Lightmass Bragg cavity polaritons in planar quantum dot lattices
Abstract
The excitonpolariton modes of a quantum dot lattice embedded in a planar optical cavity are theoretically investigated. Umklapp terms, in which an exciton interacts with many cavity modes differing by reciprocal lattice vectors, appear in the Hamiltonian due to the periodicity of the dot lattice. We focus on Bragg polariton modes obtained by tuning the exciton and the cavity modes into resonance at high symmetry points of the Brillouin Zone. Depending on the microcavity design these polaritons modes at finite inplane momentum can be guided and can have long lifetimes. Moreover, their effective mass can be extremely small, of the order of ( is the bare electron mass), and they constitute the lightest excitonlike quasiparticles in solids.
pacs:
71.36.+c, 78.20.Bh, 78.67.HcI Introduction
The observation of vacuum Rabi oscillations between a quantum dot (QD) exciton and a cavity photon represents a distinctive example of cavity quantum electrodynamics realized in a solid state system. Reithmaier et al. (2004); Yoshie et al. (2004); Peter et al. (2005) The strong lightmatter coupling in networks of cavities and dots is very attractive for the realization of scalable quantum information devices. Yao et al. (2005) Furthermore, fundamental investigations on quantum phase transitions in extended cavity systems have recently attracted a significant interest. Hartmann et al. (2006); Hartmann and Plenio (2007); Greentree et al. (2006) The strong lightmatter coupling between excitons and photons is often described in terms of polariton modes, which identify propagating electromagnetic modes in a medium with strong dispersive properties. Lately, there has been a considerable theoretical and experimental effort on polariton condensation in planar microcavities. Deng et al. (2003); Kasprzak et al. (2006); Balili et al. (2007); Keeling et al. (2007) In the simplest picture, the critical temperature for polariton condensation is inversely proportional to the polariton effective mass. Given their small effective mass (about four orders of magnitude smaller than excitons), quantum well polaritons suggest the possibility of creating a condensate at high temperature. Therefore, it is important to investigate diverse geometrical realizations of polaritonic structures, where the photon and matter excitations are confined in different ways. These diverse geometries could lead both to novel quantum information devices and to novel systems in which coherent matter states can be obtained at room temperature.
In this paper, we exploit the newly gained opportunity of growing artificial structures with spatial periodicity comparable to the wavelength of the confined quasiparticle, the exciton in our case, and their embedding in a microcavity. The main idea consists in tuning the exciton energy in such a way that at least two photon modes with different momentum are resonantly coupled to it due to the lattice symmetry. We assume that the excitonphoton coupling is stronger than the exciton inhomogeneous broadening. There has been many proposals for engineering excitonpolariton modes using periodic structures: with periodic quantum well (QW) Bragg structures,Mintsev et al. (2002); Hayes et al. (1999) cavityfree three dimensional arrays of QDs,Ivchenko et al. (2000) pointdipole crystal,Kempa et al. (2005) photonic bandgaps with antidots, Gerace and Andreani (2007) and confined QW polaritons in mesa structures. Kaitouni et al. (2006) Here we consider a system with a mismatch between the continuous symmetry of the twodimensional photons in the cavity and the discrete symmetry of a quantum dot lattice. In the proposed structure, the polariton dispersion is entirely determined by the coupling between the exciton in the dots and two dimensional photons, which leads to inplane effective masses which can be exceptionally small, of order of , for some highly symmetric points in the first Brillouin zone. The group velocity of 2D photons is large for large inplane momenta. This can be seen as the underlying reason behind such a small polariton mass at the zone edge. The extremely small mass makes this structure a promising candidate for novel approaches to high temperature polariton condensation, and for longrange coupling of spin Quinteiro et al. (2006)or exciton Tarel et al. (2007) qubits.
This paper is organized as follows: In Sec. II we introduce the system and write the Hamiltonian for the QD lattice interacting with the twodimensional photon modes. The energy dispersion of the polariton states for some highly symmetric points in the Brillouin zone of the square and hexagonal lattice is presented in Sec. III. In the same section we also discuss the excitonlight mixing, the effective masses, and the polariton lifetime. Conclusions are found in Sec. IV. In the Appendix we provide the explicit form of the excitonphoton coupling constant, we discuss an optimization procedure for the system design, and we give the exact effective mass expressions.
