Lithography - Chemical Engineering IIT Madras

Download Report

Transcript Lithography - Chemical Engineering IIT Madras

Overview
 Integration
 Matching/ Accommodating different processes to obtain a ‘passing’
or good chip
 Process Integration
 Design/ Process Integration
 Modifying the design (slightly), to suit the process
 Usually, the electrical circuit is not altered
 Assumes that process has been developed or optimized to the
maximum possible extent
 Always remember that the cost /benefit is the ‘bottom line’
2
16-Jul-15
Index
 Miscellaneous
 Electro migration
 Scaling in MOS
 Electrical Structure (mostly FEOL)
 BEOL Cu vs Al and Low-K vs Oxide
3
16-Jul-15
Electro migration
 Due to movement of electrons, the atoms in a very small
line move
 formation of voids, increase in electrical resistance, complete
breakdown
 hillocks, whiskers ==> shorts
 Minimum length needed “Blech Length” for a given
current density
 Depends on the material
 Copper, Gold good electromigration resistance
 Reliability issue
4
16-Jul-15
Electro migration
 Mean Time to Fail (MTF) depends on Current Density
E 
(J)
1  kTa 
MTF  2 e
J
 Current Density few million Amp/sq cm
 Grain size:
 movement along grain boundaries faster
 For very small lines, grain size = line width
 if av grain size is 1 um, a 10 um line will fail quickly (for the same
current density, a 1 um line will fail later, a 0.5 um line may fail more
quickly)
 also depends on arrangement and orientation of grain
5
16-Jul-15
Electro migration
 Al is more prone to electromigration, however...
 Cu does not form a stable strong oxide
Delamination from the liner is an issue
 EM failures in copper likely to be in the interface, in Al,
in the bulk
Also, W has very high resistance to electromigration
 Length of a path may not become very high
 In Cu, (especially with dual damascene), for the same
design, length may become more
16-Jul-15
6
Reliability
regime (components of
poor quality)
 Working life Regime
 Wear Out regime
Fail/unit time
 Infant Mortality
 Arrhenius equation
Time
 High Temp Operation Test - HTOT
 Highly accelerated Stress Tests (HAST): - high humidity, high
temp
16-Jul-15
7
 Different
mechanisms of fail may
have different activation
energies
 ==> Bimodal
Cumulative % Fails
Reliability
distribution of fails
Temp
 Arrhenius equation is suitable for chemical reactions
 Not always suitable for explaining physical failure mechanisms
8
16-Jul-15
Scaling
 Scaling: If you reduce all the features (or some set of
parameters) by a constant factor, will the performance be
similar? Better? Worse?
 Eg. Important dimensions are TOX, L, X
L
Gate
Source
Depletion Region
16-Jul-15
TOX
Drain
X
9
Scaling
 Information from W. Maly: Atlas of IC Technologies
 If all dimensions are reduced by 2 (for example)
CS0
CS
L/2
X/2
Gate
TOX/2
X/2
 With same diffusion process, to obtain a shallower junction, need to start
with lower initial concentration
==> very high sheet resistance (shallow Jn, low dopant)
16-Jul-15
10
Scaling
 If only some dimensions are scaled
 eg. TOX and L, but not Junction depth
CS0
L/2
X
Gate
TOX/2
X
 Depletion regions merge, short channel
11
16-Jul-15
Scaling
 If dimensions are scaled and dopant concentration
“increased”
CS
CS0
L/2
X
X/2
Gate
TOX/2
X/2
 Tolerable sheet resistance
