Plasma Basics Concepts
Plasma Basics Concepts
Plasma Basics
Objectives
After finishing this chapter, the reader will be able to:
• define plasma
• list the three basic components of plasma
• list the three key collisions in plasma and identify their importance
• list the benefits of using plasmas in CVD and etch processes
• identify major differences between plasma-enhanced CVD and plasma etch
processes
• name at least two high-density plasma systems
• define and explain the mean free path and its relationship to pressure
• define and explain the effect of magnetic fields on plasma
• describe ion bombardment and its relationship to plasma processes.
7.1 Introduction
Plasma processes are widely used in semiconductor processing. For example, all
patterned etching in IC fabrication is either a plasma etch or dry etch process.
Plasma-enhanced CVD (PECVD) and high-density plasma CVD (HDP-CVD)
processes are widely used for dielectric depositions. Ion implanters use plasma
sources to generate ions for wafer doping and to provide electrons to neutralize
ions on wafer surfaces. Plasmas are also used in physical vapor deposition
(PVD) processes, in which metal is sputtered from a target surface by ion
bombardment and deposited on a wafer surface. Remote plasma source systems are
commonly used in process chamber cleaning, film stripping, and film deposition
processes. This chapter covers the basic properties of plasma and its applications
in semiconductor processes, focusing mainly on etch and CVD.
237
Downloaded From: http://ebooks.spiedigitallibrary.org/ on 05/30/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx
238 Chapter 7
molecule inside the chamber, generating an ion and another free electron. Due to
the cascading of ionization collisions, the entire chamber fills quickly with equal
numbers of electrons and ions—this is plasma.
In plasma, some electrons and ions are continually lost and consumed by
collisions with electrodes and chamber walls, and also by recombination collisions
between electrons and ions. When the generation rate of electrons through
ionization collisions is equal to the loss rate of electrons, the plasma is said to
be stabilized.
Other plasma sources such as dc plasma, ICP, ECR, and microwave (MW)
remote plasma sources all have very similar plasma generating processes.
Question: Without the first free electron inside a chamber, there is no way to start
plasma. Where does the first free electron come from?
Answer: It is usually generated by a cosmic ray. It also can be generated by heat
(thermal electrons) or by native radioactive decay.
7.3.1 Ionization
When an electron collides with an atom or molecule, it can transfer part of its
energy to the orbital electron confined by the nucleus of that atom or molecule. If
the orbital electron gains enough energy to break free from the constraint of the
nucleus, it becomes a free electron (see Fig. 7.2). This process is called electron
e− + A → A+ + 2e− ,
7.3.2 Excitation-relaxation
Sometimes an orbital electron does not get enough energy from the impact electron
to break free from the constraint of the nucleus. The transferred energy to the
orbital electron from the collision will cause it to jump to a higher energy level of
the orbit. This process is called excitation. An excitation collision can be expressed
as
e − + A → A∗ + e − ,
A∗ → A + hν (photons),
where hν is the energy of the photon, h is Planck’s constant, and ν is the frequency
of the light emission, which determines the color of light emitted from the plasma.
Different atoms or molecules have different orbital structures and energy levels;
therefore, the light emission frequencies are different. This is why different gases
glow in various colors within the plasma. The glow of oxygen is grayish-blue,
nitrogen is pink, neon light is red, a fluorine glow is orange-red, etc.
7.3.3 Dissociation
When an electron collides with a molecule, it can break the chemical bond and
generate free radicals if the energy transferred by the impact to the molecule is
higher than the molecule’s bonding energy. This is called a dissociation collision
e− + AB → A + B + e− ,
e− + CF4 → CF3 + F + e− .
In the PECVD oxide process, silicon precursor silane (SiH4 ) and oxygen
precursor nitrous oxide (N2 O, also known as laughing gas) are used to generate
free radicals:
e− + SiH4 → SiH2 + 2H + e− ,
e− + N 2 O → N 2 + O + e− .
Free radicals such as F, SiH2 , and O are chemically very reactive. This is why
plasma can enhance chemical reactions in both CVD and etch processes.
on the target surface. Ar+ ions sputter TiN molecules from the target
surface and deposit them on a wafer surface. Sputtering deposition of
tantalum nitride (TaN) is very similar to the deposition of TiN.
