22nm Planar Logic Process Flow Guide
22nm Planar Logic Process Flow Guide
Presented
by
Threshold Systems Inc.
All rights are reserved. No part of this publication may not be reproduced, stored in a retrieval
system, or transmitted, in any form or by any means , electronic, mechanical, photocopying,
recording or otherwise, without the prior written permission of Threshold Systems Inc.
.
nc
sI
m
ste
Sy
ld
hos
re
Th
16
20
©
© 2018 Threshold Systems Inc.
Presented
by
Jerry Healey
Oxide
Ox
.
nc
sI
Ox
m
ste
Ox
Sy
P+ Sil
icon <
ld
1 Ox
ho
00>
s
re
Th
16
20
©
© 2018 Threshold Systems Inc.
Shallow Trench
Isolation (STI)
A Transistor A Shallow Trench
TEOS SiC SiC TEOS SiGe SiGe TEOS SiC SiC TEOS SiGe SiGe TEOS
Epitaxial Silicon
P+ Silicon <100>
P-Well
The starting material for this type of process consists of <100> heavily doped
P+ Silicon with a typical resistively of 0.01 Ω-cm. The front-side of the wafers
are covered in a 2 µm thick layer of Epitaxial Silicon which has a resistively of
approximately 14 Ω-cm.
Epita
xial S
sI
P+ Sil
m
icon <
ste
100>
Sy
ld
ho
s
re
Th
16
20
©
© 2018 Threshold Systems Inc.
A “wet hood” that contains four cleaning stations.
Pad Oxide
The wafers are cleaned with Piranha + HF + SC1+ SC2 and 100Å of pad
oxide is grown across their surface.
The most common cleaning agents are:
Piranha - to remove organics
Epita
xial S
.
nc
P+ Sil
m
icon <
ste
Silico
n Nitr
ide
Epita
xial S
.
nc
ilicon
sI
P+ Sil
m
icon <
ste
100>
Sy
ld
hos
re
Th
16
20
©
BARC
2,000 Å thick layer of amorphous Carbon is deposited across the
wafers surface using CVD. This is followed by a 500Å layer of
Bottom Anti-Reflective Coating (BARC). The amorphous carbon
will act as a hard mask.
Photor
es ist Next, a ~3,000Å of photoresist is spun onto the wafers and soft
Amorp baked.
h ous Ca
rbon H
Silicon ard Ma
sk
Nitrid
e
Epita
xial S
.
nc
ilicon
sI
m
P+ Sil
ste
icon <
100>
Sy
ld
hos
re
Th
16
20
©
© 2018 Threshold Systems Inc.
UV Light UV Light
Quartz Glass The wafers are exposed to 193 nm UV light through a photomask
Photomask that is made of transparent quartz and patterned with chrome
that is opaque to UV light.
Chrome Chrome Chrome Chrome
Photoresist BARC
Silicon Nitride
Epitaxial Silicon
P+ Silicon <100>
© 2016 Threshold Systems Inc.
The Lithography tools that print microchip circuits are extremely sophisticated and Lithography Equipment
are the most precise instruments in the world. They are capable of printing images as
small as 20 nm and aligning to accuracies of less than 1 nm.
They are called “scanners” because the illuminating light scans across a chip printing
one microchip at a time. (They used to be called “steppers”.)
Photolithography scanners employ excimer laser light sources at wavelengths of
193nm and super-sophisticated optical optical pathways that terminate in a column
of precisely calibrated lenses.
© 2010 ASML
The lens column shrinks the precise patten on the reticle (mask) and prints it into the
photoresist located on the wafer’s surface.
Through the use of Resolution Enhancement Techniques (RET), Optical Proximity
Correction (OPC) and other optical processing, modern scanners are able to print
images smaller than the wavelength of the illumination light (193nm) being used.
A state-of-the-art scanners can cost as much as 70 million dollars and can process up
A 193i immersion scanner.
to 200-250 wafers per hour. A modern semiconductor fab may require 20-to-30
scanners.
All process flows at the 20nm node employ 193nm immersion scanners which are
denoted as “193i”.
© 2018 Threshold Systems Inc.
A Photomask A Lithography Lens
© Carl Zeiss
© Courtesy of Zeiss
UV Light
Chrome Chrome
BARC
After patterning the wafers are soft-baked to give the resist a measure of
structural integrity and then the resist is exposed and developed. The
PR resulting photoresist pattern then undergoes a Post Exposure UV Bake
PR (PEB) to finalize the photoresist pattern and to turn the photoresist into a
PR cross-linked imidized mass that will resist etching.
Amorp PR
h ous Ca
rbon H
Silicon ard Ma
sk
Nitrid
e
Epita
xial S
.
nc
ilicon
sI
P+ Sil
m
icon <
ste
100>
Sy
ld
sho
re
Th
16
20
©
Epita
20
xial S
ilicon
s©
P+ Sil
m
icon <
ste
100>
Sy
ld
hos
re
Th
13
20
©
Patterned Carbon
Hard Mask
The photoresist and the BARC are removed leaving behind the patterned
amorphous carbon hard mask. The wafers are cleaned in Piranha to
remove any residual photoresist.
Silicon
Nitrid
e
Epita
xial S
ilicon
.
nc
sI
P+ Sil
m
icon <
ste
100>
Sy
ld
hos
re
Th
16
20
©
Silicon
nc
Trench
sI
P+ Sil
m
icon <
ste
100>
Sy
ld
sho
re
Th
16
20
©
Silicon Nitride The amorphous carbon hard mask is selectively stripped away and the
Silicon Nitride
wafers are cleaned in Piranha.
Epitaxial Silicon
P+ Silicon <100>
P+ Sil
m
icon <
ste
100>
Sy
d
ol
sh
re
Th
16
20
©
The surface of the oxide is then cleaned in P/SC1/SC2 and 4,000 Å of TEOS
(Tetra-Ethyl-Ortho-Silicate) oxide is deposited. TEOS is an oxide deposited
by CVD, as opposed to a grown (thermal) oxide. Its key advantage is that it
allows thick layers of oxide to be deposited quickly. It main disadvantage
is that a deposited TEOS oxide is not as dense as a grown oxide.
TEOS O
xide The wafers are then heated at 1,000 ˚C for 20 minutes to densify the TEOS
Nitrid
e and make it more resistant to wet etches.
Nitrid
e
Nitrid
e
Nitrid
e
.
nc
sI
m
P+ Sil
ste
icon <
100>
Sy
ld
Epitaxial
hos
re
Silicon
Th
16
20 © 2018 Threshold Systems Inc.
©
Polished Nitride
The wafers are polished back using CMP. The very hard silicon nitride layer
acts as a CMP stop. After the polish step the wafers are cleaned in P/SC1.
After the oxide has been polished, the thin layer of nitride on the surface
Oxide that acted as a CMP stop must be removed.
Ox
Ox
Ox
.
nc
sI
P+ Sil
m
icon < Ox
ste
100>
Sy
Epitaxial
ld
ho
Silicon
s
re
Th
16
20
©
There are a number of operations throughout the semiconductor manufacturing process CMP Planarization
where the surface of the wafers needs to be smoothed or “planarized.”
Planarization is required because the many oxide and Copper deposition operations create
Rough, non-planar surface
a rough upper surface on the wafer which is referred to as “topographical variation”.
Because microchip manufacturing consists of depositing and patterning a series of
structures all stacked on top of each other (known as the “planar” process) it is important Oxide
Oxide
Polishing
Pad
Platen
1. TEOS Oxide 5.
Epitaxial Silicon
P+ Silicon <100>
2.
© 2018 Threshold
6.
Systems Inc. No Dishing!
PR PR Photoresist PR
TEOS Oxide
Epitaxial Silicon
P+ Silicon <100>
P+ Silicon <100>
3. Not all trenches are uniform in size. Some trenches are very large and this poses special
TEOS problems for CMP planarity. Illustration #1 displays a very large trench that creates a
large variation in the TEOS oxide density across the surface of the wafer. This is highly
undesirable for CMP.
TEOS Oxide
Epitaxial Silicon In order to avoid this problem the trenches and their adjacent area are patterned with
P+ Silicon <100> photoresist (figure #2) and the oxide over the active areas is etched away (figure #3).
This produces an oxide density that is somewhat more uniform and which should polish
© 2009 Threshold Systems more evenly. However the very large open trenches can still end up being somewhat
dished after the CMP operation is completed (figure #4).
Dishing!
4. An alternate approach is illustrated in Method B wherein “tiles” are designed into the
TEOS Oxide middle of the large open structures (figure #5). These tiles are dummy structures that
Epitaxial Silicon perform no function other than to present a uniform oxide surface to the CMP polish
platen. This eliminates the dishing effect (figure #6) and also eliminates the photo and
P+ Silicon <100>
etch operations employed in the method A.
© 2009 Threshold Systems
Polished Nitride
After the oxide has been polished back the thin layer of nitride on the
surface that acted as a CMP stop must be removed.
Oxide
Ox
Ox
.
nc
Ox
sI
m
P+ Sil
ste
icon < Ox
Sy
100>
ld
Epitaxial
hos
re
Silicon
Th
16
20
©
© 2018 Threshold Systems Inc.
Pad Oxide Trench Oxide
The residual silicon nitride is removed by immersing the wafers in hot (140 ºC)
Phosphoric acid for 20 minutes. The Phosphoric acid has a high selectivity to
nitride over oxide and will not attack the oxide.
Oxide
Ox
Ox
.
nc
Ox
sI
m
P+ Sil
ste
icon < Ox
100>
Sy
ld
Epitaxial
ho
s
re
Silicon
Th
16
20
©
The Pad Oxide strip unintentionally removes oxide preferentially from the
trench corners because these areas are under stress and etch faster. This
results in the creation of “notches” along the sides of the trench that can
cause problems later in the process.
STI Notch
STI
STI
.
nc
STI
sI
P+ Sil
m
icon <
ste
100> STI
Sy
Epitax
ld
ial Sili
ho
con
s
re
Th
16
20
© © 2018 Threshold Systems Inc.
Sacrificial Gate Oxide
The wafers are cleaned in P/SC1/SC2 and a 50 Å sacrificial gate oxide (Sac
Ox) is grown. The sacrificial gate oxide also acts as a screen oxide.
