betaFeedback

WorksPhysics and Astronomy

Energy spectrum of trions in semiconductors

Abstract

Energy levels of trions in a bulk semiconductor crystal are calculated by two different numerical methods. On the one hand, we construct the finite-element solution of the Schrödinger equation in the total angular momentum representation. On the other hand, we use the Gaussian expansion method (GEM), applied to each of the three Faddeev components of the wave function. The results for the total angular momenta J=0J=0 and J=1J=1 of the trion were obtained for a wide range of the hole-to-electron mass ratios covering most existing semiconductor compounds. The dependencies of the energies on the mass ratio for trions in a bulk crystal qualitatively agree with the results for a simplified model of a trion in a 1D quantum wire, for which the three-body energies can be calculated with arbitrary precision as well as easily approximated using Born-Oppenheimer potential-energy curves. A comparison of our results with data from other publicly available three-body codes as well as with the energies of well-known atomic counterparts, in particular, with vibrational energy levels of the molecular hydrogen ion, shows a good agreement and confirms the accuracy of the numerical data. We found that GEM significantly improves the precision of the obtained trion energies compared to previous calculations. Our calculations demonstrate that, for most semiconductor compounds, only one trion state with zero total angular momentum exists.

Introduction

Electron-hole bound states such as excitons and more complex three- and four-body states such as trions and biexcitons are a key focus of semiconductor spectroscopy [1, 2, 3, 4, 5, 6, 7]. These few-body systems in confined geometries — particularly in heterostructures and transition-metal dichalcogenide (TMDC) monolayers [8, 9] — have attracted significant attention due to their potential use in room-temperature excitonic devices [10, 11, 12] and sensors [13, 14]. This is equally promising for the bulk structures that exhibit excitons with large binding energies [15, 16, 17, 18, 19, 20], that are higher than the characteristic energy of lattice thermal vibrations.

Although in conventional bulk semiconductors such as GaAs and CdTe, excitons have relatively small binding energies, their ground-state and excited-state peaks can be rather well observed experimentally [1, 21, 22]. Even smaller binding energies of trions and biexcitons make them more difficult to measure, as their diffuse peaks overlap with much sharper excitonic lines. This is why, despite their long-standing theoretical prediction [23], the first observation of trions (also known as charged excitons) was only possible after the development of growth techniques for heterostructures containing quantum wells (QWIs) [24]. Reducing the dimensionality increases the binding energy of electron-hole complexes [25, 26, 27], thus enabling the corresponding peaks in the spectra to be resolved experimentally [28, 29]. The large binding energies of electron-hole complexes in bulk Cu2O\mathrm{Cu}_2\mathrm{O} and ZnO

crystals make these structures promising candidates for observing trions as well. However, achieving uncompensated electrons and holes in these structures remains challenging [30].

Trions in bulk semiconductors are the simplest solid-state three-body systems. For bulk materials, many-body effects, giving rise to trions, can be successfully explained in terms of quasiparticles (or dressed particles) [31]. Therefore, a convenient framework of effective few-body Hamiltonians, allowing for an application of variational methods, is widely used [27]. However, for example, in TMDC monolayers trions appear at much higher densities of charge carriers. For such systems, the properties of trions can only be described from a proper many-body point of view, in particular within the Fermi polaron approach [20, 32, 33]. In this context, it is somewhat surprising that the spectra of trions, at least for the simplest bulk model of two bands with parabolic dispersions, have not been calculated so far. The main reason is probably that the three-body problem with the Coulomb interaction allows only a numerical solution [34]. An analytical solution for the ground state of a three-body system is only known for contact (zero-range) interactions in one dimension [35]. Short-range and long-range interactions between the particles, as well as additional spatial degrees of freedom make the problem analytically intractable. In QWIs, the situation is even more complicated: one-dimensional confinement breaks the translational invariance and introduces quantization of the center-of-mass motion. Even studying excitons in QWIs requires a computationally complex numerical solution to the three-dimensional Schrödinger equation [26, 29, 36, 37]. Furthermore, due to strong spin-orbit coupling, an accurate theoretical description of exciton spectra in bulk crystals requires one to go beyond the simple parabolic band approximation, i.e. to take into account the complex valence band structure [18, 38, 39, 40, 41, 42, 43]. A simplification can be found for very narrow QWls or quantum wires, for which the center-of-mass quantization is so strong that the relative motion of the charge carriers can be completely separated. Therefore, even within the simplest two-band model, the trion in a QWl is a multidimensional problem that can only be solved by calculating the ground state characteristics using the variational approach with predefined wave functions [44, 45, 46, 47].

In the present article, we study the energy spectra of X+X^{+} trions, consisting of an electron and two holes. The simplest two-band model with parabolic dispersions of the conduction and valence bands is assumed. We obtain spectra of trions in both the bulk crystal and the one-dimensional geometry (cylindrical quantum wire) and map out their unified dependencies on the effective mass ratio. Our study also covers the mass range of the X−X^{-} trion model, which is mathematically identical to the X+X^{+} model by a formal interchange of electrons and holes. Both X+X^{+} and X−X^{-} trion ground states in bulk materials were extensively studied theoretically [46, 48, 49], whereas the existence of excited states is still debated.

The one-dimensional system allows us to accurately determine the structure of energy levels of bound states, understand their origins and, therefore, assess the implications for the general case without spatial confinement. Similar one-dimensional three-body models have already been studied before, see Refs. [50, 51, 52, 53]. These works, however, disregarded the interaction between heavy particles, considered only zero-range interactions with the light particle, and mainly presented ground-state energies. In our article, we consider the long-range interactions in all three two-body subsystems and calculate not only the ground state but many energy levels of excited states. Thus, in contrast to Refs. [54, 55], we accurately reconstruct the energy spectrum of the one-dimensional three-body system and observe how it depends on the mass ratio. The exact energy levels are compared with the Born-Oppenheimer [56] approximation as well.

The rough estimates show that the ground-state binding energy of the three-body system with two identical particles, in particular, a trion, weakly depends on the mass ratio of two nonidentical constituents: it can be approximated by a constant value for a wide range of mass ratios [57]. However, a more precise systematic study by Combesçot [48] clearly demonstrates nonlinear deviations from a single value. It considered the trion in a bulk crystal within a two-band model with parabolic dispersions of the conduction and valence bands. The ground state energy has been calculated as a point of the kernel singularity of the momentum-space integral equation for the scattering amplitude of an electron-exciton collision. A dependence of the ground-state energy on the mass ratio has been obtained. Attention was paid to the mass ratios corresponding to the negatively charged hydrogen ion H−\mathrm{H}^{-} [58] and the positively charged hydrogen molecular ion H2+\mathrm{H}_{2}^{+} [59, 60], which is a well-studied atomic system. In particular, the lowest electronic terms of H2+\mathrm{H}_{2}^{+}, 1sσ1s\sigma and 2pσ2p\sigma, in different approximations including the infinite nuclear mass and neglecting the nuclear repulsion were calculated by many authors [61, 62, 63, 64, 65]. Analytical expressions [66] for the electronic potential curves, including the well-known Morse potential [67, 68], for different diatomic molecular ions were proposed to facilitate calculation of the vibrational levels [69]. Moreover, several high-precision variational calculations of the H2+\mathrm{H}_{2}^{+} energy levels have been reported [70, 71, 72]. The ion H2−\mathrm{H}_{2}^{-}, having only one bound state due to a strong Coulomb repulsion of the constituent electrons, has also drawn considerable attention [58, 73]. However, an evolution of the three-body spectrum during a crossover between the limiting cases of negatively charged hydrogen ion and the positively charged hydrogen molecular ion has not been studied so far.

In the present article, we extend the results of Ref. [48] by presenting the full spectrum of the bound-state energies and analyzing their dependence on the hole-to-electron mass ratio mh/mem_{h}/m_{e}. To this end, the three-body Schrödinger equation for the total angular momentum values J=0J=0 and J=1J=1 is numerically solved by two different variational methods. The first one is the finite-element solution of the Schrödinger equation in the total angular momentum representation [74, 75, 76, 77]. The second method is the Gaussian expansion method, based on the expansion of three Faddeev components of the wave function over a basis of Gaussian functions and bipolar harmonics [78, 79, 80]. The obtained dependencies of energy levels on the mass ratio are described and the vibrational levels are classified. A formation of three-body excited states, i.e. several lowest vibrational modes, for large mass ratios is shown. For the mass ratio corresponding to H2+\mathrm{H}_{2}^{+}, 20 calculated vibrational levels for J=0J=0 are compared with the high-precision data from Refs. [70, 71]. We also provide data that allow one to approximate the six lowest vibrational energies for all reasonable values of the mass ratios, including the limiting case of an infinite hole mass. These can be used to tune the parameters of the hole-hole (’interatomic’) potential to significantly facilitate simulation of the three-body system.

In contrast to Refs. [81, 82], where only the ss-wave model in all subsystems was considered, we take into account the higher partial waves and thus obtain more precise results. Our calculations improve precision of the data from Refs. [46, 48, 49] and demonstrate that, for most semiconductor compounds, only one trion state with zero total angular momentum exists. A single trion state exists if the mass ratio of the two different particles, i.e. mh/mem_{h}/m_{e} in case of X+X^{+} trion and me/mhm_{e}/m_{h} for X−X^{-}, is below 9.7. For J=1J=1, both methods show no trion bound states for mass ratios below 4.2.

It is worth noting that our numerical solvers are not unique. There are also variational [44, 45, 46, 71, 83] and finite-difference [84, 85] three-body calculations, approaches using hyperspherical harmonics [27, 57, 81, 86, 87, 88], Laguerre polynomial [89, 90, 91] and numerical splinebasis [92, 93, 94] expansion methods. However, almost all mentioned works considered mainly the minimum energy quantum state and paid little attention to the excited states, i.e., they did not study the entire spectrum of the three-body systems. Nonetheless, some of these algorithms have been released in publicly available three-body codes [91, 95, 96]. In particular, we employed the open-source SWEXPH [96] and ThreeBSCbSSR [91] programs in our study for an additional comparison of the precision of the ground-state calculations.

In our work, we calculate energies of the bound states by different methods and compare the respective results. As our computations were performed for a broad range of effective mass ratios, they cover many well-known semiconductor compounds, for example GaAs, Cu2_2O, and CdTe. Therefore, the obtained results can be interesting not only from a purely theoretical point of view, but also for practical purposes. From our calculations one can obtain more precise values of trion binding energies by scaling the numerical results for particular semiconductor compounds by their corresponding exciton Hartree energies.

The model problem of a trion in a quantum wire

In this section, we consider a model problem of a trion in a cylindrical quantum wire, which confines the motion of the charge carriers to only one dimension. This simple model allows for machine-precision solutions for various types of interparticle interactions. Its ground state has already been extensively studied in Refs. [50, 51, 52, 53, 55, 97]. Here, we calculate the energy spectrum of bound states and observe its dependence on the masses of the constituent particles. Our results underline the key implications, which are crucial for understanding the behavior of the bound-state energies for a general three-body problem, namely for a trion in a bulk crystal. For a more detailed justification of the model, including its applicability and relevance to fabricated semiconductor quantum wires, we refer the reader to Refs. [54, 55].

Formulation of the model problem

Let us consider the X+^+ trion in a one-dimensional system. For the sake of simplicity, we assume parabolic dispersions of the valence and conduction bands with effective masses m1=m2=mhm_1=m_2=m_h, m3=mem_3=m_e. We also assume strong confinement in the directions perpendicular to the quantum wire. Therefore, the motion described by these coordinates can be separated. This leads to a simple problem of a one-dimensional motion of the electron and two holes over the quantum wire direction,

denoted in the following by zz. This also makes connection to atomic physics, where a similar model of a one-dimensional three-body system was extensively studied [50, 51, 52, 53, 97]. We introduce the following dimensionless coordinates: distance is measured in units of the two-body (exciton) Bohr radius aBa_B, and energy is measured in units of the exciton Rydberg energy. The latter is Rymh=μe4/2ε2ℏ2Ry_{m_h}=\mu e^4/2\varepsilon^2\hbar^2, where μ=memh/(me+mh)\mu=m_em_h/(m_e+m_h) is the reduced mass of the electron-hole subsystem, ee is the elementary charge and ℏ\hbar is Planck’s constant. The dielectric constant ε\varepsilon is a common scale parameter which is taken to be 1. After eliminating the center-of-mass motion of the three-body system, we arrive at the Schrödinger equation for the relative motion of the electron and the holes in the quantum wire. The model Hamiltonian reads [52, 53]

−2+σ2(1+σ)∂2∂y2−2σ1+σ∂2∂x2+2μℏ2[V1(y−x/2)+V2(y+x/2)+V3(x)],(1)-\frac{2+\sigma}{2(1+\sigma)}\frac{\partial^2}{\partial y^2}-\frac{2\sigma}{1+\sigma}\frac{\partial^2}{\partial x^2}+\frac{2\mu}{\hbar^2}\left[V_1\left(y-x/2\right)+V_2\left(y+x/2\right)+V_3(x)\right], \tag*{(1)}

where we denote σ=me/mh\sigma=m_e/m_h. The Jacobi coordinates xx and yy are the distances between the two holes and between the center-of-mass of the holes (1, 2) and the electron, respectively [98].

To simulate the electron-hole interactions in a quantum wire, we use the model attractive potential [54]

V1,2R(z)=−ℏ22μ2z2+R2(2)V_{1,2}^{R}(z)=-\frac{\hbar^2}{2\mu}\frac{2}{\sqrt{z^2+R^2}} \tag*{(2)}

which, in contrast to previous studies [51, 52] of contact interactions, is long-range. It is obtained by averaging the Coulomb interaction over the radial functions, perpendicular to the wire direction [99]. In general, the parameter RR is related to the effective radius of the quantum wire, but it effectively removes the singularity of the Coulomb potential in one dimension. The hole-hole repulsion is taken into account by a potential of the same form V3R(x)=−V3R(x)V_3^{R}(x)=-V_3^{R}(x). The profile of the resulting total potential in Eq. (1) is shown in Figure 1. One can see the valleys of the potential minima along the lines y±x/2=0y \pm x/2=0 as well as two parts of the ridge of the potential maxima along the line x=0x=0 corresponding to the repulsive interaction of the two holes [35].

