acknowledgments - nci

Acknowledgments

Scientific EditingCamille Minichino

Graphic DesignJeffrey B. Bonivert

Art Production/LayoutDebbie A. MarshIrene J. ChanLucy C. DobsonKathy J. McCullough

Document Approvaland Report ServicesCynthia Tinoco

Cover:Computational graphics of particle motion and the effects of Weber number on droplet deformation.

ENGINEERING

Pushing engineering science to the Xtreme

FY01 • ETR • TechBase i

IntroductionGlenn Mara, Associate Director for Engineering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .v

Center for Complex Distributed Systems

Strategy and Capability for Communications and NetworkingA. J. Poggio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3

Characterization of Seismic Propagation and Signal Generation for Vehicle Tracking SystemsD. Harris, D. B. McCallen, D. W. Rock, P. Lewis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4

Compensated Optical CommunicationsC. A. Thompson, L. Flath, R. Hurd, R. Sawvel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5

Contact Constitutive Relationships for Interface Surface FeaturesC. R. Noble . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6

Extreme Bandwidth Security ToolsB. Bodtker, W. J. Lennon, D. Colon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7

High-Resolution Video SurveillanceC. Carrano, J. Brase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8

Radar VisionK. Romero, G. Dallum, J. E. Hernandez, J. Zumstein . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .9

Sensor-Driven Estimation of Chemical, Biological, and Nuclear Agent DispersalD. B. Harris . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10

Ultra-Wideband Communications System on a ChipC. McConaghy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .11

Evaluation of Micromachined Inertial Sensors for Distributed NetworksC. Lee, R. R. Leach, Jr., T. Woehrle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .12

Wireless Network on DemandB. Henderer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .13

Center for Computational Engineering

Development of Multi-Phase Flow Modeling CapabilityT. Dunn, L. Daily, R. Couch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .17

Engineering Visualization LabM. Loomis, R. M. Sharpe, K. Mish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .18

TechBase • ETR • FY01ii

Enhanced Fluid Dynamics Capability and Multidisciplinary Coupling in ALE3DR. McCallen, T. Dunn, G. Laskowski . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .19

Planning Tool for Site Remediation SimulationJ. Stewart, R. Glaser, A. Lamont, A. Sicherman, T. Hinkling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .20

Coupled Engineering Simulation Tools: Solids, Fluids, and ChemistryC. Hoover, A. Shapiro, R. Ferencz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .21

New Surface Loads and Display Capabilities in DYNA3DJ. Lin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .22

Advances in Implicit Finite-Element Algorithms in NIKE3DM. A. Puso . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .23

Development of Computational Capability for MEMS-Based TechnologiesM. A. Havstad, J. D. Morse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .24

Visualization and Data Management Tools: GRIZ4, Mili and XMiliCSD. Speck, E. Pierce, L. Sanford . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .25

ParaDyn Update: Parallel Interface AlgorithmsA. De Groot, R. Sherwood, C. Hoover . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .26

Center for Microtechnology

Integrated Microsyringe Arrays for Chip-Scale Fluid ControlM. Maghribi, P. Krulevitch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .29

Mesoscale NIF and Omega Laser Targets for High-Energy-Density Experimental Science fromNanofabricationR. Mariella, Jr. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .30

Microstereolithography for Fabrication of Mesoscale Structures with Microscale FeaturesV. Malba, A. F. Bernhardt, C. D. Harvey, L. Evans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .31

Optical Coating TechnologyD. Sanders, J. Wolfe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .32

Optical Pressure SensorM. D. Pocha, R. R. Miles, G. Meyer, T. C. Bond . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .33

Process Development for High-Voltage Photovoltaic ArraysG. Cooper, N. Raley, T. Graff . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .34

FY01 • ETR • TechBase iii

Rapid Fabrication of Microfluidic Devices by Replica Molding of Polydimethylsiloxane (PDMS)L. R. Brewer, K. Rose, O. Bakajin, P. Krulevitch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .35

Remote Hydrogen SensorD. R. Ciarlo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .36

Center for Nondestructive Characterization

Distributed Processing Algorithms for Reconstruction and RenderingG. P. Roberson, P. C. Schaich, H. E. Jones, M. W. Kartz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .39

Enhancements in Infrared NDE TechniquesW. O. Miller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .40

High-Precision Quantitative Tomography of Mesoscale TargetsW. Nederbragt, S. Lane, D. Schneberk, T. Barbee, J. L. Klingmann, R. Thigpen . . . . . . . . . . . . . . . . . . . . . . . .41

Linear Array Computed TomographyK. Dolan, J. Fugina, J. Haskins, R. Perry, R. D. Rikard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .42

Neutron Radiography Beam StopB. Rusnak . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .43

Rapid High-Resolution Ultrasound TomographyJ. Kallman, E. Ashby, D. R. Ciarlo, G. Thomas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .44

Synchrotron Microtomography at LBNL's Advanced Light Source FacilityK. Dolan, D. Haupt, J. Kinney, D. Schneberk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .45

Three-Dimensional ProfilometerM. W. Bowers, D. W. Swift, G. W. Johnson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .46

Center for Precision Engineering

Engineering Mesoscale InitiativeD. Meeker, R. Mariella, Jr., H. Louis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .49

Fast Tool Servo Application to Single-Point Turning Weapons Physics TargetsR. C. Montesanti, D. L. Trumper (MIT), J. L. Klingmann . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .50

TechBase • ETR • FY01iv

Other Technologies

Frame Extraction and Image Processing in Scene AnalysisA. Gooden, L. Scott . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .53

Surface and Volumetric Flaw Distribution in Brittle MaterialsR. A. Riddle, C. K. Syn, S. Duffy, J. Palko, E. Baker (Connecticut Reserve Technologies) . . . . . . . . . . . . . . . . . .54

Nitrogen Augmentation Combustion SystemsL. E. Fischer, B. Anderson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .55

Author IndexAuthor Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .59

FY01 • ETR • TechBase v

Engineering has touched on every challenge,every accomplishment, and every endeavor of

Lawrence Livermore National Laboratory during itsfifty-year history.

In this time of transition to new leadership,Engineering continues to be central to the mission ofthe Laboratory, returning to the tradition and corevalues of E. O. Lawrence: science-based engineering—turning scientific concepts into reality.

This volume of Engineering Technical Reportssummarizes progress on the projects funded for tech-nology-base efforts. Technology-base projects effect thenatural transition to reduction-to-practice of scientific orengineering methods that are well understood andestablished. They represent discipline-oriented, corecompetency activities that are multi-programmatic inapplication, nature, and scope.

Objectives of technology-base funding include: • the development and enhancement of tools and

processes to provide Engineering support capability,such as code maintenance and improved fabricationmethods;

• the support of Engineering science and technologyinfrastructure, such as the installation or integration of anew capability;

• support for technical and administrative leadershipthrough our technology Centers;

• the initial scoping and exploration of selected tech-nology areas with high strategic potential, such as assess-ment of university, laboratory, and industrial partnerships.

Five Centers focus and guide longer-term invest-ments within Engineering. The Centers attract andretain top staff, develop and maintain critical core tech-nologies, and enable programs. Through their technol-ogy-base projects, they oversee the application ofknown engineering approaches and techniques toscientific and technical problems. The Centers and theirleaders are as follows:

• Center for Complex Distributed Systems: David B. McCallen

• Center for Computational Engineering: Robert M. Sharpe

• Center for Microtechnology: Raymond P. Mariella, Jr.

• Center for Nondestructive Characterization: Harry E. Martz, Jr.

• Center for Precision Engineering: Keith Carlisle

FY2001 Center HighlightsThe Center for Complex Distributed Systems

exploits emerging information technologies to developunique communications related to data gathering,advanced signal processing, and new methodologies forassimilating measured data with computational modelsin data-constrained simulations of large systems.

Current Center technology-base activities include:construction of practical wireless communication nodesfor application to self-configuring, self-healing networksusing commercial off-the-shelf hardware; acquisitionand education on usage of commercial wirelesscommunication network simulation tools; deploymentof seismic sensors at the proposed Yucca Mountaingeologic repository to gather data for application ofarray processing software in site monitoring; computercode implementation of model update algorithms intoan application-ready package for updating and opti-mization of structural models based on observed data;modification of finite element contact surfaces toinclude omni-directional behavior necessary for model-ing contact surfaces with complex surface details; andconstruction of wireless vibration sensors for applicationin NIF ambient vibration monitoring.

These technology-base projects are delivering appli-cation-ready tools into the hands of engineers support-ing programs, and thus serve a critical link in transition-ing from research to practice.

The Center for Computational Engineeringorchestrates the research, development and deploy-ment of software technologies to aid in many facets ofLLNL's engineering mission. Computational engineer-ing has become a ubiquitous component throughoutthe engineering discipline. Current activities range fromtools to designing the next generation of mixed-signalchips (systems on a chip) to full scale analysis of keyDOE and DoD systems.

Highlights of the Center’s technology-base projectsfor FY2001 include enhancements of engineering simu-lation tools and capabilities; progress in visualizationand data management tools; and updates in parallelinterface algorithms. The Center has offered a real-world computing capability that opens the door tosolving a wide variety of fluid/solid interaction problemsin transportation, aerospace, and infrastructure settings.

The mission of the Center for Microtechnology isto invent, develop, and apply microtechnologies forLLNL programs in global security, global ecology, andbioscience. Its capabilities cover materials, devices,

IntroductionGlenn Mara, Associate Director for Engineering

TechBase • ETR • FY01vi

instruments, or systems that require microfabricatedcomponents, including microelectromechanicalsystems(MEMS), electronics, photonics, microstructures, andmicroactuators. Center staff have achieved considerablenational recognition for the successes demonstrated inChem-Bio National Security Program instrumentation,supported by the DOE and the Defense IntelligenceAgency. The majority of the Center’s technology-baseprojects support defense programs and nonprolifera-tion. These include the application of high-voltagephotovoltaics, microaccelerometers, and fabricationtechniques to physics experiments that supportStockpile Stewardship.

Highlights in FY2001 include integrated microsy-ringe arrays for chip-scale fluid control; meso-scalelaser targets for high-energy-density experimentalscience; microstereolithography for microfabrication;and process development for high-voltage photo-voltaic arrays.

The Center for Nondestructive Characterizationadvances, develops and applies nondestructive charac-terization (NDC) measurement technology to signifi-cantly impact the manner in which LLNL inspects, andthrough this, designs and refurbishes systems andcomponents. The Center plays a strategic and vital rolein the reduction-to-practice of scientific and engineer-ing NDC technologies, such as acoustic, infrared,microwave, ultrasonic, visible and x-ray imaging, toallow Engineering in the near term to incorporate thesetechnologies into LLNL and DOE programs.

This year’s technology-base projects include distrib-uted processing algorithms for reconstruction andrendering, enhancements in infrared techniques, andseveral advanced applications of computed tomogra-phy and neutron radiography.

The Center for Precision Engineering is dedicatedto the advancement of high-accuracy engineering,metrology and manufacturing. The Center is responsi-ble for developing technologies to manufacturecomponents and assemblies at high precision and lowcost; developing material removal, deposition, andmeasurement processes; and designing and construct-ing machines that embody these processes.

The scope of work includes precision-engineeredsystems supporting metrology over the full range oflength scale, from atom-based nanotechnology andadvanced lithographic technology to large-scalesystems, including optical telescopes and high energylaser systems. A new focus is the manufacturing andcharacterization of “meso-scale devices” for LLNL’s NIF.Millimeter-scale physics experiments will provide dataabout shock physics, equation of state, opacity, andother essential measurements of weapons physics.

FY2001 highlights include a fast tool servo appli-cation to single-point turning weapons physicstargets, and surface and volumetric flaw distribu-tion in brittle materials.

Science-Based EngineeringOur five Centers develop the key technologies that

make Laboratory programs successful. They providethe mechanism by which Engineering can helpprograms attract funding, while pioneering the tech-nologies that will sustain long-term investment.

Our Centers, with staff who are full partners inLaboratory programs, integrate the best of mechanicaland electronics engineering, creating a synergy thathelps turn the impossible into the doable.

Strategy and Capability for Communications and NetworkingA. J. Poggio

The purpose of this technology base project was tobring together the required elements for an Engineeringcapability in communications and networking (C&N) sothat the Directorate could better serve its customers inthe programs and in the Work-for-Others regime.

Spiros Dimolitsas, former Associate Director forEngineering, envisioned a scenario where the activities

being executed by engineers in C&N would be orga-nized, enhanced, and collocated when possible.Included in this vision was the development of C&N tothe level of an Engineering core technology.

To efficiently establish a roadmap for a C&N capability,we chose to create a catalog of LLNL strengths andweaknesses in this arena. It was defined to encompassRF, optics, photonics, networks, and security.Communications theory, analysis, and modeling werecritically assessed in terms of capability level; need, inthe context of present and anticipated projects at LLNL;and potential customers in the programs and else-where. Such a view lent itself to the definition ofrequirements in the technology base.

Similar assessments were carried out in the areas ofcommunications systems and testing. It was apparentthat LLNL’s core strength in remote sensing called for anaccompanying strength in C&N, and that capabilitiesin networks, wideband and optical communications,and systems integration warranted further attentionand advancement.

The development of a capability requires access tostate-of-the-art equipment for executing tasks andperforming R&D. To this end, we sought to create astrategic plan for equipment purchase that would becomplementary to purchases that had been made inFY01 using IGPE funds.

We also demonstrated the universality of the C&Ncapability that was being organized by assimilatingequipment from the programs that would be able toservice LLNL at large. In addition, plans were finalizedfor the establishment of a National TransparentOptical Network (NTON) lab and for an unclassifiedcommunications lab in what was then referred to asthe C&N Laboratory.

FY01 • ETR • TechBase 3

During FY01, a classified communications laboratorywas established within the confines of theElectromagnetics Laboratory. A small library was pulledtogether to support the capability represented by thelaboratories. To round out the beginning stages of thiscapability, we scoped out the needs of a communica-tions design “workstation,” and we coordinated theacquisition of a high-end computer and communica-tions design software using alternative sources of funding.

It was apparent that a core technology in C&Nwould require the establishment of relationships withprominent practitioners across the nation. We felt thatrelationships with universities would provide fruitfulsources for technology and expertise as well as recruit-ing opportunities for students. Several exchange visitswere made during FY01 with the University ofMaryland, holder of an extraordinarily strong reputa-tion in C&N. They visited LLNL and briefed us on theirinterests and ongoing projects and, in return, webriefed their staff members on the same issues.

Strong interactions also took place with key staffmembers at the University of Texas, Dallas, and theVirginia Polytechnic University. To ensure the relevancyof the capability that we were developing, we alsointeracted with several government agencies that werevery active in the C&N arena. Meetings were held withthe elements of the Joint Chiefs of Staff, the Laboratoryfor Telecommunication Science (UM and NSA), CIA,NSA (IOTC), the Navy Information Warfare Agency, theArmy’s CECOM, and the SPAWAR Network Modelingand Simulation Branch. Throughout, we receivedencouragement regarding our development strategy,and a desire to keep a close working relationship.Included in these interactions were also manyexchanges of lessons learned.

At the end of FY01 we had developed a strategy fora strong C&N capability; prepared and equipped acommunications laboratory in a classified area; laidplans and coordinated the advanced stages of prepar-ing an unclassified C&N laboratory; assembled a C&Ndesign workstation with software; and had initiatedand nurtured relationships that would help us growthis capability.

Characterization of Seismic Propagation and Signal Generation for Vehicle Tracking SystemsD. Harris, D. B. McCallen, D. W. Rock, P. Lewis

Remote detection, tracking, and identification of vehi-cles are important in battlefield and surveillance appli-cations. Networks of seismic arrays may be deployed toperform these functions. How well these systems willperform is critically dependent on how well vehiclescouple vibration energy to the ground and how effi-ciently the generated seismic energy propagates to thesensor arrays. To determine these characteristics wecarried out two experiments at the Yucca dry lakebed atthe Nevada Test Site (NTS).

In the first of our experiments we used a log-lineararray of eight seismic instruments to determine the

attenuation rate of seismic wave propagation. From thisdeployment, we learned that seismic signals propagateprimarily as surface waves and that the attenuation rateis large at high frequencies.

The second experiment consisted of 20 instrumentsdeployed in a fan-shaped array to characterize theseismic radiation pattern of tracked and wheeled vehi-cles. We learned that tracked vehicles project seismicenergy predominantly in the direction of motion, andthat this effect becomes more pronounced asfrequency increases.

Seismic sensors may be a component of an instru-mented battlefield along with radars, acoustic sensorsand optical systems. One of the principal issues is how

TechBase • ETR • FY014

to increase the “standoff” (detection range) of systemsso that relatively large areas can be observed with amodest deployment of sensors.

Our deployments demonstrated several key featuresof signal generation and propagation that may guidethe engineering of new seismic sensors and contributeto processing algorithms.

Usable signals are observed at all ranges, but arerestricted to relatively low frequencies (much lessthan 10 Hz) at the greater ranges (2000 to 4000 m).This observation, corroborated by experience withseismic arrays in Norway observing rail traffic,suggests that new ground instrumentation develop-ment for long-range vehicle tracking should focus onlow-frequency performance.

Perhaps the most interesting and new discovery camefrom work with our vehicle seismic characterization test-bed, shown in Figure 1. We determined that trackedvehicles exhibit a pronounced seismic wave radiationpattern. At low frequencies, seismic energy is projectedpredominantly fore and aft in the line of travel (Figure 2).At higher frequencies, it is even more directional, beingprojected primarily forward in the direction of travel.

This observation impacts estimates of vehicledetectability and sets a criterion to be met for real-istic simulation of seismic signals generated bymoving vehicles.

3.5 Hz component

15 Hz component

N

NENW

E200 m

342 m585 m

1000 m

W

Figure 1. Fan-shaped vehicle characterization array. Figure 2. Seismic wave radiation patterns at two frequencies for atracked vehicle.