Ii Polariton Hamiltonian
We consider the system shown in Fig. 1. A twodimensional lattice of QDs or impurities is placed at the center of a planar cavity structure, i.e. a FabryPérot resonator. There is currently an intense experimental effort for controlling the growth of QDs, based on selfassembly and growth on patterned substrates. Watanabe et al. (2005); Badolato and Petroff () Controlled impurity implantation or micropatterned electrostatic traps on a quantum well Gärtner et al. (2007); Lai () are further options to realize the proposed structure. We assume that the array consists of identical QDs and we use the effective mass, envelope function, and single sublevel approximations. Bastard (1992) Furthermore, taking into account that the confinement potential of a QD can be modelled by a harmonic potential, Kumar et al. (1990) which enables the centerofmass and relative motion separation,Sugawara (1995, 1997) we write the ground state heavy hole exciton wave function in the QD centered at in the lattice plane as
(1) 
where is the electron (hole) confinement wave function, is the exciton relative motion wave function with being exciton Bohr radius, and is the exciton centerofmass wave function
(2) 
where is the effective dot radius. By considering only the ground state we assume that the energy of the first excited state is larger than the excitonphoton coupling, which is of the order of a few meV.
We assume that the photon modes are confined by ideal planar mirrors in a cavity of length and we consider only the lowest cavity mode with the photon dispersion
(3) 
where is the speed of light, is refractive index of the cavity, is fixed to , and q is the inplane momentum. Finally, the inplane excitonphoton or polariton Hamiltonian in the second quantization reads
(4)  
where is the exciton energy, () is the QD exciton creation (annihilation) operator with polarization , () is the photon creation (annihilation) operator for a given polarization . The constant is the excitonphoton coupling constant for the th QD, which has the property
(5) 
In the following we are going to focus on TE cavity modes, and we consider only one exciton polarization. For quantum well polaritons, TM modes give smaller polariton effects due to the weaker lightmatter coupling, Savona et al. (1999) and we expect the same to be true in the lattice case. The lattice symmetry can be exploited by introducing new exciton operators
(6) 
After some algebra, the polariton Hamiltonian can be written as
(7)  
where is now restricted to the first Brillouin Zone ( BZ) due to the periodicity of the QD lattice and is a reciprocal lattice vector . The coupling constant expressed in terms of materials and structure parameters is given in Appendix A.
Moreover, the lightmatter interaction conserves the momentum only up to a reciprocal lattice vector. The terms in Eq. (7) involving the exchange of reciprocal lattice vectors are known as exciton Umklappprocesses in the literature. Bullough and Thompson (1970); Knoester and Mukamel (1989); Kempa et al. (2005) This is in contrast to the QW microcavity case where there is a onetoone correspondence between cavity and exciton modes (the inplane momentum is conserved exactly). They can be seen as interband scattering in which a valence electron with momentum scatters with a photon with momentum and creates a conduction electron . If the final momentum lies outside the BZ, then the momentum of the conduction electron is flipped back to the first BZ by adding a reciprocal lattice vector. Since the photon momentum is not restricted to the BZ, the Umklapp momentum Q for excitons can have an arbitrary large magnitude. This is a difference with respect to electronphonon Umklapp processes, Ashcroft and Mermin (1976) where only the first neighbor reciprocal lattice points are involved.
However, we remark that in Eq. (7) the quantum dotphoton coupling has a structure form factor due to the finite size of the quantum dots, i. e.
(8) 
so that we find a natural cutoff for the reciprocal lattice vectors. Furthermore, in the photonic case, excitonic corrections due to the Coulomb interaction between the conduction electron and the hole left in the valence band have to be included.
