Metalens Enabled Solar Compass
Steve Conard, Kyle Daub, Jay Komsa, Charles Lange, Joe Miragliotta, Megan Payne, David Shrekenhamer, Elizabeth Walters, Jeff Warren, Scott Biddle, Johns Hopkins Univ. / Applied Physics Lab
Location: Ballroom A
Alternate Number 2
Celestial compasses using either the sun or stars as natural reference sources are an effective method to obtain rapid and accurate azimuth information for dismounted precision targeting systems, and potentially other applications that are severely SWAP (size, weight, and power) constrained. While clouds can limit the ability to consistently obtain such a reference, the high accuracy and very rapid response of a small and low cost celestial solution make it attractive for dismounted precision targeting. This presentation will explore the potential use of metalens technology for a solar celestial compass to provide a path to reduce production cost and increase accuracy, while also further reducing size and weight compared to existing devices.
A typical solar compass is part of a targeting system which uses an azimuth and vertical angle solution from the solar compass, self-location information from GPS or another source and range to target from a laser rangefinder to determine the coordinates of a remote target. While accuracy requirements vary, this paper will discuss a composite accuracy of 10 m CE90 (90th percentile horizontal error) at 2 km range. As we will review quantitatively, the azimuth accuracy of a solar compass is strongly dependent on the solar elevation angle with no azimuth information available if the sun is directly overhead. The composite accuracy noted above would apply for a solar elevation angle of 60 degrees, although solutions with less accuracy are desired for up to about an 80 degree solar elevation angle.
A solar compass suitable for this application consists of a lens with a near hemispherical field of view, optical filtering as required, and a silicon focal plane array. The near hemispherical field of view permits operation in a “point and shoot” mode, where the targeting system is aimed at the target, and the azimuth solution can be obtained in a time synchronized fashion with the firing of the laser rangefinder.
Conventionally, the solar compass optics consist of a fisheye lens coupled with a very strongly attenuating filter. Because of the short focal length needed to obtain a near hemispherical field of view, the boresight stability must be very good to meet the accuracy requirements. A linear shift in the line of sight of the optical system, relative to the focal plane array, of just one micrometer will result in approximately 1 mrad of azimuth error for the 60 degree solar elevation angle we use as a metric. Any shift in the line of sight from the time of factory boresighting, whether from aging, temperature changes, shock, etc., will result in such azimuth errors. The total azimuth error budget from all sources to meet the desired accuracy requirement is about 4 mrad. It is possible to build 8+ element fisheye lenses which have micrometer level line of sight stability over environments and lifetime, but this requires a high tolerance, high quality design. Thus the design of the lens stability is both a strong driver of accuracy and cost.
Metasurfacess are two-dimensional (2D) analogs of conventional bulk optics. This technology promises the development of miniaturized and stable lenses and corresponding imaging systems, but has limitations associated with nano-fabrication and the inherent two-dimensional architecture. While traditional lens systems use curved glass to bend light to a focus, metalenses use a layer of subwavelength features, typically nano-scaled in dimension, on a flat glass surface to manipulate light. These can be viewed as a natural evolution of lens technology, similar to how these components transitioned to using diffractive optical elements (DOEs) as individual features sizes approached the wavelength of light. Metalenses are able to be manufactured with technology used to fabricate sub-micron sized microelectronics and much of the design and fabrication process can be done by automated methods. While the current cost per unit for modest quantities of identical metalenses is high relative to conventional optics, unit prices of hundreds of dollars or less appears to be possible for realistic military production volumes. This cost is much lower than for a traditional multi-element fisheye lens manufactured to military requirements.
Several recent papers have described metalenses for near hemispherical imaging. The general design architecture has a small central aperture on the front side of the lens substrate with a metalens phase profile pattern on the back side. Although readily available electromagnetic modeling and fabrication methods are sufficiently mature for metalens development, these 2D structures have several operational limitations. First, a single layer metalens structure has narrow band spectral performance, generally 10 to 50 nm in the visible and near infrared. A second limitation is the aperture size compared to the overall size of the lens. These factors limit the use to narrowband sources and applications that can tolerate a relatively small numerical aperture. Additionally, this near hemispherical imaging metalens suffers from extreme distortion at the edge of its fields, although most hemispherical coverage lenses have this issue.
