Uniformity of pumping, system symmetry.

For those of you depositing metals this is going to be of less interest but for those of you either reactively depositing coatings or thinking about developing a reactive process this should be of great interest.

I recently went to work on a vacuum system that had been designed as a research and development machine capable of depositing coatings by reactive deposition. Typical coatings that the machine was designed for included titania, alumina and the transparent conducting coatings such as indium tin oxide (ITO).

The system had been designed with coating uniformity in mind and the pumping was arranged around the chamber along the centreline of the web path. The gas input was through manifolds that were similarly positioned to give a uniform flow across the width of the web. Thus the general arrangement of the sputtering and other deposition sources, pumping and gas input could be expected to give uniform films.

Needless to say the reason I am writing this note is that the coatings were anything but uniform.  Originally some ITO coatings were deposited from ceramic sputtering targets and these were reasonably uniformly conducting across the width and fairly easily optimised. However when it was tried to deposit ITO coatings from a metal target it was impossible to optimise the conductivity and when the resistivity was measured in more detail it became apparent that there was a different resistivity on each edge. When a minimum was reached on one edge the other was more than 20x higher and vice versa.

On a detailed review of the system, checking the symmetry of all the equipment it became obvious what was the cause of the problem.  The original design included a cryocoil in the winding zone to getter the water vapour from the roll of polymer and this too was uniformly distributed across the web to give uniform pumping. There was some work done on the best positioning of cryocoils/cryopanels in the vacuum system and having cryocoils in both the winding and deposition zones was shown to be better than having a single cryocoil in either the winding or deposition zone.  To take advantage of this information additional cryopumping was added to the deposition zones. Unfortunately at this point the symmetry was lost as the cryocoil was positioned on only on one side between the original pumps and the chamber wall. To make this worse as the coils were behind the sputtering sources and so could be expected to see some heat they were protected by being protected by a metal box.  The net result of this was that the conductance to the original pump was restricted on the side that had the cryocoil.

We know the end result was skewed pumping to the deposition zone such that the pressure at each end of the sputtering source would be different and so the requirement for reactive gas would be different. Hence, as the reactive gas was being uniformly introduced across the whole width, the optimum point to give the correct stoichiometry for a coating with minimum resistivity would be different at each end of the sputtering target. 

The solution to the problem is simply to change the shape of the cryocoil in the deposition zone to have an equal surface area on either side of the existing pumping orifice. This would reduce the volume the existing coil takes and so the protective radiation shield could be moved further back and so most of the restriction in conductance would be removed. Thus the uniformity would be returned and the pumping speed increased at the same time.

This has yet to be done to prove the above solution is correct as the work has to be planned and fitted into programmed downtime. However we are in no doubt that we have found the source of the non-uniformity and how to correct it.

I think this offers an excellent lesson on why it is essential to have symmetry about the web centreline for all aspects of the system and process.  Adding equipment that restricts the conductance only on one side of the web or adding additional pumping such as a cryocoil and placing in a convenient space in the chamber rather than symmetrically across the web can result in non-uniform coatings.  I emphasise that this will only be a significant problem for reactive processes. Although minor variations may be measured across the web in other coatings for most applications these are negligible.  When depositing ITO from a ceramic target, where some reactive gas is included to compensate for any dissociation during the sputtering process, the dynamics of the process are such that working with an excess of reactive gas is not as critical as it is with the full reactive process converting a metal target to the desired compound.  Hence the non-uniformity of the coating will be similar to plain metal coatings rather than full reactive deposition.

Although it is less important to have symmetry of pumping in metal deposition it will become increasingly important as the width of deposition systems continues to increase. Already we have machines of 4.45m wide. If these are pumped from one end only imaging what the pressure difference will be from one end to the other. This pressure difference will become greater and greater as the width continues to increase and this will be seen as producing different deposition rates from resistance heated boats at each end of the system.  A simple solution is to pump from each end although there will still be a pressure gradient from the web centreline to each edge.  A better solution would be to distribute the pumping across the whole width. It is worth noting that as customer specifications continue to increase, particularly for high technology applications, it can be expected that there will be an increasingly tight specification for resistivity which may require an improvement in system symmetry.

As you might have guessed this is a hobbyhorse of mine. I have worked on too many systems that were never designed with symmetry in mind but were expected to produce coatings by a reactive process that have been a major challenge. Mostly the difficulties could have been eliminated by better design making the processing much easier and the productivity of the system much higher. Thus I find it disappointing that even on new machines I still see some of the same design errors.