Open Access
Add to BibSonomy
Abstract
Current imaging spectrometer forms for terrestrial remote sensing in the visible, near-, and shortwave infrared (VNR/SWIR) spectral range have been implemented in hardware and achieve a high level of performance in terms of both aberration control and signal-to-noise level. These forms are compact, relative to prior art, but more size, weight, and power optimization, while maintaining performance, is desirable for usage on small satellite platforms. Pursuant to that goal, we have developed a compact breadboard prototype VNIR/SWIR imaging spectrometer that maintains the current aberration control and has a large number of spatial samples. The new form utilizes a catadioptric lens and a flat dual-blaze immersion grating yielding a compact design that is relatively easy to manufacture.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. INTRODUCTION
In order to achieve broad usage of imaging spectrometers they must be relatively easy to manufacture for a precision optical system, optically stable, adaptable to a range of operational requirements, and have reduced size, weight, and power (SWaP) characteristics. The current work presents the design and breadboard results that reflect an effort to meet these requirements. The Chrisp Compact VNIR/SWIR (visible, near-, and shortwave infrared) Imaging Spectrometer (CCVIS) is a VNIR/SWIR imaging spectrometer with a significant reduction in SWaP while maintaining performance. Its small volume enables a modular implementation where individual spectrometers are stacked in order to acquire spectral imagery over a wide swath width when coupled with a freeform telescope. The CCVIS employs a flat dual-blaze immersion grating that is easily manufactured in bulk using grayscale photolithography, a process that is much easier to implement in grating production than similar e-beam lithography, and has the same accurate and precise metrology eliminating periodic ghosts. These gratings are relatively large and employ a tailored facet profile that optimizes optical efficiency. In addition to the SWaP advantage, the small size facilitates thermal stabilization for radiometric and spectral stability. The breadboard results demonstrate the capability of the design form for application to Earth science investigations and for planetary missions. The flexibility of the basic design and the optimization approaches are demonstrated with an aquatic remote sensing design relying on multiple CCVIS stacked behind a freeform telescope that readily fit on a smallsat platform.
Current state-of-the-art of imaging spectrometer designs are provided in the next section followed by a description of the CCVIS that uses a catadioptric lens and a flat dual-blaze immersion grating yielding factors of 10 or more smaller volumes when compared to similarly performing designs such as the Offner–Chrisp and the Dyson. A description of the technology developed to produce the grating and slit needed for CCVIS is provided in Sections 5 and 6, and Section 7 describes the alignment approach for the design as well as breadboard results demonstrating that the CCVIS controls smile to ${\lt}{0.1}$ pixels across the swath. Keystone was found to be larger than expected, being most likely due to a grating clocking error.
2. CURRENT OPTICAL FORMS
A plethora of optical designs have been developed as imaging spectrometers, and we will address only the application of two general forms based on microlithographic projectors that are in wide usage. Modern focal plane arrays (FPAs) have overcome many of the problems that plagued the first imaging spectrometer, the Airborne Imaging Spectrometer developed at the Jet Propulsion Laboratory (JPL), and have enabled the application of these novel new designs [1]. In particular, the development of substrate removed HgCdTe detectors that are sensitive to the full solar reflective spectral range on a single FPA has revolutionized imaging spectrometers 2. Additionally, pioneering work at JPL in the design and fabrication of gratings has improved the overall optical efficiency achievable.
The microlithographic projectors (Fig. 1) that have been adapted for imaging spectrometers were invented by Abe Offner and J. Dyson for mask projection applied to the production of integrated circuits and for photographic generation of gratings, respectively [3,4]. These optical forms are concentric and telecentric with the pupils projected to infinity. It is relatively simple to show that the Seidel aberrations are zero for both cases.
Fig. 1. Offner form (top) with the secondary mirror as the limiting aperture and Dyson form (bottom) with the concave mirror as the limiting aperture; ${f_1}$ is the focal length of the thick lens of index $n$. O and I signify the object and image locations, C is the center of curvature, and R is the radius of curvature of a particular surface. The grating replaces the convex mirror in the Offner form and the concave mirror in the Dyson form. Adapted from [5].
The adaptation of these projectors as imaging spectrometers requires the introduction of a grating. For the Offner form, the convex secondary mirror is replaced with a convex grating, while for the Dyson, the single concave mirror is replaced. This breaks the symmetry in both cases and introduces a host of aberrations that must be controlled. The aberrations introduced by spherical gratings have been studied by spectroscopists for mounting arrangements to support applications that require high spectral resolution ($R = \lambda /\Delta \lambda$). The primary point aberration introduced is astigmatism, and both adaptations control it at the level of the typical pixel pitch (${\sim}20 \;{{\unicode{x00B5}{\rm m}}}$). The field aberration of concern is distortion, which is also well controlled at the level of 0.05–0.1 of a linear detector element dimension. In the literature, this is referred to as high spatial–spectral uniformity and separated into “smile,” curvature of a monochromatic image of the slit, and “keystone,” variation in the magnification of the slit as a function of wavelength [6]. Of these two manifestations of distortion, keystone is the most deleterious, resulting in spatial mixing as a function of spectral channel.
Figure 2 shows examples of the imaging spectrometer forms adapted from microlithographic projectors. The adaptation of the Offner form was fully developed by Michael Chrisp and has since been extensively utilized in a variety of systems [7,8]. An example of a similar form proposed by A. Thevenon for a Littrow spectrometer is presented in [9]. Chrisp’s design, which is now standard, separates the concave primary mirror into primary and tertiary mirrors, increasing the degrees of freedom for aberration control. González-Nuñez et al. performed a thorough analytical analysis and demonstrated that astigmatism can be controlled over a broad range of wavelengths [10]. These imaging spectrometers tend to have F-numbers of 2.5 or higher. An example of an Offner–Chrisp is the JPL Hyperspectral and Infrared Imager (HyspIRI) F/2.8 design for the solar reflective spectral range (380–2500 nm) with 212 channels [11]. The distance from the slit to the primary is about 32 cm and from the top of the primary to the bottom of the tertiary about 20 cm.