Question: Which one of the collisions in Table 7.1 is most likely to happen? Why?
Answer: The collision that requires the least energy is the most likely to happen.
It is much easier for an electron to fall to a lower energy level than
climb to a higher one. Under conditions of field strength, pressure, and
temperature, an electron is more likely to be accelerated for a short
distance without a collision to attain enough energy required for that
first collision to occur (2.2 eV). It is least likely for an electron to
travel a much longer distance without a collision to gain enough energy
(15.3 eV) needed for the last reaction in Table 7.1.
1
λ= √ , (7.1)
2nσ
where n is particle density, and σ is the collision cross section. Higher particle
density causes more collisions, which shorten the MFP. Large particles collide
more easily with other particles and therefore have shorter MFPs. As shown in
Eq. (7.1), MFP is determined by chamber pressure, which determines particle
density. Gas (or gases) in the chamber can also affect MFP because different gas
molecules have different sizes or cross-sectional areas.
Question: What is the MFP for a molecule with diameter 3 Å and density
3.5 × 1016 cm−3 (the density of
√ an ideal gas at 1 torr or 1 mm Hg)?
Answer: Based on Eq. (7.1), λ = 1/[ 2 × 3.5 × 10 × π × (3 × 10−8 /2)2 ] =
16
0.029 cm.
Figure 7.6 shows that the MFP is shorter when the gas density is higher
[Fig. 7.6(a)] and longer when gas density is lower [Fig. 7.6(b)]. Large particles
have a larger cross section and sweep through more space. Therefore, they are more
likely to collide with other particles and have shorter MFPs than average sized or
smaller particles. Changing pressure changes particle density, therefore affecting
MFP:
λ ∝ 1/p.
Figure 7.6 (a) High-pressure short MFP and (b) low-pressure long MFP.
For an electron, the MFP can be twice as long, since it has a much smaller size:
F = qE,
power supply such as rf, dc, or microwave. The acceleration of the charged particle
can be expressed as
a = F/m = qE/m,
where m is the mass of the charged particle. Since the mass of electrons is less
than a factor of 1/10,000 compared with ions, electrons can be accelerated more
than 10,000 times faster than ions! This is similar to a motorcycle, which can
accelerate much faster than an 18-wheel truck. It is not very hard to imagine how
fast a motorcycle would start up using a powerful engine from an 18 wheeler, or
to visualize how slow an 18 wheeler powered by a motorcycle engine would creep
away from a traffic stop.
Most etch and CVD plasma sources use rf power. Radio frequency power
generates a varying electric field, which changes directions rapidly. Electrons can
be accelerated rapidly and start collision processes such as ionization, excitation,
and dissociation in the positive cycle of an rf field, and repeat those processes
in the negative cycle. Because ions are too heavy to respond quickly enough to
the varying electric field, electrons absorb most of the rf energy due to their light
weight and quick response. This process is very similar to two vehicles traveling
on a road with stop signs at every intersection. The motorcycle can start up and
slow down very quickly. The big truck starts very slowly and has to slow down
gradually. It is easy to see that the average speed of the motorcycle would be much
higher than the big truck on this kind of road.
When subjected to a lower rf power, ions pick up a little more energy than they
would with a higher rf power. Lower rf power allows more time for ions to respond;
therefore, it can accelerate ions to a higher energy and provide more energy for ion
bombardment.
In any case, electrons always have a much higher temperature than ions or
neutrals in plasma. Thermal velocity can be expressed as
v = (kT/m)1/2 , (7.2)
ve ≈ 4.19 × 107 T e1/2 (T in eV) ≈ 5.93 × 107 cm/sec = 1.33 × 107 mph.
F = qv × B, (7.3)
where q is the charge of the particle, v is the velocity of the particle, and B is
the magnetic field strength. Since a magnetic force is always perpendicular to
particle velocity, a charged particle will spiral around the magnetic field line. This
movement is called gyromotion.
