Photoluminescence of CdTe:In the spectral range around 1.1 eV

Journal of Luminescence 177 (2016) 71–81

Contents lists available at ScienceDirect

Journal of Luminescence journal homepage: www.elsevier.com/locate/jlumin

Full Length Article

Photoluminescence of CdTe:In the spectral range around 1.1 eV J. Zázvorka n, P. Hlídek, R. Grill, J. Franc, E. Belas Institute of Physics, Faculty of Mathematics and Physics, Charles University in Prague, Ke Karlovu 5, 121 16 Prague, Czech Republic

art ic l e i nf o

a b s t r a c t

Article history: Received 27 May 2015 Accepted 11 April 2016 Available online 14 April 2016

Temperature and excitation dependences of photoluminescence (PL) spectra of a slightly In-doped CdTe crystal were investigated, particularly in the spectral range of the “1.1 eV” PL band, i.e. in the interval of 0.9–1.25 eV. Both above-bandgap and below-bandgap excitations were employed. Three components of the “1.1 eV” band with the PL maxima at around 1.19 eV, 1.13 eV, and 1.03 eV were distinguished. The component with a PL maximum at 1.19 eV vanishes above 40 K. It can be attributed to a recombination of a thermally unstable state localized at a defect. The excitation spectrum of the second component with a maximum at 1.13 eV indicates that a transition in a deep localized center takes place. The “1.13 eV” component prevails at higher temperatures with below-bandgap excitation. The “1.03 eV” component can be related to an effective recombination center, probably to a complex of deep donor and deep acceptor. This component is the strongest one at higher temperatures and under above-bandgap excitation. Our present knowledge does not allow us to attribute the PL spectra to particular defects. & 2016 Elsevier B.V. All rights reserved.

Keywords: CdTe Photoluminescence Deep levels Defect structure

1. Introduction X-ray and gamma-ray detectors find various applications in medical, defense and security industries. CdTe or CdZnTe crystal are used as materials for production of detectors because of their unique properties such as a high average atomic number, great resistivity at room temperature achieved due to a wide bandgap and an effective compensation of defects, a high mobility-lifetime product of charge carriers etc. The most important defects in CdTe single crystals grown by an appropriate method (Bridgman or vertical gradient freeze or traveling heater method etc.) are cadmium vacancies that have to be compensated. Either chlorine or indium are commonly used for the compensation (typical dopant levels 1016 to 1018 cm
3). Nevertheless, other defects are also important for detector quality that depends extremely on the concentration of defects and impurities creating electronic energy levels in the bandgap. The deep electronic levels can help to stabilize high electrical resistance by pinning of Fermi level near the middle of the bandgap. On the other hand, properties of the detector materials are negatively influenced by a reduction of the mobility-lifetime product, caused by recombination centers. Moreover a space charge created by charging of the defects with deep levels causes an undesirable electric polarization of the detectors. n

Corresponding author. Tel.: þ 420 221 911 320. E-mail address: [email protected] (J. Zázvorka).

http://dx.doi.org/10.1016/j.jlumin.2016.04.019 0022-2313/& 2016 Elsevier B.V. All rights reserved.

Photoluminescence is one of the important tools for material characterization and mainly for the study of shallow and deep energy levels created by defects, impurities and their complexes. The luminescence spectra measured under excitation by photons of energies above the bandgap (1.606 eV at 4 K) in the spectral range of 1.2–1.6 eV represent standard characteristics of samples. The luminescence spectra measured in the range below 1.2 eV that can be used for investigation of deep levels in the bandgap are published less frequently. Numerous studies done using low-temperature photoluminescence and/or cathodoluminescence (CL) since the 1950 have shown that a broad band in the spectral region around 1.1 eV is often observed in single crystal, high resistivity materials, see Ref. [1]. Intensity and/or position of the “1.1 eV” PL band can be influenced by various treatments, e.g.:

 annealing in Cd vapor [2,3];  plastic deformation [3–5];  dislocation – e.g. dislocation induced by inclusions/precipitates [6,7];

 concentration of extended defects [8];  composition of an etching agent [9]; and other types of treatments (like electron bombardment, ion implantation, doping with some elements like Fe, V, Sn, etc.). PL spectra changes in this spectral range can be given in relation to electric conductivity type and its value [10–12]. Several types of the 1.1-eV bands were reported in [13]. A decomposition into two or three components by a fitting procedure was used in [2,14,15].

72

J. Zázvorka et al. / Journal of Luminescence 177 (2016) 71–81

Photoluminescence of semiconductor samples is commonly excited by photons with energy higher than the bandgap of the material, ℏωEXCIT 4 EG . Below gap excitation ℏωEXCIT o EG has been reported very rarely. The investigation of donor–acceptor pairs with different space distance of the acceptor from the donor (so called selective pair luminescence – SPL) has to use the below gap excitation. The SPL has been reported for CdTe crystals ([16] and references therein). An application of excitation at photon energies at the Urbach tail of the absorption edge (several meV below the bandgap) has been proven to be very useful because of PL excitation in the bulk of samples. Therefore, a higher PL signal can be obtained due to partial elimination of surface recombination [12,16]. To our knowledge, no paper has reported an investigation of PL in CdTe single crystals using an excitation ℏωEXCIT o1.4 eV. In this paper, we fix our attention on photoluminescence in the spectral range of 0.9 eV up to 1.25 eV measured on single crystal CdTe doped with a very low In concentration. We have investigated both temperature and excitation dependences of components contributing to this “1.1 eV” band. Excitation by photon energy both above the bandgap and below the bandgap was utilized. Commonly used above-bandgap excitation affects only the surface layer of the sample. Surface (nonradiative) recombination can affect the results. On the other hand, below-bandgap excitation reaches deeper into the bulk of the samples and more information on defects inside the sample can be obtained, including excitation spectroscopy. We show that at least three components (hereafter labeled as bands 9, 10, 11, respectively) with different temperature and excitation dependences contribute to the “1.1 eV” band in this sample. Section 3 of the paper reports on some experimental results obtained on the studied sample. Section 4 contains an attempt to interpret the experiments. Some results of a simple model related to position and width of a PL band are mentioned in Appendix A1. Appendix A2 compares our results with findings of other investigations. Although our current investigation cannot resolve the task of an assignment of PL bands, a short review of possible defects that have appeared in published papers as candidates for the found deep levels is given in Appendix A3.

The sample was placed in the helium flow cryostat Optistat (Oxford Instruments). The luminescence was excited either by a red line (638 nm, 1.94 eV) coming from a semiconductor laser Radius (Coherent) or from a tunable Ti:Saphire laser Millennia pro 5sJ (Spectra Physics) in the range of 1.25–1.7 eV. The luminescence spectra were measured by the Bruker IFS 66s FTIR spectrometer equipped with a CaF2 beamsplitter and a Si photodiode and cooled Ge and InSb detectors. Excitation radiation was blocked in front of the spectrometer by edge low pass filters Semrock 980 nm (1.27 eV), 830 nm (1.49 eV), 808 nm (1.53 eV) and 785 nm (1.58 eV), respectively. Only a color glass filter in front of the Si detector was used to eliminate the scattered light coming from the red laser.