For an exciton state with momentum q away from the zone boundary, many offresonant Umklapp terms give corrections which do not entail qualitative novel properties for the polariton quasiparticles. However, the situation can be very different if the structure is built in such a way that the exciton is resonant with the cavity at a which is at or near the BZ boundary. In that case we can choose, for instance, and so that the Bragg condition
(9) 
is satisfied. In general, the number of reciprocal lattice points satisfying this condition depends on the lattice symmetry and on the energy of the exciton. The mode configuration leading to the formation of Bragg cavity polaritons is illustrated in Fig. 2, where the exciton energy (red dashed horizontal line) is tuned in such a way that the polariton mixing occurs at the zone boundary. Due to the reduced symmetry of the dot lattice the cavity mode is folded giving rise to many photonic branches, each characterized by a different reciprocal lattice vector . In order to visualize this folding, the figure shows two repeated BZs. Bragg polariton modes will appear at where two or more photonic branches are crossing and are nearly degenerate with the exciton energy. These polaritons will have different properties than polaritons usually investigated at . Fig. 3 shows the high symmetry points at the zone boundary for the square and hexagonal lattice for which we will discuss Bragg polaritons in the following.
Due to the cutoff introduced by the finite quantum dot size, we can write the single polariton Hamiltonian at fixed q in the form
(10) 
Furthermore, the finite size of the dots limits the largest reciprocal lattice vector to . If we keep only the resonant terms at the zone boundary satisfying exactly the Bragg condition, the matrix Eq. (10) can be diagonalized analytically and we find the two eigenvalues
(11) 
corresponding to the strongly mixed polariton states, as well as the ()fold degenerate eigenvalue
(12) 
The normal mode splitting (Rabi energy), i. e. the difference between the lowest (lower polariton) and the highest (upper polariton) energy at resonance, is then
(13) 
and is proportional to the square root of the number of reciprocal lattice vectors involved in the Bragg condition.
In the square lattice, we focus on Bragg polaritons obtained at the and high symmetry points defined in Fig. 3 (a). The number of photonic branches that can be brought into resonance is determined by the number of points in the reciprocal lattice equidistant to or , according to Eq. (9). There are two nearest equidistant points to () and four nearest equidistant points to () (see dots in Fig. 4). However, if we do not concentrate on the lowest photonic branches and the nearestneighbors in the reciprocal lattice, but e.g. on the second nearest neighbors, then a larger number of reciprocal lattice vectors can be involved. For instance, there are eight secondnearest neighbor equidistant points to (, see crosses in Fig. 4). In the following we will label the polariton states as if and if . For the hexagonal lattice (see Fig. 3(b)) we restrict our discussion to the point, which has . The point is similar to the point of the square lattice and will not be discussed. Since it is possible to express explicitly the coupling constant for a given high symmetry point of length (see Eq. (27) in Appendix B). Then the normal mode splitting can be written as
(14) 
where is the coupling constant at the point. The values of and can then be used as scaling parameters to calculate the strength of polaritonic effects at different symmetry points, and are summarized in Tab. 1.
1  
2  4  8  3 
ii.1 Effective mass at Xpoint
One of the most interesting features of Bragg polaritons consists in their extremely small effective mass, which is a consequence of the fact that twodimensional photons at large momentum enter in the polariton formation. In this section, we investigate this property analytically for a Bragg polariton at the point. The photon energy in the vicinity of the point can be expanded to the second order in the inplane momentum as
where is the point photon energy, and with . Taking only into account the exciton modes and the two photonic branches nearly resonant in the vicinity of the point, the full Hamiltonian in Eq. (10) matrix can be simplified as
where the zero of the energy is set at . The eigenvalues of this matrix can be found analytically by solving a cubic equation. From the expansion of the eigenvalues to the second order in and we obtain
(15)  
(16) 
where the , , and index identify the upper, central, and lower polariton branches, respectively. The upper (lower) sign refers to UP (LP), and
(17) 
is the renormalized coupling constant. From the definition of the effective mass tensor it follows for the upper and lower polariton branches
(18) 
In these expressions, we have neglected the weak dependence of the coupling constant on the momentum . This dependence gives small corrections to the effective mass and is discussed in Appendix C. Since typically , and , the effective mass ratio reads
(19) 
Therefore, the polariton mass is very anisotropic at the point. We can also compare the value of these effective masses with the QW polariton mass, which is of the same order of magnitude of the cavity photon mass at
(20) 
and typically . Then, we remark that for the the mass along the direction we have and . This implies that a value of the order of is expected. Moreover, we will see in the next section that isotropic masses of such a order can be obtained at special high symmetry points.
