Real time plasma etch control by means of physical plasma parameters with HERCULES

Transcription

1 Real time plasma etch control by means of physical plasma parameters with HERCULES A. Steinbach 1) S. Bernhard 1) M. Sussiek 4) S. Wurm 2) Ch. Koelbl 3) D. Knobloch 1) Siemens, Dresden Siemens at International Sematech, Austin TX 3) Siemens, Regensburg 4) Universität Hamburg - Harburg 1) 2) -1- Introduction

2 Siemens Microelectronics Center Dresden -2- Introduction

3 Product Portfolio Feature size,35 -,15 µm DRAM 64 Mbit since Mbit since Gbit > 2 Embedded DRAM since 1998 Logic Devices since 1998 Embedded FLASH since 1998 ROS since Introduction

4 64 Mbit SDRAM Technical Data CMOS - technology Smallest feature size.24 µm (.2 µm) Chip size 62 mm² 7 million transistors (13 million devices) Supply voltage 3.3 V Storage capacity 4 DIN A4 pages PC - 1 compatible -4- Introduction

5 256 Mbit SDRAM Technical Data CMOS - technology Smallest feature size.2 µm Chip size 175 mm² 28 million transistors (52 million devices) Supply voltage 3.3 V / 2.5 V PC - 1 compatible Storage capacity 16 DIN A4 pages -5- Introduction

6 SEMICONDUCTOR 3 Joint Venture Siemens / Motorola Development Line 3 mm Part of Module 2 at Siemens Dresden Area: 18 m² Class: 1. Engineers: 15 Operators: 22 Support: 8 Invest: 45 Mio DM Technology:,25 µm -,18 µm CMOS Products: 64 DRAM / 256 DRAM -6- Introduction

7 Contents Introduction and Basics: Motivation Theory and experimental setup Process applications: Basic measurements at Contact etch Long term process monitoring Short term process monitoring Wafer effects Endpoint detection Optimisation of conditioning Maintenance applications: Chamber and tool comparison Hardware failure detection Arcing detection Production application: Detection of recipe errors Summary: Benefits Outlook -7- Introduction and Basics

8 Our way of plasma processing today an effective way? Process parameters Black Box power Process results etch called plasma rate pressure uniformity processing B field selectivity gas and statistical methods in process development - Experience flow - Time consuming Process Monitoring and toolparticles control using many test wafers - Statistical Process Control (SPC) -8- Introduction and Basics

9 The way out: Switch from SPC to APC Statistical process control (SPC) Single wafer control by real time sensors (monitoring) and model based analysis Sample based Advanced Process Control (APC) Continuously -9- Introduction and Basics

10 Measurement Techniques for in-situ real time Plasma Monitoring rf probe rf voltage rf current power Process parameters external power pressure B field gas flow body temp. Ion flux probe j (wall) + Process parameter rf voltage (wafer) rf current bias voltage effective power Chamber parameters surface temp. polymer e.g. gas ad / desorption depending on ion current We begin to measure! Plasma excitation Power balance and potential distribution electron collision rate, electron energy distribution electron density plasma potential bulk power Hercules ion density ion temperature neutral densities neutral temp. excitations Wafer Surface ion energy ion current radiation neutral flows (radicals) surface temp. layer thickness OES k*i(λ ) Process Results external measured etch rate uniformity selectivity particles Interferometry Reflectence spectroscopy layer thickness ne, ν e, PBulk Species in the volume Introduction and Basics

11 Basic HERCULES Model High Frequency Electron Resonance Current Low Pressure Spectroscopy Introduction and Basics

12 Principle and experimental setup rf current rf voltage FFT Algorithm Model SEERS Electron collision rate Electron density Bulk power DC bias voltage - Passive electrical method, no impact on the plasma - Integral measurement Introduction and Basics

13 SEERS provides reciprocally averaged parameters Self Excited Electron Resonance Spectroscopy Introduction and Basics

14 HERCULES Sensor Types Sensor surface = anodized aluminum, similar to chamber wall Introduction and Basics

15 Correlations between plasma parameters and process parameters: CT etch at MxP+ Electron collision rate vs. CF4 flow s -1 ] 9.4 pressure [mtorr] Variation of physical process parameters, e.g. pressure, rf power monotonous response, partly linear correlations collision rate patterned patterned [1 blank blank [1 s ] blank patterned collisison rate collision rate [1 s ] Electron collision rate vs. pressure CF4 flow [sccm] Variation of chemical process parameters, e.g., flow of reactive gases often strong nonlinear effects Process Applications

16 Correlations between electron density and gas flows: CT etch an MxP+ Electron density vs. CHF3 flow Electron density vs. CF4 flow blank 1.8 patterned density [1 cm ] blank patterned -3 density [1 cm ] CF4 flow [sccm] CF4, CHF3, Ar, O2 chemistry: 7 9 CHF3 [sccm] 11 CF4- or CHF3- flow increases higher F- concentration electron density decreases Process Applications