3. Results 3.1. Sample PL characterization at ℏωEXCIT 4 EG The studied sample exhibits typical structures for CdTe doped with indium usually observed in the luminescence spectra with above-bandgap excitation, see Fig. 1 and Table 2. The luminescence bands marked as 1–6 and 8 are accompanied by their wellrecognized optical phonon replicas. Dominant peak 4 is the so called C-line at 1.584 eV. The "1.45 eV" band is not single; at least two components can be distinguished. The well-resolved free and Table 1 The most abundant impurities as determined by GDMS. Element

Conc. [ppb]

Element

Conc. [ppb]

Li B Na Al Si P S

o 1.1 o5 1 12 o5 2 80

Fe Cu Zn Ag In Sn Sb

o 25 7 8 o 15 o 20 o 20 o 15

2. Experimental The In doped CdTe sample was cut off from a single crystal grown by the Vertical–Gradient–Freeze (VFG) method. The sample dimensions were 6  6  2 mm3. The intentional In-doping level was rather low: 5  1015 cm
3 in the charge. The sample was mechanically polished using a 1 mm alumina (Al2O3) abrasive and then etched by immersion into a 3% Br–methanol solution for 2 min. In spite of the low dopant concentration, the material is semi-insulating with a resistivity value of 9  108 Ω cm, Hall electron concentration n ¼107 cm
3, calculated Fermi level position EF ¼ EV þ 0:861 eV and mobility 800 cm2/V s; all values at room temperature. X-ray detectors made from the material were of average quality with mobility-lifetime product of electrons mτ E5  10–4 cm2/V. Chemical analysis (glow discharge mass spectroscopy–GDMS) shows a rather high concentration of sulfur (80 ppb), higher than the indium concentration in the sample (see Table 1). 1 ppb corresponds to concentration 2.94  1013 cm
3. Rather a low concentration of shallow acceptors was found (Li, Na, P if incorporated as substituents, but donors if in interstitial positions). Al should create shallow donor levels similar to In, whereas Zn and S are isoelectronic impurities (both ZnxCd1
xTe and CdTe1
xSx are ternary crystals), Cu and Ag are usually “deeper” acceptors. Fe, Sn, Sb have been reported to be connected with deep levels.

Fig. 1. Luminescence spectra in the range of exciton recombination (1,2,3,4), donor–shallow acceptor recombination (5,6), dislocation-bound exciton recombination (7) and recombination at A-centers/deeper acceptors (8). Changes in the spectral region of shallow acceptors recombination with increasing temperature are outlined in the inset.

J. Zázvorka et al. / Journal of Luminescence 177 (2016) 71–81

bound exciton recombination spectra are commonly attributed to samples of “good quality”. On the other hand, detection of a Y-line shows significant effect of lattice deformation (dislocations) and rather manifold spectra of deep levels indicate the presence of various types of defects. Photoluminescence at lower photon energies as detected by the InSb detector is shown in Fig. 2. The components of the "1.1 eV" PL band marked as 9, 10 and 11 are examined in more detail in the following parts of the paper. The intensity of band 12 (0.9 eV) is highly influenced by a surface damage as shown in [17] dealing with surface treatment and PL. It will be proven throughout this paper that the decomposition is necessary for the Table 2 Luminescence bands observed in CdTe doped with In at 4 K. See [17] for references. Position at 4 K [eV] 1

1.596

2 3

1.593 1.589

4

1.584

5

1.546

6

1.539

7

1.472

8

1.451

9 10 11 12 13 14

1.458 1.19 1.13 1.03 r 0.9 0.64 o 0.55

Assignment

FE; dip due to reabsorption by free excitons/excitonic polaritons (D0, X); recombination of excitons bound to neutral donors (A0, X); recombination of excitons bound to neutral acceptors C-line; recombination of excitons bound to complex defects containing In DAP-like; recombination in pairs donor–acceptor complex containing cadmium vacancy; and/or isoelectronic impurity (like oxygen and/or possibly sulfur) DAP; recombination in donor–acceptor pairs, substitution possibly NaCd Y-line; recombination of excitons at specific type of dislocations DAP recombination in pairs donor–deeper acceptors A centers (Cd vacancy þ In donor) and Cu acceptors (e, A0) recombination observed at 30 K ?, see Section 4 ? ? ? ? ?

Fig. 2. PL spectra recorded by InSb detector with excitation at the absorption edge.

73

explanation of excitation spectroscopy and temperature dependence of the observed spectra. 3.2. Temperature dependence at ℏωEXCIT 4 EG The excitonic bands (1–4) are shifted to lower energies with increasing temperature; the shift is given by temperature dependence of the bandgap EG ðT Þ [18]. The intensity of bound exciton recombination decreases monotonically with the increasing temperature; band 3 (A0, X) diminishes at the highest rate. Free exciton (FE) structure 1 weakens slowly. The replica of FE (FE – 2LO) even reaches its maximum at 20 K, as can be seen in the inset in Fig. 1. The recombination in donor–acceptor pairs (DAP) is a natural behavior of bands 5, 6 and 8:
   D0 ; A0 - D þ ; A
þ ħωLUMI ; and it is transformed to recombination of a conduction electron with an acceptor (e, A0):
   e
þ D þ ; A0 - D þ ; A
þħωLUMI : The corresponding peaks are shifted to higher energies with temperature increase up to 30 K, as indicated by shift 5-5* in the inset in Fig. 1 and by a more pronounced shift of band 8. PL spectra in the region of the “1.1 eV” band are depicted in Fig. 3 for several temperatures. Spectral position of bare PL maxima at various types of excitation (including the ℏωEXCIT 4 EG excitation) in dependence on temperature are shown in Fig. 4. This can be explained just by different temperature dependences of separate components intensities. The spectra presented in Fig. 3 can be decomposed by fitting procedures into four Gaussian bands, see inset in Fig. 3. The intensities of both components 9 and 10 reach their maxima near 13 K. It resembles the temperature dependence of the free exciton replica. This can be explained by a thermal escape of carriers from shallow levels supplying the radiative recombination channels with non–equilibrium carriers. Investigating the spectra at temperatures above 30 K, it becomes visible that peak component 11 is less quenched than components 9 and 10 and can be

Fig. 3. PL spectra recorded by the Ge LN detector with above-bandgap excitation. Decomposition of the spectrum at 24 K is presented in the inset.

74

J. Zázvorka et al. / Journal of Luminescence 177 (2016) 71–81

Fig. 4. Position of the luminescence maxima in the spectral range 1.0–1.2 eV versus temperature for several exciting photon energies. Temperature dependence of EG
0.482 eV and EG
0.577 eV are plotted as well.

