Since the revolution caused by the transistor and integrated circuits, the development of low-dimensional structures has been continually optimized for integration in on-chip architectures. Low-dimensional systems like quantum wells (two-dimensional, 2D), quantum wires (one-dimensional, 1D) and quantum dots (zero-dimensional, 0D) produced from different semiconductor materials have been used as building blocks for the development of mesoscopic devices^[1]. Heterostructures such as quantum wells have been explored as a template to control spin degree of freedom of electrons and achieve development of spintronic devices^[2] and quantum dots are potential candidates for quantum information technology^[3]. Most of these nanostructures have been fabricated employing group Ⅲ−Ⅴ semiconductors^{[4, 5]}. However, low-dimensional quantum nanostructures of the Ⅱ–Ⅵ group of semiconductors have recently attracted attention to develop quantum devices based on CdTe/(Cd, Mg)Te quantum wells (QWs)^[6] and self-assembled Cd(Se, Te)/ZnTe quantum dots (QDs)^[7]. Optical control of the electron spin coherence has been demonstrated in CdTe/(Cd, Mg)Te QWs^[8]. Spin polarization effects in negatively-charged and neutral excitons have been explored for the detection for spin injection of polarized electrons^[9] and long-living hole spin coherence in undoped CdTe QWs has been observed^[10]. Photon echo-based technique was used to study the effective in-plane hole g-factor to understand the anisotropy of the hole spin states^[6]. More recently, a theoretical approach has proposed ZnTe and CdTe monolayers as promising candidates for a new generation of spintronic devices^[11].

Thin films, QWs and QDs from Ⅱ–Ⅵ compounds for different fields of application have been produced using several substrates to achieve structures with high crystalline quality. CdTe, CdZnTe, and GaSb substrates were used for the development HgCdTe infrared detectors^[12−14]. The main problems encountered in these substrates are their high cost and limited sizes. Other substrates commonly used with more affordable cost are GaAs and Si. Many structures have been fabricated on GaAs substrates like CdTe quantum wells for investigating the persistent spin helix phenomena^[15] in the field of spintronics, CdTe/ZnTe heterostructures and CdTe nanocrystalline for the development of solar cell technologies^{[16, 17]}. Basic studies on the growth of CdTe self-assembled QDs^[18] as well as CdTe diode-type device optimization for X-ray detectors^[19] on Si substrate have also been explored. However, taking into account the values of the bulk lattice constants at 300 K for CdTe (6.482 $\text{A}$), GaAs (5.653 $\text{A}$) and Si (5.431 $\text{A}$), we observe that CdTe has a lattice mismatch of 14.7% with GaAs and 19.3% with Si. The high crystalline quality of heteroepitaxial growth with such large lattice mismatches is still a limiting factor to the improvement of the performance of electronic and optoelectronic devices using Ⅱ–Ⅵ compounds. An usual approach used to overcome the lattice mismatch problem is to grow a ZnTe or ZnSe buffer layer on GaAs^{[20, 21]} and Si substrates^[22]. As we have a lattice constant at 300 K for ZnTe of 6.101 $\text{A}$, the lattice mismatch is 7.9% with GaAs and 12.3% with Si. This reduces the incompatibility between the crystal structures and allows to improve the quality of the materials as shown for Cd(Se,Te)^[7] and CdZnTe QDs^[23] grown on GaAs substrate with a ZnTe buffer layer. Otherwise, these approaches require a more detailed growth process^{[7, 22]}. On the other side, it has been observed that good structural quality CdTe nanostructures^[18] and CdMnTe thin films^[24] can be grown directly on hydrogen passivated Si(111) substrates. However, to our knowledge, the optical properties of CdMnTe/CdTe heterostructures grown on silicon(111) have not been studied so far.

In this work, we investigate the optical response of quantum emitters in CdMnTe/CdTe/CdMnTe heterostructures grown directly (without any buffer layer) on Si(111) substrates. In addition to the heterostructure showing a usual 2D character behavior, the surface roughness arising from the large lattice mismatch favors the appearance of localized states exhibiting quantum-dot like emission spectra. Indeed, it is surprising that structures grown with lattice mismatch of almost 20% emerge as potential optical materials. We show that the high roughness which originates from the growth of CdTe or CdMnTe on Si(111) substrate can be used to form Ⅱ–Ⅵ QDs through a simple and low cost epitaxial growth.