Profile of the three-body potential between particles in the one-dimensional system

Figure 1. The profile of the three-body potential between particles in the one-dimensional system, i.e. the quantum wire. The energy is measured in units of the exciton Rydberg energy and the coordinates are in units of aBa_B. The parameter of the two-body potential is chosen as R=0.366 aBR=0.366\,a_B.

The Schrödinger equation with the Hamiltonian (1) is complemented by appropriate boundary conditions. In particular, the trivial (or zero) boundary conditions as x→±∞x \to\pm\infty and y→±∞y \to\pm\infty are employed for calculating the bound states. As the potential (2) forms a finite QWl for any finite value of RR, our model equation has, at least, one bound state solution [100].

Methods of solution

The one-dimensional model (1) allows for a numerically exact solution, which serves as a blueprint for understanding the energy spectrum of the general three-body problem. Moreover, it can be used for constructing the Born-Oppenheimer (adiabatic) potential [101, 102] of electron eigenenergies for fixed positions of the holes. This then allows one to solve the one-dimensional equation for the relative motion of the holes in the constructed potential, thereby obtaining the three-body bound states within the Born-Oppenheimer theory.

The Born-Oppenheimer approximation is valid when the kinetic energy of the heavy particles, in our case the holes, is much smaller than the energy of the heavy-light subsystem [56]. This exactly happens in Eq. (1), as mh≫mem_h \gg m_e. As σ=me/mh→0\sigma=m_e/m_h \to0, the kinetic energy of the relative motion of the heavy particles is 2σ2\sigma times smaller than that for the heavy-light subsystem. In this case, the energy difference between the Born-Oppenheimer potential-energy curves EBOi(x)E_{\mathrm{BO}}^i(x), which are obtained from the equation

(−2+σ2(1+σ)∂2∂y2+2μℏ2[V1(y−x/2)+V2(y+x/2)+V3(x)])ψi(y∣x)=EBOi(x)ψi(y∣x)(3)\left(-\frac{2+\sigma}{2(1+\sigma)}\frac{\partial^2}{\partial y^2}+\frac{2\mu}{\hbar^2}\left[V_1(y-x/2)+V_2(y+x/2)+V_3(x)\right]\right)\psi^i(y|x)=E_{\mathrm{BO}}^i(x)\psi^i(y|x) \tag*{(3)}

with xx taken as a parameter, is much larger than between the eigenenergies of the heavy-particle motion. Then, the Born-Oppenheimer approximation of the three-body energies is obtained from a solution of the one-dimensional equation over coordinate xx with the particular Born-Oppenheimer electron eigenenergy curve EBOi(x)E_{\mathrm{BO}}^i(x) taken as a potential,

(−2σ1+σ∂2∂x2+EBOi(x))ϕji(x)=Ejiϕji(x).\left(-\frac{2\sigma}{1+\sigma}\frac{\partial^2}{\partial x^2}+E_{\mathrm{BO}}^i(x)\right)\phi_j^i(x)=E_j^i\phi_j^i(x).

The opposite limit, mh≪mem_h \ll m_e, i.e. σ=me/mh→∞\sigma=m_e/m_h \to\infty, can be formally treated in a similar way. However, the kinetic energy of the heavy particle (in this case, the electron) will be just 4 times smaller. Therefore, the Born-Oppenheimer approach in this case is expected to be only approximate.

For arbitrary mass ratio σ\sigma, the Schrödinger equation with the Hamiltonian (1) and potentials of the form (2) can be solved numerically with machine precision. It can be done using the standard finite-difference method [26, 29, 103], more advanced pseudo-spectral method [52] or an expansion of the wave function over an appropriate basis of polynomials, for example, splines [37, 92, 104, 105, 106], or finite-element functions [76, 107]. In our work, we employed an expansion of the three-body wave function over a basis of cubic Hermite splines [93, 94, 108], which, due to the collocation at the Gaussian points [109], gives a fourth-order accuracy in the step hh of equidistant grid, i.e. an uncertainty ∼O(h4)\sim O(h^4). This made it possible to precisely calculate several lowest energies of the three-body bound states and to observe their dependence on the mass ratio. The size of the calculation domain and the number of collocation points in the numerical algorithm were selected in order to achieve the desired level of precision within the considered range of mass ratios.

Results and implications

First of all, we separate the calculated bound state energies from resonances. The upper energy limit of three-body bound states is determined by the two-body ground state of the one-dimensional potential (2). This energy of an electron-hole pair (exciton) in a quantum wire is the lowest scattering threshold and defines the lower boundary of the three-body continuum. As we are free to take any value of the parameter RR from a reasonable range, we have chosen it to be R=0.366 aBR=0.366\,a_B in order to obtain the two-body ground state energy of almost exactly −3-3 Ry and thus to simplify the analysis of the results. For a detailed justification of the choice of the parameter RR, see Refs. [54, 55].

We calculated the energies of the X+X^+ trion in a quantum wire and observed their dependence on the effective mass ratio mh/mem_h/m_e. The obtained energies in the units of the exciton Rydberg are shown in Figure 2(a). The green curves are the numerically exact results, whereas the dashed and solid red curves show the Born-Oppenheimer approximations. One can see that in both cases the number of bound states increases with growing mass ratio. For mh/me≲0.7m_h/m_e \lesssim0.7 there is only one bound state. In this sense, the system is similar to the negatively charged hydrogen ion, H−\mathrm{H}^{-}. The rigorous proof of the fact that H−\mathrm{H}^{-} has only one bound state has been given in Ref. [58].

Figure 2. (a) Dependence of the energies of the trion states in a quantum wire on the effective mass ratio mh/me=σ−1m_h/m_e = \sigma^{-1}. The green curves show numerically exact results, the red ones are the Born-Oppenheimer (BO) approximations: the dashed curve is a qualitative estimate of the sole energy of the system with two light holes and a heavy electron (similar to H−^-, but with a formal interchange of the electron and holes), whereas the solid ones approach energies of the system with two positively charged heavy constituents (H+^+-like). The three-body scattering threshold at −3-3 Ry is denoted by the black dashed line. (b) The lowest (solid curve) and upper (dashed curve) potential-energy curves, originating from the Born-Oppenheimer-like treatment of the trion in a one-dimensional system, as a function of the distance between center-of-mass of the holes and the electron for the mass ratio mh/me=0.01m_h/m_e = 0.01. The blue curve, i.e. the set of eigenenergies of the light holes for the fixed positions between the electron and the center-of-mass of the holes, represents a single narrow QW1. It is used as an input for Born-Oppenheimer calculation of the sole three-body energy, denoted by dashed red line. (c) The lowest (solid curve) and upper (dashed curve) potential-energy curves, originating from the Born-Oppenheimer-like treatment of the trion in a one-dimensional system, as a function of the distance between two holes for the mass ratio mh/me=100m_h/m_e = 100. The blue curve, i.e. the set of eigenenergies of the electron for the fixed distances between the holes, is used as an input for calculation of the three-body energies, i.e. hole-hole vibrational states, denoted by the red color. The potential curve, effectively, represents two weakly-coupled QW1s, thus the three-body energy levels are given by pairs, with a splitting of levels within each pair.

In our case, the Born-Oppenheimer approximation can be used to qualitatively explain the emergence of the only bound state for small values of the mass ratio, although its quantitative predictive power is still quite moderate for mh/me≪1m_h/m_e \ll1. The point is that the mass-dependent factors in the kinetic terms in Eq. (1), are of the same order: one factor, i.e. for the kinetic term of the light-particle relative motion, is only 4 times larger than the other. Nevertheless, it means that the motion of the light particles (in this case, holes) generally produces larger energies and it can be formally considered as a faster subsystem. Thus, one can construct the Born-Oppenheimer potential-energy curves of the hole-hole motion by taking yy as a parameter and averaging over xx, in a similar way as described in Sec. IIB. The analysis shows that the lowest Born-Oppenheimer potential curve obtained in this way forms a one-dimensional single QW1. Therefore, it can produce a finite number of three-body bound states [100, 110]. However, this QW1 is quite narrow and, in fact, accommodates only one energy level, see Figure 2(b). The QW1 width is related to a repulsion of the charge carriers represented, in turn, by the magnitude of the potential profile ridge in Figure 1. When averaging over xx, this ridge contributes to the barriers of the Born-Oppenheimer potential. As RR decreases, this ridge becomes higher and sharper. This causes the QW1 to be deeper and narrower. In fact, the thickness of the QW1 is less related to the variation of RR, rather than the depth. Therefore, as R→0R \to0 the QW1 becomes much deeper, while its attraction decreases not as dramatically. As a result, not only for a given value of R=0.366 aBR = 0.366\,a_B, but for all smaller ones, leading asymptotically to the Coulomb potential, the system cannot accommodate more than one bound state. For mh/me≲1m_h/m_e \lesssim1, the Born-Oppenheimer approximation gives only a qualitative description, namely the correct number of bound states, but the obtained energy values noticeably differ from the numerically exact ones. The coefficients in the kinetic terms in Eq. (1) only slightly vary for mh/me≲0.1m_h/m_e \lesssim0.1. Therefore, for such a range of mass ratios the energies in Figure 2(a) almost do not change and look like horizontal lines.

For mh/me≳0.7m_h/m_e \gtrsim0.7, there are at least two bound states. As mh/mem_h/m_e grows further, the number of bound states increases: the energy levels gradually drop down from the continuum into the discrete part of the spectrum. The three-body energy levels are given by pairs, with a fine splitting of levels within each pair for mh≫mem_h \gg m_e. These energy levels are analogues of the vibrational energies of the diatomic molecule [59]. It is natural that the Born-Oppenheimer approximation should be valid for such mass ratios as described in Sec. IIB. The green and red curves overlap better as mass ratio grows. Interestingly, for mh/me≳1m_h/m_e \gtrsim1 the Born-Oppenheimer approximation overestimates the binding energies of the two lowest energies by 4 % and 6 %. A similar overestimation was reported for zero-range potentials in Ref. [52]. It is related to the fact that the Born-Oppenheimer approach only gives a lower bound to the energy [58].

The splitting of the three-body energy levels into pairs can be understood from the plot of Born-Oppenheimer electron potential-energy curves in Figure 2(c). The electron potential curves are obtained from Eq. (3), i.e. by averaging over yy. Due to such averaging of the potential profile, the lowest potential curve EBO1E_{BO}^{1} has two QWIs separated by a narrow barrier, associated with the hole repulsion. If the separating barrier were infinite, the wave functions in each QWI would be independent and thus the energy levels would be degenerate. As the barrier is finite, the wave functions originating from each QWI are coupled and the three-body energy levels are split. This becomes more pronounced for a mass ratio mh/me∼1m_h/m_e \sim1, for which the ground state wave functions for each QWI are most extended. For larger mass ratios, the ground state wave functions become more squeezed and their overlap is decreasing. The excited states have different nodal regions of opposite signs of the wave function. Therefore, their coupling and level splitting is much smaller. Analysis shows that the first excited energy curve EBO2E_{BO}^{2} has two very shallow minima which can produce very weakly bound states, unable to be resolved with a reasonable precision.

Implications. The main result of this model study is that the increase in the mass ratio mh/me=σm_h/m_e = \sigma leads to an increase in the number of bound states. The limit mh/me→0m_h/m_e \to0 exhibits only one bound state, as predicted for the analogous atomic H−^- system. With the growth of mh/mem_h/m_e, more bound states appear. This is related to the decrease ∼σ\sim\sigma of the kinetic energy of relative hole-hole motion as mh/me→∞m_h/m_e \to\infty. As a result, more three-body bound states originating from the hole-hole vibrations appear in the spectrum. Their energies are simultaneously decreasing, gradually filling the wells of the Born-Oppenheimer potential-energy curves. In the limit of mh/me=σ−1=∞m_h/m_e = \sigma^{-1} = \infty the second kinetic term in Eq. (1) vanishes, meaning that the quantization disappears and the density of three-body states forms a continuum above the minimum of the energy curve. A similar behavior should be observed in the general case of a trion in a bulk semiconductor crystal.

Trion in a bulk crystal

In this section, we consider a three-body Coulomb system in three dimensions. A particular example of this in solid-state physics is the trion in a bulk semiconductor crystal. In atomic physics, the trion’s counterparts are the negatively charged hydrogen ion H−^- [58] and the positively charged hydrogen molecular ion H2+_2^+ [59, 60]. We further describe two distinct methods which are used to solve the three-body problem and obtain the energies of the trion. Although both methods are variational, they are based on completely different expansions of the three-body wave function. Thus, the outcomes of these methods complement one another and allow a more detailed comparison and verification of the results.

Formulation of the general three-body problem

We consider the Hamiltonian of three spinless particles (m1,m2,m3)(m_1,m_2,m_3) that interact via pairwise potentials. It is assumed that each of the potentials depends only on the radius vector between two particles. Due to translational invariance, the center-of-mass motion of such a three-body system can be eliminated. The Hamiltonian of the relative motion of these particles in the three-dimensional space is conventionally written in Jacobi coordinates [98] as

−ℏ22μ12,3Δr12,3−ℏ22μ12Δr12+V1(r12,3−m1m1+m2r12)+V2(r12,3+m2m1+m2r12)+V3(r12).(4)-\frac{\hbar^2}{2\mu_{12,3}}\Delta_{r_{12,3}}-\frac{\hbar^2}{2\mu_{12}}\Delta_{r_{12}}+V_1\left(r_{12,3}-\frac{m_1}{m_1+m_2}r_{12}\right)+V_2\left(r_{12,3}+\frac{m_2}{m_1+m_2}r_{12}\right)+V_3\left(r_{12}\right). \tag*{(4)}

Here the reduced masses are given by the masses of the particles: 1/μ12=1/m1+1/m21/\mu_{12}=1/m_1+1/m_2 and 1/μ12,3=1/(m1+m2)+1/m31/\mu_{12,3}=1/(m_1+m_2)+1/m_3. The terms ViV_i define interactions between particles of adjacent pairs (j,k)(j,k), where indices {i,j,k}\{i,j,k\} are cyclic permutations of {1,2,3}\{1,2,3\}. The vector r12r_{12} specifies the relative motion of the particles in the pair (1,2)(1,2) and the vector r12,3r_{12,3} defines the motion of the particle 3 with respect to the center-of-mass of the pair (1,2)(1,2).

