1. Introduction

Cadmium silicon phosphide (CdSiP₂) crystals have excellent nonlinear optical properties and can be grown in large sizes [1–3]. This has resulted in their wide use in optical parametric oscillators to produce coherent output in the 6.0 to 8.3 µm region [4–14]. Recently, CdSiP₂ crystals have also been used to generate terahertz radiation [15–21]. It is often noted in these reports that the performance of the crystals may be improved by reducing the intensities of residual optical absorption bands. These unwanted bands are associated with native defects that form during growth of the crystals. Thus far, silicon, cadmium, and phosphorous vacancies and silicon-on-cadmium and silicon-on-phosphorous antisites have been detected in CdSiP₂ crystals [22–26]. Similar intrinsic defects have been found in ZnGeP₂, ZnSiP₂, and CdGeP₂ crystals [27–41]. The primary experimental techniques used to identify and characterize these point defects in the II-IV-V₂ chalcopyrites are electron paramagnetic resonance (EPR), photoluminescence (PL), and optical absorption.

The present study is focused on silicon vacancies in CdSiP₂. Scherrer et al. [24] have shown that the singly ionized charge state of these vacancies ($\textrm{V}_{\textrm{Si}}^ - $) is responsible for broad optical absorption bands peaking near 0.8, 1.0, and 1.9 µm. In the Results section, we show that 1064 nm light, when applied at 80 K, significantly reduces the intensities of these unwanted optical absorption bands and eliminates nearly all of the EPR spectrum of the $\textrm{V}_{\textrm{Si}}^ - $ vacancies. The light converts half of the removed singly ionized silicon vacancies ($\textrm{V}_{\textrm{Si}}^ - $) to the neutral charge state ($\textrm{V}_{\textrm{Si}}^0$) and half to the doubly ionized charge state ($\textrm{V}_{\textrm{Si}}^{2 - }$). Warming above approximately 120 K restores the absorption bands and the EPR spectrum, as electrons are thermally excited from the valence band maximum to the $\textrm{V}_{\textrm{Si}}^0$ vacancies. Monitoring the $\textrm{V}_{\textrm{Si}}^ - $ EPR spectrum at three temperatures (116, 120, and 124 K) gives a thermal activation energy of approximately 0.27 eV for this recovery process. Unlike the 1064 nm light, an exposure to 633 nm light at 80 K increases the residual absorption across the visible and near-infrared regions. The absorption bands produced by the 633 nm light are also removed by an exposure to 1064 nm light at 80 K.

The considerable difference in the tetrahedral covalent radii of the cadmium and silicon atoms in CdSiP₂ (1.405 Å and 1.173 Å, respectively) introduces a large crystal field that splits the valence bands by approximately 0.56 eV [42–44]. In the Models section, this splitting is used to assign transitions to two of the residual optical absorption bands. The large crystal field splitting of the valence bands is also key to understanding how the 1064 nm light, when applied at low temperature, removes singly ionized ($\textrm{V}_{\textrm{Si}}^ - $) silicon vacancies. Electrons moving from the upper valence band to a $\textrm{V}_{\textrm{Si}}^ - $ acceptor are responsible for the absorption band near 1907 nm, and electrons moving from the lower valence bands to a $\textrm{V}_{\textrm{Si}}^ - $ acceptor are responsible for the absorption band near 1033 nm. The third band, at 800 nm, is assigned to an intracenter transition (ground state to excited state) of a $\textrm{V}_{\textrm{Si}}^ - $ acceptor. There may also be a small acceptor-to-donor contribution to the 800 nm band involving the $\textrm{V}_{\textrm{Si}}^ - $ acceptor.

