Tamm States

The possibilities to realize and manipulate a wide range of one-dimensional potential profiles in semiconductor
superlattices provide the driving force to study the surface state [As a result] the subject of electronic states under
an electric field has attracted much attention for its potential device applications.
16


Introduction

As the demand for smaller, faster devices increases, semiconductor device physics
continue to play an important role in the realization of these devices. The Tamm theory, for
example, can be used to predict the existence of electronic states at the surface of a lattice.
1
This
is very useful when dealing with devices that employ heterostructure design. If we can
analytically calculate or numerically solve for the affect of surface states on the potential, carrier
transport, etc., of a material, then we can gain a better understanding of how to operate a device
of this material.

In this paper I will explain, in detail, Tamms theory of surface states, give an account of
the first experimental observation of the Tamm state, and briefly discuss several developments
that have resulted from the experimental observation. However, before I begin to explain
Tamms theory of surface states, it is necessary that I give a brief overview of wave mechanics
so that Tamms theory can be fully understood.

Wave Mechanics

It is a well-known result of solid state physics that the wave function of a particle, e.g.
electron or hole, traveling in a material in one dimension (1D) can be expressed by the equation

(d
2
/dx
2
)+ = |
2
+, |
2
= m8t
2
(V(x) - c)/h
2
(1),

where (1) is the Schrodinger equation, + is equal to the particles wave function, m is equal to
the particles mass, c is equal to the particles kinetic energy, V(x) is equal to the materials
potential energy at point x, and h is equal to Plancks constant.

Another interesting result of solid state physics is that we can model a one-atom system
with a particle in a box of some finite potential. For a particle confined to potential well that is L
units in length and V
o
units below the material barrier potential, the Schrodinger equation
becomes

(d
2
/dx
2
)+ = -k
2
+, where k
2
= m8t
2
c/h
2
, (-L/2 < x < L/2), (2a)
and


(d
2
/dx
2
)+ = |
2
+, where |
2
= m8t
2
(V
o
- c)/h
2
(|x| > L/2)
14
, (2b)

and the potential well is centered at the origin. The solution to the Schrodinger equation has the
form

+
1
= Ae
|x
(x < -L/2),
2

+
2
= Bsin(kx) + Ccos(kx) (-L/2 < x < L/2),

+
3
= De
-|x
(x > L/2)
14
,

where k and | are given above. It is interesting to note how the wave function is effected if the
barrier width is made finite instead of semi-infinite. This resulting wave functions for semi-
infinite and finite barrier potential are shown in Figure 1.







Now that we understand the basics of the Schrodinger equation, let us consider a more
practical application. Suppose we have an infinite 1D lattice. This lattice can be modeled by
series of periodic potential wells [V(x) = V(x + na), where n is an interger and a is the period]
that are a result of the Coulomb potentials caused by the periodic atomic structure of the lattice.
An example of such a potential structure is shown in Figure 2 where

V(x) = V
1
(-b < x <= 0), (3a)
and
V(x) = 0 (0 < x <= a), (3b)
and
V[x + (a + b)] = V(x). (3c)




3

In order to calculate the wave function for a periodic potential structure, the Schrodinger
equation must be used. The details to the solution of this problem are very complicated;
however, the result is known and given by the expression below.

+
k
(x) = u
k
(x)e
ikx
, (4)

where u
k
(x) is a function that has period a (just like the potential) and k is the wave number to be
determined by boundary conditions. This result is known as the Bloch theorem
12
, and it can be
used to described the wave function of a particle traveling in a material of periodic potential.

