St. Catharines 2004
Technical Program

SS19:  Molecules to Planets: Infrared spectroscopy in geochemistry, exploration geochemistry and remote sensing 
Organizers / Organisateurs:  Penny King, Mike Ramsay, Gregg Swayze
Room / Salle:  AS 201

Date:  05/12/2004
Time:  10:20 AM
Presenter:  Benoit B. Rivard

Estimating bulk chemistry and modal mineralogy of igneous cores using thermal infrared spectroscopy

Feng1, J., Rivard1, B., and Gallie2, A. 1University of Alberta, Department of Earth and Atmospheric Science, Edmonton, AB, T6G 2E3 2Laurentian University, Sudbury, ON

The bulk chemistry (weight percent oxides) and modal mineralogy of igneous cores was predicted using thermal infrared reflectance (TIR) spectra. The research is motivated by the need to automate underground mining operations such as logging of rock type and ore grading in cores. To calibrate our models we examined the relationship between weight percent oxides and TIR spectra for a wide range of rock types including alkaline granite and dunite. The TIR spectra were pre-processed and transformed to a wavelet domain to isolate mineral spectral features and minimize the controls of sample physical properties (e.g. varying grain size) or varying measurement geometry across the sample suite. Results show that SiO2, Al2O3, K2O+Na2O, CaO, MgO, FeO+Fe2O3 can be estimated with a model error less than 7% absolute weight. The 100*Ca/(Na+Ca) ratio in plagioclase can be modeled with an error of less than 10 (the ratio ranges from 0-100) absolute.

A validation of the model was conducted using seventeen cut samples of felsic and mafic rocks from mines of the Sudbury basin with over thirty TIR readings per sample. The modeled modal mineralogy of each sample was calculated using the CIPW procedure and the modeled weight percent oxides as inputs. These results were compared with modal abundances estimated from thin sections showing that Quartz (< 6% error) and Orthoclase (< 7%) can be well estimated. Plagioclase shows higher errors (< 12%).