Micro-resonator laser frequency combs and Exoplanets

This past week Caltech and the Keck Observatory published the following Press Release:

http://www.caltech.edu/news/new-calibration-tool-will-help-astronomers-look-habitable-exoplanets-49624

http://www.keckobservatory.org/recent/entry/new_calibration_tool_will_help_astronomers_look_for_habitable_exoplanets

Wavelength calibration is but one part of the multi-front battle in the quest to detect the velocity reflex motion of an Earth-massed exoplanet in a Habitable Zone orbit of a Sun-like star.  Such an exoplanet causes the star to wobble back and forth with a characteristic velocity of ~10 cm/s, a very slow walking speed!  

Other battlefronts in the quest to detect another Earth around another star include maintaining the stability of the starlight illuminating the spectrograph, the optical and mechanical stability of the spectrograph itself (pressure, temperature, etc.), the properties of the detector that measures the  spectrum, how one analyzes the data, and distinguishing the behavior on the surface of the star itself (starspots, granulation, p-mode oscillations) from the motion the star experiences from an orbiting exoplanet.  All must be dealt with in tandem, but for this blog post I will focus on wavelength calibration.

Wavelength calibration involves mapping a set of particular wavelengths of light to individual pixels on a CCD (aka, the detector or digital camera) in a spectrograph.   Emission line lamps provide a means for wavelength calibration.  Thanks to quantum mechanics, light emits at very specific wavelengths from gases of particular atoms or molecules, and helps explain things like why neon lights appear red.  A common lamp used in astronomy is made of Thorium and Argon, emitting light at a wide variety of colors across the visible spectrum of light.  By measuring a spectrum of a Thorium Argon lamp with the same spectrograph that is used to measure the spectra of stars, astronomers could get a rough idea of what wavelengths corresponded to what pixels on the CCD, and apply that same relationship to the spectra of stars.  Then as stars moved towards us and away from us from the gravitational influence of an orbiting exoplanet, the spectra of the stars will shift in wavelength every so slightly to the blue or red.  By carefully measuring the wavelength shifts, using the Doppler Equation we can solve for the velocity of the star, after accounting for the fact that our telescope is also moving along with the Earth in its orbit in our Solar System.

The revolution in the discovery of exoplanets orbiting other stars was wrought through a break with tradition and advances in technology.  

Astronomers have used high-resolution spectrographs for decades. A high-resolution spectrograph can split light into many different colors; mathematically, this is represented as R = lambda/(delta lambda), where lambda is the wavelength (specific color) of light, delta lambda is the smallest two wavelengths one can distinguish, and R is the spectrograph resolution. A typical R value is 100,000.

With the first confirmed exoplanet orbiting a main sequence star found (and commonly accepted) in 1995, why didn't astronomers find exoplanets decades earlier?  The reason is three-fold: The invention and steady improvement of CCDs, changing how we perform wavelength calibration, and accepting that spectrographs are dynamic instruments - flexing with changes in orientation, pressure and temperature.   Traditionally, many spectra of stars would be obtained over the course of a night, and at the beginning and/or end of a night, astronomers would take observations of emission line lamps like Thorium Argon.  The spectrograph was assumed to be "ideal" over the course of the night (stable and unchanging), and the same wavelength solution would be used for all of the data on a given night.

This assumption of stability was an over-simplification.  As a result the best precision that could be obtained for measuring the velocities of stars was on the order of 1 km/s.  1 km/s  is comparable to c/R, where c is the speed of light, and R is again the spectrograph resolution.  This was five orders of magnitude worse than what was needed to detect another Earth.  In order to make progress, this assumption needed to be abandoned.  Astronomers took two approaches that both worked for finding exoplanets in the mid-1990s - using alternative sources for wavelength calibration that could provide simultaneous illumination of the spectrograph, such as the iodine cell technique, and stabilizing the spectrograph by making it as "ideal" as possible.

Fast forward to today, and spectrographs are built with extreme stabilization and purposefully built for the detection of exoplanets.  Thorium Argon lamps are still used today for the HARPS and HARPS-North spectrometers, arguably the best performing spectrographs for exoplanet detection in the world.  The teams of astronomers behind these instruments have recognized that the lamps can limit their ability to search for other Earths, reaching a precision of ~70 cm/s, now only one order of magnitude shy of what is needed.  In order to make future progress on this battle-front, a new wavelength calibration method is needed.

Enter the laser frequency comb.  

Originally invented for increasing bandwidth for telecommunications hardware over fiber optics cables, the devices pulse laser light at a very high fixed repetition rate such that the resulting spectrum is a series of sharp emission lines.  Hence the word "comb".  In wavelength space, the spectrum looks like a comb:

A laser frequency comb spectrum

A comb

Physicists at NIST realized they would make great wavelength rulers, and developed the technology to stabilize them using a variety of techniques.   Once the laser comb spanned an octave (a factor of two in wavelength), the use of a frequency doubler allowed the comb to be "wrapped around" in wavelength and reference one end of its spectrum to the other end of its own spectrum, a process called "self-referenced mode locking."  The resulting stability of the wavelengths of the laser comb emission lines was unprecedented, measured to be better than 1 part in 10¹⁵.   The possible application of laser frequency combs to astronomy in general and exoplanet searches in particular was obvious.  Here was a means to wavelength calibrate a spectrograph to much better than 1 cm/s.  But there was a problem - the emission lines were too closely spaced together.  So, physicists at NIST developed the hardware to filter out most of the lines - 49 out of every 50, or 99 out of every 100 using two Fabry-Perot etalons in series - leaving just enough lines that an astronomical spectrograph could tell them apart.  These filtered laser frequency combs have now been tested on HARPS and HARPS-North, as well as other spectrographs, and the results are incredibly promising.

