Entropy calculations for classical model using Quantum-Wang-Landau
Dear all, I am trying to use the Quantum-Wang-Landau (QWL) algorithm to calculate the entropy of a linear chain spin-1/2 model with Jxy=0 in a field h=1, which is classical but QWL seems to be suggested for entropy calculations. However with forty sites and down to low temperatures, I don't seem to be getting near the correct entropy density result of S ~ 0.4812 ... The parameter file is <<<<<<<<<<<<<<<<<<<<<<<<<< LATTICE="chain lattice" MODEL="spin" local_S=1/2 h=1 J = 1 Jxy = 0 CUTOFF = 500 {L=40} <<<<<<<<<<<<<<<<<<<<<<<<< and a partial output of entropy density versus temperature is <SNIP> </PARAMETERS> <legend show="true"/> <xaxis label="Temperature"/> <yaxis label="Entropy Density"/> <set label=""> <point><x>0.10000000000000001</x><y>0.065628773413218822</y></point> <point><x>0.20000000000000001</x><y>0.13394955300729872</y></point> <point><x>0.30000000000000004</x><y>0.24229259030929573</y></point> <point><x>0.40000000000000002</x><y>0.31481859757794478</y></point> <point><x>0.5</x><y>0.39826073000393336</y></point> <point><x>0.59999999999999998</x><y>0.46128739087861204</y></point> <point><x>0.69999999999999996</x><y>0.50938492873462538</y></point> <point><x>0.79999999999999993</x><y>0.5481023575910553</y></point> <point><x>0.89999999999999991</x><y>0.57776445729654646</y></point> <point><x>0.99999999999999989</x><y>0.59988758184662683</y></point> <point><x>1.0999999999999999</x><y>0.61635447545226674</y></point> <SNIP> I have tried changing the lattice length and cut-off but not much changes. Any advice on what I might be doing wrong will be greatly appreciated. With regards, Vipin
Dear Vipin, where do you get the value 0.4812 from? On Jul 26, 2013, at 4:47, Vipin Varma <varma@th.physik.uni-bonn.de> wrote:
Dear all,
I am trying to use the Quantum-Wang-Landau (QWL) algorithm to calculate the entropy of a linear chain spin-1/2 model with Jxy=0 in a field h=1, which is classical but QWL seems to be suggested for entropy calculations.
However with forty sites and down to low temperatures, I don't seem to be getting near the correct entropy density result of S ~ 0.4812 ...
The parameter file is <<<<<<<<<<<<<<<<<<<<<<<<<< LATTICE="chain lattice" MODEL="spin" local_S=1/2 h=1 J = 1 Jxy = 0 CUTOFF = 500 {L=40} <<<<<<<<<<<<<<<<<<<<<<<<<
and a partial output of entropy density versus temperature is
<SNIP> </PARAMETERS> <legend show="true"/> <xaxis label="Temperature"/> <yaxis label="Entropy Density"/> <set label=""> <point><x>0.10000000000000001</x><y>0.065628773413218822</y></point> <point><x>0.20000000000000001</x><y>0.13394955300729872</y></point> <point><x>0.30000000000000004</x><y>0.24229259030929573</y></point> <point><x>0.40000000000000002</x><y>0.31481859757794478</y></point> <point><x>0.5</x><y>0.39826073000393336</y></point> <point><x>0.59999999999999998</x><y>0.46128739087861204</y></point> <point><x>0.69999999999999996</x><y>0.50938492873462538</y></point> <point><x>0.79999999999999993</x><y>0.5481023575910553</y></point> <point><x>0.89999999999999991</x><y>0.57776445729654646</y></point> <point><x>0.99999999999999989</x><y>0.59988758184662683</y></point> <point><x>1.0999999999999999</x><y>0.61635447545226674</y></point> <SNIP>
I have tried changing the lattice length and cut-off but not much changes.
Any advice on what I might be doing wrong will be greatly appreciated.
With regards, Vipin
Dear Prof. Troyer, Thank you for the reply. I take this reference result from the paper by Jill C. Bonner and Michael E. Fisher: "The Entropy of an Antiferromagnet in a Magnetic Field", Proc. Phys. Soc. 80, 508 (1962) doi:10.1088/0370-1328/80/2/318 I do get the low temperature convergence towards the correct result from the exact diagonalization program in ALPS (with smaller lattice size L=8) but because I intend to test on other bigger 2D lattices (for which the ALPS ED crashes with std::bad_alloc()), I guessed that QWL might be a good option. Regards, Vipin On Fri, 26 Jul 2013, Matthias Troyer wrote:
Dear Vipin,
where do you get the value 0.4812 from?
