Analysis of recombination centers near an interface of a metal–SiO₂–Si structure by double carrier pulse deep-level transient spectroscopy

Crystal silicon solar cells with carrier-selective contacts (CSCs) exhibit high conversion efficiencies and reduce production costs.^1–8 In CSC structures, transition metal oxides are adopted as contact materials owing to their wide bandgaps, for high transparency, and appropriate work functions, for carrier selectivity. For example, MoOx is considered for hole contacts owing to its large work function, whereas TiOx is considered for electron contacts owing to its small size.^4,5,7 However, the high resistance of these contact materials prevents solar cells from achieving low series resistance. Therefore, transparent conductive oxides (TCO), such as indium tungsten oxide (IWO) or indium–tin oxide (ITO), should be deposited on the contact materials to decrease the resistance. Reactive plasma deposition (RPD), plasma-enhanced chemical vapor deposition, and sputtering are widely used techniques for forming TCO films.^9–11 However, the minority carrier lifetime decreases owing to TCO deposition, and the cell performance deteriorates.^9–15 This deterioration suggests the generation of recombination center into devices. Using photoluminescence, deep-level transient spectroscopy (DLTS), capacitance voltage (C–V) analysis, and transmission electron microscopy, the electric properties of the defects and formation mechanisms were studied in order to suppress process-induced degradation and increase conversion efficiency.^9,11,16–18

Based on the capacitance voltage, ITO-RPD process induces defects and they decrease minority carrier lifetime to ∼3 µs.^9,11 The density of defects was ∼10¹² cm⁻² eV at the SiO₂/Si interface. Most of the RPD-induced defects were annihilated by 200 °C heat treatment, and the minority carrier lifetime almost returned to the pre-deposition value. However, the recovery by the annealing was not complete. Regarding the defect formation mechanism, it was suggested that UV light with ∼110–180 nm wavelength induces defects.¹⁹ 

As suggested by the DLTS analysis with Bayesian optimization, three types of bulk electron traps were induced by ITO-RPD.²⁰ Because the energy levels of these defects are located at the midgap, they might be recombination centers. However, the recombination characteristics of these defects remain unclear. Double carrier-pulse DLTS (DC-DLTS) is a powerful technique to evaluate the recombination activity of defects.^21–24 Initially, the majority carriers in DC-DLTS are captured by the defects whose energy levels are located under the Fermi level (E_F). Subsequently, the minority carriers are injected by the forward bias voltage applied to the p–n junction, and some majority carriers trapped by the recombination centers are annihilated through recombination with the injected minority carriers. Therefore, the DLTS signal intensity originating from recombination-active defects decreases. However, it is difficult to apply this method to MIS structures because high minority carrier injection is challenging.^25,26

This paper proposes a DC-DLTS method for MIS structures. The inversion layer is formed by applying a pulse voltage and minority carriers are injected into the near-interface region. Subsequently, some majority carriers recombine with the injected minority carriers and the DLTS signal obtained from the recombination defects decreases. During the injection of minority carriers, some of the majority carriers are thermally emitted from the defects and the signal intensity also decreases. The recombination activity was analyzed by considering the effect of thermal emission on the change in signal intensity. This new method was applied to study the RPD-induced defects in the Al/SiO₂/Si structure. The ITO-RPD process induces defects with different energy levels around the mid-gap. These may act as a recombination center.

The electrical properties of defects induced by the ITO-RPD were analyzed by DC-DLTS. For this analysis, Al/(ITO)/SiO₂/Si/Al structured MOS diodes were fabricated. 2–5 Ω cm n-type Czochralski (Cz) Si (100) substrates were used. The single side of wafers was mirror-polished. Wafers were dipped in an H₂O, H₂O₂, and NH₃ mixture (H₂O:H₂O₂:NH₃ = 5:1:1) at 80 °C for cleaning purposes, and the surface native oxide was removed using HF solution. A relatively high concentration of P (sheet resistivity of 60 Ω/sq) layer was formed on the unpolished side of the Si substrate by thermally doping for an ohmic contact formation.

