Extended Defects in O⁺-Implanted Si Layers and Their Luminescence

Abstract

The structure and luminescence properties of Czochralski-grown n-Si samples implanted with oxygen ions have been comprehensively analyzed using photoluminescence and transmission electron microscopy (TEM). A high oxygen concentration (5 × 10¹⁹ cm^–3) in a layer at a depth of 0.3–0.8 µm was obtained in the implanted material. The samples have been annealed according to the multistage technique, including low-temperature (650/800°C) and high-temperature (1000°C) stages, to obtain oxygen precipitates and a system of various extended defects. The structure of the dislocation-photoluminescence spectrum is determined by the features of interaction of dislocations with oxygen during annealing. It is established that decoration of dislocations with oxygen precipitates leads to quenching the D1 and D2 lines. The strong D1 line is due to “pure” (without oxygen precipitates) dislocations. Free amorphous oxygen precipitates emit at a wavelength of 1476 nm.

INTRODUCTION

Currently, the fundamental problem concerning the nature of radiative centers responsible for the dislocation photoluminescence (DPL) in silicon has not been completely solved. The key question of studies is the determination of an unambiguous correspondence between the specific types of defects and characteristic DPL lines, which are traditionally denoted as D1 (0.807 eV/1535 nm), D2 (0.87 eV/1424 nm), D3 (0.95 eV/1306 nm), and D4 (0.99 eV/1234 nm). The energy positions/wavelengths are given for a temperature of photoluminescence (PL) measurements of 4.2 K. The solution of this problem is hindered by the fact that D1–D4 lines appear in the PL spectra of almost any Si crystals, independent of the way in which dislocations were introduced into them; however, an ensemble of structural defects in each material is too complex because of the presence of different types of defects. In addition, effects of interaction of dislocations with intrinsic point defects and impurity atoms lead to occurrence of a large variety of PL spectra.

To date, the following data can be assumed as reliably established. D3/D4 lines are directly related to split 60° glide dislocations [1, 2]. The radiation line (R) at a wavelength of 1370 nm (0.903 eV) is due to rod-like {113} defects [3–6]. The line at a wavelength of 1494 nm (0.83 eV) is due to oxygen precipitates [7]. Note that there is no optical activity of Frank loops containing a stacking fault [8, 9] and shuffle dislocations introduced into Si using plastic deformation under high (several GPa) hydrostatic pressure [10].

The presence of oxygen in Czochralski-grown silicon single crystals (Cz-Si), and the use of oxygen in the technology of impurity gettering in silicon wafers led to active study of the luminescence properties of Si with a high oxygen content (~10¹⁸ cm^–3). The interaction of oxygen precipitates with dislocations and formation of related microdefects as a result of agglomeration of interstitial atoms (are emitted from the strained matrix, adjacent to the precipitates, into the bulk) complicates significantly the PL spectra. It follows from the studies on silicon implanted with intrinsic ions (Si:Si⁺) [11] that the shift and broadening of the DPL peaks related to the luminescence of clusters of interstitial atoms is a fundamental phenomenon. The interaction of dislocations with oxygen precipitates leads to shift and broadening of peaks [7, 12] and occurrence of additional lines in the ranges of characteristic DPL lines [13]. As a consequence, there arises an ambiguity in solving the question about the assignment of the D1 and D2 lines to oxygen precipitates on dislocations [13] or directly to the dislocation core [8, 12]. Thus, it was noted in [13] that the dependences of the intensity of D1 and D2 lines on the time of nucleation and growth of oxygen precipitates differ; it was suggested that these lines correspond to luminescence of precipitates of different types. Vice versa, it was shown in [12] that the PL spectrum may change (from a wide band in the vicinity of 0.808 eV to a set of sharp D1/D2 peaks) in initially identical samples with an increase in the annealing temperature from 700 to 1000°C. This result was explained by the fact that the DPL spectrum is affected by secondary defects (dislocations, which are purified from small precipitates during annealing due to their gradual coarsening because of oxygen diffusion along dislocations) rather than by oxygen precipitates. Recently, some electron microscopy data on various dislocation structures in Si were presented in [9], which indicate that the D1 and D2 lines are related directly to the dislocation core and metastable interstitial microdefects {111}, {113}, and {001}, formed near the dislocation core, respectively [14–16]. As one can see, the question about the role of oxygen precipitates and their interaction with dislocations in the emission in the range of 0.8–1.0 eV is still urgent.

