# The = 5 2 fractional quantum Hall e PHYS598PTD A.J.Leggett 2016 Lecture 20 The = 5=2 fractional...

date post

16-Feb-2020Category

## Documents

view

1download

0

Embed Size (px)

### Transcript of The = 5 2 fractional quantum Hall e PHYS598PTD A.J.Leggett 2016 Lecture 20 The = 5=2 fractional...

PHYS598PTD A.J.Leggett 2016 Lecture 20 The ν = 5/2 fractional quantum Hall effect 1

The ν = 5/2 fractional quantum Hall effect 1

As we have seen, the vast majority of quantum Hall plateaux observed experimentally, whether in Si MOSFET’s, GaAlAs-GaAs heterostructures or graphene, occur either at integral values of the filling factor ν ≡ Ne/Nφ (Ne ≡ number of electrons, Nφ ≡ number of flux quanta) (IQHE) or at rational fractions ν = p/q with q an odd integer (FQHE). Within the composite-fermion picture, the FQHE is explained by setting p = k, q = 2nk±1 and viewing the FQHE as derived from the IQHE with ν = k by attaching to each electron 2n imaginary “flux quanta”; or, what is equivalent, multiplying the IQHE wave function by (zi − zj)2n. However, as we have seen in lecture 19, in the case of ν = 1/2 the same procedure yields the conclusion that the “parent” IQHE state should correspond to ν = ∞, i.e. it should be a Fermi liquid. Moreover, as we will see in the next lecture, experiments, in particular on magnetic focusing, seem consistent with this point of view.

Prima facie, one would expect the states which occur at ν = n + 1/2 in traditional QHE systems such as GaAlAs-GaAs heterostructures2 to differ from that at ν = 1/2 only by having n Landau levels completely filled; the situation in the (n + 1)-th LL should be similar to that in the LLL for ν = 1/2. IT was therefore a considerable surprise when it was discovered in 1987 that a QH plateau occurs, with all the standard characteristics, at ν = 5/2. This plateau seems to be quite robust, with a gap ∆ which in the highest-mobility samples reaches ∼ 500 mK; at temperatures � ∆ the longitudinal resistance vanishes within exponential accuracy, and the Hall conductance appears to be quantized at (5/2) e2/h to high accuracy, thereby excluding the possibility that the plateau is a case of the standard FQHE with ν = 32/13 or 33/13. So it really does look like a “proper” FQHE state.

Before examining the ν = 5/2 state in detail, let’s digress for a moment to survey the general behavior of a 2D electron gas in a strong magnetic field as a function of the filling factor ν ≡ nh/eB (n ≡ areal density). We recall that in a simple case such as GaAs heterostructures where there are no complications due to valley degeneracy etc., the highest Landau level occupied (call it N) is the integer part of ν/2; thus, in particular, occupation of the LLL (only) corresponds to 0 < ν 6 2, while N = 1 corresponds to the range 2 < ν 6 4.

It turns out that within the LLL, i.e. for ν 6 2, the system behaves much as predicted by the theory of lectures 16-20; in particular, plateaux with Σxy = νe

2/h are always centered around the corresponding odd-denominator filling fractions ν, a “Fermi- liquid-like” state occurs at the even-denominator filling factor ν = 1/2 (and probably also at ν = 3/2, 3/4 and 1/4, though these cases have been less investigated). Moreover, when the system is subjected to a tilted magnetic field, the only effect seems to be to replace the total field by its component perpendicular to the plane of the 2D electron

1A comprehensive survey of the experimental aspects is given by R. Willett, Reps. Prog. Phys. 76, 076501 (2013)

2In single-layer graphene, it is believed that a QH plateau would occur at ν = n + 1/2 if both the spin and valley degeneracies were split, but this has to do with the special nature of the Dirac spectrum and is best regarded as a variant of the IQHE.

PHYS598PTD A.J.Leggett 2016 Lecture 20 The ν = 5/2 fractional quantum Hall effect 2

gas, indicating that any orbital3 effects of the parallel component are negligible. Things are very different in the range ν > 2(N > 1): see Willett, ref cit., section

3.4 In the first place, one finds so-called “reentrant” behavior, i.e. there are plateaux corresponding to integral values of Σxy/(e

2/h), but occurring around nonintegral values of the filling factor ν (see Willett, ref. cit., fig. 5); these are usually understood as indicating some kind of phase separation, e.g. into a “striped” phase. A second unexpected feature which appears to be restricted to the range ν > 4(N > 2) is that even in a perpendicular field, the (in-plane) resistivity close to half-filling (ν ∼= n+ 1/2) shows very strong anisotropy, by a factor up to ∼3,000; see Willett, ref. cit. fig. 4. This effect, while it is very sensitive to a parallel component of the magnetic field5, which in certain circumstances can actually reverse the axis of the anisotropy, appears to be intrinsic to the sample in zero parallel field – a puzzling observation, since GaAs has the cubic ZnS structure and should ideally be completely isotropic in the xy-plane; it strongly suggests the formation of some kind of “stripes” whose orientation is fixed by small symmetry-breaking effects.

