Laser-based Scribing of CIGS Solar Cell Material Structures for Monolithic Module Integration
By Osman A. Ghandour, edited by Riad Ghandour
Introduction
Seven years into my journey of providing laser source, system, and process consulting and contract services to our valued customers, I have decided to start sharing some compelling stories from my career in laser materials processing applications. One of those stories happened about fifteen years ago, and it still carries great relevance today in choosing a laser technology that provides an optimal, innovative solution for certain material processing applications. Namely, it was about CIGS (Copper Indium Gallium Selenide) solar structure scribing for monolithic module integration.
At the time, I had no previous experience in solar cell and module technologies, but I was brought on board because of the depth and breadth of my experience in laser materials processing, including laser sources, systems, and processes. I was asked to lead a team of solar technology experts, engineers, and scientists to find the optimal laser-based technology for performing the scribes known in the solar field as P1, P2, and P3 scribes— the scribes necessary for monolithic CIGS solar module integration.
Discussions
Figure 1. A schematic illustration of a cross-sectional view of a portion of a monolithically integrated module, including interconnected cells1.
The CIGS solar material was manufactured on a roll-to-roll stainless steel web. Once a laser scribing process was determined, the coated solar cell structure and its laser scribing would be represented schematically as shown in Figure 1. A stainless-steel web would be coated with an electrical insulator layer such as SiO₂ on one side. Then a molybdenum-based back contact layer would be coated on the insulator layer. A P1 scribe would be formed on the molybdenum layer, stopping at the insulator layer. After that, a CIGS layer would be deposited on the molybdenum layer. Then, a buffer layer (made of CdS — cadmium sulfide — for example) would be deposited on the CIGS layer. A P2 scribe would then be formed in the buffer and CIGS layers, stopping on the molybdenum layer. Subsequently, a TCO (transparent conducting oxide) layer would be deposited on the buffer layer. P3 would then be formed in the TCO/buffer/CIGS stack, stopping on the molybdenum layer. With that, the solar cells would be interconnected in series, and a web section containing a number of cells would form an array used in the monolithically integrated CIGS solar module.
What makes P1, P2, and P3 scribes optimal
The scribes have to be as narrow as possible. As shown in Figure 1, the total width of P1, P2, and P3 represents a shadowed area of the solar cell structure that has no current generation capability, reducing module efficiency.
Melting on the edges of the scribe can weld the solar material layers together, causing shunting that reduces efficiency.
Particles produced during scribing can interfere with the next coating steps, causing efficiency-reducing shunts.
The front contact TCO layer would be 100 to 1000 nm thick1. The semiconductor CIGS/buffer layer would be 500 to 3000 nm thick1. The molybdenum-based back contact layer would be 100 to 1000 nm thick1.
Figure 2. FIB SEM cross-sectional view of the untouched stainless-steel-based CIGS solar cell sample sent to the outside laser apps labs1. Front contact TCO layer: 690 nm. CIGS/buffer layer: 1600 nm. Back contact molybdenum layer: 540 nm
As the laser SME (subject matter expert) on the team, I led the effort to obtain solar structure samples — sections of the solar structure web — and send them to outside companies providing laser source and laser system products. The applications teams at these outside companies received instructions from me and my team on developing and optimizing the laser scribes and then went ahead and ran the process development and optimization. The laser source technologies utilized in this effort were nanosecond, picosecond, and femtosecond lasers. We specifically asked for an optimal P3 scribe with a detailed report on the process development and optimization.
When we received the samples back, we asked our analytical lab team to perform top-view SEM (Scanning Electron Microscope) microscopy, cross-sectional FIB (Focused Ion Beam) SEM microscopy, and EDS (Energy Dispersive Spectroscopy) elemental mapping to measure and examine the scribes. Figure 2 shows a cross-sectional FIB SEM of a typical untouched solar sample similar to the samples we sent out1. In this sample, the front contact TCO layer was about 690 nm thick1, the CIGS/buffer layer was about 1600 nm thick1, and the molybdenum-based back contact layer was about 540 nm thick1.
