Quantum Mechanics

Chapter 1. Introduction

This introductory chapter briefly review the major motivations for quantum mechanics. Then its
simplest formalism – Schrödinger’s “wave mechanics” is described and its main features are
discussed.1

1.1. Experimental motivations

By the beginning of the 1900s, physics (which by that time included what we now know as
nonrelativistic classical mechanics, classical statistics and thermodynamics, and classical
electrodynamics including optics) looked as an almost completed discipline, with a lot of experimental
observations explained, and just a couple of mysterious “dark clouds”2 on the field’s horizon. However,
the rapid technological progress and the resulting fast development of experimental techniques at that
time have led to a fast multiplication of observed phenomena which could not be explained on the
classical basis. Let us list the most consequential of those experimental findings.
(i) Blackbody radiation measurements, started by G. Kirchhoff in 1859, has shown that the in
thermal equilibrium, the power of electromagnetic radiation of a fully absorbing (“black”) surface per
unit frequency interval drops exponentially at high frequencies. This is not what could be expected from
the combination of the classical electrodynamics and statistics, which predicted an infinite growth of the
radiation density with frequency. Indeed, classical statistics (see, e.g., SM3 Sec. 2.7) predicts that in the
thermal equilibrium at temperature T, the average energy of each harmonic oscillator should equal kBT ,
where kB is the Boltzmann constant.4 In turn, the classical electrodynamics indicates that
electromagnetic field modes in free space should behave as harmonic oscillators, and that the number of
the density of these modes in a large volume V >> 3 per small frequency interval is
dVk 4k 2 dk 2 8 2
dN  2  V  2 V V d   V d , (1.1)
2 3 2 3  2c3 c3
where c  2.9979×108 m/s is the free-space speed of light, k = /c is the free-space wavenumber,  is
the “cyclic” frequency (measured in Hz, i.e. cycles per second), and  = 2/k = c/ is the radiation
wavelength. (The first factor 2 in Eq. (1)5 is due to two possible polarizations of transversal EM waves.)

1 Much of this material may be found in undergraduate textbooks (which could be also used as remedial reading
throughout the course). I could especially recommend:
- R. L. Liboff, Introductory Quantum Mechanics, 3rd ed., Addison-Wesley, 1998,
- D. J. Griffith, Quantum Mechanics, 2nd ed., Pearson Prentice Hall, 2005,
- S. Gasiorowicz, Quantum Physics, 3rd ed., Wiley 2003.
2 The expression used in the 1900 talk by Lord Kelvin (W. Thomson) in reference to the blackbody radiation and
Michelson-Morley experiments.
3 Acronyms CM, SM, and EM refer to the corresponding other parts of my lecture note series.
4 In the SI units used in these notes, kB  1.3807×10-23 J/K. Also, notice that in many theoretical papers (and in the
SM part of my notes), kB in taken for 1, i.e. temperature is measured in energy units.
5 In all my notes, chapter numbers are dropped in references to formulas, figures, and tables within the same
chapter.

© K. Likharev, 2008 1
 Quantum Mechanics

Combining these two results, we get the so-called Rayleigh-Jeans formula for the average energy
per unit volume:
1 dE k B T dN 8 2
u   3 k BT , (1.2)
V d V d c
which diverges at   . On the other hand, the blackbody radiation measurements, perfected by O.
Lummer and E. Pringsheim, and H. Rubens and F. Kurlbaum with a 1%-scale accuracy, were
compatible with the phenomenological law suggested in 1900 by Max Planck:
8 2 h
u , (1.3)
c exp(h / k BT )  1
3

where h is now called the Planck constant. Since it is more natural to use the “angular” frequency  =
2 (measured in radians per second) rather than , it is now common to write the product h as , and
use the name “Plank constant”6 for the he ratio
h
  1.0545716  10 -34 J  s . (1.4)
2
(The number given in Eq. (4) corresponds to the most precise modern measurements,7 with the relative
uncertainty better than 10-7).
At low frequencies (h   << kBT), the denominator in Eq. (3) may be approximated by
(h/kBT), so that the Planck law (3) reduces to the Rayleigh-Jeans formula (2), while at higher
frequencies it describes the experimentally observed rapid decrease of the radiation density – see Fig. 1.
10

1
u
u0
0.1 Fig. 1.1. The Planck law (red line) and the
Rayleigh-Jeans formula (blue line) for the
blackbody radiation density u expressed in
reduced units u0  2kBT/2c3.
0.01
0.1 1 10

/kBT
(ii) The photoelectric effect, experimentally discovered in 1887 by H. Hertz, shows a sharp
lower bound on the frequency of light which may kick electrons out from metallic surfaces, regardless

6 In some early texts,  was called the Dirac constant.

7 The NIST Reference on Constants, Units, and Uncertainty, see online
http://physics.nist.gov/cuu/Constants/index.html.

© K. Likharev, 2008 2
 Quantum Mechanics

of the light intensity. Albert Einstein, in the first of his three famous 1905 papers,8 noticed that this
threshold min could be readily explained assuming that light consisted of certain particles (now called
photons) with energy
E    h , (1.5)
with the same Planck constant which participates in Eq. (3). Indeed, with this assumption, at the photon
absorption by the surface, its energy  is divided between a fixed energy W (now called workfunction)
of electron binding inside the metal, and the residual kinetic energy mv2/2 > 0 of the freed electron – see
Fig. 2. In this picture, the frequency threshold finds a natural explanation as min= W/.9

mv 2
 E W
2
E  
Fig. 1.2. Einstein’s explanation of the photoelectric
-e effect’s frequency threshold.