The establishment of the nature of DPL lines is especially important, with allowance for the experimental facts indicating that, at identical dislocation densities, the intensity of the D1/D2 lines is an order of magnitude stronger in dislocated Cz-Si crystals with a high oxygen concentration than in plastically deformed crystals [12, 13]. Solution of this question may facilitate development of the methods for increasing the intensity of, primarily, the D1 line, which is of practical interest for light-emitting devices.

The purpose of this study was to analyze the luminescence properties and structural defects in O⁺-implanted silicon in order to clarify the effect of oxygen interaction with dislocations. Ion implantation made it possible to obtain an elevated oxygen concentration in the surface layer of Si wafer and provide a possibility for controlling the formation of structural defects during thermal annealing. Each sample with a characteristic set of thermal-annealing conditions has been investigated in detail using PL and transmission electron microscopy (TEM).

EXPERIMENTAL

Cz-n-Si(100) wafers of KEF-4.5(100) grade with a diameter of 100 mm were used as the starting material. According to the IR-absorption measurements, the oxygen concentration in the starting material was 8 × 10¹⁷ cm^–3, and the carbon concentration was below the measurement limit (2 × 10¹⁶ cm^–3). Implantation of oxygen ions was carried out according to the multistage technique, varying the ion energy in the range from 350 to 150 keV and the implantation dose in the range from 1.5 × 10¹⁵ to 7 × 10¹⁴ cm^–2. Based on the results of calculations performed within the TRIM program, a uniform oxygen distribution at a level of 5 × 10¹⁹ cm^–3 was obtained in the wafers at depths from 0.3 to 0.8 µm. Each wafer of initial (reference) and implanted silicon was divided into several samples, subjected to annealing according to the multistage technique, which is used in the technology of integrated circuits for impurity gettering due to the formation of oxygen precipitates. The modes of multistage sample annealing are described in Table 1. The first, high-temperature (HT1) annealing was carried out at 1000°C for 15 min to dissolve growth precipitates. The second, low-temperature (LT2) annealing was carried out at 650°C for 7 or 10 h to initiate nucleation of precipitates with different density. The third (LT3) annealing was performed at 800°C for 4 h to implement outgrowth of precipitates. All these annealing stages were performed in an Ar flow. In addition to the “standard” technique, an annealing (HT4) was carried out at 1000°C for 6 or 10 h in an Ar flow or Cl-containing atmosphere; the latter was an oxygen flow saturated with carbon tetrachloride vapor (molar concentration 1%). Previously, annealing in Cl-containing atmosphere was efficiently used to increase the DPL intensity in Si due to the outgrowth of dislocations and increase in their density [17, 18]. The oxide layer formed on the wafer surface during this annealing was removed before the measurements using chemical etching. The structures of implanted layers annealed under different conditions were investigated using TEM and high-resolution transmission electron microscopy (HRTEM) on a JEM4000EX microscope (accelerating voltage 400 kV) and a TITAN80-300 microscope with a spherical-aberration corrector (accelerating voltage 300 kV). The type of extended defects and their distribution in the material were studied using TEM on longitudinal and transverse cuts of the samples using reflections 220 and 400 (the latter are favorable for simultaneous observation of Frank loops in different location planes {111} without contrast deterioration). The oxygen precipitates were revealed using dark-field weak beam microscopy (g/3g diffraction geometry). The optical properties of the same samples were investigated using PL excited by a 532-nm solid-state laser (power 64 mW) and recorded in the wavelength range of 1000−1650 nm at a temperature of 80 K applying an MDR-25 automatic monochromator and an InGaAs photodetector. The setup resolution was 7 nm. The PL spectra were analyzed taking into account the dip in the range of 1380 nm, observed in all PL spectra, which is due to the water-vapor absorption in the measuring channel.

