During the fourth year of an MPhys degree at Bath, students are required to pick a research project to join for semester one. Our names are Rebecca and Sarah and we selected to join the atomic manipulation group. After completing the STM lab in 3rd year, and being given a tour of Peter's 4W lab, we were excited to begin. This weekly blog is a lighthearted view of the trials and tribulations of our Masters Project. Week One: Tent, Tips, and Tension
So we arrive in the lab on Monday morning, filled with anticipation and excitement at the challenges ahead of us. The first thing we notice is the heat...then the tent...then the frustrated look on Duncan's face. Turns out the poor little STM isn't quite ready for our arrival, and we soon learn the term 'baking'. (Not the cupcake kind we are familiar with unfortunately). Thankfully, being at this stage worked in our favour, as it meant we got to follow step by step how to set up all the equipment. Once the tent was off and we could finally see the STM in all it's chrome glory, we had a whistle-stop tour of how it all works. To clarify to ourselves what does what, we invented a game called "Vacuum or no Vacuum", where we imagined various valves being opened or closed and had to work out what situation would be occurring. (Don't worry Peter, we didn't really touch anything!). One of the first things we got to be part of was cooling the system down with liquid nitrogen to improve the vacuum. After the simple signing of a health and safety form, and the addition to our outfits of a garish yellow glove, we were finally able to help...even though lifting the Dewar flask was quite a task in itself. The day we get bored of liquid nitrogen is a day we don't want to see. After learning a bit about flashing the pumps, we were ready to move onto our next challenge: tip etching.
To obtain a sharp enough tip, electrolysis is used to etch away at some Tungsten wire. With the help of a cleverly adapted microscope stand to keep the cathode in place, and a lot of wires, we set it up ready to watch the magic happen. Ten minutes later we were still waiting and finding it hugely less exciting than we first thought. However the joy came back when through the microscope we could see this...
Unfortunately it didn't last long as the tips began to get progressively worse in quality; some were too long, most were 'jaggerdy', all were below par. During our one last attempt to redeem ourselves...it stopped working completely. After various attempts to rectify the problem (none of which worked), it was home time.
With Friday came the next stage of equipment preparation: getting the tip in. We crept quietly round the lab as Duncan mounted it, watched expectantly for the orange glow during annealing, and held our breaths as Peter attempted to transfer the tip to the microscope... Having had a play with the wobble stick ourselves, we understood the tense atmosphere (it's not called wobble for nothing!). Unfortunately this tip was not the one, and as it fell onto the mesh below the preparation chamber, Peter and Duncan sighed and groaned. We however we just glad we weren't the first ones to drop something!
Week Two: The Waiting Game
5 tips down and no more spares: we had a problem. The tip etching station was still not playing ball so it was time to act the detective. The usual tactic of 'turn-it-off-and-on-again' wasn't helping, and changing the sodium hydroxide electrolyte made no difference, so we used our trusty friend the multimeter to look for any shorts in the circuit. This narrowed the issue down to the trip box, where Peter discovered the problem...10 Volts in...2 Volts out... Uh Oh!
After opening up the box we were greeted with this:
The little burnt resistor hiding in the centre of the image was the culprit! After Peter's re-soldering we were back up and running. Using the last of the Tungstan we cautiously made 3 last tips, being very careful at every stage not to cause any vibrations.
But our tips need something to scan, so next on the preparation list was the Silicon sample. Once it was mounted and placed under vacuum, it needed to be cleaned of any invisible muck, that would prevent good pictures being achieved at atomic scale. Said muck is blasted off by passing high current through the Silicon, causing it to reach temperatures of 960-1200 degrees (and glowing a beautiful orangey-yellow). The temperature is measured using an infrared detector gun, while balancing precariously on top of a step ladder. Keeping an eye on the temperate is necessary in working out what current is needed to achieve the temperature of gunk-blasting. This particular view point also enabled us to spot an unexpected extra glow near the sample. At first it was thought to be a mere reflection of the sample's shininess, but when we realised the temperature of this 'reflection' was actually hotter than the sample itself we decided maybe there was more to it! In fact it was a ceramic plate that was accidentally part of the circuit. So more rewiring.
However even then all was not quite well. As we tried to increase the current to raise the temperature, the power supply got stuck at 2.6A regardless of what was asked of it. Many suggestions later Peter again came to the rescue by debugging the program. This done we were finally finished with sample flashing, and ready to move on.
So: we had our tip (hoping it was sharp enough) and we had our sample (hoping it was clean enough), now time to put both together and try out the microscope! After learning how to approach and withdraw the tip so the two are a few nanometers apart, we finally got to do some imaging. Unfortunately the images were quite noisy and we've decided to change both the sample and the tip for better ones...