Fig. 2. Offner–Chrisp (top) and Dyson imaging spectrometer forms (drawing not to scale). The Dyson illustrated is the JPL CWIS design adapted from [13].
The adaptation of the Dyson form has been accomplished for both the VNIR/SWIR and the longwave infrared (approximately 7.5 to 13.5 µm) spectral ranges [6,12]. A VNIR/SWIR example of the Dyson form is the JPL Compact Wide Swath Imaging Spectrometer (CWIS) [13]. CWIS uses a doublet lens rather than the original single plano-convex lens in order to meet the aberration control requirements. The design is compact with dimensions of 32.5 cm in the long axis and a diameter of 12.5 cm, is optically fast with an F-number of 1.8 to optimize the signal-to-noise ratio (SNR), and has a wide swath of 1600 samples and finer spectral sampling (7.4 nm verses 10 nm) in comparison to the HyspIRI design. The Dyson form is more compact than the Offner–Chrisp for an equivalent F-number, spectral samples, and swath. Both forms have been implemented in hardware with the performance demonstrated.
3. CHRISP COMPACT VNIR/SWIR IMAGING SPECTROMETER
We have consulted the most recent Decadal Survey conducted by the National Academies of Sciences, Engineering, and Medicine to establish the broad requirements for a state-of-the-art imaging spectrometer for marine and terrestrial remote sensing science questions [14]. The relevant parameters that drive the spectrometer development are the spectral sampling properties, and the SNR performance. The spectral range is 380–2500 nm with 10 nm sampling. We have interpreted the sampling to be the distance between adjacent spectral channels rather than the full width at half maximum (FWHM) of an instrumental profile. The required SNR performance is 400:1 for the VNIR (380–1050 nm) and 200:1 for the SWIR (1050–2500 nm), with a reference radiance due to a spectrally flat, Lambertian reflectance, and an integration time of about 5 ms. The solar zenith angle was not specified but the intended orbit is Sun-synchronous, so a 23.5° solar zenith angle is appropriate. No other optical characteristics are addressed in the Decadal Survey.
Current dispersive imaging spectrometers have a high degree of spatial–spectral uniformity and are optically fast, maximizing the SNR performance. The challenge is to create new optical forms that maintain the current performance but reduce the SWaP requirements to enable the deployment of these powerful spectral sensors on unmanned aerial systems and small satellite platforms. The reduction in SWaP also facilitates the stabilization of the sensor since small temperature changes can lead to misalignment resulting in data calibration errors. An additional goal is comparable manufacturability. All imaging spectrometers have challenging design and alignment tolerances, and this last goal addresses both component fabrication and overall alignment. The broad imaging spectrometer goals are to
- 1. maintain the spatial–spectral uniformity to 0.1 or better of the pixel pitch,
- 2. have 1500 or greater spatial samples,
- 3. have 10 nm or better spectral sampling across the VNIR/SWIR range,
- 4. have as low an F-number as possible for SNR performance with an ultimate goal of $\le 2.5$,
- 5. reduce the SWaP,
- 6. minimize systematic errors such as scattered light,
- 7. maximize system robustness for spectral and radiometric stability, and
- 8. facilitate manufacturability.
With these goals in mind we have developed a compact breadboard prototype imaging spectrometer with a 1500 spatial sample swath width that records spectral imagery over the solar reflective spectral range at 10 nm spectral sampling. The current prototype, depicted in Fig. 3, comprises a catadioptric lens and a flat dual-blaze immersion grating [16]. The catadioptric lens consists of a concave meniscus lens with a reflective coating on the back. The lens shape creates a concave mirror that, together with the negative lens power, corrects the Petzval field curvature for the spectrometer. The grating, whose development is described below, is much simpler to manufacture in comparison to either the Offner–Chrisp or Dyson forms that employ powered gratings. The optical speed of the system is F/3.3 with a length of 7 cm and an 8 cm diameter. The CCVIS utilizes a single focal plane array, and the slit is matched to the pixel pitch since the magnification is one. The volume of the CCVIS is ${{350}}\;{\rm{cm}}^3$, which is about a factor of 11 smaller than CWIS. The comparison is not one-to-one since the CWIS is an F/1.8 system as compared to the F/3.3 CCVIS.
Fig. 3. CCVIS F/3.3 optical form. Figure previously published in [15]. The baffles illustrated in the upper figure have been added.
The CCVIS spectrometer design was developed based on previous experience in designing imaging spectrometers. Designing catadioptric infrared imaging spectrometers showed the useful technique of balancing the opposite signs of the field curvature from positive lenses and a positive mirror to generate a flat focal surface, without the need for a powered grating as in the Offner–Chrisp or Dyson designs [17–19]. Work on apochromatic VNIR/SWIR imaging spectrometers showed that the 400 to 2400 nm wavelength region could easily be focused onto a single focal plane [20]. The CCVIS spectrometer combined both these principles in order to meet the size requirements for an extremely compact imaging spectrometer.
The CCVIS optical form is well corrected for both point and field aberrations. Point aberrations were evaluated based upon an 80% encircled energy diameter metric as a function of spatial–spectral sample location through the field, as shown in Table 1 with the corresponding spot diagrams in Fig. 4. The largest diameter was about 9 µm, which is well contained within the typical detector element size, which is on the order of 20–30 µm. Distortion, the primary field aberration, was evaluated in terms of smile and keystone with the maximum values of ${\pm}0.2\;{{\unicode{x00B5}{\rm m}}}$ and ${\pm}0.3\;{{\unicode{x00B5}{\rm m}}}$, respectively. The tolerances required for the assembly and manufacture have been estimated, indicating that the as-built smile and keystone will remain within ${\pm}0.9\;{{\unicode{x00B5}{\rm m}}}$, or about a 20th of an 18 µm detector element. The contributions of stray light have also been analyzed, and baffling has been added.