 ==> High gradient of dopant in the Jn (hot electron effect) ==> LDD etc
12
16-Jul-15
Scaling
 Constant Field :
 Reduce a physical dimension by ‘k’ , then reduce
the voltage also by k
This reduces drain current, gate delay and gate-channel
capacitance by k
 Number of circuits/ sq cm increases by k2
13
16-Jul-15
Scaling
 Power per unit area remains constant
 in CMOS. NMOS and Bipolar do not follow this
scaling rules ==> Issues
 Power Delay product (PDP)
NOTE: Total power, not power per unit area:
 Reduces by cubic
 Practical Limitations by processes (eg. Litho resolution,
alignment, process marginalities)
14
16-Jul-15
Metal Gate
 MOS: using metal gates:
Gate oxide etch/growth
N
N
N
N
Si
N
N
Source/Drain Diff and
oxide growth
16-Jul-15
Contact Etch
15
Metal Gate
Source Gate
N
N
Metal Dep
N
Drain
N
Metal Etch
 Minimum Feature Size is 2l
 Alignment tolerance l
 Based on the above steps, only 5% of the area is used for transistor
channel
 Overlap of gate to source and gate to drain ==> capacitance
16
16-Jul-15
Poly Gate
 MOS: using poly gates:
 Active Mask (STI for example)
 Poly dep/etch/ implantation (self aligned)
 Contact formation
 --- Withstands high temp
 Active channel area is about 12% of total area
 lower capacitance
 better channel definition
 one more interconnect layer (poly)
17
16-Jul-15
Electrical Structures
 Electrical Structures:
 Resistor:
 Poly (silicided) for low resistance and unsilicided
for high resistance
 silicided: 5 to 10 ohm/square
 unsilicided 300 to 500 ohm/square (depends on
dopant level and type)
 undoped silicon (intrinsic) very high resistance, not
usually used
 Can be ‘created’ during gate formation
16-Jul-15
18
Electrical Structures
 Active (source/drain) areas can also be used for high resistance
 Ohmic vs Schottky contact for lightly doped n structure
 P type or highly doped N type
 CAPACITOR:
 MOS capacitor, similar to transistor
 In the BEOL in some cases (Metal Insulator Metal Capacitor )
(Use Al for better thickness control)
 Deep Trench capacitor (FEOL) for DRAM
19
16-Jul-15
Electrical Structures
 INDUCTOR:
 Metal ‘circle’
 Inductance related to number of turns, radius etc
 Parasitic resistance, capacitance
 BEOL
 Diode:
 PN Junction diode
 Schottky Diode
 high reverse bias current
 Many ‘slotted’ N+ in P well, with merging depletion
region, for good reverse bias current (appears as PN in
reverse, as Schottky in fwd)
20
16-Jul-15
FEOL Flow-I
 Review
 Sequence with more detail than what we saw before
 Start with bulk P type silicon
 Oxidize and strip (obtain clean surface)
 Pad oxide and Nitride
 Trench Lithography
 Active (source drain) and gate area definition
 CD measurement for example
 Trench Etch
 post etch CD measurement
 etch bias
21
16-Jul-15
FEOL Flow-II
 Trench Etch
 blind etch
 less than a micron depth
 post etch CD measurement
 etch bias
22
16-Jul-15
FEOL Flow-III
 Trench oxidation (very small thickness)
 TEOS Dep (0.5 micron for example, above active)+ Anneal
 Reverse Active Mask: Litho, etch (about 0.5 micron)
 To reduce the ‘load’ on STI CMP
 load here means planarization load, NOT the removal
load
 STI CMP
 target 0.5 micron removal on small active (faster polish
area) for example
 Other integration LOCOS
 OR modification of STI by Poly Dep+CMP ....
