CIGS P1, P2, P3 Scribing Processes using a Pulse Programmable Industrial Fiber Laser

CIGS P1, P2, P3 Scribing Processes using a Pulse Programmable Industrial Fiber Laser M. Rekow 1 , R. Murison 1 , C. Dunsky 2 , C. Dinkel 3 , J. Pern 4 , L. Mansfield 4 , T. Panarello 1 , and S. Nikumb 3 1 PyroPhotonics Lasers, Inc. 275 Rue Kesmark, Dollard-des-Ormeaux, Québec H9B 3J1, Canada 2 Aeos Consulting, Inc. 72 Broadway, Los Gatos, CA 95030, USA 3 NRC Canada, Industrial Materials Institute-CAMM, 800 Collip Circle, London, Ontario NG6 4X8 Canada 4 National Center for Photovoltaics, National Renewable Energy Laboratory, 1617 Cole Boulevard, Golden, CO 80401, USA ABSTRACT: We describe a novel set of laser processes for the CIGS P1, P2 and P3 scribing steps, the development of which has been enabled by a unique pulse-programmable fiber laser. We find that the unique pulse control properties of this 1064 nm wavelength laser have significant effects on the material removal dynamics of the various film layers in the CIGS material system. In the case of the P2 and P3 processes, the shaped pulses create new laser/material interaction effects that permit the material to be cleanly and precisely removed with zero Heat Affected Zone (HAZ) at the edges of the scribe. The new P2 and P3 processes we describe demonstrate the first use of infrared nanosecond laser pulses that eliminate the HAZ and the consequent localized compositional changes in the CIGS absorber material that result in poor shunt resistance. SEM micrographs and EDX compositional scans are presented. For the P1 scribe, we process the bi-layer molybdenum from the film side as well as through the glass substrate. Microscopic inspection and compositional analysis of the scribe lines are not sufficient to determine electrical and optical performance in working PV modules. Therefore, to demonstrate the applicability of the infrared pulse-programmable laser to all three scribing processes for thin-film CIGS, we fabricate small- size multiple-cell monolithically interconnected mini-modules in partnership with the National Renewable Energy Laboratory (Golden, Colorado). A total of four mini-modules are produced, two utilizing all laser scribing, and two with the P2 and P3 steps mechanically scribed (by a third party) for reference. Mini-module performance data measured at NREL is presented, and we also discuss the commercialization potential of the new single-laser CIGS scribing process. Finally we present a phenomenological model to describe this physics underlying this novel ablation process. Keywords: CIGS, Thin Film Solar Cell, Laser Scribing, P2, P3 1 Background CIGS is rapidly gaining ground as preferred material system for thin-film photovoltaics (PV). Among other advantages commercial CIGS PV modules can boast 13% [1] or more efficiency, are relatively non-toxic and environmentally benign, and have a very stable performance over time and environmental exposure. Given that the laboratory world record for CIGS efficiency is now 20.1% [2] the future for CIGS looks bright. Amid these obvious advantages, until now CIGS- based thin film PV modules have defied the traditional so-called P2 and P3 laser scribing processes [3] that have proven so valuable in creating the monolithic series interconnect structures required to achieve useful working voltage and current in large scale thin-film amorphous silicon and CdTe PV modules. With the failure of laser based processes, mechanical scribing with a force-controlled stylus has become the method of choice. This method can suffer, however, from poor edge quality (e.g., delamination, chipping, and variations over time due to mechanical wear). More importantly, the non-deterministic nature of the material removal mechanism yields wide, irregular scribe lines that necessitate large spacing between adjacent scribes. As a result, the CIGS modules suffer decreased efficiency [4]. Clearly a viable laser process would provide a tangible benefit to the mass production of CIGS thin film PV. In addition to the P2 and P3 scribe processes, CIGS modules also require a P1 process. While there is a generally accepted laser process for the CIGS P1 scribe that utilizes 1064 nm wavelength and nanosecond pulses, until now the lasers used have proven completely ineffective for the P2 and P3 scribes. In this paper we first describe a novel set of laser processes for the CIGS P1, P2 and P3 scribing steps, the development of which has been enabled by a unique pulse-programmable fiber laser [5]. Secondly we document the construction and test of monolithically integrated mini-modules utilizing these processes. Finally we present a simple phenomenological model that describes the physics of this novel ablation process and discuss the commercialization potential of these processes. 