Compensated Optical CommunicationsC. A. Thompson, L. Flath, R. Hurd, R. Sawvel

High-bandwidth, secure/covert communications is animportant piece of sensor-based systems for DOE, DoD,and intelligence community applications. Free-spaceoptical systems provide high-bandwidth, covert, wirelesscommunications, but are limited by atmospheric turbu-lence. We have proposed an Adaptive Optics (AO) test-bedsystem that allows flexible testing of new concepts andtechnology for optical communications applications.This new capability will lead to future projects fromcurrent LLNL sponsors in national security and defense,and put LLNL in a position to make significant contribu-tions to the field of free-space optical communications.

Wavefront compensation for free-space systemswill improve performance by reducing power

requirements, increasing data bandwidths, andlengthening transmission distances. Leveraging theimprovements in MEMS technology makes wavefrontcompensation feasible.

The goals for FY01 were to complete and demon-strate an integrated optical communications and wave-front control test-bed based on MEMS technology,using equipment purchased with Engineering capitalequipment funding in FY00. As shown in Figure 1, wehave designed and assembled the wavefront controlsystem. We have not yet demonstrated closed loopperformance, due to safety concerns with potentialelectrical hazards.

The electronics package shipped with the MEMSdevice is a prototype unit. After investigating how thehigh-voltage electronics box (Figure 2) was puttogether, we identified and rectified five potential elec-trical safety issues. We also collocated all electronics inone enclosure (Figure 3), which made sense to us fromboth safety and practical standpoints. The new systemhas been inspected and approved by an AHJ (Authority


Figure 2. High-voltage electronics box for MEMS system. The redsymbols indicate potential hazards we identified and rectified.

(a) (b)

Figure 1. (a) MEMS mirror integrated into the Adaptive OpticsTest-bed and (b) Adaptive Optics “Lunchbox.” MEMS technologyallows miniaturization of the system.

Having Jurisdiction) as per Document 16.3 of the LLNLEnvironment, Safety and Health Manual.

A Boston Mircomachines 140 actuator MEMS mirrorhas been successfully integrated into our AO test-bed(Figure 1a). The MEMS device has been used recentlyas an aberrator in the AO test-bed. A second wavefrontcontrol device (in this case a liquid crystal spatial lightmodulator (SLM)) is used to correct the aberrationscaused by the MEMS device. In the near future, theMEMS device will actually be used as the correctorelement for an aberration placed into the system (mostlikely by the SLM).

The addition of the MEMS device to the AO test-bedhas allowed for improvements in wavefront controlalgorithms and software. We have also built a smallerAO system as a proof of principle (Figure 1b). Becauseof its small size (approximately 40 cm × 30 cm) wehave dubbed it the “AO Lunchbox.” Currently thesystem runs with electronics shown in Figure 3 and adesktop PC. We have not yet run the AO system in thispackage. Closing the loop and modifying the systemwill be a future project.

Figure 3. New completed electronics enclosure.

The contact algorithms at our disposal currently do nothave the capability to accurately model the local topologyof a structure’s contact interface, where load is oftentransferred between components through the use ofgeometrical discontinuities. The presence of discontinuitiesis a common feature in many mechanical systems, suchas Hard and Deeply Buried Targets (HDBT) and largemulti-bodied structures such as concrete dams andbridges. Our objectives are: 1) determine a constitutivelaw that accounts for interface surface features; 2) validatethe algorithm with experimental data; 3) implement thealgorithm into FEM codes; 4) develop university collabo-rations; and 5) enhance our relationship with sponsors,the US Bureau of Reclamation (USBR).

Hardened Deeply Buried Targets consist of concretetunnels surrounded by large rocks buried deep

within a soil mass. To accurately simulate the responseof these structures, the rock joints need to account fordamage or degradation along the interface, as well asdilatancy (volume increase under simple shearingstress) and local topology effects. It is not realistic toassume that a rock joint can be modeled with asmooth, flat interface.

To model rock interfaces, a more detailed traction-displacement law is required. Similarly, concrete damstructures may consist of 30 or more contact interfacesthat include shear keys (see Figure 1). These shear keysresult in unique directionality effects along the concreteinterfaces similar to that of rock interfaces.

Our accomplishments to date have included anextensive literature review on the existing contact algo-rithms, as well as identifying the contact algorithmcapabilities needed for future work on large multi-bodied structures (HDBT and concrete dams).

The first capability needed is the ability to defineseparate slide surfaces along user-defined planes. Thereason for these planes is best described by looking atthe finite element model of Morrow Point Dam, aconcrete arch dam, in Figure 2. Because of the shearkeys located on each contraction joint, the separateconcrete blocks are free to move only vertically alongthese planes and normal to the adjacent block.

Once directionality has been added to the contactsurface, it will be useful to have separate cohesive/shearstrength terms (C1/ τ1, C2/ τ2) for the upstream/downstream direction and the vertical direction. Finally,it is imperative that the contact surface be able toprovide shear resistance as a function of the gap

Contact Constitutive Relationships for Interface Surface FeaturesC. R. Noble


Figure 2. Needed contact algorithm capabilities.

C2, τ2

C1, τ1

User-defined planeslocated along

contraction joints

distance. The reason for this is that the shear keysprovide shear resistance up until the normal gapbetween the blocks opens more than the thickness ofthe shear keys (Figure 1).

In addition to identifying the needed capabilities, theUSBR is beginning to perform shake table tests on typicalconcrete lifts used in dam structures. This data will beused to validate the new contact algorithm. Funding($80 K) was also recently secured from the USBR toperform a seismic evaluation of Morrow Point Dam.One of the primary project objectives is to incorporatethe contact features already discussed and run a suite ofsuitable test problems.

Other accomplishments include consulting with UCBerkeley and obtaining work on subterranean effects(HDBT) for LLNL’s DNT/Q Division to validate andimport finite element tools, such as contact algorithms,for use in analyzing underground facilities.

Shear keys

Figure 1. Shear keys common to concrete arch dams.

Extreme Bandwidth Security ToolsB. Bodtker, W. J. Lennon, D. Colon

While intrusion detection and firewall protection areavailable at multi-Mb/s data rates, cost-effective analysisand management of multi-Gb/s data streams arecurrently beyond the state of the art. Organizations likeLLNL will be depending on very-high-speed networks fora variety of collaborations in the near future, but can nolonger deploy networks for regular business withoutbeing able to manage their security. We are leveragingour role in high-bandwidth networking to strengthenour collaboration between nationally recognizedexperts in different aspects of high performance networkanalysis and security management. The goal is to jump-start the commercial development of security tools sothat they are readily available when LLNL and otherorganizations require multi-Gb/s networks for normalproduction activities.

As a first example of a necessary security tool, webelieve it is feasible to create a network intrusion

detector (NID) that operates at OC-48. Doing experi-ments and demonstrations at OC-12 will prove feasibilityand establish credibility to pursue a larger effort forupgrading the design to the higher rate. TheLaboratory has expertise in NIDs, having developed thesoftware currently used at DOE sites for detecting intru-sions; this software continues to be developed underthe auspices of the Nonproliferation, Arms Control, andInternational Security (NAI) Directorate.

The current capability of the software is sufficient toperform its function on the existing 100 Mb/s networksat these sites, but scaling the performance to high-bandwidth networks is not as simple as running onfaster processors. Some hardware assistance is necessary.

Network protocol analyzers on the market arecapable of viewing all packets on a network at speedsup to 10 Gb/s, which is a key requirement for intru-sion detection. Some of these devices use fieldprogrammable gate arrays (FPGA) that could bereprogrammed to perform the pattern matchingneeded for this function. We felt that collaboration

between a vendor of a programmable analyzer andNID software developers could produce a prototypeof a high-bandwidth intrusion detector.

In FY00, we acquired Boeing’s FPGA-based AS4950network protocol analyzer, and began collaborating withBoeing to combine their hardware with NID softwareto produce an intrusion detector operating at OC-12(see figure). Boeing was to modify the firmware andsoftware in the AS4950’s Generic Processing Engine(GPE) that would support sophisticated pattern match-ing and filtering functions. This would allow fordramatic reductions in NID-processed packets to levelswithin the NID’s capability. The collaboration hasproduced a design specification for the Boeing modifi-cations that is currently proprietary to Boeing.

In FY01 we developed both the application programinterface (API) and controller interface (CI) softwarepackages necessary to interact with the NID software.The CI is GUI-based and user-friendly. We opted to usethe SNORT NID package as it was both well under-stood and in the public domain and, as such, allowedus access to the source code for any modifications werequired. We successfully integrated the three softwareelements in a UNIX-based PC and were awaitingcompletion of Boeing’s development effort at fiscalyear’s end. If FY02 funding is secured we will integrateLLNL’s software with Boeing’s hardware and begin testingwith artificially generated traffic. From this we hope toleverage what we learned to extend the technology toan OC-48 rate.

The OC-12 system block diagram.


SplitterOC-12

100 base T

AS4950GPE

Subsystem

NIDSubsystem


High-Resolution Video SurveillanceC. Carrano, J. Brase

Atmospheric and optical aberrations reduce the resolutionand contrast in surveillance images recorded over long(>1 km) horizontal or slant paths. A capability ofimproving such images is of great interest to theDoD/intelligence communities. In this project, weproposed to demonstrate a prototype remote surveillancesystem that corrects these aberrations and improvesresolution in such scenes using speckle imaging techniques.

Speckle imaging techniques require multiple frameacquisition of short exposure imagery to recover the

high-resolution information that is lost in a typicalblurred long-exposure image. During FY01, we delivereda prototype remote video surveillance system consistingof the following:

Hardware:• Celestron 8-in. telescope and associated optics • Qimaging Retiga 1300 CCD camera—

1280 × 1024 pixels, 12-bit camera capable ofshort exposures down to 35 µs

• Gateway laptop with Firewire connection for thedata transfer

Software:• Data acquisition software (Empix Imaging

Northern Elite)• Data preprocessing algorithms in IDL• Speckle reconstruction softwareWe performed several horizontal-path imaging

experiments from building roof tops this year with avariety of targets including resolution targets, people,vehicles and structures. We obtained excellent resultswith the speckle processing: with a sub-resolutionpoint target, speckle processing resulted in a spot with1.3 times the telescope diffraction limit. We alsodemonstrated improved resolution in both near- andfar-range scenes of interest. Example results are shownin the figures.

To make this capability more attractive to potentialcustomers, we plan to perform more experiments andanalysis, directed at better quantifying system perfor-mance in varying conditions and system parameterconfigurations, including ranges beyond 10 km. Wealso plan to use enhanced imagery of personnel andvehicles to evaluate the accuracy of gross biometric andvehicular feature measurement.

Person at 0.5 km range

Single, raw frame

Speckle processed image

Tower at 6 km range Truck at 10 km range


Radar VisionK. Romero, G. Dallum, J. E. Hernandez, J. Zumstein

Special Forces, DARPA, DOE, and the intelligencecommunity have a need for the capability to “see”through walls and smoke. Law enforcement firstresponders, such as SWAT teams, require a portable,affordable device that will provide them real-time, full-motion images of a crime scene through exterior andinterior building walls, and through smoke. The abilityto differentiate between perpetrators and victims, and tolocate weapons, is highly desired. To get a good qualityimage useful for shape recognition, an ultra-widebandscanning radar system or an array of radars, with suit-able imaging algorithms, is needed to obtain a highresolution Synthetic Aperture Radar (SAR) image of thescene (Figure 1).

LLNL is a world leader in developing high-resolutionSAR imaging systems for NDE applications such as

the inspection of bridge decks. We have pioneeredthe use of low-power ultra-wideband impulse radarscanning systems and arrays for high-resolution imag-ing of materials.

The plan for the first year was to develop a test-bedfor scanning a single radar in a 2-D vertical plane undercomputer control to simulate different array configura-tions and electronic beam steering. This test-bedconsisted of various elements, which included thetiming circuits (Figure 2a) and software for controllingthe firing of the radars; the radars themselves (Figure 2b);and finally the data acquisition software.

The test-bed has allowed us to acquire SAR data of avariety of objects for testing and evaluating differentimaging concepts. We created a theoretical model forthe system response of scanning the vertical planealong the target. Then we varied both the lateral andlongitudinal placement of the target and acquired realdata. By comparing the real data to simulated data, wewere able to verify the functionality of the system. Bytesting different scanning patterns and radar antennaplacement, we were able to get additional insight intooptimal radar placement and scanning algorithms.

During the second year, different imaging algo-rithms and an expanded number of radar elements(transmitters, receivers) will be tested to determine thebest approach for developing a portable SAR imagingsystem for imaging objects from a distance through airand a variety of wall materials.

Our FY01 results and expected FY02 efforts willprovide the groundwork for attracting follow-up fundingfor a complete portable imaging system prototype forintelligence applications.

Figure 1. Low-power ultra-wideband impulse radar scanning system.

(a)

(b)

Figure 2. (a) Timing circuits and (b) radars for test-bed.


Sensor-Driven Estimation of Chemical, Biological, and Nuclear Agent DispersalD. B. Harris

This proposal was developed jointly with the Energyand Environment Directorate, NAI, and Computationsat LLNL. The proposal identified a new research/busi-ness opportunity for NARAC and Engineering, entailinga shift in NARAC’s current operations from open-loopprojection of future plume distributions to closed-loopestimation of past, present, and future plume distribu-tions. The proposed system is an instance of a moregeneral paradigm emerging in the physical sciences:complex numerical models for physical processes devel-oped originally to explain experimental observationsstatically are embedded in real-time, sensor-drivensystems that dynamically estimate parameters of inter-est. This paradigm represents the maturing of scientificmodels into operational systems.

Sensor-driven estimation of dispersing plumes wouldposition LLNL to compete vigorously for significant

new research and operational opportunities in a varietyof environmental, civilian, and military atmosphericrelease advisory applications. These applicationsinvolve the detection and tracking of radiological,chemical, and biological agents/pollutants releasedinto the atmosphere.

NARAC currently has sophisticated dispersionmodels for predicting patterns of mean plume concen-tration, given a source, but no systematic capability toinvert for source locations and characteristics, givenplume observations.

We proposed to develop an inversion capability byestablishing the modeling, estimation, and communi-cation framework to exploit data, first from networks ofpre-deployed sensors at cooperating sites, then fromnext-generation networks of rapidly-deployed, mobilesensors. The latter can be deployed after detection, andreconfigured as a release incident unfolds. The signifi-cance of networks of mobile sensors is that the conditionof ill-conditioned inverse problems may be improvedby judicious placement of sensors to acquire, surround,and track an evolving plume.

We envisioned a project-deploy-correct cycle, inwhich NARAC forecasts guide sensor placement in realtime, and network data is fed back to update themodels. The cycle may repeat from initial detectionuntil final plume dilution.

The main applications of interest were identified astreatment and protection of civilian and military popu-lations in scenarios of chemical, biological, and radio-logical attack; accidental industrial release of toxicmaterials; accidental release of radionuclides fromcooperating DOE and DoD facilities; and bombdamage assessment and minimization of collateral civilianexposure in strikes against WMD facilities.

New technical capabilities must be developed,acquired, or adapted to support application of basicatmospheric dispersion science to practical inverseproblems; adaptation of robust military communicationtechnologies to ensure that field data can be communi-cated to NARAC under rapid-deployment conditions;and integration of miniaturized, fast-response biologicaland chemical agent sensors under separate develop-ment into an operational forecast and inversion system.

The expected result of the proposed initiative was aset of field-validated technologies that would radicallyexpand NARAC’s operational capabilities: codes to esti-mate rapidly source location and characteristics fromreal-time concentration observations; codes to estimatepast plume concentration and dosage in space andtime; codes to forecast plume concentration anddosage in an ensemble of wind fields; and a communi-cation framework for integrating these models withcurrent and next-generation theater meteorologicaland concentration sensors.

The technologies to be developed under this initiativehave natural marketing outlets through existing LLNLprograms. Several NAI programs have interests inplume tracking and agent source estimation problems.Military applications include planning for strikes againstWMD facilities to minimize collateral civilian exposureand to perform bomb damage assessment.

The same techniques have domestic civilian applica-tions of interest to NAI: assessing and possibly mitigat-ing the impact of terrorist attacks on civilian targetsduring special events. In principle, the same servicescould be offered to the wider base of 120 municipalitiesspecified under the Nunn-Lugar-Domenici (NLD)amendment of 1997, for assistance in incidents ofdomestic terrorism. The rapid assessment and initialdetection teams designated under NLD could be thefield operational arm of an advisory system promotedby NAI and involving NARAC.


Ultra-Wideband Communications System on a ChipC. McConaghy

Ultra-Wideband (UWB) communications systems trans-mit data with pulses as short as 100 ps. These pulseshave very low probability of detection and can be usedin very-low-power communication systems. UWBcommunications systems developed previously at LLNLuse discrete electronic and microwave componentsassembled in a hybrid breadboard configuration. Theobjective of this work has been to incorporate some ofthe previous work into an application-specific integratedcircuit (ASIC).

This work included a survey of available processes,circuit topologies, and computer simulations of the

circuits with SPICE. Current resources did not permitcircuit layout and fabrica-tion. The most suitableprocesses available for thiswork include sub-µm

CMOS as well as SiGe. For example, foundry CMOSwith gate lengths of 0.35 to 0.16 µm have gate propa-gation delays ranging from 61 to 28 ps, which is morethan adequate for generating the UWB pulse.

Modeling was done of a pulse generator capable ofgenerating a pulse with 100 ps width. Additionalmodeling was done on an amplifier that could amplifythese pulses such that they could drive the 50-Ω load ofthe antenna. Circuits were also modeled for encodingdata with these pulses. The actual modeling was doneusing real transistor models from the foundry thatwould do the later circuit fabrication.

The modeling results (see figure) indicated that,indeed, short (100 ps) pulses could be generated in 0.35CMOS, data encoded, and amplified to the 50-Ω level.

