NIRCam Scattered Light Artifacts

Details about scattered light artifacts in NIRCam are provided in this article. 

On this page

Summary of scattered light artifacts

Details about each artifact are provided in the sections below.

Dragon's breath type I artifacts are due to bright sources just off the edge (within 2") of the NIRCam field of view scattering light onto the NIRCam detectors.

Dragon's breath type II artifacts are caused by bright sources ~12" from the detectors in the field of view. They are only seen in short wavelength images.

Claws are artifacts created by an extremely bright star (K ~< 3 Vega mag) ~10° from the target observation in the +V3 direction.

Wisps are stationary features that always appear in the same detector location for a given filter. They are most prominent on the B4 detector, but are also detected at a lower level on the A3, A4, and B3 detectors. During data analysis, they may be subtracted by a template that has been developed and is available in the Wisps section.

Ginkgo Leaf artifacts have been observed in long wavelength module A images due to a star out of the field to the left (+V2) and ~24.5" below the top edge (in V3) of the detector A5.

Tadpoles are artifacts in NIRCam WFSS data that may be mistaken for emission line galaxies. They are most prominent and ubiquitous in grism C module B data.

Shells are fainter WFSS artifacts that appear occasionally due to scattered light from very bright sources.

Dragon's breath

Bright sources off the edge of the JWST NIRCam field of view scatter light onto the NIRCam detectors creating image artifacts. The first of these to be identified was named dragon's breath. It is caused by stars ~2" from the field of view. Since then, other NIRCam scattered light artifacts have been identified, including a stronger NIRCam Dragon's Breath Type II due to bright objects further away (~12").

Dragon's breath (type I) is caused by light scattering from the inner wall of the mask immediately in front of the focal plane array. Its effects can be minimized by ensuring that bright objects are not located just outside the field of view.

In the long wavelength (LW) channel, the integrated intensity of the scattered light is up to ~1.5% of the total intensity from an equivalent in-field source, with the peak pixel including up to ~0.75% of the source intensity. The total intensity of the scattered light depends on the distance of the source from the edge of the detector, with the strongest effect occurring when the source is ~1.9" from the detector edge in the LW channel. The light scatters about 200–250 pixels into the field, with the peak intensity occurring within ~50 pixels (1 LW pixel = 0.063″).

The effect is significantly smaller in the short wavelength (SW) channel, where the integrated intensity of the scattered light is less than ~0.4% of the total intensity of the equivalent in-field source, with the peak pixel including ~0.01%.

Figure 1. Dragon's breath

Click on the figure for a larger view.

This is an example of dragon's breath, including the diffraction spikes. Left: a short wavelength channel dragon's breath, for a source placed 2.1″ from the frame edge. Right: a long wavelength channel dragon's breath, for a star placed ~1.7″ from the frame edge. Both images are 512 × 506 pixels; since the pixels are twice as large in the LW channel, the angular extent of the LW scattered light is larger than it is in the SW channel.

Avoidance zones  

The dragon's breath avoidance zones are regions around the perimeters of all 10 NIRCam detectors. In the LW channel, dragon's breath is strongest for sources positioned ~1.9″ from the detector edge and for those within ±0.8″ from that position, forming an avoidance annular region around each detector. In the SW channel, the avoidance zone is centered at ~2.1″ from the detector edge and has a width of ±0.5″, including the region between SW detectors in each module. 

Figure 2. Avoidance zones

Click on the figure for a larger view.

A zoomed-in view of the upper part of the module A detectors, showing the avoidance zones for the SW channel (blue) and the LW channel (red). The avoidance zones are the hatched regions, with ±0.5″ (±0.8″) width for the SW (LW) detectors.

Dragon's breath type II

Scattered light from bright stars ~10"–12" from the short wavelength detectors can produce an artifact called type II dragon's breath. Discovered during flight, it is similar to the original dragon's breath artifact identified in ground testing and caused by stars ~2" from the field of view of the detector

Type II dragon's breath was first observed in JWST commissioning program 1165 in module B images. It was characterized further by NIRCam commissioning program 1067 in module A. Avoidance zones have yet to be fully mapped.

Type II dragon's breath is caused by light scattering off a knife edge installed to block a stray light path to the short wavelength detectors. This edge is not present in the long wavelength light path, and the artifact is not seen in long wavelength images.

The artifact spans the length of 2 short wavelength detectors. Half resembles a diffraction spike extending roughly along the V2 axis (rotated 3°). The other half is a bright glint ~6" wide, containing ~2% of the flux of the star. The bright glint has been observed most prominently in short wavelength detectors B3 and A4 due to stars ~12" from the edge of the field of view of detectors B1 and A2, respectively. Less severe glint has also been observed in detector A1.

Figure 3. Type II dragon's breath

Click on the figure for a larger view.