(iii) The discrete spectra of radiation by excited atomic gases could not be explained by classical
physics. (Applied to the planetary model of atoms, proposed by E. Rutherford, it predicts the collapse of
electrons on nuclei in ~10-10 s due to electric dipole radiation of electromagnetic waves – see, e.g., EM
Sec. 9.2.) Especially challenging was the observation by J. Balmer (in 1885) that the radiation
frequencies of simple atoms may be described by simple formulas. For example, for the simplest atom,
hydrogen, all radiation frequencies may be numbered with just two positive integers n and n’:
 1 1 
 n,n '   0   2 , (1.6)
n n' 
2

with 0  1,  2.071016 s-1. The Balmer series, and in particular the value of 0, have found its
explanation in the famous theory by Niels Bohr (1913) which was a semi-phenomenological precursor
for quantum mechanics. In this theory, n,n’ is considered as the frequency of a photon which obeys the
Einstein’s formula (5), with the energy being the difference of two “quantized” (discrete) energy levels
of the atom (Fig. 3):
E n ,n '  E n '  E n  0 . (1.7)
En'
 n , n '  E n '  E n Fig. 1.3. Electromagnetic wave radiation at a
system’s transition between its two quantized
En energy levels.

8 As a reminder, Einstein received his Nobel prize (in 1922) for this work, rather than for his relativity theory!
9 For most metals, W is between 4 and 5 electron-volts (eV), so that the threshold corresponds to max = 2c/min
= ch/W  300 nm – approximately the border between the visible and ultraviolet light.

© K. Likharev, 2008 3
 Quantum Mechanics

Bohr has shown that the correct10 expression for the levels (relative to the free electron energy),
Ry
En    0, (1.8)
2n 2
and the correct value of the “Rydberg constant”11
2
m  e2 
Ry  2 0  2e    27.21138 eV , (1.9)
  4 0 
(where e  1.60217710-19 C is the electron’s charge, and me  0.91093910-30 kg is its rest mass) could
be obtained, with a one-line calculation, from the classical mechanics and the additional postulate that
the angular momentum L = mevr of the electron moving on a circular trajectory of radius r about the
nuclei (proton) at rest, is quantized as
L  me vr  n , (1.10)

where  is again the same Plank constant (4). (For that, it is sufficient to solve Eq. (10) together with the
2nd Newton’s law for the rotating electron,
v2 e2
me  , (1.11)
r 4 0 r 2
for the electron velocity v and radius r, and them plug them into the classical expression for the full
electron energy
me v 2 e2
E  . (1.12)
2 4 0 r
This nonrelativistic approach to the problem is justified a posteriori by the fact the relevant energy scale
Ry << mc2 ~ 0.5 MeV.) By the way, the value of r, corresponding to n = 1, i.e. to the lowest possible
electron orbit,
4 0  2
rB  , (1.13)
me e 2

r = rB = (m/2)(e2/40)  0.0529 nm, and called the Bohr’s radius, defines the spatial scale of all
phenomena in atomic, molecular and condensed matter physics, as well as physical and quantum
chemistry.
Now notice that Bohr’s quantization postulate (10) may be presented as having an integer
number (n) of certain “matter waves”12 along the circular orbit’s perimeter 2r = n. Dividing both parts

10 Besides small corrections due to the finite ratio of the electron mass me to that of the nuclei, as well as to spin
and relativistic effects - see Chapter 9.
11 Unfortunately, sometimes the same term (“Rydberg constant”) is frequently used for the reciprocal wavelength
(1/0 = 0/2c) corresponding to frequency 0 = Ry/2. To add to confusion, some texts call Ry the Hartree (Eh)
or the atomic energy unit, and its half, the Rydberg (!).
12 This fact was noticed and developed in 1923 by Louis de Broglie, so that instead of speaking of
wavefunctions, especially of free particles, we are still frequently speaking of “de Broglie waves”.

© K. Likharev, 2008 4
 Quantum Mechanics

of this equation by , we see that for this statement to be true, the wavenumber k  2/ of the (then
hypothetic) de Broglie waves should be proportional to the particle’s momentum p = mv:
p  k . (1.14)
(iv) The Compton effect, studied in detail by A. Compton in 1924-1925, is the change of
frequency of X-rays at their scattering on free (or nearly-free) electrons – see Fig. 4.

 ' / c

 / c 
m 
Fig. 1.4. The Compton effect.

p

The effect may be explained assuming that the X-ray photon also has a momentum which obeys the
vector-generalized Eq. (14):
   
p ph  k  n, (1.15)
c

where k is the wavevector (whose magnitude is equal to wavenumber k, and direction coincides with
 
that, n , of the wave propagation ), and that the free electron’s momentum p is related to its energy E
via the relativistic formula
E 2  (cp) 2  (mc 2 ) 2 . (1.16)
(The rest energy of a photon is zero, so that its energy is always described by Eq. (5).) Indeed, a
straightforward solution of the following system of three equations,


  mc 2   ' (cp) 2  (mc 2 ) 2 1/ 2
, (1.17)
  '
 cos   p cos  , (1.18)
c c
 '
0 sin   p sin  , (1.19)
c
(which describe, respectively, the conservation of the full energy of the photon-electron system, and two
relevant components of its full momentum, at the scattering event), immediately yields the result
1 1 1 h
  (1  cos  ) , or '    (1  cos  ) , (1.20)
 '  mc 2 mc
which is in a full agreement with experiment.13

13 The fundamental constant combination h/mc, which participates in this equation, is sometimes (and very
unfortunately) called the Compton wavelength, and is close to 2.4610-12 m.

© K. Likharev, 2008 5
 Quantum Mechanics

(v) In 1927, following the suggestion by W. Elassger (who was excited by Louis de Broglie’s
conjecture of “matter waves”), C. Davisson and L. Germer, and independently G. Thomson succeeded
to observe diffraction of electrons on crystals (Fig. 5). Specifically, they have found that the intensity of
elastic reflection from a crystal increases when angle  between the incident beam of electrons and an
atomic plane satisfies the relation
2d sin   n , (1.21)
where  = 2/k = 2/p = h/p is the de Broglie wavelength of electrons, and n is an integer. As evident
from Fig. 5, this is just the well-known condition14 that the optical path difference l = 2dsin between
the de Broglie waves reflected from two adjacent crystal planes is an integer number of , i.e. of the
constructive interference of the waves.15


d Fig. 1.5. Electron scattering from a crystal
d sin  d sin  lattice.

