VOLUME 50, +UMBER

PHYSICAL REVIEW LETTERS

I8

Anomalous

2 MAv 1983

Quantum Hall Effect: An Incompressible Quantum

with Fractionally Charged Excitations

H. B. Laughlin
Iasorence Livermore National Laboratory, University of California, Jivermore,
(Received 22 February 1983)

Fluid

California 94550

This Letter presents variational ground-state and excited-state wave functions which
describe the condensation of a two-dimensional electron gas into a new state of matter.
PACS numbers: 71.45.Nt, 72.20.My, 73.40. Lq

'" effect,

sistent with all the experimental facts and explain the effect. The ground state is a new state
of matter, a quantum fluid the elementary excitations of which, the quasielectrons and quasiholes, are fractionally charged. I have verified
the correctness of these wave functions for the
case of small numbers of electrons, where direct
numerical diagonalization of the many-body Hamiltonian is possible. I predict the existence of a
sequence of these ground states, decreasing in
density and terminating in a Wigner crystal.
Let us consider a two-dimensional electron
gas in the x-y plane subjected to a magnetic field
H, in the z direction. I adopt a symmetric gauge
vector potential A = ,'H, [xy-yx] and write the
eigenstates of the ideal single-body Hamiltonian
j (h/i)V (e/c)A ' in the manner
H, p

recently discovered by Tsui,

Stormer, and Gossard, ' results from the condensation of the two-dimensional electron gas in a
GaAs-GaAl, As heterostructure into a new type
of collective ground state. Important experimental facts are the following: (1) The electrons con' of a full Landau
dense at a particular density, ,
level. (2) They are capable of carrying electric
current with little or no resistive loss and have a
Hall conductance of ,'e'/h. (3) Small deviations
of the electron density do not affect either conductivity, but large ones do. (4) Condensation
occurs at a temperature of 1.0 K in a magnetic
field of 150 kG. (5) The effect occurs in some
samples but not in others. The purpose of this
Letter is to report variational ground-state and
excited-state wave functions that I feel are conThe

",

I
I

jm, n) =(2

'""~m!n!) 'i'exp[-;(x'+y')]

with the cyclotron energy he,

the magnetic length a, =(A/m~, )'
set to 1. We have

+i i exp [ ggx'+y')]
m

Bg

= h(equi, /mc) and

e,

m, n) = (n+

')

m,

'=(Ac/eH, )'

n).

Bx

Bg

'

B$

states of the lowest Landau level as

Im) =(2 +~em!) "i g exp( jg j )

(2)

z=x+ iy. m) is an eigenstate of angular

momentum with eigenvalue m. The many-body
Hamiltonian is

The manifold of states with energy n+ & constitutes the nth Landau level. I abbreviate the

where

jI = g { (fi /i )V, (e /c)A, ' + V(z, )j + Z e'/ ~, - z, I,

j

and 4 run over the N particles and V is a

where
potential generated by a uniform neutralizing
background.
I showed in a previous paper' that the ,' effect
could be understood in terms of the states in the
lowest I andau level solely. With e'/a, s h
the situation in the experiment, quantization of
interelectronic spacing follows from quantization
of angular momentum: The only wave functions
composed of states in the lowest Landau level
which describe orbiting with angular momentum

m about the

&=(~
My
A'

center of mass are of the form

-~.)

(~+~,)"exp[- l(j~ j'+ jz. j')] (

present theory generalizes this observation to

particles.

I write the ground state as a product of Jastrow

functions in the manner

0=
and

II f(z, z, )jexp(- 'Q,

lz, l'),
minimize the energy with respect to f.

1983 The American Physical Society

(6)

We

1395

PHYSICAL REVIEW LETTERS

VOLUME 50, +UMBER 18

observe that the condition that the electrons lie

in the lowest Landau level is tha. t f(z) be polynomial in z. The antisymmetry of g requires that
be odd. Conservation of angular momentum requires that II, ,f(z, z~) be a homogeneous polynomial of degree M, where M is the total angular
momentum.
We have, therefore, f(z) =z, with
m odd. To determine which m minimizes the energy, I write

14.1' = I{II,(z,-z, ) kexp(- -'Z, lz, l')I'

TABLE
functions

0. 999 46
0. 994 68
0.994 76
0.995 73
0.996 52

0.996 73
0.991 95
0.992 95
0.994 37
0.995 42
0.996 15

0.999 66
0.999 39
0.99981
0.999 99
0.999 96
0.999 85

Q = m,

2''=
~

'.

U,= ~

oo

[g(r) I ] rdr.