On Jul 26, 2013, at 4:47, Vipin Varma <varma@th.physik.uni-bonn.de> wrote:
Dear all,
I am trying to use the Quantum-Wang-Landau (QWL) algorithm to calculate the entropy of a linear chain spin-1/2 model with Jxy=0 in a field h=1, which is classical but QWL seems to be suggested for entropy calculations.
However with forty sites and down to low temperatures, I don't seem to be getting near the correct entropy density result of S ~ 0.4812 ...
The parameter file is <<<<<<<<<<<<<<<<<<<<<<<<<< LATTICE="chain lattice" MODEL="spin" local_S=1/2 h=1 J = 1 Jxy = 0 CUTOFF = 500 {L=40} <<<<<<<<<<<<<<<<<<<<<<<<<
and a partial output of entropy density versus temperature is
<SNIP> </PARAMETERS> <legend show="true"/> <xaxis label="Temperature"/> <yaxis label="Entropy Density"/> <set label=""> <point><x>0.10000000000000001</x><y>0.065628773413218822</y></point> <point><x>0.20000000000000001</x><y>0.13394955300729872</y></point> <point><x>0.30000000000000004</x><y>0.24229259030929573</y></point> <point><x>0.40000000000000002</x><y>0.31481859757794478</y></point> <point><x>0.5</x><y>0.39826073000393336</y></point> <point><x>0.59999999999999998</x><y>0.46128739087861204</y></point> <point><x>0.69999999999999996</x><y>0.50938492873462538</y></point> <point><x>0.79999999999999993</x><y>0.5481023575910553</y></point> <point><x>0.89999999999999991</x><y>0.57776445729654646</y></point> <point><x>0.99999999999999989</x><y>0.59988758184662683</y></point> <point><x>1.0999999999999999</x><y>0.61635447545226674</y></point> <SNIP>
I have tried changing the lattice length and cut-off but not much changes.
Any advice on what I might be doing wrong will be greatly appreciated.
With regards, Vipin
As a quick check to see what might be going on, can you send me energy as a function of temperature from your results? On Jul 26, 2013, at 6:58 AM, Vipin Varma <varma@th.physik.uni-bonn.de> wrote:
On Fri, 26 Jul 2013, Matthias Troyer wrote:
Dear Vipin,
where do you get the value 0.4812 from?
On Jul 26, 2013, at 4:47, Vipin Varma <varma@th.physik.uni-bonn.de> wrote:
Dear all,
I am trying to use the Quantum-Wang-Landau (QWL) algorithm to calculate the entropy of a linear chain spin-1/2 model with Jxy=0 in a field h=1, which is classical but QWL seems to be suggested for entropy calculations.
However with forty sites and down to low temperatures, I don't seem to be getting near the correct entropy density result of S ~ 0.4812 ...
The parameter file is <<<<<<<<<<<<<<<<<<<<<<<<<< LATTICE="chain lattice" MODEL="spin" local_S=1/2 h=1 J = 1 Jxy = 0 CUTOFF = 500 {L=40} <<<<<<<<<<<<<<<<<<<<<<<<<
and a partial output of entropy density versus temperature is
<SNIP> </PARAMETERS> <legend show="true"/> <xaxis label="Temperature"/> <yaxis label="Entropy Density"/> <set label=""> <point><x>0.10000000000000001</x><y>0.065628773413218822</y></point> <point><x>0.20000000000000001</x><y>0.13394955300729872</y></point> <point><x>0.30000000000000004</x><y>0.24229259030929573</y></point> <point><x>0.40000000000000002</x><y>0.31481859757794478</y></point> <point><x>0.5</x><y>0.39826073000393336</y></point> <point><x>0.59999999999999998</x><y>0.46128739087861204</y></point> <point><x>0.69999999999999996</x><y>0.50938492873462538</y></point> <point><x>0.79999999999999993</x><y>0.5481023575910553</y></point> <point><x>0.89999999999999991</x><y>0.57776445729654646</y></point> <point><x>0.99999999999999989</x><y>0.59988758184662683</y></point> <point><x>1.0999999999999999</x><y>0.61635447545226674</y></point> <SNIP>
I have tried changing the lattice length and cut-off but not much changes.
