Introduction
Experts predict that copper is the most likely replacement for most silver-based metal contact schemes for silicon wafer-based technologies in the future [
1]. Its high conductivity that is comparable to that of silver [
2,
3], its suitability for fine-line metallisation designs [
3–
5], its compatibility with simple plating techniques and its relative abundance and low cost [
6] are some of its desirable attributes. Nevertheless, many large-scale cell manufacturers still appear cautious about using the plating of metals directly onto the silicon. Adhesion strength and hence durability of such plated contacts appear to be the major concerns [
1].
New approaches involving the formation and use of anchor points, or grooves in the extreme case of overlapping anchor points in the silicon surface have been developed in this work for improving the adhesion and durability of plated contacts [
7,
8]. Such anchor points emulate the “Saturn technology” developed and commercialised by BP solar into large-scale manufacturing [
3], which essentially provides a continuous anchor point by plating metal into grooves buried beneath the surface. This “anchoring” of the plated metal has demonstrated superior adhesion strengths, even when compared to screen-printed contacts [
3]. This has motivated the present work on evaluating and simplifying the formation of such anchor points while simultaneously facilitating significantly higher efficiencies than achieved by Saturn. The formation of such anchor points is easily combined with laser doping processes [
9]. Laser doping for forming selective emitters in conjunction with plated contacts has been used for many years for achieving high performance cells [
10,
11] and has been successfully used in large-scale manufacturing [
4]. The formation of “anchor points” in the silicon surface is depicted in Fig. 1 [
7]. This can be achieved without complicating the laser-doping process by increasing the energy per laser pulse in the location where each anchor point is required without varying the scanning speed or frequency of the laser. The increased pulse energy allows a small amount of the molten silicon to be ablated. This results in the formation of a hole with the remaining molten silicon in the walls of the hole regrowing via liquid phase epitaxial growth during cooling while incorporating active phosphorus dopants to make the walls of such holes heavily n-type doped as shown in Fig. 1 [
7,
8].
In this work, both a stylus-based adhesion tester and an industrial metal peel tester the same as those typical used by most cell manufacturers have been compared [
12] and used to characterise the adhesion strength of various metal contact types for silicon solar cells.
The stylus-based adhesion tester shown in Fig. 2 uses a 3-axis motion control stage. The stylus is made of high-speed steel with a diameter of 0.7 mm, lapped flat on the end. The diameter of the stylus is chosen to be less than the separation of the fingers on the cell [
13]. Weights are used to apply a constant downward force on the stylus. An
X-Y stage with vacuum chuck moves the cell under the stylus at a constant programmable speed. A “load cell” senses the horizontal force on the stylus as the solar cell being tested moves under it. A load cell transmitter amplifies the signal, ready for data acquisition by Labview [
13]. The test has been described as a “dislodgement force” test [
13] with finger peeling being optionally limited by applying strips of adhesive tape 2 to 3 mm apart on each side of the line to be scanned. This prevents the stylus from peeling several fingers simultaneously by forcing each finger to break before the stylus reaches the subsequent finger. Typically, the cell is scanned at 1 mm/s.
Peel tests [
12,
14–
16] are used for adhesion testing of metallised contacts at busbar level. In parallel with the stylus-based tester, a commercially available peel tester (demonstrated in Fig. 3) that is designed based on European Adhesion Standard EN50461, is used to characterise the adhesion of metal contacts for the various cell types. The accuracy of the tool is ±0.125 N with a data capture speed of 500 measurements per second. During the measurement, the processing speed is up to 15 mm/s with a flexible ribbon width of both 1.5 mm and 2.3 mm. A 90° or 180° adhesion peel angle can be used in this peel test. In this study, 90° peel tests were conducted.
Experimental work
Five groups of cells were prepared as listed in Table 1. Group 1 was industrially produced solar cells with standard screen-printed Ag contacts. Groups 2 to 5 were solar cells utilizing LIP Ni/Cu contacts with metal fingers spaced 1 mm apart in conjunction with p-type, 1 to 3 Ω∙cm, Cz-Si (100) wafers and 100 to 120 Ω/emitters. These were all fabricated using identical industrial tool sets for processes in common. Group 2 was LDSE cells, and Group 3 was commercialised LDSE cells that were also called SuntechPLUTO cells. Groups 4 and 5 were the same as Group 3 except for the inclusion of the process for simultaneously forming the anchor points while carrying out the laser-doping process (process #7 below). In the case of Group 5, the density of the anchor points was so high as to cause them to overlap and therefore create a continuous groove equivalent to the buried contact solar cell (BCSC). The schematic of the laser-doped selective emitter (LDSE) cell design used in this work is displayed in Fig. 4.
The processing sequence used for device fabrication in this work includes wafer texturing, phosphorus 100 to 120 Ω/□ emitter diffusion, rear etch, front surface PECVD silicon nitride deposition, rear aluminium screen-printing and firing, phosphoric acid coating, front localised laser-doping and anchor point formation, Ni light-induced plating (LIP), belt furnace Ni sintering at 350 °C, CuLIP+ silver capping, and testing and sorting.
