A great deal of effort is being made to make grid conductors finer – reducing them by up to half in the immediate future; the challenge is to create narrower grid features, while at the same time increasing their height so as to maintain, or even increase, their aspect ratio.
If silicon-based photovoltaics (PV) are to claim a substantial share of the global energy market, improvements in solar cell efficiency are essential. One significant barrier to achieving this is the shadowing of the silicon wafer’s surface by its frontside solar energy collecting grid, which typically stops sunlight from reaching about 8% of the wafer. A great deal of effort has therefore gone into making the grid conductors finer, and current widths are now around 80-120µm. The industry aims to reduce these by up to half in the immediate future, but this is not without its complications. The challenge is to create narrower grid features while at the same time increasing their height so as to maintain, or even increase, their aspect ratio.
While the pursuit of higher aspect ratios is essential for improved conductor efficiencies, the solar industry would do well to combine this with work on another equally important factor: conductor uniformity. A conductor that has many high/wide and low/narrow points is less efficient than one that has the same cross section throughout. Simply put, a conductor will only carry as much energy as its narrowest, lowest point will allow, which means that silver is wasted wherever its minimum cross-sectional area is exceeded. The non-uniformity of conductors is currently at a level that the industry accepts, but as lines become finer, this, as well as aspect ratio, will become an increasingly critical factor.
|Figure 1. The effect of paste type on conductor aspect ratio.|
An in-depth study of this issue has been underway at DEK Solar since 2008, with the aim of developing a method for improving the efficiency of printed frontside silver conductor lines. As the solar industry’s favored deposition process is screen printing, the DEK team opted to concentrate its efforts on developing this technique and identifying the features necessary for an optimized, high-aspect ratio printing process. Before we look at the work and findings, let’s take a moment to look briefly at some elements of the screen printing process.
Conventional screen printing process
A conventional printing screen is made up of a frame to which a square-woven stainless steel wire mesh is applied at an angle. For frontside solar conductors, the angle used is typically 22.5° to randomize the weave with respect to the straight lines of the conductors. The mesh is coated with a photoimageable emulsion that is thicker on the underside (the wafer side) than the topside. This coated mesh is exposed and processed to create a pattern of apertures through which paste will be pumped by a squeegee that sweeps at high speed across the top of the screen during the print process.
For its initial benchmarking work, the team selected a popular industry standard paste and mesh printing screen, and designed a special wafer print pattern with multiple apertures ranging from 50 to 125µm, (later on, the team would supplement this with a further test pattern whose aperture widths ranged from 30 to 80µm). This first phase of the study indicated that process parameters – speed, print gap and print pressure – had no significant effect on conductor quality and aspect ratio. The team set out to explore the effects of changing materials and tools. As can be expected, conductor height was increased with the use of special high aspect ratio pastes (Fig. 1). It was also affected by emulsion thickness (Fig. 2) – too thin, and the conductors, too, were below optimum height. Too much emulsion also resulted in lower conductors because the print paste adhered to the higher aperture walls in the emulsion and was pulled away from the wafer below – at some points so severely that the final print was actually distorted. As is often the case, the team found the optimum emulsion thickness for efficient transfer of high aspect ratio pastes to be somewhere between the two extremes, at around 30µm.
Having found the best printing paste and optimum emulsion thickness, the team turned its attentions to the print screen itself – and to its inherent problem, which appears to be one significant obstacle to achieving the improved levels of paste transfer efficiency required for higher aspect ratio conductors, namely, the screen apertures are partly full of wire. Theoretically this must reduce paste transfer efficiency on two counts. First, the volume of screen aperture that is occupied by the wire can not be filled with paste, and second, the wire has a large surface area to which the paste can stick instead of transferring to the wafer. And herein lies one of the main reasons for poor feature uniformity. On analyzing conductor structure, it was graphically clear to the team that the intervals between the highs and lows in any conductor mirrored the intervals between the knuckles in the screen mesh.
Screen manufacturers have put a lot of work into mitigating the problem of paste transfer by reducing wire diameters from the 30-35µm that was inherited from the thick film industry, to the current PV industry standard 20-25µm. An added advantage is that while they enable more silver paste to be transferred to the wafer for any given aperture size, finer wires and higher mesh counts (wires/cm) also generally yield smoother, more uniform conductors (Fig. 3).
|Figure 2. The effect of emulsion thickness on conductor aspect ratio.|
But there is a limit to how fine the wire can be – especially considering that, in its normal lifetime, a print screen will typically be exposed to the rough and tough environment of 10,000 squeegee passes. Fine wire meshes could stretch or even break with this degree of use, distorting the print or worse, so a mesh of below 16µm is not generally considered reliable enough for high-volume manufacturing. There is also an issue of cost – fine, high mesh count screens cost more to make than large diameter, low mesh-count screens. And that’s before venturing into special alloys for making fine wires stronger.
