2 Charged Particle in a Magnetic Field

2.1 Classical Particle in B-eld

Want to remind ourselves how to write classical Hamiltonian describing
charged particle moving in E and B elds. I will sketch the more thorough
review given by, e.g. Gasirowicz in Ch. 16 or Peebles in Sec. 2-19. Griths
unfortunately does not treat this subject.
Lagrangian for charge q moving in an electromagnetic eld is (SI units)
L =
m r
2
2
+ q(A r ), (1)
where r is velocity, A is vector potential and scalar potential. Shouldnt
swallow this, but check that Euler-Lagrange eqns. L/r

= (d/dt)L/ r

produce classical Lorentz force eqns. we know & love from intro physics.
We can do this w/ straightforward but tedious algebra if we remember the
classical relations between the elds and the potentials:
B = A (2)
E =
A
t
. (3)
Eqns. of motion become (Lorentz force!)
m r = q(E + v B). (4)
The elds E and B are invariant under gauge tranformations of the
potentials A and :

/
=

t
(5)
A A
/
= A+ (6)
so the eqns. of motion (4)are manifestly gauge invariant because they
contain only elds, even though the Lagrangian L apparently does not.
1
Hamiltonian of system now dened as
H[p, r] = p r L, (7)
with canonical momentum
p =
L
r
= m r + qA. (8)
Replacing velocities with momenta yields
H =
1
2m
(p qA)
2
+ q (9)
Note: In QM have not yet worked out corresponding formalism for Lagrangian
well use dierent principle to derive Hamiltonian, which will turn out to
be exactly (9). Note we will be interpreting the canonical momentum p
as i h, following the prescription to replace the canonical momentum
in the classical theory by i h to preserve the canonical commutation
relations [x, p] = i h. To convince yourself this is really ok, see Griths,
prob. 4.59. See Peebles Ch. 2 for a more complete explanation.
2.2 Gauge Invariance in QM
Reminder: unitary transformations in QM
unitary operator

U =

U

=

U
1
.
If system described by states [), operators

Q, physically equivalent rep-
resentation given by
[
/
) =

U[), . . . ;

Q
/
=

U

Q

U
1
. (10)
Why? Because matrix elements of all operators identical:

/
[

Q
/
[
/
) = [

U
1
(

U

Q

U
1
)

U[)
= [

Q[) (11)
QM gauge transformations
2
Require that a QM gauge tranformation should take wave fctn and
potentials A, to physically equivalent set
/
, A
/
,
/
. Can accomplish
this by taking transformation
/
to be unitary, see above. Assume
form
[
/
) =

U[) = e
iq(r,t)/ h
[), (12)
obviously unitary. Note gauge function is space, time dependent =
called local gauge transformation. Now we want Hamiltonian itself, which
is observable, to be invariant wrt such transformations in the sense of
Eq. ??. Since potential energy is q, might guess H = p
2
/2m + e,
but this is obviously wrongnot gauge invariant, cant get correct eqns. of
motion.
Consider gauge-covariant momentum p qA. Note

/
[ p qAq]
/
= e
iq/ h
( p qA)
or
/
= e
iq/ h
e
iq/ h
(13)
so transforms as we would like under gauge transform, as

/
= p qA
/
. (14)
Write H =

T + q, require

T
/
= e
iq/ h

Te
iq/ h
(15)
plus we want

T to reduce to p
2
/2m when A = 0 so choose

T =
2
/2m,
or
H =
1
2m
(p qA)
2
+ q (16)
as in classical case. With this choice Schrodinger eqn.
H
/

/
= i h

/
t
(17)
will have invariant solution
/
physically equivalent to , soln. of H =
i h(/t). Check!
3
2.3 Bohm-Aharonov Eect
Suppose we have electron inside hollow conductor, or Faraday cage,
with battery which raises potential of cage and region inside, beginning at
t = 0 and ending at time t. Hamiltonian for electron is H = H
0
+ V (t),
with V (t) = e(t). Easy to see only result of varying potential will be
varying phase of electronic wave function:
(x, t) =
0
(x, t)e
iS/ h
, S =
_
t
0
V (t
/
)dt
/
(18)
where
0
is wave fctn. in absence of battery. Check:
i h

t
= i h
_
_

0
t
e
iS/ h
+
0
_
_
i
h
_
_
V (t)e
iS/ h
_
_
=
_
_
i h

0
t
+ V (t)
0
_
_
e
iS/ h
= (H
0
+ V (t))
0
e
iS/ h
= H (19)
where use was made of H
0

0
= i h

0
. But phase shift in single electrons
wave fctn. cant aect observables, since physical quantities bilinear in
and

, norm of e
iS/ h
is 1. This result might encourage sense of
complacency, since its exactly what we expect from classical E & M:
since is constant inside cage, E = 0, so no physical changes.
But in 1959 Aharanov and Bohm
1
looked at variation of above expt.
with two Faraday cages & two batteries, in which potentials on two
cages were dierent from one another. Naively, might expect that elec-
trons in two arms would experience two dierent phase shifts,
1

