FY07 Engineering Research & Technology Report

Acknowledgments Manuscript Date April 2008 Distribution Category UC-42 Scientific Editor Camille Minichino This report has been reproduced directly from the best copy available. Graphic Designer Available from Irene J. Chan National Technical Information Service 5285 Port Royal Road Project Manager Springfield, VA 22161 Debbie A. Ortega

Production Staff Or online at www-eng.llnl.gov/pubs.html Jeffrey B. Bonivert Lucy C. Dobson Kathy J. McCullough

This document was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor Lawrence Livermore National Security, LLC, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise, Cover: does not necessarily constitute or imply its endorsement, recommendation, or Graphics representing projects from favoring by the United States Government or Lawrence Livermore National Engineeringʼs technology areas. Security, LLC. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or Lawrence Livermore National Security, LLC, 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 Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344. ENG-07-0052

A Message from

Steven R. Patterson Associate Director for Engineering

his report summarizes the core research, develop- and DoD systems performing at their limits, to advanc- Tment, and technology accomplishments in Lawrence es for treating photonic and microfl uidic systems. Livermore National Laboratory’s Engineering Director- FY2007 research projects encompassed coupling ate for FY2007. These efforts exemplify Engineering’s standard fi nite element analysis methods with “mesh- more than 50-year history of developing and applying less” methods to address systems performing at and the technologies needed to support the Laboratory’s beyond failure; and multi-physics coupling of electro- national security missions. A partner in every major magnetics with structural mechanics to simulate program and project at the Laboratory throughout its systems such as electromagnetic railguns. Technology existence, Engineering has prepared for this role with projects included enhancements, verifi cation, and vali- a skilled workforce and technical resources devel- dation of engineering simulation tools and capabilities; oped through both internal and external venues. These progress in visualization and data management tools; accomplishments embody Engineering’s mission: and extensions of our competence in structural damage “Enable program success today and ensure the Labora- analysis. tory’s vitality tomorrow.” Measurement Technologies comprise activities Engineering’s mission is carried out through in nondestructive characterization, metrology, sen- research and technology. Research is the vehicle for sor systems, and ultrafast technologies for advanced creating competencies that are cutting-edge, or require diagnostics. The advances in this area are essential for discovery-class groundwork to be fully understood. the future experimental needs in Inertial Confi nement Our technology efforts are discipline-oriented, prepar- Fusion, High-Energy-Density Physics, Weapons, and ing research breakthroughs for broader application to a Department of Homeland Security programs. variety of Laboratory needs. The term commonly used FY2007 research featured probes for micrometer- for technology-based projects is “reduction to prac- scale metrology; investigations into terahertz systems tice.” for explosives and detection; illicit radionuclide detec- This report combines our work in research and tion; and investigation of the structure and properties technology into one volume, organized into thematic of nanoporous materials. Technology projects included technical areas: Engineering Modeling and Simulation; new error budgeting tools for nondestructive evaluation Measurement Technologies; Micro/ Nano-Devices and systems; x-ray system modeling; laser-based ultrasound Structures; Engineering Systems for Knowledge and applications; and tools to aid in the identifi cation of Inference; and Energy Manipulation. defects in large CT data sets. Engineering Modeling and Simulation efforts focus Micro/Nano-Devices and Structures encompass on the research, development, and deployment of com- technology efforts that fuel the commercial growth of putational technologies that provide the foundational microelectronics and sensors, while simultaneously capabilities to address most facets of Engineering’s customizing these technologies for unique, noncom- mission. Current activities range from fundamental ad- mercial applications that are mission-specifi c to the vances to enable accurate modeling of full-scale DOE Laboratory and DOE. The Laboratory’s R&D talent

ii FY07 Engineering Research and Technology Report Introduction and unique fabrication facilities have enabled highly be possible for problems as complex as the prediction innovative and custom solutions to technology needs of disease outbreaks or advance warning of terrorist in Stockpile Monitoring and Stewardship, Homeland threats. Security, and Intelligence. FY2007 research efforts were centered on the de- FY2007 research projects included systems for composition of large-scale semantic graphs. Technology defense against biothreats and for the manipulation of efforts included a testbed to evaluate hierarchical clus- biomolecules and viruses; studies of transport behav- tering; and work on a statistical approach to the design ior and crystal-driven neutron sources. Technology of complex systems in the presence of uncertainty. projects included laser pantography; pyrosequencing Energy Manipulation, a long time focus that is and validation of acoustic modeling for microfl uidic receiving increased emphasis due to newly emerging systems; construction of diagnostics for optical gating; applications, encompasses the fundamental under- applications of lasers and optical sensors; and new standing and technology deployment for many modern capabilities for micro- and nano-fabrication. pulsed-power applications. This area has broad appli- Engineering Systems for Knowledge and Inference, cations for magnetic fl ux compression generators and an emerging focus area for Engineering as well as components for modern accelerators. for the country at large, encompasses a wide variety FY2007 research focused on an ultra-high velocity of technologies. The goal is to generate new under- railgun. Technology efforts focused on railgun pulse standing or knowledge of situations, thereby allowing power systems and diagnostics; UV-induced fl ashover; anticipation or prediction of possible outcomes. With diagnostics for electrical breakdown in vacuum; and this knowledge, a more comprehensive solution may fi ber optic current measurements.

Lawrence Livermore National Laboratory iii Contents Introduction A Message from Steven R. Patterson ...... ii

Engineering Modeling and Simulation Deformation of Low Symmetry and Multiphase Materials Nathan R. Barton ...... 2

Plasticity at High Pressures and Strain Rates Using Oblique-Impact Isentropic-Compression Experiments Jeffrey N. Florando ...... 4

“Natural Neighbor” Meshless Method for Modeling Extreme Deformations and Failure Michael A. Puso ...... 6

Laser Glass Damage: Computational Analysis of Mitigation Process James S. Stölken ...... 8

Electro-Thermal-Mechanical Simulation Capability Daniel White ...... 10

Petascale Simulation Initiative John M. May ...... 12

Electromagnetics Code Management Daniel White ...... 14

Finite Element Analysis Visualization and Data Management Bob Corey ...... 16

Rotational Kinematics Output and Other Improvements in DYNA3D Jerry I. Lin ...... 18

NIDE3D Code Maintenance and Enhancement Michael A. Puso ...... 20

Validation of Reinforced Concrete Modeling Capabilities for Seismic Response Steven W. Alves ...... 22

Thermal-Structural Analysis of the MacArthur Maze Collapse Charles R. Noble ...... 24

Enhanced Composite Modeling Tools Andrew Anderson ...... 26

EMP Simulation and Measurement Data Analysis in Support of the Titan Experiments Charles G. Brown, Jr...... 28

Full-Process Computer Model of Magnetron Sputter: Testing Existing State-of-the-Art Components Christopher Walton ...... 30

iv FY07 Engineering Research and Technology Report Contents

Direct Simulation Monte Carlo Benchmarking and Code Comparison Aaron P. Wemhoff ...... 32

Photonic Doppler Velocimetry Adam White ...... 34

Measurement Technologies Standing Wave Probes for Micrometer-Scale Metrology Richard M. Seugling ...... 38

Error Budgeting and Certiﬁ cation of Dimensional Metrology Tools Jeremy J. Kroll ...... 40

Uncertainty Analysis with Inspection Shop Measurements Walter W. Nederbragt ...... 42

X-Ray System Modeling William D. Brown ...... 44

Detection, Classiﬁ cation, and Estimation of Radioactive Contraband from Uncertain, Low-Count Measurements James V. Candy ...... 46

Terahertz Spectroscopic Imaging for Standoff Detection of High Explosives Michael W. Burke ...... 48

Application of Laser-Based Ultrasound to Glovebox Enclosed Materials Robert Huber ...... 50

Defect Detection in Large CT Image Sets Douglas N. Poland ...... 52

The Structure and Properties of Nanoporous Materials Anthony Van Buuren ...... 54

Construction of a Frequency Resolved Optical Gating (FROG) Diagnostic Corey V. Bennett ...... 56

Picosecond Response, Recirculating Optical Probe for Encoding Radiation (ROPER) Stephen P. Vernon ...... 58

Serrated Light Illumination for Deﬂ ection Encoded Recording (SLIDER) John E. Heebner ...... 60

Micro/Nano-Devices and Structures Biological Sample Preparation Infrastructure N. Reginald Beer ...... 64

Rapid Defense Against the Next-Generation Biothreat Raymond P. Mariella, Jr...... 66

Lawrence Livermore National Laboratory v Validation of 3-D Acoustic Modeling of Commercial Codes for Microﬂ uidic Systems Raymond P. Mariella, Jr...... 68

Thermal-Fluidic System for Manipulation of Biomolecules and Viruses Kevin D. Ness ...... 70

Flow-Through Pyrosequencing in a Microﬂ uidic Device Klint A. Rose ...... 72

Investigation of Hybridization Times for DNA Microarrays Robin Miles ...... 74

Microfabricated Silicon Nib for Low-Cost, Multiplexed Micro-Coating Erik V. Mukerjee ...... 76

Advanced Polymerase Chain Reaction Module Satinderpall Pannu ...... 78

Tunable Optical Cavities for Gas Sensor Tiziana Bond ...... 80

Extending Lasers to Universal Gates for Photonic FPGAs Tiziana Bond ...... 82

Compact Tunable Optical Sensors Jack Kotovsky ...... 84

Transport Behavior and Conversion Efﬁ ciency in Pillar Structured Neutron Detectors Rebecca J. Nikolić ...... 86

Compact High-Intensity Crystal-Driven Neutron Source Jeffrey D. Morse ...... 88

Pillar Neutron Detector Fabrication Rebecca J. Nikolić ...... 90

Flow Programmed Mesoscale Assembly of Nanoengineered Materials and Net Shape Components Klint A. Rose ...... 92

Laser Pantography Christopher M. Spadaccini ...... 94

Bridging the Length and Time Scales from Atomistics to the Microscale Todd Weisgraber ...... 96

“HE-Less” Detonator Modeling James F. McCarrick ...... 98

Chip Slapper Detonator Processing for Rapid Prototyping and Hydrodynamic Properties Mark A. Schmidt ...... 100

vi FY07 Engineering Research and Technology Report Contents

Engineering Systems for Knowledge and Inference Semantic Graph Hierarchical Clustering and Analysis Testbed Tracy D. Lemmond ...... 104

Decomposition of Large-Scale Semantic Graphs via an Efﬁ cient Communities Algorithm Yiming Yao ...... 106

Statistical Approach to Complex Systems in the Presence of Uncertainty Lee G. Glascoe ...... 108

Energy Manipulation Ultrahigh-Velocity Railgun Jerome M. Solberg ...... 112

Pulse Power Systems and Diagnostics for the Fixed Hybrid Armature Railgun Todd J. Clancy ...... 114

Fast Diagnostic for Electrical Breakdowns in Vacuum Timothy L. Houck ...... 116

UV-Induced Insulator Flashover Jay Javedani ...... 118

Evaluation of Light Activated Thyristors for Pulsed Power Applications Laura Tully ...... 120

Fiber Optic Current Measurements Using Faraday Rotation Diagnostics Brent McHale ...... 122

Author Index ...... 125

Lawrence Livermore National Laboratory vii

Research

Nathan R. Barton Deformation of Low (925) 422-9426 Symmetry and [email protected] Multiphase Materials

aterials composed of low symmetry For example, phenomena such as shear Relevance to LLNL Mission Mcrystals or of multiple solid phases localization arise naturally from applica- In multiphase materials of interest to exhibit heterogeneous deformation at the tion of the method at the appropriate LLNL, such as Ti-6Al-4V, micro- microstructural scale, presenting signifi - scale. structure infl uences yield and fl ow cant challenges to efforts to construct stresses, ductility, high- and low-cycle macroscale constitutive models. This Project Goals fatigue, creep, fracture toughness, and deformation heterogeneity at the micro- The overarching goal of this project large crack growth properties. Micro- structural scale also produces stress is to produce effective macroscale structure has also been shown to infl u- concentration, which can lead to fracture models through novel homogenization ence response under dynamic loading or infl uence the onset and progress of methods for materials where conven- conditions. The presence of low sym- phase transformations. We are developing tional methods fail. The immediate metry phases can contribute to severe an approach that explicitly incorporates application space includes a broad class anisotropy and a tendency for strain effects of microstructure and deforma- of engineering simulations, ranging from localization. Explicit inclusion of the tion heterogeneity in a framework suited forming operations to dynamic loading microstructure allows for the effective to analysis of engineering scale compo- scenarios. Figure 1 shows an example treatment of deformation heterogeneities nents. application, with computation performed at the microstructural scale. In addition to Applications involving fully de- using similar techniques appropriate to multiphase titanium alloys, the approach veloped plastic fl ow are targeted. We single-phase cubic symmetry poly- is applicable to low symmetry crystal- explicitly represent the microstructure, crystalline materials. Software is devel- line metals such as uranium (ortho- directly treating deformation hetero- oped in a component-oriented fashion, rhombic), beryllium (hexagonal), and geneities, and we build on emerging making use of tools that enhance parallel α-plutonium (monoclinic). technologies for effectively combining load-balancing through task parallelism, microscale plasticity simulations with as shown in Fig. 2. Initial development FY2007 Accomplishments and Results macroscale models. New capabilities will is focused on the Ti-6%Al-4%V Our fi rst completed milestone is capture the impact of microstructure, (Ti-6Al-4V) alloy, given its widespread associated with the assessment of im- and thus material processing, on perfor- use and the availability of relevant portant features for microstructures of mance of engineering scale components. experimental data. interest in Ti-6Al-4V. We are able to use experimental data to inform numerical multiphase microstructure realizations (Fig. 3). New techniques have been developed to capture the morphology of the experimentally observed microstructures. Another milestone is associated Figure 1. A biaxial bulge test with new capabilities for fi ne-scale workpiece, deformed by application calculations in the ALE3D fi nite ele- of pressure to the lower surface. The false color depicts the relative ment program. We have determined a effective plastic strain rate. Due to fi ne-scale implementation strategy and have tested ALE3D in the role of the its lattice orientation distribution 1.65 function, the sheet material has fi ne-scale model. New capabilities have orthotropic symmetry and exhibits 1.32 been developed for obtaining steady macroscopic strain localization, as plastic fl ow solutions in ALE3D, with a indicated by the red region along the 1.00 specialized elasto-viscoplastic fi ne-scale nearer symmetry plane. 0.67 material model. Figure 4 depicts results from an example calculation performed 0.35 using ALE3D. Max: 1.678 Min: 0.000

2 FY07 Engineering Research and Technology Report Engineering Modeling & Simulation

Implementation is under way for 2. Barton, N. R., J. Knap, A. Arsenlis, R. new physics and software requirements Becker, R. D. Horning, and D. R. Jefferson, for the appropriate coarse/fi ne coupling “Embedded Polycrystal Plasticity and FY2008 Proposed Work strategy. This includes homogenization Adaptive Sampling,” Int. J. Plasticity, 24, In the coming fi scal year, to obtain anisotropic coarse-scale elastic pp. 242-266, 2008. software components that have been response from fi ne-scale data, new cou- 3. Barton N. R., and H.-R. Wenk, “Dauphiné developed for various aspects of the pling software for fi ne-scale models that Twinning in Polycrystalline Quartz,” Model- multiscale problem will be combined do not provide derivative information, ing Simul. Mater. Sci. Eng., 15, pp. 369-394, into a working whole. After testing, and velocity-gradient-driven parameter- 2007. algorithm refi nement, performance ized plastic fl ow. 4. Brechet, Y. J. M., P. Dawson, J. D. Em- tuning, and model calibration, the bury, C. G’sell, S. Suresh, and H.-R. Wenk, overall capability will be validated by Related References “Recommendations on Modeling Polyphase comparison with available experi- 1. Arsenlis, A., N. R. Barton, R. Becker, and Plasticity: Conclusions of Panel Discus- mental data at the appropriate length R. E. Rudd, “Generalized In Situ Adaptive sions,” Mater. Sci. Eng., A175, pp. 1-5, 1994. scale. We also plan to demonstrate Tabulation for Constitutive Model Evalu- 5. Pope, S. B., “Computationally Effi cient extension of the method to another ation in Plasticity,” Compt. Methods Appl. Implementation of Combustion Chemistry low symmetry or multiphase material Mech. & Eng., 196, pp. 1-13, 2006. Using In Situ Adaptive Tabulation,” Com- such as beryllium or uranium. bust. Theory Modeling, 1, pp. 41-63, 1997.

Finite element model (ALE3D)

Proxy

Figure 2. Schematic of Multiple-Program, Multiple-Data (MPMD) parallelism Viscoplastic polycrystal showing the Remote Method Invocation (RMI) pattern during the material model model servers evaluation. Each box indicates a separate instance of a given executable program, with box subdivisions indicating parallelism within an executable. Material model servers

Processors

Figure 3. (a) Measured, (b) and (c) numerically generated (at different magni- ﬁ cations) microstructures for Ti-6Al-4V. Red indicates α phase; blue indicates β phase. Microstructures are obtained experimentally using electron backscatter diffraction. Numerical microstructures are generated by sampling the appropriate distribution functions and using particle-packing techniques. Numerical microstructures reproduce experimentally observed morphologies, with tunable parameters to control features such as phase volume fraction.

(a) (b) (c)

Figure 4. Heterogeneity of (a) stress magnitude and (b) plastic strain rate in a two-phase Ti-6Al-4V microstructure simulated in ALE3D, with relative plastic strain rates reaching ten times the nominal applied value. The highest relative plastic strain rates are observed in the β phase, which has a more isotropic ﬂ ow behavior, while the highest stress magnitudes are observed in the more anisotropic α phase.

(a) (b)

Lawrence Livermore National Laboratory 3 Research

Jeffrey N. Florando Plasticity at High Pressures and (925) 422-0698 Strain Rates Using Oblique- fl [email protected] Impact Isentropic-Compression Experiments arious aspects of the Laboratory’s Project Goals Vnational security mission depend The goals of this project are to on accurate computer code simula- develop an oblique-impact isentropic- tions of high-strain-rate plastic fl ow compression experiment (Fig. 1) to (i.e., nonreversible deformation) under measure the strength of materials under conditions of high hydrostatic pressures. a condition of combined high strain rate While progress has been made in recent (104 to 106 s-1) and high pressure (1 to years, especially at the extreme cases of 100 GPa). The isentropic compression pressure and strain rate, there is still an allows for high pressures to be achieved uncertainty in understanding the strength over relatively long time frames of materials under conditions of com- (micro-seconds), and the oblique impact bined high strain rate (104 to 106 s-1) allows for a measurement of the strength and high pressure (1 to 100 GPa). properties under pressure. The strength Current strength models used in data will then be used to refi ne and en- simulations include some physically hance the current strength models. When based models such as the Mechanical completed, this work will increase the Threshold Stress formulation, which has Laboratory’s ability to develop predic- over 20 parameters. The uncertainty in tive strength models for use in computer the values for these parameters as well as code simulations. values for the parameters in other physically based models is under question due Relevance to LLNL Mission to the inherent diffi culties in conducting Understanding and simulating the and extracting high-quality experimental strength of materials under dynamic data in the high-pressure and high-strain- loading conditions is a major com- rate regimes. The experimental studies ponent of the Laboratory’s Stockpile of material strength at these pressure Stewardship Program and is applicable and strain rate regimes will further the to future National Ignition Facility understanding of the underlying physical (NIF) experiments. These computer strength mechanisms needed for accurate code simulations, however, require material strength models. additional experimental data in order to develop new models and validate the Keyed barrel existing codes.

5º – 25º FY2007 Accomplishments and Results Q Sample Longitudinal and The oblique-impact isentropic-com- V o Window transverse particle pression experiment requires a keyed velocity measurement barrel. The experiments are developed at Graded density Longitudinal and LLNL and tests are performed at Brown impactor-isentropic transverse waves compression are launched into University. A formal collaboration has the sample been established with a three-year contract that is now in place. The ﬁ rst year Figure 1. Oblique-impact isentropic-compression experiment.

4 FY07 Engineering Research and Technology Report Engineering Modeling & Simulation

0.0006 0.0006 Longitudinal wave/20 0.0005 0.0005 s) Exp. μ Simulation

s) transverse μ 0.0004 0.0004 longitudinal velocity Johnson-Cook velocity/20

0.0003 0.0003 Exp. Johnson- longitudinal Cook velocity/20 MTS Particle velocity (cm/ 0.0002 0.0002 MTS Particle velocity (cm/

Steinberg- 0.0001 0.0001 Guinan Steinberg- Guinan 0 0 01.60.20.40.60.8 1.0 1.2 1.4 01.60.20.40.60.8 1.0 1.2 1.4 Time (μs) Time (μs) Figure 2. 3-D hydrodynamics simulation showing the longitudinal wave and the Figure 3. Experimental data showing the measured longitudinal and transverse waves. sensitivity of the transverse waves to the different strength models. While there is good agreement between the simulation and experiment in the longitudinal wave, the experimental transverse wave is much larger, signifying a material that has higher strength than predicted. has focused on fi elding the experiment simulations are also used to help guide at low pressures, which includes sample the experiments, such as determining the preparation and characterization, simula- optimal sample and sapphire thickness. FY2008 Proposed Work tions of the proposed experiment, and Experimental Results. Experiments In FY2008, additional low- results from the initial experiments. have been performed at Brown Uni- pressure experiments and intermedi- Sample Preparation and Characteriza- versity on Cu/sapphire samples. The ate pressure experiments (150 to tion. The initial sample confi guration is a longitudinal and transverse wave results 300 kbar) on Cu, Ta, and V are Cu sample with a sapphire window, which are shown in Fig. 3. A couple of key planned at Brown University. The was chosen due to its close impedance features are that the graded density results from the tests will give us the match. A fi ne diffraction grating of 1200 impactor causes a ramp in the pressure necessary data to begin the develop- lines/mm, which is used to measure the (isentropic compression) which con- ment and refi nement of strength normal and transverse waves, is etched trols the strain rate to ~ 105, and that models. Based on the results, the into the sapphire windows, and a thin the transverse wave arrives near peak designs for a 2-in. graded density layer (~130 nm) of Cu is deposited on pressure. Also, the transverse signal impactor and a soft recovery experi- top of the grating. The Cu sample is measured is much higher than expected, ment will be explored. Simulations then attached to the deposited Cu using signifying that the Cu is stronger than will continue to be conducted to aid in epoxy on the circumference. In order to the models predict. the design of the experiments. achieve a near isentropic compression, a graded density impactor is used. Diffusion Related References bonded impactors have been character- 1. Remington, et al., Mat. Sci. and Tech., 22, ized to understand the strength and waves p. 474, 2006. imparted into the sample. 2. Follansbee, P. S., and U. F. Kocks, Acta Simulations. We have performed a Metall., 36, p. 81, 1988. series of simulations (Fig. 2), which shows 3. Nguyen, J. H., D. Orlikowski, F. H. Streitz, the validity of the proposed experiments. N. C. Holmes, and J. A. Moriarty, AIP Pro- The simulation results show that the trans- ceedings, 706, p.1225, 2004. verse wave, which is related to the strength 4. Clifton, R. J., and R. W. Klopp, ASM of the material under pressure, is very Handbook 9th ed., 8, p. 230, 1985. sensitive to the strength model used. The

Lawrence Livermore National Laboratory 5 Research

Michael A. Puso “Natural Neighbor” Meshless (925) 422-8198 Method for Modeling Extreme [email protected] Deformations and Failure

he objective of this work is to Eulerian (ALE)-type methods. On time-step calculations, and techniques Tdevelop a fully Lagrangian analysis the other hand, the standard meshless for improving effi ciency. A number of approach based on “natural neighbor” particle methods have other pathologies refereed journal articles resulted from discretization techniques to model such as instabilities and insuffi cient treat- this work and the methods are currently extreme deformation and failure for ment of boundaries and ineffi ciencies. implemented and available in Labora- analyses such as earth penetration and The application of novel shape functions tory codes. Verifi cation and validation dam failure. The standard fi nite element based on natural neighbors has been suc- and material modeling have also been a approaches often do not work in these cessfully used in this work to overcome key component of our effort. applications because of “mesh tangling” many of these pathologies. (Fig. 1). Lagrangian methods allow the tracking of particles and free surfaces, Project Goals (a) which makes handling of sophisticated The goal of this work is to fi x many material models and effects due to of the inherent problems associated debris and fragmentation much more with meshless methods. This involves straightforward and natural as compared the development of several neighbor- to Eulerian and arbitrary-Lagrangian- based approximation methods, stable

(b) x 10–6 8 72,735 nodes 7 156,028 nodes 275,952 nodes

6 Experiment 2 5

Decceleration (in/s ) 2

1 0 0 0.001 0.002250 0.003375 0.004500 Time (s) Figure 2. (a) DYNA3D simulation of concrete penetration (tensile damage shown in red); (b) comparison of simulation results to experimental results from Sandia National Laboratories. Three different nodal discretizations were Figure 1. DYNA3D particle simulation of earth moving problem. This type of analysis used in the simulation and all compare well to the Sandia would not be feasible with ﬁ nite elements due to mesh tangling. test results.

6 FY07 Engineering Research and Technology Report Engineering Modeling & Simulation

(3a) (4a) (4b) Figure 3. (a) DYNA3D simulation of penetration of reinforced concrete (tensile damage shown in red); (b) rebar (shown alone) attached to particles in the simulation (eventually fails upon exit). Simulations compare well to data in the literature.

Figure 4. (a) View of Taylor bar from new particle method simulation. An evolving (3b) elliptical kernel was used to maintain proper particle connectivity. (b) Results from classical particle methods (e.g. Smoothed Particle Hydrodynamics) where “toeing” results. This effect is often attributed to a tensile instability.

Relevance to LLNL Mission Lecture Notes in Computational Science and A number of high profi le analysis Engineering, Bonn, Germany, September areas will benefi t from this work. High 12 – 15, 2005. rate penetration dynamics is identi- (1) 4. Chen, J. S., W. Hu, and M. A. Puso, fi ed as a challenge area in engineering “Strain Smoothing for Stabilization and and validation work in this area (Fig. Regularization of Galerkin Meshfree 2) has already been done using the Methods,” Third International Workshop on new approaches with the LLNL code Meshfree Methods, Proceedings of Third In- DYNA3D. Homeland Security applica- ternational Workshop on Meshfree Methods, tions are important to the LLNL mission (2) Meshfree Methods for Partial Differential and validation has begun looking at Equations III, volume 57 of Lecture Notes the effects of penetrators on reinforced Finally, a new method for handling in Computational Science and Engineering, concrete walls (Fig. 3). principal stress damage with plasticity Bonn, Germany, September 12 – 15, 2005. was implemented in the concrete model 5. Chen, J. S., W. Hu, and M. A. Puso, FY2007 Accomplishments and Results used in the penetrator analysis shown “Orbital HP-Clouds and Higher Order Our FY2005 implementation used in Figs. 2 and 3. The remainder of our SCNI for Solving Schrodinger Equation in a natural neighbor scheme that required work was in the area of verifi cation and Quantum Mechanics,” International Jour- the computation of a Voronoi diagram validation as demonstrated in Figs. 2 nal for Numerical Methods in Engineering, at each step of an analysis. In FY2006, through 4. 67, pp. 847-867, 2007. we proposed a simplifi ed scheme where natural neighbors were used to form Related References elliptical kernel functions in a moving 1. Puso, M. A., “An Improved Linear least squares (MLS) approach for com- Tetrahedral Element for Plasticity,” Joint puting shape functions. ASME/ASCE/SES Conference on Mechanics In FY2007 this method was extended and Materials, Proceedings of McMat, Baton to the treatment of large deformation Rouge, Louisiana, June 1 – 3, 2005. problems by applying the appropriate 2. Puso, M. A., and J. Solberg, “A Formula- evolution scheme for the elliptical kernels tion and Analysis of a Stabilized Nodally FY2008 Proposed Work (Fig. 4). In addition, a maximum entropy Integrated Tetrahedral,” International Journal Our goal of developing an ac- (MAXENT) method was applied as an for Numerical Methods in Engineering, 67, curate and effi cient particle method alternative to the MLS approach to bet- pp. 841-867, 2006. has been met. Nonetheless, a number ter treat essential boundary conditions. 3. Puso, M. A., J. S. Chen, and E. Zywicz, “A of research areas still exist. For ω Given an elliptical kernel function i at New Stabilized Nodal Integration Approach,” example, initial particle placement φ node i, MAXENT shape functions i are Third International Workshop on Meshfree has a large infl uence on the accuracy chosen by minimizing the informational Methods, Proceedings of Third of the results. Continued work in the entropy function ƒ in (1), subject to the International Workshop on Meshfree area of large deformation material partition of unity and linear exactness Methods, Meshfree Methods for Partial modeling is also needed. constraints in (2). Differential Equations III, volume 57 of

Lawrence Livermore National Laboratory 7 Technology

James S. Stölken Laser Glass Damage: (925) 423-2234 Computational Analysis of [email protected] Mitigation Process

nderstanding and controlling the predictive simulation capability that A key objective is to couple the Uphysical processes that cause laser- incorporates the essential ingredients Hydro and EM simulations to facilitate induced damage to optical components of the laser-material interaction and the exploring how strongly this interaction is crucial to the success of many high- resulting coupled material response in must be implemented. The degree of energy-density experimental facilities experimentally relevant confi gurations. coupling required to adequately simulate including NIF. Experimental and theo- We are addressing the technical chal- a given phenomenon is problem- retical investigations of laser damage in lenges associated with energy deposi- dependent and greatly affects the silica glass are legion and are actively tion, dynamic material response, and computational costs and effi cacy of the being pursued within the NIF Program. the nature of the coupling between these overall simulation methodology. The insight garnered from experimental two processes. Our approach is to adapt Our technical approach has been as data is limited by the extremely short the current simulations capability within follows. time-scale (~1 ns) of the damage events the EMSolve code to simulate the time, 1. Implement spatial and material state- and the complex interplay between space, and material state dependence dependent EM properties, such as the energy deposition and hydrodynamic of the laser energy deposition process. complex dielectric constant. Using response. A further complication is the The material’s dynamic response is then a Lorenz-Lorentz formulism, we stochastic nature of damage initiation, simulated using advanced multi-phase explicitly modeled the dependence of necessitating a post priori approach equations of state (EOS) and failure the material conductivity and permit- where the damage process must be de- models within the ALE3D multi-physics tivity upon density (Fig. 1). This same duced via forensic reconstruction. code. This also provides some capabil- approach explicitly accounts for the Understanding laser-damage ity to model damage initiation in silica, laser energy frequency dependence, initiation and growth is just half of the which would provide a self-consistent thereby facilitating the investigation problem. Creating an effective strategy facility to establish initial conditions for of this important parameter in candi- to detect and mitigate laser damage is mitigation studies. date mitigation processes. To improve the essential second part. One promising the quality of the boundary conditions approach is to use infrared wavelength Project Goals of the EM simulations and enhance (CO2) laser energy to excise/anneal The primary goal is to implement their dynamic range, we adopted a the damage site. Key parameters to be a capability to perform coupled EM scattered electric-fi eld methodology optimized include the laser wavelength, simulations of laser energy deposition (Fig. 2). intensity, pulse duration, and scan pattern and subsequent material hydrodynamic 2. Adapt existing models of material in relation to the size and type of dam- response that includes the dependence deformation, phase transforma- age site. Both damage initiation/growth of the energy deposition processes tions, and damage to simulate such studies and mitigation process develop- on local variations in state-dependent processes under conditions relevant ment could benefi t from a high-fi delity material properties. to laser damage in silica glass. We adopted a unifi ed-creep model for the deformation of silica at high Total field air Scattered field glass 150 temperatures using material param- 100 eters consistent with a linear viscous 50 solid and a temperature-dependent

m 0 shear modulus. The two-phase EOS μ –50 300 is based on combining two analytic –100 200 forms for the low and high-pressure –150 100 phases with an irreversible kinetic –300 –200 –100 0 100 200 300 0 –6 relation (Fig. 3). X axis (x10 ) X –100 Scattered Total field Total Z –200 3. As a ﬁ rst step, consider only a single field air densified glass field glass Y –300 material state variable: the material Figure 1. Simulated Joule heating in glass around Figure 2. 3-D rendering of laser heat deposition similar to density. The material density plays a an artiﬁ cial defect with 20% greater permittivity and that shown in Fig. 1. Energy deposition varies by a factor central role in both the hydrodynamic conductivity than the bulk glass. of one hundred from blue to red. response and the energy deposition.

8 FY07 Engineering Research and Technology Report Engineering Modeling & Simulation

Pressure Relative volume 0 0 to the reference density, bulk modulus, and thermal expansion coefﬁ cients as- –2 –2 sociated with both the high- and low- –4 –4 pressure phases. A simple model of brittle damage has been adapted to account for –6 –6 cracking under tensile loading and suc- –8 –8 cessfully tested in conjunction with the

With densification (mm) above densifi cation EOS model. A scat- –10 –10 tered electric fi eld formulation has been –6 –4 –2 0 2 4 6 –6 –4 –2 0 2 4 6 implemented to improve the effi ciency of 0 0 the E&M simulations. –2 –2 The capability to account for spatially varying permittivity and conductivity –4 –4 has been implemented. The dependence –6 –6 of the real and imaginary parts of the refractive index on material density and –8 –8 radiation wavelength has been addressed

Without densification (mm) in the framework of a Lorenz-Lorentz –10 –10 –6 –4 –2 0 2 4 6 –6 –4 –2 0 246 model and implemented in the EMSolve mm mm code. The Hydro and E&M simulations Figure 3. Residual pressure and relative volume following indentation of glass with a smooth circular punch. were coupled via a fi le-sharing scheme by extending existing capabilities within the ALE3D software and adapting the Furthermore, under high-pressure Relevance to LLNL Mission EMSolve software to import and export loading, such as laser-induced shocks, This high fi delity predictive simula- the requisite fi les. An example of a the reference density of silica is tion capability, incorporating laser- coupled code simulation is shown in modifi ed, providing material “mem- material interactions and coupled material Fig. 5. ory” of previous history such as glass response, will provide a valuable capabil- damage. ity to help advance our understanding of A loosely coupled scheme was used the physical processes involved in optical 100 to connect the hydrodynamics simu- material damage and mitigation. lations (ALE3D) to the laser energy 50 deposition simulations (EMSolve) via FY2007 Accomplishments and Results a fi le passing mechanism (Fig. 4). This A two-phase equation-of-state (EOS) approach leverages existing capabilities model has been adapted to account for 0 ) within ALE3D to specify spatially and the permanent densifi cation of silica –6 temporally varying heat sources using glass that occurs under high-pressure X Axis (x 10 an external fi le. loading. Model parameters have been fi t -50

-100 ALE3D hydrodynamic simulation -150 Material density 100 80 60 40 20 0 100 80 60 40 20 0 Y Axis (x 10–6) Y Axis (x 10–6) Density Pseudocolor Pseudocolor 2 –qe RN 1 Var:Etot_mag Var:Joule Heat _Energy 5 m 2 2 3.499 x 10 9.00 x 1012 e A W – W0 + 2iDW Lorenz-Lorentz E(W) = 1 + 4π Heat 2.624 x 105 6.75 x 1012 4π –q2 RN 1 model 1 – e 1.749 x 104 4.501 x 1012 m 2 2 4 3 e A W – W0 + 2iDW 8.747 x 10 2.251 x 1012 X Permittivity 1.581 0.000 Laser frequency conductivity Max: 3.499 x 105 Max: 9.00 2 x 1012 Min: 1.581 Y Z Min: 0.000 EMSolve electrodynamic Figure 5. Coupled EMSolve/ALE3D simulation of a laser simulation illuminated damage site similar to that shown in Fig. 3. The total electric ﬁ eld (left) and Joule heating (right) are modulated by local variations in density captured by Figure 4. Loosely coupled scheme that includes the effects of laser frequency and material density. hydrocode simulation of the indentation process.

Lawrence Livermore National Laboratory 9 Research

Daniel White Electro-Thermal-Mechanical (925) 422-9870 Simulation Capability [email protected]

he purpose of this project is to Project Goals mission to enhance/extend the simula- Tresearch and develop numerical A coupled ETM simulation is a sim- tion capabilities and specifi cally ad- algorithms for 3-D electro-thermal- ulation that solves, in a self-consistent dresses the need for simulation capabil- mechanical (ETM) simulations. LLNL manner, the equations of electromag- ity in the area of energy manipulation. A has long been a world leader in the area netics (primarily statics and diffusion), robust ETM simulation capability will of computational solid mechanics, and heat transfer (primarily conduction), enable LLNL physicists and engineers recently several solid mechanics codes and nonlinear mechanics (elastic-plastic to better support current DOE programs, have become “multi-physics” codes deformation, and contact with friction). and will prepare LLNL for some very with the addition of fl uid dynamics, Our goal is to add electromagnetics to exciting long-term DoD opportunities. heat transfer, and chemistry. How- two existing mechanics codes, ALE3D ever, these multi-physics codes do not and Diablo. ALE3D is a heavily used FY2007 Accomplishments and Results incorporate the electromagnetics that Arbitrary-Lagrangian-Eulerian hydro- FY2007 was our fi nal six-month is required for a coupled ETM simula- dynamics code; Diablo is an implicit effort with emphasis on code enhance- tion. There are numerous applications Lagrangian thermal-mechanics code cur- ments, verifi cation, validation, pub- for an ETM simulation capability, such rently under development. lications, and initiation of follow-on as explosively driven magnetic fl ux Our objective is to develop a novel activities. The two primary code en- compressors, electromagnetic launchers, simulation capability that is not avail- hancements were the development of an inductive heating and mixing of metals, able commercially or from the other RLC circuit model and the incorporation and micro-electromechanical systems national laboratories. With this capabil- of a multigrid solver. (MEMS). ity, LLNL will have an unprecedented The RLC circuit model is used to ability to simulate, design, and optimize model capacitor banks and associated ETM systems. cables and pulse-shaping hardware. With this circuit model it is not necessary Relevance to LLNL Mission to know the applied voltage a-priori; This project is aligned with LLNL’s instead the applied voltage is computed core competency in simulation science by solving the ETM PDE’s and the RLC and engineering. It contributes to the circuit model self-consistently.

Figure 1. CAD image of the coaxial load experiment. The image shows the steel outer tube, the aluminum inner tube, the torlon bolts,and the coaxial cables. Figure 2. Photograph of the coaxial load experiment test ﬁ xture.

