2.0? This is actually a complicated question as the mechanism that creates the magnetic activity in M-Dwarfs can be much different from solar activity.

In an attempt to provide an answer, members of the HPF team–led by Penn State graduate student Robert Marchwinski–conducted a study of radial velocity (RV) variations of the Sun as a function of wavelength.  In reality, this is a very complicated measurement to make. We are actually too close to the Sun to just point a telescope at it! To do this correctly, we need the Sun to appear unresolved (like a point source). Some ambitious astronomers have tried to do this using reflected light from asteroids in the solar system. However, this introduces many additional factors you have to correct for: like the speed of the asteroid, how it is rotating, its orientation, and more. Thankfully, we have another way.

The F/F’ Technique

Astronomers have long been aware of the problem of stellar activity and much of the research has focused on modeling how activity works and how it would affect measurements. In 2012, a new approach was developed. Dr. Suzanne Aigrain and her collaborators came up with a method to use the brightness variations of a star to estimate the radial velocity variations induced by magnetic activity. This technique is called the F/F’ method. At its core, this method assumes that we have one spot or spotted region that is carried along as the star rotates. Spots emit less light and thus you lose some of the light you would normally see from the region where the spot inhabits. This missing light has an intrinsic Doppler shift thanks to the rotation of the star and the convection of the star’s surface, all of which lost. Thus you can get a slight change in your radial velocity. (See below).

A rotating starspot and its effect on the shape of a stellar absorption line.

A rotating starspot and its effect on the shape of a stellar absorption line. (Animation credit: Svetlana Berdyugina)

Using this model and combining it with high quality, well-sampled photometry (brightness measurements), you can estimate the Doppler variations you would see coming from a star. We call it the F/F’ method because it uses the flux measurements, F, and the change or derivative of the flux with time, F’ (pronounced “F-prime”).

With this method, some well-educated assumptions, and solar data, we set out to measure these variations as a function of wavelength.


Obtaining spectra of stars is normally a time intensive process and you need to have lots of data for this method to work. Thankfully, the Sun is the most well-studied star in the Universe. We use solar measurements to educate us on how stars work, how they are powered, and the effects they have on planetary companions.

SORCE Satellite in a clean room. Courtesy of LASP.

Enter SORCE. The SOlar Radiation and Climate Experiment was launched by NASA in 2003 and has been observing the Sun almost continuously. The main goal of SORCE is to measure how the variations in solar output affect the climate of the Earth. To do this, it includes an instrument that takes photometric measurements of the Sun, at different wavelengths, everyday. Exactly what we needed! We also have gotten around the issue of needing to see the Sun as a point source. Photometry is much easier to combine and get a disk totaled value then adding spectra together.

With this data in hand, we could then use the F/F’ method to estimate how the RV variations from the Sun change as you look in different wavelength regions.


We found that by looking at solar type stars in the NIR (where HPF will look), you could see a reduction in the stellar magnetic noise by up to a factor of 4! This particular result came from the analysis of solar data over the entire SORCE mission. If you focus on smaller chunks of time, you consistently see that the NIR has lower variability than other regions of the spectrum. While HPF will not have the sensitivity to detect eta-Earth, there is a very good chance that HPF could help detect other low-mass planets and could be extremely useful in identifying signals that are stellar in origin.

The main results from our analysis showing the estimated RV variability of the Sun with wavelength. We have identified the regions where HPF and 2 other optical instrument's operate/will operate. The gray region is blocked out because the detector introduces noise in this region.

The main results from our analysis showing the estimated RV variability of the Sun with wavelength. We have identified the regions where HPF and 2 other optical instruments operate/will operate. The gray region is blocked out because the detector on SORCE introduces noise that we could not correct for.

An additional result was that while the NIR is quieter overall, there are times in the solar cycle where the optical is almost as quiet as the NIR. This happens during the minimum of the solar cycle, when the Sun is the most inactive and quiet. This suggests that if you are looking for eta-Earth and you have knowledge of the target you are looking at, you could optimize your search by looking only during the star’s stellar cycle minimum.