Now lets take another look at the potential structure described by equation (3). In order
to obtain the wave function we must solve the Schrodinger equation. We know that the solution
must be in the form of equation (4). Substituting equation (4) into equation(1), we obtain the
following result.

u + 2iku (k
2
+ k
2
0
)u = 0, (-b < x <= 0) (5a)

u = Aexp[(k
0
ik)x] + Bexp[-( k
0
+ ik)x], (-b < x <= 0) (5b)
where
k
0
= (V
1
- c)
1/2
, (5c)

and
u + 2iku (k
2
+ k
2
1
)u = 0, (0 < x <= a) (6a)

u = Cexp[i(k
1
k)x] + Dexp[-i( k
1
+ k)x], 0 < x <= a) (6b)

k
1
= c
1/2
. (6c)

The additional constants in equation (2a)-- m8t
2
/h
2
--have been normalized to one in order to
make the k, c relation simpler. After applying continuity conditions to the solutions of (5) and
(6), we obtain the Konig-Penny (KP) or dispersion relation.

cos[k(a + b)] = cosh(k
0
b)cos(k
1
a) + [(k
2
0
k
2
1
)/2 k
1
k
0
]sinh(k
0
b)sin(k
1
a). (7)

By replacing the potential barriers of Figure 2 with o-potential barriers (Figure 3),
the KP relation can be approximated by the following equation:
4

cos(ka) = cos(k
1
a) + [p sin(k
1
a)]/ k
1
a, (8a)

where
p = lim
b 0, V1
k
0
ab. (8b)



We get two important results from the KP model. (1) The energy spectrum for particles
traveling in a periodic potential material consist of discrete values, and (2) the wave function is
given by
+(x) = C[exp(ik
1
x) + exp(-ik
1
x)], (0 < x <= a) (9a)
where
= D/C = {1-exp[i(k
1
- k)a}/{exp[-i(k
1
+ k)a] 1}. (9b)

At this time I will take a few moments to discuss another approach to calculating wave
functions. As stated before, the wave function of an electron in a one-atom system can be
modeled by a particle in a potential box. Now, we can assume that for a given energy, c
k
, the
particle will have an atomic wave function, v
k
(x), that describes the motion of the particle
confined to the atom.

We can see from Figure 1b that the wave function of the particle extends beyond the
potential well. As a result, if we would like to calculate the wave function of a system of
infinite, periodic potential wells, we must consider the possible coupling of wave functions from
well to well. However, if the coupling is small enough, we can approximate the wave function
of the system by a linear combination of atomic, or quasi-atomic, wave functions.
13
This
approach to wave function calculation is known at the Tight-binding approximation (TBA).
TBA can be expressed analytically by the following equation:

+
k
(x) = E
n
C
kn
o(x x
n
) (10)

where n is the number of atoms in the system, x
n
is the position of the n
th
atom, o is the quasi-
atomic wave function, and +
k
is the wave function of the system. Furthermore, it can be shown
that C
kn
= n
-

exp(ikx
n
) which, substituting into equation (10) gives

+
k
(x) = n
-
E
n
exp(ikx
n
) o(x x
n
) (11)

5
where equation (11) is of the Bloch form.
13
It is also interesting to note that the quasi-atomic
wave function, o, can be expressed as a linear combination of the atomic wave functions, v
n
.
That is, o(x) = E
n
b
n
v
n
(x).
15

Tamm (Surface State) Theory

Now that we have a basic understanding of wave mechanics for a particle in a system of
periodic potential wells, we are ready to introduce the concept of surface states. There are,
however, a few fine details that we must consider before presenting Tamms model. For an
infinite crystal system, the value of the wave number, k, must be real in order to ensure that the
wave function, +
k
, is finite. On the another hand, if the crystal system is limited by a surface
(constant potential of vacuum, V
0
, for - < x <= 0), then k may also assume a complex value.
12

The fact that k may be complex is very significant. When the value of k is complex, the
corresponding energy levels can exist in the region that is forbidden to particles of an infinite
crystal system. It is also shown that electrons of these forbidden energy levels are localized near
the crystal surface, thus the name surface states.
12


At this time we will revisit the KP potential model given in Figure 3, but now we will
terminate one end of the crystal with a semi-infinite surface potential. (Figure 4) This is the
same model Tamm used to calculate the existence of surface states.
1



Returning to the dispersion relation, we have the equation

cos(ka) = cos(c) + pc
-1
sin(c) where c = ac
-
(12).