STI
STI
STI
.
nc
STI
sI
P+ Sil
m
icon <
ste
100> STI
Sy
Epitax
ial Sili
l d
con
hos
re
Th
16
20
©
Notch Formation
Notch
TEOS TEOS
STI TEOS TEOS TEOS
STI TEOS
Notch
Epitaxial Silicon
> P+ Silicon <100> STI STI
STI
© 2016 Threshold Systems Inc.
© 2016 Threshold Systems Inc.
>
[Link] © 2018 by Threshold Systems Inc. All Rights Reserved. 21
[Link] © 2018 by Threshold Systems Inc. All Rights Reserved. 22
Well Module:
STI
P-Well
ell
N-Well
W
P-
P-Well
N-Well
performance. The Wells also augment the electrical isolation provided by the
Shallow Trench Isolation by establishing diode barriers between adjacent Epitaxial Silicon
P+ Silicon <100>
PMOS and NMOS devices.
In the well fabrication module the P-Wells and the N-Wells required for the A silicon wafer with STI
NMOS and PMOS transistors are realized using a series of masking and ion
implantation steps.
Epitaxial Silicon
P+ Silicon <100>
P-Well
The P-Well
The wafers are primed in HMDS and then coated with TARC (Top Anti-
Boron Implant Reflective Coating) and photoresist. The photoresist is soft baked to give
it a measure of structural integrity. After exposure and development the
photoresist has a pattern that covers the region where the N-Wells will
be located and which exposes the regions of the wafers where the P-
Wells are located.
TARC
Next, the wafers are implanted with a chain implant of Boron
that establishes the retrograde well P-Well (B, 5E13@200KeV, 7º),
a mid-well Field Channel Stop (B, 5E12@50KeV, 7º), and finally a
Threshold Voltage adjust (B, 5E11@5KeV 7º).
.
nc
P-Wel
transistor channel is located.
sI
l STI
m
P+ Sil
ste
icon <
100>
Sy
STI
ld
hos
re
Th
16
Retrograde Well
20
©
Dopant Concentration
Depth
Wafer Surface
Ion Implantation is the process of accelerating atoms to a very high energy and then
Dopant atoms
implanting a very exact number of them into the surface of a silicon wafer at a precise
depth. These atoms are known as “dopant” atoms.
The kinds of atoms that are implanted are generally Boron, Phosphorous or Arsenic.
These atoms alter the electrical characteristics the silicon and create localized regions that
have specific electrical characteristics (like Sources, Drains or Wells) required to form
transistors and other electrical structures.
P-Well P-Well
The big advantage of doping with an implanter is that is provides enormous control over
the depth, quantity and purity of the dopant atoms that are implanted. Ion Implantation is the introduction of precise
Implants are specified by four parameters: Dopant species, dose, energy and implant quantities of atoms into the silicon surface at specific
depths. These dopant atoms, once activated, alter
angle. the electrical characteristics of the silicon.
P-Well P-Well
The dose of the implant, expressed The implant energy expressed in Epitaxial Silicon
in “ions/cm2”. In this case 3 x 1015 Kilo-electron Volts (KeV). This P+ Silicon <100>
ions/cm2. determines how deep the ions are
implanted. The location of where the dopant atoms are placed
into the silicon can be controlled by using
photoresist. Those areas covered by the photoresist
are shielded from the implant. Exposed regions
receive the full dose of the implant.
Ion Implanters
Ion Implanters are a special kind of particle accelerator and operate in a similar
fashion to standard particle accelerators.
The dopant species to be implanted is first converted into a gas and then ionized so
that it has a charge and can be accelerated to a precise energy in an electric field.
The accelerated dopant ions are then filtered in a mass spectrum analyzer that
removes any contaminant gas atoms.
The ion beam is then steered by powerful magnets and directed toward the wafer
where is rastered across the wafer’s surface to implant the dopant species.
The exact dose of the implant is determined by “counting” the dopant atoms by
precisely measuring the accumulated charge in the ion beam as it passes a specific
point in the implanter.
Source: Axcelis
After dopant atoms have been implanted it is necessary to “activate” them by
applying heat in order to move them onto substitutional sites in the silicon lattice
where they become electrically active. Ion Source
Magnetic
Extraction Electrode Quadrupole
Assembly Triple Index
Ion Implanters are by far the largest machines in the fab and they cost several Analyzer Magnets
Lens
Electron shower
Flag Faraday
Disk
(X-Y mechanical scan)
Process Chamber
(two axis tilt capability)
Sac Oxide
The TEOS oxide in the STI trenches is now completely exposed as is the
STI Sacrificial Oxide (“Sac Ox”).
P-Wel
l STI
Epitax ell
ial ST -W
I P
Silicon
P-Wel STI
P+ Sil l
.
nc
icon <
sI
100>
m
STI
ste
Sy
ld
hos
re
Th
16
20
© © 2018 Threshold Systems Inc.
Phosphorous + Arsenic
TARC The wafers are primed in HMDS and then coated with TARC, and the photoresist
is soft baked to give it a measure of structural integrity. After exposure,
development, and hard-bake, the photoresist has a pattern that covers the
I region where the P-Wells are located and which exposes the N-Well regions.
ST
The wafers then receive a chained implant of Phosphorus and Arsenic that
STI Photor
implants the retrograde N-well (P, 5E13 @400KeV, 0º), a mid-well Field Channel
P-Wel e sist Stop (P, 5E12@100KeV, 0º) the Threshold Voltage adjust (As, 5E11@5KeV, 0º).
l STI ell
N-We W
ll STI P-
P-Wel
.
l
nc
STI
sI
P+ Sil
m
100> ll STI
Sy
ld
ho
s
re
Th
16
20
©
TARC
The photoresist from the previous operation is stripped in an ionized
oxygen plasma and the wafers are cleaned in Piranha.
I The wafers are primed with HMDS, coated in photoresist and TARC, and
ST
Photor then patterned. After patterning that portion of the active area that is
esist
designated to be the High Vt N-Well processing section is exposed, while all
STI other regions are covered with photoresist.
P-Wel
l STI Photor
esist ell
N-We -W The High Voltage N-Wells receive an implant of As,3E11@5KeV, 0º to give
ll STI P
P-Wel
l
them a higher Threshold Voltage (Vt) than the Low Voltage N-Wells.
STI
.
nc
P+ Sil
icon < N-We
sI
100> ll
m
STI
ste
Sy
ld
ho
High Vt N-Well
s
re
Th
16
20
© 2018 Threshold Systems Inc.
©
Low Vt N-Well
Sac Oxide
STI
P-Wel
l STI ell
N-We -W
ll STI P
P-Wel
.
l
nc
STI
sI
P+ Sil
m
100> ll STI
Sy
ld
ho
s
re
Th
16
20
©
.
P+ Sil l
nc
icon < STI
sI
100>
m
N-We
ste
ll STI
Sy
ld
High Vt P-Well
sho
re
Th
16
Low Vt P-Well
20
©
© 2018 Threshold Systems Inc.
The photoresist from the previous operation is stripped and the wafers
Sac Oxide are cleaned in Piranha.
The anneal makes the dopant atom electrically active and anneals out
STI the damage done to the silicon during the implants.
P-Wel
l STI ell
N-We -W
ll STI P
P-Wel
.
l
nc
STI
sI
P+ Sil
m
100> ll STI
Sy
ld
hos
re
Th
16
20
©
Front of Wafer
Back of Wafer
© Applied Materials
© Ultra-Tech
RTP abruptly raises the temperature of the front
side of the wafer for a fraction of a second
without adding a lot of heat.
Traditional RTP chambers consist of a series of very
Unlike traditional RTP tools, Laser annealing
bright lamps that raise the temperature of the wafer’s
tools activate the implanted dopant while
surface for a brief instant.
providing virtually zero heat. This is a more
sophisticated approach to annealing.
P-Well N-Well
pitaxial Silicon
con <100>
© 2016 Threshold Systems Inc.
In the Gate Module the transistor structures are fabricated. They TEOS TEOS TEOS TEOS TEOS
consist of the Source and Drains, the Halo and Extension implants, P-Well N-Well P-Well N-Well
the Nitride Spacers and the all-important Gate Electrode and gate Epitaxial Silicon
P+ Silicon <100>
dielectric.
A silicon wafer with P and N wells
The Gate Electrode initially consists of a strip of doped amorphous
silicon of precise dimensions. In the replacement gate methodology,
the amorphous silicon is merely a place-holder that is removed and
replaced with a complex series of metals.
Beneath the Gate Electrode is the transistor dielectric.
Gate Electrode
A silicon wafer with Gate electrodes formed over the P and N wells
30 Å SiO2 ESL
Hi-K
TEOS TEOS TEOS TEOS TEOS
Epitaxial Silicon
P+ Silicon
© 2016 Threshold Systems<100>
Inc.
STI
P-Wel
l STI ell
N-We -W
ll STI P
P-Wel
l STI A thin 30Å Etch Stop Layer (ESL) of SiON thermal layer oxide is grown across
P+ Sil
.
nc
icon < N-We the surface of the wafer. This oxide layer will act as an etch stop layer to
sI
100> ll STI
m
ste
protect the transistor channel when the amorphous silicon gate mandrel
Sy
(not yet deposited) is etched away.
ld
hos
re
Th
Note that there is no Hi-k dielectric deposited at this point in the process flow.
16
20
©
800 Å of Amorphous Silicon
© 2018 Threshold Systems Inc.
TiN
Epitaxial Silicon
P+ Silicon <100>
P+ Sil
sI
100> ll
ste
After it has been patterned and etched the AmSi forms the sacrificial gate
20
©
mandrel.
Note that there is no metal gate beneath the amorphous silicon gate electrode.
The amorphous silicon is implant doped with Phosphorus (5.0E15@ 5 KeV, 0º)
to make it conductive. The gate electrode is sacrificial and will be removed
later in the process. Doping the gate electrode aids in its removal but is not
necessary.
STI
P-Wel
l STI ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
icon < N-We
m
100> ll
ste
STI
Sy
ld
ho
s
re
Th
16 © 2018 Threshold Systems Inc.
20
©
BARC
A 1,500 Å thick amorphous Carbon hard mask and a 300 Å thick BARC
layer are deposited.
