4.1. Motivation and Description of the Problem
The systems proposed in this article being new, their electrical and optical properties are totally unknown. They could however be of great practical interest. Indeed, the energy of excitation between two different shells of levels is in the range of , or a few tens of meV at most. This corresponds to photon frequencies in the THz range, where emitters and detectors are not numerous. A possible application as radiation detector could use the modification of the in-plane electrical conductivity brought by an electromagnetic wave. Indeed, the overlap between wave functions on neighboring sites depends on the electron state of excitation which is changed by the incident wave. Another possible process is the ionization of the bound states which could induce a capacitive current in the central wire.
The following sections describe the linear response of the bound electrons to an ac electric field in the case of the two models introduced previously in
Section 3. This electric field is either a uniform field applied to the whole structure or that of a large-wavelength electromagnetic wave. In model 1, with wire radii
and a dc bias voltage of
, the electrons are rather strongly bound to the central wire in all directions and the response is due to quantum-mechanical transitions, either real or virtual. In model 2, with wire radii of
and in the same range of bias voltage, the electrons are almost free to move tangentially around the wire but not in the radial direction. Then, the electron states are similar to those in a quantum well except that the symmetry is almost cylindrical rather than planar. At relatively low frequencies, a mobility
could be associated with the electron tangential motion and computed, e.g., in the framework of a Drude model. This model relates mobility and collision time
through the relation
. Recall that
denotes the electron band mass. Assuming that the mobility is on the same order of magnitude as in the semiconductor bulk and using the values given in [
15] for samples of high quality, we obtain
for GaAs and
for GaN at room temperature. At
, the collision time is at least five times larger. The study of the
à la Drude motion of the electrons lies outside the scope of the present work and we focus on the quantum-mechanical response of the systems under consideration, which is expected to be dominant at frequencies larger than
, i.e., a few THz in our case. The two models discussed in this article, with a bias voltage of
, show inter-shell transitions at frequencies fulfilling this condition.
In the THz range, the radiation electric field is uniform over distances large compared to the cell size so that the interaction with the electrons is limited to the dipolar order. Therefore, the problem of the linear response of the systems studied in this article is similar to that of the polarization of atoms and molecules and the concept of polarizability can be used in our case too. Recall that the polarizability is the ratio of the amplitude of the induced dipole to that of the applied ac electric field. Due to the relatively large cell size compared to that of atoms and to the relatively weak excitation energies, the systems proposed in this article are expected to exhibit large electrical polarizabilities.
The detailed study of the interaction of a whole array with an electromagnetic wave lies beyond the purpose of the present work. Here, our aim is to obtain the linear response of an isolated center, occupied by a single electron. We use a quasi-static approximation to describe the interaction of this electron with the oscillating electric field. Within this approximation, the electric field obeys the laws of electrostatics although it is derived from a time-dependent electric potential. This results in neglecting any retardation effect, which seems justified as the cell dimensions are far shorter than the radiation wavelength in the THz range. The equivalence between this approximation and the usual semiclassical theory of radiation is discussed in the comment of complement
, Section 1c of [
16].
4.2. Effect of the Central Metal Wire on the Interaction with the Electric Field
In the present article, we do not take the electrical screening due to the wires at the square vertices into account. This is consistent with studying an electron bound to an isolated wire. However, we include the effect of the central metallic wire on the electron response function. Indeed, the electron in the state bound around the central wire induces a charge distribution at the surface of the latter which brings a non-negligible contribution to the interaction with the oscillating electric field. The easiest way to take this contribution into account is to use the image-charge method. As a first approximation, we consider that the electron charge is distributed uniformly along the
direction. Then the problem reduces to finding the image of a charge wire parallel to a conducting cylinder or, equivalently, the 2D image of a point charge with respect to a circle at the ground potential. The solution is well known. See, e.g., [
17], Section 3.5. The image has the same absolute value than the source, but the opposite sign. It is located on the same circle radius, inside the circle, and at the distance
from the center. Recall that we use the notation
R for the wire radius and
for the distance to the wire axis. As pointed out above, due to the uniformity of the applied electric field over distances far larger than the cell dimension, the interaction reduces to the dipole contribution. We only consider the case of an electric field in the plane of the semiconductor film. Then, the dipole component of interest is that parallel to this plane so that the electron dipole operator responsible for the interaction with the field is
with
being the in-plane position of the electron, while that associated with the image charge, according to the properties of the latter described above, is
. Therefore, the total electric-dipole operator is
and the Hamiltonian of interaction between the charge in the cell and the external oscillating field in the quasi-static approximation is
In this equation, Equation (
12),
denotes the externally applied electric field, which is taken oriented along the
axis and is written as
while the notation
is used for
Recall that the quasi-static electric field acting upon the cell, , is either an external uniform field applied parallel to the film plane or that of an electromagnetic wave with its polarization vector in this plane.
