Solar Cells Texturing Process Monitoring HF/HNO3/H2SiF6, KOH/IPA and NaOH/IPA Introduction Texturing is an important step in the manufacturing of a Silicon solar cell, both mono and polycrystalline. This process removes the surface damage induced by wafer sawing. It also serves to reduce the reflectance of the wafers, and incorporate light trapping in the cell (especially of longer wavelength photons) by the creation of small bumps on the surface, thus increasing cell efficiency. Monocrystalline Silicon texturing is typically performed using an aqueous solution of KOH or NaOH plus IPA or a surfactant. The anisotropic etching (the <100> and <111> oriented planes having significantly different etch rates) results in a surface that is covered with pyramids. The KOH/NaOH is consumed when wafers are etched and needs to be maintained at a steady level for constant etch rate and increased bath life. The anisotropic etch technique is not suitable for polycrystalline Si texturing, thus isotropic etchants are used, sometimes in conjunction with other techniques. A typical chemistry for this purpose is a high concentration HF/HNO3 solution. Both HF and HNO3 are consumed during the process and need to be spiked in order to maintain a constant etch rate and achieve long bath life. Too low a concentration will mean under etching, while a high concentration will result in over etching and scrapped wafers. This application note describes the use of the WetSpec200 in line process analyzer for monitoring texturing processes used in manufacturing of PV cells. Results from field installations that monitor inline processes are described, in particular polycrystalline Si texturing by HF/HNO3 and monocrystalline Si texturing by KOH/IPA and NaOH/IPA. In all cases, calibration models were first developed in CI Semi s applications laboratory and were later optimized using field data. WetSpec Product Description An In line Wet Chemistry Analyzer The WetSpec measures the absorption spectrum of a fluid sample in the near infrared wavelength range. Light from a halogen lamp is sent through optical fibers to a Teflon flow cell with sapphire windows, after which it is sent back through optical fibers to a grating based spectrometer. The spectrum is measured simultaneously by an array of photodetectors. CI Proprietary 1
Figure 1 WetSpec system connected to a flow cell A proprietary fiber optic multiplexer allows for the use of up to eight flow cells, improving the cost efficiency of the instruments. Up to eight separate and different process streams (each with multicomponent chemistry) can be measured by the analyzer. Optical fibers of up to 200 meters long connect the flow cells to the analyzer, thus allowing one instrument to measure baths/streams in different tools throughout the facility. Flow cells used for these applications have wetted parts of Teflon, Sapphire and Kalrez (O rings). These are available in different diameters and with different fittings to match the tools on which they are installed. An algorithm relates the spectral data measured by the analyzer to changes in concentration, creating a calibration model, which is then used for determining the concentration of unknown solutions in real time. CI Semi can supply models for different chemistries, different concentration ranges and different temperature ranges. In some cases real time temperature compensation is utilized, using the input from a temperature sensor embedded in the flow cell. The system is housed in a 2U 19 inch rack mounted unit and includes Ethernet and RS232 communication, 4 20mA analog outputs and digital outputs that can be used to set alarms in case of excursions. A graphical user interface is provided that allows monitoring the processes from remote computers directly or through the user s local network. Figure 2 (below) shows a typical configuration of a WetSpec installation. CI Proprietary 2
Figure 2 Typical configuration of a five channel system Case Study 1: Texturing of polycrystalline Si HF/HNO 3 /H 2 SiF 6 Texturing of polycrystalline Si using HF/HNO3 is, in some cases, performed below room temperature, typically at 6 12 C. In order to avoid moisture buildup on the cell windows, the flow cell is installed in a mini environment chamber with an input of Clean Dry Air (CDA) (Figure 3). Figure 3 flow cell installed in CDA box An initial calibration model was built for this application in CI Semi s chemistry laboratory. Process etch products were simulated in the lab by dissolving bare silicon in the bath. The H 2 SiF 6 content was calculated from the weight of the dissolved Si. Models that were built in the laboratory were a good starting point and predicted the trends very well; however, in order to improve the models to CI Proprietary 3