2 Scribe Process Thin film solar cell laser scribing processes allow division of large solar modules into an array of smaller series interconnected cells on one monolithic substrate. These processes are integral to the cell fabrication process and must be performed at specific points in the manufacturing process to successfully form the monolithic series interconnected end product [3]. In a typical CIGS process, a laser is used to segment the first conductive layer (P1), typically Molybdenum (Mo), into adjacent, electrically isolated strips (or bands). Next, the CIGS absorber, CdS, and intrinsic ZnO (iZO) layers are deposited and in the P2 step, scribed down to the first conductive layer (Mo) with a slight offset from the underlying P1 scribe. Finally, another transparent conductive layer is added and scribed again (P3) with an offset from the previous P2 scribe. The area between the P1 and the P3 scribe becomes effectively inactive for the 25th European Photovoltaic Solar Energy Conference and Exhibition / 5th World Conference on Photovoltaic Energy Conversion, 6-10 September 2010, Valencia, Spain 2862 purpose of generating electricity, and minimizing this so called “dead zone” is a primary motivation to consider a laser tool over a mechanical scribe tool. It should be noted that while it is only necessary to scribe the top conductive layer to make the P3 scribe, in practice the P3 scribe is made down to the bottom conductor layer (Mo) making it nearly identical to the P2 scribe. 2.1 Laser Source The laser ultilized for this work was the 1064 nm PyroFlex™-25 pulse programmable fiber laser from PyroPhotonics [5]. This innovative new laser technology allows pulse duration to be varied from approximately 2 to several 100s of nanoseconds, independent of the laser repetition rate, which can be varied up to 500 kHz. In addition, each pulse can be arbitrarily programmed to generate a specific desired temporal profile of instantaneous laser power. Pulse trains comprised of these shaped pulses can be applied to the scribing process at high repetition rates. The same laser source and delivery optics was used to make all scribes. All process developement work and metrology was performed at NRC in London, Ontario and process development samples were provided by the National Renewable Energy Laboratory in Golden, Colorado. 2.2 P1 Process The P1 process is generally considered a known process and many groups have reported successful and robust scribing processes utilizing Q-switched 1064nm lasers [6,7]. However it was discovered that the particular and unique bilayer structure employed at NREL for Molybdenum deposition process [8] resisted successful processing with previously published approaches. As a result substantial effort was invested in both glass-side and film-side process development. In our experimental setup, the laser beam was expanded to the desired diameter, passed through a conventional variable attenuator device, and brought to focus on the test panel surface by a 100mm focal length lens. The sample was mounted on a 2-D linear scanning table to allow the laser beam to traverse its surface at variable speed. The scanning table was capable of reaching speeds of 800mm/s over a distance of ~6 inches. The lens was mounted on a vertical motion stage to enable rapid characterization of the depth-of-focus of the scribing process. The same basic configuration was utilized for all three (P1,P2, and P3) scribes. For film side laser processing it was found that under no conditions could a single-pulse process remove the entire film cleanly. However it was determined that the first layer of this bilayer Mo structure could easily and cleanly be removed with a single 15µJ square-shaped 10ns pulse as illustrated in Figure 1-left. However, increasing the pulse energy invariably caused eruption of the underlying layer leaving a high protruding ridge as shown in Figure 1-right. Figure 1: Wyko interferometric images of ablation in bilayer Molybdenum. Left: Clean ablation of the top Mo layer only, by a single laser pulse. Right: Clean ablation of top Mo layer with “eruption” of bottom Mo layer. It was observed that the top layer was removed in its solid form by some sort of brittle fracture mechanism driven by rapid stress build up, while the bottom layer would soften due to heating, releasing the residual built- up stress. For the bottom layer, the driving forces apparently switched from stress build-up to phase transition and gas pressure accumulation, with resulting plastic deformation and rupture. Based on these observations, we hypothesized that creating a double pulse shape with a pair of 15µJ, 10ns square pulses separated by a 275ns delay would permit sufficient cooling of the bottom layer between the pulses so that