Casey Kazmierowicz 1 , Bjorn Dahle 1 , Umesh Kumar 2 and Ed Graddy 2 1 KIC, 16120 Bernardo Center Dr., San Diego, CA 92127 2 Ferro Corporation, 1395 Aspen Way, Vista, CA 92081
Presented at the 35 th IEEE PVSC Conference June 20-25, 2010 Honolulu, Hawaii
Metallization is a critical processing step in Silicon solar cell manufacturing. Typically, cells are metalized by a rapid thermal cycle in an IR belt furnace. For proper yield, it is critical to maintain the wafer thermal profile within tight tolerances. In this paper, a user friendly profiling unit, namely, e-Clipse is introduced. The thermal profile of the wafer is measured with K-type thermocouples at four locations. The frame design allows the thermocouples to make intimate contact with the wafer without any bonding material. For profiling, the e-Clipse and the SunKIC datalogger are sent through the furnace to record the thermoelectric response. The recorded data is analyzed for key properties such as peak temperature, heating and cooling rates.
The metallization process is an important step in silicon solar cell manufacturing. During this step, maintaining the proper thermal profile of the wafer is very critical for high yield manufacturing of cells with good electrical performance (Ref.1-3). Thermocouples are commonly used to record the wafer thermal profile during the fast Infrared (IR) belt furnace process. At present, two different methods are used to attach the thermocouples to the wafer. In the first method, the thermocouple beads are pinned to the surface with a spring load. In this method, placement and contact inaccuracy introduce errors. In the second method, thermocouples are cemented to the wafer surface. While it improves the contact, thermal properties of the cement greatly influence the readout. As a variation, highly polished single crystal wafers with carefully cemented thermocouple beads are used as reference wafers. These wafers allow the users to maintain and correct the furnace profiles as a function of time. Yet, it does not allow the user to measure the surface thermal profiles of the wafers being processed, which may differ considerably from that of reference wafers.
Cementing the thermocouple to the wafer can be avoided with the e-Clipse, a new frame design from KIC (Fig. 1). The e-Clipse can accommodate wafers up to 160 mm x 160 mm size.
It is very difficult to measure the wafer surface temperature in an IR belt furnace (Ref. 4, 5). When thermocouples are used for the measurements, the recorded values differ from the actual temperature noticeably. In addition to the thermocouple placement procedure, two additional factors, namely, thermocouple time constant and thermal contact resistance affect the measurement accuracy (Ref 4, 5). In the e-Clipse, low mass thermocouple junctions are flattened to increase the contact area to minimize thermal contact resistance and to improve sensitivity. In the IR heating environment, thermal contact resistance is also affected by the radiation intensity and the actual wafer temperature (Ref. 4, 5). It is generally well known that measurement accuracy is valid only over a narrow temperature range.
Fig. 1. e-Clipse with SunKIC datalogger
In this paper, a procedure used to identify optimum TC junction geometry is described. Measurement accuracy and repeatability of the commercial units are also discussed.
Thermocouple Junction Design
As mentioned, in the rapid heating environment of an IR belt furnace, the temperature readings are greatly influenced by thermal contact resistance. Primary factors affecting the thermal contact resistance are contact area between the TC and the wafer, input power to the IR lamp, and the peak temperature (Ref. 4, 5).
To optimize the TC junction design, all of these factors were considered. At present, silicon solar cells are heated to approximately 800¡ãC during the metallization step (Ref. 3). The soak time at the peak is maintained below 5 sec.
To identify optimum design, test units with various shapes of TC junctions were made. In the test units, the contact surface areas of the thermocouples were varied from approximately 0.1 mm 2 to 16 mm 2 . In Fig. 2, one such test unit with two different junction shapes is shown.
Fig. 2. An example of test unit with 0.02-in. mineral insulated sheathed Type K thermocouples from Omega (bottom right) and flattened thermocouples (top left).
To identify the actual wafer temperature, the melting phenomena of Molybdenum Trioxide (MoO 3 ) was utilized. This compound melts at 795¡ãC. It does not react with the SiNx passivation layer which is on the emitter side of the silicon wafer.
The surface temperature of a 6-in. x 6-in. bare poly silicon wafer was recorded with the test units. A small spot (d<3 mm & t~10 ¦Ìm) of MoO 3 was applied on the wafer near one of the thermocouples. The measurements were carried out in a Despatch IR belt oven model number CF-7210. The zone set points for the six zone furnace were 400-400-500-700-800-880 and the belt speed was 200 ipm. These set points were selected to reflect the rapid firing conditions used in commercial scale during metallization process. Initially, the set point for the last zone was varied to identify the minimum required setting for MoO3 melting. Molybdenum oxide melting was verified with five repeats at these set points. Five additional measurements were conducted by reducing the set point of Z6 to 870¡ãC. In all cases, MoO3 did not melt.
In Fig. 3, the recorded profiles with three different thermocouple contact areas are compared.
When the TC junction area was >3 mm 2 , in general, the recorded peak temperature was closer to the actual wafer temperature.
Fig. 3. Comparison of three different TC junction types run on the same wafer,
through the same furnace, set to the same set points. The set point temperatures for zone 1
through zone 6 were 400-400-500-700-800-880. The belt speed was 200 ipm.
In Fig. 4, the recorded profiles of silicon wafer with back Aluminum paste registered with point and flattened TC contacts are compared.
Fig. 4. Effect of TC junction shape on sensitivity.
