Carrier Dynamics and Magneto-Optical Properties of Cd1-xMnxS Nanoparticles

© 2012 Oliveira Dantas and Soares de Freitas Neto, licensee InTech. This is an open access chapter distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Carrier Dynamics and Magneto-Optical Properties of Cd1-xMnxS Nanoparticles


Introduction
Cd1-xMnxS nanoparticles (NPs) with size quantum confinement belong to the diluted magnetic semiconductor (DMS) quantum dot (QD) class of materials that has been widely studied in the last few years. The study of quasi-zero-dimensional Diluted Magnetic Semiconductors (DMS), such as Cd1-xMnxS Quantum Dots (QDs), is strongly motivated due to the localization of magnetic ions in the same places as the free-like electron and hole carriers occurring in these nanomaterials [1,2]. This interesting phenomenon causes unique properties in DMS dots that can be explored in different technological applications, such as wavelength tunable lasers [3], solar cells [4,5], or in spintronic devices [6,7]. In this context, glass matrix-encapsulated Cd1-xMnxS NPs emerge as potential candidates for several applications, given that this host transparent material is robust and provides excellent stability for DMS nanostructures. Therefore, the luminescent properties and carrier dynamics of Cd1-xMnxS NPs should be comprehensively understood in order to target optical applications. For instance, different models based on rate equations can be employed to describe the temperature-dependent carrier dynamics of DMS nanostructures, such as they have been applied to semiconductor quantum wells [8], N-impurity complexes in III-V materials [9], and self-assembled semiconductor quantum dots [10].
It is well known that the optical properties of NPs can be significantly changed by interactions between nanostructures and their host material, due mainly to the formation of surface defects [11,12]. These surface defects are heavily dependent on NP size and become more important with increasing surface-volume ratio. Generally, the comparison between the optical properties of Cd1-xMnxS QDs and their corresponding bulk is obtained in different environments. To the best of our knowledge, this study is probably the first that simultaneously investigates both the carrier dynamics and the magneto-optical properties of Cd1-xMnxS QDs and their corresponding bulk-like NC when both are embedded in the same host material.
Although the dot doped with impurities (metal and magnetic) are currently being synthesized by colloidal chemistry techniques [13,14], some possible applications require the nanoparticles (NPs) being embedded in robust and transparent host materials. In this context, the melting-nucleation approach appears as an appropriate synthesis technique since it allows the growth of DMS nanocrystals (NCs) embedded in different glass matrices. In addition to the controllable dot size and Mn 2+ ion fraction incorporated into Cd1-xMnxS dots which can be achieved by this synthesis protocol, for example, the host glass matrix provides an excellent stability to the NPs. In particular for the melting-nucleation protocol used in this chapter, it is presented a discussion on the doping of QDs with magnetic impurities reasoned in two main models [3]: the 'trapped-dopant' and 'self-purification' mechanisms.
In this chapter, we have employed the optical absorption (OA), magnetic force microscopy (MFM), photoluminescence (PL), and magnetic circularly polarized photoluminescence (MCPL) measurements in order to investigate the properties of Cd1-xMnxS NPs that were successfully grown in a glass matrix. The organization of this chapter is shown as follows. In the section 2 (next section), we present the synthesis protocol that was employed in order to grow Cd1-xMnxS NPs in a glass matrix. The results obtained from the experimental techniques are presented and discussed in the section 3, highlighting the carrier dynamics and the magneto-optical properties of nanoparticles. We conclude our study in the section 4.

Synthesis of Cd1-xMnxS nanoparticles in a glass matrix
The host glass matrix for NP growth was labeled SNAB since its nominal composition is: 40SiO2.30Na2CO3.1Al2O3.29B2O3 (mol %). Cd1-xMnxS NPs were successfully synthesized in this glass matrix by adding 2[CdO + S] (wt % of SNAB), and x[Mn] (wt % of Cd), with x = 0.0, 0.5, 5.0, and 10 %. The synthesis method consists in a two sequential melting-nucleation approach, in which it is possible obtain ensembles of nearly spherical nanoparticles embedded in a glass matrix [12]. First, the powder mixture was melted in an alumina crucible at 1200 ºC for 30 minutes. Next, the melted mixture was quickly cooled down to room temperature where diffusion of Cd 2+ , Mn 2+ , and S 2-species took place. This diffusion resulted in Cd1-xMnxS NP growth in the SNAB glass environment.
In a second stage, a sample with x = 0.100 was subjected to a thermal annealing at 560 ºC for 6 h in order to enhance the diffusion of ions within the host SNAB matrix which promotes the growth of magnetic dots. Room temperature XRD pattern of the undoped CdS NPs (x = 0) embedded in the SNAB glass matrix was recorded with a XRD-6000 Shimadzu diffractometer using monochromatic Cu-Kα1 radiation (λ = 1.54056 Å). Thus, the wurtzite structure of CdS NPs embedded in the SNAB glass matrix has been confirmed. Evidently, the Cd1-xMnxS NPs with diluted magnetic doping have this same wurtzite structure, since it is a common phase for this DMS material.

Results and discussions
We have employed several experimental techniques in order to investigate the carrier dynamics and the magneto-optical properties of Cd1-xMnxS NPs. The room temperature absorption band edge of synthesized Cd1-xMnxS NCs was obtained with a double beam UV -VIS -NIR spectrophotometer (Varian, Cary 500) operating between 250 and 800 nm and with a spectral resolution of 1 nm. Photoluminescence (PL) measurements were taken with a 405 nm (~3.06 eV) continuous wave laser focused on a ~200 μm ray spot with an excitation power of 2.5 mW. Cd1-xMnxS NP luminescence was collected using a USB4000 spectrometer from Ocean Optics equipped with a Toshiba TCD1304AP 3648-element linear CCD-array detector, in the 10 K to 300 K temperature range, with a 435 nm high-pass filter. The magnetic force microscopy images of the Cd1-xMnxS NPs doped with x = 0.100 were recorded at room temperature with a scanning probe microscope (Shimadzu, SPM -9600).
