Вы находитесь на странице: 1из 18

Semiclassical Model of Electron Dynamics

8.1 Description of the Semiclassical Model

Electrons in crystalline solids assume Bloch wave functions. The semiclassical model
deals with the dynamics of Bloch electrons. Drude assumed that electrons collide with the
fixed ions. This picture cannot account for very long mean free paths that had been found
in metals, as well as their temperature dependence. On the other hand, the Bloch theory
would have predicted infinite conductivity since the mean velocity of a Bloch state, v(k)
= (1/ħ)(/k), is nonvanishing. This can be traced to the fact that Bloch states are
stationary solutions to the Schrödinger equation incorporating the full crystal potential.
So, the interaction between the electron and the fixed periodic array of ions has been fully
accounted for and the ions can no longer be sources of scattering.
However, no real solid is a perfect crystal. Furthermore, there are always
impurities, missing ions or other imperfections that can scatter electrons. In fact, it is
these that limit the conductivity of metals at very low temperatures. At high temperatures,
there are thermally excited lattice vibrations producing deviations from the perfect crystal
structure, which can scatter electrons and limit conductivity. While the above reveals that
Drude’s picture of electron-ion scattering is inappropriate, by substituting the scattering
events by the realistic ones (defect and phonons), his approach for formulating the
electron dynamics is still valid. The formulation presented below describes the motion of
the Bloch electrons in between collisions.
The semiclassical model predicts how, in the absence of collisions, the position r
and wave vector k of an electron evolve in the presence of externally appled electric and
magnetic fields, assuming knowledge of the electron’s band structure, n(k), which
supposedly fully accounted for the crystal field. In the course of time, with the presence
of external electric and magnetic fields E(r,t) and H(r,t), the position, wave vector, and
band index are taken to evolve according to the following rules:
1. The band index n is a constant of motion. The semiclassical model ignores the
possibility of “interband transition”.
2. The time evolution of the position and wave vector of an electron with band index
n are determined by the equations of motion:
 
dr   1  n (k )
 vn ( k )   . (8.1a)
dt  k

dk      
  e[ E (r , t )  v (k )  B(r , t )]. (8.1b)
dt

We shall discuss the origin of these equations below.


3. The wave vector of an electron is only defined to within an additive reciprocal
lattice vector K. Therefore, all distinctive wave vectors for a single band lie in a
single primitive cell of the reciprocal lattice. In thermal equilibrium, the
contribution to the electronic density from those electrons in the nth band with
wave vectors in the infinitesimal volume element dk in the k-space is given by the
usual Fermi distribution:

by OKC Tsui based on A&M 1


Semiclassical Model of Electron Dynamics

 
 dk dk / 4 3
f ( n (k )) 3   (8.1c)
4 exp[( n (k )   ) / k B T ]  1

Points to Note:
- One should always bear in mind that very large fields (electric or magnetic) may
cause interband transitions. You are referred to A&M Ch. 12 for the discussion.
- At or near equilibrium, bands with all energies many kBT above the Fermi energy
will be unoccupied. As for those bands many kBT below the Fermi energy, we will
later see that they can be igored for consideration of electronic transport properties
since they are completely filled. Therefore, one often only needs to consider energy
bands within a range ~kBT about the Fermi energy. This greatly reduce the number
of bands needed to be considered.

For free electrons:


 
dr k
 (8.2a)
dt m

dk      
  e[ E (r , t )  v (k )  B(r , t )]. (8.2b)
dt

The equations of motion (8.1) for Bloch electrons within each band are the same as those
of free electrons (8.2) if we adopt ħ2k2/m for the electron energy n(k). However, while
ħk is the momention of the free electrons (and so eqn. 8.2b is anologous to Newton’s
second law), it is not for the Bloch electrons. Instead, ħk is the crystal momentum,
concerned in the momentum selection rule for scattering. To appreciate eqn. 8.1b, notice
that the total rate of change of an electron’s momentum, p, is given by the total force
acting on the electron, including the crystal field. Since the RHS of eqn. 8.1b accounts
for the forces due to the external E and B fields (without the crystal field), the equation
implies that the rate of change of p – ħk is exactly accountable for by the interactions of
the electron with the crystal field. Below is the proof:

