Other Discharge Types
Capacitive "diode" plasma reactors are simple to build and versatile. However, they have significant limitations. Increasing RF power doesn't necessarily increase plasma density: especially at low pressures, the power is wasted in increased ion bombardment and hot electron creation instead of contributing to ionization. Plasma potential (the average voltage between the plasma and the walls) can become very high, leading to sputtering of the chamber walls and contamination of the substrates. A number of alternative methods of creating discharges exist to circumvent these limitations.
Magnetic Fields and Plasmas: Magnetrons, MERIE, ECR
An electron moving in a magnetic field experiences a force perpendicular to its direction of motion:
Since the force is always perpendicular to the velocity, the electron travels in a circle "around" lines of magnetic field. Changes in the electron velocity change the radius of the orbit but not the period: the electrons all rotate at the Larmor frequency, whatever their velocity.
B (Gauss) Larmor Frequency 100 280 MHz 1000 2.8 GHz
If the pressure is low enough for electrons to complete their orbits without scattering, an electromagnetic field at the Larmor frequency will be in phase with the electrons (resonance) and add energy on each cycle. This is the principle of the cyclotron accelerator, and of Electron Cyclotron Resonance (ECR) plasmas. (Note in the figure that the electric field is directed opposite to the electron motion in order to achieve acceleration, since the electron charge is negative.)
Imposition of magnetic fields on a plasma "traps" the electrons: they are forced to circle around the field lines rather than diffusing freely to the walls. (The ions have much larger Larmor orbits and shorter mean free paths, so they are less influenced by the field.) The probability that a hot electron will ionize a molecule is increased due to the increased path length. Magnetized plasmas can be sustained at pressures of a few mTorr, where conventional capacitive plasmas are difficult to ignite: this is the principle of magnetron sputtering and magnetically-enhanced reactive-ion etching (MERIE).
Magnetized plasmas tend to have large variations in plasma density, since the strength of the magnetic field varies from place to place. Large variations in plasma potential may also result: the electrons have a hard time moving across the field lines, and thus can't easily move around to compensate for variations in potential. Such inhomogeneities have important implications for plasma damage.
Inductive Plasmas
Wrap a soilenoidal coil around a dielectric chamber (e.g. a quartz tube), and apply an RF voltage. The current flow in the coil generates a magnetic field in the vertical (z) direction:
This time-varying magnetic field creates a time-varying azimuthal electric field (wrapping around the axis of the solenoid). The field strength is proportional to the radial distance and the frequency.
The azimuthal electric field induces a circumferential current in the plasma. The electrons thereby accelerated gain energy, creating enough hot electrons to sustain the plasma through ionization.
Once the plasma forms, the magnetic fields are screened by the induced currents, just as in a metal. In operation, the magnetic field penetrates into the chamber to a depth determined by the magnetic skin depth, which is in turn set by the plasma conductivity and thus by the plasma density and the pressure.
Starting an inductive plasma presents a problem: until the plasma is present, there aren't any free electrons to respond to the induced electric field, thus no power dissipation and no plasma. This problem often solves itself: since there's no plasma, there's not much power dissipation in the coil. With a good matching network, the reactive power stored in the inductance grows quite large, resulting in a large voltage across the coil. Some of this voltage is capacitively coupled to the inner wall of the chamber, creating an electric field. When the field is large enough, conventional breakdown occurs, leading to a plasma supported by the usual capacitive coupling, albeit in an unusual geometry: the system is said to be operating in capacitive mode. As power is increased, the plasma density grows high enough to support induced circumferential currents, and the system switches (often abruptly) to a true inductive mode, typically signified by a sudden increase in brightness of the plasma.
Inductive plasmas generate electrons and ions more efficiently than capacitive plasmas, and can achieve electron densities of 1012 /cm3 at pressures of a few mTorr, as much as 100 times higher than comparable capacitive plasmas. The inductive plasma has a relatively low plasma potential, and results in little ion bombardment of surfaces. By intentionally applying an RF bias voltage to the substrate, one can modify the ion bombardment energy with almost no effect on plasma density. Biased inductive reactors allow independent adjustment of plasma parameters and ion bombardment energy to a much greater extent than dual frequency or conventional single-frequency capacitive reactor designs.
Inductive plasmas are:
- simple: basically you just need a copper coil, power supply, matching network, and cooling: parts cost is low and optimization is (relatively) simple.
- flexible: the coils can be arranged in a solenoid, a flat spiral "pancake" ("TCP") arrangement, or a spiral over a hemispherical dome
- partly capacitive: the RF current flowing in the inductor generates a voltage, some of which is capacitively coupled into the plasma. As we noted above this helps ignite the plasma, but it also causes the plasma potential to increase. You can use a Faraday shield to minimize capacitive coupling but then you may need to provide an ancillary ignition source.
- hot: the coil dissipates a lot of power. Cooling it without shorting it can be difficult. At high pressures, the plasma power will also be concentrated in close proximity to the coil; the combined effects can crack the wall insulator (typically quartz or ceramic).
Electromagnetic Fields in Plasmas
Electromagnetic fields at low frequencies can't propagate in a plasma, but are reflected. At high frequencies the plasma electrons can't respond and fields travel freely. The boundary between these regimes is the plasma frequency. Waves with frequency near the plasma frequency penetrate slightly into the plasma.