the stress build-up mechanism would again become dominant by the time the second pulse impinged. The result was clean brittle removal of both top and underlying layers. Figure 2 shows the resulting film side P1 groove and illustrates the laser double pulse configuration that achieved these results. More details of this process are given in another publication [9]. Figure 2: CIGS P1 film side process result utilizing a train of double pulses. Pulse 1 removes the first Mo layer and pulse 2 removes the second. The pulse pairs repeat at the trigger rate of the laser (PRF) and the trigger rate is set according to the sample motion speed such that adjacent laser “spots” have a slight overlap (30% typical) and form a continuous scribe line. ~ 60 µm 25th European Photovoltaic Solar Energy Conference and Exhibition / 5th World Conference on Photovoltaic Energy Conversion, 6-10 September 2010, Valencia, Spain 2863 Development of a substrate-side (glass-side) process for this film presented different challenges. Initial results on a small sample look very good using single square 10ns, 50µJ pulses, however processing larger samples revealed that the process often failed on large areas of the film. Further study revealed that there was a correlation between the orientation of cracks on the Mo film surface and the onset of failure of the process, as shown in Figure 3. We speculate that the cracks indicate that the stress state of the film strongly influences the ablation process. Figure 3: Orientation of crack pattern in area where initial P1 substrate-side ablation process failed. When crack pattern runs perpendicular to the scribe line the process begins to work. Where cracks have other orientations the ablation process fails. We utilized the broad tunability of the PyroFlex™-25 laser to survey the pulse parameter space in order to find a regime where the stress state of the film did not spoil the ablation process. To do this we first ran an experimental matrix of spot size versus energy and selected the most promising parameter zone. Next, the pulse duration was optimized and it was found that there was a very narrow window from about 8 to 12 ns over which the process produced a clean scribe using square pulses (Figure 4). Once again, the ablation mechanism was observed to be that of a brittle fracture. Figure 4: Ablation process behavior as a function of pulse duration. Clean ablation is only achieved in a narrow window from about 8ns to 12ns. With further optimization of these parameters, excellent scribe quality was observed, but there remained a noticeable „disc‟ marking of the glass by the laser which can be seen in Figure 4. Subsequent analysis revealed that this feature has no perceptible depth and there were no cracks or other overt defects left in the glass. We conclude that the “disc” marks the threshold at which there is a phase change in the Mo just before the onset of ablation. Furthermore the scribes showed excellent electrical isolation, beyond the limit of our measuring instruments. These optimized conditions, 250µJ 10ns square pulses (Figure 5), were later used to make the mini-modules. Figure 5: 200X optical image of optimized scribe process result. Maximum scribe width is about 90 µm. The ring in the center was found to have no significant depth and no cracks in the glass substrate were observed. 2.3 P2 Process Unlike the P1 process, there have not yet been any commercial implementations of a laser scribe for the P2 or the P3 process. Those processes closest to being accepted by industry generally use picosecond pulses near 500nm wavelength [4]. However these processes still result in some melting of the CIGS and a significant reduction in shunt resistance which is detrimental, particularly on the P3 scribe. Figure 6 shows a typical P3 scribe made with a ps laser. Figure 6: Typical P3 CIGS scribe with picosecond laser at 515 nm wavelength. (Used with permission.) In contrast with the existing state-of-the-art laser scribe processes, the CIGS P2 process that was developed in this work results in no edge melting and very little residue in the bottom of the groove. Figure 7 below shows SEM images of a P2 scribe that was performed with the PyroFlex™-25 pulse programmable fiber laser. Note the complete absence of melt at the edges of the scribe and the clean bottom of the scribe. 