I'm very interested in this class of approximations on consumer hardware from a different perspective: Simulating (simple) biological systems and organic molecules. Background: I'm building out the rust structural bio OSS ecosystem, and a GUI/3D CAD-style visualizer. It has molecular dynamics integrated tightly, and is designed to "just work" without fuss, setup, or errors.
The problem: It is reliant on atom-centered partial charge, and pre-calculated parameters for its Newtonian forces. These are available for a set of 30k or so organic molecules, and most protein, lipid, and nucleic acid configurations. The problems: Making it work for arbitrary systems, and doing better than these specifically-tuned models. I want the general case, and it to work in a no-fuss way.
I am now looking into the TWA, and am interested in this class of approximation in general. I have the UI and traditional MD system mostly ready; getting ready to tackle this problem. This is a much simpler setup than the ones physicists are interested, and that the article covers, since we can focus on stable covalent-bonded systems of atoms.
That's Daedalus is it? Went to look up how your project is doing, and had to go through github stars - might be worth putting a link to it in your HN profile? Assuming you don't mind making that link explicit :)
Yep! it's a grind. Good call on the profile!
Reading their paper, it does seem like this method is significantly simpler than using something like MPS, my main concern is the practical coupling regime for which this method works, I would imagine that it would fail closer to critical points in theories with phase transitions?
This is a nice practical technique for open quantum systems with relatively low entanglement. The introduction lays out exactly what regime they're aiming at:
1. Affordable (laptop scale)
2. Captures "sufficient" quantum effects (low entanglement regime; you accurately can't simulate a quantum computer with this)
3. Straightforward to implement.
From a cursory glance, it does all three. I'm slightly surprised that TWA hasn't been applied to open systems extensively before, but it was always a relatively obscure technique. I'm guessing this should be quite useful in practice for e.g. AMO and cavity systems with relatively large dissipation terms that prevent entanglement build up. However, I'd guess this wouldn't do very well near phase transitions. All-in-all, a nice new technique for a regime that didn't have too many options.
Yeah this seems like a very useful technique for ground state properties, I’m also surprised in retrospect, having never heard of TWA before now.
I'm curious, are any quantum simulators accessible to sort of hobbyists?
I have a background in solid state physics from coursework, but I've never really used QM for almost anything outside of school.
I think you will have best luck by searching for "open quantum systems" toolboxes in your language of choice. My preferences are, in order:
- QuantumOptics.jl in Julia
- QuantumToolbox.jl in Julia
- qutip in python
These are all "just" nice domain specific wrappers around linear algebra and differential equation tools. They do the "silly" exponentially expensive simulation technique that works for any quantum system. If you are interested in efficient (not exponential) simulation techniques that support only a subset of all quantum dynamics try out:
- stabilizer formalism (e.g. for error correction) with QuantumClifford.jl or stim
- Gaussian quantum optics (e.g. for laser physics) with Gabs.jl
- tensor networks (e.g. for arbitrary low-rank entanglement) with ITensors.jl
If you prefer python, tenpy is quite nice for tensor networks as well.
Thanks! I am kind of stuck in Python, or at least stuck outside of Julia.
Try ORCA and GROMACS. Accessible as you can DL them and there are lots of docs. But very high learning curve, and not graphical on their own.
With _quantum_ phase transitions, yes.