The magneto-photoluminescence (MPL) measurements were performed using superconductor coils (Oxford Instruments) with fields up to 15 T. The samples were placed into the liquid helium cryostat at 2 K and excited using a 405 nm (± 5 nm) continuous wave laser, from Laserline Laser Technology, focused on ~ 200 μm rays spot with excitation intensity values of 10 mW. The detected MPL was carried out with an ocean optics spectrometer (USB4000) and the polarization was analyzed using a λ/4 waveplate and with linear polarizer fixed parallel to the spectrometer entrance, in order to collect the photons with σ + and σcircular polarizations, respectively.

Carrier dynamics
The room temperature OA spectra of Cd1-xMnxS NPs, with different x-concentrations, are shown in Fig. 1a. The formation of two well defined groups of Cd1-xMnxS NPs of different sizes was confirmed by the two bands in the OA spectra. As indicated in Fig. 1a, these two groups of NPs were named: (i) QDs because their quantum confinement properties provoked a change in band energy around ~3 eV; and (ii) bulk-like NCs indicated by the absence of quantum confinement given the fixed band around ~2.58 eV, a value near the energy gap of bulk CdS [15,16]. At the bottom of Fig. 1a is the OA spectrum of the SNAB glass matrix where, in contrast, it can be seen that over a broad spectral range there is a complete absence of any band associated with NPs. Figure 1a shows that the undoped CdS QDs (x = 0.000) exhibit confinement energy ( conf E ) as indicated by the OA band peak at ~3.10 eV. From this value and using a confinement model based on effective mass approximation [12,[15][16][17][18], the mean QD radius R was estimated by the expression: Econf = Eg + (ħ 2 π 2 ⁄ 2μR 2 ) -1.8(e 2 ⁄ εR), where Eg is the bulk material energy gap, μ is the reduced effective mass, e is the elementary charge, and ε is the dielectric constant. From this, a mean radius of about R~2.0 nm was estimated for the CdS QDs, thus confirming strong size quantum confinement [16].
Furthermore, the increase in x-concentration clearly induced a blue shift in the OA band of the Cd1-xMnxS QDs from ~3.10 eV (x = 0.000) to ~3.22 eV for the highest magnetic doping (x = 0.100). Since these QDs were grown under identical synthesis conditions within the glass environment, it is expected that they would have the same mean size. As a result, there were no significant differences in the quantum confinements of these QDs that would cause shifts in the OA band peaks. Thus, it was concluded that the observed blue shift in OA band peak ( Fig. 1a) was a consequence of the sp-d exchange interactions between electrons confined in dot states and those located in the partially filled Mn 2+ states. This explanation is reasonable since replacing Cd 2+ with Mn 2+ ions should increase the energy gap of Cd1-xMnxS QDs [18]. In addition, it is interesting to note the weak sp-d exchange interaction in the Cd1-xMnxS bulklike NCs because their OA band remains in an almost fixed position (~2.58 eV).   quantum confinement, while the bulk-like NCs with mean radius R ~ 10.0 nm hardly exhibit any size confinement [19].
In addition, a large quantity Cd1-xMnxS NPs can be observed in Fig. 1b, as well as in the corresponding phase MFM image shown in Fig. 1c. These images reveal great proximity between the two groups of NPs (QDs and bulk-like NCs), so that strong coupling between their wave functions is expected. In Fig. 1c, the topographic signal can be neglected because its phase MFM was recorded with a 30 nm lift from the sample's surface. Thus, interaction between tip and NP magnetization induces the contrast observed in this phase MFM image.
T h e d a r k a r e a ( l i g h t a r e a ) i s c a u s e d b y attraction (repulsion) between tip and NP magnetization represented by the South (North) magnetic pole in the vertical scale bar of Fig. 1c. Evidently, the magnetization in each NP (QD or bulk-like NC) is caused by the sizedependent sp-d exchange interactions, proving that Mn 2+ ions are incorporated into the DMS nanostructures. This Mn 2+ ion incorporation in NPs has also been established by electron paramagnetic resonance (EPR) measurements and simulations with other samples synthesized in the same way as in this research [17]. In Fig. 1c, it is interesting to note that there is a relationship between the NP size and the direction of its magnetic moment: small (large) NPs have their magnetic moment oriented towards the North (South) pole.  (1) and (2) for the QDs, as well as (1) b and (2) b for the bulk-like NCs. In a recent study of other similar Cd1-xMnxS NPs with wurtzite structure, the existence of emissions from two trap levels related to the presence of deep defects was demonstrated [20]. The origin of these defects in Cd1-xMnxS NPs (and CdS NPs) with hexagonal wurtzite structure is possibly related to two energetically different VCd -VS divacancies: one oriented along the hexagonal c-axis (assigned to trap (1)), and the other oriented along the basal Cd-S bond (assigned to trap (2)) [20]. Furthermore, the size-dependence of these trap-levels, (1) and (2), has been confirmed for CdSe NCs [21], explaining the observed emissions from them in both the QDs The complete recombination aspects of these PL spectra are well-described in a diagram in Fig. 2c, where six (seven) emission bands can be identified for the CdS NPs (Cd1-xMnxS NPs with x ≠ 0). In Fig. 2b, the asymmetric shape of the emission band around 480 nm at low temperatures confirms the presence of shallow virtual levels for the QDs, and evidently there is also for the bulk-like NCs, as depicted in Fig. 2c. However, this emission band (480 nm) becomes symmetric with rising temperature, which demonstrates that the trapped carriers in the virtual levels are being released to other non-radiative channels of QDs. It is interesting to note that in Fig. 2a the excitonic emission (Eexc) of CdS QDs is almost suppressed due to the strong presence of non-radiative channels, including one related to the energy transfer from QDs to bulk-like NCs. However, a comparison between the PL spectra of the CdS and the Cd0.950Mn0.050S NPs (see Fig. 2) clearly reveals that increasing xconcentration induces gradual suppression of emissions from all trap-levels ((1), (2), (1) b , and (2) b ), since Mn 2+ ions are replacing the VCd vacancies in the NPs. Indeed, this fascinating behavior provides further evidence that the deep defects are caused by VCd -VS divacancies, and that the NPs are actually being doped by Mn 2+ ions. Hence, the non-radiative channels that supply the deep trap-levels disappear with increasing x-concentration in Cd0.950Mn0.050S NPs, as shown in Fig. 2b.
