Staying Sharp: Keeping Hubble and Webb Seeing Clearly

M. Lallo,, C. Cox,, E. Elliott,, G. Hartig,, M. Perrin,, R. Soummer,, and R. van der Marel,

Hubble and Webb are two different types of space telescopes, operating in different environments. Though the tasks of maintaining their optical alignments share many of the same basic principles, the actual techniques and processes bear less resemblance, and it is interesting to compare and contrast the two telescopes in terms of wavefront sensing and control (WFS&C).

The telescopes

OPTICS-1Hubble’s optical design is a Ritchey-Chrétien Cassegrain telescope with a monolithic, 2.4-meter primary mirror and a 0.3-meter secondary mirror. The mirrors are glass, coated with aluminum and magnesium fluoride. They sit in a graphite-epoxy ring and truss structure approximately 5 meters long (Fig. 1). The Hubble optics are optimized for light at ~0.5 micron wavelength, and form well-corrected images across a region of sky ~20 arcminutes in diameter.

Hubble operates in low-earth orbit, which means that the radiant-heat loads rapidly fluctuate, and that the outer surface of the spacecraft undergoes large temperature swings. To minimize thermal effects on the optical quality, Hubble has layers of insulation to block heat from soaking into the interior, and the two telescope mirrors are actively heated to stabilize them at room temperature, where they have been held throughout the Hubble mission.

Hubble’s single-piece primary mirror is fixed in place. Its mount offers limited low-order figure adjustment via 24 force actuators behind the mirror. Hubble’s primary mirror actuators were planned for use only if the mirror needed post-launch, low-order correction (mostly coma and astigmatism). These actuators were not used because the primary held its shape pre- and post-launch, and because the spherical aberration from Hubble’s primary mirror was outside of the correction capabilities of these mechanisms.

OPTICS-2Hubble’s secondary mirror can be positioned and oriented with six degrees of freedom. Post-launch the secondary mirror was moved multiple times in all six degrees of freedom to best align the telescope. In routine operations however, Hubble’s image quality is maintained by just one adjustment: despace, i.e., moving the secondary mirror purely towards or away from the primary mirror (see Fig. 2).

The Webb telescope is a three-mirror anastigmat—an optical design minimizing astigmatism over a wide field. Webb features a 6.6-meter primary mirror composed of 18 separate, hexagonal segments of beryllium coated in gold. A backplane assembly—a carbon-composite trellis—supports the segments. Hinged, composite tubes will deploy from the backplane assembly and support the secondary mirror. The tertiary mirror is off-axis and directs light to a fourth, flat, steering mirror, which will automatically tip and tilt frequently during an observation to stabilize the image on the science camera (see Figs. 3–4). Webb’s field of view will be similar in size to Hubble’s, about 20 arcminutes across. While the optics are optimized for light at 2 microns—four times longer wavelength than HubbleWebb’s short-wavelength sensitivity overlaps with a portion of Hubble’s, and at longer wavelengths it extends well into the thermal infrared, to beyond 20 microns.




When launched, Webb will take up station at Earth-Sun L2, about 1.5 million km away, or about four times the earth–moon distance. From this location, the earth, sun, and moon remain behind the telescope’s large sunshield, allowing the telescope to passively cool to about 35°K, or –400°F. Small fluctuations in the backplane temperature—much less than a degree—will degrade the optical alignment, and we expect to adjust the primary-mirror segments regularly. Each segment is expected to need such control in six degrees of freedom—a dramatically more complex scenario than for Hubble.

Wavefront sensing and control (WFS&C)

An ideal telescope would convert the flat wavefront of light from a star, situated anywhere in its field of view, into a perfectly spherical wavefront converging to a point on the surface of the detector. Any real telescope, however, introduces diffraction effects, contains irregularities in the mirror surfaces, and exhibits aberrations intrinsic to its design, all of which have the effect of spreading the tight natural distribution of light from a point source like a star. Additionally, errors in the alignment of the optics—in their positions or orientations—further alter the distribution of light.  This instrumental signature is the telescope’s point-spread function (PSF).

For any given optical system there is a well-known mathematical relationship between an image formed by that system and the wavefront corresponding to that image. Via techniques known as phase retrieval, this mathematical relation is algorithmically inverted to solve for wavefront errors from an image, or suite of images, collected by science instruments. In practice, noise, sampling, jitter, finite spectral bandpass, flat-fielding error, and detector dynamic range all limit the ability of phase retrieval to determine the wavefront from the images.  There are numerous approaches to mitigating these error effects, such as defocussing (discussed later) and/or collecting multiple images to span the field-of-view and range of wavelengths. These mitigation approaches are termed “diversity,” and they can significantly increase the accuracy and speed of phase-retrieval-based wavefront sensing.

Webb will use focus diversity by shifting the focus in the NIRCAM instrument through use of insertable lenses mounted in that camera’s filter and pupil wheels. It will use field diversity to separate errors in the primary mirror segments from misalignments of the secondary mirror. In routine operations, Hubble requires neither focus diversity nor field diversity to obtain wavefront information sufficient for maintaining image quality at the science instruments.

For both missions though, the wavefront sensing is or will be performed by periodically using phase retrieval to analyze star images (effectively point sources) taken with the same cameras that are used for the science to determine the wavefront error. Some part of this will be intrinsic to the telescope and unavoidable, while some amount will be due to the correctable misalignments and overall shape of the mirrors. Detailed models of the complete optical systems are used to separate these two contributions. In both missions, this wavefront sensing analysis is done on the ground and is not autonomous. Based on the results, we may decide to control the mirror positions to reduce this error, thus restoring the desired image quality.