The gyromotion of a charged particle in a magnetic field (see Fig. 7.7) is a
very important property of plasma and has many applications in semiconductor
processing. A number of capacitively coupled plasma etch chambers have magnetic
coils to generate a magnetic field and cause electron gyromotion. This helps to
generate and maintain a higher-density plasma at low pressure. One of the most
commonly used high-density plasma sources is the ECR plasma source, which uses
a magnetic field and microwave power source. When the microwave frequency is
equal to the electron gyrating frequency, the microwave power resonantly couples
to the electrons to generate high-density plasma at very low pressure.
The ion implanter is another processing tool that uses magnets. The strong
magnetic field generated by a dc current in the analyzer magnetic coils of an ion
implanter can bend the trajectories of the high-energy ions. Because ions with
different charge/mass (q/m) ratios have different trajectories in a magnetic field,
they exit the magnetic field at different places. This allows fabs to precisely select
the desired ions and terminate the undesired ones.
The frequency of a charged particle cycling around a magnetic field line is called
the gyrofrequency Ω and can be expressed as
qB
Ω= . (7.4)
m
For a charged particle with a fixed charge and specific mass, the gyrofrequency
is determined only by magnetic field strength B. The gyrofrequency for an electron
is Ωe (MHz) = 2.8B (gauss or G).
The radius of the gyration is called the gyroradius ρ and can be expressed as
ρ = v⊥ /Ω,
where v⊥ is the particle speed perpendicular to the magnetic field line. For an elec-
tron, the gyroradius is ρe (cm) = 2.38T e1/2 /B, where T e is the electron temperature
expressed in electron volts eV, and B is in gauss. Similarly, the ion gyroradius can
be expressed by ρi (cm) = 102 (AT i )1/2 /ZB. Here, A is the atomic weight of the
ion, and Z is the number of the ionization charge of the ion; both values are integer
numbers. Ion mass mion = Am p , where m p is the mass of the proton, which is equal
to 1.67 × 10−27 g. Ion charge q = Ze, where e = 1.6 × 10−19 C is the charge of an
electron.
Question: In an argon sputter etch chamber, the electron temperature is T e ≈ 2 eV,
argon ion temperature is T i ≈ 0.05 eV, and a magnetic field B = 100 G.
For an argon ion, A = 40 and Z = 1. What are the electron and ion
gyroradii?
Answer: Electron gyroradius ρe = 2.38 × 21/2 /100 = 0.034 cm; and argon ion
(Ar+ ) gyroradius ρi = 102 × (40 × 0.05)1/2 /100 = 1.44 cm.
Question: In an ion implanter with an analyzer magnetic field B = 2000 G and
argon ion (Ar+ ) energy E Ar = 200 keV, what is its gyroradius?
Answer: ρi = 102 × (40 × 200, 000)1/2 /2000 = 144 cm.
is initiated. The negatively charged electrodes will repel negative electrons and
attract positive ions; therefore, in the vicinity of the electrodes, there are fewer
electrons than ions.
The difference between the negative and positive charges creates an electric field
in that region, called the sheath potential, as illustrated in Fig. 7.9. Light emission
from this region is less intense than from the bulk plasma, since fewer electrons
are present and fewer excitation-relaxation collisions occur; thus, a dark space can
be observed near the electrodes.
The sheath potential accelerates ions toward the electrodes, causing ion
bombardment. If a wafer is placed over an electrode, ions accelerated by the sheath
potential will bombard its surface.
Ion bombardment is a very important property of plasma. Any conducting
materials close to plasma will be subject to ion bombardment. Ion bombardment
can affect etch rate, selectivity, and profiles in plasma etch processes. It also can
affect deposition rate and film stress in PECVD and HDP-CVD processes.
Ion bombardment has two parameters. One is ion energy, the other is ion flux.
Ion energy is related to external power supply, chamber pressure, electrode spacing,
and processing gases. Ion flux is related to plasma density; it is also determined by
external power supply, chamber pressure, electrode spacing, and processing gases.
