Quantum ring
Introduction
Semiconductor Quantum Dots (QDs) [Garcia, 2021] are structures made from the union of two or more semiconductor materials, like 
or 
, and are fabricated so their geometry has dimensions in the order of 100 nanometers or less. Due to the difference in energy gaps between each of the semiconductor materials forming the QD, they restrict the mobility of charge carriers like electrons or holes in all three dimensions and, effectively, confine them inside the QD. They exhibit properties that are normally associated with atoms such as quantized energy levels. For this reason QDs are sometimes called artificially fabricated atoms. The nature of the charge carriers confined inside QDs is described by the wavefunction which is obtained by solving the Schrödinger equation.
Quantum rings (QRs) are a type of QD that were first fabricated by Granados et al. [Granados, 2003] in 2003. Since its appearance QRs have been a focus of many condensed matter physicists. And, they have been proven to be useful in many applications like lasers [Cao, 2005], quantum computing [Yu, 2008], single photon emission [Gallardo, 2010], and there are still other studies that show promising qualities for other applications like optoelectronics [Fomin, 2014].
In many of the studies in the literature the authors consider 2-dimensional QRs like the one by Xie [Xie, 2009], but very few consider the 3D nature of the QRs. And this represents one of the advantages that the finite element method has for solving the Schrödinger equation in QDs with rich 3D geometry like QRs.
In this example we will consider a mono-electronic system confined in a quantum ring that is made by the junction of 
, the material of the ring, and 
, the surroundings. The objective is to find the eigenstates, which can be the starting point for further calculations like optical absorption, thermodynamical properties, among others.
An electron inside the QR will experience a finite step confinement potential, which means that inside of the ring the potential would be 0, and a finite constant potential outside the ring. In standard conditions of pressure and temperature, the potential outside the QR would be of 256.8 meV, when the concentration aluminum is 
[Culchac, 2009].
Depiction of the quantum ring and its surroundings. Here the 
vector represents the position of the electron in the Schrödinger equation.
The behavior of charge carriers inside a crystalline semiconductor, like an electron or a hole, is described by the theory of solid-state physics [Ashcroft, 1976]. In particular the motion of these particles is described by band theory [Zawadzki, 2020]. In this context, a quantity that is analogous to the electron mass is defined. It's called the effective mass and it helps describe how a particle in a periodic potential, such as an electron in a crystal, would respond to external forces, taking into account all the interactions that the particle is summited to inside the crystal. Although, this is a topic with many nuances, see for instance see: [Chang, 2014][Duque-Gomez, 2012].
Generally speaking the effective mass can depend on position and other factors such as temperature and pressure. For this example, as a first approach, we may consider a simplification that is widely used in the literature, which states that for electrons or holes with a small cristal momentum [Zawadzki, 2020], its effective mass can be considered constant. Following the approximation given by Kohn and Sham [Kohn-Sham, 1965], one can assume that every electron will experience the same average potential energy due to the many interactions inside the crystal, such that we can turn a many electron problem into a one-electron Schrödinger equation. Also, we can apply the envelope function approximation [Harrison, 2005], that let us write the time independent Schrödinger equation for the electron's wave function as in equation 1.
The time independent Schrödinger equation for the electron's wave function is given as:
Here 
represents the effective mass of the electron. For 
, 
being the free electron's mass. Also in this case, the potential 
is a piecewise function that defines a potential energy of 
outside the ring, and 
inside it.
Parameters
We define the ring's mean radius 
, width 
, height of the ring 
, boundary radius 
, and the height of the whole region 
, which all represent distances in nanometers:
These geometric parameters for this example are inspired by similar theoretical work considering a square cross-section quantum ring [Hernandez, 2022].
Domain
We will need to load NDSolve's FEM package.
We create one region for the ring and one for the bounding domain.
Then create a region made by the difference of the those two regions above. That way, we are left with a hollow part inside the domain that will define the ring in the mesh later on.
For generating the mesh the "RegionMarker" option is used to set different mesh element sizes in the two different regions, as well as "RegionHoles"->None, such that ring is part of the mesh too.
