vlsi technology 2
TRANSCRIPT
-
7/27/2019 VLSI technology 2
1/78
Lithography
-
7/27/2019 VLSI technology 2
2/78
Lithography is the process of transferring patterns of geometric
shapes in a mask to a thin layer of radiation-sensitive materials
(called resist) covering the surface of a semiconductor wafer
-
7/27/2019 VLSI technology 2
3/78
Resolution is defined to be the minimum feature
dimension that can be transferred with high fidelity to
a resist film on a semiconductor wafer
Registration is a measure of how accurately patterns
on successive masks can be aligned or overlaid with
respect to previously defined patterns on the samewafer
Throughput is the number of wafers that can beexposed per hour for a given mask level and is thus a
measure of the efficiency of the lithographic process
-
7/27/2019 VLSI technology 2
4/78
An IC fabrication facility requires a clean room, particularly in
lithography areas. Airborne particles settling on semiconductor
wafers and lithographic masks behave as opaque patterns that can besubsequently transferred to the wafers.
Particle 1 may
result in the
formation of a
pinhole in theunderlying layer
Particle 2 may
cause constriction
of current flow inthe metal runner
Particle 3 may lead to a short circuit between the two
conducting regions and render the circuit useless
-
7/27/2019 VLSI technology 2
5/78
Clean RoomIn a clean room, the total
number of dust particles
per unit volume must be
tightly controlled along
with other parameterssuch as temperature,
humidity, pressure, and so
on.
A class X clean room is
usually defined to be one
that has a dust count of
Xparticles (diameters of
0.5 m or larger) per
cubic foot.
For modern lithographic processes, a class 10 or better clean room is required
-
7/27/2019 VLSI technology 2
6/78
Optical Lithography Shadow Printing
Contact printing yields very highresolution (~ 1 m), but suffers from
major drawback caused by dust
particles or silicon specks
accidentally embedded into themask.
Proximity printing is not as prone toparticle damage. However, the small gap
between the mask and wafer (typically 10
m to 50 m) introduces optical diffraction
at the feature edges on the photomasks andthe resolution is degraded.
-
7/27/2019 VLSI technology 2
7/78
Shadow Printing
The minimum linewidth that can be printed, lm
, in shadow
printing is roughly given by
lm = (g)1/2
where is the wavelength of the exposure radiation and g is
the gap between the mask and the wafer and includes thethickness of the resist. For typical values of (~ 0.4 m) and g
(~ 50 m), lm
is on the order of 4.5 m. The above equation
imparts that the minimum linewidth can be improved byreducing the wavelength (that is, going to deep UV spectral
region) or the gap g.
-
7/27/2019 VLSI technology 2
8/78
Projection Printing
In order to circumvent
problems associated with
shadow printing, projectionprinting exposure tools have
been developed to project an
image of the mask patternsonto a resist-coated wafer
many centimeters away from
the mask. The small imagearea is scanned or stepped
over the wafer to cover the
entire surface.
-
7/27/2019 VLSI technology 2
9/78
The resolution of a projection
system is given by
lm
= /NA
where lis the wavelength of the
exposure radiation and NA isthe numerical aperture given by
NA = sin n
where denotes the refraction index of the imaging medium ( = 1 in air) and is
the half angle of the cone of light converging to a point image at the wafer.
n
The depth of focus, z, can be expressed as
z = lm/ 2 tan lm/2 sin = / [2 (NA)2]nResolution can be enhanced by reducing . This explains the trend towards
shorter wavelength in optical lithography.
-
7/27/2019 VLSI technology 2
10/78
Mask Defects
The number of mask defects has a
profound effect on the final IC yield,
which is defined as the ratio of good
chips per wafer to the total number ofchips per wafer. As a first-order
approximation,
Y exp{-DA}where Yis the yield, D is the average
number of "fatal" defects per unit
area, and A is the area of an IC chip.
If D remains the same for all masklevels, N, then
Y exp{-NDA}
-
7/27/2019 VLSI technology 2
11/78
Phase
ShiftMask
The phase-shifting layer that covers adjacent apertures reverses the sign of the
electric field. Although the absolute intensity at the mask is unchanged, the
electric field of these images at the wafer, shown by the dotted line, can be
canceled (right figure). Consequently, images that are projected close to one
another can be separated completely. A 180o phase change occurs when a
transparent layer of thickness d=
/ 2 (n 1), where n is the refraction index and is the wavelength, covers one aperture as shown in the figure.
