Hello all, I'm trying to simulate a spin glass using ALPS, and I'd like to look at the time-dependent correlation function. I can't seem to find a way to pull this out simply; is there some way to make one of the QMC codes produce this as output? Failing that, is there a way I can make one of the codes (e.g. dirloop_sse) output the state of the spins after a simulation so I can code it up myself using a temperature/imaginary time transformation?
Thanks, -Elliot Snow-Kropla
Hi,
If you're interested in real time dynamics then this will not help, since it will only give you correlation functions in "Monte Carlo time". Unlike classical systems, where local updates mimic the real time evolution, in QMC this is all just an artifact of the QMC updates used and not related to any real dynamics.
Matthias Troyer
On 26 Apr 2012, at 20:03, Elliot Snow-Kropla wrote:
Hello all, I'm trying to simulate a spin glass using ALPS, and I'd like to look at the time-dependent correlation function. I can't seem to find a way to pull this out simply; is there some way to make one of the QMC codes produce this as output? Failing that, is there a way I can make one of the codes (e.g. dirloop_sse) output the state of the spins after a simulation so I can code it up myself using a temperature/imaginary time transformation?
Thanks, -Elliot Snow-Kropla
I'm not actually interested in real time dynamics; what I'm interested in is in the long time behaviour of the correlation function, even if it's in imaginary time. On Apr 26, 2012 3:06 PM, "Matthias Troyer" troyer@phys.ethz.ch wrote:
Hi,
If you're interested in real time dynamics then this will not help, since it will only give you correlation functions in "Monte Carlo time". Unlike classical systems, where local updates mimic the real time evolution, in QMC this is all just an artifact of the QMC updates used and not related to any real dynamics.
Matthias Troyer
On 26 Apr 2012, at 20:03, Elliot Snow-Kropla wrote:
Hello all, I'm trying to simulate a spin glass using ALPS, and I'd like to look at
the time-dependent correlation function. I can't seem to find a way to pull this out simply; is there some way to make one of the QMC codes produce this as output? Failing that, is there a way I can make one of the codes (e.g. dirloop_sse) output the state of the spins after a simulation so I can code it up myself using a temperature/imaginary time transformation?
Thanks, -Elliot Snow-Kropla
You will need to go into the code and implement the measurements yourself directly. However, I don't think that you will be able to obtain the long-time behavior of the correlation function that way. We might be able to help more of you explain what exactly you want to observe.
On 26 Apr 2012, at 20:09, Elliot Snow-Kropla wrote:
I'm not actually interested in real time dynamics; what I'm interested in is in the long time behaviour of the correlation function, even if it's in imaginary time.
On Apr 26, 2012 3:06 PM, "Matthias Troyer" troyer@phys.ethz.ch wrote: Hi,
If you're interested in real time dynamics then this will not help, since it will only give you correlation functions in "Monte Carlo time". Unlike classical systems, where local updates mimic the real time evolution, in QMC this is all just an artifact of the QMC updates used and not related to any real dynamics.
Matthias Troyer
On 26 Apr 2012, at 20:03, Elliot Snow-Kropla wrote:
Hello all, I'm trying to simulate a spin glass using ALPS, and I'd like to look at the time-dependent correlation function. I can't seem to find a way to pull this out simply; is there some way to make one of the QMC codes produce this as output? Failing that, is there a way I can make one of the codes (e.g. dirloop_sse) output the state of the spins after a simulation so I can code it up myself using a temperature/imaginary time transformation?
Thanks, -Elliot Snow-Kropla
That's what I thought. I'm trying to look at location of phase transitions similar to http://arxiv.org/abs/0910.1378, so looking at how the energy gap between the first two states changes as the hamiltonian changes.
On Thu, Apr 26, 2012 at 3:12 PM, Matthias Troyer troyer@phys.ethz.chwrote:
You will need to go into the code and implement the measurements yourself directly. However, I don't think that you will be able to obtain the long-time behavior of the correlation function that way. We might be able to help more of you explain what exactly you want to observe.
On 26 Apr 2012, at 20:09, Elliot Snow-Kropla wrote:
I'm not actually interested in real time dynamics; what I'm interested in is in the long time behaviour of the correlation function, even if it's in imaginary time. On Apr 26, 2012 3:06 PM, "Matthias Troyer" troyer@phys.ethz.ch wrote:
Hi,
If you're interested in real time dynamics then this will not help, since it will only give you correlation functions in "Monte Carlo time". Unlike classical systems, where local updates mimic the real time evolution, in QMC this is all just an artifact of the QMC updates used and not related to any real dynamics.
