Dear all, I am performing some tests using spinmc (alps 2.0, Heisenberg model) on uncoupled spins changing the number and magnitude of spins at different fields. There are some things in the results that I dont understand: The energy is always zero, independently of the magnitude of the field for the uncoupled case, if I put a coupling constant, then I get as expected a non-zero value for the energy, but it is still non-dependent on field as it should be because of the Zeeman term. In a previous post, it is said that the susceptibility is calculated from magnetization^2, so it should work while M is linear on H, is there a way to get the dependence of susceptibility against field as the system approaches to saturation? And regarding the direction of field, setting H=1 means: Hz=1 Hx, Hy and Hz=1 Hx^2+Hy^2+Hz^2=1^2 , Hx=Hy=Hz or something else? I also tried to set H="0 0 1" and the program runs, but i dont know what I am actually getting. Any comments are welcome Kind Regards Daniel
Can you please send your input file? Do you use local or cluster updates? On Feb 23, 2012, at 3:24 PM, daniel.aravena@antares.qi.ub.edu wrote:
Dear all,
I am performing some tests using spinmc (alps 2.0, Heisenberg model) on uncoupled spins changing the number and magnitude of spins at different fields. There are some things in the results that I dont understand:
The energy is always zero, independently of the magnitude of the field for the uncoupled case, if I put a coupling constant, then I get as expected a non-zero value for the energy, but it is still non-dependent on field as it should be because of the Zeeman term.
In a previous post, it is said that the susceptibility is calculated from magnetization^2, so it should work while M is linear on H, is there a way to get the dependence of susceptibility against field as the system approaches to saturation?
And regarding the direction of field, setting H=1 means: Hz=1 Hx, Hy and Hz=1 Hx^2+Hy^2+Hz^2=1^2 , Hx=Hy=Hz or something else?
I also tried to set H="0 0 1" and the program runs, but i dont know what I am actually getting.
Any comments are welcome
Kind Regards Daniel
One example of input file is: LATTICE_LIBRARY=/home/g1daniel/ALPS_LATT/latt02.xml MODEL="Heisenberg" LATTICE="4_graph" UPDATE="local" CONVENTION="quantum" THERMALIZATION=1000000 SWEEPS=10000000 S0=1 J0=0 g=2.0 T=1.0 {H=0} {H=1} {H=10} {H=100} {H=1000} {H=10000} {H=100000} {H=1000000} {H=10000000} {H=100000000} {H=1000000000} {H=10000000000} the "4_graph" lattice is: <GRAPH name="4_graph" vertices="4"> <VERTEX id="1" type="0"></VERTEX> <VERTEX id="2" type="0"></VERTEX> <VERTEX id="3" type="0"></VERTEX> <VERTEX id="4" type="0"></VERTEX> <EDGE type="0" source="1" target="2"/> <EDGE type="0" source="1" target="3"/> <EDGE type="0" source="1" target="4"/> </GRAPH> Thanks for your answer. Kind regards Daniel On Wed, 29 Feb 2012 09:28:01 +0100, Matthias Troyer wrote:
Can you please send your input file? Do you use local or cluster updates?
On Feb 23, 2012, at 3:24 PM, daniel.aravena@antares.qi.ub.edu wrote:
Dear all,
I am performing some tests using spinmc (alps 2.0, Heisenberg model) on uncoupled spins changing the number and magnitude of spins at different fields. There are some things in the results that I dont understand:
The energy is always zero, independently of the magnitude of the field for the uncoupled case, if I put a coupling constant, then I get as expected a non-zero value for the energy, but it is still non-dependent on field as it should be because of the Zeeman term.
In a previous post, it is said that the susceptibility is calculated from magnetization^2, so it should work while M is linear on H, is there a way to get the dependence of susceptibility against field as the system approaches to saturation?
And regarding the direction of field, setting H=1 means: Hz=1 Hx, Hy and Hz=1 Hx^2+Hy^2+Hz^2=1^2 , Hx=Hy=Hz or something else?
I also tried to set H="0 0 1" and the program runs, but i dont know what I am actually getting.
Any comments are welcome
Kind Regards Daniel
Use h and not H for the magnetic field - does that change things? On 29 Feb 2012, at 21:42, daniel.aravena wrote:
One example of input file is:
LATTICE_LIBRARY=/home/g1daniel/ALPS_LATT/latt02.xml MODEL="Heisenberg" LATTICE="4_graph" UPDATE="local" CONVENTION="quantum" THERMALIZATION=1000000 SWEEPS=10000000 S0=1 J0=0 g=2.0 T=1.0 {H=0} {H=1} {H=10} {H=100} {H=1000} {H=10000} {H=100000} {H=1000000} {H=10000000} {H=100000000} {H=1000000000} {H=10000000000}
the "4_graph" lattice is:
<GRAPH name="4_graph" vertices="4"> <VERTEX id="1" type="0"></VERTEX> <VERTEX id="2" type="0"></VERTEX> <VERTEX id="3" type="0"></VERTEX> <VERTEX id="4" type="0"></VERTEX> <EDGE type="0" source="1" target="2"/> <EDGE type="0" source="1" target="3"/> <EDGE type="0" source="1" target="4"/> </GRAPH>
Thanks for your answer.
Kind regards Daniel
On Wed, 29 Feb 2012 09:28:01 +0100, Matthias Troyer wrote:
Can you please send your input file? Do you use local or cluster updates?
On Feb 23, 2012, at 3:24 PM, daniel.aravena@antares.qi.ub.edu wrote:
Dear all,
I am performing some tests using spinmc (alps 2.0, Heisenberg model) on uncoupled spins changing the number and magnitude of spins at different fields. There are some things in the results that I dont understand:
The energy is always zero, independently of the magnitude of the field for the uncoupled case, if I put a coupling constant, then I get as expected a non-zero value for the energy, but it is still non-dependent on field as it should be because of the Zeeman term.
In a previous post, it is said that the susceptibility is calculated from magnetization^2, so it should work while M is linear on H, is there a way to get the dependence of susceptibility against field as the system approaches to saturation?
And regarding the direction of field, setting H=1 means: Hz=1 Hx, Hy and Hz=1 Hx^2+Hy^2+Hz^2=1^2 , Hx=Hy=Hz or something else?
I also tried to set H="0 0 1" and the program runs, but i dont know what I am actually getting.
Any comments are welcome
Kind Regards Daniel
Yes, now it works, thanks! Kind regards Daniel
Use h and not H for the magnetic field - does that change things?
On 29 Feb 2012, at 21:42, daniel.aravena wrote:
One example of input file is:
LATTICE_LIBRARY=/home/g1daniel/ALPS_LATT/latt02.xml MODEL="Heisenberg" LATTICE="4_graph" UPDATE="local" CONVENTION="quantum" THERMALIZATION=1000000 SWEEPS=10000000 S0=1 J0=0 g=2.0 T=1.0 {H=0} {H=1} {H=10} {H=100} {H=1000} {H=10000} {H=100000} {H=1000000} {H=10000000} {H=100000000} {H=1000000000} {H=10000000000}
the "4_graph" lattice is:
<GRAPH name="4_graph" vertices="4"> <VERTEX id="1" type="0"></VERTEX> <VERTEX id="2" type="0"></VERTEX> <VERTEX id="3" type="0"></VERTEX> <VERTEX id="4" type="0"></VERTEX> <EDGE type="0" source="1" target="2"/> <EDGE type="0" source="1" target="3"/> <EDGE type="0" source="1" target="4"/> </GRAPH>
Thanks for your answer.
Kind regards Daniel
On Wed, 29 Feb 2012 09:28:01 +0100, Matthias Troyer wrote:
Can you please send your input file? Do you use local or cluster updates?
On Feb 23, 2012, at 3:24 PM, daniel.aravena@antares.qi.ub.edu wrote:
Dear all,
I am performing some tests using spinmc (alps 2.0, Heisenberg model) on uncoupled spins changing the number and magnitude of spins at different fields. There are some things in the results that I dont understand:
The energy is always zero, independently of the magnitude of the field for the uncoupled case, if I put a coupling constant, then I get as expected a non-zero value for the energy, but it is still non-dependent on field as it should be because of the Zeeman term.
