Vous êtes sur la page 1sur 41

Quantum Mechanics

Chapter 3. 2D and 3D Problems of Wave Mechanics

The coverage of multi-dimensional problems of wave mechanics in this course is minimal: it is limited
to a few phenomena (such as the Aharonov-Bohm effect or Landau levels) which could not be revealed
in 1D due to topological reasons, and a few renowned problems, such as the Born approximation in
scattering theory or the “Bohr” (hydrogen-like) atom, which are of key importance for applications.

3.1. Density of states at free motion


Let us start with the simplest case of free particles (U =0). As was discussed in Chapter 1, the
eigenfunctions are just plane monochromatic waves:
 
 (r )  Ce  r .
 i k (3.1)
Evidently, this expression may be factored as
 ik x x
 (r )  X ( x)Y ( y ) Z ( z ) , with X ( x)  e , etc. (3.2)
The distribution of quantum states along each axis is exactly the same as in Sec. 2.4, with
independent numbers lx, etc., so that the total number of states is the product of the numbers in each
direction. As a result, Eq. (2.206) may be generalized to d > 1 dimensions in the following way:
Vd
 f (k )  (2 )  f (k )d
k
d
d
k, (3.3)
k

where for d = 2, V2 is the system area, while for d = 3, V3 is the “real” volume V. In particular, taking
f(k) = 1, we get the state counting rule
Vd
(2 ) d k
Nd  d dk . (3.4)

Now let us calculate the density of states dNd /dE on the axis of energy
2k 2 2 d 2
E
2m
 kj .
2m j 1
(3.5)

For d = 2, n2 is the number of states with energy below

E
2k 2  kx  k y

2 2 2

,
  (3.6)
2m 2m
i.e. within a circle of radius k, so that Eq. (4) yields
A
N2  k 2 . (3.7)
(2 ) 2
Comparing this expression with Eq. (6), we can rewrite it as

© K. Likharev, 2008 1
Quantum Mechanics

mA
N2  E, (3.8)
2 2
so that in contrast with the 1D case (see Eq. (2.208)) the density of 2D states does not depend on energy:
1 dN 2 m
g2   . (3.9)
A dE 2 2
This is a unique property of quantum motion in 2 dimensions; in 3D the answer is different:
3/ 2
V 4 3 V 4  2mE 
N3  k    , (3.10)
(2 ) 3
3
(2 )3 3   2 
so that the 3D density of states
1 dN 3
g3   E1 / 2 (3.11)
V dE
grows with energy.

3.2. Quantum interference, the AB effect, and magnetic flux quantization


If particle is not free, the WKB approximation may be generalized in the evident way:
 
 r  
 a(r )
 (r )  
k (r )
exp 

i  k ( r ' )  dr ' .

(3.12)

The first major difference with the 1D case is that the (possibly, implicit) lateral quantum confinement
of the particle motion may vary along the integration path, so that the 1D arguments (probability current
conservation) leading to the constancy of coefficient a, are not more valid. Hence, in order not to be
misled by the explicit form of the denominator, it is better to use this formula in the form

   r   
 (r )  A(r ) expi  k (r ' )  dr ' , (3.13)
 

and remember that the calculation of factor A(r ) may be rather involved.
The second difference is that if a particle may move from one point to another via several
alternative routes, the wavefunction gets contributions from all of them.1 In most cases, the contribution
from the classical trajectory dominates this sum, but when we calculate the wavefunction outside that
trajectory, contributions from several paths may be comparable.
The most clear illustration of this point is the famous two-slit interference arrangement (Fig. 1),
which the wave-mechanics analog of the Young’s experiment in optics, with the light source replaced
by a source of particles with fixed energy E, and the screen, by a particle detector (counter). In the
emitted particles do not interact (which is always true if the emission rate is sufficiently low), the

counting rate is proportional to the probability density w(r ) to find a single particle at the detector’s

1 I will postpone the proof until Chapter 4.

© K. Likharev, 2008 2
Quantum Mechanics

location. To calculate it from Eq. (1.22), we need to find the total wavefunction at the detector’s
location. If the slits in the opaque partition are narrow enough, in the WKB approximation, we can just
sum up contributions from two possible particle’s paths shown in Fig. 1:
 
  r     r   
 S (r )  A1 expi  k (r ' )  dr '  A2 expi  k (r ' )  dr '. (3.14)
 L1   L2 

L1
1
C w  w(x)
2
L2
particle “screen” Fig. 3.1. Scheme of the two-slit
source (detector (Young-type) interference experiment.
partition
with slits plane)

This result may be rewritten as



 r  
 S (r )  expi  k (r ' )  dr '  i arg A1 ) A1 expi   A2  ,
 
(3.15)
 L2 
with the phase difference between two paths defined as
     
   k (r )  dr   k (r )  dr  const   k (r )  dr  const , (3.16)
L1 L2 C

where C is the contour passing along the alternative particle trajectories (see the dashed line in Fig. 1).
The resulting counting rate is proportional to
  
w(r )   S (r ) *S (r )  A1  A2  2 A1 A2 cos  .
2 2
(3.17)

The phase difference clearly depends on the detector’s position x; for example for the free-space motion
  k ( L1  L2 )  const. (3.18)
As a result, the counting rate oscillates, giving the famous interference pattern, with the amplitude
proportional to the product A1A2, and hence vanishing if any of the slots is closed.
Let us now discuss a very interesting effect of magnetic field on the quantum interference. In
classical physics, the field affects charged particle’s motion only if the field is applied at the particle’s
trajectory. In quantum mechanics, this is not so, and the field changes the interference pattern even if it
is applied deep inside contour C, where its field lines never approach any possible particle trajectory, i.e.
any region with appreciable  - see Fig. 2.

© K. Likharev, 2008 3
Quantum Mechanics


B0
1
w  w(B)
C 
2
Fig. 3.2. The AB effect.

Let us first develop a general framework for account of effects of electromagnetic fields on a
quantum particle, which will also give us some important by-product results. In order to do that, we
need to calculate the Hamiltonian operator of a charged particle in the field. From classical
electrodynamics we know that the electric (or at least electrostatic) field’s incorporation into energy is
straightforward. Indeed, the field may be presented as a gradient of its electrostatic potential ,
 
E   , (3.19)
so that if we need to describe the force exerted by the field

on a particle with electric charge q,

FE  qE , (3.20)
we can just incorporate the potential into potential energy: in the absence of other forces
U  q . (3.21)
However, magnetic field’s effect is peculiar: since the “Lorentz force”2
  
FB  qv  B , (3.22)
which it exerts on a charged particle, does no work:
      
dW  F  dr  F  v dt  q (v  B)  v dt  0,
the field cannot be presented by any potential energy. (This is also clear from the circular topology of
magnetic field lines.) But for wave mechanics, we need the system’s Hamiltonian; how to get it?
Fortunately, classical (analytical) mechanics gives us a recipe (see, e.g., CM Chapter 2). First,
we should find the Lagrangian function L as a function of generalized coordinates (in the case of
motion without mechanical constraints, the best choice of those is just the Cartesian coordinates x j ) and
velocities vj = dxj/dt. Then, considering all vj as independent arguments, we should calculate all
generalized momenta
L
pj  . (3.23)
v j
This enables us to find the Hamiltonian function of the system
H   p jv j  L . (3.24)
j

2 In SI units; in Gaussian unites, the right hand part (and generally all terms in the equations of this section, which
include either B or A) should be divided by c.

© K. Likharev, 2008 4
Quantum Mechanics

The Lagrangian function L, which is necessary to implement this program, is supplied by


classical electrodynamics (see, e.g. EM Sec. 10.8): for otherwise free, non-relativistic particle,
mr 2   m 3 3
L  qv  A  q   v 2j  q  v j A j  q , (3.25)
2 2 j 1 j 1

where A is the vector-potential which may be defined via the EM fields which, in the general (time-
dependent) case are

  A   
E    , B   A. (3.26)
t

(Notice the interesting way in which A enters the Lagrangian: it is neither in the kinetic nor in the
potential-energy part, but something in-between.) Now, we can use Eq. (25) to calculate momentum
components
L
pj   mx j  qA j . (3.27)
v j
This result may be presented in the vector form,
  
p  P  qA , (3.28)
where vector
   
P  p  qA  mv (3.29)

is called “kinematic momentum” of the particle, while vector p its “canonical momentum”.3
The distinction between these two momenta becomes even more clear if we notice that , because
the vector-potential is not “gauge-invariant”: if we make the “gauge transformation”4
  
A  A   , (3.30)
with arbitrary scalar “gauge function” , the magnetic field does not change – see the second of Eqs.
(26). Moreover, according to the first of these equations, if we make the simultaneous replacement

   , (3.31)
t
the gauge function does not affect the electric field either. With that, the gauge function does not change
 
the classical particle’s equation of motion, and hence the velocity v and momentum P . Hence, this
 
momentum is gauge-invariant, while p is not, because it changes by q .
Now, continuing our way to quantum mechanics, we use Eq. (24) to calculate


3 The origin of the first term is clear from the definition of P , while the latter name refers to the “canonical”

(here meaning “standard”) way of the derivation of p from analytical mechanics – see Eq. (23).
4 This name is purely historical, and has no clear physics context.

