| Author | Message | |||
     Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 491 Registered: 9-2001 |
Andrew - Well, I'm not a Raman expert either, but your message did remind me of the Raman offerings of Stellarnet, BW Tek, Ocean Optics, and the other similar diode array-based spectrometers that have come on the market in the last few years. Those indeed do use gratings. I was thinking of the older benchtop-type instruments, that were pretty well pushed aside when FTIR came in. I stand corrected. To see the "last word" on wavelength scale determination, though, you may want to read the tour de force on the subject that Peter Griffiths published a while back. Here are the papers: 1) JNIRS, 11(4), p.229-240 (2003) 2) JNIRS, 11(4), p.241-255 (2003) 3) Appl. Spect., 57(2), p.176-185 (2003) What he did is not something that could be applied routinely as a QC procedure, but I don't think you could argue that this doesn't represent the ultimate method of determining the precision and accuracy of your wavelength scale, regardless of the spectroscopic technology in use. \o/ /_\ | |||
| Andrew McGlone (mcglone) Senior Member Username: mcglone Post Number: 31 Registered: 2-2001 |
Dispersive raman in decline? I think it may be the other way around, that FT systems for raman are dying away. I may be wrong, been 10 years now since I was very active on the topic, but dispersive was the way it was going then, particularly for the more inexpensive and/or portable units. Partly because of the greater water tolerance, depth penetration if you like, of those lower NIR laser wavelength systems (e.g.the popular 785nm laser choice). Also the low noise condition of TE CCD arrays. Anyway, that is Raman and this is an NIR discussion group... On the back of an envelope here I calculate the formula for the bandwidths, using first order propagation of error rule, as: delta_lambda = (10^7 * delta_nu) / nu^2 for nu in cm^-1 and lambda in nm. So you are right, as per usual. It does mean a constant bandwidth on one scale will be non-linear in the other. And then won't that have consequences that should be appropriately handled if, say, an analysis done in one scale is to match that completed on the transformed scale? Have the calibration transfer experts sorted that problem out, can a model on an FT be reasonably transformed to a grating system, excepting a simple bias offset or something equally trivial? If not then there is a bit of trouble, isn't there, when talking to each other in different units? | |||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 490 Registered: 9-2001 |
Andrew - both technologies (FTIR and grating) are equally linear and have equally constant instrumental bandwidth, as long as the data display matches the innnate characteristics of the respective technology. The only time you run into trouble is when you switch the data from one basis to the other. Even then, the situation is symmetric. For instance, data collected on FTIR and changed to grating-based: wl (nm) = 10,000 / wavenumbers That's clearly symmetric. The bandwidth is a bit more complicated, but is also symmetric, and the bandwidth in one basis is proportional to the bandwidth in the other basis, and inversely proportional to the square of the [wavelength | wavenumbers], although I don't recall the exact equation offhand. As for Raman, I didn't know anybody still uses gratings to spread out the Raman spectrum, ever since Bruce Chase and Tomas Hirschfeld showed that the FT approach is viable and has all the same advantages over dispersive Raman as it does for abaorption spectroscopy. \o/ /_\ | |||
| Andrew McGlone (mcglone) Senior Member Username: mcglone Post Number: 30 Registered: 2-2001 |
The notion of a changing bandwith across a spectrum is a concern. So the user then just has to be wise enough to pre-process appropriately to handle that? I've not had to knowlingly deal with that problem with my dispersive herschel range instruments. Anyway, I think you would want to avoid that complexity, keep things as 'nice' as possible and so do your processing (e.g., chemometrics etc) in the most linear space, whichever it is. But that does not mean you have to report in that space, does it? The instrument is our servant not our boss!For example,I like the fact that the raman folk always report in reciprocal wavenumbers whether it is a grating or FT system. It's what makes sense in terms of the parallel that exists between raman scattering and MIR spectra, and the fact that molecule identification or finger-printing is often the goal. The raman system I have put together in the lab here uses a 785nm laser and measures the scattering from 797 to 953 nm. But it only ever reports reciprocal wavenumbers between about 200 and 2200 cm^-1. | |||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 489 Registered: 9-2001 |
Dave (R) - Good point. What "makes sense" sometimes depends on other factors. The flip side of what Karl pointed out, which is implied but not explicitely stated by Karl, is that besides the historical reasons, if you plot FTIR spectra vs wavenumbers, and grating spectra vs wavelengths, then the spectral bandpass is constant in each case. This gives us a physical basis for helping to decide what "makes sense". \o/ /_\ | |||
