Alkamid in the lab

How do I know that my laser works properly? Obviously, I have to measure it somehow, because a human eye (or rather any eye) can’t see this radiation. In the group, we conduct a few different experiments on QCLs. The simplest one is an LIV (Light-Current-Voltage) characteristics – resulting data indicates whether the laser emits any radiation and for what current the emission is the strongest.

Firstly, I have to mount a device on a cryostat. The end of the cryostat is called a cold finger. As both names suggest, they are used to cool down whatever we place on them – we do it by filling a cryostat with liquid helium (which has a temperature of around 4.5K or -269C), but it does not fill the chamber where the sample is, because that could spoil our measurement. The cold is transferred to it (just as heat is) by a narrow, metal part, hence the name cold finger. Before I put the cryostat on an optical bench, I have to make sure there are no broken connections by simply measuring the resistance across the device using a multimeter.

The cryostat has four windows through which radiation is coming out. Before an LIV measurement, I remove one window and mount a thermopile in its place. No, I am not talking about Greek history here – a thermopile is in fact a pile of thermocouples, which are a kind of thermometers. When a laser is emitting radiation onto this detector, there is a small (because my lasers’ output power is only of the order of milliwatts) change in its temperature, which is then converted to voltage.

What useful information I can get from such measurement? Although I will only know my device’s power in some arbitrary units, I can determine the current at which it lases with the maximum power, i.e. when it is properly aligned (see the picture in the previous post – in other words it lases best for a particular slope of the diagram, which is controlled by the current). Also, I repeat the experiment in different temperatures to see when it is too hot for the laser to work. Finally, by looking at the features of the plot, we can learn about general behaviour of the laser: what is the range of currents for which it is working, whether there are any additional wavelengths emitted, what is its resistivity and so on. Below is an LIV plot of my first QCL. It is a two colour design, so you can see a second, much smaller hill on the right – this is where the second colour (frequency) is switched on. The upper lines pertain to current-voltage characteristics.

I am afraid I haven’t explained what a QCL really is. I will try to do so in simple words. Fellow physicists: please don’t blame me for simplifications.

One of the most amazing technologies invented in the last few decades is definitely Molecular Beam Epitaxy – a technique that allows us to deposit unbelievably thin layers of elements like gallium, phosphorus or antimony. To create a wafer (a sandwich made of these elements) appropriate for a QCL, we need more precision than in other devices. A typical height of one layer of our sandwiches is of order of a few nanometres, and one sandwich consists of hundreds of layers!

Why couldn’t we just use one material and don’t worry about thickness? It’s here when quantum mechanics needs to be mentioned. In the picture below you can see many rectangular shapes. These are intuitively called wells and barriers. If we consider such a well, we get a surprising result: electrons cannot increase their energy by any portion, but are confined to discrete (not continuous) set of states. We say that their energy is quantized. Hence in each well there is a ladder of states and an electron can only be on one of its steps. It’s visible when you look at the tails of orange lines – they indicate where each electron level is. I should mention that wells’ depth and barriers’ height represents energy while the real space is drawn horizontally so when electron travels through the structure, it goes from left to right.  Now, why the diagram is diagonal? Because we apply voltage to our structure and it pushes our carriers – electrons – in the right direction.

How does an electron go through a barrier? This is the tricky part which I don’t feel competent enough to explain here. Going through the barriers is the crucial element of QCL design and researchers around the world are constantly searching for the best solution. You could think that they jump over the barriers, but this is not the case. In quantum mechanics, it’s possible that they actually go through or tunnel. In the picture you can see wavy shapes – these are wave functions of electrons. If there are two “hills” emerging from two different lines in one well, it means that an electron can mysteriously go from one well to another.

Laser emits radiation. In this case, it is radiation of a fairly low energy (200-400 times lower than that of visible light). When is it emitted? As an electron goes from left to right, with each transition they jump down, which means they lose energy. It can be passed to the electron’s environment, but sometimes it is emitted as a photon (as radiation). The whole designing process is about controlling the widths of wells and barriers. As you can imagine it is a very complex task, because one variable (width) corresponds to many effects (emitted energy, tunnelling mechanism, thermal properties etc.).

I hope it makes sense even if you are not a physicist. Next time I will describe how to make such a structure a working device that you can plug in and turn on, provided you have some spare liquid helium around.