© K. Likharev, 2008 5
Quantum Mechanics

 
H   mv j  qA j v j   v 2j  qv j A j   q  v 2  q 
m

2

3
m 1 2 1 
P  q  p  qA  q .(3.32)
j 1  2  2 2m 2m
This is the Hamiltonian to use in the Schrödinger equation. According to the postulate (iv) formulated in
 
Sec. 1.2, p should be replaced for (i) :

 2   iq  2
ˆ
H    A   q , (3.33)
2m   
so that the Schrödinger equation becomes

 2   iq  2
   A    U  E , U  q . (3.34)
2m   

We may now repeat all the calculations of Sec. 1.5 for the case A  0 , and get the following
generalized expression for the probability current density:

j
     iq  

2im  
   A
 
  c.c


 2m
1
 

P 
  c.c 
 2 
m
   
q 
A . (3.35)
   
We see that the current is gauge-invariant (as required for any observable) only if the wavefunction’s
phase  changes as
q
   . (3.36)

This may be a point of concern: the quantum interference depends on ; can it happen that the
interference pattern depends on the gauge function choice (which would not make sense). Fortunately,
this is not true, because the phase difference between two interfering beams is transformed as
q
     . (3.37)

But  (in contrast to ) has to be a single-valued function of coordinates, hence  = 0, so that  (and
hence the interference pattern) are gauge-invariant.
However, while  is not affected by the gauge function choice, it is affected by the magnetic
field, even if it is localized well inside contour C (Fig. 2). Indeed, in this case the field cannot not affect

particle’s velocity, i.e. P should not depend on the field. Hence, in classical mechanics the combination
 
p  qA should be field-independent:
  
p B  p 0  qA . (3.38)

Translating that fact to quantum mechanics, we get


  
 i B  i 0  qA . (3.39)

Presenting wavefunction  in the form exp{i}, and taking into account that in the WKB
approximation we may limit differentiation of the wavefunction to its exponent, we get

© K. Likharev, 2008 6
Quantum Mechanics

  
 B   B  qA . (3.40)

Integrating this equation along contour C, we get


q  

 B   0  A  dr . (3.41)

But from the electrodynamics we know that the last integral is just the magnetic field flux through a
surface limited by contour C, so that we may rewrite Eq. (41) as
q
 B   0  . (3.42)

In terms of the interference pattern, this means a shift of fringes, proportional to the magnetic flux (Fig.
3). This phenomenon is usually called the Aharonov-Bohm (or just “AB”) effect.5 For particles with a
single elementary charge, q = e, this result is frequently presented as

 B   
B 0
 2 , (3.43)
0

where the fundamental constant 0  2/e = h/e  4.1410-15 Wb is called the “magnetic flux
quantum”, because of the reasons we will soon discuss.

Fig. 3.3. A typical result of a two-slit-interference


experiment (by W. Bayh, Z. Phys. 169, 492
(1962)) showing the AB effect for electrons. The
interference pattern has been captured by a
photographic plate moved vertically. In the middle
of this motion, magnetic flux has been ramped up
by about 4 0.

The AB effect may be “almost” explained in terms of Faraday’s induction law. Indeed, let us
change the magnetic flux from 0 to  so fast that  does not have time to change, and consider
wavefunctions j in trajectories’ end points (j = 1 or 2) as spatially-localized. (For the experiment
sketched in Fig. 2, this is just a crude approximation, but the experimental setup may be modified in the
way that the localization assumption would be harder to criticize.) Let us now select such a gauge
function that the vector-potential remains zero everywhere on the particle trajectories, while the
Faraday’s induction e.m.f. is referred to the electric field, or more specifically to voltage V between
points 1 and 2:

5 I prefer the latter, less personable name, because the effect was actually predicted by W. Ehrenberg and R.
Siday in 1949, and merely rediscovered by Y Aharonov and D. Bohm in 1959. To be fair, it was the latter work
which had triggered a wave of interest to the phenomenon, and it was experimentally observed by R. Chambers in
1960 and several other authors soon after that – see, e.g. Fig. 3.

© K. Likharev, 2008 7
Quantum Mechanics

  d
V   E  dr   . (3.44)
C
dt

Then the spatial parts j of the wavefunctions are not affected by the field, and we still can use the
variable separation described by Eq. (1.41). For the temporal factor a(t), we still can use Eq. (1.44a):
da j
i  E ja j , (3.45)
dt
even if Ej is a function of time. Integrating this equation, we get
 i t  i
a j (t )  a j (0) exp  E j (t ' )dt '  a j (0)e j , (3.46)
  

with
t
1
 j    E j (t ' )dt ' . (3.47)

This is essentially the time-domain version of the WKB approximation. Subtracting these equations
written for points 1 and , we get
t
1

   E (t ' )dt ' . (3.48)

Now, with E = qV, and using Eq. (44), we finally get


q
  , (3.49)

i.e. the same result as before. Indeed, with our new gauge choice we have just pushed the phase factor
from the spatial part of the wave function to its temporal part.
The good news here is that the AB effect is now reduced to the Faraday’s induction. The bad
news is that this interpretation is limited. Indeed, what if the flux change is done when the there is no
particles in the system? Experiment confirms the conclusion of Eq. (42) that the particles arriving after
the change is long over (and hence there is no e.m.f. anywhere in the system) still follow the new
interference pattern, with   0. Hence there should be something in the space where the particles
propagate (i.e., outside of the magnetic field region), which holds this information in time. The standard

interpretation of this surprising phenomenon is as follows: the vector-potential A is not just a
convenient mathematical tool, but a physics reality (just as its electric counterpart ), despite a large
freedom of choice we have describing the spatial and temporal dependences of these potentials – see
Eqs. (30)-(31).
Let us briefly discuss a very interesting form the AB effect takes in superconductivity. Then our
results require two changes. The first one is simple: since superconductivity may be interpreted as the
Bose-Einstein condensate of Cooper pairs with electric charge q = 2e, 0 has to be replaced by the
“superconducting flux quantum”

© K. Likharev, 2008 8
Quantum Mechanics

h
0   2.07  10 15 Wb  2.07  10 7 Gs - cm 2 . (3.50)
2e
Second, since the pairs are Bose particles and are condensed in the same quantum state (which is
described by the same wavefunction), the electric current density, proportional to the probability current
density j, may be extremely large – in real superconducting materials, above ~1011 A/m2. In these
conditions, one cannot neglect the contribution of that current into the magnetic field and flux , which
(according to the Lenz rule) tries to compensate changes in external flux.
In order to see possible results of this contribution, let us consider a superconducting loop (Fig.
4). Due to the Meissner effect (which is just another version of the flux self-compensation), current and
magnetic field penetrate inside the superconductor by only a small distance (called the “London’s
penetration depth) L ~ 10-7 m.6 If the loop is made of a superconducting wire which is considerably
thicker than is thicker than L, we can draw contour deep inside the wire, at which the current density is
negligible. According to Eq. (35), everywhere at the contour,
 q 
  A  0 . (3.51)

Integrating this equation along the contour, we need to have  = 2n, because the wavefunction  
exp{i} should be “essentially” the same, i.e. produce the same observables. As a result, we get
h h
Φ n n  n 0 .
q 2e
This is the famous flux quantization effect (predicted in 1949 by F. London and experimentally
discovered in 1961 by B. Deaver and W. Fairbank in the US, and R. Doll and M. Näbauer in Germany)
which justifies the term “magnetic flux quantum” for 0.7

B

C
Fig. 3.4. Flux quantization in a superconducting
loop.

Now, since as a by-product of our discussion we have Eq. (34) on hand, let us use it for the
discussion of Landau levels and (very superficially) the quantum Hall effect. Let us take a uniform 2D
quantum well (say, parallel to [x,y] plane), with strong quantum confinement in the perpendicular (z)
direction. According to the earlier discussion, relaxed particles will always sit on the lowest energy level

6 For more detail, see EM Sec. 6.3.


7 Very interesting effects of “partial flux quantization”, which are used in particular for supersensitive
magnetometers (“SQUIDs”) and ultrafast computing devices, arise when a superconductor loop is closed by a
weak superconducting link (a “Josephson junction” – cf. Sec. 2.2 above). For a brief review of these effects, the
reader is referred to Sec. 6.3 of the EM part of these notes.

© K. Likharev, 2008 9
Quantum Mechanics

in that direction, (Ez)1. Adding this energy to U(x,y) = const, and taking the resulting constant energy as
the reference, for the 2D motion in [x,y] plane we get equation
2
 2   iq  
    A    E , (3.52)
2m   
Let us find its solutions for the case if magnetic field is uniform and perpendicular to the plane:
 
B  Bn z . (3.53)
According to the second of Eqs. (26), this imposes the following restriction on the choice of vector-
potential:
Ay Ax
B  , (3.54)
x y
but gauge transformations still give a lot of freedom in its choice. In 1928, L. Landau (then just 20 years
old!) realized that the energy spectrum of the system may be obtained by selection of a very simple
choice:
Ax  0, Ay  Bx, (3.55)

which evidently satisfies Eq. (54), but in a strange (and brilliant!) way does not utilize the equivalence
of the x and y directions. Now, expanding wavefunction into the Fourier integral in direction y:
iky
 ( x , y )   X k ( x )e dk , (3.56)

we see that for each component of this integral, Eq. (52) yields a separate equation
2
2  d   q 
  nx  in y  k  Bx  X k  EX k . (3.57)
2m  dx   
Since the vectors inside the parentheses are mutually perpendicular, its square has no crossterms, and we
get
2
2 d 2 2  q ~ k
 Xk   Bx  X k  EX k , where ~
x  x . (3.58)
2m d~
x2 2m    qB

But this 1D Schrödinger equation is identical to Eq. (2.229) of the 1D harmonic oscillator, with the
effective frequency
qB
c  . (3.59)
m
In this expression, it is easy to recognize the classical “cyclotron frequency” of particle motion in the
magnetic field. (The expression may be readily found from the 2nd Newton law for the particle on a
circular cyclotron trajectory of radius r,

© K. Likharev, 2008 10
Quantum Mechanics

v2
m  FL  qvB , (3.60)
r
realizing that the resulting ratio v/r = qB/m is just the angular frequency of particle rotation, i.e. its
radius-independent frequency.) Hence, the energy spectrum for each term of integral (56) is just
 1
En  c  n   , (3.61)
 2
and does not depend on k, and hence is the full spectrum of the equation.
This is our first example of a highly degenerate system: for each eigenvalue En, there are many
different eigenfunctions. They may be used to assemble a large variety of linear combinations, including
those which approach classical circular orbits, if the orbit radius r (determined, e.g., by initial
conditions) is well above the characteristic “Landau radius”
1/ 2
  
rL    (3.62)
 qB 
which results from Eq. (2.232) after the replacement 0  c, and characterizes the purely quantum
uncertainty of cyclotron motion. (For a field of 10 T, typical for experimentation with solid-state 2D
systems, rL ~ 10 nm.)
A detailed analysis (for which we would not have time) shows that magnetic field does not
change the average density of 2D states (see Eq. (9)), but just assembles them in “bundles” (see Fig. 5a),
so that the number of states on each Landau area (per unit area) is
m qB qB
n L  g 2 E   c   . (3.63)
2 2
2 h
En (a) (b)

c EF
Fig. 3.5. (a) The “assembly” of 2D
c states into Landau-level “bundles”,
and (b) filling the levels by external
0 electrons at the quantum Hall effect.