| Karl Norris (knnirs) Senior Member Username: knnirs Post Number: 67 Registered: 8-2009 |
I have enjoyed the discussion of wavenumbers vs wavelengths for NIR spectra, and I must assume some of the blame for using the wavelength scale for our NIR spectra. I moved from the visible to NIR about 1960, and at that time nearly all spectra in these two regions used wavelengths. In addition at that time I never considered the possibility that scanning spectrometers would be used as practical instruments in other than research environments. At that time the idea was that the scanning spectrometer would only be used to determine the wavelengths for filters to be used in the instrument for practical applications. A point that has not been discussed here is the change in bandpass of the scanning grating vs. the FT across the spectrum. For the grating instrument the bandpass measured in wavenumbers changes as the wavelength changes. The same is true for the FT if you measure the bandpass in wavelength This may be more important than other effects. | |||
| David Russell (russell) Senior Member Username: russell Post Number: 58 Registered: 2-2001 |
As a end user, I work in nanometers with grating instruments and wavenumbers with FT as Howard suggests since what makes most sense to the instrument maker yields the best result when all is said and done. Rather than "taking sides" in these discussions, I enjoy learning from them. | |||
| Donald J Dahm (djdahm) Senior Member Username: djdahm Post Number: 80 Registered: 2-2007 |
OVERWROGHT?? Can't we have any fun? For you non-native English speakers, Howard was suggesting that we were becoming excited over nothing. He used a word that can have a sense of being "emotionally troubled". Perhaps that was intentional, and if so, may show he has a sense of humor afterall. On the other hand, maybe I'm sicker than I thought. What do you think, Emil? OK! I'll try to write something more serious on Units. In the meantime, I still care what people who actually use this stuff think. As FT moves further into the NIR region, I'm starting to see wave numbers more often on NIR spectra in reports. | |||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 488 Registered: 9-2001 |
I think both sides in Don's (that's Don D.) discussion are becoming somewhat overwrought. There's a reason why people started using wavelength for their X-axis scale, and then at a later time started using wavenumber (a surrogate for frequency) for the X-axis scale. In both cases the reason was instrumental. Initially, the spectrometers used for mid-IR and for NIR were based on diffraction gratings, which dispersed the light (or IR, NIR or whatever) linearly with wavelength, so wavelength was the "natural" measure to plot the Y-axis (whether energy, transmittance, absorbance or whatever) values against. Later on, FTIR came into vogue and into prominence. Here the natural unit for the disperal of light was the frequency, so that became the "natural" X-scale for FTIR spectra. The application has nothing to do with it. If you want to measure cell thickness, you can obtain perfectly fine fringes from either type of instrument for measuring that cell thickness. If you happen to be using FTIR, you can count a certain number of fringes and note the two endpoint wavenumbers, then convert just those two values to wavelength and compute the cell thickness quite easily. \o/ /_\ | |||
| Donald J Dahm (djdahm) Senior Member Username: djdahm Post Number: 79 Registered: 2-2007 |
I should have said: "In wavelength space, the slope of the log(1/R) plot is approximately proportional to particle size." | |||
| Donald J Dahm (djdahm) Senior Member Username: djdahm Post Number: 78 Registered: 2-2007 |
After saying that I had no desire to fight over this issue, I recieved an irate (or at least a pretend irate) e-mail offering the following viewpoint. " You ever had to size the space in a liquid IR cell? We counted the refringence lines and calculated the distance between the two polished salt plates. Try that in reciprocal centimeters! Ha! Unless you can express particle size or distance in cm-1, I need m, cm, um, nm, etc. to measure sizes." Perhaps, the author's point could be extended to suggest that things like our spectral pre-treatments may not work as well if we use frequency instead of wavelength. For example, the argument could go: "In wavelength space, the slope is approximately proportional to log(1/R), so we should use wavelength for MSC." As I said, I don't care to argue for either side, but I am interested in the viewpoint of practioners. | |||
| David W. Hopkins (dhopkins) Senior Member Username: dhopkins Post Number: 218 Registered: 10-2002 |
