Home Search Collections Journals About Contact us My IOPscience

Kronig - Penney model: a new solution

This content has been downloaded from IOPscience. Please scroll down to see the full text.

1997 Eur. J. Phys. 18 392

(http://iopscience.iop.org/0143-0807/18/5/015)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 128.122.253.228
This content was downloaded on 03/02/2015 at 00:56

Please note that terms and conditions apply.

Eur. J. Phys. 18 (1997) 392–397. Printed in the UK
Kronig–Penney model: a new
solution
Frank Szmulowicz
Wright Laboratory, Materials Directorate, WL/MLPO, Wright Patterson AFB, OH 45433-7707, USA
Received 24 April 1997
Abstract. The one-dimensional Kronig–Penney (KP)
potential consists of a periodic array of square-well potentials.
The Schrödinger equation for an electron in this potential has
a solution in the form of the Kronig–Penney equation (KPE),
which illustrates the formation of electronic energy bands.
The KPE is routinely found from the determinant of a 4 × 4
matrix resulting from four boundary conditions on the
wavefunction and its derivative. Here, a less tedious
approach is pursued, one that is more readily adapted to
multilayer structures. The notational simplicity of the present
formulation pays off when an alternative form of the KPE is
derived. It is shown that the new form has several conceptual,
pedagogic and numerical advantages over the standard KPE.
Among them, in the limit of infinite well-to-well separation,
the present solution readily reduces to the solution for a
single square well; it readily provides the analytic relations
for the top and bottom of an energy band; for energies below
the top of the barrier, it is only on the order of unity, whereas
the KPE can vary over tens of orders of magnitude.
1. Introduction
A periodic array of square wells (henceforth called
an electronic superlattice) approximates the potential
felt by an electron in a one-dimensional atomic
lattice. The allowed electronic states in this potential
are found by solving the corresponding Schrödinger
equation. Its solution has the form of the Kronig–
Penney equation (KPE) [1] and serves as an illustration
of the formation of one-dimensional electron energy
bands [2–4]. Recently, the simple KP model was found
to be applicable to a number of problems involving
artificially structured materials, in particular, layered
semiconductor superlattices [5–9].
However, the KPE has a number of conceptual and
pedagogic disadvantages. First, it requires a tedious
computation of a determinant of a 4×4 secular equation.
Second, it does not readily reduce to the solution for an
isolated square well in the limit of infinite well-to-well
separation or readily provide the conditions for the top
0143-0807/97/050392+06$19.50
c 1997 IOP Publishing Ltd & The European Physical Society
PII: S0143-0807(97)83684-7
Résumé. La potentielle de Kronig–Penney (KP) une
dimensionelle consiste d’un arrangement périodique des
potentielles des puits carrés. L’équations de Schrödinger pour
un électron dans cette potentielle ont une solution dans la
forme d’équation de Kronig–Penney (EKP), qui illustre la
formation des bandes d’énergie électronique. L’EKP est
derivé d’habitude de le déterminant d’une matrice 4 × 4 qui
resulte de quatre conditions de limite s’appliqués à la fonction
d’onde et son dérivatif. Ici, une approche moins enneyeux est
persuit, cette approche est adopté plus facilement pour les
structures multicouché. La simplicité de notation de la
formulation présente produit des dividendes quand une forme
alternatif d’EKP est derivé. Il est montré que la forme
nouvelle a des nombreux d’avantages conceptuelle,
pédagogique, et numérique sur l’EKP standard. Entre les
avantages, la solution présente reduit facilement à la solution
pour un puit carré simple, dans la limite de séparation infinite
des puits. La forme nouvelle produit facilement les relations
analytiques pour le sommet et le fond d’une bande d’énergie.
Pour les énergies au de sous de le sommet de la barriere, la
solution présente est de l’ordre d’unité, tandis que l’EKP peut
varier sur plus de dix ordres de magnitude.
and bottom of a band [5]. Lastly, in some regimes the
KPE can vary over tens of orders of magnitude.
Section 2 establishes the notation for the present
paper. In section 3, the 4 × 4 secular determinant
is rendered diagonal, and thus trivial to evaluate. In
section 4, a novel solution, devoid of the conceptual and
pedagogic problems of the KPE, is found; its properties
are explored in section 5. Conclusions are presented in
section 6.
2. Notation
The electronic superlattice comprises barriers of material
A of width 2a and wells of material B of width 2b.
Since the potential repeats with period d = 2a + 2b,
it is possible to concentrate on only one period, say
the zeroth period, −b < z ≤ b + 2a. In the well,
the carrier has mass mB and, in the barrier, its mass is
mA , a situation which can be justified on the basis of
392
 Kronig–Penney model: a new solution 393