2. Experimental details

The CdSiP₂ crystal used in the present study was grown by the horizontal-gradient-freeze technique at BAE Systems (Nashua, NH) [1,3]. It is representative of material routinely being grown by this method. Singly ionized silicon vacancies are present in the as-grown crystal, indicating these acceptors are only partially compensated by native donors (i.e., silicon-on-cadmium antisites and phosphorous vacancies). Shallow cadmium vacancies are also present: they are fully compensated doubly ionized acceptors. An oriented sample suitable for optical absorption and EPR experiments was cut from the larger boule. Its dimensions are 5.0 × 3.0 × 2.0 mm³. The two broad sides were polished using a polyurethane pad and a 3 µm alumina slurry. CdSiP₂ crystals have tetragonal symmetry (space group $I\bar{4}$2d). Lattice constants at room temperature are a = 5.680 Å, c = 10.431 Å, and u = 0.2967 [45]. Each cadmium has four phosphorous neighbors, each silicon has four phosphorous neighbors, and each phosphorus has two cadmium neighbors and two silicon neighbors. References 23 and 41 contain detailed descriptions of the CdSiP₂ crystal structure. The c/a ratio of 1.836 for CdSiP₂ reflects the large difference in size of the cadmium and silicon atoms. In ZnGeP₂, where the zinc and germanium atoms are much closer in size, the c/a ratio is 1.962.

Optical absorption spectra were taken with a Nicolet 8700 FTIR spectrometer and a fused-silica wire-grid polarizer (Thorlabs Model WP25M-UB). The white light source was used for all wavelengths. A silicon detector and quartz beamsplitter were used for wavelengths less than 1.0 µm and a DGTS detector and CaF₂ beamsplitter were used for longer wavelengths. A Cryo Industries optical cryostat (Model 110-637-DND) with sapphire windows was used for the absorption measurements near 80 K. External light sources used to change the charge states of the defects were a HeNe laser (Thorlabs Model HNL210 L) and a Nd:YAG laser (Laserglow Model LRS-1064-PFM-00300-03). They provided 30 mW of 633 nm light and 10 mW of 1064 nm light, respectively, at the crystal. Corrections for surface reflections were made using known values of index of refraction [1,46]. The following steps were taken to reduce the amount of light on the sample from the spectrometer, thus minimizing unwanted photoinduced changes in the concentrations of the defects being monitored. For each absorption spectrum, (1) the wire-grid polarizer and the sample were oriented so that the spectrometer’s internal polarized 633 nm light was always blocked and (2) a 20% transmission internal screen was used to reduce the intensity of the white light source.

EPR spectra were taken with a Bruker EMX spectrometer operating near 9.39 GHz. The temperature of the sample was controlled with an Oxford Instruments ESR-900 flowing gas system and an ITC-503S controller. For EPR measurements at 80 K and below, the source of the cold gas was liquid helium. For the thermal recovery measurements in the 116-124 K range, the source of the cold gas was liquid nitrogen. The concentration of $\textrm{V}_{\textrm{Si}}^ - $ acceptors in the as-grown crystal was obtained by comparing the EPR spectrum to a Bruker weak-pitch standard.

3. Results

Figure 1(a) shows optical absorption spectra obtained at room temperature from the as-grown CdSiP₂ crystal. The light propagated along the [110] direction in the crystal with the electric field along either the [001] or [$\bar{1}$10] direction. This gave E ‖ c and E ⊥ c spectra, respectively. There are four broad absorption bands related to native defects in Fig. 1(a), with their peaks at room temperature near 1900, 1000, 800, and 630 nm. Note that different vertical scales are used in Fig. 1(a) (the left scale is for wavelengths longer than 730 nm and the right scale is for wavelengths shorter than 730 nm). The difference spectrum in Fig. 1(b) draws attention to the polarized bands peaking near 1900 and 630 nm that are best seen with E ‖ c. The bands near 1000 and 800 nm strongly overlap and are not polarized. The earlier study of defect-related optical absorption in CdSiP₂ by Scherrer et al. [24] suggested that these two bands, and the band near 1900 nm, are associated with singly ionized silicon vacancies ($\textrm{V}_{\textrm{Si}}^ - $). The band at 630 nm has not been previously reported (a possible assignment is suggested in the Models section). This latter absorption feature is near the band edge and is expected to play a role in two-photon absorption for OPO pump wavelengths in the 1.2-1.4 µm region. There are also two less intense broad bands peaking near 1770 and 1200 nm in the difference spectrum. These latter bands are best seen with E ‖ c and are assigned to trace amounts of Fe²⁺ and Fe⁴⁺ ions, respectively [22,47]. The Fe²⁺ band at 1770 nm appears as a shoulder on the 1900 nm native defect band.

 

Fig. 1. Optical absorption spectra taken at room temperature from an as-grown CdSiP₂ crystal. The optical path length is 2.0 mm. (a) Data were taken with light polarized parallel (red curve) and perpendicular (black curve) to the c axis. (b) The difference spectrum (E ‖ c minus E ⊥ c) shows the polarized bands.