To summarize, the last 4 effects may be explained starting from two very simple (and similarly
looking) assumptions (5) and (14), involving the same Planck constant. Moreover, the Planck radiation
law (3) may be also readily explained using Eq. (5) – see, e.g., SM Sec. 2.8. This might give an
impression of a sufficient experimental evidence to declare electrons “matter waves” rather than
particles. However, by that time (the mid 1920s) there were also very strong evidence for particle
(“corpuscular”) behavior of the same electrons and photons. It is sufficient to mention the famous oil-
drop experiments by R. Millikan and H. Fletcher (1909-13) in which singe (and whole!) electrons could
be added to a macroscopic body, changing its total electric charge by (-e). It was apparently impossible
to reconcile these observations with a purely wave picture, in which an electron and hence its charge
needed to be spread over the wave’s volume, so that its arbitrary part could be cut out using appropriate
experimental setups.
Thus the founding fathers of quantum mechanics faced a formidable task of reconciling the wave
and particle properties of electrons (and other particles). The decisive breakthrough in that task has been
achieved in 1926 by Ervin Schrödinger who (with a substantial insight from Max Born) formulated what
is now known as the Schrödinger picture of quantum mechanics in the coordinate representation, or

14 Frequently called the Bragg condition, due to the pioneering experiments by W. Bragg with X-ray scattering
from crystals (1912 on).
15 Later, spectacular experiments with diffraction and interference of heavier elementary particles, e.g. neutrons,
have also been performed – see, e.g. A. Zeilinger et al., Rev. Mod. Phys. 60, 1067 (1988). Moreover, quantum
interference between different states of truly macroscopic objects, e.g., of superconducting condensate of millions
Cooper pairs in 10-3-cm-scale metallic rings, have been observed – see, e.g., the pioneering experiments by J.
Freedman et al., Nature 406, 43 (2006), carried out at Stony Brook.

© K. Likharev, 2008 6
 Quantum Mechanics

wave mechanics. We will now formulate that picture, somewhat disregarding the actual history of its
development. 16

1.2. Wave mechanics postulates

For a single spinless,17 nonrelativistic particle, wave mechanics may be based on the following
set of axioms (which are of course comfortingly elegant, but find their final justification in the
agreement of their corollaries with experiment).
 
(i) Probability and wavefunction. Such variables as coordinate r or momentum p of a
“quantum particle”18 cannot be always measured exactly, even in the “perfect conditions” limit when all
external uncertainties (including measurement instrument imperfection, the initial state preparation
fluctuations, and unintended particle interaction with its environment) have been removed.19 Moreover,
 
r and p can never be measured exactly simultaneously. Instead, the most detailed description of the
particle’s state allowed by Nature (in these ideal conditions) is given by a certain complex function

 (r , t ) , called either wavefunction or -function, which enables only probabalistic predictions of
 
measured values of r , p , and other measurable variables (“observables”).20
(ii) Connection between  and experiment. Probability dP of finding a particle inside an
infinitesimal volume d3r may be characterized by the probability density w  dP/d3r which in turn is
related to the wavefunction as
 2  
w   (r , t )   * (r , t ) (r , t ) , (1.22)

where (…)* means the complex conjugate. More generally, for each observable A, there is a
corresponding linear operator  , such that in a perfect experiment described above, the average value
(also called the expectation value) of A is

A    * Aˆ d 3 r , (1.23)

where the integration is extended over the whole region where the particle may be found (  0), and
… means the statistical average, i.e. averaging over results of a large number of (macroscopically)
similar experiments.

16 Generally, quantum mechanics may be built on different sets of axioms. In this text, I will not try to beat down
the number of the initial postulates to the absolute minimum, because such attempts typically result in making
certain implicit assumptions hidden from the reader – a practice as common as it is regrettable.
17 Actually, the particle’s spin has not to be equal zero. Rather, we assume that the spin effects are negligible, as
they are, for example, for a nonrelativistic electron moving in a region without an appreciable magnetic field.
18 This term does not mean any special particle, just the particle whose quantum properties are substantial in the
particular circumstances.
19 We will imply these perfect conditions until the discussion of particle’s interaction with environment, and
“real” (physical) measurements in Chapter 7.
20 Sometimes this fact is referred to as the absence of “hidden variables” whose fixation could ensure the exact
values of all observables.

© K. Likharev, 2008 7
 Quantum Mechanics

(iii) Hamiltonian operator and the Schrödinger equation. One particular operator, the
Hamiltonian Ĥ , whose observable is the classical Hamiltonian function H (see, e.g., CM Sec. 2.3),21
plays in wave mechanics a special role, because it participates in the Schrödinger equation

i  Hˆ  (1.24)
t
which determines the wavefunction’s dynamics (time evolution).
(iv) Kinetic energy and momentum. For a free particle, the Hamiltonian (which in this case
corresponds to the kinetic energy of the particle) has the form
pˆ 2 2 2 2  2 2 2 
Hˆ      2  2  2  , (1.25)
2m 2m 2m  x y z 

where p̂ is the (vector) operator of particle’s momentum:

    
pˆ  i  i  , ,  . (1.26)
 x y z 
(v) Correspondence principle. In the limit when quantum effects are insignificant, e.g., when the
characteristic scale of action S (see CM Sec. 9.2) is much larger than , all wave mechanics results have
to tend to those given by classical mechanics. Mathematically, this is achieved by duplicating the
classical relations between observables by similar relations between the corresponding operators.
(Notice that the first of Eqs. (25) certainly corresponds to that requirement, giving the same functional
relation between the operators of the kinetic energy and momentum as the classical relation E = p2/2m
between the corresponding observables.)
Let us check, first of all, that these postulates indeed lead to the immediate resolution of the
apparent contradiction between the wave and corpuscular properties of particles. Indeed, for a free
particle, the Schrödinger equation (24), with the substitution of Eq. (25), takes the form
 2 2
i    , (1.27)
t 2m
and may be satisfied with a particular (but most important) plane-wave, “monochromatic” (single-
frequency) solution
 
  
 (r , t )  C exp i (k  r  t ) , (1.28)

where C, k and  are some constants. Indeed, plugging Eq. (28) into Eq. (27), we see that it is indeed a

solution, with an arbitrary constant C, but a certain “dispersion” relation between wavevector k and
frequency :
(k ) 2
  . (1.29)
2m

21 As a reminder, for many systems (including those whose kinetic energy is a called quadratic-homogeneous
function of generalized velocities, like mv2/2), H coincides with the total energy E.

