The document discusses competing magnetic instabilities that can occur in certain materials at low temperatures due to diverging susceptibilities. For half-filled bands with perfect nesting, an arbitrarily small interaction U causes an antiferromagnetic transition and the system becomes insulating. However, for a half-filled square lattice tight binding band, both the antiferromagnetic (Q) and ferromagnetic (q=0) susceptibilities diverge, so either could occur depending on which is stronger. Strictly at T=0, the susceptibilities are both infinite, but it can be shown that antiferromagnetism prevails at finite temperatures by taking the T→0 limit.
The document discusses competing magnetic instabilities that can occur in certain materials at low temperatures due to diverging susceptibilities. For half-filled bands with perfect nesting, an arbitrarily small interaction U causes an antiferromagnetic transition and the system becomes insulating. However, for a half-filled square lattice tight binding band, both the antiferromagnetic (Q) and ferromagnetic (q=0) susceptibilities diverge, so either could occur depending on which is stronger. Strictly at T=0, the susceptibilities are both infinite, but it can be shown that antiferromagnetism prevails at finite temperatures by taking the T→0 limit.
The document discusses competing magnetic instabilities that can occur in certain materials at low temperatures due to diverging susceptibilities. For half-filled bands with perfect nesting, an arbitrarily small interaction U causes an antiferromagnetic transition and the system becomes insulating. However, for a half-filled square lattice tight binding band, both the antiferromagnetic (Q) and ferromagnetic (q=0) susceptibilities diverge, so either could occur depending on which is stronger. Strictly at T=0, the susceptibilities are both infinite, but it can be shown that antiferromagnetism prevails at finite temperatures by taking the T→0 limit.
x(O)(Q)diverges logarithmically unless P ( E ) 3 0 as E + 0. It follows
that U2 = 0, i.e., for half-filled bands with the perfect nesting prop- erty, an arbitrarily small interaction U > 0 causes a transition to a two-sublattice antiferromagnetic state. We are going to see that the antiferromagnetic state is insulating. Thus in these circumstances, the metal-insulator transition occurs at U = 0. When making the above statement, we were taking it for granted that the q # Q susceptibilities are finite so that it is clear the q = Q ordering is the dominant instability. A curious case is presented by the half-filled, square lattice tight binding band. Here not only x ( Q ) diverges but, because of the logarithmic divergence of the density of states (4.31), x(q = 0) too. We have a case of competing instabilities: it appears that at an infinitesimal U > 0, the system might decide to go ei- ther ferromagnetic, or antiferromagnetic. Which instability is stronger? Strictly at T = 0, we should compare two infinite susceptibilities but coming from finite temperatures, we can easily convince ourselves that antiferromagnetism prevails (Problem 7.5). Though (7.47) often gives a good idea of what can be expected to happen, we should not forget that it was derived in a simple mean field theory, and it is certainly not exact. We are speaking about Hartree- Fock instabilities. Hartree-Fock theory is notorious for underestimating the stability of the non-ordered, symmetrical ground state. We should be especially cautious about ordering away from half-filling. It turns out, however, that the prediction of the U = 0 metal-insulator transi- tion for half-filled, perfectly nested bands is basically correct. Problem 7.4 Derive x(O)(q) for the one-dimensional tight binding band. Compare the divergence at q = 2 k to~ that obtained for free electrons. Problem 7.5 Competing instabilities: For the half-filled square lattice tight binding band at T = 0, x(O)(Q)and x ( O ) ( O ) diverge simultaneously. Decide which divergence is stronger by doing a calculation at finite temperatures, and taking the limit T + 0. (Hint: you can simplify the calculation by approxi- mating the Fermi distribution by a piecewise linear function). Compare the weak-coupling mean-field NEel temperatures for D = 2 di- mensions, and D # 2.