NIRSpec MSATA Reference Star Selection Recommended Strategies

The NIRSpec MSA Target Acquisition (MSATA) observes a set of stars (typically 5-8 objects) through the open Micro-Shutter Assembly (MSA), in order to correct for JWST's pointing and roll uncertainty.

On this page

Words in bold italics are buttons 
or parameters in GUI tools. Bold 
style represents GUI menus/
panels & data software packages.

This type of TA is the only way to ensure that science targets observed in NIRSpec's Multi-Object Spectroscopy (MOS) mode are precisely aligned in their 0.2" by 0.46" MSA shutters. This article discusses strategies for identifying suitable MSATA reference stars during the program update stage of the MOS and MSATA process. While MSATA cannot be fully defined at proposal submission (prior to orient assignment by STScI), it is important to consider whether it is feasible with presently available imaging. If MSATA cannot be carried out with existing data, NIRCam pre-imaging should be proposed alongside the NIRSpec MOS observations.

A note about terminology:

Guide star acquisition with a guide star in the Fine Guidance Sensor is different from target acquisition which this article describes. Guide star acquisition occurs at the start of each visit, delivering an expected 1-sigma radial pointing accuracy of  0.1" (100 mas). Following that, MSATA can be used to place the science sources in the small shutters of the MSA.

MSATA, uses reference stars in the MSA FOV, which are different from the single guide star used to more coarsely point the telescope during guide star acquisition.

Requirements for a successful MSATA

There are a number of requirements that must be met for MSATA to successfully align MOS targets in their respective MSA shutters. These requirements are described in detail in the NIRSpec MSA Target Acquisition article. In summary:

Astrometric Accuracy

The ideal relative astrometric accuracy for MSATA is 5-10 mas in a catalog that covers the reference stars and science targets. This relative accuracy is expected to result in TA accuracy of 20 mas, or 1/10th of a MSA shutter width. Relaxed accuracy is allowed, however, if the science case permits it. In general, MSATA will require space-based imaging, either with NIRCam, or from HST in approximately the last ten years. Stars in older HST images may have proper motions that will compromise MSATA

NIRSpec Magnitudes

NIRSpec's MSATA is carried out in one of the three TA bands, with the option to choose from a number of readout modes and patterns. The MSATA algorithm will take two exposures, with a half-shutter dither in both dispersion and cross-dispersion directions, in order to mitigate the effect of the MSA bars. Each exposure uses three groups and one integration, leading to a fixed sensitivity for each filter/readout pattern. Table 1 shows the ranges of magnitudes in the NIRSpec TA bands that will achieve S/N > 20, while also avoiding saturation. It is worth noting that any stars detected in 2MASS will saturate in NIRSpec's MSATA.

Table 1. Brightness ranges for NIRSpec MSATA filter and readout pattern options

Readout patternS/N = 20Saturation






Proposers should consider that no catalogs currently have, or will have, the exact magnitudes measured in the NIRSpec bands. However, provided that some infrared magnitudes exist, it should be reasonable to estimate NIRSpec magnitudes with sufficient accuracy. The large magnitude ranges between S/N = 20 and saturation in Table 1 implies that the accuracy of the NIRSpec magnitude estimates does not have to be particularly high. As long as the inferred magnitudes are not near the edges of the magnitude bins, there is room for uncertainty in their estimates.


The MSATA algorithm works by examining 3.2" boxes around the expected positions of the reference stars. Then, a 3 x 3 pixel box (0.3" x 0.3")  is passed over the image to find the brightest pixel and centroid the source. It follows that MSATA requires isolated stars: TA reference stars must be the brightest object within 3.2", in the chosen NIRSpec TA filter. Likewise, fainter companions within 0.3" should also be avoided, as they can skew the centroid. For the final submission of flight-ready program updates, it is highly recommended that images of the selected reference stars are examined by a human, in order to avoid close binary stars or other artifacts that might be missed by automated catalog preparation tools.  

In cases where the density of stars is too high to identify suitably isolated MSATA reference stars, MSATA is not possible. Instead, the recommended strategy is to step the MSA configuration in the dispersion direction, taking science spectra at enough positions to ensure that the desired targets fall in the slitlets at one of the planned pointings. Users interested in this approach should consult the NIRSpec Target Acquisition Recommended Strategies article to determine the required dispersion-direction spatial coverage that will account for JWST's pointing uncertainty.  More information about slitlet stepping can be found in the article on NIRSpec MOS Recommended Strategies.