The Bloch wave function is:

k =

The expectation value of the momentum of an electron in state k is:

p = <k|-iħ|k> = (8.2c)

Consider the change in p produced by the external fields:

p = (8.2d)

It shows that p has a piece coming from the plane wave components k + G in the
original wave function. When the electron state changes by k, the amplitudes of the

by OKC Tsui based on A&M 2


Semiclassical Model of Electron Dynamics

plane wave components k + G change as well. Physically, this corresponds to the


electron being reflected by the lattice and thereby causing those amplitudes to alter. If an
incident electron with plane wave component of momentum ħk is reflected with
momentum ħ(k + G), the lattice acquires the momentum -ħG. The momentum transfer to
the lattice, plat, when the state k goes over to k+G is:

(8.2e)

In eqn. 8.2e, use has been made of the fact that the portion of the plane wave component
k + G being reflected during the state change k is:

Substitute eqn. 8.2e in 8.2d, we find for the total momentum change of the system
(electron + lattice),

ptot = pplat = ħk

as implied by eqn. (8.2b).

8.2 Semiclassical Equation of Motion in a DC Applied Electric Field

The eqn. of motion, 8.1a, states that the velocity of a semiclassical Bloch electron is the
group velocity of the underlying wave packet, which is perhaps readily comprehensible.
But, eqn. 8.1b may not be as straightforward to justify. We reason it by considering
conservation of energy for an electron traversing in the field E = , whereby

n(k(t))  e(r(t)) = constant (8.3)

in the motion. The time derivative of this equation is:

 n  
  k  e  r = 0. (8.4)
k
By Eqn. 8.1a, Eqn. 8.4 becomes:

vn(k)  [ħdk/dt e] = 0. (8.5)

Since vn(k) is generally non-zero, we must have

ħdk/dt = e = eE, (8.6)

which is eqn. 8.1b, without a magnetic field. Next, observe that eqn. 8.6 is not the only
condition that gives rise to energy conservation. For example, (8.5) would still be

by OKC Tsui based on A&M 3


Semiclassical Model of Electron Dynamics

satisfied if any term perpendicular to vn(k) is added to the RHS of Eqn. 8.6. It turns out
the only appropriate term that may be added is evn(k) × B. This proves the second eqn.
of motion.
In most of the following discussions, we shall take the electronic equilibrium
distribution function to be that appropriate to zero temperature. In metals, finite
temperature effects will have negligible influence on the properties discussed below. The
spirit of the following analysis is similar to that of the analysis discussed for the Drude
model transport properties. That is, we shall describe collisions in terms of a simple
relaxation-time approximation, and focus mostly on the motion of electrons between
collisions as determined by the semiclassical equations of motion (8.1a and b).

Filled Bands Are Inert


A filled band is one in which all the energies lie below the Fermi energy. Such
bands cannot contribute to an electric or thermal current. To see this, notice that an
infinitesimal phase space volume element dk about the point k will contribute dk/43
electrons per unit volume, all with velocity v(k)=(1/ħ)((k)/k) to the current. Summing
this over all k in the Brillouin zone, we find that the total contribution to the electric and
energy (thermal) current densities from a filled band are:

(8.7)

Both of these are zero since the integral over any primitive cell of the gradient of a
periodic function must vanish, and (k) is periodic. Therefore, only partially filled bands
need to be considered in calculating the electronic properties of a solid. This explains
why Drude’s assignment to each atom of a number of conduction electrons equal to its
valence had been successful.
Clearly, a solid in which all bands are completely filled or empty will be an
electrical insulator. Since the number of levels in each band is twice the number of
primitive cells in the crystal (due to the two degenerate spin states of electrons), all bands
can be filled or empty only in solids with an even number of electrons per primitive cell.