The most famous application of the Landau levels concept is the explanation of the quantum
Hall effect (first observed in 1980 by K. von Klitzing, G. Dorda and M. Pepper). Generally, the Hall
effect is observed in the geometry, sketched in Fig. 6, where current I is passed through a sample (“Hall

bar”) with 2D electron gas placed into a magnetic field B perpendicular to the conduction plane. The
classical analysis of the effect is based on the notion of the Lorentz force (22). This force deviates
electrons from their motion from one external electrode to another electrode, bending them 
to the
isolated edges of the bar. Here electrons accumulate, generating increasing electric field E , until its
force exactly balances the Lorentz force:

© K. Likharev, 2008 11
Quantum Mechanics

qEy  qvx B (3.64)

where vx is the drift velocity of the electrons along the bar (Fig. 6), providing the sustained balance
condition Ey/vx = Bz in each point of the 2D sample.

y
W

E   I
v, j


B Fig. 3.6. The Hall effect geometry. The
darker bars show external (3D) electrodes.
0 x

With n2 carriers per unit area, in a sample of width W (Fig. 6), this gives the following classical
expression for the so-called Hall resistance RH:
Vy EyW B
RH    . (3.65)
Ix qn2 v xW qn2
This formula is broadly used in practice for the measurement of the carrier density n2, and (in
semiconductors) the carrier type – negative electrons or positive holes. However, in experiments at low
temperatures and high-quality (low-defect) 2D structures,8 the linear growth of RH with B, described by
Eq. (65), is interrupted by horizontal plateaus (Fig. 7) with constant values
1
R H  RQ , (3.66)
i

RH

Fig. 3.7. A typical record of the quantum


Hall effect. (Copied from Web site
www.prequark.org/Prequark.htm.)

B
where i (only in this context, following tradition!) is an integer, and RQ  25.81281 k is called the
“quantum unit of resistance”.

8 Very recently, the quantum Hall effect was observed at room temperature in grapheme (virtually perfect 2D
sheets of carbon atoms) – see K. S. Novoselov et al., Science 315, 1379 (2007).

© K. Likharev, 2008 12
Quantum Mechanics

Let us see what does the Landau level picture give. The 2D sample is typically in a weak contact
with 3D electrodes whose conductivity electrons form a “Fermi sea” with certain Fermi energy EF, so
that the states with E < EF are filled with electrons – see Fig. 5b. As B is increased, fewer and fewer of
these (lowest) levels are below EF and are thus filled, and within broad ranges of the field change, the
number i of filled levels is constant. (In Fig. 5b, i = 2.) In this case, plugging n2 = inL and q = e into
Eq. (65), we get
1 B 1 2 1 h
RH    , (3.67)
i qn2 i q 2 i q2
i.e. exactly the experimental result (66), with
h
RQ  . (3.68)
e2
However, this oversimplified explanation of the quantum Hall effect does not take into account
several important factors essential in experiment, including:
(i) the role of localized states which are always present due to potential energy nonuniformity,
and their surprisingly small effect on the extraordinary (~10-8) accuracy with which relation (67) holds
in real samples;
(ii) the contribution of Landau-level electrons into their potential energy (due to their electric
charge), and
(iii) the mutual Coulomb interaction of the electrons, in high-quality samples leading to the
formation of RH plateaus with not only integer, but also fractional values of l (1/3, 2/5, 3/7, etc.) – the
so-called “fractional quantum Hall effect”, discovered in 1982 by D. Tsui, H. Stormer, and A. Gossard.
The discussion of all these features is well beyond the framework of our course. 9

3.3. Scattering and diffraction


Now, after the detour, let us return to the discussion of quantum interference, and extend it to
diffraction which is essentially the interference of several (many) coherent sources. Just as in the
double-slit experiment, these sources are most frequently the elementary scatterers of some initial wave.
In this context, let us discuss the general scattering situation (Fig. 8), focusing on the most common case
when the detector of scattered particles is located on a large distance, r >> a, from the scatterer of size ~
a. (In optics, this limit is called Fraunhofer diffraction.) In this case, the main observable is the flux
(number per unit time) of particles scattered in a certain direction. If the scatterer does not change, the
flux is proportional to the incident flux of particles (per unit area), so the scattering intensity may be
characterized by the ratio of these two fluxes, which has the dimensionality of area per unit angle, and
called the differential cross-section of the scatterer:
d flux of scatterd particles (per unit solid angle)
 . (3.69)
d flux of incident particles (per unit area)

9 See, e.g., D. Yoshioka, The Quantum Hall Effect, Springer, 1998, or D. R. Yennie, Rev. Mod. Phys. 59, 781
(1987).

© K. Likharev, 2008 13
Quantum Mechanics

r  a

k' detector
 a
k scattered
b particles

incident scatterer
particles Fig. 3.8. 3D scattering (schematically).

These name and notation stem from the fact that the integral of d/d over all scattering angles,
d total flux of scattered particles
  d  , (3.70)
d incident flux in per unit area

has the dimensionality of area, and the simple interpretation as the full (effective) cross-section. For the
simplest case when an object scatters all classical particles physically hitting it, but does not affect the
particles flying by,  is just the object’s cross-section as visible from the direction of incoming
particles.
In classical mechanics (say, in the simplest task of derivation of the “Rutherford” formula for the
scattering of a point electric charges by a point fixed charge – see, e.g., CM Sec. 4.6), we first calculate
the particle scattering angle as a function of the impact parameter b (Fig. 8), and then average the result
over all values of b, considered random. In this sense the calculations in wave mechanics are simpler,
because if we present the parallel beam of incident particles by a plane wave
 
 i  Ci e  r ,
i k (3.71)
with the free-space energy
 2k 2
E , (3.72)
2m
and constant probability current density
 2  
ji  C i k, (3.73)
m
the wave already

presents a uniform distribution of the particles over the area perpendicular to their
wavevector k .
Now let us assume, for a minute, that the scattered wave, at large distances r >> a, k-1, has the
asymptotic form
 
f (k , k ' ) ik ' r
 s  Ci e , (3.74)
r

© K. Likharev, 2008 14
Quantum Mechanics

 
where k ' is the wavevector parallel to the direction r toward the observation point. (For the elastic
  
scattering we will consider, k’ = k, so that we can write k '  knr  k (r / r ) .) Generally, to calculate the
corresponding current density, we need to use Eq. (1.73), with the gradient operator, for example, in the
spherical-coordinate form
    1   1 
  nr  n  n . (3.75)
r r  r sin  
However, due to the condition kr >> 1, we can restrict the differentiation of Eq. (74) to the fast-changing
exponent, getting
  2
  2
f ( k , k ') 
js ( )  C i k', (3.76)
m r2
so that the total flux of scattered particles through a sphere of radius r is
   2
J s    j s n d 2 r  C i k  f (k , k ' ) d .
2
(3.77)
A r 2
m

From here, the total cross-section of scattering


Js   2
   f (k , k ' ) d . (3.78)
ji

Comparing this expression with Eq. (70), we get


d   2
 f (k , k ' ) . (3.79)
d
Let us see whether the Schrödinger equation really gives solution of the form (74) at large
distances, or this is just our wishful thinking. In spherical coordinates, the Laplace operator is
1   2  1     1 2
2   r    sin    . (3.80)
r 2 r  r  r 2 sin      r 2 sin 2   2
First, let us assume that the wavefunction is purely spherically-symmetric; in this case the Schrödinger
equation (for a free particle) becomes


2 1 2
2m r 2
r  ''  E , (3.81)

where prime denotes differentiation over r. We can always look for the solution of this equation in the
form
f (r )
 (r )  ; (3.82)
r
plugging it into Eq. (81), and taking into account Eq. (72), we get a very simple differential equation
f " k 2 f  0 , (3.83)

© K. Likharev, 2008 15
Quantum Mechanics

whose general solution (a linear combination of two waves exp{ikr}), we already know very well. As a
result, the general solution of Eq. (81) is
exp{ikr} exp{ikr}
 ( r )  C  C . (3.84)
r r
The first term describes a spherically-symmetric wave propagating from the center outward, while the
second one, a wave converging onto the origin from large distances. Though such solution is possible at
some special circumstances (say, when the outgoing wave is reflected back from a spherical shell), for
scattering by a limited-size object in otherwise free space, only the outgoing waves are relevant.
Now, wave (74) still differs from the spherically-symmetric solution by its angular dependence f.
However, it is evident from Eq. (80) that if f is a smooth function of  and , the contributions of its
angular derivatives to 2 drop as r-2 and are negligible at sufficiently large distances. Hence, Eq. (74)
is indeed a legitimate asymptotic form of the Schrödinger equation solution, and what remains is to
calculate function f.10
Let us write the Schrödinger equation for the scattering problem in the form
 

Hˆ 0  U (r )   E , (3.85)
where
2 2  2k 2
Hˆ 0    , and E , (3.86)
2m 2m
and present the wavefunction as a sum of the incident (i) and scattered (s) parts.11 Taking into
account that the incident wave has to satisfied the free-space equation
Hˆ 0 i  E i , (3.87)
Eq. (85) yields
Hˆ 0 s  U i  U s  E s . (3.88)
The most straightforward (and common) simplification of this problem is possible if the
scattering potential U is in some sense weak. (We will derive the exact condition of this smallness
below.) Let it be proportional to some small parameter  : U  1  0. Then since at  = 0 the scattering
wave disappears, we may expect that at small but nonvanishing , the main part of s is proportional to
. Then all terms in Eq. (88) are proportional to , besides the product Us which is proportional to 2.

