IBM Research Brazil: An Introduction

Transcription

1 : An Introduction July,

2 What does IBM do? 2012 Watson 1964 Solid Logic Technology IBM System 360 The machine that defined the computer industry and the modern IBM IBM System 360 SLT module 6 transistors,4 resistors 2 Watson System 360 Power7 chips 80KW / 80 Teraflops 1000Mflops/W Chip (POWER 7) 1.2 billion transistors/chip Embedded DRAM 190 watts max

3 Advances in Technology Source: Kurzweil 1999 Moravec

4 Complexity Science defines the future of Technology IBM has a long history of making translating, fundamental Silicon & nanotechnology discoveries and innovation into products Airgap edram 3D Chip Stacking High-k Immersion Copper SOI Strained Silicon Dual Core Chemically Amplified Resists Self Assembly Atomic Manipulation Frozen SiGe Chip Highest Resolution Carbon EM Nanotube Transistors Slowing Speed of Light Nanotube IC Atomic Molecular Storage Processing Nanophotonic Switch US$ 6B spent annually in R&D Nobel Prize, STM 4 Time

5 Scientific & Technological Achievements th Consecutive Year of Patent Leadership Watson System Nanoscale Magnetic Resonance Imaging (MRI) World s First Petaflop Supercomputer Web-scale Mining Core Extensible Markup Language (XML) Standards Services Science, Management, Engineering (SSME) Cell Broadband Engine Blue Gene/L Carbon Nanotube Transistors Java Performance Copper Interconnect Wiring Secure Internet Communication (HMAC, IPsec) Deep Blue Design Patterns Silicon Germanium (SiGe) Statistical Machine Translation High-Temperature Superconductivity Scanning Tunneling Microscope Reduced Instruction Set Computing (RISC) Speech Recognition Relational Database Fractals One-Device Memory Cell FORTRAN Random Access Memory Accounting Machine (RAMAC)

6 Distinguished scientists 5 Nobel Laureates Scanning Tunneling Microscope 10 US National Medals of Technology Copper Chip Technology High Temperature Superconductivity Electron Tunneling Effect 22 Members in National Academy of Sciences SiGe Over 400 Professional Society Fellows APS ACM AVS ACS 6 ECS 6 Turing Awards Excimer Laser DRAM AAAS 5 National Medals of Science Nuclear Magnetic Resonance Techniques Basis for MRI today 64 Members in National Academy of Engineering High Performance Computing First woman recipient in the history of this prestigious ACM award 11 Inductees in National Inventors Hall of Fame IEEE IOP OSA

7 A Diversity of Disciplines From Atoms to Service Science Mathematical Science Electrical Engineering Physics Chemistry Materials Science Computer Science Behavioral Science Service Science Science & Business & Engineering Management Technology Innovation Business Innovation Social Innovation Demand Innovation Social & Cognitive Sciences 7 Economic s& Markets

8 IBM Research: Global and Vertically Oriented Analytics Services Tools Almaden (1986/1952) San Jose, CA First New Research Lab in 12 Years Brazil (2010) Austin (1995) Sao Paulo &Rio TX de Janeiro Austin, Watson (1961) Yorktown Heights, NY Cloud Tokyo (1982) Zurich (1956) Rueschlikon, Switzerland WorkloadOptimized Systems & Dublin (2011) Supercomputing Yamato, Japan China (1995) Dublin, Ireland Beijing, China Shanghai (2008) Processing Brazil (2010) Sao Paulo &Rio de Janeiro 8 Africa (2012) Nanotechnology Nairobi, Kenya Haifa (1972) Haifa, Israel India (1998) Delhi, India Australia (2010) Melbourne, Victoria

9 The World is our Lab: 12 Labs Worldwide in 10 Countries Dublin 2012 Zürich 1956 Watson Almaden Haifa 1972 China Tokyo Austin 1995 Africa 2013 Brazil 2011 India 1998 Australia 2012 IBM Research worldwide has ~4000 research staff member with diversity of disciplines. 9

10 view from our Rio de Janeiro lab 10 Mission: To be known for our science and technology and vital to IBM, Brazil, our clients in the region and worldwide

11 IBM Research - Brazil Research Focus Areas Natural Resources Solutions Systems of Engagement Smarter Devices Social Data Analytics Underlying Research Areas: Analytics & Optimization HPC & Computational Science Distributed Systems & Cloud Computing Mobile technologies Physics, Chemistry, Mathematics & Engineering Semiconductor Packaging Service Science Social Science, Design & Human Computer Interaction A team of World Class Researchers in connection with global IBM Research as well as academic communities 11 Rua Tutóia São Paulo Av. Pasteur Rio de Janeiro

12 IBM Research - Brazil: Research Groups Natural Resources Solutions Social Data Analytics 12 Natural events, oil & gas, logistics, and sustainability Systems Of Engagement Large scale service systems operations, optimization, and integration in context of social enterprise Smarter Smarter city, citizen engagement, community integration, and education Micro- and nanotechnologies and materials aimed at addressing smarter planet challenges Devices