8.3 Semiclassical Motion in an Applied DC Electric Field


In a uniform dc electric field, the solution to the semiclassical equation of motion
for k (8.1b) is: 
  eEt
k (t )  k (0)  . (8.8)

Therefore, in time t, every electron changes its wave vector by the same amount.
Accordingly,

by OKC Tsui based on A&M 4


Semiclassical Model of Electron Dynamics


    eEt 
v (k (t ))  v  k (0)  . (8.9)
  
If the band is completely filled, this constant shift in the wave vector of all the electrons
can have no effect on the electric current. This is in contrast with the free electron case,
where v is proportional to k, and would thus grow linearly with time.
Fig. 8.1 illustrates a typical plot of v(k). It is noteworthy that it decreases with
increasing k near the two edges of the Brillouin zone. In other words, the electrons of
those k states actually decelerate with the externally applied field E. This extraordinary
behavior is a consequence of the additional force exerted by the crystal field, which,
though is no longer explicit in the semiclassical model, lies buried in it through the
dispersion relation, (k). Physically, as an electron approaches a Bragg plane, the external
electric field moves it toward levels in which it is increasingly likely to be Bragg-
reflected back in the opposite direction. For example, it is just in the vicinity of Bragg
planes that the plane-wave levels with different wave vectors are most strongly mixed in
the nearly free electron approximation. This effect leads to the curious observation that
electrons that are close enough to the zone boundary have been found to behave like
“holes”.

Fig. 8.1 (k) and v(k) vs. k (or


vs. time, via Eqn.8.8) in one
dimension (or three dimensions,
in a direction parallel to a
reciprocal lattice vector that
determines one of the first zone
faces.)

Holes
Here we shall provide a detailed account for how the transport properties of electrons in
some cases can be described by that of positive charges called “holes”. By Eqn. (8.7), the
contribution of all the electrons in a given band to the current density is:

 dk  
j  e  v (k ) (8.10)
occupied 4
3

by OKC Tsui based on A&M 5


Semiclassical Model of Electron Dynamics

where the integral is over all the occupied levels in the band. But the integral of eqn. 8.10
over the entire band should be zero. So,
  
dk   dk   dk  
0  v (k )   v (k )   v (k ), (8.11)
zone
4 3 occupied
4 3 unoccupied
4 3

Hence we can equally well write Eqn. 8.10 as:



 dk  
j  (  e) 
unoccupied
4 3
v (k ). (8.12)

It follows that the current produced by occupying a specified set of levels with electrons
is precisely the same as the current that would be produced if (i) the specified levels were
unoccupied AND (ii) all the other levels in the band were occupied but with particles of
charge +e. Such fictitious particles of positive charge occupying the levels unoccupied by
the electrons are called holes.
When one chooses to regard a current as being carried by positive holes rather
than by negative electrons, one should regard those states occupied by electrons to be
unoccupied by holes, and vice versa. However, for any given band, one should never
adopt both pictures (hole and electron) for the charge carriers. Nevertheless, one may
regard some bands by the electron picture, and the other bands by the hole picture, for
one’s convenience.
Under an applied electric field, the unoccupied levels in a band evolve precisely
as if they were occupied by real electrons (of charge –e). That is, if there are applied E
and B fields, the motion of both electrons and holes is governed by the same equation:
   1 
k  (e) E  v  H . (8.13)
 c 
This is because eqn. (8.13) describes how the occupied orbitals evolve with time; and the
unoccupied orbitals have to evolve in the same manner because a newly occupied orbital
is necessarily accompanied by a newly emptied orbital, etc., which requires the
unoccupied states to evolve in the same manner do the occupied ones. (see pp. 226-227
of A&M.) Given eqn. 8.13, how can holes be distinguished from electrons?
In classical treatment, the RHS of eqn. (8.13) is the force acting on a charged
particle due to E and B, and would have been set equal to the mass of that particle times
dv/dt. So, if dv/dt // dk/dt, the electron orbit would resemble that of a free particle with
negative charge. On the other hand, if dv/dt is // -dk/dt, the electron orbit would resemble
that of a free particle of positive charge. It turns out, it is more often the case that dv/dt is
directed opposite to dk/dt when the k orbital is unoccupied. This may be perceived from
the following: At equilibrium or near equilibrium (which is the condition found in most
cases of interest for electron transports), the unoccupied levels usually lie near the top of
the band. If the band energy (k) has its maximum value at k0, say, then if k is
sufficiently close to k0, we may expand (k) about k0. The linear term in (k  k0)
vanishes because k0 is a maximum point. If we assume that k0 is a point with sufficently
high symmetry, then

by OKC Tsui based on A&M 6


Semiclassical Model of Electron Dynamics

   
 (k )   (k0 )  A(k  k0 ) 2 , (8.14)

where A is a positive constant. Rewrite A as:


2
 A. (8.15)
2m *
Hence, for levels near k0, we have
 
 1  (k  k0 )
v(k )    , (8.16)
 k m*
 d    
So, a  v (k )   k, (8.17)
dt m*

This shows that the acceleration, a, of states near the top of a band is opposite to k . It
follows from eqn. (8.13) that acceleration of these states are opposite to the electric force,
making these electrons behave like positive-charges. By regarding those near-top
electrons as particles with charge +e, Eqn. 8.17 shows that they move as if they have an
effective mass of +m*.
We can demonstrate the above point in a more general way. Consider:
  
   d   d 1  (k ) 1 d   (k )  1  2 ( k ) 

k a  k  v  k     ki      ki  kj, (8.18)
dt dt  k  i dt  ki   ij ki k j

Then a is opposite to k if

 2 ( k ) 
ij  i k k  j  0 (for any vector ). (8.19)
i j

Suppose the local maximum of (k) is at k0. In the vicinity of k0, one can expand (k) as:

  1  2 ( k )
 (k )   (k0 )    i j . (8.20)
2 ij ki k j

It follows that Eqn. 8.19 must hold for (k0) to qualify to be a local maximum.
The quantity m* in Eqn. 8.15 determining the dynamics of the electrons near band
maxima of high symmetry is known as the “hole effective mass”. More generally, one
defines an “effective mass tensor” by:
    
 1

 1  2 ( k )
M (k ) ij   2
 ki k j
1  2 (k )
 2
 k j ki

1 vi (k )
 k j
1  v (k ) 
     , (8.21)
  k  ij

where the sign is – or + according to whether k is near a band maximum (holes) or


minimum (electrons), respectively. By using eqn. (8.21), we can write:

by OKC Tsui based on A&M 7


Semiclassical Model of Electron Dynamics

    
 dv v dk
a     M 1 (k )k . (8.22)
dt k dt
Hence, the equation of motion (8.13) takes the form:
   1  
M (k )a   e( E  v (k )  H ). (8.23)
c
If the pocket of holes (or electrons) is small enough, one can replace the mass tensor by
its value at the maximum (minimum), leading to a linear equation only slightly more
complicated than that for free particles. Such equations describe quite accurately the
dynamics of electrons and holes in semiconductors.

8.4 Semiclassical Motion in a Uniform Magnetic Field


In a uniform magnetic field, the semiclassical equations are

   1  (k )
r  v (k )   , (8.24)
 k
   
 k  ev (k )  B. (8.25)
These equations immediately evident that the component of k along the field B and the
electronic energy (k) are both constants of motion. (The latter is because d(k)/dt = F  v
= e(v  B)  v = 0.) These conservation laws dictate that electronic orbits in k-space are
curves given by the intersection of surfaces of constant energy with planes perpendicular
to the magnetic field (Fig. 8.2). The sense in which the orbit is traversed follows by

Fig. 8.2

In the example show, v is pointing away from k = 0


and so the particle is electron-like. Convince
yourself that the sense of rotation of the orbit
reverses when the particle is hole-like or v is
pointing towards k = 0.

observing that v(k), being ~k, points to higher energies in the k-space. Therefore,
closed k-space orbits surrounding levels of higher energies (hole orbits) are traversed in
the opposite sense to closed orbits surrounding levels of lower energy (electron orbits). In
other words, the sense of orbital motion determines whether the carriers are electron- or
hole-like under applied magnetic field.

