The Science Performance of JWST as Characterized in Commissioning

The Ariane 5 rocket that launched JWST on 2021 December 25 UT injected it into an orbit that was well within specification, with a semimajor axis approximately 0.5σ larger than nominal. This very slightly "hot" injection state had a semimajor axis of 542,120.1 km versus the nominal of 536,533.8 km, and a delta-v at launch vehicle separation within 3 m s⁻¹ of the target velocity of 10,089 m s⁻¹. This accurate launch, in combination with three on-time, nominal mid-course corrections, minimized propellant consumption and delivered JWST to a nominal orbit around the second Earth-Sun Lagrange point (known as L2). This orbit fully complies with all geometry requirements, and supports communications with the Science & Operations Center using the Deep Space Network.

Orbit around L2 is maintained through regular station-keeping burns, which are scheduled every three weeks. As of 2022 July 12, there have been four station-keeping burns, with typical durations of tens of seconds. During commissioning, three station-keeping burns were skipped because the computed correction was negligibly small.

There are no consumable cryogens onboard JWST; the telescope and the science instruments are passively cooled by the sunshield and radiators, and MIRI's active cryocooler recycles its helium. The only onboard consumables are propellant: fuel and oxidizer. Before launch, JWST was required to carry propellant for at least 10.5 yr of mission lifetime. Now that JWST is in orbit around L2, it is clear that the remaining propellant will last for more than 20 yr of mission lifetime. This fortunate surfeit has multiple causes: an accurate launch; launch on a day that required relatively less energy to get to L2 than most other possible launch dates; three timely and accurate mid-course corrections that sent JWST to L2 with the minimum possible propellant usage; and finally, careful stewardship of mass margins by the engineering team over the years, such that the remaining mass margin was used to add more propellant than required, until the tanks were full.

For the remainder of the mission, propellant will be used for two purposes: stationkeeping burns (using fuel and oxidizer) to maintain the orbit around L2, and momentum dumps (using only fuel) to remove momentum from the reaction wheels. Momentum accumulates as solar photons hit the sunshield and impart a net torque, which the reaction wheels resist by spinning up. The actual rate at which the observatory builds up momentum is within specifications and is well below worst-case allocations, which further contributes to propellant lifetime. While the detailed propellant usage depends on orientation, which is set by the observing schedule, the big picture is that JWST has sufficient propellant onboard to support science operations for more than 20 yr.

At this point, it is not clear what will determine the duration of JWST's mission. The mirrors and sunshield are expected to slowly degrade from micrometeoroid impacts; the detectors are expected to experience cumulative slow damage from charged particles; the sunshield and multilayer insulation will degrade from space weathering; the spacecraft was designed for a five year mission (as is standard for NASA science missions); and the science instruments include many moving parts at cryogenic temperatures. These sources of degradation were all taken into account in the design of JWST, with performance margins set so that JWST will still perform after many years of operation. At present, the largest source of uncertainty is long term effects of micrometeoroid impacts that slowly degrade the primary mirror. As discussed in Section 4.7, the single micrometeorite impact that occurred between 2022 May 22 and 24 UT exceeded prelaunch expectations of damage for a single micrometeoroid, ⁴⁹ triggering further investigation and modeling by the JWST Project. The Project is actively working this issue to ensure a long, productive science mission with JWST.

The slew speed and settle time are important drivers for efficient operations as the observatory changes pointing to look at different targets on sky and carry out orbital stationkeeping and momentum unloading maneuvers. Slews use all six reaction wheels on the spacecraft to repoint the observatory, although it is possible to control with five reaction wheels. The control system was designed to slew 90° in less than 60 minutes, which has been demonstrated during commissioning.

Achieved slew speeds and durations to repoint to new targets at the start of each visit are broadly consistent with pre-flight expectations, such as the timing model encoded in the Astronomer's Proposal Tool (APT), plus typically 2 minutes duration for control overheads and settling.

At the end of a slew, pointing transients are observed while the observatory settles. The pointing settles in ∼30 s with damping from isolators between the telescope and spacecraft bus. Re-pointing maneuvers can generate fuel slosh, which is at ∼0.045 Hz and not compensated by the fine guidance control system. The slew rate profile has been tuned to reduce the excitation of the fuel slosh mode. Measurements of line-of-sight pointing performance (Section 3.3 below) confirm that the resulting effect of fuel slosh on pointing is <0.5 milliarcseconds (mas).

After slewing to a new target, the pointing stabilizes quickly in less time than it takes the FGS guide star acquisition process to complete. This was not initially the case; in the first months of commissioning, long slew settling times and high image motion impacted many observations, and required efforts to investigate, diagnose, and mitigate. Adjustments to Attitude Control System parameters and several software patches dramatically improved slew settling performance and resolved this issue. Users examining very early commissioning data (prior to mid 2022 April), particularly images taken in coarse point mode without fine guiding, should be aware of this caveat.

The cooldown of the telescope and science instruments was nominal and closely matched predictions. As predicted, the primary mirror segments have cooled to temperatures of 35–55 K, with the hotter segments those closest to the sunshield. The secondary mirror has cooled to 29.3 K, the near-IR instruments to 35–39 K, and MIRI to 6.4 K. The MIRI cooler achieves this temperature at nominal, pre-flight predicted performance levels and has no perceptible effect on pointing stability (e.g., jitter) or on the performance of the other instruments. Since cooling to operational temperatures, these temperatures have remained extremely stable with time. For the telescope mirrors and structures, the stability is within the 40 mK noise of the temperature sensors on those components. The science instrument temperatures vary based on their activity, but they are also within the 10 mK noise for the instrument sensors. Any resulting temperature change impact can only be identified optically (see Section 4.5). No long term temperature drift has been detected since achieving final cooldown conditions. Temperatures of detector focal plane arrays are actively stabilized to a precision of a few mK.

The observatory was designed to minimize ice deposition on the optical surfaces. Sensitive components of the science instruments and the telescope's fine steering mirror were heated through the cooldown to prevent ice accumulation. Throughput measurements detect no spectral signatures of ice, which constrains any ice deposition that may have occurred to be far better (less ice) than the requirements.

Components within the spacecraft bus (computer, communications, cryocooler compressors and electronics, attitude control and propulsion systems, etc.) are comfortably within required operating temperature ranges regardless of pointing direction or telescope activities. Instrument electronics housed on the cold side of the Observatory are also within required operating range and are under tight temperature control to minimize any temperature-induced distortions (see Section 4.5.3). All heaters on the observatory are functional on prime circuits and demonstrating expected margins.

The shape of the deployed sunshield affects the temperature and thermal stability of JWST, the amount of scattered light from the Earth, Moon, and stars, and the background levels at the longer wavelengths. Telemetry (microswitch, motor, thermal, power, and inertial reference unit) indicated a successful deployment ending with a nominal deployed shape. There are no subsequent indications of any issue with the deployed shape, from the many thermal sensors onboard or from the background levels seen in the science instruments.

The shape of the deployed sunshield affects how quickly the observatory builds up angular momentum from solar photons, which is then stored in the reaction wheels and must be dumped periodically using thrusters. The sunshield geometry was designed to minimize momentum accumulation. The measured torque table on-orbit is consistent with the pre-launch model within the allocated uncertainties. As noted above, this implies a lower rate of fuel use for momentum management.

After optimization of the solar array regulator settings, JWST is now generating 1.5 kW to match the power load, with a capability of >2 kW—as such, the power margins are comfortable. JWST is projected to have a tight (11%) margin on data downlink during Cycle 1. The project is working closely with NASA's Deep Space Network (DSN) to resolve all issues and ensure DSN can adequately support all Cycle 1 science and beyond. The JWST spacecraft has all the redundancy it had at launch. Of the 344 single point failures ⁵⁰ that were present at launch—almost all of them related to deployments—only 49 remain; these are common to most science missions (for example, only one set of propellant tanks, only one high gain antenna). Fifteen of the remaining single point failure items are associated with the science instruments—any future failure with these items would degrade science performance but would not end the mission.