10 FY07 Engineering Research and Technology Report Engineering Modeling & Simulation

We incorporated a multigrid solver tests used larger current with signifi cant Related References for the curl-curl equations. The solver has plastic deformation of the inner tube. An 1. White, D. A., B. J. Fasenfest, R. N. numerous parameters and requires aux- example result is shown in Fig. 3. Over- Rieben, and M. L. Stowell, “A QR Acceler- iliary mesh information such as element all, the simulation agreed quite well with ated Volume-To-Surface Boundary Condi- connectivity and nodal coordinates. The the experimental data for early times tion for the Finite Element Solution of Eddy end result is that the solver is approxi- (t < 100 μs), showing that there are no Current Problems,” IEEE Trans. Magnetics, mately 10x faster for typical problems of major errors in the numerical algorithms 43, 5, pp. 1920-1934, 2007. interest, and has better scalability. or the software implementation. 2. Rieben, R., and D. White, “Verifi cation of The primary code validation task The fi nal experiment was with a High-Order Mixed Finite Element Solution was designing and executing a coaxial 0.030-in. inner tube with slots. The of Transient Magnetic Diffusion Problems,” load experiment and comparing experi- purpose of the slots is twofold: they IEEE Trans. Magnetics, 42, 1, pp. 25-39, mental data to computational simula- make the problem become a fully 3-D 2006. tions. A CAD image of the test fi xture problem, and the deformation of the 3. White, D., and B. Fasenfest, “Perfor- is shown in Fig. 1 and a photograph is metal should be signifi cant. In fact, we mance of Low-Rank QR Approximation of shown in Fig. 2. A sequence of tests expect a complete destruction of the the Finite Element Biot-Savart Law,” IEEE were executed, with the initial tests at tube. The simulation result is shown in Trans. Magnetics, 43, 4, pp. 1485-1488, low current such that the deformation of Fig. 4; the experimental data is still be- April 2007. the inner tube was small and the mate- ing examined. 4. White, D., R. Rieben, and B. Wallin, rial response was linear. Subsequent “‘Coupling Magnetic Fields and ALE Hydro- dynamics for 3D Simulations of MFCG’s,” IEEE International Conference on Megagauss (a) (b) x 10–3 Hoop strain 465KA x 10–3 Radial displacement Magnetic Field Generation, Santa Fe, New 1.0 3 ALE3D ALE3D Mexico, November 5-10, 2006. SG1 2 PDV1 5. Solberg, J. M., D. A. White, and R. N. 0.5 SG3 PDV2 Noise 1 PDV3 Rieben, “A Mortar-Based Method for Zero- PDV4 Conductivity Regions in a Coupled Electro- 0 0 Thermal-Mechanical Simulation Code,” 7th

cm –1 cm World Congress on Computational Mechanics, –0.5 –2 Los Angeles, California, 2006. –3 –1.0 –4 –1.5 –5 –500 50100150 200 250 300 350 400 050100150 200 250 300 350 400 μs μs Figure 3. Comparison of experimental data with computed results for the 465-kA, 0.060-in. tube shot. (a) Hoop strain measure by the two strain gauges; (b) radial displacement measure by the four PDV gauges. Note that the current peaks at 30 μs and returns to zero at 100 μs, so it is the early time data that tests the coupling of electromagnetics and FY2008 Proposed Work hydrodynamics. The late time data is determined solely by momentum and material models. We have proposed future experiments using the same coaxial test fi xture to validate kink instabilities, Figure 4. Simulation of a bucking instabilities, and electrical coaxial load test shot. contacts. This image shows both Based on our success with this the magnetic fl ux density project we were asked to participate (vectors) and the current density (pseudo color). This in the Bore Life Consortium of the particular shot had slots cut Offi ce of Naval Research (ONR) in the inner tube so that it Railgun S&T Program. The consortium was a fully 3-D problem; is tasked with gaining an understand- axial symmetry did not ing of phenomena that affect railgun apply. bore life, such as thermal ablation, gouging, and arcing. We will continue to develop the ETM simulation capability, to use the ETM simulation capability to model the notional ONR railgun, and to investigate novel materials and diagnostics.

Lawrence Livermore National Laboratory 11 Research

John M. May Petascale Simulation (925) 423-8102 Initiative [email protected]

uture supercomputers will gain combined or augmented to form larger Fperformance mainly from increased federations. parallelism, rather than faster proces- A Cooperative Parallel federation sors. To use these machines effectively, (Fig. 1) begins with the execution of a application developers must fi nd new single software component, which is a ways to exploit massive parallelism. At sequential or SPMD parallel program. present, most parallel applications use a Any process within this component may single program, multiple data (SPMD) launch additional components running model, in which each processor simul- different executables, and those may taneously applies the same algorithm launch others, and so on. Components to a different portion of the workload. communicate through remote method This approach is conceptually simple, invocation (RMI), in which a thread but balancing the workload to maximize of control within one component may effi ciency can be tricky, especially if the invoke a method (i.e., call a function) in workload distribution changes as the another component. Cooperative Paral- computation proceeds. Moreover, load lelism is a loosely-coupled model of balancing becomes harder with increas- computing that permits components to ing parallelism. be designed independently of each other The Petascale Simulation Initia- while making the interfaces through tive has developed a new programming which they interact more explicit. model called Cooperative Parallelism We have also investigated a tech- that lets applications manage and distrib- nique called Adaptive Sampling that can ute work more fl exibly. It is a multiple greatly improve performance in a broad program, multiple data (MPMD) model, class of multiscale and multi-physics so different parts of a parallel applica- simulations. This technique supports tion can work concurrently on different simulations in which coarse-scale tasks running different executables. computations are augmented with data Cooperative Parallelism is designed from costly fi ne-scale computations. to complement the SPMD model, so The application can assign the two kinds existing parallel applications can be of work to separate pools of processors

Parallel component using MPI internally Figure 1. Cooperative Par- allel application, consisting of multiple components that can be launched and terminated as needed. Each component is an executing sequential or parallel program. Components communicate Single task Ad hoc component through RMI. creation and Component wrapper communication Launch Remote method Runtime system Invocation (RMI)

12 FY07 Engineering Research and Technology Report Engineering Modeling & Simulation

(Fig. 2). That step by itself can im- can adopt incrementally to create more Its task-parallel model lets applications prove load balance, since a pool of fi ne sophisticated simulations and better exploit multiple dimensions of parallel- scale “servers” can provide data to any utilize massively parallel computers. ism at once. It also offers a natural route processor in the coarse-scale computa- The project also seeks to demonstrate to designing simulations that combine tion that needs it. However, even more that Cooperative Parallelism can benefi t multiple interacting models. Moreover, substantial performance gains arise a range of applications of interest to because cooperative parallelism can when the simulation caches fi ne-scale LLNL by allowing them to adopt a coexist with current SPMD parallelism, data and uses it to satisfy future requests federated model to complement standard applications can adopt it incrementally. from the coarse-scale model instead of SPMD parallelism. Cooperative Parallelism will also benefi t recalculating it. users of smaller-scale simulations that Relevance to LLNL Mission are executed repeatedly for parameter Project Goals Cooperative Parallelism should ben- studies or optimization studies. Our We aim to create a programming efi t large-scale codes as developers seek model greatly simplifi es the process of model that large scientifi c applications effi cient ways to use massive parallelism. orchestrating multiple instances of a simulation running concurrently.

FY2007 Accomplishments and Results The Petascale Simulation Initiative concluded its three-year term in April 2007. We designed the Cooperative Server Parallel model, implemented an proxy application-independent approach to adaptive sampling, developed a Well-balanced work Servers for unbalanced work working prototype of the Cooperative Figure 2. Adaptive sampling. Applications can be factored into well-balanced and imbalanced portions, with each portion Parallelism runtime system and demon- assigned to a different pool of nodes. When processes in the well-balanced portion need additional data, they can send strated a multiscale materials modeling a request using RMI through a server proxy, which assigns the work to an availble server component. simulation (Fig. 3) using adaptive sampling on more than 1,000 processors. The project also demonstrated its utility as a means for simplifying parameter studies through collaboration with LLNL’s PSUADE project.

Related References 1. Ashby, S., and J. May, “Multiphysics Simulations and Petascale Computing,” Petascale Computing: Algorithms and Applications, D. Bader, Ed., Chapman & Hall/CRC Press, 2007. 2. Barton, N., J. Knap, A. Arsenlis, R. Becker, R. Hornung, and D. Jefferson, “Embedded Polycrystal Plasticity and In Situ Adaptive Tabulation,” Int. J. Plasticity, in press, 2007. 3. Kumfert, G., J. Leek, and T. Epperly, “Ba- bel Remote Method Invocation,” IEEE Int. Parallel and Distributed Processing Symp., March 2007.

Figure 3. A multiscale material model of a part of an expanding cylinder, showing pressure on the left, plastic strain rate FY2008 Proposed Work in the middle, and the number of ﬁ ne-scale evaluations required on the right. Green indicates one evaluation; red indi- Further work, including closer cates two or more evaluations. The locations requiring these extra calculations track loading changes induced by passage collaborations with major LLNL simu- of waves through the material. Data are shown for a time increment over which a relatively large number of evaluations lation codes, is anticipated in FY2008. are performed. Parallel domain decomposition is also indicated by light gray boundaries in the plot on the left.

Lawrence Livermore National Laboratory 13 Technology

Daniel White Electromagnetics Code (925) 422-9870 Management [email protected]

LNL’s EMSolve code is a 3-D, paral- Relevance to LLNL Mission problems as well as a more useful and Llel, fi nite element code for solving EMSolve can perform electromag- robust interface for all of our linear solv- Maxwell’s equations. EMSolve, which netic analyses that cannot be performed ers (Fig. 1). has been used in support of many LLNL by commercial codes. Having this Effi ciency Enhancements. We imple- programs, has modules for electrostat- unique computational EM capability mented a specialized simulation code ics, magnetostatics, eigenvalues, eddy allows LLNL to have a competitive that can model electromagnetic wave currents, and wave propagation. The edge. Increasing the accuracy and ef- propagation using the scattered fi eld purpose of this project is to verify, docu- fi ciency of our CEM codes will benefi t formulation of Maxwell’s coupled fi rst- ment, and maintain the EMSolve suite of all customers. order fi eld equations. This code is tuned computational electromagnetics codes. to model blocks of materials arranged FY2007 Accomplishments and Results in Manhattan geometries with varying Project Goals AMS. In 2006 the HYPRE iterative permittivity, permeability, and electrical The goals for FY2007 are: solver library introduced an experimen- conductivity. 1. install EMSolve on the new Peleton tal multigrid solver called the Auxiliary This code was used to model several machines: Yana, Zeus, Hopi, and Space Maxwell Solver (AMS) for linear problems for the DARPA VisiBuilding Rhea; systems involving edge elements and the project which endeavors to determine 2. incorporate a new linear equation corresponding CurlCurl operator. The the location and makeup of interior solver into EMSolve; goal of this project was to evaluate this walls of buildings using radar tech- 3. set up effi ciency enhancements for solver’s effectiveness on our class of niques. EMSolve was used to evaluate an important class of applications problems and, should it prove effective, the importance of correctly modeling described by a tensor product mesh; to incorporate it into the EMSolve suite. cinderblock voids and metal reinforcing and We produced an iterative solver that rods within walls. Using these full-wave 4. collaborate with Ohio State Uni- is tuned specifi cally for edge-element techniques, we were able to show that versity (OSU) on local higher-order discretization of curl-curl equations. solid walls and fl oors produce strong absorbing boundary conditions The AMS is applicable to magnetostatic wave guiding effects that are virtually (ABCs). problems, eddy current problems, and destroyed by the presence of voids or implicit full-wave problems. The solver rebar within those structures. is based on multigrid concepts and is Using this new code we performed Mass + stiffnes solver tests 10000 almost scalable, meaning that the com- extraordinary computations (Fig. 2) on putation time increases linearly with the the Zeus machine. The computations PCG/AMS sec ) 1000 problem size. consisted of a series of four simula-

–10 PCG/DS sec 100 Incorporation of AMS into EMSolve tions of a two-story building with and was not a trivial procedure because without cinderblock voids and metal 10 AMS requires additional information reinforcing rods resolved to 1 cm. Each 1 beyond the system matrix. It also of these simulations required over 10

Solve time (tol = 10 0.1 requires information about the edge- billion mesh elements (60 billion ﬁ eld to-node connectivity of the mesh and unknowns) and 2.1 TB of memory. 0.01 1000 10000 100000 1000000 coordinates of the individual nodes. This Figure 3 shows a cut-away view of a Number of degrees of freedom necessitated augmenting the generic radar pulse propagating in the two-story solver interface and modifying a large building. Figure 4 is a close-up of the Figure 1. Improved run time of AMS solver vs. diagonally scaled preconditioned conjugate gradient (PCG). For a number of modules that make use of this geometry of the building. 1-million-unknowns problem, the AMS solver is approxi- interface. Local ABCs. Electromagnetic radia- mately 10 times faster than the existing solver. The end result is that the EMSolve tion and scattering problems require suite of codes now has a new and effec- that an ABC be applied to the bound- tive multigrid solver for electromagnetic ary of the mesh. The ABC is often the

14 FY07 Engineering Research and Technology Report Engineering Modeling & Simulation

Figure 2. Comparison of radar scattering in a solid cement wall building (left) vs. a rebar reinforced cinder block building (right). Note that this is a slice through a full 3-D simulation. The color represents the magnitude of the electric ﬁ eld: red is maximum and blue is minimum.

Figure 4. Close-up view of the modeling detail. Custom software was used to generate the mesh for the 10-billion-element simulations.

1.0 Figure 5. Verifi cation of the 1st order ABC – theoretical TE 0.8 second-order local ABC. The 2nd order ABC – theoretical solid lines are the theoretical 1st order ABC – EMSolve 0.6 results; the circles represent 2nd order ABC – EMSolve the computed results. Note 0.4 how the second-order ABC is signifi cantly better than the 0.2 fi rst-order ABC for off-normal Reflection coefficient, R Reflection coefficient, incidence, providing more 0 attenuation (y-axis) over a 040608020 Incident angle, Q (°) greater range of incidence angles (x-axis). .

FY2008 Proposed Work There are three tasks for FY2008. Figure 3. Cutaway snapshot of a radar pulse propagating through the building. We will collaborate with OSU on Note how the walls themselves act as waveguides. The fi eld propagates within 1) higher-order local ABCs, with the the wall at a slower velocity compared to free-space. goal of constructing a time-domain version of the algorithm; and 2) an algebraic domain decomposi- limiting factor in achieving an accurate boundary conditions, which require tion preconditioner for Helmholtz simulation. This task was to incorporate an integral over the entire bounding equations. We will also construct recent work on higher-order local ABCs surface. Figure 5 shows verifi cation of software infrastructure to support into EMSolve. The term “higher-order” the implementation of the second-order hp-refi nement. The challenge is refers to the performance of the ABC for local ABC in EMSolve. The addition of to implement the algorithm in an off-normal incidence. The term “local” this new capability will result in more existing code, rather than starting means that the ABC requires only local accurate simulations with little increase from scratch. This effort will result in fi eld information, in contrast to “global” in computational cost. a simulation capability that is much more robust and easier for end-users.

Lawrence Livermore National Laboratory 15 Technology

Bob Corey Finite Element Analysis (925) 423-3271 Visualization and Data [email protected] Management

Mili base API Mili extensions API key component of our project is its Xmilics is a utility used to combine (includes advanced A support for post-processing and results that are generated by our large Legacy Silo features) visualization tools, including the Griz parallel computing platforms. fi nite element visualization post-proces- MiliSilo 1.0 library sor, the Mili data management library, Project Goals MiliSilo interface (legacy) and a data utility tool called Xmilics. The project provides ongoing support These tools are used by analysts and for visualization and post-processing engineers across LLNL to interpret tools and adds new capabilities to these Silo library interface (Mili independent) data from a variety of simulation codes tools to support evolving, multi- such as DYNA3D, ParaDyn, NIKE3D, programmatic requirements. Figure 1. Mili 1.0 and 2.0 API hierarchy. TOPAZ3D and Diablo. Griz is our primary tool for visual- Relevance to LLNL Mission izing fi nite element analysis results on These post-processing tools pro- 2- and 3-D unstructured grids. Griz vide important user interfaces for our provides advanced 3-D visualization simulation capabilities and are critical techniques such as isocontours and elements in our tool suite. Analysts isosurfaces, cutting planes, vector fi eld would otherwise be severely limited in display, particle traces, and free-particles their ability to interpret the vast amounts or free-node approaches. Mili is a high- of data generated by simulations, and to level mesh I/O library intended to support synthesize key results. computational analysis and post-processing on unstructured meshes. It provides the FY2007 Accomplishments and Results primary data path between analysis codes User support continues to be a high- and Griz. Mili databases are also view- priority goal. The group is currently Figure 2. Engineering model as viewed in VisIt and able with the LLNL VisIt post-processor. supporting approximately 30 to 40 active generated from new Mili/Silo Library. users on a variety of platforms across LLNL and some off-site users, including LANL. This year we saw a high level of usage and special requests from users and a higher number of analysts using Griz and VisIt concurrently to meet their needs. We made signifi cant progress in migrating Mili to a modern fi le structure. We defi ned and documented the requirements. We then created a useful model, choosing the LLNL-supported Silo format, because of its application programming interface (API) and its compatibility with the VisIt visualization tool. The architecture is hierarchical with a separation between the Silo and Mili layers (Fig. 1). Both new and legacy interfaces are maintained in one library to provide backward compatibility and reduce the migration impacts to our simulation codes. Figure 2 shows an example of a model written using

16 FY07 Engineering Research and Technology Report Engineering Modeling & Simulation the new Mili/Silo format and rendered Collaboration continued with explor- directly with VisIt without a conver- ing “meshless” techniques having an sion utility. This is one example of our unstructured cloud of particles with no FY2008 Proposed Work ongoing effort to integrate Mili database continuum elements. The results are We will continue to provide support directly with VisIt. treated in Griz in much the same way as support for our user base. Efforts We added a variety of new visualiza- the “free-node” capability delivered in targeted for next year include: tion features to Griz, with the most sig- FY2006 to support the Department of 1) releasing the Mili/Silo Library nifi cant being: 1) a capability to render Homeland Security. This year we added for general usage with streamlined models in wireframe mode and a com- capabilities to plot any typically nodal or connections to VisIt; 2) adding bination of wireframe and solid (Fig. 3); hex result, such as displacements, onto support for higher-order elements to 2) a capability to render materials with their associated particles (Fig. 5), and Mili and Griz; 3) completing the static undefi ned results in grayscale (Fig. 4); to select individual particles and view analysis of Griz for potential defects 3) the capability to generate JPEG and a result. Software quality engineering and performing the same analysis for PNG graphic fi les on nearly all Griz- efforts expanded to include the use of Mili; and 4) implementing automated supported platforms; and 4) the ability to a commercial tool to create a baseline regression testing for Mili. use Griz on the new Windows platforms. assessment of potential defects.

(a) (b) –1 Global maximum: 6.24 x 10 . Bricks 1059 X stress Global minimum: –6.31 x 10–1. Bricks 963 7.63x10-1 6.00x10-1 Materials Displacement scale: 1.0/1.0/10 4.00x10-1 Ref frame: global 2.00x10-1 2 -8 3 1.49x10 –2.00x10-1 –4.00x10-1 –6.00x10-1 –8.08x10-8

Tubes Tubes t = 3.00004 x 10–3 (state = 4/4) t = 3.00004 x 10–3 (state = 4/4) Figure 3. (a) Rendering of model in wireframe mode. (b) Example of wireframe and solid rendering mode.

Z displacement Global maximum: 6.21. Node 24695 6.21x10 Global minimum: –2.85 x 10–1. Node 27717 5.00x10 Displacement scale: 1.0/1.0/10 4.00ex10 Ref state: initial geom 3.00x10 2.00x10 1.00x10

–2.85x10

–3

Mon Sep 17 11:14:35 2007 Mon Sep 17 11:14:35 t = 2.10003 x 10 (state = 211/211)

Figure 4. Rendering of results with inactive materials as grayscale. Figure 5. Example of plotting nodal displacement result onto particles.

Lawrence Livermore National Laboratory 17 Technology

Jerry I. Lin Rotational Kinematics Output (925) 423-0907 and Other Improvements [email protected] in DYNA3D

YNA3D is a main engineering tool documentation updates, and Software Dfor the simulation of transient Quality Assurance (SQA) compliance response of solids and structures to fast, for DYNA3D. impulsive loadings. This explicit ﬁ nite This project also supports the element code was originally created to broader interagency DYNA3D activi- support weapons activities, but over the ties through LLNL’s Collaborator Pro- years has found broader use in other gram. The Collaborator Program grants applications, such as infrastructure vul- access to selected licensed users to nerability and protection. The code also LLNL’s computational mechanics/ther- provides the mechanics functionalities mal codes in exchange for their infor- of the highly parallel ParaDyn simula- mation, results, and acknowledgement. tion tool. This project funds the ongo- These collaborative members include ing implementation of user-requested our sister laboratories, U.S. government features, general technical support, agencies, and other institutions.

Figure 1. Spinning hollow projectile Project Goals striking a layered target. The planned tasks include the implementation of functionalities for various programs needs, enhancement of existing features, the addition of new result display capabilities, and continued compliance work on SQA and the Fortran 95 standard.

Relevance to LLNL Mission Many Laboratory programs require new functionalities and technical support to complete their missions. Some of these programs and projects involve the Laboratory’s collaboration with other institutions and federal agencies, such as the Los Alamos National Laboratory, Department of Homeland Security, Bureau of Reclamation, U. S. Army Corps of Engineers, the Naval Surface Warfare Center, and selected universities.

FY2007 Accomplishments and Results Based upon multiple requests, rotational velocity and acceleration histories for an entire model, or parts of the model, were added. These quantities can be output to the full-state visualization database, or designated as

18 FY07 Engineering Research and Technology Report Engineering Modeling & Simulation

–90 Figure 2. Time history of the projectile rotational velocity about Material group 1 its principal axis. Note that the projectile slows down and partially recovers its spin as it interacts with the target. –100

–110

–120

–130

–140

–150

–160

Group velocity (rotational about chosen vector) –170

–180

–190

–200 0 1 x 10–3 2 x 10–3 3 x 10–3 4 x 10–3 5 x 10–3 6 x 10–3 Time FY2008 Proposed Work Targeted efforts include general one of the limited number of quanti- code PENCURV3D, contains layers of technical support for DYNA3D users, ties sent more frequently to a dedicated materials of different stiffness. Figure 2 the addition of user-requested capa- time history database. shows the projectile’s change of spin ve- bilities, and the ongoing moderniza- These results are made available in locity about its principal axis as it interacts tion toward Fortran 95-compatible terms of components in the global co- with the target. and SQA-compliance work. ordinate system or about a user-defi ned The Rayleigh damping features in A modifi ed Mili surface capability orientation. The user-defi ned axis may DYNA3D have been effective tools for with more essential attributes is near move arbitrarily in space over the simu- achieving quasistatic response, stress its completion. This will be integrated lation time. These quantities supplement initialization, shock noise reduction, into DYNA3D and enable many addi- their existing translational counterparts and other applications. We made the tional state variables and parameters in providing the users more detailed mass-proportional part of the damping to be included in the output databases understanding of the structure’s overall more versatile by excluding selected for visualization by analysts. motion or relative motion between parts. rigid bodies at the user’s choice. This We will evaluate the possibility A generic hollow projectile with a capability, combined with the existing of identifying element facets that blunt nose impacting an inclined fl at target rigid-deformable material switching belong to a modelʼs exterior surfaces. surface, shown in Fig. 1, is used as an feature, provides a valuable means to These facets may be assigned special example. The projectile, modeled by fi nite avoid undesirable damping-induced attributes or properties for strength- elements with isotropic elastic properties, motion slowdown. For the stiffness- ened numerical stability. Targeted is traveling with an initial translational proportional part of this feature, a more applications include augmented velocity along with a rotational velocity user-friendly input allows direct speci- hourglass control and contact surface about its principal axis. The target, mod- fi cation of fraction of critical damping, smoothing. eled by the penetration load-predicting and damping frequency is added. .

Lawrence Livermore National Laboratory 19 Technology

Michael A. Puso NIKE3D Code Maintenance (925) 422-8198 and Enhancement [email protected]

he objective of this work is to Such “just in time” enhancements in- Tenhance, maintain, and support the cluded new material model capabilities implicit structural mechanics fi nite ele- for weapons analysts and new contact ment code NIKE3D. This tool is LLNL’s analysis features. primary capability for simulating the slowly evolving or steady-state response Relevance to LLNL Mission of nonlinear solids and structures. New Structural analysis is one of the features are continually added to ac- most important functions of Engineer- commodate engineering analysis needs. ing and motivates in-house maintenance Maintenance includes bug fi xes and for LLNL’s suite of codes. NIKE3D, in code porting to new or updated plat- particular, is a premier code for handling forms. User support includes assisting diffi cult nonlinear static structural analy- analysts in model debugging and general sis problems. It has the most diverse and analysis recommendations. robust contact algorithms at the Labora- tory to handle the complex interactions Project Goals of unbonded material interfaces. Code enhancement requires continuous interactions with users as well FY2007 Accomplishments and Results as new features to meet our evolving A number of new element technolo- Figure 1. (a) Bending stress shown on 20- needs. Each year some activities are gies and discretization technologies have node model of loaded cantiliever beam with (curved) contact surface separating top and pre-planned, e.g., completing production been developed at LLNL, most notable bottom. (b) Convergence plot demonstrating versions of higher-order elements and being higher-order elements (10-node optimal quadratic convergence of the energy nodally integrated tetrahedrals. Other ef- tetrahedrals, 20- and 27-node hexahe- norm of the discretization error. forts arise in response to real-time needs. drals), nodally integrated tetrahedrals, (a) and meshless methods. Efforts are required to make them fully integrated functionalities available for production use by analysts. For example, these implementations exploited more modern, hierarchical data structures available in Fortran 95. NIKE3D’s existing technique of writing a monolithic restart database did not work with these new features. (b) –2.2 Instead, the data needed to be written 20-node brick results and read recursively to perform this –2.4 task for each data structure. The process –2.6 of writing these dedicated recursive algorithms was automated using a PERL –2.8 script adapted from one fi rst created by –3.0 the Diablo project and then adapted for new data structures in DYNA3D. Log (energy of error) –3.2 O(h2) To fi nalize the production capability –3.4 of higher-order elements, modifi cations were made for using GRIZ to visualize –3.6 –0.7–0.6–0.5–0.4–0.3 –0.2 –0.1 0 0.1 higher-order element results. Previously, Log (h)

20 FY07 Engineering Research and Technology Report Engineering Modeling & Simulation higher-order elements results could be solving the nonlinear system of equations visualized only by GID. Now TrueGrid in NIKE3D and is a keystone to its suc- and Cubit can be used to build quadratic cess. Nonetheless, there is much room for FY2008 Proposed Work element meshes, and our quadratic improvement. As part of this effort, a new Our existing mortar contact mortar contact algorithms can be used Armijo-type line search algorithm was implementation currently uses an N2 without loss of quadratic convergence added along with several small modiﬁ ca- search algorithm. In FY2008, we will rate (Fig. 1) for nonlinear analysis us- tions to improve solution robustness. implement better and faster search ing higher-order elements (Fig. 2) in algorithms such as the bucket sort. NIKE3D. Related References We will implement an improved New material model features were 1. Puso, M. A., and T. A. Laursen, “A Seg- quasi-Newton method. A number of added in support of weapons analysis. ment-To-Segment Mortar Contact Method new methods have been created in One example is a new thermal-plastic for Quadratic Elements and Large Defor- the area of optimization that could model that captures thermal material mations,” Computer Methods in Applied be converted to treat computational response over an arbitrary range of tem- Mechanics and Engineering, in press. contact mechanics. peratures (Fig. 3). Finally, a number of 2. Christensen, P. W., et al., “Formulation contact features were added. Most nota- and Comparison of Algorithms for Frictional bly, new work has begun on the improve- Contact Problems,” International Journal for ment of the quasi-Newton contact solver. Numerical Methods in Engineering, 42, The quasi-Newton method is used for pp. 145-173, 1998.

(a)

Metal puck

Graphite mold (edges only)

(b) (b) 0.04 D D' Mortar contact results E 0.02 Classic node-on-surface 6 results 200 x 10 0.06 G 0

150 Strain B –0.02 0.04 Yield stress

Stress (Pa) 100 CTE_strain –0.04 Neck effective stress 0.02 Horizontal interface force Neck plastic strain 50 –0.06 A

–0.08 00 0 5 10 15 20 25 400 600 800 1000 1200 Time Temperature (K) Figure 2. (a) Simulation of elastic sphere being forced out of Figure 3. (a) Metal “puck” cast inside graphite mold (3-D solid mesh not shown). pressurized elastic tube. Twenty-node brick elements are used. (b) Evolution of stress and strain at critical location (red circle in (a)) in casting as (b) Horizontal force on tube versus time. This force is mainly due to metal is cooled. The thermal phase changes indicated in yield stress and CTE strain frictional resistance. Force is very smooth for mortar contact results, are provided in the arbitrary format of the new material model feature. whereas node-on-surface force is not.

Lawrence Livermore National Laboratory 21 Technology

Steven W. Alves Validation of Reinforced (925) 423-2391 Concrete Modeling Capabilities [email protected] for Seismic Response

his project assesses the ability of Engineering Simulation. The structure comparison with the shake table experi- Tour explicit solid mechanics codes was subjected to four earthquake ground ments. The latter included refi nement DYNA3D/ParaDyn to simulate the motions of successively increasing of the boundary conditions and material seismic response of reinforced concrete amplitude. properties in the fi nite element model structures. These codes were originally to obtain a better representation of the created to simulate rapid, large- Project Goals structure. deformation response. For example, the Previous work has shown that there concrete model was created to assess are diffi culties in modeling seismic re- Relevance to LLNL Mission shock loadings on reinforced concrete sponse of a reinforced concrete structure Simulation of the seismic response structures. In contrast to shock loading, using the homogenized concrete/rebar of reinforced concrete structures is of structures subjected to seismic load- model implemented in DYNA3D/ importance to LLNL’s Nuclear Fuel ing will likely go through multiple load ParaDyn. An instability in the model Cycle & Reactor Program, the Global reversals over a longer response period. implementation leads to non-physical Nuclear Energy Partnership (GNEP), and Recent shake table tests at the stress states and subsequent inversion of the National Ignition Facility (NIF). In University of California, San Diego elements that terminates the simulation. general, validating this capability would (UCSD), provide a signifi cant assess- This misbehavior appears to be more allow LLNL to attract new projects in the ment case. A full-scale seven-story slice likely with the long duration and load area of seismic analysis. This project also of a reinforced concrete building was reversals necessary for a seismic simula- promotes collaboration with UCSD. constructed on the Large High- tion. Performance Outdoor Shake Table The goals of this project were to FY2007 Accomplishments and Results (LHPOST) at UCSD (Fig. 1). This facil- improve the material model and then The numerical robustness of the ity is part of the Network for Earthquake continue the assessment of the model by concrete model in DYNA3D was investigated. The code failure was determined to occur during calculation of a quantity related to the effective plastic strain increment. This error condition can be eliminated by scaling the stress in an element to zero. This improvement was implemented, but it should be noted that large effective plastic strain increments can still be calculated for an element causing large deformations. While this does not prevent simulations from running, the user must assess whether the computed response is signifi cantly affected before using the results. Other corrections to the material model, which were identifi ed during the investigation of the instability, were also implemented. A fi nite element model of the structure was generated with separate concrete properties for each level of the building. The properties were based on Figure 1. Seven-story structure on shake table. Figure 2. Finite element model of the structure and concrete test data provided by UCSD. shake table. Also, rotational stiffness of the shake table platen was included in the model.

22 FY07 Engineering Research and Technology Report Engineering Modeling & Simulation

motion (Fig. 4) unless the rotational (a) (b) stiffness of the table platen is reduced to a level that is a lower bound for the rotational stiffnesses estimated by UCSD from recorded motions. While the agreement of the modeled response with the experiment is good for the fi rst earthquake, the simulation devi- ates for the subsequent earthquakes by under-predicting the response (Fig. 4). A possible explanation for this is that as the structure is progressively damaged and the stiffness decreases, the prescribed damping becomes too high in the model. It is believed that these results partially validate the use of the homogenized concrete/rebar model in DYNA3D/ ParaDyn for seismic simulations of reinforced concrete structures. While the numerical instability causing code Figure 3. Longitudinal mode at 2.51 Hz with no table (a) and at 1.97 Hz with table rotational stiffness (b). termination has been corrected, there is still concern about the large deformations caused by the remaining numerical Both of these features are believed to the model. The fi rst longitudinal mode anomaly. The approach is believed to be be important factors in modeling the of the structure ranges from 2.51 Hz viable for simulations to determine mod- seismic response of the structure on the with no table rotation to 1.97 Hz with a erate levels of deformation and damage, shake table. relatively low rotational stiffness of the but performing simulations for long dura- A linear version of the fi nite ele- table (Fig. 3). tions after sustaining signifi cant deforma- ment model was created for running Dynamic simulations were per- tion and damage levels is of questionable static analyses and then determining the formed using ParaDyn for a fi nite value. Obviously, attention must be paid natural frequencies and mode shapes element model with concrete and rebar, to accurately defi ning boundary condi- using NIKE3D (Fig. 2). These analyses using the nonlinear material models. tions, material properties and damping. illustrate the importance of including the The stiffness of the structure is not Perhaps damping should be modifi ed as rotational stiffness of the table platen in captured during the fi rst earthquake the structure is progressively damaged.

(a) Roof Displacement – EQ1 (b) Roof Displacement – EQ2 2.5 6

-2 Measured Measured Simulated 4 Simulated -1.5

-1 2 0.5

0 0

-0.5 Displacement (in.) Displacement (in.) -2 -1-

-1.5 -4 -2

-2.5 -6 0510 15 20 25 30 0510 15 20 25 30 Time (s) Time (s) Figure 4. Comparison of simulated roof displacement to measured roof displacement for the ﬁ rst (a) and second (b) earthquakes. (Note difference in vertical scales.)

Lawrence Livermore National Laboratory 23 Technology

Charles R. Noble Thermal-Structural Analysis of (925) 422-3057 the MacArthur Maze Collapse [email protected]

t approximately 3:45 AM on the MacArthur Maze connector using LLNL Amorning of April 29, 2007, a trac- thermal and structural fi nite element tor-trailer rig carrying 8,600 gallons of simulation codes TOPAZ3D, Diablo, fuel overturned on the connector from NIKE3D, and ParaDyn. Other goals Interstate 80 to Interstate 880 in Oak- included: land, California. The accident resulted 1. validating that TOPAZ3D and in an intense fuel fi re that weakened Diablo obtain similar results for a the steel superstructure above it and thermal analysis; initiated connection failures between the 2. validating that NIKE3D, Diablo, and bridge deck and main girder, causing ParaDyn obtain similar results for a an approximately 50-yard long section structural analysis; and of the connecting ramp to collapse onto 3. performing a fully-coupled thermal- the lower connector. It was determined structural analysis using Diablo in a that studying the structural collapse of parallel computing environment. the MacArthur Maze would provide the opportunity to: Relevance to LLNL Mission 1. study the methods needed to prop- Validation of computational software erly stimulate the themal loading on for predicting progressive collapse of a structure from this type of event; large structural systems is extremely 2. determine the predictive capability of diffi cult. The ability to perform progres- LLNL’s fi nite element software for sive collapse simulations of critical progressive collapse analysis; and infrastructure is important for Homeland 3. validate this type of thermal/ Security applications, and therefore structural failure analysis, which relevant to the mission of LLNL. would then assist in determining the threat to other critical infrastructure FY2007 Accomplishments and Results in the United States. A thermal model was successfully developed for TOPAZ3D and Diablo that Project Goals showed equivalent calculated temperature The end goal of this project was values within 2% at key locations in the to simulate progressive collapse of the model. Figure 1 shows the temperature distribution at one hour for a fl ame tem- DB: m_plot.mili z Figure 1. Simulated temperature Cycle: 60 Time: 3600 perature of 1200 °C using TOPAZ3D. distribution from a stationary fi re at Pseudocolor x Var: tempC The thermal model approximated the one hour using the code TOPAZ3D. 1200. y fi re as a box-shaped region for determina- 905.0 700 Z– Axis tion of heat transfer to the structure. This 400 610.0 3 0 X– Axis (x10 ) “fl ame box” extended from the underside 315.0 12.2 3.0 of the concrete roadway down to sixteen 20.0 3.5 4.0 feet elevation below the center of the main Max: 1221. 12.4 Min: –1.959 4.5 box girder. The lateral position, orientation, 12.6 and shape of the fl ame box were adjusted as a function of time corresponding to 12.8 video footage from the East Bay Municipal Utility District. The amount of radiative 13.0 3 and convective transport from the fl ame Y– Axis (x10 ) box to a given surface facet at a given time

24 FY07 Engineering Research and Technology Report Engineering Modeling & Simulation

Pseudocolor DB: maze2plt.mili ParaDyn (32 processors)Cycle: 35 Time:1000 Diablo (24 processors) Pseudocolor Var: derived/displacement/z Var: derived/displacement/z 0.001000 0.000 –0.1250 –0.1242 –0.2500 –0.2495 –0.3750 –0.3748 –0.5000 –0.5000 Max: 0.007080 Max: 0.001081 Min: –0.5239 Min: –0.5180

0.001000 –0.1242 –0.2495 –0.3748 –0.5000 Y Max: 0.008179 Min: –0.5087 X Pseudocolor DB: m_maze2g.mili Z Cycle: 38 Time:1000 Var: derived/displacement/z

NIKE3D (8 threads)

Figure 2. Overhead view of MacArthur Maze connector under pure gravity load. The three plots show all three LLNL structural codes providing the same resulting displacement. The red is 0.0 in.; the blue represents a vertical displacement of –0.5 in. For comparison, the peak CalTrans calculated displacement was –0.5 in. for the right span. was determined based on the facet and structural analysis in a fully-coupled Related References fl ame box location and orientation, where manner. Gravity loading was applied to 1. Buyukozturk, O., and F. J. Ulm, Materi- radiative view factors and an applied ther- the structure prior to performing the ther- als and Structures, pp. 83-105, MIT Press, mal decay length for the convective gas mal-structural analysis. Figure 2 shows 2002. temperature were used. a displacement comparison of the model 2. Chen, J., and B. Young, “Experimental Structural models for the MacArthur under gravity load for the LLNL struc- Investigation of Cold-Formed Steel Mate- Maze were developed for NIKE3D, tural fi nite element codes. Not only do rial at Elevated Temperatures,” Thin-Walled Diablo, and ParaDyn. Both NIKE3D and all three fi nite element codes give similar Structures, 45, pp. 96-110, 2007. ParaDyn have the ability to incorporate results, but the codes predict similar verti- 3. Outinen, J., and P. Makelainen, “Me- the thermal loads calculated within cal displacements to the calculated dis- chanical Properties of Structural Steel at TOPAZ3D for use in the structural placements of the California Department Elevated Temperatures and After Cooling analysis. However, this methodology of Transportation (CalTrans). Figure 3 Down,” Fire Mater., 28, pp. 237-251, 2004. does not account for transient structural shows the simulated collapse of the Ma- 4. Takagi, J., and G. Deierlein, “Strength deformation effects on the calculated cArthur Maze connector using TOPAZ3D Design Criteria for Steel Members at thermal profi les. for the thermal loading and ParaDyn for Elevated Temperatures,” Journal of Diablo, on the other hand, has the structural loading, assuming a fl ame Constructional Steel Research, 63, the ability to perform the thermal and temperature of 1200 °C. pp. 1036-1050, 2007.

Figure 3. Simulated collapse of the MacArthur Maze connector using TOPAZ3D for the thermal loading and ParaDyn for the structural loading. (Photo courtesy of San Francisco Chronicle.)

Lawrence Livermore National Laboratory 25 Technology

Andrew Anderson Enhanced Composite (925) 423-9634 Modeling Tools [email protected]

omposite materials are used in compressive failure. Several experi- Cmany advanced weapons systems ments were conducted to provide data and structures at LLNL. We have previ- for the verifi cation and validation of the ously enhanced our ability to simulate model’s implementation in ALE3D. structural response and progressive failure of composite systems in ALE3D Relevance to LLNL Mission (an arbitrary Lagrange/Eulerian multi- The improved fi ber composite mate- physics code developed at LLNL) by rial models can be used in simulations porting an existing composite con- (to failure) in the many LLNL programs, stitutive model (Model 22, the Fiber such as those for composite munitions, Composite with Damage Model) from armor penetration, pressure vessels, and DYNA3D (a nonlinear, explicit, 3-D rocket motors. This project has been FEM code for solid and structural benefi cial in supporting the composite mechanics). This year, a more advanced modeling efforts within the DoD Joint model (DYNA3D Model 62, the Uni- Munition Program and the Focused Directional Elasto-Plastic Composite Lethality Munition Program. This study Model) has been implemented. Experi- supports LLNL’s engineering core ments were conducted to validate the competency in high-rate mechanical de- elastic response of the model and to give formation simulations of large complex insights and data needed for the addition structures by providing an enhanced of a failure algorithm into the model. capability to model composite structures with ALE3D. Project Goals We implemented the Uni-Directional FY2007 Accomplishments and Results Elasto-Plastic Composite Model into The implementation of the Fiber ALE3D. This included implementing Composite with Damage model into the ability to input orthotropic orienta- ALE3D, which was completed in the fi rst tion data into prescribed local volume year of this project, was verifi ed with elements. Another modeling goal was several code-to-code comparisons. The to enhance the model by incorporating hoop stresses in pressurized cylinders a failure algorithm that includes matrix from simulations run with DYNA3D, delamination, fi ber tensile, and fi ber with the new Fiber Composite model in ALE3D, and with an existing anisotropic ALE3D model, all agreed within 1%. This included both explicit and implicit ALE3D runs. The Uni-Directional Elasto-Plastic Composite model was implemented into ALE3D. An important part of this task was creating an algorithm to initialize and update material directions at the ply and element levels. The model was validated using the same pressurized-cylinder simulations described above, and the results were found to closely match the Figure 1. Fiber composite compression cylinder with 1.0-in.-diameter pin. DYNA3D predictions.

