The most distant star known

Jan 27, 2020
Seen as it was 13 billion years ago


Earendel (WHL0137-LS) is an individual star / multiple star system (binary / trinary...) that we are able to see very far away thanks to gravitational lensing. Normally at these distances, the smallest features we see are star clusters (hundreds of stars) within a galaxy. The lens and Earendel are perfectly aligned, magnifying its image by a factor of thousands.

Earendel resides in a galaxy known as the Sunrise Arc, the most highly magnified galaxy yet known in the first billion years. It is at redshift z ~ 6.2, and it light has taken 13 billion years to reach us. Earendel and the Sunrise Arc are gravitationally lensed and magnifed by the galaxy cluster WHL0137-08 (z = 0.566).

Brian Welch discovered Earendel in Hubble images (Welch et al. 2022a) and also studied the Sunrise Arc (Welch et al. 2022b), originally discovered by Brett Salmon et al. (2020) in the RELICS survey.

New JWST images reveal more details about Earendel, including its colors observed in 8 NIRCam filters spanning 0.8 – 5.0 µm. Brian Welch led the new analysis out now on the arXiv (Welch et al. 2022c)!


JWST Imaging of Earendel, the Extremely Magnified Star at Redshift z = 6.2
Excerpts of the original paper appear below:

Here we present James Webb Space Telescope (JWST) Near Infrared Camera (NIRCam) images of Earendel in 8 filters spanning 0.8–5.0μm. In these higher resolution images, Earendel remains a single unresolved point source on the lensing critical curve, increasing the lower limit on the lensing magnification to μ > 4000 and restricting the source plane radius further to r < 0.02 pc, or ∼ 4000 AU. These new observations strengthen the conclusion that Earendel is best explained by an individual star or multiple star system. The new imaging also supports the previous photometric redshift estimate, solidifying the interpretation of Earendel as a stellar source within the first billion years of the universe. Fitting stellar spectra to our photometry yields a stellar temperature of Teff ≃ 13000–16000 K assuming the light is dominated by a single star. The delensed bolometric luminosity in this case ranges from log(L) = 5.8– 6.6 L⊙, which is in the range where one expects luminous blue variable stars. Allowing for two stars with different temperatures can yield better fits to the photometry at the expense of increased free parameters, leading to non-unique well-fitting temperatures of Teff 􏰀 20000 K and Teff 􏰁 13000 K for the hotter and cooler components, respectively. Follow-up observations, including JWST NIRSpec scheduled for late 2022, are needed to further unravel the nature of this object, which presents a unique opportunity to study massive stars in the first billion years of the universe.

Massive galaxy clusters magnify the distant universe through strong gravitational lensing. These cosmic telescopes provide improved spatial resolution over what cutting-edge telescopes can provide alone, allowing the identification of small-scale structures in high redshift galaxies (e.g., Welch et al. 2022a; Vanzella et al. 2022; Meˇstri ́c et al. 2022). In certain cases of precise alignment, galaxy clusters can magnify the light from individual stars by factors of thousands, allowing these stars to be seen above the light of their host galaxies. The first of these were discovered as transients in images from the Hubble Space Telescope (HST), at redshifts ranging from z ∼ 1 − 1.5 (Kelly et al. 2018; Rodney et al. 2018; Chen et al. 2019; Kaurov et al. 2019). Recent discoveries have pushed lensed star observations to greater distances, including recent discoveries at z = 2.37 (Diego et al. 2022), another at z = 2.65 with the James Webb Space Telescope (JWST) (Chen et al. 2022), and a star at z = 6.2 discovered in HST imaging (Welch et al. 2022b).

JWST (Gardner et al. 2006), which has recently completed commissioning and begun science operations (Rigby et al. 2022), will continue improving our ability to study distant lensed stars in detail. Besides already discovering new lensed stars (Chen et al. 2022), JWST will enable more detailed study of the highest redshift lensed stars. The combination of this powerful new ob- servatory and gravitational lensing could also be our best chance at observing Population III stars directly (Windhorst et al. 2018).

Earendel was first identified in HST imaging taken as part of the Reionization Lensing Cluster Survey (RELICS; GO 14096 Coe et al. 2019) and a follow-up program (GO 15842, PI Coe), as described in Welch et al. (2022b). We recently obtained additional imaging from the newly commissioned JWST NIRCam instru- ment as part of Cycle 1 GO program 2282 (PI Coe). These images span a wavelength range of 0.8–5 μm in eight filters, presented in Table 1. A color image of the Sunrise Arc hosting Earendel is shown in Figure 1, and image stamps of Earendel in each filter are shown in Figure 2. Each filter was observed for a total of 2104 seconds of exposure time. We utilized four dithers to cover the 5′′ gaps between the short wavelength (SW; λ < 2.4μm) detectors, as well as improving the resolution of our final drizzled images and minimizing the im- pact of image artifacts and bad pixels. Additional imaging in four filters (F090W, F115W, F277W, F356W) and NIRSpec spectroscopy for GO 2282 is expected in December 2022.