JWST has a robust fault management system that makes use of its redundant components to ensure the observatory remains safe. With any new spacecraft/observatory, the first couple months after launch provide the operations team an opportunity to learn how the vehicle performs on-orbit and adjust the fault management and system parameters to fine tune the system behavior. During commissioning, JWST experienced seven safe mode entries (six safe haven and one inertial point mode). Early science operations (2022 July - December) experienced one safe haven entry and four entries into inertial point mode. The majority of these safe mode entries were a result of on-orbit learning of the nuances of the control system behavior and system level interactions. All of the safe mode entries' underlying causes are understood and several updates to the fault management and control system set items have been loaded to flight software to prevent the issues from recurring.

Likewise, flight software has placed individual instruments into safe modes on several occasions in responses to unexpected telemetry values or conflicting commands. In all cases the underlying issues were quickly understood, and appropriate steps were taken to bring the instrument back into operation and prevent recurrence of the fault. JWST's onboard event-driven operations system works as intended to allow the observatory to flexibly continue observations when an observation cannot execute due to an instrument fault or a guide star acquisition error; in such cases, JWST will automatically move on to the next observation in the onboard observation plan.

Absolute pointing accuracy after guide star acquisition is excellent. Observed pointing offsets are generally below 019 (1σ, radial). When systematic offsets are removed by improved astrometric calibration between the Guiders and the other science instruments (SIs), the residual scatter around the desired target position is generally below 010 (1σ, radial) , better than the prelaunch predicted value of 014, which is in turn significantly better than the required value of 10. Updates for such systematic offsets are in progress. This confirms expectations that many integral field spectroscopy (IFS) observations may omit target acquisition and rely solely on guide star acquisition to achieve the necessary science pointing. The excellent pointing performance can be credited to JWST's own systems, as well as the high accuracy and precision of the guide star catalog enabled by the Gaia mission.

Early in commissioning, some observations had significantly larger pointing offsets, of order ∼15, due to catalog cross-matching issues that occurred when data from Gaia and other catalogs such as 2MASS were combined to produce the current JWST guide star catalog. Updated guidelines for selecting guide/reference stars was implemented 2022 May 5, and non-stars were disallowed as reference objects 2022 June 24; both greatly reduced mis-identification. A future version of the catalog is planned for 2023 which is expected to provide further improvements.

Pointing repeatability is likewise excellent: independent separate observations returning to the same target and at the same position angle generally result in identical target pointings on a detector to within <01.

For observing scenarios requiring better final pointing accuracy than provided by the guide star catalog alone, the instruments offer onboard Target Acquisition procedures to place targets where desired within a science instrument field, e.g., centered on a coronagraphic spot or null, or in a repeatable position for NIRISS Aperture Masking Interferometry or Single Object Slitless Spectroscopy (SOSS). There are several distinct versions of target acquisition for the various instrument modes. These onboard processes have been confirmed to meet their requirements, typically yielding target positions within a few mas rms per axis in the near-IR instruments, and slightly larger in MIRI with its broader longer-wavelength PSFs. NIRCam coronagraphy, the final mode to have its Target Acquisition process evaluated, currently experiences larger offsets, which will be alleviated by future calibrations—even so, the Small Grid Dithers recommended for that mode cover the range of uncertainty and yield excellent coronagraphic contrast.

The algorithm for NIRSpec Multi-Object Spectroscopy is the most complex, because it derives a roll correction for the observatory position angle as well as the usual x, y positional offsets. That target acquisition required great care in implementation, and was the most challenging type of target acquisition to get working during commissioning; it is now working within requirements when the requisite attention is paid to the accuracy of the input reference star information.

The pointing stability of the line of sight under fine guidance control is superb, several times better than requirements. The FGS sensing precision, parameterized as the Noise Equivalent Angle (NEA, an equivalent jitter angle calculated based on centroid precision and S/N) is usually ∼1 mas (1σ per axis), as compared to the requirement of 4 mas; the NEA is usually symmetric in x and y.

The achieved line-of-sight jitter as seen in the science instruments is similarly good, even when measured at higher frequency than the 15.6 Hz cadence of the FGS measurements used in calculating the Guider NEA. Jitter measured via high-frequency sampling (every 2.2 ms) using a small NIRCam subarray has also been ∼1 mas (1σ per axis). For long observations, drift in observatory roll could generate systematic motion of sources in an SI field, as the FGS can not sense or correct roll with the single guide star used in the Fine Guide control loop. However, measurements during a commissioning thermal characterization test indicated that the roll drifts are extremely small, far below requirements, even after a worst-case hot-to-cold slew, and contribute negligibly to the total pointing error. In comparison, pre-flight predictions for pointing stability were in the range 6–7 mas (1σ per axis).

The FGS guide star magnitude scale matches the 2MASS J-band, Vega scale, with guide star brightnesses of 12.5 < J < 18 allowed for fixed targets. Dimmer stars will typically have a larger NEA, due to the lower signal to noise per time step, but with NEA that is still well within the requirement. The guide star selection system preferentially selects the brightest available guide star for a given pointing. In practice the achieved NEA can vary by ∼0.3 mas between dithers on the same source. For moving targets, the guide star faint limits are somewhat reduced to 16.5 in Guider 1 and 17.0 in Guider 2. Guiding performance can vary under some circumstances. Reasons that can cause a brighter star to appear to register lower counts than expected (and give an increased NEA) include: unflagged bad pixels in the guider box, (flagged) bad pixels exactly coincident with the guider star's position, guiding on an extended galaxy instead of a star, quantum efficiency variations due to cross hatching on the detector surface, or a guidestar falling at the edge of the centroiding box. Work is in progress to reduce these variations. In most cases, observers should expect excellent stability in their observations.

Very precise flight jitter measurements have yielded detailed insights into observatory dynamics, and confirmed many aspects of preflight integrated modeling. These data show that neither the spacecraft's reaction wheels nor the MIRI cryocooler produce measurable contributions to the line-of-sight jitter. Instead there are two signatures, detectable but at low level, that come from a 0.3 Hz oscillation mode of a vibration damper between the telescope and spacecraft bus, and a ∼0.045 Hz oscillation understood to be due to fuel slosh, both consistent with models. The amplitudes of these modes range up to ∼0.3 mas each, with variation over time that does not appear correlated with pointing history such as slew distances.

Momentum/reaction wheel operations or unplanned High Gain Antenna motion may occasionally disturb observatory dynamics and degrade pointing stability; these are generally short-lived and fine guiding is expected to be maintained. High Gain Antenna moves are not planned during any science observations except for long duration time series observations.

JWST has three methods of performing a small repointing, commonly called a dither. Dithers less than 60 mas are executed with the fine steering mirror (FSM) while maintaining closed loop on the guide star. Dithers between 60 mas and 25'' are executed by dropping closed-loop guiding, slewing, and re-entering guiding at the track mode stage. Dithers larger than 25'' are executed by dropping closed-loop guiding, slewing, and re-entering guiding at the guide star acquisition stage.

All three methods of dithers result in an offset precision as measured by the Guider of 1–2 mas. On the sky, the accuracy of the dithers is typically 2–4 mas rms per axis, due to residuals in the Guider's astrometric calibration or small systematic offsets in the Guider's calculated 3 × 3 pixel centroids. As reported by the observing SI, offsets for large dithers may differ by another few mas, due to residuals in the astrometric calibration of the SI. As SI astrometric calibrations continue to improve, these residuals may decrease.

JWST has a Level 1 requirement to track objects within the solar system at speeds up to 30 mas per second (mas s⁻¹). In commissioning, tracking was tested at rates from 5 to 67 mas s⁻¹. These tests verified tracking and science instrument performance for moving targets, including dithering and mosaicking.

All tests of moving targets during commissioning were successful. Centroids showed sub-pixel scatter in all instruments. No test showed elongation of moving targets, as would be expected in the case of poor tracking. rms jitter in the guider was typically <2 mas (1σ, radial), comparable to that seen for fixed targets. Image quality as determined by FWHM measurements of point spread functions was also comparable to that of fixed targets. Table 1 summarizes the moving targets observations during commissioning.