Cd_1−xMn_xTe/CdTe/Cd_1−xMn_xTe (x = 0.31) heterostructures were grown directly on Si(111) substrates in a home-built molecular beam epitaxy system. Mn concentration in the barrier was estimated using high-resolution X-ray diffraction as reported previously^[24]. High purity polycrystalline CdTe and Mn were evaporated from single effusion cells. The Si(111) substrates were dipped into 2% HF/H₂O solution for 2 min immediately before growth in order to have a hydrogen passivated surface. The substrate temperature was 400 °C. For clarity, we identify the three heterostructure (H) samples used in this work as H1, H2 and H3. H1 has a 15 nm nominal thick layer of CdTe between 90 nm nominal thick layers of CdMnTe. H2 and H3 have 30 nm nominal thickness of CdMnTe barriers and 9 and 3 nm nominal thick layers of CdTe, respectively. It has been observed^[25] that the surface roughness of CdMnTe layers grown on silicon increases with thickness. Therefore, a larger upper barrier, and also a lower one, was used in sample H1, for which the CdTe layer is much thicker than for H2 and H3.

Scanning electron microscopy (SEM) images and thin lamellas were prepared using a Thermofisher Helios Nanolab 650. The SEM images were collected in immersion mode and 5 kV for a better image quality of the surface. The lamellas were final polished using 2 kV accelerating potential in order to reduce sample damage. The transmission electron microscopy (TEM) images were made using a probe corrected Thermofisher Titan 80−300 working at 200 kV. The images were collected using high-angle annular dark-field detector (HAADF) in scanning transmission electron microscopy (STEM) mode. Chemical maps were performed using energy dispersive spectroscopy (EDS) also collected in STEM imaging mode.

Macro photoluminescence (PL) measurements were performed using a 520 nm solid state laser line and a T64000 system with CCD from Horiba/Jobin Yvon operating in single spectrometer mode. Micro PL (μ-PL) was performed using a 100× optical objective and a 457 nm solid state laser as excitation source. The emitted light was collected by a 55 cm single monochromator with CCD system from Horiba. For polarization measurements, a broadband quarter wavelength plate was used to generate circularly polarized incidence on the samples and a polarization displacement prism was used to simultaneously detect the right (σ+) and left (σ−) circularly polarized PL components. In macro and micro measurements, the samples were cooled down to low temperatures using a He cold finger cryostat.

Fig. 1(a) shows the SEM image of the CdMnTe surface morphology of sample H1. The observed triangular formations are typical for a cubic structure with large roughness^{[18, 25, 26]}. The growth takes place by the coalescence of 3D islands which nucleate having a high in-plane mosaicity. This leads to a large density of defects during coalescence and the resultant thin films show a considerable surface roughness which depends on the growth temperature and time. The heterostructure grows epitaxially following the (111) orientation of the silicon substrate as can be observed through HRTEM image in the Fig. 1(b). However, there are defects at the interface due to the lattice mismatch as already discussed^[25]. In Fig. 1(c), the STEM image clearly shows the CdTe layer between the layers of CdMnTe forming a heterostructure, as indicated by yellow arrows. Figs. 1(e) and 1(g) show the distribution of the elements Te, Cd, and Mn obtained by EDS. It confirms the formation of a heterostructure with a Cd rich layer between Mn rich barriers. Traditional systems such as GaAs/AlGaAs^[27] and InGaAs/InP^[28] have roughness of a few monolayers. The lattice mismatch for typical heterostructure is very small (<1% for AlGaAs/GaAs system). Our heterostructure has a lattice mismatch of 19.3% and the observed surface roughnes can be as high as 50 nm depending on growth time and temperature^[24].