Carbo
n Hard M
ask
STI Amorp
P-Wel hous S
i
l STI ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
100> ll
ste
STI
Sy
d
ol
sh
re
Th
16
20
©
Next, the wafers are primed in HMDS, coated in BARC (Bottom Anti-
Reflective Coating), photoresist and patterned. This is the most critical
Photores lithography operation in the entire process flow. It generates a series of
ist
parallel lines that will be further defined by a cut mask in a following
lithography operation.
BARC
Carbo
n Hard M
ask
STI Amorp
P-Wel hous S
i
l STI ell
N-We -W
ll STI P
P-Wel
.
nc
l STI
sI
P+ Sil
m
icon < N-We
ste
100> ll STI
Sy
ld
sho
re
Th
20
©
16
© 2018 Threshold Systems Inc.
Exposed Unexposed
Photoresist Photoresist
Next, the wafers are exposed using a grating mask. This mask prints
straight lines in one direction only.
Photoresis
t
Carbo
n Hard M
ask
STI Amorp
P-Wel hous S
i
l STI ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
100> ll
ste
STI
Sy
d
ol
sh
re
Th
16
20
©
Carbo
n Hard M
ask
STI Amorp
P-Wel hous S
i
l STI ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
icon < N-We
m
100> ll
ste
STI
Sy
ld
ho
© 2018 Threshold Systems Inc.
s
re
Th
16
20
©
Cut Mask
Patterned
Photoresist
The wafers are then loaded into an etcher and exposed to a controlled
oxygen plasma that isotropically etches the patterned photoresist. This
causes the photoresist to shrink in size to a targeted dimension.
.
nc
P+ Sil
sI
icon < N-We
m
100> ll
ste
STI
Sy
ld
ho
s
re
© 2018 Threshold Systems Inc.
Th
16
20
©
Trimmed
Photoresist
The trimmed photoresist lines are now narrower and can be used to
define smaller narrower gate electrodes.
P+ Sil
sI
100> ll
ste
STI
Sy
ld
ho
s
re
Th
16
20
©
Source: ASML
Double and triple layer patterning would not be necessary if EUV were available …
dimension the resist lines, it does not increase their density. Increased density
and smaller photoresist lines can only be achieved through advanced Polysilicon
photolithographic reduction (refer to #2 above right).
Step #1 Feedback
Too narrow; strip resist and start again
© Fujitsu
© Fujitsu
Too wide; send
TEOS TEOS TEOS TEOS TEOS TEOS TEOS TEOS TEOS TEOS TEOS TEOS TEOS TEOS TEOS
P-Well N-Well P-Well N-Well P-Well N-Well P-Well N-Well P-Well N-Well P-Well N-Well
Photoresist
Step #2 Width
Desired Width
Feed-forward
© 2018 Threshold Systems Inc.
© 2016 Threshold Systems Inc. © 2016 Threshold Systems Inc.
35 Trim Rate =
1.228 nm/sec The wafers on then sent on to have the photoresist trimmed in an etcher (step #4 ) and then the wafers are sent on so that the
30
resist after-trim dimension can be measured in a SEM (step #5).
25
20 This information is feedback to the etcher (step #6) and the etch time is adjusted to fine-tune the resist trim process. If the resist
15 is not small enough, the wafers are returned to the etcher to have their resist additionally trimmed.
10 If the resist is the correct dimension the wafers are sent on to have the polysilicon etched (Step #7). If too much resist has been
5 trimmed, then the entire photoresist pattern will be stripped and the entire photolithographic patterning and resist trim
process must be repeated.
0 5 10 15 20 25 30 35 40 45
Time (sec) This process is completely automated.
Hard Mask
The amorphous carbon hard mask is etched in a highly anisotropic etch process
that stops on the amorphous silicon.
STI Amorp
P-Wel hous S
i
l STI ell
N-We -W
ll STI P
P-Wel
.
nc
l STI
sI
P+ Sil
m
icon < N-We
ste
100> ll STI
Sy
ld
ho
s
re
Th
16
© 2018 Threshold Systems Inc.
20
©
Hard Mask
STI Amorp
P-Wel hous S
i
l STI ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
100> ll
ste
STI
Sy
ld
hos
re
Th
16
20
©
A highly anisotropic etch recipe is then used to transfer the pattern from
the hard masks into the amorphous silicon.
STI
P-Wel
l STI ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
icon < N-We
m
100> ll
ste
STI
Sy
ld
ho
s
re
Th
© 2018 Threshold Systems Inc.
16
20
©
Amorphous
Silicon
STI
P-Wel
l STI ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
100> ll
ste
STI
Sy
ld
ho
s
re
Th
16
20
©
Poly Oxide
The wafers are cleaned in Piranha and 15Å of thermal poly oxide is grown.
This is followed by the deposition of 15Å of CVD oxide. These two oxide layers
form what is referred to as the “Offset Spacer” (sometimes also referred to as
“Spacer-Zero” or the “Spacer-1”). In previous generation of devices this oxide
was much thicker and referred to as “Poly Oxide”.
STI
P-Wel
l STI ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
icon < N-We
m
100> ll
ste
STI
Sy
ld
hos
re
Th
© 2018 Threshold Systems Inc.
16
20
©
Poly Stringers
The original Pad Oxide strip (performed several operations ago after trench
fill and nitride removal) removes oxide preferentially from the trench corners
resulting in the creation of “notches” all along the sides of the trenches that
facilitate the formation of polysilicon stringers.
Notch The poly stringers form at the end of the polysilicon etch operation. The final
phase of this etch process is highly selective over-etch that is designed to
remove any small amounts of polysilicon remaining in the Source/Drain
regions and in the notches. If this phase of the etch is not carefully designed,
the thick polysilicon stringers will remain in the trench notches and will form
STI unintended conductive pathways that will short-out adjacent transistors.
STI
For this reason it is very important to minimize the size of the notches to
facilitate easy stringer removal.
© 2016 Threshold Systems Inc.
Trench Oxide A
c
Gate
Active Region t
i
v
Gate Landing
Gate e
Gate
An artificially colored section of a memory array An artificially colored section of a Logic chip displaying
displaying 3D angled views of real-world gate an angled view of the trench oxide, active areas and
electrodes running over active regions and trench gate electrodes. Note how the trench oxide is not
oxide. always at minimum dimension and that it assumes a
wide range of shapes and sizes.
Boron
TARC
The wafers are coated with photoresist and a Top Anti-Reflective Coating
(TARC) and then patterned. After patterning the photoresist covers all of
the active areas except the P-Wells.
.
nc
P+ Sil
sI
icon < N-We
m
100> ll
ste
STI
Sy
Photoresist
ld
ho
s
re
Th
16
20
©
TEOS TEOS TEOS TE
Epitaxial Silicon
TARC P+ Silicon <100>
A view of the Boron Halo implant when the wafers have been rotated 180
degrees. This would be the third of the four implants that comprise the
quad Halo implant. The others being implanted when the wafer is at
angular positions of zero, 90 and 270 degrees.
Photor
e sist
STI
P-Wel Photor
l STI e sist ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
100> ll
ste
STI
Sy
ld
ho
s
re
Th
16
20
©
Using the same photoresist pattern that was used for the Halo implant, an
Arsenic Extension implant (1E15@1 KeV, 0º ) is implanted into the P-Wels. This
implant establishes the Extension implants of the NMOS transistors.
TARC
The extension implants are at approximately the same depth as the Halo
implants (150-200 Å for a 20nm device).
Photor
e sist
STI
P-Wel Photor
l STI e sist ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
© 2018 Threshold Systems Inc.
sI
icon < N-We
m
100> ll
ste
STI
Sy
ld
hos
re
Th
16
20
©
ArsenicArsenic
Photoresist The adjacent 2D image clearly displays the Extension implant. This implant is
self-aligned to the edge of the gate electrode and this high dose implant
overwhelms the Halo implant in the area where the two implants overlap.
However, the tip of the Halo implant is tucked under the gate electrode and the
TEOS TEOS TEOS TEOS electrode shields the tip of the Halo from the Extension implant in this location.
TEOS
P-Well N-Well P-Well N-Well The Extension implant determines the final L-effective (Leff) of the transistor
Epitaxial Silicon and has a huge effect on IdSAT.
P+ Silicon <100>
© 2016 Threshold Systems Inc.
The photoresist from the previous operation is ashed away and the wafers
are cleaned in Piranha.
The wafers are then undergo a Rapid Thermal Process (~950 degree C
Spike for ~ 1 second followed by a ~1,350 degrees FLASH anneal for 1-3
milliseconds) that activates the NMOS Extension and Halo implants. In
some process flows the Flash anneal is replaced by a sub-melt Laser
STI anneal at the same temperature of ~1,350 degrees C. The location of the
ˆˆ P-Wel Extension implant fixes the final Leff of the NMOS transistors.
l STI ell
N-We -W
ll STI P
P-Wel The anneal makes the dopant atom electrically active and anneals out the
l STI
.
nc
P+ Sil damage done to the silicon during the implants.
sI
icon < N-We
m
100> ll STI
ste
Sy
ld
ho
s
re
Th
16
20
© 2018 Threshold Systems Inc.
©
Arsenic
TARC Next, the wafers are patterned with photoresist and coated with TARC. After
photoresist development only the N-Wells are exposed. A tilted 30 º Arsenic
Halo implant (2.2E13@20KeV) is applied to this well. It is a quad implant.
STI Photor
e sist
P-Wel
l STI ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
100> ll
ste
STI
Sy
ld
ho
s
re
Th
16
20
©
A view of the Arsenic Halo implant when the wafers have been rotated 180
degrees. This would be the third of the four implants that comprise the
quad Halo implant. The others being implanted when the wafer is at
angular positions of zero, 90 and 270 degrees.
Photor The Halo implant is ~ 150 - 200 Å deep for a 20nm node device.
STI e sist
P-Wel
l STI ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
icon < N-We
m
100> ll
ste
STI
Sy
ld
ho
s
re
© 2018 Threshold Systems Inc.
Th
16
20
©
Ge + B + C
TARC
A Pre-Amorphizing Implant (PAI) of Germanium (5E14@4KeV, 0º) is
applied to the N-Wells, and then a low energy Extension implant of (B,
8E14@300eV, 0º) is applied. This is followed by a Carbon implant of
5E14@3KeV which facilitates a superior Extension implant profile.