Hamiltonian and confinement potential
Potential
We define a finite step potential where the potential energy outside the ring is 256.8 meV and 0 inside it. The value of the finite potential is given without units for now, and we will make use of units later when we define the Hamiltonian.
The SetDelayed statement does not evaluate the right hand side, as a consequence, in every call to the function V the ringMemberFunction will need to be evaluated. That makes the process of solving the deferential equation inefficient. So, the Evaluate function is used in the definition of the potential to make the whole process more efficient. Consider reading this section on efficient evaluation of PDE coefficients on for more details.
Hamiltonian
First, we define the effective mass and the reduced planck constant.
Next, we specify the units of the potential with the Quantity function. Since the dimensions of the geometry are in nanometers, we set "ScaleUnits"->{"Meters"->"Nanometers"} so that all units are consistent.
In the next step we use the Schrödinger PDE component to generate all the PDE terms for us.
Boundary conditions
We set one Dirichlet condition that implies that the electron's wave function for the low-lying states that we are interested in must decay to 0 at a distance far from the ring. In this case, that is done by setting the wave function to be 0 at the outer boundary of the domain defined earlier.
Solving the eigenvalue problem
Now, NDEigensystem is used to solve the eigenvalue problem with the Hamiltonian 
and the mesh previously defined. Where the variables en and funs are defined as the energy eigenvalues and the wave functions, respectively.
It takes less than 
minute to find 
eigenstates with a normal laptop.
Visualization
Let's see what the eigenvalues look like.
To make sense of the units of the eigenvalues we need to know that the SchrodingerPDEComponent changes all units internally to base SI units. And, when we add the specification "ScaleUnits"->{"Meters"->"Nanometers"} in the parameters pars all units of meters are converted to nanometers. So we follow the reverse procedure to convert back to millielectronvolts.
From this we see that the second and third eigenvalues are basically numerically identical, and the same can be said about the fourth and fifth value. This makes sense since these are degenerate eigenstates. Degenerate eigenvalues are expected in cases where the system's geometry has symmetry, like the rotational symmetry of the quantum ring. This can be more clear by looking at the corresponding probability densities.
The probability density for the second and third states are basically the same but rotated around the 
axis. As can be seen from the top view. This is because states 
and 
, as well as 
and 
, form a degenerate subspace, meaning that any linear combination of those states would lead to a valid solution.
Also, one important feature is that by making use of NDEigensystem one obtains the wavefunctions already normalized.
On the other hand, is possible to take a closer look to each eigenfunction by doing a contour plot.
Magnetic field interaction
The effect of an external magnetic field on the eigenstates describing particles confined in low dimensional systems, such as QDs and QRs, is something that has been considered by various authors [Jahan, 2018], [Gutiérrez, 2010]. One of the reasons this is an area of interest is that the interaction with the magnetic field can greatly affect the energy spectrum and consequentially the transition energy between the states confined in the nanostructure. Also, a periodic oscillation in the eigenvalues for each state as the magnetic flux density is increased can be observed, a phenomenon known as Aharanov-Bohm (AB) oscillations. Particularly, AB oscillations are present for particles confined in a QR submitted to the presence of a threading magnetic field [Fomin, 2018]. For this reason we will consider an external constant magnetic field pointing in the 
direction.
Usually, the Hamiltonian operator for a particle of mass 
has the following form: 
, where 
is the momentum operator that in quantum mechanics is defined as 
. On the other hand, by following what is known as Peierl's substitution [Luttinger, 1951][Peierls, 1997], when a magnetic field is present the momentum changes from 
to 
. Where 
is the magnetic vector potential, which is defined in such a way that one can obtain the magnetic flux density as 
.
Therefore, the hamiltonian takes the following form:
We can expand the term 
as 
, and therefore, the hamiltonian would take the form in equation 2
Now, applying the definition of 
, replacing 
for the effective mass, and 
for the electron's charge, we can get the PDE operator for the system in question.
One particular characteristic of the magnetic vector potential is that one can add the gradient of a scalar function, for instance 
, to 
as 
, and it would lead us to the same magnetic flux density, since 
. The function 
is called gauge function, and choosing a particular 
and 
is called gauge fixing. This choice can, in many cases, lead to a simplification of the equations involved. One common way to do this is to use Coulomb's gauge fixing condition, that states that 
.