For a particle (electron) traveling in a vacuum of potential V
0
we know that the wave
function must decay, and therefore, has the form +
vac
(x) = D exp(k
0
x) where k
0
= (V
0
- c)

, c <
V
0
, and x < 0. Solving equation (12), we find that the wave number k has a complex form given
by
k = iu + mt/a, m = 0, 1, 2, (13)

and we assume u > 0 for x > 0.

Solving the Schrodinger equation for the system of Figure 4, applying continuity
conditions, and using the form of k given by equation (13), we find the following to be the
solution for the wave function in the crystal.

+
crys
(x) = A[exp(ik
1
x) + exp(-k
1
x)] (14)
6

where k
1
is given by equation (6c) and

= [(-1)
m
exp(-ua) exp(ik
1
a)]/[exp(-ik
1
a) (-1)
m
exp(-ua)] (15)

m = 0, 1, 2,

It can also be shown that +
crys
(a) = (-1)
m
+
crys
(0) exp(-ua) which supports that fact the +
crys
is
localized at the surface x = 0.
12
Using the relative probability density function defined by
Thakkar and Steslicka, we can see that the wave function is localized at the surface. (Figure 5).

We can also apply these principles to a finite crystal with potential shown in Figure 6.
The probability density function for a 10-well system is shown in Figure 7. This figure supports
our intuitive assumption that there will be localized states at both surfaces.


Tamm States First Contact

In 1990, H. Ohno et al. became the first to experimentally observe Tamms surface
states.
2,3
In this section I will discuss the details of the experiment conducted by H. Ohno et al.
as well as present the data collected from this experiment. However, before I discuss the
experiment, it is necessary that I review several concepts superlattices, minibands, and electric
fields on superlattices.

7
Although Tamms theory involved the prediction of surface states in a semi-infinite
uniform crystal, H. Ohno et al. experimentally observed these surface states in a superlattice.
For this reason, it is important that we have a basic understanding of the band structure
properties of a superlattice (more specially, GaAs/Al
y
Ga
1-y
As). In a superlattice the stacking is
periodic (see Figure 8), and such a structure can only be realized if crystal growth from one
semi-conductor to another is possible. This implies that the two semi-conductors must have the
same type of crystalline lattice and have the same size elementary cell.
4


Though these semiconductors (GaAs and Al
y
Ga
1-y
As for our case) have similar geometry,
their chemical composition, and consequently their band gaps, will be different as shown in
Figure 9.
4
Because of the difference in band gaps, a potential well will form in the conduction
and valence bands of the superlattice. As a result of wave function coupling from well to well
(for a ~ b), the energy bands widen to form minibands.
4
(Figure 10)




As part of their experiment, H. Ohno et al. applied an electric field across the superlattice
and measured the change in the surface state and bulk state energies.
2,3
For the purpose of
gaining some insight into this process, I will briefly discuss the effects of an electric on a
superlattice. If we apply a constant electric field, F, on a one-well structure (Figure 1), we will
see a change in potential. (Figure 11) The Schrodinger equation for such a structure is given by

- (h
2
/8t
2
m) (d
2
/dz
2
)+(z) + |e|Fz +(z) = c+(z), |z| <= L/2 (16a)
and
- (h
2
/8t
2
m) (d
2
/dz
2
)+(z) + (V
0
+|e|Fz) +(z) = c+(z), |z| > L/2 (16a).
9

8

After solving equation (16), we obtain a set of complex eigenvalues of the form

c = E
0
- iI/2, I > 0 (17)

where E
0
and I are the quasibound-state energy level and the resonance width respectively.
9

Figure 12 is a normalized plot of E
0
vs. F and I vs. F. E
(0)
equals the ground state energy of a
particle in an infinite potential well.

