Novel Reactor Design and Metrology Study for Tungsten ALD process Laurent Henn-Lecordier, Wei Lei, Gary W. Rubloff

Transcription

1 Novel Reactor Design and Metrology Study for Tungsten ALD process Laurent Henn-Lecordier, Wei Lei, Gary W. Rubloff Department of Materials Science and Engineering University of Maryland

2 OUTLINE Design of novel ALD reactor Motivations for new reactor design Our approach 2 ALD operational modes Integration of real-time mass-spectrometry Nucleation kinetics revealed in real-time by MS First wafer effect : understanding effects on film thickness metrology of competing wall and wafer surface reactions AVS talk, Wed. Nov. 2 nd (TF+EM-WeM3) MS-based real-time sensing in ALD processes 2

3 2 main approaches for ALD reactor designs Tube furnace (University approach) Substrate-heated production-scale tool (manufacturing approach) Low cost Reliability (simplicity of design) Fast cycle times QCM integration Small wafer size (1-2 ) High wall to wafer surface ratio Hot walls => wall reactions Contamination issue Low wafer throughput if no load lock Large wafers (up to 300 mm) Cluster tool integration High throughput Low contamination Automation Small wall to wafer surface ratio Cold wall High cost Not compatible with QCM F120 SAT SAM Microchemistry 3

4 Embedded mini-reactor concept Z-axis pneumatic actuator UHV chamber 10-5 torr Moveable cap 5 torr 100 mm wafer Substrate heater Differentially pumped MS 35 µm orifice Gas Outlet Gas Inlet ALD mini- reactor 4

5 Benefits of embedded mini-reactor concept 100 mm substrate heater 0.2 L reactor volume Wafer scale Low wall/wafer ratio Fast cycling time Low contamination In-situ diagnostic High wafer throughput Low cost Load-lock On-the-shelf UHV components Mass-spectrometry 5

6 Multiple Operation Modes Dynamic mode: steady-state flow with alternating gas species Exposure Purge 10-5 Torr Throttle Valve 10-5 Torr Reactant + carrier gas 5 torr carrier gas 5 torr Small reactor volume Viscous flow condition Short gas residence time Fast gas switching 6

7 Multiple Operation Modes Static mode: Fill-and-pumpout cycling of each gas Precursor A exposure Purge 10-5 Torr Gas inlet 5 torr Purge Precursor B exposure 5 torr Pumpout is accelerated by lifting the cap to employ the larger vacuum chamber as ballast 7

8 Real-time MS Process Sensing ALD process dynamics are directly observed by in-situ mass spectrometry 1.2x CYCLES x10-10 Intensity (A) 8.0x x x x10-11 SiH 4 exposure 1 CYCLE Purge WF 6 exposure Purge 2.0x :59:56 14:01:22 14:02:48 14:04:14 14:05:40 14:07:06 14:08:32 Time Precursors: WF 6, SiH 4 SiH 4 exposure H 2 product WF 6 exposure SiF 4 product MS response time: < 1 sec Intensity (A) 5.0x x x x x10-11 H 2 SiF 4 SiH 4 SiF Time (Sec) 8

9 Real-time Observation of Deposition Kinetics 4.0x10-10 Nucleation regime Linear growth regime 1.2x10-9 Nucleation regime Linear growth regime Integrated SiF 4 Signal Per Cycle 3.5x x x x x x x SiF 4 by-product Integrated H 2 MS Signal Per Cycle 1.0x x x x x H 2 by-product Deposition kinetics can be directly observed by integrating byproduct MS signal over each cycle and plotting it against cycle number 9

10 Effect of temperature on nucleation kinetics High process temperature accelerates the nucleation process Integrated H 2 Mass Spec Signal Per Cycle 2.5x x x x x C 250 C 175 C H 2 during SiH 4 half cycle

11 Real-time identification of nucleation kinetics for different wafer preparations Nucleation region for HF-last treated surface Nucleation region for H 2 O-last treated surface Integrated SiF 4 MS Signal Per Cycle 6.4x x x x x x x x HF Treated Surface H 2 O Treated Surface

12 Characterization of first wafer effect 1.2x nd wafer Gas Inlet Gas Outlet QMS sampling Integrated H 2 Signal Per Cycle 9.0x x x st wafer H 2 during SiH 4 ½cycle Integrated byproduct MS signal starts at much higher level for non-1 st wafer cases than 1 st wafer case 2. Integrated byproduct MS signal tends to the same level in linear growth region for both 1 st wafer and non-1 st wafer cases 12

13 Physically based model to understand 1 st wafer effect Nucleation occurs on wafer as well on walls Integrated byproduct MS signal per cycle Integrated byproduct MS signal per cycle Reaction on reactor surface Reaction on wafer surface a b 1 st wafer case Integrated byproduct MS signal per cycle Total reaction detected by MS a+b 13

14 Effect of wall conditioning on in-situ chemical diagnostic Integrated byproduct MS signal per cycle Integrated byproduct MS signal per cycle Reaction on reactor surface Reaction on wafer surface a b Non-1 st wafer cases Integrated byproduct MS signal per cycle Total reaction detected by MS For non-1 st wafers, nucleation occurs only on clean wafer surface and not on preconditioned wall surfaces b a 14

15 Real-time Film Thickness Metrology Linear correlation between total integrated byproduct MS signal and ALD film thickness 1 st wafer effect will make the 1 st wafer data deviate from the linear relationship With proper pre-process chamber treatment, 1 st wafer effect can be reduced W Film Thickness (nm) 35 (325 C) (275 C) (225 C) 1st wfr (175 C) (175 C) x x x x x10-8 Sum of Integrated SiF 4 Signal W Film Thickness (nm) (275 C) st wfr (275 C) x x x10-8 Sum of Integrated SiF 4 Signal (325 C) Without pre-process chamber treatment With pre-process chamber treatment 15

16 Conclusion 1. A novel wafer-scale ALD system has been implemented by embedding a mini-ald reactor (0.2L) inside a UHV chamber 2. Integration of in-situ mass spectrometry to ALD reactor has been optimized to achieve real-time process sensing 3. Deposition kinetics are observed in real-time using integrated MS signal data, but chamber history effect poses a challenge on the interpretation of MS results 4. The 1 st wafer data can not be used for ALD film metrology because of chamber history effects, and pre-process chamber treatment can help to reduce the effect of chamber history 16