Papers

AVS1997 Session MS+EM-ThA: Process Fundamentals

Thursday, October 23, 1997 2:00 PM in Room J1/4

Thursday Afternoon

Start	Invited?	Item
2:00 PM		MS+EM-ThA-1 An Oxygen Plasma Flash Process for the Control of Corrosive Gas Migration in a Semiconductor Wafer Plasma Etch System P.E. Brunemeier, T. Miu, W.Z. Collison, W. Klippert, C. Vetter (Lam Research Corporation); D.M. Dobkin (Enigmatics) An urgent problem in plasma etch systems for semiconductor manufacturing is the corrosion of system surfaces and consequential deposition of particles of corroded metal on wafers that undergo process in the system. Recent investigation indicates one source of these particles is corrosion of load lock components as a result of their exposure to corrosive process gases. In this work, we have characterized the migration of the corrosive gas species (Cl₂, HBr, and products) from the process chamber to the load locks; additionally, we have demonstrated a means of greatly reducing this migration, and hence, reducing system corrosion. The process system that we used in these studies was a Lam Research Corporation Alliance 9400 cluster tool etch system, using standard gate etch chemistries on a variety of typical materials. We have measured corrosive gas pressure rise during wafer processing with a Leybold residual gas analyzer (RGA) attached to a cassette load lock. In the first part of our three part study, we have identified the predominant corrosive species as well as the modes of migration, and their sensitivity to plasma process parameters. For the range of conditions tested, the data indicate that the majority of the transport is vapor diffusion through the system during process. In the second part, we have discovered a highly effective means of reducing the migration of corrosive gases, specifically, a brief oxygen plasma flash process ¹. The plasma flash step is performed in the process chamber at the end of the etch process for each wafer, has a duration of 5 seconds, and appears to reduce the concentration of corrosive species in the load lock by about 95%. In the third part of our study, we have searched for any possible production process impacts that might result from the oxygen plasma flash process such as increased process chamber particle contribution, process shift or etch profile effects, or reduced device yield or system uptime in volume production. We have found none, either in data from volume production settings or in tests done in our laboratory. ¹Patent applied for.
2:20 PM		MS+EM-ThA-2 Mechanism of Polymer Formation in Poly-Si Etching L. Liu (Micron Technology Inc., (presently at WaferTech LLC)); G. Blalock (Micron Technology Inc.) Plasma etching of gate transistor is critical in defining devices’ performance. The gate oxide is getting thinner to achieve faster speed and the linewidth is getting smaller to reduce the die size. High oxide selectivity and CD control are the basic requirements for the gate etch. The profile for the gate stack should be maintained as straight as possible. Any slope or undercut in the profile will either make the gate transistor performance inferior or the process unstable. Such straight profile is achieved by forming a protecting polymer layer on the sidewall while etching proceeds. The formation and composition of such polymer depends on what kind of film and what the etching process is. This paper concentrates on identifying the formation of polymer during poly-Si etch using HBr, Cl2, and HeO2. Adjustment of process parameters can provide excellent oxide selectivity and profile control. Such polymer, which is responsible for straight profile, is also the problem source for residual resist after clean. The polymer formed during this etch is identified to be Si-O based film by AES survey. It is done by analyzing the thin film surrounding the resist. Such polymer can be easily observed under SEM after resist strip (SEM pictures will be presented). Its thickness can be altered by the HeO2 flow, and so does the poly-Si’s profile and oxide selectivity. The effect of HeO2 on oxide selectivity will be discussed. This Si-O based film can not be removed by O2 based resist ashing process nor by the standard piranha process. This polymer can be so thick that it protects the resist from being cleaned, leaving residual resist. This Si-O polymer is found to be very porous and its thickness is between 50 to 100A. The bond strength between Si and O is low. Small amount of F containing gas can remove it effectively but it will degrade the remaining oxide. 100:1 HF dip after etch is also effective in removing it and its effectiveness for production will be discussed.