As a justification of this condition, we can assume that we have an original vector potential, 
, which divergence doesn't vanish, i.e, 
. Then we add to it the gradient of a scalar function such that 
. And, this way Coulomb's condition, 
, would lead to the equation 
. This is a Poisson type equation and it has a unique solution. Therefore, 
would be uniquely determined by applying the condition 
. That means, that Coulomb's gauge fixing condition can always be applied without any ambiguity. In summary, since the curl of 
is specified by 
, but the divergence of 
isn't specified, is possible to set the divergence of 
in a way that let us simplify our equations, in this case 
.
Considering the Coulomb's gauge fixing condition the PDE operator can be written as is shown in equation 3:
We will consider a magnetic vector potential that satisfies the condition that 
, where 
is the position vector, which ensures that 
is true. Then, we can use the SchrodingerPDEComponent to generate the new PDE operator.
Note that QuantityMagnitude is used in order to ensure that the SchrodingerPDEComponent can handle correctly the units of 
.
To generate the PDE operator specify the magnetic vector potential and the charge of the particle in the parameters argument of the SchrodingerPDEComponent function. This way two additional terms are added to the PDE.
Note that the Coulomb's gauge for the magnetic vector potential, i.e 
, is assumed as a default by the SchrodingerPDEComponent.
Solving the eigenvalue problem
We can now proceed to solve the eigenvalue problem, with the same mesh and boundary conditions as before.
Visualization
The presence of the magnetic field makes the wave functions complex valued, and its necessary to separate the real a imaginary parts of the resulting interpolating function to plot it using the Abs function.
The obtained eigenvalues are complex numbers, as the PDE is now complex valued in nature.
Upon close inspection one can see that the imaginary part is very small compared with the real part, and one can suspect that this is due to numerical noise given the fact that in quantum mechanics the energy eigenvalues are always a real quantity. For this reason we will consider the energy as being the real part of the eigenvalues obtained with NDEigensystem.
If we compare the real part of the eigenvalues when magnetic field is present with when no magnetic field is present, we find that states 
and 
, as well as 
and 
, are no longer degenerate. Thus, we can say that the presence of the magnetic field removes the degeneracy of the eigenstates we have seen when no magnetic field was present.
Varying the magnetic field
To explore how the external magnetic field affects the energy eigenvalues we can calculate the energy of each state as a function of the magnetic flux density. This can be done by just redefining the PDE operator and solving the eigenvalue problem for each value of the magnetic flux density. For this purpose, a function is defined.
Then Table is used to find the energy for a range of magnetic flux densities.
This is a lengthy calculation and takes about 
minutes to complete with the current mesh in a normal laptop.
The last graph represents the AB oscillations of the energy levels, as is often referred in the literature. One important feature is that the degeneracy in the eigenstates that can be seen when no magnetic field is present at 
is revoked by the presence of the magnetic field. The energy levels then begin to split as the magnetic flux density is increased, a phenomenon know as the Zeeman effect. Furthermore, from the graph we can see that the energy of the lowest state starts to grow until a point near 
where subsequently it starts to fall and then gets back up again at a point near 
. This oscillation is also present for the remaining states. For instance, the energy of the next state starts decreasing until near 
when it starts to increase up to a point close to 
from which it falls back again. And a similar pattern is present for higher states.
In summary, a key feature is that that mere presence of the magnetic field makes the eigenstates no longer degenerate.
On the other hand, something worth noting in the graph is the "anti-crossing" or "avoided crossing" in the energy levels. For instance, take the point close to 
where the eigenvalues for the first two states come really close and seem to "avoid" each other afterwards. This is a common feature of quantum systems when two eigenvalues won't cross when the strength of a perturbation, as is the of the magnetic field, is increased [Cohen-Tannaoudji, 1992].
To explore the effect that the external field has on the wave functions we can plot the probability densities for a magnetic flux density of 
. The choice of that specific value will be clear later.
Taking into account the steps in which we varied the magnetic flux density in the previous calculation we can get the wave functions and eigenvalues for a magnetic flux density of 
.
