Hacker News Reader

Comment by dmitrybrant | Hacker News Reader

I assume this is already being done, but I've had the following idea kicking around for a while:

Create a simulation, down to the quantum states, of a lattice of molecules at a certain simulated temperature (say, room temperature), and induce a simulated current through the lattice, and see if it superconducts. Proceed by iterating through billions of permutations of compounds in the simulated lattice, until the simulation finds a room-temperature superconductor.

Assuming this is feasible, does anyone know of organizations that are doing this?

View on HN

Not for superconductivity specifically, but for a broad range of properties of crystals, this is what the Materials Project[0] does.

Materials Project is funded by the US Department of Energy and uses supercomputing to simulate hundreds of thousands of different crystal structures on the quantum mechanical level to try and find those which have useful properties for practical applications.

This line of research is broadly called “materials discovery”, “materials design” (often “high throughput”) or even “materials genomics” depending on who you ask. These terms are provided in case anyone wants to search and read more about it.

[0] https://materialsproject.org

The computational complexity of brute-force simulating many-body quantum systems scales exponentially in the number of particles. There is no supercomputer on earth that can simulate a realistic solid.

And you can’t get out of this problem by just simulating e.g. two or three particles. That small of a system can’t simulate things like temperature in a meaningful way.

This made me think about this nice little TV show: https://www.imdb.com/title/tt8134186/

It is a very active domain, with lots and lots of techniques to do some clever approximations. Brute forcing is not feasible, even for the simplest systems.

It is anyway a lot easier to simulate a solid that has no defects. But many properties found that way turn up in defect-ridden, real materials too.

A truly periodic substrate lets you wrap it in a closed universe just big enough for what you hope to make live in it.

You can probably find a few condensed matter groups that are doing computational simulations, but I think full quantum mechanical simulations on that scale are computationally intractable so they'll have to use approximations of some sort. This is why quantum computers are predicted to exceed the computational power of classical computers.

Besides it's been fairly recent that simulations for classical thermodynamics with a decent number of particles became somewhat feasible. And even then it often uses lots of approximations as well as statistical tricks to get decent results for sizeable systems. For quantum mechanics quite a few tricks to explore the state-space stop working because the state-space is unimaginably huge, so I'm not even sure how to begin to do similar simulations, but I reckon they'll be disastrously slow for all but the simplest of systems.

The level of detail you need for a simulation that allows you to see cooper pairs come into existence is downright insane, you might be better of trying to predict next years weather.

That is essentially the idea of Random Structure Searching [0], but the precision of quantum simulation (better known as ab initio) for solids is not perfect and very time consuming.

[0] https://iopscience.iop.org/article/10.1088/0953-8984/23/5/05...

I don't know the state of the art but I think that we can't even predict the static crystal structure of simple substances---e.g. when iron is BCC vs FCC. The first-principles simulation of dynamic properties of large quantum many body systems is just not feasible. It is possible that this may change dramatically when quantum computers arrive---currently we describe quantum systems by modeling them with discrete, classical computers, and quantum computers might turn out to model the relevant quantum processes directly.

As others have pointed out, the computations are currently infeasable. The rest of the plan works though, it would just need to be done with actual atoms.

Reply on news.ycombinator.com