Transcription

Industrial Example: ANOVA Semiconductor Processing example: PhotolithographyRegistration Improvements: PROBLEM to SOLVE: We are getting a 20% rework rate and10% of contact layer wafers are getting rejected for failure to align.We have an idea that we can improve this layer’s alignment signal strength bychanging the photoresist thickness. The alignment syestem siganl strength isbased on thin films reflectivity.DATA and background:* Prolith simulations* Reflectivity at 632nm* CD swing curves for various oxide thickness* Actual data and Statistical analysis1/3/03ANOVA/IndustrialExample 5 ANOVA1

Industrial Example: ANOVA Data and backgroundALIGNMENT SIGNAL STRENGTH: CT-P1 alignment How the AMSL align632 nm illumination system “sees” the ASML marker PROLITH ModelingReflectivity. Film thickness varaiations1/3/03CT Oxide thickness over marker at CTTiSi2SiCr OvercoatTOTALTiSi2 lo SiCr lo40501350TiSi2 lo SiCr high40501650TiSi2 target SiCr lo45001350TiSi2 target SiCr high45001650TiSi2 target SiCr Target45001500TiSi2 lo SiCr Target40501500TiSi2 high SiCr Target49501500TiSi2 high SiCr lo49501350TiSi2 high SiCr OTORESISTLowTargetHighPhoto resist IX405 0 ASML marker depth114011701200Photo resist IX405 1200 ASML marker depth126012901320ANOVA/IndustrialExample 5 ANOVA2

Industrial Example: ANOVA Data and backgroundALIGNMENT SIGNAL STRENGTH: CT-P1 alignment How the AMSL align632 nm illumination system “sees” the ASML aligmnent marker PROLITHModeling Reflectivity ASML marker zero depth.CT ASML Alignment Signal AnalysisPhotoresist Reflectivity at 632 nm0.35CT target Photoresist Thickness 1170nmCT target Photoresist Thickness 1140nmCT target Photoresist Thickness 1200nmCT no 0620640660680700720740CT Total Oxide film Thickness nm1/3/03ANOVA/IndustrialExample 5 ANOVA3

Industrial Example: ANOVA Data and backgroundALIGNMENT SIGNAL STRENGTH: CT-P1 alignment How the AMSL align 632 nmillumination system “sees” the ASML marker PROLITH Modeling Reflectivity delta top ofmarker to bottom of marker 1200A depth. Desire maximum delta thinner photoresist!CT ASML Alignment Signal AnalysisPhotoresist 11400A delta ReflectivityPhotoresist 11700A delta ReflectivityPhotoresist 12000A delta ReflectivityPhotoresist Reflectivity delta AML markertop- bottom at 632 720740-0.05-0.1-0.15oxide over P2 thickness range-0.2CT Total Oxide film Thickness nm1/3/03ANOVA/IndustrialExample 5 ANOVA4

Industrial Example: ANOVA Data and backgroundALIGNMENT SIGNAL STRENGTH: CT-P1 alignment How the AMSL align632 nm illumination system “sees” the ASML marker PROLITH ModelingReflectivity ASML marker 0 and 1200A depths Minimum photoresist thicknessCT ASML Alignment Signal Analysis0.2CT target Photoresist Thickness 1140nmPhotoresist Reflectivity at 632 nm0.18CT target Photoresist Thickness 1260nm (min PR thickness marker 80600620640660680700720740CT Total Oxide film Thickness nm1/3/03ANOVA/IndustrialExample 5 ANOVA5

Industrial Example: ANOVA Data and backgroundALIGNMENT SIGNAL STRENGTH: CT-P1 alignment How the AMSL align632 nm illumination system “sees” the ASML marker PROLITH ModelingReflectivity ASML marker 0 and 1200A depths Target photoresist thicknessCT ASML Alignment Signal Analysis0.2CT target Photoresist Thickness 1170nmPhotoresist Reflectivity at 632 nm0.18CT target Photoresist Thickness 1290nm (target PR thickness marker 80600620640660680700720740CT Total Oxide film Thickness nm1/3/03ANOVA/IndustrialExample 5 ANOVA6