Now, we are ready to discuss the details of the first experimental observation of Tamm
states. The first question is, Why use a GaAs/AlGaAs superlattice? H. Ohno et al. decided on
a semiconductor superlattice because the material allowed them to construct and control a wide
range of surface potentials.
2
Because of the complicated nature of a real surface, it is difficult to
observe Tamm states. Whereas with a superlattice, an ideal surface can be realized, and the
pure Tamm states can be observed.
3
Samples used by H. Ohno et al. were grown by molecular-
beam epitaxy on n
+
GaAs (001) substrates. Starting with the substrate side, the undoped
superlattice consists of 200 Angstrom of Al
x
Ga
1-x
As, nine GaAs wells (L
W
= 40 Ang.) separated
by Al
x
Ga
1-x
As barriers (L
B
= 40 Ang.), and 200 Ang. of Al
y
Ga
1-y
As; this structure was repeated
three times. H. Ohno et al. also report that two samples were prepared, A and B. Sample A and
B were identical in geometry, but different in composition. For sample A, x = 0.2 and y = 1; for
sample B, x = 0.3 and y = 0.15.
2,3
In addition, a ~ 2 um Si-doped n
+
-type GaAs buffer layer was
grown before the superlattice structures. Also, a n
+
-type GaAs was grown on the superlattice of
sample A to minimize the effect of the band bending associated with a real sureface.
2
A p
+
-type
GaAs layer was grown on sample B, making sample A an n
+
-i-n
+
structure and sample B a p
+
-i-
n
+
structure.
3


The superlattice construct by H. Ohno et al. differs from other superlattices because the
top clad layer of AlGaAs is grown such that its potential level differs from those in the bulk. The
potential difference of this clad layer creates the ideal surface needed in order to observe
Tamm states. H. Ohno et al. refer to this surface as an internal surface, but a surface none the
less.
2,3
Figure 14 shows the first miniband energy levels as a function of the barrier difference,
AE, of the top clad layer. One can see that the highest (lowest) energy level begins to peel away
from the miniband as AE increases (decreases), but the rest of the miniband remains virtual
unchanged.
3
The probability density functions for AE = 500 meV and AE = -150 meV are also
shown in Figure 14. Notice that the highest (lowest) energy particle becomes localized in the
first quantum well while the other particles remain delocalized throughout the superlattice.
9

To observe the Tamm states, photoluminescense
5
(PL), photoluminescence excitation
(PLE), and photocurrent
5,6
(PC) spectra were measured at 5 K under excitation by a Kr
+
laser-
pumped LD700 dye laser in the range of 1.60-1.72 eV.
2
In addition, 100 Ang. of Au deposited
on the surface and an In Ohmic contact at the backside of the sample were used to apply electric
fields with magnitudes up to 40 kV/cm.
2,3


Not much data was collected by the PL spectrum; however, the PLE spectrum (under
zero electric field) gives interesting results. (Figure 15) One can immediately see the electron to
heavy-hole (HH) and electron to light-hole (LH) transitions at 1.6227 eV and 1.6419 eV,
respectively. Here, the GaAs bandgap in assumed to be 1.519 eV. The observed peak energies
agree within 1 meV of the theoretical predictions.
2,3
Another peak is observed near 1.67 eV.
The possibility of this third peak being related to excited states of the heavy-one miniband is
rejected since the transition is forbidden and cannot produce the sharp excitonic features shown
in Figure 15.
2
This third peak (1.6697 eV) is indeed a result of Tamm state electron to heavy
hole transition (Tamm HH).
2,3


The particles of a superlattice begin to localize when an electric field is applied. This
phenomenon facilitates the observation of particle transitions in a particular well. Figure 16
summarizes the energies observed from the PC spectra under various bias voltages, V
b
i.e.
10

electric fields. Solid circles are the heavy-hole-related transitions; open circles are light-hole-
related transitions, and open triangles are unidentified transitions. The 0, +/- 1, and +/- 2
correspond to direct, nearest-neighbor, and next-nearest-neighbor transitions. Also marked on
the plot are the letter A, B, C, and D. Region A shows anti-crossing of two transitions.
Monolithic decrease in transition energy for increasing electric field is observed in region B, and
new branch are found at C and D.
2
Figure 17 shows only heavy-hole related transition energies
as a function of electric field.


11
Modeling of Surface State Properties

Superlattice barrier height and length can be readily controlled.
23
Consequently,
superlattices have become instrumental in the study of Tamm states. As a result, several groups
have presented analytical and numerical solutions that further help to explain wave functions,
minibands, field effect, etc., of the superlattice Tamm states. (i.e. ideal surface state) In this
section I will discuss the work done by a few of these groups.