Hi Jerry, I would like to confirm that this particular chart was assembled by Foss-NIRSystems. The chart is a useful summary of wavelengths that you might expect to find particular absorbing groups. It is meant to show the ranges you might find, depending upon the environment of the absorbing groups. For example, the peak wavelength of water can be found from 1900 nm in a very tightly bound condition, to 1940 nm as seen in a high moisture sample of wheat or corn. Another example is that the positions of the CH, CH2 and CH3 "bands" show the sensitivity of the C-H vibrations to their molecular environments. In many compounds, the widths of the apparent bands in that 2300-nm region are relatively sharp and highly recognizable. It is now generally accepted that it is good practice to assign the bands one is using for quantification or identification, to show that there is a degree of specificity in the application. In pharmaceutical applications in particular, this is important. NIR applications can be set up without any understanding of the underlying spectroscopy, but the more knowledge that can be brought to a problem, the more likely the application will be stable and useful in the long run. At least, in my opinion. Best regards, Dave | |||
| Donald J Dahm (djdahm) Senior Member Username: djdahm Post Number: 77 Registered: 2-2007 |
You certainly are not the only person to make such a mistake. The mid-IR people think we make things needlessly complex by continuing to hang on the nanometer scale. I don't have the desire to fight that either way. | |||
| Andrew McGlone (mcglone) Senior Member Username: mcglone Post Number: 29 Registered: 2-2001 |
Yeah, I thought about that later, felt quite silly, that I was reading wavelengths, thinking energy but forgetting it is an inverse relationship between the two. So anyway, it is all trivial then, the widths in the wavelength space would be expected to scale as 1/n, where n is the order of the harmonic. And that looks to be the general case for those overtones | |||
| Donald J Dahm (djdahm) Senior Member Username: djdahm Post Number: 76 Registered: 2-2007 |
Perhaps you would like to consider the width in terms of frequency. Using the calculator on this site: 2000 nm = 5000 cm�1 1900 nm = 5263.16 cm�1 Diff = 263.16 1100 nm = 9090.91 cm�1 1000 nm = 10000.00 cm�1 Diff = 909.09 | |||
| Andrew McGlone (mcglone) Senior Member Username: mcglone Post Number: 28 Registered: 2-2001 |
A marvellous chart - I've seen other such charts a number of times but none as colourful and easy to read as this one. And so I see something that I've not seen before. The band widths on the graph appear to decrease as you move down in wavelength(right to left, or increasing overtone number). Not always but look at the H2O,ROH or CHx. I've not seen or thought about that before, perhaps because I'm typically stuck only in the boring 3rd overtone region. But I'm curious to ask why, perhaps naively expecting the widths to increase for the higher harmonics? | |||
| Gabi Levin (gabiruth) Senior Member Username: gabiruth Post Number: 74 Registered: 5-2009 |
This chart is probably the contribution made long time ago by Foss - and possibly created earlier by Karl It is not a precise tool as the one in MID-IR - but it serves a good purpose in many cases where you wish to know where to look for in the spectrum for some functional groups. The key point is - unlike MID-IR, the only way to identify a compound is by very careful analysis of a given unknown and known spectra - until you find a match that satisfies some criterion. | |||
| Don Burns (burns) Junior Member Username: burns Post Number: 8 Registered: 1-2006 |
Jerry: The 40" wide chart is from NIRSystems. I don't know when I got it, but it's been hanging in my office for many years and is one of the slides I use in my NIR course. Don Burns | |||
| Lois Weyer (lois_weyer) Senior Member Username: lois_weyer Post Number: 45 Registered: 7-2002 |
Bob Goddu published a NIR correlation chart in his Advances in Analytical Chemistry chapter in 1959. It is reproduced in the back of the Colthup, Daly, Wiberley "Introduction to Infrared and Raman Spectroscopy" book, which is the main reference used by many for mid IR interpretation. I'm not sure about whether the current edition of this text still has it. The chart you have attached is from an instrument company I think, and it is a bit sketchy. If you want more information, email me directly: [email protected]. I can send you a copy of a chapter I coauthored on the subject. | |||
| Jerry Jin (jcg2000) Senior Member Username: jcg2000 Post Number: 46 Registered: 1-2009 |
Dear all: In middle infrared spectroscopy, we use infrared correlation chart to link the absorption band to some chemical function group. I came across a similar correlation chart for near infrared, which caught my attention because I don't see it often. Please see attachment. I wonder how useful such a NIR correlation chart is in practical application. Do you identify the identity of a NIR absorption band in your NIR spectrum? If you can assign function groups to your NIR peaks, you may confuse NIR with middle infrared, I guess. Best regards. Jerry Jin
|
NIRcorrelationChart