10 Notice that function f has the dimension of length, and does not account for the incident wave. This is why in
the general theory of scattering, it is more common to work with dimensionless function S = 1 + 2ikf which is
called the scattering matrix, and is a natural generalization of the 1D matrix defied by Eq. (2.107).
11 An alternative approach is to present  in the following symbolic form  = U(E – H0)-1 + i, where operator
(E – H0)-1 is defined by the identity (E – H0)(E – H0)-1  I. Taking into account that the incident wave i obeys
Eq. (87), it is straightforward to check that  is indeed a solution of Eq. (85). This solution is indeed useful for
some applications, especially in the quantum field theory, however, is has to be improved to avoid singularities in
denominator. Such improvement results in the so-called Lipmann-Schwinger equation. Unfortunately, we will
have not time to discuss this approach, but in the limit U  0, it yields the same Born approximation which we
will explore now.

© K. Likharev, 2008 16
Quantum Mechanics

Hence, in the first approximation that term may be ignored, and Eq. (88) reduces to the famous equation
of the Born equation:
Hˆ 0 
 E  s  U i . (3.89)
A bit surprisingly, this equation may be readily solved (or rather reduced to an integral) for an arbitrary

function U (r ) . Indeed, after rewriting it as

 2
 k 2  s 
2m 
2

U (r ) i (r ) , (3.90)

we notice that s is just a response of a linear system to a certain “excitation” function (the right hand
part) which is fixed, i.e. does not depend on the solution. Hence we can use the linear superposition
principle to present s as a sum of responses to elementary excitations from elementary volumes d3r’:
 2m    
2 
 s (r )  U (r ' ) i (r ' )G (r , r ' )d 3 r ' . (3.91)

 
Here G (r , r ' ) is the Green’s function,12 defined as such an elementary response, i.e. the solution of the
following equation
 2r  k 2 G   (r  r' ) . (3.92)

But this equation is spherically-symmetric with respect to the fixed point r ' , and we already know its
relevant solution, describing the spherically-symmetric outgoing wave, is given by the first term of Eq.
  
(84), with r replaced by R – the magnitude of vector R  r  r ' :
 exp{ikR}
G ( R)  C , (3.93)
R
so that we need just to calculate coefficient C.
This can be done in several ways, for example by noticing that at r << k-1, the second term in Eq.
(3.92) is negligible, and it is reduced to the well-known Poisson equation which describes, for example,
the electrostatic potential generated by a point electric charge. Either recalling the electrostatics, or
applying the Gauss theorem, we readily get the asymptote
1
G , at kr  1, (3.94)
4R
which is compatible with Eq. (93) only if C = -1/4, i.e.
 exp{ikR}
G ( R)   . (3.95)
4R
Plugging this result into Eq. (91), we get the full solution of Eq. (90):

12 Please notice both the similarity and difference between this Green’s function and that (“the propagator”)
discussed in Sec. 2.1. In both cases, we can use the Green’s function approach due to the equation linearity, but
while Eq. (91) expresses the sum of elementary components of the right hand part of an inhomogeneous equation
(90), Eq. (2.21) expresses the sum of all elementary initial conditions of a homogeneous Schrödinger equation.

© K. Likharev, 2008 17
Quantum Mechanics

 
 m   exp{ik r  r ' } 3
2 2 
 s (r )   U (r ' ) i (r ' )   d r' . (3.96)
r  r'
Notice that the singularity in the denominator is integrable (i.e. not dangerous); indeed, the contribution

of a sphere of radius   0, with the center in point r ' , scales as
  
d 3R R 2 dR
0 R  4 0 R  4 0 RdR  2  0. (3.97)

Actually, Eq. (96) gives us a bit more than we wanted: the scattered wave at any point, including
those within of the scattering object, while our goal was to find the wave far from the scatterer.
However, before going to that limit, we can use this equation to find the quantitative criterion of the
Bohr approximation’s validity. Indeed, let us estimate the magnitude of the right hand part of this
equation, for a scatterer of linear size ~a, and the potential magnitude scale U0, in two limits:
(i) If ka << 1, then both i ~ exp{ikr} and the second exponent under the integral change slowly,
so that a crude estimate is
m
s ~ U0  i a2 . (3.98)
2 2

(ii) In the opposite limit ka >>1, integration in one of dimensions (that of the wave propagation)
if cut out on distances of the order of the de Broglie wavelength k-1, so that the integral is
correspondingly smaller:
m a2
s ~ U 0  i . (3.99)
2 2 ka
Since the reduction of Eq. (88) to Eq. (89) required s<<i, we may now formulate the conditions of
this requirement as
2
U 0  max[ka, 1] . (3.100)
ma 2
In the first factor, we readily recognize the kinetic (quantum-confinement) energy Ea of an electron
inside a quantum well of size ~ a, so that the Born approximation is valid essentially if the potential
energy of particle’s interaction with the scatterer is smaller than Ea. Notice, however, that estimates (98)
and (99) are not valid if effects of scattering accumulate in some direction. This is frequently the case
for small scattering angles in extended objects (when ka >> 1 but ka < 1), and especially in 1D (or
quasi-1D) scatterers oriented along the incident particle beam – see, e.g., Homework 7.3.
Now let us proceed to large distances r >> r’ ~ a, and simplify Eq. (96) using an approximation
similar to the dipole expansion in electrodynamics. In denominator, we can merely ignore r’ in
comparison with r, but the exponent requires more care, because even if r’ << r, the product kr’ ~ ka
may still be larger than 1. In the first approximation in r’, we can take (Fig. 9a):
   
R  r  r '  r  nr  r ' , (3.101)
 
and since directions of vectors r and k ' coincide,

© K. Likharev, 2008 18
Quantum Mechanics

 
kR  kr  k 'r ' . (3.102)
With that,
   
ik r  r ' ikr  i k 'r ' ,
e e e (3.103)
and the incident wave in form (71), the Born approximation yields
  
 m exp{ikr}   i (k ' k )  r ' 3
 s (r )  C i
2 2 r  U ( r ' )e d r' . (3.104)

(a) (b)

 R 
r' detector k' 
 q
r 
a 
0 0 
k
scatterer

Fig. 3.9. (a) The dipole-like expansion in the Born approximation and (b) the definitions of vector q and
angles  and .

But this solution has exactly the form of our assumption (74), with
   
m   iq  r 3
2 2 
f (k , k ' )   U ( r )e d r, (3.105)
 
where for the notation simplicity we have replaced r ' with r , and introduced the scattering vector
  
q  k 'k , (3.106)
 
with length q = 2k sin(/2), where  is the “scattering angle” between vectors k ' and k (Fig. 9b). For
the differential cross-section, Eq. (105) yields
2  
d  m    iq  r 3 2
d  2 2  
  U ( r )e d r , (3.107)

and the total cross-section may be now readily calculated from Eq. (70). Notice the independence of the
differential (and hence full) cross-sections of the sign of potential U.
Equation (107) is extremely simple and physically transparent: it shows that the scattering in a
certain direction is completely determined by the spatial Fourier harmonic of potential U, with the wave

vector equal to the scattering vector q . This expression may be further simplified for spherically-
symmetric objects,

U (r )  U (r ). (3.108)
Here, it is convenient to present the exponent in the Born integral as exp{-iqr’cos}, where  is the
 
angle between vectors k ' (i.e. the direction r toward the detector) and q (but not incident wave

© K. Likharev, 2008 19
Quantum Mechanics

 
direction k !) – see Fig. 9b. Now, for fixed vector q , we can take this vector’s direction as the polar axis
of a spherical coordinate system, and reduce Eq. (105) to a 1D integral:
  2 
m
2 2 
f (k , k ' )   r drU (r )  d  sin d exp{iqr ' cos  }
2

0 0
(3.109)

m
2 2
 
2 sin qr 2m
 r drU (r ) 2 qr    2 q  U (r ) sin(qr )rdr.
2

As a simple example, let us use the Born approximation to analyze scattering on the following
spherically-symmetric potential:
 r2 
U  U 0 exp 2  . (3.110)
 2a 
In this particular case, it is better to avoid the temptation to exploit the spherical symmetry by using Eq.
(109), and instead use the generic Eq. (105), breaking the integral into three similar Cartesian factors:
  mU 0
f (k , k ' )   IxIyIz , (3.111)
2 2
with

  x2 
Ix     2a 2  iq x x dx,
exp (3.112)
 
and similar integrals for Iy and Iz. From Chapter 2, we already know that “Gaussian” integrals like Ix
may be readily worked out by complementing the exponent to the full square; in our current case giving

 q 2a 2   q y2 a 2   q z2 a 2 
I x  2 a exp x , I y  2 a exp , I z  2 a exp  , (3.113)
 2   2   2 

so that, finally,
2 2
2  mU 0 a   q2a 2
d   2  mU 0  2
 f (k , k ' )   I I I 
2 x y z 
 2 a   e . (3.114)
 2 
d  2    
Now, the total cross-section  is an integral of d/d over all directions of vector k’. Since in our case
the scattering intensity does not depend on the azimuthal angle , the integration is reduced to that over
the scattering angle  (Fig. 9b):
2
d

d 2 2  mU 0 a
2
     2 
2

   d  2  sin  d  4π a   0 sin  d exp  2k sin 2  a 


d 0
d  
2
  
(3.115)
2   2
 mU 0 a 2  2π  mU 0 a 2 
 
2
    4k a .
2 2
 4π 2 a 2           
1  e
2 2
2
exp 2 k a 1 cos d (1 cos ) 2  2 
    0 k   

