Published: 2019-02-15 | Categories: [»] Engineering, [»] Opticsand[»] Chemistry.

Today’s gonna be a big chunk in terms of amateur science and I am really glad to finally write this post. Today we are going to talk about Raman Spectroscopy.

For those of you who can’t wait, a photography of the breadboard is given in Figure 2. Its working principle will be detailed in section 2 of this post. I will present experimental results in section 3 and discuss those results in section 4.

Figure 1 – Breadboard used for this post
1. What is Raman Spectroscopy?

Ok but first of all: what is Raman Spectroscopy and why am I so excited about it?

With NMR and traditional IR spectroscopy, Raman shares the podium of the most powerful techniques to probe organic matter and gives a real insight on what matter is actually made of. And when I say “matter” I do not mean some esoteric stuff that will just trigger attention from highly intellectual atomic physicists. No, no, it is “matter” as in any stuff that you encounter every day. Raman spectroscopy will allow you to actually see what the molecule you just synthesized looks like.

Every molecule is made of several atoms linked together by bonds made of a mutual exchange of electrons between the atoms. These bonds are however not immutable, they can deform into periodical patterns which we call molecular vibrations. In the classical description of the phenomenon, you can imagine atoms as metal sphere linked together by springs of different strengths. If you apply a pulse to this imaginary model, the masses will start oscillating and each bonds or group of bonds will oscillate at a deterministic frequency that is a function of the adjacent masses (atoms) and spring strengths. The study of these frequencies is vibrational spectroscopy and it is exactly what we are doing here.

What makes vibrational spectroscopy so powerful is that there are groups of atoms that seems to always vibrate at about the same frequencies. The exact frequency can shift a little bit due to the other nearby atoms but they roughly lay always in the same region. Example of these remarkable groups are, for instance, the vibrations in a -CH3 methyl group, the vibration in the carbonyl >C=O group and so on. If you observe the complete spectrum of vibration frequencies in your molecule, the presence of these specific frequencies will therefore indicate that the molecule possesses either a carbonyl group, an aromatic ring, a nitro group, a chlorate group etc. Vibrational spectroscopy even allows you to identify the presence of unsaturations like in double-linked carbon-to-carbon groups. Vibrational spectroscopy plays therefore a key role in organic chemistry.

In our though experiment, we applied a pulse to the system to put all groups into oscillation simultaneously. A different approach would be to put the imaginary steel balls molecule on a XYZ translation stage and to apply a sine pattern of given frequency to the stage along one specific axis (i.e. along X, Y or Z). Oscillations will occur only if the input frequency matches one of the vibration modes of the molecule. By sweeping the input frequency it is therefore possible to plot a graph of the oscillation intensities as a function of input frequency. This is exactly what we are doing in standard Infrared spectroscopy. The electrical field of the light will interact with the dipolar moment of the molecule and when a match exists between the light frequency (wavelength) and one of the vibrational modes of the molecule, the mode will be promoted to a higher energy level and exhibit oscillation. The photon responsible for this promotion will lose all of its energy in the process and gets absorbed by the molecule. By sweeping the wavelength (by using either white light and a monochromator or a spectrograph), one can plot a graph of light absorption as a function of wavelength (frequency). The intensity of the absorption will be a function of the likelihood of a molecule to interact with the photons. It can vary on several orders of magnitude depending on the geometry of the molecule and the vibration mode.

Raman spectroscopy is another spectroscopy method that access the vibrational data of the molecule. It is important however to acknowledge from the start that Raman spectroscopy and Infrared spectroscopy are two different techniques that will bring different information about the vibrational modes of a molecule. Sometimes they will give a redundant message about a specific vibration mode and sometimes they will deliver a complementary message because the vibrational mode is not accessible in the other spectroscopy method. In fact, when the molecule is linear with a centre of symmetry, it is possible to show that if a vibrational mode is active in Raman it will be inactive in Infrared and vice-versa. To be complete, a vibrational mode can be either (1) completely inactive, (2) active in Raman only, (3) active in Infrared only, (4) active in both Raman and Infrared.

