Liam Pedersen

NASA Ames Research Center / Carnegie Mellon University Robotics Institute

 September, 1997

There is considerable interest in using reflection spectroscopy as a means for a robotic vehicle to explore the geology of an area, and to distinguish rocks from meteorites in Antarctica. This report is a preliminary exploration of the feasibility of such an approach.

Spectra were taken using a Guided Wave spectrometer fitted with a 300 L/mm grating and a lead sulphide (PbS) detector with a range of 1100nm – 2200nm. Noise was a serious problem with this detector, especially at the lower wavelengths (see figures).

To increase the signal-to-noise ratio (SNR) the following steps were taken:

• fiber optic probe held as close to sample as possible.
• multiple readings taken and averaged at each wavelength.
• complete spectra taken multiple times and averaged.

The file readme.txt with each set of readings specifies the parameters used for each measurement. The results presented below were obtained by averaging 3 separate spectra, each point of which was obtained from 10 distinct readings. Average acquisition time was 15 minutes per sample.

The probe was held in contact with the samples, at an angle of 45° to the vertical. Holding the probe at 0° to the vertical is more likely to pick up specular reflections from the sample surface.

Calibration was done with a white spectralon standard. Ambient lighting was reduced to minimize interference, in particular from non-continuum sources such as fluorescent lights. With the exception of the enstatite spectra (see below), all spectra are divided by a spectralon spectrum obtained subsequently to the initial reference measurement, because the tungsten-halogen lamp had not yet reached thermal equilibrium (half hour settling time) and the spectra all exhibited spectral features also present in the subsequently taken spectralon spectra.

Some samples were measured twice to confirm unusual features or if there is significant surface variation.

Chosen for comparison with JPL spectral data for powdered Enstatite.

Lava Bomb (Antarctica): black igneous rock. Very similar in appearance to meteorite fusion crusts.

Murchison Meteorite: carbonaceous chondrite. Contains black fusion crust and exposed interior surfaces (dark grey with light spots.

Poltusk Meteorite: Contains black fusion crust and exposed interior surface (cut away).

Unknown Meteorite: black black fusion crust and cut away interior.

The meteorite fusion crusts and the lava bomb are almost visually indistinguishable on the basis of colour.

Imilac Meteorite (Atacama desert, Chile): pallasite. Iron skeleton with olivine matrix, mostly leached away.

Horse Mountain (near Alan Hills, Antarctica): red/ochre rock sample.

Silicified Sandstone (Antarctica): ocre and light brown colour sandstone with smooth silicified exterior.

While visually distinguishable, these samples all have similar colouration.

Note: the results have NOT been normalized to unity at a particular wavelength. However, because of uncontrollable variations in the sample – probe geometry and sample texture the scaling of the bidirectional reflection spectra obtained is not constant with each sample.

Slight absorbtion is visible at 1.4 um, but obscured by noise. Solid ample exhibits the same general trend as the spectrum produced by the powedered form, but features are distinctly less pronounced.

Note bad SNR from 1.1 um to 1.2 um. We conclude that spectrometer is unreliable at these wavelengths.

Excluding scaling, the poltusk and unknown meteorite fusion crust spectra are identical to the lava bomb. The murchison meteorite fusion crust exhibits distinct absorbtion from 1.87um to 2.00um (peak at 1.93 um), but is otherwise identical to the lava bomb spectra along these wavelengths.

Note that all the samples are dark black. The high measured reflectance is likely an artifact due to differences between the spectralon reflectance standard texture and the fusion crust texture, as well as variations in the probe/sample geometry. Note also that spectral measurements of powder from the murchison meteorite do not exceed 10%.

 Note that two spectra from different locations on the lava bomb were averaged for this result.

The above spectra were taken of the exposed interior surfaces (a fracture on murchison, transverse cuts on the others) of the same 3 meteorites before. Note the absorbtion feature at 1.658 um (width = 6nm) in the murchison interior spectra that is absent in the other two. To verify this two separate spectra from different places in the murchison meteorite were averaged to get the above result.

The two maxima in the imilac spectra occur at 1.75um and 2.12um respectively, while absorption peaks at 1.94 um. The silicified sandstone and horse mountain rocks have exactly the same features, only less pronounced, and appear to have an addition absorption feature at 1.99 um. However, this is obscured by the signal noise.

It is unlikely that these common features are artifacts, since they persist when using a different calibration measurement, and are not visible in any of the other spectra. They could be due to a mixture of olivine and pyroxene.

Although the data is too sparse to draw statistically valid conclusions, we can nevertheless hazard the following:

• Solid mineral samples may exhibit similar spectral features to the powdered forms in the NIR, but they are significantly less pronounced. Even small narrowband features will show up, but obtaining a sufficiently good SNR to reveal them may be a problem. Furthermore, even he highly weathered horse mountain and silicified sandstone sample exhibit distinctive absorption features. It is therefore feasible to consider field classifying rocks, or at least characterizing them, from their reflection spectra.

• NIR spectroscopy in the 1.7 um – 2.0 um range can distinguish some meteorite (murchison carbonaceous chondrite) fusion crusts from lava bombs. Furthermore, exposed surfaces of those meteorites have additional features that, if visible, distinguish them further from the the lava bombs.

• Accurate absolute reflectances will be difficult to obtain consistently in the field, if at all. Any classification schema must take this into account, and not depend at all on absolute reflectance data, only normalized data.
• A spectral resolution of at least 6 nm is necessary (if features such as the absorption peak in the murchison interior are to be resolved).
• Single PbS detector based spectrometers are unsuitable for field operations due to low SNR and long integration time. Rapid instruments with multiple low noise array detectors are essential if large areas are to be surveyed. Fast optics to collect light from distant samples illuminated by sunlight is essential.

With the exception of the Imilac sample which was found by L Pedersen in the Atacama desert in Chile, the meteorites here were lent out by Dr Ted Bunch of NASA Ames Research Centre. Dr Chris McKay of NASA Ames provided the Horse Shoe mountain and silicified sandstone samples from Antarctica, and Professor Bill Cassidy of the University of Pittsburgh provided the Antarctic lava bomb. Finally, Dr David Blake of NASA Ames loaned the spectrometer with which these readings were taken.