Table 1. 80% Encircled Energy Diameter (mm) as a Function of Wavelength and Slit Position
The CCVIS has also been evaluated for stray light contamination due to reflections from the front and back surfaces of the lenses and the front surface of the immersion grating. Each of these surfaces includes an anti-reflection coating that reduces the reflected light to below 1.5%; however, this can still produce measurable contamination at the FPA in the absence of mitigation steps.
The baffle positions were determined from a stray light analysis performed in FRED [21]. The model was initialized with a reflective detector surface and a reflective slit surround, and the grating was set to propagate positive and negative orders up to the third order. Multiple reflections were allowed at each lens surface. Then the ghost image paths that made their way to the detector were analyzed and mitigated through bending the optical elements, changing the angle of the grating, and by baffle placement. This analysis led to the introduction of two baffles placed roughly symmetrically on either side of the immersed grating and extending along the long axis of the spectrometer. The baffle next to the detector eliminated the path of the stray light reflection from the back surface of the second lens, which was refracted by the front surface towards the detector. A third relatively short baffle was placed in the center of the catadioptric lens extending toward the grating. The combination of these baffles eliminated almost all of the stray light due to single and multiple unintended surface reflections. As an example of stray light that does make it to the detector array, bounces from the front surface of the second lens both prior and posterior to the grating are reduced to less than 0.002%. The order-sorting filter also reduces these types of systematic errors. Additionally, a black absorbing baffle was added at the edge of the FPA for trapping the zero-order light.
One area of improvement for the CCVIS is to lower the F-number as much as possible while maintaining aberration control to increase the étendue for superior SNR performance. The redesigned imaging spectrometer is F/2.5, with 1600 spatial by 200 spectral channels utilizing a focal plane array with an 18 µm pixel pitch. The diameter has increased slightly to a new value of 8.3 cm, which represents a 7.6% volume change. Point aberration control remains excellent with the 80% encircled energy diameter being 5 µm on average, well within the pixel pitch. The maximum keystone and spectral smile were ${{\pm 0.4}}\;{{\unicode{x00B5}{\rm m}}}$ and ${{\pm 0.2}}\;{{\unicode{x00B5}{\rm m}}}$, respectively, again within the goal of less than ${{\pm 0.9}}\;{{\unicode{x00B5}{\rm m}}}$. The smaller F-number results in 1.7 times more optical throughput compared to the F/3.3 system. This is approaching the lower F-number limit for this optical form.
The SNR performance of the CCVIS is calculated by predicting the signal based upon an at-aperture radiance model and calculating the noise components [22]. The signal is modeled using the measurement equation, given by
Figure 5 shows the expected radiometric performance for F/2.5 design with 210 spectral channels coupled to a three-mirror anastigmat (TMA) foreoptic. The reference radiance employs a spectrally flat reflectance value of 25%, the mid-latitude summer standard model with a 23 km visibility, a nadir viewing geometry, and a 23.5° solar zenith angle. The SNR does not meet the Decadal Survey requirement at all wavelengths outside of the atmospheric absorption features and could be improved through the utilization of a digital FPA with superior noise performance. The optical efficiency is dominated by the grating performance with the higher frequency ripples due to measured order-sorting filter data from a previous spectrometer build. The dual-blaze profile, which produces higher optical efficiency at longer wavelengths where the solar radiance is reduced, could be modified to improve the performance as was demonstrated on the Mars Mineralogy Mapper [23].
Fig. 5. Reference radiance (top), and CCVIS signal-to-noise performance and optical efficiency. The reference radiance graph includes the high spectral resolution model and band-averaged radiance. In the bottom graph, the solid line is the SNR curve corresponding to the F/2.5 CCVIS, and the dashed line is the optical efficiency.
4. DESIGN EXCURSION IN SUPPORT OF AQUATIC ECOSYSTEM SCIENCE
The application of the CCVIS to a particular science question is illustrated by addressing the very challenging problem of remote sensing of aquatic ecosystems using a space-based imaging spectrometer. The Decadal Survey has recommended that a high-fidelity imaging spectrometer is needed to quantify the physiological dynamics of aquatic primary producers, and for the spatial and temporal quantification of the distribution of functional traits, functional types, and composition of marine biomass [14]. These are particularly difficult remote sensing problems due to the very low reflectance of water bodies with much of the signal due to atmospheric processes such as scattering, particularly in comparison to remote sensing over land masses. To meet this challenge, an imaging spectrometer must be developed with attributes that specifically address the spatial, spectral, radiometric, and temporal requirements as detailed in the Committee on Earth Observation Satellites (CEOS) “Feasibility Study for an Aquatic Ecosystem Earth Observing System” report, referred to herein as the CEOS report [24].
The CEOS report defines the requirements for an aquatic ecosystem imaging spectrometer. These include a narrower spectral range reduced to the VNIR (380–1050 nm) and an optimal spectral resolution of 5 nm for the FWHM of an instrumental profile. The signal-to-noise (SNR) requirements are similarly tailored with a very challenging threshold noise equivalent spectral radiance (NESR) of ${0.01}\;{\rm{mW/}}{{\rm{m}}^2}{\rm{sr}}/{\rm{nm}}$ for a prescribed reference radiance. Additionally, the ground sample distance (GSD) is baselined at 17 m with a threshold of 33 m. The temporal requirement is somewhat ambiguous and “is as high as is technologically possible and affordable” with a recommendation for an optimal one to five day revisit time. The reference design developed here for an aquatic ecosystem hyperspectral imager is based on the CEOS report requirements.