Table 1 Data on P3 scribes
P3 scribes on the SST/Mo/CIGS/buffer/TCO solar structure samples were performed at the outside laser apps labs using 355 nm, 532 nm, and 1064 nm lasers, with pulse widths in the nanosecond, picosecond, and femtosecond ranges. The total thickness of the CIGS/buffer/TCO layers — the material to be removed in the P3 scribe — was about 2830 nm1, with the CIGS layer alone being about 1600 nm thick1. For each laser wavelength and pulse width, laser fluence, energy, scan speed, and number of passes were optimized based on scribe width, consistency, depth, and cleanliness. Depth and cleanliness of the scribes were assessed using EDS elemental mapping. The scribes were also closely examined for evidence of melting by top-view SEM and cross-sectional FIB/SEM. Based on these criteria, data for the various scribes were organized in Table 11, showing the conditions of each scribe and whether or not the scribe had passed.
Laser scribes that successfully scribed down to the molybdenum layer were deemed to have passed1. Only the 1064 nm (scribe nos. 1 and 12) and 532 nm (scribe nos. 13, 15, 16, 17, and 18) lasers successfully scribed down to the molybdenum layer1. Of these, only the picosecond 1064 nm scribe lines displayed no evidence of melting1. Evidence of melting was most pronounced at 355 nm, followed by 532 nm1. The laser scribe at 1064 nm in the picosecond range produced very clean lines in the TCO layer, showing no evidence of melting1. P3 scribes were also performed using a 1064 nm nanosecond laser; on those scribes, cracks and melting residues were observed1.
Figure 3. 355 nm nanosecond laser P3 scribe1. Width: ~50 µm. Notice the strong melt splatter on the scribe edges and within the scribe. (a) Top-view SEM with EDS elemental map. (b) Top-view SEM.
Figure 3 shows the EDS elemental maps and SEM top views of a P3 scribe whose process was developed and optimized using a 355 nm nanosecond laser1. The scribe had excessive melting associated with it, and its width was about 50 µm. Minimizing the complete scribe width is important in order to minimize shadowing across the three scribes (P1, P2, and P3), which would otherwise negatively impact efficiency. The molybdenum signal was strong, indicating that the scribe reached the molybdenum layer; however, because of the melt splatter, the molybdenum signal within the scribe was not consistent. This scribe was considered a failure.
Figure 4. 1064 nm picosecond laser P3 scribe1. Width: ~45 µm. Notice the TCO fracture on the edge of the scribe, indicating that the scribe mechanism was mechanical. (a) Top-view SEM with EDS elemental map. (b) Cross-sectional SEM view
Figure 4 shows the EDS elemental map along with SEM top-view and cross-sectional view microscopy of a P3 scribe whose process was developed and optimized using a 1064 nm picosecond laser1. The scribe was virtually free of melting, with a width of about 45 µm. The molybdenum signal was strong, indicating that the scribe depth reached all the way down to the molybdenum layer. The cross-sectional SEM showed that the scribe had fractures on the edges, indicating that the removal mechanism was mechanical. The speed for this scribe was up to 4 m/sec.
Figure 5. 1030 nm femtosecond laser P3 scribe1. Width: ~34 µm. Notice the slight CIGS melting on the edge of the scribe, indicating that the scribe mechanism was direct vaporization with some slight melt on the edge in a transition region less than 1 µm wide. (a) Top-view SEM with EDS elemental map. (b) Cross-sectional SEM view.
Figure 5 shows the EDS elemental map along with SEM top-view and cross-sectional view microscopy of a scribe whose process was developed and optimized using a 1030 nm femtosecond laser. The width of the scribe was about 34 µm. The scribe was complete, with a strong and consistent molybdenum signal. The cross-sectional SEM showed slight melting along the scribe edge in a transition region less than 1 µm in width. The scribe speed was less than 1 mm/sec, making it economically nonviable. The slow speed and the presence of melting meant that this femtosecond scribe was a failure. Data for this scribe was not included in Table 11.