It took me more than a month to fully process my first quantum cascade laser. Do not think that it normally takes so long; I just had to do things slowly to learn them properly, but also I had to wait more than two weeks for one step to be done by someone else. It seems I am a bit too technical with my posts so I will try to reduce the amount of boring, meaningless words from now on.

I am sitting next to my experimental setup where a few minutes ago I saw my device lasing for the first time. It is fairly simple: a cryostat (to cool the sample down to 4K),  a thermopile (collects the radiation from the laser and converts it to voltage) and a few standard lab tools: an oscilloscope, a function generator and a lock-in. Oh, and obviously I needed pumps to evacuate the cryostat!

For the first time I saw LabVIEW being useful – I learned to use it at university but I cannot say we invented a wheel at that time and it was just a fun way of programming. Nevertheless, it is a powerful tool that allows me to do other things while it is collecting the data. I will try to make a few cosmetic improvements to the script later.

The final steps of processing are perhaps the most stressful, because you know that any false move can lead to destroying your samples. Wire bonding – connecting the laser with its package with a very thin gold thread - was the last of these. Fortunately, with a little help from Yash, my 250μm wide precious was given 11 beautiful bonds and it showed a few ohms of resistance across itself as a sign of happiness!

Now I need to measure it properly and then understand the figures, which will take me a few days. However, there are more devices to be processed so I won’t be bored at all in the nearest future. I present my little friend to you:

The dark rectangle with golden stripes is my laser.

I have just realized that it has been two weeks since I last wrote… that is because I started my first QCL device processing a couple of days ago and spent most of my days in the cleanroom. Now I have some experience, a few pictures and a lot to write about!

I am becoming more confident with mask aligning and developing now, but there were days when I spent 3 hours trying to move a step forward with no results. Yash leaves me alone now so that I can be independent which is great, but it also allows me to make mistakes that I would not have made if I had asked him. Let me tell you the story of the alignment…

The first one is a piece of cake. It is called an etch bead removal: we leave a big undeveloped square in the centre of our sample, removing the resist on the corners (because it forms beads that may interrupt the next steps). The next one is quite important, because the etched ridge (the core of my device) is defined by it. However, there are still no reference points to align to so it does not have to be ideal. The image below shows an n+ sample after the ridge pattern developing. n+ means that it is only a doped GaAs substrate and there is no QCL wafer structure in it. This sample serves only for tests – I do everything on this one before moving to the others to see if everything works fine.

n+ sample after 1st developing

Then comes the etching, of which I have already wrote a few words and I will write more because I know it better now. After etching, I had to align a mask for bottom contacts placed on either side of the ridges. On Friday I spent 3 hours trying – every time I looked at it under the microscope, it was completely misaligned and one of the bottom contacts was closer to the ridge than the other. You can repeat it as many times as you want, but it takes time: cleaning with acetone and IPA, spinning, aligning, 5 minutes in chlorobenzene, 2-3 minutes developing, 2-3 minutes cleaning. Altogether one repeat took me almost 1 hour. As I am not allowed to stay after 5:30pm., I went home and continued on Monday morning. Somehow I managed to align the n+, but my real samples only made me feel miserable. When Yash came, we realized that I had been using a wrong size of the pattern: there are 5 or 6, each designed for different width of the ridge. How come I had not noticed it before? I think I just forgot that Yash told me they were different sizes.

Now, when I knew which pattern to use, things began to go smoothly… Or at least a bit faster than before. Developing itself is tricky, because the time needed to fully develop the resist varies from one sample to another. For example, the image below shows an underdeveloped one:

You can see two other contacts (the top stripe and a part of the bottom one) that are fully developed, therefore it was unexpected to see only one underdeveloped. Fortunately, it only takes 5 more minutes to repeat this step. After that, I evaporated metal (AuGeNi) on my samples, but I will describe it in another post. Once it was evaporated, guess what I could practice… more aligning and developing! This time the former went smoothly – I even used a mask aligner that nobody seems to prefer, although I do not know why. See below what happened next: the stripe on the ridge was perfectly developed but the resist looked really weird. We changed chlorobenzene and developer for fresh ones. The resist was changed earlier that day.  We redid the process with the same result.