The Schrödinger equation with the Hamiltonian (4) has to be complemented by appropriate boundary conditions: vanishing wave functions are used for calculation of the bound state solutions, whereas scattering boundary conditions can be used for constructing the quantum scattering problem [98]. Note, that an alternative formulation of the scattering problem in terms of the Faddeev components of the wave function significantly simplifies its formulation and numerical solution in configuration space [111]. As we will show below, the decomposition of the wave function into the Faddeev components also serves to efficiently solve the bound state problem.

Method of solution: Total angular momentum representation

In this section, we describe our first approach to calculate energies of the trion states. It is based on the total angular momentum representation of the three-body problem and described in detail in Refs. [75, 76]. By expanding the three-body wave function over the Wigner D-functions, it allows a reduction of the initial six-dimensional Hamiltonian (4) to a system of coupled three-dimensional equations for the part of the wave function describing the relative motion of the particles within the three-body plane. These equations can be further effectively solved using the finite-element method.

Wigner D-function expansion

The quantum three-body system is characterized by a total angular momentum JJ and its projection MM. The essence is that the three-body wave function ΨMJ(R)\Psi^{J}_{M}(\mathbf{R}) can be represented in terms of Wigner D-functions [112, 113] as

ΨMJ(r12,r12,3)=∑M′=0J12+2δM′0(DM,M′J(α,β,γ)+(−1)M′DM,−M′J(α,β,γ))ψM′J(x,y,θ).(5)\Psi^{J}_{M}(\mathbf{r}_{12},\mathbf{r}_{12,3})=\sum_{M'=0}^{J}\frac{1}{\sqrt{2+2\delta_{M'0}}}\left(D^{J}_{M,M'}(\alpha,\beta,\gamma)+(-1)^{M'}D^{J}_{M,-M'}(\alpha,\beta,\gamma)\right)\psi^{J}_{M'}(x,y,\theta). \tag*{(5)}

Here (r12,r12,3)(\mathbf{r}_{12},\mathbf{r}_{12,3}) is a pair of Jacobi vectors. The angles (α,β,γ)(\alpha,\beta,\gamma) are the Euler angles defining the three-body plane, and (x,y,θ)(x,y,\theta) are the coordinates of the particles in this plane. As in § 2 A, the distance between particles 1 and 2 is denoted as x=∣r12∣x=|\mathbf{r}_{12}|, whereas the distance between particle 3 and the center-of-mass pair (1, 2) is denoted by y=∣r12,3∣y=|\mathbf{r}_{12,3}|. The angle θ\theta is between r12\mathbf{r}_{12} and r12,3\mathbf{r}_{12,3}.

Using the expansion (5) in the three-body Schrödinger equation with the Hamiltonian (4) leads to the system of equations

∑M′=0J(HMM′J−EδMM′)ψM′J(x,y,θ)=0,M=0,…,J(6)\sum_{M'=0}^{J}\left(H^{J}_{MM'}-E\delta_{MM'}\right)\psi^{J}_{M'}(x,y,\theta)=0,\qquad M=0,\ldots,J \tag*{(6)}

for the radial part ψMJ(x,y,θ)\psi^{J}_{M}(x,y,\theta) of the wave function, describing the relative motion of the particles in the three-body plane. The solution of this system gives the components of the wave functions for all projections MM of a predefined total angular momentum JJ. In Eq. (6), the diagonal matrix elements of the tridiagonal Hamiltonian matrix are

HMMJ=−ℏ22μ121x∂2∂x2x−ℏ22μ12,3[1y∂2∂y2y−J(J+1)−2M2y2]−ℏ22(1μ12x2+1μ12,3y2)(∂2∂θ2+cot⁡θ∂∂θ−M2sin⁡2θ)+V(x,y,θ).H^{J}_{MM}=-\frac{\hbar^{2}}{2\mu_{12}}\frac{1}{x}\frac{\partial^{2}}{\partial x^{2}}x-\frac{\hbar^{2}}{2\mu_{12,3}}\left[\frac{1}{y}\frac{\partial^{2}}{\partial y^{2}}y-\frac{J(J+1)-2M^{2}}{y^{2}}\right]-\frac{\hbar^{2}}{2}\left(\frac{1}{\mu_{12}x^{2}}+\frac{1}{\mu_{12,3}y^{2}}\right)\left(\frac{\partial^{2}}{\partial\theta^{2}}+\cot\theta\frac{\partial}{\partial\theta}-\frac{M^{2}}{\sin^{2}\theta}\right)+V(x,y,\theta).

By defining λ±(J,M)=J(J+1)−M(M±1)\lambda_{\pm}(J,M)=\sqrt{J(J+1)-M(M\pm1)}, the upper- and lower-diagonal matrix elements read as

HM,M±1J=−iℏ1+δM,0δM±1,1+δM,1δM±1,0λ±(J,M)2μ12,3y2(∂∂θ+(1±M)cot⁡θ).H^{J}_{M,M\pm1}=-i\hbar\sqrt{1+\delta_{M,0}\delta_{M\pm1,1}+\delta_{M,1}\delta_{M\pm1,0}}\frac{\lambda_{\pm}(J,M)}{2\mu_{12,3}y^{2}}\left(\frac{\partial}{\partial\theta}+(1\pm M)\cot\theta\right).

Note that ψ−1J(x,y,θ)≡0\psi^{J}_{-1}(x,y,\theta)\equiv0.

The potential V(x,y,θ)V(x,y,\theta) defines the interactions, i.e. the nature of the three-body system. For the X+^{+} trion under consideration, it is given by a sum of three pairwise Coulomb potentials:

V(x,y,θ)=e2ε(−1∣r23∣−1∣r13∣+1∣r12∣).V(x,y,\theta)=\frac{e^{2}}{\varepsilon}\left(-\frac{1}{|\mathbf{r}_{23}|}-\frac{1}{|\mathbf{r}_{13}|}+\frac{1}{|\mathbf{r}_{12}|}\right).

where ee is the elementary charge and ε\varepsilon is the dielectric constant taken to be 1. Due to the equal effective masses of the two holes, m1=m2=mhm_{1}=m_{2}=m_{h}, the distances are explicitly given as ∣r13∣2=x2/4+y2−xycos⁡θ|\mathbf{r}_{13}|^{2}=x^{2}/4+y^{2}-xy\cos\theta and ∣r23∣2=x2/4+y2+xycos⁡θ|\mathbf{r}_{23}|^{2}=x^{2}/4+y^{2}+xy\cos\theta.

Numerical solution

The numerical solution of Eq. (6) for a predefined total angular momentum JJ has been performed using the finite-element three-body solver, described in detail in Refs. [74, 75, 76]. Here we only briefly sketch the basic ideas.

The main advantage of the total angular momentum representation is that the structure of Eq. (6) implies coupling only three components of the wave function: those with M′=MM'=M and M′=M±1M'=M\pm1. This leads to a block-tridiagonal form of the Hamiltonian matrix, which significantly reduces the numerical complexity [114]. The computational domain over coordinates (x,y,cos⁡θ)(x,y,\cos\theta) consists of a rectangular cuboid, divided into a finite number of elements (FEM). The component ψMJ\psi^{J}_{M} of the total wave function is expanded into a sum of local basis functions as

ψMJ(x,y,cos⁡θ)=∑i,mcimJMfimJM(x,y,cos⁡θ).(7)\psi^{J}_{M}(x,y,\cos\theta)=\sum_{i,m}c^{JM}_{im}f^{JM}_{im}(x,y,\cos\theta). \tag*{(7)}

The terms cimJMc^{JM}_{im} are expansion coefficients defined at the particular element ii, whereas the index mm specifies the local basis function at this element. The basis functions over the angular variable are the associated Legendre polynomials. For the coordinate variables, the Legendre polynomials are multiplied by damping exponentials properly chosen to increase the convergence of the bound state solutions. Using the expansion (7) in the Schrödinger equation (6) leads to the generalized eigenvalue problem of the form Hc=ESc\mathcal{H}\mathbf{c}=E\mathcal{S}\mathbf{c}, where the vector c\mathbf{c} consists of the expansion coefficients.

As already mentioned, for an arbitrary total angular momentum JJ, the Hamiltonian matrix H\mathcal{H} is block-tridiagonal. Moreover, as each block of the Hamiltonian matrix H\mathcal{H} and the overlap matrix S\mathcal{S} are sparse, an efficient numerical solution of the problem can be obtained using the iterative Arnoldi algorithm, realized in ARPACK [115]. The linear system at each iteration of the algorithm is solved using the PARDISO library [116], implemented in Intel MKL [117].

Method of solution: Gaussian Expansion Method

In this section we describe our second approach, the Gaussian Expansion Method (GEM), which is used as an alternative for calculating the trion bound-state energies. In this approach, the three-body wave function is expanded in a large, non-orthogonal basis of Gaussians which allows a very efficient computation of the bound states. This variational method is described in detail in Ref. [78] and we use its implementation in the open-source package FewBodyToolkit.jl [80] developed by one of the authors.

Jacobi vectors and Faddeev components

The relative motion of three particles is described using two spatial vectors, in particular Jacobi vectors defined by corresponding coordinates. For a three-body system, there are three alternative ways of choosing these vectors, see Figure 3. Let us, for definiteness, consider the first pair of vectors (x1,y1)(\boldsymbol{x}_1,\boldsymbol{y}_1). The three-body wave function ΨMJ(x1,y1)\Psi^J_M(\boldsymbol{x}_1,\boldsymbol{y}_1) can be represented by a sum [78, 98]

Three sets of Jacobi vectors

Figure 3. Three sets of Jacobi vectors (xj,yj)(\boldsymbol{x}_j,\boldsymbol{y}_j), j=1,2,3j=1,2,3.

ΨMJ(x1,y1)=∑j=13FM(j)J(xj,yj),(8)\Psi^J_M(\boldsymbol{x}_1,\boldsymbol{y}_1)=\sum_{j=1}^{3}\mathcal{F}^{(j)J}_M(\boldsymbol{x}_j,\boldsymbol{y}_j), \tag*{(8)}

of three so-called Faddeev components FM(j)J\mathcal{F}^{(j)J}_M, each of which is expressed in terms of a corresponding pair of Jacobi coordinates (xj,yj)(\boldsymbol{x}_j,\boldsymbol{y}_j). Such a decomposition in all three sets simultaneously allows to describe each pairwise Coulomb interaction in its most suited coordinate set. This is important not only for scattering problems in order to properly treat the complicated boundary conditions of the three-body collisions but, as we will show below, also for calculation of the bound states. We note that only the three-body bound-state wave function is normalized to unity, not each of the Faddeev components individually.

In Eq. (8), the Jacobi vectors representing the second and the third Faddeev components are linear combinations of vectors (x1,y1)(\boldsymbol{x}_1,\boldsymbol{y}_1) [98]. The fact that the sets of Jacobi coordinates are not independent, provides a key advantage of the current approach. The transformation between the Jacobi sets affects the representation of the Faddeev component (and, more generally, the wave function): for instance, a component given by pure ss waves in one Jacobi set is not given by the ss waves in the others. Instead, it includes contributions from all partial waves. The representation (8) using all Faddeev components and employing only ss-wave contributions (l=L=0)(l=L=0) in each pair of the coordinates (xj,yj)(\boldsymbol{x}_j,\boldsymbol{y}_j), j=1,2,3j=1,2,3, therefore already implicitly contains a large amount of angular correlations in the total wave function. Thus, this three-set representation is usually good enough for approximation of the wave function even for a relatively small maximal number of partial waves lmax⁡l_{\max} and Lmax⁡L_{\max} per set. The main difficulty of this representation originates from the transformation between the Jacobi coordinate sets. As it will be shown in Sect. III C 2, using Gaussian basis functions allows one to overcome this issue and keep calculations efficient.

For the particular case of the trion, the representation (8) is simplified due to the identity of two out of three particles. The symmetry of the Hamiltonian under exchange of the two identical particles leads to an equivalence of two Faddeev components. Hence, it is sufficient to describe the system by only two components in (8): one where the pair consists of two identical particles, and the other where the pair is built from two different particles. Effectively, this reduces the Schrödinger equation, from a 3×33\times3 matrix-type eigenvalue equation to a 2×22\times2 one.