Tracking at faster-than-requirements rates of 30–67 mas s⁻¹ showed accuracies similar to tracking of slower-moving targets; this potentially opens up science for near-Earth asteroids (NEAs), comets closer to perihelion, and interstellar objects. This capability was pushed further early in science operations, when JWST successfully tested rates up to 110 mas s⁻¹ (396'' hr⁻¹) on near-Earth asteroid 2010 DF1, to observe and track the Double Asteroid Redirection Test (DART) mission target (65803) Didymos at the time when the spacecraft impacted the asteroid moonlet on 2022 September 26, to demonstrate the capability of modifying an object's orbit for the purpose of planetary defense (Thomas, C. et al. in preparation.)

While tracking at super-fast rates has now been demonstrated, the observatory efficiency was poor due to guide star availability and acquisition at these rates. Thus, the new maximum rate of motion being offered is 75 mas s⁻¹ without limitations, but special permission for rates up to 100 mas s⁻¹ may be requested and subjected to approval for future cycles.

Observing a bright planet and its satellites and rings was expected to be challenging, due to scattered light that may affect the science instrument employed, but also the fine guidance sensor must track guide stars near the bright planet. Therefore, commissioning included tests of moving target tracking with NIRCam, where Jupiter was incrementally moved from the NIRCam field of view (FOV) to the FGS-2 FOV. See Figure 1. These observations verified the expectation that guide star acquisition works successfully as long as Jupiter is at least 140'' away from the FGS, consistent with pre-flight modeling.

Zoom In Zoom Out Reset image size

The other SIs were also tested for efficiency with nearby scattered light, also using the Jupiter system. Preliminary results have measured scattered light contamination on the detectors for all instruments when the planet was not in the primary FOV, which will need to be considered for planning nearby satellite observations. The most notable impact for scattered light is in NIRSpec IFS mode—if a bright planet is on the microshutter array, then light seepage becomes significant.

The closed-loop fine guidance system is one of the most complex aspects of JWST operations, requiring close coordination between several subsystems and multiple control loops running in real time. Activating and tuning the fine guidance system was a significant focus during commissioning. The robustness of the system has steadily improved and continues to do so as reasons for failure are identified and mitigated.

As a snapshot of performance during the later period of commissioning, from May 25 through 2022 June 16, guiding worked successfully ∼93% of the time: guiding was successful on the first try ∼81% of the time, and 12% of the time succeeded on the second or third guide star candidate attempted. (Up to three guide star candidates can be tried in a visit.) In the same time period, guiding failed or skipped 7% of the time.

Several reasons for guiding failure were identified, with steps taken to mitigate them:

• 1.  
  The "guidestar" is actually a galaxy, but was classified as a star in the guide star catalog. These are flagged in the guide star catalog once found.
• 2.  
  Attempts to guide on known galaxies frequently failed. Using a known galaxy as a reference source for ID was disallowed starting 2022 June 24.
• 3.  
  Guide star coordinates, or occasionally brightnesses, in the catalog are incorrect. The catalog is corrected when such errors are found.
• 4.  
  The guide star catalog contained duplicate entries for some guide stars. This duplication was reduced greatly with new rules for selecting guide stars, implemented 2022 May 5. A new version of the guide star catalog expected in 2023 should largely eliminate this issue.
• 5.  
  The guide star may be placed on a bad pixel. Bad pixels are flagged once identified.
• 6.  
  The pointing may not have stabilized sufficiently at the start of the Guide Star ID process.

In Cycle 1, users should expect a closed-loop guiding success rate close to the 93% value that was achieved late in commissioning. That value is, largely by coincidence, very close to the success rate for Hubble guide star acquisitions in recent cycles. Efforts continue to optimize guiding success rates.

The achieved telescope wave front error (WFE), measured at the primary wave front sensing field point in NIRCam module A, is generally in the range 60–80 nanometers rms; it varies on multiple timescales as described below. ⁵¹ That WFE contribution sums with field-dependence of the telescope WFE and instrument internal WFE to yield total observatory WFE values which are modestly higher: 75–130 nm depending on instrument, observing mode, and field position. See Table 2. Motions of mirror segments over time can lead occasionally to wave front error levels higher than those values, which are corrected through the routine wave front sensing and control process.

JWST has exquisite image quality across the entire telescope field of view and at all available wavelengths. Expressed relative to wavelength λ, JWST ranges from ∼λ/10 for NIRCam F070W to better than λ/100 for MIRI F1000W and longer. JWST aimed to achieve at λ = 2 μm a Strehl ratio of 0.8 (corresponding to wave front rms λ/14 or better), which is considered having diffraction-limited image quality for space-based telescopes. This is achieved with substantial margin, such that the NIRCam short wavelength channel, with typical WFE ∼80 nm, in fact achieves λ/14 at λ = 1.1 μm.

That wave front quality yields optical point spread functions (PSFs) with angular resolutions set by the diffraction limit (i.e., PSF full width at half maximum ∼λ/D) across the full range of available wavelengths. See Figure 2. In particular, even in its shortest filter F070W, the NIRCam short wavelength channel achieves a Strehl ratio of ∼0.6. Though ∼40%–50% of the light is in a diffuse speckle halo at that wavelength, the PSF prior to detector sampling still has a tight core with angular resolution ∼λ/D. (In that sense, JWST's PSF quality at 0.7 μm is similar to PSFs achieved at 2 μm by adaptive optics systems on 8–10 m ground-based telescopes in good conditions.) In practice, angular resolution at the shortest wavelengths (<2 μm for NIRCam short wavelength, <4 μm for NIRCam long wavelength or NIRISS) is limited more significantly by detector pixel Nyquist sampling than by optical performance; subpixel dithering and image reconstruction ("drizzling") will be required to make use of the full resolution at these wavelengths. ⁵²

Zoom In Zoom Out Reset image size

Wave front sensing confirms the surface quality of the individual mirror segments in space matches closely the mirror surface maps measured during cryogenic testing on the ground. See Figure 3. In other words, after launch into space, and significant thermal contraction and deformation while cooling from room temperature to ∼45 K, the achieved wavefronts on each segment were just as expected.

Zoom In Zoom Out Reset image size

JWST has top-level science requirements to achieve image quality with Strehl ratio greater than 0.8 in NIRCam at a wavelength of 2 μm (equivalent to wave front error of 150 nm rms) and MIRI at a wavelength of 5.6 μm (equivalent to 420 nm rms). These and other top-level requirements flowed into detailed optical performance budgets and lower-level requirements, such as a required telescope wave front error ≤131 nm rms over the fields of view of the science instruments, a required telescope stability of ≤54 nm rms over two weeks, and so on. Though line-of-sight image jitter is distinct from WFE, it is also tracked within the same budget via a computed equivalent WFE.

Figure 4 presents an abbreviated summary comparing measured performance in flight to those budgets at high level. Note that the measured values shown here are from commissioning at observatory "beginning of life," while the required and predicted values were derived for "end of life" after a notional 5 yr mission. At the observatory top level, the achieved WFE is well below predictions by ∼30%, and the remaining performance margin relative to requirements is large. All lower-level terms are at or below requirements values, with the sole exception of the high frequency WFE which was increased by the 2022 May micrometeoroid impact. Terms with particularly notable performance relative to requirements include the telescope stability (see Section 4.5), the line of sight pointing (Section 3.3), and the wave front quality of the science instruments (which are all significantly better than their requirements).

Zoom In Zoom Out Reset image size

An initial assessment, which combined the measured beginning-of-life performance with model predictions for observatory aging in the L2 environment, predicts that JWST should meet its optical performance requirements for many years. The current largest source of uncertainty in models is the rate of mirror surface degradation from micrometeoroids, discussed below.

JWST's hexagonal aperture creates a characteristic diffraction pattern in its point spread functions, with six stronger diffraction spikes at 60° intervals created by the segment and aperture edges, plus two fainter horizontal spikes created by the vertical secondary mirror support. While these diffraction spikes can be visually dramatic in images which are deeply exposed or are plotted with log stretches, it is the case that the majority of light is focused into the PSF core (typically ∼66% within the first Airy ring).

