1. Introduction

Iron-doped crystalline zinc selenide (ZnSe:Fe) is a well-known type of active medium for tunable mid-IR lasers, which have a broad range of applications [1–4]. Lasing is usually obtained under the quasi-resonant optical pumping of Fe²⁺(d⁶) ions substituting Zn sites. The crystal field splits the atomic term of the ⁵D Fe²⁺ ground state into the ⁵E and ⁵T₂ doublet (fivefold degenerate), which is used to create an inverse population. Moreover, the spin-orbit interaction leads to the appearance of an additional fine structure, clearly observed in low-temperature emission [5] and absorption spectra [6].

A significant factor that affects the lasing efficiency of ZnSe:Fe is the scattering/absorption caused by background defects in the active medium [7–9]. Therefore, understanding the formation mechanisms of the background defects that emerged during crystal growth and/or doping is essential for improving existing laser systems [10,11]. At the same time, the crystal field should be locally distorted in the presence of point defects, thus affecting the optical properties of the iron ions.

The foregoing serves as the basis for the luminescence diagnostics of crystals using transition elements as “probes,” which enable the fully optical detection of intrinsic point defects [12,13]. The convenience of this approach is governed by the fact that iron ions—as well as other transition elements—produce narrow zero-phonon lines (ZPLs), which are sensitive to local changes in the crystal field. This means that in the presence of stable complexes with the participation of iron ions and intrinsic point defects, the introduction of the latter into the crystal will be accompanied by the appearance of one or several narrow emission lines in the luminescence spectra. At a fixed concentration of doping iron, the intensity of these lines will reflect the concentration of the embedded defects. As far as we know, the existence of such luminescent markers of intrinsic defects has not been reported for ZnSe.

In this work, we present experimental results and ab-initio calculations that indicate the presence of a stable complex that produces a narrow ZPL at 0.681 eV and is formed by an iron ion and a point defect in the zinc sublattice in ZnSe. The discovered luminescent center is of interest as a potential “optical probe” for intrinsic point defects in ZnSe.

2. Experimental details

In this work, we have studied samples of polycrystalline ZnSe doped with Fe by thermal diffusion in a Zn or Ar atmosphere. The doping procedure is shown schematically in Fig. 1(a).

 

Fig. 1. (a) ZnSe doping scheme. (b) Photoluminescence spectra of doped crystals at temperature T=80 K. The arrow marks the atmospheric absorption associated with CO₂. (c) Raman spectra for ZnSe:Fe doped in a Zn atmosphere (red curves) and an Ar atmosphere (black curves) collected from the edge and the center of the samples.

Download Full Size | PDF

In the first stage, parallelepipedal ZnSe plates with dimensions of 10×3.5×2 mm were cut from a polycrystalline druse, which was obtained by chemical vapor deposition (CVD) and had a uniform microstructure with a grain size of 37 ± 4 µm. The surfaces were mechanically polished with diamond powder and washed with acetone and distilled water. Then, an approximately 1-µm thick iron metal film was deposited on one of the surfaces using electron-beam sputtering [14]. The samples were subsequently annealed in a Zn or Ar atmosphere at 1000–1100°C for 240 h. In the course of annealing, a partial transfer of material to the opposite surface of the sample was observed [14]. After annealing, the samples were cooled in the switched-off oven mode and then removed from the ampoules and subjected to mechanical polishing [14,15]. In the samples under study, the iron concentration gradually changed by an order of magnitude along the facets located perpendicular to the surface on which the iron film was deposited. For this reason, these facets are convenient for optical measurements in which one needs to trace the impact of the iron concentration on the photoluminescence (PL) spectra [11].

To measure the PL spectra in the 0.9–2.3 µm (0.54–1.37 eV) range, the samples were mounted in a flow-through helium cryostat and cooled with helium vapor. A cooled InGaAs photodiode was used for emission detection. A flow-through optical cryostat with a zinc selenide window was utilized to operate in the 2–5 µm (0.25–0.62 eV) range. The samples were attached to a copper cold finger. The detector consisted of a cooled InSb photodiode. The signal from the photodiode was recorded using an SR830 (Stanford Instruments) synchronous detector. The excitation source was a single-frequency (532 nm) laser with a power of up to 150 mW. The laser emission was focused on a spot of approximately 300 µm. The recombination emission spectrum was analyzed by a grating monochromator with a replaceable set of diffraction gratings and interference filters adapted for measurements in the near- and mid-IR spectral ranges.

