controlled cobalt recess for advanced interconnect ... · 4/1/2019 · challenges for metal...
TRANSCRIPT
PUBLIC
CONTROLLED COBALT RECESS FOR ADVANCED
INTERCONNECT METALLIZATION.
Antoine Paccoa*, Y. Akanishib, Q.T. Lea, E. Kestersa, G. Murdocha, F. Holsteynsa
A IMEC VZW, KAPELDREEF 75, 3001 LEUVEN, BELGIUMBSCREEN, SEMICONDUCTOR SOLUTIONS CO., LTD., 480-1, TAKAMIYA-CHO, HIKONE, SHIGA 522-0292, JAPAN
*EMAIL: [email protected]
OVERVIEW
copper cobalt transition
challenges for metal recess & specifically for cobalt recess
specifications for wet recess
approach & results: Co recess by digital wet etch
effect of different oxidizing solutions
effect of oxide dissolution step & rinse step times
effect of the ambient
wet recess of cobalt for FSAV
Electrical testing
Morphological results
conclusions
2
WHEN & WHY COBALT OFFERS AN ADVANTAGE OVER COPPER
AS AN INTERCONNECT METAL
1. The electron mean free path of cobalt is
considerably lower than copper, reducing
surface scattering in small trenches
lower line resistivities compared to
copper.
2. A thinner combined barrier-nucleation
layer can be used for cobalt lower
cross-sectional resistance.
3. Cobalt can be annealed at reasonable
thermal budgets thereby enabling grain
growth which reduces resistivity.
3
Ref.: Applied Materials Internal Resistivity
Benchmark
WHEN & WHY COBALT OFFERS AN ADVANTAGE OVER COPPER
AS AN INTERCONNECT METAL
1. The electron mean free path of cobalt is
considerably lower than copper, reducing
surface scattering in small trenches
lower line resistivities compared to
copper.
2. A thinner combined barrier-nucleation
layer can be used for cobalt lower
cross-sectional resistance.
3. Cobalt can be annealed at reasonable
thermal budgets thereby enabling grain
growth which reduces resistivity.
4
will cobalt climb up the interconnect stack?
taking place....
COBALT RECESS FOR FSAV (FULLY SELF ALIGNED VIA)
M1 recess M2V1 FillM2V1 Etching
Target = 10nm
metal recess
Topography for SAV
Via confined in y-direction by M2 HM
Via-first etch lands selectively on SiCN barrier
Topography is maintained by selective removal
of SiCN
FSAV is formed
Min-distance M1-V1
is maintained by
topography
5
M2
V1
M1
CHALLENGES FOR METAL RECESS
6
NON-IDEAL (NON-UNIFORM RECESS) CASES
wet metal recess can be affected by:
1. Crystal orientations
• different crystallographic planes can have
different etching rates (anisotropy)
2. Crystal grain boundaries
• preferential etching can occur at grain
boundaries
• smaller grain size grain boundary density
increases
3. Galvanic corrosion
• Local increase in oxidation / etch rate due to
galvanic contact of the fill metal with the liner/
barrier metals
6
crys
tal
grai
n b
oundar
y
CHALLENGES FOR METAL RECESS: COBALT
7
• non-uniform (crystal-grain/boundary dependent) etch
• uncontrolled dissolution of cobalt during slightly acidic / neutral rinse alkaline solutions, NH4OH/H2O2 preferred (*,**)
• selective removal of TiN liner/barrier after or simultaneously during Co recess
SPECIFICATIONS: • control of the recess amount within the nanometer range: 10 +/- 1 nm
• small within-line roughness (Rq ~ 1nm)
• uniform within-wafer recess (center-to-edge NU < 1nm)
• avoid pattern-density-depending etching (CD-independent)
cyclic process (digital etch) = process of choice
* SPCC ’18 Y. Ogawa, [Kurita]
** SPCC ‘19 H. Iino, [Kurita]
APPROACH FOR CONTROLLED RECESS
8
“DIGITAL ETCHING”: 0-1-0-1-0-1-0-1-....
