Abstract

Selective emitter solar cells can provide a significant increase in conversion efficiency. However current approaches need many technological steps and alignment procedures. This paper reports on a preliminary attempt to reduce the number of processing steps and therefore the cost of selective emitter cells. In the developed procedure, a phosphorous glass covered with silicon nitride acts as the doping source. A laser is used to open locally the antireflection coating and at the same time achieve local phosphorus diffusion. In this process the standard chemical etching of the phosphorous glass is avoided. Sheet resistance variation from 100 Ω/sq to 40 Ω/sq is demonstrated with a nanosecond UV laser. Numerical simulation of the laser-matter interaction is discussed to understand the dopant diffusion efficiency. Preliminary solar cells results show a 0.5% improvement compared with a homogeneous emitter structure.

1. Introduction

Selective emitter technology can provide a significant increase in solar cell efficiency, but today most of the approaches need many processing steps and alignment procedure [1]. In this context, laser processing provides a good opportunity to achieve such a structure with a minimum number of technological steps [25] and without alignment when coupled with front side electrochemical metallisation [6]. Recent results have shown that around a 0.5% increase in solar cell efficiency could be obtained when using phosphosilicate glass (PSG) as a source of dopants to form highly doped areas on the emitter side according to Process A in Figure 1. In the perspective of a complete self-aligned process, a reduced number of technological steps is proposed to form selective emitters on the front surface of p-type silicon wafers (Figure 1, Process B). In this process, a homogenous emitter is formed by thermal diffusion of phosphorous. The resulting PSG layer is covered with silicon nitride (SiNx) for passivation and antireflection purposes. Ablation of the SiNx coating and dopant diffusion are then performed simultaneously with a pulsed laser before metallisation. Wet chemical etching of the PSG layer is therefore avoided and a self-aligned metallisation process can be developed with electrochemical deposition techniques. Numerical simulation of laser-matter interaction is presented to explain the heat-assisted diffusion mechanism and is compared to sheet resistance measurements. Laboratory scale solar cells are fabricated to compare homogeneous emitter and selective emitter structures. Spectral response and I-V measurements are used to characterise the solar cells.

2. Numerical Simulation of the Laser-Matter Interaction

Melting, evaporation, and dopant diffusion are the main mechanisms that result from the laser-matter interaction. In a nanosecond regime, if the optical and thermal penetration depths are much smaller than the diameter of the incident laser beam on the surface, the thermal effects can be described by a two-dimensional heat-transfer equation [7]. The diffusion process can be estimated by solving at the same time Fick’s second law. To estimate the impact of the laser parameters on the emitter formation, the finite element software COMSOL was used to solve these two equations [8]:

is the material density, the specific heat capacity, the temperature, the time, the thermal conductivity, the heat source in the volume due to the absorbed laser power, the phosphorous concentration in silicon, and the diffusion coefficient of the phosphorous atoms in silicon. The heat source term in the heat-transfer equation corresponds to the absorbed laser power and can be written as follows:

is the material absorption coefficient, the surface reflectivity, the incident laser power, and the relative intensity given by the Beer-Lambert law. The beam is considered to have a Gaussian shape in time and space:

is the peak power of the laser beam, the time shift, the pulse duration, and the beam radius at half height.

Temperature dependence of the physical properties of the materials was taken into account. The smoothed Heaviside function (flc2hs) implemented in COMSOL was used to describe the abrupt changes of material coefficients with temperatures. Latent heat of fusion was taken into account by reducing the heat-source term for the temperatures above fusion by the quantity , where is the latent fusion heat, the density and the pulse duration. In a first approximation, the texturation of the surface was not taken into account and a plan of symmetry was considered in order to minimize the calculation time. The absorption coefficient of silicon at 355 nm is around 108 m−1 corresponding to a penetration depth around 10 nm. As a consequence, the minimum element size for calculation at the silicon surface was 1 nm. The substrate was divided in different areas in order to have a finer meshing under the irradiated area. The time parameter varied from 0 to 60 ns with a step of 1 ns and 0.1 ns during the laser pulse (from 10 ns to 40 ns). The sample geometry is represented in Figure 2.

For a better understanding of the role of each layer, temperature at the surface of silicon was calculated as a function of the laser fluence (Figure 3) for the different interfaces Si/air, Si/PSG/air, and Si/PSG/SiNx/air. Temperature variation at the silicon surface can be explained by the thermal properties of the layers (mostly the thermal conductivity) and by the different reflection coefficient of the surfaces. Laser doping efficiency is dependent on the temperature reached in silicon as the substrate has to be melted to allow efficient phosphorous diffusion into the junction. Indeed, the significant increase of the diffusion coefficient in the liquid state silicon (×107) permits phosphorus diffusion during the short time of laser-matter interaction (tens of nanoseconds). As shown in Figure 3, the structure Si/PSG/SiNx/air needs more laser energy than the structure Si/PSG/air to melt the silicon substrate. Therefore a good control of the laser parameters is necessary to avoid heat-induced damages to the cell junction.

