2. Intrinsic Topological Phononic Structures
Let us consider a system composed of two coupled one-dimensional harmonic crystals as illustrated in
Figure 1.
In absence of external forces, the equations describing the motion of atoms at location
n in the two coupled 1-D harmonic crystals are given by:
here
u and
v represent the displacement in the upper and lower chains, respectively. These displacements can be visualized as being oriented along the chains. The side springs are illustrated for the sake of simplicity as vertical spring but physically they would couple masses between chains along the direction of the displacements.
In the long wavelength limit the discrete Lagrangian is expressed as a continuous second derivative of position. Taking
for the sake of simplicity and mathematical tractability, the equations of motion (1a,b) can be rewritten as:
where
is the 2 × 2 identity matrix,
and
is the displacement vector. We also have defined
and
. Equation (2) takes a form similar to the Klein-Gordon equation. Using an approach paralleling that of Dirac, one can factor Equation (2) into the following form:
In Equation (3), we have introduced the 4 × 4 matrices:
,
, . The symbol i refers to
. The wave functions
and
are 4 vector solutions of
and
, respectively.
and
are non-self-dual solutions. These equations do not satisfy time reversal symmetry (
, T-symmetry, nor parity symmetry (
separately. In the language of Quantum Field Theory,
and
represent “particles” and “anti-particles”. Let us now seek solutions of
in the plane wave form:
with
j = 1,2,3,4. This gives the eigen value problem:
where
We find two dispersion relations:
and
. The first set of dispersion relations corresponds to branches that start at the origin
k = 0 and relates to symmetric eigen modes. The second set of branches represents antisymmetric modes with a cut off frequency at k = 0 of
. Assuming that
and that
, for the symmetric waves characterized by the first set of dispersion relations, then the Equation (4) reduce to
and
which are satisfied by plane waves of arbitrary amplitudes,
and
, propagating in the forward (F) or backward (B) directions, respectively. This is the conventional character of Boson-like phonons. We now look for the eigen vectors that correspond to the second set of dispersion relations. Let us use the positive eigen value as an illustrative example:
. One of the degenerate solutions of the system of four linear Equation (4) is:
where
is some arbitrary constant. Note that the negative signs reflect the antisymmetry of the displacement. Other solutions can be found by considering the complete set of plane wave solutions
with
j = 1,2,3,4 as well as the negative frequency eigen value. The key result is that the second dispersion curve in the band structure is associated with a wave function whose amplitude shows spinorial character (Equation (5)). In this case, the displacement of the two coupled harmonic chains are constrained and the direction of propagation of waves in the two-chain system are not independent of each other. For instance, at
k = 0, the antisymmetric mode is represented by a standing wave which enforces a strict relation between the amplitude of a forward propagating wave and a backward propagating wave. This characteristic was shown [
4,
5] to be representative of Fermion-like behavior of phonons. As
,
, the first two terms in Equation (5) go to zero and only one direction of propagation (backward) is supported by the medium (third and fourth terms in Equation (5)). This example illustrates the difference in topology of elastic waves corresponding to the lower and upper bands in the band structure of the two-chain system. The constraint on the amplitude of waves in the upper band imparts a nonconventional spinorial topology to the eigen modes which does exists for modes in the lower band. The topology of the upper band can be best visualized by taking the limit
. In that case, the system of
Figure 1 becomes a single harmonic chain grounded to a substrate as shown in
Figure 2.
In that limit, the displacement
in Equation (1a,b) is negligible. Equation (2) becomes the Klein-Gordon equation:
with
and
. This equation describes only the displacement field,
u. Equation (3) can be written as the set of Dirac-like equations:
where
and
are the 2 × 2 Pauli matrices:
and
and
is the 2 × 2 identity matrix.
We now write our solutions in the form:
and
where
and
are two by one spinors. Inserting the various forms for these solutions in Equation (6a,b) lead to the same eigen values that we obtained before for the upper band of the two-chain system, namely by
. Again, let us note that the band structure has two branches corresponding to positive frequencies and negative frequencies. Negative frequencies can be visualized as representing waves that propagate in a direction opposite to that of waves with positive frequency. The spinor part of the solutions for the different plane waves is summarized in the
Table 1 below. Negative and positive k correspond to waves propagating in opposite direction.