While limitations exist, they are are not necessarily an issue for implementation within a solar compass. Since the sun is being imaged, there is a gross overabundance of available light, even near the horizon. The very narrow bandwidth and small aperture of the metalens form does not reduce light irradiance to non-saturable levels, which requires additional filtering at a level of about OD 2 (Optical Density = 2). There are a number of ways this can be implemented, in a much simpler fashion than a traditional fisheye lens requiring reduction of about OD 6. The design is also best applied to the short focal lengths required to image the entire sky on a modest sized CMOS device. The lens does have a fairly large diameter relative to its focal length, but it is still smaller than the CMOS device that receives the image and is therefore not the limiting factor for size reduction.
These metalens designs do have some very positive optical properties. Their small aperture combined with the ability to finely tune their performance over small ranges of field angles allow them to be close to diffraction limited over their full field (for monochromatic light). The geometry of their design lends well to making them nearly telecentric as well as having an extremely flat field, providing uniform performance over the full field.
As the focal length is approximately the same as for a traditional fisheye lens, there is not any reduction in optomechanical tolerance requirements between the lens assembly and the imager. However, since it is a single piece of glass, all the tolerances between individual lens elements have been eliminated. The effort required to align and calibrate each system is expected to be greatly reduced, as unit to unit repeatability is expected to be very good. The filters used have no strong influence on image location, so they have much looser tolerances than for optical elements with power.
For a solar compass, the drivers for choice of wavelength are to keep well within the band of standard Silicon sensors, use longer wavelengths to get better penetration through haze and fog, and choose wavelengths that have readily available ultra-narrow bandpass filters. These combined lead us to solutions around 850 nm. The chosen filter, from Alluxa, has a 0.75 nm bandpass and an 855.75 nm center wavelength at normal incidence. The location in the ~f/4 converging beam, just above the sensor, causes the effective center wavelength to decrease by ~5 nm for the marginal rays, still well within the performance limits of our metalens design.
Our design is optimized for an image circle diameter matched to typical small format CMOS sensors. The aperture, substrate material, substrate thickness, and approximate sensor location were modeled in Zemax to generate a phase map to optimize image size over the full field, while allowing modest changes to the substrate thickness and metalens to sensor spacing. Fused silica was chosen as the substrate as it was readily available, had an advantageous refractive index, and was previously used by our fabrication group for metalenses.
The phase maps were then transferred to our metalens fabrication team. A library of nanoposts, referred to as unit cells, were developed through electromagnetic modeling and simulation and designed to emulate the phase maps derived through Zemax. A Matlab-based algorithm was constructed to enable the selection and distribution of unit cells to the 2D Zemax-based analytical solution. Prior to the fabrication of the metalens, small metasurface test structures consisting of small sections of these surfaces were patterned, etched, and optically tested for the expected optical deflection angle and efficiency. These sample test structures showed good performance. Following this stage of development, a full metalens was constructed, with an aperture coated with aluminum on the side of the substrate opposite the meta surface. The diameter of this component was approximately 5 mm
A mechanical structure was designed and fabricated of aluminum, and a small commercial sensor board with 3.45 um pixels was procured and mounted. The commercial sensor board is expected to be the primary limit on stability over temperature (and thus azimuth performance over temperature) of this initial prototype, as it limits how stably the optomechanics can be tied into the focal plane array. Completely optimized performance over temperature will only be obtained with a more production representative prototype with custom mounting of the CMOS focal plane. The lens and bandpass filter were mounted and bonded in place using radial injection holes and RTV. A changeable ND filter was added in front of the lens to further attenuate the input light from the sun. Two high performance accelerometers for measuring inclination were mounted to the sides, allowing determination of the gravity vector necessary to project the line of sight vector to the sun down to the horizon.
The resulting prototype system is approximately 48 mm square and 24 mm high. This size is dominated by the commercial test board implementations of both the camera and accelerometers. In a production configuration, we estimate the package size could be about 20 mm square with a height of 12 to 14 mm, which would include everything except the computer processor. In many applications minimization of height is particularly critical, and a key limitation is the spacing from the top of the optics to the front surface of the CMOS camera, which is 6.2 mm in the prototype design. The required clear aperture at the outer window of the prototype is under 5 mm. Modest growth in the production package size could be used to obtain increased accuracy if that were needed for a particular application.
The initial prototype system was tested both in a laboratory and outdoors on our test range. Test results will be presented.