better fit field installation, data from an installation at a solar fab was collected. Titration results were available for all three components, including the etch product. Titration data in the fab matched the HNO3 lab model well, and that model has since been in use. Modifications were introduced to the HF and H2SiF6 models. Real time temperature compensation was also applied, as the process does operate in a rather wide temperature range and the spectra is temperature dependent. The WetSpec readings over a period of time of 3 weeks are shown in Figures 4a & 4b, as compared to titration results. Figure 4a WetSpec reading of HNO3 compared to titration Figure 4b WetSpec reading of HF and H2SiF6 compared to titration One of the main advantages of in line NIR spectroscopy over titration is its fast response time. This is essential when attempting to achieve closed loop control of bath composition, but also very important for detecting excursions rapidly. Figure 5 shows real time measurement data an HF valve was stuck open, resulting in continuous addition of HF and a much higher HF level than required. Wafers were over etched and scrapped. Had the WetSpec been used as an alarm in this case, the excursion would have been detected much earlier and most of the damage would have been avoided. CI Proprietary 4
Valve open, climbing stuck HF Problem identified, part of bath dumped and replenished Normal level of HF for stage in bath life Figure 5 HF excursion detected by the WetSpec Case Study 2: Texturing of Monocrystalline Si KOH/IPA Texturing of monocrystalline Si using a KOH/IPA aqueous solution is typically performed at elevated temperatures. The results shown in this section were obtained on tools running the process at 80 C. The KOH is the etchant and is consumed during the process, creating a residue of silicates. The added IPA improves the surface morphology. The IPA will tend to evaporate faster than the other components and its concentration will therefore drop with time, regardless of the number of processed wafers. Monitoring results from two installations are shown. Figures 6a and 6b show readings from one installation. Figure 6a shows three spikes of KOH and IPA and figure 6b shows a full cycle. CI Proprietary 5
Figure 6a Three spikes of KOH and IPA during a production run Figure 6b Full cycle of KOH and IPA during a production run After each run the bath is spiked with IPA and KOH; the spikes are clearly measured by the WetSpec200. Using the WetSpec200 s continuous measurements it can be seen that the KOH is over spiked and the IPA under spiked. Smaller spikes of KOH and larger spikes of IPA would have provided better stability of the process. Results from a second installation are shown in figure 7. This shows a test run, where known quantities of KOH and IPA were added to a known quantity of water. The test was conducted in a controlled environment, to avoid losses of IPA and without processing wafers. WetSpec readings are compared to the as mixed values. The spikes seen in the graph occur after spiking, before complete mixing is achieved and as such are real readings of the material that passes through the flow cell. CI Proprietary 6
Figure 7 KOH and IPA Preliminary results also show the ability to measure the concentration of silicate residues in KOH/IPA accurately. However, not enough reference data were available to include those data at this point. Case Study 3: Texturing of monocrystaline Si NaOH/IPA Figures 8a and 8b show results from a NaOH/IPA installation. Figure 8a shows two spikes, while 8b shows a full cycle. Figure 8a Two spikes of NaOH/IPA. Figure 8b Full cycle of NaOH/IPA CI Proprietary 7
In this example a full process cycle spans approximately 12 hours. Out of every 12 hours, more than 1 hour is wasted on bath replacement. The IPA and the NaOH are spiked together automatically every 20 30 minutes. The WetSpec200 is able to measure the small spikes of NaOH (0.08wt%) and IPA (0.3wt%). The WetSpec200 also measures the increase in Na 2 SiO 3 wt% over a full bath cycle. The long term goal in this case is to prolong bath life time, and increase up time of the tool. The WetSpec200 results suggest that increasing the IPA spikes and reducing the NaOH spikes, will help stabilize the process. Conclusions The WetSpec has demonstrated the ability to accurately monitor, in real time, chemical concentrations for the most commonly used solar cell texturing chemistries. This ability can be used in closed loop control for spiking chemicals consumed during processing and to increase bath life. The system has also shown its effectiveness in detecting real time excursions in chemical concentrations resulting from hardware failure. Early detection can save wafer scrapping and keep tool utilization at an optimal level. The WetSpec is capable of measuring chemical concentrations in a wide variety of applications with good accuracy. The instrument s multiplexing ability makes it an extremely cost effective means of monitoring chemical concentration in tools with multiple baths or in several independent tools. CI Proprietary 8