Matthias Troyer
On 26 Apr 2012, at 20:03, Elliot Snow-Kropla wrote:
Hello all, I'm trying to simulate a spin glass using ALPS, and I'd like to look at
the time-dependent correlation function. I can't seem to find a way to pull this out simply; is there some way to make one of the QMC codes produce this as output? Failing that, is there a way I can make one of the codes (e.g. dirloop_sse) output the state of the spins after a simulation so I can code it up myself using a temperature/imaginary time transformation?
Thanks, -Elliot Snow-Kropla
OK, so what you want to know is the gap. That can indeed be obtained. from the imaginary time correlation function but you should be careful to measure the right gap. Let me discuss with Sergei Isakov how much work it would be to add that functionality to the code.
Matthias
On 26 Apr 2012, at 20:15, Elliot Snow-Kropla wrote:
That's what I thought. I'm trying to look at location of phase transitions similar to http://arxiv.org/abs/0910.1378, so looking at how the energy gap between the first two states changes as the hamiltonian changes.
On Thu, Apr 26, 2012 at 3:12 PM, Matthias Troyer troyer@phys.ethz.ch wrote: You will need to go into the code and implement the measurements yourself directly. However, I don't think that you will be able to obtain the long-time behavior of the correlation function that way. We might be able to help more of you explain what exactly you want to observe.
On 26 Apr 2012, at 20:09, Elliot Snow-Kropla wrote:
I'm not actually interested in real time dynamics; what I'm interested in is in the long time behaviour of the correlation function, even if it's in imaginary time.
On Apr 26, 2012 3:06 PM, "Matthias Troyer" troyer@phys.ethz.ch wrote: Hi,
If you're interested in real time dynamics then this will not help, since it will only give you correlation functions in "Monte Carlo time". Unlike classical systems, where local updates mimic the real time evolution, in QMC this is all just an artifact of the QMC updates used and not related to any real dynamics.
Matthias Troyer
On 26 Apr 2012, at 20:03, Elliot Snow-Kropla wrote:
Hello all, I'm trying to simulate a spin glass using ALPS, and I'd like to look at the time-dependent correlation function. I can't seem to find a way to pull this out simply; is there some way to make one of the QMC codes produce this as output? Failing that, is there a way I can make one of the codes (e.g. dirloop_sse) output the state of the spins after a simulation so I can code it up myself using a temperature/imaginary time transformation?
Thanks, -Elliot Snow-Kropla
Thanks. Being able to do it (more or less) simply with ALPS saves me a lot of time (e.g. implementing MPI myself etc.)
On Thu, Apr 26, 2012 at 3:24 PM, Matthias Troyer troyer@phys.ethz.chwrote:
OK, so what you want to know is the gap. That can indeed be obtained. from the imaginary time correlation function but you should be careful to measure the right gap. Let me discuss with Sergei Isakov how much work it would be to add that functionality to the code.
Matthias
On 26 Apr 2012, at 20:15, Elliot Snow-Kropla wrote:
That's what I thought. I'm trying to look at location of phase transitions similar to http://arxiv.org/abs/0910.1378, so looking at how the energy gap between the first two states changes as the hamiltonian changes.
On Thu, Apr 26, 2012 at 3:12 PM, Matthias Troyer troyer@phys.ethz.chwrote:
You will need to go into the code and implement the measurements yourself directly. However, I don't think that you will be able to obtain the long-time behavior of the correlation function that way. We might be able to help more of you explain what exactly you want to observe.
On 26 Apr 2012, at 20:09, Elliot Snow-Kropla wrote:
I'm not actually interested in real time dynamics; what I'm interested in is in the long time behaviour of the correlation function, even if it's in imaginary time. On Apr 26, 2012 3:06 PM, "Matthias Troyer" troyer@phys.ethz.ch wrote:
Hi,
If you're interested in real time dynamics then this will not help, since it will only give you correlation functions in "Monte Carlo time". Unlike classical systems, where local updates mimic the real time evolution, in QMC this is all just an artifact of the QMC updates used and not related to any real dynamics.
Matthias Troyer
On 26 Apr 2012, at 20:03, Elliot Snow-Kropla wrote:
Hello all, I'm trying to simulate a spin glass using ALPS, and I'd like to look
at the time-dependent correlation function. I can't seem to find a way to pull this out simply; is there some way to make one of the QMC codes produce this as output? Failing that, is there a way I can make one of the codes (e.g. dirloop_sse) output the state of the spins after a simulation so I can code it up myself using a temperature/imaginary time transformation?
Thanks, -Elliot Snow-Kropla
comp-phys-alps-users@lists.phys.ethz.ch