Download Full Size | PDF

3.1 Effects of 1064 nm light

After acquiring the room-temperature spectra in Fig. 1, the CdSiP₂ crystal was cooled in the dark and two absorption spectra were recorded near 80 K with E ‖ c. These spectra, taken before (red curve) and during (black curve) an exposure to 1064 nm light, are shown in Fig. 2(a). The absorption bands from native defects are present in the spectrum taken before the exposure and are largely absent from the spectrum taken during the exposure. These removed bands are best seen in the difference spectrum in Fig. 2(b). The Gaussian-shaped dashed curves in Fig. 2(b) represent our best estimate of the peak positions and widths of the individual bands and are only meant to be guides to the eye. A detailed fitting was not attempted, as the actual shapes of the bands may deviate from Gaussian. We place the three peaks, at 80 K, at 1907, 1033, and 800 nm with full widths at half maximum (FWHM) of 0.2, 0.4, and 0.9 eV, respectively. These values for the peak positions represent small improvements on the estimates made by Scherrer et al. [24]. Following the earlier work [24] and using the EPR results described in the next paragraph, we assign these three bands to transitions involving $\textrm{V}_{\textrm{Si}}^ - $ acceptors. The absorption band near 630 nm in Fig. 1 was not affected by the 1064 nm light and is not included in Fig. 2.

 

Fig. 2. (a) Optical absorption from CdSiP₂, obtained near 80 K with E ‖ c. Spectra were taken before (red curve) and during (black curve) an exposure to 1064 nm light. (b) The difference spectrum (“before” minus “during”) shows the bands (dashed curves) eliminated by the 1064 nm light. Their peaks are at 1907, 1033, and 800 nm.

Download Full Size | PDF

The EPR spectrum in Fig. 3(a) shows that $\textrm{V}_{\textrm{Si}}^ - $ acceptors are present in the as-grown CdSiP₂ crystal (i.e., before any exposure to light). These data were obtained at 80 K with the magnetic field along the [001] direction. Equal hyperfine interactions with the four phosphorous nuclei adjacent to the vacancy are responsible for the five equally spaced lines with relative intensities 1:4:6:4:1 (³¹P nuclei are 100% abundant with I = 1/2) [23]. The concentration of $\textrm{V}_{\textrm{Si}}^ - $ acceptors in Fig. 3(a) is approximately 1.2 × 10¹⁸cm⁻³. A second spectrum, shown in Fig. 3(b), was taken at 80 K during a subsequent exposure to 1064 nm light. The two spectra in Fig. 3 can be directly compared, as the EPR spectrometer settings were the same. Nearly all the $\textrm{V}_{\textrm{Si}}^ - $ EPR spectrum is eliminated by the 1064 nm light. These results in Figs. 2(a) and 3 show that the three absorption bands (peaking at 800, 1033, and 1907 nm) and the $\textrm{V}_{\textrm{Si}}^ - $ EPR spectrum have a similar response at 80 K to 1064 nm light. This is consistent with, and further verifies, the assignment of the three absorption bands to transitions involving a singly ionized silicon vacancy ($\textrm{V}_{\textrm{Si}}^ - $).

 

Fig. 3. EPR spectra from the $\textrm{V}_{\textrm{Si}}^ - $ acceptor in CdSiP₂, taken at 80 K with the magnetic field along the [001] direction. The stick diagram above the spectra identifies the five hyperfine-split lines. (a) Before an exposure to 1064 nm light. (b) During the exposure to 1064 nm light.

Download Full Size | PDF

In Fig. 3(b), no other EPR signals appeared when the $\textrm{V}_{\textrm{Si}}^ - $ acceptors were destroyed by 1064 nm light. The same result occurs when applying the 1064 nm light at temperatures lower than 80 K. This lack of participation by a donor or another acceptor strongly suggests that silicon vacancies in the singly ionized charge state ($\textrm{V}_{\textrm{Si}}^ - $) are converted, in equal numbers, to neutral and doubly ionized charge states ($\textrm{V}_{\textrm{Si}}^0$ and $\textrm{V}_{\textrm{Si}}^{2 - }$, respectively) by 1064 nm light. These latter charge states have no unpaired spins and thus produce no EPR signals. Equally important, their formation maintains a charge neutral crystal. In Fig. 2(a), the small amount of absorption remaining in the 600 to 1000 nm region during the exposure to 1064 nm light may be caused by the $\textrm{V}_{\textrm{Si}}^{2 - }$ acceptors.