-
7/27/2019 VLSI technology 2
12/78
A photoresist is a radiation-sensitive compound. For positive resists, the
exposed region becomes more soluble and thus more readily removed in
the developing process. The net result is that the patterns formed in the
positive resist are the same as those on the mask. For negative resists,the exposed regions become less soluble, and the patterns engraved are
the reverse of the mask patterns
A positive photoresist consists of three constituents: photosensitive
compound, base resin, and organic solvent. Prior to exposure, the
photosensitive compound is insoluble in the developer solution. After
irradiation, the photosensitive compound in the exposed pattern areasabsorbs energy, changes its chemical structure, and transforms into a
more soluble species. Upon developing, the exposed areas are expunged
Photoresists
-
7/27/2019 VLSI technology 2
13/78
Negative photoresists are polymers combined with aphotosensitive compound. Following exposure, the
photosensitive compound absorbs the radiation energy and
converts it into chemical energy to initiate a chain reaction,
thereby causing cross-linking of the polymer molecules.
The cross-linked polymer has a higher molecular weight
and becomes insoluble in the developer solution. After
development, the unexposed portions are removed
One major drawback of a negative photoresist is that the
resist absorbs developer solvent and swells, thus limitingthe resolution of a negative photoresist
-
7/27/2019 VLSI technology 2
14/78
The left figure depicts a typical
exposure response curve for a positive
resist. The resist has a finite solubility
in the developer solution even prior to
exposure. At a threshold energy, ET,
the resist becomes completely soluble.
ET
therefore corresponds to the
sensitivity of the photoresist. Another
parameter, , is the contrast ratio and
is given by:1
1ln
= E
ET
A larger implies a more rapid
dissolution of the resist with an
incremental increase of exposure
energy and results in a sharper image.
The edges of the resist image are
generally blurred due to diffraction.
The right figure shows the response
curve for a negative photoresist. Thesensitivity of a negative photoresist is
defined as the energy required to retain
50% of the original resist film thickness
in the exposed region.
-
7/27/2019 VLSI technology 2
15/78
Pattern TransferThe wafer is placed in a clean room
that typically is illuminated with
yellow light as photoresists are not
sensitive to wavelengths greater than
0.5 m. The wafer is held on a
vacuum spindle, and approximately 1
cm3 of liquid resist is applied to the
center of the wafer. The wafer is than
spun for about 30 seconds. The
thickness of the resulting resist film,
lR
, is directly proportional to its
viscosity as well as the percent solid
content indigenous to the resist, and
varies inversely with the spin speed.
For spin speeds in the range of 1000
to 10000 rpm, film thicknesses on the
order of 0.5 to 1 m can be obtained.
-
7/27/2019 VLSI technology 2
16/78
The wafer is then given a pre-exposure bake (80oC to 100oC) to
remove solvent and improve adhesion. The wafer is aligned withrespect to the mask in an optical lithographic system prior to
exposure to UV or deep UV light. For a positive photoresist, the
exposed portions are dissolved in the developer solution. The wafer is
then rinsed, dried, and then put in an ambient that etches the exposedinsulating layer but does not attack the resist. Finally, the resist is
stripped, leaving behind an insulator image (or pattern) that is the
same as the opaque image on the mask. For a negative photoresist,the exposed area becomes insoluble, and the final insulator pattern is
the reverse of the opaque image on the mask.
The insulator image can be employed as a mask for subsequentprocessing. For instance, ion implantation can be performed to dope
the exposed regions selectively.
-
7/27/2019 VLSI technology 2
17/78
Lift-Off Technique
This method works if the film thickness is smaller than that of the
photoresist
-
7/27/2019 VLSI technology 2
18/78
Lithographic Techniques
El t lith h ff hi h l ti b f th ll l th f
-
7/27/2019 VLSI technology 2
19/78
Generation of micron and submicron
resist geometries
Highly automated and preciselycontrolled operation
Greater depth of focus
Direct patterning without a mask
Electron lithography offers high resolution because of the small wavelength of
electrons (< 0.1 nm for 10-50 keV electrons). The resolution is not limited by
diffraction, but by electron scattering in the resist and the various aberrations of the
electron optics. Electron lithography has the following advantages:
The biggest disadvantage of electron
lithography is its low throughput
(approximately 5 wafers / hour at less than
0.1 m resolution). Therefore, electronlithography is primarily used in the
production of photomasks and in situations
that require small number of custom
circuits.
-
7/27/2019 VLSI technology 2
20/78
Electron Resists For a positive electron resist, the
polymer - electron interactioncauses chain scission, that is,
broken chemical bonds. The
irradiated areas can be dissolved
in a developer solution that attackslow molecular - weight material.
Common positive electron resists
are poly (methyl methacrylate),
abbreviated PMMA, and poly(butene-1 sulfone), abbreviated
PBS. Positive electron resists
typically have resolution of 0.1m
or better.When electrons impact a negative electron resist, polymer linking is induced.
Poly (glycidyl methacrylate-co-ethyl-acrylate), abbreviated COP, is a common
negative electron resist. Like a negative photoresist, COP swells during
developing, and resolution is limited to about 1m.
