This document discusses improved variational methods for modeling ferromagnetism in Hubbard models, specifically for square lattices. It notes that the Stoner-Kanamori Ansatz is a simple trial ground state but that more flexible states can provide better bounds. One such improved state allows electrons to jump away from the vicinity of flipped spins, reducing the critical density. Numerical work on large finite systems showed room for improvement over the Stoner-Kanamori Ansatz results.
This document discusses improved variational methods for modeling ferromagnetism in Hubbard models, specifically for square lattices. It notes that the Stoner-Kanamori Ansatz is a simple trial ground state but that more flexible states can provide better bounds. One such improved state allows electrons to jump away from the vicinity of flipped spins, reducing the critical density. Numerical work on large finite systems showed room for improvement over the Stoner-Kanamori Ansatz results.
This document discusses improved variational methods for modeling ferromagnetism in Hubbard models, specifically for square lattices. It notes that the Stoner-Kanamori Ansatz is a simple trial ground state but that more flexible states can provide better bounds. One such improved state allows electrons to jump away from the vicinity of flipped spins, reducing the critical density. Numerical work on large finite systems showed room for improvement over the Stoner-Kanamori Ansatz results.
The Case of the Square Lattice The SKA Ansatz is the simplest of a large family of single-spin-flip trial ground states. These were usually tested by applying them to the square lattice as the simplest D > 1 model which may become ferromagnetic. Numerical work on finite systems as large as 100x100 gave ncrfi: 0.71 [433]; comparing this to the SKA value n,, x 0.51 shows that there is much room (and need) for improvement. However, impressive as these results are, numerical data for finite systems cannot give strict bounds which would be valid in the thermodynamic limit, thus it is useful to trace the development of analytic approaches which are valid for arbitrarily large systems, and at the same time give a pictorial idea of the features of the wave function. Improved bounds can be obtained from more flexible trial states whose nature has to be guessed [46, 154, 4481. For instance, one can argue that projecting out double occupation in (8.16) creates a dras- tic disturbance in the ?-electron sea at the site of the $-electron, and this makes the ?-electron kinetic energy unduly large. A better state can be obtained by letting the ?-electrons jump away (with variable amplitudes) from the vicinity of the $-electron. The U = 00, nearest- neighbour version of such a Basile-Elser-type state is [46]
(8.26)
where the m are nearest neighbours of j, and CY is a variational param-
eter. What we did is the following: removing the topmost ?-electron from the Fermi sea, we created a $-spin electron at the band bottom with wavevector q. Expanding the $-spin state in a Wannier basis, the electron is created at the site j with phase factor eid. If there is also an T-spin electron at j, we let it jump with amplitude CY to one of the neighbouring sites; if not, we do nothing. Minimizing the ground state energy with respect to a , one finds that IC? is reduced in comparison to its SKA value, while ICJ is left essentially unchanged. The procedure leads to a substantial increase of the critical density, to ncr M 0.585, which shows that the local polarization around the flipped spin is in- deed an important effect. Allowing hopping to distances up to m,