Photor
e
The PAI limits the channeling of the Boron atoms, while the Boron
sist
implant establishes the Extension implants of the PMOS transistors.
STI Photor
e sist
P-Wel
l STI ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
100> ll
ste
STI
Sy
ld
ho
s
re
Th
16
20
©
The photoresist from the previous operation is ashed away and the wafers
are cleaned in Piranha.
The wafers are then undergo a Rapid Thermal Process (~950 degree C
Spike for ~ 1 second followed by a ~1,350 degrees FLASH anneal for 1-3
milliseconds) that activates the PMOS Extension and Halo implants. In
some process flows the Flash anneal is replaced by a sub-melt Laser
anneal at the same temperature of ~1,350 degrees C.
STI
P-Wel
l STI ell The location of the Extension implant fixes the final Leff of the PMOS
N-We -W
ll STI P transistors.
P-Wel
l STI
.
nc
P+ Sil
sI
icon < N-We The anneal makes the dopant atom electrically active and anneals out the
m
100> ll STI
ste
damage done to the silicon during the implants and has a huge effect on
Sy
ld
ho
IdSAT.
s
re
Th
16
20
©
© 2018 Threshold Systems Inc.
Front of Wafer
Back of Wafer
N-Well
Halo Implant
(N-doped 1013 ions cm-2)
N-Well
(N-doped 1012 ions cm-2)
© 2016 Threshold Systems Inc.
Silicon Nitride
A 150 Å thick blanket layer of silicon nitride is deposited over the surface of
the wafers. Because of the conformal nature of silicon nitride, it is thickest
along the side-walls of the gate electrodes.
STI
P-Wel
l STI ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
icon < N-We
m
100> ll
ste
STI
Sy
ld
ho
s
© 2018 Threshold Systems Inc.
re
Th
16
20
©
Silicon Nitride
Spacers
A highly anisotropic etch removes the silicon nitride from all of the horizontal
surfaces, and but leaves nitride side-wall spacers along the sides of the gate
electrodes where the nitride is thickest. The nitride side-wall spacers act as
mini-implant masks during the contact implants that come later in the
process.
STI
P-Wel
l STI ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
100> ll
ste
STI
Sy
ld
ho
s
re
Th
16
20
©
TEOS TEOS
TEOS TEOS TEOS TEOS TEOS TEOS TEOS TEOS TEOS Because the etch is highly anisotropic (etches in one direction only, i.e.
downward) the silicon nitride is cleared everywhere except the on the
P-Well N-Well N-Well P-Well P-Well N-Well N-Well P-Well N-Well sidewalls of the gate electrode because this is where the nitride is the
Epitaxial Silicon
thickest.
Epitaxial Silicon
P+ Silicon <100> P+ Silicon <100>
© 2016 Threshold Systems Inc. © 2016 Threshold Systems Inc. © 2016 Threshold Systems Inc.
STI
ace rs
e Sp
STI Nitr
id
A highly anisotropic etch removes the silicon nitride from all of the horizontal
surfaces, and but leaves nitride side-wall spacers along the sides of the gate
ers
eS pac electrodes. The nitride side-wall spacers act as mini-implant masks during the
STI Nitrid
contact implants that come later in the process.
ers
eS pac
STI Nitrid
rs
Sp ace
ride
STI Nit
ell
P-W
STI
.
s Inc
ys tem
oldS
sh
hre
16T
[Link] 20 © 2018 by Threshold Systems Inc. All Rights Reserved. 55
©
PMOS Hard Mask Deposition & Patterning:
STI
P-Wel
l STI ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
icon < N-We
m
100> ll
ste
STI
Sy
ld
ho
s
re
© 2018 Threshold Systems Inc.
Th
16
20
©
Photores
is t
STI
P-Wel
l STI ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
100> ll
ste
STI
Sy
ld
ho
s
re
Th
16
20
©
Photores
is t
STI
P-Wel
l STI ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
icon < N-We
m
100> ll
ste
STI
Sy
© 2018 Threshold Systems Inc.
ld
ho
s
re
Th
16
20
©
STI
P-Wel
l STI ell
N-We -W
ll STI P
P-Wel
l STI
.
nc
P+ Sil
sI
100> ll
ste
STI
Sy
ld
ho
s
re
Th
16
20
©
The exposed Source/Drains of the PMOS transistor are wet etched away
using a NH4OH solution that etches along crystallographic planes. The P-
Wells and the gate electrodes are protected by the SiCN. The etch will not
attack the exposed oxide in the STI trenches.
STI
P-Wel
l STI ell
-W
N-We STI P Crystal Plane
ll P-Wel
l STI
.
nc
P+ Sil
sI
icon <
m
100>
ste
N-We STI
Sy
ll
ld
TEOS TEOS TEOS TEOS TEOS
sho
re
Th
16
20
Etched out N-Well N-Well
P-Well P-Well
Source/Drains ©
Epitaxial Silicon
SiCN Hard Mask
P+ Silicon <100>
© 2016 Threshold Systems Inc.
The wafers then undergo a selective epitaxial deposition of SiGe. The SiGe
will only nucleate on an exposed silicon surface which means that it will
only form in the PMOS Source/Drain regions (every other part of the
STI wafer consists of nitride or oxide, or is covered in SiCN). The SiGe is ~ 40%
P-Wel Ge and it will exert a high level of compressive strain on the PMOS
l STI ell
P-W channel, greatly increasing the mobility of the holes in the channel and
N-We STI
ll P-Wel therefore increase Idsat.
l STI
.
nc
P+ Sil
sI
icon <
m
100> As the Epitaxial SiGe is deposited it is heavily doped with Boron to make it
ste
N-We STI
ll
Sy
P-Type silicon.
ld
hos
re
Th
SiGe Fill
20
©
.
SiGe SiGe
nc
TEOS
P+ S SiGTEOS TEOS TEOS SiGe SiGe TEOS
sI
ilicon e
m
<100
ste
> N-We STI
Sy
ll
ld
ho
P-Well N-Well P-Well N-Well
s
re
Th
16
Epitaxial Silicon
20
©
P+ Silicon <100>
© 2016 Threshold Systems Inc.
STI
P-Wel SiGe
l STI SiGe ell
-W
N-We STI P
ll P-Wel
l STI SiGe
.
nc
P+ Sil SiGe
sI
icon <
m
100>
ste
N-We STI
Sy
ll
ld
ho
s
re
Th
16
20
© 2018 Threshold Systems Inc.
©
Photoresis
STI t
P-Wel SiGe
l STI SiGe ell
-W
N-We STI P
ll P-Wel
l STI SiGe
.
nc
P+ Sil SiGe
sI
icon <
m
100>
ste
N-We STI
Sy
ll
ld
ho
s
re
Th
16
20
©
STI
P-Wel SiGe
Crystal Plane
l STI SiGe ell
-W
N-We STI P
ll P-Wel
l STI SiGe
.
nc
P+ Sil SiGe
sI
icon < TEOS TEOS SiGe SiGe TEOS TEOS SiGe SiGe TEOS
m
100>
ste
N-We STI
Sy
ll
ld
hos
re
P-Well N-Well P-Well N-Well
Th
16
20
Epitaxial Silicon
©
P+ Silicon <100>
© 2016 Threshold Systems Inc.
SiCN Hard Mask
The oxide over the Source/Drains is etched away using a HF wet etch.
The exposed Source/Drains of the PMOS transistor are then wet etched
away using a NH4OH solution that etches along crystallographic planes.
STI
P-Wel
l
STI SiGe
N-We
SiGe
ll
STI P-
W
ell
© 2018 Threshold Systems Inc.
STI SiGe
.
P-Wel
nc
P+ Sil l SiGe
sI
icon <
m
100> STI
ste
N-We
Sy
ll
ld
ho
s
re
Th
Etched out
16
20
Source/Drains
©
The SiC exerts a very strong tensile strain on the NMOS channel and
enhances the mobility of electrons. This increases the Idsat of the NMOS
STI devices. As the SiC is deposited it is doped with Phosphorus to make it N-Type.
SiGe
P-Wel STI SiGe ell
l -W
N-We STI P
ll
STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon < Halo & Extension
m
100>
ste
N-We STI Implants
Sy
ll
ld
hos
re
Th
SiC Fill TEOS SiC SiC TEOS SiGe SiGe TEOS SiC SiC TEOS SiGe SiGe TEOS
16
20
©
P-Well N-Well P-Well N-Well
Epitaxial Silicon
P+ Silicon <100>
© 2016 Threshold Systems Inc.
STI SiC
SiC
P-Wel
l
STI SiGe
N-We
SiGe
ll
STI SiC
SiC P-W
ell © 2018 Threshold Systems Inc.
STI SiGe
.
nc
icon <
m
100>
ste
N-We STI
Sy
ll
ld
hos
re
Th
16
20
©
IdSat (mA/µm)
1.0 Other
Classic Scaling 65nm
90nm
0.5
130nm
0.0
1,00 Gate Pitch (nm) 100
Source: Schubert Chu, Applied Materials 2013
After [Link] - Intel
Source
S Drain
T S
I T
I
Silicon Subst
rate
Source: Schubert Chu, Applied Materials 2013 Source: Schubert Chu, Applied Materials 2013
The Source/Drains are selectively etched away and replaced with Silicon/Germanium. Germanium has a lattice constant
that is 4% larger than that of Silicon, and this larger lattice constant will ensure that the transistor channel is compressed,
enhancing the transistor drive current.
As previously mentioned, the amount of Ge in the PMOS SiGe Source/Drains is increased at each
new node to maintain the compressive stress on the transistor channel (currently 55% Ge at the
20nm node).
Another technique is to move the pointed edge of the SiGe crystal closer to the channel to
induce more stress. This was done at 32nm (refer to the accompanying TEMS).
The replacement gate also offers another method to increase PMOS strain. When the polysilicon
gate electrode is removed the spacers and associated structures move inward, compressing the
transistor channel and inducing compressive stress.