Iii Results and discussion
(nm)  121  300  383  161 

(nm)  118  167  373  157 
(meV)  2.2  2.2  1.4  2.0 
(meV)  8.8  8.8  5.6  8.0 
In this section we present the numerical results for three high symmetry points of the square and hexagonal lattice. The numerical calculations include all resonant and off resonant terms. We have taken parameters from AlGaAs/GaAs and CdSe/ZnSe systems, which are typical examples of IIIV and IIVI quantum dot structures. In general, IIVI semiconductors microcavities have a stronger lightmatter coupling, e.g. for CdTe microcavity the lightmatter coupling is meV,André et al. (1998) while for IIIV systems the coupling constant is meV.Savona et al. (1995) The cavity length, dot size, and lattice constant can be optimized in order to maximize the lightmatter coupling for Bragg polaritons at a given high symmetry point. The optimization procedure is discussed in the Appendix B and the parameters are given in Tab. 2.
Fig. 5 () shows contour plots of the energy dispersion of the upper polariton branch for Bragg polaritons at the point (), point ( for polariton and for polariton) of the square lattice, and point () of the hexagonal lattice. The energy dispersion near the resonance reflects the symmetry of these points. We can also see that the polariton effective mass is highly anisotropic for the polariton, as discussed in the previous section, while the effective mass for is almost isotropic. In Fig. 6 we show the energy dispersion of the Bragg polariton modes for polaritons, polaritons, and polaritons. The different behavior of the UP, CP, and LP modes at different points shows how richer Bragg polaritons are compared to the QW polaritons. Note that the total number of branches crossing or anticrossing at the different symmetry points is always given by .
At the point, we have and consequently there are three polariton modes. In the direction, the upper and lower polaritons are similar to the QW polariton (see Fig. 6b). However, there is one additional central mode degenerate with the photon mode along the zone boundary in the direction. On the contrary, the dependence along shows the existence of (i) a lower branch with a negative effective mass in one direction, (ii) an upper branch with positive effective masses but high anisotropy, and (iii) a flat branch with a large mass (consequence of the infinite exciton mass considered in our model) with energy exactly between the lower and upper branch.
Let us now look at the polaritons, which have five branches. In the direction we can see nearly flat photonic branches originating from photon modes with and , while the polariton mixing occurs mainly with photon modes having and . A similar plot for (not shown) gives nine polariton branches. Notice that there is not a saddle point at the point. Polariton mode with isotropic negative mass (lower branch) can give rise to an effective negative index of refraction. Agranovich and Gartstein (2006) In the point case there are four branches and the properties are similar to the point of the square lattice.
In the case of QW polaritons, the exitonic component is equally shared between lower and upper polaritons at the anticrossing. This is also true for the excitonic component in the point as shown in Fig. 7 ( and ). At the anticrossing, the UP and LP branches consist of half exciton and consequently, the excitonic component is completely absent in the CP branch. Nevertheless, it increases rapidly away from the anticrossing region. This excitonphoton character swap is a characteristic of the Bragg polaritons since in all the cases investigated we have found analytically and numerically that the excitonic component is always equally shared between the lower and upper polaritons at the anticrossing, the rest being purely photonic.
GaAs/AlGaAs  CdSe/ZnSe  

3.7  4.7  8.0  1.0  
8.3  4.7  6.2  1.0  
3.7  4.7  8.0  1.0  
8.3  4.7  6.1  1.0 
The effective masses calculated numerically for the two material systems are given in Table 3. As expected a strong anisotropy at the point and almost isotropic mass at point are found. The numerical values for the point are the same as calculated using the analytical formula Eq. (II.1), thus validating our numerical approach. Furthermore, the ratio in Eq. (19) for both AlGaAs/GaAs and CdSe/ZnSe systems, for which and , respectively (see Appendix B for materials parameters) is also obtained. The value of the effective mass in some cases is of the order of and is extremely small for a matter like quasiparticle. As discussed in the Introduction, the small effective mass could suggest new strategies for the realization of polariton condensation in microcavities. However, the dynamics of Bragg polaritons, including the phonon bottleneck effects,Tassone et al. (1997) and polaritonpolariton scattering is expected to be different than for quantum well polaritons, and needs further investigations. The small polariton effective mass can play an important role also in quantum information implementations. It has been recently shown that QW polaritons can mediate a spinspin coupling between two charged dots or impurities over a distance of the order of hundreds of nanometers due to the light mass of polariton.Quinteiro et al. (2006) Since Bragg polariton states have effective masses that are two to three orders of magnitude smaller, the range of the spin coupling could be further extended.
