The Stewart Mackenzie Indaba #1 Sébastien Bourdeauducq
May 25, 2020 17:52 · 3295 words · 16 minute read
and, we okay we’re not doing headphones, and three two one, and we live my friend Welcome Sebastian. This is now our second rerun, second try, the first one was a bit of a failure, we ended up too drunk to do a decent job of this. So this time we’re gonna do it again with better audio and things are gonna be more organized. Sebastian you’ve got a bit of a background in physics for a number of years but recently you’ve sort of moved into the domain of ion trapping. Could you please tell could you please tell us what what exactly is ion trapping.
Well first of all, it’s not that recent, I mean it’s we’re going 00:50 - back 2013, now it’s already been seven years. Well, I mean this is you, you specifically, has moved into ion trapping right, maybe, you can let us know. So what the ion trap is, it’s a device that lets you control single atomic particles like atoms one by one basically and completely isolate them from the environment, at which point you start getting very interesting quantum mechanical effects. Earlier proponents of quantum mechanical quantum mechanics theories were saying things like, you know yes, this theory doesn’t make any sense, this theory is completely absurd, but in practice we never experiment with a single atom, we always experiment with a large number of atoms and then the contradictions disappear when you experiment with a large amount of atoms. It’s not possible to explain with one single atom and doing this would result in ridiculous consequences and actually ion trap is the technology that lets you manipulate single atoms and actually see the actual ridiculous consequences experimentally.
01:44 - You can see exactly what? The ridiculous consequences. The ridiculous consequences. I think I think I think Schrödinger the famous scientist yes right right The Schrödinger equation, the Schrödinger cat, I think that’s the word that is used. “Ridiculous” yes, something like this, yeah. But for our audience, what exactly do you mean by “Ridiculous”? and maybe you can go into what Schrödinger… but before you do that…. can you go and get your lighter this one thing what I mean by… huh?… we should close this door if we smoke… we’ll definitely close to the door. and set the ventilation… okay okay ummm…. well what is ridiculous I mean all the, all the, things that people, all the paradoxes of quantum mechanics like things which are in superpositions of states the famous paradox of the Schrödinger cat which is alive and dead at the same time.
This kind of thing you can actually see them in the lab 02:32 - using this Ion trap technology. Okay… so prior to ARTIQ… technology… rather limiting technology… I mean there’s been great leaps and bounds in in ion trapping control systems relatively recently and this is primarily down to ARTIQ. Is that correct? Well ARTIQ is one of the most used control systems, and one of the most feature full control systems these days. I mean there are many things that limit ion traps is not just a control system control system is one of the frontiers but there are many things that make ion traps not work. It’s a multi-faceted problem I mean… ARTIQ is solving one problem but the many other ones mm-hmm what are these different sorts of facets which are problematic okay so what is the objective you want to be able to control multiple ions at the same time? One of the problems is that the problem is not in general very well-defined…
like if you ask multiple 03:38 - different academic teams they will try to solve the thing in multiple different ways so one of the problems if let’s if you want to build a quantum computer one of the points defining the architecture of such a quantum computer and there many approaches to this to this way of doing things and no two academics labs agree with each other. Okay… so while ion traps can be applied to multiple different sort of domains… is that correct? Well doubly domains, yes I mean as I said as I said already I was already assuming that you want to build a quantum computer, which is I think one of the very exciting applications. Sure! But this isn’t the only thing that people use ion traps… No no, but what are the different domains then? Well for example, there is quantum sensing like you use those single atoms as sensors, very high precision sensors I mean when you’re one of the things that make quantum computers difficult is that any kind of environmental disturbance is going to mess with your the computation you’re trying to run in that machine, but quantum sensing is actually taking this to its advantage and using the qubit as a sensor. Okay…
So you can make very high precision 04:45 - magnetometer for example. Okay so… Or you can measure gravitational fields, uhh, lot of sensing.. things, you can make like in a inertial measurement units with it, gyroscopes this kind of thing… so sensing is one domain… another big domain is the clocks, atomic clocks, you can use those qubits is very stable time references. Yeah, what’s the latest happening in this atomic clock domain? Atomic clocks? I mean they’ve managed to shrink the size of these atomic clocks significantly haven’t they? Uhhh… well…
kind of I mean 05:24 - there was the Opticlock project which just finished recently that uhh was about taking one of those room-sized atomic clocks and turning it into a device that can fit into rack units… so… Was that achieved? Yeah yeah, it works yeah. Now that project, Opticlock is finished now right? So now your partner Jordans is gonna move on to something else now right? yeah. Okay so! so we got we got the the quantum timing, we’ve got the sensing… there’s computing… yeah…… and anything else? well there are some people who do experiments to test the limits of physical theory. Okay can you expand on that…