In a previous post, it is said that the susceptibility is calculated from magnetization^2, so it should work while M is linear on H, is there a way to get the dependence of susceptibility against field as the system approaches to saturation?
And regarding the direction of field, setting H=1 means: Hz=1 Hx, Hy and Hz=1 Hx^2+Hy^2+Hz^2=1^2 , Hx=Hy=Hz or something else?
I also tried to set H="0 0 1" and the program runs, but i dont know what I am actually getting.
Any comments are welcome
Kind Regards Daniel
Dear all, Now that the field works, I find a strange scaling of energy with number of sites and spin: field: h=10000000000 Spin Number_of_sites Energy 1/2 2 -1e+10 1/2 3 -2e+10 1/2 4 -3e+10 1/2 5 -4e+10 1 2 -1e+10 1 3 -2e+10 1 4 -29999999999.5 1 5 -4e+10 3/2 2 -1e+10 3/2 3 -2e+10 3/2 4 -29999999999.8 3/2 5 -4e+10 2 2 -1e+10 2 3 -2e+10 2 4 -29999999999.7 2 5 -4e+10 5/2 2 -1e+10 5/2 3 -2e+10 5/2 4 -30000000000 5/2 5 -4e+10 looks like the energy is not dependent on the magnitude of spin (maybe it is normalized...) and that the scaling on energy with respect to the number of sites "misses" one site... i mean for 2 sites 1E10, 3 sites 2E10 and so on..., furthermore, the "Energy Density" changes depending on the number of sites for the same spin magnitude. Finally, I tried a one site simulation and got the following error: Checking if it is finished: not yet, next check in 60 seconds ( 0% done). Checking if it is finished: not yet, next check in 60 seconds ( 92% done). Checking if it is finished: not yet, next check in 60 seconds ( 0% done). Halted Simulation 1 HDF5 Error writing path '/' on type 'N4alps9scheduler12MCSimulationE': HDF5 Error writing path '/simulation/results' on type 'N4alps13ObservableSetE': HDF5 Error writing path 'Bond-type Energy' on type 'N4alps16SimpleObservableISt8valarrayIdENS_15DetailedBinningIS2_EEEE': HDF5 Error writing path '' on type 'N4alps15DetailedBinningISt8valarrayIdEEE': HDF5 Error writing path 'timeseries/logbinning' on type 'St6vectorISt8valarrayIdESaIS1_EE': Error in /scratch/jingles/alps/alps-2.0.0-r5363-src-with-boost/alps/src/alps/hdf5.hpp on 533 in resource: HDF5 error: -1 #0 H5Pdcpl.c line 858 in H5Pset_chunk(): all chunk dimensions must be positive The inputs are exactly like the last post one, only changing the fields to lowercase Kind Regards Daniel
Yes, now it works, thanks!
Kind regards Daniel
Use h and not H for the magnetic field - does that change things?
On 29 Feb 2012, at 21:42, daniel.aravena wrote:
One example of input file is:
LATTICE_LIBRARY=/home/g1daniel/ALPS_LATT/latt02.xml MODEL="Heisenberg" LATTICE="4_graph" UPDATE="local" CONVENTION="quantum" THERMALIZATION=1000000 SWEEPS=10000000 S0=1 J0=0 g=2.0 T=1.0 {H=0} {H=1} {H=10} {H=100} {H=1000} {H=10000} {H=100000} {H=1000000} {H=10000000} {H=100000000} {H=1000000000} {H=10000000000}
the "4_graph" lattice is:
<GRAPH name="4_graph" vertices="4"> <VERTEX id="1" type="0"></VERTEX> <VERTEX id="2" type="0"></VERTEX> <VERTEX id="3" type="0"></VERTEX> <VERTEX id="4" type="0"></VERTEX> <EDGE type="0" source="1" target="2"/> <EDGE type="0" source="1" target="3"/> <EDGE type="0" source="1" target="4"/> </GRAPH>
Thanks for your answer.
Kind regards Daniel
On Wed, 29 Feb 2012 09:28:01 +0100, Matthias Troyer wrote:
Can you please send your input file? Do you use local or cluster updates?
On Feb 23, 2012, at 3:24 PM, daniel.aravena@antares.qi.ub.edu wrote:
Dear all,
I am performing some tests using spinmc (alps 2.0, Heisenberg model) on uncoupled spins changing the number and magnitude of spins at different fields. There are some things in the results that I dont understand:
The energy is always zero, independently of the magnitude of the field for the uncoupled case, if I put a coupling constant, then I get as expected a non-zero value for the energy, but it is still non-dependent on field as it should be because of the Zeeman term.
In a previous post, it is said that the susceptibility is calculated from magnetization^2, so it should work while M is linear on H, is there a way to get the dependence of susceptibility against field as the system approaches to saturation?
And regarding the direction of field, setting H=1 means: Hz=1 Hx, Hy and Hz=1 Hx^2+Hy^2+Hz^2=1^2 , Hx=Hy=Hz or something else?
I also tried to set H="0 0 1" and the program runs, but i dont know what I am actually getting.
Any comments are welcome
Kind Regards Daniel
Indeed h is short for g*mu_B*S*h and it will thus not scale with S - but we can easily include that if you desire. To see the reason for the other issue I would need you to send me the lattice file that you used. On 1 Mar 2012, at 19:44, daniel.aravena@antares.qi.ub.edu wrote:
Dear all,
Now that the field works, I find a strange scaling of energy with number of sites and spin:
field: h=10000000000 Spin Number_of_sites Energy 1/2 2 -1e+10 1/2 3 -2e+10 1/2 4 -3e+10 1/2 5 -4e+10 1 2 -1e+10 1 3 -2e+10 1 4 -29999999999.5 1 5 -4e+10 3/2 2 -1e+10 3/2 3 -2e+10 3/2 4 -29999999999.8 3/2 5 -4e+10 2 2 -1e+10 2 3 -2e+10 2 4 -29999999999.7 2 5 -4e+10 5/2 2 -1e+10 5/2 3 -2e+10 5/2 4 -30000000000 5/2 5 -4e+10
looks like the energy is not dependent on the magnitude of spin (maybe it is normalized...) and that the scaling on energy with respect to the number of sites "misses" one site... i mean for 2 sites 1E10, 3 sites 2E10 and so on..., furthermore, the "Energy Density" changes depending on the number of sites for the same spin magnitude. Finally, I tried a one site simulation and got the following error:
Checking if it is finished: not yet, next check in 60 seconds ( 0% done). Checking if it is finished: not yet, next check in 60 seconds ( 92% done). Checking if it is finished: not yet, next check in 60 seconds ( 0% done). Halted Simulation 1 HDF5 Error writing path '/' on type 'N4alps9scheduler12MCSimulationE': HDF5 Error writing path '/simulation/results' on type 'N4alps13ObservableSetE': HDF5 Error writing path 'Bond-type Energy' on type 'N4alps16SimpleObservableISt8valarrayIdENS_15DetailedBinningIS2_EEEE': HDF5 Error writing path '' on type 'N4alps15DetailedBinningISt8valarrayIdEEE': HDF5 Error writing path 'timeseries/logbinning' on type 'St6vectorISt8valarrayIdESaIS1_EE': Error in /scratch/jingles/alps/alps-2.0.0-r5363-src-with-boost/alps/src/alps/hdf5.hpp on 533 in resource: HDF5 error: -1 #0 H5Pdcpl.c line 858 in H5Pset_chunk(): all chunk dimensions must be positive
The inputs are exactly like the last post one, only changing the fields to lowercase
Kind Regards Daniel
Yes, now it works, thanks!
Kind regards Daniel
Use h and not H for the magnetic field - does that change things?
