Evaluating Raman Spectroscopy to Improve Process Monitoring and Materials Characterization

Size: px
Start display at page:

Download "Evaluating Raman Spectroscopy to Improve Process Monitoring and Materials Characterization"

Transcription

1 Evaluating Raman Spectroscopy to Improve Process Monitoring and Materials Characterization Sergey Mozharov and Brian J. Marquardt Applied Physics Laboratory University of Washington Seattle, WA CPAC Meeting, 09 November 2011

2 Real-time Analysis of a Simulated Batch Fermentation Process Simulated Fermentation liquor 900 ml of synthetic sugar water was fermented with yeast for 8 hours The solution of sugar was held at 30 C and ph 6 for the duration of experiment. Glucose was added when ethanol peak equilibrated. The final total glucose concentration in the reactor was 25 g/l NeSSI Fast-Loop Kaiser Optical Systems Raman ½ ballprobe with sapphire optic, 250mW power at probe tip Average of six, 5 second exposures Mettler React-IR Fiber Optic O 2 Sensor Cassini O 2 Sensor 2

3 Glucose HPLC and Raman PC HPLC mg/ml Score Raman Time (Min) -3 3

4 Cosmic Rays th Order Polynomial 8000 Data Intensity (Arb. Units) Raman polyfit spectra Variables Raman Shift Cosmic spikes (CS) contaminate spectra and add noise to Raman data (cm-1) A universal problem that needs to be addressed Existing spike-removal methods are prone to misdetection, highly user-dependent or inappropriate for fast process analysis 4

5 CS filter: a different approach Intensity (Arb. Units) Original dataset Dataset after filtering cosmic spikes Raman Shift (cm -1 ) We designed a CS filter specifically for process analysis accurate independent of spike shape, location, width or height universal, completely automated fast 5

6 Pre-detection: derivatized dataset Differentiated along the wavenumber axis Original Raman dataset Differentiated along the sample axis 6

7 CS detection and correction The signal pattern is analyzed using 4 independent tests based on: unique behavior of cosmic spikes (randomness, abruptness) low probability of two cosmic rays will strike the same pixel in adjacent spectra spike intensity versus spectral noise and intensity variations of Raman bands Any missing point algorithm is suitable for correcting the spikes We used an average signal from two adjacent spectra (only the contaminated pixels are replaced) 7

8 Performance High speed (processing time) 8 seconds for a 331 x dataset Minimum number of spectra in the dataset 40 Accuracy All spikes were correctly detected in 7 different Raman datasets No Raman peaks were misidentified as spikes (even when their intensities were widely variable) Applicability Any process Raman dataset Convenience Fully automated, no user-dependent parameters 8

9 Development of a Raman miniprobe (1/8 o.d.) Achieved characteristics: reduced spectral background no fibers inside high optical transmission through a narrow tube (2 mm i.d.) trade-off between beam waist and divergence superb reproducibility and rheology in the sensing area ballprobe technology chemical, thermal and pressure resistance suitable materials and appropriate engineering solutions rigid design compatibility with commercial Raman instruments 9

10 1/8 mini-ballprobe 3-6 long 3.2 mm diameter (1/8 ) Materials: hastelloy/sapphire Almost as sensitive as the larger ballprobes or even more sensitive (depending on probe length and sealing method) Low working distance of 160 µm 1/16 o.d.? 1/8 o.d. 1/4 o.d. 1/2 o.d. 10

11 Sampling aspects Effective sampling volume 2.3 nanoliters 300 µm 170 µm sapphire ball surface Effective sample volume 11

12 Microliter volume sampling 3.2 mm diameter of the probe allows sampling in very small vials and testtubes Reduced sampling time Minimize sample volume Automation? Ballprobe tip (3 mm diameter sapphire ball) Liquid sample (10 microliters) 12

13 Applications of mini-ballprobes Analysis of small volume samples Continuous flow process analysis (truly micro volume reactors) In vivo analysis of tissues small volume samples continuous flow reactors tissues 13

14 Acknowledgements The Marquardt group Pfizer (ballprobe) College of Forest Resources (fermentation) University of Strathclyde (CS filter) 14