Figure 1(b) shows the luminescence spectra of the ⁵T₂ → ⁵E transition, recorded at 80 K in the area with the maximum iron concentration. The dip near 0.3 eV is due to atmospheric CO₂ absorption. The obtained luminescence bands are slightly wider than those presented in [5], which we associate with a higher iron concentration (2×10¹⁹ cm⁻³ versus 5×10¹⁸ cm⁻³). Based on PL intensity measurements at the emission band maximum, we obtained the distribution profiles of the optically active iron in the crystals. The known doping profile in combination with the measured distribution profiles of the luminescence signal at a given wavelength was used to search for centers related to the Fe ions.

To measure the Raman spectra, a portable Raman spectrometer Ramix R532 (operating wavelength of 532 nm) coupled to an Olympus CX-41 microscope was used. The spectrograph provided a recording of the Raman spectra in the 150–4000 cm⁻¹ range with a resolution of ∼4 cm⁻¹. The measurements showed that the procedures used for doping the ZnSe sample did not lead to the formation of additional phases, even in the regions with maximum Fe concentration (Fig. 1(c)). All lines observed in the spectra correspond to first-order (LO, longitudinal optical phonon; TO, transverse optical phonon) and second-order (2TA, LA + TA; where TA is transverse acoustic phonon, and LA is longitudinal acoustic phonon) Raman processes with the participation of phonons characteristic of the ZnSe lattice [16]. The changes in the relative intensities of the LO and TO peaks observed in the spectra are caused by the random orientation of the crystal grains [17].

3. Formation conditions of the non-standard luminescent center

Figure 2(a) shows the low-temperature PL spectra in the 0.6–1.0 eV region for the source ZnSe (red curve), the undoped ZnSe sample annealed in an Ar atmosphere under conditions similar to those used for Fe doping (green curve), and a sample doped with iron in a Zn atmosphere (black curve) and one in an Ar atmosphere (blue curve). It can be seen from Fig. 2(a) that a narrow line with a maximum at 0.681 eV is observed in the PL spectrum of the iron-doped sample annealed in an Ar atmosphere. In the literature, there is a mention of the spectral emission lines that lie near 0.674 eV (5440 cm⁻¹), corresponding to the V³⁺ ion in ZnSe [18]. However, the emission spectrum of a ZnSe:V³⁺ center is very different from that observed in Fig. 2(a): in addition to the zero-phonon line at 0.674 eV, there is a bright and wide vibronic band.

 

Fig. 2. (a) Low-temperature PL spectra in the 0.6–1.0 eV region for the source ZnSe (red curve), the undoped ZnSe sample subjected to annealing in an Ar atmosphere under conditions similar to those used for Fe doping (green curve), and a sample doped with iron in a Zn atmosphere (black curve) and one in an Ar atmosphere (blue curve). (b) Profiles of the PL signal for the ⁵T₂ → ⁵E transition in the optically active Fe²⁺ ions (light blue curve) and the 0.681 eV line (dark blue curve). (c) Part of the Tunabe-Sugano diagram for the d⁴ shell of Fe²⁺.

Download Full Size | PDF

As seen in Fig. 2(a), no signs of the 0.681 eV line are observed in the PL spectra of the other three samples, which also suggests that the detected center could not arise due to accidental contamination with vanadium. On the one hand, the absence of the 0.681 eV line in the source sample indicates its connection with the doping procedure. One can see that annealing in an Ar atmosphere without a dopant does not lead to the appearance of this line (black curve in Fig. 2(a)). This observation directly indicates the relation of the 0.681 eV line with the doping iron. On the other hand, the detected line is not recorded in samples doped with iron in a Zn atmosphere, which indicates its relationship with the intrinsic defects of the zinc sublattice. As is known, metal vacancies in A₂B₆ compounds are characterized by relatively low formation energies [19,20]. Therefore, this type of intrinsic defect should dominate during long-term, high-temperature annealing in an Ar atmosphere. At the same time, when using a Zn atmosphere, the formation of Zn vacancies is noticeably suppressed and their equilibrium concentration should not be significant.