Table 1.   Samples of initial (series 1) and implanted (series 2 and 3) silicon, subjected to multistage annealing in argon (Ar) and Cl-containing atmosphere (CCA)

RESULTS

The samples prepared for analysis are characterized in Table 1. The fundamental difference in the general multistage technique of sample annealing is the presence (series 1 and 2) or absence (series 3) of the initial annealing stage HT1 at 1000°C. The samples of each series were subjected to identical set of heat-treatment stages after HT1.

The Initial (Reference) Cz-n-Si (Series 1)

The reference material was subjected to “standard” annealing, complemented with the final stage (HT4). To investigate the evolution of the sample properties during annealing, they were analyzed after each stage. The PL spectra of samples 1-1 and 1-2, subjected to primary annealing at 1000°C and subsequent low-temperature stages, yielded only edge luminescence and did not contain any peaks due to oxygen precipitates or structural defects. No weak strains of the Si lattice were found using TEM. This result confirms the well-known statement that annealing of Cz-Si at a temperature in the range of 650–800°C leads to the occurrence of thermal donors, which form nonradiative centers and serve as precursors of the precipitates arising along with {113} defects at higher temperatures [19]. As one can see, the short-term (15 min) annealing at 1000°C barely affected this situation.

The samples subjected to the “standard” heat treatment also underwent final annealings (HT4) at 1000°C for various times and in atmospheres of different types. Figure 1 shows the PL spectra grouped with respect to the type of annealing atmosphere (Ar or Cl-containing). Annealing in Ar for 6 h (sample 1-3; PL spectrum in Fig. 1а) led to the occurrence of an unknown wide peak (XX) with a maximum at a wavelength of 1345 nm, which is likely due to clusters of intrinsic interstitial atoms [5], and a wide band in the range of 1420–1500 nm, which contains the D1 and D2 DPL lines and, possibly, the line of oxygen precipitates at a wavelength of 1479 nm [7]. An increase in the duration of this annealing stage to 10 h (sample 1-4) is accompanied by weakening of all peaks, except for the D1 peak, whose intensity remains at the same level. The PL spectrum radically changes in sample 1-5, whose annealing time was increased to 10 h at 650°C in the nucleation stage of oxygen precipitates (LT2). The intensity of the D2 peak increased significantly in comparison with sample 1-3, and the D1 peak remained at the same level.

Fig. 1.

PL spectra of initial Si (series 1) after multistage annealing with final stage HT4 in (a) Ar and (b) Cl-containing atmosphere. Nomenclature of the samples corresponds to Table 1.

Figure 1b shows the PL spectra of a similar set of samples, subjected to annealing HT4 in Cl-containing atmosphere. In this stage, 6-h annealing (sample 1-6) is accompanied only by the occurrence of a D1 peak (its intensity barely increased in comparison with the spectrum of sample 1-3). The peak asymmetry (smaller slope from the side of shorter wavelengths) is typical of D1 in the samples fabricated using various technological methods. An increase in the duration of the final annealing stage to 10 h (sample 1-7) almost doubles the intensity of the D1 peak and leads to occurrence of the D2 peak. In the spectrum of sample 1-8, which corresponds to a longer-term nucleation stage of oxygen precipitates (650°C/10 h), one can observe enhancement of the luminescence intensity in the range of D2. The intensity of the D1 peak remains at the previous level. It should be noted that, in both sets of samples, the increase in the intensity of the D2 peak correlates with the increase in the concentration of small oxygen precipitates and, as a consequence, interstitial microdefects in the Si matrix.

Figure 2 shows the TEM images of structural defects in sample 1-6, which was annealed in Cl-containing atmosphere in the final stage HT4. One can observe a developed system of defects in the surface layer, consisting mainly of large Frank loops and perfect dislocations (fragments of long dislocation half-loops). Dark-field images exhibit large amorphous lamellar oxygen precipitates in the wafer bulk at a depth of more than 5 µm (Figs. 2b, 2c), with rare segments of perfect dislocations formed as a result of prismatic extrusion. Obviously, these precipitates are beyond the luminescence-excitation range.