Density of MSATA Reference Objects

While MSATA typically uses 5-8 reference stars, a larger number of stars in the MOS footprint are needed since many cannot be used. For the optimally chosen science pointing, reference stars may be excluded because they fall behind the mounting plate between the MSA quadrants, or they are too close to a failed closed MSA shutter. Tests in the GOODS-S region, using the MSA Planning Tool (MPT), indicate that around 25-30% of TA reference objects in the FOV, and in a given magnitude bin (see Table 1), can be successfully used for MSATA. In other words, for a chosen readout pattern and filter in Table 1, the density of reference objects should be approximately 2 arcmin-2 or higher. This density would predict 24.5 reference objects in the 3.6 x 3.4 arcmin MSA FOV (including behind the MSA mounting plate); a 25% success rate would yield around 6 useful reference objects. Nonetheless, fluctuations around these estimates are expected, due to small number statistics and spatial variations in the distribution of stars.    

These densities of stars are not always available at high Galactic latitude. For example, in the GOODS-S extragalactic survey field, the density of stars reported in the Skelton et al. (2014) catalogs is a few times too low to reach the 5 star minimum in any of the magnitude ranges in Table 1. Hence, the usage of compact galaxies will be essential for MSATA in many cases (indeed, compact galaxies were used to estimate the success fraction calculated above). While TA has not yet been tested or simulated for non-point sources, given NIRSpec's relatively large pixel scale (100 mas), it is anticipated that compact galaxies will be adequate for MSATA.

    MultiExcerpt named 'aptgui' was not found
The multiexcerpt named 'aptgui' was not found. Please check the page name and MultiExcerpt name used in the MultiExcerpt Include macro.

When do I need to request NIRCam pre-imaging? 

There are a number of scenarios that may arise where it is not straightforward to decide whether NIRCam imaging should be proposed. Here, we offer some guidance on specific cases.

I do not have any HST or NIRCam imaging of my field:

Pre-imaging is highly recommended. Even if the TA reference stars have high precision astrometry, e.g. from GAIA, the relative astrometry between the science targets and the TA reference stars is critical. An offset between the relative astrometry of the MSATA reference stars and the science targets would cause the science targets to fall outside their intended MSA shutters. Likewise, at ground-based resolution, the centroid positions of the science targets may not be accurate enough to ensure that they are well-located in the MSA shutters.

I have HST imaging, but it is more than 10 years old:

Because the magnitude ranges for MSATA in Table 1 are relatively faint, MSATA is sensitive to nearby cool stars. As a result, high proper motion is a concern. In GAIA, proper motions of 1-2 mas/year are typical for stars around 20th magnitude. Consequently, we recommend that HST imaging used for MSATA reference stars be less than about 10 years old. Stars with larger proper motions are also commonly seen:  GAIA reports an RMS scatter reaching a few to several mas/year for faint stars. However, since MSATA uses 5-8 stars, there is some built in redundancy to allow for individual reference stars to fail. Regardless, if the only available HST imaging is more than 10 years old, NIRCam pre-imaging will likely be the safest strategy.

I have recent HST imaging at optical wavelengths, but no infrared imaging:

At Galactic latitudes where foreground extinction is low, or at least well-characterized, a single optical color can predict the infrared magnitudes with sufficient accuracy for MSATA. Therefore, if at least two bands of optical imaging are available, NIRCam pre-imaging is not necessary. Below, we discuss an algorithm for predicting the NIRSpec magnitudes. We provide reference tables of theoretical and empirical star colors, and an iPython notebook. However,  we do not recommend this method for regions where the extinction is very high, spatially variable, and not always well known. In that case, NIRCam pre-imaging would be recommended.

I have recent HST imaging in the optical, and deep ground-based near-infrared imaging:

NIRCam imaging should not be necessary, provided that the desired science targets are detected in the HST imaging. However, heavily extincted regions may have objects that are not detected at any optical wavelengths. In this case, despite the fact that the positions and infrared magnitudes of the TA reference stars are well-known, the positional accuracy of the science targets is likely inadequate for placing them in the MSA shutters. Hence, NIRCam pre-imaging would be recommended.