32nm PMOS
180 130 90 65 45 32 22
Node (nm)
Source: Schubert Chu, Applied Materials 2013 Source: Schubert Chu, Applied Materials 2013
A more specific name for Stress Memorization Technique (STM) on the NMOS device is a technique known as
“Edge Dislocation Stress”. This is a STM technique that generates a mask-edge dislocation inducing dramatic
tensile stress in the NMOS channel. It was first deployed at the 45/32nm nodes prior to the advent of P or P&C
replacement Source/Drain epitaxy for NMOS transistors.
The process involves a deep amorphization implant followed by the deposition of a stress over-liner of nitride.
Following a Rapid Thermal Anneal to induce SMT, the stress liner is removed and the mask-edge dislocations
are formed.
The initial pre-amorphization implant (PAI) for SMT creates multiple mask-edge dislocations under the
Source/Drain region which enhances short-channel mobility. Overall, a 10% increase in Idsat is achieved.
It is important to realize that different strain techniques are typically used together because their
contributions to increased performance is additive.
The wafers are then undergo a Rapid Thermal Process (~950 degree C
Spike for ~ 1 second followed by a ~1,350 degrees FLASH anneal for
~1-3 milliseconds) that activates the PMOS and NMOS dopant atoms.
STI SiC
SiC
SiGe
P-Wel STI SiGe ell
l SiC -W
N-We STI SiC P
ll
STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon <
m
100>
ste
N-We STI
Sy
ll
ld
ho
s
re
Th
16
20
©
(-V)
The wafer is placed in an evacuated chamber Target (Ni)
on top of the anode and a small amount of
Argon gas is introduced into the chamber. A Cathode
plasma is struck and the Argon is ionized. The Plasma
positively charged Argon ions then bombard Ar+
the Cathode which consists of the target Ar+
material, and in so doing, sputter off metal
that is deposited onto the wafer. Sputtered Atom
This thin and very pure coating of metal is Wafer
rapidly deposited onto the wafer’s surface.
Anode
© Applied Materials
(+V)
An open degas chamber.
After the anneal the wafers still have a layer of poly oxide on the top of
the gate electrodes. There may also be a native oxide over the SiGe and
the SiC Source/Drain regions.
STI SiC Both of these oxides must be removed to expose a bare silicon surface
SiC
P-Wel STI SiGe
SiGe ell
for the salicidation process.
l SiC -W
N-We STI SiC P
ll
STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon <
m
100>
ste
N-We STI
Sy
ll
ld
hos
re
Th
16
20
© 2018 Threshold Systems Inc.
©
The wafers are then dipped in HF to remove all of the silicon dioxide
from tops of the gate electrodes and from the surfaces of the Source/
Drain regions. This operation prepares the wafers for the upcoming
salicidation step.
STI SiC
SiC
SiGe
P-Wel STI SiGe ell
l SiC -W
N-We STI SiC P
ll
SiGe
.
nc
P-Wel STI
SiGe
sI
P+ Sil l
em
icon <
100>
st
N-We STI
Sy
ll
ld
ho
s
re
Th
16
20
©
STI SiC
SiC
SiGe
P-Wel STI SiGe ell
l SiC -W
N-We STI SiC P
ll
STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon <
m
100>
ste
N-We STI
Sy
ll
© 2018 Threshold Systems Inc.
ld
ho
s
re
Th
16
20
©
Aluminum
The wafers are patterned with photoresist that exposes only the P-Wells.
They are then implanted with Aluminum (~2E14@ 5KeV, 0º). The
aluminum is positioned so that it will be located at the interface between
the SiGe and the yet-to-be deposited Ni/Pt. This implant results in lower
contact resistance for the PMOS devices because the Aluminum
segregates to the top of the SiGe and being a P-Type dopant it lowers the
Schottky barrier height for holes from ~.4eV to ~.12eV and increases drive
current by up to 19%.
STI SiC
SiC
SiGe
P-Wel STI SiGe ell
l SiC -W
N-We STI SiC P
ll
STI SiGe
.
nc
icon <
m
100>
ste
N-We STI
Sy
ll
d
ol
sh
re
Th
16
20
©
The wafers are cleaned with P/SC1 and a blanket coating of ~125 Å
Nickel/Platinum and ~150 Å of TiN is deposited using Physical Vapor
Deposition. The TiN cap on top of the Nickel helps to prevent the Nickel
from diffusing away during the high-temperature salicidation process
and promotes more uniform NiSi formation. The 5-10% Platinum
Platinum helps to lower the barrier at the Ni-Si interface and establish a
STI SiC lower contact resistance. It also postpones the transformation of low
SiC
SiGe resistance NiSi into the higher resistance NiSi2.
P-Wel STI SiGe ell
l SiC -W
N-We STI SiC P
ll
STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon <
m
100>
ste
N-We STI
Sy
ll
TiN
Nickel/Platinum The adjacent 2D image illustrates the Nickel/Platinum/TiN stack after
initial deposition and prior to the anneal.
TEOS SiC SiC TEOS SiGe SiGe TEOS SiC SiC TEOS SiGe SiGe TEOS
Epitaxial Silicon
P+ Silicon <100>
© 2016 Threshold Systems Inc.
The wafers undergo a RTA at ~250 º C for 30 seconds that converts the
nickel-Platinum into the high resistance C49 phase. A second RTA at
~390 º C for 30 seconds converts the on the wafer’s surface into the low
resistance (15 µΩ-cm) C54 NiSi phase.
NiSi is necessary because it consumes at least 30% less silicon than the
STI SiC same thickness of CoSi2. This is extremely important because the 20nm
SiC
SiGe node junctions are very shallow. In addition, NiSi has a very smooth
P-Wel STI SiGe ell
l SiC -W surface which minimizes junction leakage.
N-We STI SiC P
ll
STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon <
m
100>
ste
N-We STI
Sy
ll
ld
hos
re
© 2018 Threshold Systems Inc.
Th
16
20
©
Reacted
Nickel
Unreacted
Nickel
The adjacent 2D image illustrates the reacted and unreacted Nickel/
Platinum on the devices. In order to form a Nickel silicide, the Nickel/
TEOS SiC SiC TEOS SiGe SiGe TEOS SiC SiC TEOS SiGe SiGe TEOS
Platinum layer must come into contact with an exposed Silicon surface.
The only place where exposed Silicon is located is on top of the gate
P-Well N-Well P-Well N-Well
electrodes or on the Source/Drain regions. The Nickel/Platinum on top
of the STI and on the Nitride spacers will remain unreacted.
Epitaxial Silicon
P+ Silicon <100>
© 2016 Threshold Systems Inc.
Nickel Silicide
The unreacted nickel located on the spacer sidewalls and on top of the
STI is etched away with a solution of Piranha and SC1. This etch is highly
selective and leaves the reacted salicide on top of the gate electrodes
and over the Source/Drains untouched. This etch also removes the TiN
capping layer from all surfaces on the wafer.
SiC
SiC
SiGe
P-Wel STI SiGe ell
l SiC -W
N-We STI SiC P
ll
.
SiGe
nc
P-Wel STI
sI
P+ Sil l SiGe
m
icon <
ste
100>
N-We STI
Sy
ll
ld
ho
© 2018 Threshold Systems Inc.
s
re
Th
16
20
©
Reacted
Nickel
The adjacent 2D image illustrates the fact that the Piranha solvent
TEOS SiC SiC TEOS SiGe SiGe TEOS SiC SiC TEOS SiGe SiGe TEOS
readily attacked and removed the unreacted Nickel/Platinum from the
nitride spacer sidewalls and from the top of the STI.
P-Well N-Well P-Well N-Well
The Piranha does not attack the reacted NiSi over the Source/Drains or
Epitaxial Silicon on top of the Gate Electrodes.
P+ Silicon <100>
© 2016 Threshold Systems Inc.
NMOS PMOS
AmSi Gate Electrode NiSi
NiSi Nitride Spacer
AmSi Gate
Electrode
Poly Oxide
NiSi
Extension Halo Implant
SiON Implant
SiON
Epitaxial Silicon
© 2016 Threshold Systems Inc. P+ Silicon <100>
Silicon Nitride
Oxide Layer
The wafers are cleaned in P/SC1 and thin 75 Å layer of silicon
dioxide followed by ~75Å of nitride is deposited. The oxide acts as a
diffusion barrier to external dopants that could contaminate the
device layer (the “active” Layer).
The silicon nitride will act as an etch-stop layer for the contact etch
STI SiC which will connect the transistor Source/Drains and gate electrodes
SiC
SiGe
P-Wel STI SiGe ell to the yet-to-be-formed Tungsten trenches.
l SiC -W
N-We STI SiC P
ll
STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon <
m
100>
ste
N-We STI Nitride Etch Stop
Sy
ll
ld
Layer (ESL)
ho
s
re
Th
16
20
©
PSG
STI SiC
P-Wel
SiC
l
STI SiGe
SiGe
STI -W
ell
© 2018 Threshold Systems Inc.
SiC P
N-We SiC
ll
STI SiGe
.
nc
icon <
m
100>
ste
N-We STI
Sy
ll
ld
ho
s
re
Th
16
20
©
PSG
A top-down view of the PSG layer prior or polishing. Note the humps in
the PSG caused by the underlying topography induced by the height of
the gate electrodes.
ms
In c. © 2018 Threshold Systems Inc.
te
Sys
sh old
hre
16T
20
©
PSG
STI SiC
SiC
STI SiGe
P-Wel SiGe ell
l STI SiC -W
N-We SiC P
ll STI SiGe
.
nc
P-Wel SiGe
sI
P+ Sil l
m
100> N-We
Sy
ll
ld
hos
re
Th
16
20
©
.
s Inc
ys tem
ld S
sho
hre
6T
201
©
Photoresist
Next, the wafers are primed in HMDS, coated in photoresist and patterned.
c.
s In
em
yst
ld S
sho
©
20 1 6 Th
re
© 2018 Threshold Systems Inc.
Photoresist
The illustration to the left is a tilted front-end view of the trench contact
photoresist pattern. Note the gate electrode contact trenches at the back of
the image.
PSG
STI SiC
SiC
STI SiGe
P-Wel SiGe ell
l STI SiC -W
N-We SiC P
.
ll
nc
STI SiGe
sI
P-Wel
m
P+ Sil l SiGe
ste
icon <
STI
Sy
100>
N-We
ld
ll
hos
re
Th
16
20
©
.
nc
STI SiGe
sI
P+ Sil P-Wel
m
l SiGe
icon <
ste
100> STI
Sy
N-We
ld
ll
ho
s
© 2018 Threshold Systems Inc.
re
Th
16
20
©
icon <
m
100>
ste
N-We STI
Sy
ll
ld
ho
s
re
Th
16
20
©
The hardMask is etched away and the wafers are cleaned in Piranha.
