Rosetta VIRTIS-V channel straylight correction

Introduction

The Visible channel of the Rosetta VIRTIS spectrometer is known to be affected by significant straylight, which makes it difficult to take advantage of its whole 230-1040 nm spectral range [1]. Straylight is by definition an extra signal due to photons reaching the detector through anomalous optical paths. The main evidence of contamination is the presence of residual solar (Fraunhofer) lines in calibrated reflectance spectra and anomalously high reflectance levels in the short-wavelength edge of the spectral range. The origin of such straylight in VIRTIS is twofold: a diffracted component, related to the IR diffraction grating, and a scattered one, related to internal reflections.

Diffracted straylight

The VIRTIS circular convex diffraction grating is formed by an outer corona optimized to disperse IR wavelength photons towards the IR detector through its diffraction order m_IR = +1 and an inner circular part optimized to disperse VIS-NIR wavelength photons towards the VIS detector, through its diffraction order m_V = -1 [2]. Both IR and VIS detectors are only sensitive to photons in their respective spectral range, but visible photons can reach the VIS detector also through diffraction on the IR grating portion. E.g., being VIS grating’s grooves 5 times denser (268 grooves/mm) than IR one (54 grooves/mm), the VIS detector is also illuminated by the order m_IR = -5, which owns the same dispersion law. These photons simply add to those from order m_V = -1, both forming the nominal VIS-NIR signal, calibrated by standard radiometric pipeline. 

However, also other diffraction orders of the IR grating illuminate the VIS detector (namely orders m_IR = -2,-3,-4,-6,-7,-8), spreading VIS wavelength photons each with its own dispersion law. As a result, the photons of a given wavelength, forming the signal at a given VIS spectel b once diffracted through the VIS grating, are actually contributing to the signal in other spectels, through IR grating higher diffraction orders [3]. The amount of spectral mixing introduced this way in the final spectrum depends on the overall shape of the radiance spectrum incident on the grating. 

This unfavourable diffraction effect has been characterized, during ground calibration, by measuring in the laboratory the spectral efficiencies of IR grating higher orders. These lab data allow to model the diffracted straylight, enabling an effective technique for removing it from the VIRTIS-V space observations. The correction amount for any given spectrum is evaluated from the whole VIS spectrum itself, using an iterative method to gradually approximate the solution. The non-linearity of the problem forces evaluating the correction independently for each spectrum, hence no simple corrective factors can be found effective for the whole dataset. 

Scattered straylight

Contrary to the diffracted one, scattered straylight is an out-of-field contamination, since the scattered photons impinging on a given detector pixel can also come from locations on the entrance slit different from that pertaining to that pixel. VIRTIS-V images across body limbs make clear the presence of such straylight, readily identifiable as an off-limb signal against the dark sky. Its spatial distribution, gradually dimming in going farther from the limb, demonstrates the interpixel correlation introduced by this effect. 

The correction developed for this kind of contamination is based on two main points: a) the stability of the spectral shape of the off-limb straylight against variation of the target body and its location in the field of view, and b) the capability of VIRTIS-V of resolving some Fraunhofer lines and measuring their depth. However, the latter is only possible when the SNR in the blue side of the spectrum is high enough, and unfortunately this condition is not often met. A systematic application of this technique to the whole VIRTIS-V dataset is therefore not possible so far.

Discussion and conclusions

We describe separately two techniques for the removal of the two straylight components identified in VIRTIS-V spectra. Of course, the two phenomena take place at the same time and a complex interplay can exist between them. In good SNR condition, the estimation of the total straylight signal proceeds by alternatively iterating the two corrections, removing the straylight modeled at each iteration step from the original measured spectrum until convergence, which is usually reached in a dozen iterations.

However, the scattered straylight model shows a high sensitivity to the noise, amplifying and propagating the fluctuations in the Fraunhofer lines region to the whole spectral range. As a consequence, a full straylight correction can only be achieved in selected spectra subsets, e.g. pertaining to bright surface regions of target bodies.

Since the straylight correction here presented has been included in a full recalibration process of the entire VIRTIS dataset, to be provided to the ESA PSA archive, only the correction for diffracted straylight has been implemented. Even if this approach keeps significant straylight levels in the data, on the other hand it represents a minimal, deterministic, correction, which can make the basis for future developments.

Finally, it is worth noting the special case of stellar targets: while in the full pixel case the amount of scattered and diffracted straylight is comparable, in the case of point sources the former is expected much lower, if any. The diffracted straylight correction is in this case very effective in recovering the unperturbed spectrum in the full V range. Since these observations are used for in-flight calibration updates, the discussed correction may have significant consequences on the radiometric calibration of the whole dataset.

References: [1] Ammannito, A. et al., RSI, 2006, doi:  10.1063/1.2349308 ; [2] Filacchione, G. PhD Thesis 2006, doi:  10.6092/UNINA/FEDOA/1462; [3] Filacchione, G. et al., RSI 2006, doi:  10.1063/1.2360786.