-
7/27/2019 VLSI technology 2
21/78
X-Ray LithographyX-ray lithography employs a shadow
printing method similar to optical
proximity printing. The x-ray
wavelength (0.4 to 5 nm) is much
shorter than that of UV light (200 to
400 nm). Hence, diffraction effects are
reduced and higher resolution can be
attained. For instance, for an x-ray
wavelength of 0.5 nm and a gap of 40
m, lm
is equal to 0.2 m. X-ray
lithography has a higher throughput
when compared to e-beam lithography
because parallel exposure can be
adopted. However, on account of the
finite size of the x-ray source and the
finite mask-to-wafer gap, a penumbral
effect results which degrades the
resolution at the edge of a feature.
-
7/27/2019 VLSI technology 2
22/78
The penumbral blur, , on the edge of the resist image is given by
= ag/L
where a is the diameter of the x-ray source, gis the gap spacing, and L is the
distance from the source to the x-ray mask. Ifa = 3 mm, g= 40 m, and L =
50 cm, is on the order of 0.2 m.
An additional geometric effect is the lateral magnification error due to the
finite mask-to-wafer gap and the non-vertical incidence of the x-ray beam.
The projected images of the mask are shifted laterally by an amount d,
called runout given by
d= rg/L
where rdenotes the radial distance from the center of the wafer. For a 125-mm wafer, the runout error can be as large as 5 m forg= 40 m and L =
50 cm. It is compensated for during the mask making process.
-
7/27/2019 VLSI technology 2
23/78
Ion Lithography
Ion lithography may achieve higher resolution than optical, x-ray, or electron
lithographic techniques because ions undergo no diffraction and scatter much less
than electrons. In addition, resists are more sensitive to ions than to electrons andthere is the possibility of a resistless wafer process. However, ion beams are larger
than electron beams. Due to the slow throughput, the most important application
of ion lithography is the repair of masks for optical or x-ray lithography.
The figure depicts the
computer trajectory of 50
H+
ions implanted at 60keV. The spread of the ion
beam at a depth of 0.4 m
is only 0.1 m.
O ti l lith h i th i t
-
7/27/2019 VLSI technology 2
24/78
Conventional optical
lithography is considered
difficult for a design rule ofmuch less than 0.1 m due to
its resolution limit. For deep
sub-micrometer structures,
the two remaining optionsare electron beam direct
writing or x-ray lithography.
Perfect x-ray masks are
difficult to make and thethroughput of electron
lithography is slow (the
throughput varies as the
reciprocal of the square ofthe minimum feature length).
For mass production, the cost
and footprint (required floor
area) of the machine mustalso be minimized.
Optical lithography is the main stream
technology and some commercial resists can
resolve down to 0.1 m or lower
-
7/27/2019 VLSI technology 2
25/78
Etching
-
7/27/2019 VLSI technology 2
26/78
-
7/27/2019 VLSI technology 2
27/78
Wet Chemical Etching
(1) Transportation of
reactants to the reactingsurface (e.g. by diffusion)
(2) Chemical reactions atthe surface
(3) Transportation of the
products from the surface
(e.g. by diffusion)
-
7/27/2019 VLSI technology 2
28/78
Silicon Wet EtchingFor semiconductor materials, wet chemical etching usually
proceeds by oxidation, accompanied by dissolution of the oxide.
For silicon, the most commonly used etchants are mixtures of
nitric acid (HNO3) and hydrofluoric acid (HF) in water or aceticacid (CH3COOH). The reaction is initiated by promoting
silicon from its initial oxidation state to a higher oxidation state:
Si + 2h+ Si2+
-
7/27/2019 VLSI technology 2
29/78
The holes (h+) are produced by the autocatalytic process:
HNO3 + HNO2 2NO2- + 2h+ + H2O2NO2
- + 2H+ 2HNO2
Si2+ combines with OH- (formed by dissociation of H2O) toform Si(OH)2 which subsequently liberates H2 to form SiO2:
Si(OH)2 SiO
2+ H
2
SiO2 then dissolves in HF:
SiO2
+ 6HF H2
SiF6
+ H2
O
The overall reaction can be written as:
Si + HNO3 + 6HF H2SiF6 + HNO2 + H2O + H2
-
7/27/2019 VLSI technology 2
30/78
Although water can be used as a diluent, acetic acid is
preferred as the dissociation of nitric acid can be retarded inorder to yield a higher concentration of the undissociated
species
At very high HF and low HNO3 concentrations, the etching
rate is controlled by HNO3, because there is an excess
amount of HF to dissolve any SiO2 formed At low HF and high HNO3 concentrations, the etch rate is
controlled by the ability of HF to remove SiO2 as it is being
formed. The etching mechanism is isotropic, that is, not
sensitive to crystallographic orientation
-
7/27/2019 VLSI technology 2
31/78
Insulating / Metal Film Etching
Etching of insulating and metal films is usually
performed with chemicals that dissolve thesematerials in bulk form and involves their
conversion into soluble salts and complexes
Film materials will etch more rapidly than their
bulk counterparts
The etching rates are higher for films that possesspoor microstructures or built-in stress, are non-