If you are wondering why some vibrational modes can be either active or inactive, you have to think about the mechanical analogy with the shaking table. Some groups may have a geometry such that the specific deformation cannot be triggered by any linear shaking of the translation stage. Such a vibration mode will therefore become inaccessible in Infrared spectroscopy. This lays down to symmetry properties of the oscillating group that falls way beyond the scope of this article. In the case of Raman spectroscopy, things get a little bit more complicated because it is not the dipolar moments of the molecule that activates the vibration modes as in IR spectroscopy but the induced dipolar moment through the polarizability of the molecule. Induced dipolar moment and dipolar moment does not have the same symmetry properties when acting on the molecule which explains why the vibrational modes accessed by Raman spectroscopy may be different than those accessed by Infrared spectroscopy even if molecules may also have modes that are accessible by both techniques. In a modern laboratory, both spectroscopy methods are therefore used to characterize a molecule.

Let us stop there for the theory of vibrational spectroscopy and dig into Raman spectroscopy itself.

Raman spectroscopy methods are very different than infrared ones, like day and night. Infrared spectroscopy requires expensive optics made of exotic materials transparent to the IR wavelengths, requires even more expensive detectors, are very sensitive to contaminants in the atmosphere (e.g. CO2), requires complex sample processing and do not tolerate the presence of water. Raman, on the other hand, works in the visible range which leads to conventional optics and detectors. Raman also offers the advantage to accept glass, water and thick samples. Until now it just looks like Raman is perfect but there is a price to pay… Molecules are pretty insensitive to the Raman effect and the system must be able to collect very few photons which stress a lot the experimental setup in terms of design and performances.

But what is this Raman effect in a first place?

I earlier said that Raman works by inducing dipolar moments in the molecule. A more conventional way of explaining the Raman effect is by discussing the Rayleigh scattering effect first. When a light ray goes by a molecule, its electrical field will excite the molecule just like an antenna will catch radio waves. Photons gets absorbed and re-emitted again in random directions. This is normal scattering, more precisely Rayleigh scattering. Once in a while, typically one in 106 (very strong emitters) to 1010 (weak emitters), a part of the energy will be transferred to a vibrational mode and the photon will be re-emitted with less energy (i.e. with a higher wavelength). This is Stokes emission. The converse is also possible but far less probable: an already excited molecule can go back to its initial state and give energy to the new photon resulting in a smaller wavelength. This is the Anti-Stokes emission. Because Anti-Stokes emissions are much less likely (and therefore intense) than the Stokes ones, I will focus on this one from now on.

2. The Setup

The overall idea of the Raman spectrometer setup is simple: we send a lot of excitation photons, collect the emitted photons only and send them to a high-resolution spectrometer to resolve the different vibrational modes. Although the principle is simple, the devil lies in the details here.

To exploit the Raman effect we need to use an excitation source of extremely pure monochromatic light with a high energetic density. We need a very thin emission line (monochromatic source) because larger source will produce blurred spectra due to the many possible excitation wavelength. If the excitation source is λe, the vibrational frequency ν and the Stokes emitted photon λ, we have the relation

So, if we measure λ experimentally, the only way to get a good estimation of ν is to have a precise, unique, value of λe.

Also, we will want to send as much light as possible to our spectrometer for the resolution of the various emitted photon frequencies. Because we will typically use a very small slit to achieve high resolution spectroscopy, all the photons emitted outside a zone of typically 10-20 µm will be lost. That means we need to focalize all of the excitation light source into a very small spot that will then be imaged on the spectrometer slit.

Hopefully there exist one type of light source that shares both of these attributes: lasers. Lasers can be extremely monochromatic and have very high power densities when focused with high quality optics. This second characteristic is what makes them very dangerous to the eyes!

