page 1 page 2 page 3 -->page 4<-- page 5 acknowledgements figures references

Fabrication process

Fabrication of the spheres from bulk material

Fused silica spheres were manufactured from 2.5" Homosil cubes by two companies: Davidson Optronics and Speedring (ref.16). Both used a similar process consisting in a diamond turning, followed by grinding with rough abrasive (silicon carbide, 50 mm average size). The silicon spheres were manufactured at the Ginzton Laboratory Crystal Shop, Stanford University, from 2"(5cm) boules, using a similar approach.

Lapping and Polishing

We have discussed the mechanical problems first because one would like to have this part under control before attempting to optimize the process. Our application was such that the surface quality was of concern mostly insofar as it affects the sphericity. Subsurface damage (visible and not) was important because it modifies the properties of the bulk material in a surface layer which is estimated to be about twice as thick as the abrasive size used (10 mm and 3 mm for lapping; 1 mm for polishing). The surface was checked at a magnification 500 times, using a good optical microscope equipped with interference contrast attachment (Reichart). Additionally, an Atomic Force Microscope was used for some observations made by Dr. V.Graham (see ref.19; both microscopes courtesy of the Materials Research Laboratory, Stanford University).

In the lapping process our main concern was to remove all damaged material from the previous steps. The thickness of this damaged layer was expected to be 0.004" (100 mm). In fact as the size of the sphere at the acceptance was 1.55 inch (39.4 mm) and the final size was 1.4959" (~38 mm), the thickness of the layer removed by lapping and polishing was over six times the estimated thickness of the layer damaged in the grinding process. Brass laps with 6 cuts were used. Abrasive slurry was made by mixing alumina powder with de-ionized water. It was held in suspension by continuous stirring. Small amounts of slurry were pumped and delivered in drops. The slurry was not recirculated and was used at rate of about 1/4 liter per hour. We used abrasive of average size 10mm (Microgrit, ref. 17) for the removal of the first .025" (.635mm) of material. This stage lasted about 15 h for fused silica and about 30h for silicon. Lapping of the remaining .002" (51mm) was done using abrasive of average size 3mm (Microgrit). When a diameter of 1.4965" (38mm) was reached, the spheres were checked, measured and transferred to polishing. At this point the typical sphericity was 5-10 micro inch (130-250 nm)p/v. The estimated thickness of the damaged layer produced by lapping with 3mm abrasive was 6mm (0.0002"), while the layer removed in polishing was 8mm (0.0003").

Two of the most important factors in determining the quality of the polishing are the choice of slurry and the selection of the material used in forming the laps. This is especially true for silicon spheres.
As polishing slurry we used commercial products: Big C (ceria slurry from Universal Photonics,  (ref.18a), H2000 (rare earths proprietary slurry from Selectox - Hastings, ref 18b) and/or fine grade alumina powders in deionized water (Baikalox, ref.18c).

Some of the products tested  contained a considerable amount of very large particles (up to 40 mm particles in a nominally "1mm" powder,( see Fig 8  ). These are hard to separate and cause scratches, particularly in silicon. Scratches affect the sphericity, as the material around the scratch is modified and doesn't polish uniformly. Such scratches, even if polished out, leave irregular traces on the surface. We therefore looked for slurries which did not have such large particles. One way to find if big particles are present in a slurry is by washing off small particles. One eventually ends up with the larger residual particles which can be easily measured under a microscope (see Fig.8 ). In this way we were able to characterize several commercial products. The slurries finally selected still had a fraction of larger particles present, but in lesser degree.

We found that matching lap material, slurry and sphere material is the key to obtaining a good surface as well as a good sphericity. We started off using the original polishing machine equipped with specially made phenolic laps sintered with paper. In fused silica, these laps produced good sphericity, but not the best surface quality. Switching to pitch laps (Gugolz 64 from Universal Photonics ref.18a) produced a better surface, but not the best sphericity. This was traced to mechanical problems: When the polishing machines were modified , as described above, good sphericity could be obtained also with pitch laps.

Polishing of silicon presents additional problems, which are related to the crystalline structure of silicon (ref. 20). The experience gained in polishing fused silica was very useful when we started polishing silicon. The anisotropic mechanical and chemical properties of single crystal silicon are used to advantage when polishing flat silicon wafers. However, when the same polishing technology was attempted on a silicon sphere, it produced peak to valley heights of the order of 100 micro-inches (2.5 mm) highlighting the crystalline structure. Commercial products for polishing silicon wafers contain KOH, which is a preferential etcher. After reformulating the process and by using appropriate abrasive slurries and process controls, as described below, we were able to reduce the peak to valley heights to below 1 micro-inch globally.

The required sphericity and surface quality was achieved with a special combination of lap material and slurry, as well as a close control of the slurry acidity. The inadequate surface quality obtained on silicon with phenolic laps prompted us to return to pitch laps. Empirically, it was finally possible to identify a combination of factors which produced a particular condition on the silicon surface, leading in turn to a uniform polishing action. The best results were obtained when the pH was 6.7 and the surface wetting was minimal, which apparently minimizes chemical action. With the pitch laps there was less tendency to wet the silicon surface; in fact, the film of slurry would break up and it was necessary to deliver the slurry to each lap separately in order to guarantee that each lap was always wet.