Fig. 2.

Dark-field TEM images of sample 1-6: (a) defects in the surface layer of the wafer in the (100) plan-view foil, (b) oxygen precipitates in the wafer bulk in the {110} cross-sectional foil, and (c) HRTEM image of the precipitate.

Implanted Cz-n-Si(O)

The multistage complex of heat treatments was applied to the implanted Si samples in such a way that the samples of series 2 were subjected to the treatment stage HT1, while the samples of series 3 were not. The PL spectra of the samples are shown in Fig. 3.

Fig. 3.

PL spectra of the Si-implanted samples of series (a, c, e) 2 and (b, d, f) 3: (a, b) stages of “standard” annealing and (c–f) multistep annealing with HT4 in (c, d) Ar and (e, f) Cl-containing atmosphere. Nomenclature of the samples corresponds to Table 1. The resolved peaks in the spectra are indicated either in the DPL notation (D1, D2) or as types of radiation sources (OP for oxygen precipitates and 113 for rod-like defects).

Samples of series 2, subjected to initial annealing stage HT1. High-temperature annealing of implanted Si layers leads generally to clusterization of intrinsic point defects, accompanied by the formation of microdefects and Frank loops. Therefore, the properties of material (sample 2-0) subjected to only one annealing stage at a temperature of 1000°C in Ar were investigated. The PL spectrum of sample 2-0 (Fig. 3a) contains only D1 and D2 peaks. According to the TEM data, extended structural defects were actively generated in the implanted layer (Figs. 4a, 4c). Frank loops with sizes in the range of 30–100 nm and a density of (1–3) × 10¹⁰ cm^–2 and perfect dislocation dipoles with lengths in the range of 300–1000 nm and a density of (1–3) × 10⁸ cm^–2 dominate in the system of defects. One can also observe a small fraction of perfect prismatic loops (wide dislocation dipoles). Independent of their sizes, Frank loops exhibit a characteristic duality, which is indicated by the pair contrast of their contouring dislocations (Fig. 4c, inset). The formation of these dual loops is related to the oxygen precipitates formed on a stacking fault of a primary Frank loop and acting as nucleation centers for smaller secondary loops [20]. The electron-microscopy contrast on defects (especially in the plan-view image of implanted layer (Fig. 4c)) indicates that all defects are decorated with oxygen precipitates. As follows from the cross-sectional image (Fig. 4a), free oxygen precipitates are formed in the Si matrix only at the lower boundary of the defect band (indicated by white arrows) and (in small amounts) near its upper boundary; however, they are absent among extended defects. This fact explains the occurrence of point contrast (corresponding to the precipitates) in the plan-view image.

Fig. 4.

Dark-field TEM images of samples (a, c) 2-0, (b) 2-1, and (d) 2-2: (a, b, d) {110} cross-sectional foils, (c) (100) plan-view foil, and (d) HRTEM image of a fragment of Frank loop. In Fig. 4a, the sample surface is partially etched. The inset in Fig. 4c shows an enlarged fragment of this image. White arrows in Figs. 4a and 4b indicate the localization domains of large oxygen precipitates.

Subsequent annealing (sample 2-1) at 650°C leads to disappearance of the D1 and D2 lines and occurrence of centers corresponding to the {113} defects and oxygen precipitates (Fig. 3a). After the annealing stage LT3 (sample 2-2) at 800°C, the intensity of the 113 peak increases, and the D1 peak of almost the same intensity appears again (however, the D2 peak is not recovered). The peak due to oxygen precipitates is not observed, possibly, because of the wide D1 peak. The TEM results (Fig. 4) show that solid oxygen solution continues to decompose in these samples. This process is accompanied by further decoration of extended defects and an increase in the density and sizes of precipitates at the lower boundary of the defect band (Fig. 4b, indicated by an arrow). Note that the ensemble of extended defects and their sizes and density change only slightly. Most of defects are Frank loops of interstitial and vacancy types. Figure 4d shows the HRTEM image (TITAN80-300) of a fragment of the Frank loop observed in sample 2-2. Its atomic structure exhibits a strongly deformed contrast, caused by deformations of the Si–Si bonds due to the oxygen incorporation into not only the dislocation core but also the stacking-fault plane, which indicates a large amount of impurity in the defect.