The setup is schematized in Figure 2. It consists of a laser light source, a dichroic beam splitter that reflects wavelengths below 550 nm and transmit wavelengths above 550 nm, a focusing lens, a sample vial containing the liquid to measure, an edge filter, an imaging lens, and a spectrometer containing itself a slit, a collimating lens, a diffraction grating, another imaging lens and a camera sensor.

Figure 2 – Setup schematics

I will now detail each group of elements and the importance they play in the experimental setup.

2.1 The laser

The laser plays an important role in the system because it brings the excitation photons. It needs to be as monochromatic as possible, as powerful as possible (staying in safe limits, i.e. <100 mW) and have a very stable and known emission frequency.

Commercially, you will find a lot of laser dedicated to Raman spectroscopy but they are all expensive. They can be found in almost any colour (blue, green, red, infrared) and have all a very stable and thin emission line. However, because they are expensive (typically starting from $1,500 US and reaching more than $10,000 US), it is often hard for the amateur to get one. We will therefore have to select much cheaper alternatives here.

The problem is that we cannot just pick the first laser we find on eBay because you have absolutely no guarantee that they are monochromatic or not. Many of them have several intense active modes separated by a fraction of a nanometer. It may sound okay at first but it will ultimately kill your resolution. For instance, if your laser has two modes separated by 1 nm, your resolution will be limited to 35 cm-1 at 1000 cm-1 without even including the size of the slit or the imaging quality of your spectrometer! With these cheap lasers you can be either lucky or not, remember it.

Then there is the choice of the colour. Blue lasers exhibit much stronger scattering than red ones and green lasers are in-between. On the other hand, blue lasers show much stronger fluorescence with some samples which can flood your Raman signal. In the past, green lasers were favoured for commercial application but are now being taken over by red and NIR ones. With these latter ones you basically have no more fluorescence effects.

Finally, not all lasers have stable frequencies. Many are built from laser diodes which change their emission frequency with temperature. Unregulated chips are therefore useless for Raman spectroscopy because they can arbitrarily shift your Raman spectra by up to 300 cm-1 which makes identification very hard. DPSS lasers, on the other hand, are usually much more stable even when used without thermal regulation.

All of this put together usually leaves the amateur with green 532 nm DPSS lasers. They are relatively cheap, can be found with integrated driving electronics and have power ranging from 1 to 100 mW or more. However, they are rarely monochromatic unless you pay the price for.

It is possible to get reliable monochromaticity from cheap DPSS lasers but you will have to build the driving electronic yourself which is more difficult than driving LEDs. Again, you can be lucky on eBay but unfortunately, I have no known vendor that will systematically sell you high quality monochromatic lasers.

I am currently building a laser driving but it is not finished yet and in the meanwhile I am using a Thorlabs CP532 DPSS laser. I had already characterized its spectrum in a [»] former post and obtained the result of Figure 3. It has several actives modes which will be the limiting factor for the resolution of this breadboard. It is also not so powerful with a power of about 4.5 mW.

Figure 3 – Laser spectrum

By the way, I need to warn you:

DISCLAIMER: The laser used in this post is rated Class 3b and can severely damage your eyes. I strongly discourage you from repeating this experiment if you did not follow an official laser safety training previously. Also, in Europe, it is illegal to purchase these laser as natural persons. Order and use the laser at your own risks.

I will also strongly urge you to purchase laser safety goggles when using the setup and have all other person or animals leave the room. Beware of windows and open door. Laser safety goggles for this laser can be bought from Thorlabs as reference LG3. Don’t fuck up with safety.

2.2 The sample vial and focalisation optics

To get some Raman signal, one must focalise the laser beam to a very small location of space to achieve a high energetic density and then to re-image the scattered light from this spot onto the slit of a spectrometer. Since the slit is typically a few tens of micrometre, any misalignment of the collection optics in regards to the focalisation optics larger than the size of the slit will lose all the signal and leave your spectrometer eager for photons. It is therefore of uttermost importance to have a perfect alignment of the focalisation of collection optics.