the envelope-function approximation (EFA) [5–7]. The Because K is diagonal, all functions of K are diagonal;
Schrödinger equation in the case of variable mass is [5] for example,
 
h̄2 d 1 dψ exp(ikA b) 0
− + V (z)ψ(z) = Eψ(z), (1) exp(iKA b) = . (11)
2 dz m(z) dz 0 exp(−ikA b)
where For future reference,
 
0 mB − b < z ≤ b,  
V (z) = m(z) = (2) 1 1 mA /kA
V0 mA b < z ≤ b + 2a. MA−1 = . (12)
2 1 −mA /kA
The solutions must obey the Bloch periodicity
condition [2] For a nontrivial solution of equation (6), the 4 × 4
determinant of the matrix must be zero. A lengthy
ψ(z + d) = e iqd
ψ(z), (3) exercise yields the solution in the form of the KPE [1],
where the wavevector −π/d < q ≤ π/d. In the zeroth
period, the solution can be written as cos qd = cos 2kA a cos 2kB b
( A ik z − 12 (α + 1/α) sin 2kA a cos 2kB b, (13)
c1 e A + c2A e−ikA z b < z ≤ b + 2a,
ψ(z) = B −ikB z
(4) where α ≡ mA kB /mB kA .
B ikB z
c1 e + c2 e −b < z ≤ b,
where [5]
q
3. Diagonalization of the secular determinant
kA = 2mA (E − V0 )/h̄2
q (5)
Using the top and bottom rows of equation (7) to solve
kB = 2mB E/h̄2 . for C B in terms of C A , one obtains, respectively,
Using the continuity of ψ and its weighted derivative C B = e−iKB b MB−1 MA eiKA b C A , (14)
ψ 0 /m at z = ±b [5] and the Bloch periodicity
condition, one obtains four boundary conditions, which C =e
B −iqd iKB b
e MB−1 MA eiKA (d−b) C A . (15)
are expressible in the form of a 4 × 4 matrix [2], Equating these two lines, one can arrive at several

eikA b e−ikA b equivalent forms, for example,
 ikA b
−kA e−ikA b /mA  

 kA e /mA µB (−b)µA (−a) − I e−iqd MA eiKA (d−b) C A = 0, (16)
  
 ik
e A (d−b)−qd
e−ikA (d−b)−qd µB (−b)µA (−a) µA (a)µB (b) − I eiqd
kA eikA (d−b)−qd /mA −kA e−ikA (d−b)−qd /mA

 A  × MA eiKA (d−b) C A = 0, (17)
−eikB b −e−ikB b c1
where I is a 2 × 2 unit matrix and
−kB eikB b /mB kB e−ikB b /mB   cA 
 2 
  B  = 0.
−e−ikB b −eikB b   c1  µA (a) ≡ MA ei2KA a MA−1 , (18a)
−kB e −ikB b
/mB kB e ikB b
/mB c2B µB (b) ≡ MB e i2KB b
MB−1 . (18b)
(6) From the definitions in equations (8)–(12), one finds that
It is advantageous to separate out the dimensional  
cos 2kA a imA sin 2kA a/kA
information, as is done below µA (a) =
  ikA sin 2kA a/mA cos 2kA a
MA exp(iKA b) −MB exp(iKB b)
e−iqd MA exp(iKA (d − b)) −MB exp(−iKB b) = µ(−a)−1 . (19)
 A
C Evaluating the 2 × 2 determinant from equa-
× B = 0, (7) tion (16) or (17), i.e. kµB (−b)µA (−a) − I e−iqd k =
C
kµA (a)µB (b) − I eiqd k = 0, one easily obtains the KPE,
where the upper/lower row refers to the boundary equation (13). The corresponding eigenvector is also
conditions at z = b, −b, respectively. In equation (7), more easily found. The method outlined above can serve
 