stoichiometric, or have been irradiated
-
7/27/2019 VLSI technology 2
32/78
Isotropic / Anisotropic Etching
Wet etching is usually isotropic,undercutting the layer underneath
the mask and resulting in a loss of
resolution in the etched pattern If patterns are required with
resolution much smaller than the
film thickness, anisotropic etching(i.e., 1 > A
f> 0) must be used:
Af
= 1 l/hf
= 1 - vl/v
v
where Af is the degree ofanisotropy, and v
land v
vare the
lateral and vertical etch rates,
respectively. For isotropic etching,
vl= v
vandA
f= 0
-
7/27/2019 VLSI technology 2
33/78
Plasma
A plasma is a fully or partially ionized gas composed
of ions, electrons, and neutrals. The plasma most
useful to ULSI processing is a weakly ionized plasma
called glow discharge containing a significant density
of neutral particles (>90% in most etchers). Althoughplasma is neutral in a macroscopic sense, it behaves
quite differently from a molecular gas, because it
consists of charged particles that can be influenced by
applied electric and magnetic fields
A plasma is produced when an
-
7/27/2019 VLSI technology 2
34/78
p p
electric field is applied across two
electrodes between which a gas is
confined at low pressure, causing
the gas to break down and become
ionized.
Simple DC (direct current) power
can be used to generate plasma,
but insulating materials require
AC (alternate current) power to
reduce charging.
In plasma etching, an RF (radio frequency) field is usually used to
generate the gas discharge. One reason for doing so is that the electrodesdo not have to be made of a conducting material. Besides, electrons can
gain sufficient energy during oscillation to cause more ionization by
electron neutral atom collisions. As a result, the plasma can be
generated at pressures lower than 10-3 Torr.
The free electrons released by photo ionization or field emission from a
-
7/27/2019 VLSI technology 2
35/78
The free electrons released by photo-ionization or field emission from a
negatively biased electrode create the plasma. The free electrons gain
kinetic energy from the applied electric field, and in the course of theirtravel through the gas, they collide with gas molecules and lose energy.
These inelastic collisions serve to further ionize or excite neutral species
in the plasma via the following reaction examples:
e- + AB A- + B+ + e- (Dissociative attachment)e- + AB A + B + e- (Dissociation)e- + A A+ + 2e- (Ionization)
Some of these collisions cause the gas molecules to be ionized and create
more electrons to sustain the plasma. Therefore, when the applied
voltage is larger than the breakdown potential, the plasma is formedthroughout the reaction chamber. Some of these inelastic collisions can
also raise neutrals and ions to excited electronic states that later decay by
photoemission, thereby causing the characteristic plasma glow.
-
7/27/2019 VLSI technology 2
36/78
Interaction of plasmas with surfaces is often
divided into two components: physical andchemical
A physical interaction refers to the surfacebombardment of energetic ions accelerated acrossthe plasma sheath, and the loss of kinetic energy by
the impinging ions causes ejection of particlesfrom the sample surface
Chemical reactions are standard electronic bondingprocesses that result in the formation ordissociation of chemical species on the surface
Plasma Etching Mechanism
-
7/27/2019 VLSI technology 2
37/78
Plasma Etching Mechanism
Sputtering Systems
-
7/27/2019 VLSI technology 2
38/78
Sputtering Systems (Top) Sputtering - etching system: Inertgas ions accelerated by the applied
electric field sputter off the surface.There is essentially no sputtering of the
sidewalls to attain a high degree of
anisotropy, but the selectivity is poor.
In reactive ion etching (RIE), the noble
gas plasma is replaced by a molecular
gas plasma similar to that in plasma
etching. Under appropriate conditions,both RIE and plasma etching can give
high selectivity and a high degree of
anisotropy
(Bottom) Parallel - plate plasma etching system: The plasma is confined between
the two closely spaced electrodes. Molecular gases containing one or more
halogen atoms are fed through the gas ring.
Ion-Assisted The etching rate can be enhanced by
-
7/27/2019 VLSI technology 2
39/78
Reaction
g y
ion-assisted chemical reactions.
This figure depicts the etch rate as a
function of flow rate of XeF2
molecules with and without 1 keV
Ne+ bombardment. The lateral etchrate depends only on the ability of
XeF2
molecules to etch silicon in the
absence of energetic ions impacting
the surface, whereas the vertical etch
rate is a synergistic effect due to both
Ne+ bombardment and XeF2
molecules.
The degree of anisotropy can
generally be enhanced by increasing
the energy of the ions
When a gas is mixed with one or
-
7/27/2019 VLSI technology 2
40/78
When a gas is mixed with one or
more additive gases, both the etch
rate and selectivity can be altered.