Notice that the presence of the metal wire leads to a decrease in the total electric dipole of the cell as well as in the strength of interaction. On the contrary, metal nanocomponents of different shapes have been used to increase the efficiency of light emission in semiconductor nanostructure devices. For more detail, see, e.g., [
18,
19,
20] and the references therein. Also, the enhancement of Raman scattering has been known for some time and is frequently used in spectroscopy. See the review by Moskovits [
21]. However, in all these cases, the photon energy is in the eV range, close to the resonance with the surface plasmons of the metal. The frequencies implied in the systems under study in the present article are about 100 times lower, far from the plasmon frequencies in the metal wires. This, together with the difference in the system geometries, explains why, in the present case, the presence of the metal has an opposite effect on the interaction with the electromagnetic radiation.
4.3. Modified Transition Matrix Elements, Oscillator Strengths, and Sum Rule
The results of the previous section show that due to the presence of the central metallic wire, the electric-dipole operator is no longer
as in an isolated atom, but
. Therefore, the probability of radiative transition between the states
i and
j is now governed by the square of the matrix element
The concept of oscillator strength is important in atomic and molecular physics. For the meaning of this concept, see, e.g., Chapter 4, Section 9
4 of [
22] and Chapter XIII, Complement A
XIII, Section 2C of [
16]. It allows the comparison between different radiative transitions of individual electrons. This justifies our interest in computing the oscillator strengths of the systems studied here. Obviously, the transition matrix elements to be used are those given by Equation (
15). This leads to the following modified form of the oscillator strengths
with
defined in Equation (
15). These modified oscillator strengths are dimensionless as the oscillator strengths used in atomic physics. They obey a sum rule slightly different from the Thomas-Reiche-Kuhn one. Indeed, in
Appendix A, we show that the sum
simply reduces to
Most often in atomic physics, the oscillator strengths referred to are those related to the ground-state. We comply with this habit and restrict ourselves to the case of transitions involving the ground-state. We denote by
the oscillator strengths of the transitions from the ground-state to the states with quantum numbers
n and
m,
their cumulative sum up to
n and
m, and
the value given by the sum rule as the limit of
for both
and
. Using the eigenfunctions obtained in
Section 3.1 and Equations (
15) and (
16), we compute these oscillator strengths for the two models chosen before. We also apply Equation (
17) to their ground-state wave functions to obtain the values of
predicted by our modified sum rule. For both models, the applied bias is 1 V. The results of the sum rule are reported in
Table 1, while
Figure 11 shows, in both models, the oscillator strengths for transitions from the ground-state as well as their cumulative sum, versus the excitation energy. The ground-state being totally symmetric with respect to reflections on both the
and
axes, the electric-dipole transitions to the states antisymmetric to
and symmetric to
are the only allowed ones and we do not represent the other transitions.
Figure 11 shows that a few transitions are largely dominant. The dominant transitions occur from the ground-state to states with
in the successive shells
. These states have a node along the
axis. This geometry explains why they produce a large electric dipole along the
axis and, therefore, why the corresponding oscillator strengths are dominant. The optical spectra of the systems studied here should show well-defined peaks at the frequencies of these transitions. We confirm this conclusion in the next section.
4.4. Polarizability and Absorption Spectrum
We now turn to computing the polarizability and the absorption spectrum of an isolated cell with an electron in the ground-state. These are important properties for the purpose of possible electromagnetic applications. Again, we assume that the intensity of the radiation field is weak and its wavelength large compared to the cell dimensions so that we can restrict the computations to the linear dipolar term in the interaction between the electron and the electromagnetic field. Therefore, the quantum-mechanical average, or expectation value, of the electron dipole momentum,
, appears as proportional to the amplitude of the electric field,
. This dipole momentum has two components, an in-phase component,
, and a quadrature component,
, such that
By analogy with the case of atoms and molecules, we call polarizability the property which has
and
as components. A complex electric field
could be used instead of the real one of Equation (
13), leading to a complex expression of the polarizability with
and
as real and imaginary parts, respectively. It is well known that the quadrature, or imaginary, component is in direct connection with the power absorbed from the electromagnetic radiation. Indeed, if
is the change of the system dipole momentum in the time interval
, then the power absorbed by the electron at the time
t is
The quantity of interest here is the power averaged over a period, which we call
. The first term in brackets in Equation (
19) has a time average equal to zero. Therefore, the net absorbed power is due to the existence of a quadrature term in the polarizability,
Similarly, in solid-state physics, the imaginary part of the dielectric constant gives the attenuation of an electromagnetic wave propagating in an absorbing medium. For a detailed description, see, e.g., [
23].