by OKC Tsui based on A&M 8


Semiclassical Model of Electron Dynamics

The projection of the real space orbit in a plane perpendicular to the field,
  ˆ ˆ 
r  r  B( B  r ) , can be found by taking the vector product of both sides of Eqn. 8.25
with a unit vector parallel to the field B. This yields


  
Bˆ  k  eB(r  Bˆ ( Bˆ  r ))  eB r , (8.26)

which integrates to:

B (8.27)
B

-------------------------------------------------------------

Below is a derivation of Eqn. (8.27) and exploration of its physical meaning. Bˆ ( Bˆ  r )

1
2

B v  (-ħ/e)B  k

-----------------------------------------------------------------------
The above derivation shows that the projection of the real space orbit in a plane
perpendicular to the field is simply the k-space orbit, rotated through 90o about the field
direction and scaled by the factor ħ/eB. But it should be noted that orbits in semiclassical
generalization need not be closed curves (Fig. 8.4).

by OKC Tsui based on A&M 9


Semiclassical Model of Electron Dynamics

It is of interest to express the rate at which the orbit is traversed. Consider an orbit
of energy  in a particular plane perpendicular to the field B in the k-space (Fig. 8.5a).
The time it takes to traverse that portion of the orbit lying between k1 and k2 is

(8.28)

By Eqns. 8.24 and 8.25,


  e    eB   
k  2 B  2   (8.29)
 k   k  

Therefore, Eqn. 8.28 can be rewritten as:

B
(8.30)

where (/k) is the component of /k perpendicular to the field, i.e., its projection in
the plane of the orbit. The quantity has the following geometrical meaning: In the plane
of the orbit for electron energy , we defined the vector (k) to be one perpendicular to
the orbit at point k, and connects the point k to a neighboring orbit in the same plane (i.e.,
same k) of energy  +  (Fig. 8.5b). When  is very small, we have

B
B

Fig. 8.3 The projection of the


r-space orbit (b) in a plane
perpendicular to the field is
obtained from the k-space
orbit (a) by scaling with the
factor ħ/eB and rotation
through 90o about the axis
determined by B.

by OKC Tsui based on A&M 10


Semiclassical Model of Electron Dynamics

Fig. 8.4 Presentation in


the repeated-zone scheme
of a constant-energy
surface with simple cubic
symmetry, capable of
giving rise to open orbits
in suitably oriented
magnetic fields. One
such orbit is shown for a
magnetic field parallel to
[101].

Fig. 8.5 The geometry of


orbit dynamics. (a) The
time of flight between k1
and k1 is given by
(ħ2/eB)k1k2 dk/|(/k)| (b)
A section of (a) in a plane
perpendicular to B
containing the orbits with
the same kz. The shaded
area is (A1,2/).

(8.31)

Since /k is perpendicular to surfaces of constant energy, the vector (/k) is


perpendicular to the orbit and B, and hence parallel to (k). So, Eqn. 8.31 can be written
as:
(8.32)

and Eqn. 8.30 becomes:

(8.33)
B

by OKC Tsui based on A&M 11


Semiclassical Model of Electron Dynamics

The integral in (8.33) gives the area of the plane between the two neighboring orbits from
k1 to k2 (Fig. 8.5b), A1,2. Hence if we take the limit of (8.33) as   0, we have

. (8.34)
B

Eqn. 8.34 is most often used in cases where the orbit is a simple closed curve, and k1 and
k2 are chosen to give a single complete circuit (k1 = k2). The quantity t2 – t1 is then the
period T of the orbit. If A is the k-space area enclosed by the orbit in its plane, then Eqn.
8.34 gives
(8.35)
B
Compared to the free electron result, i.e.,

B
(8.36)

it is customary to define a cyclotron effective mass m*(,kz):

(8.37)

8.5 Semiclassical Motion in Perpendicular Uniform Electric and Magnetic Fields

When we discussed the Hall effect in the Drude model, we only qualitatively considered
the electron motion under crossed E and B fields. (Recall: We balanced the Lorentz force
and the electric force in the transverse direction and thereby derived that RH = Ey/(jxB) =
1/ne.) Here, we discuss it in the semi-classical model.