I have recent HST imaging, but it does not contain very many stars:

In science fields that have low extinction, whether or not NIRCam imaging will identify more stars will depend on the depth of the HST imaging. In some cases,  deeper pre-imaging might help to find more stars. But in other cases, the HST imaging will already be sufficiently deep, and NIRCam will only uncover new stars fainter than the MSATA magnitude limit. (This happens to be the case for the Hubble Ultra Deep Field). Fortunately, compact galaxies should be sufficient for MSATA when there are not enough stars. However, if the HST imaging is at optical wavelengths only, NIRCam imaging is still likely required to estimate the NIRSpec magnitudes of the compact TA reference galaxies. Unlike stars, where a optical colors can predict the NIRSpec magnitudes, galaxies with unknown redshifts are considerably more difficult to extrapolate to the observed-frame near-infrared.

Predicting the NIRSpec TA-band magnitudes of stars from optical imaging

The colors of field stars with low or constant extinction form a tight sequence, with minimal scatter, as illustrated by Figure 1. This figure uses stars from the five fields of the Cosmic Assembly Near Infrared Deep Extragalactic Legacy Survey (CANDELS; Koekemoer et al. 2011, Grogin et al. 2011), demonstrating that it is possible to predict the NIRSpec TA-band magnitudes with precision that is sufficient for MSATA when the Galactic foreground extinction is low. As a result, with some analysis, it should be possible to avoid the need for NIRCam pre-imaging in high-latitude fields that lack near-infrared photometry. Here, we provide some guidance on making these predictions, as well as empirical and theoretical tables of stellar colors, and an iPython notebook to aid in this analysis. 

Figure 1.  The optical colors of stars can be used to predict NIRSpec TA-band magnitudes

The colors of stars in the Galaxy form a tight sequence when the foreground extinction is low. Here, the stellar photometry is taken (or derived) from the five CANDELS fields, as compiled by Skelton et al. (2014). Since the photometric bandpasses used differ between the five fields, the photometry shown here is a mix of observed measurements, and SED-fit based interpolations for the ACS/WFC F606W and F814W bands. The F140X band is the NIRSpec TA bandpass, is a model-based interpolation for all galaxies in this plot. Because galaxies at a given F606W - F814W occupy a relatively small locus in F814W - F140X, it is possible to estimate the F140X magnitude from measurements in F606W and F814W alone.

Strategy for predicting NIRSpec TA-band magnitudes from optical colors:

Both empirical and theoretical tables of stellar colors can be used to select samples of stars with optical colors similar to a star that is being considered for MSATA. A given sample of stars will show a distribution in their optical-to-infrared colors, which gives a statistical measurement of the likelihood that the reference star will fall in an appropriate MSATA magnitude range (Table 1). For example, a reference star might have HST/ACS F606W  = 23.1 ± 0.05 and F814W = 22.0 ± 0.05. From the CANDELS catalog of stars shown in Figure 1, we can select objects that match this color, having F606W - F814W = 1.1 ± 0.07. These stars have a color in ACS - NIRSpec, F814W - F140X = 0.51, with a scatter of 0.06 mag, so we infer that the NIRSpec F140X magnitude of this particular TA reference star is 22.5 ± 0.07. Of the 89 stars in Figure 1 that are consistent with this F606W-F814W color, the calculations show that 100% fall within the appropriate magnitude ranges for NRSRAPID and NRSRAPIDD6 with F140X (Table 1). Hence, we can conclude with high confidence that the star under consideration has a suitable magnitude for MSATA. It is straightforward to extend this technique to predict the NIRSpec TA-band magnitudes from multiple colors instead of just one.

The advantage to this empirical approach is that the stars in CANDELS reach similar magnitudes to what is used for MSATA. At these magnitudes, even small area surveys are likely to uncover cool, red stars (Ryan & Reid 2016), which are few in number at shallower depths, even in the area of the entire SDSS (e.g. Metchev et al. 2008). Hence, the colors of stars used for MSATA should be reasonably well matched to the colors of stars seen in CANDELS. However, the disadvantage to using stars in high Galactic latitude survey fields like CANDELS is that--with relatively dust-free sight-lines--this database is not applicable to regions of higher Galactic foreground extinction. Therefore, in order to include Galactic foreground extinction, we are also providing theoretical distributions of stars in the Milky Way, with Galactic foreground extinction applied as a function of their simulated distance. These models, given below, are calculated using TRILEGAL 1.6 (Girardi et al. 2012).

