1. Introduction

Hyperspectral imaging (HSI) in the IR is a proven method to study planetary mineralogy and the atmospheres. Terrestrial atmosphere blocks significant portions of the IR spectral range, leaving for the analysis the extended visible range. Airless bodies or those with faint atmospheres (Moon, Mars, asteroids, etc.) allow to sound their surfaces in the IR range offering much better diagnostic potential. Numerous orbital push-broom scanners successfully studied the composition of planets or their satellites, such as: NIMS/Galileo [1], OMEGA at Mars Express [2], Moon Mineralogy Mapper (M³) at Chandrayaan-1 [3], VIRTIS at Venus Express, Dawn, and Rosetta spacecraft [4], and others. Also an atmosphere, often imaged from afar in the IR may be sounded at different levels, to study its structure and dynamics or to map minor constituents. Special cases, such as synchronous-orbit remote observations, or surface operations favour frame instruments, allowing to acquire the scene at once, without implementation of the scanner. One example is the MicrOmega HSI microscope built for Phobos-Grunt (mission failed in 2011), Hayabusa 2 landers and ExoMars rover [5]. Another instrument is a visible-infrared imaging spectrometer that explored lunar surface from the Chang’E 3 Yutu rover [6]. The latter experiment is close to the concept proposed in the present paper, however it provides imaging only in the extended visible range (400-950 nm), and the polarimetry feature is not implemented.

Acousto-optical diffraction in crystals can be used for the HSI analysis due to birefringence and special geometry of light-sound interaction in an anisotropic media. Since 1980-s, a number of theoretical and experimental works devoted to acousto-optical HSI have been published [7–9]. Several imaging spectrometers and spectro-polarimeters have been designed on a basis of acousto-optical tunable filters (AOTF) for astronomy [10–12] and remote sensing [13,14]. Fast random-access wavelength tuning is one of advantages of AOTFs compared to other spectral imaging methods. Compactness, low mass, and reliability determine their successful implementation in space-borne spectral instruments. Almost all instruments flown to-date are pencil-beam devices, e.g. to be mounted at a robotic arm or a mast of a planetary lander [6,15,16], or to be used for orbital remote sensing [17,18]. The only example of the AOTF imager is the visible spectrometer at the Yutu rover mentioned above [6].

Spectral imaging polarimetry is an advanced variation of the HSI method providing information on scattering and light-emitting properties of remote or microscopic objects. An emerging application for the near infrared (NIR) imaging spectro-polarimeters is remote and close-up sensing of planetary surfaces, in order to discriminate between various minerals. To realize spectro-polarimetry of remote objects with acousto-optics two different principles can be used. Firstly, simultaneous processing of both polarizations can be achieved for a particular configuration when the light incidence angle provides phase matching of two eigenwaves in the crystal. In scientific instrumentation, this configuration of the AOTF was previously used in astronomical imaging applications by NASA [11,13], and for pencil-beam instruments SPICAM and SPICAV onboard Mars Express and Venus Express [17,18]. Several laboratory prototypes were also reported (e.g [19].). An alternative approach to polarimetric analysis is to use either polarization beam splitters before the AOTF or two similar AOTFs with different Bragg angles [20] followed by independent processing of orthogonal polarizations. The disadvantage of this principle is a complicated optical layout requiring two AOTFs and/or multiple additional optical elements. Such systems often suffer from deterioration of image quality. One can also apply a variable retarder before the AOTF, but the different polarizations in this design can be captured sequentially only [21].

In this paper we report design and performance of a bench prototype near-IR spectro-polarimetric imaging system with a paratellurite AOTF. A compact optical system of the spectrometer delivers the diffracted narrowband spectral images of the object with two orthogonal linear polarizations to a single focal plane array (FPA). The testing of this prototype is also described.

2. Optical design

2.1 General concept

Principles of the AOTF design and operation are well known [7]. An AOTF is based on a birefringent crystal where unpolarized light beam is split into two eigenwaves (ordinary and extraordinary one). If the beam propagates far from the optical axis of the crystal, the polarization of the eigenwaves is almost linear. Usually, in AOTF-based HSI instruments, only one eigenwave is used while the other is depleted, and further filtered out by polarizers and/or blocked angularly or spatially [8,10,14]. However, for a specific geometry of interaction (incidence light angle and the crystal cut angle), a nearly symmetric Bragg diffraction of the two eigenmodes into opposite diffraction orders can be achieved [9,11,13,19]. For example, an ordinary component of the input beam is diffracted into the –1^st order and an extraordinary one is diffracted into the + 1^st order. This approach has been developed in this work.

