Tag Archives: Toy Model

Excitonic Insulator

The state of matter dubbed the excitonic insulator was first qualitatively discussed by Mott, Keldysh and Kopaev, and others and then expanded upon more systematically by Jerome, Rice and Kohn.

The excitonic insulating state can be considered from two normal states (pictured below). Either the system must be a small-gap semiconductor or a small indirect overlap semimetal. In fact, Mott had first considered the semimetallic limit, while Kohn and others had considered the semiconducting limit.


Intuitively, one can consider the following two heuristic arguments from the different limits, as presented in the article by Rossnagel, which was cited in the previous post:

  1. Semiconducting limit: If one can somehow reduce the band gap energy, E_G, then at some point, the binding energy to form an exciton, E_B, will exceed E_G, and the system will unstable to the spontaneous formation excitons.
  2. Semimetallic limit: In this case, one considers screening effects. If one decreases the band overlap, a characteristic energy, E_1, will be reached such that particle-hole pairs will be insufficiently screened, leading to a localization of the charge carriers.

Therefore, in the regime of E_1E_G <E_B, the excitonic insulator state is expected. Properties of the excitonic insulator state are presented pedagogically in a Les Houches lecture by Kohn in this book, which is very difficult to find!

In a solid state context, it has been difficult to establish whether the excitonic insulator state has been realized because a lattice distortion is expected to accompany the transition to the excitonic insulator ground state. Therefore, it is difficult to isolate the driving mechanism behind the transition (this difficulty will be familiar to those who study high T-c superconductivity!).

There are a few materials suspected to possess excitonic insulator ground states in a solid state setting: 1T-TiSe_2, Ta_2NiSe_5 and TmSe_{0.45}Te_{0.55}. In my personal opinion, the case for 1T-TiSe_2 is probably the strongest purely because there have been far more experiments on this material than the other candidate materials.

Though this state of matter was considered almost 50 years ago, it still remains relevant today. As Landau once said,

Unfortunately, everything that is new is not interesting, and everything which is interesting, is not new.

Plasmons, the Coulomb Interaction and a Gap

In a famous 1962 paper entitled Plasmons, Gauge Invariance and Mass (pdf!), P.W. Anderson described the relationship between the gap in the plasmon spectrum and the idea of spontaneous symmetry breaking. It is an interesting historical note that Higgs cites Anderson’s paper in his landmark paper concerning the Higgs mechanism.

While there are many different formulations, including Anderson’s, of why the plasmon is gapped at zero momentum in a 3D solid, they all rely on one crucial element: the long-range nature of the Coulomb interaction (i.e. the electrons are charged particles). Of these formulations, I prefer one “cartoon-y” explanation which captures the essential physics well.

Before continuing, let me stress that it is quite unusual for a fluid medium (such as the electrons in a metal) to possess no zero frequency excitations at long wavelengths. For instance,  the dispersion relation for surface gravity waves on water (pdf!) is:

\omega^2(k)=gk \tanh kh.

Now, in 3D and in the long-wavelength limit, the plasmon sets up opposite charges on the surfaces of the solid as pictured below:

The long-wavelength plasmon therefore sets up the same electric field as in a capacitor. The electric field for a capacitor is \textbf{E} = \frac{\sigma\hat{x}}{\epsilon_0}. This expression is surprisingly independent of the distance separating the surfaces of the solid. Therefore, it takes a finite amount of energy to set up this electric field, even in the limit of infinite distance. This finite energy results in the gapping of the plasmon.

This cartoon can be extended further to 2D and 1D solids. In the 2D case, the electric field for the 1D “lines of charge” bounding the solid falls off like \textbf{E}\sim\frac{1}{\textbf{r}}. Therefore, in the infinite distance limit, it takes no energy to create this electric field and the plasmon is not gapped at \textbf{q}=0. Similarly, for the 1D case, the electric field from the points bounding the solid falls of as \frac{1}{\textbf{r}^2}, and the plasmon is again gapless.

This reasoning can be applied further to the phenomenon known as LO-TO splitting in a polar solid. Here, the longitudinal optical phonon (LO) and the transverse optical phonon (TO) branches are non-degenerate down to the very lowest (but non-zero!) momenta. Group theory predicts these modes to be degenerate at \textbf{q}=0 for the zincblende crystal structure of typical semiconducting compounds. Below is the phonon dispersion for GaAs demonstrating this phenomenon:

Again, the splitting occurs due to the long-ranged nature of the Coulomb interaction. In this case, however, it is the polar ionic degree of freedom that sets up the electric field as opposed to the electronic degrees of freedom. Using the same reasoning as above, one would predict that the LO-TO splitting would disappear in the 2D limit, and a quick check in the literature suggests this to be the case as reported in this paper about mono-layer Boron Nitride.

I very much appreciate toy models such as this that give one enough physical intuition to be able to predict the outcome of an experiment. It has its (very obvious!) limitations, but is valuable nonetheless.

Insights from the Cooper Problem

In the lead-up to the full formulation of the Bardeen-Cooper-Schrieffer Theory of superconductivity (BCS theory), Leon Cooper published a paper entitled Bound Electron Pairs in a Degenerate Fermi Gas (pdf) (referred to colloquially as “The Cooper Problem”). Its utility is not always recognized, but has been stressed by Leggett in his book Quantum Liquids where he says:

It seems not always to be appreciated how useful this “toy” model and simple generalizations of it can be, in particular in giving one a physical feel for which kinds of effects are likely to inhibit (or not) the formation of the superconducting state.

Having solved the Cooper problem in many instances, I tend to agree with Leggett. The Cooper problem, generalized to include the addition of a Zeeman field, shows the detrimental effect of a magnetic field on a Cooper pair. When generalized to include a finite center-of-mass momentum, pair-breaking is again induced.

However, it can also give one an intuition concerning effects that do not inhibit superconductivity. Such a case is where the Zeeman field and the center-of-mass momentum effects “cancel out” to yield a superconducting state (known as the FFLO state). Also, one can realize Anderson’s Theorem (pdf), which states that Cooper pairs are formed from time-reversed partners (as opposed to strictly and -k pairs), a result that is important in understanding the indifference of conventional superconductors to non-magnetic impurities.

Another instance of its usefulness is in understanding the “decoupling” of higher-order pairing (e.g. p-wave, d-wave, etc.). This is discussed in the first chapter of Introduction to Unconventional Superconductivity by Mineev and Somakhin. After solving the problem, one gets a similar result for the binding energy, \Delta to that of the Cooper Problem:

\Delta_l = -2\epsilon_l \exp(-2/N(0)V_l)

where l is the index labeling the symmetry channel (e.g. l=2 means d-wave) and \epsilon_l denotes an energy cutoff. The result demonstrates that a superconducting state will result when any of the of the angular momentum channels is unstable (at least for a spherical Fermi Surface).

The Cooper Problem: An instructive, easy-to-solve, insightful toy model.