Week 3: LabVIEW for beginners
This week involved a lot more waiting...trying to get a good tip and a clean sample at the same time was proving difficult. After trying various combinations of our tips with different samples we were still only getting sausage shaped 'atoms' and a noisey picture. As we're still getting to grips with the 'tricks of the trade' to get a clear picture, we weren't very much help with the nanonis software. So we largely left it to Peter and Duncan to play with the settings and try to get a clear image whilst we watched closely to take in all their sneaky cheats.
We tried to make ourselves useful by mounting another sample, something Duncan had shown us twice before so we were feeling confident to have a go... Apart from the massive increase in time for us to complete the same task, we successfully managed to attach the sample within the holder and didn't even lose any of the tiny nuts and bolts! Testing the resistance to make sure there were no shorts and that the current would flow across the sample correctly stumped us for an embarrassing amount of time! But we got there in the end! Here's a picture of Sarah struggling with the fiddly job...
While Peter and Duncan were tinkering away at the microscope we were set a mission (should we choose to accept it...) of programming. We're trying to use LabVIEW to create a code that performs cross-correlation on the images obtained from the STM. Once it's written this will compensate for the shifts seen in the images due to the expansion and contraction of the sample set-up (due to temperature changes). We have so far worked on this for a few days (mostly borrowing other people's code and pasting it together), and here's what we've got to:
Having never heard of LabVIEW before, to begin with the progress was very slow (mostly spent looking up terms on Wikipedia). However we are getting more of a handle of it now, and it's exciting to be learning a new skill.
Week 4: Plane pain
This week was spent buried in the depths of LabVIEW. We had so far managed to take the two images and cross-correlate them, but could not make head or tail of the result. This was apparently because the original images were too wonky and so to make it happy we should take a global plane (like a line of best fit but in 3D...not an aeroplane like Google thought!) off. After getting stuck on our own for a bit, we headed over to ask Philippe Blondel if he could spare us a moment to help. It's so lovely that our department is small and friendly enough to have lecturers know who you are and go out of their way to help with any problems! Philippe taught us all about Kernels, although after implementing this we were still left with the same problem. So back to the global plane idea...
Even after trying many different directions of thought we still ended up with a wonky background...it seemed like we just couldn't win! We would arrive in the morning and Peter would have come up with a new idea for how to fix it. Finally on Thursday he presented us with a printed booklet containing a scary looking matrix equation for creating a global plane to fit your data. Would we be able to work it out? Would LabVIEW like it? Was it even correct?? We set off to find out! By breaking it down into simpler steps we saw it wasn't nearly as bad as we thought, and after a while (and a LOT more wires) we had a picture. Still a wonky picture. AHHHHHHH!!! Flustered and confused we sent an SOS email to Peter who came down to help. The three of us spent the next 3ish hours labouring over the code. We rearranged, and added things, and took others away. We tried separate sections, and different numbers. We got the program to print out the results of every step of its calculations just to pin point the faulty section. Eventually we found the source: we were giving an array to an input box, but what the box received was crazily different!!! Much head bashing went on. The issue was with the memory of the program because of the size of the data going in. So in the end all it took was changing the way we stored some data...and hey presto a flat result with a peak in the middle! Just what we were hoping for!
Much excitement followed.
Week 5: We played around with Excel a bit. Then we read. Then we highlighted. Then atoms.
Moving on from our successful end to last week, we next needed to check we really could use the LabVIEW program to predict and alter for the drift. So we checked all the data we had and plotted various graphs to look for trends. It took us a while to get our heads round what we were even trying to test, and even after deciding on a strategy we couldn't seem to find a convincing answer. Data points seemed all over the place! We needed more data...but the microscope was still out of action.
So instead we spent a fair bit of time reading up on what we were trying to do. Google, print, read, highlight, repeat. Some papers went way over our heads, but others were exciting to read as they were discussing very similar things to what we were doing. Hooray for relevant literature!
And then everything seem to happen so quickly. We had learnt not to get our hopes too high up when the suggestion of testing the microscope came up, so when Peter came and found us to say they could see atoms it didn't really sink in for a bit. But sure enough once we had made our way down to the lab...there they were!!
Aren't they beautiful!
Our final job for the week was to make them less squishy. (Also known as calibrating the dimensions, so the size the computer thought it was matched the known unit cells size from literature.) This involved a lot of head bashing, drowning in a sea of simultaneous equations, and much typing into Wolfram (since Rebecca got so mad at the whole thing she 'threw' her calculator on the floor and it broke). We thought our simple way worked, but Duncan's complicated trig proved the victor in the end. Then all that was left to round the week off was a spectacular smash of Duncan's dinosaur coffee mug (Sarah's clumsiness 1, coffee cup 0), a visiting speaker at the nanoscience seminar, and a drink in Plug with Peter and his tutees. Hello weekend.