The implementation of a space-based system that meets these requirements would comprise a constellation of small satellites in a low Earth orbit, placing a premium on SWaP. The enabling technologies are the CCVIS with its compact form and a wide field telescope that enables side by side stacking of individual modules, as shown in Fig. 6. The three slits are positioned by a mirror field splitter to provide a slight overlap for data acquisition using a pushbroom mode. The challenge is to produce a high-quality image at the aggregated slit. This is accomplished using the Fast Accurate NURBS optimization (FANO) code developed for freeform optical designs with the surfaces under local control [25,26]. The FANO optical design code utilizes non-uniform rational basis-spline (NURBS) surfaces where a surface is defined by a set of local grid control points that can be changed independent of other grid points outside of the local area. This local control enables the optimization of complex surfaces with optical designs that have a high degree of aberration control over a wide field. The approach allows for arbitrary surface control and is more powerful than Zernike or $xy$ polynomial-based freeform designs. The alignment of a NURBS-based freeform telescope is more challenging, due to the lack of symmetry in the optical elements, requiring additional alignment fiducials and testing with computer generated holograms. The NURBS freeform design also enables relaxed tolerances for the equivalent performance of a more conventional design, or, conversely, superior optical performance for the same alignment tolerances. A preliminary design of the TMA foreoptic for the three-spectrometer design is illustrated in Fig. 6 with an 11.7º field of view in the cross-track direction. The TMA has a stop on the secondary and is designed to provide a telecentric beam for the spectrometer, which positions the pupil on the grating.
Fig. 6. Space-based wide swath imaging spectrometer comprising three CCVIS modules and a NURBS freeform TMA.
The reference design utilizes a freeform telescope with three stacked CCVISs, each with a spectral range from 380 to 1050 nm for the aquatic mission, deployed to a low Earth orbit satellite platform at 500 km. Combining the altitude with the F/2.5-CCVIS, an FPA pixel pitch of 30 µm, and the 17 m nadir GSD yields an entrance pupil diameter of 35.3 cm. Even with this optically fast system, the SNR performance over lower radiance aquatic regions necessitates a slow frame rate of 10 Hz. The imagery can be collected using a pitchback maneuver where the imaging spectrometer pointing is accomplished through satellite motion with the optical axis angled forward initially and rolled back as the satellite passes over the ground location and the data collection progresses. This GSD will vary depending upon the distance from the satellite to the projected slit image on the surface and is limited to a maximum of 33 m at the beginning and end of the pitchback maneuver and a nadir value of 17 m based on the CEOS report. Modeling the pitchback maneuver yields a 21.7 km swath length, a nadir swath width of 82 km, and an imaged area of ${{2018}}\;{{\rm{km}}^2}$. There will be gaps in the data as compared to a pushbroom sensor such as Landsat, but these can be minimized by an agile satellite bus that rapidly slews and settles for the following data collection as required for acquisition of different aquatic sites on the same orbital pass.
The CEOS radiometric performance requirement, expressed as the NESR, is based upon a reference radiance of ${{100}}\;{\rm{mW/}}{{\rm{m}}^2}{\rm{sr}}/{\rm{nm}}$ for the 380 to 650 nm spectral range and ${{20}}\;{\rm{mW/}}{{\rm{m}}^2}{\rm{sr}}/{\rm{nm}}$ for 650 to 1050 nm. These low values, as compared to the reference radiance presented in Fig. 5, reflect the challenging aquatic ecosystem measurement problem. The radiometric performance of an imaging spectrometer depends upon the étendue, which is directly proportional to the area of a detector element and inversely proportional to the square of the F-number. It may be possible to modestly decrease the CCVIS F-number, but a bigger gain is to increase the detector element size. We have modeled the CCVIS radiometric performance by utilizing a TIS FPA with Hybrid Visible Silicon Imager (HyViSI) silicon PIN detectors and an analog readout integrated circuit (ROIC) with a 30 µm pixel pitch, increasing the étendue by a factor of 2.8 over an FPA with an 18 µm pixel [27]. Our preliminary assessment is that the NESR requirement would be met for much of the 650 to 1050 nm range but fall short by a factor of 1.5 to two from 450 to 650 nm, with poorer performance in the 380–450 range. The application of the CCVIS to the aquatic problem would benefit from a larger pixel pitch such as produced by aggregating four 18 µm detector elements into a 36 µm effective pixel. A digital ROIC that supports this pixel aggregation without a readout penalty would additionally improve the NESR.
An Evolved Expendable Launch Vehicle (EELV) Secondary Payload Adapter (ESPA) class satellite is well suited for the imaging spectrometer designed to address aquatic ecosystem science since the entrance pupil diameter and the bus agility requirements make it difficult to produce a cubesat design. The baseline satellite is therefore an ESPA class with a maximum volume of ${{61}} \times 71 \times 96.5\,\,{\rm{cm}}^3$, which will have sufficient capability to perform the required maneuvers as well as the supporting infrastructure necessary for the data acquisition, processing, and downlink [28].
5. GRATING DEVELOPMENT AND FABRICATION
Both the Offner–Chrisp and Dyson forms require gratings that are challenging to manufacture due to the powered substrate and the facet profile. Gratings are by their nature inefficient optical components in terms of optical transmission. In a laboratory monochromator or spectrometer, the inefficiency is improved by using multiple gratings that are blazed at different wavelengths. A single blaze grating is efficient at the blaze wavelength and declines at both shorter and longer wavelengths relative to the design wavelength. In a laboratory system, the gratings are used sequentially to acquire the desired spectrum. The parallel implementation for an imaging spectrometer requires building multiple spectrometers to cover the desired spectral range and then go through a rigorous co-alignment process. The NASA Hyperion sensor used this method. An easier approach is to use multi-angle blaze grating facets and a single focal plane array, such as a substrate-removed HgCdTe array. The multi-angle facets improve the overall efficiency although they remain inherently inefficient devices. By thoughtfully shaping the facets, the SNR performance can be weighted so that it is approximately equal for all of the spectral channels, as shown in Fig. 5. JPL researchers have pioneered this approach manufacturing these critical components using e-beam lithography to produce high-quality gratings on curved substrates, and industry has also developed similar gratings using diamond machining techniques [29,30]. One advantage of the CCVIS is that it uses a flat grating that simplifies the manufacturing process.
