Online NIR-Spectroscopy for the Live Monitoring of Hydrocarbon Index (HI) in Contaminated Soil Material 20.12.2016 1
The Hydrocarbon Index (HI) Definition It is a cumulative parameter no absolute method Includes all hydrocarbons between C 10 H 22 and C 40 H 82 that... boil between 175 and 525 C do are not absorbed by magnesium silicate (non-polar) 20.12.2016 2
The Task Description and Requirements contaminated waste is burned in cement factory to reduce amount of toxic hydrocarbons only hydocarbons between C 10 H 22 and C 40 H 82, with boiling points of 175-525 C only apolar substances Ø Selectivity Online/Process Analysis before the oven method should be fast and exact no sample preparation should be needed no sample destruction the instrument of choice should be robust and prefereably cheap 20.12.2016 3
The Task Some Background Information Cement factory is used, because their ovens have T ~1500 C Kohlenwasserstoff-Index (KW-Index), AWEL Gewa sserschutzlabor, 2006 20.12.2016 4
The Old Methods FID/GC 2013 > Umwelt-Vollzug > Altlasten / Abfall > Analysenmethoden im Abfall- und Altlastenbereich Stand 2013 20.12.2016 5
The Old Methods FID/GC extraction of grinded/powdered sample with hexane/acetone and water filtration over Florisil (magnesium silicate) LOQ = 20 mg/kg (20 ppm) Area between n-decan and n-tetracontan used used for HI-calculation Calibration beforehand with standards/spikes 20.12.2016 6
The Old Methods FID/GC Requirement Selectivity for C 10 -C 40 exclude polar substances online/process Analysis robust fast measurement GC 20.12.2016 7
Near Infrared Spectroscopy An Overview range: 12800 4000 cm -1 (780 2500 nm) shows Overtone & Combination Bands -> lower intensities CH, OH, NH functionalities Burns, D.A.; Ciurczak, E.W., Handbook ofnear-infrared Analysis, CRC Press,Third Edition. 2007, p. 7-19 20.12.2016 8
Near Infrared Spectroscopy An Overview Band Range Wavelength (C-H) [nm] Relative Intensity [%] Fundamental Mid-IR 3380-3510 100 1st overtone Mid-IR or NIR 1690-1755 1 2nd overtone NIR 1127-1170 0.1 3rd overtone NIR 845-878 0.01 4th overtone NIR 690-770 0.005 Intensities much lower Changes in concentration harder to detect LOD = 0.1 0.5% Schwanninger, M.; Rodrigues, M.; Fackler, M.; J. Near Infrared Spectrosc. 2011, 19, p. 287 308 Workman, J.; Weyer, L.; Practical Guide and Spectral, CRC Press, Second Edition, 2012, p. 8 20.12.2016 9
Near Infrared Spectroscopy Important Features fingerprint signature of the physical state of order observable no sample preparation needed penetration depth up to a few cm less scattering than in UV-Vis Burns, D.A.; Ciurczak, E.W., Handbook ofnear-infrared Analysis, CRC Press,Third Edition. 2007, p. 7-19 20.12.2016 10
Near Infrared Spectroscopy The Spectrum Vis NIR 3rd C-H over. 850-950 nm 2nd C-H over. 1100-1225 nm 2nd C- H comb. 1300-1420 nm 1st C-H overtone 1650-1800 nm 1st C-H combination 2200-2450 nm Monograph NIR Spectroscopy; Metrohm 20.12.2016 11
Near Infrared Spectroscopy The Spectrum Monograph NIR Spectroscopy; Metrohm 20.12.2016 12
Near Infrared Spectroscopy The Spectrum Workman, J.; Weyer, L.; Practical Guide and Spectral, CRC Press, Second Edition, 2012, p. 208 20.12.2016 13
Near Infrared Spectroscopy The Spectrum The spectrum is influenced by particle size porosity and surface refractive index packing density temperature water content 20.12.2016 14
Near Infrared Spectroscopy Measurement Techniques NIR scheme Burns, D.A.; Ciurczak, E.W., Handbook ofnear-infrared Analysis, CRC Press,Third Edition. 2007, p. 71 Wetzel, D.L.; Analytical Chemistry, 1983, 55(12), p. 1166 20.12.2016 15
Factors that Influence NIR Spectra Water NaOH in water (0-50 %) top: pure water bottom: 50 %-solution wavelength shift due to water (also H-bonds) peak shape varies with water content broadness increases with water content baseline shift due to water content absorption increases with water content Workman, J.; Weyer, L.; Practical Guide and Spectral, CRC Press, Second Edition, 2012, p. 56ff Burns, D.A.; Ciurczak, E.W., Handbook ofnear-infrared Analysis, CRC Press,Third Edition. 2007, p. 334 20.12.2016 16