Table 1 can be used to identify the symmetry properties of
and
in the allowed space:
. We find the following transformation rules:
which lead to the combined transformation:
We have defined
and
as transformations that change the sign of the frequency and wave number, respectively. As one crosses the gap at the origin
k = 0, the multiplicative factor “
i” indicates that the wave function accumulated a phase of
. Also, the Pauli operator
enables the transition from the space of solutions
to the space of
. We also note the orthogonality condition
The topology of the spinorial wave functions that reflects their symmetry properties is illustrated in
Figure 3.
We now shed more light on the properties of the spinorial solutions by treating
in Equation (6a,b) as a perturbation,
. When
, Equation (6a) reduces to the two independent equations:
whose solutions correspond to plane waves propagating in the forward direction,
, with dispersion relation
and the backward direction,
with dispersion relation
. We rewrite Equation (9) in the form:
Seeking the particular solutions to first-order,
and
, which follow in frequency the driving terms on the right-side side of Equation (10a,b), we find their respective amplitude to be:
Since
at
k = 0 only, the amplitude of the first-order perturbed forward wave,
changes sign as
k varies from
to
. A similar but opposite change of sign occurs for the backward perturbed amplitude. These changes of sign are therefore associated with changes in phase of
and −
for the forward and backward waves as one crosses the origin
k = 0. These phase changes (sign changes) are characteristic of that occurring at a resonance. The gap that would occur at
k = 0 in the band structure of the harmonic chain grounded to a substrate via side springs, should we push the perturbation theory to higher orders, may therefore be visualized as resulting from a resonance of the forward waves driven by the backward propagating waves and vice versa. However, since to first-order the amplitudes given by Equation (11a,b) diverge at the only point of intersection between the dispersion relations of the forward and backward waves, we have to use analytic continuation to expand them into the complex plane:
where
continues the eigen values
and
into the complex plane.
By simple inspection, we can see that at the origin,
k = 0, both amplitudes are pure imaginary quantities and therefore exhibit a phase of
. This is expected as Equation (12a,b) are representative of the amplitude of a driven damped harmonic oscillator which also shows a phase of
with respect to the driving frequency at resonance. It is instructive to calculate the Berry connection [
31] for these perturbed amplitudes. The Berry connection determines the phase change of a wave as some parameter takes the wave function along a continuous path on the manifold that supports it. Since the Berry phase applies to continuous paths, we cannot use that concept to determine the phase change across the gap of our system,
i.e., between the positive and negative frequency branches of the band structure. Therefore, we resort to calculating the Berry connection for the first-order perturbed solution which still remains continuous but may capture the interaction between directions of propagation. Our intention is to characterize the topology of the spinorial part of the wave function Equation (12a,b), by calculating the change in phase of the waves as one crosses
k = 0. It is important first to normalize the spinor:
. This normalized spinor takes the form:
. The Berry connection is given by
. After several analytical and algebraic manipulations, we obtain:
In Equation (13), the contribution to the Berry connection of
and
are identical. Using the identities:
and
, leads to
The contribution of each direction of propagation to the spinorial part of the wave function accumulates a phase shift as one crosses the origin k = 0.
3. Extrinsic Phononic Structure
In the previous section, we revealed the spinorial character of the elastic wave function in the two-chain system and one chain coupled to the ground. The purpose of this section is to investigate the behavior of field
when the parameter
(spring stiffness,
KI) is subjected to a spatio-temporal modulation,
i.e.,
where
and
are constants (see
Figure 4). Here,
where
is the period of the modulation.
is the frequency modulation and its sign determines the direction of propagation of the modulation.
The question arises as to the effect of such a modulation on the state of the fermion-like phonons. The periodicity of the modulated one-dimensional medium suggests that we should be seeking solutions of Equation (1) in the form of Bloch waves:
where
. The wave number
is limited to the first Brillouin zone:
and
with
l being an integer. Choosing Equation (6a), we obtain the modulated Dirac-like equation in the Fourier domain:
where
.
Consistent with Quantum Field Theory (QFT) approaches, we solve Equation (15) using perturbation theory and in particular multiple time scale perturbation theory [
32] up to second-order. Nowadays, the multiple-time-scales perturbation theory (MTSPT) for differential equations is a very popular method to approximate solutions of weakly nonlinear differential equations. Several implementations of this method were proposed in various fields of mathematics, mechanics and physics [
33,
34,
35,
36,
37,
38]. Moreover, Khoo
et al. [
39] have shown that the MTSPT is a reliable theoretical tool for studying the lattice dynamics of an anharmonic crystal. More recently, Swinteck
et al. [
40] applied successfully the MTSPT, as described in Reference [
19], for solving propagation equations in a quadratically nonlinear monoatomic chain of infinite extent. Consequently, we used the MTSPT for solving Equation (15).
