Rapid Classification of the Primary Property of Chondrites by Means of Mid-Infrared Spectroscopy and Subsequent Principal Component Analysis. From Christian Göhring - Düsseldorf, Germany [email protected] Abstract in Deutsch: Proben von LL-, L- und H-Klasse sowie CK-, CV-, CO- und R-Klasse Meteoriten wurden mittels Infrarot-Spektroskopie im Bereich zwischen 4000 und 380 cm-1 vermessen. Die erhaltenen Daten wurden anschließend Hauptkomponentenanalysen unterworfen um auf diesem Weg Klassifikationsmerkmale der eingesetzten Proben zu finden und eine einfache Vorklassifikation innerhalb der Gruppe der gewöhnlichen Chondriten zu ermöglichen. Um die Ergebnisse der HKA zu verbessern wurde eine Methode zur Modifikation der Berechnung benutzt. Durch Einfügen von Spektrendaten von natürlichen Mineralen als auch synthetischer Spektren konnte die Trennschärfe und damit die Sicherheit einer korrekten Einstufung beträchtlich erhöht werden. Durch HKAs unter Kombination der Spektren von gewöhnlichen Chondriten mit denen der kohligen Chondriten konnte auch hier exemplarisch eine gute Abgrenzung zwischen diesen erreicht werden. The Infrared spectroscopy in combination with the Principal Generally these minerals are organized into coarse grains [Fig. 1] or crystals Component Analysis are powerfull tools for the data analysis for inside a surrounding matrix. example in industry, agriculture or medicine. In the chemical analytics the PCA helps to correlate statistical and/or meta data with Based on their elemental compositions, the mineralogy and the petrographical the data obtained by instrumental sample measurements. appearances meteorites are grouped into several classes. The class membership Particularly samples from native materials (crops, metabolits...) are of LL, L and H chondrites is detemined by their iron contents and the ratio excellent candidates but also challenges. Now it was obvious to use between the elementary iron Fe0 and the constituents of its oxidized forms IR spectroscopy and Principal Component Analysis as a completive Fe2+/3+. method to classify new found meteorites, as they are also a „natural product“. Class Fe0 (elem.) Fe2+ (mol Fa) The classification of found meteorites is a complex and expensive process. ____________________________________________________ Geochemical and petrological examinations are necessary and they are LL 19-22% 27-32% often not sufficent enough to take the right categorization. In these cases L 20-25% 23-26% further analytics, isotopic examinations or other, have to be done. H 25-31% 16-20% In this paper it shall be shown that the IR spectroscopy (IR) in combination Tab.1: Pure Iron versus Fayalite content in the three ordinary chondrite classes with the mathematical Principal Component Analysis (PCA) are able to differentiate between them and this combination shows an alternative way The IR spectroscopy: for fast meteorite studies. The IR spectroscopy is a technique to detect chemical compositions by the absorption of light energy inside the substances to be investigated. This absorption is caused by an electrical dipol field between two or more different kind of atoms assembling the sample. It is not essential which kind of bonding between these atoms persists. As a result vibrations e.g. along the bonds are perturbated. The energy of the absorbed light is determined by the mass of the atoms and the strengthness of the bonding. The result of an IR measurement is a diagramm with the wavelength (energy of light quantums) of the infrared light on the abscissa and the intensity of absorption on the ordinate axis [Fig. 2]. A material consisting of only one kind of atoms cannot absorb light energy and forms no spectrum. The minerals contained into a meteorite form absorptions, observable as characteristical features into the IR spectrum. These minerals are essentially Olivines (Forsterite, Fayalite), Pyroxenes, Plagioclases and Iron(Fe2/3)oxides. The elementary Fe0 and either its alloys with Nickel do not form any absorptions and are not detectable by means of IR spectroscopy. The PCA algorythm: The Principal Component Analyis is a statistical technique to evaluate data Fig.1: A section of NWA 6080 (LL4) shows a well-formed chondrule and Fe/Ni inclusions into sets with multiple connections between sample properties and observations a grey matrix (1mm/div). made by e.g. chemical analyzing. Goal of this procedure is to reduce these complex observations to a low-dimensional representation and sort out less The composition of meteorites: important raw informations. Subsequentially these results can be used for classification, qualitative or quantitative determination of the sample The compositions of meteorites are very manyfold. The main constituents constituents. are the mineral groups of Olivines, Ortho- and Clinopyroxenes as well as Plagioclases and other minor constituents. All these minerals on their part Experimental studies: consist of several cations, primarily iron, nickel, calcium, magnesium, aluminium