On 29 Feb 2012, at 21:42, daniel.aravena wrote:
One example of input file is:
LATTICE_LIBRARY=/home/g1daniel/ALPS_LATT/latt02.xml MODEL="Heisenberg" LATTICE="4_graph" UPDATE="local" CONVENTION="quantum" THERMALIZATION=1000000 SWEEPS=10000000 S0=1 J0=0 g=2.0 T=1.0 {H=0} {H=1} {H=10} {H=100} {H=1000} {H=10000} {H=100000} {H=1000000} {H=10000000} {H=100000000} {H=1000000000} {H=10000000000}
the "4_graph" lattice is:
<GRAPH name="4_graph" vertices="4"> <VERTEX id="1" type="0"></VERTEX> <VERTEX id="2" type="0"></VERTEX> <VERTEX id="3" type="0"></VERTEX> <VERTEX id="4" type="0"></VERTEX> <EDGE type="0" source="1" target="2"/> <EDGE type="0" source="1" target="3"/> <EDGE type="0" source="1" target="4"/> </GRAPH>
Thanks for your answer.
Kind regards Daniel
On Wed, 29 Feb 2012 09:28:01 +0100, Matthias Troyer wrote:
Can you please send your input file? Do you use local or cluster updates?
On Feb 23, 2012, at 3:24 PM, daniel.aravena@antares.qi.ub.edu wrote:
Dear all,
I am performing some tests using spinmc (alps 2.0, Heisenberg model) on uncoupled spins changing the number and magnitude of spins at different fields. There are some things in the results that I dont understand:
The energy is always zero, independently of the magnitude of the field for the uncoupled case, if I put a coupling constant, then I get as expected a non-zero value for the energy, but it is still non-dependent on field as it should be because of the Zeeman term.
In a previous post, it is said that the susceptibility is calculated from magnetization^2, so it should work while M is linear on H, is there a way to get the dependence of susceptibility against field as the system approaches to saturation?
And regarding the direction of field, setting H=1 means: Hz=1 Hx, Hy and Hz=1 Hx^2+Hy^2+Hz^2=1^2 , Hx=Hy=Hz or something else?
I also tried to set H="0 0 1" and the program runs, but i dont know what I am actually getting.
Any comments are welcome
Kind Regards Daniel
perfect, but, how can you handle the field dependence of a system composed of sites of different spins assigning just a single field, considering that h is g*mu_B*S*h?. The lattice is: <GRAPH name="2_graph" vertices="2"> <VERTEX id="1" type="0"></VERTEX> <VERTEX id="2" type="0"></VERTEX> <EDGE type="0" source="1" target="2"/> </GRAPH> <GRAPH name="3_graph" vertices="3"> <VERTEX id="1" type="0"></VERTEX> <VERTEX id="2" type="0"></VERTEX> <VERTEX id="3" type="0"></VERTEX> <EDGE type="0" source="1" target="2"/> <EDGE type="0" source="1" target="3"/> </GRAPH> <GRAPH name="4_graph" vertices="4"> <VERTEX id="1" type="0"></VERTEX> <VERTEX id="2" type="0"></VERTEX> <VERTEX id="3" type="0"></VERTEX> <VERTEX id="4" type="0"></VERTEX> <EDGE type="0" source="1" target="2"/> <EDGE type="0" source="1" target="3"/> <EDGE type="0" source="1" target="4"/> </GRAPH> <GRAPH name="5_graph" vertices="5"> <VERTEX id="1" type="0"></VERTEX> <VERTEX id="2" type="0"></VERTEX> <VERTEX id="3" type="0"></VERTEX> <VERTEX id="4" type="0"></VERTEX> <VERTEX id="5" type="0"></VERTEX> <EDGE type="0" source="1" target="2"/> <EDGE type="0" source="1" target="3"/> <EDGE type="0" source="1" target="4"/> <EDGE type="0" source="1" target="5"/> </GRAPH> Thanks for your help Kind regards Daniel On Thu, 01 Mar 2012 20:42:59 +0100, Matthias Troyer wrote:
Indeed h is short for g*mu_B*S*h and it will thus not scale with S - but we can easily include that if you desire. To see the reason for the other issue I would need you to send me the lattice file that you used.
On 1 Mar 2012, at 19:44, daniel.aravena@antares.qi.ub.edu wrote:
Dear all,
Now that the field works, I find a strange scaling of energy with number of sites and spin:
field: h=10000000000 Spin Number_of_sites Energy 1/2 2 -1e+10 1/2 3 -2e+10 1/2 4 -3e+10 1/2 5 -4e+10 1 2 -1e+10 1 3 -2e+10 1 4 -29999999999.5 1 5 -4e+10 3/2 2 -1e+10 3/2 3 -2e+10 3/2 4 -29999999999.8 3/2 5 -4e+10 2 2 -1e+10 2 3 -2e+10 2 4 -29999999999.7 2 5 -4e+10 5/2 2 -1e+10 5/2 3 -2e+10 5/2 4 -30000000000 5/2 5 -4e+10
looks like the energy is not dependent on the magnitude of spin (maybe it is normalized...) and that the scaling on energy with respect to the number of sites "misses" one site... i mean for 2 sites 1E10, 3 sites 2E10 and so on..., furthermore, the "Energy Density" changes depending on the number of sites for the same spin magnitude. Finally, I tried a one site simulation and got the following error:
Checking if it is finished: not yet, next check in 60 seconds ( 0% done). Checking if it is finished: not yet, next check in 60 seconds ( 92% done). Checking if it is finished: not yet, next check in 60 seconds ( 0% done). Halted Simulation 1 HDF5 Error writing path '/' on type 'N4alps9scheduler12MCSimulationE': HDF5 Error writing path '/simulation/results' on type 'N4alps13ObservableSetE': HDF5 Error writing path 'Bond-type Energy' on type
'N4alps16SimpleObservableISt8valarrayIdENS_15DetailedBinningIS2_EEEE': HDF5 Error writing path '' on type 'N4alps15DetailedBinningISt8valarrayIdEEE': HDF5 Error writing path 'timeseries/logbinning' on type 'St6vectorISt8valarrayIdESaIS1_EE': Error in
/scratch/jingles/alps/alps-2.0.0-r5363-src-with-boost/alps/src/alps/hdf5.hpp on 533 in resource: HDF5 error: -1 #0 H5Pdcpl.c line 858 in H5Pset_chunk(): all chunk dimensions must be positive
The inputs are exactly like the last post one, only changing the fields to lowercase
Kind Regards Daniel
Yes, now it works, thanks!
Kind regards Daniel
Use h and not H for the magnetic field - does that change things?
On 29 Feb 2012, at 21:42, daniel.aravena wrote:
One example of input file is:
LATTICE_LIBRARY=/home/g1daniel/ALPS_LATT/latt02.xml MODEL="Heisenberg" LATTICE="4_graph" UPDATE="local" CONVENTION="quantum" THERMALIZATION=1000000 SWEEPS=10000000 S0=1 J0=0 g=2.0 T=1.0 {H=0} {H=1} {H=10} {H=100} {H=1000} {H=10000} {H=100000} {H=1000000} {H=10000000} {H=100000000} {H=1000000000} {H=10000000000}
the "4_graph" lattice is:
<GRAPH name="4_graph" vertices="4"> <VERTEX id="1" type="0"></VERTEX> <VERTEX id="2" type="0"></VERTEX> <VERTEX id="3" type="0"></VERTEX> <VERTEX id="4" type="0"></VERTEX> <EDGE type="0" source="1" target="2"/> <EDGE type="0" source="1" target="3"/> <EDGE type="0" source="1" target="4"/> </GRAPH>
Thanks for your answer.
Kind regards Daniel
On Wed, 29 Feb 2012 09:28:01 +0100, Matthias Troyer wrote:
Can you please send your input file? Do you use local or cluster updates?