These measurements were conducted on a 6-in. X 6-in. polycrystalline silicon wafer at the same furnace set points. It is noticeable, that the flattened TC is sensitive enough to register signals related to Aluminum melting at 660¡ãC and Al-Si eutectic freezing at 577¡ãC.
Fig. 5. TC junctions in commercial unit.
After thorough data analysis, the TC junction design for the commercials units was finalized. In Fig. 5, a representative picture of the commercial unit is reproduced.
Fig. 6. Thermal profiles of bare and Aluminum metalized wafers. The furnace set points were 400-400-500-700-800-880 at 200 ipm.
In Fig. 6, representative profiles for bare and Aluminum metalized wafers recorded with a commercial unit are compared. The furnace set points for these measurements were the same as above. For these measurements also, 6-in. x 6-in. polycrystalline wafers were used. Aluminum metalized wafer registered approximately 45¡ãC lower peak temperature.
Measurement Accuracy and reproducibility
Two sets of data were collected to understand the measurement accuracy and repeatability. The first set of data was collected on bare and Aluminum metalized wafers at firing condition p1, namely, 400-400-500-700-800-x at the belt speed of 200 ipm. To understand the influence of longer soak time, the second set of data was collected at firing condition p3, namely, 400-400-500-x-x-x at the belt speed of 200 ipm. The set points for the last zones (x) were varied to identify the minimum required value for MoO 3 melting. For this measurement, the bare wafer weights of 10.5-10.6 gm and aluminum deposit weights of 1.5-1.6 gm were selected. For bare wafers, readings from five consecutive measurements were compared. On metalized wafers, six sets of data were collected. Data set 2-6 were used for the analysis.
In Fig.7, TC2 readings of the two profiles recorded with a bare wafer are compared. On the day of measurement, the minimum required set point to commence MoO 3 melting for p1 profile was 900¡ãC and for p3 profile was 870¡ãC.
Fig. 7. Thermal profiles of bare silicon wafer at two different set points. The set points for p1 are 400-400-500-500-700-800-900 at 200 ipm and for p3 are 400-400-500-870-870-870 at 200 ipm.
As seen, the thermocouples registered higher temperature when the soak time was longer.
Fig. 8. Measurement repeatability.
In Fig. 8, the recorded peak temperatures of TC2 of all the measurements are compared. Both variables, namely, presence of aluminum and the soak time at the peak temperature influence the peak temperature readout.
Discussions and Highlights
Formation of front electrical contacts during the metallization heat treatment process involves several physical phenomena such as binder removal, glass melting, SiNx etching, Si etching, Silver epitaxial film growth, etc. (Ref. 1-3, 7, 8). Some of these phenomena are sensitive to both temperature and the time at a specific temperature. In an IR furnace, all these phenomena occur at different parts of the heating and cooling segment but within a few seconds. To achieve the highest possible production yield, it is critical to maintain the entire profile as precisely as possible.
The actual temperature of the silicon wafer in an IR rapid thermal processing environment is influenced by several factors. The wafer temperature is a complex function of several factors such as IR absorption depth, thermal diffusivity, thermal mass, etc. (Ref. 6). IR absorption depth, in turn, is a function of factors such as Silicon wafer crystallanity, doping level, and the IR radiation wavelength. Thermal mass of the wafer is primarily a function of size and thickness of the wafer. (Ref. 6). In general, solar cell manufacturers metalize the wafers with various dimensions and doping levels.
The e-Clipse frame and SunKIC datalogger allow the users to test representative wafers from any manufacturing batch for the actual profile during the process to maintain quality and consistency.
In this paper, the salient properties of a user friendly thermal profiling unit for silicon solar cells during the metallization process are presented. A calibration procedure for 800¡ãC processing is described. Based on the analysis the following conclusions are drawn:
a. Commonly used sheathed thermocouples record significantly lower peak temperatures than flattened TCs.
b. Large surface area TC beads get closer to the correct peak temperature.
c. At the same furnace set points, peak surface temperature of bare and Aluminum metalized wafers differ, considerably.
d. Flattened TCs read closer to the true peak temperatures as demonstrated by the melting of spots of MoO3 .
e. The KIC e-Clipse demonstrated higher accuracy and improved repeatability compared to conventional profiling methods.
1. C. Ballif, D.M. Huljic, G. Willeke, and A. Hessler-Wyser, App. Phys. Letter, 82(12), (2003) 1878-1880.
2. C. Khadilkar, S. Sridharan, D. Gnizak, T. Pham, S. Kim and A. Shaikh, 20 th EC PV Solar Energy Conference, Barcelona, Spain (2005)
3. D. Neuhaus, A. Munzer, Advanced OptoElectronics, 2007 , Article ID 24521.
4. T. Borca-Tasciuc, D.A. Achimov, and G. Chen, Mat. Sci. Soc. Symp. Vol. 525, pp. 103-108 (1998).
5. Vandenabeele, W. Renken, Mat. Res. Soc. Symp. Proc. Vol. 525, p-109-114 (1998).
6. Ji Youn Lee, ¡°Rapid thermal Processing of Silicon Solar Cells,¡± Ph.D thesis, Fraunhofer Institute of solar energy system, (2003)
7. G. Schubert, F. Huster, and P. Fath, Proceedings of the 19 th European Photovoltaic Solar Energy Conference (EU PVSEC¡¯04), p. 813, June 2004.
8. M. M. Hilali, A. Rohatgi, and B. To, Proceedings of the 14 th Workshop on Crystalline Silicon solar Cells and Modules, 2004.