Gaussian basis functions

To solve the problem, each Faddeev component (j=1,2,3)(j=1,2,3) in its proper Jacobi coordinates is expanded into a linear combination [78]

FM(j)J(xj,yj)=∑αCα(j)∑ml,ML⟨lmlLML∣JM⟩×ϕnlml(xj)ΦNLML(yj),(9)\mathcal{F}^{(j)J}_M(\boldsymbol{x}_j,\boldsymbol{y}_j)=\sum_{\alpha}C^{(j)}_{\alpha}\sum_{m_l,M_L}\langle l m_l L M_L\mid J M\rangle\times\phi_{n l m_l}(\boldsymbol{x}_j)\Phi_{N L M_L}(\boldsymbol{y}_j), \tag*{(9)}

of basis functions, which are chosen to be the products of Gaussians and the solid harmonics [112]:

ϕnlml(x)=Nnle−νnx2xlYlml(x^),(10)\phi_{n l m_l}(\boldsymbol{x})=\mathcal{N}_{n l}e^{-\nu_n x^2}x^lY_{l m_l}(\hat{\boldsymbol{x}}), \tag*{(10)}
ΦNLML(y)=NNLe−λNy2yLYLML(y^).(11)\Phi_{N L M_L}(\boldsymbol{y})=\mathcal{N}_{N L}e^{-\lambda_N y^2}y^LY_{L M_L}(\hat{\boldsymbol{y}}). \tag*{(11)}

The multi-index α\alpha runs over {n,l,N,L}\{n,l,N,L\} describing the angular momenta ll, LL and the Gaussian-range indices nn, NN for the coordinates x\boldsymbol{x} and y\boldsymbol{y}, respectively. The Clebsch-Gordan-coefficients ⟨lmlLML∣JM⟩\langle l m_l L M_L\mid J M\rangle couple ll and LL to a fixed JJ and MM. This angular expansion is equivalent to bipolar harmonics [96, 112]. The coefficients Nnl\mathcal{N}_{n l}, NNL\mathcal{N}_{N L} are normalization factors.

The parameters νn\nu_n and λN\lambda_N define the widths of the Gaussian functions. Choosing them in a geometric progression

νn=1/xn2,xn=x1×(xnmax⁡x1)n−1nmax⁡−1,\nu_n = 1/x_n^2,\qquad x_n = x_1 \times\left(\frac{x_{n_{\max}}}{x_1}\right)^{\frac{n-1}{n_{\max}-1}},
λN=1/yN2,yN=y1×(yNmax⁡y1)N−1Nmax⁡−1,\lambda_N = 1/y_N^2,\qquad y_N = y_1 \times\left(\frac{y_{N_{\max}}}{y_1}\right)^{\frac{N-1}{N_{\max}-1}},

allows an accurate description of each Faddeev component and the entire three-body wave function both at the short-range and the long-range asymptotic domain, while reducing the variational problem to only three parameters per coordinate (nmax⁡n_{\max}, x1x_1, xnmax⁡x_{n_{\max}}, and Nmax⁡N_{\max}, y1y_1, yNmax⁡y_{N_{\max}}, respectively).

To facilitate an efficient approximation of the intermediate-range behavior of the highly-excited, strongly oscillatory states, complex-ranged Gaussians are employed. They are obtained by the substitution [78]

νn→νn×(1±iω),λN→λN×(1±iω),(12)\nu_n \to\nu_n \times(1 \pm i\omega),\qquad\lambda_N \to\lambda_N \times(1 \pm i\omega), \tag*{(12)}

in Eqs. (10) and (11). To ensure that the total basis can still represent a real wave function, basis functions associated with both signs are used. The resulting functions are damped oscillations which reproduce the nodal structure of excited states more accurately than the original real Gaussians. Since Eq. (12) keeps the Gaussian functional form unchanged, the same analytical expressions for matrix elements can be applied.

A fast computation of matrix elements for higher partial waves requires an additional modification of the basis functions. Since each Faddeev component is expanded in a basis defined in its proper coordinate set, evaluating the matrix elements involves transforming the basis between the different Jacobi sets. For ss waves this is straightforward, as the product of Gaussians in (xj,yj)(x_j,y_j) remains Gaussian, ∼exp⁡(−vTMv)\sim\exp(-\mathbf{v}^{T}M\mathbf{v}), with v=(xk,yk)\mathbf{v}=(\mathbf{x}_k,\mathbf{y}_k). The non-diagonal 2×22 \times2 matrix MM contains the information about the widths of the original Gaussians and the masses of the particles.

For higher partial waves l,L>0l,L > 0, the solid harmonics, i.e. the vector-argument functions xlYlml(x^)x^lY_{lm_l}(\hat{\mathbf{x}}) and yLYLML(y^)y^LY_{LM_L}(\hat{\mathbf{y}}), in Eqs. (10) and (11) make the transformation within their arguments laborious. To avoid this, the infinitesimally shifted Gaussian lobe (ISGL) functions [78] are used. The main idea is to represent the solid harmonics by finite superpositions of Gaussians displaced by infinitesimally shifted vectors. This allows one to perform the transformation between the Jacobi coordinate sets in these shifted Gaussians, that is much more straightforward than for the solid harmonics. The infinitesimal limit to zero can then be taken analytically and, since every integral is then of purely Gaussian type, enables finding expressions for the matrix elements in closed form for arbitrary ll and LL. In general, this depends on the choice of potential. For the Coulomb potential considered here and several other potentials, closed-form expressions for matrix elements have been obtained [118].

The angular momenta for the two relative coordinates are coupled to the total angular momentum as J=l+L\mathbf{J}=\mathbf{l}+\mathbf{L}, which restricts possible (l,L)(l,L) channels via the triangle condition ∣l−L∣≤J≤l+L|l-L|\leq J\leq l+L and the parity (−1)l+L=±1(-1)^{l+L}=\pm1. This is the case for each of the three Jacobi sets. Then, for J=0J=0, the same angular basis for xj\mathbf{x}_j and yj\mathbf{y}_j, j=1,2,3j=1,2,3, can be used, l=Ll=L. For J=1J=1, the sum l+Ll+L is restricted to be odd and hence ∣l−L∣=1|l-L|=1. For the component whose pair has two identical particles, the exchange symmetry constrains ll further: to even values for J=0J=0, and to odd values for J=1J=1.

Numerical solution

As the Gaussian basis functions are non-orthogonal, the variational principle leads to a generalized eigenvalue problem Hc=ESc\mathcal{H}c=ESc with a dense overlap matrix SS. A geometric range grid, dense enough to converge the energy, can contain nearly linearly-dependent functions, which make SS ill-conditioned. Its smallest eigenvalues reach the level of the working precision and may even become slightly negative due to rounding errors. We therefore first diagonalize SS and discard all directions whose eigenvalues fall below a relative threshold of 10−1410^{-14} of the largest one by projecting H\mathcal{H} onto the surviving subspace prior to the final diagonalization. For the basis sizes used in our calculations, this removes between 4% and 21% of the nominal dimensions while keeping the converged energies unchanged.

The numerical parameters used in our calculations are summarized in Table I. The angular basis is truncated at l≤lmax⁡l\leq l_{\max} and L≤Lmax⁡L\leq L_{\max}. The corresponding matrix sizes are Nbasis=4nmax⁡Nmax⁡nlLN_{\mathrm{basis}}=4n_{\max}N_{\max}n_{lL}. Here the factor of 4 is due to the complex-ranged Gaussians of both signs over each radial coordinate, and nlLn_{lL} is the number of (l,L)(l,L) channels taken into account, summed, for the trion, over the two Faddeev components. For example, there are 2 channels, (0,0)(0,0) and (2,2)(2,2), for the component whose pair has two identical particles and additional 3 channels, (0,0)(0,0), (1,1)(1,1), and (2,2)(2,2), for another component in case of J=0J=0 and lmax⁡=Lmax⁡=2l_{\max}=L_{\max}=2. The complex-ranged Gaussians were obtained by the parameter ω=0.9\omega=0.9 in Eq. (12).

Configurationnmax⁡n_{\max}Nmax⁡N_{\max}lmax⁡l_{\max}x1x_1xnmax⁡x_{n_{\max}}y1y_1yNmax⁡y_{N_{\max}}NbasisN_{\mathrm{basis}}
J=0J=0, r<0.5r<0.5302220.003100.51513 200
J=0J=0, r>0.5r>0.5302220.00550.51513 200
J=0J=0, H2+\mathrm{H}_2^+362620.00250.51518 720
J=0J=0, H−\mathrm{H}^-362620.002100.51518 720
J=1J=1262030.0150.51518 720

Table I. Numerical parameters used in the GEM calculation, with lmax⁡=Lmax⁡l_{\max}=L_{\max} and lengths in units of aBa_B.

The numerical algorithm is implemented in the open-source package FewBodyToolkit.jl [80], which takes advantage of the linear algebra functions in the Julia programming language largely calling the original functions from LAPACK [119].

Results

First, we will define the units of energy in which we will present the numerical results. In atomic physics, energies are measured in atomic units, a.u. or Hartree units, which are equal to 2Ry∞2\mathrm{Ry}_{\infty}, with the Rydberg energy Ry∞=mee42ε2ℏ2\mathrm{Ry}_{\infty} = \frac{m_{\mathrm{e}}e^4}{2\varepsilon^2\hbar^2}, given that the proton in the hydrogen model has infinite mass. For consistency with Refs. [48, 49], we introduce excitonic units, e.u. or excitonic Hartree units. They are defined by twice the reduced (or exciton) Rydberg energy, RyM\mathrm{Ry}_{M}, taking into account the finite mass MM of the heavy particle (hole) as RyM=μe42ε2ℏ2\mathrm{Ry}_{M} = \frac{\mu e^4}{2\varepsilon^2\hbar^2}, where μ=meM(me+M)\mu= \frac{m_{\mathrm{e}}M}{(m_{\mathrm{e}} + M)}. This gives us the following relation to recalculate the Rydberg (Hartree or a.u.) into the reduced (or exciton) Rydberg (excitonic Hartree or e.u.)

RyM=(Mme+M)Ry∞.\mathrm{Ry}_{M} = \left(\frac{M}{m_{\mathrm{e}} + M}\right)\mathrm{Ry}_{\infty}.

For atomic systems, the difference between the Rydberg and the reduced Rydberg energy is relatively small, though taking into account the finite mass of the proton shifts the hydrogen ground state energy from −0.5-0.5 a.u. to ≈−0.4997278\approx-0.4997278 a.u. In e.u. or the excitonic Hartree units, this energy will be independent of the masses, exactly −0.5-0.5 e.u. To recalculate the energy in Rydberg (Hartree or a.u.) into the energy in the reduced (or exciton) Rydberg (excitonic Hartree or e.u.) one can use the formula

ERy∞=(Mme+M)ERyM.\frac{E}{\mathrm{Ry}_{\infty}} = \left(\frac{M}{m_{\mathrm{e}} + M}\right)\frac{E}{\mathrm{Ry}_{M}}.

Zero total angular momentum J=0J = 0

Using FEM and GEM solvers described in § 5 B and § 5 C, respectively, we computed the eigenenergies of the trion for different effective masses of the constituent charge carriers. As an additional benchmark, we compared our results with the outcomes of three alternative approaches. A comparison of the ground-state energies of the X+X^{+}-trion obtained by all five methods for various mass ratios r=mh(me+mh)r = \frac{m_{\mathrm{h}}}{(m_{\mathrm{e}} + m_{\mathrm{h}})} is given in Table II. The data are obtained for a total angular momentum J=0J = 0 and are given in excitonic Hartree units (e.u.) without rounding off. The second and third columns give results of our FEM and GEM solvers, respectively. The fourth column shows results of calculations using the ThreeBSCbSSR program [91] which implements the numerical method based upon an expansion of the wave function over the Laguerre polynomials in perimetric coordinates. The fifth column contains results using the SWEXPH program [96] which, in turn, realizes a high-precision variational approach with a predefined form of the ground-state wave function. The last column shows results of the momentum-space calculations from Ref. [48]. For a particular method, the number of converged decimal digits differs for different mass ratios. Consequently, there are different numbers of decimal digits in table cells of columns 2–5. To simplify comparison, in a particular cell the energy value is given with either four decimal digits or the maximal number of common digits with another cell in a row increased by two.

The FEM solver allowed us to achieve an accuracy comparable to that of recent calculations, in particular, the ground-state energy data by Combescot [48]. We found that the convergence of this method directly relates to the number of basis functions, the Legendre polynomials, over the angular variable θ\theta in Eq. (6). The number of finite elements (60 over each spatial variable) as well as the number of basis functions over the radial coordinates (damped Legendre polynomials up to the seventh order) were fixed high enough and, thus, less affected the precision. The presented in column 2 results were obtained from the extrapolation of the values calculated for a fixed number of Legendre polynomials over the angular variable, to an infinite number of polynomials. For each mass ratio, we calculated and fitted the series with up to l=30l = 30 Legendre polynomials over θ\theta. This gave us an accuracy of less than 1.3×10−41.3 \times10^{-4} e.u. for the worst fit for r∼0.4r \sim0.4 across all studied range of mass ratios. By approaching the limiting cases r→0r \to0 and r→1r \to1, the accuracy gradually improves by, at least, one order of magnitude. Additionally, for the case of the positively charged molecular hydrogen ion H2+\mathrm{H}_{2}^{+} we used up to l=40l = 40 polynomials to improve a precision of the lowest trion eigenenergy by, at least, two orders of magnitude, see the corresponding line in Table II.