Optical system of the spectro-polarimeter was designed to meet the requirements for a remote sensing or a close-up imager. One prototype imaging spectro-polarimeter for astronomy by D. Glenar et al. [13] employed commercially available TeO₂ AOTFs. The optical system of that instrument was based on an elliptical mirror magnifying the diffracted image, so the effective focal ratio at the detector of the whole system was as low as F/26. Another system with 3 times larger field of view (provided the same focal length of the telescope) and the relay optical system, that improved focal ratio up to F/6, was developed by V. Molchanov et al. [12]. In this scheme the AOTF was placed before the primary field stop, so the zero-order beam hit the detector. This was not a problem when observing a single bright object on a dark background, but for general-purpose imaging and real-scene mineral analysis this is not acceptable. The both prototypes mentioned above were designed for ground based telescopes and the characteristic size of the instruments was about 40 cm from the input optical interface to the detector plane. Such dimensions are inappropriate for a space borne instrument, especially on board a deep space mission. Some AOTF based spectro-polarimeters with independent detectors for each of the two polarizations had been also reported [9,19,22], but the problem of compact optical layout had not been addressed.

We further describe an original system consisting of front-end optics (primary imaging objective lens), a house-made TeO₂ AOTF, and back-end optics (image relay system). The optical scheme is shown in Fig. 1. The diffracted fields for the + 1^st and the −1^st diffraction orders overlap in space at the AOTF output as shown in Fig. 1(b). To make the system more compact, we “Z”-wrapped the first-order diffracted beams out of the diffraction plane, so that the detector is placed just behind the plane mirror and above it. In addition to those elements shown in Fig. 1, an aperture stop is located in front of the objective focal plane. It provides telecentricity of the front-end optics, important for aberration reduction [8]. The optical system was designed to fit a two-octave band in the NIR-SWIR region extending from 0.75 to 3.4 µm. Thus, the same system could be used, e.g., with two different octave spanning AOTFs with tuning ranges 0.75–1.7 µm and 1.7–3.4 µm.

 

Fig. 1 General 3D layout of the spectro-polarimeter.

Download Full Size | PDF

2.2 Front-end optics

To meet the application requirements, it was necessary to develop a dedicated optical system. There are three main challenges:

The optical system of imaging objective was implemented as an apochromatic telecentric air-spaced triplet with focal length 143 mm and the focal ratio F/12. The layout is shown in Fig. 2. Telecentric aperture of the primary objective was used to provide the same angle of incidence over the linear aperture of the AOTF. Otherwise, variation of the incidence angle across the aperture would result in a shift of central transmitted wavelength. Another advantage of the confocal optical scheme is a sidelobe-free point spread function [23].

 

Fig. 2 The layout of the front-end optics.

Download Full Size | PDF

The lenses were made from CaF₂, Al₂O₃ and ZnSe transparent in the spectral range of interest, and having different dispersion Abbe numbers. Optical power of lenses are chosen so that the objective together with the TeO₂ thick slab form an apochromatic optical system [22] with minimised variation of paraxial focus position along the optical axis in the entire spectral region. The chromatic focus shift is less than 0.25 mm in the wavelength range from 0.75 to 3.4 µm that is typical for apochromatic optical systems. Longitudinal aberrations for different wavelengths demonstrate very small spherical chromatism. As the result, the combined optical system of the objective with the TeO₂ prism provides diffraction-limited image quality: the root-mean-square (RMS) diameters of the on-axis and off-axis (within ± 5 mm height) image spots are times smaller than the Airy disk diameter. Also, the Strehl number varies between 0.99 on the axis and 0.9 at 5 mm of the image height. Thus, Strehl number over the whole working field is higher than 0.8 that means the physical spatial resolution is limited by diffraction only [21].

2.3 AOTF

The paratellurite AOTF was designed and built in-house (MISIS) for the spectral range of 0.75–1.7 µm that can be covered by an InGaAs FPA detector. A shear acoustic wave propagating with the tilt angle of 7.5° relative to Z axis in a TeO₂ crystal has been chosen. That provides diffraction angles of ≥ 4.3° for both polarizations simultaneously [8, 9]. Thus, the front-end lens was stopped down to F/14 in order to prevent overlapping of the zero-order with diffracted beams. Since the AOTF is used for imaging, its TeO₂ prism has optical facets orthogonal to the input beam, providing low angular dispersion for both polarizations. In a confocal optical scheme, it is important to keep a uniform ultrasonic field in the crystal because the distribution of diffracted field at the output is imaged at the FPA. In the designed configuration of the TeO₂ wide-angle AOTF the frequency of ultrasound is below 70 MHz that ensures low attenuation of ultrasound across the aperture. Additionally, the clear aperture of the AOTF is 12 mm that is larger than the field stop diameter of 8 mm, reducing acoustic diffraction artefacts.