In the limit of large

I,

U, is

0.997 08

13

interacting via logarithmic potentials and embedded in a uniform neutralizing

background of charge density o =(27)a,') '. This
is the classical one-component plasma (OCP), a
system which has been studied in great detail.
Monte Carlo calculations' have indicated that the
OCP is a hexagonal crystal when the dimension2m is greater
less plasma parameter I'=
~' describes a
than 140 and a fluid otherwise.
system uniformly expanded to a density of o
=m '(27)a, ')
lt minimizes the energy when o
equals the charge density generating V.
In Table I, I list the projection of g for three
particles onto the lowest-energy eigenstate of
angular momentum 3m calculated numerically.
These are all nearly 1. This supports my assertion that a wave function of the form of Eq. (6) has
adequate variational freedom. I have done a similar calculation for four particles with Coulombic
repulsions and find projections of 0.979 and 0.947
for the m = 3 and m = 5 states.
has a total energy per particle which for
small m is more negative than that of a chargedensity wave (CDW). ' It is given in terms of the
radial distribution function g(r) of the OCP by

charge

(9)

within a few percent by the ion disk energy:

d2y+~
j+j
2

gm

U Ot

= (4/37)

0 2
2

e2

j+

dy
j

2'

(10)

1)2e'/R,

where the integration domain is a disk of radius

8 = (pro ) "' '. At I' = 2 we have the exact result'
')T'/'e'/
that g(r) =1 exp[- (r/fI)'], giving U, = ,
At m=3 and m=5 I have reproduced the Monte
Carlo g(r) of Caillol et al. ' using the modified
hypernetted chain technique described by them.
I obtain U, =(-0.4156+0.0012)e'/a, and U,(5)=
(-0.3340 +0.0028)e'/a, The corresponding values
for the charge-density wave4 are 0. 389e'/ao and
0. 322e'/ao. U, is a smooth function of I I
interpolate it crudely in the manner

U tot

(m)

814
0. 230
=-'0.

064 -1

This interpol. ation converges to the CD% energy

near rn = 10. The actual crystallization point cannot be determined from that of the OCP since the
CD% has a lower energy than the crystal defor m& 71
scribed by
I generate the elementary excitations of g by
piercing the fluid at z, with an infinitely thin
solenoid and passing through it a flux quantum
t) cp =- hc/e adiabatically.
The effect of this operation on the single-body wave functions is

(z-z, ) exp(-4lzl')-(z-z,
Let us take as approximate
these excited states

approximated

wave

(l)

3
5

ll

4 describes a system of N

exp(- x'/2)

where P= 1/m and 4 is a classical potential energy given by

(8)
lz, I'
identical particles of

is the lowest-energy eigenstate of angular momentum 3m calculated with V = 0 and an interelectronic potential of either 1/x, -In(x), or exp(-x'/
2).

=-Z, ,2m'lz, -z, I+ am',

I. Projection of variational three-body

in the manner (g C )/(((
) (4

(|)

x c ~)) (/'. 4

(7)

2 MAY 1983

"exp(--'. ~z~'). ()

representations

of

(13)

'4 "=&., ''4= "(--'Xl,

1396

(, l')In;(, I(n;, .(*, ,)"),

".

~o

(14)

VOLUME 50, +UMBER

PHYSICAL REVIEW LETTERS

18

for the quasihole and quasieleetron, respectively.

For four particles, I have projected these wave
functions onto the analogous ones computed numerically. I obtain 0. 998 for g, ' and 0. 994 for
I obtain 0. 982 for g, =fg, (z, -Z))g
which is P, with the center-of-mass motion re-

"

"

moved.

These excitations are particles of charge 1/m.

To see this let us write [ (+'oI' as e s ~, with
P=1/m and
O' = 4 2+, In~

z, zo(.

(15)

4'

describes an QCP interacting with a phantom

point charge at z, . The plasma will completely
screen this phantom by accumulating an equal and
opposite charge near zp. However, since the
plasma in reality consists of particles of charge
1 rather than charge rn, the real accumulated

'0
charge is 1/m. Similar reasoning applies to
if we approximate it as g, (z, z, ) 'P, gwhere
P, is a projection operator removing all configurations in which any electron is in the singlebody state (z- z, )'exp(- 4 Iz I'). The projection of
this approximate wave function onto g, 'o for four
particles is 0. 922. More generally, one observes
that far away from the solenoid, adiabatic addition of Ay moves the fluid rigidly by exactly one
state, per Eq. (12). The charge of the particles
is thus 1/m by the Schrieffer counting argument. '
The size of these particles is the distance over
which the QCP screens. Were the plasma weakly
coupled (I' & 2) this would be the Debye length AD=
a, /N. For the strongly coupled plasma, a better
estimate is the ion-disk radius associated with a
charge of 1/m: A=&2a, . From the size we can
estimate the energy required to make a particle.
The charge accumulated around the phantom in
the Debye-Huckel approximation is
t/r

where Ko is a modified Bessel function of the

second kind. The energy required to accumulate

it is
bp5p

e'

(16)

This estimate is an upper bound, since the plasma

is strongly coupled. To make a better estimate
let 5p = a inside the ion disk and zero outside, to
obtain

