Anda di halaman 1dari 17

Philosophical Magazine

ISSN: 1478-6435 (Print) 1478-6443 (Online) Journal homepage: http://www.tandfonline.com/loi/tphm20

Algebraic approximations for transcendental


equations with applications in nanophysics
Victor Barsan
To cite this article: Victor Barsan (2015) Algebraic approximations for transcendental
equations with applications in nanophysics, Philosophical Magazine, 95:27, 3023-3038, DOI:
10.1080/14786435.2015.1081425
To link to this article: http://dx.doi.org/10.1080/14786435.2015.1081425

Published online: 12 Sep 2015.

Submit your article to this journal

Article views: 42

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=tphm20
Download by: [189.234.45.160]

Date: 20 October 2015, At: 18:50

Philosophical Magazine, 2015


Vol. 95, No. 27, 30233038, http://dx.doi.org/10.1080/14786435.2015.1081425

Algebraic approximations for transcendental equations with


applications in nanophysics
Victor Barsan
Department of Theoretical Physics, Horia Hulubei National Institute for Physics and Nuclear
Engineering, Str. Reactorului 30, Magurele 077125, Romania

Downloaded by [189.234.45.160] at 18:50 20 October 2015

(Received 30 April 2015; accepted 5 August 2015)


Using algebraic approximations of trigonometric or hyperbolic functions, a class
of transcendental equations can be transformed in tractable, algebraic equations.
Studying transcendental equations this way gives the eigenvalues of Sturm
Liouville problems associated to wave equation, mainly to Schroedinger equation;
these algebraic approximations provide approximate analytical expressions for
the energy of electrons and phonons in quantum wells, quantum dots (QDs)
and quantum wires, in the frame of one-particle models of such systems. The
advantage of this approach, compared to the numerical calculations, is that the
final result preserves the functional dependence on the physical parameters of
the problem. The errors of this method, situated between some few percentages
and 105 , are carefully analysed. Several applications, for quantum wells, QDs
and quantum wires, are presented.
Keywords: nanostructures; quantum mechanical calculation; quantum dots;
quantum wells; quantum wires

1. Introduction
Quantum dots (QDs), quantum wells (QWs) and quantum wires are essential components of
a huge number of electronic and optoelectronic devices. Their understanding, manufacturing
and control depend, to a large extent, on the precise knowledge of the energy of electrons,
holes, excitons, etc. moving in these devices. The one-particle models give satisfactory
results for a large class of nanostructures. In these cases, the main task of the theory is to
solve the Schroedinger equation for the envelope function or wave function for a simple
potential, in general, described by a piecewise constant function. The wave function has
a simple form linear combination of exponentials with imaginary or real exponents, for
planar heterostructures, or spherical Bessel functions multiplied by spherical harmonics,
both of low order, for devices with spherical symmetry.
However, the eigenvalue equations, which determine the energy, are transcendental
equations, whose analytical solutions are difficult to obtain. Of course, they can be calculated
numerically, with high precision, but their dependence on the physical parameters of the
problem is totally lost. The main goal of this paper is to obtain analytical approximate
Email: vbarsan@theory.nipne.ro
2015 Taylor & Francis

3024

V. Barsan

Downloaded by [189.234.45.160] at 18:50 20 October 2015

solutions for such equations. The results will be applied at various types of QWs, QDs and
quantum wires, as illustrations of this approach.
The structure of this paper is as follows. Section 2 is a short presentation of transcendental equations. In Section 3, we shall expose the physical motivation of this study,
mentioning some applications in quantum mechanics and nanophysics. Section 4 is devoted
to the approximate solutions of the equations sin / = p, cos / = p, giving the
eigenvalues of the symmetric and slightly asymmetric finite square wells. In Section 5, the
equations tan x = ax 1 are studied, using the de Alcantara-BonfimGriffiths (dABG)
algebraization [1]. Physical applications for QDs and KronigPenny model are exposed in
Sections 6 and 7. Section 8 is devoted to a transcendental equation involving trigonometric
and hyperbolic functions, with applications to asymmetric infinite square wells and to a
class of core-shell QDs. The conclusions are presented in Section 9.
2. Transcendental equations an overview
Transcendental equations have ubiquitous presence in theoretical physics. In many cases,
they represent eigenvalue equations of SturmLiouville problems, from quantum mechanics, electromagnetism or elasticity. In this paper, we shall mainly focus on transcendental
equations which give the energy eigenvalue of several quantum mechanical and nanophysical systems.
Among the most popular transcendental equations, we mention the following:
e = p,

sin
= p,

cos
= p

(1)

