SoylentNews is people
SoylentNews
Scientists have created a transistor made up of a single molecule. Surrounded by just 12 atoms, it is likely to be the smallest possible size for a transistor – and the hard limit for Moore's law.
The transistor is made of a single molecule of phthalocyanine surrounded by ring of 12 positively charged indium atoms placed on an indium arsenide crystal, as revealed in the scientific journal Nature Physics.
The work proves that precise control of atoms to create a transistor smaller than any other quantum system available is possible and opens the door to further research into harnessing these tiny transistors for computers and systems with orders of magnitude more processing power than today's machines.
From The Guardian
The original article from Nature
(Score: 5, Interesting) by fritsd on Thursday July 23 2015, @04:28PM
The other day, in the article Why Aren't Supercomputers Getting Faster Like They Used To? [soylentnews.org], some posters wondered what the purpose is of faster and faster supercomputers.
I've never actually had calculation time on supers, only on mainframes, so maybe someone else can elaborate better, but:
The picture in the Guardian article of the phthalocyanine surrounded by Indium atoms shows the phthalocyanine as a blur. The picture is probably made with a scanning tunneling microscope, or an atomic force microscope (same thing I believe).
If you'd magically make an even stronger microscope, you wouldn't resolve that blurry picture. The blur is what it *is*. You're looking at the electron orbitals of the molecule, smeared out in true quantum fashion.
Now, why did the researchers choose phthalocyanine to build this transistor? There are myriads of molecules they could have chosen, if they chose randomly.
Phthalocyanine [wikipedia.org] is a dye, which means it changes electronic orbital structure at the drop of a hat (or more specificcally the absorption of a photon of visible light).
The molecule consists of 4 indole building blocks, each of which is an "aromatic compound" which doesn't necessarily mean that it smells, but that its electrons are smeared out nicely over the whole structure.
All this means that the molecule's behaviour at the quantum level, i.e. how its electrons behave when charge is applied to one of its various corners and orientations, is complicated and interesting and unpredictable.
So, you need a *really fucking huge* supercomputer, to run so-called ab-initio calculations [wikipedia.org], crude numerical approximations of the underlying Schrödinger equation, on this molecule.
A long time ago, in the quantum chemistry world, if you did calculations on smaller non-aromatic compounds like acetone, you needed permission of the mainframe operator, so they could allocate time and a special big scratch disk, just for your months-long calculation. Acetone has four non-hydrogen atoms.
Phthalocyanine is fourty non-hydrogen atoms, ten times as big.
Some of the more advanced and accurate ab-initio methods scaled as O(n^5) or O(n^6) if I remember correctly. Then, you're talking about 100,000 to a million times larger calculations than of small molecules.
To me, it is fascinating to imagine that maybe one day, faster supercomputers or other high-tech innovations are possible as a result of a spin-off of this current research, which probably only could be planned because the researchers had done this kind of large calculation on a current-generation supercomputer and decided from their simulation results: "we'll try phthalocyanin first, if it doesn't work after 5 years then we'll hire another PhD and choose another molecule".
Now it's time for my daily political rant:
I know people who worked in the quantum chemistry field. Every single one of them was a lot smarter than me. We're talking about "the 1 %" of chemists here. If you value your country, make sure those kind of people can always get grants to study and live for cheap, and that PhD students get paid a reasonable wage. "If you think education is expensive, try ignorance". Otherwise no shiny toys for you in 20-30 years time :-)
(Score: 5, Interesting) by opinionated_science on Thursday July 23 2015, @06:25PM
I do run on supercomputers. The overriding problem with supercomputer progress is that communication latency has not improved in at least 10 years (~ 1us). This doesn't hurt DFT calculations so much, as they are very dense and pegged to LINPACK which gets optimization. The latency does limit the biophysical calculations that need to solve multi-dimensional FFT's reflecting physical charge distribution, as these calculations are iterative.
A case-study of the gap behind "general purpose" supercomputing, and addressing a specific problem look for "Anton D.E.Shaw" , which is a machine built in 2007 for molecular dynamics calculations. It remains 2 orders of magnitude faster than any supercomputer on the top500.
We have only scratched the surface on what is achievable by computational science. Further advances in experimentation are going to require greater use of computational models to screen for better molecules.
(Score: 3, Interesting) by GeriatricGentleman on Friday July 24 2015, @02:09AM
thank you
Nice article choice and some comments that make me feel dumber (why don't I know stuff like this?) and smarter (am grateful for the chance to learn a little!)
This is the kind of thing I show my 10yr old when she asks what I am doing when I goof off on the computer.
Now, as usual in this situation, I need some time to dive down the internet rabbit hole chasing links - starting with some digestible info on quantum chemistry...as you were