Let you have the Hamiltonian "-t(i,j)*fermion_hop(i,j) + other terms" defined on any predefined in "lattices.xml" lattice, where "t" is some matrix, whose all elements are a priori different. The straightforward way to define this model is to define the graph with all its edges having different types. This solution looks fine for several defective bonds. But if all bonds are different (e.g. you want to optimize the geometry of kink-like solution of Hubbard-Peierls model), does more elegant way exist on application level? In any case it is worthwhile to place more comprehensive answer to this kind of questions in "Tutorials and Documentation/Definition of models" page. Kind regards, Andriy Zhugayevych
On Sep 11, 2009, at 6:36 PM, Andriy Zhugayevych wrote:
Let you have the Hamiltonian "-t(i,j)*fermion_hop(i,j) + other terms" defined on any predefined in "lattices.xml" lattice, where "t" is some matrix, whose all elements are a priori different. The straightforward way to define this model is to define the graph with all its edges having different types. This solution looks fine for several defective bonds. But if all bonds are different (e.g. you want to optimize the geometry of kink-like solution of Hubbard- Peierls model), does more elegant way exist on application level? In any case it is worthwhile to place more comprehensive answer to this kind of questions in "Tutorials and Documentation/Definition of models" page. Kind regards, Andriy Zhugayevych
Hi Andriy,
The straightforward way is indeed to define that graph and you can easily write a program that defines such a graph, and assigns different bond types. If you decide to write such a program or script then please consider submitting it to ALPS.
If there is a simple equation describing the t(i,j) then you can use that, as we use it for an inhomogeneous potential.
Matthias
comp-phys-alps-users@lists.phys.ethz.ch