Hoffman D.M., Singh B., Thomas J.H. (Eds). Handbook of Vacuum Science and Technology
Подождите немного. Документ загружается.
634
Chapter 5.2: Plasma Etching
Rg.3.
Plasma region( generation
of
ionic
and
neutral species)^
®
®-
• Reaction by-products
to pump
v//////////A
m
^Substrate
or
underlying layer^
Sheath region
(acceleration
of
ionic species)
(diffusion
of
neutral species)
Surface being etched (reaction) Masking layer
Ionic Species
©
Reactive neutral species
%
Reaction by-product
%
Primary processes occurring
in a
plasma etch process.
The layers
to be
etched
for
fabrication
of
semiconductor integrated circuits
can
be defined
in
three broad categories: conductors, dielectrics,
and
masking layers.
Doped polysilicon, silicides,
and
various metal layers such
as
aluminum, tita-
nium, titanium nitride, titanium tungsten,
and
indium
tin
oxide
(ITO) are
very
commonly used conductor layers. Silicon oxide
and
silicon nitride
are
very com-
monly used dielectric layers. The masking layer is often
a
photoresist layer, which
is patterned using standard photolithographic techniques.
For plasma etching
of
these layers,
the
feed
gas is
selected
so as to
generate
species that react chemically with
the
material
to be
etched,
and
whose reaction
product with
the
etched material
is
volatile.
At the
same time, removal
of the
underlying
or
masking layers
may be
negligible.
A
large variety
of
fluorine-,
chlorine-, bromine-,
and
oxygen-based etching plasmas with
a
profusion
of gas
additives
are
used
in the
semiconductor industry.
In
most cases,
a
specific
gas
mixture
is
based
on a
great deal
of
empirical evidence obtained
for a
particular
application rather than real fundamental understanding
of the
relevant plasma
chemistry. Nevertheless, certain basic insights have proven
to be
helpful
in
for-
mulating
gas
mixtures.
The
commonly used
gas
chemistries
for
etching various
materials
are
listed
in
Table
1.
A large variety
of
other primary etchant
and
additive (such
as
polymer-forming
gases)
gas
mixtures
can
also
be
used,
and the
specific formulation
can
differ from
one user
or
tool
to
another [15].
For
any
particular etch process,
a
number
of
requirements must
be met
such
as
throughput (high etch rate), high selectivity, uniformity
of
etch rate, control
of
the
etched profile (degree
of
anisotropy), control
of
critical dimensions (CD control).
5.2.3 Basic Plasma Etching Requirements
635
Table
1
Trends in Plasma Etch Chemistry
Poly-Si
Single-crystal silicon
Al, Al-Si, Al-Si-Cu
TiW, TIN,
Ti
WSi, TiSi, CoSi
Si02
Si3N4
GaAs
InP
Chemistry
CI2,
HBr, SF6/O2, CF4/O2, BCI3, SiCU
CI2,
SFe, HBr, SiCU
CI2,
BCI3, SiCl4, CHCI3 (N2 or He dihgent)
CF4,SF6,Cl2
CF4, SF6, CI2
CHF3,
C2F6, CF4/H2
CF4/O2,
SF6/O2, CHF3
SiCl4/SF6, SiCl4/NF3, SiCl4/CF4
CH4/H2, HI
low damage to the devices, low particulate or contamination, residue-free surface
after etch, process repeatability, and safety.
5.2.3.1
High Etch Rate
High etch rate is usually needed to keep the throughput (number of wafers etched
in a unit time) of the plasma etch system high. There is usually a tradeoff between
etch rate and other parameters such as selectivity and damage. For example, higher
etch rate can be obtained by increasing the pressure so that more free radicals are
available for etch. However, high pressure may cause more undercut or polymer
formation. Higher power generally also results in higher etch rate, but it can cre-
ate more damage to the substrate and may result in a less selective process.
5.2.3.2
Etch Profile Control or Anisotropy
Controlling the shape and dimension of the etched wall is crucial for realizing
very highly dense submicron circuit patterns. The term anisotropic etching gen-
erally refers to etching a feature with vertical side walls, with no lateral removal.