The relation of 0.681 eV emission line with iron is independently confirmed by the measurements of luminescence spatial profiles. Figure 2(b) shows the PL signal profiles for ⁵T₂ → ⁵E transition in optically active Fe²⁺ ions (light blue curve) and the 0.681 eV emission line (dark blue curve). The profiles were obtained perpendicular to the plane from which the doping was carried out. As can be seen from the figure, the profiles correlate with each other, thus indicating a relationship between the Fe²⁺ concentration and the concentration of centers responsible for the 0.681 eV line.

The level splitting of the Fe²⁺ ion by the tetrahedral field of the ZnSe lattice is presented in the form of a Tunabe-Sugano diagram in Fig. 2(c). It can be assumed that the detected line corresponds to the ³T₁ → ⁵T₂ transition. At the same time, the energy of this transition should be equal to the difference between the energies of the ³T₁ → ⁵E and ⁵T₂ → ⁵E transitions, which are 1.35 eV and ∼0.35 eV, respectively [21]. This means that the zero-phonon line for ³T₁ → ⁵T₂ is located near 1 eV, which is more than 0.3 eV higher than the spectral position of the detected line. It should be noted that a broad band with a blue edge near 1 eV is clearly observed in the emission spectra of doped crystals regardless of the atmosphere used for the doping procedure — see the area indicated by the dotted box in Fig. 2(a). It is worth noting, that in some cases competition of new luminescent center at 0.681 eV and typical ZnSe:Fe²⁺ emission is observed.

Thus, the detected center should be attributed to a complex defect, which comprises an iron ion and an intrinsic defect of the zinc sublattice. We exclude the possibility of an extended defect, since in this case the observation of a spectrally narrow line would be impossible due to the disorder inherent in extended defects.

4. Vibronic band and fine structure of a zero-phonon transition

Figure 3(a) compares the PL spectrum (blue curve) near the ZPL at 0.681 eV and the phonon density of states in ZnSe (black curve) [22], which allows one to identify the main peaks of the vibronic band for the detected center. Therefore, the STA(X, K) peak is a phonon replica with the participation of slow transverse acoustic phonons with quasi-momentum located at the edge of the Brillouin zone (X-K). The FTA(Σ-K) peak is due to the interaction with fast transverse acoustic phonons (from mainly the edge of the Brillouin zone, Σ-K). There is also a trace of LA and TO phonons, the quasi-momenta of which are located at the edge of the Brillouin zone near the X-K direction. TO(Γ) and LO(Γ) refer to the transverse and longitudinal optical phonons, respectively, with quasi-momenta near the Г point of the Brillouin zone.

 

Fig. 3. (a) PL spectrum (blue curve) near the ZPL at 0.681 eV and the phonon density of states in ZnSe (black curve). The zero energy corresponds to the position of the maximum of the ZPL. (b) Modification of the ZPL with temperature. (c) Comparison of the ZPLs at 5 K for the observed transition and the ⁵T₂ → ⁵E transition for Fe²⁺ ions. The fine structure of ⁵T₂ → ⁵E was taken from [5].

Download Full Size | PDF

In general, the described structure of the vibronic band is typical of intracenter transitions; on the one hand, the coupling with the lattice is relatively small, and on the other hand, the participation of not only center-zone phonons but also phonon states at the edge of the Brillouin zone is clearly recorded.

For the peaks indicated by arrows in Fig. 3(a), no pronounced features are observed in the ZnSe phonon density of states. This allows their identification as local modes. Taking into account that the Fe atom has a relatively small mass, the existence of low-frequency local modes (∼50 cm⁻¹ and ∼150 cm⁻¹) indicates the softening of bonds near the ion, the inner shell of which is responsible for the discussed electronic transition. The softening should be expected upon the formation of vacancies (e.g., [23]), which is consistent with the above identification of the discovered center.

Figure 3(b) shows the temperature dependence of the emission spectrum near the ZPL. It can be observed that an increase in temperature up to 20 K leads to the appearance of a short-wavelength satellite line at 0.683 eV, the broadening and shift of the main line, and the quenching of the overall emission intensity. The doublet structure observed at 20 K indicates the presence of at least two closely spaced sublevels in the state from which the transition occurs. In turn, luminescence quenching, significant broadening, and a shift of the main emission line — effects observed even at relatively low temperatures — are characteristic in the presence of strong coupling of the emitting state with low-frequency oscillations. According to the existing theories [24], the above-mentioned low-frequency local modes are likely candidates for such oscillations.

