November 7, 2011 — Electron-molecule collisions drive technological plasmas. Electron energies and concentrations determine the gas phase chemistry, and therefore surface effects, in all plasma etch and deposition processes. Almost all micro-electronic devices now require dry and plasma processes to achieve the necessary surface detail, yet these plasma processes are still poorly understood.
TEL (Tokyo Electron Ltd.) and Quantemol Ltd. are collaborating on using the fully quantum mechanical R-matrix methods [1] to study the electron-molecule collision processes occurring in their plasma tools. The data provided by the R-matrix codes are used in reactor-scale models to simulate plasma behavior, allowing predictions of etch or deposition rates to be made.
The authors discuss the etching of silicon dioxide by fluorocarbons, and Ar/HBr etching in a microwave plasma source.
Figure 1. The capacitively coupled plasma (CCP) etching chamber used in this article. |
Etching silicon dioxide using fluorocarbon-based plasmas
How can we understand how silicon dioxide wafer etching by fluorocarbon plasma produces surface features? How do we find out more about the plasma chemistry?
Fluorocarbon plasma etch works due to the production of fluorine, which then reacts with silicon dioxide at the surface. However, the way in which surface features, such as trench profiles, develop depends sensitively on the densities of ions (such as CF2?? and CF3?) and neutral radicals (such as CF2 and CF3).
We performed electron-CF2 collision calculations, using the Quantemol-N software [2], which provides an expert system interface to the UK Molecular R-matrix codes. The results were incorporated into a zero dimensional chemistry simulation as a global model of the capacitively-coupled plasma (CCP) etching chamber, shown in Fig. 1. The zero-dimensional reactor simulation is illustrated schematically in Fig. 2.
Figure 2. Schematic illustration of the zero-dimensional plasma model. |
This model can be applied to various plasma reactors, including plasma etch reactors, plug-flow reactors and continuously-stirred-tank reactors.
Figure 3. Simulated and experimental etch rates versus absorbed power in fluorocarbon etch. |
A comparison is made between etch rates obtained from simulation and experiment in Fig. 3. The calculated and experimental results show very good agreement. To explore the importance of the chemistry we consider simulations with and without reactions involving CFx species. The results are shown in Fig. 4. Including the extra chemistries has no impact on the overall etch rate results as the fluorine concentration is dominant in the process and is not changed by inclusion of the extra chemistry (Fig. 4, right). However, we see that including reactions with CF2 radicals into the simulation produces very different results for the CF2 and CF concentrations (Fig. 4, left and center). Concentrations of CF2 and CF play a key role in selectivity during Si and SiO etching processes. Better predicting these concentrations assists in controlling etch selectivity.
Figure 4. Concentrations of species obtained from simulations with and without CF2 radicals. From left to right, the species are: CF2, CF, and F. |
Ar/HBr etch process
We now consider Si etching performed using Ar/HBr plasmas.
TEL has experimentally determined etch profiles for the Ar/HBr etching process at a range of pressures [3]. Comparing the shapes of the etch profiles obtained with the lowest pressure (40mTorr), to those obtained with the highest (100mTorr), at the higher pressure, the edges of the etch profile are closer to vertical, the desired shape. Why does an Ar/HBr process in the TEL microwave source work well at pressures greater than 40mTorr? What is the chemical or physical process behind this?
To understand how the surface features change with pressure in the microwave source Ar/HBr etch process, we need to understand how the ion densities evolve. We suspected that etch products like SiBr2 could be important and so new quantum mechanical calculations of electron interactions with SiBr and SiBr2 using Quantemol-N were done using the software noted in the previous section.
Figure 5. 2D simulation domain and modeled gas flow stream lines for the TEL microwave antenna source. |
These new data were used as inputs to a 2D simulation [4] of a microwave Ar/HBr etch process. The details of the 2D etch simulation are shown in Fig. 5. The plasma is driven by microwaves, and the wafer sits on a stage subject to RF biasing.
Figure 6. Species concentrations (at pressure 100mTorr) for models including SiBrx etch products (upper row) and not including etch products (lower row). The numbers given in the plots are the maximum concentrations in cm-3. |
The concentrations of the various species in the chamber are shown in Fig. 6. We see that, at high pressure, SiBr and SiBr2 are ionized and return to the wafer to continue etching. These heavier ions have a lower maximum energy and therefore create neater trench profiles at high pressure. We have been able to verify this experimentally and are now designing a system based on the new insight provided by the simulation work.
Conclusion
The case studies discussed here illustrate the important role that simulation work can play in understanding fundamental plasma physics, and providing insights into process optimization and development in industrial and applied settings, in a cost-effective manner.
References:
[1] J. Tennyson,