Improvement of Laser-Induced Damage on High-Efficiency Solar Cells via Top-Hat Beam Ablation

Abstract

An important challenge in industrial laser ablation is laser-induced damage. In this study, reduced damage was achieved through the transition of the laser distribution from a Gaussian beam to a top-hat beam using diffractive optical elements (DOE), which overcome inhomogeneous irradiation. The higher peak fluence of a Gaussian beam far exceeded the ablation threshold and led to severely melted silicon at a higher depth covering the polished texture. The top-hat beam, with uniform irradiation, had a superior ablation characteristic and created a uniform square opening with the shallow melted silicon in the lift-off process. Thus, its effective minor carrier lifetime was 15.35% less at an ablated area fraction of 2% after re-passivation because of the decreased damage. After optimizing the ablation pattern with a top-hat beam, the local contacts improved the average open-circuit voltage (Voc) and short-circuit current (Isc) values of the cells due to the decreased damage and the uniform openings, but the damage induced by a Gaussian beam was too deep and can be partly restored under back surface field (BSF) formation. The overall increment in Isc and Voc enhanced the average efficiency by 0.05% of the absolute value for the PERC cells and 0.03% of the absolute value for bi-facial PERC cells.

1. Introduction

In recent years, the laser ablation of dielectric layers has been widely used in producing advanced solar cell structures, such as the laser opening of dielectrics for Ni/Cu plating cells [1,2,3], laser patterning for interdigitated back contact (IBC) cells [4,5] and passivated emitter and rear contact (PERC) cells [6,7]. Several studies about the ablation of PERC cells have introduced various laser patterns (dot, dashed line or line) to improve the cell efficiency [8,9,10]. However, the requirements for laser ablation are not only to selectively remove the layers, but also to reduce the damage induced during the laser process, such as silicon melting, heat-affected zones and dislocations [11,12,13], which can negatively affect solar cell performance [12]. The studies [13,14] have investigated the effect of the damage etching process on the performance of industrial PERC solar cells. Kim et al. [14] showed that after laser-induced damage etching using a KOH solution, the conversion efficiencies of PERC cells were improved by approximately 0.28% of the absolute value, but an additional chemical cleaning process was needed. Some alternative approaches using ultrafast picosecond (ps) laser sources have demonstrated superior ablation characteristics to minimize laser damage [15,16,17], but these techniques are more expensive than an industrial nanosecond (ns) laser.

Previous studies have investigated laser patterns [8,9,10] and various laser pulses of ns or ps lasers with different wavelengths [15,16,17]. However, a research gap regarding the shapes of the laser beam still exists. A model with two thresholds that correspond to the melting threshold of silicon and the ablation threshold of the dielectric layer was proposed for a Gaussian beam, and the threshold fluence of the ablating dielectric layer far exceeded the melting threshold fluence of silicon [18]. In order to obtain a large enough opening area, the fluence was adjusted to a high level, and its surplus peak fluence may generate a huge heat shock in the silicon underneath. The Gaussian beam is composed of inhomogeneous irradiation, so a homogeneous removal was not possible by using a Gaussian beam. The absorption of light is predicted to be dependent not only on the laser parameters but also on the geometry [19], and a homogeneous removal is also not possible since the pyramids can lead to laser intensity amplifications [20]. Compared with a textured surface, a polished surface was used to reduce the dislocations mostly formed at the pyramid tips, which absorb more energy from the laser source [11]. In addition, the electric data of the cells benefit from a flat rear surface due to decreased recombination and improved light trapping [21]. In our work, the laser ablation process with a top-hat beam, combined with a single etching process for a planer rear surface, was designed to fabricate PERC cells with improved cell efficiency. The technology in this study is a cost-effective way of using an industrial ns laser instead of an expensive ps laser.

Our study aims to understand, develop and control the top-hat laser ablation conditions and patterns for PERC cell processing. Average efficiencies of 22.523% and 22.43% were obtained from 400 pcs PERC cells and 400 pcs bi-facial PERC cells, respectively. These rates exceeded those of the reference cells by 0.05% and 0.03%, mainly because of the higher V_oc values and I_sc values.

2. Experimental Process and Details

In our experiment, PERC solar cells were manufactured on Czochralski silicon wafers in accordance with the process flow A displayed in Figure 1. The wafers that featured a resistivity of 1.5–1.6 Ω·cm were Ga-doped with an area of 252 cm² and a thickness of 185 μm.