The degree of anisotropy (Figure 4) can be defined by relating the etching in ver-
tical and lateral directions as
A = 1 - //v (1)
where v is the thickness of the layer etched in the vertical dimension
/ is the lateral distance etched underneath the mask
Majority of the etch processes used today require the patterns to be etched
to have much smaller dimensions than the thickness of the film. This requirement
necessitates the etch process to be anisotropic. Anisotropy can be induced by en-
636
Chapter
S.2:
Plasma Etching
Rg.4.
Thin film
^Substrate or underlying layei^^
J Thin film
^^^^^^Substrate or underlying layer '
^^5ubstrate or underlying layer^^
Anisotropic etch profile
A=l
Illustration
of
anisotropic
and
isotropic
etch
profiles.
Isotropic etch profile
A=0
ergetic ion-bombardment-enhanced reaction, by inhibitor layer formation on the
side walls, by lowering the pressure, or by reducing the temperature of the wafer.
ENERGETIC ION BOMBARDMENT ETCHING
A good example of ion-bombardment-enhanced etching is the etching of Si in
chlorine. An undoped single-crystal Si surface is not etched by CI2 or CI atoms
at room temperature [16]. When the surface is simultaneously exposed to high-
energy ion flux, silicon chlorides are formed in a rapid reaction. Normally, higher
the energy of the ions bombarding the surface, the etched profiles are more
anisotropic. However, higher ion bombardment energy can result in poor selectiv-
ity and damage.
INHIBITOR ION-ENHANCED ETCHING
The inhibitor ion-enhanced etching requires two conceptually different species:
etchants and inhibitors. The inhibitor species form a thin film on the surfaces be-
ing etched. Only the bottom surfaces experience energetic ion bombardment, due
to directionality of ions accelerated by the sheath. The film is continuously re-
moved from the bottom surfaces by sputtering, and etching takes place. Vertical
side walls experience little or no ion bombardment, and no etching takes place on
vertical side walls. The resultant etch profile is anisotropic. In fluorocarbon plas-
mas,
CF2 radicals can be used to form thin polymeric inhibitor films [17]. The
concept of inhibitor-ion-enhanced etching or "side wall passivation" for achiev-
ing high anisotropy is illustrated in Figure 5.
LOW-PRESSURE ETCHING
Lowering the pressure of the reactor, in general, improves the degree of anisot-
ropy. This is due to the fact that the mean free path increases with decreasing
pressure. As the mean free path is increased, the scattering in ion directionality
and energy in the sheath region is minimized and better anisotropy is achieved.
5.2.3
Basic Plasma Etching Requirements
637
Fig.
5.
Radical reaction
not
occurring
due to
side-wall polymer film
Radical reaction occurring
Thin film
iSubstrate
or
underlying layer!
Anisotropy
by
polymeric inhibitor films.
EFFECT
OF
WAFER TEMPERATURE
Any basic etch process
can be
categorized
in two
simple components: chemical
reaction
by
reactive radicals (isotropic)
and
ion-assisted reaction (anisotropic).
By lowering
the
temperature
of the
substrate
and
hence
the
surface
on
which
re-
action takes place,
the
chemical component
of
etch
can be
minimized. Only
the
bottom
of the
features that experience directional
ion
flux
are
etched. With little
or
no ion
bombardment
on the
sidewall,
the
resultant lateral etch rate
is
very slow,
because the chemical etch rate
is
reduced. The concept
of
achieving anisotropy
by
lowering substrate temperature
is
illustrated
in
Figure
6.
Fig.
6.
Radical reaction not occurring
Radical reaction occurring
/
K
•
•
Thin film
M\-MJ^
Thin film
l^;^^///////////^^^^^
'^^7^7777777777777777777777777777777777777^^
VM^////////////////////////////////////////,
^Substrate
or
underlying layei^^ ^^Substrate
or
underlying laye]
Anisotropic etch profile Isotropic etch profile
Radical
•
Anisotropy
at lower
temperature.
Positive
ion
638 Chapter 5.2: Plasma Etching
The concept of low-temperature etch is similar to the side wall inhibitor film
approach. However, the chemistries or the feed gas mixture that can be used are
much simpler. The disadvantages of the low-temperature etch are lower etch
rates,
condensation on the wafer, and complexity involved in the design and oper-
ation of the wafer holder for maintaining high throughput of the system.