G
PS
STI SiC
SiC
SiGe
P-Wel STI SiGe ell
l -W
N-We STI SiC P
ll SiC
STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon <
m
100>
ste
N-We STI
Sy
ll
ld
ho
s
re
Th
© 2018 Threshold Systems Inc.
16
20
©
Ti/TiN
The wafers are degassed and then exposed to a soft Argon sputter pre-
clean to remove any native oxides present on top of the NiSi at the
bottom of the contact holes.
P+ Sil P-Wel
l SiGe Nitride and set the resistance of the glue layer.
sI
icon <
m
100> STI
ste
Sy
ld
ho
s
re
Th
16
20
©
Next, a seed layer of Tungsten is deposited that lines the inside of the
contact holes and ensures conformal coverage of the bulk Tungsten
deposition.
G
PS
STI SiC
SiC
STI ell
SiGe
P-Wel SiGe
W
l STI SiC P-
N-We SiC
ll
STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon <
m
100> STI
ste
Sy
ld
© 2018 Threshold Systems Inc.
sho
re
Th
16
20
©
Tungsten
Trench
The wafers are again polished using CMP. In this operation the
Tungsten overburden is removed and the Tungsten trenches are
polished back tops of the PSG.
icon <
m
100> STI
ste
Sy
ld
hos
re
Th
16
20
©
Tungsten
Trench These figures illustrate what the Tungsten trenches would look like if
the PSG oxide were removed. Clearly visible are the Tungsten trenches,
clad in Ti/TiN, making electrical contact to the Source Drain regions
and to the Gate Electrodes.
.
nc
P+ Sil P-Wel SiGe
l
sI
icon <
m
100>
ste
N-We STI
Sy
ll
ld
ho
s
re
Th
16
© 2018 Threshold Systems Inc.
20
©
Gate Contact
STI
P-W
ell
P+
Sili
con
<10
ell 0> STI
P-W c.
Source/Drain s In
em
Contacts yst
ld S
Inc
. sho
s hre
tem 6T
ys 201
ld S ©
e sho
6 Thr
201
©
[Link] © 2018 by Threshold Systems Inc. All Rights Reserved. 84
Contact Module:
Contact Module:
Polished-Back
TEOS SiC SiC TEOS SiGe SiGe TEOS SiC SiC TEOS SiGe SiGe TEOS
Epitaxial Silicon
P+ Silicon <100>
Polishing the Tungsten plugs down to be coplanar with the tops of the
gate electrodes is something new and was first introduced at the 32nm
Height (L)
node. Shorter Tungsten plugs have a lower contact resistance. This
feature, coupled with the raised Source/Drains for both the NMOS and
PMOS transistors, ensure that the contact resistance is kept to
TEOS SiGe SiGe TEOS SiC SiC TEOS
Poly Gate SiGe acceptableSiGe
values. TEOS
Cu Line
It is difficult to eliminate Tungsten plugs completely and adopt Cu
plugs because it is too dangerous to bring the Cu all the way down to
N-Well P-Well N-Well
the Source/Drains regions.
W Trench
© 2016 Threshold Systems Inc.
Epitaxial Silicon Source: ICE
Conductor
P+ Silicon <100> Resistance
© 2018 Threshold Systems Inc.
Length
L
R=ρ Conductor
Ax Cross-sectional The data displayed on the graph to the left makes it plain why contact
Area plugs must be polished back to such a short height at the 20nm node.
The resistivity of Tungsten is such that at the contact diameters
Resistivity required at the 20nm node the plug resistance soars to unacceptable
levels. However, because resistance is a function of length, shortening
22nm the length (i.e. the height) of the plug reduces the plugs overall
resistance.
The Tungsten overburden is then polished back until the plugs are so
short that they are coplanar with the tops of the gate electrodes.
TEOS SiC SiC TEOS SiGe SiGe TEOS SiC SiC TEOS SiGe SiGe TEOS
Epitaxial Silicon
P+ Silicon <100>
© 2016 Threshold Systems Inc.
Epitaxial Silicon The Tungsten plugs cannot be eliminated altogether and replaced with
P+ Silicon <100> Cu plugs because it is too dangerous to bring the Cu all the way down
© 2016 Threshold Systems Inc. to the Source/Drains.
Device Module:
Gate Last
P-Well N-Well
pitaxial Silicon
con <100>
© 2016 Threshold Systems Inc.
The Tungsten overburden is then polished back until the trenches are
so short that they are coplanar with the tops of the gate electrodes.
G
PS
STI SiC
SiC
STI ell
SiGe
P-Wel SiGe
W
l STI SiC P-
N-We SiC
ll
STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon <
m
100> STI
ste
N-We
ll
Sy
ld
hos
re
Th
© 2018 Threshold Systems Inc.
16
20
©
Gate Electrode
Tungsten Trench
Polishing the Tungsten trenches down to be coplanar with the tops of the
gate electrodes is relatively new and was first introduced at the 32nm
node. Shorter Tungsten trenches have a lower contact resistance. This
feature, coupled with the raised Source/Drains for both the NMOS and
PMOS transistors, ensure that the contact resistance is kept to acceptable
Nitride
STI values.
SiC
SiC Spacer
STI SiGe
P-Wel SiGe ell The Tungsten trenches cannot be eliminated altogether and replaced with
l -W
N-We STI SiC P
ll SiC Copper plugs because it is too dangerous to bring the Copper lines all the
STI SiGe way down to the Source/Drains because Copper is a transistor poison.
.
nc
icon <
m
100>
ste
N-We STI
Sy
ll
ld
sho
re
Th
16
20
©
Once the top of the gate electrode has been opened-up, the doped
amorphous silicon gate mandrel is etched away. Thus, the gate electrode
is said to be “sacrificial”. Its only purpose was as a place-holder,
commonly referred to as a “Mandrel”.
SiC G
STI
SiC PS This etch also removes the oxide offset spacer that was grown on the
SiGe
P-Wel STI SiGe ell sidewalls of the gate cavity as well as the oxide etch stop layer at the
l -W
N-We STI SiC P bottom of the gate cavity.
ll SiC
STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon <
m
100>
ste
N-We STI
Sy
ll
ld
ho
s
re
Th
16
20
Gate Electrode
REMOVED! © © 2018 Threshold Systems Inc.
N-Well This etch also removes the oxide offset spacer that was grown on the
sidewalls of the gate cavity as well as the oxide etch stop layer at the
bottom of the gate cavity.
con <100>
© 2016 Threshold Systems Inc.
These factors, combined with the fact that the SiGe has been moved
closer to the gate edge, increased the PMOS drive current substantially.
Source: Intel Source: Intel
pitaxial Silicon
con <100>
© 2016 Threshold Systems Inc.
Hi-k (HfO2)
Hi-k Deposition:
Next, a thin ~18Å thick layer of Hafnium oxide (HfO2) is deposited using
TEOS SiC SiC TEOS SiGe SiGe TEOS ALD.
P-Well N-Well
pitaxial Silicon
con <100>
©
© 2010
2016Threshold
ThresholdSystems
Systems Inc.
TaN ESL
A ~1.0 nm layer of TaN is deposited using PVD. This layer will serve as an
SiC SiC SiGe SiGe Etch Stop layer (ESL).
TEOS TEOS TEOS
P-Well N-Well
pitaxial Silicon
con <100>
© 2016 Threshold Systems Inc.
A ~5.0 nm layer of TiN is deposited using PVD. This is the first layer of
back-fill metals in the gate stack that will provide strain and electrical
TEOS SiC SiC TEOS SiGe SiGe TEOS contact to the bottom of the gate stack.
pitaxial Silicon
con <100>
© 2016 Threshold Systems Inc.
P-Well N-Well
pitaxial Silicon
con <100>
© 2016 Threshold Systems Inc.
The exposed TiN layer is etched away. The TaN acts as an etch-stop layer
SiC SiC SiGe SiGe to this etch. The integrity of the very thin (<10Å) TaN layer is compromised
TEOS TEOS TEOS
by the TiN etch as indicated.
The fact that the integrity of the TaN layer is compromised is helpful when
P-Well N-Well
we diffuse Aluminum into the underlying TiN layer in the NMOS transistor
to form the NMOS TiAlN Work Function metal later in the process.
pitaxial Silicon
con <100>
© 2016 Threshold Systems Inc.
Photoresist Strip:
The photoresist is stripped off and the wafers are cleaned.
P-Well N-Well
pitaxial Silicon
con <100>
© 2016 Threshold Systems Inc.
P-Well N-Well
pitaxial Silicon
con <100>
© 2016 Threshold Systems Inc.
TiAl
Anneal:
pitaxial Silicon
con <100>
© 2016 Threshold Systems Inc.
The anneal that causes the formation of the TiAlN Work Function metal
is performed at a temperature below 400 º C. Aluminum is very
diffusive at this temperature and since the TaN which separates it from
the underlying TiN in the NMOS device was previously compromises (it
was used as an etch stop layer) the Aluminum freely diffuses into the
TiN and forms TiAlN which is the desired NMOS Work Function metal.
SiC SiC SiGe SiGe This metal acts as an electrical contact from the top of the gate electrode
TEOS TEOS TEOS
to the bottom.
P-Well N-Well
pitaxial Silicon
con <100>
© 2016 Threshold Systems Inc.
AlTiO Layer
CMP Polish:
.
nc
P+ Sil P-Wel SiGe
l
sI
icon <
m
100> N-We STI
ste
ll
Sy
ld
ho
s
re
Th
16
20
©
© 2018 Threshold Systems Inc.
AlTiO Layer
AlTiO Layer
TEOS SiC SiC TEOS SiGe SiGe TEOS
P-Well N-Well
pitaxial Silicon
con <100>
© 2016 Threshold Systems Inc.
NMOS PMOS
TiN
TaN TaN
TiN AlTiO
HfO2
HfO2
AlTiO
AlTi TiN WF
Metal
TiN
TiAl
P-Well
P-Well
© 2016 Threshold Systems Inc.