It is interesting to compare the structure of the observed transition with the fine structure of the well-studied ⁵T₂ → ⁵E transition for the Fe²⁺ ion (Fig. 3(c)). This transition is characterized by the presence of four peaks, which correspond to the allowed transitions from the lower state (Γ₅) ⁵T₂ to the ⁵E states (γ₁, γ₃, γ₄, γ₅) [5]. For the ZPL of the detected center, no signs of such a fine structure were recorded. This means that the final state of the system is different from the ⁵E state of the Fe²⁺ ion in a tetrahedral field.

5. Model of the emitting states

To develop a qualitative model of the emitting states, ab-initio calculations with the QuantumESPRESSO software package were performed [25]. The Fe_Zn center was modeled in a 54-atom periodic supercell. For the density functional calculation, we employed the Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional with a projected augmented-wave pseudopotential with the energy cutoff at 75 Ry and the charge-density cutoff at 500 Ry. For the integration over the Brillouin zone, the Γ-point approximation was used. Before calculating the electronic states, the atom positions at fixed cell dimensions (which were taken to be equal to the experimental ones) were fully optimized until the residual force on every atom did not exceed 0.001 Ry/bohr. The resulting stress was under 15 kbar. Additional calculations were subsequently performed in the resulting geometry using the HSE06 hybrid functional to clarify the positions of the impurity levels and the electronic bandgap. For pristine ZnSe, the initial PBE calculations yielded a bandgap equal to 1.3 eV, which is far below the experimental value of 2.8 eV. HSE06 calculations with standard parameters for the non-local part significantly improved the band gap value, which became equal to 2.5 eV. The precision of this method is given by a 300 meV difference between the calculated and experimental band gaps.

To determine the structural stability of the defects, we calculated the formation energy of substitutional Fe_Zn and zinc vacancies because (as mentioned above) both are relevant for describing the optical properties of the samples. The results obtained at the PBE level of theory are presented in Fig. 4(a). The formation energy of iron impurity in charge state q can be calculated by means of DFT methods:

 

Fig. 4. (a) Formation energy of the zinc vacancy (red dashed line) and substitutional iron Fe_Zn (blue line) in various charge states. (b) The Kohn-Sham orbitals of Fe_Zn in the spin-majority and spin-minority channels (marked by upward and downward arrows, respectively). The Fe_Zn d-orbitals in the vicinity of the bandgap (whose borders are marked by the CBM and VBM labels) are shown by dashed lines. Blue labels on the left represent the symmetry of the orbitals of the pristine ZnSe crystal at the Γ-point, while the red labels show the symmetry of the Fe_Zn levels.

Download Full Size | PDF

It is interesting to note that according to the calculation, the formation energy of Fe_Zn is negative in the whole range of attainable Fermi energies. At the same time, if the Fermi level were located near the midgap (0.65 eV in the PBE approximation), Fe_Zn should be present in the Fe²⁺ state. This is not surprising, taking into account that zinc and iron are isoelectronic in ZnSe. However, this can change drastically if a significant amount of vacancies is present in the sample. According to the charge state diagram in Fig. 4(a), the zinc vacancies in ZnSe are strong electron acceptors and are likely to be found in a double-negatively charged state. Therefore, when both types of defects are located near one another, the Fe_Zn inner shells can provide electrons for the vacancy, which is why the iron is prone to acquire an additional positive charge. Depending on the acquired charge, this process can significantly influence the optical properties of the iron ions.

It is interesting to compare our system with CdSiP₂ one with chalcopyrite structure [26,27] which also can be doped with iron. It was demonstrated that iron can substitute either silicon or cadmium positions and naturally accept various charge states [27]. In Fe-doped CdSiP₂ crystals two various optical absorption bands were found, one of which was ascribed to the absorption of Fe²⁺ center with the ZPL around 0.666 eV and the second one to the transition in the Fe⁴⁺ center with the ZPL around 1.158 eV [26]. Interesting to note that in contrast to ZnSe, the differentiating between two charge states of substitutional impurity is induced by crystallographic constraints on the two crystallographically different sites that the impurity atom might occupy. According to our model, in ZnSe, where all Zn sites are crystallographically equivalent, the corresponding difference is imposed by the presence in the environment of impurity atom the charged vacancy which acquires the electrons from impurity atom and converts it into doubly positively charged state. The resulting Fe⁴⁺ color center also has higher ZPL energy than its Fe²⁺ counterpart.