© K. Likharev, 2008 8
 Quantum Mechanics

Constant C may be calculated assuming that solution (28) is extended over a certain volume V, while
beyond it  = 0. It the particle is certainly inside that volume, according to Eq. (22) the integral of *
over this volume should equal 1 - the so-called “normalization condition”. From here, using Eq. (28),
we get
C V  1.
2
(1.30)

Now we can calculate the expectation value of particle’s momentum p and energy E (which, for
a free particle, coincides with its Hamiltonian function H), using Eqs. (23), (25) and (26). The result is
  (k ) 2
p  k , E  H  ; (1.31)
2m
according to Eq. (29), the last result may be rewritten as E = . Next, equation (23) enables one to
calculate not only the average (in math speak, the “first moment”) of an observable, but its higher
moments, notably the second moment (in physics, usually called the variance or dispersion):

A 2  A  A   A 2  2 A A  A  A 2  A ,
~ 2 2 2
(1.32)

and hence its “root mean square (r.m.s.) fluctuation”

~
A  A2 (1.33)
~
which characterizes the intensity of deviations A  A  A of measurement results from the average, i.e.
the uncertainty of observable A. Applying these relations to the wavefunction (28), we get E = 0,

p  0 , meaning that in that plane-wave state the energy and momentum of the particle are exactly
defined, so that the signs of statistical average in Eq. (31) may be removed.
Hence Eqs. (5) and (14) are indeed valid, and may explain the observed wave properties of
electrons. (For photons, we would need a relativistic formalism – see Ch. 9.) On the other hand, due to
the linearity of the Schrödinger equation (24), any sum of its solutions is also a solution. (The so-called
linear superposition

principle.) For a free particle, this means that a set of plane waves (28) with close
values of k (and hence  – see Eq. (29)) is also a solution of this equation. It is well known that such
sets may be used to describe spatially localized rf pulses, called wave packets –– see Fig. 6. (We will
study wave packet properties in detail in the beginning of the next chapter.) For such a packet, Eq. (22)
describes the particle location probability spread over a finite, possibly very small volume, so that the
wave packets may be used to describe the observed corpuscular properties of particles.

© K. Likharev, 2008 9
 Quantum Mechanics

(a) p (b)
x c(p)
Re 
Im 

0 0
x p0 p  k
the particle is Fig. 1.6. (a) Snapshot of a wave packet propagating
(somewhere:-) along axis x, and (b) the corresponding distribution of
here! wave numbers k, i.e. momenta p.

The elementary wave theory, based on the Fourier expansion, teaches us that the spatial
extension x of a wave packet in the direction of its propagation (say, x) is related to the width kx of its
wavenumber distribution as xkx  ½. Applying Eq. (14), we arrive at the famous Heisenberg’s
uncertainty principle:

x  p x  (1.34)
2
which tells us that the particle’s coordinate and momentum cannot be defined exactly simultaneously. In
the next section, we will prove that Eq. (34) is actually valid for the particle motion in an arbitrary field.
Moreover, in Chapter 4 we will see that similar uncertainty relations are valid for many other pairs of
observables.
In a nutshell, this is the main idea of the wave mechanics, and the first part of this course
(Chapters 1-3) will be essentially various elaborations on this idea.

1.3. Discussion
The postulates listed in the previous section seem very simple, and they are hopefully familiar to
the students from their undergraduate studies. However, actually the physics of these axioms are very
deep, and they lead to several counter-intuitive conclusions. This is why the balance of this section will
give only an initial, admittedly something superficial discussion of the postulates, and we will be
repeatedly returning to the conceptual foundations of quantum mechanics throughout the course,
especially in Chapter 7.
First of all, the notion of fundamental uncertainty of observables, which is in the core of
postulate (i), is very foreign to the basic ideas of classical mechanics, and historically this fact has made
quantum mechanics so hard to swallow for many star physicists, notably including A. Einstein.
However, it has been confirmed by numerous experiments, and (more importantly) there have not been a
single confirmed experiment which would contradict to this fact, so that quantum mechanics moved
from a theoretical hypothesis to the rank of a reliable scientific theory long ago.
One more necessary remark concerning postulate (i) is that in many classical statistical physics

situations, the probability density distribution w(r , t ) may be also calculated from a deterministic
differential equation, e.g., the Boltzmann equation or the Fokker-Planck equations. Quantum-
mechanical description differs from those situations in two important aspects. First, in the perfect

© K. Likharev, 2008 10
 Quantum Mechanics

conditions outlined above (exact initial state preparation, no interaction with environment, the best
possible measurement), the Fokker-Planck equation reduces to the 2nd Newton law, while in wave
 
mechanics this is not true. Second, the wave function  (r , t ) gives more information than just w(r , t ) ,
because besides the modulus  involved in Eq. (22), this complex function also has “phase”  
arg, and as evident from Eq. (23), this phase does affect some observables.
Next, it is very important to understand that the relation between the quantum mechanics to
experiment, given by postulate (ii), necessarily involves another key notion: that of the corresponding
statistical ensemble. Indeed, the probability of a certain (n-th) outcome of an experiment may be only
defined as the limit
N
Nn
Pn  lim M  , M   Nn , (1.35)
M n 1

