1. Introduction
The conductivity sum rule is an important tool in the analysis of the dynamical charge transport properties in the strongly correlated systems [
1,
2]. In this paper, we show, by using the simple electron gas model exposed to periodic potential on the atomic scale, how the electron transport concentrations (which are of fundamental importance in experimental data analysis), depend on doping under conditions of the pseudogap formation. Due to the presence of multiple electron-scattering channels and a significant number of bands around Fermi level, it is a challenge to separate intraband from interband contributions in the dynamical conductivity measurements. We demonstrate the sum rule connection with the charge conservation in the two-dimensional (2D) system of nearly free electrons (NFE) in an additional weak periodic potential. We assume that the weak periodic potential is uniaxial, having a single Fourier component with amplitude
and the modulation wave vector
, and that the Fermi energy can be easily changed. A possible onset of such potential in the real system is stabilization of the charge density wave (CDW). Hence, we call this model by abbreviation UniAxNFE. The conservation of charge that participates in electric transport, or the sum rule, amounts to evaluating the contributions from partial spectral weights which originate from the real parts of the intraband and interband conductivity. The general form of the multiband conductivity tensor [
3] is approximated by an expression consisting of a bare single-particle electron-hole energies and a phenomenological electron-hole scattering constant. This approximate form of the multiband conductivity tensor is then divided into its intra- and interband parts, which are then evaluated in the long wavelength limit of a perturbating electric field. In this way, we obtain a simple form of the intraband part of the multiband conductivity tensor, known as the Drude conductivity formula [
4], whereas the contributions to the conductivity originating from the interband excitations, will be referred to as optical conductivity [
5]. In the limit of the vanishing intra- and interband relaxation constant, the expressions for the real parts of intra- and interband conductivity become simplified and, for the particular model of the electron ground state used in this work, they are given in a closed form. The sum rule is defined in Ref. [
6] as well as the way by which it connects the real part of the conductivity and the various types of electron transport concentrations. The latter depends on the direction in which the macroscopic electric field is pointing.
The electron energies of the resulting two-band UniAxNFE model [
4,
7] retain, to some extent, the free electron-like properties in the direction perpendicular to
, whereas in the parallel direction, the bands are strongly influenced by the potential; hence, as a result a pseudogap is opened. The pseudogap position and its width is defined by two of the four characteristic energy points of the electron bands. These points also define the intervals of different energy dependence of the transport concentrations. Finally, we explicitly show how we can identify the intraband and interband effective electron concentration of the UniAxNFE model, which when added together, yield the total charge concentration. Moreover, we demonstrate a different energy dependence of these effective transport concentrations when measured in the direction parallel to the wave vector
, and direction perpendicular to it.
2. Multiband Conductivity Tensor
Here we give a simple overview of the general form of the conductivity tensor in a simple cubic type of crystal without the infrared active phonons. The conductivity tensor includes contributions from all types of single-particle excitation in a multiband system. Within the linear response theory, the conductivity tensor
[
3] is defined as a connection between the Cartesian component
of the induced current density
with respect to the same
component of the macroscopic electric field,
where
are the wave vector and frequency of the macroscopic electric field
. In its general form, as shown in
Figure 1, the conductivity tensor is in fact a current-dipole correlation function. This object is too complicated to be used as a tool for a simple analysis of the dynamical transport properties, so a simplified version has to be introduced. In this simplification of the conductivity tensor [
8], the electron and hole self-energy renormalisation is neglected and electron-hole scattering is approximated by the phenomenological constant
, which depends on the band index
s,
In (
2),
is the
s-band energy with the corresponding Fermi–Dirac distribution
,
V is the volume (area) of the crystal. The first term within the
-summation, which extends over the first Brillouin zone in (
2), is a product of current and dipole matrix elements, which are connected via continuity equation [
10]. This is why the product of current and dipole matrix elements is written by using just the current elements (
) and the single-particle energies. Thus,
is the
matrix element of an
-dependent matrix
, which is derived by using the unitary matrix
needed to diagonalize the electron Hamiltonian and the derivative of the Hamiltonian matrix [
7,
8,
11],
In general, the intraband
current matrix elements are model-independent and given by
whereas the interband (
s ≠
s′) current matrix elements are model-dependent. Taking the
q→0 limit in the Equation (2) and dividing it in the intraband (
s =
s′) and interband (
s ≠
s′) channel and assuming that the intraband phenomenological relaxation constants are equal to γ we get
In Equation (5), we have introduced the effective concentrations of electrons
that participate in the intraband transport [
12]. There are two equivalent forms of
[
13] obtainable from one another by partial integration,
We now turn to the real part of the total conductivity tensor (5) in the limit of a vanishing relaxation constants
. For the intraband conductivity, we obtain
and, correspondingly, for the interband conductivity,
The sum rule states that by integrating the real part of the conductivity (7) and (8), which are even function of
ω, he conserved quantity called the spectral weight is obtained [
6,
12,
14],
where we have defined, analogously to the intraband effective concentration of electrons (6),
an interband effective concentration of electrons
For any multiband system, the sum rule (9) states that the concentration of electrons
engaged in the dynamical charge transport is distributed among electrons participating in the intraband, and those participating in the interband excitations,
Here, we stress that
can be different from the total concentration of electrons
needed to fill the bands up to the Fermi energy. Whether or not
and
are equal, depends on the electron model in question. We will show that for the UniAxNFE model indeed
. This, however, does not apply for the systems with linear electron dispersions like the
d-dimensional Dirac systems [
3,
15,
16] where
. Moreover, evaluating the relation (6a) for the single-band system described by
, we can easily obtain the following general relation:
Thus, we see that only for the parabolic-like electron dispersion (
),
and
might be equal. We now turn our attention to the example of the two-band system in which (
11) is conserved.
