Stepping Up to Challenges in FR4 Microcircuit Imaging Technologies
July 11, 2006 |Estimated reading time: 6 minutes
<?xml:namespace prefix = o ns = "urn:schemas-microsoft-com:office:office" />
As the electronics industry's huge, well-financed R&D machine creates ever-smaller electronic wonder products, the sector consistently seeks smaller and smaller packaging. The current shortcoming in packaging is the printed circuit board's limits on line width and via size. To create the next generation of fine-line circuits at Sierra Proto Express, we experimented with new ideas on different very-fine-line manufacturing methods, including various imaging technologies to break the 2mil line (50 microns) bottleneck.
Pushing the optical envelope
We found a physical optical limit using our standard, noncollimated, 356nm light-imaging source, the 7mil photo tool and 1.5mil dry film. At about 1.7 mils (42.5 microns) of space, the straight lines start to form wavy lines due to the diffraction of light at each interface and harmonics of the frequency. It's known as the "Venetian blind effect" or Bohr's complementarity principle, and Fraunhofer diffraction slit phenomena. To decrease the line space at which this effect turns the spaces wavy, we can decrease the number of interfacing connections through which the light must pass, or decrease the thickness of the photo tool, decrease the period of the light, decrease the thickness of the dryfilm, or collimate the light to decrease the angle of diffraction.
To decrease the number of optical connections, the dryfilm cover sheet can be removed; although problems exist with sticky dryfilm and oxygen absorption in the dryfilm, both problems can be overcome. There is a .25mil (6 microns) improvement obtained by removing the dryfilm cover sheet--not as much as we need, but a worthwhile improvement nevertheless. A thinner dryfilm can be used; the thinnest we could procure was .8 mils (20 microns). Surprisingly, we found no major improvement between 1.5 mils (37 microns) and .8 mils dryfilm (20 microns).
With proper exposure, the different dryfilm thickness resolved both line width and space (although the thicker dryfilm did seem to break away from the copper sooner, due to its extra height). This result was quite unexpected, as it was always understood that thicker dryfilm would decrease line width; however, all the experiments showed equal resolution. In fact, in one experiment with 4 mils of dryfilm, 2mil lines were, surprisingly, resolved quite easily but were very fragile.
Note: There are 7 interfacing diffraction points in the typical printed circuit exposure system on each side: exposure system Mylar (2), photo tool (2), dryfilm release sheet (2), dryfilm (1). To produce lines/spaces below 2 mils, the number of interface points must be reduced.
The typical 7mil (150 microns) photo tool used in printed circuit manufacturing is not accurate enough for very fine lines. Photo 2 shows pin holes, a dip in the edge and splatter around the edge of the silver, and a large scrape mark at 10,000 dpi.
The silver coating on polyester film has small pinholes in the surface, creating possible opens or shorts. The finer-resolution (and more expensive) sputtered chrome on glass is a significantly thinner coating, but it offers a completely pinhole-free surface. The manufacturing method of laser-exposed photo tools produces a side effect of laser splatter; it's impossible not to have some reflection of the laser as it hits the surface. These reflected light beams can cause small dots alongside the main line edge, reducing the accuracy of the subsequent exposure. The laser transport system has large enough bumps in the table or laser system to allow small dips in the line edge. Angled lines also create irregular edge dips when the laser has to switch raster lines to create the angled line. As well, the soft silver on polyester film scratches easily, as seen in Photo 2.
To test the "thickness of the photo tool" theory, 4mil film was plotted and exposed on 1.2mil FX 930 dryfilm. No major difference was noted in line width: Lines were still limited at 1.5 mils. There was a small improvement but not a large enough one to resolve 1mil lines or less. We found no suitable method to discover whether a shorter-wavelength light would create thinner lines, as the light source to which we had access was 359nm. An experiment with a collimated light source did prove that imaging down to a 1.5mil (37 microns) line is possible with 7mil photo tools. However, the dryfilm did not prove to be usable, as the adhesion of the thin dryfilm lines to the copper surface was too low to survive manufacturing properly. Additionally, the surface of the copper was too rough and showed too much fiberglass weave from below to allow the dryfilm to follow the copper surface properly. Wet dryfilm lamination may improve this situation but not solve it. The collimated light exposure system requires a very clean room with impeccable photo tool management, as any dust scratches or fingerprints result in collimated light reducing yield.
Working the 2mil limit
Even our best technology ideas could not get significantly below the 2mil lines and spaces. Time to change tactics: We looked around at industries manufacturing very fine lines such as 40nm. We researched their technologies with a few experiments to ascertain whether they were applicable to PCB circuit manufacturing methods. Our results? Even though the integrated circuit imaging technology produced very fine resolution, their resists are too thin and fragile for circuit manufacturing; as well, their glass masks are prohibitively expensive. We researched other photolithography methods, with little in the way of promising results--except one. The best technology we found uses a fairly thick resist of 15 microns, which provides excellent adhesion to copper and is able to resolve 300nm lines in both etch and plate production methods. A collimated exposure projection unit allows for less expensive, simpler-to-manufacture glass masks on cheaper glass frames. The projection imaging technology does limit the maximum circuit size, but the size is well within the anticipated size of the microcircuits to be manufactured.
A second method of laser photo exposure on dryfilm was experimented on with an ESI Yag laser and produced 9micron lines; however, the laser is too slow for production methods.
To improve the copper roughness and imprinted weave of FR4, we conducted several experiments, resulting in a very smooth surface. Note: The copper surface is still too rough, and a polishing machine to remove the fragmented roughness will be incorporated. The upper corner of Photo 5 shows the rough copper.
For the ultimate fine-line width on FR4, a projection collimated light source was used with a glass chrome photomask. The finest line resolved was .015 mil, .361 microns or 361 Nanometers, on standard Fr4 with thin laminated copper. At 500x magnification, the photos clearly show how rough the Fr4 copper surface is. The lines are visible with a few copper specks left, signifying more experiments on dendrite oxide size (tooth) and etching technologies. The copper thickness as an etched circuit was 1-micron-thick copper with .361-micron resolution. The plated/etched experiment plated 5-micron copper/gold and resolved 2-micron lines and spaces.
2 mil limit on the way out
Micro FR4/flex circuits are not only possible today but are needed to allow for future very-small silicon die creations. With multiple-chip packages emerging, stacked die cavity packaging, and very small die at .3 mm pitch, the present limit of 2mil lines and spaces is huge compared to the technology required now and in the near future. Our manufacturing facility is in production of 1mil ( 25-micron) lines and spaces at present, with future plans of 5-micron traces and spaces.
Ken Bahl is the CEO of Sierra Proto Express.
Robert Tarzwell is the director of technology at Sierra Proto Express.