All of these results were extremely promising and we hope to continue building on these results and making the F/F’ method more accurate. For more information, the results will soon be published in the Astrophysical Journal in a paper submitted by team members Robert Marchwinski, Suvrath Mahadevan, Paul Robertson, Larry Ramsey, and collaborator Jerry Harder.

Posted in HPF Science, Stellar Science | Leave a comment

MLI Blankets

In recent weeks much progress has been made in making Multi-Layer Insulation (MLI) blankets for the HPF spectrograph. Below follows a video of some of the fabrication steps in the HPF clean-room lab: but what are MLI-blankets, how do they work, and why are they needed for the HPF?

What are MLI blankets?

MLI blankets are blankets made out of multiple alternating thin sheets of a highly reflective material – in our case we use 6 micron aluminized Mylar (looks a lot like aluminum foil but more durable) – and a netted spacer material – Tulle in our case which is, yes, commonly used to make bridal veils. MLI primarily reduces heat loss by thermal radiation, but is ineffective at reducing thermal losses by heat conduction and convection. They are therefore widely used as thermal control elements for vacuum applications where radiation losses dominate. Satellites are a great example: MLI gives them the characteristic appearance of being covered with aluminum, or sometimes gold foil.

<b>The Hubble Space Telescope (HST)</b> - With MLI blankets (Image from Wikipedia.)

The Hubble Space Telescope (HST) – With MLI blankets (Image from Wikipedia.)

How do they work?

The main idea behind MLI blankets is the principle of radiation balance and the Stefan-Boltzmann law. Ideally, the perfectly insulating blanket would be a blanket that reflects 100% of the incident radiation. It is, however, very hard to fabricate a single-sheet blanket that accomplishes this, but by stacking many highly reflective layers on top of each other we can achieve higher and higher reflectivity, and reduce radiation losses further. The individual reflective layers can’t touch, as then they would transfer heat between them resulting in no added insulation benefit (that is, we short-circuit the heat transfer); we need a spacer material to space them apart.  A netted plastic material such as tulle (bridal veil) is great for this purpose. It is very thin and light allowing for very easy handling of the overall multi-layered blanket.

Why do we need them for HPF?

HPF is an infrared spectrograph which will operate at cryogenic temperatures of 180K, cooled with liquid nitrogen.These low temperatures are needed so we don’t saturate the infrared detector with background radiation. Moreover, the spectrograph operates under vacuum so radiative thermal losses will dominate, which makes MLI a great choice for thermal insulation.

<b>MLI blanket making</b> - The blankets are assembled in a clean-room to reduce particulates, which are unfavorable for vacuum pump-down.

MLI blanket making – The blankets are assembled in a clean room to reduce particulates, which are unfavorable for vacuum pump-down.

Traditionally, MLI blankets are sewn together; the multi-layered blanket being held together by stitches. However, any kind of hole that punches through the layers tends to degrade the overall thermal performance of the blanket. Another method, of using tag-pins – the small nylon “I” looking pins that are used to hook price tags to clothes in stores – to fix the layers in place, has been mentioned in the literature, see this paper by R. Hatakenaka. That way you don’t need to punch as many holes as when you are sewing, and tagging – a few inches between tags – is faster and less error-prone than sewing around the whole perimeter of the blanket. Moreover, the tag-pins allow you to fasten the layers together without compressing them, which reduces stress around the holes. Lastly, the blankets tend to contract in the direction of sewing, which might lead them to be too small if not over-sized properly.

The blankets need to be properly sized, aligned, and held together to cover the whole radiation shield, liquid nitrogen tank, and copper thermal straps. Strategically placed Velcro-pads are used to align and hold the blankets in place on the instrument. We sew the Velcro onto the blankets to strongly fix them to the blankets. This results in more holes punched on the area of the Velcro than the tag-only method, but sewing fixes them better to the blankets.

See the video at the top of this post for all the details on how we put everything together!