When a uniform electric field E is added with a perpendicular uniform magnetic field B
with |E| < |B|, Eqn. 8.27 for the projection of the real space orbit in a plane perpendicular
to B acquires an additional term

B (8.38)

where
B (8.39)
B
Note that w is the the velocity of the frame of reference in which the electric field
vanishes (see Jackson’s book pp. 582-584). It follows that electrons in crossed E and B
fields move in a superposition of circular orbits (as if only the B field is present) plus a
linear drift at the velocity w in the r-space (see Fig. 8.6).

by OKC Tsui based on A&M 12


Semiclassical Model of Electron Dynamics

Bigger
Smaller radius
radius when v//-
when E. Fig. 8.6 E  B
v//E. drift of charges +e
and –e in crossed
E and B fields.
w
B

Recall that d/dt = (/k)·(dk/dt) = ħv·[(-e/ħ)(vB + E)] = -ev·E. Since v is in general


not perpendicular to E, d(k)/dt  0 and (k) may vary with time.

When E and B are perpendicular, one can show that the equation of motion (8.1b) can be
written in the form (see below):

B (8.40)

where
(8.41)

One may appreciate the physical origin of these equations by perceiving that Eqn. 8.41 is
the energy of a free electron in the reference frame moving with velocity w, and Eqn.
8.40 is the equation of motion an electron would have if only the magnetic field B were
present, and the band structure given by (k. ) Given Eqns. (8.40) & (8.41), the k-space
orbits are given by intersections of surfaces of constant. (k) with planes perpendicular to
the magnetic field.
Note that in cases where |E| > |B|, the orbit will be hyperbolic and not closed.
---------------------------------------------------------------------
Below is a derivation of Eqn. (8.41):

by OKC Tsui based on A&M 13


Semiclassical Model of Electron Dynamics

B B
The triple product would not be E-hat had E

been not perpendicular to B.

This gives .

------------------------------------------------------------------

8.6 High-Field Hall Effect and Magnetoresistance (in theor long  limit)
In this section, we shall analyse the case for crossed E and B field where B is very large
of the order of 1 T or more (so |w| = E/B << c), and (k) differs only slightly from (k).
The limiting behavior of the current for these cases turn out to be quite different
depending on whether (a) all occupied (or all unoccupied) electronic levels lie on orbits
that are closed curves or (b) some of the occupied and unoccupied levels lie on orbits that
do not close on themseleves, but are “open” in k-space.

(a) Cases where all occupied (or all unoccupied) orbits are closed. We shall take the
high magnetic field condition to mean that these orbits can traverse many times between
successive collisions. In the free electron case, this reduces to the condition c >> 1,
where  is the relaxation time and c is the cyclotron frequency (see Chapter 1).
Suppose that the period T is small compared with the relaxation time, , for every
orbit containing occupied levels. To calculate the current, j, one uses j = ne<v>. Here,
<v> is the average velocity of an electron between two collisions, which may be taken to
be from t =  to t = . By Eqn. 8.38, we have

B (8.42)
B
Since we are considering for the case where all the occupied orbits are closed, k =
k(0) – k() is bounded in time. So for sufficiently large , the contribution from the
drift velocity w dominates the average velocity. This provides the long- limit to the
current

by OKC Tsui based on A&M 14


Semiclassical Model of Electron Dynamics

B
B (8.43)

But if it is the unoccupied levels that all lie on closed orbits, the corresponding result is

B
B (8.44)

Eqns. 8.43 and 8.44 suggest that when all relevant orbits are closed, the deflection of the
Lorentz force is so effective in preventing electrons from acquiring energy from the
electric field that the uniform drift velocity w perpendicular to E gives the dominant
contribution to the current. Recall that the definition of the Hall coefficient is the
component of the electric field perpendicular to the current, divided by the product of the
magnetic field and the current density. We have RH = Etransverse/(jtotB) = E/(newB) =
1/(ne):

(8.45)