On Feb 23, 2012, at 3:24 PM, daniel.aravena@antares.qi.ub.edu wrote:
> Dear all, > > I am performing some tests using spinmc (alps 2.0, Heisenberg > model) > on > uncoupled spins changing the number and magnitude of spins at > different > fields. There are some things in the results that I dont > understand: > > The energy is always zero, independently of the magnitude of > the > field > for > the uncoupled case, if I put a coupling constant, then I get as > expected a > non-zero value for the energy, but it is still non-dependent on > field > as > it should be because of the Zeeman term. > > In a previous post, it is said that the susceptibility is > calculated > from > magnetization^2, so it should work while M is linear on H, is > there a > way > to get the dependence of susceptibility against field as the > system > approaches to saturation? > > And regarding the direction of field, setting H=1 means: > Hz=1 > Hx, Hy and Hz=1 > Hx^2+Hy^2+Hz^2=1^2 , Hx=Hy=Hz > or something else? > > I also tried to set H="0 0 1" and the program runs, but i dont > know > what I > am actually getting. > > Any comments are welcome > > > Kind Regards > Daniel > >
No, I'm sorry - I checked it and we actually do include S in the field term as we should. TO look into this Iw ould like to ask you to send me your complete parameter file for one of the cases, e.g. for
Spin Number_of_sites Energy 1/2 2 -1e+10
On 1 Mar 2012, at 22:14, daniel.aravena wrote:
perfect, but, how can you handle the field dependence of a system composed of sites of different spins assigning just a single field
Matthias
ok, here it is the parameter file for S=1/2 and 2 sites: LATTICE_LIBRARY=/home/g1daniel/ALPS_LATT/latt02.xml MODEL="Heisenberg" LATTICE="2_graph" UPDATE="local" CONVENTION="quantum" THERMALIZATION=1000000 SWEEPS=10000000 S0=1/2 J0=0 g=1.0 T=1.0 {h=0} {h=1} {h=10} {h=100} {h=1000} {h=10000} {h=100000} {h=1000000} {h=10000000} {h=100000000} {h=1000000000} {h=10000000000} the last field (h=10000000000) gives -1e+10 for the energy Kind regards Daniel On Thu, 01 Mar 2012 22:31:43 +0100, Matthias Troyer wrote:
No, I'm sorry - I checked it and we actually do include S in the field term as we should. TO look into this Iw ould like to ask you to send me your complete parameter file for one of the cases, e.g. for
Spin Number_of_sites Energy 1/2 2 -1e+10
On 1 Mar 2012, at 22:14, daniel.aravena wrote:
perfect, but, how can you handle the field dependence of a system composed of sites of different spins assigning just a single field
Matthias
Hi Daniel, Can you please try to apply the following patch, and turn the quantum convention off. It will not give sensible results in magnetic fields. Matthias Index: applications/mc/spins/abstractspinsim.h =================================================================== --- applications/mc/spins/abstractspinsim.h (revision 6011) +++ applications/mc/spins/abstractspinsim.h (working copy) @@ -136,6 +136,11 @@ else use_error_limit = false; + if (has_magnetic_field_ && quantum_convention_) + boost::throw_exception( + std::runtime_error("Cannot use quantum convention with nonzero magnetic field .")); + + if (parms["UPDATE"]=="cluster" && has_magnetic_field_) {boost::throw_exception(std::runtime_error("Illegal update type " + std::string(parms["UPDATE"]) @@ -159,7 +164,7 @@ spinvalue = quantum_convention_ ? 0.5 : 1.; else spinvalue = alps::evaluate<double>(parms[spin],parms); - spinfactor_[*si]=std::sqrt(spinvalue*(spinvalue+(quantum_convention_ ? 1. : 0.)))*g_; + spinfactor_[*si]=std::sqrt(spinvalue*(spinvalue+(quantum_convention_ ? 1. : 0.))); // set the values for the diagonal matrix terms site_type_visited[type]=true; @@ -176,7 +181,7 @@ selfinteraction_[*si].setMatrix(diag_string,parms); general_case_=true; } - selfinteraction_[*si] = selfinteraction_[*si]*spinvalue*(spinvalue+(quantum_convention_ ? 1. : 0.)); + selfinteraction_[*si] = selfinteraction_[*si]*spinvalue*spinvalue; if (!selfinteraction_[*si].is_symmetric()) boost::throw_exception(std::runtime_error("Invalid matrix for D")); Index: applications/mc/spins/spinsim.h =================================================================== --- applications/mc/spins/spinsim.h (revision 6011) +++ applications/mc/spins/spinsim.h (working copy) @@ -246,14 +246,14 @@ m_h = this->spinfactor_[*si]*spins_[*si].mag_h(this->h_normalized); if (this->has_magnetic_field_) - en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_; double par=this->parity(*si); ++si; for (; si !=this->sites().second ; ++si) { if (this->has_magnetic_field_) - en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_; if (this->general_case_) en+=onsite_energy(spins_[*si],this->selfinteraction_[*si]); @@ -321,7 +321,6 @@ if (this->print_sweeps_) { int good_sweeps = this->sweeps_done_-this->thermalization_sweeps_; if (good_sweeps > 0 && (good_sweeps % this->print_sweeps_ ==0)) { - std::cout << "Dumping configuration: \n"; for (int i=0;i<this->num_sites();++i) { std::cout << i << " "; std::vector<double> coord = this->coordinate(i); @@ -350,7 +349,7 @@ bt_energy.resize(bt+1); bt_count.resize(bt+1); } - double be = bond_energy(spins_[this->source(*bi)],spins_[this->target(*bi)], + double be = this->spinfactor_[this->source(*bi)]*this->spinfactor_[this->target(*bi)]*bond_energy(spins_[this->source(*bi)],spins_[this->target(*bi)], this->couplings_[*bi]); en -= be; @@ -374,13 +373,16 @@ mst *= this->spinfactor_[*si]; double m_h = 0.; if (this->has_magnetic_field_) + { m_h = this->spinfactor_[*si]*spins_[*si].mag_h(this->h_normalized); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_; + } double par=this->parity(*si); ++si; for (; si !=this->sites().second ; ++si) { if (this->has_magnetic_field_) - en+=site_energy(spins_[*si],this->h_); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_; if (this->general_case_) en+=onsite_energy(spins_[*si],this->selfinteraction_[*si]); On Mar 1, 2012, at 10:39 PM, daniel.aravena wrote:
ok, here it is the parameter file for S=1/2 and 2 sites:
LATTICE_LIBRARY=/home/g1daniel/ALPS_LATT/latt02.xml MODEL="Heisenberg" LATTICE="2_graph" UPDATE="local" CONVENTION="quantum" THERMALIZATION=1000000 SWEEPS=10000000 S0=1/2 J0=0 g=1.0 T=1.0 {h=0} {h=1} {h=10} {h=100} {h=1000} {h=10000} {h=100000} {h=1000000} {h=10000000} {h=100000000} {h=1000000000} {h=10000000000}
the last field (h=10000000000) gives -1e+10 for the energy
Kind regards Daniel
On Thu, 01 Mar 2012 22:31:43 +0100, Matthias Troyer wrote:
No, I'm sorry - I checked it and we actually do include S in the field term as we should. TO look into this Iw ould like to ask you to send me your complete parameter file for one of the cases, e.g. for
Spin Number_of_sites Energy 1/2 2 -1e+10
On 1 Mar 2012, at 22:14, daniel.aravena wrote:
perfect, but, how can you handle the field dependence of a system composed of sites of different spins assigning just a single field
Matthias
Hi Matthias, I tried the new patch and it works correctly, the only thing I find strange is that the "Magnetization along field" vs. magnetic field plot doesn't look like dependent on the g factor but the energies are dependent on g, so i would expect a higher magnetization for higher g at the same field when far from saturation while i get the same magnetization regardless of the value of g (i tried g=1.0 and 2.0 and h from 0 to 11). Thanks Kind Regards Daniel
Hi Daniel,
Can you please try to apply the following patch, and turn the quantum convention off. It will not give sensible results in magnetic fields.