In Fig. 2c, the wavy arrows represent non-radiative channels from the excitonic states of QDs, and from the conduction band (CB) of bulk-like NCs. Here, non-radiative energy transfer (ET) is given by the rate 1 n ET  (with n = A, B, C, A', and B'), where n ET  is the carrier escape time from an NP to one of these five non-radiative transitions. In our model, we have assumed that the non-radiative paths from the excitonic states of QDs, as well as from the conduction band of bulk-like NCs to the deep trap-levels ((1), (2), (1) b , and (2) b ) can be disregarded. However, it is evident that these deep trap-levels may be filled by carriers from: (i) the shallow virtual levels of QDs and bulk-like NCs; and (ii) the 4 T1 levels of Mn 2+ ions [20]. Energy transfers from the excitonic states of QDs follow three paths: (A) to virtual levels (QDs); (B) to the conduction band of bulk-like NCs; and (C) to the 4 T1 level of Mn 2+ ions. On the other hand, the energy transfers from the conduction band of bulk-like NCs follow two paths: (A') to virtual levels (bulk); and (B') to the 4 T1 level of Mn 2+ ions. It is well known that the very fast energy transfer from a NP to Mn 2+ ions is generally resonant due to the high density of states above the emissive 4 T1 level, 1 as shown by the 2,4 Γ levels in Fig. 2c. However, size quantum confinement can play an important role in this process that, besides being mediated by the sp-d exchange interactions, is strongly dependent on the Mn 2+ fraction in Cd1-xMnxS NPs. In other words, QDs and bulk-like NCs are expected to behave differently due to the strong confinement of the QDs with a small mean radius of about R~2.0 nm.
The excitonic states of QDs can be denoted by 1 , and the CB of bulk-like NCs by 1 b . The carrier number (depending on temperature T) of these two states is given by   NT, respectively. Since carriers are thermally distributed each one of the three nonradiative channels related to QDs is supplied by where En (with n = A, B, and C) is the corresponding activation energy of the non-radiative n channel. Similarly, carriers are transferred to each one of the two non-radiative channels related to bulk-like NCs, where n = A', and B'. Furthermore, as shown in Fig. 2c by the straight, downward pointing arrows, radiative emissions are also present from both QDs and bulk-like NCs in the PL spectra which are related to 1 QD r  and 1 b r  rates, respectively.
The straight, upward pointing arrow, indicated by g (g'), represents photo-excitation of the QDs (bulk-like NCs) caused by the laser pump. The carrier dynamics that take into account Carrier Dynamics and Magneto-Optical Properties of Cd1-x MnxS Nanoparticles 7 these transitions from the 1 (QD) and 1 b (bulk-like NC) levels can be described by the following rate equations: Figure 2. PL spectra of both (a) the CdS NPs (x = 0.000) and (b) Cd0.950Mn0.050S NPs at several temperatures, from 20 K (top) to 300 K (bottom), as indicated by the downward pointing arrows. Their recombination aspects are depicted in panel (c), where the emissions from both the QDs and the bulklike NCs are clearly identified. In addition, the characteristic emission of Mn 2+ ions ( 4 T1-6 A1), EMn ~2.12 eV, when substitutionally incorporated in II-VI semiconductors is also evident. In the present energy scale, the 6 A1 level of the Mn 2+ ions is located at top of the QD ground state.
QDs virtual levels (QDs) QDs bulk-like NCs radiative QDs ions emission QDs bulk-like NCs bulk-like NCs virtual levels (bulk) radiative bulk-like NCs ions emission ' ;  1) and (2), we get: The carriers' number in the QDs ( 1 level) as a function of temperature T is given by Eq.
Evidently, the second term on the right side of Eq. (5) is related to the carrier-mediated energy transfer from QDs to bulk-like NCs, and can be defined by: Carrier Dynamics and Magneto-Optical Properties of Cd1-x MnxS Nanoparticles 9 This represents temperature-dependent excitation of bulk-like NCs caused by carriers transferred from the QDs. Thus, Eq. (5) can be solved, resulting in the following expression: where   NT is given by Eq.
(6). From Eqs. (3) and (7), we can find the steady-state intensities of the QDs and bulk-like NCs, respectively, which results in: In Eq. (8), it is interesting to note that the term   (6) since the carrier-mediated energy transfer from QDs to bulk-like NCs is strongly temperature dependent. Evidently, there is coupling between Eqs. (8) and (9), and they can be fit to the experimental integrated PL intensity. This in turn, permits the deduction of activation energies related to the non-radiative channels of QDs (EA, EB, and EC) as well as of bulk-like NCs (EA', and EB').
Figures 3a, b, and c show integrated PL intensity behavior for the doped Cd1-xMnxS NPs (x ≠ 0) as a function of temperature. Here, the solid and open triangle symbols represent the bulk-like NCs and QDs, respectively. At low temperatures, QD emission intensity decreases quickly while bulk-like NC emissions remain almost constant except for a small increase at x = 0.100 (Fig. 3c). This behavior is due to the trapping of excited carriers from the excitonic states to the shallow virtual levels of QDs, where temperature increases induce a gradual release of these carriers to other electronic states, including the CB of bulk-like NCs. This carrier-mediated energy transfer from QDs to bulk-like NCs is a tunnelling phenomenon that is strongly dependent on the coupling between the wave functions of these NPs [24]. This effect is expected, given the high proximity between QDs and bulk-like NCs as confirmed by the MFM images (Figs. 1b and c). The ratio between these PL peak intensities (bulk-like NCs/QDs) as a function of temperature is shown in the insets of Fig.3. Here, it can be seen that the ratio increases at low temperatures and then decreases as QD emissions remain constant and bulk-like NC emissions decrease.