This is easy on paper but very hard in practice because the light must also go through a sample vial first. Tilting or moving the vial will shift the focalisation spot and so the collection optics will not be aligned properly anymore. Getting the spot from 90 degrees on the side is therefore attractive in a first place but most of the people who try this solution struggle after with non-repeatable setups. This is why this setup was based on a back-scattering design.

In a backscattering design, the setup uses the same lens to focalise the laser spot into the sample and to collect the light back for the spectrometer. It eases up the positioning of the sample because the light will always be collected from the same position as were it was focalised.

Typically, an achromatic lens will produce a tight spot which is why I am using a Thorlabs AC127-019A-ML here. Performances will however be lowered when passing through the vial but any spot that is below 20-25 µm will do the job here.

Concerning the vials, most people use a square cuvette. I did try this solution in a first place but I quickly changed my mind because the glass of the cuvette tends to adsorb fluorescent chemicals which will produce a background blur on top of your spectra. Because glass cuvettes are expensive (typ. ~100€) you will be tempted to reuse them as much as possible. This is where you start adsorbing all sort of fluorescent compounds. After some unsatisfactory tests, I decided to switch to cheaper test tubes with narrow 0.5 mm glass walls. If the tube start showing some fluorescence, I can just throw it away (or re-use it for other purposes than Raman spectroscopy). Since they are about 0.10€ per tube, it is an economically viable solution. In practice, I re-use the same tube several time and clean it with distilled water every time after recording a spectrum. I change it only for difficult measurements or when I see some fluorescence building up in the spectra.

However, going from a square cuvette to a cylindrical test tube is not without its problem. Along one direction the light will not be deviated because it arrives nearly perpendicular to the wall (no refraction occurs then). In the second axis however, refraction occurs which alters the focusing property of the optics. This is represented in Figure 4. Please note that the schematic is the actual raytracing, not a dramatization of the effect!

Figure 4 – Astigmatism in laser focalisation

This is unfortunately a problem because the focalisation spot will be spread between the sagittal and meridional focci. The whole volume will generate the inelastic scattering but each point in this volume will generate rays in all directions, leading to a blurred spot in the spectrometer slit plane.

One way to circumvent this would be to place a negative cylindrical lens of about the same power as the test tube. However, since the power of the test tube will change depending on the liquid being measured, only an approximation can be made. This is something I am planning to implement in an upgraded version of the spectrometer.

2.3 The beamsplitter

As stated previously, this setup is based on a backscattering design. This help in positioning the sample and collecting the light but also bring the disadvantage that the light is sent back to the laser source. As a consequence, we need a mechanism to separate the excitation wavelength from the Stokes scattered wavelengths.

There are several alternatives possible. Here I have chosen to use a dichroic beamsplitter.

The dichroic beamsplitter will reflect all light whose wavelength is below some threshold and let larger wavelengths pass through. Unfortunately, the dichroic beamsplitter I am using here is not perfectly adapted to Raman spectroscopy. There exist filters designed specifically for known Raman excitation sources such as the 532 nm with very sharp transition but they are horribly expensive. The one I am using here is made for fluorescence experiments and has its transition at 550 nm.

It may seem that it is therefore not recommended for Raman spectroscopy but when compared to a non-dichroic beamsplitter which will throw away 75% of the light (50% of the excitation and 50% of the Stokes signal), it already outperforms the conventional beamsplitter from 540 nm (278 cm-1) and reaches 95% efficiency from 557 nm (840 cm-1). I would therefore always recommend a dichroic beamsplitter over a conventional beamsplitter since most of the interesting stuff is above 500 cm-1.

However, the dichroic beamsplitter is not strong enough to reject all the Rayleigh scattered excitation wavelength since it still let about 2% of the 532 nm to pass through and reach the spectrometer. 2% might not seems a lot but it is actually huge when compared to the 1:106-1:1010 fraction of scattered photons which will experience the Raman effect. If we do not take care of these remaining photons, they will flood the spectrometer and make it blind to the Raman spectra.