Science-image-based phase retrieval has a history that precedes Hubble, and was proposed by Perkin-Elmer for use with the Hubble mission. Post-launch this approach quickly become the method of choice for characterizing Hubble by groups both inside and outside of the Institute, accurately validating Hubble’s corrective optics axial replacement instrument (COSTAR), by verifying through image analysis the predictions of pre- and post-COSTAR wavefront errors. These successes led ultimately to exploration of the overall approach and its selection and validation for Webb by groups at NASA/GSFC, NASA/JPL, the Institute, BATC, and in academia.


Hubble WFS&C

Over Hubble’s long mission, we have found that the primary and secondary mirrors do not change shape enough to be noticed in the PSF. Tips, tilts, and off-center motions of these mirrors are also so small that they go unnoticed. We do, however, see significant changes in focus caused by variations in the distance between the primary and secondary mirrors. These disparities are caused by small temperature changes in the truss as Hubble orbits the earth and points in different directions (see Fig. 5). The small but measurable focus errors due to temperature changes have been reasonably well understood and modeled, and no mirror control is performed to compensate for this effect, which occurs over hours to days.



A more significant effect on Hubble’s focus is the long-term, gradual shrinkage of the entire truss due to outgassing of water and other molecules from the graphite-epoxy material. This outgassing has caused Hubble’s focus to trend throughout its mission, as shown in Figure 6. If left uncorrected, by now this effect would have built up a wavefront error of many times the design specification. To compensate, we have commanded dozens of focus adjustments, each time backing the secondary mirror away from the primary mirror. These adjustments are typically 3–4 microns every few years, based on monthly measurements of the PSF.

For Hubble, the phase retrieval method used to determine wavefront error from the observed PSF expresses the result in terms of a polynomial of aberration terms. The software is typically set to solve for only certain “low-order” aberrations like focus, astigmatism, and sometimes coma, but among these, only the focus error varies enough to require periodic corrections. The Hubble phase retrieval operates on science images that are very close to being “in focus.” The sharp star images formed by the telescope are also undersampled at the detector, meaning that the PSF falls on fewer pixels than needed to optimally represent it (Fig. 7). This can present a challenge to a fitting algorithm like phase retrieval, but due in large part to the limited parameters that are being fit, we can accurately determine the amount of defocus even in a sharp Hubble PSF.


Webb WFS&C

Unlike Hubble’s truss, the large, cold, carbon-composite structure holding Webb’s secondary mirror and 18 primary mirror segments is not expected to shrink significantly over time. The structure will, however, unavoidably deform when repointing of the telescope changes the angle of the sunshield with respect to the sun. Over the mission of five or more years, the sunshade material will degrade, causing temperatures to rise, which will produce a long-term evolution of the PSF. We may also find the PSF affected by unanticipated mechanical changes in the structure. The Webb WFS&C program is designed to address these issues and deliver an optimal PSF to the science instruments.

Compared with Hubble, the more complex telescope design of Webb requires the phase-retrieval algorithm to extract much more information about the wavefront from the observed PSF in order to guide more complex control of the mirrors. Instead of simply determining a single parameter (i.e., focus aberration), which scales directly with secondary mirror despace, the sophisticated algorithms developed to support Webb will produce a map of the best fitting observed wavefront, compare that with the required or a desired target map, then determine the set of commands needed to control the positions and orientations (“poses”) of each of the 18 segments to achieve that target wavefront. The relationship between the mirror poses and the wavefront is much more complicated for Webb than with Hubble, and sometimes the solutions will be degenerate, with various sets of adjustments producing indistinguishable PSFs. In such cases, practical considerations— economy of motion, conservation of the actuators, and ensuring that mirror segments avoid contact with each other—will influence the final choice of mirror commands.

Star images taken through NIRCam’s special lenses will increase the amount of information measurable in the PSF by introducing a precisely known defocus. This spreads the PSF out over more pixels, facilitating a more accurate and detailed phase retrieval (see Fig. 8).


Without corrections, Webb’s total wavefront error could wander outside of its specification over periods of days to weeks, depending on the amount of thermal perturbations. Therefore, the current plan is to correct the aberrations as needed on an approximately biweekly timescale, based on wavefront sensing every two days. Figure 9 shows one particular model of possible fluctuations of the wavefront error during a year of representative science observations. As was the case with Hubble, much about Webb’s stability will not be known until it is operating on orbit. We will continue to refine models before launch, and utilize a control scheme that is flexible enough to adapt as we better understand behavior in flight and develop optimal control cadences.

Webb’s initial optical alignment

We have described how Webb’s WFS&C program will maintain the optical quality of the PSF during routine science operations, and compared the program to Hubble’s. The first test of these WFS&C procedures will be the initial alignment of the optics after the backplane assembly and secondary-mirror supports are deployed, when gross alignment errors at the millimeter level will be removed. These are very large adjustments compared with the tens of nanometers of fine control typically expected for the routine WFS&C program to follow. Nevertheless, the same wavefront sensing and mirror management tools will be used in both cases. The one-time WFS&C program after deployment will also involve special operations to support the unique conditions and scenarios that are part of the observatory commissioning.

The approach to performing Webb’s WFS&C resulted from a joint effort of NASA/GSFC, NASA/JPL, Ball Aerospace and Technology, Northrop Grumman Aerospace Systems, and the Institute. Ball will also be responsible for the initial alignment of the telescope.



PSF modeling tools

The Institute provides tools to the community to model Hubble and Webb PSFs:

Tiny Tim