In the insets of Fig. 3, a fitting procedure with a Gaussian-like component, gives the temperature that yields the maximum ratio for each x-concentration: 122 K (x = 0.005); 134 K (x = 0.050); and 127 K (x = 0.100). Moreover, the FWHM (Full Width at Half Maximum) of the Gaussian-like component broadens with increasing x-concentration: 63 K (x = 0.005); 70 K (x = 0.050); and 74 K (x = 0.100), thus confirming that emission intensity from bulk-like NCs decreases more slowly after the maximum ratio is reached. It is interesting to note that the peak ratio between the PL intensities of bulk-like NCs/QDs is related to the inflection point of the corresponding integrated PL intensity of bulk-like NCs. This is indicated by the dashed vertical lines in Figs. 3a, b, and c. The inflection point temperatures were attributed to the maximum thermal energy transfer process from QDs to bulk-like NCs.
It can be seen that the temperatures obtained by the Gaussian fitting (T = 122 K, 134 K, and 127 K) can be related to delocalization thermal energies (like KBT) [25], which are needed to release the trapped carriers at shallow virtual levels (surface defects, for example) of QDs. Thus, the aforementioned EA activation energy coupled to these virtual levels (QDs) could be found by using the following expression: A B EK T  ; where KB is the Boltzmann constant, and T is the temperature obtained by the Gaussian fitting. As a result, the x-concentration dependent behavior of this EA activation energy is given by: 10.51 meV (x = 0.005); 11.54 meV (x = 0.050); and 10.94 meV (x = 0.100), where the deduced values remain almost invariable. This result can take into account two effects caused by the increasing x-concentration of Cd1-xMnxS QDs: (i) the increasing energy gap that was observed in OA spectra of Fig. 1a; and (ii) possible density amplification of virtual levels associated to shallow defects of QDs [23]. Therefore, the combination of these effects in the electronic structure of Cd1-xMnxS QDs (x ≠ 0) explains the nearly constant values obtained for the EA activation energy.
In order to deduce the additional activation energies (EB and EC) related to other nonradiative channels of doped Cd1-xMnxS NPs (x ≠ 0), Eqs. (8) and (9)    For the undoped NPs (x = 0), only a PL emission band associated with the bulk-like NCs could clearly be observed (see Fig. 2a). Therefore, it was not possible to use Eq. (8) to find the activation energies associated with the QDs. In addition, since there was no magnetic doping for these NPs (x = 0), it is expected that the non-radiative channels related to Mn 2+ ions would not exist. With these alterations, a modified Eq. (9) was used in fitting the experimental integrated PL intensity of CdS bulk-like NCs, where the EA, EB, and EA' activation energies were considered as parameters of fit. Figure 4a shows that good fit of the experimental data was achieved which confirms the absence of non-radiative channels related to Mn 2+ ions. Thus, even though any PL emissions from CdS QDs were not observed (see Fig. 2a), the EA and EB activation energies associated with them could be indirectly determined in this fitting procedure due to the carrier-mediated energy transfer from the QDs to the bulk-like NCs that is also present in the modified Eq. (9).
Furthermore, in Figs. 4b, c, and d, the fittings for both the QDs (Eq. (8)) and bulk-like NCs (Eq. (9)) are in excellent agreement with the experimental data. However, these concordances were not achieved by further fittings given: (i) one or two non-radiative channels for QDs (x ≠ 0), and (ii) one non-radiative channel for bulk-like NCs (x ≠ 0). Therefore, all these fittings evidently demonstrate that the Eqs. (8) and (9) for Cd1-xMnxS NPs (x ≠ 0), as well as the modified Eq. (9) for CdS NPs, are satisfactorily suitable for describing the temperature-dependent carrier dynamics of the 1 or 1 b levels. Table 1 shows all the activation energies found that are related to non-radiative channels of Cd1-xMnxS NPs (QDs and bulk-like NCs) where, for doped NPs (x ≠ 0), the EA remains almost constant as previously explained. Moreover, for undoped NPs the value EA ~ 16.88 meV is slightly larger than that for doped NPs (EA ~ 11 meV). This proves that increases in xconcentration enhance the density of the virtual levels associated with the shallow defects of QDs. The carrier-mediated energy transfer from QDs to bulk-like NCs, a tunnelling phenomenon, is evidently being hampered due to EB rising with increases in x-concentration (see Table I). In heavily doped NPs, there are many Mn 2+ ions incorporated near the surface of both groups of NPs (QDs and bulk-like NCs) [22], an effect that enhances Mn-Mn interactions [17,20]. Therefore, we can conclude that high quantities of Mn 2+ ions near the surface of these NPs weakens the coupling between their wave functions which hampers the tunnelling process from the QDs to bulk-like NCs. Consequently, this effect also contributes to the excitonic emission (Eexc) of Cd1-xMnxS QDs, as observed in Fig. 2b The non-radiative energy transfers from NPs to Mn 2+ ions are related to the following activation energies: EC for the QDs; and EB' for the bulk-like NCs. In Table I, it can be seen that EC increases and EB decreases with rising x-concentration. This opposite behavior between QDs and bulk-like NCs demonstrates that the sp-d exchange interactions are strongly dependent on the size quantum confinement of the NPs. Increasing x-concentration from 0.000 to 0.100 induces considerable blue shift in the energy gap of the QDs, which can be disregarded for the bulk-like NCs (see Fig. 1a), while the density of the 2,4 Γ levels of Mn 2+ ions is being amplified. Thus, the depth of 2,4 Γ levels is increasing in relation to the excitonic states of QDs, while remaining almost constant for the CB of bulk-like NCs. Therefore, the combination of these effects explains very well the observed increase (decrease) in EC (EB') activation energy with increasing x-concentration.