In this example, the etching rate of
SiO2
is approximately constant for
addition of up to 40% hydrogen,
while the etch rate for silicon drops
monotonically and is almost zero at
40% H2.
Selectivity (ratio of the etch rate of
silicon dioxide to silicon) exceeding
45:1 can be achieved with CF4 - H2reactive ion etching. This process is
thus useful when etching a SiO2
layer that covers a polysilicon gate.
-
7/27/2019 VLSI technology 2
41/78
Lift Off versus Direct Etching
-
7/27/2019 VLSI technology 2
42/78
Lift-Off versus Direct Etching
Disadvantages of the
lift-off technique:
(1) Rounded feature
profile
(2) Temperature
limitations
(> 200oC to 300oC)
-
7/27/2019 VLSI technology 2
43/78
Polysilicon and Dielectric Film
Deposition
-
7/27/2019 VLSI technology 2
44/78
Thin Films in Microelectronics
Polycrystalline silicon or polysilicon
Doped or undoped silicon dioxide
Stoichiometric or plasma-deposited silicon
nitride
Polysilicon
-
7/27/2019 VLSI technology 2
45/78
Polysilicon
Gate electrode material in MOS devices
Conducting materials for multilevel
metallization
Contact materials for devices with shallow
junctions
Polysilicon can be undoped or doped with elements such as As,
P, or B to reduce the resistivity. The dopant can beincorporated in-situ during deposition, or later by diffusion or
ion implantation. Polysilicon consisting of several percent
oxygen is a semi-insulating material for circuit passivation.
-
7/27/2019 VLSI technology 2
46/78
Dielectric Thin Films
Insulation between conducting layers
Diffusion and ion implantation masks
Diffusion sources (doped oxide)
Capping doped films to prevent dopant loss
Gettering impurities
Passivation to protect devices fromimpurities, moisture, and scratches
Ph h d d ili di id l f d t
-
7/27/2019 VLSI technology 2
47/78
Phosphorus-doped silicon dioxide, commonly referred to
as P-glass or phosphosilicate glass (PSG), is especiallyuseful as a passivation layer because it inhibits the
diffusion of impurities (such as Na), and it softens and
flows at 950o
C to 1100o
C to create a smooth topographythat is beneficial for depositing metals
Borophosphosilicate glass (BPSG), formed by
incorporating both boron and phosphorus into the glass,flows at even lower temperatures between 850oC and
950oC
The smaller phosphorus content in BPSG reduces theseverity of aluminum corrosion in the presence of
moisture
-
7/27/2019 VLSI technology 2
48/78
Silicon nitride is a barrier to sodium diffusion, isnearly impervious to moisture, and has a low
oxidation rate
The local oxidation of silicon (LOCOS) processalso uses silicon nitride as a mask
The patterned silicon nitride will prevent the
underlying silicon from oxidation but leave the
exposed silicon to be oxidized
Silicon nitride is also used as the dielectric forDRAM MOS capacitors when it combines with
silicon dioxide
Physical Vapor Deposition (PVD)
-
7/27/2019 VLSI technology 2
49/78
Physical Vapor Deposition (PVD)
Evaporation: A vapor is first generated by evaporating a
source material in a vacuum chamber and then transported
from the source to the substrate and condensed to a solid filmon the substrate surface.
Sputtering: The process involves the ejection of surfaceatoms from an electrode surface by momentum transfer from
the bombarding ions to the electrode surface atoms. The
generated vapor of electrode material is then deposited on thesubstrate. Sputtering processes, unlike evaporation, are very
well controlled and generally applicable to all materials such
as metals, insulators, semiconductors, and alloys.
Typical Sputtering System
-
7/27/2019 VLSI technology 2
50/78
Typical Sputtering System
Chemical Vapor Deposition (CVD)
-
7/27/2019 VLSI technology 2
51/78
Chemical Vapor Deposition (CVD)
Atmospheric-pressure chemical vapor deposition (APCVD) Low-pressure chemical vapor deposition (LPCVD)
Plasma-enhanced chemical vapor deposition (PECVD)
Since the steps in a CVD process are sequential the one that
-
7/27/2019 VLSI technology 2
52/78
Since the steps in a CVD process are sequential, the one that
occurs at the slowest rate will determine the deposition rateand the rate-determining steps can be grouped into gas-phaseand surface processes
Gas-phase processes dictate the rate at which gases impinge on
the surface and since such transport processes occur by gas-phase diffusion proportional to the diffusivity of the gas andthe concentration gradient across the boundary layer, they are
only weakly influenced by the deposition temperature The surface reaction rate is greatly affected by the depositiontemperature. At low temperature, the surface reaction rate isreduced so much that the arrival rate of reactants can exceed
the rate at which they are consumed by the surface reactionprocess. Under such conditions, the deposition rate is surface-reaction-rate-limited, and at high temperature, it is usuallymass-transport-limited.