where M is the number of experiments carried out at constant conditions, Nk is the number of outcomes
of the k-th type, and N is the number of different outcomes.22 In mathematical language, this infinite (or
at least very large) set of similar experiments is an example of a statistical ensemble. It is clear that a
particular choice of an ensemble may affect probabilities Pk very significantly. For example, if we pull
cards at random from a pack of 52 different cards, the probability Pk of getting a certain card (e.g., the
queen of spades) is 1/52. However, if cards of a certain suit (say, hearts) had been taken out from the
pack in advance, the probability of getting the queen of spades will be higher, 1/39. It is important that
we would also get the same probability is we used the full 52-card pack, but by some reason ignored
results of all experiments giving us any rank of hearts.
Similarly, in quantum mechanics, the probability distributions (and hence expectation values of
particle coordinate and other observables) depend not only on the conditions of experiments, but also on
the set of outcomes we count. Because of the unique relation (22) between w and , this means the
wavefunction also depends on those factors, i.e. on the chosen statistical ensemble, i.e. both on the
experiment preparation and the subset of outcomes taken into account. While the first restriction still
may be compatible with the attribution of a wavefunction (calculated for the whole ensemble, with no
discarded measurement results) to every single experiment of the set, it is important to understand that
the function (if it exists, see below) also depends on the way the experiment results are counted. The
insistence on the continuous attribution of the wavefunction to a single experiment, both before and after
the measurement, may leads to very unphysical interpretations of some experiments, including
wavefunction evolution not described by the Schrödinger equation (the so-called “wave packet
reduction”), subluminal action on distance, etc. In Chapter 7 we will see that minding the statistical
nature of the quantum mechanics, and in particular the dependence of the wavefunction on the statistical
ensemble definition, may readily explain such apparent paradoxes of quantum measurements.23

22 Here we imply that the outcomes are discrete (e.g., may be reliably distinguished), i.e. their total number M is
finite, but all the reasoning may be readily generalized to the case of continuous observable (M  ). Notice also
that due to the definition (34), the sum of all possible Pn equals 1. This is the normalization condition for the
discrete outcome case.
23 “Quantum measurements” is also a very unfortunate term; it would be more reasonable to speak about
experimental measurements of observables in quantum mechanics. However, the former term is so common that I
will use it through the course, in particular in Chapter 7.

© K. Likharev, 2008 11
 Quantum Mechanics

Let me also emphasize that statistics is intimately related to the information theory (and not only
via their common mathematical background, the probability theory). For example, the question, “What
subset of experimental results we count?” may be replaced by the question, “What subset of results do
we have information about?”. As a result, the reader has to be ready to the use of information theory
notions for the discussion quantum mechanics, or at least is relation to experiment (i.e. to “reality”).
This feature of quantum mechanics also makes some physicists uncomfortable, because much of
classical mechanics and electrodynamics may be discussed without any reference to information. In
quantum mechanics, such abstraction is impossible.
The ideal condition specification in postulate (i) also invites a natural question: what if (at least
some of) these conditions do not hold? As will be discussed in Chapter 7, in this case the statistical
ensemble of quantum particles cannot be described with a certain wavefunction, but requires a more
general description, e.g., using the density matrix. These cases are actually a bridge between the
quantum mechanics and the “usual” (classical) statistical physics, and their study is usually called the
quantum statistics.
Proceeding to Eq. (23), it also relies on the probability theory for the definition of the statistical
average (the expectation value). Namely, let each of N possible outcomes in a set of M macroscopically
similar experiments give a certain value An of observable A; then, by definition,
N N
1
A  lim M 
M
 An N n   An Pn ;
n 1 n 1
(1.36)

Equation (23) specifies how this expectation value should be calculated in wave mechanics. Notice that
in order to have that equation compatible with Eq. (22), we should associate the identity (unit) operator
Iˆ , defined as
Iˆ   (1.37)
for any function , with the detection of particle’s presence.
One more remark concerning postulate (ii) is that in quantum mechanics uses only linear
operators. This means that all operators  obey the following rules
Aˆ  Aˆ   Aˆ   Aˆ ,
1 2 1 2 (1.38)

Aˆ c1 1  c 2 2   Aˆ c1 1   Aˆ c 2 2   c1 Aˆ 1  c 2 Aˆ 2 , (1.39)

where n are arbitrary wavefunctions, and cn, arbitrary constants (in quantum mechanics, frequently
called “c-numbers”, to distinguish them from operators and wavefunctions).24 The examples of linear
operators are given by:
(i) a multiplication of the wavefunction by a function (e.g., expression x2 may be understood
as the action of operator x2 on wavefunction ), and

24 By the way, if any equality involving operators is valid for an arbitrary wavefunction, the latter is frequently
dropped from notation, resulting is an operator equation. For example, Eqs. (38) and (39) may readily used to
  
prove the following operator properties: Aˆ 2  Aˆ1  Aˆ1  Aˆ 2 (“commutativity”) and Aˆ1  Aˆ 2  Aˆ 3  Aˆ1  Aˆ 2  Aˆ 3 
(“associativity”).

© K. Likharev, 2008 12
 Quantum Mechanics

(ii) a scalar or vector differentiation of the wavefunction, like in Eqs. (25) and (26).25
Now proceeding to postulate (iii), it is very important that the Schrödinger equation (24) is also
linear. (Actually, we have already used this fact when we discussed wave packets in the previous
section.) This means that if each of functions n are (particular) solutions to Eq. (24) with a certain
Hamiltonian, there arbitrary linear combination
   cn n (1.40)
n

is also a solution to the same equation.26 Let us use this feature to accomplish an apparently impossible
feat: immediately find the general solution to the Schrödinger equation for the most important case
when operator Ĥ does not depend on time. (An example of such operator is provided by Eq. (1.25).)