The femtosecond laser scribe had a minimal heat-affected zone (HAZ), but it still showed some melting. For the femtosecond laser at a reasonable spot size — comparable to that of the picosecond laser — the peak intensity is high enough to induce nonlinear absorption, rendering the removal mechanism direct vaporization across most of the scribe with a tiny (<1 µm) melt-dominated transition region on the edges. The picosecond laser had a larger HAZ and might have been expected to produce some melting, but its material removal was mechanical in nature, removing material at the scribe and only slightly outside of it.
As shown in Figure 4, the 1064 nm laser provided melt-free and particle-free scribing that passed the scribe quality criteria1. In conventional laser scribe processes, film is removed by a mechanism that includes melting: laser pulse energy is absorbed, leading to a temperature increase, which in turn leads to thermal expansion and melting. Laser energy on the molten material causes splatter, which, when solidified, results in particles welded to other layers1. The solidified melt pools and particles can be a major cause of shunt formation between front and back contacts in a cell, which is detrimental to device performance. This is the case with nanosecond lasers. The femtosecond laser, because of its extremely high peak intensity, drives energy absorption via nonlinear absorption, producing direct vaporization across most of the scribe with slight melting in a tiny ~1 µm transition region on the edges.
The 1064 nm picosecond laser successfully scribed the solar structure with virtually no melting or particle formation1. Using ultra-short picosecond pulse laser beams of the right wavelength, P1, P2, and P3 scribes in CIGS solar structures were produced without the melting or particles associated with typical scribing processes. Specifically, at 1064 nm and in the picosecond range (12–15 picoseconds), the optical penetration depth is on the order of, or greater than, the overlying film thickness. Once the laser beam passes through the overlying layer and is absorbed by the layer beneath, a strong ablation shockwave is produced in the underlying layer, resulting in mechanical shock removal of the overlying layer with virtually no melting or particles — as shown in the P3 scribe in Figure 4. P1 scribing of molybdenum coatings on a flexible SiO₂ layer over a stainless-steel web using a 1064 nm picosecond pulse would similarly produce scribe lines with virtually no melting or particles. This was determined based on using this type of laser to scribe molybdenum layers similar to those in the CIGS solar structure described here on glass plates1.
It is believed that the improved scribing is due to producing the scribes via a shockwave generated at the interface of the layer to be removed and the underlying layer. This mechanism is only possible by selecting the right wavelength and peak intensity. The right wavelength yields an optical penetration depth on the order of, or greater than, the thickness of the overlayer to be removed. The right peak intensity yields strong ablation of the underlying layer without onsetting nonlinear absorption. Together, the right wavelength and peak intensity produce overlying-layer removal dominated by mechanical action — without the melting seen with the nanosecond laser1, and without the direct vaporization with slight melting seen with the femtosecond laser1.
To produce an ablative shockwave that removes a top layer without melting, the following conditions must be met1:
The laser beam has a wavelength that provides an optical penetration depth on the order of, or greater than, the thickness of the top layer.
The absorption coefficient of the underlying layer at that wavelength is greater than a threshold minimum necessary for the underlying layer to absorb the energy and allow a high enough absorbed peak intensity to produce the shockwave.
The pulse width is small enough to provide the requisite peak intensity to produce the shockwave, but not so small that it results in nonlinear absorption in the top overlying layer.
Optical penetration depths for various example solar cell layers and laser beam wavelengths are provided in Table 21. The optical penetration depth is calculated as:
Optical penetration depth = 1/α = λ / (4·π·k)1
where α is the absorption coefficient, λ is the wavelength, and k is the extinction coefficient for the material at that wavelength.
Table 2 presents examples of optical penetration depths for typical solar cell layers at various wavelengths1. Determining a laser beam wavelength that can produce an ablative shockwave may be done with reference to Table 2 or by calculating the optical penetration depth for a particular material and wavelength. For example, to remove a CIGS layer of 1000 nm, Table 2 shows that wavelengths of 1000 nm or 1040 nm may be used 1(both having optical penetration depths greater than 1000 nm).