10ns, 250 µJ ~90µm 25th European Photovoltaic Solar Energy Conference and Exhibition / 5th World Conference on Photovoltaic Energy Conversion, 6-10 September 2010, Valencia, Spain 2864 Figure 7: SEM images of edge quality for the PyroFlex CIGS P2 process. SEM images of PyroFlex CIGS P2 scribes (a) top view showing cleanliness of groove and fractured edges, (b) close up of fractured edge indicating no melting of the CIGS absorber layer, and(c) angle view of edge scallop between adjacent laser pulses. In Figure 7 it is clear that there is some residue left at the bottom of the scribe. As seen in Figure 7(c), the surface texture in the trench bottom is suggestive of that which occurs when a mass of solid material is lifted from an underlying solid surface, with a fluid gel or viscous liquid film sandwiched between the two solid layers. This is suggestive of the underlying processes responsible for this ablation to be discussed later. For the P2 scribe, the material that remains at the bottom of the groove is of particular interest since the top contact layer (typically Aluminum doped ZnO or AZO [8]) must make the contact to the bottom layer in this groove to make the cell-to-cell series interconnect. Any residue left in the groove could negatively affect the contact resistance and therefore the overall series resistance. With this in mind EDX scans (Figure 8) were performed with an emphasis on determining the composition of the residue at the bottom of the groove. Figure 8: EDX analysis reveals excess selenium in the scribe trench and at the trench edges. The EDX analysis reveals that the trench bottom contains only trace amounts of Cu, In, and Ga. The Se levels, however, are comparable to those on the surface of the bulk CIGS film. While this analysis gives a good indication that the residue in the trench is primarily selenium it does not indicate what the electrical properties of the material may be. As will be discussed later in the section on the modules created with these scribe processes, there does not appear to be any significant increase in series resistance for laser scribe grooves vs mechanically scribed grooves. Also of interest is the process window over which the process behaves well in terms of pulse duration, pulse energy and focus depth. As part of this work we characterized these aspects of the process window thoroughly. We determined that nominal pulse duration was less than 5 ns and that both the rise and fall time of the pulse were critical parameters for achieving a robust result with a wide process window. The results of the pulse shape study are published in another work [9]. Figure 9 shows the process window in terms of pulse energy for a spot size of about 50 µm in diameter. The limit for a stable process runs from about 10 µJ to about 16 µJ which indicates a  25 % process window in energy. At lower energies the process simply fails to complete and at higher energies the amount of residue in the groove increases markedly and damage to the Mo layer in the form of pinholes and cracks appear. Figure 9: Robust process window with pulse energy. Robust process window extends from about 10 µJ to 16 µJ. Nominal scribe width is about 50µm, laser pulse rate = 20 kHz, and stage speed is 800 mm/s. In addition to the energy process window, a key parameter that indicates that a process will be robust in an industrial environment is the depth of focus (DOF). To characterize the DOF we configure the laser system for the nominal energy of about 12 µJ and adjust the focus position of the delivery lens until the process begins to degrade noticeably in both the positive and negative z directions. Figure 10 shows the laser scribes as the focusing lens is adjusted in 0.2 mm steps through focus. As can be clearly seen the quality of the scribe lines is very good and effectively unchanged over an entire 2 mm range of motion of the delivery lens. This is a good indicator that the process can likely be robust over very large panels where the panel surface flatness may be an issue [10]. This DOF also speaks to the excellent M2 value of the fiber laser, typically < 1.3. Furthermore the large DOF implies that a system design may not require autofocus optics for a robust process. GaGaInInCuCu SeSe MoMo GaGaInInCuCu SeSe MoMo (c) (a) (b) 25th European Photovoltaic Solar Energy Conference and Exhibition / 5th World Conference on Photovoltaic Energy Conversion, 6-10 September 2010, Valencia, Spain 2865 Figure 10: DOF of PyroFlex CIGS P2 process. Note that the process remains approximately unchanged over a 1 mm focal depth with a 100 mm focal length delivery lens. Finally the optimized scribing process was applied to the P2 step in fabrication of the mini-modules in collaboration with NREL. The picture in Figure 11 shows the resulting P2 scribe (right) next to the P1 scribe (left). More details with regard to the structure, fabrication and test of the resulting modules can be found in ref. [8, 11]. Also, it should be noted that the large scribe spacing (~210 µm) was chosen for consistency with typical mechanical scribing systems and is not limited by the laser process. Figure 11: Sample of P2 scribe with optimized process as it was applied to the NREL mini modules.