17 Correlations between plasma parameters and etch results: CT etch at MxP+ Contact angle vs. electron density 68 9 CHF3 Variation CF4 Variation increasing gas flow contact angle [ ] etch rate [nm / min] Etch rate vs. electron density 88 CHF3 Variation increasing gas flow electron density [1 / cm ³] CF4- flow increases: CF4 Variation 86 electron density [1 / cm ³] higher concentration of F, CF2 radicals and ions higher etch rate and steeper contact angle CHF3- flow increases: higher F, CF2 - concentration higher etch rate higher CHFx- concentration higher polymerization, less steeper contact angle Process Applications

18 Long term process stability: Tool related effects on CT etch at MxP+ 7-1 collision rate [1 s ] Collision rate vs. rf hours WC1 WC3 WC Wet clean (WC) depending drift effect, hardware reason not found yet: WC2 WC WC1, WC3 chamber drift 7.5 one point - one lot rf hours - WC2, WC4, WC5 stable chamber conditions on varying level Process monitoring of product wafers for 5 wet clean cycles, more than 6 months Process Applications

19 Long term conditioning effect: CT etch at MxP+ Step 1 2 Pr1 BPSG etch collision rate [1 s ] Electron collision rate vs. rf hours 11 Pr1 BPSG Pr2 BPSG Pr2 Nitride Pr3 Oxide 1 9 one point - one wafer rf hours [h] Process monitoring of 3 products covering the period between two wet cleans Pr2 N2 / O2 step BPSG etch Nitride etch Process 2: deconditioning caused by steps 1 and 3 Electron collision rate is very sensitive to etch chemistry Process Applications

20 Short term chamber drift: CT etch at MxP+ Electrical failure counts at contact etch 9.9 bad chamber failure counts Idle time min 45 min 5h collision rate [1 s ] Electron collision rate vs. wafer wafer one point - one wafer wafer one point - one wafer - Collision rate shows dependence on chamber idle time. - Constant chamber conditions after about 4 min! - In some cases a change in electron collision rate corresponds to a change in electrical failure counts Process Applications

21 First wafer effect: Al etch at LAM TCP Al etching in Cl2 - first wafer effect - LAM TCP 96 Product wafer - resist mask on Al (appr. 5%) electron density [1/cm 3 ] 8.19 main etch 7.19 Endpoint limited process: First wafer effect in main etch is connected with higher etch time, lower etch rate. first wafer second 4.19 third wafer process time [s] Process Applications

22 Wafer effects - Monitoring of lot mean values: CT etch at MxP+ Electron density depends on: electron density [1 8/cm3] Electron density vs. rf hours hard ware effect: WC3 - WC4 WC3 WC one point - one lot rf hours [h] wafer effect: product 1 product 2 the same etch process is used on two different products with different open area Process Applications

23 Wafer effects - Monitoring of wafer mean values: Contact etch at MxP+ Electron collision rate vs. wafer in July collision rate [1 s ] collision rate [1 s ] Electron collision rate vs. wafer in August Electron density vs. wafer Electron density vs. wafer 17 / cm³] 17 in July density [1 8 2 wafer wafer density [1 /cm³] Single wafer control of process stability and pre-processes in August wafer wafer In every diagram one point - one wafer Process Applications

24 Wafer effects - Monitoring of time resolved values: Contact etch at MxP+ Electron collision rate vs. etch time collision rate 7-1 [1 s ] Time resolved values show wafer dependent process variations, see wafer 6,7, Electron density vs. etch time 9.5 r 25 Wafe wafer wafer 6,7,8 W 1 afe r 1 15 density etch time [s] 8 [1 / cm³] Electron collision rate and electron density detect different process variations, see wafer r Wafe wafer wafer 6,7, r1 Wafe etch time [s] 1 Process Applications

25 Comparison of HERCULES and AMAT HOT Pack results: Contact etch at MxP+ Hercules AMAT HOT Pack AMAT HOT Pack mean optical emission intensity of contact main etch vs. wafer Electron collision rate and electron density of contact main etch vs. wafer one point one wafer collision rate 12 density intensity [arbitrary units] density [1 / cm³] collision rate [1 s ] one point one wafer CO CN wafer wafer - All measured parameters detect wafer dependent process variations. - No correlation to stable and high yield at this process - Process is robust, measurements are very sensitive Process Applications

26 Endpoint detection: Al etch at LAM TCP Endpoint signal of main etch caused by 15 nm Ti layer below the- Al layer. Al etching with/without barrier (TiN, Ti) - LAM TCP 96 each curve averaged from five testwafers break through (Al2O3) 4.17 collision rate [1/s] 3 step recipe: Break through Main etch Over etch Ti layer nm AlSiCu 17 SiO2 with TiN (1 nm), Ti (15 nm) process time [s] 1 12 Joint project Siemens - ASI - Lam Process Applications

27 Optimization of conditioning: CT etch at MxP+ - Wet clean at 11.7 rf hours - Effect of chamber clean shows up in the electron collision rate - About 1 wafers are necessary to reach stabile chamber conditions again collision rate [1 s ] Electron collision rate vs. rf hours 7 one point - one wafer rf hours [h] 1.5 Non - Productive Wafer reduction one point - one wafer density [1 cm ] - Optimization of conditioning procedures Electron density vs. rf hours rf hours [h] Process Applications