The results of the GEM solver appeared to be more precise than those of the FEM solver. The GEM energies are variational upper bounds with an accuracy better than 10−710^{-7} e.u. for all studied mass ratios and order 10−910^{-9} e.u. for H−\mathrm{H}^{-} and H2+\mathrm{H}_{2}^{+}, see Table II. Therefore, tracking of the results as a function of the basis size and further extrapolation was not required. The results were obtained with the basis parameters listed in Table I. For the ground state calculations, the convergence is most sensitive to the number of partial waves taken into account, i.e. values of lmax⁡l_{\max} and Lmax⁡L_{\max}, and to the short-range parameter of the grid, x1x_{1}. For the given Nmax⁡N_{\max} and nmax⁡n_{\max}, using only ss waves already produces decent, with an accuracy of less than 10−510^{-5} e.u., results for the ground state and a few lowest excited states of H2+\mathrm{H}_{2}^{+}. With lmax⁡=Lmax⁡=2l_{\max} = L_{\max} = 2 the precision of the lowest energy values is improved by almost three orders of magnitude.

r=mh/(me+mh)r = m_h/(m_e + m_h)Our FEM solver (e.u.)Our GEM solver (e.u.)ThreeBSCbSSR [91] (e.u.)SWEXPH [96] (e.u.)Combescot [48] (e.u.)
0-0.5277509-0.5277510-0.5282
0.0001-0.52774760-0.5277477274-0.527747729157-0.527747729164
H−^{-} 0.0005443206-0.52773300-0.5277331360-0.527733137107-0.527733137114
0.001-0.52771802-0.5277181948-0.527718196518-0.527718196525
0.01-0.5274252-0.5274281626-0.527428163600-0.527428163607
0.05-0.526234-0.5262523345-0.52625233932-0.52625233948-0.5267
0.10-0.52496-0.525029358-0.525029370-0.525030-0.5256
0.15-0.524014-0.524063618-0.524063638-0.524078-0.5247
0.20-0.52326-0.523341609-0.523341636-0.52341-0.5240
0.25-0.522846-0.522854649-0.522854683-0.5231-0.5235
0.30-0.522565-0.5225989-0.5225990-0.52289-0.5233
0.35-0.5225748-0.5225759-0.5225760-0.5231-0.5233
0.40-0.52281-0.522792835-0.522792880-0.5241-0.5235
0.45-0.523227-0.523263504-0.523263546-0.5253-0.5239
0.50-0.52399-0.524010130-0.524010140-0.5256-0.5246
0.55-0.525059-0.525065354-0.525065362-0.5241-0.5256
0.60-0.5264787-0.526476186-0.526476193-0.52618-0.5270
0.65-0.5283181-0.5283099-0.5283100-0.52824-0.5287
0.70-0.530675-0.530664943-0.530664955-0.530615-0.5310
0.75-0.533699-0.533689379-0.533689385-0.533623-0.5340
0.80-0.5376184-0.5376202929-0.5376202985-0.5376115-0.5378
0.85-0.5428769-0.542871317-0.542871323-0.54259-0.5430
0.90-0.5502789-0.5502762613-0.5502762668-0.55046-0.5503
0.95-0.5620578-0.562056105-0.562056110-0.56226-0.5621
0.96-0.5654825-0.56548099-0.56548100-0.56566
0.97-0.5695923-0.5695918508-0.5695918556-0.56974-0.5696
0.98-0.574789406-0.574789496-0.57478950-0.57488-0.5748
0.99-0.5821145-0.5821151234-0.5821151274-0.582153-0.5821
0.999-0.5956839-0.595685027-0.59568503133-0.59568503143
H2+_2^{+} 0.9994556794-0.597464260-0.597464273720-0.59746427506-0.59746427521a^{a}
0.9999-0.600369-0.6003883-0.6003851-0.5959
1-0.602628-0.5971-0.602633765b^{b}

Table 2. Comparison of the X+X^{+} trion ground-state energies (total angular momentum J=0J = 0) for different effective mass ratios r=mh/(me+mh)r = m_h/(m_e + m_h). The energies are given in excitonic Hartree units (e.u.). The exciton energy in these units is exactly −0.5-0.5 e.u. for all mass ratios. There are data obtained by our FEM and GEM solvers (second and third columns), the results of using the ThreeBSCbSSR program [91] (fourth column) and the SWEXPH program [96] (fifth column) as well as outputs of the momentum space calculations from Ref. [48] (last column). Each value is given with either four decimal digits or a maximal number of common digits with another cell in a row plus two digits. The precision of our FEM calculation for arbitrary rr is about 10−410^{-4} e.u., whereas for rr corresponding to H−^{-} and H2+_2^{+}, it is below 10−510^{-5} e.u. The precision of our GEM calculation is of about 10−710^{-7} e.u., whereas for rr corresponding to H−^{-} and H2+_2^{+}, it is of order 10−910^{-9} e.u. The results for r=0r = 0 and r=1r = 1 are calculated for rr values shifted by 10−1210^{-12}.

The method of expansion over the Laguerre polynomials in perimetric coordinates realized in the ThreeBSCbSSR program [91] with 28 basis functions was employed to produce the ground state energy with a very good precision of eight digits for intermediate values of rr and minimum five decimal digits for very limiting case of r=0.9999r = 0.9999. For H2+_2^{+}, 36 basis functions were used in order to increase the accuracy even more. One can see from Table II that our FEM results and the results of the ThreeBSCbSSR program coincide reasonably well, with a maximal discrepancy less than 10−410^{-4} e.u. for r∼0.2r \sim0.2. Moreover, for the cases of H−^{-} and H2+_2^{+}, the agreement is by two orders of magnitude better. The difference between our GEM results and the ThreeBSCbSSR data does not exceed 5×10−85 \times10^{-8} e.u. and decreases to a few 10−910^{-9} e.u. for H−^{-} and H2+_2^{+}. It also confirms the high accuracy of the GEM.

The results of the SWEXPH program [96] were obtained with the variational parameters optimized for H2+_2^{+} and H−^{-}, for those it gives the ground state energy with a remarkable precision of minimum 40 decimal digits. Thus, for such rr they are the reference values. For an arbitrary value of the mass ratio, the results of the SWEXPH solver are less superior, but still suitable for comparison. For r≤0.8r \le0.8, we used the ansatz optimized for H−^{-}, whereas for other values the ansatz appropriate for H2+\mathrm{H}_{2}^{+}. The reason for choosing this value of rr is that, at r∼0.9r \sim0.9, as we will demonstrate below, the upper bound state levels appear in the spectrum, resulting in a significant change to the internal structure of the wave function. The low precision of the SWEXPH program for r→1r \to1 can also be caused by this effect. For H−\mathrm{H}^{-} and H2+\mathrm{H}_{2}^{+}, our FEM results and the data from the SWEXPH agree within an accuracy of order of 10−710^{-7} and 10−810^{-8} e.u., respectively. For intermediate values of rr, the difference is larger, reaching its maximal value of about 2×10−32 \times10^{-3} e.u. in the middle of [0,1][0,1], for r∼0.45r \sim0.45. Our GEM results for H−\mathrm{H}^{-} and H2+\mathrm{H}_{2}^{+} coincide with SWEXPH results with an excellent precision of 4×10−94 \times10^{-9} e.u. For intermediate values of rr, the difference reaches 2×10−32 \times10^{-3} e.u., that is comparable with the difference between FEM and SWEXPH. Similar to a comparison to our data, the discrepancy between the SWEXPH outcomes and ThreeBSCbSSR data is almost negligible for H−\mathrm{H}^{-} and H2+\mathrm{H}_{2}^{+}, whereas it reaches the maximum of the same 2×10−32 \times10^{-3} e.u. for r∼0.45r \sim0.45. As a result, we conclude that for intermediate values of the mass ratios the predefined for H−\mathrm{H}^{-} and H2+\mathrm{H}_{2}^{+} variational parameters of SWEXPH are less appropriate.

The results of Ref. [48] agree with our FEM and GEM data with an uncertainty of less than 10−410^{-4} e.u. for r≥0.9r \geq0.9, but, as rr becomes smaller, the difference to our data increases reaching its peak discrepancy of about 10−310^{-3} e.u. for r∼0.15−0.40r \sim0.15 - 0.40. It seems that for such mass ratios the method by Combescot overestimates the ground-state binding energies. Thus, for such rr the momentum-space calculations may be less precise.

ν\nu (vibr. q. n.)Our FEM result (a.u.)Our GEM result (a.u.)High-accuracy results (a.u.)ThreeBSCbSSR [91] (a.u.)
0-0.597139049-0.5971390616-0.597139063123...-0.5971390629
1-0.5871546-0.5871556771-0.587155679212...-0.587155665
2-0.57775129-0.5777519014-0.57775190459547-0.57775178
3-0.56890807-0.5689084930-0.56890849896677-0.5689078
4-0.5606091931-0.5606091963-0.56060922113307-0.5606062
5-0.5528421-0.552840656-0.55284075021966-0.552840676
6-0.5455956-0.54559236-0.54559265134900-0.545508
7-0.538862-0.5388557-0.53885738734702-
8-0.5326377-0.5326253-0.53263037975264-
9-0.526922-0.52689-0.52691012442161-
10-0.52172-0.521661-0.52169836942035-
11-0.517027-0.5169-0.51700036567716-
12-0.512867-0.51267-0.51282520352719-
13-0.50925-0.5089-0.50918624872335-
14-0.50621-0.5057-0.50610168128651-
15-0.50378-0.50323-0.50359508526777-
16-0.50198-0.50137-0.50169577359316-
17-0.50082-0.50022-0.50043704058981-
18-0.5002-0.49977-0.49983743207558-
19-0.5001-0.499726-0.4997312306558-

Table 3. Calculated energies of the states of H2+\mathrm{H}_{2}^{+} with the vibrational quantum numbers ν=0−19\nu= 0 - 19 for the total angular momentum J=0J = 0 in comparison with other results. The energies are given in a.u. The high-precision results for ν=0\nu= 0 are from Refs. [71, 72, 120], for ν=1\nu= 1 from Ref. [72], and for higher energies they are from Ref. [70]. The energies in the last column were obtained by the ThreeBSCbSSR program using 36 basis functions. For ν≥7\nu\geq7, results of the program did not converge.

Table III shows a comparison of our results on the vibrational energy levels of H2+\mathrm{H}_{2}^{+} with the high-precision data from different publications and the outputs from the ThreeBSCbSSR program. The energies for a full set of quantum numbers ν=0−19\nu= 0 - 19 are given in a.u. The distribution of the finite elements over radial variables in the FEM solver was optimized for the ground state calculations. Therefore, for the lower energy levels, the FEM shows a good agreement with alternative approaches, with an accuracy of less than 10−610^{-6} a.u. As ν\nu increases, the discrepancy with the reference values gradually grows. It reaches its maximum of 4×10−44 \times10^{-4} a.u. for the last bound state with ν=19\nu= 19. Our GEM results are more precise than the FEM data for ν≤8\nu\leq8 and ν≥17\nu\geq17. For 9≤ν≤169 \leq\nu\leq16 the FEM values are closer to the reference values, but lie below the exact energies due to the extrapolation. It is worth noting that the FEM energies for fixed numbers of the Legendre polynomials over the angular variable θ\theta are above the reference values. The GEM basis was also optimized for the ground state, which deviates from the reference by 1.6×10−91.6 \times10^{-9} a.u. The difference grows with ν\nu, up to 3.6×10−43.6 \times10^{-4} a.u. for ν=15\nu= 15. With further increase of ν\nu, the deviation becomes gradually smaller, though, relative to the binding energy, the uncertainty of GEM data grows monotonically. For the last calculated state, ν=19\nu= 19, the accuracy is about 5×10−65 \times10^{-6} a.u. One could see that the obtained energy lies slightly above the three-body threshold. These deviations stem from the density of the basis covering intermediate distances: increasing the minimum range (x1=0.02x_1 = 0.02) yields all 20 levels bound, at the cost of a less accurate ground state (with the difference from the reference value of order 10−710^{-7} a.u.). The accuracy of the calculated weakly bound states can be improved also by increasing the long-range parameter yNmax⁡y_{N\max} and a number of Gaussians.

The ThreeBSCbSSR program with 28 basis functions allows one to obtain only the six lowest energy levels. The precision of the data gradually decreases with an increase in the level number. The increase of the size of the basis set up to 36 allows one to calculate one more level. For larger numbers of basis functions, the calculation time dramatically increases. The comparison of our results and the data from the ThreeBSCbSSR program with 36 basis functions shows a good agreement of lower energy levels. The advantage of our methods is that they allow accurate calculations of all excited bound states.

Figure 4 shows the calculated energy levels of the X+X^+ trion for the total angular momentum J=0J = 0 as a function of the mass ratio mh/mem_h/m_e. It spans several orders of magnitude: from very large values, including the mass ratio of 1836.152701 for H2+\mathrm{H}_2^+, down to almost those values for which the system is similar to H−\mathrm{H}^-. In fact, the data for H−\mathrm{H}^- is not shown because there is only one bound state with the energy given in Table II. As in the table, the energies in the figure are given in excitonic Hartree units (e.u.). The data from the FEM and GEM calculations are precise enough to visually overlap in the plot. The dashed horizontal line shows the three-body scattering threshold, defined exactly by the exciton ground-state energy of −0.5-0.5 e.u.

Figure 4. Calculated energy levels of the X+X^+ trion in a bulk crystal as a function of the effective mass ratio for a total angular momentum J=0J = 0. The energies are given in e.u. The mass ratio corresponding to the case of the molecular hydrogen ion H2+\mathrm{H}_2^+ is denoted by a vertical line. The dashed horizontal line is the lower boundary of the continuum spectrum. The energy levels originating from the lowest (1sσ1s\sigma and 2pσ2p\sigma) subbands are shown. The quantum numbers ν\nu of three lowest vibrational levels are given. The inset shows the energies as a function of the ratio r=mh/(me+mh)r = m_h/(m_e + m_h).

For a mass ratio mh/me≲10m_h/m_e \lesssim10, there is only one bound state. This is related to the significant repulsion between the particles of the same charge, whose contribution, with respect to the kinetic energy, increases as mh/me→0m_h/m_e \to0, see the discussion in § 2 C. An appearance of the only bound state for a H−\mathrm{H}^--like system with me/mp≲0.21m_e/m_p \lesssim0.21 was rigorously shown by Hill [58]. However, the proved estimate of the mass ratio is not an exact bound. Our calculations show that the first excited bound state disappears already for mh/me≈9.7m_h/m_e \approx9.7, i.e., when the two identical holes are about one order of magnitude heavier than the electron. In the model of X+\mathrm{X}^{+} trion, its two holes play the role of the electrons in H−\mathrm{H}^{-}, whereas its electron plays the role of the proton.

For larger mh/mem_h/m_e, more energy levels of bound states appear in the spectrum. In contrast to the one-dimensional case (see discussion in § 2 C), there are no pairs of energy levels here. This originates from the fact that in 3D the coordinates of the relative motion are only nonnegative. Thus, the Born-Oppenheimer potential curves have only a single QW and the levels are single.