After the texturing process, a phosphor-diffused emitter was fabricated in a tube furnace for a sheet resistance (Rsq) of 110 Ω/sq. Subsequently, a 532 nm pulse laser of pulse duration 40 ns, with a squared top-hat beam size of 100 µm and a fluence of 1.0 J/cm², was utilized to create local heavily doped emitters in the regions of busbars and fingers, where the value of Rsq decreased to 80 Ω/sq. Before the subsequent wet-etch processing, a barrier of 2 nm-thick SiO₂ was formed through thermal oxidation at 780 °C for 30 min to protect the laser doping regions. During the single-side wet-etch processing, the removal of the remaining phosphor silicate glass (PSG) in the rear and edge was firstly accomplished via an inline wet-bench. Then, a 6-vol% alkaline solution, with the assistance of functional additives at 80 °C for about 3.5 min, was applied to clear the rear and edge emitter. After that, the protective PSG was etched off using a HF solution. The cells were then passivated with thermally grown 2 nm-thick SiO₂ layers on both sides. Subsequently, an 80 nm SiN_x (n = 2.10) antireflection layer was deposited by plasma-enhanced chemical vapor deposition (PECVD) on the front side. The rear side was coated by a passivation stack of 10 nm-thick Al₂O₃ and 80 nm-thick SiN_x deposited by another multifunctional PECVD. The pattern of rear side stack opening was carried out with a 2 ns pulse laser with a 532 nm wavelength and a top-hat beam by DOE or a Gaussian beam, and the fluences varied from 0 to 3 J/cm². Finally, front silver (Ag) paste, rear Ag paste and rear aluminum (Al) paste containing boron additives were screen printed and co-fired for front contacts, rear contacts and rear busbars, respectively. To reduce the series resistance, a pattern of five 0.9 mm-wide busbars and 110 fingers was exploited for the front metallization. The difference between PERC cells and bi-facial PERC cells was the existence of various Al metallization patterns. A bi-facial PERC cell was fabricated with Al fingers instead of the full-area Al layer.

The four-point probe method was used to measure the sheet resistance and each wafer had five testing points. A laser ellipsometer (EMPro-PV) was used to measure the thickness of the dielectric layer. The cell characteristics were measured under a simulated AM1.5 (air mass 1.5) spectrum at 25 °C, generated by a solar simulator on a h.a.l.m. system (halm elektronik gmbh). The internal quantum efficiency was measured by QE-R (Enlitech).

To estimate the damage induced by laser ablation, symmetrical test samples were prepared on chemically polished Ga-doped wafers according to the process sequence as shown in Figure 1 (process B). Thermally grown 2 nm-thick SiO₂ layers and stacks of 10 nm-thick Al₂O₃ and 80 nm-thick SiN_x on both sides were formed using a tube furnace and multifunctional PECVD, respectively. Thereafter, each wafer was divided into four areas with various open-area fractions of 0%, 2%, 4% and 8%, as shown in Figure 2. The wafers were ablated using a laser with a Gaussian beam or a top-hat beam separately. Then, the original passivation layer was removed by HF, but the damage induced during the laser ablation process remained. Finally, the samples were re-passivated through the deposition of a new stack of 10 nm-thick Al₂O₃ and 80 nm-thick SiN_x on both sides. The laser-induced damage was evaluated based on a WCT-1200-calibrated minor carrier lifetime. The residual lifetime ratio is defined as the ratio between the lifetime of the non-ablated area and the ablated area. The laser ablation was assessed through optical microscopy and scanning electron microscopy (SEM).

3. Results and Discussion

3.1. Characterization of Ablation Quality

The interaction between the laser irradiation and matter depends strongly on the parameters of the laser and the physical and chemical properties of the material [17]. Thus, optimizing the laser parameters, such as the fluences and pulse overlap ratios (POR), is crucial to decrease laser-induced damage [17]. The POR is defined as the ratio between the laser pulse overlap length and the laser pulse diameter. The values of the pulse overlap ratio were set to zero and a negative value for two tangent pulses and two isolated pulses, respectively. In these experiments, we set the value of POR as zero.