Note that the use of the symbol R is to indicate that the expression is valid under the
condition /T  . It is remarkable that Eqn. 8.45 gives an identical result as that found
in the free free electron case, which preserves the notion that the high-field Hall
coefficient R provides a valuable measure of the electron (or hole) density. It is also
remarkable that Eqn. 8.45 allows for the possibility of a positive Hall coefficient. If
several bands contribute to the current density and each of them has only closed electron
(or hole) orbits, then Eqn. 8.43 or 8.44 holds separately for each band, and the total
current density in the high-field limit will be
 neff e 
lim j    ( E  Bˆ ) , (8.46)
 / T  B

where neff is the total density of electrons minus the total density of holes. The
corresponding expression for R is
1
R   . (8.47)
neff e
Furthermore, it can be shown that the corrections to the high-field current densities (Eqns.
8.43 & 8.44) are smaller by a factor of order 1/c (Problem 5 of A&M Ch. 12), the
transverse magnetoresistance approaches a field-independent constant in the high-field
limit.

(b) Case 2 Some orbits are open. More specifically, we refer to those cases where there
are energy bands near the Fermi surface containing open orbits (Fig. 8.4). Electrons in
open orbits are not forced by the magnetic field to undergo a periodic motion along the
direction of E. (This is in contrast with the case of closed orbits where the electrons
spend equal amount of time traveling along and opposite to E (Fig. 8.6).) Therefore, the

by OKC Tsui based on A&M 15


Semiclassical Model of Electron Dynamics

electrons may acquire energy from the E field. In particular, if the unbounded orbit
stretches in a real space direction n̂ , one would expect to find a finite contribution to the
current directed along n̂ and proportional to the projection of E along n̂ :

(8.48)
Contribution from projection Contribution from components of
of open orbit parallel to E the open orbit perpendicular to E,
and from the closed orbits, if any.

Fig. 8.7 illustrates the physical origin of how an open orbit may give rise to a net electric
current along n̂ . The possibility of a non-vansihing term not parallel to w in the high-
field limit is also in accord with the general result of Eqn. (8.42). For an electron
traversing in an open orbit, the growth in its wave vector k due to the first term is
unbounded, enabling its contribution to j to dominate that of the second term. Because
the rate at which the orbit is traversed is proportional to B, one expects this contribution
to the average velocity to be independent of B and directed along the real space direction
of the open orbit n̂ .
To appreciate the limiting behavior implied by Eqn. (8.48) on the high-field
magnetoresistance, consider an experiment in which the direction of current flow does
not lie along the direction of the open orbit in real space n̂ (which happens when n̂ does
not lie parallel to the direction where the circuit is closed, see Fig. 8.8). When the B field
is high, by Eqn. (8.48) j can be misaligned from n̂ only if E  n̂ = 0. So, if we write

j, n

vk

Fig. 8.7

8.41

by OKC Tsui based on A&M 16


Semiclassical Model of Electron Dynamics

Fig. 8.8

the E field in the following form:

(8.49)

where n̂ ’ is a unit vector perpendicular to both n̂ and Ĥ ( n̂ ’= n̂  Ĥ), E(0)  constant and
E(1)  0 as H  . By definition, the magnetoresistance, , is :

(8.50)

When j is not parallel to n̂ , E  E(0) n̂ ’ in the high-field limit, which gives the
corresponding limit of :

(8.51)

To find (E(0)/j), subs. Eqn. 8.49 in Eqn. 8.48:

(8.52)

Next, take the dot product with n̂ ’on both sides, and use the fact that n̂ ’ n̂ = 0, one has:

(8.53)

which gives

(8.54)

by OKC Tsui based on A&M 17


Semiclassical Model of Electron Dynamics

Subs. this result in Eqn. 8.51, we find:

(8.55)

Since (1) vanishes in the high-field limit, this gives a magnetoresistance that grows
without limit with increasing field, and is proportional to the square of the sine of the
angle between j and the open orbit n̂ (since n̂ is perpendicular to n̂ ’ in the same plane
with j). Therefore, the semiclassical model resolves another anomaly of free electron
theory, providing possible mechanisms by which the magnetoresistance can grow without
limit with increasing magnetic field.

by OKC Tsui based on A&M 18

Вам также может понравиться