In addition, increasing x-concentration induces the density amplification of the virtual levels associated with the shallow defects of bulk-like NCs. This also occurs in the QDs [26]. However, since the change in energy gap of bulk-like NCs can be disregarded, these virtual levels become shallower for the conduction band (CB). Hence, in Table 1, the decrease in EA' activation energy with the increase in x-concentration can be adequately explained by taking into account this effect in the electronic structure of the bulk-like NCs. Figure 5 shows the OA spectra, taken at room temperature, of Cd1-xMnxS magnetic NPs that were grown into the glass matrix environment. In Fig. 5a, the spectra were taken from NP samples with three different Mn-concentrations: x = 0.000, 0.050, and 0.100, and these samples did not have any thermal treatment. Only for comparison, the OA spectrum of the SNAB matrix is also shown in bottom of Fig. 5a and is clear the absence of any absorption band in the range between 350-650 nm. However, the OA spectra of all NP samples revealed the formation of two well defined groups of Cd1-xMnxS NPs with different sizes: (i) one group displaying a fixed band around 2.58 eV (near the energy gap of bulk CdS) and denominated as bulk-like NCs; (ii) the other displaying changing band energy due to quantum confinement properties and denominated as QDs.

Magneto-optical properties
A careful analysis of the bands attributed to Cd1-xMnxS QDs with concentrations x = 0.000; 0.050; and 0.100, clearly reveals a width of about 65 nm for each OA band which is due to a size distribution of the nanoparticles. From the OA peak at 3.13 eV, in Fig. 5a, and using the effective mass approximation [12,17], an average radius around R~2.0 nm was estimated for CdS QDs (x = 0.000), which confirms the strong size quantum confinement. It is noted that an increase of Mn-concentration induces a blueshift on the QDs band, from 3.13 eV, for x = 0.000, to 3.22 eV, for x = 0.100. Since these magnetic QDs were synthesized under the same thermodynamic conditions, one should expect that they have the same average dot size (R~2.0 nm) and, thus, no significant differences in their quantum confinements that would cause shift among the OA band peaks. Here, can be also inferred that the growth kinetics of these dots is not influenced by the magnetic ions, since the amount of Mn dispersed in the glass environment is actually very small. Therefore, we attribute the blueshift on the peaks of Fig. 5a to the sp-d exchange interaction between electrons confined in the dot and located in the partially filled Mn 2+ ion states. This explanation is quite reasonable [17] since the replacement of Cd by Mn, in Cd1-xMnxS NPs, should change the energy gap between 2.58 eV, for CdS buk (x = 0) and 3.5 eV, for MnS bulk (x = 1).

Figure 5.
Room temperature OA spectra of Cd1-xMnxS NPs embedded in the SNAB matrix. Panel (a) shows the spectra of the as-grown samples which did not receive any thermal treatment. For comparison, it is also shown the OA spectrum of the SNAB matrix, at the bottom. Panel (b) shows spectra of two identical samples containing the same magnetic ion doping (x = 0.100), but one before thermal annealing (BTA) and the other after thermal annealing (ATA) at T = 560 ºC for 6 h. Observe that the peak at 2.58 eV, attributed to bulk-like NCs, does not change with doping or annealing.
We also compare the optical spectra of two identical Cd0.900Mn0.100S samples, one before thermal annealing (BTA) and another after thermal annealing (ATA) at 560 ºC for 6h. The effect of this thermal annealing on the two OA band peaks of these NPs is shown in Fig. 5b. As expected for NPs without size quantum confinement, the OA band related to Cd0.900Mn0.100S buk-like NCs does not show any shift in the samples with or without thermal annealing. At the same time, the Cd0.900Mn0.100S QDs peak shows a redshift from 3.22 eV, in the sample without treatment (BTA), to ~3.17 eV, in the annealed sample (ATA). This shift can be ascribed to two possible annealing effects: (i) the size increase of the magnetic dot and thus, inducing weakening on the quantum confinement, and (ii) the decrease in the effective concentration of Mn 2+ ions incorporated into the dots during growth thus, inducing decreasing on the energy gap.
The dot size increase for increasing annealing time is a well known phenomenon governing the growth kinetic of nanoparticles in glass matrices [18,27,28]. However, a recent study of thermal treatments on undoped CdSe QDs [12] embedded in this same glass matrix (SNAB) showed a much smaller redshift (~0.03 eV) when annealed for 6 h. Since CdSe and CdS structures display great similarities, as well as Cd1-xMnxS with dilute Mn-concentration, it is reasonable to assume that they have the same growth kinetic in the same glass matrix. Thus, the higher shift (~0.05 eV) observed in Cd0.900Mn0.100S QDs annealed for 6 h provides strong evidences that the observed higher redshift must also be ascribed to a decrease of the effective concentration of Mn 2+ incorporated to the dots, and this decrease takes place during the thermal treatment of the sample. We shall return to this evaporation-like process later, since its understanding is still an opened subject on doping processes in semiconductor QDs [3]. Figure 6 presents the magnetic circularly polarized PL spectra, taken at 2.0 K and 15 T, of CdS ( Fig. 6a) and of Cd0.900Mn0.100S (Fig. 6b) NP samples. A decreasing from 300 K to temperatures near ~2.0 K causes an increase in the energy gap of CdS QDs, as well as in bulk CdS, of about ~ 85 meV [29]. Certainly, a similar temperature-dependent behaviour is expected for Cd1-xMnxS QDs with dot size R ~ 2 nm and for diluted magnetic doping. However, even with this large blueshift in the OA bands, the Cd1-xMnxS QDs with absorption around 405 nm, as well as all bulk-like NCs, were excited during the measurements at 2.0 K, due the large OA band width (~ 65 nm) of QDs as well as the wavelength width (± 5 nm) of the 405 nm excitation laser source. Therefore, according to OA spectra shown in Fig. 5, the emissions attributed to the two well-defined bulk-like NCs and QDs groups are observed in each spectrum of Fig. 6. The presence of virtual levels in these structures, as due to surface defects for example, can explain the asymmetric character of the emission band around 480 nm (Fig. 6b). Also, the broad emission band near 580 nm cannot be fitted by only one