Fig. 6. PL excitation spectra-contributions of components 9, 10, 11 and 12 to the luminescence at 4.6 K obtained by fitting procedures.

are changed as well. Very different characteristics for components 9 and 10 are revealed contrary to excitation at absorption edge and at higher excitation energies. The components 9 and 10 become well-distinguishable at excitation below 1.5 eV. In various ranges of the excitation photon energy one of the two independent bands dominates as can be seen in Fig. 5, where the positions of the bare PL maxima are depicted as a dependence on the exciting photon energy. Component 10 (PL near 1.13 eV) becomes stronger in a rather narrow interval of excitation photon energies 1.34–1.42 eV, whereas component 9 (PL near 1.19 eV) prevails outside of this interval as shown in Fig. 6, where results of fitting decomposition of the spectra are shown. A noteworthy threshold for components 9 and 10 at 1.45 eV coincidences with ionization energy of “deeper” acceptors like A-centers or Cu impurity. 3.4. Temperature dependence at ℏωEXCIT o EG

Fig. 5. Position of the bare luminescence maxima (in the spectral range of components 9 and 10) versus exciting photon energy at 4.6 K.

regarded as an independent band. Both components 9 and 10 are thermally quenched quite rapidly; they become very weak above 30 K. On the contrary, the intensity of the “1.4 eV” band decreases monotonically with increasing temperature under ℏωEXCIT 4 EG excitation. 3.3. Excitation spectroscopy at 4.6 K, ℏωEXCIT oEG At excitation well below the absorption edge, the intensity of luminescence rapidly decreases with the drop of exciting photon energy due to the fall of the absorption. The forms of the spectra

The shapes of the PL spectra are changed with temperature (see Fig. 7 for ℏωEXCIT ¼ 1.39 eV). Component 10 is dominant at a low temperature, then contribution of component 9 reaches its maximum at 20 K and a rapid fall down of intensity of component 9 comes with increasing temperature. It is demonstrated also by a shift of the luminescence maxima displayed in Fig. 4 where spectral positions of the measured PL maxima (no decomposition) are shown in dependence on temperature and excitation photon energy. In the case of excitation at the absorption edge and at higher photon energies, the “1.4 eV” band (not shown in Fig. 4) and component 11 of the 1.1 eV band are dominant above 30 K, whereas the excitation ħωEXCIT oEG induces component 10 as the main contribution above 40 K. Results of fitting procedures for excitation photon energies 1.32 eV and 1.39 eV are shown in Fig. 8. Moreover, the thermal activation of component 10 becomes apparent with excitation 1.32 eV oħωEXCIT, MAX E 1.38 eV in the case of insufficient exciting photon energy to reach the maximal PL of component 10, i.e. below 1.36 eV. No such effect appears at higher excitation photon energy. Activation energy estimated for very narrow interval at 24 K is 40 meV.

J. Zázvorka et al. / Journal of Luminescence 177 (2016) 71–81

75

Fig. 9. One of the possible modifications of the schematic energy diagram of the simplest model of a localized center with two localized electron states (ground GES and excited EES). A relaxation of atomic arrangement comes after the optical excitation (red arrow) and the peak of the PL band is represented by the blue downward arrow. NRR tands for nonradiative relaxation. E01 represents the difference between the excited electronic state and the ground electronic state, both in the case of relaxed atomic arrangement. ΔEINT1 indicates activation energy (height of a barrier) for internal thermal PL quenching. The red dashed arrow shows the excitation possible at a higher temperature when even lower photon energy can excite the electronic state. The Escape and Capture arrows symbolize the possibilities of charge changes of the center. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 7. Luminescence spectra excited at 1.39 eV.

Fig. 8. Temperature dependence of fitted intensities of the respective components of the luminescence excited at 1.94 eV, 1.39 eV and 1.32 eV. Component 10 at excitation 1.32 eV is four times amplified.

pairs with shallow acceptors (spectral range at 1.5 eV) and in the “1.4 eV” band (at least in crystals of sufficient quality at low temperatures), see Fig. 1. The PL intensity of the respective replicas is given by the Poisson distribution with the Huang–Rhys factor S being a parameter characterizing electron–phonon interaction. It is expected that the electron–phonon interaction is stronger for deep, well-localized states of defects. Thus ZPL energy can be remarkably different from the peak of the PL band. Usually no replica structure is observed for deep defect levels including the “1.1 eV”. No band with replica pattern was reported in published papers on CdTe in this spectral range. Therefore, the S parameter (needed for estimation of the difference between the position of the PL peak and ZPL) cannot be determined directly. An important parameter is the width of PL bands. However, there are some other contributions to the band broadening like fluctuations of structure and potential in the locality of the defect. Some comments on the broadening are given in Appendix A1. The simplest energy diagram for both the excited and the ground state is shown in Fig. 9. It uses the Franck–Condon principle (vertical arrows represent optical transitions with no phonon emission or absorption). The diagram is drawn for exciting photon energy lower than ionization energy of the ground state ΔEION0 . The optical excitation is indicated by the red upward arrow (in principle this energy should be observable in excitation spectra PLE). Then a relaxation of atomic positions to a new equilibrium position happens. A radiative emission is drawn as the blue downward arrow and it corresponds to the PL peak. The energy of the zero-phonon line (ZPL) can be approximately identified with the energy difference between the excited and ground states of a defect in equilibrium (E01 in Fig. 9). As one can see in this oversimplified model, the probable value for E01 is located in the interval between the PL peak and the PLE band. Increasing the excitation photon energy a regime of center ionization is realized (bound-to-band transition). This means that:

4. Discussion

 either the system of energy levels is changed and the center

4.1. Spectral position of PL peaks and PLE spectra

 or a charge recapture happens and a relaxation either by a

falls out of this absorption/emission channel,

radiative or nonradiative process takes place. It is widely accepted that PL bands are composed of phonon replicas of zero phonon lines (ZPL). The replicas are welldistinguished in the case of recombination in donor–acceptor

Of course, a close link of PL and PLE is then broken. Whereas a relatively narrow excitation band is expected for an inner excitation