As the mass ratio increases, more vibrational energy levels, originating from the hole-hole vibrations, appear in the spectrum. They correspond to the lower electronic potential-energy curves. The energy levels originating from the lowest, 1sσ1s\sigma, and upper 2pσ2p\sigma subbands are clearly seen in the figure. The vibrational energy levels are indexed by ν=0,1,2…\nu= 0, 1, 2 \ldots. The 1sσ1s\sigma states are characterized by an even (’gerade’) wave function, and they are the bound states. The 2pσ2p\sigma subband is within the three-body continuum and thus beyond the scope of the current study. Whether these levels are proper resonances or numerical artifacts due to rigid-wall boundary conditions requires a dedicated analysis based on the complex-scaling method [122, 123].

As it was shown by Bethe and Salpeter [124], for mh/me→∞m_h/m_e \to\infty the vibrational energies scale as mh−1/2m_h^{-1/2}. Assuming this dependence as a leading term of the asymptotic expansion over a small parameter 1−r\sqrt{1-r}, we can extend a series introduced in Ref. [48] as

E(r)=E(1)+A1−r+B(1−r)+C(1−r)3/2+D(1−r)2+….(13)E(r) = E(1) + A\sqrt{1-r} + B(1-r) + C(1-r)^{3/2} + D(1-r)^2 + \ldots. \tag*{(13)}

This series represents a deviation of the energy from a constant value E(1)E(1) as rr becomes smaller. From Figure 4, one may surmise that the number of reasonable values of the expansion is defined by the form of a particular vibrational curve. The analysis shows that the first five terms of this series approximate the ground state energies with an average accuracy of about 10−410^{-4} e.u. Interestingly enough, the upper vibrational curves are approximated by one order of magnitude better because they are localized in the vicinity of r=1r = 1. The results of the fits to the most precisely calculated data are shown in Table IV. There are the first five parameters of Eq. (13), including E(1)E(1), fitted for the six lowest energy curves. The fitted values are statistically significant. The allowed ranges of rr are also given.

ν\nuE(1)E(1)AABBCCDDallowed rr
0−0.602726-0.6027260.2304220.230422−0.250908-0.2509080.1685900.168590−0.073041-0.073041r∈[0,1]r \in[0,1]
1−0.602629-0.6026290.6804590.680459−1.353728-1.3537281.3544791.354479−2.013583-2.013583r>0.907r > 0.907
2−0.602611-0.6026111.1320591.132059−3.544812-3.5448124.9517844.951784−14.60841-14.60841r>0.97r > 0.97
3−0.602638-0.6026381.5891841.589184−7.049382-7.04938215.8637315.86373−70.56026-70.56026r>0.985r > 0.985
4−0.602640-0.6026402.0434422.043442−11.59381-11.5938132.8322632.83226−187.5359-187.5359r>0.99r > 0.99
5−0.602654-0.6026542.5002442.500244−17.45588-17.4558864.3553764.35537−464.5818-464.5818r>0.994r > 0.994

Table IV. Results of the fit of Eq. (13) to the calculated data in Figure 4. The first five terms of the series are used. The last column shows the allowed values of the ratio r=mh/(me+mh)r = m_h/(m_e + m_h), for which the approximation is reasonable. The values are given in e.u.. The fitted values of energy in e.u. for r=1r = 1 coincide with the values in a.u.. All obtained values are statistically significant.

One can see in Table IV that, up to a precision of calculation, the energy E(1)E(1) is constant for all the fitted vibrational levels. This means that all the levels converge to the same value as r→1r \to1 (or mh/me→∞m_h/m_e \to\infty), i.e. to the ground state of H2+\mathrm{H}_2^{+} with infinite nucleus mass. Moreover, our extrapolated energies are close to the calculated value in Table II, as well as to the precise value of -0.602633765597 a.u., reported in Ref. [121]. This additionally ensures correctness of the applied expansion.

Total angular momentum J=1J = 1

We calculated the energy levels of the trion states for the total angular momentum J=1J = 1 as well. Due to the increased number of coupled equations and partial waves taken into account, calculations for J=1J = 1 required significantly more computational resources to store the matrices and to solve the eigenvalue problem.

For the FEM solver, because of limitations introduced by PARDISO to a number of nonzero elements of the matrices, the number of finite elements over each spatial variable was taken to be 30. It was only possible to achieve a convergence of the results with an accuracy of 4×10−54 \times10^{-5} e.u. for mass ratios r≳0.9r \gtrsim0.9. For lower mass ratios, the precision of our calculation is by one order of magnitude worse. In particular, the convergence of the lowest energy for 0.1≲mh/me≲50.1 \lesssim m_h/m_e \lesssim5, when, as it will be clear below, the ground state wave function is large enough, requires an increased number (up to 40) of Legendre polynomials over the angular variable. To compensate this increase and keeping the same matrix sizes, the number of finite elements over the spatial variables was decreased down to 20. As a result, for 0.1≲mh/me≲50.1 \lesssim m_h/m_e \lesssim5, the precision of our FEM calculation is 3×10−33 \times10^{-3} e.u. For lower mass ratios, including the one corresponding to H−\mathrm{H}^-, our precision is again below 10−310^{-3} e.u. .

For the GEM calculations, memory constraints also restricted the sizes of the matrices. The larger number of (l,L)(l,L) channels, nlL=9n_{lL}=9, required for J=1J=1, forced us to use fewer Gaussians per radial coordinates than for J=0J=0. Nevertheless, the precision of GEM results for H2+\mathrm{H}_2^+ ground-state energy is of order 10−810^{-8} e.u. . For 0.99≤r≤0.9990.99 \leq r \leq0.999 the estimated accuracy is by one order of magnitude worse and for 0.85<r≤0.980.85 < r \leq0.98 it is of order 10−610^{-6} e.u. For smaller values of rr, the calculated ground state is either very weakly bound or completely delocalized: its wave function extends beyond the basis range. Thus, the precision of 2×10−52 \times10^{-5} e.u. is estimated for r∼0.81r \sim0.81.

The calculated energy levels for the total angular momentum J=1J=1 as a function of mh/mem_h/m_e are shown in Figure 5. Similar to the case of J=0J=0, for large mass ratios there are many vibrational energy levels. As mh/mem_h/m_e decreases, the energies of the bound states gradually increase. Therefore, the bound state energies transit into the continuous part of the spectrum. The number of bound states decreases and for mh/me<4.2m_h/m_e < 4.2 there are no bound states in the system at all. The achieved precision of minimum 3×10−33 \times10^{-3} e.u. for FEM calculations and, moreover, of order 10−510^{-5} e.u. for GEM calculations allows us to reasonably argue about the absence of bound states for such mass ratios. This additionally confirms the well-known fact that H−\mathrm{H}^--like systems have only one bound state [58], corresponding to the case J=0J=0. In fact, this means that for semiconductor compounds with effective masses of the same order of magnitude, for instance, for Cu2_2O, there is only one trion state with total angular momentum J=0J = 0. All the energies calculated by our methods for mh/me<4.2m_h/m_e < 4.2 are in the three-body continuum. As it is unclear whether they correspond to proper resonant states or artificially discretized continuum states, they are beyond the energy range shown in Figure 5. In any case, they lie far enough above the scattering threshold. This can be seen by a narrow strip of the continuum region above the dashed line in Figure 5.

Figure 5. Calculated energy levels of the X+X^+ trion in a bulk crystal as a function of the effective mass ratio for the total angular momentum J=1J=1. The energies are given in e.u. . The mass ratio corresponding to the case of the molecular hydrogen ion H2+\mathrm{H}_2^+ is denoted. The dashed horizontal line is the lower boundary of the continuous spectrum. The inset shows the energies as a function of the ratio r=mh/(me+mh)r=m_h/(m_e+m_h).

The comparison of the energies of the lowest state of the X+X^+ trion with total angular momentum J=1J = 1 is given in Table V. The data are only given for values r≥0.8r \ge0.8, since for most of these values both methods find a bound state. The second and third columns show our results, whereas the fourth column lists the data from the SWEXPH program [96]. The energy values are given with the number of decimal digits corresponding to the order of the estimated uncertainty plus two digits, but not less than four digits. For the mass ratio corresponding to H2+_2^+, the additional high-precision result from Ref. [70] is given in e.u. as a reference. The results of SWEXPH are obtained with the variational parameters optimized for the ground state of H2+_2^+ with J=0J = 0.

J=1J = 1
r=mh/(me+mh)r = m_h/(m_e + m_h) (e.u.)Our FEM result (e.u.)Our GEM result (e.u.)SWEXPH [96] (e.u.)Reference value [70] (e.u.)
0.80-0.49927-0.4991609-0.4883
0.81-0.50054-0.5002221-0.4980
0.85-0.50762-0.50729494-0.5056
0.90-0.521762-0.52164086-0.5235
0.95-0.544489-0.54447612-0.5462
0.96-0.550755-0.55074404-0.5522
0.97-0.557957-0.55796718-0.5590
0.98-0.566572-0.56659133-0.5672
0.99-0.577706-0.577730515-0.57789
0.999-0.595174-0.595202342-0.595202318
H2+_2^+ 0.9994556794-0.597170-0.5971987935-0.5971987953-0.5971988064
0.9999-0.600308-0.600338716-0.5959
1-0.602479

Table V. Comparison of the X+X^+ trion lowest-state energies (the total angular momentum J=1J = 1) for different effective mass ratios r=mh/(me+mh)r = m_h/(m_e + m_h). The energies are given in excitonic Hartree units (e.u.). For the mass ratio corresponding to H2+_2^+, the high-precision data from Ref. [70] is additionally given. The energy for the mass ratio r=1r = 1 is obtained from the extrapolation. For r=0.81r = 0.81 the GEM calculation used an extended basis with yNmax⁡=129y_{N_{\max}} = 129. The energy for r=1r = 1 is calculated for rr value shifted by 10−1210^{-12}.

It can be seen that, similar to zero total angular momentum data, the GEM results are more accurate than the FEM results. For the mass ratio corresponding to H2+_2^+, the results of GEM and the SWEXPH program remarkably agree up to the eighth decimal digit. Moreover, they coincide with the reference value from Ref. [70] with a similar precision of 10−810^{-8} e.u. The FEM results are by three orders of magnitude less precise. For smaller values of rr, the discrepancy between FEM and GEM data gradually increases as the mass ratio approaches r∼0.81r \sim0.81. For this rr, both the results agree with an uncertainty of about 3×10−43 \times10^{-4} e.u. For such ratios, the wave function extends very far from the origin and further, as the energy crosses the scattering threshold, becomes completely delocalized. A calculation of weakly bound states requires using the extended angular basis and an increased value of the parameter yNmax⁡y_{N_{\max}} to achieve convergence in our FEM and GEM computations, respectively. In the other limit, r→1r \to1, the FEM and GEM results agree well for r=0.9999r = 0.9999, but the SWEXPH program gives much less precise energy. In Table V the energy for the mass ratio r=1r = 1 is obtained for r=1−10−12r = 1 - 10^{-12}. The value is slightly above the related energy for J=0J = 0.

It is worth noting that for small mass ratios, mh/me≲1m_h/m_e \lesssim1 or r∼0.5r \sim0.5, the SWEXPH program gives the energy corresponding to the discrete part of the spectrum. In particular, for the negatively charged hydrogen ion H−^-, it yields an energy of about −0.5071-0.5071 e.u. Both our FEM and GEM results do not show any bound states for mh/me<4.2m_h/m_e < 4.2. All our results are in the continuous part of the spectrum. Therefore, we believe that the SWEXPH program with the predefined form of the ground-state wave function provides only approximate data for this range of the mass ratios.

Conclusion and Outlook

In this article, we calculated several of the lowest energy levels of the semiconductor trion for a wide range of hole-to-electron mass ratios, covering most of the existing semiconductor compounds. We considered trions in a quantum wire as well as in a bulk crystal. We obtained qualitatively unified dependencies of the boundstate energies on mh/mem_h/m_e and observed changes of the energy spectra during a crossover from small to large mass ratios.

The trion energy levels in the bulk were computed by two different methods, FEM and GEM, which produced similar results. We show the advantage of the GEM: its results are more precise than the FEM data. For both the 1D and 3D geometries of the system there is only one bound state for the small values of the mass ratio, as predicted for the analogous atomic H−^{-} system [58]. Our calculations improve the precision of the data from Refs. [46, 48, 49] and demonstrate that for most semiconductor compounds only one trion state with zero total angular momentum exists. The sole trion state takes place for the mass ratio of two different particles, i.e. mh/mem_h/m_e in case of X+X^{+} trion and me/mhm_e/m_h for X−X^{-}, to be below 9.7. For J=1J = 1, both applied methods show no trion bound states for the ratio below 4.2. With increasing mass ratio, more three-body bound states originating from the hole-hole vibrations appear in the spectrum. Their energies are simultaneously decreasing, gradually filling all the energy range between the ground state and the continuum. For mh/me∼1836.152701m_h/m_e \sim1836.152701, the studied system is identical to H2+_2^{+} with 20 well-known vibrational states. In the limit of mh/me→∞m_h/m_e \to\infty the quantization disappears and the density of three-body states forms a continuum above the lowest energy of the H2+_2^{+} with the infinite nucleus mass.

The binding energy of the trion (depending on whether it is X+X^{+} or X−X^{-} trion) is of about 5–10% of the exciton binding energy. From our calculations one can obtain more precise values by scaling the numerical results for particular semiconductor compounds by their corresponding Hartree energies.

The results obtained here within the simplest two-band model show a unified dependence of bound state energies on the mass ratio. In contrast to earlier works [57], this dependence is nonlinear: one can observe the crossover of the energy levels based on the same theoretical footing. Nevertheless, the energies above the three-body scattering threshold, the three-body resonant states [125], can be further studied using a more advanced complex-scaling method [122, 123]. Our results can also be improved for the particular semiconductors by taking into account the complex valence-band structure [126], nonparabolicity of the conduction band [36] and other effects, i.e., the central cell corrections [16], screening of the potential [127, 128], and so on.

