Low-Temperature Plasma Physics
Because of its undisputed advantages in a large variety of applications, plasma processing will undoubtedly play an ever larger role in manufacturing, especially in the semiconductor and display industries. The science of low-temperature, partially ionized plasmas, on which this technology depends, has roots in the field of gaseous electronics-arcs and glow discharges, switches, and diodes. The immense interest in high-temperature plasma physics since the 1950s has spawned the development of experimental and theoretical techniques undreamed of in the days of Townsend and Faraday. Only rarely, as in the case of electron-cyclotron resonance (ECR) sources, have these developments been applied to low-temperature plasmas. The advances in analysis and computational techniques, and in the knowledge of waves and instabilities resulting from fusion and space physics research, can now be brought to bear on the problem of partially ionized, multispecies plasmas. These studies would differ from the modeling work discussed in Chapter 2 in that they concern isolated problems rather than the whole final product. The discovery of new physical phenomena and evaluation of the importance of previous known effects would serve as inputs to the modeling studies. Given below are several examples of interesting topics of this nature that are suitable for both theoretical and experimental research.
Diffusion in weak magnetic fields. In gas discharges with a magnetic field between 10 and 1000 gauss (G), the thermal electrons are the magnetic field, but the ions and primary electrons are not. Ambipolar diffusion of the plasma across the magnetic field in that case does not follow standard formulas. When the ion and primary electron Larmor radii are of the order of the discharge radius, the problem is particularly difficult. What is needed is a simple treatment, even an approximate one, which does not require a full-blown computation for each set of parameters.
Contours of electric potential in a finite cylinder. When a magnetized plasma is bounded both radially and axially, the equipotential contours are expected to be saddle-shaped, having a potential minimum across a diameter and a potential maximum along the axis. In practice, however, measurements usually show a potential hill in both directions, indicating that some other mechanism is operative. This mechanism could be high-frequency waves or low-frequency drift waves, which can redistribute the electrons differently from classical collisions.
Stochastic heating in rf sheaths. In rf plasmas, the potential drop across the wall sheaths can have large oscillations at the rf frequency. Electrons entering and reflecting from the sheath can gain or lose energy, depending on the rf phase at the time they entered. This effect leads to a broadening of the electron energy distribution, thus affecting the formation of various molecular species in the plasma.
Ion temperature in low-pressure discharges. The spread of ion energies is important for anisotropic etching of semiconductors. Normally, the ion temperature is determined by energy gain from collisions with electrons and energy loss from collisions with neutral atoms. In low-pressure discharges of a few mTorr or below, however, the ion distribution is usually broader. This energy gain could be from ambipolar electric fields, sheath fields, plasma instabilities, or some mechanism yet to be discovered.
Ionization processes. In classical gas discharge theory, the electron-ion pairs forming the plasma are imagined to be produced by electron impact on neutral atoms. These could be fast electrons in the “primary” electron distribution, or they could be electrons so far in the tail of the thermal distribution that their energies exceed the ionization potential. In the dense plasmas of today, however, particularly the high-pressure discharges used in deposition, multistep ionization via metastables can occur. In fact, in high-pressure discharges such as used in the lighting industry, the propagation of ionization energy occurs via transport of resonant photons rather than electrons. This subject requires both theoretical and experimental study, and the results for etchant gases should yield information on the cross-section data that need to be obtained.
Electron runaway in high-pressure rf discharges. In direct current discharges between a cathode and an anode, applying too large an electric field will cause the discharge to break into an arc. In electrodeless rf discharges, however, extremely large electric fields can be applied without arcing. In that case, some electrons can “run away” into a velocity region of decreasing cross section and therefore reach ionizing energies even when the mean free path for a thermal electron is extremely short. This phenomenon would be expected to occur preferentially in gases, such as hydrogen, that do not have a large peak in the cross-section
Landau and cyclotron damping in rf discharges. Production of primary electrons by Landau damping has been postulated for helicon discharges, but there has been no prediction or measurement of the number of electrons accelerated in this manner. More recently, cyclotron damping has been proposed as an additional mechanism that can be important in low magnetic fields. These kinetic absorption mechanisms may be important in low-pressure plasmas used for materials processing.
Particle confinement by multidipole magnetic fields. Large uniform plasmas can be produced by letting plasma stream from a source into a “magnetic bucket” with permanent magnets lining the walls. Previous studies of confinement by such surface fields were concerned with the overall confinement time of the plasma. In etching tools, however, it is the confinement of each velocity class of electrons that matters, since the electron distribution has a large effect on the production of the various molecular species and on the damage incurred on thin oxide layers. This problem should be reexamined in the light of the new requirements.
Expansion of plasma in rapidly diverging magnetic fields. In “remote” plasma sources, the substrate to be processed is exposed to plasma that has come from the ionization source along magnetic field lines that are sharply curved. The usual adiabatic invariants are not preserved in such an environment, but there may be other invariants if the system is axisymmetric. The manner in which the electrons and ions of various energies move will determine the potential and density gradients in the downstream plasma, as well as the ionization occurring there. Though numerical modeling may ultimately be needed to treat this complicated problem, insight into the physics can be gained by considering general principles such as the invariants mentioned above.
Plasma instabilities. All plasmas, particularly magnetized ones, are subject to instabilities. In industrial devices, there can be drift instabilities due to gradients in density or temperature; gravitational instabilities due to curving magnetic field lines; or streaming instabilities due to non-Maxwellian distributions, such as in ECR. No devastating instabilities have yet been seen, but they will no doubt be found someday. In that case, stabilizing measures, such as minimum-B fields, are well in hand because of what has been learned in magnetic fusion.
A ROLE FOR NRL
The theoretical and experimental study of basic plasma phenomena in industrially relevant plasmas will benefit from the experience of personnel with extensive knowledge of plasma physics as well as considerable insight and experience in finding and solving simple, tractable problems within a complicated system. NRL has such expertise. An opportunity exists for NRL to draw on this expertise and focus its basic research on understanding the intrinsic behavior of low-temperature plasmas.