Gaussian-like component which provides further evidence for its complex nature associated to several emissions. We may conclude that besides the radiative recombination of excitons, labelled as EQD (Eb) for the QDs (bulk-like NCs), there are also the emissions from deep defect levels, labelled as (1) and (2) for QDs and (1)b and (2)b for bulklike NCs. These emissions detected on the PL spectra are qualitatively described in the diagram depicted in Fig. 6c where the seven emission bands are identified. The deep defect levels in CdS and Cd1-xMnxS NPs with hexagonal wurtzite structure, a common phase for these materials, [30][31][32] are possibly related to two energetically different divacancy defects, VCd -VS, associated to the absence of Cd 2+ and S 2-ions in the crystalline NP structure [20]. One divacancy is oriented along the hexagonal c-axis of the wurtzite CdS structure and assigned to trap(1), whereas the other is oriented along the basal Cd-S bond directions and assigned to trap(2) [20,21]. The size dependence of these trapping levels has been confirmed for CdSe NCs [21], and is used to explain the detected emissions from QDs ( The comparison between the emissions and absorptions of Cd1-xMnxS NPs, by taking into account the mentioned increase in energy gap at low temperature (~ 2.0 K), revels that these nanostructures embedded in a glass matrix exhibit an anomalously large Stokes shift (ΔSS) given by: ΔSS ~ 0.65 eV for QDs, and ΔSS ~ 0.54 eV for bulk-like NCs. Possibly, the origin for this large Stokes shift can be attributed to radiative recombination due the many-body effects on the excitonic states of the NPs, a phenomenon that was recently demonstrated for PbS nanocrystals [21] and should be considered for the Cd1-xMnxS NPs. Certainly, further investigations are required in order to reach a comprehensive explanation for these observed large Stokes shifts in the Cd1-xMnxS NPs embedded in a glass matrix. The characteristic emissions 4 T1 → 6 A1 between levels of Mn 2+ ions, labelled as E(Mn 2+ ) in Figs. 6b and 6c and with transition energy ~2.12 eV, also confirm that these magnetic impurities were substitutionally incorporated in the Cd1-xMnxS NPs [1,22,23]. This incorporation of Mn 2+ ions in NPs has also been proved by electron paramagnetic resonance (EPR) measurements and simulations in other samples which were synthesized by the same method used in this work [17,20].
Note that the emissions from deep defect levels observed in bulk-like NCs, and shown in Figs. 6a and 6c with labels 1 b E and 2 b E , become almost suppressed in samples with magnetic doping (see Fig. 6b). Thus, Mn 2+ ions must be filling out the Cd vacancies, VCd, during doping and this interesting fact provides further evidence not only for the existence of the deep divacancies, VCd -VS, but also for the incorporation of Mn 2+ ions in the NPs.
Figures 7a and 7b present the circularly polarized (σand σ + ) PL spectra of Cd0.900Mn0.100S NP samples without thermal treatment and taken at 2.0 K, for several magnetic field values between 0.0 and 15.0 T. The magnetic subcomponent emissions from QDs and from bulk-like NCs can be clearly observed in all PL spectra. It is also noted that σ + increase the intensity faster than σemissions, thus resulting in the strong PL circular polarization. The relative intensity ratio between the polarized emissions from QDs and from bulk-like NCs is shown in Fig. 7c as a function of magnetic field for two different Mn-concentrations (x = 0.050 and 0.100).
The internal optical transition ( 4 T1→ 6 A1) occurring within excited 3d 5 shells of the Mn 2+ ions is highly sensitive to the presence of external magnetic fields [33,34]. After electron-hole pair creation by laser excitation, the band-edge exciton can either recombine radiatively or transfer its energy to a Mn 2+ ion via an Auger-like process that depends on the exciton-Mn coupling. At low temperatures and in magnetic fields, this Mn 2+ PL band remains unpolarized and, eventually, becomes suppressed while the circularly polarized band-edge excitonic emissions increase the intensity. This is a universal behaviour that has been observed in DMS crystals, epilayers, quantum wells, quantum wires, and in self-assembled epitaxial quantum dots [33][34][35][36][37][38][39][40][41][42]. Although the precise mechanism of energy transfer from excitons to electrons in the Mn 2+ 3d 5 shell is still debated in the scientific community, [37,38,[42][43][44] the marked field dependence of this process indicates a spindependent excitation transfer as described by Nawrocki [45] and by Chernenko [42,43].
The energy transfer from QDs to Mn 2+ ions is also highly sensitive to the presence of external magnetic fields [2]. For example, in self-organized Cd1-xMnxSe QD samples, the 4 T1 → 6 A1 emissions from the incorporated Mn 2+ ions are completely suppressed at a magnetic field values near 3-4 T [40]. According to Nawrocki model [45], this occurs because the Mn 2+ magnetization freezes out the electron population in the Ms = -5/2 ground state Zeeman sublevels of the Mn 2+ ions (labelled 6 A1) [2, [42][43][44][45]. In this model, the transition of an electron from the conduction to the valence band occurs without change of its spin, and the transition is allowed if the total spin of the combined system of Mn-ion+electron is conserved. In particular, no Auger recombination is possible with participation of Mn-ion ground state ( 6 A1) with spin S = 5/2 and Sz = ± 5/2 since the excited state ( 4 T1) has spin S = 3/2 and Sz = ± 3/2. The suppression of the Auger recombination in the high magnetic field can be Carrier Dynamics and Magneto-Optical Properties of Cd1-x MnxS Nanoparticles 19 explained by assuming thermalization of Mn-ions in the lowest state with Sz = -5/2. Thus, the band-edge excitonic emission intensity saturates with increasing magnetic field, indicating alignment of the QD exciton spins by the magnetic field. Figure 7. Circularly polarized σ -(panel (a)) and σ + (panel (b)) PL spectra taken at 2K for Cd0.900Mn0.100S NPs without thermal treatment and excited with line 405 nm of a laser source. Panel (c): Comparison between the ratio of σ -(filled symbols) and σ + (opened symbols) emission intensities from QDs and from bulk-like NCs in samples with concentrations x = 0.050 (circles) and x = 0.100 (triangles) for increasing magnetic field values. Notice that magnetic doping affects strongly the magnetic dependence of these intensities.