26 FY07 Engineering Research and Technology Report Engineering Modeling & Simulation

(a) (b) Strain concentration factors due to fo- % (tech.) cused shear in composites were measured Epsilon Y (0-2) 0.15 0.10 using the specimen shown in Fig. 3. 0 This sample was loaded in compression to produce a concentrated shear band in –0.10 the composite sample. The Aramis load –0.20 strain curve is shown in Fig. 4. –0.30 Related References y –0.40 1. Christensen, R., and E. Zywicz, “A Three z x –0.50 Dimensional Constitutive Theory for Fiber –0.60 Composite Laminants,” Journal of Applied Figure 2. (a) Aramis axial stain results for fi ber composite compression cylinder with no pin at 300,000 lbs of load. Mechanics, 57, pp. 948-955, December (b) The ALE3D simulation. 1990. 2. Chang, F. K., and K. Y. Chang, “A Composite failure mechanisms can and resins. The data collected on the Progressive Damage Model for Laminated be divided into two types: intra-ply stiffness, Poisson’s ratio, and ultimate Composites Containing Stress Concentra- failure mechanisms, such as fi ber break- strength of each specimen provide tion,” Journal of Composite Materials, 21, age, matrix failure (cracking/crushing), model validation data for the newly pp. 834-855, 1987. and fi ber buckling; and inter-ply failure implemented models 22 and 62. The mechanisms involving ply delamination. data also provide an expanded source Intra-ply failure can be applied at of failure data for upcoming failure the ply level and so fi ts in well with this model validation in ALE3D. FY2008 Proposed Work model’s “unit cell” approach. Inter- Strain concentration factors in fi ber In a proposed follow-on project, ply failure that includes crack opening composite cylinders with holes and bond- we will continue to improve fi ber com- between plies and plies sliding relative ed pins were measured using the Aramis posite modeling in ALE3D, with an to each other affects all layers simul- video strain measurement system. The emphasis on local bending response taneously, and so is more diffi cult to basic fi ber composite cylinder with pin and progressive damage. We plan to implement. All the relevant mathemati- confi guration is shown in Fig. 1. implement ply-level capabilities and cal expressions necessary for these func- Figure 2 shows a comparison for the damage algorithms taken from a tionalities have been derived, and the case of no pin (open hole) between the specialized LLNL ply-level composite corresponding changes to the existing measured experimental data and the code known as ORTHO3D, and verify code have outlined. Implementation will simulated response from ALE3D. The their implementation experimentally. be undertaken next year. results appear to be very similar. A series of compression tests to failure were conducted on eight different composite cylinder specimens MK82 shear 1 “0.01, 0.02 in./min with different fi ber, fi ber orientations, –10000 Strain averaged over: Center box Small diamond –8000

–6000

55 kip ld (lb) –4000

–2000

0 0 –0.5 –1.0 –1.5 Average axial strain (%) Figure 3. Composite shear specimen from section of a Mk82. Figure 4. Shear strain concentration in composite Mk82 shear specimen.

Lawrence Livermore National Laboratory 27 Technology

Charles G. EMP Simulation and Measurement Brown Jr. (925) 423-4435 Data Analysis in Support of the [email protected] Titan Experiments

lectromagnetic pulse (EMP) is a and analysis in this project, will be Eknown issue for a variety of situa- extrapolated to NIF and other short- tions including short-pulse laser facili- pulse lasers around the world to better ties such as LLNL’s Titan and Omega, mitigate EMP effects. and experts believe EMP will also be an issue with NIF. During FY2007, this FY2007 Accomplishments and Results project focused on improving simulation During FY2007, we gained further capabilities of EM fi elds due to elec- experience in performing simulations trons from laser/target interactions, and using the EMSolve processing chain, providing post-processing algorithms which consists of Cubit, EMSolve, and and software. VisIt. Cubit is a meshing tool from San- dia National Laboratories that is used to Project Goals construct models for EMSolve. VisIt is The overarching goal of this project an LLNL code for visualizing simulation is to simulate EMP in situations like the data. Titan short-pulse laser due to electrons We created more realistic CAD mod- from laser/target interactions and to els of the Titan chamber, including fi ner compare measurement results with the details such as optics stands modeled as simulations. To accomplish the simu- rods. Further, we augmented EMSolve lation task, we used EMSolve 3, an so that both the magnetic and electric LLNL EM solver. Unlike most avail- fi elds of electron beam sources can be able codes, EMSolve’s architecture simulated accurately. We also advanced allows seamless integration of user- our ability to post-process data measured created sources and boundary condi- by B- and D-dot probes, in particular tions, and EMSolve has been augmented deconvolution of the probe response. with electron beam sources. The deconvolution method we used is an optimal method based on the Relevance to LLNL Mission Wiener fi lter, which is implemented in An understanding of the EMP processing software. We have extend- in Titan, through the simulation ed the software tool and created a user

28 FY07 Engineering Research and Technology Report Engineering Modeling & Simulation

~1 m

0 ns 0.9 ns 1.8 ns

2.7 ns 3.6 ns 4.5 ns Time sequence of simulation of magnetic ﬁ eld due to propagating electron bunch.

interface called Time-Series Analysis Related References Tool (TSAT) that allows quick and 1. Mead, M. J., et al. “Electromagnetic Pulse easy manipulation of time-series data Generation within a Petawatt Laser Target FY2008 Proposed Work and includes data input, signal pro- Chamber,” Review of Scientifi c Instruments, Our FY2008 work will include cessing, and display capabilities. The 75, 10, pp. 4225-4227, October 2004. improving our CAD models to perform simulation and post-processing capa- 2. “VisIt Visualization Tool,” http://www.llnl. more realistic simulations; continu- bilities originating from this project gov/VisIt/home.html. ing to perform EMSolve simulations; have allowed us recently to publish 3. Candy, J. V., Model-Based Signal Process- assisting the experimental team in a comparison of a simulation of an ing, Wiley-IEEE Press, 2006. the measurement process to better electron beam in Titan and actual, 4. Brown, C. G., Jr., et al., “Electromagnetic understand the measurement system; measured data. Pulses at Short-Pulse Laser Facilities,” Pro- and working with signal processing The fi gure depicts frames of a time ceedings of the 5th International Conference experts to calibrate and analyze the sequence of the magnetic fi elds due on Inertial Fusion Sciences and Applications, data, particularly the data obtained to 1012 electrons in a Gaussian bunch Kobe, Japan, September 9-14, 2007. from the September 2007 dedicated with a full-width at half maximum of shots in Titan. 100 ps.

Lawrence Livermore National Laboratory 29 Technology

Christopher Walton Full-Process Computer Model (925) 423-2834 of Magnetron Sputter: Testing [email protected] Existing State-of-the-Art Components his work is the fi rst part of a larger the results (such as fi lm stress, density, FY2007 Accomplishments and Results Tproject to create a modeling capabil- and stoichiometry.) The simulation In the plasma simulations, we ity for magnetron sputter deposition. must use existing codes, either open- demonstrated an approach to a realistic The process is divided into four steps: source or low in cost. In FY2007 we plasma model, eliminated computational plasma transport, target sputter, neutral identifi ed and tested the best avail- heating, and discovered the computa- gas and sputtered atom transport, and able code for each process step, then tional limitations of the plasma code. fi lm growth, shown schematically in determined if it could cover the size Figure 1 shows the largest obtainable Fig. 1. Each of these is simulated sepa- and time scales we need for reason- domain size, 1% of real dimensions, rately in this part of the project. able computation times. We also had which features a grid of cells, 340 x 638. The plasma modeling was performed to determine if the process steps are We discovered that a 5% scale simulation using the Object-Oriented Particle-In- suffi ciently decoupled that they can using a grid of cells 1702 x 3192 crashes Cell code (OOPIC) from the University be treated separately, and identify any the code on a Windows platform, which of California, Berkeley. Figure 2 shows issues preventing practical use of these calls for future work in implementing the the electron density in the simulated codes. The second part of the project code on high performance machines and region, using magnetic fi eld strength will consider whether the codes can creating a parallel electrostatic solver. input from experiments where a scale be (or need to be) made to talk to each In the larger project we partly of 1% was used. Figures 3 and 4 depict other and integrated into a whole. validated models of the pressure and the magnetic fi eld components that were temperature distribution in the gas in generated using 2-D linear interpolation Relevance to LLNL Mission front of the sputter gun. This model of the experimental data. LLNL has multiple projects that enabled us to calculate arrival angles depend on precision fi lm deposition for and energies for sputtered Be atoms Project Goals critical goals. A prime example is con- passing through the gas. These were The goal of the overall modeling trol of fi lm microstructure and stress in then used as inputs for simulations tool is to understand, and later predict, 180-mm-thick sputtered Be metal shells of the growth of the Be fi lms. These relationships among parameters of fi lm for NIF targets. Others include stress simulations show columnar micro- deposition we can change (such as gas control in large-area sputtered nanolami- structure approaching a limiting grain pressure, gun voltage, and target- nate adaptive optics, and fi lm stoichiom- size, with dome-shaped grains and a substrate distance) and key properties of etry in high-power capacitors. gap developing at the grain boundaries, all consistent with experiment and of key importance in understanding fi lm 5 cm density. An example of this qualitative agreement is shown in Fig. 5. Target (Cu) In the larger goal of testing all the separate codes for suitability, this project succeeded in identifying the plasma simulation as the most diffi cult of the four steps, showing us we must focus B

Ar ~ 2 mTorr resources on it for the second year. We are negotiating a consulting contract with one of the authors of the plasma E code, which will include parallel- processing capability since a single CPU seems to be insufﬁ cient for the system size we must simulate. To pump

Substrate Figure 1. Schematic of magnetron deposition system to be simulated in overall project. The engineering part consists of simulating charge species and interaction with E and B ﬁ elds.

30 FY07 Engineering Research and Technology Report Engineering Modeling & Simulation

8.00 x 10–4 Substrate Related References 1. Verboncoeur, J. P., A. B. Langdon, and 6.67 x 10–4 N. T. Gladd, “An Object-Oriented Electro- magnetic PIC Code,” Comp. Phys. Comm., Target 5.33 x 10–4 87, pp. 199-211, 1995. 2. Bohlmark, J., U. Helmersson, M. Van-

r Zeeland, I. Axnas, J. Alami, and 4.00 x 10–4 N. Brenning, “Measurement of the Magnet- ic Field Discharge in a Pulsed High Current Magnetron Discharge,” Plasma Sources Sci. 2.67 x 10–4 Technol., 13, pp. 654-661, 2004.

1.33 x 10–4 Axis of symmetry FY2008 Proposed Work 0 All parts of this project are 0 3.75 x 10–4 7.50 x 10–4 1.13 x 10–3 1.50 x 10–3 continuing in FY2008. z Figure 2. Test domain for plasma simulation, scaled 1%. Red points represent simulated clusters of electrons (argon ions and neutral gas not shown).

3.42 x 10–2 6.00 x 10–2

1.19 x 10–2 1.45 x 10–2 Br Br

–1.04 x 10–2 –3.10 x 10–2 –4 –4 8.00 x 10 8.00 x 10

–4 –4 r r 4.00 x 10 4.00 x 10 –3 –3 –4 1.50 x 10 –4 1.50 x 10 0 7.50 x 10 0 7.50 x 10 0 z 0 z

Figure 3. Radial magnetic ﬁ eld component used in plasma simulation. Figure 4. Axial magnetic ﬁ eld component used in plasma simulation.

(a) (b)

200 nm

60 nm

Figure 5. (a) Simulated and (b) experimental microstructure of sputtered Be ﬁ lms. Agreement of the columnar grains, reaching a steady-state width with dome-shaped tops and apparent gaps at the grain boundaries is our key qualitative result. Quantitative agreement remains as a future goal.

Lawrence Livermore National Laboratory 31 Technology

Aaron P. Wemhoff Direct Simulation (925) 423-9839 Monte Carlo Benchmarking [email protected] and Code Comparison

wide variety of applications require their opinions regarding the usage of the A a core capability to model fl uid aforementioned codes for certain types fl ow on the subcontinuum scale. Direct of problems. Simulation Monte Carlo (DSMC) is the standard methodology used to model Project Goals these types of fl ows. The major goal of our study is to The purpose of our project is to ap- implement the codes on both classes ply a series of available DSMC codes of problems mentioned above in order to two classes of problems. The fi rst set to assess the accuracy of various code comprises simple benchmark problems features. Systems ranged from internal that contain either known or analytical fl ow problems such as a simple com- solutions, which provides us confi dence parison of a temperature gradient in a in our ability to use the codes effec- microchannel using DSMC to an ana- tively. The second set contains problems lytical temperature slip model (Fig. 1), that are pertinent to LLNL applications, or a model of an experimental chemical which allows us to determine which vapor deposition apparatus compared to code contains the best capabilities to experimental results (Fig. 2). handle a given type of problem. Three We also modeled chemically react- suites of freely obtained codes were ing external re-entry fl ows (Mach num- used: DS2V/DS3V; the DSMC Analysis ber > 18) for a cylinder (Fig. 3) and a Code (DAC); and MONACO2d/3d. sphere (Fig. 4). To supplement our study, we created fl uid mesh generation tools to allow for Relevance to LLNL Mission effi cient implementation of the codes. Systems containing subcontinuum In addition, we performed traditional fl uids are seen in a wide variety of continuum CFD analysis of several of applications around the Laboratory, the benchmark problems at noncon- which generally feature very small tinuum fl ow conditions. The results of length scales or low density (rarefi ed) this analysis demonstrated the loss of gas conditions. Examples of the former convergence ability in the CFD solver group of systems include fl ow near as the fl ow conditions deviate from the sensors or inside microchannels, while continuum limit. the latter group includes microfabrica- Finally, we interviewed known tion processes and vehicle re-entry leaders in the DSMC fi eld to obtain scenarios. One of the LLNL-specifi c benchmarks we applied the codes to is a magnetron sputtering chamber used for Figure 1. DSMC (lines) and analytical 310 growing fi lms on large scale adaptive (markers) solutions to the steady-state 305 optics, which also has relevance to the temperature fi eld of Ar in a 1-μm channel 300 microfabrication cluster tool and NIF for a Knudsen number of 0.01 (brown) 295 target capsule manufacturing. and 10 (green), where the wall temperatures are 270 K and 310 K. 290 285 FY2007 Accomplishments and Results Temperature (K) Temperature 280 We successfully modeled six general problems and three LLNL-related 275 problems using the codes DAC and 270 0 0.2 0.4 0.6 0.8 1.0 DS2V. We also examined the use of Flow position in channel (μm)

32 FY07 Engineering Research and Technology Report Engineering Modeling & Simulation

DS3V in the magnetron sputtering modifi ed continuum simulations of the type of problem, and which DSMC chamber analysis, and we applied the CVD chamber shown in Fig. 2. We also code options are the most appropriate MONACO2d code for use in a re-entry showed good code agreement between for a given system. chemistry sensitivity study. We ob- DAC and DS2V in a variety of external tained good agreement with analytical fl ow problems. All of our fi ndings are Related References solutions for the classic Fourier prob- available in a report that provides guid- 1. Davis, D. H., “Monte Carlo Calculation of lem (Fig. 1) and fl ow in a microchan- ance on the advantages and disadvan- Molecular Flow Rates Through a Cylindrical nel. Good agreement was also achieved tages of using each DSMC code, when Elbow and Pipes of Other Shapes,” J. Appl. when compared to experimental and a continuum code is proper for a given Phys., 31, 7, pp. 1169-1176, 1960. 2. LeBeau, G. J., and F. E. Lumpkin, “Ap- plication Highlights of the DSMC Analysis 0.5 Nd (1/m3) Code (DAC) Software for Simulating Rar- 4.5 x 1020 efi ed Flows,” Comp. Meth. Appl. Mech. Eng., 4.2 x 1020 191, 6-7, pp. 595-609, 2001. 20 0.4 3.9 x 10 3. Dietrich, S., and I. D. Boyd, “Scalar and 20 3.6 x 10 Parallel Optimized Implementation of the 20 3.3 x 10 Direct Simulation Monte Carlo Method,” 3.0 x 1020 20 J. Comp. Phys., 126, pp. 328-342, 1996. 0.3 2.7 x 10 20 4. Singh, V., B. Berney, and A. Krishnan, 2.4 x 10 2.1 x 1020 “Designing Low Pressure Systems With 20 Y (m) 1.8 x 10 Continuum Models,” J. Vac. Sci. Technol. A, 0.2 1.5 x 1020 14, 3, pp. 1252-1257, 1996. 1.2 x 1020 5. McNenly, M. J., A. P. Wemhoff, and 9.0 x 1019 H. Hsieh, “Sensitivity of a DSMC Solu- 6.0 x 1019 19 tion to the Thermochemistry Model for 0.1 3.0 x 10 Non-Equilibrium Reentry Flows,” DSMC Theory, Methods, and Applications Conference, Santa Fe, New Mexico, 0 Sept. 30 – Oct. 3, 2007. 0 0.1 0.2 0.3 0.4 0.5 Z (m) Figure 2. Simulation of Ar number density in a chemical vapor deposition chamber following the study of Singh et al.

6x10–05 6x10–05

4x10–05

ND 25 2x10–05 2x10 1.8x1025 4x10–05 1.6x1025 1.4x1025 1.2x1025

Y 0 25 1x10 Y MA 8x1024 20 6x1024 18 24 –2x10–05 4x10 16 2x1024 2x10–05 14 12 10 –05 8 –4x10 6 4 2 –6x10–05 0 –5x10–05 –6x10–05 –4x10–05 –2x10–05 0 X X Figure 3. Number density for air impinging upon a cylinder at Mach 18. The up- Figure 4. The simulated local Mach number for reacting airﬂ ow at 5901.38 m/s, 216.7 K, and per and lower plots show results for reacting and nonreacting ﬂ ow, respectively. 1.8487x1024 m-3 around a 0.07-mm-diameter sphere.

Lawrence Livermore National Laboratory 33 Technology

Adam White Photonic Doppler Velocimetry (925) 424-5935 [email protected]

ptical velocimeters are noncontact electromagnetically induced collapse of Odiagnostics widely used to measure the tube. The measured PDV data would the velocities of explosively driven be used to validate and critique modeled metal surfaces in single-shot shock results from ALE3D, an LLNL multi- physics experiments. Photonic Doppler physics hydrocode. Velocimetry (PDV) is a novel optical velocimetric technique, developed at Relevance to LLNL Mission LLNL, that uses optical heterodyning Velocimetry is one of the primary to measure the beat frequency between shock physics diagnostics. PDV is a light incident on and light refl ected by a particularly attractive diagnostic for ex- moving metal surface. In this project, we periments involving signifi cant quanti- have used PDV in an atypical role on the ties of radiated electromagnetic energy coaxial load experiment with ALE3D. or high-explosives because 1) the PDV probes and fi bers that are exposed in the Project Goals experimental environment are immune Our goal was to use PDV on the to electromagnetic interference; 2) PDV Figure 1. Block diagram of PDV system and probe layout. ALE3D coaxial load experiment to requires no direct mechanical contact An optical circulator and power splitter within the PDV provide velocity and displacement ver- with the measurement surface; and detector chassis routed optical power from the laser, to sus time data for four locations on the 3) PDV does not require any electrical the probes, and into the detectors. sidewall of an aluminum tube during the connections to be made on or near the measured surface. Specifi c programs re- Screen Pulsed power lab quiring this expertise include explosive room high-current test cell pulsed power for high-energy-density physics research, high-velocity impact Laser experiments for materials, EM launcher/ shaker experiments for military applica- 23tions, and subcritical component testing.

FY2007 Accomplishments and Results PDV Four PDV probes were installed on the detector ALE3D coaxial load testbed to monitor chassis the motion of the tube at four azimuthal positions separated by 90° at the mid- plane of the tube (Fig.1). The interface between the PDV probes and coaxial load 14test stand were machined adapters that were locally designed and fabricated. Digitizer The PDV laser transmitted 1550- nm light over optical ﬁ ber to each Aluminum tube PDV probe. A fraction of light incident Trigger on the ﬁ ber endface inside the probe

34 FY07 Engineering Research and Technology Report Engineering Modeling & Simulation

Figure 2. Raw PDV data. The beat Probe 1 Probe 2 Probe 3 Probe 4 respect to time to provide displacement frequency is clearly visible in the versus time data. The processed veloc- plot on the right, which depicts 1.5 only a small sample in time (be- ity and displacement data are shown in tween 20.0 and 20.5 μs) of the 1.0 Fig. 3. total waveform on the left. The Our application of PDV was slightly 0.5 waveforms are offset vertically to unusual due to the low surface velocities improve visibility. 0 measured—tens of meters per second rather than kilometers per second. We –0.5 experienced no diffi culty measuring Beat amplitude (v) these relatively low velocities. –1.0 In addition, our experiment con- –1.5 fronted one of the disadvantages of 0100 200 300 400 20.5 21.0 PDV, that the magnitude of the Doppler Time (μs) shift can be detected, but not the direction of surface movement. Although was refl ected back along the fi ber; 1 m/s surface velocity. The maximum this problem was prevalent in our data a comparable amount was launched velocity detectable by the PDV system due to the oscillatory response of the by the probe onto the moving tube we used was around 5 km/s, limited tube walls, the issue was minimized by surface, refl ected and Doppler-shifted, by the bandwidth of the detectors and assuming that after the initial compres- and collected by the probe to also digitizers, and well beyond our expected sion, each velocity zero corresponded return along the fi ber. needs. to a reversal in direction of motion. Both the endface refl ected (unshift- PDV data were taken on fi ve shots These assumptions were supported by ed) and refl ected Doppler-shifted signals of the ALE3D coaxial load experiment. data from auxiliary diagnostics. were coupled via an optical circulator Figure 2 shows raw data from one shot. onto the detectors. The beat signal (a The raw data, representing the beat Related Reference signal that is amplitude modulated at waveform, is processed using a slid- Strand, O. T, D. R Goosman, C. Martinez, a frequency equal to the difference in ing Fast Fourier Transform (FFT) to T. L. Whitworth, and W. W. Kuhlow, frequency between the two refl ected sig- determine beat frequency as a function “A Novel System for High-Speed nals) was recorded and digitized. With of time. The transformed data are scaled Velocimetry Using Heterodyne Techniques,” unshifted light at 1550 nm, this beat by 1 m/s per 1.29 MHz to provide veloc- Rev. Sci. Instrum., 77, 083108, 2006. frequency corresponds to 1.29 MHz per ity versus time data, and integrated with

(a) (b) Probe 1 Probe 2 Probe 3 Probe 4 30 0.2 0 20 –0.2 10 –0.4

0 –0.6

Velocity (m/s) Velocity –0.8 –10 Displacement (mm) –1.0 –20 –1.2 –30 –1.4 0 100 200 300 400 0 100 200 300 400 Time (μs) Time (μs)

Figure 3. (a) Velocity and (b) displacement of the monitored surfaces as a function of time. Velocity is calculated by applying a sliding FFT to the raw data in Fig.2, then scaling the transformed results. Displacement is calcualted by integrating the velocity data with respect to time.

Lawrence Livermore National Laboratory 35

Research

Richard M. Seugling Standing Wave Probes for (925) 423-8525 Micrometer-Scale Metrology [email protected]

n this project we are developing a low Relevance to LLNL Mission The near-term impact of this project Iforce, high-aspect-ratio, mechanical Currently, LLNL’s ICF and HEDP is the expansion of LLNL’s ability to probe for the nondestructive character- targets comprise components with accurately perform dimensional mea- ization of manufactured components dimensions in the millimeter range, surements of micrometer-scale features and assemblies. The key concept for with micrometer-scale, high-aspect- using a low force, high-aspect-ratio the probe is the correlation between ratio functional features, including probe system. the dynamic response of an oscillating fill-tube holes and counterbores, hohl- cantilever rod (probe) and the interac- raum starburst patterns, and step-joint FY2007 Accomplishments and Results tion of its tip with a surface (Fig. 1). geometry on hemispherical targets. In A prototype probe system has been The applications for this probe begin addition to the small-scale, complex built and is operational at LLNL includ- with surface location, but may encom- geometry, these structures are often ing advanced electronics for increased pass the characterization of the material thin walled and manufactured out of sensitivity and upgraded mechanics for properties of the surface, and perhaps delicate materials such as aerogels integration to any number of arbitrary branch into the modifi cation of the and/or metallic foams, which are sus- motion control platforms, as shown in surface. This project is a collaborative ceptible to deformation and fracture Fig. 2. The probe has been integrated effort between LLNL, the University during contact. into a profi ling station with nanometer- of North Carolina at Charlotte, the (a) National Institute of Standards and Input excitation Information contained in Input excitation Technology (NIST), and an industrial X , W Y(t) the frequency response i i of the system partner, InsituTec. Tuning fork Project Goals excitation We will provide the scientifi c under- Current carbon fiber probes are standing of a low force, contact probe excited to mode 5. capable of being used on a number of High-aspect-ratio machine tools and metrology platforms probe 0(500:1) with a characterized uncertainty based The tip interrogates the on the fundamental understanding of Low contact force surface for location and the probing process. The exit strategy 0(1–100 nN) property information. includes continued collaboration with 7 μm diameter x 5 mm long carbon filter 32 kHz; ~20 μm total tip motion A (oscillation amplitude) the UNC-Charlotte team, rigorous calibration efforts with NIST, engag- (b) ing a commercial source of probing instruments, and the probe’s practical High sensitivity region (area we are exploring) application at LLNL and its vendors for Inertial Confi nement Fusion (ICF) and

High Energy Density Physics (HEDP) Region where oscillation amplitude target fabrication. A prototype instru- decreases with contact ment developed during this project will provide an excellent platform for con- Current tinual reduction-to-practice efforts. operating The ﬁ nal goal of this work is to region

Signal amplitude (A.U.) Heavy contact: mode changes? enhance LLNL’s precision metrology capabilities, which encompass key core 0 2 4 6 8 1012 14 16 18 20 technologies supporting NIF, Weapons Probe-surface displacement (μm) Complex Integration, and nanofabrication. Figure 1. (a) Schematic description of probe system including tuning fork excitation and high-aspect-ratio rod. An ω input amplitude (Xi) at a frequency ( i) is applied to the tuning fork while the output amplitude (Y) is monitored as the probe comes into contact with the surface. (b) Output response of probe as it comes into contact with a surface.

38 FY07 Engineering Research and Technology Report Measurement Technologies level positioning capability and data A typical sensitivity experiment of a stand-alone model in Fortran. This acquisition. Experimental evaluation the probe contacting a steel gage block model consists of an Euler-Bernoulli of probe performance is being quanti- is shown in Fig. 3. Results of these ex- beam driven at one boundary by a tun- fi ed on this system. Probe sensitivity periments indicate that probe sensitivity ing fork drive system; at the other end has been measured on three different varies based on materials with similar of the beam, we apply nonlinear contact materials including a metallic glass surface fi nishes, as illustrated in Fig. 4. conditions. Theoretical approxima- (Vitreloy 105), gold, and a hardened A model of the probe dynamics tions of near surface (van der Waals, steel gage block. including contact has been derived as and electrostatic) and contact (impact, adhesion, and meniscus) forces have (a) (b) been developed and are applied to the numerical model. Year one of the proposed effort comprises both the simulation of the probe’s response when characterizing well- defi ned surfaces, and the acquisition of data using an experimental apparatus.

Related References 1. Bauza, M. B., R. J. Hocken, S. T. Smith, and S. C. Woody, “Development of a Virtual Probe Tip with an Application to High Aspect Ratio Microscale Features,” Rev. Sci. Instrum., 76, pp. 095112, 2005. 2. Burnham, N. A., O. P. Behrend, F. Oulevey, Figure 2. (a) Standing wave produced in probe by G. Gremaud, P.-J. Gallo, D. Gourdon, E. oscillating tuning fork; (b) probe mounted on LLNL Dupas, A. J. Kulik, H. M. Pollock, and G. A. motion control platform. D. Briggs, “How Does a Tip Tap?” Nanotech- nology, 8, pp. 67-75, 1997. (a) (b) 3. Weckenmann, A., T. Estler, G. Peggs, and 3.5 0.5 D. McMurtry, “Probing Systems in Dimen- 3.0 Sequence 3 Sequence 3 Sequence 4 0.4 Sequence 4 sional Metrology,” CIRP Annals – Manufac- 2.5 Sequence 5 Sequence 5 Sequence 6 Sequence 6 turing Technology, 53, 2, pp. 657-684, 2004. 2.0 Sequence 7 0.3 Sequence 7 Sequence 8 Sequence 8 4. Grifﬁ th, J. E., and D. A. Grigg, 1.5 Sequence 9 Sequence 9 “Dimensional Metrology with Scanning 1.0 0.2 Uncertainty Probe Microscopes,” J. Appl. Phys., 74, 9,

0.5 Filtered sensor signal (V) 0.1 pp. R83-R109, 1993. Filtered sensor signal (V) 0 –0.5 0 –4 –2 0 2 4 6 8 10 12 –1.0 –0.5 0 0.5 1.0 Nanopositioner (μm) Nanopositioner (μm)

Figure 3. (a) Probe output contacting a steel gage block; (b) measurement uncertainty related to the variation in location of surface illustrated by the deviation between contact measurements. This deviation is a result of surface force interactions combined with environmental effects, such as thermal instability of the system and electronics. FY2008 Proposed Work 0.1 We plan on leveraging our under- 0 y = –0.13x + 0.0335 standing of the current probe system 2 –0.1 R = 0.9686 to develop a scaled version capable Gold of operating on micrometer features –0.2 with sub-micrometer-level sensitivity. –0.3 y = –0.18x + 0.0051 We are also continuing analytical R2 = 0.9911 assessment of the probe system to

Signal amplitude (V) –0.4 Steel gage block incorporate actuation and sensitivity –0.5 in multiple dimensions, and the feasi- –0.6 bility of using the probe geometry for 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 determination of material properties Sample displacement (μm) and/or surface modiﬁ cation. Figure 4. Sensitivity response as the probe comes into contact with a steel gage block and a gold sample of nominally the same surface ﬁ nish.

Lawrence Livermore National Laboratory 39 Technology

Jeremy J. Kroll Error Budgeting and (925) 422-6437 Certiﬁ cation of Dimensional [email protected] Metrology tools

adiography and ultrasonic inspection (x-ray computed tomography) and the Rhave tremendous value in identify- Laser UT (ultrasonic testing) systems, ing features and defects in physical and we prepared a framework for error biological samples and engineering budgets comprising the source, object, structures. The historical focus of these and detector. inspection methods has been quantita- In Fig. 1 we show how information tive but without a clear understanding fl ows in these systems and where errors of uncertainty in the measurements. For can be accumulated. many applications at LLNL, it would In this project, we will form predic- be of signifi cant value to use these tions of the uncertainty of dimensional inspection techniques as quantitative measurements for NDE instruments, metrology tools with well understood based on analyzing the physics of the uncertainties. These could be used to information fl ow in the instruments obtain dimensional information on the and sources of uncertainty. We will internal structure of engineered compo- also fabricate calibration artifacts that nents and assemblies, to be compared to can be measured on other independent specifi ed tolerances. An example of the instruments, such as CMMs, which have latter requirement would be in validat- existing uncertainty values for dimen- ing the precision of centering an inner sional measurements. In comparing our component within an outer shell, which predicted uncertainty with measure- precludes access with a coordinate mea- ments on pre-characterized artifacts, suring machine (CMM) or visible light we will validate our ability to associate inspection. uncertainties with dimensional measure- In FY2006, we initiated a project ments on these instruments. to formulate error budgets for key LLNL radiographic and ultrasonic tools Project Goals that would provide rigorously defi ned The goal is to produce validated uncertainties in associating dimensional quantitative error budgets for the Xradia information with their acquired data. Micro-XCT and the Laser UT systems, Specifi cally, for the Xradia Micro-XCT which will enable a structured approach for improving the capabilities of these machines, as well as provide insight into • Alignment error • Thermal stability the effect of individual error sources. • Phase effects • Algorithm uncertainty Relevance to LLNL Mission Stage error motion The result of this project will be a broader view of dimensional metrology Signal/image that extends beyond the traditional tools Source Object Detector processing used in LLNL’s precision engineering. Weapons Complex Integration (WCI) and NIF obtain improved quantifi cation • Wavelength • Focusing optics Image • Spot size • Detector uniformity reconstruction of the uncertainties in the fabrication of targets or other components. Both HEDP and ICF target fabrication will benefi t Analysis/results from an improved understanding of the measurement uncertainties involved Figure 1. Error diagram showing possible sources of system uncertainty. with these metrology tools.

40 FY07 Engineering Research and Technology Report Measurement Technologies

and tilt motion of the rotary axis. Figure 3 shows that an object would move Identify Project goal fundamental 2.7 μm due to the rotary axis tilt error Measure or error sources at a typical fi xturing location above the estimate uncertainty Mapping of of error sources axis. Data from the testing will be used error sources Determine to complete the population of the error to measurand combinatory Generate budget, which will allow the propa- rule error Redesign or upgrade gation of these errors through object budget fundamental error retrieval algorithms to determine system sources to meet sensitivity. uncertainty goals Compare Related References Error budgeting process with desired uncertainty 1. Donaldson, R. R., “Error Budgets in Technology of Machine Tools,” Technology Figure 2. The error budgeting process. of Machine Tools – Machine Tool Accuracy, pp. 9.14-1-9.14-14, 1980. 2. BIPM, IEC, IFCC, ISO, IUPAC, IU- FY2007 Accomplishments and Results control stability of these axes are di- PAP, and OIML, Guide to the Expression Error budgets for the Xradia Micro- rectly coupled to the uncertainty in the of Uncertainty in Measurement, Geneva, XCT and the Laser UT systems have fi nal CT data. Measurements of posi- Switzerland: International Organization for been produced. Figure 2 shows the tional stability have been made between Standardization, 1995. project goal of creating validated error the source and object as well as between 3. Huber, R. H., D. J. Chinn, O. O. Balogun, budgets for the NDE machines. The the object and detector. and T. W. Murray, “High Frequency Laser- feedback portion of Fig. 2 indicates the Typically, the Xradia Micro-XCT Based Ultrasound,” Review of Progress in value in having an error budget to iden- system uses only the rotary axis to Quantitative Nondestructive Evaluation, tify key areas of a metrology instrument rotate the object with respect to a fi xed August 2005. to gain the most return on investment for source, and the detector to generate 2-D 4 Martz, H. E., Jr., and G. F. Albrecht, “Non- upgrades or redesigns. projections (radiographs) for computed destructive Characterization Technologies for The Xradia Micro-XCT system tomographic image reconstruction of the Metrology of Micro/Mesoscale Assemblies,” consists of multiple sets of stacked axes object. Proceedings of Machines and Processes for between the source, object, and detec- The fi rst set of tests performed Microscale and Mesoscale Fabrication, Me- tor. Even though the source and detector showed the angular positioning accuracy trology, and Assembly, ASPE Winter Topical axes are typically stationary during a of the axis. Subsequent tests are being Meeting, Gainesville, Florida, pp. 131-141, CT scan, the thermal, vibrational, and performed to measure the radial, axial, January 22-23, 2003.

0.0015

0.0010 FY2008 Proposed Work For FY2008, we will use the 0.0005 sensitivity information from the error budgets to choose characteristic 2.7 μm artifacts to validate the error budget 0 predictions. These artifacts will be measured on both CMM and NDE Run 1A –0.0005 tools to create validated error budgets Run 1B

Position error (mm) for the Xradia Micro-XCT and Laser Run 2A UT. These quantitative error budgets –0.0010 Run 2B will be used to state uncertainties in Run 3A dimensional measurements and will –0.0015 Run 3B lead to improved uncertainties in future generations of these instruments. –0.0020 –90 –70 –50 –30 –10 10 30 50 70 90 Angular position (°)

Figure 3. Position error of a part 25 mm above the rotary axis, due to tilt motion of the axis, for the Xradia Micro-XCT.

Lawrence Livermore National Laboratory 41 Technology

Walter W. Uncertainty Analysis with Nederbragt (925) 424-2807 Inspection Shop Measurements [email protected]

he LLNL inspection shop is chartered Tto make dimensional measurements of components for critical programmatic experiments. These measurements ensure that components are within tolerance, and provide geometric details that can be used to further refi ne simulations. For these measurements to be useful, they must be signifi cantly more accurate than the tolerances that are being checked. For example, if a part has a specifi ed dimension of 100 mm and a tolerance of 1 mm, then the precision and/or accuracy of the measurement Figure 2. The Z-Mike measurement instrument, which uses a laser to measure diameters and lengths. should be less than 1 mm. Using the “10-to-1 gagemaker’s rule of thumb,” that infl uence measurement uncertainty. Project Goals the desired precision of the measurement This project aims to augment the efforts The goal of this project is to begin should be less than 100 μm. Currently, within the LLNL inspection shop with providing measurement uncertainty the process for associating measurement a standardized and commensurately statements with critical measurements uncertainty with data is not standard- rigorous approach to determining and performed in the inspection shop. To ized, nor is the uncertainty based on a reporting uncertainty. accomplish this task, comprehensive thorough analysis. During FY2006, a fundamental knowledge about the underlying sources The National Institute of Standards understanding of inspection shop opera- of uncertainty for measurement in- and Technology (NIST) has developed tions and equipment was gained so that struments need to be understood and methods for analyzing measurement un- measurement uncertainty analysis could quantifi ed. Moreover, measurements of certainty. Figure 1 shows the key factors proceed in FY2007. elemental uncertainties for each physical source need to be combined in a meaningful way to obtain an overall measure- Thermal effects ment uncertainty. Relevance to LLNL Mission The measurements being made by Long term Master gage the inspection shop are used to make reproducibility calibration decisions about accepting or rejecting critical parts. The inspection shop is widely used and the measurements are Uncertainty typically accepted as being “suffi cient- analysis ly” accurate. This assumption should be verifi ed by a measurement uncertainty Elastic Scale analysis, which is the accepted practice deformation calibration at all of the other NNSA sites. There is a signifi cant risk to Laboratory programs if measurement data is in error, Artifact Instrument which could lead to the use of compo- effects geometry nents in experiments that are outside of Figure 1. Schematic of the many factors that infl uence the uncertainty of a measurement. specifi cations.

42 FY07 Engineering Research and Technology Report Measurement Technologies

FY2007 Accomplishments and Results CMM #2 CMM #3 During FY2007, four milestones were met: the reports written in FY2006 CMM #1 CMM #4 were ﬁ nalized; a tutorial on uncertainty analysis was given to the inspectors in the inspection shop; an uncertainty analysis was performed on the inspection shop’s Z-Mike measurement instrument; and an uncertainty analysis was performed on the inspection shop’s four Coordinate Measurement Machines Part CMM (CMMs). Uncertainty Probe The Z-Mike measurement instru- analysis specifications ment (Fig. 2) is a new instrument in the software inspection shop for measuring diameters and lengths using a laser. Reproduc- Room temperature ibility data was collected on the Z-Mike instrument; this data was analyzed and an uncertainty analysis was completed using the methods described in the references. Applying this same analysis method to the four CMMs in the inspection shop would be extremely cumbersome; hence,

a different method was used. A commer- Probability of error cial vendor has created a CMM program that uses algorithms developed at NIST Error to analyze the measurement uncertain- Figure 3. Performance analysis of CMMs. The software creates a report detailing the uncertainty of each feature ties associated with the measurement measurement on each part for each CMM. of speciﬁ c part features/geometries. Using this program, a generic part with common geometric features was input into the program along with the inspection shop’s CMM calibration data, probe Task: determine the uncertainty data, and temperature data. Using the associated with measuring the data, the program created uncertainty position of the large hole relative reports for each part feature (Fig. 3). to the part datums The analysis shows that the performance of the four CMMs in the inspection shop varies considerably (Fig. 4). This result is not a surprise, but it does show the importance of choosing the correct metrology tool when making a critical measurement.