76

J. Zázvorka et al. / Journal of Luminescence 177 (2016) 71–81

of defect states, a threshold of excitation spectra for the ionization of deep levels is expected near the energy difference between the defect and bottom (top) of the CB (VB), whereas the maximum of the absorption is expected at much higher photon energy. In an oversimplified model of photoionization by Lucovsky [19] (the case of extremely localized deep level state), the maximum would appear at ℏωEXCITMAX  2EIONIZ . Of course, such a maximum in the PLE spectra is overridden by much more intensive interband absorption. The below-gap excitation produces a low excitation in the bulk of the sample (because of the low absorption coefficient) and processes like internal excitation inside defects or ionization of defect states (e.g. A
þhν-A0 þe
for acceptor) or both can dominate. Therefore, the PLE local maximum observed for component 10 in Fig. 6 is a strong argument for the opinion that the PL component 10 is caused by an internal transition (bound-tobound). Finally, the mostly used experimental implementation is bandto-band excitation ℏωEXCIT 4 EG A simple version of processes in a donor–acceptor pair at above-gap excitation is outlined illustratively in Fig. 1 in Ref. [20] where repeated captures/escapes of the charge are taken into account as well as internal PL emission and relaxation. Generally for bulk crystals with a commonly treated surface, the excitation spectra reach their maxima in the spectral region of the absorption edge (Urbach tail); then sharp structures due to absorption/reflection spectra of excitons are observed. The PLE spectra in the region of higher ℏωEXCIT are not very spectrally dependent and PL is reduced because of surface nonradiative recombination. 4.2. Temperature and excitation dependence As stated in the Introduction, the absolute majority of PL measurements is realized at above-bandgap excitation, where generation of electron–hole pairs in the respective bands (CB, VB) is dominant. A non-equilibrium population of defect states with energy levels in the energy gap is created by a capture of the electrons and holes. With increasing temperature, a PL intensity decrease (thermal quenching) appears. Two main limit mechanisms can be distinguished: 1. The quenching is caused by the thermal excitation of an electron (hole) bound to the defect that enables to overcome a potential barrier ΔEINT1 given by the dependence of total energies of the excited and ground states on atomic space arrangement as shown in Fig. 9. This mechanism is also possible in the case of above-gap excitation; see Fig. 1b in Ref. [20]. The carriers in VB/CB are captured in a defect state and then they can recombine nonradiatively by breaking the barrier ΔEINT1 . A multiple phonon emission assists the relaxation/recombination. Because this process can proceed in one defect, this mechanism is reported as the “one-center model” or “internal recombination model” (the schema proposed by Seitz and Mott). 2. The quenching is due to a thermal escape of the carriers from the defect state back to the energy bands, e.g. escape of a hole from acceptor A0-A
þh. The escaped carriers can be captured by other defects where a recombination (nonradiative or radiative) proceeds. The carriers in CB/VB constitute a common reservoir for various recombination channels and a competition of individual channels comes into play. This is sometimes associated with the effect when thermal quenching of a PL band is accompanied with an increase of PL in another band. This can result in a non-monotonic temperature dependence of PL intensity: a thermal activation of the intensity at low temperatures in the latter band turns into the thermal quenching at higher temperatures. This type of the quenching mechanism is

called the “multi-center model” or the Schön–Klasens mechanism [20]. Temperature dependence of PL in an acceptor band using a three-level model (one shallow donor, one acceptor and one nonradiative recombination center near the middle of the bandgap) under above-gap excitation was extensively discussed by Reshchikov [20] for wide-gap semiconductors and for various assumptions on the ionization energies of donors and acceptors, their concentrations, capture coefficients of the carriers and generation rates of electron–hole pairs. The “multi-center models” result in a great variety of PL temperature dependences: some of the dependence corresponds to an ingenuous concept that “activation energy” of the quenching should be equal to the ionization energy. However, sometimes very different behavior (much lower “activation energy” for the thermal quenching) can be obtained, particularly when concentration and/or capture coefficients for the middle-gap recombination center are high compared to the values for the donors and acceptors. It is the case of high resistivity samples with Fermi level near the middle of the bandgap and a result of the above mentioned common reservoir of carriers for various recombination processes. A possible scheme of some basic processes in our sample is drawn in Fig. 10. The luminescence transitions are supposed to be connected with electrons in the CB or with electrons bound to states near the CB bottom. A similar chart can be sketched for holes at the VB top as indicated for #9. We assume that bands 8– 11 are independent, i.e. that they are created by different defects/ impurities. Of course, the population of band states and localized states is given by the competition of radiative and nonradiative recombination channels and by the capture and escape of electrons (holes) from traps. 4.3. Comments on individual components 4.3.1. Component 9 Components 9 and 10 are not well-resolved at ℏωEXCIT 4 EG and band 9 and band 10 dominates at temperatures below 30 K. On the contrary, the components can be well distinguished at the excitation ℏωEXCIT o EG (Figs. 5–7). This encouraged us to decompose components 9 and 10 by a fitting procedure, see inset in Fig. 3. We can compare properties of the component 9 with the band 8. It is commonly accepted that band 8 (“1.4 eV band”) is caused by a recombination in shallow donor–deeper acceptor pairs at low temperatures and by a recombination of an electron in CB with a hole localized at the acceptor at higher temperatures. The position

Fig. 10. Some of possible processes: a) excitation ℏωEXCIT 4 EG ; b) excitation at absorption edge; c) excitation ℏωEXCIT o EG , ionization of the deep level; d) internal excitation of component 10. Black arrows outline charge carriers relaxation and transport including capture; NRR symbolizes nonradiative recombination. Inclined arrows #8 and #11 indicate possible DAP transitions.

J. Zázvorka et al. / Journal of Luminescence 177 (2016) 71–81

of the PL maximum is associated with this modification; a shift of several meV is observed (7 meV for our sample). On the contrary, no shift of several meV to a higher energy was recognized in component 9 with increasing temperature. At above-gap excitation, thermal quenching of the PL intensity of component 9 is much faster than that of the whole band 8 (together both the DAP and (e, A0) recombination, see Fig. 8). The temperature dependence of the PL intensity of component 9 could resemble rather only the component DAP of band 8 (that is effectively quenched above 20 K) than a sum DAPþ(e, A0). Thus component 9 can be caused by the transition from an excited state that is activated by carrier capture. If we follow the scheme given in Fig. 1 in Ref. 20 two competitive quenching mechanisms described above and denoted as 4 and 5 in the figure can occur. 1. Escape of the carrier to the CB follows the excitation, i.e. the excited state is very effectively thermally depopulated at temperatures above 20 K. Either a shallow “one-electron” trap state with a level near the CB bottom (VB top) or a metastable state (including a possibility of the so called “negative-U center”) should be taken into consideration. 2. Or there is a very low energy barrier for nonradiative recombination inside the center resulting in an effective thermal quenching over the potential barrier. Of course, both mechanisms can be combined. Taking into account that the steep quenching of component 9 is accompanied by noticeable changes in other PL components (see Fig. 8) we would prefer the possibility of an intensive exchange of the charges with the CB (or VB). A threshold in PLE spectrum near 1.45 eV is recognizable in Fig. 6. It is the energy where ionization of deeper acceptors (Acenters, Cu impurities) can make the concentration of carriers in bands higher. It is possible that defects connected with PL component 9 capture the electrons preferentially giving raise to excited states of the defect very near the CB bottom. Thermal depopulation of the excited state with increasing temperature causes the rapid PL decrease. 4.3.2. Component 10 The dependence of the PL maxima positions (Fig. 5) and results of the PL spectra decomposition procedures (Fig. 6) on excitation photon energy can be explained using the simplest model with a strong electron–phonon interaction (at least at low photon excitation energy, process (d) in Fig. 10), scheme depicted in Fig. 9. No charge transfer between the CB or the VB and the defect center may be assumed. Thus the excitation in the localized center can be while excitation at dominant for ℏωEXCIT o 1.45 eV, ℏωEXCIT 4 1.45 eV can cause photoionization (process (c) in Fig. 10). The photoionization is followed by a capture of carriers into the excited state of the center and then PL emission in component 10 appears.