Something happened to the resist

Confused, we tried another resit (1828, the thickest we use) and finally we got what we wanted. although still not understanding why it went wrong. Then I faced yet another strange behaviour: I had to redevelop it for 5 times before the resist completely went off! So it took 8-9 minutes, whereas normally 1-2 is sufficient. Ladies and gentlemen, I proudly present my sample ready for top contact evaporation! If you look close, you can see that there is a bright stripe developed on the ridge (the middle rectangle). On the right side of the ridge, there is a small dot of remaining resist, but it does not matter as it will be cleaved later.

After three days of processing

It is the second part of this story. When the mask and the sample are aligned, the latter has to be exposed to UV light for a few seconds. We use a positive resist, which means that we want exposed parts of the sample to vanish during the next step, namely “developing” (sounds a bit like old-school photography, doesn’t it).

All the chemical steps of the process are done on a wet bench, that is a small, partially closed area with conditioning system to remove fumes and toxic gases. The sample is immersed in a developer which dissolves the UV-exposed parts of the resist. Then I can check if everything went well under a microscope. There is an interesting invention involved in this part of processing: a DI weir. It checks the purity of deionized water by measuring its resistivity. Pure DI water should show around 18MΩ·cm, so we put the sample on a tray and insert it into the weir, obviously contaminating the water and increasing its resistivity. We usually wait until it goes back over 10MΩ·cm – then we can assume that the sample is clean.

Another interesting fact about the UV lithography room is that it is yellow. It has filters on all windows (a kind of filtering foil I think) to prevent the high-energy part of the Sun radiation from penetrating our samples. One feels as if in one of van Gogh’s paintings. I am sure people sometimes leave the room with their samples in their hands before developing… Fortunately, you only have to clean the resist and spin it again if you do that.

Developed sample with some resist debris.

Next comes a step in which timing is essential: etching. Sometimes you have to re-etch a few times before you can proceed, because etching time is not always the same. I haven’t figured out yet what exactly can affect the etching rate, but the trivial answers should be: impurities and material properties. Wet etching, as opposed to dry etching, is anisotropic – it means that the walls of my device will be diagonal (see the picture below). To check if our structure was etched properly, we use a clever tool called DEKTAK, a stylus profilometer. I had thought that measuring surface roughness with nanometre resolution was done only with AFM/STM so I was surprised when I saw their ancestor in operation! I wonder what drawbacks it has: does it damage the sample? How is the movement of the stylus detected? I have to find out.

Notice that the etched walls are sloped

It is time to write about what I do in the lab, because you would soon conclude that the only thing I am doing here is sightseeing. I am sorry that there are no photos, but you are not encouraged to take a camera with you. Hopefully I will manage to smuggle it once or twice later.

Last Thursday my cleanroom training began. It should have started earlier, but even here sometimes you have to ask a few times before something is done. At first Graham showed me the cleanroom and most of the equipment that I will be using. Then Melanie and Shruti started showing me the processing steps, but only today I got hands-on and began making my first HEMT (I do not even know how it works yet, but it is only for training purposes, my structures will be different).

Scribing and cleaving is the first step and is relatively easy: all you have to do is a tiny scratch with a diamond scriber parallel to the [100] plane, push it with you finger and watch how the crystal breaks along one line. Then the real fun begins. I prepared the etch (which is H2SO4, H2O2 and H2O) and left it for some time. The next step is to spin the resist on the sample; we have two resists that result in different thickness but as far as I know I will only be using the thicker one. It is an oil-like magenta liquid in a syringe – you only have to cover the sample with three or four droplets. Spinning creates a thin layer of resist coating (the resist I use is marked 1813, which means the thickness should be around 1.3 microns when span at 5000rpm). Do not ask me what 18 means or what the resist is made of – it is a polymer, that is all I know now and nobody seems to show much interest about its structure. It works.

There are two lithography methods that we use: UV and electron beam, but I do not know if I will have the opportunity to use the latter. You have to choose a mask, position it properly, and do the same with your sample. At the moment, I think that UV lithography is the most difficult part of processing, the reason being the alignment of the mask and the device. It is  relatively easy when you are doing the first pattern – it just has to be aligned with crystallographic planes. However, if you are processing a sample that has already a pattern on it (e.g. evaporating top contact when the bottom ones have been done), it is a meticulous job. Fortunately someone invented alignment marks: little “+” around the pattern that help you. However, I was trying to align one of Shruti’s samples on Monday and after 10 minutes I had an impression that it was worse than the initial (quite random) state.