Related References 1. Doiron, T., and J. Stoup, “Uncertainty and Dimensional Calibrations,” Journal of Re- search of the National Institute of Standards and Technology, 120, 6, pp. 647-676, 1997. Results 2. Taylor, B. N., and C. E. Kuyatt, “Guidelines CMM #1: 5.5 μm CMM #2: 3.3 μm for Evaluating and Expressing the Uncertainty CMM #3: 1.6 μm CMM #4: 7.5 μm of NIST Measurement Results,” National In- Figure 4. Uncertainty results for the four inspection shop CMMs when measuring the position of a hole using the stitute of Standards and Technology, Technical uncertainty analysis software. Note 1297, 1994.

Lawrence Livermore National Laboratory 43 Technology

William D. X-Ray System Modeling Brown (925) 422-7933 [email protected]

e are implementing a forward optically coupled to a scientifi c grade a 100-kV potential and a tube current of Wmodel for x-ray system response cooled CCD camera. We will compare 40 μA. The focal spot of the source at the that will enable us to predict the ca- the modeled x-ray spectrum, detec- 100-kV potential is 7 μm2 nominal. To pability of our systems and allow us tor, and x-ray transport to a previously validate the source model, measured data to choose optimal system parameters. imaged double-shell Omega target that were obtained from an identical source The system model will include the is used for high-energy-density physics using a high-purity germanium detector. major components of a system: x-ray (HEDP) experiments. Figure 1 compares the model spectra producing source, photon transport, and and the measured data. The fi gure indi- detector. The modeled components will Relevance to LLNL Mission cates the measured spectra E< 20 keV be used in conjunction with an LLNL A forward model for system re- has less fl ux than the model. This could software program, HADES, to model sponse will enable us to better perform indicate possible inherent fi ltering in the complete system. experiments for WCI, NIF and Global the x-ray source that was not attributed Security (GS). to the model. The three K-lines (58, 59 Project Goals and 67 keV) correspond correctly in the The overall project goal is to model FY2007 Accomplishments and Results model and measured data. However, the the x-ray source, detector and x-ray The fi rst task was to create a model intensity of the modeled lines is greater. transport of the Micro-XCT Xradia sys- for the Brehmstrahlung source. We used The 9-keV L-line is barely evident in tem and compare it to a known sample. Monte Carlo N-Particle (MCNP) to model the measured data. This may be due to The modeling techniques can later be the x-ray source. The Hamamatsu source inherent source fi ltration. expanded to other x-ray systems with is a conventional tungsten anode micro- To model the x-ray energy deposited different x-ray sources, detectors, and focus system with a Brehmstrahlung in the Xradia CsI scintillator we used transport characteristics. The Xradia sys- continuous spectrum and characteristic MCNP. The MCNP calculations were tem uses a 150-kV Hamamatsu Brehm- K- and L-lines. The modeling parameters used to compute the scintillator blur strahlung source and a CsI scintillator for the source were chosen from the function and the DQE. To validate the mechanically coupled to a microscope parameters used to image the double-shell detector model, Modulation Transfer objective. The microscope objective is Omega targets. These parameters include Functions (MTF) for the model and measured data were to be compared, but we were unable to obtain the modeled Measured/MCNP comparison MTF in time. Future efforts would in- 4 clude modeling the MTF and comparing with measured results. Measured The sample modeled was the HEDP Model 3 Omega double-shell (DS) target, consisting of two concentric shells. The outer shell is a CH (Br-doped) ablator consisting of upper and lower hemi- 2 shell components that are joined togeth-

Fluence (A.U.) er with an S-shaped step joint. Because of the complexity of the step joint, the 1 model assumes a uniform ablator without the joint. The inner shell is made of glass, and is held in place within 3 0 the outer shell with 50-mg/cm SiO2 0 20 40 60 80 100 aerogel. The HADES software, incor- Energy (keV) porates the source and detector models Figure 1. Comparison of x-ray source model and measured spectra. with its ray tracing techniques to output a radiographic model of the sample.

44 FY07 Engineering Research and Technology Report Measurement Technologies

(a) (b) Figure 2. (a) Double-shell digital radio- 0.5 mm 0.5 mm graph; (b) model. (The pedestal in (a) was not modeled in (b)).

Source blurring and detector resolution lines. The model, both with and without An ASNT Topical Conference, Mashantucket, are computed for each energy bin of the lines, shows a stronger phase effect at ma- Connecticut, July 24-26, 2006. spectra. The total intensity is computed terial transition than is measured. Overall, 3. Hibbard, R. L., et al., “Precision Manufac- from the sum of the doses at each DQE- a comparison of the model and measured turing of Inertial Conﬁ nement Fusion Double weighted energy bin. data present a close agreement. Shell Laser Targets for OMEGA,” Fusion Figure 2 shows a Micro-XCT digital Science and Technology, 45, 2, pp. 117-123, radiograph using 0.6-μm pixel pitch of the Related References 2004. double-shell target and a HADES simu- 1. Aufderheide, M. B., et al., “Analysis and 4. Martz, Jr., H. E., and G. F. Albrecht, “Non- lated radiographic model. A qualitative Modeling of Phase Contrast Radiography of destructive Characterization Technologies for comparison of the images shows a close Gradient Density Laser Targets,” Symposium Metrology of Micro/Mesoscale Assemblies,” comparison in resolution and contrast. on Radiation Measurements and Applications, Proceedings of: Machines and Processes Figure 3 shows a lineout through the Ann Arbor, Michigan, May 23-25, 2006. for Micro-scale and Mesoscale Fabrication, ablator, aerogel, and inner glass sphere of 2. Brown, W. D., and H. E. Martz, Jr., “X-ray Metrology, and Assembly, American Society the radiograph measurement and model. Digital Radiography and Computed Tomog- for Precision Engineering, Raleigh, North The lineouts of the model were obtained raphy of ICF and HEDP Materials, Subas- Carolina, pp. 131-141, 2003. with and without adding the characteristic semblies and Targets,” Digital Imaging IX, 5. Xradia website: http//www:xradia.com.

0.5 FY2008 Proposed Work 0.4 Additional work is needed to compare the MTF of the model with the empirical MTF from the Xradia 0.3 )

0 system. The MTF comparison will allow us to quantitatively compare –In (l/l 0.2 the contrast and spatial resolution of the model and empirical data. In addition, we need to expand the 0.1 XRadia modeling to other x-ray sources up Brem + lines HADES to 450 kV, additional detectors and Brem only HADES ﬁ nally full systems.

0 –400 –300 –200 –100 0 Location (μm) Figure 3. Lineouts as shown in Fig. 2 across ablator, aerogel, and inner shell.

Lawrence Livermore National Laboratory 45 Research

Detection, Classiﬁ cation, and James V. Candy (925) 422-8675 Estimation of Radioactive [email protected] Contraband from Uncertain, Low-Count Measurements

he detection of radioactive contra- Bayesian inference and physics-based Tband smuggled into our nation is signal processing, as illustrated in Fig. 1 a serious national security issue. The for a cargo scanning application. objective of this project is to investigate fast, reliable radiation detection methods Project Goals capable of making a more rapid decision We expect to develop solutions for with higher confi dence along with the the detection, classifi cation, and estima- ability to quantify performance. Today’s tion of a moving radionuclide source high-speed, high-throughput computers and/or detector. Our goal is to reliably enable physics-based statistical mod- detect kilograms of shielded Pu with els that capture essential signatures of a 95% detection probability at a 5% radionuclides and incorporate them into false alarm rate in less than a minute. a Bayesian sequential scheme capable The Bayesian approach enables the of online, near real-time operation. This development of a sequential framework project is focused on the detection, clas- establishing the foundation for future sifi cation, and estimation of illicit radio- problems that are both time- and space- active contraband from highly uncertain, varying or equivalently statistically low-count radionuclide measurements nonstationary. using a statistical approach based on

Figure 1. A scenario of high importance in Alarm National Security: the processing of gamma- ray measurements for cargo containers using detection instrumentation along with sophisticated Bayesian processing techniques to detect, classify and estimate Threat illicit radionuclides. processor • Detect Type • Classify • Estimate

Mass

46 FY07 Engineering Research and Technology Report Measurement Technologies

(a)Measured input μ(t) & deconvolved estimate Xμ(t) (b) Histogram of true input 4. development of sequential Bayesian x10–3 60 processors for a high-purity germa- 18 40 nium (HPG) detector. 16 Counts 20 14 We expect to ﬁ nish similar runs for 12 0 0.0130 0.0144 0.0158 0.0172 0.0186 0.0200 sodium-iodide detectors as well. 10 These results are quite promising 8 Histogram of established deconvolved input Outputs (v) Input 60 6 and demonstrate the potential capability 4 40 of the Bayesian model-based approach 2

Counts 20 to solving a variety of radiation detec- 0 –2 0 tion problems. 5500560057005800 5900 6000 6100 6200 0.0130 0.0144 0.0158 0.0172 0.0186 0.0200 Samples Input energy level (v) Related References 1. Candy, J. V., “Nonlinear Statistical Signal (c) (d) Histogram of scaled (relative to input) measured output Processing: A Particle Filtering Approach,” Measured & estimated outputs 15 0.025 Imaging for Detection and Identiﬁ cation, J. 10 0.020 Measured Byrnes (Ed.), Springer, pp. 129-150, 2007.

Counts 5 0.015 Estimated 2. Candy, J. V., “Bootstrap Particle Filter- 0.010 0 0.0130 0.0144 0.0158 0.0172 0.0186 0.0200 ing,” IEEE Signal Process. Magaz., 24, 4, 0.005 Output Histogram of scaled (relative to input) KF output pp. 73-85, 2007. Outputs (v) 0 15 3. Candy, J. V., K. Sale, B. Guidry, D. –0.005 10 Manatt, and A. Meyer, “Bayesian Processing

–0.010 Counts 5 for the Detection of Radioactive Contra- –0.015 0 band from Uncertain Measurements,” IEEE 2500255026002650 2700 2750 2800 0.0130 0.0144 0.0158 0.0172 0.0186 0.0200 Samples Input energy level (v) Computational Advances in Multi-Channel Sensor Array Processing Proc., pp. 49-52, Figure 2. Bayesian deconvolution processor outputs from experimental data: (a) actual test input excitation pulse 2007. sequence and estimated (deconvolved) input (overlaid) to the preamplifi er; (b) true and estimated (deconvolved) 4. Sale, K., J. V. Candy, B. Guidry, D. input excitation histograms; (c) measured and estimated preamplifi er outputs; (d) measured and estimated output histograms. Manatt, and A. Meyer, “A Bayesian Signal Processor Approach to Spectroscopic Portal System Measurements,” Proc. S.P.I.E., pp. 670701-670723, 2007. Relevance to LLNL Mission 3. validation through simulation/ The detection of illicit radionuclides experiments; is a top priority in furthering the national 4. application to known controlled security mission of the Laboratory. Ra- experimental data; publication of the dionuclide detection, classifi cation, and results demonstrating the capability FY2008 Proposed Work estimation are critical for detecting the of the Bayesian processor to extract Our proposed work for FY2008 transportation of radiological materials the unknown input (deconvolution) consists of completing the sodium- by terrorists, an important goal in both from measured radiation data iodide experiments and initiating a national and international security. This (Fig. 2); and set of validated (based on our experi- technology also supports the life exten- 5. publication of a record of invention ments) COG simulations to design sion program because of its potential describing the initial detection/ the Bayesian processor incorporating application in defect detection. classifi cation system design. more and more of the physics. Here Having just started this effort at we plan to incorporate more of the FY2007 Accomplishments and Results mid-year we have performed the initial transport physics into the processor Our FY2007 accomplishments planned tasks of: and begin to validate and refi ne our included: 1. sophisticated modeling using COG, initial design concepts. 1. development of a radionuclide signal a Monte Carlo code developed at processing model (mathematically) LLNL, for a variety of well-known by decomposition into monoener- radionuclides; getic sources; 2. controlled experiments; 2. incorporation of these embedded 3. development of the required signal physics-based models into a sequen- processing models such as instru- tial Bayesian processor; mentation and detectors; and

Lawrence Livermore National Laboratory 47 Research

Michael W. Burke Terahertz Spectroscopic (925) 422-0192 Imaging for Standoff [email protected] Detection of High Explosives

DX (1,3,5-trinitro-1,3,5-triazacy- from common background materi- Rclohexane), a component in plastic als (Fig. 1). Since THz radiation can explosives, including C-4 and Semtex-H, penetrate common dielectric concealants is extremely diffi cult to detect using air (e.g., fabrics and leather), and provides sampling due to its low vapor pressure reasonable spatial resolution for imaging (10 ppt). This project is examining the applications, spectral imaging near 800 feasibility of using terahertz (THz) GHz may provide a solution to the RDX radiation in the standoff detection and detection problem. identifi cation of this high explosive Our primary goal is to assess the (HE) compound. Our approach uses utility of THz spectral imaging for the emerging spectroscopic and imaging detection of concealed, RDX-based technologies in the THz frequency explosives, for reliable RDX screening regime. We propose to fi rst develop a of people at safe (30- to 50-m) standoff system-level analysis, a system simula- distances. Our approach is to develop tion, and an experimental program. a systems concept and multispectral The THz portion of the electromag- detection algorithms, and simulate netic (EM) spectrum is rich with spec- the behavior of such a system in the troscopic information about small- and presence of atmospheric absorption, medium-size molecules. RDX-based HE obscurant losses, and system noise. Our exhibit a distinctive sub-THz signature fi nal goal is to demonstrate detection of near 800 GHz that distinguishes them RDX-based explosives at 30 to 50 m through a concealing material, using multispectral imaging. Scaled to match peak amplitude at 800 GHz 1.2 The algorithm is a two-channel ap- Teraview (RDX) proach that compares the return signal SRI (Semtex) from two points on a target to remove 1.0 Yamamoto (C-4) the effects of the intervening atmosphere PETN additive

” (Fig. 2), and then compares the return E 0.8 signals to a database of spectral sig- Semtex-H natures. Our strategy is to collect THz spectral data for materials of interest 0.6 from both experiments and the literature, and combine it with systems analysis C-4 0.4 and simulation to assess the viability of the remote sensing capability.

Imaginary dielectric constant RDX 0.2 Relevance to LLNL Mission Explosives detection that enables the 0 interdiction of terrorists is an important 300 800 1300 1800 2300 capability for LLNL’s missions in both Frequency (GHz) national security and homeland security. Figure 1. RDX-based HE signature. A distinctive sub-THz signature near 800 GHz distinguishes them from common background materials.

48 FY07 Engineering Research and Technology Report Measurement Technologies

Reference signal rR(t) s(t) THz detection rT(t) system Atmosphere Target signal h(t) Target Target signature

Figure 2. Two-channel approach. The return signals from two points on a target are compared to remove the effects of the intervening atmosphere. The return signals are then compared to a database of specral signatures.

FY2007 Accomplishments and Results system using atmospheric absorption Related References In FY2007, we developed a systems data from HITRAN (a standard database 1. National Academy of Sciences, Existing concept (link budget) and supporting of atmospheric absorption properties), and Potential Standoff Explosives Detection algorithms for standoff THz spectral and material parameters we obtained. Techniques, National Academies Press, 2004. imaging. In particular, we: 1) completed Results of these simulations show 2. Woolard, D. L., E. R. Brown, M. Pepper, collection of THz data on common HE the proposed system can achieve and M. Kemp, “Terahertz Frequency Sensing materials; 2) developed a systems con- detection of bulk HE at safe standoff and Imaging: A Time of Reckoning Future cept and link budget for a THz spectral distances (~30 to 50 m), even when the Applications?” Proc. IEEE, 93, pp. 1722- imager; 3) developed supporting algo- explosive is concealed by a few layers 1743, 2005. rithms for THz spectral imaging; and of fabric. Simulated receiver opera- 3. Yamamoto, K., et al., “Noninvasive 4) executed several Monte Carlo simu- tion characteristic curves, which show Inspection of C-4 Explosive in Mails by lations of system performance. the probabilities of detection and false Terahertz Time-Domain Spectroscopy,” Jap. The algorithms developed were for alarm, show excellent discrimination of J. Appl. Phys., 43, L414, 2004. deconvolving the spectral modiﬁ cations C-4 against innocuous materials such as 4. Bjarnason, J. E., et al., “Millimeter-Wave, induced by atmospheric propagation. skin and lactose (Fig. 3). Terahertz, and Mid-Infrared Transmission We simulated the performance of this Through Common Clothing,” Appl. Phys. Lett., 85, p. 519, 2004.

1.0 0.9 FY2008 Proposed Work 0.8 ROCs for bare material: In FY2008 we will seek outside distance = 30m funding to verify experimentally our 0.7 algorithms for spectral detection of RDX in the presence of atmospheric 0.6 losses and concealants. We will ﬁ rst explore near-THz (800 to 900 GHz) 0.5 bsem-skin1-SEM bsk1-skin1-SEM imaging at the necessary standoff 0.4 bsk2-skin1-SEM distances; then, if results are favor- lactos1-skin1-SEM able, extend measurements to two 0.3 lactos2-skin1-SEM lower frequency bands to obtain the Probability of detection/false alarm bsem-skin2-SEM required multispectral coverage. bsk1-skin2-SEM 0.2 bsk2-skin1-SEM lactos1-skin2-SEM 0.1 lactos2-skin2-SEM

0 –1.0 –0.8 –0.6 –0.4 –0.2 0 0.2 0.4 0.6 0.8 1.0 Decision threshold – correlation coefficient Figure 3. ROC-type curves for different targets (HE, skin, and lactose interferents), no covering material, at 30 m standoff. Skin 1 and 2 and lactose 1 and 2 are four different targets differing in their spectral reﬂ ectance.

Lawrence Livermore National Laboratory 49 Technology

Robert Huber Application of Laser-Based (925) 424-2002 Ultrasound to Glovebox [email protected] Enclosed Materials

e have investigated laser-based ratio of the ultrasound generated and Wultrasound applied to materials detected using laser-based systems? enclosed in gloveboxes or other types of enclosures. Applications include materi- Relevance to LLNL Mission als at elevated temperatures in furnaces, Joining technologies such as weld- radioactive materials in gloveboxes and ing remain critical to manufacturing materials being joined in welding enclo- critical components at LLNL, which has sures. Laser-based ultrasound is ideal for state-of-the-art laser welding facilities. these remote applications because the Nondestructive evaluation of the welds technique does not require contact with using a technique such as laser-based the surface, and the laser beams can be ultrasound can ensure that the welds transmitted through windows. meet the requirements. Provided that laser-based ultrasound can be used Project Goals through the windows of a laser welding The main goal for this project is to station, the potential exists for process perform initial studies to determine if control whereby the ultrasonic signal laser-based ultrasound could be applied can determine when the appropriate to materials important to LLNL, inside welding depth has been reached. an enclosed laser welding station. Two main questions are to be answered by FY2007 Accomplishments and Results this project: 1) how well does ultrasound Contact ultrasonic tests (UT) were propagate in materials that are cur- performed on vanadium (Va) and tita- rently being welded by a laser welding nium (Ti) plates. These plates were the process; and 2) will glovebox windows same alloys and the same thicknesses signiﬁ cantly affect the signal-to-noise currently being welded in the LLNL

1.5

3.2-mm-thick Va 10 MHz Longitudinal 1 Pulse-Echo

0.5

0 Amplitude (V)

-0.5

-1

-1.5 00.2x10-4 0.4x10-4 0.6x10-4 0.8x10-4 1x10-4 1.2x10-4 Time (s)

Figure 1. Amplitude vs. time trace in Va using a contact UT.

50 FY07 Engineering Research and Technology Report Measurement Technologies

setup allowed a silica window, similar to 0.8 those used in the welding station, to be placed in front of the sample on the side 0.6 being hit by the pulsed laser to look for Va Filtered a reduction in signal-to-noise over the Laser Gen. PZT DHz 0.4 case when no window was present. The drop-off in signal level was negligible. 0.2 Figure 2 shows a laser generated signal obtained in Va. At this point, it 0 was clear that laser-based ultrasound could indeed be applied to the Va and Ti Amplitude (V) -0.2 materials being welded at LLNL. The fi nal tests involved laser- -0.4 based detection. Since the materials as welded have smooth but not specularly -0.6 refl ecting surfaces, tests using a low laser power Michelson interferometer -0.8 -1x10-4 0 1x10-4 2x10-4 3x10-4 4x10-4 5x10-4 had a very low signal-to-noise ratio. Time (s) Photorefractive interferometers, unlike Michelson interferometers, can work with somewhat diffusely reflecting Figure 2. Laser generated ultrasound and contact detected in VA through a glovebox-like window. surfaces and are more suitable for dif- fuse surfaces. laser welding stations. The piezoelec- high frequency ultrasound could be used Figure 3 shows a signal-averaged tric transducers chosen for these tests for characterization of parts made from waveform obtained using the pho- had center frequencies of 10 MHz. For the Va and Ti alloys. torefractive interferometer. A contact both of the materials, 10 MHz has a Figure 1 shows a typical amplitude transducer was used to generate the wavelength of approximately 0.6 mm, vs. time ultrasound trace for the Va waveform in Fig. 3. This particular test and should be suffi cient to see voids sample. Next, a through transmission was not done through a window. or cracks on the order of 0.3 mm. The test was set-up, whereby a pulsed laser This project shows the potential for 10-MHz ultrasound propagated very was used to generate ultrasound in the the application of laser-based ultrasound well through the plates with very little plates, and a piezoelectric transducer to Va and Ti alloys of interest to LLNL attenuation, demonstrating that fairly was used to detect the ultrasound. This in laser welding stations.

0.1

0.05

-0.05

-0.1 Amplitude (V) -0.15

-0.2

-0.25

-0.3 -2 x10-4 0 2x10-4 4x10-4 6x10-4 8x10-4 10x10-4 Time (s)

Figure 3. Contact transducer generated UT and detection of ultrasound using photorefractive interferometer.

Lawrence Livermore National Laboratory 51 Technology

Douglas N. Poland Defect Detection in (925) 422-4980 Large CT Image Sets [email protected]

his image analysis project has Project Goals in metal parts. Of the 13 holes in this Tconstructed a tool for performing The enhanced surveillance program ring, 9 (ranging from 305 to over 1000 computer-assisted detection of small will produce CT data sets that are up to mm in diameter) are well resolved by the voids in CT data sets of metal compo- several thousand voxels on a side CT system and are readily discernible in nents. We fi rst constructed algorithms (i.e., 8000 x 8000 x 8000). The current the processed CT data (the remaining 4 using known approaches and measured method of analysis requires a tomogra- are less than 300 mm in diameter). One their performance on a test object, pher to view sequences of several thou- key success metric is that our algorithm described below. Noise reduction and sand images, where each image occupies must place these 9 voids at the top of our void detection algorithms based on several computer screens at full resolu- ranked list of candidate voids. mathematical morphology demonstrated tion. One of the primary objectives is to good performance on the test object and identify voids of any size. The work of Relevance to LLNL Mission excellent performance on a small but searching for voids at the resolution lim- This project has produced a tool that diverse collection of programmatic data it of the system is extremely demanding will increase the effi ciency of enhanced sets that had been previously analyzed and tedious in these data sets. The goal surveillance program tomography manually. Implementation issues, such of this project is to create a tool that will analysts by focusing their attention on as handling large data sets and auto- reduce these data sets to a ranked set of suspect voids in a given data set, thereby mated parameter selection, have been candidate voids that can be quickly vali- allowing them to forego an exhaustive explicitly addressed in the course of dated or rejected by the tomographer; search of many thousands of screens of working with these data. The result is a the broader goal is to strengthen the largely defect-free data. This tool could standalone C++ application, delivered dialog and technical exchange between also make these data sets more acces- to the enhanced surveillance program tomography and image analysis experts sible to other technical staff, who may that directs a tomography analyst’s so that emerging image analysis issues not be tomography experts, by perform- attention to candidate void regions in can be effectively addressed. ing necessary preprocessing and quickly large data sets. The enhanced surveillance program guiding them to regions with suspect uses a tungsten ring with holes drilled voids. This application should be gener- into it (Fig. 1) to study the ability of their ally applicable to the detection of small systems and analysts to detect small voids defects (i.e., voids and inclusions) in homogeneous media. Figure 1. Tungsten ring test 5 4 3 2 1 7 6 object, a cylinder with 13 holes 13 FY2007 Accomplishments and Results drilled into its inner surface. This 12 cutaway sketch shows the hole We accomplished four interrelated 11 locations, while the table lists their tasks in FY2007: 1) algorithm testing sizes. Note that in the images used 10 and refi nement using programmatic data Hole Diameter Depth for this work, holes #4 through index (μm) (μm) sets; 2) automated parameter selection; 9 #7 are very diffi cult to impossible 1 1952 400 3) implementation in C++; and 4) iden- to detect manually. They are not 2 1020 930 8 tifi cation of issues that warrant future found by our algorithms. 3 1035 460 work. We have delivered a standalone 4 244 247 C++ application that effectively fo- 5 257 115 6 138 52 cuses an analyst’s attention on localized 7 138 124 anomalies in large grayscale 3-D CT 8 508 508 data sets. 9 508 406 In the course of testing with these 10 406 406 11 406 305 data sets, we have refi ned the way 12 305 406 that we calculate ranking metrics for 13 305 305 candidate voids, using, for example, a contrast calculation that handles region boundaries (Fig. 2). Our revised volume- weighted contrast ranking performs well on the tungsten ring defects (9 detected 52 FY07 Engineering Research and Technology Report Measurement Technologies holes ranked in order of decreasing size, with the most conspicuous false alarm ranked number 10 (Fig. 3)). We have implemented a minimum entropy approach to calculating the void detection threshold parameter, and also a second threshold to yield a more complete void reconstruction (Fig. 4). We have also achieved a 4x speedup, as well as portability and improved data and memory management, with our optimized C++ implementation leveraging published methods.

Related References 1. Haralick, R., S. Sternberg, and X. Zhuang, “Image Analysis Using Mathematical Figure 2. Three candidate defects ranked 1, 6 and 11 by the detection code. For each defect we display (from left to Morphology,” IEEE Transactions on Pattern right): x-y, x-z and y-z views; defect voxel intensity histogram; and immediate background voxel intensity histogram. Analysis and Machine Intelligence, 9, Defects 1 and 6 are true positives, (corresponding to holes #2 and #11 from Fig. 1, respectively) that occur on a region pp. 532-550, 1987. boundary. Their background histograms are bi-modal, with the relevant background intensity captured by the primary 2. Rosenberg, A., and J. Hirschberg, “V- mode (deﬁ ned by the vertical red lines that are generated automatically). Using this primary mode in the contrast calculation improves the ranking metric. The size of the W ring CT data set is 1908 x 1908 x 56. The thumbnails shown Measure: A Conditional Entropy-based are all 64 voxels on a side, displayed at full resolution. Cluster Evaluation Measure,” Proceedings of the 2007 Joint Conference of Empirical Methods in Natural Language Process- ing and Computational Natural Language Learning, Prague, pp. 410-420, June 2007. 3. van Herk, M., “A Fast Algorithm for Local Minimum and Maximum Filters on Rect- angular and Octagonal Kernels,” Patt. Rec. Letters, 13, pp. 517-521, 1992. 4. Bloomberg, D., “Implementation Efﬁ - ciency of Binary Morphology,” International Symp. for Math. Morphology VI, Sydney, i0, 1002i1, 520 i2, 404 i3, 173 i4, 134 i5, 93 i6, 54 i7, 28 i8, 37 i9, 29 Australia, April 3-5, 2002. Figure 3. The top 10 ranked tungsten ring candidate defects (in 10 columns above) produced by our application. Rank 5. van den Boomgard, R., and R. van Balen, proceeds from left to right, with x-y, x-z and y-z views (top to bottom) shown for each candidate. “Methods for Fast Morphological Image Transforms Using Bitmapped Images,” Computer Vision, Graphics, and Image Processing: Graphical Models and Image 1.4 Processing, 54, pp. 254-258, May 1992. 1.2 )

3 1.0

FY2008 Proposed Work 0.8 We have identiﬁ ed ideas for future 0.6 work that may lead to enhanced

detection sensitivity and false Measured volume (mm 0.4 alarm rejection by performing more sophisticated true and false positive 0.2 characterization (e.g., classiﬁ cation techniques operating on shape, 0 density, texture, and other features 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 in both the candidate defect and Design volume (mm3) surrounding background regions). Figure 4. Plot of measured tungsten (brown line) defect volumes vs. design (green dashed line) volumes. Measured volumes were calculated using “double-threshold” morphological reconstruction.

Lawrence Livermore National Laboratory 53 Research

Anthony The Structure and Properties Van Buuren (925) 423-5639 of Nanoporous Materials [email protected]

ur goal is to quantify the microstruc- We will characterize the pore struc- Oture of highly porous materials, and ture of metal oxide foams immersed in to determine how processing of the po- a cryogenic fl uid. Knowledge of how rous material relates to the structure and the aerogel wets in a cryogenic fl uid is ultimately to the mechanical behavior. unknown but it is needed information if We will quantify structural changes with these foams are to be used in laser targets. a combination of small angle x-ray scat- We will use fi nite element modeling tering (SAXS) and high-resolution x-ray to study the effects of mechanical load- tomography. Finite element modeling, ing on the cell structures, and to map out using the structures determined above, relationships among processing, density, will be used to study how a change in and strength. pore structure effects mechanical proper- We will determine the extent of any ties. anisotropy in lattice architecture, and improve spatial resolution to examine and Project Goals characterize graded density structures. We plan to make SAXS measurements to understand the effect of Relevance to LLNL Mission synthesis conditions on the change in This project develops critical experi- structure. In particular we are interested mental technologies for many LLNL in how synthesis may affect the amount applications. A key deliverable will be of mass at the nodal points in the aerogel the ability to predict the mechanical and then provide feedback for improved properties of nanoporous materials and processing to control strength (network characterize gradient density foam mi- morphology), uniformity, and density. crostructures for future laser targets.

0.055

0.050

0.045

–1 0.040

0.035

0.030

0.025 1/ Correlation range (nm)

0.020

0.015

0.010 050100 150 200 250 Aerogel density (mg/cc) Figure 1. Change in the correlation range (pore diameter) measured by SAXS as a function of Ta2O5 aerogel density.

54 FY07 Engineering Research and Technology Report Measurement Technologies

(a) (b)

100

200 mg/cc ) 6 10 40 mg/cc Count/s)/(Mon x 10 1

0.1 10–4 10–3 10–2 q(A–1)

Figure 2. (a) SANS spectra measured on a 40- and 200-mg/cc Ta2O5 aerogel; (b) high-resolution rendering of 500-nm cube of Ta2O5 aerogel.

FY2007 Accomplishments and Results image result (Fig. 2b) that shows large One of the main objectives of this voids in the low-density aerogel. As project was to quantify the pore struc- aerogel density is decreased bellow 100 ture as measured from the SAXS experi- mg/cc, the nanoporous structure remains ments due to changes in the metal oxide constant and the decreased density is foam preparation condition. In Fig. 1, obtained by the formation of ~ 1000-nm the correlation range or the distance voids. This may be a fundamental limit between two ligaments in the aerogel on the density of the materials. measured in the SAXS experiment is We have also been successful in our related to the density of the aerogel. We second goal, to measure the pore struc- see that the inverse of the correlation ture of the metal oxide foam immersed range scales linearly with density until in a cyroﬂ uids. We have been able to a density of 100 mg/cc. Below this, the measure the SAXS of a porous material density decreases with no corresponding immersed in liquid nitrogen and measure change in correlation range. the deformation of the foam. An explanation of this surprising result X-ray tomography was used to was given by measuring the scattering at measure density ﬂ uctuation in the metal smaller q not possible with SAXS, using oxide foams at better than 1 μm. small angle neutron scattering (SANS). The small angle scattering capabil- SANS results for 40- and 200-mg/cc are ity developed in this project has suc- shown in Fig. 2a. The scattering curves cessfully lead to a number of follow up show that in the 40- mg/cc sample there projects in support of various programs is scattering from spherical like objects at LLNL. A number of examples are the that are approximately 1 μm in diameter. measurement of void structure in highly No such features are observed in the energetic materials, the study of pore aerogels above 100 mg/cc. The scatter- structure and volume in Be capsules, ing on the μm-length scale comes from and the in-situ SAXS study of carbon large voids or holes in the aerogel. This aerogels immersed in liquid hydrogen. result is in agreement with our lensless

Lawrence Livermore National Laboratory 55 Technology

Corey V. Bennett Construction of a Frequency (925) 422-9394 Resolved Optical Gating [email protected] (FROG) Diagnostic

any efforts in the area of single-shot, waveforms. This repetitively averaged Mreal-time recording of complex measurement system would enable a ultrafast waveforms (those with 100 fs more accurate characterization of the to 1 ps detail and 100 ps to 1 ns record performance of the real-time recording length) have no way of measuring the systems in progress. true response of the diagnostic systems because of diffi culty independently Project Goals characterizing the true input waveform. FROG is measurement technique When no other measurement technique that produces a frequency vs. time map exists, it is diffi cult to know if details of an optical signal in a manner similar of the recorded waveform were real or to the way a musical score describes an artifacts of the recording instrument. acoustical signal. It does this by time- Frequency Resolved Optical Gating gating the optical fi eld with a time- (FROG) is a well known and accepted delayed replica of itself and recording “gold standard” technique for measur- the resulting output spectra vs. this de- ing such complex ultrafast waveforms. lay. It is essentially a spectrally resolved While a single-shot FROG system autocorrelation. This analogy is shown does not exist with the desired sensi- in Fig. 1a, with Fig. 1b depicting one tivity and time-bandwidth product, a typical implementation in which the repetitively averaged, slow-scan system, time gating is done through noncolinear could be constructed to measure these second harmonic generation.

(a) pp ff pp

Spectrogram, time–frequency analysis Frequency Time 2 ∞ E(t) contributes E(t) g(t – T) contributes 3g(W T) E(t) g(t – T) exp(–iWt) dt phase (i.e., color), g(t – T) only intensity, not –∞ to the signal pulse. phase (i.e., color), to the signal pulse. E(t) g(t – T)

0 T Time

(b) Pulse to be measured FROG is a self gated spectrogram, or spectrally Beam resolved autocorrelation splitter E(t – T) SHG Camera crystal Spectro- meter Esig(t, T) Variable E(t) delay, T Figure 1. (a) Analogy between FROG and acoustic frequency analysis; (b) typical implementation in which the time gating is done through noncolinear second harmonic generation.

56 FY07 Engineering Research and Technology Report Measurement Technologies

FROG control software in LabView Retrieved pulse FROG analysis software set/control scan parameters 1.0 5

0.8 4

0.6 Phase Intensity 3

Intensity 0.4 Phase 2 0.2

0 1 –2 0 2 4 Time (ps)

FROG hardware FROG data Measured Retrieved Frequency

Fiber or free-space input Time Figure 2. The FROG system. The process fl ow starts with the computer control windows in the upper left and fl ows in a counter-clockwise direction, resulting in the fi nal pulse shape at the top. The actual hardware we constructed is shown in the lower left. The lower right shows a typical experimentally measured FROG trace along side a retrieved, or best matched, FROG trace obtained using retrieval software shown in the upper right. The desired information, the actual intensity and phase of the pulse vs. time is reported by the software and shown at the middle top of the fi gure.

The goal of this project was to build tools, even if repetitively averaged, Related References and test a FROG system with time-band- are required to aid in the creation and 1. Trebino, R., et al., Georgia Institute of width product, and sensitivity necessary characterization of these new real-time Technology, http://www.physics.gatech. to measure ultrafast test signals in mul- recording instruments. edu/gcuo/subIndex.html tiple projects within LLNL, i.e., to add 2. Trebino, R., Frequency-Resolved Optical FROG to its measurement tool kit. FY2007 Accomplishments and Results Gating: The Measurement of Ultrashort Typical signals have energies as low We have successfully built and Laser Pulses, Kluwer Academic Publishers, as 5 pJ/pulse, pulse durations from 100 fs tested, and are in regular use of the 2000. to nearly 1 ns, time-bandwidth products FROG system shown in Fig. 2. One 3. Kane, D. J., and R. Trebino, “Character- up to 4000, and are approximately cen- example of the accomplishments of this ization of Arbitrary Femtosecond Pulses tered at a 1550-nm optical wavelength. FROG system was the measurement of Using Frequency-Resolved Optical Gating,” pulses matching in duration to within IEEE J. Quant. Electron., 29, pp. 571-579, Relevance to LLNL Mission 1 ps, and frequency chirp (slope of the 1993. Future high-energy-density science optical frequency vs. time) to within 2% 4. Trebino, R., K. W. DeLong, D. N. Fitting- experiments at NIF and other facilities of that reported by our temporal imaging- hoff, J. N. Sweetser, M. A. Krumbu, and B. will require ultrafast real-time measure- based diagnostic. The temporal imaging A. Richman, “Measuring Ultrashort Laser ment systems with performance well system makes a complete measurement Pulses in the Time-Frequency Domain Us- beyond current technology, i.e., ps and on one pulse at a time and can repeat the ing Frequency-Resolved Optical Gating,” even faster temporal resolution with measurement at 155 million measure- Rev. Sci. Instrum., 68, 9, pp. 3277-3295, 10,000:1 dynamic range. There are also ments/s. The FROG system is an average September 1997. externally funded national security ap- of trillions of assumed to be identical plications that require faster real-time pulses. The good agreement between diagnostics. What is needed is a 1-THz these two different diagnostics gives us real-time oscilloscope, 50 times faster high conﬁ dence in the accuracy of the FY2008 Proposed Work than the existing state of the art. Better new temporal imaging technology. FROG will continue to be used as a tool in further work on SLIDER, ROPER, and the temporal imaging systems.

Lawrence Livermore National Laboratory 57 Technology

Stephen P. Vernon Picosecond Response, (925) 423-7836 Recirculating Optical Probe for [email protected] Encoding Radiation (ROPER)

LNL is performing physics ex- The sensor architecture can be opti- Lperiments on the NIF, OMEGA and mized for detection of γ rays, x rays and Phoenix platforms, including those neutrons. We project that integration of addressing deuterium-tritium (DT) burn these sensors with advanced optical data physics, equations of state, and dynamic recorders offers the possibility of prob- material properties. The dynamic range, ing the DT burn rate with good detec- spatial resolution, bandwidth, and noise- tion sensitivity (4 x 103 g/cm2) and ~ ps robustness required in these experi- temporal resolution. ments are extremely challenging and surpass present capabilities. This project Project Goals is undertaking engineering reduction to At the end of the project we hope to practice of a picosecond response time, successfully demonstrate a fully char- radiation to optical down-converting acterized, optimized, prototype, high- detector that can address these require- bandwidth, x-ray sensor, and provide ments. a plan for integrating the sensor with a The Picosecond Response, Recir- next-generation optical recorder, to ad- culating Optical Probe for Encoding dress the possibility of creating an Radiation (ROPER) sensor is a resonant ~ ps response, γ detector for Weapons optical cavity consisting of high- Complex Integration (WCI) experi- reﬂ ectance mirrors that surround a ments on NIF. direct band-gap semiconductor detection medium (Figs. 1 and 2). Radiation Relevance to LLNL Mission absorption within the detection medium This project speciﬁ cally addresses induces a change in its optical refrac- the instrumentation requirements of tive index. The index change is detected WCI. It is well aligned with LLNL with an optical probe. The perturbation engineering focus areas and enhances in the refractive index modulates the LLNL’s core competency in measure- phase of the probe beam. Interferometry ment science at extreme dimensions. is used to convert the phase modulation We anticipate that, when available, ICF to amplitude modulation, down-convert- and HEDS experimental programs, and ing the radiation signature to the optical NIF will identify applications for these domain. detectors.