 Nitride Removal
16-Jul-15
23
FEOL Flow-IV
 Pad oxide removal
 Sacrificial oxidation (10 nm)
 Nwell implant
 Pwell implant
 Gate Oxidation
 Poly deposition (perhaps thin cap oxide for hard mask)
 Gate Litho(use ARC)/ Etch
 NMOS implant (LDD, Halo): Not the main implants
 PMOS implant (LDD, Halo): Not the main implants
 RTP
24
16-Jul-15
FEOL Flow-V
 Spacer Formation
 Nitride
 Spacer etch
 NMOS (main) implant
 PMOS (main) implant
 RTP
 If you need unsilicided areas, you need to mask them with oxide/nitride
 Pre-silicide clean (eg. Sputter the surface, or light chemical etch)
 Co dep
 Anneal (about a minute, 500 C)
 Un-reacted Co removal
 Nitride Dep (for contact etch)
25
16-Jul-15
FEOL / BEOL
 BPSG Dep + Anneal + CMP (0.5 um thick remaining)
 Cap Ox (to withstand W CMP in the next steps)
 Contact photo + etch
 Ti/TiN Liner dep (IMP Ti, CVD TiN)
 W CVD
 W CMP
 ILD/PMD Oxide for Copper, Barrier Dep for Al
26
16-Jul-15
BEOL (Aluminum)
 Barrier Dep (Ti/TiN)
 Al Dep (PVD) + Ti/TiN
 Al Photo + Etch (M1 levl)
 Oxide Dep
 Oxide CMP (blind)
 Via12 etch and so on...
27
16-Jul-15
BEOL (Copper)
 Oxide Dep
 M1 photo + Etch
 Liner (Ta/TaN) Dep
 Cu (seed + electrochem) Dep
 Cu CMP
 Nitride (via etch stop layer) dep
 Oxide Dep
 V12-photo + etch (stop on nitride)
 M2 photo + etch (blind) and so on...
OR M2 photo + blind etch followed by V12-photo+ stop on nitride
etch
16-Jul-15
28
BEOL (Copper)
 Via first integration
 Resist removal from the bottom of via can be an issue
 Metal first integration
 Resist pooling an issue
 Some Other schemes:
 Single damascene
 Intermediate (stop on) nitride
 increased cost, lower throughput
29
16-Jul-15
BEOL Integration
 Some Issues:
 Over etch: defectivity
 under etch: opens
 etch uniformity (center edge)
 CMP
 In Copper CMP, too much dishing/erosion ==> oxide
needs to be planarized
 increased cost
30
16-Jul-15
Low -K
 Oxide vs Low K
 For RC delay reduction
 Need
 thermal stability
 chemical inertness (selective etch)
 mechanical strength (for Cu CMP)
 adhesion to barrier
 low stress
 Thermal expansion matching
 low diffusivity of contaminants (or copper)
31
16-Jul-15
Low -K
 Only if k < 3, it is considered as ‘real low-k’
 FSG (fluorinated silicate glass) (k=3.5)
 not very low
 absorbs water (can create HF and hence havoc)
 Carbon Doped Silica (k = 3)
 commercial: AMAT Black Diamond , NOVELUS Coral
 uniform carbon doping is an issue
 Organic polymers
 low k, but soft (mechanical strength), temp stability, chemical
reactivity (photoresist etc)
 thermal expansion mismatch ==> crack
32
16-Jul-15
Low -K
 Organic polymers
 low k, but soft (mechanical strength), temp stability, chemical
reactivity (photoresist etc)
 thermal expansion mismatch ==> crack
 low thermal conductivity ==> more stress on metal
 Adhesion, Delamination
 eg. PTFE adhesion, delamination during CMP
 Absorbance of chemicals during process
 and subsequent release
 --> metal lines and transistors affected
 DLC (diamond like carbon)
16-Jul-15
 etch etc not tested
33
Low -K
 Porous silica (or other materials)
 Thermal, chemical stability is good
 Diffusion of materials is high
 Mechanical strength (CMP) may be an issue
 Absorbance of chemicals an issue
 Good thermal insulator ==> very high stress on metal
 --> need to take heat conduction path in design
 Repeatability (of process and material quality) etc need to be
tested
 New material: Not tested very well yet
34
16-Jul-15
Low -K
 CAP layer used to ‘increase’ mechanical strength
 eg SiC. (Has high K, so only very thin films can be used)
 DLC may also be used
35
16-Jul-15