3. Two-Dimensional UniAxNFE Model
As a testing ground for the sum rule (9), we analyze the charge transport properties of the two-dimensional, nearly free electron gas in the presence of weak periodic uniaxial potential. The crystal potential is of the form
with the wave vector
, thus becoming the new reciprocal lattice vector. The electronic two-band Hamiltonian in its matrix form in the basis of
and
, where the latter are the states near a single Bragg’s plane defined by
, is
where
where
,
is electron wave vector,
is the bare electron mass [
4]. The diagonalisation of Equation (
13) gives two electron bands labeled by index
within the newly defined Brillouin zone with the periodicity determined by
,
In the electron dispersion Equation (
14),
is defined relatively to
by decomposition
where
is parallel to the
-direction. Moreover, the origin of the newly formed Brillouin zone is shifted by
. Implementing these changes in Equation (
14), we get
The two electron dispersions (
15) are shown in
Figure 2 (left) and are scaled with the
, the electron energy at the Bragg’s plane, where the dispersions cross, prior to the pseudogap opening. In addition, a dimensionless gap parameter
is introduced as a dimensionless order parameter, i.e., the measure of strength of the periodic potential. In the weak periodic potential approach we expect
. The Brillioun zone, over which Equation (
15) is spanned, resembles an infinitely long stripe in
direction of total width
. Four characteristic energy points related to the bands (
15) are identified and designated by the yellow circle in
Figure 2. The bottom (
B) energy of the
band and the top (
T) energy of the
band, within the
cross-section of the Brillouin zone, are located at
(see
Figure 2 (left)). We have
In addition, the extent of the pseudogap region at
is determined by the upper (
U) and lower (
L) energy
Calculated properties related to the charge transport will be presented as functions of the scaled Fermi energy .
4. Charge Transport Concentrations of the 2D UniAxNFE Model
Dispersion (
15) implies different response to the dynamical electric field pointing parallel, or perpendicular to the direction of the vector
. To demonstrate this, we calculate the intraband concentration
(6), for electrons described by the bands (
15), filed up to the scaled Fermi energy
at zero temperature for two distinct directions
.
Let us inspect the ⊥ direction first. It is easy to check from (
15) that
, and thus Equation (6b) reduces to the standard definition of the total concentration of electrons
,
which is shown in
Figure 2 (right) in green as a function of
in units of concentration
.
should be compared to the total concentration of 2D free electrons which is
and is shown as a red dashed line in
Figure 2 (right). We see a small deviation between the two concentrations (
18) and (
19) within the pseudogap region and for high
values. This illustrates a small but observable influence of the uniaxial potential on the transport in the perpendicular direction. Moreover, from the Hamiltonian (
13) and from the definition of the current matrix elements (
4), we can derive [
7] the interband current matrix elements
The only Cartesian component
, for which (
20) is finite, is
. By inspecting Equation (10), we conclude that the interband concentration of electrons is
Therefore, the interband conductivity in the electron model derived in
Section 3 is finite only when the macroscopic electric field is applied in the direction parallel to the vector
. Comparing expressions (
21), (
18) and (
11), we get
If the macroscopic electric field is oriented parallel to
, the calculated
is shown in
Figure 2 (right) in blue. As is seen for
, the two concentrations
and
nearly coincide. The biggest difference between them is when
is within the pseudogap region
. The
is calculated by inserting (
20) into Equation (10). The result is also shown in
Figure 2 (right) in orange. The main feature of
is that it is nearly zero for
and that it has a maximum at
. Furthermore, by inspecting the
Figure 2 (right), we observe that
The above expression can be shown explicitly by adding the terms in (6b) and (10) with (
20) together under the same
.
Another way of showing that a simple two-band model (
15) obeys the sum rule (
23) is by integrating the real part of the optical conductivity. To demonstrate this, we plot the parallel component of the real part of the interband conductivity which for the present model, has been derived in [
7],
In the above expression,
is the scaled photon energy
, and
is the Heaviside unit step function and
is the two-dimensional conductivity constant. Equation (
24) is plotted in
Figure 3 (left) for three values of
whose positions are indicated by correspondingly colored arrows on the right part of
Figure 3. Expression (
24) is finite only for frequencies within the interval
as shown in
Figure 3. These limits originate from the energy conservation (
8) with bands (
15) that are spanned between the endpoints (
16). Moreover, if
, then
(blue curve) is finite, whereas if
, then
develops a one-over a square-root type of divergence (red and dashed green line
Figure 3 (left)). The common feature of the Equation (
24) is the
decay as we increase
. Using
as the new frequency variable, Equation (9) gives
Because of the
-restrictions in the
, the upper limit in the integral is set to
. After finding the area under the conductivity curve numerically, as noted by the shaded region of the blue curve in
Figure 3 (left), we obtain
as shown in the
Figure 2 (right).
As mentioned, the approach where we simply equate
, does not apply for the 2D or 3D Dirac system [
16,
17] where the interband current matrix elements (
20) are constant for any
, and correspondingly,
. In order to apply the sum rule (9) with
, a cutoff frequency
has to be introduced in the upper boundary of the integral in (
25) to prevent
from diverging. This will in turn lead to a self-consistency problem in which the parameters, like the Fermi energy
, will be tied to
.