4

2

0

–25004003002001000

Time (ps)

Volt

s

V(U6:Out)

V(U1:In)

Simulation results of the UWB pulse generator in 0.35-µm CMOS. The blue trace is the step input to thegenerator; the red trace is the 100-ps pulse output.


Evaluation of Micromachined Inertial Sensors for Distributed NetworksC. Lee, R. R. Leach, Jr., T. Woehrle

Distributed wireless sensor networks made up ofhundreds to thousands of individual sensor nodes canbe deployed in locations or on structures in severe, inac-cessible, remote environments. These networks can beused to collect information; monitor the state/health ofa structure or mechanical system; or validate large-scale computational simulation models. This projecthas developed a test capability for evaluatingaccelerometer sensor packages for implementation indistributed networks.

Two requirements for accelerometer sensors indistributed networks are 1) that the overall package be

as small as possible, and 2) that the sensors be sensitiveenough to measure motion in a near DC-levelfrequency range. Capacitive-type micromachinedaccelerometers can meet these requirements, due totheir miniature size and frequency bandwidth.However, accelerometers with small inertial proofmasses may have trouble differentiating signals frombackground noise at low excitation levels and frequencies.

The primary component of the test-bed is an air-bearing horizontal shaker table (Figure 1). The peak-to-peak stroke of the shaker is 6.25 in. Its performanceenvelope is approximately 0.2 Hz to 200 Hz and 0.01 gto 1.5 g (for single sine frequencies). Input to theshaker table is by a controller that can specify random,single/swept sine, or user-defined time series inputs.Various control or reference accelerometers were useddepending on the desired input signal frequency andamplitude range.

A series of off-the-shelf commercial, microma-chined (or MEMS) type capacitive accelerometers

Figure 1. Air-bearing shaker table in test-bed.

were evaluated and their performance was recordedto form a data “library.”

Performance of the capacitive sensors with respectto the reference accelerometers was based, initially,upon five classes of input signals: random, burstrandom, single frequency sine, swept sine, and burstsine. The frequency bandwidth was approximately 0.1 Hzto 200 Hz with various frequency amplitudes. It wasdetermined that three types of input, random, singlefrequency sine, and swept sine, were sufficient to evaluatethe accelerometers.

Results for accelerometer performance are summa-rized by the (complex) transfer function and by thecoherence between the response of the accelerometerunder test with respect to the reference accelerometer.These values are presented as a 3-D plot with excitationfrequency and excitation amplitude as two abscissaaxes. This generates a “surface of sensitivity” that char-acterizes the accelerometer. For clarity of presentation,these surfaces can be presented as color intensityimages. Regions of “good” or “bad” performance caneasily be picked out over the range of input excitationfrequency and amplitude. Figure 2 shows an image ofthe real part of the transfer function for the Kistler8304B2 accelerometer. The data and images of theaccelerometers are stored as MATLAB files, allowingeasy access and portability.

Note that this evaluation procedure can be appliedto any type of accelerometer over any frequency band-width and amplitude, provided the vibration shakerequipment can deliver the appropriate excitation.Accelerometer performance can also be evaluated withrespect to actual pre-recorded field measurementsusing the vibration controller’s playback capability.

Figure 2. Performance image for test accelerometer (real part of thetransfer function).

log

10 (

G)

0

–10–1.2

–1.4

–1.6

–1.8

–2.0

–2.2

–2.4

–2.6

–2.8

–20

–30

–40

–50

–60

–70

–90

–80

200

roll-off

“bad” region

150100Frequency (Hz)

500


Wireless Network on DemandB. Henderer

The goal of this technology base work was to constructa wireless, low-power communications network thatwould maintain itself. The network would configureitself with the nodes available, add new nodes as theyarrived, and reroute paths to handle lost nodes. All ofthis was to be accomplished with no user intervention. Athree-node network was to demonstrate the basics ofthe design for future work. Commercial, off-the-shelfequipment was to be used to reduce costs.

The Wireless Network on Demand (WNOD) technol-ogy base project accomplished all the goals stated

above, plus extra features to further test the capabilitiesof the system (see figures). We purchased severalembedded 486 computers and installed Linux on them.We also purchased 802.11b wireless Ethernet cards inPC card format and configured them to operate withthe Linux environment on the embedded controller.

Software was developed that accomplished theself-configuring and self-healing aspects of thesystem. All the nodes have the networking softwareinstalled on them. The software was also loadedonto laptop computers running Linux to demon-strate larger network capabilities. This software hassuccessfully operated in five demonstrations and onefield deployment.

During pre-testing and the field deployment, rangetesting was conducted to determine feasibility of the

system. Any distance less than two miles will result inno loss of data. Power consumption is 1.25 W, and thesystem can run off a hand-sized battery for 1.5 days.

In addition to accomplishing the goals of the tech-nology base work, we demonstrated the ability of thecommunications node to control sensors. An analog-to-digital converter PC card was installed in the commu-nications node. During the field test, the node gathereddata from a ground sensor and reported it back to abase station for display.

Future enhancements of the system would solidify theunits as ready-to-use, with limited start-up requirements.A two-PC cardholder PCB is being built and needs thor-ough testing. This allows the node to operate the wire-less network and have space for sensor controls or extramemory. Testing with a larger network (10+ nodes)would help understand engineering requirements forbandwidth usage and network stability. Adding morepower-conserving technology to lower the powerconsumption to less than 1 W is also desirable.

Many projects have shown interest in the WNODtechnology. The Cooperative UAV Networks will beusing the communications nodes as the primary meansof data communications; the Ground Sensing Arraysproject can use them to gather data in real time toallow faster analysis; Integrated Optic CapillaryElectrophoresis (IOCE) plans to use them to operateequipment remotely; and Lasers Engineering Divisionhas shown interest in the project as a diagnostics tool.

Figure 2. Wireless network node with added 16-channel A/Dconverter loaded into a second PCMCIA card slot. The second cardslot adds versatility.

Figure 1. Wireless network node with 486 computer, wirelessEthernet, 10 Mb on-board flash memory for user applications. It iscapable of creating a self-configuring, self-healing wireless network.


Development of Multi-Phase Flow Modeling CapabilityT. Dunn, L. Daily, R. Couch

The motion of fluids plays a significant role in our physicalworld. Three states (liquid, gas, and plasma) of the fourstates of matter are classified as a fluid. Even solidsbehave as a fluid in many situations. Therefore, it is notsurprising that the physics describing fluid behavior playsa critical role in the analysis of many realistic engineeringproblems. The purpose of this project was to enhanceLLNL’s fluid dynamics capability in the area of multi-phase flow.

Although most flows can be approximated as a singlehomogeneous fluid, many problems of interest to

industry and government agencies are dominated bymore complex phenomena. Some complex flows maycontain multiple components; reside in multiplephases; involve chemical reactions; or be a combinationof all of the above.

Our code development activities built upon thecurrent tools available in the ALE3D multi-physicshydrocode. During the first year of this project, theprimary focus has been on discrete particulate transportand free-surface flows.

The particle-tracking module simulates problemsinvolving particulate transport within a carrier fluid.Dilute particle flow is assumed where local aerodynamicforces control the particle’s motion. Momentum istransferred from the particle to the fluid through thefluid’s continuum equations. Each particle is individuallytracked in a Lagrangian reference frame using an equa-tion of motion derived from Newton’s second law. The

equation accounts for the steady-state aerodynamicand gravity/buoyancy forces acting on the particle. Thetrajectories are computed with a generic algorithm, somore specific physics may be easily added to the modelthrough additional terms to the equation of motion.The parallel methodology of ALE3D provides fastcomputations for large problems with many particles.

Bounce models have been added to account forparticle-wall interaction. A hard-sphere model is usedfor the mechanical behavior associated with the collision.The interaction depends on the inertia of the particle.Kinetic energy loss due to friction and inelasticity areaccounted for. Figure 1 shows a snapshot of a valida-tion case where a large number of particles are releasedin a walled container.

Droplet dynamics has been the main thrust of thefree-surface modeling effort. The deformation andbreakup of droplets involves complex couplingbetween aerodynamic forces and the droplet response.The Arbitrary-Lagrangian-Eulerian (ALE) formulation inALE3D is ideally suited to capture the droplet motion. Asurface-tension model was added to ALE3D to simulatethe intermolecular forces within the droplet. The modelis based on minimizing the element energy around thesurface of the droplet. In addition, a mesh-relaxationscheme was added to ALE3D, such that the computa-tional grid surrounding the droplet is mapped to thedroplet motion, and a constant level of grid refinementnear the surface is obtained throughout the simulation.

Figure 2 shows the effects of surface tension on liquiddroplets traveling through a gas. Note that the dropletwith a small surface-tension coefficient (We = 3.5 x 108)exhibits much more deformation than the droplet witha large surface-tension coefficient (We = 3.5 x 102).

Figure 1. Computation of particle motion in an enclosed box withwall interaction. The particles bounce off the walls until all themomentum is dissipated.

Figure 2. The effects of Weber number on droplet deformation(Weber number is inversely proportional to surface tension).Snapshots at 5 instances of time are shown for each Weber number.

We = 3.5 × 108

We = 3.5 × 102


Figure 2. EVL after the initial hardware integration.Figure 1. Artist‘s conception of EVL before this project.

Engineering Visualization LabM. Loomis, R. M. Sharpe, K. Mish

Our goal was to establish a venue in which to highlightthe work of LLNL‘s Engineering personnel. We wouldaccomplish this by creating an appropriate space,assembling the necessary hardware and software, andproviding visualization expertise for the production andpresentation of high-quality audio/visual media. At theclose of the project‘s first year, the EngineeringVisualization Lab (EVL) has realized its initial goals.

As a presentation theater, the facility integrates alarge projection video screen and surround audio

system, with a number of media sources, includingcomputer display output, digital disc recorder, VCR,DVD, and a video network feed from LLTN. Presenterscan easily patch a laptop computer into the system oruse the resident hardware to display their material onthe large display.

The room provides seating for 10 to 15 people, andprovides an unclassified environment suitable for small-group collaborations, software demonstrations, orvideo presentations. High-end animation, compositing,and image processing software staffed by knowledgeableoperators provide tools and expertise.

The presentation system is based on a large-screen,rear-projection, video cube and a SGI Onyx 2computer, which is located in an adjacent machineroom to isolate noise and heat. These units were theninterfaced with a rack of conventional analog and digitalvideo equipment, and video routing hardware wasadded to manage signal flow. An Apple G4 computerwas added later and integrated into the system fordisplay on the large screen.

Production and authoring software allow thecreation of engineering video presentation material inanalog or digital formats. A suite of software providesthe cornerstone of this capability with an open-ended

and flexible 3-D animation solution. The resulting videocan be output as conventional VHS videotape, digitalformats such as Quicktime, or DVD media.

A list of key hardware and application softwareappear below. “Before” (artists conception) and “after”(end of initial hardware integration) images of the EVLare shown in Figures 1 and 2, respectively.Hardware:

Computer 1 SGI Onyx 2Computer 2 Macintosh G4Large Screen Display Clarity Visual Systems LionDigital disk recorder Accom WSD 2XtremeS-VHS video recorder Panasonic AG-1980Video Monitor Panasonic BT-H1390YDVD player Panasonic DVD-A120Surround sound Klipsch Synergy 6

speakersAV Receiver Denon VR-3300Video switcher

Software:3-D Animation Alias/Wavefront MayaCompositing, Image Alias/Wavefront Composer

ProcessingVideo Editing Apple Final Cut ProDVD Authoring Apple DVD Studio ProFormat Conversion Equilibrium DeBabelizerDigital Video Terran Media Cleaner

CompressionOur primary plans for the future are to generate

representative content to demonstrate the capabilities ofthe EVL, and to prepare materials to assist engineers toeffectively use the facility. A general distributed platformcommunication scenario is being developed; a finalproduct would animate the communication paths andintegrate simulation data. The goal is to turn largequantities of data into actual information that is readilyassimilated by a technical audience.

19

The objective of this project is to provide an improvedmodeling capability by incorporating an incompressibleflow module in ALE3D, a multi-physics hydrocode devel-oped at LLNL. This addition, coupled to its structural,heat transfer, and chemistry models, will push us intopreviously unattainable areas of computational modeling.

The need for robust multidisciplinary modeling toolsis increasing. ALE3D provides many of the physics

options required, but not all. The ALE3D code is at theforefront of multi-physics modeling.

Low Mach number flow is one area where modelenhancements are needed. To address this, anincompressible flow model has been implemented tohandle cases where compressible flow modeling isinefficient. With this capability and additionalimprovements in ALE3D algorithms, many program-matic modeling problems of immediate and futureconcern can be solved.

The implementation was completed in FY99.Moving towards the goal of multi-physics modeling,the heat transfer model was coupled to the flow duringFY00. Significant improvements in turbulence modelingcapability, model accuracy, and solution speed wereachieved in FY01.

The flow is solved for an Eulerian formulation of thetime-dependent incompressible Navier-Stokes equa-tions using a finite-element method. Coupling of theheat transfer model requires the addition of an advectionterm to the thermal transport equation and aBoussinesq term to the flow momentum equationrepresenting the buoyancy force. The temperature issolved independent of the flow equations.

The resulting matrices for the flow and heat transferequations are assembled using the Finite ElementInterface (FEI) developed by Sandia NationalLaboratories in collaboration with LLNL. The system ofequations is solved using the HYPRE parallel solver pack-age developed at LLNL’s Center for Advanced ScientificComputing. The library packages allow the use of manyadvanced iterative linear solvers and preconditionersdesigned for efficient matrix solutions on massivelyparallel computer systems.

Significant improvements in speed were achieved byimplementation of a pressure stabilization option; anA-conjugate acceleration (or minimum residual projec-tion) scheme; and a linearized implicit time integration

Enhanced Fluid Dynamics Capability and Multidisciplinary Coupling in ALE3DR. McCallen, T. Dunn, G. Laskowski

Figure 2. Simulation capturing the time-dependent vortex sheddingpast a circular cylinder.

FY01 • ETR • TechBase

method; and by allowing for the reuse of the staticmatrix in the FEI. As shown in Figure 1, the pressurestabilization adds coupling between neighboringelements. This improves the solver speed of conver-gence by reducing the number of iterations.Stabilization also allows the use of fast algebraic multi-grid solvers.

Accuracy improvements were recognizedthrough algorithm reformulation that providedflexibility for the addition of full spatial integrationof the matrix coefficients.

The turbulence modeling capability was enhancedwith the addition of a one-equation Reynolds-averagedNavier-Stokes (RANS) model called the Spallart-Allmaras model.

These code enhancements were validated by simu-lating several flow geometries that have characteristicsof common engineering flow applications. For example,the code was found to accurately capture the separating,time-dependent flow past a circular cylinder as shownin Figure 2.

Figure 1. Simulated pressures for a lid driven cavity (a) withoutstabilization and (b) with stabilization.


Planning Tool for Site Remediation SimulationJ. Stewart, R. Glaser, A. Lamont, A. Sicherman, T. Hinkling

Site Remediation Simulation (SRS) is a computer-baseddecision aid and predictive model for site remediation.This management tool is designed to assist sitemanagers and DOE HQ with environmental planningand budgeting. It is designed to optimize budget alloca-tions, level of expertise, technological investment, andconsequences of low-probability, unexpected events.

SRS selects the optimal remediation strategy for agiven set of constraints. It specifies when to start

treating an area, the type of action to take (e.g., pumpand treat, excavate, investigate) and the intensity of theapplication (e.g., how much to excavate, the rate ofpumping, the number of wells). Users of SRS canchange treatment plans, budget, and regulations todetermine other possible scenarios.

The SRS model has been adapted for network use toallow remote sites to operate the model. Each site canbe given access to another site’s database and operateSRS. The SRS tool facilitates budget negotiations usingtransparent data and assumptions for the first time inenvironmental management. It also reduces theresponse time to a DOE request for a clean-up plan,from days to minutes.

This technology can be used in any long-term clean-up such as a Superfund site. The model (see figure) hasbeen or will be presented to the following: 1) DoDenvironmental conference in May 2002; 2) EPA confer-ence in May 2002; 3) Japanese chemical weaponsexcavation and clean-up project managers, January2002 and Summer 2002; and 4) PetroleumEnvironmental Research Forum, February 2002.

Cost

Utility

Remediation Site

Budgets

Plans Improved Plan

Evaluationof Plan

Outcomes (uncertain over time)($, concentrations, cleanup time,

regulatory compliance,stakeholder concerns, etc.)

Physical Sub-Model • Plume concentration, extent, migration

Cost Sub-Model • Short-term and long-term budgets

Planning Sub-Model • Regulatory and technology selections, expertise levels, research and development

Valuation Sub-Model • Multi-attribute utility model

Estimates outcomes based on actions • Outcomes are probabilistic and multi-attribute

Consolidates this information and evaluates plans for decision maker • Satisfies constraints • Maximizes expected utility of outcomes • Accommodates changed conditions

Allows user to test impacts of assumptions • Uncertainties about the state of nature • Budget levels • Technology characteristics

Components of Model Capabilities of Model

OU2OU1 OU3

plumes

SRS assists decision makers with optimizing complex decisions.


Coupled Engineering Simulation Tools: Solids, Fluids, and ChemistryC. Hoover, A. Shapiro, R. Ferencz

The Methods Development and Thermal Fluids Groupscollaborate on code development for the LLNL finite-element computer programs. The codes model nonlin-ear thermal and mechanical system responses. Inrecent years, the LLNL massively-parallel ASCI comput-ers and high-speed workstation clusters have providedthe opportunity to extend this suite of codes, NIKE,DYNA, and TOPAZ, with a new generation ofcomputer programs.

Our extended software incorporates multidisciplinarycoupling of solid and fluid mechanics, thermo-

dynamics with chemical kinetics, and transport algo-rithms. The strong algorithm development effort in ourgroups, supported mostly by the Center forComputational Engineering and technology basefunding, focuses on contact interface development,element technology, material models, coupledfluid/solid mechanics algorithms, and chemical kinetics.These activities provide the underlying state-of-the-artmodeling capability for the coupled-analysis simulationsof the future.