The catch - it takes a team of well qualified physicists to run them, along with $1 million in hardware.  Today you can buy one of these combs from Menlo Systems, if you've got a cool million to spare.  With modern astronomical spectrographs now costing $5-$30M apiece, an extra $1M is a lot, but not totally unreasonable.

Enter the micro-resonator laser frequency comb. 

So, the physicists at NIST started working on new technology to build combs that natively have wider spacing between emission lines, called a micro-resonator.  It would simplify the hardware, and perhaps lower the costs.  Collaborating with Kerry Valhala and graduate student Xu Yi at Caltech, scientists at NIST built a prototype that fits on a moderate sized optical breadboard (~24"x24").  But would it work on a telescope?

That is where I come in.  In the press release, I'm listed at the end as "other authors of the paper" for my five seconds of fame:

"Peter Plavchan (BS '01), formerly at Caltech and now a professor at Missouri State University;"

Yup, not only did I get my Bachelors degree in physics from Caltech, but after earning my PhD from UCLA in 2006, I spent another 8 years as a postdoctoral scholar and research scientist at Caltech (JPL and the NASA Exoplanet Science Institute in particular).   Those 8 years of effort and a lot of elbow grease are going to have to suffice in place of any future alumni donations!

It was a great experience to be part of this project bridging two fields - Astronomy and Applied Physics.  My contribution was providing the hardware and the expertise for getting first light with the micro-resonator comb at the NASA IRTF telescope with the CSHELL spectrograph. We had simultaneous star and laser comb light going down to the spectrograph via fibers from one of my instruments.  Basically, I built part b in the diagram below back in 2012, which I published a paper on in 2013:  

Figure 2 from Xu Yi's Nature Communications paper

In 2013 and 2014, I helped design and build a custom part that place two fibers - one carrying starlight and the other carrying the micro-resonator laser comb light - very close to each other in the above diagram.  How close?  Let's just say it required dipping the tip of one of the fibers in acid, an X-ray imaging device, and a 3-D printed part, all courtesy of the engineers and scientists at JPL.

In July of 2014, I went out to IRTF atop Mauna Kea for a few nights for our first engineering run using the comb on the telescope.  It was only one month after my infant son Lincoln was born and after we had moved halfway across the country.  It's cold at the summit of Mauna Kea, nearly 14,000 feet above sea level.  And your brain doesn't work quite right, getting only about 60% of the oxygen you would get at sea level.  Since it was my instrument that we were adding this new dual-fiber part to, I had to be there to train the others on the team with how it worked.  Unfortunately while we were mounting the two bare fibers into the hardware, the fiber broke clean off.  The fibers weren't protected by a "sleeve" and thus were very fragile.  As a result, we could only put unfocused light from the microresonator comb down into the spectrograph.  At least we verified that worked, and this gave us confidence to come back and try again. The comb was working fine, but we identified a few ways to improve it.  And we needed to repair the fiber.

By the time we got more engineering time on IRTF in September 2014, I was in Pasadena for my last three days before starting my job as an assistant professor at Missouri State University.  I called in remotely via Skype to help with the commissioning.  We brought the comb back out to the telescope, this time with some further improvements.  Other members of our team that had gone in July were back on site atop Mauna Kea.  This time we had protected the fibers from breaking, and it worked!

Figure 4 from Yi et al. 2016.  In the upper right, you can see two simultaneous spectra.  The top spectrum is from the micro-resonator laser frequency comb, and the bottom is from the starlight of SV Peg, a bright K=0 red giant. 

Unfortunately, in September 2014 the weather was so bad, we barely collected a few hours worth of data at night time.  We had to look at the very bright red giant star SV Peg, because clouds were reducing the apparent brightness of stars by over a factor of 100!  So, the radial velocity measurements were not ideal.  But we demonstrated that the technology was feasible, repeatable, and that the spectrum of the microresonator comb could be captured simultaneously with starlight.  Several months later we would successfully test it again on a different spectrograph during the daytime - the NIRSPEC spectrograph on the Keck Telescope (one of the largest in the world) - showing the versatility of the comb and how "turn-key" the device was after months of being in storage.

From the NIRSPEC tests, we learned that the micro-resonator comb is at least good enough to measure radial velocities at a precision of 30 cm/s, with room for improvement in the future.  Currently, it's only limited by the stability of the reference laser we use to "lock" the wavelengths of the comb.  In the future, we may be able to get the device to span an octave in wavelength, and thus obtain a "self-referencing" lock, which will hopefully be stable to better than 1 cm/s.  With time, we believe this device will be more cost effective, and less complex, enabling routine use of the micro-resonator comb for the wavelength calibration of spectrographs in our search for another Earth.  During our last test it only required one highly skilled applied physicist to set it up, and we envision a future where these will be as easy to use as Thorium Argon lamps are today.

Here's the full content of the scientific paper describing our results:

http://www.nature.com/ncomms/2016/160127/ncomms10436/full/ncomms10436.html