(e/ h)
_
t
0
dt
/

1
(t
/
), and
2
(e/ h)
_
t
0
dt
/

2
(t
/
). Wave fctn. would
then be =
1
0
e
i
1
+
2
0
e
i
2
. This phase dierence,
1

2
would then be observable because it would shift interference pattern at
screen as shown. (To see how comes in, calculate [[
2
.) Ob-
servable consequence of nonzero would then be found even though
region of space electrons travelled through has electric eld E = 0,
1
Phys. Rev. 115, 485 (1959)
4
i.e. electrons never subject to classical force! Worth noting that
obvious interpretationnamely that electromagnetic potentials have inde-
pendent signicance in QM, more fundamental than elds perhapsat
least consistent with observation that eect observed (interference pattern
for electrons) has no classical analogue.
Problem: cylinders in gure cant be true Faraday cages if electrons pass
through themmust be holes, elds could leak in at edges. . . Another, per-
haps more convincing, demonstration uses magnetic eects.
Pass 2 e

beams around long solenoid as shown, such that paths fully

enclose solenoid, interfere at screen. Note B ,= 0 only inside solenoid, but
A ,= 0 everywhere.
2
Might expect (see below) that eect is similar to
2
In useful gauge, a solenoidal vector potential can be written (cylindrical coordinates , z, )
A =
_
A
z
= A

= 0, A

= B
0
/2 < a
A
z
= A

= 0, A

= B
0
a
2
/(2) > a
(20)
5
electrostatic potential case, i.e. that =
1
+
2
, and each

acquires a
phase factor

0
e
iS

/ h
in the presence of the vector potential.
3
Then
a B-dependent shift in interference at screen may produce a measurable
eect of magnetic ux, even though e

s never pass through it! We may

say vector potential A is more fundamental than eld B in QM, or note
that this is simply another example of nonlocality in QM!
Substitute
0
e
iS/ h
into t-dependent S.-eqn.:
i h

t
=
1
2m
(i h+ eA)
2
(22)
to nd (assuming
0
satises zero-eld S.-eqn i h(
0
/t) =
h
2
2m

0
)
that satises the full S-eqn. if we take
4
S(r) = e
_
r
A(r
/
) dr
/
(23)
Easy to see qualitatively now that since A points in opposite directions
on opposite sides of solenoid, beams arrive with dierent phases S
1
/ h and
S
2
/ h for nonzero eld. Note that the phase dierence at a point P on the
screen will depend on the dierence of the phases accumulated along each
trajectory:
= S
1
(P)/ h S
2
/ h =
e
h
_
whole path
A(r
/
) dr
/
(24)
where whole path means along path 1 to P and back along path 2 to
electron gun. Stokess theorem then gives
e
h
_
whole path
A(r
/
) dr
/
=
e
h
_
area encl.
da A (25)
such that
B = A = z
1

(A

) =
_
B
0
< a
0 > a
(21)
3
It is important to realize that we are not talking necessarily about 2 electrons interfering with
each other, although we could. We say beams of electrons, but then in principle we should worry
about antisymmetrizing the many-electron wave-function, etc. Think of one electron interfering with
itself, in the same sense as a 2-slit experiment. Then
1
is the amplitude the electron follows path 1,
and
2
is the amplitude it follows path 2.
4
Note in electric case we had path integral in time, now we have one in space. Not hard to guess that this
generalizes to path in spacetime.
6
=
e
h
_
area encl.
da B (26)
=
e
h
(ux thru solenoid) (27)
So phase dierence is observable, & related to gauge invariant quantity,
magnetic ux through solenoid.
AB eect measured many times, starting with Chambers
5
and Furry and
Ramsey
6
The latter authors used a long magnetic whisker in place of a
solenoid. The search for AB eects in mesoscopic systems (small semi-
conductor devices) led to the discovery of weak localization in disordered
metals. Recently an eect analogous to the AB eect, but for neutral
particles, was predicted by Aharonov and Casher
7
and has also been mea-
sured.
8
2.4 Landau Levels
Consider 2D electron system in x y plane with eld B | z. Convenient
to choose Landau gauge A = Bx y, check that B = A = B z.
With this choice Hamiltonian is (convention: electron has charge -e)
H =
1
2m
( p + eA)
2
(28)
=
1
2m
_
p
2
x
+ p
2
y
+ 2eBx p
y
+ (eB)
2
x
2
_
(29)
Note that [H, p
y
] = 0, so we may write all eigenfctns. of H as eigenfctns
of p
y
, namely
(x, y) = e
ik
y
y
X(x) (30)
Substitute, nd X satises
5
R.G. Chambers, Physical Review Letters 5, 3 (1960).
6
W.H. Furry and N.F. Ramsey, Physical Review 118, 623 (1960)
7
Y. Aharonov and A. Casher, Physical Review Letters 53,.319 (1984).
8
I dont know the reference hereanybody?
7
1
2m
_
_
_ h
2