A Gaussian beam and a top-hat beam have different irradiation distributions. A Gaussian beam has a peak power and a top-hat beam has a uniform irradiation, as shown in Figure 3. The profiles were obtained from the numerical simulations performed using Matlab software through the theoretical models of super-Gaussian beams [22].

The optical microscope images of the ablation characteristics for a Gaussian laser beam are shown in Figure 4. The size of the ablated area grew as the fluence increased. At a low fluence of 0.7 J/cm² (Figure 4a), the opening of the stack started from the area irradiated by the center of the beam as a result of it having the strongest peak power. However, the fluence of the surrounding area was lower than the threshold. The passivating stack was only partly ablated at such a fluence, where a high contact resistance would be generated. As shown in Figure 4b, when the laser fluence increased to 1.4 J/cm², an increase in the size of the ablated area was obtained and the polished texture was observed clearly, so less melted silicon had occurred but the ablated area consisted of a row of isolated points. At a fluence of 2.06 J/cm², the molten silicon in the center of the ablated area and the heat-affected region of the passivation layer around the opening were observed, as shown in Figure 4c. Further increasing the fluence to a higher value was critical to ablate the full area. The final condition was the strongest ablation applied, with a fluence of 2.79 J/cm². Figure 4d reveals a row of tangent opening points and an increased depth of severely melted silicon in the center without a polished texture.

For a ns laser, the laser was absorbed by silicon, since the band gap energy of silicon and the passivating stack were below and above the photon energy, respectively. Thus, the silicon melted locally and, induced by mechanical stress inside the layer, lifted off the dielectric layer [23]. For a Gaussian beam, the laser fluences exceeded the threshold of silicon melting and the dielectric layer ablation, respectively [18]. The ablation process started when the peak fluence of the Gaussian beam exceeded the ablation threshold, which was higher than the threshold for the melting of silicon. If the laser fluence further increased to a higher value than the other area that was ablated, then a greater depth of melted silicon was found in the central opening. Such a condition indicated that the fluence had already far exceeded the melting threshold. Thus, a uniform and complete opening was difficult to ablate by applying a Gaussian beam. A Gaussian beam ablation had a main drawback that the depth of the central ablated area was too large, which limited the conversion efficiency of the PERC cells.

Figure 5 depicts the ablation characteristics of a top-hat beam ablation from low fluence to high fluence. On the contrary, the top-hat laser ablation characteristics showed a different dependence on laser fluence. When the laser fluence was between 2.34 and 2.84 J/cm², the ablated areas almost remained the same for a top-hat laser beam. At low fluences of 1.57 and 1.95 J/cm², the shape of a square with unfilled corners occurred, as shown in Figure 5a,b. In addition, a polished texture was observed across all samples, from which we predict the occurrence of less molten silicon for a top-hat beam. The complete ablation of the dielectric stack with reduced laser-induced damage was obtained at a fluence of 2.34 J/cm², shown in Figure 5c.

In Figure 6, the sharp edges of the ablated areas for a top-hat beam were detected, while the severely melted silicon in the center and the surrounding burrs were found for a Gaussian beam, where the melted silicon was so deep as to cover the polished texture even at a low fluence of 2.06 J/cm². Increasing the fluence to 2.79 J/cm² was necessary to obtain enough opening area, however, an enlarged melted silicon area was inevitable, as shown in Figure 4d. The ablation threshold fluence of a Gaussian laser beam was higher than that of a top-hat beam because of its surplus peak fluence, which exceeds the ablation threshold fluence and leads to the melting of the silicon, as shown in Figure 6a. However, a top-hat beam can produce uniform irradiation to create lower damage and clean openings in the dielectric stack with less melted silicon, and a polished texture was observed as shown in Figure 6b.

3.2. Laser-Induced Damage in Si

To quantify the defect density level, we fabricated the samples on silicon wafers from the same silicon rod according to the process sequence in Figure 1 (process B). Then, lifetime testing was used to estimate the surface recombination caused by the impact of laser-induced defects. Indeed, some studies [11,17] found that the laser-induced damage remained when the surface was re-passivated after the ablated dielectric stack was removed without post-laser etching. As shown in Figure 2, each wafer was divided into four areas. For the wafer ablated by a Gaussian laser beam, the number 1 area was the non-ablated area, and the numbers 2–4 areas were ablated with different ablated area fractions, which were regulated in the range of 2%, 4% and 8% by adjusting the line spacing from 0.6 mm to 1.8 mm at various negative overlap values (−15% to −45%). Similarly, the same layouts were applied to the wafer ablated by a top-hat beam.

