| Author | Message | ||
     Kenneth Gallaher (ken_g) Intermediate Member Username: ken_g Post Number: 18 Registered: 7-2006 |
Yes, I did not mean to put good work down - there is progress for sure. I was involved in a marketing study of process Raman a number of years ago - involving Raman customer fans and the numbers just were not there. | ||
| Andrew McGlone (mcglone) Junior Member Username: mcglone Post Number: 10 Registered: 2-2001 |
I endorse what Ken says except to add that there are people doing the hard work around raman and making progress towards reliable quantification with the technique. The work of the Van Duyne group at NorthWestern springs to mind. Amongst other things they are developing a quantitative blood sugar system. Traditionally a hard (impossible?) measure by NIR, the glucose signal being low and matrix effects confounding calibrations. Van Duyne's group has good success with Raman. They use surface-enhanced raman to get sensitivity and a chemical capture layer to get the analyte into the measurement zone. They still need to use some solid chemometrical analysis to get good predictive models. It shows promise as a practical biosensor. But it also demonstrates the sort of trouble you have to go to, to get good quantitative raman results in practical situations (e.g., away from an analytical lab and in a timely fashion). Here is a link to one of their recent papers on the topic: http://chemgroups.northwestern.edu/vanduyne/pdf/Glucose_Sensing_with_SERS.pdf | ||
| Kenneth Gallaher (ken_g) Intermediate Member Username: ken_g Post Number: 17 Registered: 7-2006 |
Raman is very good at looking at silica supported catalysts (think glass). It does not see the support. I think Raman is still ... and may always be a niche technique - appropriate for maybe 5% of the applications compared with 95% for IR/NIR - there are too many issues with quantitative Raman. Where you see the growth in Raman is in qualitative applications - is it an explosive or shampoo etc. etc. It is not that Raman cannot be shown to work in feasability studies - but then reality sets in. | ||
| James Malone (jmalone) Junior Member Username: jmalone Post Number: 7 Registered: 8-2006 |
I just recently had an interesting situation in which I was developing a calibration equation for an organic material in wastewater. It seems that NIR gave very good equations, (1 or 2 factors; correlation coefficients of ~0.997) but very poor predictions. (The prediction samples were a second set of samples, collected at various times and confirmed using a reference method. Calibration samples were prepared using wastewater with standard addition of the organic component.) When I ran the same samples with Raman (785 nm excitation), I also got very good equations but also very good predictions - usually within 0.1% of actual values on the prediction set. I imagine that the fact that wastewater has many components that vary considerably, makes it difficult to develop a calibration equation that is robust. The OH bands seemed to be significant in the NIR calibration equations. Raman, on the other hand, does not respond to water (very much!), so it only measured the organic component of interest. Variations in the OH bands were not significant in these equations. This might be a special case in which Raman outperforms NIR for a water based matrix. I'd be interested in knowing if anyone else has seen anything similar. | ||
| Stephen Medlin |
Hello. I've been very interested for awhile in the maturity of Raman with respect to true process measurements. Now that the technology is almost completely here, I would like to know if anyone has done a critical comparison of NIR (preferably FT) and Raman (both high resolution and low resolution) for process applications. Not only am I interested in the typical metrics (e.g., SECV, # factors), but also in the practical considerations that must be made for process measurements. For example, model ruggedness, minimum number of calibration samples, chemometric complexity (e.g., to correct for laser fluctuations, sample inhomogeneities, turbidity). References or personal experiences would be very interesting. Thanks! | ||
| Tony Davies (Td) |
Stephen, You should have a look at Heinz Siesler's article in NIR news 11(3) 9 (2000). Tony Davies | ||
| Theo van Kempen |
We recently completed a comparison between Raman, FTIR (with an ATR), and NIRS for the prediction of amino acids in animal meal samples (feed grade meat products). Both NIRS and FTIR yielded good quality calibrations, with the FTIR having a slight edge of the NIRS, especially for sulfur amino acids. The FTIR was also easier to run (with the ATR). Note that we felt that the ATR was not an ideal accesory here, but we didn't know what would give us better data (without sacrificing convenience). Raman was very disappointing. A 10 minute! spectral acquisition time was used but the signal to noise ratio was still terrible. These samples were losely packed in glass vials and actually ended up charring during acquisition. Like you, I was very enthusiastic about Raman till I ran this experiment. | ||
| Kathryn Lee |
To Theo van Kempen, Can you provide more details about the Raman spectrometer that you used for your meal samples? The success of Raman often depends on the instrument, the laser wavelength, and how the sample is presented to the instrument. | ||
| weyer |
These recent responses to the April message have prompted me to respond to the original. We have a Kaiser Raman instrument (785 nm) multiplexed to four locations in a process, and have had very good results. My original models have not needed to be changed for 2-3 years. The components I'm measuring are minor ones and we have seen many variations in spectral intensity and interferences. I found that using a second derivative followed by SNV has corrected for the bulk of the variations. A Mahalonobis distance check on the spectra differentiates between usable (noisy, some sunlight, some fluorescence) and unusable (pure noise, lots of sunlight, etc.) spectra. The calibration did cover a 6-9 month period, with fluctuations in sunlight, intensity, noise, fluorescence, process variations, etc. |