Table 2. Optical penetration depths of various solar cell layers
Multiple layers may be scribed simultaneously if they are transparent at the laser beam wavelength and the optical penetration depth exceeds the stack thickness1. For example, a TCO/buffer/absorber stack is P3-scribed with a 1064 nm laser, leaving the underlying back contact layer intact. For the purposes of this description, a "top layer" or "overlayer" may include multiple layers that are optically transparent at a particular wavelength. Optical transparency may be characterized by the extinction coefficient k, which characterizes absorption of electromagnetic energy by a wave propagating through a material1. As indicated above, optical penetration depth is 1/α = λ/(4·π·k). Transparent materials are characterized by k close to zero and a correspondingly high optical penetration depth.
While an optical penetration depth large enough to act on the underlying layer is a prerequisite for the ablative shockwave removal mechanism illustrated in Figure 4, the peak intensity of the laser pulse must also be appropriately chosen to produce the ablative shockwave. Peak intensity is inversely correlated to pulse width:
Peak intensity = Peak Power / Spot area
Peak power = Energy / (full-width half-max pulse width)
This represents one way to define peak power. Base-to-base pulse width or 90% pulse width are examples of other definitions that can be used in the peak power formula1.
Above a threshold pulse width for a particular laser energy and spot size, the peak intensity is not high enough to generate a shockwave. Scribing a CIGS/buffer/TCO layer with a total thickness on the order of 1000 nm at a wavelength of 1064 nm was attempted using pulse widths in the nanosecond (10⁻⁹ s) and picosecond (10⁻¹² s) ranges. Descriptions of pulse widths in these units include tens of each unit — e.g., tens of picoseconds. It was found that even with a wavelength of 1064 nm (and thus the requisite optical penetration depth), melt-free scribing was obtained only with pulse widths in the picosecond range. . In the nanosecond range, the peak intensity of the pulse is not high enough to produce the strong shockwave. If the pulse width is decreased too much — e.g., into the femtosecond range — nonlinear absorption in the overlying layer occurs due to the extremely high peak intensities. Due to that strong absorption, the optical penetration depth does not extend beyond the CIGS layer to produce the same mechanism.
Summary
Although this work took place about 15 years ago, the laser sources evaluated at that time remain highly relevant in today's laser industry.
One of the most critical requirements for the P1, P2, and P3 scribes used in monolithic CIGS solar module integration is the absence of melting. Nanosecond lasers were ruled out early on due to excessive melting and thermal damage. From there, one might intuitively expect that femtosecond lasers — known for their minimal heat-affected zones (HAZ) — would be the ideal choice. However, it's important not to make assumptions when it comes to laser applications. Instead, each application must be thoroughly investigated through systematic experimentation, including, when necessary, collaboration with external laser applications laboratories.
In this particular case, the best-quality scribe — completely free of melting and particles, while maintaining a reasonably narrow scribe width of approximately 45 µm — was surprisingly achieved using a 1064 nm picosecond laser, rather than a 1030 nm femtosecond laser. The 1064 nm picosecond laser also proved to be the most economically viable option, supporting scan speeds of up to several meters per second, possibly reaching as high as 4000 mm/sec.
According to this study, the optimal scribe quality was achieved through a laser-induced mechanical removal mechanism. At 1064 nm, the optical penetration depth was comparable to or greater than the thickness of the layer to be removed. This allowed the laser energy to pass through the weakly absorbing or even transparent top layer and interact with the absorbing layer underneath. The laser's peak intensity was sufficient to generate a strong ablation shockwave at the interface between these layers. This resulted in a "shock-off" effect — mechanically ejecting the top layer without significant thermal damage. The resulting scribe was relatively narrow and virtually free of melting and particles.
By contrast, the femtosecond laser — though operating with extremely short pulse durations — produced peak intensities that were too high, leading to nonlinear absorption in the top layer. The primary material removal mechanism became direct vaporization, which, while effective, also introduced some melting. This melting was observed in a very narrow transition region at the scribe edges, approximately 1 µm in width. As a result, the femtosecond laser scribes exhibited slight but measurable thermal effects.