A microfabrication-based approach using grayscale photolithography as reported by Smith et al. [31] was used for the grating development. Depending on the end application, the resist can either be coated with a reflective metal or an etching process can be used to transfer the shapes in the resist into a more stable and robust underlying film, and then coated with a reflective metal. One manufacturing advantage of this approach is that once the process has been perfected, grating production is rapid and robust.
The scattering properties of a grating have long been of interest to absorption spectroscopists where absorption features are measured against a spectrally broad source. Even though the degree of scatter at a particular wavelength is small, the resulting measurement will be contaminated by scatter from the totality of wavelengths accessible in the measurement. This problem effects data products from imaging spectrometers by filling in absorption features and rounding off peaks. These atmospheric absorption features are used in data analysis to quantify the amount of a given molecular species in order to perform reflectance retrieval. Spectroscopists have attacked this problem in hardware, using, for example, double monochromators, and through correction factors [32]. Correction techniques are applicable to imaging spectrometer data; however, it is highly desirable to reduce the scatter as much as possible to minimize the correction.
The grating is the dominant source of scatter in an imaging spectrometer where the other optics are manufactured with best practices and the device is thoroughly baffled. Measurements of gratings indicate that scatter falls rapidly from a maximum value for the design order as a function of distance or rotation angle. This falloff will be about five orders of magnitude for a quality grating.
Fig. 7. Grating scatter measurements as a function of grating angle displayed on linear (upper) and semi-logarithmic (lower) plots. Inset in the upper graph is the dual blaze facet shape. The different orders, positive and negative first and zeroth, are labeled.
Fig. 8. Comparison of grayscale photolithography gratings (left two images) to a precision machined grating. Arrows indicate evidence of errors in each grating. A cleaner diffraction pattern is observed in the grayscale photolithography grating (middle) when multiple exposures and mask shifts are implemented in the patterning process.
We have made measurements using an apparatus based upon that presented by Woods et al [33]. A red helium–neon laser (632.8 nm) illuminates a pinhole to produce a point source, which is subsequently collimated using an off-axis parabola (OAP). The collimated light is reflected from the grating under test, focused by another OAP onto a pinhole, and measured using an avalanche photodiode. The noise floor is between five and six orders of magnitude below the peak value based upon measurements using a high-quality optical flat. The relative results are presented in Fig. 7 for the dual facet profile indicated. The design order is ${-}{{1}}$, and the data are normalized to the order peak. These results indicate a level of performance comparable to other gratings.
To evaluate the quality of the gratings that were manufactured using grayscale photolithography, another grating was fabricated using industry standard precision diamond machining for comparison. The diffraction patterns and surface roughness were measured for each of the gratings and are shown in Fig. 8. The precision machined grating has superior surface roughness and therefore lower scattered light contamination; however, there is substantial noise between the diffraction peaks. For the single exposure shown to the left in Fig. 8, the grayscale grating shows clear diffraction peaks, but with undesirable ghost diffraction. While the roughness resulting from many photolithographic and etching processes tends to leave surfaces rougher than traditional machining techniques [34–36], grayscale photolithographic approaches are typically at least an order of magnitude more precise and accurate than traditional precision machining techniques [37,38]. While both types of gratings show non-idealities that are attributable to manufacturing artifacts, there is an opportunity to improve RMS surface roughness for the grating fabricated using grayscale photolithography through mask redesign or resist process optimization.
The observed ghost diffraction patterns can be attributed to periodic errors in the patterned resist that are most likely the consequence of stitching or beam deflection errors. Similar errors were also seen in diffraction gratings fabricated with scanning electron beam lithography by Wilson et al. [39]. As our photomask was made using a direct write technique, one can expect errors in that write process to be captured in the photomask. There are a number of approaches that can reduce both surface roughness and periodic errors. For example, Wilson et al. showed that using multiple field sizes during the write process resulted in decoherence of diffraction ghosts caused by field boundaries [39]. Resist patterning process modeling enabled by PROLITH [31] suggests that increasing the post-exposure bake temperature or reducing the pitch between subresolution features of the photomask may work to reduce surface roughness. Another approach, which was implemented in this work, is to run multiple exposures while inducing a shift in the mask location for each exposure. The net effect of this approach is the same decoherence that Wilson et al. [39] showed by using multiple field sizes. The resulting optical profile and diffraction pattern are shown in Fig. 8. The grayscale grating with multiple exposures and shifts shows a dramatic improvement in surface roughness (3.2 nm) and far fewer false diffraction peaks, which indicates a reduction in periodic errors.
6. SLIT FABRICATION
Dispersive imaging spectrometers typically employ slits as spatial masks at the output of the foreoptic. For a unit magnification imaging spectrometer such as the CCVIS, the slit width matches the width of a detector element at the focal plane array. The slit side parallelism is maintained at a small fraction of the width. For the CCVIS, the design goal is to maintain this parallelism to within about 100 nm. Variation in the width, if severe enough, will lead to variation in the spectral calibration even for the most well-corrected sensor. We have developed a lithographic slit that meets this requirement.