1 1 as an alternative textbook treatment of the KP problem.
MA,B = , (8)
kA,B /mA,B −kA,B /mA,B However, the problem can be simplified even
 A,B  further. From time-reversal invariance (or the inversion
c
C A,B = 1A,B , (9) symmetry of the KP potential), E(−q) = E(q), which
c2 allows us to switch q → −q, so that the determinant
from equation (17) is equivalent to
and
  kI − e−iqd µA (a)µB (b)k = kI − eiqd µA (a)µB (b)k = 0.
kA,B 0
KA,B = . (10)
0 −kA,B (20)
394 F Szmulowicz

Multiplying the determinant in equation (16) by the after the curly brackets, which leaves
determinant in equation (20), one obtains a pleasing,
symmetric, KPE-like form, (MB tan(KB b)MB−1 +MA tan(KA a − qd/2)MA−1 )X = 0,
1
kµB (−b)µA (−a) − I e−iqd kkI − eiqd µA (a)µB (b)k (27)
2
= k 12 [µA (a)µB (b) + µB (−b)µA (−a)] where
−I cos qdk = 0. (21)  
χ1
X=
Performing the indicated multiplications, the matrix in χ2
equation (21) turns out to be diagonal, with the KPE,
≡ MA cos(KA a − qd/2) exp(i(KA − q)d/2)C A .
cos 2kA a cos 2kB b − 1
2
(α + 1/α) sin 2kA a cos 2kB b (28)
− cos qd = 0, (22)
Using equations (14), (15) and (28) one also finds that
twice along the diagonals. Thus the problem is solved.
This approach is easily generalized to any number of X = MB cos(KB b)C B . (29)
layers per superlattice period.
This new system of equations has a nontrivial solution
when
4. New solution to the Kronig–Penney problem
kMB tan(KB b)MB−1 + MA tan(KA a − qd/2)MA−1 k = 0.
4.1. Tangents-only form (30)
Consider equation (16). After factoring out the Once the zeros of equation (30) are found, the
exponential eiqd , equation (16) can be written out as eigenvector is found from equation (27), and then C A
 
MB e−2iKB b MB−1 MA e−2i(KA a−qd/2) MA−1 − I and C B from equations (28) and (29), respectively.


× MA ei[KA (d−b)−qd] C A = 0. (23)
For energies below the top of the well, the exponential 4.2. Explicit solution
factors can become very large. One way to avoid these Inserting the definitions from equations (8)–(12) into
numerical problems is to replace these exponentials by equation (27) gives rise to the following 2 × 2 equation:
hyperbolic tangents (which are bounded by ±1), as is
 
done next. The labour expended here will be well worth A− − A+ mA A− + A+ mB
it.  + B 
 2 kA 2 kB 
Next, consider the little known operator identity [10]  kA A− + A+ 
kB A − − A+
1 − i tan Ka + B
e−i2Ka = mA
 
2 mB 2
1 + i tan Ka χ1
= (1 + i tan Ka)−1 (1 − i tan Ka); (24) × = 0, (31)
χ2
also, it is easily proved that
where A± = tan(kA a ± qd/2) and B = tan kB b. The
(1 + i tan Ka)−1 = cos2 (Ka)[1 − i tan Ka]. (25) determinant of equation (31) provides the eigenvalue
equation
Observe that the numerator and the denominator
in equation (24) commute. With these identities, DSZ = (A+ + αB)(A− + B/α)
equation (23) becomes
  +(A− + αB)(A+ + B/α) = 0, (32)
1 − i tan(KB b) −1
MB MB where α = mA kB /mB kA , in terms of tangents only.
1 + i tan(KB b)
   Equation (32) is the new solution to the Kronig–Penney
1 − i tan(KA a − qd/2) −1 model. (For numerical purposes, the discontinuities
× MA MA − I
1 + i tan(KA a − qd/2)) due to the tangents are easily removed by multiplying