Arrhenius Equation
-
7/27/2019 VLSI technology 2
53/78
q
R = A exp {-qEa/kT}where R is the deposition rate, A is the
frequency factor, q is the electronic
charge, Ea
is the activation energy, k is
the Boltzmann's constant, and T is the
absolute temperature
The activation energy calculated from the
slope of the straight-line plots is roughly1.7 eV.
At high temperature, the reaction becomes faster than the rate at which unreacted
silane arrives at the surface. The reaction in this temperature regime is mass-transport limited, as exemplified by the high temperature (or small 1/T) data. The
linear portion of the lines show the surface-reaction limited conditions, that is, the
rate of reaction is slower than the rate of reactant arrival.
Characteristics and Applications of CVD Processes
-
7/27/2019 VLSI technology 2
54/78
Process Advantages Disadvantages Applications
APCVD Simple reactor, Poor step coverage, Doped/undoped
(low T) fast deposition, particle contamination, low T oxides
low temperature low throughput
LPCVD Excellent purity High temperature, Doped/undoped
& uniformity, low deposition rate high T oxides,
conformal step silicon nitride,
coverage, large polysilicon,wafer capacity, tungsten,
high throughput WSi2
PECVD Low temperature, Chemical (e.g. H2
) Passivation
fast deposition, and particle (nitride), low T
good step contamination insulators over
coverage metals
Simple Commercial Hot-Wall, Low-
-
7/27/2019 VLSI technology 2
55/78
S p e Co e c a ot Wa , ow
Pressure Reactor for High WaferCapacity Deposition of Polysilicon
Horizontal Tube
-
7/27/2019 VLSI technology 2
56/78
Horizontal Tube
APCVD Reactor
Gas Injection-TypeContinuous-Processing
APCVD Reactor
Plenum-Type
Continuous-ProcessingAPCVD Reactor
-
7/27/2019 VLSI technology 2
57/78
Parallel Plate
PECVD Reactor
Single Wafer
PECVD Reactor
Typical Commercial PECVD System
-
7/27/2019 VLSI technology 2
58/78
Common Gases Used in CVD
-
7/27/2019 VLSI technology 2
59/78
Flammable
Exposure ToxicLimits in Air Limit
Gas Formula Hazards (vol %) (ppm)
Ammonia NH3 Toxic, corrosive 16-25 25Argon Ar Inert -- --
Arsine AsH3 Toxic -- 0.05
Diborane B2H6 Toxic, flammable 1-98 0.1
Dichlorosilane SiH2Cl2 Toxic, flammable 4-99 5Hydrogen H2 Flammable 4-74 --
Hydrogen chloride HCl Toxic, corrosive -- --
Nitrogen N2 Inert -- --
Nitrogen Oxide N2O Oxidizer -- --
Oxygen O2 Oxidizer -- --
Phosphine PH3 Toxic, flammable Pyrophoric 0.3
Silane SiH4 Toxic, flammable Pyrophoric 0.5
Polysilicon Deposition
-
7/27/2019 VLSI technology 2
60/78
Polysilicon is deposited by pyrolyzing silane between 575oC and 650oC in alow-pressure reaction:
SiH4 (g) Si (s) + 2H2 (gas)
Either pure silane or 20 to 30% silane in nitrogen is bled into the LPCVD
system at a pressure of 0.2 to 1.0 torr. The properties of the LPCVD
polysilicon films are determined by the deposition pressure, silane
concentration, deposition temperature, and dopant content.
Polysilicon can be doped by adding phosphine, arsine, or diborane to the
reactants (in-situ doping). Adding diborane causes a large increase in the
deposition rate because diborane forms borane radicals, BH3, that catalyzegas-phase reactions and increase the deposition rate. In contrast, adding
phosphine or arsine causes a rapid reduction in the deposition rate, because
phosphine or arsine is strongly adsorbed on the silicon substrate surface
thereby inhibiting the dissociative chemisorption of SiH4.
The dopant concentration in diffused polysilicon often exceeds the
solid solubility limit, with the excess dopant atoms segregated at the
-
7/27/2019 VLSI technology 2
61/78
grain boundaries.
The high resistivity
observed for lightly
implanted polysilicon is
caused by carrier traps
at the grain boundaries.
Once these traps aresaturated with dopants,
the resistivity decreases
rapidly and approachesthat for implanted
single-crystal silicon.