In rf plasma systems, rf frequency also affects ion energy. At a high frequency,
such as 13.56 MHz, electrons pick up the majority of the energy, and ions are
left “cold.” At a much lower frequency, such as 350 kHz, although electrons still
acquire most of the energy, ions have a better chance to pick up some energy
from the rf power due to the slower change rate of the electric field. By using the
earlier analogy of the two vehicles, this would be akin to increasing the distance
between stop signs from one every block to one every kilometer. In this case, the
motorcycle (electron) still has the advantage at an average speed because of its
fast acceleration and quick stops, but the average speed of the truck (ion) will be
significantly increased because this scenario enables the big truck to reach and
sustain a high speed.
Question: Why is 13.56 MHz the most commonly used frequency for rf systems?
Answer: Because the government regulates applications of rfs under an interna-
tional treaty, it is necessary to regulate the applications of rfs to avoid
interference with different applications. If some rf interferes with an
air traffic control radio signal, it could have disastrous consequences.
Industrial manufacturing, medical applications, and scientific research
have 13.56 MHz assigned to them. Such rf generators have been com-
mercially available for a long time and are more cost effective than other
rf sources such as 2 MHz, 1.8 MHz, etc.
Figure 7.11 dc biases with (a) lower and (b) higher rf power.
species. When rf power increases, the amplitude of the rf potential will increase
correspondingly, and the plasma potential and dc bias increases (see Fig. 7.11).
Plasma potential depends on the rf power, pressure, and spacing between the
electrodes. For a symmetrical system with two electrodes having the same area
(shown in Fig. 7.12), the dc bias is about 10 to 20 V. Most PECVD systems have
this type of structure. Since rf power also affects plasma density, a capacitively
coupled (parallel plate) plasma source cannot independently control ion energy
and flux.
Figure 7.13 shows the voltage in an asymmetric electrode plasma source. In this
case, the two electrodes have different areas. Current continuity induces a self-bias
when the negative bias voltage builds on the smaller electrode. Sheath voltages
depend on the ratio of the electrode areas, as shown in Fig. 7.14. In an ideal case
(no collisions in the sheath region), voltages and electrode areas have the following
relation:
where V1 is the dc bias, which is the potential difference between the bulk plasma
and the rf hot electrode; and V2 is the plasma potential (plasma to ground). The self-
bias equals V1 − V2 (rf hot to ground). The smaller electrode has a larger sheath
voltage and thus receives more energetic ion bombardment.
Most etch chambers use asymmetric electrodes, and a wafer is always placed on
the smaller rf hot electrode to receive more energetic ion bombardment during
the entire rf cycle. Since etch processes require more ion bombardment, self-
bias on the smaller electrode can help increase ion bombardment energy. For an
rf system with symmetric parallel electrodes, the amount of ion bombardment
on both electrodes is about the same. Therefore, if the wafer receives heavy ion
bombardment, the other electrode will receive the same amount of bombardment,
which is very undesirable, as this can cause particle contamination and short
lifetimes of the processing chamber parts.
Question: If an electrode area ratio is 1:3, what is the difference between the dc
bias and the self-bias, compared to the dc bias?
Answer: The dc bias is V1 , and the self-bias is V1 − V2 ; therefore, the difference
is:
e− + SiH4 → SiH2 + 2H + e−
e− + N 2 O → N 2 + O + e− .
Both SiH2 and O are free radicals with two unpaired electrons; they are very
reactive. On a heated wafer surface, they react very quickly and form silicon
dioxide:
In comparison, the LPCVD silane oxide deposition process uses SiH4 and O2 .
SiH4 thermally dissociates to SiH2 when it comes close to a heated wafer surface.
SiH2 then chemisorbs on the wafer surface, reacts with oxygen, and forms silicon
dioxide:
SiH4 → SiH2 + H2 ,
SiH2 + O2 → SiO2 + other volatiles.
Without plasma, the chemical reaction rate of this CVD process at ∼400 ◦ C is
very low, which causes a very low deposition rate. Thus, low-temperature LPCVD
processes for ILD applications must be performed in a batch system to achieve
reasonable throughput. Table 7.2 compares the properties of PECVD and LPCVD
processes for ILD silane oxide deposition at 400 ◦ C.
e− + CF4 → CF3 + F + e−
e− + C2 F6 → C2 F5 + F + e− .