Similar equations may contain other trigonometric functions (tan, cot), hyperbolic functions (sinh, cosh, tanh) or more complicated algebraic expressions. To solve the aforementioned equations means to find the explicit form of the function ( p), which is equivalent
to obtaining the inverse of each of the functions p = p ( ), defined by (1).
There are at least three general methods of solving such equations. The first is to apply
the LagrangeBurmann [2] inversion theorem. This method can be conveniently used in
few cases, when the derivative of the expressions given in the l.h.s. of Equations (1), or of
similar transcendental equations, has a simple form. This is the case with the first equation
of (1), when the solution ( p) is the Lambert function W ( p), i.e. ( p) = W ( p) . Its series
expansion can be easily obtained using the inversion theorem.
The second general method is based on a domain of the theory of complex functions
the theory of singular integral equations [3] and it was systematically applied by Siewert
and his co-workers (see, for instance [4]). Unfortunately, the results obtained this way are
given in an implicit form, and their use in practical cases is extremely difficult.
The third method, applicable to an important class of transcendental equations, consists
in writing them in differential form, and constructing, from the differential equation obtained
this way, the series expansion of the desired function. This method, invented and re-invented
by several authors [5,6], has been recently applied (see [6]) in order to obtain the first 16
terms of the power series of the solutions of the second and third equation from (1).
As the exact solutions of transcendental equations are, in general, of limited practical use,
it is important to find analytical approximations of such solutions. An attractive approach
is to approximate the trigonometric or hyperbolic functions by algebraic functions and
to transform the transcendental equation into an algebraic equation. For instance, if we

Philosophical Magazine

3025

Downloaded by [189.234.45.160] at 18:50 20 October 2015

approximate with cubic polynomials the restrictions of the functions sin /, cos / on
their intervals of monotony, the respective equations in (1) will be approximately transformed in third-order algebraic equations [7,8]. This is why we shall refer to this approach
as cubic approximation. However, this method does not work for the first root of the
third equation in (1), as the function cos / cannot be conveniently approximated by a
polynomial, for small .
More efficient than the cubic approximation is the approach which has been proposed by
de Alcantara Bonfin and Griffiths [1], with a quite precise algebraization of trigonometric
and hyperbolic functions, for instance:
1 (2x/ )2
cos x  f (s, c; x) = 
s , /2 < x < /2
1 + cx 2
0.45x
, 0 < x < /2
tan x  
f (x) =
1 (2x/ )

(2)
(3)

where s in (2) is 1/2 or 1, and the parameter c depends on the choice of s. They can be
extended to the entire domain of definition of cos and tan using the parity and periodicity
properties of the respective functions. Also, they can be used in order to invert these
functions, for instance:
arctan x 

x
0.45 + (2/ ) x

The algebraic approximations of the trigonometric functions, mentioned in this section,


are obtained by a two-step procedure. In the first step, one finds a simple algebraic function,
which shares some evident similarities with the trigonometric one; for instance, in (2), the
polynomial in the denominator describes qualitatively the bump of the cos function. In
the second step, a corrective factor is added; it has a simple form, compatible with the
symmetry of the exact function, and contains one or more numerical constants, obtained
after a numerical optimization; to refer to the same example, Equation (2), the nominator
is an even function, the value of s is chosen to give a simple factor (s = 1/2 or 1), and the
value of c is given by a least-squares fit. For the approximation given in Equation (2),
in the first step, one simulates the singularity of the tan function with the expression
1/ (1 (2x/ )); the number 0.45 appearing at the denominator is given by a nonlinear
curve fit [1], done in the second step. Of course, the result must be simple enough in order
to transform the transcendental equation in a tractable algebraic one.
Functional dependencies in the solutions ( p) of Equation (1), or of similar ones, can
be obtained by solving numerically the transcendental equation for a set of values of p, and
from the obtained set of data, find an expression that gives the 
best fit for it. This way, one
can obtain, for instance, a polynomial approximation, ( p)  an n . The disadvantages
of this blind method, compared to the algebraization, are that the result is not intuitive
and that we have to keep a large number of terms in the polynomial in order to get a good
precision. The algebraization reduces the number of numerical constants entering in the
approximate solution and gives an intuitive idea about its behaviour.
The main goal of this paper is to use algebraic approximations of trigonometric and
hyperbolic functions in order to obtain reasonably accurate analytical approximations for
the solutions of several transcendental equations relevant for quantum mechanics and
nanophysics.