To confirm the level of laser-induced damage, the residual lifetime ratio was measured as a function of various ablated area fractions. The residual lifetime ratio is defined as the ratio between the lifetime of the non-ablated area and the ablated area. The residual lifetime ratio plays an important role because the presence of laser-induced defects can be assessed by a significant mirror lifetime loss. The residual lifetime ratios as a function of the ablated area fractions of 0%, 2%, 4% and 8% are shown in

Table 1. Detailed minor carrier lifetime data of different ablations from a Gaussian beam or a top-hat beam.

Splits	Laser Ablated Area Fraction
	0%	2%	4%	8%
Lifetime with a Gaussian beam (µs)	110.24	70.69	40.28	33.52
Residual lifetime ratio with a Gaussian beam (%)	100	64.12	36.54	30.41
Lifetime with a top-hat beam (µs)	108.88	86.53	60.31	40.68
Residual lifetime ratio with a top-hat beam (%)	100	79.47	55.39	37.36

. For a Gaussian beam, a remarkable decrease was observed in the values of the residual lifetime ratios as the ablated area fractions became large, where the residual lifetime ratio decreased by approximately 69.59%, from 100% in the non-ablated area to 30.41% in the 8% ablated area fraction. Similarly, the same trend was observed in a top-hat beam ablation process. Compared with the sample with a Gaussian beam, the sample with a top-hat beam showed a higher residual lifetime ratio at the same ablated area fraction because of less detrimental surface damage that ensured good surface passivation. By comparing the residual lifetime ratios, we found that a top-hat laser beam had a significant effect in decreasing the damage.

As shown above, the ablation quality was estimated by the residual lifetime ratios resulting from two types of laser beams. The laser-induced damage decreased in the silicon where the removal of the passivation layer was achieved by a top-hat laser beam ablation. By contrast, the results from the Gaussian laser beam indicated that more defects had been introduced compared to the top-hat beam. In addition, the highest conversion efficiency was obtained when the ablated area fraction varied from 2% to 3%. For the top-hat beam, the ablation at the optimized ablated area fraction of approximately 2% had a superior ablation characteristic, and the residual lifetime ratio was 79.47%, which is 15.35% higher than the Gaussian laser beam. The surface carrier lifetime (τ) is directly related to the carrier recombination process through the damage, and mainly affects the value of Voc as shown in Equation (1), where Jsc is the short circuit current density, W is the thickness of a wafer, ni is the intrinsic carrier density of silicon, k is the Boltzmann constant, q is the electronic charge and T is the temperature [12]. As the value of the damage density decreases, the value of τ increases, and the value of Voc is improved.

$$V_{0C}=\frac{2kT}{q}\ln \left(\frac{I_{sc}\times \tau }{q{n}_{i}W}\right)$$

(1)

3.3. Influence of Laser Ablation Fraction on PERC Solar Cells

Low-damage laser ablation was studied for the ablation of the PERC cells. However, the performance of a cell was heavily dependent on the laser pattern and the ablated area fraction [10]. The rear surface contact geometry was optimized to a dashed line (a:b) pattern of a μm opening to b μm distance. The radio (a:b) was used to control the ablated area fraction while keeping the same contact pitch.

Wafers from the same silicon rod were divided into four groups, each containing 400 wafers. Group 1 was used to fabricate the reference cells. The ablation used a Gaussian laser beam with a dashed line (500:500). Groups 2–4 were applied with a top-hat beam under the best ablation parameters as shown in Section 3.1 but different ablated area fractions. The effect of laser patterns on cell efficiencies was studied as a function of the ablated area fraction or the metallization fraction. The effects of the metallization factors on the cell electric data such as Voc, Isc value, fill factor (FF), series resistance (Rs) and cell efficiency (Eff) were studied in fabricated PERC cells.

Table 2. Detailed electric data of PERC cells with ablation of a Gaussian beam or a top-hat beam.