Atomic fluorine is one of the most reactive free radicals; with ion bombardment
from plasma, it can react with silicon oxide rapidly and form the gaseous
compound SiF4 , which can be easily pumped out of the processing chamber:
SF6 and NF3 are commonly used fluorine-source gases for tungsten CVD
chamber cleaning. Fluorine radicals can react with tungsten to form volatile
tungsten hexafluoride, WF6 , which can be evacuated from the chamber with a
vacuum pump.
Plasma chamber cleaning processes can be automatically endpointed (to prevent
overcleaning) by monitoring the characteristic fluorine light emissions from the
plasma. This is further discussed in Chapter 10.
7.7.1.4 Gap fill
When gaps between metal lines shrink to 0.25 µm with an aspect ratio of 4:1, most
CVD methods cannot fill them without voids. The best known method for filling
such narrow gaps without voids is the HDP-CVD process, shown in Fig. 7.15. The
HDP-CVD process is covered in Chapter 10.
Figure 7.15 HDP-CVD oxide fills a 0.25-µm (aspect ratio 4:1) gap without voids.
plasma uniformity and particle contamination, it replaced the clamp ring in most
advanced plasma systems.
Argon-sputtering etch chambers were widely used in dielectric thin-film
processes for tapering the corners of gaps prior to gap filling and thin-film surface
planarization. Since the sputtering etch rate is relatively insensitive to wafer
temperature, it does not need the helium backside cooling system with a clamp
ring or E-chuck.
bombardment is not necessary in these cases. Remote plasma etch systems are
designed as dry chemical etching systems that can compete with wet etching
processes for these applications. There was a trend toward replacing all wet
processes in IC fabs with dry processes; however, this improvement proved to
be impossible because of the wide application of CMP in advanced IC chip
manufacturing processes.
Like an ICP system, the ECR system also has a bias rf system to control ion
bombardment energy, and an E-chuck with a helium backside cooling system to
control wafer temperature [as shown in Fig. 7.22(b)]. Ion bombardment flux is
mainly controlled by microwave power. One advantage of an ECR system is that,
by changing the electric current in the magnetic coils, the position of resonance can
be adjusted. Thus, process uniformity can be optimized by adjusting the current in
the magnetic coils, which control plasma position.
7.11 Summary
• Plasma consists of ions, electrons, and neutrals.
• Three important collisions are ionization, excitation-relaxation, and disassocia-
tion.
• MFP is the average distance a particle can travel before it collides with another
particle. It is inversely proportional to pressure.
• Free radicals from dissociation collisions enhance chemical reactions in CVD,
etch, and dry cleaning processes.
• Plasma always has higher potential than the electrodes. Higher plasma potential
causes ion bombardment.
• Increasing rf power increases both ion bombardment flux and ion bombardment
energy in capacitively coupled plasmas.
• Low-frequency rf power gives ions more energy, leading to heavier ion
bombardment.
• Etch processes need much more ion bombardment than PECVD processes. Etch
chambers use a magnetic field to increase plasma density at low pressure.
• Capacitively coupled plasma sources cannot generate HDP.
• HDP at low pressure is desirable for both etch and CVD processes.
• ICP and ECR are the two most commonly used HDP sources.
• Both ICP and ECR plasma sources can independently control ion bombarding
flux and energy.
7.12 Bibliography
B. Chapman, Glow Discharge Process, John Wiley & Sons, New York (1980).
F. F. Chen, Introduction to Plasma Physics and Controlled Fusion, Volume 1:
Plasma Physics, 2nd ed., Plenum Press, New York (1984).
S. Dushman, Scientific Foundations of Vacuum Technique, J. M. Lafferty, Ed., John
Wiley and Sons, New York (1962).
M. A. Lieberman and A. J. Lichtenberg Principles of Plasma Discharges and
Materials Processing, John Wiley & Sons, New York (1994).
11. Why does an etch chamber need a helium backside cooling system with an
E-chuck?
12. What are the advantages of an E-chuck over a clamp ring?
13. When the etch rate goes awry in a plasma etch system, why do fabs always
check the rf system first?
14. Can a capacitively coupled plasma source generate HDP?
15. List the two most commonly used HDP systems.