× MA ei[KA (d−b)−qd] C A = 0. (26) equation (32) by cosines.) Notice that A+ ↔
A− as q → −q, so that equation (32) is even
To remove the denominators, multiply equation (26) on in q. The interchange (A, a) ↔ (B, b) is also
the left-hand side by MB (1 + i tan KB b)MB−1 and insert a symmetry operation. Lastly, using trigonometric
the unit matrix identities, equation (32) can be reduced to the KPE,
   so both methods produce the same results, as they
I = MA 1 + i tan(KA a − qd/2) MA−1
   −1 must. However, some physics may be clearer with
× MA 1 + i tan(KA a − qd/2) MA−1 equation (32), as will be shown next.
 Kronig–Penney model: a new solution 395

Figure 1. Energy dependence of two alternative secular equations for the case of an electron in a GaAs–Alx Ga1−x As
superlattice with x = 0.4, at the centre of the Brillouin zone, q = 0, for a barrier width 2a = 500 Å and three well widths
2b = 67, 133 and 200 Å. (a ) DSZ cos2 (kB b ), using the present formalism, equation (42); (b ) log |DKP |, using the
Kronig–Penney formalism, equation (43). Cusps labelled 1, 2 and 3 indicate the location of the zeros for the cases of
67, 133 and 200 Å wide wells, respectively.

4.3. Wavefunctions which when substituted into equation (4) leads to

unnormalized even-parity solutions,
Wavefunctions are easily found from the eigenvector of
equation (31) and from the definitions in equations (28) ψA = cos kB b cos kA (z − d/2), (34)
and (29). For example, at q = 0, one solution is ψB = cos kA a cos kB z; (35)
the other solution is
   
1 0
X= , X= ,
0 1
   
1 exp(−ikA d/2) mA exp(−ikA d/2)
CA = , (33) CA = , (36)
2 cos kA a exp(ikA d/2) 2kA cos kA a − exp(ikA d/2)
   
1 1 mB 1
CB = , CB = ,
2 cos kB b 1 2kB cos kB b 1
396 F Szmulowicz

which leads to unnormalized, odd-parity solutions, where for numerical convenience DSZ cos2 (kB b) is
shown in order to eliminate the singularities due to the
ψA = mA kB cos kB b sin kA (z − d/2), (37)
tangents. For the same situation, the KPE, equation (13),
ψB = mB kA cos kA a sin kB z. (38) gives
DKP = 1 − cos 2kB b cosh 2κA a
 
5. Properties of the new solution 1 mA kB mB κA
+ − sin 2kB b sinh 2κA a. (43)
2 mB κA mA kB
5.1. Band extrema and a square-well limit
For the present example, DKP ranges over 18 orders of
Consider the conditions for band extrema at q = magnitude, so figure 1(b) actually displays log |DKP |.
0, ±π/d for energies below the top of the well, Each cusp corresponds to a root of DKP . In the same
p ≤ V0 ; define β ≡ mB κA /mAp
E kB , where kA = energy range, DSZ cos2 (kB b) in figure 1(a) is of the
2mA (E − V0 )/h̄2 ≡ iκA and kB = 2mB E/h̄2 . From order of unity (because hyperbolic tangents are bounded
equation (32), for even-parity bands one immediately by ±1) and its zeros are easily located, making it more
finds convenient computationally for any q. Clearly, both
q=0 tan kB b = β tanh κA a −→ β, methods produce the same energy roots.
a→∞
(39)
q = ±π/d tan kB b = β coth κA a −→ β; 6. Conclusions
a→∞