1.4. Eigenstates and eigenvalues

First of all, let us check whether the following product

n  an (t ) n (r ) (1.41)
qualifies as a (particular) solution to the equation. Plugging Eq. (41) into Eq. (24), using the fact that for
a time-independent Hamiltonian
 
Hˆ an (t ) n (r )  an (t ) Hˆ  n (r ) , (1.42)

and dividing both parts of the equation by n = ann, we get

i dan Hˆ  n
 . (1.43)
an dt n
The left hand side of this equation may depend only on time, while the right hand one, only on
coordinates. These facts may be only reconciled if we assume that each of these parts is equal to (the
same) constant, which we will denote as En. As a result, we are getting two separate equations for the
temporal and spatial parts of the wavefunction:
dan
i  E n an , (1.44a)
dt
Hˆ  n  En n . (1.44b)
(This trick, called the variable separation, is very common in many fields of mathematical physics.)
The first of these equations is readily integrable, giving

25 Let me hope that the reader understands that a double differentiation like 2/x2 is still a linear operation as
defined by Eqs. (38) and (39).
26 It is amazing that the linear Schrödinger equation correctly describes quantum properties of systems
whose classical dynamics is described by nonlinear equations of motion (e.g., an anharmonic oscillator).
We will discuss this apparent contradiction later in the course.

© K. Likharev, 2008 13
 Quantum Mechanics

 E 
an  const  exp  i n t  , (1.45)
  
returning us again to the fundamental relation (5) between energy and frequency. Plugging Eqs. (41) and
(46) into Eq. (22), we see that if the system is in the state (41), the probability of finding the particle at
certain location does not depend on time, while doing the same with Eq. (23) shows that the same is true
for the expectation value of any operator which does not depend on time explicitly:

A n
  n* A n d 3 r = const. (1.46)

Due to this property, states (41) are called stationary. In contrast to the time dependence, the spatial

distributions  n (r ) of the stationary states are hard to find, besides some simple cases, and the solution
of the stationary (or time-independent) Schrödinger equation (45)27 which describes the distributions,
will be the main task of the initial part of this course.
To make that equation specific, one needs, first of all, to find the Hamiltonian operator
describing the particular problem. For that, substantial relief comes from postulate (v), the
correspondence principle (v). It implies, in particular, that for a particle moving in a constant field of
conservative forces, the Hamiltonian may be presented as a sum of part (25), describing the kinetic

energy of the particle, and operator U (r ) of its potential energy, acting on the wavefunction as a
multiplier:
pˆ 2 
Hˆ   U (r ). (1.47)
2m
(Historically, this was the main step done by Ervin Schrödinger on the background of L. de Broglie’s
work.) In mathematics, the corresponding stationary Schrödinger equation,
2 2 
   n  U (r ) n  En n , (1.48)
2m
falls into the category of linear “eigenproblems”,28 in which self-consistent eigenfunctions n and
eigenvalues En should be simultaneously found. Mathematics also tells us that for the problems in which
all n are space-restricted (“quantum-confined”), i.e. tend to zero at r  , the spectrum of eigenvalues
is discrete. It is also easy to prove that the eigenfunctions corresponding to different eigenvalues are
orthogonal, i.e. that integrals of the products n*n’ vanish for all pair with n  n. Moreover, due to the
linearity of the Schrödinger equation linearity, each of these functions may be multiplied by a constant
coefficient to make this set orthonormal:

* d 3 r    1, if n  n' ,
 n n' n ,n ' 
0, if n  n'.
(1.49)

27 In contrast, the initial Eq. (24) is frequently called the time-dependent or nonstationary Schrödinger equation.
28 From German root “eigen” meaning “particular” or “characteristic”.

© K. Likharev, 2008 14
 Quantum Mechanics


Also, the eigenfunctions form a full set, meaning that an arbitrary function  (r ) , for example

the actual wavefunction of the system  (r , t ) in the initial moment of its evolution (which we will take
for t = 0), may be presented as an expansion over the eigenfunction set:29
 
 (r ,0)   cn n (r ) . (1.50)
n

Let us now consider the wavefunction

   E  
 (r , t )   c n a k (t ) k (r )   c n exp  i n t  n (r ) , (1.51)
n n   
where we took the constant in Eq. (46) for unity, as we have every right to do. Since each term of the
sum has the form (41) and satisfies the Schrödinger equation, so is the sum as the whole. Moreover, Eq
(1.50) that it evidently satisfies the initial conditions as well. At this moment we can again use a help by
mathematicians who tell us that is the partial differential equation of type (24) with the Hamiltonian of
type (47),
 2 2 
i     U (r ) , (1.52)
t 2m
with fixed initial conditions, can only have one (“unique”) solution. This means that Eq. (51) gives the
general solution to the general (“time-dependent”) Schrödinger equation (52). We will use this key fact
through the course, though in many cases, following the physical sense of particular problems, will be
rather interested in particular solutions.
In order to get some feeling of these general results, let us consider an elementary example
(which nevertheless will be the basis for discussion of many less trivial problems): a particle in a
“quantum well” with sharp (“hard”) and infinitely high walls:
 0, for 0  x  a x , 0  y  a y , and 0  z  a z ,
U (r )   (1.53)
  , otherwise.
In order to keep the product Un is Eq. (48) finite outside of the well, in these regions the wavefunction
should equal zero. Also, the function have to be continuous everywhere, to avoid the divergence of its
Laplace operator. Hence, we may solve the stationary the Schrödinger equation (48) inside the well,
2 2
   n  E n n , (1.54)
2m
with zero boundary conditions on all the walls.
At this point, the same method which was used earlier in this section to separate time and space
variables, is very useful to separate Cartesian spatial variables from each other. Let us look for a
particular solution of Eq. (54) in the form

29 The expansion coefficients ck may be readily found by multiplying both parts of Eq. (50) by *n’, integrating
the result over space, and using Eq. (49):
 
cn   ( r ,0) n* (r ) d 3 r .