Week 6: Meet me halfway
50% through the project...
Progress is steadily being made to get the stm back to full working order. Eventually the whole set-up needs to be clean and stable enough to perform experiments on and collect data for several hours at a time. Since the main issue stopping that right now is just tweaking the sample and tip to get them as clean as possible, so our images are nice and noise free, this week we were introduced to the next stage of preparation for actual experiments: chemicals. The group's experimental focus at the moment is studying the manipulation of benzene and toluene on the surface of the Silicon sample. For this, they need these chemicals in pure gaseous form, ready to inject on a nice flat Silicon sample (when we eventually find one we're happy with!).
Enter the freeze-pump-thaw method. The name does a pretty good job of explaining what this is. First the test tube containing liquid benzene is bolted onto the bottom of the stm. Then it is frozen solid using liquid nitrogen, and the valve which connects it to the pump is opened to suck out any air left in the tube. Then we thaw it. The oxygen melts first, so is sucked out while the benzene is still a solid, and once the benzene has itself melted, we repeat the process. This is done until we see no more bubbles of oxygen escaping, and no change in pressure (so we know nothing more is being sucked out). This leaves us with a sample of pure benzene. (We will eventually check this with a mass spectrometer to check we really just have the good stuff.) Here's a picture of Rebecca doing this...
The remainder of the week was spent back in the world of LabVIEW. (Slightly more exciting these days, as we have learnt how to change the background colour of our windows. So much nicer staring helplessly at a sky blue screen than dull grey!) Putting some of the new images taken through our cross-correlation program illustrated noise was an issue that stopped us from being able to correctly evaluate the shifts. Boo. Peter pointed out that what we wanted was a low-pass filter, to get rid of all the yucky uber-high data (from the tip regurgitating all over the sample), and leave us with just the bits we wanted to look at. But we couldn't work out how to connect the pre-made vi to our current program...so we decided to make our own from scratch. This involved lots of histograms, averaging, and a return to the dreaded plane game! We're still working on it.
Week 7: Movie Makers in the Making
This week was very exciting. We're nearly there with the drift LabVIEW program, but we gave up on the low pass filter, as this was deemed unnecessary and takes experts ages...let alone us! Meanwhile Peter has managed to get us some very stable images, so good in fact the thermal drift was pretty negligible! So for us to test our program on it he shifted images randomly to mimic drift. However when we came to cross correlating them, it turned out they were so big that they induced some x-direction drift, causing a parabolic dip in our data (a.k.a it looked like a bowl). Hmm, how do we get rid of this one?! Our global plane fitting didn't fix it, so we thought maybe we needed to fit something else. We pointlessly implemented a rather disgusting matrix to the program, before deciding this didn't help. Then we learnt all about case structures, and put a load of them in the program to let us choose which fit to try. When none worked on their own, we just started trying different combinations. Funnily enough this didn't work either! Then Duncan came to the rescue and told us not to bother...we'd never be using images this big for real data (as this was way too big to actually see the atoms). So we went back to the original. All that was now needed (we hoped) was to bolt it onto the scanning software.
While we were fiddling with all our case structures, Peter had been on the microscope. He had just been scanning a tiny section, and pointed out what appeared to be an atom (or dirt) moving between adjacent sites. It bopped backwards and forwards quite happily for quite a number of images. The next day, when everyone else was busy we decided to have a little fun with these pictures. After a lot of converting individual files to jpgs, changing the colours, and importing to Windows Movie Maker, we produced this...
We like to think of it as an atom, but truth is it's quite possible that the black blob is a bit of 'dirt' that can't make it's mind up where to settle. Although we did it just for fun, it's been really cool having something like this to show people when trying to explain what we're doing. It also illustrates the thermal drift well, and so shows why it needs correcting. It also proves how awesome physics is. We showed off this video at the Physics Department Welcome Evening to all our friends (yes, we're that chuffed with it), but it didn't manage to eclipse the enjoyment of seeing and hearing lecturers hit the karaoke machine!
This week we've again been trying to learn bits of chemistry so we can understand better exactly what we're seeing and what is going on. As we get ready for the next stage (adding the chemicals) we have been reading papers (and googling lots) on the electron configuration of silicon and benzene, and how they bond. While we were struggling with doughnut electron clouds, Peter was putting the finishing touches to the drift correction program. Finally this was done, and it's goodbye thermal drift.
Next on the to do list:
* Fix the x/y calibration (again)
* Stop the vibrations from people in the cafe upstairs mucking up our scans.