Silicon Dioxide Deposition
-
7/27/2019 VLSI technology 2
62/78
Silicon dioxide films can be deposited at lower than 500oC
by reacting silane, dopant (phosphorus in this example), and
oxygen under reduced pressure or atmospheric pressure
SiH4 (g) + O2 (g) SiO2 (s) + 2H2 (g)
4PH3 (g) + 5O2 (g) 2P2O5 (s) + 6H2 (g)
The process can be conducted in an APCVD or LPCVD
chamber. The main advantage of silane-oxygen reactions is
the low deposition temperature allowing films to be depositedover aluminum metallization. The primary disadvantages are
poor step coverage and high particle contamination caused by
loosely adhering deposits on the reactor walls.
Silicon dioxide can be deposited at 650oC to 750oC in an
-
7/27/2019 VLSI technology 2
63/78
p
LPCVD reactor by pyrolyzing tetraethoxysilane,Si(OC2H5)4. This compound, abbreviated TEOS, is
vaporized from a liquid source
Si(OC2H5)4 (l) SiO2 (s) + by-products (g)
The by-products are organic and organosilicon
compounds. LPCVD TEOS is often used to deposit the
spacers beside the polysilicon gates. The process offers
good uniformity and step coverage, but the high
temperature limits its application on aluminuminterconnects.
Sili b d it d b LPCVD t b t 900oC b
-
7/27/2019 VLSI technology 2
64/78
Silicon can be deposited by LPCVD at about 900oC by
reacting dichlorosilane with nitrous oxide
SiCl2H2 (g) + 2N2O (g) SiO2 (s) + 2N2 (g) + 2HCl (g)
This deposition technique provides excellent uniformity,
and like LPCVD TEOS, it is employed to depositinsulating layers over polysilicon. However, this oxide is
frequently contaminated with small amounts of chlorine
that may react with polysilicon causing film cracking.
Plasma-assisted CVD requires the control and optimization of
h d i f d d l i ddi i
-
7/27/2019 VLSI technology 2
65/78
the RF power density, frequency, and duty cycle in addition to
the conditions similar to those of an LPCVD process such as gas
composition, flow rate, deposition temperature, and pressure.
Like LPCVD at low temperature, the PECVD process is
surface-reaction-limited, and adequate substrate temperaturecontrol is necessary to ensure film thickness uniformity.
By reacting silane and oxygen or nitrous oxide in plasma,silicon dioxide films can be formed by the following reactions:
SiH4 (g) + O2 (g) SiO2 (s) + 2H2 (g)
SiH4 (g) + 4N2O (g) SiO2 (s) + 4N2 (g) + 2H2O (g)
Step CoverageA uniform or conformal step coverage depicted in (a)
results when reactants or reactive intermediates
adsorb and then migrate promptly along the surface
-
7/27/2019 VLSI technology 2
66/78
before reacting.When the reactants adsorb and react without
significant surface migration, the deposition rate is
proportional to the arrival angle of the gas molecules.
If the mean free path of the gas is much larger thanthe dimensions of the step (b), the arrival angle and
consequently the film thickness vary. The film is thin
along the vertical walls and may have a crack at the
bottom of the step caused by self-shadowing.
When there is minimal surface mobility and the mean
free path is short, the arrival angle at the top of the
step is 270o
, thus giving a thicker deposit. The arrivalangle at the bottom of the step is only 90o, and so the
film is thin (c). The thick cusp at the top of the step
and the thin crevice at the bottom combine to give a
re-entrant shape that is particularly difficult to coverwith metal.
Reflow
-
7/27/2019 VLSI technology 2
67/78
Doped oxides used as diffusion sources contain 5 to 15
weight % of the dopant
Doped oxides for passivation or interlevel insulation
contain 2 to 8 wt. % phosphorus to prevent the diffusion of
ionic impurities to the device
Phosphosilicate glass (PSG) used for the reflow process
contains 6 to 8 wt. % phosphorus
Reflow
O id ith l h h t ti ill t
-
7/27/2019 VLSI technology 2
68/78
Oxides with lower phosphorus concentrations will not
soften and flow, but higher phosphorus concentrations can
give rise to deleterious effects for phosphorus can react
with atmospheric moisture to form phosphoric acid which
can consequently corrode the aluminum metallization
Addition of boron to PSG further reduces the reflow
temperature without exacerbating this corrosion problem
Borophosphosilicate glass (BPSG) typically contains 4 to 6
wt. % P and 1 to 4 wt. % B
SEM photographs (3200x) showing
surfaces of 4.6 wt. % P-glass annealed
-
7/27/2019 VLSI technology 2
69/78
g
for different times in steam at 1100oC
0 minute 20 minutes
40 minutes 60 minutes
SEM cross sections (10,000x) of samples with
different weight percent of phosphorus
-
7/27/2019 VLSI technology 2
70/78
different weight percent of phosphorus
annealed in steam at 1100oC for 20 minutes
0.0 wt. % P 2.2 wt. % P
4.6 wt. % P 7.2 wt. % P
Planarization The step coverage ofdeposited oxides can be
-
7/27/2019 VLSI technology 2
71/78
p
improved by planarization oretch-back techniques.