Downloaded by [189.234.45.160] at 18:50 20 October 2015

3026

V. Barsan

3. Transcendental equations for QWs, QDs and quantum wires


We shall examine some systems for which we can find the energy of electrons, holes or
phonons using approximate solutions of the transcendental equations studied in the present
paper.
A thin GaAs layer sandwiched between thick Al x Ga1x As layers is a typical example
of a QW and is the most common heterostructure [9]. The electron envelope function (we
shall call it wave function, for sake of simplicity) satisfies the Schroedinger equation with
a symmetric finite square well potential. A thin layer sandwiched between two thick layers
of different composition can be described by an asymmetric finite square well.
The radial wave function of an electron in a binary core-shell spherical QD satisfies the
Schroedinger equation with a radial asymmetric square well (the quotation marks intend to
diminish the oxymoronic effect of radial ... square). The problem can be simplified further
by assuming infinitely high barriers at the semiconductor/matrix interface, which is a good
approximation for nanocrystals embedded in silica or similar dielectrics. In the opposite
direction, more complex structures, so-called quantum dotsquantum wells (QDQWs), with
the architecture core A, thin shell B, thin shell A, thin shell B and outer thin shell A, can
be described with more complex potential stepping [10]. The energy of an electron in such
potentials can be conveniently obtained applying the DalgarnoLewis perturbation theory
to the energy levels of simpler square potentials [19,20].
An interesting class of nanowires, consisting in a succession of small, cylindrical QDs
of InAs alternating with GaAs (vertically stacked self-assembled QDs), was obtained in
1995 [21]. This nanosystem is the first realistic illustration of the KronigPenny model,
considered until recently a toy model for the existence of energy bands in solids, and is
important both from fundamental and applicative perspective.
But the interest for piecewise constant potentials is not restricted to nanophysics.
A rectangular potential hole in the middle of a large square well (the case of an impenetrable
well corresponds to the problem 26 of [11]) exhibits the level rearrangement (Zeldovich
effect), which is very pronounced if the distance between walls is much larger than the
dimension of the hole [12]. Few particle systems contained in such a potential (with
impenetrable outer walls) have a negative heat capacity [13]. Several variants of rectangular
potentials are discussed in detail in [14,22,23].
So, there are serious reasons for investigating the transcendental equations satisfied by
the eigenvalues of simple quantum mechanical problems.

4. The equations sin / = p, cos / = p and the finite square wells


4.1. The symmetric nite square well
In order to make more transparent the physical meaning of these equations, let us mention
that the solutions , of
sin = p,

cos = p

(4)

are related to the eigenenergies of a particle in a symmetric finite square well (the sign
alternation will be not discussed here; the issue is not trivial, see, for instance [6]). The
symmetric finite square well is defined by the potential:


a
(5)
|z|
V (z) = V0 1
2

Philosophical Magazine
U

3027

Downloaded by [189.234.45.160] at 18:50 20 October 2015

Figure 1. The asymmetric finite square well (see (15)).

It corresponds to the potential plotted in Figure 1, if = U  = 0, U = V0 . The energy


levels of a particle of mass m, moving in the potential (5) and having the same value inside
and outside the well, is given by the expressions:
E 2n1 =

22
[n ( p)]2 ,
ma 2

E 2n =

22
[n ( p)]2 ,
ma 2

n = 1, 2, . . .

(6)

Actually, the quantities n , n are proportional to the wave vectors:


n =

1
ak2n1 ,
2

n =

1
ak2n
2

(7)

In this context, the parameter p is:


2
p=
a 2mV0

(8)

and it characterizes globally the well (a, V0 ) and the particle (m); a deep (and/or large) well
corresponds to a small p and to a large number of roots n , n (to a large n max ). Other
forms of the eigenvalue equations, equivalent to (4), are:

(1/ p)2 2

, tan =

(9)
tan =

2
2
(1/ p)

3028

V. Barsan

or:

n
X n = arcsin p X n , X 2n1 = n , X 2n = n , n = 1, 2, . . .
(10)
2
It is easy to check (for instance, comparing the plots of functions f (s, c; x) / cos x, for
all s and c, with the constant function C (x) = 1) that the most precise algebraization of
cos is given by f (1, 0.1010164; x); however, as noticed in [1], the function f (1/2, c; x)
has the advantage of transforming the equation for 1 ,
cos 1
= p,
1

0  1 

(11)

Downloaded by [189.234.45.160] at 18:50 20 October 2015

2
in a second-order
 algebraic equation in p . Its solution can be approximated as 1 
1
8
f 2 , 1 2 ; p (see Equation (2)). It has the correct behaviour in the limit of very shallow
wells, which corresponds to p :

1
p

(12)