Material interaction Radiation & carrier excitation Amplitude modulation

Phase modulation & interferometry Optical probing Figure 1. The ROPER system. Radiation absorption within the ROPER sensor modiﬁ es the sensor optical properties. Interferometric detection of the phase-modulated probe beam converts the radiation signature to an amplitude-modulated optical signal.

58 FY07 Engineering Research and Technology Report Measurement Technologies

FY2007 Accomplishments and Results Pump-probe optical testbeds with both free-space and fi ber optical systems were implemented for sensor characterization. High fi nesse (f ~ 20) Fabry-Perot sensors, operating at or near 1550 nm, using radiation damaged InxG1-xAsyP1-y sensor material were fabricated and characterized. Proton ion, H+, implantation was used to tailor the carrier recombination time in the sensor medium and, therefore the temporal response of the detector (Fig. 3). Measurements made with 140-fs pulse duration, 800-nm Ti:sapphire excitation, and a 1550-nm CW probe show that sensor bandwidth can be tuned with ion dose. For H+ at 175 keV and 1013/cm2, an instrument-limited sensor temporal response of ~ 100 ps is observed. Current engineering efforts are focused on improving the fi delity and Figure 2. Fiber-based ROPER, pump-probe, optical testbed. A Ti:sapphire pump laser (140-fs pulse duration at 800 nm) is S/N in pump-probe characterization of used as a surrogate excitation source and 1550-nm CW fi ber-coupled probe is used for readout. The sensor is mounted on radiation damaged sensors using 800-nm the translation stage at the top center of the fi gure. Ti:sapphire excitation and a spectrally fi ltered slave optical parameter amplifi er at 1550 nm.

Influence of ion dose on sensor recovery time 0.040 Related References 1. Lowry, M. E., C. V. Bennett, S. P. Vernon, + H at 175 keV R. Stewart, R. J. Welty, J. Heebner, O. L. 0.035 H+ dose 1010 Landen, and P. M. Bell, Rev. Sci. Inst., 75, 0.030 pp. 3995-3997, 2004. H+ dose 1011 2. Bennett, C. V., and B. H. Kolner, IEEE J. 0.025 Quant. Electr., 36, pp. 430-437, 2000. 3. Lambsdorff, M., J. Kuhl, J. Rosenzweig, 0.020 A. Axmann, and J. Schneider, Appl. Phys.

Signal (v) H+ dose 1012 Lett., 58, 1881, 1991. 0.015

0.010 FY2008 Proposed Work 0.005 In FY2008 optimized ROPER + 13 H dose 10 sensors will be engineered for 0 1500 1700 1900 2100 2300 2500 2700 2900 integration with an advanced optical Time (ps) recorder. Our goal is demonstration of a working “front-end” for a high- Figure 3. Sensor characterization. Radiation damage (H+ ion implantation) is used to tailor the sensor response time. bandwidth x-ray detection system. Note that the sensor recovery time varies from 10 ns to 100 ps as the H+ dose is increased from 1010 (purple curve) to 1013 (green curve) ions/cm2. In parallel, we will work on an implementation plan for integrating the ROPER sensor and a recording system to address the possibility of γ implementing an ~ ps response detector for WCI experiments at NIF.

Lawrence Livermore National Laboratory 59 Research

John E. Heebner Serrated Light Illumination (925) 422-5474 for Deﬂ ection Encoded [email protected] Recording (SLIDER)

ltrafast, high dynamic range, Usingle-shot diagnostics are critical to performing experiments in high- energy-density physics (HEDP). In recent years, conventional single-shot recording instrumentation has under- gone minor incremental performance improvements. This instrumentation is largely based upon oscilloscopes and streak cameras that are fundamentally limited to trading off dynamic range for temporal resolution and do not scale well into the ps regime. More recently developed ultrafast optical methods can achieve 0.01 ps resolution but only over short record lengths of a few ps. Figure 2. Fabricated 1-cm-x-2-cm SLIDER defl ector consisting of a gold patterned coating on a GaAs No current technology is aptly suited waveguide. The prism array is manifested as a continuous gradient because the 60-μm pitch is too for high dynamic range measurement in fi ne to be resolved. the 1 to 100 ps regime. In this project, we address this tech- angular channels. A slow camera in the single-shot, high dynamic range record- nology gap by demonstrating a novel back focal plane of a cylindrical lens is ing (>8 bits) in a temporal regime (1 to technique for single-shot recording of used to record the swept signal in paral- 100 ps) that currently does not have any optical signals that is extendable to lel with high dynamic range (Fig. 1). strong technology base. The technique x rays. Using an auxiliary pump beam leverages a wide body of research and a serrated mask, a prism array Project Goals on ultrafast all-optical phenomena in pattern is optically imprinted onto the We aim to demonstrate an all-optical semiconductors and combines it with guiding channel of a semiconductor analog of the conventional streak camera well-established high dynamic range waveguide. The transient prism array that can meet upcoming HEDP end-user camera technology. defl ects signals in a stepwise manner requirements. Specifi cally, our goal is linearly encoding time into multiple to demonstrate the SLIDER concept for

Serrated Light Illumination Deflection Encoded Recording

Pump-activated Temporal Spatially-encoded UItrafast virtual prism signal signal deflection array Recording in parallel on high dynamic range camera Signal input Slab waveguide Serrated gold mask

Achieves Temporal Resolution Achieves Dynamic Range Figure 1. Schematic illustrating the SLIDER concept. The concept is based on the optically-induced deﬂ ection of an optical signal propagating in a slab waveguide. The deﬂ ection is caused by a sequential array of transient prisms that are created by a pump beam applied perpendicular to the guide through a serrated mask. The tradeoff between temporal resolution and dynamic range in traditional recording instruments is decoupled here by implementing ultrafast nonlinear mechanisms for the sweep followed by recording on a high dynamic range camera.

60 FY07 Engineering Research and Technology Report Measurement Technologies

optimize the pump-induced modifi cation of the refractive index of GaAs resulting Pump-activated from the interplay of the plasma effect, virtual prism charge screening, and the Burstein-Moss array effect. A ringdown test pattern consisting of ~1 ps pulses separated at 10-ps Ultrafast was used for a proof-of-concept. deflection Figure 3 shows the experimental Signal testbed setup. A cylindrical lens was used input to couple the 1-cm test signal beam into the waveguide. Cylindrical lenses on the output end focused the beam to a nearly To high dynamic diffraction limited spot. The pump beam range camera was expanded and applied perpendicu- larly to the guiding channel passing fi rst through the serrated mask. The pump and signal pulses were synchronized by use of a motorized delay stage. Figure 3. Experimental testbed setup showing the SLIDER defl ector mounted in a multi-axis coupling stage. The signal A recorded single-shot trace is shown and pump beam pathways are illustrated. in Fig. 4.

Relevance to LLNL Mission was negligible. A 60-μm prism array Related References The proposed work directly addresses pitch ensured sub-ps temporal discreti- 1. Walden, R. H., “Analog-to-Digital Con- instrumentation performance gaps in zation into defl ection encoded channels. verter Survey and Analysis,” IEEE J. Sel. engineering at LLNL. Future WCI and An ultrafast Ti:Sapphire based Area. Comm., 17, p. 539, 1999. NIF experiments will require single-shot, oscillator and chirped pulse regenerative 2. Heebner, J. E., et al., “Enhanced Linear high-dynamic range characterization of amplifi er provided a 0.15-ps pump pulse and Nonlinear Optical Phase Response of fusion burn histories. at 800 nm. This wavelength is suitably AlGaAs Microring Resonators,” Optics Let- above the bandgap of GaAs to be strongly ters, 29, p. 769, 2004. FY2007 Accomplishments and Results absorbing in the guiding layer generating 3. Hubner, S., et al., “Ultrafast Defl ection We fabricated and tested a SLIDER electron-hole pairs that subsequently alter of Spatial Solitons in AlGaAs Slab Wave- device (Fig. 2) constructed from a the refractive index in a transient manner. guides,” Optics Letters, 30, p. 3168, 2005. 0.6-μm core GaAs planar slab wave- Part of the pulse energy was tapped 4. Lowry, M. E., et al., “X-ray Detection guide. A gold serrated mask was pat- off to drive an optical parametric ampli- by Direct Modulation of an Optical Probe terned suffi ciently far above (1.2 μm) fi er (OPA) that generated an idler beam at Beam-Radsensor: Progress on Develop- the core so as to not perturb the guided 1920 nm that was subsequently frequency- ment for Imaging Applications,” Rev. Sci. signal, yet close enough such that dif- doubled to 960 nm for use as a test signal Instrum., 75, p. 3995, 2004. fraction of the illuminating pump beam beam. This wavelength was chosen to

SLIDER single-shot recording of ringdown test pattern 10000 FY2008 Proposed Work We plan to conduct a detailed characterization of the merits and 1000 drawbacks of the SLIDER concept with a reﬁ ned device. Numerous 100 technical challenges remain before the proof-of-concept can be turned

Intensity (counts) into a working instrument. Primary 10 technical challenges include overcom- ing dynamic absorption (free-carrier and two-photon) and pump beam 1 0 10 20 30 40 50 nonuniformities. We will further study Time (ps) the feasibility of combining SLIDER with an x-ray-to-optical converter for Figure 4. Recorded single-shot trace of a ringdown test pattern of ~1-ps pulses derived from a Gires-Tournois cavity with a 10-ps round trip time. The measured resolution is 3–4 ps characterizing fusion burn histories FWHM and the dynamic range is in excess of 1000. on NIF.

Lawrence Livermore National Laboratory 61

Technology

N. Reginald Beer Biological Sample (925) 424-2232 Preparation Infrastructure [email protected]

roplet-based microfl uidic systems detecting a specifi ed portion of a sample Dallow the ability to manipulate DNA via PCR. independent microdroplets as individual Isolating the PCR reaction in such reactors, enabling reliable and quantita- small volumes provides an order of tive sample processing and detection. magnitude reduction in overall detec- Aqueous droplet formation in an oil tion time by signifi cantly reducing the crossfl ow at low Reynolds numbers is total number of cycles required. Dilution highly reproducible using extremely of the fl uorescently generated signal simple microfl uidic devices. is largely eliminated in such a small We are now using an LLNL-created volume, allowing much earlier detec- chip (to generate and stop microdroplets) tion, since the brightness of the small to perform polymerase chain reaction droplet, for a fi xed quantity of product, (PCR) on viral nucleic acid samples. is much greater than the brightness of PCR is an enzymatic process that ampli- a larger sample with the same quantity fi es a specifi c DNA target sequence in of product. This capability will enable response to temperature cycling that the next generation of biothreat detec- leads to denaturation, annealing, and ex- tors to operate faster with lower reagent tension. Each microdroplet contains the consumption, and provide the ability necessary biochemical constituents for to amplify individual viral or bacterial selectively amplifying and fl uorescently genomes without interference.

Figure 1. Integrated and Labview-automated microﬂ uidic workstation used to optimize on-chip PCR devices. Shown are the computer, programmed pumps, imaging electronics, and thermal control system hardware.

64 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

tem integrated Watlow PID controllers and power supplies for Peltier Thermo- electric Cooling (TEC), a Coolsnap HQ2 CCD camera for fl uorescence imaging (data acquisition), a Redlake high-speed CMOS camera for droplet production dynamics analysis, two programmable KD Scientifi c infusion syringe pumps, an Excite metal-halide arc-lamp light source, a Nikon TE-2000U fl uorescence microscope, a Uniblitz shutter and controller, a cooling fan, and three Valco Instruments multiport valves. Integration and automation allowed repeatable and effi cient experiment processing, resulting in more data and faster iteration cycles. This work also provided an optimized fabrication and device surface coating process for the silicon substrate (Fig. 2). The optimized device used a Deep Reactive Ion Etched (DRIE) Figure 2. Silicon wafer fabrication process after DRIE, performed at LLNL. The wafer shown awaits anodic bonding to 0.5-mm-thick silicon wafer with topside Pyrex, dicing into individual devices, and hydrophobic surface coating. etching for the microchannels and backside etching for the fl uidic vias. Project Goals Relevance to LLNL Mission The device was then anodically bonded The result of this work will be the Optimized microfl uidic devices and to a 0.5-mm Pyrex wafer and diced into capability to fabricate the microfl uidic instrumentation systems enable greatly individual devices. devices, including an optimized mo- improved chemical and biological detec- Furthermore, device coating optimi- lecular vapor deposition (MVD) coating tors for countering emergent chemical zation was performed, with the results process for the microfl uidic channel and biological threats. The optimization that liquid treatment with SigmaCote surfaces, as well as the implementation of these technologies advances the core was the most cost-effi cient method, but of a Labview-automated instrumenta- microfabrication competencies at LLNL, molecular vapor deposition (MVD) with tion system to operate the microfl uidic while supporting a vital counterterrorism FDTS was the most resilient and effec- experiments. Deliverables include: program. tive method. 1. complete benchtop system integra- The results of the device optimization and optimization; FY2007 Accomplishments and Results tion are visible in Fig. 3, where a stream 2. automate data acquisition and con- As a result of this work, LLNL now of monodisperse droplets with diameters trol system; has an integrated and automated bench- of only 27 μm are visible after droplet 3. fabricate microfl uidic channels; top system for conducting microfl uidic generation and stopping on-chip. The 4. optimize microfl uidic model; and testing, including on-chip PCR amplifi - droplets are now ready to undergo the 5. optimize fabrication processes. cation and detection (Fig. 1). This sys- PCR amplifi cation and real-time detection process. Due to the optimized chan- Droplet diameter ~ 27 μm nel surface coating, the droplets do not wet the channel walls.

300 μm Figure 3. Image of the device in operation showing the generation and trapping of a monodisperse stream of 27-μm droplets. After droplet-stopping the device is thermally cycled to power the PCR ampliﬁ cation process.

Lawrence Livermore National Laboratory 65 Research

Raymond P. Rapid Defense Against the Mariella Jr. (925) 422-8905 Next-Generation Biothreat [email protected]

ioengineered and emerging patho- ligation-dependent probe amplifi cation Bgens represent a signifi cant threat to (MLPA). human health. The best defense against a rapidly expanding pandemic is to isolate Relevance to LLNL Mission the pathogen quickly from biological Sample preparation is a critical samples so that it may be identifi ed, char- requirement for biological assays and acterized, and have treatments developed is a major bottleneck in the process of against it. The one persistent technology detecting and identifying biological gap in the process of identifying and agents, particularly unknown/engineered/ quantifying the presence of pathogenic emerging pathogens. Viruses are an im- agents has been sample handling and portant category of pathogens because preparation that must precede any assay. some of its members, such as infl uenza Also, we need higher-performance, and smallpox, are extremely infectious multiplex assays for families of rapidly- and very virulent forms could result in mutating RNA viruses. sudden, massive pandemics. Viruses are often diffi cult to isolate due to their Project Goals small size (diameter < 200 nm), com- The objective of this project is to pared with the bulk of the particles in a replace burdensome manual techniques sample. Multiplex assays are also a key for sample handling and preparation part of this work. This project strongly with new automated technologies. We aligns with LLNL’s missions in national will use microfl uidics with ultrasonics, security. and electrokinetics to separate and purify viruses from biological and envi- FY2007 Accomplishments and Results ronmental samples. We will also create We have created new simulation less costly, but more general multiplex capabilities, along with experimental assays for viruses, using multiplex, validation, including full 3-D models

(a) 0 V (b) 2 V

Top of microchannel Buffer input MS2, transported

MS2 input

Bottom of microchannel 0.7 mm

Figure 1. Monochrome photographs of a low-conductivity solution of fl uorescently-labeled virus MS2, fl owing in a microchannel with top and bottom electrodes (not shown). Two solutions are being introduced from the left of the channel: the upper solution is a clean buffer and the lower solution contains the labeled MS2. In (a) there is no voltage applied and the laminar fl ow and relatively low diffusivity of the MS2 confi ne the MS2 to the bottom half of the fl ow. In (b) with 2 V applied perpendicular to the fl ow, the MS2 has been almost entirely transported up into the buffer solution, as desired.

66 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

Ultrasonic transducer is outside of viewing area We have initiated an extensive effort to collect published data on the electrophoretic mobilities of DNA, proteins, viruses, bacteria, and other biological particles that we will need to manipulate. These data, coupled with our modeling of transport versus pH and solution conductivity, have driven our Top electrode designs of microfl uidics-based sample (not used) processing systems. We used cyclic voltammetry to Clarified sample characterize the electrode processes, so that we could avoid the extensive electrolysis of water, with its collat- eral generation of bubbles. We still are Yeast To waste addressing the effects of pH changes, when we operate at or slightly above the overpotential. Non-specifi c binding of non-ligated, Top electrode PCR-amplifi ed probes to bead oligos (not used) has been a signifi cant and on-going problem for the MLPA assay. This problem has been addressed by optimi- Figure 2. Photograph of yeast fl owing in a 0.5-mm microchannel that has an ultrasonic standing wave, perpendicular to the image plane and running along the center of the fl ow channel. Ultrasonic transducer is outside of viewing area. zation of the probe design and MLPA assay parameters (probe concentration, nucleic acid concentration, ligation and with Monte Carlo models for transport for EBV, Adeno C and TTV as well as hybridization buffer properties, and of biological particles. In our model- spikes of EBV, AdenoC and murine ligation parameters). None of these ing and experimental research, we have herpes virus. modifi cations resulted in adequate re- investigated acoustics, electrokinetics We designed, fabricated, and tested duction of background signal and there- (including effects of electroosmosis), numerous confi gurations of microfl uid- fore the assay probes were redesigned electrochemistry, including local pH ics that demonstrated that we could use into a Molecular Inversion Probe (MIP) changes due to wall/electrode processes ultrasonic standing waves with high-Q format to allow digestion of unligated in our microfl uidic systems. at 10-μl/min fl ow rates to drive yeast probes that may cause a non-specifi c We have established quantitative (model organism for human cells) to the signal. assays for the titers of MS2, E. coli, S. energy/pressure-wave nodes and that A polymerase step was added to the cerevisiae, murine herpesvirus, torque we could use electrophoresis at 1.4 V protocol to adjust for the 5-ft truncation teno virus (TTV), Epstein-Barr Virus or slightly less to transport virus-sized of probes that commonly occurs during (EBV) and Adeno virus subgroup C (40-nm-diameter) latex beads. We have probe synthesis. A DNA polymerase (types 1 and 5). Prior to using human conceived of new separation confi gu- has been identifi ed for the PCR ampli- nasopharyngeal samples (NPS), we rations that use continuous fl ow with fi cation step that greatly increases the have used prepared mixtures of viruses dielectrophoretic forces (negative dielec- signal-to-noise ratio. The probe hybrid- both RNA and DNA with bacteria and trophoresis). We have established a ization and ligation protocols have been eukaryotic cells in our research on micro- quantitative fl ow-cytometer-based assay optimized so that the hybridization-liga- fl uidic separation techniques. Some for the yeast and bacteria with diameters tion reaction time is decreased from 4 h examples of these mixtures are the ranging from 0.5 μm to 10 μm. We to 90 min. This modifi ed protocol yields bacteriophage MS2, a RNA virus, with designed a fl uidic system that is capable a very high signal-to-noise ratio in a its prokaryotic host bacteria, E. coli; of excluding the host DNA/RNA from 10-plex assay format and is capable of and BSL-1, Risk-Group-1 virus such as well-characterized biological samples detecting Adenovirus in a clinical naso- the murine herpes virus, a DNA virus; consisting of cells, bacteria, and viruses pharyngeal sample. This is the fi rst time and eukaryotic S. cerevisiae (Baker’s using Isotachophoresis. the MIP technology has been modifi ed yeast). NPS collection from volunteers Our general research effort has been for use on a bead-based microarray. has commenced and the samples have coordinated with the Viral Identifi ca- Figures 1 and 2 illustrate our ac- been analyzed with quantitative assays tion and Characterization Initiative. complishments.

Lawrence Livermore National Laboratory 67 Technology

Raymond P. Validation of 3-D Acoustic Mariella, Jr. (925) 422-8905 Modeling of Commercial Codes [email protected] for Microﬂ uidic Systems

early-ideal, single-node, planar microchannel have never been realized Nstanding waves can be modeled and experimentally. It is unclear whether the established experimentally for waves non-ideality arises from intrinsic defects that are centered in rectangular micro- and inhomogeneities in the materials ﬂ uidic ﬂ ow channels. However, similar that are used in such microsystems or single-node, planar standing waves that whether it arises from non-ideality in the are proximate to one wall of such a assembly/bonding/packaging.

1000 μm x

500 μm 300 μm 200 μm

400 μm ~ V(W)

Water Glass Silicon PZT NOT TO SCALE Figure 1. Two-dimensional computational region for the ATILA fi nite element code. The planar structure is comprised of an upper glass layer, and a silicon wafer in the middle with a fl uidic channel etched into the upper surface. The structure is acoustically driven into resonance by a thin thickness-poled PZT layer, bonded to the bottom of the silicon wafer. The fl uid fl ow is perpendicular into the modeled plane. Dynamic fl ow behavior of the fl uid is not taken into account in this approximation.

Figure 2. The experimental microfl uidic (a) Bottom (silicon side) (b) Top (glass side) chip (H-bridge) test package, owing to similar designs found in the literature. Here, the PZT is bonded to the silicon layer, and plastic surface-mounted fl uidic Inlet ports ports are used to couple the fl uid and beads from a syringe pump into the focusing channel. The channel splits at the ends resulting in two inlet and two exit ports. A wide variety of separation, mixing, and Flow fractionation schemes can be realized with this design.

PZT driver

Focusing channel Exit ports Silicon layer Fluidic ports 10 mm

68 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

Project Goals tion of particles will signifi cantly en- the resonant frequencies of the ultra- Our goal is to use an existing code, hance LLNL’s engineering capabilities. sound for our confi gurations, where ATILA, that can be used to perform larger particles are captured at the 1-D, 2-D, and 3-D modeling of acoustic FY2007 Accomplishments and Results nodes of standing waves; and 2) the waves at ultrasonic frequencies within We used a commercial code, ATILA, fact that sub-1-μm scale particles are microsystems. We can use this capabil- which is already available, to perform largely unaffected by non-cavitating ity to simulate the ultrasonic pressure 2-D and 3-D modeling of relevant power levels of standing waves in the fi elds in a variety of structures with possible structures with microchannels microfl uidic channels. microchannels that could be used to and bonded piezoelectric transducers. The ATILA code and our results are manipulate particles of scale 2 μm or This enabled the fabrication and testing shown in Figs. 1 to 4. larger in water or similar aqueous fl uids, of several generations of microfl uidic including possible generation of nearly- systems with ultrasonic transducers Related Reference proximate standing waves. We also have bonded to enable the manipulation of Coakley et al., “Spore And Micro-Particle fabrication expertise to test and validate particles larger than 2-μm diameter Capture on an Immunosensor Surface in an the accuracy of these simulations. We via acoustic standing waves. We have Ultrasound Standing Wave System,” Biosens. compared these simulations with experi- experimentally validated 1) the code at Bioelectron., 21, pp. 758-767, 2005. mental measurements on microchannels in a variety of structures with micro- 106 channels passing particle-bearing fl uids Figure 3. Theoretical energy density as a in order to validate these simulations on 105 function of frequency for the 2-D channel relevant microsystems. described in Fig. 1. The amplitude of the 104 drive voltage is 1 V. Several frequencies show energy densities in excess of 5x103 103 Relevance to LLNL Mission ) 3

3 J/m . Large energy densities indicate Acoustic manipulation of μm-scale 102 potentially useful operating conditions ) (J/m particles is important for biosecurity o for manipulation of particles. Conversely, applications and for biological sample (E 101 there are regions where there is very little energy in the ﬂ uid, implying that all of handling and processing. Sample pro- 100 cessing is still performed in a tedious, the energy is in the elastic structure. manual manner today, requiring large 10–1 amounts of time from skilled technical 10–2 personnel. Having a validated simulation 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 capability to model acoustic manipula- Frequency (MHz)

1.454 MHz 2.939 MHz

Flow Flow

z 0.1 mmz 0.1 mm

x x

Pressure and force fields orthogonal to flow Pressure and force fields orthogonal to flow 20 2.0 15 1.5 10 1.0 y 0.5 y 5 Pressure (p0) Pressure (p2) x 4.8 4.9 5.0 5.1 5.2 5.3 5.4 5.5 5.6 5.7 x 4.8 4.9 5.0 5.1 5.2 5.3 5.4 5.5 5.6 5.7 Figure 4. Two representative operating conditions of the microfl uidic channel ƒ = 1.454 MHz and 2.939 MHz. The upper images are microscope views of the channel through the glass substrate as the fl uid and latex spheres fl ow from the inlet to the exit through the standing wave pattern in the channel. The lower images are the 2-D numerical estimate of the pressure fi eld from ATILA and a vector plot corresponding to the direction of the calculated acoustic force fi elds. Here the dark regions represent nodes in the sound fi eld and the light regions are antinodes.

Lawrence Livermore National Laboratory 69 Research

Kevin D. Ness Thermal-Fluidic System (925) 423-1856 for the Manipulation of [email protected] Biomolecules and Viruses

e are developing a reconfi gurable analytes at a specifi c location (Fig. 1). Top view of microchannel Wfl uidic system that demonstrates The analytes are then separated based the ability to simultaneously perform on their free solution electrophoretic separations, concentrations, and puri- mobility. fi cations of biomolecules and viruses using temperature gradient focusing Project Goals (TGF). Many projects throughout The project goals are to develop an LLNL, particularly those related to automated TGF instrument to improve pathogen detection, mitigation, and the separation resolution and throughput Bulk flow Electrophoresis protection, require the manipulation of when applied to front-end sample pro- biomolecules or viruses to accurately cessing of biological samples. Two spe- Figure 2. Automated TGF capture and concentration 1) quantify the presence of a particular cifi c application areas are identifi ed to (>2000-fold) of a fl uorescent dye. substance; or 2) synthesize and investi- demonstrate the novel sample manipula- gate the function of a molecule. tion capabilities inherent to TGF: 1) the This novel microfl uidic technology purifi cation and separation of different tions through enhanced collection, sepa- is an equilibrium gradient version of virus strains in complex samples; and ration, and purifi cation strategies. This is capillary electrophoresis (CE) that al- 2) improving the performance of protein facilitated by performing the necessary lows for the stationary fractionation and concentration and separation for in vitro front-end sample preparation through concentration (up to 10,000 x) of target transcription/translation (IVT) protein concentration procedures and removing analytes on the dimension of bulk or free expressions. noisy background signals/contaminants. solution electrophoretic mobility. In this This project supplies LLNL engineers technique a delicate balance is achieved Relevance to LLNL Mission with a novel capability to perform bio- in a microchannel between a net fl uid TGF specifi cally addresses LLNL’s molecular, viral, and cellular control in fl ow and an opposing electrophoretic interests in the detection of biomole- a fl exible format to address a wide range velocity gradient to capture charged cules, viruses, or cells at low concentra- of programmatic assay conditions.

FY2007 Accomplishments and Results Fluid TGF: The Technology Fluid Automated TGF capture. We dem- well well onstrated the successful capture and Micro-channel concentration (> 2000-fold) of a small ﬂ uorescent dye molecule in an auto- Fluid motion Electrophoresis mated sequential injection analysis system (Fig. 2). We improved the overall Hot Applied temperature gradient Cold instrument performance by improving the stability of the applied electric ﬁ eld

μEx u and fl ow rate, improving the linearity of the temperature fi eld, and increasing the E Net velocity x μEx u magnitude of the temperature gradient. Stable fl ow. We designed a pressure- μE u x controlled fl ow system with integrated Net velocity Capture point fl ow rate feedback to achieve the desired ~1 nl/min fl ow rate control. Our fl ow x rate control system enables steps, ramps, Figure 1. Schematic of temperature gradient focusing (TGF). Bulk fl uid motion (yellow arrow) is balanced and other functions (Fig. 3) necessary by an opposing electrophoretic velocity (red arrow) to capture analytes at a unique spatial location within for our project goals. a specifi c electrophoretic mobility range.

70 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

Linear ramping flow rate versus time 600 500 400 300 200 Flow (nl/min) 100 0 –100 0 200 400 600 800 1000 1200 1400 1600 1800 Time (s)

Figure 3. Flow rate stability of LLNL pump control system in a varying ﬂ ow rate ramping mode.

Linear temperature gradients. We used Modeling effort. For the initial mod- numerical modeling of the governing eling of the relevant fi eld variables equations (energy equation and Stokes (temperature, velocity, and voltage) we FY2008 Proposed Work fl ow) and multi-physics simulations use the commercially available fi nite In FY2008, we plan to 1) move (COMSOL Multiphysics) to determine element modeling package, COMSOL from fl uorescent analytes to biologi- optimal device designs for the genera- Multiphysics. To solve for the analyte cal samples (virus and protein); tion of thermal gradients (50 ˚C/mm) in concentration profi les in a TGF system, 2) build up a fl uorescently labeled a high-throughput microfl uidic structure. these known fi eld variables feed into a sample library (virus and protein) An example of the temperature gradi- Brownian dynamics simulation. From to characterize system performance ent as a function of applied electric there specifi c metrics, such as peak with biological samples; 3) organize fi eld is shown in Fig. 4. Our models height, peak width, throughput, and limit the library into biologically relevant demonstrate the need for “thick walled” of detection, are extracted to determine groupings; 4) determine how bio- capillaries and robust thermal mount- the effi ciency and resolving power of logically relevant groupings of our ing to the heat source. We achieve these a particular TGF run (Fig. 5). Design sample library relate to electropho- requirements using microfabricated guidelines and rules (geometric and op- retic mobility; 5) demonstrate an glass chips with etched microchannels erating conditions) were constructed for ability to separate three spiked viral in a copper package using compres- various applications to aid in the design samples in a “simplifi ed” back- sion mounting and a high performance and testing of TGF devices. ground (repeat with proteins) and thermal interface material. purify spiked virus from a “dirty” background of multiple contaminants (repeat with proteins); and Temperature at chip edge and fluid centerline 6) move to more relevant and com- 80 plex sample matrixes (nasopharyn- E m) geal and IVT protein productions). 70 μ

60 E = 0 V/cm 50 E = 50 V/cm E = 100 V/cm

40 Channel width (200 Temperature (°C) Temperature E = 150 V/cm 30 E = 200 V/cm E = 250 V/cm 20 –6 –4 –2 0 2 4 6 8 Microchannel gap (1 mm) Nondimensional axial distance (x/Lgap) Flow Electrophoresis

Figure 4. Numerical simulation results showing the Figure 5. Monte Carlo particle tracking simulations, varying temperature profi le in the focusing region of predicting the steady state shape and band location of our microfl uidic chip. The model includes externally three analytes with different electrophoretic mobilities applied temperature gradients, natural convective losses that are captured, concentrated, and separated using TGF. through the packaging, nonuniform Joule heating, and The top image is scaled to aid in visualization. convective/conductive conjugate heat transfer through the fl uid/wall structure.

Lawrence Livermore National Laboratory 71 Technology

Klint A. Rose Flow-Through Pyrosequencing (925) 423-1926 in a Microﬂ uidic Device [email protected]

yrosequencing is a recent DNA gel electrophoresis, are being pursued at Psequencing technique based on DNA LLNL. These processes, however, are replication. Using special cascading less amenable to the rapid sample analy- chemistry, a light-emitting reaction is sis needed to characterize the viral load triggered each time a nucleotide, A, C, G within a single clinical sample. or T, is incorporated into the complemen- By reducing to practice the micro- tary strand of DNA. The sequence of the fl uidic approach in which beads carrying DNA can be determined by iteratively a DNA template are washed with 200-nl introducing one type of nucleotide and volumes of the sequencing chemicals, we detecting the emitted signal. This method can reduce the sequencing cycle time per is primarily used for sequencing short nucleotide, reduce the volume of expen- lengths of DNA (up to 100 bases) and de- sive reagents used, and potentially increase tecting single nucleotide polymorphism. the number of bases sequenced. This Sequencing provides one option for sequencing method is shown in Fig. 1. the fi nal identifi cation step in an ongoing effort to start with a clinical sample and Project Goals identify all known and unknown viruses The four major goals of this project within it. Although sequence data is con- were: 1) create a microfl uidic chip for sidered the “gold standard” for microbial rapid, small-volume pyrosequencing; identifi cation, the implementation of this 2) replicate the biochemistry necessary technique is currently limited by time to bind DNA to a magnetic polystyrene and cost. Since identifi cation of a virus is bead; 3) replicate the sequencing bio- improved with longer sequences, the 100- chemistry; and 4) assemble an auto- base cutoff is also a limitation. Therefore, mated system to control sample loading, other identifi cation approaches, such sequential nucleotide injection, and as fragment length analysis in capillary signal detection.

G Step 1: load beads with DNAG G Step 2: inject dNTPs, detect light signal

3’ AG Polymerase 5’ GA 3 T T TC A CCA T TGAGGC AGTT TCCT C AGAC3

PPi A Sulfurytase A A ATP Luciferase

C C C

Figure 1. Description of microﬂ uidic pyrosequencing technique. Step 1: The DNA template to be sequenced is bound to a magnetic polystyrene bead then loaded into the center sequencing chamber of the microﬂ uidic chip. The particles are held in this region using an external permanent magnet. Step 2: Solutions containing dCTP, dATP, dTTP, and dGTP are sequentially injected into the chamber and washed over the bound DNA. Successful incorporation of a nucleotide initiates a cascade reaction resulting in the release of light.

72 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

40000

35000 dGTP

30000 100 μm 25000

20000

Intensity (A.U.) dTTP dTTP 15000

10000

5000 5 mm Water 0 0 200 400 600 800 1000 Time (s) Figure 2. Example of the multiport microﬂ uidic sequencing chip. The sequencing chamber Figure 3. Example of the detected intensity signal from a DNA sequencing reaction. has a 200-μm diameter and total volume of approximately 6 nl. The chips are microfabri- The height of each peak corresponds to the number of nucleotides incorporated. cated in a silicon substrate for high reﬂ ectivity and bonded to glass.

Relevance to LLNL Mission the detected intensity signal from a Detecting Single Nucleotide Polymorphisms This project addresses needs identi- DNA sequencing reaction. in the Dihydrofolate Reductase and Dihy- fi ed by LLNL for the detection of vi- 3. Successfully bound DNA template dropteroate Synthetase Genes of Plasmodium ruses at low concentrations (10 to 1000 to magnetic polystyrene beads using Falciparum,” Journal of Clinical Microbiol- copies/ml). Specifi cally, this project a biotin-streptavidin linker. ogy, 44, 11, pp. 3900-3910, 2006. addresses the need for rapid identifi ca- 4. Demonstrated magnetic capture of 4. Russom, A., N. Tooke, et al., “Pyrose- tion of individual viruses based on a the template-covered beads in the quencing in a Microfl uidic Flow-Through low number of DNA sequences. This sequencing chamber. Device,” Analytical Chemistry, 77, work has direct impact on LLNL’s Viral 5. Replicated the bioluminescence pp. 7505-7511, 2005. Identifi cation and Characterization reporter system for identifying Initiative (VICI). incorporation of dNTPs onto the existing DNA template. FY2007 Accomplishments and Results 6. Assembled an automated sequenc- We accomplished the overall goals ing system with fi ne fl uidic control, of this project by delivering an automat- multiple sample injection, and a de- ed microfl uidic system for sequencing tection component for low signals. bead-bound DNA. Specifi c results and accomplishments include the following. 1. Modeled sequencing chamber plans Related References to optimize light capture from se- 1. Margulies, M., M. Egholm, et al., quencing reactions. “Genome Sequencing in Microfabricated 2. Implemented two microfl uidic High-Density Picolitre Reactors,” Nature, sequencing chips. The fi rst has dual 437, 7057, pp. 376-380, 2005. input and output ports and a 1-μL 2. Ronaghi, M., “Pyrosequencing Sheds volume. The second version has Light on DNA Sequencing,” Genome multiple input ports (up to 8) and Research, 11, pp. 3-11, 2001. two output ports with a 6-nl volume 3. Zhou, Z., A. C. Poe, et al., “Pyrose- (Fig. 2). Figure 3 is an example of quencing, a High-Throughput Method for

Lawrence Livermore National Laboratory 73 Technology

Robin Miles Investigation of Hybridization (925) 422-8872 Times for DNA Microarrays [email protected]

arge DNA hybridization arrays (micro- Table 2. Factors affecting hyybridzation times. Larrays) are used to detect the presence of specifi c genetic sequences in samples. Factors affecting Effect Mitigation steps They are used for many applications hybridization time including pathogen detection, detection of Temperature Activity Maintain about 40 °C genetic mutations, and sequencing. Oligos, Concentration of nonbound Interaction time Increase concentration, mixing short strands of single-stranded DNA, are Number oligos bound Detection limit System design bound to a surface in spots. Complemen- Optics Detection limit System design Background Detection limit Wash step tary DNA strands present in the samples Competitive assay Interference (2) Design of assay bind (hybridize) to the surface-bound Hairpins in oligo Lower association rate (1) Design of oligos strands and are often optically-labeled Free vs. bound hybridization Steric effects lowers activity to indicate that the binding event has oc- for bound oligos(1) curred. Detection is determined through correlation of the location of known oligo Relevance to LLNL Mission achieved though increased concentra- sequences in the microarray and the indi- Microarrays are used in the biodetection/mixing. cation of a binding event with the DNA in tion program at LLNL to identify known Microarrays can have tens of thou- the sample. For a pathogen detection ap- pathogens in fi eld samples and ultimate- sands of spots per array. The Nimblegen plication, for example, the microarray can ly to help identify emerging pathogens system used by LLNL builds the microar- be populated with oligos with sequences through their RNA sequences. ray on a microscope slide over an area unique to a given pathogen. Positive measuring approximately 18 mm x 13 optical detection of binding implies the FY2007 Accomplishments and Results mm. The oligo strands arrive at the sur- presence of the pathogen in the sample. Typical assay processing steps in- face through a combination of convec- Microarrays can consist of thousands clude culturing of the sample, purifi cation and diffusion. A typical diffusion of different oligos, making microarrays tion of the DNA or RNA, converting coeffi cient for DNA strands is a very powerful tool for multiplexed RNA to cDNA, performing PCR ampli- 1 x10-7 cm2/s. Using this number, one pathogen detection or sequencing. One fi cation, labeling the DNA, performing can calculate the time required for drawback of microarrays for pathogen hybridization to the microarray, wash- strands to migrate via diffusion from detection is that they can exhibit long ing the array and reading the results. the edge of a hemisphere of a radius hybridization times, up to several hours, The processing time can be lengthy. calculated such that the enclosed volume making them unattractive for fast detec- Typical assay processing times for each contains the number of strands for a tion applications. step are listed in Table 1. given concentration of the strands in It is clear from Table 1 that cultur- solution. The results of this calculation Project Goals ing and hybridization times are the assuming a 106 oligo detection limit The long hybridization time makes main obstacles to obtaining timely as- is shown in Fig. 1 as the “diffusion” the microarrays less attractive for detec- say results. Culturing is used for initial curve. Similar results are plotted for the tion when thousands of samples must be amplifi cation and purifi cation and PCR “Maui” mixer, which is currently used to processed. Factors affecting the hybrid- can be used in some circumstances to enhance the mixing of the solution. The ization time were investigated to assess substitute for these functions. The elim- Maui mixer is a laminar fl ow device that if the process time could be minimized. ination of the culturing step will reduce pushes the sample fl uid back and forth the overall assay time. Several of the past the chip. While there is convection Table 1. Typical sample processing steps. factors affecting hybridization time are in the fl ow direction, vertical and lateral Processing step Typical processing time shown in Table 2. Of the factors shown mixing is likely to be the result of diffu- Culture 16 h in Table 2, increased concentration and sion only. Purification 1 h mixing can have a substantial effect on As the concentration of strands in PCR 1 h hybridization times without changing solution decreases, a greater volume of Label 30 min the assay or the hardware. That said, liquid is required to provide the required Hybridize 16 h assay/oligo models, including the ex- strands for detection and the reaction Wash 15 min Read array 15 min istence of hairpins in the oligos, which time increases to allow the strands to reduce activity, can eliminate gains migrate from the limits of the volume to

74 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

10 the bound oligos. The knee of the curve 9 is at about 10-12 to 10-11 M. Above this Rect. mixer 8 concentration, reactions happen in less Diffusion 7 than 1 h. Below this concentration, many hours may be needed to complete 6 the reaction. 5 Mixing can aid in reducing the 4 hybridization time between the knee and 3 the ultimate limit-of-use, which occurs Hybridization time (h) 2 when the number of oligos in the sample 1 is equal to the detection limit. Improve- ments over the Maui mixer would consist 0 10–14 10–13 10–12 10–11 10–10 10–9 10–8 10–7 10–6 of a device that provides convective mix- Concentration (M) ing from all parts of the sample volume. On the microscale, acoustic streaming, Normal chaotic advection, and manipulation Limit of available DNA Apparent limit of detection through electric ﬁ elds represent methods Figure 1. Calculated hybridization times for diffusion (diamonds) and Maui (rectangles) mixer. for effecting mixing in an inherently low- Reynold’s number regime. A mixer based on acoustic streaming instigated at the air-water interface of an array of bubbles, as shown in Fig. 2, was built and tested along the Maui mixer. Test data in Table 3 shows that acoustic streaming is at least PZT as effective as the Maui mixer. Future Air bubble effort will focus on understanding the effects of mixing in both the Maui mixer PZT as well as the acoustic streaming device on the surface hybridization and detection Air bubble Polycarbonate process for various concentrations of the analyte/sample. Microstreaming DNA probe Related References 1. Gao, Y., et al., “Secondary Structure Effects on DNA Hybridization Kinetics: Tape A Solution Versus Surface Comparisons,” Glass Nucleic Acid Research, 34, pp. 3370-3377, 2006. Figure 2. Schematic of acoustic-streaming mixer. 2. Zhang, Y., et al., “Competitive Hy- bridization Kinetics Reveals Unexpected Behavior Patterns,” Biophysical Journal, 89, pp. 2950-2959, 2005. 3. Gadgil, C., et. al., “A Diffusion- Table 3. Data of hybridization times for various conditions. Reaction Model for DNA Microassay Assays,” Journal of Biotechnology, 114, DNA Sample ConcentrationMixer Type Hyb. Time (h) Detected (Y/N) pp. 31-45, 2004. Source 4. Liu, R. H., et al., “Hybridization E. coli 4.6 x 10-11 Maui 2 Y Enhancement Using Cavitation -11 E. coli 4.6 x 10 Maui 1 Y Microstreaming,” Anal. Chem., 75, pp. E. coli 4.6 x 10-11 Acoustic-streaming 2 Y E. coli 4.6 x 10-11 Acoustic-streaming 1 Y 1911-1917, 2003. E. faecalis 6.5 x 10-12 Maui 2 Y 5. Michel, W., “Optical Study of DNA E. faecalis 2.2 x 10-12 Maui 2 Y – on verge Surface Hybridization Reveals DNA of non-detection Surface Density as a Key Parameter for Microarray Hybridization Kinetics,” Bio- physical Journal, 92, pp. 999-1004, 2007.

