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 > >