Compared to the diffraction patterns from circular apertures which many astronomers are more familiar with, such as Hubble, the hexagonal geometry of JWST concentrates wide-angle diffracted light more strongly into the diffraction spikes, while the areas between those spikes are relatively darker. For bright sources the diffraction spikes can be seen to a separation of many arc minutes, falling off as ∼R^−1.5, including diffraction spikes from sources outside of an instrument's field of view.

Note that the position angles of the six bright diffraction features are different between the PSF core region (2–5 λ/D, dominated by diffraction from the overall outer hexagonal outline) and the outer wings (>5 λ/D, dominated by diffraction from the individual segments). See Figure 2. This can be understood intuitively from the pupil geometry: when assembling a larger hexagon from smaller hexagons, the outline of the larger hexagon which is created is rotated 30 deg relative to the smaller ones.

Additional PSF details per instrument:

• 1.  
  MIRI imager and MRS PSFs, particularly at wavelengths ≤10 μm, show additional diffraction spikes in vertical and horizontal directions in the instrument coordinate frame, called the "cruciform" or "cross artifact." These arise due to diffraction that occurs internal to the detector at the detector electrical contacts. This effect was seen in Spitzer's similar Si:As detectors, and was modeled and expected for MIRI (Gaspar et al. 2021). Improved models based on flight data are able to reproduce this artifact in detail including the field dependence. Pipeline steps are being developed to compensate for its impact on MIRI MRS data cubes.
• 2.  
  NIRISS imaging PSFs at wavelengths ≥2.7 μm have stronger diffraction spikes due to additional pupil obscuration from the CLEARP pupil wheel position which must be used with those filters.
• 3.  
  NIRISS imager PSFs at wavelengths ≤2.0 μm show additional anomalous or "extra" diffraction spikes, which are more diffuse than the ordinary spikes and have a position angle that varies strongly with field position. The intensity of the anomalous spikes decreases with wavelength, in F090W containing ∼70% of the flux of the normal vertical diffraction spike, becoming barely detectable by F200W. These spikes are now understood as due to diamond-turning tool marks on off-axis mirrors within NIRISS. Retroactive examination of some ground test data also shows their presence at lower S/N. A similar effect is also seen in FGS PSFs.
• 4.  
  The rotation of NIRSpec in the focal plane means that, in the detector coordinate system, NIRSpec PSFs are rotated by 139 deg relative to the other instruments. MIRI is similarly rotated 5 deg.
• 5.  
  Instrument modes with specialized optical components, such as NIRCam coronagraphy, MIRI coronagraphy, NIRISS aperture masking interferometry, and NIRISS SOSS, each have their own PSF properties, as described in Girard et al. (2022), Kammerer et al. (2022), Kammerer et al. (2023), Boccaletti et al. (2022). These observed PSFs are in general highly consistent with model predictions.

The telescope's effective area, the product of the telescope area and its transmission, is a key optical performance parameter that was tracked through development and characterized on orbit. The JWST unobscured telescope area was required to be >25 m². The measured value using the NIRCam pupil imaging lens was 25.44 m². The telescope's wavelength-dependent transmission ranged from 0.786 at 0.8 μm to 0.933 at 28 μm, better than requirements at each wavelength. Even though the JWST optics spent significantly more time in ground facilities than originally anticipated, these high transmissions were maintained with careful control of contamination throughout the integration and test phases and during preparation for launch. Brush cleaning was also carried out on the telescope primary mirror segments and secondary mirror to remove particulates.

Commissioning observations characterized the telescope's optical stability and variations on multiple timescales. Highly precise wave front sensing allows measurement of very small changes (<10 nm). The overall amplitudes of variation are close to predicted values, however some timescales are faster than expected. These variations have small but measurable effects on PSF properties, which should be taken into account for measurements at the highest precision, but will be negligible for many science cases (similar to the effect of the "breathing" variations seen in Hubble PSFs on orbital timescales).

Overall stability: During the last month of commissioning, 2022 June, the change in wave front measured between successive observations roughly 2 days apart was typically ≤25 nm rms, and frequently less than half that (range 8–50 nm rms per 2 days). During this time period, the observatory was conducting a wide range of commissioning and Early Release Observations typical of science activities, so this level of stability should be representative of what can be expected during Cycle 1. Indeed, during the first half of Cycle 1 the median change in wavefront between successive observations was only 10 nm rms, just slightly above the typical wavefront sensing measurement uncertainty of 7 nm rms.

Though we provide details below on observed variations over time, we wish to emphasize that these are small variations. The absolute amplitude of drifts seen between successive wave front monitoring visits is comparable to the Hubble Space Telescope's ∼18 nm rms of focus variation typical on orbital timescales (see e.g., Lallo 2012). Yet for JWST the variations are observed to mostly occur on significantly longer timescales, and the effect on images at longer wavelengths is correspondingly reduced. (Hubble is typically stable to λ/30 at λ = 0.5 μm over 90 minutes; JWST is often stable to λ/100 at λ = 2 μm over 2 days.) Also similar to Hubble, the amplitude of wave front variations over time is comparable to and generally less than the field-dependent variations within a given instrument.

Factors contributing to stability levels:The telescope is very thermally stable, but small changes in equilibrium temperature (<0.1 K, within the temperature sensor noise) can still occur in response to changes in attitude with respect to the Sun. Variations from instrument heat sources can also affect telescope structures. A thermal stability exercise measured these effects by moving between the extremes of JWST's field of view, with a week-long soak at "cold" attitude (−40° pitch relative to the Sun) sandwiched between an equal amount of time at the "hot" attitude (0° pitch); this test is intentionally more stressing than typical attitude profiles during science operations. The thermal slew exercise (PIDs 1445, 1446) included a wide range of wave front sensing, imaging, and sensor telemetry investigations to characterize many aspects of JWST's performance on multiple timescales. Modes of variation observed with JWST include the following.

4.5.1. Telescope Backplane Thermal Distortion

4.5.2. Telescope Soft Structure Thermal Distortion

4.5.3. Fast Oscillations from Heaters

4.5.4. Segment Tilt Events and other Drifts

Since the completion of telescope alignment, regular wave front sensing and control has monitored and maintained telescope alignment, as will continue throughout the mission. During science operations, wave front sensing measurements are conducted every two days (roughly, with flexible cadence around science observations). The resulting measurements of mirror alignment state versus time are automatically made available in the MAST archive; software for making use of these measurements to produce time-dependent PSF models is now included in the WebbPSF package. ⁵⁴

Wavefront control is nominally applied whenever the total wavefront error (telescope + NIRCam, short wave channel, at the reference field point) reaches 80 nm rms due to drift or any other instability. Over the period 2022 November through 2023 January, this total has varied over a narrow range between about 65--77 nm rms, with only two wavefront corrections needed.

Because of the every-two-days cadence of wave front sensing measurements, wave front states at times between those measurements are not directly measured (e.g., if a tilt event or micrometeoroid impact occurs within some 2 days, its exact time of occurrence is not measured). Some science modes do allow for more frequent wave front measurements, in particular NIRCam time series using the weak lens in the short wave channel, and to some extent the NIRISS SOSS mode.

Inevitably, any spacecraft will encounter micrometeoroids. The part of JWST that is most vulnerable is the primary mirror. Some of the resulting wavefront degradation from micrometeoroid strikes is correctable through regular wavefront control, while some of it comprises high frequency terms that cannot be corrected; this latter, cumulative damage was incorporated into the prelaunch JWST wavefront error budget (Feinberg et al. 2022). Over the period 2022 March 11 to 2023 January 12, wavefront sensing recorded a total of 25 localized surface deformations on the primary mirror that are attributed to impact by micrometeoroids (a hit rate of 2.5 per month.) With one major exception, after correction these micrometeoroid strikes have had no measurable impact on the overall wavefront error.