So far, the lifetime of Bragg polaritons has not been discussed. There is an essential difference with respect to the QW polariton case, where the light matter mixing occurs close to . The quantum well polariton lifetime depends on the photon lifetime in the cavity, which is determined by the reflectance of the mirrors in the normal direction and is of the order of picoseconds. However, at finite q, depending on the external index of refraction and on the propagation angle , a cavity photon can either be transmitted through the mirrors, or it can remain confined due to total internal reflection. This means that structures can be engineered where the polariton mixing occurs at large q with guided modes, which is never possible in the quantum well case. Since guided photons have in principle a much longer lifetime, the polariton lifetime will also be much longer, likely in the order of micro or milliseconds, depending on the properties of the lateral surfaces of the cavity structures. However, the details of the cavity modes at large inplane momentum, in particular leaky modes Tassone et al. (1996) due to the distributed Bragg mirrors, need to be carefully taken into account, since they will affect both the dynamics and the radiative properties of these polaritons.
Iv Conclusions and outlook
We have investigated polariton modes in a planar quantum dot lattice embedded in a microcavity. Assuming small exciton and photon homogeneous broadening, we have considered systems in the strong coupling regime. We have studied Bragg polariton modes formed at high symmetry points of the Brillouin Zone. Truly polaritonic quasiparticles with a fiftyfifty lighmatter character appear at these points in upper and lower branches. Analytical and numerical calculations of the polariton effective masses give extremely small values, of the order of , and in some cases strong anisotropy. The small mass is due to the mixing of excitons with nearly linear twodimensional photon modes at large momenta. There is a full transfer of the excitonic wavefunction component from the flat exciton dispersion to photonic branches in the strong coupling region. We have also discussed the possibility of polariton coupling with guided modes, which could lead to long polariton lifetime.
Future extensions of this work could include: (i) an explicit theoretical treatment of the homogeneous broadening with a discussion of the strong to weak coupling transition, (ii) realistic structure modelling of the distributed Bragg mirrors with leaky modes, which could be close to the Bragg polariton resonances, and (iii) sensitivity to inhomogeneous broadening effects. The polaritonphonons, polaritonpolaritons, and polaritonspin dynamics for these Bragg quasiparticles also needs further investigation, since it is expected to be very different than in the quantum well case.
Acknowledgements.
This research was supported by the National Science Foundation, Grant No. DMR0608501.Appendix A Coupling constant
The lattice excitonphoton coupling in the cavity for the TE mode readsSavona et al. (1999)
(21) 
where is the volume of the cavity, being the surface of the cavity, is the dipole matrix element between conduction and valence band, is the number of dots in the cavity, is the centerofmass form factor
and is the form factor
We can assume that the QW that contains the quantum dots is much narrower than the cavity length , which implies . The coupling constant for TM modes
(22) 
is always smaller than for TE modes at finite inplane momentum. We can relate the QD coupling constant in Eq. (21) to the equivalent quantity in the QW case as
(23) 
where is a QW exciton energy, is the QD density, () is a Bohr radius of the QW (QD) exciton. We note that recent numerical calculations show that for confined excitons the effective Bohr radius defined by the ansatz of Eq. (1) is typically smaller that the quantum well value. Grochol et al. (2005a, b) The quantity can be understood as a filling or packing factor. For a fixed dot size and exciton energy the hexagonal lattice has the largest packing factor and the largest coupling at . The value of the interband matrix element can be calculated conviniently from the Kane energy parameter using the relationRosencher and Vinter (2002)
(24) 
In order have a strong light matter coupling, a system with a large dipole matrix element and with small QD Bohr radius is desired. From this point of view IIVI semiconductors are better candidates than IIIV. Our scheme could be also implemented with impuritybound excitons, which usually have very strong oscillator strength, large spatial extension, and very small inhomogeneous broadening. Stébé et al. (1996) The materials parameters used in the calculations are given in Tab. 4.