Tables of Stellar Colors and Example Calculations:

Empirical Method: Stars in CANDELS

In order to compute an empirical database of stellar colors, we take stars from the CANDELS photometric tables that were compiled by Skelton et al. (2014), which are available from the 3D-HST project. These stars reach a depth of 25th magnitude (AB) in WFC3/F160W, so they are well-matched in brightness to the MSATA reference star magnitude bins. In order to predict the magnitudes of the stars in bands that are not used by CANDELS, we fit Phoenix model spectra (Allard et al. 2016) to these stars, using the spectral energy distribution (SED) fitting tool Chorizos (Maíz-Apellániz 2004). Since the imaging coverage of the CANDELS fields is inhomogeneous, we also used these best-fitting SEDs to calculate synthetic magnitudes predicted by the fit, for the full set of optical and near-infrared bands in the various CANDELS observations. As such, the ACS F606W and F814W photometry in Figure 1 is a mixture of observed and predicted magnitudes, since these bands were not observed in all five of the CANDELS fields.  

The lookup table that we have derived for the stars in CANDELS can be downloaded here. The table includes:

  • The ID of the source, including the field name and the ID from Skelton et al. (2014)
  • Parameters from the Chorizos SED fits; stellar effective temperature (Temperature), stellar surface gravity (log_g), stellar metallicity (Z), and foreground extinction (EBV; constrained to be < 0.1). Errors are also included (sigma_Temperature, sigma_log_g, sigma_Z, sigma_EBV).
  • Residuals from the fit for the 3 worst photometric points for each object (color_uncert, color_resid, and color_normerr). The quantities color_uncert are uncertainties in colors used by the fit, color_resid are the residuals from the fit, and color_normerr is the ratio of the two. This information can also be calculated (for all bands) from the observed and predicted photometry that is included in the table.
  • The χ2 and number of degrees of freedom in the fit (chisq, ndf) from Chorizos
  • Photometry in more than sixty bands, in AB magnitudes, including the observed magnitudes, the measured error, and a prediction from integrating the best-fitting SED under the desired bandpass (e.g. acs_f435w, acs_f435w_err, acs_f435w_pred). When a desired photometric band is not observed in a particular field, the measurement and the error are set to -1, but predicted magnitude is still included. Note that for NIRSpec and NIRCam magnitudes, only the predicted magnitude columns are included, since no observations exist yet.

We assessed the accuracy of the NIRSpec TA-band predictions by comparing the observed and SED-fit predicted magnitudes in WFC3/IR F125W, F140W, and F160W, as well as the IRAC 1 (3.6 μm) and IRAC 2 (4.5 μm) channels. For the WFC3/IR bands, which cover similar wavelengths as NIRSpec's F110W and F140X, we find no systematic difference between the observed and predicted magnitudes, and an RMS of 0.05-0.06 magnitudes. We take this to imply a similar level of accuracy for NIRSpec's F110W and F140X TA bands. The IRAC bands, on the other hand, show worse residuals, with systematic offsets around 0.1-0.2 magnitudes, and scatter of a similar amount. Hence, we expect that the prediction of the flux in NIRSpec's CLEAR band is likely 0.1-0.3 magnitudes too bright for the ~2.5-5 μm. Since this inaccuracy only impacts around half of the wavelengths covered by the CLEAR band, we iner that the CLEAR magnitude predictions are somewhat worse than the WFC3/IR bands but not as bad as the IRAC magnitudes. These uncertainties should be considered when using CLEAR magnitudes that are predicted to be near the edges of the MSATA reference star magnitude ranges in Table 1. Note that the lookup table for CANDELS stars also includes predicted magnitudes for all medium and broadband filters for NIRCam; the longer wavelength NIRCam predictions are subject to the same uncertainties.