13 Natural Resources Solutions Academic Areas of Interest Applied Math Computational Sciences High Performance Computing Data visualization Mission: Create industry leading solutions and platforms with innovative data-driven, physically-driven and people-driven analytics 13

14 Systems of Engagement Academic Areas of Interest Education & Universal Design Healthcare Biodiversity & Sustainability Mobile and Ubiquitous Computing Mission: Promote social engagement, health, urban mobility, the inclusion of people with disabilities, and economic development. 14

15 Social Data Analytics Academic Areas of Interest Service Sciences & Design Distributed Computing Data & Graph Mining Information Visualization Analytics & Optimization AI Simulation & Machine Learning Human Computer Interaction, Design & Social Sciences Mission: Reinvent large scale service systems, operations, and enterprises. 15

16 Smarter Devices Academic Areas of Interest Physics Chemistry Fluidics Nanotechnology Electrical Engineering Electronics Mission: To conduct research in micro- and nanotechnologies and materials supporting Brazilian and global industries. 16

17 Areas of Interest Microfluidics in Health Care Enhanced Oil Recovery Nanotechnology Micro and Nano Technology Porous Rock Pore/Network modeling and microfludics Prototypes and Sensors CMOS based devices, MEMS and Sensors Electronic Packaging Computational Modeling Multiscale Modeling, Fludics 17 Smarter Materials Polymer Design EOR materials

18 Microfluidics in Healthcare SIX Semicondutores (HQ: Rio d.j., plant: Belo Horizonte, MG) Founded in 2012 in a public private partnership which includes IBM Most advanced semiconductor mfg. company of the Southern hemisphere with 130 & 90 nm (IBM s 7RF & 8RF technology and MEMS) with 360 WSPD (wafer starts per day) on 200 mm wafers. Products are customized integrated circuits with mixed signal /hybrid technology for industry and health care Wafer fabrication (130 nm) will begin in Brazil in 2015 Joint program BRL with SIX Semi Technology development project for microfluidics bio- & environmental sensor devices Technology is developed in IBM Research labs (BRL e ZRL) and focuses on control of reagent flow and fixing analytes in a specific place Industrial complex of SIX Semi in Ribeirão das Neves, MG (2013) Total sample volume: ~2µL Microfluidics device manufactured at ZRL Modeling of electrodes in a microfluidics device 18

19 Microfluidics for Rock Characterization Test simple and complex fluids in microfluidic devices of various wettability characteristics, chemistries and complexities Single Channel 19 Multi-Channel Reservoir on a Chip: Actual Rock Structure

20 Multiscale Modeling for Enhanced Oil Recovery Quantum Mechanics; Molecules Molecular Dynamics Porous media Fluid Flow Surfaces Reservoir simulation 20

21 Quantitative evaluation of flow fields using μpiv measurements and LBM simulations P. W. Bryant1 R. F. Neumann1 M. J. B. Moura1 M. Steiner1 M. S. Carvalho2 C. Feger PUC Rio

22 Outline 22 Introduction Literature Review Computational Methods Experimental Methods Results Examples Conclusion

23 Introduction What is Microscopic Particle Image Velocimetry (µpiv)? An experimental method for measuring fluid flow in microscale. How does it work? Fluid seeded with tracer (fluorescent) particles. Particles are excited with a laser and emit light. Emitted light is collected by a CCD. Consecutive snapshots allow determination of particle velocities. Flow velocity field is obtained. Where is it used? Microfluidics Microelectronics Healthcare Oil & Gas Chemistry... 23

24 Introduction Syringe pump MICRO-PIV In Out Fluid with particles Microcapillary Objective Dichromatic Mirror Laser Microscopic Synchronizer 24 Camera

25 Introduction Syringe pump MICRO-PIV In Out Fluid with particles Microcapillary Objective 1st Image (t) Dichromatic Mirror Laser Microscopic Synchronizer 25 Camera

26 Introduction Syringe pump MICRO-PIV In Out Fluid with particles Microcapillary Objective 1st Image (t) Dichromatic Mirror Laser Microscopic Synchronizer 26 Camera 2nd Image (t + t) Computer

27 Introduction 2) Pre-processing step 1) Image acquirement step 1st Image (t) particles + background 2nd Image (t+ t) background particles Background subtraction = 3) Processing step 4) Post-processing step outlier vectors local mean vectors Peak at the net particle displacement 27

28 Introduction Ensemble Average over Velocity Vectors Algorithm Image Sequence Frame A (t = t0 ) 1 A1 Frame B (t = t0+ t ) Correlation RAB B1 RA1B1 Peak Search + 2 A2 B2 RA2B2 + 3 A3 B3 RA3B3 + + N AN BN RANBN Average Velocity 28