The function 1  f 12 , 0.2120126; p has a similar expression and an identical
asymptotic behaviour. It represents an even better approximation for 1 ( p) , especially
4
for 0 < p < 1; for p  2, the error
 is about
 10 , and decreases strongly for larger values
of p, for both functions 1  f 12 , c; p (see [1]). The fact that the dABG approximation
for 1 , (12) is so precise is important because the shallow wells play a significant role
in practical applications. As previously explained, it cannot be obtained with the cubic
approximation, used in [6,7].
The QW model for semiconductor heterojunction is largely used in order to interpret
the absorption and photoluminiscence data. Several authors [25,2830], obtained approximate analytical formulas for several quantities of interest using the Barker approximation
[27], which is obtained keeping the first two terms of the exact series expansions of the
wavevectors n ( p) , n ( p), [6,8]. The use of the approximations just described would
increase the precision of their results. The same remark is valid for the analysis of resonant
tunnelling diodes done in [31].
In many realistic cases of semiconductor physics, the effective mass of the electron
has different values, outside and inside the well. The quantum mechanics of particles
with position-dependent mass can be treated within a general frame (see, for instance
[26] and references given there), but if the position dependence can be described by a
step function, an elementary approach can be followed. More precisely, in this situation,
the usual boundary conditions of quantum mechanics (the continuity of the wavefunction
and of its derivative) are replaced with the BenDanielDuke boundary conditions, and the
eigenvalue equations for the wavevector are slightly modified. Let us adapt the previous
approach to the heterostructure already mentioned in Section 3, when a thin GaAs layer
(material B) is sandwiched between thicker Al0.3 Ga0.7 As layers (material A). We shall
consider that the growth direction is Oz, so the layer A intersects the Oz axis on the interval
(a/2, a/2), and the layer B intersects the same axis outside this interval. The conduction
band edge of this heterostructure can be obtained from Figure 1, for the particular case when
(A)
(B)
(A)
(B)
= 0, E c = U, E c = U  , U U  = V0 , so the band offset is E c E c = V0
(see also [9], Fig. 5.5, p. 67). We shall obtain an analytical approximate expression for
(A)
(B)
the energy E of the electronic bound states, E c > E > E c . With the usual notation

Philosophical Magazine

3029

= m B /m A , the Equations (4), valid in the case of usual boundary conditions, are replaced,
for BenDanielDuke boundary conditions, by:
p
cos x
=
,
x
+ (1 ) p2 x 2

p
sin x
=
x
+ (1 ) p2 x 2

(13)

with

Downloaded by [189.234.45.160] at 18:50 20 October 2015

x=

1
ak A ,
2

2
p=
a 2m A V0

(14)

The sign rule is similar to that corresponding to usual boundary conditions ( = 1) .


The first root of each equation can be obtained with the dABG algebraization; the other
roots using a variant of the parabolic approximation for the l.h.s. of (13) [8]. For > 1, the
equations have a finite number of roots.
The physics of heterostructures using the BenDanielDuke boundary conditions has
been analysed in detail by Singh and co-workers [32,33]. They obtain analytic approximate
results using low-order series expansions of trigonometric functions. In these cases, the
algebraization would produce much more precise results, valid on all range of parameters
p and .

4.2. The asymmetric square well


Let us consider an asymmetric well, described by the potential (see Figure 1):


 a
a
, 0
V (x) = V (1 + ) x + V x
2
2

(15)

The parameter measures the well asymmetry; the case = 0 corresponds to a


symmetric well. As previously mentioned, it represents a good model for a heterostructure
composed of a layer sandwiched between two thick layers of different composition.

If the
bound state energy E is written, as usual, in terms of the wavevector k, E = 2 k 2 /2m,
the eigenvalue equation satisfied by k is:
n ka = arcsin

ka
ka
+ arcsin

2P
2P 1 +

(16)

As expected, for = 0, the Equation (16) becomes identical to (10).

4.2.1. The ground state energy


For n = 1 and ka < /2, the transcendental equation (16) can be transformed in a quartic
algebraic equation using the following approximation for the cos function:
cos x 

1 (2x/ )2
 ,


1 + 1 8/ 2 x 2

x 
2
2

(17)

proposed by dABG [1]. We shall not write here the solution of this quartic equation, but it
is however important that an elementary expression, giving a very precise approximation
of the ground state energy, exists.

3030

V. Barsan

4.2.2. Slightly asymmetric well


If the asymmetry parameter is small,
1,

Downloaded by [189.234.45.160] at 18:50 20 October 2015

the potential can describe a heterostructure composed of a thin layer sandwiched between
two thicker, slightly different layers, with a small band offset.
In this case, we obtain, instead of (16), the equation:


n
ka
ka 

sin

=
1
(18)
2
2
2P
4
which is a variant of the Equation (10) of the symmetric well, corresponding to a somewhat
stronger potential. So, in the first order in the asymmetry parameter , the presence of a
slightly higher wall of the well has the effect of the replacement:


P
(19)
P
 P 1+
4
1 4
All the energy levels are consequently given by formulas obtained from (6), making the
replacement (19).
Like in the previous subsections, the results (18) and (19) can be easily extended to
envelope functions in the conduction bands at heterointerfaces, where the usual quantum
mechanical boundary conditions are replaced by BenDanielDuke ones [9,15].
5. The equations tan x = ax 1
In order to evaluate the advantages and disadvantages of the dABG algebraization of the
tangent function (Equation (3)), let us firstly study the equation:
tan x = ax,

a > 0,

x >0

(20)