Each of the above processes was implemented through the use of standard commercial manufacturing tools except for the laser system as described above. The texturing baths, rear-etch tool and the Ni and Cu plating baths were manufactured by Kuttler and were the same as those used on the Pluto production lines at Suntech. More specifically, Group 4 wafers were fabricated by laser-ablating anchor points uniformly distributed along laser-doped line with spacing of about 0.2 mm. This was done for simplified testing despite earlier analysis and evaluation showing that the adhesion strength towards the ends of each finger is most important and sufficient in determining the reliability and durability of such metallisation scheme [
9]. Following processing, cells in the five groups were adhesion strength tested and analysed using both the stylus-based and industrial testers. Figures 2 and 3 show images of a stylus-based tester and a peel tester respectively, typical of those used for the testing in this work.
Results and discussion
Histograms showing the distribution of adhesion forces for screen-printed cells and LDSE cells are presented in Figs. 5 and 6 using the stylus-based tester. The adhesion force for the Group 1 (screen-printed) cells in Fig. 4 is close to a normal distribution although the distribution shown in Fig.6 for the LDSE plated contact cell is an unusual distribution with a strong positive tail. This is likely to be an artefact of the measurement technique when used in conjunction with textured surfaces. This highlights the need for caution when measuring the adhesion using such a stylus-based adhesion tester. The significant positive tail might be largely influenced by the textured pyramids near the laser-doped regions where the lateral metal dislodgement is prevented by the metal being “pushed” against the structure of the pyramids by the sideways force applied by the stylus. Such interference appears to periodically give much higher adhesion strengths compared to the peel test, which is much lower and reasonably consistent. Such inconsistencies appear to be able to reach values as high as 300% of the value indicated by the peel test although the median value from the measurements appear to give results more consistent with peel test measurements.
Figure 7 exhibits the underneath side of the metal following dislodgement from the Si. The clear pattern of the pyramids can be seen impregnated in the metal layer. The plated metal does not plate to the dielectric coated silicon, with minimal contribution to the adhesion being contributed by these regions of the metal impregnated by the pyramids where the metal has simply grown across the wafer surface. Consequently, as can be seen there is minimal or no damage to the metal in Fig. 6 (a) in the regions adjacent to the laser-doped region as a result of dislodging the metal. Nevertheless, as noted, such regions might be significantly contributing to the measured adhesion strength for some of the measurements when using the stylus-based approach as the lateral force causes the metal over the pyramids to be wedged against the actual pyramids despite the fact that there is no significant adhesion between them.
To identify and study the peeling interface for these plated LDSE cells shown in Fig. 7, the scanning electron microscopy (SEM) with energy dispersive X-ray spectrometry (EDS) was used to analyse the underside of a peeled finger. As seen in Fig.7(b), Ni remained on the underside finger surface after testing in the region where the Si surface was exposed for nucleation of the plating, suggesting that the Si-Ni interface was disrupted by the test and therefore represented the weakness or failure point in the adhesion strength. In comparison, there was no Ni evident in the regions away from where the Si surface was exposed, confirming the masking ability of the SiNx layer to prevent plating to the regions coated by the SiNx. This however exacerbates the adhesion problem since it implies there is minimal adhesion between the metal and Si in the textured regions of Fig. 7(a) which are SiNx coated and yet such areas can be seen to comprise nearly 70% of the interfacial area between the metal and the solar cell surface. The expectation therefore is that the weakness in adhesion strength for these contacts arises significantly as a result of this poor or negligible adhesion between the plated metal and the dielectric coated pyramids and the fact that in this cell design, such regions are so large in area compared to the actual metal/Si interface area that is relied upon for the adhesion strength.
In order to avoid this problem, the Group 3 cells received significantly less plating than the Group 2 cells as indicated by the metal contact heights in Table 1 with the latter receiving about 13 μm of plating while the Group 3 cells were plated with only about 7 to 8 μm of Cu. The significance of this is that the height also tends to represent the amount or distance the Cu plates laterally across the SiN
x coated pyramid regions since the conventional LIP process [
17] tends to plate reasonably uniformly in all directions. This explains the reason why in Fig.7, the textured regions of Cu juxtaposed to the smooth region (where the metal directly contacted the silicon) are about 13 μm in width. Therefore, of the total 36 to 38 μm of width in the underneath surface of the metal contact, 26 μm are simply Cu as shown in Fig. 6(c) with no adhesion to the underlying surface, while in comparison, the Group 3 cells, using the same laser-doped lines to expose the Si surface to nucleate the plating, have a metal contact width of only about 26 μm, with only just over half of this being the regions of minimal adhesion strength where the Cu interfaces with the SiN
x. As seen in Table 1, the adhesion strength of contacts designed in this way greatly exceeds the adhesion strength of the Group 2 cells by a factor of about 6 to 7 times as measured with the “peel tester”. Interestingly, if the adhesion strength is normalised to the interface area between the plated metal and the silicon, these plated cells match the adhesion strength of the screen-printed contacts. However, since the plated cells formed in Groups 2 and 3 do not achieve a high interface area with the silicon in the same way that conventional fired screen-printed contacts can, concern remains for the former in terms of reliability and durability.
