5 ms·

Simulating Physics

12 points by stefanpie 9y ago

selimthegrim 9y ago
I worked with the UCSB team on some of the thermalization calculations that didn't end up making it into the paper (I am at a different university). It's really cool to see how quickly phenomena we study in simulations came to the forefront of experiment. The actual paper can be found here - https://arxiv.org/abs/1709.06678 https://arxiv.org/abs/1709.06678
Koshkin 9y ago
In this case, it looks like the goal was to demonstrate the power of quantum computers rather than to use them in order gain a new insight into the physics of the otherwise well-known phenomenon (the so-called "Anderson localization", which is a "quantum-mechanical part" of the effect that impurities have on the conductivity of a metal). Speaking of the "non-goal", I am left wondering how much information is there to be gained from looking this closely (albeit by means of simulation) at things that have been or, if not, perhaps could be described well enough following the more traditional lines of statistical or phenomenological investigation.
- selimthegrim 9y ago
  Anderson localization and many-body localization are distinct. AL is a single-particle phenomenon, considering wavefunctions and eigenstates of a single electron at a time. MBL involves ensembles of manifestly many-electron/many-body eigenstates. If you take a subsystem (a small window of eigenstates or energy if you are in the microcanonical ensemble) and resample it, if you are in the thermal regime it will recover the entire distribution of the extensive quantity being sampled for the whole system (let's take entropy as an example). The goal is to eventually see how many-body localization scales (does the experimental appearance vanish at large but finite time scales as is common in implementations of the Bose-Hubbard model) and see if it can be harnessed as either a form of passive quantum information protection/error correction and/or useful in assisting speedup in solving certain adiabatic quantum computing problems (this last has nothing to do with the Google experiment at hand and more with the intersection of the problem with spin glasses)

12 points by stefanpie 9y ago

I worked with the UCSB team on some of the thermalization calculations that didn't end up making it into the paper (I am at a different university). It's really cool to see how quickly phenomena we study in simulations came to the forefront of experiment. The actual paper can be found here - https://arxiv.org/abs/1709.06678 https://arxiv.org/abs/1709.06678

Koshkin 9y ago

In this case, it looks like the goal was to demonstrate the power of quantum computers rather than to use them in order gain a new insight into the physics of the otherwise well-known phenomenon (the so-called "Anderson localization", which is a "quantum-mechanical part" of the effect that impurities have on the conductivity of a metal). Speaking of the "non-goal", I am left wondering how much information is there to be gained from looking this closely (albeit by means of simulation) at things that have been or, if not, perhaps could be described well enough following the more traditional lines of statistical or phenomenological investigation.

selimthegrim 9y ago
Anderson localization and many-body localization are distinct. AL is a single-particle phenomenon, considering wavefunctions and eigenstates of a single electron at a time. MBL involves ensembles of manifestly many-electron/many-body eigenstates. If you take a subsystem (a small window of eigenstates or energy if you are in the microcanonical ensemble) and resample it, if you are in the thermal regime it will recover the entire distribution of the extensive quantity being sampled for the whole system (let's take entropy as an example). The goal is to eventually see how many-body localization scales (does the experimental appearance vanish at large but finite time scales as is common in implementations of the Bose-Hubbard model) and see if it can be harnessed as either a form of passive quantum information protection/error correction and/or useful in assisting speedup in solving certain adiabatic quantum computing problems (this last has nothing to do with the Google experiment at hand and more with the intersection of the problem with spin glasses)