Silver‐ and Silver–Hydrogen‐Related Defects in Silicon

Herein, the electrical and structural properties of Ag‐ and AgH‐related defects in n‐ and p‐type Si are reinvestigated using conventional deep‐level transient spectroscopy (DLTS) and high‐resolution Laplace DLTS. It is evidenced that a peak corresponding to substitutional Ag does not always appear in as‐grown Si and additional heat treatments are necessary to observe this defect. Several AgH‐related peaks, which were not previously reported in the literature, are observed in hydrogenated n‐ and p‐type Si. The electrical properties of these defects are determined and their origin is discussed. By using high‐resolution Laplace DLTS, the depth profiles of AgH‐related defects previously assigned to AgSH and AgSH2 are analyzed, and their assignments are questioned.


Introduction
The study of the electrical properties of transition metals (TM) in Si and their interaction with hydrogen still attracts significant interest since these impurities could significantly decrease the lifetime of minority carriers. In modern microelectronics silicon wafers can be often contaminated with TM in various technological steps. The electrical and optical properties of silver-related defects in crystalline Si have been extensively studied in the literature. [1][2][3][4][5][6][7][8][9][10][11][12][13][14] Similar to gold substitutional silver should also introduce two deep levels in the band gap of Si: the single acceptor state at around 0.5-0.59 eV below the conduction band [2,3,7,8,10,12] and the single donor state at about 0.34-0.41 eV above the valence band . [2,3,7,8,10,12] Until now the origin of the scatter of the activation energy of Ag S observed in different studies is not understood. In addition, some shallower Ag-related levels were also reported using Hall effect measurements or by the metal oxide semiconductor technique. [1,4,5] In hydrogenated samples, H interacts with Ag forming various AgH-related complexes (E2, E3, E5, H2, E6, and H3). Their electrical properties were investigated in nand p-type Si in Yarykin et al. [12] By analyzing depth profiles of the defects E3, E2, and H2 were attributed to different charge states of AgH, whereas E6 and H3 were assigned to different charge states of AgH 2 . One should note that E6 has never been explicitly observed by deep-level transient spectroscopy (DLTS), and its existence was deduced from the comparison of the depth profiles of E2 and E3, which were not identical at depths close to the surface of n-type Si. Yarykin et al. [12] proposed that E6 could have similar electrical properties to those of E2 and, therefore, these defects could not be properly resolved by the conventional DLTS technique. [12] This hypothesis was used as an explanation for the different depth profiles of E2 and E3.
In the present study, we reinvestigate the electrical properties of Ag and AgH-related defects in as-grown samples after wet chemical etching and subsequent annealing at 500 C by using the high-resolution Laplace DLTS technique. [15] 2. Results Figure 1 shows DLTS spectra recorded in three different samples (#1, #2, and #3) after WCE and after annealing at 500 C. All spectra were recorded close to the surface with a reverse bias of -4 V and a filling pulse of 0 V. Four peaks, labelled E35 AgH , E44 AgH , CH, and E100 AgH , were observed in sample #1 after WCE. By comparing the electrical properties of the peaks with those known from the literature [16] we assign CH to the single acceptor state of the carbon-hydrogen complex. E35 AgH , E44 AgH , and E100 AgH have not been previously reported in the literature. In addition to these levels we did not observe any other DLTS peaks deeper in the bulk of Si by measuring with a reverse bias of -8 V and a filling pulse of À4 V. The Ag related peaks E35 AgH , E44 AgH , and E100 AgH disappear after the annealing at 500 C and another defect labelled Ag(-/0) becomes dominant in the DLTS spectrum. Similar DLTS peaks as Ag (-/0) have been already reported in the previous studies. [2,3,7,8,10,12] and we attribute this level to the single acceptor state of substitutional Ag. In addition to this dominant defect, several small DLTS peaks E108 Ag and E148 Ag can be distinguished at lower temperatures in the annealed sample. These defects were observed close to the surface as well as deeper in the bulk. The electrical and annealing properties of E148 Ag are very similar to that of Ag* observed in previous studies, [10,12] whereas E108 Ag has not been reported in the literature so far.
DLTS spectra in sample #2 differ significantly from those observed in sample #1. E54 AgH and E250 AgH are dominant DLTS peaks in this sample after WCE. The electrical properties of the defects corresponding to these peaks are similar to those labelled E2 and E3 in a previous study. [12] In addition to E54 AgH and E250 AgH , we also observed CH and E158 Ag in this sample. By measuring deeper in the bulk of the sample, Ag (À/0) becomes dominant whereas the intensity of E54 AgH , CH, and E250 AgH reduces significantly. After annealing at 500 C three pronounced DLTS peaks E108 Ag , E148 Ag , and Ag (À/0) were observed in sample #2 and they are similar to those observed in sample #1. In addition, a small DLTS peak can be distinguished at around 55 K. The electrical properties of the defect corresponding to this small peak are close to those of thermal donors, which usually appear after heat treatments at such temperatures. [17] Two DLTS peaks H170 Ag and H225 Ag were observed close to the surface in p-type Si (sample #3) after WCE and after annealing at 500 C ( Figure 1c). These peaks were also dominant when measuring deeper in the bulk of these samples. The activation enthalpy of H170 Ag is similar to that of Ag (0/þ) reported in a previous study. [12] However, its apparent capture cross section differs by a factor of 100. In addition, several small DLTS peaks H85 AgH and H150 Ag were observed in sample #3 after WCE whereas H83 Ag and H105 Ag appeared in the annealed sample.
The inset to Figure 1c shows the distribution of the net free carrier concentration obtained from C-V measurements in sample #3 after WCE and after annealing at 500 C. After WCE, the net free carrier concentration reduces towards the surface. We correlate this effect with the passivation of shallow acceptors by hydrogen due to the formation of electrically inactive BH complexes close to the surface. The annealing at 500 C leads to the destruction of the BH pairs and the net free carrier concentration increases close to the surface.
Laplace DLTS measurements were used to investigate the electrical properties of the defects observed in Figure 1. Only single Laplace DLTS peaks were observed for the defects in sample #1 after WCE and after annealing at 500 C. Figure 2 shows the Laplace DLTS spectra measured after WCE of sample #2 ( Figure 2a) and sample #3 (Figure 2b,c). The spectra were recorded with two filling pulses close to the surface and deeper in the bulk at 255 K a) for sample #2 and with different reverse biases at 160 K b) and 225 K c) for sample #3. At 255 K, two distinct Laplace DLTS peaks were resolved and they were attributed to E250 AgH and Ag (À/0) in n-type Si (sample #2). Close to the surface the intensity of E250 AgH is significantly larger in comparison with that of Ag (À/0) whereas Ag (À/0) becomes dominant deeper in the bulk. We did not observe any other peaks at this temperature. The spectra recorded at 160 K in sample #3 also revealed the presence of two different Laplace DLTS peaks, labelled H150 Ag and H170 Ag , respectively. Similarly, two peaks H210 AgH and H225 Ag , which were not resolved in the conventional DLTS spectra were observed at 225 K in sample #3. By varying the reverse bias we did not observe any enhancement of the emission rates of the defects in sample #3. The activation enthalpy and the apparent capture cross sections of the defects were determined from Arrhenius plots derived from the Laplace DLTS measurements and they are combined in Table 1. Figure 3a shows the concentration depth profiles of the defects observed in sample #2 after WCE. The concentration of PH was calculated by subtraction of the net free carrier concentration in the hydrogenated sample from the original free carrier concentration taken from the bulk. The concentrations of E54 AgH and E250 AgH are descending toward the bulk with the same slope as PH. Similarly, to the findings of Yarykin et al., [12] a large discrepancy between these depth profiles was observed close to the surface and this difference becomes smaller in the bulk of Si. The concentration of Ag (À/0) reduces toward the surface and it reaches a constant value of around 1.2 Â 10 13 cm À3 in the bulk  www.advancedsciencenews.com www.pss-a.com of Si. The concentration of Ag (À/0) in the annealed sample was almost a constant at around 3 Â 10 13 cm À3 . Figure 3b shows the depth profiles of the deep level defects observed in p-type Si (compare Figure. 1c). The concentration of BH was also obtained by subtraction of the net free carrier concentration in the hydrogenated sample from the free carrier concentration taken from the bulk of the as-grown sample. The slope of the reduction of the concentration of H85 AgH is similar to that of BH. The depth profile of H225 Ag is also decreasing toward the bulk but with a smaller slope than that of BH. In contrast, the concentration of H150 Ag does not vary with depth. The concentration of H220 AgH was too small to reliably determine the depth profile of this defect. In samples annealed at 500 C the concentration of H105 Ag , H170 Ag , and H225 Ag does not vary with depth. For H170 Ag , it is around 2 Â 10 13 cm À3 . In sample #3, the spectra were recorded with two filling pulses (0 and À1 V) by varying the reverse bias. E35 AgH E C -0.08 4 Â 10 À13 E44 AgH E C -0.09 2 Â 10 À13  www.advancedsciencenews.com www.pss-a.com

Discussion
Different Ag-related defects were observed in n-type and p-type Si depending on the growth conditions and cooling rates of samples (compare Figure 1 and previous studies. [2,3,7,8,12] ) In the asgrown sample #1 we did not observe any levels which could be correlated with Ag S , whereas it was detected in the bulk of sample #2. However, the DLTS spectra recorded in these samples after the annealing at 500 C are similar, they exhibit three DLTS peaks E108 Ag , E148 Ag , and Ag (À/0). The ratio between the intensities of E108 Ag or E148 Ag and Ag (À/0) was significantly larger in the annealed sample #2 in comparison with that in sample #1. We explain this observation by a larger amount of Ag introduced into sample #2 during the growth. Similar to the findings of the previous studies [10,12] we assign E148 Ag to the Ag-Ag pair (A*) whereas E108 Ag can be attributed to unknown Agrelated complex.
In p-type Si we also observed two dominant defects H170 Ag and H225 Ag which appear after the growth and two additional defects H83 Ag and H105 Ag detected after the annealing at 500 C. One should also note that the intensity of H225 Ag decreases in respect to H170 Ag after the annealing at 500 C. The electrical and annealing properties of H170 Ag are very similar to that of Ag (0/þ) reported in Yarykin et al. [12] and they might have originated from the same defect, but in this case, we cannot explain the different apparent capture cross section. We attribute H83 Ag , H105 Ag , and H225 Ag to unknown Ag-related complexes.
Several AgH-related peaks were observed in n-and p-type Si after WCE. E35 AgH , E44 AgH , and E100 AgH were dominant close to the surface in sample #1. None of these defects could be correlated with complexes between H and isolated substitutional Ag since we did not observe any traces of Ag (À/0) in this sample after WCE. Therefore, we tentatively attribute E35 AgH , E44 AgH , and E100 AgH to complexes containing interstitial Ag and hydrogen. The existence of interstitial Ag was also confirmed in n-and p-type Si in Refs. [18,19] By using electron paramagnetic resonance, Son et al. [18] observed three Ag-related defects in p-type Si after the introduction of Ag at 1250 C and subsequent quenching. The authors correlated one of the defects with an isolated interstitial silver atom whereas the two others were attributed to complexes, which contain interstitial Ag. Rollert et al. [19] investigated concentration depth profiles of Ag in dislocation free n-type Si using neutron activation analysis. The diffusivity of Ag was around 7 Â 10 À9 m 2 s À1 at 1473 K in these samples and it corresponds to the diffusion of interstitial Ag rather than Ag S . Two DLTS levels E54 AgH and E250 AgH were observed close to the surface in sample #2 after WCE. As mentioned earlier, the electrical properties of the defects are similar to E3 and E2, which were previously assigned to different charge states of substitutional Ag with one H atom. [12] According to the findings of Feklisova and Yarykin , [20] the concentration of H-related defects should reduce with the same slope if they contain an identical number of H in samples after WCE. We observed that the concentration profile of E54 AgH and E250 AgH has the same slope as the PH and CH profiles and, therefore, we indeed attribute these defects to complexes that comprise a single H atom. However, the depth profiles of E54 AgH and E250 AgH as determined with high-resolution Laplace DLTS are not identical (Figure 3). This contradicts to the attribution of these defects to different charge states of AgH 2 . The origin of these defects is still not clear and further studies are necessary to understand the nature of the defects or to explain the difference in their depth profiles close to the surface.
In p-type Si, we observed only one defect with a descending depth profile (H85 AgH ), which could be correlated with H-related defects. The concentration profile of H85 AgH has the same slope as BH in p-type Si after WCE and therefore H85 AgH should also contain a single H atom. This defect was not observed in the study of Yarykin et al. [12] and we tentatively correlate it with an unknown Ag-H complex. However, one could also assume that interstitial Ag is a part of this defect. The electrical properties of H210 AgH are similar to those of H3 , [12] but we did not find any traces of H2 in our samples although H2 and H3 were always observed together in the study of Yarykin et al. [12]

Conclusion
In the present study, we reinvestigated the electrical properties of Ag-and AgH-related defects in n-and p-type Si by DLTS and Laplace DLTS measurements. Different defects were observed in these samples depending on their growth conditions and heat treatments. Several AgH-related peaks, which were not previously reported in the literature were also observed in hydrogenated n-and p-type Si. We investigated their depth profiles and discussed their origin. By using Laplace DLTS measurements, we questioned the origin of the peaks (E6, E2, and E3) previously attributed to Ag S H and Ag S H 2 in Yarykin et al. [12] 5. Experimental Section N-type and p-type float-zone Si samples with a doping concentration of 2 Â 10 14 cm À3 of P or B were used. Silver was incorporated during the crystal growth ([Au S ] <1 Â 10 14 cm À3 ). Samples were cut from crystals grown with different concentrations of Ag (15 mg into sample #1 and 30 mg into samples #2 and #3). The as-grown samples were polished by wet chemical etching (WCE) in CP4A (HF:HNO 3 :CH 3 COOH 3:5:3) at room temperature for 10 min. Several samples were annealed at 500 C in air. Schottky-contacts were fabricated by resistive evaporation of Au (n-type) or Al (p-type) through a shadow mask. Ohmic contacts were made by rubbing a eutectic InGa-alloy into the backside of the samples. The quality of the Schottky and Ohmic contacts was checked by currentvoltage and capacitance-voltage (C-V ) measurements at room temperature and at 50 K. The C-V curves were recorded at 1 MHz. Conventional DLTS and Laplace DLTS were used to investigate the properties of the defects. The concentration depth profiles were obtained using the double pulse Laplace DLTS technique as described in the previous study. [15] For the calculation of the concentration the temperature dependent λ-layer was taken into account. [21] The electric field was calculated from the C-V measurements as described in Ref. [21]. The levels are labelled using E or H (for electron or hole traps, respectively) followed by the temperature at which the peaks appear in the DLTS spectrum for an emission rate of 50 s À1 .