I’ve been looking forward to this for quite some time – two weeks of work experience in a lab at Imperial College working with some PhD students with an ultimate goal of making some progress towards the construction of a quantum computer that doesn’t take tens of man-hours to perform each calculation. After a week I feel I’ve learnt a lot about formal lab work and large-scale experiments (*slightly* different from the 20-minute assessed practicals from AS!) and about the general physics and concepts behind some of the experiments and equipment – I’ve been (not so) conscientiously filling pages of my notebook with messy notes and cryptic diagrams so hopefully some of the stuff I write here will make some vague sense and not quite directly contradict truth.
There are also several interesting bits and pieces lying around the place. There’s an enormous Newton’s Cradle in which each ball looks like it could be heavy enough to be a ship’s anchor. There are also enormous capacitors lying around everywhere for the people working on high-density fluxes.
The Physics / Maths / Computer Science part of Imperial is somewhat bizzarre – from what I gather it consists basically of two adjacent buildings which were built at different times and were haphazardly connected together by knocking down bits of walls. Unfortunately the actual floor levels are out of alignment and the floor heights are also different, which means there’s a crazy staircase joining the two buildings together and floors 6 and 7 in one of the buildings had to be rechristened 6 and 6M for the sake of keeping the numbering consistent with the lifts. There are also only connections on certain floors of each building so it’s possible to leave the Blackett Lab on the bottom floor, go up one level and be confronted with a solid-looking wall where the connection should be. As if things aren’t crazy enough, there’s a set of lifts placed almost exactly at the junction (so to speak) between the two buildings, making it really confusing to navigate the whole 3D maze. It’s pretty good fun actually!
The Lab and Equipment
I was quite surprised when I first saw the lab – I was expecting a Leonard Hofstadter style lab (as seen on The Big Bang Theory) but actually quantum computing with ions (ions therefore being the main focus of most of the projects which I’ll come to later) involves a fairly large amount of optics work, so each of two adjacent, connected labs I was working in has an optical table as its centrepiece littered with lasers and ridiculously complicated setups of mirrors and lenses which have been tuned very accurately to direct laser beams into tiny optical fibres and whatnot. Speaking of accuracy, the setups are so sensitive to small shifts that they need to be tuned almost constantly. The PhD students told me they detect a lot more drift during the daytime when other experiments are going on in other labs which release radio waves and traffic is rumbling overhead (despite being two floors below ground level and over a block away from a small road) than at night when there is less activity.
The equipment is also sensitive to tiny temperature fluctuations. Most of the lasers are basically diode lasers:
It’s a fairly standard laser setup in which electrons and holes come together in the depleted region between n and p type semiconductors and then either wait for a nanosecond or so before annihilating and releasing a photon (spontaneous emission) or get hit by a photon, resulting in stimulated emission. What was experimentally interesting was that the cavity length in fact determines the wavelength of the laser owing to the fact that a standing wave needs to be created which ‘fits’ exactly in the cavity (a whole number of half-wavelengths need to fit in the cavity) and this is sensitive to temperature. So each laser box has four BNC sockets: one for providing the laser with electricity, one for a thermistor which is hooked up to a feedback loop system which regulates temperature using a Peltier junction heat pump (which occupies another socket on the laser), and one for a piece of Piezo (placed on the diffraction grating) which can change width depending on the voltage across it (or maybe current through it, or something) thus allowing the cavity length to be adjusted, though my suggestion to manipulate the piezo in the feedback loop to compensate for temperature changes would fail since the temperature-dependent expansion of the cavity is several orders of magnitude greater than anything the piezo can correct. When I heard that I was pretty astonished the laser cavity had to be adjusted to such an exact length – several orders of magnitude more exact than the expansion of a bit of metal when raised by a few degrees. The entire laser is covered by a thick black piece of foam to protect it from temperature fluctuations in the room.
It was pretty cool to find out that I was to be working in the same room as a 2.5 Tesla electromagnet! Ion trapping, as I will also come to later, involves not only charge and potential fields but also magnetic fields, so the Penning Trap the researchers there were using was sitting inside an enormous superconducting electromagnet.