© K. Likharev, 2008 20
Quantum Mechanics

Let us analyze these formulas. In the low-energy limit, ka << 1 (and hence qa << 1 for any
scattering angle), the scattered wave is virtually isotropic: d/d  const – a very typical feature of
scattering, in any approximation. Notice that in this limit, the Born expression for ,
2
 mU 0 a 2 
  8 a 
2 2
2
 ,
 (3.116)
  
is only valid if it is much smaller than the “visible cross-section” ~a2 of the scatterer.
In the opposite, high-energy limit ka >>1, the scattering is dominated by small angles   q/k ~
1/ka ~ /a:
2
d 2  mU 0 a   2
2
 2a   exp k 2 a 2
 . (3.117)
d  
2
  2
This is, again, very typical for diffraction. Notice, however, that due to the smooth character of the
Gaussian potential (110), the diffraction pattern exhibits no oscillations; such oscillations of d/d as
function of angle naturally appear for potentials with sharp border – see, e.g. Homework Problem 7.2.
The Born approximation, while being very simple and used more often than any other scattering
theory, is not without shortcomings. First, it cannot describe such a straightforward effect as a dark
shadow of an opaque object (say, with U >> E). On a more fundamental level, another drawback is
important. It is not difficult to prove the following general “optical theorem” which should be valid for
an arbitrary scatterer:
k
Im f q 0  . (3.118)
4
However, Eq. (109) shows that in the Born approximation, function f is real for any q for at least any
spherically-symmetric scatterer, and hence cannot satisfy the optical theorem.
There are several ways to improve the Born approximation, while still holding the general idea
of approximate treatment of U.
(i) Instead of our main assumption  = i + s, with s  1, we can use a complete perturbation
series (cf. Chapter 5 below):
   i   1   2  ... (3.119)

with j  j, and find successive approximations j one by one. In the first approximation we of course
return to the Born formula, but already the second approximation yields
k
Im f 2 q 0  1 , (3.120)
4
so that the optical theorem is “almost” satisfied.
(ii) As has mentioned above, the Born approximation does not work well for small-angle
scattering by extended objects. This deficiency may be corrected by the so-called “eikonal
approximation” (from the Greek , meaning “icon”) which replaces the plain-wave exponent

© K. Likharev, 2008 21
Quantum Mechanics

exp{ikx} representation of the incident wave by a straight-line WKB-like exponent, though still in the
first nonvanishing approximation in U 0:
 x   x 2mE  U ( x)  m
x
e ikx  expi  k ( x' )dx   expi  dx   e ikx  2  U ( x' )dx' . (3.121)
      k

This approximation gives results which satisfy the optical theorem already in the first approximation in
U. Unfortunately, in this course we will not have time to consider these and other approaches to
scattering in detail.

3.4. Energy bands in higher dimensions


In Sec. 2.5, we have discussed, in substantial detail, the 1D band theory, for a potential profile
which obeys the periodicity condition (2.166). Notice that this condition may be rewritten as
  
U (r   )  U (r ) , (3.122)
with
 
  a . (3.123)
where  is any integer. However, the most important example of periodic systems, crystals, have 3D
periodicity, and we will now discuss, in brief, how the band theory may be generalized to this case.
The 3D periodicity may be also described by Eq. (122), but with Eq. (123) generalized as
   
   1 a1   2 a3   3 a3 , (3.124)
 
with three integer numbers j and a set of three “primitive vectors” a j . The system of all points  ,
described by this equation, is called the Bravais lattice. Its simplest example is given by the “simple

cubic lattice” (Fig. 10a), which may be described by a system of mutually perpendicular vectors a j of
equal length. However, not in any lattice these vectors are perpendicular; Fig. 10b shows a possible set
of the primitive vectors describing the “face-centered cubic” lattice (fcc). The science of crystallography
(based on the group theory) distinguishes 14 topologically different Bravais lattices in 7 different
“crystal systems”.13

13 A clear, nicely illustrated introduction to crystallography, from the standpoint of physics rather than math, may
be found, for example, in Ch. 7 of N. Ashcroft and N. Mermin, Solid State Physics, Saunders College, 1976.

© K. Likharev, 2008 22
Quantum Mechanics

(a) (b)


a3 Fig. 3.10. The (a) simple cubic and (b)
 face-centered cubic Bravais lattices and
 a3  a possible choice of their primitive
 a2  a2
a1 a1 vectors.

Now we are ready for the formulation of the 3D version of the Bloch theorem – see Eqs. (2.167)
and (2.171). Any eigenfunction of the Schrödinger equation describing particle’s motion in the periodic
potential (122) may be presented either as
 
   iq  
 q (r  l )   q (r )e (3.125)

or as
 
  iq  r   
 ( r )  u q ( r )e

q
 , u q (r   )  u q (r ), (3.126)

where quasi-momentum q is again a constant of motion, but now it is a vector.
The next key notion of the band theory is the reciprocal lattice in the wavevector space:
   
g  2 (l1b1  l 2 b2  l3 b3 ) , (3.127)

with integer lj, and with vectors b j selected so that the following equality
 
ig  
e 1 (3.128)
The importance of this notion is immediately clear from the first formulation of the Bloch theorem, see
 
Eq. (125): if we add to the quasi-momentum q any vector g of the reciprocal lattice, the wavefunction
does not change. This means that all information about the system is contained in just one elementary
cell of the lattice, closest to the origin, called the 1st Brillouin zone.
Let us check that the primitive vectors of the reciprocal lattice may be constructed from those of
the initial, direct lattice as
  
a 2  a3
b1     , etc. (3.129)
a1  a 2  a3 
   
do satisfy Eq. (128). From Eq. (129), it is evident that a1  b1  1 , while a 2  b1  0, etc. Hence, the
exponent in the left-hand part of Eq. (128) is reduced to
 
ig  
e  exp2i l1 1  l 2 2  l3 3  . (3.130)

Since all lj and j are integers, the expression in parentheses is also an integer, so the exponent indeed
equals 1.

© K. Likharev, 2008 23
Quantum Mechanics

As the simplest example, consider for a simple cubic lattice of period a (Fig. 10), oriented so that
     
a1  an x , a 2  an y , a3  an z , (3.131)

According to Eq. (129), its reciprocal lattice is of course also cubic:


 2   
g (l x n x  l y n y  l z n z ) , (3.132)
a
so that the 1st Brillouin zone is a cube with a side 2/a. Figure 11 shows a more complex 1st Brillouin
zone of the fcc lattice.

qz

qy
qx
Fig. 3.11. The first Brillouin zone of the fcc
lattice, and its main directions. Adapted from
http://en.wikipedia.org/wiki/Band_structure.

The band theory in multiple dimensions is actually not much more complex than that in 1D;
what is more complex is the presentation (and hence the comprehension:-), of its result. Typically, it is
limited to plotting energy as a function of components of vector q along certain special directions the
reciprocal space of quasi-momentum – see, e.g., the lines shown in Fig. 11. Figure 12 shows possibly
the most famous (and certainly the most practically important) of such plots, the band structure of
crystallic silicon. The dashed lines show the “indirect” band gap of ~1.12 eV between the “valence” and
“conduction” bands, which is the playground of virtually all modern electronics.

Fig. 3.12. Band structure of silicon, along the special


directions shown in Fig. 11. (Copied from
http://en.wikipedia.org/wiki/Band_structure.) Silicon
has the “diamond” lattice, which may be interpreted as
the fcc with a two-atom “basis” (meaning that the
two-atom potential is repeated in accordance with rule
(122)). As a result, the direction and special point
nomenclature is the same as for the fcc lattice.

© K. Likharev, 2008 24
Quantum Mechanics

In order to understand the reason of this band structure complexity, let us see how we would start
building the weak-coupling approximation for the simplest case of a 2D square lattice (which is a subset
of the cubic lattice (131) for 3 = 0). Its 1st Brillouin zone is of course also a square, of area (2/a)2. Let
us draw the lines of constant energy of a free particle (U = 0) in this zone. Repeating the arguments of
Sec. 2.5, we may conclude that Eq. (2.189) should now be generalized as follows,

 2 k 2  2  
2
2
2 l x   2 l y 
E  E l x ,l y    q x     q y   , (3.133)
2m 2m  a   a  

with all possible integers lx and ly. Considering the result only within the first Brillouin zone, we see that
as energy E grows, the lines of equal energy evolve as shown in Fig. 13. Just like in 1D, the weak-
coupling effects are only important of the Brillouin zone boundaries, and may be crudely reduced to the
appearance of narrow energy gaps, but one can see that the band structure is complex enough even
without these effects.
qy (a) (b) (c)

Fig. 3.13. Lines of constant energy


2 E of a free particle, within the 1st
a 0 qx Brillouin zone of a square Bravais
lattice, for: (a) E/E1  0.95, (b) E/E1
 1.05; and (c) E/E1  2.05, where
E1  22/2ma2.
2 / a

The tight-binding approximation is usually easier. For example, for the square 2D lattice, we
may repeat arguments which have led us to Eq. (2.179) to write14
da j ,k
i   n a j 1,k  a j 1,k  a j ,k 1  a j ,k 1  , (3.134)
dt
where indices j and k list x- and y-positions of the potential minima located in Bravais lattice nodes (and
hence wavefunction “humps” which are quasi-localized at these minima). Now, looking for the
stationary solution of these equations, which corresponds to the Bloch theorem (128),

a j (t )  expi q x x j  q y y k   i t  const  ,
 
(3.135)
  
instead of Eq. (2.182) we get
E  E n    E n  2 n cos q x a  cos q y a . (3.136)

Figure 12 shows this result, within the 1st Brillouin zone, in two popular forms: as the color-
coded lines of equal energy and as a 3D plot. One can see that on the zone boundaries, the slope of the
energy profile is horizontal:

14 Actually, using the same values of n in both directions implies some sort of symmetry of the quasi-localized
states. For example, s-states (see Sec. 6 below) always have such symmetry.

© K. Likharev, 2008 25
Quantum Mechanics

E n
Brillouin zone boundary  0. (3.137)
q n
As one could expect from our 1D results (see, e.g., Figs. 2.26, 2.28, 2.29), this result is general and does
not depend on coupling strength – see, e.g., Fig. 12.

qy 
 4 n

2 qy
0 qx 0
a
Fig. 3.14. Allowed band
energy  = E – En for a square
 4 n qx 2D lattice, in the tight-binding
approximation.