© K. Likharev, 2008 15
 Quantum Mechanics


 n (r )  X ( x)Y ( y ) Z ( z ) . (1.55)
(We will prescribe proper indices to the participating factors later.) Plugging this expression into the Eq.
(54) and dividing by  = XYZ  0, we get
 2  X " Y" Z" 
    E. (1.56)
2m  X Y Z 
Repeating the above argumentation, since each term in the parentheses may be only a function of the
corresponding argument, the equality is possible only if each term is a constant, with the dimensionality
of energy. Calling them Ex, etc., we get three 1D equations
2 X" 2 Y"  2 Z"
  Ex ,   Ey ,   Ez , (1.57)
2m X 2m Y 2m Z
with the matching condition
Ex  E y  Ez  E . (1.58)

All three resulting ordinary differential equations, and their solutions, are similar. For example, for X(x),
we get
2 d 2 X
  Ex X , (1.59)
2m dx 2
with the simple boundary conditions, X(0) = X(ax) = 0. This is a well-known boundary problem
(describing, for example, waves on a guitar string), with sinusoidal standing-wave solutions30
2 2 x
n  sin k nx x  sin n x , n x  1, 2,... , (1.60)
a a a
corresponding to eigenenergies
 2 2  2 2 2
En  kn  2
n  E1 n 2 . (1.61)
2m 2ma
Figure 7 shows this result using a somewhat odd but convenient way when the eigenenergy
values (“energy levels”) are used as horizontal axes for plotting eigenfunctions.

E nx / E1
nx  3
X nx ( x)
9

5
nx  2
4
nx  1 Fig. 1.7. Eigenfunctions (solid lines) and eigenvalues
10 (dashed lines) of the 1D wave equation. Solid black lines
show the potential energy profile.
0 ax x

30 The coefficient before the sine is selected to satisfy the orthonormality condition (49).

© K. Likharev, 2008 16
 Quantum Mechanics

Y (y) and Z(z) are similar functions of their arguments, and may also be numbered by integers (say, ny
and nz) independent of nx, so that the total eigenenergy spectrum is

 2  2  n x2 n y2 n z2 
.
E nx ,ny ,nz    (1.62)
2m  a x2 a y2 a z2 

and our former index n should be understood as a set of three “quantum numbers” (integers) nx, ny, and
nz. Since functions (60) are orthogonal, the general solution of Eq. (54) may be presented as the sum of
all possible products (55):

 
n x x n y y n z z
 (r )   C nx ,ny ,nz sin
ax
sin
ay
sin
az
. (1.63)
n ,n ,n  1
x y z

This simplest problem is a good illustration of the basic features of wave mechanics for a spatially-
confined motion, including the discrete energy spectrum, and (in this case, evidently) orthogonal
eigenfunctions.
An example of the opposite limit of a continuous spectrum for unconfined motion is given by
plane waves (28) which may be viewed as the product of the time-dependent factor (46) by
eigenfunction
 
pr
 p  C p exp{i } (1.64)

which is the solution to the stationary Schrödinger equation (54). Since there are no boundary conditions
to satisfy, the hence the particle momentum p can take any real values. (This is why I prefer to index the
wavefunctions and eigenenergies

p2 p2
Ep   0 (1.65)
2m 2 m
with their momentum rather than n.) Systems with continuous spectrum require some math caution:
summation (51) oven n should be replaced by integration over p, and the normalization condition should
be refined to avoid integral divergence. (We will discuss these details later in the course.) However,
conceptually the basic philosophy of wave mechanics remains the same.

1.5. Continuity equation and probability current

This philosophy survives one more sanity check: it satisfies the natural requirement that the
particle does not appear or vanish in the course of the quantum evolution.31 Let us use Eq. (22) to
calculate the time rate of change of the probability w to find the particle within a certain volume V:
dP d d
  wd 3 r    *d 3 r . (1.66)
dt dt V dt V

31 This property is not absolute: it disappears in the relativistic quantum mechanics of interacting particles – see
Chapters 9 and 10 below.

© K. Likharev, 2008 17
 Quantum Mechanics

Assuming for simplicity that the boundaries of volume V do not move, it is sufficient to carry out the
partial differentiation of the product * inside the integral. Using the time-dependent Schrödinger
equation (24) and its complex conjugate

 *
 i  ( Hˆ  )* , (1.67)
t
we get
  *  3
dP 
    * d 3 r     *
dt V t   

t

t 
1 
i V 
   *
d r    * Hˆ    Hˆ  d 3 r. (1.68)
V  

At this stage we may notice that due to property (39), the Hamiltonian operator (47),
2 2 
H 
ˆ   U (r ) , (1.69)
2m
acting on a real function, gives a real function again. Hence, the result of its action on an arbitrary
complex function  = a + ib (where a and b are real) is Hˆ   Hˆ (a  ib)  Hˆ a  iHˆ b , where Ĥa and
Ĥb are also real, while ( Hˆ  )*  ( Hˆ a  iHˆ b)*  Hˆ a  iHˆ b  Hˆ (a  ib)  Hˆ  * .32 This means that

dP 1  * ˆ 2 1  * 2
   H  Hˆ  *  d 3 r        2  *  d 3 r .
2m i V 
(1.70)
dt i V   

. Now, applying the general mathematical identity

  
    *   *    * 2    2  * , (1.71)
 
(which may be readily proved by differentiation by parts), we finally get
dP  
   (  j )d 3 r , (1.72)
dt V

where vector j is defined as
 i   * *  *
j        Im   . (1.73)
2m   m  
Vector algebra tells us that a volume integral of a divergence of a vector, like that in the right hand part
of Eq. (72), can be transformed to a surface integral of the normal of the vector over surface S of that
volume. As a result, Eq. (72) may be rewritten as the continuity equation
dP
  jn d 2 r  0 . (1.74)
dt S