To investigate the optical properties of Fe_Zn, we draw its electron orbital diagram in the HSE06 approximation accounting for spin polarization (Fig. 4(b)). The tetrahedral crystal fields split each d-orbital of iron into two sublevels with E and T₂ symmetry occupied by six d-electrons (in the case of Fe²⁺). Due to the exchange interaction, these two sublevels—corresponding to the spin-up and spin-down channels—are significantly shifted with respect to each other. Therefore, in the spin-up channel, the Fe_Zn levels are close to the valence band maximum, while in the spin-down channel, they are close to the conduction band minimum. This leads to a significant spin polarization of the optical center and fulfillment of Hund rule—5 electrons occupy the impurity levels in the spin-up (majority spin) channel, and only one is present on the E level of the minority spin channel. Due to the contribution of the exact exchange-correlation term in the hybrid HSE06 functional, the partial occupation of the degenerate levels leads to further splitting, which is shown in Fig. 4(b). As the Fe²⁺ levels in the majority spin channel are occupied, all optically active transitions take place in the minority spin channel. There are two possibilities—a transition from the partially occupied E level to the Fe²⁺ T₂ level or one to the conduction band minimum A (slightly more energetically favorable). Surprisingly, it turns out that both transitions produce approximately equal (within a 50-meV margin) vertical excitation energies of 550 meV. The relaxation energy in both cases is also small and equals 30 meV.

An increase in the positive charge of the Fe²⁺ ions (for example, due to the presence of a nearby zinc vacancy) leads to the depopulation of Fe-related levels in the ZnSe bandgap. The Fe³⁺ ion is optically inactive (all levels in the spin-down channel are empty), but for the Fe⁴⁺ ion, the hole is formed in the T₂ level in the majority spin channel. Therefore, optical activity is possible, and the electron transition from the fully occupied impurity level (E) can take place. The corresponding quantum energy of ∼0.6 eV, estimated using the data in Fig. 4(b), fits well the spectral position of the detected ZPL at 0.681 eV. Consequently, we believe that the scenario described above is plausible for the ZnSe:Fe sample synthesized in an Ar atmosphere.

As noted above, an alternative interpretation of the detected luminescent center could be the emission of tetrahedrally coordinated Fe⁴⁺ ions (without the participation of vacancies). Nevertheless, our estimates of the spectral position of the ⁵E-⁵D emission line of isolated Fe⁴⁺ ion, based on the Tanabe-Sugano diagrams, give a value of about 0.45 eV. This is noticeably lower than the registered line at 0.681 eV. Thus, the distortions of the crystal field in which the Fe⁴⁺ ion is located seem to play a significant role. Therefore, we consider the association of the discovered line with the vacancy complex as more reliable.

6. Conclusion

In ZnSe:Fe crystals, we report a new luminescent center, which forms a narrow ZPL near 0.681 eV at a 5-K temperature. The fine structure of the zero-phonon transition excludes any connection between the discovered center and the ⁵E level of Fe²⁺. In addition to the ZPL, the spectrum contains phonon replicas caused by the interaction with the ZnSe lattice and two satellite lines associated with low-frequency localized vibrations. The presence of the latter indicates the existence of bond softening, an effect characteristic for vacancy centers.

The connection between the discovered center and the Fe dopant is confirmed by the correlation of its intensity with that of the well-known ⁵E → ⁵T₂ transition of Fe²⁺ ions, as shown by experiments that measure the luminescence intensity profiles. In turn, the key role of intrinsic point defects in the zinc sublattice is confirmed by experimental investigations of different ZnSe crystals doped in a variety of atmospheres. Using ab-initio calculations, we propose a simple model of emitting states produced by the spin majority channel of an Fe ion attached to a Zn vacancy.

The data obtained indicate the presence of a luminescence-active vacancy complex, which can be used as an “optical marker” for intrinsic point defects in crystalline ZnSe.