2 / a

3.5. Central- and spherically-symmetric systems: The brute force approach


Most multi-dimensional problems of wave mechanics do not allow analytical solutions.
Substantial relief may come from symmetry, in particular the central symmetry in 2D problems and the
spherical symmetry in 3D problems. In rare cases such symmetry may be exploited by the separation of
Cartesian variables. The famous example is the 3D harmonic oscillator, i.e. a particle moving inside the
potential:15
m 02 2
U r . (3.138)
2
Presenting the potential as
m 02 2
U  Ux U y Uz , with U x  x , etc., (3.139)
2
and separating the variables exactly as we did for the rectangular quantum well (see Eqs. (1.55)-(1.58)),
for each degree of freedom we get the Schrödinger equation (2.229) of a 1D oscillator, whose solution
are eigenfunctions (2.241) and the energy spectrum (2.228). As a result, the total energy spectrum may
be indexed by three independent quantum numbers:
 3
E nx ,n y ,nz   0  n x  n y  n z   , (3.140)
 2

15 The corresponding 2D problem is solved absolutely similarly, giving Enx,ny = 0(nx + ny + 1).

© K. Likharev, 2008 26
Quantum Mechanics

all of them ranging from 0 to . Notice that every energy level of this system, with the only exception of
the ground state
3/ 4
 m 0   r2 
 0 ( x)  X 0 ( x)Y0 ( y ) Z 0 ( z )    exp 2  , (3.141)
    2 x0 
is degenerate: different wavefunctions, with different sets of quantum numbers but the same value of
their sum, have the same energy.
However, the oscillator problem is an exception; for other central- and spherically-symmetric
problems the solution is made easier by using more appropriate coordinates. Let us start with the
simplest axially-symmetric problem: a “2D rotator”, i.e. a particle constrained (quantum-confined) to
move along a round circle of radius R (Fig. 15).16

R l  R

0
Fig. 3.15. 2D rotator.

We can, of course, solve this problem as a 1D one, taking

ˆ pˆ 2 
H , with pˆ  i , (3.142)
2m l
where l = R is the arc passed by the particle (Fig. 15). This Hamiltonian has the same form as that for
the straight, free 1D motion, and hence has the same eigenfunction structure:

  Ce ikl . (3.143)
The only new feature is that all observables should be 2R-periodic functions of l, and hence, as we
have already discussed in the context of flux quantization, the wavefunction’s phase kl has to change by
at most 2n, with an arbitrary integer n (from - to +), as the particle makes one turn about the center:

 n (l  2R )   n (l )e 2in . (3.144)

With wavefunctions (143), this gives k2R = 2n. Thus, wavevector k may take only quantized values
kn = n/R, so that eigenfunctions (143) should be indexed by n:
 l
 n  C n expin  , (3.145)
 R
and the energy spectrum is discrete:
p n2  2 k n2 2n2
En    . (3.146)
2m 2m 2mR 2

16 This is a reasonable model for the confinement of light atoms, notably hydrogen, in some organic compounds.

© K. Likharev, 2008 27
Quantum Mechanics

Now let us consider properties of the 2D rotator (carrying electric charge q) in a uniform
magnetic field perpendicular to the rotation axis – see Eq. (53). According to Eq. (33), in this case we
have to generalize Eq. (142) as

1     2
Hˆ    in  qA  . (3.147)
2m  l 
Now, in contrast to the gauge choice (55), which was so instrumental in the Landau level problem, it is
clearly beneficial to take the vector-potential in a manifestly-symmetric form
 
A  n A(  ) , (3.148)

where   x, y is the 2D radial coordinate. Vector algebra says that for the axially-symmetric field,
   1 
  A  nz A  , (3.149)
 
and we can readily check that Eq. (53) may be satisfied with
  B
A  n . (3.150)
2
For the 2D rotator, ρ = R = const, so that the Schrödinger equation becomes
2
1   BR 
  i  q   n  E n n . (3.151)
2m  l 2 
A little bit surprisingly, we see that this equation is still satisfied with the sine-wave
eigenfunctions (143). Moreover, since the periodicity condition (144) is also unaffected by the applied
magnetic field, we return to the eigenfunctions (146). However, the field does affect the system’s
energy:
2 2
1  n BR  2   
En    q    n   , (3.152)
2m  R 2  2mR 2  0 
where   R2B is the magnetic flux through the area limited by the particle’s trajectory, and 0  h/q is
the “non-superconducting” magnetic flux quantum - see Sec. 2 above. The field also changes the electric
current of the particle in n-th state:
  *   iqRB    2   
In  q n    n  complex conjugate   q Cn  n   . (3.153)
2im   l 2   mR  0 
Normalizing wavefunctions (145) to have P = 1, we get Cn 2 = 1/2R, and
   q
I n  I 0  n  , I0  . (3.154)
 0  2mR 2

© K. Likharev, 2008 28
Quantum Mechanics

Notice that since 0  1/q, for any sign of the particle’s charge, dIn/d <0. It is easy to check that this
means that the current is “diamagnetic”:17 in correspondence with the Lenz rule, it flows in such
direction that its own magnetic field tries to compensate the external magnetic flux applied to the loop.
The functions En() and In () are shown in Fig. 16.

En

n  1 n0 n  1
0 1  / 0
In Fig. 3.16. Effect of magnetic field on a
charged 2D rotator. Dashed lines show
possible transitions between metastable and
ground states, due to interaction with
environment, as the magnetic field is being
0 1  / 0 increased.
n  1 n0 n  1

It is easy to check that Eqs. (152) and (153) remain valid even if the magnetic field lines do not
touch the particle’s trajectory, and the field is localized well inside the ring. So, these results may be
interpreted as just another manifestation of the AB effect. However, in contrast to the two-slit
interference experiment discussed in Sec. 2, in the situation shown in Fig. 15 the particle is not absorbed
by the detector, but travels around the ring continuously. As a result, its wavefunction is “rigid”: due to
the boundary condition (144), the topological quantum number n is discrete, and magnetic field cannot
change the wavefunction gradually. In this sense, the system is similar to the superconducting loop (see
Fig. 4 and its discussion). The difference between these systems is two-fold:
(i) For a single charged particle, with practicable values of q, R, and m, the current scale I0 is
very small: for example, for m = me, q = e, and R = 1 m, Eq. (154) yields I0  3 pA. The contribution LI
~ 0RI0 ~ 10-24 Wb of the current so small into the net magnetic flux is negligible in comparison with 0
~ 10-15 Wb, so that the quantization of n does not lead to the magnetic flux quantization.18 Some authors
describe this situation as quantization of “fluxoid” n0, with the diamagnetic current I (153) proportional
to the difference between the real magnetic flux and the current value of the fluxoid.
(ii) As soon as the magnetic field raises the eigenstate energy En above that of another eigenstate
En’, the former state becomes metastable, and certain interactions of the system with its environment
(which is neglected in our simple model) may induce a quantum transition of the system to the lower

17 This effect is frequently referred to as the orbital diamagnetism.


18 The persistent diamagnetic currents I may be experimentally observed, for example, by measuring the weak
magnetic field generated by electrons in a system of a large number (~107) of similar conducting rings – see L.
Levy et al., Phys. Rev. Lett. 64, 2074 (1990). Due to the dephasing effects of mutual electron scattering, and their
scattering by phonons, the effect is only observed in submicron samples, at millikelvin temperatures.

© K. Likharev, 2008 29
Quantum Mechanics

state, thus reducing its energy and the diamagnetic current’s magnitude – see the dashed lines in Fig.
16.19
Now let us return to Eq. (145), and notice that the eigenfunctions may be presented just as
functions of angle  = l/R (Fig. 13):

 n   n    C n e in , (3.155)

with the only difference that the normalization constant is now different: Cn 2 = 1/2. This equation
hints that we could get these results in conceptually simpler way by exploiting the classical notion of
angular momentum20
  
L  r  p, (3.156)

and its extension to quantum mechanics. In our current problem, vector L evidently may have just one
component,
L z  Rp , (3.157)
in the direction perpendicular the rotator plane, so that our result kn = n/R may be rewritten as
L z  ( L z ) n  Rk n  n . (3.158)
So, the angular momentum (at least in this geometry) is naturally quantized in units of Planck’s constant
. With that, Eq. (146) may be also rewritten in the angular form:
( Lz ) 2n
En  . (3.159)
2I
where I  mR2 is the moment of inertia of the rotator.
The evident similarities between Eqs. (143) and (155), and Eqs. (146) and (159), suggests that
the system’s Hamiltonian may be presented as
Lˆ2
Hˆ  z , (3.160)
2I
with Lz considered as a quantum-mechanical operator of angular momentum (or rather of its z
component),

Lˆ z  i . (3.161)

Indeed, considering n as functions of , which should satisfy the corresponding Schrödinger equation,
 2 d 2 n
  E n n , (3.162)
2mR 2 d 2

19 Some weakly coupled superconducting systems (e. g., the so-called SQUIDs - superconducting loops
interrupted by a Josephson junction) behave in the same way – see, e.g., EM Sec. 6.3.
20 See, e.g., CM Sec. 1.4.