Factors that Influence NIR Spectra Particle Size cellulose samples a: 24 µm b: 45.8 µm c: 93.4 µm d: 261 µm e: 406 µm baseline shifts with particle size larger particles absorb more radiation Monograph NIR Spectroscopy; Metrohm 20.12.2016 17
Interesting Example egg samples 1: 20 min 2: 40 min 3: 60 min 4: 120 min 5: 240 min eggs measured directly after certain times after laying Norris, K.H.; J. NearInfrared Spectrosc., 1996, 4, p. 31-37 20.12.2016 18
Feasability Study Experimental Part The calibration is the most important part for quantitative analysis take several hundred samples (inhomogeneous) cover large concentration region grind and dry samples spike samples with oil of known concentration measure and assign bands of importance use data to get a calibration curve cross-check HI with GC method Data evaluation and calibration works by pattern recognition instead of single peak observation. 20.12.2016 19
Statistical Techniques Principle Component Analysis Linear data transformation Goals: Information about relevance (variance) Independent from each other (orthogonal) 1. Centering & Scaling 2. Matrix transformation X = t $ p & $ + t ( p & ( + So that all t x are orthogonal and each t describes the maximum variance 3. The Eigenvalue of p x describes the fraction of the total variance 4. Selection of relevant Eigenvalues & plotting in the corresponding dimensions (t x ) Norgaard, L.; Rasmus, B.; Engelsen, S.B.; Principal ComponentAnalysis and Near Infrared Spectroscopy, paperfrom FOSS A/S Analytical. 20.12.2016 20
NIR of glucose/fructose/sucrose mixture Norgaard, L.; Rasmus, B.; Engelsen, S.B.; Principal ComponentAnalysis and Near Infrared Spectroscopy, paperfrom FOSS A/S Analytical. 20.12.2016 21
Statistical Techniques Principle Component Analysis Norgaard, L.; Rasmus, B.; Engelsen, S.B.; Principal ComponentAnalysis and Near Infrared Spectroscopy, paperfrom FOSS A/S Analytical. 20.12.2016 22
Statistical Techniques Principle Component Analysis Norgaard, L.; Rasmus, B.; Engelsen, S.B.; Principal ComponentAnalysis and Near Infrared Spectroscopy, paperfrom FOSS A/S Analytical. 20.12.2016 23
Statistical Techniques Univariate Linear Regression 20.12.2016 24
Statistical Techniques Multiple Linear Regression (MLR) Burns, D.A.; Ciurczak, E.W., Handbook ofnear-infrared Analysis, CRC Press,Third Edition. Boca Raton, 2008. 20.12.2016 25
Statistical Techniques Partial Least Squares (PLS) Burns, D.A.; Ciurczak, E.W., Handbook ofnear-infrared Analysis, CRC Press, Third Edition. Boca Raton, 2008. Lindon, J.C.; Encyclopedia ofspectroscopy and Spectrometry, Elsevier Academic Press, 2. Edition, 2010. Mark, H.; Workman, J.; Chemometrics in spectroscopy, Academic Press, London, 2007. 20.12.2016 26
Statistical Techniques Comparison MLR - PLS MLR Proper choice of absorbances at different wavelengths is required. Not too many should be considered. In particular no collinear absorbances. Struggles with multicollinearity i.e. strong correlations between absorbances at different wavelengths. PLS Visual results (loadings) might help for the interpretation. However physical meaning of the new coordinates are hard (if not impossible) to interpret, since they are linear combinations of the initial coordinates. Can deal with multicollinearity. Model requires more observations than predictor variables (absorbances). Burns, D.A.; Ciurczak, E.W., Handbook ofnear-infrared Analysis, CRC Press, Third Edition. Boca Raton, 2008. Lindon, J.C.; Encyclopedia ofspectroscopy and Spectrometry, Elsevier Academic Press, 2. Edition, 2010. Mark, H.; Workman, J.; Chemometrics in spectroscopy, Academic Press, London, 2007. 20.12.2016 27
Alternative Methods Requirement NIR IR Raman GC Selectivity for C 10 -C 40 exclude polar substances online/process Analysis ( ) robust fast measurement no sample preparation ( ) limitations/problems no structure elucidation water interference broad fluorescent baseline saturation of detector time-consuming 20.12.2016 28
Thank you for your attention 20.12.2016 29