This last energies are the real part of the eigenvalues provided by NDEigensystem in millielectronvolts.
There are a few things that one can notice here. First, the wave functions for states 
and 
have interchange roles compared with the wave functions previously calculated for a magnetic flux density of 
. And, the same can be said about states 
and 
. Moreover, the wave function for state 
has a morphology that did not appear in the first 
states for the 
case. To explain this, its useful to analyze the AB oscillations graph.
We can think about the AB oscillations pattern as a set of displaced parabolas, each representing a state. This way, the energy for a particular eigenstate at 
, with its correspondent wave function, will evolve following a parabola. In the case of 
, the probability densities morphologies are the same as in the no magnetic field case, which can be understood by looking at the last figure, and noticing that the ordering of the eigenstates is the same for a value of 
as for a 
case. Inversely, for the case with magnetic flux density of 
the parabolas have intertwined, and consequentially the ordering of the eigenstates has changed, which is reflected in the probability densities as described earlier. Furthermore, state 
in the 
case can be traced back, following a parabola, up to the point with a magnetic flux density of 
, and, as can be seen in the previous graph, in that case it would correspond to state 
. That is the reason why in the morphology of the probability density for state 
in the 
case differs from the previously studied cases.
The choice of the specific magnetic flux density of 
, is clearer now. Any value between 
and 
would let us see how the states have intertwined and how the wave functions have interchanged order due to the magnetic field presence, as can be seen in the graph.
Optical absorption
Once the eigenstates are computed, additional properties, such as the optical absorption can be obtained. To calculate the optical absorption we consider the first two eigenstates obtained for a magnetic flux density of 
.
We will consider the presence of a magnetic field to do the optical absorption calculation, because in the no-magnetic-field-case states 
and 
degenerate, which means that the real physical system will be in a linear combination of the two. Therefore, we will take advantage of the fact that the presence of the magnetic field removes the degeneracy of the eigenstates. And in this case a magnetic flux density of 
will be sufficient to have the first two eigenstates being non-degenerate.
One way to compute the optical absorption is to follow a procedure described by Karabulut et al. [Karabulut, 2005], that is based on the density matrix formalism and perturbation theory. The details on how the interaction with light is modeled are in Karabulut's paper [Karabulut, 2005]. We can then apply the expressions for the linear optical absorption 
, the non-linear optical absorption 
and total optical absorption 
, that are give in equations 4, 5, and 6 respectively.
Here 
is the permeability of 
, 
and 
are vacuum and relative electrical permittivity, respectively. The intensity of the incident light beam is 
and 
is its frequency. The 
is the inverse of the decay time between states 1 and 2, due to the interaction between the system and the environment. 
is the carrier density of states of 
, 
is the refractive index and 
is the speed of light. 
are the components of the electric dipole matrix defined as 
= 
where 
is the electron's charge and 
represents the 
coordinate, which is also in the direction in which the incident light beam is polarized. And finally, 
is the difference between the two energy levels considered, 
.
Now we can proceed defining the necessary quantities.
The components of the electric dipole matrix must have units of [
]. But the geometry's and the wave functions's length units are nanometers. So, to have the correct units a factor of 
is introduced in the integral above. Also, note that 
is symmetrical.
We set up the other parameters.
As for the intensity of the light beam and the absorption expressions we will use a similar intensity as the one used in the work by Xie [Xie, 2009].
To plot the absorption a factor of 
is used to have the 
axis be in meV.
In the latter plot we can see that the optical absorption has a peak around the transition energy 
, which means that the system is better able to absorb optical waves with an energy that can promote the ground state to the first excited state. Also, we can see that the absorption has a maximum for a frequency that is in the Terahertz range, and that can be one of the characteristics to consider this type of nanostructure in an optoelectronic application.
Variable effective mass model extension
In the previous approach the effective mass of the electron in the 
quantum ring was used throughout the whole region. Although this procedure is widely used in the literature, a more appropriate way to solve the problem is to take into account the difference in effective mass in the ring region and the outside region. So, we can write what is commonly know as a BenDaniel-Duke Hamiltonian [BenDaniel, 1966], as in equation 7.