2.4 Back-end optics

Functions the back-end relay system are:

The latter problem can be efficiently solved placing a beam stop in the back focal plane of a finite-conjugate objective lens, but spatial separation of two diffracted beams cannot be provided by an axial lens system. One possible solution is to use a multiple lens system with spatial filtering as described in [13], but a compact design can hardly be provided. A Fourier lens used in our system forms an entrance pupil in its back focus. In such a way, all three diffraction orders are spatially separated in the pupil plane and the 0-order beam stop can be placed there.

The back-end optical scheme consists of a common CaF₂ Fourier lens (F = 75 mm), and two symmetric “Z”-shaped channels (see Fig. 1 and Fig. 3). The optical axis of the Fourier lens, and the axis of symmetry of the two channels is displaced by 3 mm with respect to the axis of the paratellurite crystal to compensate for the shift of an extraordinary beam. Thus, the + 1^st and the −1^st diffraction orders are symmetric relative to the relay system. Zero-order beam stop is located at the back focal plane of the Fourier lens. Each channel of the relay system consists of an off-axial spherical mirror (R = 132 mm) tilted by 1.5° in the vertical plane, and a flat mirror inclined by 4.5° in both vertical and horizontal planes (the folding plane mirror in Fig. 1 and Fig. 3). Magnification of the relay system was 0.88 providing the focal ratio of the whole optical system F/12.3. The focal plane is located behind the flat mirror. Both diffracted channels are imaged on the same FPA detector without beam crossing. The dimensions of this scheme are only 100 × 150 × 80 mm. For comparison, the back-end optical system of the spectro-polarimeter described in [22] had the footprint of approximately 250 × 300 mm that is 2 times larger. An optimal diffraction-limited spot scattering at the FPA (RMS diameter) found with ZEMAX is less than the correspondent Airy disk for wavelengths from 0.75 µm to 3.4 µm.

 

Fig. 3 Bench-top prototype of the AOTF spectro-polarimeter.

Download Full Size | PDF

3. Laboratory prototype testing

3.1 Prototype configuration

Laboratory prototype of the spectro-polarimeter was fabricated according to the optical design described above. The dimensions of the instrument are 100 × 200 × 80 mm from the input field stop to the detector plane. This envelope does not include the front-end optics, which extends at 300 mm from the telecentric diaphragm to the input field stop, and the size of the detector, which is mainly determined by the laboratory cooling and readout systems. The photograph of the prototype is shown in Fig. 3. The optical design is suitable for the spectral range of 0.74-3.4 µm, while the further described Vis-NIR configuration of the spectro-polarimeter with the tuning interval from 0.8 to 1.75 µm is determined by the AOTF and FPA detector spectral range. A more long-wavelength configuration of the instrument requiring different AOTF and the FPA detector and will be a subject of future prototyping.

The AOTF unit has the dimensions 50 × 30 × 20 mm. Acousto-optical interaction occurs in the bench plane, and then a Fourier lens and two spherical mirrors (one for each polarization) focus light beams onto the FPA. A Vis-NIR array detector with 640 × 512 elements, and pixel pitch of 20 µm from Xenics, Belgium (XSW-640) was used as the FPA. This configuration allows to capture the images in the both polarizations simultaneously onto the two halves of the array (320 × 512 pixels per each image). The first AOTF unit described in Section 2.3 operates in spectral range from 0.8 to 1.75 µm that corresponds to the feeding RF signal from 30 to 65 MHz at the power consumption variable around 0.5–2 W. Main characteristics of the prototype are summarized in Table 1.

Table 1. Characteristics of the AOTF spectro-polarimeter.