Screen Shot 2022-08-24 at 5.23.32 PM.png
Figure 1. JWST NIRCam image of the z ∼ 6.2 Sunrise Arc, including the lensed star Earendel marked with an arrow. This 15. 2 × 12. 4 color image combines all 8 NIRCam images on a 0. 02 pixel scale.

We processed the JWST Level 2 data products using the grizli pipeline4* (Brammer & Matharu 2021). This processing pipeline reduces striping from 1/f noise and masks “snowballs” in the images. All images are then aligned to a common WCS registered to GAIA DR3 (Gaia Collaboration et al. 2021). The pipeline next drizzles the images to a common pixel grid of 0. 04
per pixel using the astrodrizzle software (Koekemoer et al. 2003; Gonzaga et al. 2012; Hoffmann et al. 2021). The short wavelength NIRCam images are drizzled to a higher resolution grid of 0.02 per pixel, aligned to the lower resolution grid (with each low-resolution pixel corresponding to 2 × 2 high-resolution pixels)

Sources are then detected in a weighted sum of the drizzled NIRCam images in all filters using a Python implementation of SourceExtractor called SEP (Barbary 2016; Bertin & Arnouts 1996).


* grizli pipeline4 - Much of the background related to this question in the context of the currently available software tools was discussed in a document by Brammer, Pirzkal and Ryan (2014). Along with a detailed description of the format of the configuration files originally developed for the aXe software, we provided a compact Python script to address exactly the question above and strip away all of the many layers of bookkeeping, file-IO, etc. in existing analysis pipelines such as aXe (Kummel et al. 2009) and “THREEDHST” (Brammer et al. 2012, Momcheva et al. 2015). In fact, that relatively simple script serves as the low-level kernel for the way grizli computes the grism dispersion.

Eventually, grizli is intended to encourage and enable general users to move away from simple “data reduction” (e.g., extracting a 1D spectrum of a single object akin to standard slit spectroscopy) and toward more quantitative and comprehensive modeling and fitting of slitless spectroscopic observations, which typically involve overlapping spectra of hundreds or thousands of objects in exposures taken with one or more separate grisms and at multiple dispersion position angles. The products of this type of analysis will be complete and uniform characterization of the spectral properties (e.g., continuum shape, redshifts, line fluxes) of all objects in a given exposure taken in the slitless spectroscopic mode.

  • Grizli-Pipeline : End-to-end processing of WFC3/IR data.
    1. Query the MAST archive and automatically download files
    2. Image pre-processing (astrometric alignment & background subtraction)
    3. Field contamination modeling
    4. Spectral extractions
    5. Redshift & emission line fits (multiple grisms)
  • Fit-with-Photometry : Demonstrate simultaneous fitting with grism spectra + ancillary photometry
  • NewSpectrumFits: Demonstration of the lower-level fitting tools
    1. Unify the fitting tools between the stacked and exposure-level 2D cutouts (“beams”).
  • Fit-Optimization (09.14.17): Custom fitting (hasn’t been tested recently)
    1. Demonstrate some of the workings behind the fitting wrapper scripts by defining custom model functions with parameters to optimize.
The notebooks below are deprecated and haven’t been tested against the master branch since perhaps late 2017.

  • Grizli Demo: Simple interaction with WFC3/IR spectra
  • Basic-Sim (5.5.16): Basic simulations based on single WFC3/IR grism and direct exposures
  • multimission-simulation (5.11.16):
    1. Demonstration of more advanced simulation techniques using deep image mosaics and external catalogs/segmentation images as reference.
    2. Provide a comparison between dispersed spectra from WFC3/G141, JWST/NIRISS and WFIRST.
  • WFC3IR_Reduction (9.6.16): End-to-end processing of WFC3/IR data.
    1. Pre-processing of files downloaded from MAST (astrometric alignment & background subtraction)
    2. Field contamination modeling
    3. Spectral extractions
    4. Redshift & emission line fits (multiple grisms)
  • NIRISS-simulation (11.11.16): Simulation and analysis of JWST/NIRISS observations
    1. Simulate NIRISS spectra in three blocking filters and two orients offset by 90 degrees.
    2. Simulation field taken from the Hubble WFC3/IR Ultra-Deep Field, with the WFC3 F140W image as the morphological reference and photo-z templates taken as the spectral models.
    3. Extract spectra and fit redshifts and emission lines from the combined six exposures.


In the case of the star named Earandel, or (WHL0137-LS), the James Webb Space Telescope (JWST) has further added information on its characteristics since it was first identified by the Hubble Space Telescope (HST) by Brian Welch in the Hubble images (Welch et al. 2022a) and also studied the Sunrise Arc (Welch et al. 2022b), originally discovered by Brett Salmon et al. (2020) in the RELICS survey.
To calculate the distance of z=6.2, use the below chart for an estimate:


Amazing times as we look back farther into the age of the cosmos. My thanks, as always, to NASA, the JWST and the Hubble teams and to JPL.