Similar to the grating development, microfabrication techniques are also used to build the air slit. Starting with a 200 mm Si wafer, a 2000 nm plama-enhanced chemical vapor deposition oxide hard mask was deposited. An 18 µm space for the slit was defined and etched using photolithography and a Bosch deep silicon etch process. Flowing the slit definition, a bevel was also patterned and etching using the same hard mask, photolithography, and a Bosch deep silicon etch process. Last, the backside of the wafer was removed by either mechanical grinding or silicon wet chemical etching. The fabrication of the air slits is illustrated in Fig. 9.
7. ALIGNMENT, SMILE, AND KEYSTONE
Like many high-performance optical designs, imaging spectrometers have alignment tolerances on the order of micrometers. This tight tolerance is required to achieve the spatial–spectral uniformity characteristics of the optical form. The CCVIS has some advantages over the typical Offner–Chrisp since two out of the three elements are axially symmetric and easily aligned to one another. The critical slit-grating-FPA alignment that is universal to all grating-based imaging spectrometers remains in this design. Our task was simplified compared to a flight prototype, as we demonstratied the optical form only on a breadboard and did not have the challenge of building the full housing. The large catadioptic lens pair were first aligned together with a TriOptics Lens Alignment station. We achieved an alignment of $5 \pm 2\;{{\unicode{x00B5}{\rm m}}}$ in decenter and $2 \pm 1\;{{\unicode{x00B5}{\rm m}}}$ in axial displacement on the lens pair. The grating was bonded to a powered wedged lens and clocked to ensure that the orders were in plane with the wedged optic using an alignment laser source. The slit and grating were then positioned and tilted to the optical axis using a Brunson Alignment Scope with a point illumination projector along with various long throw micrometers. Finally the camera was iteratively positioned to achieve the correct focus and clocking.
Due to funding constraints, a CCD silicon array was used as a surrogate with the longer wavelength performance evaluated through the use of higher orders. The testing for smile was accomplished using a tungsten source with a series of 10 nm wide monochromatic filters in order to measure both the parallelism and curvature of the slit image. Keystone was evaluated using a white light source and bandpass filter (600–1000 nm to reduce order overlap) with a pinhole array. The full measurement of the instrumental profiles was been accomplished due to the unavailability of an appropriate spectral source.
A tungsten halogen bulb with a condensing reflector provided a broadband source that was concentrated onto the slit of the spectrometer. The condensed beam was filtered through a series of 10 nm bandwidth filters to provide discrete wavelengths for spectral analysis. An off-the-shelf silicon focal plane CCD alignment camera was chosen for its 5.5 µm pixel size, to provide oversampling analysis of the line spread function, as well as an array size that would allow the full field and spectrum recording of the slit images. The spectrometer was aligned in real time using streaming CCD frame spectral slit images. At various stages of system alignment, images were captured for analysis, at integration times that exercised the full dynamic range without saturation. No non-uniformity correction was performed for the image analysis. It was necessary to baffle, to the extent possible, the background light from entering the system. Residual background images were captured and subtracted from the analyzed frames.
Measurement of the smile was accomplished using a series of 10 nm bandwidth filtered wavelength sets separated by at least 100 nm and over the silicon focal plane spectral response range. A threshold was applied to the background-subtracted images, and detection of the first-order diffracted spectral lines was performed. A Gaussian fit of the line spread function along the slit image region was used to determine the approximate line width and cross-sectional centroid positions at each point along a focal plane row. Cross-sectional centroid positions were filtered using a goodness of Gaussian fit metric and plotted against their spatial field position. The spatial position centroids were then fit with linear and second-order functions to determine the tilt and curvature of the smile deviation in the spectral direction, with the linear fit illustrated in Fig. 10. Maximum spectral smile deviation of the second-order fit relative to the linear fit, over the field, was found to be less than 10% of the alignment camera pixel for all wavelengths measured. This scales to about 0.5 µm deviation or about 3% of the intended system design focal plane pixel size of 18 µm.
Fig. 10. Spectral smile measurements (dots) and linear fit (dashed lines) plotted on a relative scale. The filter center wavelengths are listed above the graph. All of the data are referenced to spatial pixel 2500, and a small amount of residual clocking, approximately 0.4 mrad, was removed for the plot. The ordinate is in pixel units for the CCD’s 5.5 µm pitch. The data were subsampled for plotting clarity.
Measurement of the system keystone was more challenging and required the construction of a three-pinhole aperture that was aligned in place of the entrance slit. The pinhole diameters were chosen to be less than the size of an intended focal plane pixel but large enough to allow sufficient light through the system. The pinholes were aligned linearly along the slit direction. One pinhole was aligned to the center of the spatial field and two pinholes sampling either side at approximately 90% of the maximum field. Broadband dispersion of the pinhole images creates three spectral lines. A 600–1000 nm bandpass filter was used after the condensed broadband source to prevent the diffracted dispersion lines of successive orders from overlapping. The three filtered keystone dispersion lines were analyzed down columns of the alignment focal plane. A threshold was applied to the background-subtracted images to detect all three first-order spectral keystone line segments, and Gaussian cross-sectional fits were performed to determine the centroids along each spectral line segment. The goodness-of-fit filtered points for each line were then fit linearly. An extrapolation of the linear fit was performed out to the position where first-order diffraction of 2500 nm wavelength would occur. Keystone dispersion line deviation in the spatial direction along the outer field positions, relative to the central line, was determined to be the specification metric, to make the analysis independent of residual focal plane clocking errors. The maximum deviation of any field keystone line, relative to the central line, over the extrapolated spectrum was determined to be approximately 1.25 pixels of the intended design focal plane array. This deviation was larger than expected and is due to the uncertainty in the linear fit and extrapolation, residual grating clocking error, and clocking error due to the positioning of the three pinholes. The largest alignment challenge was due to the limitations in the off-the-shelf optical mounts that were utilized for the breadboard, preventing independent grating clocking relative to the wedged lens. A future prototype will utilize custom mounts that enable precise rotation of the grating and adjustment of the FPA along the Cartesian axes and rotations about these axes in order to meet the spatial–spectral uniformity characteristics of the optical form [23]. The breadboard CCVIS in all other respects performed as predicted from the optical design.