View Table

3.2 Calibrations

A stabilized infrared lamp with the color temperature of 1900 K (Thorlabs SLS202/M; see Fig. 8) was used for illumination of tested patterns and samples when calibrating and validating the AOTF performance. We also took the 1523 nm line of a He-Ne laser in order to measure the AOTF spectral transmittance function and to estimate its passband, i.e. instrument spectral resolution δλ (Fig. 4). It was retrieved as a full width at half maximum (FWHM) while fitting the transmittance profile by the sinc² function: δλ = 25.7 nm for the −1^st diffraction order (Pol1) and δλ = 19.4 nm for the + 1^st one (Pol2). Such a difference between two polarization channels is expectable since geometry of the acousto-optical interaction differs a bit between the −1^st and + 1^st orders [17, 18]. On average, the invariant wavenumber passband equals to ~100 cm⁻¹, which is 10 nm at wavelength 1 μm and 30 nm at 1.6 μm.

 

Fig. 4 AOTF spectral transmittance function for the two orthogonal polarizations (Pol1 and Pol2) of the diffracted light. Blue dots – experimental data retrieved from laser source at 1523 nm; red line – fitted curve by the sinc² function.

Download Full Size | PDF

One more calibration characteristic, the frequency-wavelength assignment, was retrieved experimentally from six reference points: 1523 nm from the laser, and 850, 1000, 1300, 1500, 1600 nm from a set of narrow band filters illuminated by the lamp. The following formula was then applied for approximation between those points: λ = 10⁴ / (A·F_AOTF² + B·F_AOTF + C). Here, F_AOTF [MHz] is the frequency of the radio signal applied to the AOTF cell; λ [µm] is the wavelength of the diffracted light; coefficients A, B and C were fitted by the least squares method for the both polarizations separately (Fig. 5). Standard deviation of λ, i.e. accuracy of the wavelength setting, was estimated to be 0.5-1 nm that corresponds to ~5% of the spectral resolution.

 

Fig. 5 Frequency-wavelength assignment retrieved on a basis of six experimental points (left plot). Right plot shows difference (residuals in [nm]) between two polarization cases.

Download Full Size | PDF

In order to study sensitivity to the linear polarization of light, a Glan-Laser Calcite Polarizer (GL10P-B from Thorlabs) was mounted in front of the prototype telescope. The prism holder SM1PV10 adjusted orientation of the polarization plane. In such a manner we verified how the AOTF diffraction attenuates the −1^st order while the input beam is linearly polarized in the plane of the + 1^st order (i.e. perpendicularly), and vice versa [Fig. 6]. The polarization sensitivity can be estimated for the both orders as 100% × (I_max-I_min)/(I_max + I_min), where the attenuated light signal I_min is polarized orthogonally to the polarization of the incident light, giving the signal I_max [Fig. 6(c)]. The resulting sensitivity is ~0.5%, an appropriate accuracy for the linear polarization analysis.

 

Fig. 6 Image of 20 µm narrow slit illuminated by the light linearly polarized perpendicularly to the diffraction plane (a) and in the diffraction plane (b). Slit cross section at the matrix line #255 is indicated on panel (c) by black curve for the case (a) (Pol1) and by the blue curve for the case (b) (Pol2). The slit PSF fitted by the Gaussian profile (in red) and its FWHM are shown as well.

Download Full Size | PDF

In parallel with the linear polarization test, the image point spread function (PFS) was measured when observing a narrow slit with the width of 20 µm that is equal to the FPA pixel size. The cross section of the slit is shown on Fig. 6(c) for the both linear polarizations. Fitting the peak by the Gaussian profile resulted in estimation of the PSF FWHM: 2.8 pixels for Pol1 and 2.5 pixels for Pol2. Imaging capabilities of the spectro-polarimeter were also validated when observing a test pattern from the distance of one meter (illumination by unpolarized light). The smallest resolved element of the pattern, i.e. the spatial resolution, was 0.6 mm: 5 lines on 3 mm square are well resolved in diagonal and vertical/horizontal directions [Fig. 7].

 

Fig. 7 Image of a test pattern recorded in both polarizations (Pol1 and Pol2) at wavelength 1.25 µm from the distance of one meter. Gray scale indicates pixel counts.

