Possible Paths for Cu CMP

Transcription

1 Possible Paths for Cu CMP J.S. Drewery, V. Hardikar, S.T. Mayer, H. Meinhold, F. Juarez, and J. Svirchevski Presented by Julia Svirchevski

2 Agenda Perceived Need for ECMP Technology Differentiation Profile Control Electrical Contact Design Residual Cu Planarization Capability Feature Level Planarization Die Level Planarization Perception and Reality of Ecmp 2

3 Perceived Need for ECMP What does it enable? Enables: Ultra-low k material integration Low down force ( psi) may be required for dielectrics with k= High removal rate at low down force Throughput High planarization efficiency process Dishing and erosion requirements Lower CoC 3

4 Removal Rate Response Impact of Current and Pressure Copper Removal Rate Copper Removal Rate (A/min) R 2 = P=1psi Current (A) (A/min) Pressure (psi) I=9A ECMP removal rate proportional to current Pressure insensitive Pressure Novellus insensitive Confidential 4

5 Differentiated Profile Control Rotary versus Orbital Platform Rotary ECMP system Orbital ECMP system Single Anodic point of contact Zone 3 Multi-zone Cathode Multiple Anodic Pad Contacts Zone 2 Zone 1 wafer Single contact terminal effect, edge burn Complexity of zonal control Removal rate variations across the wafer Multiple contacts minimize terminal effect Orbital motion enables easy zonal control Ideally suited for global & local uniformity control Orbital format is ideal for minimizing global/local scale uniformity 5 Orbital format is ideal for minimizing global/local scale uniformity

6 Electrical Contact Comparison Impact on Cu global and local uniformity Novellus Multiple Contacts Single Contact 1.8 inch 12 inches Voltage Drop Voltage Drop Contacts Thinner residual Cu <1,500 Å Uniform removal ( voltage drop V) Contact 2,000-1,500 Å residual Cu, Edge burn Non uniform removal (voltage drop 0.51V) Multi-contact system enables minimal voltage drop => uniform Cu removal Edge burn with single-contact systems 6 Edge burn with Novellus single-contact Confidential systems

7 Residual Cu Multi-contact approach compared to single point Minimum rate/avg rate (Minimum / average removal rate) vs. residual thickness Multi-contact ECMP Point contact Remaining Cu thickness (A) A clear path to enable 10x less remaining Cu with 7 Smart Pad multi-contact system Smart Pad multi-contact system Based on Modeled calculations

8 Segmented Cathode 0.3psi, Current Density 30mA/cm Outer Zone ON Center Zone OFF Outer Zone OFF Center Zone ON Removal Rate Å/min Position mm Segmented Cathode allows wide range center-to-edge profile adjustment center-to-edge profile adjustment 8

9 Feature Level Planarization Typical planarization of 100um x 100um line structures Height difference center-edge = 0±51Å 50 microns 0.65 microns Feature level planarization of <100A w/nvls process 9 Feature level planarization of <100A w/nvls process

10 Die Level Planarization DLP Intensified downforce ECMP exhibits pattern sensitivity ( DLP ) High removal rate in areas of low pattern density Analogous to CMP Pad Intensified electric current Slurry Cu Electrolyte Cu CMP ECMP Downforce, pad structure and electrolyte composition are key controls allowing DLP to be optimized Low down force ECMP enables better topography and planarization than CMP 10 and planarization than CMP

11 Die Level Planarization Down force Impact Thickness variation (Angstrom) Necessary low down force Nominal Pressure (psi) ECMP requires low down force processing similar 11 to conventional CMP to conventional CMP CMP DLP NVLS DLP target

12 Electrolyte Flow Rate Studies Flow rate range ml/min Removal (Angstrom) ml/min 75 ml/min 100 ml/min Diameter Scan Stable Removal rate with reduced electrolyte flow <100 ml/min Transparent reduction of slurry/electrolyte on low CoC Orbital platform 12 Transparent reduction of slurry/electrolyte on low CoC Orbital platform

13 Die Level Planarization Optimized Process 150ml/min Proprietary Electrolyte flow rate; 0.2psi 3040Å 3040Å 3040Å 3040Å 0.25x0.25um 0.5x0.5um 9x1um 100x100um DLP variation of 0Å Full Planarization with Optimized approach!! Full Planarization with Novellus Optimized Confidential approach!! 13

14 Comparison of Advanced Cu slurries performance to ECMP Selected Attributes *Published ECMP Performance Removal Rate (A/min) *~6000 Down Force (psi) *0.3 Ecmp Comments NVLS Ecmp NVLS Cu CMP Widely used Cu slurries Advanced Cu slurries RR dow nforce insensitive Low dow nforce required for better DLP Planarization Efficiency (%) *95% Marginally better 95% 80% 85% Remaining Cu thickness after Cu bulk (A) * issues w ith preexisting topo DLP or thickness variation after Bulk removal or Ecmp, (100X100um, 9x1um, *200 DLP is challenging <100, (some sites 0) Final Dishing after topo correction (A) * TBD <300 Total System (inculdes h/w, barrier step) CoC ($/w ) Electrolyte contains abrasives *Published Ecmp Performance (Source IITC 2003 paper L. Economikos et. Integrated Electro-Chemical Mechanical Planarization (Ecmp) for Future Generation Device Technology) Performance of Advanced Cu slurries is comparable with Ecmp is comparable with Ecmp 14 meets requirements marginal

15 Perception & Reality of ECMP Summary Realities of Ecmp is not an enabling technology yet.. Can not remove all Cu, underlying topography defines Cu thickness after Ecmp Ultra low down force (0.3psi) is not required for low k integration psi is suitable for (k= ) dielectrics Ecmp contact reliability is not mature for volume manufacturing Ecmp is not as cheap as advertised Reality is $12-14/wafer pass at low metal levels (M1-M5), higher at (M6-M10) Advanced emerging Cu slurries yield similar performance Substantial improvements in planarization efficiency and topography performance Dishing performance similar with Ecmp Adequate removal rates at lower downforce Availability of cheap Cu slurries of similar price range as Ecmp electrolyte Advanced slurry Cu CMP is a viable & safer choice selected over Ecmp for volume manufacturing at 65nm and 45nm 15

16 Acknowledgments Thanks to the authors and the team that conducted this work: John Drewery Steve Mayer Vishwas Hardikar Henner Meinhold Francisco Juarez 16

17 ULK CORAL k=2.5 CMP Integration Capability Sematech Mask 400AZ Dielectric ULK 5000 Å HMS Coral cap 700 Å Cap after CMP 200 Å kstop 700 Å Passivation TEOS 1 um 15% reduction in RC delay No CMP related delamination Consistent 92% yield Measured on 120-dies for each of 8 wafers Successful Integration of ULK Coral TM k=2.5 Accomplished ULK Coral TM k=2.5 Novellus Accomplished Confidential 17