for odd-parity bands,
q=0 tan kB b = − tanh κA a/β −→ −1/β,
a→∞
(40)
q = ±π/d tan kB b = − coth κA a/β −→ −1/β.
a→∞
In the limit of isolated square wells, a → ∞, one
immediately obtains the well known conditions for the
even- and odd-parity levels of a square well of width 2b
and depth V0 [3, 5, 9]. For E > V0 , replace κA → kA /i
and β → α/i, which changes the hyperbolic functions
into regular trigonometric functions. These relations
for the band extrema and the single-well limits are not
readily obtainable from the KPE.
Lastly, equation (32) is a quadratic equation which
can be solved for B. The resulting expression has
a previously unknown solution at q = ±π/2d, ‘the
middle’ of a band:
(α + α −1 )
tan kB b = −
 2  barrier, the present secular equation is more convenient
s
 2
2 cot 2k Aa
× 1 ± 1 +  tan 2kA a. (41)
α + α −1
5.2. Computational convenience
Consider the case of an electron in a GaAs–Alx Ga1−x As
electronic superlattice with x = 0.4. The experimental
parameters characterizing this system are [5]: mA =
0.100m0 , mB = 0.067m0 and V = 0.303 eV, where
m0 is the free-electron mass. Three well widths 2b =
67, 133 and 200 Å, and one barrier width 2a = 500 Å,
will be used.
For q = 0 and for energies below the top of the
barrier, E ≤ V0 , equation (32) becomes
DSZ = [tan(kB b) + tanh(κA a)/β]
×[tan(kB b) − β tanh(κA a)], (42)
where the two factors correspond to the odd- and even-
parity solutions. This equation is plotted in figure 1(a),
This paper demonstrates a new approach to the solution
of the KP problem for an electronic superlattice. First,
the original 4 × 4 eigenvalue problem was rewritten
in a form in which the dimensional information was
separated from the rest of the problem, which made
it possible to reduce the problem to a 2 × 2 equation,
equation (16). Next, the 2 × 2 secular determinant
was made diagonal, equation (21), with the KPE along
the diagonals. This approach should be accessible for
most students with a matrix algebra background and
could serve as an alternative to the existing textbook
treatments of the KP problem.
Second, this paper derives an eigenvalue equation
in terms of tangents only, equation (32). The new
equation immediately reduces to the solution for an
isolated square well in the limit of infinite barrier
thickness, equations (39) and (40). It also immediately
provides the solutions for the top and bottom of a band,
equations (39) and (40). For states below the top of the
computationally, being of the order of unity, whereas
the KPE varies over many orders of magnitude for
a typical superlattice example. Overall, the approach
and the solution presented in this paper have several
conceptual, computational and pedagogic advantages
over the standard KPE and can serve as an alternative
to the existing treatments.
Acknowledgment
This work was performed under Air Force contract
no F33615-91-C-5603 at the Materials Laboratory,
WPAFB, OH, USA.
References
[1] Kronig R de L and Penney W G 1931 Proc. R. Soc. A
130 499
[2] Kittel C 1976 Introduction to Solid State Physics 5th
edn (New York: Wiley) p 192
 Kronig–Penney model: a new solution
[3] Merzbacher E 1970 Quantum Mechanics 2nd edn (New
York: Wiley) p 102 and problem 5, p 115
[4] Leighton R B 1959 Principles of Modern Physics (New
York: McGraw-Hill) pp 396–400
[5] Bastard G 1988 Wave Mechanics Applied to
Semiconductor Heterostructures (New York: Wiley)
[6] Bastard G, Brum J A and Ferreira R 1991 Solid State
Physics: Semiconductor Heterostructures, and
Nanostructures ed H Ehrenreich and D Turnbull
(New York: Academic) vol 44
397
[7] Altarelli M 1986 Heterojunctions and Semiconductor
Superlattices ed G Allan, G Bastard, N Boccora and
M Voos (Berlin: Springer) p 12
[8] Jaros M 1989 Physics and Applications of
Semiconductor Microstructures (Oxford: Clarendon)
pp 220–2
[9] Ivchenko E L and Pikus G 1995 Superlattices and Other
Heterostructures (New York: Springer) pp 75, 112
[10] Merzbacher E 1970 Quantum Mechanics 2nd edn (New
York: Wiley) p 322