OPTIMIZING THE FRONT SIDE METALLIZATION PROCESS USING THE CORESCAN A.S.H. van der Heide, J.H. Bultman, J. Hoornstra, A. Schönecker, G.P. Wyers, W.C. Sinke ECN Solar Energy, P.O. Box 1, 1755 ZG Petten, The Netherlands ABSTRACT In solar cell manufacturing it is difficult to optimize screen printed front side metallization, in particular to obtain good contact over the entire cell. The Corescan enables mapping of contact resistance of full cell surfaces. Using these mappings, it is possible to diagnose and exclude reasons for poor contact formation. In this study, several process conditions have been studied. Cross-belt temperature non-uniformity in the firing furnace turns out to be a reason for large contact resistance differences. This shows that the acceptable firing temperature range can be considerably increased if a constant temperature profile across the belt width is realized. Further, it is shown that plasma etching for isolation gives cause for an increase of the contact resistance of the fingers at the cell edges. Finally, contact resistance differences related to emitter non-uniformity are found. To summarize, the Corescan is a powerful tool that enables easy optimization of front side metallization. far, only the Transmission Line Method (TLM) was available. This technique, however, is slow and it is not able to give information as a function of position. The technique is even based on the assumption of a positionindependent contact resistance. In contrast to the TLM method, the Corescan does not have these limitations. With this instrument (see Fig. 1), the contact resistance is locally measured, and a complete scan for a 10 cm x 10 cm cell takes only 6 minutes. The Corescan is commercially available from ECN's daughter company SunLab [4]. INTRODUCTION To reach optimum solar cell efficiencies, it is necessary to reduce losses to a minimum. One type of loss is the resistance loss, which is subdivided in series resistance and parallel resistance losses. This contribution focuses on the localization and reduction of the most problematic series resistance contributor in industrial solar cells: the contact resistance of the front side metallization. The localization is done with the recently introduced Corescan [1,2,3], derived from COntact REsistance scan. For approximately 90% of solar cells produced in industry, metallization is applied by screen printing of metal paste, followed by electrical and mechanical contact formation performed by sintering at high temperature in a belt furnace. Important parameters determining the quality of the contact formation are e.g. screen print parameters like print pressure, paste rheology and screen quality, belt furnace parameters like temperature profile, uniformity and stability, but also emitter sheet resistance and anti-reflection coating thickness. Although fill factors around 78% are achieved in the laboratory, fill factors in industry are regularly found to be 5-10% lower due to the process sensitivity of the front side contact formation. To reduce the contact resistance, it is important to be able to measure it, preferably as a function of position. So Fig. 1. Picture of Corescan By varying cell processing conditions and measuring the resulting contact resistance distributions, it is now easy to determine the reason for not optimally contacted cells, and realize process optimization. Parameters investigated in this paper are firing furnace temperature settings and uniformity, the influence of plasma etching for edge isolation and emitter non-uniformity. Screen print influences have been reported in [2]. THE CORESCAN In general, the Corescan is an instrument to determine where resistance loss in a solar cell occurs, for both parallel and series resistance. The patented methods [5] used by the Corescan are based on the analysis of the potential distribution on a cell while a current flow is induced in the cell. For parallel resistance (shunt locating) the cell is slightly forward biased in the dark. For contact resistance the cell is short-circuited and a local illumination is applied around the potential probe. The shunt method is described in detail in [1] and is compared with lock-in thermography in [6]. As contact resistance is the main subject of this paper, shunt resistance will not further be discussed in this paper. To compare the conventional TLM method with the Corescan, the TLM method is explained in more detail here. The TLM method was first published in the sixties and is still used today. For more information see e.g. [7]. The contact resistance is proportional to the potential difference Vce between the edge of the finger and the silicon adjacent to the finger. To be more precise, the contact resistance for fingers is best expressed in terms of the line contact resistance Rcl, which is defined as R cl ≡ Fig. 2. The Transmission Line Method (TLM) In the method (see Fig. 2) the resistance R is measured between