Rapid Cure of Polyimide Coatings for Packaging Applications using Variable Frequency Microwave Irradiation Mel Zussman (HD MicroSystems), Bob Hubbard and Keith Hicks (Lambda Technologies) Are polyimides more like pizza or popcorn?
VFM has been widely studied for cure of polyimides and PBs 2 VFM cure method has been demonstrated to have no harmful effects on any electronic devices or circuitry. Variable Frequency irradiation produces spatial uniformity (no hot spots ) and eliminates arcing associated with single frequency microwave heating VFM has been shown to cure polyimides and PB s at much lower temperatures than oven or furnace cures Lower thermal budget reduces yield loss in devices Lower stress from CTE mismatch
Does rapid microwave cure give the same results as a slow oven cure? 3 Rapid microwave cooking Microwave popcorn tastes great Microwave pizza is hot, but the outer crust is tough and the middle gets soggy How about rapid cure of polyimide? Does rapid VFM cure deliver the same key properties as oven/furnace cure? Can we define a VFM cure that reproduces the results of a specific oven cure? If the results are equivalent, are there cost advantages to rapid VFM cure?
Phase 1: Compare key properties of PI s with Rapid VFM or Standard ven Cure Selected two polyimides for initial testing Both are used extensively for stress buffer and RDL PI-5878G non-photosensitive polyimide Coat & soft-bake HD4004 photosensitive polyimide Coat, soft-bake & blanket expose Films coated over cured PI-2611 for easy release after cure VFM cure in Microcure 3100-2000 single wafer ven cure in Koyo CLH-21CD-S ven 60 min hold @ 350 C 5 hr cycle time 4
Phase 1: A designed experiment looks at four process variables in VFM cure 5 All process times are very far from standard oven cure 2 level kept below 50 ppm Total Cycle Time (min) Time @ 350 C Ramp to 350 C Time @ 200 C Ramp to 200 C 16.6 27.5 min 10-15 min 0.25 0.63 min 5-10 min 0.25-0.71 min Equivalent process condition 0 0.2 0.4 0.6 0.8 1 1.2 Ratio VFM/ven Cure
Phase 1: VFM cure results similar to oven cure Both PI5878G and HD-4000 appear well cured by rapid VFM process Properties show small dependence on the VFM process variables tested Spread in elongation results due to experimental method Ramp rate to 200 C has least effect VFM results over process range tested Tg ( C) % Imidization Elong. at Break (%) Tens. Strength (MPa) Modulus (GPa) 358-365 279 302 91 99 84 104 8 20 204 274 154 175 2.5 2.9 3.9 4.6 Error bars show range of results Equivalent properties 6 PI5878G avg. HD4004 avg. 0 0.2 0.4 0.6 0.8 1 1.2 Ratio VFM/ven Cure
Phase 2: Focus on HD4110 & find VFM cure that replicates oven cure at 350 C 7 Factors: three VFM cure process variables Hold @ 200 C, Ramp to 350 C and Hold at 350 C Test more properties Tensile modulus, strength, elongation photo-imaged tensile strips Film stress wafer bow before and after cure Tg DSC Decomposition temperature - TGA Adhesion stud pull testing Via shape: height, slope and crowning SEM, Tencor on 30µm channels Chemical resistance (CR) % dimension change after exposure CR1 - NMP, RT 30 min CR2-10% H2S4, 3 hr, RT CR3 - NH4H/H22/H2 1:1:5 80 C, 30 min CR4 - TMAH/H2/DMS 1/3/96 80 C, 30 min
Phase 2: VFM cure at 350 C meets or exceeds oven cure Range of results for VFM cure overlaps oven cure for most properties Responses are significantly different for: Decomposition temperature Thickness Crowning Differences are consistent with higher cure by VFM Modulus (GPa) Elongation (%) Strength (Mpa) Stress (Mpa) Td - 1% ( C) Line Height (µm) Crown Height (µm) Adhesion (lbs) CR1*100+100 CR2*100+100 CR3*100+100 CR4*100+100 ven Avg. VFM Avg. 1.0 10.0 100.0 1000.0 Note: Error bars for ven cure results show standard deviation in data. Error bars for VFM cure show range of responses 8