Some people let the excitation wavelength pass and tune their spectrometer so that its wavelength fall just on the side of the sensor but it is not a good idea. On paper you can certainly do that but a real spectrometer will produce stray light from that excitation source which will ultimately find a way to your sensor and create an annoying background which will decrease your dynamic range or create annoying ghost lines in your spectra.

The origin of stray light is multiple and very difficult to control. A first source of straylight are the dust particles on your optics or in the air that will scatter light out of its nominal path. This scattered light will then reflect inside the spectrometer, even on black surfaces, and ultimately reach your sensor producing a background offset on your spectra. A second source of straylight is the light reflected on the imaging optics or the sensor itself and reflected back on your sensor as well. A third source of straylight, very typical of spectrometers, are imperfections in the ruling frequency or local defects in your diffraction grating which will diffract the wavelength with a different angle, generating one or more ghost lines in your spectra. There may be more sources of straylight in your system but these are the main ones I can think about on the moment.

In a good (professional) spectrometer, it is considered that straylight is on the order of 0.1%. Going below that value usually requires to implement double-passes spectrometers or to work in extremely clean and controlled conditions. Again, this 0.1% may not seems like a lot but it is still enough radiation to compete with the Raman photons (0.1% of 2% is still about 20 times larger than our best expectations of Raman signal!).

Again, there are multiple ways of dealing with that remaining excitation signal but the most popular one is to use either an edge or a notch filter that will remove the unwanted wavelengths. Here, I have chosen the Thorlabs FEL550 edge filter which will cut-off any frequency below 550 nm by a factor of 10-6. This is now safe enough to start Raman spectroscopy. It is possible to tune a little bit the cut-off frequency by tilting the filter but I have found this to be not so easy in practice. I will show some experimental results in Section 3 of this post.

2.4 The spectrometer

The spectrometer is the part of the setup responsible for separating the various wavelengths composing the Raman spectra. I have already dedicated extensive posts on [»] how to build and align a spectrometer which you can read for all the details. I will therefore directly jump to the selected design parameters.

Keeping the same spectrometer layout as in my previous post, I got the best results using a 1200 lines/mm grating with a blaze angle of 17° 27’ (Thorlabs GR25-1205), 50 mm 1” achromats (Thorlabs AC254-050A-ML) and a slit of 20 µm. The angle between the two arms of the spectrometer is 40° and the grating is tilted by about 42.5° relative to the input beam to create a span of 540-660 nm (330-3600 cm-1) with the camera FLIR BFLY-PGE-31S4M-C. A 12 mm aperture was put in front of the collimating lens to limit the numerical aperture to 0.12 or f/4.

When aligning the imaging optics on the centre spot (as I did in my previous post), the intrinsic resolution of the spectrometer is expected to go from 0.35 nm on the centre to 1.3 nm on the edge of the sensor. This translates into a resolution going from ~35 cm-1 at 500 cm-1 to slightly less than 10 cm-1 at 2000 cm-1 before it increases again at 2500-3000 cm-1. Taking into account that the laser I am using has a bandwidth of about 0.4 nm, the resolution is expected to decrease to an average of ~20 cm-1 between 500-3000 cm-1 as can be seen in Figure 5.

Figure 5 – Expected Resolution of the Raman spectrometer

Please note that the resolution figures given here are very approximates at the moment since I used my shorthand formula to predict RMS spot size and that I inferred the FWHM of the system PSF from the RMS spot size and slit width, neglecting any diffraction effect. Consequently, the resolutions given here are more of an order of magnitude than a precise theoretical estimation.

I will cover a more precise resolution evaluation in section 4. Also, I am currently working on a better imaging system for this spectrometer that should achieve a ~15 cm-1 resolution on the complete spectrum. Stay tuned for updates as this is going to be a big piece too!