Now, the effective mass 
is a function of the position, however, still constant in each sub region. Thus, the effective mass will be a piecewise function.
NDSolve can handle PDE coefficients that are position dependent. The only thing that needs to be done is to make use of the piecewise effective mass in the parameters pars, redefine the magnetic flux density and magnetic vector potential, and generate a new Hamiltonian. The equation can be solved as before with the same geometry and mesh.
By default a hamiltonian with the diffusion term in the BenDaniel-Duke form is generated.
The boundary condition remains the same as before.
It is important to note that the energies have changed and are lower in magnitude than before. To see this more clearly we can plot the energy eigenvalues for each approach.
Let's explore the difference between the probability densities for the model with constant effective mass with the ones in the BenDaniel-Duke approach.
Now, let's explore now what happens to the absorption.
The absorption behaves in a very similar manner as before. And maximum frequency has changed by an amount close to 
. To compare how the absorption has changed we visualize them.
These graphs illustrate that there are not drastic changes in the absorption. Nevertheless, the absorption coefficients' maxima have decreased and a small redshift can be observed in the peak frequency, when the difference in effective masses between regions is considered.
On the other hand, its also interesting to explore the changes in the eigenstates. By looking at the plots of the wavefunctions for each approach it is difficult to see any difference. Therefore we calculate the integral of the product between an specific eigenstate considering constant effective mass and the same eigenstate considering variable effective mass.
The integrals are close to unity, which is expected when one calculates the integral of a given eigenstate squared, which make one suspect that the eigenfunctions didn't change much. Also, from the plot of the difference in the probability density for each approach is clear that the functional form of the probability density stayed the same.
We can conclude that the changes in the energy eigenvalues are appreciable, close to 
of the energy of the ground state. The changes in the position of the absorption peak and the maximum absorption value are minor but noticeable. And, this small change can be explained if we take a look at the eigenfunctions: their value and functional form didn't change much. That means that the electric dipole moment matrix components didn't suffered a huge change, and they are a big part of the absorption expressions we have used.
For this problem, we can say that making use of the BenDaniel-Duke Hamiltonian might be dispensable if one is interested in the optical absorption only, but its addition to the model can be essential for calculating the energies properly.
References
1. García de Arquer, F. P., Talapin, D. v., Klimov, V. I., Arakawa, Y., Bayer, M., Sargent, E. H. (2021). Semiconductor quantum dots: Technological progress and future challenges. Science, 373 (6555). https://doi.org/10.1126/science.aaz8541
2. Granados, D., Garcia, J. M. (2003). Customized nanostructures MBE growth: From quantum dots to quantum rings. Journal of Crystal Growth, 251 (1–4), 213–217. https://doi.org/10.1016/S0022-0248(02)02512-5
3. Cao, H., Deng, H., Ling, H., Liu, C., Smagley, V. A., Caldwell, R. B., Smolyakov, G. A., Gray, A. L., Lester, L. F., Eliseev, P. G., Osiński, M. (2005). Highly unidirectional InAsInGaAsGaAs quantum-dot ring lasers. Applied Physics Letters, 86(20), 1–3. https://doi.org/10.1063/1.1931044
4. Yu, L., Voskoboynikov, O. (2008). Ballistic Aharonov–Bohm quantum bits and quantum gates. Solid State Communications, 145 (9–10), 447–450. https://doi.org/10.1016/J.SSC.2007.12.017
5. Gallardo, E., Martínez, L. J., Nowak, A. K., Sarkar, D., Sanvitto, D., van der Meulen, H. P., Calleja, J. M., Prieto, I., Granados, D., Taboada, A. G., García, J. M., Postigo, P. A. (2010). Single-photon emission by semiconductor quantum rings in a photonic crystal. Journal of the Optical Society of America B, 27(6), A21. https://doi.org/10.1364/JOSAB.27.000A21
6. Fomin, V. M. (2014). Physiscs of Quantum Rings (Vladimir M Fomin, Ed.). Springer Berlin Heidelberg. https://doi.org/10.1007/978-3-642-39197-2
7. Xie, W. (2009). Absorption spectra of a donor impurity in a quantum ring. Physica Status Solidi (b), 246 (6), 1313–1317. https://doi.org/10.1002/pssb.200844349
8. Culchac, F. J., Porras-Montenegro, N. Latgé, A. (2009). Hydrostatic pressure effects on electron states in GaAs–(Ga,Al)As double quantum rings. Journal of Applied Physics, 105 (9), 094324. https://doi.org/10.1063/1.3124643