In retrospect it is perhaps not so surprising that the behavior for N > 1) (and particularly for N > 2) is more complicated than that of the LLL: In the latter case, if we adopt the symmetric gauge, the radial dependence of the single-electron wave functions has a simple Gaussian form, while for N > 1 it is a more complicated Hermite polynomial with zeroes; thus the Coulomb interaction (whose Fock term, we recall, is attractive for parallel spins) may be able to organize rearrangements which are not available in the LLL. Indeed, theorectical work predating the experiments anticipated such rearrangements.

While for N > 2(ν > 4) no standard FQHE has to date been seen, the second LL (N = 1, 2 < ν 6 4) shows a complicated pattern: along with the “reentrant” quantum Hall effect described above, there appear several plateaux which appear prima facie to have the “standard” FQHE structure, not only as observed at ν = 5/2, but also at 2 + n/3 and 2 + n/5 (see Willett, ref. cit., fig 5). It appears that the second LL is in some sense a “battleground” between the traditional “composite-fermion” behavior of the LLL and the Coulomb effects which appear to dominate for N > 2.

With this in mind, let us now focus specifically on the ν = 5/2 state. We first review some of its basic properties:

1. Robustness: both plateau in RH and the zero of Rxx extend over a range ∼ 0.1 in ν. (in the highest-mobility samples)6

2. Excitation gap ∆: this is measured by fitting Rxx to Rxx ∼ const. exp−∆/T . ∆ appears to be a strong (∼ exponential) func- tion of disorder: the highest measured value to date ∼ 0.45 K, and extrapolation to zero disorder (µ→∞) gives ∆ ∼ 0.6K (∼ 0.006Vc) (Vc = e

2/4π��0lM) (Note: Vc is the natural unit to measure qp gap, as ∆qp ≡ 0 for the FQHE if Vc is neglected). 3. Magnetic field dependence: If B⊥ (hence ν) is held constant, and B‖ is varied, ∆qp

3Of course the Zeeman splitting remains proportional to the total field. 4It should be noted that much of the behavior described occurs only at T . 50mK, so that the data

is quite recent. 5Lilly, et al., PRL 83, 824 (1999) 6See fig. 1 of Xia et al., PRL 93, 176809 (2004).

PHYS598PTD A.J.Leggett 2016 Lecture 20 The ν = 5/2 fractional quantum Hall effect 3

decreases linearly7 with B‖, extrapolating to 0 at B‖ = 1.5−2.5 T. In the same geometry, ν = 7/2 FQHE shows similar behavior, but for ν = 7/3 (“Laughlin” state) ∆qp increases with B‖. 4. At least to date, the only system in which the ν = 5/2 QH plateau has been seen

is GaAs heterostructures: here typical parameters are ns ∼ 1 − 3 × 1011 cm−2 (so B ∼ 2− 6 T), µ ∼ 3× 107 cm2/V sec, T ∼ 5− 100 mK.

In trying to understand what is going on at ν = 5/2 it seems very natural to make (as we have above) the default assumption that the LLL is filled both for ↑ and ↓ spin states; then prima facie the behavior in the n = 1, ↑ spin LL should be identical to that in the n = 0, ↑ spin LL for ν = 1/2. However, this need not necessarily be the case, because as already observed in view of the different behavior of the wave functions for n = 0 and n = 1 the relevant matrix elements of the Coulomb interaction could be appreciably different in the two LL’s. An immediate question which arises then is: given that the LLL is completely filled for both spins, i.e. unpolarized, is the n = 1 LL spin-polarized or not? As we shall see below, the answer to this question is crucial to the identification of the nature of the wave function and thus to the possibility of using the ν = 5/2 FQHE for TQC; unfortunately, it has not so far proved possible to determine the answer experimentally. Originally, it was found that the plateau is suppressed by a substantial magnetic field component parallel to the plane, and the most obvious explanation is that the relevant state is a spin singlet (hence energetically disadvantaged by the Zeeman field). However, subsequent experiments, combined with numerical theory, tend to suggest that the effect is actually of orbital origin; because the parallel component of the magnetic field affects the motion perpendicular to the plane, it can change the relevant Coulomb matrix elements. For the moment, let us assume that the ν = 5/2 state seen in GaAs heterostructures is in fact fully spin polarized (we will return to this question below) and ask what is its nature? In particular, why does a QH system at ν = 5/2 not behave, as it seems to at ν = 1/2, as a “disguised” Fermi liquid (of composite fermions)?

A relevant question is: What do we know about the possible instabilities of a Fermi liquid? There are of course a great many, but most of them, such as crystallization, tend to occur either not at all or at tempera

Recommended

*View more*