If exposure to 1064 nm light occurs at temperatures higher than 80 K, the $\textrm{V}_{\textrm{Si}}^ - $ EPR spectrum recovers soon after removal of the light. This behavior, at 120 K, is shown in Fig. 4. When the 1064 nm light is turned on, the intensity of the EPR signal decreases. Then, after the light is removed, the EPR signal returns to near its initial intensity in several hundred seconds. To acquire these data, the spectrometer was operated in a kinetics mode (a time sweep with a fixed magnetic field corresponding to the peak of the central line in Fig. 3). This experiment was repeated at 116 and 124 K, with all three recovery curves shown in Fig. 5.

 

Fig. 4. Effect of 1064 nm light, at 120 K, on the $\textrm{V}_{\textrm{Si}}^ - $ EPR spectrum in a CdSiP₂ crystal. The decrease with light and the recovery after removing the light are shown.

Download Full Size | PDF

 

Fig. 5. Thermal recovery of the EPR signal from the $\textrm{V}_{\textrm{Si}}^ - $ acceptors, following a brief exposure to 1064 nm light at 116, 120, and 124 K. The light is removed at t = 0.

Download Full Size | PDF

The recovery of the $\textrm{V}_{\textrm{Si}}^ - $ EPR signal occurs when electrons are thermally excited from the valence band maximum to neutral silicon vacancies ($\textrm{V}_{\textrm{Si}}^0$). A general-order kinetics model [48–54] is used to extract a thermal activation energy from the set of recovery curves in Fig. 5. Following the steps described in Reference 53, we find that the kinetics is second order with an activation energy of 0.27 eV (± 0.02 eV). An equivalent statement of this result places the (0/−) level of the silicon vacancy 0.27 eV above the valence band maximum. Second order kinetics indicates that the excited electrons repeatedly move back and forth between the valence band and the neutral vacancies ($\textrm{V}_{\textrm{Si}}^0$) until all the holes being formed in the valence band have been captured by doubly ionized vacancies $(\textrm{V}_{\textrm{Si}}^{2 - }$). At room temperature, this activation energy predicts that the recovery of 50% of the $\textrm{V}_{\textrm{Si}}^ - $ acceptors, after a pulse of 1064 nm light, will occur in about 5 µs [54]. Note: The pump repetition rate for some OPO’s is well above 1 MHz.

3.2 Effects of 633 nm light

The effects of 633 nm light on CdSiP₂ were also studied with EPR and optical absorption. Figure 6(a) shows the EPR spectrum taken at 80 K during an exposure to 633 nm light. The magnetic field was along the [001] direction. Three native defects, two acceptors and one donor, contribute to the spectrum. Silicon-vacancy ($\textrm{V}_{\textrm{Si}}^ - $) and silicon-on-phosphorous ($\textrm{Si}_\textrm{P}^0$) acceptors and silicon-on-cadmium ($\textrm{Si}_{\textrm{Cd}}^ + $) donors are identified by the stick diagrams above the spectrum [23,26]. The concentrations of these defects were not precisely determined since their spectra strongly overlap and they have different microwave saturation behaviors. Rough estimates place the concentrations near 1 × 10¹⁸cm⁻³. The 633 nm light produces the paramagnetic charge states of these native defects when electrons move from acceptors to the conduction band and then are trapped at donors. Figure 6(b) shows the EPR spectrum taken during a subsequent exposure to 1064 nm light. The charge states of the three native defects seen in Fig. 6(a) are eliminated by this light (i.e., the defects are converted to nonparamagnetic charge states).

 

Fig. 6. EPR spectra from CdSiP₂, taken at 80 K with the magnetic field along the [001] direction and the same spectrometer settings. Stick diagrams above the spectra identify the lines from two acceptors and one donor. (a) During an exposure to 633 nm light. (b) During a subsequent exposure to 1064 nm light.