Other samples were subjected to final annealing HT4 at a temperature of 1000°C for various times and in different annealing atmospheres after the “standard” heat treatment. Figure 3c shows the spectra of the samples, annealed in an Ar atmosphere. A 6-h annealing (sample 2-3) gives rise to the D1 and D2 peaks in the PL spectrum. An increase in the annealing time to 10 h in the nucleation stage of oxygen precipitates (sample 2-5), with application of a similar HT4 stage, decreases the intensity of all peaks. The intensities of the DPL peaks obey the inequality D2 > D1, which is characteristic of these two samples. The inverse intensity ratio D1 > D2 is observed for sample 2-4 when the final annealing stage is increased to 10 h. A specific feature of this spectrum is the twofold increase in the intensity of the D1 peak, whereas the intensities of the other peaks are retained.

A set of the PL spectra of the samples annealed in a Cl-containing atmosphere (Fig. 3e) exhibits similar behavior of curves and all aforementioned peaks. The ratio of peak intensities D1 > D2 is valid for all three spectra and the D1 peak in these samples is stronger by a factor of almost 2 than the corresponding peaks in the same samples after annealing in Ar. The fundamental difference between the PL spectra of the samples annealed in different atmospheres is that the strongest D1 peak is observed at a shorter annealing stage HT4 performed in a Cl-containing atmosphere (6 h; sample 2-6) than in the case of Ar (sample 2-4). It turned out that the longer term (10 h) annealing in a Cl-containing atmosphere decreases significantly the intensity of this peak. The use of a longer term annealing in stage LT2 of oxygen precipitate nucleation weakens the D1 peak in the final annealing stage, independent of the atmosphere type (samples 2-5, 2-8). Concerning the D2 peak, its intensity decreases significantly in both cases of longer term annealing stages, LT2 or HT4.

The ensemble of structural defects in the samples under study undergoes characteristic changes in the final annealing stage (Fig. 5). Under these conditions, Frank loops mainly outgrow and are transformed into perfect prismatic loops, due to which an irregular spatial dislocation network is formed. In the case of Ar atmosphere (sample 2-3), one can observe a significant increase in the sizes of individual Frank loops with retained dual configuration and their drift towards the wafer surface (Figs. 5a, 5c). In this sample, all structural defects remain decorated with oxygen precipitates, and free precipitates in the matrix are significantly enlarged on the bottom of the defect band (i.e., irrespective of dislocations) (Fig. 5a). These are amorphous lamellar precipitates up to 60 nm in size, which demonstrate the same electron-microscopy contrast under different diffraction conditions as the oxygen precipitates in the initial Cz-Si, which are shown in Fig. 2.

Fig. 5.

Dark-field TEM images of samples (a, c) 2-3 and (b, d) 2-6: (a, b) {110} cross-sectional foils and (c, d) (100) plan-view foils. White arrows in Fig. 4d indicate segments of pure (without oxygen precipitates) dislocations.

During annealing in a Cl-containing atmosphere, similar processes in Si occur more intensively due to the flux of point defects from the wafer surface. Due to this, the transformation of Frank loops into perfect loops and their growth also occur more actively. As a result, a high density of perfect dislocations arises in the surface layer, which is free of oxygen precipitates due to the oxygen distribution characteristic of an implanted impurity and subsequent oxygen gettering by defects (Fig. 5b). The expansion of perfect loops in the defect band induces occurrence of dislocation segments free of oxygen precipitates (indicated by arrows in Fig. 5d). The formation of pure (precipitate-free) dislocations correlates with the significant increase in the intensity of the D1 peak in the spectra of these samples.