Lawrence Livermore National Laboratory 75 Technology

Erik V. Mukerjee Microfabricated Silicon Nib (925) 423-4841 for Low-Cost, Mutiplexed [email protected] Micro-Coating

hemical and biological detection large reservoir capacity is governed by Csystems are key areas of interest at macroscale machining and therefore the LLNL, directly supporting its mission of dispensing volume is controlled by dif- nonproliferation and homeland security. ferences in surface energies and not by Many modalities of sensor platforms total reservoir volume. are presently being pursued (e.g,. The effort of this work was to use BAMS, IMS, GC-MS), including novel a modifi ed coating structure similar to microscale detection systems. Current the meso-contact printing mechanism LLNL micro-cantilever-based detec- currently in use. Creating an in-plane tion systems have shown promise for silicon array of micro-coating structures both biological and chemical detection will enable much faster throughput of and identifi cation. This system couples polymer fi lm selection and trials. Film organic polymer (polyolefi n) swelling thickness uniformity across devices is 3473 -222.8 to mechanical stress causing resistance essential for device functionality. The changes in the sensing element as a sheer volume of polymers trials/selec- 3493.5 3493.5 function of analyte exposure. tion necessary for sensor viability and However, the fabrication and inte- selectivity almost certainly precludes gration of these systems involves time- serial coating methods. consuming serial processes for coating each micro-cantilever with the chemical- Project Goals ly sensitive polymer. The current coating The goal of this project is to fab- set-up consists of a single-machined ricate a multiplexed polymer micro- stainless steel nib (tip of a fountain pen) coating system. Near term application structure engineered to wick and store is specifi c to the GS chemical agent Major tick = 1000 Minor 50 (um) STS 1 fl uid into a reservoir until dispensed via detection project; however, the ap- -269 -269 contact with a higher surface energy plication space can be expanded to 3473 -222.8 material. Mechanical contact between include chemical/biological assays and Figure 1. Schematic of nib geometries including the the nib and cantilevers often causes microscale organic materials deposi- current stainless steel confi guration. the cantilevers to break. The fi xed tion processes.

-3968.5 Major tick = 1000 Minor tick = 50 (UM) 4031.5 1774.5 1774.5 -1725.5 -1725.5

-3968.5 Metal 4031.5 STS 1 STS 2

Figure 2. Illustration of ﬁ nal nib structure and mask (insert). Figure 3. Exploded schematic of nib, reservoir, and universal socket structures.

76 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

and 4) a set of larger demonstration structures (4x larger). The nib geometries varied in length (0.5 mm to 1mm, and 3.5 mm for stainless steeel); total reservoir volume (4 nl to 25 nl, and 400 nl for stainless steel); and volume to surface area (15:1 to 50:1, and 35:1 for stainless steel). The universal sockets facilitated rapid switching of nib confi guration during characterization runs and fi t both the nibs and the reservoir structures. The testing assembly consisted of two micromanipulator stages (X-Y cantilever mounted stage, X-Z nib mounted stage) and a stereoscope for visualization of Figure 4. Wafer level view of various nib confi gurations. cantilever coating. Results are as follows: Relevance to LLNL Mission overarching technical challenge was 1. Maximum fracture load Work on this device advances several thickness uniformity of the individual a. Average nib 1 through 4: 1.3 N processing techniques to extend LLNL’s fi nal nib elements. Cross-wafer etch b. Average nib 5 through 8: 3.8 N capabilities and will produce a signifi cant rates of the deep reactive ion etcher 2. Average spot size (0.1% polyolefi n/ technological impact on the instrumenta- (DRIE) varied as a function of position dioxane, nibs 5 through 8 on planer tion community. The resulting technolo- on the silicon substrate (1.8 μm/min silicon nitride surface): 423 μm +/- 5% gies will aid in the fabrication of deploy- to 2.4 μm/min, ~25%). This resulted 3. Nib thicknesses 20 μm to 30 μm able detection systems that are well in fi nal device thicknesses ranging most consistent dispensing aligned with the national security mission between 5μm to 125 μm on a single 4. Cantilever coating of LLNL, e.g., biodetection of pathogens silicon substrate. Changing localized a. Dummy cantilevers structures and chemical warfare agent detection. loading effects (i.e., amount of exposed (nibs 6 and 7 most consistent) silicon in a given region) improved etch b. Real cantilever sets ( nibs 6 and 7 FY2007 Accomplishments and Results uniformity to an acceptable level of to be tested and run against various In FY2007 the milestones that were ~10%. analytes) achieved include layout, fabrication, Each lithographic mask set produced From these results, optimal nib ge- characterization, and testing of sev- 1) eight nib geometries; 2) three reser- ometries (nibs 6 and 7) and thicknesses eral silicon micro-nib structures. An voir geometries; 3) universal sockets; (25 μm+/- 5 μm) were determined for consistent/uniform polyolefi n dispensing. This preliminary data suggest that the microfabricated silicon nib arrays can be used as a cost/time effective alternative to the serial stainless steel system currently in use.

Related References 1. Baller, M. K., “A Cantilever Array-Based Artiﬁ cial Nose,” Ultramicroscopy, 82, pp. 1-9, 2000. 2. Porter, T. L., “Sensor Based on Piezoresis- tive Microcantilever Technology,” Sensors and Actuators A-Physical, 88, pp. 47-51, 2001. 3. Sepaniak, M. “Microcantilever Transducers: A New Approach to Sensor Technology,” Ana- lytical Chemistry, 74, pp. 568A-575A, 2002. 4. Then, D., “A Highly Sensitive Self- Oscillating Cantilever Array for the Quan- titative and Qualitative Analysis of Organic Figure 5. Demonstration of 1–to–1 registering of nib elements with micro-cantilevers and Vapor Mixtures,” Sensors and Actuators size comparison with current stainless steel nib structure. B-Chemical, 117, pp. 1-9, 2006.

Lawrence Livermore National Laboratory 77 Technology

Satinderpall Advanced Polymerase Chain Pannu (925) 422-5095 Reaction Module [email protected]

olymerase chain reaction (PCR) is a performance of the APDS and ASTEP Pwidely used technique for amplify- programs by improving the reliability ing DNA signatures. It is used in many and minimizing the power requirements biodetection systems such as LLNL’s of the module. These programs are vital Autonomous Pathogen Detection System to the Laboratory’s mission in home- (APDS). To create a fi eld deployable land security. system, the current PCR reactors have to be miniaturized in terms of size, weight, Project Goals reagents, and detection times. This project implemented a fully This project describes the imple- characterized PCR module with integrat- mentation of an advanced PCR module ed polyimide heaters and temperature with an integrated polyimide heater sensors. The characterization will focus and an integrated temperature sensor. on measuring the thermal performance An older version of this PCR module of each module and the thermal stability is currently an integral component of as a function of time. the APDS and ASTEP programs. The current PCR module uses a series of Relevance to LLNL Mission discrete resistors to heat a thermal mass The need for advanced PCR modules during the PCR cycle. However, these that allow for the detection of biological discrete resistors and the temperature and chemical agents is directly aligned sensor are mounted on the outside with LLNL focus areas that outline of the thermal mass. The heater and key technology needed for biological the temperature sensor can be inte- security/defense of the nation. This grated into a polyimide fl ex cable and project will enable a robust, reliable mounted inside the thermal mass near PCR module that is fi eld deployable the reagent tubing. This implementa- in systems such as the APDS. There is tion will improve the thermal effi ciency also a signifi cant opportunity to produce and temperature sensor reading. This intellectual property in novel integration new PCR module will improve the and assembly techniques.

Sample tube

Discrete resistor Temperature PC board sensor

Copper thermal mass

Figure 1. Previous version of the PCR module with discrete resistive heaters and temperature sensor.

78 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

Sample tube

Resistive heaters Copper thermal mass Polyimide substrate

Figure 2. New version of the polyimide substrate for the PCR module with integrated Figure 3. Complete PCR system with control electronics used to test resistive heaters and temperature sensor. the PCR module.

FY2007 Accomplishments and Results these issues, the discrete resistors were For the fi rst test, two positive The previous version of the PCR replaced with a high-resistance, thin-fi lm controls and one negative control were module is shown in Fig. 1. It consists of nickel alloy laminated on polyimide. pumped into the reagent tubes. Forty- discrete resistors mounted to a printed Further, the temperature sensor was fi ve PCR cycles were run to verify that circuit board. The temperature sensor mounted onto the polyimide substrate the PCR was successful. As expected, is also mounted to the printed circuit as well to place it in contact with the signals from the positive controls in- board. The discrete resistors and the reagent tubing. The polyimide substrate creased with each PCR cycle and the neg- temperature sensor are mounted to the with integrated resistive heating elements ative control signal remained fl at. Next, copper thermal mass that has a through and temperature sensor was fabricated a target DNA molecule was inserted into hole for the reagent tube. The reactants (Fig. 2). the sample and a new test was performed. for PCR are contained in this tubing The integrated heaters on the poly- Again, the test was successful. The and reactants are kept in the middle of imide substrate were tested for long- signal from the positive control increased the thermal mass to provide a uniform term reliability, and passed success- with each cycle as well as the target temperature to the reactants. fully. The temperature sensor was also signal. The negative control performed as The previous version of the PCR integrated onto the polyimide substrate expected. module has several failure modes includ- and tested successfully. This polyimide Upon the successful completion ing delamination of the discrete resistors substrate was then integrated into the of these tests, this technology was from the thermal mass and a varying PCR system for testing and evaluation. transferred to the APDS and ASTEP temperature offset from the temperature The complete PCR system is shown in programs, and these programs are cur- sensor since it is mounted away from Fig. 3. rently integrating this technology into the reagent tubing. In order to correct their systems.

(a) (b) 4000 4000 ch1 ch1 3500 ch2 3500 ch2 ch3 ch3 3000 ch4 3000 ch4

2500 2500

2000 2000

1500 1500 Opt. signal (ADC) Opt. signal (ADC)

1000 1000

500 500

0 0 0 10 20 30 40 50 0 10 20 30 40 50 Cycle Cycle Figure 4. (a) Performance of the PCR module with only control samples and no target; (b) performance of the PCR module with the controls and a target.

Lawrence Livermore National Laboratory 79 Technology

Tiziana Bond Tunable Optical Cavities (925) 423-2205 for Gas Sensor [email protected]

race-gas analysis is critical for real- molecules, and to extend it to a miniatur- Ttime environmental monitoring, ized gas in-situ detection platform with weapons surveillance, combustion stud- built-in multiplexed detection potential. ies, and space exploration. Tunable diode Ground or avionic systems that re- laser absorption spectroscopy (TDLAS) spectively need unattended, fl exible or is a powerful approach for in-fi eld IR lightweight, highly sensitive sensors will chemical detection and identifi cation. greatly benefi t from this approach. Recently, MEMS tunable vertical cavity The technology relies on extended surface emitting lasers (VCSELs) have coupled cavity (ECC) MEMS-tunable been implemented for NIR TDLAS. We vertical cavity devices: the epitaxial are exploiting this technology to bridge material is engineered to align the laser a capability gap in sensing low weight emission to a specifi c absorption wavelength (coarse tuning), and the top suspended mirror causes a continuous scanning across the absorption lines of Figure 1. Cross-sectional schematic Tunable laser the gas (fi ne tuning) when defl ected by an of MEMS tunable ECC-VCSEL for output applied voltage. The device will be used 0 sensing. The presence of gas 2 TiO /SiO in the air gap quenches the laser Aluminum 2 2 for standard TDLAS. Ultimately, it can DBR pillar be described as a multipass cell with opti- emission when the resonance wave- SiNx length is tuned to correspond with cal gain (Fig. 1). In operation, the laser is an appropriate absorption line. Air gap electrically driven above threshold; the p+ GaAs gas fl owing through the air gap spoils the AlGaAs Gain gain-loss balance necessary for lasing by 3 QW active medium increasing the absorption losses within the cavity. Absorption is enhanced as the 40.5 period light is refl ected several times within the AlGaAs DBR resonant cavity, between the top and bot- n-doped GaAs tom distributed Bragg refl ectors (DBRs). substrate n-contact Ge/Au/Ni/Au

(a) (b) 560

) 550 –1 540 ) 530 –24 8 520 6 510 4 500

2 Threshold gain (cm 490 0 480 cm/molecule (10 760 762 764 766 768 750 755 760 765 770 775 Wavelength (nm) Wavelength (nm)

Figure 2. (a) Signature of 02 at ~760– HITRAN source; (b) modeled laser threshold gain displaying continuous emission tuning for 02 sensing.

80 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

(a) (b) Undercut Figure 3. (a) 3-D model of the tunable MEMS top contact protection VCSEL p-contact/ VCSEL. Note that the scale of the air (aluminum) electrostatic actuator gap has been exaggerated to clarify the lower electrode free-standing nature of the micomechani- VCSEL cal structure. (b) SEM picture of tunable p-contact Suspended membrane. TiO2/SiO2 DBR Membrane structure Suspended VCSEL epi & SiO2/TiO2 bottom n-contact DBR

Project Goals sustain the generation of a new class epistructure (Fig. 4), which performed Our goal is to establish complex fab- of compact, fi ber compatible optical well, demonstrating the good quality rication procedures to reliably reproduce gas sensors for real-time detection of of the material. We have set up optical electrostatically tunable MEMs vertical chemical agents. This will facilitate and mechanical measurement systems lasers. The project focuses on reduc- minimally invasive trace-gas analysis for (tunable) MEMS characterization. ing them to effi cient and miniature gas for next-generation weapons with built- Fixed and tunable VCSELs have been sensors. Finally, we wish to explore their in persistent surveillance; monitoring of fabricated, tested and are going through effectiveness, in selectively detecting the in-fi eld explosive detection; nuclear ma- troubleshooting for optimization. signature absorption lines of the gas of terial production activities and environ- interest, and the limit of detection (LOD) mental pollution; and healthcare. The Related References both theoretically and experimentally. 2-D nature of the technology enables 1. Welre, P., et al., “Near and Mid-Infrared We expect to be able to scan with a very other interesting applications such as Laser Optical Sensors for Gas Analysis,” Op- narrow linewidth (< 1 pm) the full 10- to multiplexed smart detection systems, tics and Lasers Engin., 37, pp. 101-4, 2002. 20-nm gas spectrum with actuation volt- adaptive imaging, beam forming, opti- 2. Kogel, B., et al., “Micromechanically ages < 10 V, and power consumption of a cal computing, and high power lasers. Widely Tunable VCSEL for Absorption few mWs. Spectroscopy at Around 1.55,” IEEE MEMs An example of gas spectral signature FY2007 Accomplishments and Results Conference, MA2, pp. 7-8, 2006. is given in Fig. 2 (a) for O2. Some initial In the fi rst year we started, as 3. Cole, G. D., E. S. Björlin, Q. Chen, C.-Y. estimates of the sensitivity show that planned, with a survey of the state-of- Chan, S. Wu, C. S. Wang, N. C. MacDonald, LODs of 100s of ppm (for a system level the-art technology reported in literature and J. E. Bowers, “MEMS-Tunable Vertical- resolution ΔP/P = 10-3) are achievable for low-weight molecules detection Cavity SOAs,” IEEE J. Quantum Elect., 41, σ -22 for gases with cross-sections NIR ~10 and investigation of gas species spec- pp. 390-407, March 2005. 2 cm /molecule. tral properties. We down selected O2 for a proof-of-concept at λ = 760 nm, Relevance to LLNL Mission based on membrane and epilayers Our project supports several appli- requirements and constraints. We have cations at the core of LLNL’s national assessed and established the required FY2008 Proposed Work security missions, from Stockpile Stew- processes on dummy samples (Fig. 3). We will complete the establish- ardship to Homeland Security. It will We have fabricated 760-nm LEDs on the ment of the enabling capability by fully characterizing the MEMS membrane and tunable VCSELs in (a) (b) 5 terms of refl ectivity, emitted power, 4 spectra, voltage tuning, and power consumption. Preliminary gas testing 3 experiments for O2 TDLAS at 760 2 nm will be attempted using existing Voltage (V) Voltage 1 facilities. The spectral and intensity cross-sensitivity to different gases will 0 0 12345 also be investigated. We will focus the Current (mA) efforts on gas testing and determi- Figure 4. (a) Light emission from LED before completing full VCELS process; (b) IV curve with the turn on <2V. nation of spectral and amplitude sensitivity of the devices.

Lawrence Livermore National Laboratory 81 Technology

Tiziana Bond Extending Lasers to Universal (925) 423-2205 Gates for Photonic FPGAs [email protected]

he photonic fi eld programmable gate reprogramming speeds are one to two and The basic array element of the Tarray (P-FPGA) is an all-optical three to four orders of magnitude faster, P-FPGA is the reconfi gurable universal integrated circuit that has the reprogram- respectively, and so we expect a pro- logic gate (Fig. 2). With feedback, the mability of software and the algorithm portional speedup in signal processing, gate can act as an optical clock source acceleration of application-specifi c encryption, graphics rendering, or genetic or as a SET-RESET fl ip-fl op, i.e., a hardware. The P-FPGA is a core enabling algorithms. Other advantages are effi cient single bit of memory. The gate is a mul- technology because of its speed and ver- gate usage, EMI insensitivity, and direct tiple section quantum well laser, with satility (Fig. 1). Compared to electronic compatibility with optical signals used in three lateral input optical ports with a FPGAs, computational and functional communications and photonic sensors. gate selection, and two lasing outputs. The specifi c logic operation performed by the gate is optically selectable. The Adaptive Transmit photonic analyzed mirrors are sampled grating distributed sensors data Bragg refl ectors (SG-DBRs) confi gured Numerical Network Optical signal simulations to let the output facet alternate as the analysis for processing laser wavelength is tuned by the optical routing Adaptive inputs. The mirrors and phase control decisions computing Dynamic Hardware sections are blue-shifted by quantum optical routers acceleration well intermixing (QWI) and thus are Image non-absorbing at the laser wavelength. analysis The control and slave sections provide Automated Photonic Graphics optical gain for the circulating power. design & FPGA rendering engineering In summary, the critical effects in the device operation are: 1) special AR/HR Solve NP Genetic Agile Revolving mirrors to create an alternating output problems algorithms encryption algorithms facet laser; 2) vernier effect to reduce tuning power, so that small refractive Optimization Field upgradable Fixing index change causes alignment of phase modules cracked codes with round-trip gain to jump to the next Avoids wavelength; and 3) gain-index lever obsolescence to reduce tuning power and generate enhanced wavelength modulation ef- Figure 1. The P-FPGA, a core enabling technology that will enhance system performance. fi ciency in multiple section lasers.

L = 1728 μm Mirror 2 sampled grating gate Mirror 1 sampled grating repeats 7 times total repeats 7 times total AR facet No L/4 No L/4 Slave Phase Control Phase Slave L/4 No L/4 No AR facet coating modulation grating modulation grating section section section section section grating modulation grating modulation coating m μ

l r Lnm = 94.89 μm Ls/2 = 50 μm Lo = 40 μm Ls/2 = 50 μm Lnm = 94.69 μm W = 2.5 l r Lb = 13.56 μmLLp/2 = 35.7 μm Lp/2 = 35.7 μm b = 13.53 μm Figure 2. Layout (not to scale) for a universal logic gate. Laser output is the red ellipse. The active layer is shown in red. The active layers for the mirrors and the phase control section have been slightly blue-shifted using quantum well intermixing and are shown in medium blue. Heavily intermixed section boundaries are shown in dark blue.

82 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

5 p-clad 77K 50 Ta O with O DQW As-grown T 2 5 2 T 40 3 4 Ta O no O 1 30 2 5 2 n-clad T2 T2 T 20 SiN 3 10 T1 0

3 peak shift (nm) PL Annealing temperature

Substrate 2 PL intensity (A.U.) PL Reflection (%)

λ/4 ARC 700 750 800 850 900 950 1000 0 Wavelength (nm) 0.8 0.9 1.0 1.1 Wavelength (µm) Figure 4. Photoluminescence (PL) of QWI for different annealing temperatures. PL excitation is 532 nm DPSS laser with excitation 2 Figure 3. AR coating ellipsometry measurement results. The value n = 1.904 for Ta205 with the 05 backfi ll is density 200 W/cm . Visible are the fi rst peak (left) from GaAs and λ the best (Rmin<0.1% and R<1% over 150 nm allows slight inaccuracies in /4 fi lm thickness). the second peak from InGaAs DQW. PL peak intensities increase after intermixing; a 38-nm shift is achieved at T3 (inset).

Project Goals expand the capabilities of all photonics- and used automatic measurement and This project focuses on establish- based sensors created at the Laboratory. post-processing techniques to evaluate ing and troubleshooting some of the For example, the sensors can adapt or the gain-index lever effect. fundamental processing steps in a fairly make decisions based on collected data, complex fabrication procedure in order or could transmit analyzed data rather Related References to reduce the universal gate to practice. than raw images when the downlink 1. Vahala, K., et al., “The Optical Gain Several of these processing steps (e.g., bandwidth is limited. Lever: A Novel Gain Mechanism in the Di- etched facet lasers, low-loss passive rect Modulation of Quantum Well Semicon- waveguides, mode-matched laser to FY2007 Accomplishments and Results ductor Lasers,” Appl. Phys. Lett., 54, waveguide interfaces, and waveguide We have accomplished the fol- pp. 2506-2508, June 1989. turning mirrors) are routinely per- lowing: 1) completed AR coating and 2. Stohs, J., et al., “Gain, Refractive Index formed on site. The four other neces- gratings confi gurations; 2) character- Change, and Linewidth Enhancement Factor sary capabilities are: deposition of AR ized AR coatings and demonstrated in Broad-Area GaAs and InGaAs Quantum- λ/4 coatings on etched facet lasers; a vertical facet deposition procedure Well Lasers,” IEEE J. Quantum Electron., mounting samples on patterned heat (Fig. 3); 3) purchased, patterned, and 37, pp. 1449-1459, November 2001. sinks; performing QWI; and defi ning evaluated heat sinks; 4) specifi ed and 3. Todt, R., et al., “Demonstration of Vernier SG-DBRs. The acquired techniques and procured a fi ve-mask set; 5) established Effect Tuning in Tunable Twin-Guide Laser professional relationships will expand a collaboration to perform QWI and Diodes,” IEEE J. Proceed., pp. 66-71, April LLNL’s processing capabilities and will observe induced blue shift (Fig. 4); 6) 2005. be valuable to future projects. fabricated multi-section lasers with new 4. Goddard, L., et al., “Rapidly Reconfi gu- engineered epitaxial material and a new rable All-Optical Universal Logic Gates,” Relevance to LLNL Mission chemical etch recipe to handle an InGaP Proc. SPIE, 6368, pp. 6368 OH-1-13, The two applications of the P-FPGA stop layer (Fig. 5); and 7) implemented October 2006. of most interest to LLNL are implementing hardware-accelerated algorithms and (a) (b) processing data from photonics-based sensors because of their benefi t to many Laboratory core applications. Some sample hardware-accelerated algorithms are 1) self-evolving genetic algorithms for effi ciently solving NP problems; 2) adaptive computing for enhanced speed in computationally intensive simulations or graphics rendering; and 3) reconfi gurable encryption algorithms for secure communications, e.g., constantly changing encryption algorithms or repair- ing broken algorithms in fi eld deployed units. On-chip data processing would Figure 5. SEM of (a) 3-side input InGaAs QW laser; and (b) lateral ridge etch.

Lawrence Livermore National Laboratory 83 Technology

Jack Kotovsky Compact Tunable (925) 424-3298 Optical Sensors [email protected]

his project reduces to practice a light source for illumination, since the FY2007 Accomplishments and Results Tscanning Fabry-Perot (FP) cavity- optical intensity in the fi ber is very low For the optical conditioner, a new based absolute conditioner, an “optical with these sources. Furthermore, this amplifi er was built that gave a 10-dB force probe” (OFP), and an “optical amplifi er will help make more accurate improvement in signal/noise ratio. gap gauge” (OGG). Two tasks were at- measurements when using the brighter A multi-LED package was obtained tempted to establish the FP conditioner: LED light sources. to experiment with the synthesized light 1) fabrication of a transimpedance ampli- The goals of the OFP and OGG are source. It consisted of LEDs ranging in fi er to reduce detection circuitry noise for to demonstrate tunable sensors appro- wavelengths from 570 nm to 850 nm, observing fringes; and 2) experimental priate for weapons and other load and covering most of the bandwidth of inter- verifi cation of a synthesized broadband displacement measurements. The sen- est. Two problems with this new light light source using two or three different sors must be inherently safe and able to source were a signifi cant difference in wavelength LEDs. We also demonstrate accurately perform measurements over intensity between the shortest wave- experimental results for the fi ber-based long periods of time without access to length LED and the longest wavelength OFP that is capable of sensing compres- the sensor. LEDs, and a spacing of the LEDs that sive loading transverse to the long-axis would not allow simple coupling of of an optical fi ber. We have achieved Relevance to LLNL Mission multiple LEDs into one fi ber. Several experimental results for a fi ber-based LLNL’s stockpile stewardship experimental setups were tried to solve OGG capable of measuring displace- mission demands microsensors and these problems. ments transverse to the long axis of a signal conditioning that are able to Uniform intensity can be achieved fi ber. These devices constitute a class of survive harsh environments of tem- at the cost of some loss of intensity compact optical sensors realized through perature, vibration, and radiation while by simply attenuating the light from the integration of microfabricated and maintaining very small form factors, the brighter LEDs; however, coupling machined components and strain sensi- inherent safety, and long-term accu- multiple LEDs into a fi ber proved to be tive fi ber Bragg grating (FBG) elements. racy. These constraints pose signifi cant a more diffi cult problem. It will require challenges and require low-power the implementation of a special coupling Project Goals methods, microscale features and care- structure, which will be addressed in a For the conditioner, a new ampli- fully confi gured devices and material future effort. fi er is needed primarily when using choices. These technologies address Figure 1 shows a 3-D schematic of an incandescent bulb or broadband these performance goals. the OFP. The device is a hybrid optical

Single-mode 1550-nm fiber Figure 1. Three-dimensional schematic with strain sensing and thermal of the OFP. compensation gratings

Etched groove for fiber alignment

6 mm Side view Compressive load Imparted axial strain

μm 750 140 μm

84 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

Figure 2. A microfabricated silicon component of the OFP, pictured next to a US penny for scale. Figure 4. SEM of a completed OFP. This view was generated by stitching together two separate images.

fi ber sensor consisting of a fi ber-based sive load to a tensile strain in the fi ber. single-transducer sensor is presented strain-sensing element integrated with Finite element simulations are used to in Fig. 5. In this experiment, the sen- micromachined transducers. The strain- properly dimension the fl exure elements. sor was subjected to a compressive sensing element consists of a single- A typical axial strain response is load resulting in an applied pressure mode 1550-nm optical fi ber containing included in Fig. 3. Under an applied varying from 40 to 180 psi. The sensor two FBGs (one for load sensing, one for compressive load, the transducers exhibits a repeatable linear response thermal compensation). The transducer generate a tensile strain along the fi ber, in the range of 60 to 180 psi with a elements, an example of which can be increasing the periodicity of the FBG sensitivity of 6.52 pm/psi. Assuming a seen in Fig. 2, consist of single-crystal and resulting in a red shift in the peak strain-induced peak shift of 1.21 pm/με silicon (SCS) structures that are micro- refl ectance that can be recorded via at 1550 nm, this converts to an axial machined to defi ne geometry for fi ber standard FBG interrogation hardware. strain transduction of 5.38 με/psi for alignment as well as mechanical trans- An SEM of a completed OFP this geometry. duction, converting an applied compres- is presented in Fig. 4. Data from a

Nodal solution

Step = 1 Sub = 1 Time = 1 Eptox (avg) RYS = 0 DMX = .002753 SMN = –.002251 1556.0 900 SMX = .002284 800 1555.8

700 ) E μ 1555.6 600 500 1555.4 400

300 Axial strain (

Bragg wavelength (nm) 1555.2 200 1555.0 100 –.002251 –.001243 –.235E–03 .773E–03 .001781 40 60 80 100 120 140 160 180 –.001747 –.739E–03 .269E–03 .001277 .002284 Pressure (psi)

Figure 3. Finite element simulation of the OFP strain response. The device modeled here has Figure 5. Typical response of the single-transducer OFP to a loading three sets of 40-μm ﬂ exures. At 200 psi, 773 με is imparted to the optical ﬁ ber sensor. cycle up to 180 psi.

Lawrence Livermore National Laboratory 85 Research

` Rebecca J. Nikolic Transport Behavior and (925) 423-7389 Conversion Efﬁ ciency in Pillar [email protected] Structured Neutron Detectors

oom temperature, high effi ciency and 3He neutron detectors are currently structure. The converter material (10B) Rscalable radiation detectors can be used for typical neutron detection. How- is deposited on the sample in the space realized by manipulating materials at the ever, they have major fi eldability issues, between the pillars by chemical vapor microscale. With micro-semiconductor- such as microphotonic interference, high deposition (CVD) (Fig. 2). pillars, a thermal neutron detector with voltage, and large device footprint. Also, In FY2007 we have achieved a fi ll fac- an effi ciency of over 70% is possible. the 3He gas makes air transport diffi cult. tor of 95+ % for pillar diameter of 2 μm, New material science, new transport New materials and device structures pitch of 4 μm, and height of 12 μm with behavior, neutron-to-alpha conversion are needed for revolutionary improve- 10B. By using this inter-digitated device dynamics and their relationship with ment in radiation detectors. By applying structure, charged particles (alpha and 7Li) neutron detection will be discovered with microtechnology methods to the area from the thermal neutron - 10B reaction the proposed structures. of neutron detection, we will be able to have a higher probability of impinging on make revolutionary improvements in the the Si P-I-N structures, where electrons Project Goals device effi ciency and fi eldability. and holes are created, which results in Our goals for FY2007 include: The proposed solution has a large high effi ciency. 1) neutron-to-alpha conversion effi cien- potential payoff in the area of national The pillar detector roadmap was fur- cy in high resistivity CVD 10B fi lms security. If all of the requirements in our ther developed in FY2007 using Monte and understanding of impurities; proposed device can be met, these detec- Carlo N Particle (MCNP) transport 2) thermal neutron detector demonstra- tors would be manufactured by partner- code to simulate a near-by neutron point tion with 20+ μm pillar height; ing with an industrial collaborator and source interaction with the structure, 3) electron and hole transport in verti- readily deployed to many agencies. after which the range of the charged par- cal high-aspect-ratio structures; and ticles was calculated using Ion-Range- 4) model of pillar detector that includes FY2007 Accomplishments and Results In-MAtter (IRMA) code (Fig. 3). both nuclear physics and semiconduc- A schematic diagram of the detector Further refi nement of the pillar detector device physics. is shown in Fig. 1. The platform consists tor model was carried out by determining of a P-I-N layer structure grown epi- the energy loss within the semiconductor Relevance to LLNL Mission taxially on an n+ silicon (Si) substrate. pillar. Silvaco’s Atlas was then used to Radiation detection requires both The epitaxial Si is dry-etched utilizing calculate the transport of the electrons neutron and gamma detectors that high-density plasma to form the pillars and holes generated by the alpha and 7Li are inexpensive, operable at ambient temperature, high effi ciency and robust. Currently available detectors typically fall short in at least one of these areas.

Neutrons

Pitch 10 Top of + – B silicon 10B pillars 7 Pillar height (a, Li) Charged particles Si + –

+ – Electron –hole pairs

Figure 1. Schematic diagram of pillar detector design. Figure 2. SEM of pillar platform ﬁ lled with CVD 10B.

86 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

100 “Chemical Vapor Deposition of Boron for 90 Neutron Detector Application,” Materials 80 Research Society (MRS) Spring Meeting, 70 San Francisco, California, 2007. 60 2. Deo, N., J. Brewer, C. L. Cheung, R. J. Nikolic, C. E. Reinhardt, and T. F. Wang 50 “Growth of Boron Thin Films by LPCVD

Efficiency (%) Efficiency 40 from Decaborane,” American Chemical 30 1 μm dia 2 μm pitch Society, Quincy, Illinois, October 2006. 20 2 μm dia 4 μm pitch 3. Welty, R. J., C. L. Cheung, C. E. Re- inhardt, T. F. Wang, “Roadmap for High 10 5 μm dia 10μm pitch Effi ciency Solid-State Neutron Detectors,” 0 0 20 40 60 80 100 SPIE Optics East, Boston, Massachusetts, Pillar height (μm) October 2005. Figure 3. MCNP and IRMA simulations of thermal neutron capture effi ciency versus pillar 4. Welty, R. J., C. L. Cheung, C. E. Rein- height for several pillar diameters and pitches. hardt, and T. F. Wang, “Solid-State Pillar Structured Thermal Neutron Detector,” IEEE Nuclear Science Symposium, Puerto Rico, particles in the Si diode portion of the without including the energy loss within October 2005. detector at sampled particle energies and the silicon pillars. 5. Nikolic, R. J., C. L. Cheung, T. F. Wang, heights within the pillar. The collected Neutron measurements were per- and C. E. Reinhardt, “Semi-Conductor Ma- charge was then calculated by scaling formed using a 12-μm pillar detector terials Matrix for Neutron Detection,” patent the energy deposited by the ionized fi lled with the 10B converter material, as fi led April 27, 2005. particles in the Si pillars by the percent- shown in Fig. 5. A 252Cf neutron source age of carriers that recombine before with neutron fl ux of 2.3 x 106 n/s was reaching the contacts, as determined embedded in 15-cm-thick polyethylene from Atlas charge transport simulations. blocks for the measurement. The mea- FY2008 Proposed Work The pillar geometry simulated in this surement time was 20 h. To obtain the In FY2008 we will pursue: work had a height of 20 μm, diameter of thermal neutron effi ciency, the observed 1) fabrication of Pillar Detector 2 μm and a pitch of 4 μm. The resulting events were compared with results from (20 μm pillar); 2) 10B coatings of energy spectrum for the pillar detector a Monte Carlo simulation. The thermal high-aspect-ratio structures and is shown in Fig. 4, for 4 million neutron neutron detection effi ciency is estimated understanding of stress origin and histories and a minimum energy thresh- to be 7.3 % with an error of 1.4 %. mechanics; 3) electrical, alpha, and old for detection of 100 keV. For the neutron measurements of Pillar geometry used the effi ciency simulated Related References Detector; and 4) the addition of re- is 18%. This is compared to the 24 % ef- 1. Deo, N., J. Brewer, C. L. Cheung, R. J. combination to Pillar Detector model. fi ciency simulated for the same structure Nikolic, C. E. Reinhardt, and T. F. Wang,

Long data set 105 106

5 104 10

104 103

103 Counts 102 Counts 102

101 101

100 100 00.51 1.5 2.0 2.5 0100 200 300 400 Deposited energy (MeV) Channel Figure 4. Simulated total energy spectrum (alpha + 7Li) for neutron interaction with Figure 5. Energy response of a 10B ﬁ lled 12-μm-tall pillar detector from neutron interaction 20-μm-tall pillars with 2-μm diameter and 4-μm pitch. with 10B.

Lawrence Livermore National Laboratory 87 Research

Jeffrey D. Morse Compact High-Intensity (925) 423-4864 Crystal-Driven Neutron [email protected] Source

ctive neutron interrogation is a high-voltage power supplies. A crystal- length and tip diameter is positioned on Apromising technique for identifi cation driven neutron source offers a means to the positively charged surface of the crys- of threats such as explosives or shielded compact the high-voltage power supply, tal. As the charge and voltage build up nuclear material. Neutrons are gener- ion source, and accelerator structure in during thermal ramping, the electric fi eld ated from fusion reactions when ions of an integrated design one to two orders intensifi es at the radius of curvature of the appropriate species and energy impinge of magnitude smaller than alternative needle tip until it becomes suffi cient to upon a target of a specifi c material having sources having similar neutron yields. cause fi eld ionization of background deu- adequate cross-section to yield a neutron. Pyrofusion has been demonstrated in terium molecules (3 mTorr) in proximity For example the d(D,n), or d(T,n) reac- recent years using pyroelectric crystals to the crystal surface. Once ionized, the tions can be facilitated by deuteron beams that generate large surface charge and deuterons are accelerated away from the hitting a deuterated or tritiated target, voltages when subjected to thermal ramp- positively charged high-voltage surface respectively. Accelerator-based technolo- ing. This effect is illustrated in Fig. 1. In of the crystal. gies are rather large, and include large this case, a tungsten needle of suffi cient Project Goals The objective of this project is to Pyroelectric crystal provide a comprehensive understand- D2+, D+ ing of the operation and limitations of n0 pyrofusion as a means of generating neutrons in applications for identifying potential threats. As such, a compact n0 (<1 liter), high-intensity (>107 n/s) source is desired for rapid detection of n0 hidden or shielded materials of concern. The goal of this effort is to demonstrate neutron yields in excess of 106 n/s using n0 the d(T,n) fusion reactions. The targeted Heat problems include effective generation Deuterated target D2 and control of deuteron ion beams, stable generation of acceleration volt- Figure 1. Illustration of pyrofusion effect. ages in the 100 to 200 kV range, and system integration and scaling.