Download Full Size | PDF

Figure 7(a) shows additional optical absorption results obtained at 80 K from the CdSiP₂ crystal. Spectra were taken before exposure to light (spectrum 1), then during an exposure to 633 nm light (spectrum 2), and finally during an exposure to 1064 nm light (spectrum 3). The 633 nm light increased the absorption across the entire region from 0.6 to 1.9 eV. Some of this increase in absorption is due to the $\textrm{Si}_\textrm{P}^0$ acceptors formed by the 633 nm light [26]. During the subsequent exposure to 1064 nm light, the absorption was reduced below the values seen before the 633 nm exposure. In other words, the 1064 nm light removed the absorption created by the 633 nm light plus a significant portion of the absorption present before any light. The difference spectrum (“during 633” minus “during 1064”) in Fig. 7(b) shows the removed absorption. This ability of 1064 nm light to eliminate optical absorption, either initially present in an as-grown crystal or produced with 633 nm light, is consistently seen in Figs. 2 and 7.

 

Fig. 7. (a) Optical absorption taken at 80 K with E ‖ c. (1) Before exposure to light, (2) During exposure to 633 nm light, (3) During exposure to 1064 nm light. (b) Difference spectrum (“during 633 nm” minus “during 1064 nm”).

Download Full Size | PDF

4. Models for the absorption bands

CdSiP₂ is a pseudodirect gap semiconductor with a “soft” band edge extending from 2.2 to 2.4 eV (i.e., there is not a sharp rise in absorption at the fundamental edge) [44]. Electronic structure calculations confirm that CdSiP₂ (and ZnGeP₂) crystals have several conduction-band minima with similar energies at different points in the Brillouin zone and thus are not simple direct gap materials [44,55]. The three valence bands in CdSiP₂ are split by a large internal crystal field and a smaller spin-orbit interaction. Goryunova et al. [43] predicted that the splitting due to the crystal field (Δ_cf) is 0.56 eV and Hübner and Unger [56] calculated a value of approximately 0.1 eV for the spin-orbit splitting (Δ_so). The crystal-field splitting for CdSiP₂ is considerably larger than for other II-IV-P₂ chalcopyrites because of the mismatch in size of the cadmium and silicon atoms [42]. As illustrated in the band diagram in Fig. 8, there is one upper valence band (labeled VB1) and two closely spaced, spin-orbit split, lower valence bands (labeled VB2 and VB3). The relative positions of the three charge states of the silicon vacancy ($\textrm{V}_{\textrm{Si}}^0$, $\textrm{V}_{\textrm{Si}}^ - $, and $\textrm{V}_{\textrm{Si}}^{2 - }$) are included in Fig. 8.

 

Fig. 8. Band diagram for a CdSiP₂ crystal, including the charge states of the silicon vacancy ($\textrm{V}_{\textrm{Si}}^0$, $\textrm{V}_{\textrm{Si}}^ - $, and $\textrm{V}_{\textrm{Si}}^{2 - }$). Transitions responsible for the three residual optical absorption bands are shown as blue vertical lines and the thermally excited transition is shown as a red vertical line. The position of the $\textrm{V}_{\textrm{Si}}^{2 - }$ state is an estimate.

Download Full Size | PDF

The experimental results in Section 3.1 show that the singly ionized silicon vacancy ($\textrm{V}_{\textrm{Si}}^ - $) is the native defect responsible for the three residual absorption bands peaking near 800, 1033, and 1907 nm. These bands are broadened (see Fig. 2) by strong electron-phonon interactions. In Fig. 8, the 1907 nm absorption band is assigned to the transition of an electron from the upper valence band (VB1) to a $\textrm{V}_{\textrm{Si}}^ - $ acceptor and the 1033 nm band is assigned to the transition of an electron from the closely spaced lower valence bands (VB2 and VB3) to a $\textrm{V}_{\textrm{Si}}^ - $ acceptor. The difference in energy of the 1033 and 1907 nm peaks is 0.55 eV. This is nearly the same as the 0.56 eV crystal-field splitting of the valence bands predicted by Goryunova et al. [43]. We suggest that the 800 nm absorption band is an intracenter transition (ground state to excited state) of a $\textrm{V}_{\textrm{Si}}^ - $ acceptor with, perhaps, a small contribution from an acceptor-to-donor transition also originating from the $\textrm{V}_{\textrm{Si}}^ - $ acceptor. The unrelaxed first excited state of the $\textrm{V}_{\textrm{Si}}^ - $ acceptor will be just below the conduction band and, if formed, will quickly relax to a slightly lower energy state before returning nonradiatively to the ground state. In support of this model, we note that intracenter transitions have been observed for singly ionized vacancies in ZnSe [57]. Also, in ZnGeP₂, an intracenter transition and an acceptor-to-donor transition are invoked to explain the dominant 1200 nm absorption band associated with singly ionized zinc vacancies ($\textrm{V}_{\textrm{Zn}}^ - $) [31,33]. Future studies will refine the model for the 800 nm absorption band in CdSiP₂.