1e'

disk

3
2v2p m' a p

(17)

2 MAv 1983

For m--3, these estimates are 0.062e'/a, and

0.038e'/a, . This compares well with the value
0.033e'/ao estimated from the numerical fourpartiele solution in the manner
~ =-.'(E(q. -') E(e. 2E(y, )),

")

(18)

where E(g, ) denotes the eigenvalue of the numerical analog of P, . This expression averages the
electron and hole creation energies while subtracting off the error due to the absence of V.
I have performed two-component hypernetted
chain calculations for the energies of P, "o and
I obtain (0.022 a 0.002)e'/ao and (0.025
+ 0.005)e'/a,
If we assume a value e = 13 for
the dielectric constant of GaAs, we obtain 0.02e'/
cap ~4 K when Ho = 150 kG.
The energy to make a particle does not depend
on
long as its distance from the boundary
is greater than its size. Thus, as in the singlepartiele problem, the states are degenerate and
there is no kinetic energy. We can expand the
creation operator as a power series in z, :

zso

~, = j=0
gX, (,

~,

z, )z, '- .

These A. , are the elementary symmetric polynomials, ' the algebra of which is known to span
the set of symmetric functions. Since every antisymmetric function ean be written as a sym. metric function times y these operators and
their adjoints generate the entire state space.
It is thus appropriate to consider them X linearly independent particle creation operators.
The state described by g is incompressible
because compressing or expanding it is tantamount to injecting particles. If the area of the
system is reduced or increased by ~A the energy rises by &U = o ss M ~. Were this an elastic solid characterized by a bulk modulus B, we
'B(5A)'/A. Incompressibility
would have 5U= ,
causes the longitudinal collective excitation
roughly equivalent to a compressional sound
wave to be absent, or more precisely, to have
an energy - 4 in the long-wavelength limit. This
facilitates current conduction with no resistive
loss at zero temperature. Our prototype for
this behavior is full Landau level (m = 1) for
which this collective excitation occurs at R~, .
The response of this system to compressive
stresses is analogous to the response of a typeII superconductor to the application of a magnetic
field. The system first generates Hall currents
without compressing, and then at a critical
stress collapses by an area quantum ppl2~QO'
~

1397

VOLUME 50, +UMBER

18

PHYSICAL REVIEW LETTERS

and nucleates a particle. This, like a flux line,

is surrounded by a vortex of Hall current rotat-

ing in a sense opposite to that induced by the

stress.
The role of sample impurities and inhomogeneities in this theory is the same as that in my
theory of the ordinary quantum Hall effect. ' The
electron and hole bands, separated in the impurity-free case by a gap 2A, are broadened
into a continuum consisting of two bands of extended states separated by a band of localized
ones. Small variations of the electron density
move the Fermi level within this localized state
band as the extra quasiparticles become trapped
at impurity sites. The Hall conductance is (1/m)
& (e'/h) because it is related by gauge invariance
to the charge of the quasiparticles e* by o H, z
=e*e/h, whenever the Fermi level lies in a
localized state band. As in the ordinary quantum
Hall effect, disorder sufficient to localize all the
states destroys the effect. This occurs when the
collision time w in the sample in the absence of
a magnetic field becomes smaller than T &h/4.
I wish to thank H. DeWitt for calling my attention to the Monte Carlo work and D. Boercker

1398

2 MA+ 198$

for helpful discussions. I also wish to thank

P. A. Lee, D. Yoshioka, and B. I. Halperin for

helpful criticism. This work was performed

under the auspices of the U. S. Department of
Energy by Lawrence Livermore National Laboratory under Contract No. W-7405-Eng-48.

'D. C. Tsui, H. L. Stormer, and A. C. Qossard, Phys.

Rev. Lett. 48, 1559 (1982).
'R. B. Laughlin, Phys. Rev. B 27, 3383 (1983).
3J. M. Caillol, D. Levesque, J. J. leis, and J. P.
Hansen, J. Stat. Phys. 28, 325 (1982).
D. Yoshioka and H. Fukuyama, J. Phys. Soc. Jpn.
47, 394 (1979}; D. Yoshioka and P. A. Lee, Phys. Rev.
B 27, 4986 {1983), and private communication.
~B. Jancovici, Phys. Rev. Lett. 46, 386 {1981). -I=-2
corresponds to a full Landau level, for which the total
energy equals the Hartree-Fock energy P~/8 e2/ao. This correspondence may be viewed as the underlying
reason an exact solution at 1 = 2 exists.
6W. P. Su and J. R. Schrieffer, Phys. Rev. Lett. 46,
738 (1981).
S. Lang, Algebra (Addison-Wesley, Reading, Mass. ,

1965}, p. 132.
'R. B. Laughlin, Phys. Rev. B 23, 5632 (1981).