Industrial Example: ANOVA Data and backgroundALIGNMENT SIGNAL STRENGTH: CT-P1 alignment How the AMSL align632 nm illumination system “sees” the ASML marker PROLITH ModelingReflectivity ASML marker 0 and 1200A depths maximum photoresist thicknessCT ASML Alignment Signal Analysis0.2Photoresist Reflectivity at 632 nm0.18CT target Photoresist Thickness 1200nmCT target Photoresist Thickness 1320nm (Max PR thickness marker 80600620640660680700720740CT Total Oxide film Thickness nm1/3/03ANOVA/IndustrialExample 5 ANOVA7

Industrial Example: ANOVA Data and backgroundALIGNMENT SIGNAL STRENGTH: CT-P1 alignment signal improves withthinner or thicker photoresist film. We’d prefer to go thicker: 12300A but fornominal oxide thickness the swing curve is not optimal ( desire a CD minimumthickness): hence we need to run an actual swing curve for all three contact types:CT over DI, P1, and P2/P1.Prolith: CT layer Swing: Contact CD Various oxideThicknesses8805490A6090A6690A6230A860Contact CD 606406206001120 1130 1140 1150 1160 1170 1180 1190 1200 1210 1220 1230 1240 1250 1260 1270 1280IX405 Thickness nm1/3/03ANOVA/IndustrialExample 5 ANOVA8

Industrial Example: ANOVA Actual data Experimental data Photoresist impact on alignment and registration Run experiment: look at the effect of five different photoresist thicknesses on alignment varaibles: Independent variables or factors: Photoresist thickness: Levels 5 Response variables: Alignment signal strength %; X axis registration error amd Y axiserrorASML Signal Strength AL marker %CT-P1 ASML Alignment Signal Strength ASML 13 Vs CT PhotoresistThickness WC19Y Lot 069242220M%Poly. (M%)18y 3E-05x 2 - 0.7993x 4663.1R2 220012400IX405 Photoresist Thickness A1/3/03ANOVA/IndustrialExample 5 ANOVA9

Industrial Example: ANOVA Actual data Experimental data Photoresist impact on alignment and registration Maximum registration errorCT-P1 Registration Error Vs CT PhotoresistThickness WC19Y Lot 069CT-P1 max Reg error um0.2100.1900.1700.150Max Error Poly. (Max Error )0.1300.110y -1E-07x 2 0.0027x - 15.756R2 01220012400IX405 Photoresist Thickness A1/3/03ANOVA/IndustrialExample 5 ANOVA10

Industrial Example: ANOVA Actual data Experimental data Photoresist impact on alignment and registration X axis registration errorCT-P1 X Reg Error um Vs Resist Thickness AMSL Align to AL markerRegistration Error microns0.150.10.050-0.05-0.1-0.1511000X Reg Error umPoly. (X Reg Error um )11200114001160011800120001220012400IX405 Thickness A1/3/03ANOVA/IndustrialExample 5 ANOVA11

Industrial Example: ANOVA Actual data Experimental data Photoresist impact on alignment and registration Y axis registration errorCT-P1 Y Reg Error um Vs Resist Thickness AMSL Align to AL marker0.25Registration Error microns0.2Y Reg Error umPoly. (Y Reg Error um 00120001220012400IX405 Thickness A1/3/03ANOVA/IndustrialExample 5 ANOVA12

Industrial Example: ANOVA Actual data* Experimental data Photoresist impact on alignment and registrationNORMSINVGST40 CT-P1 Registration Error Vs IX405 4.1%0.569.2%050%-0.530.9%X Reg Error um 11700A-115.9%Y Reg Error um 11700A-1.56.7%X Reg Error um 12000A-22.3%Y Reg Error um 12000A-2.50.6%X Reg Error um 12300A0.1%Y Reg Error um 12300A-3-0.15-0.1-0.0500.050.10.15X Reg Error um 11200AY Reg Error um 11200AX Reg Error um 11400AY Reg Error um 11400A0.2Registration Error microns1/3/03ANOVA/IndustrialExample 5 ANOVA13

