sample_presenta

sample_presenta - MIT OpenCourseWare http://ocw.mit.edu...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: MIT OpenCourseWare http://ocw.mit.edu _____________ Spring 2008 2.830J / 6.780J / ESD.63J Control of Manufacturing Processes (SMA 6303) For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms . ___________________ Manufacturing 2.830/6.780J Control of Manufacturing Processes “An Industrial Example of Oxide Etch An Process Control and Optimization” Process Spring 2007 Jing Yao Kai Meng Yi Qian Manufacturing Agenda • Plasma Etch Process physics • Industrial Practices – SPC Practice – A Process Improvement Experiment • Proposed DOE and RSM methods • Process control improvements and Process recommendations recommendations 2/1/05 2 Manufacturing Layered Wafer Manufacturing Process • 3 basic operations: basic – Film Deposition – Photolithography – Etch • This cycle is repeated This to build up various layers in the devices. layers Image removed due to copyright restrictions. Please see http://dot.che.gatech.edu/henderson/Introductions/Image55.gif 2/1/05 3 Manufacturing Types of Etching – Dry etch / Plasma etch (Anisotropic) • Etch techniques – Wet etch (Isotropic) Αnisotropy is critical in submicron feature fabrication! 2/1/05 1 Manufacturing Plasma Etching Steps • Plasma etching uses RF power to drive material removal by chemical reaction • Steps: – Formation of active gas species, e.g. CF4+ e- CF3++ F+ 2eCF – Transport of the active species to the wafer surface – Reaction at the surface SiO2 + 4F SiF4 + O2 SiF – Pump away volatile products 2/1/05 Gas in To pump Absorption Desorbtion Reaction RF Power 5 Manufacturing Physical vs Chemical Etching Physical vs Mechanism Etch Rate Selectivity Bombardmentinduced damage Anisotropy Physical Method Chemical Method Ion Bombardment Chemical Reaction Low Low High High High High Low Low • Industry often uses hybrid technique: physical Industry method to enhance chemical etching method • This gives anisotropic etch profile, reasonably good This selectivity, and moderate bombardment-iinduced nduced selectivity, damage. damage. 2/1/05 6 Manufacturing Plasma Etch Parameters • Gas chemistry – Fluorocarbon gases (C4F6, CF4, C4F8,etc) Atomic F is active etchant for SiO2 SiO2 + 4F 4F SiF4 + O2 SiF Carbon reacts with oxygen to form Carbon provides passivation llayer on Si passivation ayer provides selectivity – O2: Under certain level, O2 scavenge C in scavenge Fluorocarbon, results in higher F Higher etch rate concentration Higher – Ar: Ar+ ion beam enhances chemical Ar reaction reaction 2/1/05 7 Manufacturing Plasma Etch Parameters • Pressure – Low pressure reduces ion-neutral neutral collision on sidewalls (lateral etch), enhances anisotropic etching enhances • Bias Power – Increase bias power enhances physical Increase bombardment of ions bombardment • Etch Time • Temperature 2/1/05 8 Manufacturing Critical Issues • • • • • Anisotropy Selectivity Microscopic Uniformity Etch Depth Critical Dimension (CD) Image removed due to copyright restrictions. Please see http://www.memsexchange.org/catalog/P3431/file/f38826bf4266f95d6e054553/thumbnail?600 2/1/05 9 Manufacturing Background • Industry Practices in a DRAM wafer Industry fabrication plant in Singapore fabrication • Current Technology: Current – 95nm 1GB DRAM on 200mm wafers – 78nm 1GB DRAM on 300mm wafers • Information source – Interview with process engineer – Scaled data based on experiments data Scaled (actual data unknown) (actual 2/1/05 10 Manufacturing Focused Output • Etch Depth – Measuring Method • Test wafer ONLY! – Over-etch on test wafer – Cost • 5 sites measurement 100% SiO2 SiO2 – Percentage over-etch on test wafer • 20%-60% over-etch on test wafer • Selectivity 150% Nitride/Silicon SiO2 Test Wafer Production Wafer • Critical Dimension – Measuring Method • Test or production wafer • 5 sites measurement Image removed due to copyright restrictions. Please see http://www.memsexchange.org/catalog/P3431/file/f38826bf4266f95d6e054553/thumbnail?600 2/1/05 11 Manufacturing SPC Practice • SPC analysis tools are installed in all SPC production machines production – X-bar chart and R chart • Different test methods for different outputs – Etch Depth • • • Insert test wafer into production lots Infrequent: ~200 hours Increase frequency when special attention needed – Critical Dimension • Test 1 wafer per lot (25 wafers) • 5 sites average 12 2/1/05 Manufacturing SPC Practice • Rules: similar to Western Electrical Rules: Handbook rules Handbook • UCL/LCL are set by process engineer – – – – Based on USL/LSL UCL/LCL are little bit tighter than USL/LSL Tighten UCL/LCL based on experience UCL/LCL are not based on standard deviation! • Process pass SPC most of the time • Stop a machine when a measurement is Stop outside UCL/LCL, other rules mostly ignored ignored • Slow response 2/1/05 13 Manufacturing SPC Improvement • Set UCL/LCL based on sample Set standard deviation standard • Use more effective control chart, like Use CUSUM or EWMA chart, to improve response time response • Use multivariate process control 2/1/05 14 Manufacturing A Process Improvement Process Experiment Experiment • Problem – – – Under-etch Discovered by quality assurance from finished Discovered products products Process improvement is necessary because no Process issues found on the machine issues Focus on two inputs (C4F6 Flow Rate, Bias Power) Vary inputs one step away from current value Test with all inputs combinations Test Change third input (Time) Repeat 1 to 3 Find the best result 15 • Approach 1. 2. 3. 4. 4. 5. 6. 2/1/05 Manufacturing A Process Improvement Process Experiment Experiment • • • 1 wafer, no replicates 5 sites average Goal: – – CD: 100 ± 5 nm CD: Etch Depth: 1.4 um with 60%~70% over etch on test wafer [2.25um, 2.4 um] C4F6 (sccm) C4F6 (sccm) Etch Depth (um) 15.5 1.56 1.91 2.41 Bias Power (W) 1300 1400 1500 14.5 1.63 2.00 2.50 15 1.60 1.95 2.37 15.5 1.50 1.87 2.28 Etch Depth (um) 14.5 Bias Power (W) 1300 1400 1500 1.72 2.08 2.56 15 1.68 2.01 2.45 200 sec C4F6 (sccm) CD (nm) 14.5 Bias Power (W) 1300 1400 1500 100 110 118 15 95 103 110 15.5 88 96 104 Bias Power (W) 1300 1400 1500 CD (nm) 190 sec C4F6 (sccm) 14.5 98 106 114 15 93 100 106 15.5 85 94 100 2/1/05 sccm : Standard Cubic Centimeters per Minute 16 Manufacturing A Process Improvement Process Experiment Experiment • A combination of DOE and OFAT – Rely on theoretical study and experience • Find an optimal based on tested input Find combinations combinations • No Response Surface analysis No • No replicates or center points – Hard to prove model accuracy • No variance study • Confidence Level unknown! 