3. Experimental

Monocrystalline Cz silicon wafers with textured surfaces were used. Substrates were p-type with a thickness of 250 μm and a resistivity of 1 Ω·cm. Homogeneous n-doped emitters with sheet resistance of 80 Ω/sq and 100 Ω/sq corresponding to surface doping concentration of  cm−3 and  cm−3, respectively, were fabricated by low-pressure thermal diffusion of POCl3 using the Lydop technology. The PSG layer formed at the wafer surface during thermal diffusion was preserved for use as a doping source during the laser-assisted diffusion process on the 100 Ω/sq samples. The 80 Ω/sq were kept as homogeneous emitters for comparison. An 80 nm thick SiNx layer was deposited over the PSG layer by conventional PECVD process for passivation and antireflection purposes. A frequency tripled Nd : YAG laser from ROFIN with a wavelength of 355 nm and a pulse duration of 10 ns was used for heat-assisted diffusion.

Simples test structures were fabricated to optimise the laser parameters. Test structures were characterised by optical microscopy, 4-point probe equipment, Transmission Line Method (TLM), and Quasi-Steady-State Photoconductance. Laboratory scale solar cells were characterised by illuminated I-V and spectral response measurements.

3.1. Sheet Resistance Measurements

Figure 4 shows the typical evolution of the sheet resistance with laser fluence and pulse overlap measured on samples with a 100 Ω/sq initial emitter. Complete ablation of SiNx without debris was observed between 0.4 J·cm−2 and 0.7 J·cm−2. As expected from the simulation, laser-assisted doping was more efficient at low fluence for samples with PSG only. A decrease of the sheet resistance was measured at low fluence (0.25 J·cm−2) in PSG-capped silicon while no laser doping was measured for fluence under 0.4 J·cm−2 in PSG/SiNx-capped silicon. Moreover, 0.2 J·cm−2 and 50% overlap was sufficient to reach 65 Ω/sq for a PSG-capped silicon while around 0.45 J·cm−2 and 50% overlap was necessary to reach the same doping level for a PSG/SiNx-capped silicon. On the other hand, at high fluence the minimum sheet resistance was lower for PSG/SiNx-capped silicon (20 Ω/sq) than for PSG-capped silicon (30 Ω/sq).

3.2. Passivation Properties

The influence of the PSG layer on the carrier lifetime was studied on different samples by transient decay photoconductance. FZ wafers with a thickness of 400 μm and a resistivity of 800 Ω·cm were used for thermal diffusion followed by a standard SiNx deposition on both surfaces. The PSG layer was etched on one sample before SiNx deposition. FZ samples with or without a PSG layer under the SiNx coating showed similar lifetime around 200 μs for a 60 Ω/sq emitter. A similar procedure was applied to a lower quality multicrystalline silicon wafer with a thickness of 250 μm and a resistivity of 1 Ω·cm. Samples with PSG, PSG/ SiNx, and SiNx on both faces were fabricated. All showed a similar lifetime around 10 μs. These first observations on high and low quality silicon substrates are promising for practical use of PSG/ SiNx as passivating bilayer.

3.3. Laboratory Scale Solar Cell Results

Selective emitter solar cells were made following a modified version of the Process B presented in Figure 1. Front side metallization was made by Ti/Pd/Ag evaporation preceded by a photolithography step. Homogeneous emitter solar cells were also fabricated following the same procedure without any laser step and with a selective chemical etching of SiNx. Selective emitter cells were realised on substrate with an initial emitter of 100 Ω/sq and homogeneous emitter cells with an initial emitter of 80 Ω/sq. Laser parameters were chosen to reach a sheet resistance of 40 Ω/sq under the metal contacts of the selective emitter cells. Table 1 shows the average parameters extracted from the illuminated I-V measurements with and without laser step. Nine cells (4.1 cm2) were measured for each process.

The selective emitter cells showed a 0.5% absolute increase in efficiency compared to the homogeneous emitter cells. The overall low efficiency of the solar cells was attributed to nonoptimized technological steps. Figure 5 shows an example of front contact metallisation misalignment that was observed in some of the cells. The laser-treated zone extends under the photoresist on the right hand side of the image. After metal deposition and chemical etching this area without passivation will be exposed to strong recombination. Lack of Ti/Pd/Ag metal sintering and nonoptimised antireflection properties of the PSG/SiNx stack might also be responsible of the global loss of efficiency. For all cells, shunt resistance appears as the main cause of parasitic resistive losses with an averaged value around 670 Ω·cm2. Series resistance is reasonably low with an averaged value around 0.6 Ω·cm2. The low values of are mainly attributed to recombination at the front surface.

Figure 6 shows the typical internal quantum efficiencies (IQEs) measured for homogeneous emitter (HE) and selective emitter (SE) cells. As expected, an increase of IQE is visible in the blue part of the spectrum for the selective emitter solar cells suggesting that the surface passivation is not degraded by the presence of the PSG layer at the surface.

4. Conclusion

Selective emitter solar cells were fabricated with a reduced number of technological steps. Wet etching of the PSG was not necessary as this layer was preserved and covered by silicon nitride before laser-assisted diffusion. This process is potentially self-aligned if electrochemical front side metallisation is used. Preliminary results show reasonably good cell results with a 0.5% improvement compared with a homogeneous emitter structure. Several technological adjustments are foreseen on the antireflection and passivation properties of the PSG/SiNx stack and on the front metal contact geometry to improve these first results. Long-term stability of the PSG layer has also to be investigated. Furthermore, numerical simulation of the laser-matter interaction is shown to be very useful to predict and understand the heat-assisted diffusion mechanism.

Acknowledgments

This work was sponsored by the French Agency for Environment and Energy Management (ADEME) and Agence Nationale pour la Recherche (ANR).