and further trace metals. Iron and Nickel often appear as pure From many internet supported sources meteorite samples were acquired. Most metals as well as their oxides and sulfides [Ref. 6], too. As the counterpart of them in form of slices and many as coarse fragments. All of the slices were silicates and alumosilicates are the main anions which can be found in sawed without hydrocarbon containing oils, but some of them with water or meteorites. In low concentrations carbonate, sulfate and nitrate are often ethanol. The surfaces of the water treated examples show sometimes a present (as well as a great repertory of organic constituents especially brownish rusty area around its inclusions. [Fig.1] found in carbonaceous chondrites). Page 1 / 7 The obtained spectra show absorptions of the main components Mg/Fe- Name: Class: Fayalite Ferrosilite containing Olivines (Forsterite and Fayalite) at 885, 606, 499 and 416 cm-1 mol-% mol-% [Fig. 2]. The exact positions of the maxima are depending from the magnesium/iron ratio built into the crystal lattice of this mineral. Furthermore Chergach H5 18 16 in the range between 1000 to 1150 cm-1 the absorptions of Ortho- and DHO 2007 H5 15 14 Clinopyroxenes are visible. Their shapes and positions are also depending Gruver H4 n/a n/a from the cation ratios. In the range between 700 cm-1 and lower the Malotas H5 n/a n/a absorptions of iron and nickel oxides are overlaid. NWA 6624 H5-6 19 17 Tamdakht H4-6 18 16 Additionally to the signals of the iron minerals the formed absorptions are also superimposed by them of these minerals which are not decisive for the DHO 1706 L6 24 20 classification, for example calcium silicate (Wollastonite) and sulfate. An JAH 055 L4-5 26 22 Mt. Tazerzait L5 extreme example shown above is the spectrum of Gruver H4. It shows further 25 21 -1 Sahara 97001 L6 24 20 absorptions between 1100 to 1200 cm of noniron and nonsilicate minerals. NWA 869 L3-6 n/a n/a Data pretreatment: Benguerir LL6 29 25 NWA 5768 LL5 30 25 After measurement the spectra were baseline corrected with a hand driven NWA 7867 LL7 n/a n/a polynominal baseline modell. Loaded into the PCA software the spectra were NWA 6042 *** LL6 n/a n/a subsequently maximum normalized and then derived by the Norris algorythm. Chelyabinsk LL5 28 23 In a pretest calculation the loadings plots [Fig. 7 appendix] were used to Bjurböle L/LL4 n/a n/a commit to a useful frequency range and number of principal components. As also seen in the IR spectra the absorptions of carbonate and matrix water from -1 NWA 5371 *** Prov. L6 n/a n/a the KBr-pellet appear into the range above 4000 to 1300 cm . This range was NWA 6080 LL4 29% 23% excluded in the next calculations. Futhermore a maximum of six Principal Estacado H6 n/a n/a Components were found to be sufficient enough to differentiate between all SaU 582 L6 26% 22% the measured meteorite samples. Tab. 2: List of the used H/L/LL meteorite samples. Black printed individuals were used for The final parameters were now: calibration, red printed for test. Values are rounded. Meteorites with marked (*** ) abbreviations are not adequate evaluated. Items are listed in MBE with inofficial Y-Scale: Absorption denominations Normalization: Tallest band to 1 AU Derivation: Norris Derivation - Smoothing 3, Step 2 Instrumentation: Frequency Range: 1300 to 380 cm-1 No. of Principal Components: 6 The measurements were carried out by using an IR-spectrophotometer Validation: Full Cross Validation IFS66v from Bruker Optik (Karlsruhe/Germany). The spectrometer is Modell: Full equipped with a DTGS detector. For the sample spectra an effective No noise addition wavenumber range between 4000 to 380 cm-1 was choosen. For the subsequent PCA analysis the software “The Unscrambler” from All further calculations were carried out with these new parameters. CAMO AS (Oslo/Norway) was used. It uses the NIPALS algorythm (Herman World 1966) and also carries out an internal validation routine Results and Discussions: [Ref. 1]. Sample preparation and measurement: From each sample 100 mg were in-depth pulverized in an agarte mortar. From this powder now approx. 1 to 5 mg were once more grinded and mixed with Potassium bromide (KBr, e.g. Fluka, No.: 34919) in a ball mill. The amount of the meteoritic dust depends on the maximum absorption in the later obtained IR spectrum. It was measured so that the tallest band reaches between 0.8 and 1.4 absorption units. After mixing the blend was pressed with around 10 tons in a 13 mm die. The obtained pellets are semi opaque because of the recrystallization of the KBr matrix. Often fine grained particles are visible. PCA1a: The PC1/2 scores plot shows widely spread point groups for the three classes LL (green), L (red), H (blue) and the intermediate L/LL (light blue). The first calculation PCA1 was performed under using the set of calibration samples [Tab.2 black]. The scores plot of the first calculation shows a widespread distribution of all samples.