2/1/05 17 Manufacturing Experimental Design • Bias Power and C4F6 – Central composite design Central – 3 levels • Etching Time – 2 levels Factor X1-Bias Power X2-C4F6 X3-Etching Time Actual test levels (coded test level) (-1) 1300 14.5 190 (0) 1400 15.0 (1) 1500 15.5 200 W sccm sec 2/1/05 18 Manufacturing Run Data Bias Power -1 0 1 -1 0 1 -1 0 1 -1 0 1 -1 0 1 -1 0 1 C4F6 -1 -1 -1 0 0 0 1 1 1 -1 -1 -1 0 0 0 1 1 1 Time 1 1 1 1 1 1 1 1 1 -1 -1 -1 -1 -1 -1 -1 -1 -1 Etch Depth (um) 1.72 2.08 2.56 1.68 2.01 2.45 1.56 1.91 2.41 1.63 2.00 2.50 1.60 1.95 2.37 1.50 1.87 2.28 Critical Dimension (nm) 100 110 118 95 103 110 88 96 104 98 106 114 93 100 106 85 94 100 -1 -1 Bias Power 0 1 -1 1 0 C4F6 Time 0 1 Trial 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Note: each run data is the mean of 5 sites average on 1 wafer 2/1/05 19 Manufacturing Response Models • Second order polynomial models – models built using coded variables – no transformations of output variables no attempted attempted Y = b0 + ∑ bi X i + i =1 3 j = i +1 i =1 ∑ ∑b X X + ∑b X ij i j i =1 ii 3 3 3 2 i 2/1/05 20 Manufacturing Model Evaluation • RSM fitting – ANOVA performed – Each output model claimed significant at Each >99.8% confidence level or higher >99.8% • Regression coefficients shown for Regression significant terms significant 2/1/05 21 Manufacturing Etch Depth • Response Surface model ED = 1.970 + 0.407x1 -0.080x2 + 0.038x3 + 0.052x12 • Residual Resi dual Pl ot s f or Et ch Dept h No rmal Pro b ab ilit y Plo t o f t h e Resid u als 99 90 Residual Per cent 50 10 1 -0.050 -0.025 0.000 Residual 0.025 0.050 0.030 0.015 0.000 -0.015 -0.030 1.50 1.75 2.00 2.25 Fit t ed Value 2.50 Resid u als Versu s t h e Fit t ed Valu es Hist o g ram o f t h e Resid u als 4 Fr equency 3 2 1 0 -0.03 -0.02 -0.01 0.00 0.01 Residual 0.02 0.03 Resid u als Versu s t h e Ord er o f t h e Dat a 0.030 0.015 Residual 0.000 -0.015 -0.030 2 4 6 8 10 12 14 Obser vat ion Or der 16 18 2/1/05 22 Manufacturing Etch Depth – Contour Plot Etch Cont our Pl ot of Et ch Dept h vs Ti me, Bi as Pow er Etch Depth < 1.6 1.6 - 1.8 1.8 - 2.0 2.0 - 2.2 2.2 - 2.4 > 2.4 Cont our Pl ot of Et ch Dept h vs C4 F6 , Bi as Pow er 1.0 1.0 0.5 0.5 Etch Depth < 1.6 1.6 - 1.8 1.8 - 2.0 2.0 - 2.2 2.2 - 2.4 > 2.4 Hold Values C4F6 0 C4F6 0.0 Time Hold Values Time 0 0.0 -0.5 -0.5 -1.0 -1.0 -0.5 0.0 Bias Pow er 0.5 1.0 -1.0 -1.0 -0.5 0.0 Bias Pow er 0.5 1.0 Cont our Pl ot of Et ch Dept h vs Ti me, C4 F6 1.0 Etch Depth < 1.90 1.90 - 1.95 1.95 - 2.00 2.00 - 2.05 > 2.05 Hold Values Bias Power 0 0.5 0.0 • Etch Depth most sensitive to Bias Power • Bias Power , or Time , or C4F6 Etch Depth Time -0.5 -1.0 -1.0 -0.5 0.0 C4F6 0.5 1.0 2/1/05 23 Manufacturing Critical Dimension • Response Surface model CD = 101.111 + 7.750x1 -6.583x2 + 1.556x3 • Residual Resi dual Pl ot s f or CD No rmal Pro b ab ilit y Plo t o f t h e Resid u als 99 90 Residual Per cent 50 10 1 -2 -1 0 Residual 1 2 1 0 -1 90 100 Fit t ed Value 110 120 Resid u als Versu s t h e Fit t ed Valu es Hist o g ram o f t h e Resid u als 6.0 Fr equency 4.5 3.0 1.5 0.0 -1.5 -1.0 -0.5 0.0 Residual 0.5 1.0 1 Residual 0 -1 Resid u als Versu s t h e Ord er o f t h e Dat a 2 4 6 8 10 12 14 Obser v at ion Or der 16 18 2/1/05 24 Manufacturing Critical Dimension – Contour Plot Critical Cont our Pl ot of CD vs C4 F6 , Bi as Pow er Cont our Pl ot of CD vs Ti me, Bi as Pow er CD < 90 95 100 105 110 > 1.0 1.0 90 95 100 105 110 115 115 0.5 0.5 CD < 95 100 105 > 95 100 105 110 110 Hold Values C4F6 0 C4F6 Time 0.0 Hold Values Time 0 0.0 -0.5 -0.5 -1.0 -1.0 -0.5 Cont our Pl ot of CD v s Ti me, C4 F6 1.0 CD < 95.0 97.5 100.0 102.5 105.0 > 95.0 97.5 100.0 102.5 105.0 107.5 107.5 0.0 Bias Pow er 0.5 1.0 -1.0 -1.0 -0.5 0.0 Bias Pow er 0.5 1.0 0.5 Time 0.0 Hold Values Bias Power 0 -0.5 • CD most sensitive to Bias Power & C4F6 • Bias Power , or Time , or C4F6 CD -1.0 -1.0 -0.5 0.0 C4F6 0.5 1.0 2/1/05 25 Manufacturing Process Optimization • Optimization criteria for Oxide etch and the Optimization best values attainable within the resulting optimized factor space optimized Factor Etch Depth Critical Dimention Optimization Criteria Best Values 2.25 μ m ≤ CD ≤ 2.40 μ m 100 ± 5nm 2.25 μ m 100 nm • Optimal Input X1-Bias Power Model Actual 1487 W 1500 W X2-C4F6 15.48 sccm 15.5 sccm X3-Etching Time 190 sec 190 sec 2/1/05 26 Manufacturing 23 Full Factorial Design Full 1 • Only consider linear relationships • Drop other 10 test points (possible test points for lack-of-fit) Trial 1 2 3 4 5 6 7 8 Bias Power -1 1 -1 1 -1 1 -1 1 C4F6 -1 -1 1 1 -1 -1 1 1 Time 1 1 1 1 -1 -1 -1 -1 Etch Depth (um) 1.72 2.56 1.56 2.41 1.63 2.50 1.50 2.28 Time 0 -1 -1 0 Bias Power 1 -1 1 0 C4F6 Critical Dimension (nm) 100 118 88 104 98 114 85 100 27 2/1/05 Manufacturing Etch Depth • Predicted Value (p<0.01) ED = 2.020 + 0.418x1 -0.083x2 + 0.043x3 • Residual Residual Plots for Etch Depth Normal Probability Plot of the Residuals 99 90 Residual Percent 50 10 1 -0.050 -0.025 0.000 Residual 0.025 0.050 0.00 -0.02 -0.04 0.02 Residuals Versus the Fitted Values 1.50 1.75 2.00 2.25 Fitted Value 2.50 Histogram of the Residuals 2.0 Frequency Residual 1.5 1.0 0.5 0.0 -0.03 -0.02 -0.01 0.00 Residual 0.01 0.02 Residuals Versus the Order of the Data 0.02 0.00 -0.02 -0.04 1 2 3 4 5 6 Observation Order 7 8 2/1/05 28 Manufacturing Critical Dimension • Predicted Value (p<0.01) CD = 100.875 + 8.125 x1 -6.625x2 + 1.625x3 • Residual Residual Plots for CD Normal Probability Plot of the Residuals 99 90 Residual Percent 50 10 1 -1 0 Residual 1 1.0 0.5 0.0 -0.5 -1.0 80 90 100 Fitted Value 110 120 Residuals Versus the Fitted Values Histogram of the Residuals 2.0 Frequency Residual 1.5 1.0 0.5 0.0 -1.00 -0.75 -0.50 -0.25 0.00 0.25 0.50 0.75 Residuals Versus the Order of the Data 1.0 0.5 0.0 -0.5 -1.0 Residual 1 2 3 4 5 6 Observation Order 7 8 2/1/05 29 Manufacturing DOE Improvement • Adding replicates at center points – Use to assess pure error (‘noise’) as percentage as of the response of – Assess lack of fit • Use Factorial Design – – – Current practice 18 trails 23 with 4 center points 12 trails with 12 33-1III with 6 center points 15 trails with 15 • Analyze Variation – consider variation at the desired value • Randomize run order Randomize – Esp. in replicates to minimize the trend 2/1/05 30 Manufacturing Process Control Recommendations • SPC Analysis – Use more effective control chart, like Use CUSUM or EWMA chart CUSUM – Use multivariate process control • DOE and RSM optimization – Adding replicates at center points – Use Factorial Design – Analyze Variation – Randomize run order 2/1/05 31 Manufacturing Thank You! 2/1/05 32 ...
View Full Document

This note was uploaded on 09/24/2010 for the course MECHE 2.830J taught by Professor Davidhardt during the Spring '08 term at MIT.

Ask a homework question - tutors are online