Posted in HPF Hardware | Leave a comment

More on stellar activity: an investigation of Gliese 667C

A few posts back, we explored how even with a highly precise planet-hunting instrument such as HPF, radial velocity noise from magnetic activity on our target stars may hinder our ability to detect exoplanet signals.  In that post, we used Gliese 581 and its planetary system as an example of how stellar activity such as the spots we see on our own Sun can create signals that look deceptively like exoplanets.  The results of that study generated a lot of public and media interest, and one of the most frequent questions presented to our team was whether any of the other well-known exoplanet systems might be secretly harboring any activity-induced false positive planets.  It is a great question, and one we very much wanted to pursue!  HPF team members Paul Robertson and Suvrath Mahadevan published a new research paper on the arXiv preprint server today (it will appear in the Astrophysical Journal Letters shortly) discussing a stellar activity investigation of Gliese 667C–the “other” most-famous red dwarf planet system.  There is a lot of technical information in the research article, but one bit of good news jumps out immediately: Gliese 667Cc, widely believed to be a promising candidate for low-mass exoplanets in the habitable zone, looks to be real!  Let’s look briefly at the results in more detail.

About Gliese 667C

An artists' rendition of what sunset might look like on Gliese 667Cc.  The host star, Gliese 667C, orbits a distant stellar binary (Gliese 667AB), seen in the background.  Image credit: ESO/L. Calçada

An artist’s rendition of what sunset might look like on Gliese 667Cc. The host star, Gliese 667C, orbits a distant stellar binary (Gliese 667AB), seen in the background. Image credit: ESO/L. Calçada

Gliese 667C, or GJ 667C, is an M dwarf star that is part of a triple star system about 22 light-years away from Earth.  It is a distant companion to GJ 667AB, a binary pair of more massive stars.  In 2013, astronomers using the HARPS spectrograph discovered a pair of planets around this star–GJ 667Cb and GJ 667Cc.  Both of these planets have super-Earth masses, but planet c is especially interesting, since it has a minimum mass just 4 times that of Earth and it lies in the stellar habitable zone!  This makes Gliese 667Cc one of the best candidates yet for a potentially habitable planet.

To make things even more compelling, a separate group of researchers using different data reduction and frequency analysis techniques announced as many as five additional super-Earth planets around GJ 667C shortly thereafter.  This announcement left the system with up to three small planets in the star’s habitable zone!

But there is a catch.  It was clear from the original HARPS analysis that stellar activity is present in the data for GJ 667C.  Aside from planet b, essentially all the planet candidates have periods close to the star’s 105-day rotation period or its aliases.  If we are to be truly confident that Gliese 667C hosts any habitable zone planets, we must correct for this activity and see how many signals remain.

Our Results

We are fortunate that a large amount of data exists for this star, including a good number of stellar activity indicators.  For Gliese 667C, we found the most sensitive indicator was the width of the HARPS cross-correlation function, or FWHM for short.  This is a technically complicated measurement, but it can be understood simply as the average width of the star’s spectral absorption lines.  If magnetic features like starspots are altering the spectrum in a way that would produce false Doppler shifts, FWHM should change.

As for Gliese 581, we found that Doppler measurements made during certain active periods were contaminated by stellar activity signals, as revealed by correlations between the star’s velocity and FWHM.  Fortunately, we were again able to model and remove the stellar activity signals and re-model the star’s planetary system.

So, on to the obvious question: are the planets real?  The good news is that the signals of planets b and c were still present after we removed the stellar activity effects, suggesting those planets are real.  For the rest of the planet candidates, the outcome is not so exciting.  “Planet d,” which was believed to orbit near the outer edge of the habitable zone, is definitely a stellar activity signal created by the star’s rotation.  As for the rest of the planets, we cannot be so sure.  The signals associated with them are so small that they cannot be seen with “industry-standard” analysis techniques, regardless of whether we have corrected for activity.  However, considering how successful our activity correction has been at boosting the signals of real planets, the fact that we see no sign of any of these planet candidates after the activity correction leads us to strongly doubt their existence.

Posted in Exoplanet Science, HPF Science, Stellar Science | 3 Comments