You are here


, Updated


MAROON-X is high-resolution (R~80,000), optical (500-920nm), fiber-fed echelle spectrograph.

The instrument is designed to have the capability to detect Earth-size planets in the habitable zones of mid- to late-M dwarfs using the radial velocity (RV) method. MAROON-X is highly stabilized, both opto-mechanically (main optical elements in vacuum, the whole spectrograph inside an environmental enclosure with mK temperature stability), and via its fiber feed (octogonal and rectangular fibers, double-scrambler). A stabilized etalon is used to illuminate a calibration fiber to provide a dense comb of emission lines for simultaneous wavelength calibration.

Spectral range and resolution

MAROON-X has no movable parts and only one mode. The main disperser is a R4 echelle with 31.6 gr/mm. A dichroic splits the light into a ‘blue’ (full range: 500-678nm, order 122-91; useful range: 500-663nm, order 122-93) and a ‘red’ (654-920nm, order 93-67) arm. Each arm has a VPH grism cross-disperser and camera optics optimized for their respective spectral grasp.The two echellograms are shown below.


Figure B1. Left: Spectral format of the visual arm of MAROON-X (roughly true color); Right: Spectral format of the NIR arm of MAROON-X (false colors). Increasing wavelengths run from the lower left to the upper right in each image.

The resolving power is typically 80k-88k (see Figure B2), measured as the FWHM of an unresolved line in the extracted 1D spectrum. These measurements were obtained by fitting etalon lines from an observation during the commissioning run in May 2020 with a combination of a box and an asymmetric gaussian profile. The variability in the resolving power is entirely driven by the field- and color-depended aberrations of the two cameras and by the focus (incl. tip-tilt term) of the CCD detector. 

Resolving power of MAROON-X
Figure B2. Resolving power of MAROON-X for all three science fibers for the blue and red arm, respectively. The thin traces are points of individually measured etalon line profiles along each order for each fiber.The symbols show the mean resolving power within the FSR of each order.

Slit Format

The spectrograph simultaneously records the spectra of the star, the sky background, and light from a calibration source. The stellar light is sliced three times by a pupil slicer (for details see Components page) and fed into three rectangular fibers. The sky background light is reimaged onto a forth rectangular fiber. Light from a calibration source is injected into a fifth rectangular fiber. These five fibers form the entrance slit of MAROON-X (see image below). 

There is sufficient space in between the fibers and in between orders to keep any crosstalk well under 0.1%. Slit tilt and slit curvature are neglectable and the full resolving power is recovered without interpolation along the slit. The flux distribution between the three science fibers is roughly 30%, 42%, 28% of the total flux for fibers 2, 3, and 4 respectively.

Each fiber is sampled by 10 spatial pixels. To enable high-precision RV measurements, each resolution element is sampled on average by 3.2 pixels (varying from about 2.9 to 3.6 pix within the FSR of each order due to the anamorphic magnification of the echelle grating). A resolution element is thus sampled by almost 100 pixels (3 x 10 x 3.2).

Figure B3. Small cutout from a typical MAROON-X spectrum recorded with the blue arm (around 540nm) to illustrate how the five rectangular fibers are stacked to form a pseudo slit.


MAROON-X receives its light from a 100um octagonal fiber, fed at f/3.3. The FOV is thus 0.77” on sky. This is at the order of the IQ75-percentile seeing (0.75”) and makes the throughput of the instrument highly dependent on the seeing conditions. During the commissioning run in December 2019, we encountered different atmospheric conditions with seeing ranging from 0.7” but very unstable (MASS seeing >1”) to better than average (<0.5”) and stable. We have recorded peak efficiencies of up to 11.5% in the red channel for a few observations. However, most of the observations recorded at median conditions (0.4”-0.6” seeing) show a typical peak throughput of 8% in the red and 7% in the blue channel. This is lower than expected based on the instrument throughput measured in the lab and the nominal variability of the fiber coupling efficiency with the image FWHM (e.g., pure geometric losses would suggest 50% coupling efficiency for 0.75” seeing conditions). We are working on isolating the cause of the throughput shortfall and will work on improving the throughput for 2020B.