Appendix B Parameter optimization
GaAs/AlGaAs  CdSe/ZnSe  
(eV)  1.68 
2.50 

(eV)  1.6  2.4 
(nm)  11  3 
(nm)  13  6 
(meV)  1.2 
2.7 
3.5 
2.5 

(eV)  23 
21 
(nm)  35  35 
(m)  2.0  2.0 
Let us now fix the QD size and ground state exciton energy. The lattice constant and cavity length are then adjustable parameters which can be optimized to increase the polariton mixing. The coupling constant at a high symmetry point with can be written as
(25) 
where is a constant which does not depend on , , and . Note that the dependence on the lattice constant is nonmonotonic due to the competition between the quantum dot density and the structure factor. Moreover, the excitonphoton Bragg resonance condition gives the constraint
(26) 
In order to take into account that distributed Bragg reflectors are typically used as mirrors, we have added an effective to the nominal cavity spacer length in the coupling constant. Savona et al. (1995) By expressing as a function of from Eq. (26), the coupling constant becomes a function of only one variable . The optimal lattice spacing can then be found by analytical or numerical maximization. Defining , we obtain from Eq. (25)
(27) 
Thus one needs to optimize only for one high symmetry point of each material combination and then rescale it by for other points.
Appendix C Effective Mass Corrections
Here we include the corrections to the effective mass due to the dependence of the coupling constant on q. This provides the exact second order expansion in and of Bragg polaritons and therefore the exact expression for their effective mass tensor. Only close to the point is considered here. The coupling constant written as a function of small around the point reads
Using and the expansion of the exponential function we obtain
where
The exact expansion for the polaritonic modes is therefore
(28) 
The effective masses for the UP(LP) can be written as
(29) 
while for the CP we have
(30) 
The upper and lower polaritons small mass in the direction is mainly due to the term and the q dependence of the coupling constant provides only small corrections. The numerical calculations of the effective mass given in Table 3 include the explicit dependence of the coupling constant on the wavevector.
Footnotes
References
 Present Address: MaxPlanck Institute for Quantum Optics, Garching, Germany.
 J. P. Reithmaier, G. Sȩk, A. Löffler, S. K. C. Hofmann, S. Reitzenstein, L. V. Keldysh, V. D. Kulakovskii, T. L. Reinecke, and A. Forchel, Nature 432, 197 (2004).
 T. Yoshie, A. Scherer, J. Hendrickson, G. Khitrova, H. M. Gibbs, G. Rupper, C. Ell, O. B. Shchekin, and D. G. Deppe, Nature 432, 200 (2004).
 E. Peter, P. Senellart, D. Martrou, A. Lemaitre, J. Hours, J. M. Gérard, and J. Bloch, Phys. Rev. Lett. 95, 067401 (2005).
 W. Yao, R. B. Liu, and L. J. Sham, Phys. Rev. Lett. 95, 030504 (2005).
 M. J. Hartmann, F. G. S. L. Brandão, and M. Plenio, Nature Physics 2, 849 (2006).
 M. J. Hartmann and M. B. Plenio, Phys. Rev. Lett. 99, 103601 (2007).
 A. D. Greentree, C. Tahan, J. H. Cole, and L. C. L. Hollenberg, Nature Physics 2, 856 (2006).
 H. Deng, G. Weihs, D. Snoke, J. Bloch, and Y. Yamamoto, PNAS 100, 15318 (2003).
 J. Kasprzak, M. Richard, S. Kundermann, A. Baas, P. Jeambrun, J. M. J. Keeling, F. M. Marchetti, M. H. Szymanska, R. André, J. L. Staehli, et al., Nature 443, 409 (2006).
 R. Balili, V. Hartwell, D. Snoke, L. Pfeiffer, and K. West, Science 316, 1007 (2007).
 J. Keeling, F. M. Marchetti, M. H. Szymanska, and P. B. Littlewood, Semic. Sci. Tech. 22, R1 (2007), and references therein.