It gives, for instance, the bound state energies of a particle moving in a cavity bounded
by a rigid wall and a function (problem 27, [11]), or in an infinite square well, with a
negative potential in the middle (problem 19, [11]), or the normal modes of acoustic
phonons in a spherical QD [16], etc.
It has an infinity of roots, corresponding to the intersections of the line y = ax with
the branches of the function y = tan x. We can reduce the equation to the first quadrant,
noticing that, if x0n is the nth non-trivial root of (20) and X 0n is the first quadrant solution
of the equation:
tan x = ax + n
(21)
they satisfy the condition:
x0n = X 0n + n

(22)

We can see how the errors in the evaluation of roots of (18) are generated by the
approximation (3), considering a particular case, namely a = 1:
tan x = x, x > 0

(23)

Plotting the functions tan x, 


f (x) = 0.45x/ (1 (2x/ )) (see Equation (3)) and
y (x) = x, for 0  x < /2, it is easy to see that the algebraization produces a spurious

Philosophical Magazine

3031

x
Log
x

0.001

5 10

2 10

1 10

5 10

Downloaded by [189.234.45.160] at 18:50 20 October 2015

10

Log n

20

50

Figure 2. (colour online) Loglog plot of the absolute value of relative errors of the approximate
roots of Equation (23) obtained through the algebraization (3).

root in this interval. Denoting by x n (a = 1) xn (1), n  1, the exact roots of (27)


corresponding to the intersection of the first bisectrix with the (n + 1)th branch of the
appr
tangent, and comparing them with the roots x n (1) of the equation obtained from (21)
through algebraization of tan function, according to (3):

f (x) =

0.45x
= x + n
1 (2x/ )

we can evaluate the errors


appr

xn (1) xn
xn (1)

(1)

x
x

(24)

(25)

The plot of the value of the relative errors, as a function of n (Figure 2), shows a decrease
of ( x/x)n when n increases, from about 102 for n = 1 at about 105 for n = 50.
The apparently unexpectedly small values of ( x/x)2 , ( x/x)3 , ( x/x)4 are due to
the fact that tan x = 
f (x) at x = 1.40111, while x2 (1) = 1.442 + 2. For n  6, the
error decreases exponentially, for large n. So, we can conclude that the roots x n (1) , n  1
are given by algebraization with reasonable accuracy, and this accuracy increases while n
increases asymptotically.
Concerning the root of (20) from the first quadrant for 0 < a < 1.2, it can be obtained
quite precisely replacing the tangent with its Taylor series near x = 0, cut at the 5th or
7th power; this algebraization gives a quadratic or cubic equation in x 2 . The dABG
algebraization becomes semi-quantitatively applicable for somewhat larger values of a; it
gives an error which decreases from 4%, if a = 2, to less than 1%, if a = 10.
A more subtle algebraization, also proposed by dABG (note 24 of [1]),
 


x 1 1 82 x 2
tan x  f (x) =
(26)
1 (2x/ )2
gives, for tan x = ax, a third-order equation, which does not produce any spurious root.

3032

V. Barsan

Downloaded by [189.234.45.160] at 18:50 20 October 2015

This more subtle approximation is better in the domain where the simpler one is
unsatisfactory, 0  x  1.1, but it fails for larger arguments; it is not a global solution for
the algebraization of tan, but merely a complementary variant. So, this case shows that a
more sophisticated approximation is not necessarily better than a simpler one.
It is useful to extend the dABG algebraization of tangent for an arbitrary value of the
argument. We get:


1
0.45 (x n )
, n
< x < n
(27)
tan x =
2x (2n 1)
2


0.45 (x n )
1
tan x =

(28)
, n < x < n +
2
(2n + 1) 2x
The formulas (26)(28) can be extended to negative arguments using the fact that tan x is an
odd function. For x < 0, the algebraization does not produce, in this case, any spurious
roots. They are also absent for the equations tan x = a/x.
Another application of the algebraization (3) is connected to the fact that the intervals
of monotony of the functions sin x/x, cos x/x, important for a precise description of the
solutions n, n (7), are given by the roots rsn , rcn of the equations:
1
(29)
x
respectively. The advantage of the approximate roots, obtained with (3), compared to
the exact numerical roots, consists in the fact that the approximate ones are explicitly
n-dependent, n being the order of the root. Let us also mention that the roots of the equation
tan x = x are also the roots of the spherical Bessel function j1 (x), and the roots of
tan x = x, tan x =

tan x =

3x
,
3 x2

which can be transformed, through algebraization, in a quadratic or cubic equation, are


the roots of j2 (x).
6. Physical applications: electrons and phonons in a spherical QD
Let us see now how these results can be applied in the study of QDs. In spite of the complexity
of semiconductor QDs, the one-particle approach is satisfactory in many relevant cases. For
wide-gap semiconductors, such as CdSe, the confined electron and hole levels may be treated
independently [16]. For electrons near the bottom of the conduction band, it is sufficient
to consider a single parabolic band. For very high barriers at the semiconductor/matrix
interface, the radial wave function can be assimilated with the solution of a Schroedinger
equation for an asymmetric infinite square well.
These situations lead to simple expressions for the radial envelope wave function for the
electrons in a QD, which is the solution of a Schrodinger-type equation with a rectangular
potential:




ln

r Ylm (, )
(30)
r jl
R
where R is the radius of the QD, and ln are the roots of the spherical Bessel function jl .
The radial envelope function for holes is more complicated (but still a combination of
spherical Bessel functions and spherical harmonics of low order (l  2), and their radial