This type II dragon's breath observed during commissioning is due to a bright star (K ~ 9 Vega mag) 12" from the edge of the B1 detector (in the negative V2 direction). The bright glint shown here is 6" wide, 45" long in B3 plus another 10" in B1, and contains ~2% of the flux of the star. A gap of a few arcseconds in B3 is a shadow due to stray light being blocked by the FPA mask. Another linear feature spans the length of B1 pointing back to the star. It resembles the two PSF diffraction spikes also pointing to the star. The full B3 and B1 detectors are shown here, each 64" across. This image is from commissioning program 1165, observation 3. Type II dragon's breath has been observed in other short wavelength detectors, produced by stars 10"–12" away.


A type of scattered light phenomenon (Figures 4 and 5), known as claws, is observed in NIRCam when an extremely bright star (K ~< 3 Vega mag) lands in a specific susceptibility region. This region is located very far from the actual instrument field of view, at about +10° in the V3 direction, a huge distance compared to the ~2' effective extent of the NIRCam field of view (Figure 6). Light from this region enters the NIRCam instrument directly through the aft-optic-system (AOS) mask, located in front of the JWST tertiary mirror, without first bouncing off the primary and secondary mirror. This light path is called the rogue path. The collecting area through the AOS mask is effectively small (about 0.01 square meters), moreover the light in this case is not focused by the primary and secondary, and it undergoes several reflections within the NIRCam instrument before hitting the detector. As a consequence, the total amount of scattered light is reduced by about 18 magnitudes with respect to the bright star that causes it. There is some variability in this dimming factor due to the exact placement of the stars within the rogue path. 

Figure 4. Claws and wisps

Click on the figure for a larger view.

Examples of the morphology and detector location of the stray light features known as wisps and claws are shown in this figure. Multiple dithered exposures are median-stacked to remove astronomical sources, and to enhance the stray light features. Note that while the wisps are fixed in detector space, and their morphology does not change across observations (although their strength does), the claws visible in the figure are peculiar to this particular observation and differ between observations (and are completely absent in most of them).

The resulting feature is an extended image with a claw-mark-like appearance. The characteristic shape is imprinted by machining features on one of the surfaces involved. The claws are observed exclusively in the short wavelength channel. When present, claws occur primarily in the A1 or B4 detectors, affecting roughly 5% of the pixels on those detectors.

Claws will have different appearances and morphologies depending on the location of the bright star in the susceptibility region. For a set of exposures with similar on-sky positions (e.g., multiple dithers or small mosaics), their appearance and location on the detectors may remain roughly constant.

The claws do not always appear in the same areas of the NIRCam detectors, but they move as the responsible source moves within the susceptibility region. Given that the claw images are not observed via the nominal telescope path, the amount of claw motion is not 1:1 with the amount of motion of the source within the rogue path. Moving the telescope boresight by several arcminutes results in a claw motion of a few arcseconds. Typically, claws do not move in an appreciable manner within dithered sequences, but they are observed to move across multiple mosaic tiles.

The effect of the claws is to increase the background by about 10% on a typical area of 50,000 pixels. This factor is estimated at the ecliptic (larger than average background) using the F200W filter and a very red star. Variations are expected as a function of the SED of the bright star and of the background specific to each observation. Note that the affected area is a few percent of a single NIRCam short wavelength detector (2048 × 2048 pixels).

Given the size of the susceptibility region and based on a conservative estimate of the minimum stellar brightness necessary to produce this feature, the NIRCam team estimates that up to about 10% of observations may be affected. In practice, during the JWST commissioning period, the occurrence of claws has been significantly lower than this. The NIRCam team is both developing mitigation strategies and actively monitoring data as they are acquired, to further characterize this phenomenon and further reduce its potential impacts to science.

Figure 5. Claw stacked image

Click on the figure for a larger view.

Example of a claw on detector A4. This image is obtained by stacking multiple exposures in a dithered sequence, in order to remove stars and galaxies and increase the SNR in the claw feature itself. The colorbar indicates the ratio with respect to the median value of the image. In this specific instance of a very strong claw, the surface brightness in the claw region is about 15%–20% higher than the average background.

Figure 6. Claws susceptibility region

Click on the figure for a larger view.

The shape of the susceptibilty zone for claws. Given the different optomechanical characteristics of the 2 NIRCam modules, the regions for modules A and B are similar and almost specular, but they do differ in the details. Note the axes are in degrees; the NIRCam field of view would appear very small (~5' × 2') at the origin.


Wisps (see Figure 6) are faint, diffuse stray light features present in the same detector locations in all NIRCam exposures. They are most prominent on the B4 detector, with variable brightness that is typically about 10% of the zodiacal background. Fainter wisps have also been seen in detectors A3, A4, and B3, affecting a few percent of the pixels in each detector.

Similarly to claws, the wisps are caused by light going through a rogue path. The light bypasses the primary and secondary mirrors and enters NIRCam through the aft-optic-sytem (AOS) mask in front of the tertiary mirror.