The next four articles in this report detail FY01progress on these code development activities. The fiftharticle provides an update on the ParaDyn program,the parallel version of DYNA3D.

The simulation of dams, bridges, and other structuresundergoing loads from earthquakes and blasts hasbeen a Laboratory activity for more than ten years. Ofparticular interest in FY01 were the energy absorptionand structural damage predictions resulting from watersloshing against a dam. Significant algorithm develop-ment for both NIKE3D and DYNA3D, in collaboration

with the Center for Complex Distributed Systems, gavebetter predictions for the low-frequency deformationmodes of dams.

Chemical reaction kinetics and surface-ablationmodels have been developed to simulate fuel cells andreentry vehicles. These models are being coupled withfinite-volume micro-scale transport models to simulatefuel cells. The TOPAZ3D program has been modified tomodel surface chemistry by adding finite-volume trans-port algorithms. These calculations support devicesdesigned and built at the Center for Microtechnology.

The LLNL ASCI program supports our coupledmechanics simulations and parallel algorithm develop-ment. Massively-parallel calculations with our production-explicit finite-element program, ParaDyn, can now usea thousand processors with problem sizes up to tenmillion elements. Representative applications includebuilding designs and retrofits, blast-structure interactions,and full-system weapons simulations with detailedmodeling added as needed.

In FY01 we provided DYNA3D/ParaDyn executablesto our DoD collaborators. These programs were used tosimulate homeland security and other defense-relatedproblems. The challenge to develop massively-parallelnonlinear implicit solid mechanics algorithms is beingaddressed in our new scalable implicit program, Diablo.This code will grow to encompass simulation ofcoupled physical phenomena. For example, the surfaceablation model prototyped in TOPAZ3D will beincluded in Diablo. Future capabilities of Diablo willsupport modeling of the thermomechanical response ofweapons and reentry vehicle dynamics, manufacturing,and other processes of Engineering interest.


New Surface Loads and Display Capabilities in DYNA3DJ. Lin

In structural/continuum mechanics problems, a distrib-uted force over a surface area is one of the mostcommon boundary conditions. DYNA3D users candefine a surface by a collection of four-noded quadrilat-eral segments and associate this surface with a specificforce function. In the past, the users had two choicesfor the force functions: a user-input time-pressure tableor pressure generated by a user-defined explosive mate-rial. In FY00 a third choice, a pressure load modulepredicting the resistance of soil and underground struc-tures, was added in DYNA3D. The force functionchoices were further expanded in FY01 to include pres-sures from hydrostatic, hydrodynamic, and closed-volume gas law calculations. Traditionally, surfaceloads can be applied only in the direction of the normalto the surface. In the latest version of DYNA3D, surfaceloads are applicable in directions tangent to the surface.

The hydrostatic and hydrodynamic loading optionsare for fluid-imposed pressure. The hydrodynamic

pressure is based on the classical Westergaard formulafor dams with pressure loads arising from groundmotion. The user must specify the locations of the dambase and the free surface to have this feature properlyactivated.

The closed-volume gas pressure load is designed tosimulate the pressure variation caused by the volumechange of air/gas-filled chambers in a structure system.The possible applications of this feature include thesimulation of airbag, piston, gun barrel, and any othercompressible fluid-filled enclosure in a structure.

The user must provide a collection of quadrilateralsegments that forms a closed volume. DYNA3D calculates

the current volume of the enclosure, based on theDivergence Theorem, and in turn the pressure accordingto the user-specified gas law. By properly selecting theprojection plane for the Divergence Theorem calcula-tion, the algorithm can also accommodate enclosuresbounded by boundary surfaces such as fixed boundaryplane or plane of symmetry.

Instead of limiting the pressure forces always actingnormal to the surface, the latest DYNA3D gives theusers the flexibility to apply surface traction in a tangentialdirection as well. The tangential direction can bedefined in a number of ways that are detailed in theDYNA3D User Manual.

With the expanded choices for the surface loadsources, the need to be able to display the surface loaddistribution becomes evident. In the past year, weadded the capability, upon users’ requests, to havesurface loads and contact forces on sliding interfacesincluded in the DYNA3D output database. During apost-processing session, a user can display either all thesurface loads and contact forces, or individual loadfunctions and sliding interfaces separately. The nodaltemperature for temperature-dependent analysis andnodal rotational velocities and displacements were alsomade available for post-processing display.

In addition to the new surface loads and displaycapabilities, eight material models were either added orupgraded in FY01. Most of these enhancements are anintegration of LANL engineers’ work. As always in thepast and probably for the future years, the mainte-nance of DYNA3D and its manual, user support, andCollaborator Program Administration are ongoing.

This year, the author also refereed three technicalpapers for journal publication.


Advances in Implicit Finite-Element Algorithms in NIKE3DM. A. Puso

The highlights of the NIKE3D implicit algorithm devel-opment in FY01 include an infinite fluid boundarycondition, contact algorithms, new elements, and animproved hyper-visco-elasto-plastic material model.

AWestergaard added-fluid mass method was devel-oped for NIKE3D to simulate the effects of water

sloshing against the surface of a dam (see figure). Thismethod simulates an infinite fluid boundary conditionwith an added fluid mass discretized at the surface ofthe dam.

This added-mass algorithm can be used for both atime-domain analysis and an eigenvalue analysis. C. Noble performed an eigenanalysis with this newfeature using a model of the Morrow Point Dam andfound that the natural frequency of the dam was50% lower and agreed much better with experimen-tal measurements.

Other new features were added to NIKE3D to facilitatea static fluid force and gravity initialization coupled to adynamics simulation with DYNA3D/ParaDyn.

Contact algorithms1) A new contact algorithm has been implemented

with a proper Lagrange multiplier treatment ofthe contact surfaceconstraint. This methodrequires a variablenumber of equations inthe linear solver andintroduces non-zerodiagonal terms andindefinite matrices.Further generalization ofthe method will bepursued in future work.

2) The iterative augmentedLagrangian contact algo-rithm was modified toallow penalty factormagnification for each

augmentation step. This technique providestighter gap tolerance control for the interface.

3) The contact surface extension algorithm has beenimproved with better internal logic for choosingthe extension length and an input parameter tofurther influence the surface extensions. The newalgorithm prevents spurious contact detectionwhen the extension length is too long.

Element technology and material model development1) A new shell element was developed to replace the

default Hughes-Liu shell element. This shellelement eliminates the zero energy mode whichoccurs when point loads are applied.

2) Both a C1 beam element and a linear beamelement were implemented successfully. C1 beamformulation is often more efficient and robustthan C0 beam elements. In addition, the C1beam elements are consistently linearized andtherefore converge better. The linear beam isuseful for making comparisons to analytical solu-tions and for checking model implementation.

3) A hybrid-element version of the hyper-visco-elasto-plastic material model was developed. Theoriginal version was limited to the enhancedstrain element.

(b)(a)

Simulation of Morrow Point Dam free vibration: (a) fundamental, and (b) second modes. NIKE3Dwith added mass feature computes a fundamental frequency of 2.76 Hz, which compares well withthe 2.82 Hz computed by Tan and Chopra with a full numerical treatment of the water. Thefrequency was measured experimentally at 2.95 Hz with a symmetric excitation. This work is partof a study for the US Bureau of Reclamation performed through the Center for ComplexDistributed Systems.


Development of Computational Capability for MEMS-Based TechnologiesM. A. Havstad, J. D. Morse

Two distinct modeling capabilities have been added toTOPAZ3D in support of both micro- and macroscale fuelcell projects at LLNL. First, an overall reactor rate modelingcapability has been integrated into the finite-elementstructure of TOPAZ3D. Second, a detailed surface chemi-cal kinetic and thermal transport approach is beingimplemented for microscale modeling.

Our reactor rate work integrates design and thermalmanagement with concepts in LLNL’s

Microtechnology Center and Energy Directorate. A“plug flow reactor” approach has been mated toTOPAZ3D so that both critical chemical performanceparameters (such as fuel conversion efficiency) andthermal parameters (such as start-up or trimming heat)can be calculated consistently.

Our microscale modeling work uses our reentry-vehicle-based surface chemical kinetic effort, andadds detailed transport calculations using finite-volume methods.

A surface chemical kinetic model of a methanolprocessing microreactor on a chip was exercised, givena flow field solution. Device heat loss, reaction rates,and microheater power were computed. The plug flowreactor approach was also exercised with a similar reactorpacked with a porous fuel reforming catalyst. Devicefuel conversion efficiency and the evolution of the reac-tant and product molar flow rates down the length ofthe micro-channel were computed (Figure 1).

The plug flow reactor approach was also applied toa microscale device fabricated at the MicrotechnologyCenter. Projected thermal performance is shown inFigure 2. Measurements of exhaust gas compositionfor model validation are in progress. Once agreementis demonstrated, the concept can be scaled up andheat exchange between entering and exiting streamscan be optimized.

The detailed transport modeling approach has beenbuilt up from a number of capabilities recently addedto TOPAZ3D for other efforts (primarily automaticdetection of slidelines and mesh rezoning). From theseand other efforts it has been straightforward to add thegeometric aspects of an unstructured grid finite-volumeformulation. Thus the facet areas, neighbor detection,gradient operators, and other strictly geometric itemsare fully debugged.

Much of the required debugging of the coupledsolution of the momentum and mass conservationequations is also done.

Our general and flexible surface chemical kineticapproach has been exercised with six chemical kineticsystems. The coding and input formats allow rapidcomparison of the thermal consequences of varyingchemical systems. Though this work is motivated byDNT, our comparison of the six systems to data and toother published computations serves as a significantverification and validation effort, applicable to fuel celldevelopment as well.

35 × 10–7

30 × 10–7

25 × 10–7

20 × 10–7

15 × 10–7

10 × 10–7

5 × 10–7

00.090 0.03 0.050.01 0.07

Position (m)

Mo

lar

flo

w (

g-m

ole

/s)

H2OCO2

CH3OH

COH2

Figure 1. Product and reactant flow rates.

TemperatureMax: 582Min: 389 560

500

450

389

Figure 2. Heated microreactor thermal profile (methanol reformerfor fuel cell on a silicon wafer).


Visualization and Data Management Tools: GRIZ4, Mili and XMiliCSD. Speck, E. Pierce, L. Sanford

In FY01there were several major developments in thevisualization and data management tools.

GRIZThe GRIZ User Manual was completely rewritten to

document Version 4 changes. It is available online inPDF format on the Engineering network and alsothrough the MDG web page. Several images from thenew manual are reproduced below in Figures 1 and 2.GRIZ4 executables are now supported on theCompaq and IBM platforms at the LivermoreComputing Center and on Sun and SGI platforms onthe Engineering networks.

Feature enhancements implemented in GRIZ includethe following: desktop window management has beenenhanced to group and display sets of windows foreach current execution of GRIZ4; GRIZ menus wereenhanced to support multi-dimensional state variabletypes in the Mili I/O library (Figure 3); both RGB andJPEG image formats can be generated automatically foranimations; custom specification of colors used torender mesh edge lines, element outlines, and text andtime series plot curves can be selected with the setcolcommand; the outmm command from GRIZ2 wasported to GRIZ4; new commands include and excludewere implemented, which combine the existing func-tionality of vis/invis and enable/disable.

MiliThe XMiliCS utility was completed and the Mili

library has been updated to support IEEE big/littleendian format conversions and double precision.Automatic and targeted endian formats are available forMili databases. This multi-purpose application providescapabilities to support post-processing and restarts ofparallel analysis runs.

1.2

ymax

Crush Some Spheres

1.078711e0.000000eymin

01

8 × 10–1

6 × 10–1

4 × 10–1

2 × 10–1

2 × 10–4 4 × 10–4 6 × 10–4 8 × 10–40

0Time

EQ

PS

HEX 8020HEX 5428HEX 9627

Figure 1. Time history plot of data from Exodus II database, showingcursor coordinates tracking display.

Figure 3. GRIZ menus, illustrating the mapping of a multi-dimensional“vector array” state variable, from a demonstration Mili databaseinto GRIZ’s pulldown menus.

Figure 2. X Strain vs. Y Strain (top) and parametric X Stress vs. X Strain (bottom) for the same three elements.

0–5 × 10–3

0

5 × 10–3

2 × 10–5 4 × 10–5 6 × 10–5 8 × 10–5 1 × 10–4

Time

–4 × 10–3 –2 × 10–3 2 × 10–3 4 × 10–30

0

–2 × 105

–1 × 105

1 × 105

X S

tres

s

X Strain (infinitesimal)

Bricks 256Bricks 255Bricks 254


Y Strain X Strain



ParaDyn Update: Parallel Interface AlgorithmsA. De Groot, R. Sherwood, C. Hoover

ParaDyn is an implementation of DYNA3D onmassively-parallel, distributed-memory computers.DYNA3D/ParaDyn algorithm design is a team effortinvolving computational mechanics and parallel algo-rithm developers. Over the past two years DYNA3D/ParaDyn code developers have collaborated on efficientnew methods for modeling contact interfaces. Theyhave successfully designed algorithms for target appli-cations in which materials in the domain undergo largerelative motion.

The version 2.x release of ParaDyn represents a signif-icant advance in the parallel contact algorithm

capability. The highly optimized parallel contact tech-nique for allocating a full surface to a processor is nowcomplemented with parallel forms of the automaticcontact and erosion algorithms in DYNA3D. Analystscan use multiple instances of the automatic contactalgorithm to limit expensive contact searches, andthese algorithms can be mixed in with the full-surface-to-processor algorithms. Material and domain limitingselection in DYNA3D provide options for optimizingparallel performance.

The parallel implementation of the automaticcontact algorithms uses a separate domain decomposi-tion for contact surfaces, and includes vertex and edgeweights. Dynamic load balancing is implemented to

redistribute the contact surfaces over the processors.Cost functions are used to estimate the frequency forsorting contact surface nodes and redistributing thesurfaces over the processors. The initial mesh partitioningis also constrained to allocate nearby interfaces into asingle processor as nearly as possible.

ParaDyn calculations for full-system weapons modelswere routinely carried out on 700 to 1000 processorson the Los Alamos ASCI Blue Mountain system. DuringFY01 they accrued over 12 CPU years on their system,running ParaDyn simulations to perform sensitivitystudies and to generate system response surfaces.

ParaDyn has been used for building design andretrofit applications. The figure illustrates a timesequence of one of several building design simulationscarried out by our collaborators at the Army ResearchLaboratory and the Engineering Research andDevelopment Center. Related calculations model brickand mortar walls with tied-surfaces-with-failure andthen use the erosion algorithms to track the debris afterfailure occurs.

Other algorithm developments completed in FY01include the parallel implementation of the deformable-rigid material switching and the coupling of ParaDynanalyses with stress-data restart files. These develop-ments in DYNA3D/ParaDyn enable a new capability formodeling manufacturing assembly processes that willbe used in future weapons applications.

Time sequence illustrating the collapse of a CMU wall with support posts and a textile liner used in a retrofit design. The ParaDyn simulationfor this problem used the parallel automatic contact algorithm with material erosion. The problem size was roughly 500,000 elements.


Integrated Microsyringe Arrays for Chip-Scale Fluid ControlM. Maghribi, P. Krulevitch

We have developed microfabricated piston arrays capableof controlling fluids (pumping, metering, and valving) inmicrofluidic devices. Our approach integrates low-power, thermopneumatic sources with piston sealsenclosed within microchannels, analogous to microscalesyringes. When the pressure source is activated, themetal piston slides within the channel, pumping thefluid on the opposite side of the piston without allowingfluid to leak around the piston. The challenges of thisproject were fabricating the channels, integrating thepistons, and developing thermopneumatic sources.

Acrucial aspect to the development of the microsy-ringe array is fabricating channels with circular

cross-sectional geometry to minimize friction andprovide a uniform, leak-proof seal with a spherical orcylindrical piston. We developed an innovative tech-nique for rapidly fabricating circular, smooth channels inpolydimethylsiloxane (PDMS), as seen in Figures 1 and 2.

The process is capable of producing channels assmall as 30 µm in diameter, but our current microsy-ringes range from 254 to 508 µm in diameter, due topiston size. These microsyringes can be adapted tomeet desired volumetric needs from nanoliters tomicroliters. The flexible PDMS conforms to an over-sized spherical metal piston inserted into the channel,forming a leak-proof seal. This is the reverse ofconventional syringes, which use a rigid vessel andcompliant plunger.

Reservoirs that contain the actuation fluid weremolded along with the channels. The thermopneumaticactuation reservoir is filled with a high-coefficient-of-expansion fluid (Fluorinert 77 from 3M) that is activatedby a thin film resistive heater. The heaters are covalently

bonded to the PDMS containing the reservoirs andchannels, using an oxygen surface treatment. PDMSsurface treatment in an oxygen plasma produces ahydrophilic surface that is necessary for bonding andfluid loading.

The initial prototype of the microsyringe arrays isshown in Figure 1, along with a controlled time sequencemovement of the piston within the PDMS microchannel.Optimal channel diameter was found through multipleiterations to be ~60 to 65% of the piston diameter, toachieve a leak-proof seal without obstructing pistonmovement within the channel. The pressure require-ment to initiate piston movement was between 65 and100 psi, varying with microsyringe diameter.

These microsyringes offer the capability of truly inte-grated fluid handling for miniature and field-portablesystems. This project has resulted in the developmentof new techniques that are being used by others, andhas also generated a patent application (IL-10630)titled “Low Power Integrated Pumping and ValvingArrays for Microfluidic Systems.” This research has alsobeen presented at the Micro Total Analysis Systems2001 conference in Monterey.

Further funding is being pursued through anumber of NIH SBIR proposals, in collaboration withPhoenix Biosciences. All milestones for this projecthave been completed.

T=15s

Piston

T=5sT=0s

~254 µm diameterT=10s

Figure 1. SEM of Microsyringe and time sequence. Figure 2. Integrated microsyringe.