Any advice on what I might be doing wrong will be greatly appreciated.
With regards, Vipin
Hi Vipin, If you look at the documentation you will see that the QWL code is only for Heisenberg models while you try to simulate an Ising model in a field. Best regards Matthias On Jul 26, 2013, at 6:58 AM, Vipin Varma <varma@th.physik.uni-bonn.de> wrote:
Dear Prof. Troyer,
Thank you for the reply.
I take this reference result from the paper by Jill C. Bonner and Michael E. Fisher: "The Entropy of an Antiferromagnet in a Magnetic Field", Proc. Phys. Soc. 80, 508 (1962) doi:10.1088/0370-1328/80/2/318
I do get the low temperature convergence towards the correct result from the exact diagonalization program in ALPS (with smaller lattice size L=8) but because I intend to test on other bigger 2D lattices (for which the ALPS ED crashes with std::bad_alloc()), I guessed that QWL might be a good option.
Regards, Vipin
On Fri, 26 Jul 2013, Matthias Troyer wrote:
Dear Vipin,
where do you get the value 0.4812 from?
On Jul 26, 2013, at 4:47, Vipin Varma <varma@th.physik.uni-bonn.de> wrote:
Dear all,
I am trying to use the Quantum-Wang-Landau (QWL) algorithm to calculate the entropy of a linear chain spin-1/2 model with Jxy=0 in a field h=1, which is classical but QWL seems to be suggested for entropy calculations.
However with forty sites and down to low temperatures, I don't seem to be getting near the correct entropy density result of S ~ 0.4812 ...
The parameter file is <<<<<<<<<<<<<<<<<<<<<<<<<< LATTICE="chain lattice" MODEL="spin" local_S=1/2 h=1 J = 1 Jxy = 0 CUTOFF = 500 {L=40} <<<<<<<<<<<<<<<<<<<<<<<<<
and a partial output of entropy density versus temperature is
<SNIP> </PARAMETERS> <legend show="true"/> <xaxis label="Temperature"/> <yaxis label="Entropy Density"/> <set label=""> <point><x>0.10000000000000001</x><y>0.065628773413218822</y></point> <point><x>0.20000000000000001</x><y>0.13394955300729872</y></point> <point><x>0.30000000000000004</x><y>0.24229259030929573</y></point> <point><x>0.40000000000000002</x><y>0.31481859757794478</y></point> <point><x>0.5</x><y>0.39826073000393336</y></point> <point><x>0.59999999999999998</x><y>0.46128739087861204</y></point> <point><x>0.69999999999999996</x><y>0.50938492873462538</y></point> <point><x>0.79999999999999993</x><y>0.5481023575910553</y></point> <point><x>0.89999999999999991</x><y>0.57776445729654646</y></point> <point><x>0.99999999999999989</x><y>0.59988758184662683</y></point> <point><x>1.0999999999999999</x><y>0.61635447545226674</y></point> <SNIP>
I have tried changing the lattice length and cut-off but not much changes.
Any advice on what I might be doing wrong will be greatly appreciated.