Samples of series 3, subjected to initial annealing stage LT2. For sample 3-1, the initial annealing stage at a temperature of 650°C is the only one. The PL spectrum of this sample contains a wide band with a gradually rising intensity in the range from 1350 to 1550 nm. This spectrum is characteristic of as-implanted samples and samples subjected to annealing with insufficiently high temperature and/or duration (Fig. 3b). Subsequent annealing at 800°C (sample 3-2) induces occurrence of relatively weak peaks in the spectrum, corresponding to the {113} defects and oxygen precipitates (1476 nm). A TEM analysis showed that, as it was suggested, sample 3-1 does not contain any structural defects, while a developed ensemble of defects is formed in sample 3-2. The dark-field TEM image of sample 3-2 (Figs. 6a, 6c) indicates that high-density free oxygen precipitates in the Si matrix (~10¹¹ cm^–2), dual Frank loops, and dislocation dipoles are formed in the implanted region and that all extended defects are decorated with precipitates.

Fig. 6.

Dark-field TEM images of samples (a, c) 3-2 and (b, d) 3-6: (a, b) {110} cross-sectional foils and (c, d) (100) plan-view foils.

Figure 7 shows the HRTEM image of a Frank loop in implanted sample 3-2, annealed at temperatures 650/800°C. Despite the low temperature, classical interstitial Frank loops are formed under these conditions: they have a specific three-layer structure, formed in {111} planes by an incorporated layer of interstitial atoms, which is in the twin configuration with respect to the Si matrix, with the formation of two hexagonal silicon layers (Fig. 7a). There are few pure Frank loops without any signs of decoration with oxygen and the majority of loops exhibits a complex contrast, which is indicative of oxygen segregation in the form of small SiO_x clusters in the defect plane (Fig. 7b). Obviously, the fraction of oxygen, deposited onto defects in this sample is much lower than in the sample subjected to the first annealing stage at 1000°C, because of the presence of free precipitates in the Si matrix. The long-term annealing at 800°C is accompanied by diffusion of oxygen and point defects, which occur with almost the same rate; therefore, oxygen precipitates are formed and increase in size; Frank loops grow as well. As a result, both nucleation of Frank loops on precipitates and decoration of growing loops with oxygen precipitates occur. An HRTEM image of a small amorphous SiO_x precipitate formed near a Frank loop is shown in Fig. 8a. The average size of free precipitates was determined (based on TEM weak beam images) to be 10 nm. Based on the density and size of precipitates, the concentration of silicon and oxygen atoms in them can be estimated as lower than 3 × 10¹⁸ and 6 × 10¹⁸ cm^–3, respectively. Since 5 × 10¹⁹ cm^–3 oxygen atoms were incorporated during implantation, oxygen is concentrated mostly on extended defects (in particular, in the plane of Frank loops) rather than in free precipitates. This estimate explains why there are no fundamental differences in the HRTEM images of Frank loops in all samples of this series.

Fig. 7.

HRTEM images (TITAN 80-300) of (a) “pure” and (b) “contaminated” with oxygen dislocation Frank loops on the {110} cross-sectional foil in sample 3-2.

Fig. 8.

HRTEM image (TITAN 80-300) of oxygen precipitates in implanted Si: (a) precipitate related to the Frank loop (indicated by arrows), annealing at 800°C (sample 3-2) and (b) free precipitate, retained in the Si matrix after annealing in a Cl-containing atmosphere at 1000°C (sample 3-6).

As follows from the analysis of the luminescence and structural data, DPL (D1/D2) is absent completely in sample 3-2 in a layer with a sufficiently high density of Frank loops and dislocation dipoles, heavily decorated with oxygen. In this case, one can conclude that the signal recorded at a wavelength of 1476 nm corresponds to free oxygen precipitates in the Si matrix.

