Correlation between Boron–Silicon Bonding Coordination, Oxygen Complexes and Electrical Properties for n-Type c-Si Solar Cell Applications

Abstract

In this paper, the relationship between coordination complexes and electrical properties according to the bonding structure of boron and silicon was analyzed to optimize the p–n junction quality for high-efficiency n-type crystalline solar cells. The p⁺ emitter layer was formed using boron tribromide (BBr₃). The etch-back process was carried out with HF-HNO₃-CH₃COOH solution to vary the sheet resistance (R_sheet). The correlation between boron–silicon bonding in coordination complexes and electrical properties according to the R_sheet was analyzed. Changes in the boron coordination complex and boron–oxygen (B–O) bonding in the p⁺ diffused layer were measured through X-ray photoelectron spectroscopy (XPS). The correlation between electrical properties, such as minority carrier lifetime (τ_eff), implied open-circuit voltage (iV_oc) and saturation current density (J₀), according to the change in element bonding, was analyzed. For the interstitial defect, the boron ratio was over 1.8 and the iV_oc exceeded 660 mV. Additional gains of 670 and 680 mV were obtained for the passivation layer AlO_x/SiN_x stack and SiO₂/SiN_x stack, respectively. The blue response of the optimized p⁺ was analyzed through spectral response measurements. The optimized solar cell parameters were incorporated into the TCAD tool, and the loss analysis was studied by varying the key parameters to improve the conversion efficiency over 23%.

1. Introduction

The n-type crystalline silicon solar cells possess advantages in terms of high efficiency due to their higher minority carrier lifetime, robustness and light-induced degradation (LID) in comparison with p-type substrates silicon solar cells [1]. However, p-type-based solar cells dominate the photovoltaic market with the advantage of material cost due to their cheaper wafer cost and process compatibility and the fact that they have an easily formed emitter and local back-surface field (LBSF) using phosphorus diffusion and Al–Si-alloy formation through laser-opening screen printing—a fast-firing process [2].

Despite this situation, the International Technology Roadmap for Photovoltaic (ITRPV) predicts that n-type products, which currently have about 10% market share, will have an equal market share with p-type products by 2029 [3]. This means the benefits of using n-types are similar to those of existing p-types. Though the n-type wafer price is higher than that of the p-type, the market predicts that it can be compensated with high-efficiency potential. The mass production of various n-type high-efficiency technologies, such as the TOPCon (also known as passivated contact) and the monopoly solar cell [4], which is under continuous development, will accelerate this situation [5,6,7,8]. In this regard, optimization of diffusion technology to form good-quality p–n junctions is the basis for realizing the high-efficiency potential of n-types and must be emphasized in the future. This is because the formation of p⁺ diffused layers using BBr₃ is expected to maintain about 85% of the market share until 2029, as there is no viable alternative available [3,9,10,11].

In this paper, the effect of the boron–silicon (B–Si) coordination complex and boron–oxygen (B–O) pair on the electrical characteristics of the emitter was found through bonding peak analysis for the optimization of the p⁺ diffused layer. In addition, trend analysis of solar cell output characteristics was carried out according to the change of the passivation layer on the p⁺ diffused layer. Assuming that the sheet resistance (R_sheet) is higher than the current as predicted in the market, determination of the dominant factor for improving the efficiency and loss factor in whole solar cell structures, such as recombination front/rear surfaces and contact loss, was analyzed using the TCAD tool [12,13].

2. Experimental Details

The n-type solar-grade Czochralski (Cz-Si) wafers with (100) crystal orientation, a resistivity of 1.5 ohm-cm and a thickness of 200 μm were used as the substrate for solar cell fabrication. To reduce the reflectance of the silicon wafer, texturing was carried out with 2% NaOH with 8.75% IPA at a temperature of 84–86 °C and an etching time of 5 min. The resultant pyramid size was under 2 μm. The textured n-type c-Si wafers were doped with boron using a conventional thermal diffusion process in a furnace at 910 °C with a 5-min pre-deposition and at 1000 °C with a 7.5-min drive-in step using the BBr₃ dopant source, respectively. After the diffusion process, removal of the borosilicate glass (BSG) was carried out by dipping in a diluted HF solution for 30 s, followed by DI-water rinsing and drying. The heavily diffused p⁺ layer ending in a R_sheet of 70 ohm/sq was used as the control group. The p⁺ emitter was then etched back with the HF-HNO₃-CH₃COOH mixture. The etch-back time was varied so that the R_sheet ended with 80, 100 and 200 ohm/sq. XPS was used to analyze the formation characteristics of impurities, such as boron, silicon and oxygen within the boron-diffused layer. The p⁺ layers were passivated with three different passivation layers, such as SiN_x with 72 nm thickness, and AlO_x/SiN_x and SiO₂/SiN_x with 10/62 nm thickness, respectively. The AlO_x and SiN_x layer was deposited using PECVD (plasma-enhanced chemical vapor deposition), and SiO₂ was formed in a high-temperature furnace at 850 °C. The passivation quality and interface characteristics of the deposited layers were evaluated by the minority carrier lifetime (τ_eff) and open-circuit voltage (iV_oc). The τ_eff of the symmetrical test structures was measured by a contactless photoconductance tool (WCT-120, Sinton Instruments) at an excess carrier density of n = 10¹⁵ cm⁻³.