The dry oxidation process resulted in the formation of a 12 nm thick SiO₂ layer on the polished-side of Si wafer under an O₂ atmosphere at 950 °C. An ITO was deposited on the SiO₂ surface during RPD process.^12,14 Argon (Ar) gas was delivered into the plasma gun for generating Ar plasma. The plasma was injected into the growth chamber, and a magnetic field guided it to the ITO target. The plasma evaporated ITO target and an ITO film were grown on the SiO₂ substrate surface. The source material was 5 wt. % Sn-doped ITO target. O₂ was also injected into the growth chamber. Ar and O₂ were supplied with 85 and 20 SCCM flowrate, respectively, and the growth pressure was controlled to 0.3 Pa. The chamber or processed wafers were not intentionally heated; however, the actual sample temperature was ∼80 °C due to the Ar plasma during the ITO-RPD process. The samples were transferred in the growth chamber, and a uniform ITO film with a 90 nm thick was deposited. The front-side ITO of the fabricated ITO/SiO₂/Si/SiO₂ structured samples was removed using HCl solution, and the back-side SiO₂ layers was removed using HF solution. Al electrodes were then deposited on both sides by using a resistance heating evaporation system.

The concept of the proposed DC-DLTS to analyze near-interface recombination centers in the n-MIS structure is schematically shown in Fig. 1. Initially, a positive pulse voltage V_p1 is applied to the sample, the energy level Et is located under the E_F as shown in Fig. 1(a), and all defects capture the majority of carrier electrons. Subsequently, the second pulse voltage V_p2 was applied to the MIS structure for t_mp, and the minority carrier holes were injected to the near-surface, where E_F was below the intrinsic Fermi level Ei at the interface by applying V_p2. The defect level near the interface was above E_F. If the defects were recombination centers, some of them would capture the holes through recombination [Fig. 1(b)], resulting in thermal electron emission, simultaneously. The recombination and thermal emission processes reduce the number of electrons trapped in the defects. Finally, a reverse voltage Vr was applied, and the DC-DLTS signal was obtained. Under this condition, E_F was located near Ei at the interface [Fig. 1(c)], and the number of injected holes was negligible. A few electrons released from the defect may be captured again by the defect. During DC-DLTS measurements, a few electrons emitted from the defects may be captured again by another defect. This phenomenon must be considered in the DLTS analysis. Here, the decay curves were theoretically obtained using the estimated energy levels and capture cross sections of RPD-induced defects, and by comparing the theoretical and experimental results, the recapture effect on the DC-DLTS signal was observed.

The recombination properties of the defects were studied using the decay curves as a function of the second pulse duration t_mp. The DC-DLTS signal is expressed in terms of the duration of the inversion state as follows:

where tw is the period width, T is temperature, t₀ is the start time of the measurement, and τ(T) is the time constant of carrier emission from defects at T. For t_mp = 0 s, the capacitance decreased, as indicated in Fig. 2. The value of A denotes the transient intensity, and these values are used to evaluate the recombination properties. When there is no recombination during the injection of minority carriers, the DC-DLTS signal is the same as that with t_mp = 0 s. When recombination occurs, the capacitance decreases more swiftly at t_mp and then decreases according to the curve without recombination. Therefore, the value of A was estimated using the transient curves after the t_mp duration. If the defect acts as a recombination center, the value of A decreases with increasing t_mp.

DC-DLTS spectra were composed of several signals originating from different types of defects. First, the number of defect types was determined. Next, the defect properties were predicted using Bayesian optimization. The difference between the predicted and measured spectra was quantified using the evaluation functions.²⁰ This evaluation value was maximized using Bayesian optimization to explain the DC-DLTS spectrum with the predicted defect characteristic values.