Final annealings at 1000°C were also performed for various times and in different atmospheres. The corresponding PL spectra are shown in Figs. 3d and 3f. The run of the curves in these plots is similar to that for the corresponding samples of series 2 subjected to the annealing stage HT1. A similar tendency of change in the intensity of the D1 and D2 peaks with an increase in the durations of the precipitate nucleation stage (LT2) and final annealing stage is retained. Among the samples annealed at stage HT4 in Ar (Fig. 3d), the strongest D1 peak is obtained at an increased duration of the final stage. However, the situation changed for the annealing in a Cl-containing atmosphere (Fig. 3f). The strongest D1 peak in the samples of series 3 was obtained at an increased duration of the final stage (sample 3-7), whereas in the samples of series 2, it was observed at a short (6 h) stage (sample 2-6). The TEM data on the samples of series 3, subjected to final annealing at a temperature of 1000°C, also indicate (as in the case of the samples of series 2) that the sizes of Frank loops and perfect dislocation loops, whose segments propagate to the sample surface crossing the region of pure (without oxygen precipitates) Si layer, increased (Figs. 6b, 6d). In addition, a significant decrease in the density of free precipitates in the Si matrix is observed within the defect band, which indicates that some precipitates were dissolved and that oxygen was transported to structural defects. Residual free precipitates also form an amorphous aggregate (Fig. 8b).

DISCUSSION

The results obtained indicate that the observed variety of the DPL spectra is determined mainly by the interaction of dislocations with oxygen precipitates. The key factors of this interaction are the concentration of precipitates in the Si matrix and the order of their interaction with dislocations and Frank loops: expansion of loops or propagation of dislocations in the Si matrix saturated with precipitates; simultaneous nucleation and growth of structural defects and precipitates; and dissolution of small precipitates and oxygen diffusion on the structural defects, already formed in Si.

Interaction of oxygen precipitates with defects according to the first mechanism is implemented at a relatively low oxygen concentration (8 × 10¹⁷ cm^–3) in the initial single-crystal Cz-Si during final annealing at a temperature of 1000°C. The formation of Frank loops and their transformation into perfect dislocation loops (Fig. 2) gives rise to the D1 peak in the spectra, which is stronger in the case of annealing in a Cl-containing atmosphere (because defect-formation processes are more intense in this case). An increase in the annealing time in a Cl-containing atmosphere doubles the intensity of the D1 peak (Fig. 1b) because of the increase in the dislocation density at the same concentration of oxygen precipitates. The increase in the concentration of thermal donors (as precursors of oxygen precipitates) in the Si matrix due to the increase in the annealing time at 650°C increases the intensity of the D2 peak and broadens it, independent of the annealing atmosphere in the final stage, whereas the intensity of the D1 peak barely changes in comparison with that for samples 1-3 and 1-6, subjected to the nominal annealing procedure (1000°C (15 min)/650°C (7 h)/800°C (4 h)/1000°C (6 h)).

A more complex interaction of dislocations with oxygen precipitates (according to the second and third mechanisms) is implemented in implanted Si with an oxygen concentration of 5 × 10¹⁹ cm^–3, subjected to multistage annealing with the initial stage HT1. Due to the high concentrations of oxygen and point defects in the implanted material, the first short (15 min) annealing at 1000°C immediately leads to the formation of a developed system of extended defects, heavily decorated with oxygen. Most of oxygen atoms precipitate onto defects from the solid solution in this stage. Free oxygen precipitates are formed only at the upper and lower boundaries of the defect band. The D1 and D2 lines dominate in the PL spectrum of this sample (Fig. 3a). Decomposition of oxygen solid solution continues in the next stage LT2 of a rather long-term (7 h) annealing at a temperature of 650°C, which enhances the decoration of defects with oxygen under the conditions when the system of extended defects barely changes. As a result of this process, quenching of the D1/D2 lines occurs.

