AVS 47th International Symposium
    Incorporating Principles of Industrial Ecology Wednesday Sessions
       Session IE+PS+MS+SE-WeM

Paper IE+PS+MS+SE-WeM3
Characterization of Remote Plasma Clean Process for Plasma CVD Chamber

Wednesday, October 4, 2000, 9:00 am, Room 304

Session: Environmentally Friendly Process Development
Presenter: T. Tanaka, Applied Materials Inc.
Authors: T. Tanaka, Applied Materials Inc.
T. Nowak, Applied Materials Inc.
M. Seamons, Applied Materials Inc.
B.H. Kim, Applied Materials Inc.
K. Lai, Applied Materials Inc.
M. Cox, Applied Materials Inc.
P. Loewenhardt, Applied Materials Inc.
D. Silvetti, Applied Materials Inc.
S. Shamouilian, Applied Materials Inc.
Correspondent: Click to Email
Remote plasma cleaning of CVD process chambers has proven to be more efficient than conventional in-situ plasma cleaning in terms of higher throughput and higher gas breakdown efficiency. It is still important, however, to maximize the efficiency of the remote plasma clean process because of the potential environmental impact and the cost of process gases. The remote clean process involves three steps: generation of reactive species (mostly fluorine atoms) in a remote plasma source, transport of the reactive gas, and the cleaning reaction in the CVD chamber. We studied the efficiency of the process in each step. Since accurate direct measurement of the atomic fluorine concentration in the various parts of the CVD reactor is difficult, we used etching of thermal oxide wafer coupons to estimate the relative distribution of atomic fluorine within the reactor. Source dissociation efficiency was studied using an indirect technique based on correlation of pressure to effluent composition. We found that it requires approximately 24eV to break down each NF@sub 3@ molecule. This translates to 1.7W/sccm of NF@sub 3@ flow. This was seen to be approximately the same for both a microwave discharge operating at 2.45GHz and an inductively coupled plasma at 13.56MHz. Results characterizing the transport step demonstrate the importance of system design on minimizing recombination losses of the reactive species which, for a parallel plate reactor, can be as high as 50% of the atomic fluorine generated in the remote plasma source. The experimental results are compared with a simple model, which describes the general behavior of the cleaning process.