Matthias
Index: applications/mc/spins/abstractspinsim.h =================================================================== --- applications/mc/spins/abstractspinsim.h (revision 6011) +++ applications/mc/spins/abstractspinsim.h (working copy) @@ -136,6 +136,11 @@ else use_error_limit = false;
+ if (has_magnetic_field_ && quantum_convention_) + boost::throw_exception( + std::runtime_error("Cannot use quantum convention with nonzero magnetic field .")); + + if (parms["UPDATE"]=="cluster" && has_magnetic_field_) {boost::throw_exception(std::runtime_error("Illegal update type " + std::string(parms["UPDATE"]) @@ -159,7 +164,7 @@ spinvalue = quantum_convention_ ? 0.5 : 1.; else spinvalue = alps::evaluate<double>(parms[spin],parms); - spinfactor_[*si]=std::sqrt(spinvalue*(spinvalue+(quantum_convention_ ? 1. : 0.)))*g_; + spinfactor_[*si]=std::sqrt(spinvalue*(spinvalue+(quantum_convention_ ? 1. : 0.)));
// set the values for the diagonal matrix terms site_type_visited[type]=true; @@ -176,7 +181,7 @@ selfinteraction_[*si].setMatrix(diag_string,parms); general_case_=true; } - selfinteraction_[*si] = selfinteraction_[*si]*spinvalue*(spinvalue+(quantum_convention_ ? 1. : 0.)); + selfinteraction_[*si] = selfinteraction_[*si]*spinvalue*spinvalue;
if (!selfinteraction_[*si].is_symmetric()) boost::throw_exception(std::runtime_error("Invalid matrix for D")); Index: applications/mc/spins/spinsim.h =================================================================== --- applications/mc/spins/spinsim.h (revision 6011) +++ applications/mc/spins/spinsim.h (working copy) @@ -246,14 +246,14 @@ m_h = this->spinfactor_[*si]*spins_[*si].mag_h(this->h_normalized);
if (this->has_magnetic_field_) - en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_;
double par=this->parity(*si); ++si;
for (; si !=this->sites().second ; ++si) { if (this->has_magnetic_field_) - en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_;
if (this->general_case_) en+=onsite_energy(spins_[*si],this->selfinteraction_[*si]); @@ -321,7 +321,6 @@ if (this->print_sweeps_) { int good_sweeps = this->sweeps_done_-this->thermalization_sweeps_; if (good_sweeps > 0 && (good_sweeps % this->print_sweeps_ ==0)) { - std::cout << "Dumping configuration: \n"; for (int i=0;i<this->num_sites();++i) { std::cout << i << " "; std::vector<double> coord = this->coordinate(i); @@ -350,7 +349,7 @@ bt_energy.resize(bt+1); bt_count.resize(bt+1); } - double be = bond_energy(spins_[this->source(*bi)],spins_[this->target(*bi)], + double be = this->spinfactor_[this->source(*bi)]*this->spinfactor_[this->target(*bi)]*bond_energy(spins_[this->source(*bi)],spins_[this->target(*bi)], this->couplings_[*bi]);
en -= be; @@ -374,13 +373,16 @@ mst *= this->spinfactor_[*si]; double m_h = 0.; if (this->has_magnetic_field_) + { m_h = this->spinfactor_[*si]*spins_[*si].mag_h(this->h_normalized); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_; + } double par=this->parity(*si); ++si;
for (; si !=this->sites().second ; ++si) { if (this->has_magnetic_field_) - en+=site_energy(spins_[*si],this->h_); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_; if (this->general_case_) en+=onsite_energy(spins_[*si],this->selfinteraction_[*si]);
On Mar 1, 2012, at 10:39 PM, daniel.aravena wrote:
ok, here it is the parameter file for S=1/2 and 2 sites:
LATTICE_LIBRARY=/home/g1daniel/ALPS_LATT/latt02.xml MODEL="Heisenberg" LATTICE="2_graph" UPDATE="local" CONVENTION="quantum" THERMALIZATION=1000000 SWEEPS=10000000 S0=1/2 J0=0 g=1.0 T=1.0 {h=0} {h=1} {h=10} {h=100} {h=1000} {h=10000} {h=100000} {h=1000000} {h=10000000} {h=100000000} {h=1000000000} {h=10000000000}
the last field (h=10000000000) gives -1e+10 for the energy
Kind regards Daniel
On Thu, 01 Mar 2012 22:31:43 +0100, Matthias Troyer wrote:
No, I'm sorry - I checked it and we actually do include S in the field term as we should. TO look into this Iw ould like to ask you to send me your complete parameter file for one of the cases, e.g. for
Spin Number_of_sites Energy 1/2 2 -1e+10
On 1 Mar 2012, at 22:14, daniel.aravena wrote:
perfect, but, how can you handle the field dependence of a system composed of sites of different spins assigning just a single field
Matthias
Hi Daniel, Indeed nobody ever wanted to change g and some modification dropped the energy dependence on g. The following patch should fix it. Matthias Index: localupdate.h =================================================================== --- localupdate.h (revision 6036) +++ localupdate.h (working copy) @@ -42,7 +42,7 @@ template <class Graph, class MomentMap, class RNG, class CouplingMap, class SpinfactorMap> void local_update(const Graph& graph, MomentMap& moment, double beta, RNG& rng, const CouplingMap& coupling, - const SpinfactorMap& spinfactor, + const SpinfactorMap& spinfactor, double g, TinyVector<double,CouplingMap::value_type::dim> h = TinyVector<double,CouplingMap::value_type::dim>(0.0)) @@ -72,7 +72,7 @@ delta_E += bond_energy_change(moment[boost::source(*edge,graph)], moment[boost::target(*edge,graph)],J,update); } - delta_E += spinfactor[vertex]*site_energy_change(moment[vertex], h, update); + delta_E += spinfactor[vertex]*site_energy_change(moment[vertex], h, update) * g; // step iv) if (delta_E<0 || rng() < exp(-beta*delta_E)) @@ -86,7 +86,7 @@ void local_update_self(const Graph& graph, MomentMap& moment, double beta, RNG& rng, const CouplingMap& coupling, const SelfinteractionMap& selfinteraction, - const SpinfactorMap& spinfactor, + const SpinfactorMap& spinfactor, double g, TinyVector<double,CouplingMap::value_type::dim> h = TinyVector<double,CouplingMap::value_type::dim>(0.0)) { @@ -115,7 +115,7 @@ delta_E += bond_energy_change(moment[boost::source(*edge,graph)], moment[boost::target(*edge,graph)],J,update); } - delta_E += spinfactor[vertex]*site_energy_change(moment[vertex], h, update); + delta_E += spinfactor[vertex]*site_energy_change(moment[vertex], h, update) * g; delta_E += onsite_energy_change(moment[vertex], selfinteraction[vertex],update); Index: abstractspinsim.h =================================================================== --- abstractspinsim.h (revision 6036) +++ abstractspinsim.h (working copy) @@ -205,7 +205,7 @@ MAT this_J; this_J.setMatrix(J_name,parms); bond_type[*b] = num_bond_types; - this_J = this_J /(g_*g_) * (quantum_convention_ ? 1. : -1.); + this_J = this_J * (quantum_convention_ ? 1. : -1.); couplings_[*b]=this_J; double det = this_J.det(); couplings_det[*b] = det; Index: spinsim.h =================================================================== --- spinsim.h (revision 6036) +++ spinsim.h (working copy) @@ -166,10 +166,10 @@ typedef AbstractSpinSim<MAT> parent; if (this->general_case_) local_update_self(this->graph(),spins_, this->beta_,this->random_01, - this->couplings_,this->selfinteraction_,this->spinfactor_,this->h_); + this->couplings_,this->selfinteraction_,this->spinfactor_,this->g_,this->h_); else local_update(this->graph(),spins_, this->beta_,this->random_01, - this->couplings_,this->spinfactor_,this->h_); + this->couplings_,this->spinfactor_,this->g_,this->h_); update_info_type up_info; up_info.clustersize = this->num_sites(); up_info.m2_ = 0; @@ -190,10 +190,10 @@ else { if (this->general_case_) local_update_self(this->graph(),spins_,this->beta_,this->random_01, - this->couplings_,this->selfinteraction_,this->spinfactor_,this->h_); + this->couplings_,this->selfinteraction_,this->spinfactor_,this->g_,this->h_); else local_update(this->graph(),spins_,this->beta_,this->random_01, - this->couplings_,this->spinfactor_,this->h_); + this->couplings_,this->spinfactor_,this->g_,this->h_); update_info_type up_info; up_info.clustersize = this->num_sites(); up_info.m2_ = 0; On 7 Mar 2012, at 07:43, daniel.aravena@antares.qi.ub.edu wrote:
Hi Matthias,
I tried the new patch and it works correctly, the only thing I find strange is that the "Magnetization along field" vs. magnetic field plot doesn't look like dependent on the g factor but the energies are dependent on g, so i would expect a higher magnetization for higher g at the same field when far from saturation while i get the same magnetization regardless of the value of g (i tried g=1.0 and 2.0 and h from 0 to 11).