2:40 PM		MS+EM-ThA-3 Gas Purity Requirements for Semiconductor Processing: Tungsten CVD and Aluminum Etch A.D. Johnson (Air Products and Chemicals, Inc.); A. Glew, R. Rajogopalan, S. Nanjangud, S. Ghanayem (Applied Materials); J. Ammenheuser (SEMATECH); R.V. Pearce (Air Products and Chemicals, Inc.) Successive generations of semiconductors put increasing demands on the purity of process gases. However, there is a disconnect between these gas specifications and the needs of semiconductor manufacturing. Preliminary results on the effect of gas impurities on the tungsten deposition and aluminum etch processes are presented. Mass spectrometry was used to measure baseline impurity levels in the process chamber and monitor impurity additions to the process gas. Known impurity levels were then introduced immediately before the process chamber with either an Ar (W CVD) or N₂ (Al etch) carrier gas. Tungsten deposition in a WxZ chamber consists of two steps: a nucleation step involving SiH₄ reduction of WF₆, and a deposition step involving the reduction of WF₆ by H₂. Aluminum etch uses a Cl₂ and BCl₃ plasma chemistry. Carbon dioxide, moisture, oxygen and hydrocarbons (C₂H₄) are common impurities in these process gases and were used in this study. Tungsten deposition was insensitive to moisture up to 100 ppm and CO₂ up to 500 ppm. Oxygen has minimal effect below 10 ppm, whereas at 500 ppm was found to inhibit deposition. Ethylene, up to a concentration of 10 ppm, increases the deposition rate, and, at a level of 500 ppm, inhibits deposition. The Al process (Al etch rate, uniformity, and corrosion) was found insensitive to these gas impurities.
3:00 PM		MS+EM-ThA-4 Study of E-H Mode Transition in a Large Planar Geometry ICP Reactor J. Shon, A. Ellingbow, P. Vitello (Lawrence Livermore National Laboratory) Inductively coupled radio frequency (rf) discharges are known to operate in two modes: E-mode, when capacitive coupling dominates, and H-mode, when inductive coupling dominates. Recent work in cylindrical geometry has found that the transition occurs when the inductive skin-depth is approximately twice the characteristic discharge length.[1] In this work,we investigate the dependence of E-H mode transition on electron density and scale length of the reactor. We use a large planar geometry ICP reactor with 50cm diameter and 10-20cm height. The experimental range of input power and pressure are up to 2kW at 13.56kHz driving frequency and 1-100mTorr pressure. The rf magnetic field in the plasma is measured using an inductive loop array, aligned with the chamber height. rf magnetic field data is used to distinguish E or H mode coupling. A 35GHz microwave interferometer measures line integrated electron density, also aligned with the chamber height. A two dimensional plasma reactor model solves temperature and continuity equations for electrons, and momentum and continuity equations for ions, coupled with poisson's equation, which is used to predict electron density variations in the experiment.[2] An effective collision frequency is applied to account for stochastic heating of electrons. Preliminary results have shown that the E-H mode transition in the large ICP reactor, which is sensitive to pressure and input power. [1] U. Kortshagen, N. D. Gibson, and J. E. Lawler, J. Phys. D: Appl. Phys. 29 (1996) 1224-1236. [2] P. Vitello, J. N. Bardsley, G. Dipeso, and G. J. Parker, IEEE Trans. on Plasma Sci., 24, 123, 1996.
3:20 PM		MS+EM-ThA-5 Effects of Lateral Confinement and Ti Underlayer Thickness on C49 and C54 TiSi₂ Phase Formation in TiN/Ti Bilayers on Flat and Patterned Si(001) J.R.A. Carlsson, I.G. Petrov, D. Gall, J.E. Greene (University of Illinois, Urbana-Champaign); J. Givens (Micron Technology Inc.) The effect of initial Ti thickness and 2D lateral confinement in device vias with diameters of 0.6 µm on the microstructure and formation of the C49 and C54 TiSi₂ phases have been investigated for TiN/Ti/Si(001) structures. Ti films were grown on flat Si(001) and patterned SiO₂/Si(001) wafers by collimated magnetron sputter deposition to total thicknesses t_Ti ranging from 8 to 130 nm. TiN overlayers, 15 nm thick, were then deposited by CVD without breaking vacuum. The microstructure and phase composition of as-deposited and annealed films were examined by TEM and XRD. In both blanket and patterned structures, the as-deposited CVD-TiN layers were amorphous whereas the Ti layers were polycrystalline with a strong (0002) preferred orientation and an average grain size Ti> of 20 nm. Rapid two-step thermal annealing to 715 °C resulted in the formation of a TiSi₂ layer at the original Ti/Si(001) interface and a polycrystalline TiN layer at the CVD-TiN/Ti interface. The CVD-TiN layer remained amorphous after annealing. Blanket bilayers with different thicknesses on unpatterned Si(001) contained C49 (t_Ti≤14 nm), a mixture of C49+C54 (20≤t_Ti≤30 nm), and C54 (t_Ti≥40 nm) TiSi₂ after annealing. Patterned structures, on the other hand, contained only the C49 TiSi₂ phase for all t_Ti. Lateral confinement also had a dramatic effect on the average grain size C49> of the C49 phase. While C49> in the blanket layers increased from 80 to 170 nm as t_Ti was increased from 13 to 30 nm, C49> in the patterned structures was found to decrease from 120 nm for t_Ti=14 to 60 nm for t_Ti=30 nm.