one finger and the next ones, and a graph of R against distance is constructed. The data points will be located on a straight line, assuming that the emitter sheet resistance is constant and that the contact resistance is the same for each finger. The interception of this line with the R axis is proportional to the contact resistance. The busbar must be disconnected from the fingers to avoid parallel conduction. So, in practice small samples have to be cut from a full cell. Main disadvantages of the TLM method are that it only works when the contact resistance is the same for all fingers (which is often not the case), that it only reveals information on a limited cell surface and, that the method is rather time consuming. The Corescan method for contact resistance is based on the measurement of the potential jump between a finger and the adjacent silicon while current flows through the cell (see Fig. 3). This current is generated locally with a light beam, while the cell is short-circuited externally. A probe, centered in the light beam detects the potential at the front surface and is displaced together with the beam over the cell. The probe is pressed against the surface by the probe holder weight and scratches through the anti-reflection coating to enable electrical contact with the silicon surface. Fig. 3. Corescan in contact resistance mode Vce , J sc d (1) where Jsc is the local short circuit current density generated by the beam and d is the distance between the fingers. There are several reasons to prefer Rcl above the conventionally used specific contact resistance ρc in case of screen printed fingers. In the first place, the contact interface is very inhomogeneous, secondly the sheet resistance of the silicon below the fingers is not known. Therefore, the model to calculate the current pattern below a contact collecting lateral currents does not apply and it is impossible to calculate ρc correctly (in fact, we cannot even speak of a single ρc value). It is better to incorporate the entire situation below the contact in a single value, as is accomplished by using the Rcl definition given above. This value can be directly calculated from the Corescan data: Vce can be determined from the measured potential data, while the light intensity is adjusted before each scan in such way that Jsc has a predetermined value. Fig. 4. Corescan (part of one scan line) An example scan performed at Jsc = 30 mA/cm2 (see Fig. 4) shows the following typical features of a Corescan: 1. Fingers are at the lowest potential because they are in contact with the grounded busbars, 2. Between the fingers the potential has a parabolic shape due to emitter sheet resistance, 3. Potential jumps proportional to the contact resistance occur at the finger edges where the probe moves from metal to the adjacent silicon, 4. A finger interruption causes a 4 times higher parabola due to effectively doubled finger spacing. Although line scans are instructive, complete cell scans must be presented in 2D/3D graphs; examples of 2D graphs will be shown in the next section. Typical advantages of the Corescan method are its speed (typically 6 min for a 10 cm x 10 cm cell), the full cell surface scanning, and the automation of the measurement. METALLIZATION PROCESS OPTIMIZATION To demonstrate the application of the Corescan for contact formation optimization, example scans on cells with SiN anti-reflection coating and screen printed contacts are given in this section. In the first example, cells were IR fired at the same belt speed but with different furnace temperature settings. The temperature setting ranged from T-45 °C to T+30 °C. The Corescan results are given in Fig. 6. The measured Fill Factor is indicated, and also a picture of the aluminum backside at T+45 °C is shown. FF 72 %, T-45 °C FF 74 %, T-30 °C FF 74%, T-15 °C FF 75 %, T °C In this example, maximum FF is reached at temperature T. The degradation in FF for the other temperature settings is caused by increased contact resistance losses. The first observation is that it is possible to deduce from the contact resistance distribution whether a cell is over-fired or under-fired. This enables easy optimization of the temperature setting. The second observation is the brownish color (or darker gray in B&W) on the left and right sides of the cell. This reveals that the aluminum backside reached a much higher temperature on the sides than in the center. The similarity between the color pattern of the aluminum backside and the contact resistance figure for the same temperature is striking. The higher temperature apparently results in a strongly increased contact resistance. Comparing the color of the aluminum backside of different cells, the temperature difference over one cell was estimated to be 45±10 °C. Work is underway to investigate the cross-belt temperature non-uniformity in the furnace. A large improvement could be gained, if the temperature difference