Fig. 2(a) illustrates the structure expected for the CdTe/CdMnTe when the CdTe layer is thick. For a sufficiently small CdTe thickness, the growth shall lead to structures with three-dimensional confinement, as shown in Fig. 2(b). As will be shown later by the μ-PL measurements, this is possibly the case of sample H3. Fig. 2(c) shows the low temperature (T = 8 K) normalized macro-PL spectra of samples H1, H2, and H3. The most prominent emission in sample H1 is observed around 1.61 eV which is close to what would be expected for a CdTe/CdMnTe QW of the same thickness and can be attributed to the electron−heavy hole recombination in the system^[29]. Recombinations in CdTe QW associated with the heavy hole^[29] transition is treated as stronger in CdTe(111) than CdTe(100) quantum wells^[30]. In addition to the main peak, there is a broad band centered around 1.48 eV. The region between 1.40−1.50 eV is usually assigned to defects^{[31, 32]} and bound exciton recombination. In particular, the transition at 1.48 eV has been associated with exciton recombination from bound states due to dislocations induced by the large lattice mismatch growth^[33]. The shoulder above the main peak (highest energy) in H1 spectrum is associated to a non-homogeneous emission band which appears due to confinement and the strong roughness of the CdTe film in sample H1, as will be discussed later. The blueshift of the spectra in Fig. 2(c) is consistent with the reduction of the CdTe layer thickness from sample H1 to H3. Moreover, the (full width at half maximum) FWHM increases with the reduction of the CdTe layer thickness reflecting an increase in the inhomogeneity of the confinement potential which act as carrier localization centers (also discussed in the sequence). It is also interesting to note that despite the large lattice mismatch and the expected high density of defects, we still observe a reasonable emission at relatively low power densities for H1 and H3 (1.4 and 2.8 W/cm², respectively) as shown in Figs. 2 (d) and 2(e).

The Fig. 3(a) shows the integrated PL emission for sample H1 as a function of the laser excitation power. Black filled circles are associated with the 1.61 eV emission and the black filled triangles with the 1.48 eV one. The results were obtained using Voigt functions for the fitting. In Fig. 3(b), we show the power dependence of the main emission for sample H3 at 1.79 eV (from here on we will concentrate on the results obtained for samples H1 and H3). The origin of the recombination mechanism can be identified by the dependence between the integrated PL intensity (I) and the excitation laser power (P). Such dependence is represented by I ∝ P^k, where the exponent k = 1 indicates exciton recombination (radiative recombinations) while k = 2 is associated to free carriers recombination (non-radiative recombinations)^{[31, 34]}. For H1 sample, for the 1.61 eV emission, we have k = 1.21 ± 0.03 while k = 1.08 ± 0.05 for the 1.48 eV peak. For the H3 sample, we have k = 1.23 ± 0.01. Therefore, we have excitonic character for emissions from H1 and H3 heterostructures, which indicates that the band at 1.48 eV for sample H1 is possibly dominated by bound exciton states instead of defect-related recombination.

The Fig. 3(c) shows representative spectra of the temperature dependence of the macro PL for sample H1. The spectra are normalized and vertically shifted for a better visualization. The behavior of the main emission (at 10 K) and its FWHM as a function of temperature for sample H1 are shown in Fig. 3(d). The same result for the lower energy band is presented in Fig. 3(e). The solid red lines in Figs. 3 (d) and 3(e) represent Varshni equation E = E₀ − (αT²)/(T + β). The equation was calculated using the CdTe parameters^[35] (α = 4.35 × 10⁻⁴ eV·K⁻¹ and β = 126.8 K) and E₀ as the experimental emissions at 10 K. The main emission (QW-like emission around 1.61 eV at 8 K) does not follow Varshini’s equation, as observed in Fig. 3(d), while the lower energy emission (defects and bound exciton states around 1.48 eV at 8 K) in Fig. 3(e) does. The behavior observed in Fig. 3(d) is very similar to what has been reported for excitons in InGaAs^[36] QDs structures, CdTe/ZnTe^[37] self-assembled QDs and quasi-0D excitons in CdTe/CdMnTe^[38] and QWs on GaAs and GaSb substrates. It arises from an equilibrium and non-equilibrium distribution mechanism^[36] of the carriers in localized states in interface states and potential fluctuations associated to the roughness observed in Fig. 1(c). The maximum for the FWHM observed around 80 K reflects a competition between localized and delocalized exciton dynamics^[39], thus indicating that, besides emitting at the same energy expected for the electron−heavy hole recombination in a CdTe/CdMnTe QW^[29], the main emission at 1.61 eV for sample H1 has a non negligble 0D character. The FWHM behavior for lower energy emission in Fig. 3(e) may be associated to thermally induced redistribution between the localized states. The PL integral in Fig. 3(f) summarizes the temperature-dependent behavior already shown in Fig. 3(c) for QW-like emission (1.61 eV at 8 K) and defects and bound exciton states (1.48 eV at 8 K). In 2D QW systems, localized states emission intensity usually decreases faster with temperature due to thermal populations of the upper energy levels of a nearby QW. The opposite behaviour observed in Fig. 3(f) is typical of systems with strong carrier localization and disorder^[40].