The outlier is the micrometeoroid hit to segment C3 in the period 2022 May 22–24, which caused significant uncorrectable change in the overall figure of that segment. However, the effect was small at the full telescope level because only a small portion of the telescope area was affected. After two subsequent realignment steps, the telescope was aligned to a minimum of 59 nm rms, which is about 9 nm rms above the previous best wave front error rms values. ⁵⁵ Since the NIRCam shortwave channel has the least wavefront error of all the modes (see Table 2 of McElwain et al. (2023), this issue), the increase in wavefront error for all other instruments and modes from the C3 strike should be comparable to or less than NIRCam. 

Images of the telescope optics using the NIRCam pupil imaging lens can reveal smaller impacts below the threshold detectable by wave front sensing. Comparison of pupil images taken 23 February and 2022 May 26 show evidence for 19 such minor strikes over that 92 days period. Regular monitoring of the pupil may help constrain the micrometeoroid hit rate and power spectrum. Micrometeoroid impacts should also slightly lower the telescope throughput; this effect is not yet measurable.

It appears the 2022 May hit to segment C3 was a fairly rare event. Still an open question is how rare --- every year versus every few years. Ten such large hits would degrade the mirror such that it is no longer diffraction limited at lambda < 2 micron. Therefore, after further investigation of the micrometeoroid population (see Section 6.2 of McElwain et al. (2023), this issue), the JWST Project has decided, starting in Cycle 2, to limit the amount of time the telescope spends pointed toward the direction of orbital motion, as those directions statistically have higher micrometeoroid rates and energies.

The level of background emission is critical to the depth of many JWST imaging and low-spectral-resolution spectroscopic observations. In addition to the unavoidable in-field backgrounds from our solar system and our galaxy, as an unbaffled telescope with a non-zero temperature, JWST sees two additional sources of background: (a) the astronomical stray light, mainly affecting the near-infrared wavelengths, in which light is scattered into the field of view; and (b) thermal self-emission from the glowing mirrors, sunshield, and other observatory components, which affects the mid-infrared. Before launch, the predicted levels of these components carried uncertainties of ∼20% and ∼50%, respectively. These backgrounds, now measured from commissioning and early science data, are described in a companion PASP paper (Rigby et al. 2023). Here, we briefly summarize the key results.

First, the astronomical stray light is observed at levels considerably below the requirements, and 20% lower than the pre-launch predictions. As a result, deep fields at high ecliptic latitude will go deeper faster than expected for wavelengths <5 μm.

In the mid-infrared, the main JWST background is a combination of thermal emission from the primary mirror and scattered thermal emission from the sunshield and other parts of the spacecraft. The measured background spectrum is very close to the predicted thermal spectrum that was incorporated into the exposure time calculator before launch, with somewhat higher backgrounds at at 5–15 μm. The measured thermal background at 10 and 20 μm is close to the requirements values. The mid-infrared backgrounds in the F2550W filter are variable at the level of 7% over timescales of weeks to months, which is less variability than expected, due to greater than expected thermal stability of the primary mirrors.

Given these measurements, JWST is indeed, as it was designed to be, limited by the irreducible astronomical background emission, not by stray light or its own self-emission, for all wavelengths <12.5 μm. The control of stray light and thermal emission is an engineering triumph that will translate into substantially better than expected scientific performance for many applications.

During commissioning, a few unexpected stray light features were discovered, characterized and understood, and mitigation plans were developed. Early science operations have shown the success of these mitigations.

There exists a stray light path, termed the "rogue path," that bypasses the primary and secondary mirrors, directly passes through the aft optics system (AOS) aperture just over the back of the fine steering mirror, and reaches the science instrument pick off mirrors (Lightsey et al. 2014). While the rogue path was anticipated, and steps taken to block it to the extent possible, the NIRCam and NIRSS instruments at times show some unexpected scattered light features which have been identified as being due to this path. The resulting features are relatively faint, but for some observations such as deep fields these could be significant noise terms if not mitigated. More information may be found at the JDox Data Features and Image Artifacts page. Here we summarize these features and their mitigations.

5.2.1. Wisps

A few percent of the pixels in four of the eight NIRCam short wavelength detectors show small, faint, diffuse features that are termed "wisps." See Figure 5. In the B4 detector wisps are always present, with variable brightness that is typically about 10% of the zodiacal background. Fainter wisps have also been seen in detectors A3, A4, and B3. Wisps occur at fixed detector positions. The origin of wisps has been traced to reflections from the upper strut that supports the secondary mirror. Wisps have not been seen in the NIRCam long wavelength channel.

Zoom In Zoom Out Reset image size

5.2.2. Claws

5.2.3. The Lightsaber

Some NIRISS images show a narrow band of excess stray light running almost horizontally across the detector. Dubbed "the lightsaber," this light originates from the rogue path (see NIRCam claws above), grazes off the NIRISS entrance housing wall and then experiences double reflections off two mirrors inside NIRISS. See Figure 6. The light comes from a susceptibility region mapped and modeled to be far away from the NIRISS field-of-view (+20 < V2 < +50, +124 < V3 < +128). The zodiacal light and any stars in this region contribute to the intensity. When the light is dominated by the zodiacal light the observed brightness is typically up to 1.5% per pixel of the in-field background. When bright (H_Vega ∼ 0) stars are in that region, the observed brightness is up to 10% per pixel. The lightsaber from bright stars is somewhat narrower and shifts position within the broader band caused by the zodiacal light.

Zoom In Zoom Out Reset image size

5.2.4. Mitigation of the Stray Light Features

JWST has 17 science instrument modes. Near the end of commissioning, the science performance of each mode in turn was reviewed against criteria developed pre-launch for such parameters as sensitivity, image quality, wavelength calibration, astrometric calibration, ghosts, stability, and so on. As of 2022 July 10, all 17 of the 17 modes have been approved to begin science operations. With the performance of the instruments being uniformly excellent versus requirements and typically also better than pre-flight estimates, the assessments of modes as being ready for Cycle 1 science have been quite straightforward. Below we summarize the performance and any known issues for each instrument mode.

A key result of science instrument commissioning is that overall, the JWST science instruments have substantially better sensitivity than was predicted pre-launch. This result is due to higher science instrument throughput, sharper point spread functions, cleaner mirrors, and lower levels of near-infrared stray light background compared to pre-launch expectations. The exposure time calculator (ETC) and its underlying Pandeia engine were overhauled in v2.0 (released 2022 December) to reflect on-orbit performance, timed to support the Cycle 2 Call for Proposals. The JDox documentation has been similarly updated. The ETC and its Pandeia engine are the definitive reference as to sensitivity. For now, we quote several representative measurements and calculations that were determined during commissioning.

In Figures 7 and 8, we summarize the sensitivity of JWST in two common modes: imaging and emission line spectroscopy, and compare to previous and current observatories. We convert from the continuum sensitivity calculated by Pandeia to the more intuitive limiting emission line flux, using the following equation:

$$\begin{eqnarray*}&&{f}_{\mathrm{line}}={f}_{\mathrm{cont}}^{\mathrm{pix}}{10}^{-23}w\sqrt{l},\end{eqnarray*}$$

where f_line is the limiting emission line flux in erg s⁻¹ cm⁻², f_cont ^pix is the per-pixel continuum sensitivity in Janskies, w is the pixel width in Hz, and l is the line width in pixels. 

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Across all instruments, JWST has multiple time series observation (TSO) modes to support observations of transiting exoplanets. During commissioning, observations of the exoplanet HAT-P-14-b were obtained with three NIR spectroscopic modes—the NIRCam grism time series mode, the NIRISS SOSS mode, and the NIRSpec bright object time series mode—to enable cross-comparison between instruments and assessment of astrophysical versus instrumental systematics. This target was chosen because, as a massive planet with a high surface gravity, it is expected to have a flat transmission spectrum. It also has a relatively bright host star (K_s,Vega = 8.9). In addition the transiting exoplanet L 168-9 b was observed with the MIRI LRS time series mode. Results are given in the subsections below for each instrument; the overall picture is that JWST is returning precise transit spectra with minimal processing.

