Analysis Of Large Pad Surface Contact Area In Copper CMP

Analysis of Large Pad SurfaceContact Area inCopper CMPX. Liao 1, Y. Zhuang 1,2, L. Borucki 2, Y. Sampurno 1,2and A. Philipossian 1,21University of Arizona, Tucson AZ USA2 Araca Incorporated, Tucson AZ USA

Introduction, Previous Workand Objective2

Why Do We Need CMP?Multilevel (10-level) Interconnect NetworkSource: T. Ohba, Fujitsu Sci. Technol. J., 38, 17 (2002).3

How Does CMP Work?Carrier HeadSlurry TubePolishing PadWaferPad ConditionerPlatenGeneralized Schematic of a Rotary CMP Tool4

Why is Pad-Wafer Contact Area Important? There have been numerous reports that pad-wafer contactstrongly impacts material removal rate. Majority of previous work has shown that:THE SMALLER THE PAD-WAFER CONTACT AREA,THE HIGHER THE MATERIAL REMOVAL RATE However one investigation has claimed the opposite effectto be true!5

Previous WorkMejia* et al.: J. Electrochem. Soc. 150 (2003) G76 Found that:Total pad-wafer contact area increases as a function of padimmersion time in pH-adjusted water.Local pressure on the wafer surface is reduced asimmersion time and contact area increase.PETEOS material removal rate is consequently reduced.* Sematech/SRC ERC Ex-Student6

Previous WorkBushan: Principles and Applications of Tribology,John and Wiley & Sons, New York (1999) Found that:Real contact pressure in CMP is closely related with padsurface roughness and its composite elastic modulus.Real contact pressure is not related to the applied load. Accordingly:Rougher surfaces cause faster asperity wear and thusincrease the real contact area.Removal rate drops due to declining real contact pressure.7

Previous WorkJeong et al.: J. Adv. Mech. Dsgn. Sys. Mfg. 6 (2012) 113 Found that:Pad surface wear increases the total real contact area.This in turn lowers the average of the real contact pressureat a given applied down force.The reduction of the real contact pressure leads to a declinein material removal rate.8

Previous WorkSun* et al.: Jpn. J. Appl. Phys. 49 (2010) 026501 Found that:The more aggressive the conditioning process, the lower thetotal pad-wafer contact area.Pads with lower contact areas cause PETEOS removal rateto go up during CMP.* Sematech/SRC ERC Ex-Student9

Previous WorkLiao* et al.: Jpn. J. Appl. Phys. 52 (2013) 018001 Found that:Total pad-wafer contact area decreased dramatically athigher conditioning forces.This led to a sharp increase in the local contact pressure.This consequently caused significantly higher PETEOSremoval rates during CMP.* Sematech/SRC ERC Student10

Previous WorkNair et al.: ICPT 2012, October 15-17, 2012, Grenoble, France Found that:*After a certain critical point, total pad-wafer contact areadoes not contribute to TEOS material removal rate. Did not specify as to contact size is considered to be‘large’ and what is considered to be ‘small’.* Blue symbol on Slide 1311

Previous WorkPark et al.: J. Electrochem. Soc. 155 (2008) H595 Found totally opposite trends:*The lower the total pad-wafer contact area, the lower thePETEOS removal rate.* Green symbol on Slide 1312

Removal RateSummary of Previous WorkSomewhat of an‘unknown’ region!TOTAL Contact Area13

Why Are We Focusing on ‘Large’ Contact Area? The relationship between contact area and materialremoval rate in CMP can be complicated. All previous work has focused on the TOTAL contact area,even though contact areas are very different in size andtheir characteristics. As such, each contact area shouldnot be treated the same way. We believe that contact areas need to be studied andclassified more carefully. In this study we are going to classify contact areas asbeing ‘SMALL’ and ‘LARGE’. Sum of small and large areasis assumed to equal the total contact area.14

Laser Confocal Microscopy15

Laser Confocal MicroscopyZeiss LSM 510 Meta NLOPad surface contact area and topography analyses wereperformed throughlaser confocal microscopy.16

How Does a Laser Confocal Microscope Work?Laser LightDetectorBeam splitterPinholeLensOptical SliceFocal PlaneReflected light near thefocal plane reaches theDetector, but out-of-focus17light does not.

Confocal Microscope Pad Sample HolderMounting RingSapphire WindowPad SampleMiniature StageBall BearingSpringLoad CellThreaded ScrewSource: L. Borucki,US Patent Nos. 7,869,027 and 7,839,49618

Confocal Contact Area MeasurementsNear contact reflection or interference fringes (zebras)LoadSapphire windowConfocal optical sliceContactNear contactReflection. Notflat enough tofringe!Far from contactPadNo reflected imageBlack areaContact19

Topography and Contact Area ImagesMean contact percentage is 0.076 at 1.5 PSIIC 1000 Pad450x450 µm0.29% contact450x450 µm0.10% contact20

Example of LARGE Contact Area50 µm21

Topographic Analysis of LARGE Contact Area50 µm22

SEM Analysis of a LARGE Contact Area23

Detailed SEM Image of a LARGE Contact AreaThe SEM image shows that the individual large contact area corresponds tocollapsed pore walls and loosely attached pad debris, suggesting that thelarge contact area may not be fully supported.24