9. Ashcroft, N. W. and M. N. D. and M. N. D. (1976). Solid State Physics. Holt, Rinehart and Winston.
10. Zawadzki, W. (2020). Wave and Uncertainty Properties of Electrons in Crystalline Solids. Physica Status Solidi (B) Basic Research, 257 (6). https://doi.org/10.1002/pssb.201900517
11. Chang, R., Potnis, S., Ramos, R., Zhuang, C., Hallaji, M., Hayat, A., Duque-Gomez, F., Sipe, J. E., Steinberg, A. M. (2014). Observing the Onset of Effective Mass. Physical Review Letters, 112 (17), 170404. https://doi.org/10.1103/PhysRevLett.112.170404
12. Duque-Gomez, F., Sipe, J. E. (2012). Response of a particle in a one-dimensional lattice to an applied force: Dynamics of the effective mass. Physical Review A, 85 (5), 053412. https://doi.org/10.1103/PhysRevA.85.053412
13. Kohn, W., Sham, L. J. (1965). Self-Consistent Equations Including Exchange and Correlation Effects. Physical Review, 140 (4A), A1133–A1138. https://doi.org/10.1103/PhysRev.140.A1133
14. Harrison, P. (2005). Quantum Wells, Wires and Dots. John Wiley Sons, Ltd. https://doi.org/10.1002/0470010827
15. Hernández, N., López-Doria, R. A., Rivera, I. E., Fulla, M. R. (2022). Refractive index change of a D+2 complex in GaAs/AlxGa1−xAs quantum ring. Journal of Materials Science, 57 (18), 8417–8424. https://doi.org/10.1007/s10853-021-06763-8
16. Jahan, K. L., Boda, A., Shankar, I. v., Raju, Ch. N., Chatterjee, A. (2018). Magnetic field effect on the energy levels of an exciton in a GaAs quantum dot: Application for excitonic lasers. Scientific Reports, 8 (1), 5073. https://doi.org/10.1038/s41598-018-23348-9
17. Gutiérrez, W., García, L. F., Mikhailov, I. D. (2010). Coupled donors in quantum ring in a threading magnetic field. Physica E: Low-Dimensional Systems and Nanostructures, 43 (1), 559–566. https://doi.org/https://doi.org/10.1016/j.physe.2010.09.015
18. Fomin, V. M. (2018). Quantum Ring: A Unique Playground for the Quantum-Mechanical Paradigm. https://doi.org/10.1007/978-3-319-95159-1_1
19. Luttinger, J. M. (1951). The Effect of a Magnetic Field on Electrons in a Periodic Potential. Physical Review, 84 (4), 814–817. https://doi.org/10.1103/PhysRev.84.814
20. Peierls, R. (1997). On the Theory of the Diamagnetism of Conduction Electrons (pp. 97–120). https://doi.org/10.1142/9789812795779_0010
21. Feynman, Richard P., Leighton, Robert B., Sands, Matthew (1965). The Feynman Lectures on Physics. Vol. 3. Addison-Wesley. ISBN 0-201-02115-3. https://www.feynmanlectures.caltech.edu/III_12.html#Ch12-S4
22. Cohen-Tannaoudji,Claude et al.(1992),Quantum Mechanics( Vol. 1), p.409
23. Karabulut, İ., Ünlü, S., Şafak, H. (2005). Calculation of the changes in the absorption and refractive index for intersubband optical transitions in a quantum box. Physica Status Solidi (b), 242(14), 2902–2909. https://doi.org/10.1002/pssb.200541093
24. BenDaniel, D. J., Duke, C. B. (1966). Space-Charge Effects on Electron Tunneling. Physical Review, 152 (2), 683–692. https://doi.org/10.1103/PhysRev.152.683