Potential benefits of this approach:
• recess amount can be controlled within the nanometer range
• uniform within-wafer recess (single wafer processing)
• uniform within-structures recess = non pattern-density-depending etching
Co Co CoCo
CoOx
initial after X cycles“0” oxidation “1” oxide dissolution
X cycles
0 = self-limiting metal-oxide growth on the metal surface
1 = metal-oxide removal selective towards metal surface
EFFECT OF DIFFERENT OXIDIZING SOLUTIONS
9
FOR THE SAME OXIDE DISSOLUTION STEP: 10S HF 0.05%
• increasing the peroxide concentration in
both slightly acidic H2O2/H2O and alkaline
NH4OH/H2O2/H2O solutions does not
significantly affect the total cobalt loss
• removing the oxidizing agent, being it H2O2
or atmospheric O2 does reduce the total
cobalt loss.
• self-limiting oxide film formation when a
sufficiently oxidizing solution is used (H2O2).
# oxidizing
solutions
0
0.5
1
1.5
2
2.5
ref CMP SC1/HF H2O2/HF DIW/HF
Ra
(nm
)COBALT ROUGHNESS AFTER DIFFERENT OXIDIZING SOLUTIONS
10
AFM (ECD COBALT)
ref CMP SC1/HF
H2O2/HF DIW/HF
recess depth approx. 10nm
• Slight (<1nm) roughness increase after recess.
• No significant difference between different oxidizing solutions.
# oxidizing solutions
IMPACT OF DISSOLVED [dO2] IN DILUTED HF (0.05%)
11
EXAGGERATED TIMES & AGGRAVATED CONDITIONS ...
WHAT ABOUT THE IMPACT OF [O2, ATM] IN THE PROCESS AMBIENT ?
+ high roughness
non controllable
recess
EFFECT OF OXIDE DISSOLUTION STEP & RINSE STEP TIMES
12
STD(NON-CONTROLLED) VS. LOW [O2] AMBIENT
• DIW time has a small impact on cobalt loss.
• Increasing the HF time has a more noticeable
impact on the total cobalt loss.
• The parameter that has highest impact is the O2
content in the ambient.
TIN LINER REMOVAL
13
TWO APPROACHES: 1-STEP VS. 2-STEPS
TiN removal most effective
by using a final 60°C APM
1-STEP: combined TiN liner removal & Co
recess etch
2-STEPS: Co recess etch followed by
TiN liner removal
(APM RT /HF) x4 + 30s APM 40°C
(APM RT /HF) x4 + 40s APM 60°C
(APM 40°C /HF) x4
(APM 60°C /HF) x4
ELECTRICAL TESTING
14
TARGET: 10 NM CO RECESS FOR FSAV
Line resistance of 10 µm long lines
21 nm 30 nm CD
Process: “digital” Co recess etch [APM followed by HF] x times
+ TiN barrier etch APM 60°C
Uniform distributions
> 95% yield for smallest CD
21 nm
Electrically active lines
CMP
RECESS
ELECTRICAL TESTING
15
TARGET: 10 NM CO RECESS FOR FSAV
Line resistance of 10 µm long lines
21 nm 30 nm CD
Data follows the expected trend: line resistance increase for
smaller lines due to line X-section reduction (R ~ 1/A)
Limited pattern loading
Measured vs. predicted line resistance before and after 10 nm Co recess
60 nm 50 nm
Co
MORPHOLOGICAL ANALYSIS
16
TEM/EDX
CD ~ 21 nm
CD ~ 30 nm
Limited pattern loading ~10 nm Co recess & TiN liner removal
MORPHOLOGICAL ANALYSIS
17
TEM/EDX
EDGE of the wafer CENTER of the wafer
• Good within-wafer uniformity
• TiN removal height ~ Co recess
CONCLUSIONS
18
Digital etch of cobalt-metal way to go:
1. Promising morphological & electrical results with commodity chemistries
2. Oxidizing step less critical (~ self limiting with H2O2)
3. Predominant role of the oxide dissolving agent (acid) step & DO2
• The local line roughness was relatively high for the ECD cobalt post-CMP & post-recess
etch
• Limited pattern loading observed for CVD cobalt
• Cobalt recess was controlled in the nanometer range (10 +/- 1 nm)
• Good WIW uniformity and yield obtained
PUBLIC