In the implanted Si subjected to multistage annealing with the initial stage LT2 (samples of series 3), another mixed version of interaction of dislocations and oxygen precipitates is implemented. In the next stage of low-temperature annealing (800°C), growth of precipitates, nucleation of extended structural defects on them, and their further growth occur simultaneously in the implanted region (however, with different diffusion rates of oxygen and interstitial atoms). As a result, all structural defects (Frank loops, dislocation dipoles, and perfect dislocation loops) are also decorated with oxygen precipitates. The degree of decoration of these defects is so high that the D1 and D2 peaks are not observed in the PL spectra. A further annealing of these samples in stage HT4 at a temperature of 1000°C is accompanied by dissolution of precipitates in the Si matrix and oxygen diffusion to dislocations. However, the expansion of dislocation loops (including their propagation to the surface) induces the occurrence of oxygen-free segments of dislocations. As a result, strong D1 and D2 peaks appear in the PL spectra.

The statement that the strong peaks D1/D2 arise in the DPL spectra due to the formation of pure (oxygen-free) dislocations or dislocation segments is confirmed by the specific features of the change in the intensity of the D1 peak, depending on the multistage-annealing conditions. Thus, upon the HT4 annealing in Ar, the strongest D1 peak is obtained with an increase in the duration of this stage to 10 h, independent of the presence or absence of the first annealing stage at a temperature of 1000°C in the general annealing scheme (Fig. 3b). This effect is explained by the fact that the growth rate of dislocation loops in Ar is not so high as in the case of Cl-containing atmosphere, and it takes more time for loops to reach pure (precipitate-free) portions of the Si matrix. The final annealing stage in a Cl-containing atmosphere facilitates a more rapid growth of dislocation loops due to the additional flux of intrinsic point defects from the wafer surface. Therefore, the strongest D1 peak in sample 2-6 (in which the Si matrix is almost precipitate-free in the defect band because of the HT1 annealing stage) is obtained at 6-h duration of the HT4 stage. At high density of free oxygen precipitates in the defect band (sample 3-7), the strongest D1 peak is obtained at a longer (10 h) HT4 stage (Figs. 3e, 3f), which is also related to the additional obstacles hindering the dislocation outcrop onto pure matrix portions. And, vice versa, an increase in the precipitate concentration in the implanted region due to the longer annealing stage at 650°C (LT2) decreases significantly the DPL intensities in all samples (i.e., for all annealing schemes).

Concerning the intrinsic emission of oxygen precipitates, the 1476-nm peak was observed in two samples of implanted Si in the initial annealing stages: short-term annealing at 1000°C (sample 2-0) and long-term low-temperature annealing at 650°C (7 h)/800°C (4 h) (sample 3-2). This emission is related directly to free amorphous lamellar precipitates. It should be noted that quenching of the D1/D2 lines was observed simultaneously in both cases, because of the decoration of dislocations with oxygen precipitates, which clearly facilitated occurrence of a weak peak due to the oxygen precipitates. This fact indicates that broadening of the peaks in the region between the D1 and D2 lines in most of DPL spectra may be due to the emission of oxygen precipitates.

Note that the PL spectra of all investigated samples do not contain the D3/D4 peaks, which are characteristic of split glide dislocations, because all extended defects in the material under study are prismatic. Under identical heat-treatment conditions, the intensity of the D1 line for oxygen-implanted Si exceeds that for single-crystal Cz-Si by a factor of more than 5.

CONCLUSIONS

The complex study of the structure and luminescence properties of Cz-Si samples implanted with oxygen ions and subjected to multistage thermal annealing was performed in order to refine the role of oxygen precipitates in formation of the DPL spectra of Si. The evolution of the PL spectra and oxygen precipitates and structural defects in each annealing stage was traced using the PL and TEM methods.

It was established that the structure of the PL spectra of the samples subjected to complex heat treatment at different combinations of the annealing temperature, duration, and atmosphere in different stages is determined by the specific features of the interaction between dislocations and oxygen precipitates. The decoration of dislocations with oxygen precipitates weakens the D1 and D2 lines up to their complete elimination during decoration. The strong D1 line is due to pure (without oxygen precipitates) dislocations. The oxygen precipitates decorating dislocations do not exhibit intrinsic luminescence. The luminescence at a wavelength of 1476 nm is due to free amorphous oxygen precipitates, which do not interact with dislocations.