The front-side-emitter-doped wafers were diffused using pentavalent impurity (phosphorus) to form an n⁺ back surface field, with a phosphoryl chloride (POCl₃) diffusion at 810 °C for a 7-min pre-deposition and at 880 °C for a 5-min drive-in step using the dopant source. PECVD-SiNx passivation stacks with refractive indices of 2.7 and 2.1 were deposited on the rear. High-throughput metal contacts were formed on the rear surface using a silver paste. Additionally, fingers and busbars were formed on the front surface using a silver and aluminum mixture paste with a screen-printing process. The samples were then baked at 150 °C for 4 min and cofired in a conveyer belt furnace at 865 °C with 110 IPM (inches per meter), respectively. The boron-diffused layer characteristics with fabricated cells were analyzed by using an IPCE (incident photon-to-current conversion efficiency) measurement system, QEX-7, in the wavelength region of 300 to 1100 nm and an XPS system, K-Alpha (Thermo Electron), in the detection depth under 10 nm resolution. Finally, by extracting the characteristics of the boron-diffused layer, we analyzed the key parameters needed for high efficiency, such as loss in emitter, bulk, BSF and contact region after optimizing the boron emitter with the TCAD tool.

3. Results

Figure 1 shows the normalized peak analysis of B(1s) using XPS on a sample with variable R_sheet of the emitter using an etched-back process. The black line is the XPS raw signal and the red line is the line after fitting. As a result of deconvolution of the fitted XPS peak, when excluding trivalent boron (B(III)), which is broadly distributed in binding energy [14], it is divided into 186.7 and 187.8 eV. The peak with a binding energy of 186.7 eV is mono/divacancy-substitutional B, and it mainly acts as an interstitial defect in the p⁺ diffused layer due to the combination of boron and vacancy (V₄B₂-type defect) [15,16]. In the case of a peak having a binding energy of 187.8 eV, the boron atom is sufficiently activated in Si and activated electrically. Afterwards, when the binding energy increases further, it shows the peak of BEH (highest binding energy, 189.6 eV) [17], which is considered to appear above the solubility in silicon [18]. This is the basis for the formation of an inactive boron in the p⁺ diffused layer, but it is excluded from the discussion because no peak appeared in this paper. It can be interpreted that as the binding energy shifts, the interstitial defect decreases, and the sufficiently activated boron in silicon increases. Thus, it shows an increase in conductivity [19], which affects the electrical properties of the p⁺ diffused emitter and the output properties of the solar cell.

Figure 2 shows the results of XPS peak analysis of Si2p and O1s orbitals. The Si₀ (Si–Si Bonding) peak with a binding energy of 100 eV [20] does not show a significant difference with respect to the change in the R_sheet, but in the case of boron–oxygen bonding binding energy with 533 eV [21], the peak decreases with an increase in the R_sheet. In the case of boron–oxygen bonding, the B_i–O_i (interstitial boron–interstitial oxygen dimer) or B_S–O_i (substitutional boron–interstitial oxygen dimer) complex acts as a defect site. The B_i–O_i defect recombination model can exist only in the stable state. However, in the metastable state, it is difficult to describe the recombination defect center as the B_i–O_i defect mechanism. The model developed to describe the defect model in the metastable state is a recombination defect model by B_S–O_2i, a substitutional boron (B_S) and an oxygen atom dimer (O_2i) complex. Substitutional boron is largely immobile in the silicon lattice, and there is a negligibly small amount of interstitial boron. Therefore, substitutional boron and oxygen atoms mainly react. In addition, O_2i reacts preferentially to the substitutional boron, which is explained by two reasons. The first reason is that the diffusivity in silicon is several times higher in the case of O_2i, which is the agglomerates of two interstitial oxygen atoms, compared to interstitial atomic oxygen, and secondly, the tetrahedral covalent radius of the B atom (0.88 Å) is 25% smaller than the Si host atom (1.17 Å) [22]. As the peak of the boron–oxygen complex changes, the recombination center caused by the B–O complex decreases, resulting in the minority carrier lifetime, iV_oc, and the saturation current density of the p⁺ diffused layer. This leads to the change in the electrical properties and affects the output characteristics of the solar cell.