Download Full Size | PDF

3.3 Functional testing

In order to validate spectral fidelity of the AOTF prototype we measured spectra of two minerals in diffused light. Gypsum and kaolinite powders (particles size <100 µm) have been chosen to simulate martian or lunar regolith. The samples were put onto a sample pad and illuminated by the Vis-NIR lamp at the phase angle 30° [Fig. 8(a)]. The image of both samples simultaneously captured by the FPA is shown in Fig. 8(b) for the both polarizations at the wavelength λ = 1 µm. The step of the AOTF frequency scanning (i.e. grid of spectral points) during the measurements of sample spectra, I_samp, was 5 times less than the instrumental passband. To measure the reference signal I_ref the sample pad was equivalently replaced by a Lambertian screen that provided homogeneous 95-100% reflection in the NIR-SWIR spectral range (from the Labsphere Spectraflect, see on Fig. 8(a)). The reflectance of a sample was estimated as a ratio I_samp/I_ref, where I is an integral over a group of the FPA pixels corresponded to investigated zone (see Fig. 8(b)). The resulting reflectance spectra are presented in Figs. 8c and 8d for the both minerals and polarizations. We compared the measured spectra with ones from the USGS database of minerals [24]. The database spectra were convolved with the AOTF transmittance profile from Fig. 4 that was approximated by the sinc² function for different wavelengths. One can conclude that the instrument spectral resolution is good enough to distinguish characteristic spectral features of the tested minerals.

 

Fig. 8 Measurements of reflectance spectra for gypsum and kaolinite powders: (a) pad with the samples illuminated by the Vis-NIR lamp; (b) spectral image of the samples at wavelength of 1 µm; (c, d) reflectance spectra of minerals (black and blue dots with error bars for polarizations Pol1 and Pol2) compared with library spectra from the USGS database [26] (red solid line). Gray profiles in fragment (c) are the AOTF transmittance functions approximated by sinc² with the passband 10 nm at 1 µm and 30 nm at 1.6 µm.

Download Full Size | PDF

Error bars of the measured reflectance depend on the illumination conditions, the detector sensitivity, and integration time. For instance, it took us 10 accumulated images with 10 msec per each one to measure the peak of the gypsum 1.43 µm absorption band with the signal-to-noise ratio SNR~200. Totally it took around 30 sec to scan over 300 spectral points through whole the wavelength range. One can note that the SNR<20 at wavelength shorter than 0.9 µm since the AOTF transmittance and the lamp power are rather low. The described case corresponds to the light power 10 µW/cm² reflected by the gypsum sample and integrated in 20 nm passband around 1.43 µm. To compare, the Moon surface reflects 50 µW/cm² of solar light in the same spectral interval, taking into account 7% of the Moon mean albedo. Thus, being on a planetary surface, the prototype sensitivity would good enough to perform a complete wavelength scan of the regolith reflectance within few minutes.

4. Discussion and conclusions

Optical reflection and scattering by the planetary regolith depend on its grains properties, and polarization effects can be used for advanced close-up analysis of planetary surfaces. Spectral imaging in two orthogonal polarizations at different phase angles can significantly augment the diagnostic potential of the imaging near-IR spectroscopy. Detailed analysis and laboratory experiments simulating spectro-polarimetric measurements of scattering by solid rocks or by the regolith deserves a separate study. Here we just briefly describe some ideas that could be implemented with our spectro-polarimeter.

While the incident light is unpolarized, the AOTF cell allows analyzing two output light beams, which are linearly polarized: in the plane of the acousto-optical interaction, and perpendicularly. It gives us possibility to estimate the degree of linear polarization P for the light scattered from a particulate surface. This value is defined as follows: P = (I_┴-I_||)/(I_┴ + I_||). Here I_|| and I_┴ are intensities of the scattered light that is polarized in the scattering plane and perpendicular to the scattering plane, respectively. Depending on the phase angle and on a surface reflectance, parameter P can serve to investigate the polarization contrast of scattering particles, in order to constrain the grain sizes (surface roughness), and to distinguish grains from different regolith species, as it was reported in papers [25,26]. Analogous linear polarization study of scattering by Venus clouds was recently performed by the NIR AOTF spectrometer SPICAV onboard Venus Express orbiter [27].

In conclusion, we have designed and validated a first-generation prototype of an imaging spectro-polarimeter that is aimed for remote or close-up mineralogical studies of planetary surfaces. The concept is based on an acousto-optical tunable filter that is able to process near-IR images in two orthogonal polarization planes simultaneously. With the help of a new designed compact scheme of the back-end optics the two images are located side by side on the flat FPA and are captured during a single exposure. This scheme was developed to be more compact with respect to previously reported analogues [9,19,22], and is suitable for a spaceborn instrument. To obtain high image quality in a NIR spectral range covering two octaves, we combined an apochromatic telecentric F/# = 12 front-end imaging lens with a mirror-based relay system. The laboratory prototype features the field of view of 5 cm at 1 m distance to an object, and spectral resolution around 100 cm⁻¹ over the NIR bandwidth 0.8–1.75 µm. The development of the spectro-polarimeter sensitive in the longer wavelength range up to 2.6 or 3.4 µm is planned for the future.