32 In Chapter 3, we will discuss properties of such “Hermitian” operators in more detail.

© K. Likharev, 2008 18
 Quantum Mechanics

  
(The formal definition of jn is the scalar product j  n , where n is the externally directed normal vector
of unit length.)
Equations (73) and (74) show that if the wavefunction on the surface vanishes, or at least is
sufficiently small, the total probability w of finding the particle within the volume does not  change,
providing the required sanity check. In the opposite case, dP/dt equals to the flux of vector j through
the surface, with the minus sign. It is clear that this vector may be interpreted as the probability current
density, and its flux, as the probability current. This interpretation may be further supported by
rewriting Eq. (73) using definition (26) of the momentum operator:

 i   * *   1  * pˆ 
pˆ * * 
j             . (1.75)
2m   2 m m 
 

Comparison of this equation with Eq. (23) shows that j cannot be expressed in a regular way via the
momentum operator. (Later we will see that its operator may be presented as an average of the velocity
operator and its Hermitian conjugate.)
To get a better feeling of this observable, it is useful to calculate it for a wavefunction presented
in the amplitude-phase form
  A exp(i ) , (1.76)
with real A and . For this presentation, Eq. (73) yields
 2  
j A  , (1.77)
m
evidently a real magnitude. Notice that for a real wavefunction, or even for that with an arbitrary but
space-constant phase , the probability current vanishes. On the contrary, for the traveling wave (28),
with a constant probability density w  C , Eq. (77) yields the final (and very transparent) result:
2

 
 k p 
jw  w  wv . (1.78)
m m
If multiplied by the particle’s mass m, the probability density w turns into the (average) mass density ,

and the probability current density into the mass flux density v (CM Sec.7.2), while if multiplied by

the total electric charge q of the particle, with w turning into the charge density , j becomes the
electric current density (EM, Sec. 4.1), both satisfying the classical continuity equations similar to Eq.
(74).
By the way, let us recast that equation to a different form. In order to get it, we may rewrite Eq.
(74) as
  
 ( t w    j )d r  0,
3
(1.79)
V

and argue that this equality may is true for any choice of volume V only if the expression under the
integral vanishes in all points:

© K. Likharev, 2008 19
 Quantum Mechanics

w  
   j  0. (1.80)
t
This differential form of the continuity equation is frequently more convenient than its integral form
(74).

1.6. Partial quantum confinement and the dimensionality reduction

Based on our discussion of the rectangular quantum well in Sec. 4, let us discuss the conditions
when a spatial dimensionality of the wave mechanics problem may be reduced. Let us consider a
potential which forms a rectangular quantum well in just one dimension, say x:
 U ( y, x), for 0  x  a x ,
U (r )   (1.81)
  , otherwize.
In this case, we can separate variables as

 n ( x, y, z )   C nx X nx ( x) ( y, z ), (1.82)
nx 1

where functions Xnx(x) are again given by Eq. (60). For each component of sum (82), the 3D
Schrödinger equation yields
2 2
  y , z  U ef ( y, z )  E y , z , (1.83)
2m
where
 2  2 nx2
U ef ( y, z )  U ( y, z )  E x  U ( y, z )  . (1.84)
2ma x2
The last term is sometimes referred to as the (partial) “confinement energy”, but it is important to remember its
kinetic energy origin. Also, if all Xnx are normalized to unity, the normalization condition becomes

P    ( y, z ) * ( y, z )dydz . (1.85)

So, the dimensionality is reduced from 3 to 2; notice, however, that index nx in the effective potential
energy Uef(y,z). There are three possible uses of this result:
(i) If no strong relations between energies are available, accept that (sad:-) reality, solve the 2D
problem for each n, and then sum up the solutions – see Eq. (82). (In this general case, it is not quite fair
to speak about the dimensionality reduction.)
(ii) Ex is much smaller than U(y,z), and may be neglected. This may be the case, for example, if
the potential profile is more steep along axes y and z, than along x. Notice, however, that condition ax 
 does not guarantee the smallness of Ex, because it may be compensated by large values of nx. In this
case (typical for solid state problems), summation over nx is still needed, but frequently may be carried
out analytically, because functions (y,z) are simple sinusoidal waves.
(iii) Counter-intuitively, the most straightforward dimensionality reduction is possible in the
opposite limit, when Ex is much higher than the energy of motion in {y,z} plane. Indeed, in this case the

© K. Likharev, 2008 20
 Quantum Mechanics

distance between levels Exn is so high that in the system was initially prepared to be on (or allowed to
relax to) the lowest, ground level Ex1, the “soft” motion along y, z cannot excite it to higher levels of Ex.
Hence, the system stays at that level, and nx = 1 through the motion, so that the confinement energy Ex1
is constant and may be treated as a potential energy offset.
Another advantage of the last case is that the quantum well in x direction should not necessarily
be rectangular, because this would only change Ex1. For example, much 2D quantum mechanics, has
been implemented in quantum semiconductor heterojunctions (e.g., GaAs-AlxGa1-xAs) where the
potential well in the direction perpendicular to the interface has a nearly triangular shape, with 1D level
splitting for conduction electrons of the order of 10-2 eV. This splitting corresponds to ~100 K, so that
careful experimentation at liquid helium temperatures (4K and lower) may keep the electrons at the
ground 1D level.
If the quantum well is formed in two dimensions (say, y and z), and the conditions formulated
above are fulfilled along both of them, the quantum confinement makes the system one-dimensional,
with effective potential energy
U ef ( x)  U ( x)  ( E y , z )1 . (1.86)

Though such confinement to 1D is difficult to implement in practice, we will start our detailed
exploration of wave mechanics in the next chapter from this case, because it allows a study of most
basic phenomena and concepts of wave mechanics, without involving overly complex math. In that
chapter, for the notation simplicity, we will refer to Uef(x) just as U(x), but one should always remember
about the existence of two other degrees of freedom! We will also consider each wavefunction
component Y(y) and Z(z) normalized, so that the wavefunction normalization condition becomes
P(t )    ( x, t ) * ( x, t )dx , (1.87)

and Eq. (23), for an operator depending only on coordinate x, is reduced to

A (t )    * ( x, t ) A ( x, t )dx . (1.88)

It is also useful to introduce the probability current along the x-axis,

 2 
J ( x, t )   j x dydz 
 ( x, t ) , (1.89)
m x

where jx is x-component of the current density vector j ( x, y, z , t ) . Then the integral form (74) of the
continuity equation takes the form
dPAB
 J ( x A )  J ( x B )  0, (1.90)
dt
where PAB is the probability to find the particle on the segment [xA, xB]:
xB

PAB (t )    ( x, t ) * ( x, t )dx . (1.91)

xA

© K. Likharev, 2008 21