Well for example measuring 06:15 - fine-structure constant, this kind of thing, with very high precision, and see if there was any discrepancy if there is any anything weird with it there’s this, or just better knowledge of it… uhh… Right well go into the quantum computing side… there are also things, there are also experiments that try to find variations of the standard model Ahh, that was, recall that lecture that we went to was it here no no it was in Boulder Colorado, the LOCC That was in Hong Kong! Oh that was in Hong Kong yeah yeah that’s right That wasn’t about ion traps.. it was very theoretical quantum information talk… yeah yeah theoretical… yeah okay okay but this this talk already assumed the basic principles of quantum information and then just build abstractions on top of it. Yeah… There was nothing really exciting in there. Yeah another test of fundamental properties of the universe is people try to understand gravity so they would for example put iron traps in free fall and see if there is any uhhh… well that’s fascinating…
Yup I mean these these devices are rather sensitive like 07:26 - sensitive to any sort of thing So you are going to drop them down a tower… isn’t that experiment done in the world’s largest vacuum chamber, which is literally room sized vacuum chamber with these massive doors. Well I don’t think is the world largest vacuum chamber, but it’s a pretty large vacuum chamber yes it’s a large tower, a large cylinder which is under vacuum because you don’t want air to stop the free fall of what you’re dropping in there, so it’s a pretty large vacuum chamber.. yes and umm then you drop the experiment in there… they make the joke that you are dropping your thesis…
Oh okay okay 08:57 - all right, well maybe we can go into the more interesting side of ion trapping which is the quantum computing side… but before we do that before we open up that chapter. Shall we smoke a cigar? We can close the doors and all that stuff We can do that sure. Just take it out… alright it’s boring, I mean who gives a shit we are smoking cigars smoking and drinking, those are the most stupid occupations in the world smoking and drinking are some of the most stupid occupations of the world. well, they’re not exactly occupations, bro.
Well, not occupations, yes they’re activities 11:06 - all right throw the lighter my way… dude, that was a naff throw, you know about underhand throws right? You are going to ruin your cameras. No I don’t think so… it doesn’t matter. So the fan is running right? The fan is running yeah, but do you want the bathroom door open? Oh the bathroom door is closed? Yeah you should open the bathroom door. and run the extractor fan. Actually, I’m not sure if the door is open or not… fan is running… door is open… Alright! Okay, let’s open up the next chapter. Quantum computing, so Sebastian now there are multiple different pathways to being able to achieve quantum computing mm-hmm you know, each of them is difficult but one of the higher, one of the higher probabilities or one of the pathways of greatest likelihood of success is that of ion traps. Well it’s one of the competing technologies yes. Yeah, yeah yeah…
Why? Why? 12:32 - because I ion are pretty stable, like if you store quantum information in an ion, it’s going to stay there for quite a long time, without decaying or anything like this and also when you have interaction between… if you make a quantum gate between two ions, the fidelity, or that the accuracy of the gate is going to be quite high and higher than other technologies. But that’s just two ions, I mean to do a realistic quantum computing setup you’d need to have…. Well that’s the thing, if you.. of course you need many of them but.. uhh if you just this system which has many ions and many operations is just you know accumulation of very simple operations between pairs of ions and qubits in general, but each of those basic operations has a certain error rate after you have a number of operations, those errors accumulate.. in the the end you just have noise and garbage out of your quantum computer and the amount of garbage that you get, the amount of noise is less with ion traps than it is with most other technologies.
You can actually see it very 13:32 - clearly I mean there was this IBM Q. Well I think it’s you can still join it this is IBM Q quantum computer in the cloud I mean they did a lot of PR on this You can actually use a… superconducting qubit machine and run very simple quantum circuits on them and you can see that once you’re past just a few gates the output vectors from your experiment is going to be very far from the theoretical values and this error is pretty high for those superconducting qubits from IBM the IBM Q is not really the world-class machine superconducting qubits but it gives you an idea that there are serious issues with this kind of technology there are also issues with ion traps but they’re not as bad. okay so I know that also Google’s got this D-Wave. Ah but D-Wave is doing something else entirely It’s not quantum computing is it? Well it’s using quantum effects but it’s not a general-purpose quantum computer.
It’s just the same thing as 14:31 - Google this they claim they achieve as they call it “quantum supremacy” which is basically like a speed-up that couldn’t be achieved on a classical computer and again don’t do it as a general-purpose quantum computer they just showed that they had quantum they had some some better results for a very specific algorithm for a specific problem… than a classical computer…. here’s the lighter.. I mean at what point do you call your thing quantum computer? right? I mean you could just take a neon light and determining the spectrum of a neon bulb very high accuracy is very difficult on a classical computer but a “quantum computer” made with just simple neon bulb is.. is… achieves also “quantum supremacy” because it gives you the wavelength with better accuracy than your classical computer so… but people are not calling neon bulbs “quantum computers” right? So this is an absurd example you know it gives you the general problem with those claims of quantum computing of quantum supremacy which are not really built on a general-purpose quantum computer but only on a very constrained problem that’s only one little part of the story. well I mean I suppose by rather like.. you know… can we save more traditional definition of computing is like you have an import transformation and an output.