5.233 Selectivity
Selectivity is defined as the etch rate ratio of the material to be etched to the over-
laying masking layer or the underlying material. Having a process with high se-
lectivity is critical in cases where a significant overetch (additional etch time after
the nominal thickness of the layer is etched away) is required to remove all mate-
rials from all topological features. For example, during formation of contacts
to the substrate and polysilicon layers, the glass layer can have different thickness
in different regions in a single masking step. The etch process must be able to etch
the thickest layer without damaging the underlying layers. High selectivity is also
critical during etching of gate stacks where the stopping layer may be a very thin
oxide (~100A) [18].
In an ideal chemical etching, the only function of the plasma is to maintain a
supply of gaseous etchant species. Chemical etching is the most selective mecha-
nism, because unwanted reactions will not take place at all when the thermo-
dynamics are unfavorable. In reality, a combination of chemical and physical
(such as ion-bombardment-enhanced) etch mechanisms is used to obtain the de-
sirable etch rate, selectivity, and anisotropy.
In many etch processes, higher selectivity is obtained by adding nonetchant
gases to the basic etchant chemistry. A classical example of using gas additives to
obtain desired selectivity is the addition of
O2
or H2 to the CF4 feed gas. The ad-
dition of small amounts of O2 to CF4 plasma is known to increase the F radical
concentration in the discharge, dramatically increasing both the etch rate of Si
etching and selectivity of Si to oxide etching. This is caused by the reaction of
oxygen with CFx radicals, forming CO, CO2, and COF2 and producing more free
fluorine by reducing the recombination of F atoms with
CF3
[19].
When H2 is added to CF4, the Si etch rate decreases monotonically as the per-
centage of H2 is raised, because of formation of HF by gas-phase reaction be-
tween atomic H and atomic F
[20].
Also, adding H2 increases selective deposition
of polymer on Si surface. In this case, the etch rate of the underlying layer is de-
termined by a balance between polymer deposition and etching. Similar results
can also be obtained by using C2F6/H2 or CHF3 gas chemistries, and these gases
are widely used to etch Si02 more selective to Si.
Chlorine- and bromine-based chemistries are also widely used for Si and Al
etching. Chlorine and bromine atoms have relatively high selectivity to the under-
5.23 Basic Plasma Etching Requirements 639
lying oxide layers; chlorine is more widely used because it has much higher vapor
pressure, is less corrosive, and tends to form somewhat less toxic by-products. O2
is widely used for photoresist stripping without damaging underlying layers and
for cleaning polymers from the reactor walls.
5.2.3.4 Etch Uniformity
Good etch uniformity (across the wafer and also from wafer to wafer) is a very
desired feature for any etch process. Uniformity problems often worsen as wafer
size increases or new high-etch-rate processes are developed. Nonuniform etch-
ing necessitates extensive overetching and creates a stringent demand for a high-
selectivity process.
Nonuniformity can be caused by depletion of etchant concentration or gradi-
ents in ion bombardment flux and energy [21]. For an etch process to be uniform,
diffusion of the reactant species and the ion flux to the wafer surface must be uni-
form. The removal of the volatile by-product by the pumping system also must be
uniform. In many reactor configurations, nonuniform etching tends to develop at
the boundary of the wafer. This is commonly known as the "bull's-eye" clearing
pattern, where the etch rate monotonically decreases from the wafer periphery to
its center. The other nonuniformity-related common problem is known as micro-
loading or aspect ratio dependent etching (ARDE) [22], which results from varia-
tion of etch rates across the wafer due to the feature size and density. Normally,
the smaller and denser the feature becomes, the slower it etches. This phenome-
non arises from the fact that the mean free path of the etchant species can be com-
parable to the geometry being etched. The flux of etchant species to the surface
and diffusion of the volatile by-product back into plasma can be disturbed by the
high-aspect-ratio (ratio of the width to the depth) features.
In practice, pressure, power, interelectrode distance, and gas flow are interac-
tively adjusted for a specific application until an acceptable uniformity is achieved.
This empirical approach is taken because the underlying mechanisms responsible
for nonuniform etching are not known quantitatively.
Another term that is used very often in production to assess uniformity is criti-
cal dimension or CD control. In production, the line widths of etched profiles are
measured routinely, and any significant statistical variations in critical dimension
are often related to a machine or a process problem. Control of line width is cru-
cial for high-yield manufacturing. For example, any major deviations in line width
of a etched gate layer can cause significant variations of the leakage or drive cur-
rents of the metal-oxide-semiconductor (MOS) transistors.