Side view

Desired reagent

Channels

Top view

Thermopneumaticchamber

PDMS

Piston

Microchipcontroller

Resistive heater


Mesoscale NIF and Omega Laser Targets for High-Energy-Density Experimental Sciencefrom NanofabricationR. Mariella, Jr.

Currently, the targets that are fabricated for High-Energy-Density Experimental Science (HEDES) physicsexperiments (performed on the Omega laser at theUniversity of Rochester) are hand-made and cannot befabricated, assembled, and characterized with theprecision, accuracy, and throughput that is required.This Engineering TechBase project investigates the useof a focused-ion-beam etching (FIBE) process to addressthe problem.

HEDES experiments play an important role in corrob-orating the improved physics codes that underlie

LLNL’s Stockpile Stewardship mission. A method ofmaterial removal and deposition is required that canproduce µm-size features on mm-size components in avariety of materials including copper, polyimide, CRFfoam, aerogel, and beryllium.

FIBE routinely demonstrates material removal on thesub-µm scale (see figures). The work described in thisreport is LLNL’s first attempt to evaluate the perfor-mance of FIBE at commercial/service houses.

In the long term, we believe that LLNL needs toconstruct a facility, designed from the ground up, tohouse this capability. Part of this process will be tradi-tional precision machining, but another part willinclude novel fabrication procedures such as FIBE toremove material with 75-nm precision and accuracy.The use of an ion beam enables a much longer workingdepth of focus that can be attained with light beams, aswell as finer features, in general.

We have identified and visited four FIBE contractorswhere we have taken parts to machine.

To evaluate the applicability of FIBE to LLNL problems,we prepared a number of chips with silicon nitride onsilicon for etch studies, as well as polished naturaldiamond tips and EUVL optics (Mo/Si nanolayers onfused quartz).

A few observations became immediately evident,consistent at all four facilities:

1. The manipulation stages are designed to supportonly the silicon-integrated-circuits industry, i.e.,they have precise x-y motion only.

2. All four facilities handled all samples in unfilteredair, and surfaces quickly became contaminatedwith dust.

3. No facility had SEM on the same chamber asthe FIB.

4. On the positive side, sub-10-nm beam diametersare available, and remarkable precision can beachieved in the etching.

5. Specialized 3-D deposition of platinum or othermetals is possible.

Figure 1. Our first attempt to use FIBE to produce a cutting edge in adiamond tip. This would be used as a precision bit for an end mill.

Figure 2. SEM micrographs of pinholes etched into the EUVL optics.

Cutting edge

70 µm


Microstereolithography for Fabrication of Mesoscale Structures with Microscale FeaturesV. Malba, A. F. Bernhardt, C. D. Harvey, L. Evans

High-energy-density experiments performed with thenew NIF facility will play an important role in corrobo-rating the improved physics codes that underlie LLNL’sStockpile Stewardship mission. These experiments willrequire radical improvements in material removal anddeposition methods capable of fabricating fine featureson millimeter-size components. We have examined theapplicability of microstereolithography (µSL) to themanufacture of millimeter-sized weapons physicstargets for NIF.

Stereolithography is a rapid prototyping techniqueused commercially to produce macroscale 3-D

polymer structures by laser exposure of a photopolymerizable liquid resin in successive layers. Currenttechnology uses a laser focused to a few hundredmicrons. The size limitation of the technique resultsfrom the laser wavelength, the quality and characteristicsof the focusing objective, the dynamic control of laserfluence, and the accuracy and repeatability of themechanical stages.

Our project plan was to convert an existing laserpantography system to a µSL system by the addition ofa z-elevator stage, the retrofit of optics components for351-nm light, the reduction of the laser spot size, andthe repair of the laser.

Of the photoresin materials we evaluated, the best(1,6-hexanediol diacrylate with 4wt.% benzoin ethylether as sensitizer) was used to produce a resolutionpattern with 10-µm lines and 10-µm spaces (Figure 1),the limit of resolution of optics.

Multilayer structures were also fabricated (Figure 2),and considerable effort was directed toward reducingthe z-axis layer thickness. The minimum layer thicknessachieved was 10 µm. We are currently experimentingwith 1- to 3-µm lines and spaces with an ad hoc,replacement arrangement.

The smallest feature size that can be formed with aphotopolymerization system is the product of thediffraction limit of the optics times the solidificationfactor of the resin. The smallest spot size is given by thediffraction limit (d = 2.44 λ/F/#, where λ is 351 nm).For a “fast” objective (F/# = 1), the smallest spot sizewould be 856 nm.

The solidification factor is the fraction of the exposedvolume that is actually solidified. The ultimate resolutionof a system with a 0.5 numerical aperture (NA) objectiveis 856 × 0.8 = 685 nm.

An objective with a NA of 1.0 could theoreticallyproduce a polymerized voxel 343 nm in diameter.However, UV optics that perform near the diffractionlimit are almost impossible to find, and are difficult touse because of the short focal length.

The resolution of a photopolymerization system isalso dependent upon voxel overlap. Even the mostcarefully controlled motion equipment is hard pressedto provide seamless stitching of voxels. Seams result inunintended surface features that limit the surface finishof the part.

The present system handles the drawing of allfeatures as rectangles, which leads to stair-step linesand jagged-edged circles, which result in features thatare out-of-spec dimensionally. We have purchasedmodern DSP-based motion controllers, to be configuredin FY02.

Figure 1. Resolution pattern. The multilayerpyramidal structure clearly shows overlap seams.

Figure 2. Four-by-four arrays of (a) high- and (b) low-aspect ratio multilayer structures.

(a) (b)


Optical Coating TechnologyD. Sanders, J. Wolfe

In earlier work, sponsored by NASA, we were able todemonstrate a durable silver coating with an averagereflectance of greater than 94%, from 280 nm togreater than 2500 nm. Due to equipment limitations,we were limited in our ability to coat parts larger than6 in. in diameter. The focus of the current effort was amodified coating process that would allow us to coatlarger mirrors with the same formulation, without theneed for significant capital equipment expenditure. Wewere able to demonstrate this new capability bysuccessfully coating a 22-in.-diameter optic under asubsequent contract from the Keck Observatory.

To obtain adequate coating uniformity without the useof uniformity masking, we used a computer model

that allowed us to site our deposition sources in thecorrect locations with respect to the part to be coated.

In addition, we introduced an ion source to enhancedeposition energy at the substrate surface. Such anenhancement was necessary to compensate for thereduction in energy of the deposition species thatresulted from the increase in source-to-substratedistance. We found we needed to increase the source-to-

substrate distance to allow us to achieve coating unifor-mity and adequate optical properties simultaneously.

The layout of this successful configuration is shownin Figure 1. The system consists of three 6-in.-diametersputter sources, an electron beam evaporation source,and a linear ion gun.

After making the modifications, we carried out aseries of experiments to convince ourselves that wecould reliably repeat our coating process on a 21-in.-diameter part. Once we satisfied ourselves that wecould meet both uniformity and optical specifications,we accepted the mirror from Keck Observatory andcoated it, along with two witness samples mounted atthe perimeter of the substrate fixture.

As can be seen in Figure 2, which represents theactual reflection spectra measured on these witnesssamples prepared at the same time as the Keck mirrorwas coated, our approach for coating a relatively largepart with small sources proved to be successful.

We conclude from this experience that it is possibleto coat larger parts with smaller sources if themagnetron-to-substrate is increased, and a suitable ionsource is used to supply an appropriate energetic envi-ronment at the substrate surface.

Figure 1. Photograph of successful single rotation layout with Keckpart in place.

Figure 2. Spectra of witness samples prepared during the depositionof the Keck mirror.

100

90

92

94

96

98

12001000800600400Wavelength (nm)

Final Keck Uniformity Test Across 22 In.

5.2% Run off across 22 in.

EdgeMiddleCenter

757 nm723 nm717 nm%

Ref

lect

ion


Optical Pressure SensorM. D. Pocha, R. R. Miles, G. Meyer, T. C. Bond

The purpose of this technology-base project was toexplore the feasibility of a pressure sensor, based on thecoupling or modulation of light in optical waveguidesembedded in a plastic film.

The problem to be solved is to develop an array ofsensors that can be placed between two hard

objects under load, and measure the pressure distribu-tion without significantly changing the existing geom-etry of the objects and their mounting frame. Theseobjects are nominally 25 to 30 cm on a side, or 25 to30 cm in diameter.

The idea for this project was to embed wave-guides in Mylar and use changes in index or dimen-sion to measure the effect of pressure on the phase,amplitude, or other parameters of light propagatingin the waveguides.

A series of theoretical and experimental tasks wascarried out to explore the original concept as well asnew ideas. What is needed is a linear, repeatable sensorthat is imbedded in the polymer and will measure itsbehavior in terms of pressure redistribution over time

without affecting that behavior. We have found apotential solution in silicon-based pressure sensors.

Figure 1 shows the overall geometry for two 6 µm ×6 µm waveguides spaced 4 µm apart; Figure 2 showshow the sensitivity of the spacing changes as the gapbetween the waveguides changes. We used a compact,and extremely sensitive read-out scheme with a Fizeauinterferometer to read out a Fabry-Perot cavity gap toan accuracy of +/– 1 to 2 nm.

An experiment was designed to measure the gapchange of a thin film of plastic on a glass substrate. Alarge number of measurements were made over severalweeks. A typical measurement result is shown in Figure 3.

One outcome of our study was that a more sensitivesensor material would be beneficial. Several of ourconcepts could be developed into a measurementsystem, but they all suffer from the nonlinearities of theresponse of plastic materials to pressure. These sensors,if calibrated, can be used for short-term measurements.

While the initial work looks promising, there is agreat deal of additional work to be done.

Horizontal direction (µm)–10

–10

0

0

10

10

Contour map of transverse index profile at Z = 0

Ver

tica

l dir

ecti

on

(µm

)

6 µm × 6 µmnclad = 1.5∆n = 0.0055λ = 0.8 µmα = 1.1 j 10–7

gap = 4 µm

Length (µm)2000

0.0

0.2

0.4

0.6

0.8

1.0

4000 6000 8000 10000No

rmal

ized

wav

egu

ide

po

wer

Launching fiber mode in the bottom waveguide (blue)

Figure 1. Overall geometry for two 6-µm-×-6-µm waveguidesspaced 4 µm apart.

10.0

0.2

0.4

0.6

0.8

1.0

2 3 4 5Gap (µm)

Sen

siti

vity

(µm

)

w = 6 µm, l = 1 cm, λ = 0.8 µm, ∆n = 0.0055

Steps =0.01 µm

Figure 2. Effect of reducing gap.

Cured polyimide

Gap

(n

m)

P (psi)0 5000

9600

9700

9800

9900

10000

10000 15000 20000 25000

Gap_load (nm)Gap_unload (nm)

Figure 3. Typical polyimide thickness vs. pressure measurement.


We have worked on device designs and processing tech-niques that can be used to fabricate high-voltagearrays of photovoltaic diodes in gallium-arsenide(GaAs) materials. Of particular interest to us is the needfor low-resistance electrical contacts for both N-typeand P-type GaAs.

Many unusual processing problems had to be over-come this year to make progress towards our

technical objectives. The overall device fabrication requires that our

diodes be isolated as physical mesas to sustain thegeneration of more than 1000 V. Using our sputteredmetal deposition process (which should be able tobridge across the projections on the sidewalls of ourGaAs etched mesa structures) required the develop-ment of a metal etch process to remove the metal fromthe unwanted areas. The metalization we had chosenwas a variation of the Ti-Pt-Au metal used successfullyon GaAs in the past. Our variation was to remove the Ptlayer because there is no simple method to etch Ptwithout attacking the other layers.

We chose buffered HF for the etchant, despite thefact that it will slowly etch our oxide layer, because ofits high etch rate for Ti and zero effect on the GaAs orAlGaAs layers. We completed a test wafer using theSpire, Inc., device designs with the sputtered metalprocess. Although we achieved good electrical continuityacross the sidewall protrusions, the electrical perfor-mance of the devices and arrays was disappointing.

We also pursued alternate mesa etch processes,hoping to eliminate the sidewall protrusions and enablea return to the lift-off metalization technique. The mostunusual effect we observed is that the age of the citric-water solution greatly affects the etching characteristicsof the 30% AlGaAs. We have determined empiricallythat 26-day-old citric-water solution provides thesmoothest sidewalls.

We were also able to show that agitation of the etchsolution while etching produces more protrusions onthe sidewalls. We believe that this is due to the differentetch rates of different crystal planes. We observed thatthe protrusions seem to be faceted and correspondedto known crystal planes in GaAs.

We also learned that while 80% AlGaAs workedvery well as an etch stop and could be easily removedin hot, concentrated HCl, 30% AlGaAs did notprovide as much selectivity and was not as easily

Process Development for High-Voltage Photovoltaic ArraysG. Cooper, N. Raley, T. Graff

removed. It was around this point that it became clearthat there were fundamental problems with the Spire,Inc., array design.

We developed test chips, the most important contri-bution of which was to verify a fundamentally differentand easier-to-fabricate diode design. Diodes fabricatedusing the new design showed very good light and darkcurrent-versus-voltage (IV) characteristics. Actual N-typecontact IV data from the test chip was incorporatedinto a diode simulation model that produced diode IVcurves that very closely matched the IV curves wemeasured on the test chips (see figure).

Although we feel we can improve the N-typecontact and eliminate the “knee” in the IV curve, webelieve the existing characteristics are sufficient to meetthe demands of our near-term objectives. We did,however, examine the characteristics of AuGeNicontact metalization in case it was required to reducethe N-contact resistance. The N-type resistance withthis metal system was four orders of magnitude lowerthan that achieved with CrAu, as we expected, butthere were adhesion problems on the oxide layer.

Using the data obtained from the test chips and thediode model, we were able to design a new array ofdiodes, currently being fabricated into testable circuits.

Future plans include an examination of 1) AuGeNimetal adhesion on oxide; 2) why citric-water mixtureschange over time; 3) the behavior of sulfuric-basedetchants when not agitated; 4) maximum doping levelsachievable in the LLNL MOCVD growth system; 5) other etch stop layers; and 6) the 80% AlGaAs –GaAs interface.

Comparison of IV data measured on a single test diode under laserillumination with a mathematical model that incorporates real,measured, nonlinear N-contact data.

0.0004

0.0003

0.0002

0.0001

0

–0.0001

–0.0002

Cu

rren

t (A

)

2.00.5

Measured diode

Model with measured contact data

0.0 1.51.0V


Rapid Fabrication of Microfluidic Devices by Replica Molding of Polydimethylsiloxane (PDMS)L. R. Brewer, K. Rose, O. Bakajin, P. Krulevitch

We originally proposed constructing multichannel lami-nar flow cells for single DNA molecule-protein work fromreplica molded polydimethlysiloxane (PDMS) because ofits many appealing properties, including ease of fabrica-tion, low fluorescence, and optical transparency.However, for our application, there were some unforeseendifficulties, and instead we pursued an alternateapproach that was previously demonstrated by Bakajin.

The multichannel flow devices fabricated for our orig-inal proposal were not structurally rigid enough to

maintain the sub-micron positional accuracy (along themicroscope optical axis) necessary for DNA forcemeasurements using optical trapping. In addition, wewere not able to fabricate a fluid interconnect that wasleak-tight under pressure.

For our alternate approach, a thin (<10 µm) layer ofPDMS was spun onto a microscope cover slide andactivated in an oxygen plasma. It was bonded to glassinto which channels had been wet-etched 40 µm deep.Modified Upchurch fittings, used as fluid interconnects,were attached to the glass using a sheet-like thermalepoxy (Ablestik). The resulting device was structurallyrigid and the fluid interconnects did not leak evenunder high pressure.

Perhaps the most exciting aspect of this technique isthat materials with different thermal expansion coeffi-cients can be bonded together because the bondingoccurs at room temperature. We constructed a three-channel flow cell, shown in the figure. Red and bluefood dyes were injected into the outer channels using adual syringe pump, and water was injected into themiddle channel using a second syringe pump. Therewas very little mixing between the channels even atrelatively low flow speeds (100 µL/h). We can noweasily fabricate many different configurations of flowcells using this technique.

Future plans include incorporating a micropipetteinto the flow cell for holding 1-µm beads attached toDNA molecules by applying suction. We anticipate thatit will be easy to insert the pipette into a channelmachined into the cell and backfill it with PDMS. Weplan to replace the syringe pumps with a pressuremanifold and high-resolution pressure regulator.Compressed nitrogen will move different liquidconstituents into their respective channels with exactlythe same force. This will allow for very stable operationof the cell.

A three-channel laminar-flow cell with red and blue dye flowing inthe outer channels and water in the inner channel. The depth of thedevice is 40 mm; the width of the individual channels before theyjoin is 1.5 mm.


Remote Hydrogen SensorD. R. Ciarlo

Preliminary results have demonstrated the feasibility ofa novel remote hydrogen sensor. The sensor makes useof a bi-morph structure consisting of a 1.4-µm-thickpalladium film on a 200-µm-thick strip of glass. In thepresence of hydrogen, a stress is chemically induced inthe palladium film, causing the glass to bend. Thisbending is measured remotely, either with optics orradiography.

The remote detection of hydrogen generated inside aclosed container is an important problem that

occurs often in scientific, industrial, and military appara-tus. Hydrogen is a by-product of the deterioration oforganic components and it needs to be detected atlevels of a few ppm and above. This problem can bereadily solved if a tube can be inserted to periodicallyextract gas for analysis. However, such openings are notalways permitted.

The hydrogen sensor in this work can detect hydrogeninside a closed container without any physical access orat worst with the use of non-electrical fiber optics.

The operating principle for this sensor relies on thebehavior of a bi-morph structure. One element of thebi-morph reacts with hydrogen; the other does not.The reaction causes a chemically induced strain, whichbows the bi-morph structure. This action is similar tothe behavior of a temperature-activated bi-morph. Thedifference is that for this sensor, the strain is chemically,not thermally, induced. The sensitivity of the sensor isdetermined by the selection of materials and by the sizeand shape of the elements. We decided to use palla-dium (Pd) for the reactive component since it is knownto readily absorb hydrogen to form palladium hydride(Pd2H).

