Interactions of Hydrogen Atoms with Acceptor–Dioxygen Complexes in Czochralski‐Grown Silicon

It is debated in the silicon PV community whether or not the presence of hydrogen is essential for the permanent suppression (“regeneration”) of the recombination activity of the boron–oxygen (BO) defect, which is responsible for light‐induced degradation (LID) of solar cells produced from B‐doped oxygen‐rich silicon. The BO‐LID defect has been identified as a BsO2 complex which has negative‐U properties. This study focuses on the interactions of hydrogen with the BsO2 defect to elucidate the BO‐LID regeneration mechanism. With the use of junction spectroscopy techniques, the changes in concentration of the BsO2 donor state in diodes which are fabricated on Czochralski‐grown (Cz) B‐doped Si and subjected to hydrogenation and subsequent heat treatments have been monitored. It is found that annealing of the hydrogenated Cz‐Si:B diodes in the temperature range 398–448 K under the application of reverse bias (RBA) results in nearly total disappearance of the BsO2 defect. It is argued that electrically neutral BsO2–H complexes have been formed upon the RBA treatments. According to ab initio calculations, the binding energy of H+ to BsO2− exceeds that of H+ to Bs− by at least 0.1 eV, and the resulting BsO2–H complexes are electrically inactive.


Introduction
Solar cells made from crystalline silicon doped with boron and containing significant amounts of oxygen suffer from the degradation of minority carrier lifetime and conversion efficiency upon the early stages of operation under illumination, commonly called "light-induced degradation (LID)." The detrimental effect of this degradation mechanism on the performance of the Si solar cells (solar conversion efficiency drops up to 10% relative or 2% absolute over time) has driven intense research efforts to eliminate the impact of recombination activity associated with the formation of the boron-oxygen (BO) defect, particularly in Czochralski-grown boron-doped silicon material (Cz-Si:B). It is well established that the minority carrier lifetime and conversion efficiency degradation process is associated with a complex of one boron atom and two oxygen atoms. [1,2] Compelling evidence was recently presented that the complex, which consists of a substitutional boron atom and the oxygen dimer (B s O 2 defect), is a defect with negative-U properties. [3][4][5] The B s O 2 defect undergoes a transformation from the deep donor state, which is recombination inactive, into a shallow acceptor state with enhanced recombination activity upon prolonged minority carrier injection. [3,4] It has been argued that the B s O 2 defect in the shallow acceptor state is responsible for BO-LID. [3,4] The interactions between other substitutional acceptor atoms, A s (A s being either Ga, Al, or In), and the interstitial oxygen dimer in p-type Cz-grown silicon have been also reported resulting in the formation of A s O 2 defects. [5] All members of the family of defects formed due to the interactions of group III elements with the oxygen dimer in silicon exhibit negative-U properties and are deep donors in the ground state in p-type Si.
Upon illumination, the BO defect, which gives rise to LID, is converted from an initial state often named the "annealed state" to a degraded state which is metastable and can be transformed back to the annealed state (unstable recovery) by dark annealing at 200°C for several minutes. [1,2] Importantly, the degraded state can be converted into a stabilized state, [2,6,7] also called the "regenerated state" [2] via illuminated annealing at slightly elevated temperatures. [6][7][8][9] This state is characterized by the stable recovery of the minority carrier lifetime back to predegradation values via the regeneration treatment. Unfortunately, the destabilization, which is the reverse reaction from the regenerated state to the annealed state, can still occur via dark annealing at 200°C for about 100 min.
The hypothesis that hydrogen plays a key role in the regeneration mechanism of the BO-LID defect has been proposed by many research groups after being reported in 2009. [9] It has been shown that the presence of mobile atomic hydrogen in the silicon bulk is necessary for regeneration and that no or only extremely slow regeneration is possible without hydrogen. [10][11][12][13][14] The hydrogen is usually introduced into the bulk regions of Si solar cells via diffusion from hydrogen-rich silicon nitride layers during hightemperature firing, which is necessary for formation of metal contacts for the cells. The effective BO-LID regeneration requires an appropriate amount of hydrogen to be present in the bulk of the Si material. From the analysis of BO-LID regeneration kinetics, the following influencing factors have been determined: 1) the presence of a hydrogen source to supply enough amount of hydrogen for complete regeneration (e.g., H-rich silicon nitride layers); [11][12][13][14] 2) firing temperature, which affects the introduction of hydrogen into the bulk; [11,13,14] and 3) cooling rate after firing. [11] The properties of the nitride layers, i.e., their thickness, SiH 4 and NH 3 gas flow ratio, and layer composition, were found to affect the amount of hydrogen diffused into the bulk upon firing. This is in agreement with the finding that the different structures of surface coating (SiO 2 /SiN x , Al 2 O 3 /SiN x ) have no influence on the BO-LID regeneration rate as long as they introduce the same amount of hydrogen into the bulk. [13] The introduction of hydrogen from the passivation layers can also be enhanced by the firing conditions. It has been found that the regeneration rate is accelerated by a higher firing temperature and a longer duration at this temperature (known as the peak width). [11] Further, a fast cooling down after firing is needed for effective regeneration to occur. It has been argued that the fast cooling down to 500°C is necessary to stop the net effusion of hydrogen from the silicon bulk, which decreases the hydrogen content available for the passivation of the BO defect. [13,15] It has been suggested that oxygen also plays a role in the BO deactivation process. [15] The evidence presented for this is that the kinetics of deactivation reaction of the BO defect is inversely affected by the presence of interstitial oxygen and/or oxygen clusters (thermal donors), which result in a lower deactivation rate and smaller value of lifetime after complete deactivation. [15] Moreover, it has been argued that hydrogen is responsible for passivation of background unknown defects, which affect the lifetime, and it is not required for the BO-LID regeneration. [14] In a more recent study, it was proposed that the overall regeneration of the BO defects is composed of two different mechanisms. [16] It was found that while hydrogen plays no role in the BO-related lifetime degradation mechanism, it is only involved in one of the regeneration mechanisms while the other can occur without hydrogen. [16] A possible source of hydrogen within the Si, important for regeneration, is the B-H pairs. It was proposed that during regeneration conditions, splitting of B-H pairs can occur (under carrier injection) which increases the concentration of bulk hydrogen available to contribute to regeneration. [17] Any thermal conditions (e.g., slower cooling rate after high-temperature firing), which result in the formation of more stable hydrogen complexes (compared to B-H) such as H 2 molecules, can have a detrimental effect on the kinetics of the BO regeneration. It was proposed that the passivation of BO defects by neutral hydrogen atoms, H 0 , is responsible for the regeneration. [17] The regeneration reaction can be accelerated by hightemperature/high-injection conditions. It has been argued that these conditions are needed to enhance the defect degradation rate, rather than the defect passivation rate, subsequently improving the effectiveness of regeneration. [18][19][20] With the aim of using the regeneration treatments commercially, it has been shown that regeneration of hydrogenated samples can be completed in less than 10 s at 200-230°C under 2.7 suns illumination. [12,18,20] The BO-LID regeneration can be further accelerated (completed in 0.5 s) via a laser illumination treatment, which provides simultaneous heat and carrier injection. [13] In this work, we present experimental results which support the arguments that the regeneration of the BO defect is not possible without hydrogen. Further, we report experimental evidence of the formation of electrically inert B s O 2 -H complexes due to the interaction of hydrogen atoms with the B s O 2 defect. We use capacitance-voltage (C-V ) measurements to monitor the changes in the uncompensated ionized shallow acceptor profiles N À A (W ), which can be correlated with the movement of hydrogen in the samples upon different heat treatments. Conventional deep-level transient spectroscopy (DLTS) and high-resolution Laplace DLTS measurements were carried out on hydrogen-free diodes and on diodes hydrogenated either via wet chemical etching or via treatments in remote H plasma at room temperature. Measurements were carried out before and after reverse bias annealing (RBA) treatments of the hydrogenated samples at about 150°C, which resulted in 1) the transformation of the The n þ -p diodes used in this study have been fabricated with no significant amount of hydrogen introduced into the base regions during processing. The goal of this part of our work is to investigate the effects of regeneration-like treatments (heat treatments in the range of 403-473 K with simultaneous minority carrier injection induced by forward current flow) on the changes in concentration of the B s O 2 defects in the samples with negligible concentrations of hydrogen. In the conventional DLTS spectra recorded on all n þ -p diodes studied in this work, a hole-emission related peak with its maximum at about 390 K has been detected. This peak is assigned to hole emission from the donor state of the B s O 2 complex, which is a precursor of the recombination active BO-LID defect (annealed state of the BO defect). [3,4] The details of the electronic structure of the B s O 2 defect, parameters of the carrier emission and capture processes for its different configurations, peculiarities of its observation by junction spectroscopy techniques, and changes in the DLTS spectra induced by minority carrier injections and dark annealing treatments at different temperatures have been reported earlier in refs. [3,4]. In this work, the changes in the concentration of the B s O 2 donor state have been monitored with the use of high-resolution DLTS measurements upon minority carrier injection (MCI) treatments with a current density J FB ¼ 1.5 A cm À2 at 350 and 423 K and upon dark annealing at 473 K after the carrier-injection treatments. The treatment at 350 K corresponds to a BO-related carrierinduced degradation regime and the treatment at 423 K to a BO-LID regeneration regime. Figure 1 compares the evolution of the B s O 2 donor state concentration upon carrier-injection treatments at 350 and 423 K.
The BO degradation mechanism is activated by minority carrier injection at 350 K and has been shown to be related to the transformation of the B s O 2 defect from the deep donor to the shallow acceptor state. [3,4] This transformation is shown in Figure 1 by the exponential decay of the B s O 2 donor state concentration upon continuous injection of minority carriers. The complete recovery of the defect concentration after dark annealing at 473 K for 10 min is another well-established feature of the BO-LID related processes. Further, the regeneration-like treatment, i.e., injection of minority carriers at slightly elevated temperature (423 K in this case), [7,20,21] resulted in a similar but faster decrease in the concentration of the B s O 2 donor state. Importantly, upon short time dark annealing at 473 K, a nearly complete recovery of the B s O 2 donor state has been observed in the sample subjected to the regeneration treatment. This indicates that this treatment has not resulted in the true regeneration, i.e., the transformation of the B s O 2 defect into the stable regenerated state (if the regenerated state is formed, no recovery of the B s O 2 defect occurs after short time annealing at 473 K). It should be mentioned that the regeneration treatment in this case has been started when the B s O 2 defect was in the annealed state. According to the literature results, the regeneration process occurs when the BO-LID defect is in the degraded state. [22] To study the effect of the initial state of the B s O 2 defect on the results of the regeneration treatments, further DLTS and highresolution Laplace DLTS measurements have been carried out on n þ Àp diodes from 10 Ω cm Cz-Si:B material. We have detected a similar hole emission-related peak with its maximum at 390 K in the conventional DLTS spectra of these diodes. The magnitude of the detected peak is slightly lower than that in the n þ -p diodes with a 3 Ω cm base that is expectable due to the lower boron concentration. The changes in concentration of the B s O 2 donor state versus the minority carrier injection time at 423 K are shown in Figure 2, left. The effects of subsequent different injection/dark annealing treatments on the concentration of the B s O 2 donor state are also presented in colored-filled part of Figure 2.
Upon minority carrier injection at 423 K, the concentration of the B s O 2 donor state decays exponentially with time down to a saturation value after which no further decrease is attained upon longer forward bias injection. This can be explained by the competition between the degradation and annealing reactions at this temperature. [22] The annealing reaction results in the recovery of the concentration of the B s O 2 donor state, while degradation reaction leads to the disappearance of this state. As the degradation reaction is slightly faster than the annealing reaction at 423 K, the resultant saturation value of the B s O 2 donor state concentration in this case is fixed at about 3.5 Â 10 12 cm À3 . The www.advancedsciencenews.com www.pss-a.com same effect can be observed in Figure 1 for the diode from 3 Ω cm material. However, in this material the degradation rate, which is proportional to squared hole concentration, [1,2] is significantly faster than the annealing rate, so nearly all the B s O 2 defects are transformed to the degraded state after a long treatment.
After the first regeneration treatment, it can be observed that the dark annealing results in the total recovery of the B s O 2 donor state concentration in agreement with the results presented in Figure 1. To convert the annealed state into the degraded state, a one-step degradation treatment (forward-bias-induced minority carrier injection at 370 K for 150 min) has been carried out which results in the complete disappearance of the B s O 2 defect deep donor state. At 370 K, the annealing rate is very slow and thus the degradation reaction results in the complete disappearance of the donor state of the B s O 2 defect.
Subsequently, the one-step regeneration treatment has been applied to the diode with the B s O 2 defect in the degraded state (to compare its result with that of the treatment which started with the B s O 2 defect in the annealed state in the previous experiment). This treatment resulted in the partial recovery of the B s O 2 defect donor state to the saturation value which is close to that obtained after the first detailed regeneration treatment. The stability of the obtained decrease in the concentration of the B s O 2 defect donor state (relative to the initial concentration) has been examined via dark annealing at 443 K. Again, the total recovery of the B s O 2 defect concentration has been achieved after the annealing treatment.
Overall, the total recovery of the concentration of the deep donor state of the BO defect by annealing means that the degradation will occur again upon minority carrier injection at room temperature. It can be concluded that what has been referred to as "regeneration" in the experiments described above and presented in Figure 1 and 2 by the blue squares is not an actual regeneration reaction, but rather a degradation reaction with partial recovery due to the annealing reaction. The initial state of the BO defect, to which the regeneration treatment is applied (annealed state or degraded state), has no effect on the resultant state after the regeneration-like treatments ( Figure 2). Importantly, we have not obtained any evidence of the permanent disappearance of the B s O 2 defect upon regeneration-like treatments in the Cz-Si:B samples, which contain negligible concentrations of hydrogen.

Hydrogen Introduced via Wet Chemical Etching
To investigate the interaction between hydrogen atoms and the B s O 2 defect, Schottky barrier diodes (SBDs) have been fabricated on the boron-doped Cz-Si materials. For this presentation, we have chosen a material with high initial concentration of the B s O 2 defect for the sake of clear monitoring of any changes in the defect concentration after treatments. It is well known that the chemical treatments of p-type Si materials we have used introduce some hydrogen, which passivates electrical activity of shallow acceptor impurities in the subsurface region of samples. [23] Particularly, we have carried out wet chemical etching of some samples intentionally to increase the concentration of hydrogen, which is positively charged in p-type materials, in the subsurface regions. We have monitored the changes in the concentration of ionized uncompensated boron atoms by an analysis of N À A (W ) dependencies obtained from the capacitance-voltage (C-V ) measurements as shown in Figure 3a. The strong drop in the concentration of uncompensated ionized boron atoms close to the surface in the initial profile (black line in Figure 3a) is due to the interaction of H þ with B À s , thus forming neutral BH complexes. [23] The N À A (W ) values for depth > 1.1 μm, i.e., in the deeper bulk region, are nearly constant at about 1.1 Â 10 16 cm À3 .
We have used C-V measurements to choose the reverse bias and the filling pulse voltages to be used in conventional DLTS measurements in order to probe the bulk region which is hydrogen-free initially. The resultant DLTS spectrum in the  www.advancedsciencenews.com www.pss-a.com temperature range 300-425 K on the initial (as-processed) diode is presented in Figure 3b as the black line (spectrum 1). We have detected a hole emission signal with its maximum at about 390 K related to the deep donor state of the B s O 2 defect with relatively high concentration of about 2 Â 10 14 cm À3 . Subsequently, the diode was kept at 425 K with the applied reverse bias of 12 V for 10 min. This treatment, referred to as RBA treatment, provides the thermal energy needed to release hydrogen from the B s -H complexes close to the surface and promote the electric field-induced drift of positively charged hydrogen atoms into the bulk region. It can be observed in Figure 3a that the RBA treatment caused significant changes in the N À A (W ) profile. The concentration of the uncompensated ionized shallow acceptor atoms is increased in the subsurface region but is decreased deeper in the bulk region. This indicates that positively charged hydrogen atoms were moved into the bulk region where they interact with the negatively charged boron atoms. To monitor the changes in the concentration of the B s O 2 defect after the RBA treatment, we have recorded another conventional DLTS spectrum in the temperature range 300-425 K in the same depth region, which was probed in the DLTS measurements before the RBA treatment (Figure 3b). A significant drop in the concentration of the deep donor state of the B s O 2 defect from %1.8 Â 10 14 to %4 Â 10 13 cm À3 has been induced by the RBA treatment. To examine the stability of the RBA-induced effect, we have carried out annealing of the sample at 443 K for 30 min without bias. This annealing treatment is known to result in the recovery of the annealed state of the B s O 2 defect after LID. The N À A (W ) profile, shown in Figure 3a, indicates that the hydrogen content has been redistributed during the annealing treatment between the subsurface region and the bulk region that resulted in an approximately constant uncompensated ionized shallow acceptor concentration of about 9 Â 10 15 cm À3 all over the depletion region. For the bulk region (depth >1.1 μm), this value is less than the initial bulk N À A value before any treatments but higher than the value reached after RBA which means that hydrogen boron pairs are dissociated in the bulk region upon the annealing treatment. For the subsurface region (depth < 1.1 μm), the N À A value after annealing is higher than the initial value but less than the value after RBA. So, a comparison of the N À A (W ) profiles in Figure 3a indicates that hydrogen diffused in all directions when annealing is performed without bias at 443 K. Importantly, no recovery of the DLTS signal was obtained after the annealing treatment, but we have observed a further reduction in the concentration of the deep donor state of the B s O 2 defect to a very low concentration of %9.3 Â 10 11 cm À3 (Figure 3b). The reduction in the concentration of the B s O 2 defect upon the heat treatment at 443 K is related to its passivation by hydrogen atoms released from the unstable B-H pairs at this temperature, that is consistent with the observed change in the N À A (W ) profiles in Figure 3a. This finding shows that the observed drop in the B s O 2 defect concentration after RBA at 425 K is stable at a temperature of 443 K.

Hydrogen Introduced via Remote Plasma
To further investigate the effect of hydrogenation on the B s O 2 defect concentration, we have carried out similar measurements on another set of SBDs fabricated on the same boron-doped material. The samples from this set were hydrogenated via remote radiofrequency (RF) hydrogen plasma prior the deposition of metals for fabrication of Schottky diodes. The hydrogen plasma treatment was carried out to introduce significant amounts of hydrogen exceeding those introduced via wet chemical etching. The initial (as-processed) concentration-depth profile of a H plasma-treated sample is shown in Figure 4a. It can be observed that the hydrogenation via the H plasma treatment introduced significant amounts of hydrogen into the subsurface region which results in the steep drop of concentration of N À A from %1.5 Â 10 16 cm À3 in the bulk region to nearly zero at %0.7 μm from the surface. The N À A drop in the subsurface region was less powerful in samples hydrogenated via wet chemical etching compared to that in the samples hydrogenated via H plasma as can be seen from a comparison of the N À A (W ) initial profiles in Figure 3a and 4a. This confirms that the treatment in , in a boron-doped Cz-Si sample, which was hydrogenated via wet chemical etching (initial) and subjected to RBA at 425 K and a heat treatment (HT) at 443 K without bias applied. The profiles have been calculated from the C-V dependences measured at 320 K. b) Conventional DLTS spectra recorded in the temperature range 300-425 K with rate window R ¼ 10 s À1 and filling pulse width t p ¼ 200 ms on the same sample before and after the aforementioned treatments. The bias and pulse voltages have been selected according to C-V measurements to probe the same region between 1.1 and 1.45 μm from the surface in the three cases.
www.advancedsciencenews.com www.pss-a.com remote hydrogen plasma introduces higher concentration of hydrogen into the subsurface region.
We have carried out the same experiments on the samples hydrogenated via the treatment in H plasma. Figure 4b shows the conventional DLTS spectra recorded in the temperature range 300-425 K on a H plasma hydrogenated diode before and after an RBA treatment at 425 K and annealing treatment at 443 K. A very sharp drop in the concentration of B s O 2 defect occurs as a result of the RBA treatment as can be seen in Figure 4b. The RBA treatment induced the dissociation of the initially formed BH pairs in the subsurface region and the drift of positively charged hydrogen atoms into the bulk region (as monitored via N À A (W ) profiles in Figure 4a). Relative to the initial magnitude of the hole emission-related signal with its maximum at about 390 K in the DLTS spectra, the sample hydrogenated via H plasma has shown about 25 times smaller peak magnitude after the RBA treatment compared to about 4.5 times drop in the peak magnitude in the sample hydrogenated via wet chemical etching (Figure 3b vs Figure 4b). A further small drop in the concentration of the B s O 2 defect has been observed after dark annealing at 443 K of the H plasma-hydrogenated sample, indicating that the RBA-induced decrease in the B s O 2 defect concentration is thermally stable upon dark annealing at 443 K (Figure 4b).
We have carried out detailed Laplace DLTS measurements to obtain the spatial profiles of the B s O 2 defect concentration in the H plasma-hydrogenated sample. Figure 5 shows the concentration profiles of the B s O 2 defect over the depletion region depth before and after an RBA treatment. It can be observed that the RBA treatment results in the decrease in the B s O 2 defect concentration in all the studied regions compared to approximately uniform profile in the initial hydrogenated sample. However, the scale of the RBA-induced decrease in the B s O 2 defect concentration is not uniform and at least three different regions can be identified: the deep bulk region at depth > 1.  Figure 4. a) Depth profiles of the ionized uncompensated dopant concentration, N À A (W ), in 1 Ω cm boron-doped Cz-Si sample, which was hydrogenated via remote RF hydrogen plasma (initial) and subjected to RBA at 425 K and a HT at 443 K without bias applied. The profiles have been calculated from the C-V dependences measured at 320 K. b) Conventional DLTS spectra recorded in the temperature range 300-425 K with rate window R ¼ 10 s À1 and filling pulse width t p ¼ 200 ms on the same sample before and after the aforementioned treatments. The bias and pulse voltages have been selected according to C-V measurements to probe the same region between 1.0 and 1.35 μm from the surface in the three cases. www.advancedsciencenews.com www.pss-a.com pairs as can be seen in the initial N À A (W ) profile in the same figure. Thus, only a part of the available hydrogen was involved in the formation of BH pairs in the bulk region, while another part interacted with the B s O 2 complexes. In the subsurface region, the concentration of hydrogen introduced via plasma was very high initially, but the majority of these hydrogen atoms were bound to boron in the form of BH pairs. This means that hydrogen was indeed not available for the interaction with B s O 2 complexes in this region. During RBA treatment, hydrogen atoms were released from the BH pairs but the majority of H þ species drifted in the electric field into the bulk region to a depth which depends on the applied reverse bias. This gives only small chance for the interaction of the free hydrogen atoms with the B s O 2 defects in this region and, thus, the observed drop in the B s O 2 defects concentration is minimal. In between these two regions, a transition in the concentration of hydrogen from a minimum value to a maximum value occurs. This is reflected by the inverse trend of the decrease in the concentration of the B s O 2 defects from a maximum to a minimum value upon the application of RBA unlike the initial case where it was almost uniform.

Modeling of Hydrogen Interactions with the B s O 2 Complexes
Previous work indicated that the most stable B s O 2 complexes in Si comprise an interstitial oxygen dimer next to a substitutional boron atom, B s , with no direct B─O bonds formed. [3,5,24] Importantly, B s O 2 is bistable, while in p-type Si it is a deep donor (D-state), where the oxygen dimer adopts a "square" configuration; under reverse bias conditions (or under minority carrier injection) the defect converts to a shallow acceptor state (A-state) with the O 2 unit in a staggered configuration. Besides the A-form, at least two metastable shallow acceptor structures, named A 0 and A 00 , with energy 0.23 and 0.38 eV above A were found. It has been suggested that the B s O 2 complex in A 0 form is responsible for the lifetime degradation. [3] Our calculations indicate that the interactions between hydrogen and the B s O 2 complex in D-state (deep donor) results either in complexes with marginal binding energies or in the breaking of the squared O 2 units, which ultimately convert to a staggered form after atomic relaxation. These results suggest that besides the hindering effect of Coulomb repulsion between BO þ 2 ðDÞ and H þ , thermodynamics also works against the formation of B s O 2 (D)-H in p-type Si.
The attachment of H þ to structures A, A 0 , and A 00 of BO À 2 leads to the formation of stable B s O 2 -H complexes. The two lowest energy configurations are depicted in Figure 6a, (1)-(3) were already studied in ref. [25]. They show that the binding energy of a bond-centered proton (ground state of hydrogen in p-type Si) to B À s is 0.73 eV. We also found, as reaction (2) shows, that boron can bind at least to two H atoms (ΔE ¼ À0.40 eV for the capture of the second proton), and that the exchange of H between BH pairs is not energetically favored (ΔE ¼ 0.34 eV). More importantly in the present context, reaction (4) and (5) show that the binding energy of H to B s O 2 exceeds the analogous quantity of H to substitutional boron by about 0.1 eV (see reaction (6) and (7)).

Formation of A s O 2 -H Complexes in Al-and Ga-Doped Cz-Si Materials
It has been previously shown that the oxygen dimer interacts with any substitutional group III acceptor atoms in silicon  Table 1. Energetics of several reactions of interest involving hydrogen, boron, and BO 2 complexes in Si. Negative energies balances refer to exothermic reactions. Defects are referred using the usual notation-superscripted symbols refer the charge state, while HBC stands for bond-centered hydrogen. www.advancedsciencenews.com www.pss-a.com and forms A s O 2 (A s ¼ B, Al, Ga, or In) complexes with very similar electronic properties. [5,26] All the A s O 2 complexes are found to be defects with negative-U properties. The values of the E(-/þ) occupancy level, activation energy for hole emission from the donor state, energy barriers, and frequencies for transition between different configurations of the different A s O 2 complexes have been determined and compared in refs. [5,26]. Evidence of the interactions between hydrogen and the complexes made of a substitutional acceptor atom and the oxygen dimer are obtained in this study by means of conventional DLTS measurements on hydrogenated silicon crystals doped with aluminum and gallium impurities.
We have processed the samples from Cz-Si wafers doped with either Ga or Al in a similar way as the wafers doped with boron were processed. Hydrogen was introduced into the Cz-Si:Ga and Cz-Si:Al samples via wet chemical etching. A similar hole emission-related signal with its maximum at about 390 K has been detected in the conventional DLTS spectra of the Ga-and Al-doped materials as shown in Figure 7 and 8. Figure 7 shows the DLTS spectra recorded in the temperature range 300-425 K before and after RBA of a diode made on a hydrogenated Ga-doped Cz-Si sample. A decrease in the concentration of Ga s O 2 complexes has been observed, but the decrease is much less significant than in the case of boron-doped samples. The N À A (W ) profiles (shown in the inset of Figure 7) indicate that large amounts of hydrogen atoms have moved into the bulk region after RBA and passivated Ga À s atoms in the bulk. A comparison of the ratios of passivated Ga À s and Ga s O À 2 defects to their initial values in the bulk regions suggests that the hydrogen interacts more efficiently with single gallium atoms than with the Ga s O 2 complexes. Figure 8 shows the changes in the concentration of the Al s O 2 complexes in a hydrogenated Cz-Si:Al sample after RBA treatments at different temperatures. It can be seen that the RBA treatment at 420 K of the Al-doped Cz-Si sample results in a relatively small decrease in the concentration of Al s O 2 complexes (see Figure 8). Only after RBA with the same reverse bias voltage (À10 V) at 440 K is applied, a significant drop in the concentration of Al s O 2 complexes has been observed. The N À A (W ) profiles (shown in the inset of Figure 8) indicate that relatively small concentration of hydrogen moved into the bulk from sub-surface regions after RBA at 420 K. The concentration of hydrogen relocated into the bulk has increased significantly after RBA at 440 K and the concentration of the Al-H pairs near the surface has decreased. The results obtained show that the Al-H pairs are dissociated more effectively at 440 K, and hydrogen atoms, which have been moved into the bulk regions upon RBA, interact effectively with both Al À s and Al s O À 2 defects.

Discussion
Recently, the BO-LID has been linked with the structural transformation of the B s O 2 complex, and the peak with its maximum at about 390 K detected in conventional DLTS spectra has been assigned to hole emission from the donor state of this complex. [3,4] The details of the electronic properties of the B s O 2 defect and its transformations upon the injection of minority carriers have been reported. [3] The B s O 2 complex in the donor state can be considered as a precursor of the recombination active BO-LID defect, and its concentration can be monitored by measurements of the strength of the emission signal with its maximum at about 390 K. Disappearance of this hole emission signal can be caused by either: 1) the degradation reaction, by which the B s O 2 defect is converted from the deep donor (annealed state) into the shallow acceptor state (degraded state) upon minority carrier injection at room temperature or slightly above it; and 2) regeneration reaction, by which the B s O 2 defect is converted into the regenerated state upon minority carrier injection at slightly elevated temperature (398-473 K). Subsequent annealing in dark at about 473 K for about 10 min can be used to evaluate which of these reactions have occurred. The recovery of the B s O 2 -related hole emission signal after dark annealing at 473 K shows that the degradation reaction occurred, while the  www.advancedsciencenews.com www.pss-a.com absence of the dark-annealing-induced recovery of the signal indicates that the regeneration reaction happened. The total dark-annealing-induced recovery in the concentration of the donor state of the B s O 2 defect in non-hydrogenated Cz-Si:B samples after the application of forward-bias-induced minority carrier injection at 423 K indicates that the MCI treatment has not initiated the regeneration reaction but rather the degradation reaction, which is faster in the material with lower resistivity (Figure 1 and 2).
It has been argued in a few studies that the BO-LID regeneration consists of two processes. [14,16] Hydrogen plays a key role in one of them, while the second one occurs without hydrogen. [14,16] Our results on the effects of regeneration-like treatments on the concentration of donor state of the B s O 2 defect in n þ -p diodes show that no permanent deactivation of the B s O 2 defect can happen in the non-hydrogenated samples. These findings are not consistent with the suggested non-hydrogen related regeneration mechanism of BO-LID, [14,16] and support the arguments that the presence of hydrogen in the bulk is crucial for occurrence of the regeneration process. [12,13] The goals of the RBA treatments applied in our study to the hydrogenated B-doped Cz-Si samples were the following: 1) to transform the B s O 2 defects into the shallow acceptor state; 2) to release H atoms from the B s -H complexes which were formed upon hydrogenation; and 3) to initiate the interaction of H þ atoms with the B s O À 2 defect. In the depleted regions of Schottky diodes created by the application of reverse bias, the Fermi level position is at midgap, so the B s O 2 complex, having the E(À/þ) occupancy level at about E v þ 0.31 eV, [3][4][5] is in the negatively charged state, B s O À 2 . Strong Coulombic force-induced attraction occurs between the positively charged hydrogen atoms H þ and the negatively charged B s O À 2 complex if hydrogen is available. We have not observed any newly formed electrically active defects in the DLTS spectra recorded after the RBA treatments. Thus, RBA initiates the interaction of H þ with the B s O À 2 complex and results in the formation of B s O 2 -H complexes that explains the significant drop in the concentration of deep donor state of the B s O 2 defect or even complete passivation of the defect if more hydrogen is incorporated. Experimental and ab initio modeling results above clearly support a picture according to which regeneration involves the passivation of the B s O 2 complexes (either precursor or recombination active forms A or A', respectively), via the reaction with H released from BH pairs upon heat, illumination, and/or injection treatments.
In the boron-doped material, the complete passivation of the B s O 2 defect by hydrogen means, in the context of LID, that the BO defect precursor (annealed state of the BO-LID defect) is no longer available to be converted into the shallow acceptor state which is recombination active upon minority carrier injection (the regenerated state has been formed upon RBA).
Similar interactions occur between hydrogen and the Al s O 2 and Ga s O 2 complexes, but the observed passivation effects were only partial. It seems that even though hydrogen is available in the bulk, the hydrogen interactions with single acceptor atoms A s (Ga s and Al s ) are more efficient than the interactions with the A s O 2 complexes. This can be attributed to different structural configurations of the Al s O 2 and Ga s O 2 complexes compared to those for B s O 2 , [5] which could lead to weaker binding energies to H, and favor the interactions of hydrogen with single Al and Ga acceptor atoms unlike the case of boron.
If the concentration of hydrogen introduced into the bulk region is high, such as that introduced via hydrogen plasma compared to that introduced via wet chemical etching, the passivation effect of the B s O 2 complex is more pronounced. We have observed nearly complete disappearance of the B s O 2 related DLTS signal after RBA in the samples hydrogenated with remote H plasma. This is in agreement with the results reported in the literature about more powerful regeneration or accelerated passivation of the BO-LID defect when the concentration of hydrogen in the bulk regions is higher, [11,13,17,27] even though the method of hydrogen incorporation is different. A clear correlation exists between the presence of free hydrogen at a certain region in the sample and the decrease in the concentration of the B s O 2 complex. Therefore, our results support the suggestion that hydrogen is directly involved in the regeneration mechanism of the BO-LID defect, which can be identified as the H-induced passivation of the B s O 2 complexes via the formation of stable B s O 2 -H complexes.
The hydrogen atoms must be free for any interactions with the A s O 2 complexes to occur. This means that the A s -H pairs, which act as the source of hydrogen, should be separated. The binding energies of hydrogen to the acceptor atoms are different for different acceptor atoms, [28] and the temperature needed to overcome these energies varies between 420 and 440 K for B-, Ga-, and Al-doped Si. According to our annealing experiments, the electrically neutral B s O 2 -H complexes are stable at 443 K, whereas the BH pairs formed in the bulk after RBA treatments dissociate at this temperature. This corroborates the ab initio modeling results, which predict a higher binding energy of the B s O 2 -H complexes than the analogous quantity for the BH pairs. In contrast, the other A s O 2 -H complexes (with A s being either Ga or Al) seem to have binding energies comparable to those for the respective A s -H pairs. This can explain the less pronounced decrease in the concentration of A s O 2 defects after the RBA treatments where the dissociation rates of the A s -H pairs and A s O 2 -H complexes determines the effectiveness of the passivation of the A s O 2 defects.

Conclusion
In this article, we have attempted to understand the details of hydrogen interactions with the A s O 2 (A ¼ B, Al, Ga) complexes in p-type Cz-Si materials and, importantly, to elucidate how the interaction between hydrogen atoms and the B s O 2 defect suppresses its recombination activity by the formation of the regenerated state. The investigation of changes in the B s O 2 defect concentration upon regeneration-like treatments in n þ -p hydrogen-free diodes has shown that regeneration is not possible without the presence of hydrogen in the active region in agreement with results from previous studies. We found that without hydrogen, the application of the regeneration treatment (injection of minority carriers at slightly elevated temperature) results in the transformation of the defect into the degraded state rather than the regenerated state which then is converted back to the unstable annealed state upon dark annealing for few minutes at 473 K. exceeds that of H þ to B À s by at least 0.1 eV. The comparison between hydrogen introduction methods, i.e., wet chemical etching versus hydrogen plasma treatment, has revealed that the passivation of the B s O 2 defects is more effective when more hydrogen is introduced into the bulk region. The investigation of spatial profiles of the B s O 2 defects has shown a clear correlation between the presence of hydrogen at a certain region and the passivation effect where complete passivation of the B s O À 2 defects has been achieved in the region which contains the maximum concentration of hydrogen atoms.
Evidence of similar interactions between hydrogen and other acceptor-dioxygen complexes has been obtained from DLTS spectra recorded on samples from Cz-Si:Ga and Cz-Si:Al materials. However, the observed hydrogen-induced passivation of the Al s O 2 and Ga s O 2 defects is not as effective as for the Cz-Si:B case. A possible explanation of the different behavior after RBA treatments is that the B s O 2 -H complexes are more stable than BH pairs whereas other A s O 2 -H complexes have binding energies close to that of A s -H pairs. Further ab initio calculations and experiments are needed to elucidate the difference in the interactions of hydrogen with different acceptor atoms and acceptor-dioxygen complexes.

Experimental Section
N þ -p and SBDs have been used to study the interactions of hydrogen atoms with A s O 2 defects in Czochralski-grown silicon materials doped with either boron, aluminum, or gallium.
The n þ region of the n þ -p diodes was formed by in-diffusion of phosphorous atoms from PCl 3 gas at high temperature into 3 and 10 Ω cm borondoped Cz-Si wafers. The interstitial oxygen concentration [O i ] in the 3 and 10 Ω cm diodes was about (7.5 AE 1) Â 10 17 and (9.5 AE 1) Â 10 17 cm À3 , respectively, as estimated from measurements of the capture rate of interstitial carbon atoms by the O i atoms in the diodes, which were irradiated with alpha particles at 260 K. [29,30] Some of the diodes were annealed at 400°C for 10-15 h to increase the concentration of the BO defect. It is found that the concentration of hydrogen introduced into the base regions of the n þ -p diodes during fabrication is negligible.
For SBDs, plasma sputtering was used to deposit titanium/aluminum stack via a shadow mask on samples from p-type Cz-Si doped with either B, Al, or Ga, and gold was deposited via thermal evaporation on the back surface to act as an Ohmic contact. Samples were cleaned with trichloroethylene, acetone, and methanol and subjected to 1 min dipping in diluted (10%) HF before deposition of the metals. The diode area was 0.79 mm 2 with a leakage current of the order of %10 À7 A at 16 V reverse bias voltage.
Hydrogen was introduced into the samples before fabrication of the Schottky diodes either via wet chemical etching in a mixture of HF:HNO 3 (1:7) or via treatments in remote RF hydrogen plasma at room temperature for 60 min. The H plasma treatments were carried out with 50 W RF power, 1-2 mbar chamber pressure, and 200-250 cc min À1 hydrogen gas flow.
Current-voltage (I-V ) and capacitance-voltage (C-V ) measurements were carried out to assess the quality of the diodes and determine the concentrations of uncompensated shallow acceptor atoms, the depth of the depletion region, and the bias voltage and pulse voltage used to probe the bulk and subsurface regions in electrical measurements. Conventional DLTS and high-resolution Laplace DLTS techniques were used to detect and characterize deep-level traps in the diodes. [31] Measurements were carried out before and after RBA treatments in the temperature range 398-448 K as well as after dark annealing at 443 K which was done in the furnace. The effect of these treatments on the concentration profiles of uncompensated shallow acceptor atoms has been monitored using C-V measurements while junction spectroscopy techniques have been used to monitor the changes in the concentration of deep-level defects.
The interactions of hydrogen with BO 2 complexes in silicon were further investigated using density functional theory within the projectoraugmented wave/plane wave formalism [32] as implemented by the Vienna Ab initio Simulation Package (VASP). [33,34] Total energies were obtained using the electronic exchange-correlation functional of Heyd, Scuseria, and Ernzerhof (often referred to as HSE06), [35] which mixes a portion of exact Fock exchange for short-ranged interactions. This leads to a calculated bandgap of E g ¼ 1.1 eV, thus avoiding the severe underestimation of this quantity when using cheaper local and semilocal functionals.
The valence states were described by plane waves with a cutoff energy E cut ¼ 400 eV. Boron, oxygen, and hydrogen impurities were inserted in supercells with cubic shape and 512 Si atoms (with calculated lattice constant a ¼ 5.4318 Å). For such large cells, we found that a Γ-point sampling of the Brillouin zone resulted in sufficiently accurate energy differences, including forces. The latter were evaluated within the generalized gradient approximation (GGA) to the exchange correlation potential, [36] from which minimum energy structures were found with a maximum residual force of 0.01 eV Å À1 . Total energies of defects with nonzero charge included a periodic charge correction. [37]