The DC-DLTS spectra obtained using different t_mp values are demonstrated in Fig. 3. Here, t_mp was 0, 5, and 10 ms; t₀ = 7.81 ms, V_p1 = 0 V, V_p2 = −2 V, and Vr = −1.3 V. These spectra correspond to electron traps because the n-type Si substrate was used here. This measurement condition allows for the observation of defects that have energy levels around the mid-gap. They exhibit peaks at ∼270 K. The signal intensities decrease with increasing t_mp. Therefore, thermal emission of electrons captured by the RPD-induced defects occurred, and DLTS spectra were changed. However, it is unclear whether the trapped electrons recombine with the injected holes.

To confirm the possibility of recombination, the relationship between the DLTS signal and t_mp was obtained as a function of V_p2, as shown in Fig. 4. Here, V_p2 was −1.3 and −2.0 V. Both the signal intensities decreased as t_mp increased. When V_p2 = −1.3 V, holes were not injected and the signal intensity decay resulted only from the thermal emission of electrons captured by the defects to the conduction band. Conversely, when V_p2 = −2.0 V, the signal intensity decreased more steeply. This implies the recombination of trapped electrons with the injected holes occurred and the difference between the signals corresponded to the recombination process.

The spectra in Fig. 3 are not explained using the single-peak spectra. This implies each spectrum may contain several signals from different defects. Therefore, the spectrum was deconvoluted using Bayesian optimization. The results are demonstrated in Fig. 5. The spectra are composed of three signals, corresponding to three types of defects, and the peak positions of the decomposed signals are constant and independent of the V_p1 values. This result suggests that these defects are located in bulk Si near the interface.²¹ The energy levels and capture cross sections of the defects were analyzed, and the results are summarized in Table I. They have defect levels located in the mid-gap and capture cross sections of ∼10⁻¹⁵ cm². These values indicate that these defects can act as recombination centers.

To evaluate the effects of hole emission and electron recapture on the decay signals, the decay curves of E1, E2, and E3 defects were theoretically obtained using defect properties, such as energy levels and capture cross sections obtained by Bayesian optimization. The obtained curves are shown in Fig. 6. Here, the signal intensity of the decay originating from E3 is too small and is neglected in this figure. The transient curve obtained through experiments was well explained by the sum of the signals corresponding to E1 and E2. This result indicates that the hole emission and electron recapture were negligible when evaluating the A values in Eq. (1).

The dependence of the A values originating from each defect on t_mp is shown in Fig. 7. The A values for E2 and E3 are almost constant and independent of the t_mp values. In contrast, the A value for E1 decreased with increasing t_mp. The A value corresponds to the number of RPD-induced defects. The concentration of E1 defects was the highest. This indicates that the E1 defect is a recombination center and mainly causes a short minority carrier lifetime. Although the formation of defects with similar energy levels to E1 was reported and these structures might be related to vacancy or interstitial Si, the actual structure of RPD induced defects is not clear yet.^27,28

Three types of electron traps were induced using the RPD process to fabricate CSC solar cells. The recombination properties of these defects near the SiO₂/n-Si interface in the CSC structure were evaluated using a new DC-DLTS method. In the proposed method, after the application of a positive pulse voltage for the injection of majority carriers, a negative pulse voltage was applied and minority carriers were injected into the MIS structure. DC-DLTS spectra were composed of three types of bulk defect signals, which were decomposed into three spectra using Bayesian optimization. Owing to the application of the second pulse, the intensities of the signals obtained from these defects decreased. Regarding the effect of thermal emission on the change in signal intensity, the defect with Ec-0.57 eV energy level and ∼10⁻¹⁵ cm² capture cross section was suggested to act as a recombination center.

This work was supported by the Research Center for Smart Energy Technology under the Ministry of Education, Culture, Sports, Science and Technology, Japan. The authors thank Motoo Morimura and Mayuko Ohsugi at Toyota Technological Institute for the help in sample fabrication.

The authors have no conflicts to disclose.

Tomohiko Hara: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Software (equal); Validation (equal); Visualization (equal); Writing – original draft (equal). Yoshio Ohshita: Funding acquisition (equal); Methodology (equal); Supervision (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.