One way to remotely read out this sensor is to useradiography. The resolution of radiography in a highcontrast environment is 10 to 20 µm. If the contrast islow, as it would be when looking through other material,the resolution is only 75 to 100 µm. This resolution canbe improved by using high contrast beads and/or byusing a set of bi-morphs, attached at their ends so thatgaps are measured.

During FY00 we performed a number of experi-ments to test the feasibility of this sensor on bi-morphsshown in the figure. We used type 0211 borosilicateglass for the unreactive element of the bi-morph. The

glass was 200 µm thick, 0.5 cm wide, and 4 cm long.For the reactive metal we chose 1.4-µm-thick Pd filmsbecause of the known reaction of Pd with hydrogen.Following the deposition, the Pd film was clearly in astate of tensile stress as could be seen from the directionof bow in the glass substrate. The radius of curvaturefor the bow was 0.85 m as measured with a laser scanner.The bi-morph was then placed in a 4% hydrogen innitrogen atmosphere at room temperature for onehour. Following this treatment the film was clearly in astate of compressive stress as could be determined fromthe direction of bow in the glass substrate. The radiusof curvature of the glass substrate was 1.18 m.

The stress in a thin film on a much thicker substratecan be calculated from the following equation:

where σ = stress in the thin film (MPa), ν = Poisson’sratio for glass, 0.17, E = Young’s Modulus for glass, 50 ×103 MPa, H = substrate thickness (m), R = radius ofcurvature, and T = film thickness.

Using this equation, the calculated tensile stress inthe Pd film after deposition is 337 MPa tensile.Following the one-hour treatment in 4% hydrogen, thestress in the Pd film was found to be 243 MPa,compressive. The center of the glass strip before andafter hydrogen treatment moved a total of 370 µm.

The above experiments demonstrate the feasibility ofthis remote hydrogen sensor. Further experiments areneeded to optimize the shape of the sensor and to cali-brate its response.

σ =(1 – ν)

1 EH2

6RT( )

Palladium-glass bi-morph strips for the measurement of hydrogen.


Distributed Processing Algorithms for Reconstruction and RenderingG. P. Roberson, P. C. Schaich, H. E. Jones, M. W. Kartz

The ability to acquire or generate large volume datasets is currently outrunning our ability to process,analyze, and visualize them. Good examples are thelarge-volume computed tomography (CT) systems beingdeveloped for DOE by the Enhanced SurveillanceCampaign (ESC). It is expected that in the next fewyears these systems will be acquiring data sets that areseveral terabytes in size, requiring hundreds of processormonths to handle. In the past, we have used expensivecustom hardware systems that implement graphics-pipeand shared-memory architectures that do not economi-cally scale up as our data sizes increase. Because of this,we have turned to distributed computing systems.

In the last year we have ported three of our heavilyused nondestructive evaluation (NDE) codes to a

distributed computing system. The small-distributedsystem that is used in this work was designed to besimilar to systems that have been developed by theAdvanced Strategic Computing Initiative (ASCI). Thegoal of this effort is to implement our codes on ascaled- down version of an ASCI type system (~1/300in size) so that we do not impede everyday operationsand we are guaranteed computer time for a processthat is interactive and iterative in nature.

We have implemented three codes on our distributedsystem and have run benchmarks to evaluate theirperformance. These include two reconstruction codesand one rendering code. The two reconstruction codesinclude a fan-beam algorithm called the ConvolutionBack Projection (CBP), and Feldkamp (FDK), an inexactcone-beam algorithm that reconstructs to a limitedcone angle. The rendering code is a volumetric ray-tracing algorithm called cell tracer, developed at LLNL.

The CBP algorithm is implemented by transmittingidentical code to all the processing nodes (broadcasting)and equally dividing and distributing the projectiondata. The memory requirement for each node is notsevere for this algorithm, as the image slices along thez-axis can be stored onto disk as they are reconstructed.After the reconstruction process, a portion of the volu-metric image can reside on each of the processingnodes. This is ideal for further processing of the 3-D

image if it is required. Implementing the CBP algorithmin this way provides code that runs very efficiently on adistributed system.

We can distribute the problem over a number ofprocessors equal to the number of projections. Thenumber of projections is typically on the order of 1000to 1500 and will reach 8000 in the near future.

The FDK algorithm is implemented by broadcastingthe algorithm and the projection data set to all theprocessing nodes. The image that is reconstructed isequally divided and distributed over the processingnodes. The projection data is broadcast to all theprocessing nodes, one projection at a time. The portionof the volume image that resides on each node isupdated by each projection as it arrives. Like the CBPalgorithm, the volume image is distributed for furtherprocessing after the reconstruction.

Memory requirements are higher for FDK than theCBP because the image portion that is being recon-structed must reside in memory to attain an idealprocessing speed. The reconstructed image volume canget very large; however, the image is distributed overall the processors and with an increasing number ofnodes, less memory is required per node. The FDKalgorithm also runs very efficiently on a distributedsystem. The reconstruction problem can be distributedover a number of processors equal to the number ofslices along the x-axis of the image. This number is typi-cally 1000 to 2000 but will soon be 4000 to 8000.

The cell-tracer algorithm is implemented by broad-casting the code and the volume image that is beingrendered. Different projections are produced at eachprocessing node. The projections are produced fordifferent angles around the volume image and representframes of a movie. In this implementation, the memoryrequirements are severe as the entire image volumeresides on each processor node. The algorithm wasimplemented in this way because of its simplicity,requiring less modification of the code.

We hope to implement a more practical renderingalgorithm in the future, where the image is distributedand a portion of each projection is produced by eachprocessing node, thereby reducing the memoryrequirements per node.


Enhancements in Infrared NDE TechniquesW. O. Miller

Infrared methods for nondestructive evaluation (NDE)are being enhanced as part of the overall LLNL Centerfor Nondestructive Characterization (NDC) effort. SonicIR is being pursued with both theoretical and practicalefforts. Sonic IR shows promise as a new IR NDE tech-nique that can detect flaws, such as tightly closed cracksthat are difficult to detect with other methods. We havesuccessfully evaluated Sonic IR on several materials andflaw types. Quantitative NDC is a new effort that willapply and correlate several NDE techniques, includingIR, to evaluate microstructural defects in ceramics.

NDE methods are necessary tools for many program-matic efforts, with applications such as parts certifi-

cation and materials characterization. IR NDE methodshave several unique advantages, including noninvasiveand noncontact inspection, portability, simplicity, andrelatively low cost. Further, certain material characteristicsand flaw types are more readily characterized by IRNDE than by other methods. Sonic IR is a new tech-nique for NDE that works by dynamically exciting thepart being tested with an acoustic probe that is in phys-ical contact with the part. Any resulting differentialmotion across a crack face creates heat by friction, anda traditional IR camera images the transient tempera-ture rise at the crack. We found that the method andequipment used for this study were generally effectivefor all except the smallest flaws.

The initial effort was to see if Sonic IR could detectsmall cracks in aluminum oxide. Three types ofaluminum oxide coupons with flaws were made. First,surface grinding produced a large field of surfacedamage of approximately 0.2-mm closed cracks.Second, a Vickers hardness test bit was pressed into the

material to produce a radial array of closed surfacecracks, each approximately 1 mm long. Third, notchedbeams were cracked by three-point bending, producingcracks approximately 4 mm long.

Flaws were imaged in all but the small surfaceground defects (Figures 1 and 2). The inability to detectthe flaws in the surface ground coupons may be due toboth inadequate spatial and temporal resolution of theIR system, and a poor match of input forcing betweenthe acoustic probe and the aluminum oxide parts.Additional effort to resolve this is underway.

Sonic IR tests were made on carbon compositetensile specimens that had been fatigued to failure. TheSonic IR images of the damage areas were clear, and inexcellent agreement with ultrasonic images.

New efforts in Sonic IR include studies to examinethe response to alternative input parameters (broad-band, sweep), and image processing for reconstructionof subsurface detail from surface thermal images.

Qualitative NDC is a new effort that will provideinformation on microstructural variations in ceramics. Amajor component of this effort will be the applicationand correlation of results from several NDE techniques.Passive and heated IR methods will be used to measurelocal thermal property variations in ceramics in supportof this effort.

In FY02 the project will include experiments in arange of acoustic input devices to rapidly reveal flaws ofmany types while preventing part damage. Signalprocessing of the thermal images will include efforts informal deconvolution to reveal 3-D subsurface detail.

The effort in quantitative NDC will include thepreliminary investigation of suitable IR techniques, witha focus on the traveling heat source method fromNASA Langley.

Figure 1. Sonic IR image of a 4-mm crack in a notched beam coupon. Figure 2. Sonic IR image of four 1-mm cracks in a Vickers coupon.


High-Precision Quantitative Tomography of Mesoscale TargetsW. Nederbragt, S. Lane, D. Schneberk, T. Barbee, J. L. Klingmann, R. Thigpen

High-Energy-Density Experiments play an importantrole in corroborating the improved physics codes thatunderlie LLNL’s Stockpile Stewardship mission.Conducting these experiments, whether on NIF oranother national facility such as Omega, will require notonly improvement in the diagnostics for measuring theexperiment, but also the fabrication and characteriza-tion of the target assemblies.

The characterization of the target assemblies is asimportant as their fabrication: just as the actual

target assembly is the input to the experiment, thecharacterization of the target assembly is the input tothe physics simulation. With this in mind, a radicalimprovement is needed in characterizing the targetassemblies that comprise millimeter-sized componentswith micrometer-sized features.

X-ray radiography/tomography is one technique thatholds promise for characterizing the internal structureof target assemblies. Sub-micrometer spatial resolutionhas been demonstrated using a synchrotron x-raysource. This approach ill accommodates our require-ments for a high throughput and low cost approach forcharacterizing production quantities of target assemblies.

This TechBase project is funding the construction ofa Kirkpatrick-Baez (K-B) microscope. This is a classic x-ray optics design commonly used for focusing x rays. Ituses two grazing-incidence mirrors. Each mirror focusesthe x rays in one direction. Although this microscopedoes not meet all of our goals, it has a proven trackrecord. It will be a chief resource as we continue toimprove our target characterization capabilities.

We are making one improvement over the standardK-B design: we are using multi-layer mirrors to improvethe throughput of the instrument.

During FY01, we completed the instrument design,fabricated or purchased all of the necessary hardware,and began assembling the system. Most of the yearwas dedicated to the optical design. Several differentvariations on the standard K-B design were consid-ered, including the use of extra mirrors and the use of

non-spherical surfaces. We also investigated theperformance-enhancing characteristics of apertureswhen used with the standard K-B microscope. Weconcluded that the standard K-B with optimized aper-ture placement was the best choice because of itssimplicity and good performance.

After choosing the optical design, we conducted asensitivity analysis of all of the instrument componentsto determine the effect of position errors on the instru-ment’s imaging performance. With this information itbecame apparent that we needed to have four degreesof freedom on the instrument to correct any misalign-ments created during manufacturing and assembly.The mechanical design (see figure) of the instrumentfollowed the completion of the sensitivity analysis.

The instrument components were fabricated andpurchased based on the finalized mechanical design.We now have the majority of the parts in our posses-sion, and we have started the assembly process.

In the beginning of FY02, we will complete theassembly of the instrument and begin using it.

Geometric model of the base of our Kirkpatrick-Baez microscope.The instrument should provide sub-micrometer resolution of anobject over a quarter-millimeter field of view.

X-ray source

Sample to beradiographed

High precisionbase plate

Sample holder

Scintillator

Mirror translation stage

Mirrorsupport block

Opticstable

CCDcamera

Aperture translation stage(apertures are inside stage)


Linear Array Computed TomographyK. Dolan, J. Fugina, J. Haskins, R. Perry, R. D. Rikard

The objective of this project was to improve our capabil-ities for fan beam computed tomography (CT) byadapting new x-ray sensitive, linear diode-array tech-nology to an existing CT platform. A linear array CTsystem was upgraded with improved digital linear arrayx-ray camera technology. The new system provides 80-µmpixel pitch with square pixel format, higher detectionefficiency, and extended operating range. The lineararray system uses collimated fan beam geometry toreduce x-ray scatter noise and increase quantitativeanalysis capabilities.

Use of fan beam geometry for CT has the advantageof greatly reducing scatter at the detector array to

improve signal-to-noise ratio and reduce quantumnoise. This is done by slit collimation to limit exposureof the object to just the region of interest.

The new system realizes a factor of 5 improvementin spatial resolution with a 5500-element array at apixel pitch of 80 µm (6 line sets/mm). It provides afactor of 100 improvement in detection efficiency,which more than compensates for the smaller pixelsize, and extends the operating range by a factor of 2 upto an x-ray energy of 450 keV. (External shielding isrequired above 225 keV.) Readout time per line is 1.0 ms,which is a factor of 4 improvement, even though thenew array has 5 times the number of elements.

Image integration time can be varied from 1 ms to4000 ms, which extends our image integration capabilityby a factor of 40. The system retains 12-bit dynamicrange (4096 gray levels), and 470-mm linear arraylength. The new system on the existing platform isshown in Figure 1.

New capabilities with the upgraded system are as follows:1) Pixel and Line Binning. Statistical uncertainties can be

decreased. Up to 16 pixels or 8000 lines can be combined. 2) Fast Digital Radiography. This mode can perform

single radiographs, multiple sequential radiographs atincremented object rotational orientation, or up to 3-DCT data scans.

3) Fast Sinogram. This mode can perform a singlesinogram at a desired location, multiple sinogramsseparated by selected Z-increment distance, or multiplesinograms for sequential Z-axis locations.

4) Increment Average Radiograph. This modeconstructs single radiographs or provides up to 3-D CTdata sets by obtaining averaged digital radiographs atas many projection angles as desired.

5) Increment Average Sinogram. This mode canperform a single averaged sinogram at a desired Z-axislocation, multiple averaged sinograms separated byselected Z-increment distances, or multiple averagedsinograms at sequential Z-axis locations corresponding toprojected Y-pixel dimension for 3-D CT reconstruction.

Examples of a digital radiograph and a CT recon-struction obtained with the upgraded system are givenin Figures 2 and 3.

Figure 1. LCAT linear array CT system showing linear array andstaging with test object at left, and x-ray tube at right on tubularframe stand.

Figure 2. Digital radiograph of line pair gauge (left) and traceanalysis (center and right).

Figure 3. CT reconstruction of cylindrical hole set phantom, fullimage (left) and enlargement (right).

10 5 2 lp/mm10 5 2 lp/mm

Hole Dia.

2.5 mm

1.8 mm

1.3 mm0.8 mm0.5 mm


Neutron Radiography Beam StopB. Rusnak

A commonly used method for producing 10- to 13-MeVneutrons uses a deuterium-deuterium (D-D) nuclearreaction: a 7- to 10-MeV beam of D ions from an accel-erator impinges on a small volume of high-pressure Dgas confined in a vessel. A very thin metal window onone side of the vessel allows the beam to shoot through,but still physically separates the D gas from the vacuumenvironment needed for the accelerator to run. Thisapproach has traditionally limited the beam intensity toa few µA of average current, which limits neutronproduction to 108 to 109 neutrons/s. To image larger,denser objects, we are building a high-power neutronimaging machine that uses a virtual aperture to replacethe thin window. We have also determined that using apressurized high-atomic-number gas beam stop is aviable approach for stopping an intense 7- to 10-MeVdeuteron beam.

Our imaging machine, shown in the figure, uses arotating aperture system — a series of six rotating

and stationary disks with small holes — to replace thethin beam window. The holes act as a dynamic shuttersystem that allows the beam to enter 2% of the time,and is effectively closed to seal off the gas area from thevacuum 98% of the time. While this system promises toincrease neutron production by a factor of 20 overwindowed systems, it also creates the problem ofhaving to dispose of the more intense beam after itgoes through the gas.

Traditional methods of stopping the beam on ametal plate are complicated by very high power densitiesthat would literally blow away a solid metal surface nomatter how well it was cooled. If the plate were slantedso the beam was at a grazing incidence, or if the platewere rotating, the resulting mechanical bulk of thedevice would create more scatteredneutrons that would degrade the overallimaging system resolution.

A way to address this problem is tostop the beam in a high-pressure, high-atomic-number noble gas like Xe or Ar.With this approach, there are severaladvantages: we stop the beam within8 to 10 in.; we are at or below theCoulomb barrier for the resultingnuclear collision, precluding by-productneutrons or radionuclides; the gas canhandle the heat load better than a solidmaterial; the gas is self-healing, since

the volume of gas that is heated by the beam andrarefied will re-equilibrate and cool on its own. Finally,having a noble gas contiguous to the high-pressure Dgas that is producing neutrons creates a situation wherethe produced neutrons are nearly monoenergetic, sincethe beam only transfers about 10% of its energy in theD to make neutrons.

Our objective was to evaluate this concept of using ahigh-pressure, high-atomic number gas for a high-power beam stop.

A data set was established for the anticipated gases,including D, Xe, and Ar. Using values for the density,the critical pressure and temperature, the thermalconductivity, and the heat capacity as a function ofpressure, we estimated the sound speed for the D flowinginto the high-pressure vacuum chamber when theapertures opened. Calculations showed if the first andsecond rotating apertures were separated by a suffi-cient distance, the advancing gas from the opening ofthe first aperture could not reach the second aperturebefore it closed, thereby greatly improving the sealingefficiency of the overall system.

Gas hydrodynamic simulations were started, usingthe ALE3D code to evaluate how the high-energy particlebeam would affect the neutral gas in both the D gascell and the Xe gas beam stop. Our work has led to abetter understanding of the energized beam focuschannel hydrodynamics. Implemented design changesbased on this work are expected to greatly improve theperformance of this system.