Since the organic resistmaterial has a low viscosity,
reflow occurs during
application or the subsequent
bake. The sample is then
plasma etched to remove all
the organic coating and part of
the PSG, as long as theetching conditions are selected
to remove the organic material
and PSG at equal rates.
Silicon Nitride Deposition
-
7/27/2019 VLSI technology 2
72/78
(1) Stoichiometric silicon nitride (Si3N4) can be deposited at
700oC to 800oC at atmospheric pressure:
3SiH4 (g) + 4NH3 (g) Si3N4 (s) + 12H2 (g)
(2) Using LPCVD, silicon nitride can be produced by reactingdichlorosilane and ammonia at temperature between 700oC and
800oC:
3SiCl2H2 (g) + 4NH3 (g) Si3N4 (s) + 6H2 (g)The reduced-pressure technique has the advantage of yielding
good uniformity and higher wafer throughput.
Silicon Nitride Plasma Deposition
-
7/27/2019 VLSI technology 2
73/78
Hydrogenated silicon nitride films can be deposited by reactingsilane and ammonia or nitrogen in plasma at reduced temperature:
SiH4 (g) + NH3 (g) SiN : H (s) + 3H2 (g)
SiH4 (g) + N2 (g) 2SiN : H (s) + 3H2 (g)
Plasma-assisted deposition is conducted at low temperature by
reacting the gases in a glow discharge. Two plasma depositedmaterials, plasma deposited silicon nitride (SiNH) and plasma
deposited silicon dioxide, are useful in VLSI. On account of the
low deposition temperature, 300oC to 350oC, plasma nitride can be
deposited over the final device metallization. Plasma-deposited
films contain large amounts of hydrogen (10 to 35 atomic %)
bonded to silicon as Si-H, to nitrogen as N-H, and to oxygen as Si-
OH and H2O.
Properties of Silicon Nitride Films
-
7/27/2019 VLSI technology 2
74/78
Deposition LPCVD Plasma
Temperature (o
C) 700-800 250-350Composition Si
3N
4(H) SiN
xH
y
Si/N ratio 0.75 0.8-1.2
Atom % H 4-8 20-25
Refractive index 2.01 1.8-2.5Density (g/cm3) 2.9-3.1 2.4-2.8
Dielectric constant 6-7 6-9
Resistivity (ohm-cm) 1016 106-1015
Dielectric strength (106V/cm) 10 5Energy gap (eV) 5 4-5
Stress (109 dyne/cm2) 10T 2C-5T
Rapid Thermal Chemical Vapor
Deposition (RTCVD) System
-
7/27/2019 VLSI technology 2
75/78
Deposition (RTCVD) System
RTCVD requires higher
temperature but shorter time
to deposit the same materialcompared to LPCVD.
Temperature is the switchthat turns the RTP
deposition process on or off,
avoiding the long ramp-up
and ramp-down times
required in conventional
methods.
To be commercially viable, RTCVD systems
-
7/27/2019 VLSI technology 2
76/78
y , y
must be able to process a single wafer in 1 to 2minutes, giving a corresponding throughput of 30to 60 wafers per hour
For relatively thick deposited films of 100 to 200nm, this requires deposition rates greater than 100nm/min
In comparison, the typical deposition rate inconventional batch LPCVD processes is about 10nm/min
The RTCVD of SiO2 by pyrolysis of TEOS is believed to
-
7/27/2019 VLSI technology 2
77/78
occur by the reaction:
Si(OC2H5)4 SiO2 (s) + 2H2O + 4C2H4
Below 800oC, the deposition rate is controlled by surface
reaction processes with activation energy of 3.3 eV. This
large sensitivity to temperature thus demands tight
temperature control for thin oxide deposition at lowtemperature. Above 800oC, the deposition rate of SiO2approaches 100 nm/min with a lower activation energy,
which meets the throughput and deposition controlrequirements of the RTCVD system. The high operating
temperature makes it infeasible for back end steps in
multilevel-metallization technologies that use aluminium.
Thin oxide applications include in-situ deposition of MOS gate
structures where Si surface cleaning, gate oxide formation, and
polysilicon gate electrode deposition would all occur in a low-
-
7/27/2019 VLSI technology 2
78/78
polysilicon gate electrode deposition would all occur in a low
pressure, multi-chamber cluster tool. RTCVD yields high-quality
films through the use of ultraclean gases in a chamber with a highly
regulated ambient, and examples of these reactions are:
SiH4 + O2 SiO2 + 2H2
SiH4 + 2N2O SiO2 + 2N2 + 2H2O
3SiH2Cl2 + 10NH3 Si3N4 + 6NH4Cl + 6H2
3SiH4 + 4NH3 Si3N4 + 12H2
SiH4 Si + 2H2