8. CONCLUSION
We have developed and demonstrated a compact VNIR/SWIR imaging spectrometer form, the CCVIS, that has a significant reduction in SWaP while maintaining performance over a wide swath width. The form has several advantages over both the Offner–Chrisp and the Dyson forms, with the most innovative being the small size and the employment of a flat grating, which is much easier to manufacture when compared to powered gratings. The SWaP reduction is best illustrated in comparison to CWIS, which shares some attributes with the CCVIS, such as refractive optical elements and a large number of spatial samples supporting a wide swath width. The F/2.5 CCVIS design, which is close to the lower limit possible for this optical form, is about 10.5 times smaller in volume than the F/1.8 CWIS (${{379 }}\;{{\rm{cm}}^3}$ versus ${{3988 }}\;{{\rm{cm}}^3}$). In addition to the SWaP advantage, the small size facilitates thermal stabilization for radiometric and spectral stability, and it is relatively simple to align. As an example of a particular application, a wide-swath imaging spectrometer design for remote sensing of aquatic ecosystems was developed that employs multiple CCVIS modules in conjunction with a NURBS freeform optical telescope. We chose the aquatic ecosystem problem to illustrate this imaging spectrometer concept, as opposed to a similar design that would support a Landsat-like pushbroom imaging spectrometer with a 30 m GSD as recommended in the Decadal Survey. We have also developed and demonstrated the fabrication process for rapid and accurate production of dual-faceted blaze gratings through the application of grayscale lithographic manufacturing techniques. These relatively large gratings have tailored facet profiles that optimize the optical efficiency and are produced without stitching and with precise metrology so there are no periodic errors that could generate ghosts. This breadboard demonstration points to the tremendous capability of the design form for application to Earth science investigations and for planetary missions.
Funding
National Aeronautics and Space Administration Earth Science Technology Office Instrument Incubator Program (IIP) (NNG17HH60I).
Acknowledgment
We thank Sachidananda Babu, Greg Berthiaume, Laura Kennedy, Lalitha Parameswaran, Lawrence J. Petrilli, Mordechai Rothschild, Christopher Semisch, and Robert Smith for their assistance and support. Gregory Asner, Charles Bachmann, and Kevin Turpie advised us on the application of the CCVIS to aquatic ecosystem remote sensing. Any opinions, findings, conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Aeronautics and Space Administration. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise, does not constitute or imply its endorsement by the United States Government or Lincoln Laboratory, Massachusetts Institute of Technology.
Disclosures
The authors declare no conflicts of interest.
REFERENCES
1. G. Vane, A. F. H. Goetz, and J. B. Wellman, “Airborne imaging spectrometer: A new tool for remote sensing,” IEEE Trans. Geosci. Electron. GE-22, 546–549 (1984).
2. P. Jerram and J. Beletic, “Teledyne’s high performance infrared detectors for space missions,” Proc. SPIE 11180, 111803D (2019). [CrossRef]
3. A. Offner, “New concepts in projection mask aligners,” Opt. Eng. 14, 130–132 (1975). [CrossRef]
4. J. Dyson, “Unit magnification optical system without Seidel aberrations,” J. Opt. Soc. Am. 49, 713–716 (1959). [CrossRef]
5. D. Manolakis, R. Lockwood, and T. Cooley, Hyperspectral Imaging Remote Sensing: Physics, Sensors, and Algorithms, 1st ed. (Cambridge University, 2016).
6. P. Mouroulis, R. O. Green, and T. G. Chrien, “Design of pushbroom imaging spectrometers for optimum recovery of spectroscopic and spatial information,” Appl. Opt. 39, 2210–2220 (2000). [CrossRef]
7. M. P. Chrisp, “Convex diffraction grating imaging spectrometer,” U.S. patent 5880834 (9 March 1999).
8. D. Kwo, G. Lawrence, and M. Chrisp, “Design of a grating spectrometer from a 1:1 Offner mirror system,” Proc. SPIE 0818, 275–279 (1987). [CrossRef]
9. L. Mertz, “Concentric spectrographs,” Appl. Opt. 16, 3122–3124 (1977). [CrossRef]
10. H. González-Nuñez, X. Prieto-Blanco, and R. de la Fuente, “Pupil aberrations in offner spectrometers,” J. Opt. Soc. Am. 29, 442–449 (2012). [CrossRef]
11. H. A. Bender, P. Mouroulis, R. J. Korniski, R. O. Green, and D. W. Wilson, “Wide-field imaging spectrometer for the Hyperspectral Infrared Imager (HyspIRI) mission,” Proc. SPIE 9222, 92220E1 (2014). [CrossRef]
12. D. W. Warren, D. J. Gutierrez, and E. R. Keim, “Dyson spectrometers for high-performance infrared applications,” Opt. Eng. 47, 103601 (2008). [CrossRef]
13. B. V. Gorp, P. Mouroulis, D. W. Wilson, and R. O. Green, “Design of the compact wide swath imaging spectrometer (CWIS),” Proc. SPIE 9222, 92220C1 (2014). [CrossRef]
14. “Thriving on our changing planet: a decadal strategy for Earth observation from space,” Tech. rep. (National Academies of Sciences, Engineering, and Medicine, 2018).
15. M. P. Chrisp, “Visible-infrared plane grating imaging spectrometer,” U.S. patent 9689744 (27 June 2017).
16. M. P. Chrisp, “Imaging spectrometer wide field catadioptric design,” U.S. patent 7414719 B2 (19 August 2008).
17. M. P. Chrisp, “Wide swath imaging spectrometer utilizing a multi-modular design,” U.S. patent 7808635 B2 (5 October 2010).