Theoretical Method: Stellar population models from TRILEGAL

The same strategy can be applied using a simulated database of stellar colors. In this way, it is possible to select a simulation with a non-zero Galactic foreground extinction, in order to match a chosen science target. A set of simulated stellar colors, created with TRILEGAL 1.6 (Girardi et al. 2012), can be downloaded below. Each simulation predicts the magnitudes of around 1400 stars in an area of 0.1 square degrees, to limiting magnitude of J=27.5 mag, in the SDSS ugriz + 2MASS photometric bands. Stellar binaries are included in the simulated stellar population. These models presently include foreground extinction values of Av = 0.0, 0.1, ...5.0, 6.0, ...10.0; additional extinction values can be simulated using the TRILEGAL web interface. An ipython notebook, showing how to predict the Ks-band magnitude of an individual star from its SDSS r- and i-band magnitudes is also provided below. Transformations between J,H,and Ks bands and NIRSpec's TA bands should be small relative to the magnitude ranges in Table 1, and rough estimates can be made from the empirical CANDELS database given above. The NIRSpec TA-bands will be included in the TRILEGAL-simulated stellar colors in advance of Cycle 1 operations, when MSATA flight-ready updates are due to STScI. 

The TRILEGAL models can be used to determine how increasing Galactic foreground extinction increases the uncertainty in the predicted near-infrared magnitudes. Figure 2 shows an example for Galactic extinction of AV = 1.0 (gold points), comparing to the case with no foreground extinction (blue points). In the former case, the extinction seen by the stars varies between AV = 0.54 - 1.0, depending on the distance to the star. For the simulation with higher Galactic extinction, when r-i = 1.6 ± 0.07,  the 95% confidence interval on the inferred Ks-magnitude spans 0.63 mags. This uncertainty is larger than the uncertainty for the case with no extinction. In cases like this, with larger uncertainties on the inferred infrared magnitudes, candidate MSATA reference stars should be excluded if their predicted magnitude is close to the edge of the magnitude bins listed in Table 1. As with the empirical approach, multiple optical colors, if available, can be used to select a sample most closely matched to the MSATA reference star under consideration. 

Downloads for the Theoretical Method:

Figure 2.  Using Theoretical Models to Predict Infrared Magnitudes

Simulated stellar colors, created with TRILEGAL 1.6 (Girardi et al. 2012), show a tight sequence, implying that infrared magnitudes can be predicted from one or more optical colors. The left two panels show the same color-color plot, covering different ranges in their r - i and i - Ks colors. The right-most panel shows a histogram of the i-Ks magnitudes for stars that match a candidate MSATA reference star with r = 25.0 ± 0.03 and i = 23.40 ± 0.03. Blue markers correspond to no Galactic foreground extinction, while gold indicates AV = 1.0. When AV = 1.0 for the full column of Galactic dust, the stars simulated by TRILEGAL have a range of AV = 0.54 - 1.0, with nearby objects experiencing less extinction. This variation gives rise to more scatter in the i - Ks colors at a given r-i, leading to larger uncertainties in the predicted infrared magnitudes.


Allard, F. 2016, SF2A-2016: Proceedings of the Annual Meeting of the French Society of Astronomy and Astrophysics, 223
The Phoenix Model Atmosphere Grid for Stars

Girardi, L., et al. 2012, Astrophysics and Space Science Proceedings, 26, 165
TRILEGAL, a TRIdimensional modeL of thE GALaxy: Status and Future

Grogin, N. A., Kocevski, D. D., Faber, S. M. 2011, ApJS, 197, 35
CANDELS: The Cosmic Assembly Near-infrared Deep Extragalactic Legacy Survey

Koekemoer, A., Faber, S. M., Ferguson, H. C., et al. 2011, ApJS, 197, 36
CANDELS: The Cosmic Assembly Near-infrared Deep Extragalactic Legacy Survey—The Hubble Space Telescope Observations, Imaging Data Products, and Mosaics

Maíz-Apellániz, J. 2004, PASP, 116, 859
CHORIZOS: A χ2 Code for Paremterized Modeling and Characterization of Photometry and Spectrophotometry

Metchev, S. A., Kirkpatrick, J. D., Berriman, G. B., & Looper, D., 2008, ApJ, 676, 1281
A Cross-Match of 2MASS and SDSS:  Newly Found L and T Dwarfs and an estimate of the Space Density of T-Dwarfs

Ryan, R. E., Jr. & Reid, I. N. 2016, AJ, 151, 92
The Surface Densities of Disk Brown Dwarfs in JWST Surveys

Skelton, R., et al. 2014, ApJS, 214, 24
3D-HST WFC3-selected Photometric Catalogs in the Five CANDELS/3D-HST Fields: Photometry, Photometric Redshifts, and Stellar Masses

Latest updates
    Minor formatting fixes, and added the Publication date and Latest Updates box.
Originally published