Relevance to LLNL Mission Figure 2. Pyroelectric crystal mounting Active neutron interrogation is be- assembly including thermal control coming a key approach for the detection element. and screening of hidden threats including explosives and shielded nuclear materials. This results from the penetrating nature of fast neutrons and subsequent speciﬁ city of the resulting gammas produced through inelastic scattering and capture of the neutrons from materials exposed to the neutrons. As such, this technology will make a signiﬁ cant impact for applications in homeland security, the military, and intelligence gathering needs. Furthermore, it has the potential

88 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

Figure 3. Pryoelectric crystal mounted with arc sup- yields produced by other groups using pression dielectric guard ring at base of assembly. this technique. This is mainly due to the precision thermal control, along with the arc suppression approaches that enable long, persistent ion beams generated from a single thermal cycle.

Related References 1. Naranjo, Gimzewski, and Putterman, Nature, 434, 28, pp. 1115-1117, April 2005. 2. Geuther and Danon, J. Appl. Phys., 97, to provide new methods of interrogation a thermal electric element for rapid heat- 074109, 2005. that are not presently possible due to the ing and cooling of the crystal. Figure 3 3. Geuther, Danon, and Saglime, Phys. Rev. nature of existing neutron sources. In this shows a LiTaO3 crystal (1 cm thick by Lett., 96, 5, pp. 054803-1-4, February 10, sense, a crystal-driven neutron source 3 cm diameter) with a dielectric guard 2006. represents a new paradigm for active ring formed around the base to sup- interrogation of threats.. press fl ashover at the base electrode and crystal interface where high electric fi elds FY2007 Accomplishments and Results build up during thermal ramping. This We have made signifi cant progress has resulted in signifi cant improvement FY2008 Proposed Work on experimental efforts to understand in reproducing the ion beams for each During FY2008 efforts will focus the limits of pyrofusion for specifi ed thermal cycle. Figure 4 illustrates the ion on the inverse confi guration in which confi gurations, and further extend these beam current, and subsequent neutrons the target is mounted on the high- limits to useful operational regimes. The generated during a single thermal cycle. voltage surface of the crystal. This key contributions to the fi eld have been In this experiment, a deuterated tar- confi guration has the advantage that the design and assembly of a precision get (ErD2) was placed in series with a the ion beam will be better focused thermal control assembly for crystal Faraday cup to simultaneously measure on target without the need for focus- actuation and thermal cycling; high-volt- the ion beam while generating neutrons. ing optics due to the electric fi eld age engineering designs implemented to A He3 scintillator detector was used to distribution. Additional efforts will suppress high electric fi elds that result measure the neutron yield. Figure 5 continue the development of a gated in electron cascading and subsequent shows the x-ray spectra measured during ion source that will decouple fi eld breakdown events; and demonstration the crystal thermal cycle, which provides ionization from the crystal charge of record neutron yields using ungated, a measure of the voltage that the crystal is state, providing a separate, low volt- crystal-driven fi eld ionization sources. charged to, in this case >80 kV. Incorpo- age (<500 V) power supply for either Figure 2 illustrates the crystal rating active thermal control, these results pulsed or DC ion beam operation. mounting assembly, which incorporates have demonstrated >2 times the neutron

12 105 105 w/bg+att Neutrons 10 w/bg 4 raw 90 10 detected bg per 10 s 75 8 103 n/10s 60 6 >80 keV peak Counts 102 T(C) or 45

4 Beam current (nA) 30 101 2 15

100 0 0 0 200 400 600 20 40 60 80 100 120 140 Time (s) Energy (keV) Figure 4. Crystal-generated ion beam and resulting neutron yield during thermal Figure 5. Resulting x-ray spectra generated during ion beam measurements ramping cycle. indicating a crystal voltage on the order of >80 kV achieved during thermal ramping.

Lawrence Livermore National Laboratory 89 Technology

` Rebecca J. Nikolic Pillar Neutron (925) 423-7389 Detector Fabrication [email protected]

he goal of this project is to imple- FY2007 Accomplishments and Results Tment a set of process recipes for the The work carried out in FY2007 fabrication of the “pillar detector” de- was focused on the processing of the vice that identifi es special nuclear ma- pillar detector device. The process fl ow terials by detection of thermal neutrons. is shown in Fig. 1. We are using contact This new device can meet the demands photolithography to defi ne 2-μm- of high effi ciency and fi eldability while diameter pillars with 2-μm separation. maintaining a small device footprint, Next, plasma etching is used to etch enabling covert applications. the pillars with varying etch depth. The pillars are then fi lled with boron-10. Relevance to LLNL Mission This boron-10 material is responsible Thermal neutron detectors are used for converting the neutrons to alpha and to determine the presence of special lithium particles. nuclear materials. This falls under the The next step in this process is the Laboratory’s mission area of global planarization of the boron. This is a major security. task of this project. We have evaluated

Multi-step Process PR 10Boron

1. Pattern PR mask 2. Deep RIE etching 3. Boron deposition

Electrode

4. Boron planarization 5. Deposit contacts

Figure 1. Multi-step fabrication process for pillar detector.

90 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

Top of Silicon pillars 10Boron

Figure 2. Pillar array after CF4 /H2 /O2 plasma etch-back process. chemical mechanical polishing, wet Figure 2 shows a representative pillar etching and plasma processing to deter- chip after the boron etch step. Once the mine an appropriate recipe for this pla- device is planarized, electrical contacts narization. Plasma etching using a mixed can be placed on the top and bottom of gas composition of CF4/O2/H2 gave the the device. Aluminum is used for both best results in terms of a fast etch rate of the top and bottom electrode. For the 0.2 μm/min while maintaining a smooth top electrode, a higher pressure (25 mT and planar surface. vs. 6 mT) sputter deposition is required Each component of the reaction to coat the uneven topography. In pack- gas mixture is required for the fast etch aging the devices, we used conductive rate. The O2 removes the carbon layer silver epoxy to attach the pillar chip to that forms on the surface. If the carbon the package and simultaneously make layer is not removed the etch process electrical connection to the bottom will terminate. The CF4 and H2 are both contact of the pillar chip. Wire bond- responsible for the chemical etching of ing was then used to connect the top the boron by reacting with the ﬂ uoride electrode to the package leads. and hydrogen.

Lawrence Livermore National Laboratory 91 Technology

Klint A. Rose Flow Programmed Mesoscale (925) 423-1926 Assembly of Nanoengineered [email protected] Materials and Net Shape Components e have reduced to practice a bottom- powerful, low-cost capabilities to the 2) demonstrate the synthesis of porous Wup mesoscale assembly tool capable fabrication of density- and composi- polymer layers with defi ned, varying of integrating highly dissimilar materi- tion-varying layers. The highly con- density profi les, via the computer- als and producing features with critical formal nature of the deposition allows controlled electrophoretic deposition of dimensions in the submicron range. This the fabrication of near-net shape high submicron polystyrene spheres; and system enables nanometer-scale precision precision parts. The lack of need for 3) convert an existing Brownian dynam- synthesis of millimeter-scale materials vacuum conditions allows high-speed ics simulation code for this application. and parts in a simple, low-cost system. and low-cost nanomaterial synthesis in We perform this synthesis by a simple benchtop system. Relevance to LLNL Mission computer-controlled electrophoretic The system is available for follow-on The proposed project is a seed effort deposition in a miniature aqueous work to capitalize on the broader capa- for a new, integrated, mutually support- deposition cell interfaced to computer- bilities it makes accessible, including the ing, and dynamic multi-project port- controlled automated fl uidics. Nanopar- programmed synthesis of parts incorpo- folio in novel nanomaterial synthesis ticle solutions are available in a wide rating dissimilar and custom materials, capabilities for mesoscale manufactur- and ever-growing variety of material the in-situ fabrication of hemispherical ing. This portfolio is directed to LLNL compositions, morphologies, and surface and other net-shape parts, and ultimately application areas in target materials and chemistry states. We introduce mixtures the electronically-controlled biochemi- structures, sensor materials, and biode- of various precursor solutions into the cally templated nanoscale self-assembly tection devices. The short-term payoff deposition cell, and deposit nm- to μm- of nanostructured composite and func- is the creation of new nanoscale target scale layers by the pulsed-fi eld electro- tional materials. structures. The proposed technology is phoretic deposition of particles onto sub- ideal for combining novel nanomate- strates. We use post-deposition sintering Project Goals rial components such as engineered to achieve densifi cation of these fi lms. The goals of this project were to: nanoparticles, carbon nanotubes, and The high degree of control and 1) produce a powerful and fl exible pro- others into complex materials and wide range of heterogeneous materials totype system for mesoscale assembly, structures. made accessible by this approach brings ready to be leveraged for future work;

Power supply Figure 1. Schematic of the mesoscale assembly system. A FloPro unit provides automated valving and pumping to move particle solutions FloPro unit through the deposition cell. The power supply provides either constant voltage or constant current across the particle-laden ﬂ uid. The ﬂ uidics and the power supply are controlled via Labview software for full automation. • Computer • LabView • DAQ interface

• Fluidic cell • Electrophoretic deposition geometry Nanoparticle colloid percursors • Variable substrates

92 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

FY2007 Accomplishments and Results custom built miniature electropho- 3. Film synthesis: Demonstrated We have accomplished the overall retic deposition cell. the deposition of thick (> 50 μm) goal of this project by delivering the 2. Process testing and reduction to fi lms of polystyrene particles using automated mesoscale assembly system. practice: Carried out a series of particles of a single diameter; also Specifi c results and accomplishments parametric tests to quantify and demonstrated stacked fi lm layers as include the following. optimize the aqueous electropho- shown in Fig. 3. 1. Instrumentation (Figs. 1 and retic deposition process to allow 4. Density-gradient materials. Pro- 2): Built a prototype system by predictive control; tested multiple duced a two-layer fi lm with a combining a mini FloPro system electrode materials, particle materi- base layer of 200-μm polystyrene (Global FIA) for automated control als, solution properties, and applied particles and 80-nm gold particles of the sample solutions, LabView- electric fi eld. (Figs. 4 and 5). based control architecture, and a

4 μm

Banana plugs 5 mm

4 μm Pt coated Si chips Spacers 5 mm

Figure 2. Exploded view of the electrophoretic deposition cell. The cell holds the platinum Figure 3. Examples of particle fi lms of single particle types. Top: 1-μm-diameter yellow- coated electrodes at a fi xed distance to maintain a uniform and constant electric fi eld. green polystyrene particles deposited approximately 150 μm thick. Bottom: 200-nm- Depending on their charge, the particles are deposited onto either the cathode or the diameter red polystyrene particles deposited approximately 60 μm thick. Insets show SEM anode. images of the top surface of the respective fi lms.

4 μm 1 μm

500 nm particles

Gold

200 nm Polymer particles

Figure 4. SEM cross-section of a two-layer particle ﬁ lm. The bottom layer is approxi- Figure 5. SEM cross-section of a two-layer particle ﬁ lm. The bottom layer is approximately mately 60 μm thick and comprised of 200-nm-diameter polystyrene particles. The top 17 μm thick and comprised of 200-nm-diameter polystyrene particles. The top layer is layer is approximately 15 μm thick and comprised of 500-nm-diameter polystyrene approximately 300 nm thick and comprised of 80-nm-diameter gold particles. particles.

Lawrence Livermore National Laboratory 93 Technology

Christopher M. Laser Pantography Spadaccini (925) 423-3185 [email protected]

aser Pantography (LP) is a unique a portable NMR spectrometer (Fig. 2), Llithographic technology created at and the patterning of tungsten sensors on LLNL in support of a number of projects the tips of diamond anvils for the study and programs. LP is the only technol- of material properties at high pressure ogy that can routinely draw patterns on (Fig. 3). 3-D objects such as cylinders, spheres, For over two decades the various LP fl ats, and diamonds. These patterns are systems have been physically separate usually metal, and are typically used as from most of the micro- and nanofabri- circuits for sensors and actuators. cation facilities and personnel. This proj- In one example, LP is being used to ect endeavors to collocate the systems pattern coils on the tips of microcath- with other capabilities and integrate with eters intended for use in the brains of existing programs and projects in the stroke patients (Fig. 1). These coils can micro- and nanotechnology area. be energized with a continuous current, creating a magnetic moment that will Project Goals attempt to align with an exterior fi eld, The purpose of this project is to such as the magnetic fi eld of an MRI. ensure the future vitality of this unique This magnetic interaction will cause capability. Specifi c goals include: the catheter to defl ect, in effect steering 1. moving the two LP systems and all it. By having three orthogonal coils on associated support hardware to the the catheter, the tip can be defl ected in micro- and nanofabrication center any direction, providing a mechanism at LLNL in order to better integrate for guiding the catheter into a desired with and leverage other capabilities; blood vessel. Other uses of LP include 2. training new technical staff in the the fabrication of microreceiver coils for technique; and

Figure 1. Orthogonal saddle coils on polyimide cylinders for MRI guided catheters.

94 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

Figure 2. 150-mm OD helical coil for a portable NMR spectrometer. Figure 3. Designer diamond anvil with LP-generated electrodes and contact pads.

3. educating the LLNL micro- and Training of new personnel and edu- customers and other microfabrication nanofabrication community so that cation of the micro- and nanofabrication personnel via a series of seminars and LP can be integrated with existing community at LLNL has been ongoing individual meetings and tours to give and future projects and program- throughout this project. New engineers the community at large a better under- matic work. and technicians have been involved standing of the capability. with patterning 3-D components for As a result of this work, the LP Relevance to LLNL Mission the projects that LP currently supports: capabilities have been both physi- The ability to fabricate microscale portable NMR, MRI guided catheters, cally and intellectually integrated patterns on arbitrary 3-D surfaces and diamond anvils. These personnel with the micro- and nanofabrication (slopes, spheres, cylinders, and com- have also participated in planning and community at LLNL. This is critical pound curvatures) is a unique capability strategy meetings with customers to to maintaining and advancing this im- available only via the LLNL LP process. better understand individual project portant and unique technology that is These systems and processes have ser- needs. LP experts have actively mar- capable of patterning features on any viced critical program needs in the past keted the technology to potential new 3-D surface. and the vitality of the technique needs to be maintained in order to do so in the future. The availability and advancement of LP will contribute to the core microfabrication competencies at LLNL that provide vital support to both internal and external customers.

FY2007 Accomplishments and Results Two LP systems have been moved to the LLNL microfabrication facility. These systems include a full fi ve-axis tool with a diode laser and three-axis system with an Ar-ion laser. Also included were laser system enclosures, fl oating optical tables, and associated computer controls and electronics. Figure 4 shows the fi ve-axis LP system in its enclosure with computer controls at its new location. In addition to the lasers and associated equipment, an entire set of other sup- Figure 4. An LP system in its new location. port equipment, including a wet processing bench and sputter tool, was moved.

Lawrence Livermore National Laboratory 95 Technology

Todd Weisgraber Bridging the Length and (925) 423-6349 Time Scales from Atomistics [email protected] to the Microscale

olymer networks of interest to Project Goals PLLNL’s mission undergo structural The goal of this project is to bridge and mechanical changes over time. the length and time scales from atomis- These changes occur due to radiation tics to microscale. damage, contaminant transport and reaction, and prolonged deformation due to Relevance to LLNL Mission physical loads. To date, useful molecular LLNL has an interest in polymer dynamics (MD) and ﬁ nite element anal- networks and their structural changes. ysis (FEA) efforts have been performed. A fundamental numerical approach to These efforts, however, focus on the understand and characterize what nano and continuum scales, respec- happens to the micro and nanostruc- tively. The missing link in this effort is tures of polymeric materials is a central a “meso” or micron-to-centimeter scale need of LLNL’s commitment to DOE’s that uses interaction potentials gener- Enhanced Surveillance Campaign. ated by MD simulations and supplies time-dependent material properties to FY2007 Accomplishments and Results continuum models. This last year has The mechanical properties of resulted in a capability to generate 3-D polymeric materials are a function of foam networks with the ability to track the bonds between crosslinks and the Brownian walkers with steric effects crosslink density of the material. Given and local interaction potentials derived that the bulk material length scales are from MD simulations. typically on the order of centimeters and

Figure 1. 3-D vertex model of a polymer foam. The ﬁ nal structure (shown in red) is evolved from an initial cubic lattice (shown in green).

0.20

0.15

0.10

0.05

0.20 0.15 0.20 0.15 0.10 0.10 0.05 0.05

96 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures the polymer structure is on the order This approach can also be modiﬁ ed of nanometers, modeling of polymeric to track the interaction of radiation with materials is inherently a multiscale the polymeric medium while accounting FY2008 Proposed Work problem. In the work being performed for the degradation in bonds, crosslinks, We have submitted a proposal to on this project, a 3-D representation of and resulting mechanical properties. incorporate these enhancements and the microstructure with a quantiﬁ able Furthermore, the vertex model provides a validate with experiments. Continuing crosslink density was adapted from a starting point to get a glimpse of poly- the work would have high impact to vertex model. This model allows the user meric microstructure of real systems. the program since it aligns with their to prescribe mechanical properties to With some calibration to experimental goals and would also strengthen the the bonds between crosslinks, such as data, this model can be used to postulate relationship between Engineering and Hookean springs, and allows automatic microstructural properties. the Enhanced Surveillance Campaign. tracking of crosslink density. Addition- To date, we have implemented a 3-D ally, the numeric polymeric media can vertex polymer foam model that enables be mechanically assayed, using tension, a new way to assess bulk material proper- compression and shear, to yield con- ties (Fig. 1). The model also includes a tinuum property information. In parallel capability to characterize contaminant with this adaptation of the vertex diffusion in the foam (Fig. 2). Our pro- model, a Brownian dynamics simulation grammatic customers see promise in this was adapted to track random walkers mesoscale approach since it is less costly through the 3-D medium. than MD simulations and they would like These random walkers “react” with to further enhance the current model. the medium at crosslinks in accord with the interaction potentials generated Related References by MD simulation. It has been shown 1. Grimson, M. J., “Structure and Elasticity that the modulus of polymeric material of Model Gel Networks,” Mol. Phys., 74, is directly dependent on the crosslink pp. 1097-1114, 1991. density of the material. As the ran- 2. Flory, P. J., Principles of Polymer dom walkers traverse the medium as a Chemistry, Cornell University Press, Ithaca, function of time, they sever crosslinks, New York, 1953. modifying the modulus of the material and yielding a time-dependent change in mechanical properties.

Figure 2. Contaminant particle (gold) represented by a random walker in the vertex model. Crosslinks within the containmentʼs range of inﬂ uence are highlighted in red.

0.20

0.15

0.10

0.05

0.20 0.15 0.20 0.15 0.10 0.10 0.05 0.05

Lawrence Livermore National Laboratory 97 Technology

James F. McCarrick “HE-Less” Detonator (925) 423-8182 Modeling [email protected]

combined modeling and experimental FY2007 Accomplishments and Results A effort was undertaken to see if exist- In the modeling portion of the ing simulation tools at LLNL contain a project, an existing 2-D ALE hydrocode combination of physics models suitable model of pentaerythritol tetranitrate to study the initiation of high explosives (PETN)-based exploding bridge wire (HE) by an electrical arc. (EBW) detonators was modifi ed to simulate an arc-driven “air bridge” in Project Goals place of the metal bridge wire. To avoid Our goals were to 1) determine if the attempting a fi rst-principles model of shock initiation model implemented in the arc itself, the electrical behavior of a particular ALE hydrocode is capable the air bridge was matched to experi- of reproducing previously measured arc mental measurements. An extensive initiation data; and 2) if so, to map out matrix of studies was performed in a pa- the space for an arc-based detonator that rameter space containing the initial arc would not require its own internal HE radius, the source capacitance, source (an “HE-less” detonator). The experi- inductance, and source voltage. Figure mental effort provides a map from fi reset 1 shows a typical result of the simulated parameters to shock pressures, and the pressure fi eld in PETN that was suc- modeling effort gives a map from shock cessfully ignited by the air bridge. pressures to initiation. The intent was to see if the simulated “go vs. no-go” results resembled Relevance to LLNL Mission data taken from a set of measurements Arc initiation phenomena are of detonator sensitivity to electrostatic directly relevant to LLNL stockpile discharge (ESD). In particular, we safety studies and to improved initiation looked to reproducing the observed system plans. trend of lowered threshold with decreased circuit inductance, which we chose as our indicator that the shock- 20 ns: based initiation model was suffi cient. 11 kbar Further simulations would then map out the arc-driven shock pressures useful in a detonator intentionally engineered to 50 ns: use arc initiation. 5.7 kbar However, as shown in Fig. 2, while the simulations do show arc initiation of PETN, they do not reproduce the observed dependence of threshold 90 ns: energy on inductance. It has not yet 4.7 kbar been determined if this is due to the assumption that only shock pressure participates in the initiation mechanism, or to the lack of a signifi cant energy transport mechanism in this particular Figure 1. Air bridge model of arc-initiated PETN. mode of operation of the hydrocode, or to a purely numerical effect not discernible in simple baseline tests.

98 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

Shot 12 9 kV 1.0 2000 1800 0.9 No-go Go 0.8 1600

0.7 1400

0.6 1200 Noisy higher 0.5 1000 order frequency components E (mJ/Mil)

0.4 (m/s) Velocity Simulations 800 0.3 600

0.2 400 Main 0.1 signal 200 below 200 m/s 0 0 50 100150 200 250 300 350 1234 56789 L (nH) Time x10-7 Figure 2. Simulated and measured energy thresholds. The agreement at 150 nH is forced by fi xing the Figure 4. Sample PDV results, showing noisy return signal and low energy and fi nding the threshold in initial arc radius. The resulting radius (0.9 mil) is then used to fi nd expansion speed. the energy thresholds at 100 and 250 nH.

In the complementary experiments, build a map between electrical param- signal-to-noise is not strong and the er- photonic Doppler velocimetry (PDV) eters and the pressures produced in the ror bars in the expansion speed are fairly was used to measure the velocity front resulting arcs. This map could then be large. of an arc in air, using fi reset parameters combined with the modeling study to Furthermore, even at the high end scaled to match the detonator ESD fi nd fi reset parameters that would proof v~1000 m/s, an order-of-magnitude safety study. The setup is shown in Fig. duce initiation in a variety of useful HEs estimate of the pressure yields P ~ nmv2 3. The purpose was to determine if the (since the model itself lacks a fi rst- ~ 12 atm for a mass corresponding to shock pressure required to drive the ex- principles coupling between the fi re- molecular nitrogen and ambient density. pansion of the air was of the right order set and the resulting arc). However, as This is several orders of magnitude below for shock initiation of HE; and, if so, to shown by the typical results in Fig. 4, typical shock initiation thresholds. Tests were done for 20-, 40-, and 150-mil arc lengths with source voltages in the 8 to 10 kV range, a 1-nF source capacitance, and a circuit inductance of about 250 nH. If the peak pressure in the arc volume resulted from all of the energy initially Fireset stored in the fi reset, it would be of the order 600 atm at the 40-mil spacing, assuming an initial arc diameter even as large as 1 mm. Gap 40 mils EBW no wire Thus, these results again point to the importance of understanding the correct energy transport, since a considerable amount of energy is not accounted for.

PDV Prove Related References 25.4 cm 1. Lee, E. L., and C. M. Tarver, “Phenom- enological Model of Shock Initiation in Het- erogeneous Explosives,” Physics of Fluids, 12, pp. 2362-72, 1980. 2. Strand, O. T., et al., “Compact System for High-Speed Velocimetry Using Heterodyne Techniques,” Review of Scientiﬁ c Instru- Figure 3. Experimental setup for PDV measurement of arc-driven air expansion speeds. ments, 8, pp. 83108-1-8, 2006.

Lawrence Livermore National Laboratory 99 Technology

Mark A. Schmidt Chip Slapper Detonator (925) 422-9118 Processing for Rapid Prototyping [email protected] and Hydrodynamic Properties

ext generation weapons require vapor deposition followed by an In-situ Nreliable, compact, energy-effi cient Polymerization (SLIP) process. initiation systems that will perform for the lifetime of the stockpile. Chip slap- Relevance to LLNL Mission per detonators are now being evaluated Effi cient chip slapper fabrication = 0.0078” for future use. There are limitations can enable future initiation systems with in the present technology and high improved safety and performance that costs associated with tooling for each would be relevant to both DOE and DoD confi guration as we strive to produce missions at the Laboratory. reliable, low-energy slappers. The present manufacturing process uses thick- FY2007 Accomplishments and Results Figure 3. Prototype slappers laser-machined into Au on fi lm, hybrid-circuit, photolithographic Fabrication of the initial rapid pro- an alumina substrate. processing with solvent-based polymers. totyping plan included thick fi lm PVD Two main problem areas exist: coatings, gold electroplating combined 1. for the multilayer metallization pro- with photolithography, and femtosecond between runs. Full-scale slapper models cess—poor alignment tolerances and laser machining to defi ne fi nal geom- are initially laser machined from paper high tooling costs; and etries (Fig. 1). All coatings and laser to check dimensions, and repeatability is 2. for the fl yer polymer process— defi nition were within manufacturing well within requirements. solvent-based aging issues and in- tolerances for a slapper device. The processed substrates must be adequate hydrodynamic properties. Software was written to “direct coated with a protective fi lm before write” slapper arrays on a processed laser-machining operations to avoid re- Project Goals substrate (Figs. 2-3) with the fem- coating and contamination of deposited This three-year project has two tosecond laser (Fig. 4) and is in its fi lms. It takes approximately 20 s to cut goals: reduce the costs associated with preliminary stage of implementation. a simple slapper pattern on a substrate. creating chip slapper patterns by ap- Once the software for rapid prototyping A substrate full of fi nished slapper pat- plying rapid prototyping techniques; is complete, it will include a feature that terns can be machined in approximately and evaluate alternative detonator fl yer allows the user to arbitrarily change the 6 min. materials by using LLNL’s SolventLess slapper dimensions quickly (seconds) The SLIP process consists of mono- mers in the vapor phase being mixed in a nozzle system and deposited onto a substrate (Fig. 5). Deposited fi lms range in thickness from 6 μm to 100 μm. Sub- strates remain at room temperature or can be heated. Stoichiometric polyimide Au Au has been confi rmed with a PMDA/ODA Al molar ratio of 0.83 for fl at substrates. Al2O3 This is quite different from the molar ratio of 1.17 that produced stoichiometric fi lms for ICF targets (Fig. 6). The process parameters for fl at substrates are not the same as those for spherical substrates, which made it challenging to fi nd the proper operating conditions Figure 1. Schematic of thick-fi lm deposition and pattern- Figure 2. Processed substrate with various bridge widths since we started with the ratios for ICF ing of Au over Al using femtosecond laser machining to defi ned. applications. defi ne fi nal geometries.

100 FY07 Engineering Research and Technology Report Micro/Nano-Devices and Structures

Deposition chamber PDMA ODA

Deposited layer

Substrate

Figure 4. Femtosecond laser micromachining labora- Figure 5. Solvent-less process for deposition of polyimide ﬁ lms. tory, which includes a 4-axis motion control system that enables rapid prototyping of slappers and slapper arrays.

Signifi cant progress was made model accepts an electrical stimulus (in extending past work on exploding the form of a charged external capacitor bridgewire models into a small slapper – a fi reset) and predicts burst time, burst FY2008 Proposed Work detonator initiation model. Early at- current, and fl yer speed as functions of Next year we will test the rapid tempts looked at fairly sophisticated 2-D chip slapper materials, geometry, and prototyping pattern and compare representations, but have been reduced excitation. Recent work focuses on the it to a standard part made by pres- to simple 1-D models for the sake of proper initiation response from the ac- ent manufacturing methods. The model robustness, and for verifi cation ceptor HE, building toward comparison experimental data will be evaluated with other modeling attempts and vali- with experiments planned for the fi rst alongside the model. Efforts will also dation against experiments. The present quarter of FY2008. include implementing a robust 2-D slapper detonator initiation model and comparing it to the 1-D model for 2500 consistency. For the polyimide fl yer material, we need to determine the temperature(s) that will produce the 2000 desired mechanical fi lm properties, establish adhesion characteristics on m)

μ multiple materials, and investigate 1500 adhesion promoters, if required.

1000 Crystallite diameter (

500 Stoichiometric ICF targets range for RP 0 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 PMDA/ODA molar ratio Figure 6. Stoichiometry per molar ratio as determined by crystallite formation.

Lawrence Livermore National Laboratory 101

Technology

Tracy D. Lemmond Semantic Graph (925) 422-0219 Hierarchical Clustering [email protected] and Analysis Testbed

LNL has invested more than a non-numerical algorithm output, such as Ldecade in inference methodologies the dendrogram shown in Fig. 1. for semantic graph analysis to facili- For hierarchical clustering tech- tate knowledge discovery. However, niques, performance evaluation frequent- knowledge discovery systems based on ly requires highly subjective assessment semantic graphs are rarely optimal for and, therefore, extensive analyst inter- enabling the construction and testing of action. We have provided tools to help these algorithms. guide analysts in evaluating many aspects We have addressed this defi ciency of algorithm performance. by building a testbed to serve as a companion to analysts for the rapid Relevance to LLNL Mission prototyping of graph-based algorithms New large knowledge discovery in an environment equipped to evalu- systems may revolutionize our ability ate and compare their effi ciency and to perform real-time inference activi- performance. Due to the unique needs ties, since massive graphs are capable of LLNL to process massive graphs, of fusing terabytes of multisource data we have constructed this environment that conceal complex relationships. to emphasize hierarchical clustering Effective graph analysis techniques can methodologies as the foundation of the expose these relationships, leading to analysis process. more competent decisions through a more thorough understanding of vital, Project Goals and frequently obscured, signature The testbed provides a suite of behaviors. This testbed makes the cre- modular algorithm components, catego- ation of these analysis techniques more rized according to their typical function effi cient and cost-effective, leading to in graph analysis algorithms, which may more productive use of semantic graphs be combined as desired to create distinct and the knowledge discovery sys- algorithms. Algorithm evaluation takes tems that leverage them in support of place within a testing framework suitable LLNL’s intelligence/security mission. for the evaluation of numerical algorithm results as well as for the visualization of FY2007 Accomplishments and Results Our testbed has been built as a plug- Figure 1. A dendrogram representation of in to Everest, a relatively mature graph the hierarchical decomposition of a semantic visualization environment at LLNL. graph. Each node in the dendrogram During FY2007, we completed the represents a cluster. algorithm interface, called the Algorithm Builder, accompanied by a library of “tasks” comprising metrics and operations commonly used by graph analysis algorithms. In addition, we have completed an interactive environment for exploring algorithm results. The Algorithm Builder has been constructed on the premise that most graph analysis algorithms are composed of multipurpose metrics and

104 FY07 Engineering Research and Technology Report Engineering Systems for Knowledge & Inference graph operations. These algorithms can the capability to visualize/drill down into algorithms. The accompanying exten- be modularly represented in a fashion the partitions produced by the algorithm sible task library will form a solid foun- consistent with a plug-and-play para- process and track intermediate results. dation for future enhancement, with the digm, such that individual algorithm Figure 3 shows the primary results expectation that it will prove benefi cial components can be easily modifi ed. In window, depicting pertinent information for the next generation of semantic the spirit of the fl ow diagram concept, a relating to a cluster at the 20th partition graph inference algorithms at LLNL. classical approach to algorithm represen- of Newman’s agglomerative decomposi- tation, we found it convenient to model tion algorithm. The information shown Related References our algorithms using semantic graphs. includes the cluster size; its node type 1. Duch, J., and A. Arenas, “Community Figure 2 shows a screenshot of the distribution; high degree nodes of its Detection in Complex Networks Using Algorithm Builder, in which the task child, parent, and sibling clusters; the Extremal Optimization,” Phys. Rev. E, 72, library is depicted to the left (catego- point in the algorithm at which it was 027104, 2005. rized by function, e.g., metrics, fl ow, or created; the point at which it formed 2. Fortunato, S., V. Latora, and M. Marchiori, diagnostics), and the Girvan-Newman two new clusters; and the high degree “Method to Find Community Structures community decomposition algorithm has nodes of every other cluster within the Based on Information Centrality,” Proc. Natl. been constructed to the right. partition. Other information that is avail- Acad. Sci., 70, 056104, 2004. After executing an algorithm, an able includes a breakdown of runtime 3. Guimerà, R., and L. A. N. Amaral, analyst must assess both its computational for individual algorithm tasks, a graph “Cartography of Complex Networks: Mod- effi ciency and the quality of the results of modularity (partition quality) as it ules and Universal Roles,” J. Stat. Mech., produced. Many decomposition algo- evolves throughout the algorithm, and P02001, 2005. rithms are hierarchical, i.e., they proceed the ability to track nodes of interest from 4. Newman, M. E. J., “Fast algorithm for De- through a series of operations that incre- one partition to the next. tecting Community Structure in Networks,” mentally break the graph into clusters. We All of these analysis capabilities Phys. Rev. E 69, 066133, 2004. refer to each stage of such an algorithm combine with others to provide a com- 5. Newman, M. E. J., and M. Girvan, “Find- as a partition. We have built an interac- prehensive view of algorithm output by ing and Evaluating Community Structure in tive result analysis system that attempts to which an analyst may effi ciently make Networks,” Phys. Rev. E 69, 026113, 2004. facilitate quality assessment by providing performance assessments of multiple

Figure 2. Screenshot of the Algorithm Builder. Girvan-Newman decomposition is shown on the right. Potential Modul. Generalize Mo. Now Partition Cr...

Modularity Matr.. Modularity Yes No Calculate Edge.. Local Between Yes

Adjacency Matr.. Calculate Infor.. Yes Yes Yes Start Calculate Edge... Maximum Remove Link Modularity Minimum Leading Eigenv.. No Yes Maximum

Yes Biconnected C.. Split Graph Stopping Condition Stop

Figure 3. Primary results window. Left Cluster ‘al-qaeda’ Dendrogram level is 20 to right: cluster list by partition; local dendrogram; and node type frequency, given al-qaeda (parent) organization agent a selected cluster. knowledge al-qaeda morocco (sibling) resource (focus)

task al-qaeda (child) algeria (child) location

al-qaeda (grand child) intellig (grand child)

Lawrence Livermore National Laboratory 105 Research

Yiming Yao Decomposition of Large-Scale (925) 422-1922 Semantic Graphs via an [email protected] Effi cient Communities Algorithm emantic graphs have become key Newman makes it impractical for graphs Scomponents in analyzing complex with more than approximately 104 systems such as the Internet or biologi- nodes. Our goal for this project was to cal and social networks. These types develop methodologies and correspond- of graphs generally consist of sparsely ing algorithms capable of effectively connected clusters or “communities” processing graphs with up to 109 nodes. whose nodes are more densely con- From a practical standpoint, we expect nected to each other than to other nodes the developed scalable algorithms to in the graph. The identifi cation of these help resolve a variety of operational communities is invaluable in facilitating issues associated with the productive use the visualization, understanding, and of semantic graphs at LLNL. analysis of large graphs by producing subgraphs of related data whose inter- Relevance to LLNL Mission relationships can be readily character- In recent years, LLNL has developed ized. Unfortunately, the ability of LLNL semantic graph technologies capable to effectively analyze the terabytes of of fusing disparate facts from diverse multi-source data at its disposal has sources into massive semantic graphs remained elusive, since existing decom- to facilitate inference of complex and position algorithms become computa- anomalous behaviors embedded within tionally prohibitive for graphs of this the data. A critical challenge in effective- size. We have addressed this limitation ly applying this technology to the Labora- by developing more effi cient algorithms tory’s mission is to decompose massive for discerning community structure that graphs into meaningful subgraphs that can effectively process massive graphs. an analyst can effi ciently interrogate to identify these behaviors. This research Project Goals represents a signifi cant contribution to Current algorithms for detecting LLNL’s counterterrorism, biodefense, community structure are capable of and nonproliferation missions, because processing only relatively small graphs. effi cient decomposition methodologies The cubic complexity of Girvan and will provide the foundation for information analysis environments enabling large-scale data mining, and information Table 1. Computation time reduction over Girvan & Newmanʼs method. discovery and visualization.

Graph NodesLinks Tgn T R(%) FY2007 Accomplishments and Results Erdos972 5488 8972 1180.1 129.6 89.0 During FY2007, we completed a Hep-Th 7610 15751 1524.5 670.0 56.1 graph clustering implementation that Kohonen 4470 12720 22.8 8.7 62.0 leverages a dynamic graph transforma- Power 4941 6594 723.0 550.4 23.9 tion to more efﬁ ciently decompose large graphs. In essence, our approach Tgn: Girvan & Newman’s time (min); T: our time (min); R = 100(T-Tgn)/Tgn dynamically transforms the graph (or

106 FY07 Engineering Research and Technology Report Engineering Systems for Knowledge & Inference

Table 2. Computation time for parallel graph clustering. We have developed a verifi cation tool for decomposition algorithms, based Graph NodesLinks CPUs T Q on a novel integer linear programming G10m 10000000 43749984 4 28.9 0.40 (ILP) approach, that computes an exact G100m 100000000 298437392 32 706.0 0.72 solution. We have used ILP methodol- G1000m 1000000000 2624753446 512 710.0 0.78 ogy to fi nd the maximum modularity and corresponding optimal commu- T: computation time (min); Q: modularity nity structure for several well-studied graphs in the literature (see fi gure). The above approaches assume that subgraphs) into a tree structure con- computing, we have exploited the fact modularity is the best measure of qual- sisting of bi-connected components that large semantic graphs tend to be ity for community structure. In an effort interconnected by bridge links. This sparse, comprising loosely connected to enhance this quality metric, we have isomorphism allows us to compute dense node clusters. When imple- also generalized Newman’s modularity edge betweenness, the chief source of mented on distributed memory com- based upon an insightful random walk ineffi ciency in Girvan and Newman’s puters, our approach performed well interpretation that allows us to vary decomposition algorithm, much more on several large graphs with up to one the scope of the metric. Generalized effi ciently, leading to signifi cantly billion nodes, as shown in Table 2. The modularity has enabled us to develop reduced computation time. Test runs on rightmost column of Table 2 contains new, more fl exible versions of our a desktop computer have shown reduc- the associated Newman’s modularity, a algorithms. tions of up to 89% (Table 1). metric that is widely used to assess the In developing these methodologies, Our focus this year has been on quality of community structure. we have made several contributions to the implementation of parallel graph Existing algorithms produce results both theoretical graph algorithms and clustering on one of LLNL’s supercom- that merely approximate the optimal software engineering. puters. To achieve effi ciency in parallel solution, i.e., maximum modularity. Related References 1. Newman, M. E. J., and M. Girvan, “Find- 19 ing and Evaluating Community Structure in 10 16 3 Networks,” Phys. Rev. E 69, 026113, 2004. 12 2. Zachary, W., “An Information Flow Model 23 14 for Confl ict and Fission in Small Groups,” Journal of Anthropological Research, 33, 33 8 pp. 452-473, 1977. 1 15 3. Yao, Y., T. L. Hickling, W. G. Hanley, and J. 31 4 S. Lenderman, “Graph Clustering Evaluation 2 via Integer Linear Programming,” Proceed- 21 13 ings of the 2007 International Conference on 34 Data Mining, pp. 369-375, 2007. 18 22 9 20 30 27 28 6 24 32 11 26 17

5 7 25 29 Optimal community structure for Zacharyʼs karate club.