Also,
note that uniform control of wafer temperature is required for many ap-
plications for maintaining uniformity and reproducibility of the etch rate and etch
profiles [23,24]. In many reactor configurations, this is achieved by clamping the
640 Chapter 5.2: Plasma Etching
wafer to the chuck and by applying He gas on the backside of
the
wafer to provide
thermal conductivity between the wafer and the temperature-controlled chuck.
5.2.3.5 Damage
For high-yield integrated-circuit manufacturing, it is desirable to have an etch
process that causes no or minimal damage to the wafers. In plasma etch processes,
positive ions from plasma volume are accelerated by the sheath voltage. These
ions can impinge on the surface of the semiconductor wafer with energies in ex-
cess of a few hundred eVs and can cause physical damage by sputtering. In gen-
eral, the bias on the wafer is optimized to obtain desired etch results with minimal
damage. Also, the high level of UV radiation present in the plasma can generate
defects in the semiconductor oxides [25].
Various features on semiconductor wafers are usually exposed to plasma di-
rectly and collect charges from the plasma. The induced electrostatic stress on
isolated features due to any nonuniformity in plasma can also degrade the quality
of gate oxide. The stress can cause surface states at Si02-Si interface as well as
trapped charges in the oxide, decreasing the breakdown voltage and deforming
the CV characteristics of the gate oxide [26].
The exact mechanisms responsible for plasma-etch-induced damage are not
well understood at present. Special antenna test structures are commonly incor-
porated on semiconductor wafers to aid process development and to minimize
yield loss due to plasma-induced damage [27].
5.2.3.6 Particulate or Contamination Control
Particulate or contamination is a major concern in modern high-density integrated-
circuit fabrication facilities. Particulates can severely affect the yield and reliabil-
ity of the devices. With increasing size of dies and wafers, this can translate into
a major financial concern. Plasma processes are among the "dirtiest" processes
among the various processes (such as implantation, diffusion, photolithography,
and so on) used for semiconductor circuit fabrication [28]. In plasma etching
apparatus, the by-product films deposited on chamber wall and electrodes can
fracture and can generate particulate. Often, a plasma clean step is used between
processing of every wafer, or of a selected number of wafers, to minimize particle
generation. Particles can also be formed by gas phase reactions within a plasma
volume [29]. Particles present in plasma can acquire negative charge from elec-
trons and can be suspended at the sheath boundary by electrostatic repulsion. As
the discharge is turned off, these particles can fall on the wafer. Sputtering of ma-
terials from the reactor walls by the discharge is another concern.
5.2.4 Plasma Diagnostics 641
In recent years, the reactor designs have been significantly modified to mini-
mize particles. The reactor materials are selected to be inert to process reactants
and with high resistance to sputtering. Particle traps such as the grooved electrode
design are also being incorporated in some reactor designs to electrostatically
trap the charged particles [30]. "Cleaner" plasma chemistries are also preferred,
to minimize particle formation.
Another source of contamination is the polymers formed on the side walls of
the features on the wafers
itself.
In many cases, the wafers are subjected to a wet
chemical clean followed by a plasma etch step to remove these polymers.
During the etching of Al-based metallization, the chlorine residue from the etch
process can cause serious postetch corrosion and can affect yield, failure rate, de-
vice performance, and its long-term reliability. Many techniques, such as water or
solvent rinse [31], passivation using fluorocarbon plasma [32], annealing or oxi-
dation [33], and nitric acid treatment [34] are generally used to reduce postetch
corrosion.
5.2.4
PLASMA DIAGNOSTICS
Over the years, various plasma diagnostic techniques have been used to elucidate
some fundamental properties of the plasma processes. Some of these basic tech-
niques are summarized next.
IN SITU ETCH RATE MEASUREMENTS
The two major techniques used for in situ measurements of etch rates are laser
reflection interferometry and ellipsometry. In laser reflection interferometry, a
laser beam is reflected back from an unpatterned area on the wafer. The interfer-
ence pattern originating from the thickness change of the layer being etched is
used to measure the etch rate [35]. Standard ellipsometric techniques have also
been applied to measure in situ the etch rate of the film being etched.