To compute the polarizability, we use the semiclassical theory of radiation and treat its interaction with the electron in the framework of the quasi-static approximation. The time-dependent perturbation theory suits our purpose of obtaining the linear response of the system studied here. See, e.g., Chapter XIII and its complements in [
16] for a description of the perturbation theory applied to the interaction between electrons and the electromagnetic radiation. Recall that we take the ac electric field acting upon the electron oriented along the
axis. The electric-dipole operator, including the contribution of the image charge in the metal wire, is that of Equation (
11). By reason of symmetry, its quantum-mechanical average is also aligned with the
axis so that we only need its
x component which is
. The unperturbed Hamiltonian
, defined in Equation (
2), is the one used in the discussion of the eigenstates in
Section 3.1 and its eigenvalues are denoted by
. Here, for the sake of easiness, we use a single index
i to represent the quantum numbers
n and
m.
The perturbation theory gives for the ground-state wave function restricted to the linear order
with
We have added imaginary terms to the denominators to describe the effects of the finite lifetime of the states reached in the transitions. The parameter
is the inverse of the relaxation time
of the electrons in these states,
. The causes of decay are numerous, for example radiative decay, phonon emissions, collisions with defects, Auger effects, etc. Their importance in the present case is not known. As there are no experimental results to compare with, the chosen value of the relaxation time is not crucial. We suggest using
, which is in the range of the collision times deduced from bulk-mobility measurements. See the discussion in
Section 4.1.
Using Equations (
21) and (
22), we obtain the expectation value of the system electric dipole computed at the linear order in
,
Call
and
,
. Then, the polarizability components can be written as
In case of no damping, i.e., if , the imaginary part of is zero as well as the average absorbed power, as expected.
Introducing a complete set
of eigenfunctions of
, we can write the polarizability components as
in which equations
denotes the matrix element
This allows computation of the polarizability term by term. Recall that the transitions from the ground-state to the states which are not antisymmetric to
and symmetric to
are symmetry forbidden. However, the complete polarizability can be computed at once in the following way. From Equation (
24) it can be deduced that
obeys the following partial differential equation
Separating the real and imaginary parts of
in Equation (
28) leads to a system of two real partial differential equations, namely
This system of equations is solved in COMSOL with boundary conditions appropriate to the symmetry of
and the results are used in Equations (
26) to obtain the polarizability components. The spectrum of absorbed power,
, is easily deduced from the values obtained for
by means of Equation (
20).
First, consider model 1 with
,
, and the bias voltage
.
Figure 12 shows the in-phase and quadrature components of the polarizability obtained as explained above, versus the photon energy of the electromagnetic radiation. Recall that a frequency of
corresponds to about
. The contribution due to the transition to the
state, in the same shell as the initial state, is largely dominant at low photon energy, reaching values above
. This leads us to use two different scales in the graph of
Figure 12, a first one for photon energies below
, a second one for photon energies above
. The resonance next in frequency appears at about
, i.e., in the mid-infrared region of the electromagnetic radiation. Close to this frequency, the polarizability reaches a large value of nearly
. This resonance is due to the transition to the
state. This large value of the polarizability agrees with that of the corresponding oscillator strength,
, shown in the top graph of
Figure 11.
Figure 13 shows the product
which, according to Equation (
20), is a measure, in arbitrary units, of the power absorbed from the electromagnetic field in the cell under consideration. The contribution from the transition to the
state is strongly reduced in the power spectrum, due to the presence of the factor
in the definition of
. The dominant transition now is that to the
state at about
.
The upper graph in
Figure 14 shows both components of the polarizability in the case of model 2 (
,
) while the lower graph shows
, i.e., the absorbed power in arbitrary units in the same model. The resonance occurring at about 52 meV, due to the transition to the state
, gives a contribution largely dominant in the polarizability as well as in the absorbed power.
In both models, the resonance in the polarizability and the main peak in the absorbed-power spectrum due to the transition to the state are relatively sharp and well defined. This makes the systems discussed in the present article good candidates for the development of new IR detectors, notwithstanding the possible difficulties in setting them up practically.
Figure 15 shows the product
, i.e., the absorbed power
vs. photon energy
for four different bias voltages from
to
. The upper and lower graphs are related to models 1 and 2, respectively. The figure shows that if photodetectors based on the present systems could be built, they could be easily made tunable in a range from
to
approximately, in the case of model 1 and from
to
in that of model 2.