Differently from the claws, however, the wisps are not caused by light coming directly from a bright source. Wisps are instead caused by light coming off-axis, and bouncing off the top secondary mirror strut. Therefore, with respect to the NIRCam FOV, the illuminating source is always at the same location, albeit its brightness varies.

Wisps can be thought of as an additive feature that can be subtracted from each exposure by using a universal wisp template whose strength can be scaled to match the observed strength. The NIRCam team developed initial templates provided here, along with preliminary guidance on how to process NIRCam images to remove this feature.

Wisp templates available here may be subtracted from count rate (slope) files (stage 1 output) using code provided:

Ginkgo leaf

This wispy artifact shaped like a ginkgo leaf was observed in long wavelength images in module A (detector A5). It was produced by scattered light from a star to the left (+V2) of the detector and ~24.5" below (-V3) the top edge. This is a rarely observed feature that has so far only been reported for one observing program. Dithers should be sufficient to mitigate the effects of this scattered light.

Figure 7. NIRCam ginkgo leaf

Click on the figure for a larger view.

The wispy ginkgo leaf artifact in the dashed box was observed in NIRCam commissioning program 1067, observation 453 in detector A5. It appeared in all 10 images during a scan: as the star moved further off the detector (to the left), the glint moved further onto the detector (to the right). The star was off the left side (+V2) of detector A5 and ~24.5" below the top edge (in V3). A similar scan of module B did not yield a similar feature.

WFSS tadpoles and shells

Tadpoles are scattered light artifacts that may be mistaken for emission line sources in NIRCam WFSS data. They appear most prominently and ubiquitously in Grism C module B data.

Shells are fainter diffuse features seen occasionally due to scattered light from very bright sources.

Both tadpoles and shells are seen primarily in module B (Figure 8). The module B grism has anti-reflective coating on only one side. The module A grism has the coating on both sides, significantly suppressing these artifacts.

Figure 8. NIRCam WFSS tadpole and shell

Click on the figure for a larger view.

Example of tadpole and shell artifacts observed in F322W2 Grism C module B data from NIRCam commissioning program 1076: jw01076126001_02101_00003_nrcblong (see JWST file naming conventions).


Each tadpole consists of either one bright knot or multiple bright knots in a line parallel to the spectral traces in Grism C data (Figure 9) or at an angle in Grism R module B.

Tadpole brightness is correlated with source continuum brightness. A source with F322W2 continuum AB mag 17 (19.5) yields a tadpole head with brightness ~10 (1) DN/s.

Figure 9. Tadpole examples

Click on the figure for a larger view.

Tadpole artifacts observed in F322W2 Grism C module B data from NIRCam commissioning program 1076. Green circles on most panels indicate sources identified by DAOFIND that could be mistaken for bright emission lines with faint or no continuum visible in its spectrum.

Tadpole locations

The tadpoles are observed near the spectral trace, far (~1.5 arcmin) from the direct-imaging position (the position of the source if the pupil  were set clear). The tadpole head is offset from the direct-imaging position by roughly (-40, 1385) pixels in (x, y) in F322W data and (36, 1350) pixels in F444W data. This position varies by up to 10 pixels depending primarily on the x position. Note the tadpole is offset to the left in F322W2 and the right in F444W.

Figure 10. Tadpoles observed in F322W2 and F444W


Click on the figures for a larger view.

Tadpoles observed in F322W2 (top) and F444W (bottom) Grism C module B data. Red stars mark the direct-imaging positions to the right of the 1st-order spectrum at 3.95 µm. Red squares mark the tadpole positions to the left of the 2nd-order spectrum in F332W2 and to the right of the spectral trace in F444W. Data are from NIRCam commissioning program 1076: jw01076126002_02101_00004_nrcblong and  jw01076119001_02101_00004_nrcblong.

Weaker tadpoles observed in Grism R modules A and B

While tadpoles are strongest in Grism C module B, they are also observed more weakly in Grism R and in module A (Figure 11).

Grism R module B tadpoles are not parallel to the spectral traces, but rather at an angle. They appear to be isolated to the upper few hundred pixels in y on the detector. Unlike the Grism C tadpoles, these move with respect to the spectra when dithering.

Tadpoles may also appear in module A data, though significantly suppressed by the anti-reflective coating. One possible example is shown in Figure 11. It is parallel to the spectral traces.

Figure 11. Weaker tadpoles observed in Grism R


Click on the figures for a larger view.

Weaker tadpoles are observed in Grism R data from module B (top) and perhaps module A (bottom). Note the module B Grism R tadpoles are not parallel to the spectral traces, as they are in Grism C data. The tadpole in module A was tentatively identified by visual inspection in NIRCam commissioning program 1076, observation jw01076113001_02101_00001_nrcalong.

Notable updates
Originally published