The superconducting electromagnet uses liquid He to keep cool – as a sidenote I asked why they (and CERN) use cool superconductors (more expensive liquid He) instead of the more recently discovered crazy warm ones (cheap liquid N2); the reason is because above a certain current, superconductors end up failing and develop some resistance causing heat to be produced resulting in a quench (the He boils off, expands to something like 15x its volume and the whole can explodes in a fit of freezing fury), and the cool superconductors can carry a much higher current before this happens, allowing more powerful electromagnets. Of course this comes at a very high cost. As can be seen from the diagram, the He (at ~4K) is shielded from room temperature by a layer of liquid N2 (at a balmy ~77K). The He needs to be replaced about once every couple of months, while the N2 is replaced about twice a week. The superconducting coil, power supply and cables are eventually going to have 80A coursing through them – a truly formidable current!
Apparently the way they get the electromagnet to start conducting current is to simply arrange the coil in a loop – they can’t expose the 4K superconductor to air, so it is necessary to induce the current in the superconducting coil. Once this is done, the power supply can be switched off and the current in the superconductor just keeps going round (owing to the lack of resistance), allowing a very strong noise-less magnetic flux to be produced (the flux’s precision is something like 10-6%)
This was particularly new to me. I’d never worked with 2.5 Tesla or 5W lasers before, but while I’ve come across magnets and lasers in experiments, I’ve never really observed experiments involving vacuums before (apart from the bell-ringing-in-a-jar/gerbil-squeaking-in-a-jar one to show sound doesn’t travel well through a near-vacuum). There’s a lot of novel (to me) and interesting experimental stuff that goes on here.
Basically the idea of creating a seal when joining two flanges together is to use a copper O-ring. Each flange has a ‘knife edge’ (90° very sharp edge) and when they’re pressed together with a Cu O-ring in between, the knife edges cut into the soft Cu; thus the Cu itself becomes the seal.
While flicking through Inward Bound by Abraham Pais (recommended by CAPS) I read about various attempts at making a good vacuum pump. Modern technology has come a long way since the mercury-filled jar, and now creating a very good vacuum is a multi-stage process. First all the equipment is cleaned thoroughly – for some reason fats and oils from people’s hands (for example) are disastrous for a vacuum so everything needs to be wiped squeaky clean with something like acetone or isopropanol. Then everything is sat on the optical table which has a source of clean dust-free air on the ceiling which constantly blows on the equipment, keeping dust off and constantly cleaning it of bits of dust that have settled. Then everything is put together using gloves, nuts, bolts, Cu O-rings and *a lot* of effort (believe me, putting flanges on sideways while stopping the Cu O-ring from slipping out is infinitely more difficult than measuring SHM of a cork in a tub of water – reference to AS practical; one of the researchers also described putting He into the cannister having first cooled it sufficiently to stop everything boiling off immediately as a dark art rather than a science). The air is pumped out using a conventional pump until the pressure inside is something like 10-6 millibars, at which point an ion pump is turned on to essentially evacuate the remaining air molecule by molecule. The pump essentially ionises the gases inside the chamber and use charged plates to attract them out. The final result is a very good vacuum.
There was a MSC researcher from Germany sharing the lab with the PhD students from Imperial, and he was working on a different method of ionisation. The supervising prof, Dr Danny Segal (a reader in Quantum Optics), explained that the previous approach to getting ions was to use a ‘splat gun’ approach – basically a stream of neutral atoms from an oven hits a stream of electrons from an electron gun, and those electrons will tend to knock out some electrons from the stream of particles, resulting in a few ions. This has a few problems: lots of atoms never get ionised so end up getting deposited on the side of the chamber, screwing up the shape of the potential well in the ion trap; lots of electrons end up floating around in the chamber and get deposited on insulators, again causing irregularities in charge distribution.