18. M. P. Chrisp, “Dual waveband compact catadioptric imaging spectrometer,” U.S. patent 8339600 B2 (25 December 2012).
19. M. Chrisp, Developments in Remote Sensing Instrument Optical Designs (IEEE, 2000), pp. III23–III24.
20. M. P. Chrisp, R. B. Lockwood, M. A. Smith, C. Holtsberg, G. Balonek, K. J. Thome, S. R. Babu, and P. Ghuman, “A novel imaging spectrometer form for the solar reflective spectral range for size, weight and power limited applications,” Proc. SPIE 10780, 107800L (2018). [CrossRef]
21. Photon Engineering, FRED (2017).
22. T. S. Lomheim and E. D. Hernández-Baquero, “Translation of spectral radiance levels, band choices, and signal-to-noise requirements to focal plane array specifications and design constraints,” Proc. SPIE 4486, 263–307 (2002). [CrossRef]
23. R. O. Green, C. Pieters, P. Mouroulis, M. Eastwood, J. Boardman, T. Glavich, P. Isaacson, M. Annadurai, S. Besse, D. Barr, B. Buratti, D. Cate, A. Chatterjee, R. Clark, L. Cheek, J. Combe, D. Dhingra, V. Essandoh, S. Geier, J. N. Goswami, R. Green, V. Haemmerle, J. Head, L. Hovland, S. Hyman, R. Klima, T. Koch, G. Kramer, A. S. K. Kumar, K. Lee, S. Lundeen, E. Malaret, T. McCord, S. McLaughlin, J. Mustard, J. Nettles, N. Petro, K. Plourde, C. Racho, J. Rodriquez, C. Runyon, G. Sellar, C. Smith, H. Sobel, M. Staid, J. Sunshine, L. Taylor, K. Thaisen, S. Tompkins, H. Tseng, G. Vane, P. Varanasi, M. White, and D. Wilson, “The moon mineralogy mapper (M3) imaging spectrometer for lunar science: instrument description, calibration, on–orbit measurements, science data calibration and on-orbit validation,” J. Geophys. Res. 116, E00G19 (2011). [CrossRef]
24. “Feasibility study for an aquatic ecosystem Earth observing system,” Tech. rep. (Committee on Earth Observation Satellites, 2018).
25. M. P. Chrisp, B. Primeau, and M. A. Echter, “Imaging freeform optical systems designed with NURBS surfaces,” Opt. Eng. 55, 071208 (2016). [CrossRef]
26. M. P. Chrisp, “Method of and system for optimizing nurbs surfaces for an imaging system,” U.S. patent 10437943 (8 October 2019).
27. Y. Bai, J. Bajaj, J. W. Beletic, M. C. Farris, A. Joshi, S. Lauxtermann, A. Petersen, and G. Williams, “Teledyne imaging sensors: silicon CMOS imaging technologies for x-ray, UV, visible, and near infrared,” Proc. SPIE 7021, 702102 (2008). [CrossRef]
28. “Evolved expendable launch vehicle rideshare user’s guide,” Tech. rep. (Space and Missile Systems Center, Launch Systems Enterprise Directorate (SMC/LE), 2016).
29. P. Mouroulis, D. W. Wilson, P. D. Maker, and R. E. Muller, “Convex grating types for concentric imaging spectrometers,” Appl. Opt. 37, 7200–7208 (1998). [CrossRef]
30. J. M. Cobb, L. E. Comstock, P. G. Dewa, M. M. Dunn, and S. D. Flint, “Advances in diamond turned surfaces enable unique cost effective optical system solutions,” Proc. SPIE 6269, 62691L (2006). [CrossRef]
31. M. A. Smith, S. Berry, L. Parameswaran, C. Holtsberg, N. Siegel, R. Lockwood, M. P. Chrisp, D. Freeman, and M. Rothschild, “Design, simulation, and fabrication of three-dimensional microsystem components using grayscale photolithography,” J. Micro/Nanolithogr. MEMS MOEMS 18, 043507 (2019). [CrossRef]
32. Y. Zong, S. W. Brown, B. C. Johnson, K. R. Lykke, and Y. Ohno, “Simple spectral stray light correction method for array spectoradiometers,” Appl. Opt. 45, 1111–1119 (2006). [CrossRef]
33. T. N. Woods, R. T. Wrigley, G. J. Rottman, and R. E. Haring, “Scattered-light properties of diffraction gratings,” Appl. Opt. 33, 4273–4285 (1994). [CrossRef]
34. C. A. Mack, “Stochastic modeling in lithography: use of dynamical scaling in photoresist development,” J. Micro/Nanolithogr. MEMS MOEMS 8, 033001 (2009). [CrossRef]
35. T. H. Fedynyshyn, R. F. Sinta, D. K. Astolfi, A. Cabral, J. Roberts, and R. Meagley, “Deconstructing the resist to probe innate material roughness,” Proc. SPIE 6153, 615315 (2006). [CrossRef]
36. K. K. Lee, D. R. Lim, L. C. Kimerling, J. Shin, and F. Cerrina, “Fabrication of ultralow-loss Si/SiO2 waveguides by roughness reduction,” Opt. Lett. 26, 1888–1890 (2001). [CrossRef]
37. P. Shore and P. Morantz, “Ultra-precision: enabling our future,” Philos. Trans. R. Soc. A 370, 3993–4014 (2012). [CrossRef]
38. N. Taniguchi, “Current status in, and future trends of, ultraprecision machining and ultrafine materials processing,” CIRP Ann. 32, 573–582 (1983). [CrossRef]
39. D. W. Wilson, R. E. Muller, P. M. Echternach, and J. P. Backlund, “Electron-beam lithography for micro- and nano-optical applications,” Proc. SPIE 5720, 68–77 (2005). [CrossRef]