The parameter
is treated as a perturbation
. The wave function is written as a second-order power series in
, namely:
here
with
are wave functions expressed to zeroth, first and second-order. We have also replaced the single time variable,
, by three variables representing different time scales:
,
, and
. We can subsequently decompose Equation (15) into equations to zeroth, first and second-order in
. The zeroth-order equation consists of the Dirac-like equation in absence of modulation:
As seen in
Section 2, its solutions take the form
with
.
and
represent the orbital and the spinorial parts of the solution, respectively. We have the usual eigen values:
. Inserting the zeroth-order solution into Equation (15) expressed to first-order leads to
The solutions of the first-order Dirac equation are the sum of solutions of the homogeneous equation and particular solutions. The homogenous solution is isomorphic to the zeroth-order solution, it will be corrected in a way similar to the zeroth-order solution as one accounts for higher and higher terms in the perturbation series. Under these conditions and to ensure that there are no secular terms (
i.e., terms that grow with time and that are incompatible with the assumption that
must be a correction to
in the particular solution of Equation (17), the pre-factors of terms like
are forced to be zero [
39]. Subsequently, the derivative of the amplitudes
with respect to
must vanish and these amplitudes only depend on
. The right-hand side of Equation (17) reduces to the second term only. The particular solution of that simplified equation contains frequency shifted terms given by:
The coefficients
are resonant terms:
In the preceding relations, we have defined: and . We have also omitted the time dependencies for the sake of compactness.
To second order, the Dirac-like equation is written as:
The derivative
leads to secular terms. The homogeneous part of the first-order solution does not contribute secular terms but the particular solution does. Combining all secular terms and setting them to zero lead to the conditions:
where we have defined:
We note the asymmetry of these quantities. The terms G, G’ and F diverge within the Brillouin zone of the modulated systems when the condition is satisfied but not when . This asymmetry reflects a breaking of symmetry in wave number space due to the directionality of the modulation.
Equation (19a,b) impose second-order corrections onto the zeroth-order solution. We multiply the relations (18a,b) by
to obtain them in terms of
and subsequently recombine them with the zeroth-order Equation (3). This procedure reconstructs the perturbative series of Equation (15) in terms of
only:
with
and
. To obtain Equation (20), we have also used:
. Equation (20) shows that Equation (15) describing the dynamics of elastic waves in a harmonic chain grounded to a substrate via side springs, whose stiffness is modulated in space and time, is to second-order isomorphic to Dirac equation in Fourier domain for a charged quasiparticle including an electromagnetic field. The quantity
plays the role of the electrostatic potential and
the role of a scalar form of the vector potential. The parentheses
and
are the Fourier transforms of the usual minimal substitution rule.
is the dressed mass of the quasiparticle. The mechanical system provides a mechanism for exchange of energy between the main chain modes and the side springs. The side springs lead to the formation of a fermion-like quasiparticle while their modulation provides a field through which quasiparticles interact. The strength and nature of the interaction is controllable through the independent modulation parameters,
,
and K.
The mechanical system allows for the exploration of a large parameter space of scalar QFT as the functions and can be varied by manipulating the spatio-temporal modulation of the side spring stiffness. We imagine, therefore, that this classical phononic system can be employed to examine the behavior of scalar QFT from weak to strong coupling regimes, as well as, at all intermediate couplings. Further, the capacity to separate the ratio of the effective potentials opens avenues for the experimental realization of scalar fields whose behavior could previously only have been theorized.
We now seek solutions of Equation (20). These are solutions of Equation (18) with spinorial part
satisfying the second-order conditions given by Equation (19a,b). These conditions can be reformulated as
where the vector
and the matrix
. Solutions of this 2 × 2 system of first-order linear equations are easily obtained as:
where
, and
are the eigen values and eigen vectors of the matrix
. The coefficients
C and
C’ are determined by the boundary condition:
. We find the eigen values
and
. The respective eigen vectors are
and
. The coefficients are given by
and
. We note that although the quantities G, G’ and F may diverge the ratio
remains finite. The directed spatio-temporal modulation impacts both the orbital part and the spinor part of the zeroth-order modes. The orbital part of the wave function is frequency shifted to
and
. The quantities
and
represent phase shifts analogous to those associated with the Aharonov-Bohm effect [
41] resulting from electrostatic and vector potentials
and
. Near the resonant condition:
, it is the eigen value
which diverges. This divergence is indicative of the formation of a gap in the dispersion relation [
31]. The eigen value
does not diverge. Since the frequency shift,
, is expected to be small compared to
, the orbital term
. Considering that the lowest frequency
is
, this condition would occur for all k*. Therefore, the term
will essentially contribute to the band structure in a perturbative way similar to that of the uncorrected zeroth-order solution or homogeneous parts of the first or second-order equations. The spinorial part of the zeroth-order solution is also modified through the coupling between the orbital and “spin” part of the wave function as seen in the expressions for
and
. This coupling suggests an approach for the manipulation of the “spin” part of the elastic wave function by exciting the medium using a spatio-temporal modulation. Again, these alterations can be achieved by manipulating independently the magnitude of the modulation,
as well as the spatio-temporal characteristics
and K.