These results are also confirmed by the stylus tester although again caution needs to be exercised with the latter as there are questionable values on the high side which suggest that it is possible for the best/strongest of the Group 2 cells to have adhesion strengths that match or even exceed the contact adhesion strengths of many of the Group 3 cells. This is inconsistent with the “peel test” results of Table 1 where the lower limit to the adhesion strengths measured for the Group 3 cells is more than four times greater than the highest adhesion strengths measured for the Group 2 cells. It is concluded that roughened surfaces such as those existed between the metal and textured silicon surfaces make it difficult to measure the actual adhesion strength when applying a force parallel to the silicon surface. It is likely that the same argument can be applied to the screen-printed (SP) contacts of the Group 1 cells, where all the contacts are formed onto textured surfaces. This increase in surface area would of course be expected to increase the adhesion strength between the metal and the silicon surface, although this is difficult to quantify when using the application of a force parallel to the Si surface. In fact, the testing in this way leads to a completely different failure mode for SP contacts with the disintegration of the metal itself occurring ahead of failure at the metal/Si interface. In comparison, through conventional peel testing, there is no evidence of disintegration of the SP metal, with failure either occurring due to the metal/Si interface giving way or else chunks of Si being removed prior to the metal/Si interface giving way. For these reasons there are no attempts being made in this work to try and assign absolute values to the adhesion strengths when measured using the adhesion testing approaches in this work and so instead, everything is normalised against the SP cells. In conclusion, the two adhesion testing approaches, each having their limitations, appear to be measuring different aspects of the strength of the metal contacts, making it difficult to conclude which is most appropriate for indicating the reliability and durability of the metal contacts. Using the stylus method in addition to the standard peel testing allows further information about plated metal adhesion and durability of plated contacts to be gathered.
In general, this work showed that for all the plated cells, Ni and its interface with the silicon appears to be the key factor that affects the adhesion strength in a Ni/Cu contact LDSE cell. This same conclusion has been drawn by others such as Mondonetal [
15] who reported the existence of voids at the Si-Ni interface which reduce interfacial contact area and lower the adhesion strength. Therefore, enlargement of Si and Ni contact areas (e.g., modification of surface patterning features) and chemical treatments, e.g., surface treatments, metallization process and post-treatments (e.g., Ni strip-off and replate), of Ni on Si surface can potentially be used to enhance the adhesion strength in a Ni/Cu contact silicon solar cell, as well as plating lesser amounts of metal.
Table 1 shows the excellent adhesion strength achieved for the plated contacts when using the anchor points, with more than an order of magnitude increase compared to conventional plated contacts onto planar silicon. In fact in this work, the adhesion strength measurements did not even determine the strength of the metal/silicon interface, but rather the force necessary to remove large chunks of silicon that pulled away with the plated metal as shown in Fig. 8. The use of such anchor points appears to more than compensate for the potential problems created by increased volumes of plated copper, poor interfaces between the Ni and silicon, and the lack of adhesion between the SiNx and the overlying Cuthat plates across the surface. With regard to the latter, the EDS result in Fig.7(c) confirms the low interfacial adhesion at the SiNx/Cu interface with the evidence of Cu remaining on the underside of peeled fingers adjacent to the remaining nickel. Without the use of anchor points, it appears that both the SiNx/Cu interfacial area as well as the height of the plated metal need to be carefully optimised to achieve adhesion strength approaching that of SP cells.
Conclusions
Adhesion strength and durability of plated contacts are probably the main issues causing many cell manufacturers to be cautious about considering such metallisation for large-scale manufacturing. It appears that conventional light-induced plated Ni/Cu contacts onto planar silicon can only achieve metal adhesion strength comparable to that of screen-printed cells if the plated height is restricted to below 8 µm. For contacts requiring metal heights greater than this for increased metal conductivity, laser-ablated anchor points or grooves in the heavily doped silicon surface can ensure good mechanical adhesion strength. LAASE cells showed superior adhesion strength to other types of cells with plated metallisation, and were also superior to SP cells, with peel tests showing adhesive strengths that are 100%–150% of that achieved by SP cells. BCSCs also demonstrated metal adhesion strengths superior to those of SP cells and only slightly worse than for the LAASE cells.
The evaluation included the use of the new stylus-based adhesion tester as well as the favoured industrial “peel test” approach for comparison. The testing shows that the new stylus-based adhesion tester needs to be used with caution in conjunction with textured or roughened surfaces where the application of the force parallel to the silicon surface tends to force the metal against elevated regions of the silicon surface rather than testing its adhesion, but this new technique allows further information about plated metal adhesion and durability of plated contacts to be gathered.
Higher Education Press and Springer-Verlag Berlin Heidelberg
PDF
(240KB)