N-Well
N-Well
© 2016 Threshold Systems Inc.
The metal stack in both the gate cavities is very compressive in the Z-
direction. It pushes down on the transistor channel and introduces
tensile strain which enhances the drive current of the NMOS device.
P-Well
© 2016 Threshold Systems Inc.
N-Well
This tensile strain in the PMOS device degrades the drive current of the
PMOS device somewhat, but the induced tensile force is overwhelmed
© 2018 Threshold Systems Inc. by the compressive force introduced by the SiGe located in the Source/
Drains.
licon PMOS Stack of AlTiO + AiT + TiN + TiAl + TiN + TaN + TiN
n <100>
This complex metallization structure is what
makes the Replacement Gate so difficult!
TEOS SiGe SiGe TEOS
N-Well
© 2016 Threshold Systems Inc.
Next, four layers of Silicon Nitride is deposited over the surface of the
wafer. This seals the tops of the transistors and will acts as an etch
STI SiC G stop layer for the contact trenches that have yet to be fabricated.
SiC PS ell
STI SiGe
W
P-Wel
l
SiGe P-
STI SiC
N-We SiC
ll
STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon < STI
m
100> N-We
ste
ll
Sy
ld
hos
re
Th
16
20
©
© 2018 Threshold Systems Inc.
CESL
P-Well N-Well
pitaxial Silicon
con <100>
© 2016 Threshold Systems Inc.
TiN
TEOS
The first Inter-layer Dielectric (ILD) is deposited. It consists of a ~1,000Å
layer of Carbon Doped Oxide (CDO) that is UV cured after deposition to
bake out porogens and lower its k value.
This is followed by the deposition of a 400Å TEOS CMP stop layer and
Carbon
D CDO followed by a 300Å TiN hard mask.
oped O
xide G
STI SiC PS
SiC
P-Wel
STI SiGe ell
l
SiGe
-W
STI SiC P
N-We SiC
ll
STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon <
m
100> STI
ste
N-We
ll
Sy
ld
ho
s
re
Th
14
©
20 © 2018 Threshold Systems Inc.
Next, the wafers are primed in HMDS (to promote photoresist adhesion)
and coated with Bottom Anti-Reflective Coating (BARC) and a layer of
photoresist.
Photores
is t
O
CD
CDO
G
STI SiC PS
SiC
STI ell
SiGe
P-Wel SiGe
W
l STI SiC P-
N-We SiC
ll
STI SiGe
.
nc
icon <
m
100> STI
ste
N-We
ll
Sy
ld
s ho
re
Th
14
20
The wafers are then soft-baked to give the resist a measure of structural
integrity, then the resist is exposed and developed. The resulting
TEOS photoresist pattern then undergoes a Post Exposure UV Bake (PEB) to
finalize the photoresist pattern and to turn the photoresist into a cross-
PR BARC
PR linked imidized mass that will resist etching.
PR
PR
O
CD
CDO
G
STI SiC PS
SiC
STI ell
SiGe -W
P-Wel SiGe P
l STI SiC
N-We SiC
ll STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon < STI
m
100>
ste
N-We
ll
Sy
ld
hos
re
Th
14
20
© © 2018 Threshold Systems Inc.
TEOS
O
CDO CD
G
STI SiC PS
SiC
STI ell
SiGe W
P-Wel SiGe P-
l STI SiC
N-We SiC
ll STI SiGe
.
nc
100>
ste
N-We
ll
Sy
ld
ho
s
re
Th
14
20
105
©
The photoresist and the BARC are stripped away and the wafers are
cleaned. The TiN hard mask has now been fully formed.
O
CDO CD
G
STI SiC PS
SiC
STI ell
SiGe W
P-Wel SiGe P-
l STI SiC
N-We SiC
ll STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon < STI
m
100>
ste
N-We
ll
Sy
ld
ho
s
re
Th
14
20
©
© 2018 Threshold Systems Inc.
TiN Mask
Using the TiN hard mask as a pattern, a highly anisotropic etch cuts
through the TEOS and CDO layers to define the metal-1 trenches.
The etch stops on the SiN etch-stop layer.
Nitride
barrier
G
SiC PS
SiC
STI ell
SiGe W
STI P-Wel SiGe P-
l STI SiC
N-We SiC
ll STI SiGe
.
nc
100>
ste
N-We
ll
Sy
ld
hos
re
Th
14
20
©
The wafers are cleaned and soft Argon sputter etch opens up the
Silicon Nitride etch stop layer at the bottom of the trench.
G Ta/TaN
STI SiC PS
SiC
STI ell
SiGe -W
P-Wel SiGe P
l STI SiC
N-We SiC
ll STI SiGe CDO CDO
.
nc
P+ Sil P-Wel SiGe
l
sI
icon < STI
m
100>
ste
N-We
ll
Sy
ld
BLoK
hos
re
Th
14
20
©
Source: Applied Materials
100>
ste
N-We
Sy
ll
ld
hos
re
Th
14
20
©
The barrier material of choice is TaN to entomb the Copper in the Dual Damascene trenches and vias. Ta is the liner that
facilitates the Copper fill of these structures.
However, Ta cannot be directly deposited onto an oxide or it will configure as the high resistance beta-phase. Since the Ta
is competing with Cu for volume in the trench, the Beta phase of Ta must be avoided.
TaN adheres relatively well to oxide and is deposited first. TaN is also chemically stable and unreactive with Copper.
Ta is deposited second because when it is deposited on top of TaN it manifests itself in the low resistivity alpha phase
which is highly desirable
Copper is a transistor poison and cannot be allowed to contaminate the active layer. Therefore it must be entombed on all
sides by diffusion barriers (Ta/TaN on five sides & a SiN seal on the top).
TaN/Ta Diffusion
Barrier
Trench
Cu Seed
G
STI SiC PS
SiC
STI ell Ta/TaN
SiGe W
P-Wel SiGe P-
l STI SiC
N-We SiC
ll STI SiGe
Cu Seed
.
nc
P+ Sil P-Wel SiGe
l
sI
icon < STI
m
100> N-We CDO
ste
ll
Sy
ld
hos
re
Th
14
20
©
© 2018 Threshold Systems Inc. Underlying Cu Line
100>
ste
N-We
ll
Sy
ld
hos
re
Th
14
20
©
TEOS
The Copper is polished back using CMP. This polishing also removes the
Tantalum that was on the upper surfaces of the TEOS. The softer TEOS
acts as a polish stop for the CMP process.
.
nc
P+ Sil P-Wel SiGe
l
sI
icon < STI
m
100>
ste
N-We
Sy
ll
ld
hos
re
Th
14
20
©
© 2018 Threshold Systems Inc.
30 Å of SRN < 30 Å of CuSix < 30 Å of SiN
The Cu surface is cleaned of organic impurities left behind from the CMP
clean. This is accomplished by employing a multiple step plasma assisted
reducing chemical that eliminates Cu oxide and other organic impurities.
Copper Copper Copper Next, a thin 30Å layer of Silicon Rich Nitride (SRN) is deposited that reacts
with the Cu to form a Cu silicide (CuSix). This is a self-limiting reaction and
the CuSix limits the Cu diffusion by locking-up the Cu on the upper surface
of the trench in the SRN matrix. The CuSix is electrically leaky so it is next
subjected to a nitridization step to convert it to SiN film.
30 Å of SRN
A thin 30Å layer of Silicon Rich Nitride (SRN) is deposited across the wafer.
Cu
Cu
Cu
G
STI SiC Cu PS
SiC
STI ell
SiGe
Cu -W
P-Wel SiGe P
l STI SiC
N-We SiC
ll STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon < STI
m
100>
ste
N-We
ll
Sy
ld
ho
s
re
Th
14
20
©
© 2018 Threshold Systems Inc.
30 Å of SRN
< 30 Å of CuSix A thin 30Å layer of Silicon Rich Nitride (SRN) reacts with the Copper metal
to form a Copper silicide (CuSix).
Cu
Cu
Cu
G
STI SiC Cu PS
SiC
STI ell
SiGe
Cu -W
P-Wel SiGe P
l STI SiC
N-We SiC
ll STI SiGe
.
nc
100>
ste
N-We
ll
Sy
ld
ho
s
re
Th
14
20
©
SiN
The CuSix is then subjected to a nitridization step to convert it to SiN
film.
Cu
Cu
Cu
G
STI SiC Cu PS
SiC
STI ell
SiGe
Cu -W
P-Wel SiGe P
l STI SiC
N-We SiC
ll STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon < STI
m
100>
ste
N-We
ll
Sy
ld
hos
re
Th
14
20
©
© 2018 Threshold Systems Inc.
SiCN
Cu
Cu
Cu
G
STI SiC Cu PS
SiC
STI ell
SiGe
Cu W
P-Wel SiGe P-
l STI SiC
N-We SiC
ll STI SiGe
.
nc
100>
ste
N-We
ll
Sy
ld
hos
re
Th
14
20
©
Input
Vdd
C
NMOS pulls Input
up to Vdd
GND
B
Output
Cross-Talk
Dielectric Capacitance
ῖ = RC
Time Delay
Conductor
Resistance
1 2 3
© 2018 Threshold Systems Inc. Low-k Dielectric
Interconnect delay can be nearly 20 times greater than gate delay at the © 2011 Threshold Systems Inc.
Copper alone is insufficient to minimize interconnect delay. Low-k Line 1 Low Voltage
dielectrics must be used between the Cu lines to minimize capacitive
coupling between adjacent metal lines, and to a lesser extent, between
Line 2 High Voltage
stacked metal lines.
Low Voltage
The use of Cu metal lines and Low-k dielectrics dramatically reduces the RC
time constant enhancing chip speed.
High Voltage
Low-k materials will not easily support an electric field, and therefore inhibit Line 3 Low Voltage
adjacent metal lines from capacitively coupling. This minimizes parasitic Time
capacitances and lowers the probability of cross-talk between adjacent
metal lines, a mechanism that can result in data corruption. Adjacent metal lines embedded in a dielectric will capacitively couple with each
other. At very small geometries, and especially at the 22nm node, this capacitive
coupling will lead to cross-talk between the lines. This results in signal distortion.