The German MSC researcher was working on using photons to create these ions – a much more tenuous stream of neutral particles is projected into a beam of photons which, via the photoelectric effect, knock out electrons creating ions. This should have a higher rate of ionisation leaving fewer ‘waste’ atoms sticking to the inside, and the number of photoelectrons knocking around the chamber should be much lower than the number of electrons being shot from the electron gun. I suggested the photons might knock electrons off other bits of the apparatus, again screwing up the flux; apparently this should happen infrequently enough to allow a reasonably controllable flux, though some researchers using a Paul trap (involving an oscillating EM field) apparently detected ionisation using photons of the wrong frequency for direct ionisation leading them to believe electrons were being knocked from the apparatus and these, accelerated by the oscillating field, slammed into atoms causing ionisation.
Anyways the setup was more or less thus (a picture is worth a thousand words):
The way to ionise these Ca atoms is to first use a laser to excite the atoms – push some electrons up to a higher energy level. The UV LED then does the actual ionisation from that energy level. The picture above is of a half-finished setup (optics haven’t been sorted out yet and there are two unsealed flanges).
GSM/KPZ gave us an article in class last year about laser cooling (‘Cool things to do with lasers’, Ifan G Hughes et al 2007), and it turns out it’s useful for Quantum Computing – a jittery ion is presumably pretty bad for physicists who want a stable wavefunction. Well, here’s the setup.
It’s in fact mostly about using lasers to manipulate electron energy levels in a Ca+ ion:
The Ca+ ion has an energy level electrons can fall down to (RHS of diagram) where they would stay for quite long before falling back down which is undesirable considering the cooling involves shuttling electrons between the leftmost levels (in the diagram). So four red lasers are required to pump those back up to the top energy level.
I’m not really sure what’s going on in this experiment but basically, since QED is only significant at high charges (something like that), the researchers go to GSI to conduct this research. The idea at GSI is to slam super-high energy ions through a gold foil which apparently strips them of all electrons. Different ions are separated via a very similar system to how a mass spec works.
Some Other Physics-ey Stuff
There are of course lots of different methods of trapping ions – I mentioned one in my post about the UCL antimatter lecture. Apparently it’s provable from Maxwell’s equations that it is impossible to create a static 3-dimensional potential well to trap ions, so there are currently two main methods: using a purely electromagnetic system (using either some feedback system to wobble the ions towards the centre of the trap or a constantly oscillating field like in a Paul trap), or to use magnets:
A cation is sitting at the bottom of a potential well in the z direction. It is surrounded in the xy plane by oppositely charged plates. As it is attracted to the plates, the z-directional magnetic field causes it to move in a circular motion (as seen in cloud / bubble chambers to determine momenta of ejected charged particles), represented on the diagram by ‘micro OOO’. The charge on the plates are then somehow tuned to make the large-scale motion of the particle resemble a circle and so it eventually loops back on itself, so its path shape looks like what is labelled in the diagram as ‘tuned, get O’
The Ca+ Ion
The actual quantum computation to be done with the Ca+ ion (not a typo: just one +; this isn’t chemistry!) involves electron energy levels. An electron can be in one of two energy levels, and that is the qubit. In Ca+ there are two more or less independent distinct situations in which an electron can be in one of two energy levels, allowing two qubits to be encoded into one ion.
The use of this isn’t only to cram more qubits into fewer ions (I read a research group somewhere is making base-5 ‘qudits’ using microwaves and superconducting things) but also to allow easier entanglement – since both qubits are in the same ion it’s supposedly easier to make them interfere in a predictable manner, which allows a quantum NOT gate to be set up which is critical to quantum computing; supposedly only two research groups in the world have managed to get this quantum NOT gate to work.
The biggest limitation apart from the sheer fiddly-ness and slowness of everything in the quantum computing world is the fact that it’s impossible to put more than about 8 ions in one trap before they start screwing up each others’ wavefunctions. The PhD researchers had previously been working on a solution to this problem that the theorists came up with – getting the ions to hop around in the trap, thus manipulating each ion more or less individually. This has already been done using Paul traps (I think) but the researchers here were trying to use Penning traps and show they are in fact better for quantum computing (or at least can do the same things as Paul traps).
There’s a lot more I would say if I had the time but as with all blog posts, you’ve got to stop somewhere. But overall I’ve never done modern practical physics before (at UCL we looked at some particle traces on the computers which is the closest I’ve really got so far) so this is a pretty damn amazing experience for me, hence the mega-post.