Mode readiness reviews also included the documentation of any remaining work needed to enable particular use cases (e.g., nudging a subarray location slightly to not clip a spectrum; updating an astrometric file needed for target acquisition) or the identification of performance features that observers should be made aware of before final planning of their observations (e.g., alerts regarding faster-than-expected saturation due to higher-than-expected throughput, or warnings regarding stray light features such as the "claws" and "lightsaber," with tips for mitigation). These issues were captured as "liens" on the mode readiness that must be addressed for some specific observing programs (or that are being addressed by the operations team in the near term). Liens have not prevented any Cycle 1 observations from executing, nor have any observing windows been missed due to a lien. Liens have caused delays in scheduling some programs, and some have required workarounds to make some programs executable before a permanent fix for the lien was ready.

Observations not affected by liens are ready for immediate insertion into the observing plan. Following the first several mode readiness reviews, cycle 1 observations of early release science targets began on 2022 June 20. General Observer and Guaranteed Time Observer cycle 1 programs followed.

6.1. NIRCam Performance

The Near Infrared Camera (NIRCam) instrument is described in the companion PASP special issue paper Rieke et al. (2023). Here, we summarize the science performance of NIRCam as characterized during commissioning.

6.1.1. NIRCam Imaging

The throughput of NIRCam meets or slightly exceeds pre-launch expectations for all but a few of the filters in the short wavelength channel. In the long wavelength channel, the throughput is systematically 20% higher than expected for most filters. Figure 9 shows the throughput of NIRCam compared to what was assumed in the pre-launch version of the ETC.

Zoom In Zoom Out Reset image size

The point-spread function is better than expected, as parameterized by encircled energy or full width at half maximum. The photometric stability is stable to at least 4%, and is likely much better. The residual astrometric errors are 2–4 mas per filter across a detector. Optical ghosts are consistent with expectations from ground tests. Some scattered light features are seen (the "claws" and "wisps," described in Section 5.2), which may have some impact on deep imaging.

Table 3 compares the required, predicted, and on-orbit limiting point-source sensitivity of NIRCam imaging, using the flux calibration of 2022∼October. The limiting sensitivity is the flux density of the faintest point source that can be detected at signal to noise ratio S/N = 10 in an integration time of 10,000 s. For a representative wavelength for each of the short and long-wavelength channels, the table quotes the requirements values, the pre-launch predictions from the exposure time calculator, and the predicted performance assuming the measured on-orbit throughput, PSF, and detector noise levels. Table 3 shows that NIRCam imaging is substantially more sensitive than pre-launch expectations.

Table 3. NIRCam Limiting Point Source Sensitivity

Wavelength (μm)	2	3.5
filter	F200W	F356W
Requirement (nJy)	11.4	13.8
ETC prediction (nJy)	10	14.1
Actual (nJy)	6.2	8.9

Note. What is quoted is the faintest flux density (in nanojanskies) that can be detected at S/N = 10 in 10,000 s, for imaging in broad-band filters, assuming a background that is 1.2 times the minimum zodiacal light level. Smaller numbers are better. The equation to convert from flux density f in nJy to AB magnitudes is: mAB = −2.5 log₁₀[f(nJy) × 10⁻³²] −48.57. Actual sensitivities are from Table 2 of Rieke et al. (2023), this issue).

Download table as:  ASCIITypeset image

For the common case of background–limited broadband imaging of a point source, the integration time required (to reach a given S/N on a target of a given brightness) scales as the square of the sensitivity. As such, given Table 3, NIRCam deep imaging should proceed 3.3 and 2.4 times faster at 2.0 and 3.5 μm, respectively, than a system that just met requirements.

We quickly compare this limiting point source sensitivity to Hubble and Spitzer, again considering the faintest point source detectable at S/N = 10 in 10,000 s. For NIRCam at 1.5 μm, the sensitivity is 9 times better than WFC3-IR on Hubble. ⁵⁶ For NIRCam at 3.5 μm, it is 68 times better than IRAC on Spitzer. Again, the integration time required to achieve a given limiting sensitivity, relative to these previous instruments, scales roughly as the square of these advantage factors for background-limited broadband imaging. Clearly, NIRCam imaging should detect faint objects substantially faster than pre-launch expectations.

6.1.2. NIRCam Grism Time-series

Observations of exoplanet HAT-P-14 b taken during commissioning (PID 1442) demonstrated that NIRCam grism time series spectroscopy is working well, meeting performance requirements (Schlawin et al. 2023, this issue) after only simple removal of systematic instrumental noise, known as detrending, by fitting an astrophysical lightcurve model times a polynomial and exponential model that are both functions of time. The noise for this mode is within 150% of the theoretical photon noise. Additional detrending and analysis will presumably move even closer to the photon noise limit.

With simple detrending, the standard deviation of the transit spectrum (as fit by a limb darkened transit model) was 91 ppm at R = 100, compared to the expected photon noise of 55 ppm. The settling time was observed to be 5–15 minutes. The throughput for NIRCam grism spectroscopy is 20%–40% higher than expected for most wavelengths. This may make saturation more likely for approved programs, particularly at long wavelengths.

Figures 10 and 11 illustrate the utility of monitoring the star in NIRCam's short wavelength channel with the weak lens, during such grism transit observations in the long wavelength channel. In this example, such monitoring with the weak lens captured two distinct tilts of primary mirror segments (see Section 4.5.4), which caused jumps in the grism data of 200–1000 ppm, depending on wavelength and extraction aperture size.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

6.1.3. NIRCam Photometric Time-series

This mode is not heavily used in Cycle 1. In commissioning, the mode was checked out using a J_Vega = 14 star with a short 500 s observation, in PID 1068. The performance was nominal: the standard deviation in the normalized flux was measured to be 0.62% (long wavelength channel) and 1% (short wavelength channel), both close to theoretical expectations.

6.1.4. NIRCam Wide Field Slitless Spectroscopy

NIRCam's Wide Field Slitless Spectroscopy (WFSS) has shown excellent performance through commissioning observations (see Rieke et al., this issue). With the F322W2 filter at 3.5 μm, in an integration of 10⁴ s, the continuum sensitivity (S/N=10 per resolution element) is 4 microJy, and the emission line sensitivity (line flux S/N = 10) is 2.8 × 10⁻¹⁸ erg s⁻¹ cm⁻². Figure 12 illustrates the power of the NIRCam/WFSS mode, by showing a serendipitous detection of a line-emitting galaxy at z = 4.39 in the commissioning data. Note the strong detections of the [O iii] 5007 Å and Hα lines in the spectrum, which securely identify the redshift of this galaxy. A similar serendipitous detection of a z = 6.11 galaxy was reported by Sun et al. (2022).

Zoom In Zoom Out Reset image size

6.1.5. NIRCam Coronagraphy

As described in Girard et al. (2022), the performance of the NIRCam coronagraphs exceed expectations. JWST commissioning demonstrated the 5σ PSF-subtracted contrast of the 335R mask at 1'' to be roughly ten times better than requirements, achieving contrasts of ∼4 × 10⁻⁵ to ∼4 × 10⁻⁶. Coronagraphic target acquisition was demonstrated to be on par with requirements.

Stray light features have not been observed in the NIRCam coronagraph fields of view. At this time there is no evidence that stray light will impact coronagraphy or that mitigation strategies need to be taken.

6.2. NIRISS Performance

The Near Infrared Imager and Slitless Spectrograph (NIRISS) instrument is described in the companion PASP special issue paper Doyon et al. (2022). Here, we summarize the science performance of NIRISS as characterized during commissioning.

6.2.1. NIRISS Imaging (Parallel Only)

The imaging performance of NIRISS matches or exceeds expectations. Shortward of 2 μm, the measured throughput is about 20% higher than the instrument team's expectations, and about 25%–30% higher than the ETC's predictions. Longward of 2 μm, the throughput is about 5%–10% better than the instrument team's expectations, and 25% better than the prelaunch ETC predicted. Detector noise properties are similar to as measured on the ground, with a slightly higher "dark current" and total noise, likely due to residuals from uncorrected cosmic ray events. The sky background is lower than predicted before flight.