In 1990, Steslicka and Kucharczyk presented a paper whose aim was to propose a
Kronig-Penny-like model appropriate for a terminated compositional supperlattice (e.g.,
vaccum/GaAs/Al
x
Ga
1-x
As/GaAs/) and a mass-modulation superlattice (e.g.,
vaccum/In
0.69
Ga
0.31
As/InP/In
0.69
Ga
0.31
As/).
17
By using direct matching procedure, they
analytically calculated and numerically solved the energy and localization properties of surface
states.
17
The model used for this calculation has the potential profile given in Figure 18 where
m
1
is the electron effective mass in GaAs (0.067m
0
) and m
2
is the electron effective mass in
Al
x
Ga
1-x
As (m
2
= m
1
+ 0.083x).
17


Using the direct matching procedure with Bastards
19
boundary condition at z = a,
Steslicka and Kucharczyk have found the wave function to be of the form

+
SL
(z) = A[sin(kz) + cos(kz)], c < V, (0 < z < a) (18a)

with A as a normalization constant and

= [sin(k
1
a) + Ksinh(k
2
b)exp(ikd)]/[cosh(k
2
b)exp(ikd) cos(k
1
a)] (18b)
where
K = k
1
m
2
/k
2
m
1
, (18c)
k
1
= (2m
1
c)

, k
2
= (2m
2
|c-V|)

, (18d)
d = a + b.
17
(18e)

For electronic states that lie in the m
th
miniband gap, the wave number, k, is of the form given by
equation (13). In the vacuum region (z < 0), the wave function have the form

+
V
(z) = Fexp(k
3
z), where k
3
= [2(U - c)]

. (19)

Using the probability density function, P(z) = |+(z)|
2
/|+(0)|, Steslicka and Kucharczyk have
plotted the particle localization for the first and second minigaps as a function of space. (Figure
12
19) Notice that the first minigap has only one maxima per period while the second minigap has
two maxima per period. It is also true that the n
th
minigap while have n maxima per period.
Other calculations done by Steslicka and Kucharczyk include miniband and surface state energy
as a function of x composition for Al
x
Ga
1-x
As and the calculation of first minigap surface state
energy as a function of a and b in In
0.69
Ga
0.31
As/InP effective-mass superlattice. (Not shown)

As a follow up to the experimental observation, Glessner and Ru presented calculations
of the electronic structure of a finite AlGaAs/GaAs superlattice using the parameters reported by
H. Ohno et al.
18
Glessner and Ru reported that, because of the translational invariance parallel to
the surface, the three dimensional (3D) conduction electron wave function has the form

+
k||
(x
||
, z) = 1/A

exp(ik
||
x
||
)o
i
(z) (20)

where x
||
and k
||
are position and wave vector in the plane parallel to the surface, and A is the
surface area. The quantized motion of an electron is the z direction in characterized by the wave
function o
i
(z). For convenience they have also assumed the both ends of the superlattice meet a
hard wall.

Solving the Schrodinger equation using the wave function

o
i
(z), electron energies are
found to have the form

c
i
(k
||
) = (hk
||
2
)/(8t
2
m*) + c
I
(21)

where m* is the effective mass of an electron. Glessner and Yu also report that the wave
function o
i
(z) can be expanded by a series of sine functions which vanish at z = 0 and z = L.
22


13
The results of Glessner and Yus calculations are presented in Figure 20. In this figure,
the Eigenenergy for the first and second minibands are plotted as a function of the surface
potential barrier difference AE. Also shown in Figure 20 are the probability density functions of
the lowest 15 states at AE = -0.2 eV and AE = -0.13 eV. Glessner and Yus results are not only
in good agreement with experimental observation but also predict the existence of Tamm states
not seen by H. Ohno et al.
18


Work done by M. M. Vrubel predicts the electron eigenenergies while under the
influence of a constant electric field. In his first paper, Vrubels model makes it possible to
analyze the behavior of two surface states of a quantum well in an electrostatic field and their
interaction with size-quantized levels and with each other.
7
Supposedly, his model can be
generalized to those models considered in references 2 and 17.