Philosophical Magazine

3033

argument contains, instead of ln , a parameter , which is the first root of the equation
[16,17]:
j0 (b ) j2 ( ) + j0 ( ) j2 (b ) = 0

(31)

Here, b is a dimensionless parameter, (b2 is a mass ratio light hole/heavy hole) and the
quantity to be found is . In the particular case mentioned by Efros [17], b2 = 0.115, one
finds the value = 5.18961 (actually, Efros gives a somewhat smaller value, = 5.21).
It can be put in the form:
 



b
+ b2
+ b2 2 1 2 = 0
(32)
2
tan b
tan

Downloaded by [189.234.45.160] at 18:50 20 October 2015

As, in this case, 1.5 < = 5.19 < 2, /2 < b = 1. 76 < , the algebraization of
tangent functions gives, according to (27) and (28):
tan bx =

0.45 (x 2 )
0.45 (bx )
, tan x =
2bx
2x 3

(33)

Replacing in the exact equation, we get a 4th-order algebraic equation; its solution
(we
do not give explicitly its solution, which is elementary, but cumbersome), for b = 0.115,
is = 5.14, quite close to the exact value 5.19 (so, with an error of about102 ). We can
obtain a third-order equation, after an algebraization of (32), only when one of the arguments
of the tan functions belongs to the first or fourth quadrant. Knowing the value of b, in terms
of the parameters of the problem, we know the analytic form of holes enveloping function.
The phonons are also essential ingredients for understanding the exciton and polaron
physics, which finally gives the optical properties of the QDs [16]. The quantized frequencies
of the acoustic phonons in a spherical QD are the roots of the equations for Lambs l p = 0
modes for a free-standing sphere:
2
cT
4
j1 (k R) k R j0 (k R) = 0
(34)
cL
Putting

cT
a=4
cL

2
,

kR = x

(35)

one obtains the following equation:


tan x =

ax
a x2

(36)

Through the algebraization of tangent, it becomes a quadratic or cubic equation in x.

7. Vertically stacked self-assembled QDs and the KronigPenny model


A simple model describing the 1D motion of an electron in a periodic potential is generated
by the translation of a rectangular well of depth V0 and length b, with the lattice constant
a, over the real axis (the well-known KronigPenny model [11,18]). In the limit
b 0, V0 , V0 b = const

(37)

3034

V. Barsan
1.0

0.8

0.6

0.4

0.2

10

15

Downloaded by [189.234.45.160] at 18:50 20 October 2015

Figure 3. (colour online) Graphical solutions of Equation (39).

the wells become functions, and the potential takes the form:
V (x) =


2

(x + na)
m n=

(38)

The model provides a simple way of understanding the formation of energy bands in a
periodic solid and of the evaluation of their edges. According to standard books of quantum
mechanics (see, for instance [11]), the values of the wave vector k, which give the band
edges, are obtained from the equation (see Figure 3):




1
cos ka arctan  =

(39)


k
1 + ( /k)2
where ka = K , and a = . Denoting by K 1 , K 2 , . . . the non-zero roots of Equation
(39), written in ascending order, we can see that the even indices correspond to trivial roots,
determined by intersections of the square root (the function in the r.h.s. of (39)) with the
ascending part of the bumps of the |cos| function:
K 2 = , K 4 = 2,

K 6 = 3,

K 8 = 4, ...

(40)

The odd indices correspond to the non-trivial ones, K 4 p+1 , K 4 p+3 , determined similarly
by the intersections of the same function with the descendent part of the bumps, and given,
respectively, by the equations:
a
ka
ka
ka
=
, tan
=
2
ka
2
a
By algebraization of the tan function, we obtain:

1 2
1
K 4 p+1 = 2 p +

+
...
, p>0
0.45 p
0.9 2 p3
2

2 + 0.45
+O
K 4 p+3 = (2 p + 1) +
4
p
p3
tan

(41)

(42)
(43)

Philosophical Magazine


 
2
20

+ 0.9
K1 =
+
9

Also,

3035

(44)

The width of the allowed bands, in terms of the dimensionless wavevector K is:
2

1
(a)
(45)
+
O

K
=

=
K
()
4 p+2
4 p+1
p
0.45 p
p3
Similarly, for the forbidden ones:

Downloaded by [189.234.45.160] at 18:50 20 October 2015

(f)
p ()