Thanks Kind Regards Daniel
Hi Daniel,
Can you please try to apply the following patch, and turn the quantum convention off. It will not give sensible results in magnetic fields.
Matthias
Index: applications/mc/spins/abstractspinsim.h =================================================================== --- applications/mc/spins/abstractspinsim.h (revision 6011) +++ applications/mc/spins/abstractspinsim.h (working copy) @@ -136,6 +136,11 @@ else use_error_limit = false;
+ if (has_magnetic_field_ && quantum_convention_) + boost::throw_exception( + std::runtime_error("Cannot use quantum convention with nonzero magnetic field .")); + + if (parms["UPDATE"]=="cluster" && has_magnetic_field_) {boost::throw_exception(std::runtime_error("Illegal update type " + std::string(parms["UPDATE"]) @@ -159,7 +164,7 @@ spinvalue = quantum_convention_ ? 0.5 : 1.; else spinvalue = alps::evaluate<double>(parms[spin],parms); - spinfactor_[*si]=std::sqrt(spinvalue*(spinvalue+(quantum_convention_ ? 1. : 0.)))*g_; + spinfactor_[*si]=std::sqrt(spinvalue*(spinvalue+(quantum_convention_ ? 1. : 0.)));
// set the values for the diagonal matrix terms site_type_visited[type]=true; @@ -176,7 +181,7 @@ selfinteraction_[*si].setMatrix(diag_string,parms); general_case_=true; } - selfinteraction_[*si] = selfinteraction_[*si]*spinvalue*(spinvalue+(quantum_convention_ ? 1. : 0.)); + selfinteraction_[*si] = selfinteraction_[*si]*spinvalue*spinvalue;
if (!selfinteraction_[*si].is_symmetric()) boost::throw_exception(std::runtime_error("Invalid matrix for D")); Index: applications/mc/spins/spinsim.h =================================================================== --- applications/mc/spins/spinsim.h (revision 6011) +++ applications/mc/spins/spinsim.h (working copy) @@ -246,14 +246,14 @@ m_h = this->spinfactor_[*si]*spins_[*si].mag_h(this->h_normalized);
if (this->has_magnetic_field_) - en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_;
double par=this->parity(*si); ++si;
for (; si !=this->sites().second ; ++si) { if (this->has_magnetic_field_) - en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_;
if (this->general_case_) en+=onsite_energy(spins_[*si],this->selfinteraction_[*si]); @@ -321,7 +321,6 @@ if (this->print_sweeps_) { int good_sweeps = this->sweeps_done_-this->thermalization_sweeps_; if (good_sweeps > 0 && (good_sweeps % this->print_sweeps_ ==0)) { - std::cout << "Dumping configuration: \n"; for (int i=0;i<this->num_sites();++i) { std::cout << i << " "; std::vector<double> coord = this->coordinate(i); @@ -350,7 +349,7 @@ bt_energy.resize(bt+1); bt_count.resize(bt+1); } - double be = bond_energy(spins_[this->source(*bi)],spins_[this->target(*bi)], + double be = this->spinfactor_[this->source(*bi)]*this->spinfactor_[this->target(*bi)]*bond_energy(spins_[this->source(*bi)],spins_[this->target(*bi)], this->couplings_[*bi]);
en -= be; @@ -374,13 +373,16 @@ mst *= this->spinfactor_[*si]; double m_h = 0.; if (this->has_magnetic_field_) + { m_h = this->spinfactor_[*si]*spins_[*si].mag_h(this->h_normalized); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_; + } double par=this->parity(*si); ++si;
for (; si !=this->sites().second ; ++si) { if (this->has_magnetic_field_) - en+=site_energy(spins_[*si],this->h_); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_; if (this->general_case_) en+=onsite_energy(spins_[*si],this->selfinteraction_[*si]);
On Mar 1, 2012, at 10:39 PM, daniel.aravena wrote:
ok, here it is the parameter file for S=1/2 and 2 sites:
LATTICE_LIBRARY=/home/g1daniel/ALPS_LATT/latt02.xml MODEL="Heisenberg" LATTICE="2_graph" UPDATE="local" CONVENTION="quantum" THERMALIZATION=1000000 SWEEPS=10000000 S0=1/2 J0=0 g=1.0 T=1.0 {h=0} {h=1} {h=10} {h=100} {h=1000} {h=10000} {h=100000} {h=1000000} {h=10000000} {h=100000000} {h=1000000000} {h=10000000000}
the last field (h=10000000000) gives -1e+10 for the energy
Kind regards Daniel
On Thu, 01 Mar 2012 22:31:43 +0100, Matthias Troyer wrote:
No, I'm sorry - I checked it and we actually do include S in the field term as we should. TO look into this Iw ould like to ask you to send me your complete parameter file for one of the cases, e.g. for
> Spin Number_of_sites Energy > 1/2 2 -1e+10
On 1 Mar 2012, at 22:14, daniel.aravena wrote:
perfect, but, how can you handle the field dependence of a system composed of sites of different spins assigning just a single field
Matthias
Hi Daniel, Indeed nobody ever wanted to change g and some modification dropped the energy dependence on g. The following patch should fix it. Matthias Index: localupdate.h =================================================================== --- localupdate.h (revision 6036) +++ localupdate.h (working copy) @@ -42,7 +42,7 @@ template <class Graph, class MomentMap, class RNG, class CouplingMap, class SpinfactorMap> void local_update(const Graph& graph, MomentMap& moment, double beta, RNG& rng, const CouplingMap& coupling, - const SpinfactorMap& spinfactor, + const SpinfactorMap& spinfactor, double g, TinyVector<double,CouplingMap::value_type::dim> h = TinyVector<double,CouplingMap::value_type::dim>(0.0)) @@ -72,7 +72,7 @@ delta_E += bond_energy_change(moment[boost::source(*edge,graph)], moment[boost::target(*edge,graph)],J,update); } - delta_E += spinfactor[vertex]*site_energy_change(moment[vertex], h, update); + delta_E += spinfactor[vertex]*site_energy_change(moment[vertex], h, update) * g; // step iv) if (delta_E<0 || rng() < exp(-beta*delta_E)) @@ -86,7 +86,7 @@ void local_update_self(const Graph& graph, MomentMap& moment, double beta, RNG& rng, const CouplingMap& coupling, const SelfinteractionMap& selfinteraction, - const SpinfactorMap& spinfactor, + const SpinfactorMap& spinfactor, double g, TinyVector<double,CouplingMap::value_type::dim> h = TinyVector<double,CouplingMap::value_type::dim>(0.0)) { @@ -115,7 +115,7 @@ delta_E += bond_energy_change(moment[boost::source(*edge,graph)], moment[boost::target(*edge,graph)],J,update); } - delta_E += spinfactor[vertex]*site_energy_change(moment[vertex], h, update); + delta_E += spinfactor[vertex]*site_energy_change(moment[vertex], h, update) * g; delta_E += onsite_energy_change(moment[vertex], selfinteraction[vertex],update); Index: abstractspinsim.h =================================================================== --- abstractspinsim.h (revision 6036) +++ abstractspinsim.h (working copy) @@ -205,7 +205,7 @@ MAT this_J; this_J.setMatrix(J_name,parms); bond_type[*b] = num_bond_types; - this_J = this_J /(g_*g_) * (quantum_convention_ ? 1. : -1.); + this_J = this_J * (quantum_convention_ ? 1. : -1.); couplings_[*b]=this_J; double det = this_J.det(); couplings_det[*b] = det; Index: spinsim.h =================================================================== --- spinsim.h (revision 6036) +++ spinsim.h (working copy) @@ -166,10 +166,10 @@ typedef AbstractSpinSim<MAT> parent; if (this->general_case_) local_update_self(this->graph(),spins_, this->beta_,this->random_01, - this->couplings_,this->selfinteraction_,this->spinfactor_,this->h_); + this->couplings_,this->selfinteraction_,this->spinfactor_,this->g_,this->h_); else local_update(this->graph(),spins_, this->beta_,this->random_01, - this->couplings_,this->spinfactor_,this->h_); + this->couplings_,this->spinfactor_,this->g_,this->h_); update_info_type up_info; up_info.clustersize = this->num_sites(); up_info.m2_ = 0; @@ -190,10 +190,10 @@ else { if (this->general_case_) local_update_self(this->graph(),spins_,this->beta_,this->random_01, - this->couplings_,this->selfinteraction_,this->spinfactor_,this->h_); + this->couplings_,this->selfinteraction_,this->spinfactor_,this->g_,this->h_); else local_update(this->graph(),spins_,this->beta_,this->random_01, - this->couplings_,this->spinfactor_,this->h_); + this->couplings_,this->spinfactor_,this->g_,this->h_); update_info_type up_info; up_info.clustersize = this->num_sites(); up_info.m2_ = 0; On 7 Mar 2012, at 07:43, daniel.aravena@antares.qi.ub.edu wrote:
Hi Matthias,
I tried the new patch and it works correctly, the only thing I find strange is that the "Magnetization along field" vs. magnetic field plot doesn't look like dependent on the g factor but the energies are dependent on g, so i would expect a higher magnetization for higher g at the same field when far from saturation while i get the same magnetization regardless of the value of g (i tried g=1.0 and 2.0 and h from 0 to 11).