77

Assuming: 1) main broadening of the PL band is given by the phonon emission mechanism; 2) peak position for absorption (PL excitation) at 1.38 eV follows the symmetry rule concerning Stokes–anti-Stokes transitions; 3) both excitation and emission processes involve the same ground and excited electron states but in changing atomic arrangement. The ZPL position can then be estimated to be at 1.26 eV, i.e. the difference EZPL
ℏωPEAK  6  ℏωLO : In Fig. 8 we can see a striking difference between temperature dependences of the intensity of component 10 measured at excitation near the maximum of the PLE spectra (ℏωEXCIT ¼ 1.39 eV) and measured below the maximum (1.32 eV). It is possible to explain the difference by a thermal activation of the ground state; see the dashed upward arrow in the energy scheme of a localized center in Fig. 9. The thermal energy supplies a contribution to the lacking optical excitation energy. The activation energy 0.04 eV for the PL growth near 24 K at excitation of 1.32 eV was determined from the steepness of the temperature dependence. It is in a reasonable agreement with the difference 0.07 eV of the two excitations (considering the simultaneous thermal quenching). Component 10 is dominant at higher temperatures under below-bandgap excitation when a concentration of nonequilibrium carriers in the CB and VB is low. The excited state of the related defects changes the charge with much lower probability than component 9. Component 10 wins a competition for charge carriers under the condition of low optical generation rate at higher temperatures; the excited state is deeper than that of component 9. 4.3.3. Component 11 Unlike components 9 and 10, component 11 can be observed up to relatively high temperatures under near- and above-EG excitation when carriers in both energy bands (VB and CB) are generated with a high concentration (in comparison with below-bandgap excitation). Component 11 could be attributed to an effective recombination center. Relatively high concentration of both electrons and holes seems to be the condition for dominance of this mechanism. 4.4. PL peak position, width of PL band and PL quenching activation energies The spectral position of ZPL energy could be in principle determined using the width of the PL band. A simple model of broadening given by Huang–Rhys factor S and by Gaussian broadening of individual replicas is given in Appendix A1. A very tentative estimation of ZPL energy intervals is given in Table 3. The estimation was based on an assumption that both Poisson-like and Gaussian-like broadening contributes by a similar

Table 3 Activation energies deduced from PL thermal quenching at relatively high temperatures at above-gap excitation compared to estimated ZPL energies at 4 K. The values 1.430 and 1.450 are for the stronger component of the band 8. Component 8 9 10 11 12

temperature range [K]

activation energy [eV]

ħωPEAK [eV]

FWHM [eV]

ħωZPL [eV]

71–115 120–145 36–60 50–80 125–145 60–125 125–145

0.073 0.077 0.068 0.068 0.095 0.070 0.078

1.430

0.075

1.450 7 0.001

1.19 1.13 1.03 0.83

0.125 0.162 0.154 0.172

1.277 0.04 1.26 7 0.01 1.147 0.08 1.0 7 0.1

78

J. Zázvorka et al. / Journal of Luminescence 177 (2016) 71–81

degree. It should be noted that the problem of ZPL determination from the PL spectra is usually simply “solved” by a choice ℏωPEAK ¼ ℏωZPL , i.e. the (unrevealed) assumption S ¼ 0 is used and a dependence of electron energy levels on atomic arrangement is disproved. Moreover, an assumption that the luminescence is due to bandto-bound transitions is often “automatically” accepted and the peak energy of the PL spectrum is believed to be directly the ionization energy. This may not be true in many cases and may be the source of differences between results of photoluminescence and other methods working with thermal excitation of carriers (e.g. PICTS). In some cases an apparent but improper agreement may be observed. The activation energy of thermal quenching is also assumed to be the ionization energy in an oversimplified concept. Both experimental results (see Section 3.2) and model calculations [20] show that this is not correct, particularly for high resistivity samples. This inconsistency can be a source of differences between results various measurements (e.g. PL, PICTS, etc.). The temperature dependences of the integrated PL intensity are complex, so a fitting using a simple model is doubtful and questionable. The “activation energies” of quenching deduced from short temperature intervals at relatively high temperatures are summarized in Table 3. Only data obtained for above-bandgap excitation were usable. If we assumed that the luminescence would be caused by band-to-bound transitions and the quenching would be caused by a thermal escape of electrons (holes) into CB (VB), the values near either ℏωZPL or rather at EG
ℏωZPL should be obtained according to a simple estimation. However, much lower values have been received. As mentioned above, if a high-rate recombination (possibly nonradiative) in materials with Fermi level near the middle of the bandgap is taken into account [20], much lower “activation energy” can be obtained than the one corresponding to ionization energy of the center for radiative recombination. Another possibility is a quenching caused by overcoming a barrier either inside the defect (like in Fig. 9) or more probably in the case of high concentration of various defects by breaking barriers between neighboring defects. However, similar values of “activation energy” have been received for various components; and an interlink between temperature dependences of the respective components is obvious. Thus we prefer the above mentioned “multicenter-model.”

Component 9 (1.19 eV) is thermally quenched rapidly both at above-gap excitation and at below-gap excitation. This can be attributed to radiative relaxation of a thermally unstable state localized at the defect caused by a very low energy difference between the excited state of the defect and the CB (or VB). The antisite TeCd could be a hypothetical candidate for the PL component. On the contrary, component 10 (1.13 eV) is the strongest PL component at an increased temperature under below-bandgap excitation in a wide temperature interval. Its excitation spectrum shows that component 10 is probably caused by a transition in a deep localized center. Component 11 (1.03 eV) is dominant at higher temperatures under above-bandgap excitation when both electrons and holes are generated with a high rate. It seems to be related to an effective recombination center, possibly a complex of a deep donor and a deep acceptor. The possibility that this is connected with tellurium vacancy cannot be excluded. The concentration of defects can become higher in the vicinity of extended defects like dislocations.

5. Conclusions

where integral intensity of the band I INT is multiplied by the normalized form of the band. The FWHM of the band is sensitive to S and to the width of Gaussian components Δ. The dependence FWHM(S, Δ) is demonstrated in Fig. A1 together with FWHM of components 8–12. The individual replicas are completely blurred for Δ 4 1. No phonon replicas were resolved for components 9–12. Thus Δ  1 can be taken as the low limit for Δ. For comparison, the value Δ E 0.45, ωPHONON ¼ ωLO  21 meV was determined from a measured “modulation” of the replica pattern in the “1.45 eV” PL band shown in Fig. 1. The use of ωLO for the phonon value in other PL band is questionable, because important local phonons could be of a different value. However, for another value of dominant phonon energy, a rescaling is possible. The position of PL band peak is given (for Δ Z1) by formula    ℏωPEAK ¼ ℏωZPL
S
δS S; Δ ℏωPHONON ;

The studied CdTe single crystal sample was characterized by a relatively high resistivity (9  108 Ω cm) in spite of a very low Indoping. It means that the Fermi level is pinned near the middle of the gap by “deep defects” (approximately 0.86 eV above the valence band at room temperature). The luminescence spectrum of the sample observed in the “standard” spectral range is usually said to be typical of a high quality material (except for a dislocation-related Y-line that is clearly observed in the sample). We have found that at least three different components with distinct dependences on temperature and exciting photon energy are combined in the so called “1.1 eV” PL band. We can recognize component 9 with its maximum near 1.19 eV, component 10 with the PL peak at 1.13 eV and component 11 (1.03 eV). Assuming the phonon emission broadening of the components we estimate PL zero-phonon-lines to be (1.27 70.04) eV, (1.26 70.01) eV, and (1.14 70.08) eV, respectively. Temperature dependences of PL intensity are complex with quenching activation energies above 100 K, much lower than expected ionization energies. It is in a qualitative agreement with models for wide-gap high-resistivity materials with a high recombination rate through the level near the middle of the gap.