Splits	AF (%)	Eff (%)	Voc (V)	Isc (A)	FF (%)	Rs (Ω)
Group 1 **	2.1	22.475	0.6788	10.442	79.91	0.00263
Group 2 *	2.1	22.520	0.6790	10.465	79.87	0.00264
Group 3 *	1.89	22.523	0.6792	10.473	79.80	0.00271
Group 4 *	1.68	22.501	0.6795	10.472	79.68	0.00280
Best cell in Group1	2.1	22.563	0.6791	10.499	79.75	0.00281
Best cell in Group3	1.89	22.636	0.6804	10.515	79.73	0.00281

** represents the Gaussian laser beam, * represents the top-hat laser beam, and AF represents the ablated area fraction. The deviation for E_ff, V_oc, I_sc, FF or R_s was ±0.037, ±0.0002, ±0.011, ±0.12, and ±0.0001, respectively.

summarizes the electric data of cells with ablation of a Gaussian beam or a top-hat beam at various ablated area fractions. The best electrical data for the PERC cells with the top-hat ablation were in Group 3. Compared with the best cell from the conventional cells in Group 1, the greatest improvements in Voc, Isc and Eff were seen in the best cell in Group 3 due to the top-hat ablation process. The increments in Voc and Isc were 1.3 mV and 16 mA, respectively. The Eff of the best cell with the top-hat beam ablation was 22.636%, exceeding that of the best cell with the Gaussian beam ablation by 0.07%. The presented types of defects caused by laser ablation are silicon melting, heat-affected zones and dislocation. The impact of laser-induced surface defects on the cell performance was studied: when reduced damage was achieved by post-annealing, the value of Voc significantly improved, resulting in an increased conversion efficiency [12].

As shown in Table 2, compared with the baseline of group 1, an average efficiency enhancement of 0.045% in group 2 solar cells was realized, originating from higher values of Isc and Voc. The increments in the Voc value and the Isc value were 0.2 mV and 23 mA, respectively. The major difference between the cells of the two groups was the existence of the opening of the rear dielectric stack by various distributions of the laser beams. The top-hat laser beam process seems to have improved cell performance. As shown in Section 3.2, the damage decreased when the opening was exposed to the uniform irradiation of a top-hat beam due to the shallowly melted silicon. Bounaas et al. [16] studied the effect of surface damage with a Gaussian beam on the performance of cells, showing that it was somehow harmful to the cell performance when the depth of the damage was too deep, although it was partly restored under back surface field (BSF) formation. However, the depth of laser-induced damage by a top-hat beam was shallow enough so that an average efficiency enhancement was achieved as shown in Group 2.

To further improve the cell performance, the best ablated area fraction needs to be studied for a top-hat beam since it created an effective and uniform opening across the entire ablated area compared with a Gaussian beam. The proposed ablation patterns of groups 3–4 were extensions of group 2 and had lower ablated area fractions via control of the (a:b) ratios. The average Eff of the cells in group 3 was 0.003% higher than that of the cells in group 2 due to the lower ablated area fraction. The corresponding increases in Voc and Isc resulted from the suppressive rear-side carrier recombination and improved optical reflectivity. However, the cells of group 3 represented a slightly lower average FF value than the cells of group 2 due to the reduced total BSF area. If the ablated area fraction value was incorrect as in group 4, the average efficiency of the solar cells greatly reduced to even lower than the average efficiency of the solar cells in group 2, mainly due to a larger drop in the average FF value, although increased values were observed in the Voc and Isc. The ablated area fraction was a key parameter for the PERC cells, showing a trade-off between the value of Voc and the value of FF as the ablated area fraction varied. This result means that optimizing the ablated area fraction is needed to achieve high efficiency cells.

In addition, the best average Eff for a top-hat laser beam ablation we obtained was 22.523% in group 3. Compared with the conventional cells in group 1 with a Gaussian laser beam, an average efficiency enhancement of 0.05% was achieved for the cells that applied a top-hat laser ablation, originating from the improved values of Isc and Voc. The increments in the Voc value and the Isc value were 0.4 mV and 31 mA, respectively.

Figure 7 shows the results of internal quantum efficiency (IQE) of the i-PERC cell and the conventional PERC cell. Compared with the IQE of the conventional PERC cell, one of the i-PERC cells improved in the long wavelength from 800 nm to 1100 nm. The improvement in the value of IQE indicated a beneficial local BSF due to a top-hat beam ablation. We concluded that the reduced laser damage suppressed the carrier recombination at the rear side.