© K. Likharev, 2008 30
Quantum Mechanics

we readily recover results (155) and (159), without using arc or any other linear coordinate. However,
we will postpone the active using of the notion of angular momentum until the next chapter, and for now
will continue our “brute force” approach based on wave mechanics.
Let us proceed to the more general problem of a 2D particle moving in a centrally-symmetric
potential

U ( )  U ( ) . (3.163)
In polar coordinates, the 2D Laplace operator is
1     1 2
2    , (3.164)
      2  2
so that the stationary Schrödinger equation takes the form
2  1     1 2 
      U (  )  E . (3.165)
2m        2  2 

(At this stage, I do not want to mark eigenfunctions  and eigenenergies E by any index, because we
already may suspect that in a 2D problem that the role of this index will be played by two integers -
quantum numbers.) Separating the radial and angular variables as
  R (  ) F ( ) , (3.166)
we get, after the division by  and multiplication by ρ2, equation
2  
  R '' F "    2U (  )   2 E , (3.167)
2m  R F
where each prime means the differentiation of every function over is only argument. It is clear that the
fraction F”/F should be a constant (because all other terms of the equation should be functions of ρ
alone, so that we get for function F() an ordinary differential equation identical to Eq. (162). Since this
function should again satisfy the periodicity condition identical to Eq. (144),

F (  2 )  F ( )e 2in , (3.168)
we again get a result identical to Eq. (155):
in
Fn  C n e , i.e. Fn" / Fn  n 2 , (3.169)
with integer quantum number n. Returning to Eq. (167), we may now rewrite is as
2  1 n2 
   R ' '  U ( )  E . (3.170)
2m  R  2 
The physical interpretation of this equation is that the full energy is a sum,
E  E   E , (3.171)

of the angular part

© K. Likharev, 2008 31
Quantum Mechanics

E 
 2n2

L2z n , (3.172)
2m 2 2m 2
and the radial part
2 1
E   R ''  U (  ) . (3.173)
2m R
The similar separation exists in classical mechanics, because the total energy of a particle moving in a
central field may be presented, within the plane of motion, as
L2z  p 2 
E  v  U (  )       U (  ) 
m 2 m 2 2
   U (  ) . (3.174)
2 2 2m 2  2m 
Notice, however that (just as in classical mechanics) the components Eρ, E are not integrals of motion,
only their sum is.
Returning to Eq. (170), on the basis of our experience with 1D wave mechanics, we expect that
this ordinary, linear, second-order differential equation should have (for a motion confined to a certain
final region of ρ) a discrete energy spectrum described by some integer quantum number (say, l),
independent on n. This means that eigenfunctions (166), and corresponding eigenenergies (171) should
be indexed by two quantum numbers. Notice, however, that since the radial function obeys equation
(170) which already depends on n, R(ρ) should carry two indices, so the variable separation is not so
“clean” as it was for the rectangular quantum well (cf. Eq. (1.63)):
1 in
 n ,l  Rn,l (  ) Fn ( )  Rn,l (  )e . (3.175)
2
A good (and important) example of a solvable problem of this type is a free 2D particle confined
to a disk of radius R:
0, for 0    R,
U ( )   (3.176)
  , for R   .
In this case, the solutions to Eq. (170) are Bessel functions Jn(klρ), where the spectrum of possible
values of parameter kl should found the boundary condition R(R) = 0.
Leaving this problem for the reader (see Homework Problem 8.1), let me rush to the
(mathematically more involved) case of 3D motion, with spherically-symmetric potential

U (r )  U (r ). (3.177)
All the conceptual physics here remain the same, but math is more complex. Let us start, again, with a
rotator, now a 3D one, i.e. a particle confined to move on the surface of a sphere of radius R. Then the
kinetic energy is limited to its angular part, so that for the Laplace operator in spherical coordinates (80)
we take only that part, with fixed r = R. Then the stationary Schrödinger equation becomes
2  1     1 2 
   sin    2 
  E . (3.178)
2mR 2  sin      sin 2
  

© K. Likharev, 2008 32
Quantum Mechanics

(Again, I abstain from attaching any indices to  and E for the time being.) By the way, comparing this
equation with Eq. (172), it is evident that the expression is square brackets, multiplied by (-2), may be
considered as the quantum-mechanical operator of the square of angular momentum, now with arbitrary
orientation:
 1     1 2 
Lˆ2   2   sin   . (3.179)
 sin      sin 2   2 
However, as before, we will leave this observation alone until Chapter 4, and continue our analysis in
the purely wave-mechanics spirit.
With the usual variable separation assumption,
  ( ) F ( ) , (3.180)
Eq. (178) yields
2  1 1 F"

2mR 2   sin   ' sin  ' sin 2  F   E . (3.181)
 
Just as in the previous problem, the fraction F”/F has to be constant, giving for it an equation similar to
Eq. (162). So, the azimuthal functions are just sine waves (143) again, and we can repeat the same
periodicity condition (168) to write them in the form21
1 im
Fm ( )  e . (3.182)
2
With that, the fraction F”/F equals (-m2), and Eq. (181) is reduced to the following ordinary, linear
differential equation for function ():
2
 2 

1
 ' sin  ' m2   ,   E /  2 , (3.183)
sin  sin   2mR 
It is customary to recast it into an equation for P()  (), with   cos  :
 m2 
(1   2

) P ' ' l (l  1)  P  0 ,
1 2 
(3.184)

where prime now means the differentiation over , and a new notation for energy is introduced: l(l+1) 
. The sense of this notation is that, according to a mathematical analysis, Eq. (184) with integer m, has
solutions only for an integer l = 0, 1, 2,…, and only if that integer is not smaller than m, i.e. if
 l  m  l . (3.185)
This immediately gives the following energy spectrum of the 3D rotator:
 2 l (l  1)
El  , (3.186)
2mR 2

21 Here, very regrettably, I had to replace the notation of the integer from n to m, in order to comply with the
generally accepted convention for this “magnetic” quantum number.

© K. Likharev, 2008 33
Quantum Mechanics

so that the only effect of the magnetic quantum number m is imposing restriction (185) on the “orbital
quantum number” l. This means, in particular, that each of energy level (186) is (2l + 1) – degenerate.
To understand the physics of this degeneracy, we need to explore the corresponding
eigenfunctions of Eq. (184). They are naturally numbered by two integers, m and l, and are called the
associated Legendre functions Pl n . If m = 0, these functions are just (“Legendre”) polynomials which
may be defined either as the solutions of the “ordinary Legendre equation”:

(1   2

) P ' 'l (l  1) P  0 , (3.187)
or calculated explicitly from the following Rodrigues formula:
1 dl
Pl ( )  l ( 2  1) l , l  0, 1, 2,... . (3.188)
2 l! d l

Using this formula, we can readily derive explicit expressions for a few lowest Legendre polynomials:
P0 ( )  1,
P1 ( )   ,

P2 ( ) 
1
2

3 2  1 ,  (3.189)

1

P3 ( )  5 3  3 , ...,
2

though the expressions become more and more bulky as l is increased. As Fig. 17 shows, all these
functions start in one point, Pl(+1) = + 1, and end up either in the same point or in the opposite point:
Pl(-1) = (-1)l. On the way between these two end points, the l-th polynomial crosses the horizontal axis l
times. It can be shown that on the segment [-1,+1], these polynomials form a full orthogonal set of
functions, with the following normalization rule:
1
2
 P ( ) P ( )d  2l  1 
1
l l' ll ' . (3.190)

0.5

Pl ( ) 0
4
3
0.5
2
l 1
1 Fig. 3.17. A few lowest Legendre polynomials.
1  0.5 0 0.5 1
  cos 

© K. Likharev, 2008 34
Quantum Mechanics

For m > 0, the associated Legendre functions may be defined via these polynomials using the
following formula:
dm
Pl m  (1) m (1   2 ) m / 2 Pl ( ) , (3.191)
d m
while if the index m is negative, we can use the following simple relation:
(l  m)! m
Pl  m ( )  (1) m Pl ( ) . (3.192)
(l  m)!
On the segment  = [-1, +1], each set of the associated Legendre functions with fixed index m forms a
full orthogonal set, with the normalization relation,
1
2 (l  n)!
P ( ) Pl m' ( )dx   ll ' ,
m
l (3.193)
1
2l  1 (l  n)!
which is evidently a generalization of Eq. (190).
Since the difference between angles  and  is purely artificial, physicists prefer to use not the
functions () = Pl m (cos) and Fm() separately, but their products (180), which are called the spherical
harmonics:
1/ 2
 2l  1 (l  m)! im
Yl ( ,  )  
m
 Pl m (cos  )e . (3.193)
 4 (l  m)!
The specific coefficient in Eq. (193) is chosen in a way to simplify the following two relations: the
equation for negative m,

 
Yl  m ( ,  )  (1) m Yl m ( ,  ) ,  (3.194)
and the normalization relation

Y l
m
  
( ,  ) Yl m' ' ( ,  ) d   ll ' mm ' . (3.195)

The last relation shows that the spherical harmonics form an orthonormal set of functions of two
arguments  and ; so that any function defined on a sphere may be uniquely presented as their linear
combination. Despite a somewhat intimidating formulas given above, they yield rather simple
expressions for the lowest spherical harmonics:
l  0: Y00  1 / 4 , (3.196)

 Y 1   3 / 8 sin  e i ,
 1
l  1:  Y10  3 / 4 cos  , (3.197)
 1  i
 Y1   3 / 8 sin  e ,

© K. Likharev, 2008 35
Quantum Mechanics

 Y 2   15 / 32 sin 2  e 2i ,


 2

 Y2   15 / 8 sin  cos  e i ,
1


l  2 :  Y20  3 / 16 (3 cos 2   1), (3.198)
 1  i
 Y2   15 / 8 sin  cos  e ,
 2  2i 
 Y2   15 / 32 sin  e
2
.
It is important to understand the symmetry of these functions. Since they are complex, the most popular
way of their graphical representation is to plot one of their parts (e.g., the real one) as a distance from
the plot’s origin, while using bi-color to show the sign of this part – see Fig. 18.

Fig. 3.18. Several lowest spherical


harmonics. (Copied from Web site
http://people.csail.mit.edu/sparis/sh/.)

l = 0 (“s-state”): m=0

l = 1 (“p-states”): m = -1 m=0 m=+1

l = 2 (“d-states”):
m = -2 m = -1 m=0 m=+1 m = +2

Starting from the simplest case (l = 0), according to Eq. (185), there could be only one such “s-
state”,22 with m = 0. The spherical harmonic corresponding to that state is just a constant, so that the
wavefunction is uniformly distributed over the sphere. Since the functions does not have gradient in any
direction, the kinetic energy (186) of the particle equals is zero.
For l = 1, there could be 3 “p-states”, with m = -1, 0, and +1. As Fig. 18 shows, these states are
essentially identical, with equal energies, but differently oriented in space.

