Dear Chun-Hsu,
The reason is that the SSE code can only deal with one particle type. The workaround is to define a lattice with two sites in the unit cell and two site types. One for the atomic excitations the other for the photonic ones. Each site type gets a different basis with its associated operators, one the atomic, the other the photonic ones. The mixed terms will then be bond-terms between the two sites. I can help you do that if you have problems.
Another important comment is that the SSE code works in the grand canonical ensemble. You cannot fix the total particle number but need to introduce a chemical potential.
Matthias
On 2 Jun 2009, at 01:42, Su Chun-Hsu wrote:
Dear Matthias,
Thank you for pointing out that obvious mistake. After reducing the maximum occupancy to 10, there is an error message "Offdiagonal site terms are not implemented in this SSE code" when using dirloop. It is unclear to me when atomic projection operators and bosonic operators are defined accordingly (attached). Please advise and thank you again.
Regards, Chun-Hsu
From: troyer@phys.ethz.ch To: comp-phys-alps-users@phys.ethz.ch Date: Sat, 30 May 2009 16:49:02 -0600 Subject: Re: [ALPS-users] Setting up polaritonic model in ALPS
On 29 May 2009, at 19:09, Su Chun-Hsu wrote:
Dear all, I'm new to using ALPS and not entirely confident with setting up
my
model, which didn't appear to work at the moment. I'm interested in polaritonic system -- evanescently coupled-cavity system, each cavity
containing a
two-level atom. I have set up my model in the attached xml, Sminus and Splus are atomic projection operators and b are photon annihilation operator. The total number of excitations (atomic and photonic) is
conserved.
When I try running it, there is an bad_alloc error. Will anyone be kindly give me some pointers? Much appreciated and thank you
very
much in advance.
Not knowing exactly what you have in mind I can only say that the model looks reasonable. You have to be very careful though as you
have
a large number of local states and already a 2x2 lattice without constraints has a Hilbert space of 202^4, which is far too large
to be
solved by exact diagonalization.
With your constraints the Hilbert space is smaller but there is
still
the problem that the ALPS program wants to build the full
Hamiltonian
matrix for two sites. The 2-site Hilbert space in your model is
202^2
= 40804 states. The matrix for two sites thus 40804 x 40804 = 1 664 966 416 elements large, which needs 12.5 GB memory. The allocation
of
this matrix fails and thus there is a bad_alloc exception. Such a model is to big to be studied by exact diagonalization. If you
reduce
the maximum occupancy t0 10 or 20 it will work
Matthias
Make ninemsn your homepage! Get the latest news, goss and sport <JCHmodel.xml><parmJCHSquare>