Figure 3 analyzes the electrical properties of the p⁺ diffused emitter by the relative ratio of the activated boron (binding energy (BE) = 187.8 eV) and interstitial defect boron (BE = 186.7 eV) peaks, obtained through the previous XPS analysis. The ratio of the activation/interstitial boron peaks shown on the x-axis is 0.58, 1.81, 1.89 and 2.24, and the R_sheet of the p⁺ diffused layers are 70, 80, 100 and 200 ohm/sq, respectively.

The electrical properties of the p⁺ diffused layer, such as effective lifetime, iV_oc and saturation current are obtained by the following formula [23,24].

$$V_{OC}=\frac{kT}{q}\ln \left(\frac{J_{SC}}{J_{o,total}}+1\right)$$

(1)

$$\frac{1}{{\tau }_{eff} \left(\Delta n\right)}-\frac{1}{{\tau }_{Auger}}=\frac{1}{{\tau }_{SRH} \left(\Delta n\right)}+\frac{1}{q{n}_i^{2}W}·J_{0,total}·\Delta n$$

(2)

where k is the Boltzmann constant, T is Kelvin temperature, J_SC is current density at 1-sun illumination, τ_SRH is the bulk Shockley–Read–Hall (SRH) lifetime, τ_Auger is the bulk Auger lifetime, J_0,total is the total saturation current density, q is the elementary charge, n_i is the intrinsic carrier density, Δn is the excess carrier concentration and W is the wafer thickness. The open-circuit voltage, in which the final output characteristics of the solar cell are directly proportional to the minority carrier lifetime, saturation current density of the bulk, emitter, back-surface field and total (J_0,bulk + J_0,emitter + J_0,BSF = J_0,total), is inversely related, respectively. The decrease in saturation current density in the p⁺ diffused layer reveals that the activated boron has diffused significantly so that the electrical property degradation caused by the defect site is minimized, leading to an improvement in the minority carrier lifetime and an increase in the open-circuit voltage and, hence, conversion efficiency of the solar cell.

Figure 4 shows the results of the minority carrier lifetime and iV_oc obtained by depositing SiN_x with a positive charge, an AlO_x thin film with a negative charge and a thermally grown SiO₂ thin film with a neutral charge as a passivation layer [25] on the p⁺ diffused layer with variable R_sheet (activation/interstitial boron peak ratio change). For passivation of the p⁺ diffused layer, it is effective to improve the carrier lifetime and open-circuit voltage using AlO_x with a negative charge or an SiO₂ layer with a neutral charge. Conversely, the SiN_x layer that has a positive charge is not suitable for passivation of the p⁺ diffused layer due to the deterioration of its properties. In the case of the SiN_x single layer, the iV_oc is limited to the maximum 655 mV, whereas in the case of the sample applied with AlO_x/SiN_x, it can be confirmed that when the FGA (forming gas annealing) process [26] is performed after fast-firing, it reaches from a minimum of 661 to a maximum of 670 mV. Based on this result, the passivation of the p⁺ diffused layer and the difference in iV_oc of about 6 to 15 mV remains dependent on the charge polarity, which is the difference between the SiN_x and AlO_x layers.

It can be seen that the passivation quality of the thermally grown SiO₂ thin film is superior in terms of lifetime and iV_oc to that of AlO_x/SiN_x, which is due to the difference in interface trap density between the thin film formed in the PECVD and the thermal furnace.