Acknowledgments This paper was financially supported by the Grant Agency of the Czech Republic under No. GAČR 15-05259S, the Grant Agency of Charles University (Project no. 1054213) and student project SVV-2015-260216.

Appendix A A1. Model for peak position and width of a PL band A simple model of broadening given by Huang–Rhys factor S and by broadening of individual replicas by Gaussian gives for a PL spectrum   1 I ωLUMI ; S; Δ ¼ I INT expð
SÞpffiffiffiffi π ωPHONON :Δ " # 1 X Sn
ðωLUMI
ωZPL þ nωPHONON Þ2 exp  ;   n! ωPHONON :Δ 2 n¼0

where the “correction” δS is depicted in Fig. A2. A2 Comparison to results in other papers A2.1 Spectral position of the components of the “1.1 eV” band In some papers, the “1.1 eV” band was decomposed into two components, typically one with a maximum in the interval

J. Zázvorka et al. / Journal of Luminescence 177 (2016) 71–81

Fig. A 1. Calculated band widths depending on Huang–Rhys parameter S for several values of Gaussian broadening. Horizontal lines represent widths of components obtained by PL spectra fitting. Open circles represent the value S obtained from phonon replicas and from the difference between PL and PLE maxima for the component 8 and 10, respectively. Full circles indicate the “mean” values for estimation of ħωZPL given in Table 3.

Fig. A2. Dependence of correction δS for peak position on Huang Rhys factor S for several values of Gaussian broadening Δ.

1.1 eVo ℏωLUMI o 1.15 eV and the second maximum at 1.0 eV o ℏωLUMI o 1.1 eV: e.g. 1.13 eV and 1.03 eV in samples damaged by Fe ion implantation [21], 1.12 eV and 1.05 eV in polycrystalline samples doped with Cl [15], 1.12 eV and 1.03 eV [11]. The two components are reported to reveal different characteristics: different activation energies of thermal quenching, distinct dependences of the two components on the excitation power [15], various time constants at chopped excitation [21].

79

Stadler et al. [13] observed several components of the “1.1 eV” band in various samples: 1.145 eV, 1.135 eV, 1.11 eV, 1.05 eV and no correlation with In-doping was found. The “1.145 eV” type is shifted to a lower energy with rising temperature and to a higher energy with increasing x similarly to the gap EG ðxÞ in the system Cd1
xZnxTe. The width of the PL band at low temperature in CdTe has been reported as FWHME92 meV. In the localized model the temperature dependence of FWHM for the “1.145 eV” band was interpreted so that the ZPL of the PL band is at (1.2570.02) eV. If it were a band-to-bound transition, the binding energy would be EB ¼ EG
EZPL ¼0.35 eV. Stadler et al. considered it to be just the band-to-bound transition. If we assume that the Stokes shift for PL emission and the anti-Stokes shift for absorption (excitation) are approximately equal, the absorption causing the excitation of the defect states could be expected near 1.36 eV, which is in good agreement with our measured excitation spectrum for component 10. Stadler et al. [13] found that the latter type of the PL band (“1.135 eV”) reveals a shift to higher energies with increasing temperature in some samples. They interpreted it as internal recombination (both excited and ground levels are deep). It was explained as a subsequence of difference between local vibration energy of the ground and the excited state in a well-localized center ℏωGR  15 meV, ℏωEXCIT  14 meV. In our measurement, no component revealing such a shift was recognized. The FWHM E127 meV at low temperature is very similar to our values. Krustok et al. [15] considered the PL peak 1.12 and 1.05 eV to be connected with ZPL at 1.17 eV and 1.08 eV, respectively. They assumed a high value of the Gaussian components width, namely 40 meV and 130 meV. Thus they use low values for the parameter S ¼ 2:2 and 1.5, respectively. Consequently the energy difference ℏωZPL
ℏωPEAK was estimated lower than ours and those reported in [13]. Armani et al. [11] observed (cathodoluminescence at 77 K) that contribution of the component at 1.12 eV is relatively more intensive in the conductive p-type sample (more acceptors like VCd expected) than the component at 1.02 eV. Both components were equal for samples revealing resistivity 108 Ω cm. In our sample of the same resistivity, the 1.03 eV component is more intensive than that of the 1.13 eV at 80 K and above-gap excitation. PL spectra with maxima above 1.15 eV (as component 9) are reported relatively seldom: 1.2 eV in In-doped CdTe after Xe ion implantation [22]; 1.17 eV in some unspecified samples [13,23]; 1.18 eV after annealing of nominally undoped material in an inert atmosphere. A strongly asymmetric band with a maximum at 1.17 eV was observed for the bulk crystal after mechanical polishing and chemical etching [24]. A dependence of the PL peak position on resistivity in semi-insulating CdMnTe:In was noticed in [10]: the peak is shifted from 1.05 eV for samples with RT resistivity above 1010 Ω cm up to 1.21 eV for samples with resistivity below 108 Ω cm. No correlations of the peak position to Mn content or In concentration was reported. Let us note again that PL peak at 1.19 eV was obtained for our sample of resistivity almost 109 Ω cm. A2.2 Thermal quenching We have found only little information in published papers on temperature dependence of the intensity of the PL “1.1 eV” band. Krustok et al. [15] quoted values of quenching activation energies above 100 K as 174 meV and 113 meV for the PL components peaking at low temperature at 1.03 eV and 1.13 eV, respectively. Their sample was a sintered polycrystalline material, p-type with concentration 1.7  1016 cm
3. This is a remarkably higher concentration than that found for our sample of the n-type. Therefore we have compared the results with the more investigated “1.45 eV” band that is caused by radiative recombination of an electron in the CB with a hole bound to the “deeper” acceptor with ionization energy of about 150 meV. Stadler et al. [13]

80

J. Zázvorka et al. / Journal of Luminescence 177 (2016) 71–81

reported the activation energy (95 7 10) meV for a single crystal sample doped with In (neither In concentration nor electron concentration was specified for the sample). Lee et al. [25] found the value 125 meV for MBE epilayer doped with iodine and electron concentration 1.8  1018 cm
3.

