Right - the advent of comparatively low-cost (as predicted in the attached article) violet laser-diodes, means that it is possible that el-cheapo Raman Spectroscopy is on the horizon. If simple laser-diodes are used, then the elaborate laser control & cooling requirements go out the window, the use of 1800 lines/mm gratings (see the article) means that holographic gratings can compete equally with their more expensive ruled counterparts. Actually, if we could find the right laser diode (ie. one that only radiated on a very narrow wavelength), it wouldn't be strictly necessary to have either monochromator / gratings...

The difficulty I see arising is mainly what to do with the returned signal - I have NO FUCKING IDEA how to turn that into something the PC can utilize so as to map the peaks of the Raman Scattered light received by the CCD/Optics...

The lenses, for mine, are merely convex lenses, with specific focal lengths, that at least should be solveable...

If two or more, narrow-range laser diodes were utilized, one at a time, then there would be no need for notch filters, etc. as the reflected/incidental light (ie. that at the same wavelength as that emitted by the diode) wouldn't have to be blocked physically (I'm pretty sure the PC could be instructed to ignore all light within 10nm (+/-)5nm from the emitted light from that diode, same with the other diode(s)). Would merely mean programming the PC to ignore the specific region - separate them fairly well, say 50-100nm apart, anything that shows up in both is a legitimate peak, whatever doesn't might not be, although if it doesn't correspond with wavelengths X or Y (being the 10nm range of Diode X and Diode Y) whichever is in use at the time, then it should still be plotted.

The real problem is how to amp up the power of the available violet laser diodes so that the actual Raman Scatter Signal is able to be discerned effectively.

Here is another article on the basics of Raman Spectroscopy

What would be necessary though is paired biconvex lenses, firstly to capture then focus the laser light input to the cuvette, then a second set, to capture the scatter, then focus it upon the CCD/optical fibre/other capture device.*

Another thing would be the programming itself, to program a small robotic turntable to turn (so as to bring the other diode(s) into line with the sample - lightpath), then to program the capture device/CCD/optical fibre, to not only determine the wavelength of the light collected, but also to measure the intensity thereof. These together, with the added logic (alluded to above), would allow for the Raman Spectra to be plotted.

* Although not_important (over @SM) suggested a CCD could (he was suggesting a scanner/digital camera), by itself, be utilized to capture the entirety of the Raman Scatter - although how that would be programmed to identify the various colors, and then determine the relative population of each color (ie. part of the wavelength and intensity thereof when dealing with the focused version) remains a challenge. Of course, both scanners & digital cameras can and do (that is what they are designed for after all) determine the specific wavelength of the light incident upon them (that is what the colors in a photo / scanned copy of a document are), but how to get them to measure the intensity - ie the number of pixels of every specific color would allow us to determine the intensity of the peak corresponding to that color/wavelength. I suspect it "SHOULD" theoretically be possible to get a CCD of that type and use the data collected to spit out a graph showing the wavelengths of light captured and the volume/intensity thereof, after all, it is all data isn't it? What is given as data can be manipulated as data.

PS In the second paper, they used a 405nM diode laser of 5mW (such laser diodes are available ALL OVER THE INTERNET).

PPS I have added yet another article, which shows the idea, but they have improved it immensely - they have gotten rid of the lenses & used optical fibre probes to deliver the laser light and collect the reflected light from the sample, they use a bandpath filter to limit the range of light from the laser to a narrow range and a notch filter between the optical fibre (the collection one) and the spectrograph to remove the reflected light on the same wavelength as the laser. Thus all that gets to the spectrometer is the Stokes/Anti-Stokes or Raman scattered light...

Now what we need is some thought put into how to take the input from a CCD, which gives colors a value and their intensity is determined by a voltage, into a PC so as to remove the need for the most expensive remaining part of the system, the spectrometer itself. Good thing is, if we can work it out, probably by working out a way to set up a spectrograph that can not only split the light into its component parts, but also determine the intensity thereof (probably by the length of the line). The other alternative is to try and write some sort of program to take the raw data from the CCD itself (which splits the light into wavelengths & intensities all by itself without the need for spectrometry) and use that to graph the spectrum of the sample.

There are now two more papers, one on Students making their own spectrometers (cheap as shit) and viewing the results via webcam on the PC, and another on the use of a cheap grating spectrometer, CCD and software (which converts the digital image of the spectrum into the relevant wavelengths & intensities). I cannot find any reference to the software yet, hopefully shareware is available because this will make Raman Spectroscopy available to pretty much everyone for a couple of hundred bucks

The simplest route would be to utilise either the CCD from either (1) a scanner = rectangular - long, but only a few pixels high; or (2) a digital camera = rectangular, not as long, but a hell of a lot higher, plus huge number of pixels

The easiest way to determine the wavelengths we are dealing with would be to use a grating to split the light into its spectrum and then take a "photo" of that using the CCD - that would give the spectral wavelengths/colors. How to determine intensity is going to be harder

This looks nice & easy, the code is essentially written for us

This article (I have printed it in PDF in case the link dies), describes the use of a 10mW, 532nm  (very, VERY OTC) laser diode/pointer for actual Raman Spectroscopy, using a CCD device.