Simulating the Young’s Modulus25

Steps Involved in Simulating the Young’s ModulusLevel ahLevel b Measure pad topography using laser confocalmicroscopy without applying any load. Assume elastic contact between the wafer and the padasperities.Borucki (2009).26

Steps Involved in Simulating the Young’s Modulus Perform a load balance based on the Greenwood andWilliamson elastic model.Assume a certain value for the Young’s modulus.Calculate the pressed displacement ‘h’ of the summits fromheight level ‘a’ to some height level ‘b’ at a certain load. Contour the contact at height level ‘b’. Obtain thesimulated contact area for each asperity at level ‘b’. Sum up the areas of all contact contours to obtain thesimulated total contact area. Compare simulated vs. measured contact areas. Increaseor decrease the Young’s modulus and simulate again untilthe two contact areas are close enough.Borucki (2009).27

Simulation Results540 MPa surface300 MPa surface150 MPa surface101 µm2193 µm2396 µm250 µm50 µm100 MPa surface50 MPa surface594 µm2843 µm250 µm793 µm250 µm50 µmBorucki (2009).28

Assumed Young's Modulus (MPa)Young’s Modulus vs. Surface Contact Area600450300150002004006008001000Simulated Pad Surface Contact Area (µm 2)Borucki (2009).29

Implications of Individual ‘LARGE’ Contacts Individual large contact area corresponds to very lowvalues of Young’s modulus (i.e. about 50 MPa). Such low values indicate that the pad material is soft andthe summit underlying the individual large contact area isnot fully supported (as seen in the SEM image) and likelyconsists of fractured pore walls and loosely attached paddebris. As the soft material is very compliant, the large contactarea implies much lower contact pressures compared tosmall contact area induced by fully supported padasperities. As a result, large ,and low-pressure, individual contactsbecome easily lubricated and contribute less to removalrate than small, high-pressure contacts.30

An Experimental Case Study31

Araca APD – 800 Polisher & Tribometer32

Experimental Conditions PadIC1000 K-groove pad withSuba IV sub-pad Slurry7 volume parts of HitachiChemical HS 2H635-12 slurry 7 volume parts of DI water 6 volume parts of ultra pure30% hydrogen peroxideFlow rate: 300 ml/min Wafer Pad Conditioning3M A2810 disc rotatingat 95 RPM and sweepingat 10 times/minIn-situ pad conditioningat 25.8 and 44.5 N Wafer PolishingPolishing pressure: 1.5PSISliding velocity: 1.0 m/swafer polishing time: 1.5minute200-mm blanket copperwafers33

Removal Rate, Total Contact Areaand Pad Surface AbruptnessRR 25.8 N RR 44.5 NCA 25.8 N CA 44.5 Nλ 25.8 N λ 44.5 N34

Pad Surface Abruptness (Lambda) When asperity summitshave exponentiallydistributed heights, thenthe right hand tail of thePDF will be linear on a logscale. The pad abruptness factor(λ) is the distance overwhich the tail drops by afactor of e. A pad with larger λ meansa rougher pad contactingsurface. This shouldresult in higher removalrates. In our case (previousslide), RR trends areconsistent with Lambdatrends.1/eλ35

Measured Total Large Contact Area(µm2)Measured Total Large Contact Area Based onContacting Asperity Threshold Size1000Mostly due to collapsed walls and looselyattached pad debris25.8 N44.5 N100Collapsed walls and loosely attached pad debris seem tobe removed by the high conditioning force!10 3 4 5 6 7 8 9 10 11 12 13Threshold Size of Large Individual Contacting Asperity (µm2)Regardless of threshold size selected to define ‘LARGE’ individual contactingasperities, there is a significant difference in the large contact area measuredbetween 25.8 and 44.5 N conditioning forces, but no change in RR.36

Measured Total Small Contact Area(µm2)Measured Total Small Contact Area Based onContacting Asperity Threshold Size807025.8 N6044.5 d Size of Small Individual Contacting Asperity (µm2)Pads conditioned at 25.8 and 44.5 N, share similar small contact area valueswhen 0 – 8 or 0 – 9 square micron ranges are selected to define the size of‘SMALL’ contacting asperities.37

Summary Individual large contact area seems to be induced by collapsedpore walls and loosely attached pad debris. Simulations indicate that individual large contact areas correspondto very low values of Young’s modulus. A case study was presented to illustrate the role of the individuallarge contact area in copper CMP. Results confirmed that individual large contact area had minimalcontribution to removal rate and indicated that the removal rate wasmainly caused by small contact area for copper CMP. The case study showed that (on an IC-1000 pad) individual contactareas smaller than 9 square micron contributed to removal rate. Threshold values that may define ‘SMALL’ and ‘LARGE’ individualcontact areas for different pads and processes need to be furtherinvestigated.38

Acknowledgement SRC Engineering Research Center for EnvironmentallyBenign Semiconductor Manufacturing39

Analysis of Large Pad Surface Contact Area in Copper CMP X. Liao 1, Y. Zhuang 1,2, L. Borucki 2, Y. Sampurno 1,2 and A. Philipossian 1,2 1 University of Arizona, Tucson AZ USA 2 Araca Incorporated, Tucson AZ USA