The main application for this technology is as abeam stop for fairly intense (1- to 20-mA), fairly low-energy (5- to 20-MeV) particle beams from pulsedaccelerators. This technology can also be used forattenuating extremely intense beams of light that comeoff accelerator-driven coherent light sources.

Mechanical design of the rotating aperture system, the D gas cell, and beam stop.

Low-pressure vacuumpumping stage

Xe gas beam stop

Deuteriumgas cell

Neutrons forradiograph

High-pressurepumping stage Rotating apertures

and rotor assembly

Statorassembly7-MeV

deuterium beam


Rapid High-Resolution Ultrasound TomographyJ. Kallman, E. Ashby, D. R. Ciarlo, G. Thomas

Transmission ultrasound imaging is performed by passingultrasound through an object of interest and mechani-cally scanning a point sensor to determine the sound fieldtransmitted. This process is slow, typically taking 20 minper acquired field. We have produced a sensor thatcurrently acquires the acoustic field over the 2-D sensoraperture in 20 s. This, and anticipated further increases inspeed, make feasible new uses for transmission ultrasoundin nondestructive evaluation and medical imaging. Thissensor technology will make possible applications such asnear-real-time inspection of parts as they come off assem-bly lines, and rapid transmission ultrasound tomographyfor breast cancer screening.

The objective of this project is to speed the acquisitionof transmission ultrasound data for tomography

through the use of a new kind of acoustic sensor. TheOptically Parallel Ultrasound Sensor (OPUS) images anacoustic pressure wave over an entire surface byconverting sound pressure into an optical modulation.

The key to this conversion is evanescent fieldcoupling. When light encounters the interface betweena slow medium (high index of refraction) and a fasterone (lower index of refraction), light is both reflectedand transmitted. The transmitted light is refracted asper Snell’s Law: n1sin(θ1) = n2sin(θ2), where n1 and n2are the slow and fast media indices respectively, θ1 isthe angle of incidence, and θ2 is the angle of refraction.As θ1 approaches the critical angle θc (where n1sin(θc)/n2 = 1), θ2 approaches 90° and the light that istransmitted decreases. Beyond the critical angle, all thelight is reflected.

There is, however, an evanescent field that extendsbeyond the slow medium and into the faster one. Ifanother piece of slow medium intercepts this field,some of the light tunnels across the gap and is trans-mitted, with a corresponding decrease in the reflectionfrom the original total internal reflection (TIR) interface.The amount of light transmitted is sensitively dependenton the gap size (by the time the gap is approximately1 wavelength, the evanescent field has dropped almostto zero).

We exploit this phenomenon by suspending a flexiblemembrane over the TIR surface and exposing theopposite surface of the membrane to an acousticmedium. An acoustic wave hitting the membrane willcause it to deflect, changing the gap between the

membrane and the TIR surface, and causing a changein the reflection. By illuminating the TIR surface with astrobe light and observing it with a video camera, wecan observe and record the membrane deflections.

We have applied LLNL’s expertise in computationalmodeling and microfabrication to the development ofthe OPUS.

In FY00 we built robust, low tension membranesthat produced 0.7-cm-×-0.7-cm sensors that cansurvive indefinitely; upgraded our camera to speed dataacquisition by a factor of 100; acquired multiple viewsof a 2.5-D phantom; and performed diffraction tomo-graphic reconstructions using these data.

In FY01 we were unable to scale up the aperture ona membrane-based sensor, and therefore were unableto complete our fabrication plans. We investigated aparaxial-wave-equation-based adjoint method forreconstruction with transmission data and were able touse previously acquired data to perform 2.5-D tomo-graphic reconstructions.

In FY02 we plan to examine a new method of fabri-cating the sensor that will not require suspending amembrane on the optical substrate. This methodshould be inherently scalable and produce a sensor thatis more sensitive and less expensive to produce. We alsoplan to produce a full 3-D adjoint reconstruction codeand to write an invention disclosure on the nonmem-brane based sensor design to update the patentawarded in FY00.

Two-and-a-half-dimensional diffraction-tomography reconstructionof phantom. The phantom consists of three pieces of monofilamentfishing line, each 0.5 mm in diameter, arranged at 120° intervalson a circle 1.5 mm in radius.

Volume image


Synchrotron Microtomography at LBNL's Advanced Light Source FacilityK. Dolan, D. Haupt, J. Kinney, D. Schneberk

A microtomography system for nondestructivemicrostructural characterization and internal dimen-sioning at small length scales will be installed on thededicated computed tomography (CT) beamline at theAdvanced Light Source (ALS) synchrotron facility atLBNL. We have adapted previously developed LLNLsoftware for the user interface to provide motioncontrol, data acquisition, data transfer, image analysis,and reconstruction.

The ALS CT beamline has two experimental stationsbeing built into shielded hutches, one for x-ray

microtomography and one for x-ray microscopy. Themicrotomography system consists of translation stag-ing, a lens-coupled camera-based imaging system, anddata acquisition and control systems. It was designed toobtain spatial resolution as small as 1 to 2 µm in themicrotomography hutch with parallel beam geometryat x-ray energies in the range 4 to 100 keV.

The camera-based imaging system is also appropri-ate for image acquisition in the x-ray microscopyhutch. For this application, the design is expected toprovide spatial resolution as small as 0.1 µm with x-rayenergies in the range 3 to 12 keV. When translationstages become available in the x-ray microscopyhutch, the imaging and data acquisition systems willbe available for microtomography at these smallerlength scales as well.

Figure 1. Microtomography system mounted on optical benchshowing 3-axis positioning system (front), scintillator mount andextension tube (center), and ring rotational camera mount (rear).

The microtomography imaging detector is based ona lens-coupled design consisting of a cadmium-tungstate scintillator, high-quality optics, and a CCDcamera. The scintillator is 50 mm diameter by 0.75 mmthick, and is selected from high-grade scintillator materialto be nearly flaw-free. The optics is a high-quality lenstested during a previous program and demonstrated toprovide 180 lp/mm. The cooled CCD is a 3-k-×-2-k chipwith sensitive elements 9 µm × 9 µm.

At nominal magnification of 1.0, the systemaccommodates objects up to 27 mm in diameter and100 mm vertical, and provides spatial resolution ofapproximately 20 µm. At the maximum optical magni-fication of 5.4 provided by the high-quality lens, thesystem accommodates objects up to 5 mm diameterand 20 mm vertical with spatial resolution of approxi-mately 3 µm. The horizontal and vertical mechanicalstage positioning and repositioning accuracies havebeen verified to be ±0.5 µm.

The microtomography system as assembled on anoptics table for the field work is shown in Figure 1.Examples of CT reconstructions from data taken atSSRL, where the system has been used successfully, areshown in Figure 2.

Significant improvements in image resolution andcontrast and extended capabilities are expected sincethe beam divergence and beam intensity at ALS areboth improved over SSRL data by nearly a factor of 10.

Three areas that have benefited from synchrotrontomography, and are in urgent need of the greaterprecision this beamline will allow are: 3-omega damagestudies, shock spallation studies, and NIF target charac-terization. Other areas will benefit from the monochro-matic radiation and high signal-to-noise provided bythe ALS synchrotron source, which is expected to comeon line in February, 2003.

Figure 2. (a) Tomogram of laser beam damage in silica, showinglocal densification near edges of cavity (white region). (b)Tomogram of composite bond test sample, showing bond depth(mid-tone region between light and dark regions) and microcracksin higher density (light colored) region.

(a) (b)


The primary goal of this project was to show that wecould take commercially available telecommunications,cell phone, and survey instrumentation technologiesand merge them to build a more accurate coordinatemeasurement machine. For this we needed to showthat it was possible to achieve better than 10-µmmeasurement resolution with better than 50-µm trans-verse resolution, and that it was possible to rapidly andaccurately scan the measurement beam across thetarget area to be measured.

Our design (Figure 1) is based on a microwave inter-ferometer. For this system we use an 8-GHz RF

source. The RF power is transmitted to a microwavesplitter, allowing for the two arms ofthe microwave interferometer. Onearm becomes the reference arm andthe other is the measurement armthat is used to measure the distanceto a target point.

Since we cannot focus themicrowaves down to a small pointon the target, we must use an opticalfrequency as the carrier for themicrowaves. We used a telecommu-nications laser diode at an eye-safe1550-nm wavelength. The RF powermodulates the optical beam using anintegral Mach-Zender modulator.The optical beam passes through anacousto-optical (AO) deflector,deflecting the beam to various pointson the target. An AO deflector waschosen to allow for very precisesteering of the beam under a widevariety of conditions while still beingvery repeatable.

The beam then returns to an opticaldetector that converts the modulatedoptical beam into RF power. Thephase of this returned measurement

Three-Dimensional ProfilometerM. W. Bowers, D. W. Swift, G. W. Johnson

Figure 2. A scan across three gauge blocks arranged in a stairstep pattern.

arm signal beam is compared with that of the referencebeam using a telecommunications “IQ” demodulator.The phase can be measured very precisely using thismethod. The difference in the number of RF phasecycles between the reference and the signal can bedetermined by changing to a slightly differentfrequency and using a simple linear equation. The“measurement resolution” can be very precise usingthis method.

An example of a scan across three gauge blocksarranged in a stairstep fashion is shown in Figure 2.

We are pursuing higher precision and accuracyduring the next fiscal year. We are also planning ontesting the system on several different applications thathave high importance to the missions of LLNL.

Referencearm

Microwavesplitter

8 GHzsine wave

AO deflector

Measurement arm

Lasertransmitter

Beam shaping optics

RF ramp generatorSine wave generator

Measurement headassembly

Computer

Auto-focus

Phasedetector

Low noise amplifier

Fast detector

BA

A B

Figure 1. Three-dimensional optical profilometer system showing the system components.

Scan beamError due to edgeof gauge block

2.286 mm(0.09 in.)

2.34 mm(0.0921 in.)

2.99 mm(0.1177 in.)

3.048 mm(0.12 in.)

24.689

24.688

24.687

24.686

24.685

24.682

24.6811500 100

Lateral position (arb units)20050 250

24.684

24.683

Dis

tan

ce (

m)


Engineering Mesoscale InitiativeD. Meeker, R. Mariella, Jr., H. Louis

The Engineering Directorate’s Mesoscale Initiative has asits goal the development of capabilities needed to fabri-cate target assemblies for the Defense and NuclearTechnologies High-Energy-Density Experimental Science(HEDES) Program. Many of these targets will be fieldedon the Omega and National Ignition Facility (NIF)lasers, and a number of these targets will require surfacefinishes in the 50-to 100-nm range, and surface featuresapproaching 1 µm. We are coordinating efforts toensure that the proper capabilities are being developed,and performing feasibility studies of new approaches toyield the needed capabilities.

The HEDES share of the NIF shot schedule amountsto 200 to 300 shots a year. Of that number, we esti-

mate that 80% to 90% of the targets can be producedwith extensions to existing capabilities; the MesoscaleInitiative is focused on the more difficult targets remain-ing. Therefore, a successful Mesoscale Initiative willresult in a capability that can produce this wide rangeof targets in sufficient numbers.

Targets could have planar, cylindrical, or sphericalgeometries, in materials that vary from very-low-densityfoams such as aerogels, to high-density, high-Z shellsserving as high-pressure fuel capsules. The targets couldhave 1-, 2-, or 3-D features on their surfaces, withfeature size in the µm to tens of µm range.

To reach that end, we are coordinating variousresearch and technology efforts to ensure that theproper capabilities are being developed, and to dofeasibility studies of new approaches to yield thesecapabilities. Finally, the Initiative is responsible forcoordinating Engineering’s efforts with those of theNIF Ignition Target Fabrication Group, and for keep-ing Engineering management aware of requirementsand progress.

Fabricating these targets is just part of the process.Once fabricated, the targets, which are usually madeup of several components, must be assembled andcharacterized. Thus, the Initiative has six elements:material synthesis; material removal; material deposi-tion; metrology of components; assembly; and charac-terization of the assembly. This year, the projectfocused on gathering requirements from the experi-mentalists and theoreticians and initiating severalLDRD and TechBase projects in the areas of materialremoval and characterization.

The gathering of requirements provided a vast rangeof possible target shapes, materials, and feature size. Asmany as twenty separate proposals for experiments onNIF and Omega are presently being considered, cover-ing areas as diverse as hydrodynamics (preheat, mix,instabilities); opacity; radiation transport; materialstrength; and equation-of-state studies.

Materials in these targets range from very lowdensity (~0.04 g/cc) plastics, with and without dopantor tracer elements embedded, to high-Z metal capsulesused as pressure vessels to contain DT fuel. Sometargets require surface finishes of the order of 50 nmrms, while others need 3-D surface features with wave-lengths of a few to tens of µm, and a few µm andlarger amplitudes. We are told these requirements verylikely will change in time, but we can get a prettygood estimate from the present needs. Thus, we willbe developing capabilities to machine a few µm tomany µm amplitude features with 50- to 100-nmfinishes in materials as varied as low density foams tohigh-Z metals.

In FY01, Engineering successfully proposed threeLDRD projects that are critical to the MesoscaleInitiative. The first LDRD project is a continuation of amid-year proposal to design and build an x-ray micro-scope capable of characterizing small (~1 cm) targetsto sub-µm resolution.

A second LDRD is studying the enhancement of afast servo tool to be added to the z axis of a diamondturning machine. By increasing the speed this tool canmove in and out, we can greatly enhance the materialremoval on complex, 3-D features.

The third LDRD is studying the material removalcharacteristics of an ultra-short-pulse laser. Here, experi-ments are coupled to modeling efforts to see if a deter-ministic behavior can be identified, and if these lasersare capable of delivering the surface finishes required.

During FY02, we are expanding our study to sub-µmaccuracy. We also expect to learn from our ultra-short-pulse laser LDRD project the capabilities of femtosecondlasers as tools to machine surfaces to required surfacefinishes. Our LDRD on the x-ray microscope should bethrough its design stage and ready for construction.The fast tool servo LDRD is expected to yield results inincreasing the repetition rate of the servo when matedto a diamond turning machine. We will also assess thefeasibility of adapting FIBEs to a multi-ion, variablecurrent machine that could be used to remove ordeposit material.


Fast Tool Servo Application to Single-Point Turning Weapons Physics TargetsR. C. Montesanti, D. L. Trumper (MIT), J. L. Klingmann

High-Energy-Density Experiments play an important rolein corroborating the improved physics codes that underlieLLNL’s Stockpile Stewardship mission. Conducting theseexperiments presently on Omega at Rochester, and in thefuture on NIF, requires improving the fabrication capabil-ity for the target assemblies. These targets are mm-sizedassemblies of components with µm-sized features. A faith-ful mapping of experiment to computer simulationrequires an ability to produce parts with a high surfacecontour accuracy and low surface roughness.

Single-point diamond turning is currently the bestmethod for a high material removal rate with non-

iterative accuracy and low surface roughness. Currentsingle-point turning can handle only special cases ofcontoured surfaces. New experiments for StockpileStewardship require surfaces that exceed LLNL’s currentsingle-point fabrication capability. The two-directionalsinusoidal surface shown in Figure 1 is one example.

Single-point turning requires the tool to accuratelyfollow a rapidly changing trajectory as the workpiecepasses by. A fast tool servo (FTS) is a small, fast-movingmachine axis that can be added to a single-point turningmachine to accomplish this.

The goal of this project is the improvement of a 100-Hz bandwidth commercial FTS, shown in Figure 2, thatLLNL purchased in FY00. Preliminary work at LLNL indi-cated that this FTS did not perform as anticipated, andwould severely limit the fabrication process by requiringextremely low spindle rotation speeds.

During FY01 the FTS was shipped to the PrecisionMotion Control Laboratory at MIT to become part ofthe work being conducting there. The initial assessmentat MIT was that the FTS controller is embedded withinthe proprietary circuit boards of the system, making itimpractical to directly modify the controller behavior.We decided instead to treat the entire system (mechan-ical and proprietary controller) as the plant (a dynamicsub-system), and plan to wrap a compensating looparound it, using a Simulink-based controller imple-mented on a dSpace controller card.

Preparation for this work was accomplished by a re-commissioning effort on a two-axis diamond turningmachine at MIT; the FTS will be integrated with this forcutting tests. Specifically, the electrical, electronic, andcontrol systems for the machine were modified andupdated to allow integration of a dSpace controlsystem, and safe operation of the machine.

During FY02 we will complete the following:1. Measure the step response of the FTS amplifiers;2. Measure the step response of FTS position feed-

back device;3. Characterize the open-loop response of the FTS;4. Measure the closed-loop response of the FTS;5. Design the outer loop controller using Simulink;6. Implement the controller using dSpace;7. Integrate the FTS with the MIT diamond turning

machine and produce a test part.

Figure 2. 100-Hz commercial fast tool servo owned by LLNL.

Figure 1. Example of a desired two-directional sinusoidal surface,70-µm spacing, 5-µm height.

50

50

0y (µm)

x (µm)

z (µm)

0

0–1–2

–5

–3–4

–50

–50


Frame Extraction and Image Processing in Scene AnalysisA. Gooden, L. Scott

This project has two phases: first, to determine an effec-tive and efficient method for frame extraction from anMPEG-2 video stream; second, to apply image processingtechniques to the frames for the enhancement of imagesfor artificial neural network processing.

This real-time scene analysis project consists of aseries of modules connecting several components,

including a smart camera system, frame extraction andimage processing software, and neural networks. Thecamera will capture video streams, which will bebroken into sets of frames that will be manipulated byimage processing algorithms. Neural networks will beadded to produce a system by which objects can beidentified from remote locations.

As the video is streamed over the network, softwarecaptures and decodes the stream, allows the stream tobe viewed on the viewer system, controls the cameraover the network, and can capture single frames fromthe stream or record a video loop instead. This softwareis upgradeable and can trigger alarms remotely, andwill be used to send video information to neuralnetwork systems, or to image processing modules. Thesoftware is one of the image acquirement, control, andtransmission components.