could be reduced to a few degrees, and especially for wide belts. This would increase the firing window by tens of degrees. Another example of temperature non-uniformity is shown in Fig. 5. Fig. 5. Corescan on a cell supported at 4 points during firing. FF 74 %, T+15 °C FF 69 %, T+30 °C FF 62 %, T+45 °C T+45 °C, backside Fig. 6. Corescan potential mappings on cells fired at different set temperatures and a picture of the Al backside for the last cell. Lighter areas have higher contact resistance. Four regular oriented spots of poor contact are seen in the center of a specially prepared test cell. Also the area between the four spots shows enhanced contact resistance as compared to the outside areas of the cell. Due to the presence of supports placed underneath the cell during firing the temperature distribution was influenced, and as a consequence non-uniformly distributed. The four supports are clearly revealed in the scan by a higher contact resistance. In Fig. 7 the Corescan shows a narrow band of high contact resistance along the edges of the cell (the fact that there are no high potentials visible at the lower edge is due to that the probe scanned exactly over the outer metallization line). The plasma etching of the edges as used for cell isolation causes the high contact resistance. Fig. 7. Increased contact resistance of the fingers at the cell edges caused by plasma etching used for edge isolation. The last example shows contact resistance nonuniformity caused by a non-uniform emitter doping. One of the methods to apply e.g. phosphorous dopant on the solar cell is by spinning. In this example, a cell on both sides doped by spinning has higher contact resistance at the rim of a circle with the same diameter as the spinning head (see Fig. 8.). accomplish, because the TLM method is time-consuming and assumes a uniform contact resistance. The latter is often an invalid assumption as shown by the examples in this paper. From the given examples, the influence of the individual process steps on the contact resistance is clearly shown. The firing temperature and non-uniformity of the temperature distribution in the furnace appear to have a large impact on the contact resistance. This leads to the important conclusion that the range of acceptable firing temperature settings increases by tens of degrees, if a uniform cross-belt temperature is realized. The other examples demonstrate that also other process steps, such as plasma etching of cell edges for isolation, and non-uniform emitter doping, can locally increase the contact resistance. The optimization of the front side metallization process is much easier when using the Corescan. It is no longer necessary to optimize without knowledge of the contact resistance distribution, which is often the key to find the reason for poor contact formation. REFERENCES [1] A.S.H. van der Heide, A. Schönecker, G.P. Wyers and th W.C. Sinke, Proceedings 16 European Photovoltaic Solar Energy Conference, Glasgow (United Kingdom), 1438 (2000) [2] A.S.H. van der Heide, J.H. Bultman, J. Hoornstra and th A. Schönecker, 11 NREL workshop on crystalline silicon solar cell materials and processes, Estes Park (Colorado, USA), 293 (2001) Fig. 8. Increased contact resistance in a circular pattern caused by the spinning head used for spinning of dopant fluid. Apparently, some of the dopant material has been removed from the front side when placing the cell upside down to spin the cell backside, resulting in lower doping and a higher contact resistance. In addition to the circle, there is a high contact resistance region in the center. This is also caused by the furnace temperature nonuniformity discussed earlier. At the upper part of the figure the higher potential is not caused by increased contact resistance, but by finger interruptions. Other methods of emitter doping can also lead to non-uniform sheet resistance, for example the doping level on cells made in tube furnaces usually decreases to the cell center due to a decreasing gas concentration. Whether these doping differences are so large that the contact resistance is influenced can be easily verified by performing a Corescan. CONCLUSIONS With the Corescan, that enables mapping of contact resistance on full cell surfaces, a new tool is available to optimize the front side metallization. With the traditional TLM method, optimization is difficult or even impossible to [3] A.S.H. van der Heide, J.H. Bultman, J. Hoornstra and th A. Schönecker, Proceedings 17 European Photovoltaic Solar Energy Conference, Munich (Germany), in press [4] http://www.sunlab.nl [5] A.S.H. van der Heide, Dutch patent 1013204, 5 April 2001, worldwide patent pending [6] O. Breitenstein, J.P. Rakotoniaina and A.S.H. van der th Heide, 11 NREL workshop on crystalline silicon solar cell materials and processes, Estes Park (Colorado, USA), 253 (2001) [7] G.K. Reeves, H.B. Harrison, IEEE Electron Device Letters, 3, 111 (1982)