Tuesday, March 23, 2010

Deconstructing Anderson's radical proposal?

In 1987 Anderson made the radical proposal concerning doping Mott insulators:
"The preexisting magnetic singlet pairs of the insulating state become charged superconducting pairs when the insulator is doped sufficiently strongly."
The material SrCu2(BO3)2 has been argued to be a Mott insulator on the Shastry-Sutherland lattice (see Figure below). In the corresponding Heisenberg model there is an exchange interaction J along all vertical and horizontal bonds and a diagonal interaction J' along every second plaquette. It can be shown that for J'/J > 1.44 +/- 0.02 that the exact ground state is a product of singlets along the diagonals.

Liu et al. studied the corresponding t-J model (including three site hopping terms) away from half-filling using a projected BCS wave function. They considered the model with t'=+/- 1.25 t and J=0.3t. They found the following (summarised in the Figure below).
(i) There is significant particle-hole asymmetry. This is because one sign of t' corresponds to electron doping and the other to hole doping while t does not change.

(ii) Hole doping produces d-wave superconductivity. But, this is NOT the result of delocalisation of the pre-existing singlets in the Mott insulator since they were along the diagonals.

(iii) Electron doping does not produce superconductivity, but only a correlated metal with singlet pairing along the diagonal, as in the parent Mott insulator.

(iv) The hole-doped superconducting state co-exists with plaquette bond order where all the nearest neighbour spins have antiferromagnetic correlations. Thus the spin correlations are qualitatively different from in the parent Mott insulator.

I feel this paper has not received the attention that it deserves. It shows that the competition between superconductivity and antiferromagnetism and resonating valenc bonds that occurs when doping a frustrated Mott insulator is more subtle (and confusing) than suggested by Anderson's original conjecture.

On the other hand, one might argue that the parent Mott insulator is very different from the cuprates and organics because there are NO resonating valence bonds in the parent insulator.

Aside: I am confused about (i). I thought that when one goes from particles to holes one changes the sign of both t and t'. But, is this difference because they include the three site hopping term in the t-J model?

No comments:

Post a Comment