Vrubel begins with the form of the Schrodinger equation given by equation (16a) and
then performs a change of variable given by

Z = - [2m/(ekF)
2
]
1/3
(c - eFz), (22)

and the Schrodinger equation becomes

(d
2
/dz
2
) +(Z) - Z+(Z) = 0,
7
(23)

Equation (23) has the solution of the form

+(Z) = aAi(Z) + bBi(Z) (24)

where Ai(Z) and Bi(Z) are Airy functions.

Using the conditions given by Volkov and Pinsker
20,21
and incorporating Vrubels change
of variable, we obtain the following boundary conditions.

14
{1 R[2meF/k
2
]
1/3
(d/dZ)}+(Z
0
) = 0 (25a)

{1 R[2meF/k
2
]
1/3
(d/dZ)}+(Z
L
) = 0 (25b)

where Z
0
and Z
L
are associated with left and right well edges, respectively, and R is a material
constant given by

R = (k
2
/2mE
g
), where E
g
is the material band gap.
7
(25c)

Included in Figure 21 are plots for first miniband energies verses electric field for various values
of R.

In his second paper, Vrubel explicitly calculated the affects of an electric field on the
surface state energies.
8
In agreement with others, Vrubel find the eigenenergies to be of the form
given by equation (17). His results are shown in Figure 22.



From Surface (Tamm) States to Intersurface States
15

The superlattice continues to be an excellent structure for the realization of Tamms
surface states, but what will happen if the internal surface potential is made finite. (See Figure
23) In this section I will discuss a concept very similar to Tamm surface statesintersurface
states.

When identical superlattices are coupled through a thin barrier, the strong coupling
between superlattices causes each surface state to split into two states, i.e. symmetric and
antisymmetric.
24
However, if the two superlattices are different from each other, the interface
states will appear to be almost identical regardless of barrier thickness.

Using Figure 23 as a reference, and solving the Schrodinger equation, we have the
solution

+
I
= R[sin(k
1
z) + l cos(k
1
z)], in region I (26a)

+
II
= C
1
exp(-k
3
z) + C
2
exp(k
3
z), in region II (26b)

+
III
= A[sin(k
1
z) + cos(k
1
z)], in region III (26c)

and

k
1
= 2t(2m
1
*
c)

/h, k
3
= 2t[2m
3
*
(V- c)]

/h, (27)

where expressions for l and are given by Ihm et al. Results from this solution can be seen in
Figure 24. The study of intersurface states is very important when designing heterostructure
devices.
16

Conclusion

The experiment done by Ohno et al. serves to validate Tamms surface state theory;
however, there are some secondary effects of this experimental system which were not discussed
by Ohno et al.

Figure 13 shows the effect of a constant electric field being applied to a superlattice. For
electric field, F, equal to zero, the transitions
5
between electron-heavy hole (T
1
) and electron-
light hole (T
2
) remain constant throughout the superlattice because the particles are allowed to
move to any available state. However, as F increase, the minibands of the superlattice separate
into discrete energy bands, (The collection of these discrete energy levels is called a Stark
ladder.) and the particles begin to localize themselves in a particle well, having decaying wave
functions that may extend for two or three periods.
10


There will be some coupling between Stark
ladder particles and Tamm state particles; however, at very high electric fields, the Stark
particles will localize in the potential wells of the superlattice, and the Tamm particles are
localized at the surface.
22


In the experiment done by Ohno et al. there was no discussion explaining how Stark
particles near the surface were distinguished from the Tamm states. Since this point is not
mentioned in Ohnos paper, I am not completely convinced that the state observed were pure
17
Tamm states. However, H. Ohno et al. do report observing states in which the origin is
unknown.

Because the demand for faster devices is always increasing, physicist are constantly
exploring the option of multi-material devices, so understanding and predicting surface states
and intersurface states will continues to be a critical part of device physics.