When AlO_x is deposited on the surface of the p⁺ diffused layer, an accumulation layer is formed by a negative charge, and field effect passivation is performed on the p⁺ diffused layer. As can be seen in Figure 4, the SiN_x layer passivation characteristic represented by the yellow line and bar is inferior to the AlO_x/SiN_x double-stacked layer passivation characteristic represented by the blue line and bar. When the fast-firing process is performed after the deposition of the AlO_x/SiN_x layer, degradation of the minority carrier lifetime and iV_oc characteristics occurs. Additionally, after undergoing the FGA process, the degradation due to firing is recovered. The quality of passivation between the AlO_x and p⁺ diffused layer is lowered by thermal stress when fast-firing, and the lifetime and iV_oc are recovered due to the hydrogen covering the dangling bond of the p⁺ surface through the FGA process.

Figure 5 shows the external quantum efficiency of the solar cells made with different R_sheet. These samples were passivated with the optimized AlO_x/SiN_x layer. When the ratio of quantum efficiency at 300 nm wavelength was assumed to be 100% with the 200 ohm/sq sample, the blue response also increases with 88%, 92% and 94% as the R_sheet increases, respectively. A higher R_sheet will contribute to increase the short wavelength absorption, and when the emitter is heavily doped (lower R_sheet), the spectral response at short wavelength characteristics is decreased. As can be seen in the analysis of the boron–silicon–oxygen bonding phenomenon through the XPS, minority carrier lifetime and iV_oc, the change of the active/interstitial defect boron ratio and B–O bonding complex decreased in boron–silicon bonding coordination. The quantum efficiency of the p⁺ diffused layer, which acts as an emitter in the n-type-based c-Si solar cell, can also be confirmed by the improvement of properties as the R_sheet increases.

Figure 6 shows the loss analysis carried out for optimization and contributed to the efficiency improvement, such as the loss in emitter, bulk, BSF and contact region. The loss analysis was carried out using the TCAD program. The optimized values obtained from Figure 3 for the 70, 80, 100 and 200 ohm/sq R_sheet were inputted into the TCAD program. The front surface recombination velocity values were varied, and we obtained a conversion efficiency of 20.15%, 20.52%, 20.59% and 20.61%, respectively.

In this TCAD simulation, five key parameters were selected for efficiency improvement and optimization: the “charge effect”, which applies negative charge to improve the characteristics of the front passivation layer; “front and rear recombination loss”, which is based on the quality of the interface trap density of the front and rear passivation layer; wafer quality, to reduce the “bulk recombination loss”; and “contact loss”, indicating the contact characteristics between the p⁺ diffused layer and front electrode. In the TCAD simulation, the baseline contact resistance was assumed to be 5 mΩ∙cm⁻² and varied the contact resistance in steps of 0.5 mΩ∙cm⁻² during optimization [27]. The front and rear recombination loss parameters were optimized by varying the front and rear surface recombination velocities from 900 to 600 cm/s in steps of 10 cm/s, respectively [28,29,30].

Though there is potential for improvement of conversion efficiency with the increase in the emitter R_sheet, there are few other key parameters needed to be optimized based on the R_sheet. The major parameter showing the largest change is the loss due to the contact resistivity of the p⁺ emitter and the front finger. In the case of the 70 ohm/sq, the ratio of contact loss to the overall efficiency improvement range is 9.25%, but in the 200 ohm/sq, it accounts for 36.07%. This is because the surface concentration related to the contact resistance is relatively low as the R_sheet increases. To compensate the contact loss caused by the increase in the R_sheet, optimization of the Ag–Al (Silver–Aluminum) paste [31], selective emitter technology [32] and plating using a seed layer [33] can be considered.

Next is the front recombination loss and charge effect. On average, the conversion efficiency of both parameters increased by 0.75% and 0.47%. In order to achieve high efficiency, the passivation layer, either the AlO_x layer deposited by ALD (atomic layer deposition) or the thermally grown SiO₂ thin films containing a negative or neutral charge are suitable to reduce the recombination compared to the PECVD process. The rear-side recombination loss and bulk properties were 0.44% and 0.425%, respectively, which contributed to the improvement of conversion efficiency.

4. Conclusions

For the high efficiency of crystalline silicon solar cells using n-type substrates, the optimization of electrical properties of boron diffusion and the emitter, which are the most basic properties, was examined in terms of boron–silicon coordination bonding and oxygen complexes. The change in the electrical properties depends on the ratio of activation/interstitial boron peaks in the boron-diffused layer, the charge polarity and the passivation layer of the p⁺ diffused surface. The experimental values were incorporated into the TCAD simulation, and loss analysis was carried out. The recombination losses at the front and rear and the contact loss must be minimized to achieve high-efficiency silicon solar cells.