= K 4 p+3 K 4 p+2

1 + 0.45
+O
=
4
p
p3

(46)

The KronigPenny model has been considered until recently a toy model in the
absence of any 1D (or quasi-1D) physical system it could describe: more exactly, until
1995, when a nanowire, composed by pieces of InAs alternating with GaAs, has been
produced [21]. The incorporation of BenDanielDuke boundary conditions in the model
has been worked out in [24] and will be not repeated here. The algebraization allows a
deeper understanding of the physical significance of the KronigPenny model results.
8. Equations involving trigonometric and hyperbolic functions
Such an equation appears in the elementary problem of an asymmetric infinite square well
(see, for instance [14], problem 13), defined by the potential:
V (x < 0) = , V (0 < x < a) = V1 ,
V (a < x < b) = V2 > V1 , V (x > b) =

(47)

(see Figure 4).


As already mentioned, it is relevant for a class of QDs embedded in a matrix with a high
energy gap. Presuming that
0 < E < V2 V1
and denoting by

2m
k=
(E V1 ),
2


K =

2m
(V2 E),
2

2m
(V2 V1 ) = k 2 + K 2
2
b a = b, k0 b = p, kb = x (48)
k02 =

we can write the eigenvalue equation for the energy in the form:

x cot x (1 ) = p2 x 2 coth p2 x 2

(49)

One could try to solve this equation using an algebraization of the hyperbolic functions
proposed in dABG (Equation (45) of [1]), which gives:
coth x  F (x) =

0.0572x 2 + 0.286x + 1
x (1 0.0572x)

(50)

Just comparing the plots of the function coth x/F (x) and of the constant function
C (x) = 1, it is clear that the previous approximation is a poor one. However, a key

3036

V. Barsan

Downloaded by [189.234.45.160] at 18:50 20 October 2015

V2

V1

Figure 4. The asymmetric infinite square well potential (see (47)).

0.4

0.2

10

12

14

0.2

0.4

Figure 5. (colour online) Plot of the two sides of (49), for = 0.1, p = 14.

for solving the Equation (49) emerges from noticing the very elementary fact that the
trigonometric (hyperbolic) function has a rapid (slow) variation for a large range of values
of physical parameters entering in the respective relations. For instance, let us consider
(Figure 5) the realistic case = 0.1, corresponding to a spherical QD, where the shell
represents 10% of the total radius (this is the only place where we use the fact that is
a small quantity), p = 14; 1/ p is the potential strength of height V2 V1 and range a,
corresponds to a mass m (see Figure 5). A simple numerical exercise shows that replacing the

Philosophical Magazine

3037

function in the r.h.s. of (49) with a conveniently chosen segment of its tangent, we obtain
the roots of this equation with an error smaller than 103 . So, considering the simplest
algebraization of the r.h.s. as
an (, p) x + bn (, p)

(51)

Downloaded by [189.234.45.160] at 18:50 20 October 2015

and the dABG algebraization of the tangent, we obtain the approximate roots of (49) from
a third-order algebraic equation, with errors decreasing from 5% (for the smallest root) to
1.6% (for the largest one).
9. Conclusions
Using algebraic approximations of trigonometric or hyperbolic functions, it is possible to
transform a class of transcendental equations in approximate, tractable algebraic equations.
As the algebraization used in this paper is, to a certain extent, an ad hoc procedure, this
approximation must be used with a certain caution in order to avoid the appearance of
spurious roots or of roots with too large errors. A more sophisticated approximation is not
necessarily better than a simpler one. The conclusions based on eye naked examinations
of indistinguishable curves may be tricky. The frequent check of the results by numerical
examples or by graphical methods is highly recommended. Such issues have been studied
in detail in Section 5, for the equation tan x = ax.
The transcendental equations solved this way are, in general, eigenvalue equations
of the energy, with relevance for quantum mechanics and nanophysics; they are part of
SturmLiouville problems, a theory covering a huge area of research, from optoelectronics
to elasticity or quantum graph problems [34]. They can be used in a large number of
models for heterojunctions, QDs and quantum wires, when the one-particle description
is applicable, in terms of wave functions or enveloping functions. These models contain
intrinsic approximations (for instance, the spherical QDs are not perfectly spherical),
producing errors of the order of a few percentages; consequently, an exact expression for
the eigenvalues produced by such a model would not present a practical interest. So, if
the advantage of an exact solution, compared to an approximate one, obtained through
algebraization is decisive for a mathematician, it is of secondary interest for a researcher
working in applied physics. Also, an important benefit of approximate analytical solutions
is the fact that they provide a semi-quantitatively precise image of the dependence of the
results on the physical parameters of the problem (effective masses, potential barriers, etc.)
Several applications of this approach to multiple heterostructures, core-shell spherical
QDs, KronigPenny model and quantum wires are worked out, as an illustration of the
method. Similar approaches can be used to many other problems of nanophysics and
quantum mechanics.
Disclosure statement
No potential conflict of interest was reported by the author.