3.4. Influence of Ablation on Bi-Facial PERC Solar Cell

The bi-facial PERC is widely used for its power generation ability and high device compatibility with conventional PERC manufacturing lines. Compared with a full-area Al rear layer in the PERC cells, the Al fingers in the bi-facial PERC cells were prepared in order to absorb more sunlight from the rear side and produce much power from both sides. The wafers from the same lot were divided into two groups that applied either a Gaussian or a top-hat laser beam. The ablation pattern was dashed lines. The line pitch was 900 μm and the line width was 42 μm. The ablated area fraction was set to 2.3% for both of the groups. The alignment of the local printing to the openings was a challenge for this local printing method. The width of Al fingers of 120 μm was big enough to cover the ablation pattern. This region extended 78 μm beyond the width of the ablation line to incorporate the alignment tolerance between the ablated and metallization areas. The electrical data from the cells were tested for the front side of the cells according to the requirements of the customers, since an improvement in front Eff is important and the cells are sold using the power of the front side.

Table 3. Average electric data of bi-facial cells with ablation of a Gaussian beam or a top-hat beam.

Splits	AF (%)	Eff (%)	Voc (V)	Isc (A)	FF (%)	Rs (Ω)
Group 5 **	2.3	22.40	0.6778	10.490	79.41	0.00293
Group 6 *	2.3	22.43	0.6782	10.492	79.43	0.00292
Best cell in Group 5	2.3	22.58	0.6787	10.524	79.65	0.00274
Best cell in Group 6	2.3	22.60	0.6801	10.526	79.57	0.00284

** represents the Gaussian laser beam, * represents the top-hat laser beam, and AF represents the ablated area fraction. The deviation for E_ff, V_oc, I_sc, FF, or R_s was ±0.031, ±0.0002, ±0.008, ±0.11, ±0.0001, respectively.

shows the best cell ablated with a Gaussian beam or a top-hat beam, respectively. The improvements in the Voc and Isc values were seen in the best cell with a top-hat beam. The Voc of the best cell ablated with a top-hat beam was 1.4 mV higher than that of the best cell ablated with a Gaussian beam. The improvement in Voc was due to the impact of lower laser-induced damage.

Table 3 summarizes the average electrical data of the Gaussian beam and top-hat beam laser-processed cells. The average values were taken from at least 400 cells from the same batch. The average Eff of the top-hat beam laser-processed cells was 22.43%, which exceeded the one of the Gaussian beam-processed cells by approximately 0.03%. The higher average Eff, higher average Voc and higher Isc for the top-hat beam laser processed cells demonstrated the superior ablation characteristics of the top-hat beam laser dielectric ablation.

4. Conclusions

In this study, the improved ablation characteristics achieved by a top-hat beam ablation, combined with a rear polishing process, were used for the openings of the rear passivation stack of a PERC cell. The ablation threshold fluence of a Gaussian beam was higher than that of a top-hat beam, because its peak fluence was harmful and wasteful. The laser with a uniform top-hat beam had superior ablation characteristics and created low-damage dielectric openings with shallowly melted silicon. To determine the level of laser-induced damage in Si, after removing the ablated dielectric layers, a wafer was re-passivated with a new dielectric stack, and the sample applied with a top-hat laser beam had a 15.35% larger residual minor lifetime than the one ablated with a Gaussian beam at an ablated area fraction of 2%, which indicated the damage decreased in the sample ablated by a top-hat beam. We used a top-hat beam with optimized laser fluence and ablation pattern to fabricate PERC cells and bi-facial PERC cells. Average efficiencies of 22.523% from 400 pcs PERC cells and 22.43% from 400 pcs bi-facial PERC cells were obtained, which exceeded those of the reference cells by 0.05% and 0.03% in absolute value, respectively. The improvements were observed in the open-circuit voltage (Voc) and the short-circuit current (Isc) values, mainly because of the decreased laser-induced damage. In the future, top-hat laser ablation will have application prospects for other advanced solar cells. In the future, dot matrices instead of dashed lines and a picosecond laser will be used to enhance the performance of a top-hat ablation. The improved ablated surface quality has application prospects for other advanced solar cells, such as local openings of interdigitated back contact (IBC) cells.