The present report shows that 1064 nm light, when applied at low temperature, can destroy singly ionized silicon vacancies ($\textrm{V}_{\textrm{Si}}^ - $) in CdSiP₂ and thus eliminate residual absorption bands. We propose the following mechanism to explain how 1064 nm light removes $\textrm{V}_{\textrm{Si}}^ - $ acceptors and produces $\textrm{V}_{\textrm{Si}}^0$ and $\textrm{V}_{\textrm{Si}}^{2 - }$ acceptors. As a result of the crystal-field splitting of the valence bands, 1064 nm light can directly “pump” electrons from the VB2 and VB3 valence bands to $\textrm{V}_{\textrm{Si}}^ - $ acceptors and form $\textrm{V}_{\textrm{Si}}^{2 - }$ acceptors. The energy of the 1064 nm photons (1.165 eV) is very close to the 1.20 eV separation between the VB2 and VB3 valence bands and the $\textrm{V}_{\textrm{Si}}^ - $ state in Fig. 8. Holes left behind in the two lower valence bands will move to the upper valence band and then become trapped at $\textrm{V}_{\textrm{Si}}^ - $ acceptors and form $\textrm{V}_{\textrm{Si}}^0$ acceptors. At temperatures near and below 80 K, the $\textrm{V}_{\textrm{Si}}^0$ and $\textrm{V}_{\textrm{Si}}^{2 - }$ acceptors produced by this mechanism will be thermally stable. If the 1064 nm light is sufficiently intense and the temperature is sufficiently low, this conversion process will quickly reach an equilibrium state where half of the removed $\textrm{V}_{\textrm{Si}}^ - $ acceptors have become $\textrm{V}_{\textrm{Si}}^0$ acceptors and half have become $\textrm{V}_{\textrm{Si}}^{2 - }$ acceptors. Warming above approximately 120 K, as seen in Fig. 4, restores the initial charge state of the silicon vacancies.

Finally, we turn to a model for the 630 nm optical absorption band seen in Fig. 1. Cadmium vacancies are present in all CdSiP₂ crystals with concentrations approximately the same as the silicon vacancies [23]. These cadmium vacancies are shallow acceptors and are in the doubly ionized charge state ($\textrm{V}_{\textrm{Cd}}^{2 - }$) due to compensation [23,25]. A likely candidate for the transition responsible for the 630 nm band is the excitation of an electron from a $\textrm{V}_{\textrm{Cd}}^{2 - }$ acceptor to the minimum of the conduction band. This could occur if the $\textrm{V}_{\textrm{Cd}}^{2 - }$ state is approximately 0.20 eV above the maximum of the valence bands [25].

5. Summary

Singly ionized silicon vacancies ($\textrm{V}_{\textrm{Si}}^ - $) are responsible for residual optical absorption bands peaking near 800, 1033, and 1907 nm in CdSiP₂ crystals. These absorption bands affect the performance of the crystals when used in optical parametric oscillators. In CdSiP₂, the cadmium and silicon atoms have different covalent radii. This introduces a large internal crystal field and causes the valence bands to be split by 0.56 eV. The 1033 and 1907 nm absorption bands are assigned to transitions from the lower and upper valence bands, respectively, to a $\textrm{V}_{\textrm{Si}}^ - $ acceptor. The 800 nm band is assigned to an intracenter (ground state to excited state) transition of $\textrm{V}_{\textrm{Si}}^ - $ acceptors, with a possible contribution from an acceptor-to-donor transition. By identifying silicon vacancies as the native defect responsible for residual optical absorption in CdSiP₂, efforts can now be made, during growth or in post-growth treatments, to remove the unwanted charge state. This could occur either (1) by producing crystals with fewer silicon vacancies or (2) by doping with a shallow donor that would compensate all the silicon vacancies (i.e., convert the silicon vacancies to the more benign doubly ionized charge state).