Funding
This work was supported by Autoritatea Nationala pentru Cercetare Stiintifica [PN 09 37 01 02/
2009], [grant number 04-4-1121-2015/2017]; HOPE Network [2013-3710_540130-LLP-1-2013-1FR-ERASMUS-ENW].

3038

V. Barsan

Downloaded by [189.234.45.160] at 18:50 20 October 2015

References
[1] O.F. de Alcantara Bonfim and D.J. Griffiths, Am. J. Phys. 74 (2006) p.43.
[2] E.T. Whittaker and G.N. Watson, A Course of Modern Analysis, Cambridge University Press,
Cambridge, 1996.
[3] N.I. Muskhelishvili, Singular Integral Equations, Nordhoff, Groningen, 1953.
[4] C.E. Siewert, C.J. Essig and Z. Angew, Math. Phys. 24 (1973) p.281.
[5] V.V. Pogosov, V.P. Kurbatsky and E.V. Vasyutin, Phys. Rev. B 71(2005) (2005) p.195410.
[6] V. Barsan, Philos. Mag. 94 (2014) p.190.
[7] V. Barsan, Rom. Rep. Phys. 64 (2012) p.685.
[8] V. Barsan and R. Dragomir, Optoelectron. Adv. Mater. Rapid Commun. 6 (2012) p.917.
[9] T. Ihn, Semiconductor Nanostructures, Oxford University Press, Oxford, 2010.
[10] D. Dorfs and A. Eychmueller, Multishell semiconductors, in Semiconductor Nanocrystal
Quantum Dots, A. Rogach, ed., Springer, Wien, 2008, p.101.
[11] S. Fluegge, Practical Quantum Mechanics, I, Springer, Berlin, 1971.
[12] M. Combescure, A. Khare, A. Raina, J.-M. Richard and C. Weydert, Int. J. Mod. Phys. 22 (2007)
p.3765.
[13] P. Serra, M.A. Carignano, F.H. Alharbi and S. Kais, Europhys. Lett. 104 (2013) p.16004.
[14] F. Constantinescu and E. Magyari, Problems in Quantum Mechanics, Pergamon Press, Oxford,
1971.
[15] G. Bastard, Wave Mechanics Applied to Semiconductor Heterostructures, Les Editions de
Physique, Paris, 1989.
[16] M.I. Vasilevskiy, Exciton-phonon interaction in semiconductor nanocrystals, in Semiconductor
Nanocrystal Quantum Dots, A. Rogach, ed., Springer, Wien, 2008, p.217.
[17] A.I.L. Efros, Phys. Rev. B 46 (1992) p.7448.
[18] C. Kittel, Introduction to Solid State Physics, 8th ed., Wiley, New York, 2005.
[19] H.A. Mavromatis, Exercises in Quantum Mechanics, Kluver Academic Publishers, Dordrecht,
1992.
[20] H.A. Mavromatis, Am. J. Phys. 59 (1991) p.738.
[21] Q. Xie, A. Madhukar, P. Chen and N.P. Kobayashi, Phys. Rev. Lett. 75 (1995) p.2542.
[22] C.A. Duque, M.E. Mora-Ramos, E. Kasapoglu, H. Sari and I. Sokmen, Eur. Phys. J. B 81 (2011)
p.441.
[23] R.P. Lungu, Phys. Scr. 75 (2007) p.206; Rom. J. Phys. 44 (1999) p.939; 45 (2000) p.427; 45
(2000) p.25; Rom. Rep. Phys. 52 (2000) p.265; 52 (2000) p.233.
[24] M.S. Kushwaha, J. Chem. Phys. 135 (2011) p.124704.
[25] S. Kabi, S. Panda and D. Biaswas, J. Appl. Phys. 109 (2011) p.053110.
[26] A.G. Nikitin and T.M. Zasadko, J. Math. Phys. 56 (2015) p.042101.
[27] B.I. Barker, G.H. Rayborn and J.W. Ioup, Am. J. Phys. 59 (1991) p.1038.
[28] D. Biswas, S. Kumar and T. Das, J. Appl. Phys. 101 (2007) p.026108.
[29] D. Biswas, S. Kumar and T. Das, Thin Solid Films 515 (2007) p.4488.
[30] D. Biswas, S. Kumar and T. Das, Mater. Lett. 61 (2007) p.5282.
[31] A. Dargys and P. Cimmperman, Solid-State Electron. 45 (2001) p.526.
[32] V.A. Singh and L. Kumar, Am. J. Phys. 74 (2006) p.412.
[33] S. Singh, P. Pathak and V.A. Singh, Am. J. Phys. 32 (2011) p.1701.
[34] F.W. Williams, W.P. Howson and A.J. Jones, Proc. R. Soc. A 468 (2007) p.3195.

Anda mungkin juga menyukai