28 Evaluation of shadow rings at MxP Electron collision rate vs. etch time Parameter Collision rate [1 7 s -1] 15 Conditioning with resist wafers Nitride etch rate Wafer temp. Inverse ring temp. El. collision rate Electron density Inverse ratio of the cathode areas time [s] Chamber A, quartz ring Chamber A, Si ring Ratio Si ring / Quartz ring Chamber B, Si ring Comparison at chamber A: - Quartz ring isolating - Si ring rf conducting increase of effective cathode area decrease of rf power density Maintenance Applications

29 Tool and chamber comparison at MxP Electron collision rate vs. etch time Etch rate test of Nitride etch at MxP 8 Chamber B, Si ring etch rate [nm / min] collision rate [1 7 s -1] Chamber A, Si ring Conditioning with resist wafers Chamber A Nitride etch rate ratio correlates with plasma parameter ratios. - Lower etch rate caused by lower power density one point one wafer date time [s] Comparison of Chamber A and Chamber B with Si shadow ring. Chamber B Parameter Ratio Ch A/ Ch B Nitride etch rate 1,17 Electron collision rate 1,19 Electron density 1,2 Bulk power 1, Maintenance Applications

30 Detection of tool failure: Al etch at LAM TCP Cl2 - MFC failure was detected before hardware alarm. Monitoring of main clean Al etching - trend analysis main etch - LAM TCP one point one lot quick clean Cl2-MFC error 2.17 main clean Lot No collsion rate [1/s] optical emission (EP) *3 etch time [s] 12 electron density [1/cm3] Cl2-MFC drift/error Joint project Siemens - ASI - Lam Trend analysis of Al main etch Maintenance Applications

31 Detection of tool failure: CT etch at MxP+ Bulk power vs. rf hours Higher mean values and higher variance of dissipated power (also electron collision rate and electron density. bulk power [mw/cm²] 4 Product Resist blank Si blank 3 Oxide blank 2 Caused by process instabilities. one point one lot rf hours bulk power [mw / cm²] Reason: Parasitic plasma inside the He feedthrough of the wafer backside cooling, below the powered electrode He leakage. Bulk power vs. etch time time [s] Maintenance Applications

32 Comparison of electron collision rate & etch rate 7-1 collision rate [1 s ] Electron collision rate vs. date WC1 WC3 WC5 Electron collision rate (also electron density and bulk power), measured on product wafers, detect the hardware failure. WC2 WC4 WC Oxide etch rate, measured on blank test wafers does not show any significant variation. etch rate [nm/min] Etch rate vs. date 72 7 WC1 WC3 WC WC2 WC4 WC Maintenance Applications

33 Arcing detection: Conditioning of emxp+ Electron collision rate (mean) 7-1 collision rate [1 s ] 3. Electron collision rate was the most sensitive of all measured parameters including reflected power. one point one wafer Electron collision rate vs. time wafer 25 wafer collision rate [1 s ] 3 Heavy arcing detected between e - chuck and wafer backside time [s] Maintenance Applications

34 Recipe control: CT etch at MxP+ 7-1 collision rate [1 s ] Electron collision rate vs. time Three product lots were etched with the test recipe for etch rate measurement to short, scrap. This error was not detected by any other control system. 8 product blank oxide time [s] - Electron collision rate differs on blank oxide and product wafers. - Hercules measures the etch time independently. - With an automatic alarm, scrap can be reduced to one wafer! Production Applications

35 Demonstrated applications of HERCULES - Long term process stability Short term process stability Detection of wafer effects Development and optimization of processes Endpoint detection Error prevention, no scraped lots Tool and chamber matching Monitoring of chamber cleaning Control of power coupling into plasma Detection of hardware failures Arcing detection Summary

36 Demands on process monitoring tools for industrial applications - Insensitive to insulating layers, e.g. polymers Independent of chemistry (different neutrals, neg. ions) Passive method, no impact on the plasma Applicable to existing tools Measurement of absolute parameters Real time analysis during the process Very high stability and reliability Easy to handle, plug and play tool Stable connection to fab network and data base Use of standard software for further analysis Minimise the quantity of data Cause only a minimum of additional work to the staff! Summary

37 Estimated Benefits Parameter Value (estimated) - Reduction of scrap lots: - Yield improvement at critical processes: Reduction of test wafers and test time for: etch rate particle monitor - Increase of OEE - Improvement of preventive maintenance Increase of MTBC about 5 % in the order of... 5 % %... 2 %... 3 %... 1 % % Summary

38 Outlook - Installation of HERCULES cluster tool at four chambers, long term monitoring of chambers and processes - Software improvements: reliable connection to fab network and data base - Measurements on other tools and processes: LAM TCP, AMAT DPS, TEL - Comparison with other measuring techniques, e.g. OES Summary