Crowning and film thickness are sensitive to cure 9 VFM cure 7 min 350 C ven cure - 60 min 350 C Line Thickness (µm) Crown Height (µm) Via Slope ( ) 10.2 3.0 128 12.2 1.8 123 ven cured sample was re-baked at 400 C: Line thickness reduced to 10.7 µm Crown height increased to 2.4 µm Conclude: Difference in line thickness & crown height reflects higher cure level in VFM cured samples
Cure of HD4110 involves cyclization and photopackage burnout 10 HD4110 process chemistry Photo-exposure yields crosslinked network Imidization forms linear polyimide, burn-out removes crosslinker Ar 2 N H HD-4110 R Ar 1 R H N Ar 2 Ar 2 Crosslink R H N Ar 1 N H R Ar 2 Ar 1, Ar 2 are aromatic nuclei R is an acrylate ester UV Expose Negative Tone Image R N H R N H Ar 1 Ar 1 R R H N H N Interpenetrating Network HR Ar 2 Ar 1 Heat H Cyclize R R H Ar 2 Ar 1 H R Polyimide Ar 2 Ar 1 Heat "Burn ut" Ar 2 Ar 1
350 C cure: oven cure leaves residual IPN, VFM cure does not 11 0.5 Residual polyacrylate indicated by DMA TGA Weight loss from polyacrylate burn-out 741280_ven #3 HD4110 ven Cure 741281_VFM #10 HD4110 VFM Cure tan_delta ( ) [ ] 0.4 0.3 0.2 DMA shows Tg from polyacrylate and polyimide DSC Tg of polyacrylate 0.1 0.0 0.0 100.0 200.0 300.0 400.0 500.0 Temp [ C] Polyimide Tg is not seen in DSC
12 IR method to monitor cure in HD4110 TGA is a sensitive monitor of cure in HD4110 Use Td-1%, Td-5% or weight loss at 400 C But TGA is a destructive test IR method commonly used for polyimide cure I 1367 /I 1473 monitors imide content I 1720 monitors total C= content, reduced by acrylate burn-out IR Absorbance Ratio 1.2 1.1 1 0.9 0.8 0.7 0.6 0.5 IR (I1367/I1473), norm'd IR (I1367/I1720), norm'd Td - 1% ( C) 150 200 250 300 350 400 Cure Temperature ( C) I 1367 /I 1720 correlates with TGA result, captures both imide formation and burn-out 450 400 350 300 250 200 Decomposition Temp. ( C)
VFM cure of HD4110 at 340 C close to oven cure at 350 C Mechanical and thermal properties very similar Results averaged for 8 VFM cured films, 9 oven cured films Comparable standard deviations found Conclude: VFM process capable of same cure as oven with 94% reduction in cycle time Strength (Mpa) Modulus x 100 (GPa) Elongation x 10 (%) Td - 1% ( C, TGA) Tg ( C, DSC) IR cure (I1360/I1720 x 500) ven: 60 min@350 C, 5.0 hr cycle VFM: 8 min@340 C, 0.3 hr cycle ven VFM 13 0 100 200 300 400 500
VFM shown to lower manufacturing and energy costs 14 Time reduction Cycle time reduced from >4 hrs to <20 minutes Major benefit for process development or mixed lot production Prototype batch VFM (300 mm wafers) shows excellent uniformity Example of energy savings in a non-ic application Energy reduction 160 140 120 100 80 60 40 20 0 Energy transferred efficiently to the wafers no energy wasted heating the oven No energy required to cool the oven after cure KW per Week in Production Example Convection VFM Parts Fixtures Air Chamber
15 Conclusions Does rapid VFM cure deliver the same key properties as oven/furnace cure? Yes, and at a somewhat lower temperature. Can we define a VFM cure that reproduces the results of a specific oven cure? Yes Defined key tests to monitor cure: TGA, IR method (I 1630 /I 1720 ) Temperature is most significant VFM process variable. Ramp rates and hold times appear to have small impact on final properties. Are there cost advantages to rapid VFM cure? $ Significant cycle time reduction is demonstrated Need to demonstrate energy cost reductions for cure of 300 mm wafers
16 Acknowledgments A number of individuals assisted in the experimental aspects of this project. Significant contributions were made by: Chuck gbuawa HD MicroSystems John McFarland HD MicroSystems Tomoko Kawamura HD MicroSystems Robert Pryor DuPont CCAS Project funding and guidance were provided by: John Malloy HD MicroSystems