Thanks Kind Regards Daniel
Hi Daniel,
Can you please try to apply the following patch, and turn the quantum convention off. It will not give sensible results in magnetic fields.
Matthias
Index: applications/mc/spins/abstractspinsim.h =================================================================== --- applications/mc/spins/abstractspinsim.h (revision 6011) +++ applications/mc/spins/abstractspinsim.h (working copy) @@ -136,6 +136,11 @@ else use_error_limit = false;
+ if (has_magnetic_field_ && quantum_convention_) + boost::throw_exception( + std::runtime_error("Cannot use quantum convention with nonzero magnetic field .")); + + if (parms["UPDATE"]=="cluster" && has_magnetic_field_) {boost::throw_exception(std::runtime_error("Illegal update type " + std::string(parms["UPDATE"]) @@ -159,7 +164,7 @@ spinvalue = quantum_convention_ ? 0.5 : 1.; else spinvalue = alps::evaluate<double>(parms[spin],parms); - spinfactor_[*si]=std::sqrt(spinvalue*(spinvalue+(quantum_convention_ ? 1. : 0.)))*g_; + spinfactor_[*si]=std::sqrt(spinvalue*(spinvalue+(quantum_convention_ ? 1. : 0.)));
// set the values for the diagonal matrix terms site_type_visited[type]=true; @@ -176,7 +181,7 @@ selfinteraction_[*si].setMatrix(diag_string,parms); general_case_=true; } - selfinteraction_[*si] = selfinteraction_[*si]*spinvalue*(spinvalue+(quantum_convention_ ? 1. : 0.)); + selfinteraction_[*si] = selfinteraction_[*si]*spinvalue*spinvalue;
if (!selfinteraction_[*si].is_symmetric()) boost::throw_exception(std::runtime_error("Invalid matrix for D")); Index: applications/mc/spins/spinsim.h =================================================================== --- applications/mc/spins/spinsim.h (revision 6011) +++ applications/mc/spins/spinsim.h (working copy) @@ -246,14 +246,14 @@ m_h = this->spinfactor_[*si]*spins_[*si].mag_h(this->h_normalized);
if (this->has_magnetic_field_) - en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_;
double par=this->parity(*si); ++si;
for (; si !=this->sites().second ; ++si) { if (this->has_magnetic_field_) - en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_;
if (this->general_case_) en+=onsite_energy(spins_[*si],this->selfinteraction_[*si]); @@ -321,7 +321,6 @@ if (this->print_sweeps_) { int good_sweeps = this->sweeps_done_-this->thermalization_sweeps_; if (good_sweeps > 0 && (good_sweeps % this->print_sweeps_ ==0)) { - std::cout << "Dumping configuration: \n"; for (int i=0;i<this->num_sites();++i) { std::cout << i << " "; std::vector<double> coord = this->coordinate(i); @@ -350,7 +349,7 @@ bt_energy.resize(bt+1); bt_count.resize(bt+1); } - double be = bond_energy(spins_[this->source(*bi)],spins_[this->target(*bi)], + double be = this->spinfactor_[this->source(*bi)]*this->spinfactor_[this->target(*bi)]*bond_energy(spins_[this->source(*bi)],spins_[this->target(*bi)], this->couplings_[*bi]);
en -= be; @@ -374,13 +373,16 @@ mst *= this->spinfactor_[*si]; double m_h = 0.; if (this->has_magnetic_field_) + { m_h = this->spinfactor_[*si]*spins_[*si].mag_h(this->h_normalized); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_; + } double par=this->parity(*si); ++si;
for (; si !=this->sites().second ; ++si) { if (this->has_magnetic_field_) - en+=site_energy(spins_[*si],this->h_); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_; if (this->general_case_) en+=onsite_energy(spins_[*si],this->selfinteraction_[*si]);
On Mar 1, 2012, at 10:39 PM, daniel.aravena wrote:
ok, here it is the parameter file for S=1/2 and 2 sites:
LATTICE_LIBRARY=/home/g1daniel/ALPS_LATT/latt02.xml MODEL="Heisenberg" LATTICE="2_graph" UPDATE="local" CONVENTION="quantum" THERMALIZATION=1000000 SWEEPS=10000000 S0=1/2 J0=0 g=1.0 T=1.0 {h=0} {h=1} {h=10} {h=100} {h=1000} {h=10000} {h=100000} {h=1000000} {h=10000000} {h=100000000} {h=1000000000} {h=10000000000}
the last field (h=10000000000) gives -1e+10 for the energy
Kind regards Daniel
On Thu, 01 Mar 2012 22:31:43 +0100, Matthias Troyer wrote:
No, I'm sorry - I checked it and we actually do include S in the field term as we should. TO look into this Iw ould like to ask you to send me your complete parameter file for one of the cases, e.g. for
> Spin Number_of_sites Energy > 1/2 2 -1e+10
On 1 Mar 2012, at 22:14, daniel.aravena wrote:
perfect, but, how can you handle the field dependence of a system composed of sites of different spins assigning just a single field
Matthias
Hi Matthias, I tried the new patch and magnetization now varies with g, so everything is ok! Thank you Kind Regards Daniel
Hi Daniel,
Indeed nobody ever wanted to change g and some modification dropped the energy dependence on g. The following patch should fix it.