Hence, in our samples is also expected that the electron-hole radiative recombinations (EQD and Eb) show increasing intensities while the E(Mn 2+ ) emissions show decreasing change of intensities for increasing magnetic field. Because the overlap with the Eb emissions in these samples, the E(Mn 2+ ) emissions cannot be clearly resolved in the PL spectra (Figs. 7a and b) but should show a change in the relative intensity (Fig. 7c). For sample with Mnconcentration x = 0.100, the ratio between polarized PL intensities from QDs and from bulklike NCs (triangles) displays non-monotonic behaviour from 0 to 12 T with a saturation tendency, where the increased exciton emissions increase occurring at the expenses of Mn PL emissions decrease, as the magnetic fields increase. In the sample with smaller concentration, x = 0.050, the ratio between PL intensities (circles) increases almost linearly up to 15 T, or even with slight intensity change. In Fig. 7c, it is also noted a significant change in the relative intensities of emissions for σpolarization, starting at B ~ 4 T, in the sample with higher Mn-concentration (x = 0.100) and a much lower intensity change in the sample with Mn-concentration x = 0.050. It is our understanding that this effect is related to the suppressed E(Mn 2+ ) emissions.
Furthermore, in Fig. 7c, the x-concentration dependent behaviour of the exciton intensity variation with the magnetic field indicates a strong modification of the Auger energy transfer rate from the excitons to Mn 2+ ions. Therefore, in the low Mn-concentration this energy transfer does not occur as strongly as in the case of the high Mn-concentration. It has been shown that Auger energy transfer is sensitively dependent on carrier densityexcitation power [40,42] and Mn-concentration [36]. In the high power excitation, the suppression of the Auger process does not take place as strongly as in the case of the weak power excitation even in the high magnetic field region. In the low excitation intensity the PL intensity curve behaves very different as in the high excitation intensity. Based on the approach of Nawrocki et al. [45], Chernenko et al. [42,43]  E , become almost suppressed in samples with magnetic doping (see Fig. 6b). Thus, Mn 2+ ions must be filling out the Cd vacancies, VCd, during doping and this interesting fact provides evidence that Mn 2+ doping can alter the carrier density in the NPs.
The Zeeman energy splitting in the electronic structure of the NPs is also other important effect caused by the increasing in the magnetic field. It is well known that in DMS structures the Zeeman energy splitting can be considerably altered by the exchange interaction between the carrier spins and the substitutional doping magnetic ions [46]. Thus, in our Cd1-xMnxS NP samples, it is quite expected different Zeeman energy splitting for QDs and bulklike NCs with the increase in the magnetic field. This in turn should also increase the separation between the excited electronic levels of QDs and bulk-like NCs, so that the nonradiative energy transfer between them (as depicted in Fig. 6b) is being weakened up to be completely interrupted at a given magnetic field value. We understand that this phenomenon occurs at a magnetic field B ~ 12 T for the Cd1-xMnxS NP sample with x = 0.100, contributing thus to the abrupt increase in relative PL intensity (QDs/bulk-like NCs) shown in Fig. 7c. Since this effect cannot be clearly observed for the Cd1-xMnxS NP samples with x = 0.050, we can conclude that non-radiative energy transfer involving the excited electronic levels of QDs and bulk-like NCs was not completely broken off in the investigated magnetic field range, favouring the almost linear behaviour observed in Fig. 7c. It is important to mention that the energy transfers involving the excitonic states of QDs, the conduction band of bulk-like NCs, and the shallow virtual levels of NPs was demonstrated in section 3.1 (carrier dynamics) [47]. bulk-like NC emissions appear to increase quadratically with B and, at higher magnetic fields, show a saturation tendency near 35%. However, the degree of polarization for CdS bulk-like NCs (x = 0.000) shows a much slower increase with saturation value near 15%. In the Mn-doped samples (x ≠ 0), the bulk-like NCs exhibit a higher degree of polarization than the QDs, thus evidencing that the amount of Mn 2+ ions that are substitutionally incorporated into QDs is smaller than in the bulk-like NC. Evidently, this effect is related to the well known difficulty in doping semiconductor QDs with magnetic impurities. Furthermore, the mentioned non-radiative energy transfer from QDs to bulk-like NCs is also a cause for the lower degree of polarization for the QDs. It is fascinating to note that the degree of polarization for the Cd1-xMnxS QDs with x = 0.100 becomes higher than for the undoped bulk-like NCs (x = 0) at the same magnetic field in which the abrupt increase in relative PL intensity (Fig. 7c) takes place, i. e., B ~ 12 T. For the Cd1-xMnxS QDs with x = 0.050 this effect is less pronounced and occurs at a higher magnetic field (see Fig. 8). In addition, the different magneto-optical properties of the bulk-like NCs and QDs can be explained by taking into account a considerable change of exchange interaction between the carrier spins and the substitutional doping of magnetic ions incorporated into the NPs with different sizes. This fact confirms that the size quantum confinement plays important role on the magneto-optical properties of Cd1-xMnxS NPs. We note that the Mn-doped with x = 0.050 and undoped CdS NP samples does not exhibit any zero-field degree of polarization. However, the samples doped with x = 0.100 presented negative zero-field polarization, (0 )5 % B    , which is ascribed to a change in the Zeeman ground state character. This interesting phenomenon is related to an intrinsic magnetism of Cd1-xMnxS NPs caused by the change in the sp-d exchange interaction strength, which is strongly dependent on the doping mole fraction x of incorporated magnetic ions.