PLASMA IMPEDANCE MEASUREMENTS
Radio frequency current and voltage probes are often incorporated on various re-
actors for measuring plasma impedance and true power dissipated [36,37]. Since
these probes are installed externally, they are nonevasive, making the measure-
ments useful for process monitoring and tool fingerprinting. For inferring the
plasma characteristics from the externally measured parameters, it is necessary to
account for stray impedances between the point of measurement and the plasma-
electrode interface.
642 Chapter 5.2: Plasma Etching
OPTICAL EMISSION SPECTROSCOPY
Emission spectroscopy has played an important role in plasma diagnosis [38,39].
In optical emission spectroscopy, the optical spectrum emitted by etch plasma is
analyzed. This is generally accomplished by a detector and an optical system con-
sisting of
a
monocrometer. Various species present in the plasma can be identified
by analyzing the spectrum and the concentration of various species can be in-
ferred qualitatively.
LASER ABSORPTION SPECTROSCOPY
Various laser-based absorption techniques have been used to determine concen-
trations of various species in the plasma [40]. This is achieved by using wave-
length-tunable lasers and by monitoring absorptance at
a
characteristic wavelength
of a specific molecule. Various frequency modulation schemes have also been ap-
plied to enhance the detection limit of this technique [41].
MASS AND ENERGY ANALYZER
Mass and energy sampling of glow discharges has yielded a rich source of infor-
mation for developing some fundamental understanding of overall discharge be-
havior [42,43]. A mass/energy analyzer normally samples the discharge through
a small orifice and provides information about mass and energy of various ionic
and neutral species in the plasma. These analyzers are normally differentially
pumped when used in conjunction with process plasma.
MICROWAVE INTERFEROMETRY
Microwave interferometry has been used to infer electron properties in process
plasma by measuring interference to the applied microwaves propagating through
the plasma. In this technique, the applied frequency is selected to be higher than
the electron plasma frequency, and the resulting phase shift, which is proportional
to the average electron density, the path length, and the wavelength, is measured
[44].
LANGMUIR PROBES
Electrical measurements using a Langmuir probe is a powerful and experimen-
tally simple means of determining key internal discharge parameters, such as
charged particle concentrations, plasma potential, and electron energy distribu-
tion function (EEDF). Since the pioneering work of Irving Langmuir nearly sev-
enty years ago [45], numerous papers and excellent reviews of the subject have
5.2.5 Basic Plasma Etch Reactors
643
been published [46,47]. Basically, an electrical probe consisting of one or more
small metallic electrodes is inserted into the plasma. The I-V characteristic of the
probe is then used to deduce important plasma parameters.
LASER-INDUCED FLUORESCENCE
Laser-induced fluorescence has been used by many researchers to yield informa-
tion on various species present in the plasma [14]. Normally, a laser is used to
excite the species, and the resultant fluorescence is detected. Measurement of
fluorescence spectrum is used to identify laser-excited species. The state, temper-
ature, and density of these excited species can be inferred from the spectrum.
5.2.5
BASIC PLASMA ETCH REACTORS
5.2.5.1 Introduction
The ultimate goal for an etch tool would be to translate the etch requirements into
proper settings for the tool control parameters—e.g., power, pressure, feed gas
mixture, flow, bias voltages,
etc.
— and then proceed to etch with closed-loop
control. Plasma-activated processes are, however, far more complex, involving a
myriad of nonequilibrium mechanisms. The resultant etching rates and profiles
depend on the feed gas, gas flow, power, pressure, substrate temperature, reactor
geometry, and many other parameters in a fairly complex and interrelated man-
ner. In a practical etching process, the end results are optimized through an artful
balance of these parameters. The underlying mechanisms in plasma processes are
not well understood, and there is much less understanding of how the tool control
parameters relate to the plasma discharge characteristics that actually determine
the etch process. These effects further add to the complexity in determining the
connections among the tool control parameters, the plasma discharge parameters,
and the etch characteristics. There is, at present, only an empirical understanding
of how to set the tool control parameters to achieve a given set of etch character-
istics.
The development and optimization of various reactor configurations have
been based on this empirical understanding. The basic considerations and inter-
actions for development of an etch tool are illustrated in Figure 7.
The requirement for larger wafers and thinner oxides for modern high-density
high-yield high-throughput integrated-circuit manufacturing is also putting strin-
gent demands on manufacturing requirements such as cost of ownership (COO),
up-time, reliability, process control, ease of process development, and automa-