The plasma frequency varies from about 100 MHz to a few GHz for typical laboratory plasma densities. (The rarefied plasma of the ionosphere reflects low frequency electromagnetic waves, allowing AM radio stations to reach listeners far beyond their line of sight.)
Note that high density plasmas have very thin "skins": a low frequency wave can not be used to heat the bulk of the plasma.
It is interesting to note that the phase velocity of a wave propagating in the plasma is "infinite" at a frequency just above the plasma frequency, and slows back to the speed of light as the frequency increases. The group velocity (the actual speed at which a signal is carried) is zero at fp and increases with increasing frequency above.
Microwave Plasmas
Use of a high frequency electromagnetic wave (a microwave) allows direct heating of plasma electrons without reflection by the plasma: that is, we choose f > fp so that the wave can enter the plasma readily.
In order for the wave to continuously add energy to the plasma, collisions must occur. Otherwise, the energy received by an electron each half-cycle is taken back by the field on the next half-cycle. The frequency of collisions increases with pressure.
As a consequence, for any particular frequency, there's an optimum operating pressure for maximum power transfer.
Microwave plasmas are usually implemented as an evacuated dielectric tube placed within a microwave resonator. Low losses (high "Q" factors) are needed to achieve the high initial fields required to cause breakdown.
The size of the cavity must be adjusted to support a resonance at the desired operating frequency. If the plasma volume is small, its effects on the resonant frequency are modest and one can estimate the required dimensions from the unperturbed values. For example, the lowest mode of a "tall" cavity (height > radius) is a TE(111) mode, with frequency shown below.
d = height; a = radius;
p11 = first zero of the derivative of the first Bessel function
The mode frequency goes roughly as (c/d) where d is a characteristic size. To get a cavity resonance at 2.45 GHz (a typical industrial frequency) we need dimensions on the order of 6 cm (half a wavelength).
Characteristics of microwave plasmas:
- clean: it is easy to make a system in which only e.g. quartz is exposed to the plasma ("electrodeless"); no metal walls touch the plasma, no high potentials exist to cause sputtering.
- compact: very small sources can easily be made; these are conveniently mounted at the side of a larger chamber to provide reactive gases for cleaning or remote plasma treatment.
- high density: the high frequency waves can continue to penetrate the plasma and heat electrons even when the plasma density is very high. An excitation frequency of 2.45 GHz allows densities to approach 1011/cm3.
- non-uniform: mode fields go to zero and walls and nodes present for higher resonances. It is difficult to create a uniform plasma over a large area.
ECR Plasmas
Recall that electrons in a magnetic field spin in orbit with fixed frequency regardless of speed (and energy); thus, an accelerating field at the resonant frequency continually accelerates all electrons: Electron Cyclotron Resonance. The electrons in the magnetic field spin in circular orbits. Recall that a linearly polarized electromagnetic wave can be regarded as the sum of two circularly polarized waves; the circularly polarized electric field rotating in the same sense as the electrons can continually add energy to them if the microwave frequency and the Larmor frequency agree. This will happen in a resonance region, typically chosen to lie in the center of the solenoidal field.
The plasma electrons are trapped along the magnetic field lines at low pressures, and ions follow them to maintain charge neutrality. Thus ECR plasma properties near the wafer are influenced by the magnetic field configuration. A second coil is often placed near the wafer holder to allow manipulation of the field lines.
Note that, unlike the case of the conventional microwave discharge, no collisions are needed to ensure energy transfer. Thus ECR plasmas can be sustained at very low pressures.
Characteristics of ECR plasmas:
GOOD:
- low pressure OK: Energy doesn't "return" to field as in simple microwave case => frequent collisions not needed, works fine at mTorr pressures.
- uniformity: "Resonant" region determined by magnetic field not by microwave reflections: can have large fairly uniform (flat) plasma.
- high density: Continuous electron heating in resonance region gives very efficient energy transfer to plasma; magnetic field prevents plasma from shielding incoming waves completely even for f
- adjustable: Fields can be adjusted to "tune" uniformity at wafer plane, since electrons and ions diffusing towards the wafer are constrained to the field lines
and BAD:
- transistors crunched: Presence of fields limits lateral electron diffusion: plasma potential can vary from place to place, causing plasma damage [discussed later]
- we said cyclotron, didn't we?: A small number of electrons don't collide, reach very high energies: x-rays!
- complex to operate: the magnetic fields in the plasma create a very complex environment for electromagnetic field propagation; the left-circular waves can also propagate under some circumstances, causing hysteretic behavior and irreproducibility. Reference: "Multiple steady states in electron cyclotron resonance plasma reactors" E. Aydil, J. Gregus and R. Gottscho, J. Vac. Sci. Technol. A11 2883 (1993)
- complex to design and build: to build a reactor big enough to process a large wafer, you need a big electromagnet and big power supply (we're talking fields of 1000 Gauss = .1 Tesla), a high-power microwave generator and matching network, and a huge vacuum pump. Big bucks, lots of things to go wrong.
Return to Tutorial Table of Contents
Book version of the CVD Tutorial