Fig. 4 shows the low temperature μ-PL results for H1 and H3. The spectra in Figs. 4(a) and 4(b) were normalized and vertically shifted for a better visualization. Fig. 4(a) shows that for higher excitation powers, the μ-PL spectrum of sample H1 becomes similar to the macro PL one. As the laser power is decreased we observe a series of bright and sharp emission lines which change as we probe different places on the sample. We see that there are two distributions of sharp lines, one around 1.61 eV and another one right above 1.70 eV, as in the macro PL spectrum of Fig. 2(c). Again, this result evidences the localized component of the two higher energy emissions from sample H1 which is induced by local confinement. Fig. 4(b) shows that sample H3, in comparison to H1, display more spaced optical emissions when different spatial possitions are probed. This result reflects the sample structure illustrated in Fig. 2(b). The surface roughness caused by the Volmer−Weber growth mode does not allow the formation of a complete CdTe layer between the CdMnTe barriers, and three-dimensional structures arise, as a random ensemble of quantum-dots, emiting in a wide range of about 200 meV. Fig. 4(c) compares the FWHM for randomly choose features in the spectra of samples H1 and H3. As we observe, the single lines in sample H1 range from 1.3 to 4.2 meV in linewidth, while in sample H3 from 4.5 to 12 meV. The larger values of FWHM of the optical emissions detected for sample H3 can possibly be associated to interdifusion in the thin CdTe layer. Fig. 4(d) presents an intensity map obtained from a 20 × 20 μm² in steps of 1 μm across sample H3. The strong spatial dependence of the optical emission demonstrates the local characteristics of the confined states in the heterostructure.

The Fig. 5(a) shows circular polarization-resolved μ-PL measurements on two different places on sample H1. The degree of spin polarization ρ is defined as ρ = (I⁺ − I⁻)/(I⁺ + I⁻), where I⁺, I⁻ are the integrated intensities for each PL component of corresponding to σ+ and σ− polarizations. There is clear a non-zero spin polarization excited along the whole emission range of the sample. Fig. 5(b) shows a normalized μ-PL map obtained after scanning the sample surface along an area of 10 × 10 μm² in steps of 1 μm. Fig. 5(c) is the corresponding spin polarization map. As we observe, the sharp emission lines of this sample are strongly polarized with degrees of spin polarization which can exceed 40%. This result may be important for new studies of spin polarized injection and tunneling in double quantum well structures based in non-magnetic and magnetic quantum wells^[41] as well as exchange interaction and properties of excitons excited in CdTe islands^[42]. The sharp and circularly polarized emission lines detected in our heterostructures based on diluted magnetic semiconductors may also create perspectives for the development of polarized single photon sources^[43] for quantum information technology. Furthermore, the nanostructures studied here can be employed for basic physics investigations and explore potential applications in single quantum emitters^{[44, 45]} and quantum communication^[46] using a low-cost silicon- based technology platform.

We have used the large lattice mismatch of almost 20% between Si(111) substrates and Ⅱ–Ⅵ dilute magnetic semiconductors to produce bright emitting quantum structures. By sandwiching thin CdTe layers between CdMnTe barriers, we studied the optical properties of confined energy levels in theses structures as a function of the CdTe layer thickness. For a 15 nm CdTe layer, we identify three main emission bands at 1.48, 1.61, and 1.71 eV at 8 K. Macro PL measurements as function of laser excitation power and temperature indicate that the lower emission band is related to defects and bound exciton states. The same measurements also indicate that main emission band at 1.61 eV has weak 2D character while the higher energy emission band has essentially a strong 3D nature. As the CdTe layer is reduced, the emissions blue shift as expected. Even for the lower CdTe thickness of 3 nm, μ-PL measurements identify strongly emitting isolated lines which span along a 190 meV emission range. However, these lines are slightly broader than the single lines identified in the other samples possibly due to intermixing in the thin CdTe layer. Circular polarization-resolved μ-PL measurements show that the isolated emission lines are strongly polarized with degree of polarization which can be higher than 40%. Our heterostructures, although expected to present a high density of defects originated from the Volmer−Weber growth mode, still show a very bright quantum well-like emission for thick enough CdTe layer. To our knowledge, it is the first time that optical properties of such Ⅱ–Ⅵ heterostructures grown on silicon are reported. Besides that, when the CdTe is thin enough, the interface roughness leads the appearance of 0D localized states, exhibiting quantum-dot like spectra, also with very high intensity. This opens up the possibility of many applications for these structures in silicon-based photonics, spintronic devices and single photon sources.