What I am proposing is not exactly earth shatteringly new - simple facts are that even optical fibre (useful - with it, you don't need to worry so much about ensuring light paths, fibre will take it round in circles if needed). It is simply that I am determining whether we can use OTC components (cheap as chips at that), to build commercial quality instrumentation, to allow us to determine what exactly is in the flask and even to monitor the reactions.

I'm seriously thinking that if a cheap brand/model of USB connected minicam is available, and we can work out how to remove the UV/NIR masks/filters, then with the low-price lasers and similarly low priced optical components - coupled with free software, then this could potentially take the guesswork out of amateur chemistry.

The code is already written, it is working out what items are needed, there is at least one online site where the author built a Raman instrument, capable of pulling up the Raman scatter for Toluene using a 5mW laser pointer. The code they used is freeware. As a 532nm laser line filter is about $44 and the edge filters are about $90, this is looking like it might be something more than a pipe-dream. Couple that with the potential for UV-adsorption spectrometry and there is a real benefit.

Or Carbonyl/Alcohol, or other bands of significance... It ain't perfect although it is something, especially when coupled with Raman as well. When these groups get their shit together (and we can get access to the magnets, they have to be highly polished, super-magnets, also the nano-scale quartz tubing) life might be perfect. It ain't this week but, is it? (although it is) getting closer all the time.

Still badly need that software (I'd love the one mentioned in the attached article) - anything that is point, click and read... Nobody wants to fuck around any more than they have to

As to Raman Spectroscopy, I just had a massive brain explosion, why is it that ONLY lasers are ever used? It is certain that Raman himself didn't have one, he had a monochromatic source of light, but it was not a laser.

I was reminded of this when I just turned on the 7 LEDS in a headlamp, it is insanely bright (more than 100W headlights) and I had a search around, turns out that aqua (~530nm) and royal blue (~400 odd) high-powered LEDS are available, are a lot easier to use than a laser module, don't have the heat issues and use less current (plus being all round easier to work with).

Of course, one then looks at how to focus the LED light into optical fiber - which would give a coherent collimated light-source, every bit as useful (although being wider - I'd have to wonder if it were more effective for Raman, more incident light = more scattered light)... Especially if we put the laser-line filter (to remove any wavelengths outside our desired band) between the 'Torch' and the lens/fiber-optic coupling...

Anyone got a cheap-arse fix for getting the output from a circular bank of LEDS into a multi-strand cable? Even hooking each LED to one cable gives a SHITLOAD of light when you get to 30-odd fibers, which also avoids having to deal with linepaths, mirrors, etc.

To take the RGB image, turn it into an 8-bit Grayscale image, draw a histogram therefrom (shade v intensity) and provide the cross-section picture (grey) and from that the histogram showing the intensity of light and the relation to the peaks on the cross-section picture

In order to determine the wavelength of the light - it will require reverse engineering the spectrum of a neon globe (1 of the 3 sub-pictures), to determine the spectral order of various pixels (ie. assign them absolute wavelengths), from that we can assign wavelengths to EVERY other pixel (column) by mathematical deduction of the diffraction/refraction of the neon light from the holographic grating slide.

Then we use the greyscale image (0-256 shades which are purely a measurement of intensity, nothing else) and find the median (middle point) of the column array (all of which are integers 0-256), also find the mode (ie. the integer that recurs most frequently in the whole 1 pixel column array), then we replace every pixel that is equal to, or exceeds 1.5 times the median value or is equal to or less than 0.75 the median amount and replacing that with the mode. Then we take the mean/average of the entire column array, to get a single integer (representing intensity) for that pixel position.

Then we work from the known transmission characteristics of the OG-550 Colored Glass Filter from Schott Glass and mathematically convert the values we get from the Reference (Solvent only) and Analytical (Solvent + Analyte) Spectra, to equalise them (ie. work it out so that the Spectra is corrected as if the transmission through the OG-550 filter was 100% across the ENTIRE range). Then we convert the numbers to a percentage (NB Prior to that they were still averages (0-256), plus the correction for transmission.

At the end of that, we have a very good outcome, we effectively have an average of 300 individual spectra (each row in the image is effectively a spectra in its own right), corrected for noise and transmission.

To do that, we need to be able to access the camera from an installed, dedicated program, which interacts with the default TWAIN interface (for pictures/images), gets the images, saves them to a specific spot (within a reserved folder made in the install) and then split the image into its 3 component parts. As the grating from which each image is generated is identical to the that used for the other 2 images, then the numbers from one are equivalent to those from the others.

We then use whatever program (display one tab with the individual images, converted to greyscale), then another with a tickbox (whether or not the user wants to convert the data to correct for transmission), then display the full grayscale image directly under the spectra derived therefrom. There will be NO user input required (except for the tickbox). The program will be WEB 2.0 compliant, so it will have direct web access to the Free Online Raman DB, that will allow the user to examine their spectra against close matches, etc. (in another tab). This will allow for rapid evaluation of the results and thus real-time information would be available.

The programming isn't a cinch (it is actually an interesting problem, C and C++ work best with TWAIN & they can also interact with XULRunner, etc.).