With regards, Vipin
Dear all, I am trying to simulate the hard-core boson model on a square lattice using the ALPS's (alps-2.0.2-r5790-src-with-boost) dirloop_sse as follows:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> LATTICE="square lattice" L=12
MODEL="hardcore boson" V = 1 T = 0.05 t = 1 SWEEPS=500000 THERMALIZATION=20000 { mu=-4; } { mu=-3.5; } { mu=-3; } { mu=-2.5; } { mu=-2; } { mu=-1.5; } { mu=-1; } { mu=-0.5; } { mu=0; } { mu=0.1; } { mu=0.2; } { mu=0.3; } { mu=0.4; }
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
However, I get almost the same problem as was reported by another user on O2 May 2012, at 08:20 (no subject): <SNIP> Created run 1 locally Starting task 9. Checking if it is finished: not yet, next check in 10 seconds ( 0% done). Checking if it is finished: not yet, next check in 10 seconds ( 1% done). Checking if it is finished: not yet, next check in 180 seconds ( 3% done). NUMBER_OF_WORMS_PER_SWEEP: 10 Checking if it is finished: not yet, next check in 143 seconds ( 26% done). Internal error: size of prob table is zero. <SNIP> I mention that I did use the same ALPS version some time back for a very similar run but faced no issues then. Moreover, the hopping t is not zero in my case, as was seemingly the problem for the other user. I have tried adjusting the thermalization and sweeps a little, but to no effect unfortunately. Any advice on what I am doing wrong is greatly appreciated. Regards, Vipin
Hi, can you try the nightly snapshot release? Matthias On Jul 30, 2013, at 10:02 AM, Vipin Varma <varma@th.physik.uni-bonn.de> wrote:
Dear all,
I am trying to simulate the hard-core boson model on a square lattice using the ALPS's (alps-2.0.2-r5790-src-with-boost) dirloop_sse as follows:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> LATTICE="square lattice" L=12
MODEL="hardcore boson" V = 1 T = 0.05 t = 1
SWEEPS=500000 THERMALIZATION=20000
{ mu=-4; } { mu=-3.5; } { mu=-3; } { mu=-2.5; } { mu=-2; } { mu=-1.5; } { mu=-1; } { mu=-0.5; } { mu=0; } { mu=0.1; } { mu=0.2; } { mu=0.3; } { mu=0.4; }
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
However, I get almost the same problem as was reported by another user on O2 May 2012, at 08:20 (no subject):
<SNIP>
Created run 1 locally Starting task 9. Checking if it is finished: not yet, next check in 10 seconds ( 0% done). Checking if it is finished: not yet, next check in 10 seconds ( 1% done). Checking if it is finished: not yet, next check in 180 seconds ( 3% done). NUMBER_OF_WORMS_PER_SWEEP: 10 Checking if it is finished: not yet, next check in 143 seconds ( 26% done).
Internal error: size of prob table is zero.
<SNIP>
I mention that I did use the same ALPS version some time back for a very similar run but faced no issues then. Moreover, the hopping t is not zero in my case, as was seemingly the problem for the other user. I have tried adjusting the thermalization and sweeps a little, but to no effect unfortunately.
Any advice on what I am doing wrong is greatly appreciated.
Regards, Vipin
Dear Prof. Troyer, Dr. Isakov suggested I include EPSILON=0.01 in the parameter file and that fixed the issue. But I will also try the release you suggest ... Thanks, Vipin On Thu, 1 Aug 2013, Matthias Troyer wrote:
Hi,
can you try the nightly snapshot release?
Matthias
On Jul 30, 2013, at 10:02 AM, Vipin Varma <varma@th.physik.uni-bonn.de> wrote:
Dear all,
I am trying to simulate the hard-core boson model on a square lattice using the ALPS's (alps-2.0.2-r5790-src-with-boost) dirloop_sse as follows:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> LATTICE="square lattice" L=12
MODEL="hardcore boson" V = 1 T = 0.05 t = 1
SWEEPS=500000 THERMALIZATION=20000
{ mu=-4; } { mu=-3.5; } { mu=-3; } { mu=-2.5; } { mu=-2; } { mu=-1.5; } { mu=-1; } { mu=-0.5; } { mu=0; } { mu=0.1; } { mu=0.2; } { mu=0.3; } { mu=0.4; }
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
However, I get almost the same problem as was reported by another user on O2 May 2012, at 08:20 (no subject):
<SNIP>
Created run 1 locally Starting task 9. Checking if it is finished: not yet, next check in 10 seconds ( 0% done). Checking if it is finished: not yet, next check in 10 seconds ( 1% done). Checking if it is finished: not yet, next check in 180 seconds ( 3% done). NUMBER_OF_WORMS_PER_SWEEP: 10 Checking if it is finished: not yet, next check in 143 seconds ( 26% done).
Internal error: size of prob table is zero.
<SNIP>
I mention that I did use the same ALPS version some time back for a very similar run but faced no issues then. Moreover, the hopping t is not zero in my case, as was seemingly the problem for the other user. I have tried adjusting the thermalization and sweeps a little, but to no effect unfortunately.
Any advice on what I am doing wrong is greatly appreciated.
Regards, Vipin
participants (2)
-
Matthias Troyer
-
Vipin Varma