Acknowledgments

The authors are grateful to I. N. Filikhin, R. Ya. Kezerashvili, and V. I. Korobov for fruitful discussions. P. A. B. and S. S. acknowledge financial support by the National Research Council Canada, grant No. QSP-129-1.

References

  1. [1]E. Iychenko, Optical Spectroscopy of Semiconductor Nanostructures (Alpha Science, Harrow, 2005).
  2. [2]C. Klingshirn, Semiconductor Optics, Advanced Texts in Physics (Springer, Berlin, Heidelberg, 2007).
  3. [3]T. Kazimierczuk, D. Fröhlich, S. Scheel, H. Stolz, and M. Bayer, Giant Rydberg excitons in the copper oxide Cu2O, Nature 514, 343 (2014).DOI
  4. [4]G. Plechinger, P. Nagler, A. Arora, R. Schmidt, A. Chernikov, A. G. del Águila, P. C. Christianen, R. Bratschitsch, C. Schüller, and T. Korn, Trion fine structure and coupled spin–valley dynamics in monolayer tungsten disulfide, Nature Communications 7, 12715 (2016).DOI
  5. [5]K.-Q. Lin, J. D. Ziegler, M. A. Semina, J. V. Mamedov, K. Watanabe, T. Taniguchi, S. Bange, A. Chernikov, M. M. Glazov, and J. M. Lupton, High-lying valley-polarized trions in 2D semiconductors, Nature Communications 13, 6980 (2022).DOI
  6. [6]R. Qi, Q. Li, Z. Zhang, S. Chen, J. Xie, Y. Ou, Z. Cui, D. D. Dai, A. Y. Joe, T. Taniguchi, K. Watanabe, S. Tongay, A. Zettl, L. Fu, and F. Wang, Electrically controlled interlayer trion fluid in electron-hole bilayers, Science 390, 299 (2025).arxiv.org/abs/2312.03251
  7. [7]P. X. Nguyen, R. Chatzurvedi, L. Ma, P. Knuppel, K. Watanabe, T. Taniguchi, K. F. Mak, and J. Shan, An equilibrium trion liquid in atomic double layers, Science 390, 304 (2025).DOI
  8. [8]G. Wang, A. Chernikov, M. M. Glazov, T. F. Heinz, X. Marie, T. Amand, and B. Urbaszek, Colloquium: Excitons in atomically thin transition metal dichalcogenides, Rev. Mod. Phys. 90, 021001 (2018).DOI
  9. [9]C. Fey, P. Schmelcher, A. Imamoglu, and R. Schmidt, Theory of exciton-electron scattering in atomically thin semiconductors, Phys. Rev. B 101, 195417 (2020).DOI
  10. [10]G. Grosso, J. Graves, A. T. Hammack, A. A. High, L. V. Butov, M. Hanson, and A. C. Gossard, Excitonic switches operating at around 100 K, Nature Photonics 3, 577 (2009).
  11. [11]P. Andreaukov, S. V. Poltavtsev, J. R. Leonard, E. V. Calman, M. Remeika, Y. Y. Kuznetsova, L. V. Butov, J. Wilkes, M. Hanson, and A. C. Gossard, Optically controlled excitonic transistor, Applied Physics Letters 104, 091101 (2014).
  12. [12]L. Butov, Excitonic devices, Superlattices and Microstructures 108, 2 (2017).
  13. [13]D. H. Meyer, Z. A. Castillo, K. C. Cox, and P. D. Kunz, Assessment of Rydberg atoms for wideband electric field sensing, Journal of Physics B: Atomic, Molecular and Optical Physics 53, 034001 (2020).DOI
  14. [14]M. A. Chukeev, S. Zheng, E. S. Khramtsov, I. V. Ignatiev, S. A. Eliseev, V. A. Lovtcius, Y. P. Efimov, and M. A. Lozhkin, Excitonic sensor of electric field in quantum-well heterostructures, Phys. Rev. B 109, 235305 (2024).DOI
  15. [15]T. Godde, D. Schmidt, J. Schmutzler, M. Abmann, J. Debus, F. Withers, E. M. Alexeev, O. Del Pozo-Zamudio, O. V. Skrypka, K. S. Novoselov, M. Bayer, and A. I. Tartakovskii, Exciton and trion dynamics in atomically thin MoSe₂ and WSe₂: Effect of localization, Phys. Rev. B 94, 165301 (2016).
  16. [16]F. Schweiner, J. Main, M. Feldmaier, G. Wunner, and C. Uihlein, Impact of the valence band structure of Cu₂O on excitonic spectra, Phys. Rev. B 93, 195203 (2016).DOI
  17. [17]J. Heckötter, M. Freitag, D. Fröhlich, M. Aßmann, M. Bayer, M. A. Semina, and M. M. Glazov, High-resolution study of the yellow excitons in Cu₂O subject to an electric field, Phys. Rev. B 95, 035210 (2017).DOI
  18. [18]J. Heckötter, A. Farbrunch, D. Fröhlich, M. Aßmann, D. Yakovlev, M. Bayer, M. Semina, M. Glazov, P. Rommel, J. Ertl, J. Main, and H. Stolz, The energy level spectrum of the yellow excitons in cuprous oxide, Physics Reports 1100, 1 (2025).DOI
  19. [19]J. Jasiński, J. Hagel, S. Brem, E. Wietek, T. Taniguchi, K. Watanabe, A. Chernikov, N. Bruyant, M. Dydsik, A. Surrente, M. Baranowski, D. K. Maude, E. Malic, and P. Plockocha, Quadrupolar excitons in MoSe₂ bilayers, Nature Communications 16, 1382 (2025).DOI
  20. [20]P. Massignan, R. Schmidt, G. E. Astrakharchik, A. Imamoglu, M. Zwierlein, J. J. Arlt, and G. M. Bruun, Polarons in atomic gases and two-dimensional semiconductors, Rev. Mod. Phys. 98, 035002 (2026).DOI
  21. [21]J. Lee, N. C. Giles, D. Rajavel, and C. J. Summers, Room-temperature band-edge photoluminescence from cadmium telluride, Phys. Rev. B 49, 1668 (1994).pubmed.ncbi.nlm.nih.gov/10010956
  22. [22]P. Horodyský, R. Grill, and P. Hlídek, Band-edge photoluminescence in CdTe, Phys. Status Solidi B 243, 2882 (2006).DOI
  23. [23]M. A. Lampert, Mobile and immobile effective-mass-particle complexes in nonmetallic solids, Phys. Rev. Lett. 1, 450 (1958).DOI
  24. [24]K. Kheng, R. T. Cox, M. Y. d’ Aubigné, F. Bassani, K. Saminadayar, and S. Tatarenko, Observation of negatively charged excitons X− in semiconductor quantum wells, Phys. Rev. Lett. 71, 1752 (1993).pubmed.ncbi.nlm.nih.gov/10054489
  25. [25]J. H. Davies, The Physics of Low-dimensional Semiconductors: An Introduction (Cambridge University Press, Cambridge, 1997).
  26. [26]P. A. Belov, Energy spectrum of excitons in square quantum wells, Physica E: Low-dimensional Systems and Nanostructures 112, 96 (2019).DOI
  27. [27]R. Y. Kezerashvili, Few-body systems in condensed matter physics, Few-Body Systems 60, 52 (2019).DOI
  28. [28]A. Shields, J. Osborne, M. Simmons, M. Pepper, and D. Ritchie, Magneto-optical spectroscopy of positively charged excitons in GaAs quantum wells, Phys. Rev. B 52, R5523(R) (1995).pubmed.ncbi.nlm.nih.gov/9981820
  29. [29]E. S. Khramtsov, P. A. Belov, P. S. Grigoryev, I. V. Ignatiev, S. Y. Verbin, Y. P. Efimov, S. A. Eliseev, V. A. Lovtcius, V. V. Petrov, and S. L. Yakovlev, Radiative decay rate of excitons in square quantum wells: Microscopic modeling and experiment, Journal of Applied Physics 119, 184301 (2016).DOI
  30. [30]H. Stolz, private communication (2026).
  31. [31]R. S. Knox, Theory of excitons (Academic Press, New York, 1963).
  32. [32]D. K. Efimkin and A. H. MacDonald, Many-body theory of trion absorption features in two-dimensional semiconductors, Phys. Rev. B 95, 035417 (2017).DOI
  33. [33]A. M. Shentsev and M. M. Glazov, Terahertz radiation induced attractive-repulsive Fermi polaron conversion in transition metal dichalcogenide monolayers, Phys. Rev. B 113, 125307 (2026).arxiv.org/abs/2509.15708
  34. [34]A. M. Frolov, Coulomb three-body problem in the second quantized form, Phys. Rev. A 64, 042503 (2001).DOI
  35. [35]J. B. McGuire, Study of exactly soluble one-dimensional N-body problems, Journal of Mathematical Physics 5, 622 (1964).DOI
  36. [36]P. A. Belov and E. S. Khramtsov, The binding energy of excitons in narrow quantum wells, Journal of Physics: Conference Series 816, 012018 (2017).DOI
  37. [37]P. A. Belov, F. Morawetz, S. O. Krüger, N. Scheuler, P. Rommel, J. Main, H. Giessen, and S. Scheel, Energy states of Rydberg excitons in finite crystals: From weak to strong confinement, Phys. Rev. B 109, 235404 (2024).DOI
  38. [38]J. Thewes, J. Heckötter, T. Kazimierczuk, M. Aßmann, D. Fröhlich, M. Bayer, M. A. Semina, and M. M. Glazov, Observation of high angular momentum excitons in cuprous oxide, Phys. Rev. Lett. 115, 027402 (2015).DOI
  39. [39]F. Schöne, S.-O. Krüger, P. Grünwald, H. Stolz, S. Scheel, M. Aßmann, J. Heckötter, J. Thewes, D. Fröhlich, and M. Bayer, Deviations of the exciton level spectrum in Cu₂O from the hydrogen series, Phys. Rev. B 93, 075203 (2016).DOI
  40. [40]F. Schweiner, J. Main, and G. Wunner, Linewidths in excitonic absorption spectra of cuprous oxide, Phys. Rev. B 93, 085203 (2016).DOI
  41. [41]F. Schweiner, J. Main, G. Wunner, M. Freitag, J. Heckötter, C. Uihlein, M. Aßmann, D. Fröhlich, and M. Bayer, Magnetoexcitons in cuprous oxide, Phys. Rev. B 95, 035202 (2017).DOI
  42. [42]P. Rommel, F. Schweiner, J. Main, J. Heckötter, M. Freitag, D. Fröhlich, K. Lehninger, M. Aßmann, and M. Bayer, Magneto-Stark effect of yellow excitons in cuprous oxide, Phys. Rev. B 98, 085206 (2018).DOI
  43. [43]J. Heckötter, M. Freitag, D. Fröhlich, M. Aßmann, M. Bayer, M. A. Semina, and M. M. Glazov, Dissociation of excitons in Cu₂O by an electric field, Phys. Rev. B 98, 035150 (2018).DOI
  44. [44]B. Stébé, G. Munschy, L. Stauffer, F. Dujardin, and J. Murat, Excitonic trion X− in semiconductor quantum wells, Phys. Rev. B 56, 12454 (1997).DOI
  45. [45]Y. Suzuki and K. Varga, Stochastic Variational Approach to Quantum-Mechanical Few-Body Problems, Lecture Notes in Physics. Monographs (Springer-Verlag Berlin Heidelberg, 1998).
  46. [46]J. Usukura, Y. Suzuki, and K. Varga, Stability of two- and three-dimensional excitonic complexes, Phys. Rev. B 59, 5652 (1999).DOI
  47. [47]A. Wójs and J. J. Quinn, Exact-diagonalization studies of three energy spectra in high magnetic fields, Phys. Rev. B 75, 085318 (2007).DOI
  48. [48]R. Combescot, Trion ground-state energy: Simple results, Phys. Rev. B 100, 245201 (2019).DOI
  49. [49]F. Marsusiak, E. Mostaani, and N. D. Drummond, Quantum Monte Carlo study of three-dimensional Coulomb complexes: Trions and biexcitons, hydrogen molecules and ions, helium hydride cations, and positronic and muonic complexes, Phys. Rev. A 106, 062822 (2022).DOI
  50. [50]N. P. Mehta, B. D. Esry, and C. H. Greene, Three-body recombination in one dimension, Phys. Rev. A 76, 022711 (2007).DOI
  51. [51]N. P. Mehta, Born-Oppenheimer study of two-component few-particle systems under one-dimensional confinement, Phys. Rev. A 89, 052706 (2014).DOI
  52. [52]L. Happ, M. Zimmermann, S. I. Betelu, W. P. Schleich, and M. A. Efremov, University in a one-dimensional three-body system, Phys. Rev. A 100, 012709 (2019).DOI
  53. [53]L. Happ, M. Zimmermann, and M. A. Efremov, Universality of excited three-body bound states in one dimension, Journal of Physics B: Atomic, Molecular and Optical Physics 55, 015301 (2022).DOI
  54. [54]M. A. Semina, R. A. Seeger, and R. A. Suris, The binding energy of excitons and X+ and X− trions in one-dimensional systems, Semiconductors 42, 1427 (2008).DOI
  55. [55]M. A. Semina and R. A. Suris, Localized excitons and trions in semiconductor nanosystems, Phys. Usp. 65, 111 (2022).
  56. [56]M. Born and R. Oppenheimer, Zur Quantentheorie der Molekeln, Annalen der Physik 389, 457 (1927).
  57. [57]C. Lin, Hyperspherical coordinate approach to atomic and other Coulombic three-body systems, Physics Reports 257, 1 (1995).DOI
  58. [58]R. N. Hill, Proof that the H− ion has only one bound state. Details and extension to finite nuclear mass, Journal of Mathematical Physics 18, 2316 (1977).DOI
  59. [59]L. D. Landau and L. M. Lifshitz, Quantum Mechanics Non-Relativistic Theory: Volume 3 (Butterworth-Heinemann, Oxford, 1981).
  60. [60]A. Carrington, I. R. McNab, and C. A. Montgomerie, Spectroscopy of the hydrogen molecular ion, Journal of Physics B: Atomic, Molecular and Optical Physics 22, 3551 (1989).
  61. [61]E. Teller, Über das Wasserstoffmolekül, Zeitschrift für Physik 61, 458 (1930).
  62. [62]F. A. Matsen, The united atom treatment of H₂+, The Journal of Chemical Physics 21, 928 (1953).DOI