Part two of this system consists of components thatwill be useful in image processing applications. Thesesoftware packages are customized for use in file typeconversion, image compression, frame extraction,automation, filtering, and pattern recognition. Thelong-term goal is to create modules that will be usedin conjunction with existing software for advancedimage processing of files retrieved from part one—data acquisition.

It is proposed that this part of the system serve asthe mediator between the acquisition system and theneural network. This part will not only prepare theacquired images for the network, but also removedetected blockages and send audio extractions.

The design and implementation of neural networksis part three of this system. The neural networks willdecipher images captured by the smart camera andcategorize the predetermined objects within the imageframe. This is done by implementing learning algo-rithms on the neural networks.

A neural network is a collection of nodes thatemulates neurons in a human brain. Human learning

involves adjustments to the synaptic connectionsbetween neurons. This is also how the neural networkworks. Learning usually occurs by example throughtraining where some algorithm repeatedly adjusts theconnection weights. These connection weights storethe knowledge necessary to solve specific problems.

Today, neural networks are being applied to anincreasing number of complex problems in the world.They are good pattern recognition tools and very effec-tive categorizers. They have the ability to generalize inmaking decisions about inaccurate data, and offerhypothetical solutions to a variety of classification prob-lems such as speech, and character and signal recogni-tion. The advantage of neural networks lies in theirresilience against distortions in the input data and theircapability to learn.

The second phase of the project is to apply imageprocessing techniques. Image processing is the manip-ulation of images by restoring degradation, enhancingappearance, and extracting features. Three major typesof image processing techniques were applied: sharpeningfilters, edge detection, and object removal.

The input of this research was streaming MPEG-2video. Each video averages one minute. The videoswere broken into frames and put into sets. Therewere 41 sets; each set contained 1800 frames. Theseframes were used in Stuttgart Neural NetworkSimulator back propagation and clustering algo-rithms for training and testing the network. The colorsingle frames were converted to 8-b monochromebitmap images. The grayscale images were runthrough the mentioned algorithms. Each algorithmwas used to enhance the images.

More time is needed to teach the neural networkdifferent image patterns. This will allow one to seewhat type of image patterns the neural network canuse for better generalization. Time is also needed toproduce programs to give a statistical analysis of theneural network, as well as training the neural networkmore efficiently. It has been proven that it is feasiblefor the neural network to differentiate and identifyobjects using visual data. The system as a whole willcapture a scene using the CCD camera, extract frames,manipulate the data, and analyze the data using theneural networks.

Once the system is functional, it will provide LLNLwith a tool that can be used for surveillance or manyother applications.


Surface and Volumetric Flaw Distribution in Brittle MaterialsR. A. Riddle, C. K. Syn, S. Duffy, J. Palko, E. Baker (Connecticut Reserve Technologies)

Materials are called "brittle" when their failure involvesvery little plastic deformation and widely variable failurestrengths—glasses and ceramics are two very differentexamples. Critical applied stresses, i.e., those that causefailure, depend on the size of the flaws, the orientationof the flaws in the material with respect to the appliedloads, and the flaw growth resistance of the material.We have previously fabricated and tested a set of flexurespecimens taken from an oxide ceramic component. Thisyear we have microscopically examined the failuresurfaces to characterize the critical flaws.

When the size and orientation of the flaws incomponents are unknown, statistical methods are

used to predict the adequacy of structural componentsto survive or to have a certain lifetime under appliedloads. Statistical methods are often used with brittlematerials because the low fracture toughness of thematerials implies that the critical flaws are very small,and hence undetected. Critical flaw sizes for brittlematerials are often on the order of a few microns.

Weibull statistics is the most common statisticalmethod applied to the prediction of strength in brittlematerials, with two- and three-parameter Weibull

distributions used to correlate probability of failurewith stress. A differentiation is made, depending onwhether the flaws are distributed randomly through-out the volume of the material, or the defects arerestricted to surfaces.

With few exceptions, surface flaws produced bymachining or handling are the predominant source offailure in glass. In contrast, ceramic materials have flawdistributions characterized by the presence of bothvolumetric and surface flaws. When failure data isgrouped according to whether the specimen criticalflaw was on the surface or within the volume, it is saidto be censored.

The location of the critical flaw in a broken specimen,and whether it was a surface or volume flaw, must bedetermined by microscopic examination of the failuresurface. This year we have microscopically examinedthe failure surfaces to characterize the critical flaws offlexure specimens taken from an oxide ceramic compo-nent (Figure 1).

We have developed algorithms implemented in thecomputer program WeibPar to develop two-and three-parameter Weibull distributions to correlate to thecensored data. The values of the Weibull parametersderived are very sensitive to the details of the datapartitioning. (See Figure 2 for an example.)

The Weibull parameters thus derived are readily usedto predict the reliability of structures analyzed usingfinite-element analysis. In previous years, LLNL finite-element codes were modified to perform reliabilityanalysis on material with volumetrically distributedflaws. This year, both NIKE2D and NIKE3D were modifiedto offer reliability analysis for material with surface flawsas well.

Figure 2. Cumulative Weibull distribution line. The data set hasbeen censored according to flaws found on the surface of the speci-men and within the volume.

21

0

–1

–2

–3

–4

–5

–618 26 32 3422 282420 30

Fracture stress

LN

(L

N(1

/(1–

Pf)

)) SurfaceUnknown

Volume

Figure 1. Specimen from low-density white area. (a) Low-magnifica-tion micrograph indicating presence of oversized powder grains,shown as black particles. (b) Cross-sectional view, indicating presenceof internal porosity. (c) Large particles. Scratch marks on the leftparticle indicate that it was polished, while the surrounding smallparticles were crumbling during the polishing. (d) Enlarged view oftwo porosities in (b).

(a) (b)

(c) (d)


Nitrogen Augmentation Combustion SystemsL. E. Fischer, B. Anderson

This work combines multi-stage combustion technologywith nitrogen-enriched air technology to 1) control thecombustion temperature and products; 2) extend themaintenance and lifetime cycles of materials in contactwith combustion products; and 3) reduce pollutants,while maintaining relatively high combustion and thermalcycle efficiencies.

Increasing fuel efficiency in combustion engines, whilesimultaneously reducing polluting exhaust emissions,

has been researched over the past 25 years and subsi-dized by the federal government.

Maximum fuel efficiency normally occurs at or nearstoichiometric conditions, where the fuel is completelyoxidized. In practice, the combustion process in anengine is usually with air and not with pure oxygen.When oxygen is supplied by dry air, 3.76 moles ofnitrogen will accompany one mole of oxygen. Thestoichiometric air-fuel ratio is the ratio of the mass of airto the mass of fuel to result in stoichiometric combustion.The actual operating condition of an engine is usuallyexpressed in terms of the equivalence ratio, which isthe ratio of the stoichiometric air-fuel ratio to the actualair-fuel ratio. The equivalence ratio is 1.0 when theengine is operating at stoichiometric conditions.

When an engine operates at an equivalence ratiogreater than 1.0, it is operating fuel-rich and producespollutants such as hydrocarbons (HC), carbon monox-ide (CO) and particulate matter. At equivalence ratiosless than 1.0, the engine produces oxides of nitrogen(NOx), which are a major source of photochemicalsmog, and are regulated. Also the combustion gasescan be very corrosive with the excess oxygen, and canreduce the life of the engine.

This TechBase work combines multi-stage combustiontechnology with nitrogen-enriched air technology tocontrol the combustion temperature and products. Thefirst stage of combustion operates fuel-rich, where mostof the heat of combustion is released by burning it withnitrogen-enriched air. Part of the energy from the

combustion gases is used to perform work or toprovide heat. The cooled combustion gases arereheated by additional stages of combustion until thelast stage is at or near stoichiometric conditions, andadditional energy is extracted from each stage to resultin relatively high thermal cycle efficiency. The air isenriched with nitrogen using air separation technologiessuch as diffusion, permeable membrane, absorption,and cryogenics.

Our computer modeling has indicated that the useof multi-staged combustion coupled with nitrogen-enriched air could result in extremely low NOx pollutionlevels and still achieve relatively high combustion effi-ciency. Two white papers were written documentingthe computer modeling and results, one on gas turbineperformance and the other on coal burning systems. Inboth cases the NOx was reduced by two orders ofmagnitude and the free oxygen by two to three orders.

Potential industrial partners, such as OEMs for powerproduction, OEMs for NOx sensors and controlsystems, and the EPA have been contacted to gaugethe level of interest in this technology. Two OEMs forpower production—GE Turbines and ABB PowerSystems—along with Marathon Sensors and the EPAhave also indicated interest in this technology. Contactsare being developed with OEMs for nitrogen-enrichmentprocesses. A provisional patent on this new technologyhas been filed to allow open discussions with industry.

Staged Combustion with Nitrogen-Enriched Air(SCNEA) development was originally funded in FY01 asa technology base project. SCNEA is viewed as a possi-ble alternative combustion technology for stationarycombustion processes such as boilers, furnaces, and gasturbines. SCNEA can be implemented relatively easily incoal-burning power plants that already use two-stagedcombustion to lower NOx. By enriching the nitrogen inthe air to 86% in the first stage of combustion the NOxcan be further reduced by two orders of magnitude.Estimated capital and operational costs for implementingthis technology appear to be very competitive.


Author Index

Anderson, B. . . . . . . . . . . . . . . . . . . . . . . . . . 55Ashby, E. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44Bakajin, O. . . . . . . . . . . . . . . . . . . . . . . . . . . . 35Baker, E. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54Barbee, T.. . . . . . . . . . . . . . . . . . . . . . . . . . . . 41Bernhardt, A. F. . . . . . . . . . . . . . . . . . . . . . . . 31Bodtker, B. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7Bond, T. C.. . . . . . . . . . . . . . . . . . . . . . . . . . . 33Bowers, M. W. . . . . . . . . . . . . . . . . . . . . . . . . 46Brase, J. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8Brewer, L. R. . . . . . . . . . . . . . . . . . . . . . . . . . . 35Carrano, C. . . . . . . . . . . . . . . . . . . . . . . . . . . . 8Ciarlo, D. R. . . . . . . . . . . . . . . . . . . . . . . . 36, 44Colon, D. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7Cooper, G. . . . . . . . . . . . . . . . . . . . . . . . . . . . 34Couch, R. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17Daily, L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17Dallum, G. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9De Groot, A.. . . . . . . . . . . . . . . . . . . . . . . . . . 26Dolan, K. . . . . . . . . . . . . . . . . . . . . . . . . . 42, 45Duffy, S. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54Dunn, T.. . . . . . . . . . . . . . . . . . . . . . . . . . 17, 19Evans, L.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31Ferencz, R. . . . . . . . . . . . . . . . . . . . . . . . . . . . 21Fischer, L. E. . . . . . . . . . . . . . . . . . . . . . . . . . . 55Flath, L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5Fugina, J. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42Glaser, R. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 Gooden, A. . . . . . . . . . . . . . . . . . . . . . . . . . . 53 Graff, T. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Harris, D. B. . . . . . . . . . . . . . . . . . . . . . . . . 4, 10Harvey, C. D. . . . . . . . . . . . . . . . . . . . . . . . . . 31Haskins, J.. . . . . . . . . . . . . . . . . . . . . . . . . . . . 42Haupt, D. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45Havstad, M. A. . . . . . . . . . . . . . . . . . . . . . . . . 24Henderer, B.. . . . . . . . . . . . . . . . . . . . . . . . . . 13Hernandez, J. E. . . . . . . . . . . . . . . . . . . . . . . . . 9Hinkling, T.. . . . . . . . . . . . . . . . . . . . . . . . . . . 20Hoover, C. . . . . . . . . . . . . . . . . . . . . . . . . 21, 26Hurd, R. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5Johnson, G. W. . . . . . . . . . . . . . . . . . . . . . . . . 46Jones, H. E.. . . . . . . . . . . . . . . . . . . . . . . . . . . 39Kallman, J. . . . . . . . . . . . . . . . . . . . . . . . . . . . 44Kartz, M. W. . . . . . . . . . . . . . . . . . . . . . . . . . . 39Kinney, J. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45Klingmann, J. L. . . . . . . . . . . . . . . . . . . . . 41, 50Krulevitch, P. . . . . . . . . . . . . . . . . . . . . . . 29, 35Lamont, A. . . . . . . . . . . . . . . . . . . . . . . . . . . . 20Lane, S. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

Laskowski, G. . . . . . . . . . . . . . . . . . . . . . . . . . 19Leach, Jr., R. R. . . . . . . . . . . . . . . . . . . . . . . . . 12Lee, C. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12Lennon, W. J. . . . . . . . . . . . . . . . . . . . . . . . . . . 7Lewis, P. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4Lin, J. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22Loomis, M.. . . . . . . . . . . . . . . . . . . . . . . . . . . 18Louis, H.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49Maghribi, M. . . . . . . . . . . . . . . . . . . . . . . . . . 29Malba, V. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31Mariella, Jr., R. . . . . . . . . . . . . . . . . . . . . . 30, 49McCallen, D. B. . . . . . . . . . . . . . . . . . . . . . . . . 4McCallen, R. . . . . . . . . . . . . . . . . . . . . . . . . . 19McConaghy, C. . . . . . . . . . . . . . . . . . . . . . . . 11Meeker, D. . . . . . . . . . . . . . . . . . . . . . . . . . . . 49Meyer, G.. . . . . . . . . . . . . . . . . . . . . . . . . . . . 33Miles, R. R. . . . . . . . . . . . . . . . . . . . . . . . . . . . 33Miller, W. O.. . . . . . . . . . . . . . . . . . . . . . . . . . 40Mish, K. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18Montesanti, R. C. . . . . . . . . . . . . . . . . . . . . . . 50Morse, J. D. . . . . . . . . . . . . . . . . . . . . . . . . . . 24Nederbragt, W. . . . . . . . . . . . . . . . . . . . . . . . 41Noble, C. R. . . . . . . . . . . . . . . . . . . . . . . . . . . . 6Palko, J. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54Perry, R. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42Pierce, E. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25Pocha, M. D. . . . . . . . . . . . . . . . . . . . . . . . . . 33Poggio, A. J. . . . . . . . . . . . . . . . . . . . . . . . . . . . 3Puso, M. A. . . . . . . . . . . . . . . . . . . . . . . . . . . 23Raley, N. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34Riddle, R. A. . . . . . . . . . . . . . . . . . . . . . . . . . . 54Rikard, R. D. . . . . . . . . . . . . . . . . . . . . . . . . . . 42Roberson, G. P.. . . . . . . . . . . . . . . . . . . . . . . . 39Rock, D. W. . . . . . . . . . . . . . . . . . . . . . . . . . . . 4Romero, K.. . . . . . . . . . . . . . . . . . . . . . . . . . . . 9Rose, K. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35Rusnak, B. . . . . . . . . . . . . . . . . . . . . . . . . . . . 43Sanders, D. . . . . . . . . . . . . . . . . . . . . . . . . . . 32Sanford, L. . . . . . . . . . . . . . . . . . . . . . . . . . . . 25Sawvel, R. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5Schaich, P. C. . . . . . . . . . . . . . . . . . . . . . . . . . 39Schneberk, D. . . . . . . . . . . . . . . . . . . . . . 41, 45Scott, L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53Shapiro, A. . . . . . . . . . . . . . . . . . . . . . . . . . . . 21Sharpe, R. M. . . . . . . . . . . . . . . . . . . . . . . . . . 18Sherwood, R. . . . . . . . . . . . . . . . . . . . . . . . . . 26Sicherman, A.. . . . . . . . . . . . . . . . . . . . . . . . . 20Speck, D. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25Stewart, J. . . . . . . . . . . . . . . . . . . . . . . . . . . . 20


Author Index (continued)

Swift, D. W. . . . . . . . . . . . . . . . . . . . . . . . . . . 46Syn, C. K. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54Thigpen, R. . . . . . . . . . . . . . . . . . . . . . . . . . . 41Thomas, G. . . . . . . . . . . . . . . . . . . . . . . . . . . 44Thompson, C. A. . . . . . . . . . . . . . . . . . . . . . . . 5

Trumper, D. L. . . . . . . . . . . . . . . . . . . . . . . . . 50Woehrle, T.. . . . . . . . . . . . . . . . . . . . . . . . . . . 12Wolfe, J. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32Zumstein, J. . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Manuscript Date September 2002Distribution Category UC-42

This report has been reproduced directly from the bestavailable copy.

Available for a processing fee to U.S. Department of Energyand its contractors in paper fromU.S. Department of EnergyOffice of Scientific and Technical InformationP.O. Box 62Oak Ridge, TN 37831-0062Telephone: (865) 576-8401Facsimile: (865) 576-5728E-mail: [email protected] for sale to the public fromU.S. Department of CommerceNational Technical Information Service5285 Port Royal RoadSpringfield, VA 22161Telephone: (800) 553-6847Facsimile: (703) 605-6900E-mail: [email protected] ordering: http://www.ntis.gov/ordering.htm/

OR

Lawrence Livermore National LaboratoryTechnical Information Department’s Digital Libraryhttp://www.llnl.gov/tid/Library.html

This document was prepared as an account of work sponsored by an agency of the

United States Government. Neither the United States Government nor the University of

California nor any of their employees, makes any warranty, express or implied, or

assumes any legal liability or responsibility for the accuracy, completeness, or useful-

ness of any information, apparatus, product, or process disclosed, or represents that its

use would not infringe privately owned rights. Reference herein to any specific com-

mercial products, process, or service by trade name, trademark, manufacturer, or other-

wise, does not necessarily constitute or imply its endorsement, recommendation, or

favoring by the United States Government or the University of California. The views

and opinions of authors expressed herein do not necessarily state or reflect those of the

United States Government or the University of California, and shall not be used for

advertising or product endorsement purposes.

This work was performed under the auspices of the U.S. Department of Energy by the

University of California, Lawrence Livermore National Laboratory under Contract

W–7405–Eng–48.

ENG-01-0043b-AD

acknowledgments - nci

Documents