3. Results

The complete setup was assembled as shown in Figure 1. I will create a dedicated step-by-step article on how to assemble and align the setup later but the main steps are:

1. Align the spectrometer as explained in my previous post using a Neon lamp for the calibration.

2. Align the refocusing lens (the one that image the spot on the slit) using an [»] autocollimator.

3. Tilt the edge filter until no more signal from the laser pass through.

4. Use a solution of dilute Rhodamine B to tilt the laser and bring the spot on the slit.

The last step is essential as the laser spot must be at the exact place where the image of the slit is. If you fail to bring your laser spot at the right place, you will record no signal. The idea is to use rhodamine B which is strongly fluorescent in the red when illuminated in the green as a guide for the alignment. That way, you can even use your eyes to see where the light is going into the setup (wear laser safety goggles!!). Aligning the setup with Rhodamine will really change your life and make the assembly much simpler.

DISCLAIMER: Rhodamine B is extremely irritating to the respiratory system and will taint anything it goes in contact with. Deal with it as it was deadly plutonium !

Just for the record, I bought my Rhodamine B on eBay and it was shipped in a ziplock pack. I decided to carefully transfer it to a more secure PE bottle, just as I used to do when I was working in the chemistry department at the university. Doing that, even carefully, had put a very small quantity (micrograms ?) of rhodamine into the air which tinkled my nose a little. At that exact moment a friend of mine just told me “Dude, you are bleeding from your nose!”. Except that it was not blood but a creepy pink liquid that was flowing out of my nose… Damned Rhodamine! I cleaned my nose with water for 10 minutes and it was still pink! I even spent 3 full hours at the swimming pool and still found some pink stuff in my ears… Months later I still found deep pink Rhodamine spots in my lab despite we washed it thoroughly with water. So, please, be cautious when handling that stuff :-)

That being said, I have tested the setup with several solvents that I had at home including ethyl acetate, acetone, methyl-ethyl ketone, nitromethane, iso-propanol, methanol and ethanol. I will now compare the results obtained with spectra from the [∞] SDBS online database. All the exposure time were set to 1 minute with my 4.5 mW laser.

The spectrum of three common alcoholic solvents are given in Figure 6 (methanol), Figure 7 (ethanol) and Figure 8 (iso-propanol). They are the three most simple alcohols in organic chemistry and each one has one more carbon atom than the previous one. I chose them because they are simple molecules that will help draw the resolution properties of our spectrometer.

Figure 6 - Methanol spectrum
Figure 7 – Denaturated ethanol spectrum
Figure 8 – iso-Propanol spectrum

As can be seen in Figure 6, methanol has two large peaks in the 3000 cm-1 region (in-phase and out-of-phase -CH3 stretches). These peaks are 90 cm-1 apart and are clearly resolved. The two other peaks are found at the correct location with the right shape. We also see that the spectrometer does not detect anything below 600 cm-1. The sharp peak at 1050 cm-1 has a FWHM of 34 cm-1.

Ethanol, in Figure 7, makes the spectrum a bit more complex. The structure at 3000 cm-1 becomes more compact and the three peaks distant of 46 cm-1 and 51 cm-1 are barely defined. The two peaks at 1097 cm-1 and 1053 cm-1, distant of 44 cm-1 are also at the limit of resolution. Interestingly, the peak at 900 cm-1 is doubled with a separation of about 15 cm-1. I will come back to this later.

Iso-propanol, in Figure 8, make us draw the same conclusions. Its peaks at 3000 cm-1 are distant from 34 cm-1 and 29 cm-1 and are barely resolved too, but can still be distinguished.

This is in line with the predicted resolution of about 30 cm-1 although, for some reasons, the resolution in the ethanol spectrum is a bit less satisfactory. This might be due to the fact that I used denatured alcohol instead of pure one and that some other compounds are messing with the -CH3 stretch region.