A worst case scenario occurs when one metal line is transitioning into a low
voltage state, when lines on either side of it are transitioning to a high voltage
state. The line that is attempting to enter the low-voltage state will experience
strong cross talk as a result of capacitive coupling, and will distort the digital
signal that it is carrying. A low-k dielectric is extremely helpful in avoiding this
kind of problem because it will not readily support the electric field that
facilitates cross-talk.
The baseline generation of backend dielectrics was SiO2 which served the industry well
for several generations. It was hard, easy to etch, had good coefficients of thermal Low-K Dielectric Generations
expansion and conduction, a high breakdown voltage and was an excellent insulator.
However, with a K-value of 3.9 - 4.2 it was an unacceptable dielectric for advanced Dielectric Constant Vs. Hardness
technology nodes. 5
SiO2
90 nm node. It is simply SiO2 that has been doped with Fluorine to make it less polar and 3
OSG
reduce its k-value to ~ 3.5
!
UL-k
The second generation of low-k dielectrics was carbon doped oxide (CDO), which are
also known as Organo-Silicate Glass (OSG), or more specifically as SiCOH (marketed as 1
0 1 2 3 4 5
“Black Diamond”). This material has a k-value of ~ 3.0 and saw widespread use at the 90 Hardness (GPa)
nm and 65 nm nodes. The k-value can be lowered to 2.7 by increasing the Carbon
content of the film, and a porous version of this dielectric is used at the 45/32/28nm As the k-value of a dielectric film declines so does its hardness.
nodes having a k-value of ~2.5 This reduction in mechanical strength introduces a range of
serious processing issues that has become one of the biggest
A third generation of low-k dielectrics consisting of porous polymers with k-values <2.0 challenges for sub-32nm node technology.
has proven to be elusive.
Photoresist
Carbon D
o ped Oxide
(CDO)
Cu
Cu
Cu
G
PS
STI SiC
P-Wel
SiC
l
STI SiGe
SiGe
STI SiC
Cu
Cu P-
W
ell
© 2018 Threshold Systems Inc.
N-We SiC
ll STI SiGe
.
nc
P+ Sil P-Wel SiGe
l
sI
icon < STI
m
100> N-We ste
ll
Sy
ld
hos
re
Th
14
20
©
Carbon
D oped O
xide (C
Cu DO)
Cu
Cu G
PS ll
e
Cu W
P-
Cu
.
nc
sI
m
ste
Sy
ld
hos
re
20
Th
14
© 2018 Threshold Systems Inc.
©
Carbon
D oped O
xide (C
Cu DO)
Cu
Cu G
PS l
el
Cu P-
W
Cu
.
nc
sI
m
ste
Sy
STI
ld
N-We
ho
ll
s
re
Th
14
20
.
nc
sI
m
ste
Sy
ld
hos
re
Th
14
20
© © 2018 Threshold Systems Inc.
The vias are partially etched in a highly anisotropic etch recipe.
They are not etched all the way down to the SiCN etch-stop layer.
Photores
ist
Photores
is t
Photores
is t x ide
dO
Via o pe
Hole nD
rbo
Cu Ca
Cu
Cu
Cu
Cu
.
nc
sI
m
ste
Sy
ld
ho
s
re
Th
14
20
©
The BARC and the photoresist are stripped and the wafers are
cleaned.
id e
Ox
Via
p ed
Hole Do Trench
on
Cu rb
Ca
Cu
Cu Via
Cu
Cu
.
nc
sI
m
ste
Sy
Source: Lam Research
ld
ho
s
re
Th
14
20
© © 2018 Threshold Systems Inc.
Using the TiN hard mask pattern as a template, the Metal-2 trenches
are etched. During this operation the vias are fully etched. The via
etch stops on the SiCN etch stop layer and may etch slightly into this
layer.
Trench ide
Ox
p ed Trench
Via Do
n
Cu
Hole
SiCN rbo
Ca
Cu Via
Cu
Cu
Cu
.
nc
sI
Note that the SiCN barrier at the bottom of the vias has not yet been
punctured.
Cu Cu Cu Cu Cu
Cu
sI
m
Cu
ste
Sy
ld
Cu
hos
re
Cu
Th
14
20
©
Etch-Stop
Opening
Degas Clean Sputter Clean TaN Deposition TaN Re-sputter Ta Deposition Cu Seed Deposition
The wafers are heated up A clean is applied The bottom of the via A thin layer of TaN is The TaN is sputtered in & Re-sputter A thin ~ 300Å layer of
to ~ 200 C to boil off water (usually a HN03/HF structure is then sputter deposited using an Ar sputter etch. This A very thin Layer of Ta Cu is deposited via
and other contaminants blend) to further remove clean with a “soft” Ar Ionized Metal Plasma redistributes the TaN is then deposited via PVD
any etch residues in the sputter etch to remove (IMP) from the bottom of the IMP PVD & then re- Critical steps
Dual Damascene any residual via and trench & sputtered for 28 nm
structure. The wafers are contaminants. improves side-wall
then dried. coverage
.
nc
sI
m
ste
Sy
ld
hos
re
Th
14
20
©
© 2018 Threshold Systems Inc.
Tan/Ta
A head-on view of the Dual Damascene structure displaying the trench
and numerous via holes coated in Tan/Ta.
Like the contact holes, via holes are very tiny and difficult to fill with
barrier metal. However, this problem is especially important for via
holes since it is critical that they be continuously lined with Ta/TaN in
order to entomb the copper that will later fill them. Any discontinuity
in the Ta/TaN barrier metal will allow copper atoms to leak out of the
Via Hole via and will result in transistor poisoning.
Carbon Doped Oxide
Dedicated Ta/TaN ionized PVD tools combined with a sputter
Cu Cu Cu Cu Cu operation ensure that proper barrier coverage is achieved.
x ide The Copper seed is required to ensure conformal void-free bulk Copper
dO deposition (the following operation) and is a necessary precondition for
o pe
nD the electroplating process.
rbo
Cu Ca
Cu
Cu
Cu
Cu
.
nc
sI
m
ste
Sy
ld
ho
s
re
Th
14
20
©
Cu
nc
sI
m
ste
Sy
ld
ho
s
re
Th
14
20
©
The Copper is polished back using CMP. This polishing also removes the
Tantalum that was on the upper surfaces of the TEOS. The TEOS acts as a
Cu x ide polish stop for the CMP process.
Cu dO
Cu o pe Next, an oxide buff operation is conducted to smooth out the surface of the
Cu nD
rbo TEOS oxide and ensure that the surface of the TEOS is coplanar with the top
Cu Cu Ca of the Copper lines.
Cu
Cu After each CMP step the wafers are cleaned.
Cu
Cu
.
nc
sI
m
ste
Sy
ld
hos
re
Th
14
20
©
© 2018 Threshold Systems Inc.
SiCN
After the same SRN layer is deposited, converted to CuSi and then to
SiN, a 300Å layer of SiCN is deposited to seal off the top of the second
layer of metal. The Copper in the second layer of metal (and its via) is
now entombed with TaN/Ta on five sides and SiCN on the top.
Cu x ide NoteUnderlying
that in thisCu
Dual
LineDamascene integration methodology the
Cu dO photoresist never touched the low-k dielectric.
Cu o pe
D
Cu
b on
r Seven further layers of metal trenches and vias are defined in a similar
Cu Cu Ca
fashion.
Cu
Cu
Cu
Cu
.
nc
sI
m
ste
Sy
ld
hos
re
Th
14
20
©
Conductor
Length
Resistance
Metal 9
L © 2018 Threshold Systems Inc.
R=ρ Conductor
Ax Cross-sectional
Area
Resistivity M8
We want R to be LOW! ῖ = RC V
i V
a i
a
M7
M6
M5
© 2011 Threshold Systems
M4
M3
Trench Trench Trench Trench
V V V V V V V M2
i i i i i i i
a a a a a a a M1
Trench Trench Trench Trench
© 2014 Thre
shold Systems
Inc.
© 2011 Threshold Systems
M7
M5
M2 Cu via
© 2014 Thre
shold Systems
Inc.
Copper Pillars
Cu Pillar
Cu Pillars Cu Pillars
The conventional solder balls have been replaced at the 45/32 nm node with “Copper Pillars” to
© 2010 Chipworks - Courtesy of Chipworks comply with RoHS (Restriction of Hazardous Substances) and to act as a Peltier effect cooling conduit
to drain heat away from the chip.
Incorporated 3rd generation Metal Gate High-k technology This generation of microchips demanded the extensive use of 193
with a more aggressively scaled dielectric (EOT reduced from immersion lithography and extremely sophisticated and
10Å to 9Å). expensive reticles sets.
Strained silicon continued to provided a dramatic speed boost The requirement for damascene-like replacement gates and their
and this technology node employed 4th generation strain associated complex internal metallization morphologies
technology that incorporating 40% Ge in the SiGe of the PMOS dramatically increased processing complexity.
and the use of 1st generation SiC in the NMOS Source/Drains.
Ultra-thin barriers for copper metallization faced new deposition
The lithography process moved from 193 dry litho to 193i challenges at the 20nm nodes as the metal lines scaled to
immersion lithography. extremely small dimensions.
NMOS S/D leakage has been reduced 5X; PMOS S/D leakage The extremely narrow gate electrodes present at the 20nm node
has been reduced 10X. posed special etch uniformity problems that required innovative
The use of porous CDO (Carbon Doped Oxide) significantly forward/feedback mechanisms to control.
lowered the k value of the inter-layer dielectric stack, which in
The ultra-shallow Source/Drain regions required complex and
turn permitted higher back-end speeds.
careful implant schemes and millisecond spike anneal
TiN metal hard masks substantially improved trench technology.
morphology and enabled very narrow trenches to be cleanly
defined. The cost of a manufacturing facility dedicated to the production
of 20 nm devices easily approached or exceeded seven billion+
Amorphous carbon hard masks greatly enhanced the dollars.
definition of STI, gate electrodes and narrow contact holes.
The high level of processing complexity associated with the
The use of double patterning dramatically improved gate replacement metal gate significantly increased overall cost.
electrode scaling.
There are significant questions regarding the scalability of the
Transistor counts at the 20 nm node reached 1.7+ billion.
replacement gate methodology, particularly at the 7 nm node.