We thus advise adopting an 8% peak throughput estimate for semester 2020B.


Figure B4. Typical throughput of MAROON-X calculated from observations of a G0V star after airmass correction.


The sensitivity is canonically expressed as the limiting magnitude to reach a (read-noise limited) signal-to-noise ratio of one per resolution element in a one hour integration. Given the per-pixel read noise (3e-), the number of pixels per 1D spectral bin (3 fibers x 10 pixels each) and the number of 1D pixel bins per resolution element (~3.2 pixels near the blaze peak of each order), we find a limiting magnitude of Ic_limit = 19.5 mag for the peak of the red arm and V_limit = 19.7 mag for the peak of the blue arm.

Radial Velocity Precision

MAROON-X follows the instrument concept of high opto-mechanical and illumination stability pioneered by the HARPS instrument. This approach has now been adopted for all recent and future RV spectrographs. However, even the latest generation of highly stabilized spectrographs exhibit drifts exceeding m/s levels on timescales from hours to months due to various practical limitations in the instrumental design and in the environmental control of these instruments. Instrumental drifts need to be monitored and observations corrected accordingly. This requires a calibrator that provides a uniform spectrum with high local information density (many equally bright unresolved emission lines per order) and is long-term stable. MAROON-X is using a stabilized Fabry-Perot etalon illuminated by a white light fiber laser that provides a dense comb of emission lines with a spacing of 15 GHz and an unresolved line-width of 340 MHz. This translates to a line spacing of 2 - 3.6 resolution elements and a completely unresolved line-width of 1/20 to 1/10 of a resolution element between 500 and 920nm, respectively. 

Typical exposures taken with the etalon exhibit a photon noise domited rms of ~5 m/s for the centroid of the individual etalon comb lines (~400 lines per order). The zero point drift of each order can thus be determined to about 25 cm/s. This is more than one order of magnitude lower than the intrinsic RV content per order of a typical stellar spectrum with high SNR. We find no deviations from a smooth wavelength solution at the m/s level confirming the absence of CCD stitching effects.

A second, equally important factor is the illumination stability of the spectrograph, as changes in the illumination of the entrance slit directly translate into measured RV shifts. MAROON-X is using a combination of octagonal and rectangular fibers as well a double scrambler (exchanging near- and far-field) to suppress all spatial information in the light distribution originating from systematic guiding errors at the telescope. This effect is best characterized by observing stars that show no significant RV signal based on measurements from other instruments, i.e. so called ‘RV standards’.

Preliminary on-sky RV stability tests were conducted using observations taken during a commissioning campaign in December 2019. The data were reduced using standard procedures for echelle spectra and analyzed for their RVs using the SERVAL code (see Zechmeister et al. 2018, A&A 609, A12). The code derives a high SNR template by stacking the individual observations of the star and uses a forward-modeling approach to fit each individual observation by applying a RV shift to the template. This technique has been shown to deliver better results than the classical binary mask cross-correlation technique for complex, feature-rich spectra, such as M dwarfs.

In the December run, our quietest RV standard GJ908 (M0V) exhibits a nightly rms of 0.9 m/s over 25 observations (on average three observations per night over a 8 night campaign). See plot below. A preliminary analysis of the same dataset using the wobble algorithm (Bedell et al. 2019, AJ 158, 164) results in quantitatively similar RVs. 


Figure B5. Radial velocities for the ‘quiet’ M0V star GJ908 during a 8-night observing campaign in December 2019. The nightly rms is 0.9 m/s and is dominated by the first two nights. The rms of the last 5 nights is under 0.5 m/s. The individual error bars on the datapoints are the error in the mean obtained from 30 spectral orders in the blue arm and 26 spectral orders in the red arm and represent internal errors. The error of the nightly means is simply the rms of the datapoints. High and variable cloud coverage badly affected the SNR for observations during nights 5, 6 and 7, which explains their larger error bars and higher nightly spread. But no systematic RV shift was found with lower SNR and adverse seeing.