  63. [63]D. R. Bates, K. Ledsham, and A. L. Stewart, Wave functions of the hydrogen molecular ion, Philosophical Transactions of the Royal Society of London, Series A: Mathematical and Physical Sciences 246, 215 (1953).DOI
  64. [64]J. M. Peek, Eigenparameters for the 1sσg and 2pσu orbitals of H₂+, The Journal of Chemical Physics 43, 3004 (1965).DOI
  65. [65]I. V. Komarov, L. I. Ponomarev, and S. Y. Slavyanov, Spheroidal and Coulombian Spheroidal Functions (Nauka, Moscow, 1976).
  66. [66]R.-H. Xie and J. Gong, Simple three-parameter model potential for diatomic systems: From weakly and strongly bound molecules to metastable molecular ions, Phys. Rev. Lett. 95, 263202 (2005).pubmed.ncbi.nlm.nih.gov/16486350
  67. [67]P. M. Morse and E. C. G. Stueckelberg, Diatomic molecules according to the wave mechanics I: Electronic levels of the hydrogen molecular ion, Phys. Rev. 33, 932 (1929).DOI
  68. [68]P. M. Morse, Diatomic molecules according to the wave mechanics. II. Vibrational levels, Phys. Rev. 34, 57 (1929).DOI
  69. [69]H. Wind, Vibrational states of the hydrogen molecular ion, The Journal of Chemical Physics 43, 2956 (1965).
  70. [70]Hilico, L., Billy, N., Grémaud, B., and Delande, D., Ab initio calculation of the J = 0 and J = 1 states of the H₂+, D₂+ molecular ions, Eur. Phys. J. D 12, 449 (2000).
  71. [71]V. I. Korobov, Coulomb three-body bound-state problem: Variational calculations of nonrelativistic energies, Phys. Rev. A 61, 064503 (2000).DOI
  72. [72]H. Li, J. Wu, B.-L. Zhou, J.-M. Zhu, and Z.-C. Yan, Calculations of energies of the hydrogen molecular ion, Phys. Rev. A 75, 012504 (2007).DOI
  73. [73]E. Petrimoulx, A. Bondy, E. Ene, L. Sati, and G. Drake, Ground-state energy of H−: a critical test of triple basis sets, Canadian Journal of Physics 103, 60 (2025).DOI
  74. [74]A. Scrinzi, A 3-dimensional finite elements procedure for quantum mechanical applications, Computer Physics Communications 86, 67 (1995).
  75. [75]E. Yarevsky and N. Elander, A finite-element approach to transient wavelengths in antiprotonic helium, Europhysics Letters 37, 453 (1997).DOI
  76. [76]N. Elander and E. Yarevsky, Finite-element calculations of the antiprotonic helium atom including relativistic and qed corrections, Phys. Rev. A 56, 1855 (1997).DOI
  77. [77]A. Gagin, E. Yarevsky, M. Salci, and N. Elander, Eigen energies and the statistical distributions of the rovibrational levels of the bosonic van der waals argon trimer, The Journal of Physical Chemistry A 113, 14979 (2009).DOI
  78. [78]E. Hiyama, Y. Kino, and M. Kamimura, Gaussian expansion method for few-body systems, Progress in Particle and Nuclear Physics 51, 223 (2003).DOI
  79. [79]E. Hiyama, R. Lazauskas, and J. Carbonell, 7H ground state as a 3H+4n resonance, Physics Letters B 833, 137367 (2022).
  80. [80]L. Happ, Fewbodytoolkit.jl: A Julia package for solving quantum few-body problems, SciPost Phys. Codebases, 81 (2026).DOI
  81. [81]I. Filikhin, R. Y. Kezerashvili, S. M. Tsiklauri, and B. Vlahovic, Trions in bulk and monolayer materials: Faddeev equations and hyperspherical harmonics, Nanotechnology 29, 124002 (2018).DOI
  82. [82]I. Filikhin, R. Y. Kezerashvili, and B. Vlahovic, On binding energy of trions in bulk materials, Physics Letters A 382, 787 (2018).DOI
  83. [83]D. T. Aznabaev, A. K. Bekbaev, and V. I. Korobov, Nonrelativistic energy levels of helium atoms, Phys. Rev. A 98, 012510 (2018).DOI
  84. [84]E. A. Kolganova, A. K. Motovilov, and S. A. Sofianos, Three-body configuration space calculations with hardcore potentials, Journal of Physics B: Atomic, Molecular and Optical Physics 31, 1279 (1998).DOI
  85. [85]S. Edvardsson, K. Karlsson, and H. Olin, corr3p__tr: A particle approach for the general three-body problem, Computer Physics Communications 200, 259 (2016).
  86. [86]N. K. Timofeyuk, Convergence of the hyperspherical-harmonics expansion with increasing number of particles for bosonic systems, Phys. Rev. A 86, 032507 (2012).DOI
  87. [87]N. K. Timofeyuk and D. Baye, Hyperspherical harmonics expansion on lagrange meshes for bosonic systems in one dimension, Few-Body Systems 58, 157 (2017).DOI
  88. [88]R. Y. Kezerashvili, S. M. Tsiklauri, and A. Dublin, Trions in two-dimensional monolayers within the hyperspherical harmonics method: Application to transition metal dichalcogenides, Phys. Rev. B 109, 085406 (2024).arxiv.org/abs/2310.19196
  89. [89]P. Maris, J. P. Vary, and A. M. Shirokov, Ab initio no-core full configuration calculations of light nuclei, Phys. Rev. C 79, 014308 (2009).DOI
  90. [90]M. A. Caprio, P. Maris, and J. P. Vary, Coulomb-Sturmian basis for the nuclear many-body problem, Phys. Rev. C 86, 034312 (2012).DOI
  91. [91]E. Z. Liverts and N. Barnea, Three-body systems with Coulomb interaction. Bound and quasi-bound S-states, Computer Physics Communications 184, 2596 (2013).
  92. [92]H. Bachau, E. Cormier, P. Deleclova, J. E. Hansen, and F. Martín, Applications of B-splines in atomic and molecular physics, Reports on Progress in Physics 64, 1815 (2001).
  93. [93]V. A. Roudnev, S. L. Yakovlev, and S. A. Sofianos, Bound-state calculations for three atoms without explicit partial wave decomposition, Few-Body Systems 37, 179 (2005).DOI
  94. [94]P. A. Belov and S. L. Yakovlev, Asymptotic method for determining the amplitude for three-particle breakup: Neutron-deuteron scattering, Physics of Atomic Nuclei 76, 126 (2013).DOI
  95. [95]V. Roudnev and M. Cavagnero, Benchmark helium dimer and trimer calculations with a public few-body code, Journal of Physics B: Atomic, Molecular and Optical Physics 45, 025101 (2011).
  96. [96]V. Korobov and J. Buša, SWEXPHC: Variational bound state solution for the three-body nonrelativistic Schrödinger equation, Computer Physics Communications 319, 109920 (2026).
  97. [97]O. I. Kartavtsev, A. V. Malykh, and S. A. Sofianos, Bound states and scattering lengths of three two-component particles with zero-range interactions under one-dimensional confinement, Journal of Experimental and Theoretical Physics 108, 365 (2009).
  98. [98]L. D. Faddeev and S. P. Merkuriev, Quantum Scattering Theory for Several Particle Systems, Mathematical Physics and Applied Mathematics (Springer, Dordrecht, 1993).DOI
  99. [99]R. Elliott and R. Loudon, Theory of the absorption edge in semiconductors in a high magnetic field, Journal of Physics and Chemistry of Solids 15, 196 (1960).DOI
  100. [100]K. Chadan, N. N. Khuri, A. Martin, and T. Tsun Wu, Bound states in one and two spatial dimensions, Journal of Mathematical Physics 44, 406 (2003).DOI
  101. [101]H. Mustroph, Potential-energy surfaces, the Born–Oppenheimer approximations, and the Franck–Condon principle: Back to the roots, ChemPhysChem 17, 2616 (2016).
  102. [102]M. S. Schuurman and A. Stolow, Dynamics at Conical Intersections, Annual Review of Physical Chemistry 69, 427 (2018).DOI
  103. [103]A. A. Samarskii, The Theory of Difference Schemes, Pure & Applied Mathematics (CRC Press, Boca Raton, 2001).
  104. [104]N. Scheuler, P. Rommel, J. Main, and P. A. Belov, Resonance energies and linewidths of Rydberg excitons in Cu2O quantum wells, Phys. Rev. B 109, 165440 (2024).
  105. [105]A. Aslanidis, J. Main, P. Rommel, S. Scheel, and P. A. Belov, Bound states in the continuum in cuprous oxide quantum wells, Phys. Rev. B 111, L121103 (2025).DOI
  106. [106]L. Kühner, P. Rommel, J. Main, S. Scheel, and P. A. Belov, Wave functions and oscillator strengths of Rydberg excitons in cuprous oxide quantum wells, Phys. Rev. B 112, 075409 (2025).DOI
  107. [107]O. C. Zienkiewicz and K. Morgan, Finite Elements and Approximation (Dover Publications, New York, 2006).
  108. [108]V. Roudnev and M. Cavagnero, Automatic grid construction for few-body quantum-mechanical calculations, Computer Physics Communications 182, 2099 (2011).DOI
  109. [109]C. de Boor and B. Swartz, Collocation at gaussian points, SIAM Journal on Numerical Analysis 10, 582 (1973).DOI
  110. [110]K. Yang and M. de Llano, Simple variational proof that any two-dimensional potential well supports at least one bound state, American Journal of Physics 57, 85 (1989).
  111. [111]S. Merkuriev, C. Gignoux, and A. Laverne, Three-body scattering in configuration space, Annals of Physics 99, 30 (1976).
  112. [112]V. K. Khersonskii, A. N. Moskalev, and D. A. Varshalovich, Quantum Theory Of Angular Momentum (World Scientific Publishing Company, 1988).
  113. [113]D. I. Abramov, Hyperspherical coordinates in the Coulomb three-body problem, Journal of Physics B: Atomic, Molecular and Optical Physics 41, 175201 (2008).arxiv.org/abs/0803.2456
  114. [114]P. A. Belov, E. R. Nुगуманов, and S. L. Yakovlev, The arrowhead decomposition method for a block-tridiagonal system of linear equations, Journal of Physics: Conference Series 929, 012035 (2017).DOI
  115. [115]R. B. Lehoucq, D. C. Sorensen, and C. Yang, ARPACK Users’ Guide (Society for Industrial and Applied Mathematics, Philadelphia, 1998).
  116. [116]O. Schenk and K. Gärtner, Solving unsymmetric sparse systems of linear equations with PARDISO, Future Generation Computer Systems 20, 475 (2004).
  117. [117]Intel® oneAPI Math Kernel Library 2025.1 (2025).
  118. [118]E. Hiyama and M. Kamimura, Study of various few-body systems using gaussian expansion method (GEM), Frontiers of Physics 13, 132106 (2018).DOI
  119. [119]E. Anderson, Z. Bai, C. Bischof, L. S. Blackford, J. Demmel, J. J. Dongarra, J. Du Croz, S. Hammarling, A. Greenbaum, A. McKenney, and D. Sorensen, LAPACK Users’ Guide (Third Ed.) (Society for Industrial and Applied Mathematics, Philadelphia, 1999).
  120. [120]V. I. Korobov, Variational methods in the quantum mechanical three-body problem with a Coulomb interaction, Physics of Particles and Nuclei 53, 1 (2022).
  121. [121]A. M. Frolov, Highly accurate bound state calculations of the two-center molecular ions by using the universal variational expansion for three-body systems, The European Physical Journal D 72, 52 (2018).
  122. [122]N. Moiseyev, Quantum theory of resonances: Calculating energies, widths and cross-sections by complex scaling, Phys. Rep. 302, 211 (1998).
  123. [123]P. A. Belov, Linewidths and energy shifts of electron-impurity resonant states in quantum wells with infinite barriers, Phys. Rev. B 105, 155417 (2022).
  124. [124]H. A. Bethe and E. E. Salpeter, Quantum Mechanics of One- and Two-Electron Atoms (Dover Publications, 1957).
  125. [125]L. Happ and P. Naidon, Stabilization of three-body resonances to bound states in a continuum, Phys. Rev. A 113, 032203 (2026).arxiv.org/abs/2503.02037
  126. [126]N. Scheuler, J. Main, P. Rommel, F. Pfeiffer, S. Scheel, and P. A. Belov, Impact of the valence band on Rydberg excitons in cuprous oxide quantum wells, Phys. Rev. B 113, 115413 (2026).arxiv.org/abs/2511.20441
  127. [127]N. S. Rytova, The screened potential of a point charge in a thin film, Moscow Uni. Phys. Bull. 22, 18 (1967) [Vestnik Moskovskogo Universiteta. Fiz. 22, 30 (1967)].
  128. [128]L. V. Keldysh, Coulomb interaction in thin semiconductor and semimetal films, JETP Lett. 29, 658 (1979), [Pis’ma Zh. Eksp. Teor. Fiz. 29, 716 (1979)].

Paper details

vlnc:W_pxvtpxs7
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents