Lifetime degradation of n-type Czochralski silicon after hydrogenation

Hydrogen plays an important role in the passivation of interface states in silicon-based metal-oxide semiconductor technologies and passivation of surface and interface states in solar silicon. We have shown recently [Vaqueiro-Contreras et al., Phys. Status Solidi RRL 11, 1700133 (2017)] that hydrogenation of n-type silicon slices containing relatively large concentrations of carbon and oxygen impurity atoms {[Cs] ≥ 1 × 1016 cm−3 and [Oi] ≥ 1017 cm−3} can produce a family of C-O-H defects, which act as powerful recombination centres reducing the minority carrier lifetime. In this work, evidence of the silicon's lifetime deterioration after hydrogen injection from SiNx coating, which is widely used in solar cell manufacturing, has been obtained from microwave photoconductance decay measurements. We have characterised the hydrogenation induced deep level defects in n-type Czochralski-grown Si samples through a series of deep level transient spectroscopy (DLTS), minority carrier transient spectroscopy (MCTS), and high-resolution Laplace DLTS/MCTS measurements. It has been found that along with the hydrogen-related hole traps, H1 and H2, in the lower half of the gap reported by us previously, hydrogenation gives rise to two electron traps, E1 and E2, in the upper half of the gap. The activation energies for electron emission from the E1 and E2 trap levels have been determined as 0.12, and 0.14 eV, respectively. We argue that the E1/H1 and E2/H2 pairs of electron/hole traps are related to two energy levels of two complexes, each incorporating carbon, oxygen, and hydrogen atoms. Our results show that the detrimental effect of the C-O-H defects on the minority carrier lifetime in n-type Si:O + C materials can be very significant, and the carbon concentration in Czochralski-grown silicon is a key parameter in the formation of the recombination centers.Hydrogen plays an important role in the passivation of interface states in silicon-based metal-oxide semiconductor technologies and passivation of surface and interface states in solar silicon. We have shown recently [Vaqueiro-Contreras et al., Phys. Status Solidi RRL 11, 1700133 (2017)] that hydrogenation of n-type silicon slices containing relatively large concentrations of carbon and oxygen impurity atoms {[Cs] ≥ 1 × 1016 cm−3 and [Oi] ≥ 1017 cm−3} can produce a family of C-O-H defects, which act as powerful recombination centres reducing the minority carrier lifetime. In this work, evidence of the silicon's lifetime deterioration after hydrogen injection from SiNx coating, which is widely used in solar cell manufacturing, has been obtained from microwave photoconductance decay measurements. We have characterised the hydrogenation induced deep level defects in n-type Czochralski-grown Si ...


I. INTRODUCTION
Hydrogen has been used for over 50 years in the semiconductor industry to passivate surface and interface states by satisfying dangling bonds. This was the important breakthrough in the 1970s, 1 which enabled metal-oxide-semiconductor (MOS) and complementary MOS (CMOS) signal processing devices to be made reproducibly and is still an important part in the processing of modern microelectronic devices ranging from semiconductor memory to processor chips. 2 Hydrogen passivation of interface states and recombination centers in multicrystalline and crystalline silicon materials has also been successfully used in solar cell processing [3][4][5][6] and has become a part of solar silicon research and technologies. 7,8 Such passivation results in fewer carriers being lost at the surface of the silicon solar cell, at the interface with the anti-reflection dielectric layers, and in the solar material bulk thereby contributing to increased cell efficiencies. Achieving higher efficiencies of solar cells has become the prime target for the solar cell industry as discussed in a recent review paper by Green. 9 However, somewhat surprisingly we have found recently that hydrogen atoms can react with carbon-oxygen-related defects in silicon to produce powerful recombination centers, which reduce the minority carrier lifetime in n-type silicon containing O impurity atoms and higher than usual concentrations of C, thereby reducing the efficiency of any solar cells made from such material. 10,11 We have reported the observation of two very powerful recombination centers, H 1 and H 2 , formed by carbon-oxygen and hydrogen (C-O-H) interactions in a range of n-type Czochralski (Cz) Si materials grown by conventional and continuous Czochralski (C-Cz) techniques. 11 These types of materials are gaining ground in the solar industry for high efficiency products because of a number of beneficial properties compared to p-type silicon. They have, particularly, a higher "as grown" lifetime, lesser sensitivity to metallic contaminations, and negligible light induced degradation. 12,13 However, the oxygen concentration is high (!5 Â 10 17 cm À3 ) in Czochralski-grown silicon, and some Czand C-Cz-Si materials grown for solar applications contain rather high (>2 Â 10 16 cm À3 ) concentration of carbon atoms. As hydrogen is frequently encountered in the silicon solar materials either by deliberate introduction or as a result of wafer processing, the likelihood of the formation of the C-O-H complexes is quite significant in n-type Cz-Si and C-Cz-Si, thereby making the solar cells from these materials more susceptible to an efficiency decrease if the C-O-H reaction can progress and the complex retained in the finished product. It has been found that like many other H-related defects in silicon, the H 1 /H 2 recombination centers anneal out at quite low temperatures, and they have been shown to be eliminated after a 30-min heat-treatment at 175 C. 10 However, as we have suggested in Refs. 10 and 11, when enough hydrogen is present in hydrogenated solar cells, the C-O-H complexes can reform upon cooling the cells after heat treatments at T ! 200 C and thereby reduce the cell efficiency.
In the present work, we show the results of a series of microwave photoconductance decay (lPCD) measurements, which validate the earlier suggestion and demonstrate lifetime deterioration in n-type Cz-Si samples after hydrogen injection from SiN x coating, which is widely used in solar cell manufacturing, and after the introduction of hydrogen by other methods. Further, it is shown that along with the hydrogenrelated H 1 and H 2 hole traps in the lower half of the gap, hydrogenation of n-type Cz-Si samples give rise to two electron traps (E 1 and E 2 ) in the upper half of the gap. These electron traps have been characterized and their origin is discussed.

II. EXPERIMENTAL DETAILS
We have carried out measurements of minority carrier lifetime and characterization of deep level defects on samples from a wide range of n-type phosphorus doped Cz-and C-Cz-Si wafers with resistivity between 1 and 8 X cm. The concentration of interstitial oxygen atoms, [O i ], was in the range from 12 to 20 ppma (1 ppma ¼ 5 Â 10 16 cm À3 ), while the concentration of substitutional carbon atoms, [C s ], varied from 0.01 to 9 ppma in the samples studied. Also a few float zone (FZ) grown silicon samples with [O i ] < 0.2 ppma have been studied for a comparison. The samples were cut and RCA cleaned, and hydrogen was then introduced into the silicon by three different techniques: (1) wet etching with either KOH or HF/HNO 3 solution, (2) 50 W remote H plasma exposure for 30 min at temperatures between 25 and 250 C, and (3) in-diffusion of hydrogen from H-rich silicon nitride (SiN x :H) films. SiN x :H films were deposited by the Plasma-Enhanced Chemical Vapour Deposition (PECVD) technique, in a PlasmaPro 100 PECVD system, using Oxford Instruments. The deposition parameters were: 20:20 sccm SiH 4 :NH 3 gas flow, 300 mTorr chamber pressure, 400 C table temperature, 13.56 MHz plasma generated frequency, and 50 W of power. These parameters produced a 90 nm SiN x :H film after 2 min of plasma exposure, which was then etched off in an Oxford Instruments System 100 180 RIE system before measurements of minority carrier lifetime. For etching the SiN x :H films, the samples were exposed for 50 s to a 6 W RF plasma plus a 600 W inductively coupled plasma (ICP) of O 2 :CHF 3 :CF 4 (2:10:23 sccm) gas, at 13.56 MHz. The annealing of the samples, aimed at the dissociation of the hydrogen-related recombination active defects, has been carried at 200 C in an oven in air ambient.
The microwave photoconductivity decay (lPCD) lifetime mapping was carried out with the use of a Semilab WT-2000 PVN machine, which has 905 nm LASER excitation and a microwave source operating at $10 GHz. The lifetime maps thus represent an averaged lifetime in the sample at a depth of a few tens of microns from the surface. For suppression of surface recombination, the maps were recorded with both surfaces of the samples covered with iodine/ethanol solution. 14 For deep level transient spectroscopy (DLTS) and minority carrier transient spectroscopy (MCTS) measurements, 15 the samples were cleaned and subjected to evaporation of Au through a shadow mask from the front side and to Al evaporation without mask on the back side, to produce Schottky barrier diodes and an Ohmic contact, respectively. An un-metalized region of the Ohmic contact was left on the back of the MCTS samples to allow an optical excitation from a 940 nm light emitting diode (LED). Each diode was characterized first with the current-voltage and capacitancevoltage measurements, and subsequently DLTS, MCTS, and high resolution Laplace DLTS/MCTS 16 measurements were carried out for the detection and characterization of electrically active defects.

A. Effect of hydrogenation treatments on minority carrier lifetime in n-type Cz-Si
We have carried out a series of lPCD measurements to observe the reduction of the minority carrier lifetime induced by hydrogen introduction with different methods into an n-Si sample with [O i ] ¼ 7.5 Â 10 17 cm À3 , [C s ] ¼ 4.5 Â 10 17 cm À3 , and 2 mm thickness. Figure 1 shows a set of minority carrier lifetime maps recorded on the Si:O þ C sample, which was subjected to the following subsequent hydrogenation treatments: (a) wet etching in 1HF þ 7HNO 3 solution for 3 min at room temperature (RT), (b) 50 W remote H plasma exposure for 30 min at RT, (c) wet etching in 20% KOH solution for 10 min at 60 C. Figure 1 also shows a lifetime map recorded on the other side of the same sample after in-diffusion of hydrogen from H-rich silicon nitride (SiN x :H) films (the deposition temperature was 400 C). All the maps were recorded immediately after each hydrogenation treatment and then after annealing at 200 C for 20 min to destroy the C-O-H complexes.
An inspection of the lifetime maps in Fig. 1 shows that each hydrogenation step resulted in the lifetime degradation of the material and its partial recovery after the heat treatments at 200 C. Further, some decrease in the averaged lifetime (s av ) occurred in the course of the subsequent hydrogenationannealing steps from the first cycle with HF/HNO 3 etching (s av ¼ 273 ls) to the last one with KOH etching (s av ¼ 111 ls). Such a reduction can be associated with the accumulated introduction of hydrogen into the sample after each hydrogenation step, thereby making the thermally induced recovery of lifetime more difficult after each of the subsequent hydrogenationannealing cycles. The depth of the hydrogen penetration into the sample upon the SiN x :H film deposition procedure is greater than in any other hydrogenation treatments used in this study because of the higher processing temperature, so it is likely that the recombination active C-O-H complexes have been re-formed in the SiN-processed sample during cooling down after the 200 C annealing. The re-formed C-O-H complexes are responsible for the low s av value, 117 ls, after the recovery annealing treatment. The results of lifetime mapping presented in Fig. 1  In our previous paper, 10 the DLTS and MCTS spectra for one of the hydrogenated n-type Si:C þ O samples have been shown. In the DLTS spectra, four peaks due to electron traps labelled as E 1 to E 4 have been detected, and in the MCTS spectra, four negative peaks due to hole traps labelled as H 1 to H 4 have occurred. These peaks can be seen in Figs. 2, 4, and 5 in the present paper. In Ref. 11, we have characterized and identified the dominant hole traps, H 1 and H 2 , in hydrogenated n-type Cz-Si samples. From a series of Laplace MCTS measurements, the positions of energy levels of the H 1 and H 2 traps have been determined as E v þ 0.38 eV and E v þ 0.36 eV, respectively, and the ratios of the capture cross sections of minority (holes) and majority (electrons) carriers as 50 and 40, respectively. 11 The measured parameters indicate that both the H 1 and H 2 traps are related to the acceptor levels of two defects, which are powerful recombination centers for holes in n-type Si crystals and are responsible for the lifetime degradation in the hydrogenated n-type Cz-Si crystals. We have demonstrated the involvement of carbon, oxygen, and hydrogen atoms in the formation of centers responsible for the H 1 and H 2 traps, and with the use of low temperature infrared absorption measurements, it has been found that a complex consisting of a substitutional carbon and an interstitial oxygen atom (C-O) is a pre-cursor of these defects. 11 The possible structures of the C-O-H complex with electronic properties close to those for the H 1 and H 2 traps have been found from the first principles calculations. 11,17 Here, we present results of a study of the electronic properties and thermal stability of the E 1 , E 2, and E 4 electron traps and some results on the relation of these traps to the most abundant oxygen, carbon, and hydrogen impurities. It should be noted, however, that the E 1 -E 4 traps do not appear to be the significant recombination centers for holes in n-type Si crystals because of the closeness of their energy levels to the conduction band edge. Figure 2 shows the development of the DLTS and MCTS spectra upon 30-min isochronal annealing of a hydrogenated n-type Cz-Si sample in the temperature range from 50 C to 200 C. It can be seen that the annealing behaviour of the E 1 and E 2 electron traps is similar to that of the H 1 /H 2 hole traps. Both pairs of traps disappear upon heattreatments in the temperature range of 150-175 C. In contrast, the magnitude of the peak related to the E 4 trap has increased after the 150 C and 175 C annealing steps. The E 4 trap has been assigned in our previous work (Ref. 10) to the second donor level of the oxygen-related thermal double donors (TDDs). 18 Even though hydrogen can passivate TDDs, the binding energy of H atoms to the TDDs centers is not high, and the electrical activity of TDDs is restored after heat-treatments in the temperature range of 150-175 C. 19 We have roughly estimated the activation energy for the E 4 trap recovery from the isochronal annealing data in Fig. 2 and obtained the value of 1.65 6 0.05 eV, which is close to that found in Ref. 19 for the elimination of the TDD-H complexes. So, our results on the annealing behavior of the E 4 trap validate the assignment of this trap to TDDs and are consistent with the previous DLTS results on TDDs and their passivation with hydrogen. 18,19 The similarity in the elimination behaviour of the E 1 /H 1 and E 2 /H 2 trap pairs (Fig. 2) suggests that these pairs can be related to two energy levels of two defects. Later on, we present some results, which confirm this suggestion. Figure 3 shows the concentration profiles of the H 2 and E 2 traps in a Cz-Si sample, which was subjected to etching in 1HF þ 7HNO 3 solution for 3 min at room temperature. The profiles have been obtained with the use of Laplace DLTS/MCTS techniques; however, the measurement conditions for the E 2 and H 2 traps were different. The concentration profile for the E 2 trap has been obtained using L-DLTS by changing both bias and pulse voltages (U b and U p , respectively) and a fixed difference between U b and U p . So, the concentration values for the E 2 trap have been determined in nearly equidistant relatively narrow probing areas. 20,21 In MCTS measurements, only the reverse bias can be changed, and the probing area for minority carrier traps extends from the junction (W 1 ¼ 0) to the depletion depth under reverse bias (W 2 ¼ W b ). So, the concentration values for the H 2 trap in Fig.  3 are the mean concentration values for the relatively wide probing areas (e.g. the deepest point for the H 2 trap concentration in Fig. 3 corresponds to the mean concentration in a region between 0 and 2.85 lm). Taking into account the earlier and other possible sources of errors in the determination of minority carrier trap concentration using MCTS measurements, 15,22 it appears that the profiles presented in Fig. 3 are almost certainly related to the same defect. The profiles are characteristics of the defects induced by hydrogen in-diffusion during wet chemical etching of n-type Si samples. 10,[23][24][25] The E 1 and E 2 electron traps have been observed in the DLTS spectra of hydrogenated Cz-Si samples but not in the spectra of hydrogenated FZ-Si samples. This observation indicates an involvement of oxygen impurity atoms into the corresponding defect structures. We have further investigated the possible links of the E 1 and E 2 electron traps to carbon and oxygen impurities. In Fig. 4, the DLTS and MCTS spectra are compared for four hydrogenated as-grown n-type Cz samples with different C s and nearly equal O i concentrations. In order to take into account the differences in resistivity of the samples studied, we have plotted the value of 2 Â N d Â DC/C b on the vertical axis of Fig. 4, where DC is the magnitude of a capacitance change related to carrier emission from traps, N d is the concentration of doping atoms, and C b is the bias capacitance.  can be associated with two energy levels of two complexes each incorporating one oxygen, one carbon, and one hydrogen atom. We have argued in our previous study (Ref. 11) that the H 1 and H 2 hole traps are related to the first acceptor levels of two C-O-H complexes. So, if the E 1 /H 1 and E 2 /H 2 pairs are related to two energy levels of two C-O-H defects, it is reasonable to associate the E 1 and E 2 traps with the second acceptor levels of the defects.
We have carried out Laplace DLTS measurements of electron emission rates (e n ) and capture cross sections (r n ) for the E 1 and E 2 traps to determine their electronic characteristics. From the Arrhenius plots of T 2 -corrected emission rates, we have derived the activation energies for the electron emission from the defect levels to the conduction band and pre-exponential factors as follows: 0.12 6 0.005 eV and 3.3 Â 10 7 s À1 K À2 ; 0.14 6 0.005 eV and 2.1 Â 10 7 s À1 K À2 for the E 1 , and E 2 traps, respectively. From direct measurements of the majority carrier capture cross sections, we found the following r n values for the E 1 and E 2 traps 5.7 Â 10 À16 cm 2 and 7.35 Â 10 À16 cm 2 , respectively. No significant temperature dependences of r n have been detected for both traps.
The E 1 and E 2 electron traps have been observed and characterized in previous works by Yoneta et al. 24 and by Gwozdz et al. 25 The results presented in Refs. 24 and 25 are, in general, consistent with our results on the E 1 and E 2 electron traps. There are, however, some discrepancies related to the identification of the charge state of the E 1 and E 2 traps. Yoneta et al. 24 have said that "the results on depth profiles of these traps and carriers suggested the donor character of the traps, but they hardly exhibited the Poole-Frenkel effect." On the other hand, Gwozdz et al. on the basis of their analysis of dependencies of electron emission rates from the E 1 and E 2 traps (labelled as E65 and E75 traps in Ref. 25) on the electric field strength have related these traps to single acceptor levels. 25 It should be noted that the values of the directly measured electron capture cross sections for the E 1 and E 2 traps given in Ref. 25 (r n ¼ 2 Â 10 À18 cm 2 for both the traps) indicate the electron capture by a negatively charged defect (and so the second acceptor level) rather than a neutral one.
To obtain some further data to resolve the question regarding the charge state of the E 1 and E 2 traps, we have studied the dependencies of the electron emission rates from these traps on the electric field strength with the use of Laplace double DLTS technique. 16 The obtained dependencies are shown in Fig. 6. It appears that the dependencies can be described well by a e n / exp(E 2 ) law, thereby indicating the phonon-assisted tunnelling as the mechanism for the electric field enhancement of electron emission. 29,30 The phonon-assisted tunnelling has been shown to be the mechanism of the electric field enhancement of carrier emission for a number of defect levels related to the double to single charge state transitions in Si and Ge. [31][32][33] So, the results obtained support our assignment of the E 1 and E 2 electron traps to the second acceptor levels of two C-O-H complexes.
The modelling studies are needed to identify configurations and explain the electronic structures of the two C-O-H complexes responsible for the E 1 /H 1 and E 2 /H 2 electron/hole traps responsible for the lifetime degradation in hydrogenated n-type Si crystals containing a relatively large carbon concentration.

IV. CONCLUSIONS
A number of minority and majority carrier traps are observed in phosphorous doped silicon containing oxygen and carbon impurity atoms after the introduction of hydrogen. Some of the traps act as powerful recombination centres, which severely decrease minority carrier lifetime in the Si:O þ C materials. We demonstrate such an effect by lPCD mapping of the effective lifetime in a carbon and oxygen rich Cz-Si sample, which was subjected to hydrogenation by four different techniques. The dominant recombination active defects are identified as two complexes each incorporating carbon, oxygen, and hydrogen atoms. Our results indicate that even though the defects are disassociated by heat-treatments at relatively low temperatures ($200 C), they are likely to reform upon cooling the samples down when enough hydrogen is retained in the silicon. Particularly, it is shown that such a scenario can be realized in the case of Cz-Si-based solar cells covered with SiN x :H dielectric films, which are widely used in solar cell industry for anti-reflection and surface passivation purposes. The rise in C s concentrations in recent commercial Cz and C-Cz solar silicon increases the probability of the formation of the C-O-H complexes and the associated decrease in efficiency of the solar cell produced from such materials. So, care should be taken for controlling the carbon concentration in solar grade silicon as it is one of the key parameters for the formation of the detrimental C-O-H traps.