A model for antisite TeCd was presented in [30]. The lowest empty electronic levels of the states TeCd0 and TeCd þ lie near the CB bottom. One of possible conceivable recombination channels at low temperatures could be described as a capture of a CB electron followed by electronic and lattice relaxation

þ

Te þ Cd ða; Q 1 Þ þ ħωEXCIT -Te þ Cd ða; Q 1 Þ þ e– þ h -;   þ þ -Te0 Cd ðae; Q 1 Þ þ h -Te0Cd a2 ; Q 2 þ h

e–capture and lattice relaxation

-Te þ Cd ða; Q 2 Þ þ ħωLUMI -

h–capture; PL emission

-Te þ Cd ða; Q 1 Þ;

The activation energy 0.17 0.01 eV was found for the 1.45 eV band for films grown by vapor transport deposition and subsequently annealed in a CdCl2 containing atmosphere and doped with Cu. [26] Our value gained for the high-resistivity sample is significantly lower: 73 meV and 77 meV, i.e. about ½ of the acceptor ionization energy. It is in a qualitative agreement with the results of the Reshchikov [20] models: the quenching activation energy could be approximately equal to the acceptor ionization energy for a conductive n-type material with Fermi level near the CB bottom. On the other hand, higher quenching activation energies in a broad interval 100–200 meV were obtained for highly defect polycrystalline CdTe:Cl [27]. This was connected with a space fluctuation of potential for charge carriers. A space separation of electrons and holes can influence the quenching. Considering these results for the “1.45 eV” band, one cannot be surprised that a great discrepancy between the quenching activation energy and expected ionization energies is observed for components of the “1.1 eV” band as well, see Table 3 (if mechanism of radiative band-to-bound recombination is assumed). A3 Origin of the relevant deep levels: an overview of candidates Although our very limited PL measurement of a particular sample cannot give any significant contribution to the discussion about the nature of the “1.1 eV” band, we add some examples of previously published opinions. The components of the “1.1 eV” band were tentatively attributed to native defects, impurities and extended defects (dislocations and defects in their neighborhood). All “simple” native defects like vacancies VTe [13], VCd [28], antisite TeCd [29], interstitials Tei and their complexes [15] were proposed to be the origin of the PL band. An example of assignment of the “1.1 eV” band to cation vacancy consists in results of photo-EPR measurements [28] with threshold photon energy 0.47 eV. It was interpreted by photoionization of a hole þ

V – Cd þ ħωEXCIT -V 2– Cd þ h ;

ħωEXCIT Z 0:47 eV;

where symmetry of the atom arrangement is C3V for both charge states. Recombination channel after above-bandgap excitation þ

þ

V – Cd þ ħωEXCIT -V – Cd þ e– þ h -V 2– Cd þh þ ħωLUMI can be followed by a rapid capture of electrons. Position of the PL band can be estimated as a complement ħωLUMI  EG –EðV – Cd ; C 3V Þ  1:61–0:47 ¼ 1:14 eV: The hole h þ can be captured later or can recombine at another defect. However, according to [29], the similar EPR spectra were attributed to the antisite TeCd.

e–h pair generation;

and lattice relaxation;

where a and e stand for electronic states of respective symmetry and Q1, Q2 represent atomic arrangements. The capture becomes less effective with increasing temperature. Maxima in the range of 0.97–1.25 eV were observed [31] for the PL (80 K) measured on layers treated using various thermal and atmosphere regimes. The unstable (long-time RT storage) “1.25 eV” band was attributed to the complex of vacancies VCd
VTe, whereas the more stable “1.0 eV” band could be connected rather with Tei and Tei – complexes. A recombination in donor–acceptor pairs (where both donor and acceptor are deep) was proposed in [15]. The splitting of the “1.1 eV” band was suggested to be caused by the different space distance of a deep donor and a deep acceptor; 1.046 eV and 1.129 eV for the closest arrangement and for the second closest neighbors, respectively. As possible candidates for deep donors and deep acceptors vacancies VTe and interstitials Tei were suggested, respectively. Babentsov and James [32] discussed properties of anion vacancies in Zn- and Cd-related chalcogenides with the conclusion that the deeper PL band (at 0.8 eV in CdTe) is due to a recombination at the vacancy VTe, whereas the “1.1 eV” PL band could be caused by either a recombination at VCd
VTe vacancy pairs or by some impurity-related level. A lot of candidates for defects causing deep levels can be found in papers dealing with calculation of formation energy for various defects [30,33,34]. Unfortunately, the results still remain controversial. Differences between results of various authors and the accuracy (usually much worse than 0.1 eV) of the calculation do not allow us the use of the results directly for a credible assignment of particular deep levels obtained in experiments. Calculated formation energies of some native defects seemed rather high. Thus their equilibrium concentrations were assumed to be low, e.g. a complex TeCd
VCd , references in [33]. Newer calculations [34] resulted in lower formation energies (e.g. for TeCd) or for some “new” atomic arrangements: TeTe þ Tei - (Te– Te)split where a pair of Te atoms is in the regular position instead of a single atom. Du et al. [34] calculated that the formation energy of TeCd under Te- rich growth condition is relatively low (E1.5 eV). The energy of a deep donor-like state corresponding to the þ2/0 charge change was computed to be 0.35 eV above the top of VB. The neutral state is realized in an equilibrium with Fermi level near the middle of gap, optical excitation causes positively charged state. The difference EG–E(TeCd, 2/0) ¼ 1.26 eV corresponds to our estimation of ZPL for the components 9 and 10. The calculated levels of VTe are approximately 0.9 eV and 1.1 eV for charge changes þ2/0 and 0/-2, respectively. These values fall into the region of the component 12. Results of calculations presented by Ma et al. [35,36] place ionization energy of Tei about 0.5 eV above VB and ionization energies of TeCd about 0.55 eV and 1.25 eV above VB for charge

J. Zázvorka et al. / Journal of Luminescence 177 (2016) 71–81

changes þ2/0 and 0/
2, respectively. On contrary, no ionization energy of VTe has been found in the energy gap and vacancies VTe þ 2 seem to be stable or rather create shallow donor states. Ionization energy 0/
2 for vacancies VCd was calculated to be at 0.38 eV. Moreover, the crystal disorder as a result of implantation, deformation, annealing far from equilibrium solid CdTe – gaseous phase etc. can cause that the defect concentration can be higher than the calculated one. Some impurities have been found to create PL bands near 1 eV, as mentioned in the Introduction. Moreover, not only “heavy” atoms like Fe, Sn or Ge can cause deep levels. Also “light” atoms (oxygen, hydrogen) can take part in the generation of a deep level. Du et al. [34] investigated various configurations of hydrogen bound to oxygen. Complexes of the OTe–H type could play an important role in pinning the Fermi level near the middle of the gap [33,34] because of a lower formation energy than those for native defects. Deformation of the crystals is expected to create various types of dislocations. Hümmelgen and Schröter [37] found that the indentation of the p-type CdTe generates glide dislocations marked as Te(g) and Cd(g). The dislocations induce defect levels with ionization energy 0.44 eV and 0.43 eV, respectively. The complementary energy is EG
0.44 eV E1.17 eV. Let us mention that the Y-line (1.47 eV) was assigned to a recombination of excitons bound to Te(g). Kim et al. [10] reported that a characteristic behavior of the DLTS-peak near 1.1 eV (observed for CdMnTe and CdZnTe) corresponds to free carrier trapping on extended defects (dislocations) because the activation energy varies with the bias. The intensity of the I–DLTS (current DLTS) peak follows a logarithmic dependence on the trap filling pulse width. It was connected with the PL “1.1 eV” band as well. Babentsov et al. [38] discussed the effects of uniaxial plastic deformation, indentation, and scribing and stated that moving dislocations introduce one-dimensional chains of defects (e.g. vacancies) that are not stable at room temperature. The result is a higher concentration of “usual” point defects with levels in the bandgap. A detailed investigation [39] of the Shockley partial dislocation connected with ends of stacking faults shows that no localized states are introduced, but local band-bending results in a space separation of excited electrons and holes at the Te-core and Cdcore, respectively. Counter-intuitively, the charge separation induced by dislocations can reduce the rate of carrier recombination. Thus, stacking faults work rather as traps. A thermally induced escape of carriers from these and other shallow traps can contribute to the PL thermal activation below 20 K as observed in our measured temperature dependences (components 9 and 10 in Fig. 8).

