Majorana 1: Microsoft Makes Quantum Computing Breakthrough With New Chip

Microsoft's big quantum gamble pays off: they were harder to build (and about as hard to understand!) but Microsoft's topological qubits are more effective, easier to control - and will probably help design their own replacements.

I've been tracking this work since 2016 when Peter Lee (who runs MSR) had just presented to billg and I caught him at just the right moment to get a brain dumb on the science and history of Microsoft and Majorana anyons; it goes back to an Italian physicist and a Fields medal winning mathematician

along the way you've got investment: Craig Mundie funding six experimental physics labs and pioneers of processor architecture like Cray cofounder Burton Smith and Doug Carmean of Pentium fame. you've got hope: so many times the research lab is sure they've really seen this exotic quasiparticle

you've got despair: the time the team was so sure they had the data and they really didn't. you've got not giving up after having to retract the paper and doubling down on getting every step of the research independently verified to be sure the data actually says what you hope it says

I think that's why there have been a whole bunch of slightly understated announcements over the last few years, because when I look at the 2022 paper on how Microsoft thinks quantum error correction should work, the actual structure of this new topological qubit is *right there* in the diagrams!

but this time it's not just a measurement, not just a paper: it's the actual design of the topological qubits and a physical chip that starts with 8 qubits in an architecture that can go to a million qubits - and the DARPA backup to build that prototype, maybe by DARPA's target of 2033

also, for this brief period of time I *understand* superconductors: electrons pal up into cooper pairs that share the same lowest energy quantum state and the energy gate to get to the next quantum energy state is too high so they stay there

because they are in the lowest quantum energy state they don't collide with the crystal matrix structure of the material, so there is no resistance when they move: that's why the current flows through superconductors without resistance (the super conducting bit)

superconducting qubits work because if there's an odd number of electrons, it takes more energy to move them because there's that one electron slouching along at the back with its hands in its pockets, scuffing its feet. measure which is spiffy fast electron flow and which is slouchier = 1s &0s

topological qubits use superconductors that make a topological phase of matter that hides quantum state so they also have the even/odd electron thing but it's stuffed in a locker so the teacher can't look out in the hall to see it until the class bell rings and the computation is finished

also, I couldn't put the animation in my piece but I love the hat tip to Eames' Powers of Ten and unlike a lot of video it actually helped me understand the chip architecture because those 8 H shapes on the first frame are the qubit: what you see in the window is just connectors to the nanowires

glad you liked it! I have had three goes at understanding and writing up topological qubits and *this* time I felt like I saw enough the research and physics to actually grasp how it works on the same way I can think about electrons flooding into a FinFET gate; just so many clever things in this

the physics of building this stuff is so freaking cool! sending a single electron down a 5 micrometer wire!

it's so freaking elegant! the same system does the qubit priming and the error correction and the measurement and the measurement is what *does the braiding* by transformation, and it's nice easy digital pulse not hand cranked demanding analog ones

and elegant here = accurate and easy to control and easy to scale and that also has to keep the complexity and cost down (the whole thing must be crazy expensive because of the fridges but anywhere you can reduce the complexity is good)

so digital computers are binary: the transistors store either 1 or 0 and you do computations with that. qubits are 1 and 0 and everything in between; it's a smeary kind of statistical information that you have to interpret. but they're very fast: one operation in a quantum computer takes 100ns

people are understandably cautious about Microsoft's claims about Majoranas especially as a researcher got way too excited about some eventually inconclusive results a few years back: Chetan Nayak confirming that the Arxiv paper (link in my piece) does show actual measurements of Majorana Zero Modes

I'm really hoping topological qubits are real; a qubit that lasts for on average 20 seconds or even a minute is massively stable compared to what else is out there. also a proof point for Microsoft Discovery science platform because they used it to get to Majorana 2 in just over a year