Other RV standards show stellar activity signals in our data and are thus of limited use in determining the intrinsic instrument stability. 

We note that during the December run two important parts of our drift correction scheme were not ready: 

  • We did not use the simultaneous calibration fiber but determined the drift of the spectrograph only from etalon exposures bracketing the science observations. This limits our ability to determine non-linear parts of the instrumental drift during the science observations. The simultaneous calibration fiber is now working and will be used for all future observations.
  • The independent stability tracking of our etalon calibrator against a Rubidium transition was not working in December. We have thus not corrected the etalon for any potential temperature drifts or known intrinsic variations. However, the monitored temperature of the etalon suggests stability at the cm/s level during the observing run. The known drift of the etalon from material shrinkage is well under 1/ms per week. The etalon tracking is foreseen to be functional for semester 2020B.

Acquisition and Guiding

MAROON-X has a frontend unit that converts the f/16 beam from the telescope to an f/3.3 beam to feed the 100um octagonal science fiber. Primary (i.e. telescope guiding) is using the PWFS2 of Gemini. The frontend unit is mounted on the bottom port of the ISS, a port shared with NIFS and NIRI.

In the frontend unit, the light passes through an atmospheric dispersion compensator (ADC) and a tip-tilt mirror provides flexure compensation between the telescope and the fiber face to provide < 0.1” average centroid stability. To achieve this, ~1% of the light that is sent towards the fiber is picked up by a beam splitter and imaged by a fast and sensitive acquisition and guide camera (sCMOS with >90% peak QE, 400-950nm passband). Three back-illuminated single-mode fibers surrounding the object fiber allow to track the apparent position of the science fiber during the integration. A slow (f<1Hz) guide loop is steering the tip-tilt mirror to keep the centroid of the star on the measured fiber position. Exposure times as short as 0.05s can be chosen. ND Filters allow a further reduction of the flux for bright stars. Exposure times of up to 30sec allow guiding on targets as faint as ~16mag. 

A second 100um octagonal fiber 20” from the object fiber is used to record the sky background.

The frontend also facilitates the illumination of the object and sky fibers with calibration light. Calibration light (flatfield, ThAr, etalon) is sent up from the spectrograph via an additional fiber integrated into the fiber bundle. The tip-tilt mirror is used to steer the light onto the object or the sky fiber.

Target acquisition: The target can be centered on the fiber in typically less than 1min after the PWFS2 guide loop has been closed by the telescope operator. Depending on seeing conditions, the PWFS2 loop needs to settle for a couple minutes before the exposure begins. If the true target position is off by more than about 5” from its specified position, the PWFS2 guide loop needs to be opened to send larger offsets to the telescope which slows down the acquisition process. The acquisition and guide camera has a vignetting-free FOV of about 60” in diameter and allows to monitor the position of the sky fiber for any contamination by field stars.

ADC: The ADC is designed to provide compensation to better than 50mas from 500-900nm down to airmass 2.5. The ADC can be used in three modes:

  • ADC off: no ADC correction. 
  • ADC fixed: The ADC is pre-set for the average airmass of the observation and kept static.
  • ADC tracking: The ADC updates during the observation to compensate for the actual airmass.

In all three modes, the telescope de-rotator can either be fixed (Alt-Az mode) or following (RA-DEC mode). Since the object fiber is on the nominal optical axis of the field, field rotation has no direct impact on the fiber coupling. A fixed telescope de-rotator offers potentially improved image quality through more stable aberration compensation of the telescope. The added advantage is that the parallactic angle is constant and the ADC does not have to follow the de-roator. The only downside is that in a crowded field, background stars might rotate over the sky background fiber and contaminate the sky background spectrum. However, we strongly advise to keep the telescope de-rotator fixed whenever possible.

Guider images: Guider images can be saved if the user requests this option.