81

References [1] D. de Nobel, Philips Res. Rep. 14 (1959) 361. [2] S. Gurumurthy, K.S.R.K. Rao, A.K. Sreedhar, H.L. Bhat, B. Sundersheshu, R. K. Bagai, V. Kumar, Bull. Mater. Sci. 17 (1994) 1057. [3] K. Guergouri, N. Brihi, Y. Marfaing, R. Triboulet, J. Cryst. Growth 256 (2003) 230. [4] P. Fernádez, J. Optoelectron. Adv. Mater. 5 (2003) 369. [5] V. Babentsov, V. Boiko, G.A. Schepelskii, R.B. James, J. Franc, J. Procházka, P. Hlídek, J. Lumin. 130 (2010) 1425. [6] N. Zambelli, L. Marchini, G. Benassi, D. calestani, A. Zappettini, IEEE Trans. Nucl. Sci. 59 (2012) 1526. [7] K.H. Kim, J.H. Choi, A.E. Bolotnikov, G.S. Camarda, A. Hossain, G. Yang, Y. Cui, R. B. James, J. Korean Phys. Soc. 62 (2013) 623. [8] A.A. Pruchkina, N.S. Nikolaev, V.S. Krivobok, V.S. Bagaev, E.E. Onoshchenko, Y. V. Klevkov, S.A. Kolosov, Semiconductors 48 (2014) 292. [9] Z. Sobiersierski, I.M. Dharamdasa, R.H. Williams, Appl. Phys. Lett. 53 (1988) 2623. [10] K.H. Kim, A.E. Bolotnikov, G.S. Camarda, A. Hossain, R. Gul, G. Yang, Y. Cui, J. Procházka, J. Franc, J. Hong, R.B. James, J. Appl. Phys. 109 (2011) 113715. [11] N. Armani, C. Ferrari, G. Salviati, F. Bissoli, M. Zha, L. Zanotti, Mater. Sci. Eng. B 91–92 (2002) 353. [12] J. Zázvorka, J. Franc, P. Hlídek, R. Grill, J. Lumin. 143 (2013) 382. [13] W. Stadler, D.M. Hofmann, H.C. Alt, T. Muschik, B.K. Meyer, E. Weigel, G. Müller–Vogt, M. Salk, E. Rupp, K.W. Benz, Phys. Rev. B 51 (1995) 10619. [14] R.C. Bowman, D.E. Cooper, Appl. Phys. Lett. 53 (1988) 1521. [15] J. Krustok, V. Valdna, K. Hjelt, H. Collan, J. Appl. Phys. 80 (1996) 1757. [16] J. Procházka, P. Hlídek, J. Franc, R. Grill, E. Belas, M. Bugár, V. Babentsov, R. B. James, J. Appl. Phys. 110 (2011) 093103. [17] J. Zázvorka, P. Hlídek, J. Franc, J. Pekárek, R. Grill, Semicond. Sci. Technol. 31 (2016) 025014. [18] P. Horodyský, P. Hlídek, Phys. Status Solidi B 243 (2006) 494. [19] G. Lucovsky, Solid State Commun. 3 (1965) 299. [20] M.A. Reshchikov, J. Appl. Phys. 115 (2014) 012010. [21] R. Kernöcker, K. Lischka, L. Palmetshofer, J. Cryst. Growth 86 (1988) 625. [22] C.B. Norris, C.E. Barnes, K.R. Zanio, J. Appl. Phys. 48 (1977) 1659. [23] B.K. Meyer, H. Linke, P. Omling, M. Salk, K.W. Benz, Mater. Sci. Eng. B 16 (1993) 243. [24] S.R. Das, J.G. Cook, N.L. Rowell, M.S. Aouadi, J. Appl. Phys. 68 (1990) 5796. [25] J. Lee, N.C. Giles, D. Rajavel, C.J. Summers, J. Appl. Phys. 78 (1995) 5669. [26] D. Kuciauskas, P. Dippo, Z. Zhao, L. Cheng, A. Kanevce, W.K. Metzger, M. Gloeckler, IEEE J. Photovolt. 6 (2016) 313. [27] J. Krustok, J. Mädasson, K. Hjelt, H. Collan, J. Mater. Sci. 32 (1997) 1545. [28] P. Emanuelson, P. Omling, B.K. Meyer, M. Wienecke, M. Schenk, Phys. Rev. B 47 (1993) 15578. [29] D. Verstraeten, G. Longeaud, A. Ben Mahmoud, H.J. von Bardeleben, J. C. Launay, O. Viraphong, Ph.C. Lemaire, Semicond. Sci. Technol. 18 (2003) 919. [30] A. Carvalho, A.K. Tagantsev, S. Öberg, P.R. Briddon, N. Setter, Phys. Rev. B 81 (2010) 075215. [31] V. Valdna, Phys. Scr. T69 (1997) 315. [32] V. Babentsov, R.B. James, J. Cryst. Growth 379 (2013) 21. [33] K. Biswas, M.H. Du, New J. Phys. 14 (2012) 063020. [34] M.H. Du, H. Takenaka, D.J. Singh, Phys. Rev. B 77 (2008) 094122. [35] J. Ma, D. Kuciauskas, D. Albin, R. Bhattacharya, M. Reese, T. Barnes, J.V. Li, T. Gessert, S.-H. Wei, Phys. Rev. Lett. 111 (2013) 067402. [36] M.O. Reese, C.L. Perkins, J.M. Burst, S. Farrell, T.M. Barnes, S.W. Johnston, D. Kuciauskas, T.A. Gessert, W.K. Metzger, J. Appl. Phys. 118 (2015) 155305. [37] I.A. Hümmelgen, W. Schröter, Appl. Phys. Lett. 62 (1993) 2703. [38] V. Babentsov, V. Boiko, R. Grill, J. Franc, G.A. Shepelskii, Phys. Status Solidi C 7 (2010) 1492. [39] C. Li, Y. Wu, T.J. Pennycook, A.R. Lupini, D.N. Leonard, W. Yin, N. Paudel, M. Al– Jassim, Y. Yan, S.J. Pennycook, Phys. Rev. Lett. 111 (2013) 096403.