29 Introduction RA1B1 Ensemble Average over Correlation Functions Algorithm Image Sequence Frame A (t = t0 ) 1 A1 Frame B (t = t0+ t ) Correlation RAB B1 RA1B1 RA2B2 + 2 A2 B2 RA3B3 RA2B2 + 3 A3 B3 RA3B3 + RANBN + N AN BN Average Correlation 29 RANBN Peak Search <RAB> <RAB>

30 Literature Review 1999 Exp. setup Curve fit parameters: - flow rate (pump) - width (channel) 450 nm resolution How can we explain the discrepancies...??? 30

31 Literature Review 2000 Exp. setup Out-of-focus particle images Depth of Field Typical values for δzm Measurement depth Contributions from out-of-focus particles...??? 31

32 Literature Review 2000 Exp. setup Weighting function Finite sampling region Depth of correlation 2 Finite sampling region... OK!!! =) Is this cumbersome formula the ultimate truth...??? 32

33 Literature Review 2011 Convoluted correlation function Velocity decrease as a function of DOC Theory x Experiments Spatial average does not work...??? 33

34 Literature Review 2012 Normalized profile Several experimental center Several pre-/post-processing walls Agreement between theory and experiments...??? 34

35 Literature Review Tracers vs Red Blood Cells 2012 Exp. setup Flow rate determination Depth of correlation Flow rate determination is parameter-dependent...??? 35

36 Literature Review 2013 Straight channels Experiment Exp. setup Simulation rescaled profile Sierpiński pattern - rescaled simulation to the experimental average. - velocity data taken at center Qualitative comparison...??? 36 Experiment Simulation

37 Computational Methods Boltzmann Equation Boltzmann Equation with BGK approximation Transforming Boltzmann Equation as dimensionless 37

38 Computational Methods Lattice Boltzmann Method Computational algorithm Collision Streaming 38

39 Computational Methods Collision Streaming 39

40 Computational Methods Laminar flow Porous media Higher Reynolds number 40

41 Experimental Methods µpiv System Manufactured by TSI Incorporated Controlled by Insight 3G/4G software by TSI Inc. Inverted microscope IX71S1F-3 by Olympus 10x/0.3 air objective UPlanFL-N by Olympus 2x projection lens by Olympus 1376 x 1024 pixels CCD Sensicam by PowerView 2 pulsed Nd:YAG lasers Gemini PIV-15 by NEW WAVE Laser pulse synchronizer by TSI Inc. Microfluidic chip Glass microfluidic device by Dolomite Centre Ltd: straight channel with an elliptical cross section (50 µm and 55 µm semi-axes) 14% aqueous solution of 1 µm fluorescent particles by Thermo Scientific Syringe pump 11 Elite by Harvard Apparatus: flow rates from 25 to 100 µl/h 41

42 Results Exp. setup Sampling Volume SEM image 50 µm 55 µm t = 500 µs Depth of Field ~20 µm 42

43 Results Flow field measurement 100 µl/h 32x32 pixel windows ~ 5.12 µm Simulated velocity Finding c and δ SV-channel intersections Fit residual minima Best fit kinks Average over SV 43

44 Results 44 Flow rate Channel geometry Theoretical velocity field Sampling Volume

45 Examples Robustness against camera misalignment camera x misaligned SV Determine c(x) θ =

46 Examples Robustness against noisy data Dust particle on microfluidic chip 46

47 Examples Robustness against irreproducibility Before disconnecting the pump After reconnecting the pump and refocusing the microscope Two measurements on the same channel and with the same flow rate Moving the channel and refocusing changes the location of the SV and, hence, the measured velocity profile. 47

48 Examples Changing processing algorithms Ensemble average over correlation functions Ensemble average over velocity vectors Velocity profile processed from the exact same set of images, but with different algorithms 48

49 Examples Analysis of Scanning PIV Poiseuille profile Focal plane Microscope objective Maximize velocity for c Laser beam Kloosterman et al., Maximum velocity

50 Conclusion A simple spatial average over the Sampling Volume suffices to explain the discrepancies between expected and measured velocity profiles. The near-wall features such as kinks provide extra information that allow the full determination of flow rates unknown to the experimenter. The Sampling Volume approach provided a straightforward interpretation of the measured data and was able to reproduce the experimental profile from wall to wall. µpiv measurements can be made quantitative without using post-processing. This approach is robust against the most common sources of experimental uncertainty. 50 The Scanning PIV procedure fails to locate the center of the channel for large DOC.

51 Acknowledgements Michael Engel from IBM Research Watson for the SEM images. Diney Ether from LPO UFRJ for helping with the calibration. José Florián from PUC Rio for help with the µpiv equipment. Angelo Gobbi from LMF - LNNano for profilometer measurements Contact: rneumann@br.ibm.com 51

52 Thank You! 52