I also recorded the spectra for nitromethane (Figure 9), acetone (Figure 10), butanone (methyl-ethyl-ketone, Figure 11) and ethyl acetate (Figure 12). I will not detail them further as they share the same conclusive evidences as the previous figures. As a side note, ethyl acetate (Figure 12) was my very first Raman spectrum :-) I was just so excited when I saw all these little peaks appearing in front of me!

Figure 9 – Nitromethane spectrum
Figure 10 – Acetone spectrum
Figure 11 – Methyl-Ethyl-Ketone (MEK) spectrum
Figure 12 – Ethylacetate spectrum

When looking at the previous results, I was extremely pleased to see that some substances that are chemically extremely close could be distinguished in their Raman spectra even when the change was in the carbon chain only. An example is given in Figure 13 with the spectra of acetone and methyl-ethyl-ketone which only differs by one methyl (-CH3) group. A similar result is obtained with ethanol and methanol in Figure 14. Substances that differ in oxidation stages such as acetone and iso-propanol can also be distinguished easily, as can be seen in Figure 15. Substituting groups has even more impact on spectra as can be seen in Figure 16 with a comparison of methanol and nitromethane.

Figure 13 – Acetone vs. MEK spectra
Figure 14 – Ethanol vs. Methanol spectra
Figure 15 – Acetone vs. iso-propanol spectra
Figure 16 – Methanol vs. nitromethane spectra

This is really what makes Raman spectroscopy so powerful for any chemist and why I am convinced that we are making a big step forward with this DIY setup. And since the results were promising I also tested some solid substances dissolved in water. Urea at ~50 mg/5 ml has shown a characteristic peak at 1000 cm-1 with a SNR of 1:5 which was just enough to bring it out of the noise floor. The result is shown in Figure 17. This is not so bad at all considering that this spectrum was recorded on a dilute solution with a 4.5 mW laser and 1 minute exposure! It opens up new doors even further…

Figure 17 – Dilute urea solution
4. Discussions

The first thing we notice on the different spectra of section 3 is that the setup is blind below 600 cm-1. This comes from the filtering of both the dichroic beamsplitter and the edge filter. I did not test them separately but I made a recording of a white LED through the system and got the results of Figure 18.

Figure 18 – Setup Transmission

The behaviour between 2500 and 3500 cm-1 is not important because it is probably due to a decrease in the LED intensity at high wavelengths. What is more important is the gap between the expected and experimental values at the small wavenumbers. The curves are shifted by about 200 cm-1 which does not allow any access to the 300-600 cm-1 range which is important for the characterization of halogen compounds. I think the effect is due to the edge pass filter which was not tilted with the appropriate angle and hence cut at a too-high wavelength. I am not very satisfied with this value and I will try to improve this later. By the way, 600 cm-1 corresponds to about 550 nm which is the nominal value of the edge filter.

Also, the setup goes from -91 cm-1 to 3350 cm-1 which is not optimal. A better alignment of the grating would expand a little bit the range on the higher wavenumbers. By adjusting the grating angle, it should be possible to cover a range of 300-3570 cm-1 which is just perfect for Raman spectrometry. A slightly smaller focal lens could reach 4000 cm-1 which might be needed in some application but will be done at the expense of resolution.

Resolution will now be a big part of this discussion as it somehow differs from the theoretical expectations.

The quick assessment of Figure 5 was made with an approximation formula of the RMS spot size and the measured bandwidth of the laser as shown in Figure 3 plus the effect of the slit. Although it explains relatively well the resolution near 3000 cm-1 with a predicted ~30 cm-1 resolution, it does not explain the poorer resolution at the centre of the sensor which was supposed to go down to 15 cm-1 but which was experimentally measured to be on the order of ~35 cm-1.

To break down experimentally the resolution between the contribution of the spectrometer and the laser, I first used a Neon spectrum to estimate the resolution of the spectrometer itself. The results are shown in Figure 19.