The perturbative approach used here is showing the capacity of a spatio-temporal modulation to control the “spin-orbit” characteristics of elastic modes in a manner analogous to electromagnetic waves enabling the manipulation of the spin state of electrons [
42]. However, the pertubative method is not able to give a complete picture of the effect of the modulation on the entire band structure of the elastic modes. For this, the vibrational properties of the mechanical system are also investigated numerically beyond perturbation theory. We calculate the phonon band structure of the modulated elastic Klein-Gordon equation since its eigen values are identical to those of the modulated Dirac-like equation. We use a one-dimensional chain that contains
N = 2400 masses,
m = 4.361 × 10
−9 kg, with Born-Von Karman boundary conditions. The masses are equally spaced by
h = 0.1 mm. The parameters
= 0.018363 kg·m
2·s
−2 and
= 2295 kg·s
−2. The spatial modulation has a period L = 100 h and an angular frequency
= 1.934 × 10
5 rad/s. We have also chosen the magnitude of the modulation:
. The dynamics of the modulated system is amenable to the method of molecular dynamics (MD). The integration time step is
dt = 1.624 × 10
-9 s. The dynamical trajectories generated by the MD simulation are analyzed within the framework of the Spectral Energy Density (SED) method [
43] for generating the band structure. To ensure adequate sampling of the system’s phase-space the SED calculations are averaged over 4 individual MD simulations, each simulation lasting 2
20 time steps and starting from randomly generated initial conditions. We report in
Figure 5, the calculated band structure of the modulated system.
Figure 5 retains the essential features of the unperturbed band structure but for frequency shifted Bloch modes
and two band gaps in the positive half of the Brillouin zone. The periodicity of the band structure in wave number space is retained but its symmetry is broken. The frequency shifted modes are illustrative of the first-order particular solutions. Second-order frequency shifted modes
do not show in the figure due to their very weak amplitude. The two band gaps occur at the wave vector
defined by the condition
for
g = 0 and
g = K. It is the band folding due to the spatial modulation which enables overlap and hybridization between the frequency-shifted Bloch modes and the original Bloch modes of the lattice without the time dependency of the spatial modulation. The hybridization opens gaps in a band structure that has lost its mirror symmetry about the origin of the Brillouin zone. Considering the first gap and following a path in
k space, starting at
k = 0 at the bottom of the lowest branch, the wave function transitions from a state corresponding to a zeroth-order type wave, with orbital part (
) and spinor part
to a wave having the characteristics of the first-order wave with orbital part (
) and spinor part
. The control of the position of the gap through
and
K enables strategies for tuning the spinorial character of the elastic wave. The effect of these “spin-orbit” manipulations of the elastic system can be measured by examining the transmission of plane waves as shown in [
4].
It is also instructive to consider the symmetry of the Dirac-like equations in the presence of a spatio-temporal modulation to best understand its effect on the spinorial character of the wave function. In the case of a modulation with a general phase
φ, Equation (6a,b) take the overall form:
Applying the joint T-symmetry and parity symmetry to Equation (21a) does not result in Equation (21b) for all phases
φ but a few special values. The modulated Equations (21a,b) have lost the symmetry properties of the unmodulated Dirac equations Equation (6a,b). The gap that formed at
in
Figure 4 is, therefore, not a Dirac point. The transformations
and
do not apply near
. The constraints imposed on the spinorial component of the elastic wave function may be released in the vicinity of that wavenumber. This constraint was associated with Fermion-like wave functions which have the character of quasistanding waves,
i.e., composed of forward and backward waves with a very specific proportion of their respective amplitudes. The release of the Dirac constraint associated with the impossibility for the medium to support forward propagating waves (+
kgap) but only backward propagating waves (−
kgap), may lead again to Boson-like behavior with no restriction on the amplitude of the backward propagating waves.