Lawrence Livermore National Laboratory 107 Technology

Lee G. Glascoe Statistical Approach to (925) 423-2922 Complex Systems in the [email protected] Presence of Uncertainty

he optimization of engineering sys- Project Goals Ttems often inappropriately assumes The purpose of the project is to deterministic input parameters. The establish a stochastic framework that effect of such assumptions is generally enables the effi cient calculation of re- unknown particularly for complex sys- sponse statistics over the entire variable tems. We address this problem using a space. The framework consists of two probabilistic framework accounting for components. parametric uncertainties and allowing First, directly parallelizable ad- for quantifi cation in system response vanced sampling techniques are used to or performance. In this framework the evaluate the system response as a func- performance functions of interest are tion of design and random variables random, and their optimization is per- using the least number of deterministic formed with respect to the design vari- simulations. Second, the output of this ables. We adapted and implemented an sampling process is statistically classi- advanced statistical learning technique, fi ed using robust probit regression and sequential importance sampling (SIS), boosting schemes to derive the depen- to make predictions about the response dence of the performance statistics as of complex uncertain systems, includ- functions of the variables. ing an HE-detonation model and an energy policy model, and their sensitivity to variations of the variables.

(a) (b) 1.0

0.4 0.8

0.3 0.6 Pressure

0.2 P(detonation) 0.4

0.2 0.1

0 0 0 1 2 3 4 5 2000 4000 6000 8000 Time Plate velocity Figure 1. (a) Pressure vs. time at different gages for a deterministic ALE3D detonation simulation. (b) Probability of detonation as a function of plate velocity for three spatially ﬁ xed HE mass-fraction values (dotted lines) and accounting for zonal HE mass-fraction uncertainty using SIS (solid line).

108 FY07 Engineering Research and Technology Report Engineering Systems for Knowledge & Inference

Low investment 1.0 High investment 0.9 1.0 1.0 0.8 0.7 0.6 0.5 0.5 0.5 P (success) P 0.4 (success) P 0 0 0.4 0.3 0.4 Nucl 0.6 0.2 Nuclear generation0.6 7.2 ear 7.0 0.1 gener 0.8 6.8 0.8 6.6 7.2 ation 6.4 6.8 7.0 1.0 6.2 4 1.0 6.6 Allowed CO2 x 10 6.2 6.4 Allowed CO2 Figure 2. Probability of policy success as a function of three variables.

Relevance to LLNL Mission the zonal HE mass fraction. Illustrated Related References Our project addresses the need for in Fig. 1b is one possible response space 1. Doucet, A. J., F. G. de Freitas, and N. J. uncertainty quantifi cation and sensitivity using SIS along with three deterministic, Gordon, Eds., Sequential Monte Carlo Meth- analysis of complex systems and enhanc- i.e., spatially fi xed HE mass-fraction, ods in Practice, Springer, New York, 2001. es LLNL’s ability to apply advanced sto- ALE3D simulations. 2. Griffi ths, W. E., R. C. Hill, and P. J. Pope, chastic techniques to existing determin- The mathematical modeling of “Small Samples Properties of Probit Model istic codes. The advocated techniques energy systems is used to optimize Estimators,” Journal of the American Statis- increase effi ciency and predictive ability policy strategies but is hampered by large tical Association, 82, pp. 929-937, 1987. of existing computational codes and are uncertainties associated with several key 3. Neal, R. M., “Annealed Importance Sam- applicable to problems in infrastructure parameters. Policy decisions made in the pling,” Statistics and Computing, 1, 1, protection, energy security, and national presence of uncertainty work best with a pp. 125-139, 2001. security. probabilistic objective function, e.g., the 4. Urtiew, P., K. Vandersall, C. Tarver, F. probability that the cost will not exceed a Garcia, and J. Forbes, “Shock Initiation FY2007 Accomplishments and Results certain threshold. Experiments and Modeling of COMPB and Our accomplishments have been In this example we use MARKAL, C-4”, 13th International Detonation Sympo- in the area of detonation modeling and a deterministic model for energy sium, July 2006. energy policy. systems that accounts for a very large . Modeling HE detonation is poten- number of uncertain parameters. We tially a multivariate nonlinear problem consider two random variables (gas with uncertainty in several parameters. and oil prices) as well as three design The ALE3D hydrodynamics code variables (national R&D investment, deterministically predicts detonation nuclear generation, and CO2 allowance) based on peak overpressure, assuming and predict the probability of success- a spatially fi xed HE mass-fraction ratio fully attaining a cost-effective CO2 as a function of fl yer plate velocity emission reduction. Figure 2 depicts the (Fig. 1a). Using SIS we effi ciently probability of success as a function of compute and optimize the likelihood of the three design variables. detonation accounting for uncertainty in

Lawrence Livermore National Laboratory 109

Research

Jerome M. Solberg Ultrahigh-Velocity (925) 422-5971 Railgun [email protected]

railgun is a very simple device, essen- railgun would have contemporary applica- generating projectile velocities above Atially a single-turn coil (Fig. 1). Elec- tions in equation-of-state research, space 8 km/s. tric current is sent down one rail, through launch, and kinetic energy weapons. the moveable armature, and returns down Our UHV railgun project aims to Relevance to LLNL Mission the other rail. The magnetic “pressure” understand the physics that currently The development of a 3-D model- developed in the railgun bore behind the limits railgun performance, by leveraging capability tool for materials in the armature creates a driving force to ac- ing the history of railgun development warm-dense-matter regime characteristic celerate it down the rails. For speeds less with our world-class leadership in of railgun plasmas, including the phys- than about 2.5 km/s, designs typically use modeling and simulation. ics of melt and transition to the ionized a solid armature made of aluminum. At state, has a wide range of application speeds higher than that plasma forms, and Project Goals within the Laboratory’s mission. Related therefore either a hybrid armature (solid Today’s railgun technology is limited applications include the modeling of armature with plasma brushes) is used, or by poorly understood mechanisms at ve- exploding detonator bridge wires, explo- the solid portion of the armature is elimi- locities between 5 and 6 km/s. The most sive pulsed power applications, and arc nated a priori and a plasma armature is commonly accepted mechanism for this welding/cutting. Diagnostic techniques allowed to form. poor performance is “restrike,” defi ned developed and used in the validation During the late 1980’s, Lawrence as the short-circuiting of the main experiments have similar applications. Livermore National Laboratory took a current path at a distance signifi cantly A UHV railgun that exceeds current leading role in the development of rail- behind the armature. The ultimate cause velocity limits would be well-suited to guns as a means of launching projectiles of restrike has been variously attributed high-velocity impact-driven equation-of- at ultrahigh velocities, i.e., > 8 km/s. to viscous/ablative drag, thermally-driven state experiments, for which the present A practical ultrahigh-velocity (UHV) avalanche, or 3-D MHD effects (Fig. 2). technology (two-stage gas guns) has Progress toward isolating the mecha- limitations above 8 km/s. Current nisms had been impeded by the limited + simulation and diagnostic capabilities FY2007 Accomplishments and Results Magnetic available in the 1980’s when most railgun We augmented ALE3D with plasma field – tests were conducted. conductivity models, primarily from This project aims to overcome these the Laboratory’s Purgatorio code, and Projectile twin hurdles, by 1) developing ALE3D SESAME conductivity models from Rail B as a useful tool for the simulation of Sandia National Laboratories. Valida- Force UHV railguns; and 2) performing exper- tion studies comparing ALE3D to the F I Rail iments adequate to verify the capability legacy 2-D code CALE began. Both Figure 1. Basic railgun physics. I is the current; B is the of ALE3D in railgun-relevant regimes. CALE and ALE3D were used to design magnetic fi eld; and F is the resultant Lorentz force. Maxi- Ultimately, an understanding of the the Fixed Hybrid Armature experimental mum effi ciency occurs when the current is concentrated at limiting physics of railguns could lead facility (Figs. 3 and 4). the rear of the projectile. to candidate railgun designs capable of The Fixed Hybrid Armature test facility was constructed leveraging legacy hardware from the early 1990’s. In this Restrike Neutral Plasma Propulsive experiment, a railgun with the hybrid ar- plasma region tail plasma Projectile mature held in place is coupled to a long- pulse power source. Two plasma brushes are initiated from exploding aluminum foils. Since no gross motion of the armature occurs, this experiment allows Ablation Ablation Boundary detailed analysis of the plasma behavior, Viscous and inertial drag layer Figure 2. Restrike Hypothesis. The restrike plasma short-circuits the propulsive plasma, resulting in a reduction in the Lorentz force.

112 FY07 Engineering Research and Technology Report Energy Manipulation

2.6 2.4 Material temp 0.009 2.2 Current density Radiation temp 0.008 DB: FixedArmRailGun1D_008.04414 2.0 Cycle: 4414 Time:0.0300014 1 copper_1 2 air_2 0.007 Filled boundary 3 air_3 1.8 2 Var: material 4 aluminum–4 1.6 0.006 Density 5 insulator–5 A/cm 1.4 4 0.005 Conductivity x 0.1 1.2 Total pressure 0.004 1.0 gm/cc 10 0.8 0.003 Magnetic pressure 0.6 0.002 Y Temperature (eV) and pressure (kb) Temperature 0.4 Material pressure 0.2 0.001 Z X 0 0 0255351015 20 3040 45 0255351015 20 30 40 45 Figure 4. ALE3D Fixed Armature Simulations. The alumi- Position (mm) Position (mm) num in light blue has exploded and is being contained by Figure 3. 2D CALE Fixed Armature Simulations, used both as a reference for ALE3D development and to help design the magnetic pressure against the insulator wall (pink).The Fixed Armature Experiment. red bars are the rails; the dark blue is the air. including measurements of currents and Related References magnetic fi elds, plasma pressure, and 1. Hawke, R. S., et al., “Measurements of various optical diagnostics. Experimental Plasma Brush Behavior in FY2008 Proposed Work The experiment (Fig. 5) permits in- Hybrid Armatures,” IEEE Transactions on Simulation efforts will focus vestigation of armature plasma dynam- Magnetics 29, 1, pp. 814-818, 1993. largely on improving and understand- ics due to both ablation and high plasma 2. Crawford, R., D. Keefer, and R. Tipton, ing ALE3D as it applies to railgun ejection speeds, as well as investigating “Velocity Limiting Magnetohydrodynamic physics. This will include further near steady-state conditions. Diagnos- Effects in Railgun Plasma Armatures,” IEEE comparisons with CALE and application tics includes a fi ne array of B-dots for Transactions on Magnetics, 29, 1, pp. 781- to railguns with moving armatures. the magnetic fi eld, Rogowski coils for 786, 1993. Ultimately, the goal is to model legacy currents, and piezo-electric pressure 3. Boyton, G. C., et al., “Secondary Arcs in LLNL railgun experiments with a view sensors. Fiber optics captured the light 2-D MHD Numerical Simulations of EML toward understanding the observed emission data in preparation both for Plasma Armatures,” IEEE Transactions on velocity limitations. future use of fi ber-optic-based pressure Magnetics 31, 1, pp. 564-569, 1995. The Fixed Hybrid Armature experi- sensors and for optical techniques to 4. Parker, J. V., “Why Plasma Armatures ments will continue. A number of diag- measure the plasma emission character- Don’t Work (and What Can Be Done About nostics will be fully vetted within the istics as a function of time and position. It),” IEEE Transactions on Magnetics 25, 1, context of the experiment. Ultimately, We initiated a project with the U.S. pp. 418-424, 1999. the results of these experiments will Navy for modeling and simulation ef- be used to increase confi dence in the forts that focus on a different portion simulation results. of the mass/velocity space. This effort aims to replace conventional ship-based artillery with electrically launched kinetic energy weapons.

(a) (b)

Figure 5. Fixed Armature Experiment Shot 1, 290 kA. (a) As the aluminum foils explode, part of the plasma is lost out the back of the experiment. (b) The remainder is contained by the magnetic forces, and forms plasma brushes that conduct current through the ﬁ xed armature.

Lawrence Livermore National Laboratory 113 Technology

Todd J. Clancy Pulse Power Systems and (925) 422-8571 Diagnostics for the Fixed [email protected] Hybrid Armature Railgun

lasma physics within the confi nes of Project Goals Prailgun science is not well under- The primary goal of this project is stood. Since plasmas and “restrike” are to provide diagnostic data with suf- primarily responsible for preventing fi cient resolution and relevance to ultrahigh velocities (UHV) in such guns, validate a plasma model in ALE3D that there is a great need to predict and con- is applicable in the railgun regime. To trol that behavior. achieve this, we assembled the railgun Given the potential role of UHV and confi gured a testbed for the ap- railguns in equation of state (EOS) plication of high-energy pulsed power studies, we must regain our expertise in and sensitive diagnostics. We intend all aspects of railgun science. Working to perform multiple experiments, with toward this goal we rebuilt the Fixed multiple parameter modifi cations simi- Hybrid Armature Railgun (pictured lar to those performed a decade ago, in the previous report in this volume), along with new confi gurations. implemented advanced diagnostics, and began to recreate past experiments. Relevance to LLNL Mission Experimental data will be provided to This project directly addresses diag- the UHV railgun project and applied to nostic systems for a railgun testbed and the validation of a new plasma model in also addresses the feasibility of UHV ALE3D, a 3-D multi-physics computa- railguns, which would be well suited to tional platform. shock physics and EOS studies. Both of these goals are expressed as immediate needs for LLNL, and impact all pulsed power programs at LLNL.

FY2007 Accomplishments and Results By leveraging legacy hardware from the early 1990’s, we assembled the Fixed Hybrid Armature test facility and performed the fi rst experiment in late FY2007. In this experiment, plasma brushes were formed from exploding aluminum foils, which then provided a current path through the armature. Peak current was 290 kA and a post-shot inspection of the armature exhibited evidence of plasma brush creation (Fig. 1). Since the armature was fi xed in place, a pseudo-steady state was Figure 1. Post-shot picture of a small hybrid armature that had been held stationary between rails by the rod and stop achieved and diagnostic investiga- block. The plasma brush creation caused material ablation, evident on the armature. tion was simplifi ed. This shot included

114 FY07 Engineering Research and Technology Report Energy Manipulation

many diagnostic sensors. Two arrays of Figure 2. One of two B-dot fi ve B-dots each were used to measure arrays positioned to sense magnetic fi elds produced by the plasma. plasma magnetic fi eld gen- These small loop antennas were ori- eration. These are 15-turn loop antennas with an ID of ented with their loop normals orthogo- 0.063 in. and a measured nal to the rail magnetic fi elds, and were effective area of 33mm.2 thus unaffected by rail current (Fig. 2). They are soldered to semi- The array was used to diagnose plasma rigid coax and encased in currents as a function of position. transparent shrink-wrap. Two counter-wound Rogowski coils were positioned around the upper rail for measurement of the total input current, and plasma pressure was obtained with a piezoelectric quartz sensor (Fig. 3). The rail voltage was measured with a differential probe, and the pulsed power system included additional current and voltage diagnostics. Fiber optic pressure and optical Related References emission sensors were used as a proof of 1. Drobyshevski, E. M., et al., “Physics of concept in determining plasma charac- Solid Armature Launch Transition into Arc FY2008 Proposed Work teristics, and will be implemented fully Mode,” IEEE Transactions on Magnetics, We intend to continue experiments in future experiments. The experiment 37, 1, pp. 62-66, 2001. on the Fixed Hybrid Armature Railgun. produced a total of 45 fi les of data that we 2. Hawke, R. S., et al., “Summary of EM Data from FY2007 will be processed are currently analyzing. A sample of B-dot Launcher Experiments Performed at LLNL,” and analyzed, providing guidance data is shown in Fig. 4. IEEE Transactions on Magnetics, 22, 6, for additional shots. Fiber optic diag- pp. 1510-1515, 1986. nostics will be implemented, allowing 3. Hawke, R. S., et al., “Plasma Armature faster response times and less physical Formation in High Pressure, High-Velocity intrusion, thus providing additional Hydrogen, Starfi re: Hypervelocity Rail Gun precision to ALE3D validation. Development of High Pressure Research,” IEEE Transactions on Magnetics, 25, 1, p. 219, 1989.

x 1010 5

1 Plasma current (A/s)

–1 0 5 10 15 20 25 30 35 40 45 50 Time (μs) Figure 3. Piezoelectric quartz pressure sensor with Figure 4. Eight traces of scaled B-dot data from an early experiment shot. The initiation of each of the three capacitor integrated electronics that has a form factor capable of banks is evident at 8.5, 12.5, and 14.7 μs, followed by cable ringing. pressure measurements up to 30 kpsi. This post-shot photograph shows evidence of hardened copper and aluminum debris on the sensing diaphragm which is just 0.1 in. in diameter.

Lawrence Livermore National Laboratory 115 Technology

Timothy L. Houck Fast Diagnostic for Electrical (925) 423-7905 Breakdowns in Vacuum [email protected]

n important area of pulsed-power phenomenon. Typical vacuum elec- Project Goals AR&D is high-voltage vacuum in- trode spacing is dependent on voltage The deliverable for this project is sulator breakdown, often referred to as standoff expectations with spacings on a compact, high-bandwidth diagnostic fl ashover due to the sudden avalanche the order of a centimeter experiencing that can measure the electric fi eld in the of electrons across the insulator surface. fl ashover in a fraction of a nanosecond. immediate vicinity of an insulator fl ash- This phenomenon is often the limiting To record such fl ashover we require over with a temporal resolution capable factor in attaining the highest possible diagnostics with response times on the of capturing the initiating mechanism. performance in pulsed-power devices. order of a few hundred picoseconds or In addition, the design methodology for While several hypotheses attempt to faster. Equally important, the diagnostic the diagnostic is to be evaluated and explain the initiating mechanism(s), must be compatible with the geometry documented so that similar diagnostics fl ashover is not well understood. Com- of the experiment. In most applica- can be fabricated for other projects. putational modeling is limited to estab- tions and test stands, the insulator is There are presently multiple experi- lishing the electromagnetic fi eld and located between parallel electrodes a ments and proposals directed toward radiation imposed on the insulator. The few centimeters to tens of centimeters in the understanding of vacuum insula- designer then makes adjustments to the transverse dimension and separated by tor breakdown/fl ashover. All of these confi guration and/or materials until the distances of a few millimeters to a few studies require a fast diagnostic to modeling results are within safety mar- centimeters. Diagnostics that are located record the initial buildup of the electron gins based on empirical data. This design away from the electrode gap, e.g., on the avalanche. approach is believed to be overly conser- power feeds, record a signal delayed and vative since the bulk dielectric strength of distorted by the intervening inductance Relevance to LLNL Mission the insulating material can be an order of and capacitance. These inductances and This project provides LLNL with magnitude greater than the fl ashover limi- capacitances are normally very small, a more capable diagnostic for pulsed tation. However, without a quantitative but for the extremely fast signals associ- power R&D, specifi cally vacuum understanding of the phenomenon this is ated with fl ashover they act as fi lters that insulator studies. Our diagnostic can be the only approach to ensure reliability. remove critical spectral components of included in the suite of diagnostics used A major obstacle to gaining a the signal. for monitoring the performance of new quantitative understanding of insula- The goal of this project is to provide hardware under extreme conditions, tor fl ashover is the time scale for the an appropriate diagnostic. e.g., explosively driven fl ux generators.

Tips of probes penetrating anode

Cathode

Insulator

Anode with Diagnostic cables probes inserted

Figure 1. Anode plate with capacitively coupled diagnostic probes installed. Lines on the anode indicate the footprint of the test insulator. All but one probe are positioned under the insulator during testing.

Figure 2. Electrodes positioned in vacuum test chamber.

116 FY07 Engineering Research and Technology Report Energy Manipulation

At a more basic level, this fast FY2007 Accomplishments and Results 4. A prototype was fabricated and diagnostic is a tool for understanding Diagnostic probes were fabricated tested/calibrated. fl ashover and benefi ts projects studying and demonstrated involving insulator 5. Five probes were installed in the fl ashover in general. If the information fl ashover testing. Several steps were anode of an insulator test chamber acquired by this diagnostic leads to a involved in the successful fi elding of (Figs. 1 and 2). fundamental understanding of fl ashover, the probes: The frequency response of the probe all high power, pulsed power systems 1. A general design based on a D-dot was computationally estimated to be would be impacted. These systems probe for measuring the time rate in excess of 100 GHz. Figure 3 is a include high-current particle-beam ac- of change of the electric field was sample probe signal. The response of the celerators, high-power radio frequency chosen, based on measurement probes exceeds commercially available and microwave sources, high-power quantity and physical size con- digitizing capability. The two traces cor- laser sources, pulsed neutron sources, straints. respond to a fl ashover on the insulator nuclear weapons effects simulators, 2. A survey of commercially available surface immediately above the probe lightning and electromagnetic pulse components was undertaken. and a breakdown event distant from the effects simulators, x-ray and proton 3. Electrostatic simulations were probe. The changing electric fi eld due to radiography machines, inertial fusion performed to calculate the probe’s the electron avalanche is noted and par- drivers, directed energy weapons, and response and estimate its sensitivity tially resolved. The raw probe signal has electromagnetic launchers. and potential bandwidth. been converted to diode voltage in Fig. 4. The conversion parameters used were calculated from simulations and agreed 5 well with the known charge voltage.

4 Red trace is for probe directly Green trace is for probe beneath flashover. Trace removed from area of Related References 3 indicates effect of electrons flashover. Trace indicates 1. Humphries, S. Jr., Principles of Charged 2 avalanching on surface only the average field Particle Acceleration, John Wiley and Sons, and average electric field. between the electrodes. New York, 1999, Chapter 9.14. 1 2. Wetzer, J. M., “Vacuum Insulator Flash- Probe 1 over: Mechanisms, Diagnostics and Design 0 Implications,” Proc. XVIIth International –1 Symposium on Discharges and Electrical Insulation in Vacuum, Berkeley, California, –2 pp. 449-458, 1996. –3 3. Enloe, C. L., and R. M. Gilgenbach, “Mi- 0 2 41661481210 18 20 croscopic and Macroscopic Material Prop- Time (ns) erty Effects on Ultraviolet-Laser-Induced Figure 3. Raw signal from fast diagnostic probe. The signal is proportional to the electric ﬁ eld on the surface of the Flashover of Angled Insulators in Vacuum,” inner conductor of the probe. IEEE Trans. Plasma Sci., 16, 3, pp. 379-389, June 1988. 20

0 FY2008 Proposed Work –10 With our conﬁ dence in the success Electrodes have been of this project, we will explore placing –20 discharged by flashover probes in Flux Compression Gen- –30 erators. Further R&D related to the Diode voltage (kV) application of the probes is expected –40 to be associated with techniques for –50 Initial charge on cathode processing/analyzing the output signals to take advantage of the large –60 0 10 20 30 40 50 60 bandwidth. Time (ns) Figure 4. The raw probe signal, converted to show the potential difference between the electrodes with the cathode at -52.5 kV prior to discharge.

Lawrence Livermore National Laboratory 117 Technology

Jay Javedani UV-Induced Insulator (925) 422-8569 Flashover [email protected]

nsulators are critical components Project Goals Iin high-energy, pulsed power sys- The objective of this project is to tems. The insulator’s main function in measure the UV fl uence required for these devices is to provide an interface fl ashover of insulator as a function between regions of insulating fl uid/gas of material and cone angles. The data and vacuum that separate electrodes at gathered would validate the measure- different high-voltage potentials. It is ments reported in the 1980’s and known that the vacuum surface of the extend the knowledge base to present insulator will fl ashover when illuminated parameters. These results will be useful by ultraviolet (UV) radiation at voltages to electrically stressed, UV-radiated depending on the insulator material, in- systems in many pulsed power applica- sulator cone angle, applied voltage, and tions. insulator shot-history. Surface fl ashover refers to a cascade of electrons along the Relevance to LLNL Mission insulator’s surface leading to the collapse For many systems the delivery of of the voltage between the conductors. pulsed power into a vacuum region The UV radiation may be generated is the most critical factor impacting by ohmic heating of metal surfaces, performance and reliability. The ap- coronas in high electrical fi eld regions, plicability of our investigation in UV or plasmas from explosive emission. As fl ashover performance addresses issues the power of the pulsed power system related to power fl ow channels for fl ux is increased, so is the UV fl uence. An compression generators. As such, the accurate knowledge of the UV fl uence results have a signifi cant impact on the (energy per unit area) required for fl ash- LLNL national security mission. over is critical for the next generation of high power fl ow systems.

Figure 1. Photograph of the experimental apparatus with an opened cylindrical vacuum test chamber.

118 FY07 Engineering Research and Technology Report Energy Manipulation

corrected earlier published studies and provided design guidance for ongoing pulsed power programs. A sample of the data is shown in Fig. 3. A new effect was observed related to the UV intensity on fl ashover that had not been previously reported. It appears that UV pulses with intensity greater than the minimum UV fl uence need more than the established minimum energy/fl uence to induce surface fl ashover. In other words, the fl uence required for fl ashover is also a function of the intensity of the UV pulse. This effect would bias the data toward higher minimum fl ashover fl uence and lead to an incorrect interpretation of the previously reported data. We have proposed doing further studies to understand and quantify this effect. Figure 2. Photograph of insulators (clear HDPE, milky Rexolite ®). The HDPE insulator on its edge displays a rectangular damaged area from the UV laser beam. Related References 1. Enloe, C., et al., “Vacuum Ultra-Violet FY2007 Accomplishments and Results Testing included two types of insula- Effects on Power Transport Across a Vacuum/ A testbed comprised of an ex- tor materials: high-density polyethylene Solid Dielectric Interface,” Proc. Xth Interna- cimer laser (KrF, 248 nm), a vacuum (HDPE), and a crossed linked polysty- tional Symposium on Discharges and Electri- chamber (<1.0 x 10-6 Torr), and a dc rene (Rexolite®). Insulator angles of cal Insulation in Vacuum, Columbia, South high-voltage power supply (<60 kV) 0°, ±30°, and ±45° were extensively Carolina, I, pp. 308-314, October 1982. was established. Fast capacitive probes tested with a maximum UV fl uence of 2. Enloe, C. L., and R. E. Reinovsky, “Ultra- (D-dot) were embedded in the anode 70 mJ/cm2 at varying static charge on Violet Induced Insulator Flashover as a Func- electrode underneath the insulator to the insulator. Figure 2 is a photograph of tion of Material Properties,” Proc. 4th IEEE give the time of arrival of UV pulse tested insulators. Pulsed Power Conf., Albuquerque, New and fl ashover. A photograph of the test Over 700 data records were ob- Mexico, I, pp. 679-682, June 1983. equipment is provided in Fig. 1. tained. This information clarifi ed/ 3. Enloe, C. L., and R. M. Gilgenbach, “Microscopic and Macroscopic Material +45 degree insulator Property Effects on Ultraviolet-Laser-Induced 80 Flashover of Angled Insulators in Vacuum,” IEEE Trans. Plasma Sci., 16, 3, pp. 379-389, +45 polyethylene June 1988. +45 Rexolite® 70 ) 2 FY2008 Proposed Work We propose to extend our studies 60 to include other typically used insulator materials such as Macor and Micalex (machinable ceramics), Critical fluence (mJ/cm polyurethane and Mica-fi lled epoxies. 50 These materials have desirable mechanical properties for proposed applications, but their electrical properties in a UV radiation environ- 40 ment are not quantitatively known. 0501020 30 40 60 Field (kV/cm) We also propose to study the effect of the UV intensity on fl ashover. Figure 3. Critical fl uence at which surface fl ashover occurs versus the electric fi eld strength for two materials. The relatively complex relationship between the two variables is evident.

Lawrence Livermore National Laboratory 119 Technology

Laura Tully Evaluation of Light Activated (925) 424-2235 Thyristors for Pulsed Power [email protected] Applications

n pulsed power system design, the A high peak current handling capabil- Iswitch is arguably the most critical ity in a switch decreases the number of fundamental element. By deﬁ nition, parallel switches required in a high peak pulsed power is the compression of current system, thus reducing system energy in space and time. Reliable cost and complexity. High di/dt capabil- switching of stored energy is essential ity increases the range of possible loads to the performance of a pulsed power to the system. Both criteria are important system. Solid-state switch technology for pulsed power switch selection. offers the demonstrated advantage of reliability under a variety of condi- Project Goals tions. Light-triggered switches operate The goal of this project was to de- with a reduced susceptibility to electro- termine the suitability of light-triggered magnetic interference commonly found thyristors for pulsed power applications within pulsed power environments. beyond manufacturers’ ratings. To test Despite the advantages, commercially the applicability of the most recent available solid-state switches are typi- commercial light-triggered solid-state cally not designed for extreme pulsed designs, an adjustable high current power requirements. switch test stand was assembled in the High current switching poses a LLNL Pulsed Power Lab. A typical unique set of challenges. This effort switch stack is shown in Fig. 1. The focused on two key switch characteris- selected devices (Eupec T2563NH) were tics: high peak current and high di/dt. stressed beyond the di/dt rating of 5 kA/μs while maintaining low peak cur- Figure 1. Switch stack with optical trigger rents and, conversely, stressed beyond ﬁ bers inside orange jackets. the peak current rating of 93 kA while maintaining a low current rate of rise.

Relevance to LLNL Mission Solid-state switching is the future direction for many pulsed power applications with high-current, high-energy capacitor discharge units. LLNL has many switching needs including magnetic fl ux compression generators, fl ash- lamp banks, pulsed high-fi eld magnets, compact electric power conversion, and electromagnetic launchers.

FY2007 Accomplishments and Results Switch testing was conducted in two customizable testbeds. A resistive load (Fig. 2) was implemented for high di/dt testing while maintaining low peak current. An inductive load (Fig. 3) was constructed for high peak current testing while maintaining low di/dt. Each load

120 FY07 Engineering Research and Technology Report Energy Manipulation

Figure 2. High di/dt test stand with resistive load. Figure 3. High peak current test stand with inductive load. consisted of a variable number of resis- over rated values (95 kA) despite low not be operated outside of the speci- tors and inductors, respectively, to target di/dt (1.6 kA/ μs). The optical trigger fi ed ratings. With the current state of the bounds of operation for the device. pulsewidth was optimized such that no the technology, light-triggered solid- Full pulsewidth was minimized to less observable decrease in turn-on time could state switches may be self-limiting for than 10% of nominal 10 ms rating. be achieved. The trigger pulse amplitude demanding high current pulsed power Switch action was kept orders of mag- was maximized until it was driver-board applications. The increased cost and nitude below specifi ed rating. Switch limited. The decrease in power dissipated complexities introduced with series/ performance diagnostics included by the switch due to the slight reduction parallel combinations to achieve opera- Rogowski coils to measure current into in device turn-on time appeared to reduce tion within ratings may prove too cum- the switch and a differential voltage the magnitude of destruction evident dur- bersome for high current applications in probe to measure the voltage drop across ing dissection post-mortem. No increased the near future. the switch. peak current-carrying ability was demon- During high di/dt testing (5.6 to strated as a result of modifi cation of the Related References 7.2 kA/μs), the device failed after four trigger pulse. Detailed investigation of 1. General Electric Corporation, SCR Manual: shots operating above rated di/dt despite the failed devices was undertaken to de- Including Triacs and Other Thyristors, Sixth minimal current injection (28 to 36 kA). termine the failure mechanisms (Fig. 4). Edition, Prentice-Hall, Inc., Englewood Cliffs, High peak current testing indicated For high reliability, we have con- New Jersey, 1979. failure imminent at currents slightly cluded that the tested devices should 2. Ruff, M., H. Schulze, and U. Kellner, “Progress in the Development of an 8-kV Light-Triggered Thyristor with Integrated Protection Functions,” IEEE Trans. on Electron Devices, 46, 8, August 1999. 3. Podelsak, T. F., S. Schneider, and F. M. Simon, “Single Shot and Burst Repetitive Operation of Involute Gate 125 mm Symmetric Thyristors up to 221 kA with a di/dt of 2.0 kA/μs,” IEEE Trans. on Plasma Science, 28, 5, October 2000. 4. Wisken, H. G., F. Podeyn, T. H. Weise, J. Dorn, and D. Westerholt, “Light Activat- ed Semiconductors for ETC Pulsed Power Applications,” IEEE Trans. on Magnetics, 37, 1, January 2001. 5. Podelsak, T. F., F. M Simon, and S. Schneider, “Single Shot and Burst Re- petitive Operation of Thyristors for Electric Launch Applications,” IEEE Trans. on Magnetics, 37, 1, January 2001. Figure 4. Microscopic views of the gate structures. These images assist in determining exact mode of failure in the switching device.

Lawrence Livermore National Laboratory 121 Technology

Brent McHale Fiber Optic Current (925) 422-8730 Measurements Using Faraday [email protected] Rotation Diagnostics

t is often necessary to measure Project Goals Iextremely large pulsed electric cur- Our goal for FY2007 was to install rents when conducting pulsed power, and operate a FRD in the PPL high- explosively driven pulsed power, or current test cell, and to acquire the controlled fusion experiments. There are capabilities and knowledge base to a limited number of diagnostics that can support all aspects of FRD implemen- be used to accurately measure currents tation, including sensor fabrication, at these levels. The most common are experimental installation and operation, calibrated inductive fi eld sensors which and data analysis. are very susceptible to undesirable fi eld coupling and EMI. Relevance to LLNL Mission Faraday Rotation Diagnostics (FRDs) FRD sensors have excellent linear- rely on magneto-optical rather than ity and bandwidth characteristics, and inductive phenomena and are largely are optically isolated. These qualities immune to EMI. The polarization of the make FRDs particularly well suited for light in a magneto-optical material is application in experiments that involve rotated due to a circular birefringence. large quantities of guided or radiated This rotation is directly proportional electromagnetic energy. Since failure to the magnetic fi eld strength and the modes of FRDs differ from those of length over which the magnetic fi eld conventional inductive fi eld sensors, and optical fi eld interact in the mate- FRDs offer a level of data redundancy rial. FRDs have been used as a method for high-value single-shot experiments of measuring large pulsed currents for that is not easily achievable otherwise. more than 40 years and are presently Numerous programs at LLNL stand to used at other institutions in a range of benefi t from this expertise, including high-value pulsed power experiments. explosive pulsed power for high energy A FRD of proven confi guration was density physics research, EM launcher/ successfully installed in LLNL’s Pulsed shaker experiments for military applica- Power Lab (PPL) and used to measure tions, and operations at the NIF. current on the ALE3D coaxial load experiment. FY2007 Accomplishments and Results FRD hardware was acquired and installed in the PPL high-current test Screen room Sensor fiber cell on the ALE3D coaxial load testbed. Laser A graphical depiction of this system is shown in Fig. 1. A diode laser launches Digitizer PDs a few mW of linearly polarized light at Polarizers Beamsplitter 850 nm into a FRD sensor fi ber, which is a single-mode fi ber that is wrapped 50 to 100 times around the coaxial load Current carrying input (Fig. 2). The magnetic fi eld in conductor the vicinity of the sensor fi ber induces Bulk optics assembly Faraday rotation of the linearly polarized Figure 1. Block diagram of the FRD system.

122 FY07 Engineering Research and Technology Report Energy Manipulation

Optical Fiber Current Sensors,” J. Lightwave Technology, 15, pp. 803-807, May 1997. 2. Day, G. W., and A. H. Rose, “Faraday Effect Sensors: The State of the Art,” Proc. SPIE, 985, pp. 138–150, September 1988. 3. Bland, S. N., D. J. Ampleford, S. C. Bott, A. Guite, G. N. Hall, S. M. Hardy, S. V. Lebedev, P. Shardlow, A. Harvey-Thompson, F. Suzuki, and K. H. Kwek, “Use of Faraday Probing to Estimate Current Distribution in Wire Array Z Pinches,” Rev. Sci. Instrum., 77, 10E, p. 315, 2006.

FY2008 Proposed Work We will continue working on Figure 2. FRD sensor ﬁ ber installed on the ALE3D coaxial load testbed. a FRD at 635 nm that will provide improved sensitivity and noise characteristics over the present light. The polarization-rotated signal The total Faraday rotation is extracted implementation. Motivating this is coupled into a bulk optics assembly from the digitized data and is scaled by wavelength change is the fact that the that splits the beam via a non-polariz- material constants and geometric factors sensitivity of the device varies with ing beam splitter and passes each beam to yield current. We have created an the inverse of wavelength squared. through polarizers that are at a known algorithm that automates this process. The system will be deployable in relative angle. Both signals are then A time history of current obtained using support of pulsed power experiments coupled through multimode ﬁ ber onto this process is shown in Fig. 4, where it at LLNL, NTS, LANL and Site 300. We photodetectors and a digitizer in the is compared to integrated Rogowski coil will continue to work on ancillary screen room. data. These data demonstrate excellent aspects of FRD implementation such We have implemented the FRD on agreement, within 1% at peak current. as sensor fabrication, modeling, and four shots of the ALE3D coaxial load data analysis that permit improve- validation test. An example of raw data Related References ment of the precision and accuracy of representing the two components of 1. Rose, A. H., S. M. Etzel, and C. M. Wang, the diagnostic. the polarized signal is shown in Fig. 3. “Verdet Constant Dispersion in Annealed

3.0

2.5 800 700 FRD 2.0 600 Rogowski 500 1.5 400 1.0 300 Current (kA) 200 Photodetector voltage (V) 0.5 100 0 0 –100 0 20 40 6080 100 120 0 20 40 60 80 100 120 Time (μs) Time (μs) Figure 3. Raw FRD data representing two components of the Faraday-rotated polarized light. Figure 4. FRD and Rogowski measurements of a large amplitude, time varying current.

Lawrence Livermore National Laboratory 123

Author Index Author Index

Alves, Steven W...... 22 Miles, Robin ...... 74 Anderson, Andrew ...... 26 Morse, Jeffrey D...... 88 Barton, Nathan R...... 2 Mukerjee, Erik V...... 76 Beer, N. Reginald ...... 64 Nederbragt, Walter W...... 42 Bennett, Corey V...... 56 Ness, Kevin D...... 70 Bond, Tiziana ...... 80, 82 Nikolić, Rebecca J...... 86, 90 Brown, Charles G., Jr...... 28 Noble, Charles R...... 24 Brown, William D...... 44 Pannu, Satinderpall ...... 78 Burke, Michael W...... 48 Poland, Douglas N...... 52 Candy, James V...... 46 Puso, Michael A...... 6, 20 Clancy, Todd J...... 114 Rose, Klint A...... 72, 92 Corey, Bob ...... 16 Schmidt, Mark A...... 100 Florando, Jeffrey N...... 4 Seugling, Richard M...... 38 Glascoe, Lee G...... 108 Solberg, Jerome M...... 112 Heebner, John E...... 60 Spadaccini, Christopher M...... 94 Houck, Timothy L...... 116 Stölken, James S...... 8 Huber, Robert ...... 50 Tully, Laura ...... 120 Javedani, Jay ...... 118 Van Buuren, Anthony ...... 54 Kotovsky, Jack ...... 84 Vernon, Stephen P...... 58 Kroll, Jeremy J...... 40 Walton, Christopher ...... 30 Lemmond, Tracy D...... 104 Weisgraber, Todd ...... 96 Lin, Jerry I...... 18 Wemhoff, Aaron P...... 32 Mariella, Raymond P., Jr...... 66, 68 White, Adam ...... 34 May, John M...... 12 White, Daniel ...... 10, 14 McCarrick, James F...... 98 Yao, Yiming ...... 106 McHale, Brent ...... 122

Lawrence Livermore National Laboratory 125

Production Staff Or online at www-eng.llnl.gov/pubs.html Jeffrey B. Bonivert Lucy C. Dobson Kathy J. McCullough

This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344. ENG-07-0052