Industrial Example: ANOVA Descriptive statistics Experimental data Photoresist impact on alignment and registration Descriptive statisticsStatisticMean error micronsMean Error nmStandard ErrorMedianModeStandard DeviationStd Deviation nm3 sigmamean 3 SigmaSpecSample t1/3/03X RegError um11200AY RegError um11200AX RegError um11400AY RegError um11400AX RegY RegError um Error um11700A11700AX RegError um12000AY RegError um12000A0.007682 0.037253 0.007973 0.018339 0.013398 0.0647180.017130.067392 0.014404 0.0595788378181365170.003827 0.004213 0.005065 0.004991 0.003127 0.004863 0570.01030.02570.00330.01210.01580.05520.042784 0.047106 0.043866 0.043224 0.034965 0.054372 46X RegError um12300A14Y RegError um12300A600.004112 0.0037920.01250.05820.0080.00950.045978 390.00183 0.002219 0.001924 0.001868 0.001223 0.002956 0.001571 0.002111 0.002114 0.0017980.084449 -0.353122 0.461491 0.480213 -0.3367 0.247761 0.17415 -0.399276 0.345385 -0.49423-0.488511 -0.229216 -0.493331 -0.678253 0.14626 -0.20979 -0.134692 -0.258289 -0.377258 ample 5 ANOVA14

Industrial Example: ANOVA Single Factor X error* Experimental data Analysis ANOVA Photoresist impact on alignment andregistration run using EXCEL Single Factor ANOVAHi: µ1 µ2 µ3 µ4 µ5Ho: µ1 µ2 µ3 µ4 µ5F cal Fcrit and P 0.34 so we accept Ho. Means if we reject Ho we have a 34% chance of being wrong!Anova: Single FactorSUMMARYGroupsX Reg Error um 11200AX Reg Error um 11400AX Reg Error um 11700AX Reg Error um 12000AX Reg Error um 12300ACount12575125125125ANOVASource of VariationBetween GroupsW ithin age Variance0.9602 0.0076820.001830.598 0.007973 0.0019241.6748 0.013398 0.0012232.14120.01713 0.0015711.8005 0.014404 347089F crit2.387566574ANOVA/IndustrialExample 5 ANOVA15

Industrial Example: ANOVA Single Factor Y error* Experimental data Analysis ANOVA Photoresist impact on alignment andregistration using EXCEL Single Factor ANOVAHi: µ1 µ2 µ3 µ4 µ5Ho: µ1 µ2 µ3 µ4 µ5F cal Fcrit and P 0.00 so we reject Ho. Means if we reject Ho we have a 0%chance of being wrong! We know at least one mean is different!Anova: Single Factor1/3/03SUMMARYGroupsY Reg Error um 11200AY Reg Error um 11400AY Reg Error um 11700AY Reg Error um 12000AY Reg Error um 12300ACount12575125125125ANOVASource of VariationBetween GroupsW ithin GroupsSS0.169161.264698Total1.433858SumAverage Variance4.6566 0.037253 0.0022191.3754 0.018339 0.0018688.0897 0.064718 0.0029568.424 0.067392 0.0021117.4472 0.059578 14F crit2.387566574ANOVA/IndustrialExample 5 ANOVA16

Industrial Example: ANOVA Single Factor What did we learn? 1, Endure you randomize experiment 2. Plot the data different ways graphically look at it before running statisticaanalysis. 3, Test data for normalcy 4. When a difference is found Look at how large the difference is with the Pvalue. 5. Remember this ANOVA just tells us : Yes at least one of the means isdifferent, but not which one and not how much. EXCEL Single Factor ANOVAHi: µ1 µ2 µ3 µ4 µ5Ho: µ1 µ2 µ3 µ4 µ5If F cal Fcrit we reject Ho. Means if we reject Ho we have a P% chance ofbeing wrong! We know at least one mean is different!1/3/03ANOVA/IndustrialExample 5 ANOVA17

1/3/03 ANOVA/Industrial Example_5_ANOVA 1 Industrial Example: ANOVA Semiconductor Processing example: Photolithography Registration Improvements: PROBLEM to SOLVE: We