Matthias
Index: localupdate.h =================================================================== --- localupdate.h (revision 6036) +++ localupdate.h (working copy) @@ -42,7 +42,7 @@ template <class Graph, class MomentMap, class RNG, class CouplingMap, class SpinfactorMap> void local_update(const Graph& graph, MomentMap& moment, double beta, RNG& rng, const CouplingMap& coupling, - const SpinfactorMap& spinfactor, + const SpinfactorMap& spinfactor, double g, TinyVector<double,CouplingMap::value_type::dim> h = TinyVector<double,CouplingMap::value_type::dim>(0.0))
@@ -72,7 +72,7 @@ delta_E += bond_energy_change(moment[boost::source(*edge,graph)], moment[boost::target(*edge,graph)],J,update); } - delta_E += spinfactor[vertex]*site_energy_change(moment[vertex], h, update); + delta_E += spinfactor[vertex]*site_energy_change(moment[vertex], h, update) * g;
// step iv) if (delta_E<0 || rng() < exp(-beta*delta_E)) @@ -86,7 +86,7 @@ void local_update_self(const Graph& graph, MomentMap& moment, double beta, RNG& rng, const CouplingMap& coupling, const SelfinteractionMap& selfinteraction, - const SpinfactorMap& spinfactor, + const SpinfactorMap& spinfactor, double g, TinyVector<double,CouplingMap::value_type::dim> h = TinyVector<double,CouplingMap::value_type::dim>(0.0)) { @@ -115,7 +115,7 @@ delta_E += bond_energy_change(moment[boost::source(*edge,graph)], moment[boost::target(*edge,graph)],J,update); } - delta_E += spinfactor[vertex]*site_energy_change(moment[vertex], h, update); + delta_E += spinfactor[vertex]*site_energy_change(moment[vertex], h, update) * g;
delta_E += onsite_energy_change(moment[vertex], selfinteraction[vertex],update);
Index: abstractspinsim.h =================================================================== --- abstractspinsim.h (revision 6036) +++ abstractspinsim.h (working copy) @@ -205,7 +205,7 @@ MAT this_J; this_J.setMatrix(J_name,parms); bond_type[*b] = num_bond_types; - this_J = this_J /(g_*g_) * (quantum_convention_ ? 1. : -1.); + this_J = this_J * (quantum_convention_ ? 1. : -1.); couplings_[*b]=this_J; double det = this_J.det(); couplings_det[*b] = det; Index: spinsim.h =================================================================== --- spinsim.h (revision 6036) +++ spinsim.h (working copy) @@ -166,10 +166,10 @@ typedef AbstractSpinSim<MAT> parent; if (this->general_case_) local_update_self(this->graph(),spins_, this->beta_,this->random_01, - this->couplings_,this->selfinteraction_,this->spinfactor_,this->h_); + this->couplings_,this->selfinteraction_,this->spinfactor_,this->g_,this->h_); else local_update(this->graph(),spins_, this->beta_,this->random_01, - this->couplings_,this->spinfactor_,this->h_); + this->couplings_,this->spinfactor_,this->g_,this->h_); update_info_type up_info; up_info.clustersize = this->num_sites(); up_info.m2_ = 0; @@ -190,10 +190,10 @@ else { if (this->general_case_) local_update_self(this->graph(),spins_,this->beta_,this->random_01, - this->couplings_,this->selfinteraction_,this->spinfactor_,this->h_); + this->couplings_,this->selfinteraction_,this->spinfactor_,this->g_,this->h_); else local_update(this->graph(),spins_,this->beta_,this->random_01, - this->couplings_,this->spinfactor_,this->h_); + this->couplings_,this->spinfactor_,this->g_,this->h_); update_info_type up_info; up_info.clustersize = this->num_sites(); up_info.m2_ = 0;
On 7 Mar 2012, at 07:43, daniel.aravena@antares.qi.ub.edu wrote:
Hi Matthias,
I tried the new patch and it works correctly, the only thing I find strange is that the "Magnetization along field" vs. magnetic field plot doesn't look like dependent on the g factor but the energies are dependent on g, so i would expect a higher magnetization for higher g at the same field when far from saturation while i get the same magnetization regardless of the value of g (i tried g=1.0 and 2.0 and h from 0 to 11).
Thanks Kind Regards Daniel
Hi Daniel,
Can you please try to apply the following patch, and turn the quantum convention off. It will not give sensible results in magnetic fields.
Matthias
Index: applications/mc/spins/abstractspinsim.h =================================================================== --- applications/mc/spins/abstractspinsim.h (revision 6011) +++ applications/mc/spins/abstractspinsim.h (working copy) @@ -136,6 +136,11 @@ else use_error_limit = false;
+ if (has_magnetic_field_ && quantum_convention_) + boost::throw_exception( + std::runtime_error("Cannot use quantum convention with nonzero magnetic field .")); + + if (parms["UPDATE"]=="cluster" && has_magnetic_field_) {boost::throw_exception(std::runtime_error("Illegal update type " + std::string(parms["UPDATE"]) @@ -159,7 +164,7 @@ spinvalue = quantum_convention_ ? 0.5 : 1.; else spinvalue = alps::evaluate<double>(parms[spin],parms); - spinfactor_[*si]=std::sqrt(spinvalue*(spinvalue+(quantum_convention_ ? 1. : 0.)))*g_; + spinfactor_[*si]=std::sqrt(spinvalue*(spinvalue+(quantum_convention_ ? 1. : 0.)));
// set the values for the diagonal matrix terms site_type_visited[type]=true; @@ -176,7 +181,7 @@ selfinteraction_[*si].setMatrix(diag_string,parms); general_case_=true; } - selfinteraction_[*si] = selfinteraction_[*si]*spinvalue*(spinvalue+(quantum_convention_ ? 1. : 0.)); + selfinteraction_[*si] = selfinteraction_[*si]*spinvalue*spinvalue;
if (!selfinteraction_[*si].is_symmetric()) boost::throw_exception(std::runtime_error("Invalid matrix for D")); Index: applications/mc/spins/spinsim.h =================================================================== --- applications/mc/spins/spinsim.h (revision 6011) +++ applications/mc/spins/spinsim.h (working copy) @@ -246,14 +246,14 @@ m_h = this->spinfactor_[*si]*spins_[*si].mag_h(this->h_normalized);
if (this->has_magnetic_field_) - en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_;
double par=this->parity(*si); ++si;
for (; si !=this->sites().second ; ++si) { if (this->has_magnetic_field_) - en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_;
if (this->general_case_) en+=onsite_energy(spins_[*si],this->selfinteraction_[*si]); @@ -321,7 +321,6 @@ if (this->print_sweeps_) { int good_sweeps = this->sweeps_done_-this->thermalization_sweeps_; if (good_sweeps > 0 && (good_sweeps % this->print_sweeps_ ==0)) { - std::cout << "Dumping configuration: \n"; for (int i=0;i<this->num_sites();++i) { std::cout << i << " "; std::vector<double> coord = this->coordinate(i); @@ -350,7 +349,7 @@ bt_energy.resize(bt+1); bt_count.resize(bt+1); } - double be = bond_energy(spins_[this->source(*bi)],spins_[this->target(*bi)], + double be = this->spinfactor_[this->source(*bi)]*this->spinfactor_[this->target(*bi)]*bond_energy(spins_[this->source(*bi)],spins_[this->target(*bi)], this->couplings_[*bi]);
en -= be; @@ -374,13 +373,16 @@ mst *= this->spinfactor_[*si]; double m_h = 0.; if (this->has_magnetic_field_) + { m_h = this->spinfactor_[*si]*spins_[*si].mag_h(this->h_normalized); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_; + } double par=this->parity(*si); ++si;
for (; si !=this->sites().second ; ++si) { if (this->has_magnetic_field_) - en+=site_energy(spins_[*si],this->h_); + en+=this->spinfactor_[*si]*site_energy(spins_[*si],this->h_)*this->g_; if (this->general_case_) en+=onsite_energy(spins_[*si],this->selfinteraction_[*si]);
On Mar 1, 2012, at 10:39 PM, daniel.aravena wrote:
ok, here it is the parameter file for S=1/2 and 2 sites:
LATTICE_LIBRARY=/home/g1daniel/ALPS_LATT/latt02.xml MODEL="Heisenberg" LATTICE="2_graph" UPDATE="local" CONVENTION="quantum" THERMALIZATION=1000000 SWEEPS=10000000 S0=1/2 J0=0 g=1.0 T=1.0 {h=0} {h=1} {h=10} {h=100} {h=1000} {h=10000} {h=100000} {h=1000000} {h=10000000} {h=100000000} {h=1000000000} {h=10000000000}
the last field (h=10000000000) gives -1e+10 for the energy
Kind regards Daniel
On Thu, 01 Mar 2012 22:31:43 +0100, Matthias Troyer wrote:
No, I'm sorry - I checked it and we actually do include S in the field term as we should. TO look into this Iw ould like to ask you to send me your complete parameter file for one of the cases, e.g. for
>> Spin Number_of_sites Energy >> 1/2 2 -1e+10
On 1 Mar 2012, at 22:14, daniel.aravena wrote:
perfect, but, how can you handle the field dependence of a system composed of sites of different spins assigning just a single field
Matthias
participants (3)
-
daniel.aravena
-
daniel.aravena＠antares.qi.ub.edu
-
Matthias Troyer