APPLICATIONS OF RAMAN AND AND BIOSYSTEMS APPLICATION: WESLEY THOMPSON JULY 17 TH, 2008

Transcription

1 APPLICATIONS OF RAMAN AND MINIATURIZATION IN INDUSTRIAL AND BIOSYSTEMS APPLICATION: BRIAN MARQUARDT CPAC SUMMER INSTITUTE WESLEY THOMPSON JULY 17 TH, 2008

2 Applied Optical Sensing Lab

3 Raman Sampling Applications

4 Raman Scattering Two types of scatter of electromagnetic radiation occur Elastic (Rayleigh scatter, very intense) Inelastic (Raman scatter, weak phenomenon) vibrational information used to identify if molecules l

5 Diagram of a Raman Instrument Excitation ti Optical Fiber Laser Filtered Probe Spectrograph CCD Detector Lens Sample Computer

6 Raman and IR Spectra Raman spectroscopy gives us information about the vibrational energies of molecules Raman is complementary to IR, producing relatively stronger peaks for symmetrical stretching vs. anti-symmetrical stretching modes Raman spectra tend to be less cluttered than IR, much less affected by water and it is usually easier to implement/sample than IR

7 Advantages of Raman Spectroscopy Little or no sample preparation is required Water is a weak scatterer - no special accessories are needed for measuring aqueous solutions Water and CO 2 vapors are very weak scatterers - purging is unnecessary Inexpensive glass sample holders, non-invasive probes and immersion probes are ideal in most cases Fiber optics (up to 100's of meters in length) can be used for remote analyses Since fundamental modes are measured, Raman bands can be easily related to chemical structure (very good for fingerprinting) The standard spectral range reaches well below 400 cm -1, making the technique ideal for both organic and inorganic species Raman spectroscopy can be used to measure bands of symmetric linkages which are weak in an infrared spectrum (e.g. -S-S-, -C-S-, -C=C-)

8 Disadvantages of Raman Inherently not sensitive (need ~ 1 million incident photons to generate 1 Raman scattered photon) Fluorescence is a common background issue Typical detection limits in the parts per thousand range Fluorescence Probability versus Probability of Raman Scatter ( 1in vs 1 in ) Requires expensive lasers, detectors and filters

9 Analysis of a Batch Fermentation Process Real-time Fermentation Monitoring Yeast Fermentation Process Image from Purves et al., Life: The Science of Biology, 4th Edition

10 Raw Raman Data for Fermentation Batch Reaction (8 day run) F Fermentation t ti Raw R Raman R S Spectra t Inttensity Raman Shift (cm-1) 10 second acquisition, 20 accumulations, sample every 10 minutes Analysis was run continuously for 8 days

11 Raman Data After Fluorescence Correction Algorithm Applied Ferm entation #3 Raman an Fluorescence Corrected Spectra Fluore escence Corre ected Intensity R am an S h ift (cm -1 )

12 3D Plot of Corrected Raman Data for Fermentation ti Batch Reaction Ethanol Intensity Maltose Raman Shift (cm -1 ) Maltose Spectrum Number (~time)

13 Reactants/Products Curves Flourescence Corrected Raman Signals for 3 Maltose and 1 Ehtanol Peaks Intensity Corrected I Product Maltose 540 cm - 1 Maltose 911 cm - 1 Maltose 1120 cm - 1 Ethanol 875 cm Reactants Elapsed Hours

14 Experiment Description Created a designed experiment that spanned the concentrations of interest 5 sugars at 4 different concentrations: Arabinose, Cellobiose, Galactose, Glucose, Xylose 0, 0.5, 1 and 2 % of each sugar (D-Optimal mixture model) Collected spectra of these 43 solutions on each instrument: Agilent Dielectric Network Analyzer (500 MHz to 50 GHz) 785nm Kaiser Holoplex Raman 532nm Kaiser Rxn3 Raman ReactIR FT-IR