Right? Using a light bulb to sort of 16:00 - determine…. Well yeah you can charge the current in the light bulb and then.. Ah all right! Okay fine! you have an input right? Or the temperature, something like this! Well I mean the input is very limited in that way, but having some sort of a programming language… Yeah but you don’t! Huh? You don’t I mean most of those DWave Google whatever… They don’t have programming language inputs? Well not really I mean you can you can change certain things but it’s not a general purpose programming language no. IBM Q can do…
16:31 - IBM Q is so general-purpose quantum computer but not the D-Wave stuff and not Google quantum supremacy thing. mm-hmm They are more like light bulbs than quantum computers… oh so D-Wave is more like a light bulb? yeah it’s a programmable light bulb. it’s a probability light bulb a a programmable light bulb… oh I see all right. It does something called quantum annealing which is more a crystallization something like this… rearranging things…
What does that mean? Quantum annealing, which is like crystallization 17:04 - crystallization of what? well it’s an algorithm which is similar to simulated annealing. Simulated annealing is uhh…. Wait, what the hell is annealing Annealing? yeah… It’s when you heat a material and cool it slowly… and then every bonds right? No, it doesn’t bond, it becomes soft well in general it becomes soft because its internal molecules are reorganised to minimize energy. For example, it would crystallize I mean while it becomes soft depends it metals become soft but if it’s a crystal then it becomes a bigger crystal and not something which is made of smaller crystals.
So it’s something which can be modeled I mean if 17:52 - you look at the temperature curve and behavior of atomic particles… it’s uhh How to explain this? oh it’s so hard to explain…. I’m not Feynman, Feynman is very good explaining this not me. No that’s fine I mean… okay so DWave is this quantum annealing sort of system It’s like a light bulb basically How do you explain this annealing? yeah when some material has some thermal energy, temperature all it’s particle its fundamental like all the atoms molecules are like jiggling around and the higher the temperature the more jiggling happens yeah and if the temperature is high more molecules of higher energy and there is more jiggling and more rearrangement of the material structure that can happen and when you lower the temperature there is less and less, fewer and fewer of those rearrangements become permitted because the particles don’t have so much energy so they cannot really overcome shocks or the barriers between them and so when you when cool a material slowly it would, for example, make bigger crystals because since they had to spent more time in a higher temperature region the the interaction would allow larger structures to form are permitted whereas if you cool it more slowly, only like short range interactions are possible and you get lots of very small crystals but people have used this physical phenomenon to build an algorithm called simulated annealing where you try to optimize some difficult problem, like for example the optimal placement of certain graph like you have certain nodes which are connected and have certain interactions between them and you want to rearrange this graph by placing nodes in such a way that minimizes all the length of the edges between the nodes of this graph. That’s a pretty difficult problem to find out optimal positions for this and the simulated annealing is a good optimization technique for this like you would basically jiggle around those nodes randomly and then you’d look, if the move that was selected randomly is going to increase or reduce the total length of cabling of the graph The total length of what? Let’s say the total length of cabling (the edges of the graph) for example to sum of all the all the distances between the nodes which are connected.
All the edges and then you understand 20:45 - probability of accepting or rejecting this this move if it increases the length of cabling which is what you want to reduce and then you decrease this probability slowly just like you are cooling a crystal to make very nice cooling material to make a very nice we order of matter very ordered structure and it’s actually quite a remarkable thing because this this physical phenomenon that you see which is microscopic phenomenon that produces those very ordered materials on a scale when you cool them properly, can be simulated in a computer and applied to other domains then crystallizing solids but can be also applied to optimizing graphs or there are types of things just using this very basic thing. People sometimes call this God’s algorithm. It’s something which happens at a very fundamental level in microscopic systems and produces macroscopic results. If you look at a snowflake, for example, it has this very beautiful structure but this structure is just built using those very microscopic interactions between water molecules and using a certain temperature profile that allows or disallows certain motions of the molecules at the macroscopic level and… that’s simulated annealing and what d-wave has been doing is something quite similar but instead of using classical statistical mechanics which covers the movement of those molecules, they use certain quantum effects which also have this kind of random nature, with acceptance or rejection of certain moves.
So it’s something rather similar to simulated 22:22 - annealing, that you can run on D-Wave machines, but it’s not a general-purpose quantum computer!.