by Bob Gunn, Applications Team Leader, OIPT
A key element in achieving a repeatable process is to ensure that the chamber is kept as clean as possible. Some processes produce by-products that build up on chamber components over a period of time, and can either lead to a process drifting out of specification and/or a particle issue.
One means of removing this build up of product is to open the chamber and physically clean the components with the appropriate solvents and cleaning agents. This is costly in terms of man-power, time, the system being non-productive etc. Once the chamber is cleaned, the time taken to reach base pressure, re-qualify the system for production etc must be taken into account. Anything that can be done to increase the time between a mechanical clean of the chamber has a great benefit on system uptime and productivity.
Changing the process chemistry such that it eliminates the non-volatile byproducts is one solution, but unfortunately this is not always possible. For instance, one case is Sapphire etching, primarily used in the production of HBLED’s where the optimum process is based on using BCl3 chemistry. After 10-15 etch runs the chamber needs to be opened as the build up of by products becomes so great it starts to produce particulates which are detrimental to the device performance.
Another approach to increasing the time between mechanical cleans is to develop an in situ clean, which effectively uses another plasma gas chemistry that etches away the non-volatile products, leaving the chamber clean. This in situ clean is used periodically before the chamber build up gets too thick to become problematical. After performing the clean, which often uses a gas chemistry which may affect the device process, a conditioning run is carried out using the device process chemistry and also dummy wafers, to ensure the system is ready to run production wafers. This condition process also needs optimisation to minimise the time to producing wafers again.
OIPT has been developing an in situ clean to extend the mean time between mechanical cleans for a Sapphire etch, with a large HBLED manufacturer. The customer is using a System133 ICP380 to etch batches of 2” sapphire wafers. This requires a mechanical clean every 2-3 days, takes the system out of operation for 16-24 hours, and is unacceptable as the production of HBLED’s increases.
Initial trials were based on a CF4/O2 chemistry. This did have limited success, increasing the meantime between mechanical cleans to 18 process runs before requiring a mechanical clean. To improve this further a rigorous study was undertaken.
EDAX and SEM analysis of the material deposited showed a crystalline type structure and the material was BxOy. Based on this analysis O2 was removed from the clean as it was thought this would aid the formation of more BxOy. SF6 was chosen as the basis for the clean chemistry to try and form volatile by-products. To maximise the amount of free F in the plasma high ICP powers and high flows of SF6 were selected.
To ensure that the process clean was run under the same conditions each time, ten sapphire etch processes were performed prior to a clean being run. Once the initial clean time had been completed, the chamber was opened for visual inspection. Depending on what was observed when the chamber was open either extra time was given to clean, or a mechanical clean was performed to start another experiment from a clean chamber. It was found in this initial study phase that higher ICP powers with moderate RF bias gave the best results. In consultation with an HBLED manufacturer the time was considered too long, and they were concerned about long term running of their systems at higher powers.
A new approach was needed. To shorten the time of the clean required that fewer sapphire etch runs were performed prior to running it, as it was found the clean time did not increase linearly with the number of sapphire runs. Experiments were also carried out with various changes to system configuration as well as process parameters.
After many experimental iterations and discussions with the customer a running regime was agreed of three process runs, clean, condition then repeat. Optimisation of both the clean process, still based on SF6, and the conditioning has to be carried out to ensure a rapid return to running production wafers. The process performance ie etch depth, CD etc has been checked after each iteration of the process cycle (process, clean, condition, process), to ensure that this was within specification. After several cycles of this the chamber was opened which showed it was very clean.
This new process regime is now under trial at the customer site with very encouraging results.
For more information on Oxford Instruments Plasma Technology processes and process techniques click here