2
+ (eB)
2
_
_
x +
hk
y
eB
_
_
2
_
_
_ X = EX. (31)
Note this eqn. is exactly of harmonic oscillator form, with x shifted by
x
0
= hk
y
/eB. So we can immediately write down the eigensolns for this
problem:
(x, y) = e
ik
y
y
u
n
(x + x
0
) = e
ieBx
0
y/ h
u
n
(x + x
0
), (32)
where u
n
is nth eigenfctn. of the SHO, with eigenvalue E
n
= h(n +
1/2), and can be read o by comparison with standard SHO potential
m
2
x
2
/2, to nd
=
eB
m
(cyclotron frequency). (33)
This is just classical frequency of orbital motion of chged. particle in
magnetic eld. Energy levels labeled by n called Landau levels because
Landau solved this problem 1st (see his QM book!). What is degeneracy
of each level? Note we can have many dierent k
/
y
s all with same E
n
. If
width of system in y-direction is L
y
, assume periodic boundary conditions,
(y) = (y + L
y
), then allowed k
y
s are given by k
y
L
y
= 2, =
0, 1, 2, 3 . . .. May also translate condition into one on x
0
, classical center
of electron orbit, x
0
= 2 h/(eBL
y
). Note must have
0 x
0
L
x
, (34)
where L
x
is width of sample in x-direction, so that all e

s are orbiting
inside sample. This gives upper bound on ,
0
eB
2 h
L
x
L
y

max
(35)
Natural unit of length size of orbit

B
=

_
h
eB
(36)
So maximum number of electrons which can occupy given Landau level is
8

max
=
L
x
L
y
2
2
B
. (37)
. Remarks:
1. Note
max
depends on eld: bigger eld, more electrons can be t into
each Landau level.
2. Landau levels split by spin Zeeman coupling, so (37) applies to one
spin only.
3. Although we treated x and y asymmetrically for convenience of calcu-
lation, no physical quantity should dierentiate between the two due
to symmetry of original problem with eld in z direction!
2.5 Integer Quantum Hall Eect
Now take 2D electron system (can be really manufactured in semicon-
ductor heterostructures!) and apply E-eld in y direction as shown.
Longitudinal current density proportional to applied E-eld E
y
(Ohms
law):
j
y
=
0
E
y
(38)
9
Classically, electron trajectories curved by Lorentz force F = ev B,
can think of as extra electric eld
E
/
= v B =
j B
ne
(39)
where I used j = nev. Total current is j =
0
(E
y
y + E
/
), or
j =
0
E
0
j B
ne
(40)
which may be written
_
_
_
1

0
B
ne

0
B
ne
1
_
_
_
_
_
_
j
x
j
y
_
_
_ =
0
_
_
_
0
E
y
_
_
_ (41)
which can be inverted to nd
j
x
=

2
0
B
ne
1 + (

0
B
ne
)
2
. .
E
y
(42)

xy
and j
y
=

0
1 + (

0
B
ne
)
2
. .
E
y
. (43)

yy
These are classical expressions for the longitudinal and Hall conductivities,

xy
and
yy
in eld B. Note classical proportionality of
xy
and B!
How does QM change this picture? Assumption of classical Drude model
is that electrons scatter randomly & elastically o imperfections, leading
to constant drift velocity in presence of electric eld. This argument yields

0
=
ne
2

m
, (44)
where is mean time between collisions, n is number density of e

s.
Now suppose the magnetic eld is chosen so that number of electrons
exactly lls all the Landau levels up to some N, i.e.
10
nL
x
L
y
= N
max
=n = N
eB
h
, (45)
where last step follows from Eqs. (35-37).
9
(Ask now what happens if
level is lled. If electron scatters o imperfection, must go into another
quantum state. But all such states of the same energy are lled, so elastic
scattering impossible. Inelastic scattering frozen out, i.e. next accessible
Landau level a nite energy h away, at low T thermal energy not enough
to jump this gap. So no scattering can occur, due to Pauli principle! This
means at special values of eld, so comparing with (42-43) nd

yy
0, and

xy

ne
B
= N
e
2
h
(46)
At critical values of eld, conductivity is quantized
10
units of e
2
/h. Exptl
measurement of these values provides best determination of fundamental
ratio e
2
/h, better than 1 part in 10
7
. Nobel prize awarded 1985 to von
Klitzing for this discovery.
9
Careful: n is the total number of electrons per unit area,
max
is the number of electrons per Landau level, and
N is the number of lled levels.
10
This is the extent of the naive argument for IQHE. Note there is no discussion of what happens for elds just
above or below critical elds. Understanding why
xy
doesnt change over a nonzero range of B, i.e. existence of
quantum Hall plateaus, crucial to understanding how eect can be measured at all. Ultimate explanation relies on
eect of disorder on electronic wave functions, so-called localization eects.
11
12