Table 4 compares the ETC-predicted and actual limiting point-source sensitivity of NIRISS imaging. The limiting sensitivity is the faintest point source that can be detected at signal to noise ratio S/N = 10 in an integration time of ten thousand seconds. The table quotes the pre-launch predictions from the exposure time calculator, and the predicted performance assuming the measured on-orbit throughput, PSF, and detector noise levels, and pre-flight background. This is predicted performance, not measured performance, as commissioning in general did not involve long integrations. Table 4 predicts that NIRISS parallel imaging should detect faint objects substantially faster than pre-launch expectations. The photometric stability is better than 1%, based on two measurements of the same standard star made 16 days apart.

Table 4. NIRISS Limiting Point Source Sensitivity

Wavelength (μm)	1.15	2	3.5	4.4
ETC prediction (nJy)	13	10.2	14.5	22.8
Actual (nJy)	10.0	8.4	11.8	17.9

Note. What is quoted is the faintest flux density that can be detected for a point source at S/N = 10 in 10,000 s. Values are for wide-band filters. Smaller numbers are better. The requirement level was set at 13 nJy for the 3.5 μm filter.

Download table as:  ASCIITypeset image

The field distortion of NIRISS was calibrated using thousands of stars in the LMC astrometric field. Residual astrometric errors with respect to the catalog are 3 mas per axis. This is better than the requirement for accurate targeting of NIRISS sources with NIRSpec multi-object spectroscopy.

The NIRISS PSF is better or similar to pre-flight WebbPSF predictions as defined by encircled energy, FWHM, and ellipticity. There is very little field-dependence of the PSF as measured by these parameters. NIRISS imaging of very bright stars shows an extra diffraction spike offset from the vertical by between −10° and +20° with an angle that rotates smoothly as one moves from the right to the left edge of the detector. This spike is stronger at shorter wavelengths with an integrated intensity ∼70% of the vertical diffraction spike for the shortest wavelength filter, F090W. The cause of this spike has been traced to diamond turning residual wave front errors in some of the NIRISS off-axis mirrors.

NIRISS images show a narrow band of excess stray light running almost horizontally across the detector, dubbed "the lightsaber." See Section 5.2.

Imaging ghosts were seen in ground testing of NIRISS and similar behavior is seen in flight. They are the result of internal reflections in the optical system. Ghost positions are predictable for each filter as they form at a position that is symmetric around the ghost axis point (GAP). The GAPs for each filter were measured during commissioning and the intensity of the ghosts was found to be ∼1% of the original source intensity.

The cosmic ray rate at L2 is similar to predictions. However there is a much higher rate of large "snowballs" that appear as diffuse, mostly circular, events that usually saturate in their centers. See Section 6.6.

6.2.2. NIRISS Single Object Slitless Spectroscopy

Time series observations of a spectrophotometric standard A-star (BD+601753, K_s,Vega = 9.6, PID 1091, duration 5 hr) were used for flux calibration during commissioning. The median precision obtained in order 1 was 147 ppm and order 2 was 206 ppm, binned to a 22 s integration time. This provides an independent test of the stability and precision achieved for a non-variable star, with only a low-level polynomial trend observed in spectral order 1 and no significant trends observed in orders 2 and 3. This observation was affected by a tilt event approximately in the middle of the time series resulting in a flux jump of a few 100 s ppm. The tilt event was easily detected by monitoring the PSF shape along the spatial direction, more specifically by measuring the second derivative of the PSF which is a good proxy of the FWHM. The flux jump was demonstrated to be achromatic.

Observations of the exoplanet HAT-P-14 b (PID 1541, duration 6 hr) returned a point-per-point median precision in the transit depth of 85 ppm at R = 100 for order 1, and 90 ppm at R = 100 for order 2. The weighted scatter of the spectrum itself was 92 ppm for order 1 and 85 ppm for order 2. Errors in the transit depth are within <10%–20% from expectations at this resolution. A tilt event was also noted early in the sequence well before ingress. The NIRISS SOSS mode readily meets performance requirements.

Throughput in this mode is 25% better for order 1 near the blaze wavelength at 1.3 μm, and ∼50% better for order 2. There appears to be no significant noise penalty to operate at 70%–75% of the saturation level. Observations should generally not exceed 35,000 ADUs (56,000 e-) to avoid extra noise. There is also a trade-off with observing efficiency; users may opt to saturate part of the spectrum to improve the duty cycle.

6.2.3. NIRISS Wide Field Slitless Spectroscopy

The throughput for NIRISS wide field slitless spectroscopy (WFSS, Willott et al. 2022; Figure 13) is generally 30% better than expected from the prelaunch ETC. As a result, the predicted on-sky sensitivities (assuming the pre-flight background model) are better than the prelaunch ETC predictions by 7%–20%, similar to the results in imaging mode at these wavelengths. The GR150C filter has higher throughput than GR150R at wavelengths below 1.2 μm, likely due to a different anti-reflective coating, so is preferred for F090W and F115W if only one grism is used (normally both are used to mitigate contamination).

Zoom In Zoom Out Reset image size

Trace positions, curvature, dispersion and spectral resolution for WFSS are close to those measured on the ground. In addition to dispersed versions of imaging ghosts, additional ghosts due to the grisms are seen, but at a relatively low intensity level. The lightsaber scattered light feature is also apparent for WFSS. For the typical intensity where the lightsaber is dominated by zodiacal light, this emission is included in the WFSS background reference files, so will be subtracted off in the pipeline.

6.2.4. NIRISS Aperture Masking Interferometry

The AMI mode (Sivaramakrishnan et al. 2022) features a seven-hole non-redundant mask that enables high-contrast (10⁻³–10⁻⁴) imaging at sub λ/D angular separations (01–05) over three medium-band filters (F380M, F430M, F480M). This mode was successfully demonstrated through the easy detection of AB Dor C, a companion with a separation of ∼03 with a contrast ratio of 4.5 mag. The noise floor of this data set is 6.5–7.0 mag (3σ), very close to the photon noise floor limit of ∼7.5 mag. This is the first space-based demonstration of both infrared interferometry and non-redundant aperture masking. The nominal operational concept for AMI requires staring (rather than dithering) on the science target, followed by a similar observation on an isolated reference star. Target acquisition (TA) places targets at the same detector location, and TA accuracy well within 0.1 pixel was demonstrated. Kernel Phase Interferometry (KPI) is the full pupil generalization of AMI used without the NRM mask but using a similar Fourier-based removal of instrument effects from the data (Kammerer et al. 2023). Like AMI, KPI also enables sub λ/D imaging but with better throughput at the price of a lower contrast at small separation. Observations were performed on 4 targets, all presumed to be single stars, but one was found to have a companion at 015 with a contrast of 1.7 mag. Interferometry with JWST has shown unsurpassed fringe amplitude stability, promising valuable complementarity to ground-based interferometry's significantly higher resolution.

6.3. NIRSpec Performance

A detailed description of the Near-Infrared Spectrograph (NIRSpec) instrument and of its pre-launch performance can be found in a series of four papers: overview (Jakobsen et al. 2022); multi-object spectroscopy (MOS) mode (Ferruit et al. 2022); integral field spectroscopy (IFS) mode (Böker et al. 2022); and exoplanet time series (Birkmann et al. 2022). The on-orbit science performance of NIRSpec is described in the companion PASP special issue paper Böker et al. (2023). Here we briefly summarize those results.

All modes of NIRSpec are working well, in general better than pre-launch expectations. Both of NIRSpec's two detectors show noise levels, in the actual cosmic ray environment of L2, similar to or lower than ETC predictions. The excellent optical quality of the telescope translates to lower-than-expected slit losses in the multi-object and fixed slit modes of NIRSpec. For the 200 mas wide microshutters, this translates to increased photon conversion efficiency of 2.5% at 5 μm, >7.5% below 3 μm, and >10% below 1 μm. NIRSpec bright object time series mode has demonstrated precision in the transit depth of 50–60 ppm per point (Espinoza et al. 2023). Both target acquisition methods that are specific to NIRSpec, wide-aperture target acquisition (WATA) and microshutter assembly target acquisition (MSATA), are working well.