A comparison between the degrees of polarization for two sets of identical Mn-doped samples with x = 0.100, is shown in Fig. 8, one before thermal annealing, labelled BTA and represented by triangles symbols; another undergone a thermal annealing at T = 560 ºC for 6 h, labelled ATA and represented by star symbols. After thermal annealing, the Cd0.900Mn0.100S bulk-like NCs sample showed a degree of polarization very similar to the Cd0.950Mn0.050S bulk-like NCs sample. This fact indicates that during the thermal annealing at T = 560 ºC occurred a decrease in the effective concentration (xeff) of the Cd1-xMnxS bulk-like NCs. In agreement with observed redshift for OA band shown in Fig. 5b, this same dynamical doping process should also occur in the Cd1-xMnxS QDs. However, as shown in Fig. 8 (see open triangle and star symbols), the change in degree of polarization for QDs induced by thermal annealing is small due to two effects: (i) the strong localization of magnetic Mn 2+ ions in the same place as the charged carries confined to the dots, and (ii) the mentioned smaller amount of magnetic impurity that is incorporated into QDs. Figures 9 (a, b, and c) presents two-dimensional phase MFM images (room temperature) of the Cd1-xMnxS NPs (x = 0.100) that are located at the samples surface, where it is possible to investigate the thermal annealing effect on the total magnetic moment of NPs. The images (150 x 150 nm) with a lift of 20 nm were recorded in two situations: in Fig. 9a before thermal annealing (BTA); and in Fig. 9c after thermal annealing (ATA). The contrast between these MFM images is a result of the interactions between the tip and the NP magnetization. However, there is also a small influence of the sample topography because the probe is close to the sample surface (20 nm). As a result of magnetic interaction between tip and surface, the bright area (dark area) of the phase MFM image displays repulsive (attractive) interaction. Thus, the clear contrast that is observed in Fig. 9a, which is mainly caused by magnetic interactions with the NPs, is almost vanished after thermal annealing as shown in Fig. 5c. Hence, we may conclude that the total magnetic moment of each NP (observed in Fig. 9a) is caused by the sp-d exchange interactions and can be tuned by a suitable thermal annealing of the Cd1-xMnxS NP samples. In agreement with the results of Figs. 5 and 8, this behaviour can ascribed to diffusion of Mn 2+ ions from the core to a position near the NP surface and, therefore, decreasing the effective concentration (xeff) in Cd1-xMnxS samples. In addition, Figure 9b shows the phase MFM image (30 x 30 nm) obtained with no lift, of a Cd1-xMnxS NP (BTA) that is at the sample surface. Since this image was recorded with no lift, there is a strong influence of the sample topography and allowing the observation of the characteristic hexagon of the wurtzite structure. The main doping models for QDs that are used to explain the incorporation of impurities, including Mn 2+ as in our samples, are known as 'trapped-dopant' and 'self-purification' mechanisms, which have being largely discussed in last few years [3,[48][49][50][51][52]. The trappeddopant mechanism is governed by the growth kinetics, where the impurity is adsorbed on the dot surface and then covered by additional material, [3] while the self-purification mechanism is governed by a diffusion process of impurities to more stable and stronger binding energy sites near the surface of dots [48].
It becomes clear that the trapped-dopant mechanism is occurring in the course of the thermal annealing at T = 560 ºC of our sample, since the Cd1-xMnxS QDs are growing due to increasing annealing time (see Fig. 5b). However, it is necessary to answer the question: Is this mechanism responsible for the decreasing x-concentration of Mn 2+ ions in the QDs? In our conception, it is reasonable to assume that the trapped-dopant mechanism does not account for a significant change of Mn-concentration during the Cd1-xMnxS dot growth by the melting-nucleation synthesis. Since there is a relative homogeneous distribution of Cd 2+ , Mn 2+ and S 2-species into the glass environment in each moment of the thermal annealing, it is expected to observe nearly constant Mn-concentration in the Cd1-xMnxS dot growth process. Furthermore, the trapped-dopant mechanism is generally dominant for QD synthesis based on liquid phase approach, as the colloidal chemistry, where the temperatures are generally below 350 ºC and, in some case, even as low as room temperature [3,53]. In contrast, the energetic argument related to the self-purification mechanism imposes a relative instability for the impurity species due to increasing formation energy for decreasing dot-size. In Mn-doped CdSe NCs, for example, it is known that the diffusion of Mn 2+ ions occurs at a synthesis temperature around ~550 K (277 ºC), due to this instability [54,55]. Therefore, we are convinced that the relatively high temperature used in thermal annealing of our sample (560 ºC) is able to provide enough energy to provoke impurity diffusion toward surface region, a site having stronger binding energy, or even to evaporate the magnetic impurity ions from the Cd1-xMnxS QDs. In other words, our results confirm that self-purification is the dominant mechanism that controls the doping in semiconductor QDs grown by melting-nucleation synthesis approach.

Conclusions
In conclusion, we have recorded optical absorption (OA), photoluminescence (PL), and magnetic circularly polarized photoluminescence (MCPL) spectra, as well as magnetic force microscopy (MFM) images, in order to investigate Cd1-xMnxS NPs that were synthesized in a glass matrix. Room temperature OA spectra revealed the growth of two groups of NPs with different sizes: QDs and bulk-like NCs, a result confirmed by MFM images. Several emissions were observed in the temperature dependent PL spectra of Cd1-xMnxS NPs, including those from deep defect levels that were attributed to two energetically different divacancies, VCd-VS, in the wurtzite structure. Moreover, the emissions from these deep defect levels were suppressed with increasing x-concentration, providing further evidence not only of the incorporation of Mn 2+ ions in the NPs, but also for the existence of deep divacancy defects VCd -VS. Therefore, we have demonstrated that the density of NP defects can be controlled by magnetic doping. From the temperature dependent PL spectra of these NPs, we have deduced, based on rate equation, expressions in order to describe the carrier dynamics between excitonic states of QDs and conduction band of bulk-like NCs. Fitting procedures with these coupled expressions achieved satisfactory agreement with the integrated PL intensity of both the QDs and bulk-like NCs provided activation energies of non-radiative channels observed in Cd1-xMnxS NPs.
Our results confirm that the magnetic doping, Mn 2+ ions localization, and quantum confinement play important roles on the magneto-optical properties of these NPs. The different behaviour observed between the two groups of NPs with different sizes, QDs and bulk-like NCs, were ascribed to a considerable change of exchange interaction between the carrier spins and the substitutional doping magnetic ions incorporated into the NPs. In addition, we have demonstrated that the relatively high temperature that was used in the thermal annealing of the samples provides enough energy to provoke magnetic impurity diffusion toward surface region of NPs. Therefore, for semiconductor QDs grown by the melting-nucleation synthesis approach, the doping process is dominated by the selfpurification mechanism. We believe that the main results of this chapter can motivate further investigations and applications of other systems containing DMS NPs.