Doing away with the mesh
The above discussion begs a question: why not remove the mesh from the apertures entirely? Several attempts have already been made to do just this, and in some cases, these have been quite successful, firstly in demonstrating feasibility, and then in higher volume environments, using two-layer electroformed stencils. Having decided to test this option as part of its study, the team worked to produce a number of specially designed high precision stencils.
|Figure 3. The effect of mesh type on conductor uniformity.|
Two-layer stencils typically use a bottom nickel aperture layer in which the wafer pattern is formed. As this is made up of numerous fine, long parallel apertures that alternate with what are essentially very thin metal strips, it is extremely delicate, and would not be able to withstand the physical stresses of multiple squeegee passes. Hence the top layer, which protects and lends stability to the aperture layer, is key. In its coarsest form, this top layer is a perforated foil that simply replaces the traditional 3D mesh screen with a 2D version. While the level of paste transfer and conductor uniformity is better than can be achieved using mesh screens, it is still far from optimal. In more refined versions, the top side is a solid foil with apertures similar to those in the bottom layer, but in this case they are strengthened by bridges, so that the long apertures in the bottom layer correspond with what in effect are broken lines in the top layer. Here, the challenge lies in precisely aligning the two layers and in ensuring that the apertures are well engineered – the top apertures will typically be narrower than those underneath to ensure good paste pull through.
Using this more sophisticated approach to stencil design, together with appropriate pastes, the team was able to improve aspect ratios as line widths decreased down to 50μm, and conductor uniformity was significantly better than anything achieved using mesh screens. Nevertheless, there was still a slight undulation in line height and width, in correspondence with the stencil’s bridges. Tests showed that this undulation was influenced by three factors: 1) the thickness of each layer; 2) bridge width; and 3) bridge pitch. This is to be the subject of further work as the team believes that a thicker bottom layer will allow more paste to flow under the bridges, effectively smoothing out the conductors.
Given these results, the team noted that two-layer metal stencils could potentially outperform the best emulsion mesh screens, but that several obstacles must be overcome before this technology can be implemented widely. First, great care must be taken to eliminate surface irregularities on the wafer side of the stencil. Defects such as nickel nodules or high spots can result in very high pressures being applied locally to the wafer, introducing the risk of wafer cracking and even breakage. Furthermore, given that the wafer surface has a rough pyramid-textured surface, and the nickel aperture layer, unlike screen printing emulsion, is unyielding, it is much more difficult to create an effective gasket between the two. This means that silver paste may bleed out between the stencil and the wafer during the print process, obviating the efforts made to decrease line width. Consider too that as conductor widths decrease, accurate alignment of the stencil’s two layers becomes ever more critical, and few vendors are capable of the high level of accuracy required for sub-50μm apertures. All this added to the costs of electroformed stencils led the team to conclude that while they offer some undeniable advantages, further developments are necessary to optimize their use for mainstream solar cell manufacture.
Figure 4. High aspect ratio conductor printed with DEK’s Hybrid Screen.
Certainly, the electroformed stencil route does offer the attractive prospect of almost wire-free apertures, a condition that is essential to being able to print uniform solar cell grid conductors reliably and repeatably. These factors and the considerable advantages of mesh screens led the team to consider a third option: a hybrid solution that would combine the considerable benefits of these two tried and tested technologies. An alternative was accordingly developed that is based on a conventional screen, but that replaces its wire mesh with a prefabricated electro-formed nickel top layer with “bridged” apertures similar to those used in two-layer electroformed stencils. This is coated on the underside with traditional photo-imageable screen emulsion, enabling the team to maintain the excellent ‘soft contact’ gasketing properties of screen emulsion, while removing a significant amount of metal wire from the apertures (Fig. 4).
The hybrid screen solution design incorporates bridges that, at 30µm wide, are as narrow as the wires in an equivalent screen, and that are placed every 500µm along the conductor apertures. This effectively creates a printing screen with almost no wire in the apertures, or just 6% coverage compared with the 35% coverage that would be found in a typical mesh screen.
Another benefit of the hybrid stencil is that the metal layer, at 30µm thick or less, is considerably thinner than a conventional 280 mesh. Provided that the nickel and emulsion thicknesses are optimized, the team concluded that this solution will theoretically offer better paste transfer efficiency, aspect ratio and cross-sectional area uniformity than the alternatives from which it is derived.
Tom Falcon is senior process development engineer at DEK International, 11 Albany Road, Granby Industrial Estate, Weymouth DT4 9TH United Kingdom; ph.: +44 1305 208415; email [email protected]