Figure 4 of Böker et al. (2022) shows the measured on-orbit sensitivity of NIRSpec for both multi-object spectroscopy and integral field spectroscopy. Across the board, the sensitivity is better than pre-launch predictions.

The operability rate of the un-vignetted microshutters is 82.5% (Rawle et al. 2022), with electrical short masking now the primary cause of non-operable shutters. The resulting multiplexing levels are within 10% of the pre-launch levels and are still excellent, with the possibility, for high target densities, to observe more than 200 scientific targets at a time at low spectral resolution or close to 60 at medium / high spectral resolution (see Figure 14).

Zoom In Zoom Out Reset image size

6.4. MIRI Performance

The Mid-Infrared Instrument (MIRI; Wright et al. 2023, this issue) is the only instrument on JWST that operates beyond 5 μm; as such it supports a broad range of measurement types: (1) standard imaging; (2) high contrast (coronagraphic) imaging; (3) low resolution spectroscopy (LRS) with and without a slit; and (4) medium resolution integral field unit spectroscopy (MRS).

The MIRI imager uses most of a 1024 × 1024 pixel detector array to provide a field of view of 74'' × 113'' with eight broadband filters, providing bands starting at 5.6 μm and spaced every factor of ∼1.2–1.3 to 25.5 μm. An additional 6% bandwidth filter is centered on the aromatic feature at 11.3 μm. The images are all diffraction limited, and they are Nyquist sampled for wavelengths longer than 6.25 μm. The overall throughput of the OTE and MIRI imager exceeds the prelaunch expectations, particularly for wavelengths at 10 μm and longer. The result is that the imager sensitivity is improved over pre-launch predictions.

JWST provides subarcsec imaging in the mid-infrared, with a beam 50 times smaller in area than that of the Spitzer Space Telescope, the previous most capable space infrared telescope. Figure 15 illustrates this capability, which opens an entirely new realm of study of the structure of mid-infrared sources. Where MIRI is limited by natural backgrounds (for wavelengths of 5 to at least 12.5 μm, Rigby et al. 2023, this issue), its sensitivity for point sources is about 50 times better than that of the Spitzer Space Telescope. This gain is reduced at longer wavelengths due to telescope emission, but in deep exposures the lack of confusion noise with the small beam of MIRI still provides significant gains over Spitzer.

Zoom In Zoom Out Reset image size

Four coronagraphs lie along a side of the imaging field of view, optimized for wavelengths of 10.58, 11.30, 15.50, and 23 μm, and providing raw contrasts of ∼10⁴ (at 6λ/D). The MIRI coronagraphs are performing significantly better than anticipated (Boccaletti et al. 2022), in part due to the excellent image quality of the telescope. Other key factors are the achieved precision of alignment of MIRI and the ISIM to the telescope in pupil shear and focus, both of which are better than the budget allocations. In common with other coronagraphs on JWST, subtraction of a PSF reference star is necessary to achieve the best performance. This technique is very sensitive to high order wave front error (i.e., the shape of the PSF wings) and thus scheduling needs to take account of tilt events and routine telescope mirror alignments.

The MIRI low resolution spectrometer (LRS) also lies to one side of the imager. It has very high throughput (∼80%) and nominal resolution of λ/Δλ = 100, designed for spectroscopy of very faint objects. Its performance is optimized for 5–10 μm, but it is usable to 14 μm. There is also a slitless mode of LRS for exoplanet transits. During commissioning, a transit spectrum of the exoplanet L168-9b was obtained, demonstrating calibration to ∼25 ppm with a spectral resolution of ∼50 at 7.5 μm (Bouwman et al. 2023). LRS transit spectroscopy will be used to study organics and water in the atmospheres of exoplanets, helping determine abundances of these molecules and whether non-equilibrium chemistry is at work influencing them.

The MIRI medium resolution spectrometer (MRS) covers by design the full 5–28.3 μm range (currently calibrated to 27.9 μm). It uses integral field units as its inputs, with fields increasing with increasing wavelength from 32 × 37 to 66 × 77. The spectra and spatial information are arranged over two 1024 × 1024 pixel detector arrays to provide spectral resolution, λ/Δλ, of ≳ 3000 for wavelengths shorter than ∼11.7 μm, and ≳1500 out to 28.3 μm, along with near-diffraction-limited imaging. The high spectral resolution, sensitivity, and imaging capability of the MRS together provide breakthrough capabilities for mid-infrared spectroscopy. For example, MRS gives access to many molecular transitions (e.g., organic molecules, H2) and to the full suite of neon fine structure lines: [Ne ii] 12.81 μm, [Ne iii] 15.56 μm, [Ne v] 14.32 μm, and [Ne vi] 7.64 μm, which are very powerful for determining the excitation mechanism of emission line objects. Figure 16 illustrates the use of these capabilities to dissect the planetary nebula NGC 6543. These data also illustrate well the quality of the calibration of the MRS distortion, fields of view and astrometry achieved during commissioning. While the point source illumination used during ISIM testing on the ground was too faint at mid-IR wavelengths to characterize well these aspects of the MRS before launch, the commissioning data confirm the design expectations.

Zoom In Zoom Out Reset image size

JWST supports parallel observing, in which two science instruments are used simultaneously. There are two types of parallel: coordinated parallels and pure parallels.

6.5.1. Coordinated Parallels

Coordinated parallels are when two science instruments are used together for the same observing program. A total of 170 coordinated parallel visits were successfully executed during science instrument commissioning (Figure 17), and worked well with only one issue (discussed just below). Parallel operations are designed such that mechanism movements of each instrument do not disturb the operation of the other; this coordination is working as designed.

Zoom In Zoom Out Reset image size

While there was no requirement that all of the eleven coordinated templates be exercised during commissioning, in the end seven of the templates were. The four templates that were not exercised are: NIRCam Imaging + NIRISS WFSS; NIRCam WFSS + MIRI Imaging; NIRCam WFSS + NIRISS Imaging; and MIRI Imaging + NIRISS WFSS. We see no reason why these templates should not work, as these WFSS templates are identical to the imaging templates except for putting a grism in the beam instead of a filter.

The one issue with parallel observing that was identified was this: when all 10 NIRCam detectors were used in parallel with another instrument, data from one instrument could under some circumstances be partially overwritten by data from the other instrument. Flight software was patched on 2022 June 24 to fix this issue. Data taken earlier may be affected by partial data overwrite.

6.5.2. Pure Parallels

Observed cosmic ray rates and properties are largely in line with expectations. The vast majority of cosmic ray impacts directly affect only one or 2 pixels, but there are also uncommon events that affect hundreds of pixels. These large events are colloquially termed snowballs. There are also large radiation events that affect the MIRI detectors, which are called "shower" events. ⁵⁷

The current JWST data reduction pipeline handles the first order effects from cosmic ray events but the large number of electrons that result from a snowball event have secondary effects that are not currently corrected in the pipeline. Residuals have a circular appearance with alternating light and dark bands a few tens of pixels across. Dithering exposures is the current recommended mitigation strategy. A four-point or larger dither pattern will allow the pipeline outlier detection routine to significantly improve the final combined image. Work is in progress to improve the calibration pipeline's detection and handling of regions affected by snowballs.

The JWST User Documentation system (JDox) provides comprehensive information about the JWST Proposing process, the Observatory and science instruments, science data characteristics and data access, data pipeline processing and calibration, as well as introductions to post pipeline data analysis tools and training materials.

JDox was updated on 2022 July 12 to describe data features and image artifacts seen in flight data, known shortcomings in current data pipeline products, and other articles to help users understand the status of the Observatory and science instruments. The next major release of JDox coincided with the release of the Cycle 2 Call for Proposals on 2022 November 15.

Users can refer to the Latest Update information box at the bottom of each JDox article to see when that article was updated. JDox also maintains a summary page showing recently changed articles by titles and when they were updated.