Figure 19 – Neon Spectrum

There is no peak in the Neon spectrum below the wavelength corresponding to 1700 cm-1 in the Raman spectra so I cannot estimate the behaviour below that value. On the other hand, I have found the spectrometer resolution to be about 0.4 nm up to ~2700 cm-1 where the peaks start to suffer from asymmetric broadening. This is actually well in line with the result predicted by the RMS formula which predicted a 0.4 nm resolution in the range 1500-2700 cm-1. The formula also predicted the resolution to drop after 2700 cm-1 to reach 0.85 nm at 3300 cm-1.

Although this explains the ~30 cm-1 resolution in the high wave number range, it does not explain anything in the medium wave number range, especially the doubling of the peaks.

Trying to explain the behaviour in the 600-2500 cm-1 range forced me to run some more accurate simulations.

I used an optical simulation software to generate the actual PSF of the spectrometer at different wavelengths. I then convolved these PSFs with the slit function and converted the result to a response in nm scale using the known dispersion of the system. I have then convolved this result with the measured spectrum of the laser as shown in Figure 3 and converted the wavelength scale to a wavenumber scale. I finally used these physical responses to create simulated Raman spectra for chemicals based on their reported peak positions and intensities from the SDBS database, assuming they were infinitely thin.

The simulated Raman spectrum of nitromethane is shown in Figure 20 together with the experimental measurement. The spectrum consists of two peaks at 1402 cm-1 and 1376 cm-1 as well as two other separate peaks at 918 cm-1 and 656 cm-1.

Figure 20 – Simulated nitromethane spectra vs. experimental

The simulated spectrum is extremely close to the experimental one and even manage to predict the doubling of the peak measured experimentally at 918 cm-1 and 656 cm-1. The simulations also predicted the small shoulder that can be seen near the 918 cm-1 and 1402 cm-2 peak.

Simulations also teaches us that the peak doubling effect does not originate from the multiple modes of the laser but from the spectrometer PSF itself. The PSF with and without the laser effect are shown in Figure 21 for the 918 cm-1 peak. The doubling already appears (even tripling!) in the original PSF.

Figure 21 – Simulated PSF with and without laser effect

The poor PSF of the system is directly linked to the relatively simple design I used. The only way to achieve better resolution will be to work on the optical formula for the imaging lens.

5. Conclusions

To summarize everything, in this post I have shown how to build a homemade Raman spectrometer in a back-scattering configuration with a resolution of ~35 cm-1 on a range of 600-3200 cm-1. I have detailed each part of the setup and how they affect the overall performances. Most of all, I have highlighted several aspects that would benefit from an upgrade:

1. The laser must be replaced by one that has a thinner bandwidth and, possibly, a higher power output. I have currently identified a laser diode that is extremely monochromatic and about ten times more powerful. The laser diode however needs driving electronics which I am currently working on. I expect this to take me about 6 weeks to complete.

2. The imaging optics of the spectrometer must be replaced by a better one. I have identified an optical formula for an objective that should enhance the resolution to about an uniform 15 cm-1. The formula is almost ready and I still need to make some tolerance analysis to be sure it will behave properly once built. Also, I am a bit low moneywise at the moment and I will have to wait before I can place the order for the components. I expect this to take me another 6 weeks.

3. Check if placing a negative cylindrical lens in front of the sample vial could improve the signal strength. It should be relatively easy to implement but requires to place orders for both the optics and the mechanical part to replace the sample holder. I think it should take less than 4 weeks to complete this point.

I also have some fellows visiting shortly to do some experimentations which will set me back too by a few weeks. I therefore hope to have completed everything by the end of the summer!

Since I am very excited about this stuff and that I am convinced that it will really help a lot of fellow amateur scientists or real scientists in emergent countries, teachers… who cannot buy commercial Raman spectrometers, I have decided to open a dedicated website to share this DIY Raman Spectrometer as an open-source system. The website has already been created but I haven’t had the time to write anything at the moment. I will continue posting on The Pulsar but will also make dedicated posts on the new website too.

So, stay tuned for updates and the new website!

[⇈] Top of Page