22 The “personal” names for states with different values of l come from the history of optical spectroscopy.

© K. Likharev, 2008 36
Quantum Mechanics

This is not quite true for 5 “d-states”(l = 2): while states with m  0 differ only by the special
orientation, the wavefunctions with m = 0 has more gradient in the  direction than in the  direction.
This trend continues for higher values of l: the states with m = 0 have gradient only in the  direction,
while the states with the ultimate values of m (l) change only gradually (as sinl) in the polar direction,
while strongly oscillating in the azimuthal direction. (The states with intermediate values of m provide a
crossover between these two extremes, oscillating in both directions, stronger and stronger in the
direction of  as m is increased.)
Let me conclude with the following statement: though our analysis of the 3D rotator is formally
(mathematically) complete, it is as unsatisfactory on the physics level, just as the harmonic oscillator
analysis in Sec. 2.6. In particular, it does not explain the meaning of extremely simple relations (185)
and (186) on the backdrop of rather complicated wavefunctions. Also, the relation between the
eigenvalues of the total angular momentum,
L2   2 l (l  1) , (3.199)
and the z-component of the momentum23
L z  m , (3.200)
remains unexplained. (For example, why for any state with l >0, (Lz)2 is less than L2? In other words,
what prevents the angular momentum vector to be fully oriented along axis z?)
We will obtain natural answers to these questions in Chapter 4, but for now let us complete our
survey of wave mechanics by extending it to 3D motion in an arbitrary spherically-symmetric potential
(177). In this case we have to use the full form (80) of the Laplace operator. The variable separation
procedure is an evident generalization of what we have done before:
  R (  )( ) F ( ), (3.201)
and gives
 2  (r 2 R ' )' 1 
 2 
  ' sin  ' 12 F "   U (r )  E . (3.202)
2mr  R  sin  sin  F 
It is evident that the angular part (the two last terms in square brackets) separate from the radial part,
and for them we have Eq. (178) again, with the only change, R  r. This change does not alter the fact
that the eigenfunctions of that equation are spherical harmonics (193), and the eigenenergies are given
by Eq. (186), again with the replacement R  r. This means that for the radial function we get equation
 2  (r 2 R ' )' 
 2 
 l (l  1)  U (r )  E. (3.203)
2mr  R 
Notice that no information about the magnetic quantum number m has crept into the radial equation, so
that this equation depends only on l.
This equation becomes relatively simple for U(r) = 0 within a certain range of r, for example,
for in case of a free 3D motion of a particle inside the sphere of radius R. Leaving that problem for the

23 Equation (200) follows from the calculation of the average value of operator (161) in state (193), but frankly
the very applicability of Eq. (161) to 3D motion is questionable at this stage.

© K. Likharev, 2008 37
Quantum Mechanics

reader’s exercise (see Homework Problem 8.2), I will proceed to the most important case of the “Bohr’s
atom”, i.e. the motion in a spherically-symmetric potential well with:24
C
U (r )   , C  0. (3.204)
r
The natural scales of r and E are, 25
2
2 C 
r0  and E 0  m  . (3.205)
Cm 
(Notice that these two units are naturally related as
2 C
E0  2
 ; (3.206)
mr0 r0
I can recommend this relation, without any coefficients, as the best mnemonic rule for both scales.) In
the normalized units   E/E0 and   r/r0, Eq. (204) looks simpler,
d 2 R 2 dR l (l  1)  1
   2   R  0, (3.207)
d 2
 d r  
but unfortunately its eigenfunctions may be called elementary only in the most generous meaning of the
term. With adequate normalization,

R
0
n ,l Rn ',l r 2 dr   nn ' , (3.208)

these (mutually orthogonal) functions may be presented as


1/ 2
 2  (n  l  1)! 
3
 r  2r
l
 2l 1  2r 
Rn ,l (r )    3
exp   Ln l 1  . (3.209)
 nr0  2n(n  l )!   nr0  nr0   nr0 

Here Lqp ( ) are the so-called associated Laguerre polynomials which may be derived as

dq
Lqp ( )  (1) q L p  q ( ) . (3.210)
d q
from “simple Laguerre polynomials” Lp().26 In turn, the easiest way to obtain Lp() is to use a
Rodrigues formula (named after the same person, and belonging to the same math class as Eq. (188)):

24 Historically, the solution of this problem, which has recovered the main result (1.8)-(1.9) of the “old” quantum
theory developed by N. Bohr in 1912, was the decisive step for the general acceptance of Schrödinger’s wave
mechanics.
25 For the most important case of the hydrogen atom, C = e2/40, these scales are reduced, respectively, to the
Bohr radius rB (1.13) and the Rydberg constant Ry (1.9).
26 In Eqs. (211)-(212), p and q are non-negative integers, with no relation whatsoever to particle’s momentum or
electric charge. Sorry for this notation, but it is absolutely common, and can hardly result in any confusion.

© K. Likharev, 2008 38
Quantum Mechanics

p
L p ( )  e
 d  p e   . (3.211)
 
d p  

Notice that in contrast with the associated Legendre functions Pl m (191) participating in spherical
harmonics, Lqp are just polynomials, and those with small indices p and q are really simple.

Returning to Eq. (209), we see that the natural quantization of the radial Eq. (207) has brought us
a new quantum number (integer) n. In order to understand its range, we may notice that according to Eq.
(212), the highest power of terms in polynomial Lp+q is (p + q), and hence that of Lqp is p, so that of the
highest power in polynomial L2nll11 is (n – l – 1). Since the power cannot be negative (to avoid the
divergence of wavefunctions at r  0), the radial quantum number n has to obey the restriction n  l +
1. Since l, as we already know, may take values l = 0, 1, 2,…, we conclude that n may only take values
n  1, 2,... (3.212)
What makes this relation important is the following, most surprising result of the theory: the
eigenenergies corresponding to wavefunctions (201), which are indexed with 3 quantum numbers:
 n ,l .m  Rn ,l (r )Yl m ( ,  ) , (3.213)

depend only on n and agrees with Bohr’s formula (1.8):


2
E 1 C 
E   02   2 m  . (3.214)
2n 2n   
Because of this reason, n is usually called the principal quantum number, and the above relation
between it and (more subordinate) l is rewritten as
l  n  1. (3.215)
Together with the similar result (186) for spherical harmonics, this gives us the well-known hierarchy of
the 3 quantum numbers:27
1  n    0  l  n  1   l  m  l , (3.216)
which may be also presented as the following classification of the lowest states:
n  1 : l  0 (1s state) m  0 . (3.217)
n  2: l  0 (2s state) m  0,
(3.218)
l  1 (2 p states) m  0,  1.

27 Taking into account the (2l +1)-degeneracy related to the magnetic number m, this means that each energy
n 1
level of the system is n2-degenerate:  (2l  1)  n
l 0
2
.

© K. Likharev, 2008 39
Quantum Mechanics

n  3: l  0 (3s state) m  0,
l  1 (3 p states) m  0,  1, (3.219)
l  2 (3d states) m  0,  1,  2,
etc.
This classification needs only the addition of the notion of spin and the Pauli principle, to explain, at
least on a semi-quantitative level, the whole periodic table of chemical elements.
Figure 19 shows plots of radial functions of the listed states. The most important of them is of
course the ground (1s) state with n = 1 and hence E = - E0/2, whose radial distribution is a simple
exponent,
2  r / r0
R1,0 (r )  3/ 2
e , (3.220)
r
0

and the angular distribution is uniform - see Eq. (196). It is very fortunate for physics that such simple
results describe the systems most frequently met in Mother Nature. Indeed, the gap between the ground
energy and the energy E = - E0/8 of the lowest excited states (with n = 2) in a hydrogen atom is as large
as ~ 10 eV, so that their excitation requires temperature as high as ~105 K, so that the overwhelming part
of all hydrogen atoms around us are in their ground state.

1 0.5
n 1 n2
1s (l  0)
2 p (l  1)

R1, 0 ) 0 R2 ,l 0

2 s (l  0 )

1  0.5
0 1 2 3 4 5 0 2 4 6 8 10
r/r0 r/r0
0.25
n3
3 p (l  1)
3d (l  2)

R3,l 0

Fig. 3.19. The lowest radial functions


3 s (l  0 )
of the Bohr atom problem. Notice the
change of vertical scale from frame to
frame.
 0.25
0 2 4 6 8 10
r/r0

© K. Likharev, 2008 40
Quantum Mechanics

The radial functions of the next states, 2s and 2p, are also not that complex:
1  r   r / 2r0 1 r  r / 2r0
R2, 0 (r )   2  e , R2,1 (r )  e . (3.131)
2r0  3/ 2
 r0  2r0  3/ 2
3 r0
(Notice again that the former of these states (2s) can only have a uniform angular distribution, while
three 2p states have different values of m = 0, 1, and hence have different angular distributions – see
Eq. (197) and the second row of Fig. 18. ) The most important trend here is the larger radial extension of
the states with larger n, which is only slightly affected by l:

r n ,l

r0
2

3n 2  l (l  1) .  (3.132)

The second important trend is that at fixed n, the orbital quantum number l determines how fast does the
wavefunction arise with r, and how much it oscillates in the radial direction. For example, the 2s
eigenfunction is finite at r = 0, and makes one “wiggle” (and has one root) in the radial direction, while
eigenfunctions 2p equal zero at r = 0, and do not oscillate at all in the radial direction. Instead, those
wavefunctions always oscillate as functions of some angle – see the second row of Fig. 18. The same
trend in clearly visible for n = 3 – see Fig. 19, and continues for the higher values of n.
Besides Eq. (214), mathematics yields one more surprisingly simple relation for the radial
functions Rn,l (and, since the spherical harmonics are normalized to 1, for the eigenfunctions as the
whole):
1 1
 2
. (3.133)
r n ,l n r0

This means, in particular, that for any eigenfunction n,l,m, with all its complicated radial and angular
dependencies, there is a simple relation between the potential and full energies:
1 C E
U n ,l
 C  2
  20  2 E n , (3.134)
r n ,l n r0 n
so that the average kinetic energy of the particle is also simply related to En:
K n ,l
 En  U n .l
 En . (3.135)
These simple results are in a sharp contrast with the complex expressions for the eigenfunctions,
and motivate a search for more refined methods which would replace, or at least complement our brute-
force (wave-mechanics) approach, to reveal their real nature. Such an approach is the main topic
discussed in the next chapter.

© K. Likharev, 2008 41

Vous aimerez peut-être aussi