 A. V. Mintsev, L. V. Butov, C. Ell, S. Mosor, G. Khitrova, and H. M. Gibbs, JETP Lett. 76, 637 (2002).
 G. R. Hayes, J. L. Staehli, U. Oesterle, B. Deveaud, R. T. Phillips, and C. Ciuti, Phys. Rev. Lett. 83, 2837 (1999).
 E. L. Ivchenko, Y. Fu, and M. Willander, Phys. Solid State 42, 1756 (2000).
 K. Kempa, R. Ruppin, and J. B. Pendry, Phys. Rev. B 72, 205103 (2005).
 D. Gerace and L. C. Andreani, Phys. Rev. B 75, 235325 (2007).
 R. I. Kaitouni, O. E. Daïf, A. Baas, M. Richard, T. Paraiso, T. G. P. Lugan, F. MorierGenoud, J. D. Ganiére, J. L. Staehli, V. Savona, et al., Phys. Rev. B 74, 155311 (2006).
 G. F. Quinteiro, J. FernándezRossier, and C. Piermarocchi, Phys. Rev. Lett. 97, 097401 (2006).
 G. Tarel, G. Parascandolo, and V. Savona, arXiv:0708.1920 (2007).
 S. Watanabe, E. Pelucchi, K. Leifer, A. Malko, B. Dwir, and E. Kapon, Appl. Phys. Lett. 86, 243105 (2005).
 A. Badolato and P. M. Petroff, to be published.
 A. Gärtner, L. Prechtel, D. Schuh, A. W. Hilleitner, and J. P. Kotthaus, condmat/07052428 (2007).
 C. W. Lai, unpublished.
 G. Bastard, Wave mechanics applied to semiconductor heterostructures (Les éditions de physique, Paris, 1992).
 A. Kumar, S. E. Laux, and F. Stern, Phys. Rev. B 42, 5166 (1990).
 M. Sugawara, Phys. Rev. B 51, 10743 (1995).
 M. Sugawara, Jpn. J. Appl. Phys. 36, 2151 (1997).
 V. Savona, C. Piermarocchi, A. Quattropani, P. Schwendimann, and F. Tassone, Phase Transitions 68, 169 (1999).
 R. K. Bullough and V. B. Thompson, J. of Phys. C 3, 1780 (1970).
 J. Knoester and S. Mukamel, J. Chem. Phys 91, 989 (1989).
 N. W. Ashcroft and N. D. Mermin, Solid State Physics (Brooks/Cole,Thomson Learning, Cornell, 1976).
 R. André, D. Heger, L. S. Dang, and Y. M. dâAubigne, J. Cryst. Growth 185, 758 (1998).
 V. Savona, L. C. Andreani, P. Schwendimann, and A. Quattropani, Solid State Comm. 93, 733 (1995).
 V. M. Agranovich and Y. N. Gartstein, PhysicsUspekhi 49, 1029 (2006).
 F. Tassone, C. Piermarocchi, V. Savona, A. Quattropani, and P. Schwendimann, Phys. Rev. B 56, 7554 (1997).
 F. Tassone, C. Piermarocchi, V. Savona, A. Quattropani, and P. Schwendimann, Phys. Rev. B 53, R7642 (1996).
 M. Grochol, F. Grosse, and R. Zimmermann, J. Lumin. 112, 208 (2005a).
 M. Grochol, F. Grosse, and R. Zimmermann, Phys. Rev. B 71, 125339 (2005b).
 E. Rosencher and B. Vinter, Optoelectronics (Cambridge University Press, Cambridge, 2002).
 B. Stébé, E. Assaid, S. L. Goff, and F. Dujardin, Solid State Comm. 100, 217 (1996).
 D. Gammon, E. S. Snow, B. V. Shanabrook, D. S. Katzer, and D. Park, Phys. Rev. Lett. 76, 3005 (1996).
 I.C. Robin, R. André, and J.M. Gérard, Phys. Rev. B 74, 155318 (2006).
 P. Borri, J. R. Jensen, W. Langbein, and J. M. Hvam, phys. stat. sol. (b) 221, 143 (2000).
 O. Madelung, Semiconductors: Data Handbook (Springer, New York, 2003).
 M. Cardona, J. Phys. Chem. Solids 24, 1543 (1963).