15 Dielectric i Spectroscopy 500 MHz to 50 GHz 1000 points per spectrum (50MHz spacing)

16 Dielectric Theory κ = ε ε 0 = ε r = ε ' r jε " r e'50 e A measure of how much energy from an external electric field is stored in a material Called the loss factor, a measure of how dissipative or lossy a material is to an external electric field e''20 e GHz Ionic Polarization Cl- Na+ Cl- Na+ Cl- Na+ Cl- Na+ Na+ Cl- Na+ Cl- Na+ Cl- Na+ Cl- Electronic or Atomic Polarization GHz + E - Orientation Polarization (Dipole Rotation) F T + E F

17 Dielectric e spectra from full model Not sure if this is calibration shifting or due to sugars affecting dielectric

18 Glucose e Model DS

19 Arabinose e Model DS Spread in Arabinose is typical of the other four sugars

20 Water e Model DS Modeling just the concentration of water is a much tighter model. Could be useful as a total sugars measurement.

21 Dielectric e spectra from full model Same as e, could be calibration shift.

22 Glucose e Model DS

23 Cellobiose e Model DS Spread typical of the other sugars.

24 Water e Model DS Again, total sugar concentration is tighter model

25 Rd Red Laser Raman 785nm 170mW at sample 10 second exposure, 5 accumulations, 5 replicates U d fi i i d b d 4 th Used fingerprint region and subtracted 4 th order polynomial to improve model fit

26 Raman 785nm Raw Spectra Used these peaks for fingerprinting. Subtracted a 4 th order polynomial from data to baseline before modeling.

27 Raman 785nm Corrected Spectra

28 Glucose Model 785nm

29 Galactose Model 785nm Typical of all five sugars Water model correlation similar to sugar models

30 Green Laser Raman 532nm 170mW at sample 10 second exposure, 5 accumulations, 5 replicates Used U d fingerprint i region Attempted both baseline and SNV correction to remove fluorescence before modeling

31 Raman 532nm Raw Spectra

32 Baseline Corrected 532 Base

33 Glucose Model 532 Base Typical of all five sugars Water model correlation similar to sugar models

34 SNV Corrected 532 SNV

35 Glucose Model 532 SNV

36 Xylose Model 532 SNV Typical of all five sugars Water model correlation similar to sugar models

37 IR Spectroscopy Used U d parts of spectrum related to the C-H and C-O stretch to improve model

38 IR Spectra -OH stretch Water -OH from Sugar C-O Stretch region CH Stretching H2O bending H2O vapor CO2

39 Glucose Model IR Typical of all five sugars Water model correlation similar to sugar models

40 Water Model IR

41 Summary For these simple solutions, all four instruments are able to collect usable data The Raman data requires some preprocessing to remove background before tight PLS models can be created Dielectric data is more scattered, possibly could tell concentration of total sugars but identification of individual sugars may prove challenging

42 PLS Correlations Dielectric Raman 532 e' e'' Baselined SNV Correlation R^2 Correlation R^2 Correlation R^2 Correlation R^2 Arabinose Cellobiose Galactose Glucose Xylose Water Raman 785 IR From these data it seems that the optical Correlation R^2 Correlation R^2 measurements outperform the dielectric. However Arabinose it must be remembered that in an actual Cellobiose hydrolysate or fermentation liquor there may be Galactose fluorescence problems that overwhelm the optical Glucose signal of the sugars Xylose Water

43 Future Work Create unknown solutions for analysis with current PLS models for validation Evaluate different processing steps to improve models. Options include multiple preprocessing steps (Polynomial subtraction and SNV correction, etc.) and focus on smaller regions of the spectrum Assess feasibility of these instruments in more complex solutions

44 All work and no play

45 Ak Acknowledgements ld CPAC Charlie Branham CPAC/APL DuPont John Steichen James Cronin Agilent Roger Stancliff Shelley Begley UW Forestry Rick Gustafson Renata Bura